Growth model of the reared sea urchin Paracentrotus ... - SciViews
Growth model of the reared sea urchin Paracentrotus ... - SciViews
Growth model of the reared sea urchin Paracentrotus ... - SciViews
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
U L<br />
B<br />
bio mar<br />
UNIVERSITE LIBRE DE BRUXELLES<br />
FACULTE DES SCIENCES<br />
LABORATOIRE DE BIOLOGIE MARINE<br />
<strong>Growth</strong> <strong>model</strong> <strong>of</strong> <strong>the</strong> <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong><br />
<strong>Paracentrotus</strong> lividus (Lamarck, 1816)<br />
Committee:<br />
Pr<strong>of</strong>. G. Josens (president)<br />
Pr<strong>of</strong>. Ph. Dubois (secretary)<br />
Pr<strong>of</strong>. M. Jangoux<br />
Pr<strong>of</strong>. M. Russell<br />
Pr<strong>of</strong>. J.-L. Deneubourg<br />
Pr<strong>of</strong>. Ch. Lancelot<br />
Thesis submitted<br />
in fulfillment <strong>of</strong><br />
<strong>the</strong> degree <strong>of</strong> Doctor<br />
in Agronomic Sciences<br />
and Biological Engineering<br />
Supervisor: Pr<strong>of</strong>. M. JANGOUX<br />
Philippe GROSJEAN – September 2001
To my mo<strong>the</strong>r for her patience<br />
To my fa<strong>the</strong>r for showing me <strong>the</strong> way<br />
To 'Zazouille' for all she gave during 7 years<br />
1
Acknowledgements<br />
ACKNOWLEDGEMENTS<br />
Well, well, well… this <strong>the</strong>sis is written in English, fine! But I do not<br />
feel confident enough with Shakespeare's language to express my feelings.<br />
So, let's switch to French for one page or two…<br />
En tout premier lieu, je tiens à exprimer toute ma gratitude au<br />
Pr<strong>of</strong>esseur Michel JANGOUX pour m'avoir accueilli dans son laboratoire<br />
(ou devrais-je dire, dans l'annexe ô combien humide et salée de nos locaux<br />
à la Station Marine de Luc-sur-mer). Je le remercie de m'avoir témoigné<br />
toute sa confiance et d'avoir tout fait pour que mes conditions de travail<br />
soient aussi optimales que possible.<br />
Je tiens également à remercier les pr<strong>of</strong>esseurs Claude LARSONNEUR,<br />
Jacques AVOINE et Marie-Paule CHICHERY pour m'avoir accueilli au<br />
Centre Régional d'Etudes Côtières. Merci à Didier BUCAILLE pour s'être<br />
occupé des élevages avec tant de minutie. Les techniciens de la Station<br />
Marine (Jean-Paul LEHODEY, Jean-Pierre DESMASURES et Alain<br />
SAVINELLI) méritent un énorme bravo pour leur travail de qualité et pour<br />
leur aide précieuse dans la construction de matériel échinicole spécialisé.<br />
Je me dois également de signaler combien le travail des animaliers de tout<br />
poil (objecteurs, C.E.S., étudiants) a été vital et je les en remercie, en<br />
particulier Alexis DECTOT. Enfin, je remercie Brigitte GARCIA pour<br />
s'être acquittée de son travail d'intendance –et même plus– avec efficacité<br />
et… humour.<br />
A toute l'équipe "oursin", j'adresse mes remerciements du fond du cœur<br />
tant pour la coopération sur le plan pr<strong>of</strong>essionnel, que pour les aprèsboulots<br />
mémorables: Christine SPIRLET, Pol GOSSELIN, Devaragen<br />
VAITILINGON, Jean-Marc OUIN, Cristina DE AMARAL, Raphaël<br />
MORGAN, Corentin CAM, Yolaine BEYENS, Hélène RABAHIE, ainsi<br />
que les étudiants et étudiantes qui ont transité de façon plus brève dans<br />
l'équipe, trop nombreux que pour être cités tous, qu'ils m'en excusent.<br />
3
Acknowledgements<br />
Je voudrais également exprimer toute ma reconnaissance à Michel,<br />
Marie-Pierre, Ludo, Véronique, Joël, Alexandra, Patrice, Bernard, Jeloul,<br />
Jeff, Céline, Jean-Paul, Julie, Laurent, François, Laurence, Roseline,<br />
Stéphane, et bien sûr à Isabelle, pour tous les bons moments passés en leur<br />
compagnie. Je tiens aussi à remercier les laboratoires de Biologie Marine<br />
de Bruxelles et de Mons pour leur acceuil.<br />
Je remercie Christian VAN OSSELAER pour ses encouragements, ses<br />
discussions fructueuses et aussi pour LE conseil qui m'a permis de finir<br />
cette thèse: "Saint-John's Wort". A ma famille, j'exprime ma<br />
reconnaissance pour m'avoir soutenu dans mon travail et pour sa présence<br />
dans les moments difficiles.<br />
Enfin, bien que ce ne soit pas usuel, je crois utile de signaler qu'un<br />
certain nombre de personnes ont rendu ce travail possible de manière<br />
indirecte. Ainsi, c'est en parcourant les écrits de Ludwig VON<br />
BERTALANFFY, de Thomas EBERT et de Roger KOENKER que… plaf<br />
(bruit de la main qui frappe le front), bon sang, mais c'est bien sûr…! Des<br />
trois, je n'ai eu l'occasion de rencontrer que Thomas EBERT, et je me<br />
souviens encore de ses yeux écarquillés comme des billes de loto lorsque<br />
j'ai essayé de lui expliquer comment un modèle flou défuzzifié pouvait être<br />
une solution au problème qui nous préoccupait… A la réflexion, j'espère<br />
être plus explicite par écrit et après avoir maturé la question,… sinon, je<br />
risque bien de me retrouver de nouveau face à des billes de loto à la<br />
soutenance! Enfin, je voudrais adresser un très grand merci à tous les<br />
programmeurs qui ont fait de "R" un logiciel statistique aussi fantastique.<br />
Ce travail a été rendu possible par la collaboration entre le laboratoire<br />
de Biologie Marine de l'Université Libre de Bruxelles (Belgique) et le<br />
Centre Régional d'Etudes Côtières de l'Université de Caen (France). Il a pu<br />
être réalisé grâce à l'appui financier de la Commission Européenne<br />
(contrats FAR AQ2.530 BFE "Sea <strong>urchin</strong>s cultivation" et FAIR CT96-<br />
1623 BFN "Biology <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s under intensive cultivation [closed<br />
cycle echiniculture]").<br />
4
Abstract<br />
ABSTRACT<br />
A rearing protocol for <strong>the</strong> edible European <strong>sea</strong> <strong>urchin</strong> <strong>Paracentrotus</strong><br />
lividus in a closed cycle (control <strong>of</strong> <strong>the</strong> whole life cycle <strong>of</strong> <strong>the</strong> echinoid)<br />
and in a recirculating system (control <strong>of</strong> <strong>the</strong> environment around <strong>the</strong><br />
echinoid) is set up and tested at a pilot scale. This protocol is used to<br />
experiment on growing postmetamorphics whose age and genetic origin<br />
are perfectly known. Among <strong>the</strong> various measurements <strong>of</strong> size, we<br />
determined that <strong>the</strong> test diameter is both rapid and accurate for quantifying<br />
somatic growth. Causes and mechanisms <strong>of</strong> asymmetrical, or even<br />
sometimes multimodal, size distributions among previously homogeneous<br />
cohorts are studied. Results evidence <strong>the</strong> existence <strong>of</strong> a size-based<br />
intraspecific competition, causing a reversible growth inhibition <strong>of</strong> smaller<br />
individuals. A new growth <strong>model</strong> (called 'fuzzy-remanent'), including a<br />
component <strong>of</strong> intraspecific competition, is elaborated by defuzzifying a<br />
fuzzy <strong>model</strong>. Traditional least-square regression is abandoned in favor <strong>of</strong><br />
quantile regression to fit it. Both <strong>the</strong> <strong>model</strong> and <strong>the</strong> regression method are<br />
adapted to include individual variations (we call this an 'envelope <strong>model</strong>').<br />
This envelope <strong>model</strong> has functionally interpretable parameters. One <strong>of</strong><br />
<strong>the</strong>m quantifies <strong>the</strong> degree <strong>of</strong> inhibition caused by intraspecific<br />
competition. Since many similar fuzzy-remanent functions can be designed<br />
and fitted with this method, this approach is promising to <strong>model</strong> growth <strong>of</strong><br />
o<strong>the</strong>r organisms in a functional way. This <strong>model</strong> rehabilitates von<br />
Bertalanffy's <strong>the</strong>ory on individual growth. Moreover, <strong>the</strong> latter <strong>the</strong>ory is<br />
now verified for <strong>Paracentrotus</strong> lividus, despite <strong>the</strong> observation <strong>of</strong> an initial<br />
lag phase in growth. A functional classification <strong>of</strong> growth curves is<br />
proposed.<br />
Keywords: <strong>sea</strong> <strong>urchin</strong>, growth <strong>model</strong>, intraspecific competition, quantile<br />
regression, fuzzy logic, aquaculture, <strong>Paracentrotus</strong> lividus.<br />
5
Abstract<br />
6
Résumé<br />
RESUME<br />
Un protocole d'élevage pour l'oursin comestible européen <strong>Paracentrotus</strong><br />
lividus en cycle fermé (contrôle de tout le cycle de vie de l'échinide) et dans un<br />
système à recirculation d'eau (contrôle de l'environnement autour de<br />
l'échinide) est mis au point et testé à l'échelle pilote. Ce protocole est utilisé<br />
pour effectuer des expériences sur des individus postmétamorphiques en<br />
croissance dont l'âge et l'origine génétique sont parfaitement connus. Parmi les<br />
différentes manières de mesurer la taille de l'oursin, nous avons déterminé que<br />
le diamètre de son test est à la fois une mesure rapide et précise pour quantifier<br />
la croissance somatique. Les causes et les mécanismes responsables de<br />
distributions de tailles asymétriques, voire parfois multimodales au sein de<br />
cohortes initialement homogènes sont étudiés. Les résultats démontrent la<br />
présence d'une compétition intraspécifique basée sur la taille. Cette<br />
compétition entraîne une inhibition réversible des plus petits individus. Un<br />
nouveau modèle de croissance (dit 'à rémanence floue'), incluant une<br />
composante de compétition intraspécifique, est élaboré par défuzzification<br />
d'un modèle flou. La traditionnelle régression par les moindres carrés est<br />
abandonnée au pr<strong>of</strong>it de la régression quantile pour son ajustement. Tant le<br />
modèle que la méthode de régression sont modifiés pour inclure les variations<br />
individuelles (ce que nous appelons un 'modèle enveloppe'). Ce modèle<br />
enveloppe présente des paramètres que l'on peut interpréter fonctionnellement.<br />
L'un d'eux quantifie le degré d'inhibition occasionnée par la compétition<br />
intraspécifique. Etant donné que beaucoup de modèles à rémanence floue<br />
peuvent être conçus et ajustés de la sorte, cette approche est prometteuse pour<br />
modéliser la croissance d'autres organismes de manière fonctionnelle. Ce<br />
modèle réhabilite la théorie de von Bertalanffy sur la croissance des<br />
organismes. Cette théorie se vérifie par ailleurs dans le cas de <strong>Paracentrotus</strong><br />
lividus, malgré l'observation d'une phase de latence initiale dans sa croissance.<br />
Une classification fonctionnelle des courbes de croissance est proposée.<br />
Mots clefs: oursin, modèle de croissance, compétition intraspécifique,<br />
régression quantile, logique floue, aquaculture, <strong>Paracentrotus</strong> lividus.<br />
7
Résumé<br />
8
Table <strong>of</strong> contents<br />
TABLE OF CONTENTS<br />
ACKNOWLEDGEMENTS.............................................................................. 3<br />
ABSTRACT.................................................................................................. 5<br />
RESUME ..................................................................................................... 7<br />
TABLE OF CONTENTS................................................................................. 9<br />
LIST OF FIGURES...................................................................................... 13<br />
LIST OF TABLES ....................................................................................... 17<br />
LIST OF EQUATIONS................................................................................. 19<br />
LIST OF SYMBOLS .................................................................................... 23<br />
FOREWORD.............................................................................................. 29<br />
GENERAL INTRODUCTION ....................................................................... 31<br />
Economical interest <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s .........................................................................32<br />
a. Sea <strong>urchin</strong> markets and fisheries .........................................................................32<br />
b. Aquaculture potentials .........................................................................................34<br />
Overview <strong>of</strong> <strong>the</strong> biology <strong>of</strong> <strong>Paracentrotus</strong> lividus ..................................................35<br />
<strong>Growth</strong> <strong>model</strong>s .........................................................................................................42<br />
a. The exponential curve, a simple Malthusian growth <strong>model</strong> ................................42<br />
b. The logistic function for asymptotic growth ........................................................44<br />
c. The Gompertz <strong>model</strong>, an asymmetrical sigmoidal curve .....................................45<br />
d. The von Bertalanffy curves ..................................................................................46<br />
e. The Richards <strong>model</strong>, a flexible curve that contains many o<strong>the</strong>rs.........................47<br />
f. The Weibull <strong>model</strong>, a polyvalent and flexible function.........................................48<br />
g. The Jolicoeur curve, ano<strong>the</strong>r flexible <strong>model</strong>........................................................49<br />
h. The Johnson <strong>model</strong>, a heavily asymmetrical sigmoid..........................................50<br />
i. The Preece-Baines 1 <strong>model</strong> for human growth.....................................................51<br />
j. The Tanaka <strong>model</strong> for indeterminate growth........................................................51<br />
Modelling <strong>sea</strong> <strong>urchin</strong>s growth.................................................................................52<br />
a. Choice <strong>of</strong> <strong>the</strong> growth <strong>model</strong> for <strong>sea</strong> <strong>urchin</strong>s ........................................................53<br />
b. Fitting <strong>of</strong> growth <strong>model</strong>s on real data for echinoids ...........................................56<br />
AIM OF THE THESIS.................................................................................. 61<br />
PART I: SET UP OF AN EXPERIMENTAL REARING PROCEDURE FOR<br />
ECHINOIDS ............................................................................................... 65<br />
9
Land-based closed-cycle echiniculture <strong>of</strong> <strong>Paracentrotus</strong> lividus (Lamarck)<br />
(Echinoidea: Echinodermata): a long-term experiment at a pilot scale .............67<br />
a. Abstract ................................................................................................................67<br />
b. Introduction..........................................................................................................67<br />
c. Material and methods...........................................................................................69<br />
d. Results ..................................................................................................................78<br />
e. Discussion ............................................................................................................83<br />
f. Conclusions...........................................................................................................90<br />
g. Acknowledgements...............................................................................................90<br />
PART II: MEASUREMENT FOR SIZE IN THE SEA URCHIN ........................ 95<br />
Comparison <strong>of</strong> three body-size measurements for echinoids ..............................97<br />
a. Abstract ................................................................................................................97<br />
b. Introduction..........................................................................................................97<br />
c. Material and methods...........................................................................................98<br />
d. Results and discussion .......................................................................................100<br />
e. Conclusions ........................................................................................................104<br />
f. Acknowledgements..............................................................................................104<br />
Choice <strong>of</strong> measurement .........................................................................................105<br />
PART III: EXPERIMENTAL STUDIES OF THE INTRASPECIFIC<br />
COMPETITION ........................................................................................ 111<br />
Experimental study <strong>of</strong> growth in <strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus (Lamarck,<br />
1816) (Echinodermata)...........................................................................................113<br />
a. Abstract ..............................................................................................................113<br />
b. Introduction........................................................................................................113<br />
c. Material and methods.........................................................................................115<br />
d. Results ................................................................................................................118<br />
e. Discussion ..........................................................................................................124<br />
f. Acknowledgements..............................................................................................126<br />
Intraspecific competition: an additional experiment .........................................127<br />
PART IV: A GROWTH MODEL WITH INTRASPECIFIC COMPETITION.... 135<br />
A functional growth <strong>model</strong> with intraspecific competition applied to a <strong>sea</strong><br />
<strong>urchin</strong>, <strong>Paracentrotus</strong> lividus (Lamarck, 1816)....................................................137<br />
a. Abstract ..............................................................................................................137<br />
b. Introduction........................................................................................................138<br />
c. Material..............................................................................................................141<br />
d. Results ................................................................................................................144<br />
e. Discussion ..........................................................................................................164<br />
f. Conclusions.........................................................................................................177<br />
g. Acknowledgments...............................................................................................178<br />
GENERAL CONCLUSIONS ....................................................................... 181<br />
REFERENCES.......................................................................................... 189<br />
Table <strong>of</strong> contents<br />
10
ANNEXES................................................................................................ 213<br />
Annex I: R code for fitting growth <strong>model</strong>s ..........................................................215<br />
a. Code for analyzing data and fitting envelope <strong>model</strong>s........................................215<br />
b. The 'nlrq' package for nonlinear quantile regression........................................246<br />
Annex II: dataset <strong>of</strong> <strong>the</strong> cohort measured during seven years ..........................255<br />
Annex III: abstracts <strong>of</strong> publications and symposia ............................................257<br />
a. International journals ........................................................................................257<br />
b. Reports and o<strong>the</strong>r publications..........................................................................262<br />
c. International symposia.......................................................................................265<br />
Table <strong>of</strong> contents<br />
11
Table <strong>of</strong> contents<br />
12
List <strong>of</strong> figures<br />
LIST OF FIGURES<br />
Figure 1. <strong>Paracentrotus</strong> lividus in a tidal pool in Morgat, Brittany,<br />
France. Page 36.<br />
Figure 2. <strong>Paracentrotus</strong> lividus. A. Echinopluteus. B. Postlava a few<br />
days after metamorphosis. Page 37.<br />
Figure 3. External anatomy <strong>of</strong> a regular <strong>sea</strong> <strong>urchin</strong>. A. Oral view. B.<br />
Aboral view. Page 38.<br />
Figure 4. Internal anatomy <strong>of</strong> a regular <strong>sea</strong> <strong>urchin</strong>, side view. Page 39.<br />
Figure 5. Mature adult <strong>Paracentrotus</strong> lividus with oral region removed<br />
showing <strong>the</strong> five gonads. Page 40.<br />
Figure 6. Location <strong>of</strong> <strong>the</strong> sampled <strong>Paracentrotus</strong> lividus population.<br />
Page 41.<br />
Figure 7. Example <strong>of</strong> an exponential curve. Page 43.<br />
Figure 8. Example <strong>of</strong> a logistic curve. Page 44.<br />
Figure 9. Example <strong>of</strong> a Gompertz curve. Page 45.<br />
Figure 10. Both von Bertalanffy 1 and von Bertalanffy 2 curves.<br />
Page 47.<br />
Figure 11. Shape <strong>of</strong> Richards curves depending on values <strong>of</strong> m. Page 48.<br />
Figure 12. Examples <strong>of</strong> Weibull curves with different values for m.<br />
Page 49.<br />
Figure 13. Examples <strong>of</strong> Jolicoeur curves with different values for m.<br />
Page 50.<br />
Figure 14. Example <strong>of</strong> a Johnson curve. Page 50.<br />
Figure 15. Example <strong>of</strong> a Preece-Baines 1 curve. Page 51.<br />
13
List <strong>of</strong> figures<br />
Figure 16. Example <strong>of</strong> a Tanaka curve. Page 52.<br />
Figure 17. Overview <strong>of</strong> <strong>the</strong> closed-cycle process and devices used to<br />
produce <strong>sea</strong> <strong>urchin</strong>s on land at a pilot scale. Page 71.<br />
Figure 18. Changes with time in <strong>the</strong> size distribution and survival rate <strong>of</strong><br />
one fertilization issued from a single larval rearing tank and<br />
followed over 7 years. Page 80.<br />
Figure 19. Change with time in <strong>the</strong> biomass <strong>of</strong> a <strong>reared</strong> cohort <strong>of</strong> <strong>sea</strong><br />
<strong>urchin</strong>s (<strong>the</strong> same batch as shown in Fig. 18). Page 81.<br />
Figure 20. Data acquisition system. Page 99.<br />
Figure 21. Principal components analysis <strong>of</strong> <strong>the</strong> 14 measurements.<br />
Page 106.<br />
Figure 22. Evolution <strong>of</strong> a single cohort <strong>of</strong> P. lividus (Fb) <strong>reared</strong> in stable<br />
environmental conditions according to time. Page 120.<br />
Figure 23. Size distribution <strong>of</strong> Fc juveniles in each batch in <strong>the</strong><br />
beginning <strong>of</strong> <strong>the</strong> experiment (A) and 4 months later (B).<br />
Page 121.<br />
Figure 24. Size distribution <strong>of</strong> Fd individuals in each batch at <strong>the</strong><br />
beginning <strong>of</strong> <strong>the</strong> experiment (A) and 4 months later (B).<br />
Page 122.<br />
Figure 25. Size distributions <strong>of</strong> <strong>the</strong> two different fertilizations used in <strong>the</strong><br />
additional experiment (Ff and Fg). Page 128.<br />
Figure 26. Change in size distributions <strong>of</strong> <strong>the</strong> Ff and Fg batches (large,<br />
mixed and small) with time. Page 130.<br />
Figure 27. Change in size distributions <strong>of</strong> <strong>the</strong> Ff and Fg batches with<br />
time after large individuals were removed from <strong>the</strong> mixed<br />
batches. Page 131.<br />
14
List <strong>of</strong> figures<br />
Figure 28. A. Histograms <strong>of</strong> size distributions <strong>of</strong> a cohort <strong>of</strong> <strong>reared</strong> P.<br />
lividus with time. Top <strong>of</strong> <strong>the</strong> box: a projection <strong>of</strong> three<br />
quantiles (0.025, 0.5 and 0.075) issued from those size<br />
distributions and also presented in B. Page 143.<br />
Figure 29. Survival with time <strong>of</strong> <strong>the</strong> same <strong>reared</strong> cohort <strong>of</strong> P. lividus as<br />
in Fig. 28A. Page 143.<br />
Figure 30. Construction <strong>of</strong> <strong>the</strong> fuzzy growth <strong>model</strong>. Page 150.<br />
Figure 31. First step <strong>of</strong> constraining parameters <strong>of</strong> <strong>the</strong> new growth <strong>model</strong><br />
(origin forced to {t0, D0}). Page 157.<br />
Figure 32. A. Variation <strong>of</strong> l as a function <strong>of</strong> 1-τ for several quantile<br />
regressions. B. Variation <strong>of</strong> k1 (black squares) and k2 (white<br />
triangles) as functions <strong>of</strong> 1-τ. Page 159.<br />
Figure 33. Envelope <strong>model</strong> fitted (upper surface) to <strong>the</strong> whole dataset<br />
(lower surface). Page 162.<br />
Figure 34. Diagnostic <strong>of</strong> <strong>the</strong> envelope <strong>model</strong> fitted in Fig. 33. A.<br />
Contour plot <strong>of</strong> <strong>the</strong> residuals. B. Three "slices" cut in <strong>the</strong> 3Dsurfaces<br />
<strong>of</strong> Fig. 33 at t' = 300 (1), 600 (2) and 1800 (3) days.<br />
Page 163.<br />
Figure 35. A classification <strong>of</strong> growth <strong>model</strong>s based on <strong>the</strong>ir functional<br />
features. Page 184.<br />
15
List <strong>of</strong> figures<br />
16
List <strong>of</strong> tables<br />
LIST OF TABLES<br />
Table 1. Models used to fit <strong>sea</strong> <strong>urchin</strong>s growth data. Page 54.<br />
Table 2. Age, density, number, and survival rate <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s at each<br />
rearing stage. Page 79.<br />
Table 3. Gonadal and maturity indices <strong>of</strong> wild and <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s.<br />
Page 82.<br />
Table 4. Comparison <strong>of</strong> <strong>the</strong> three selected measurements for body size.<br />
Page 101.<br />
Table 5. Allometric relations between parameters for <strong>Paracentrotus</strong><br />
lividus from Morgat. Page 103.<br />
Table 6. Principal components analysis: contribution <strong>of</strong> <strong>the</strong> parameters<br />
to <strong>the</strong> three first axes. Page 107.<br />
Table 7. Statistical analysis <strong>of</strong> <strong>the</strong> size frequency distribution <strong>of</strong> single<br />
cohorts <strong>of</strong> P. lividus at different ages. Page 119.<br />
Table 8. Statistics on <strong>the</strong> four batches <strong>of</strong> Fc in <strong>the</strong> beginning <strong>of</strong> <strong>the</strong><br />
experiment and after 4 months. Page 121.<br />
Table 9. Statistics on <strong>the</strong> six batches <strong>of</strong> Fd in <strong>the</strong> beginning <strong>of</strong> <strong>the</strong><br />
experiment and after 4 months. Page 122.<br />
Table 10. Size distribution <strong>of</strong> Fe echinoids after having been <strong>reared</strong><br />
individually (control) or toge<strong>the</strong>r (experimental batches) for 4<br />
months. Page 124.<br />
Table 11. Statistics on <strong>the</strong> small, large and mixed batches (fertilizations<br />
Ff, 'small' and Fg, 'large'). Page 129.<br />
Table 12. Results <strong>of</strong> quantile regressions for three values <strong>of</strong> τ, using<br />
different growth <strong>model</strong>s. Page 154.<br />
17
List <strong>of</strong> tables<br />
Table 13. Results <strong>of</strong> quantile regressions for three values <strong>of</strong> τ , using <strong>the</strong><br />
new growth <strong>model</strong>. Page 155.<br />
Table 14. Results <strong>of</strong> quantile regressions for different values <strong>of</strong> τ, using<br />
<strong>the</strong> new growth <strong>model</strong> constrained to <strong>the</strong> origin. Page 156.<br />
18
List <strong>of</strong> equations<br />
LIST OF EQUATIONS<br />
Equation 1. Exponential growth <strong>model</strong>: population increase by a fixed<br />
proportion. Page 43.<br />
Equation 2. Exponential growth <strong>model</strong>: differential equation. Page 43.<br />
Equation 3. Exponential growth <strong>model</strong>: equation. Page 43.<br />
Equation 4. Logistic <strong>model</strong>: differential equation. Page 44.<br />
Equation 5. Logistic <strong>model</strong>: equation. Page 44.<br />
Equation 6. 4-parameter logistic <strong>model</strong>. Page 45.<br />
Equation 7. Gompertz <strong>model</strong>: differential equation. Page 45.<br />
Equation 8. Gompertz <strong>model</strong>: equation. Page 45.<br />
Equation 9. von Bertalanffy 2 <strong>model</strong>: differential equation. Page 46.<br />
Equation 10. von Bertalanffy 2 <strong>model</strong>: equation. Page 46.<br />
Equation 11. von Bertalanffy 1 <strong>model</strong>. Page 46.<br />
Equation 12. Richards <strong>model</strong>. Page 47.<br />
Equation 13. Weibull <strong>model</strong>. Page 48.<br />
Equation 14. Jolicoeur <strong>model</strong>. Page 49.<br />
Equation 15. Johnson <strong>model</strong>. Page 50.<br />
Equation 16. Preece-Baines <strong>model</strong> 1. Page 51.<br />
Equation 17. Tanaka <strong>model</strong>. Page 52.<br />
Equation 18. Definition <strong>of</strong> SIW. Page 100.<br />
19
List <strong>of</strong> equations<br />
Equation 19. Calculation <strong>of</strong> ds, <strong>the</strong> apparent mean density <strong>of</strong> <strong>the</strong> skeleton<br />
<strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> using DWs / IW relationship. Page 101.<br />
Equation 20. Moving average smoothing applied to size-frequency<br />
distributions. Page 117.<br />
Equation 21. Objective function (deviance δ1) <strong>of</strong> <strong>the</strong> quantile regression.<br />
Page 146.<br />
Equation 22. Piece-wise linear function used to calculate <strong>the</strong> deviance δ1<br />
in eq. 21. Page 147.<br />
Equation 23. Definition <strong>of</strong> <strong>the</strong> relative time-scale t'. Page 149.<br />
Equation 24. Function <strong>of</strong> size D with relative time t' for <strong>the</strong> set S in <strong>the</strong><br />
fuzzy <strong>model</strong>. Page 150.<br />
Equation 25. Function <strong>of</strong> size D with relative time t' for <strong>the</strong> set L in <strong>the</strong><br />
fuzzy <strong>model</strong>. Page 151.<br />
Equation 26. Membership function to <strong>the</strong> set L with relative time t'.<br />
Page 151.<br />
Equation 27. Membership function to <strong>the</strong> set S with relative time t'.<br />
Page 151.<br />
Equation 28. Defuzzification <strong>of</strong> <strong>the</strong> fuzzy <strong>model</strong>. Page 152.<br />
Equation 29. Equation <strong>of</strong> <strong>the</strong> defuzzified <strong>model</strong> after simplification.<br />
Page 152.<br />
Equation 30. Definition <strong>of</strong> <strong>the</strong> relative size-scale D'. Page 155.<br />
Equation 31. The new growth <strong>model</strong> with relative size D' and relative<br />
time t'. Page 155.<br />
Equation 32. Parameter l in function <strong>of</strong> τ in <strong>the</strong> envelope <strong>model</strong>.<br />
Page 158.<br />
20
List <strong>of</strong> equations<br />
Equation 33. Parameters k1 and k2 in function <strong>of</strong> τ in <strong>the</strong> envelope<br />
<strong>model</strong>. Page 158.<br />
Equation 34. Parameter ∆D∞ in function τ in <strong>the</strong> envelope <strong>model</strong>.<br />
Page 160.<br />
Equation 35. Analytic function <strong>of</strong> <strong>the</strong> envelope <strong>model</strong>. Page 160.<br />
Equation 36. Calculation <strong>of</strong> ˆ τ . Page 160.<br />
Equation 37. Objective function (deviance δ2) for <strong>the</strong> quantile regression<br />
modified for envelope <strong>model</strong>s. Page 161.<br />
Equation 38. Reparameterization <strong>of</strong> <strong>the</strong> 4-parameter logistic function.<br />
Page 174.<br />
Equation 39. Reparameterized 4-parameter logistic function after<br />
simplification. Page 174.<br />
Equation 40. A growth <strong>model</strong> that is both 'dimensional' and 'transitional'<br />
at <strong>the</strong> same time. Page 175.<br />
Equation 41. Equation defining <strong>the</strong> relation between metabolic time tM<br />
and time t'. Page 175.<br />
Equation 42. Generalized von Bertalanffy <strong>model</strong>. Page 175.<br />
21
List <strong>of</strong> equations<br />
22
List <strong>of</strong> symbols<br />
LIST OF SYMBOLS<br />
Variables and parameters are in italic; function names are in roman<br />
type according to standard ma<strong>the</strong>matical notation. For instance, e (in<br />
italic) is <strong>the</strong> name <strong>of</strong> a variable and e (in roman) is <strong>the</strong> exponentiation<br />
function as in y = e x and thus e equals Euler constant: 2.71282.<br />
α, α, α, α, ββ<br />
ββ<br />
Parameters <strong>of</strong> <strong>the</strong> Huxley's allometric equation y = α·x β .<br />
∆D∞ Maximum increase in diameter <strong>of</strong> <strong>the</strong> test <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> from a<br />
defined initial value (usually, at metamorphosis) D0 to <strong>the</strong> asymptotic<br />
maximum diameter D∞ (in mm).<br />
∆Y∞ Maximum increase in size from a defined initial size (at birth, at<br />
metamorphosis…) Y0 to <strong>the</strong> asymptotic size Y∞ (same unit as Y).<br />
δδδδ1 Deviance <strong>of</strong> quantile regression, according to Koenker & Bassett<br />
(1978) (same unit as <strong>the</strong> dependent variable y in <strong>the</strong> <strong>model</strong>).<br />
δδδδ2 Deviance <strong>of</strong> quantile regression; modified version for envelope<br />
<strong>model</strong>s (same unit as <strong>the</strong> dependent variable y in <strong>the</strong> <strong>model</strong>).<br />
δδδδs Apparent mean density <strong>of</strong> <strong>the</strong> skeleton <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> (in g/l).<br />
ττττ Quantile (0 < τ < 1) <strong>of</strong> a distribution or <strong>of</strong> a quantile regression<br />
(dimensionless).<br />
ˆ ττττ Estimator <strong>of</strong> <strong>the</strong> quantile τ. Value calculated after a sample <strong>of</strong> <strong>the</strong><br />
distribution (dimensionless).<br />
ωωωω Parameter corresponding to k·Y∞ in <strong>the</strong> von Bertalanffy equation<br />
Y = Y∞·(1 – e -k·(t – t 0 ) ), as proposed by Gallucci et al (1979) to solve<br />
<strong>the</strong> problem <strong>of</strong> intercorrelation between k and Y∞ (same unit as Y·t -1 ).<br />
ξξξξ1 Value returned by a function <strong>of</strong> <strong>the</strong> form Y = f(t) at time t and for <strong>the</strong><br />
current solution for <strong>the</strong> parameters (same unit as Y).<br />
23
List <strong>of</strong> symbols<br />
ξξξξ2 Value returned by a function <strong>of</strong> <strong>the</strong> form Y = f(t, τ) (envelope <strong>model</strong>)<br />
at time t and quantile τ and for <strong>the</strong> current solution for <strong>the</strong> parameters<br />
(same unit as Y).<br />
a, b, c, d, e Parameters in growth <strong>model</strong>s with no particular functional<br />
meaning or used in a context where <strong>the</strong> possible functional meaning<br />
has no importance (fitting for descriptive purpose only) (units are<br />
context-dependent).<br />
D Diameter <strong>of</strong> <strong>the</strong> test <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> measured at <strong>the</strong> ambitus and<br />
considered without spines (in mm).<br />
D' Relative diameter <strong>of</strong> <strong>the</strong> test <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong>: increase <strong>of</strong> <strong>the</strong> diameter<br />
from a defined initial value (usually at metamorphosis) D0 to <strong>the</strong><br />
current value D (in mm).<br />
D0 Diameter <strong>of</strong> <strong>the</strong> test <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> at time t = t0, usually at<br />
metamorphosis (in mm).<br />
D∞ Maximum asymptotic diameter <strong>of</strong> <strong>the</strong> test <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong><br />
(determinate growth) (in mm).<br />
DWs Dry weight <strong>of</strong> <strong>the</strong> skeleton <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> (in g).<br />
Fx A <strong>reared</strong> batch <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s issued from a single fertilization x used<br />
in an experiment. If several batches issued from <strong>the</strong> same fertilization<br />
are used, <strong>the</strong>y are fur<strong>the</strong>r labeled with numbers (Fx1, Fx2…).<br />
fi<br />
A frequency observed in <strong>the</strong> i th class <strong>of</strong> a size distribution<br />
(dimensionless).<br />
fsi Same as fi but after applying a moving average smoothing<br />
(dimensionless).<br />
GI Gonad index: <strong>the</strong> ratio between <strong>the</strong> weight <strong>of</strong> <strong>the</strong> gonads and <strong>the</strong> total<br />
weight <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong>, ei<strong>the</strong>r in wet or in dry weight (dimensionless).<br />
24
List <strong>of</strong> symbols<br />
IW Immersed weight: <strong>the</strong> weight <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> measured when<br />
immersed in <strong>sea</strong>water (in g). See also SIW.<br />
k Kinetic parameter in a growth <strong>model</strong>. If <strong>the</strong>re are different kinetic<br />
parameters in <strong>the</strong> same <strong>model</strong>, <strong>the</strong>y are fur<strong>the</strong>r labeled with numbers:<br />
k1, k2… (in day -1 ).<br />
L Fuzzy set representing <strong>the</strong> largest possible size (in mm). See also ML<br />
and S.<br />
l Lag parameter in a growth <strong>model</strong> with an intraspecific competition<br />
component. Indicate <strong>the</strong> length <strong>of</strong> <strong>the</strong> lag phase, i.e., <strong>the</strong> degree <strong>of</strong><br />
inhibition (dimensionless).<br />
m Parameter in a dimensional <strong>model</strong> that indicates <strong>the</strong> power<br />
transformation to apply to be in <strong>the</strong> best dimension for describing<br />
growth (dimensionless?).<br />
Md Mass density <strong>of</strong> <strong>sea</strong>water (in g/l).<br />
MI Maturity index: <strong>the</strong> arithmetic mean <strong>of</strong> all maturity stages observed<br />
in a batch. An eight-stage scale <strong>of</strong> maturity was defined by Spirlet et<br />
al (1998a) for <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> gonads (dimensionless).<br />
ML Membership function for <strong>the</strong> set L in a fuzzy <strong>model</strong> with 0 ≤ ML ≤ 1<br />
(dimensionless). See also L and MS.<br />
MS Membership function for <strong>the</strong> set S in a fuzzy <strong>model</strong> with 0 ≤ MS ≤ 1<br />
(dimensionless). See also S and ML.<br />
n Number <strong>of</strong> observations in a dataset (dimensionless).<br />
p Probability <strong>of</strong> a statistical test, or probability <strong>of</strong> a value according to<br />
a statistical distribution with 0 ≤ p ≤ 1 (dimensionless).<br />
S Fuzzy set representing <strong>the</strong> smallest possible size (in mm). See also<br />
MS and L.<br />
25
List <strong>of</strong> symbols<br />
s Parameter <strong>of</strong> <strong>the</strong> envelope <strong>model</strong>. Slope <strong>of</strong> <strong>the</strong> linear relationship<br />
l = s·(1 - τ). (dimensionless). See also l.<br />
SCT Standard competence test. Determine if <strong>sea</strong> <strong>urchin</strong> larvae are ready to<br />
metamorphose; adapted from Gosselin & Jangoux, 1996).<br />
SIW Standard immersed weight. The apparent weight <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> in<br />
<strong>sea</strong>water, standardized according to eq. 18 (in g).<br />
t Chronological time with an arbitrary origin (in days).<br />
t' Relative chronological time, that is, with a defined origin (usually<br />
corresponding to <strong>the</strong> metamorphosis event) (in days).<br />
t0<br />
Time at a defined origin (usually corresponding to <strong>the</strong><br />
metamorphosis event) (in days).<br />
tM Metabolic or physiologic time, that is, a time-scale that is modulated<br />
by environmental parameters in <strong>the</strong> same proportions as <strong>the</strong>y change<br />
<strong>the</strong> metabolic rate <strong>of</strong> <strong>the</strong> organism (in days).<br />
W A measure <strong>of</strong> <strong>the</strong> size <strong>of</strong> an animal using a weight measurement (in<br />
g).<br />
Y A measure <strong>of</strong> <strong>the</strong> size <strong>of</strong> an animal, using any kind <strong>of</strong> measurement.<br />
Note that y corresponds to any kind <strong>of</strong> dependent variable, while Y is<br />
a dependent variable that expresses <strong>the</strong> size <strong>of</strong> an individual, or <strong>of</strong> a<br />
population (unit si context-dependent).<br />
Y' Relative size: increase <strong>of</strong> <strong>the</strong> size from a defined initial value (at<br />
birth, at metamorphosis…) Y0 to <strong>the</strong> current value Y (same unit as Y).<br />
Y0 The size <strong>of</strong> an animal at time t = t0 (usually at birth or<br />
metamorphosis) (same unit as Y).<br />
Y∞ Maximum asymptotic size (determinate growth) (same unit as Y).<br />
26
Introduction<br />
27
Foreword<br />
FOREWORD<br />
This <strong>the</strong>sis is organized in four successive parts accordingly to a<br />
logical progression in <strong>the</strong> scientific approach. Published or submitted<br />
papers constitute <strong>the</strong> body <strong>of</strong> each section. Some supplemental material is<br />
added when fur<strong>the</strong>r exploration was made after <strong>the</strong> publication <strong>of</strong> <strong>the</strong><br />
papers.<br />
In this document, results <strong>of</strong> some statistical tests are presented inline in<br />
an abridged form, like: Student test, p < 0.001. Current tendency –at least<br />
in international journals– is to favor a more detailed presentation, e.g.,<br />
Student test, t = 3.858, df = 84, p < 0.001. We believe it does not bring<br />
additional vital information, but it just surcharges text. A small table<br />
summarizes much better statistical results where required.<br />
Also I actively participated in scientific works that are marginal to <strong>the</strong><br />
main body <strong>of</strong> <strong>the</strong> present <strong>the</strong>sis. These works are shortly presented<br />
(abstract <strong>of</strong> publications) as annexes (see Annex III).<br />
29
Foreword<br />
30
General introduction<br />
GENERAL INTRODUCTION<br />
This work was initiated in <strong>the</strong> context <strong>of</strong> <strong>the</strong> global overexploitation <strong>of</strong><br />
<strong>the</strong> natural resources <strong>of</strong> <strong>sea</strong> <strong>urchin</strong> fisheries (Allain, 1972a, 1972b; Le<br />
Gall, 1987, 1990; Ledireac'h, 1987; Conand & Sloan, 1989; Hagen,<br />
1996a). Recently this issue raised considerable interest in <strong>sea</strong> <strong>urchin</strong><br />
aquaculture (echiniculture) (Le Gall, 1990; Cellario & Fenaux, 1990; de<br />
Jong-Westman, 1995a, 1995b; Fernandez, 1996; Hagen, 1996a; Blin,<br />
1997; Kelly et al, 1998; Spirlet et al, 2000, 2001). As a consequence <strong>of</strong><br />
<strong>the</strong>ir differences [<strong>sea</strong> <strong>urchin</strong>s are radically different than most marine<br />
species usually farmed (finfishes, molluscs or crustaceans)], specific<br />
rearing methods had to be developed. Our knowledge about <strong>the</strong> biology <strong>of</strong><br />
<strong>the</strong>se animals was too fragmentary and several national or international<br />
programs were established to lead ecophysiological studies <strong>of</strong> fished<br />
echinoid populations and <strong>of</strong> <strong>the</strong> echinoids in culture. In this context, we<br />
worked on two successive European contracts: FAR AQ2.530 BFE "Sea<br />
<strong>urchin</strong>s cultivation" and FAIR CT96-1623 BFN "Biology <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
under intensive cultivation (closed cycle echiniculture)".<br />
This project presented an opportunity to build an experimental rearing<br />
facility in <strong>the</strong> Marine Station <strong>of</strong> Luc-sur-mer, Normandy, France ("Centre<br />
de Recherche et d'Etude Côtière"), where it was possible to grow<br />
thousands <strong>of</strong> echinoids in strictly controlled food and environmental<br />
conditions. If experiments conducted in this facility shed light on critical<br />
aspects <strong>of</strong> echiniculture, <strong>the</strong>y are also beneficial to fundamental re<strong>sea</strong>rch<br />
because <strong>of</strong> <strong>the</strong> opportunity to experiment at a larger scale than in <strong>the</strong><br />
laboratory. This project allowed us to collect exhaustive data on <strong>sea</strong> <strong>urchin</strong><br />
grow when age and genetic origin (artificial fertilizations) are known.<br />
These results revive <strong>the</strong> "organic growth" (sensu von Bertalanffy, 1938)<br />
<strong>model</strong>s in echinoids.<br />
This introduction is organized in four parts. First, we provide a<br />
summary <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> fishery, aquaculture potentials, and markets for<br />
<strong>Paracentrotus</strong> lividus (Lamarck), in <strong>the</strong> economical context that motivated<br />
31
General introduction<br />
this study. Second, a brief review <strong>of</strong> selected aspects <strong>of</strong> its biology<br />
provides essential background for this work. The third section reviews <strong>the</strong><br />
historical development <strong>of</strong> growth <strong>model</strong>s. Finally, use <strong>of</strong> <strong>the</strong>se growth<br />
curves with <strong>sea</strong> <strong>urchin</strong>s is summarized and discussed.<br />
Economical interest <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
a. Sea <strong>urchin</strong> markets and fisheries<br />
The most important market in <strong>the</strong> world is Japan. Sea <strong>urchin</strong> roe (both<br />
male and female gonads, uni in Japanese) are marketed under different<br />
forms: fresh (65%), but also dried, salted, frozen or cooked (35%) (Saito,<br />
1992; Hagen, 1996a). According to various authors, <strong>the</strong> main species<br />
exploited in Japan are Strongylocentrotus intermedius (A. Agassiz), S.<br />
nudus (A. Agassiz), Heterocentrotus pulcherrimus (A. Agassiz),<br />
Pseudocentrotus depressus (A. Agassiz), Anthocidaris crassispina (A.<br />
Agassiz) and Tripneustes gratilla (L.) (Fuji, 1967; Fuji & Kamura, 1970;<br />
Fernandez, 1996; Hagen, 1996a). Both Strongylocentrotus droebachiensis<br />
(Müller) and S. franciscanus (A. Agassiz) are imported from North<br />
America (Kato, 1972; Sloan, 1985), while Loxechinus albus Molina is<br />
imported from Chile (Gonzalez et al, 1993; Lawrence et al, 1997). The<br />
Japanese market is quite stable, around 60,000 tons <strong>of</strong> fresh echinoids per<br />
annum (Hagen, 1996a), since several years and accounts for more than<br />
95% <strong>of</strong> <strong>the</strong> whole world <strong>sea</strong> <strong>urchin</strong> market. Current landings in Japan<br />
average 14,000 tons per year. Japanese imports total approximately 5,000<br />
tons <strong>of</strong> roe under different forms, corresponding to 40 to 50,000 tons <strong>of</strong><br />
live <strong>sea</strong> <strong>urchin</strong>s.<br />
The average price <strong>of</strong> roe on <strong>the</strong> Japanese market ranges from 18.6 €/kg<br />
(750 BEF/kg) for <strong>the</strong> local production (fresh animals considered as top<br />
quality), to 7.9 €/kg (320 BEF/kg) <strong>of</strong> fresh imported echinoids (Hagen,<br />
1996a). These figures equate to a total market <strong>of</strong> approximately 657<br />
32
General introduction<br />
millions €/y (26.5 milliard BEF/y), 397 millions €/y <strong>of</strong> which is<br />
importated.<br />
The second largest market is France. Its landings are much smaller:<br />
about 1,000 tons <strong>of</strong> live echinoids per year in <strong>the</strong> 1960s and 1970s. Since<br />
<strong>the</strong>se peak levels harvests have dropped to 250 to 350 tons per annum<br />
(Allain, 1972a; Ledireac'h, 1987; Le Gall, 1987, 1990). Spain, Ireland and<br />
Greece export to France and compensate for <strong>the</strong> reduced local production,<br />
so <strong>the</strong> market was kept at 500 to 600 tons from 1988 to 1990 (Fernandez,<br />
1996). According to <strong>the</strong> same author, 185 tons transited by Rungis (Paris)<br />
in 1991. The major species in <strong>the</strong> French market is <strong>Paracentrotus</strong> lividus<br />
(Lamarck), but Psammechinus miliaris (Gmelin) and Sphaerechinus<br />
granularis (Lamarck) are also sold. In France, most <strong>sea</strong> <strong>urchin</strong>s are<br />
consumed fresh during <strong>the</strong> period when gonads are in an adequate<br />
reproductive stage, i.e., between December and March (Ledireac'h, 1987).<br />
The <strong>sea</strong>son limits <strong>the</strong> importation market.<br />
Wholesale prices in Rungis fluctuate according to <strong>the</strong> roe quality<br />
(freshness, size, color, maturity stage and taste). It ranges from 4.5 €/y<br />
(180 BEF/y) to 17.8 €/y (720 BEF/kg) (Fernandez, 1996; Grosjean et al,<br />
1998, see Part I). Briton, and to a lesser extent, Irish <strong>sea</strong> <strong>urchin</strong>s are most<br />
valued. Mediterranean strains are <strong>of</strong> lower quality because <strong>the</strong>y do not<br />
withstand travel as well as Briton or Irish strains.<br />
Fishing <strong>sea</strong> <strong>urchin</strong>s is very pr<strong>of</strong>itable during <strong>the</strong> 5 to 10 years after<br />
starting harvesting new stocks. But after that short period <strong>of</strong> time, wild<br />
populations decline due to <strong>the</strong> high efficiency and selectivity <strong>of</strong> fishing<br />
techniques: most exploitable natural stocks are easily picked by hand at<br />
low tide, or at least using simple equipment at shallow depths (Allain,<br />
1972b; Ledireac'h, 1987; Le Gall, 1987). <strong>Growth</strong> speed is also too low in<br />
some harvested species to allow replacement <strong>of</strong> large adults (M. Russell,<br />
pers. com.). Few rules exist to limit overexploitation. Indeed, <strong>the</strong> biggest<br />
problem is that animals must be collected before <strong>the</strong>y fully mature, and<br />
<strong>the</strong>y have no opportunity to spawn. A lack <strong>of</strong> recruitment results from<br />
33
General introduction<br />
intense fishing and, consequently, a rapid decline <strong>of</strong> <strong>the</strong> standing stock<br />
(Allain 1971, 1972a; Le Gall 1990; Campbell & Harbo, 1991). In addition,<br />
removing most adults from a site probably has a negative impact on <strong>the</strong><br />
survival <strong>of</strong> <strong>the</strong> remaining juveniles. The later could be more susceptible to<br />
predation because <strong>the</strong>y are no longer protected by <strong>the</strong> "spine canopy <strong>of</strong><br />
adults" (Tegner & Dayton, 1977).<br />
An example <strong>of</strong> such a decline in landings is <strong>the</strong> Japanese fisheries<br />
which produced 23 to 28,000 tons <strong>of</strong> whole live <strong>sea</strong> <strong>urchin</strong>s per year from<br />
1967 to 1982 (Hagen, 1996a). Landings dropped to 14,000 since 1991,<br />
despite <strong>the</strong> establishment <strong>of</strong> hatcheries (to seed in <strong>the</strong> field) and <strong>of</strong><br />
artificial feeding <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s in harvested areas (Saito, 1992). Chile and<br />
<strong>the</strong> U.S.A. also produce less than before, and only Canada and Korea are<br />
still increasing harvests (Fernandez, 1996) because <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> fisheries<br />
are more recent <strong>the</strong>re. Average worldwide landings are still stable but are<br />
obviously not sustainable in a near future. Aquaculture is a necessary<br />
alternative in all countries with <strong>sea</strong> <strong>urchin</strong> fisheries.<br />
b. Aquaculture potentials<br />
Japan was <strong>the</strong> first country to address <strong>the</strong> issue <strong>of</strong> overexploitation, and<br />
initiated stock enhancement programs very early (Saito, 1992; Hagen,<br />
1996a). These techniques include habitat enhancement (artificial reefs),<br />
artificial feeding, translocation and building <strong>of</strong> hatcheries that produce<br />
several millions <strong>of</strong> seed a year that are transplanted to <strong>the</strong> field. For<br />
instance, a single hatchery in Hokkaido produces 11 million juveniles per<br />
year (Hagen, 1996a). Hatcheries may be a solution to ensure recruitment<br />
where harvesting eliminates adults before <strong>the</strong>y spawn, but good natural<br />
habitats are required, like large tidal pools, to give enough protection to<br />
juveniles released in <strong>the</strong> field (Saito, 1992, Hagen, 1996a).<br />
Ano<strong>the</strong>r way to enhance production is through gonad enhancement.<br />
With an adequate artificial diet, it is possible to increase gonad size (de<br />
Jong-Westman, 1995a; Spirlet, 1999; Spirlet et al, 2000), particularly with<br />
34
General introduction<br />
diets rich in proteins (Klinger et al, 1997; Spirlet, 2001). Gonad<br />
enhancement in culture is a necessity in Canada because <strong>sea</strong> <strong>urchin</strong>s are at<br />
<strong>the</strong> right stage <strong>of</strong> maturity during <strong>the</strong> winter. At this time, <strong>the</strong> <strong>sea</strong> is frozen<br />
and <strong>the</strong> collection <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s under <strong>the</strong> ice by scuba divers is a painful<br />
and dangerous activity. One solution is to collect animals during autumn,<br />
store <strong>the</strong>m in tanks, and feed <strong>the</strong>m with an adequate diet before marketing<br />
<strong>the</strong>m (Motnikar et al, 1997).<br />
The use <strong>of</strong> cages in <strong>sea</strong> ranching operations is also an alternative and<br />
may be used in mono- or polycultures (Keats et al, 1983; Kelly et al,<br />
1998). As for any mariculture activity, degradation <strong>of</strong> cages by waves and<br />
storms is a major problem, and site location is critical. Suitable sites are<br />
limited, and <strong>the</strong>re is <strong>of</strong>ten strong competition for space with o<strong>the</strong>r<br />
mariculture activities like salmoniculture or mytiliculture on long lines.<br />
Because <strong>of</strong> <strong>the</strong>ir grazing activity, <strong>sea</strong> <strong>urchin</strong>s erode <strong>the</strong> cage nets and are<br />
also a direct cause <strong>of</strong> depredation which increases maintenance costs<br />
(Kelly et al, 1998).<br />
The ultimate step in <strong>the</strong> aquaculture production <strong>of</strong> <strong>sea</strong> <strong>urchin</strong> is<br />
independence from natural resources, that is, to control <strong>the</strong> whole life cycle<br />
in culture, from spawning to gonad enhancement (Le Gall, 1990; Hagen,<br />
1996a). This is <strong>the</strong> goal we established for <strong>the</strong> experimental facility in<br />
Normandy. It is called "closed-cycle echiniculture" (Grosjean et al, 1998,<br />
see Part I). Somatic growth <strong>of</strong> juveniles untill <strong>the</strong>y reach market size is a<br />
process that requires major improvements in current technology and is key<br />
to <strong>the</strong> successful development <strong>of</strong> closed-cycle echiniculture. The present<br />
work is devoted to achieving this goal.<br />
Overview <strong>of</strong> <strong>the</strong> biology <strong>of</strong> <strong>Paracentrotus</strong> lividus<br />
The common European <strong>sea</strong> <strong>urchin</strong>, <strong>Paracentrotus</strong> lividus (Lamarck,<br />
1816) (Echinodermata : Echinoidea : Echinidae) is a marine invertebrate<br />
that lives along European coasts <strong>of</strong> <strong>the</strong> North Atlantic (Ireland, Brittany,<br />
Spain) and troughout <strong>the</strong> Mediterranean Sea. It colonizes two types <strong>of</strong><br />
35
General introduction<br />
habitats: intertidal (or sometimes subtidal) rocky shores (Atlantic,<br />
Mediterranean <strong>sea</strong>) and Posidonia oceanica (L.) beds (Mediterranean <strong>sea</strong>)<br />
(Fernandez, 1996).<br />
Figure 1. <strong>Paracentrotus</strong> lividus in a tidal pool in Morgat, Brittany, France. Echinoids are<br />
hardly visible (dark patches) being partly burrowed in cracks and holes in <strong>the</strong> rock and<br />
hidden by stones, empty patellid shells and <strong>sea</strong>weed fragments (covering behavior).<br />
In rocky shores, echinoids have a burrowing behavior (Fig. 1). The<br />
holes <strong>the</strong>y bore in <strong>the</strong> rock protect <strong>the</strong>m from waves and predators.<br />
Juveniles and adults are abundant in tidal pools close to algal fields<br />
(Laminaria spp., but mainly Laminaria digitata Lamouroux in Briton and<br />
Irish coasts). These sedentary populations feed on drift algae fragments<br />
brought in <strong>the</strong> pools by waves and water currents. Some infratidal dense<br />
populations also exist, but <strong>the</strong>y exhibit <strong>the</strong> same sedentary and hidden<br />
behavior (Grosjean, pers. obs.).<br />
In contrast, in <strong>the</strong> Mediterranean Sea, most populations grow in<br />
Posidonia oceanica beds where <strong>the</strong>y exhibit a circadian cycle <strong>of</strong> activity.<br />
36
General introduction<br />
During <strong>the</strong> day, <strong>the</strong>y stay near <strong>the</strong> roots <strong>of</strong> <strong>the</strong> plants, away from predators.<br />
At night, <strong>the</strong>y climb to <strong>the</strong> top <strong>of</strong> <strong>the</strong> fronds and feed on <strong>the</strong>ir s<strong>of</strong>test parts<br />
(Nedelec et al, 1981).<br />
Like almost all echinoids, P. lividus is gonochoristic and fertilization is<br />
external. When animals are ripe in early spring (Allain, 1975; Spirlet et al,<br />
1998a) and in some localities in autumn (Crapp & Willis, 1975;<br />
Fernandez, 1996), spawning is synchronized and triggered by an external<br />
signal, e.g., temperature change or disturbance (Spirlet et al, 1998a;<br />
Spirlet, 1999).<br />
Figure 2. <strong>Paracentrotus</strong> lividus. A. Echinopluteus. B. Postlava a few days after<br />
metamorphosis.<br />
Homolecithal eggs are fertilized in <strong>the</strong> water column and develop into<br />
free-swimming planktotrophic larva characteristic <strong>of</strong> echinoids: an<br />
echinopluteus (Fig. 2A). This larva develops 4, 6 and <strong>the</strong>n 8 arms<br />
supported by calcareous skeletal rods. After a few weeks, <strong>the</strong><br />
echinopluteus will develop a rudiment inside <strong>the</strong> wall <strong>of</strong> an epidermic<br />
invagination (<strong>the</strong> vestibule) located on <strong>the</strong> right-hand side <strong>of</strong> <strong>the</strong> body<br />
(Strathmann, 1978). When <strong>the</strong> larva becomes competent, it seeks a solid<br />
37
General introduction<br />
substrate to settle and metamorphose. An adequate chemical stimulus is<br />
required (Gosselin & Jangoux, 1996). Metamorphosis lasts less than one<br />
hour: <strong>the</strong> echinoid rudiment is evaginated and most larval tissues are<br />
resorbed. The postlarva resembles a miniaturized adult (Fig. 2B) but has<br />
no mouth and no anus, and is thus endotrophic (Gosselin & Jangoux,<br />
1998). After a week <strong>the</strong> postlava has undergone some major changes and<br />
becomes an exotrophic juvenile with a fully developed and functional<br />
digestive tract. It <strong>the</strong>n begins foraging.<br />
Figure 3. External anatomy <strong>of</strong> a regular <strong>sea</strong> <strong>urchin</strong>. A. Oral view. B. Aboral view. (after<br />
Reid, W.M,. In: Ruppert & Barnes, 1994).<br />
From an anatomical point <strong>of</strong> view, <strong>the</strong> body <strong>of</strong> <strong>the</strong> postmetamorphic<br />
echinoid has a quasi-spherical shape (for regular <strong>sea</strong> <strong>urchin</strong>s such as P.<br />
lividus) with a pentaradial symmetry (Fig. 3). Its shape is constrained by<br />
