Impacts des changements climatiques sur les flux de carbone ...

Thèse de l’Université Pierre et Marie Curie – Paris VI
Laboratoire d’Océanographie de Villefranche (UMR 7093)
Spécialité
Océanologie – Météorologie – Environnement
Présenté par
Simon Bélanger
Pour l’obtention du grade de Docteur de l’Université Pierre et Marie Curie
Sujet de thèse:
Impacts des changements climatiques sur les flux de carbone
stimulés par la lumière dans l'Océan Arctique:
Quantification et suivi de la photo-oxydation de la matière organique dissoute dans
la Mer de Beaufort par télédétection spatiale
Response of light-related carbon fluxes in the Arctic Ocean to
climate change:
Quantification and monitoring of dissolved organic matter photo-oxidation in the
Beaufort Sea using satellite remote sensing
Devant le jury compose de :
Rapporteurs :
Dr. Rainer M. W. Amon
Pr. Dave A. Siegel
Soutenue le 28 novembre 2006
Éxaminateurs :
Dr. Ingrid Obernosterer
Pr. Alain Saliot (Président)
Pr. Warwick F. Vincent (co-directeur de thèse)
Dr. Marcel Babin (directeur de thèse)

Remerciements
On ne vient pas à bout d’une thèse de doctorat sans le soutient financier de différent
organismes subventionnaires, de celui professionnel de nos pairs et personnel de nos
proches. Donc pour commencer je tiens à exprimer ma gratitude envers les décideurs du
programme de bourse de thèse du Fond Québécois pour la Recherche sur la Nature et les
Technologies (FQRNT). En plus de m’avoir attribué un salaire raisonnable durant mes trois
années de thèse, ils m’ont surtout donné carte blanche pour réaliser mon projet de recherche
dans un pays étranger. Projet qui fût aussi soutenu à travers le programme de recherche
Canadian Arctic Shelf Exchange Study (CASES), financé par le Conseil Canadien des
Sciences Naturelles et Génie (CRSNG) et dirigé par le Dr Louis Fortier. Cette plate-forme
m’a permis d’obtenir les données nécessaires à mes études et d’acquérir une expérience
« Arctique » si enrichissante. Un autre acteur important dans la réalisation de la campagne
CASES, que je remercie pour la place qu’il m’a réservée sur le navire, est Mr Pierre Larouche
de l’Institut Maurice-Lamontagne à Mont-Joli. Aussi à propos de la campagne en mer à bord
du brise-glace canadien CCGS Amundsen, j’en profite aussi pour remercier ses capitaines et
ses membres d’équipages qui ont su mener à bon port nos aventures entre mer, terre et glace,
ainsi que mes collègues scientifiques et amis dont l’enthousiasme m’a permi de passer 12
semaines inoubliables.
Sur le plan scientifique, plusieurs personnes ont consacré nombre d’heures de leur
temps pour m’aider dans l’interprétation et la publication de mes résultats. Je remercie tous
ces personnes, et plus particulièrement ceux qui, malgré leur éloignement, ont contribué de
façon très significative à mes travaux : Dr Warwick Vincent, Dr Huixiang Xie, Dr Nick
Krotkov et Jens Ehn. Je dois ajouter au passage que la présence de Warwick a été une source
de motivation importante lors de mes passages à Québec. Merci Warwick d’avoir accepté de
me co-diriger. Tu seras toujours un modèle de motivation, et je compte bien m’en inspirer
durant mon nouveau travail de professeur à Rimouski.
Plus près de moi, j’ai aussi eu une chance immense de bénéficier de l’expertise d’un
certain nombre de mes collègues du Laboratoire d’Océanographie de Villefranche-sur-mer.
Plus qu’une chance, ce fût un privilège de côtoyer l’un des plus grand chercheur de
l’océanographie moderne, le Prof André Morel. Je n’oublierai jamais les nombreux matins
III

passés à discuter de tout et de rien dans sa « caverne d’Ali Baba », laquelle est pratiquement
toujours éclairée par le chaud soleil de Villefranche...
Marcel, quand j’ai pris contact avec toi il y a quatre ans, je ne doutais pas que tes
qualités professionnelles m’aideraient à mener à bien mon projet de recherche. Mais j’étais
loin d’imaginer à quel point te côtoyer allait m’apporter au niveau personnel. Merci pour le
temps que tu m’as consacré, pour ta patience et ta disponibilité, pour ton sens de la
perfection et ta rigueur scientifique, pour m’avoir appris à mieux gérer mon stress (et mes
gestes durant mes présentations!), pour avoir su comment me redonner confiance quand il le
fallait, etc., etc.. Je pourrais continuer ainsi pour des pages! Je te considère avant tout comme
un ami, et ensuite comme un collègue. Ma porte sera toujours ouverte pour toi à Rimouski…
Il y a plusieurs personnes avec lesquelles je me suis lié d’amitié entre le quai de la
darse, la Colle St-Michel, le Mont Boron, le cap Ferrat, Nice, la Corse, le Hockey, Monaco,
Roma, Barcelone, Lavello, Paris, etc. Je les quitte avec un gros pincement au cœur. Ces
personnes sauront se reconnaître en lisant ces lignes, nul besoin d’en ajouter.
Il y a une personne en particulier à qui je dois beaucoup plus que tout au monde,
c’est ma blonde adorée. Sans toi, je ne m’en serais jamais sorti!! Merci d’avoir toujours été là,
prête à me couper un peu de jambon, ou à me cuisiner une p’tite crêpe bretonne.
Pour terminer, je tiens à m’excuser auprès de mes proches du Québec qui m’ont très
peu vu durant ces années passées sur la côte d’azur. Je les retrouve avec empressement et
près à rattraper le temps passé outre Atlantique.
IV

V
À la mémoire de père

Résumé de la thèse
L’oxydation photochimique de la matière organique dissoute colorée (CDOM), et la
production de CO 2 qui en résulte, est maintenant reconnue comme étant un processus
majeur dans le cycle du carbone dans le système ocean-atmosphère. L’Océan Arctique fait
parti des environnements où ce processus risque de jouer un rôle de plus en plus important
dans le contexte des changements climatiques à cause de : 1) l’augmentation du CDOM
d’origine terrigène transporté à l’Océan côtier en réponse à la fonte du pergélisol et à
l’augmentation du débit des rivières; 2) la diminution du couvert de glace estivale permettant
au rayonnement solaire de pénétrer dans la colonne d’eau; et 3) l’augmentation du
rayonnement ultraviolet (UV) dans cette région.
Un modèle couplé « optique-photochimique » a été utilisé pour examiner le rôle de la
photooxydation dans le cycle du carbone organique de l’Océan Arctique. Pour calculer la
photoproduction de CO 2 (P DIC), l’éclairement spectral incident, incluant les UV, a été
modélisé avec un modèle de transfert radiatif nécessitant les concentrations de glace de mer,
d’ozone, d’aérosols atmosphériques et le couvert nuageux observés par satellite entre 1979 et
2004. Les rendements quantiques apparents pour calculer la P DIC ont été déterminés en
laboratoire sur des échantillons provenant de la Mer de Beaufort et du Fleuve Mackenzie. La
contribution du CDOM au coefficient d’absorption totale ([a CDOM/a t]), un paramètre clé du
modèle, a été déterminée soit à partir des mesures in situ, soit dérivée des données de la
couleur de l’océan à l’aide d’un nouvel algorithme empirique.
Contrairement aux méthodes semi-analytiques trouvées dans la littérature,
l’algorithme empirique proposé permet de séparer, à l’échelle régionale, les coefficients
d’absorption du CDOM et des particules non-algales. Cependant, l’utilisation des données de
la couleur de l’océan dans les hautes latitudes est souvent compromise par la présence de
glace de mer qui contamine les données. Ce problème a été abordé dans le cadre de la
présente étude, et une méthode a été proposée afin de détecter et éliminer les données
contaminées par la glace de mer.
Enfin, il a été démontré que les valeurs de P DIC sont similaires aux taux de carbone
organique provenant de la production primaire marine qui est séquestré dans les sédiments
océaniques; et que l’augmentation des UV et la diminution de la glace de mer estivale au
cours des 26 dernières années a conduit à une augmentation de P DIC d’environ 15%. Ces
résultats indiquent que la photooxydation joue un rôle significatif dans le cycle du carbone
actuel, et que la diminution de l'étendue de la glace de mer, prévue par les modèles
climatiques, risque d'amplifier significativement la reminéralisation photochimique du
carbone organique dissous d'origine terrigène dans l'Océan Arctique.
VI

Thesis abstract
Photochemical oxidation of colored dissolved organic matter (CDOM), and the
resulting production of CO2, is now known to be a significant process in the cycling of
carbon in the ocean-atmosphere system. One environment where that process may play a
major role in the context of climate change is the Arctic ocean because of: 1) the
increasing amount of terrestrial CDOM released by the melting permafrost and brought to
coastal ocean by rivers, 2) the decreasing summer ice cover that allows more solar
radiation to penetrate the water column, and 3) the continuing increase in UV radiation
over that region.
A coupled optical-photochemical model was used to assess the role of photooxidation
in the carbon cycle of the Arctic Ocean. To calculate the photoproduction of
CO2 (PDIC), the incoming spectral irradiance, including UV, was modeled with a radiative
transfer model that uses satellite observations of sea ice, ozone, aerosols and cloud cover
covering the 1979 to 2004 period. In situ determinations of the apparent quantum yield
for the photoproduction of CO2 made in the Beaufort Sea were used for the calculations.
A key parameter in the model was the contribution of CDOM to the total absorption
coefficient. It was either obtained from in situ measurements or derived from Ocean
Color imagery using a new empirical algorithm.
Unlike most semi-analytical approaches found in the literature, the proposed
empirical algorithm provides a mean to separate CDOM absorption coefficient from nonalgal
particles absorption coefficient at the regional scale. The use of Ocean Color remote
sensing at high latitude is, however, compromised by the presence of sea ice that
contaminates the data. This problem was addressed in the present study, and a method
was proposed to detect and eliminate contaminated pixels.
Finally, it was shown that the level of PDIC is similar to the level of sequestered
rates of organic carbon in the ocean sediments, which was produced through marine
photosynthesis; and that the increase in UV and decrease in summer sea ice over the last
26 years have led to an increased in PDIC by about 15%. These results indicate that the
predicted trend of ongoing contraction of sea ice cover will greatly accelerate the
photomineralization of CDOM in Arctic surface waters.
VII

Table des matières
Remerciements ....................................................................................................... III
Résumé de la thèse ..................................................................................................VI
Thesis abstract....................................................................................................... VII
Table des matières................................................................................................VIII
Liste des figures..................................................................................................... XII
Liste des tableaux .............................................................................................. XVIII
Avant propos .........................................................................................................XIX
Chapitre I : Introduction et problématique générale .................................1
I.1. Introduction ......................................................................................................... 2
I.2. Problématique ..................................................................................................... 5
I.2.1. Importance de la matière organique dissoute d’origine terrigène dans l’Océan
Arctique et impact du réchauffement des écosystèmes terrestres nordiques .................... 5
I.2.2. Importance de la photooxydation et impact de la fonte de la glace de mer et de
l’augmentation du rayonnement ultraviolet dans l’Arctique ................................................ 8
I.2.3. Quantification des processus photochimiques........................................................... 12
I.2.4. La télédétection de la couleur de l’océan : un outil de choix pour quantifier les
processus photochimiques aux échelles océaniques............................................................ 15
I.2.4.1. Propriétés optiques des eaux arctiques et algorithmes pour l’estimation du coefficient
d’absorption du CDOM........................................................................................................... 17
I.2.4.2. La glace de mer : un problème pour la télédétection de la couleur de l’océan..................... 22
I.3. Objectifs et Organisation de la Thèse.............................................................. 25
I.4. Thesis objectives and organization................................................................... 27
Chapitre II : L’Océan Arctique, la Mer de Beaufort et le Plateau du
Mackenzie................................................................................................. 29
II.1. Introduction...................................................................................................... 30
II.2. Généralités sur l’Océan Arctique..................................................................... 32
II.3. Généralités sur la Mer de Beaufort et le Plateau du Mackenzie ..................... 35
II.3.1. Cycles Saisonniers sur le Plateau du Mackenzie ....................................................... 37
II.3.2. Bilan de Carbone Organique sur le Plateau du Mackenzie..................................... 41
Chapitre III: Photominéralisation de la matière organique dissoute
d’origine terrigène dans les eaux côtières arctiques entre 1979 et 2003:
Variabilité interannuelle et implications des changements climatiques 43
III.A Résumé ........................................................................................................... 44
III.B Article publié dans la revue Global Biogeochemical Cycles :
“Photomineralization of terrigenous dissolved organic matter in Arctic coastal
waters from 1979 to 2003: Interannual variability and implications of climate
change” .................................................................................................................... 45
Acknowledgments ................................................................................................... 46
VIII

Abstract .................................................................................................................... 47
III.1. Introduction .................................................................................................... 48
III.2. Materials and methods ................................................................................... 51
III.2.1. Sample collection and storage ................................................................................... 51
III.2.2. Optical measurements ................................................................................................ 52
III.2.3. Irradiation experiments .............................................................................................. 53
III.2.4. DIC measurement and calculation of φ DIC .............................................................. 54
III.2.5. Modeling DIC photoproduction............................................................................... 55
III.3. Results and discussion ................................................................................... 59
III.3.1. Apparent Quantum Yield for DIC photoproduction............................................ 59
III.3.2. Role of DIC photoproduction in the organic carbon cycling .............................. 62
III.3.3. Interannual variability in DIC photoproduction .................................................... 65
III.3.4. Implications for tDOC cycling in Arctic coastal waters........................................ 69
III.4. Summary and conclusions ............................................................................. 72
III.5. Auxiliary material: On the determination of the Apparent Quantum Yield for
DIC photoprodcution .............................................................................................. 73
III.6. References ...................................................................................................... 77
Chapitre IV : Quantification améliorée de la photooxydation de la
matière organique dissoute colorée dans les eaux côtières à l’aide des
propriétés optiques inhérentes dérivées des données de couleur de
l’océan ....................................................................................................... 83
IV.A Résumé............................................................................................................ 84
IV.B. Article soumis à la revue Journal of Geophysical Research - Oceans (28
décembre 2006): “Improved quantification of Chromophoric Dissolved Organic
Matter photooxidation in coastal waters using satellite-derived inherent optical
properties” ............................................................................................................... 85
Abstract .................................................................................................................... 86
IV.1. Introduction .................................................................................................... 87
IV.2. Materials and Methods ................................................................................... 89
IV.2.1. Data sets description and measurements................................................................. 89
IV.2.2. Description of the [aCDOM/at] algorithm............................................................. 96
IV.3. Results and Discussion................................................................................... 96
IV.3.1. Variability of [a CDOM/a t] and [a CDOM/a CDM] in coastal waters.................................. 96
IV.3.2. Retrieval of [a CDOM/a t] from remote sensing reflectance ....................................... 98
IV.3.4. Application to SeaWiFS imagery............................................................................. 113
IV.4. Summary and conclusions.............................................................................116
IV.5. Notations........................................................................................................120
IV.6. References......................................................................................................121
Chapitre V: Impact de la glace de mer sur les estimations de la
réflectance marine, la concentration de chlorophylle a et les propriétés
optiques inhérentes dérivés des données satellitales de la Couleur de
l’Océan .....................................................................................................125
IX

V.A. Résumé ...........................................................................................................126
V.B. Article soumis à la revue Remote Sensing of Environment (1 novembre 2006):
“Impact of sea ice on the retrieval of water-leaving reflectance, chlorophyll a
concentration and inherent optical properties from satellite Ocean Color data”..128
Abstract ...................................................................................................................129
V.1. Introduction .....................................................................................................130
V.2. Methodology....................................................................................................132
V.2.1. Definitions ................................................................................................................... 132
V.2.2. Description of the surface reflectance spectra........................................................ 133
V.2.3. Simulations of the adjacency effect .......................................................................... 136
V.2.4. Simulations of sea ice contamination within an ocean pixel................................. 136
V.2.5. Atmospheric correction and bio-optical processing.............................................. 137
V.3. Results and discussion ....................................................................................138
V.3.1. Impact of the adjacency effect on the retrieval of [ρ w] N, CHL and IOPs .......... 138
V.3.2. Impact of sub-pixel sea ice contamination on the retrieval of [ρ w] N, CHL and
IOPs.......................................................................................................................................... 147
V.3.3. Detection of pixels affected by sea ice (adjacency effect and sub-pixel
contamination) ........................................................................................................................ 153
V.4. Summary and conclusions...............................................................................157
Acknowledgment ....................................................................................................158
Appendix V.1. Water reflectance model .................................................................158
Appendix V.2. Post-treatment of 6s output ............................................................161
Appendix V.3. Modification of the QAA ................................................................163
V.5. References .......................................................................................................165
Chapitre VI : Utilisation de l’imagerie « couleur de l’océan » pour la
quantification de la photooxydation dans la Mer de Beaufort...............169
VI.A Résumé...........................................................................................................170
VI.B. Article en preparation: “The use of Ocean Color imagery for the
quantification of photooxidation in the Beaufort Sea” ..........................................171
Abstract ...................................................................................................................172
VI.1. Introduction ...................................................................................................173
VI.2. Methods .........................................................................................................174
VI.2.1. SeaWiFS Level 2 and 3 processing.......................................................................... 174
VI.2.2. DIC photoproduction calculation........................................................................... 177
VI.3. Results............................................................................................................179
VI.3.1. Validation of the SeaWiFS water-leaving reflectance........................................... 179
VI.3.2. Spatial-temporal variability in [a CDOM/a t](412)....................................................... 183
VI.3.3. Spatial-temporal variability in DIC photoproduction.......................................... 186
VI.4. Discussion......................................................................................................189
VI.4.1. SeaWiFS data quality................................................................................................. 189
VI.4.2. Quantification of DIC photoproduction using satellite Ocean Color .............. 190
VI.5. Conclusions....................................................................................................193
VI.6. References......................................................................................................193
Chapitre VII: Conclusion générale et perspectives ................................196
X

VII.A. Version anglaise: Overall Conclusion and Perspectives ............................ 204
Bibliographie générale ............................................................................ 211
Annexe 1 : Modification de l’algorithme Quasi-Analytique pour
l’application à la Mer de Beaufort...........................................................231
Annex A1. Modification of the Quasi-Analytical Algorithm for an application to
the Beaufort Sea..................................................................................................... 232
A1.1. Introduction and motivation......................................................................... 232
A1.2. Quasi-Analytical Algorithm description....................................................... 232
A1.3. Results and discussion.................................................................................. 237
A1.3.1. Application of QAA to in situ data set.................................................................. 237
A1.3.2. Estimation of a CDOM.................................................................................................. 240
A1.4. References..................................................................................................... 242
XI

Liste des figures
Figure I.1. Anomalie débit d’eau douce (trait noir) des six plus grands fleuves sibériens qui
se déchargent dans l’Océan Arctique depuis 1940. L’augmentation (7%) est corrélée à
l’augmentation de la température de surface à l’échelle globale (trait bleu), ainsi qu’avec
l’indice climatique NAO (trait rouge) (source : Peterson et al., 2002)................................. 7
Figure I.2. Observations de l’étendue moyenne du couvert de glace saisonnier au cours du
20 ième siècle (source des données : ACIA, 2005). La diminution de la glace est plus
importante au printemps et en été, qu’en automne et en hiver. .......................................... 9
Figure I.3. Étendue de la banquise Arctique observée par satellite annuellement au mois de
septembre entre 1979 et 2005. (Source des données: National Snow and Ice Data Center,
2006)........................................................................................................................................... 10
Figure I.4. Éclairement spectral dans les parties UV-A et UV-B pour trois concentrations
en ozone telles qu’indiquées sur la figure (DU = Dobson Units), et un exemple de
rendement quantique apparent (voir la section suivante pour plus d’explications)
exprimé en unités relatives, représentant la réactivité photochimique du CDOM en
fonction de la longueur d’onde (courbe grisée).................................................................... 11
Figure I.5. Trajectoires possibles dans le système atmosphère-océan pour un photon émis
par le soleil par temps dégagé (voir aussi encadré 4). De la quantité de photon diffusée
dans l’océan qui parvient à sortir de l’océan en direction du satellite (a; L w), une partie
sera transmise à travers l’atmosphère jusqu’au capteur (b) alors que l’autre partie sera
perdue par diffusion ou absorption dans l’atmosphère (c). Certains photons en
provenance du soleil sont transmis directement (d) ou indirectement (e) dans le champ
de visée du capteur après avoir été réfléchis de façon spéculaire par l’interface air-mer
(L r, en anglais sun glint ou glitter). Comme pour L w, une partie sera transmise (g) et l’autre
sera perdue dans l’atmosphère (f). Dans la partie visible du spectre, plus de 90% de
l’énergie reçue au satellite provient de l’atmosphère (L p) et comprend : les photons
diffusés une (h) ou plusieurs (i) fois par les molécules ou les aérosols atmosphériques.
Dans les régions polaires où la glace de mer est présente, il faut ajouter à ce
schéma la contribution des photons réfléchis par la glace à l’extérieur du champ de
visée qui sont redirigés vers le capteur (l; effet de l’environnement). Enfin, quand la
glace se trouvant dans le champ de visée du capteur, une partie des photons qu’elle
réfléchira sera transmise (m; contamination sub-pixel) et l’autre sera perdue dans
l’atmosphère (n). Les contributions indirectes (l) et directes (m) posent problème
puisqu’ils ne sont pas pris en compte dans le traitement des données de couleur de
l’océan......................................................................................................................................... 23
Figure II.1. Patrons de circulation générale des eaux de surface dans l’Océan Arctique. Les
eaux Atlantique et Pacifique (en rouge) pénètrent dans l’Arctique via la Mer de Barents
et le détroit de Bering respectivement. Le Tourbillon de Beaufort et le courant
Transpolaire (en bleu) sont les deux principaux chemins empruntés par les eaux
Arctiques, lesquelles sont éventuellement exportées vers l’Atlantique Nord via de
détroit de Fram et l’Archipel Canadien. Ces eaux de surface sont fortement influencées
par les grands fleuves Sibériens et Nord Américains. Les huit plus importants sont
indiqués avec, entre parenthèses, leur débit annuel (volume d’eau douce en km 3 ans -1 ).
Le cadre noir localise la zone d’étude. (Source des informations : ACIA, 2005)............ 31
XII

Figure II.2. Carte de localisation de la Mer de Beaufort avec la direction des principaux
courants : la Gyre de Beaufort (en gris); le sous-courant de Beaufort (c.f., Beaufort
Undercurrent; en bleu); la dispersion attendue du panache du Mackenzie (en rouge) qui
varie selon la direction et l’intensité des vents dominants. L’encadré en gris montre les
limites de l’échantillonnage effectué durant le programme CASES.................................. 36
Figure II.3. Cycle saisonnier de l’eau douce sur le Plateau du Mackenzie caractérisé par les
apports continentaux et la glace de mer : A) englacement (apports continentaux faibles,
la glace commence à se former); B) fin de l’hiver (apports continentaux faibles, la glace
a atteint son épaisseur maximale et a arrêté de croître); C) débâcle printanière (apports
continentaux élevés, la glace demeure intacte et reste près de la côte); D) été (apports
continentaux élevés, la glace a fondue ou a été exportée vers l’intérieur de l’Océan).
(Modifiée de Macdonald, 2000).............................................................................................. 37
Figure II.4. Variabilité journalière entre 1976 et 1998 (A) du débit du fleuve Mackenzie, (B)
du flux de particules en suspension d’origine terrigène, et (C) de l’éclairement
photosynthétique incident à 71°N et la température de l’air mesurés à Tuktoyaktuk situé
à l’embouchure du fleuve. (Modifiée de O’Brien et al., 2006)............................................ 38
Figure II.5. Image SeaWiFS acquise le 15 Juin 2004 montrant le début de la débâcle.
Rapidement les eaux turbides du Mackenzie occupent une large partie des eaux
ouvertes du plateau continental (~15 000-20 000 km²)...................................................... 40
Figure II.6. Coupe transversale effectuée à partir de l’embouchure du fleuve Mackenzie
(Kugmalit Bay) jusqu’à la limite du plateau continental en juillet 2004 dans le cadre du
programme CASES : a) densité; b) salinité........................................................................... 41
Figure III.1. Location of CASES stations sampled for absorption measurements in 2004
over the Mackenzie Shelf (open circles), in Amundsen Gulf (open triangles), and in
Canada Basin (open squares). ARDEX and CASES samples collected for φDIC determination are shown as closed squares. ......................................................................... 50
Figure III.2. φDIC spectra determined on water samples collected during ARDEX (R5a, R5d,
and R9; solid thin lines) and CASES (108, 406 and 409; dashed thin lines). The grey
curves are the average φDIC spectra published by Vähätalo et al. [2000] for a boreal lake,
and by Johannessen and Miller [2001] for inshore, coastal and offshore waters.................. 58
Figure III.3. (a) Variation of aCDOM(330) with salinity for water samples collected in
Mackenzie Shelf (open circles), Amundsen Gulf (open triangles), Canada Basin (open
squares). Samples collected for φDIC are shown as closed squares. (b) Variation of
weighted quantum yield normalized to the integrated irradiance, φ DIC , with salinity. .. 61
Figure III.4. Average DIC photoproduction rates, P DIC, calculated using the φ DIC
representing each sub-region: R5a, R5d and R9 for the Mackenzie Shelf; 108 for the
Amundsen Gulf; 406 and 409 for the Canada Basin........................................................... 62
Figure III.5. Average spectral light fraction absorbed by CDOM in the surface waters at
stations (Fig. III.1) sampled over the Mackenzie Shelf (dashed line), in Amundsen Gulf
(dash-dotted line), and in Canada Basin (thin line).............................................................. 63
Figure III.6. Trends in the annual DIC photoproduction for period 1979-2003 for the
Mackenzie Shelf, the Amundsen Gulf, the Canada Basin, and the sums of the three
sub-regions. ............................................................................................................................... 65
Figure III.7. Correlation between the annual DIC photoproduction and the annual average
area of open water for: (a) Mackenzie Shelf, DIC = 0.91x10 6 * X; (b) Amundsen Gulf,
XIII

DIC = 0.52x10 6 * X ; (c) Canada Basin, DIC = 0.63x10 6 * X. (X = open water area in
km 2 ; slope in Gg C y -1 km -2 ; n=25). ....................................................................................... 66
Figure III.8. Trends in open water area for the three sub-regions as observed by
SMMR/SSMI between 1979 and 2003. The left panels show the day of the year when
>50% of the surface area is ice free. The right panels show the number of days having
open water. ................................................................................................................................ 67
Figure III.9. Trends in monthly O3 concentration over the study area as observed by
TOMS between 1979 and 2001. Note the changes of scale among the panels. ............. 68
Figure III.A.1. Absorption spectra of CDOM for before (black line) and after the
irradiation for the sample under WG280 cutoff filter (red line). The time of irradiation
and the irradiance level under the cutoff filter are also shown.......................................... 75
Figure III.A.2. Modeled versus measured DIC photoproduction rates for the Single
exponential (squares) and Quasi-exponential (circles) functional forms.......................... 76
Figure IV.1. Map of the study area showing location of stations where IOPs (dots), SPMR
(triangles) and ASD (squares) measurements were made. Contour lines of sea ice
concentration of 50% are also shown for June 1 st (red), July 1 st (blue) and August 1 st
(green)......................................................................................................................................... 91
Figure IV.2. Frequency distributions of the contribution of CDOM to (a) the total light
absorption at 412 nm ([aCDOM/at]), and (b) to the colored detrital material ([aCDOM/aCDM]). Note that the graph includes 54 additional COASTlOOC or CASES stations where
only IOP were available........................................................................................................... 98
Figure IV.3. Comparison between retrieved and measured [aCDOM/at] at 412 nm for the
COASTlOOC and CASES datasets. [aCDOM/at](412) was calculated using equation 6
with coefficients obtained using the whole data set (N=255; Table IV.2). ..................... 99
Figure IV.4. Error analysis of [aCDOM/at] at 412, calculated using region-independent
coefficients, as a function of: a) the ratio of ap(412) to Rrs(555); b) the relative
contribution of CDOM to the total CDM absorption coefficient at 412 nm. Only the
statistically significant relationships are shown. ................................................................. 103
Figure IV.5. Same as Fig. IV.3 but for region specific coefficients for Eq. 6 (Table IV.2).104
Figure IV.6. Apparent Quantum Yield spectra for DIC production determined for two
stations located in the southeastern Beaufort Sea. AQY1 was determined on surface
waters influenced the by the Mackenzie River (salinity = 8.2 ‰; Station R5a), while
AQY2 was determined on surface waters collected in the Amundsen Gulf, away from
direct riverine influence (salinity = 30.0 ‰; Station 108) [Bélanger et al., 2006]. ............ 106
Figure IV.7. Spectral production of DIC for different values of [aCDOM/at](412) and for A)
AQY1 and B) AQY2. .............................................................................................................. 107
Figure IV.8. Relative difference in DIC production as a function of the absolute difference
between the retrieved and true [aCDOM/at](412) values for the two AQYs and two initial
values for [aCDOM/at] true (412) of 0.3 and 0.7. The thick line at the bottom illustrate the
95% confidence interval in the [aCDOM/at] retrieval using Eq 6........................................ 108
Figure IV.9. Variability of SCDOM as a function of the aCDOM(412) value measured in the
surface waters of the southeastern Beaufort Sea during summer 2004 (n=65). A
second-order polynomial fit between the SCDOM values and the logarithm of aCDOM(412) provided a strong description of the observations (r 2 =0.79; n=65): .............................. 110
2
S CDOM = 0. 0186 − 0.
0019log[
aCDOM
( 412)
] + 0.
0031log[
aCDOM
( 412)]
................................ 110
Figure IV.10. Spectral absorption by particulate matter normalized to 440 nm value
measured in the surface waters of the southeastern Beaufort Sea during summer 2004
XIV

N
(thin grey lines). The thick black line represent the average ap (λ) values. The
sensitivity the PDIC (eq 1) to the extrapolation to the UV domain is explore within the
limits given by 95% probability of the normal distribution (i.e. ±1.96*Standard
Deviation (σ); dashed lines). The ap(λ) values for λ

Figure V.10. Retrieved Angström exponent parameter (α) using the atmospheric correction
algorithm, as a function of σR ice, for τ a=0.1 and four types of sea ice............................149
Figure V.11. Ratio of CHL estimated from OC4v4 (solid line) and OC4L (dotted line) when
sea ice is present within the water pixel to CHL estimated without sea ice as a function
of σR ice(865). The results are presented for three initial concentrations of chlorophyll
(with f CDOM =50% and SPM NAP =0.0 g m -3 ), as indicated, for τ a=0.1 and four types of sea
ice. The dashed lines represent a 35% error range. ........................................................... 150
Figure V.12. Ratio of a t estimated at 443 nm using QAA when sea ice is present within the
water pixel to a t estimated with no sea ice as a function of σR ice(865). The results are
presented for each water type as shown in figure 2 and for the landfast ice covered by
fresh snow. The dashed lines represent a 35% error range.............................................. 151
Figure V.13. Same as Fig. V.12, but for b bp estimated at 555 nm using QAA....................... 151
Figure V.14. Same as Fig. V.12, but for the ratio [a CDOM/a t] estimated at 412 nm................ 152
Figure V.15. Example of sea ice contamination on the CHL estimation (OC4v4) for a
SeaWiFS scene of the Beaufort Sea acquired on the 8 th of September 2002. Upper panel
is the true color composite showing the sea ice field offshore........................................ 153
Figure V.16. Reflectance ratio between [ρ w] N at 412 and [ρ w] N at 443 nm as a function of
[ρ w] N at 555 nm for: 1) in situ measurements of the water-leaving reflectance conducted
in the Southeastern Beaufort Sea [Bélanger et al., submitted manuscript. 2006] (closed
circles); 2) modeled water-leaving reflectance with the bio-optical model presented in
Appendix V.1 (closed squares); and 3) SeaWiFS pixels extracted from the image
presented on Fig. V.17 (light grey dots).............................................................................. 155
Figure V.17. Example of application of the adjacency flag to a SeaWiFS scene from the
Southeastern Beaufort Sea acquired on the 16 th of June 1998. The R rs at 555 nm is
shown on colored log scale as indicated on the right; the flagged pixels for adjacency
are identified with dark red color. The sea ice and land are shown in dark and light grey
respectively. Because of known problem with the standard SeaWiFS atmospheric
correction over the turbid waters, we apply the algorithm proposed by Ruddick et al.
[2000]........................................................................................................................................ 156
Figure V.A1. Schematic representation of a water pixel located at distance D 0 from an ice
edge........................................................................................................................................... 162
Figure VI.1. A schematic representation of the DIC photoproduction modeling. The model
ingests four inputs: (1) For a given day, when data are available, the spatio-temporal
variability in the [a CDOM/a t] parameter is specified by interpolating the values of the
months before and after. If monthly IOP are not available (due to either sea ice or
cloud cover), the data from the climatology is taken. The spectral interpolation of the
[a CDOM/a t](412) is perform as described in section IV.3.4. (2) The daily spatial variability
in the area of open water. (3) The spectral daily downward irradiance calculated as
described in section III.2.5. (4) The spectral apparent quantum yield for the DIC
production (AQY, or φ DIC). Two methods are compared for the choice of the φ DIC for a
given day and pixel. Method 1 is based on the regional definition described in section
III.2.5. In the Method 2, the φ DIC varies as a function of the magnitude of the CDOM
(see text for details). ............................................................................................................... 178
Figure VI.2. A) CDOM photoreactivity as a function of a CDOM(412). B) Spectral variability
of the pooled φ DIC spectra for low and high a CDOM(412) value.......................................... 179
XVI

Figure VI.3. Comparison of in situ measurements of water-leaving radiance with coincident
spectra from the SeaWiFS image processed with the standard and the turbid water AC
algorithms. ............................................................................................................................... 181
Figure VI.4. Scatterplots of the in situ measurements versus SeaWiFS-derived variables
employed in the empirical algorithm used to estimate [a CDOM/a t](412) (eq. IV.6): a) the
logarithm of water-leaving reflectance ratio of 412 to 555 nm; b) the logarithm of
water-leaving reflectance ratio of 490 to 555 nm............................................................... 182
Figure VI.5. Seasonal variability of the SeaWiFS-derived [a CDOM/a t](412) for 1998 and 2004
(1 x 1 km resolution). Sea ice (or cloud) is shown in grey................................................ 184
Figure VI.6. Variability of the SeaWiFS-derived b bp(555) for June and July 1998 and 2004 (1
x 1 km resolution)................................................................................................................... 185
Figure VI.7. Variability of the SeaWiFS-derived a t(412) for June and July 1998 and 2004 (1 x
1 km resolution)...................................................................................................................... 185
Figure VI.8. Annual maps of DIC photoproduction rates from 1998 to 2004 in
southeastern Beaufort Sea..................................................................................................... 188
Figure A1. Schematic representation of the quasi-analytical algorithm developed to derive
the particles backscattering coefficient at 555 nm and the spectral total absorption
coefficient. The data (parallelogram) necessary for each steps of the algorithm
(rectangular boxes) are identified by the dotted arrows, while the outputs are identified
by the thin line with filled arrows. The final outputs, i.e. b bp(555) and a t(λ), are identified
by the think line with filled arrows at bottom right of the diagram. The equations
(identified with upper script number) are listed in Table A1........................................... 236
Figure A2. Comparison between retrieved and measured a t with the ac9 at A) 412 nm and B)
440 nm. Total absorption coefficients were retrieved using the optimized version of
QAA proposed by Lee et al. [2002] that use λ 0 555 and 640 nm (open circles) and the
tuned QAA with the CASES dataset (inverted triangles)................................................. 238
Figure A3. Comparison between QAA-derived b bp(555) and measured b p(555)................... 240
Figure A4. Retrieved versus measured a CDOM(412) in the Beaufort Sea. Statistical parameters
of the Type II regression are also shown............................................................................ 241
XVII

Liste des tableaux
Table III.1. Chemical and physical properties of samples used for the φDIC........................... 52
Table III.2. CDOM properties and model parameters for φDIC. .............................................. 58
Table III.3. Comparison between daily DIC photoproduction, and total and new primary
production rates in open water (in mg C m -2 d -1 )................................................................. 64
Table III.4. Mean annual DIC photoproduction and estimates of tDOC mineralized by
photooxidation.......................................................................................................................... 71
Table IV.1. Relative contribution of CDOM to the colored detrital matter
(CDM=CDOM+NAP) and to the total absorption coefficients at 412 nm, respectively,
for the different regions covered by our data set................................................................. 97
Table IV.2. Empirical coefficients of Eq. 6 determined by multiple regression for the
different coastal environments. .............................................................................................. 99
Table IV.3. Performance of Eq. 6 for the retrieval of [aCDOM/at] at 412 nm. For each region,
independent data set is used to obtain the empirical coefficients for Eq. 6. ................. 100
Table VI.4. Coefficient of determination of the linear regression between Δ[aCDOM/at] (shown in Fig. IV.4) and 1) the logarithm of the ratio [ap(412)/Rrs(555)] and 2) the ratio
[ a CDOM / aCDM
] at 412 nm. NS = the slope is not significantly different from 0 at 95%
confidence level (Type II regression). ................................................................................. 101
Table IV.5. ΔPDIC ( in %), due to error in the spectral extrapolation of [aCDOM/at](412) to the
true
300-600 nm range resulting from variation in the magnitude of at(412). PDIC is
N
calculated with averaged SCDOM , ap (λ), and for different at(412) values. PDIC is
calculated assuming aw(λ) is null........................................................................................... 112
Table IV.6. ΔPDIC ( in %), due to error in the spectral extrapolation of [aCDOM/at](412) to the
N true
300-600 nm range resulting from variation in SCDOM and ap (λ) spectra. PDIC is
N
calculated with averaged SCDOM , ap (λ), and at(412)........................................................... 112
Table V.1. Distance (in km) from ice egde within which the error on CHL > 35% (Rice fresh
snow). Values in parenthesis are for the OCL4 algorithm............................................... 141
Table VI.1. Number of SeaWiFS Level 1A MLAC images processed................................... 175
Table VI.2. Location, date, time interval, and the sun zenith angle at the time of sampling of
the match-ups.......................................................................................................................... 180
Table VI.3. Comparison of in situ with SeaWiFS-derived [aCDOM/at](412) values using
Standard and MUMM AC algorithms. ................................................................................ 183
Table VI.4. Annual DIC photoproduction estimation using various methods. The value in
parenthesis for methods 1 and 2 is the difference relative (in %) to GBC method. .... 186
Table A2. Performance of the QAA to retrieved at(λ) with empirical relationships of Lee et
al. [2002] versus QAA tuned with CASES dataset (n=46)............................................... 239
XVIII

Avant propos
Cette thèse fût d’abord et avant tout motivée par mon désir d’apprendre, mon
émerveillement face à la beauté de la Nature, ainsi que pour mon goût d’aventures. En
particulier, vivre l’expérience de l’Arctique, où en été les jours sont interminables, était pour
moi un rêve qui est devenu réalité. Mais au fur et à mesure que je me suis plongé dans mes
études, mon sentiment d’inquiétude face à l’avenir de notre planète s’est fait ressentir de plus
en plus fort. Si les changements climatiques font partie de l’histoire de la Terre et sont, par
conséquent, tout à fait naturels, la situation dans laquelle nous nous trouvons aujourd’hui est
sans précédent dans cette histoire vieille de quatre milliards d’années. En exploitant les
ressources énergétiques fossilisées, nous avons, avec une inconscience presque légitime,
modifié notre environnement au point qu’il déstabilise le fragile équilibre de la planète. Mon
sentiment d’inquiétude provient justement du fait que les conséquences de ce déséquilibre
restent hors de portée pour nous, « bipèdes bien intentionnés ». Dans l’état actuel de nos
connaissances il est impossible de savoir si nous allons ou pas vers une catastrophe naturelle
irrévocable, et ce malgré toutes nos technologies aussi avancées soient-elles. Reste
qu’aujourd’hui, lorsque nous faisons une rétrospective environnementale du siècle dernier
(e.g., déforestation, destruction de la couche d’ozone, contamination des systèmes aquatiques,
augmentation du dioxyde de carbone dans l’atmosphère, etc, etc), nous pouvons nous
attendre au pire 1 . En effet, il a été démontré que la mauvaise gestion de l’environnent a
souvent été l’une des causes premières (pas la seule bien sûre!) expliquant l’effondrement de
plusieurs sociétés humaines du passé (e.g., l’île de Pâques, Viking en Amérique) 2 .
Le sujet que je traite dans ce manuscrit concerne les changements climatiques. Si
cette thèse apporte de nouveaux indices qui montrent l’importance des changements
climatiques, en particulier ceux qui se produisent dans l’Océan Arctique, elle ne propose
aucune solution qui permettrait d’en atténuer ses conséquences. De plus, la quantité
1 Plusieurs scientifiques renommés se sont aujourd’hui engagés dans une campagne de conscientisation de la
population. À ce sujet, je recommanderais notamment les essais d’Hubert Revees (Mal de terre, édition Le seuil),
de David Suzuki (L’équilibre sacré – redécouvrir sa place dans la nature, édition Fides), et de Jared Diamond
(Effondrement, édition Gallimard).
2 J. Diamond, Effondrement.
XIX

phénoménale de gaz à effet de serre qui fut utilisée pour réaliser ce projet de thèse (les
nombreux aller-retour France-Québec, navire océanographique, satellite, Fiat Panda, MBK,
etc., etc.) pèse lourd sur ma bonne conscience. Néanmoins, comme le dit si bien David
Suzuki, « le plus important, à l’heure actuelle, est l’échange d’idées; il faut se passer le mot en travaillant
tous à réduire notre effet négatif sur la planète, en développant des infrastructures qui permettront d’adopter
un mode de vie durable et de susciter dans le public un appui qui changera à la fin les priorités politiques ».
J’espère que mon travail de thèse puisse me donner au moins la crédibilité nécessaire pour
éduquer et sensibiliser les générations futures aux conséquences éventuelles que court notre
planète. En tant que « futur » professeur de géographie, je compte jouer un rôle en ce sens,
aussi minime soit-il.
Dans mon premier chapitre, qui s’adresse à un large public, j’ai tenté d’expliquer avec
le plus de détails possible, dans un nombre limité de pages, la problématique générale de
l’ensemble de ma thèse. Ensuite, après un court chapitre qui introduit la zone d’étude où ont
été réalisés mes travaux, on retrouvera trois chapitres qui constituent ce qu’on pourrait
appeler « le cœur de la thèse ». La science le voulant ainsi, ces trois chapitres sont assez
pointus et sont écrits en anglais, la langue avec laquelle mes travaux seront le mieux
disséminés au niveau international. Chacun des chapitres écrits en anglais est précédé d’un
résumé en français qui décrit l’essentiel du contenu. Enfin, une discussion axée sur les
problèmes majeurs qui restent à explorer complètera le travail en guise de conclusion
générale. Sur ce, je vous souhaite une lecture agréable et instructive.
XX

Chapitre I : Introduction et problématique
générale
1

I.1. Introduction
La vitesse avec laquelle l’Arctique se réchauffe est sans précédent dans l’histoire
moderne de la société humaine. Les nombreuses conséquences du réchauffement de
l’Arctique, et d’autres changements environnementaux (e.g., augmentation du rayonnement
ultraviolet), ont récemment fait l’objet d’un important rapport scientifique présenté par
l’Arctic Council (Arctic Climate Impact Assessment, ACIA, 2005). Bien que les causes exactes de
ce réchauffement restent mal comprises, ce d’après le dernier rapport de l’Intergovernmental
Panel on Climate Change (IPCC, 2001), il semble que les émissions anthropiques de gaz à effet
de serre soient, en grande partie, responsables de la situation qu’on observe aujourd’hui. En
particulier, l’augmentation de la concentration atmosphérique en dioxyde de carbone (CO 2)
est significative depuis le milieu des années 50 (Keeling, 1986; Schlesinger, 1990). Le CO 2 a
la particularité, comme d’autre gaz à effet de serre tel que le méthane (CH 4), d’absorber le
rayonnement infrarouge qui est réémis par la surface de la Terre. Une meilleure
compréhension des processus qui influencent la concentration de CO 2 dans l’atmosphère est
donc primordiale dans le contexte des changements climatiques. Le sujet de cette thèse
s’intègre dans l’effort international qui vise à mieux comprendre le cycle global du carbone
afin de mieux appréhender l’évolution du climat.
L’Océan Arctique joue un rôle primordial dans le système climatique global en
permettant à la chaleur accumulée aux basses latitudes et transportée dans les hautes latitudes
de s’échapper, pour ainsi équilibrer le bilan énergétique de la planète. C’est aussi une région
qui agit comme un « moteur » pour la circulation océanique globale avec la production d’eau
dense. Par exemple, l’Atlantique Nord est le site où se forme la masse d’eau profonde
connue sous le nom North Atlantic Deep Water (NADW), laquelle tapisse les fonds d’une vaste
partie de l’Océan Mondial. La circulation générée par la formation de cette masse d’eau est
de type thermohaline car elle résulte d’un changement de densité causé par des changements
de température (thermo-) ou de salinité (-haline). À l’aide de modèle de circulation global,
Rahmstorf (2002) a mis en évidence l’impact de la production de NADW sur le transport de
chaleur à l’échelle planétaire. Par exemple, les températures de l’air en surface dans
l’hémisphère Nord seraient jusqu’à 6°C inférieures sans la formation de cette masse d’eau
(Rahmstorf, 2002).
2

En plus d’influencer la circulation océanique globale, les phénomènes dits de
«rétroactions» 1 sont très importants dans l’Arctique. Les rétroactions peuvent soit accélérer le
réchauffement global actuel (positive), soit l’inverser (négative). Les trois plus importantes,
identifiées dans le rapport de l’ACIA (2005) sont l’albédo de surface 2 , la circulation
thermohaline, et l’émission de gaz à effet de serre.
1. Rétroaction due à l’albédo : la glace de mer, et son couvert de neige, réfléchie
pratiquement 90% du rayonnement solaire vers l’espace. En réduisant l’étendue du
couvert de glace, une plus grande quantité d’énergie pénètre dans l’océan, induisant
une augmentation de la température de la couche d’eau superficielle qui accélère la
fonte de la glace. C’est une rétroaction positive.
2. Rétroaction liée à la circulation thermohaline : l’augmentation du volume d’eau
douce dans la couche superficielle de l’Arctique (en réponse à l’augmentation du
débit des rivières Arctiques et de la fonte des glaces) intensifie la stratification des
eaux de surface de l’Atlantique Nord, ce qui diminue la formation d’eaux denses et
un ralentissement global de la circulation thermohaline profonde. De tels
changements risquent d’entraîner un refroidissement généralisé dans l’Hémisphère
Nord (localisé), et peut être même à l’échelle globale. On parlera donc d’une
rétroaction négative, mais cela reste très incertain à ce jour.
3. Rétroaction due à l’émission de gaz à effet de serre d’origine naturelle : une grande
quantité de carbone organique se trouve « séquestré» dans le pergélisol (c.f. permafrost
en anglais; le sol qui est gelé en permanence) situé en marge de l’Océan Arctique. Le
pergélisol, qui représente 25% de la surface continentale de l’hémisphère nord, subit
depuis une quarantaine d’années une augmentation de sa température et, à plusieurs
endroits, une fonte progressive. Cette fonte conduit à une mobilisation du carbone
organique séquestré, qui est alors accessible à la minéralisation 3 microbienne
aérobique (production de CO 2) et anaérobique (production de CH 4). Comme on
s’attend à ce que la minéralisation excède la production primaire (fixation de CO 2),
on parlera de rétroaction positive.
1
Dans le contexte particulier de l’étude des changements climatiques, on utilise le terme feedback pour signifier
un phénomène physique, biologique ou chimique qui, soumis à un réchauffement climatique, produit une
action qui augmente (positif) ou diminue (négatif) le réchauffement (Woodwell et al., 1998).
2
Albédo planétaire : proportion de l’énergie solaire incidente qui est simplement réfléchie vers l’espace par le
système terre.
3
Minéralisation : transformation de la matière organique en composés inorganiques (e.g. CO2, CH4).
3

Dans cette thèse je m’intéresserai surtout au dernier point, i.e. à la rétroaction
positive due à l’émission de gaz à effet de serre d’origine naturelle. Dans les sections
suivantes, on détaillera la problématique de la thèse. Brièvement, la fonte du pergélisol et
l’augmentation du débit des rivières de l’Arctique devraient mobiliser une partie du carbone
organique sous forme dissoute vers l’Océan Arctique (section I.2.1). Dans l’océan, le carbone
organique dissous d’origine terrigène peut être minéralisé soit par les microorganismes, soit
par la lumière. La minéralisation par la lumière est un phénomène physico-chimique
complexe appelé photooxydation (pour oxydation photochimique) (section I.2.2). Les
rayons ultraviolets (UV) sont les principaux responsables de la photooxidation. Dans
l’Arctique, les eaux de surface sont « protégées » des UV grâce à la présence de la glace de
mer. En parallèle, la diminution de l’ozone stratosphérique permet à une plus grande
quantité de rayons UV d’atteindre la surface de la Terre (section I.2.2). L’hypothèse centrale
de cette thèse est que la photooxydation, suite à l’augmentation des apports
continentaux et du rayonnement UV pénétrant dans la colonne d’eau, est un
mécanisme clé qui contribue à accélérer la minéralisation du carbone organique
terrigène, produisant du CO 2 dans la couche superficielle de l’océan Arctique, qui
autrement aurait été « séquestré » dans l’océan profond.
L’objectif fixé est de quantifier les processus de photooxydation dans une région
de l’Arctique afin d’établir l’importance de ce phénomène dans le contexte actuel et
d’appréhender les conséquences potentielles du réchauffement climatique sur cette
rétroaction positive.
4

I.2. Problématique
I.2.1. Importance de la matière organique dissoute d’origine terrigène dans l’Océan
Arctique et impact du réchauffement des écosystèmes terrestres nordiques
Bien qu’il soit connu depuis le début des années 70 que les eaux de surface de
l’Arctique sont riches en carbone organique dissous (DOC 1 ) (Kinney et al., 1971), ce n’est
que récemment qu’on a démontré l’importance de la contribution des apports terrigènes. En
effet, de 5 à 40% du DOC des eaux de surface de l’Arctique serait d’origine terrigène
(Wheeler et al., 1997; Guay et al., 1999; Opsahl et al., 1999). L’Océan Arctique ne représente
que 1% du volume de l’Océan global mais reçoit 11% des apports fluviaux mondiaux
annuels (Rachold et al., 2004). En plus de contribuer au maintient de la stratification haline
caractéristique de l’Arctique (voir Chap. 2), ces eaux douces, riches en carbone organique
dissous, expliquent donc en partie les fortes concentrations de DOC observées dans
plusieurs secteurs de l’Arctique (e.g., Kinney et al., 1971; Anderson et al., 1998; Kattner et al.,
1999; Bussmann et Kattner, 2000; Anderson 2002; Amon, 2004; Benner et al., 2005). En plus
de la grande quantité de DOC déchargée dans l’Arctique, le mélange conservatif du DOC
observé dans les estuaires – e.g. Lena (Cauwet et Sidorov, 1996), Yenisei et Ob (Köhler et al.,
2003; Amon et Meon, 2004) - suggère que les processus de minéralisation sont pratiquement
inexistants dans les environnements estuariens polaires. En aval de la zone de floculation, où
les changements ioniques provoquent la transformation d’une fraction du DOC en particules
(i.e. dans les eaux de salinité ~0-4; Droppo et al., 1998; Köhler et al., 2003), le DOC est
éliminé soit par l’oxydation microbienne, soit par la photooxydation. L’importance
relative de ces deux processus de minéralisation est inconnue. Pour expliquer que seulement
~40% des apports terrigènes annuels en DOC est exporté à l’extérieur de l’Arctique, Hansell
et al. (2004) ont suggéré que les processus microbiens devaient être responsables de la
majeure partie de la dégradation observée. Cependant, aucune étude n’a permis de vérifier
cette hypothèse. Au contraire, d’après l’étude de Meon et Amon (2004), le DOC d’origine
terrigène est très réfractaire et ne constitue pas une source significative de carbone organique
pour les bactéries (Mer de Kara). Concernant la photooxydation, les études d’Amon et al.
(2003), Amon et Meon (2004) et de Benner et al. (2004; 2005) suggèrent que ce mécanisme
n’est pas important dans les zones polaires, sans toutefois le démontrer quantitativement.
1 DOC : dissolved organic carbon.
5

Les processus responsables de la minéralisation de la matière organique d’origine terrigène
dans l’Océan Arctique restent énigmatiques.
Une meilleure compréhension des processus de minéralisation de la matière
organique d’origine terrigène (et/ou de sa préservation) est cruciale dans le contexte
des changements environnementaux récemment observés dans les écosystèmes
terrestres nordiques. En effet, la tendance selon laquelle les écosystèmes terrestres de
l’Arctique agissent comme puits net de CO 2 et de CH 4 atmosphérique depuis la fin de la
dernière glaciation pourrait être renversée avec le réchauffement. Ces écosystèmes pourraient
alors devenir une source significative de carbone pour l’atmosphère (e.g. Oechel et al., 2000).
Les tourbières arctiques ont accumulées approximativement 455 x 10 15 g de carbone
organique depuis la dernière glaciation (Gorham, 1991), soit près d’un tiers de la quantité de
carbone organique contenu dans les sols à l’échelle globale (Dixon et al., 1994). La majeur
partie de ce carbone (~400 x 10 15 g C) se trouve actuellement emprisonnée dans le pergélisol
(Davidson et Janssens, 2006; et références citées).
Plusieurs études récentes suggèrent que la mobilisation d’une partie du carbone
contenu dans les sols arctiques se fera sous forme dissoute par le réseau hydrographique
(Freeman et al., 2001, 2004; Mack et al., 2004; Billett et al., 2004). En accord avec ces études,
Frey et Smith (2005) prévoient que la quantité de DOC exportée de la Sibérie vers l’Océan
Arctique augmentera de 29 à 46% au cours du siècle présent. Ce scénario n’est pas
surprenant vu l’augmentation prévue du débit des rivières (jusqu’à 31%; Arnell, 2005) et la
fonte progressive du pergélisol (Camill, 2005; Hinzman et al., 2005). Bien que la mobilisation
du carbone « séquestré » n’a pas été détectée avec les analyses d’isotopes stables du DOC
dans les fleuves arctiques (Benner et al., 2004; Amon et Meon, 2004; Guo et Macdonald,
2006), l’augmentation du débit des fleuves observée depuis les années 60 suggère une
augmentation significative de l’export de carbone des écosystèmes terrestres vers l’océan
(Figure I.1; Peterson et al., 2002, 2006; McClelland et al., 2006). Dans ce contexte, on peut se
poser les questions suivantes :
• Quel est le destin de ce carbone organique dissous d’origine terrigène dans
l’océan Arctique?
• Quels sont les processus de dégradation dominants?
6

Figure I.1. Anomalie débit d’eau douce (trait noir) des six plus grands
fleuves sibériens qui se déchargent dans l’Océan Arctique depuis 1940.
L’augmentation (7%) est corrélée à l’augmentation de la température de
surface à l’échelle globale (trait bleu), ainsi qu’avec l’indice climatique NAO 1
(trait rouge) (source : Peterson et al., 2002).
1 NAO: north atlantic oscillation. L’indice NAO décrit la variabilité de pression atmosphérique entre l’Islande
et les Açores. Le NAO est positif quand la pression atmosphérique aux Açores est plus élevé que la normale et
plus faible en Islande que la normale.
7

I.2.2. Importance de la photooxydation et impact de la fonte de la glace de mer et de
l’augmentation du rayonnement ultraviolet dans l’Arctique
En raison de son caractère réfractaire, l’absence d’un fort signal indiquant la présence
de matière organique dissoute d’origine terrigène (tDOM 1 ) dans l’océan (Meyers-Schulte et
Hedges, 1986) intrigue les océanographes depuis des décennies (voir Hedges et al., 1997 et
réf. citées). À l’aide d’un traceur moléculaire caractéristique des végétaux terrestres, la lignine,
Opsahl et Benner (1997) et Hernes et Benner (2002; 2006) ont estimé le temps moyen de
résidence de la tDOM à 21-93 et ~90 ans dans les océans Atlantique et Pacifique,
respectivement, ce qui suggère une minéralisation relativement rapide de cette matière
organique dans le milieu océanique.
En 1977, Oliver C. Zafiriou suggérait que les processus photochimiques pouvaient
jouer un rôle significatif dans le maintient de la transparence de l’eau de mer, et comme puits
pour le DOC réfractaire dans l’océan. Ce n’est que récemment qu’on a pu mettre en
évidence le rôle prépondérant que jouaient les réactions photochimiques dans la
transformation et la minéralisation du DOC d’origine terrigène dans l’océan côtier (Kieber et
al., 1990; Miller et Zepp, 1995; Amon et Benner, 1996; Opsahl et Benner 1998; Benner et
Opsahl 2001; Hernes et Benner 2003). Puisque les processus photochimiques n’affectent que
la couche superficielle de l’océan (i.e.

