Figure 6.10. Frequency of occurrence (FO) of capelin (dotted line) and cod (solid line) in cod stom‐
achs in the Barents Sea 1947–2005.
6.5 Newfoundland Shelf – progress in the study of the diet of marine predators with emphasis on diet interannual variability
Although the diet of many predators of the Newfoundland Shelf and Grand Bank, Northwest Atlantic, have received considerable attention over the years (Lilly and Fleming 1981, Lilly 1991, Bowering and Lilly 1992, Pedersen and Riget 1993, Pedersen 1994, Orr and Bowering 1997, Lilly 1994, Stenson et al., 1997, Lilly et al., 2000, Stenson and Perry 2001, Stenson and Hammill 2006), there are a surprisingly few studies at‐
tempting comparative analyses among them.
Given the importance that diet changes can have on the estimations of consumption by predators, as well as their impact in the parameterization of multispecies and eco‐
system models, studies on the diet of several predators have been re‐initiated and/or enhanced in the Newfoundland and Labrador Region of Fisheries and Oceans Can‐
ada (DFO). This work is being carried out as part of a national DFO Ecosystem Re‐
search Initiative (ERI), implemented in this region through the ERI‐NEREUS program.
In collaboration with other initiatives and DFO Centres of Expertise ) like the Centre of Expertise in Marine Mammals (CEMAM), a series of studies involving the re‐analysis and modelling of archived diet data, as well as collection and analysis are currently underway. As part of this process, some preliminary results on a novel approach to model diet data (Buren et al., 2008a) and a comparative diet analysis among cod, turbot and harp seals (Buren et al., 2008b) have been recently presented at the ICES‐NAFO‐NAMMCO Symposium on “The Role of Marine Mammals in the Ecosystem in the 21st Century” held in Dartmouth, Canada, on Sep 29–Oct 1 2008.
The first of these aforementioned papers describes a new method to model diet com‐
position by weight (Buren et al., 2008a). This approach is based on the fact that the classical calculation of proportion by weight in the diet can also be represented in terms of probability of finding a given prey in a stomach and the mean weight of that prey in the stomach when it is present.
For example, if N is the total number of predators in a sample, nx is the number that actually contains a given prey x, and Bx is the total biomass of x in the collection, then
( )
and where px is the probability of finding prey x in a stomach and mx is the mean weight of x in a stomach when x is actually present. The usefulness of this reparame‐
terization is that allows for estimating pi and mi independently.
Under the assumption that each prey species in a stomach is consumed independ‐
ently, using the above parameterization when considering all prey species in a stom‐
ach or just one of them should render similar results if the sample size is large enough. Buren et al. (2008) tested this concept using harp seal diet data by comparing the classical diet analysis with the proportions obtained by only considering the most important prey in biomass in each stomach. Of course, the pi and mi calculated were adjusted accordingly (i.e. pi was defined as the probability of being a main prey while
mi was calculated as the mean weight of prey when it is a main prey). The results ob‐
tained indicated that the approach was sensible (Figure 6.11).
Figure 6.11. Comparison between the prey proportions by weight in the diet of harp seal from Newfoundland, Canada, using classical diet analysis and using only the main prey (see text for details).
Considering only main prey in the stomach allows for a more credible support for the assumption that individual data are statistically independent. Furthermore, individ‐
ual stomachs can be thought as the realization of a multinomial experiment (one prey species is selected as main prey from a finite set of potential prey), and hence, the probability pi can be estimated using multinomial regression models (a.k.a. multi‐
category logit models) (Agresti 2002). These probabilities, in conjunction with esti‐
mated (or reasonably approximated) mi values can be used to reconstruct the predator diet. If time is one of the significant factors in these regressions, then the expected changes in diet over time can be visualized (Figure 6.12).
Classic Main Prey
N seals: 5039
Classic Main Prey
Arctic cod Atlantic cod Atlantic herring Capelin Gadiforme Other Fish Other Invertebrates Pandalus Pleuronectidae Sandlances Other Shrimp Zooplankton
Figure 6.12. Predicted change over time for adult harp seals in the inshore waters of Newfound‐
land as estimated using the multinomial regression approach (MR) compared with the classic diet analysis.
This approach was then applied to cod (Gadus morhua), turbot (Reinhardtius hippoglos‐
soides), and harp seals (Pagophilus groenlandicus) diet data to compare the predicted diets for these three top predators in the offshore of the Southern Labrador and Northern Newfoundland Shelf (NAFO Divisions 2J3KL) before and after the collapse of groundfish stocks (Figure 6.13) (Buren et al., 2008b).
Figure 6.13. Modelled diet for cod, turbot, and harp seals in the offshore waters of NAFO divi‐
sions 2J3KL during 1986–1996. Years are grouped in two periods (1986–1991 and 1991–1996) asso‐
ciated with the pre and post groundfish stock collapses (most notably, but not exclusively, cod).
This analysis indicated that the diet of cod was the most stable among all three preda‐
tors, while the diet of both harp seals and turbot showed a higher variability over time (Figure 6.13). Turbot was the predator that had the highest trophic plasticity, showing an important reduction of capelin and redfish (Sebastes sp.) in the diet and an increase of invertebrates (Figure 6.13). Harp seals, on the other side, showed a de‐
crease in Arctic cod (Boreogadus saida) and an increase in sandlance (Ammodytes
0
1986 1991 1996 2001
Classic
1986 1991 1996 2001
MR
Arctic Cod Atlantic Cod Atlantic Herring Capelin Gadiforms Other Fish Other Invertebrates Pandalus shrimp Flatfish Sandlance Other Shrimp Zooplankton
Small Medium Large Cod 1992-1996
Small Medium Large Harp seal 1992-1996
Small Medium Large Turbot 1992-1996 Cod
1986-1991
Harp seal 1986-1991
Arctic Cod Atlantic Cod Capelin Sandlances Gadiforme Pleuronectidae Redfishes Other Fish Unidentified fish Shrimp Amphipods Zooplankton Other Invertebrates
7 ToR e) – Review further progress in multispecies and ecosystem modelling throughout the ICES region
In 2007 the working group provided a broad overview of the multispecies models currently used within each ICES Ecoregion. This chapter of the WGSAM 2008 report offers an update on further developments and progress made within the different geographic areas: