08.06.2013 Views

Bottom Trawl Surveys - Proceedings of a Workshop Held at Ottawa ...

Bottom Trawl Surveys - Proceedings of a Workshop Held at Ottawa ...

Bottom Trawl Surveys - Proceedings of a Workshop Held at Ottawa ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

108<br />

The uncertainties <strong>of</strong> weight estim<strong>at</strong>es are, on<br />

the other hand, a result <strong>of</strong> the sampling design<br />

and sampling r<strong>at</strong>e <strong>of</strong> commercial c<strong>at</strong>ch.<br />

The above analysis considers th<strong>at</strong> the Fo.1<br />

fishing levels and the partial recruitment<br />

coefficients, i.e. our controls, are known<br />

quantities. For instance, our results are valid<br />

for an instantaneous r<strong>at</strong>e <strong>of</strong> fishing mortality<br />

<strong>of</strong> 0.195. In practice, the Fo.1 fishing level<br />

is estim<strong>at</strong>ed from mean weight-<strong>at</strong>-age and from<br />

partial recruitment coefficients; this estim<strong>at</strong>e<br />

is not without errors. We can assess the effect<br />

<strong>of</strong> the uncertainties <strong>of</strong> the FQ.1 reference level<br />

on c<strong>at</strong>ch projections by extending the projection<br />

algorithm so as to include the simultaneous<br />

evalu<strong>at</strong>ion <strong>of</strong> FQ.1· By definition, Fo.1 is the<br />

value <strong>of</strong> F.,T which ensures th<strong>at</strong> the first<br />

partial deriv<strong>at</strong>ive <strong>of</strong> yield per recruit with<br />

respect to F.,T is equal to one tenth the value<br />

<strong>of</strong> the slope <strong>of</strong> the yield per recruit curve <strong>at</strong><br />

the origin. Rivard (1980) suggests the use <strong>of</strong><br />

the method <strong>of</strong> false position for estim<strong>at</strong>ing Fo.1<br />

from age-specific weights and from the<br />

coefficients <strong>of</strong> partial recruitment; we used<br />

th<strong>at</strong> method for estim<strong>at</strong>ing the Fo.1 reference<br />

level in the projection algorithm. Variances <strong>of</strong><br />

c<strong>at</strong>ch projections with the extended method are<br />

given in Table 9. While the rel<strong>at</strong>ive error for<br />

the 1981 stock size estim<strong>at</strong>e is not modified by<br />

this extension <strong>of</strong> the projection algorithm, the<br />

rel<strong>at</strong>ive error <strong>of</strong> the 1981 c<strong>at</strong>ch biomass (TAG)<br />

goes from 15% to 18%. This 3% difference<br />

represents the increase in rel<strong>at</strong>ive error due to<br />

the fact th<strong>at</strong> Fo.1 is now estim<strong>at</strong>ed from input<br />

d<strong>at</strong>a. It is interesting to note th<strong>at</strong> the<br />

variances <strong>of</strong> age-specific weights now account<br />

for 74% <strong>of</strong> the total variance <strong>of</strong> 1981 c<strong>at</strong>ch<br />

biomass, in comparison to 60% with 'fixed'<br />

projections. These results re-enforce the<br />

conclusions presented in the preceding paragraph<br />

regarding the importance <strong>of</strong> age-specific weight<br />

estim<strong>at</strong>es for c<strong>at</strong>ch projections.<br />

Similarly, various methods are used for<br />

estim<strong>at</strong>ing the coefficients <strong>of</strong> partial<br />

recruitment but the precision and the accuracy<br />

<strong>of</strong> such estim<strong>at</strong>es are generally unknown. In<br />

general, we expect a neg<strong>at</strong>ive correl<strong>at</strong>ion<br />

between estim<strong>at</strong>es <strong>of</strong> partial recruitment and<br />

stock size estim<strong>at</strong>es for the current year.<br />

Consequently, the results presented hereabove<br />

regarding the uncertainties <strong>of</strong> c<strong>at</strong>ch projections<br />

will not be affected if the covariance term due<br />

to this neg<strong>at</strong>ive correl<strong>at</strong>ion is <strong>of</strong> the same<br />

order <strong>of</strong> magnitude than the variance term due to<br />

partial recruitment estim<strong>at</strong>es. There is,<br />

however, an additional source <strong>of</strong> error which is<br />

rel<strong>at</strong>ed to the use <strong>of</strong> current estim<strong>at</strong>es <strong>of</strong><br />

partial recruitment for the years <strong>of</strong> projection:<br />

the current approach assumes th<strong>at</strong> the fleet<br />

composition, the gears being used and fish<br />

availability are constant for the period <strong>of</strong><br />

projection. Such error, which is not random and<br />

which is independent <strong>of</strong> stock size estim<strong>at</strong>es,<br />

does not cancel out and could introduce<br />

inaccuracies in c<strong>at</strong>ch projections. In general,<br />

such an error will influence more seriously the<br />

estim<strong>at</strong>es <strong>of</strong> partial recruitment for the<br />

youngest age-groups, which age-groups do not<br />

TABLE 9<br />

Effect <strong>of</strong> various options on the rel<strong>at</strong>ive error<br />

<strong>of</strong> calcul<strong>at</strong>ed quantities in c<strong>at</strong>ch projections:<br />

example is for cod in 4T-4Vn.<br />

A. Current methods (24%)*<br />

B. Projecting from<br />

estim<strong>at</strong>es <strong>of</strong> stock<br />

size <strong>at</strong> the beginning<br />

<strong>of</strong> the current year<br />

(Doubleday's method):<br />

a. FQ·1 known<br />

Rel<strong>at</strong>ive error for<br />

1981 1981<br />

stock size c<strong>at</strong>ch<br />

biomass<br />

1. with historical d<strong>at</strong>a 11% 15%<br />

2. annual mean<br />

weight <strong>at</strong> age<br />

calcul<strong>at</strong>ed from a<br />

von Bertalanff<br />

growth curve 11% 14%<br />

3. 25% reduction <strong>of</strong> the<br />

within-year variance<br />

<strong>of</strong> research survey<br />

estim<strong>at</strong>es 10% 14%<br />

4. the rel<strong>at</strong>ive error<br />

for mean weight <strong>at</strong><br />

age reduced to 10% 11% 11%<br />

5. Applic<strong>at</strong>ion <strong>of</strong> 3<br />

and 4 above 10% 10%<br />

b. FQ·1 expressed as a<br />

function <strong>of</strong> input<br />

inform<strong>at</strong>ion (with<br />

historical d<strong>at</strong>al 11% 18%<br />

c. Effect <strong>of</strong> some system<strong>at</strong>ic<br />

(non random) errors when<br />

method B is applied.<br />

1. 100% increase <strong>of</strong><br />

partial recruitment<br />

coefficient <strong>at</strong> age 3<br />

and 45%, <strong>at</strong> age 4<br />

for 1980 and 1981** 0.1% -2%<br />

2. In 1979, unreported<br />

50% discard <strong>at</strong> sea<br />

for age 3, 25% for<br />

age 4 0.1% 0.1%<br />

3. Unreported 80%<br />

discard <strong>at</strong> sea for<br />

age 3, from 1970<br />

to 1979 -0.5% -0.6%<br />

4. In 1979, 20% 'underreporting'<br />

for<br />

a 11 age groups 0.7% 2.5%<br />

*See Table 1; refers to the 1977 stock size.<br />

**Results also valid for method A; considers<br />

the effect <strong>of</strong> a change in r 1 on Fo.l·

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!