2003), lesquels sont les principaux responsables des réactions photochimiques en milieu
aquatique (Zafiriou, 1977).
On assiste depuis la fin des années 50 à une réduction de l’étendue du couvert estival
de glace dans l’Arctique (Fig. I.2). La diminution observée est particulièrement importante en
été, saison pendant laquelle l’étendue moyenne est passée de ~11 à ~7.5 10 6 km 2 en un demi
siècle. Cette tendance semble même s’accélérer d’après les observations satellitales qui ont
enregistré une étendue minimale du couvert de glace pluriannuelle 1 durant quatre années
consécutives depuis 2002 (Serreze et al., 2003; Stroeve et al., 2005; National Snow and Ice Data
Center, 2006). En septembre 2005, le couvert de glace atteignait seulement ~5.5 10 6 km 2 , ce
qui représente une réduction de l’ordre de 27% depuis le début des observations satellitales,
il y a 26 ans (Figure I.3).
Figure I.2. Observations de l’étendue moyenne du couvert de glace
saisonnier au cours du 20 ième siècle (source des données : ACIA, 2005). La
diminution de la glace est plus importante au printemps et en été, qu’en
automne et en hiver.
1 La glace pluriannuelle est celle qui survit à au moins un cycle de gel-dégel. Elle est observée de manière
distincte quand l’étendue du couvert est minimale en septembre.
9

Figure I.3. Étendue de la banquise Arctique observée par satellite
annuellement au mois de septembre entre 1979 et 2005. (Source des données:
National Snow and Ice Data Center, 2006).
En parallèle au rétrécissement du couvert de glace, on assiste à une diminution de la
concentration en ozone stratosphérique en réponse aux émissions anthropiques de
chlorofluorocarbures (CFC) qui, par une série de réactions chimiques avec les composés
chlorés et bromés, conduit à la destruction de l’ozone stratosphérique. Entre 1979 et 2000, la
concentration d’ozone a diminué de 11% en moyenne au printemps au dessus de l’Arctique
(ACIA, 2005). L’augmentation du rayonnement ultraviolet reçu à la surface en est une
conséquence directe. La Figure I.4 montre un exemple de l’impact de l’ozone sur
l’éclairement UV reçu à la surface de l’océan au midi solaire, à la latitude de 70°N, durant le
solstice d’été. Dans la littérature, le rayonnement ultraviolet comprend les longueurs d’onde
allant de 100 nm à 400 nm et se divise en trois sous-groupes : UV-C (100-280 nm), UV-B
(280-315 nm) et UV-A (315-400 nm). Si l’ozone absorbe presque entièrement les UV-C, elle
laisse pénétrer une partie des UV-B et n’affecte pratiquement pas les UV-A.
En milieu aquatique, le rayonnement UV-B agit sur les cycles biogéochimiques. Par
exemple, il diminue la productivité primaire, accélère la décomposition du DOC et du
tDOM, entraîne la production des gaz volatiles et favorise le recyclage des micronutriments
tel que le fer et le cuivre (voir la récente revue par Zepp et al., 2003). Gibson et al. (2000) se
10

sont intéressés aux facteurs contrôlants l’exposition des microorganismes aquatiques aux
radiations UV dans l’Arctique. Ils ont démontré que les variations de la concentration en
matière organique dissoutes colorée (CDOM 1 ) jouent un rôle plus important que l’ozone sur
cette exposition. L’impact des changements environnementaux sur les réactions
photochimiques impliquant le CDOM reste cependant inexploré.
Récemment, Amon et Meon (2004), en se basant sur des expériences de
photooxidation réalisées sur des eaux provenant des fleuves Ob et Yenisei, confirmèrent la
forte réactivité du CDOM, et suggèrent que « […] with reduced ice cover in the Arctic Ocean the fate
of terrestrial DOM could be significantly different compared to present situation ». Ces diverses
considérations m’amènent à formuler les questions suivantes :
• Peut-on quantifier l’impact du rétrécissement du couvert de glace et de
l’augmentation du rayonnement UV-B sur la photooxidation du CDOM dans
l’Arctique?
• Quels sont les principaux facteurs (i.e. couvert de glace, UV-b, nébulosité,
photoréactivité et concentration du CDOM) à prendre en compte?
Figure I.4. Éclairement spectral dans les parties UV-A et UV-B pour trois
concentrations en ozone telles qu’indiquées sur la figure (DU = Dobson
Units), et un exemple de rendement quantique apparent (voir la section
suivante pour plus d’explications) exprimé en unités relatives, représentant la
réactivité photochimique du CDOM en fonction de la longueur d’onde
(courbe grisée).
1 CDOM : chromophoric dissolved organique matter. On quantifie sa concentration en mesurant les propriétés
d’absorption de l’eau de mer filtrée avec des filtres dont la porosité est d’environ 0.2 – 0.7 μm (on parlera du
coefficient d’absorption du CDOM : a CDOM).
11

I.2.3. Quantification des processus photochimiques
Dans les bilans de carbone globaux ou régionaux, il est primordial de quantifier les
flux de carbone entre les réservoirs organique et inorganique, et particulaire et dissous. Dans
le présent travail, je me suis intéressé à quantifier la transformation du carbone organique
dissous (DOC) en carbone inorganique. Pour se faire, j’ai quantifié la production
photochimique (i.e. photoproduction) de CO 2, le composé inorganique le plus important des
réactions photochimiques en milieux aquatiques (e.g. Miller et Zepp 1995). En réalité, la
photoproduction de CO 2 est légèrement inférieure à minéralisation du DOC par
photooxydation en milieu marin (Miller et Zepp 1995; Miller et Moran, 1997; Gao et Zepp,
1998; Mopper et Kieber, 2002). En effet, le CO 2 représente le produit final d’une série de
réactions physiques et chimiques déclanchés par l’absorption de l’énergie radiative par le
CDOM (Zafiriou, 1977). Parmi les mécanismes connus conduisant à la production
photochimique de CO 2, on retrouve la photodecarboxylation (Miles et Brezonik, 1981;
Budac et Wan, 1992), où des groupes carboxyliques (-COOH) rattachés à des chaînes de
carbone organique sont clivés. Par ailleurs, il a récemment été démontré que le clivage des
structures aromatiques des substances humiques 1 pouvait aussi être responsable de la
production de CO 2 dans les eaux riches en DOC d’origine terrigène (Vähätalo et al., 1999;
Xie et al., 2004). Si les processus exacts menant à la production de CO 2 sont complexes et
toujours mal connus (Mopper et Kieber, 2002), il est néanmoins possible d’estimer la
photooxydation du DOC en milieu marin si les paramètres suivant sont connus (Zafiriou,
1977) (voir aussi encadré 1.1):
• L’éclairement solaire incident à la surface de l’océan, E(λ), qui dépend
principalement des conditions atmosphériques (concentration en H 2O, O 3, aérosol,
nuage, etc);
• le spectre du coefficient d’absorption des chromophores organiques dissous, a CDOM;
• l’efficacité avec laquelle le CO 2 est produit suite à l’excitation des chromophores par
les photons absorbés, i.e. le rendement quantique apparent (AQY 2 ou φ).
1 Substances humiques : mélange de matières organiques, de couleur foncée, fortement polymérisées, que l'on
peut extraire de l'humus du sol par l'action de solutions alcalines diluées et qui sont précipitables par les acides.
2 AQY: Apparent Quantum Yield.
12

Encadré 1. Modèle de production photochimique de carbone inorganique à partir du
carbone organique dissous
En milieu aquatique, le CO2 se dissous et forme des ions bicabonatés (H2CO3 et HCO3 -). L’ensemble
forme le carbone inorganique dissous ou DIC1. Le taux de photoproduction de DIC, PDIC, à une
profondeur z et suite à l’absorption de photons de longueur d’onde λ, peut être exprimé
mathématiquement par :
0
( λ , z)
= E ( λ,
z)
a ( λ,
z)
φ ( λ,
z)
, (E.1)
PDIC CDOM DIC
où E 0 est l’éclairement scalaire en moles de photon m -2 s -1 (i.e., l’intégrale de l’énergie provenant de
toute les directions dans l’espace tridimensionnel), aCDOM est le coefficient d’absorption du CDOM en
m -1(voir Encadré 2) et φDIC est le rendement quantique pour la photoproduction de DIC en moles
DIC (moles photon) -1. φDIC est le rapport entre le nombre de moles de DIC produit par moles de
photons absorbés par le CDOM.
Dans le cadre de cette étude, je m’intéresserai à la photoproduction intégrée dans toute la colonne
d’eau (sur la verticale) par unité de surface. Comme on ne connaît pas la variabilité verticale de tous
les paramètres de l’équation E.1, j’adopterai des hypothèses simplificatrices, sans perdre pour autant
trop de précision. Tout d’abord, les propriétés optiques de l’eau de mer sont supposées homogènes
dans la couche où les réactions photochimiques prennent place. Cette hypothèse est valide tant que la
couche de mélange est plus grande que la couche éclairée par le rayonnement UV et bleu, et que φDIC
est indépendant de l’intensité de E 0 (contrairement à la production primaire par exemple). Ainsi, il
suffira de connaître la quantité d’énergie qui pénètre la surface de l’océan et la fraction qui est
absorbée par le CDOM dans la colonne d’eau. En abandonnant la dimension verticale, on peut
réécrire plus simplement l’équation E.1 comme :
− aCDOM
( λ)
PDIC ( λ) = Ed
( λ,
0 ) [ 1−
R(
λ)
] φDIC
( λ)
, (E.2)
at
( λ)
où Ed est l’éclairement descendant juste sous l’interface air-mer (indiqué par 0-), R est la réflectance
de l’océan (i.e. le pourcentage d’énergie rétrodiffusée vers l’atmosphère, voir Encadré 2) et at est le
coefficient d’absorption totale de tout les constituants optiquement efficaces (i.e. l’eau de mer pure,
les particules détritiques et minérales, le phytoplancton et le CDOM). Dans la plupart des conditions
océaniques, le facteur [1-R] approche 1 et peut donc être abandonné. Enfin, en intégrant l’équation
E.2 dans le domaine spectral photochimiquement efficace, i.e. entre 290 nm et 600 nm, on obtient la
production intégrée de DIC dans la colonne d’eau,
600
− aCDOM
( λ)
PDIC = ∫ Ed
( λ,
0 ) φDIC
( λ)
dλ
. (E.3)
λ=
290
a ( λ)
Parmi les paramètres de l’Eq. E.3, le rendement quantique apparent pour la
photoproduction de CO 2 (ou DIC), φ DIC(λ), est de loin le moins bien connu. Le manque de
données sur φ DIC dans les eaux océaniques s’explique par des difficultés expérimentales
difficiles à surmonter. Tout d’abord, étant donné que la concentration de DIC dans l’eau de
mer s’éleve à ~2000 μM, une production photochimique de moins de 10 μM de DIC
1 DIC: dissolved inorganic carbon.
13
t

(pendant une expérience d’une durée de ~12 h) s’avère indétectable (Johannessen, 2000). Les
premières mesures de φ DIC(λ) proviennent donc des eaux douces récoltées soit dans les
rivières (Gao et Zepp, 1998), soit dans des lacs (Vähätalo et al., 2000) où les concentrations
initiales en DIC sont faibles, ce qui rend possible la mesure directe de la production
photochimique de DIC. Gao et Zepp (1998) ont publié les premières valeurs de φ DIC à trois
longueurs d’onde dans la partie UV du spectre (290, 350, et 390 nm) pour des eaux très
riches en DOC provenant de la rivière Satilla (Georgia, USA). Vähätalo et al. (2000) ont
proposé une méthode pour estimer le spectre de φ DIC à partir d’incubation in situ à différentes
profondeurs. Grâce à un modèle spectral de propagation de la lumière couplé à un modèle
de photooxidation, la photoproduction de DIC est calculée en exprimant φ DIC(λ)=x x 10 yλ .
Les coefficients x et y sont retrouvés en minimisant la différence entre les productions
calculées et mesurées (par une procédure d’optimisation mathématique). La méthode de
Vähätalo et al. (2000), qui demande des temps d’incubation in situ de l’ordre d’une semaine,
n’est pas adaptée aux applications océanographiques où le temps passé à une station
d’échantillonnage est relativement court (souvent

I.2.4. La télédétection de la couleur de l’océan : un outil de choix pour quantifier les
processus photochimiques aux échelles océaniques
Durant les étés de 1967 et 1968, les chercheurs américains George L. Clarke, Gifford
C. Ewing et Carl J. Lorenzen effectuaient des survols depuis Cape Cod jusqu’à la mer des
Sargasses à bord d’un avion équipé d’un spectromètre mesurant la lumière rétrodiffusée par
l’océan. Ainsi les premiers spectres de lumière sortant de la mer étaient mesurés à distance; la
télédétection de la couleur de l’océan voyait le jour. Avec des mesures in situ simultanées de
concentration en chlorophylle a, Clarke et al. (1970) démontraient la possibilité d’utiliser la
télédétection pour cartographier, depuis l’espace, la distribution de ce pigment à des échelles
spatiales encore jamais explorées à cette époque. Pour expliquer les anomalies des spectres
ne pouvant pas être attribuées à la chlorophylle a, Clarke et al. (1970) suggéraient entre autre
la présence de particules en suspension et de « biochromes ». Ces travaux pionniers
ouvraient la voie, avec le lancement en octobre 1978 du satellite Nimbus-7 équipé du capteur
Coastal Zone Color Scanner (CZCS) (Hovis et al., 1980; Gordon et al., 1980), à la télédétection
spatiale de la couleur de l’océan. Depuis, les applications utilisant les données « couleur de
l’océan » acquises depuis l’espace se sont multipliées. Ne sont citées ici que les plus
courantes : étude de la distribution spatiale et de l’évolution temporelle de la biomasse
phytoplanctonique aux échelles régionale et globale, production primaire à l’échelle globale,
étude des structures physiques océaniques tels les tourbillons et les méandres du Gulf Stream.
À l’heure actuelle, on compte neuf capteurs en orbite autour de la terre qui fournissent des
données de façon quasi-continue (voir site internet de l’IOCCG 1 ).
En 1997, Cullen et al. proposaient l’idée d’utiliser des données couleur de l’océan
pour estimer l’atténuation des UV en milieu marin afin de quantifier les flux photochimiques
et les dommages causés par les UV aux microorganismes planctoniques. Quelques années
plus tard, Johannessen (2000) développait une méthode empirique pour estimer l’atténuation
diffuse (voir Encadré 2) de la lumière dans l’UV à partir des mesures du capteur SeaWiFS
(Sea Wide Field-of-View Sensor), un capteur en orbite depuis septembre 1997. Les valeurs du
coefficient d’atténuation diffuse dans l’UV permettaient ainsi d’identifier trois types d’eaux le
long de la côte Est américaine, pour lesquelles les spectres de φ DIC(λ) avaient été déterminés
(Johannessen et Miller, 2001). Il s’agissait là de la première tentative d’utiliser la télédétection
1 IOCCG : International Ocean Colour Coordinating Group, www.ioccg.org/sensors/current.html .
15

de la couleur de l’océan comme un outil pour la quantification des processus
photochimiques. Plus récemment, Cédric G. Fichot proposait une méthode statistique
robuste permettant d’estimer, à partir des données SeaWiFS, l’atténuation des UV,
l’absorption du CDOM et le taux de photoproduction de monoxyde de carbone dans la
colonne d’eau à l’échelle globale. Les travaux de Fichot (2004), bien que peu connus,
représentent probablement le meilleur exemple de l’utilité des données de couleur de l’océan
pour l’estimation des flux photochimiques à partir de l’espace. Il s’agit, néanmoins, d’une
application encore peu exploitée dont le plein potentiel reste à découvrir.
Les données couleur de l’océan peuvent servir à améliorer la quantification de
la photoproduction de CO 2 (ou DIC) dans l’Arctique. Cependant, l’exploitation de ces
données dans les régions polaires pose plusieurs problèmes qui demandent des études plus
approfondies. Dans le cas de l’Arctique, on peut mentionner :
1) le manque de connaissance des propriétés optiques des eaux Arctiques, plus
particulièrement des eaux des plateaux continentaux, et l’absence d’algorithmes pour
estimer le CDOM;
2) la présence de glace de mer qui empêche la télédétection des propriétés optiques de
l’eau sous-jacente et complique celle des eaux environnantes;
3) l’éclairement solaire faible, voire nul, pour la majeur partie de l’année limite la
période de détection entre mai et août;
4) l’angle zénithal solaire 1 élevé qui représente un problème particulier pour l’estimation
de la contribution de l’atmosphère au signal mesuré au niveau du satellite, laquelle est
essentielle afin de retrouver le signal marin.
Dans mon travail de thèse, j’ai examiné les deux premiers points. Je présenterai d’abord une
revue de la littérature sur les propriétés optiques mesurées dans les eaux de l’Arctique. Je
montrerai pourquoi il est difficile d’estimer le coefficient d’absorption du CDOM à partir de
la télédétection. Enfin, j’expliquerai comment la glace de mer peut ruiner la qualité des
données couleur de l’océan.
1 L’angle zénithal solaire est l’angle qui se mesure entre la normale sortant de la surface de la mer et la direction
du soleil. Un angle de 0° correspond à un soleil au zénith.
16

I.2.4.1. Propriétés optiques des eaux arctiques et algorithmes pour l’estimation du
coefficient d’absorption du CDOM
Les premières mesures optiques réalisées dans l’océan Arctique ont été rapportées
par Raymond C. Smith en 1973, qui a déterminé les valeurs spectrales du coefficient
d’atténuation diffuse, K d(λ) et d’atténuation (c(λ)) (voir encadré 2). Dans la partie verte du
spectre, ces valeurs s’avéraient comparables aux eaux les plus claires connues à l’époque.
Avec ces mesures, toutes réalisées sous le couvert de glace en mai, on a longtemps cru que
les eaux arctiques faisaient partie des eaux les plus transparentes au monde.
Il a fallu attendre plus de quinze ans avant que d’autres mesures spectrales soient
réalisées dans l’Arctique. Topliss et al. (1989) ont analysé les mesures de K d(λ) acquises dans
le nord de la Baie de Baffin (Ouest du Groënland) suggérant que ces eaux diffèrent des eaux
du Cas 1 telles que définies par Morel et Prieur (1977). Topliss et al. (1989) ont noté que le
coefficient de diffusion des particules était plus élevé que pour les eaux du Cas 1, et que la
contribution des produits de dégradation était relativement importante. Peu après, les
données de K d(λ) acquises dans la Mer de Barents et dans le détroit de Fram (Fram Strait),
publiées par Mitchell (1992), montraient clairement la présence d’un fort signal provenant de
matière non-corrélée avec le phytoplancton (voir Figure 3 de Mitchell, 1992). En comparant
ces observations avec d’autres faites en Antarctique, Mitchell (1992) concluait que les eaux
de l’Arctique étaient probablement plus influencées par le CDOM d’origine terrigène. Cette
hypothèse a largement été vérifiée récemment grâce aux mesures in situ de fluorescence 1 du
CDOM (F CDOM) et d’absorption du CDOM (a CDOM) réalisées dans différents secteurs de
l’Arctique :
• Arctique central (F CDOM, Guay et al., 1999; a CDOM Pegau 2002);
• Détroit de Fram (F CDOM, Amon et al., 2003);
• Mers de Chukchi et Beaufort (a CDOM, Wang et al., 2005a; F CDOM et a CDOM,Guéguen et al.,
2005);
• Polynie des eaux du Nord (a CDOM, Scully et Miller, 2000).
1 Une fraction de l’énergie absorbée par le CDOM fait passer les électrons d’un chromophore à un niveau
d’énergie supérieur. En retournant très rapidement à leur niveau d’énergie initial, ces molécules émettent des
photons. C’est la fluorescence. La fluorescence du CDOM est souvent utilisée pour estimer la concentration en
DOC (e.g. Vodacek et al.,1997; Ferrari, 2000; Guay et al., 1999; Amon et al., 2003)
17

Dans la partie Ouest de l’Arctique, i.e. la région qui nous intéressera dans cette thèse
(Chap. 2), les études de Pegau (2002), Wang et al., (2005a) et Guéguen et al. (2005) ont
confirmé que l’absorption dans la partie bleue du spectre était largement dominée par le
CDOM. Pegau (2002) attribue les fortes valeurs de a CDOM dans les eaux de surface à
l’abondance des apports terrigènes et les faibles taux de photooxydation dans ces eaux
couvertes de glace.
Une forte absorption de la lumière par le CDOM caractérise les propriétés
optiques des eaux Arctiques; par rapport aux autres bassins océaniques (e.g., Barnard et al.,
1998). Il est reconnu que la présence de CDOM d’origine terrigène influence la réflectance
de la mer (e.g., Carder et al., 1989, 1991), et que l’estimation de a CDOM à partir des images
satellitales est possible (e.g., Sathyendranath et al., 1989). Cependant aucune méthode n’est
actuellement disponible pour estimer a CDOM à partir du spectre de réflectance. Or il s’agit d’un
paramètre nécessaire, avec le coefficient d’absorption totale (en fait, le rapport [a CDOM/a t]),
pour quantifier les flux photochimiques (Eq. E.2).
18

Encadré 2. Définition des propriétés optiques inhérentes et apparentes
En optique marine il est utile de distinguer les propriétés optiques dites inhérentes et apparentes.
Cette distinction fut proposée par Preisendorfer (1961) afin de faciliter l’interprétation des mesures
optiques et de la théorie du transfert radiatif en milieu marin. Les propriétés inhérentes (IOP1) sont
indépendantes des changements dans la distribution du champ radiatif et ne dépendent que des
propriétés intrinsèques de la matière. Il s’agit des coefficients d’absorption (a) et de l’indicatrice
angulaire de diffusion (β(ψ, Φ); avec ψ, l’angle de diffusion, compris entre 0 et π, et Φ, l’angle
azimutal compris entre 0 et 2π). Le coefficient de diffusion (b) se calcule en intégrant l’indicatrice de
diffusion dans toutes les directions :
∫
π
b ( λ)
= 2π
β ( λ,
ψ ) sin( ψ ) dψ
(E.4)
0
Le coefficient de rétrodiffusion (bb) se calcule de la même manière, mais en n’intégrant que la partie
arrière de β, i.e. avec π/2 < ψ < π. La somme de a et b donne le coefficient d’atténuation (c). a, b et c
varient en fonction de la longueur d’onde (λ) et l’unité de mesure est le m -1. Dans les eaux naturelles,
les IOP totales peuvent être obtenues en additionnant les coefficients de l’eau pure, du
phytoplancton, des détritus, des particules minérales, et du CDOM.
Les propriétés optiques apparentes (AOP1) dépendent à la fois des IOP et des changements dans la
distribution du champ radiatif (e.g., élévation du soleil, présence de nuages, état de la surface de la
mer). Les AOP s’obtiennent par la combinaison de quantité d’énergie mesurable tels l’éclairement (E)
et la luminance (L). La luminance est le rayonnement électromagnétique enregistré dans de petits
angles solides pointés sur des directions données, alors que l’éclairement est l’intégrale des
luminances dans un espace tridimensionnel donné. Par exemple, l’éclairement descendant, Ed(λ),
mesuré par un capteur cosinus est
2π
π / 2
∫ = 0 ∫ =
Ed ( λ) = L(
θ,
ϕ,
λ)
cosθ
sinθdθdϕ
(E.5)
ϕ
θ 0
où θ et ϕ représentent les angles zénithaux et azimutaux respectivement. L’éclairement ascendant,
Eu(λ), se calcule de la même façon mais en intégrant L avec θ allant de π/2 à π. Le coefficient
d’atténuation diffus, Kd(λ) est souvent utilisé en optique marine pour désigner le taux de décroissance
verticale de Ed(λ) dans la colonne d’eau :
d ln( Ed
( λ))
K d ( λ) = − (m
dz
-1). (E.6)
Pour quantifier les variations spectrales de la couleur de la mer, on fait référence à la réflectance qui
se calcule soit comme le rapport entre l’éclairement montant et l’éclairement descendant (sans unité):
Eu
( λ)
R ( λ)
= , (E.7)
E ( λ)
d
soit comme le rapport entre la luminance sortant de l’eau (Lw) dans une direction donnée (θ, ϕ) et
l’éclairement descendant qui est en quelque sorte une réflectance « directionnelle » (en anglais remote
sensing reflenctance):
Lw
( λ,
θ , ϕ)
R rs ( λ)
= . (sr
E ( λ)
-1) (E.8)
d
À partir d’un jeu limité de données acquises dans les mers de Chukchi et de Beaufort,
Wang et Cota (2003) ont évalué deux algorithmes permettant de retrouver les IOP des
1 IOP and AOP : Inherent and Apparent Optical Property.
19

principaux constituants de l’eau. Ces méthodes, dites semi-analytiques, se basent sur des
formes simplifiées des équations du transfert radiatif qui relient la réflectance de la mer aux
IOP (voir encadré 3 et aussi Mobley, 1994). Les algorithmes semi-analytiques de Lee et al.
(1999, 2001) et de Maritorena et al. (2002) testés par Wang et Cota (2003) se sont avérés
fiables pour l’estimation des coefficients d’absorption du phytoplancton et de rétrodiffusion
des particules en suspension, mais peu performantes quant à l’estimation du coefficient
d’absorption par les matières détritiques colorées (CDM 1 ). En effet, la plupart des méthodes
d’inversion regroupent le a CDOM avec l’absorption par les particules non-algales (NAPs)
(a CDM=a CDOM+a NAP) en raison de la similarité de leur spectre d’absorption. Ceux-ci peuvent
être décris avec une fonction exponentielle du type (Bricaud et al., 1981; Roesler et al., 1989),
−S
( λ−λ0
)
a cdm = acdm
( λ0
) e
(I.1)
où λ 0 est la longueur d’onde de référence et le paramètre S est connu comme étant la pente
spectrale. Seigel et al. (2002) publiaient récemment des cartes de distribution globale de l’a CDM
(dont la variabilité était attribuée principalement au CDOM qui représente en moyenne
~82% du a CDM) qui apportèrent de nouveaux indices sur les processus régulant sa
concentration dans les eaux de surface (i.e. transport vertical, blanchiment). Dans l’objectif
d’utiliser les données de couleur de l’océan pour l’étude des processus
photochimiques, il est essentiel de distinguer a CDOM de a NAP. Parmi les 12 algorithmes
semi-analytiques récemment évalués par l’IOCCG (2005) pour estimer les IOPs, aucun ne
permettait de faire cette distinction. Quelques études ont proposé des algorithmes
empiriques régionaux pour estimer a CDOM (Kahru et Michtell, 2001; D’Sa et Miller, 2003;
Johannessen et al., 2003; Kowalczuk et al. 2005). Cependant, dans les eaux côtières
optiquement complexes où la contribution des NAPs à l’absorption peut être supérieure à
celle du CDOM (Babin et al., 2003b), les algorithmes empiriques sont en général peu fiables
(IOCCG, 2006). Enfin pour pouvoir calculer la photoproduction de DIC intégrée dans la
colonne d’eau (encadré E.1), il nous faudra connaître le rapport entre l’absorption du
CDOM et l’absorption totale. Par conséquent, une méthode pour estimer la contribution
du CDOM à l’absorption totale se doit d’être mise au point pour améliorer les
estimations du taux de photoproduction de DIC intégrée dans la colonne d’eau.
1 CDM : colored detrital material.
20

Encadré 3. Modèles semi-analytiques de réflectance de la mer.
Des solutions numériques (e.g. Monte Carlo) aux équations de transfert radiatif permettent de
déduire des relations simplifiées entre les IOPs et les AOPs (Gordon et al., 1975; Morel et Prieur,
1977). Plusieurs auteurs ont proposé des expressions simplifiées pour exprimer la réflectance (Rrs;
voir encadré 2), en fonction du rapport bb/a+bb en se basant sur des analyses théoriques et des
simulations numériques. Les plus connues ont été proposées par Morel et ses collègues (Morel et
Gentili, 1991, 1993, 1996; Loisel et Morel, 2001; Morel et al., 2002),
− f ⎛ bb
⎞ f ' ⎛ bb
⎞
R = ⎜ ⎟ = ⎜
⎟
rs ( 0 )
(E.9)
Q ⎝ a ⎠ Q ⎝ a + bb
⎠
et par Gordon et al. (1988a),
− ⎛ bb
⎞ ⎛ bb
⎞
R ( 0 ) = 1 ⎜
⎟ + 2 ⎜
⎟
rs g g
, (E.10)
⎝ a + bb
⎠ ⎝ a + bb
⎠
où les facteurs f/Q et f’/Q dépendent de la géométrie de visé, de l’angle zénithal solaire et de la
concentration en chlorophylle (valable pour les eaux du Cas 1; Morel et al., 2002), et g1 et g2 sont des
constantes (0.0949, 0.0794; Gordon et al., 1988a). Plusieurs modèles d’inversion utilisés en
télédétection de la couleur de l’océan, i.e. les modèles permettant de retrouver les IOPs à partir des
mesures d’AOPs, s’appuient sur la relation proposée par Gordon et al., (1988a). On peut citer à titre
d’exemple les méthodes : d’optimisation spectrale de Garver et Siegel (1997) et de Lee et al., (1996;
1999); d’inversion matricielle de Hoge et al. (1995) et de Hoge et Lyon (1996); et d’algorithme quasianalytique
de Lee et al. (2002).
Les IOP de l’équation E.10 sont simplement la somme des IOP de chacun des constituants de l’eau.
En générale, l’absorption est
a = a + a + a = a + a + a + a
(E.11)
w
p
CDOM
où les indices w, p, NAP, phy et CDOM représentent, l’eau, les particules, les particules non-algales,
le phytoplancton et le CDOM, respectivement. Quant à la rétrodiffusion, on fait la somme entre les
coefficients des particules et de l’eau pure :
b = b + b
(E.12)
b
bw
21
w
bp
NAP
2
phy
CDOM

I.2.4.2. La glace de mer : un problème pour la télédétection de la couleur de l’océan
La télédétection de la couleur de l’océan dans les hautes latitudes fait face à un
problème additionnel, comparé aux basses et moyennes latitudes, directement lié à la
présence de glace de mer. En effet, la glace de mer est la plupart du temps recouverte de
neige pouvant réfléchir plus de 90% de la lumière visible et du proche-infrarouge incidente
(e.g. Grenfell et Maykut, 1977; Perovich et al., 1998, 2002). En comparaison, l’océan réfléchi
en général moins ~10% de la lumière dans le visible, et moins de 3% de la lumière du proche
infrarouge (e.g. Morel, 1988; Morel et Maritorena, 2001). À l’époque du capteur CZCS, cette
différence de réflectance causait des problèmes pour la télédétection des eaux le long de la
glace en « aveuglant » les détecteurs du capteur (Comiso et al., 1990). L’amélioration des
performances radiométriques des nouveaux capteurs (e.g., le capteur SeaWiFS; Gordon et
Wang, 1994) permet aujourd’hui de faire des observations même à proximité de la glace de
mer (Wang et al., 2005b). Cependant, la présence de glace génère des problèmes d’autres
natures. En comparant les estimations de chlorophylle a obtenues par SeaWiFS avec des
mesures in situ couvrant la plupart des bassins océaniques, Gregg et Casey (2004) ont
identifié un biais de l’ordre de 100% dans les eaux situées en marge de l’océan Arctique (mer
de Barents), soit la pire performance de SeaWiFS à l’échelle régionale. Pour expliquer cette
piètre performance, ces auteurs suggéraient que « […] contamination by drifting, subpixel-scale sea
ice undetected in the processing algorithms […] » devait en être la cause. Pour illustrer le problème,
j’ai représenté schématiquement les trajectoires possibles d’un photon émis par le soleil par
temps dégagé, dans un système atmosphère-océan influencé par la présence de glace de mer
(Fig. I.5). En plus des trajectoires connues, et prises en compte dans le traitement des
données couleur de l’océan (voir encadré 4), la glace de mer peut contribuer au signal détecté
au niveau du satellite de façon 1) directe, quand, de taille bien inférieure à celle d’un pixel,
elle se trouve à l’intérieure dans un pixel (contamination sub-pixel), ou 2) indirecte, quand
elle se trouve à proximité et que la lumière qu’elle réfléchie est redirigée par diffusion vers le
capteur qui vise un pixel occupé que par de l’eau (effet dit de l’environnement; Tanré et al.,
1979, 1981).
L’impact de la contamination des données couleur de l’océan par la glace de
mer n’a jamais fait l’objet d’étude. Ayant comme objectif d’utiliser les données acquises
par télédétection pour estimer le taux de photoproduction de DIC dans les eaux arctiques,
22

une étude dédiée au problème spécifique que pose la glace de mer vis-à-vis de l’exploitation
des données de couleur de l’océan, était nécessaire.
Figure I.5. Trajectoires possibles dans le système atmosphère-océan pour un
photon émis par le soleil par temps dégagé (voir aussi encadré 4). De la
quantité de photon diffusée dans l’océan qui parvient à sortir de l’océan en
direction du satellite (a; L w), une partie sera transmise à travers l’atmosphère
jusqu’au capteur (b) alors que l’autre partie sera perdue par diffusion ou
absorption dans l’atmosphère (c). Certains photons en provenance du soleil
sont transmis directement (d) ou indirectement (e) dans le champ de visée du
capteur après avoir été réfléchis de façon spéculaire par l’interface air-mer (L r,
en anglais sun glint ou glitter). Comme pour L w, une partie sera transmise (g) et
l’autre sera perdue dans l’atmosphère (f). Dans la partie visible du spectre,
plus de 90% de l’énergie reçue au satellite provient de l’atmosphère (L p) et
comprend : les photons diffusés une (h) ou plusieurs (i) fois par les
molécules ou les aérosols atmosphériques. Dans les régions polaires où la
glace de mer est présente, il faut ajouter à ce schéma la contribution des
photons réfléchis par la glace à l’extérieur du champ de visée qui sont
redirigés vers le capteur (l; effet de l’environnement). Enfin, quand la glace
se trouvant dans le champ de visée du capteur, une partie des photons qu’elle
réfléchira sera transmise (m; contamination sub-pixel) et l’autre sera
perdue dans l’atmosphère (n). Les contributions indirectes (l) et directes (m)
posent problème puisqu’elles ne sont pas prises en compte dans le traitement
des données de couleur de l’océan.
23

Encadré 4. Introduction au traitement des données couleur de l’océan
Un instrument dédié à la télédétection de la couleur de l’océan doit enregistrer une certaine quantité
d’énergie radiative en la convertissant en un courant électrique dont l’intensité sera fonction de
l’énergie reçue. Le signal ainsi créé est ensuite digitalisé et traité pour être converti en une quantité
physique en W m -2 nm -1 (corrections radiométriques élémentaires: première calibration, application
des gains, etc.) La quantité d’énergie enregistrée dépend de la portion du spectre électromagnétique
concernée et également du champ de vue instantané (IFOV en anglais pour Instantaneous field-of-view),
qui correspond à l’élément de base observable (pixel). Le rayonnement électromagnétique enregistré,
i.e. la luminance, est qualifiée de luminance totale (Lt), dans la mesure où son intensité dans une
direction donnée est déterminée aussi bien par les propriétés optiques de l’océan que par celles de
l’interface air-mer et surtout celles de l’atmosphère (y compris les nuages ou la glace s’ils sont
présents). Comme cette luminance contient des informations sur la totalité du système océanatmosphère,
il faut en premier lieu retrancher de ce signal la contribution atmosphérique.
L’ensemble des opérations nécessaires pour extraire la « luminance marine » (Lw) du signal total est
connu sous le terme de « corrections atmosphériques ». Elles consistent en effet à éliminer la partie
du rayonnement qui a été rétrodiffusée par les molécules (LR) et les aérosols de l’atmosphère (LA et
LRA), éventuellement aussi réfléchie par la surface de l’océan (Lr), mais qui n’a jamais pénétré l’océan.
Les corrections atmosphériques sont en général basées sur une décomposition du signal en
différentes contributions telles que
Lt = (LR + LA + LRA) + TLr + tLw . (E.13)
Les deux contributions majeures au signal mesuré proviennent de la diffusion par les molécules de
l’air et de la diffusion par les aérosols (LR + LA + LRA = Lp). L’estimation de la part du signal due à la
diffusion par les molécules (LR) ne pose pas de problèmes majeurs (e.g., Gordon et al., 1988b ;
Gordon et Wang, 1992a,b). Le terme comprenant la réflexion spéculaire du soleil (TLr) par l’interface
air-mer peut être estimé en connaissant la vitesse et la direction du vent, le coefficient de réflexion de
Fresnel, et l’angle par rapport au soleil (Cox et Munk, 1954; Ebuchi et Kizu, 2002).
L’estimation de la part du signal dû à la diffusion par les aérosols (LA + LRA) est, en revanche, la
difficulté majeure de la correction atmosphérique, puisque ni la concentration, ni l’indicatrice de
diffusion sont connues a priori, c’est-à-dire avant de réaliser la correction du signal. Pour déterminer
ces deux inconnues, au moins deux équations doivent être posées, ce qui signifie dans la pratique que
des informations doivent être obtenues à au moins deux longueurs d’onde. La plupart des techniques
actuellement utilisées reposent sur l’observation du système océan + atmosphère dans au moins deux
canaux du proche infrarouge, pour lesquels le signal océanique est nul (en tout cas dans les eaux du
Cas 1). Une fois corrigé de l’effet de la diffusion moléculaire et de la réflexion spéculaire, le signal
restant est entièrement dû aux aérosols. À partir de l’intensité de ce signal et de sa dépendance
spectrale entre les deux longueurs d’onde considérées, on obtient suffisamment d’informations sur
l’aérosol en présence pour pouvoir en extrapoler la contribution vers les longueurs d’onde du
domaine visible, et ainsi corriger les luminances mesurées des effets atmosphériques (pour plus
détails voir : Gordon et Wang, 1994 ; Gordon, 1997 ; Antoine et Morel 1999).
24

I.3. Objectifs et Organisation de la Thèse
La problématique décrite ci-dessus a été abordée dans le cadre du programme
CASES (Canadian Arctic Shelf Exchange Study). CASES est un vaste programme de recherche
canadien mis sur pied en 2000 avec la collaboration de scientifiques provenant de sept pays
différents (Canada, État-Unis, Japon, Danemark, Norvège, Espagne, France). Le but ultime
de CASES est de mieux comprendre les conséquences biogéochimiques et écologiques
résultant de la fonte de la glace présente sur le plateau continental du Mackenzie. Entre
octobre 2003 et août 2004, le navire de recherche canadien CCGS Amundsen est resté sur le
site d’étude de CASES. Les données optiques et photochimiques qui furent acquises dans ce
cadre ont permis de poursuivre les objectifs spécifiques suivants :
1. Déterminer, pour la première fois, le rendement quantique pour la production de
DIC sur des échantillons d’eau provenant de l’Arctique.
2. Afin de déterminer l’importance relative de la photooxydation dans le bilan de
carbone actuel, comparer le taux de photoproduction de DIC aux taux de
production primaire totale, nouvelle et séquestrée, ainsi qu’à la demande en carbone
par les bactéries.
3. À partir de l’archive d’observations satellitales de la glace de mer, de l’ozone
stratosphérique et du couvert nuageux obtenue entre 1979 et 2003, évaluer l’impact
des changements climatiques sur la photoproduction annuelle de DIC.
4. Développer une méthode pour estimer le rapport entre les coefficients d’absorption
du CDOM et total ([a CDOM/a t]) dans les eaux côtières à partir des données de couleur
de l’océan. Évaluer la sensibilité du modèle de photoproduction de DIC à
l’extrapolation [a CDOM/a t], estimé à 412 nm, entre l’UV et le vert.
5. Décrire et comprendre l’influence de la glace de mer sur l’estimation de la luminance
marine à partir des données de couleur de l’océan, ainsi que les erreurs résultantes sur
les produits géophysiques qui sont dérivés de ces luminances contaminées (i.e. la
chlorophylle, les IOPs). Proposer une méthode pour détecter les pixels contaminés
par la glace de mer.
6. Quantifier la photoproduction de DIC en utilisant les données de couleur de l’océan.
Afin de familiariser le lecteur à la zone d’étude, le Chapitre II présente une brève
revue des caractéristiques générales de l’Océan Arctique et du sud-est de la Mer de Beaufort.
25

Ensuite, les objectifs énumérés ci-dessus seront abordés à travers quatre chapitres
relativement indépendants et présentés sous forme d’articles scientifiques. Chacun est
constitué des sections introduction, méthodes, résultats, discussion et conclusion. Le
Chapitre III répond aux trois premiers objectifs concernant le problème de la
photooxidation du CDOM dans les eaux arctiques. Dans le chapitre IV, je propose une
méthode qui permet d’estimer la fraction de l’énergie radiative absorbée par le CDOM dans
la colonne d’eau à partir des données de couleur de l’océan. Le problème supplémentaire que
pose la présence de glace de mer sur la qualité des données de couleur de l’océan acquises en
milieu polaire est développé dans le chapitre V. Le Chapitre VI démontre l’utilité des
données couleur de l’océan pour l’estimation de la production photochimique de DIC dans
les eaux côtières de la mer de Beaufort.
En guise de conclusion, quelques perspectives d’avenir sur les travaux nécessaires à
faire dans les domaines d’études abordés au cours de ma thèse seront discutées.
26

I.4. Thesis objectives and organization
The problematic described above has been addressed within the frame work of the
CASES program (Canadian Arctic Shelf Exchange Study). CASES is an important
Canadian initiative created in 2000 with the collaboration of scientists coming from seven
different countries (Canada, United States, Japan, Norway, Spain and France). The
ultimate objective of CASES is to better understand the ecological and biogeochemical
consequences resulting from the reduction in the sea ice cover over the Mackenzie Shelf
and in the Amundsen Gulf. Between October 2003 and August 2004, the Canadian
research vessel CCGS Amundsen stayed in the CASES study area. The optical and
photochemical measurements collected during this field campaign allowed me to address
the following specific objectives:
1. To determine, for the first time, the apparent quantum yield for the
photoproduction of dissolved inorganic carbon (DIC) on samples collected in the
Arctic.
2. To assess the relative importance of CDOM photooxidation in the carbon budget,
by comparing the rate of DIC photoproduction to 1) the rates of the total, new and
sequestered primary production, and 2) the rate of bacterial carbon demand.
3. To assess the impact of environmental changes on the annual DIC
photoproduction using archived satellite observations of sea ice, ozone and cloud
cover from 1979 to 2003.
4. To develop a method to derive the ratio between chromophoric dissolved organic
matter (CDOM) and total absorption coefficients ([aCDOM/at]) from Ocean Color
data in coastal waters, which is a parameter of the DIC photoproduction model.
To assess the impact of the spectral extrapolation of [aCDOM/at], measured at 412
nm, from the UV to the green part of the spectrum on the DIC photoproduction
estimation.
5. To describe and to better understand the effect of sea ice on the retrieval of the
water leaving radiance from satellite Ocean Color data, and the resulting errors on
the geophysical parameters derived from the contaminated spectrum (i.e.
27

chlorophyll a concentration, IOP, [aCDOM/at]). To propose a method to detect the
contaminated pixels by the sea ice.
6. To quantify the depth-integrated DIC photoproduction using Ocean Color data.
Except the Chapter II, which provides a brief overview (in french) of the main
characteristics of the Arctic Ocean, the Beaufort Sea and the Mackenzie Shelf, each
Chapter is presented under the form of an usual scientific paper with: abstract,
introduction, materials and methods, results, discussion and conclusions sections. Chapter
III addresses the first three objectives enumerated above, which are related to the role of
CDOM photooxidation in the Arctic waters. In Chapter IV, I propose a method to
estimate the contribution of CDOM to the total light absorption coefficient from Ocean
Color data in coastal waters. The specific problems posed by the presence of sea ice in
the exploitation on Ocean Color imagery at high latitude are studied in Chapter V. Using
the results of Chapters III to V, the Chapter VI demonstrates the utility of Ocean Color
remote sensing to quantify the DIC photoproduction in the coastal waters of the Beaufort
Sea.
In the overall conclusions and perspectives, I will discuss a number of issues
related to Ocean Color remote sensing and to the quantification of the CDOM
photooxidation in the Arctic waters that required more investigation in the future.
28

Chapitre II : L’Océan Arctique, la Mer de Beaufort
et le Plateau du Mackenzie
29

II.1. Introduction
La Mer de Beaufort est située dans la partie ouest de l’Arctique, le long des côtes des
Territoires du nord-ouest du Canada et de l’Alaska (Fig. II.1). Afin de mieux comprendre le
contexte géographique et océanographique de la Mer de Beaufort, ce chapitre offre d’abord
une description générale de l’Océan Arctique : les masses d’eau, la circulation générale, les
différents types de glace et la biologie au sens large. Ensuite je décris un cycle annuel typique
des conditions physiques qui caractérisent le Plateau du Mackenzie, qui se situe dans le sud-
est de la Mer de Beaufort. L’emphase sera mise sur la variabilité de la glace de mer et de la
distribution des eaux douces qui proviennent du Fleuve Mackenzie et/ou de la fonte de la
glace. Enfin je résume, en quelques lignes, le bilan de carbone organique du Plateau du
Mackenzie. Ce bilan, tel que synthétisé par Macdonald et al. (1998 ; 2004b), intègre les
connaissances accumulées depuis les années 70 à travers de nombreuses études réalisées dans
cette région connue pour ses réserves en hydrocarbures.
30

Figure II.1. Patrons de circulation générale des eaux de surface dans l’Océan
Arctique. Les eaux Atlantique et Pacifique (en rouge) pénètrent dans
l’Arctique via la Mer de Barents et le détroit de Bering respectivement. Le
Tourbillon de Beaufort et le courant Transpolaire (en bleu) sont les deux
principaux chemins empruntés par les eaux Arctiques, lesquelles sont
éventuellement exportées vers l’Atlantique Nord via de détroit de Fram et
l’Archipel Canadien. Ces eaux de surface sont fortement influencées par les
grands fleuves Sibériens et Nord Américains. Les huit plus importants sont
indiqués avec, entre parenthèses, leur débit annuel (volume d’eau douce en
km 3 ans -1 ). Le cadre noir localise la zone d’étude. (Source des informations :
ACIA, 2005)
31

II.2. Généralités sur l’Océan Arctique
L’Océan Arctique représente 2.6% de la surface de l’Océan mondial, mais seulement
1% en terme de volume. Cela s’explique par sa géomorphologie caractérisée par un large
plateau continental occupant 52.7% de son étendue (Jakobsson et al., 2004). Les échanges
avec les océans Pacifique et Atlantique sont relativement limités et confinées entre les
continents Eurasien, Américain et le Groenland (Fig. II.1), ce qui fait de l’Arctique un Océan
dit « méditerranéen ». La majeure partie de ces échanges se fait avec l’Atlantique à travers le
détroit de Fram (Fram Strait; voir Rudels et al., 2005) et la mer de Barents (voir Indvaldsen et
al., 2004). Les eaux d’origine Atlantique sont plus salées (S~35; Indvaldsen et al., 2004) que
celles d’origine Pacifique (31.9-33; Woodgate et al., 2005b). Lors de son passage dans la mer
de Barents et le détroit de Fram, une partie des eaux Atlantiques se refroidit et plonge pour
renouveler les masses d’eau profondes de l’Arctique. Le temps de résidence des eaux
profondes arctiques est de ~25 ans entre 200 et 1000 m, et de ~300 ans au delà de 1000 m
de profondeur (Macdonald et al., 2004a). Les eaux profondes sont isolées de la surface par la
une couche connue sous le nom de halocline (50 à 200 m). Cette dernière est alimentée par
des eaux formées sur les plateaux durant la formation des glaces (e.g., Aagaard et al., 1981), et
par les eaux du Pacifique qui pénètrent dans l’Arctique via le détroit de Bering (e.g.,
Woodgate et al., 2005a). La halocline joue un rôle important dans la stabilité de l’Arctique car
elle isole l’atmosphère et la glace superficielle d’un important réservoir de chaleur sous-jacent
(e.g., Aagaard et al., 1981; Rudels et al., 1996).
La couche de surface, appelée « couche mélangée polaire » (Polar Mixed Layer, PML),
s’étend en 0 et 50 m avec une salinité caractéristique de ~31.6 (Aagaard et al., 1981). Elle est
maintenue en grande partie par les apports continentaux (e.g., Bauch et al., 1995). Les grands
fleuves de Sibérie et de l’Amérique du Nord déchargent annuellement un volume d’eau
douce de 4000 km 3 dans les différentes mers environnantes (Fig. II.1; Barents, Kara, Laptev,
Est Sibérienne, Chukchi, Bering et Beaufort), soit ~11% des apports continentaux globaux
(Aagaard et Carmack, 1989; Rachold et al., 2004). Ces propriétés varient saisonnièrement
avec la formation et la fonte des glaces, ainsi qu’avec la crue estivale. En été, la surface est
souvent occupée par une couche de 5 à 10 mètres d’eau relativement douce résultant de la
fonte de la glace de mer et des apports fluviaux (Macdonald et al., 2004a).
32

La circulation des eaux de surface dans l’Arctique est relativement bien connue grâce
au suivi des glaces dérivantes. Les trois composantes principales de la circulation sont (Fig.
II.1) :
1) le courant anticyclonique de la Gyre de Beaufort située dans la partie
ouest de l’Arctique;
2) le courant « Transpolaire » (Transpolar Drift) qui s’écoule depuis la côte
Sibérienne jusqu’au détroit de Fram;
3) les courants cycloniques côtiers induits par la force de Coriolis.
Le Tourbillon de Beaufort et le Courant Transpolaire sont influencés fortement par les
différentes phases (cyclonique et anticyclonique) de la circulation atmosphérique. Lorsque la
circulation anticyclonique prédomine (e.g. dans les années 80), le Tourbillon de Beaufort
s’élargit et tend à faire recirculer les eaux surfaces du Courant Transpolaire dans le Bassin
Canadien. Sous ces conditions, les glaces s’accumulent dû à la convergence et à la formation
de crêtes de pression (ice ridges). Lorsque la circulation cyclonique prédomine (e.g. de 1989 à
1996), le Tourbillon de Beaufort s’affaiblit, les glaces sont transportées le long des côtes de
l’Alaska et de l’Eurasie et sont exportées hors de l’Arctique via le Courant Transpolaire (voir
Rigor et al., 2002). Sous le PML, les courants sont cycloniques et sont influencés par les
apports provenant de l’Atlantique et du Pacifique (e.g., Pickart, 2004).
Une caractéristique unique de l’Arctique par rapport aux autres océans est la
présence d’un couvert de glace permanent (multi-year ice) dont l’épaisseur varie de 1.8 à 5 m
(e.g., Wadhams et Davis, 2000). Aujourd’hui, la banquise pluriannuelle s’étend sur ~5.5 10 6
km 2 (Fig. I.3). La variabilité saisonnière du couvert de glace résulte de la formation/fonte de
glace annuelle (first-year ice) dont l’étendue peut atteindre entre 7 et 9 millions de km 2 au
début du printemps. Comme la glace pluriannuelle, la majeure partie de la glace annuelle est
mobile. Le long des côtes arctiques, cependant, une banquise continentale (landfast ice),
immobile, se développe progressivement à partir d’octobre ou novembre. En hiver,
néanmoins, une frange libre de glace se forme entre la banquise continentale et le couvert
mobile de glace présent plus au large. L’étendue de ces zones varie selon le mouvement des
glaces provoqué par les vents et les courants. Ces zones d’échanges directs de chaleur entre
l’océan et l’atmosphère se nomment polynies ou leads (Smith et al., 1990).
La glace de mer joue un rôle limitant pour: 1) la pénétration de la lumière dans la
colonne d’eau; 2) les échanges de chaleur entre l’atmosphère et l’océan; 3) l’action mécanique
33

des vents et le mélange vertical. À première vue, la présence de glace dans l’Arctique est
défavorable au développement de la vie marine. Bien que la présence de glace de mer limite
la pénétration de la lumière dans la majeur partie de l’Océan Arctique, la production
biologique y est malgré tout significative (Wheeler et al., 1996). En mesurant les taux de
production primaire et bactérienne, Wheeler et al. (1996) mettaient en évidence le rôle
prédominant que jouaient les microorganismes dans le cycle du carbone dans les eaux de
l’Arctique. En particulier, les travaux de Wheeler et al. (1996) démontrent l’importance des
organismes hétérotrophes dans la reminéralisation relativement rapide des nutriments en
surface. D’après Anderson et al. (2003), >90% de la production primaire totale (~15 g C m -2
an -1 ; Gosselin et al., 1997) est minéralisée dans les 50 premiers mètres de la colonne d’eau. La
production primaire est, cependant, beaucoup plus élevée dans les mers situées en marge de
la banquise centrale (voir Sakshaug, 2004 et réf. citées). Pour des superficies similaires, la
production primaire sur les plateaux continentaux Arctiques est environ quatre fois plus
élevée que la production dans les bassins profonds (329 vs 50 Tg C an -1 ; Sakshaug, 2004). À
titre d’exemple, dans la polynie des eaux du Nord (en anglais North Water Polynya), les taux de
production primaire totale peuvent atteindre 377 g C m -2 an -1 (Klein et al., 2002), soit des
valeurs comparables à la plupart des zones côtières des basses latitudes (e.g., Antoine et al.,
1996). Ces observations indiquent que, régionalement, la production biologique varie de plus
d’un ordre de grandeur dans l’Arctique.
34

II.3. Généralités sur la Mer de Beaufort et le Plateau du Mackenzie
La Figure II.2 montre les principaux courants rencontrés dans la Mer de Beaufort.
En surface, au large, les eaux se déplacent d’est en ouest dans le mouvement anticyclonique
généré par le Tourbillon de Beaufort. L’intensité de ce courrant dépend fortement de la
direction générale des vents de surface (section précédente; Rigor et al., 2002). Près des côtes,
les eaux de surface peuvent circuler soit vers l’Est par temps calme sous l’influence de la
force de Coriolis, soit vers l’Ouest quand les vents dominants proviennent de l’Est ou du
Sud-Est (Carmack et Macdonald, 2002). À une centaine de mètres de profondeur, à la limite
du Plateau continental de l’Alaska, un fort courant cyclonique de sub-surface transporte des
eaux provenant du Pacifique et de la Mer de Chukchi (Pickart, 2004).
Le sud-est de la Mer de Beaufort est caractérisé par la présence d’un large plateau
continental occupant plus de 60 000 km 2 , bordé à l’est par le Golfe d’Amundsen, à l’ouest
par le canyon du Mackenzie, au sud par le delta du fleuve Mackenzie, et par la Mer de
Beaufort au nord (Fig. II.2). D’après Macdonald (2000), le Plateau du Mackenzie est le plus
estuarien des plateaux Arctiques où le temps de résidence des eaux de surface est inférieur à
une année. La circulation y est très variable, souvent reliée à l’importante saisonnalité du
débit du fleuve Mackenzie. Ce dernier décharge ~330 km 3 d’eau douce chaque année
(Macdonald, 2000), soit le quatrième plus important à l’échelle de l’Arctique et le quinzième à
l’échelle global (Mckee et al., 2004).
35

Figure II.2. Carte de localisation de la Mer de Beaufort avec la direction des
principaux courants : la Gyre de Beaufort (en gris); le sous-courant de
Beaufort (c.f., Beaufort Undercurrent; en bleu); la dispersion attendue du panache
du Mackenzie (en rouge) qui varie selon la direction et l’intensité des vents
dominants. L’encadré en gris montre les limites de l’échantillonnage effectué
durant le programme CASES.
36

II.3.1. Cycles Saisonniers sur le Plateau du Mackenzie
Inspirée du modèle conceptuel présenté par Macdonald et al. (1995), Macdonald
(2000) et Carmack et Macdonald (2002), cette section présente un cycle typique de la
dispersion de l’eau douce et la formation/fonte de la glace de mer sur le Plateau du
Mackenzie (Fig. II.3).
Figure II.3. Cycle saisonnier de l’eau douce sur le Plateau du Mackenzie
caractérisé par les apports continentaux et la glace de mer : A) englacement
(apports continentaux faibles, la glace commence à se former); B) fin de
l’hiver (apports continentaux faibles, la glace a atteint son épaisseur maximale
et a arrêté de croître); C) débâcle printanière (apports continentaux élevés, la
glace demeure intacte et reste près de la côte); D) été (apports continentaux
élevés, la glace a fondue ou a été exportée vers l’intérieur de l’Océan).
(Modifiée de Macdonald, 2000).
Avant la formation des premières glace à l’automne, i.e. en septembre, l’eau douce
provenant de la fonte des glace et du fleuve est mélangée dans les 15-20 premiers mètres de
la colonne d’eau sous l’action des vents (Carmack et al., 1989). L’intensité du mélange
influence la salinité des eaux de surface, et par conséquent, la formation d’eau profonde
(Macdonald et al., 1995). Par exemple, si le mélange est intense, les eaux de surface seront
plus salées et, par conséquent, la densité des eaux saumâtres produites lors de la formation
de la glace sera plus élevée (formation d’eau profonde plus efficace).
37

Figure II.4. Variabilité journalière entre 1976 et 1998 (A) du débit du fleuve
Mackenzie, (B) du flux de particules en suspension d’origine terrigène, et (C)
de l’éclairement photosynthétique incident à 71°N et la température de l’air
mesurés à Tuktoyaktuk situé à l’embouchure du fleuve. (Modifiée de O’Brien
et al., 2006).
38

À partir du mois d’octobre, la glace se forme rapidement à mesure que les
températures diminuent (Fig. II.4C). Pendant ce temps, le débit du fleuve diminue et ses
eaux se dispersent sous la glace. Au large, la divergence des glaces crée des zones d’eau
ouverte. À la fin de l’hiver, la banquise continentale fait environ 2 m d’épaisseur et s’étend
jusqu’à l’isobathe d’environ 20 m (Fig. II.3B). La marge externe de la banquise continentale
est soumise aux mouvements de la banquise Arctique qui se déplace sous l’effet des vents.
Ces mouvements favorisent l’accumulation de glace de mer pouvant atteindre plusieurs
mètres d’épaisseur (>20m). La zone d’accumulation, communément appelé stamukhi, forme
un barrage naturel de glace retenant près de la côte les eaux du Fleuve Mackenzie. Au début
du printemps, les eaux accumulées forment un « lac » saisonnier contenant jusqu’à 70 km 3
d’eau douce (le lac Herlinveaux). Au-delà du stamukhi, on retrouve un corridor d’eau ouverte
entre les banquises continentale et Arctique qui varie selon les conditions de vents et qui fait
quelques dizaines de km de large (flaw lead polynya). Cette zone, libre de glace et en contact
direct avec l’atmosphère, est le lieu d’importants échanges de chaleur favorisant la
convection profonde des eaux situées à la marge continentale. La production primaire y est
faible au cours de cette période car limitée par la lumière (cycle solaire, glace, mélange
vertical intensif).
La débâcle printanière commence au sud par la crue du fleuve Mackenzie et
progresse graduellement vers le nord (Fig. II.3C). En juin, le débit du fleuve est maximal (Fig.
II.4A) et inonde la zone côtière jusqu’à ce que le stamukhi cède sous la pression générée par
la masse d’eau douce se trouvant en amont. Cette poussée d’eau relativement chaude et très
turbide intervient souvent avant même que la glace n’ait commencé à fondre sous l’action du
rayonnement solaire. Rapidement, le panache s’étend sur plusieurs milliers de kilomètres
carrés au-dessus du plateau continental. Une image acquise par le capteur SeaWiFS au
moment de la débâcle illustre bien la dispersion des eaux du Fleuve Mackenzie au moment
de la débâcle (Fig. II.5). La couleur brunâtre des eaux du Mackenzie indique la présence
d’une forte concentration de particules en suspension à cette époque de l’année. L’étendue
du panache arrive même au-delà du plateau continental, sous la banquise Arctique.
39

Figure II.5. Image SeaWiFS acquise le 15 Juin 2004 montrant le début de la
débâcle. Rapidement les eaux turbides du Mackenzie occupent une large
partie des eaux ouvertes du plateau continental (~15 000-20 000 km²).
En été, la glace de mer font graduellement avec la chaleur apportée par les eaux du
Fleuve Mackenzie (~15°C) et l’énergie solaire absorbée en surface. En règle générale, les
eaux sont fortement stratifiées, dû à l’importance de l’eau douce présente (Macdonald et al.,
1989; Carmack et al., 1989). La Figure II.6 montre l’étendue de la stratification des 40
premiers mètres de la colonne d’eau au-dessus du plateau continental. Le déplacement du
panache turbide du Fleuve Mackenzie dépend principalement des conditions de vent (Fig.
II.2). Quand les vents sont calmes, la force de Coriolis tend à maintenir les flots d’eau douce
le long de la côte en direction est, avant d’éventuellement d’entrer dans le Golfe
d’Amundsen (Macdonald et al., 1987; Carmack et Macdonald, 2002). Par vent d’est,
cependant, les eaux du Mackenzie sont transportées au large dans le Bassin Canadien (e.g.
Macdonald et al. 1999, 2002). La direction que prennent les eaux du Fleuve Mackenzie est
importante car elle influence le temps de résidence des eaux douces dans l’Arctique. Il
semble que les changements dans la circulation atmosphérique (dominante cyclonique),
observés depuis la fin des années 80, aient favorisé l’accumulation des eaux douces dans le
Bassin Canadien (voir les discussions de Macdonald et al., 2002 et de Proshutinsky et al., 2002
sur les conséquences potentielles de ce changement).
40