an endoskeleton, located just under <strong>the</strong> epidermis, composed <strong>of</strong> calcareous<br />
ossicles sutured toge<strong>the</strong>r in a solid test. This test supports movable spines<br />
that cover <strong>the</strong> body <strong>of</strong> <strong>the</strong> animal and are <strong>the</strong> origin <strong>of</strong> <strong>the</strong> name<br />
Echinoidea, "like a hedgehog (porcupine)" (Ruppert & Barnes, 1994). P.<br />
38
General introduction<br />
lividus reaches a maximal test diameter <strong>of</strong> 65 to 70 mm (Grosjean, pers.<br />
obs.).<br />
Figure 4. Internal anatomy <strong>of</strong> a regular <strong>sea</strong> <strong>urchin</strong>, side view. (modified after Reid, W.M., In:<br />
Ruppert & Barnes, 1994).<br />
The regular <strong>sea</strong> <strong>urchin</strong> body can be divided in two hemispheres: an oral<br />
pole where <strong>the</strong> mouth opens, directed towards <strong>the</strong> substratum, and an<br />
opposed aboral pole bearing <strong>the</strong> anus. The mouth opens in a short pharynx<br />
surrounded by a complex scraping apparatus –recall that P. lividus is a<br />
grazer– called Aristotle's lantern (Fig. 4). It is composed <strong>of</strong> 5 pyramids<br />
radially arranged around <strong>the</strong> mouth and each holds one tooth (Fig. 3A).<br />
The digestive tract forms two complete turns around <strong>the</strong> inner side <strong>of</strong> <strong>the</strong><br />
test wall, one in one way and <strong>the</strong> o<strong>the</strong>r one in <strong>the</strong> opposite direction,<br />
leaving much space in <strong>the</strong> internal cavity for gonads (Fig. 4).<br />
The anatomy <strong>of</strong> <strong>the</strong> reproductive organs reflects <strong>the</strong> radial symmetry <strong>of</strong><br />
<strong>the</strong> animal (Fig. 5). Five gonads open in genital pores close to <strong>the</strong> anus and<br />
are disposed radially in <strong>the</strong> coelomic cavity along <strong>the</strong> ambulacral zones.<br />
They start developing when <strong>the</strong> echinoid is still very small, around 4 to 6<br />
39
General introduction<br />
mm (Spirlet et al, 1994). P. lividus becomes mature when it reaches a test<br />
diameter <strong>of</strong> 20 to 25 mm (Grosjean, pers. obs.).<br />
Figure 5. Mature adult <strong>Paracentrotus</strong> lividus with oral region removed showing <strong>the</strong> five<br />
gonads. The individual at <strong>the</strong> top is a male, <strong>the</strong> two o<strong>the</strong>rs are females (gonads <strong>of</strong> brighter<br />
color).<br />
Like any echinoderm, <strong>sea</strong> <strong>urchin</strong>s have a water-vascular system which<br />
is used for locomotion. Tube feet (podia) elongate and retract and <strong>the</strong>ir<br />
sucker-shaped tip can glue and unglue to <strong>the</strong> substratum (Flammang, 1996;<br />
Flammang et al, 1998). Locomotion in any direction is sometimes aided by<br />
<strong>the</strong> movements <strong>of</strong> spines. P. lividus exhibits a gregarious behavior and<br />
lives in aggregates where small individuals tend to stay under larger ones<br />
(including inside holes, for populations living in rocky shores). This<br />
behavior was clearly identified as a protective mechanism against<br />
predators (Tegner & Dayton, 1977). Many <strong>sea</strong> <strong>urchin</strong>s species, including<br />
P. lividus, also cover <strong>the</strong>ir exposed aboral surfaces with shells, stones, and<br />
algae for camouflage (Crook et al, 1999; see also Fig. 1).<br />
Echinoids are key-species in several ecosystems such as kelp forests,<br />
barren grounds or Posidonia beds (Tegner & Dayton, 1981; Rowley, 1989;<br />
Fernandez, 1996; Leinass & Christie, 1996). Emson (1984) suggested that<br />
40
Brittany<br />
General introduction<br />
some features <strong>of</strong> echinoderms, namely <strong>the</strong> "position, size and <strong>the</strong> method<br />
<strong>of</strong> formation <strong>of</strong> <strong>the</strong> skeleton" and "<strong>the</strong> substitution <strong>of</strong> collagenous tissues<br />
for muscles", in addition to <strong>the</strong>ir low metabolic rate may have given <strong>the</strong>m<br />
competitive advantages over o<strong>the</strong>r animals. Among echinoderms, <strong>sea</strong><br />
<strong>urchin</strong>s are <strong>the</strong> more mineralized ones, and <strong>the</strong>ir competitive advantage in<br />
some marine ecosystems could probably be summarized by "bone idle – a<br />
recipe for success" as proposed by Emson.<br />
Morgat<br />
Brest<br />
Figure 6: Location <strong>of</strong> <strong>the</strong> sampled <strong>Paracentrotus</strong> lividus population.<br />
Echinoids <strong>reared</strong> in <strong>the</strong> Luc-sur-mer facility originated from a single<br />
population in Morgat, Brittany, France (Fig. 6, but see also Fig. 1). They<br />
were collected at low tides from tidal pools. Individuals used in <strong>the</strong>se<br />
experiments were first or second generation <strong>sea</strong> <strong>urchin</strong>s <strong>reared</strong> from <strong>the</strong><br />
egg.<br />
41
<strong>Growth</strong> <strong>model</strong>s<br />
General introduction<br />
The previous section on <strong>the</strong> biology <strong>of</strong> P. lividus was brief because we<br />
do not need much information to <strong>model</strong> growth. Indeed, when <strong>model</strong>ling<br />
growth, <strong>the</strong> animal is mainly regarded as a black box. We care about its<br />
global change through time, but we do not have to detail all <strong>the</strong> complex<br />
biochemical and physiological processes (feeding, digestion, assimilation,<br />
respiration, excretion, etc…) or anatomical modifications that lead to such<br />
changes. On <strong>the</strong> o<strong>the</strong>r hand, a good understanding <strong>of</strong> <strong>the</strong> various growth<br />
curves and <strong>the</strong>ir respective shapes and properties is required. This section<br />
is thus dedicated to a "portrait gallery" <strong>of</strong> all growth <strong>model</strong>s that have been<br />
used for <strong>sea</strong> <strong>urchin</strong>s.<br />
There are two types <strong>of</strong> growth <strong>model</strong>s in biology: population and<br />
individual. Population <strong>model</strong>s describe <strong>the</strong> change in <strong>the</strong> number <strong>of</strong><br />
individuals through time. A typical population growth <strong>model</strong> is <strong>the</strong> logistic<br />
curve (Verhulst, 1838). Ano<strong>the</strong>r <strong>model</strong> <strong>of</strong>ten used to describe decreasing<br />
number <strong>of</strong> individuals, i.e., mortality, is <strong>the</strong> Weibull function (Weibull,<br />
1951).<br />
Individual growth <strong>model</strong>s represent changes in <strong>the</strong> size <strong>of</strong> a single<br />
individual with time. A typical individual growth <strong>model</strong> is <strong>the</strong> von<br />
Bertalanffy curve (von Bertalanffy, 1938, 1957). Although we will deal<br />
exclusively with individual growth in this work, population growth <strong>model</strong>s<br />
are also <strong>of</strong>ten used to <strong>model</strong> individuals, particularly Gompertz or logistic<br />
curves (for examples that used <strong>the</strong>m for <strong>sea</strong> <strong>urchin</strong>s, see Gage et al, 1986;<br />
Gage, 1987; Ebert, 1999) and we will do so as well.<br />
a. The exponential curve, a simple Malthusian growth <strong>model</strong><br />
In 1798, Thomas Malthus described a ma<strong>the</strong>matical <strong>model</strong> for growth<br />
<strong>of</strong> human populations. According to Murray (1993), this <strong>model</strong> was<br />
previously suggested by Euler. Today this <strong>model</strong> is not used much,<br />
however its historical significance should not be overlooked. It is <strong>the</strong> first<br />
42
20<br />
15<br />
10<br />
Y<br />
5<br />
Y 0<br />
General introduction<br />
ma<strong>the</strong>matical formulation <strong>of</strong> one <strong>of</strong> <strong>the</strong> most fundamental aspects <strong>of</strong><br />
growth: its exponential nature (positive or negative). Malthus observed<br />
that <strong>the</strong> U.S. population doubled every 25 years. He suggested that human<br />
populations increase by a fixed proportion r on a given time interval, when<br />
<strong>the</strong>y are not affected by environmental or social constraints, and this<br />
proportion is not dependent on <strong>the</strong> initial size <strong>of</strong> <strong>the</strong> population:<br />
Yt+ 1 (1 r) Yt k Yt<br />
= + ⋅ = ⋅ (1)<br />
One obtains a continuous <strong>model</strong> by expression eq. 1 as a differential<br />
equation:<br />
dY () t<br />
= Y'( t) = k⋅ Y( t)<br />
(2)<br />
dt<br />
This differential equation results in <strong>the</strong> following solution:<br />
() e kt ⋅<br />
Yt = Y⋅<br />
(3)<br />
0<br />
with Y0 being <strong>the</strong> initial size <strong>of</strong> <strong>the</strong> population at time t = 0 (Fig. 7). This 2parameter<br />
<strong>model</strong> is also interesting because it demonstrates how to build a<br />
growth <strong>model</strong>: write a differential equation <strong>of</strong> <strong>the</strong> variation <strong>of</strong> size with<br />
time and solve it (dynamic <strong>model</strong>ling). Almost all existing growth <strong>model</strong>s<br />
have been constructed this way. They correspond, as a consequence, to a<br />
simple differential equation.<br />
0.5 1 1.5 2 2.5 3 t<br />
Figure 7. Example <strong>of</strong> an exponential curve with Y0 = 1.5 and k = 0.9.<br />
43
. The logistic function for asymptotic growth<br />
Y<br />
1<br />
Y•<br />
0.8<br />
0.6<br />
Y•ê2<br />
0.4<br />
0.2<br />
General introduction<br />
The exponential <strong>model</strong> describes infinite growth without constraints.<br />
This is not a realistic hypo<strong>the</strong>sis. In practice, growth is limited by available<br />
resources. Verhulst (1838), working also on population growth, proposed a<br />
<strong>model</strong> containing an auto-limitation term [Y∞ – Y(t)]/Y∞ that represents<br />
some <strong>the</strong>oretical limiting resource:<br />
dY () t Y∞ −Y<br />
() t k<br />
= k⋅ ⋅ Y t =− Y t + k⋅ Y t<br />
(4)<br />
dt Y Y<br />
()<br />
2<br />
() ()<br />
∞ ∞<br />
Solving and simplifying this differential equation yields:<br />
Y<br />
∞ () −k⋅( t−t0) 1 e<br />
Yt<br />
= +<br />
Eq. 5 is one form <strong>of</strong> <strong>the</strong> logistic function. The function has two<br />
horizontal asymptotes at Y(t) = 0 and Y(t) = Y∞ (Fig. 8) and it is a<br />
symmetrical sigmoid (<strong>the</strong> two limbs <strong>of</strong> <strong>the</strong> S are similar).<br />
i<br />
2 4 6 8 10 t<br />
t<br />
t<br />
0<br />
Figure 8. Example <strong>of</strong> a logistic curve with k = 1, Y∞ = 0.95, t0 = 5. This sigmoidal curve is<br />
asymptotic in 0 and Y∞, and is also symmetrical around its inflexion point i at {t0, Y∞ /2}.<br />
As a generalization <strong>of</strong> this <strong>model</strong>, it is easy to define a logistic function<br />
whose lower asymptote is different from 0. If this lower asymptote is Y0,<br />
we obtain equation:<br />
(5)<br />
44
General introduction<br />
Y∞−Y Yt () = Y+ 1+ e<br />
0<br />
0 −k⋅( t−t0) We will refer to it as <strong>the</strong> 4-parameter logistic <strong>model</strong>.<br />
c. The Gompertz <strong>model</strong>, an asymmetrical sigmoidal curve<br />
Y<br />
1<br />
Y•<br />
0.8<br />
0.6<br />
0.4<br />
Y•êe<br />
0.2<br />
Gompertz (1825) empirically observed that survival rate <strong>of</strong>ten<br />
decreases proportionally to <strong>the</strong> logarithm <strong>of</strong> survival. Although this <strong>model</strong><br />
is still used with survival data (Ebert, 1999), it has many applications for<br />
growth data as well (Winsor, 1932, Ebert, 1999). The differential equation<br />
<strong>of</strong> Gompertz <strong>model</strong> is:<br />
which simplifies to:<br />
dY () t<br />
= k⋅[ lnY∞−ln Y( t) ] ⋅ Y( t)<br />
(7)<br />
dt<br />
−k⋅( t−t0) e<br />
−kt ⋅<br />
e<br />
t<br />
b<br />
∞ ∞ ∞<br />
Yt () = Y.e = Y. a = Y. a<br />
(8)<br />
The last parameterization is simpler and used more <strong>of</strong>ten. The first one is<br />
derived from <strong>the</strong> differential equation (eq. 7) and gives a better comparison<br />
with <strong>the</strong> logistic <strong>model</strong>, since t0 also corresponds to <strong>the</strong> abscissa <strong>of</strong> <strong>the</strong><br />
inflexion point, which is not in a symmetrical position here (Fig. 9).<br />
t 0<br />
i<br />
1 2 3 4 5 6<br />
Figure 9. Example <strong>of</strong> a Gompertz curve with k = 1, Y∞ = 0.95, t0 = 1.5. Inflexion point i, at<br />
{t0, Y∞ /e}, is lower than in <strong>the</strong> logistic curve.<br />
t<br />
(6)<br />
45
d. The von Bertalanffy curves<br />
General introduction<br />
The von Bertalanffy <strong>model</strong>, sometimes called Brody-Bertalanffy<br />
(according to works <strong>of</strong> von Bertalanffy, 1938, 1957, and Brody, 1945) or<br />
Pütter (in Ricker, 1979), is <strong>the</strong> first growth <strong>model</strong> specifically designed to<br />
describe individuals. It is based on a simple bioenergetic analysis. An<br />
individual is regarded as a simple dynamic chemical reactor where inputs<br />
(anabolism) compete with outputs (catabolism). The result <strong>of</strong> <strong>the</strong>se two<br />
fluxes is growth. Anabolism is more or less proportional to respiration, and<br />
respiration is surface-proportional for many animals (von Bertalanffy,<br />
1957). Catabolism is always proportional to <strong>the</strong> volume or weight. These<br />
mechanistic relationships are collected toge<strong>the</strong>r in <strong>the</strong> following<br />
differential equation when Y(t) measures a volume or a weight with time:<br />
dY () t<br />
2/3 1/3 2/3<br />
= aYt ⋅ () −bYt ⋅ () = 3 k⋅⎡Y∞⋅Yt () −Yt<br />
() ⎤<br />
dt<br />
⎣ ⎦ (9)<br />
Solving this equation, we obtain <strong>the</strong> von Bertalanffy <strong>model</strong> in weight, also<br />
called "von Bertalanffy 2" in <strong>the</strong> next part <strong>of</strong> this work:<br />
−k⋅( t−t0) ( )<br />
3<br />
Yt () = Y ⋅ 1− e (10)<br />
∞<br />
The simplest form <strong>of</strong> this <strong>model</strong> occurs when one measures a linear<br />
dimension for <strong>the</strong> body size, since a linear dimension is <strong>the</strong> cubic root <strong>of</strong> a<br />
volume or a weight (not considering a possible allometry). The von<br />
Bertalanffy for linear measurements, called von Bertalanffy 1 in <strong>the</strong><br />
present work, is simply:<br />
−k⋅( t−t0) ( )<br />
Yt () = Y⋅<br />
1− e (11)<br />
∞<br />
A graph <strong>of</strong> both <strong>model</strong>s is shown in Fig. 10. Von Bertalanffy 1 <strong>model</strong><br />
has no inflexion point. <strong>Growth</strong> is fastest at <strong>the</strong> outset, gradually<br />
diminishes, and finally reaches zero. <strong>Growth</strong> is determinate and size<br />
cannot exceed <strong>the</strong> horizontal asymptote <strong>of</strong> <strong>the</strong> curve at Y(t) = Y∞. Due to<br />
46
Y<br />
Y<br />
1<br />
Y• 0.8<br />
0.6<br />
0.4<br />
0.2<br />
General introduction<br />
<strong>the</strong> cubic power transformation <strong>of</strong> von Bertalanffy 1, von Bertalanffy 2 is<br />
an asymmetrical sigmoid like <strong>the</strong> Gompertz <strong>model</strong>.<br />
1 2 3 4 5 6<br />
Figure 10. Both von Bertalanffy 1 (curve in bold) and von Bertalanffy 2 (plain curve) <strong>model</strong>s<br />
with k = 1, Y∞ = 0.95 and t0 = 0. Both <strong>model</strong>s describe asymptotic growth, but von<br />
Bertalanffy 1 has no inflexion point, while von Bertalanffy 2 is sigmoidal.<br />
e. The Richards <strong>model</strong>, a flexible curve that contains many<br />
o<strong>the</strong>rs<br />
The general scheme for von Bertalanffy <strong>model</strong>s is:<br />
t<br />
−k⋅( t−t0) ( )<br />
m<br />
Yt () = Y⋅<br />
1− e<br />
(12)<br />
∞<br />
Von Bertalanffy (1938, 1957) set m at ei<strong>the</strong>r 1 or 3. Richards (1959) lets m<br />
vary freely and thus his <strong>model</strong> has an additional parameter. This curve is<br />
very flexible (Fig. 11) and one can demonstrate several o<strong>the</strong>r growth<br />
<strong>model</strong>s are just special cases with different m values. We have already<br />
observed it reduces to von Bertalanffy 1 when m = 1, and to von<br />
Bertalanffy 2 when m = 3. It also reduces to <strong>the</strong> logistic curve when m = -1<br />
and one can demonstrate it reduces to <strong>the</strong> Gompertz <strong>model</strong> when |m| → ∞<br />
(Tomassone et al, 1993; Ebert, 1999).<br />
47
Y<br />
Y<br />
1<br />
Y•<br />
0.8<br />
0.6<br />
0.4<br />
0.2<br />
General introduction<br />
2 4 6 8 10 t<br />
Figure 11. Shape <strong>of</strong> Richards curves depending on values <strong>of</strong> m. From left to right: m = 0.5,<br />
1, 3, 6 and 10; with k = 0.5, Y∞ = 0.95 and t0 = 0 for all curves. The curve in bold, with<br />
m = 1, is equivalent to <strong>the</strong> von Bertalanffy 1 <strong>model</strong>.<br />
f. The Weibull <strong>model</strong>, a polyvalent and flexible function<br />
Since it was introduced, <strong>the</strong> Weibull (1951) function was presented as<br />
a polyvalent <strong>model</strong>. Originally, it was described as a statistical distribution.<br />
It has many applications in population (negative) growth, and is used also<br />
to describe survival in cases <strong>of</strong> injury or di<strong>sea</strong>se, or in population dynamic<br />
studies (Fahrmeir & Tutz, 1994; Ebert, 1985, 1999). It is sometimes used<br />
as a growth <strong>model</strong> (Tomassone et al, 1993). The most general form <strong>of</strong> this<br />
<strong>model</strong> is:<br />
m<br />
∞ ∞<br />
−kt ⋅<br />
Yt ( ) = Y −d⋅ e with d= Y − Y<br />
(13)<br />
A 3-parameter <strong>model</strong> is used as well with Y0 = 0 (Tomassone et al, 1993).<br />
The function is sigmoidal when m > 1, o<strong>the</strong>rwise it has no inflexion point<br />
(Fig. 12).<br />
0<br />
48
Y<br />
1<br />
Y• 0.8<br />
0.6<br />
Y•-d.e 0.4<br />
-k<br />
0.2<br />
Y0<br />
General introduction<br />
1<br />
i<br />
2 4 6 8 10 tt<br />
Figure 12. Examples <strong>of</strong> Weibull curves for m = 5, 2, 1 and 0.5 respectively, with k = 0.6,<br />
Y∞ = 0.95 and Y0 = 0.05. In bold, <strong>the</strong> curve with m = 1, equivalent to a von Bertalanffy 1<br />
<strong>model</strong>. All curves start from Y0 and pass by Y∞ - d·e -k which is also <strong>the</strong> inflexion point for <strong>the</strong><br />
sigmoidal curves when m > 1.<br />
g. The Jolicoeur curve, ano<strong>the</strong>r flexible <strong>model</strong><br />
Ano<strong>the</strong>r curve that can possibly be sigmoid or not is Jolicoeur <strong>model</strong><br />
(Jolicoeur, 1985). It is derived from a logistic curve, but using <strong>the</strong><br />
logarithm <strong>of</strong> time instead <strong>of</strong> time itself:<br />
Y∞<br />
Yt () =<br />
1+<br />
bt ⋅<br />
−m<br />
(14)<br />
Like <strong>the</strong> Weibull function, it is sigmoidal when m > 1, but with an<br />
asymmetry between <strong>the</strong> limbs <strong>of</strong> <strong>the</strong> S that can vary independently,<br />
according to <strong>the</strong> value <strong>of</strong> parameter b. When m ≤ 1, <strong>the</strong>re is no inflexion<br />
point (Fig. 13).<br />
49
Y<br />
1<br />
Y•<br />
0.8<br />
0.6<br />
Y•êH1+bL<br />
0.4<br />
0.2<br />
General introduction<br />
1<br />
i<br />
2 4 6 8 10 t<br />
t<br />
Figure 13. Examples <strong>of</strong> Jolicoeur curves with m = 3, 1 (bold curve) and 0.5 respectively from<br />
highest to lowest curve; Y∞ = 0.95 and b = 0.9. When m > 1, <strong>the</strong> curve is sigmoidal.<br />
h. The Johnson <strong>model</strong>, a heavily asymmetrical sigmoid<br />
Y<br />
Y<br />
1<br />
Y• 0.8<br />
0.6<br />
0.4<br />
0.2<br />
The Johnson growth curve (see Ricker, 1979) uses 1/t instead <strong>of</strong> t:<br />
Yt () Y e<br />
2 4 6 8 10 t<br />
1 k⋅( t−t0) = ∞ ⋅ (15)<br />
Figure 14. Example <strong>of</strong> a Johnson curve, with k = 0.7, Y∞ = 0.95 and t0 = 0.<br />
It is sigmoidal with a very strong asymmetry, <strong>the</strong> inflexion point being<br />
very low and close to 0 (and thus hardly visible in Fig. 14).<br />
50
i. The Preece-Baines 1 <strong>model</strong> for human growth<br />
Y<br />
1<br />
Y• 0.8<br />
0.6<br />
0.4<br />
0.2<br />
General introduction<br />
Preece and Baines (1978) described various growth <strong>model</strong>s specific to<br />
human growth. These <strong>model</strong>s combine two different exponential growth<br />
phases to represent <strong>the</strong> gradual growth <strong>of</strong> infants followed by a faster<br />
growth <strong>of</strong> adolescents, but becoming rapidly asymptotic (Fig. 15). These<br />
are indeed very flexible <strong>model</strong>s. Among <strong>the</strong>se curves, <strong>model</strong> 1 was used<br />
for a <strong>sea</strong> <strong>urchin</strong> by Gage and Tyler (1985) and is:<br />
2( ⋅ Y∞−d) Yt () = Y −<br />
e + e<br />
∞ k1⋅( t−t0) k2⋅( t−t0) 2 4 6 8 10 t<br />
Figure 15. Example <strong>of</strong> a Preece-Baines 1 curve with k1 = 0.19, k2 = 2.5, Y∞ = 0.95, d = 0.8<br />
and t0 = 6.<br />
j. The Tanaka <strong>model</strong> for indeterminate growth<br />
(16)<br />
All <strong>the</strong> previous <strong>model</strong>s are asymptotic, except <strong>the</strong> exponential one<br />
(but it is only usable for <strong>the</strong> initial stages <strong>of</strong> a growth process). They<br />
describe determinate growth that can never exceed horizontal asymptote at<br />
Y(t) = Y∞. Knight (1968) questioned whe<strong>the</strong>r it is a biological fact or just a<br />
ma<strong>the</strong>matical artifact. In <strong>the</strong> later case, growth seems to be determinate<br />
only because ma<strong>the</strong>matical <strong>model</strong>s used to represent it are asymptotical.<br />
To overcome this constraint, Tanaka (1982, 1988) elaborated a new <strong>model</strong><br />
that allows for indeterminate growth:<br />
51
YY<br />
3<br />
2.5<br />
2<br />
1.5<br />
1<br />
0.5<br />
General introduction<br />
⎛ 1 ⎞<br />
2 2<br />
Yt () = ⎜ ⎟⋅ln2<br />
b⋅( t− t0) + 2 b⋅( t− t0) + ab ⋅ + d<br />
⎝ b ⎠<br />
(17)<br />
This complex 4-parameter <strong>model</strong> has an initial period <strong>of</strong> slow growth, a<br />
period <strong>of</strong> exponential growth followed by an indefinite period <strong>of</strong> slow<br />
growth (Fig. 16).<br />
2 4 6 8 10 tt<br />
Figure 16. Example <strong>of</strong> a Tanaka curve with a = 3, b = 2.5, d = -0.2 and t0 = 2.<br />
Modelling <strong>sea</strong> <strong>urchin</strong>s growth<br />
Table 1 summarizes <strong>sea</strong> <strong>urchin</strong> studies using growth <strong>model</strong>s.<br />
Numerous works using growth rate only (final – initial size) are not<br />
included. Most species <strong>of</strong> economic interest (Tripneustes gratilla,<br />
Sphaerechinus granularis, Psammechinus miliaris, <strong>Paracentrotus</strong> lividus,<br />
Strongylocentrotus droebachiensis, S. intermedius, S. nudus, S.<br />
franciscanus, Hemicentrotus pulcherrimus) were considered by one or<br />
more authors. They focused ei<strong>the</strong>r on population dynamics aiming to<br />
provide management criteria for fisheries, or on growth in cultivation.<br />
O<strong>the</strong>r species studied were ei<strong>the</strong>r key-species in some biotopes (Diadema<br />
antillarum Philippi, Echinus esculentus L.), or animals occupying some<br />
particular biotopes [for instance, deep-<strong>sea</strong> <strong>urchin</strong>s like Echinosigra phiale<br />
(Thompson) or Hemiaster expergitus Loven].<br />
52
a. Choice <strong>of</strong> <strong>the</strong> growth <strong>model</strong> for <strong>sea</strong> <strong>urchin</strong>s<br />
General introduction<br />
Von Bertalanffy 1 is <strong>the</strong> <strong>model</strong> most <strong>of</strong>ten used. Among 69 studies <strong>of</strong><br />
<strong>sea</strong> <strong>urchin</strong>s (regardless <strong>of</strong> species), this <strong>model</strong> was used 32 times, <strong>the</strong><br />
Richards <strong>model</strong> 17 times, <strong>the</strong> Gompertz <strong>model</strong> 9 times and o<strong>the</strong>r <strong>model</strong>s<br />
11 times (Table 1). However, in 13 studies where several <strong>model</strong>s were<br />
tested in addition to von Bertalanffy 1, <strong>the</strong> latter was considered as being<br />
<strong>the</strong> best one only twice. The reason invoked to reject <strong>the</strong> von Bertalanffy<br />
<strong>model</strong> was <strong>the</strong> initial lag phase in growth that is correctly represented<br />
solely by a sigmoid like in <strong>the</strong> Richards, Gompertz or logistic <strong>model</strong>s<br />
(Yamagushi, 1975). In many studies where no o<strong>the</strong>r <strong>model</strong> was tested, it<br />
seems that <strong>the</strong> von Bertalanffy 1 curve was just a default choice: it<br />
represents <strong>the</strong> <strong>model</strong> "usually" fitted on such kind <strong>of</strong> data.<br />
The Richards <strong>model</strong> was first proposed by Ebert (1973) as a better<br />
alternative to <strong>the</strong> von Bertalanffy 1 curve to fit echinoid growth data. It<br />
was intensively used by <strong>the</strong> same author (Ebert, 1973, 1980a, 1982, 1999;<br />
Ebert & Russell, 1992, 1993) as well as by some o<strong>the</strong>rs (Gage & Tyler,<br />
1985; Russell, 1987; Kenner, 1992; Turon et al, 1995; Lamare &<br />
Mladenov, 2000). The Gompertz <strong>model</strong> is also a favorite when <strong>the</strong>re seems<br />
to be a lag phase in growth and it has been used in various studies (Gage et<br />
al, 1986; Gage, 1987; Cellario & Fenaux, 1990; Dafni, 1992; Turon et al,<br />
1995; Ebert, 1999). Each <strong>of</strong> <strong>the</strong>se <strong>model</strong>s (i.e., Richards or Gompertz) was<br />
preferred in 50% <strong>of</strong> <strong>the</strong> multi-<strong>model</strong> studies which considered <strong>the</strong>m. The<br />
logistic curve, although also sigmoidal, was systematically rejected in<br />
multi-<strong>model</strong>s studies <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s, except for <strong>the</strong> irregular echinoid<br />
Echinocardium pennatifidum Norman (Gage, 1987).<br />
53
Table 1. Models used to fit <strong>sea</strong> <strong>urchin</strong>s growth data (<strong>the</strong> one preferred by authors in multi<strong>model</strong>s<br />
studies is in bold).<br />
Species (1)<br />
<strong>Growth</strong> <strong>model</strong> (2)<br />
Family Cidaridae<br />
Reference<br />
Eucidaris tribuloides (Lamarck) von Bertalanffy1 McPherson, 1968<br />
Family Diadematidae<br />
Diadema setosum (Leske) Richards Ebert, 1980a<br />
von Bertalanffy1, Richards, Gompertz, logistic Ebert, 1999<br />
Diadema antillarum Philippi von Bertalanffy1 Ebert, 1975<br />
Echinotrix diadema (L.) von Bertalanffy1 Ebert, 1975<br />
Richards Ebert, 1982<br />
Centrostephanus rodgersii (A. Agassiz) Richards Ebert, 1982<br />
Family Stomopneustidae<br />
Stomopneustes variolaris (Lamarck) Richards Ebert, 1982<br />
Family Temnopleuridae<br />
Salmacis belli Döderlein Richards Ebert, 1982<br />
Family Toxopneustidae<br />
Lytechinus variegatus (Lamarck) von Bertalanffy1 Ebert, 1975<br />
Tripneustes gratilla (L.) von Bertalanffy1, Gompertz, logistic, Johnson Dafni, 1992<br />
Tripneustes ventricosus (Lamarck) von Bertalanffy1 McPherson, 1965<br />
Sphaerechinus granularis (Lamarck) von Bertalanffy1 Lumingas & Guillou, 1994<br />
von Bertalanffy1 Jordana et al, 1997<br />
Family Echinidae<br />
Echinus esculentus L. Richards Ebert, 1973<br />
logistic Nichols et al, 1985<br />
logistic Sime & Cranmer, 1985<br />
Gompertz Gage et al, 1986<br />
von Bertalanffy1 Gage, 1992<br />
Echinus acutus Lamarck logistic Sime & Cranmer, 1985<br />
von Bertalanffy1, Gompertz, logistic Gage et al, 1986<br />
Echinus elegans Düben & Koren von Bertalanffy1, Gompertz, logistic Gage et al, 1986<br />
Echinus affinis Mortensen von Bertalanffy1, Richards, Gompertz,<br />
logistic, Preece-Baines 1<br />
Gage & Tyler, 1985<br />
Gompertz Gage et al, 1986<br />
linear Middleton et al, 1998<br />
Psammechinus miliaris (Gmelin) von Bertalanffy1 Jensen, 1969a<br />
von Bertalanffy1 Allain, 1978<br />
von Bertalanffy1 Gage, 1991<br />
<strong>Paracentrotus</strong> lividus (Lamarck) von Bertalanffy1 Allain, 1978<br />
(von Bertalanffy1), Gompertz Cellario & Fenaux, 1990<br />
Gompertz, logistic, Richards Turon et al, 1995<br />
von Bertalanffy1 Sellem et al, 2000<br />
von Bertalanffy1, von Bertalanffy2, Gompertz, Grosjean et al (submitted),<br />
logistic, 4p-logistic, Weibull, original <strong>model</strong> see Part. IV<br />
Loxechinus albus Molina von Bertalanffy1 Gebauer & Moreno, 1995<br />
Family Strongylocentrotidae<br />
Strongylocentrotus droebachiensis (Müller) von Bertalanffy1 Munk, 1992<br />
von Bertalanffy1 Hagen, 1996b<br />
logistic Meidel & Scheibling, 1998<br />
Tanaka Russell et al, 1998<br />
Strongylocentrotus intermedius (A. Agassiz) von Bertalanffy1, Gompertz Fuji, 1967<br />
Strongylocentrotus nudus (A. Agassiz) von Bertalanffy1 Ebert, 1975<br />
Strongylocentrotus purpuratus (Stimpson) von Bertalanffy1 Ebert, 1977<br />
Richards Russell, 1987<br />
Richards Kenner, 1992<br />
Strongylocentrotus franciscanus (A. Agassiz) von Bertalanffy1 Ebert, 1977<br />
Richards Ebert & Russell, 1992<br />
Richards, Tanaka, Jolicoeur Ebert & Russell, 1993<br />
Tanaka Ebert, 1998<br />
Richards, Tanaka Ebert, 1999<br />
Hemicentrotus pulcherrimus (A. Agassiz) von Bertalanffy1 Fuji, 1963<br />
Allocentrotus fragilis (Jackson) von Bertalanffy1 Sumich & McCauley, 1973<br />
General introduction<br />
54
(Table 1, next part)<br />
Species (1)<br />
<strong>Growth</strong> <strong>model</strong> (2)<br />
Family Echinometridae<br />
Reference<br />
Evechinus chloroticus (Valenciennes) von Bertalanffy1, Richards, Tanaka, Jolicoeur Lamare & Mladenov, 2000<br />
Anthocidaris crassispina (A. Agassiz) von Bertalanffy1 Chiu, 1990<br />
Heliocidaris erythrogamma (Valenciennes) Richards Ebert, 1982<br />
Echinometra mathaei (de Blainville) von Bertalanffy1 Ebert, 1975<br />
Richards Ebert, 1982<br />
Echinometra oblonga (de Blainville) von Bertalanffy1 Ebert, 1975<br />
Richards Ebert, 1982<br />
Heterocentrotus mammillatus (Klein)<br />
Heterocentrotus trigonarius (Lamarck)<br />
Richards<br />
Richards<br />
Ebert, 1982<br />
Ebert, 1982<br />
Colobocentrotus atratus (L.) von Bertalanffy1 Ebert, 1975<br />
Richards Ebert, 1982<br />
Family Mellitidae<br />
Mellita quinquiesperforata (Leske) von Bertalanffy1 Lane & Lawrence, 1980<br />
Mellita grantii Mortensen von Bertalanffy1 Ebert & Dexter, 1975<br />
Encope grantis L. Agassiz von Bertalanffy1 Ebert & Dexter, 1975<br />
Family Pourtalesiidae<br />
Echinosigra phiale (Thompson) von Bertalanffy1, Gompertz, logistic Gage, 1987<br />
Family Hemiasteridae<br />
Hemiaster expergitus Loven von Bertalanffy1, Gompertz, logistic Gage, 1987<br />
Family Spatangidae<br />
Spatangus purpureus Müller von Bertalanffy1, Gompertz, logistic Gage, 1987<br />
Family Loveniidae<br />
Echinocardium cordatum (Pennant) von Bertalanffy1 Duineveld & Jenness, 1984<br />
Echinocardium pennatifidum Norman von Bertalanffy1, Gompertz, logistic Gage, 1987<br />
(1) Classification according to Mortensen (1950) and Durham (1955).<br />
(2) von Bertalanffy1: D = a·(1 - e -b·(t – c) ), von Bertalanffy2: D = a·(1 - e -b·(t – c) ) 3 , Richards: D = a·(1 - e -b·(t – c) ) d , Gompertz:<br />
t<br />
c<br />
= ⋅ , logistic: D = a/(1 + e -b·(t - c) ), 4p-logistic: D = (a – d)/(1 + e -b·(t - c) ) + d, Johnson: D = a·e -1/b·(t – c) , Preece-Baines 1:<br />
D a b<br />
D = a – 2·(a – d)/(e b·(t – c) + e e·(t – c) c<br />
−bt ⋅<br />
), linear: D = a·t + b, Weibull: D= a−d⋅ e , original <strong>model</strong>: D = e + a·(1 - e -b·t )/(1 + d·e -c·t )<br />
(see Part IV), Tanaka: D = (1/b 1/2 )·ln(|2b·(t – c) + 2·(b 2 ·(t – c) 2 + a·b) 1/2 | + d), Jolicoeur: D = a/(1 - c·t -b ).<br />
General introduction<br />
Questioning asymptotic growth in <strong>the</strong> largest regular <strong>sea</strong> <strong>urchin</strong>,<br />
Strongylocentrotus franciscanus, Ebert & Russell (1993) introduced <strong>the</strong><br />
indeterminate growth <strong>model</strong> <strong>of</strong> Tanaka as a better representation <strong>of</strong> <strong>the</strong><br />
continuous growth <strong>of</strong> large individuals. However, this species seems to be<br />
a special case, even inside <strong>the</strong> Strongilocentrotidae family (Lawrence et al,<br />
1995). The Tanaka <strong>model</strong> was not used much for o<strong>the</strong>r species. Lamare &<br />
Mladenov (2000) tested it on Evechinus chloroticus (Valenciennes), but<br />
concluded it is not <strong>the</strong> more appropriate one in this particular case.<br />
In an attempt to find a better <strong>model</strong> to fit echinoid growth data, various<br />
"exotic" curves were also tested. They were sometimes successful, such as<br />
in <strong>the</strong> works by Gage & Tyler (1985) that introduced <strong>the</strong> Preece & Baines<br />
55
General introduction<br />
<strong>model</strong> 1 for Echinus affinis Mortensen or by Dafni (1992) that proposed<br />
<strong>the</strong> Johnson <strong>model</strong> to fit rapid growth with a very small initial lag phase <strong>of</strong><br />
<strong>the</strong> Toxopneustidae Tripneustes gratilla at Elat.<br />
Based on this review (Table 1), it seems <strong>the</strong>re is no ideal <strong>model</strong> to fit<br />
growth data in <strong>sea</strong> <strong>urchin</strong>s and, consequently, <strong>the</strong> use <strong>of</strong> one particular<br />
<strong>model</strong> is more a question <strong>of</strong> personal preference. For instance, Ebert uses<br />
<strong>the</strong> Richards <strong>model</strong> most <strong>of</strong> <strong>the</strong> time (Ebert, 1973, 1980a, 1982, 1999),<br />
while Gage favors <strong>the</strong> Gompertz <strong>model</strong> (Gage & Tyler, 1985; Gage et al,<br />
1986; Gage, 1987). This situation is problematic since parameters derived<br />
from <strong>the</strong>se <strong>model</strong>s –and o<strong>the</strong>rs– are not comparables. Using a single <strong>model</strong><br />
to fit all growth data would be preferable for comparison purposes (Turon<br />
et al, 1995).<br />
b. Fitting <strong>of</strong> growth <strong>model</strong>s on real data for echinoids<br />
In selecting a growth <strong>model</strong>, several conditions should be met when<br />
fitting real data. First, animals sampled from a single homogeneous<br />
population should be measured at various ages. Second, one particular<br />
individual should be measured only once to ensure independence <strong>of</strong> <strong>the</strong><br />
errors (since authors consider individual variation as part <strong>of</strong> <strong>the</strong> error term:<br />
<strong>the</strong>y look for growth <strong>of</strong> a virtual "mean individual" among <strong>the</strong> population).<br />
Third, as most regression methods assume no error on <strong>the</strong> dependent<br />
variable –that is, time– (Sokal & Rohlf, 1981; Sen & Srivastava, 1990;<br />
Draper & Smith, 1998; Zar, 1999), <strong>the</strong> age <strong>of</strong> each measured individual<br />
should be known. Fourth, no interaction should exist between individuals<br />
in <strong>the</strong> population. Such ideal conditions are so restrictive that <strong>the</strong>y are<br />
never met.<br />
When <strong>the</strong> age <strong>of</strong> individuals can be determined precisely, such as for<br />
<strong>sea</strong> <strong>urchin</strong>s <strong>reared</strong> from <strong>the</strong> egg, it is common to measure <strong>the</strong> same<br />
specimens several times (Bull, 1938; Michel, 1984; Cellario & Fenaux,<br />
1990; Basuyaux & Blin, 1998; Lamare & Mladenov, 2000). Errors are<br />
individual-dependent in such cases and this interaction is <strong>of</strong>ten ignored.<br />
56
General introduction<br />
Among various methods developed to fit growth curves (Walford, 1946;<br />
Fabens, 1965; Allen, 1966; Causton, 1969; Ebert, 1980a; Kaufmann,<br />
1981), <strong>the</strong> most powerful one is considered to be <strong>the</strong> nonlinear least-square<br />
regression (Gallucci & Quinn, 1979; Vaughan & Kanciruk, 1982). All<br />
studies listed in Table 1 use it, except Grosjean et al (submitted, see Part<br />
IV). However, when using field-collected data, it is not possible to<br />
determine precisely <strong>the</strong> age <strong>of</strong> individuals and thus it is estimated. Yet, <strong>the</strong><br />
same nonlinear least-square regression method is still used despite <strong>the</strong><br />
violation <strong>of</strong> <strong>the</strong> assumption that <strong>the</strong>re is no error on <strong>the</strong> time variable. Two<br />
methods coexist to estimate <strong>the</strong> age <strong>of</strong> echinoids in <strong>the</strong> field: cohort<br />
separation and growth rings analysis.<br />
The cohort separation method is commonly used with species<br />
displaying annual recruitment (Hasselbald, 1966; McDonald & Pitcher,<br />
1979; Ebert et al, 1993, 1999; Aksland, 1994; Smith et al, 1998). Cohorts<br />
are separated by time increments <strong>of</strong> one year. Usually, <strong>the</strong> youngest<br />
individuals recruited in <strong>the</strong> year form a well-defined cohort whose peak<br />
displacement with time can be used to estimate mean growth rate<br />
(Raymond & Scheibling, 1987; Dafni, 1992; Munk, 1992; Lumingas &<br />
Guillou, 1994; Gebauer & Moreno, 1995). The cohorts are –implicitly–<br />
assumed to be unimodal and normally, or at least, symmetrically<br />
distributed. No authors working on <strong>sea</strong> <strong>urchin</strong>s tested <strong>the</strong> validity <strong>of</strong> this<br />
fundamental assumption, nor do <strong>the</strong>y discuss implications <strong>of</strong> violations <strong>of</strong><br />
this assumption. When individuals interact, cohorts can be asymmetrical,<br />
or even multimodal (Grosjean et al, 1996, see Part III). In this case, <strong>the</strong><br />
study is biased because several modes <strong>of</strong> a single cohort are interpreted<br />
as multiple separate cohorts. Ebert (1968) observed wide variations in<br />
growth rate <strong>of</strong> Strongylocentrotus purpuratus (Stimpson) in some habitats,<br />
and attributed it to food limitation. However, <strong>the</strong>re is also some evidence<br />
<strong>of</strong> growth inhibition <strong>of</strong> juveniles in <strong>the</strong> field for Strongylocentrotus<br />
droebachiensis (Himmelman, 1986). Kenner (1992), Munk (1992) and<br />
Turon et al (1995) discuss Himmelman's conclusions for <strong>the</strong>ir biological<br />
implications but do not question <strong>the</strong> validity <strong>of</strong> <strong>the</strong> cohorts separation<br />
57
General introduction<br />
method used. In addition, Levitan (1988) demonstrated that interactions<br />
exist between adult Diadema antillarum as maximal size is densitydependent.<br />
The second method to estimate age uses <strong>the</strong> natural growth bands. The<br />
trabecules within <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> skeleton are more or less densely packed<br />
depending on growth rate (Pearse & Pearse, 1975). A succession <strong>of</strong> fast<br />
and slow growth stages results in light and dark bands, respectively, in <strong>the</strong><br />
stereom <strong>of</strong> <strong>the</strong> ossicles (Jensen, 1969b; Pearse & Pearse, 1975; Sime,<br />
1981; Gage, 1991, 1992; Lumingas & Guillou, 1994). It is postulated that<br />
<strong>the</strong>re is only one period <strong>of</strong> fast growth and ano<strong>the</strong>r period <strong>of</strong> slow growth<br />
per year. If this is true, counting <strong>the</strong>se growth bands allows determining<br />
<strong>the</strong> ages <strong>of</strong> <strong>the</strong> echinoids. If <strong>the</strong>re is a single recruitment in a narrow time<br />
window during <strong>the</strong> year (Ebert, 1983), precision is even better. Not all<br />
authors agree with <strong>the</strong> validity <strong>of</strong> this method. Ebert (1986) questioned it<br />
and Russell & Meredith (2000) experimentally demonstrated it is not valid<br />
for Strongylocentrotus droebachiensis. However, Gage (1992) validated it<br />
for Echinus esculentus with an experiment using echinoids kept in cages in<br />
<strong>the</strong> <strong>sea</strong> during two years.<br />
Many authors consider that if <strong>the</strong>y use both methods simultaneously –<br />
cohort separation and growth rings analysis– and get <strong>the</strong> same result, each<br />
method is validated by <strong>the</strong> o<strong>the</strong>r one (Duineveld & Jenness, 1984;<br />
Lumingas & Guillou, 1994; Gebauer & Moreno, 1995; Turon et al, 1995;<br />
Jordana et al, 1997). Yet, if <strong>the</strong> number <strong>of</strong> growth rings is correlated with<br />
<strong>the</strong> size, not <strong>the</strong> age, one would interpret a group <strong>of</strong> fast-growing<br />
individuals as being older, and a group <strong>of</strong> slow-growing ones as being<br />
younger and eventually mix animals <strong>of</strong> different age in a single cohort.<br />
This would result in an agreement between both methods although<br />
conclusions on size at age are incorrect.<br />
Measuring relative growth (without knowing age) is an alternative to<br />
calculating growth rate <strong>of</strong> individuals in <strong>the</strong> field. Animals are tagged,<br />
field-released and captured again one year later (Ebert, 1977, 1988a;<br />
58
General introduction<br />
Russell, 1987; Rowley, 1990; Kenner, 1992; Ebert & Russell, 1992, 1993;<br />
Russell et al, 1998; Lamare & Mladenov, 2000). This way, <strong>sea</strong>sonal<br />
variation in growth is also partly eliminated. For <strong>sea</strong> <strong>urchin</strong>s, it is <strong>the</strong><br />
skeleton that is tagged using tetracycline (Kobayashi & Taki, 1969; Taki,<br />
1972). Size increase <strong>of</strong> <strong>the</strong> ossicles can <strong>the</strong>n be determined because a band<br />
<strong>of</strong> tetracycline-labeled skeleton, visible under ultraviolet light, indicates its<br />
size at tagging time. An allometric relationship between <strong>the</strong> size <strong>of</strong> <strong>the</strong><br />
given ossicles and <strong>the</strong> body size allow estimating <strong>the</strong> latter. To fit such<br />
data, growth <strong>model</strong>s need to be reworked, using a so-called Ford-Walford<br />
representation, or Walford plot (Ford, 1933; Walford, 1946; Ebert, 1999)<br />
where size at time t + 1 year (at recapture) is expressed as a function <strong>of</strong><br />
size at time t (at capture). Ebert (1999) reviewed such transformations for<br />
von Bertalanffy, Gompertz, logistic, Richards and Tanaka <strong>model</strong>s.<br />
Sainsbury (1980) demonstrated <strong>the</strong> strong biases that could occur with<br />
such a method when <strong>the</strong>re is individual variation in growth in a population.<br />
In fact, relative growth does not take <strong>the</strong> age into account, by definition.<br />
Hence, due to individual variations in growth rate, some fast-growing but<br />
young individuals have same size as slower-growing but older ones at a<br />
given time. The method mixes all animals having <strong>the</strong> same size, no matter<br />
<strong>the</strong>ir age, and calculates a mean growth rate at that size. This mixing <strong>of</strong><br />
age-cohorts is troublesome when growth variation is high in <strong>the</strong><br />
population. Rejection <strong>of</strong> a particular growth <strong>model</strong> could result from <strong>the</strong>se<br />
biases as Sainsbury evidenced, using a <strong>the</strong>oretical analysis with <strong>the</strong> von<br />
Bertalanffy 1 <strong>model</strong>. As individual variation in growth pattern is probable,<br />
one should be very careful using this method. Among all authors using<br />
tagged <strong>sea</strong> <strong>urchin</strong>s, Gage (1992) was <strong>the</strong> only one that cited Sainsbury's<br />
work and care about a possible bias due to individual variation in growth<br />
rate.<br />
Clearly, <strong>the</strong>re is no fool-pro<strong>of</strong> method for <strong>model</strong>ling individual growth<br />
<strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s in <strong>the</strong> field. In such a context, one can rely on experiments<br />
conducted in aquaria, although it is evident that growth patterns observed<br />
in artificial conditions could be very different to what happens in <strong>the</strong> field.<br />
59
General introduction<br />
Experiments in cages in <strong>the</strong> <strong>sea</strong> are closer to conditions <strong>of</strong> wild<br />
populations, but in <strong>the</strong> case <strong>of</strong> P. lividus populations living in tidal pools,<br />
<strong>the</strong>y are impossible to perform in practice.<br />
60
Aim <strong>of</strong> <strong>the</strong> <strong>the</strong>sis<br />
AIM OF THE THESIS<br />
The ultimate goal <strong>of</strong> this work is to formulate a mechanistic <strong>model</strong> –that is, whose<br />
parameters are functionally interpretable– <strong>of</strong> somatic growth <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
<strong>Paracentrotus</strong> lividus in cultivation. Artificial culture conditions <strong>of</strong>fer <strong>the</strong> opportunity<br />
to work with echinoids whose age, genetic origin, environmental and food conditions<br />
are perfectly known. However, we need to set up a rearing protocol adapted to P.<br />
lividus. We have also to define which is <strong>the</strong> best measurement <strong>of</strong> body size, and what<br />
it exactly represents (an overall trend in growth, or just <strong>the</strong> growth <strong>of</strong> one<br />
compartment or one organ <strong>of</strong> <strong>the</strong> echinoid). Once <strong>the</strong>se issues are solved, we will<br />
have to conduct experiments on <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s to reveal underlying processes that<br />
influence growth. Using <strong>the</strong> results <strong>of</strong> <strong>the</strong>se experiments we can <strong>the</strong>n elaborate an<br />
original ma<strong>the</strong>matical <strong>model</strong> that functionally describes growth <strong>of</strong> P. lividus <strong>reared</strong> in<br />
an aquaculture system, including those underlying processes.<br />
61
Aim <strong>of</strong> <strong>the</strong> <strong>the</strong>sis<br />
62
PART I<br />
Set up <strong>of</strong> an experimental rearing procedure<br />
for echinoids<br />
63
PART I: SET UP OF AN EXPERIMENTAL<br />
REARING PROCEDURE FOR ECHINOIDS<br />
A 230 m 2 experimental facility (with ca. 20,700 l <strong>of</strong> circulating<br />
<strong>sea</strong>water) was designed. A precise standard rearing method was set up for<br />
P. lividus aiming to study different aspects <strong>of</strong> <strong>the</strong> biology <strong>of</strong> this species in<br />
cultivation, ranging from larval biology to reproduction, including<br />
metamorphosis, somatic growth, feeding and ecophysiological reactions. It<br />
was adapted from Le Gall's method (Le Gall & Bucaille, 1989; Le Gall,<br />
1990). It allows controlling <strong>the</strong> whole life cycle <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> (closedcycle)<br />
in artificial conditions (that is, land-based systems with water<br />
recirculation) on a pilot scale.<br />
Performances <strong>of</strong> this rearing method were quantified. In particular,<br />
surviving rates and timing <strong>of</strong> <strong>the</strong> different stages were recorded for a large<br />
number <strong>of</strong> replicates (a grand total <strong>of</strong> ca. 65,000 echinoids were followed<br />
in <strong>the</strong> rearing devices, some <strong>of</strong> <strong>the</strong>m during 7 years; more than 225,000<br />
measurements were performed). Somatic growth and production were also<br />
recorded. Finally, gonadal production and maturation <strong>of</strong> <strong>the</strong> gonads were<br />
examined. In <strong>the</strong> framework <strong>of</strong> <strong>the</strong> present work, all <strong>the</strong>se measurements<br />
were used to verify that <strong>the</strong> rearing method is in adequacy with <strong>the</strong> biology<br />
<strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> and ensures its correct development –not just its survival–<br />
in <strong>the</strong> cultivation conditions.<br />
One <strong>of</strong> <strong>the</strong> goals <strong>of</strong> <strong>the</strong> re<strong>sea</strong>rch was to provide a basic methodology to<br />
be expanded on a commercial scale for industrial <strong>sea</strong> <strong>urchin</strong> farming<br />
(echiniculture). Accordingly, implications, advantages and drawbacks <strong>of</strong><br />
<strong>the</strong> method on a large-scale are widely discussed in <strong>the</strong> paper. This<br />
approach, however, is not <strong>of</strong> major concern in <strong>the</strong> present work. The latter<br />
aims basically to obtain a well-calibrated rearing method for experimental<br />
studies on somatic growth <strong>of</strong> postmetamorphic echinoids whose genetic<br />
origin, age, environmental and food conditions are perfectly known.<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
65
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
66
Land-based closed-cycle echiniculture <strong>of</strong> <strong>Paracentrotus</strong><br />
lividus (Lamarck) (Echinoidea: Echinodermata): a long-term<br />
experiment at a pilot scale<br />
a. Abstract<br />
b. Introduction<br />
Ph. Grosjean, Ch. Spirlet, P. Gosselin, D. Vaïtilingon & M. Jangoux,<br />
1998. Journal <strong>of</strong> Shellfish Re<strong>sea</strong>rch, 17(5):1523-1531<br />
Today, most worldwide <strong>sea</strong> <strong>urchin</strong>s fisheries must deal with<br />
overexploitation. Better management <strong>of</strong> exploited field populations and/or<br />
aquaculture is increasingly considered necessary to sustain <strong>sea</strong> <strong>urchin</strong><br />
production in <strong>the</strong> future. In this context, we evaluate <strong>the</strong> potential <strong>of</strong> landbased,<br />
closed-cycle echiniculture. A long-term experiment with <strong>the</strong> edible<br />
<strong>sea</strong> <strong>urchin</strong> <strong>Paracentrotus</strong> lividus has been done on a pilot scale. The<br />
process allows total independence from natural resources because <strong>the</strong><br />
entire biological cycle <strong>of</strong> <strong>the</strong> echinoids is under control (closed-cycle<br />
echiniculture) and all activities are performed on land. Fur<strong>the</strong>rmore, a<br />
method has been set up to control <strong>the</strong> reproductive cycle with <strong>the</strong> aim to<br />
produce marketable individuals all year long. Performances obtained on<br />
each stage <strong>of</strong> <strong>the</strong> rearing process are quantified and analyzed. Overall, <strong>the</strong><br />
results <strong>of</strong> this experiment are promising; however, some problems remain<br />
to be solved before we can claim pr<strong>of</strong>itability. The most important finding<br />
is that land-based, closed-cycle echiniculture is a potential viable<br />
supplement to fisheries to sustain worldwide <strong>sea</strong> <strong>urchin</strong> roe production.<br />
Keywords: <strong>sea</strong> <strong>urchin</strong>, <strong>Paracentrotus</strong> lividus, aquaculture, larval culture,<br />
metamorphosis, growth, roe enhancement.<br />
Depending upon <strong>the</strong>ir respective gastronomic culture, people consider<br />
<strong>sea</strong> <strong>urchin</strong> gonads (both male and female gonads are collectively referred<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
67
to as roe) as ei<strong>the</strong>r a fine and delicate <strong>sea</strong>food or as absolutely inedible.<br />
However, its economic value is well established given <strong>the</strong> price consumers<br />
are willing to pay. The wholesale price <strong>of</strong> live <strong>sea</strong> <strong>urchin</strong>s in France ranges<br />
from 30 to 120 FF/kg (price range in <strong>the</strong> 1990s at Rungis, Paris), and fresh<br />
roe in Japan from 6,000 to 14,000 ¥/kg (price in Japan in <strong>the</strong> 1990s, see<br />
Hagen 1996a). Both market prices are roughly equivalent in terms <strong>of</strong> fresh<br />
roe, making <strong>sea</strong> <strong>urchin</strong> roe one <strong>of</strong> <strong>the</strong> most valuable <strong>sea</strong>foods in <strong>the</strong> world.<br />
In both markets, <strong>the</strong> lowest prices are those <strong>of</strong> imported <strong>sea</strong> <strong>urchin</strong>s, which<br />
are considered to be <strong>of</strong> poorer quality.<br />
The most important market, Japan, imports approximately five<br />
thousand tons <strong>of</strong> <strong>sea</strong> <strong>urchin</strong> gonads per year, <strong>the</strong> equivalent <strong>of</strong> 40 to 50<br />
thousand tons <strong>of</strong> live <strong>sea</strong> <strong>urchin</strong>s (Hagen, 1996a). According to <strong>the</strong> same<br />
author, <strong>the</strong> Japanese consumes approximately 60,000 tons <strong>of</strong> whole <strong>sea</strong><br />
<strong>urchin</strong>s per year. The second largest consumer is France, with an annual<br />
consumption <strong>of</strong> approximately 1,000 tons <strong>of</strong> whole <strong>sea</strong> <strong>urchin</strong>s (Le Gall,<br />
1990).<br />
Increasing demand for <strong>sea</strong> <strong>urchin</strong> roe and a steady rise in price have led<br />
to worldwide intensification <strong>of</strong> <strong>sea</strong> <strong>urchin</strong> fisheries (Conand and Sloan,<br />
1989; Le Gall, 1990; Saito, 1992) which has now (1998) probably reached<br />
its maximum. This production cannot be sustained at current levels<br />
because <strong>the</strong> declining productivity <strong>of</strong> overexploited existing stocks can no<br />
longer be compensated by harvest <strong>of</strong> new stocks, as was possible over <strong>the</strong><br />
last three decades (most exploitable natural populations have already been<br />
fished today). In Japan, this decline occurred despite <strong>the</strong> development and<br />
implementation <strong>of</strong> extensive domestic fishery enhancement techniques<br />
which include <strong>the</strong> annual release <strong>of</strong> 60 million juvenile <strong>sea</strong> <strong>urchin</strong>s into <strong>the</strong><br />
wild (Saito, 1992; Hagen, 1996a). Consequently, <strong>the</strong> worldwide supply <strong>of</strong><br />
high quality <strong>sea</strong> <strong>urchin</strong> roe will be unable to meet market demand unless<br />
commercial <strong>sea</strong> <strong>urchin</strong> aquaculture develops to partially replace <strong>the</strong> steady<br />
decrease in natural captures.<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
68
Aquaculture <strong>of</strong> echinoderms, including <strong>sea</strong> <strong>urchin</strong>s and <strong>sea</strong> cucumbers<br />
is known as echinoculture (Le Gall & Bucaille, 1989; Le Gall, 1990;<br />
Hagen, 1996a). We prefer to use <strong>the</strong> term echiniculture to describe <strong>sea</strong><br />
<strong>urchin</strong> aquaculture solely (Echinoidea); thus, it is more accurate in this<br />
context. This activity is not yet fully developed. Maintenance or rearing <strong>of</strong><br />
<strong>sea</strong> <strong>urchin</strong>s in <strong>the</strong> laboratory has been successfully performed for different<br />
species (Hinegardner, 1969; Fridberger et al, 1979; Cellario & Fenaux,<br />
1990). Several different processes are being experimented on a larger<br />
scale, ranging from <strong>sea</strong> <strong>urchin</strong> ranching (cultivation in <strong>the</strong> field, see<br />
Fernandez & Caltagirone, 1994; Fernandez, 1996), to land-based systems<br />
(Le Gall & Bucaille, 1989; Le Gall, 1990; Fernandez, 1996) or polyculture<br />
(<strong>sea</strong> <strong>urchin</strong>s cultivated in cages with fish, see Kelly et al, 1998).<br />
Never<strong>the</strong>less, considering <strong>the</strong> limited carrying capacity <strong>of</strong> natural sites that<br />
are already largely exploited by fisheries, only land-based or cage<br />
techniques will help to sustain worldwide <strong>sea</strong> <strong>urchin</strong> roe production.<br />
Similarly, only cultivation processes totally independent <strong>of</strong> natural stocks,<br />
that is, by controlling <strong>the</strong> complete life cycle <strong>of</strong> <strong>the</strong> echinoid, will lower<br />
<strong>the</strong> pressure imposed by fisheries upon natural populations. In this context,<br />
this paper presents a 7-year experimental rearing method to produce <strong>the</strong><br />
edible <strong>sea</strong> <strong>urchin</strong> P. lividus on a pilot scale, and discusses <strong>the</strong> biological<br />
and technological issues that emerged from this cultivation method.<br />
c. Material and methods<br />
The aim <strong>of</strong> land-based, closed-cycle echiniculture is to get maximum<br />
control over each phase <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>'s life cycle by controlling <strong>the</strong><br />
major environmental parameters (temperature, photoperiod, water quality,<br />
quality and quantity <strong>of</strong> food). A land-based system has advantages over<br />
methods performed directly in <strong>the</strong> <strong>sea</strong>. The greatest <strong>of</strong> <strong>the</strong>se is <strong>the</strong> ability<br />
to control <strong>the</strong> whole life cycle <strong>of</strong> <strong>the</strong> animals (closed cycle), thus <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong> never depends, at any <strong>of</strong> its stages, on a supply <strong>of</strong> animals<br />
originating from <strong>the</strong> field.<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
69
The method used here is adapted from Le Gall (Le Gall & Bucaille,<br />
1989; Le Gall, 1990) with some fine-tuning and modifications that allow a<br />
routine output <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s on a pilot scale. An experimental facility has<br />
been set up at <strong>the</strong> "Centre de Recherche et d'Etude Côtière" (CREC,<br />
Normandy, France) in which several generations <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s have been<br />
<strong>reared</strong> according to a thoroughly defined experimental procedure.<br />
Pilot echiniculture facility<br />
The experimental facility includes a hatchery (30 m 3 ) and a cultivation<br />
room (160 m 3 ). The hatchery is equipped with 11 200-l larval rearing tanks<br />
(see below) and a system for phytoplankton production (classical devices<br />
for large-scale production).<br />
The cultivation room is insulated, <strong>the</strong>rmoregulated at 22°C ± 1°C,<br />
correctly aerated, and exposed to a 12h/12h photoperiod. It is equipped<br />
with 10 autonomous rearing structures with ei<strong>the</strong>r three or six superposed<br />
4-m long and 60-cm wide ponds called toboggans. Each set <strong>of</strong> toboggans<br />
hangs over a reserve/settling tank <strong>of</strong> <strong>the</strong> same length, 80 cm wide and<br />
80 cm deep. The water depth in <strong>the</strong> toboggans varies between 5 and 10 cm.<br />
A centrifugal pump transfers water from <strong>the</strong> reserve tank to <strong>the</strong> top level<br />
with a flow <strong>of</strong> 8 to 10 m 3 /h (4 to 5 m 3 /h for <strong>the</strong> pregrowth structure, see<br />
below). The water <strong>the</strong>n recirculates by gravity from one level to <strong>the</strong> o<strong>the</strong>r<br />
(each toboggan has a gentle slope to help water run into it and is connected<br />
to <strong>the</strong> previous and <strong>the</strong> next one at its opposite ends, see Fig. 17). This<br />
device, specifically designed for <strong>sea</strong> <strong>urchin</strong> cultivation, optimizes both <strong>the</strong><br />
surface available for <strong>the</strong> postmetamorphic individuals and <strong>the</strong> water<br />
current around <strong>the</strong>m. It also facilitates access to <strong>the</strong> animals and <strong>the</strong>ir<br />
visual control. The 10 rearing structures are organized as follows:<br />
(1) One pregrowth structure <strong>of</strong> 3 toboggans with a capacity <strong>of</strong> 1500 l <strong>of</strong><br />
circulating water <strong>the</strong>rmoregulated at 20°C ± 1°C. The water is renewed at<br />
a rate <strong>of</strong> 150% per day. This structure can hold a biomass ranged between<br />
0.2 and 1 kg/m 2 <strong>of</strong> toboggans.<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
70
6b. exploitation<br />
KCl<br />
water renewal<br />
toboggans<br />
KCl<br />
6a. broodstock cond.<br />
reserve tank<br />
