Biogeochemical Model 

The development of a biogeochemical model coupled to an Ocean General Circulation model at the LODYC (Stoens et al, 1998, 1999) allowed to test the impact of interannual variability on the forage in the equatorial 20N-20S area (Lehodey et al., 1998).

In 1999, an important development in the model concerned the interaction between predicted tuna density and forage density, that is the coupling between the forage (prey) and tuna (predator) population. Indeed, in the first version, the forage mortality due to the tuna species was assumed to be included in the total mortality ll. However, the use of the predicted distribution of forage to constrain the movement of tunas without considering the feedback effect of tuna density distribution has implicit hypotheses.Either the forage mortality due to the tuna predation is assumed very low and relatively negligible to the total mortality of forage, or the predators present an "ideal distribution". In this ideal distribution, the forage predators would have a natural ideal distribution, such that forage mortality would be the same everywhere and equal to l. This hypothetical situation presents an analogy with the "ideal free distribution" proposed by (Fretwell and Lucas 1970) and (Fretwell 1972) for the density dependent habitat selection theory, whereby individuals differentially occupy available habitats so that realized "suitability" is equal for all occupied habitats.

Forage Mortality

To have more realistic predator-prey interactions, and eventually to test these hypotheses, the forage mortality has to be dependent on the tuna density. The approach used was to calculate and apply first a specific local mortality w_i,j,t ,j,t due to the food requirements of tuna population described in the model, then a mean residual mortality l' which is the difference between the total mortality l and the mean specific mortality over the area occupied by the tuna species. Therefore, the total forage biomass over the whole area remains equal to the total forage biomass calculated in the case of a constant l, but the spatial distribution linking the density of tuna may be different. An increasing density of tuna increases the forage mortality, and if there is no additional supply of forage, decreases the habitat index value. Since the movement of tuna is based on the gradient of the index, tuna will start to leave the zone when forage is not abundant enough to support the local tuna population density. On the other hand, tuna will continue to concentrate if the index value remains higher than in neighboring zones. This approach seems appropriate for reproducing tuna behavior, particularly the frequent huge aggregations of tuna feeding on large patches of prey organisms.

The new version called SEPODYM was first tested with the previous LODYC’s run of new primary production. Despite some limitations in the input data set and a simplified parameterization, this first simulation improved the prediction of catch and was helpful to interpret the observed ENSO-related spatio-temporal changes in the distribution of skipjack population (Lehodey, 2001).

However, the run presented a short time-series (1992-95) and was limited to the 20^oN–20^oS equatorial region when tuna stocks extend to sub-tropical and temperate oceanic regions. Furthermore, both observations relative to the development of the recent El Niño – La Niña sequence (1997-2000) and results from Multifan-CL providing the first statistical recruitment estimates suggested that a simple temperature constrain was not sufficient to explain the fluctuation of the recruitment.

The opportunity to consider a longer time-series over the whole Pacific basin arose with the recent developments in NPZ models. Other environmental effects than temperature in the spawning habitat index are also investigated. With temperature and physical constraints like the advection creating favourable zones of retention for larvae, and that are already considered in the model, the food availability and the predation are likely the other major factors that affect larval survival and pelagic fish recruitment. Food of larvae is directly depending of primary production, while their predators are organisms included in the tuna forage. Therefore, it was interesting to investigate these two opposite effects on the recruitment by using the ratio P/F.

The use of this new input data and spawning habitat index set provide spectacular results. The skipjack recruitment defined by environmental constrains in SEPODYM is converging with the independent statistical estimate from Multifan-CL and the spatial correlation between observed and predicted catch is high.

When the spawning habitat index is only dependent of SST the recruitment and biomass do not show any large fluctuation as suggested by the MULTIFAN-CL analysis. Using P/F ratio in the spawning habitat index produces fluctuations showing an evident correlation with the MULTIFAN-CL estimates. The high spatial correlation between observed and predicted catch even with a relatively homogeneous recruitment depending of SST only (a = 0) suggests that the dynamics are relatively well described in the model, at least at this large spatial scale. In particular, the redistribution of larvae by currents and the movement of young and adults tuna constrained by the habitat index are fundamental processes to explain the distribution of the population.

The large interannual variations in the recruitment are related to the ENSO (El Niño Southern Oscillation) events, the recruitment being the highest during El Niño years (1972, 1982-83, 1987, 1990) and the lowest during La Niña years (1974-76, 1988-89). The simulations predict that the main skipjack spawning ground occurs in the western Pacific between Indonesia and Papua New Guinea, with a general decreasing gradient from west to east. During El Niño events the recruitment of juvenile increases drastically with a spatial extension to the central Pacific, while during La Niña it is contracted in the western Pacific and with a lower level. A large proportion of these recruits moves either to Japan following the Kuroshio or to the central Pacific in the convergence zone between the warm pool and the cold tongue as already described in previous studies (Lehodey et al. 1997, 1998, Lehodey 2001) and in agreement with the results of the skipjack MULTIFAN-CL analysis (Bigelow, 2000, Hampton et al 2001).

Monthly spatial (1 degree square) correlation between observed and predicted skipjack catch for Japanese pole and line fleet and purse seiners fleets in the western central and eastern Pacific Ocean.

Recruitment mechanisms

Despite some limitations, the use of a new set of predicted primary production and physical data has considerably extended the analysis, as the simulations cover all the Pacific Ocean since 1955, and also because prediction of primary production are improving. The next simulations should include the period between 1992 to day, allowing comparisons with recent and numerous observed data (e.g., SeaWiFS) including the strong El Niño of 1997-98 followed by the long La Niña event of 1998-2001.

Recruitment mechanisms are fundamental processes in population dynamics. The spatial model SEPODYM is particularly well suited to test hypotheses on these mechanisms and on the stock-recruitment relationships. The results concerning the skipjack recruitment are more than encouraging. However, with the improvement of recruitment prediction it is necessary to improve the parameterization of young and adult movements. In particular, advection and diffusion coefficients of tuna could be age-dependent to have a more realistic description of the physical capability of movement associated to the size of fish.

These results confirm the impact of the ENSO variability with a positive (negative) effect of El Niño (La Niña) events on the recruitment that is propagated into the stock in the following two years. Therefore, after the 1997-98 El Niño event and the associated high record of skipjack catch in 1998-2000, the last La Niña episode of 1998-2001 should lead to a decrease of the skipjack stock biomass in the next two years.
   

Skipjack population and catch	Skipjack recruitment
Forage

Animations (click on the image to see the animation)
You can download a free video viewer (753 Kb) at http://www.gromada.com/Moyager.html