Figure II.6. Coupe transversale effectuée à partir de l’embouchure du fleuve
Mackenzie (Kugmalit Bay) jusqu’à la limite du plateau continental en juillet
2004 dans le cadre du programme CASES : a) densité; b) salinité.
II.3.2. Bilan de Carbone Organique sur le Plateau du Mackenzie
Un bilan de carbone organique pour le Plateau du Mackenzie a été proposé par
Macdonald et al. (1998, 2004b). Ce budget ne concerne que la fraction particulaire du
carbone. Selon ce budget, le Plateau du Mackenzie est une zone majeure d'enfouissement de
carbone organique à l'échelle de l'Arctique. Bien que la production primaire génère plus de
carbone organique particulaire que l'apport du Mackenzie, le carbone enfoui est presque
entièrement d'origine terrigène. En effet, le Mackenzie exporte annuellement 2.1 x 10 12 g de
POC d’origine terrigène, soit l’équivalant d’environ deux tiers de la production primaire
marine (3.3 x 10 12 g de C an -1 ; Macdonald et al., 1998). À l’échelle régionale, si 65% du
carbone terrigène est préservé dans les sédiments du delta (40%) et du plateau continental
(25%), 97% de la production primaire est minéralisé (Macdonald et al. 1998, 2004b).
Cependant, on connaît encore peu de chose sur les processus exacts de minéralisation de la
41

matière organique (activité hétérotrophique dans la colonne d’eau et les sédiments), et/ou le
transfert vers les échelons supérieurs de la chaîne alimentaire (e.g. zooplankton, etc). En
faisant le bilan, le Plateau du Mackenzie représente une source de carbone inorganique (i.e.
CO 2) de l’ordre 0.64 x 10 12 g de C an -1 (l’écosystème est dit « hétérotrophe net »).
Il est intéressant de noter que le carbone particulaire terrigène transporté par le fleuve
Mackenzie est en majeure partie « ancien » (>7000 ans) (Goñi et al. 2005). Il semble donc
qu’il proviendrait essentiellement du pergélisol (Goñi et al. 2005). Guo et Macdonald (2006)
suggéraient récemment que la dégradation du pergélisol pourrait entraîner la mobilisation de
carbone « ancien », principalement sous forme particulaire. De plus, la modification de la
toundra arctique vers un écosystème de type « forêt boréale » sous un climat plus chaud
(Hinzman et al., 2005), risque d’augmenter les apports en carbone « moderne ». Se dernier est
facilement dégradable en milieu côtier (Goñi et al. 2005). Par conséquent, le réchauffement
global risque de modifier sensiblement le cycle du carbone particulaire sur le Plateau du
Mackenzie en modifiant la quantité et la qualité des apports allochtones.
Contrairement au carbone particulaire, le cycle du carbone dissous est pratiquement
inconnu sur le Plateau du Mackenzie. D’après Telang et al. (1991), le fleuve Mackenzie
transporte 1.3 x 10 12 g de C organique dissous, soit le quatrième plus important à l’échelle de
l’Arctique. Le destin de ce carbone terrigène reste pratiquement inconnu, mais si l’on se fit
au mélange conservatif du DOC observé ailleurs en Arctique (Cauwet et Sidorov 1996 ;
Köhler et al., 2003), la majeur partie serait exportée vers l’océan Arctique ou l’Archipel
Canadien. D’après Gosselin et al. (1997), la production de DOC par les algues de glace
Arctique représente entre 20 et 65 % de la production primaire totale. La production
primaire marine est donc certainement une source significative de DOC sur le Plateau du
Mackenzie. Cependant, il semble que ce carbone soit rapidement utilisé par les organismes
hétérotrophes (Davis et Benner, 2005).
42

Chapitre III: Photominéralisation de la matière
organique dissoute d’origine terrigène dans les
eaux côtières arctiques entre 1979 et 2003:
Variabilité interannuelle et implications des
changements climatiques
43

III.A Résumé
La minéralisation photochimique (photominéralisation) de la matière organique dissoute
d’origine terrigène (tDOM) dans l’Océan Arctique est limitée par la présence d’un couvert de
glace permanant qui réduit la quantité d’énergie ultraviolette (UV) atteignant la colonne d’eau
sous-jacente. En réponse à la diminution de l’étendue et de l’épaisseur du couvert de glace
causée par le réchauffement global, les processus biogéochimiques qui dépendent de
l’énergie UV devraient s’amplifier dans cette région. Dans cette étude, nous avons réalisé
pour la première fois des estimations quantitatives de la photominéralisation de la tDOM
dans un écosystème côtier Arctique sous des conditions de glace actuelles et futures. Nous
avons employé un modèle couplé optique-photochimique utilisant en entrée les propriétés
optiques inhérentes de la colonne d’eau et des mesures de production photochimique
(photoproduction) de carbone inorganique dissous (DIC), le principal produit de la
photooxidation de la DOM. Les rendements quantiques apparents de photoproduction de
DIC ont été déterminés sur des échantillons provenant de l’estuaire du Fleuve Mackenzie, du
Plateau du Mackenzie, et du Golfe d’Amundsen. Grâce à un modèle de transfert radiatif
atmosphérique, l’éclairement UV journalier incident juste sous l’interface air-mer a été estimé
en combinant des mesures 1) de luminance rétrodiffusée par l’atmosphère et la surface et 2)
de micro-ondes passives émises par la surface. La photoproduction annuelle de DIC dans les
eaux de surface du sud-est de la Mer de Beaufort où l’absorption des UV est dominée par la
tDOM colorée déchargée par le Fleuve Mackenzie, est, en moyenne entre 1979 et 2003,
estimé à 66.5 ± 18.5 Gg de carbone. Cette valeur est équivalente à environ 10% du taux de le
respiration bactérienne dans les eaux de surface, 8% du taux de production primaire, et 2.8 ±
0.6% des 1.3 Tg de carbone organique dissous (DOC) déchargé annuellement par le Fleuve
Mackenzie dans cette région. Durant les périodes où le couvert de glace est fortement réduit
(comme en 1998), cette dernière estimation peut atteindre 5.1% de la masse du DOC
d’origine terrigène déchargé annuellement par le Fleuve. En simulant un été où la région
serait pratiquement libre de glace durant la période estivale, le modèle prédit une
photoproduction annuelle de 150.5 Gg de DIC, minéralisant 6.2% des apports annuels en
DOC par le Fleuve Mackenzie. Ces résultats indiquent que si les prédictions des modèles
climatiques concernant la réduction du couvert de la Arctique sont correctes, la
photominéralisation de la tDOM dans les eaux de surface Arctiques s’accélérera grandement
dans le futur.
44

III.B Article publié dans la revue Global Biogeochemical Cycles :
“Photomineralization of terrigenous dissolved organic matter in Arctic coastal
waters from 1979 to 2003: Interannual variability and implications of climate
change”
Simon Bélanger 1 , Huixiang Xie 2 , Nickolay Krotkov 3 , Pierre Larouche 4 , Warwick F.
Vincent 5 and Marcel Babin 1
1 Université Pierre et Marie Curie-Paris 6, Laboratoire d'océanographie de Villefranche,
06230 Villefranche-sur-Mer, France ; CNRS, Laboratoire d'océanographie de
Villefranche, 06230 Villefranche-sur-Mer, France
2 Institut des Sciences de la Mer de Rimouski, Université du Québec à Rimouski,
Rimouski, Canada
3 Goddard Earth Sciences and Technology Center, University of Maryland Baltimore
County, USA
4 Institut Maurice-Lamontagne, Pêches et Océans Canada, Mont-Joli, QC, Canada G5H
3Z4
5 Département de Biologie & Centre d’Études Nordiques, Université Laval, Québec City,
Canada
45

Acknowledgments
This study was made possible with financial support from the Natural Sciences and Engineering
Research Council of Canada (NSERC), the Fonds Québécois pour la Recherche sur la Nature et
les Technologies (FQRNT) (to HX), and the Fonds France Canada pour la Recherche (FFCR to
MB and WFV) and Indian and Northern Affairs Canada (to SB). SB receives a doctoral
fellowship from the FQRNT. We thank Dr S.H. Lee for providing the primary production rates
measured in the Gulf of Amundsen. We are grateful to Dr. C. Osburn and Dr. M. Tzortziou for
their constructive comments on the manuscript. We appreciate the efforts of S. Cizmeli in CASES
field measurements, T. Lou in DIC measurement and J. Caveen in Matlab programming. We
thank the crews of the CCGS Amundsen cruises for their enthusiastic help onboard the ship. This
is a contribution to the Canadian Arctic Shelf Exchange Study (CASES) under the overall
direction of L. Fortier. S. C. Johannessen and one anonymous referee are acknowledged for
constructive comments on the manuscript.
46

Abstract
Photomineralization of terrigenous dissolved organic matter (tDOM) in the Arctic Ocean
is limited by persistent sea ice cover that reduces the amount of ultraviolet (UV) radiation
reaching the underlying water column. UV-dependent processes are likely to accelerate
as a result of shrinking sea ice extent and decreasing ice thickness caused by climatic
warming over this region. In this study, we made the first quantitative estimates of
photomineralization of tDOM in a coastal Arctic ecosystem under current and future sea
ice regimes. We used an optical-photochemical coupled model incorporating water
column optics and experimental measurements of photoproduction of dissolved inorganic
carbon (DIC), the main carbon product of DOM photochemistry. Apparent quantum
yields of DIC photoproduction were determined on water samples from the Mackenzie
River estuary, the Mackenzie Shelf, and Amundsen Gulf. UV irradiances just below the
sea surface were estimated by combining satellite backscattered and passive microwave
radiance measurements with a radiative transfer model. The mean annual DIC
photoproduction between 1979 and 2003 was estimated as 66.5 ± 18.5 Gg carbon in the
surface waters of the southeastern Beaufort Sea, where UV absorption is dominated by
chromophoric dissolved organic matter discharged by the Mackenzie River. This value is
equivalent to 10% of bacterial respiration rates, 8% of primary production rates and 2.8 ±
0.6% of the 1.3 Tg of dissolved organic carbon (DOC) discharged annually by the
Mackenzie River into the area. During periods of reduced ice cover such as 1998, the
latter value could rise to 5.1% of the annual riverine DOC discharge. Under an ice-free
scenario, the model predicted that 150.5 Gg of DIC would be photochemically produced,
mineralizing 6.2% of the DOC input from the Mackenzie River. These results show that
the predicted trend of ongoing contraction of sea ice cover will greatly accelerate the
photomineralization of tDOM in Arctic surface waters.
47

III.1. Introduction
Terrigenous dissolved organic matter (tDOM) is an abundant component of Arctic
polar surface waters [see Benner et al., 2005 and references therein] as a result of the
large amounts of riverine tDOM that enter the Arctic Ocean relative to its volume
[Aagaard and Carmack, 1989]. The refractory character of tDOM, as observed in several
Arctic marginal seas [e.g., Dittmar and Kattner, 2003; Meon and Amon, 2004], may also
contribute to the high concentrations of Arctic Ocean tDOM. Recent studies suggest that
slightly more than 50% of the annual input of terrigenous dissolved organic carbon
(tDOC) to the Arctic Ocean is remineralized before its transport to the North Atlantic
[Hansell et al., 2004]. However, quantitative information on major tDOM loss processes
in the Arctic is still lacking.
There is growing evidence indicating a major role of photooxidation in the
remineralization of tDOM in the coastal oceanic environment [e.g., Benner and Opsahl,
2001; Hernes and Benner, 2003; Kieber et al., 1990]. Other studies suggested that a large
fraction of the marine plankton-derived dissolved organic carbon (DOC) pool also can be
photochemically mineralized in the open ocean [Johannessen, 2000; Johannessen and
Miller, 2001]. Although there is molecular evidence that photochemical transformation of
tDOM occurs in the Arctic Ocean [Benner et al., 2005], this mechanism is not considered
of importance because of the prevailing low solar angle, cloudy conditions and
permanent sea ice cover [Amon and Meon, 2004; Benner et al., 2004]. Sea ice and its
overlying snow cover is known to be a strong attenuator of ultraviolet (UV) radiation in
both polar regions [Vincent and Belzile, 2003].
Recent observations in the Arctic reveal that sea ice conditions are changing, with
a 20% decrease in the summer areal extent of sea ice over the last three decades [e.g.,
Cavalieri et al., 2003]. In addition, the thinning of sea ice has also been observed in
several parts of the Arctic Ocean [e.g., Wadhams and Davis, 2000], permitting more UV
radiation to reach the water column. Moreover, a complete loss of summer ice is
predicted for later this century [ACIA, 2005]. In parallel, a decrease in stratospheric ozone
concentration (O3), which results in an increase of UVB (280-320 nm) radiation at sea
surface, was recently reported over the high northern latitudes [Fioletov et al., 2004; Rex
48

et al., 2004]. These observations suggest that the amount of photochemically active UV
radiation (280-400 nm) penetrating into the Arctic waters could have significantly
increased in the last decades, and that this effect may accelerate in the coming years.
In addition to alteration in sea ice dynamics, a number of other recent
observations suggest that the Arctic is currently experiencing rapid and extensive
environmental changes in response to global climate warming [Overland et al., 2004;
Serreze et al., 2000]. Arctic rivers account for about 10% of global terrestrial organic
carbon export to the ocean [Rachold et al., 2004]. This percentage is likely to increase
markedly during the next century [Frey and Smith, 2005] as a result of increasing river
discharge [Peterson et al., 2002] and thawing of the permafrost [Camill, 2005]. The latter
stores one third of the global soil carbon and contains 7 to 26% of all terrestrial organic
carbon stored since the last glacial maximum [Smith et al., 2004]. To assess and predict
the impact of climate change on the marine carbon cycle, it is critical to determine the
transformation and fate of the exported terrestrial carbon, including photochemical
production of dissolved inorganic carbon (DIC), which is the main carbon product of
chromophoric DOM (CDOM) photochemistry [Miller and Zepp, 1995; Mopper and
Kieber, 2002].
The objectives of this study are to assess the present significance of
photooxidation of CDOM in the tDOM cycling in ice-free Arctic seawaters, and to assess
the potential impact of the reduction of sea ice extent on the removal of tDOM through
this process. Our field study was conducted in the Canadian Shelf of the Beaufort Sea
where a large amount of tDOM is discharged by the Mackenzie River (Figure III.1). In
terms of dissolved organic carbon export, the Mackenzie River is the fourth largest
among the Arctic rivers, discharging ~1.3 Tg of tDOC per year [Telang et al., 1991]. This
large tDOC flux explains the high concentration of terrestrial CDOM found in the entire
Western Arctic Ocean [Guay et al., 1999; Guéguen et al., 2005]. As this area has
experienced a significant decrease in sea ice areal extent between 1979 and 2000 [Barber
and Hanesiak, 2004], photooxidation may have become a significant component of the
tDOM cycling in this Arctic coastal ecosystem. In this study, experimentally determined
wavelength-specific apparent quantum yields (AQY) of DIC production, along with
remote sensing-derived spectral irradiances (300 to 600 nm) and surface areas of open
49

waters, were used to model the depth-integrated photochemical production rates of DIC
in the ice-free waters of the southeastern Beaufort Sea . To assess recent trends in the
DIC production, we calculated the annual photochemical production of DIC for the
period 1979 to 2003.
Figure III.1. Location of CASES stations sampled for absorption
measurements in 2004 over the Mackenzie Shelf (open circles), in
Amundsen Gulf (open triangles), and in Canada Basin (open squares).
ARDEX and CASES samples collected for φDIC determination are shown
as closed squares.
50

III.2. Materials and methods
III.2.1. Sample collection and storage
Sampling was conducted in June and July 2004 onboard the CCGS Amundsen in
the Mackenzie Shelf, the Gulf of Amundsen, and the Arctic Canada basin (Fig. III.1) as
part as the Canadian Arctic Shelf Exchange Study (CASES) program, and onboard the
CCGS Nahidik in the Mackenzie River estuary in late July during the Arctic River-Delta
Experiment (ARDEX). Surface water samples were collected at 79 locations during
CASES (Fig. III.1) for measurement of the absorption coefficients of CDOM (aCDOM) and
suspended particles (ap). The samples were filtered through 0.2-μm Anotop® syringe
filters (Whatman) and collected into 100-mL acid-cleaned amber glass bottles.
Suspended particles were retained by 25-mm GF/F glass fiber filters (Whatman) by
filtering 0.1 to 3.5 L of seawater. The glass bottles and GF/F filters were stored frozen
(seawater: -20 °C; particle: -80 °C) in the dark until being analyzed ca. four months later
in the land-based laboratory at Rimouski (aCDOM) and Villefranche-sur-mer (ap).
Six stations were sampled for determination of the apparent quantum yield (AQY)
spectra of DIC photoproduction: two in the Mackenzie River estuary, one on the
Mackenzie Shelf (ARDEX cruise), and three in the Amundsen Gulf (CASES cruise) (Fig.
III.1; Table III.1). Surface water (∼5 m deep) in the Amundsen Gulf was taken with
Teflon-coated Niskin bottles and gravity-filtered through a Pall AcroPak 1000 Capsule
sequentially containing 0.8-μm and 0.2-μm polyethersulfone membrane filters. During
the ARDEX cruise, surface water was collected using an acid-cleaned plastic bucket and
transferred to acid-cleaned carboys. Both the bucket and carboys were profusely rinsed
three times with sample water. The surface water was then filtered through sequential 3-
μm and 0.2-μm pore-size polysulfone filters (PALL corporation). All filtered samples
were stored cold (4 °C) in the dark in acid-cleaned 4-L clear glass bottles and shipped to
the laboratory at Rimouski.
51

Table III.1. Chemical and physical properties of samples used for the φDIC.
Station- Date of Lat Lon Depth S T
Mission sampling (°N) (°W) (m)
(°C)
R5d-ARDEX 31-07 69.28 133.97 0 1.57 16.1
R5a-ARDEX 31-07 69.42 133.5 0 8.19 12.2
R9-ARDEX 28-07 70.05 133.42 0 25.81 9.23
108-CASES 07-06 70.63 123.22 10 30.09 -1.08
406-CASES 16-06 71.30 127.75 5 30.77 0.51
409-CASES 23-07 71.46 127.3 5 28.79 3.46
III.2.2. Optical measurements
The frozen seawater samples were thawed and warmed to room temperature.
Their optical densities (OD), referenced to pure water, were measured in a 10-cm quartz
cuvette between 250 and 800 nm with 1-nm increments using a Perkin-Elmer Lambda 35
dual beam spectrophotometer. A background correction was applied by subtracting the
absorbance value averaged over an interval of 5 nm around 685 nm from all the spectral
values [Babin et al., 2003]. The spectral absorption coefficient of CDOM, aCDOM(λ) (m -1 ),
was calculated as :
a CDOM
2.
303 OD(
λ)
( λ ) = (1)
l
where l is the optical pathlength (0.1 m). Note that aCDOM(λ) spectra of frozen samples
were lower in magnitude (~10%) than the refrigerated samples used for irradiation
experiments.
The spectral absorption coefficients of the particles retained on the GF/F filters,
ap(λ) (m -1 ), were determined at 1-nm resolution between 350 and 750 nm according to
the transmittance-reflectance technique described by Tassan and Ferrari [1995] using a
Perkin-Elmer Lamda-19 spectrophotometer equipped with a 60-mm integrating sphere.
Before the measurements of the OD, the frozen filters were hydrated with 1-2 ml of
filtered seawater and warmed to room temperature to avoid phytoplankton cells lyses.
The OD of particles on filter, obtained by subtracting the mean OD of 10 hydrated blank
GF/F filters at each wavelength from the total measured OD with the reference beam in
air, was converted to ap using the expressions given by Tassan and Ferrari [2002].
Assuming that particle absorption is negligible in the near-IR [Babin and Stramski, 2002],
ap(750) was subtracted from ap at all wavelengths < 750 nm.
52

III.2.3. Irradiation experiments
The irradiation setup and procedure for determining the AQY of DIC production,
φDIC (mol C (mol photons) -1 ), were similar to those adopted by Johannessen and Miller
[2001]. Briefly, stored samples were refiltered through 0.2-μm polycarbonate membrane
filters (Millipore), acidified with HCl to pH ~3 and purged with CO2-free air for ~24 h to
reduce the background DIC to undetectable levels. The purging gas was humidified
(100% relative humidity) so that no net loss of water occurred due to evaporation. The
DIC-free water samples were buffered back to approximately the original pH values with
powdered sodium borate (ACS grade) and transferred into pre-combusted (420 °C), gas-
tight quartz-windowed cylindrical cells (internal diameter: 0.02 m; length: 0.14 m) and
irradiated in a thermostated incubator (0 ± 0.5 °C) using a SUNTEST CPS solar
simulator equipped with a 1-kW Xe lamp. Eight spectral treatments were examined
employing successive Schott long-pass glass filters; the Schott numbers of these filters
are WG280, WG295, WG305, WG320, WG345, GG395, GG435 and GG495. Spectral
irradiance under each filter was measured with an Optronics OL-754 UV-vis
spectroradiometer fitted with a fibreoptic cable. The irradiance level under the WG280
filter was set to 190 W m -2 for stations R5a, R5d and R9, and to 585 W m -2 for stations
108, 406 and 409, and the irradiation time varied from 24.5 to 46 h, depending on the
abundance of CDOM in the irradiated samples (Fig 1 in Supplementary Materials). The
rationale for choosing these irradiation times and light intensities was to obtain
measurable DIC while minimizing the extent of photobleaching. Parallel incubations in
the dark were conducted to correct for potential nonphotochemical production of DIC.
Extreme caution was taken to minimize bacterial contamination during sample handling
and transfer. Small increases of DIC in the dark controls were often observed, but they
were always similar to or less than the amounts of DIC formed in the quartz cell under
the GG495 filter. The dark control values were subtracted during the calculation of DIC
photoproduction in each irradiated vessel.
53

III.2.4. DIC measurement and calculation of φDIC
After irradiation, 1.50 mL of sample (in triplicates) were mixed with 1 mL of 10%
DIC-free H3PO4 in an acid trap and purged with ultra-high-purity nitrogen. DIC in the
sample was converted into CO2 which was introduced by a nitrogen carrier stream to a
LI-COR 6262 CO2/H2O analyzer for quantification. The whole system was calibrated
immediately before and after sample measurement with varying volumes (usually 0.20,
0.40, 0.60, and 0.8 mL) of a 12-μM DIC standard. This working standard was daily
prepared by diluting a 600-μM sodium carbonate stock solution; the stock solution was
made by dissolving dried sodium carbonate (Na2CO3, BDH AnalaR) in DIC-free pure
water. All calibration curves resulted in R 2 > 0.999. The analytical precision was
determined to be ∼2 % at a concentration of 4 μM DIC.
The spectral apparent quantum yield of DIC production, φDIC(λ), was defined as
the moles of DIC photochemically produced per mole of photons absorbed by CDOM at
wavelength λ. To derive the φDIC(λ) values, the iterative curve-fitting method of
Johannessen and Miller [2001] was applied. Briefly, this method assumes an appropriate
mathematical functional form with unknown parameters to express the change of φDIC as
a function of wavelength. Decreasing exponential functional forms are usually chosen for
photochemical quantum yield spectra. The amount of DIC produced in an irradiation cell
over the exposure time can then be predicted as a product of three terms: the assumed
φDIC(λ) functional form, the measured incident irradiance, and the sample absorption
coefficient. The optimum values of the unknown parameters in the assumed φDIC(λ)
functional form are obtained by varying these parameters from initial estimates until
minimum difference between the measured and predicted DIC production is achieved.
The absorption coefficient was measured before and after irradiation and was averaged
according to first-order kinetic decay, which well describes photobleaching [Del Vecchio
and Blough, 2002; Xie et al., 2004]. We followed the recommendation of Hu et al. [2002]
for the calculation of the number of photons absorbed by CDOM and adopted the
following modified exponential form to fit the data:
where k1, k2 and k3 are fitting parameters.
k2
/ ( λ + k3)
φ DIC ( λ)
= k1
e
(2)
54

III.2.5. Modeling DIC photoproduction
Assuming that the seawater optical properties are vertically constant in the photic
zone and that the flux of photons backscattered to the atmosphere is negligible, the daily
depth-integrated DIC photoproduction rate, PDIC (mol C m -2 d -1 ), can be calculated as:
a ( λ)
DIC λ ( λ)
dλ
(3)
P
600
−
= ∫ Ed
( 0 ,
λ=
300
)
CDOM
at
( λ)
φDIC
where at(λ) (m -1 ) is the total absorption coefficient (i.e., sum of absorption by seawater,
−
CDOM and particles) and ( 0 , λ)
E d (mol photons m -2 d -1 ) is the spectral daily
downward irradiance just beneath sea surface. Eq. 3 was used to calculate the temporal
changes in PDIC in different regions of the Beaufort Sea over the period between 1979 and
2003. Constant regional values of aCDOM(λ)/at(λ) and φDIC(λ) determined from our field
and laboratory work were assumed (see below). Note that in the open ocean where the
fraction of light absorbed by the CDOM (i.e., the ratio aCDOM(λ)/at(λ)) approaches the
value of 1 in the UV domain, PDIC can be approximated by integrating the product
−
of E ( 0 , λ)
and φDIC(λ) over λ [e.g. Johannessen, 2000]. When particles compete for
d
light absorption in the UV domain, as in coastal waters, the parameter aCDOM(λ)/at(λ) of
Eq. 3 is necessary to avoid overestimation of PDIC.
By using constant regional value for φDIC(λ), our approach did not explicitly take
into account any potential dependence on light dose of this quantity; i.e., the effect of
CDOM light history on φDIC(λ). Johannessen and Miller [2001] observed a 2.3 times
increase in the φDIC(λ) after a coastal water sample was extensively photobleached (loss
of absorbance in the UV wavelengths > 50%) while Vähätalo and Wetzel [2004] found a
13% decrease in the DOC loss AQY of a lake water sample with a 9% increase in the
photons absorbed by CDOM. Therefore, both the direction and magnitude of change in
photomineralization AQY in response to light dose remain unclear. In the present study,
significant photobleaching was incurred during the AQY experiments (Fig. III1 in
Supplementary Materials) while no appreciable photobleaching seemed to have taken
place during the transport of CDOM from the estuary of the Mackenzie River to the Shelf
or even further seaward (see Results and Discussion). Ideally, experimentally determined
55

AQYs only apply to timescales over which the amount and spectral quality of the light
absorbed by CDOM in the field are equivalent to those of the light absorbed by CDOM in
irradiation experiments. On larger geochemically significant timescales (e.g., the
residence times of the Mackenzie River’s runoff in the Canada Basin or the entire Arctic
Ocean), the uncertainties in the calculation of PDIC associated with the dose dependence
of φDIC(λ) should be smaller, since the extent of photobleaching in the field on these
timescales, as compared to the timescales of land-to-sea CDOM transport, are likely
closer to that occurring in our laboratory irradiations.
The spectral daily downward irradiance received at the sea surface was calculated
using a radiative transfer method that uses as inputs satellite measurements of total
atmospheric ozone column content and of cloud cover derived from Total Ozone
Mapping Spectrometer (TOMS) data [Herman et al., 1999; Krotkov et al., 2002; Krotkov
et al., 1998; Krotkov et al., 2001]. The calculations were made for each day between
February 1979 to April 1993, and between August 1996 to November 2001, which
correspond to the periods when TOMS data were available. Calculations were done for
conditions with and without cloud cover. The method was modified in two respects. First,
instead of using the monthly minimum Lambertian equivalent reflectivity (MLER) for
the surface albedo (As) from climatology, we used a linear combination of albedos for
open water and sea ice:
As water
ice
= A ( 1−
SIC)
+ A SIC
(4)
where SIC is the fractional sea ice surface concentration (dimensionless, from 0 to 1), and
Awater and Aice are the water and sea ice albedos, respectively. Note that As must be
accounted for in the calculation of E (λ)
because multiple scattering between sea
d
surface and atmosphere can be significant in the UV domain. Daily SIC data were
obtained from the National Snow and Ice Data Center (NSIDC) and were derived from
observations of the Scanning Multi channel Microwave Radiometer (SMMR; 1979-1987)
or Special Sensor Microwave Imager (SSM/I; 1987-2003) using the NASA Team
algorithm [Cavalieri et al., 1990]. The values for Awater (0.05) and Aice (0.76) were
adopted from the MLER climatology for summer (SIC = 0) and winter (SIC = 1) seasons,
respectively, published by Herman and Celarier [1997] and Herman et al. . Second, the
calculations of irradiance were spectrally extended to cover the range from 300 to 600 nm
56

(with 0.5 nm resolution). As the above procedure yields downward irradiance just above
sea surface, we applied a factor of 0.95 to account for specular reflection and to
−
obtain E ( 0 , λ)
just beneath surface as needed in Eq. 3.
d
For the calculation of total DIC photoproduction, the Southeastern Beaufort Sea
was divided into three sub-regions: the Mackenzie Shelf (MS; 80 000 km 2 ; depth < 200m;
west of Cape Bathurst), the Amundsen Gulf (AG; 87 000 km 2 ; >118°W) and the Canada
Basin (CB; 240 000 km 2 ;

Figure III.2. φDIC spectra determined on water samples collected during
ARDEX (R5a, R5d, and R9; solid thin lines) and CASES (108, 406 and
409; dashed thin lines). The grey curves are the average φDIC spectra
published by Vähätalo et al. [2000] for a boreal lake, and by Johannessen
and Miller [2001] for inshore, coastal and offshore waters.
Table III.2. CDOM properties and model parameters for φDIC.
φ ( λ)
= k e
Station aCDOM330 SCDOM a DIC
1
k2
/ ( λ + k3
)
(m -1 ) k1 k2 k3
2 b
r φ DIC (x10 -6 )
R5d 5.44 0.0190 1.2x10 -6 726. -176. .998 20.3
R5a 4.50 0.0201 4.9x10 -6 352. -224. 1.00 25.8
R9 1.50 0.0204 9.6x10 -6 128. -257. .984 18.9
108 0.72 0.0200 2.7x10 -10 5232. 88.7 .976 8.8
406 0.69 0.0191 6.7x10 -8 1538. -111. .996 6.6
409 0.62 0.0217 7.44x10 -6 128. -256. .994 14.6
a
Spectral slope of the the model aCDOM(λ)=aCDOM(λ0) Scdom*(λ0-λ) .
b 2
The correlation coefficient (r ) is from the linear regression between modeled and measured
DIC production.
58

III.3. Results and discussion
III.3.1. Apparent Quantum Yield for DIC photoproduction
The six φDIC spectra determined during this study are shown in Fig. III.2 and the
parameters of the quasi-exponential model (Eq. 2) are given in Table III.2. The
exponential form adopted here has a steeper slope and fits better the measured production
(r 2 > 0.97) than the two-parameter single exponential form used by Johannessen and
Miller and Vähätalo et al. . A detailed comparison between these two functional forms is
provided in the supplementary online material. To compare the photoreactivity of the
CDOM, we calculated the mean φDIC value weighted by an incident solar irradiance
spectrum:
600
∫
E
( 0
=
280
d
600
∫
280
59
−
, λ)
φ
( λ)
dλ
φ DIC . (5)
E
−
( 0 , λ)
dλ
Here Ed (0 - , λ) was modeled using the Tropospheric Ultraviolet Visible (TUV) model
[Madronich and Flocke, 1999] for summer solstice at latitude of 70°N with clear sky,
moderate aerosols concentration (optical thickness at 550 nm of 0.1), and total ozone
column content of 330 DU. Note that only the spectral shape of Ed (0 - , λ) influenced the
value of φ DIC (e.g., for a total ozone column content of 480 DU, the difference in φDIC
35) waters with 174.0 and 786.5 × 10 -6 mol C
(mol photons) -1 , respectively. However, our φDIC
DIC
values for low-salinity water samples
(R5a and R5d, salinity < 9) are comparable to the inshore (S < 31) value of 30.4 × 10 -6
mol C (mol photons) -1 measured by Johannessen and Miller [2001] and to the average
is

φDIC
of 25.9 x 10-6 mol C (mol photons) -1 reported by Vähätalo et al. [2000] for water
samples from Valkea-Kotinen, a humic lake located in coniferous forest of Finland
(61°14′N; 25°04′E). While CDOM was mixed conservatively over the shelf estuary, as
suggested by the negative linear correlation between aCDOM and salinity (Fig. III.3a), we
observed a nonlinear decrease of φ DIC with increasing salinity (Fig. III3b). This behavior
possibly resulted from a conservative mixing of high-φDIC inshore waters with low-φDIC
offshore waters and a non-conservative changes in the photoreactivity of CDOM across
the salinity gradient. The decrease of φ DIC with salinity observed in this study is opposite
to the trend observed by Johannessen and Miller [2001] who proposed that, as the
terrestrial aromatic structures in terrestrial CDOM are eliminated by photochemical
fading or as marine-derived CDOM increasingly dominates the light absorption, the
proportion of DIC-producing to non-DIC-producing chromophores in CDOM increases.
In the present study, photochemical fading of terrestrial CDOM is not supported by the
conservative behavior of aCDOM(λ) (Fig. III.3a). In contrast, the lower φ DIC values were
observed at stations 108 and 406 (S = 30-31) sampled in June and away from direct input
of the Mackenzie River. Overall, Fig III.3b suggests that the photoreactivity of tDOM
drained into the Mackenzie River changed when it was mixed into coastal and offshore
waters (S

Figure III.3. (a) Variation of aCDOM(330) with salinity for water samples
collected in Mackenzie Shelf (open circles), Amundsen Gulf (open
triangles), Canada Basin (open squares). Samples collected for φDIC are
shown as closed squares. (b) Variation of weighted quantum yield
normalized to the integrated irradiance, φ DIC , with salinity.
61

III.3.2. Role of DIC photoproduction in the organic carbon cycling
The yearly pattern of DIC photoproduction (PDIC), averaged over the period
corresponding to TOMS data availability (01/1979 to 04/1993; 08/1996 to11/2001), is
shown in Fig. III.4. At summer solstice, PDIC reached 6.1, 2.9 and 4.1 mg C m -2 d -1 for
MS, AG and CB, respectively. Given that the spectral irradiance was similar for each
sub-region, the difference in PDIC was mainly due to regional differences in φDIC (Fig.
III.2) and in the fraction of light absorbed by CDOM (Fig. III.5). While CDOM largely
dominated the total light absorption at wavelengths 90%), the
high concentration of particulate matter in the MS reduced light availability to CDOM by
10 to 20% compared with the other two sub-regions. Nevertheless, PDIC was larger in the
MS due to the higher φDIC in the intermediate salinity range (Fig. III.3b).
Figure III.4. Average DIC photoproduction rates, PDIC, calculated using
the φDIC representing each sub-region: R5a, R5d and R9 for the Mackenzie
Shelf; 108 for the Amundsen Gulf; 406 and 409 for the Canada Basin.
62

Figure III.5. Average spectral light fraction absorbed by CDOM in the
surface waters at stations (Fig. III.1) sampled over the Mackenzie Shelf
(dashed line), in Amundsen Gulf (dash-dotted line), and in Canada Basin
(thin line).
To assess the importance of photomineralization relative to other biogeochemical
processes affecting the marine carbon cycle, our DIC photoproduction rates were
compared with published bacterial respiration and primary production data for the same
areas. Based on leucine uptake rates, Garneau et al. [2006] reported low bacterial
production rates of 12 to 112 μg C m -3 d -1 (mean: 48 μg C m -3 d -1 ) in the fall of 2002 in
the vicinity of the Mackenzie River estuary. Assuming a typical growth efficiency of
27% for coastal Arctic bacteria [Meon and Amon, 2004], the bacterial respiration rate
would be ~3.2 mg C m -2 d -1 in the upper 25 m of the water column. For comparison, PDIC
was 0.6 mg C m -2 d -1 at that time of the year (Fig. III.4), equivalent to ~19% of the
bacterial respiration. In late June and early July 2004, when primary productivity and
temperature are higher, the bacterial respiration rate in the upper 25 m ranged from 20.2
to 77.8 mg C m -2 d -1 (mean: 49.0 mg C m -2 d -1 ) on the shelf [M.-È. Garneau, unpub data].
During the same period, PDIC in the MS was 5.1 mg C m -2 d -1 (Fig. III.4; Table III.3),
equivalent to 10.4% of the bacterial respiration.
Wavelength (nm)
63

Table III.3. Comparison between daily DIC photoproduction, and total and new primary
production rates in open water (in mg C m -2 d -1 ).
Sub-region Period Primary Production PDIC
Mackenzie Shelf
(80 000 km 2 )
Canada Basin
(240 000 km 2 )
Late July
Total New
67-134 (1) 3.5-5.0
100-200 (1)
Late August 79-145 (106) (2)
64
14-26 (19) (2) 1.3-1.8
Amundsen Gulf
(87 000 km 2 )
Late August
13.2 (2) 1.0
(1) From Carmack et al. [2004].
(2) From Lee and Whitledge [2005], personal communication of S. Lee for AG values.
91.9 (2)
Table III.3 compares our PDIC values with recent published values of primary
production (PP) and new production (sensus Dugdale and Goering [1967]) for open
waters in the same areas. PP in the MS is typically 100 to 200 mg C m -2 d -1 in late July
while new production (based on nitrate drawdown) is 67 to 134 mg C m -2 d -1 [Carmack et
al., 2004]. Values reported for late August in the Canada Basin range from 79 to 145 mg
C m -2 d -1 for PP, with a mean of 106 mg C m -2 d -1 , and from 14 to 26 mg C m -2 d -1 for
new production (based on nitrate and ammonium uptake), with a mean of 19 mg C m -2 d -1
[Lee and Whitledge, 2005]. In the AG, primary and new productions in late August 2002
were 91.9 and 13.2 mg C m -2 d -1 , respectively [S. H. Lee, personal communication].
Based on our estimates of PDIC, photomineralization of DOC occurs at a rate equivalent
to 2.6-7.8, 6.8-9.4 and 7.6% of the amount of organic carbon produced by new
production in MS, CB, and AG, respectively. Note that for the small area near the Cape
Bathurst, however, satellite-based PP rates can reach up to 2800 mg C m -2 d -1 during the
spring phytoplankton bloom [Arrigo and van Dijken, 2004]. Therefore the relative
contribution of photomineralization to the organic carbon cycling could be much lower
during short periods of bloom conditions.

III.3.3. Interannual variability in DIC photoproduction
Figure III.6 shows the annual photoproduction of DIC from 1979 to 2003. The
mean annual production (± s. d.) was 24.9 ± 5.2, 13.7 ± 3.3 and 27.8 ± 11.8 Gg C in the
MS, AG and CB, respectively. When all regions are pooled together, the photoproduction
was on average 66.5 ± 18.4 Gg C y -1 . Because of the large interannual variability, the
positive slopes of the linear regressions between the production rate and time are not
significantly different from zero. Nevertheless, the positive trends observed over the 25
years are higher for the AG and CB (+ 1.5 and + 2.1 Gg C decade -1 , respectively) than for
the MS (+ 0.5 Gg C decade -1 ). Based on these trends, the DIC photoproduction increased
by 5.0%, 26.3%, and 18.1% between 1979 and 2003 for the MS, AG, and CB,
respectively, with an overall increase of 14.8%.
Figure III.6. Trends in the annual DIC photoproduction for period 1979-
2003 for the Mackenzie Shelf, the Amundsen Gulf, the Canada Basin, and
the sums of the three sub-regions.
65

Figure III.7. Correlation between the annual DIC photoproduction and
the annual average area of open water for: (a) Mackenzie Shelf, DIC =
0.91x10 6 * X; (b) Amundsen Gulf, DIC = 0.52x10 6 * X ; (c) Canada Basin,
DIC = 0.63x10 6 * X. (X = open water area in km 2 ; slope in Gg C y -1 km -2 ;
n=25).
The large interannual variability of the DIC photoproduction is mainly due to
variability in sea ice cover. Figure III.7 shows the strong positive correlation between the
annual photoproduction of DIC and the average area of open water for each sub-region.
Note that the timing of the opening of sea ice is also important due to the strong seasonal
variability of PDIC (Fig. III.4). For instance, the presence or absence of an ice cover in
June has a stronger impact on the annual DIC photoproduction than it does in September.
66

Figure III.8 shows the changes in the number of days with open water (defined as >50 %
of the area free of ice) and in the date when the opening begins (i.e., first appearance of
open water) for each sub-region. The number of days with open water was on average (±
s. d.) 116.8 ± 25.7, 125.7 ± 26.9 and 52.6 ± 47.6, and increased at a rate of 3.3, 12.0, and
9.3 days per decade for MS, AG and CB, respectively. Concurrently, the opening
occurred on average at day 170.9 ± 20.3, 167.1 ± 18.9 and 181.6 ± 74.3, and decreased at
rates of 1.9, 14.1 and 6.8 days per decade, for MS, AG and CB, respectively. As expected,
we observed a larger interannual variability in sea ice concentration in the Canada Basin,
which was due in part to the movement of the Arctic central packed ice associated with
dominant anticyclonic or cyclonic circulation regime [e.g., Comiso et al., 2003]. For
instance, the open-water area in the CB remained 120 days of fully open water were observed in 1987,
1993 and 1998.
Figure III.8. Trends in open water area for the three sub-regions as
observed by SMMR/SSMI between 1979 and 2003. The left panels show
the day of the year when >50% of the surface area is ice free. The right
panels show the number of days having open water.
67

In addition to the variability in sea-ice cover and in the timing of its opening, the
interannual variation of the atmospheric ozone concentration and cloud cover also
contributed to the variability of DIC photoproduction. Figure III.9 gives the trends
between 1979 and 2001 in monthly averaged ozone column content as observed by
TOMS over the southeast Beaufort Sea. In May, the ozone content was high and varied
within a narrow range (379.5 ± 4.1 DU). In June and July, it decreased down to 322.4 ±
10.5 and 287.8±7.22 DU, respectively. In August, the year-to-year variability was higher,
as indicated by a higher standard deviation, 344.1 ± 37.2 DU, and a larger range, 270 to
400 DU. Linear regression gave a significant negative slope (p < 0.05) for July and
August, -5.4 and -24 DU decade -1 , respectively. Nevertheless, the influence of the
decreasing trends in the summer ozone column content on the DIC photoproduction was
largely overshadowed by the variability in sea-ice cover (Fig. III.7).
Figure III.9. Trends in monthly O3 concentration over the study area as
observed by TOMS between 1979 and 2001. Note the changes of scale
among the panels.
68

These results show that variation of sea-ice cover is the most important factor
controlling the photochemical production of DIC in the Arctic surface waters. Our
calculations did not account for seasonal variability in φDIC and optical properties of
surface waters since they are unknown. Fortunately, our field sampling took place at the
time of year (June and July) when both irradiance and river plume extent were maximal,
and therefore when photooxidation processes were most intense. Our sampling time,
however, might not adequately capture the influence of particulate matter dynamics on
the availability of light to CDOM photochemistry. For instance, the continental shelf was
sampled in early July, shortly after the ice breakup when the particle-rich river plume was
spread over most of the shelf area [Carmack and Macdonald, 2002]. Later in the season
the river discharge and suspended particulate matter abundance decrease sharply,
allowing more light available for photochemical processes. The PDIC for the MS, based
on the ratio of aCDOM / at obtained in early July, might therefore be underestimated for late
summer. In contrast, spring bloom events, as observed in the area of Cape Bathurst
polynya [Arrigo and van Dijken, 2004], could temporarily decrease aCDOM / at in some
areas.
The trend in sea-ice cover extent observed in the southeastern Beaufort Sea is
coherent with the trend observed over the whole Arctic Ocean [e.g., Cavalieri et al., 2003;
Rothrock and Zhang, 2005]. The rate of change in the northern hemisphere sea-ice cover
extent between 1979 and 2002 was estimated to be -0.36 ± 0.05 x 10 6 km 2 decade -1
[Cavalieri et al., 2003]. If the PDIC calculated in this study is applicable to the rest of the
Arctic Ocean, the increase in DIC production since 1979 can be estimated for the whole
Arctic. Using the area-normalized slope (in g C y -1 km -2 ) for the CB (Fig. III.7), we
estimated that photochemical production of DIC in the northern hemisphere increased by
19.6-26.0 Gg C y -1 between 1979 and 2002.
III.3.4. Implications for tDOC cycling in Arctic coastal waters
Physical transport appears the main mechanism responsible for the removal of
tDOC in the Arctic Ocean. Benner et al. [2005] reported that physical transport of tDOC
to the northeast Atlantic Ocean represents 25 to 33% of the 25 Tg tDOC y -1 input into the
Arctic Ocean. In addition, a similar fraction of tDOC may be exported through the
69

Canadian archipelago into the northwest Atlantic [Amon, 2004] and into the Arctic deep
basin resulting from the sinking of brine-enriched waters formed over the shelves
[Dittmar, 2004]. The remaining fraction (~34 to 50%) should accumulate or be removed
by biological and photochemical processes in the Polar Surface Water (PSW). The
relative importance for the removal of tDOC by microbial and photochemical processes
is unknown [Amon, 2004]. Our study is a first attempt to quantify the photochemical sink
of tDOC in an Arctic coastal environment affected by large inputs of terrestrial DOM.
The main source of CDOM in the study area is the terrestrial runoff from the
Mackenzie River [Guay et al., 1999; Guéguen et al., 2005]. To estimate tDOM
mineralization, one needs to know the fraction of light absorbed by the terrigenous
CDOM (aCDOM ter ). This fraction (in %) was approximated as
a
a
ter
CDOM
CDOM
⎛
= ⎜
⎝
6.
5
6.
5 f ⎞
⎟ * 100 , (6)
f + 0.
5(
1−
f ) ⎠
where f is the fraction of runoff in the surface layer, and 6.5 and 0.5 are the aCDOM(330)
values for the riverine and marine end-members, respectively (Fig. III.3a). Based on
conservative estimates for f of ~25% in the MS [Macdonald et al., 1995] and ~8% in the
CB [Macdonald et al., 2002], we calculated that 80 and 50% of the aCDOM is of
terrigenous origin in the MS and CB, respectively (Table III.4). The contribution of
aCDOM ter in the Amundsen Gulf is less certain and depends on the surface currents and the
autochthonous production of CDOM. Under calm wind conditions, the Mackenzie River
plume tends to flow eastward and enter the gulf [Carmack and Macdonald, 2002].
However, since PP appears relatively important around Cape Bathurst [Arrigo and van
Dijken, 2004], we assume that only 20% of the CDOM in the AG is of terrigenous origin.
This number represents a typical contribution of tDOC to PSW [Benner et al., 2005].
Based on these assumptions, tDOC photomineralization is estimated to be 19.9 ± 4.2,
13.1 ± 5.9 and 2.8 ± 0.7 Gg C y -1 for the MS, CB and AG, respectively (Table III.4),
representing 1.53 ± 0.32%, 1.0 ± 0.5% and 0.21 ± 0.04% of the tDOC flux into the
southeastern Beaufort Sea via the Mackenzie River. Put together, these estimates
represent 2.8 ± 0.55% of the annual input of tDOC (1.3 Tg) [Telang et al., 1991].
Therefore, our results corroborate the speculation by Benner et al. [2005] that
70

photochemical transformation of tDOM occurs in the PSW but that this removal process
is a small fraction of the total organic carbon flux.
Table III.4. Mean annual DIC photoproduction and estimates of tDOC mineralized by
photooxidation.
Sub-region DIC
photoproduction (1)
Gg C y -1
a
a
ter
CDOM
71
CDOM
(%)
tDOC
mineralized (1)
Gg C y -1
Mackenzie Shelf 24.9 ± 5.2 80 19.9 ± 4.2
Canada Basin 27.8 ± 11.8 50 13.9 ± 5.9
Amundsen Gulf 13.7 ± 3.3 20 2.8 ± 0.7
Total 66.5 ± 18.5 36.6 ± 7.1
(1) Average ± s. d. calculated from 1979 to 2003.
Amon and Meon [2004] suggested that the fate of tDOM under reduced ice cover
could become significantly different from the present situation, and our results allow
further analysis of this prediction for photomineralization losses. Under the minimal sea
ice conditions observed during spring-summer 1998, DIC photoproduction nearly
doubled relative to the averaged value for the period from 1979 to 2003 (123.0 vs. 66.5
Gg C y -1 , P

30.0, 32.5 and 4.2 Gg y -1 of tDOC to be photomineralized in 1998 in the MS, CB, and
AG respectively, the sum of these numbers representing 5.1% of the annual tDOC release
from the Mackenzie River. The ice-free scenario gives an equivalent estimate of 6.2%
(assuming no change in tDOC input). For these conditions, we estimated that 43.5 Gg y -1
of allochthonous DOC would be photomineralized in the Canada Basin, which is higher
than the estimates for autochthonous organic carbon sequestered in the deep ocean basin
(22.8 to 42 Gg C y -1 ). This range is based on the assumptions that 1% of the annual PP is
buried in sediments [Stein and Macdonald, 2004] and that PP rate ranges from 9.5 to 17.4
g C m -2 in the open water of CB [Lee and Whitledge, 2005].
In this study, photooxidation of CDOM within or under sea ice was not
considered because most of the radiative energy is either backscattered to the atmosphere
or absorbed by the freshly produced biogenic material accumulated at the bottom of sea
ice . Although photochemical reactions within the organic-rich bottom ice layer is likely
to play a role in the marine DOM cycling [Xie and Gosselin, 2005], it does not affect the
underlying terrigenous DOM. Due to the lack of tDOM data in sea ice [Amon, 2004], it is
impossible at present to estimate the photomineralization of tDOC in sea ice.
III.4. Summary and conclusions
This study represents the first attempt to quantify photomineralization of DOM in
an Arctic coastal ecosystem that receives large riverine inputs of terrigenous materials.
The quantum yield spectra for DIC photoproduction (φDIC) are the first to be obtained for
the Arctic, and among the first for marine waters in general. The non-conservative
behavior of φDIC in the freshwater-saltwater transitional zone suggests that the
photoreactivity of tDOM transported by Mackenzie River changed when it was mixed
into coastal and offshore waters, and/or that marine algae-derived DOM was less
photoreactive than tDOM in terms of photoproduction of DIC.
Photomineralization of DOC in the study area, as estimated from DIC
photoproduction, occurred at a rate equivalent to ~10% of the bacterial respiration rate
and

esponse to the decreasing sea ice extent. Our estimates of tDOC photomineralization
confirm that photooxidation presently is not a major removal pathway for tDOM in the
Arctic. The photochemical sink for tDOM is, however, anticipated to grow in the future
due to decrease in sea-ice extent as predicted by the Global Circulation Model [ACIA,
2005]. In addition to remineralizing CDOM to CO2, photolysis also produces biologically
labile DOM compounds [e.g., Kieber et al., 1989; Mopper et al., 1991]. The latter may
provide another pathway by which Arctic tDOM is removed, thereby amplifying the
effect of decrease in sea ice cover on tDOM cycling in the Arctic Ocean.
III.5. Auxiliary material: On the determination of the Apparent Quantum Yield for
DIC photoprodcution
In sections III.2.3 and III.2.4, we described the method adopted to determine of
the Apparent Quantum Yield for DIC photoproduction. Briefly, we exposed filtered
seawater to artificial sunlight. To obtain spectral value for φDIC, the incident light was
filtered using eight Schott long-pass glass filters. For each of the six experiments, spectral
absorption by colored dissolved organic matter (aCDOM) was measured before and after
the irradiation. Figure III.A.1 presents the measured aCDOM for the six stations before
(black) and after (red) the irradiation experiments. The loss of absorption observed is due
to photobleaching. This change was accounted for in the calculation of the AQY by using
an appropriately averaged absorption coefficient. To calculated the amount of energy
absorbed by the CDOM during the course of the experiment, Qa, the following equation
was used:
⎛ a ⎞ CDOM
= E(
0)
× ⎜ × S × [ 1−
exp( −a
× L)
]
a ⎟
, (mol photons s
⎝ t ⎠
-1 nm -1 ),
Qa t
S: cross-section of the irradiation cells;
L: pathlength of the irradiation cells;
E(0): irradiance just below the upper window of the irradiation cells
at: total absorption coefficient, which is only acdom+awater since the sample was filtered
before irradiation.
73

Note that in all samples, the fraction aCDOM/atotal is >90% in the UV-blue domain.
aCDOM was averaged according to first-order kinetic decay (i.e., mean aCDOM =
exp((ln(ainitial)+ln(aend))/2)).
To derive the φDIC(λ) values, the iterative curve-fitting method of Johannessen
and Miller [2001] was applied. Briefly, this method assumes an appropriate mathematical
functional form with unknown parameters to express the change of φDIC as a function of
wavelength. Decreasing exponential functional forms are usually chosen for
photochemical quantum yield spectra. The amount of DIC produced in an irradiation cell
over the exposure time can then be predicted as a product of three terms: the assumed
φDIC(λ) functional form, the measured incident irradiance, and the sample absorption
coefficient. The optimum values of the unknown parameters in the assumed φDIC(λ)
functional form are obtained by varying these parameters from initial estimates until
minimum difference between the measured and predicted DIC production is achieved.
We tested two exponential form to fit the data
1.
2.
φ
−(
m1
+ m2
( λ + 290)
DIC ( λ)
= e (Single exponential, as in Johannessen and
Miller [2001]);
k2
/ ( λ + k3)
φ DIC ( λ)
= k1
e (Quasi-exponential).
where m1, m2, k1, k2 and k3 are fitting parameters. Figure III.A.2 shows the measured
versus the modeled DIC within each irradiated cells. Table III.A.1 gives the regression
coefficients for each model and station. These show that the performance of Quasi-
exponential form model was better than the single exponential form as previously used
by Johannessen and Miller [2001].
74

Figure III.A.1. Absorption spectra of CDOM for before (black line) and
after the irradiation for the sample under WG280 cutoff filter (red line).
The time of irradiation and the irradiance level under the cutoff filter are
also shown.
75

Table III.A.1. Comparison of statistical results between the single exponential and quasiexponential
functional forms
Station R5d R5a R9 108 406 409
r Single exponential 0.990 0.994 0.964 0.956 0.994 0.988
2
Quasi-exponential 0.998 1.00 0.984 0.976 0.996 0.994
∑(modeled- Single exponential 0.173 0.065 0.028 0.075 0.010 0.011
measured) 2 Quasi-exponential 0.025 0.002 0.009 0.020 0.006 0.003
Figure III.A.2. Modeled versus measured DIC photoproduction rates for
the Single exponential (squares) and Quasi-exponential (circles) functional
forms.
76

III.6. References
Aagaard, K., and E. C. Carmack (1989), The role of sea ice and other fresh water in the
arctic circulation, J. Geophys. Res., 94, 14485-14498.
ACIA (2005), Arctic Climate Impact Assessment, 1042 pp., Cambridge Press University.
Amon, R. M. W. (2004), The role of dissolved organic matter for the organic carbon
cycle in the Arctic Ocean, in The organic carbon cycle in the Arctic Ocean, edited
by R. Stein, and R. W. MacDonald, pp. 83-99, Springer, New York.
Amon, R. M. W., and B. Meon (2004), The biogeochemistry of dissolved organic matter
and nutrients in two large Arctic estuaries and potential implications for our
understanding of the Arctic Ocean system, Mar. Chem., 92, 311-330.
Arrigo, K. R., and G. L. van Dijken (2004), Annual cycles of sea ice and phytoplankton
in Cape Bathurst polynya, southeastern Beaufort Sea, Canadian Arctic, Geophys.
Res. Lett., 31, L08304, doi:10.1029/2003GL013978.
Babin, M., and D. Stramski (2002), Light absorption by aquatic particles in the nearinfrared
spectral region, Limnol. Oceanogr., 47, 911-915.
Babin, M., D. Stramski, G. M. Ferrari, H. Claustre, A. Bricaud, G. Obolensky, and N.
Hoepffner (2003), Variations in the light absorption coefficients of phytoplankton,
nonalgal particles, and dissolved organic matter in coastal waters around Europe,
J. Geophys. Res., 108(C7), 3211, doi: 10.1029/2001JC000882.
Barber, D. G., and J. M. Hanesiak (2004), Meteorological forcing of sea ice
concentrations in the southern Beaufort Sea over the period 1979 to 2000, J.
Geophys. Res., 109(C6), C06014, doi:10.1029/2003JC002027.
Benner, R., B. Benitez-Nelson, K. Kaiser, and R. M. W. Amon (2004), Export of young
terrigenous dissolved organic carbon from rivers to the Arctic Ocean, Geophys.
Res. Lett., 31, L05305.
Benner, R., P. Louchouarn, and R. M. W. Amon (2005), Terrigenous dissolved organic
matter in the Arctic Ocean and its transport to surface and deep waters of the
North Atlantic, Global Biogeochem. Cycles, 19, GB2025,
doi:10.1029/2004GB002398.
Benner, R., and S. Opsahl (2001), Molecular indicators of the sources and
transformations of dissolved organic matter in the Mississippi river plume, Org.
Geochem., 32, 597-611.
77

Camill, P. (2005), Permafrost thaw accelerates in boreal peatlands during late-20th
century climate warming, Climatic Change, 68, 135-152.
Carmack, E. C., and R. W. Macdonald (2002), Oceanography of the Canadian Shelf of
the Beaufort Sea: A setting for Marine Life, Arctic, 55, 29-45.
Carmack, E. C., R. W. MacDonald, and S. Jasper (2004), Phytoplankton productivity on
the Canadian Shelf of the Beaufort Sea, Mar. Ecol. Prog. Ser., 277, 37-50.
Cavalieri, D. J., P. Gloersen, and J. Zwally (1990), DMSP SSMI/I daily polar gridded sea
ice concentrations, edited by J. A. Maslanik, and J. Stroeve, pp. Digital media,
National Snow and Ice Data Center, Boulder, CO.
Cavalieri, D. J., C. L. Parkinson, and K. Y. Vinnikov (2003), 30-Year satellite record
reveals contrasting Arctic and Antarctic decadal sea ice variability, Geophys. Res.
Lett., 30, 1970, doi:10.1029/2003GL01803.
Comiso, J. C., J. Yang, H. Susumo, and R. A. Krishfield (2003), Detection change in the
Arctic using satellite and in situ data, J. Geophys. Res., 108(C12), 3384,
doi:10.1029/2002JC001347.
Del Vecchio, R., and N. V. Blough (2002), Photobleaching of chromophoric dissolved
organic matter in natural waters: kinetics and modelling, Mar. Chem., 78, 231-253.
Dittmar, T. (2004), Evidence for terrigenous dissolved organic nitrogen in the Arctic
deep sea, Limnol. Oceanogr., 49, 148-156.
Dittmar, T., and G. Kattner (2003), The biogeochemistry of the river and shelf ecosystem
of the Arctic Ocean: a review, Mar. Chem., 83, 105-120.
Dugdale, R. C., and J. J. Goering (1967), Uptake of new and regenerated forms of
nitrogen in primary production, Limnol. Oceanogr., 12, 196-206.
Fioletov, V. E., M. G. Kimlin, N. Krotkov, L. J. B. McArthur, J. B. Kerr, D. I. Wardle, J.
R. Herman, R. Meltzer, T. W. Mathews, and J. Kaurola (2004), UV index
climatology over the United States and Canada from ground-based and satellite
estimates, J. Geophys. Res., 109(D22), D22308, doi:10.1029/2004JD004820.
Frey, K. E., and L. C. Smith (2005), Amplified carbon release from vast West Siberian
peatlands by 2100, Geophys. Res. Lett., 32, L09401, doi:10.1029/2004GL022025.
Garneau, M.-È., W. F. Vincent, L. Alonso-Saez, Y. Gratton, and C. Lovejoy (2006),
Prokaryotic community structure and hetereotrophic production in a riverinfluence
arctic ecosystem, Aquat. Microb. Ecol., 42, 27-40.
78

Guay, C. K., G. P. Klinkhammer, K. K. Falkner, R. Benner, P. G. Coble, T. E. Whitledge,
B. Black, F. J. Bussell, and T. A. Wagner (1999), High-resolution measurements
of dissolved organic carbon in the Arctic Ocean by in situ fiber-optic
spectrometry, Geophys. Res. Lett., 26(8), 1007-1010.
Guéguen, C., L. Guo, and N. Tanaka (2005), Distributions and characteristics of colored
dissolved organic matter in the Western Arctic Ocean, Cont. Shelf Res., 25, 1195-
1207.
Hansell, D. A., D. Kadko, and N. R. Bates (2004), Degradation of Terrigenous Dissolved
Organic Carbon in the Western Arctic Ocean, Science, 304, 858-861.
Herman, J. R., and E. A. Celarier (1997), Earth surface reflectivity climatology at 340-
380 nm from TOMS data, J. Geophys. Res., 102(D23), 28003-28011.
Herman, J. R., N. Krotkov, E. Celarier, D. Larko, and G. Labow (1999), Distribution of
UV radiation at the Earth's surface from TOMS-measured UV-backscattered
radiances, J. Geophys. Res., 104(D10), 12059-12076.
Hernes, P. J., and R. Benner (2003), Photochemical and microbial degradation of
dissolved lignin phenols: Implications for the fate of terrigenous dissolved organic
matter in marine environments, J. Geophys. Res., 108(C9), 3291, doi:
1029/2002JC001421.
Hu, C., F. E. Muller-Karger, and R. G. Zepp (2002), Absorbance, absorption coefficient,
and apparent quantum yield: A comment on common ambiguity in the use of
these optical concepts, Limnol. Oceanogr., 47, 1261-1267.
Johannessen, S. C. (2000), A photochemical sink for dissolved organic carbon in the
ocean, Ph. D. thesis, Dalhousie University, Halifax.
Johannessen, S. C., and W. L. Miller (2001), Quantum yield for the photochemical
production of dissolved inorganic carbon in seawater, Mar. Chem., 76, 271-283.
Kieber, D. J., J. McDaniel, and K. Mopper (1989), Photochemical source of biological
substrates in sea water: implications for carbon cycling, Nature, 341, 637-639.
Kieber, R. J., X. Zhou, and K. Mopper (1990), Formation of carbonyl compounds from
UV-induced photodegradation of humic substances in natural waters: Fate of
riverine carbon in the sea, Limnol. Oceanogr., 35(7), 1503-1515.
Krotkov, N., J. Herman, P. K. Bhartia, C. Seftor, A. Arola, S. Kalliskota, P. Taalas, and I.
V. Geogdzhaev (2002), Version 2 total ozone mapping spectrometer ultraviolet
algorithm: problems and enhancements, Opt. Eng., 41, 3028-3039.
79