5. growth <strong>of</strong> subadults<br />
1. fertilization<br />
FERTILIZING<br />
TUB<br />
4. growth <strong>of</strong> juveniles<br />
18 to 20°C<br />
PREGROW TH & GROWTH STRUCTURES<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
2. larvae culture<br />
200 l<br />
20°C<br />
LARVAL REARING TANK<br />
3. metamorphosis<br />
water<br />
pump<br />
Figure 17. Overview <strong>of</strong> <strong>the</strong> closed-cycle process and devices used to produce <strong>sea</strong> <strong>urchin</strong>s on<br />
land at a pilot scale.<br />
(2) Two growth structures made <strong>of</strong> six toboggans each. The capacity <strong>of</strong><br />
each structure is 3000 l <strong>of</strong> circulating water <strong>the</strong>rmoregulated at 18°C ± 1°C<br />
and with a water renewal ranged between 100 and 600% per day,<br />
depending upon <strong>the</strong> density <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s present in <strong>the</strong> structures. These<br />
structures can hold a maximum biomass <strong>of</strong> 7 kg <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s per m 2<br />
without supplemental filtration <strong>of</strong> <strong>the</strong> water.<br />
71
(3) Seven experimental / conditioning structures <strong>of</strong> three toboggans each<br />
with a capacity <strong>of</strong> 1500 l <strong>of</strong> circulating water. These are isolated from one<br />
ano<strong>the</strong>r so that <strong>the</strong>y can be <strong>the</strong>rmoregulated individually from 10°C to<br />
25°C, and each has up to six different photoperiods (a dark separation<br />
divides <strong>the</strong> toboggans in <strong>the</strong>ir center). An electronic system allows <strong>the</strong><br />
transition <strong>of</strong> light to darkness and vice versa, thus simulating dawn and<br />
dusk. The rate <strong>of</strong> water renewal can be fixed between 50 and 600% per<br />
day. Biomass varies following criteria imposed by <strong>the</strong> experiments.<br />
Additional devices are grouped in a technical room containing a central<br />
<strong>the</strong>rmoregulation system (a <strong>the</strong>rmorefrigerating pump providing ei<strong>the</strong>r<br />
cold or hot water to <strong>the</strong> heat exchangers that equip <strong>the</strong> rearing structures),<br />
a water pumping and filtration system, an emergency electric generator<br />
and a central alarm. The water is pumped directly from <strong>the</strong> <strong>sea</strong> at high tide<br />
and is stored in a reservoir <strong>of</strong> 60 m 2 . It is filtered before being used (30 µm<br />
mesh cartridge mechanical filtration, followed by a 14 m 3 biological filter<br />
and two settling tanks <strong>of</strong> 8 m 3 each).<br />
Origin <strong>of</strong> <strong>the</strong> animals<br />
The species cultivated is <strong>Paracentrotus</strong> lividus (Lamarck, 1816). This<br />
species is found all along <strong>the</strong> European coast from <strong>the</strong> nor<strong>the</strong>rn Atlantic<br />
Irish coast to <strong>the</strong> Mediterranean Sea. All individuals used come from a<br />
single population located in Morgat, Brittany, France. Some were directly<br />
collected in <strong>the</strong> small tidepools that spread all along <strong>the</strong> rocky shores <strong>of</strong><br />
Douarnenez Bay (emerged only during high coefficient tides). The<br />
remaining animals come from artificial fertilizations in <strong>the</strong> laboratory and<br />
were grown in <strong>the</strong> structures (cross fertilizations <strong>of</strong> first (F1) or second<br />
(F2) generation <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s collected in <strong>the</strong> field). By so doing, <strong>the</strong> age<br />
and <strong>the</strong> parental origin <strong>of</strong> <strong>the</strong> F1 and F2 <strong>sea</strong> <strong>urchin</strong>s are known precisely.<br />
Rearing method<br />
Aiming at closely matching <strong>the</strong> echinoid requirements along <strong>the</strong>ir life<br />
history and minimizing technical constraints, <strong>the</strong> dissociation <strong>of</strong> <strong>the</strong> whole<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
72
earing cycle into seven stages is essential (Fig. 17). These stages are: (1)<br />
fertilization, (2) larval culture, (3) metamorphosis, (4) growth <strong>of</strong> juveniles,<br />
(5) growth <strong>of</strong> subadults and (6) growth <strong>of</strong> adults, which is fur<strong>the</strong>r divided<br />
into (6a) conditioning for <strong>the</strong> marketing <strong>of</strong> roe (exploitation) and (6b)<br />
providing gametes (broodstock).<br />
Stage 1: Fertilization is performed using gametes issued from healthy<br />
animals that restored <strong>the</strong>ir gamete potential as described below in<br />
broodstock conditioning (see stage 6b). The gametes are obtained by<br />
stimulating <strong>the</strong> parents to spawn with 0.5 N KCl (injection <strong>of</strong> 50 µl per g<br />
<strong>of</strong> body weight through <strong>the</strong> peristomial membrane). The gametes <strong>of</strong> each<br />
individual are collected in a small jar <strong>of</strong> 50 ml in 20°C filtered natural<br />
<strong>sea</strong>water (on a 1 µm cartridge filter, referred hereafter as "larval rearing<br />
water").<br />
When <strong>the</strong> spawning is over, <strong>the</strong> volume <strong>of</strong> <strong>the</strong> gametes is evaluated.<br />
The ova <strong>of</strong> a single female are transferred in a fertilization tub, that is, a<br />
shallow polyethylene container. The volume is brought to 800 ml with <strong>the</strong><br />
same water. One fifth <strong>of</strong> <strong>the</strong> spermatozoa <strong>of</strong> a single male is added to <strong>the</strong><br />
ova. The mixture is kept at 20 ± 1°C during 4 h and <strong>the</strong> tub is gently stirred<br />
three or four times during that period. After that, <strong>the</strong> success <strong>of</strong> <strong>the</strong><br />
fertilization is checked and <strong>the</strong> fertilized eggs are counted (most <strong>of</strong>ten over<br />
90% <strong>of</strong> <strong>the</strong> eggs are fertilized).<br />
Stage 2: Rearing <strong>of</strong> <strong>the</strong> larvae is done in a 200-l polyethylene<br />
cylindrical tank where larval rearing water is introduced 24 h beforehand<br />
and stabilized at 20 ± 1°C. The embryos (in <strong>the</strong> gastrula stage) are<br />
introduced at a concentration <strong>of</strong> 250 per liter. This density is low enough<br />
to allow <strong>the</strong> entire rearing <strong>of</strong> <strong>the</strong> larvae to be conducted without renewing<br />
<strong>the</strong> water. The food (Phaeodactylum tricornutum Bohlin issued from<br />
cultivation in Erdschreiber medium) is introduced from <strong>the</strong> third day<br />
postfertilization (acquisition <strong>of</strong> larval exotrophy). The larvae are fed once a<br />
day with 600 ml algal cultivation (concentration around 10·10 6 cells/ml).<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
73
The whole is kept in dim light with a 12h/12h photoperiod and is gently<br />
mixed and aerated by a central bubbling.<br />
Stage 3: From <strong>the</strong> sixteenth day onward, competence to<br />
metamorphosis is checked daily (Standard Competence Test or SCT,<br />
adapted from Gosselin & Jangoux, 1996). One hundred larvae are<br />
transferred in a clean SCT sieve (a 10 cm high, 20 cm 2 sieve with a bottom<br />
mesh <strong>of</strong> 250 µm placed 1.5 cm above <strong>the</strong> water floor). This SCT sieve is<br />
placed in <strong>the</strong> pregrowth structure in <strong>the</strong> presence <strong>of</strong> a metamorphosis<br />
stimulating factor (living Corallina elongata Ellis & Sollander, freshly<br />
collected from <strong>the</strong> field). The percentage <strong>of</strong> metamorphosed larvae is<br />
determined 24 h later. If this value lies around 80%, <strong>the</strong> whole batch is<br />
transferred in <strong>the</strong> pregrowth structure aiming at its fixation on one or two<br />
metamorphosis sieves (similar to SCT sieves but each covering 1800 cm 2 ).<br />
Batches containing large amounts <strong>of</strong> larvae exhibiting bad development,<br />
abnormalities or too low metamorphosis ratios are discarded.<br />
Stage 4: <strong>Growth</strong> <strong>of</strong> juveniles. The postmetamorphic period begins with<br />
a short endotrophic stage. During this period, <strong>the</strong> postmetamorphic<br />
individuals, also called postlarvae, reorganize <strong>the</strong>ir digestive tract<br />
(Gosselin & Jangoux, 1998). The mouth and anus <strong>of</strong> <strong>the</strong> future juvenile are<br />
not yet pierced. This postlarval stage lasts for up to 8 days, after which <strong>the</strong><br />
echinoids become exotrophic juveniles. One or two days before<br />
development <strong>of</strong> exotrophy, 100 g fresh weight <strong>of</strong> Enteromorpha linza (L.)<br />
Agardh collected in <strong>the</strong> field are distributed in each sieve. From this<br />
moment onward, <strong>the</strong> same food quantity is given every time it is<br />
completely consumed. Some Gammarus locusta L. are also introduced to<br />
clean <strong>the</strong> sieves from decomposing parts <strong>of</strong> <strong>the</strong> algae.<br />
The juveniles are left in <strong>the</strong>se sieves until <strong>the</strong> mean individual size in<br />
<strong>the</strong> batch reaches 2 mm. The entire batch is <strong>the</strong>n transferred in 500 µm<br />
mesh pregrowth sieves. A homogeneous bed <strong>of</strong> E. linza is maintained in<br />
<strong>the</strong> sieves. The bottoms <strong>of</strong> <strong>the</strong> sieves are cleaned every week using filtered<br />
<strong>sea</strong>water. Because <strong>the</strong> growth <strong>of</strong> <strong>the</strong> juveniles is not homogeneous<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
74
(Grosjean et al, 1996, see Part III), <strong>the</strong> animals are graded every month,<br />
and those with a diameter larger than 5 mm are transferred into a 1-mm<br />
mesh pregrowth sieve. The E. linza diet is maintained and <strong>the</strong> sieves<br />
remain in <strong>the</strong> same pregrowth structure.<br />
Stage 5: <strong>Growth</strong> <strong>of</strong> subadults. Every month, a sorting <strong>of</strong> size is done to<br />
collect all individuals bigger than 10 mm. The individuals whose size<br />
exceeds 10 mm but is below <strong>the</strong> minimum market size <strong>of</strong> around 40 mm<br />
for P. lividus are defined as subadults. They are potentially mature but not<br />
large enough for <strong>the</strong> market. Consequently, <strong>the</strong>ir somatic growth<br />
performances must be promoted while <strong>the</strong>ir gonadal growth should be kept<br />
as low as possible to optimize food allocation to <strong>the</strong> soma.<br />
Subadults are placed in rectangular rearing baskets, with all sides made<br />
out <strong>of</strong> 5-mm mesh. These rearing baskets are placed 1.5 cm above <strong>the</strong><br />
bottom <strong>of</strong> <strong>the</strong> toboggans and are just slightly narrower. This is important to<br />
allow good water circulation around and inside <strong>the</strong>m, and good elimination<br />
<strong>of</strong> solid wastes produced by <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s. Their surface ranges between<br />
1200 and 2400 cm 2 . When <strong>the</strong> size <strong>of</strong> <strong>the</strong> animals increases above 15 mm<br />
in test diameter, <strong>the</strong>y are transferred in <strong>the</strong> same type <strong>of</strong> rearing baskets,<br />
but with 10-mm mesh, which allows even better water circulation.<br />
Subadults, inside <strong>the</strong>ir rearing baskets, are transferred to a growth<br />
structure. From this time onward, and twice a week, <strong>the</strong>y are fed ad libitum<br />
with fresh kelp, Laminaria digitata. Cleaning <strong>of</strong> <strong>the</strong> baskets and toboggans<br />
is also done twice weekly. Dead or dying animals are removed daily. Each<br />
month, sorting by size is done to separate <strong>the</strong> batches into different size<br />
categories from 5 to 5 mm. The entire cultivation is kept in 12h/12h<br />
photoperiod.<br />
Stage 6a: Conditioning <strong>of</strong> <strong>the</strong> adults for market. When <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
reach 40 mm, <strong>the</strong>y are prepared to get marketable gonads in conditioning<br />
structures. For <strong>the</strong> commercialization, it is <strong>of</strong> utmost importance that <strong>the</strong><br />
echinoids' gonadal cycle is synchronous, presents <strong>the</strong> right stage <strong>of</strong><br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
75
maturity (reproductive stages 4 and 5, growing and premature, according<br />
to Spirlet et al, 1998a) and is <strong>of</strong> acceptable texture (firm and not leaking),<br />
size (as large as possible), good color (yellow-orange to bright orange) and<br />
taste. P. lividus has an annual reproductive cycle that tends to fade in<br />
constant artificial conditions: lacking <strong>the</strong> "usual" stressors (low<br />
temperature, lighting variation, lower quality or lack <strong>of</strong> food during<br />
winter) <strong>the</strong> echinoids tend to bypass <strong>the</strong> growth phase <strong>of</strong> <strong>the</strong> gonads and<br />
have permanent gametogenesis, giving rise to flabby gonads with few<br />
nutritive phagocytes. Such gonads are unacceptable in <strong>the</strong> market. To<br />
counteract this, <strong>the</strong> echinoids are starved at a temperature <strong>of</strong> 12-14°C and<br />
at a 12h/12h photoperiod. This leads to consumption <strong>of</strong> <strong>the</strong> possible<br />
content <strong>of</strong> <strong>the</strong> gonads, which also act as storage organs, in order for <strong>the</strong><br />
animals to get in phase regarding <strong>the</strong>ir reproductive cycle (reproductive<br />
stages 1 to 3, spent and recovering, Spirlet et al, 1998a). When <strong>the</strong> content<br />
<strong>of</strong> <strong>the</strong> gonads is fully consumed, that is, between 1 and 2 months later,<br />
depending on <strong>the</strong>ir initial state, <strong>sea</strong> <strong>urchin</strong>s are fed ad libitum with ei<strong>the</strong>r<br />
Laminaria digitata or an appropriate artificial food rich in proteins<br />
(Klinger et al, 1994, 1997, 1998; Williams & Harris, 1998) at a higher<br />
temperature (at least 16°C). The duration <strong>of</strong> this feeding stage is dictated<br />
by <strong>the</strong> maturation <strong>of</strong> <strong>the</strong> gonads and lasts for 6 weeks to 3 months, mainly<br />
depending on <strong>the</strong> food quality. Usually, both <strong>the</strong> size and <strong>the</strong> maturation<br />
stage simultaneously reach acceptable values, and gonads are ready for <strong>the</strong><br />
market at <strong>the</strong> end <strong>of</strong> this starving-feeding treatment (see results).<br />
Stage 6b: Conditioning broodstock. Maintaining mature broodstock <strong>of</strong><br />
P. lividus all year long is done by keeping individuals at high temperature<br />
(between 18°C and 20°C) and under ei<strong>the</strong>r a fixed photoperiod <strong>of</strong> 12h/12h<br />
(directly in <strong>the</strong> growth structures) or, better, in total darkness (in a<br />
conditioning structure), leading to <strong>the</strong> disruption <strong>of</strong> <strong>the</strong>ir reproductive<br />
cycle. In <strong>the</strong>se conditions, food is <strong>the</strong> most important factor to get large<br />
quantities <strong>of</strong> good quality gametes. Feeding adults ad libitum with fresh<br />
Laminaria digitata ensures both <strong>the</strong> quality and <strong>the</strong> quantity <strong>of</strong> sexual<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
76
output. The quality <strong>of</strong> gametes is <strong>of</strong>ten a little bit lower from December till<br />
February, though still usable most <strong>of</strong> <strong>the</strong> time.<br />
Measurements <strong>of</strong> <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
Essentially two criteria are used to quantify <strong>the</strong> performances <strong>of</strong> <strong>the</strong><br />
rearing method: (1) <strong>the</strong> survival rate with time and (2) <strong>the</strong> growth rate, that<br />
is, <strong>the</strong> change <strong>of</strong> test diameter <strong>of</strong> <strong>the</strong> <strong>urchin</strong>s with time (gonadal size and<br />
quality are taken into account only after <strong>the</strong> minimal market size has been<br />
reached). The first is determined by counting survivals in a single batch<br />
(issued from a single fertilization and a single larval rearing tank) at<br />
various times. The counting <strong>of</strong> eggs, embryos and larvae is performed on<br />
at least five samples <strong>of</strong> <strong>the</strong> homogenized batch (<strong>the</strong> volume chosen to<br />
count each time is at least one hundred individuals) and <strong>the</strong> total amount is<br />
estimated by extrapolating <strong>the</strong> mean concentration found to <strong>the</strong> whole<br />
volume. The survival rate <strong>of</strong> competent larvae, postlarvae and juveniles is<br />
determined by rearing subsamples <strong>of</strong> 50 to 100 individuals in SCT sieves.<br />
Several replicates (at least five) are sacrificed and counted at each time.<br />
All subadults and adults <strong>of</strong> a batch are counted and measured every 3<br />
months (typically between a few hundred to a few thousand individuals in<br />
each batch) during size sorting. Measurements <strong>of</strong> subadults and adults do<br />
not induce additional stress or mortality o<strong>the</strong>r than those occurring during<br />
<strong>the</strong> normal size grading operation (no additional manipulations). Mortality<br />
caused by manipulations could thus be attributed to <strong>the</strong> rearing method<br />
itself.<br />
Size is evaluated by means <strong>of</strong> <strong>the</strong> diameter, which is measured to <strong>the</strong><br />
ambitus <strong>of</strong> <strong>the</strong> test (its largest part) considered without spines. To prevent<br />
errors caused by a possible slightly oval shape, we measure two<br />
perpendicular diameters, both to <strong>the</strong> ambitus, and only <strong>the</strong> average is<br />
considered. The diameter <strong>of</strong> juveniles, after fixing <strong>the</strong>m (glutaraldehyde<br />
3%), is measured on digitized microphotographs using a specific image<br />
analysis s<strong>of</strong>tware (Grosjean et al, 1996, see Part III). The diameter <strong>of</strong><br />
subadults and adults is measured with a sliding caliper. Fresh weight, used<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
77
d. Results<br />
to evaluate biomass, is measured after draining residual water on absorbent<br />
paper during 5 minutes.<br />
The relative size <strong>of</strong> <strong>the</strong> gonads is quantified by means <strong>of</strong> <strong>the</strong> fresh and<br />
dry weight gonadal indices (GI, also called gonadosomatic indices). These<br />
indices are defined as <strong>the</strong> ratio between <strong>the</strong> fresh (or dry) weight <strong>of</strong> <strong>the</strong><br />
gonads and <strong>the</strong> total fresh (or dry) weight <strong>of</strong> <strong>the</strong> <strong>urchin</strong>s. First, fresh weight<br />
<strong>of</strong> <strong>the</strong> <strong>urchin</strong>s is determined after drying <strong>the</strong>m for 5 minutes on absorbent<br />
paper. The animals are <strong>the</strong>n dissected, and <strong>the</strong> five gonads are extracted<br />
and weighed. One gonad is fixed in Bouin's fluid for fur<strong>the</strong>r determination<br />
<strong>of</strong> its gametogenic stage (see below). The remaining four gonads are<br />
weighed again, and <strong>the</strong> difference is computed to allow correction <strong>of</strong> <strong>the</strong><br />
dry weight for <strong>the</strong> missing gonad. The remaining gonads and <strong>the</strong> soma are<br />
<strong>the</strong>n dried at 70°C during 48 h (constant weight) before being separately<br />
weighed. Dry weight gonad index is more accurate but has been found to<br />
be less representative <strong>of</strong> <strong>the</strong> "filling" <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s (how much space<br />
<strong>the</strong> gonads occupy inside <strong>the</strong> coelomic cavity), especially when comparing<br />
various maturity stages and/or various diets (unpublished results). Both<br />
indices are provided to allow comparisons.<br />
The maturity stage is determined on histological sections <strong>of</strong> <strong>the</strong> fixed<br />
gonad following an 8-stages scale defined by Spirlet et al (1998a). The<br />
maturity index (MI) corresponds to <strong>the</strong> arithmetic mean <strong>of</strong> all <strong>the</strong> observed<br />
maturity stages. Male and female data are pooled for both <strong>the</strong> GI and <strong>the</strong><br />
MI, since differences between sexes are not significant (Spirlet et al,<br />
1998a, 1998b).<br />
Table 2 shows <strong>the</strong> age, <strong>the</strong> density and <strong>the</strong> survival rate for each stage<br />
described in rearing conditions. These data come from 29 fertilizations<br />
studied during several years taking into account, among o<strong>the</strong>r things, <strong>the</strong><br />
<strong>sea</strong>sonal variations. The survival rate for larvae is about 56%. Competence<br />
is reached most <strong>of</strong>ten in 18 days (mode and median value), with an<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
78
average value <strong>of</strong> 19.5 days, a minimal time <strong>of</strong> 16 days and a maximal time<br />
<strong>of</strong> 25 days. The mean metamorphosis rate is 80.4% when larvae are<br />
competent. This rate was reached in almost two-third <strong>of</strong> <strong>the</strong> fertilizations<br />
that attained <strong>the</strong> competent stage (non-symmetrical distribution). Thirty<br />
percent <strong>of</strong> <strong>the</strong> larvae were discarded, ei<strong>the</strong>r because <strong>of</strong> an incomplete<br />
development or too low metamorphosis rate. The remaining larvae were<br />
used for studies on postlarval or juvenile stages (and, thus, sacrificed<br />
whenever measured) or were <strong>reared</strong> to <strong>the</strong> adult stage. Overall, <strong>the</strong> survival<br />
rate is homogeneous from one fertilization to <strong>the</strong> o<strong>the</strong>r and for all stages,<br />
except during and after <strong>the</strong> acquisition <strong>of</strong> exotrophy (transition from <strong>the</strong><br />
postlarval to <strong>the</strong> juvenile stage): <strong>the</strong> average rate is 54.5%, but extremes<br />
are close to 0 and 100% (13% and 94.5% respectively). Whatever <strong>the</strong><br />
success <strong>of</strong> this transition, <strong>the</strong> most critical period for survival is <strong>the</strong><br />
juvenile stage, with a very low survival rate <strong>of</strong> 5%. Most <strong>of</strong> <strong>the</strong> mortality<br />
occurs during <strong>the</strong> few first months <strong>of</strong> <strong>the</strong> juvenile's life (and even probably<br />
during <strong>the</strong> few first weeks), with a progressive decrease around 8 to 9<br />
months <strong>of</strong> age.<br />
Table 2. Age, density, number, and survival rate <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s at each rearing stage.<br />
Rearing<br />
stage<br />
Developmental<br />
stage<br />
No.<br />
replicated<br />
fertilizations<br />
1 embryos 29 (a)<br />
2 competent larvae 29 (a)<br />
3 postlarvae 18 (b)<br />
4 juveniles 9 (b)<br />
5 subadults 6 (c)<br />
6a & b adults 5 (c)<br />
Age Mean density Mean no. Survival from Mean<br />
(min/ median / (no. ind./vol. individuals in previous stage global<br />
max) or /surf. unit) 1 batch (%) survival<br />
mean ± SD rate (%)<br />
4 h 250 / l 50,000 - 100<br />
16 d / 18 d / 25 d 141 / l 28,200 56.4 ± 11.6 56.4<br />
idem + 1 d 6.5·10 4 / m 2<br />
idem + 10 d 3.5·10 4 / m 2<br />
2 (d)<br />
ca. 9 months 4,000 / m<br />
2 (d)<br />
1.7 y / 2.6 y / 3.5 y 250 / m<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
22,700 80.4 ± 14.4 45.3<br />
12,400 54.5 ± 26.8 24.7<br />
600 4.9 ± 1.5 1.2<br />
310 51.5 ± 3.0 0.6<br />
(a)<br />
Total number <strong>of</strong> larval rearing tanks: 103, from which 72 have produced enough usable competent larvae.<br />
(b)<br />
In <strong>the</strong> pregrowth structure, 5 to 15 replicates are measured at key times for each fertilization (see Material and Methods).<br />
(c)<br />
In <strong>the</strong> growth or conditioning structures. Each batch is issued from a single larval rearing tank and is followed over 2 to 7<br />
years.<br />
(d)<br />
Densities in <strong>the</strong> rearing structures are adjusted during sorting operations according to both <strong>the</strong> individual size and <strong>the</strong> survival<br />
rate.<br />
79
survival rate (%)<br />
test<br />
diameter<br />
(mm)<br />
age (years)<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
nbr <strong>of</strong> individuals<br />
Figure 18. Changes with time in <strong>the</strong> size distribution and survival rate <strong>of</strong> one fertilization<br />
issued from a single larval rearing tank and followed over 7 years. Note <strong>the</strong> leading group<br />
that singles out (represented by dark bars in <strong>the</strong> histograms).<br />
Figure 18 shows both <strong>the</strong> survival rate and <strong>the</strong> size distribution over<br />
time <strong>of</strong> a batch followed for 7 years, far beyond <strong>the</strong> minimal marketable<br />
size and age. For <strong>the</strong> sake <strong>of</strong> clarity, only data taken every 6 months are<br />
represented, although measurements were made every 3 months beginning<br />
at 6 months <strong>of</strong> age. The trends observed on this single cohort are<br />
representative <strong>of</strong> <strong>the</strong> way animals grow in cultivation, as confirmed by <strong>the</strong><br />
five o<strong>the</strong>r independent batches measured over 2 to 4 years (for an<br />
illustrated example <strong>of</strong> ano<strong>the</strong>r batch, see Grosjean et al, 1996, Part III).<br />
Mortality (represented on <strong>the</strong> backwall <strong>of</strong> <strong>the</strong> 3-D box in Fig. 18)<br />
remains very high until about 9 months <strong>of</strong> age in <strong>the</strong> pregrowth structure.<br />
In <strong>the</strong> figured case, from around 12,400 juveniles issued from one rearing<br />
tank, only 725 individuals where counted after 6 months and 507 remained<br />
after ano<strong>the</strong>r 3 months. Mortality dropped after this critical period, and 491<br />
80
individuals were still alive 3 months later (1 year <strong>of</strong> age). This<br />
corresponds, respectively, to a mortality <strong>of</strong> 94% (between <strong>the</strong> acquisition<br />
<strong>of</strong> exotrophy by <strong>the</strong> juvenile to 6 months old), 30% (during <strong>the</strong> next 3<br />
months) and 3% (after <strong>the</strong> following 3 months). The mortality rate <strong>of</strong><br />
subadults stabilizes around 5.4% per trimester until 6 years <strong>of</strong> age, but<br />
ranges from 0.9% per trimester to 12.7% per trimester. Most <strong>of</strong> this<br />
variation is correlated with <strong>sea</strong>son: mortality is higher during winter;<br />
whereas, summer mortality nearly reaches 0%. Most <strong>of</strong> winter mortality<br />
occurs by waves that start unpredictably and last for 2 to 3 days.<br />
Juvenile's individual growth in test diameter is slow. It accelerates for<br />
subadults but <strong>the</strong>n scatters for intermediate sizes (15 to 35 mm), even<br />
inside a presumably homogeneous batch. This scattering <strong>of</strong>ten results in<br />
bimodal or trimodal size distributions (see Fig. 18 for an example and<br />
Grosjean et al. 1996 –Part III– for an analysis). When echinoids approach<br />
asymptotic size, <strong>the</strong>ir growth rate drops. Hence, <strong>the</strong> leading group is<br />
eventually caught up by <strong>the</strong> trailers around or slightly above <strong>the</strong> minimal<br />
market size. This minimal market size is attained between 1.7 and 3.5<br />
years old (respectively 10% and 90% <strong>of</strong> <strong>the</strong> individuals are larger than 40<br />
mm) with a median value <strong>of</strong> 2.6 years.<br />
biomass (kg)<br />
40<br />
30<br />
20<br />
10<br />
0<br />
1 2 3 4 5 6 7<br />
age (years)<br />
Figure 19. Change with time in <strong>the</strong> biomass <strong>of</strong> a <strong>reared</strong> cohort <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s (<strong>the</strong> same batch<br />
as shown in Fig. 18).<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
81
Biomass variations (Fig. 19, same batch as in Fig. 18) are correlated to<br />
both <strong>the</strong> survival rate and <strong>the</strong> growth speed <strong>of</strong> <strong>reared</strong> echinoids. The higher<br />
mortality observed in winter overrides growth speed, and biomass tends to<br />
decrease slightly. Summer biomass is highest during <strong>the</strong> third and <strong>the</strong><br />
fourth years in this case. The first peak <strong>of</strong> biomass (around 3.5 years old in<br />
<strong>the</strong> figured case, between 2.8 and 3.5 years old for <strong>the</strong> o<strong>the</strong>r batches<br />
depending on <strong>the</strong> <strong>sea</strong>son) corresponds to reaching <strong>of</strong> <strong>the</strong> minimal market<br />
size by more than 90% <strong>of</strong> <strong>the</strong> individuals and seems to be <strong>the</strong> best time to<br />
commercialize <strong>the</strong>m after conditioning <strong>the</strong>ir gonads (stage 6a) from a strict<br />
biological point <strong>of</strong> view. At that time, between 35 and 40 kg <strong>of</strong> fresh<br />
weight <strong>sea</strong> <strong>urchin</strong>s are produced in a single batch. This represents an overall<br />
yield per surface unit <strong>of</strong> <strong>the</strong> growth structures <strong>of</strong> 4 to 7 kg / m 2 <strong>of</strong><br />
toboggans / year. To obtain this result, roughly 400 kg <strong>of</strong> kelp was<br />
provided to <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s. Thus, over-all food conversion efficiency lies<br />
around 10%.<br />
Table 3. Gonadal and maturity indices <strong>of</strong> wild and <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s (pooled results for<br />
males and females).<br />
Origin Month Food Treatment Tempe-<br />
rature<br />
field March natural diet collected in Morgat (b)<br />
Photoperiod<br />
(L/D)<br />
Wet w. GI (%)<br />
mean ± SD<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
Dry w. GI (%)<br />
mean ± SD<br />
MI (a)<br />
mean ±<br />
SD<br />
10°C 13h / 11h 11.6 ± 4.2 7.1 ± 2.6 4.3 ± 0.5<br />
cultiv. June L. digitata 2 mo starving/3 mo feeding 16°C 12h / 12h 11.1 ± 2.6 7.3 ± 1.7 4.4 ± 0.5<br />
cultiv. May pellets (c)<br />
cultiv. June pellets (c)<br />
cultiv. Oct. pellets (c)<br />
2 mo starving/2 mo feeding 16°C 12h / 12h 11.2 ± 3.3 6.7 ± 2.1 4.2 ± 1.3<br />
2 mo starving/3 mo feeding 16°C 12h / 12h 17.5 ± 2.4 11.3 ± 1.5 6.2 ± 1.0<br />
2 mo starving/1.5 mo feeding 16°C 17h / 7h 13.9 ± 1.5 9.7 ± 1.0 4.8 ± 0.9<br />
(a) Best MI values for <strong>the</strong> market range from 4 to 5 (growing and premature reproductive stages).<br />
(b) Mean values obtained on samplings during 3 consecutive years.<br />
(c) For <strong>the</strong> composition <strong>of</strong> this food, see Williams and Harris, 1998 (<strong>the</strong>ir Table 1, "new diet").<br />
Table 3 presents some results obtained after conditioning <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong>s for <strong>the</strong> market with <strong>the</strong> starving-feeding method. To allow<br />
comparisons, GI and MI <strong>of</strong> field echinoids issued from Brittany are also<br />
provided. In <strong>the</strong> field, best GI was observed in March and reaches a mean<br />
value <strong>of</strong> 11.6% in fresh weight. Sea <strong>urchin</strong>s conditioned in cultivation<br />
82
e. Discussion<br />
show similar GI and MI. The feeding period must be extended to 3 months<br />
when using L. digitata, whereas 2 months are sufficient with <strong>the</strong> artificial<br />
diet to obtain <strong>the</strong> same results. Feeding 3 months with <strong>the</strong> pellets leads to a<br />
remarkable mean GI <strong>of</strong> 17.5% in fresh weight. Such a GI has never been<br />
observed in <strong>the</strong> field and corresponds to <strong>the</strong> complete filling <strong>of</strong> <strong>the</strong><br />
coelomic cavity with <strong>the</strong> gonads, <strong>the</strong> digestive tract, almost empty, being<br />
compressed against <strong>the</strong> body wall. However, <strong>the</strong> MI is too high and <strong>the</strong><br />
gonads contain too many gametes to satisfy market criteria. Fur<strong>the</strong>rmore,<br />
<strong>the</strong> color obtained with <strong>the</strong> pellets is too pale (white to beige) and <strong>the</strong> taste<br />
does not match wild roe, whereas gonads produced with L. digitata are <strong>of</strong><br />
good quality. An out-<strong>of</strong>-<strong>sea</strong>son conditioning was initiated in July with a<br />
long-day photoperiod (17h / 7h). Very large gonads (GI around 14%) with<br />
an adequate MI were obtained in October, after only 6 weeks <strong>of</strong> feeding<br />
with <strong>the</strong> artificial food. Hence, <strong>the</strong> starving-feeding method could be used<br />
to produce marketable gonads all year long.<br />
Both <strong>the</strong> increasing demand for roe and systematic overexploitation <strong>of</strong><br />
wild populations support <strong>the</strong> need for a <strong>sea</strong> <strong>urchin</strong> cultivation independent<br />
<strong>of</strong> field resources. The method presented here is one design <strong>of</strong> a rearing<br />
process that satisfies this criterion. It appears to be successful at any life<br />
stages <strong>of</strong> P. lividus on a pilot scale.<br />
Obtaining gametes <strong>of</strong> P. lividus in large amounts is an easy task, as is<br />
<strong>the</strong> rearing <strong>of</strong> its larvae with <strong>the</strong> proposed method (rudimentary devices,<br />
low maintenance and feeding with one <strong>of</strong> <strong>the</strong> easiest microalgae to grow:<br />
Phaeodactylum tricornutum). Metamorphosis is a little bit more critical<br />
but can be achieved with care and use <strong>of</strong> a good inductant (fresh coralline<br />
algae). Rearing <strong>of</strong> juveniles, subadults and adults is feasible if five major<br />
constraints are simultaneously respected. A specific design <strong>of</strong> <strong>the</strong> rearing<br />
structures and baskets provides (1) correct water flow around <strong>the</strong> echinoids<br />
(for gas exchanges and removal <strong>of</strong> solid wastes) and (2) sufficient bottom<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
83
surface on which those benthic animals can settle (stacked toboggans). The<br />
maintenance <strong>of</strong> good water quality is ensured by (3) <strong>the</strong> adaptation <strong>of</strong> <strong>the</strong><br />
<strong>sea</strong> <strong>urchin</strong> density at each life stage and (4) water renewal fixed at a<br />
sufficient rate to minimize pollution and avoid depletion in carbonates (see<br />
rearing method for tolerable values for both parameters without<br />
supplemental filtration). Finally, (5) providing adequate food ad libitum<br />
promotes somatic and gonadal growth. Fur<strong>the</strong>r regulation <strong>of</strong> resources<br />
allocation is possible by diet (starving-feeding method), temperature and<br />
photoperiod conditions, leading to good quality <strong>of</strong> <strong>the</strong> final product –<strong>the</strong><br />
roe–, which could be obtained all year long.<br />
If all <strong>the</strong>se five conditions are met, P. lividus behaves fairly well in<br />
cultivation and seems highly resistant to di<strong>sea</strong>ses. The only cases <strong>of</strong><br />
di<strong>sea</strong>se observed (mainly necrosis on <strong>the</strong> test or spines) were attributed to<br />
opportunistic bacterial or fungal infections attributable to poor rearing<br />
conditions, that is, when one or several <strong>of</strong> <strong>the</strong>se five parameters were<br />
poorly controlled. Cannibalism was also observed when <strong>the</strong> quality <strong>of</strong> food<br />
was low or when carbonates concentration or pH dropped ("foraging"<br />
behavior to compensate <strong>the</strong> lack in calcium carbonates?) or on dying<br />
animals, but never on healthy individuals maintained in good condition.<br />
<strong>Growth</strong> is perfectly asymptotic, and <strong>the</strong> maximal size <strong>of</strong> 45 to 65 mm<br />
(individual variation) is reached around 3.5 to 4 years old in <strong>the</strong> rearing<br />
conditions mentioned above. This size is similar to that observed among<br />
<strong>the</strong> field population <strong>of</strong> Morgat, from which <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s descend<br />
directly or indirectly. In Brittany, <strong>the</strong> most precise estimation <strong>of</strong> size at age<br />
for wild populations <strong>of</strong> P. lividus has been performed by Allain (1978) by<br />
analysis <strong>of</strong> <strong>the</strong> growth bands in <strong>the</strong> skeleton. According to this author, wild<br />
<strong>sea</strong> <strong>urchin</strong>s reach <strong>the</strong> size <strong>of</strong> 40 to 50 mm in 4 years, which is a little bit<br />
longer than in <strong>the</strong> present rearing conditions (between 2 and 3.5 years).<br />
The gain could probably be attributed to <strong>the</strong> food distributed ad libitum all<br />
year long and to <strong>the</strong> water temperature (heating <strong>of</strong> <strong>the</strong> water in <strong>the</strong> winter)<br />
as already suggested by Le Gall (1990).<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
84
The success <strong>of</strong> <strong>the</strong> present method leads to optimistic forecasting for<br />
<strong>the</strong> future <strong>of</strong> echiniculture. However, we should probably expect slightly<br />
different results with large-scale, intensive cultivation. With <strong>the</strong> experience<br />
acquired during this long-term trial and some informal observations<br />
performed at a larger scale, we can predict some problems that could<br />
potentially arise when scaling up or when considering pr<strong>of</strong>it. These<br />
problems can be ranged into four different categories: (1) loss <strong>of</strong> pr<strong>of</strong>it due<br />
to high and/or uncontrolled mortality <strong>of</strong> juveniles and subadults; (2)<br />
unevenly distributed growth rates due to intraspecific competition; (3) lack<br />
<strong>of</strong> carbonates and accumulation <strong>of</strong> CO2 because <strong>of</strong> skeletogenesis in<br />
intensive closed-circuit systems; and (4) problems linked to <strong>the</strong> quality <strong>of</strong><br />
food, water pollution or poor color and/or taste <strong>of</strong> gonads produced with<br />
artificial diets.<br />
Survival rates around <strong>the</strong> critical period when <strong>the</strong> postlarva acquires<br />
exotrophy to become fully functional juvenile are highly unpredictable. To<br />
get over <strong>the</strong> difficult phase <strong>of</strong> endotrophy, <strong>the</strong> larvae must store enough<br />
reserves before undergoing metamorphosis. In addition, <strong>the</strong> early juveniles<br />
must promptly find suitable food when <strong>the</strong>ir digestive tract becomes<br />
functional. It seems that one or both parameters are not always optimal in<br />
rearing conditions. In any way, with a mean 55% success rate, we obtain<br />
over 12,000 viable juveniles per 200-l tank which is enough for our use but<br />
can probably be improved. Indeed, gametes are not limited: a female <strong>of</strong><br />
40-mm diameter usually produces around 5 to 7 millions <strong>of</strong> eggs. Thus, <strong>the</strong><br />
50,000 embryos introduced in one larval rearing tank represent only about<br />
1% <strong>of</strong> a whole spawn (about 0.2% <strong>of</strong> <strong>the</strong> sperm produced by a single<br />
male). Hence, only a few dozen mature adults are necessary to produce<br />
enough gametes for mass production <strong>of</strong> larvae.<br />
However, after <strong>the</strong> critical phase <strong>of</strong> exotrophic acquisition, <strong>the</strong><br />
mortality <strong>of</strong> juveniles remains very high until <strong>the</strong>y reach about 10 mm in<br />
test diameter. To minimize this, juveniles are <strong>reared</strong> in specific structures<br />
referred to as pregrowth structures where biomass is kept at a low level<br />
and where water quality is <strong>of</strong> prime importance. Moreover, quality <strong>of</strong> <strong>the</strong><br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
85
immediate environment <strong>of</strong> juveniles is improved by use <strong>of</strong> a good "waterresistant"<br />
diet (Enteromorpha linza) and by means <strong>of</strong> cleaners (Gammarus<br />
locusta). In any case, <strong>the</strong> space occupied in <strong>the</strong> pregrowth structure by<br />
juveniles and <strong>the</strong> total care <strong>the</strong>y need remain much lower as compared to<br />
subadults and adults (compare densities in Table 2). This minimizes <strong>the</strong><br />
cost <strong>of</strong> losing many juveniles from <strong>the</strong> point <strong>of</strong> view <strong>of</strong> <strong>the</strong> total<br />
productivity <strong>of</strong> <strong>the</strong> cultivation.<br />
More insidious is <strong>the</strong> effect <strong>of</strong> winter mortality <strong>of</strong> subadults and adults.<br />
Its cumulative value is ten times lower than juvenile mortality, but its cost<br />
is much higher, because it concerns individuals occupying a significant<br />
space in <strong>the</strong> growth structures and having already consumed a significant<br />
amount <strong>of</strong> food (drop <strong>of</strong> <strong>the</strong> overall yield per surface unit and food<br />
conversion efficiency). However, <strong>the</strong> cause <strong>of</strong> this <strong>sea</strong>sonal variability<br />
cannot be explained. It could be because <strong>of</strong> lower quality <strong>of</strong> food (fresh<br />
kelp with a <strong>sea</strong>sonal variation in <strong>the</strong>ir composition, Gayral & Cosson,<br />
1973; Abe et al, 1983), or to any pollution <strong>of</strong> <strong>the</strong> water probably induced<br />
by <strong>the</strong> food itself (bad quality food is less ingested and decomposes more<br />
easily), or to ano<strong>the</strong>r undetermined cause. For <strong>the</strong> moment, waves <strong>of</strong> mass<br />
mortality have not been correlated with ei<strong>the</strong>r temperature variability <strong>of</strong><br />
<strong>the</strong> natural <strong>sea</strong>water, meteorological conditions (atmospheric pressure,<br />
rain) or feeding. However, any correlation will be difficult to assess,<br />
because <strong>of</strong> <strong>the</strong> scarcity <strong>of</strong> <strong>the</strong>se mass mortality waves and <strong>the</strong> probable but<br />
not quantified delay between <strong>the</strong> stress and <strong>the</strong> observed mortality. Total<br />
productivity could undoubtedly be enhanced if this winter mortality was<br />
lowered or eliminated. To suppress or minimize <strong>the</strong> winter decrease in <strong>the</strong><br />
biomass is also worth considering when one intends to produce marketable<br />
gonads all year long.<br />
Mortality is not <strong>the</strong> only problem inhibiting steady productivity:<br />
widespread distribution <strong>of</strong> growth speed among individuals expands <strong>the</strong><br />
time interval when largest and smallest individuals are exploitable and<br />
constrains to sort batches frequently. <strong>Growth</strong> <strong>of</strong> P. lividus is very slow at<br />
<strong>the</strong> juvenile stage. This "lag-phase" has also been observed by Cellario &<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
86
Fenaux (1990) for <strong>the</strong> same species in cultivation and by Ebert & Russell<br />
(1993) for wild populations <strong>of</strong> Strongylocentrotus franciscanus. When<br />
growth initiates in term <strong>of</strong> test diameter, size distribution expands. This<br />
individual variability is not genetic but is attributable to a reversible sizebased<br />
intraspecific competition (Grosjean et al, 1996, see Part III) that<br />
takes place rapidly, even in size-sorted batches, although sorting reduces<br />
its effect. Presently, <strong>the</strong> exact impact <strong>of</strong> this competition on productivity<br />
and <strong>the</strong> best way to avoid it (if it should be avoided at all) are still<br />
unknown.<br />
A third problem that will probably occur when considering fur<strong>the</strong>r<br />
intensification <strong>of</strong> echiniculture in closed or semiclosed systems is <strong>the</strong><br />
depletion <strong>of</strong> dissolved carbonates and <strong>the</strong> accumulation <strong>of</strong> CO2 in<br />
<strong>sea</strong>water. In growing, <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> builds a magnesium-calcite skeleton.<br />
This skeleton represents an important fraction <strong>of</strong> <strong>the</strong> body weight: between<br />
28% and 31% <strong>of</strong> <strong>the</strong> total fresh weight for P. lividus (measured on animals<br />
issued from <strong>the</strong> <strong>reared</strong> strain, after digestion <strong>of</strong> organic tissues with a<br />
12°Chl bleaching agent under gentle agitation, n = 356). Thus, for each kg<br />
<strong>of</strong> fresh weight produced, about one-third has to be supplied in one or <strong>the</strong><br />
o<strong>the</strong>r form <strong>of</strong> calcium carbonate. However, P. lividus is unable to<br />
assimilate efficiently carbonates provided as a solid substrate (calcareous<br />
rocks, algae or cuttlefish bones for instance) because <strong>the</strong> pH <strong>of</strong> its<br />
digestive tract is too high to dissolve large amounts <strong>of</strong> solid calcite<br />
(between six and eight, for a review see Lawrence, 1982; for data<br />
concerning P. lividus see Claerebout & Jangoux, 1985). The main usable<br />
source <strong>of</strong> magnesium/calcium carbonates is thus present under a dissolved<br />
form in <strong>sea</strong>water. If calcium and magnesium ions (respectively 400 mg<br />
and 1,350 mg per kg <strong>sea</strong>water at a salinity <strong>of</strong> 35‰, Spotte, 1991) are not<br />
limiting, <strong>the</strong> quantity <strong>of</strong> dissolved carbonates available could be consumed<br />
very quickly in intensive closed or semiclosed systems (unpublished data).<br />
Most <strong>of</strong> <strong>the</strong> carbonate alkalinity (about 2.3 - 2.4 meq / kg <strong>sea</strong>water,<br />
corresponding to 140 mg <strong>of</strong> HCO3 - ) remains unavailable for<br />
skeletogenesis, <strong>the</strong> pH dropping too much when <strong>sea</strong> <strong>urchin</strong>s consume it<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
87
(carbonate and bicarbonate are <strong>the</strong> most important chemical components<br />
that buffer pH in <strong>sea</strong>water, Stumm & Morgan, 1981). The actual fraction<br />
<strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s can use is still unknown, but is probably under 10% <strong>of</strong> <strong>the</strong><br />
total carbonate alkalinity. To illustrate this, without supplemental chemical<br />
filtration and with a usable fraction <strong>of</strong> 10% <strong>of</strong> <strong>the</strong> dissolved carbonates to<br />
produce skeleton that final weight represents 30% <strong>of</strong> <strong>the</strong> total <strong>sea</strong> <strong>urchin</strong><br />
fresh weight, one must provide at least 24,500 m 3 <strong>of</strong> <strong>sea</strong>water per ton <strong>of</strong><br />
<strong>sea</strong> <strong>urchin</strong> fresh weight produced. However, this optimistic calculation<br />
does not consider mortality that o<strong>the</strong>rwise also exports carbonates.<br />
Precipitation <strong>of</strong> bicarbonates (<strong>the</strong> main form <strong>of</strong> dissolved carbonates in<br />
<strong>sea</strong>water at usual pH) into calcium carbonate is a dismutation reaction that<br />
liberates a stoichiometric amount <strong>of</strong> carbonic acid in <strong>the</strong> water column.<br />
This carbonic acid, toge<strong>the</strong>r with <strong>the</strong> CO2 produced by <strong>the</strong> respiration <strong>of</strong><br />
<strong>sea</strong> <strong>urchin</strong>s, algae and bacteria in <strong>the</strong> rearing structures tends to reach<br />
rapidly undesired levels in a large-scale intensive cultivation. We have<br />
observed <strong>sea</strong> <strong>urchin</strong>s whose skeleton growth was totally inhibited in <strong>the</strong>se<br />
conditions. CO2 partial pressure was recorded to be 5 to 9 times higher<br />
than usual in <strong>sea</strong>water (despite a strong aeration <strong>of</strong> <strong>the</strong> water) and was<br />
presumed to be <strong>the</strong> direct cause <strong>of</strong> <strong>the</strong> inhibition <strong>of</strong> <strong>the</strong> skeletogenesis.<br />
These limitations force us to choose ei<strong>the</strong>r a flow-through system or to<br />
provide a chemical filtration to level carbonates and carbonic acid<br />
concentrations. The present method could be considered as a semiintensive,<br />
semiclosed system where both <strong>sea</strong> <strong>urchin</strong> densities and water<br />
renewals remain compatible with <strong>the</strong> equilibrium <strong>of</strong> <strong>the</strong> inorganic carbon<br />
in <strong>sea</strong>water without supplemental filtration. However, such a trade-<strong>of</strong>f<br />
would not be compatible with a rearing strategy aiming to raise pr<strong>of</strong>it on a<br />
large scale.<br />
For <strong>the</strong> moment, fresh algae used as food form part <strong>of</strong> <strong>the</strong> natural diet<br />
<strong>of</strong> P. lividus. The composition <strong>of</strong> this food is presumably correct, although<br />
it might not be necessarily optimal (Frantzis & Grémare, 1992; Gonzalez<br />
et al, 1993; Fernandez & Boudouresque, 1998). The major problem<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
88
encountered with food is its stability once put in <strong>the</strong> rearing structures<br />
because this echinoid, being a grazer, ingests it slowly. Uneaten food could<br />
easily give rise to undesired pollution. Hence, we recommend <strong>the</strong> use <strong>of</strong> a<br />
stable diet (Enteromorpha linza) in <strong>the</strong> present rearing method instead <strong>of</strong><br />
higher quality algae (Laminaria digitata, L. saccharina Lamouroux or<br />
Rhodymenia palmata (L.) Greville, unpublished results) for juveniles. We<br />
also avoid using artificial diets at water temperature above 16°C without<br />
<strong>the</strong> presence <strong>of</strong> an efficient bi<strong>of</strong>ilter in <strong>the</strong> rearing structures.<br />
The use <strong>of</strong> fresh algae is not always possible or pr<strong>of</strong>itable on a large<br />
scale (Fernandez, 1996). Hence, an artificial diet designed specifically for<br />
<strong>sea</strong> <strong>urchin</strong>s seems necessary for intensified echiniculture and is presently<br />
under investigation by several authors (Fernandez & Caltagirone, 1994;<br />
Klinger et al, 1994, 1997, 1998; de Jong-Westman et al, 1995a, 1995b;<br />
Fernandez, 1996). Results obtained so far are encouraging, especially in<br />
term <strong>of</strong> GI but <strong>the</strong> food we were able to test gave unsatisfactory results in<br />
terms <strong>of</strong> color and palatability <strong>of</strong> <strong>the</strong> roe. Recent testing <strong>of</strong> semimoist diets<br />
on Strongylocentrotus droebachiensis (Motnikar et al, 1997) seems to<br />
confirm <strong>the</strong> positive effect <strong>of</strong> <strong>the</strong> artificial diet on <strong>the</strong> gonadosomatic index<br />
and <strong>the</strong> failure to obtain high quality gonads in terms <strong>of</strong> color and taste.<br />
Trials with carotenoids-enriched artificial food to enhance <strong>the</strong> color do not<br />
yet produce high quality gonads (Goebel & Barker, 1998). Thus, a better<br />
formulation <strong>of</strong> <strong>the</strong> food is basic to achieve a correct taste and color for<br />
exploitation.<br />
Finally we should mention that <strong>the</strong> rearing method described here is<br />
labor-intensive. Hence, manpower cost could be too high when<br />
considering pr<strong>of</strong>it. This would require some adaptation or mechanization<br />
<strong>of</strong> <strong>the</strong> most time-consuming operations: feeding subadults and adults,<br />
cleaning <strong>the</strong> growth structures, grading <strong>the</strong> batches or extracting <strong>the</strong><br />
gonads if <strong>sea</strong> <strong>urchin</strong>s are not commercialized alive (exportation to Japan).<br />
However, <strong>the</strong>se are technical problems that could be solved by <strong>the</strong><br />
industry.<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
89
f. Conclusions<br />
This rearing method constitutes a good working basis to design a<br />
closed-cycle, land-based echiniculture. We suggest it could be used as a<br />
standard method to evaluate improvement obtained by adaptations or<br />
modifications aimed at intensification or pr<strong>of</strong>itability <strong>of</strong> echiniculture. This<br />
method could possibly be adapted to o<strong>the</strong>r species, allowing better<br />
comparisons <strong>of</strong> <strong>the</strong> biology <strong>of</strong> respective species as well as <strong>the</strong>ir<br />
aquaculture potentials.<br />
Latent remaining problems when scaling up and intensifying<br />
cultivation, aiming at raising pr<strong>of</strong>it, should not be regarded as unavoidable<br />
limitations, but should be considered as challenges to address in fur<strong>the</strong>r<br />
studies. Being "new" cultivated species, it is not surprising that <strong>the</strong>se<br />
obstacles mostly concern less known life stages or "biological features or<br />
characteristics" <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s: <strong>the</strong> transition between <strong>the</strong> endotrophic<br />
postlarva and <strong>the</strong> exotrophic juvenile, <strong>the</strong> mechanism <strong>of</strong> <strong>the</strong> intraspecific<br />
competition, <strong>the</strong> carbonate budget needed for skeletogenesis and <strong>the</strong><br />
biochemical pathways in gametogenesis and in stocking reserve material in<br />
<strong>the</strong> gonads. Thus, it is probable that advances in fundamental biology <strong>of</strong><br />
echinoderms, and more particularly <strong>of</strong> echinoids, will suggest solutions to<br />
<strong>the</strong>se problems in <strong>the</strong> future.<br />
It would seem that fur<strong>the</strong>r development <strong>of</strong> closed-cycle, land-based <strong>sea</strong><br />
<strong>urchin</strong> cultivation is worthwhile and will undoubtedly promote<br />
diversification <strong>of</strong> aquaculture and production <strong>of</strong> high quality <strong>sea</strong>food. This<br />
will, secondarily, lead to <strong>the</strong> conservation <strong>of</strong> <strong>the</strong> natural environment by<br />
limiting <strong>the</strong> fisheries impact on natural populations <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s.<br />
g. Acknowledgements<br />
This study was conducted in <strong>the</strong> framework <strong>of</strong> EEC contracts in <strong>the</strong><br />
"AIR" and "FAR" aquaculture program (ref. AQ2.530 BFE & CT96.1623<br />
BFN). This re<strong>sea</strong>rch was also supported by an EC re<strong>sea</strong>rch grant attributed<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
90
to Christine Spirlet (ref. ERB 4001 GT92 0223), in <strong>the</strong> framework <strong>of</strong> <strong>the</strong><br />
Sea Urchin Cultivation contract n° AQ 2.530 BFE. We thank Didier<br />
Bucaille for his help in <strong>the</strong> laboratory work and <strong>the</strong> CREC (University <strong>of</strong><br />
Caen) for its financial contribution in <strong>the</strong> building <strong>of</strong> <strong>the</strong> specific <strong>sea</strong><br />
<strong>urchin</strong>s facility. We are grateful to John Lawrence for providing <strong>the</strong><br />
artificial diet and to Addison Lawrence and <strong>the</strong> Wenger Company for<br />
designing and producing it. Thanks to Raphaël Morgan for pro<strong>of</strong>reading<br />
<strong>the</strong> manuscript. This paper is a contribution to <strong>the</strong> "Centre<br />
Interuniversitaire de Biologie Marine" (CIBIM).<br />
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
91
Part I: Set up <strong>of</strong> an experimental rearing procedure for echinoids<br />
92
PART II<br />
Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
93
PART II: MEASUREMENT FOR SIZE IN THE SEA URCHIN<br />
Having a rearing method to grow P. lividus in aquaria, we still have to<br />
decide how to measure growth, that is, size increase <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
between various time intervals. Clearly <strong>the</strong> best measurement method<br />
should be both rapid (to allow measuring hundreds <strong>of</strong> individuals in a<br />
reasonable time) and as accurate and reproducible as possible. That<br />
measurement should be also most representative <strong>of</strong> somatic growth.<br />
Finally it should be harmless, since successive measures <strong>of</strong> <strong>the</strong> same<br />
animals will be performed.<br />
Various direct measurements <strong>of</strong> body size are available for <strong>sea</strong> <strong>urchin</strong>s:<br />
diameter or height <strong>of</strong> <strong>the</strong> test, volume, total fresh weight or 'immersed<br />
weight'. Since <strong>sea</strong> <strong>urchin</strong> has a rigid endoskeleton, its shape is constrained<br />
and it is easy to take a linear measurement: ei<strong>the</strong>r <strong>the</strong> diameter or <strong>the</strong><br />
height <strong>of</strong> its test. However, mouth is not rigidly fixed to <strong>the</strong> test. A<br />
measure <strong>of</strong> height, that is from mouth to anus, is thus less precise than a<br />
measure <strong>of</strong> diameter, and we did not consider it. Volume measurement was<br />
also eliminated for it is too inaccurate: <strong>the</strong> volume <strong>of</strong> 15 <strong>sea</strong> <strong>urchin</strong>s<br />
measured 8 times each using <strong>the</strong> 'displacement apparatus' described in<br />
Comely & Ansell (1988, <strong>the</strong>ir Fig. 2) has an accuracy (expressed as<br />
standard deviation in percent <strong>of</strong> <strong>the</strong> mean) <strong>of</strong> ca. 10.9% when accuracies<br />
for diameter, fresh weight or immersed weight range from 0.6 to 2.5% (see<br />
Table 4, p. 101).<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
95
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
96
Comparison <strong>of</strong> three body-size measurements for echinoids<br />
a. Abstract<br />
b. Introduction<br />
Ph. Grosjean, Ch. Spirlet & M. Jangoux, 1999. Echinoderm Re<strong>sea</strong>rch<br />
1998. M.D. Candia Carnevali & F. Bonasoro (eds). Balkema,<br />
Rotterdam. Pp 31-35.<br />
Several measurements can be employed to quantify <strong>the</strong> body size <strong>of</strong><br />
echinoids. We evaluate here <strong>the</strong> accuracy <strong>of</strong> three measurements on <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong> <strong>Paracentrotus</strong> lividus (test diameter, fresh body weight and<br />
immersed weight –<strong>the</strong> weight <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> when immersed in<br />
<strong>sea</strong>water–) and discuss <strong>the</strong>ir respective potentials. The immersed weight<br />
appears to be by far <strong>the</strong> most accurate, providing it is standardized, but<br />
also <strong>the</strong> most time-costly measurement. Allometric relationships and<br />
