Abstract
This study provides an approach to the comprehension of the mesoscale structure in the habitat patterns of three major tropical tunas species, yellowfin (Thunnus albacares; YFT), skipjack (Katsuwonus pelamis; SKJ) and bigeye (Thunnus obesus: BET). These species, mainly caught by the purse seiner's fishery worldwide, represent 20% and 13% of the world total production for the Indian and the Eastern Pacific oceans respectively. Single set records from this fishing gear in both, the Indian and in the Eastern Pacific oceans, were used to separately evaluate the environmental characteristics of their fishing grounds for three fishing modes: free school (FS), log (LOG) and fish aggregated devices (FADs) and several individual sizes. Prior to the analyses, a statistical expert-based method was applied to detect and classify thermal fronts at basin scale whereas the anticyclonic and cyclonic eddies are detected, based on the winding angle method. The distances of the catch positions from these structures were then calculated. We found that strong fronts are mostly found in coastal regions and weak fronts in the open ocean. This allows the consideration of the frontal intensity helps to spatially differentiate mechanisms of frontogenesis than may attract tunas.In addition to these mesoscale components, classic, temporal and fishery-related variables were added to describe the fishing environment. We used the Boosted Regression Tree (BRT) method to create a three step modeling scheme for each study area in order to explore the responses of the catch level and size for the different fishing modes, also considering the effect of randomly distributed catch positions, in order to separately estimate the mesoscale effects. The relative dominance of each species is also explored and globally shows similar results than for the catch level.All BRT models show that the catch level was better explained by the environment for free school (FS) than for FADs fishing modes and that the differences in distribution were more important among fish-sizes than among species. We quantify for the first time the strong influence of the mesoscale in determining tuna's habitat, confirmed by the poor explanation obtained in all random models, mainly for the IO. For both oceans, small individual were strongly related with the proximity to mesoscale eddies (<200 km) whereas the larger individuals area found at larger distances. A low influence of mesoscale fronts was found mainly for the Indian ocean whereas the opposite pattern was observed for the EPO, except in the coastal regions where strong fronts become important. Different environmental conditions were observed in well defined sub-regions as the coastal upwelling, the equatorial tongue, and the Costa Rica Dome. Contrarily to the Indian ocean were the fishing grounds are relatively homogeneous in term of Sea Surface Temperature and chlorophyll concentration, these two parameters highly influence the tuna distribution in the EPO. In the Indian Ocean, the total percentages of relative contribution by category of variables for all models were 34% for the mesoscale, 39% for the classical variables and 27% for others (temporal and fisheries related). For the Eastern Pacific, even if mesoscale remains important (37%), the most relevant ones were the classical variables with 55% of the total relative contribution (8% for others). Key word: Tropical tunas, Indian Ocean, Eastern Pacific Ocean, mesoscale, purse seiner, species, fish-size, fishing mode.