Krotkov, N. A., P. K. Bhartia, J. R. Herman, V. Fioletov, and J. Kerr (1998), Satellite
estimation of spectral surface UV irradiance in the presence of tropospheric
aerosols - 1. Cloud-free case, J. Geophys. Res., 103(D8), 8779-8793.
Krotkov, N. A., J. R. Herman, P. K. Bhartia, V. Fioletov, and Z. Ahmad (2001), Satellite
estimation of spectral surface UV irradiance 2. Effects of homogeneous clouds
and snow, J. Geophys. Res., 106(D11), 11743-11759.
Lee, S. H., and T. R. Whitledge (2005), Primary and new production in the deep Canada
Basin during summer 2002, Polar Biology, 28, 190-197.
Macdonald, R. W., F. A. McLaughlin, and E. C. Carmack (2002), Fresh water and its
sources during the SHEBA drift in the Canada Basin of the Arctic Ocean, Deep
Sea Res., Part I, 49, 1769-1785.
Macdonald, R. W., D. W. Paton, E. C. Carmack, and A. Omstedt (1995), The freshwater
budget and under-ice spreading of Mackenzie River Water in the Canadian
Beaufort Sea based on salinity and 18 O/ 16 O measurements in water and ice., J.
Geophys. Res., 100(C1), 895-919.
Madronich, S., and S. Flocke (1999), The role of solar radiation in atmospheric chemistry,
in Handbook of Environmental Chemistry, edited by P. Boule, pp. 1-26, Springer-
Verlag, Heidelberg, Germany.
Meon, B., and R. M. W. Amon (2004), Heterotrophic bacterial activity and fluxes of
dissolved free amino acids and glucose in the Arctic rivers Ob, Yenisei and the
adjacent Kara Sea, Aquat. Microb. Ecol., 37, 121-135.
Miller, W. L., and R. G. Zepp (1995), Photochemical production of dissolved inorganic
carbon from terrestrial organic matter : Significance to the oceanic carbon cycle,
Geophys. Res. Lett., 22, 417-420.
Mopper, K., and D. J. Kieber (2002), Photochemistry and the Cycling of Carbon, Sulfur,
Nitrogen and Phosphorus, in Biogeochemistry of Marine Dissolved Organic
Matter, edited by D. A. Hansell, and C. A. Carlson, pp. 455-507, Academic Press,
San Diego.
Mopper, K., X. Zhou, R. J. Kieber, D. J. Kieber, R. J. Sikorski, and R. D. Jones (1991),
Photochemical degradation of dissolved organic carbon and its impact on the
oceanic carbon cycle, Nature, 353, 60-62.
Overland, J. E., M. C. Spillane, and N. N. Soreide (2004), Integrated analysis of physical
and biological Pan-Arctic change, Climatic Change, 63, 291-322.
80

Peterson, B. J., R. M. Holmes, J. W. McClelland, C. J. Vorosmarty, R. B. Lammers, A. I.
Shiklomanov, and S. Rahmstorf (2002), Increasing river discharge to the Arctic
Ocean, Science, 298, 2171-2173.
Rachold, V., H. Eicken, V. V. Gordeev, M. N. Grigoriev, H.-W. Hubberten, A. P. Lisitzin,
V. P. Shevchenko, and L. Schirrmeister (2004), Modern terrigenous organic
carbon input to the Arctic Ocean, in The organic carbon cycle in the Arctic Ocean,
edited by R. Stein, and R. W. MacDonald, pp. 33-41, Springer, New York.
Rex, M., R. J. Salawitch, P. von der Gathen, N. R. P. Harris, M. P. Chipperfield, and B.
Naujokat (2004), Arctic ozone loss and climate change, Geophys. Res. Lett., 31,
L04116, doi:10.1029/2003GL018844.
Rothrock, D. A., and J. Zhang (2005), Arctic Ocean sea ice volume: What explains its
recent depletion?, J. Geophys. Res., 110(C1).
Serreze, M. C., J. E. Walsh, F. S. Chapin, T. Osterkamp, M. Dyurgerov, V. Romanovsky,
W. C. Oechel, J. Morison, T. Zhang, and R. G. Barry (2000), Observational
evidence of recent change in the northern high-latitude environment, Climatic
Change, 46, 159-207.
Smith, L. C., G. M. MacDonald, A. A. Velichko, D. W. Beilman, O. K. Borisova, K. E.
Frey, K. V. Kremenetski, and Y. Sheng (2004), Siberian peatlands a net carbon
sink and global methane source since the early Holocene, Science, 303(5656),
353-356.
Stein, R., and R. W. Macdonald (2004), Organic carbon budget: Arctic Ocean vs. Global
Ocean, in The Organic Carbon Cycle in the Arctic Ocean, edited by R. Stein, and
R. W. MacDonald, pp. 315-322, Springer, New York.
Tassan, S., and G. M. Ferrari (1995), An alternative approach to absorption
measurements of aquatic particles retained on filters, Limnol. Oceanogr., 40,
1358-1368.
Tassan, S., and G. M. Ferrari (2002), A sensitivity analysis of the 'transmittancereflectance'
method for measuring light absorption by aquatic particles, J.
Phytoplankton. Res., 24, 757-776.
Telang, S. A., R. Pocklington, A. S. Naidu, E. A. Romankevich, I. I. Gitelson, and
Glacyshev (1991), Carbon and mineral transport in Major North American,
Russian Arctic, and Siberian Rivers: the St Lawrence, the Mackenzie, the Yukon,
the Arctic Alaskan Rivers, the Arctic Basin rivers in the Soviet Union, and the
Yenisei, in Biogeochemistry of Major World Rivers. SCOPE 42, edited by E. T.
Degens, S. Kempe, and J. E. Richey, pp. 74-104, Wiley, New York.
81

Vähätalo, A. V., M. Salkinoja-Salonen, P. Taalas, and K. Salonen (2000), Spectrum of
the Quantum yield for the photochemical mineralization of dissolved organic
carbon in a humic lake, Limnol. Oceanogr., 45, 664-676.
Vähätalo, A. V., and R. G. Wetzel (2004), Photochemical and microbial decomposition
of chromophoric dissolved organic matter during long (months-years) exposures,
Mar. Chem., 89, 313-326.
Vincent, W. F., and C. Belzile (2003), Biological UV exposure in the polar oceans:
Arctic-Antarctic comparisons, in Antarctic Biology in a Global Context, edited by
W. J. Wolff, pp. 176-181., Backhuys Publishers, Leiden, Netherlands.
Wadhams, P., and N. R. Davis (2000), Further evidence of ice thinning in the Arctic
Ocean, Geophys. Res. Lett., 27, 3973-3975.
Xie, H., and M. Gosselin (2005), Photoproduction of carbon monoxide in first-year sea
ice in Franklin Bay, southeastern Beaufort Sea, Geophys. Res. Lett., 32, L12606,
doi:10.1029/2005GL022803.
Xie, H., O. C. Zafiriou, W.-J. Cai, R. G. Zepp, and Y. Wang (2004), Photooxidation and
its effects on the carboxyl content of dissolved organic matter in two coastal
rivers in the southeastern United States, Environ. Sci. Tech., 38(15), 4113-4119.
82

Chapitre IV : Quantification améliorée de la
photooxydation de la matière organique dissoute
colorée dans les eaux côtières à l’aide des
propriétés optiques inhérentes dérivées des
données de couleur de l’océan
83

IV.A Résumé
Pour estimer le taux de photooxydation de la matière organique dans les eaux côtières, le
rapport entre les coefficients d’absorption de la matière organique dissoute colorée (CDOM)
et total ([a CDOM/a t]) se doit d’être connu depuis l’ultraviolet (UV) jusqu’à la partie verte du
spectre électromagnétique. La variabilité de ce rapport à 412 nm, observée à 307 stations
visitées dans plusieurs environnements côtiers, est très importante, soit entre 20 et 94%. Une
relation empirique est proposée pour estimer [a CDOM/a t](412) à partir des mesures spatiales de
la réflectance de l’océan. L’incertitude absolue de l’algorithme de [a CDOM/a t](412) est de ±
14%. Dans des environnements tels que la Mer Baltique et la Mer du Nord, l’algorithme
permet de séparer le CDOM des particules non-algales à l’échelle régionale. À l’aide d’un
modèle spectral de photoproduction de carbone inorganique dissous (DIC) intégrée dans la
colonne d’eau, on a évalué l’impact de 1) l’erreur de [a CDOM/a t](412) retrouvé, et de 2)
l’extrapolation de ce rapport entre l’UV et le vert, sur la photooxydation. L’incertitude sur la
valeur absolue de [a CDOM/a t](412) retrouvée résulte en une incertitude sur la photoproduction
de DIC inférieure à 50%. Si la valeur du coefficient d’absorption total à 412 nm est connue,
l’erreur due à l’extrapolation spectrale de [a CDOM/a t](412) est inférieure à 20%. L’algorithme a
été appliqué à une image SeaWiFS du sud-est de la Mer de Beaufort où des rendements
quantiques apparents spectraux (AQY) pour la photoproduction de DIC furent déterminés
dans les principales masses d’eau. En se basant sur ces spectres de AQY, la variabilité
spatiale du taux de photoproduction de DIC résultant des variations du spectre de [a CDOM/a t]
était de l’ordre d’un facteur trois au-dessus du plateau continental et au-delà. Ces résultats
démontrent clairement qu’il est nécessaire de tenir compte des variations spatiales de
[a CDOM/a t] pour quantifier les processus de photooxydation dans les eaux côtières.
84

IV.B. Article soumis à la revue Journal of Geophysical Research - Oceans (28
décembre 2006): “Improved quantification of Chromophoric Dissolved Organic
Matter photooxidation in coastal waters using satellite-derived inherent optical
properties”
Simon Bélanger 1 , Marcel Babin 1 and Pierre Larouche 2
1 Laboratoire d’Océanographie de Villefranche, Centre National de la Recherche
Scientifique, Université Pierre et Marie Curie - Paris 6, 06230 Villefranche-sur-mer,
France
2 Institut Maurice-Lamontagne, Pêches et Océans Canada, Mont-Joli, QC, Canada G5H
3Z4
85

Abstract
To estimate the rate of organic matter photooxidation in coastal waters, the ratio between
chromophoric dissolved organic matter (CDOM) and total absorption coefficients
([aCDOM/at]) must be known from the UV to the green spectral range. At 307 sites
sampled in various coastal marine environments, the [aCDOM/at] at 412 nm was found to
vary within wide range, between 20 and 94%. An empirical algorithm was developed to
retrieve [aCDOM/at](412) from satellite remote sensing reflectance. The absolute
uncertainty on the [aCDOM/at] retrieval was ±14%. As exampled with the data from the
Baltic and North Seas, the algorithm provides a mean to separate the CDOM from the
colored detrital material (i.e. CDM=CDOM+non-algal particles) absorption coefficient at
the regional scale. Using a depth-integrated spectral dissolved inorganic carbon (DIC)
photoproduction model, we assessed the impact of 1) the error on the derived [aCDOM/at]
ratio, and 2) the extrapolation of this ratio over the green and UV spectral domains, on
photooxidation. The uncertainty associated with retrieved [aCDOM/at](412) results in an
uncertainty on DIC photoproduction estimates of

IV.1. Introduction
Dissolved organic carbon (DOC) in the ocean is a major pool of carbon in the
earth system of 685 Gt C [Hansell and Carlson, 1998], a value comparable to the mass of
atmospheric carbon dioxide (CO2). In the context of global warming, it is important to
better understand and quantify the sources and sinks of DOC in the ocean [Carlson,
2002]. The role of photochemical oxidation (c.f. photooxidation) of the chromophoric
dissolved organic matter (CDOM) has received increasing attention because it may
represent a major sink for DOC from terrestrial [Kieber et al., 1990; Miller and Zepp,
1995; Benner and Opsahl, 2001] and marine [Johannessen, 2000] origins, in the ocean.
Nevertheless, quantitative assessment for this carbon sink at global- or regional-scale are
scarce [e.g., Johannessen, 2000; Bélanger et al., 2006].
Quantitative assessment of CDOM photooxidation requires the knowledge of the
incident irradiance in the ultraviolet (UV) and blue part of the spectrum, the spectral
absorption properties of seawater and its constituents, and the spectral apparent quantum
yield of the reaction (AQY). The latter is defined as the ratio of the number of
photoproduct molecules formed per photon absorbed by the CDOM. Carbon monoxide
(CO) and carbon dioxide (CO2, measured as dissolved inorganic carbon (DIC)) are the
two major photoproducts of CDOM photooxidation [see review by Mopper and Kieber,
2002]. Assuming that all incident radiation penetrating an optically homogenous water
column is absorbed underneath, the depth-integrated photoproduction rate of DIC (or CO)
can be calculated as:
P
=
λ max
∫
λ min
E
( 0
a ( λ)
, λ) AQY ( λ)
dλ
− CDOM
d , (1)
at
( λ)
where λmin - λmax is the spectral range within which photooxidation is efficient, aCDOM(λ)
and at(λ) are the spectral CDOM and total absorption coefficients, respectively, and
−
E ( 0 , λ)
is the spectral downward irradiance just beneath sea surface (see the list of
d
Notation). In Eq. 1, the ratio of aCDOM(λ) to at(λ) is the fraction of the incident light
absorbed by the CDOM in the whole water column. While this ratio approaches 0.9-1.0
in the UV domain in the open ocean [Johannessen, 2000], absorbing particles may
compete for light absorption in coastal environments. Because this is where
87

photooxidation may be particularly important as rivers discharge form buoyant and
shallow plumes that promote the exposure of CDOM to solar radiation [see Dagg et al.,
2004 and references therein], information on the ratio of aCDOM(λ) to at(λ) is highly
desirable to improve quantification of photochemical processes.
On a volume basis, the Arctic Ocean receives the highest amount of terrigenous
DOC relative to the world ocean [Rachold et al., 2004]. During the past decade, dramatic
changes have been observed in the Arctic ice climate, ocean circulation, storage of
freshwater, ozone concentration, permafrost thawing and riverine discharge [ACIA, 2005].
Recently, Bélanger et al. [2006] have shown that CDOM photooxidation in the
southeastern Beaufort Sea (western Arctic) could increase significantly in the future in
response to the predicted trend of ongoing reduction of sea ice cover. Because this region
is influenced by the Mackenzie River, which annually supplies enormous amount of
runoff (330 km 3 ), suspended particles (124 Tg) and DOC (1.3 Tg) [Telang et al., 1991;
Macdonald et al., 1998], optical properties of the surface waters are largely controlled by
terrigenous inputs. The spatial and temporal variability of the optical properties of the
surface waters are, therefore, expected to be high during the spring-summer season when
the photooxidation occurs and the river plume spreads over the continental shelf.
Bélanger et al. [2006] found the ratio [aCDOM/at] in the UV-blue part of the spectrum to
vary between 30 and 95%. In their calculation of photoxidation at the scale of the
Beaufort Sea, they could only use regionally averaged values for [aCDOM/at], and
therefore only partly account for that variability in the optical properties of the surface
waters. This is a limitation of their approach that perhaps impacts on large-scale
photooxidation estimates, and certainly does small ones.
Satellite remote sensing is particularly suited for monitoring CDOM
photooxidation in the Arctic Ocean. Remote observations of sea ice concentration, cloud
cover and total atmospheric ozone content allow the calculation of UV and visible
radiation fluxes penetrating the water column. For example, satellite-derived UV
radiation has been used to study the impact of the decline in ozone on the oceanic
primary production in Antarctica [Arrigo et al., 2003], and the impact of the decline in
sea ice concentration on the CDOM photooxidation in the Arctic [Bélanger et al., 2006].
Ocean Color remote sensing can provide relevant bio-optical indicators to refine the
88

quantification of CDOM photooxidation. During the last two decades, several algorithms
have been proposed to derive inherent optical properties (IOP) from the remote sensing
reflectance spectrum [see review by IOCCG, 2006]. None of them, however, are
specifically designed to derive separately the total and CDOM absorption coefficients.
Indeed, because CDOM and non-algal particles (NAP) are characterized by similar
exponentially decreasing absorption spectrum with increasing wavelength [Bricaud et al.,
1981; Roesler et al., 1989], current algorithms fail to distinguish them and, as a practical
solution, combine them into a unique absorption coefficient referred to as colored detrital
material [CDM=CDOM+NAP, see Siegel et al., 2002]. While CDM is dominated by
CDOM at the global scale [~0.817±0.14 at 440 nm; Siegel et al., 2002], the relative
contribution of CDOM to CDM varies within a wide range in coastal waters, i.e. from
0.25 to 0.95 (see below).
In view of modeling depth-integrated photooxidation in oceanic coastal waters,
our first objective was to document the variability in the ratio [aCDOM/at]. Using an
extensive data set covering various coastal environments, we observed that [aCDOM/at] at
412 nm varies between 0.20 to 0.90, with an average of 0.53 and a standard deviation (σ)
of 0.16, which stresses the importance of the variability in [aCDOM/at](412) in coastal
waters. The second objective was to propose an algorithm for the retrieval of [aCDOM/at]
at 412 from the remote sensing reflectance spectrum. In what follows, we first present the
in situ data sets used to develop our empirical algorithm. Next we quantify the sensitivity
of the depth-integrated photoproduction of DIC to 1) the inaccuracy of the retrieved
[aCDOM/at](412), and 2) to the extrapolation of [aCDOM/at] from 412 nm over the whole
spectral range efficient for CDOM photooxidation (300 to 600 nm). Finally, using Sea-
viewing Wide Field-of-view Sensor (SeaWiFS) data, we demonstrate the benefit of the
proposed method for the quantification of photooxidation in the Southeastern Beaufort
Sea.
IV.2. Materials and Methods
IV.2.1. Data sets description and measurements
Two different data sets collected in various coastal environments are used in this
study: the Coastal Surveillance Through Observation of Ocean Color (COASTlOOC) and
89

the Canadian Arctic Shelf Exchange Study (CASES) data sets. Details on the time period
and sampling locations of COASTlOOC can be found in Babin et al. [2003a; 2003b].
Only the COASTlOOC data collected in coastal (so-called Case 2) waters were retained
for the algorithm development and validation: 29 stations in the northern Adriatic Sea; 53
stations in the southwestern Baltic Sea; 68 stations in the North Sea; and 58 stations in
the English Channel along the southern coast of England. Although our prime interest is
photooxidation in the Arctic ocean, we also used the COASTlOOC data to reach more
general conclusions about [aCDOM/at] variability in coastal waters, and to develop a more
robust [aCDOM/at] algorithm.
CASES sampling was conducted in June and July 2004 onboard the CCGS
Amundsen in the southeastern Beaufort Sea (Fig. IV.1). The Amundsen Gulf was first
sampled after the sea ice opening in June and was revisited in late July when most of the
area was free of ice . The surface waters were, therefore, influenced by melt waters at few
stations. The Mackenzie Shelf was sampled during late June and early July when the
Mackenzie River discharge peaks [Macdonald et al., 1998], making those stations typical
Case 2 waters. At 47 stations, our whole set of in situ optical measurements was
achieved, while at other stations only IOP measurements (52 and 25 spectra of the
particle and CDOM absorption coefficient, respectively) were made because of either sun
elevation too high for proper apparent optical properties (AOP) measurements (sun zenith
angle >70°) or occasional technical problems with instruments (Fig. IV.1).
90

Figure IV.1. Map of the study area showing location of stations where
IOPs (dots), SPMR (triangles) and ASD (squares) measurements were
made. Contour lines of sea ice concentration of 50% are also shown for
June 1 st (red), July 1 st (blue) and August 1 st (green).
IV.2.1.1. Inherent optical properties
Details on the IOP measurements for COASTlOOC can be found in Babin et al.
[2003a; 2003b]. Similar protocols were adopted during CASES including a few
modifications as described below. At each station a sample of ~20 L of surface water was
collected with a clean bucket for spectrophotometric analyses. Subsamples for the
determination of aCDOM were filtered through 0.2-μm Anotop® syringe filters (Whatman)
and kept into 100-mL acid-cleaned amber glass bottles. For the determination of the
absorption coefficient of particles, ap, suspended particles were retained onto 25-mm
GF/F glass fiber filters (Whatman; pore size of 0.7-μm) by filtering 0.1 to 3.5 L of
seawater. The glass bottles and GF/F filters were stored frozen (seawater: -20 °C; particle:
-80 °C) in the dark until being analyzed two to four months later in the land based
laboratory. Samples treatment and methods applied to determine the ap and aCDOM spectra
are detailed in Bélanger et al. [2006]. Briefly, ap(λ) was determined at 1-nm resolution
between 350 and 750 nm according to the transmittance-reflectance protocol developed
by Tassan and Ferrari [1995; 2002]. After ap measurements, the filters were soaked
during ~30 minutes in 90% methanol to extract phytoplankton pigments [Kishino et al.,
91

1985], and the transmittance-reflectance measurements were then repeated for the
determination of non-algal absorption (aNAP). The absorption coefficient by the
phytoplanktonic pigments (aφ) is calculated as ap(λ) – aNAP(λ). The aCDOM(λ) was
measured in 10-cm quartz cuvettes between 250 and 800 nm with 1-nm increments using
a dual beam spectrophotometer (Perkin-Elmer Lambda 35). A baseline correction was
applied by subtracting the absorbance value averaged over an interval of 5 nm around
685 nm from all the spectral values [Babin et al., 2003b]. After applying a baseline
correction to the aCDOM(λ) spectrum, a non-linear least squares curve fitting (Levenberg-
Marquardt) procedure was used to fit an exponential model to the data between 300 and
500 nm:
SCDOM
*( λ0
−λ
)
a CDOM ( λ)
= aCDOM
( λ0
) e , (2)
where λ0 is a reference wavelength (e.g. 412 nm) and SCDOM is the spectral slope of the
aCDOM(λ) spectrum.
The spectral absorption and beam attenuation coefficients of seawater constituents
(i.e. excluding pure seawater itself), at-w and ct-w, were determined at nine wavelengths
(412, 440, 488, 510, 532, 555, 650, 676 and 715 nm) using a submersible
spectrophotometer (ac-9, WET Labs Inc.). The ac-9 was either operated onboard the ship
laboratory where ~20 L of surface water were passed through the instrument by gravity,
or deployed from the deck or from a zodiac to obtain vertical profiles from surface down
to ~40 m. The manufacturer calibration was checked daily using ~10 L of water purified
onboard (Milli-Q Gradient A10, pre-filtered with a Millipore RiOs 8). Temperature and
salinity corrections were applied using the latest coefficients determined by the
manufacturer [M. Twardowski, pers. comm.]. The ac-9 overestimates absorption
coefficients due to the loss of scattered photons that do not reach the detector [Zaneveld
et al., 1994]. To correct for that error, the following expression was applied:
m
m
a − ( λ)
= a ( λ)
− εb
( λ)
, (3)
t
w
where a m (λ) is the measured absorption coefficient, b m (λ) is the measured scattering
coefficient calculated as the difference ct-w(λ) - a m (λ)), and ε is the fraction of the
scattering coefficient that corresponds to photons not detected by the sensor. We used the
matrix inversion procedure proposed by Gallegos and Neale [2002] to estimate ε. The
92

spectral shapes for aCDOM(λ), aNAP(λ) and aφ(λ), and the relationship between aNAP(440)
and b m (440), which are necessary in the inversion, were determined using measurements
on discrete water samples (described above) following the statistically augmented method
described by Gallegos and Neale [2002]. When available, the vertical at-w(λ) profiles
were optically averaged from surface down to the first attenuation length using the
weighting function approach proposed by Gordon [1992].
The ratio [aCDOM/at](412) was computed using the spectrophotometrically
determined IOPs, excepted for seven of the CASES stations where aCDOM(412) was not
available and was computed as at-w(412) – ap(412). A comparison of the two methods
indicates that the [aCDOM/at](412) agree within ±0.11 at 95% confidence level.
IV.2.1.2. Remote sensing reflectance
During COASTlOOC, irradiance measurements were carried out from helicopter
(202) and from ship (6). Vertical profiles of the spectral upwelling irradiance, Eu(z, λ),
and the spectral downwelling irradiance just above the sea surface, Es(0 + , λ), were
measured at 13 wavebands (412, 443, 456, 490, 510, 532, 560, 620, 665, 683, 705, 779
and 866 nm) with a Satlantic free-fall SPMR (SeaWiFS Profiling Multichannel
Radiometer) and SMSR (SeaWiFS Multichannel Surface Reference), respectively. Both
radiometers had been calibrated less than 3 months before each cruise by the instrument
manufacturer. Platform (ship or helicopter) shadow was avoided and, in the specific case
of helicopter, instruments were deployed from an altitude high enough to avoid
perturbation at sea surface created by the rotor air flux. After excluding data with an
instrument tilt > 5°, the in-water profiles start beyond ca. 2 m when the radiometer was
deployed in free-falling mode from a ship, and beyond ca. 0.2 m when deployed from a
helicopter using a winch. To estimate upward irradiance just below the sea surface (at 0 -
m), Eu(z, λ) profiles were extrapolated up to surface by fitting the data to an exponential
function. Eu(0 - , λ) was propagated through water-air interface using a factor of 0.543
[Mueller et al., 2003] to obtained Eu(0 + , λ). The spectral remote sensing reflectance is
calculated as,
93

+
1 Eu
( 0 , λ)
R rs ( λ ) =
, (4a)
+
Q E ( 0 , λ)
where Q factor is the ratio between Eu and Lu. Here a Q value of 3.8 was used, which falls
within calculated range for turbid coastal waters [Loisel and Morel, 2001].
94
s
During CASES, the remote sensing reflectance at 13 wavebands (405, 412, 434,
442, 490, 510, 520, 532, 555, 590, 665, 683 and 700 nm), Rrs(λ), was calculated from the
vertical profile of Lu(z, λ), and Es(0 + , λ) measured with a SPMR and SMSR, respectively.
At each station, three to five SPMR casts were performed at least 50 m away from the
ship. The upwelling radiances were extrapolated to the sub-surface, Lu(0 - , λ), using a
linear fit to all ln[Lu(z, λ)] data points measured within a depth interval of 1-2 m within
the surface layer, and propagated through water-air interface to obtained Lw(0+, λ). Note
that both Eu(0 - , λ) and Lu(0 - , λ) were corrected for instrument self-shading following the
method proposed by Gordon and Ding [1992] as applied by Zibordi and Ferrari [1995]
using at-w(λ) measurements made with the ac-9 and the aw(λ) values published by Pope
and Fry [1997]. The spectral remote sensing reflectance is calculated as,
+
Lw(
0 , λ)
R rs ( λ ) = .
(4b)
+
E ( 0 , λ)
s
For each station, we averaged Rrs(λ) obtained from different casts (2 to 4) after
elimination of the spectra different from the mean value by more than 10% in the blue.
Out of 40 stations with SPMR measurements, 34 Rrs(λ) were retained for the algorithm
development (section 2.2). Note that four spectra were eliminated because the difference
among all replicates was > 10%, and two because of the surface waters stratification.
In shallow water areas or when surface waters were stratified (i.e. 14 stations), the
Rrs(λ) was measured above the sea surface using a hyperspectral radiometer (336 to 1062
nm with Δλ = 1.42 nm; Analytical Spectral Device, ASD). To make measurements away
from the ship shadow and bubble clouds, the instrument was mounted at the end of the
ship bow and deployed either on the port or starboard side depending of the sun position
and the sea state. The measurements were made only when incident irradiance conditions
were stable (clear or uniformly cloudy skies). To avoid calibration issues that generally
result from the use of several sensors, the measurements of the radiance (acceptance

angle of 10°) coming from sea surface (Lt), downwelling sky (Lsky), and a lambertian
reflector (grey Spectralon) (Lp) were performed successively, within a few minutes, using
an unique sensor. To obtain one Rrs spectrum, five to ten scans of each quantity were
quality checked and averaged. Note that the ASD automatically adjusts the integration
time of the measurements to optimize the signal-to-noise ratio (1.09 to 4.35 s for Lt, and <
1 s for Lsky and Lp). The above-water remote sensing reflectance is calculated as [Mobley,
1999]:
R
rs
( Lt
( λ)
− ρ sky ⋅ Lsky
( λ))
⋅ R p
( λ)
= , (5)
πL
( λ)
where ρsky is the air-water interface specular reflection coefficient for radiance, and Rp is
the albedo of the lambertian panel (88.3%). The value of Rp, spectrally constant in the
visible and near infrared parts of the spectrum, was determined using a calibrated
lambertian panel (P. Minnett, U. of Miami; calibration performed on 4 th of May 2004 by
Avian Technologies LLC inc.). As recommended by Mobley [1999], Lt was measured
with a viewing zenith angle of ~40° and a relative azimuth angle of ~135°. For this
viewing configuration and for a solar zenith angle >45°, i.e. the conditions encountered
during this study, ρsky depends only on wind speed (Figure 9 in Mobley, 1999). ρsky varies
from ~0.0256 to ~0.038 for wind speed of 0 and 15 m s -1 , respectively, under clear sky
condition [Mobley, 1999]. For cloudy sky condition, a value 0.0256 was adopted for ρsky.
An additional spectrally neutral correction was applied to Rrs(λ) based on the known
spectral shape of Rrs in the near infrared (NIR) part of the spectrum [Ruddick et al., 2006].
This correction aims to correct the Rrs spectra for uncertainty due to the air-sea reflection
correction (e.g. glitter) and the instrument-related errors (e.g. integration time, sky
radiance distribution) [for details see web appendix 2 in Ruddick et al., 2006]. The
correction assumes that the ratio of water leaving reflectance between two wavelengths in
the NIR is known (here 780 and 870 nm were used), allowing the computation of a
residual error that is then subtracted to Rrs(λ).
95
p

IV.2.2. Description of the [aCDOM/at] algorithm
To improve the quantification of photooxidation (Eq. 1), we need to know the
spectral contribution of aCDOM to the total light absorption coefficient of seawater. As
mentioned in the Introduction, because of the similar spectral absorption spectra of
CDOM and NAP, semi-analytical algorithms currently found in the literature fail to
discriminate them. Here we propose an empirical method to obtain the ratio [aCDOM/at] at
412 nm, directly from the Rrs spectrum. The [aCDOM/at](412) was regressed against
several combinations of Rrs and ratios of Rrs selected according to the following rationales:
in coastal waters, 1) reflectance at 412 nm is the most affected by CDOM absorption
compared with other channels, 2) reflectance at 490 nm is the most affected by
phytoplankton absorption, 3) variations in reflectance at 555 nm is most driven by light
scattering by particles, and 4) a major distinctive property of CDOM is that it does not
contribute to particle scattering. These considerations led us to propose the following
empirical algorithm for the retrieval of [aCDOM/at], based on multiple regression
conducted using measured [aCDOM/at](412) and Rrs(λ):
⎡a
⎢
⎣ at
⎤
⎛ R ( 412)
⎞ ⎛
rs
R ( 490)
⎞
rs
⎥(
412)
= α + β ⋅log10
⎜ + ⋅log10
+ ⋅log10
( rs ( 555)
)
Rrs
( 555)
⎟ χ ⎜
Rrs
( 555)
⎟ δ
, (6)
⎦
⎝ ⎠ ⎝ ⎠
CDOM R
where α, β, χ and δ are empirical coefficients. When CDOM increases, for instance, the
ratio Rrs(412)/Rrs(555) decreases and [aCDOM/at](412) tends to increase (negative slope for
β). In contrast, when phytoplankton pigments increase, the ratio Rrs(490)/Rrs(555)
decreases and [aCDOM/at](412) tends to decrease (positive slope for χ). Finally, when the
total amount of suspended particles increases, Rrs(555) increases and [aCDOM/at](412)
tends to decrease (negative slope for δ).
IV.3. Results and Discussion
IV.3.1. Variability of [aCDOM/at] and [aCDOM/aCDM] in coastal waters
Figure IV.2a illustrates the observed variability in the ratio [aCDOM/at] in the violet
part of the spectrum in various coastal environments including the Adriatic Sea, Baltic
Sea, English Channel, Gulf of Lyon, North Sea, and Beaufort Sea. Overall, [aCDOM/at] at
412 nm varies between 0.2 and 0.9, with an average of 0.53 and a standard deviation (σ)
96

of 0.16, which confirms the large variability of [aCDOM/at](412) in coastal waters. The
frequency distribution is slightly flat compared with a normal distribution (negative
kurtosis).
The contribution of CDOM to the total dissolved and detrital material (CDM) also
varies over a wide range in coastal waters, from 0.26 to 0.95 (Fig. IV.2b; Table IV.1),
even at regional scale (e.g. North Sea, 0.26 to 0.91). For comparison, Siegel et al. [2002]
reported a mean value for the ratio [aCDOM/aCDM] of 0.817 at 440 nm (with a σ of 0.13)
from a global collection of absorption measurements of aCDOM and aNAP. This is
significantly higher than most values found in coastal waters influenced by river runoff or
sediment resuspension. The GSM01 semi-analytical ocean color model [Garver and
Siegel, 1997; Maritorena et al., 2002] is currently implemented in the SeaWiFS data
processing chain to retrieve aCDM(440). The use of this algorithm is of limited utility for
quantifying CDOM photooxidation in coastal waters because it provides no information
either on the ratio [aCDOM/aCDM] nor [aCDOM/at]. Since coastal zones may be major sites
for terrigenous CDOM photooxidation [Kieber et al., 1990; Miller and Zepp, 1995;
Benner and Opsahl, 2001], the following method is specifically designed to estimate
[aCDOM/at] in turbid coastal waters.
Table IV.1. Relative contribution of CDOM to the colored detrital matter
(CDM=CDOM+NAP) and to the total absorption coefficients at 412 nm, respectively, for
the different regions covered by our data set.
[ a CDOM / aCDM
] (412) [ a CDOM / at
] (412)
Area N Min. Max. Average Min. Max. Average
± σ
± σ
Adriatic 29 .38 .87 .63 ± .12 .22 .66 .40 ± .09
Baltic 53 .60 .84 .76 ± .05 .38 .72 .60 ± .06
Eng. Channel 58 .30 .91 .72 ± .13 .24 .69 .50 ± .10
North Sea 68 .26 .91 .61 ± .16 .20 .81 .49 ± .13
Beaufort Sea 47 .41 .95 .82 ± .13 .39 .94 .75 ±.13
All 255 .26 .95 .71 ± .14 .20 .94 .55 ± .15
97

Figure IV.2. Frequency distributions of the contribution of CDOM to (a)
the total light absorption at 412 nm ([aCDOM/at]), and (b) to the colored
detrital material ([aCDOM/aCDM]). Note that the graph includes 54 additional
COASTlOOC or CASES stations where only IOP were available.
IV.3.2. Retrieval of [aCDOM/at] from remote sensing reflectance
Table IV.2 provides the multiple regression results for the empirical coefficients
of Eq. 6 and the determination coefficients (R 2 ) obtained for the whole dataset (global
algorithm), and separately for the five different regions of the data set (region-specific
98

algorithms). Figure IV.3 compares the [aCDOM/at] at 412 nm retrieved using the global
algorithm with the measured values.
Table IV.2. Empirical coefficients of Eq. 6 determined by multiple regression for the
different coastal environments.
Area N α β χ δ R 2
Adriatic 29 -.015 -.321 .691 -.223 .46
Baltic 53 .078 -.133 .674 -.280 .81
Eng. Channel 58 -.048 -.423 .539 -.204 .33
North Sea 68 -.480 -.255 .526 -.483 .65
Beaufort Sea 47 -.514 -.546 .480 -.454 .42
All 255 -.387 -.387 .577 -.390 .70
Figure IV.3. Comparison between retrieved and measured [aCDOM/at] at
412 nm for the COASTlOOC and CASES datasets. [aCDOM/at](412) was
calculated using equation 6 with coefficients obtained using the whole
data set (N=255; Table IV.2).
To evaluate rigorously the ability of our global algorithm to retrieve
[aCDOM/at](412) in a given region, we conducted the following validation exercise: the
empirical coefficients of Eq. 6 for a given region were also derived using the data from
the four other regions. The algorithm could thereby be tested with a data set different
99

from that used for algorithm tuning. For the Beaufort Sea, for example, the coefficients of
Eq. 6 were derived using the whole COASTlOOC data set (i.e. Adriatic Sea, Baltic Sea,
English Channel and North Sea). Note that these coefficients are not provided as they
were derived for validation purpose only. Table IV.3 provides the results from a
comparison between retrieved and measured [aCDOM/at] for each region, achieved using
Type-II regression. The algorithm captures part of the variability within each region as
the regression slopes are all significantly positive, (p < 0.05), although they are
significantly < 1 (p < 0.05), except for the English Channel. This result indicates that
further tuning is necessary to get the variability of [aCDOM/at](412) right within a given
region. Nevertheless, the overall absolute uncertainty of the global algorithm with the
coefficient derived independently is ±0.18 at the 95% confidence interval (Table IV.3).
Despite the relatively high heterogeneity among the five coastal areas considered here,
the algorithm is able to retrieve the average [aCDOM/at](412) value for each region
(Δ[aCDOM/at]±0.04), except for the Beaufort Sea where the estimation is slightly biased (-
0.078).
Table IV.3. Performance of Eq. 6 for the retrieval of [aCDOM/at] at 412 nm. For each
region, independent data set is used to obtain the empirical coefficients for Eq. 6.
Area N Intercept a
Slope a
Δ[aCDOM/at] b CI 95%c
Adriatic 29 .22 [.11, .31] .44 [.21, .71] 0.001 ±0.13
Baltic 53 .28 [.20, .34] .54 [.43, .67] 0.000 ±0.07
Eng. Channel 58 .00 [-.30,.17] 1.06 [.69, 1.64] 0.039 ±0.20
North Sea 68 .27[.22, .32] .45 [.36, .54] 0.013 ±0.17
Beaufort Sea 47 .30 [.13, .46] .56 [.36, .80] -.078 ±0.20
All d 255 .16 [.12, .20] .70 [.63, .76] -.002 ±0.18
a
Numbers in brackets are for the range of 95 % Confidence Interval computed for a Type II
regression.
b
[ ] [ ] [ ] measured
1 n
retrieved
.
Δ a CDOM / at
= ∑ aCDOM
/ at
− aCDOM
/ at
n
i
a CDOM / at
retrieval.
d
These results were obtained by comparing the [aCDOM/at] values retrieved within each region
using the global algorithm derived independently, with the measured values (see text).
c 95% Confidence Interval of the [ ]
In an attempt to understand the above results, we further analyzed the absolute
difference between the retrieved and measured [aCDOM/at](412) values (Δ[aCDOM/at]).
Again for each region, [aCDOM/at](412) was retrieved using Eq. 6 with the coefficients
derived independently. Figure IV.4a compares Δ[aCDOM/at] with the ratio between the
100

particulate absorption at 412 nm and the remote sensing reflectance at 555 nm
([ap(412)/Rrs(555)]). Positive Δ[aCDOM/at] values indicate an overestimation of
[aCDOM/at](412) by the empirical algorithm. Low values of [ap(412)/Rrs(555)] indicate
that the particles are weakly absorbing blue photons relative to the their scattering
properties in the green, while high values indicate the contrary. The former may indicate
the presence of mineral particles, and the latter the presence of phytoplankton and/or
detritus [Gallegos and Neale, 2002]. Since the absolute value of Rrs(555) is used in the
algorithm as an indicator of the total amount of particles, variability in the
[ap(412)/Rrs(555)] will affect the performance of the algorithm. Statistically significant
positive slopes of the regression (Type II) between Δ[aCDOM/at] and the logarithm of
[ap(412)/Rrs(555)] were observed in the Adriatic Sea, Baltic Sea, English Channel and the
North Sea (Table IV.4). These results suggest that the [aCDOM/at](412) value is, as argued
above, underestimated in presence of weakly absorbing particles, while it is
overestimated in presence of highly absorbing ones. This is because the enhance of
Rrs(555) in presence of weakly absorbing particles results, for instance, in an
overestimation of the contribution of particles to the total light absorption at 412 nm.
Based on these observations, one can ask why in the Baltic Sea, where particles are
predominantly organic (87%; Babin et al. [2003b]) and highly absorbing (Fig. IV.4a), the
[aCDOM/at](412) is not systematically overestimated? The answer can be found in Fig.
IV.4b.
Table VI.4. Coefficient of determination of the linear regression between Δ[aCDOM/at]
(shown in Fig. IV.4) and 1) the logarithm of the ratio [ap(412)/Rrs(555)] and 2) the ratio
[ a CDOM / aCDM
] at 412 nm. NS = the slope is not significantly different from 0 at 95%
confidence level (Type II regression).
a / (412) a
Area N [ap(412)/Rrs(555)] a [ ]
101
CDOM aCDM
Adriatic 29 .23 (NS) .18 (.17)
Baltic 53 .51 (NS) .34 (NS)
Eng. Channel 58 .46 (.14) NS (.23)
North Sea 68 .46 (.33) .58 (.25)
Beaufort Sea 47 NS (.21) .22 (.27)
All 255 .15 (.08) .21 (.14)
a Values in parenthesis are for the regionally tuned algorithm, while the bold numbers indicate the
parameter that explains most the error for a given region.

Figure IV.4b plots the Δ[aCDOM/at] as a function of the ratio between aCDOM to
aCDM at 412 nm. Statistically significant negative slopes of the linear regression (Type II)
between Δ[aCDOM/at] and [aCDOM/aCDM](412) were observed in the Adriatic Sea, Baltic
Sea, North Sea and Beaufort Sea (Table IV.4). Altogether, the R 2 of the regression
reaches 21%. Generally, when NAP dominates the CDM the empirical algorithm tends to
overestimate the contribution CDOM to the total light absorption. This is because the
ratio of Rrs(412)/Rrs(555) used in the algorithm is an indicator of the total CDM rather
than just the CDOM. The fact that relatively high [aCDOM/aCDM](412) is found in the
Baltic Sea partly explains why we do not observed a systematic overestimation of
[aCDOM/at](412). In the Beaufort Sea, the systematic underestimation of [aCDOM/at](412)
by the empirical algorithm (-0.078; Table IV.3) is probably due to the dominance of
CDOM to CDM (Table IV.1).
102

Figure IV.4. Error analysis of [aCDOM/at] at 412, calculated using regionindependent
coefficients, as a function of: a) the ratio of ap(412) to
Rrs(555); b) the relative contribution of CDOM to the total CDM
absorption coefficient at 412 nm. Only the statistically significant
relationships are shown. The color coding is the same as Fig. IV.3.
The trends described above can be attenuated if the algorithm is tuned with the
region-specific data. Figure IV.5 compares the retrieved and measured [aCDOM/at](412)
values when the coefficients of Eq. 6 are derived for each region individually (Table
IV.2). For the Baltic Sea and North Sea, the regional tuning leads to R 2 > 0.65. In the
103

North Sea, in particular, where the composition of CDM is highly variable (%CDOM =
26 to 91%; Table IV.1), the algorithm is able to discriminate the CDOM from the total
absorption within a CI 95% of 0.13. In the Baltic Sea, where the optical properties of
particles are relatively homogeneous (Fig. IV.4), the algorithm performance is excellent
(R 2 = 0.81, CI 95% = ±0.05). For the English Channel, in contrast, the algorithm poorly
performs as indicated by the low R 2 of 0.33. For the Adriatic Sea and Beaufort Sea,
because most of the measured [aCDOM/at](412) values felt within a narrow range, it is
difficult to draw clear conclusions from this analysis. The absolute uncertainty of the
region-specific algorithm is ±0.14 at the 95% confidence interval.
Figure IV.5. Same as Fig. IV.3 but for region specific coefficients for Eq.
6 (Table IV.2).
In general, an algorithm based on the ratios Rrs(412)/Rrs(555) and
Rrs(490)/Rrs(555), and on Rrs(555) seems suitable to discriminate aCDOM from at at 412 nm
with good accuracy. We also applied this combination to the synthetic data set developed
by the International Ocean Color Coordinating Group to test and compare algorithms
(available at: http://www.ioccg.org/groups/OCAG_data.html). The performance of the
algorithm was similar to that found with our in situ data set (R 2 =0.65, CI 95% ±0.15), but
the β and χ coefficients of Eq. 6 were quite different (-.385, -1.105, 1.33 and -.342 for
104

α, β, χ and δ respectively). This difference may be ascribed to the wide range of IOPs
covered by the synthetic data set, and in particular to the highly variable ratio between the
scattering and absorption coefficients of NAP (more than one order of magnitude). In
addition, the synthetic data set includes clear waters for which the pure water absorption
is no longer negligible at 412 nm (up to 33% of at). This fourth optically active
component complicates the estimation of [aCDOM/at](412) with the proposed algorithm.
Our method was also tested on the NASA bio-Optical Marine Algorithm Data set
(NOMAD) [Werdell and Bailey, 2005]. The algorithm did not perform as well as on our
in situ data set or on the IOCCG data set (R 2 =0.30, CI 95% ±0.22). Two possible
explanations for this poor result is 1) the relatively low influence of terrestrial inputs in
most of the NOMAD dataset compare to our dataset, and 2) the lack of consistency
among the different sub-datasets that form the NOMAD dataset, in terms of methods to
measure the absorption coefficients adopted by different investigator. For some specific
sub-datasets found in NOMAD, however, results similar to the one presented above were
obtained (e.g., R 2 of 0.81 and CI 95% of 0.14 for the ONR-Chesapeake program). These
results suggest that our algorithm works properly only in coastal waters influenced by
terrestrial inputs, and that more data measured following an unique protocols, for the
absorption in particular, are necessary to validate the algorithm.
3.3. Sensitivity of the depth-integrated photooxidation model to errors on [aCDOM/at](λ)
In this section, we assess the relative error made on the depth-integrated
production of dissolved inorganic carbon (PDIC) resulting from the error on the
[aCDOM/at](412) estimate, and from the uncertainty in its extrapolation over the 300-600
nm range. As mentioned in the Introduction, DIC is the main product of CDOM
photooxidation. For our purpose, the sub-surface spectral irradiance, Ed (0 - , λ), was
calculated using the Tropospheric Ultraviolet Visible (TUV) model [Madronich and
Flocke, 1999] for the summer solstice, at latitude of 70°N, under clear sky, moderate
aerosols concentration (aerosol optical thickness at 550 nm of 0.1), and total ozone
column content of 330 DU. Because the magnitude and spectral shape of apparent
quantum yield for DIC photoproduction vary widely between coastal and open ocean
[Johannessen and Miller, 2001; Bélanger et al., 2006], the calculations were made for
105

two contrasting AQY spectra measured in the Beaufort Sea (Fig. IV.6). The two AQY
spectra show marked differences in shape, with the relative contribution of longer
wavelengths being more important for AQY1.
Wavelength (nm)
Figure IV.6. Apparent Quantum Yield spectra for DIC production
determined for two stations located in the southeastern Beaufort Sea.
AQY1 was determined on surface waters influenced the by the Mackenzie
River (salinity = 8.2; Station R5a), while AQY2 was determined on
surface waters collected in the Amundsen Gulf, away from direct riverine
influence (salinity = 30.0; Station 108) [Bélanger et al., 2006].
Figure IV.7 shows the spectral PDIC(λ) calculated using Eq. 1 with different
values of [aCDOM/at](412). While the maximum PDIC occurs in the spectral range from
325 to 335 nm for both AQYs, the relative contribution of the visible radiation to DIC
photoproduction is different. The decrease in [aCDOM/at](412) has a greater impact on
PDIC(λ) when using AQY1 than when using AQY2 (Fig. IV.7). This is because
[aCDOM/at](λ) variability is more pronounced in the visible relative to the UV. So, the
magnitude of the error on PDIC resulting from the uncertainty on [aCDOM/at](412), depends
on the spectral shape of the AQY. To illustrate that, Fig. IV.8 shows the relative error in
spectrally integrated DIC photoproduction (ΔPDIC=100*[PDIC-PDIC true ]/ PDIC true ) as a
106

function of Δ[aCDOM/at](412). If the true [aCDOM/at](412) value of 0.7 is overestimated (or
underestimated) by 0.18, the overall uncertainty of our algorithm, the ΔPDIC reaches ~±24
and ~±18% for AQY1 and AQY2, respectively. When the initial [aCDOM/at](412) value is
0.3, ΔPDIC can reach as much as ~±40%. Overall, PDIC can be predicted with an
uncertainty of 50% (Fig. IV.8). Note that ignoring the impact of particles by using
constant [aCDOM/at](λ) values of one, e.g. as in Johannessen [2000], can lead to an error
in PDIC of up to 3-fold.
Wavelength (nm)
Figure IV.7. Spectral production of DIC for different values of
[aCDOM/at](412) and for A) AQY1 and B) AQY2.
107

Figure IV.8. Relative difference in DIC production as a function of the
absolute difference between the retrieved and true [aCDOM/at](412) values
for the two AQYs and two initial values for [aCDOM/at] true (412) of 0.3 and
0.7. The thick line at the bottom illustrate the 95% confidence interval in
the [aCDOM/at] retrieval using Eq 6.
The above error analysis assumes that a perfect spectral extrapolation of [aCDOM/at]
from 600 nm to the UV domain is achieved. Here we conduct a sensitivity study of
spectral extrapolation of [aCDOM/at] on the depth-integrated DIC photoproduction. The
extrapolation is achieved as follows:
where
and
a
CDOM
a
t
aCDOM
( λ)
( λ)
= , (7)
a ( λ)
+ a ( λ)
+ a ( λ)
CDOM
p
{ [ a / a ] ( 412)
* a ( 412)
} exp[
S * ( 412 ) ]
a ( λ) − λ
(7a)
CDOM
= CDOM t
t
CDOM
N
ap(λ) = [ at(412) − aCDOM (412) − aw(412) ]* ap (λ). (7b)
In Eq. 7, aw(λ) is the pure water absorption spectrum, SCDOM is the slope of the aCDOM
N
spectrum, and ap (λ) is the particulate absorption spectra normalized at 412 nm
N
( ap (λ) = ap (λ) /a p(412)). As recommended by Fry [2000], aw(λ) values were taken
108
w

from Quickenden and Irvin [1980] for 300 nm ≤ λ ≤ 320 nm, Pope and Fry [1997] for
380 nm ≤ λ ≤ 600 nm, and were extrapolated for the λ between those ranges. The
N
sensitivity study is conducted for various values for SCDOM, ap (λ) and at(412). These
parameters are required, in addition to [aCDOM/at](412), to operate Eq. 7.
Figure IV.9 shows the variability SCDOM as a function of aCDOM(412) in the
Beaufort Sea. SCDOM ranged from 0.0177 to 0.0303 nm -1 with an average value of 0.022
nm -1 , and a standard deviation of 0.002 nm -1 . The SCDOM values observed over the
Mackenzie shelf (~0.018 nm -1 ) were within the range reported by Guéguen et al. [2005]
for this area, and similar to others reported for coastal environments [e.g., Blough and
Del Vecchio, 2002; Babin et al., 2003b]. Offshore, near the Arctic packed ice, SCDOM
reached 0.025 to 0.030 nm -1 , which is higher than the values previously reported for the
Canada Basin [Guéguen et al., 2005], but within the upper range found in the literature
for oligotrophic seawaters [Blough and Del Vecchio, 2002]. Nevertheless, the high SCDOM
may also be due the freezing of our sample during the storage, which is known to affect
to some extent the CDOM absorption spectra (C. Belzile, personal communication, 2005).
Furthermore, when aCDOM is low, the spectral range choosen (e.g. 300 to 500 nm) to
calculate SCDOM using a single exponential greatly affects the results [see Twardowski et
al., 2004]. Interestingly, the relationship we observed between SCDOM and aCDOM(412) is
very similar to the one reported for the Baltic Sea by Kowalczuk et al. [2006] (see their
Figure 4). The sensitivity analysis of the UV extrapolation is conducted using 0.018 and
0.026 nm -1 as the lower and upper values (corresponding to the 95% confidence interval).
Figure IV.10 shows the 99 ap(λ) spectra normalized to the value at 412 nm
measured in the surface waters during the CASES field campaign. Except for a few
number of spectra where the presence of phytoplankton pigments was dominant (with a
ap(λ) peak at ~440 nm), most spectra show an exponential increase toward shorter
wavelengths, which indicates the dominance of the aNAP in the total particulate matter
N
absorption (62 ± 18% at 443 nm). The average ap (λ) values ± 1.96*σ (i.e. a 95%
confidence interval) was chosen to represent the limits of the ap spectrum for low and
high spectral dependency, respectively.
109

Figure IV.9. Variability of SCDOM as a function of the aCDOM(412) value
measured in the surface waters of the southeastern Beaufort Sea during
summer 2004 (n=65). A second-order polynomial fit between the SCDOM
values and the logarithm of aCDOM(412) provided a strong description of
the observations (r 2 =0.79; n=65):
2
S CDOM = 0. 0186 − 0.
0019log[
aCDOM
( 412)
] + 0.
0031log[
aCDOM
( 412)]
.
Figure IV.10. Spectral absorption by particulate matter normalized to 412
nm value measured in the surface waters of the southeastern Beaufort Sea
during summer 2004 (thin grey lines). The thick black line represents the
N
averaged ap (λ) values. The sensitivity the PDIC (eq 1) to the extrapolation
to the UV domain is explored within the limits given by 95% probability
of the normal distribution (i.e. ±1.96*σ; dashed lines). The ap(λ) values
for λ

If we assume that the pure water absorption coefficient is negligible over the 300
to 600 nm range, the spectral extrapolation of [aCDOM/at](412) can be achieve without the
N
absolute value of at(412) by using SCDOM and ap (λ) only. In the Beaufort Sea, the total
absorption coefficient at 412 nm measured in the surface waters varied between 0.048
and 2.7 m -1 with an averaged value of 0.41 m -1 and a standard deviation of 0.61 m -1 . So,
these values are used to examine the impact of at(412) on the spectral extrapolation and
PDIC. The null aw(λ) assumption can lead to error > 25% when at(412) is low and the
[aCDOM/at](412) is high (Table IV.5). In most situations, however, the error on PDIC made
using the null aw(λ) assumption is < 10%. In turbid coastal waters where at(412) is larger
than ~0.2 m -1 , aw(λ) could be neglected without losing to much accuracy in PDIC.
N
The variability in the SCDOM and ap (λ) can also lead to some error in PDIC. For
extreme cases, i.e. when [aCDOM/at](412) is low and the spectral dependency of both
aCDOM(λ) and ap(λ) are opposite, the ΔPDIC reaches ~10% and ~18% for AQY1 and AQY2
respectively (Table IV.6). Figure IV.11 shows several examples of PDIC(λ) spectra
calculated for relatively high proportions of particulate matter to the total absorption, i.e.
[aCDOM/at](412) of 0.3. Differences in the PDIC(λ) spectra are notable for both AQYs
(note the different y-axis scales). As expected, the sensitivity of PDIC to the spectral
extrapolation decreases when the relative contribution of the CDOM to the total
absorption increases. In fact, for a [aCDOM/at](412) value of 0.7, ΔPDIC is

Table IV.5. ΔPDIC ( in %), due to error in the spectral extrapolation of [aCDOM/at](412) to
the 300-600 nm range resulting from variation in the magnitude of at(412). PDIC true is
N
calculated with averaged SCDOM , ap (λ) , and for different at(412) values. PDIC is
calculated assuming aw(λ) is null.
AQY1 AQY2
[aCDOM/at] [aCDOM/at] [aCDOM/at] [aCDOM/at]
= 0.3 = 0.7 = 0.3 = 0.7
at(412) = 2.7 m -1 1.3 3.4 0.4 0.8
at(412) = 0.41 m -1 at(412) = 0.048 m
4.7 10.7 1.5 2.6
-1 14.5 26.4 6.0 8.1
Table IV.6. ΔPDIC ( in %), due to error in the spectral extrapolation of [aCDOM/at](412) to
N
the 300-600 nm range resulting from variation in SCDOM and ap (λ) spectra. PDIC true is
N
calculated with averaged SCDOM , ap (λ), and at(412).
[aCDOM/at]
= 0.3
112
AQY1 AQY2
[aCDOM/at]
= 0.7
[aCDOM/at]
= 0.3
[aCDOM/at]
= 0.7
N
ap (λ) ; SCDOM = 0.018 nm -1 -2.0 4.1 -8.2 -0.9
N
ap (λ) ; SCDOM = 0.026 nm -1 3.2 -3.4 8.1 0.6
N
ap (λ) - 1.96σ ; SCDOM = 0.022 nm -1 7.2 -0.3 11.3 1.9
N
ap (λ) + 1.96σ ; SCDOM = 0.022 nm -1 -3.8 0.9 -8.0 -1.5
N
ap (λ) - 1.96σ ; SCDOM = 0.018 nm -1 4.9 4.1 3.4 1.5
N
ap (λ) + 1.96σ ; SCDOM = 0.018 nm -1 -5.0 4.8 15.9 -2.9
N
ap (λ) - 1.96σ ; SCDOM = 0.026 nm -1 10.2 -3.8 18.7 -2.0
N
ap (λ) + 1.96σ ; SCDOM = 0.026 nm -1 -1.0 -2.3 0.3 -0.4

Figure IV.11. Example of spectral production of DIC for
[aCDOM/at] true (412) value of 0.3 for various spectral shape for aCDOM(λ) and
ap(λ) and for AQY1 (a-c) and AQY2 (d-f). The grey curves on the bottom
N
panels (b-c; e-f) is PDIC(λ) computed using averaged SCDOM and ap (λ).
IV.3.4. Application to SeaWiFS imagery
We applied our algorithm to a full resolution SeaWiFS image of the southeastern
Beaufort Sea acquired on June 21 st 1998 (day 172). On that year, sea ice cover was
exceptionally reduced in early spring over the study area, which allowed good ocean
color observations over the whole Mackenzie Shelf and a large portion of the Canada
Basin. The application of a turbid-water detection algorithm [Morel and Bélanger, 2006]
revealed that almost two thirds of the ice-free waters were turbid. Because the standard
113

atmospheric correction, which is based on invalid assumptions of null water-leaving
radiance in the NIR bands (765 and 865 nm) [Gordon and Wang, 1994], generally gives
rise to negative water leaving radiances in the blue, the scene was processed using the
turbid-water atmospheric correction algorithm described by Ruddick et al. [2000]. The
[aCDOM/at](412), calculated using Eq. 6 with region-specific coefficients (Table 2), shows
coherent patterns over the continental shelf and beyond (Fig. IV.12). First, [aCDOM/at](412)
values increase from the river mouth to the border of the continental shelf (~0.40 to
~0.90). This reflects the decreasing contribution of the terrigenous particles to the total
light absorption in the surface waters when the particles gradually settle over the shelf.
Interestingly, the highest values (> 0.85) are found beyond the continental shelf in waters
influence by the Mackenzie runoff. Further offshore in the Canada Basin and in the
Amundsen Gulf, the [aCDOM/at](412) is relatively uniform, with values between 0.7 and
0.8. These lower offshore values probably reflect the absence of direct influence of river
runoff. The values observed there fall within the range observed in those areas during
CASES.
Figure IV.12. Spatial distribution of the [aCDOM/at](412) as derived from
SeaWiFS data acquired the 21 st of June 1998. The grey color code
represents sea ice.
Figure IV.13 shows the spatial distribution of the depth-integrated DIC
photoproduction rate (in mg C m -2 d -1 ) for both AQYs, as calculated using the satellite-
derived [aCDOM/at](412) values (Fig. IV.12). The spectral of [aCDOM/at](412) was
114

achieved with the averaged spectral shape of aCDOM and at, and using the total absorption
at 412 derived from the Rrs spectrum using a modified version of the quasi-analytical
algorithm proposed by Lee et al. [2002] (see Annex A1). The daily spectral irradiance
−
just below the sea surface, E ( 0 , λ)
, was obtained using the TUV model with the
d
ozone observed by TOMS and aerosols concentrations derived from SeaWiFS data. The
spatial distribution of the PDIC is similar to that of [aCDOM/at](412) with the highest
values of ~10-12 and ~3.6-3.8 mg C m -2 d -1 for AQY1 and AQY2, respectively, observed
at the border of the continental shelf. The spatial variability in PDIC is slightly higher for
AQY1 (~2.5-fold) than for AQY2 (~2-fold). These results confirm that the spatial
variability in PDIC resulting from the optical properties of the water column alone is
important.
115