formula for calculating a standard immersed weight for P. lividus are also<br />
provided.<br />
In vivo determination <strong>of</strong> <strong>the</strong> body size is a basic approach used in many<br />
fields in biology, including individual growth (Ebert, 1967; Grosjean et al,<br />
1996; Régis, 1969), population dynamics (Allain, 1972a; Régis & Arfi,<br />
1978), morphometry or biometry (Ebert, 1968, 1981, 1988b; Lawrence et<br />
al, 1995; Moss & Meehan, 1968), physiology (through calculation <strong>of</strong><br />
gonadal or repletion indices, for instance Agatsuma & Sugawara, 1988;<br />
Giese, 1966; Nedelec, 1983; Spirlet et al, 1998a). Various kinds <strong>of</strong><br />
measurements are available, from lengths to volumes or weights. For<br />
echinoids, this task is facilitated by <strong>the</strong> presence <strong>of</strong> a rigid endoskeleton<br />
that restrains both <strong>the</strong>ir external dimensions and <strong>the</strong>ir total<br />
volume / weight. Hence, several measurements are accessible and used to<br />
quantify <strong>the</strong> body size.<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
97
Few studies have compared and discussed <strong>the</strong> accuracy and suitability<br />
<strong>of</strong> <strong>the</strong>se various measurements, except for <strong>the</strong> classical relationship<br />
between <strong>the</strong> test diameter and <strong>the</strong> body weight (Agatsuma & Sugawara,<br />
1988; Allain, 1972a; Kaneko et al, 1981). Consequently, authors use<br />
different body size measurements that are not always optimal for <strong>the</strong>ir<br />
studies.<br />
In <strong>the</strong> present study, <strong>the</strong> accuracy, reproducibility and suitability <strong>of</strong><br />
some measurements that can be performed in vivo on <strong>sea</strong> <strong>urchin</strong>s were<br />
assessed. The three selected measurements are test diameter, fresh body<br />
weight and immersed weight.<br />
c. Material and methods<br />
A stratified sample <strong>of</strong> 224 <strong>Paracentrotus</strong> lividus <strong>sea</strong> <strong>urchin</strong>s <strong>of</strong> various<br />
sizes (20 to 25 individuals in each 5 mm size-class ranging from 10 to 60<br />
mm in test diameter) was measured. Half <strong>of</strong> <strong>the</strong>m where directly collected<br />
in <strong>the</strong> field in Morgat, Brittany (France). The o<strong>the</strong>rs where cultivated<br />
specimens from <strong>the</strong> Marine Station <strong>of</strong> Luc-sur-Mer, Normandy (France)<br />
(see Grosjean et al, 1998 –Part I– for <strong>the</strong> protocol), but which field parents<br />
originated also from Morgat. In both field and cultivated individuals, half<br />
population was measured in September, and half was measured in March<br />
in order to assess a possible <strong>sea</strong>sonal variation (Spirlet et al, 1998a).<br />
- The diameter is measured to <strong>the</strong> nearest 0.1 mm with a sliding caliper to<br />
<strong>the</strong> widest part <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> (<strong>the</strong> ambitus), and without considering <strong>the</strong><br />
spines. In case <strong>of</strong> a possible oval shape, it is <strong>the</strong> average <strong>of</strong> two<br />
perpendicular diameters taken to <strong>the</strong> ambitus that is considered.<br />
- The total fresh weight is measured to <strong>the</strong> nearest 0.001 g after leaving <strong>the</strong><br />
<strong>sea</strong> <strong>urchin</strong>s for 5 minutes (stabilization <strong>of</strong> weight) on absorbent paper,<br />
which prevents possible fluctuations due to residual water on <strong>the</strong><br />
integument surface.<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
98
- The immersed weight is a much less customary measurement and its use<br />
is fur<strong>the</strong>r discussed. In <strong>the</strong> present case, it corresponds to <strong>the</strong> apparent<br />
weight <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong> in <strong>sea</strong>water. It is assessed with a scale (precision <strong>of</strong><br />
0.1%) provided with a plate or basket immerged in a tank <strong>of</strong> <strong>sea</strong>water and<br />
containing <strong>the</strong> individuals to be weighed (see Fig. 20).<br />
1<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
2<br />
3<br />
4<br />
EXCEL<br />
Figure 20. Data acquisition system. A scale (1), an electronic sliding caliper (2) and a second<br />
scale equipped with a plate immersed in a tank filled with <strong>sea</strong> water (3) are connected to a<br />
computer (4) for fast data treatment (a s<strong>of</strong>tware to acquire data directly from scales and<br />
calipers into a PC and to calculate <strong>the</strong> SIW is freely available from <strong>the</strong> authors).<br />
Each specimen was measured only once. However, to assess <strong>the</strong><br />
reproducibility <strong>of</strong> <strong>the</strong> 3 types <strong>of</strong> measurement and <strong>the</strong> possible variation<br />
due to <strong>the</strong> experimenter, an additional 15 echinoids were measured 3 times<br />
at a 4 h interval by 7 different people. The individuals were distributed<br />
randomly to <strong>the</strong> experimenters. Data were collected by means <strong>of</strong> a data<br />
acquisition system composed <strong>of</strong> an electronic sliding caliper and electronic<br />
scales connected to a computer (Fig. 20).<br />
99
d. Results and discussion<br />
The ambital test diameter and <strong>the</strong> total fresh weight measurements are<br />
exploitable directly. The values <strong>of</strong> <strong>the</strong> immersed weight can be compared<br />
only if <strong>the</strong> density <strong>of</strong> <strong>the</strong> <strong>sea</strong>water (depending upon salinity and<br />
temperature) is constant between measurements, o<strong>the</strong>rwise a correction<br />
factor must be introduced. The immersed weight is <strong>the</strong> resultant <strong>of</strong> 2<br />
opposite forces, <strong>the</strong> weight and <strong>the</strong> buoyancy, which compensate each<br />
o<strong>the</strong>r for organs <strong>of</strong> <strong>the</strong> same density as <strong>sea</strong> water: gonads, digestive tract,<br />
and coelomic fluid (Stickle & Ahokas, 1974). Their apparent weight in<br />
<strong>sea</strong>water is thus close to zero. Conversely, <strong>the</strong> calcareous skeleton has a<br />
significant positive apparent weight which means that <strong>the</strong> immerged<br />
weight is primarily a measure <strong>of</strong> <strong>the</strong> apparent weight <strong>of</strong> <strong>the</strong> skeleton in<br />
<strong>sea</strong>water. This is also evidenced in Table 5, showing <strong>the</strong> immersed weight<br />
is directly proportionate to <strong>the</strong> dry weight <strong>of</strong> <strong>the</strong> skeleton (allometric<br />
coefficient = 1.00 = perfect isometry).<br />
We define <strong>the</strong> standard immersed weight (SIW) as <strong>the</strong> immersed<br />
weight that would have been measured in a liquid which density is strictly<br />
equal to 1.00·10 3 g/l; we calculate it as follows:<br />
2.80 −1.00<br />
SIW = IW.<br />
2.80 − Md /1000<br />
where: - SIW is <strong>the</strong> standard immersed weight in g,<br />
- IW is <strong>the</strong> measured immersed weight in g,<br />
- Md is <strong>the</strong> mass density <strong>of</strong> <strong>sea</strong> water where echinoids are<br />
measured, in g/l,<br />
- 2.80 is <strong>the</strong> apparent mean density <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> skeleton in<br />
10 3 g/l (δs in eq. 19).<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
(18)<br />
The mass density <strong>of</strong> <strong>the</strong> <strong>sea</strong>water can be determined ei<strong>the</strong>r directly<br />
(with a densitometer), or by calculation (Cox et al, 1970; UNESCO, 1981).<br />
In <strong>the</strong> second case, both <strong>the</strong> salinity and <strong>the</strong> temperature <strong>of</strong> <strong>the</strong> water are<br />
needed.<br />
100
The apparent mean density <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> skeleton δs is calculated<br />
from <strong>the</strong> isometric relationship between <strong>the</strong> immersed weight measured at<br />
a constant <strong>sea</strong>water density (1.023·10 3 g/l) and <strong>the</strong> dry weight <strong>of</strong> <strong>the</strong><br />
skeleton DWs (n = 63, R 2 = 0.999):<br />
δ s<br />
DWs = 1.576⋅ IW = ⋅IW⇔δs ≈2.80<br />
δ −1.023<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
s<br />
(19)<br />
The SIW is usually 2 to 3% higher than <strong>the</strong> immersed weight actually<br />
measured.<br />
General comparison<br />
The diameter is <strong>the</strong> fastest and easiest measurement in <strong>the</strong> field (see<br />
Table 4). It is <strong>the</strong> most convenient parameter for separating <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
in size categories or for measuring individually large amounts <strong>of</strong><br />
echinoids. The o<strong>the</strong>rs are suited more for batch evaluation (total biomass<br />
for instance). Fresh weight and SIW take more time. Since several<br />
individuals can be drought simultaneously for <strong>the</strong> fresh weight, time spent<br />
for each measurement drops to around 40 s, instead <strong>of</strong> <strong>the</strong> overall 5 min.<br />
Hence, <strong>the</strong> SIW is <strong>the</strong> longest measurement because <strong>the</strong> scale takes a while<br />
to stabilize. However, it is both <strong>the</strong> most accurate and <strong>the</strong> less stressful<br />
measurement, which can be <strong>of</strong> importance when working on sexually<br />
mature individuals that can spawn when handled.<br />
Table 4. Comparison <strong>of</strong> <strong>the</strong> three selected parameters.<br />
Parameter Diameter Fresh weight SIW<br />
Timing < 30s 40s (5min) ca. 2min<br />
Accuracy (a) 1.33-2.52% (b)<br />
> 1.31% < 0.62%<br />
Possible bias (b)<br />
yes no no<br />
Stress medium medium low<br />
Batch measure no yes yes<br />
Field measure (c)<br />
yes no no<br />
(a)<br />
Standard deviation expressed in percent <strong>of</strong> <strong>the</strong> mean.<br />
(b)<br />
Depending on <strong>the</strong> experimenter.<br />
(c)<br />
Easily usable in <strong>the</strong> field and underwater.<br />
101
Reproducibility <strong>of</strong> measurements<br />
The diameter <strong>of</strong> <strong>the</strong> test to <strong>the</strong> ambitus is reproducible when it is done<br />
by <strong>the</strong> same person. A two-way ANOVA (measurement order versus<br />
experimenter) indicates that <strong>the</strong>re is no significant difference between<br />
measures from a single experimenter and no interaction between<br />
measurements and experimenters (p > 0.05 in both cases). However, <strong>the</strong><br />
experimenters have a great influence (p < 0.01) on <strong>the</strong> values recorded.<br />
The same analysis done on fresh weight and SIW reveals <strong>the</strong>re is only<br />
one significant effect (p < 0.01): <strong>the</strong> order <strong>of</strong> measurements. The<br />
difference between 2 successive measurements is steady for all animals, in<br />
all cases. The SIW decreases by an average <strong>of</strong> –0.63% <strong>the</strong>n –0.49%<br />
between measurements which can be due to <strong>the</strong> accidental breaking and<br />
loss <strong>of</strong> spines during handling. Conversely, <strong>the</strong>re is an average increase in<br />
<strong>the</strong> fresh weight <strong>of</strong> successively +1.70% and +0.64% between 2 series.<br />
Such high variation in 4 h intervals can be explained only by slight<br />
variations in <strong>the</strong> volume <strong>of</strong> <strong>the</strong> water confined in <strong>the</strong> echinoid (perivisceral<br />
fluid and/or intradigestive fluid; protrusion more or less important <strong>of</strong> <strong>the</strong><br />
Aristotle's lantern). This is a drawback that would lower both <strong>the</strong> accuracy<br />
and reproducibility <strong>of</strong> fresh weight measure in echinoids.<br />
This analysis reveals that <strong>the</strong> SIW is by far <strong>the</strong> most reproducible and<br />
thus reliable measurement. Its reliability is probably even higher than<br />
shown in Table 4 where <strong>the</strong> loss due to broken spines between<br />
measurements was not deduced from <strong>the</strong> overall recorded variation.<br />
Allometry and measurements relationship<br />
An ANCOVA on log-transformed data (p > 0.05) indicates <strong>the</strong>re is no<br />
effect <strong>of</strong> <strong>the</strong> origin (field or cultivated), or <strong>the</strong> <strong>sea</strong>sons on <strong>the</strong> allometric<br />
relationship between <strong>the</strong> three measurements. Hence, data are pooled.<br />
Table 5 presents <strong>model</strong> I allometric relations between <strong>the</strong> 3 measurements<br />
considered. All regressions are highly significant (R 2 ≥ 98%) for this<br />
species in <strong>the</strong> size range explored. In all cases, <strong>the</strong> double log data<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
102
transformation leads to linear regressions with homoscedasticity <strong>of</strong><br />
variance and random distribution <strong>of</strong> <strong>the</strong> residuals. However, caution is <strong>the</strong><br />
rule as <strong>model</strong> I is not verified (<strong>the</strong> independent variable should be<br />
measured without error which is not <strong>the</strong> case here). A non-biased <strong>model</strong> II<br />
would be more adequate (Ebert, 1981, 1994; Laws & Archie, 1981;<br />
Tessier, 1948) but only biased <strong>model</strong> II are available for such data sets<br />
(Sokal & Rohlf, 1981, p. 549). Since <strong>the</strong> explained variance is higher than<br />
98%, <strong>the</strong> bias remains negligible, whatever <strong>the</strong> <strong>model</strong> chosen. Thus, in this<br />
case, a <strong>model</strong> I is to be preferred for prediction purposes with independent<br />
regressions for reciprocal relationships (Sokal & Rohlf, 1981).<br />
Table 5. Allometric relations (<strong>model</strong> I linear regressions on double log transformed data)<br />
between parameters for <strong>Paracentrotus</strong> lividus from Morgat, n = 224. Verifications are needed<br />
when applying on o<strong>the</strong>r strains, or out <strong>of</strong> <strong>the</strong> announced validity range.<br />
Measured (x) Estimated (y) Allometry R 2<br />
Std err<br />
log(y)<br />
Validity<br />
Diameter (mm) Fresh weight (g) y = 5.50·10 -4 x 2.94<br />
0.997 0.037 10 < x < 60<br />
Diameter (mm) SIW (g) y = 2.40·10 -4 x 2.70<br />
0.986 0.053 10 < x < 60<br />
Fresh weight (g) Diameter (mm) y = 12.7 x 0.35<br />
0.995 0.011 0.5 < x < 90<br />
Fresh weight (g) SIW (g) y = 0.22 x 0.95<br />
0.994 0.034 0.5 < x < 90<br />
SIW (g) Diameter (mm) y = 22.1 x 0.37<br />
0.984 0.019 0.1 < x < 15<br />
SIW (g) Fresh weight (g) y = 4.95 x 1.05<br />
0.994 0.036 0.1 < x < 15<br />
SIW (g) Skeleton(dry w. g) (a)<br />
y = 1.56 x 0.998 0.021 0.1 < x < 15<br />
SIW (g) Soma (dry w. g) y = 1.74 x 0.98<br />
0.999 0.010 0.1 < x < 15<br />
(a)<br />
Measured after digestion <strong>of</strong> <strong>the</strong> organic matter with sodium hypochloride 10% under gentle agitation<br />
and drying for 48h at 70°C.<br />
The SIW being a direct in vivo measurement <strong>of</strong> <strong>the</strong> skeleton weight <strong>of</strong><br />
<strong>the</strong> <strong>sea</strong> <strong>urchin</strong>, <strong>the</strong> latter can be calculated by <strong>the</strong> formula in Table 5. As<br />
<strong>the</strong> soma is composed <strong>of</strong> ca. 90% <strong>of</strong> skeleton (in dry weight, Grosjean,<br />
unpubl.), it is also a reasonably good in vivo estimation <strong>of</strong> <strong>the</strong> somatic dry<br />
weight, after applying possibly a correction calculated after <strong>the</strong> SIW-soma<br />
allometric relationship (Table 5). As such, it allows to follow most<br />
accurately <strong>the</strong> somatic growth <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s. In some experiments<br />
(Grosjean et al, in prep.), we were able to quantify somatic growth <strong>of</strong> <strong>sea</strong><br />
<strong>urchin</strong>s within a 7-days period using <strong>the</strong> SIW, while it would require at<br />
least a 1 or 2 months period to get <strong>the</strong> same accuracy with test diameter or<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
103
e. Conclusions<br />
fresh weight! Caution must be taken, <strong>of</strong> course, when applying<br />
conversions on echinoids in particular physiological state that lead to<br />
variations in allometric relationships, such with starved individuals (Ebert,<br />
1968; Kaneko et al, 1981).<br />
The SIW should be used whenever possible both as a reference<br />
measurement for indices (gonadal index, repletion index…) and for studies<br />
involving somatic growth. Test diameter remains <strong>the</strong> fastest measure, and<br />
thus preferred for measuring large number <strong>of</strong> individuals when accuracy is<br />
not <strong>of</strong> prime importance, or for measures in <strong>the</strong> field. Fresh weight use<br />
should be restricted to <strong>the</strong> determination <strong>of</strong> <strong>the</strong> biomass.<br />
f. Acknowledgements<br />
This work was supported by an EC re<strong>sea</strong>rch grand attributed to Ch.<br />
Spirlet (ref. ERB 4001 GT92 0223), in <strong>the</strong> framework <strong>of</strong> <strong>the</strong> contract No.<br />
AQ2.530 BFE ("Sea <strong>urchin</strong> cultivation"). We thank D. Bucaille, P.<br />
Gosselin, F. Benard & F. Louise for help in measurements. This paper is a<br />
contribution to <strong>the</strong> Centre Interuniversitaire de Biologie Marine (CIBIM).<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
104
Choice <strong>of</strong> measurement<br />
Immersed weight is definitely <strong>the</strong> most accurate and less stressing<br />
measurement <strong>of</strong> <strong>sea</strong> <strong>urchin</strong> body size. However, it is not possible, using<br />
this method, to measure hundreds <strong>of</strong> echinoids in a reasonable period <strong>of</strong><br />
time. We thus used <strong>the</strong> test diameter, as <strong>the</strong> optimal compromise between<br />
accuracy and speed.<br />
The diameter has ano<strong>the</strong>r advantage: it is <strong>the</strong> only parameter that can<br />
be measured on very small animals <strong>of</strong> a few hundreds <strong>of</strong> microns (that is,<br />
<strong>the</strong> size <strong>of</strong> P. lividus just after metamorphosis, see Fig. 2B, p. 37). To do<br />
this, we have to kill and fix <strong>the</strong> individuals (3% glutaraldehyde) in order to<br />
manipulate <strong>the</strong>m and transfer <strong>the</strong>m on a millimeter-graduated support to be<br />
photographed. The picture is digitalized and analyzed with a custom image<br />
analysis s<strong>of</strong>tware (ShellAxis, available at http://www.sciviews.org). For a<br />
description <strong>of</strong> <strong>the</strong> program, see Van Osselaer & Grosjean (2000) and for<br />
extensive tests <strong>of</strong> accuracy and reproducibility <strong>of</strong> measurement with this<br />
s<strong>of</strong>tware, see Van Osselaer (2001).<br />
Measuring body size with good accuracy is one aspect, knowing what<br />
it really means in terms <strong>of</strong> relative sizes <strong>of</strong> <strong>the</strong> different organs during<br />
growth is ano<strong>the</strong>r one. Indeed, successive measurements would be really<br />
representative <strong>of</strong> somatic growth if <strong>the</strong>y were strongly correlated with<br />
growth <strong>of</strong> all somatic organs.<br />
To determine how different compartments <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> vary with<br />
body size, a stratified sample (from 5 to 60 mm every 5 mm) <strong>of</strong> 440<br />
animals was analyzed. 220 <strong>sea</strong> <strong>urchin</strong>s were dissected in spring and 220 in<br />
autumn, which correspond to two contrasting reproductive stages in <strong>the</strong><br />
field: empty or full gonads, in order to make sure maximum variance is<br />
included in <strong>the</strong> dataset. Measurements done on each individual were:<br />
immersed weight; two perpendicular diameters at <strong>the</strong> ambitus; height; total<br />
fresh weight; fresh "drained" weight (that is, <strong>the</strong> test is opened by cutting<br />
around <strong>the</strong> ambitus and <strong>the</strong> two resulting parts are left upside down on<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
105
principal axis 3<br />
1<br />
0.8<br />
0.6<br />
0.4<br />
0.2<br />
absorbent paper for 5 min to eliminate most <strong>of</strong> <strong>the</strong> coelomic fluid); fresh<br />
and dry weight (drying at 70°C until constant weight) <strong>of</strong> <strong>the</strong> digestive tract<br />
with its contents, <strong>of</strong> <strong>the</strong> gonads and <strong>of</strong> <strong>the</strong> integuments; dry weight <strong>of</strong><br />
Aristotle's lantern, spines and test after elimination <strong>of</strong> most organic part<br />
with 12°Chl bleach under slow agitation.<br />
0<br />
0<br />
0.2<br />
0.4<br />
principal<br />
axis 2<br />
0.6<br />
0.8<br />
1 1<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
0.8<br />
0.6<br />
0.2<br />
0.4<br />
principal<br />
axis 1<br />
Figure 21. Principal components analysis: orientation in <strong>the</strong> first 3D-space <strong>of</strong> <strong>the</strong> vectors<br />
representing <strong>the</strong> 14 measurements. Different colors have been chosen to symbolize <strong>the</strong> 4<br />
groups <strong>of</strong> measurements: for general body size measurements, for <strong>the</strong> integuments and<br />
skeleton, for <strong>the</strong> digestive tract and its content and for <strong>the</strong> gonads.<br />
Principal component analysis (PCA) indicated that all body size<br />
measurements are well correlated with <strong>the</strong> first axis (trend corresponding<br />
0<br />
106
to general growth) that explains 55.0% <strong>of</strong> <strong>the</strong> whole variance (Fig. 21).<br />
Integument and skeleton measurements are also well correlated with <strong>the</strong><br />
first axis. The second axis explains 30.2% <strong>of</strong> <strong>the</strong> total variance. Gonad<br />
measurements are highly correlated with this axis that represents change<br />
with <strong>the</strong> reproductive cycle. Note that body size measurements are not<br />
much correlated with gonads measurements in any axis. Indeed, due to <strong>the</strong><br />
rigidity <strong>of</strong> <strong>the</strong> test, diameter, height and total weights are not affected by<br />
<strong>the</strong> size <strong>of</strong> <strong>the</strong> gonads (when <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> has small gonads, its general<br />
cavity is filled with coelomic fluid with about <strong>the</strong> same density than <strong>the</strong><br />
gonads). This is convenient because body size measurements can be<br />
considered as soma measurements, independently <strong>of</strong> <strong>the</strong> size <strong>of</strong> <strong>the</strong><br />
gonads. The third axis amounts for 12.9% <strong>of</strong> <strong>the</strong> total variance and is<br />
slightly represented by digestive tract measurements. It should be due to its<br />
contents, as a consequence <strong>of</strong> <strong>the</strong> feeding activity during <strong>the</strong> last day<br />
before dissection. 98.2% <strong>of</strong> <strong>the</strong> variance is explained by <strong>the</strong> first three axes<br />
that are kept (Table 6).<br />
Table 6. Principal components analysis: contribution <strong>of</strong> <strong>the</strong> parameters to <strong>the</strong> three first axes.<br />
Parameter Axis 1 Axis 2 Axis 3<br />
Height 0.895 0.311 0.246<br />
Diameter 0.869 0.369 0.269<br />
Weight <strong>of</strong> Aristotle's lantern 0.852 0.365 0.323<br />
Weight <strong>of</strong> spines 0.833 0.450 0.248<br />
Weight <strong>of</strong> test 0.804 0.450 0.347<br />
Immersed weight 0.799 0.510 0.298<br />
Fresh weight <strong>of</strong> integuments 0.789 0.507 0.339<br />
Dry weight <strong>of</strong> integuments 0.769 0.574 0.291<br />
Total fresh weight 0.729 0.552 0.390<br />
Drained weight 0.729 0.575 0.371<br />
Dry weight <strong>of</strong> digestive tract and its content 0.699 0.407 0.567<br />
Fresh weight <strong>of</strong> digestive tract and its content 0.629 0.454 0.626<br />
Dry weight <strong>of</strong> gonads 0.360 0.900 0.231<br />
Fresh weight <strong>of</strong> gonads 0.392 0.891 0.215<br />
We are now confident that a body size measurement, e.g., <strong>the</strong> diameter<br />
<strong>of</strong> <strong>the</strong> test, is an adequate representation <strong>of</strong> <strong>the</strong> growth achieved by all <strong>the</strong><br />
somatic organs. Sea <strong>urchin</strong> appears to be a good experimental subject for<br />
studying growth because size is easy to measure with accuracy thanks to<br />
<strong>the</strong> rigidity <strong>of</strong> its skeleton that constraints its shape; also because it exhibits<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
107
a relative homogeneous growth <strong>of</strong> all its somatic organs. For its <strong>model</strong>, <strong>the</strong><br />
<strong>sea</strong> <strong>urchin</strong> could be considered as a "system with three degrees <strong>of</strong><br />
freedom": (1) <strong>the</strong> body wall and most somatic organs, (2) <strong>the</strong> gonads and<br />
(3) <strong>the</strong> digestive tract and its content. This means we can fully characterize<br />
<strong>the</strong> echinoid by three, easily performed, measurements such as <strong>the</strong> body<br />
size, <strong>the</strong> gonad index (Spirlet et al, 2000, 2001) and <strong>the</strong> repletion index<br />
(Nedelec, 1983). With such three parameters, it should be possible to<br />
calculate <strong>the</strong> size and <strong>the</strong> weight <strong>of</strong> all organs, once <strong>the</strong> corresponding<br />
allometric relationships are established.<br />
Part II: Measurement for size in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
108
PART III<br />
Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
109
110
PART III: EXPERIMENTAL STUDIES OF THE<br />
INTRASPECIFIC COMPETITION<br />
Working on P. lividus, we were puzzled by asymmetry and even<br />
multimodality in size distributions <strong>of</strong> previously homogeneous batches <strong>of</strong><br />
<strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s (see Part I). However, almost no author, except<br />
Himmelman (1986) and Levitan (1988), considered that intraspecific<br />
competition exists among populations <strong>of</strong> aggregative <strong>sea</strong> <strong>urchin</strong>s (see <strong>the</strong><br />
general introduction).<br />
In fact, field observations do not easily allow <strong>the</strong> study <strong>of</strong> intraspecific<br />
competition because one can only describe size distributions <strong>of</strong> wild<br />
populations. Shape <strong>of</strong> size distributions, being bimodal for instance, does<br />
not tell which is <strong>the</strong> cause <strong>of</strong> this bimodality. One can speculate on<br />
possible mechanisms involved without bringing evidences. Huston & De<br />
Angelis identified, among o<strong>the</strong>r possible causes, at least 23 biological<br />
mechanisms that can produce bimodality (1987, see <strong>the</strong>ir Table 1). In such<br />
circumstances, only targeted experiments can bring clues on what really<br />
happens. Our rearing system provided a good basis to start from. So, we<br />
were able to explore a little deeper <strong>the</strong>se puzzling asymmetric size<br />
distributions that appear in batches <strong>of</strong> <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
111
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
112
Experimental study <strong>of</strong> growth in <strong>the</strong> echinoid <strong>Paracentrotus</strong><br />
lividus (Lamarck, 1816) (Echinodermata).<br />
a. Abstract<br />
b. Introduction<br />
Ph. Grosjean, Ch. Spirlet & M. Jangoux, 1996. Journal <strong>of</strong><br />
Experimental Marine Biology and Ecology, 201:173-184.<br />
Multimodal size frequency distribution (that is, a few individuals<br />
growing very fast and a few individuals growing very slowly) among an<br />
originally homogeneous cohort <strong>of</strong> juveniles <strong>Paracentrotus</strong> lividus is<br />
observed in <strong>reared</strong> conditions when <strong>the</strong>y are 6 to 24 months old. The<br />
splitting <strong>of</strong> this cohort into homogeneous size-classed subgroups results in<br />
an increased growth <strong>of</strong> <strong>the</strong> smaller animals that catch up with <strong>the</strong> bigger<br />
ones in 4 months time. This indicates that <strong>the</strong> smaller animals are not<br />
genetically less productive and suggests <strong>the</strong>y were inhibited in <strong>the</strong>ir<br />
growth due to <strong>the</strong> presence <strong>of</strong> larger ones. Supposing such growth<br />
inhibition also occurs in <strong>the</strong> natural environment, <strong>the</strong> observed mechanism<br />
could be very efficient in stabilizing field populations <strong>of</strong> aggregative<br />
echinoid species by maintaining a protected pool <strong>of</strong> small individuals with<br />
high growth potential but inhibited by <strong>the</strong> density <strong>of</strong> larger ones.<br />
Keywords: Echinoids, growth, population dynamics, size frequency<br />
distribution.<br />
Echinoid surveys in <strong>the</strong> field are <strong>of</strong>ten based on size frequency<br />
distribution studies which <strong>the</strong>oretically allow <strong>the</strong> separation <strong>of</strong> different<br />
cohorts (Ebert, 1973; Kenner, 1992; Guillou & Michel, 1993). When<br />
proceeding so, authors necessarily assume that <strong>the</strong> size frequency<br />
distribution in a single cohort is normal or at least unimodal (Ebert, 1981;<br />
Ebert et al, 1993; Botsford et al, 1994). When animals can be aged, <strong>the</strong><br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
113
assumption <strong>of</strong> normality can be tested because <strong>the</strong> different cohorts are<br />
unambiguously separated. Since size is not a reliable indication <strong>of</strong> <strong>the</strong> age<br />
<strong>of</strong> echinoid (Ebert, 1967; Levitan, 1988) and since <strong>the</strong> possibility to age<br />
<strong>the</strong>m with growth lines remains disputed (Ebert, 1986; Gage, 1992; Ebert<br />
& Russell, 1993; Gebauer & Moreno, 1995), <strong>the</strong> interpretations <strong>of</strong> size<br />
frequency distributions among natural populations <strong>of</strong> echinoids remain<br />
ra<strong>the</strong>r speculative unless <strong>the</strong> assumption <strong>of</strong> normality can be verified. The<br />
difficulty may be bypassed by rearing animals under controlled conditions<br />
without <strong>the</strong> pressure <strong>of</strong> predators which should allow a good follow-up <strong>of</strong><br />
<strong>the</strong>ir growth and enable <strong>the</strong> observation <strong>of</strong> a cohort’s distribution. Indeed,<br />
parameters like recruitment, mortality and age <strong>of</strong> <strong>the</strong> individuals can be<br />
known precisely. Small-scale cultivation <strong>of</strong> echinoids under artificial<br />
conditions has been successfully performed for several years (e.g.,<br />
Hinegardner, 1969; Fridberger et al, 1979; Le Gall, 1990) and among <strong>the</strong><br />
various aspects tackled by different authors, growth seems to be a<br />
privileged one (Ebert, 1975; Fridberger et al, 1979; Frantzis & Grémare,<br />
1992). Yet, data remain ra<strong>the</strong>r limited and little interpretation <strong>of</strong> <strong>the</strong> size<br />
distribution itself has been performed. Cellario and Fenaux (1990)<br />
observed that <strong>the</strong>re was a "wide size dispersion pattern with age progress"<br />
in P. lividus rearing. They also observed that <strong>the</strong> relative spreading <strong>of</strong> size<br />
(standard deviation <strong>of</strong> test diameters / average test diameter ratio)<br />
increases rapidly in early postmetamorphic life, reaches a constant level at<br />
2-3 mm <strong>of</strong> mean test diameter (<strong>the</strong> distribution is <strong>the</strong>n at its widest) and<br />
begins to decrease when <strong>the</strong> animals become larger than 10 mm in<br />
diameter. As far as we know, no study treated more precisely <strong>the</strong> size<br />
distribution shape <strong>of</strong> a cohort <strong>of</strong> <strong>reared</strong> echinoids.<br />
The present paper focuses on how size distribution <strong>of</strong> <strong>reared</strong><br />
<strong>Paracentrotus</strong> lividus changes with time. It aims at testing if this<br />
distribution is normal and attempts to determine <strong>the</strong> factors that lead to its<br />
spreading.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
114
c. Material and methods<br />
All <strong>the</strong> echinoids used in this work were produced in laboratory. They<br />
were cultivated with a method adapted from Le Gall (1990). The original<br />
strain comes from <strong>the</strong> rocky shore <strong>of</strong> Morgat (Brittany, France).<br />
Rearing procedure<br />
Spawning was induced by injecting KCl 0.5 M in <strong>the</strong> body cavity <strong>of</strong><br />
adult individuals. Eggs <strong>of</strong> one female were transferred in a small plastic jar<br />
containing 800 ml <strong>of</strong> <strong>sea</strong>water. A quantity <strong>of</strong> sperm equivalent to 0.5 ml <strong>of</strong><br />
milt was added to <strong>the</strong> eggs. The fertilization was controlled after 4 h and<br />
<strong>the</strong> number <strong>of</strong> fertilized eggs was evaluated. The embryos (in <strong>the</strong> gastrula<br />
stage) were <strong>the</strong>n transferred in a 200-l larval rearing tank to a<br />
concentration <strong>of</strong> 250 embryos per liter. Larvae were fed daily with<br />
Pheodactylum tricornutum from <strong>the</strong> third day on. The water remained<br />
unchanged for <strong>the</strong> whole larval period.<br />
About twenty days later, competent larvae were transferred in clean<br />
sieves with a 250-µm mesh. Sieves with larvae were placed in toboggans<br />
(see Le Gall, 1990) with 10 cm depth <strong>of</strong> recirculating water providing a<br />
gentle uniform current around <strong>the</strong>m. Metamorphosis was induced by<br />
introducing coralline algae in <strong>the</strong> sieves. Larval fixation and<br />
metamorphosis took less than 24 hours.<br />
The day before juveniles become exotrophic, 5 g (fresh weight) <strong>of</strong><br />
green alga Enteromorpha linza per 100 cm 2 sieve surface were distributed.<br />
From this moment on, <strong>the</strong> same food quantity was given every 15 days.<br />
The treatment remained identical during <strong>the</strong> first year, except that <strong>the</strong> sieve<br />
diameter and mesh size were progressively increased according to <strong>the</strong><br />
growing diameter <strong>of</strong> <strong>the</strong> individuals. After one year, when <strong>the</strong> smallest<br />
echinoids reached more than 5 mm in test diameter, all <strong>the</strong> individuals<br />
were put in a basket with a 5 mm mesh and transferred in ano<strong>the</strong>r<br />
toboggan where stronger water current and higher echinoid biomass<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
115
occurred. From this moment on, and twice a week, individuals were fed ad<br />
libitum with fresh kelp Laminaria digitata.<br />
The entire rearing was carried out in natural dim light and to a constant<br />
temperature <strong>of</strong> 18 ± 2°C all year long. The water was renewed<br />
continuously at a rate <strong>of</strong> 200% to 300% <strong>of</strong> <strong>the</strong> total volume per day with<br />
fresh natural <strong>sea</strong>water allowed to settle for at least 30 h beforehand.<br />
Size frequency distribution just after <strong>the</strong> metamorphosis<br />
All <strong>the</strong> larvae issued from a single fertilization (Fa) and <strong>reared</strong> as<br />
described above in <strong>the</strong> same tank were induced to metamorphosis <strong>the</strong> same<br />
day. Postmetamorphics were not fed. They were fixed (3% glutaraldehyde)<br />
and photographed 7 days after metamorphosis. Pictures were transferred<br />
on a Kodak photoCD and <strong>the</strong> test diameter <strong>of</strong> every individual computed<br />
by image analysis s<strong>of</strong>tware. Actual size was determined by using a<br />
graduated background. The frequencies <strong>of</strong> observed sizes were tested<br />
against a normal and a log-normal distribution with a Kolmogorov-<br />
Smirnov test adapted by Lilliefors for intrinsic comparison (Sokal &<br />
Rohlf, 1981).<br />
Follow up <strong>of</strong> a single <strong>reared</strong> strain <strong>of</strong> juveniles during 30<br />
months<br />
A whole strain issued from a single fertilization (Fb) was cultivated<br />
over 30 months. The test diameter <strong>of</strong> each individual was measured every<br />
6 months with a sliding caliper. The first set <strong>of</strong> measurements was<br />
performed when echinoids were 6 months old (no measurements were<br />
done just after metamorphosis because <strong>of</strong> <strong>the</strong> extreme fragility <strong>of</strong> early<br />
postmetamorphics). The total number <strong>of</strong> individuals was 536 at 6 months<br />
old, and dropped progressively to 280 at 30 months old. The mean<br />
mortality was thus 48% on a 2-year period.<br />
Size frequency data obtained were <strong>the</strong>n tested against a normal and a<br />
log-normal distribution. Possible multimodality was checked by <strong>the</strong><br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
116
graphical method <strong>of</strong> Bhattacharya (1967). Since <strong>the</strong> number <strong>of</strong> individuals<br />
is low, data need to be smoo<strong>the</strong>d before applying <strong>the</strong> method:<br />
fs = f + f + f<br />
(20)<br />
1 1 1<br />
i 4 ( i− 1) 2 i 4 ( i+<br />
1)<br />
where: fi = frequency observed in <strong>the</strong> size class i,<br />
fsii = smoo<strong>the</strong>d frequency for <strong>the</strong> class i.<br />
The minimal number <strong>of</strong> modes that match <strong>the</strong> observed size frequency<br />
distribution (i.e: when a χ 2 test gives a probability higher than 0.05) was<br />
determined by <strong>the</strong> technique <strong>of</strong> maximum-likelihood estimator using<br />
NORMSEP (Hasselblad, 1966; modified by McDonald & Pitcher, 1979).<br />
Effect <strong>of</strong> size sorting on <strong>the</strong> growth <strong>of</strong> juveniles and<br />
interactions among <strong>the</strong>m<br />
Three additional fertilizations (Fc, Fd and Fe) were done at different<br />
times with parents not genetically related. Produced individuals were<br />
sorted twice before <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> experiment so as to have several<br />
homogeneous batches in terms <strong>of</strong> size. The experiment started with<br />
populations <strong>of</strong> Fc, Fd and Fe being respectively 4, 6 and 8 months old.<br />
Four batches (Fc1 to Fc4) and six batches (Fd1 to Fd6) <strong>of</strong> 50 size-sorted<br />
echinoids were set up for Fc and Fd respectively. Mean-sized individuals<br />
<strong>of</strong> Fe were separated into nine batches <strong>of</strong> 20 echinoids (Fe1 to Fe8 plus a<br />
control batch randomly chosen).<br />
Each Fc to Fe batch was cultivated in a sieve <strong>of</strong> 20 x 15 cm <strong>of</strong> surface<br />
with a 2-mm mesh except <strong>the</strong> control batch <strong>of</strong> Fe whose 20 individuals<br />
were <strong>reared</strong> separately in 20 different sieves. Individuals were fed ad<br />
libitum with Enteromorpha linza exclusively. The <strong>sea</strong>weeds covered <strong>the</strong><br />
entire surface <strong>of</strong> <strong>the</strong> sieve so <strong>the</strong>re was no competition for food.<br />
Experiments lasted for 4 months and <strong>the</strong> final test diameter <strong>of</strong> all echinoids<br />
from each batch was measured with a sliding caliper.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
117
d. Results<br />
Size distribution among early postmetamorphics and change<br />
through time<br />
Statistics on <strong>the</strong> distribution <strong>of</strong> <strong>the</strong> test diameters among early<br />
postmetamorphics (Fa) is presented in Table 7. Temporal evolution in <strong>the</strong><br />
size frequency <strong>of</strong> all <strong>the</strong> individuals <strong>of</strong> <strong>the</strong> cohort followed during 30<br />
months (Fb) is presented in Table 7 and Fig. 22. The test diameter <strong>of</strong> early<br />
postmetamorphics distribute along a normal curve characterized by a mean<br />
<strong>of</strong> 497 µm and a standard deviation <strong>of</strong> 56 µm (Kolmogorov-Smirnov /<br />
Lilliefors test, p > 0.05).<br />
At 6 months old, distribution is nei<strong>the</strong>r normal, nor log-normal<br />
(p < 0.001, Kolmogorov-Smirnov / Lilliefors test); multimodality appears,<br />
although not clearly yet. Two modes can be identified by both graphical<br />
and numerical analysis. However, this kind <strong>of</strong> graph might also be<br />
obtained with a unimodal distribution very skewed to <strong>the</strong> right (skewness =<br />
0.748).<br />
After 12 months, <strong>the</strong> distribution does not match a normal nor lognormal<br />
one. This time, a "head" portion is clearly distinguishable. It<br />
concerns 18 individuals, that is 5.1% <strong>of</strong> <strong>the</strong> total. At least two o<strong>the</strong>r classes<br />
could be separated although not clearly isolated from each o<strong>the</strong>r. The<br />
distribution is widely spread. The ratio between <strong>the</strong> 10% larger and <strong>the</strong><br />
10% smaller is 3.2 in test diameter and 30.9 in wet weight.<br />
After 18 months, <strong>the</strong> general shape remains <strong>the</strong> same with a head<br />
clearly detached, except that <strong>the</strong> two o<strong>the</strong>r classes seems to have merged.<br />
The head is represented by 26 animals (8.7% <strong>of</strong> <strong>the</strong> total number) which<br />
means <strong>the</strong>re are newcomers. The distribution is always widely spread. The<br />
mean size <strong>of</strong> <strong>the</strong> heading class is 39.6 mm, which is already more than <strong>the</strong><br />
mean size (36.9 mm) <strong>of</strong> all <strong>the</strong> individuals one year later, when <strong>the</strong>y will<br />
be 30 months old.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
118
Table 7. Statistical analysis <strong>of</strong> <strong>the</strong> size frequency distribution <strong>of</strong> single cohorts <strong>of</strong> P. lividus at<br />
different ages.<br />
Age (months) 1 6 12 18 24 30<br />
Fertilization Fa Fb Fb Fb Fb Fb<br />
treatment Killed 7d. after<br />
metamorphosis<br />
Same cohort <strong>of</strong> <strong>reared</strong> echinoids followed in controlled conditions<br />
General descriptive statistics on size frequencies (individual horizontal test diameter in mm)<br />
Number <strong>of</strong> individuals 296 536 361 296 285 280<br />
Median 0.501 4 15 23 31 36<br />
Mean 0.497 4.81 17.1 24.6 32.3 36.9<br />
Standard deviation 0.056 2.28 5.51 6.98 6.12 5.56<br />
Skewness -0.275 0.748 0.851 0.599 0.136 0.042<br />
Kurtosis 0.226 -0.044 0.373 0.094 -0.024 0.139<br />
Intrinsic goodness <strong>of</strong> fit test to a normal curve (Kolmogorov-Smirnov / Lilliefors)<br />
Maximum difference 0.046 0.172 0.125 0.097 0.077 0.064<br />
Probability 0.127 **<br />
0.000 0.000 0.000 0.000 0.011 *<br />
Intrinsic goodness <strong>of</strong> fit to a log-normal curve (Kolmogorov-Smirnov / Lilliefors)<br />
Maximum difference 0.068 0.130 0.075 0.064 0.059 0.095<br />
Probability 0.002 0.000 0.000 0.005 0.018 *<br />
0.000<br />
Components analysis by <strong>the</strong> graphical Bhattacharya's method<br />
Number <strong>of</strong> modes 1 2 3 or 4 2 3 or 4 1 or 2<br />
Groups clearly<br />
separated<br />
- - 2 2 2 -<br />
Characterization <strong>of</strong> <strong>the</strong> groups (maximum-likelihood estimator method, NORMSEP)<br />
(minimal number <strong>of</strong> groups giving a χ 2 probability > 0.05)<br />
Number <strong>of</strong> groups - 2 3 2 3 1<br />
χ 2 value - 8.41 14.50 21.44 24.48 26.71<br />
Probability - 0.077 **<br />
0.488 **<br />
0.554 **<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
0.140 **<br />
0.181 **<br />
Mean <strong>of</strong> group 1 3.43 13.3 23.1 18.4 36.9<br />
SD group 1 1.15 2.49 5.39 1.82 5.56<br />
Percentage in 1 58.2 53.2 91.3 2.3 100<br />
No. <strong>of</strong> ind. in 1 312 192 269 7 280<br />
Mean <strong>of</strong> group 2 6.73 20.2 39.6 27.7<br />
SD group 2 2.07 3.61 1.97 2.54<br />
Percentage in 2 41.8 41.7 8.7 34.4<br />
No. <strong>of</strong> ind. in 2 224 151 26 98<br />
Mean <strong>of</strong> group 3 31.2 35.3<br />
SD group 3 1.34 5.00<br />
Percentage in 3 5.1 63.3<br />
No. <strong>of</strong> ind. in 3 18 179<br />
* Fitting is significant (p > 0.01); ** Fitting is very significant (p > 0.05).<br />
After 24 months, <strong>the</strong> individuals forming <strong>the</strong> head seem to be caught<br />
up by <strong>the</strong> mean sized ones. However, some individuals do not follow this<br />
general movement, and a tail remains amounting to 2.7% <strong>of</strong> <strong>the</strong> total<br />
population. The shape <strong>of</strong> <strong>the</strong> distribution is completely changed: skewness<br />
decreases (0.136) and <strong>the</strong> whole distribution matches possibly a lognormal<br />
one (Kolmogorov-Smirnov / Lilliefors, p > 0.01).<br />
119
Nber<br />
<strong>of</strong><br />
ind.<br />
120<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
0 5 10 15 20 25 30 35 40 45 50 55<br />
Test diameter (mm)<br />
6 months old<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
18 months old<br />
12 months old<br />
30 months old<br />
24 months old<br />
Figure 22. Evolution <strong>of</strong> a single cohort <strong>of</strong> P. lividus (Fb) <strong>reared</strong> in stable environmental<br />
conditions according to time.<br />
After 30 months, <strong>the</strong> latecomers forming <strong>the</strong> tail catch up <strong>the</strong> mean<br />
sized ones, and <strong>the</strong> distribution loses its significant multimodal shape. The<br />
shape becomes symmetrical: low skewness <strong>of</strong> 0.042 and approach a<br />
normal curve (0.01 < p < 0.05, Kolmogorov-Smirnov / Lilliefors). The<br />
spreading decreases also: <strong>the</strong> ratio 10% larger / 10% smaller drops to 1.8<br />
in size and to 5.3 in wet weight.<br />
Effect <strong>of</strong> size sorting on juveniles’ growth<br />
Figure 23 and Table 8 show <strong>the</strong> initial and final size distributions <strong>of</strong> <strong>the</strong><br />
four Fc size-sorted batches. A Kolmogorov-Smirnov / Lilliefors test<br />
applied on each distribution showed that 2 cases out <strong>of</strong> <strong>the</strong> 4 did not match<br />
a Gaussian curve (p < 0.01). Since extremes were eliminated before <strong>the</strong><br />
experiment began (only mean-sized individuals were used), non normality<br />
in <strong>the</strong> distribution <strong>of</strong> <strong>the</strong> whole cohort is not due to <strong>the</strong> spreading <strong>of</strong> <strong>the</strong>se<br />
extremes, producing "head" and "tail" classes, but could be <strong>the</strong><br />
120
consequence <strong>of</strong> differential growth rates among all individuals, including<br />
mean-sized ones.<br />
Table 8. Statistics on <strong>the</strong> four batches <strong>of</strong> Fc in <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> experiment (initial) and<br />
after 4 months (final).<br />
Batch Fc1 Fc2 Fc3 Fc4<br />
Treatment Size sorted batches <strong>of</strong> mean-sized individuals only (no "head" already<br />
differentiated)<br />
Initial mean size (mm) 6.0 6.0 6.0 6.0<br />
Final mean size (mm) 14.3 15.1 14.0 14.9<br />
Increase in size (mm) 8.3 9.1 8.0 8.9<br />
Kolmogorov-Smirnov / Lilliefors on final sizes (mm)<br />
Maximum difference 0.161 0.148 0.106 0.124<br />
Probability 0.003 **<br />
0.008 **<br />
0.175 0.061<br />
**<br />
Does not fit a normal curve (p < 0.01)<br />
A<br />
Nber<br />
<strong>of</strong><br />
ind.<br />
B<br />
Nber<br />
<strong>of</strong><br />
ind.<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
5 7 9 11 13 15 17<br />
Test diameter (mm)<br />
5 7 9 11 13 15 17<br />
Test diameter (mm)<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
Fc4<br />
Fc3<br />
Batch<br />
Fc2<br />
Fc1<br />
Fc4<br />
Fc3<br />
Batch<br />
Fc2<br />
Figure 23. Size distribution <strong>of</strong> Fc juveniles in each batch in <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> experiment<br />
(A) (extreme individuals have been eliminated) and 4 months later (B).<br />
Fc1<br />
121
Table 9. Statistics on <strong>the</strong> six batches <strong>of</strong> Fd in <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> experiment (initial) and<br />
after 4 months (final).<br />
Batch Fd1 Fd2 Fd3 Fd4 Fd5 Fd6<br />
Treatment Size sorted batches ranging from "head" to mean-sized individuals<br />
<strong>of</strong> a fertilization that already differentiated a heading group.<br />
Initial mean size (mm) 10.5 9.5 8.5 7.0 6.0 6.0<br />
Final mean size (mm) 14.0 14.0 13.2 12.7 13.5 13.6<br />
Increase in size (mm) 3.5 4.5 4.7 5.7 7.5 7.6<br />
Comparison <strong>of</strong> final mean sizes <strong>of</strong> <strong>the</strong> six batches (Kruskal-Wallis)<br />
Statistic value 11.85 Associated probability 0.037 *<br />
* Difference slightly significant (0.01 < p < 0.05).<br />
A<br />
Nber<br />
<strong>of</strong><br />
ind.<br />
B<br />
Nber<br />
<strong>of</strong><br />
ind.<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
5 7 9 11 13 15 17<br />
Test diameter (mm)<br />
5 7 9 11 13 15 17<br />
Test diameter (mm)<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
Fd4<br />
Fd3<br />
Batch<br />
Fd2<br />
Fd1<br />
Fd1<br />
Fd6<br />
Fd5<br />
Fd6<br />
Fd5<br />
Fd4<br />
Fd3<br />
Batch<br />
Figure 24. Size distribution <strong>of</strong> Fd individuals in each batch (ranging from "head" individuals,<br />
Fd1, to mean-sized ones, Fd5 and Fd6) at <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> experiment (A) and 4 months<br />
later (B).<br />
Figure 24 shows <strong>the</strong> size distributions <strong>of</strong> <strong>the</strong> Fd batches at <strong>the</strong><br />
beginning <strong>of</strong> <strong>the</strong> experiment (Fig. 24A) and after 4 months <strong>of</strong> controlled<br />
Fd2<br />
122
earing (Fig. 24B). Table 9 presents <strong>the</strong> statistics on <strong>the</strong>se data. The<br />
difference between <strong>the</strong> Fd batches after 4 months is only <strong>of</strong> slight<br />
significance (Kruskal-Wallis, 0.01 < p < 0.05) although each batch <strong>of</strong> Fd<br />
was made up initially <strong>of</strong> individuals from a different size class, going from<br />
10.5 mm in Fd1 ("head" individuals) to 6 mm in Fd5 and Fd6 (mean-sized<br />
individuals). Thus, it seems that size sorting eliminates <strong>the</strong> factor(s) that<br />
maintained <strong>the</strong> difference in sizes between <strong>the</strong> batches so that smaller<br />
individuals catch up rapidly to <strong>the</strong> larger ones.<br />
Interactions between individuals<br />
Table 10 presents <strong>the</strong> size frequency distributions <strong>of</strong> Fe batches<br />
obtained after rearing <strong>the</strong>m 4 months. The control shows <strong>the</strong> distribution <strong>of</strong><br />
<strong>the</strong> sizes that can be expected without interactions between <strong>the</strong> individuals<br />
because each echinoid was cultivated independently. This distribution<br />
matches very well a normal curve (intrinsic Kolmogorov-<br />
Smirnov / Lilliefors test, p >> 0.05) characterized by a mean <strong>of</strong> 18.5 mm<br />
(test diameter) and a standard deviation <strong>of</strong> 1.52 mm.<br />
The distribution <strong>of</strong> <strong>the</strong> o<strong>the</strong>r eight experimental Fe batches is compared<br />
to <strong>the</strong> <strong>the</strong>oretical normal curve fitted on <strong>the</strong> control with its estimated<br />
mean and standard deviation. In <strong>the</strong> absence <strong>of</strong> interactions, no more than<br />
5% <strong>of</strong> <strong>the</strong> individuals should be smaller than 15.5 mm or greater than 21.5<br />
mm (considered as potential outliers, grayed zones in Table 10). The<br />
number <strong>of</strong> small "outliers" is much more important than this prediction in<br />
all eight batches, ranging from 11% to 35% <strong>of</strong> <strong>the</strong> total which indicates a<br />
consistent tendency: some <strong>of</strong> <strong>the</strong> individuals <strong>reared</strong> toge<strong>the</strong>r were<br />
apparently inhibited in <strong>the</strong>ir growth. Moreover, whole batches Fe4, Fe6<br />
and Fe7 do not match <strong>the</strong> control’s distribution (p < 0.05, extrinsic<br />
Kolmogorov-Smirnov test). Consequent to this inhibition, <strong>the</strong> mean sizes<br />
in <strong>the</strong> experimental batches are systematically lower than <strong>the</strong> mean size <strong>of</strong><br />
<strong>the</strong> control.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
123
Table 10. Size distribution <strong>of</strong> Fe echinoids () after having been <strong>reared</strong> individually (control)<br />
or toge<strong>the</strong>r (experimental batches) for 4 months.<br />
class size (mm) control experimental batches<br />
min ≤ ≤ ≤ ∅ ∅ ∅ < max batch Fe1 Fe2 Fe3 Fe4 Fe5 Fe6 Fe7 Fe8<br />
11 - 11.5<br />
11.5 - 12<br />
<br />
12 - 12.5<br />
12.5 - 13<br />
<br />
13 - 13.5 <br />
13.5 - 14 <br />
14 - 14.5 <br />
14.5 - 15 p = 5% <br />
15 - 15.5 <br />
15.5 - 16<br />
16 - 16.5<br />
<br />
CD<br />
<br />
<br />
<br />
16.5 - 17 <br />
17 - 17.5 <br />
17.5 - 18 <br />
18 - 18.5 <br />
18.5 - 19 <br />
19 - 19.5 <br />
19.5 - 20 <br />
20 - 20.5 <br />
20.5 - 21 <br />
21 - 21.5 <br />
21.5 - 22 <br />
22 - 22.5<br />
22.5 - 23<br />
23 - 23.5<br />
p = 5% <br />
23.5 - 24 <br />
mean size 18.5 17.5 18.3 17.8 18.1 18.2 17.2 17.5 18.0<br />
standard dev. 1.52 2.28 2.52 2.37 2.49 3.07 2.05 2.38 2.23<br />
Lilliefors test <strong>of</strong> normality Kolmogorov-Smirnov extrinsic test and probability to fit <strong>the</strong> control curve<br />
max. dif. (K-S) 0.128 0.283 0.180 0.257 0.316 0.172 0.389 0.348 0.246<br />
proba. to fit 0.543 **<br />
0.066 0.614 0.177 0.042 *<br />
0.637 0.003 **<br />
0.015 *<br />
0.170<br />
Small "outliers" (5% expected) 30% 20% 24% 17% 18% 35% 11% 28%<br />
Highlighted cells are <strong>the</strong> mean classes <strong>of</strong> each group. Grayed zones show 'outliers' (i.e., large and small sizes with p < 0.05<br />
according to <strong>the</strong> control's fitted distribution CD). * Test significant; ** test highly significant.<br />
e. Discussion<br />
<strong>Growth</strong> <strong>of</strong> <strong>the</strong> regular echinoid <strong>Paracentrotus</strong> lividus was<br />
experimentally studied using homogeneous cohorts kept in controlled<br />
conditions. This homogeneity was obtained by using <strong>reared</strong> individuals<br />
from <strong>the</strong> same parental origin, induced to metamorphosis <strong>the</strong> same day, fed<br />
ad libitum with <strong>the</strong> same food and kept in <strong>the</strong> same environment. Size<br />
frequencies observed among such a cohort distribute along a unimodal<br />
normal curve just after metamorphosis but present a multimodal shape<br />
with at least two, sometimes three, distinct subgroups (i.e.: "head", meansized<br />
and "tail" individuals) when individuals were 6 to 24 months old.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
124
Multimodality disappeared around 30 months <strong>of</strong> age and size frequency<br />
distribution recovered a near-normal shape.<br />
Such differences in growth between juveniles leading to non-normality<br />
in <strong>the</strong> size frequency distribution could probably be attributed to every<br />
individual, not only to extreme ones, as shown with mean-sized<br />
individuals <strong>of</strong> Fc batches. Moreover, <strong>the</strong> difference in speed <strong>of</strong> growth<br />
between heading and mean-sized echinoids cannot be attributed to <strong>the</strong>ir<br />
respective genetic potentials, for size sorting eliminates rapidly <strong>the</strong><br />
differences in size between <strong>the</strong>se subgroups (Fd). Hence, this difference is<br />
<strong>the</strong> consequence <strong>of</strong> an intraspecific competition between echinoids having<br />
different sizes. Fur<strong>the</strong>rmore, evidence <strong>of</strong> inhibition in <strong>the</strong> growth <strong>of</strong><br />
smaller individuals is provided by <strong>the</strong> comparison <strong>of</strong> <strong>the</strong> size distribution<br />
<strong>of</strong> echinoids <strong>reared</strong> toge<strong>the</strong>r (Fe1 to Fe8) or individually (Fe, control).<br />
We observed that when echinoids <strong>of</strong> different sizes are <strong>reared</strong> toge<strong>the</strong>r,<br />
<strong>the</strong> smaller ones tend to insert <strong>the</strong>mselves between <strong>the</strong> larger ones along<br />
<strong>the</strong> walls and corners <strong>of</strong> <strong>the</strong> rearing baskets. In those aggregates, <strong>the</strong> water<br />
is relatively stagnant and <strong>the</strong> pH <strong>the</strong>re is lower than <strong>the</strong> one measured in<br />
<strong>the</strong> running water <strong>of</strong> <strong>the</strong> tanks due to CO2 accumulation (pH NBS <strong>of</strong> 7.1-<br />
7.7 and 7.8-8.0 respectively). Poor water quality could <strong>the</strong>n contribute to<br />
<strong>the</strong> slower growth rate <strong>of</strong> smaller juveniles. The difference in sizes<br />
between large "inhibitor" and small "inhibited" individuals does not need<br />
to be very important to engage <strong>the</strong> process <strong>of</strong> differential growth. Indeed,<br />
<strong>the</strong> spreading in sizes that occurs from originally homogeneous batches<br />
seems to be sufficient to trigger this intraspecific competition, as observed<br />
in <strong>the</strong> eight experimental Fe batches <strong>reared</strong> toge<strong>the</strong>r.<br />
Whe<strong>the</strong>r a multimodal size distribution inside a single cohort <strong>of</strong><br />
juvenile echinoids could be possible in <strong>the</strong> field is worth questioning. If it<br />
is <strong>the</strong> case, <strong>the</strong>n all studies using analysis <strong>of</strong> size frequency distributions<br />
and based on <strong>the</strong> separation <strong>of</strong> presumed unimodal cohorts from a whole<br />
population could be biased. Aggregative behavior is currently observed<br />
among various species <strong>of</strong> echinoids (Ebert, 1977: Strongylocentrotus<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
125
purpuratus; Dafni & Tobol, 1987: Tripneustes gratilla elatensis; Levitan<br />
& Genovese, 1989: Diadema antillarum). In those cases, small juveniles<br />
are <strong>of</strong>ten found under larger conspecifics or between <strong>the</strong>ir spines where<br />
<strong>the</strong>ir rate <strong>of</strong> survival has proven to be higher thanks to <strong>the</strong> protection<br />
provided against predators (Tegner & Levin, 1983; Levitan & Genovese,<br />
1989). Yet, <strong>the</strong> chemical conditions in this environment could vary a lot, as<br />
observed in our rearing devices, and growth could be greatly reduced for<br />
some individuals, causing <strong>the</strong> spreading <strong>of</strong> <strong>the</strong> size distribution <strong>of</strong> <strong>the</strong><br />
cohort or transforming it into a multimodal one. The extension <strong>of</strong> <strong>the</strong><br />
critical period when <strong>the</strong> echinoid is small and thus vulnerable to predators<br />
limits somewhat <strong>the</strong> benefits gained by <strong>the</strong> protection provided by <strong>the</strong><br />
adult’s spine canopy. However, if <strong>the</strong> adults’ density decreases, <strong>the</strong><br />
inhibition <strong>of</strong> a juvenile’s growth is <strong>the</strong>n removed. Some <strong>of</strong> <strong>the</strong>m can grow<br />