Figure IV.13. Spatial distribution of depth-integrated DIC
photoproduction as calculated using both AQYs and the satellite-derived
[aCDOM/at](412) as shown in Fig. IV.12.
IV.4. Summary and conclusions
This study was motivated by the need to improve the quantification of the
depth-integrated DIC photoproduction in the Arctic coastal waters where a rapid decrease
in the summer sea ice cover is observed, and expected to culminate with totally open
waters during September before the end of this century [ACIA, 2005]. This major change
will expose river runoff to light, including UV radiations [Bélanger et al., 2006]. Depth-
integrated DIC photoproduction was assessed here using a simple spectral model that
116

needs as input the ratio between the CDOM and total absorption coefficients ([aCDOM/at]).
To account for the important variability in this ratio in coastal waters, an empirical
algorithm was proposed to estimate the [aCDOM/at] ratio at 412 nm from the remote
sensing reflectance spectrum. The algorithm was developed and validated using an
extensive data set of in situ measurements of spectral absorption coefficients and
reflectance made in contrasting coastal waters. The absolute uncertainty of the algorithm
was found to be ±0.18 (±0.14 for regional tuning). This uncertainty in [aCDOM/at](412)
results in a potential error on the depth-integrated photoproduction of DIC

aspects that need more attention for future semi-analytical algorithms to be successful in
discriminating aCDOM from the total absorption coefficient in coastal waters: the IOP
model formulation and the inversion approach.
While the fact that NAP scatters light and CDOM do not is a source of ambiguity
when pooling CDOM and NAP, it provides discrimination power when CDOM and NAP
are distinguished in an IOP model. So, semi-analytical algorithms need to be based on an
IOP model that includes a good link between aNAP and bNAP (or bbNAP). Gallegos and
Neale [2002] successfully applied such an approach to deconvolve the absorption
contribution by phytoplankton, CDOM and NAP in total absorption spectra measured in
situ. But, obviously, the inversion of remotely sensed reflectance is more challenging
than inverting absorption spectra measured in situ.
Again, the fact that our empirical algorithm succeeded in discriminating aCDOM
from aNAP means that the necessary information is contained in the reflectance spectrum.
This information, however, must be extracted properly, and this is probably one
weakness of current semi-analytical models. In coastal waters, variations in reflectance
are primarily due to variations in turbidity [Sathyendranath et al., 1989]. Subtle changes
in the shape of the reflectance spectrum that may result from changes in the proportions
of phytoplankton, CDOM and NAP, are overwhelmed by the change in the magnitude of
reflectance at all wavelengths due to variations in turbidity. So, when some specific
information is needed from the reflectance spectrum, such as the ratio of aCDOM to at,
some specific spectral regions provide less ambiguous information. The rationales
considered in the development of our empirical algorithm are valid for a semi-analytical
algorithm. The latter must go beyond finding the best fit between calculated and observed
reflectance spectra, at all wavelengths [e.g., Roesler and Perry, 1995; Lee et al., 1996;
Garver and Siegel, 1997]. This procedure must not be blurred by the large changes in
reflectance magnitude due to variations in turbidity, and be misled by reflectance at
wavelengths where the information is ambiguous.
The simple method proposed in this study represents, to our knowledge, a first
step toward the integration of Ocean Color information to quantify of CDOM
photooxidation in coastal waters. Further research studies in that field needs to be
conducted co-jointly with specialists of both optics/remote sensing and photochemistry
118

sciences in order to document in parallel AOPs, IOPs and CDOM photoreactivity (i.e.
AQY) in coastal waters. The latter is critical for the quantification of CDOM
photooxidation and is definitely the least documented at the moment [Johannessen and
Miller, 2001; Mopper and Keiber, 2002; Bélanger et al., 2006].
119

IV.5. Notations
Parameter Description
λ Light wavelength (nm)
a m (λ) Measured absorption coefficient with the ac-9 (m -1 )
ax(λ) Absorption coefficient for constituent x (m -1 ),
where subscript x is either t (total), w (water), CDOM (Chromophoric Dissolved Organic
Matter), p (particles), NAP (Non-algal particles),CDM (Colored Dissolved and Detritic
Material) or φ (phytoplankton).
b(λ) Total scattering coefficient (m -1 )
b m (λ) Measured scattering coefficient with the ac-9 (m -1 ), i.e. c(λ)-a m (λ).
bp(λ) Particulate scattering coefficient (m -1 )
bb(λ) Total backscattering coefficient (m -1 )
bbp(λ) Particulate backscattering coefficient (m -1 )
bbw(λ) Water backscattering coefficient (m -1 )
c(λ) Beam attenuation (m -1 )
Es(λ) Irradiance just above the sea surface (W m -2 nm -1 )
Ed(λ,z) Downwelling irradiance at depth z (W m -2 nm -1 or mol photon m -2 nm -1 )
Lu(λ,z) Upwelling radiance at depth z (W m -2 nm -1 sr -1 )
Lw(λ) Water-leaving radiance (W m -2 nm -1 sr -1 )
Rrs(λ) Remote sensing reflectance above the sea surface (nm -1 sr -1 )
AQY (λ) Apparent quantum yield for dissolved inorganic carbon (mol C nm -1 (mol
PDIC
120
photon) -1 )
Depth-integrated photochemical production of dissolved inorganic carbon
(mol C m -2 s -1 )

IV.6. References
ACIA (2005), Arctic Climate Impact Assessment, 1042 pp., Cambridge Press University.
Arrigo, K. R., D. Lubin, G. L. van Dijken, O. Holm-Hansen, and E. Morrow (2003),
Impact of a deep ozone hole on Southern Ocean primary production, J. Geophys.
Res., 108(C5), doi:1029/2001JC001226.
Babin, M., A. Morel, V. Fournier-Sicre, F. Fell, and D. Stramski (2003a), Light
Scattering properties of marine particles in coastal and open ocean waters as
related to the particle mass concentration, Limnol. Oceanogr., 48(2), 843-859.
Babin, M., D. Stramski, G. M. Ferrari, H. Claustre, A. Bricaud, G. Obolensky, and N.
Hoepffner (2003b), Variations in the light absorption coefficients of
phytoplankton, nonalgal particles, and dissolved organic matter in coastal waters
around Europe, J. Geophys. Res., 108(C7), 3211, doi: 10.1029/2001JC000882.
Bélanger, S., H. Xie, N. Krotkov, P. Larouche, W. F. Vincent, and M. Babin (2006),
Photomineralization of terrigenous dissolved organic matter in Arctic coastal
waters from 1979 to 2003: Interannual variability and implications of climate
change, Global Biogeochem. Cycles, in press, doi:10.1029/2006GB002708.
Benner, R., and S. Opsahl (2001), Molecular indicators of the sources and
transformations of dissolved organic matter in the Mississippi river plume, Org.
Geochem., 32(4), 597-611.
Blough, N. V., and R. Del Vecchio (2002), Chromophoric DOM in the coastal
environment, in Biogeochemistry of Marine Dissolved Organic Matter, edited by
D. A. Hansell, and C. A. Carlson, pp. 509-546, Academic Press, San Diego.
Bricaud, A., A. Morel, and L. Prieur (1981), Absorption by dissolved organic matter of
the sea (yellow substance) in the UV and visible domains, Limnol. Oceanogr.,
26(1), 43-53.
Carlson, C. A. (2002), Production and Removal Processes, in Biogeochemistry of Marine
Dissolved Organic Matter, edited by D. A. Hansell, and C. A. Carlson, pp. 91-
151, Academic Press, San Diego.
Dagg, M., R. Benner, S. Lohrenz, and D. Lawrence (2004), Transformation of dissolved
and particulate materials on continental shelves influenced by large rivers: plume
processes, Cont. Shelf Res., 24, 833-858.
Fry, E. S. (2000), Visible and near-ultraviolet absorption spectrum of liquid water:
comment, Appl. Opt., 39(16), 2743-2744.
Gallegos, C. L., and P. J. Neale (2002), Partitioning spectral absorption in case 2 waters:
discrimination of dissolved and particulates components, Appl. Opt., 41(21),
4220-4233.
121

Garver, D. L., and D. A. Siegel (1997), Inherent optical property inversion of ocean color
spectra and its biogeochemical interpretation, 1. Time series from the Sargasso
Sea, J. Geophys. Res., 102(C8), 18607-18625.
Guéguen, C., L. Guo, and N. Tanaka (2005), Distributions and characteristics of colored
dissolved organic matter in the Western Arctic Ocean, Cont. Shelf Res., 25(10),
1195-1207.
Gordon, H. R. (1992), Diffuse reflectance of the ocean: influence of nonuniform
phytoplankton pigment profile, Appl. Opt., 31(12), 2116-2129.
Gordon, H. R., and K. Ding (1992), Self-shading of in-water optical instruments, Limnol.
Oceanogr., 37(3), 491-500.
Gordon, H. R., and M. Wang (1994), Retrieval of water-leaving radiance and aerosol
optical thickness over the oceans with SeaWiFS: a preliminary algorithm, Appl.
Opt., 33(3), 443-452.
Hansell, D. A., and C. A. Carlson (1998), Deep-ocean gradients in the concentration of
dissolved organic carbon, Nature, 395(6699), 263-266.
IOCCG (2006), Remote sensing of inherent optical properties: Fundamentals, tests of
algorithms, and applications, edited by Z.-P. Lee, pp. 126, Reports of the
International Ocean-Colour Coordinating Group, No. 5, IOCCG, Dartmouth,
Canada.
Johannessen, S. C. (2000), A photochemical sink for dissolved organic carbon in the
ocean, Ph.D. thesis, 176 pp., Dalhousie University, Halifax, N. S., Canada.
Johannessen, S. C., and W. L. Miller (2001), Quantum yield for the photochemical
production of dissolved inorganic carbon in seawater, Mar. Chem., 76, 271-283.
Kieber, R. J., X. Zhou, and K. Mopper (1990), Formation of carbonyl compounds from
UV-induced photodegradation of humic substances in natural waters: Fate of
riverine carbon in the sea, Limnol. Oceanogr., 35(7), 1503-1515.
Kishino, M., M. Takahashi, N. Okami, and S. Ichimura (1985), Estimation of the spectral
absorption coefficients of phytoplankton in the sea, Bull. Mar. Sc., 37(2), 634-642.
Laurion, I., F. Blouin, and S. Roy (2003), The quantitative filter technique for measuring
phytoplankton absorption : Interference by MAAs in the UV waveband, Limnol.
Oceanogr.: Methods, 1(1), 1-9.
Lee, Z.-P., K. L. Carder, and R. A. Arnone (2002), Deriving inherent optical properties
from water color : a multiband quasi-analytical algorithm for optically deep
waters, Appl. Opt., 41(27), 5755-5772.
Lee, Z. P., K. L. Carder, T. G. Peacock, C. O. Davis, and J. L. Mueller (1996), Method to
derive ocean absorption coefficients from remote-sensing reflectance, Appl. Opt.,
35(3), 453-462.
Loisel, H., and A. Morel (2001), Non-isotropy of the upward radiance field in typical
coastal (Case 2) waters, Int. J. Remote Sens., 22(2&3), 275-295.
122

Macdonald, R. W., S. M. Solomon, R. E. Cranston, H. E. Welsh, M. B. Yunker, and C.
Gobeil (1998), A sediment and organic carbon budget for the Canadian Beaufort
Shelf, Mar. Geol., 144, 255-273.
Madronich, S., and S. Flocke (1999), The role of solar radiation in atmospheric chemistry,
in Handbook of Environmental Chemistry, edited by P. Boule, pp. 1-26, Springer-
Verlag, Heidelberg, Germany.
Maritorena, S., D. A. Siegel, and A. R. Peterson (2002), Optimization of a semianalitical
ocean color model for global-scale applications, Appl. Opt., 41(15), 2705-2714.
Miller, W. L., and R. G. Zepp (1995), Photochemical production of dissolved inorganic
carbon from terrestrial organic matter : Significance to the oceanic carbon cycle,
Geophys. Res. Let., 22(4), 417-420.
Mobley, C. D. (1999), Estimation of the remote-sensing reflectances from above-water
measurements, Appl. Opt., 38(36), 7442-7455.
Mopper, K., and D. J. Kieber (2002), Photochemistry and the Cycling of Carbon, Sulfur,
Nitrogen and Phosphorus, in Biogeochemistry of Marine Dissolved Organic
Matter, edited by D. A. Hansell, and C. A. Carlson, pp. 455-507, Academic Press,
San Diego.
Morel, A., and S. Bélanger (2006), Improved detection of turbid waters from ocean color
sensors information, Remote Sens. Environ., 102, 237-249.
Morrison, J. R., and N. B. Nelson (2004), Seasonal cycle of phytoplankton UV
absorption at Bermuda Atlantic Time-series Study (BATS) site, Limnol.
Oceanogr., 49(1), 215-224.
Mueller, J. L., A. Morel, R. Frouin, C. Davis, R. A. Arnone, K. L. Carder, Z.-P. Lee, R. G.
Stewart, S. B. Hooker, B. Holben, C. D. Mobley, S. McLean, M. Miller, C.
Pietras, G. S. Fargion, K. D. Knobelspiesse, and K. Voss (2003), Radiometric
Measurements and Data Analysis protocols, in NASA Tech. Memo. 2003-21621,
rev. 4, vol. 3, edited by J. L. Mueller, G. S. Fargion, and C. R. McClain, pp. 78,
NASA Goddard Space Flight Center, Greenbelt, Md.
Pope, R. M., and E. S. Fry (1997), Absorption spectrum (380-700 nm) of pure water. II.
Integrating cavity measurements, Appl. Opt., 36, 8710-8723.
Quickenden, T. I., and J. A. Irvin (1980), The ultraviolet absorption spectrum of liquid
water, J. Chem. Phys., 72(8), 4416-4428.
Rachold, V., H. Eicken, V. V. Gordeev, M. N. Grigoriev, H.-W. Hubberten, A. P. Lisitzin,
V. P. Shevchenko, and L. Schirrmeister (2004), Modern Terrigenous Organic
Carbon input to the Arctic Ocean, in The organic carbon cycle in the Arctic
Ocean, edited by R. Stein, and R. W. MacDonald, pp. 33-41, Springer, Berlin.
Roesler, C. S., and M. J. Perry (1995), In situ phytoplankton absorption, fluorescence
emission, and particulate backscattering spectra determined from reflectance, J.
Geophys. Res., 100(C7), 13279-13294.
123

Roesler, C. S., M. J. Perry, and K. L. Carder (1989), Modeling in situ phytoplankton
absorption from total absorption spectra in productive inland marine waters,
Limnol. Oceanogr., 34, 1510-1523.
Ruddick, K. G., V. De Cauwer, Y.-J. Park, and G. Moore (2006), Seaborne
measurements of near infrared water-leaving reflectance: The similarity spectrum
for turbid waters, Limnol. Oceanogr., 51(2), 1167-1179.
Ruddick, K. G., F. Ovidio, and M. Rijkeboer (2000), Atmospheric correction of SeaWiFS
imagery for turbid coastal and inland waters, Appl. Opt., 39(6), 897-895.
Sathyendranath, S., L. Prieur, and A. Morel (1989), A three-component model of ocean
colour and its application to remote sensing of phytoplankton pigments in coastal
waters, Int. J. Remote Sens., 10(8), 1373-1394.
Siegel, D. A., S. Maritorena, N. B. Nelson, D. A. Hansell, and M. Lorenzi-Kayser (2002),
Global distribution and dynamics of colored dissolved and detrital organic
materials, J. Geophys. Res., 107(C12), 3228, doi:10.1029/2001JC000965.
Tassan, S., and G. M. Ferrari (1995), An alternative approach to absorption
measurements of aquatic particles retained on filters, Limnol. Oceanogr., 40(8),
1358-1368.
Tassan, S., and G. M. Ferrari (2002), A sensitivity analysis of the 'transmittancereflectance'
method for measuring light absorption by aquatic particles, J.
Phytoplank. Res., 24(8), 757-776.
Telang, S. A., R. Pocklington, A. S. Naidu, E. A. Romankevich, I. I. Gitelson, and
Glacyshev (1991), Carbon and Mineral Transport in Major North American,
Russian Arctic, and Siberian Rivers: the St Lawrence, the Mackenzie, the Yukon,
the Arctic Alaskan Rivers, the Arctic Basin Rivers in the Soviet Union, and the
Yenisei, in Biogeochemistry of Major World Rivers. SCOPE 42, edited by E. T.
Degens, S. Kempe, and J. E. Richey, pp. 74-104, Wiley, New York.
Twardowski, M. S., E. Boss, J. M. Sullivan, and P. L. Donaghay (2004), Modeling the
spectral shape of absorption by chromophoric dissolved organic matter. Marine
Chemistry, 89, 69-88.
Werdell, P. J., and S. W. Bailey (2005), An improved in-situ bio-optical data set for
ocean color algorithm development and satellite product validation, Remote Sens.
Environ., 98, 122-140.
Zaneveld, J. R. V., J. C. Kitchen, and C. Moore (1994), The scattering Error Correction
of Reflecting-Tube Absorption Meters, in Ocean Optics XII, edited by J. S. Jaffe,
pp. 44-55, SPIE, Bergen.
Zibordi, G., and G. M. Ferrari (1995), Instrument Self-Shading in Underwater Optical
Measurements - Experimental-Data, Appl. Opt., 34(15), 2750-2754.
124

Chapitre V: Impact de la glace de mer sur les
estimations de la réflectance marine, la
concentration de chlorophylle a et les propriétés
optiques inhérentes dérivés des données
satellitales de la Couleur de l’Océan
125

V.A. Résumé
À l’aide de simulations numériques et d’observations, on a décrit deux phénomènes
physiques impliquant la glace de mer qui contaminent les mesures spatiales de la couleur de
l’océan dans les hautes latitudes : 1) l’effet de l’environnement qui se produit le long de la
bordure de la banquise de glace et 2) la contamination sub-pixel due à la présence d’une
faible quantité de glace à l’intérieur d’un pixel classifié océanique. Le signal reçu au sommet
de l’atmosphère (TOA) a été simulé à l’aide un modèle de transfert radiatif permettant la
modélisation de l’effet de l’environnement pour divers types de glace de mer entourant une
surface d’eau ouverte. Ont été utilisés pour ces simulations des spectres de réflectance de la
glace de mer obtenus avant et pendant la période de fonte de la glace en 2004 dans la Mer de
Beaufort. Pour la contamination sub-pixel, le signal TOA a été simulé en utilisant les
propriétés atmosphériques utilisées dans le traitement des données SeaWiFS, pour différents
spectres de réflectance de surface obtenus en combinant linéairement les spectres de
réflectance marine (ρ w) et de glace de mer. Les algorithmes de correction atmosphérique
standard de SeaWiFS ont été appliqués aux spectres TOA simulés afin de retrouver le
spectre ρ w, à partir duquel les concentrations en chlorophylle a (CHL) et les propriétés
optiques inhérentes (IOP) sont calculées. Lorsque que les conditions atmosphériques sont
claires, l’effet de l’environnement introduit une surestimation significative (>0.002) de ρ w
dans la partie bleue du spectre (i.e. 443 nm), et ce à une distance aussi grande que 24 Km de
la bordure de glace. La CHL, calculée empiriquement à partir d’un algorithme utilisant des
rapports de réflectance entre le bleu et le vert, était soit sous-estimée, soit surestimée,
dépendamment de la concentration réelle en CHL. Pour des CHL de modérées à élevées (>
0.5 mg m -3 ), tout les pixels se trouvant à une distance inférieure à 10-20 Km de la bordure de
glace doivent être éliminés. Il devient crucial de considérer l’effet de l’environnement lorsque
l’on veut estimer le coefficient d’absorption total (a t) de l’eau de mer à partir d’un algorithme
d’inversion semi-analytique. À 443 nm, a t est sous-estimé par plus de 35% à ~20 Km de la
bordure de glace quand CHL est > 0.5 mg m -3 . L’effet sur l’estimation du coefficient de
rétrodiffusion par les particules (b bp) n’est important que lorsque que les eaux sont
relativement peu turbides (i.e. CHL < 0.05 mg m -3 et matière en suspension < 0.2 g m -3 ).
L’impact sur le rapport entre les coefficients d’absorption par la matière organique dissoute
colorée (CDOM) et total ([a CDOM/a t]) à 412 nm, calculé empiriquement, est similaire à celui
126

sur a t (i.e. sous-estimation). À l’opposé, la contamination sub-pixel produit une sous-
estimation systématique de ρ w dans la partie bleue du spectre. En générale, la contamination
sub-pixel résulte en une surestimation de la CHL, de a t, et de [a CDOM/a t], et en une
sousestimation de b bp. Une méthode simple a été proposée pour identifier les pixels
contaminés par les effets d’environnement.
127

V.B. Article soumis à la revue Remote Sensing of Environment (1 novembre 2006):
“Impact of sea ice on the retrieval of water-leaving reflectance, chlorophyll a
concentration and inherent optical properties from satellite Ocean Color data”
Simon Bélanger a , Jens K. Ehn b and Marcel Babin a
a CNRS, Laboratoire d'océanographie de Villefranche, 06230 Villefranche-sur-Mer,
France ; Université Pierre et Marie Curie-Paris 6, Laboratoire d'Océanographie de
Villefranche, 06230 Villefranche-sur-Mer, France
b Centre for Earth Observation Science, Department of Environment and Geography,
Clayton H. Riddell Faculty of Earth, Environment and Resources, 467 Wallace Building,
University of Manitoba, Winnipeg, MB, R3T 2N2, phone: 204-474-6961, fax 272-1532,
umehnjjk@cc.umanitoba.ca
128

Abstract
Through simulations and observations, we described two physical phenomena by which
satellite remotely sensed ocean color data are contaminated by sea ice at high latitudes: 1)
the adjacency effect that occurs along sea ice margins and 2) the sub-pixel contamination
by a small amount of sea ice within an ocean pixel. The signal at the top of the
atmosphere (TOA) was simulated using a radiative transfer code that allows modeling of
the adjacency effect for various types of sea ice surrounding an open water area. In situ
sea ice reflectance spectra obtained prior to and during the melt period are used in the
simulations. For sub-pixel contamination, the TOA signal was simulated using the
SeaWiFS atmospheric look-up tables with various surface reflectances obtained by linear
mixture of both sea ice and water-leaving reflectances (ρw). The standard atmospheric
correction algorithms were applied to the simulated TOA spectra to retrieve the ρw
spectrum from which chlorophyll a concentration (CHL) and inherent optical properties
(IOP) are derived. Under clear atmospheric conditions, the adjacency effect introduced
large overestimation in ρw (>0.002), as far as 24 km from an ice edge in the blue part of
the spectrum (443 nm). The CHL, estimated using standard blue-to-green reflectance
ratio, is either over- or underestimated depending on the actual concentration. For
moderate to high CHL (>0.5 mg m -3 ), any pixel located within a distance of ~10-20 km
from the ice edge should be discarded. It is very important to consider the adjacency
effect when the total absorption coefficient (at) is to be retrieved using a semi-analytical
algorithm. The at(443) was underestimated by more than 35% at a distance of ~20 km
from an ice edge for CHL > 0.5 mg m -3 . The effect on the particles backscattering
coefficient (bbp) retrieval was important only for clear waters (CHL ~0.05 mg m -3 ). The
impact on the ratio between chromophoric dissolved organic matter (CDOM) and total
absorption coefficients ([aCDOM/at]) at 412 nm, estimated empirically, is similar to at(443)
(i.e. underestimation). In contrast, sub-pixel contamination by small amount of sea ice
produces systematic underestimation of ρw in the blue. In general, sub-pixel
contamination results in an overestimation of CHL, at and [aCDOM/at], and an
underestimation of bbp. A simple method was proposed to flag the pixels contaminated by
adjacency effect.
129

V.1. Introduction
Polynyas and flaw leads are areas of open water, or much reduced sea ice cover,
at high latitudes where full sea ice cover is expected [Smith et al., 1990]. Due to their
high biological productivity, polynyas and leads are sites of reproduction for several
species of birds, fishes and mammals [Stirling, 1997]. At high latitudes, intense primary
productivity also occur in Marginal Ice Zones (MIZ) where melt water provides enough
vertical stability in the water column for phytoplankton to grow under high-light and
high-nutrients conditions [Smith and Nelson, 1986; Sakshaug and Skjoldal, 1989].
Because polar regions are especially susceptible to global warming (i.e., + ~3°C in 2050)
[ACIA, 2005], the timing and magnitude of the primary productivity in polynyas, leads
and MIZ are likely to be affected in the coming years.
Global warming can also alter the amount and fate of terrigenous material (e.g.,
organic matter, nutrients, and sediments) input into the Arctic Ocean, which is a key
component of the Arctic carbon cycle [Stein and MacDonald, 2004]. For example,
increase in river discharge [McClelland et al., 2006] and accelerate permafrost thawing
[Camill, 2005] could lead to the mobilization of significant amount of terrigenous
dissolved and particulate carbon toward to coastal Arctic ocean during the next century
[Frey and Smith, 2005; Guo and Macdonald, 2006]. Furthermore, amplified
photomineralization of terrigenous dissolved organic carbon in the Arctic coastal waters
is expected as a response to the decrease in summer ice cover that allows more solar
radiation to penetrate the water column [Bélanger et al., 2006]. Since global warming can
modify either marine or terrigenous components of the carbon cycle, continuous
monitoring systems for the biological state of north polar seas are urgently needed.
Satellite remote sensing of the ocean color provides a means for studying long-
term and large-scale changes in the biological processes occurring in the polar waters.
The most common bio-optical parameter derived from ocean color is the concentration of
chlorophyll a (CHL), the key molecule of photosynthesis and a common index of
phytoplankton biomass in the ocean. With a surface estimate of CHL, the gross primary
130

productivity can be calculated using various models [e.g., Antoine and Morel, 1996;
Behrenfeld and Falkowski, 1997; Arrigo et al., 1998]. Furthermore, the present
generation of bio-optical algorithms [Lee et al., 2002; Maritorena et al., 2002] can
provide, in addition to CHL, other optical indicators like some of the inherent optical
properties (IOP).
Good retrieval of CHL and IOP from remotely sensed ocean color data requires (1)
a reliable retrieval of the spectral water leaving radiance, Lw, or reflectance, ρw, which
relies mainly on the quality of the sensor calibration and of the atmospheric correction,
and (2) robust bio-optical algorithms. In polar seas, the former may be strongly
compromised by the presence of sea ice [e.g. Gregg and Casey, 2004]. Despite the fact
that this problem might represent a major obstacle to the use of ocean color in Polar
Regions, that problem has never been addressed quantitatively.
In view of using satellite remote sensing of bio-optical properties to study
biogeochemical processes occurring in polar seas, the purpose of this study is to quantify
the error on the estimate of ρw due to the presence of sea ice, and its consequences on the
retrieval of CHL and IOP. Sea ice can introduce biases in the interpretation of ocean
color data in two ways: 1) the adjacency effect is the process by which a photon, reflected
from a surface adjacent to a targeted pixel, is scattered by the atmosphere between the
sensor and the target, blurring the sharp boundary between sea ice and open water [e.g.,
Tanré et al., 1979; Tanré et al., 1981]; 2) sub-pixel contamination when a small amount
of sea ice within an ocean pixel (a pixel not flagged as cloud or ice), may be wrongly
interpreted as a high aerosol concentration and/or high seawater turbidity. The first
phenomenon is observed throughout the year, but may be more important in spring time
when leads and polynyas start to open and adjacent sea ice is still covered by highly
reflective fresh snow; while the second can occurs during summer when ice floes,
transported by water currents and winds, melt progressively in the ocean.
In what follows, we first describe the methodology which is adopted to simulate
the top-of-atmosphere (TOA) reflectance spectra, ρTOA, with and without the effects of
sea ice. Then we apply the Sea-viewing Wide Field-of-view Sensor (SeaWiFS)
atmospheric correction [Gordon and Wang, 1994] to our simulated TOA spectra to
quantify the errors in water-leaving reflectance resulting from the presence of ice. Next
131

the CHL is calculated using two empirical algorithms: 1) the standard NASA global
algorithm OC4v4 used in both SeaWiFS and MODIS processing [O'Reilly et al., 2000]
and 2) the Arctic’s regionally “tuned” OC4L algorithm [Cota et al., 2004]. For testing the
effect of sea ice on the retrieval of IOPs, we applied an improved version of the algorithm
of Lee at al. [2002] designed to estimate the total absorption (at) and particulate
backscattering coefficients (bbp). Finally, we tested an algorithm recently developed to
retrieve the ratio [aCDOM/at], which is necessary to estimate CDOM photooxidation
[Bélanger et al., Improved quantification of Chromophoric Dissolved Organic Matter
photooxidation in coastal waters using satellite-derived inherent optical properties,
submitted manuscript, 2006; (hereinafter referred to as Bélanger et al., submitted
manuscript, 2006)].
V.2. Methodology
V.2.1. Definitions
The radiance received by a sensor at the TOA in a given direction, LTOA, is
converted into remote sensing reflectance, ρTOA, using the following expression [Gordon,
1997] (note that geometrical dependency of the following quantities is omitted for
simplicity):
π LTOA
( λ)
ρ TOA(
λ)
= , (1)
F ( λ)
cos( θ )
132
0 s
where F0 is the extraterrestrial solar irradiance, θs is the sun zenith angle and λ is the
wavelength (in nm). Ocean Color applications commonly use the “normalized” water-
leaving reflectance (or radiance), [ρw]N (or [Lw]N), defined as [Gordon and Clark, 1981]:
π
π L ( λ)
ρ w = ( λ)
=
, (2)
N
) t
[ ] ( λ)
[ L ]
F0 ( λ)
w N
w
F0
( λ)
cos( θs
where Lw is the water-leaving radiance and t0 is the diffuse transmittance from sun to
pixel. Note that [Lw]N/F0 is equivalent to the commonly used remote sensing reflectance
(Rrs) defined as the ratio of water-leaving radiance to the downwelling plane irradiance
just above the sea surface [see Morel and Mueller, 2002].
0

V.2.2. Description of the surface reflectance spectra
V.2.2.1. Sea ice reflectance spectra
The spectral irradiance reflectance spectrum of sea ice (Rice) was determined at
three locations during the spring-summer 2004 onboard the CCGS Amundsen in the
Southeastern Beaufort Sea as part as the Canadian Arctic Shelf Exchange Study (CASES)
program. Radiometric measurements were performed at 512 wavelengths ranging from
340 to 1060 nm with a dual-headed spectroradiometer (FieldSpec, Analytical Spectral
Devices Inc., Boulder, Colorado) equipped with a cosine collector that was placed 0.7 m
above the sea ice surface, yielding measurement of upwelling and downwelling
irradiance, Eu and Ed, respectively, and a second cosine collector for simultaneous
measurements of
ref
E d to correct for changes in incident solar flux during the Eu and Ed
measurement. The irradiance reflectance is calculated as
ref
Eu
( λ)
Ed
( λ)
R ice ( λ) = . (3)
ref
E ( λ)
E ( λ)
d
d
The Rice spectrum was determined over four different surfaces: 1) 2-m-thick
landfast ice overlayed by a 15-cm-thick dry and fresh snow cover measured in early May
at 70.25°N and 126°W (in Franklin Bay, Canadian Arctic); 2) an ice floe covered by
melting snow [Perovich et al., 1998]; 3) a white ice floe with a drained high-scattering
surface layer; 4) a mixed white-grey ice floe with a drained low-scattering surface layers.
The ice floes were measured in the Amundsen Gulf (71.25°N, 127.66°W). The additional
spectrum from Perovich et al. [1998] was included to cover the conditions that may be
observed later during melting season (June-July) on landfast ice when the snow cover
vanishes. The five Rice spectra (Fig. V.1) are representative of Arctic sea ice at different
stages of their seasonal evolution. In early spring, the landfast ice is covered by fresh
snow that reflects >91% at all visible wavelengths. During summer time, the reflectance
can decrease below 10% and the spectral variations may become more pronounced (e.g.
the ice floe). During the melting season, however, the surface of an ice floe is a mixture
of different ice surface types. The surface of the ice floes encountered in the Amundsen
Gulf at the end of June was typically composed of about 30% of white ice and the rest of
grey ice (i.e. White and Grey ice spectra shown in Fig. V.1). Therefore, it was more
representative to average the two ice floe spectra measured on June 26 as: 0.3* Rwhite ice +
133

0.7*Rgrey_ice. This spectral composite is called hereinafter “grey ice”. An extensive study
of the sea ice surface reflectance is out of the scope of this study.
Figure V.1. Reflectance spectra for five types of Arctic sea ice at different
stages during the melting season. The spectrum of landfast ice covered by
melting snow was taken from Perovich et al. [1998], while the other
spectra were measured in the Southeastern Beaufort Sea during the
CASES expedition.
V.2.2.2. Water-leaving reflectance spectra
We calculated 20 water-leaving reflectance spectra using the semi-analytical
forward model described in the Appendix V.1 (Fig. V.2). The parameterization of the
model was designed to simulate spectra typical of both Open Ocean (Case 1) and coastal
(Case 2) waters. The three input parameters used to set the spectral IOP (i.e. absorption
and backscattering coefficients) are: 1) the phytoplankton chlorophyll a concentration
(CHL); 2) the dry weight of particles that do not co-vary with phytoplankton (SPM NAP )
(in g m -3 ); and 3) the contribution of CDOM to the non-water absorption coefficient at
443 nm [fCDOM = aCDOM/(aφ+aNAP+aCDOM)]. To generate Case 1 water reflectance spectra,
fCDOM and SPM NAP are set to 0, and the IOPs are a function of CHL as in Morel and
Maritorena [2001] (for details, see Appendix V.1). To represent a wide range of trophic
134

and turbidity conditions that could be encountered at high latitude, the following input
parameters were used:
1. 4 values for SPM NAP (0, 0.2, 2.0 and 20 g m -3 );
2. 3 values for CHL (0.05, 0.5, 5.0 mg m -3 ) (excepted when SPM = 20.0 g m -3
because [ρw]N spectra are almost identical for any CHL values);
3. 2 levels for fCDOM (50% and 80%).
To be more representative of polar waters where the contribution of the colored dissolved
and detrital material (CDM) to the total light absorption at 440 nm is high (>50%) [Siegel
et al., 2005, see their Figure 2a], high values for fCDOM were chosen. In the Arctic, in
particular, the strong CDOM signal is due to the high concentration of DOC that is
maintained by the large amount of river runoff entering into polar surface waters annually
[see Benner et al., 2005 and ref. therein].
Figure V.2. Modeled normalized water-leaving reflectance spectra for
various chlorophyll concentration (mg m -3 ), as indicated. The solid curves
are produced by introducing an fCDOM value of 50% while dotted curves
are produced with fCDOM =80%. Four concentrations of non algal particles
(SPM NAP ) are presented: a) 0.0 g m -3 ; b) 0.2 g m -3 ; c) 2.0 g m -3 ; d) 20 g m -3 .
135

V.2.3. Simulations of the adjacency effect
The adjacency effect is significant at high latitude because the sea ice, that
reflects up to 90% of the incident light, are adjacent to seawater with a reflectance < 5-
10%. The Second Simulation of the Satellite Signal in the Solar Spectrum (6S) radiative
transfer code [available online at http://www-loa.univ-
lille1.fr/SOFTWARE/Msixs/msixs_gb.html, Vermote et al., 1997] was used to simulate
ρTOA for the eight SeaWiFS spectral bands (i.e. 412, 443, 490, 510, 555, 670, 765, and
865 nm). The ρTOA spectra are simulated for water pixels located at 1 to 50 km from the
coast to offshore along a transect perpendicular to an hypothetical linear ice edge of
infinite length (for details, see Appendix V.2). The gas concentration and the
atmospheric pressure, temperature, ozone and water vapor profiles for the U.S. standard
atmosphere [Mc Clatchey et al., 1971] were used as input parameters. The 6s maritime
aerosol model composed of 95 and 5% of oceanic and water-soluble components,
respectively [Vermote et al., 1997], was used in the simulations. The aerosols optical
thickness at 550 nm (τa(550)) was set to 0.03, 0.1 and 0.2 to cover very clear to turbid
atmospheric conditions (note that global mean is 0.1). Finally, the illumination and
viewing geometry is set for typical SeaWiFS observation in mid-June at 70°N, for a pixel
located in the center of the field of view (i.e., θ0 = 50°; satellite viewing angle from nadir,
θv = 22°; azimuth angle between the sun-pixel and pixel-sensor half planes, Δφ = 8°). The
calculations were performed for four different types of ice and for various water-leaving
spectra as described above (Section V.2.2).
V.2.4. Simulations of sea ice contamination within an ocean pixel
During the summer season, after the breakup of the ice cover, oceanic pixels may
be contaminated by the presence of sea ice. Physically, the phenomenon can be compared
to the effect of sea foam or whitecaps [Frouin et al., 1996]. In such a case, the TOA
reflectance can be expressed as:
TOA
path
[ ( 1 − σ )[ ρ R ]
ρ = ρ + t ] + σ
0 tv
w N ice , (4)
where ρpath is the atmospheric path reflectance, tv is the diffuse atmospheric transmittance
from sea surface to the sensor and σ is the fractional sea ice surface concentration within
136

the pixel (dimensionless, from 0 to 1). Thus in equation 4, the bracketed term represent
surface reflectance, which is simply a linear mixture of both sea ice and water-leaving
reflectances. Here, ρpath, tv and t0 were taken from the SeaWiFS lookup tables for an
atmosphere containing maritime aerosol with relative humidity of 90%. The illumination
and viewing geometry and aerosol concentrations were the same as described in Section
2.3. Simulations were done for several values of σRice(865) between 0 and 3.2%, a range
at which the pixel may not be masked as cloud. Above this value, the pixel is generally
masked as a cloud. For example, SeaWiFS cloud masking is perform using a single
threshold (here set to 3%) on the TOA reflectance at 865 nm corrected for the molecular
scattering contribution (Rayleigh). So, depending on the magnitude the ice albedo (Fig.
V.1), the σ values ranged from 0 to ~3.5% for the fresh snow, and to ~25% for the grey
ice.
V.2.5. Atmospheric correction and bio-optical processing
In order to apply the SeaWiFS atmospheric algorithm to the simulated spectra, the
ρTOA were converted into LTOA (Eq. 1) and processed using the SeaWiFS Data Analysis
Software (SeaDAS 4.6). The standard atmospheric correction algorithm based on the
black pixel assumption in the near infrared (NIR) [Gordon and Wang, 1994] was used to
process most simulated spectra. An atmospheric correction algorithm designed for turbid
waters [Stumpf et al., 2003] was used to process the simulated spectra under moderately
turbid water conditions (Fig. V.2d). For chlorophyll a concentration estimation, we used
the global OC4v4 algorithm, which is a forth-order polynomial function that uses the
maximum reflectance band ratio between 443, 490 and 510 nm upon 555 [O'Reilly et al.,
2000]:
with,
CHL
a
+ a * R + a * R
2
3
4
= 10 0 1
2
3
4 ,
R =
log 10
⎛ max
⎜
⎝
137
+ a
* R
+ a
{ [ ρ ] ( 443,
490,
510)
}
* R
[ ] ⎟ ⎞
w N
, (5a)
ρw
( 555)
N
⎠

and where ai are empirical coefficients (0.366, -3.067, 1.93, 0.649, -1.532, respectively).
We also tested the regionally tuned OC4L algorithm proposed by Cota et al. [2004] for
the Arctic ocean,
b0
b1
* R
CHL 10
+
= , (5b)
where R is the same maximum band ratio that is used in OC4v4 (0.592 and -3.609 for b1
and b2, respectively).
For testing the influence of sea ice on the IOP retrieval, we examine the quasi-
analytical algorithm (QAA) [Lee et al., 2002], as modified here (see Appendix V.3).
Finally, we applied an empirical algorithm designed to estimate the ratio between the
CDOM and total absorption coefficients [Bélanger et al., submitted manuscript, 2006]:
⎡ a
[ ρ ] ( 412)
[ ρ ] ( 490)
CDOM ⎤
⎛ w ⎞ ⎛ w ⎞
N
N
( 412)
= c0
+ c1
⋅ log ⎜ ⎟
10
+ c2
⋅ log ⎜ ⎟
10
+ c3
⋅ log10
( [ ρ w ] ( 555)
/ π ) , (6)
⎥
N
a
⎜ [ ρ ] ( 555)
⎟ ⎜ [ ρ ] ( 555)
⎟
⎢
⎣
t
⎦
⎝
w
N
where ci are empirical coefficients (-0.51, -0.54, 0.48 and -0.45, respectively).
V.3. Results and discussion
V.3.1. Impact of the adjacency effect on the retrieval of [ρw]N, CHL and IOPs
⎠
Figure V.3 shows the absolute difference between the retrieved [ρw]N after
application of the atmospheric correction algorithm to the simulated ρTOA in presence and
absence of adjacency effect (Δ[ρw]N), at 443 and 555 nm for clear water (CHL=0.05 mg
m -3 ; fCDOM=50%), five types of sea ice, and for low and high concentrations of aerosols
(τa(550) = 0.03 and 0.2 respectively). The difference is always larger at 443 than at 555
nm because of the properties of the atmospheric light scattering. Indeed, the adjacency
effect is a result of increased scattering by atmospheric molecules and aerosols with
decreasing wavelength (referred to as Rayleigh and Mie scattering) [Tanré et al., 1979].
Therefore, a larger number of blue photons reflected by the ice environment are scattered
above the targeted water surface compared to the number of green photons. Accordingly,
the adjacency effect is even smaller in the near infrared part of the spectrum, where the
atmospheric correction algorithm estimates the aerosol type and concentration. The
adjacency effect is only partly removed by atmospheric correction. Although the increase
in the NIR signal is interpreted as an increase in aerosol concentration, and thus
138
⎝
w
N
⎠

subtracted from ρTOA, the atmospheric correction algorithm does not totally remove the
additional signal at shorter wavelengths that result from the adjacency effect. This is
because the adjacency effect is mainly due to the molecular scattering, which has a
steeper spectral dependency than aerosols scattering (λ -4 versus ~λ 0 to λ −2 ). This also
explains why the adjacency effect has less impact when the aerosol concentration
increases and Rayleigh scattering becomes relatively less important (Fig. V.3). Under
clear skies (i.e., τa(550) = 0.03), the total atmospheric scattering approaches the
molecular spectral dependency resulting in a strong spectral dependence for the
adjacency effect. Under a turbid sky (i.e., τa(550) = 0.2), in contrast, the spectral
dependency of the adjacency effect approaches the aerosol scattering dependency (∼λ −1 )
and the contamination is better removed by the atmospheric correction algorithm.
Figure V.3. Error on [ρw]N (denoted as Δ[ρw]N) at 443 and 555 nm for four
types of sea ice and two concentrations of aerosol (τa=0.03 and 0.2), as a
function of the distance from the ice edge. The dashed line represents the
acceptable error limit on [ρw]N at 443 nm (i.e. ±2 x 10 -3 ) for precise
chlorophyll estimation [Antoine and Morel, 1999].
139

As expected, the extent of the adjacency contamination also depends on the
magnitude of sea ice reflectance. Δ[ρw]N is larger when the landfast ice is covered by
fresh snow that reflects >90% of the incident light, and decreases by a factor of ~4 for
less reflective sea ice surfaces (Figs. V.3a-d). If we take 0.002 as an acceptable error limit
on [ρw]N(443) for good CHL retrieval [Antoine and Morel, 1999], then the adjacency
effect would be significant within 20-25 km of the landfast ice surface covered by fresh
or melting snow (Figs. V.3a-b), and within 5-15 km for an ice floe surface (Figs. V.3c-d).
The impact of the adjacency effect on the bio-optical parameters retrieval is
examine for τa(550) = 0.1 with an adjacent target covered by dry fresh snow. Therefore,
with this setting, the effect is nearly at its maximum. Figure V.4 shows the impact CHL
estimation. Only the results for water-leaving spectra generated with fCDOM = 50% and
SPM NAP = 0 g m -3 are presented here because they are more representative of the high
latitude open ocean waters (see Section V.2.2.2). The dashed lines represent SeaWiFS
mission requirement of 35% error on CHL [Hooker and McClain, 2000]. Interestingly,
the adjacency effect can lead to an over- and under-estimation of chlorophyll
concentration depending on the actual chlorophyll a concentration. When the
concentration is low (0.05 mg m -3 ), the CHL is increasingly overestimated when
approaching the ice edge. This is because, at low CHL, the blue-to-green radiance ratio
is higher for seawater than for the photons originating from ice and scattered by the
atmosphere to the sensor. In this case, the adjacency effect tends to decrease the blue-to-
green radiance ratio, which leads to CHL overestimates. In contrast, when the initial CHL
is >0.5 mg m -3 , the blue-to-green radiance ratio is lower for seawater than for the photons
reflected by ice and scattered by the atmosphere to the sensor. In this case, the adjacency
effect tends to increase the blue-to-green reflectance ratio, which leads to CHL
underestimates. In general, the error on CHL falls within the 35% range for low initial
concentration (0.05 mg m -3 ) but becomes significant when CHL is >0.5 mg m -3 . In the
worst case simulated (i.e. when CHL = 5.0 mg m -3 , τa = 0.03, and sea ice is covered by
fresh snow), the retrieved CHL are of unacceptable reliability (i.e., error >35%) within 20
km from an ice edge (Table V.1). Compared to OC4v4 algorithm, the Arctic OC4L
algorithm shows somewhat higher sensitivity to the adjacency contamination (Fig. V.4;
Table V.1).
140

Figure V.4. Ratio of CHL estimated from OC4v4 (solid line) and Arctic
OC4L (dotted line) when adjacency effect are present to CHL estimated
without adjacency effect as a function of the distance from the ice edge.
The results are presented for three initial concentrations of chlorophyll
(with fCDOM =50% and SPM NAP =0.0 g m -3 ) as indicated, τa=0.1 and for
four types of sea ice. The dashed lines represent a 35% error range.
Table V.1. Distance (in km) from ice egde within which the error on CHL > 35% (Rice
fresh snow). Values in parenthesis are for the OCL4 algorithm.
Initial CHL τa
(mg m -3 ) 0.03 0.1 0.2
0.05 0 (0) 0 (0) 0 (10)
0.5 12 (14) 10 (17) 8.5 (11)
5.0 19 (20) 17 (19) 17 (18)
Figure V.5 shows the impact of the adjacency effect on at retrieval at 443 nm
when the nearby ice is covered by fresh snow and aerosols are at moderate concentration
(τa(550) = 0.1). Here the error on at(443) is examined for the 20 water-leaving spectra
presented on Fig. V.2. As for the CHL retrieval, we assume an error acceptable when it is
141

less than 35%. In all cases, at(443) is increasingly underestimated when approaching the
ice edge. At low CHL (0.05 mg m -3 ) and fCDOM (50%), the errors are

Figure V.6 shows results similar to those in Fig. V.5, but for the retrieval of the
particulate backscattering coefficient (bbp) at 555 nm. In general, bbp(555) is increasingly
overestimated when approaching the ice edge when CHL is

Figure V.6. Same as Fig. V.5, but for bbp estimated at 555 nm using the
QAA.
Figure V.7. Same as Fig. V.5, but for the ratio [aCDOM/at] estimated at 412
nm.
144

The impact of the adjacency effect on ocean color data is illustrated by the pixels
extracted along a 37-km transect starting from 70.24°N and 124.6°W to 70.54°N and
124.3°W taken from the SeaWiFS scene acquired on the 31 st of May 2004 over the
Amundsen Gulf (70.25°N; 124.5°W) (Fig. V.8). The SeaWiFS level 1a image with
Local-Area-Coverage (LAC) resolution (1 km) received by the University of Fairbanks in
Alaska was obtained from the Distributed Active Archive Center (DAAC) and processed
to Level 2 using SeaDAS 4.7. Note that this transect is located in clear waters adjacent to
the offshore ice edge. Therefore, it can be reasonably assumed that all bio-optical
variables do not vary much in that area and that any change in them can be ascribed to the
adjacency effect. The aerosol optical thickness (τa) ranges between 0.10 and 0.11 away
from the ice edge (> 40 km), while it increases to ~0.125 toward the ice edge (Fig. V.8a).
This increase in τa is most likely due to the effect of adjacent ice edge, which had not
started to melt significantly at that time, as corroborated by in situ observations in this
area during the CASES expedition (data not shown). At 443 nm, the [ρw]N increased from
~0.01 offshore to 0.038 close to the ice edge, while [ρw]N(555) increased from 0.008 to
0.013 (Fig. V.8b). Consequently, the CHL estimation decreases from ~0.55 to ~0.22 mg
m -3 and from ~0.55 to 0.1 mg m -3 using OC4v4 and OC4L, respectively (Fig. V.8c). Note
that if the chlorophyll concentration was really decreasing toward the ice edge, [ρw]N at
555 nm should also decrease (see Fig. V.2). These results therefore confirm that the
adjacency effect actually occurs and is not corrected for by current SeaWiFS algorithms.
Interestingly, if the correct concentration is ~0.55 mg m -3 (i.e. the offshore value) the
error on CHL is >35% within the first 10 and 15 km from the ice edge for OC4v4 and
OC4L respectively, as predicted by our numerical simulations (see Fig. V.4). The impact
on the IOPs and [aCDOM/at](412) retrievals is also evident on this transect (Figs. V.8d-e),
and coherent with the predicted impact of the adjacency effect.
145

Figure V.8. SeaWiFS along-track transect extracted from an image of the
Amundsen Gulf acquired on the 31 st of May 2004: (a) τa; (b) [ρw]N at 443
and 555 nm; (c) CHL; (d) at and bbp at 443 and 555 nm, respectively; (e)
[aCDOM/at] at 412 nm.
146

V.3.2. Impact of sub-pixel sea ice contamination on the retrieval of [ρw]N, CHL and IOPs
The impact of the sub-pixel sea ice contamination is presented here as a function
of the product between the fraction of ice within a given pixel and its reflectance at 865
nm (σRice(865)). This way, the shape of the Rice rather than its magnitude, produce
different results on the retrieval of ρw. Therefore, results presented in this section must be
interpreted with keeping in mind that ice floes can have a fresh snow- or a melting snow-
liked reflectance spectrum (Fig. V.1).
In general, the absolute error in the normalized water-leaving reflectance, Δ[ρw]N,
decreases with increasing σRice(865) (Fig. V.9). This is explained by the overestimation
of the aerosol contribution resulting from the enhanced reflectance in the NIR due to sea
ice. As for the adjacency effect, the contamination is more pronounced at 443 nm than at
555 nm, except in presence of grey ice (Fig. V.9d). The larger negative Δ[ρw]N values at
the shorter wavelengths result from the extrapolation of the aerosol reflectance to the
visible from the NIR. Note that under turbid atmospheric condition (e.g., τa = 0.2), pixels
become flagged as clouds when σRice(865) gets larger than ca. 0.017 if a one-band
threshold test at 865 nm, as for SeaWiFS, is used for cloud detection (resulting in the
truncation of the black full lines in Fig. V.9). Because the spectral shape of Rice in the
NIR influences the selection by the atmospheric correction algorithm of the aerosol
models, the type of sea ice has a significant impact on the retrieval of [ρw]N. For instance,
because the landfast ice covered by melting snow and the grey ice have a steeper
decrease towards longer wavelengths (see Fig. V.1), they have the greatest impact on the
retrieval of [ρw]N at all wavelengths (Fig. V.9b,d). Note that the landfast ice cases can
only occur for pixels which are partly overlapping landfast ice, in which case the
adjacency effect would completely mask the sub-pixel contamination effect (e.g., Fig.
V.8) that are computed assuming no contribution from adjacent pixels.
147

Figure V.9. Error on [ρw]N at 443 and 555 nm for four types of sea ice and
two concentrations of aerosol (τa=0.03 and 0.2), as a function of the
product of fractional sea ice surface concentration and its reflectance
values at 865 nm, σRice(865). The dashed line represents the acceptable
error limit on [ρw]N at 443 nm.
To illustrate the impact of sub-pixel contamination on the aerosol model selection,
Fig. V.10 shows the angstrom exponent (α) as a function of σRice(865) for the different
types of sea ice. Here, the α refers to the exponent of the power law in λ −α for the spectral
variation in τa. The α increases by a factor >2 for the landfast ice covered by melting
snow and the grey ice floe, which explain the greatest Δ[ρw]N values obtained for these
type of ice (Figs. V.9b,d). For spectrally neutral Rice spectra such as for the landfast ice
covered by fresh snow and the ice floe covered by melting snow (see Fig. V.1), α
remains relatively unchanged in presence of sea ice. Although these sea ice types do not
significantly affect the selected aerosol models, they still introduce an error in [ρw]N. In
fact at TOA, σRice is modulated by the diffuse transmittance (tv and t0 in Eq. 4) which
decreases exponentially with decreasing wavelengths. For example, the tvt0 values are
148

0.71, 0.86 and 0.96 at 443, 555 and 865 nm, respectively, for a τa(865) value of 0.1 and
the sun-viewing geometry adopted here.
Figure V.10. Retrieved Angström exponent parameter (α) using the
atmospheric correction algorithm, as a function of σRice, for τa=0.1 and
four types of sea ice.
Figure V.11 shows the implication of errors in [ρw]N on the CHL estimation for
the four sea ice types (with τa(550) = 0.1). As for [ρw]N, the error on CHL is dramatic in
presence of the landfast ice (Figs. 11a-b). For CHL > 5.0 mg m -3 , a 10-fold
overestimation could be reached, even for a small amount of sea ice within the pixel
(σRice~0.05-0.01; Fig. V.11b). For low CHL (0.05 mg m -3 ), the underestimation can be
important in the presence of melting snow (Fig. 11b) due to the fact that the band ratio
tends toward infinity. In presence of ice floes, strong overestimation of CHL occur when
the actual concentration is >0.5 mg m -3 (Figs. 11c-d). As for the adjacency effect, the
OC4L algorithm is more sensitive to the contamination than OC4v4 is.
Figures V.12 and V.13 show the impact of the sub-pixel contamination by sea ice
covered by fresh snow on the retrievals of the at(443) and bbp(555), respectively. In
general, the negative Δ[ρw]N result in an overestimation of at(443) (Fig. V.12). The error
increase when both CHL and fCDOM increase. At low CHL (0.05 mg m -3 ), the error is

y

Figure V.12. Ratio of at estimated at 443 nm using QAA when sea ice is
present within the water pixel to at estimated with no sea ice as a function
of σRice(865). The results are presented for each water type as shown in
figure 2 and for the landfast ice covered by fresh snow. The dashed lines
represent a 35% error range.
Figure V.13. Same as Fig. V.12, but for bbp estimated at 555 nm using
QAA.
151

The impact of sub-pixel contamination on the [aCDOM/at](412) retrieval is shown
in Fig. V.14. The underestimation of the blue reflectance produces an overestimation of
[aCDOM/at](412) in all situations. When CHL is >0.5 mg m -3 and SPM NAP is

high aerosol optical thickness (not shown) suggests that it is an artifact resulting from sea
ice contamination. The artifact is, in counterpart, restricted to only a few pixels within the
ice field.
Figure V.15. Example of sea ice contamination on the CHL estimation
(OC4v4) for a SeaWiFS scene of the Beaufort Sea acquired on the 8 th of
September 2002. Upper panel is the true color composite showing the sea
ice field offshore.
V.3.3. Detection of pixels affected by sea ice (adjacency effect and sub-pixel
contamination)
In this section, we propose a simple method to detect pixels contaminated by the
adjacency effect. Due to the highly dynamic nature of summertime sea ice and its quickly
153

changing optical properties, the adjacency effect could not easily be corrected or
predicted without external information. Although sea ice distribution and reflectance
properties may be obtained from other satellite sensors (e.g., SSMI, AVHRR, ERS), the
sensitivity, or time and space scales do not always match those of Ocean Color data.
Therefore, a spectral test on the retrieved [ρw]N spectrum is a straightforward way to
detect and eliminate contaminated pixels. As the adjacency effect increase with
decreasing wavelengths (see section V.3.1), the blue part of the spectrum is most
appropriate to detect such contamination. As mentioned above, [ρw]N spectra observed at
high latitude show a typical depression at λ < 450 nm [Reynolds et al., 2001; Wang and
Cota, 2003]. Figure V.16 shows the variation of the ratio [ρw]N(412)/[ρw]N(443) as a
function of [ρw]N(555) (which is an indicator of the water turbidity), for the CASES data
set and for modeled reflectance spectra generated using the reflectance model described
above (with values for CHL and SPM NAP as specified above, and with fCDOM = 20, 50
and 80%). So the ratio of [ρw]N(412)/[ρw]N(443) is generally < 1, and decrease as the
turbidity increase. This behavior is opposite to the adjacency contamination for which the
increase in [ρw]N at 555 nm is accompanied by an increase in the reflectance ratio
[ρw]N(412)/[ρw]N(443). The black line on Fig. V.16 represents an upper limit, though
arbitrary, below which the ratio [ρw]N(412)/[ρw]N(443) should be for a given [ρw]N(555)
value. So the contaminated pixels are identified as soon as the inequality
[ ρ ]
[ ρ ]
is satisfied. The threshold value is
w N ≥
w
N
( 412)
( 443)
Threshold
Threshold = . 64 − 0.
14log
([ ρ ] ( 555))
. (7)
0 10 w N
An example of application to a SeaWiFS image of the Southeastern Beaufort Sea
is shown in Figs. V.16 and V.17. The image presents a wide variety of water types from
turbid waters inshore to relatively clear waters offshore. Pixels surrounding the central
Arctic packed ice are well identified. As expected, the violet-to-blue reflectance ratio of
those pixels increases together with [ρw]N(555) (Fig. V.16). In addition, a number of
turbid pixels, located on the continental shelf east of the Mackenzie River and adjacent to
the remaining landfast ice, are also flagged for adjacency contamination (Fig. V.17).
154

Interestingly, those pixels would not have been detected if a fix threshold (i.e. 1.0) had
been adopted instead of using equation 7. Some pixels within the Amundsen Gulf, away
from the ice edges, were nevertheless flagged, which is likely a result from the adjacency
effect due to the presence ice floe apart from a continuous sea ice cover. The robustness
of the method was examined by applying it to several SeaWiFS scenes from the Beaufort
Sea. In some situations when the actual ratio [ρw]N(412)/[ρw]N(443) was very low (i.e. in
high CDM waters), only pixels where the contamination was extreme were detected.
Clearly, for an application of this flag to images from Antarctica or North Atlantic, an
adjustment of the threshold is likely necessary since the abundance of CDM is lower in
those regions compared to the Arctic coastal waters [Siegel et al., 2005].
Figure V.16. Reflectance ratio between [ρw]N at 412 and [ρw]N at 443 nm
as a function of [ρw]N at 555 nm for: 1) in situ measurements of the waterleaving
reflectance conducted in the Southeastern Beaufort Sea [Bélanger
et al., submitted manuscript. 2006] (closed circles); 2) modeled waterleaving
reflectance with the bio-optical model presented in Appendix V.1
(closed squares); and 3) SeaWiFS pixels extracted from the image
presented on Fig. V.17 (light grey dots).
155

Figure V.17. Example of application of the adjacency flag to a SeaWiFS
scene from the Southeastern Beaufort Sea acquired on the 16 th of June
1998. The Rrs at 555 nm is shown on colored log scale as indicated on the
right; the flagged pixels for adjacency are identified with dark red color.
The sea ice and land are shown in dark and light grey respectively.
Because of known problem with the standard SeaWiFS atmospheric
correction over the turbid waters, we apply the algorithm proposed by
Ruddick et al. [2000].
In this study we do not propose a method to detect the sub-pixel contamination
specifically. It is, however, possible to identify contaminated pixel when they are
relatively isolated by applying a spatial analysis of the distribution of the aerosol optical
thickness. Indeed, sub-pixel contamination will introduce some noise in the τa which may
be removed using a filtering scheme similar to the one proposed by Franz et al. [2003]
for reducing noise in the SeaWiFS NIR bands. In addition, the overcorrection for aerosol
due to sea ice results in negative water-leaving reflectance at 412 nm which can be used
as a data quality control flag. A visual examination of several individual full resolution
SeaWiFS scenes from the Southeastern Beaufort Sea led us to the conclusion that the
sub-pixel contamination is not an extensive phenomenon in this area. It appears that the
melting sea ice field is generally dense enough to be flagged as cloud, or to produce
adjacency effect. Clearly from this visual analysis, we conclude that adjacency effect is
far more important than sub-pixel contamination and must be accounted for when
processing and interpreting Ocean Color data.
156

V.4. Summary and conclusions
The adjacency effect and the sub-pixel contamination by a small amount of sea
ice within an ocean pixel affect the quality of the water-leaving reflectance retrieval from
satellite Ocean Color data at high latitude. Both contaminations introduce large errors in
[ρw]N in the blue wavelength region reducing the quality of the bio-optical parameters
derived from the water-leaving spectrum. The bio-optical parameters retrieval evaluated
in this study are the chlorophyll a, total absorption and particles backscattering
coefficients respectively, as well as the ratio between CDOM to total absorption
coefficient. While adjacency effect results in an overestimation of the [ρw]N, sub-pixel
contamination result in a systematic underestimation of the water-leaving reflectance in
the blue. These errors result either in under- and over-estimation of the CHL depending
on the actual chlorophyll a concentration. The impact on IOP retrieval is more
straightforward: 1) adjacency effect results in an underestimation of at and [aCDOM/at],
and an overestimation of bbp; 2) sub-pixel contamination results in an overestimation of at
and [aCDOM/at], and an underestimation of bbp. Because adjacency effect appear to be
more frequent compared to the sub-pixel contamination, a simple spectral test on the
retrieved [ρw]N was proposed to flag contaminated pixels. The method relies on the fact
that adjacency contamination increases both the reflectance ratio between 412 and 443
nm and the [ρw]N value at 555 nm, which is opposite to trend observed when seawater
turbidity increases (i.e. [ρw]N(412)/[ρw]N(443) decreases when [ρw]N(555) increases).
Clearly the adjacency effect must be accounted for when the aim is to use Ocean
Color data for quantifying bio-optical properties and biogeochemical processes in ice
covered seas. The method proposed to eliminate contaminated pixel by this effect could
be easily included in the data processing chain, but more data in polar seas are needed to
adjust the threshold. Nevertheless, because open water nearby sea ice are important
biologically speaking, an extension of the atmospheric correction scheme to include a
correction for adjacency effect is desirable. This might be done by including additional
LUTs generated accounting for the adjacency effect. Those LUTs would be used in the
processing only when contaminated pixels are detected. On the other hand, the sub-pixel
157

contamination, which is not an extensive problem, could probably identified using
simultaneous high resolution measurements (e.g. MODIS 250-m and MERIS 330-m
channels).
Finally, it must be stressed that similar effect could also be observed world-wide
in the surrounding of thick cloud. Indeed the adjacency-flagging procedure proposed
above frequently eliminates the pixels located nearby clouds in the study area. Further
study on the adjacency contamination due to thick cloud found at different altitude in the
atmosphere is necessary. The contamination might be particularly important in the
equatorial zones where irradiance and could cover are high.
Acknowledgment
This study was made possible with financial support from the Fonds Québécois pour la
Recherche sur la Nature et les Technologies (FQRNT) (to SB), the Fonds France-Canada
pour la Recherche (FFCR to MB) and Indian and Northern Affairs Canada (to SB). We
are grateful to the Laboratoire d’Optique Atmosphérique (LOA) for providing the 6S
radiative transfert code. Pierre Larouche and Dave Barber are kindly acknowledged for
providing instrumentation and field assistance. Yannick Huot, Fabrizio D’Ortenzio and
André Morel are specially thanked for their valuable comments on the manuscript. We
thank the CCGS Amundsen cruise members for their enthusiastic help onboard the ship.
This is a contribution to the Canadian Arctic Shelf Exchange Study (CASES) under the
overall direction of L. Fortier.
Appendix V.1. Water reflectance model
The water-leaving reflectance spectra were generated using the semi-analytical
model proposed by Park and Ruddick [2005] valid for both case 1 and case 2 waters. The
remote sensing reflectance (Rrs) is expressed using a forth-order polynomial parameter as:
158

4 ⎡ b ⎤ b
Rrs ( λ)
= ∑ gi
( λ)
⎢ ⎥ ( λ)
, (A1)
i=
1 ⎣at
+ bb
⎦
where bb and at are the total backscattering and absorption coefficients respectively, and
gi are coefficients tabulated in a look-up table (LUT) for as a function of Sun-Viewing
geometry and of a phase function parameter, γb(λ). γb(λ) is defined as the relative
contribution of particles to the total backscattering (i.e. bbp(λ)/bb(λ)).
The spectral total absorption coefficient is subdivided into four additive
components as:
159
i
a ( ) = a ( λ)
+ a ( λ)
+ a ( λ)
+ a ( λ)
, (A2)
t λ w ϕ NAP
CDOM
where subscripts w, φ, NAP and CDOM stand for water, phytoplankton, non-algal
particles, and colored dissolved organic matter (CDOM), respectively. Similarly, the
spectral total backscattering coefficient is express with three components as:
b ( ) = b
~
( λ)
+ b
~
b ( λ)
+ b b ( λ)
, (A3)
b λ bw bϕ
φ
bNAP NAP
where bφ and bNAP are the volume scattering coefficients for phytoplankton and NAP
respectively, and b bϕ
~ and b bNAP
~
are the ratio of backscattering to scattering coefficients
for phytoplankton and NAP respectively. Spectral values for aw and bbw are taken from
Pope and Fry [1997] and Morel [1974], respectively.
Spectral absorption coefficient for phytoplankton and its co-varying materials are
computed as a function of CHL using the bio-optical model proposed by Morel and
Maritorena [2001] (hereafter referred as MM01). To avoid confusion with current
literature on the phytoplankton absorption [Bricaud et al., 1995], the IOP computed using
MM01 will be denoted as acase1 and bbcase1 (thus co-varying CDOM and NAP absorption
are included in acase1(λ)). Non-algal particle absorption varies as a function of SPM NAP
according to:
[ S ( 443 λ)
]
*
NAP
a ( λ) a ( 443)
SPM exp − , (A4)
NAP
= NAP
NAP
*
where a ( 443)
is the mass-specific absorption coefficient defined as ratio of aNAP(443)
NAP
to the SPM NAP (in m 2 g -1 ), and SNAP is the spectral slope of the aNAP(λ) spectrum. Here,
*
a ( 443)
= 0.0534 m 2 g -1 and SNAP = 0.010 nm -1 . Those values were observed in the
NAP
turbid waters of the Mackenzie River [S. Bélanger and P. Larouche, unpublished data].

Spectral CDOM absorption is generated assuming a given contribution of the
CDOM to the non-water constituents absorption coefficients at 443 nm (fCDOM) as:
and
a
CDOM
( 443)
[ a ( 443)
+ a ( 443)
]
f CDOM NAP
case1
= , (A5)
( 1 − f )
CDOM
160
[ S ( 443 ) ]
a ( λ) a ( 443)
exp − λ . (A6)
CDOM
= CDOM
CDOM
Here, SCDOM is fixed at 0.022 nm -1 , the average value measured in the Southeastern
Beaufort Sea [Bélanger et al., submitted manuscript, 2006].
The scattering coefficient at 555 nm for the phytoplankton and its co-varying
particles is obtained using a statistical relationship established for Case 1 Loisel and
Morel [Loisel and Morel, 1998]:
−0.
766
bcase 1(
555)
= 0.
41CHL
. (A7)
To obtain the backscattering coefficient for phytoplankton and its co-varying particles,
the backscattering ratio is computed following MM01 as:
~
= 0.
002 + 0.
01 0.
5 − 0.
25log
( CHL)
. (A8)
bbcase1 [ ]
The spectral backscattering coefficients are obtained using a power law function in
λ ν where the exponent increases from -1 to 0 with respect to CHL as:
calculated as:
= 0. 5*
(log10
( CHL)
− 0.
3)
ν , 0.03 < CHL < 2.0 mg m -3
10
ν = 0, CHL > 2.0 mg m -3 . (A9)
For NAP, which does not covary with phytoplankton, the scattering coefficient is
b ( 555)
*
NAP
*
NAP
= bpNAP
( 555)
SPM , (A10)
*
where b ( 555)
is the mass-specific scattering coefficient defined as ratio of bNAP(443)
pNAP
to the SPM NAP (in m 2 g -1 *
). A b ( 555)
value of 0.5 m 2 g -1 , typical of coastal waters
pNAP
[Babin et al., 2003], was adopted. The backscattering ratio for NAP was assumed to be
1.83%, as measured by Petzold [1972] in the San Diego Harbor. The spectral bbNAP(λ)
values were assumed to follow a λ -1 law.

Appendix V.2. Post-treatment of 6s output
The 6s code computes the ρTOA for a circular target of radius r, surrounded by an
environment of different reflectivity (see Tanré et al. [1981]). Briefly, the reflectance at
the TOA, ρTOA, is express as
TOA
[ ρ + ρ ( r)
ρ ( r)
]
ρ = + , (A11)
tg path `t
arg et env
where tg is the total gas transmission (i.e. O2, H2O and O3), ρpath is the atmospheric path
reflectance, ρtarget is the target reflectance, and ρenv is the environment reflectance. The
reflectances of the target (e.g. water-leaving, ρw) and the environment (e.g. sea ice, ρice)
are expressed as
−τ
T ( θs
) ⎛ cos( θ ) ⎞ v
ρ t arg et ( r) = ⎜ ρwe
⎟
(A12)
1−
〈 ρ〉
( r)
S ⎝ ⎠
and
T ( θs
)
ρ env(
r) = ( 〈 ρ〉
( r)
td
( θv
) ) , (A13)
1−
〈 ρ〉
( r)
S
where T(θs) is the total transmittance (direct + diffuse), S is the spherical albedo of the
atmosphere,
e
−τ
cos( θv
)
is the direct upward transmittance, td is the diffuse transmittance,
and 〈ρ 〉 (r)
is a spatial average of each pixel reflectance over the whole surface. Here
〈ρ 〉 (r)
is a function of the atmospheric conditions and the surface reflectance which is
given by
〈 ρ 〉 ( r ) = ρ F(
r)
+ ( 1−
F(
r))
R , (A14)
w
ice
where F(r) is the environment function, which depends on the atmospheric conditions
(for details see Tanré et al. [1981] and Vermote et al. [1997]). ρtarget and ρenv were
computed for several r (0.5 to 100 km), and four types of sea ice (Section 2.2.1). Since
more or less linear ice edges rather than circular target is expected, ρtarget(r) and ρenv(r)
were used to estimate the contribution of the environment and the target at TOA of a
pixel located at distance D0 from an ice edge as (see Fig. V.A1)
π
2
edge 1
arg et ( D0) = ∫ ρt
arg et
2π
−π
2
ρt ( D)
dα
(A15)
and
161

π
2
1
D0 ) =
2 ∫ ρenv
π −π
2
edge
ρenv (
( D)
dα
, (A16)
with cos ( )
1 −
D = D α .
0
When 6s is run for an inhomogeneous surface, the specular reflection of the diffuse
skylight toward the sensor is not account for. The sky radiance reflected by the air-water
interface transmitted to the TOA, ρsky, can be estimated as
ρ
sky
π Lsky
⋅ ρ Fresnel
= tv
(A17)
ε F θ )
0
cos( 0
where ρFresnel is the reflectance of a flat air-water interface given by Fresnel’s formula
(typically ~2.3% for satellite viewing angle), Lsky is the radiance at surface coming from
the diffuse sky, and tv is the upward transmittance from pixel to sensor. Assuming a
perfectly isotropic sky, then
diff
Ed
L sky = , (A18)
0.
5π
where
diff
E d is the diffuse downward irradiance at surface and 0.5 is the mean cosine for
an isotropic sky. Since
path reflectance.
diff
Ed and tv are given by 6s, ρsky was estimated and added to the
Figure V.A1. Schematic representation of a water pixel located at distance
D0 from an ice edge.
162

Appendix V.3. Modification of the QAA
The quasi-analytical algorithm (QAA) proposed by Lee et al. [2002] was
modified in respect to 1) the formulation of Rrs versus the IOP, 2) the empirical
formulations for the absorption coefficient at reference wavelength (λ0), and 3) the use of
a single value for the exponent of the power law used to model the particulate
backscattering spectrum (Y).
First, a semi-analytical remote-sensing reflectance model designed to account for
the bidirectional effects (Eq. A1) [Park and Ruddick, 2005] is used instead of the Gordon
et al. [1988] formulation. Equation A1 requires a priori information on backscattering
nature of the medium (i.e. molecular versus particulate) to operate. Therefore, the original
QAA is used to provide an initial estimate of γb at a reference wavelength. With the initial
value of γb(λ0) and Rrs(λ0, θs, θv, Δφ, W), [bb/(at+bb)](λ0) is estimated by inverting the
fourth-order polynomial using a numerical solution [Laguerre's method, or the zroots
function in Chap 9.5 of Press et al., 1992]. Two iterations are sufficient to obtain a
difference between two consecutives estimate of γb(λ0)

available on our ac-9. The value of at(650) is calculated empirically as (r 2 = 0.71, N = 14
at-w(555)>0.5aw(555)):
164
⎡ Rrs
( 650)
⎤
−0.
69+
1.
41log10
⎢ ⎥
⎣ Rrs
( 555)
⎦
a ( 650)
= a ( 650)
+ 10
(A20)
t
w
where aw(650)=0.34 [Pope and Fry, 1997]. For continuity in the at retrieval, when at-
w(555) falls between 0.03 and 0.06 m -1 the a linear combination of the two results
obtained with λ0 = 555 and 650 nm is performed following Eq. 20 in Lee et al. [2002].
Third, the spectral particles backscattering coefficient (bbp) is calculated using λ −Y
law with Y = 1.

V.5. References
ACIA (2005), Arctic Climate Impact Assessment, 1042 pp., Cambridge Press University.
Antoine, D., and A. Morel (1996), Oceanic primary production 1. Adaptation of a
spectral light-photosynthesis model in view of application to satellite chlorophyll
observations, Global Biogeochem. Cycles, 10(1), 43-55.
Antoine, D., and A. Morel (1999), A multiple scattering algorithm for atmospheric
correction of remotely sensed ocean colour (MERIS instrument): principle and
implementation for atmospheres carrying various aerosols including absorbing
ones, Int. J. Remote Sens., 20(9), 1875-1916.
Arrigo, K. R., D. Worthen, A. Schnell, and M. P. Lizotte (1998), Primary production in
Southern Ocean waters, J. Geophys. Res., 103(C8), 15587-15600.
Babin, M., A. Morel, V. Fournier-Sicre, F. Fell, and D. Stramski (2003), Light Scattering
properties of marine particles in coastal and open ocean waters as related to the
particle mass concentration, Limnol. Oceanogr., 48(2), 843-859.
Behrenfeld, M. J., and P. G. Falkowski (1997), Photosynthetic rates derived from
satellite-based chlorophyll concentration, Limnol. Oceanogr., 42(1), 1-20.
Bélanger, S., H. Xie, N. Krotkov, P. Larouche, W. F. Vincent, and M. Babin (2006),
Photomineralization of terrigenous dissolved organic matter in Arctic coastal
waters from 1979 to 2003: Interannual variability and implications of climate
change, Global Biogeochem. Cycles, doi:10.1029/2006GB002708.
Benner, R., P. Louchouarn, and R. M. W. Amon (2005), Terrigenous dissolved organic
matter in the Arctic Ocean and its transport to surface and deep waters of the
North Atlantic, Global Biogeochem. Cycles, 19, GB2025,
doi:10.1029/2004GB002398.
Bricaud, A., M. Babin, A. Morel, and H. Claustre (1995), Variability in the chlorophyllspecific
absorption coefficients of natural phytoplankton: Analysis and
parameterization, J. Geophys. Res., 100(C7), 13321-13332.
Camill, P. (2005), Permafrost thaw accelerates in boreal peatlands during late-20th
century climate warming, Climatic Change, 68(1-2), 135-152.
Cota, G. F., J. Wang, and J. C. Comiso (2004), Transformation of global satellite
chlorophyll retrievals with a regionally tuned algorithm, Remote Sens. Environ.,
90, 373-377.
Franz, B. A., R. E. Eplee, S. W. Bailey, and M. Wang (2003), Changes to the
atmospheric correction algorithm and retrieval of oceanic optical properties, Chap.
5, in NASA Tech. Memo. 2003-206892, vol. 22, edited by S. B. Hooker, and E. R.
Firestone, pp. 29-33, NASA Goddard Space Flight Center, Greenbelt, Md.
Frey, K. E., and L. C. Smith (2005), Amplified carbon release from vast West Siberian
peatlands by 2100, Geophys. Res. Let., 32(9).
165

Frouin, R., M. Schwindling, and P. Y. Deschamps (1996), Spectral reflectance of sea
foam in the visible and near-infrared: In situ measurements and remote sensing
implications, J. Geophys. Res., 101(C6), 14361-14371.
Gordon, H. R. (1997), Atmospheric correction of ocean color imagery in the Earth
Observation System era, J. Geophys. Res., 102(D14), 17081-17106.
Gordon, H. R., O. B. Brown, R. H. Evans, J. W. Brown, R. C. Smith, K. S. Baker, and D.
K. Clark (1988), A semianalytic radiance model of ocean color, J. Geophys. Res.,
93(D9), 10909-10924.
Gordon, H. R., and D. K. Clark (1981), Clear Water Radiances for Atmospheric
Correction of Coastal Zone Color Scanner Imagery, Appl. Opt., 20(24), 4175-
4180.
Gordon, H. R., and M. Wang (1994), Retrieval of water-leaving radiance and aerosol
optical thisckness over the oceans with SeaWiFS: a preliminary algorithm, Appl.
Opt., 33(3), 443-452.
Gregg, W. W., and N. W. Casey (2004), Global and regional evaluation of the SeaWiFS
chlorophyll data set, Remote Sens. Environ., 93, 463-479.
Guo, L., and R. W. Macdonald (2006), Source and transport of terrigenous organic matter
in the upper Yukon River: Evidence from isotope (δ 13 C, Δ 14 C, and δ 15 N)
composition of dissolved colloidal, and particulate phases, Global Biogeochem.
Cycles, 20, GB2011, doi:10.1029/2005GB002593.
Hooker, S. B., and C. R. McClain (2000), The calibration and validation of SeaWiFS data,
Prog. Oceanog., 45, 427-465.
Lee, Z.-P., K. L. Carder, and R. A. Arnone (2002), Deriving inherent optical properties
from water color : a multiband quasi-analytical algorithm for optically deep
waters, Appl. Opt., 41(27), 5755-5772.
Loisel, H., and A. Morel (1998), Light scattering and chlorophyll concentration in case 1
waters: A reexamination, Limnol. Oceanogr., 43(5), 847-858.
Maritorena, S., D. A. Siegel, and A. R. Peterson (2002), Optimization of a semianalytical
ocean color model for global-scale applications, Appl. Opt., 41(15), 2705-2714.
Mc Clatchey, R. A., R. W. Fenn, J. E. A. Selby, F. E. Volz, and J. S. Garing, Optical
properties of the Atmosphere, AFCRL-TR-71-079, L.G. Hancom Fiel, Bedford,
Ms., 1971.
McClelland, J. W., S. J. Dery, B. J. Peterson, R. M. Holmes, and E. F. Wood (2006), A
pan-arctic evaluation of changes in river discharge during the latter half of the
20th century, Geophys. Res. Let., 33(6), L06715, doi:10.1029/2006GL025753.
Morel, A. (1974), Optical properties of pure water and pure sea water, in Optical aspects
of Oceanography, edited by E. S. Nielsen, pp. 1-24, Academic, New York.
Morel, A., and S. Maritorena (2001), Bio-optical properties of oceanic waters: A
reappraisal, J. Geophys. Res., 106(C4), 7163-7180.
166

Morel, A., and J. L. Mueller (2002), Normalized Water-Leaving Radiance and Remote
Sensing Reflectance: Bidirectional Reflectance and Other Factors, in Ocean
Optics Protocols For Satellite Ocean Color Sensor Validation, Revision 4,
Volume III, edited by C. R. McClain, pp. 32-59, NASA Goddard Space Flight
Space Center, Greebelt.
O'Reilly, J. E., S. Maritorena, G. Mitchell, D. A. Siegel, K. L. Carder, D. L. Garver, M.
Kahru, and C. R. McClain (1998), Ocean color chlorophyll algorithms for
SeaWiFS, Journal of Geophysical Research, 103(C11), 24937-24950.
O'Reilly, J. E., S. Maritorena, D. Siegel, M. C. O’Brien, D. Toole, B. G. Mitchell, M.
Kahru, F. P. Chavez, P. Strutton, G. Cota, S. B. Hooker, C. R. McClain, K. L.
Carder, F. Muller-Karger, L. Harding, A. Magnuson, D. Phinney, G. F. Moore, J.
Aiken, K. R. Arrigo, R. Letelier, and M. Culver (2000), Ocean color chlorophyll a
algorithms for SeaWiFS, OC2, and OC4: version 4., in NASA Tech. Memo. 2000-
206892, vol. 11, edited by S. B. Hooker, and E. R. Firestone, pp. 9-23, NASA
Goddard Space Flight Center, Greenbelt, Md.
Park, Y.-J., and K. G. Ruddick (2005), Model of remote-sensing reflectance including
bidirectional effects for case 1 and case 2 waters, Applied Optics, 44(7), 1236-
1249.
Perovich, D. K., C. S. Roesler, and W. S. Pegau (1998), Variability in Arctic sea ice
optical properties, J. Geophys. Res., 103(C1), 1193-1208.
Petzold, T. J., Volume scattering functions for selected ocean waters, Tech. Rep. SIO, 72-
78 pp., Scripps Institution of Oceanography, San Diego, La Jolla, Ca., 1972.
Pope, R. M., and E. S. Fry (1997), Absorption spectrum (380-700nm) of pure water. II.
Integrating cavity measurements, Appl. Opt., 36, 8710-8723.
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery (1992), Numerical
recipes in Fortran 77: The art of scientific computing, Chap 9 pp., Cambridge U.
Press, New York.
Reynolds, R. A., D. Stramski, and B. G. Mitchell (2001), A chlorophyll-dependent
semianalytical reflectance model derived from field measurements of absorption
and backscattering coefficients within the Southern Ocean, J. Geophys. Res.,
106(C4), 7125-7138.
Sakshaug, E., and H. R. Skjoldal (1989), Life at the ice edge, Ambio., 18(1), 60-67.
Siegel, D. A., S. Maritorena, and N. B. Nelson (2005), Independence and
interdependencies among global ocean color properties: Reassessing the biooptical
assumption, J. Geophys. Res., 110(C7), C07011,
doi:10.1029/2004JC002527.
Smith, S. D., R. D. Muench, and C. H. Pease (1990), Polynyas and leads: An overview of
physical processes and environment, J. Geophys. Res. , 95(C6), 9461-9479.
Smith, W. O., and D. M. Nelson (1986), The importance of ice-edge phytoplankton
production in the Southern Ocean, BioScience, 36, 251-257.
167

Stein, R., and R. W. MacDonald (2004), Organic Carbon Budget: Arctic Ocean vs.
Global Ocean, in The Organic Carbon Cycle in the Arctic Ocean, edited by R.
Stein, and R. W. MacDonald, pp. 315-322, Springer, Berlin.
Stirling, I. (1997), The importance of polynyas, ice edges, and leads to marine mammals
and birds, J. Mar. Syst., 10(1-4), 9-21.
Stumpf, R. P., R. A. Arnone, R. W. Gould, R. M. Martinolich, and V. Ransibrahmanakul
(2003), A partially coupled ocean-atmosphere model for retrieval of water-leaving
radiance from SeaWiFS in coastal waters, Chap. 9, in NASA/TM-2003-206892,
edited by S. B. Hooker, and E. R. Firestone, pp. 51-59, Greenbelt, MD.
Tanré, D., M. Herman, and P. Y. Deschamps (1981), Influence of the background
contribution upon space measurements of ground reflectance, Appl. Opt., 20(20),
3676-3684.
Tanré, D., M. Herman, P. Y. Deschamps, and A. de Leffe (1979), Atmospheric modeling
for space measurements of ground reflectances, including bidirectional properties,
Appl. Opt., 18(21), 3587-3594.
Vermote, E., D. Tanré, J. L. Deuzé, M. Herman, and J. J. Morcrette, 6s User Guide
Version 2, 218 pp., Laboratoire d'Optique Atmosphérique, Lille, France, 1997.
Wang, J., and G. F. Cota (2003), Remote-sensing reflectance in the Beaufort and Chukchi
seas: observations and models, Appl. Opt., 42(15), 2754-2765.
168

Chapitre VI : Utilisation de l’imagerie « couleur de
l’océan » pour la quantification de la
photooxydation dans la Mer de Beaufort
169

VI.A Résumé
La production photochimique de carbone inorganique dissous (DIC) intégrée dans la
colonne d’eau a été estimée l’aide des propriétés optiques inhérentes dérivées des données
SeaWiFS de la Mer de Beaufort obtenues entre 1998 et 2004. Plus de 300 images SeaWiFS
ont été traitées à l’aide d’un algorithme de correction atmosphérique adapté pour les eaux
turbides qui a permis de retrouver des spectres de réflectance marine de bonne qualité. Après
l’élimination des données contaminées par les effets d’environnement dus à la glace de mer,
le rapport entre les coefficients d’absorption de la matière organique dissoute colorée
(CDOM) et totale ([a CDOM/a t]) a été calculé à partir du spectre de réflectance marine à l’aide
d’un algorithme empirique. La photoproduction annuelle de DIC estimée à l’aide de
[a CDOM/a t] dérivé des données SeaWiFS est faiblement inférieure (~10%) aux estimations
réalisées avec les valeurs régionales constantes de [a CDOM/a t] obtenues à partir des mesures in
situ. En faisant l’hypothèse que le rapport [a CDOM/a t] est de 0.9 sur tout le spectre, la
photoproduction annuelle de DIC était surestimée d’environ 60%. Ces résultats suggèrent
que les données de Couleur de l’Océan peuvent être utilisées pour estimer [a CDOM/a t] des les
eaux côtières arctiques, lequelles sont souvent inaccessibles et les mesures in situ n’y sont pas
toujours disponibles. En plus de fournir [a CDOM/a t], l’estimation de la valeure absolue de
a CDOM, obtenue grâce à l’estimation semi-analytique de a t, a aussi permis de tenir compte des
variations spatiales du rendement quantique apparent pour la photoproduction de DIC (φ DIC).
Dans le futur, d’autres produits dérivés des données de couleur de l’océan pourront être
utilisés afin d’évaluer la variabilité de φ DIC. Le manque de données sur ce dernier limite,
cependant, le développement d’un tel modèle prédictif de φ DIC.
170

VI.B. Article en preparation: “The use of Ocean Color imagery for the
quantification of photooxidation in the Beaufort Sea”
Simon Bélanger and Marcel Babin
Laboratoire d’Océanographie de Villefranche, Centre National de la Recherche
Scientifique, Université Pierre et Marie Curie - Paris 6, 06230 Villefranche-sur-mer,
France
171

Abstract
Depth-integrated photochemical production of dissolved inorganic carbon (DIC) in the
Southeastern Beaufort Sea was assessed for the period from 1998 to 2004 using
SeaWiFS-derived water inherent optical properties. More than 300 SeaWiFS images
were processed using a turbid-water atmospheric correction algorithm that provides
reliable water-leaving reflectance retrieval. After eliminating data contaminated by the
adjacency effect due to sea ice, the ratio between chromophoric dissolved organic matter
(CDOM) and the total absorption coefficients ([aCDOM/at]) was calculated from the water-
leaving reflectance spectrum using an empirical algorithm. The annual DIC
photoproduction estimated using SeaWiFS-derived [aCDOM/at] was slightly lower (~10%)
than the estimation made using regionally constant [aCDOM/at] spectra determined from in
situ measurements. In contrast, assuming a spectrally constant value for [aCDOM/at] of 0.9,
the annual DIC photoproduction was overestimated by ~60%. These results suggest that
satellite Ocean Color data could be used to estimate [aCDOM/at] in the remote Arctic
coastal waters where in situ measurements are still unavailable. In addition, the
estimation of the absolute value in aCDOM, as obtained from at calculated using a semi-
analytical algorithm, was used to account for the spatio-temporal variability in the
apparent quantum yield for the DIC photoproduction (φDIC). In the future, other Ocean
Color products may also be used to account for the variability in φDIC, but such data are
currently lacking to develop a robust parameterization of φDIC.
172

VI.1. Introduction
In Chapter III I quantified, for the first time, the amount of dissolved organic
carbon mineralized through CDOM photooxidation in the Beaufort Sea. A coupled
optical-photochemical was used to calculate the depth-integrated photoproduction of DIC
in ice-free Arctic waters. I showed that CDOM photooxidation in the southeastern
Beaufort Sea (western Arctic) may increase significantly in the future in response to the
predicted trend of ongoing reduction of sea ice cover. Interestingly, the predicted
increased in terrigenous dissolved organic carbon (tDOC) mineralization is nearly
equivalent to the amount of particulate marine organic carbon actually buried into the
deep sediments. This result confirms the relative importance of this flux in the balance of
the Arctic Organic Carbon Budget.
In our initial approach, satellite observations of sea ice, ozone, aerosols and cloud
cover were used with a radiative transfer model to estimate the spectral irradiance that
penetrates the water column. The regional variability of the surface water optical
properties as derived from in situ measurements was used in the calculation, which partly
controlled the rates of CDOM photooxidation in the water column. This variability is
expressed by the [aCDOM/at] parameter, which is the light fraction absorbed by the CDOM
in an optically homogenous water column. In Chapter IV, we developed an Ocean Color
algorithm to retrieve this parameter from the remote sensing reflectance (Rrs)
measurements. The absolute accuracy of the regionally tuned [aCDOM/at] algorithm is
±14%. Interestingly, despite UV light are the most efficient radiation in terms of DIC
photoproduction, the variability in [aCDOM/at] in the violet part of the spectrum (412 nm)
is nearly proportional to the spectrally-integrated photoproduction of DIC. We have
shown that using averaged spectral shapes for CDOM and particulate matter, spectral
[aCDOM/at] values can be obtained. The error in terms of DIC photoproduction attributed
to the spectral extrapolation is

to examine the specific problem posed to ocean color remote sensing by the presence of
this almost perfect reflector (Chap. V). The adjacency effect that is due to nearby sea ice
cover is only partly removed by atmospheric correction, resulting in a contamination of
the water-leaving reflectance retrieval and the subsequent estimation of [aCDOM/at](412)
from it. In general, [aCDOM/at](412) is underestimated by more than 0.15 at a distance of
10-15 km from an ice edge. As the adjacency effect increases with decreasing
wavelengths, the blue part of the spectrum was used to detect and discard pixels affected
by such contamination.
The main goal of the present chapter is to estimate the DIC photoproduction using
satellite-derived [aCDOM/at] in open waters free of adjacency contamination. The revised
estimations are compare to the ones obtained using in situ and constant [aCDOM/at] values.
Specifically, I want to know if, when in situ measurements of [aCDOM/at] in a given
coastal region are not available, ocean color data can be used instead. I first described the
data processing of a seven years time-series of Sea wide Field-of-View (SeaWiFS)
acquired between 1998 and 2004. Two different methods in which satellite-derived
information is used to calculate the DIC photoproduction are described. The first one
make used of [aCDOM/at] only, whereas in the second method, the magnitude of the
CDOM absorption coefficient at 412 nm is used to identify two water classes with
different photoreactivity. In the Results section, I present a validation of SeaWiFS water-
leaving reflectance by comparing coincident satellite and in situ measurements obtained
in 2004 in the study area. Then the observed spatio-temporal variability of some IOP is
presented. Finally, I discuss the quality of the SeaWiFS data and the utility of ocean color
imagery to quantify depth-integrated DIC photoproduction in the Arctic coastal zones.
VI.2. Methods
VI.2.1. SeaWiFS Level 2 and 3 processing
SeaWiFS Level 1A MLAC data (merged Local Area Coverage) gathering all
available SeaWiFS HRPT (High Resolution Picture Transmission) and LAC data for a
given orbit were downloaded from the NASA Ocean Color Web site
(www.oceancolor.gsfc.nasa.gov). The Level 1A MLAC data contains raw radiance
values for each of the eight SeaWiFS bands (412, 443, 490, 510, 555, 670, 765, and 865
174

nm) and have a spatial resolution of 1.1 km at nadir. At the latitude of 70°N, a daily
temporal coverage is available with SeaWiFS (up to 3 images per day). However, due to
the presence of sea ice or clouds, only the scenes where open water were present were
processed to Level 2. A total of 335 scenes acquired during the summer months from
1998 to 2004 were processed to Level 2 (Table VI.1).
Table VI.1. Number of SeaWiFS Level 1A MLAC images processed
May June July August Total
1998 14 19 28 17 78
1999 10 11 19 19 59
2000 0 0 9 16 25
2001 6 9 17 15 47
2002 5 10 9 25 39
2003 5 16 11 10 42
2004 1 22 13 9 45
Total 335
The Level 2 processing was performed image-by-image in two steps: 1)
atmospheric corrections (AC) were achieved to subtract the atmospheric scattering
components from the top-of-atmosphere (TOA) reflectance and obtain the water-leaving
reflectance for the visible bands; and 2) the water-leaving reflectance spectrum was
analyzed to derive inherent optical properties (IOP).
The SeaWiFS standard atmospheric correction scheme over clear open ocean
waters is based on the “black pixel assumption” [e.g., Gordon and Wang, 1994; Gordon,
1997; Siegel et al., 2000]. This assumption is based on fact that water-leaving reflectance
for the near-infrared bands at 765 nm and 865 nm is null, and that the measured signal at
those bands in entirely due to the atmosphere. Then the atmospheric aerosols properties
obtained in the NIR can be extrapolated to the visible using recomputed look-up-tables.
Typically over turbid waters, the black pixel assumption results in negative water-leaving
reflectance in the blue part of the spectrum (< 443 nm) due to an overestimation of the
aerosol signal extrapolated from the NIR bands. Because of the violation of the “black
pixel assumption” in turbid coastal waters, the AC was performed using a modified
version of the Multi-Sensor Level-1 to Level-2 (MSL12) software. The modified MSL12
175

was provided by the Belgian Management Unit of the North Sea Mathematical Models
(MUMM; www.mumm.ac.be/OceanColour/). In the MUMM’s MSL12, the assumption
of null water-leaving reflectance in the NIR bands is replaced by the assumptions of
spatial homogeneity of the 765:865-nm ratios for aerosol reflectance and for water-
leaving reflectance [Ruddick et al., 2000]. These ratios can be fixed as calibration
parameters during the processing. While the ratio of water-leaving reflectance is expected
to be rather independent of region and time [Ruddick et al., 2006], the aerosol reflectance
ratio is expected to vary in space and time in response to variations in aerosol particle
type and concentration. But here, it is assumed that the aerosol type is constante and that
only the amount of aeraosol is variable for a given image. Therefore, the aerosol
reflectance ratio (denoted hereafter as "eps78") needs to be assessed on an image-by-
image basis by inspection of the Rayleigh-corrected reflectance scatterplot [Ruddick et al.,
2000]. Instead of inspecting the Rayleigh-corrected reflectance scatterplot, which could
be contaminated by clouds or sea ice, I obtained eps78 by the inspection of the
atmospheric products derived using the standard AC algorithm. In general I chose an
averaged eps78 value obtained in an area away from the turbid waters, but as close as
possible. In general, the aerosol eps78 values were within the range from 1.0 to 1.15,
which is typical of marine aerosols. When large areas were still showing negative water-
leaving reflectance after the application of turbid water AC algorithm, the image was
reprocessed using a slightly lower value for eps78. To avoid negative reflectance in the
blue, therefore, I often had to choose lower values for eps78.
After AC processing, the pixels affected by the adjacency effect (Chap. V), as
well as standard flagged pixels [e.g, Bailey and Werdell, 2006], were not considered for
further bio-optical processing. Typically, when the sea ice melting begins in early season,
more than 50% of the open water pixels must be eliminated. This number decreases to
~15% during the summer season. Note that the cloud borders were also detected by the
adjacency effect detection algorithm (Section V.3.3.), resulting sometime in more than
50% of invalid pixels even in August. For each valid pixels, the ratio [aCDOM/at] at 412
nm was calculated using the empirical algorithm described in Chapter IV, with the
region-specific coefficients (Table IV.2). The total absorption coefficient at 412 nm (at)
176

and particulate backscattering coefficient at 555 nm (bbp) were computed using a
modified version of the quasi-analytical algorithm (QAA) proposed by Lee et al. [2002]
(see Annex A1). The former coefficient is further used in the DIC photoproduction
processing to extrapolate [aCDOM/at](412) to the full UV-to-green spectrum (300 and 600
nm) (see section IV.3.3) and to estimate the magnitude of aCDOM at 412 nm.
The Level 3 processing consists in binning the Level 2 geophysical products in
time and space. First, the Level 2 data were remapped on an azimuthal projection grid
with pixel resolution of 1 x 1 km. Then, the monthly averages were produced for each
pixel of the grid with all valid pixels available. Finally, for the DIC photoproduction
calculation, the SeaWiFS monthly averages were remapped onto the grid used by the
National Snow and Ice Data Center (NSIDC) for the sea ice concentration data
distribution. The NSIDC grid is polar stereographic projection with grid elements of
approximately 25x25 km resolution.
VI.2.2. DIC photoproduction calculation
The DIC photoproduction calculation is essentially the same as the one described
in section III.2.5. Here the spatially integrated DIC photoproduction (in mol DIC d -1 ),
however, is done for each element of the NSIDC grid as:
P
×
DIC
600
∫
λ = 300
( x,
y,
d)
= OpenWaterArea(
x,
y,
d)
− ⎡a
⎤ CDOM
Ed
( 0 , λ, d)
⎢ ⎥
⎣ at
⎦
,
( λ,
x,
y,
d)
φDIC
( λ)
dλ
(1)
where the OpenWaterArea (m 2 ) is the area of open water of a given NSIDC pixel,
−
E ( 0 , λ)
(mol photons m -2 d -1 ) is the spectral daily downward irradiance just beneath
d
sea surface, φDIC is the spectral apparent quantum yield for the DIC production, x and y
are the longitude and the latitude, respectively, and d is the day of the year. A schematic
representation of the DIC modeling is shown and detailed in Fig. VI.1.
In Chapter III, the southeastern Beaufort Sea was sub-divided in three sub-regions:
Canada Basin, Mackenzie Shelf, and Amundsen Gulf. For each of those, the φDIC spectra
were assumed constant based on the location of the in situ sampling. Here an alternative
method is proposed to select the φDIC as a function of the magnitude of the aCDOM(412). It
appears, at a first approximation, that the CDOM photoreactivity in the study area is
177

higher when the CDOM concentration is higher (Fig. VI.2a). For the low values of
CDOM absorption coefficient at 412 nm (i.e. aCDOM(412) < 0.2 m -1 ) the φDIC for station
108, 406 and 409 were pooled together, while R5a, R5d and R9 were average to represent
the high CDOM concentration (see Fig. III.1 for station locations; Fig. VI.2b). In the
processing, aCDOM(412) is calculated as the product of the empirically-derived
[aCDOM/at](412) and the QAA-derived at(412) (see Annex A1).
Figure VI.1. A schematic representation of the DIC photoproduction
modeling. The model ingests four inputs: (1) For a given day, when data
are available, the spatio-temporal variability in the [aCDOM/at] parameter is
specified by interpolating the values of the months before and after. If
monthly IOP are not available (due to either sea ice or cloud cover), the
data from the climatology is taken. The spectral interpolation of the
[aCDOM/at](412) is perform as described in section IV.3.4. (2) The daily
spatial variability in the area of open water. (3) The spectral daily
downward irradiance calculated as described in section III.2.5. (4) The
spectral apparent quantum yield for the DIC production (AQY, or φDIC).
Two methods are compared for the choice of the φDIC for a given day and
pixel. Method 1 is based on the regional definition described in section
III.2.5. In the Method 2, the φDIC varies as a function of the magnitude of
the CDOM (see text for details).
178

VI.3. Results
Figure VI.2. A) CDOM photoreactivity as a function of aCDOM(412). B)
Spectral variability of the pooled φDIC spectra for low and high aCDOM(412)
value.
VI.3.1. Validation of the SeaWiFS water-leaving reflectance
To assess the quality of the retrieved water-leaving reflectance (Rrs) with the
MUMM’s algorithm, coincident satellite and in situ observations (from hereafter called
“match-ups”) are compared. Out of 48 Rrs spectra measured during CASES (section
IV.2.1.2), six match-ups of good quality were retained for the analysis. The match-ups
used for the assessment were collected within a lag shorter than three hours between in
situ sampling and the satellite overpass, to minimize perturbations due to the temporal
179

variability of seawater optical properties (Table VI.2). Satellite data were obtained from
the average of the 3 x 3 image elements centered on the measurement site, and were
considered as valid for comparison when all of the 9 elements were not flagged for cloud
or adjacency effect. Table VI.2 provides the details on the match-ups location, date, time
interval, and the sun zenith angle at the time of sampling.
Table VI.2. Location, date, time interval, and the sun zenith angle at the time of sampling
of the match-ups
CASES
Station
Date Lat.
(°N)
Long.
(°W)
SeaWiFS image
(SYYYYDDDHHMMSS)
Time interval
(hh:mm)
Sun
zenith
ID
(°)*
CA-20 17-Jul 70.43 126.35 S2004199221315 00:25 51.9 (50.9)
309-1 18-Jul 71.12 125.84 S2004200225405 02:10 54.8 (50.3)
309-2 19-Jul 71.17 126.2 S2004201215611 00:05 52.3 (52.2)
721 26-Jul 69.84 133.29 S2004208232426 01:40 55.5 (62.9)
115 29-Jul 70.84 125.09 S2004211220929 02:40 57.5 (54.9)
215 30-Jul 70.95 123.52 S2004212225019 01:10 54.8 (61.9)
* The value in parenthesis is the sun zenith angle at the time of the in situ measurements.
The in situ measurements of the water-leaving reflectance spectra are compared to
SeaWiFS data processed with both the standard and the turbid water AC algorithms (Fig.
VI.3). Moderate turbidity was observed at station 721 located in the Mackenzie Shelf
(Rrs(555)=0.005 sr -1 ), while the turbidity was relatively low at the stations located in the
Amundsen Gulf. The results of the match-ups analysis can be summarized as follows:
1. When the turbidity is low (i.e. Rrs(555) < 0.002 sr -1 ; stations 309 and 215), the
water-leaving reflectance at wavelengths larger than 490 nm retrieved using the
standard algorithm is in good agreement with the in situ measurements.
2. The standard AC algorithm produced a systematic underestimation of the
reflectance at 412 and 443 nm, regardless of the water turbidity. For the most
turbid station (721), negative reflectance at 412 nm was return by the standard AC
algorithm, certainly because of the violation of the black pixel assumption.
3. The turbid water AC algorithm (MUMM) produced an overestimation of the Rrs
spectrum all wavelengths (except at stations CA-20 and 115). The overestimation
is in general more pronounced when the turbidity is low (stations 309 and 215).
180

4. In all cases, the spectral shape of the Rrs spectrum retrieved using the turbid water
AC algorithm is good agreement with the in situ measurements.
Figure VI.3. Comparison of in situ measurements of water-leaving
radiance with coincident spectra from the SeaWiFS image processed with
the standard and the turbid water AC algorithms.
Figure VI.4 compares satellite-derived versus in situ reflectance ratios employed
in the [aCDOM/at](412) empirical algorithm, i.e. the log[Rrs(412)/Rrs(555)] and
log[Rrs(490)/Rrs(555)]. The SeaWiFS-derived ratios from both the standard and the turbid
water AC algorithms are shown. Larger discrepancies (> 3 fold) in the
log[Rrs(412)/Rrs(555)] are obtained with the standard relative to the turbid water AC
181

algorithm (Fig. VI.4a). On the other hand, both algorithm provide similar results for the
retrieval of the log[Rrs(490)/Rrs(555)] (Fig. VI.4b).
Figure VI.4. Scatterplots of the in situ measurements versus SeaWiFSderived
variables employed in the empirical algorithm used to estimate
[aCDOM/at](412) (eq. IV.6): a) the logarithm of water-leaving reflectance
ratio of 412 to 555 nm; b) the logarithm of water-leaving reflectance ratio
of 490 to 555 nm.
The mean absolute difference between in situ and satellite-derived [aCDOM/at](412)
was, on average, +0.188 and -0.105 for the standard and turbid water AC algorithms,
respectively (Table VI.3). While the use of the standard AC results in systmatic
overestimation in [aCDOM/at](412), the overestimation of the retrieved water-leaving
reflectance at 555 nm using the turbid water AC algorithm results in a systematic
underestimation of [aCDOM/at](412). The [aCDOM/at](412) computed with the reflectance
182

spectrum derived using the standard AC can reach values above 1, which is physically
impossible. Using the turbid water AC algorithm, the [aCDOM/at](412) retrieval falls
within the algorithm accuracy (±0.14 see Chap. IV), except for stations 309-2 and 721.
The latter station was sampled in an area where the spatial variability was very high and
the vertical distribution of IOPs was not homogenous (data not shown). Under such
conditions large discrepancy between satellite retrievals and in situ Rrs spectra can be
expected (Fig. VI.3). In the next section, only the results using MUMM algorithm are
presented as they appears more reliable than the one obtained using standard AC (see
Discussion VI.4.1 section).
Table VI.3. Comparison of in situ with SeaWiFS-derived [aCDOM/at](412) values using
Standard and MUMM AC algorithms.
CASES Station
ID
[aCDOM/at](412)
in situ Standard MUMM
CA-20 0.88 1.07 (+.19) 0.83 (-.05)
309-1 0.81 1.14 (+.33) 0.69 (-.13)
309-2 0.82 0.96 (+.14) 0.66 (-.16)
721 0.81 NA 0.63 (-.18)
115 0.71 0.83 (+.12) 0.68 (-.03)
215 0.76 0.92 (+.16) 0.68 (-.08)
VI.3.2. Spatial-temporal variability in [aCDOM/at](412)
An example of the seasonal variability in the [aCDOM/at](412) parameter for the
summers of 1998 and 2004 (lowest sea ice concentration of the time serie) is shown on
Fig. VI.5. Most of the study area is characterized by [aCDOM/at](412) values ranging from
0.6 to 0.7. These values may be slightly underestimated for the reasons explained above.
For example, the SeaWiFS averaged value in the Amundsen Gulf in July 2004 was 0.66,
while the in situ measurements obtained in this area during the same period range
between 0.60 and 0.94, with an average value of 0.78. In the sector of the Mackenzie
Shelf where the Mackenzie river input are high, however, the SeaWiFS-derived values
ranging from 0.30 to 0.40 are similar to those observed in 2004 near the Mackenzie River
Delta. To better see the turbid river plume, Fig. VI.6 shows the particulate backscattering
coefficient at 555 nm (bbp) observed during June and July. In fact, bbp(555) is a robust
indicator of the total suspended material. For both years, the plume extent was maximum
183

in June and decrease in July. The decrease in suspended material explains the increase in
[aCDOM/at](412) observed over the Mackenzie Shelf during summer, as the river discharge
declines (Fig. VI.5).
Figure VI.5. Seasonal variability of the SeaWiFS-derived [aCDOM/at](412)
for 1998 and 2004 (1 x 1 km resolution). Sea ice (or cloud) is shown in
grey.
184

Figure VI.6. Variability of the SeaWiFS-derived bbp(555) for June and
July 1998 and 2004 (1 x 1 km resolution).
Figure VI.7. Variability of the SeaWiFS-derived at(412) for June and July
1998 and 2004 (1 x 1 km resolution).
185

Further offshore at the Shelf break or along southern coasts of the Amundsen Gulf
and Alaskan Shelf, higher [aCDOM/at](412) values (>0.8) can be observed. These high
values are observed within highly absorbing waters likely coming from the rivers (Fig.
VI.7), but where the amount of suspended particles is relatively low (Fig. VI.6). The
striking extent of the Mackenzie River plume toward the west in the Canada Basin during
July 1998 is obvious on both [aCDOM/at](412) and at(412) maps. Interestingly, a little
amount of particles is transported beyond the continental shelf at that time (Fig. VI.6). In
2004, the presence of a persistent sea ice cover obscures the visualization of the
Mackenzie River plume extent, but the bbp(555) pattern observed in June and the high
at(412) values around the location 72°N and 140°W (~ 0.2 m -1 ) in July suggest that the
plume was taking approximately the same westward “route” as in 1998.
VI.3.3. Spatial-temporal variability in DIC photoproduction
The DIC photoproduction was reassessed with taking into account the spatio-
temporal variability in [aCDOM/at]. Table VI.4 compares the annual DIC photoproduction
calculated using four different approaches: 1) with regionally constant values for
[aCDOM/at](λ) and φDIC(λ), as defined in our previous study (Chapter III; the “GBC
method”); 2) with SeaWiFS-derived [aCDOM/at](λ) and regional φDIC(λ) values (“Method
1” in Fig. VI.1); 3) with SeaWiFS-derived [aCDOM/at](λ) and variable φDIC(λ) as a
function of aCDOM(412) (“Method 2” in Fig. VI.1); 4) spectrally constant [aCDOM/at] value
of 0.9 and regional φDIC(λ) (“Constant method”).
Table VI.4. Annual DIC photoproduction estimation using various methods. The value in
parenthesis for methods 1 and 2 is the difference relative (in %) to GBC method.
Year GBC Method 1 Method 2 Constant
1998 122.3 108.2 (-11.5) 115.5 (-5.6) 195.7 (+60.0)
1999 71.3 65.6 (-8.0) 72.3 (+1.4) 114.5 (+60.6)
2000 47.9 43.0 (-10.2) 47.2 (-1.5) 77.1 (+61.0)
2001 52.6 48.7 (-7.4) 56.4 (+7.2) 82.3 (+56.5)
2002 60.8 53.5 (-12.0) 54.1 (-11.0) 99.7 (+64.0)
2003 67.7 61.0 (-9.9) 61.5 (-9.1) 109.8 (+62.2)
2004 73.6 69.1 (-6.1) 75.4 (+2.4) 120.6 (+63.9)
186

Comparison between GBC method and Method 1, which differ only in terms
[aCDOM/at], suggest that our former annual DIC photoproduction estimations were
overestimated by 6 to 12%. However, this could be partly attributed to the
underestimation of [aCDOM/at](412) by SeaWiFS (see above). But a regional comparison
reveals that the largest differences were found mainly in the Canada Basin (~25%, data
not shown). This is probably due to the fact that the stations used to represent the Canada
Basin in our former estimation were biased toward higher [aCDOM/at] values. In fact, the
Canada Basin’s stations are located at the fringe of the shelf (see Fig. III.1), where
[aCDOM/at] is generally higher than over the deep basin (e.g., June and July 1998; Fig.
VI.5). In contrast, using a spectrally constant [aCDOM/at] value of 0.90 results in a severe
overestimation of the DIC photoproduction (~+60% compare to GBC method).
Method 2 is an alternative method to account for the variability of φDIC(λ) based
on derived aCDOM(412) magnitude The annual DIC photoproduction is slightly higher
than using Method 1 (Table VI.4). In particular, the production is significantly higher in
the Amundsen Gulf (~30 %) where most of the southern part appears to be affected river
runoff. The spatial variability in the DIC photoproduction rate (in gC m -2 y -1 ) derived
using Method 2 is shown in Fig. VI.8. Although most of the spatial variability is
explained by the sea ice concentration, higher DIC photoproduction is found in the waters
influenced by the river runoff. With the regionally constant φDIC(λ), the runoff was
blurred on the DIC photoproduction maps (data not shown). Finally, the variability in
DIC photoproduction observed between 1998 and 2004 is extremely important both in
time and space, with rates ranging between ~0.05 and 0.65 gC m -2 y -1 . Most of this
variability is, nevertheless, explained by the variability in sea ice concentration.
187

Figure VI.8. Annual maps of DIC photoproduction rates from 1998 to 2004 in southeastern Beaufort Sea.
188

VI.4. Discussion
VI.4.1. SeaWiFS data quality
The visual inspection of several SeaWiFS images confirms that the Rrs(412)
retrieved using the standard AC algorithm was often negative, or unrealistically too low,
even several kilometers away from the turbid waters. It is not clear why the standard AC
produces systematic underestimation of Rrs(412) over clear waters. Because this
systematic underestimation of the water-leaving reflectance at 412 nm results in negative
values or severe underestimation of the log[Rrs(412)/Rrs(555)] ratio (Fig. VI.4a), the
whole SeaWiFS data set was processed using the turbid water AC. Ideally, the turbid
water atmospheric correction should be applied only over area where seawater is actually
turbid. For the three match-ups where clear waters were present, the aerosols optical
thickness at 865 nm was reduced by a value of ~0.01 (from ~0.04 to 0.03) using the
turbid water AC, resulting into an underestimation of the aerosol radiances over the
whole visible spectrum and an overestimation of the magnitude of the Rrs (Fig. VI.3).
Several reasons can be invoked to explain this results. First, the radiometric calibration of
the NIR channels may not be perfect, leading to the selection of an aerosol type with
higher spectral dependency than it should. This would explained the systematic
underestimation of Rrs < 490 nm, and the relatively good retrieval for channels ≥ 490 nm
obtained with the standard AC algorithm (Fig. VI.3). Second, in clear waters, the
765:865-nm water-leaving reflectance ratio tends to be higher than the value of 1.72 that
is given as a constant in the turbid water AC algorithm (see Figure 4 in Ruddick et al.
[2006]). Consequently, when applying the turbid water AC algorithm over clear water
pixels, the concentration of aerosol is underestimated, obtained by solving the set of two
equations for the near infrared bands (for details see Ruddick et al. [2000]). While fixing
the type of aerosol significantly reduced the amount of negative reflectance at 412 nm, it
causes an underestimation of the aerosol concentration over the clear waters.
One way to obtain consistent reflectance in the blue part of the spectrum may be
to revise the vicarious calibration coefficients. The vicarious calibration coefficients are
factors applied to the TOA radiances that have been estimated a posteriori to improve the
quality of the water-leaving radiance. The current coefficients were determined using
189

MOBY (Marine Optical Buoy) in situ match-up located west of Hawaiian island of Lanai
characterised by Case 1 water with low pigment concentration. It is known that if the
environmental conditions of a given region, in particular the type and the concentration of
atmospheric aerosols, depart from the “typical” conditions encountered at the MOBY site,
then vicarious coefficients may not be valid [e.g. of the Baltic Sea, Ohde et al., 2002]. In
addition, the sun zenith angle encountered at the MOBY site is always lower than 50°
(except at the winter). For example, a recent validation of satellite reflectances obtained
over the Adriatic Sea indicates a degradation of the data quality when the sun zenith
angle exceed 40° [Zibordi et al., 2006]. To revise the SeaWiFS vicarious calibration for
the Arctic Ocean, however, a large number of match-ups is required, which are scarce in
this region. Clearly, more work is needed to better understand the reasons of the failure of
the standard AC algorithm over the clear water area.
VI.4.2. Quantification of DIC photoproduction using satellite Ocean Color
To my knowledge, there is only three other studies that proposed the used of
Ocean Color data to calculate photochemical rates in the marine environment. The first
two methods, proposed by Cullen et al. [1997] and Johannessen et al. [2003], were never
tested nor validated on satellite imagery to derive depth-integrated photoproduction of
DIC. More recently, Fichot [2004] proposed a method to derive empirically the diffuse
attenuation coefficients at several UV wavelengths and CDOM absorption coefficient at
320 nm. A data set gathering 335 in situ measurements was used to develop his method,
which was validate using a synthetic data set (IOCCG) and 217 SeaWiFS match-ups data
set. Fichot [2004] further apply his method to SeaWiFS imagery to derive depth-resolved
and depth-integrated photoproduction of carbon monoxide (CO) at the global scale.
These methods are briefly described and discussed below.
Cullen et al. [1997] suggested an approach in several steps which makes use of
the SeaWiFS 412 nm sensor: (1) find empirical relationships between the ratio of
reflectance at two visible wavelengths (412 nm and 555 nm) and diffuse attenuation
coefficient (Kd) at several UV wavelengths; (2) calculate spectrally resolved total UV
absorption from Kd(UV); (3) determine the magnitude of absorption by particles,
particularly phytoplankton; (4) subtract absorption due to particles and water from the
190

total to calculate the CDOM absorption; and (5) apply an action spectrum to calculate
photochemical production. The estimates of absorption and attenuation could be
combined with surface irradiance to calculate absorption by CDOM as a function of
depth.
Johannessen et al. [2003] modified Cullen’s approach by finding direct,
empirical relationships between the Rrs(412)/Rrs(555) and Kd(UV), and then by using a
constant ratio between Kd(UV) and aCDOM(UV). These authors did not, however, used
their method to estimate the depth-integrated DIC photoproduction. But for her DIC
photoproduction calculation, Johannessen [2000] only used the magnitude of the
SeaWiFS-derived Kd to define three water classes for which φDIC(λ) were determined
[Johannessen and Miller, 2001]. Johannessen [2000] then computed the depth-integrated
DIC photoproduction in the wavelength range going from 300 to 450 nm assuming a
constant [aCDOM/at](λ) value of 1.0.
Fichot [2004] proposed two methods to estimated Kd at 320, 340, 380, 412, 443,
and 490 nm based on the principal components (PC) of the visible Rrs spectrum (at six
SeaWiFS bands 412, 443, 490, 510, 555 and 670 nm). In the first one (SeaUV), the log-
transformed Rrs spectrum is reduced into the first four PCs which are correlated to Kd at
different wavelengths through a multiple regression. Next, aCDOM(λ) is obtained assuming
a constant ratio between Kd(320) and aCDOM(320) and a spectral slope for the CDOM
absorption spectrum (SCDOM) of 0.0194 nm -1 . In the second method (SeaUVc), the water
is first classified using the first 2 PCs into one of the seven predefined classes. The Kd at
different wavelengths is then calculated empirically using the coefficients determined for
the given class. For each class, a different value for the ratio Kd(323)/aCDOM(323) is used
with SCDOM = 0.0194 nm -1 to obtain aCDOM(λ). With both aCDOM(λ) and Kd(λ), and further
assumptions on the light field distribution (i.e., the conversion of the planar into scalar
irradiance), Fichot [2004] estimated the depth-resolved photoproduction of carbon
monoxide (CO). However, the global maps of depth-integrated CO photoproduction
presented by Fichot [2004] follow the incident irradiance distribution and are mostly
independent of the water column optics. This is because the proposed method by Fichot
[2004] only captures part of the variability of the contribution of CDOM to the total light
absorption. There are some evidences suggesting limitations in Fichot’s method for the
191

estimation of the depth-integrated photoproduction of CO (or DIC). First, in large sub-
tropical ultraoligotrophic gyre, CDOM is almost impossible to measure and pure water is
likely completely dominating the full UV-visible [Morel et al., in press in L&O, 2006].
Thus the depth-integrated CO photoproduction should be much lower in those area.
Similarly, the depth-integrated CO photoproduction should be reduced in coastal waters
rich in particulate matter, and higher in river plumes where CDOM is concentrated at the
surface (e.g. Amazone and Oricono Rivers). Those general trends were not visible on the
maps presented by Fichot [2004], likely because of the assumptions made on the ratio
Kd(323)/aCDOM(323). Therefore, if one would like to adopt this method to estimate the
depth-integrated CO (or DIC) photoproduction, the validity of the above assumptions on
the ratio Kd(323)/aCDOM(323) adopted by Fichot [2004] need to be evaluated in more
details.
We proposed an alternative method to use Ocean Color data for photooxidation
quantification in which depth-integrated DIC photoproduction varies as function of
[aCDOM/at]. Because assuming constant proportion [aCDOM/at], as for example 0.90, may
lead to significant overestimation of the DIC photoproduction (Table VI.4), I recommend
to use either in situ or satellite-derived [aCDOM/at]. Note that a value of 0.9 may appear
unrealistically high, but based on in situ measurements of [aCDOM/at], this value is
generally valid for wavelength < 350 nm, where maximum photooxidation occurs (see
Fig. IV.7).
The difference between total annual DIC photoproduction calculated in the
Beaufort Sea using the satellite-derived and in situ [aCDOM/at] is not significant (i.e. less
than the algorithm accuracy). Because in situ measurements of [aCDOM/at] are not
available for every season and region of the Arctic Ocean, then the use of ocean color
data could be useful. If one would like to use ocean color data for DIC photoproduction
calculation, a particular attention should be paid to the data processing. First, it is
strongly recommended to process the satellite data on an image-by-image basis in order
to check the consistency of the water-leaving reflectance retrieval in the blue part of the
spectrum. Second, adjacency effect due to sea ice could be extremely important, in
particular at the beginning of the summer when a significant snow cover is still present.
192