very fast if food is available (as "head" ones in <strong>the</strong> currently studied<br />
cohort) and rapidly replace missing adults. This mechanism could be very<br />
efficient in stabilizing field populations <strong>of</strong> aggregative species <strong>of</strong> echinoids<br />
by maintaining a protected pool <strong>of</strong> small individuals with high growth<br />
potential but inhibited by <strong>the</strong> density <strong>of</strong> larger ones.<br />
f. Acknowledgements<br />
This study was conducted in <strong>the</strong> framework <strong>of</strong> an EEC contract in <strong>the</strong><br />
FAR aquaculture program (ref. AQ2.530). We thank Didier Bucaille for<br />
his help in <strong>the</strong> laboratory work. We also thank <strong>the</strong> CREC and <strong>the</strong><br />
University <strong>of</strong> Caen for <strong>the</strong>ir contribution in building a specific echinoid<br />
rearing facility. The "Centre Interuniversitaire de Biologie Marine" also<br />
contributed to this study.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
126
Intraspecific competition: an additional experiment<br />
To investigate intraspecific competition in batches with a large initial<br />
dispersion <strong>of</strong> sizes, two series <strong>of</strong> 6 replicates <strong>of</strong> 60 'small' <strong>sea</strong> <strong>urchin</strong>s (7.5<br />
to 9 mm test diameter; 720 individuals in <strong>the</strong> total) coming from <strong>the</strong><br />
heading group <strong>of</strong> a single fertilization (Ff), and two series <strong>of</strong> 6 replicates <strong>of</strong><br />
10 'large' <strong>sea</strong> <strong>urchin</strong>s (20 to 24 mm test diameter; 120 individuals in <strong>the</strong><br />
total) extracted also from <strong>the</strong> heading group <strong>of</strong> ano<strong>the</strong>r single fertilization<br />
(Fg) were set up at <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> experiment (Fig. 25). Small <strong>sea</strong><br />
<strong>urchin</strong>s <strong>of</strong> Ff were 5 months old at <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> experiment while<br />
those <strong>of</strong> Fg were 13 months old. One series <strong>of</strong> small and one series <strong>of</strong><br />
large individuals were <strong>reared</strong> separately. The two remaining series <strong>of</strong> small<br />
and large individuals were <strong>reared</strong> toge<strong>the</strong>r ('mixed' series). The surface <strong>of</strong><br />
<strong>the</strong> rearing baskets used was <strong>the</strong> same for all batches (20 x 30 cm).<br />
Figure 26 presents <strong>the</strong> change <strong>of</strong> size distributions in <strong>the</strong> three<br />
experimental series. Table 11 presents corresponding statistics. The six<br />
replicates from <strong>the</strong> same series are pooled after verifying that <strong>the</strong><br />
differences between <strong>the</strong>m are not significant (Kruskal-Wallis test, p >><br />
0.05, except for <strong>the</strong> small individuals in <strong>the</strong> mixed batch at 2 months were<br />
0.01 < p < 0.05, see Table 11). The presence or absence <strong>of</strong> smaller <strong>sea</strong><br />
<strong>urchin</strong>s did not influence very significantly <strong>the</strong> growth <strong>of</strong> larger<br />
individuals. The latter were thus not inhibited whatsoever. It is, however,<br />
interesting to note that size distribution <strong>of</strong> larger echinoids did not spread<br />
much with time, which means <strong>the</strong>ir density was low enough to avoid<br />
competition among <strong>the</strong>m and allowed a homogeneous growth <strong>of</strong> <strong>the</strong> whole<br />
set. On <strong>the</strong> o<strong>the</strong>r hand, growth <strong>of</strong> smaller <strong>sea</strong> <strong>urchin</strong>s was strongly affected<br />
by <strong>the</strong> presence <strong>of</strong> larger individuals: <strong>the</strong>ir growth was limited and, as a<br />
consequence, <strong>the</strong>y remained more grouped than batches <strong>of</strong> small<br />
individuals alone. In batches where small <strong>sea</strong> <strong>urchin</strong>s occurred alone,<br />
competition took place and size distribution spread. This is an effect<br />
already identified in <strong>the</strong> previous experiments.<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
127
Nbr. <strong>of</strong> ind.<br />
Nbr. <strong>of</strong> ind.<br />
A. Size distribution <strong>of</strong> <strong>the</strong> "small" <strong>urchin</strong>s just before <strong>the</strong><br />
experiment starts (fertilization Ff, 5 months old)<br />
1000<br />
900<br />
800<br />
700<br />
600<br />
500<br />
400<br />
300<br />
200<br />
100<br />
0<br />
0<br />
0<br />
3<br />
1.5<br />
6<br />
3<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
4.5<br />
6<br />
7.5<br />
Diameter (mm)<br />
B. Size distribution <strong>of</strong> <strong>the</strong> "large" <strong>urchin</strong>s just before <strong>the</strong><br />
experiment starts (fertilization Fg, 13 months old)<br />
45<br />
40<br />
35<br />
30<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
9<br />
12<br />
15<br />
Diameter (mm)<br />
Figure 25. Size distributions <strong>of</strong> <strong>the</strong> two different fertilizations used in <strong>the</strong> additional<br />
experiment (Ff and Fg). Dark bars indicate <strong>the</strong> portion <strong>of</strong> animals that were actually used.<br />
These were chosen in <strong>the</strong> heading part <strong>of</strong> <strong>the</strong> distribution for both fertilizations.<br />
9<br />
18<br />
10.5<br />
21<br />
12<br />
24<br />
13.5<br />
27<br />
15<br />
30<br />
16.5<br />
33<br />
18<br />
128
Table 11. Statistics on <strong>the</strong> small, large and mixed batches (fertilizations Ff, 'small' and Fg,<br />
'large'): mean sizes (pooled replicates) and comparison <strong>of</strong> mean sizes <strong>of</strong> separate versus<br />
mixed batches.<br />
Time<br />
(months)<br />
Small<br />
separate<br />
mean size<br />
(mm)<br />
Small<br />
mixed<br />
mean size<br />
(mm)<br />
Kruskal-<br />
Wallis<br />
stat.<br />
K.-W.<br />
proba.<br />
Large<br />
separate<br />
mean size<br />
(mm)<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
Large<br />
mixed<br />
mean size<br />
(mm)<br />
Kruskal-<br />
Wallis<br />
stat.<br />
Treatment Two fertilizations <strong>reared</strong> ei<strong>the</strong>r separately or mixed during 6 months.<br />
Then, 'large' individuals are discarded<br />
and 'small' individuals are <strong>reared</strong> alone for ano<strong>the</strong>r 6 months.<br />
0 8.5 8.5 - - 21.7 21.7 - -<br />
2 12.9 10.6 #<br />
K.-W.<br />
proba.<br />
179.6 < 0.001 28.6 28 3.32 0.068<br />
4 16.5 13.1 152.2 < 0.001 32.1 31.4 4.99 0.025 *<br />
6 18.0 13.6 156.0 < 0.001 34.3 33.7 4.07 0.044 *<br />
8 20.1 17.8 21.8 < 0.001<br />
10 22.3 20.9 3.91 0.048 *<br />
12 23.9 22.2 4.54 0.033 *<br />
# Difference between <strong>the</strong> 6 replicates inside a group is slightly significant (Kruskal-Wallis,<br />
0.01 < p < 0.05)<br />
*<br />
Difference slightly significant (0.01 < p < 0.05).<br />
After 6 months (Fig. 27), large individuals were taken away, while <strong>the</strong><br />
smaller ones were left in place. Inhibited individuals <strong>of</strong> <strong>the</strong> mixed group<br />
almost caught up with <strong>the</strong> o<strong>the</strong>rs within 4 months when inhibition was<br />
removed (Kruskal-Wallis, 0.01 < p < 0.05, test only slightly significant,<br />
Table 11). Just after <strong>the</strong> large echinoids have been removed, <strong>the</strong> heading<br />
group that rapidly formed in <strong>the</strong> mixed series had a remarkably high<br />
growth speed (compare Fig. 26, 6 months to Fig. 27, 8 months: some<br />
specimens grew from 18 to almost 30 mm within only two months, which<br />
meant an individual weight increase from about four time, viz. from ca. 2.7<br />
to ca. 11.0 g)! In <strong>the</strong> "mixed" series, when large <strong>sea</strong> <strong>urchin</strong>s were removed,<br />
sizes distribution spread much with a heading group that clearly detaches,<br />
as it was <strong>the</strong> case for small animals alone, six months before. This<br />
experiment demonstrates <strong>the</strong> high growth potential <strong>of</strong> inhibited<br />
individuals, a potential that was expressed when inhibitors were removed.<br />
129
Nbr.<br />
<strong>of</strong> ind.<br />
Nbr.<br />
<strong>of</strong> ind.<br />
Nbr.<br />
d'ind.<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
70<br />
0<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
70<br />
60<br />
50<br />
5<br />
0<br />
5<br />
40<br />
30<br />
20<br />
10<br />
0<br />
5<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
10<br />
15<br />
20<br />
Diameter (mm)<br />
10<br />
15<br />
20<br />
Diameter (mm)<br />
10<br />
15<br />
20<br />
Diameter (mm)<br />
2 months<br />
25<br />
25<br />
30<br />
30<br />
35<br />
35<br />
40<br />
4 months<br />
6 months<br />
25<br />
30<br />
35<br />
40<br />
40<br />
Large<br />
Large<br />
Large<br />
Mixed<br />
Mixed<br />
Mixed<br />
Figure 26. Change in size distributions <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> batches (large, mixed and small)<br />
with time (replicates are pooled).<br />
Small<br />
Small<br />
Small<br />
Large<br />
Small<br />
Large<br />
Small<br />
Large<br />
Small<br />
130
Nbr.<br />
<strong>of</strong> ind.<br />
Nbr.<br />
<strong>of</strong> ind.<br />
Nbr.<br />
<strong>of</strong> ind.<br />
60<br />
40<br />
20<br />
70<br />
60<br />
0<br />
60<br />
40<br />
20<br />
50<br />
40<br />
30<br />
0<br />
20<br />
10<br />
0<br />
5<br />
5<br />
5<br />
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
10<br />
10<br />
15<br />
20<br />
Diameter (mm)<br />
15<br />
8 months<br />
20<br />
Diameter (mm)<br />
10<br />
15<br />
25<br />
25<br />
30<br />
30<br />
35<br />
10 months<br />
20<br />
Diameter (mm)<br />
25<br />
30<br />
35<br />
12 months<br />
35<br />
40<br />
40<br />
40<br />
Mixed<br />
Small<br />
Mixed<br />
Mixed<br />
Figure 27. Change in size distributions <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> batches with time. Large individuals<br />
were removed from <strong>the</strong> mixed batches and <strong>the</strong> growth <strong>of</strong> small individuals (from ei<strong>the</strong>r <strong>the</strong><br />
"mixed" or <strong>the</strong> small batches) was recorded during an additional 6 months (pooled<br />
replicates).<br />
Small<br />
Small<br />
131
Part III: Experimental studies <strong>of</strong> <strong>the</strong> intraspecific competition<br />
132
PART IV<br />
A growth <strong>model</strong> with intraspecific competition<br />
133
134
PART IV: A GROWTH MODEL WITH<br />
INTRASPECIFIC COMPETITION<br />
We have now collected all elements required to describe <strong>the</strong> processes<br />
implied in <strong>the</strong> somatic growth <strong>of</strong> P. lividus in cultivation. We can thus<br />
elaborate a <strong>model</strong>. Since <strong>the</strong>re is no growth <strong>model</strong> that includes a<br />
component <strong>of</strong> intraspecific competition, we proposed an original approach<br />
for <strong>model</strong>ling growth. Statistical methods for fitting growth curves were<br />
also adapted.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
135
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
136
A functional growth <strong>model</strong> with intraspecific competition<br />
applied to a <strong>sea</strong> <strong>urchin</strong>, <strong>Paracentrotus</strong> lividus (Lamarck, 1816)<br />
a. Abstract<br />
Ph. Grosjean, Ch. Spirlet & M. Jangoux (submitted).<br />
An original <strong>model</strong> obtained by defuzzifying a fuzzy <strong>model</strong> is fitted on<br />
data from <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s, <strong>Paracentrotus</strong> lividus. Quantile regressions<br />
are used instead <strong>of</strong> least-square, for <strong>the</strong>y are insensitive to <strong>the</strong> dimension <strong>of</strong><br />
<strong>the</strong> measurement and accommodate more than just symmetrical<br />
distributions. Quantile regressions allow comparison <strong>of</strong> fittings on various<br />
parts <strong>of</strong> <strong>the</strong> size distributions, including large competitors versus small,<br />
inhibited animals, in <strong>the</strong> presence <strong>of</strong> a size-based intraspecific competition.<br />
The <strong>model</strong> has functionally interpretable parameters and allows<br />
quantifying <strong>of</strong> <strong>the</strong> intensity <strong>of</strong> inhibition. An extension <strong>of</strong> this <strong>model</strong>,<br />
called 'envelope <strong>model</strong>', fits <strong>the</strong> whole dataset at once, including size<br />
distributions. Its parameters are constrained using information about<br />
underlying biological processes involved, namely asymptotic growth with<br />
inhibition in early ages due to intraspecific competition whose intensity<br />
depends on <strong>the</strong> relative size <strong>of</strong> <strong>the</strong> individual in <strong>the</strong> cohort. The new <strong>model</strong><br />
appears most adequate to describe growth <strong>of</strong> <strong>Paracentrotus</strong> lividus and<br />
probably <strong>of</strong> many o<strong>the</strong>r <strong>sea</strong> <strong>urchin</strong>s species as well as o<strong>the</strong>r animals or<br />
plants. It is an intermediary <strong>model</strong> in a hierarchy <strong>of</strong> asymptotic growth<br />
<strong>model</strong>s ranging from <strong>the</strong> simplest one (von Bertalanffy 1) to more complex<br />
'dimensional' and 'transitional' groups. A general asymptotic growth<br />
<strong>model</strong>, being both 'dimensional' and 'transitional', is proposed. Most o<strong>the</strong>r<br />
<strong>model</strong>s, including <strong>the</strong> new one, are just special cases <strong>of</strong> this general <strong>model</strong>.<br />
However, it is only <strong>the</strong> visible tip <strong>of</strong> <strong>the</strong> iceberg. Many similar functions<br />
can be designed by defuzzifying simple fuzzy <strong>model</strong>s, including <strong>model</strong>s<br />
that do not derive from <strong>the</strong> von Bertalanffy curve. The family <strong>of</strong> so-called<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
137
. Introduction<br />
fuzzy-remanent functions represents a powerful alternative to dynamic<br />
<strong>model</strong>ling (using differential equations) for describing nonlinear<br />
phenomena widely observed in sciences.<br />
Keywords: growth <strong>model</strong>, intraspecific competition, fuzzy logic, quantile<br />
regression, <strong>Paracentrotus</strong> lividus, <strong>sea</strong> <strong>urchin</strong>, population dynamic,<br />
aquaculture.<br />
For <strong>the</strong> last two centuries, after Malthus (1798) discovered <strong>the</strong><br />
exponential nature <strong>of</strong> growth, <strong>the</strong> diversity <strong>of</strong> growth <strong>model</strong>s has steadily<br />
increased (Gompertz, 1825; Verhulst, 1838; Winsor, 1932; von<br />
Bertalanffy, 1938; Brody, 1945; Weibull, 1951; Richards, 1959; Preece &<br />
Baines, 1978; Schnute, 1981; Tanaka, 1982; Jolicoeur, 1985). However,<br />
<strong>the</strong>se <strong>model</strong>s compete ra<strong>the</strong>r than complement each o<strong>the</strong>r. Choosing a<br />
growth <strong>model</strong> <strong>of</strong>ten remains arbitrary (Fletcher, 1974). Sea <strong>urchin</strong><br />
individual growth is an example <strong>of</strong> such a problem (Gage & Tyler, 1985;<br />
Gage et al, 1986; Gage, 1987; Dafni, 1992; Ebert & Russell, 1993; Lamare<br />
& Mladenov, 2000). Papers on growth <strong>model</strong>s applied to <strong>sea</strong> <strong>urchin</strong>s<br />
compare <strong>the</strong> fit <strong>of</strong> different curves and are limited to <strong>the</strong>ir advantages and<br />
drawbacks in representing <strong>the</strong> growth <strong>of</strong> a "mean individual" (Gage and<br />
Tyler, 1985; Ebert & Russell, 1993; Lamare & Mladenov, 2000). They all<br />
conclude that none <strong>of</strong> <strong>the</strong>m is fully satisfactory. Ebert (1999, p. 200)<br />
wrote: "<strong>sea</strong>rching for <strong>the</strong> [growth] <strong>model</strong> that will provide <strong>the</strong> "best fit"<br />
can become a <strong>sea</strong>rch for <strong>the</strong> Grail with all <strong>of</strong> <strong>the</strong> fun being in <strong>the</strong> <strong>sea</strong>rch<br />
because <strong>the</strong>re may be no end."<br />
The key problem with <strong>the</strong>se <strong>model</strong>s is that parameters are not all<br />
comparable, and lack biological meaning (or lose it when applied to real<br />
data). Moreover, elaborated growth <strong>model</strong>s have three or more parameters<br />
that are not independent from each o<strong>the</strong>r (intercorrelations). Thus it is not<br />
possible to extract <strong>the</strong> estimator <strong>of</strong> one parameter from one fitting and<br />
compare it with <strong>the</strong> value obtained for <strong>the</strong> same parameter with ano<strong>the</strong>r<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
138
dataset. Indeed, its value depends on <strong>the</strong> estimation <strong>of</strong> all o<strong>the</strong>r parameters<br />
in each respective fitting. This contrasts with linear <strong>model</strong>s where slopes<br />
can be compared and usually convey a biological or physical meaning<br />
about <strong>the</strong> relationship between <strong>the</strong> considered variables (proportionality).<br />
Some authors have tried to combine parameters into a single one.<br />
Duineveld & Jenness (1984) used ω = k·Y∞ (Gallucci & Quinn, 1979) <strong>of</strong><br />
<strong>the</strong> von Bertalanffy equation Y(t) = Y∞·(1 – e -k·(t – t 0 ) ) to compare growth <strong>of</strong><br />
two populations <strong>of</strong> <strong>the</strong> irregular <strong>sea</strong> <strong>urchin</strong> Echinocardium cordatum<br />
(Pennant). <strong>Growth</strong> being an emergent property <strong>of</strong> various physiological<br />
processes like feeding, digestion, respiration, etc., it is dubious that it could<br />
be summarized into a single coefficient, and such a practice is probably<br />
error-prone.<br />
A few authors (e.g., Richards, 1959) have built up flexible and general<br />
growth <strong>model</strong>s that include some o<strong>the</strong>r existing <strong>model</strong>s as special cases.<br />
Schnute (1981) developed a general <strong>model</strong> whose parameters have a better<br />
biological meaning. None <strong>of</strong> <strong>the</strong>se <strong>model</strong>s have proved to be fully efficient<br />
when fitting real data, partly due to <strong>the</strong> problem <strong>of</strong> intercorrelation<br />
between parameters.<br />
It is thus only possible ei<strong>the</strong>r to carry on a global comparison <strong>of</strong><br />
various <strong>model</strong>s for <strong>the</strong> same dataset, or to use a single <strong>model</strong> applied to<br />
various datasets. Usually, ei<strong>the</strong>r <strong>the</strong> R 2 value or <strong>the</strong> residual sum <strong>of</strong> squares<br />
<strong>of</strong> <strong>the</strong> nonlinear least-square regression is used to evaluate how well a<br />
<strong>model</strong> fits <strong>the</strong> data (Gage & Tyler, 1985; Cellario & Fenaux, 1990; Lamare<br />
& Mladenov, 2000). A visual comparison <strong>of</strong> graphs is also commonly used<br />
(Gage & Tyler, 1985; Dafni, 1992; Ebert & Russell, 1993). These two<br />
techniques are not considered rigorous by statisticians. Among all papers<br />
cited, only Lamare & Mladenov (2000) performed a complete residual<br />
analysis. In this context, a growth <strong>model</strong> only summarizes <strong>the</strong> data cloud<br />
into a "best-fit" curve supposedly representing <strong>the</strong> growth <strong>of</strong> a mean<br />
individual. It is very difficult to use such a growth curve as a tool to<br />
investigate underlying processes (e.g., to understand <strong>the</strong> impact <strong>of</strong> a<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
139
considered factor on <strong>the</strong> curve's shape). For instance, current <strong>model</strong>s<br />
hardly cope with a superimposed effect <strong>of</strong> intraspecific competition on<br />
growth, though <strong>the</strong> latter was evidenced in <strong>sea</strong> <strong>urchin</strong>s (Ebert, 1977;<br />
Himmelman, 1986; Levitan, 1988; Grosjean et al, 1996) as in many o<strong>the</strong>r<br />
species (Branch, 1974, for <strong>the</strong> limpet Patella cochlear Born; Timmons &<br />
Shelton, 1980, for <strong>the</strong> largemouth bass Micropterus salmoides (Lacepede);<br />
Kautsky, 1982, for <strong>the</strong> mussel Mytilus edulis L.).<br />
One method to <strong>model</strong> growth in a functional way is by building<br />
equations that directly represent underlying processes, i.e., by elaborating a<br />
dynamic <strong>model</strong> that balances inputs (food, oxygen intake…) and outputs<br />
(carbon dioxide, feces…) from which it is possible to calculate variation <strong>of</strong><br />
size with time, thus yielding an estimation <strong>of</strong> growth. This is <strong>the</strong><br />
bioenergetic and/or ecophysiologic approach, using <strong>the</strong> scope for growth<br />
concept, which has proved very successful for filter feeders (Willows,<br />
1992). Such an approach requires a lot <strong>of</strong> measurements and equations. It<br />
is most <strong>of</strong>ten used in <strong>the</strong> simplest cases, where environmental conditions<br />
are constant, or vary in a very predictable way, like in a protected reef<br />
lagoon for cultivated pearl oysters (Pouvreau et al, 2000). Indeed, similar<br />
studies on European oyster cultures in <strong>the</strong> intertidal zone –a very changing<br />
and unpredictable environment– lead to a much more complex <strong>model</strong><br />
(Bacher et al, 1991). As far as we know, no such <strong>model</strong> has been<br />
completely successful when applied to <strong>sea</strong> <strong>urchin</strong>s because a large part <strong>of</strong><br />
<strong>the</strong> carbon or energy absorbed is lost as dissolved organic matter that is<br />
hard to quantify and to enter in equations (Miller & Mann, 1973; Lawrence<br />
& Lane, 1982).<br />
Being a basic feature <strong>of</strong> life, growth has been widely explored but it<br />
still remains unsatisfactorily <strong>model</strong>led in a functional way, possibly<br />
because <strong>of</strong> <strong>the</strong> approach used. Most growth <strong>model</strong>s were elaborated from<br />
<strong>the</strong>ir differential equations. Functions obtained by solving <strong>the</strong>se equations<br />
were <strong>the</strong>n systematically used to determine growth <strong>of</strong> a "mean individual"<br />
(by least-square regression) and no parameter <strong>of</strong> <strong>the</strong> <strong>model</strong> was<br />
constrained using particular knowledge (such as size at birth or at<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
140
c. Material<br />
metamorphosis) or hypo<strong>the</strong>ses that can be formulated about changes in<br />
individual growth (such as intra- or interspecific competition).<br />
The aim <strong>of</strong> this paper is to propose a growth <strong>model</strong> taking into account<br />
usually neglected aspects such as individual variations or intraspecific<br />
competition. This means we will question <strong>the</strong> concept <strong>of</strong> "mean<br />
individual" and tentatively build up a growth function with parameters<br />
carrying high biological meaning.<br />
The dataset we used results from a growth study <strong>of</strong> a single cohort <strong>of</strong><br />
<strong>sea</strong> <strong>urchin</strong>s, <strong>Paracentrotus</strong> lividus, <strong>reared</strong> over a period <strong>of</strong> 7 years in a<br />
controlled environment (see Grosjean et al, 1998, see Part I, for a detailed<br />
description <strong>of</strong> <strong>the</strong> rearing protocol). Echinoids were never size-sorted, nor<br />
individually tagged. All <strong>sea</strong> <strong>urchin</strong>s in <strong>the</strong> cohort were measured every 3<br />
months starting at 6 months old (younger echinoids are too fragile to be<br />
measured alive) until 4.5 years old, and <strong>the</strong>n every 6 months until 7 years<br />
old (Fig. 28A, see also Annex II). Due to mortality, <strong>the</strong> total number <strong>of</strong><br />
individuals in <strong>the</strong> cohort dropped from 725 at 6 months old to 221 at 4.5<br />
years old and to 67 at 7 years old (Fig. 29). Size is expressed by <strong>the</strong><br />
ambital test diameter D which corresponds to <strong>the</strong> external diameter <strong>of</strong> <strong>the</strong><br />
test at its largest region (<strong>the</strong> ambitus) excluding spines. D is measured with<br />
an electronic sliding caliper at <strong>the</strong> nearest 0.1 mm (Grosjean et al, 1999,<br />
see Part II) and recorded into 1 mm-wide size classes. Note that between<br />
400 and 1200 days, size distributions were heavily skewed, or even<br />
multimodal. This is <strong>the</strong> effect <strong>of</strong> an intraspecific competition (Grosjean et<br />
al, 1996, see Part III).<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
141
Diameter<br />
D 20<br />
in mm<br />
Diameter D in mm<br />
60<br />
0 10 20 30 40 50 60<br />
40<br />
A<br />
B<br />
0<br />
500<br />
1000<br />
1500<br />
Time t in days<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
2000<br />
2500<br />
0<br />
150<br />
100<br />
50<br />
Nbr.<br />
<strong>of</strong><br />
ind.<br />
500 1000 1500 2000 2500<br />
Time t in days<br />
quantile 0.975<br />
quantile 0.5 (median)<br />
quantile 0.025<br />
142
Figure 28. Left page. A. Histograms <strong>of</strong> size distributions <strong>of</strong> a cohort <strong>of</strong> <strong>reared</strong> P. lividus with<br />
time. Only 6-month interval histograms are presented, although measurements were<br />
performed every 3 months during <strong>the</strong> first 4.5 years (1600 days). Top <strong>of</strong> <strong>the</strong> box: a projection<br />
<strong>of</strong> three quantiles (0.025, 0.5 and 0.075) issued from those size distributions. They are also<br />
presented in (B) where points are unconditional quantiles extracted from each size<br />
distribution considered separately, and lines are conditional quantile regressions, thus<br />
considering <strong>the</strong> whole dataset (using all 6988 initial data points). Best fitting growth <strong>model</strong>s<br />
(according to δ1 values in Table 12) for each quantile are used: von Bertalanffy 1 for<br />
τ = 0.975, 4-parameter logistic for τ = 0.5 and Weibull for τ = 0.025.<br />
Nbr <strong>of</strong> individuals<br />
100 200 300 400 500 600 700<br />
500 1000 1500 2000 2500<br />
Time t in days<br />
Figure 29. Survival with time <strong>of</strong> <strong>the</strong> same <strong>reared</strong> cohort <strong>of</strong> P. lividus as in Fig. 28A. Line is a<br />
spline interpolation <strong>of</strong> observed values.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
143
d. Results<br />
Theoretical considerations: use <strong>of</strong> quantile regression instead<br />
<strong>of</strong> least-square regression<br />
The most widespread method to fit a curve is <strong>the</strong> use <strong>of</strong> a least-square<br />
method with one <strong>of</strong> <strong>the</strong> many minimization algorithms available [simplex,<br />
(quasi-)Newton, etc.; Sen & Srivastava, 1990; Draper & Smith, 1998;<br />
Nocedal & Wright, 1999]. The algorithm finds <strong>the</strong> combination <strong>of</strong> values<br />
for <strong>the</strong> various parameters in <strong>the</strong> <strong>model</strong> (<strong>the</strong> solution) that leads to a<br />
minimal value for <strong>the</strong> objective function, which is here <strong>the</strong> sum <strong>of</strong> <strong>the</strong><br />
square <strong>of</strong> <strong>the</strong> residuals (that is, <strong>the</strong> sum <strong>of</strong> squared distances between<br />
observed values for <strong>the</strong> dependent variable and values predicted by <strong>the</strong><br />
<strong>model</strong> at <strong>the</strong> same levels for <strong>the</strong> independent variables).<br />
Least-square regression has many advantages over o<strong>the</strong>r methods. In<br />
particular, when partial first (gradient matrix) and second (Hessian matrix)<br />
derivatives <strong>of</strong> <strong>the</strong> function are calculable for each parameter, convergence<br />
through a solution is accelerated and can be verified (at least for a local<br />
solution, Nocedal & Wright, 1999). In <strong>the</strong> counterpart, that regression<br />
supposes that <strong>the</strong> fluctuation around <strong>the</strong> <strong>model</strong> (called <strong>the</strong> error term) is<br />
additive, independent, normally distributed and with a constant standard<br />
deviation (heteroscedasticity). It is also very sensitive to outliers because it<br />
uses <strong>the</strong> squared residuals. Those constraints, even if not strictly met every<br />
time, particularly in many nonlinear phenomena like growth, appear to be<br />
<strong>of</strong> minor importance for many authors. Indeed, outliers are eliminated, or<br />
weighing methods are applied to limit <strong>the</strong>ir impact.<br />
Yet, <strong>the</strong>re are two arguments against <strong>the</strong> least-square regression used in<br />
<strong>the</strong> framework <strong>of</strong> growth <strong>model</strong>s, particularly when individuals' growth is<br />
influenced by <strong>the</strong> presence <strong>of</strong> conspecifics or <strong>of</strong> o<strong>the</strong>r species (indeed a<br />
general case to be verified in each study, except when a single individual is<br />
grown alone in a cage or an aquarium!): first it is sensitive to <strong>the</strong><br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
144
dimension <strong>of</strong> <strong>the</strong> size measurement, and second it can only <strong>model</strong> mean<br />
individuals.<br />
If least-square regression is influenced by <strong>the</strong> dimension <strong>of</strong> <strong>the</strong><br />
dependent variable used to quantify size in time, what is a good<br />
measurement <strong>of</strong> size? Is it a linear measurement (height, width, diameter,<br />
etc…) or is it a volume, a weight, or any o<strong>the</strong>r tri-dimensional measure? It<br />
could also be a two-dimensional measure such as, e.g., <strong>the</strong> surface covered<br />
by a colony <strong>of</strong> sponges or corals. Is <strong>the</strong>re a privileged dimension (1, 2 or<br />
3D) to measure growth? Except for changes in <strong>the</strong> curve shape due to<br />
distortion introduced by a power transformation –compare von<br />
Bertalanffy's <strong>model</strong> in size and in weight (von Bertalanffy, 1957, his<br />
Fig. 3, and Fig. 10 p 47)–, no criterion exists for choosing <strong>the</strong> best<br />
dimension to describe growth. Accordingly, a length, a surface or a<br />
volume/weight can each be acceptable, and <strong>the</strong> final choice will be<br />
dictated by practical considerations: which measurement is <strong>the</strong> easiest to<br />
obtain with <strong>the</strong> highest possible accuracy (see Grosjean et al, 1999, see<br />
Part II, for a discussion <strong>of</strong> this problem in P. lividus). Thus if a regression<br />
method is highly sensitive to power transformation, it will be less desirable<br />
because results <strong>of</strong> <strong>the</strong> regression will vary according to <strong>the</strong> measure used.<br />
By contrast, <strong>the</strong> median, as well as <strong>the</strong> quantiles, are insensitive to power<br />
transformation. Hence, quantile regression is generally insensitive to any<br />
transformation by a monotonous function (Koenker, 2001) and will not be<br />
influenced by <strong>the</strong> dimension (1, 2 or 3D) <strong>of</strong> <strong>the</strong> dependent measurement.<br />
Accordingly, a median individual in a regression <strong>of</strong> length against time<br />
will remain <strong>the</strong> same median individual in a regression <strong>of</strong> a volume (as<br />
length 3 , or any allometry coefficient) against time with a quantile<br />
regression, while this is not true for a mean individual using a least-square<br />
regression. Using median/quantiles instead <strong>of</strong> mean allows remaining <strong>the</strong><br />
independence <strong>of</strong> <strong>the</strong> dimension <strong>of</strong> size measurement in a context where <strong>the</strong><br />
dimension <strong>of</strong> <strong>the</strong> studied phenomenon (growth) is undetermined. As<br />
suggested by von Bertalanffy (1938, 1957), growth could be a<br />
multidimensional process, since it is a balance between anabolism (with<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
145
some two-dimensional processes like respiration or digestion, i.e.,<br />
exchanges across surfaces) and catabolism (proportional to <strong>the</strong> volume <strong>of</strong><br />
<strong>the</strong> animal, and thus a process quantified in 3D).<br />
The second argument against <strong>the</strong> least-square regression method is its<br />
limitation to <strong>model</strong> a "mean individual" in <strong>the</strong> presence <strong>of</strong> individual<br />
variations in <strong>the</strong> dataset. Yet, <strong>the</strong> concept <strong>of</strong> mean individual lacks<br />
meaning when <strong>the</strong> distribution <strong>of</strong> <strong>the</strong> error is not symmetrical and even,<br />
sometimes, multimodal. Considering a bimodal distribution with two<br />
similar, but well-separated modes, <strong>the</strong> mean value is located just in <strong>the</strong><br />
middle <strong>of</strong> <strong>the</strong> two groups, where <strong>the</strong>re are no individuals! In this example,<br />
<strong>the</strong> median is located at <strong>the</strong> same place and will also be a poor<br />
representation <strong>of</strong> <strong>the</strong> dataset. Quantiles, however, can be used more<br />
efficiently (1 st and 3 rd quartiles will be located around each <strong>of</strong> <strong>the</strong> two<br />
modes). Clearly, a regression that can fit a curve on ano<strong>the</strong>r part <strong>of</strong> <strong>the</strong><br />
distribution than a symmetrical position can be useful in situations where<br />
<strong>the</strong> distribution is asymmetrical or multimodal. For instance when sizebased<br />
intraspecific competition occurs, <strong>the</strong> largest animals are inhibitors,<br />
while <strong>the</strong> smallest ones represent <strong>the</strong> most inhibited fraction. <strong>Growth</strong><br />
curves fitted on large and small individuals thus contain information about<br />
<strong>the</strong> impact <strong>of</strong> <strong>the</strong> interspecific competition (by comparison). This<br />
information is not available using a least-square regression, but it is when<br />
two quantile regressions are used, based on large and small quantiles<br />
respectively.<br />
Quantile regression, as defined by Koenker & Bassett (1978) is an<br />
extension <strong>of</strong> <strong>the</strong> least-absolute deviation regression that fits a function for a<br />
median individual. In quantile regression, <strong>the</strong> objective function (called<br />
here deviance δ1) to be minimized is:<br />
with:<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
n<br />
δ 1= ∑ ρτ( Di−ξ1) (21)<br />
i=<br />
1<br />
146
- ξ1 being <strong>the</strong> solution returned by <strong>the</strong> <strong>model</strong>,<br />
- Di being each actual observation (diameter <strong>of</strong> a <strong>sea</strong> <strong>urchin</strong>) with i = 1…n<br />
observations.<br />
- ρτ (u) being a piece-wise linear function defined as:<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
ρ ( u) = u( τ − I( u<<br />
0))<br />
(22)<br />
τ<br />
and where τ is <strong>the</strong> quantile and I(u < 0) equals 1 if (u < 0) is true and 0 if it<br />
is false. Thus, quantile τ defines <strong>the</strong> fraction <strong>of</strong> all observations that lie<br />
beneath <strong>the</strong> curve. If τ = 0.5, half <strong>of</strong> <strong>the</strong> observations are beneath, and <strong>the</strong><br />
o<strong>the</strong>r half are above <strong>the</strong> curve, and <strong>the</strong> objective function (eq. 21)<br />
simplifies to half <strong>the</strong> least-absolute deviation<br />
∑<br />
1 δ 1= 2 | Di−ξ1| . With τ<br />
values different than 0.5 (0 < τ < 1), it is possible to fit a curve that will<br />
represent ano<strong>the</strong>r fraction <strong>of</strong> <strong>the</strong> population. For instance, τ = 0.975 fits a<br />
curve that represents <strong>the</strong> 2.5% largest individuals in <strong>the</strong> cohort; here <strong>the</strong>y<br />
are <strong>the</strong> fastest growing competitors. Similarly, using τ = 0.025 fits a curve<br />
that represents <strong>the</strong> 2.5% smallest individuals, here, <strong>the</strong> slowest growing<br />
fraction that undergoes <strong>the</strong> strongest competition. The surface between<br />
<strong>the</strong>se two curves represents 95% <strong>of</strong> <strong>the</strong> whole batch. A third curve fitted<br />
on median individuals using τ = 0.5 indicates asymmetry in <strong>the</strong><br />
distribution. If this last curve is located in <strong>the</strong> middle <strong>of</strong> <strong>the</strong> two previous<br />
ones, <strong>the</strong> distribution is symmetrical. If it is closer to <strong>the</strong> lowest curve, <strong>the</strong><br />
distribution is skewed toward small individuals. If it is closer to <strong>the</strong> highest<br />
curve, it is skewed toward large individuals. This is <strong>the</strong> representation we<br />
will keep for <strong>the</strong> following sections in this paper; it is more descriptive<br />
than just <strong>model</strong>ling a median (or a mean) individual.<br />
As far as we know, only one method is currently available to fit a<br />
nonlinear <strong>model</strong> using quantile regression (Koenker & Park, 1996), using a<br />
particular interior point algorithm (Nocedal & Wright, 1999). Presently,<br />
only R (Ihaka & Gentleman, 1996), a fast S-Plus clone freely available<br />
under <strong>the</strong> GNU license (http://cran.r-project.org) proposes a package that<br />
147
implements this method ('nlrq' package, by R. Koenker & Ph. Grosjean,<br />
see Annex I) from <strong>the</strong> same site. It runs on several platforms (several Unix,<br />
Linux, Windows, MacOS). Analyses in this paper are performed using this<br />
package, and also some experimental R code developed by <strong>the</strong> authors and<br />
available at http://www.sciviews.org/_phgrosjean/growth/index.htm. A<br />
script for complete treatment <strong>of</strong> <strong>the</strong> example presented in this paper is also<br />
provided.<br />
Elaboration <strong>of</strong> a new growth <strong>model</strong> which includes<br />
intraspecific competition<br />
Fuzzy logic can deal efficiently with complex nonlinear problems<br />
(Zimmermann, 1991; Passino & Yurkovich, 1998; Cox, 1999) and is thus<br />
ano<strong>the</strong>r possible approach for creating growth <strong>model</strong>s than dynamic<br />
<strong>model</strong>ling (differential equations) commonly used in this field, though, it<br />
has not been employed yet so far (Salski et al, 1995). Fuzzy systems can<br />
<strong>of</strong>ten be formulated ra<strong>the</strong>r intuitively in a linguistic way (Zimmermann,<br />
1991) and we will start this way, introducing ma<strong>the</strong>matics subsequently.<br />
For individual growth without competition, a trivial semantic<br />
description could be: "a young, small individual gradually becomes larger<br />
with age". In terms <strong>of</strong> fuzzy sets, this translates into a temporal transition<br />
between two sets: 'small' and 'large' (or S and L). Young animals belong to<br />
<strong>the</strong> S set, old ones belong to <strong>the</strong> L set, and "middle-aged" individuals<br />
belong partly to each <strong>of</strong> <strong>the</strong>se sets. The "degree <strong>of</strong> membership" to each set<br />
depends upon <strong>the</strong> amount <strong>of</strong> growth achieved and thus gradually shifts<br />
with time form S to L sets. This change is characterized by a membership<br />
function to each set, thus MS and ML respectively. The sum <strong>of</strong> all<br />
membership values at any given time is one, since we deal with a single<br />
individual in its integrity.<br />
The next step is to incorporate <strong>the</strong> concept <strong>of</strong> growth inhibition to<br />
represent <strong>the</strong> effect <strong>of</strong> intraspecific competition. Grosjean et al (1996, see<br />
Part III) showed that this competition is size-based in <strong>the</strong> case <strong>of</strong> <strong>reared</strong> P.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
148
lividus. 10 to 15% <strong>of</strong> <strong>the</strong> largest individuals (<strong>the</strong> inhibitors) in <strong>the</strong><br />
populations grow at <strong>the</strong>ir maximal speed (i.e. <strong>the</strong> growth speed <strong>the</strong>y would<br />
have if <strong>the</strong>y were alone in <strong>the</strong> same food and environmental conditions).<br />
The growth <strong>of</strong> o<strong>the</strong>rs depends upon <strong>the</strong>ir relative size in <strong>the</strong> population<br />
(<strong>the</strong> smaller <strong>the</strong>y are, <strong>the</strong> slower <strong>the</strong>y grow). The cause <strong>of</strong> this inhibition is<br />
not known yet, but it does not seem to be food-related: in <strong>the</strong> experiments,<br />
all individuals had access to food ad libitum. Inhibition progressively fades<br />
out when larger individuals reach <strong>the</strong>ir asymptotic size and are caught up<br />
with smaller ones that are still growing.<br />
According to <strong>the</strong>se observations, a semantic formulation <strong>of</strong> <strong>the</strong><br />
problem becomes: "a young, small individual is potentially inhibited in its<br />
growth, but gradually reaches its maximum size with age". It can also be<br />
represented by two sets and one transition, but now <strong>the</strong> S set is <strong>the</strong> minimal<br />
size with time (with maximum inhibition) and L set is <strong>the</strong> size at <strong>the</strong> age<br />
where <strong>the</strong> growth speed is maximal (with no inhibition at all). The<br />
transition is now <strong>the</strong> expression <strong>of</strong> a progressive release <strong>of</strong> <strong>the</strong> inhibition,<br />
instead <strong>of</strong> a representation <strong>of</strong> <strong>the</strong> entire growth process. The difficulty<br />
resides in <strong>the</strong> proper characterization <strong>of</strong> sets and membership functions<br />
with age.<br />
First <strong>of</strong> all, we use a time-scale t' with a well-defined origin that really<br />
coincides with <strong>the</strong> initiation <strong>of</strong> <strong>the</strong> growth process. Until now, we used<br />
(time-scale t) <strong>the</strong> age <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s (since fertilization, thus including<br />
larval life). However, growth <strong>of</strong> postmetamorphic <strong>sea</strong> <strong>urchin</strong>s really starts<br />
after metamorphosis. Knowing <strong>the</strong> age at metamorphosis (t0), t' is simply:<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
t' = t − t0<br />
(23)<br />
For <strong>the</strong> studied dataset, metamorphosis was artificially induced for all<br />
echinoids in <strong>the</strong> batch at <strong>the</strong> same time at 30 days old. t' scale is thus<br />
shifted to <strong>the</strong> left by t0 = 30 days.<br />
Set S corresponds to <strong>the</strong> minimum possible growth, that is simply no<br />
growth at all. Thus, in set S, size remains constant at its minimum initial<br />
149
value just after metamorphosis (D0, see Fig. 30). In this <strong>model</strong>, we do not<br />
consider negative growth (observed by Régis, 1979, on P. lividus; Ebert,<br />
1967, on Strongylocentrotus purpuratus; Levitan, 1988, on Diadema<br />
antillarum) because echinoids are fed ad libitum and <strong>the</strong>refore size<br />
shrinking should not occur. Moreover, negative growth was observed only<br />
for large, full-grown adults but not for juveniles that die instead <strong>of</strong><br />
reducing size in <strong>the</strong> absence <strong>of</strong> enough food to maintain <strong>the</strong>ir basic<br />
metabolism (unpubl. res.). Equation for S set is:<br />
10<br />
8<br />
6<br />
4<br />
2<br />
0.5<br />
D<br />
0<br />
0<br />
0<br />
2<br />
Membership<br />
1<br />
0<br />
2<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
D( t') = D<br />
(24)<br />
4<br />
4<br />
D∞<br />
D0<br />
0<br />
Dmax<br />
D0<br />
6<br />
6<br />
large (L)<br />
8<br />
8<br />
∆Dmax<br />
∆D∞<br />
small (S)<br />
fuzzy<br />
large (ML)<br />
small (MS)<br />
Figure 30. Construction <strong>of</strong> <strong>the</strong> fuzzy growth <strong>model</strong>. Two sets, called 'small' and 'large' (or S<br />
and L) are used. Actual size will always be located between <strong>the</strong>se two curves ('fuzzy'). The<br />
membership functions (ML and MS) <strong>model</strong> <strong>the</strong> belonging to each set with logistic functions. If<br />
a membership is close to 1, like MS for young juveniles or ML for large adults, <strong>the</strong> resulting<br />
fuzzy curve is close to <strong>the</strong> corresponding set. When memberships are 0.5 for each set, here at<br />
t' ≈ 3.9, <strong>the</strong> value returned by <strong>the</strong> fuzzy <strong>model</strong> is in <strong>the</strong> middle <strong>of</strong> values returned by both sets.<br />
10<br />
10<br />
t'<br />
t'<br />
150
Set L describes <strong>the</strong> largest size reached by <strong>sea</strong> <strong>urchin</strong>s at maximum<br />
growth speed with time. P. lividus having a determinate or asymptotic<br />
growth (see Fig. 28A), final increase <strong>of</strong> size ∆D∞ = D∞ – D0 is finite for<br />
t' → ∞. However, maximum size cannot be reached instantaneously. If <strong>the</strong><br />
largest individuals in <strong>the</strong> actual dataset are not inhibited at all, <strong>the</strong>y can be<br />
used as a reference for <strong>the</strong> whole cohort to define this maximum growth<br />
curve. Supposing that a von Bertalanffy 1 curve best fit <strong>the</strong> largest fraction<br />
<strong>of</strong> <strong>the</strong> size distributions (see next section), it is an appropriate <strong>model</strong> for set<br />
L. Because we use <strong>the</strong> t' time-scale here, we choose a parameterization <strong>of</strong><br />
<strong>the</strong> <strong>model</strong> such as D'(t' = 0) = D0:<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
−k1⋅t' Dt' ( ) = D +∆D⋅(1− e )<br />
(25)<br />
0<br />
∞<br />
Membership to <strong>the</strong> L set with time, ML(t'), is <strong>model</strong>led with a logistic<br />
function (a classical <strong>model</strong> for a transition in fuzzy sets, Cox, 1999)<br />
(Fig. 30):<br />
1<br />
M ( t')<br />
=<br />
1+ l ⋅e<br />
L −k2⋅t' (26)<br />
Membership to <strong>the</strong> S set with time, MS(t'), is complementary so that MS<br />
and ML add up to one:<br />
1<br />
M ( t') = 1 − M ( t')<br />
= 1− 1+ l ⋅e<br />
S L −k2⋅t' (27)<br />
Thus, full-growing individuals belong to <strong>the</strong> L set from <strong>the</strong> beginning. The<br />
stronger <strong>the</strong> growth inhibition, <strong>the</strong> longer o<strong>the</strong>r individuals remain in <strong>the</strong> S<br />
set before gradually shifting to <strong>the</strong> L set. The fuzzy <strong>model</strong> integrates <strong>the</strong><br />
effect <strong>of</strong> intraspecific competition (or any o<strong>the</strong>r inhibition mechanism<br />
having a similar effect on growth) as a delayed transition from S to L sets,<br />
as explicitly quantified by parameter l (<strong>the</strong> lag or position <strong>of</strong> <strong>the</strong> inflexion<br />
point in <strong>the</strong> membership curves, see Fig. 30). If l = 0, <strong>the</strong>re is no inflexion<br />
point and ML(t') = 1; growth occurs at maximum speed from start and<br />
follows set L, that is, a von Bertalanffy curve. While parameter k1<br />
151
quantifies maximum speed growth, k2 represents <strong>the</strong> speed at which<br />
inhibition is released with time. All parameters in this <strong>model</strong> carry a clear<br />
biological meaning, considering hypo<strong>the</strong>ses that were formulated to build<br />
it.<br />
Usually, a fuzzy <strong>model</strong> is treated with fuzzy arithmetic. The output is<br />
<strong>the</strong>n "defuzzified" by one <strong>of</strong> several methods (Cox, 1999) to provide a<br />
crisp number (<strong>the</strong> most probable size <strong>of</strong> an individual at a determined age).<br />
Being simple enough, <strong>the</strong> current <strong>model</strong> can also be transformed into a<br />
classical analytic equation:<br />
D( t') = M ( t') ⋅ S( t') + M ( t') ⋅ L( t')<br />
(28)<br />
S L<br />
which gives, after combination <strong>of</strong> eqs 24-28 and simplification:<br />
Dt' ( ) = D +∆D<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
0<br />
1−e 1+ l ⋅e<br />
−k1⋅t' ∞ −k2⋅t' (29)<br />
This way <strong>the</strong> <strong>model</strong> can be treated with classical (crisp) arithmetic that<br />
<strong>of</strong>fers a larger panel <strong>of</strong> statistical tools than fuzzy arithmetic.<br />
Fitting <strong>the</strong> dataset<br />
Since echinoids are not tagged individually, it is not possible to track<br />
animals across measurement sets. Consequently, one will consider virtual<br />
individuals according to <strong>the</strong>ir relative position in <strong>the</strong> entire size<br />
distribution at each sampled time, that is, virtual individuals corresponding<br />
to fixed quantiles (or percentiles) in each size distribution. It should be<br />
noted also that, if mortality is not randomly distributed among individuals,<br />
actual growth speed could be different from <strong>the</strong> one calculated on virtual<br />
individuals. This means that if mortality preferably affects small<br />
individuals, growth speed is overestimated; conversely, if mortality affects<br />
ra<strong>the</strong>r larger animals, growth speed is underevaluated. In absence <strong>of</strong><br />
individual tagging, we will thus consider <strong>the</strong> apparent growth speed <strong>of</strong> <strong>the</strong><br />
virtual individuals as defined here above.<br />
152
Ano<strong>the</strong>r feature <strong>of</strong> this dataset is that <strong>the</strong> error terms are time- and<br />
individual-dependents. Since <strong>the</strong> same (surviving) individuals are<br />
measured at each sampling time, this constitutes a time-series thus having<br />
autocorrelated errors. Moreover, even if artificial rearing conditions are<br />
kept as constant as possible (Grosjean et al, 1998, see Part I), some<br />
<strong>sea</strong>sonal variations are possible, partly because animals are fed with<br />
freshly field-collected kelp whose chemical composition is <strong>sea</strong>sondependent<br />
(Abe et al, 1983). To be rigorous, autocorrelation terms and<br />
some <strong>sea</strong>sonal variation should be introduced into <strong>the</strong> <strong>model</strong>. However, to<br />
simplify <strong>the</strong> <strong>model</strong> as much as possible, and also because <strong>the</strong>se effects are<br />
very limited (see fur<strong>the</strong>r), we decide to ignore <strong>the</strong>m here and we will thus<br />
fit growth curves without autocorrelation terms.<br />
Table 12 and Fig. 28B illustrate quantile regressions on P. lividus<br />
dataset with some usual growth <strong>model</strong>s. 4-parameters <strong>model</strong>s fit all 3<br />
quantiles while 3-parameters <strong>model</strong>s seem adequate for some quantiles<br />
only. Logistic function yields unreliable results in all cases. Criteria to<br />
decide which <strong>model</strong> best fits <strong>the</strong> data (deviance δ1 and visual impression<br />
on a graph) are nei<strong>the</strong>r rigorous nor discriminant. Two to four <strong>model</strong>s<br />
among <strong>the</strong> six tested seem adequate in each situation with <strong>the</strong>se criteria.<br />
Indeed, <strong>the</strong> increasing lag-phase for quantiles 0.5 and 0.025 compared to<br />
quantile 0.975 (more pronounced S-shape, see Fig. 28B) was<br />
experimentally demonstrated to be an inhibition in growth (Grosjean et al,<br />
1996, see Part III). None <strong>of</strong> <strong>the</strong>se <strong>model</strong>s, no more than many o<strong>the</strong>rs like<br />
Richards (1959), Preece & Baines (1978), Johnson (Ricker 1979), Schnute<br />
(1981), Tanaka (1982), Jolicoeur (1985)… contain explicit parameters that<br />
quantify such an inhibition and thus none <strong>of</strong> <strong>the</strong>m are really adequate in<br />
this case. Good fitting <strong>of</strong> data does not imply that <strong>the</strong> <strong>model</strong> is correct.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
153
Table 12. Results <strong>of</strong> quantile regressions with different growth <strong>model</strong>s for quantiles<br />
τ = 0.975, 0.5 and 0.025.<br />
<strong>model</strong> (a)<br />
a b c d deviance δδδδ1 fitting (b)<br />
τ = 0.975<br />
Gompertz 63.1 7.06·10 -2 0.998 . 2247 -<br />
von Bertalanffy 1 68.1 1.44·10 -3 81.8 . 2186 + +<br />
von Bertalanffy 2 63.9 2.24·10 -3 -178 . 2232 -<br />
logistic 61.8 3.45·10 -3 525 . 2334 - -<br />
4-param. logistic 67.3 1.55·10 -3 -1488 -761 2188 + +<br />
Weibull 67.1 1.04·10 -3 1.05 74.0 2186 + +<br />
τ = 0.5<br />
Gompertz 57.5 1.39·10 -2 0.977 . 14964 +<br />
von Bertalanffy 1 68.4 9.67·10 -4 185 . 15750 - -<br />
von Bertalanffy 2 59.4 1.87·10 -3 -54.3 . 14980 +<br />
logistic 54.6 3.77·10 -3 776 . 15328 - -<br />
4-param. logistic 56.9 2.70·10 -3 626 -13.9 14854 + +<br />
Weibull 56.5 5.79·10 -6 1.77 57.0 14870 + +<br />
τ = 0.025<br />
Gompertz 47.3 4.74·10 -4 0.998 . 1869 + +<br />
von Bertalanffy 1 60.9 8.14·10 -4 296 . 2020 - -<br />
von Bertalanffy 2 49.7 1.95·10 -3 114 . 1848 + +<br />
logistic 47.0 4.21·10 -3 929 . 2152 - -<br />
4-param. logistic 47.9 3.06·10 -3 809 -8.18 1853 + +<br />
Weibull 47.0 2.16·10 -7 2.21 48.2 1843 + +<br />
(a) Gompertz <strong>model</strong> is<br />
t<br />
c<br />
D = ab ⋅ , von Bertalanffy 1 is D = a·(1 - e -b·(t – c) ), von Bertalanffy 2 is<br />
D = a·(1 - e -b·(t – c) ) 3 , logistic is D = a/(1 + e -b·(t - c) ), 4-parameter logistic is D = (a – d)/(1 + e -b·(t - c) ) + d,<br />
c<br />
−bt ⋅<br />
and Weibull is D = a−d⋅ e (parameters are not all comparable between <strong>model</strong>s).<br />
(b) 'Fitting' is a visual impression <strong>of</strong> adequacy <strong>of</strong> <strong>the</strong> <strong>model</strong> on a graph (as presented in Fig. 28B for<br />
<strong>model</strong>s with lowest deviance δ1, in bold in <strong>the</strong> table).<br />
Fitting <strong>of</strong> <strong>the</strong> original <strong>model</strong> which includes intraspecific competition<br />
(eq. 29) on <strong>the</strong> same data and for <strong>the</strong> same quantiles τ = 0.975, 0.5 and<br />
0.025 is presented in Table 13. Deviance δ1 and visual inspection on a<br />
graph (not shown but very close to Fig. 28B) indicate that this <strong>model</strong> is<br />
one <strong>of</strong> <strong>the</strong> most adequate, with all cautions previously formulated about<br />
<strong>the</strong>se criteria. The major difference between this <strong>model</strong> and previous ones<br />
(in Table 12) is <strong>the</strong> better biological meaning <strong>of</strong> its parameters. However,<br />
unconstrained regression leads to meaningless estimations <strong>of</strong> some<br />
parameters: all D0 values are negative, meaning a negative size at<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
154
metamorphosis! Biological meaning <strong>of</strong> parameters in <strong>the</strong>ory does not<br />
imply meaningful estimates when using unconstrained regression on real<br />
datasets. Constraints should be formulated according to background<br />
knowledge or hypo<strong>the</strong>ses about <strong>the</strong> phenomenon studied. For instance, it is<br />
logical to constrain D0 to be a positive value, since negative size is<br />
meaningless.<br />
Table 13. Results <strong>of</strong> quantile regressions for three values <strong>of</strong> τ using <strong>the</strong> new growth <strong>model</strong><br />
(eq. 29). Curves are not shown in a graph, but <strong>the</strong>y are quasi identical to those in Fig. 28B.<br />
ττττ D0 ∆D∞ k1 k2 l deviance δδδδ1<br />
0.975 -8.34 74.6 5.42·10 -3 2.03·10 -3 326 2183<br />
0.5 -5.09 61.5 8.26·10 -3 2.95·10 -3 693 14827<br />
0.025 -3.79 52.3 2.59·10 -3 2.86·10 -3 749 1855<br />
Constraining parameters <strong>of</strong> <strong>the</strong> <strong>model</strong><br />
It is possible to do a little better for D0 than just forcing it to be<br />
positive. If it is known, it can be replaced in eq. 29 by its real value.<br />
Ambital test diameter was measured on a large number <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
(n = 296) just after metamorphosis in similar rearing conditions by<br />
Grosjean et al (1996, see Part III). Its values are normally distributed, with<br />
a mean <strong>of</strong> 0.497 mm and a standard deviation <strong>of</strong> 0.056 mm. Since this<br />
spreading in initial sizes is negligible compared to <strong>the</strong> size-scale during <strong>the</strong><br />
whole growth process (compare 0.056 mm with 0.50 to 50-65 mm), one<br />
can simplify <strong>the</strong> <strong>model</strong> and consider that <strong>the</strong> initial size <strong>of</strong> echinoids just<br />
after metamorphosis, D0, is about 0.5 mm for any individual. Accepting<br />
this simplification, it is possible to eliminate D0 from <strong>the</strong> equations by<br />
working with size increase D' instead <strong>of</strong> absolute size D:<br />
and:<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
D'( t') = Dt' ( ) − D<br />
(30)<br />
D'( t') =∆D<br />
1−e 1+ l ⋅e<br />
0<br />
−k1⋅t' ∞ −k2⋅t' (31)<br />
155
This way, one obtains a 4-parameter <strong>model</strong> with an origin constrained to<br />
be D'(t' = 0) = 0, size increase being null just after metamorphosis for all<br />
quantiles.<br />
Fitting <strong>the</strong> modified growth <strong>model</strong> <strong>of</strong> eq. 31 with quantile regression<br />
for various quantiles is shown in Table 14 and Fig. 31. The <strong>model</strong> is<br />
flexible enough to accommodate any quantile. Adding a constraint on one<br />
parameter leads to a slightly poorer fitting according to δ1 and visual<br />
impression, which is not surprising.<br />
Table 14. Results <strong>of</strong> quantile regressions for different values <strong>of</strong> τ, using <strong>the</strong> new growth <strong>model</strong><br />
constrained to <strong>the</strong> origin (eq. 31). Curves for quantiles τ = 0.025, 0.5 and 0.975 (in bold) are<br />
also presented graphically in Fig. 31. Relations between some parameters <strong>of</strong> <strong>the</strong> <strong>model</strong> and τ<br />
are shown in Fig. 32.<br />
ττττ ∆D∞ k1 k2 l deviance δδδδ1<br />
0.975 64.5 1.97·10 -3 1.89·10 -3 0.806 2251<br />
0.95 68.2 1.25·10 -3 7.51·10 -3 1.60 4048<br />
0.9 60.9 2.16·10 -3 2.80·10 -3 1.96 7105<br />
0.85 59.8 2.39·10 -3 2.76·10 -3 2.65 9542<br />
0.8 58.8 2.31·10 -3 2.82·10 -3 3.09 11448<br />
0.75 58.3 2.39·10 -3 2.82·10 -3 3.77 12871<br />
0.7 57.5 2.35·10 -3 2.84·10 -3 4.07 13898<br />
0.65 57.9 1.88·10 -3 3.02·10 -3 3.99 14577<br />
0.6 57.0 2.21·10 -3 2.89·10 -3 4.93 14968<br />
0.55 57.2 1.70·10 -3 3.33·10 -3 5.04 15094<br />
0.5 56.6 1.89·10 -3 3.11·10 -3 5.38 14943<br />
0.45 55.9 1.93·10 -3 3.15·10 -3 6.16 14616<br />
0.4 55.8 1.96·10 -3 3.17·10 -3 7.00 14049<br />
0.35 55.6 1.73·10 -3 3.38·10 -3 7.28 13296<br />
0.3 55.1 1.74·10 -3 3.43·10 -3 8.11 12338<br />
0.25 55.6 1.44·10 -3 3.65·10 -3 8.25 11152<br />
0.2 55.4 1.35·10 -3 3.84·10 -3 9.57 9702<br />
0.15 55.4 1.25·10 -3 3.90·10 -3 9.80 7993<br />
0.1 57.7 9.92·10 -4 4.58·10 -3 13.8 5948<br />
0.05 54.0 1.03·10 -3 4.83·10 -3 19.4 3429<br />
0.025 53.3 9.24·10 -4 5.84·10 -3 39.9 1943<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
156
Diameter increase D' in mm<br />
0 10 20 30 40 50 60<br />
500 1000 1500 2000 2500<br />
Time t' in days<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
quantile 0.975<br />