VI.5. Conclusions
I conclude that the proposed method on the use of satellite-derived [aCDOM/at]
could improve the estimation of depth-integrated DIC photoproduction in regions where
in situ measurements are lacking. Nevertheless, good estimation can be obtained if
realistic assumption on [aCDOM/at] are made. Such assumptions, however, should be based
on a sufficient number in situ measurements to cover the spatio-temporal variability in
[aCDOM/at]. At the moment, in situ optical measurements are lacking for most of the
Arctic coastal area (e.g. Kara and Laptev seas).
The proposed method was tested using SeaWiFS imagery, but it could be applied
to any ocean color sensors having ~412, ~490 and ~555 nm channels (e.g., MERIS and
MODIS). Future improvements in the radiometric calibration and in the atmospheric
correction over turbid waters should simplify the application of the method to satellite
imagery.
One important parameter to take into account for the estimation of DIC
photoproduction is the variability in the apparent quantum yield. Beside providing
[aCDOM/at], Ocean Color data may be useful to predict the variability of φDIC(λ). Here, a
method was tested to account for the variability in φDIC(λ) based on the magnitude of the
CDOM absorption coefficient at 412 nm (Method 2). The lack of information on the
φDIC(λ) precludes any further attempt to model, or predict, its spatio-temporal variability
from other external environmental conditions (e.g. mixed layer vs photic layer depth,
temperature, turbidity, salinity, etc.). Therefore, the AQY needs to be extensively
documented and its variations understood. Beside the origin of CDOM, its light history,
including the role of vertical mixing, is certainly a major aspect of the problem.
VI.6. References
Bailey, S. W., and P. J. Werdell (2006), A multi-sensor approach for the on-orbit
validation of ocean color satellite data products, Remote Sens. Environ., 102, 12-
23.
Cullen, J. J., R. F. Davis, J. S. Barlett, and W. L. Miller (1997), Toward remote sensing
of UV attenuation, photochemical fluxes, and biological effects of UV in surface
193

waters, in Current and Emerging Issues in Aquatic Science: Aquatic Science
meeting, Am. Soc. Limnol. and Oceanogr., Santa Fe, N. M.
Fichot, C. G. (2004), Marine photochemistry from space: Algorithms for the Retreival of
Diffuse Attenuation and CDOM absorption Coefficients (320-490 nm) from
Ocean Color and Estimation of Depth-resolved photoproduction Rates of Carbon
Monoxide (CO) at Global Scales using SeaWiFS imagery, Master of Science
thesis, 217 pp., Dalhousie, Halifax, Canada.
Gordon, H. R. (1997), Atmospheric correction of ocean color imagery in the Earth
Observation System era, J. Geophys. Res., 102(D14), 17081-17106.
Gordon, H. R., and M. Wang (1994), Retrieval of water-leaving radiance and aerosol
optical thickness over the oceans with SeaWiFS: a preliminary algorithm, App.
Opt., 33(3), 443-452.
Johannessen, S. C. (2000), A photochemical sink for dissolved organic carbon in the
ocean,176 pp., Dalhousie University, Halifax.
Johannessen, S. C., and W. L. Miller (2001), Quantum yield for the photochemical
production of dissolved inorganic carbon in seawater, Mar. Chem., 76, 271-283.
Johannessen, S. C., W. L. Miller, and J. J. Cullen (2003), Calculation of UV attenuation
and colored dissolved organic matter absorption spectra from measurements of
ocean color, J. Geophys. Res., 108(C9), 3301, doi:10.1029/2000JC000514.
Lee, Z.-P., K. L. Carder, and R. A. Arnone (2002), Deriving inherent optical properties
from water color: a multiband quasi-analytical algorithm for optically deep waters,
App. Opt., 41(27), 5755-5772.
Morel, A., B. Gentili, H. Claustre, M. Babin, A. Bricaud, J. Ras, and F. Tièche (in press,
2006), Optical properties of the "clearest" natural waters, Limnol. Oceanogr.
Ohde, T., B. Sturm, and H. Siegel (2002), Derivation of SeaWiFS vicarious calibration
coefficients using in situ measurements in Case 2 water of the Baltic Sea, Remote
Sens. Environ., 80, 248-255.
Ruddick, K. G., V. De Cauwer, Y.-J. Park, and G. Moore (2006), Seaborne
measurements of near infrared water-leaving reflectance: The similarity spectrum
for turbid waters, Limnol. Oceanogr., 51(2), 1167-1179.
Ruddick, K. G., F. Ovidio, and M. Rijkeboer (2000), Atmospheric correction of SeaWiFS
imagery for turbid coastal and inland waters, App. Opt., 39(6), 897-895.
Siegel, D. A., M. Wang, S. Maritorena, and W. Robinson (2000), Atmospheric correction
of satellite ocean color imagery: the black pixel assumption, App. Opt., 39(21),
3582-3591.
194

Zibordi, G., F. Mélin, and J.-F. Berthon (2006), Comparison of SeaWiFS, MODIS and
MERIS radiometric products at a coastal site, Geophys. Res. Lett., 33, L06617,
doi:10.1029/2006GL025778.
195

Chapitre VII: Conclusion générale et perspectives
196

Le couvert de glace pluriannuel de l’Océan Arctique se rétréci en étendu et en
épaisseur, laissant derrière des étendues d’eau ouverte de plus en plus grandes. En parallèle,
au cours des dernières décennies, on a assisté à une augmentation significative du
rayonnement UV reçu à la surface de la mer en réponse à la diminution de l’ozone
stratosphérique. Le travail réalisé dans le cadre de cette thèse représente un effort afin
d’appréhender l’impact des ces changements environnementaux sur le cycle du carbone
organique dissous (DOC) dans l’Océan Arctique. En particulier, je me suis intéressé au
carbone d’origine terrigène (tDOC) car il représente une importante fraction du pool de
DOC dans les eaux côtières de l’Océan Arctiques et que la fonte du pergélisol risque
d’augmenter grandement cet apport dans le futur. Afin de mieux comprendre le rôle relatif
que joue la lumière dans le cycle du DOC mon objectif principal était d’estimer la quantité
de carbone organique qui est minéralisée à la surface de l’océan par les processus de
photooxydation. Grâce à un modèle couplé optique/photochimique simple nécessitant en
entré de l’information dérivée d’observations satellitales, j’ai quantifié pour la première fois la
production photochimique de carbone inorganique dissous (DIC) dans une région de
l’Arctique.
Les conclusions majeures de ce travail sont que : 1) la photoproduction de DIC dans
le Sud-est de la Mer de Beaufort a augmenté de ~15% au cours de la période allant de 1979 à
2004, principalement en réponse à la diminution de l’étendue de glace de mer, et 2) les taux
de minéralisation du tDOC sont pratiquement équivalents aux taux actuels d’enfouissement
de carbone organique particulaire dans les sédiments marins. Cette dernière conclusion
confirme l’importance relative de ce flux pour la balance nette du Bilan de Carbone
Organique Arctique. En raison de l’intensité de sa stratification, les eaux de surface arctiques
sont généralement riches en tDOC et pauvres en nutriments inorganiques, expliquant en
partie sa faible production primaire (e.g., couche euphotique plus mince due à la présence de
CDOM, mélange verticale faible, etc.). Si la minéralisation de carbone organique allochtone
(i.e., terrigène) par les processus de « photooxydation/respiration bactérienne » est plus
grande que la séquestration de carbone organique autochtone (i.e. marin) dans les sédiments
profonds, alors l’Océan Arctique doit être considéré comme une source nette de CO 2 pour
la biosphère (i.e. hétérotrophie nette). La question à savoir si l’Océan Arctique agit comme
un puits ou une source de carbone reste entièrement ouverte à ce jour.
197

Cette thèse représente un premier pas vers une meilleure compréhension du rôle que
joue la photooxydation du CDOM dans le bilan de carbone organique de l’Arctique. Elle
fournit aussi des pistes quant à l’utilisation simultanée de paramètres divers dérivés des
données acquises depuis des plateformes satellitales pour quantifier, à de larges échelles
spatiales, des processus physico-chimiques telle que la photooxydation du CDOM. Dans ce
qui suit, sont discutés certains aspects liés à la télédétection de la couleur de l’Océan, et à la
quantification de la photooxydation dans les eaux de l’Océan Arctique.
Vers une estimation semi-analytique du coefficient d’absorption du CDOM à partir
de la couleur de l’océan
Puisque la modélisation de la photooxydation du CDOM nécessite la connaissance
de certaines propriétés optiques de l’eau de mer, un effort particulier a été mis sur
l’extraction des données de couleur de l’océan des paramètres bio-optiques pertinents. Or
dans la littérature, les algorithmes semi-analytiques d’inversion proposés ne permettent pas
de distinguer les coefficients d’absorption du CDOM et des particules non-algales (NAP).
Ceci s’explique par la similarité de leurs spectres d’absorption respectifs. J’ai donc proposé
un simple algorithme empirique permettant de retrouver directement la contribution du
CDOM au coefficient d’absorption totale ([a CDOM/a t]) directement à partir du spectre de
réflectance de l’eau.
Cependant, les relations empiriques peuvent empêcher notre compréhension des
fondements physiques, et par conséquent, limiter les futurs avancements dans le domaine de
la télédétection de la couleur de l’océan (IOCCG, 2006). Pour cette raison, les approches
semi-analytiques basées sur les fondements physiques de l’optique hydrologique devraient
êtres préférées aux méthodes empiriques. Or pour estimer a CDOM à l’aide d’algorithmes semi-
analytiques, les modèles bio-optiques se doivent d’être formulés de manière différente de
celle présentement adoptée. Le fait que, contrairement au CDOM, les NAP diffusent la
lumière offre un pouvoir discriminant lorsque ceux-ci sont considérés individuellement dans
la formulation du modèle d’IOPs. Ainsi, pour que les algorithmes semi-analytiques soient
performant dans la discrimination de a CDOM et a NAP, ils devront se baser sur un modèle
d’IOPs qui inclura un lien robuste entre les propriétés de rétrodiffusion et d’absorption des
NAP (e.g. Gallegos et Neale, 2002). Une difficulté majeure d’un tel paramétrage est qu’il
n’est présentement pas possible de mesurer distinctement les propriétés de rétrodiffusion des
198

particules algales et non-algales. Le succès de notre algorithme empirique suggère néanmoins
que ce type de paramétrage est possible dans certain environnement (e.g. Mer Baltique et
Mer du Nord).
Un autre aspect du problème lié à l’utilisation des modèles semi-analytiques concerne
la méthode avec laquelle l’information contenue dans le spectre de l’eau est extraite. Dans les
eaux côtières, les variations dans les spectres de réflectance sont principalement dues aux
variations de la turbidité de l’eau (e.g., Sathyendranath et al., 1989). Les subtiles variations
pouvant résultées des changements de proportions entre le phytoplancton, le CDOM et les
NAP, sont « camouflés » par les changements d’amplitude de la réflectance causés par les
variations dans la turbidité à toutes les longueurs d’onde. Par conséquent, lorsqu’une
information bien spécifique se doit d’être dérivée d’un spectre de réflectance, comme par
exemple [a CDOM/a t], seules les parties du spectre fournissant l’information la moins ambiguë
possible se doivent d’être considérées par le modèle d’inversion. La justification du choix
des longueurs d’ondes utilisées dans notre algorithme empirique est donc tout autant valable
pour les algorithmes semi-analytiques.
Est-ce que les effets d’environnement dus à la présence de glace de mer peuvent être
corrigés?
La télédétection spatiale de la couleur de l’océan peut fournir des informations utiles
pour la quantification des processus photochimiques dans les eaux de l’Arctique souvent
inaccessibles. Mais l’utilisation des données de couleur de l’océan dans les hautes latitudes est
compliquée par la présence de glace de mer. Au début de la période estivale, près de la moitié
des eaux ouvertes observéee depuis l’espace sont contaminées par les effets d’environnement.
Ce phénomène comprommets sévèrement l’utilisation des données de couleur de l’océan
dans les mers polaires où la glace est omniprésente. Par conséquent, il serait souhaitable de
mettre au point une méthode permettant de corriger les effets d’environement. Ceci pourrait
être fait en générant des tables pré-calculées (LUTs) additionnelles pour lesquelles les effets
d’environnement seraient inclus. Dans ce cadre, une étude plus approfondie de la variabilité
de la réflectance de la glace mer, aux échelles correspondant aux mesures de couleur de
l’océan (i.e. ~1 Km de résolution), serait utile pour définir les types de glace les plus
fréquents dans l’Arctique. Pour chaque type de glace, une LUT pourrait être générée. La
réflectance de la glace pourrait éventuellement être mesurée à partir du même instrument, et
199

ainsi le type de glace identifié depuis l’espace. Finalement, ces LUTs pourraient être utilisées
dans le traitement des données uniquement lorsque des pixels contaminés par les effets
d’environnement sont détectés avec la méthode proposée dans cette thèse.
À propos de la variabilité naturelle du rendement quantique apparent pour la
photoproduction de DIC, et de sa détermination
Le φ DIC est un paramètre fondamental pour la modélisation de la photooxydation du
CDOM. Il est, à l’heure actuelle, de loin le paramètre du modèle de photooxydation le moins
bien documenté. Quelle est la représentativité des spectres de φ DIC mesurés dans les eaux du
Fleuves Mackenzie, du Plateau du Mackenzie, et du Golfe d’Amundsen, du pool de CDOM
de l’Arctique? Pour étendre notre étude régionale à l’échelle de l’Océan Arctique, il est
nécessaire de déterminer φ DIC dans les autres régions côtières de l’Arctique. En général, on
s’attend à ce que la composition de la matière organique dissoute (DOM) provenant des
quatre plus grands fleuves arctiques (i.e, Ob, Yenisei, Lena and Mackenzie) diffère
significativement. D’une part, les structures phénoliques de la lignine (Benner et al., 2005), les
acides aminées et les sucres neutres (Amon et Benner, 2003), les propriétés de fluorescence
(Amon et al., 2003; Retamal et al., 2006), tous indiquent des différences dans la composition
moléculaire de la DOM provenant ces différents systèmes. D’autre part, les caractéristiques
physiques et chimiques du Fleuve Mackenzie semblent être particulièrement différentes
comparées aux autres grands fleuves sibériens : les plus frappantes sont ces fortes
concentrations en particules en suspension (> 4.3 fois; Rachold et al., 2004), son alcalinité et
son pH élevés (e.g., Millot et al., 2003). Sous des conditions aussi différentes, on peut
s’attendre à des différences significatives quant à la photoréactivité du CDOM entre ces
environnements côtiers. Comme il a été mentionné précédemment, l’augmentation attendue
de la photominéralisation du tDOC est importante pour déterminer si l’Océan Arctique agira
comme un puits ou une source de CO 2 pour la biosphère dans le futur. Pour mieux évaluer
cet impact, il est nécessaire d’estimer la proportion relative de CDOM d’origine terrigène
versus marine, ainsi que de connaître leur photoréactivité respective. La proportion de
chacun de ces composants est actuellement inconnue. Dans cette étude, on a mesuré φ DIC de
l’ensemble du pool de CDOM. Il nous a donc par été possible d’estimer avec précision
combien de DIC fut réellement produit suite à la photooxydation du tDOC. Par conséquent,
200

il semble que φ DIC doit être déterminé en parallèle avec une caractérisation moléculaire du
pool de CDOM.
Outre la composition moléculaire du CDOM, la dose de lumière absorbée par une
molécule de CDOM précédent son échantillonnage (i.e. son « histoire lumineuse ») est
certainement importante pour déterminer sa photoréactivité. Comme il est pratiquement
impossible de connaître l’histoire lumineuse d’un échantillon récolté en mer à un temps t,
l’effet de l’histoire lumineuse sur φ DIC se doit d’être évaluée en laboratoire. Les études sur les
rendement quantique apparent (AQYs) autres que pour le CO 2 fournissent des pistes quant à
l’effet de l’histoire lumineuse sur la photoréactivité du CDOM. Par exemple, Andrews et al.
(2000) rapportaient une diminution des AQYs pour la consommation photochimique d’O 2
et la photoproduction de H 2O 2, alors que Zhang et Xie (2006) observaient des tendances
similaires pour les AQYs pour la photoproduction de CO. Cependant, l’effet de l’histoire
lumineuse sur φ DIC est inconnu, et les résultats trouvés dans la littérature sont conflictuels.
Vähätälo et Wetzel (2004) concluaient que les AQYs pour la perte en DOC (i.e. ~φ DIC)
diminuaient en fonction de la dose de lumière absorbée par le CDOM. À l’opposé,
Johannessen et Miller (2001) ont trouvé que le φ DIC avait pratiquement doublé après le
photo-blanchissement intensif de l’échantillon (i.e., diminution de a CDOM de > 50%). Puisse
qu’il est impossible de tirer des conclusions claires de ces observations, plus d’études sont
nécessaires afin de déterminer l’effet de l’histoire lumineuse sur φ DIC.
Parmi les autres aspects qui demandent plus d’étude, l’impact de l’acidification que
doivent subir les échantillons avant leurs incubations semble être cruciale. Cette étape est
essentiel pour éliminer le DIC initial contenu dans l’échantillon qui masquerait totalement la
photoproduction durant les incubations. Une façon d’évaluer les effets de l’acidification est
de regarder si des changements dans le AQY pour la photoproduction de CO (i.e. un proxy
pour φ DIC) sur des échantillons traités de la même manière que ceux pour la détermination de
φDIC. Si l’acidification induit des changements majeurs dans les valeurs de CO-AQY, on peut
donc s’attendre à des changements similaires de φDIC. Une autre méthode pourrait être de
travailler avec des échantillons provenant de lacs où les concentrations en DIC sont
naturellement faibles. Ainsi, l’impact de l’acidification pourrait être évalué directement sur
φDIC. À côté de l’acidification, les effets de l’intensité de la lumière utilisé durant les
incubations et le stockage des échantillons pendant de longue période de temps (i.e. > 4
201

mois) peuvent aussi avoir un impact sur l’amplitude de φ DIC. Il est clair que ces aspects
méthodologiques sur la détermination de φ DIC méritent plus d’attention.
Enfin, si on arrive à mieux comprendre quels sont les paramètres biologiques,
chimiques et physiques qui contrôlent φ DIC, ainsi il sera possible d’estimer, ou d’approximer,
sa variabilité à partir d’observations environnementales.
Sur la modélisation de φφφφ DIC à partir d’informations dérivées des données satellitales
Le plein potentiel des observations satellitales pour la quantification de la
photoproduction de DIC n’a pas été totalement exploré. En plus de fournir le rapport entre
le a CDOM et a t, les données de couleur de l’océan peuvent servir à modéliser la variabilité de
φ DIC. Par exemple, il semble qu’une relation entre φ DIC et a CDOM(412) pourrait exister (Chap.
VI). D’autres produits dérivés des données de couleur de l’océan, comme la chlorophylle et
le coefficient d’atténuation diffus, peuvent certainement fournir des informations clés. Alors
que la chlorophylle est un proxy de la productivité primaire, et donc de la production de
CDOM marin, l’atténuation diffuse peut servir à déterminer la profondeur de la couche
euphotique dans laquelle la photoproduction de DIC prend place.
Un modèle prédictif de φ DIC, basé sur des observations satellitales, doit probablement
être développé en parallèle avec un modèle de mélange vertical. En effet, les taux de mélange
verticaux et la profondeur de la couche euphotique doivent être connus pour évaluer
l’histoire lumineuse du CDOM, et ainsi pour prédire les changements de φ DIC. Dans un
premier temps, la profondeur de la couche mélangée peut être dérivée des données
climatologiques de température et salinité (e.g. Steele et al., 2001). Mais un modèle de
mélange vertical utilisant les forcages dérivés d’observations satellitales (vitesse de vent,
température de surface, éclairement solaire, glace de mer, etc.) pourait être développé dans le
futur pour l’Océan Arctique. Encore une fois, le manque d’information sur φ DIC empêche,
pour le moment, toute tentative de modélisation de sa variabilité spatio-temporelle à partir
d’autres observations environnementales.
Implication des changements climatiques sur la minéralisation du Carbone
Organique Terrigène dans l’Océan Arctique
202

La couche mélangée polaire Arctique (PML) contient les plus fortes concentrations
de tDOC par rapport aux autres océans. Ceci est en partie le résultat de la grande quantité
d’apport terrigène qui alimente la PML par rapport à son volume. Une autre explication est
la présence de ce couvert de glace permanent qui réduit fortement la pénétration des UV
dans la colonne d’eau, limitant ainsi la dégradation du tDOC par les processus de
photooxydation.
Selon plusieurs études récentes (e.g., Opsahl et al., 1999; Amon et al., 2003; Hansell et
al., 2004), environ la moitié des apports annuels en tDOC dans la PML sont exportés vers
l’Océan Atlantique Nord via les Détroit de Fram et l’Archipel Canadien. D’autres évidences
récentes suggèrent qu’une fraction significative de ce matériel est ensuite exportée hors de la
surface de l’océan durant la formation de la masse d’Eau Profonde Nord Atlantique
(NADW) (e.g., Benner et al., 2005). Dans les profondeurs de l’océan, le temps de résidence
du tDOC est de plusieurs centaines d’années (Hernes et Benner, 2006). À l’opposé, la
minéralisation du tDOC dans la couche de surface est très rapide en raison de l’efficacité des
processus de photooxydation et de respiration microbienne (e.g. Kieber et al. 1990; Miller et
Zepp 1995; Hernes et Benner 2003), et ainsi ce carbone peut en partie retourner en dans
l’atmosphère sous forme de CO 2.
En me basant sur ces observations, je suggère que la réduction du couvert de glace
pluriannuel accélèrera la minéralisation du tDOC dans la couche de surface Arctique,
réduisant l’export de ce carbone dans les eaux profondes océaniques, et augmentant la
quantité de CO 2 pouvant retourner dans l’atmosphère. Cependant, comme il a été suggéré
par Gobeil et al. (2001), la réduction du couvert de glace peut en contre partie augmenter la
productivité primaire et la séquestration de carbone organique dans l’Océan Arctique. Par
conséquent, l’augmentation prévue de l’oxydation du tDOC en surface suite à une
diminution de la glace pourrait être compensée, et peut être même surpassée en terme de
flux de carbone, par l’augmentation de la production primaire. Dans mes travaux futurs, les
conditions de glace prévues par les modèles globaux de circulation générale Atmosphère-
Ocean (ACIA, 2005) serviront à évaluer la balance nette entre les processus stimulés par la
lumière, c’est-à-dire la photooxydation du tDOC et la production primaire.
203

VII.A. Version anglaise: Overall Conclusion and Perspectives
The permanent sea ice cover of the Arctic Ocean is currently experiencing an
important reduction in its extent and thickness, leaving behind increasingly wider area of
open water. In parallel, the UV radiation reaching the sea surface has increased
significantly during the last decade as a response of the stratospheric ozone depletion.
The work achieved in this thesis represents an effort to evaluate the consequence of these
environmental changes on the cycling of dissolved organic carbon (DOC) in the Arctic
Ocean. In order to better understand relative role of the light into the DOC cycling, in
particular the terrigenous component (tDOC) as it represent a significant fraction of the
DOC pool in the coastal Arctic Ocean, my main objective was to estimate how much
organic carbon is mineralized in the surface waters through photooxidation. Using a
simplified model that uses as inputs information derived from satellite remote sensing
instruments, I quantified for the first time the photoproduction of dissolved inorganic
carbon in a region of the Arctic.
The major conclusions of the present work are that 1) the DIC photoproduction in
the southeastern Beaufort Sea has increased over the period from 1979 to 2004 by ~15%,
mainly in response to the decreasing sea ice extent, and 2) the estimated rate of tDOC
mineralization is nearly equivalent to that of particulate marine organic carbon currently
buried into the deep sediments. The latter conclusion confirms the relative importance of
this flux in the net balance of the Arctic Organic Carbon Budget. Due to their strong
stratification, the surface Arctic waters are generally rich in tDOC and poor in inorganic
nutrients, which results in low primary production (e.g., thinner euphotic zone due to
high CDOM, weak mixing, etc.). If the mineralization of allochthonus organic carbon (i.e.
terrigenous) through photo-oxidation/bacterial respiration is higher than the sequestration
of autochtonous organic carbon (i.e. marine) in the deep sediment, then the Arctic Ocean
should be consider as a net source of CO2 for the biosphere (i.e. net heterotrophic).
Whether the Arctic acts as a net source or sink of carbon still remains an open question.
This thesis represents a first step toward a better understanding of the role of
CDOM photooxidation in the Arctic Ocean Organic Carbon Budget. It also provides
valuable insights on how various satellites remote sensing information can be merged to
204

quantify large scale processes such as CDOM photooxidation. Below, a number of
unresolved issues related to the ocean color remote sensing and to the quantification of
CDOM photooxidation in the Arctic Ocean are discussed.
Toward semi-analytical estimation of CDOM absorption coefficient from ocean
color
Because modeling CDOM photooxidation involved in-water optical properties, a
particular effort was made to extract relevant bio-optical parameters from satellite Ocean
Color data. The current semi-analytical ocean color inversion algorithms found in the
literature do not distinguish the CDOM and non-algal particles (NAP) absorption
coefficients because of the similarity in their respective spectral shapes. I proposed a
simple empirical algorithm to retrieve the contribution of CDOM to the total absorption
coefficient ([aCDOM/at]) from the remote sensing reflectance.
Simple empirical relationships may prevent understanding of the basics and,
therefore, limit advancement in ocean color remote sensing [IOGGC, 2005]. For this
reason semi-analytical approaches based on fundamentals of hydrological optics should
be preferred. To derive aCDOM using semi-analytical algorithms, however, the bio-optical
model must be formulated in a way different from that presently adopted. The fact that
NAP scatters light, and CDOM do not, provides discrimination power when CDOM and
NAP are distinguished in an IOP model. So, for future semi-analytical algorithms to be
successful in discriminating aCDOM from aNAP using water reflectance, they will need to be
based on an IOP model that includes a good link between aNAP and bNAP (or bbNAP) [e.g.,
Gallegos and Neale, 2002]. One major difficulty is that the particles backscattering
measurements do not make the difference between the algal and non-algal particles. The
success of our empirical algorithm suggests that regional parameterization of the NAP
optical properties is possible in certain environmental conditions (e.g. Baltic Sea and
North Sea).
The necessary information contained in the reflectance spectrum to discriminate
aCDOM from aNAP, however, must be extracted properly, and this is probably one weakness
of current semi-analytical models. In coastal waters, variations in reflectance are
primarily due to variations in turbidity [Sathyendranath et al., 1989]. Subtle changes in
205

the shape of the reflectance spectrum that may result from changes in the proportions of
phytoplankton, CDOM and NAP, are overwhelmed by the change in the magnitude of
reflectance at all wavelengths due to variations in turbidity. Also, when some specific
information is needed from the reflectance spectrum, such as the ratio of aCDOM to at, only
the spectral regions that provide less ambiguous information must be considered in the
inversion. The rationales considered in the development of our empirical algorithm are
valid for a semi-empirical algorithm.
Is the correction of the adjacency effect due to sea ice possible?
Ocean Color remote sensing can provide valuable information for the
quantification of photochemical processes in the remote Arctic waters. But the use of
ocean color remote sensing at high latitude is complicated by the presence of sea ice.
During spring time, more than the half of the open water area observed from space is
contaminated by the adjacency effect. This phenomenon severely compromises the use of
ocean color data over ice covered seas. Therefore, an extension of the atmospheric
correction scheme to include a correction for adjacency effect is highly desirable. This
might be done by generating additional look-up-tables (LUTs) that would account for the
adjacency effect. For this purpose, further study dedicated to the variability of the sea ice
reflectance would be necessary to define a number of typical Arctic sea ice reflectance
spectra. For each type of ice, a LUT could be generated. In turns the sea ice reflectance
could be measures using the same Ocean Color instrument, and thus the type of ice
obtained from space for the correction. Those additional LUTs could be used in the
processing only when contaminated pixels are detected, for instance, using the flagging
approach developed in this thesis.
On the natural variability of the apparent quantum yield for the DIC
photoproduction, and its determination
The φDIC is fundamental for modeling CDOM photooxidation and is definitely the
least documented parameter of photooxidation models. How representative are the φDIC
spectra for the Mackenzie River and Shelf and from the Amundsen Gulf, of the whole
Arctic CDOM pool? To extend our regional study to the whole Arctic Ocean, it is
206

necessary to determine φDIC in other regions of the coastal Arctic. In general, the
composition of DOM from the four major Arctic river systems (i.e. the Ob, Yenisei, Lena
and Mackenzie) is expected to differ to some extent. For example, the lignin phenols
[Benner et al., 2005], amino acids and neutral sugars analysis [Amon and Benner, 2003]
and excitation/emission fluorescence spectrometry [Amon et al., 2003; Retamal et al.,
2006] indicate varying composition of the DOM in these river systems. In addition, the
Mackenzie River is also very different in terms of its physical and chemical environments:
the most striking differences are its high concentration of particulate matter [> 4.3-fold
the Siberian Rivers; Rachold et al., 2004], high alkalinity and pH [e.g., Millot et al.,
2003]. Under these different conditions, one should expect important differences in their
CDOM photoreactivity. As mentioned above, the expected increase in tDOC
photomineralization is important to determine if the Arctic Ocean will act as a net source
or sink of CO2 for the biosphere in the future. To better evaluate this impact, it is
necessary to assess the relative proportion of terrigenous versus marine CDOM, as well
as their respective photoreactivity. The exact proportion each component is actually
unknown. In this study, we measured the φDIC of the whole CDOM pool. We could not
assess precisely how much of the DIC photoproduced came actually from the terrigenous
component. Therefore, the φDIC needs to be documented in parallel with molecular
characterization of the CDOM pool.
Beside the molecular composition of CDOM itself, the light history of CDOM is
certainly important for determining its photoreactivity. The light history refers to the light
dose received by the CDOM prior to sampling. Because it is almost impossible to know
the light history of water sample at the time of sampling, role of light history in the φDIC
variations should be studied in the laboratory. The studies of the Apparent Quantum
Yield (AQY) other than CO2 provided some insights regarding the impact of the light
history on the CDOM photoreactivity. For example, Andrews et al. [2000] reported
decreasing photochemical O2 consumption AQY and H2O2 photoproduction with the
light dose absorbed by CDOM. Recently, Zhang and Xie [2006, in press in Environ. Sci.
Technol.] reported a similar trend for CO photoproduction AQY. However, the effect of
the light history on φDIC is still unknown, and the results found in the literature are
conflictual. Vähätälo and Wetzel [2004] concluded that DOC loss AQY (i.e. ~φDIC)
207

decreases with increasing aborbed dose. In contrast, Johannessen and Miller [2001]
found that the φDIC nearly doubled after the sample was extensively photobleached (i.e., >
50 %). Therefore, no clear conclusions can be drawn from the above observations. Hence,
further studies are required to clarify the effect of light history of CDOM on φDIC. Again,
to reduce the uncertainty on variability of φDIC resulting from CDOM light history in the
modeling DIC photoproduction, an extensive data set on φDIC is necessary.
Among the other aspects that require more investigation is the impact of
acidification applied to the sample prior the incubation for φDIC determination. This step
was essential to remove the background in DIC that would mostly mask the DIC
photoproduction during the experiment. One way to assess the effects of acidification is
to look at the changes in the carbon monoxide (CO) AQY (i.e. as a proxy of φDIC) on
samples treated with the same manner as DIC ones. If acidification induces major
changes in the CO AQY, then it likely affects the φDIC as well. Another method would be
first to determine φDIC of a fresh water sample that contains naturally small amount of
DIC (i.e. in a lake), and then, after a 24-h acidification and resetting the original pH, re-
assess φDIC. In addition to the acidification, the effect of light intensity used during the
incubation and sample storage may also have an impact on the magnitude of φDIC. It is
clear that these experimental aspects for the φDIC determination deserve more study.
Finally, if the main biological, chemical and physical parameters controlling φDIC
are better understood, then its variability may be assessed, or approximated, from
environmental observations.
On the modeling of φφφφDIC from space-derived information
The potential of satellite remote sensing observations to quantify DIC
photoproduction has not been fully explored. Besides providing the contribution of
CDOM to the total light absorption coefficient, ocean color data may be used to constrain
the variability of φDIC. For example, it appears that a relationship between φDIC(λ) and the
magnitude of the CDOM absorption coefficient at 412 nm might exist (Chap. VI). Other
ocean color products like the chlorophyll a concentration and the diffuse attenuation
coefficient can certainly provide key information. While chlorophyll is a proxy of
208

primary productivity and thus the production of fresh marine CDOM, the diffuse
attenuation can be used to assess the depth of the photic layer within which DIC
photoproduction takes place.
But a predictive model of φDIC based on satellite-derived parameters probably
needs to be developed in parallel with a vertical mixing model. Both the rate of vertical
mixing and the photic layer depth must be known to assess the light history of CDOM,
and thus to predict changes in the φDIC. At a first approximation, mixed-layer depth may
be derived from temperature and salinity climatology [e.g., Steele et al., 2001]. But a
vertical mixing model using as input satellite observations (surface winds, sea surface
temperature, solar radiation, sea ice, etc) may also be developed in the future for the
Arctic Ocean. Again, the lack of information on the φDIC precludes any attempt to model
or predicts its spatio-temporal variability from observations of the environmental
conditions.
Implication of Climate Change on the mineralization of the Terrigenous Dissolved
Organic Carbon in the Arctic
The Arctic Polar Mixed Layer (PML) contains the highest concentration of tDOC
relative to the other ocean basins. This is mainly a result of the large amount of riverine
input that enter the Arctic Ocean relative to its volume. Another explanation may be the
presence of a permanent sea ice cover strongly reduces the UV radiation penetration in
the water column, limiting the degradation of tDOC by photooxidation processes.
According to a number of recent studies [e.g. Opshal et al., 1999; Amon et al.,
2003; Hansell et al. 2004], about one half of the annual input of tDOC into the PML is
exported to the North Atlantic Ocean via Fram Strait and the Canadian Archipelago.
Recent evidences further suggest that a significant fraction of this material is further
exported from the ocean surface during the formation of the North Atlantic Deep Water
(NADW) [e.g., Benner et al., 2005]. Interestingly, the residence time of the tDOC in the
deep ocean may be of order several centuries [Hernes and Benner, 2006]. In contrast, in
the surface layer, tDOC is rapidly mineralized due to photooxidation and microbial
processes [e.g., Kieber et al. 1990; Miller and Zepp, 1995; Hernes and Benner, 2003] and
thus partly released to the atmosphere as CO2.
209

Based on these observations, I conclude that the reduction in the sea ice cover will
accelerate the mineralization of tDOC within the surface layer of the Arctic Ocean,
reducing the export of organic carbon to the deep ocean, and increasing the amount of
CO2 potentially released to the atmosphere. However, as suggested by Gobeil et al.
[2001], the reduction in sea ice cover could also increase primary productivity and carbon
sequestration in the Arctic Ocean. Therefore, the expected increase in tDOC oxidation
resulting from the decline in sea ice could be compensated for, and perhaps even
overtaken, by the increase in primary production in terms of carbon fluxes. In my future
work, the sea cover conditions predicted by the global coupled Atmosphere-Ocean
General Circulation Models (AOGCMs) [ACIA, 2005] will be used to assess the net
balance between light-related carbon fluxes, i.e. tDOC photooxidation and primary
production.
210

Bibliographie générale
211

Aagaard, K., and E. C. Carmack (1989), The role of sea ice and other fresh water in the
Arctic circulation, J. Geophys. Res., 94(C10), 14485-14498.
Aagaard, K., L. K. Coachman, and E. C. Carmack (1981), On the halocline of the Arctic
Ocean, Deep Sea Res., 28A(6), 529-545.
ACIA (2005), Arctic Climate Impact Assessment, 1042 pp., Cambridge Press University.
Amon, R. M. W. (2004), The role of dissolved organic matter for the organic carbon cycle
in the Arctic Ocean, in The organic carbon cycle in the Arctic Ocean, edited by R. Stein, and
R. W. MacDonald, pp. 83-99, Springer, Berlin.
Amon, R. M. W., and R. Benner (1996), Photochemical and microbial consumption of
dissolved organic carbon and dissolved oxygen in the Amazon River System, Geochim.
Cosmochim. Acta, 60(10), 1783-1792.
Amon, R. M. W., and R. Benner (2003), Combined neutral sugars as indicators of the
diagenetic state of dissolved organic matter in the Arctic Ocean, Deep Sea Res., 50(1),
151-169.
Amon, R. M. W., G. Budéus, and B. Meon (2003), Dissolved organic carbon distribution
and origin in the Nordic Seas: Exchanges with the Arctic Ocean and North Atlantic, J.
Geophys. Res., 108, 3221, doi: 10.1029/2002JC001594.
Amon, R. M. W., and B. Meon (2004), The biogeochemistry of dissolved organic matter
and nutrients in two large Arctic estuaries and potential implications for our
understanding of the Arctic Ocean system, Mar. Chem., 92, 311-330.
Anderson, L. G. (2002), DOC in the Arctic Ocean, in Biogeochemistry of Marine Dissolved
Organic Matter, edited by D. A. Hansell, and C. A. Carlson, pp. 665-683, Academic
Press, San Diego.
Anderson, L. G., E. P. Jones, and J. H. Swift (2003), Export production in the central
Arctic Ocean evaluated from phosphate deficits, J. Geophys. Res., 108(C6).
Anderson, L. G., K. Olsson, and M. Chierici (1998), A carbon budget for the Arctic Ocean,
Global Biogeochem. Cycles, 12(3), 455-465.
Andrews, S. S., S. Caron, and O. C. Zafiriou (2000), Photochemical oxygen consumption
in marine waters: A major sink for colored dissolved organic matter?, Limnol. Oceanogr.,
45(2), 267-277.
Antoine, D., and A. Morel (1999), A multiple scattering algorithm for atmospheric
correction of remotely sensed ocean colour (MERIS instrument): principle and
implementation for atmospheres carrying various aerosols including absorbing ones,
Int. J. of Remote Sens., 20(9), 1875-1916.
Antoine, D., and A. Morel (1996), Oceanic primary production : I. Adaptation of a spectral
light-photosynthesis model in view of application to satellite chlorophyll observations,
Global Biogeochem. Cycles, 10(1), 43-55
Antoine, D., A. Morel, and J. M. André (1996), Oceanic primary production : II.
Estimation at global scale from satellite (Coastal Zone Color Scanner) chlorophyll,
Global Biogeochem. Cycles, 10(1), 57-69.
212

Arnell, N. W. (2005), Implications of climate change for freshwater inflows to the Arctic
Ocean, J. Geophys. Res., 110(D7).
Arrigo, K. R., and G. L. van Dijken (2004), Annual cycles of sea ice and phytoplankton in
Cape Bathurst polynya, southeastern Beaufort Sea, Canadian Arctic, Geophys. Res. Lett.,
31, L088304.
Arrigo, K. R., D. Worthen, A. Schnell, and M. P. Lizotte (1998), Primary production in
Southern Ocean waters, J. Geophys. Res., 103(C8), 15587-15600.
Arrigo, K. R., D. L. Worthen, and D. H. Robinson (2003), A coupled ocean-ecosystem
model of the Ross Sea : 2. Iron regulation of phytoplankton taxonomic variability and
primary production, J. Geophys. Res., 108(C7), 3231, doi:10.1029/2001JC000856.
Babin, M., and D. Stramski (2002), Light absorption by aquatic particles in the nearinfrared
spectral region, Limnol. Oceanogr., 47(3), 911-915.
Babin, M., A. Morel, V. Fournier-Sicre, F. Fell, and D. Stramski (2003a), Light scattering
properties of marine particles in coastal and open ocean waters as related to the
particle mass concentration, Limnol. Oceanogr., 48(2), 843-859.
Babin, M., D. Stramski, G. M. Ferrari, H. Claustre, A. Bricaud, G. Obolensky, and N.
Hoepffner (2003b), Variations in the light absorption coefficients of phytoplankton,
nonalgal particles, and dissolved organic matter in coastal waters around Europe, J.
Geophys. Res., 108(C7), 3211, doi: 10.1029/2001JC000882.
Bailey, S. W., and P. J. Werdell (2006), A multi-sensor approach for the on-orbit validation
of ocean color satellite data products, Remote Sens. Environ., 102, 12-23.
Barber, D. G., and J. M. Hanesiak (2004), Meteorological forcing of sea ice concentrations
in the southern Beaufort Sea over the period 1979 to 2000, J. Geophys. Res., 109(C6),
C06014, doi:10.1029/2003JC002027.
Barnard, A. H., W. S. Pegau, and J. R. V. Zaneveld (1998), Global relationships of the
inherent optical properties of the oceans, J. Geophys. Res., 103(C11), 24955-24968.
Bauch, D., P. Schlosser, and R. G. Fairbanks (1995), Freshwater balance and the sources
of deep and bottom waters in the Arctic Ocean inferred from the distribution of H2 Prog. Oceanogr., 35, 53-80.
Behrenfeld, M. J., E. Boss, D. A. Siegel, and D. S. Shea (2005), Carbon-based ocean
productivity and phytoplankton physiology from space, Global Biogeochem. Cycles, 19,
GB1006, doi:10.1029/2004GB002299.
Behrenfeld, M. J., and P. G. Falkowski (1997), Photosynthetic rates derived from satellitebased
chlorophyll concentration, Limnol. Oceanogr., 42(1), 1-20.
Bélanger, S., H. Xie, N. Krotkov, P. Larouche, W. F. Vincent, and M. Babin (2006),
Photomineralization of terrigenous dissolved organic matter in Arctic coastal waters
from 1979 to 2003: Interannual variability and implications of climate change, Global
Biogeochem. Cycles, 20, GB4005, doi:10.1029/2006GB002708.
Benner, R., B. Benitez-Nelson, K. Kaiser, and R. M. W. Amon (2004), Export of young
terrigenous dissolved organic carbon from rivers to the Arctic Ocean, Geophys. Res.
Lett., 31, L05305.
213
18 O,

Benner, R., P. Louchouarn, and R. M. W. Amon (2005), Terrigenous dissolved organic
matter in the Arctic Ocean and its transport to surface and deep waters of the North
Atlantic, Global Biogeochem. Cycles, 19, GB2025, doi:10.1029/2004GB002398.
Benner, R., and S. Opsahl (2001), Molecular indicators of the sources and transformations
of dissolved organic matter in the Mississippi river plume, Org. Geochem., 32(4), 597-611.
Billett, M. F., S. M. Palmer, D. Hope, C. Deacon, R. Storeton-West, K. J. Hargreaves, C.
Flechard, and D. Fowler (2004), Linking land-atmosphere-stream carbon fluxes in a
lowland peatland system, Global Biogeochem. Cycles, 18(1).
Blough, N. V., and R. Del Vecchio (2002), Chromophoric DOM in the coastal
environment, in Biogeochemistry of Marine Dissolved Organic Matter, edited by D. A. Hansell,
and C. A. Carlson, pp. 509-546, Academic Press, San Diego.
Bricaud, A., M. Babin, A. Morel, and H. Claustre (1995), Variability in the chlorophyllspecific
absorption coefficients of natural phytoplankton: Analysis and
parameterization, J. Geophys. Res., 100(C7), 13321-13332.
Bricaud, A., A. Morel, and L. Prieur (1981), Absorption by dissolved organic matter of the
sea (yellow substance) in the UV and visible domains, Limnol. Oceanogr., 26(1), 43-53.
Budac, D., and P. Wan (1992), Photodecarboxylation: mecanism and synthetic utility, J. of
Photochem. Photobiology, 67, 135-166.
Bussmann, I., and G. Kattner (2000), Distribution of dissolved organic carbon in the
central Arctic Ocean: the influence of physical and biological properties, J. Mar. Syst.,
27, 209-219.
Camill, P. (2005), Permafrost thaw accelerates in boreal peatlands during late-20th century
climate warming, Climatic Change, 68(1-2), 135-152.
Carder, Hawes, Baker, Smith, Steward, and Mitchell (1991), Reflectance model for
quantifying chlorophyll a in the presence of productivity degradation products, J.
Geophys. Res., 96(C11), 20599.
Carder, K. L., R. G. Steward, G. R. Harvey, and P. B. Ortner (1989), Marine humic and
fulvic acids: Their effects on remote sensing of ocean chlorophyll, Limnol. Oceanogr., 34,
86-81.
Carlson, C. A. (2002), Production and Removal Processes, in Biogeochemistry of Marine
Dissolved Organic Matter, edited by D. A. Hansell, and C. A. Carlson, pp. 91-151,
Academic Press, San Diego.
Carmack, E. C., and R. W. MacDonald (2002), Oceanography of the Canadian Shelf of the
Beaufort Sea: A setting for marine life, Arctic, 55(1), 29-45.
Carmack, E. C., R. W. MacDonald, and S. Jasper (2004), Phytoplankton productivity on
the Canadian Shelf of the Beaufort Sea, Mar. Ecol. Prog. Ser., 277, 37-50.
Carmack, E. C., R. W. MacDonald, and J. E. Papadakis (1989), Water mass structure and
boundaries in the Mackenzie Shelf estuary, J. Geophys. Res., 94(C12), 18043-18055.
Cauwet, G., and I. Sidorov (1996), The biogeochemistry of Lena River: organic carbon and
nutrients distribution, Mar. Chem., 53(3-4), 211-227.
214

Cavalieri, D. J., P. Gloersen, and J. Zwally (1990), DMSP SSMI/I daily polar gridded sea
ice concentrations, edited by J. A. Maslanik, and J. Stroeve, pp. Digital media, National
Snow and Ice Data Center, Boulder, CO.
Cavalieri, D. J., C. L. Parkinson, and K. Y. Vinnikov (2003), 30-Year satellite record reveals
contrasting Arctic and Antarctic decadal sea ice variability, Geophys. Res. Lett., 30(18).
Chami, M., E. B. Shybanov, T. Y. Churilova, G. A. Khomenko, M. E. G. Lee, O. V.
Martynov, G. A. Berseneva, and G. K. Korotaev (2005), Optical properties of the
particles in the Crimea coastal waters (Black Sea), J. Geophys. Res.,
doi:10.1029/2005JC003008, 110(C11), C11020.
Clarke, G. L., G. C. Ewing, and C. J. Lorenzen (1970), Spectra of backscattered light from
the sea obtained from aircraft as a measure of chlorophyll concentration, Science, 167,
1119-1121.
Comiso, J. C., N. G. Maynard, W. O. Smith Jr., and C. W. Sullivan (1990), Satellite ocean
color studies of Antarctic ice edges in summer and autumn, J. Geophys. Res., 95(C6),
9481-9486.
Comiso, J. C., J. Yang, H. Susumo, and R. A. Krishfield (2003), Detection change in the
Arctic using satellite and in situ data, J. Geophys. Res., 108(C12), 3384.
Cota, G. F., J. Wang, and J. C. Comiso (2004), Transformation of global satellite
chlorophyll retreivals with a regionally tuned algorithm, Remote Sens. Environ., 90, 373-
377.
Cox, C., and W. H. Munk (1954), Statistics of the sea surface derived from sun glitter, J. of
Marine Res., 13, 198-227.
Cullen, J. J., R. F. Davis, J. S. Barlett, and W. L. Miller (1997), Toward remote sensing of
UV attenuation, photochemical fluxes, and biological effects of UV in surface waters,
in Current and Emerging Issues in Aquatic Science: Aquatic Science meeting, Am. Soc. Limnol.
and Oceanogr., Santa Fe, N. M.
Dagg, M., R. Benner, S. Lohrenz, and D. Lawrence (2004), Transformation of dissolved
and particulate materials on continental shelves influenced by large rivers: plume
processes, Cont. Shelf Res., 24, 833-858.
Davidson, E. A., and I. A. Janssens (2006), Temperature sensitivity of soil carbon
decomposition and feedbacks to climate change, Nature, 440, 165-173.
Davis, J., and R. Benner (2005), Seasonal trends in the aboundance, composition and
bioavailability of particulate and dissolved organic matter in the Chukchi/Beaufort
Seas and western Canada Basin, Deep Sea Res. II, 52, 3396-3410.
Del Vecchio, R., and N. V. Blough (2002), Photobleaching of chromophoric dissolved
organic matter in natural waters: kinetics and modelling, Mar. Chem., 78, 231-253.
Dittmar, T. (2004), Evidence for terrigenous dissolved organic nitrogen in the Arctic deep
sea, Limnol. Oceanogr., 49(1), 148-156.
Dittmar, T., and G. Kattner (2003), The biogeochemistry of the river and shelf ecosystem
of the Arctic Ocean: a review, Mar. Chem., 83, 105-120.
215

Dixon, R. K., S. Brown, R. A. Houghton, A. M. Solomon, M. C. Trexler, and J. Wisniewski
(1994), Carbon pools and flux of global forest ecosystems., Science, 263(5144), 185-190.
Droppo, I. G., D. Jeffries, C. Jaskot, and S. Backus (1998), The prevalence of freshwater
flocculation in cold regions: A case study from the Mackenzie River Delta, Northwest
Territories, Canada, Arctic, 51(2), 155-164.
D'Sa, E., and R. L. Miller (2003), Bio-optical properties in waters influenced by the
Mississippi River during low flow conditions, Remote Sens. Environ., 84, 538-549.
Dugdale, R. C., and J. J. Goering (1967), Uptake of new and regenerated forms of nitrogen
in primary production, Limnol. Oceanogr., 12, 196-206.
Ebuchi, N., and S. Kizu (2002), Probability distribution of surface wave slope derived
using sun glitter images from geostastionay meteorological satellite and surface vector
winds from scatterometers, J. Oceanogr., 58, 477-486.
Ferrari, G. M. (2000), The relationship between chromophoric dissolved organic matter
and dissolved organic carbon in the European Atlantic coastal area and in the West
Mediterranean Sea (Gulf of Lions), Mar. Chem., 70, 339-357.
Fichot, C. G. (2004), Marine photochemistry from space: Algorithms for the retreival of
diffuse attenuation and CDOM absorption coefficients (320-490 nm) from Ocean
Color and estimation of depth-resolved photoproduction rates of carbon monoxide
(CO) at global scales using SeaWiFS imagery, Master of Science thesis,217 pp.,
Dalhousie, Halifax, Canada.
Fioletov, V. E., M. G. Kimlin, N. Krotkov, L. J. B. McArthur, J. B. Kerr, D. I. Wardle, J. R.
Herman, R. Meltzer, T. W. Mathews, and J. Kaurola (2004), UV index climatology
over the United States and Canada from ground-based and satellite estimates,
doi:10.1029/2004JD004820, J. Geophys. Res., 109(22).
Franz, B. A., R. E. Eplee, S. W. Bailey, and M. Wang (2003), Changes to the atmospheric
correction algorithm and retrieval of oceanic optical properties, Chap. 5, in NASA
Tech. Memo. 2003-206892, vol. 22, edited by S. B. Hooker, and E. R. Firestone, pp. 29-
33, NASA Goddard Space Flight Center, Greenbelt, Md.
Freeman, C., C. D. Evans, D. T. Monteith, B. Reynolds, and N. Fenner (2001), Export of
organic carbon from peat soils, Nature, 412(6849), 785-785.
Freeman, C., N. Fenner, N. J. Ostle, H. Kang, D. J. Dowrick, B. Reynolds, M. A. Lock, D.
Sleep, S. Hughes, and J. Hudson (2004), Export of dissolved organic carbon from
peatlands under elevated carbon dioxide levels, Nature, 430(6996), 195-198.
Frey, K. E., and L. C. Smith (2005), Amplified carbon release from vast West Siberian
peatlands by 2100, Geophys. Res. Lett., 32(9).
Frouin, R., M. Schwindling, and P. Y. Deschamps (1996), Spectral reflectance of sea foam
in the visible and near-infrared: In situ measurements and remote sensing implications,
J. Geophys. Res., 101(C6), 14361-14371.
Fry, E. S. (2000), Visible and near-ultraviolet absorption spectrum of liquid water:
comment, Appl. Optics., 39(16), 2743-2744.
216

Gallegos, C. L., and P. J. Neale (2002), Partitioning spectral absorption in case 2 waters:
discrimination of dissolved and particulates components, App. Opt., 41(21), 4220-4233.
Gao, H., and R. G. Zepp (1998), Factors influencing photoreactions of dissolved organic
matter in a coastal river of southeastern United States, Environ. Sci. Techn., 32(19),
2940-2946.
Garneau, M.-È., W. F. Vincent, L. Alonso-Saez, Y. Gratton, and C. Lovejoy (2006),
Prokaryotic community structure and hetereotrophic production in a river-influence
arctic ecosystem, Aquat. Microb. Ecol., 42, 27-40.
Garver, D. L., and D. A. Siegel (1997), Inherent optical property inversion of ocean color
spectra and its Biogeochem. interpretation, 1. Time series from the Sargasso Sea, J.
Geophys. Res., 102(C8), 18607-18625.
Gibson, J. A. E., W. F. Vincent, B. Nieke, and R. Pienitz (2000), Control of biological
exposure to UV radiation in the Arctic Ocean: Comparison of the roles of ozone and
riverine dissolved organic matter, Arctic, 53(5), 372-382.
Gobeil, C., B. Sundby, R. W. Macdonald, and J. N. Smith (2001), Recent change in organic
carbon flux to Arctic Ocean deep basins: Evidence from acid volatile silfide,
manganese and rhenium discord in sediments, Geophys. Res. Lett., 28(9), 1743-1746.
Goñi, M. A., M. B. Yunker, R. W. Macdonald, and T. I. Eglinton (2005), The supply and
preservation of ancient and modem components of organic carbon in the Canadian
Beaufort Shelf of the Arctic Ocean, Mar. Chem., 93(1), 53-73.
Gordon, H. R. (1992), Diffuse reflectance of the ocean: influence of nonuniform
phytoplankton pigment profile, App. Opt., 31(12), 2116-2129.
Gordon, H. R. (1997), Atmospheric correction of ocean color imagery in the Earth
Observation System era, J. Geophys. Res., 102(D14), 17081-17106.
Gordon, H. R., J. W. Brown, and R. H. Evans (1988b), Exact Rayleigh scattering
calculations for use with the Nimbus-7 Coastal Zone Color Scanner, Appl. Optics., 27,
862-871.
Gordon, H. R., O. B. Brown, R. H. Evans, J. W. Brown, R. C. Smith, K. S. Baker, and D.
K. Clark (1988a), A semianalytic radiance model of ocean color, J. Geophys. Res., 93(D9),
10909-10924.
Gordon, H. R., O. B. Brown, and M. M. Jacobs (1975), Computed relationships between
the inherent and apparent optical properties of a flat homogenous ocean, Appl. Optics.,
14(2), 417-427.
Gordon, H. R., and D. K. Clark (1981), Clear water radiances for atmospheric correction
of coastal zone color scanner imagery, App. Opt., 20(24), 4175-4180.
Gordon, H. R., D. K. Clark, J. L. Mueller, and W. A. Hovis (1980), Phytoplankton
pigments from the Nimbus-7 Coastal Zone Color Scanner: comparisons with surface
measurements, Science, 210, 63-66.
Gordon, H. R., and K. Ding (1992), Self-shading of in-water optical instruments, Limnol.
Oceanogr., 37(3), 491-500.
217

Gordon, H. R., and M. Wang (1992a), Surface-roughness considerations for atmospheric
correction of ocean color sensors. I : the Rayleigh- scattering component, Appl. Optics.,
31, 4247-4260.
Gordon, H. R., and M. Wang (1992b), Surface-roughness considerations for atmospheric
correction of ocean color sensors. II : error in the retrieved water-leaving radiance,
Appl. Optics., 31, 4261-4267.
Gordon, H. R., and M. Wang (1994), Retrieval of water-leaving radiance and aerosol
optical thickness over the oceans with SeaWiFS: a preliminary algorithm, App. Opt.,
33(3), 443-452.
Gorham, E. (1991), Northern peatlands: role in the carbon cycle and probable responses
to climatic warming, Ecological Applications, 1(2), 182-195.
Gosselin, M., M. Levasseur, P. A. Wheeler, R. A. Horner, and B. C. Booth (1997), New
measurements of phytoplankton and ice algal production in the Arctic Ocean, Deep-Sea
Res. II, 44(8), 1623-1644.
Gregg, W. W., and N. W. Casey (2004), Global and regional evaluation of the SeaWiFS
chlorophyll data set, Remote Sens. Environ., 93, 463-479.
Grenfell, T. C., and G. A. Maykut (1977), The optical properties of ice and snow in the
Arctic Basin, J. of Glaciol., 18, 445-463.
Guay, C. K., G. P. Klinkhammer, K. K. Falkner, R. Benner, P. G. Coble, T. E. Whitledge,
B. Black, F. J. Bussell, and T. A. Wagner (1999), High-resolution measurements of
dissolved organic carbon in the Arctic Ocean by in situ fiber-optic spectrometry,
Geophys. Res. Lett., 26(8), 1007-1010.
Guéguen, C., L. Guo, and N. Tanaka (2005), Distributions and characteristics of colored
dissolved organic matter in the Western Arctic Ocean, Cont. Shelf Res., 25(10), 1195-
1207.
Guo, L., and R. W. Macdonald (2006), Source and transport of terrigenous organic matter
in the upper Yukon River: Evidence from isotope (δ 13 C, Δ 14 C, and δ �5 N) composition
of dissolved colloïdal, and particulate phases, Global Biogeochem. Cycles, 20, GB2011,
doi:10.1029/2005GB002593.
Hansell, D. A., and C. A. Carlson (1998), Deep-ocean gradients in the concentration of
dissolved organic carbon, Nature, 395(6699), 263-266.
Hansell, D. A., D. Kadko, and N. R. Bates (2004), Degradation of terrigenous dissolved
organic carbon in the western Arctic Ocean, Science, 304, 858-861.
Hedges, J. I., R. G. Keil, and R. Benner (1997), What happens to terrestrial organic matter
in the ocean?, Org. Geochem., 27(5/6), 196-212.
Herman, J. R., and E. A. Celarier (1997), Earth surface reflectivity climatology at 340-380
nm from TOMS data, J. Geophys. Res., 102(D23), 28003-28011.
Herman, J. R., N. Krotkov, E. Celarier, D. Larko, and G. Labow (1999), Distribution of
UV radiation at the Earth's surface from TOMS-measured UV-backscattered radiances,
J. Geophys. Res., 104(D10), 12059-12076.
218

Hernes, P. J., and R. Benner (2002), Transport and diagenesis of dissolved and particulate
terrigenous organic matter in the North Pacific Ocean, Deep-Sea Res., 49(12), 2119-2132.
Hernes, P. J., and R. Benner (2003), Photochemical and microbial degradation of dissolved
lignin phenols: Implications for the fate of terrigenous dissolved organic matter in
marine environments, J. Geophys. Res., 108(C9), 3291, doi: 1029/2002JC001421, 2003.
Hernes, P. J., and R. Benner (2006), Terrigenous organic matter sources and reactivity in
the North Atlantic Ocean and a comparison to the Arctic and Pacific oceans, Mar.
Chem., 100(1-2), 66-79.
Hinzman, L., N. D. Bettez, W. R. Bolton, F. S. Chapin, III, M. Dyurgerov, C. L. Fastie, B.
Griffith, R. D. Hollister, A. Hope, H. P. Huntington, A. M. Jensen, G. J. Jia, T.
Jorgenson, D. L. Kane, D. R. Klein, G. Kofinas, A. H. Lynch, A. H. Lloyd, A. D.
McGuire, F. E. Nelson, W. C. Oechel, T. Osterkamp, C. H. Racine, V. E.
Romankevich, R. S. Stone, D. A. Stow, M. Sturm, G. E. Tweedie, G. L. Vourlitis, M. D.
Walker, D. A. Walker, P. J. Webber, J. M. Welker, K. S. Winker, and K. Yoshikawa
(2005), Evidence and implications of recent climate change in northern Alaska and
other Arctic regions, Climatic Change, 72, 251-298.
Hoge, F. E., and P. E. Lyon (1996), Satellite retreival of inherent optical properties by
linear matrix inversion of oceananic radiance models: An analysis of model and
radiance measurement errors, J. Geophys. Res., 101, 16631-16648.
Hoge, F. E., M. E. Williams, R. N. Swift, J. K. Yungel, and A. Vodacek (1995), Satellite
retrieval of the absorption coefficient of chromophoric dissolved organic matter in
continental margins, J. Geophys. Res., 100(C12), 24847-24854.
Hooker, S. B., and C. R. McClain (2000), The calibration and validation of SeaWiFS data,
Prog. Oceanogr., 45, 427-465.
Hovis, W. A., D. K. Clark, F. Anderson, R. W. Austin, W. H. Wilson, E. T. Baker, D. Ball,
H. R. Gordon, J. L. Mueller, S. Z. El-Sayed, B. Sturm, R. C. Wrigley, and C. S. Yentsch
(1980), Nimbus-7 Coastal Zone Color Scanner: System description and initial imagery,
Science, 210, 60-63.
Hu, C., F. E. Muller-Karger, and R. G. Zepp (2002), Absorbance, absorption coefficient,
and apparent quantum yield: A comment on common ambiguity in the use of these
optical concepts, Limnol. Oceanogr., 47(4), 1261-1267.
Huot, Y., C. A. Brown, and J. J. Cullen (2005), New algorithms for MODIS sun-induced
chlorophyll fluorescence and a comparison with present data products, Limnol. Oceanogr.
Meth., 3, 108-130.
Ingvaldsen, R. B., L. Asplin, and H. Loeng (2004), The seasonal cycle in the Atlantic
transport to the Barents Sea during the years 1997-2001, Cont. Shelf Res., 24(9), 1015-
1032.
IOCCG (2006), Remote sensing of Inherent Optical Properties: Fundamentals, Tests of algorithms, and
Applications, edited by Z.-P. Lee, pp. 126, Reports of the International Ocean-Colour
Coordinating Group, No. 5, IOCCG, Dartmouth, Canada.
219

IPCC (2001), Climate Change 2001: The scientific basis. Contribution of Working Group I to the
Third Assessment Report of the Intergovernmental Panel on Climate Change, 881 pp., Cambridge
Press University.
Jakobsson, M., A. Grantz, Y. Kristoffersen, and R. Macnab (2004), Bathymetry and
physiography of the Arctic Ocean ans its constituent seas, in The organic carbon cycle in the
Arctic Ocean, edited by R. Stein, and R. W. Macdonald, pp. 1-6, Springer, Berlin.
Johannessen, S. C. (2000), A photochemical sink for dissolved organic carbon in the ocean,
176 pp., Dalhousie University, Halifax.
Johannessen, S. C., and W. L. Miller (2001), Quantum yield for the photochemical
production of dissolved inorganic carbon in seawater, Mar. Chem., 76, 271-283.
Johannessen, S. C., W. L. Miller, and J. J. Cullen (2003), Calculation of UV attenuation and
colored dissolved organic matter absorption spectra from measurements of ocean
color, J. Geophys. Res., 108(C9), 3301, doi:10.1029/2000JC000514.
Kattner, G., J. M. Lobbes, H. P. Fitznar, R. Engbrodt, E.-M. Nöthig, and R. J. Lara (1999),
Tracing dissolved organic substances and nutrients from the Lena River through
Laptev Sea (Arctic), Mar. Chem., 65, 25-39.
Keeling, C. D., Atmospheric CO 2 concentration - Mauna Loa Observatory, Hawaii 1956-
1986, NDP-001/R1 Carbon Dioxide Information Center, Oak Ridge National
Laboratory, Oak Ridge Tennessee, 1986.
Kharu, M., and G. B. Mitchell (2001), Seasonal and nonseasonal variability of satellitederived
chlorophyll and colored dissolved organic matter concentration in the
California Current, J. Geophys. Res., 106(C2), 2517-2529.
Kieber, R. J., X. Zhou, and K. Mopper (1990), Formation of carbonyl compounds from
UV-induced photodegradation of humic substances in natural waters: Fate of riverine
carbon in the sea, Limnol. Oceanogr., 35(7), 1503-1515.
Kieber, D. J., J. McDaniel, and K. Mopper (1989), Photochemical source of biological
substrates in sea water: implications for carbon cycling, Nature, 341, 637-639.
Kinney, P. J., T. C. Loder, and J. Groves (1971), Particulate and dissolved organic matter
in the Amerasian Basin of the Arctic Ocean, Limnol. Oceanogr., 16(1), 132-137.
Kishino, M., M. Takahashi, N. Okami, and S. Ichimura (1985), Estimation of the spectral
absorption coefficients of phytoplankton in the sea, Bull. Mar. Sci., 37(2), 634-642.
Klein, B., B. LeBlanc, Z. P. Mei, R. Beret, J. Michaud, C. J. Mundy, C. H. von Quillfeldt, M.
E. Garneau, S. Roy, Y. Gratton, J. K. Cochran, S. Bélanger, P. Larouche, J. D. Pakulski,
R. B. Rivkin, and L. Legendre (2002), Phytoplankton biomass, production and
potential export in the North Water, Deep-Sea Res., 49(22-23), 4983-5002.
Köhler, H., B. Meon, V. V. Gordeev, A. Spitzy, and R. M. W. Amon (2003), Dissolved
organic matter (DOM) in the estuaries of Ob and Yenisei and the adjacent Kara Sea,
Russia, in Siberian river runoff in the Kara Sea: Characterization, Quantification, Variability and
Environmental Significance, Proceedings in Marine Science, edited by R. Stein, K. Fahl, D. K.
Fütterer, E. M. Galimov, and O. V. Stepanets, pp. 281-308, Elsevier, Amsterdam.
220

Kowalczuk, P., J. Olszewski, M. Darecki, and S. Kaczmarek (2005), Empirical relationships
between coloured dissolved organic matter (CDOM) absorption and apparent optical
properties in Baltic Sea waters, Int. J. Remote Sens., 26(2), 345-370.
Krotkov, N., J. Herman, P. K. Bhartia, C. Seftor, A. Arola, S. Kalliskota, P. Taalas, and I.
V. Geogdzhaev (2002), Version 2 total ozone mapping spectrometer ultraviolet
algorithm: problems and enhancements, Opt. Engin., 41(12), 3028-3039.
Krotkov, N. A., P. K. Bhartia, J. R. Herman, V. Fioletov, and J. Kerr (1998), Satellite
estimation of spectral surface UV irradiance in the presence of tropospheric aerosols -
1. Cloud-free case, J. Geophys. Res., 103(D8), 8779-8793.
Krotkov, N. A., J. R. Herman, P. K. Bhartia, V. Fioletov, and Z. Ahmad (2001), Satellite
estimation of spectral surface UV irradiance 2. Effects of homogeneous clouds and
snow, J. Geophys. Res., 106(D11), 11743-11759.
Laurion, I., F. Blouin, and S. Roy (2003), The quantitative filter technique for measuring
phytoplankton absorption: Interference by MAAs in the UV waveband, Limnol.
Oceanogr.: Methods, 1(1), 1-9.
Lee, S. H., and T. R. Whitledge (2005), Primary and new production in the deep Canada
Basin during summer 2002, Polar Biology, 28(3), 190-197.
Lee, Z.-P., K. L. Carder, and R. A. Arnone (2002), Deriving inherent optical properties
from water color : a multiband quasi-analytical algorithm for optically deep waters, App.
Opt., 41(27), 5755-5772.
Lee, Z. P., K. L. Carder, R. F. Chen, and T. G. Peacock (2001), Properties of water column
and bottom derived from Airborne Visible Infrared Imaging Spectrometer (AVIRIS)
data, J. Geophys. Res., 106(C6), 11639-11651.
Lee, Z. P., K. L. Carder, C. D. Mobley, R. G. Steward, and J. S. Patch (1999),
Hyperspectral remote sensing for shallow waters: 2. Deriving bottom depths and water
properties by optimization, App. Opt., 38(18), 3831-3843.
Lee, Z. P., K. L. Carder, T. G. Peacock, C. O. Davis, and J. L. Mueller (1996), Method to
derive ocean absorption coefficients from remote-sensing reflectance, Appl. Optics.,
35(3), 453-462.
Lee, Z.-P., K. L. Carder, and K. Du (2004), Effects of molecular and particle scatterings on
the model parameter for remote-sensing reflectance, App. Opt., 43(25), 4957-4964.
Loisel, H., and A. Morel (1998), Light scattering and chlorophyll concentration in case 1
waters: A reexamination, Limnol. Oceanogr., 43(5), 847-858.
Loisel, H., and A. Morel (2001), Non-isotropy of the upward radiance field in typical
coastal (Case 2) waters, Int. J. Remote Sens., 22(2&3), 275-295.
Macdonald, R. W. (2000), Arctic estuaries and ice: a positive-negative estuarine couple, in
The Freshwater Budget of the Arctic Ocean, edited by E. L. Lewis, pp. 383-407, Kluwer
Academic Publishers, Dordrecht, Netherlands.
Macdonald, R. W., E. C. Carmack, F. A. McLaughlin, K. K. Falkner, and J. H. Swift (1999),
Connections among ice, runoff and atmospheric forcing in the Beaufort Gyre, Geophys.
Res. Lett., 26(15), 2223-2226.
221

Macdonald, R. W., E. C. Carmack, F. A. McLaughlin, K. Iseki, D. M. MacDonald, and M.
C. O'Brien (1989), Composition and modification of water mmasses in the Mackenzie
Shelf Estuary, J. Geophys. Res., 94(C12), 18057-18070.
Macdonald, R. W., F. A. McLaughlin, and E. C. Carmack (2002), Fresh water and its
sources during the SHEBA drift in the Canada Basin of the Arctic Ocean, Deep-Sea Res.
I, 49, 1769-1785.
Macdonald, R. W., A. S. Naidu, M. B. Yunker, and C. Gobeil (2004b), The Beaufort Sea:
distribution, sources, fluxes and burial of organic carbon, in The organic carbon cycle in the
Arctic Ocean, edited by R. Stein, and R. W. Macdonald, pp. 167-193, Springer, Berlin.
Macdonald, R. W., D. W. Paton, E. C. Carmack, and A. Omstedt (1995), The freshwater
budget and under-ice spreading of Mackenzie River Water in the Canadian Beaufort
Sea based on salinity and 18 O/ 16 O measurements in water and ice., J. Geophys. Res.,
100(C1), 895-919.
Macdonald, R. W., E. Sakshaug, and R. Stein (2004a), The Arctic Ocean: modern status
and recent climate change, in The organic carbon cycle in the Arctic Ocean, edited by R. Stein,
and R. W. MacDonald, pp. 6-21, Springer, Berlin.
Macdonald, R. W., S. M. Solomon, R. E. Cranston, H. E. Welsh, M. B. Yunker, and C.
Gobeil (1998), A sediment and organic carbon budget for the Canadian Beaufort Shelf,
Mar. Geol., 144, 255-273.
Macdonald, R. W., C. S. Wong, and P. E. Erickson (1987), The distribution of nutrients in
the Southeastern Beaufort Sea : Implications for the water circulation and primary
production, J. Geophys. Res., 92(C3), 2939-2952.
Mack, M. C., E. A. G. Schuur, M. S. Bret-Harte, G. R. Shaver, and F. S. Chapin, III (2004),
Ecosystem carbon storage in arctic tundra reduced by long-term nutrient fertilization,
Nature, 431(7007), 440-443.
Madronich, S., and S. Flocke (1999), The role of solar radiation in atmospheric chemistry,
in Handbook of Environmental Chemistry, edited by P. Boule, pp. 1-26, Springer-Verlag,
Heidelberg, Germany.
Maritorena, S., D. A. Siegel, and A. R. Peterson (2002), Optimization of a semianalitical
ocean color model for global-scale applications, App. Opt., 41(15), 2705-2714.
Mc Clatchey, R. A., R. W. Fenn, J. E. A. Selby, F. E. Volz, and J. S. Garing, Optical
properties of the atmosphere, AFCRL-TR-71-079, L.G. Hancom Fiel, Bedford, Ms.,
1971.
McClelland, J. W., S. J. Dery, B. J. Peterson, R. M. Holmes, and E. F. Wood (2006), A panarctic
evaluation of changes in river discharge during the latter half of the 20th century,
Geophys. Res. Lett., 33(6), L06715.
McKee, B. A., R. C. Aller, M. A. Allison, T. S. Bianchi, and G. C. Kineke (2004), Transport
and transformation of dissolved and particulate materials on continental margins
influenced by major rivers: benthic boundary layer and seabed processes, Cont. Shelf
Res., 24(7-8), 899-926.
222

Meon, B., and R. M. W. Amon (2004), Heterotrophic bacterial activity and fluxes of
dissolved free amino acids and glucose in the Arctic rivers Ob, Yenisei and the
adjacent Kara Sea, Aquat. Microb. Ecol., 37(2), 121-135.
Meyers-Schulte, K. J., and J. I. Hedges (1986), Molecular evidence for a terrestrial
component of organic matter dissolved in ocean water, Nature, 321, 61-63.
Miles, C. J., and P. L. Brezonik (1981), Oxygen consumption in humic-colored waters by
photochemical ferrous-ferric catalytic cycle, Environ. Sci. Technol., 15(9), 1089-1095.
Miller, W. L., and M. A. Moran (1997), Interaction of photochemical and microbial
processes in the degradation of refractory dissolved organic matter from coastal marine
environment, Limnol. Oceanogr., 42(6), 1317-1324.
Miller, W. L., and R. G. Zepp (1995), Photochemical production of dissolved inorganic
carbon from terrestrial organic matter: Significance to the oceanic carbon cycle, Geophys.
Res. Lett., 22(4), 417-420.
Millot, R., J. Gaillardet, B. Dupre, and C. J. Allegre (2003), Northern latitude chemical
weathering rates: Clues from the Mackenzie River Basin, Canada, Geochim. Cosmochim.
Acta, 67(7), 1305-1329.
Mitchell, B. G. (1992), Predictive bio-optical relationships for polar oceans and marginal
ice zones, J. Mar. Syst., 3, 91-105.
Mobley, C. D. (1994), Light and water: Radiative Transfer in Natural Waters, 1st ed., 592 pp.,
Academic Press, San Diego, California.
Mobley, C. D. (1999), Estimation of the remote-sensing réflectances from above-water
measurements, App. Opt., 38(36), 7442-7455.
Mopper, K., and D. J. Kieber (2002), Photochemistry and the cycling of carbon, sulfur,
nitrogen and phosphorus, in Biogeochemistry of Marine Dissolved Organic Matter, edited by
D. A. Hansell, and C. A. Carlson, pp. 455-507, Academic Press, San Diego.
Mopper, K., X. Zhou, R. J. Kieber, D. J. Kieber, R. J. Sikorski, and R. D. Jones (1991),
Photochemical degradation of dissolved organic carbon and its impact on the oceanic
carbon cycle, Nature, 353, 60-62.
Morel, A. (1974), Optical properties of pure water and pure sea water, in Optical aspects of
Oceanography, edited by N. G. Jerlov, and E. S. Nielsen, pp. 1-24, Academic, New York.
Morel, A. (1988), Optical modeling of the upper ocean in relation to its biogenous matter
content (Case 1 Waters), J. Geophys. Res., 93(C9), 10749-10768.
Morel, A., D. Antoine, and B. Gentili (2002), Bidirectional reflectance of oceanic waters:
accounting for Raman emission and varying particle scattering phase function, App.
Opt., 41(30), 6289-6306.
Morel, A., and S. Bélanger (2006), Improved detection of turbid waters from ocean color
sensors information, Remote Sens. Environ., 102, 237-249.
Morel, A., and B. Gentili (1991), Diffuse reflectance of oceanic waters: its dependance on
Sun angle as influenced the mollecular scattering contribution, App. Opt., 30(30), 4427-
4438.
223

Morel, A., and B. Gentili (1993), Diffuse reflectance of oceanic waters. II. Bidirectional
aspects, App. Opt., 32(33), 6864-6879.
Morel, A., and B. Gentili (1996), Diffuse reflectance of oceanic waters. III. Implication of
bidirectionality for the remote-sensing problem, App. Opt., 35(24), 4850-4862.
Morel, A., B. Gentili, H. Claustre, M. Babin, A. Bricaud, J. Ras, and F. Tièche (in press,
2006), Optical properties of the "clearest" natural waters, Limnol. Oceanogr.
Morel, A., and S. Maritorena (2001), Bio-optical properties of oceanic waters: A reappraisal,
J. Geophys. Res., 106(C4), 7163-7180.
Morel, A., and J. L. Mueller (2002), Normalized Water-Leaving Radiance and Remote
Sensing Reflectance: Bidirectional Reflectance and Other Factors, in Ocean Optics
Protocols For Satellite Ocean Color Sensor Validation, Revision 4, Volume III, edited by J. L.
Mueller, G. S. Fargion, and C. R. McClain, pp. 32-59, NASA Goddard Space Flight
Space Center, Greebelt.
Morel, A., and L. Prieur (1977), Analysis of variations in ocean color, Limnol. Oceanogr.,
22(4), 709-722.
Morrison, J. R., and N. B. Nelson (2004), Seasonal cycle of phytoplankton UV absorption
at Bermuda Atlantic Time-series Study (BATS) site, Limnol. Oceanogr., 49(1), 215-224.
Mueller, J. L., A. Morel, R. Frouin, C. Davis, R. A. Arnone, K. L. Carder, Z.-P. Lee, R. G.
Stewart, S. B. Hooker, B. Holben, C. D. Mobley, S. McLean, M. Miller, C. Pietras, G. S.
Fargion, K. D. Knobelspiesse, and K. Voss (2003), Radiometric measurements and
data analysis protocols, in Ocean Optics Protocols For Satellite Ocean Color Sensor Validation,
Revision 4, Volume III, edited by J. L. Mueller, G. S. Fargion, and C. R. McClain, pp. 1-
78, NASA Goddard Space Flight Center, Greenbelt.
O'Brien, M. C., R. W. MacDonald, H. Melling, and K. Iseki (2006), Particles fluxes and
geochemistry on the Canadian Beaufort Shelf: Implications for sediment transport and
deposition, Cont. Shelf Res., 26, 41-81.
Oechel, W. C., G. L. Vourlitis, S. J. Hastings, R. C. Zulueta, L. Hinzman, and D. Kane
(2000), Acclimation of ecosystem CO 2 exchange in the Alaskan Arctic in response to
decadal climate warming, Nature, 406(6799), 978-981.
Ohde, T., B. Sturm, and H. Siegel (2002), Derivation of SeaWiFS vicarious calibration
coefficients using in situ measurements in Case 2 water of the Baltic Sea, Remote Sens.
Environ., 80, 248-255.
Opsahl, S., and R. Benner (1997), Distribution and cycling of terrigenous dissolved organic
matter in the ocean, Nature, 386, 480-482.
Opsahl, S., and R. Benner (1998), Photochemical reactivity of dissolved lignin and ocean
waters, Limnol. Oceanogr., 43(6), 1297-1304.
Opsahl, S., R. Benner, and R. M. W. Amon (1999), Major flux of terrigenous dissolved
organic matter throught the Arctic Ocean, Limnol. Oceanogr., 44(8), 2017-2023.
O'Reilly, J. E., S. Maritorena, G. Mitchell, D. A. Siegel, K. L. Carder, D. L. Garver, M.
Kahru, and C. R. McClain (1998), Ocean color chlorophyll algorithms for SeaWiFS, J.
Geophys. Res., 103(C11), 24937-24950.
224

O'Reilly, J. E., S. Maritorena, D. Siegel, M. C. O’Brien, D. Toole, B. G. Mitchell, M. Kahru,
F. P. Chavez, P. Strutton, G. Cota, S. B. Hooker, C. R. McClain, K. L. Carder, F.
Muller-Karger, L. Harding, A. Magnuson, D. Phinney, G. F. Moore, J. Aiken, K. R.
Arrigo, R. Letelier, and M. Culver (2000), Ocean color chlorophyll a algorithms for
SeaWiFS, OC2, and OC4: version 4., in SeaWiFS Postlaunch Technical Report Series, vol.11.
SeaWiFS postlaunch calibration and validation analyses: part 3, edited by S. B. Hooker, and E.
R. Firestone, pp. 9-23, NASA Goddard Space Flight Center., Greenbelt, MD:.
Overland, J. E., M. C. Spillane, and N. N. Soreide (2004), Integrated analysis of physical
and biological Pan-Arctic change, Climatic Change, 63(3), 291-322.
Park, Y.-J., and K. G. Ruddick (2005), Model of remote-sensing reflectance including
bidirectional effects for case 1 and case 2 waters, App. Opt., 44(7), 1236-1249.
Pegau, W. S. (2002), Inherent optical properties of the central Arctic surface waters, J.
Geophys. Res., 107(C10), doi: 10.1029/2000JC000382.
Perovich, D. K., T. C. Grenfell, B. Light, and P. V. Hobbs (2002), Seasonal evolution of
the albedo of multiyear Arctic sea ice, J. Geophys. Res., 107(C10).
Perovich, D. K., C. S. Roesler, and W. S. Pegau (1998), Variability in Arctic sea ice optical
properties, J. Geophys. Res., 103(C1), 1193-1208.
Peterson, B. J., R. M. Holmes, J. W. McClelland, C. J. Vorosmarty, R. B. Lammers, A. I.
Shiklomanov, and S. Rahmstorf (2002), Increasing river discharge to the Arctic Ocean,
Science, 298, 2171-2173.
Peterson, B. J., McClelland, J. W., Curry, R., Holmes, R. M., Walsh, J. E., & Aagaard, K.
(2006). Trajectory shifts in the Arctic and Subarctic freshwater cycle. Science, 313, 1061-
1066.
Petzold, T. J., Volume scattering functions for selected ocean waters, Tech. Rep. SIO, 72-
78 pp., Scripps Institution of Oceanography, San Diego, La Jolla, Ca., 1972.
Pickart, R. S. (2004), Shelfbreak circulation in the Alaskan Beaufort Sea: Mean structure
and variability, J. Geophys. Res., 109(C4), C04024.
Pope, R. M., and E. S. Fry (1997), Absorption spectrum (380-700nm) of pure water. II.
Integrating cavity measurements, Appl. Optics., 36, 8710-8723.
Preisendorfer, R. W. (1961), Application of radiative transfert theory to light
measurements in the sea, Union Géod. Géophys. Ints., Monogr., 10, 11-30.
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery (1992), Numerical recipes
in Fortran 77: The art of scientific computing, Chap 9 pp., Cambridge U. Press, New York.
Proshutinsky, A., R. H. Bourke, and F. A. McLaughlin (2002), The role of the Beaufort
Gyre in Arctic climate variability: Seasonal to decadal climate scales, Geophys. Res. Lett.,
29(23), 2100, doi:10.1029/2002GL015847.
Quickenden, T. I., and J. A. Irvin (1980), The ultraviolet absorption spectrum of liquid
water, J. Chem. Phys., 72(8), 4416-4428.
Rachold, V., H. Eicken, V. V. Gordeev, M. N. Grigoriev, H.-W. Hubberten, A. P. Lisitzin,
V. P. Shevchenko, and L. Schirrmeister (2004), Modern terrigenous organic carbon
225

input to the Arctic Ocean, in The organic carbon cycle in the Arctic Ocean, edited by R. Stein,
and R. W. MacDonald, pp. 33-41, Springer, Berlin.
Rahmstorf, S. (2002), Ocean circulation and climate during the past 120,000 years, Nature,
419(6903), 207-214.
Retamal, L., W. F. Vincent, C. Martineau, and C. L. Osburn (in review, 2006), Comparison
of optical properties of dissolved organic matter in two river-influence coastal regions
of the Canadian Arctic, Estuarine Coast. Shelf Sci.
Rex, M., R. J. Salawitch, P. von der Gathen, N. R. P. Harris, M. P. Chipperfield, and B.
Naujokat (2004), Arctic ozone loss and climate change, Geophys. Res. Lett., 31(4).
Reynolds, R. A., D. Stramski, and B. G. Mitchell (2001), A chlorophyll-dependent
semianalytical reflectance model derived from field measurements of absorption and
backscattering coefficients within the Southern Ocean, J. Geophys. Res., 106(C4), 7125-
7138.
Rigor, I. G., J. M. Wallace, and R. L. Colony (2002), Response of sea ice to the Arctic
Oscillation, J. Climate, 15(18), 2648-2668.
Roesler, C. S., and M. J. Perry (1995), In situ phytoplankton absorption, fluorescence
emission, and particulate backscattering spectra determined from reflectance, J. Geophys.
Res., 100(C7), 13279-13294.
Roesler, C. S., M. J. Perry, and K. L. Carder (1989), Modeling in situ phytoplankton
absorption from total absorption spectra in productive inland marine waters, Limnol.
Oceanogr., 34, 1510-1523.
Rothrock, D. A., and J. Zhang (2005), Arctic Ocean sea ice volume: what explains its
recent depletion?, J. Geophys. Res., 110, C01002, doi:10.1029/2004JC002282.
Ruddick, K. G., V. De Cauwer, Y.-J. Park, and G. Moore (2006), Seaborne measurements
of near infrared water-leaving reflectance: the similarity spectrum for turbid waters,
Limnol. Oceanogr., 51(2), 1167-1179.
Ruddick, K. G., F. Ovidio, and M. Rijkeboer (2000), Atmospheric correction of SeaWiFS
imagery for turbid coastal and inland waters, App. Opt., 39(6), 897-895.
Rudels, B., L. G. Anderson, and E. P. Jones (1996), Formation and evolution of the
surface mixed layer and halocline of the Arctic Ocean, J. Geophys. Res., 101(C4), 8807-
8821.
Rudels, B., G. Bjork, J. Nilsson, P. Winsor, I. Lake, and C. Nohr (2005), The interaction
between waters from the Arctic Ocean and the Nordic Seas north of Fram Strait and
along the East Greenland Current: results from the Arctic Ocean-02 Oden expedition,
J. Mar. Syst., 55(1-2), 1-30.
Sakshaug, E. (2004), Primary and secondary production in the Arctic seas, in The organic
carbon cycle in the Arctic Ocean, edited by R. Stein, and R. W. MacDonald, pp. 57-81,
Springer, Berlin.
Sakshaug, E., and H. R. Skjoldal (1989), Life at the ice edge, Ambio., 18(1), 60-67.
Saliot, A., G. Cauwet, G. Cahet, D. Mazaudier, and R. Daumas (1996), Microbial activities
in the Lena River delta and Laptev Sea, Mar. Chem., 53, 247-254.
226

Sathyendranath, S., L. Prieur, and A. Morel (1989), A three-component model of ocean
colour and its application to remote sensing of phytoplankton pigments in coastal
waters, Int. J. of Remote Sensing, 10(8), 1373-1394.
Schlesinger, W. H. (1991), Biogeochemistry: an analysis of global change, 443 pp., Academic press,
San Diego.
Scully, N. M., and W. L. Miller (2000), Spatial and temporal dynamics of colored dissolved
organic matter in the North Water Polynya, Geophys. Res. Lett., 27(7), 1009-1011.
Serreze, M. C., J. A. Maslanik, T. A. Scambos, F. Fetterer, J. Stroeve, K. Knowles, C.
Fowler, S. Drobot, R. G. Barry, and T. M. Haran (2003), A record minimum arctic sea
ice extent and area in 2002, Geophys. Res. Lett., 30(3), 1110,
doi:10.1029/2002GL016406.
Serreze, M. C., J. E. Walsh, F. S. Chapin, T. Osterkamp, M. Dyurgerov, V. Romanovsky,
W. C. Oechel, J. Morison, T. Zhang, and R. G. Barry (2000), Observational evidence
of recent change in the northern high-latitude environment, Climatic Change, 46(1-2),
159-207.
Siegel, D. A., S. Maritorena, and N. B. Nelson (2005), Independence and
interdependencies among global ocean color properties: Reassessing the bio-optical
assumption, J. Geophys. Res., 110(C7).
Siegel, D. A., S. Maritorena, N. B. Nelson, D. A. Hansell, and M. Lorenzi-Kayser (2002),
Global distribution and dynamics of colored dissolved and detrital organic materials, J.
Geophys. Res., 107(C12), 3228, doi:10.1029/2001JC000965.
Siegel, D. A., M. Wang, S. Maritorena, and W. Robinson (2000), Atmospheric correction
of satellite ocean color imagery: the black pixel assumption, App. Opt., 39(21), 3582-
3591.
Smith, L. C., G. M. MacDonald, A. A. Velichko, D. W. Beilman, O. K. Borisova, K. E.
Frey, K. V. Kremenetski, and Y. Sheng (2004), Siberian peatlands a net carbon sink
and global methane source since the early Holocene, Science, 303(5656), 353-356.
Smith, R. C. (1973), Optical properties of the Arctic upper water, Arctic, 26(4), 303-313.
Smith, S. D., R. D. Muench, and C. H. Pease (1990), Polynyas and leads: An overview of
physical processes and environment, J. Geophys. Res., 95(C6), 9461-9479.
Smith, W. O., and D. M. Nelson (1986), The importance of ice-edge phytoplankton
production in the Southern Ocean, BioScience, 36, 251-257.
Sokal, R. R., and F. J. Rohlf (1995), Biometry: principles and practice of statistic in biological research,
Third ed., 887 pp., Freeman, New York.
Steele, M., R. Morley, and W. Ermold (2001), A global ocean hydrography with a high
quality Arctic Ocean, J. Climate, 14, 2079-2087.
Stein, R., and R. W. MacDonald (2004), Organic carbon budget: Arctic Ocean vs. Global
Ocean, in The Organic Carbon Cycle in the Arctic Ocean, edited by R. Stein, and R. W.
MacDonald, pp. 315-322, Springer, Berlin.
227

Stramska, M., D. Stramski, R. Hapter, S. Kaczmarek, and J. Ston (2003), Bio-optical
relationships and ocean color algorithms for the north polar region of the Atlantic, J.
Geophys. Res., 108(C5), 3143, doi: 10.1029/2001JC001195.
Stirling, I. (1997), The importance of polynyas, ice edges, and leads to marine mammals
and birds, J. Mar. Syst., 10(1-4), 9-21.
Stroeve, J. C., M. C. Serreze, F. Fetterer, T. Arbetter, W. Meier, J. Maslanik, and K.
Knowles (2005), Tracking the Arctic's shrinking ice cover: Another extreme September
minimum in 2004, Geophys. Res. Lett., 32(L04501), doi:10.1029/2004GL021810.
Stumpf, R. P., R. A. Arnone, R. W. Gould, R. M. Martinolich, and V. Ransibrahmanakul
(2003), A partially coupled ocean-atmosphere model for retrieval of water-leaving
radiance from SeaWiFS in coastal waters, Chap. 9, in NASA/TM-2003-206892, edited
by S. B. Hooker, and E. R. Firestone, pp. 51-59, Greenbelt, MD.
Tanré, D., M. Herman, and P. Y. Deschamps (1981), Influence of the background
contribution upon space measurements of ground reflectance, App. Opt., 20(20), 3676-
3684.
Tanré, D., M. Herman, P. Y. Deschamps, and A. de Leffe (1979), Atmospheric modeling
for space measurements of ground reflectances, including bidirectional properties, App.
Opt., 18(21), 3587-3594.
Tassan, S., and G. M. Ferrari (1995), An alternative approach to absorption measurements
of aquatic particles retained on filters, Limnol. Oceanogr., 40(8), 1358-1368.
Tassan, S., and G. M. Ferrari (2002), A sensitivity analysis of the 'transmittance-reflectance'
method for measuring light absorption by aquatic particles, J. Plankton Res., 24(8), 757-
776.
Telang, S. A., R. Pocklington, A. S. Naidu, E. A. Romankevich, I. I. Gitelson, and
Glacyshev (1991), Carbon and mineral transport in major North American, Russian
Arctic, and Siberian Rivers: The St Lawrence, the Mackenzie, the Yukon, the Arctic
Alaskan Rivers, the Arctic Basin Rivers in the Soviet Union, and the Yenisei, in
Biogeochemistry of Major World Rivers. SCOPE 42, edited by E. T. Degens, S. Kempe, and
J. E. Richey, pp. 74-104, Wiley, New York.
Topliss, B. J., J. R. Miller, and P. W. Horne (1989), Ocean optical measurements-II.
Statistical analysis of data from Canadian eastern Arctic waters, Cont. Shelf Res., 9(2),
135-142.
Twardowski, M. S., E. Boss, J. B. Macdonald, W. S. Pegau, A. H. Barnard, and J. R. V.
Zaneveld (2001), A model for estimating bulk refractive index from the optical
backscattering ratio and the implication for understanding particle composition in case
I and case II waters, J. Geophys. Res., 106(C7), 14129-14142.
Twardowski, M. S., E. Boss, J. M. Sullivan, and Donaghay, P. L. (2004). Modeling the
spectral shape of absorption by chromophoric dissolved organic matter. Marine
Chemistry, 89, 69-88.
Vähätalo, A. V., M. Salkinoja-Salonen, P. Taalas, and K. Salonen (2000), Spectrum of the
quantum yield for the photochemical mineralization of dissolved organic carbon in a
humic lake, Limnol. Oceanogr., 45(3), 664-676.
228

Vähätalo, A. V., K. Salonen, M. Salkinoja-Salonen, and A. Hatakka (1999), Photochemical
mineralization of synthetic lignin in lake water indicates enhanced of aromatic organic
matter under solar radiation, Biodegradation, 10(6), 415-420.
Vähätalo, A. V., and R. G. Wetzel (2004), Photochemical and microbial decomposition of
chromophoric dissolved organic matter during long (months-years) exposures, Mar.
Chem., 89, 313-326.
Vermote, E., D. Tanré, J. L. Deuzé, M. Herman, and J. J. Morcrette, 6s User Guide
Version 2, 218 pp., Laboratoire d'Optique Atmosphérique, Lille, France, 1997.
Vincent, W. F., and C. Belzile (2003), Biological UV exposure in the polar oceans: Arctic-
Antarctic comparisons, in Antarctic Biology in a Global Context, edited by A.H.L. Huiskes,
W.W.C. Gieskes, J. Rozema, R.M.L. Schorno, S. M. v. d. Vies, and W. J. Wolff, pp.
176-181., Backhuys Publishers, Leiden, Netherlands.
Vodacek, A., N. V. Blough, M. D. DeGrandpre, E. T. Peltzer, and R. K. Nelson (1997),
Seasonal variation of CDOM and DOC in the Middle Atlantic Bight: Terrestrial inputs
and photooxidation, Limnol. Oceanogr., 42(4), 674-686.
Wadhams, P., and N. R. Davis (2000), Further evidence of ice thinning in the Arctic
Ocean, Geophys. Res. Lett., 27, 3973-3975.
Wang, J., G. Cota, and D. A. Ruble (2005a), Absorption and backscattering in the Beaufort
and Chukchi Seas, J. Geophys. Res., 110(C4), C04014, doi: 10.1029/2002JC001653.
Wang, J., and G. F. Cota (2003), Remote-sensing reflectance in the Beaufort and Chukchi
seas: observations and models, App. Opt., 42(15), 2754-2765.
Wang, J., G. F. Cota, and J. C. Comiso (2005b), Phytoplankton in the Beaufort and
Chukchi Seas: Distribution dynamics, and environmental forcing, Deep Sea Res. II, 52,
3355-3368.
Werdell, P. J., and S. W. Bailey (2005), An improved in-situ bio-optical data set for ocean
color algorithm development and satellite product validation, Remote Sens. Environ., 98,
122-140.
Wheeler, P. A., M. Gosselin, E. Sherr, D. Thibault, D. L. Kirchman, R. Benner, and T. E.
Whitledge (1996), Active cycling of organic carbon in the central Arctic Ocean, Nature,
380(6576), 697-699.
Wheeler, P. A., J. M. Watkins, and R. L. Hansing (1997), Nutrients, organic carbon and
organic nitrogen in the upper water column of the Arctic Ocean: implications for
sources of dissolved organic carbon, Deep Sea Res. II, 44(8), 1571-1592.
Woodgate, R. A., K. Aagaard, J. H. Swift, K. K. Falkner, and W. M. Smethie (2005), Pacific
ventilation of the Arctic Ocean's lower halocline by upwelling and diapycnal mixing
over the continental margin, Geophys. Res. Lett., 32(18).
Woodgate, R. A., K. Aagaard, and T. J. Weingartner (2005), Monthly temperature, salinity,
and transport variability of the Bering Strait through flow, Geophys. Res. Lett., 32(4).
Woodwell, G. M., F. T. Mackenzie, R. A. Houghton, M. Apps, E. Gorham, and E.
Davidson (1998), Biotic feedbacks in the warming of the earth, Climatic Change, 40,
495-518.
229

Xie, H., O. C. Zafiriou, W.-J. Cai, R. G. Zepp, and Y. Wang (2004), Photooxidation and its
effects on the carboxyl content of dissolved organic matter in two coastal rivers in the
Southeastern United States, Environ. Sci. Techn., 38(15), 4113-4119.
Xie, H. X., and M. Gosselin (2005), Photoproduction of carbon monoxide in first-year sea
ice in Franklin Bay, southeastern Beaufort Sea, Geophys. Res. Lett., 32(12).
Zafiriou, O. C. (1977), Marine organic photochemistry previewed, Mar. Chem., 5, 497-522.
Zaneveld, J. R. V., J. C. Kitchen, and C. Moore (1994), The scattering error correction of
reflecting-tube absorption meters, in Ocean Optics XII, edited by J. S. Jaffe, pp. 44-55,
SPIE, Bergen.
Zepp, R. G., T. V. Callaghan, and D. Erickson III, J. (2003), Interactive effects of ozone
depletion and climate change on biogeochemical cycles, Photochem. Photobiol. Sci., 2, 51-
61.
Zhang, Y., and H. X. Xie (in press 2006), Factors affecting the efficiency of carbon
monoxide photoproduction in the St Lawrence estuarine system (Canada), Environ. Sci.
Tech.
Zibordi, G., and G. M. Ferrari (1995), Instrument Self-Shading in Underwater Optical
Measurements - Experimental-Data, App. Opt., 34(15), 2750-2754.
Zibordi, G., F. Mélin, and J.-F. Berthon (2006), Comparison of SeaWiFS, MODIS and
MERIS radiometric products at a coastal site, Geophys. Res. Lett., 33, L06617,
doi:10.1029/2006GL025778.
230

Annexe 1 : Modification de l’algorithme Quasi-
Analytique pour l’application à la Mer de Beaufort
231

Annex A1. Modification of the Quasi-Analytical Algorithm for an application to
the Beaufort Sea
A1.1. Introduction and motivation
In Chapter IV, we presented a method to derived the contribution of CDOM to the
total absorption coefficient ([aCDOM/at]) at 412 nm. To extrapolate [aCDOM/at] from 412 to
the full spectrum between 300 and 600 nm, the magnitude of at(412) is necessary (see
section VI.3.3). It has been shown that the total absorption coefficient could be derived
with a relatively good precision from remote sensing reflectance [IOCCG, 2006].
In addition, little is known on the spatial variability of the different surface water
masses in the southeastern Beaufort Sea. In particular, the distribution of river runoff,
which constitutes a major part of the freshwater budget in the western part of the Arctic
[Macdonald et al., 2002], remains poorly documented. Spatial-temporal variability of
satellite-derived inherent optical properties (IOP) may provide new insights on river
runoff distribution in the southeastern part of the Beaufort Sea.
For those reasons, here I present the modifications to the method proposed by Lee
et al. [2002] to derive IOPs in the southern Beaufort Sea from SeaWiFS imagery. The
objectives were 1) to include recent developments achieved the field of Ocean Color, and
2) to tune the empirical relationships involved in the algorithm with the CASES data set.
The results of the application of the algorithm to the CASES data set are presented and
discussed. The application to SeaWiFS imagery is shown and discussed in Chapter VI
(Figs. VI.6 and VI.7).
A1.2. Quasi-Analytical Algorithm description
Lee et al. [2002] recently proposed a quasi-analytical algorithm (hereafter denoted
as QAA02) that calculates the bulk IOPs, i.e. the total absorption (at) and backscattering
(bb) coefficients, from the remote-sensing reflectance spectrum (Rrs). The later is defined
as the ratio of water-leaving radiance to the above-surface downwelling irradiance. Our
algorithm differs from the QAA02 with respect to 1) the formulation of Rrs versus the
IOP, 2) the empirical formulations for the absorption coefficient at reference wavelength
232

(λ0), and 3) the use of a single value for the exponent of the power law used to model the
particulate backscattering spectrum (Y).
First, a semi-analytical remote-sensing reflectance model designed to account for
the bidirectional effects [Park and Ruddick, 2005] is used instead of the Gordon et al.
[1988] formulation. Since the Park and Ruddick [2005] model required a priori
information on backscattering nature of the medium (i.e. molecular versus particulate) to
operate, additional steps are needed in the implementation of our algorithm with respect
to the QAA02. Those are illustrated in Fig. A1, which shows the inputs/outputs data flow
through each step of the processing chain. Table A1 provides a summary of all the
equations used in the different steps. Briefly, the correction factors for both air-water
interface reflection and bidirectional effects for the observation geometry are expressed
by a single fourth-polynomial model as
4
Rrs ( λ,
θ s , θ v , Δ φ,
W , γ b ( λ))
= ∑ gi
( θ s , θ v , Δφ,
W , γ b ( λ))
ωb
( λ)
, (A1)
i=
1
where ωb is the ratio bb/(at+bb), and gi are coefficients tabulated in look-up table (LUT)
for given values of the Sun and sensor zenith angles (θs and θv), the azimuth difference
between the Sun-pixel and pixel-sensor half vertical plane (Δφ), the wind speed (W) and
the phase function parameter, γb(λ). The latter is defined as the relative contribution of
particles to the total backscattering (i.e. bbp(λ)/bb(λ)). Thus the first three steps of the
algorithm (eqs. no. 1-4 in Table A1 and Fig. A1), similar to the QAA02 (steps 0-3 and 5
in Lee et al. 2002), provide an initial estimate of γb at a reference wavelength. With the
initial value of γb(λ0) and Rrs(λ0, θs, θv, Δφ, W), ωb(λ0) is estimated by inverting the
fourth-order polynomial using a numerical solution (no. 5) [Laguerre's method, or the
zroots function in Chap 9.5 of Press et al., 1992]. Two iterations are sufficient to
obtain a difference between two consecutives estimate of γb(λ0)

chlorophyll concentration in case 1 water [e.g., O'Reilly et al., 1998] that uses the
maximum reflectance band ratio between 443, 490 and 510 nm upon 555 nm (r 2 = 0.785,
N = 35; for at-w(555)0.5aw(555)):
234
⎡ Rrs
( 650)
⎤
−0.
69+
1.
41log10
⎢ ⎥
⎣ Rrs
( 555)
⎦
a ( 650)
= a ( 650)
+ 10
(8)
t
w
where aw(650)=0.34 [Pope and Fry, 1997]. For continuity in the at retrieval, when at-
w(555) falls between 0.03 and 0.06 m -1 the a linear combination of the two results
obtained with λ0 = 555 and 650 nm is performed following equation 20 in Lee et al.
[2002].
Third, the spectral particles backscattering coefficient (bbp) is calculated using λ −Y
law with Y = 1. Note that in the QAA02, an empirical relationship based on the
reflectance ratio Rrs(443)/Rrs(555) is used to estimate Y (developed with data from the
Gulf of Mexico).
In Chapter IV, an empirical algorithm was proposed to estimate the ratio between
the CDOM and the total absorption coefficients at 412 nm ([aCDOM/at](412)). Multiplying

the QAA-derived at(412) by the empirically derived [aCDOM/at](412) yield the CDOM
absorption coefficient.
Table A1. List of equations find in Fig. A1.
Equation
number 1
1
2
3
4
If λ0=555 then
a
t
( 555)
Equations Ref.
a
= w
( 555)
235
2
−1.
3 − 2.
53ρ
+ 0.
46ρ
+ 10
⎛ max[
Rrs
( 443,
490,
510)
] ⎞
with ρ = log ⎜
⎟ 10
.
⎝ Rrs
( 555)
⎠
If λ0=650 then
a
t
( 650)
= a
w
( 650)
+ 10
⎡ Rrs
( 650)
⎤
−0.
69+
1.
41log10
⎢ ⎥
⎣ Rrs
( 555)
⎦
with Rrs(650)= 1.082*Rrs(665) or 1.101* Rrs(670)
r
rs
( λ)
=
1
0.
52
2
g 0 + g 0 + 4g1rrs
R
rs
( λ)
+ 1.
7R
rs
( λ)
−
( λ)
ωb
( λ)
= , g0=0.0945 and g1=0.0794
2g
ωb
( λ)
at
( λ)
bb
( λ)
= and bbp(λ)=bb(λ) - bbw(λ)
1−
ω ( λ)
5 ωb(λ)=zroots(Rrs(λ), gi; i=1..4)
6
b
Y
⎛ λ0
⎞
bbp ( λ)
= bbp
( λ0
) ⎜ ⎟
⎝ λ ⎠
7 ( 1 − ω ( λ)
)
b bb
( λ)
at
( λ)
=
ωb
( λ)
1
Numbers correspond to the upper subscript in Fig. A1.
This study
Lee et al.
[2002]
Gordon et al.
[1988]
Press et al.
[1992]
e.g., Gordon
et al. [1988]

Figure A1. Schematic representation of the quasi-analytical algorithm
developed to derive the particles backscattering coefficient at 555 nm and
the spectral total absorption coefficient. The data (parallelogram)
necessary for each steps of the algorithm (rectangular boxes) are identified
by the dotted arrows, while the outputs are identified by the thin line with
filled arrows. The final outputs, i.e. bbp(555) and at(λ), are identified by
the think line with filled arrows at bottom right of the diagram. The
equations (identified with upper script number) are listed in Table A1.
236

A1.3. Results and discussion
A1.3.1. Application of QAA to in situ data set
The QAA02 and its modified version described above (QAA-CASES) were
applied to the remote-sensing reflectance measurements obtained during the CASES field
campaign (Fig. A2). Table A2 provides the statistics on the performance to retrieve at of
both algorithms. The criteria presented includes the 95% confidence interval range for
intercept and the slope of the linear regression (Type 2) between retrieved and measured
at [Sokal and Rohlf, 1995], the geometric standard deviation (GSD 95% ), and the mean
relative difference (AD). A perfect performance of the inverse algorithm would give an
intercept of 0 and a slope of 1. The GSD 95% is calculated as 10 1.966*σ , where σ is the
standard deviation of the difference between measured and retrieved estimate computed
on the log10-transformed data. Note that GSD 95% is almost equivalent to 10 1.966*RMSE
where RMSE is the root-mean-square-error which is commonly used is algorithm
evaluation [IOCCG, 2006]. The interpretation of GSD 95% is easier than the RMSE since it
means that a retrieved value of x (x retrieved ) falls within the range between x true / GSD 95%
and x true * GSD 95% at 95% confidence interval.
As expected, the tuned algorithm (QAA-CASES) provides more accurate estimate
of the total absorption than the QAA02, as revealed by the lower GSD 95% and AD at all
wavelengths (Table A2). For the QAA02, both the intercept and slope parameter deviated
significantly from 0 and 1 respectively, indicating a bias in the retrieval of at. The low
slope of the regression obtained using the QAA02 is largely explained by the use of an
inappropriate relationship for Y in the inversion. In fact, when the QAA02’s empirical
formulation for Y is used in instead of 1.0 in the QAA-CASES, the slope of the regression
between retrieved and measured at(412) was significantly lower than 1. In addition, the
performance of the QAA-CASES decreases with decreasing wavelength as indicated by
the higher values for GSD 95% and AD. These results also suggest that the value of Y is not
constant in this area. While studies carried out in the Arctic found empirical relationships
between Y and the magnitude of bb(555) [Stramska et al., 2003; Wang et al., 2005], no
significant relationships between neither Rrs ratios nor the magnitude of bp(555) and the
measured Y values were find in this study (not shown). Hence the highest Y values (~1.4),
as derived from the bp(λ) measured with the ac-9, were observed near the Mackenzie
237

River mouth, while offshore the values ranged between 0.3 to 1.0. This trend is opposite
to the one proposed by Lee et al. [2002] and Morel and Maritorena [2001] where Y
decreases with increasing turbidity. Here a value of 1.0 was found to give the best results
in the retrieval of at at 412 nm.
Figure A2. Comparison between retrieved and measured at with the ac9 at
A) 412 nm and B) 440 nm. Total absorption coefficients were retrieved
using the optimized version of QAA proposed by Lee et al. [2002] that use
λ0 555 and 640 nm (open circles) and the tuned QAA with the CASES
dataset (inverted triangles).
238

Table A2. Performance of the QAA to retrieved at(λ) with empirical relationships of Lee
et al. [2002] versus QAA tuned with CASES dataset (n=46).
Wavelength Approach Intercept
(nm)
a Slope a 95% b
GSD AD c
(%)
412 QAA02 [-0.138, -0.043] [0.809, 0.945] 1.51 16.2
QAA-CASES [-0.019, 0.054] [0.965, 1.069] 1.38 12.9
440 QAA02 [-0.163, -0.062] [0.811, 0.928] 1.42 14.3
QAA-CASES [-0.030, 0.042] [0.963, 1.046] 1.28 10.1
490 QAA02 [-0.179, -0.059] [0.799, 0.909] 1.35 18.9
QAA-CASES [-0.044, 0.053] [0.930, 1.019] 1.29 12.9
510 QAA02 [0.736, 0.839] [0.736, 0.839] 1.27 9.9
QAA-CASES [-0.028, 0.075] [0.997, 1.094] 1.25 10.1
555 QAA02 [-0.174, -0.026] [0.818, 0.962] 1.25 9.4
QAA-CASES [-0.006, 0.124] [0.992, 1.119] 1.22 7.4
a
95% Confidence interval calculated on the log10-transformed data using Type II regression
model with the major axis (MA method based on Sokal and Rohlf [1995],described by Pierre
Legendre, 2001: http://www.bio.umontreal.ca/casgrain/fr/labo/model-ii.html).
b 95%
1.
966*
σ log
Geometric Standard Deviation (95% confidence interval): GSD = 10 , where σlog is the
standard deviation calculated on the log10-transformed difference between retrieved and measured
values.
c Mean absolute difference in %,
retrieved measured
n ai
− ai
AD = ∑
measured
n a
1 .
i
i
The inaccuracy of the QAA-CASES could also become from the empirical
relationships used to obtain the total absorption at the reference wavelength (Eqs. 7 and
8). The mean absolute error made on at(555) and at(650) was only 5.1% and 2.5%
respectively. Those are lower than the error made using the QAA02’s empirical
relationships (~10%).
Unlike most semi-analytical models currently available [see review by IOCCG,
2006], the proposed version of the QAA accounts for the non-isotropic nature of the
radiant light field that emerged from the ocean [Loisel and Morel, 2001; Morel et al.,
2002; Park and Ruddick, 2005]. This dependence is largely explained by the nature of
light scattering (i.e. molecular vs particulate) [Lee et al., 2004; Morel et al., 2002] which
is included in the model form adopted here. According to Park and Ruddick [2005],
bidirectional effects (BRDF) due to illumination-viewing geometry and the nature of the
scattering can explain ~10% of the variability in the relationship between Rrs and the ratio
bb/(at+bb). When satellite data acquired in various illumination conditions are averaged to
239

produce level 3 data, the proposed algorithm should minimize the undesirable variability
resulting from the ocean BRDF.
Unfortunately, no in situ measurements of bbp were performed during CASES. So
I compared the retrieved bbp(555) values versus the measured bp(555) values (Fig. A3). A
relatively strong correlation was obtained for the linear regression between the logarithm-
transformed data (r²=0.86; N=46). Interestingly, the average b bp
~ calculated using
retrieved bbp(555) and measured bp(555) was 1.4% with a standard deviation of 0.7%.
Those are within the expected range for particles found in coastal waters [e.g., Chami et
al., 2005; Twardowski et al., 2001; Wang et al., 2005].
Figure A3. Comparison between QAA-derived bbp(555) and measured
bp(555).
A1.3.2. Estimation of aCDOM
To estimate aCDOM, several investigators have proposed empirical relationships
based on a single water-leaving radiance or reflectance ratio [D'Sa and Miller, 2003;
Johannessen et al., 2003; Kahru and Mitchell, 2001; Kowalczuk et al., 2005]. We fitted
all possible ratios from the 13 Rrs SPMR channels with the aCDOM values at 412 nm. The
240

egression between the ratio of Rrs at 412 nm to Rrs at 665 nm with aCDOM at 412 nm gives
the highest correlation (aCDOM(412)=10 -.301-.656*[Rrs(412)/Rrs(665)] ; r 2 =0.924, GSD 95% =1.63,
AD=26.8%). For comparison, we estimate aCDOM at 412 nm multiplying the QAA-derived
at(412) by the empirically-derived [aCDOM/at](412) (Chapter IV). Figure A1.4 compares
the retrieved aCDOM(412) with its measured value. Combining the results from the QAA
and the empirical algorithm provides a more robust estimate of the CDOM absorption
coefficient than an empirical algorithm based on a single Rrs ratio in this region
(GSD 95% =1.45, AD=14.8%).
Figure A4. Retrieved versus measured aCDOM(412) in the Beaufort Sea.
Statistical parameters of the Type II regression are also shown.
A1.4. Conclusion
In view of deriving IOPs from SeaWiFS imagery in the Beaufort Sea, a semi-
analytical algorithm was modified 1) to account for the anisotropic nature of the oceanic
upwelling light field and 2) for the regionally distinct relationships between absorption
coefficients at reference wavelength (i.e. 555 or 650 nm) and various reflectance ratios.
The method provides an estimation of the total absorption coefficient at 412 and 443 nm
within a precision of about 38 and 28%, respectively, at confidence level of 95%.
Combined with the [aCDOM/at](412) algorithm (Chap. IV) the CDOM absorption
coefficient could be estimated within a precision of 45% at confidence level of 95%. For
241

the Beaufort Sea, the latter performance is better than any other single band ratio found in
the literature.
Satellite-derived IOPs (bbp and at) could be further used to study the spatio-
temporal variability of the Makcenzie River plume and to quantify the particles
distribution in the surface waters (Forest et al., Particulate organic carbon fluxes on the
Mackenzie Shelf slope: physical and biological forcing shelf basin exchanges, submitted
manuscript to Journal of Marine Systems, 2006). Finally, the magnitude of the CDOM
absorption coefficient can also be use as an indicator of the DOM photoreactivity (see
Chap. VI).
A1.4. References
Chami, M., E. B. Shybanov, T. Y. Churilova, G. A. Khomenko, M. E. G. Lee, O. V.
Martynov, G. A. Berseneva, and G. K. Korotaev (2005), Optical properties of the
particles in the Crimea coastal waters (Black Sea), J. Geophys. Res., 110(C11),
C11020, doi:10.1029/2005JC003008.
D'Sa, E., and R. L. Miller (2003), Bio-optical properties in waters influenced by the
Mississippi River during low flow conditions, Remote Sens. Environ., 84, 538-
549.
Huot, Y., C. A. Brown, and J. J. Cullen (2005), New algorithms for MODIS sun-induced
chlorophyll fluorescence and a comparison with present data products, Limnol.
Oceanogr.: Methods, 3, 108-130.
IOCCG (2006), Remote sensing of inherent optical properties: Fundamentals, tests of
algorithms, and applications, edited by Z.-P. Lee, pp. 126, Reports of the
International Ocean-Colour Coordinating Group, No. 5, IOCCG, Dartmouth,
Canada.
Johannessen, S. C., W. L. Miller, and J. J. Cullen (2003), Calculation of UV attenuation
and colored dissolved organic matter absorption spectra from measurements of
ocean color, J. Geophys. Res., 108(C9), 3301, doi:10.1029/2000JC000514.
Kahru, M., and G. B. Mitchell (2001), Seasonal and nonseasonal variability of satellitederived
chlorophyll and colored dissolved organic matter concentration in the
California Current, J. Geophys. Res., 106(C2), 2517-2529.
Kowalczuk, P., J. Olszewski, M. Darecki, and S. Kaczmarek (2005), Empirical
relationships between coloured dissolved organic matter (CDOM) absorption and
apparent optical properties in Baltic Sea waters, Int. J. Remote Sens., 26(2), 345-
370.
Lee, Z.-P., K. L. Carder, and R. A. Arnone (2002), Deriving inherent optical properties
from water color : a multiband quasi-analytical algorithm for optically deep
waters, Appl. Opt., 41(27), 5755-5772.
242

Lee, Z.-P., K. L. Carder, and K. Du (2004), Effects of molecular and particle scatterings
on the model parameter for remote-sensing reflectance, Appl. Opt., 43(25), 4957-
4964.
Loisel, H., and A. Morel (2001), Non-isotropy of the upward radiance field in typical
coastal (Case 2) waters, Int. J. Remote Sens., 22(2&3), 275-295.
Macdonald, R. W., F. A. McLaughlin, and E. C. Carmack (2002), Fresh water and its
sources during the SHEBA drift in the Canada Basin of the Arctic Ocean, Deep-
Sea Research I, 49, 1769-1785.
Morel, A., D. Antoine, and B. Gentili (2002), Bidirectional reflectance of oceanic waters:
accounting for Raman emission and varying particle scattering phase function,
Appl. Opt., 41(30), 6289-6306.
O'Reilly, J. E., S. Maritorena, G. Mitchell, D. A. Siegel, K. L. Carder, D. L. Garver, M.
Kahru, and C. R. McClain (1998), Ocean color chlorophyll algorithms for
SeaWiFS, J. Geophys. Res., 103(C11), 24937-24950.
Park, Y.-J., and K. G. Ruddick (2005), Model of remote-sensing reflectance including
bidirectional effects for case 1 and case 2 waters, Appl. Opt., 44(7), 1236-1249.
Pope, R. M., and E. S. Fry (1997), Absorption spectrum (380-700 nm) of pure water. II.
Integrating cavity measurements, Appl. Opt., 36, 8710-8723.
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery (1992), Numerical
recipes in Fortran 77: The art of scientific computing, Chap 9 pp., Cambridge U.
Press, New York.
Sokal, R. R., and F. J. Rohlf (1995), Biometry: principles and practice of statistic in
biological research, Third ed., 887 pp., Freeman, New York.
Stramska, M., D. Stramski, R. Hapter, S. Kaczmarek, and J. Ston (2003), Bio-optical
relationships and ocean color algorithms for the north polar region of the Atlantic,
J. Geophys. Res., 108(C5), 3143, doi: 10.1029/2001JC001195.
Twardowski, M. S., E. Boss, J. B. Macdonald, W. S. Pegau, A. H. Barnard, and J. R. V.
Zaneveld (2001), A model for estimating bulk refractive index from the optical
backscattering ratio and the implication for understanding particle composition in
case I and case II waters, J. Geophys. Res., 106(C7), 14129-14142.
Wang, J., G. Cota, and D. A. Ruble (2005), Absorption and backscattering in the
Beaufort and Chukchi Seas, J. Geophys. Res., 110(C4), C04014, doi:
10.1029/2002JC001653.
243