quantile 0.5 (median)<br />
quantile 0.025<br />
Figure 31. First step <strong>of</strong> constraining parameters <strong>of</strong> <strong>the</strong> new growth <strong>model</strong> (eq. 31, origin<br />
forced to {t0, D0}). At this stage, fitting seems slightly poorer than with some unconstrained<br />
<strong>model</strong>s (compare with Fig. 28B). As a consequence <strong>of</strong> independence <strong>of</strong> <strong>the</strong> three quantile<br />
regressions, curvatures do not appear "homogeneous" between curves.<br />
Some inconsistencies are observed for extreme quantiles. This is partly<br />
due to a less satisfactory convergence <strong>of</strong> <strong>the</strong> quantile regression because a<br />
lower number <strong>of</strong> data points have a greater influence on <strong>the</strong> fitting (due to<br />
ρτ(u), see eq. 22) for large and small quantiles.<br />
As a consequence <strong>of</strong> independence <strong>of</strong> <strong>the</strong> fitting for <strong>the</strong> different<br />
quantiles, curves do not appear very harmonious in Fig. 31. This can be<br />
solved by linking regressions, that is, by adding relationships between <strong>the</strong><br />
4 parameters and τ. Intuitively, <strong>the</strong>re should be a relationship between each<br />
<strong>of</strong> <strong>the</strong>se curves because <strong>the</strong>y originate from a single dataset with a single<br />
conditional distribution and also because <strong>the</strong>y represent growth <strong>of</strong> virtual<br />
157
individuals related to <strong>the</strong> presence <strong>of</strong> o<strong>the</strong>r virtual individuals (inhibitorsinhibited<br />
interactions). With current <strong>model</strong> (eq. 31) and quantile regression<br />
method (eqs. 21 and 22), it is only possible to fit one curve at a time. An<br />
extension or adaptation <strong>of</strong> both <strong>the</strong> <strong>model</strong> and <strong>the</strong> regression method are<br />
required to link curves for different quantiles.<br />
In <strong>the</strong> case <strong>of</strong> parameter l, we have already mentioned that we expect<br />
l = 0 for τ = 1, according to <strong>the</strong> hypo<strong>the</strong>sis that larger animals in <strong>the</strong> batch<br />
are not inhibited at all (see above, construction <strong>of</strong> <strong>the</strong> <strong>model</strong>). We would<br />
also expect a monotonous increase <strong>of</strong> l with a decrease <strong>of</strong> τ because<br />
fractions <strong>of</strong> smaller individuals should be more inhibited than fractions <strong>of</strong><br />
larger ones (recall this is a size-based competition mechanism). Fig. 32A<br />
highlights a linear relationship between l and τ, except for <strong>the</strong> 10% smaller<br />
fraction. Consequently, we constrain l(τ) as:<br />
where s is <strong>the</strong> slope <strong>of</strong> <strong>the</strong> linear relation.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
l( τ ) = s⋅(1<br />
− τ )<br />
(32)<br />
k1 and k2 appear negatively correlated in Fig. 32B but are quite<br />
constant along τ values, except for extreme quantiles. Moreover, large<br />
values <strong>of</strong> k2 for <strong>the</strong> three rightmost points in <strong>the</strong> graph at Fig. 32B seem<br />
associated with a potential overestimation <strong>of</strong> corresponding l values in<br />
Fig. 28A (outliers). In regard with <strong>the</strong>se considerations, reasonable<br />
relationships between k1/k2 and τ could be:<br />
k1( τ ) = cste = k1; k2( τ ) = cste = k2<br />
(33)<br />
with k1 probably different (and lower) than k2.<br />
158
k<br />
l<br />
45<br />
40<br />
35<br />
30<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
0.008<br />
0.007<br />
0.006<br />
0.005<br />
0.004<br />
0.003<br />
0.002<br />
0.001<br />
0<br />
l<br />
l (outliers)<br />
<strong>model</strong><br />
l = 11.644 (1 - τ )<br />
R 2 = 0.9595<br />
0 0.2 0.4 0.6 0.8 1<br />
1 - τ<br />
k2<br />
k1<br />
0 0.2 0.4 0.6 0.8 1<br />
1 - τ<br />
Figure 32. A. Variation <strong>of</strong> l as a function <strong>of</strong> 1-τ for several quantile regressions performed<br />
separately with <strong>the</strong> new growth <strong>model</strong> constrained to <strong>the</strong> origin (eq. 31, Table 14 and<br />
Fig. 31). A simple linear relationship appears suitable to define l in function <strong>of</strong> τ , except for<br />
<strong>the</strong> 10% smallest individuals (black dots, "outliers"). 'Model' is a least-square linear<br />
regression, after eliminating <strong>the</strong>se outliers, with s = 11.6 (R 2 = 0.960). It is just indicative. B.<br />
Variation <strong>of</strong> k1 (black squares) and k2 (white triangles) as functions <strong>of</strong> 1-τ.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
A<br />
B<br />
159
Finally, ∆D∞ should follow a normal distribution, as size distributions<br />
when approaching asymptotic maximum size are normal or close to<br />
normal (see Fig. 28A, t > 1500 days, and also Grosjean et al, 1996, see<br />
Part III):<br />
with µ D∞<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
∆D ( τ) ∼ N ( µ , σ )<br />
(34)<br />
∞ ∆D ∆D<br />
∆ being <strong>the</strong> mean and σ ∆D∞<br />
normal distribution <strong>of</strong> ∆D∞(τ).<br />
∞ ∞<br />
Replacing eqs. 32-34 into eq. 31, we get:<br />
−k1⋅t' 1+ e<br />
D'( t', τ) =∆D<br />
( τ)<br />
1 −s⋅(1 −τ) ⋅e<br />
being <strong>the</strong> standard deviation <strong>of</strong> <strong>the</strong><br />
∞ −k2⋅t' (35)<br />
which links curves for all quantiles 0 < τ < 1 and has 5 parameters to be<br />
estimated: k1, k2, s, µ D∞<br />
∆ and σ ∆ D∞<br />
. It includes individual variations into<br />
<strong>the</strong> <strong>model</strong> and is a kind <strong>of</strong> 3D-surface that envelops data (see Fig. 33). For<br />
this reason, it will be called an 'envelope <strong>model</strong>'.<br />
The quantile regression method is modified as follows. Considering<br />
that every individual present in <strong>the</strong> batch is measured at each sampling<br />
time, unconditional quantiles at each size distribution can be regarded as<br />
estimators <strong>of</strong> conditional quantiles τ at corresponding time t' in eq. 35<br />
(note that this is fundamentally different than <strong>the</strong> previous quantile<br />
regression method in eqs. 21-22 where unconditional quantiles were not<br />
used at all in <strong>the</strong> regression). Estimators <strong>of</strong> τ, noted ˆ τ , are <strong>the</strong>n calculated<br />
as:<br />
ˆ τ =<br />
i<br />
nt' ( i )<br />
∑<br />
j=<br />
1<br />
( D'j t'i < D'i)<br />
I ( )<br />
nt' ( )<br />
i<br />
(36)<br />
where t'i is t' corresponding to <strong>the</strong> i th observation, n(t'i) is <strong>the</strong> total number<br />
<strong>of</strong> individuals measured at time t'i, D'j(t'i) is <strong>the</strong> j th observation among all<br />
160
measures made at time t'i and I(u < v) returns 1 if true and 0 if false, as in<br />
eq. 22. Parameters <strong>of</strong> <strong>the</strong> envelope <strong>model</strong> to fit, ξ2, are estimated by<br />
minimizing <strong>the</strong> following objective function δ2:<br />
δ 2 =<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
n<br />
∑<br />
i=<br />
1<br />
| D' −ξ<br />
2( t' , ˆ τ )|<br />
i i i<br />
n<br />
(37)<br />
δ2 is indeed <strong>the</strong> mean absolute deviation between observed and predicted<br />
sizes for all observations. A robust simplex minimization algorithm is used<br />
to converge to <strong>the</strong> solution (Nelder & Mead, 1965; Nocedal & Wright,<br />
1999).<br />
Fitting <strong>of</strong> <strong>the</strong> envelope <strong>model</strong> (eq. 35) by minimizing δ2 is presented in<br />
Fig. 33. This graph emphasizes how individual variation is now included<br />
in <strong>the</strong> <strong>model</strong> itself. Gain is obvious by comparing it to Fig. 28A, where <strong>the</strong><br />
same dataset is summarized into less informative 2D-curves. Parameters <strong>of</strong><br />
<strong>the</strong> <strong>model</strong> are: k1 = 1.53 10 -3 , k2 = 3.65 10 -3 , s = 12.7, µ ∆ D = 57.0 and<br />
∞<br />
σ ∆ D = 4.28, deviance δ2 = 1.18.<br />
∞<br />
Fig. 34 is a diagnostic <strong>of</strong> this regression. Fig. 34A shows residuals (as<br />
differences between observed and predicted values) using a contour plot.<br />
There are only small patches <strong>of</strong> residuals above 2 mm or below –2 mm,<br />
attesting a good fitting <strong>of</strong> <strong>the</strong> <strong>model</strong>. Residuals are not randomly<br />
distributed. This is probably due to some autocorrelation in <strong>the</strong> dataset, to<br />
some subtle environmental fluctuations in <strong>the</strong> rearing system (<strong>sea</strong>sonal<br />
variations…), or possibly to some lack <strong>of</strong> fit <strong>of</strong> <strong>the</strong> <strong>model</strong>.<br />
161
60<br />
Diameter<br />
increase<br />
D'<br />
in mm<br />
40<br />
20<br />
0<br />
500<br />
1000<br />
1500<br />
Time t' in days<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
2000<br />
2500<br />
Figure 33. Envelope <strong>model</strong> (eq. 35) fitted (upper surface) to <strong>the</strong> whole dataset (lower<br />
surface). The dataset is <strong>the</strong> same as <strong>the</strong> histograms <strong>of</strong> Fig. 28A, but represented differently<br />
here to facilitate comparison with <strong>the</strong> <strong>model</strong>. Elevations (z-values, number <strong>of</strong> individuals) in<br />
<strong>the</strong> <strong>model</strong>'s surface are weighted according to <strong>the</strong> number <strong>of</strong> individuals surviving with time<br />
in <strong>the</strong> dataset (spline interpolation <strong>of</strong> actual values, see Fig. 29).<br />
0<br />
50<br />
100<br />
150<br />
Nbr.<br />
<strong>of</strong><br />
ind.<br />
162
Nbr <strong>of</strong> individuals<br />
0 20 40 60 80 100<br />
A<br />
Diameter increase D' in mm<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
1<br />
0 10 20 30 40 50 60<br />
Diameter increase D' in mm<br />
Figure 34. Diagnostic <strong>of</strong> <strong>the</strong> envelope <strong>model</strong> (eq. 35) fitted in Fig. 33. A. Contour plot <strong>of</strong> <strong>the</strong><br />
residuals as [observed D' - predicted D'] (shades according to <strong>the</strong> scale at right, in mm).<br />
Quantiles 0.025, 0.5 and 0.975, obtained from <strong>the</strong> initial dataset (points) and corresponding<br />
curves extracted from <strong>the</strong> envelope <strong>model</strong> by fixing τ (lines) are superimposed. Three "slices"<br />
are cut in <strong>the</strong> 3D-surfaces <strong>of</strong> Fig. 33 at t' = 300 (1), 600 (2) and 1800 (3) days (vertical dotted<br />
lines in Fig. 34A) and represented as size distributions in B. For each pair <strong>of</strong> distributions,<br />
<strong>the</strong> smoo<strong>the</strong>st one is <strong>the</strong> <strong>model</strong>. Differences are clearly visible and correspond to positive or<br />
negatives patches in <strong>the</strong> residuals in A.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
500 1000 1500 2000 2500<br />
Time t' in days<br />
2 3<br />
quantile 0.975<br />
quantile 0.5<br />
quantile 0.025<br />
B<br />
163<br />
4<br />
2<br />
0<br />
-2<br />
-4
e. Discussion<br />
Fig. 34B shows three sections across <strong>the</strong> surfaces <strong>of</strong> Fig. 33 at three<br />
given times t' = 300, 600 and 1800 days. Overall size spreading and<br />
asymmetries in <strong>the</strong> original dataset are quite well respected by <strong>the</strong> <strong>model</strong>.<br />
The higher peak in <strong>the</strong> first section <strong>of</strong> <strong>the</strong> <strong>model</strong> at 300 days is partly a<br />
consequence <strong>of</strong> considering D0 as strictly equivalent for all individuals<br />
while, in reality, it is normally distributed. With <strong>the</strong> simple relation<br />
between l and τ, as established in eq. 32, multiples modes are just<br />
approximated by a unimodal, skewed distribution. This is most visible in<br />
<strong>the</strong> second section, at 600 days. A more complex <strong>model</strong> would be required<br />
to fit multimodal distributions.<br />
Fitting methods<br />
We have argued in favor <strong>of</strong> quantile regression instead <strong>of</strong> least-square<br />
regression in <strong>model</strong>ling growth. Distribution <strong>of</strong> <strong>the</strong> "error", that is, mainly<br />
individual variation in <strong>the</strong> present case, can be ei<strong>the</strong>r asymmetrical or<br />
multimodal and this is a violation <strong>of</strong> basic assumptions <strong>of</strong> <strong>the</strong> least-square<br />
<strong>model</strong>. In <strong>the</strong> same circumstance, a mean effect obtained by least-square is<br />
less representative <strong>of</strong> <strong>the</strong> tendency. Quantile regression fits any part <strong>of</strong> <strong>the</strong><br />
distribution, including extremes, and accommodates any kind <strong>of</strong> size<br />
distribution. It is also robust against any power transformation and it is a<br />
guarantee that <strong>the</strong> regression remains independent from <strong>the</strong> dimension <strong>of</strong><br />
measurements for size (linear versus surface versus volume/weight). Leastsquare<br />
regression is very sensitive to any power transformation, and thus<br />
to <strong>the</strong> dimension <strong>of</strong> <strong>the</strong> variables.<br />
For purely descriptive fittings, we introduced <strong>the</strong> triple<br />
τ = 0.975 / 0.5 / 0.025 quantile regression representation as an informative<br />
summary <strong>of</strong> <strong>the</strong> data. 5% is a commonly used critical level in statistics.<br />
The curves for τ = 0.975 and t = 0.025 materialize a kind <strong>of</strong> two-tailed 5%<br />
nonparametric conditional confidence interval <strong>of</strong> <strong>the</strong> dataset: 95% <strong>of</strong> data<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
164
are inside this interval. Additionally, a τ = 0.5 median curve visualizes<br />
possible asymmetry and its change with time. When a representative<br />
sample <strong>of</strong> <strong>the</strong> whole size distribution at each time is available, typically<br />
with n > 50, points representing unconditional quantiles can be superposed<br />
on <strong>the</strong> graph to show how well quantile regressions match <strong>the</strong>m, as we did<br />
in Figs. 28B, 31 and 34A. There is no residuals analysis for this kind <strong>of</strong><br />
quantile regression, at least in <strong>the</strong> way it is conceived for least-square<br />
regression.<br />
Nonlinear quantile regression using interior point algorithm proposed<br />
by Koenker & Park (1996) is compatible with many growth <strong>model</strong>s. For<br />
<strong>the</strong> cases where representative samples are measured at each time interval,<br />
we have proposed a modified method to use with so-called 'envelope<br />
<strong>model</strong>s', that is, <strong>model</strong>s <strong>of</strong> <strong>the</strong> form Y = f(t, τ) as eq. 35. These <strong>model</strong>s<br />
calculate a 3D-surface enveloping data (Fig. 33). Curves for all quantiles<br />
0 < τ < 1 are calculated at once. They contain most <strong>of</strong> <strong>the</strong> information in<br />
<strong>the</strong> initial dataset, including individual variability. They are particularly<br />
useful when individual variability in itself expresses some aspects <strong>of</strong> <strong>the</strong><br />
phenomenon studied, like growth in <strong>the</strong> presence <strong>of</strong> an intraspecific<br />
competition. We have developed basic tools to fit and diagnose <strong>the</strong>m<br />
(namely, basic graphical residuals analysis, Fig. 34). Tests <strong>of</strong> significance<br />
<strong>of</strong> <strong>the</strong> fitting can be elaborated as extensions <strong>of</strong> some tests for<br />
unconditional size-distributions, like χ 2 or Kolmogorov-Smirnov tests or<br />
by o<strong>the</strong>r means (Koenker & Machado, 1999). Confidence intervals for <strong>the</strong><br />
parameters do not yet exist. Tests for comparing various <strong>model</strong>s fitted on<br />
<strong>the</strong> same dataset, or to compare a single <strong>model</strong> fitted on various datasets<br />
remain also to be developed. However, one difficulty in conceptualizing<br />
such tests for envelope <strong>model</strong>s is that one <strong>of</strong> <strong>the</strong> "independent variables",<br />
τ, is estimated according to <strong>the</strong> dependent variable y (eq. 36) and is thus<br />
not really independent.<br />
Flexible, unconstrained envelope <strong>model</strong>s can be incredibly complex,<br />
with dozens <strong>of</strong> parameters. They are impossible to fit in practice.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
165
Constraining parameters as we did in eq. 30-34 leads to a double benefit.<br />
First, it simplifies <strong>the</strong> <strong>model</strong>. In <strong>the</strong> present case, we started with a 5parameter<br />
unconstrained classical <strong>model</strong> (eq. 29) and we ended with a 5parameter<br />
constrained envelope <strong>model</strong> (eq. 35). Yet, <strong>the</strong> latter contains<br />
much more information than <strong>the</strong> former. Second, if constraints are<br />
formulated according to some knowledge about <strong>the</strong> underlying<br />
phenomenon (value <strong>of</strong> <strong>the</strong> intercept corresponding to actual initial size) or<br />
to some reasonable hypo<strong>the</strong>ses (relation between l and τ, in regard with<br />
information in <strong>the</strong> literature), parameters remain meaningful in <strong>the</strong> fitted<br />
<strong>model</strong>. A correct formulation <strong>of</strong> both <strong>the</strong> initial unconstrained <strong>model</strong> and<br />
<strong>of</strong> superimposed constraints is a bit <strong>of</strong> an art. It requires many trials and<br />
errors, much patience and perseverance. But at <strong>the</strong> end, it pays <strong>of</strong>f with a<br />
<strong>model</strong> whose parameters are fully functionally interpretable, even on real<br />
data.<br />
Fitting and individual variations in growth<br />
A good fitting is not a criterion for deciding if a <strong>model</strong> is adequate<br />
(Fletcher, 1974). Choosing an inadequate <strong>model</strong> that fit <strong>the</strong> data very well<br />
is not problematic if that <strong>model</strong> is just used for descriptive purposes. It<br />
turns out to be a problem when parameters are functionally interpreted or<br />
if it is used in population dynamic simulations. In this case, individual<br />
variation should not be simply considered as an independent, normally<br />
distributed "error" whenever it is not. Individual variation has <strong>of</strong>ten been<br />
overlooked in <strong>the</strong> literature. The most aberrant calculations could result<br />
from such mistakes. For instance, Basuyaux & Blin (1998) extrapolated<br />
over 4 years a growth <strong>model</strong> for P. lividus where size distributions were<br />
supposed to be normal and using measurements from 7 to 23 months only.<br />
They calculated <strong>the</strong> fraction <strong>of</strong> <strong>the</strong> size distribution that would reach 40<br />
mm (<strong>the</strong> minimal market size) with time on basis <strong>of</strong> this extrapolation.<br />
Many techniques for population dynamic analyses are based on <strong>the</strong><br />
assertion that cohorts should distribute normally, including most recent<br />
ones (Smith & Botsford, 1998; Morgan et al, 2000) and could be also<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
166
iased for <strong>the</strong> same reason. On <strong>the</strong> contrary, <strong>the</strong> envelope <strong>model</strong> is an<br />
elegant alternative that considers non-symmetrical individual variations in<br />
<strong>the</strong> <strong>model</strong> itself.<br />
Not considering individual variation and asymmetries in <strong>the</strong> size<br />
distributions could lead to <strong>the</strong> rejection <strong>of</strong> <strong>the</strong> von Bertalanffy <strong>model</strong><br />
(Sainsbury, 1980). In <strong>the</strong> case <strong>of</strong> P. lividus, Cellario & Fenaux (1990) for<br />
<strong>reared</strong> and Turon et al (1995) for wild populations both rejected <strong>the</strong> von<br />
Bertalanffy <strong>model</strong> in favor <strong>of</strong> <strong>the</strong> Gompertz curve. We would conclude to<br />
<strong>the</strong> same rejection <strong>of</strong> <strong>the</strong> von Bertalanffy 1 <strong>model</strong> if we considered only<br />
median quantile regression with τ = 0.5 in <strong>the</strong> present study (see Table 12).<br />
However, taking intraspecific competition into account using <strong>the</strong> new<br />
growth <strong>model</strong> leads to a different conclusion when inhibition is eliminated.<br />
Functional interpretation <strong>of</strong> <strong>the</strong> von Bertalanffy <strong>model</strong> in <strong>sea</strong><br />
<strong>urchin</strong>s<br />
From a functional point <strong>of</strong> view, von Bertalanffy growth implies that<br />
metabolism should be surface-proportional (see von Bertalanffy, 1957, his<br />
Table 6). This is also known as <strong>the</strong> Rubner's surface rule <strong>of</strong> metabolism<br />
(Fletcher, 1974). The relation between body-weight (W) and respiratory<br />
rate (R), which is proportional to metabolic rate in aerobic organisms,<br />
should thus vary as R = α·W β , with β ≈ 0.67 (surface:volume). There is no<br />
indication <strong>of</strong> β for P. lividus in <strong>the</strong> literature but for o<strong>the</strong>r <strong>sea</strong> <strong>urchin</strong>s:<br />
β = 0.620-0.685 (Percy, 1972), β = 0.708-0.866 (Miller & Mann, 1973) for<br />
Strongylocentrotus droebachiensis; β = 0.65 (Webster & Giese, 1975) for<br />
Strongylocentrotus purpuratus. Lawrence & Lane (1982), after<br />
summarizing similar studies on echinoderms in general, concluded: "most<br />
values <strong>of</strong> β […] are between 0.6 and 0.8 regardless <strong>of</strong> body form or<br />
taxonomic group". In a review, Shick (1983) gives a value <strong>of</strong> 0.64 for<br />
echinoids and considers that, among <strong>the</strong> echinoderms, <strong>the</strong>y are closest to<br />
<strong>the</strong> expected value <strong>of</strong> 0.67. It should be noted that <strong>the</strong> prediction <strong>of</strong><br />
β = 0.67 in von Bertalanffy's <strong>the</strong>ory does only account for somatic growth.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
167
It does not consider respiration associated with gonadal growth or<br />
gametogenesis. There are two possibilities for mature individuals: ei<strong>the</strong>r<br />
gonadal growth competes with somatic growth (and <strong>the</strong> later should be<br />
lower than predicted while β remains 0.67) or it just adds to it (and β<br />
should be somewhat larger). Giese et al (1966) measured no difference in<br />
<strong>the</strong> respiration <strong>of</strong> S. purpuratus in function <strong>of</strong> gonad index. This could<br />
indicate a competition between somatic and gonadal growth in <strong>the</strong><br />
presence <strong>of</strong> a limiting factor. However, it should imply a different somatic<br />
growth <strong>model</strong> for juveniles and adults. In <strong>the</strong> present case, a single <strong>model</strong><br />
fits both juvenile and adult stages for <strong>reared</strong> P. lividus. Measurements <strong>of</strong><br />
respiration and a more accurate comparison between both phases are<br />
required to evidence a possible effect <strong>of</strong> maturity on <strong>the</strong> somatic growth<br />
curve. In any case, β-values between 0.6 and 0.8 do not contradict von<br />
Bertalanffy's law. Thus, <strong>the</strong> present study rehabilitates its <strong>the</strong>ory that was<br />
too <strong>of</strong>ten rejected after observation <strong>of</strong> a sigmoidal growth (and now we<br />
know it could result from an inhibition).<br />
Functional analysis <strong>of</strong> <strong>the</strong> constrained parameters<br />
Constraining <strong>the</strong> <strong>model</strong> to <strong>the</strong> origin is very easy, in <strong>the</strong>ory. Most<br />
<strong>model</strong>s in Table 12 and also eq. 29 have a free intercept. It could be<br />
formally expressed: D0 in eq. 29, or it could be hidden in <strong>the</strong><br />
parameterization: a⋅(1 – e bc ) for von Bertalanffy 1, a - d for Weibull, for<br />
<strong>model</strong>s <strong>of</strong> Table 12. Unconstrained intercept means <strong>the</strong> parameter<br />
representing size when <strong>the</strong> growth process initiates is estimated at <strong>the</strong> same<br />
time as all o<strong>the</strong>rs, and is thus influenced by <strong>the</strong>ir values (intercorrelation).<br />
In real life, <strong>the</strong> initial size can influence following growth (for some<br />
experimental studies on P. lividus, see Vaïtilingon et al, 2001). We believe<br />
that a meaningful <strong>model</strong> should follow <strong>the</strong> same logic: initial size is fixed<br />
first and parameters that characterize growth are estimated afterwards.<br />
This is done in eqs. 30-31. Of course, "initial" size just after<br />
metamorphosis is <strong>the</strong> result <strong>of</strong> ano<strong>the</strong>r growth process during larval life but<br />
<strong>the</strong> <strong>model</strong> describes postmetamorphic growth, not larval growth.<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
168
If a <strong>model</strong> does not fit correctly after fixing <strong>the</strong> origin, it means that it<br />
is not adapted to describe growth in this case. The only good reason to<br />
avoid constraint is when time (t0), size (D0) or both are unknown at <strong>the</strong><br />
origin <strong>of</strong> <strong>the</strong> growth process. It is unfortunately common with data<br />
collected in <strong>the</strong> field (Ebert, 1973, 1980a), when it is not possible to<br />
estimate age accurately (Ebert, 1998; Russell & Meredith, 2000). In this<br />
case, only relative growth can be studied and <strong>the</strong> problem <strong>of</strong> origin is thus<br />
eliminated de facto. However, <strong>the</strong> <strong>model</strong> must be reworked to fit relative<br />
growth data.<br />
In <strong>the</strong> present case, we have <strong>the</strong> information necessary to characterize<br />
<strong>the</strong> whole size distribution at <strong>the</strong> origin because we worked in aquaria:<br />
metamorphosis was artificially induced (same t0 for all individuals, see<br />
Grosjean et al, 1998, see Part I), and we have measurements <strong>of</strong> initial sizes<br />
just after it (Grosjean et al, 1996, see Part III). However, by using actual<br />
distribution instead <strong>of</strong> approximating D0 by <strong>the</strong> mean value for all<br />
quantiles, this parameter cannot be eliminated from eq. 31. The <strong>model</strong> is<br />
still viable, and perhaps a little bit more accurate for small sizes (see<br />
pr<strong>of</strong>ile 1 in Fig. 34B). Yet, we preferred to keep <strong>the</strong> simplest <strong>model</strong> in <strong>the</strong><br />
present case.<br />
At <strong>the</strong> o<strong>the</strong>r extreme <strong>of</strong> <strong>the</strong> growth process, a single parameter<br />
characterizes its completion when growth is asymptotic in all <strong>model</strong>s. It is<br />
parameter a in <strong>model</strong>s <strong>of</strong> Table 12 and ∆D∞ in eqs. 29, 31 and 35. Several<br />
authors questioned whe<strong>the</strong>r asymptotic growth is a biological reality, or<br />
just a ma<strong>the</strong>matical artifact. Ricker (1979) wrote a section untitled<br />
"asymptotic growth: is it real?" in a chapter <strong>of</strong> a book; Knight (1968)<br />
devoted a whole article to demonstrate it is a biological non-sense. Some<br />
<strong>model</strong>s with infinite growth appeared (for instance, Tanaka 1982, 1988).<br />
They were also tested on <strong>sea</strong> <strong>urchin</strong>s (Ebert, 1998, 1999; Ebert & Russell,<br />
1993). Some P. lividus were <strong>reared</strong> in our installations for 15 years. They<br />
reached <strong>the</strong>ir maximum size at 4 to 5 years old. They thus kept exactly <strong>the</strong><br />
same size for more than 10 years, proving asymptotic growth is a fact for<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
169
this species. For o<strong>the</strong>r species, where no plateau is observed, lifetime could<br />
be simply too short to reach it. Yet, it is <strong>the</strong>n impossible to tell if growth is<br />
determinate or indeterminate. Anyway, if maximum size is not actually<br />
reached, it is very difficult to estimate <strong>the</strong> corresponding parameter in <strong>the</strong><br />
<strong>model</strong>.<br />
We constrained ∆D∞ to be normally distributed in <strong>the</strong> envelope <strong>model</strong><br />
(eq. 35). It is in agreement with <strong>the</strong> analysis <strong>of</strong> size distributions for fullgrown<br />
animals (Grosjean et al, 1996, see Part III; current dataset). It is also<br />
a consequence <strong>of</strong> <strong>the</strong> genetic homogeneity <strong>of</strong> <strong>the</strong> batch as all individuals<br />
are issued from a single artificial fertilization, i.e., from one male and one<br />
female. In case where D0 is also considered as normally distributed, <strong>the</strong><br />
<strong>model</strong> relates individuals with largest D0 with individuals with largest<br />
∆D∞. But remember <strong>the</strong>se are virtual individuals. This could be <strong>the</strong> case<br />
for real echinoids or not. We cannot verify it without tagging individuals to<br />
track <strong>the</strong>m through time in <strong>the</strong> cohort.<br />
As a consequence <strong>of</strong> fixing k1 (eq. 33), ∆D∞ is <strong>the</strong> only parameter to<br />
contain information on relative growth potential <strong>of</strong> <strong>the</strong> individuals among<br />
<strong>the</strong> cohort in eq. 35. The kinetic parameter k1 could be viewed as<br />
environment-dependent (temperature, food, water quality, etc…). Since<br />
<strong>the</strong>se are <strong>the</strong> same for all animals because <strong>the</strong>y are in <strong>the</strong> same aquarium,<br />
<strong>the</strong>y are fed ad libitum and have access to <strong>the</strong> food <strong>the</strong> same way, it<br />
appears logical to fix k1. Fixing k2 is motivated by a similar reason: we<br />
want it to express one global aspect <strong>of</strong> <strong>the</strong> inhibition. When homogeneous<br />
batches <strong>of</strong> animals <strong>of</strong> same age and same genetic origin are <strong>reared</strong><br />
toge<strong>the</strong>r, speed at which inhibition is released is supposed to be about <strong>the</strong><br />
same for all individuals. This way, only l quantifies changes between<br />
virtual individuals (inhibitors versus inhibited). Of course, many o<strong>the</strong>r<br />
variants are possible, but at <strong>the</strong> cost <strong>of</strong> an increasing complexity <strong>of</strong> <strong>the</strong><br />
<strong>model</strong>.<br />
Indeed, as discussed by Grosjean et al (1996, see Part III), water<br />
quality is not exactly <strong>the</strong> same for all echinoids in culture because <strong>the</strong>y<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
170
tend to form aggregates where a pH gradient is measurable from outside to<br />
inside. Smaller animals are more likely found inside and larger <strong>sea</strong> <strong>urchin</strong>s<br />
outside. This was described as a protective behavior against predators in<br />
<strong>the</strong> field (Tegner & Dayton, 1977; Tegner & Levin, 1983; Levitan &<br />
Genovese, 1989; Ebert, 1998). But <strong>the</strong>n, parallel gradients in both pH and<br />
size distribution could be <strong>the</strong> explanation <strong>of</strong> <strong>the</strong> size-based inhibition <strong>of</strong><br />
growth: a lower pH possibly lowers <strong>the</strong> speed <strong>of</strong> skeletogenesis and,<br />
consequently, <strong>the</strong> growth rate (<strong>the</strong> soma <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s is made <strong>of</strong> ca. 90%<br />
<strong>of</strong> mineralized tissues). No matter what <strong>the</strong> cause <strong>of</strong> <strong>the</strong> inhibition may be,<br />
in such simple experimental conditions, <strong>the</strong> relationship between l and τ<br />
appears amazingly simple. A precise measure <strong>of</strong> <strong>the</strong> pH gradient among<br />
aggregates and <strong>the</strong> quantification <strong>of</strong> skeletogenesis speed with pH drop<br />
should bring some more insight into <strong>the</strong>se relations and causalities.<br />
Lacking such information (<strong>the</strong> only experiment on <strong>sea</strong> <strong>urchin</strong>s growth<br />
related to pH was performed on larvae, Bouxin, 1926), eq. 32 is currently<br />
<strong>the</strong> only way to quantify <strong>the</strong> degree <strong>of</strong> inhibition.<br />
Even in <strong>the</strong> present example, eq. 32 seems to be a very simplified<br />
relationship between l and τ. The 10% smallest animals do not follow it.<br />
Pr<strong>of</strong>ile 2 <strong>of</strong> Fig. 34B at 600 days shows that <strong>the</strong> <strong>model</strong> overestimates <strong>the</strong><br />
smallest fraction (and thus underestimates <strong>the</strong> peak <strong>of</strong> mid-sized animals)<br />
at this age. Adaptation <strong>of</strong> l, or even k2, is perhaps necessary to better fit <strong>the</strong><br />
smallest fraction. Again, <strong>the</strong> simplest possible <strong>model</strong> was presented but<br />
many variations can be conceived.<br />
Relation between l and τ could be even more complex in o<strong>the</strong>r<br />
circumstances: different rearing methods, large and small animals <strong>of</strong><br />
different ages and/or genetic origins maintained toge<strong>the</strong>r, periodic size<br />
sorting in culture, etc. Pr<strong>of</strong>iles <strong>of</strong> l in function <strong>of</strong> τ must be studied in each<br />
particular case. It is also probably very different in <strong>the</strong> field, or for o<strong>the</strong>r<br />
species. The <strong>model</strong> still needs to be reformulated and tested before being<br />
used with field-collected data (mainly cohort separation, or mark-<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
171
ecapture, McDonald & Pitcher, 1979; Baker et al, 1991; Francis, 1995;<br />
Ebert 1999) to confirm it.<br />
An interesting potential <strong>of</strong> this <strong>model</strong>, thanks to variations in <strong>the</strong><br />
relations between l and τ, is <strong>the</strong> possibility <strong>of</strong> predicting growth <strong>of</strong> <strong>the</strong><br />
remaining fraction after elimination <strong>of</strong> largest animals (fisheries or<br />
harvesting <strong>of</strong> largest fraction in aquaculture). Virtual individuals just<br />
below minimum harvesting size suddenly become <strong>the</strong> largest fraction and<br />
will exhibit a very rapid catch up growth to reach <strong>the</strong> maximum growth<br />
speed curve (as evidenced by Grosjean et al, 1996, see Part III). This goes<br />
far beyond <strong>the</strong> scope <strong>of</strong> this paper.<br />
Relations with o<strong>the</strong>r growth <strong>model</strong>s<br />
Several authors have formulated general growth <strong>model</strong>s, <strong>of</strong> which<br />
many o<strong>the</strong>r <strong>model</strong>s are just particular instances. Richards' <strong>model</strong><br />
D = a·(1 – e -b·(t – c) ) d is an extension <strong>of</strong> a von Bertalanffy 1 to a von<br />
Bertalanffy 2 <strong>model</strong> with a variable exponent as an additional parameter d<br />
(Richards, 1959; Ebert, 1980a). Depending on <strong>the</strong> value <strong>of</strong> d, it reduces to<br />
one <strong>of</strong> <strong>the</strong> two von Bertalanffy's <strong>model</strong>s (d = 1 or d = 3), to a logistic (d = -<br />
1), or to a Gompertz curve (|d| → ∞) (Ebert 1999). Schnute (1981), from a<br />
formulation <strong>of</strong> <strong>the</strong> derivative <strong>of</strong> growth rate with time, developed a<br />
sophisticate <strong>model</strong> that contains most o<strong>the</strong>r ones, and also some<br />
unexplored functions. These studies are most useful to show relations<br />
between <strong>model</strong>s that are sometimes hidden by different parameterizations.<br />
For instance, it is hard to tell which is <strong>the</strong> relation between <strong>the</strong> Gompertz<br />
(1825) curve and o<strong>the</strong>r <strong>model</strong>s in Table 12 just by looking at <strong>the</strong>ir<br />
equations.<br />
Starting from von Bertalanffy 1 as <strong>the</strong> simplest asymptotic growth<br />
<strong>model</strong> with no inflexion point, one possible contrasting classification <strong>of</strong><br />
derived <strong>model</strong>s is 'dimensional' versus 'transitional'. A typical<br />
'dimensional' <strong>model</strong> is Richards'. Parameter d is an exponent that changes<br />
<strong>the</strong> dimension <strong>of</strong> <strong>the</strong> value returned by <strong>the</strong> function. Hence, with d = 1, we<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
172
have <strong>the</strong> basic von Bertalanffy 1 <strong>model</strong> that was designed for linear<br />
measurements <strong>of</strong> size (1938, his eq. 26). With d = 3, we have <strong>the</strong> cube <strong>of</strong> a<br />
linear measurement, that is, a volume or a weight (not considering possible<br />
allometries) and Richards <strong>model</strong> reduces to von Bertalanffy <strong>model</strong> 2,<br />
designed for weight measurements. Weibull's <strong>model</strong> (Weibull, 1951)<br />
belongs probably to this category too, though <strong>the</strong> exponent c applies only<br />
to time t. When c = 1, it also reduces to <strong>the</strong> basic von Bertalanffy 1 <strong>model</strong>,<br />
with a slightly different parameterization. The effect <strong>of</strong> <strong>the</strong> exponent,<br />
being d in Richards or c in Weibull, is to transform <strong>the</strong> von Bertalanffy 1<br />
curve into a S-shaped one, or sigmoid, by means <strong>of</strong> a power<br />
transformation.<br />
'Transitional' <strong>model</strong>s fit <strong>the</strong> S-shape as a transition between two states.<br />
The logistic function describes a transition between two constant states<br />
corresponding to its two horizontal asymptotes: D = 0 and D = a. In regard<br />
to <strong>the</strong> results obtained in Table 12, this <strong>model</strong> is not adapted here and it<br />
has no affinity with <strong>the</strong> von Bertalanffy 1 <strong>model</strong>. This <strong>model</strong> was initially<br />
designed to <strong>model</strong> population growth, not individual growth (Verhulst,<br />
1838). The 4-parameter logistic is ano<strong>the</strong>r 'transitional' <strong>model</strong> and we will<br />
demonstrate later its relation with von Bertalanffy 1. With <strong>the</strong> current<br />
parameterization, it also represents a transition between two constant states<br />
materialized by two horizontal asymptotes at D = a and D = d. It fits P.<br />
lividus data very well.<br />
The original growth <strong>model</strong> <strong>of</strong> eq. 29 is a third 'transitional' <strong>model</strong> and<br />
is our missing link as a general equivalent for 'transitional' <strong>model</strong>s to <strong>the</strong><br />
Richards' curve for 'dimensional' <strong>model</strong> (if we except d = -1 and d → ∞<br />
that are physically and biologically meaningless, and thus probably<br />
ma<strong>the</strong>matic artifacts in this context). When l = 0, it reduces to <strong>the</strong> von<br />
Bertalanffy 1 <strong>model</strong>. We now have to demonstrate it is a generalization <strong>of</strong><br />
<strong>the</strong> 4-parameter logistic function. If, in D = d + (a - d)/(1+e -b(t-c) ) we<br />
perform <strong>the</strong> following replacements: t ⇒ t', a ⇒ D0 + ∆D∞, b ⇒ k,<br />
c ⇒ ln(l)/k and d ⇒ D0 – ∆D∞/l, we obtain:<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
173
∆D∞<br />
D0 +∆D∞ − D0 +∆D∞<br />
/ l<br />
Dt (') = D0−<br />
+ ⇔<br />
−kt ⋅ '+ ln( l)<br />
l 1+ e<br />
Dt (') = D +<br />
−kt ⋅ '<br />
∆D∞⋅( −1−l⋅ e + 1 + l)<br />
0 −kt ⋅ '<br />
l⋅ (1+ l⋅e<br />
)<br />
which gives, after fur<strong>the</strong>r simplification:<br />
Dt (') = D +∆D<br />
1−e 1+ l ⋅e<br />
−kt ⋅ '<br />
0 ∞ −kt ⋅ '<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
(38)<br />
(39)<br />
Eq. 39 is equivalent to eq. 29 where k1 = k2 = k. Thus <strong>the</strong> 4-parameter<br />
logistic is ano<strong>the</strong>r parameterization <strong>of</strong> <strong>the</strong> new growth <strong>model</strong> where both<br />
growth speed constants k1 and k2 are equal. With this new<br />
parameterization, reduction to a von Bertalanffy 1 <strong>model</strong> when l = 0 is<br />
now obvious. The 4-parameter logistic <strong>model</strong> also represents a transition<br />
between same sets S and L as our fuzzy <strong>model</strong>, but with k1 = k2. It is a<br />
ma<strong>the</strong>matical coincidence that <strong>the</strong> same <strong>model</strong> represents also, with<br />
ano<strong>the</strong>r parameterization, a transition between two constant states,… a<br />
misleading coincidence as it hides its affinity with <strong>the</strong> von Bertalanffy 1<br />
<strong>model</strong>!<br />
Whe<strong>the</strong>r eq. 29 or eq. 39 is more appropriate to describe P. lividus<br />
growth is hard to tell. From a biological point <strong>of</strong> view, we do not see any<br />
reason why k1 should equal k2, but we perhaps miss it. Without<br />
confidence intervals on parameters, it is not possible to show if k1 is<br />
significantly different <strong>of</strong> k2 for <strong>the</strong> dataset studied (see Fig. 32B).<br />
The distinction between 'dimensional' and 'transitional' <strong>model</strong>s does not<br />
help to explain why one <strong>model</strong> fits <strong>the</strong> data better than ano<strong>the</strong>r. On <strong>the</strong><br />
contrary, both types fit data very well. Indeed, dimension change<br />
('dimensional' <strong>model</strong>s) and inhibition <strong>of</strong> growth ('transitional' <strong>model</strong>s) both<br />
have <strong>the</strong> same effect on <strong>the</strong> shape <strong>of</strong> <strong>the</strong> von Bertalanffy 1 function: <strong>the</strong>y<br />
transform a curve without inflexion point into a sigmoid. P. lividus growth<br />
data are S-shaped for low τ values because <strong>of</strong> an inhibition caused by an<br />
intraspecific competition (according to background knowledge on <strong>the</strong><br />
174
involved processes). It is thus appropriately described by a 'transitional'<br />
<strong>model</strong>. However, by fitting data, only <strong>the</strong> shape that is matched or not by<br />
<strong>the</strong> <strong>model</strong> is considered. Consequently, 'dimensional' <strong>model</strong>s fit equally<br />
well, although <strong>the</strong>ir ma<strong>the</strong>matical formulation does not match biological<br />
observation.<br />
Considering this distinction, we can now formulate a <strong>model</strong> that is both<br />
'dimensional' and 'transitional':<br />
⎛ −k1⋅t' 1−e m<br />
⎞<br />
0 ∞ −k2⋅t' Yt (') = Y+∆Y⎜ ⎟<br />
⎝1+ l ⋅e<br />
⎠<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
(40)<br />
Y being any kind <strong>of</strong> measurement <strong>of</strong> size and m (corresponding to d in <strong>the</strong><br />
Richards <strong>model</strong>) indicating <strong>the</strong> power transformation required to be in <strong>the</strong><br />
best 'dimension <strong>of</strong> growth'. To fur<strong>the</strong>r generalize eq. 40, we could also<br />
replace k1·t' (<strong>the</strong> chronological time modulated by a constant kinetic<br />
parameter) with tM, <strong>the</strong> metabolic –or physiologic– time (Brody, 1937),<br />
using:<br />
t = f( t, x , x ,..., x )<br />
(41)<br />
M 1 2 n<br />
where x1-n are environmental variables that modulate growth, e.g., <strong>sea</strong>son<br />
(Cloern & Nichols, 1978) or temperature (Muller-Feuga, 1990). This<br />
gives:<br />
−t<br />
m<br />
M ⎛ 1−e ⎞<br />
= 0 +∆ ∞ ⎜ −kt ⋅ ⎟ M<br />
Yt (') Y Y<br />
⎝1+ l ⋅e<br />
⎠<br />
(42)<br />
where k = k2/k1. This is a general functional <strong>model</strong> for an asymptotic<br />
growth that derives from <strong>the</strong> von Bertalanffy 1 curve. Therefore, we call it<br />
a generalized von Bertalanffy <strong>model</strong>. This <strong>model</strong> is impossible to fit<br />
without some precautions because <strong>the</strong> two effects, 'dimension' (m) and<br />
'transition' (l and k), are impossible to separate with solely a shape<br />
criterion. From a fitting point <strong>of</strong> view, this <strong>model</strong> is overparameterized<br />
(Draper & Smith, 1998). The dimension parameter m must first be<br />
175
evaluated on basis <strong>of</strong> biological knowledge: determine <strong>the</strong> relation<br />
between <strong>the</strong> dimension <strong>of</strong> growth and that <strong>of</strong> <strong>the</strong> measurement Y that<br />
evaluates it.<br />
But is <strong>the</strong>re a privileged dimension when measuring growth? We have<br />
already asked this question and must come back to it now, because two<br />
concurrent observations suggest that <strong>the</strong> best dimension to describe growth<br />
in <strong>the</strong> case <strong>of</strong> P. lividus is linear. First, using a linear measurement (<strong>the</strong><br />
diameter D), basic shape <strong>of</strong> growth curve, in absence <strong>of</strong> inhibition, is<br />
exactly <strong>the</strong> von Bertalanffy 1 <strong>model</strong>, without inflexion point. With a<br />
weight or a volume to evaluate size, growth <strong>of</strong> <strong>the</strong> largest (not inhibited)<br />
fraction in <strong>the</strong> batch would have followed a more complex von Bertalanffy<br />
2 sigmoidal <strong>model</strong> and it would have been difficult to distinguish <strong>the</strong> Sshape<br />
due to dimension from <strong>the</strong> S-shape due to inhibition. With a linear<br />
measurement, everything is clear: no S-shape means no inhibition and Sshape<br />
means inhibition. Second, <strong>the</strong> diameters are normally distributed<br />
before (at t' = 0) and after <strong>the</strong> growth process (when <strong>the</strong> asymptotic size is<br />
reached that is, above 1600 days, see Fig. 28A). Normal distribution for<br />
linear measurement means that size distribution <strong>of</strong> corresponding weight<br />
or volume measures must be asymmetrical. In this circumstance, <strong>the</strong><br />
envelope <strong>model</strong> (eq. 35) would have been more difficult to fit because<br />
eq. 34, normal distribution <strong>of</strong> ∆Y∞(t), is not correct any more. Hence, <strong>the</strong><br />
preferred dimension to describe growth <strong>of</strong> P. lividus seems linear. Using a<br />
linear measurement <strong>of</strong> size, like <strong>the</strong> diameter D, we are in <strong>the</strong> right<br />
dimension and parameter m in eq. 40 equals one, and thus, <strong>the</strong> <strong>model</strong><br />
reduces to eq. 29. If we had chosen to measure weight, we would have<br />
been in a "wrong dimension" to describe growth and a transformation to<br />
<strong>the</strong> right dimension would imply m ≈ 3 (or more precisely, <strong>the</strong> allometric<br />
coefficient between weight and diameter).<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
176
f. Conclusions<br />
The new growth <strong>model</strong> with intraspecific competition is a very flexible<br />
one. It can accommodate different situations and has meaningful<br />
parameters that allow exploring and quantifying various aspects <strong>of</strong> growth.<br />
Using a quantile regression method, modified for envelope <strong>model</strong>ling, and<br />
constraining parameters ensures <strong>the</strong> meaning <strong>of</strong> <strong>the</strong> latter is saved into <strong>the</strong><br />
fitted <strong>model</strong>. It takes also individual variability into account. This is<br />
particularly useful to <strong>model</strong> growth <strong>of</strong> P. lividus and probably <strong>of</strong> many<br />
o<strong>the</strong>r <strong>sea</strong> <strong>urchin</strong>s species and o<strong>the</strong>r animals or plants. It is original in many<br />
aspects, including <strong>the</strong> way it was designed, by defuzzifying a fuzzy <strong>model</strong><br />
where most <strong>of</strong> <strong>the</strong> o<strong>the</strong>r growth <strong>model</strong>s were built from <strong>the</strong>ir differential<br />
equations. It is a general 'transitional' growth <strong>model</strong>. By distinguishing<br />
'dimensional' and 'transitional' growth <strong>model</strong>s, a duality in sigmoidal<br />
growth curves comes to light. A preferred dimension for <strong>model</strong>ling growth<br />
seems to exist. It is linear in <strong>the</strong> case <strong>of</strong> P. lividus but it should be most<br />
interesting to check it for o<strong>the</strong>r species.<br />
A generalized von Bertalanffy growth <strong>model</strong>, which is both<br />
'dimensional' and 'transitional' and includes varying environmental effects<br />
on growth, thanks to <strong>the</strong> use <strong>of</strong> metabolic time, was proposed (eq. 42). It<br />
is, however, <strong>the</strong> visible tip <strong>of</strong> <strong>the</strong> iceberg. Fuzzy sets and transitions<br />
(membership functions) can be combined in countless ways to create many<br />
o<strong>the</strong>r similar <strong>model</strong>s. In <strong>the</strong> present work, we studied <strong>model</strong>s deriving<br />
from von Bertalanffy 1 curve, because <strong>the</strong> latter seemed to be a good basis<br />
for describing growth <strong>of</strong> P. lividus. O<strong>the</strong>r <strong>model</strong>s incorporating an<br />
inhibition component or any o<strong>the</strong>r 'transitional' feature can be derived from<br />
o<strong>the</strong>r growth <strong>model</strong>s, including non-asymptotic ones, and would perhaps<br />
be more adapted for o<strong>the</strong>r species. We propose to call this family <strong>of</strong><br />
functions 'fuzzy-remanent' <strong>model</strong>s. From <strong>the</strong>ir fuzzy origin, <strong>the</strong>y keep<br />
nothing in appearance, but <strong>the</strong> biological meaning <strong>of</strong> <strong>the</strong>ir parameters is<br />
still <strong>the</strong>re. Recalling <strong>the</strong> fuzzy <strong>model</strong> <strong>the</strong>y come from, one has a much<br />
clearer idea <strong>of</strong> how various components –sets and membership functions–<br />
interact to produce <strong>the</strong> final result. Fuzzy logic is closer to <strong>the</strong> way human<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
177
ain conceptualizes complex objects. Statistical tools handle defuzzified<br />
analytic functions more conveniently. By <strong>the</strong>ir bivalence, fuzzy-remanent<br />
functions promise to be powerful tools for <strong>model</strong>ling complex nonlinear<br />
phenomena, like growth, in a functional way.<br />
g. Acknowledgments<br />
We thank <strong>the</strong> CREC and <strong>the</strong> University <strong>of</strong> Caen for <strong>the</strong>ir contribution<br />
in building a specific <strong>sea</strong> <strong>urchin</strong> rearing facility. We are grateful to Didier<br />
Bucaille who performed <strong>the</strong> laboratory measurements. We thank also <strong>the</strong><br />
Pr<strong>of</strong>. Michael Russell for constructive criticisms and Pr<strong>of</strong>. Jean-Pierre Van<br />
Noppen for pro<strong>of</strong>reading <strong>the</strong> manuscript. This study was conducted in <strong>the</strong><br />
framework <strong>of</strong> <strong>the</strong> European Contracts AQ2.530, "Sea <strong>urchin</strong>s cultivation"<br />
and FAIR-CT96-1623, "Biology <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s under intensive cultivation<br />
(closed cycle echiniculture)". This is a contribution <strong>of</strong> <strong>the</strong> "Centre<br />
Interuniversitaire de Biologie Marine".<br />
Part IV: A growth <strong>model</strong> with intraspecific competition<br />
178
General conclusions<br />
179
180
General conclusions<br />
GENERAL CONCLUSIONS<br />
By focusing on variability and interactions in individual growth <strong>of</strong> <strong>the</strong><br />
<strong>reared</strong> <strong>sea</strong> <strong>urchin</strong> <strong>Paracentrotus</strong> lividus, we raised questions on <strong>the</strong><br />
adequacy <strong>of</strong> existing <strong>model</strong>s and methods. To resolve <strong>the</strong>se problems we<br />
developed a new growth <strong>model</strong> with incorporating interspecific<br />
competition by defuzzifying a fuzzy <strong>model</strong> and a quantile regression<br />
method was adapted to account for individual variability (envelope<br />
<strong>model</strong>ling). The <strong>model</strong> appears to be an appropriate functional description<br />
<strong>of</strong> <strong>the</strong> process, as it is in agreement with all experimental results. It allows<br />
quantifying <strong>the</strong> degree <strong>of</strong> growth inhibition in <strong>the</strong> P. lividus echinoids in<br />
cultivation.<br />
Similarly, one should question <strong>the</strong> validity <strong>of</strong> growth <strong>model</strong>s, <strong>of</strong> fitting<br />
methods and <strong>of</strong> calculation <strong>of</strong> size at age (growth ring analysis, sizefrequency<br />
data analysis, mark and recapture) in all studies on ei<strong>the</strong>r <strong>reared</strong><br />
or wild echinoids. Yet, if <strong>the</strong>re is some interaction between individuals or<br />
if individual variability is large (as both can be suspected in most if not all<br />
cases), all <strong>the</strong>se methods could lead to biased estimations <strong>of</strong> growth and,<br />
consequently, to erroneous inferences about population dynamics.<br />
Adequate tools remain to be developed for field data where age is not<br />
measurable without error. A modification <strong>of</strong> <strong>the</strong> new <strong>model</strong> for relative<br />
growth rate data would be a logical starting point.<br />
The new <strong>model</strong> clearly has application both in <strong>sea</strong> <strong>urchin</strong> aquaculture<br />
and in fisheries management. Indeed, <strong>the</strong> experiment with mixed Ff and Fg<br />
batches (see Part III, p. 130) indicated a great growth potential <strong>of</strong> smaller,<br />
inhibited individuals in a heterogeneous population. Yield per surface unit<br />
should improve in cultivation when using mixed batches because <strong>the</strong>y are<br />
almost as productive as small plus large batches <strong>reared</strong> separately and thus,<br />
on <strong>the</strong> double <strong>of</strong> <strong>the</strong> surface. If <strong>the</strong> mechanism <strong>of</strong> inhibition/catch up<br />
growth also occurs in <strong>the</strong> field, a bimodal size distribution could be <strong>the</strong><br />
most efficient configuration to maintain a wild population <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s.<br />
When <strong>the</strong> largest fraction <strong>of</strong> <strong>the</strong> population is harvested, some mid-sized<br />
181
General conclusions<br />
individuals, whose inhibition is suddenly eliminated, could quickly replace<br />
missing adults. Two conditions should be met, however, to obtain this<br />
result on <strong>the</strong> long term. First, mortality <strong>of</strong> small and mid-sized individuals<br />
should not increase when large adults are removed (indeed, when<br />
removing large adults, <strong>the</strong> passive protection <strong>of</strong> small individuals against<br />
predators disappears). If necessary, part <strong>of</strong> <strong>the</strong> adults should be left in<br />
place during harvesting. Second, recruitment should not be a limiting<br />
factor. By harvesting large adults before <strong>the</strong>y spawn, recruitment is de<br />
facto lowered. An artificial production <strong>of</strong> a large amount <strong>of</strong> seed in<br />
hatcheries is one way to maintain recruitment levels. Beyond <strong>the</strong>se general<br />
considerations, it is difficult to define rules for sustainable fishery<br />
practices. If <strong>the</strong> growth <strong>model</strong> with intraspecific competition were<br />
calibrated against field population data, it would be possible to quantify<br />
<strong>the</strong> impact <strong>of</strong> various fishery methods, and to provide objective criteria for<br />
sustainable <strong>sea</strong> <strong>urchin</strong> fisheries (Grosjean & Jangoux, 2000).<br />
From a <strong>the</strong>oretical point <strong>of</strong> view, <strong>the</strong> new growth <strong>model</strong> rehabilitates a<br />
60-year old <strong>the</strong>ory <strong>of</strong> growth elaborated by von Bertalanffy (1938). Since<br />
<strong>the</strong>n, several authors have questioned its validity (Knight, 1968; R<strong>of</strong>f,<br />
1980; Frontier & Pichot-Viale, 1993). Many o<strong>the</strong>r works have indicated<br />
that <strong>the</strong> von Bertalanffy 1 <strong>model</strong> is probably not acceptable to describe <strong>the</strong><br />
growth <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s (Gage & Tyler, 1985; Gage et al, 1986; Gage, 1987;<br />
Dafni, 1992; Ebert, 1980a; Ebert & Russell, 1993; Lamare & Mladenov,<br />
2000), including P. lividus (Cellario & Fenaux, 1990; Turon et al, 1995).<br />
Here we have shown that sigmoidal growth does not necessarily means<br />
that von Bertalanffy's <strong>the</strong>ory is invalid. An S-shape could result from<br />
inhibition <strong>of</strong> growth at small sizes/ages. P. lividus follows <strong>the</strong> von<br />
Bertalanffy's law for its somatic growth when it is not inhibited. In a<br />
cohort, <strong>the</strong> non-inhibited fraction amounts for less than 10% <strong>of</strong> all <strong>the</strong><br />
individuals. Using least-square regression leads to <strong>the</strong> rejection <strong>of</strong> <strong>the</strong> von<br />
Bertalanffy 1 <strong>model</strong>. Using quantile regression validates it for <strong>the</strong> largest<br />
fraction. Using an envelope <strong>model</strong> with intraspecific competition<br />
182
General conclusions<br />
component validates it as <strong>the</strong> basic process for <strong>the</strong> whole cohort and shows<br />
how intraspecific competition delays actual growth.<br />
<strong>Growth</strong> is <strong>the</strong> result <strong>of</strong> many complex mechanisms that interact:<br />
feeding, digestion, respiration, building up <strong>of</strong> new somatic tissues,<br />
maintenance, reproduction, etc. A general consensus is that growth is too<br />
complicated and could only be reliably described by complex <strong>model</strong>s.<br />
Indeed, it is surprising that a simple 2-parameter <strong>model</strong> like von<br />
Bertalanffy 1 (D(t) = D∞·[1 – e -k·t ]) could represent individual growth <strong>of</strong><br />
many organisms. It is also surprising that, considering interactions and<br />
individual variability in addition to <strong>the</strong> basic processes, a quite simple 5parameter<br />
<strong>model</strong> (our envelope curve, eq. 35 p 160) represents growth <strong>of</strong> a<br />
whole cohort <strong>of</strong> <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s.<br />
Most <strong>of</strong> <strong>the</strong> simple <strong>model</strong>s <strong>of</strong> <strong>the</strong> first half <strong>of</strong> <strong>the</strong> twentieth century,<br />
attempt to interpret, in a functional way, ei<strong>the</strong>r individual growth (von<br />
Bertalanffy, 1938, 1957; Brody, 1945) or size/shape at age (Huxley, 1932;<br />
Teissier, 1934, 1948; d'Arcy Thompson, 1961). In <strong>the</strong> last half <strong>of</strong> <strong>the</strong><br />
century, <strong>the</strong>se <strong>model</strong>s have been replaced by more complex, but purely<br />
descriptive and/or speculative ones (Richards, 1959; Schnute, 1981;<br />
Tanaka, 1982, 1988; Jolicoeur, 1985). Clearly, it is worth revisiting old<br />
concepts using new tools, e.g., fuzzy logic or quantile regression. Such<br />
revisitation was accomplished for instance by van Osselaer & Grosjean<br />
(2000) in <strong>the</strong>ir study <strong>of</strong> <strong>the</strong> suture <strong>of</strong> coiled shells reflecting <strong>the</strong> ontogeny<br />
<strong>of</strong> molluscs. Despite <strong>the</strong> conclusion <strong>of</strong> Tursh (1998) that <strong>the</strong> shape <strong>of</strong> <strong>the</strong><br />
suture is too complex to be <strong>model</strong>led with a simple equation, <strong>the</strong>y<br />
demonstrated that a simple 4-parameter helicospiral <strong>model</strong> was <strong>the</strong> best<br />
descriptor for most coiled shells and that some methodological errors<br />
prevailed when using much more complex 8- to 16-parameter <strong>model</strong>s (for<br />
a review, see Stone, 1996). This example demonstrates ano<strong>the</strong>r case where<br />
growth is less complex in reality than in <strong>the</strong>ory. The discipline (call it<br />
"ontogenology") <strong>of</strong> describing individual growth or ontogeny with <strong>model</strong>s<br />
that are both reasonably simple and functional may reveal one day how<br />
183
Root <strong>model</strong>s:<br />
y’ = ay m -by n<br />
(exponential,<br />
von Bertal. 1,<br />
logistic)<br />
General conclusions<br />
various mechanisms constrain "organic growth" <strong>of</strong> metazoans. These<br />
constraints, <strong>model</strong>s, and "rules <strong>of</strong> growth and development" may turn out<br />
to be simpler than expected.<br />
As a final conclusion, we propose an original classification <strong>of</strong> growth<br />
<strong>model</strong>s based on <strong>the</strong>ir functional features ra<strong>the</strong>r than <strong>the</strong>ir pure ma<strong>the</strong>matic<br />
affinities (though similar functions are <strong>of</strong>ten expressed by similar<br />
equations). In this typology, general growth <strong>model</strong>s derive from simpler<br />
predecessors thanks to one or more additional features having a biological<br />
(or physical) meaning (Fig. 35).<br />
transitional<br />
dimensional<br />
metabolic time<br />
diauxic<br />
Preece-Baines<br />
polyphasic<br />
4-p. logistic Fuzzy-reman.<br />
von Bertal. 2 Richa rds<br />
Weibull?<br />
Seasonal VB<br />
T(°C) VB<br />
General. logis.<br />
monophasic<br />
von Bertal.<br />
with<br />
t M = f(t, x i ..)<br />
Go mpe rtz<br />
?<br />
Johnson<br />
Generalized<br />
von Bertal.<br />
with t M<br />
Jolicoeur<br />
?<br />
Tanaka<br />
…<br />
Figure 35. A classification <strong>of</strong> growth <strong>model</strong>s based on <strong>the</strong>ir functional features (see text for<br />
fur<strong>the</strong>r explanations).<br />
184
General conclusions<br />
In this classification, all <strong>model</strong>s derive from simplest forms described<br />
by <strong>the</strong> general differential equation y' = ay m - by n , that is, a damped<br />
exponential growth with <strong>the</strong> first term being <strong>the</strong> limiting factor and <strong>the</strong><br />
second one representing <strong>the</strong> exponential growth (Fletcher, 1974; Mueller-<br />
Feuga, 1990). The various forms <strong>of</strong> basic growth <strong>model</strong>s are obtained from<br />
different particular values for m and n. With m = 1 and n = 1, we obtain <strong>the</strong><br />
exponential growth, which is a simple indeterminate growth <strong>model</strong>. If<br />
m = 0 and n = 1, we get <strong>the</strong> von Bertalanffy 1 <strong>model</strong> (von Bertal. 1),<br />
which seems to be <strong>the</strong> simplest determinate growth <strong>model</strong> for individuals.<br />
If m = 1 and n = 2, we obtain <strong>the</strong> logistic curve, which is probably <strong>the</strong><br />
simplest determinate growth <strong>model</strong> for populations (Verhulst, 1838). It is<br />
not clear if o<strong>the</strong>r solutions <strong>of</strong> <strong>the</strong> differential equation could be considered<br />
as useful roots in <strong>the</strong> classification. Some solutions correspond to more<br />
complex <strong>model</strong>s that are best placed in a stem instead <strong>of</strong> as a root in this<br />
classification. For instance, using m = 2/3 and n = 1, we obtain <strong>the</strong> von<br />
Bertalanffy 2 <strong>model</strong> that we prefer to position inside <strong>the</strong> "von Bertalanffy<br />
1 family" (see Fig. 35). Some authors have generalized this differential<br />
equation (Fletcher, 1974; Schnute, 1981) to a point that it describes almost<br />
all existing <strong>model</strong>s. Our concern here is just to derive <strong>the</strong> simplest <strong>model</strong>s<br />
as roots <strong>of</strong> our classification tree and consider a di- or a polychotomous<br />
system to position more complex <strong>model</strong>s in <strong>the</strong> tree as generalizations <strong>of</strong><br />
<strong>the</strong> simplest root <strong>model</strong>s. One should consider each <strong>of</strong> <strong>the</strong> simplest <strong>model</strong>s<br />
as <strong>the</strong> root <strong>of</strong> a distinct tree. However, in our presentation (Fig. 35), we<br />
mixed all existing <strong>model</strong>s in a single tree for conciseness (because, except<br />
for <strong>the</strong> "von Bertalanffy 1 family", <strong>the</strong> trees would not been much<br />
populated).<br />
Starting from those simplest <strong>model</strong>s, one could consider a first<br />
dichotomous separation: monophasic versus polyphasic <strong>model</strong>s. While<br />
<strong>the</strong> group <strong>of</strong> monophasic <strong>model</strong>s is more populated (because <strong>the</strong>y probably<br />
represent more common growth processes), <strong>the</strong> second group contains two<br />
items:<br />
185
General conclusions<br />
- The diauxic growth <strong>model</strong> <strong>of</strong> Liquori et al (1981). This is a biphasic<br />
<strong>model</strong> used to describe <strong>the</strong> increase in cell numbers in an embryo that<br />
results in a slow and a fast division processes.<br />
- The Preece-Baines (1978) <strong>model</strong> that describes human growth (and<br />
perhaps also <strong>the</strong> growth <strong>of</strong> some o<strong>the</strong>r mammals) that we presented in<br />
<strong>the</strong> introduction, p. 51.<br />
In <strong>the</strong> monophasic growth group, we have various sub-groups that<br />
correspond each to a distinct feature added to <strong>the</strong> basic <strong>model</strong>. In Part IV,<br />
we discussed <strong>the</strong> two antagonist features that transform a von Bertalanffy 1<br />
<strong>model</strong> without inflexion point into a sigmoid: <strong>the</strong> dimensional and<br />
transitional sub-groups. Models in <strong>the</strong> dimensional sub-group account for<br />
a change in <strong>the</strong> dimension <strong>of</strong> size measurement (length, surface or<br />
volume/weight) thanks to an additional parameter m. Von Bertalanffy 2<br />
<strong>model</strong> (von Bertal. 2) is a particular case with m = 3 and Richards <strong>model</strong><br />
is <strong>the</strong> general equation <strong>of</strong> this sub-group. In <strong>the</strong> transitional sub-group,<br />
<strong>the</strong>re is an inhibition <strong>of</strong> growth that is progressively released with age.<br />
This is probably <strong>the</strong> kingdom <strong>of</strong> 'fuzzy-remanent functions' as <strong>the</strong>y are<br />
convenient descriptors for transitions. Both a reparameterized form <strong>of</strong> <strong>the</strong><br />
4-parameter logistic function (4-p. logistic, eq. 39, p. 174) and <strong>the</strong> <strong>model</strong><br />
with intraspecific competition component that we propose here for<br />
<strong>model</strong>ling <strong>the</strong> growth <strong>of</strong> <strong>reared</strong> P. lividus (Fuzzy-reman., eq. 29, p. 152)<br />
belong to this category. The former being a special case <strong>of</strong> <strong>the</strong> latter, with<br />
both kinetic parameters k1 and k2 being equal.<br />
Ano<strong>the</strong>r sub-group can be created for <strong>model</strong>s that incorporate <strong>the</strong> effect<br />
<strong>of</strong> environmental variation on growth, through <strong>the</strong> concept <strong>of</strong> metabolic<br />
time. One such <strong>model</strong> was proposed by Cloern & Nichols (1978) for<br />
considering <strong>sea</strong>sonal variations (as a global summary <strong>of</strong> various<br />
environmental variables with a <strong>sea</strong>sonal cycle like temperature,<br />
photoperiod, food availability…) in <strong>the</strong> von Bertalanffy 1 growth <strong>model</strong><br />
(Seasonal VB). Mueller-Feuga (1990) proposed ano<strong>the</strong>r <strong>model</strong> <strong>of</strong> this<br />
group which account for temperature only (T(°C) VB). Both <strong>of</strong> <strong>the</strong>se<br />
186
General conclusions<br />
<strong>model</strong>s obviously belong to a more general family where metabolic time tM<br />
is a function <strong>of</strong> time t as well as several o<strong>the</strong>r environmental<br />
(meta)variables [von Bertal. with tM = f(t, xi…)].<br />
There are perhaps o<strong>the</strong>r unidentified sub-groups. The Weibull <strong>model</strong> is<br />
a generalization <strong>of</strong> von Bertalanffy 1 with an exponent applied to time t.<br />
As such, it could belong to both <strong>the</strong> dimensional and <strong>the</strong> metabolic time<br />
sub-groups. We do not see any functional meaning <strong>of</strong> <strong>the</strong>m, but it could<br />
exist. A generalized logisitic <strong>model</strong> was proposed by Nelder (1961) and<br />
Turner et al (1969). It is a logistic function where an additional parameter<br />
m is applied as a global exponent. From its analytic form, it is a<br />
dimensional <strong>model</strong>, which makes sense only when it is applied to<br />
individual growth. Turner (1969) indicates that, when applied to<br />
populations, this <strong>model</strong> accounts for growth with a maximum population<br />
size that is allowed to vary. It this context, it may belong to ano<strong>the</strong>r<br />
unidentified sub-group.<br />
The generalized von Bertalanffy with tM (eq. 42, p. 175) is at <strong>the</strong><br />
same time a 'dimensional', a 'transitional' and a 'metabolic time' <strong>model</strong>. It<br />
represents <strong>the</strong> highest level <strong>of</strong> generalization in <strong>the</strong> "von Bertalanffy 1"<br />
family but it spans a large space for deriving o<strong>the</strong>r kinds <strong>of</strong> <strong>model</strong>s.<br />
Finally, <strong>the</strong>re are some unclassifiable <strong>model</strong>s, such as Gompertz,<br />
Johnson, Jolicoeur and Tanaka. Ei<strong>the</strong>r <strong>the</strong>y are purely descriptive<br />
<strong>model</strong>s that just mimic <strong>the</strong> shape <strong>of</strong> some functional <strong>model</strong>s, or <strong>the</strong>ir<br />
affinity is not established yet. In <strong>the</strong> first case, <strong>the</strong>y have clearly no place<br />
in <strong>the</strong> proposed functional classification, as it is <strong>the</strong> case for o<strong>the</strong>r purely<br />
descriptive <strong>model</strong>s (e.g., Rao's polynomial growth <strong>model</strong>; Rao, 1965;<br />
Basu, 1999). In <strong>the</strong> o<strong>the</strong>r case, a reparameterization or a good example <strong>of</strong> a<br />
functional use is probably required to reveal <strong>the</strong>ir real nature.<br />
Much space is left empty in this typology for adding new functional<br />
<strong>model</strong>s when growth <strong>of</strong> o<strong>the</strong>r animals and plants will be described in a<br />
functional way. Such a classification should help choosing <strong>the</strong> right growth<br />
187
General conclusions<br />
<strong>model</strong> not on a shape criterion (does it fit data according to <strong>the</strong> R 2 , or sum<br />
<strong>of</strong> square <strong>of</strong> residuals, or to an analysis <strong>of</strong> residuals, or to a visual<br />
inspection on a graph), but on its ability to represent at best underlying<br />
processes <strong>of</strong> growth, as <strong>the</strong>y are experimentally evidenced.<br />
188
References<br />
REFERENCES<br />
Abe E., M. Kakiuchi, K. Matsuyama & T. Kaneko, 1983. Seasonal variations in<br />
growth and chemical components in <strong>the</strong> blade <strong>of</strong> Laminaria religiosa<br />
Miyabe in Oshoro Bay, Hokkaido. Sci. Rep. Hokkaido Fish. Exp. St.,<br />
25:47-60.<br />
Agatsuma, Y. & Y. Sugawara, 1988. Reproductive cycle and food ingestion <strong>of</strong><br />
<strong>the</strong> <strong>sea</strong> <strong>urchin</strong>, Strongylocentrotus nudus (A. Agassiz), in Sou<strong>the</strong>rn<br />
Hokkaido. II. Seasonal changes <strong>of</strong> <strong>the</strong> gut content and test weight. Sci.<br />
Rep. Hokkaido Fish. Exp. St., 30:43-49.<br />
Aksland, M., 1994. A general cohort analysis method. Biometrics, 50:917-932.<br />
Allain, J.-Y, 1971. Note sur la pêche et la commercialisation des oursins en<br />
Bretagne nord. Trav. Labo. Biol. Halieut. Univ. Rennes, 5:59-69.<br />
Allain, J.-Y., 1972a. Structure des populations de <strong>Paracentrotus</strong> lividus<br />
(Lamarck) (Echinodermata, Echinoidea) soumises à la pêche sur les côtes<br />
nord de Bretagne. Rev. Trav. Inst. Pêches Marit., 39(2):171-212.<br />
Allain, J.-Y., 1972b. La pêche aux oursins dans le monde. La Pêche Maritime,<br />
1133:625-630.<br />
Allain, J.-Y., 1975. Structure des populations de <strong>Paracentrotus</strong> lividus soumises à<br />
la pêche sur les côtes nord de Bretagne. Rev. Trav. Inst. Pêches Marit.,<br />
39(2):171-212.<br />
Allain J.-Y., 1978. Age et croissance de <strong>Paracentrotus</strong> lividus (Lmk) et de<br />
Psammechinus miliaris (Gmelin) des côtes nord de Bretagne<br />
(Echinoidea). Cah. Biol. Mar., 19(1):11-21.<br />
Allen, R.K., 1966. A method <strong>of</strong> fitting growth curves <strong>of</strong> <strong>the</strong> von Bertalanffy type<br />
to observed data. J. Fish. Res. Board Can., 23(2):163-179.<br />
Bacher, C., M. Héral, J.M. Deslous-Paoli & D. Razet, 1991. Modèle énergétique<br />
uniboite de la croissance des huîtres (Crassostrea gigas) dans le bassin de<br />
Marennes-Oléron. Can. J. Fish. Aquat. Sci., 48:391-404.<br />
189
References<br />
Baker, T.T., R. Lafferty & T.J. Quinn II, 1991. A general growth <strong>model</strong> for markrecapture<br />
data. Fish. Res., 11:257-281.<br />
Basu, T.K., 1999. Application <strong>of</strong> Rao's polynomial growth curve <strong>model</strong> to<br />
analyse <strong>the</strong> length growth data <strong>of</strong> a freshwater major carp <strong>of</strong> India, Catla<br />
catla (Ham) during <strong>the</strong> early development period. Fish. Res., 42:303-307.<br />
Basuyaux, O. & J.L. Blin, 1998. Use <strong>of</strong> maize as a food source for <strong>sea</strong> <strong>urchin</strong>s in<br />
a recirculating rearing system. Aqua. International, 6(3):233-247.<br />
Bhattacharya, C.G., 1967. A simple method <strong>of</strong> resolution <strong>of</strong> a distribution into<br />
Gaussian components. Biometrics, 23:115-135.<br />
Blin, J.-L., 1997. Culturing <strong>the</strong> purple <strong>sea</strong> <strong>urchin</strong>, <strong>Paracentrotus</strong> lividus, in a<br />
recirculation system. Bul. Aqua. Assoc. Canada, 97(1):8-13.<br />
Botsford, L.W., B.D. Smith & J.F. Quinn, 1994. Bimodality in size distributions:<br />
<strong>the</strong> red <strong>sea</strong> <strong>urchin</strong> Strongylocentrotus franciscanus as an example. Ecol.<br />
Applic., 4:42-50.<br />
Bouxin, H., 1926. Action des acides sur le squelette des larves de l'oursin<br />
<strong>Paracentrotus</strong> lividus. Influence du pH. C. R. Soc. Biol., 94:453-455.<br />
Branch, G.M., 1974. Intraspecific competition in Patella cochlear Born. J. Anim.<br />
Biol., 44:263-281.<br />
Brody, S., 1937. Relativity <strong>of</strong> physiologic time and physiologic weight. <strong>Growth</strong>,<br />
1:60-67.<br />
Brody, S., 1945. Bioenergetics and growth. Reinhold, New York.<br />
Bull, H.O., 1938. The growth <strong>of</strong> Psammechinus miliaris (Gmelin) under<br />
aquarium conditions. Rep. Dove Mar. Lab., 3(6):39-42.<br />
Campbell, A. & R.M. Harbo, 1991. The <strong>sea</strong> <strong>urchin</strong> fisheries in British Columbia,<br />
Canada. In: Y. Yanagisawa et al (eds). Biology <strong>of</strong> Echinodermata.<br />
Balkema, Rotterdam. Pp. 191-199.<br />
Causton, D.R., 1969. A computer program for fitting <strong>the</strong> Richards function.<br />
Biometrics, 25:401-409.<br />
190
References<br />
Cellario, Ch. & L. Fenaux, 1990. <strong>Paracentrotus</strong> lividus (Lamarck) in culture<br />
(larval and benthic phases): Parameters <strong>of</strong> growth observed during two<br />
years following metamorphosis. Aquaculture, 84:173-188.<br />
Chiu, S.T., 1990. Age and growth <strong>of</strong> Anthocidaris crassispina (Echinodermata:<br />
Echinoidea) in Hong Kong. Bull. Mar. Sci., 47(1):94-103.<br />
Claerebout, M. & M. Jangoux, 1985. Conditions de digestion et activité de<br />
quelques polysaccharidases dans le tube digestif de l'oursin <strong>Paracentrotus</strong><br />
lividus (Echinodermata). Biochem. Syst. Ecol., 13(1):51-54.<br />
Cloern, J.E. & F.. Nichols, 1978. A von Bertalanffy growth <strong>model</strong> with a<br />
<strong>sea</strong>sonally varying coefficient. J. Fish. Res. Board Can., 35:1479-1482.<br />
Comely, C.A. & A.D. Ansell, 1988. Population density and growth <strong>of</strong> Echinus<br />
esculentus L. on <strong>the</strong> Scottish West Coast. Estuar. Coast. Shelf Sci.,<br />
27:311-334.<br />
Conand, C. & N.A. Sloan, 1989. World fisheries for echinoderms. In: J.F. Caddy<br />
(ed.). Marine invertebrate fisheries: <strong>the</strong>ir assessment and management.<br />
Wiley & Sons, New York. Pp. 647-663.<br />
Cox, E., 1999. The fuzzy systems handbook, 2 nd ed. Academic Press, San Diego.<br />
Cox, R.A., M.J. McCartney & F. Culkin, 1970. The specific gravity/salinity/<br />
temperature relationship in natural <strong>sea</strong> water. Deep-Sea Res., 17:679-689.<br />
Crapp, G.B. & M.E. Willis, 1975. Age determination in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
<strong>Paracentrotus</strong> lividus (Lamarck), with notes on <strong>the</strong> reproductive cycle. J.<br />
Exp. Mar. Biol. Ecol., 20:157-178.<br />
Crook, A.C., E. Verling & D.K.A. Barnes, 1999. Comparative study <strong>of</strong> <strong>the</strong><br />
covering reaction <strong>of</strong> <strong>the</strong> purple <strong>sea</strong> <strong>urchin</strong>, <strong>Paracentrotus</strong> lividus, under<br />
laboratory and fied conditions. J. Mar. Biol. Assoc. U.K., 79(6):1117-<br />
1121.<br />
Dafni, J., 1992. <strong>Growth</strong> rate <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Tripneustes gratilla elatensis. Israel<br />
J. Zool., 38:25-33.<br />
191
References<br />
Dafni, J. & R. Tobol, 1987. Population structure patterns <strong>of</strong> a common Red Sea<br />
echinoid (Tripneustes gratilla elatensis). Israel J. Zool., 34:191-204.<br />
d'Arcy Thompson, W., 1961. On growth and form, 4 th ed. Cambridge University<br />
Press, Cambridge.<br />
de Jong-Westman, M., B.E. March & T.H. Carefoot, 1995a. The effect <strong>of</strong><br />
different nutrient formulations in artificial diets on gonad growth in <strong>the</strong><br />
<strong>sea</strong>-<strong>urchin</strong> Strongylocentrotus droebachiensis. Can. J. Zool., 73:1495-<br />
1502.<br />
de Jong-Westman, M., P.-Y. Quian, B.E March & T.H. Carefoot, 1995b.<br />
Artificial diets in <strong>sea</strong> <strong>urchin</strong> culture: effects <strong>of</strong> dietary protein level and<br />
o<strong>the</strong>r additives on egg quality, larval morphometrics, and larval survival<br />
in <strong>the</strong> green <strong>sea</strong> <strong>urchin</strong>, Strongylocentrotus droebachiensis. Can. J. Zool.,<br />
73:2080-2090.<br />
Draper, N.R. & H. Smith, 1998. Applied regression analysis, 3 rd ed. Whiley &<br />
Sons, New York.<br />
Duineveld, G.C.A. & M.I. Jenness, 1984. Differences in growth rates <strong>of</strong> <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong> Echinocardium cordatum as estimated by <strong>the</strong> parameter ω <strong>of</strong> <strong>the</strong><br />
von Bertalanffy equation applied to skeletal rings. Mar. Ecol. Prog. Ser.,<br />
19:65-72.<br />
Durham, J.W., 1955. Classification <strong>of</strong> Clypeasteroid echinoids. University <strong>of</strong><br />
California, Museum <strong>of</strong> Paleontology. Pp. 73-192.<br />
Ebert, T.A., 1967. Negative growth and longevity in <strong>the</strong> purple <strong>sea</strong> <strong>urchin</strong><br />
Strongylocentrotus purpuratus (Stimpson). Science, 157(3788):557-558.<br />
Ebert, T.A., 1968. <strong>Growth</strong> rates <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Strongylocentrotus purpuratus<br />
related to food availability and spine abrasion. Ecology, 49:1075-1091.<br />
Ebert, T.A., 1973. Estimating growth and mortality rates from size data.<br />
Oecologia, 11:281-298.<br />
Ebert, T.A., 1975. <strong>Growth</strong> and mortality <strong>of</strong> post-larval echinoids. Amer. Zool.,<br />
15:755-775.<br />
192
References<br />
Ebert, T.A., 1977. An experimental analysis <strong>of</strong> <strong>sea</strong> <strong>urchin</strong> dynamics and<br />
community interactions on a rock jetty. J. Exp. Mar. Biol. Ecol., 27:1-22.<br />
Ebert, T.A., 1980a. Estimating parameters in a flexible growth equation, <strong>the</strong><br />
Richards function. Can. J. Fish. Aquat. Sci., 37(4):687-692.<br />
Ebert, T.A., 1980b. Relative growth <strong>of</strong> <strong>sea</strong> <strong>urchin</strong> jaws: an example <strong>of</strong> plastic<br />
response allocation. Bull. Mar. Sci., 30(2):467-474.<br />
Ebert, T.A., 1981. Estimating mortality from growth parameters and a size<br />
distribution when recruitment is periodic. Limnol. Oceanogr., 26:764-769.<br />
Ebert, T.A., 1982. Longevity, life history, and relative body wall size in <strong>sea</strong><br />
<strong>urchin</strong>s. Ecol. Monographs, 52(4):353-394.<br />
Ebert, T., 1983. Recruitment in echinoderms. In: M. Jangoux & J.M. Lawrence<br />
(eds). Echinoderm studies , vol. 1. Balkema, Rotterdam. Pp. 169-203.<br />
Ebert, T.A., 1985. Sensitivity <strong>of</strong> fitness to macroparameter changes: an analysis<br />
<strong>of</strong> survivorship and individual growth in <strong>sea</strong> <strong>urchin</strong> life histories.<br />
Oecologia, 65:461-467.<br />
Ebert, T.A., 1986. A new <strong>the</strong>ory to explain <strong>the</strong> origin <strong>of</strong> growth lines in <strong>sea</strong><br />
<strong>urchin</strong> spines. Mar. Ecol. Prog. Ser., 34:197-199.<br />
Ebert, T.A., 1988a. Calibration <strong>of</strong> natural growth lines in ossicles <strong>of</strong> two <strong>sea</strong><br />
<strong>urchin</strong>s, Strongylocentrotus purpuratus and Echinometra mathaei, using<br />
tetracycline. In: R.D. Burke, P.V. Mladenov, P. Lambert, R.L. Parsley<br />
(eds). Echinoderm biology. Balkema, Rotterdam. Pp. 435-443.<br />
Ebert, T.A., 1988b. Allometry, design and constraint <strong>of</strong> body components and <strong>of</strong><br />
shape in <strong>sea</strong> <strong>urchin</strong>s. J. Nat. Hist., 22:1407-1425.<br />
Ebert, T.A., 1994. Allometry and <strong>model</strong> II non-linear regression. J. Theo. Biol.,<br />
168:367-372.<br />
Ebert, T.A., 1998. An analysis <strong>of</strong> <strong>the</strong> importance <strong>of</strong> Allee effects in management<br />
<strong>of</strong> <strong>the</strong> red <strong>sea</strong> <strong>urchin</strong> Strongylocentrotus franciscanus. In: R. Mooi & M.<br />
Telford (eds). Echinoderms: San Francisco. Balkema, Rotterdam.<br />
Pp. 619-627.<br />
193
References<br />
Ebert, T.A., 1999. Plant and animal populations. Methods in demography.<br />
Academic Press, San Diego.<br />
Ebert, T.A. & D.M. Dexter, 1975. A natural history study <strong>of</strong> Encope grandis and<br />
Mellita grantii, two sand dollars in <strong>the</strong> nor<strong>the</strong>rn Gulf <strong>of</strong> California,<br />
Mexico. Mar. Biol., 32:397-407.<br />
Ebert, T.A., S.C. Schroeter & J.D. Dixon, 1993. Inferring demographic processes<br />
from size-frequency distributions: effect <strong>of</strong> pulsed recruitment on simple<br />
<strong>model</strong>s. Fish. Bull. U.S., 91:237-243.<br />
Ebert, T.A. & M.P. Russell, 1992. <strong>Growth</strong> and mortality estimates for red <strong>sea</strong><br />
<strong>urchin</strong> Strongylocentrotus franciscanus from San Nicolas Island,<br />
California. Mar. Ecol. Prog. Ser., 81:31-41.<br />
Ebert, T.A. & M.P. Russell, 1993. <strong>Growth</strong> and mortality <strong>of</strong> subtidal red <strong>sea</strong><br />
<strong>urchin</strong>s (Strongylocentrotus franciscanus) at San Nicolas Island,<br />
California, USA: problems with <strong>model</strong>s. Mar. Biol., 117:79-89.<br />
Emson, R.H., 1984. Bone idle – a recipe for success? In: B.F. Keegan & B.D.S.<br />
O'Connor (eds). Echinodermata. Balkema, Rotterdam. Pp. 25-30.<br />
Fabens, A.J., 1965. Properties and fitting <strong>of</strong> <strong>the</strong> von Bertalanffy growth curve.<br />
<strong>Growth</strong>, 29:265-289.<br />
Fahrmeir, L. & G. Tutz, 1994. Multivariate statistical <strong>model</strong>ling based on general<br />
linear <strong>model</strong>s. Springer-Verlag, New York.<br />
Fernandez, C., 1996. Croissance et nutrition de <strong>Paracentrotus</strong> lividus dans le<br />
cadre d’un projet aquacole avec alimentation artificielle. PhD Thesis,<br />
Université de Corse, France.<br />
Fernandez, C. & C.-F. Boudouresque, 1998. Evaluating artificial diet for small<br />
<strong>Paracentrotus</strong> lividus (Echinodermata: Echinoidea). In: R. Mooi & M.<br />
Telford (eds). Echinoderms: San Francisco. Balkema, Rotterdam. Pp.<br />
651-656.<br />
Fernandez, C. & A. Caltagirone, 1994. <strong>Growth</strong> rate <strong>of</strong> adult <strong>sea</strong> <strong>urchin</strong>s,<br />
<strong>Paracentrotus</strong> lividus in a lagoon environment: <strong>the</strong> effect <strong>of</strong> different diet<br />
194
References<br />
types. In: B. David, A. Guille, J.-P. Féral & M. Roux (eds.). Echinoderms<br />
through Time. Balkema, Rotterdam. Pp. 655-660.<br />
Flammang, P., 1996. Adhesion in echinoderms. In: M. Jangoux & J.M. Lawrence<br />
(eds). Echinoderm studies, vol. 5. Balkema, Rotterdam. Pp. 1-60.<br />
Flammang, P., P. Gosselin & M. Jangoux, 1998. The podia, organs <strong>of</strong> adhesion<br />
and sensory perception in larvae and post-metamorphic stages <strong>of</strong> <strong>the</strong><br />
echinoid <strong>Paracentrotus</strong> lividus (Echinodermata). Bi<strong>of</strong>ouling, 12(1-3):161-<br />
171.<br />
Fletcher, R.I., 1974. The quadric law <strong>of</strong> damped exponential growth. Biometrics,<br />
30:111-124.<br />
Ford, E., 1933. An accounting <strong>of</strong> <strong>the</strong> herring investigations conducted at<br />
Plymouth during <strong>the</strong> years from 1924-1933. J. Mar. Biol. Assoc. U.K.,<br />
19:305-384.<br />
Francis, R.I.C.C., 1995. An alternative mark-recapture analogue <strong>of</strong> Schnute's<br />
growth <strong>model</strong>. Fish. Res., 23:95-111.<br />
Frantzis, A. & A. Gremare, 1992. Ingestion, absorption and growth rates <strong>of</strong><br />
<strong>Paracentrotus</strong> lividus (Echinodermata: Echinoidea) fed different<br />
macrophytes. Mar. Ecol. Prog. Ser., 95:169-183.<br />
Fridberger, A., T. Fridberger & L.-G. Lundin, 1979. Cultivation <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s <strong>of</strong><br />
five different species under strict artificial conditions. Zoon, 7:149-151.<br />
Frontier, S. & D. Pichot-Viale, 1993. Ecosystèmes: structure, fonctionnement,<br />
évolution, 2 nd ed. Masson, Paris.<br />
Fuji, A., 1963. On <strong>the</strong> growth <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>, Hemicentrotus pulcherrimus (A.<br />
Agassiz). Bull. Jap. Soc. Sci. Fish., 29(2):118-126.<br />
Fuji, A., 1967. Ecological studies on <strong>the</strong> growth and food consumption <strong>of</strong><br />
Japanese common littoral <strong>sea</strong> <strong>urchin</strong>, Strongylocentrotus intermedius (A.<br />
Agassiz). Mem. Fac. Fish. Hokkaido Univ., 15(2):83-160.<br />
Fuji, A. & K. Kawamura, 1970. Studies on <strong>the</strong> biology <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>. IV.<br />
Habitat structure and regional distribution <strong>of</strong> Strongylocentrotus<br />
195
References<br />
intermedius on a rocky shore <strong>of</strong> sou<strong>the</strong>rn Hokkaido. Bull. Jap. Soc. Sci.<br />
Fish., 36(8):755-762.<br />
Gage, J.D., 1987. <strong>Growth</strong> <strong>of</strong> <strong>the</strong> deep-<strong>sea</strong> irregular <strong>sea</strong> <strong>urchin</strong>s Echinosigra phiale<br />
and Hemiaster expergitus in <strong>the</strong> Rockall Trough (N.E. Atlantic Ocean).<br />
Mar. Biol., 96:19-30.<br />
Gage, J.D., 1991. Skeletal growth zones as age-markers in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
Psammechinus miliaris. Mar. Biol., 110:217-228.<br />
Gage, J.D., 1992. Natural growth bands and growth variability in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
Echinus esculentus: Results from tetracycline tagging. Mar. Biol.,<br />
114:607-616.<br />
Gage, J.D. & P.A. Tyler, 1985. <strong>Growth</strong> and recruitment <strong>of</strong> <strong>the</strong> deep-<strong>sea</strong> <strong>urchin</strong><br />
Echinus affinis. Mar. Biol., 90:41-53.<br />
Gage, J.D., P.A. Tyler & D. Nichols, 1986. Reproduction and growth <strong>of</strong> Echinus<br />
acutus var. norvegicus Düben & Koren and E. elegans Düben & Koren<br />
on <strong>the</strong> continental slope <strong>of</strong>f Scotland. J. Exp. Mar. Biol. Ecol., 101:61-83.<br />
Gallucci, V.F. & T.J. Quinn, 1979. Reparameterizing, fitting, and testing a simple<br />
growth <strong>model</strong>. Trans. Am. Fish. Soc., 108:14-25.<br />
Gayral, P., J. Cosson, 1973. Exposé synoptique des données biologiques sur la<br />
laminaire digitée Laminaria digitata. In: Synopsis FAO sur les pêches,<br />
n°89.<br />
Gebauer, P. & C.A. Moreno, 1995. Experimental validation <strong>of</strong> <strong>the</strong> growth rings <strong>of</strong><br />
Loxechinus albus (Molina, 1782) in Sou<strong>the</strong>rn Chile (Echinodermata:<br />
Echinoidea). Fish. Res., 21:423-435.<br />
Giese, A.C., 1966. Changes in body-component indexes and respiration with size<br />
in <strong>the</strong> purple <strong>sea</strong> <strong>urchin</strong> Strongylocentrotus purpuratus. Physiol. Zool.,<br />
40:194-200.<br />
Giese, A.C., A. Farmanfarmaian, S. Hilden & P. Doezema, 1966. Respiration<br />
during <strong>the</strong> reproductive cycle in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>, Strongylocentrotus<br />
purpuratus. Biol. Bull., 130:192-201.<br />
196
References<br />
Goebel, N. & M.F. Barker, 1998. Artificial diets supplemented with carotenoid<br />
pigments as feeds for <strong>sea</strong> <strong>urchin</strong>s. In: R. Mooi & M. Telford (eds).<br />
Echinoderms: San Francisco. Balkema, Rotterdam. Pp. 667-672.<br />
Gompertz, B., 1825. On <strong>the</strong> nature <strong>of</strong> <strong>the</strong> function expressive <strong>of</strong> <strong>the</strong> law <strong>of</strong> human<br />
mortality and a new mode <strong>of</strong> determining <strong>the</strong> value <strong>of</strong> life contingencies.<br />
Phil. Trans. Roy. Soc., 115:513-585.<br />
Gonzalez, M.L., M.C. Pérez, D.A. López & C.A Pino, 1993. Effects <strong>of</strong> algal diet<br />
on <strong>the</strong> energy available for growth <strong>of</strong> juvenile <strong>sea</strong> <strong>urchin</strong>s Loxechinus<br />
albus (Molina, 1782). Aquaculture, 115: 87-95.<br />
Gosselin, P. & M. Jangoux, 1996. Induction <strong>of</strong> metamorphosis in <strong>Paracentrotus</strong><br />
lividus larvae (Echinodermata, Echinoidea). Oceano. Acta, 19(3-4):293-<br />
296.<br />
Gosselin, P. & M. Jangoux, 1998. From competent larva to exotrophic juvenile: a<br />
morph<strong>of</strong>unctional study <strong>of</strong> <strong>the</strong> perimetamorphic period <strong>of</strong> <strong>Paracentrotus</strong><br />
lividus (Echinodermata, Echinoidea). Zoomorphology, 118:31-43.<br />
Grosjean, Ph. & M. Jangoux, 2000. <strong>Growth</strong> <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> <strong>Paracentrotus</strong><br />
lividus: <strong>model</strong> and optimization. Green <strong>sea</strong> <strong>urchin</strong> workshop, Moncton,<br />
Canada. Inline: http://crdpm.cus.ca/oursin/pdf/gros.pdf (last consulted<br />
Sept. 8 th 2001).<br />
Grosjean Ph., Ch. Spirlet & M. Jangoux, 1996. Experimental study <strong>of</strong> growth in<br />
<strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus (Lamarck, 1816) (Echinodermata). J.<br />
Exp. Mar. Biol. Ecol., 201:173-184.<br />
Grosjean, Ph., Ch. Spirlet & M. Jangoux, 1999. Comparison <strong>of</strong> three body-size<br />
measurements for echinoids. In: M.D. Candia Carnevali & F. Bonasoro<br />
(eds). Echinoderm Re<strong>sea</strong>rch 1998, Balkema, Rotterdam. Pp. 31-35.<br />
Grosjean, Ph., Ch. Spirlet, P. Gosselin, D. Vaïtilingon & M. Jangoux, 1998.<br />
Land-based closed cycle echiniculture <strong>of</strong> <strong>Paracentrotus</strong> lividus Lamarck<br />
(Echinoidea: Echinodermata): a long-term experiment at a pilot scale. J.<br />
Shellfish Res, 17(5):1523-1531.<br />
197
References<br />
Guillou, M. & Ch. Michel, 1993. Reproduction and growth <strong>of</strong> Sphaerechinus<br />
granularis (Echinodermata: Echinoidea) in <strong>the</strong> Glenan archipelago<br />
(Brittany). J. Mar. Biol. Ass. U.K., 73:172-193.<br />
Hagen, N.T., 1996a. Echinoculture: from fishery enhancement to closed cycle<br />
cultivation. World Aquaculture, Dec. 1996, pp. 7-19.<br />
Hagen, N.T., 1996b. Tagging <strong>sea</strong> <strong>urchin</strong>s: a new technique for individual<br />
identification. Aquaculture, 139:271-284.<br />
Hasselblad, V., 1966. Estimation <strong>of</strong> parameters for a mixture <strong>of</strong> normal<br />
distributions. Technometrics, 8:431-444.<br />
Himmelman, J.H., 1986. Population biology <strong>of</strong> green <strong>sea</strong> <strong>urchin</strong>s on rocky<br />
barrens. Mar. Ecol. Prog. Ser., 33:295-306.<br />
Hinegardner, R.T., 1969. <strong>Growth</strong> and development <strong>of</strong> <strong>the</strong> laboratory cultured <strong>sea</strong><br />
<strong>urchin</strong>. Biol. Bull., 137:465-475.<br />
Huston, M.A. & D.L. De Angelis, 1987. Size bimodality in monospecific<br />
populations: a critical review <strong>of</strong> potential mechanisms. Amer. Nat.,<br />
129(5):678-707.<br />
Huxley, J.S., 1932. Problems <strong>of</strong> relative growth. Metuen and Co, London.<br />
Ihaka, R. & R. Gentleman, 1996. A language for data analysis and graphics. J.<br />
Comput. Graphic. Stat., 5(3):299-314.<br />
Jensen, M., 1969a. Breeding and growth <strong>of</strong> Psammechinus miliaris (Gmelin).<br />
Ophelia, 7:65-78.<br />
Jensen, M., 1969b. Age determination <strong>of</strong> echinoids. Sarsia, 37:41-44.<br />
Jolicoeur, P., 1985. A flexible 3-parameter curve for limited or unlimited somatic<br />
growth. <strong>Growth</strong>, 49:271-281.<br />
Jordana, E., M. Guillou & L.J.L. Lumingas, 1997. Age and growth <strong>of</strong> <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong> Sphaerechinus granularis in Sou<strong>the</strong>rn Brittany. J. Mar. Biol. Ass.<br />
U.K., 77:1199-1212.<br />
198
References<br />
Kaneko, I., Y. Ikeda & H. Ozaki, 1981. Biometrical relations between body<br />
weight and organ weights in freshly sampled and starved <strong>sea</strong> <strong>urchin</strong>. Bull.<br />
Jap. Soc. Sci. Fish., 47(5):593-597.<br />
Kato, S., 1972. Sea <strong>urchin</strong>s: a new fishery develops in California. Mar. Fish.<br />
Rev., 34(9-10):23-30.<br />
Kaufmann, K.W., 1981. Fitting and using growth curves. Oecologia, 49:293-299.<br />
Kautsky, N., 1982. <strong>Growth</strong> and size structure in a Baltic Mytilus edulis<br />
population. Mar. Biol., 68:117-133.<br />
Keats, D.W., D.H. Steele & G.R. South, 1983. Food relations and short term<br />
aquaculture potential <strong>of</strong> <strong>the</strong> green <strong>sea</strong> <strong>urchin</strong> (Strongylocentrotus<br />
droebachiensis) in Newfoundland. MSRL Tech. Rep., 24:1-24.<br />
Kelly, M.S., J.D. McKenzie & C.C. Brodie, 1998. Sea <strong>urchin</strong>s in polyculture: The<br />
way to enhanced gonad growth? In: R. Mooi & M. Telford (eds).<br />
Echinoderms: San Francisco. Balkema, Rotterdam. Pp. 707-711.<br />
Kenner, M.C., 1992. Population dynamics <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Strongylocentrotus<br />
purpuratus in a Central California kelp forest: recruitment, mortality,<br />
growth and diet. Mar. Biol., 112:107-118.<br />
Klinger, T.S., J.M. Lawrence & A.L. Lawrence, 1994. Digestive characteristics <strong>of</strong><br />
<strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Lytechinus variegatus (Lamarck) (Echinodermata:<br />
Echinoidea) fed prepared feeds. J. World Aqua. Soc., 25(4):489-496.<br />
Klinger, T.S., J.M. Lawrence & A.L. Lawrence, 1997. Gonad and somatic<br />
production <strong>of</strong> Strongylocentrotus droebachiensis fed manufactured feeds.<br />
Bull. Aqua.. Assoc. Canada, 1:35-37.<br />
Klinger, T.S., J.M. Lawrence & A.L. Lawrence, 1998. Digestion, absorption, and<br />
assimilation <strong>of</strong> prepared feeds by echinoids. In: R. Mooi & M. Telford<br />
(eds). Echinoderms: San Francisco. Balkema, Rotterdam. Pp. 713-721.<br />
Knight, W., 1968. Asymptotic growth: an example <strong>of</strong> non-sense disguised as<br />
ma<strong>the</strong>matics. J. Fish. Res. Board Canada, 25(6):1303-1307.<br />
199
References<br />
Kobayashi, S. & J. Taki, 1969. Calcification in <strong>sea</strong> <strong>urchin</strong>s. I. A tetracycline<br />
investigation <strong>of</strong> growth <strong>of</strong> <strong>the</strong> mature test in Strongylocentrotus<br />
intermedius. Calcif. Tiss. Res., 4:607-621.<br />
Koenker, R., 2001. Quantile regression. In: S. Fienberg & J. Kadane (eds).<br />
International encyclopedia <strong>of</strong> <strong>the</strong> social sciences. (in press)<br />
Koenker, R. & G. Bassett, 1978. Regression quantiles. Econometrica, 46:33-50.<br />
Koenker, R. & J. Machado, 1999. Goodness <strong>of</strong> fit and related inference processes<br />
for quantile regression. J. Am. Stat. Assoc., 94:1296-1310.<br />
Koenker, R. & B.J. Park, 1996. An interior point algorithm for nonlinear quantile<br />
regression. J. Econometrics, 71(1-2):265-283.<br />
Lamare, M.D. & P.V. Mladenov, 2000. Modelling somatic growth in <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong> Evechinus chloroticus (Echinoidea: Echinometridae). J. Exp. Mar.<br />
Biol. Ecol., 243:17-43.<br />
Lane, J.E.M. & J.M. Lawrence, 1980. Seasonal variation in body growth, density<br />
and distribution <strong>of</strong> a population <strong>of</strong> sand dollars, Mellita<br />
quinquiesperforata (Leske). Bull. Mar. Sci., 30(4):871-882.<br />
Lawrence, J.M., 1982. Digestion. In: M. Jangoux & J.M. Lawrence (eds.).<br />
Echinoderm nutrition. Balkema, Rotterdam. Pp. 283-316.<br />
Lawrence, J.M. & J.M. Lane, 1982. The utilization <strong>of</strong> nutrients by<br />
postmetamorphic echinoderms. In: M. Jangoux & J.M. Lawrence (eds).<br />
Echinoderm nutrition. Balkema, Rotterdam. Pp. 331-371.<br />
Lawrence, J.M., S. Olave, R. Otaiza, A.L. Lawrence & E. Bustos, 1997.<br />
Enhancement <strong>of</strong> gonadal production in <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Loxechinus albus in<br />
Chile fed extruded feeds. J. World Aqua. Soc., 28:91-96.<br />
Lawrence, J.M., B.D. Robbins & S.S. Bell, 1995. Scaling <strong>of</strong> <strong>the</strong> pieces <strong>of</strong> <strong>the</strong><br />
Aristotle's lantern in five species <strong>of</strong> Strongylocentrotus (Echinodermata:<br />
Echinoidea). J. Nat. Hist., 29:243-247.<br />
Laws, E.A & J.W. Archie, 1981. Appropriate use <strong>of</strong> regression analysis in marine<br />
biology. Mar. Biol., 65:13-16.<br />
200
References<br />
Ledireac'h, J.-P., 1987. La pêche des oursins en Méditerranée: historique,<br />
techniques, législation, production. In: C.F. Boudoresque (ed). Colloque<br />
international sur <strong>Paracentrotus</strong> lividus et les oursins comestibles. GIS<br />
Posidonie Publ., Marseille. Pp. 335-362.<br />
Le Gall, P., 1987. La pêche des oursins en Bretagne. In: C.F. Boudoresque (ed).<br />
Colloque international sur <strong>Paracentrotus</strong> lividus et les oursins<br />
comestibles. GIS Posidonie Publ., Marseille. Pp. 311-324.<br />
Le Gall, P., 1990. Culture <strong>of</strong> echinoderms, In: G. Barnabé (ed.). Aquaculture,<br />
Vol. 1. Ellis Horwood, New York. Pp. 443-462.<br />
Le Gall, P. & D. Bucaille, 1989. Sea <strong>urchin</strong>s production by inland farming. In: N.<br />
De Pauw, E. Jaspers, H. Ackefors & N. Wilkins (eds.). Aquaculture - a<br />
biotechnology in progress. European Aquaculture Society, Bredene,<br />
Belgium. Pp. 53-59.<br />
Leinass, H.P. & H. Christie, 1996. Effects <strong>of</strong> removing <strong>sea</strong> <strong>urchin</strong>s<br />
(Strongylocentrotus droebachiensis): stability <strong>of</strong> <strong>the</strong> barren state and<br />
succession <strong>of</strong> kelp forest recovery in <strong>the</strong> East Atlantic. Oecologia,<br />
105:524-536.<br />
Levitan, D.R., 1988. Density-dependent size regulation and negative growth in<br />
<strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Diadema antillarum Philippi. Oecologia, 76:627-629.<br />
Levitan, D.R. & S.J. Genovese, 1989. Substratum-dependent predator-prey<br />
dynamics: Patch reefs as refuges from gastropod predation. J. Exp. Mar.<br />
Biol. Ecol., 130:111-118.<br />
Liquori, A.M., A. Monroy, E. Parisi & A. Tripiciano, 1981. A <strong>the</strong>oretical<br />
equation for diauxic growth and its application to <strong>the</strong> kinetics <strong>of</strong> <strong>the</strong> early<br />
development <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> embryo. Differentiation, 20:174-175.<br />
Lumingas, L.J.L. & M. Guillou, 1994. <strong>Growth</strong> zones and back-calculation for <strong>the</strong><br />
<strong>sea</strong> <strong>urchin</strong> Sphaerechinus granularis from <strong>the</strong> bay <strong>of</strong> Brest, France. J.<br />
Mar. Biol. Assoc. U.K., 74:671-686.<br />
Malthus, T.R., 1798. An essay on <strong>the</strong> principal <strong>of</strong> population. Reedition (1970),<br />
Penguin Books, New York.<br />
201
References<br />
McDonald, P.D. & T.J. Pitcher, 1979. Age-groups from size-frequency data: A<br />
versatile and efficient method <strong>of</strong> analyzing distribution mixtures. J. Fish.<br />
Res. Board Can., 36:987-1001.<br />
McPherson, B.F., 1965. Contributions to <strong>the</strong> biology <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Tripneustes<br />
ventricosus. Bull. Mar. Sci., 15(1):228-244.<br />
McPherson, B.F., 1968. Contributions to <strong>the</strong> biology <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> Eucidaris<br />
tribuloides (Lamarck). Bull. Mar. Sci., 18(2):400-443.<br />
Meidel, S.K. & R.E. Scheibling, 1998. Size and age structure <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong><br />
Strongylocentrotus droebachiensis in different habitats. In: R. Mooi & M.<br />
Telford (eds). Echinoderms: San Francisco. Balkema, Rotterdam. Pp.<br />
737-742.<br />
Michel, H.B., 1984. Culture <strong>of</strong> Lytechinus variegatus (Lamarck) (Echinodermata:<br />
Echinoidea) from egg to young adult. Bull. Mar. Sci., 34(2):312-314.<br />
Middleton, D.A.J., S.C. Gurney & J.D. Gage, 1998. <strong>Growth</strong> and energy allocation<br />
in <strong>the</strong> deep-<strong>sea</strong> <strong>urchin</strong> Echinus affinis. Biol. J. Lin. Soc., 64:315-336.<br />
Miller, R.J. & K.H. Mann, 1973. Ecological energetics <strong>of</strong> <strong>the</strong> <strong>sea</strong>weed zone in a<br />
marine bay on <strong>the</strong> Atlantic coast <strong>of</strong> Canada. III. Energy transformations<br />
by <strong>sea</strong> <strong>urchin</strong>s. Mar. Biol., 18:99-114.<br />
Morgan, L.E., L.W. Botsford, S.R. Wing & B.D. Smith, 2000. Spatial variability<br />
in growth and mortality <strong>of</strong> <strong>the</strong> red <strong>sea</strong> <strong>urchin</strong>, Strongylocentrotus<br />
franciscanus, in nor<strong>the</strong>rn California. Can. J. Fish. Aquat. Sci., 57:980-<br />
992.<br />
Mortensen, T., 1950. A monograph <strong>of</strong> <strong>the</strong> Echinoidea. 5 vol., Reitzel edit.,<br />
Copenhagen.<br />
Moss, M.L. & M. Meehan, 1968. <strong>Growth</strong> <strong>of</strong> <strong>the</strong> echinoid test. Acta Anat., 69:409-<br />
444.<br />
Motnikar S., P. Bryl & J. Boyer, 1997. Conditioning green <strong>sea</strong> <strong>urchin</strong>s in tanks :<br />
<strong>the</strong> effect <strong>of</strong> semi-moist diets on gonad quality. Bull. Aqua. Assoc.<br />
Canada, 97-1:21-25.<br />
202
References<br />
Muller-Feuga, A., 1990. Modélisation de la croissance des poissons en élevage.<br />
Rapp. Sci. Tech. IFREMER, n° 21.<br />
Munk, J.E., 1992. Reproduction and growth <strong>of</strong> green <strong>sea</strong> <strong>urchin</strong><br />
Strongylocentrotus droebachiensis (Müller) near Kodiak, Alaska. J.<br />
Shellfish Res., 11(2):245-254.<br />
Murray, J.D., 1993. Ma<strong>the</strong>matical biology. 2 nd ed. Springer-Verlag, Berlin.<br />
Nedelec, H., 1983. Sur un nouvel indice de réplétion pour les oursins réguliers.<br />
Rapp. Comm. Int. Mer Médit., 28(3):149-151.<br />
Nedelec, H., M. Verlaque & A. Dialopoulis, 1981. Preliminary data on Posidonia<br />
consumption by <strong>Paracentrotus</strong> lividus in Corsica (France). Rap. Comm.<br />
Int. Mer Médit., 27(2):203-204.<br />
Nelder, J.A., 1961. The fitting <strong>of</strong> a generalization <strong>of</strong> <strong>the</strong> logistic curve.<br />
Biometrics, 17:89-110.<br />
Nelder, J.A. & R. Mead, 1965. A simplex algorithm for function minimization.<br />
Comput. J., 7: 308-313.<br />
Nichols, D., A.A.T. Sime & G.M. Bishop, 1985. <strong>Growth</strong> in populations <strong>of</strong> <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong> Echinus esculentus L. (Echinodermata: Echinoidea) from <strong>the</strong><br />
English Channel and Firth <strong>of</strong> Clyde. J. Exp. Mar. Biol. Ecol., 86:219-228.<br />
Nocedal, J. & S.J. Wright, 1999. Numerical Optimization. Springer-Verlag, New<br />
York.<br />
Passino, K.M. & S. Yurkovich, 1988. Fuzzy control. Addison-Wesley, Menlo<br />
Park.<br />
Pearse, J.S. & V.B Pearse, 1975. <strong>Growth</strong> zones in <strong>the</strong> echinoid skeleton. Amer.<br />
Nat., 15:731-753.<br />
Percy, J.A., 1972. Thermal adaptation in <strong>the</strong> boreo-arctic echinoid,<br />
Strongylocentrotus droebachiensis (O.F. Muller 1776). I. Seasonal<br />
acclimatization <strong>of</strong> respiration. Physiol. Zool., 45(4):277-289.<br />
203
References<br />
Pouvreau, S., C. Bacher & M. Héral, 2000. Ecophysiological <strong>model</strong> <strong>of</strong> growth<br />
and reproduction <strong>of</strong> <strong>the</strong> black pearl oyster, Pinctada margaritifera:<br />
potential applications for pearl farming in French Polynesia. Aquaculture,<br />
186(1-2):117-144.<br />
Preece, M.A. & M.J. Baines, 1978. A new family <strong>of</strong> ma<strong>the</strong>matical <strong>model</strong>s<br />
describing <strong>the</strong> human growth curve. Ann. Hum. Biol., 5:1-24.<br />
Rao, C.R., 1965. The <strong>the</strong>ory <strong>of</strong> least squares when parameters are stochastic and<br />
its application to <strong>the</strong> analysis <strong>of</strong> growth curves Biometrika, 52:477-458.<br />
Raymond, B.G. & R.E. Scheibling, 1987. Recruitment and growth <strong>of</strong> <strong>the</strong> <strong>sea</strong><br />
<strong>urchin</strong> Strongylocentrotus droebachiensis (Müller) following mass<br />
mortalities <strong>of</strong>f Nova Scotia, Canada. J. Exp. Mar. Biol. Ecol., 108:31-54.<br />
Régis, M.-B., 1969. Premières données sur la croissance de <strong>Paracentrotus</strong> lividus<br />
Lmk. Téthys, 1(4):1049-1056.<br />
Régis, M.-B., 1979. Croissance négative de l'oursin <strong>Paracentrotus</strong> lividus<br />
(Lamarck) (Echinoidea-Echinidea). C. R. Acad. Sci. Paris D, 188:355-<br />
358.<br />
Régis, M.-B. & R. Arfi, 1978. Etude comparée de la croissance de trois<br />
populations de <strong>Paracentrotus</strong> lividus (Lamarck), occupant des biotopes<br />
différents, dans le golfe de Marseille. C. R. Acad. Sc. Paris D, 286:1211-<br />
1214.<br />
Richards, F.J., 1959. A flexible growth function for empirical use. J. Exp. Bot.,<br />
10(29):290-300.<br />
Ricker, W.E., 1979. <strong>Growth</strong> rates and <strong>model</strong>s. In: Fish physiology, vol. 8.<br />
Academic Press, New York.<br />
R<strong>of</strong>f, D.A., 1980. A motion for <strong>the</strong> retirement <strong>of</strong> <strong>the</strong> von Bertalanffy function.<br />
Can. J. Fish. Aquat. Sci., 37:127-129.<br />
Rowley, R.J., 1989. Settlement and recruitment <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s<br />
(Strongylocentrotus spp.) in a <strong>sea</strong> <strong>urchin</strong> barren ground and a kelp bed:<br />
204
References<br />
are populations regulated by settlement or post-settlement? Mar. Biol.,<br />
100:485-494.<br />
Rowley, R.J., 1990. Newly settled <strong>sea</strong> <strong>urchin</strong>s in a kelp bed and <strong>urchin</strong> barren<br />
ground: a comparison <strong>of</strong> growth and mortality. Mar. Ecol. Prog. Ser.,<br />
62:229-240.<br />
Ruppert, E.E. & R.D. Barnes, 1994. Invertebrate zoology. 6 th ed. Saunders<br />
College Publishers, Philadelphia.<br />
Russell, M.P., 1987. Life history traits and resource allocation in <strong>the</strong> purple <strong>sea</strong><br />
<strong>urchin</strong> Strongylocentrotus purpuratus (Stimpson). J. Exp. Mar. Biol.<br />
Ecol., 108:199-216.<br />
Russell, M.P., T.A. Ebert & P.S. Petraitis, 1998. Field estimates <strong>of</strong> growth and<br />
mortality <strong>of</strong> <strong>the</strong> green <strong>sea</strong> <strong>urchin</strong>, Strongylocentrotus droebachiensis.<br />
Ophelia, 48(2):137-153.<br />
Russell, M.P. & R.W. Meredith, 2000. Natural growth lines in echinoid ossicles<br />
are not reliable indicators <strong>of</strong> age: a test using Strongylocentrotus<br />
droebachiensis. Invert. Biol., 119(4):410-420.<br />
Sainsbury, K.J., 1980. Effect <strong>of</strong> individual variability on <strong>the</strong> von Bertalanffy<br />
growth equation. Can. J. Fish. Aquat. Sci., 37:241-247.<br />
Saito, K., 1992. Japan's <strong>sea</strong> <strong>urchin</strong> enhancement experience. In: C.M. Dewees<br />
(ed.). The Management and enhancement <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s and o<strong>the</strong>r kelp<br />
bed resources: a Pacific rim perspective. Publication no. T-CSGCP-028,<br />
California Sea Grant College, University <strong>of</strong> California, La Jolla, CA<br />
92093-0232. Pp. 1-38.<br />
Salski, A., O. Franzle & P. Kandzia (eds), 1995. Fuzzy logic in ecological<br />
<strong>model</strong>ling. Papers presented at a workshop held in Kiel, Germany, 12-13<br />
October 1993. Ecol. Model., 85:1-97.<br />
Schnute, J., 1981. A versatile growth <strong>model</strong> with statistically stable parameters.<br />
Can. J. Fish. Aquat. Sci., 38:1128-1140.<br />
205
References<br />
Sellem, F., H. Langar & D. Pesando, 2000. Age et croissance de l'oursin<br />
<strong>Paracentrotus</strong> lividus Lamarck, 1816 (Echinodermata-Echinoidea) dans le<br />
golfe de Tunis (Méditerranée). Oceanol. Acta, 23(5):607-613.<br />
Sen, A. & M. Srivastava, 1990. Regression analysis. Theory, methods, and<br />
applications. Springer-Verlag, New York.<br />
Shick, J.M., 1983. Respiratory gas exchange in echinoderms. In: M. Jangoux &<br />
J.M. Lawrence (eds). Echinoderm studies, vol. 1. Balkema, Rotterdam.<br />
Pp. 67-110.<br />
Sime, A.A.T., 1981. <strong>Growth</strong> ring analyses in regular echinoids. Prog. Underwat.<br />
Sci., 7:7-14.<br />
Sime, A.A.T. & G.J. Cranmer, 1985. Age and growth <strong>of</strong> North Sea echinoids. J.<br />
Mar. Biol. Ass. U.K., 65:583-588.<br />
Sloan, N.A., 1985. Echinoderm fisheries <strong>of</strong> <strong>the</strong> world. In: B.F. Keegan & B.D.F.<br />
O'Connor (eds). Proceedings <strong>of</strong> <strong>the</strong> fifth international echinoderm<br />
conference, Galway. Balkema, Rotterdam. Pp. 109-124.<br />
Smith, B.D. & L.W. Botsford, 1998. Interpretation <strong>of</strong> growth, mortality, and<br />
recruitment patterns in size-at-age, growth increment, and size frequency<br />
data. In: G.S. Jamieson & A. Campbell (eds). Proceedings <strong>of</strong> <strong>the</strong> North<br />
Pacific workshop on invertebrate stock assessment and management. Can.<br />
Publ. Fish. Aquat. Sci., 125:125-139.<br />
Smith, B.D., L.W. Botsford & S.R. Wing, 1998. Estimation <strong>of</strong> growth and<br />
mortality parameters from size frequency distributions lacking age<br />
patterns: <strong>the</strong> red <strong>sea</strong> <strong>urchin</strong> (Strongylocentrotus franciscanus) as an<br />
example. Can. J. Fish. Aquat. Sci., 55:1236-1247.<br />
Sokal, R.R. & F.J. Rohlf, 1981. Biometry, 2 nd ed. Freeman and co, New York.<br />
Spirlet, Ch., 1999. Biologie de l'oursin comestible (<strong>Paracentrotus</strong> lividus):<br />
contrôle du cycle reproducteur et optimalisation de la phase de<br />
remplissage gonadique. PhD Thesis, Université Libre de Bruxelles,<br />
Belgium.<br />
206
References<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 1994. Differentiation <strong>of</strong> <strong>the</strong> genital<br />
apparatus in a juvenile echinoid (<strong>Paracentrotus</strong> lividus). In: B. David,<br />
A. Guille, J.-P. Féral & M. Roux (eds). Echinoderms through Time,<br />
Balkema, Rotterdam. Pp. 881-886.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 1998a. Reproductive cycle <strong>of</strong> <strong>the</strong><br />
echinoid <strong>Paracentrotus</strong> lividus: analysis by means <strong>of</strong> <strong>the</strong> maturity index.<br />
Invert. Reprod. Develop., 34(1):69-81.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 1998b. Optimizing food distribution in<br />
closed-circuit cultivation <strong>of</strong> edible <strong>sea</strong> <strong>urchin</strong>s (<strong>Paracentrotus</strong> lividus:<br />
Echinoidea). Aquat. Living Resour., 11(4):273-277.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 2000. Optimization <strong>of</strong> gonad growth by<br />
manipulation <strong>of</strong> temperature and photoperiod in cultivated <strong>sea</strong> <strong>urchin</strong>,<br />
<strong>Paracentrotus</strong> lividus (Lamarck) (Echinodermata). Aquaculture, 185:85-<br />
99.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 2001. Cultivation <strong>of</strong> <strong>Paracentrotus</strong><br />
lividus (Echinodermata: Echinoidea) fed extruded feeds: digestion<br />
efficiency, somatic production and gonadal growth. Aqua. Nutr., 7(2):91-<br />
99.<br />
Spotte, S., 1991. Captive <strong>sea</strong>water fishes. Science and technology. Wiley & Sons,<br />
New York.<br />
Stickle, W.B. & R. Ahokas, 1974. The effects <strong>of</strong> tidal fluctuation <strong>of</strong> salinity on<br />
<strong>the</strong> perivisceral fluid composition <strong>of</strong> several echinoderms. Comp.<br />
Biochem. Physiol. A, 47:469-476.<br />
Stone, J.R., 1996. The evolution <strong>of</strong> ideas: a phylogeny <strong>of</strong> shell <strong>model</strong>s. Amer.<br />
Nat., 148:904-929.<br />
Strathmann, R.R., 1978. Larval settlement in echinoderms. In F.S. Chia & M.E.<br />
Rice (eds). Settlement and metamorphosis <strong>of</strong> marine invertebrate larvae.<br />
Elsevier, New York. Pp. 235-246.<br />
207
References<br />
Stumm, W. & J.J. Morgan, 1981. Aquatic chemistry. An introduction<br />
emphasizing chemical equilibria in natural waters, 2 nd ed. Wiley & Sons,<br />
New York.<br />
Sumich, J.L. & J.E. McCauley, 1973. <strong>Growth</strong> <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>, Allocentrotus<br />
fragilis, <strong>of</strong>f Oregon coast. Pacif. Sci., 27(2):156-167.<br />
Taki, J., 1972. A tetracycline labelling observation on growth zones in <strong>the</strong> test<br />
plates <strong>of</strong> Strongylocentrotus intermedius. Bull. Jap. Soc. Sci. Fish.,<br />
38:117-121.<br />
Tanaka, M., 1982. A new growth curve which expresses infinite increase. Publ.<br />
Amakusa Mar. Biol. Lab., 6:167-177.<br />
Tanaka, M., 1988. Eco-physiological meaning <strong>of</strong> parameters <strong>of</strong> ALOG growth<br />
curve. Publ. Amakusa Mar. Biol. Lab., 9:103-106.<br />
Tegner, M.J. & P.K. Dayton, 1977. Sea <strong>urchin</strong> recruitment patterns and<br />
implication <strong>of</strong> commercial fishing. Science, 196:324-326.<br />
Tegner, M.J. & P.K. Dayton, 1981. Population structure, recruitment and<br />
mortality <strong>of</strong> two <strong>sea</strong> <strong>urchin</strong>s (Strongylocentrotus franciscanus and S.<br />
purpuratus) in a kelp forest. Mar. Ecol. Prog. Ser., 5:255-268.<br />
Tegner, M.J. & L.A. Levin, 1983. Spiny lobsters and <strong>sea</strong> <strong>urchin</strong>s: analysis <strong>of</strong> a<br />
predator-prey interaction. J. Exp. Mar. Biol. Ecol., 73:125-150.<br />
Tessier, G., 1934. Dysharmonies et discontinuités dans la croissance. Exposés de<br />
biométrie et de statistique biologique, vol. 95. Hermann and Cie, Paris.<br />
Tessier, G., 1948. La relation d'allométrie. Sa signification statistique et<br />
biologique. Biometrics, 4:14-48.<br />
Timmons, T.J. & W.L. Shelton, 1980. Differential growth <strong>of</strong> largemouth bass in<br />
West Point reservoir, Alabama-Georgia. Trans. Amer. Fish. Soc.,<br />
109:176-186.<br />
Tomassone, R., C. Dervin & J.P. Masson, 1993. Biométrie: modélisation de<br />
phénomènes biologiques. Masson, Paris.<br />
208
References<br />
Turch, B., 1998. A simple shell <strong>model</strong>: applications and implications. Apex,<br />
13:161-176.<br />
Turner, M.E., Jr., B.A. Blumenstein & J.L. Sebaugh, 1969. A generalization <strong>of</strong><br />
<strong>the</strong> logistic growth. Biometrics, 25:577-580.<br />
Turon, X., G. Giribet, S. Lopez & C. Palacin, 1995. <strong>Growth</strong> and population<br />
structure <strong>of</strong> <strong>Paracentrotus</strong> lividus (Echinodermata: Echinoidea) in two<br />
contrasting habitats. Mar. Ecol. Prog. Ser., 122:193-204.<br />
UNESCO, 1981. Tables océanographiques internationales, Vol. 3. UNESCO<br />
Tech. Pap. Mar. Sci., 39.<br />
Vaïtilingon, D., R. Morgan, Ph. Grosjean, P. Gosselin & M. Jangoux, 2001.<br />
Influence <strong>of</strong> delayed metamorphosis and food intake on <strong>the</strong><br />
perimetamorphic period <strong>of</strong> <strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus. J. Exp.<br />
Mar. Biol. Ecol., 262(1):41-60.<br />
Van Osselaer, Ch., 2001. Differentiation morphométrique et génétique de Helix<br />
pomatia L., 1758 et Helix lucorum L., 1758. PhD Thesis, Université Libre<br />
de Bruxelles, Belgium.<br />
Van Osselaer, Ch. & Ph. Grosjean, 2000. Suture and location <strong>of</strong> <strong>the</strong> coiling axis<br />
in gastropod shells. Paleobiology, 26(2):238-257.<br />
Vaughan, D.S. & P. Kanciruk, 1982. An empirical comparison <strong>of</strong> estimation<br />
procedures for <strong>the</strong> von Bertalanffy growth equation. J. Cons. Int. Explor.<br />
Mer, 40:211-219.<br />
Verhulst, P.F., 1838. Notice sur la loi que suit la population dans son<br />
accroissement. Cah. Math. Phys., 10:113-121.<br />
von Bertalanffy, L., 1938. A quantitative <strong>the</strong>ory <strong>of</strong> organic growth (inquiries <strong>of</strong><br />
growth laws II). Hum. Biol., 10(2):181-213.<br />
von Bertalanffy, L., 1957. Quantitative laws in metabolism and growth. Quart.<br />
Rev. Biol., 32(3):217-231.<br />
Walford, L.A., 1946. A new graphic method <strong>of</strong> describing <strong>the</strong> growth <strong>of</strong> animals.<br />
Biol. Bull., 90:141-147.<br />
209
References<br />
Webster, S.K. & A.C. Giese, 1975. Oxygen consumption <strong>of</strong> <strong>the</strong> purple <strong>sea</strong> <strong>urchin</strong><br />
with special reference to <strong>the</strong> reproductive cycle. Biol. Bull., 148:165-180.<br />
Weibull, W., 1951. A statistical distribution function <strong>of</strong> wide applicability. J.<br />
Appl. Mech., 18:293-296.<br />
Williams, C.T. & L.G. Harris, 1998. <strong>Growth</strong> <strong>of</strong> juvenile green <strong>sea</strong> <strong>urchin</strong>s on<br />
natural and artificial diets. In: R. Mooi & M. Telford (eds). Echinoderms:<br />
San Francisco. Balkema, Rotterdam. Pp. 887-892.<br />
Willows, R.I., 1992. Optimal digestive investment: a <strong>model</strong> for filter feeders<br />
experiencing variable diets. Limnol. Oceanogr., 37(4):829-847.<br />
Winsor, C.P., 1932. The Gompertz curve as a growth curve. Proc. Nat. Acad.<br />
Sci., 18:1-8.<br />
Yamaguchi, M., 1975. Estimating growth parameters from growth rate data.<br />
Oecologia, 20:321-332.<br />
Zar, J.H., 1999. Biostatistical analysis, 4th ed. Prentice Hall, London.<br />
Zimmermann, H.J., 1991. Fuzzy set <strong>the</strong>ory and its applications. 2 nd ed. Kluwer<br />
Academic, Boston.<br />
210
Annexes<br />
211
212
Annexes<br />
ANNEXES<br />
Annex I: R code for fitting growth <strong>model</strong>s.<br />
Annex II: dataset <strong>of</strong> <strong>the</strong> cohort measured during seven years.<br />
Annex III: abstracts <strong>of</strong> publications and symposia.<br />
213
Annexes<br />
214
Annex I: R code for fitting growth <strong>model</strong>s<br />
Annexes<br />
Code (as well as dataset presented in annex II) is available at:<br />
http://www.sciviews.org/_phgrosjean/growth/index.htm. This code<br />
runs under <strong>the</strong> free (GNU Public License) statistical s<strong>of</strong>tware R, which<br />
is downloadable at: http://cran.r-project.org. It is available for almost<br />
all plateforms (Unixes, Linux, Windows, MacOS). The 'nlrq' package<br />
for nonlinear quantile regression is also downloadable from <strong>the</strong>re.<br />
Rem: LaboKit and ShellAxis used to assist in measurements <strong>of</strong> <strong>sea</strong><br />
<strong>urchin</strong>s are available for free (GPL) at: http://www.sciviews.org.<br />
a. Code for analyzing data and fitting envelope <strong>model</strong>s<br />
Main script file<br />
This script runs a complete analysis <strong>of</strong> <strong>the</strong> dataset presented in annex<br />
II, and discussed in Part IV. The dataset is first explored (distribution <strong>of</strong><br />
sizes, growth pattern…). Then, quantile regressions are fitted with<br />
traditional <strong>model</strong>s and with <strong>the</strong> original growth <strong>model</strong>. Finally, <strong>the</strong><br />
envelope <strong>model</strong> is designed, tested, and fitted on <strong>the</strong> same dataset.<br />
## Demonstration <strong>of</strong> using R for analyzing growth data as in <strong>the</strong> paper:<br />
# A functional growth <strong>model</strong> with intraspecific competition applied to<br />
# <strong>sea</strong> <strong>urchin</strong>s. Grosjean, Ph., Ch. Spirlet & M. Jangoux (in preparation)<br />
#<br />
# version 1.0 (30/08/2001)<br />
#<br />
# by Ph. Grosjean (phgrosjean@sciviews.org)<br />
# GNU Public License v. 2 or above at your convenience<br />
# Use at your own risks!<br />
# You need:<br />
# R v. 1.3.0 or above (tested only under Windows, please, report o<strong>the</strong>r)<br />
# libraries nls, nlrq, akima (see http://cran.r-project.org)<br />
# files Plividus.txt, <strong>Growth</strong>Fun.R, nlModels.R<br />
# (see http://www.sciviews.org/_phgrosjean/growth/index.htm)<br />
# Put all files in a common directory<br />
# In R, change current directory to that one<br />
# Enter: source("<strong>Growth</strong>.R", print.eval=TRUE)<br />
cat("\n\n ===== DEMONSTRATION OF ANALYSIS OF GROWTH DATA =====\n")<br />
# To do: put here a more detailed introduction!!!<br />
library(nls)<br />
library(nlrq)<br />
source("<strong>Growth</strong>Fun.R")<br />
source("nlModels.R")<br />
215
# If no graphic device currently open, create one for <strong>the</strong> demo<br />
if (is.null(dev.list())) windows()<br />
# Define a pause function<br />
Pause
Pause()<br />
# Rem: o<strong>the</strong>r graphs not used here...<br />
#hist(Pl$sizes[, s
lines(edatq[, 1], predict(edat.025, newdata=list(age=edatq[, 1])), col=4)<br />
legend(1500, 20, c("quantile 0.975", "quantile 0.5 (median)", "quantile 0.025"),<br />
col=c(1,2,4), lty=1, pch=1)<br />
#Rem: for <strong>the</strong> graph <strong>of</strong> Fig. 1B in <strong>the</strong> paper, it is:<br />
#windows(8, 6)<br />
#plot(edatq[, "age"], edatq[, "0.975"], ylim=c(0,65), xlab=expression(paste("Time ",<br />
italic("t"), " in days")), ylab=expression(paste("Diameter ", italic("D"), " in mm")),<br />
pch=2)<br />
#points(edatq[, "age"], edatq[, "0.5"], pch=1)<br />
#points(edatq[, "age"], edatq[, "0.025"], pch=6)<br />
#lines(edatq[, 1], predict(edat.975, newdata=list(age=edatq[, 1])), lty=2)<br />
#lines(edatq[, 1], predict(edat.5, newdata=list(age=edatq[, 1])), lty=1)<br />
#lines(edatq[, 1], predict(edat.025, newdata=list(age=edatq[, 1])), lty=4)<br />
#legend(1700, 18, c("quantile 0.975", "quantile 0.5 (median)", "quantile 0.025"),<br />
lty=c(2,1,4), pch=c(2,1,6))<br />
Pause()<br />
cat("\n ---- Quantile regression with <strong>the</strong> fuzzy-remanent growth function ----\n\n")<br />
cat("... it takes some time, please, be patient...\n")<br />
# Rem: not able to calculate self-starting conditions with this data set... give<br />
initial plausible values!<br />
# starting values are very important here! We <strong>of</strong>ten get stuck in a local minimum or<br />
<strong>the</strong> algorithm fails!<br />
# Rem: time-scale starts at metamorphosis, and is thus shifted by 30 days (see paper)<br />
edatf.975
plot(edat0q[, "age"], edat0q[, "0.975"], ylim=c(0,65), xlab="Time elapsed from<br />
metamorphosis in days", ylab="Diameter increase in mm", main="Quantile regressions<br />
with SSfuzremOrig1")<br />
points(edat0q[, "age"], edat0q[, "0.5"], col=2)<br />
points(edat0q[, "age"], edat0q[, "0.025"], col=4)<br />
lines(edat0q[, 1], predict(edat0f.975, newdata=list(age=edat0q[, 1])), col=1)<br />
lines(edat0q[, 1], predict(edat0f.5, newdata=list(age=edat0q[, 1])), col=2)<br />
lines(edat0q[, 1], predict(edat0f.025, newdata=list(age=edat0q[, 1])), col=4)<br />
legend(1500, 20, c("quantile 0.975", "quantile 0.5 (median)", "quantile 0.025"),<br />
col=c(1,2,4), lty=1, pch=1)<br />
#Rem: for <strong>the</strong> graph <strong>of</strong> Fig.4 in <strong>the</strong> paper, it is:<br />
#windows(8, 6)<br />
#plot(edat0q[, "age"], edat0q[, "0.975"], ylim=c(0,65), xlab=expression(paste("Time ",<br />
italic("t'"), " in days")), ylab=expression(paste("Diameter increase ", italic("D'"),<br />
" in mm")), pch=2)<br />
#points(edat0q[, "age"], edat0q[, "0.5"], pch=1)<br />
#points(edat0q[, "age"], edat0q[, "0.025"], pch=6)<br />
#lines(edat0q[, 1], predict(edat0f.975, newdata=list(age=edat0q[, 1])), lty=2)<br />
#lines(edat0q[, 1], predict(edat0f.5, newdata=list(age=edat0q[, 1])), lty=1)<br />
#lines(edat0q[, 1], predict(edat0f.025, newdata=list(age=edat0q[, 1])), lty=4)<br />
#legend(1700, 18, c("quantile 0.975", "quantile 0.5 (median)", "quantile 0.025"),<br />
lty=c(2,1,4), pch=c(2,1,6))<br />
Pause()<br />
#Rem: <strong>the</strong> following code takes very long to run, so it is deactivated<br />
#just remove comment marks to run it...<br />
#cat("\n ---- Calculating parameters estimation for quantiles every 5% step ----\n\n")<br />
#cat("... it takes some time, please, be patient...\n")<br />
#edat0f.pars
SimRes
Thetas2[Thetas2 < 0.002] 0.998]
Annexes<br />
Code <strong>of</strong> functions required to run <strong>the</strong> script<br />
These functions perform various data manipulations and implement <strong>the</strong><br />
object-oriented R code for fitting envelope <strong>model</strong>s. Utilities to convert<br />
data, to generate artificial datasets and to run simulations are also<br />
provided.<br />
#===============================================================================#<br />
# #<br />
# Envelope <strong>Growth</strong> Models v. 1.0. #<br />
# #<br />
#===============================================================================#<br />
#<br />
# by Ph. Grosjean, 2001 (phgrosjean@sciviews.org)<br />
#<br />
# Parameters estimation, analyses and simulations <strong>of</strong> envelope growth <strong>model</strong>s<br />
# including <strong>the</strong> fuzzy-remanent growth curve presented in:<br />
# Grosjean, Ph., Ch. Spirlet & M. Jangoux, A functional growth <strong>model</strong> with<br />
# intraspecific competition applied to <strong>sea</strong> <strong>urchin</strong>s (in preparation).<br />
# Regression methods include nonlinear quantile regression with an interior<br />
# point algorithm <strong>of</strong> Koenker, and a custom nonlinear quantile regression<br />
# over <strong>the</strong> whole quantile range 0 -> 1 specifically developped for envelope<br />
# growth <strong>model</strong>s (<strong>model</strong>s that envelop size distributions with time).<br />
#<br />
# This is a free s<strong>of</strong>tware distributed under <strong>the</strong> terms <strong>of</strong> <strong>the</strong> GNU Public<br />
# License version 2 or above at your convenience (see licence.txt).<br />
#<br />
#This version is still in development. Use at your own risks!<br />
# To run <strong>the</strong>se samples:<br />
# Source this file in R, version 1.3.0 or higher<br />
# Change current directory to <strong>the</strong> one containing Plividus.txt and <strong>Growth</strong>.R<br />
# and enter source("<strong>Growth</strong>.R", print.eval=TRUE) in R<br />
#(see <strong>Growth</strong>.R for more details)<br />
###==== Basic data manipulation ===========================================<br />
# Data are issued from Plividus.txt and stored in variable 'dat':<br />
# dat is: - col #1: class (1 by 1 mm in test diameter)<br />
# - col #2: mean diameter for each class (ex: class 1 = 0 to 1 mm, mean diam.<br />
= 0.5 mm)<br />
# - cols #3 to 27: counts <strong>of</strong> individuals in each class at various increasing<br />
ages (Xxxx where xxx is age from fertilisation in days)<br />
## Transformations <strong>of</strong> raw data<br />
# Sum per column<br />
colsum
# o<strong>the</strong>rwise, <strong>the</strong> latter function must be overloaded (a better alternative => should be<br />
done later!!)<br />
expfreq
plot(cumsum(dat[,columns[1]])/sum(dat[,columns[1]])*100, type="s",<br />
col=cols[1],...)<br />
for (i in 1:length(columns-1)) {<br />
lines(cumsum(dat[,columns[i]]/sum(dat[,columns[i]])*100),<br />
type="s", col=cols[i])<br />
}<br />
} else { # Relative==FALSE<br />
plot(cumsum(dat[,columns[1]]), type="s", col=cols[1],...)<br />
for (i in 1:length(columns-1)) {<br />
lines(cumsum(dat[,columns[i]]), type="s", col=cols[i])<br />
}<br />
}<br />
}<br />
###=== Basic functions for envelope <strong>model</strong> =================================<br />
# Data are stored in an EGMData presentation which is a list containing:<br />
# - x: <strong>the</strong> time values in day, from birthday (or metamorphosis day) being 0<br />
# time is assumed to be known without error (follow up <strong>of</strong> animal born<br />
# in captivity and tagged for instance).<br />
# - y: <strong>the</strong> vector <strong>of</strong> all classes upper bounds (lower bound <strong>of</strong> first class is<br />
# always 0, and <strong>of</strong> course, lower bound <strong>of</strong> following classes are upper<br />
# bounds <strong>of</strong> <strong>the</strong> preceeding classes). Classes limits are: ] x ].<br />
# y is <strong>the</strong> size increase, and is thus Y - Yini where Y is <strong>the</strong> measured<br />
# size and Yini is <strong>the</strong> initial size at birth<br />
# - freq: <strong>the</strong> table <strong>of</strong> all frequencies <strong>of</strong> individuals measured at time t and<br />
# whose y is included in <strong>the</strong> corresponding class y<br />
# - n: <strong>the</strong> vector <strong>of</strong> <strong>the</strong> number <strong>of</strong> individuals counted in <strong>the</strong> batch at each<br />
# sampled time. It only correspond to <strong>the</strong> number <strong>of</strong> individuals in freq<br />
# if ALL <strong>the</strong> batch is measured at each sampled time (if possible!).<br />
# - sizes: (facultative) <strong>the</strong> table <strong>of</strong> all measured sizes (or its approximation<br />
# as back-calculated from freq table<br />
# - units: a list with time and size units, ex: c("days", "mm")<br />
# - desc: (facultative) a description <strong>of</strong> <strong>the</strong> data set<br />
# - info: (facultative) information about sizes. How it was obtained (measured<br />
# or back-calculated (and with which algorithm), and do <strong>the</strong> data in a<br />
# row correspond to <strong>the</strong> same animal (individual tagging) or not.<br />
# An example dataset is a whole batch <strong>of</strong> <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s <strong>Paracentrotus</strong><br />
# lividus grown in aquaculture, and is presented in Grosjean et al (in prep.)<br />
# The next function create <strong>the</strong> EGMData corresponding to this dataset<br />
# Usage: Pl
EGMData.Sizescalc
for <strong>the</strong> CI!!!<br />
Annexes<br />
SEMean
### Some additional graphes to visualize raw data<br />
# Plot <strong>the</strong> mortality curve for <strong>the</strong> data<br />
plotmort
}<br />
Annexes<br />
# - y, a vector <strong>of</strong> classes upper bound <strong>of</strong> length j<br />
# - freq, a matrix <strong>of</strong> i columns and j rows with<br />
# frequencies in each class at each time<br />
#<br />
# Return a list with:<br />
# - xtable being x repeated along j rows<br />
# - ytable being y repeated along i columns<br />
# - freq as in freq input<br />
# - <strong>the</strong>ta being quantiles for each distribution<br />
# at each class upper bound and at each time<br />
# Rem: does not check sizes <strong>of</strong> x, y and freq!!!<br />
i
###===Plots for data visualisation=============================================<br />
# Plot an "image" <strong>of</strong> <strong>the</strong> frequency table (color <strong>of</strong> each cell according to<br />
log(frequency))<br />
# Note: one can superpose data points and <strong>model</strong> lines on this graph using adddual=T.<br />
In this case only, fuzres must be provided<br />
plotimage
plot(x, y, type="l", col=col, lty=lty, xlab="Time (in days)",<br />
ylab="Size (in mm)", main=paste("<strong>Growth</strong> <strong>model</strong> for quantile", <strong>the</strong>ta))<br />
}<br />
}<br />
# Plot original points corresponding to a quantile.<br />
# This function requires fuzdata$sizes, obtained using fuzgenerate(ReturnSizes=T)<br />
plotpoints
sizes
}<br />
Annexes<br />
}<br />
results[i, 1:5]
# library. See <strong>the</strong> R documentation and also:<br />
# Pinherio, J.C. & D.M. Bates, 2000. Mixed-effects <strong>model</strong>s in S and Splus.<br />
# Springer, New York. Appendix C, p. 511-521.<br />
# Since R v. 1.2.3 <strong>the</strong>re is also SSweibull and SSgompertz in nls library<br />
library(nls)<br />
# This is for <strong>the</strong> artificial data generators<br />
AddError
}<br />
.value<br />
Annexes<br />
dimnames(.grad)
lrc
#{<br />
# # This is adapted from SSasympOrig<br />
# xy
.actualArgs
Exp.ival
SSexpAB
SSallo
}<br />
.value<br />
Annexes<br />
.grad[, "y0"]
.expr4
.expr4
.expr9
.expr9
. The 'nlrq' package for nonlinear quantile regression<br />
Annexes<br />
This package was developed by R. Koenker and Ph. Grosjean and it is<br />
now available as part <strong>of</strong> <strong>the</strong> <strong>of</strong>ficial distribution <strong>of</strong> <strong>the</strong> R s<strong>of</strong>tware.<br />
Description file<br />
Package: nlrq<br />
Version: 0.1-1<br />
Date: 2000/05/14<br />
Title: Nonlinear quantile regression<br />
Author: Roger Koenker ,<br />
Philippe Grosjean <br />
Maintainer: Roger Koenker<br />
Depends: R (>= 1.2.3)<br />
Description: Nonlinear quantile regression routines<br />
License: GPL version 2 or later<br />
URL: http://www.econ.uiuc.edu/~roger/re<strong>sea</strong>rch/nlrq/nlrq.html<br />
This package contains functions and methods for nonlinear quantile regression<br />
coef.nlrq extract coefficients<br />
deviance.nlrq deviance at solution<br />
fitted.nlrq response <strong>of</strong> <strong>the</strong> fitted <strong>model</strong><br />
formula.nlrq formula used in <strong>the</strong> nlrq object<br />
nlrq nonlinear quantile regression<br />
nlrq.control construct a control list for using with nlrq<br />
predict.nlrq predict data according to <strong>the</strong> <strong>model</strong><br />
residuals.nlrq extract residuals<br />
summary.nlrq display summary <strong>of</strong> an nlrq object<br />
tau.nlrq quantile used in <strong>the</strong> nlrq object<br />
Online manual pages<br />
nlrq package:nlrq R Documentation<br />
Function to compute nonlinear quantile regression estimates<br />
Description:<br />
Usage:<br />
This function implements an R version <strong>of</strong> an interior point method<br />
for computing <strong>the</strong> solution to quantile regression problems which<br />
are nonlinear in <strong>the</strong> parameters. The algorithm is based on<br />
interior point ideas described in Koenker and Park (1994).<br />
nlrq(formula, data=parent.frame(), start, tau=0.5, control, trace=FALSE)<br />
Arguments:<br />
formula: formula for <strong>model</strong> in nls format; accept self-starting <strong>model</strong>s<br />
data: an optional data frame in which to evaluate <strong>the</strong> variables in<br />
`formula'<br />
start: a named list or named numeric vector <strong>of</strong> starting estimates<br />
tau: a vector <strong>of</strong> quantiles to be estimated<br />
control: an optional list <strong>of</strong> control settings. See `nlrq.control' for<br />
246
Annexes<br />
<strong>the</strong> names <strong>of</strong> <strong>the</strong> settable control values and <strong>the</strong>ir effect.<br />
trace: logical value indicating if a trace <strong>of</strong> <strong>the</strong> iteration progress<br />
should be printed. Default is `FALSE'. If `TRUE'<br />
intermediary results are printed at <strong>the</strong> end <strong>of</strong> each<br />
iteration.<br />
Details:<br />
Value:<br />
An `nlrq' object is a type <strong>of</strong> fitted <strong>model</strong> object. It has methods<br />
for <strong>the</strong> generic functions `coef' (parameters estimation at best<br />
solution), `formula' (<strong>model</strong> used), `deviance' (value <strong>of</strong> <strong>the</strong><br />
objective function at best solution), `print', `summary',<br />
`fitted' (vector <strong>of</strong> fitted variable according to <strong>the</strong> <strong>model</strong>),<br />
`predict' (vector <strong>of</strong> data points predicted by <strong>the</strong> <strong>model</strong>, using a<br />
different matrix for <strong>the</strong> independent variables) and also for <strong>the</strong><br />
function `tau' (quantile used for fitting <strong>the</strong> <strong>model</strong>, as <strong>the</strong> tau<br />
argument <strong>of</strong> <strong>the</strong> function). Fur<strong>the</strong>r help is also available for <strong>the</strong><br />
method `residuals'.<br />
A list consisting <strong>of</strong>:<br />
m: an `nlrqModel' object similar to an `nlsModel' in package nls<br />
data: <strong>the</strong> expression that was passed to `nlrq' as <strong>the</strong> data<br />
argument. The actual data values are present in <strong>the</strong><br />
environment <strong>of</strong> <strong>the</strong> `m' component.<br />
Author(s):<br />
Based on S code by Roger Koenker modified for R and to accept same<br />
<strong>model</strong>s as nls by Philippe Grosjean.<br />
References:<br />
See Also:<br />
Examples:<br />
Koenker, R. and Park, B.J. (1994). An Interior Point Algorithm for<br />
Nonlinear Quantile Regression, Journal <strong>of</strong> Econometrics, 71(1-2):<br />
265-283.<br />
`nlrq.control' , `residuals.nlrq'<br />
# Example using <strong>model</strong> defined in <strong>the</strong> nls library<br />
library(nls)<br />
# build artificial data with multiplicative error<br />
Dat
nlrq.control package:nlrq R Documentation<br />
Set control parameters for nlrq<br />
Description:<br />
Usage:<br />
Annexes<br />
Set algorithmic parameters for nlrq (nonlinear quantile regression<br />
function)<br />
nlrq.control(maxiter=100, k=2, big=1e+20, eps=1e-07, beta=0.97)<br />
Arguments:<br />
maxiter: maximum number <strong>of</strong> allowed iterations<br />
k: <strong>the</strong> number <strong>of</strong> iterations <strong>of</strong> <strong>the</strong> Meketon algorithm to be<br />
calculated in each step, usually 2 is reasonable,<br />
occasionally it may be helpful to set k=1<br />
big: a large scalar<br />
eps: tolerance for convergence <strong>of</strong> <strong>the</strong> algorithm<br />
beta: a shrinkage parameter which controls <strong>the</strong> recentering process<br />
in <strong>the</strong> interior point algorithm.<br />
See Also:<br />
`nlrq'<br />
residuals.nlrq package:nlrq R Documentation<br />
Return residuals <strong>of</strong> an nlrq object<br />
Description:<br />
Usage:<br />
Get residuals from an nlrq (nonlinear quantile regression) object<br />
residuals.nlrq(nlrqObject, type = c("response", "rho"), ...)<br />
Arguments:<br />
nlrqObject: an `nlrq' object as returned by function `nlrq'<br />
type: <strong>the</strong> type <strong>of</strong> residuals to return: "response" is <strong>the</strong> distance<br />
between observed and predicted values; "rho" is <strong>the</strong> weighted<br />
distance used to calculate <strong>the</strong> objective function in <strong>the</strong><br />
minimisation algorithm as tau * pmax(resid, 0) + (1 - tau) *<br />
pmin(resid, 0), where resid are <strong>the</strong> simple residuals as above<br />
(with type="response").<br />
See Also:<br />
`nlrq'<br />
R code<br />
###===Nonlinear quantile regression with an interior point algorithm===<br />
# see: Koenker, R. & B.J. Park, 1996. An interior point algorithm<br />
# for nonlinear quantile regression. J. Econom., 71(1-2): 265-283.<br />
# adapted from nlrq routine <strong>of</strong> Koenker, R.<br />
# to be compatible with R nls <strong>model</strong>s<br />
# by Ph. Grosjean, 2001 (phgrosjean@sciviews.org)<br />
# large parts <strong>of</strong> code are reused from <strong>the</strong> nls library <strong>of</strong> R v. 1.2.3<br />
248
# It is made available under <strong>the</strong> terms <strong>of</strong> <strong>the</strong> GNU General Public<br />
# License, version 2, or at your option, any later version<br />
#<br />
# This program is distributed in <strong>the</strong> hope that it will be<br />
# useful, but WITHOUT ANY WARRANTY; without even <strong>the</strong> implied<br />
# warranty <strong>of</strong> MERCHANTABILITY or FITNESS FOR A PARTICULAR<br />
# PURPOSE. See <strong>the</strong> GNU General Public License for more details.<br />
#<br />
# You should have received a copy <strong>of</strong> <strong>the</strong> GNU General Public<br />
# License along with this program; if not, write to <strong>the</strong> Free<br />
# S<strong>of</strong>tware Foundation, Inc., 59 Temple Place - Suite 330, Boston,<br />
# MA 02111-1307, USA<br />
# TO DO:<br />
# - nlrq should return a code 0 = convergence, 1 = lambda -> 0, etc..<br />
# - Extensive diagnostic for summary() (Roger, what would you propose?)<br />
# - Calculate with a list <strong>of</strong> tau values at once (currently accept 1 value)<br />
# - When providing several tau values, allow calculating a single value<br />
# for one or more parameters across all <strong>model</strong>s fitted to all tau values<br />
# ...but I have ano<strong>the</strong>r idea for doing that more efficiently.<br />
"nlrq.control"
gradSetArgs[[2]]), call("[", gradSetArgs[[1]], gradSetArgs[[2]],<br />
gradSetArgs[[2]]), call("[", gradSetArgs[[1]], gradSetArgs[[2]],<br />
gradSetArgs[[2]], gradSetArgs[[3]]), call("[", gradSetArgs[[1]],<br />
gradSetArgs[[2]], gradSetArgs[[2]], gradSetArgs[[3]],<br />
gradSetArgs[[4]]))<br />
getRHS.varying
}<br />
": ", format(getPars()), "\n"), Rmat = function() qr.R(QR),<br />
predict = function(newdata = list(), qr = FALSE) {<br />
Env
<strong>model</strong>.step
object$m$deviance()<br />
"tau.nlrq"
Annexes<br />
254
Annex II: dataset <strong>of</strong> <strong>the</strong> cohort measured during seven years<br />
Age<br />
(days)<br />
Annexes<br />
210<br />
306<br />
364<br />
456<br />
546<br />
636<br />
726<br />
818<br />
911<br />
1006<br />
Size class<br />
(mm)<br />
Count <strong>of</strong> <strong>the</strong> no. <strong>of</strong> individuals in each size class<br />
0 – 1 65 5<br />
1 – 2 142 30 11<br />
2 – 3 114 29 22<br />
3 – 4 115 46 19<br />
4 – 5 65 45 27 3<br />
5 - 6 62 50 32 3<br />
6 - 7 31 47 42 10<br />
7 - 8 34 56 46 13<br />
8 - 9 31 31 44 18 2<br />
9 - 10 20 27 38 17 4<br />
10 - 11 17 22 28 28 6<br />
11 - 12 18 19 22 26 10 1<br />
12 - 13 7 21 25 33 22 2<br />
13 - 14 4 20 16 33 31 8 2<br />
14 - 15 12 16 21 30 13<br />
15 - 16 10 15 29 35 23 3<br />
16 - 17 12 23 37 29 26 10 1<br />
17 - 18 10 15 26 35 45 6 1<br />
18 - 19 11 13 26 25 29 12<br />
19 - 20 4 11 19 31 28 20 4 1<br />
20 - 21 5 16 25 27 25 6<br />
21 - 22 6 22 14 27 26 8 2<br />
22 - 23 10 19 21 29 36 19 1 1 1<br />
23 - 24 5 13 27 20 36 15 5<br />
24 - 25 15 19 22 35 31 6 1<br />
25 - 26 4 17 19 10 23 15 2 1<br />
26 - 27 6 7 21 23 28 18 7<br />
27 - 28 9 8 21 27 28 15 5 2 1 1<br />
28 - 29 5 12 6 18 29 12 6 1<br />
29 - 30 8 3 5 18 27 23 13 4 1<br />
30 - 31 3 8 8 16 19 29 14 9<br />
31 - 32 3 7 8 7 17 22 20 16 1 1<br />
32 - 33 2 6 9 4 17 33 18 9 2<br />
33 - 34 8 4 7 15 23 16 16 3 1<br />
34 - 35 5 5 7 12 17 16 9 3<br />
35 - 36 7 4 10 9 16 16 13 10 1<br />
36 - 37 4 10 9 8 19 26 17 10 1 1<br />
37 - 38 3 9 8 8 11 28 15 8 1 1 1<br />
38 - 39 5 7 10 15 20 13 9 5 1 1 1 1 1 1 1<br />
39 - 40 6 10 12 29 28 10 5 1<br />
40 - 41 2 11 6 13 9 21 16 9 10 1<br />
41 - 42 1 1 12 3 14 22 12 8 5 8<br />
42 - 43 2 6 16 10 19 37 19 8 3 1<br />
43 - 44 8 11 18 14 25 16 7 1 5 1 1<br />
44 - 45 1 7 12 17 17 30 17 15 15 4 6 2<br />
45 - 46 1 13 10 12 21 31 8 13 6 3 3 2 1<br />
46 - 47 2 10 9 13 19 28 16 15 11 7 2 1 2 1 1<br />
47 - 48 7 15 10 12 24 28 16 16 8 2 4 3 3 1<br />
48 - 49 3 17 8 20 30 23 9 13 9 6 3 2 1<br />
49 - 50 2 7 11 14 22 33 34 17 12 10 1 2 2<br />
50 - 51 1 2 14 5 16 24 22 19 21 8 5 3 1 2<br />
51 - 52 2 2 8 12 15 16 22 26 25 10 6 4 1 2<br />
52 - 53 15 8 17 16 22 28 14 8 7 5 2<br />
53 - 54 1 10 13 10 6 14 13 16 12 1 4 3<br />
54 - 55 1 7 11 12 14 15 21 19 14 13 8 12<br />
55 - 56 1 1 12 13 13 10 17 12 10 12 13 5<br />
56 - 57 1 8 11 8 8 9 9 4 7 7 11<br />
57 - 58 4 7 7 9 9 6 6 8 13 8<br />
58 - 59 1 4 2 6 5 5 6 6 7 5<br />
59 - 60 2 2 4 10 11 4 3 5 5 5<br />
60 - 61 1 2 3 7 1 3 4 2<br />
61 - 62 1 5 4 5 3 1 2 3<br />
62 - 63 1 1 2 1 1 1 1 1<br />
63 - 64 1<br />
64 - 65 1 2 1 2 1<br />
65 - 66 1 2 1<br />
66 - 67 1<br />
Total no. 725 507 491 467 461 437 403 386 387 371 324 315 309 275 233 223 221 137 92 85 82 67<br />
1097<br />
1188<br />
1281<br />
1369<br />
1461<br />
1551<br />
1642<br />
1825<br />
2008<br />
2190<br />
2377<br />
255<br />
2554
Annexes<br />
256
Annex III: abstracts <strong>of</strong> publications and symposia<br />
Annexes<br />
Since my scientific activity was very diversified, a part <strong>of</strong> my work<br />
was included in this <strong>the</strong>sis. Here are <strong>the</strong> abstracts <strong>of</strong> all publications and<br />
symposia where I have participated.<br />
a. International journals<br />
Vaïtilingon, D., R. Morgan, Ph. Grosjean, P. Gosselin & M. Jangoux.<br />
Influence <strong>of</strong> delayed metamorphosis and food intake on <strong>the</strong><br />
perimetamorphic period <strong>of</strong> <strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus. J. Exp.<br />
Mar. Biol. Ecol., 262(1):41-60.<br />
ABSTRACT: Effect <strong>of</strong> delayed metamorphosis and food ration on late<br />
(competent) larvae and postlarvae <strong>of</strong> <strong>Paracentrotus</strong> lividus were<br />
investigated. Metamorphosis <strong>of</strong> competent larvae was ei<strong>the</strong>r not delayed or<br />
delayed from 1 up to 4 days. Larvae were starved or submitted to two<br />
different food rations <strong>of</strong> <strong>the</strong> algal species Phaeodactylum tricornutum.<br />
Larvae during <strong>the</strong> prolonged competence period and <strong>the</strong> resulting<br />
postlarvae were characterized by: (1) <strong>the</strong> size <strong>of</strong> <strong>the</strong> larval body, (2) <strong>the</strong><br />
size <strong>of</strong> <strong>the</strong> rudiment, (3) <strong>the</strong> rate <strong>of</strong> metamorphosis, (4) <strong>the</strong> size <strong>of</strong><br />
postlarvae 24 h after metamorphosis, (5) <strong>the</strong> rate <strong>of</strong> opening <strong>of</strong> mouth and<br />
anus, (6) <strong>the</strong> rate <strong>of</strong> survival, and (7) <strong>the</strong> growth rate <strong>of</strong> early<br />
postmetamorphic individuals. Both <strong>the</strong> width <strong>of</strong> <strong>the</strong> larval body and <strong>the</strong><br />
diameter <strong>of</strong> <strong>the</strong> echinus rudiment grew in competent larvae that were fed.<br />
Unfed larvae did not grow. There was no significant difference in growth<br />
between <strong>the</strong> two food rations. The rate <strong>of</strong> metamorphosis was higher with<br />
larvae that metamorphosed soon after <strong>the</strong>y became competent. Lower<br />
capacity <strong>of</strong> larvae to metamorphose during <strong>the</strong> delay period was associated<br />
with treatments yielding a greater larval width and rudiment diameter<br />
during <strong>the</strong> same period. Postlarval development was affected by a delayed<br />
metamorphosis treatment inflicted on competent larvae before<br />
257
Annexes<br />
metamorphosis. Acquisition <strong>of</strong> exotrophy happened earlier in postlarvae<br />
issued from prolonged competent larvae whatever <strong>the</strong> larval food rations.<br />
The delay treatment negatively affected <strong>the</strong> development <strong>of</strong> <strong>the</strong> digestive<br />
tract through it positively affected <strong>the</strong> growth <strong>of</strong> early postmetamorphic<br />
individuals during <strong>the</strong> first 6 days following metamorphosis. However,<br />
selective mortality occurred afterwards as bigger individuals died<br />
preferentially.<br />
KEYWORDS: Larvae, metamorphosis, <strong>Paracentrotus</strong> lividus, plutei,<br />
<strong>sea</strong> <strong>urchin</strong>.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 2000. Cultivation <strong>of</strong><br />
<strong>Paracentrotus</strong> lividus (Echinodermata: Echinoidea) fed extruded feeds:<br />
digestion efficiency, somatic production and gonadal growth.<br />
Aquaculture Nutrition, 7(2):91-99.<br />
ABSTRACT: This study assessed <strong>the</strong> use <strong>of</strong> extruded feeds, in <strong>the</strong><br />
form <strong>of</strong> pellets, for growing <strong>of</strong> <strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus within a<br />
closed culture system. Two feeds types, one with soybean protein, <strong>the</strong><br />
o<strong>the</strong>r with both soybean and fish protein were compared to dried Lessonia<br />
sp. and fresh Laminaria sp. as food sources. Pellets present a very high<br />
conversion efficiency (about 80%) against about 50% for Laminaria and<br />
35% for Lessonia. However, since pellets are less absorbed, somatic<br />
growth is statistically equivalent for <strong>the</strong> <strong>sea</strong> <strong>urchin</strong>s fed with pellets and<br />
Laminaria: between 2 and 2.2%.g <strong>of</strong> soma.day -1 . Sea <strong>urchin</strong>s fed pellets<br />
produced significantly more gonadal tissue in a shorter time. Resulting in a<br />
gonadal index twice higher (6.5%) than Laminaria (3%) in <strong>the</strong> second<br />
month <strong>of</strong> <strong>the</strong> experiment. Dry Lessonia does not promote gonadal growth.<br />
This study shows that extruded feeds are well assimilated by P. lividus and<br />
promote both somatic growth and production <strong>of</strong> gonadal tissue.<br />
258
Annexes<br />
KEYWORDS: Sea <strong>urchin</strong>, aquaculture, artificial food, somatic<br />
growth, roe, digestion.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 2000. Optimization <strong>of</strong> gonad<br />
growth by manipulation <strong>of</strong> temperature and photoperiod in cultivated<br />
<strong>sea</strong> <strong>urchin</strong>, <strong>Paracentrotus</strong> lividus (Lamarck) (Echinodermata).<br />
Aquaculture, 185:85-99.<br />
ABSTRACT: A starvation and <strong>the</strong>n feeding method was developed to<br />
produce about 100% marketable <strong>sea</strong> <strong>urchin</strong>s, <strong>Paracentrotus</strong> lividus, in 3 ½<br />
months. This method is needed because <strong>the</strong> reproduction cycle is<br />
desynchronized in <strong>the</strong> conditions imposed during <strong>the</strong> somatic growth stage<br />
in land-based closed systems. The major advantages <strong>of</strong> starving <strong>the</strong><br />
animals are resetting <strong>the</strong> reproductive cycle to <strong>the</strong> spend stage (gonads<br />
almost devoid <strong>of</strong> sexual cells) and stressing <strong>the</strong> individuals so that <strong>the</strong>y<br />
mobilize and restore <strong>the</strong> nutritive phagocytes, filling <strong>the</strong>m with nutrients.<br />
Batches <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s starved for 2 months beforehand were fed ad<br />
libitum for 45 days with enriched food under eight combinations <strong>of</strong> four<br />
temperatures (12°C, 16°C, 20°C and 24°C) and two photoperiods (9 and<br />
17 h daylight). In our system, <strong>the</strong> best combination was 24°C and 9 h<br />
daylight for growth as well as for gonad quality. The gonadal indices<br />
obtained (in dry weight) were over 9% at 16°C and over 12% at 24°C,<br />
which are better than what is found in <strong>the</strong> field for this population.<br />
KEYWORDS: Gonad, growth, temperature, photoperiod, <strong>sea</strong> <strong>urchin</strong>,<br />
<strong>Paracentrotus</strong> lividus.<br />
Van Osselaer, Ch. & Ph. Grosjean, 2000. Suture and location <strong>of</strong> <strong>the</strong><br />
coiling axis in gastropod shells. Paleobiology, 26(2):238-257.<br />
259
Annexes<br />
ABSTRACT: The general allometric equations for <strong>the</strong> logarithmic<br />
helicospiral can fit many extraconical shapes, but <strong>the</strong> isometric conditions<br />
traditionally used limits study only to conical growth. We present evidence<br />
to show that in real gastropod shells, <strong>the</strong> logarithmic helicospiral equations<br />
fit <strong>the</strong> suture. Poor location <strong>of</strong> <strong>the</strong> coiling axis and/or an inappropriate pole<br />
for <strong>the</strong> logarithmic helicospiral has <strong>of</strong>ten led to <strong>the</strong> rejection <strong>of</strong> this <strong>model</strong>.<br />
The differences between <strong>the</strong> errors associated with measurements or<br />
previously available <strong>model</strong>s, are discussed. Two methods, based on suture<br />
trace measurements, are proposed to locate <strong>the</strong> coiling axis both in apical<br />
and lateral views. The first is a graphical method based on an elementary<br />
property <strong>of</strong> <strong>the</strong> logarithmic spiral. The second, computational method is<br />
based on iterative reprojections <strong>of</strong> <strong>the</strong> suture. It is shown that <strong>the</strong><br />
protoconch and <strong>the</strong> teleloconch must be treated separately. The precision<br />
<strong>of</strong> <strong>the</strong> new methods (especially <strong>the</strong> computing method) enables deviations<br />
from logarithmic helicospiral trajectory to be identified and differentiated<br />
from irregularities <strong>of</strong> <strong>the</strong> shell and sequential growth phases. Application<br />
<strong>of</strong> <strong>the</strong>se methods may be useful not only for o<strong>the</strong>r gastropod<br />
morphological features, but also for o<strong>the</strong>r taxa such as brachiopods and<br />
o<strong>the</strong>r molluscs.<br />
KEYWORDS: Coiled shell, coiling axis, gastropod, mollusc,<br />
morphometry, suture.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 1998a. Reproductive cycle<br />
<strong>of</strong> <strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus: analysis by means <strong>of</strong> <strong>the</strong> maturity<br />
index. Invert. Reprod. Develop., 34(1):69-81.<br />
ABSTRACT: The gonad maturity index cycles <strong>of</strong> <strong>the</strong> echinoid<br />
<strong>Paracentrotus</strong> lividus and <strong>the</strong>ir relations with environmental abiotic<br />
parameters are assessed after 2 years <strong>of</strong> observation in sou<strong>the</strong>rn Brittany,<br />
France. The gonadal cycle is briefly described and eight gonadal stages are<br />
characterized. The annual cycle, <strong>the</strong> time <strong>of</strong> spawning and <strong>the</strong> period <strong>of</strong><br />
260
Annexes<br />
gonadal growth are well established, suggesting <strong>the</strong>y are controlled<br />
externally. The reproductive cycle has three main phases: <strong>the</strong> growing<br />
phase (late autumn and winter) when gonads accumulate reserve material;<br />
<strong>the</strong> maturation phase (spring and early summer) in which gametogenesis<br />
<strong>the</strong>n spawning take place; and <strong>the</strong> spent/regenerating phase when relict<br />
gametes are resorbed by <strong>the</strong> nutritive phagocytes, <strong>the</strong> gonads being<br />
virtually devoid <strong>of</strong> sexual cells. The maturity index based on <strong>the</strong><br />
histological diagnosis <strong>of</strong> gonads and <strong>the</strong> use <strong>of</strong> circular data and polar<br />
graphical representation make it possible to reliably determine <strong>the</strong><br />
spawning period, <strong>the</strong> rate <strong>of</strong> gametogenesis and <strong>the</strong> synchronization <strong>of</strong><br />
males and females among <strong>the</strong> echinoid population. From this analysis, we<br />
can reasonably say that <strong>the</strong> gonadal cycle (represented by <strong>the</strong> gonad<br />
index), <strong>the</strong> rate <strong>of</strong> gametogenesis, and <strong>the</strong> end <strong>of</strong> <strong>the</strong> spawning period are<br />
influenced by temperature whereas <strong>the</strong> first spawning event appears to be<br />
triggered by day length.<br />
KEYWORDS: Echinoid, reproduction, maturity index, gonadal cycle,<br />
abiotic parameters.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 1998b. Optimizing Food<br />
distribution in closed-circuit cultivation <strong>of</strong> edible <strong>sea</strong> <strong>urchin</strong>s<br />
(<strong>Paracentrotus</strong> lividus: Echinoidea). Aquat. Living Resour., 11(4):273-<br />
277.<br />
ABSTRACT: In <strong>the</strong> framework <strong>of</strong> echinoid cultivation, whose<br />
objective is to succeed in continuously producing large amounts <strong>of</strong> edible<br />
<strong>sea</strong> <strong>urchin</strong>s (Paracentroutus lividus) under controlled conditions<br />
(aquaculture), gonadal growth is to be optimized. Among <strong>the</strong> various<br />
parameters influencing <strong>the</strong> production <strong>of</strong> roe, <strong>the</strong> quantity <strong>of</strong> food<br />
distributed was tested for optimization. After a 1-month fast, echinoids<br />
were fed artificial food pellets (enriched in soybean and fish proteins) for<br />
different periods <strong>of</strong> time over 48 h, <strong>the</strong> food thus being available ad<br />
261
Annexes<br />
libitum for 8, 16, 24, 32, 40 and 48 h; <strong>the</strong> cycles were repeated for a<br />
month. The results show that <strong>the</strong> quantity <strong>of</strong> food intake and <strong>the</strong> gonad<br />
index peak after about 35 h <strong>of</strong> food availability. This suggests food should<br />
be distributed discontinuously for optimal gonad production and minimal<br />
waste.<br />
KEYWORDS: Aquaculture, food ration, gonad growth, artificial diet,<br />
<strong>sea</strong> <strong>urchin</strong>.<br />
Grosjean, Ph., Ch. Spirlet, P. Gosselin, D. Vaïtilingon & M. Jangoux,<br />
1998. Land-based closed cycle echiniculture <strong>of</strong> <strong>Paracentrotus</strong> lividus<br />
(Lamarck) (Echinoidea: Echinodermata): a long-term experiment at a<br />
pilot scale. J. Shellfish Res., 17(5):1523-1531.<br />
See Part I.<br />
Grosjean, Ph., Ch. Spirlet & M. Jangoux, 1996. Experimental study <strong>of</strong><br />
growth in <strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus (Lamarck, 1816)<br />
(Echinodermata). J. Exp. Mar. Biol. Ecol., 201:173-184.<br />
See Part III.<br />
b. Reports and o<strong>the</strong>r publications<br />
Jangoux, M., Ph. Grosjean, D. Vaïtilingon, C. Cam, J. Cosson, Ch.<br />
Billard, D. Bucaille, J.M. Ouin, C. Rebours, N.T. Hagen, C. Solberg &<br />
H.H. Ludvigsen, 2000. Biology <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s under intensive<br />
cultivation (closed-cycle echiniculture). European Contract FAIR<br />
CT96-1623 BFN, final report. 163 pp.<br />
262
Annexes<br />
EXECUTIVE SUMMARY: The ultimate objective <strong>of</strong> <strong>the</strong> project is to<br />
control every life stage <strong>of</strong> <strong>the</strong> most valuable species <strong>of</strong> European edible <strong>sea</strong><br />
<strong>urchin</strong>s (<strong>Paracentrotus</strong> lividus and Strongylocentrotus droebachiensis)<br />
under intensive cultivation (closed-cycle echiniculture) to produce high<br />
quality gonads (roe, i.e., <strong>the</strong> edible part <strong>of</strong> <strong>the</strong> animal) at a pilot scale. The<br />
obstacles that prevent <strong>the</strong> intensification <strong>of</strong> echiniculture have been clearly<br />
identified: (1) post-settlement survival and growth rate need to be<br />
improved, and (2) <strong>the</strong> carrying capacity <strong>of</strong> <strong>the</strong> rearing structures needs to<br />
be increased by bypassing main limiting factors, i.e., depletion in dissolved<br />
carbonates and accumulation <strong>of</strong> carbonic acid. Moreover, <strong>the</strong> quality<br />
control <strong>of</strong> gonads and optimization <strong>of</strong> gonad growth are key factors that<br />
have to be addressed. The proposed work aims to investigate aspects <strong>of</strong> <strong>the</strong><br />
biology <strong>of</strong> cultivated <strong>sea</strong> <strong>urchin</strong>s related to <strong>the</strong>se obstacles, to finalize<br />
technical enhancements <strong>of</strong> <strong>the</strong> cultivation procedure in ei<strong>the</strong>r eliminating<br />
or bypassing <strong>the</strong>se obstacles, and to adapt <strong>the</strong> rearing method presently<br />
used for <strong>Paracentrotus</strong> lividus to Strongylocentrotus droebachiensis.<br />
Grosjean, Ph., Ch. Spirlet & M. Jangoux, 1999. Comparison <strong>of</strong> three<br />
body-size measurements for echinoids. In: M.D. Candia Carnevali &<br />
F. Bonasoro (eds). Echinoderm Re<strong>sea</strong>rch 1998, Balkema, Rotterdam.<br />
Pp. 31-35.<br />
See Part II.<br />
Jangoux, M., P. Gosselin, Ph. Grosjean, M. Larsonneur, Ch. Spirlet,<br />
D. Bucaille, M. Catoira Gomez, 1996. Sea-<strong>urchin</strong> cultivation.<br />
European Contract FAR AQ 2.530 BFE, final report. 103 pp +<br />
annexes.<br />
263
Annexes<br />
EXECUTIVE SUMMARY: The aim <strong>of</strong> <strong>the</strong> re<strong>sea</strong>rch is to succeed in<br />
continuously producing large amount <strong>of</strong> edible <strong>sea</strong> <strong>urchin</strong>s with high<br />
gonadal productivity under controlled conditions (aquaculture). The<br />
selected species is <strong>Paracentrotus</strong> lividus (viz. <strong>the</strong> edible echinoid species<br />
from Europe). The re<strong>sea</strong>rch focuses on <strong>the</strong> following aspects:<br />
a. Investigation on metamorphic events, with a special interest in<br />
<strong>the</strong> morphological changes affecting metamorphic larvae and in<br />
<strong>the</strong> metamorphosis inducing factors (this approach needs to<br />
perform routinely mass-cultivation <strong>of</strong> larvae);<br />
b. Optimization <strong>of</strong> juveniles' somatic growth and adults' gonadal<br />
productivity under rearing conditions;<br />
c. Study <strong>of</strong> reproductive periodicities in field and cultivated<br />
individuals (investigations on parameters responsible for <strong>the</strong><br />
cycling <strong>of</strong> reproduction and <strong>the</strong> duration <strong>of</strong> <strong>the</strong> spawning period)<br />
and attempt for a continuous reproduction under aquaculture<br />
conditions.<br />
Spirlet, Ch., Ph. Grosjean & M. Jangoux, 1994. Differentiation <strong>of</strong> <strong>the</strong><br />
genital apparatus in a juvenile echinoid (<strong>Paracentrotus</strong> lividus). In: B.<br />
David, A. Guille, J.-P. Féral & M. Roux (eds). Echinoderms through<br />
Time, Balkema, Rotterdam. Pp. 881-886.<br />
ABSTRACT: Gonad and genital pore development was observed on<br />
field and cultivated juveniles <strong>of</strong> <strong>the</strong> echinoid <strong>Paracentrotus</strong> lividus. The<br />
gonad condition was evaluated by counting <strong>the</strong> number <strong>of</strong> acini per gonad<br />
following dissection. Presence <strong>of</strong> <strong>the</strong> genital pores was determined for<br />
each individual, plate by plate, after partial digestion <strong>of</strong> <strong>the</strong> tissues.<br />
Progressive and relative development was determined. The gonads first<br />
appear as a filament which starts to bud and develop acini that eventually<br />
fill up with genital material. The pores are pierced from <strong>the</strong> inside out<br />
264
Annexes<br />
when <strong>the</strong> gonads have reached a certain growth stage. Both gonads and<br />
pores do not develop simultaneously but in a certain order. In addition,<br />
statistical analysis shows that size has more influence on <strong>the</strong> condition <strong>of</strong><br />
<strong>the</strong> genital apparatus than age.<br />
KEYWORDS: Echinoid, gonads, growth, reproduction, development.<br />
c. International symposia<br />
Green Sea Urchin Workshop, Moncton, Canada, 2000. Talk. Ph.<br />
Grosjean & M. Jangoux. <strong>Growth</strong> <strong>of</strong> <strong>the</strong> <strong>sea</strong> <strong>urchin</strong> <strong>Paracentrotus</strong><br />
lividus: <strong>model</strong> and optimization.<br />
Inline: http://crdpm.cus.ca/oursin/pdf/gros.pdf<br />
(last consulted Sept. 8 th 2001).<br />
ABSTRACT: Echiniculture, or <strong>sea</strong> <strong>urchin</strong> aquaculture, is more and<br />
more considered as an alternative or a complement to fisheries but it is not<br />
implemented yet on a commercial scale. This is because rearing methods<br />
still have to be optimized. We developed a ma<strong>the</strong>matical <strong>model</strong> to simulate<br />
and predict <strong>sea</strong> <strong>urchin</strong>s' production according to various rearing methods<br />
and exploitation strategies. Simulations using this <strong>model</strong> demonstrate how<br />
<strong>the</strong> variation <strong>of</strong> production has a complex and sometimes counterintuitive<br />
relationship with both <strong>the</strong> rearing method and <strong>the</strong> exploitation strategy.<br />
Intraspecific competition and spread in sizes in <strong>reared</strong> batches are taken<br />
into account in <strong>the</strong> <strong>model</strong>. Hence, an accurate estimation <strong>of</strong> <strong>the</strong> time<br />
required to grow a whole cohort <strong>of</strong> <strong>reared</strong> echinoids to <strong>the</strong> market size is<br />
obtained. This tool allows a rapid determination <strong>of</strong> best rearing methods<br />
and should speed up <strong>the</strong> optimization process. At a latter date, this <strong>model</strong><br />
could help in rationalizing stock management in future <strong>sea</strong> <strong>urchin</strong>s farming<br />
activities. By applying this <strong>model</strong> to field <strong>sea</strong> <strong>urchin</strong>s populations, it could<br />
also help in <strong>the</strong> establishment <strong>of</strong> sustainable fishery policies.<br />
265
Annexes<br />
KEYWORDS: Somatic growth, dynamic <strong>model</strong>, aquaculture,<br />
intraspecific competition.<br />
5 th European Echinoderm Colloquium, Milano, 1998. Poster. Ph.<br />
Grosjean, Ch. Spirlet & M. Jangoux. Comparison <strong>of</strong> three body-size<br />
measurements for echinoids and <strong>the</strong>ir use in growth and<br />
gonadosomatic calculations.<br />
See Part II.<br />
Aquaculture '98, Las Vegas, 1998. Talk in a special session. Ph.<br />
Grosjean, Ch. Spirlet & M. Jangoux. Is land-based closed cycle<br />
echiniculture (<strong>sea</strong> <strong>urchin</strong>s aquaculture) a viable alternative to fisheries<br />
today?<br />
ABSTRACT: Today, most world <strong>sea</strong> <strong>urchin</strong>s fisheries have to deal<br />
with overexploitation or yields drop problems. Better management <strong>of</strong><br />
exploited field populations and/or aquaculture is more and more<br />
considered as necessities to sustain <strong>sea</strong> <strong>urchin</strong>s’ production in <strong>the</strong> near<br />
future. In this context, we evaluate here <strong>the</strong> potentials <strong>of</strong> land-based closed<br />
cycle echiniculture.<br />
A long-term experiment with <strong>the</strong> edible violet <strong>sea</strong> <strong>urchin</strong><br />
(<strong>Paracentrotus</strong> lividus) has been done at a pilot scale in France. The<br />
process used allows total independence against natural resources, since <strong>the</strong><br />
whole biological cycle <strong>of</strong> <strong>the</strong> echinoids is under control (closed cycle<br />
echiniculture) and all activities are performed on land. Also, a method has<br />
been set up to gain control over <strong>the</strong> reproductive cycle <strong>of</strong> <strong>the</strong>se animals<br />
and to produce marketable individuals all year long.<br />
266
Annexes<br />
Overall conclusions <strong>of</strong> this experiment reveal great potentials, but also<br />
point out some pitfalls that remain to be eliminated before pretending for<br />
pr<strong>of</strong>itability. The most critical pitfalls identified are (1) poor control <strong>of</strong><br />
extremely variable growth rates due to intraspecific competition, (2) poor<br />
control on inorganic carbon in closed or semi-closed systems due to a high<br />
demand in carbonates for skeletogenesis and (3) needs for increased<br />
quality <strong>of</strong> gonads (<strong>the</strong> edible part <strong>of</strong> <strong>the</strong> <strong>urchin</strong>s) thanks to a specific<br />
artificial diet that remains to be formulated.<br />
One important aspect comes to light: land-based closed cycle<br />
echiniculture should have a very low impact on o<strong>the</strong>r mariculture or<br />
touristic activities that usually compete strongly for space on <strong>the</strong> coastline<br />
in many places. This should be a major advantage considering tomorrow’s<br />
aquaculture diversification.<br />
KEYWORDS: Sea <strong>urchin</strong>, <strong>Paracentrotus</strong> lividus, aquaculture, larval<br />
culture, metamorphosis, growth, roe enhancement.<br />
3 th International Symposium on Nutritional Strategies and<br />
Management <strong>of</strong> Aquaculture Waste, Porto, 1997. Talk. Ph. Grosjean,<br />
Ch. Spirlet, J. M. Lawrence & M. Jangoux. Optimizing somatic growth<br />
<strong>of</strong> <strong>the</strong> edible <strong>sea</strong> <strong>urchin</strong> (<strong>Paracentrotus</strong> lividus Lmk) (Echinodermata:<br />
Echinoidea) in closed-circuit cultivation with artificial diet.<br />
ABSTRACT: The closed-circuit cultivation <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s <strong>of</strong>fers <strong>the</strong><br />
opportunity to optimize <strong>the</strong>ir growth by controlling <strong>the</strong> rearing parameters,<br />
but <strong>the</strong> question whe<strong>the</strong>r water pollution would result from <strong>the</strong> waste<br />
produced is critical. Little is known about <strong>the</strong> requirements <strong>of</strong> cultivated<br />
<strong>sea</strong> <strong>urchin</strong>s in terms <strong>of</strong> food composition. The effect <strong>of</strong> two prepared feeds<br />
(soybeans and soybeans-fish pellets) versus fresh and dried kelp (<strong>the</strong><br />
natural food <strong>of</strong> <strong>Paracentrotus</strong> lividus) on feeding, digestion and somatic<br />
growth has been investigated under semi-intensive cultivation. The total<br />
267
Annexes<br />
amount <strong>of</strong> waste produced at different levels <strong>of</strong> feeding, digestion, and<br />
assimilation has been measured. Both prepared feeds are used as<br />
efficiently as fresh kelp for somatic growth, but wastes are reduced by<br />
20% due to a much higher conversion efficiency. These preliminary results<br />
suggest that a drastic improvement <strong>of</strong> feeding strategies in <strong>the</strong> culture <strong>of</strong><br />
<strong>sea</strong> <strong>urchin</strong>s would be obtained soon by appropriate formulations <strong>of</strong><br />
prepared feeds. On-land aquaculture <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s with prepared feeds<br />
provides a means <strong>of</strong> controlling <strong>the</strong> wastes produced. It would prevent<br />
environmental pollution and would allow recovery <strong>of</strong> <strong>the</strong> large amounts <strong>of</strong><br />
material still rich in organic material that may be used for o<strong>the</strong>r purposes.<br />
KEYWORDS: Artificial food, digestion, somatic growth, <strong>sea</strong> <strong>urchin</strong>,<br />
aquaculture.<br />
9 th International Echinoderm Conference, San Francisco, 1996. Talk<br />
in a special session. Ph. Grosjean, Ch. Spirlet & M. Jangoux. Closedcircuit<br />
cultivation <strong>of</strong> <strong>the</strong> edible <strong>sea</strong> <strong>urchin</strong> <strong>Paracentrotus</strong> lividus:<br />
optimization <strong>of</strong> somatic growth through <strong>the</strong> control <strong>of</strong> abiotic<br />
environment.<br />
ABSTRACT: Among <strong>the</strong> technological choices to develop<br />
aquaculture <strong>of</strong> new species, open-<strong>sea</strong> versus "on-land" cultivation is a<br />
major one. On-land based systems are more expensive but <strong>of</strong>fer <strong>the</strong><br />
possibility to control <strong>the</strong> environmental conditions, possibly leading to<br />
better performance. In <strong>the</strong> case <strong>of</strong> echinoderms, ecophysiological<br />
responses are insufficiently understood to decide at <strong>the</strong> present time which<br />
is <strong>the</strong> best strategy. To investigate this crucial question, one should (1) set<br />
up a good cultivation system, (2) develop an experimental methodology<br />
adapted to <strong>the</strong> specificity <strong>of</strong> echinoderm biology, and (3) quantify <strong>the</strong><br />
responses <strong>of</strong> <strong>the</strong> animals against gradients <strong>of</strong> environmental parameters.<br />
As an illustration <strong>of</strong> <strong>the</strong> promising perspectives this approach <strong>of</strong>fers, <strong>the</strong><br />
case <strong>of</strong> <strong>Paracentrotus</strong> lividus is discussed.<br />
268
Annexes<br />
A pilot system has been set up to control all life stages <strong>of</strong> <strong>reared</strong> <strong>sea</strong><br />
<strong>urchin</strong>s, from fertilization to gonad filling; <strong>the</strong> used protocol is being<br />
carefully standardized.<br />
A rapid (3 weeks) but accurate method has been developed to measure<br />
feeding, digestion, and somatic growth <strong>of</strong> <strong>reared</strong> <strong>sea</strong> <strong>urchin</strong>s under various<br />
environmental conditions. Both technical and statistical improvements<br />
have been made to increase <strong>the</strong> significance <strong>of</strong> <strong>the</strong> results.<br />
Two <strong>of</strong> <strong>the</strong> most important abiotic factors (photoperiod and<br />
temperature) have been investigated and let us to <strong>model</strong> <strong>the</strong>ir effects on <strong>sea</strong><br />
<strong>urchin</strong>s. Photoperiod has an impact on <strong>the</strong> feeding rate, but not on<br />
absorption or somatic growth. Optimal temperature for juveniles appears<br />
to be higher than for adults (respectively 23-24°C and 19°C). Moreover,<br />
juveniles are more sensitive to departure from this optimum. Hence, a<br />
strict control <strong>of</strong> temperature is a more critical issue for juveniles than for<br />
adults.<br />
Integration <strong>of</strong> our results in a wider <strong>model</strong> concerning <strong>the</strong> entire<br />
rearing structure shows that <strong>the</strong> apparent food conversion efficiency results<br />
in several complex phenomena: feeding and digestion <strong>of</strong> course, but also<br />
degradation <strong>of</strong> food that in addition depends on temperature. In cultivation,<br />
<strong>the</strong> highest productivity is obtained by making a compromise between a<br />
high somatic growth at optimal temperature and a high apparent food<br />
conversion efficiency at a lower temperature.<br />
KEYWORDS: Echinoid, aquaculture, food conversion, temperature,<br />
photoperiod.<br />
4 th European Echinoderm Colloquium, London, 1995. Poster. Ph.<br />
Grosjean, Ch. Spirlet & M. Jangoux. Establishment and presumed<br />
causes <strong>of</strong> <strong>the</strong> multimodal distribution commonly observed in cultivated<br />
populations <strong>of</strong> juvenile <strong>Paracentrotus</strong> lividus.<br />
269
Annexes<br />
ABSTRACT: Multimodal distribution (i.e., few individuals growing<br />
very fast and few individuals growing very slowly among an originally<br />
homogeneous group <strong>of</strong> <strong>sea</strong> <strong>urchin</strong>s P. lividus <strong>of</strong> <strong>the</strong> same strain) is <strong>of</strong>ten<br />
observed in controlled cultivation. The splitting <strong>of</strong> this group into<br />
homogeneous size-classed subgroups induces an increased growth <strong>of</strong> <strong>the</strong><br />
smaller individuals that achieve <strong>the</strong> same size than <strong>the</strong> o<strong>the</strong>rs in 3 months.<br />
This indicates that <strong>the</strong> smaller animals are not genetically less productive<br />
and suggests <strong>the</strong>y are inhibited in <strong>the</strong>ir growth due to some environmental<br />
constraints.<br />
KEYWORDS:<br />
competition, growth.<br />
Echinoid, population dynamics, intraspecific<br />
3 rd European Aquaculture Symposium, Bordeaux, 1994. Poster. Ph.<br />
Grosjean, Ch. Spirlet & M. Jangoux. First approach <strong>of</strong> <strong>the</strong><br />
performances <strong>of</strong> a closed-circuit <strong>sea</strong> <strong>urchin</strong> rearing structure.<br />
ABSTRACT: Within <strong>the</strong> context <strong>of</strong> an ECC financed re<strong>sea</strong>rch<br />
program on echinoid cultivation which objective is to succeed in<br />
continuously producing large amounts <strong>of</strong> edible <strong>sea</strong> <strong>urchin</strong>s<br />
(<strong>Paracentrotus</strong> lividus) under controlled conditions (aquaculture), <strong>the</strong><br />
performances <strong>of</strong> a closed-circuit rearing structure was tested. The rearing<br />
structure consists <strong>of</strong> toboggans measuring 4 m long and 60 cm wide on 3<br />
levels, <strong>the</strong> whole overhanging a reserve/settling tank <strong>of</strong> <strong>the</strong> same length, 80<br />
cm wide and 80 cm high. The <strong>sea</strong> <strong>urchin</strong>s are placed on <strong>the</strong> toboggans in 5<br />
to 10 cm running <strong>sea</strong>water. A 2.5 months follow-up <strong>of</strong> 2 structures do not<br />
show any significant increase <strong>of</strong> <strong>the</strong> biomass, mortality being barely<br />
compensated. An assessment <strong>of</strong> <strong>the</strong> quality <strong>of</strong> <strong>the</strong> water, done<br />
simultaneously reveals that CO2 is continuously oversaturated from 300 to<br />
400 %. This factor could be likely <strong>the</strong> cause <strong>of</strong> <strong>the</strong> poor growth <strong>of</strong> <strong>the</strong><br />
echinoids.<br />
270
Annexes<br />
KEYWORDS: Sea <strong>urchin</strong>, aquaculture, closed-circuit system, growth,<br />
CO2.<br />
8 th International Echinoderm Conference, Dijon, 1993. Poster. Ph.<br />
Grosjean & M. Jangoux. Effect <strong>of</strong> light on feeding in cultivated<br />
echinoids (<strong>Paracentrotus</strong> lividus).<br />
ABSTRACT: Within <strong>the</strong> context <strong>of</strong> a re<strong>sea</strong>rch on <strong>sea</strong> <strong>urchin</strong><br />
cultivation (<strong>Paracentrotus</strong> lividus), <strong>the</strong> effect <strong>of</strong> light on <strong>the</strong> amount <strong>of</strong><br />
food ingested and <strong>of</strong> faeces produced per echinoid per day has been<br />
investigated. Three sets <strong>of</strong> 20 adults were subjected to particular light<br />
conditions (constant light, constant darkness or 12 hours <strong>of</strong> light per day)<br />
for 6 days. Measurements were done for each <strong>of</strong> <strong>the</strong> 60 investigated<br />
echinoids. Calculation <strong>of</strong> <strong>the</strong> mean daily feeding and absorption rates for<br />
each set <strong>of</strong> individuals indicates that <strong>the</strong> highest values were obtained for<br />
echinoids at constant darkness and <strong>the</strong> lowest for echinoids subjected to<br />
light/darkness alternation (values for <strong>the</strong> three sets <strong>of</strong> individuals are<br />
significantly different).<br />
KEYWORDS: Echinoid, feeding rate, absorption rate, light,<br />
photoperiod.<br />
271