Resumen del producto

food-web relationships so that community ecology may contribute to the process of setting management policy. Changes over time in the structure of pelagic open-ocean food webs are difficult to quantify, and fisheries-independent trawl surveys are expensive and biased against larger micronekton. Tropical tunas, however, are considered opportunistic, non-selective predators due to their broad diet and high energy requirements in oligotrophic habitats. Assuming tropical tunas are effective biological samplers of middle-trophic-level communities, comparative diet studies across temporal scales have the potential to provide valuable clues of changing ecological interactions within the food web. Yet, the analysis of diet data has suffered from the lack of robust statistical methods due to their multivariate nature, non-linearity, unbalanced distributions, and missing values. To examine the feeding ecology of yellowfin tuna, Thunnus albacares, in the eastern Pacific Ocean, we applied a bagged classification tree approach to explore the relationships of spatial, temporal, environmental, and biological predictor variables to yellowfin diet composition. Stomach samples from 6810 yellowfin tuna were taken from 433 purse-seine sets during 2 sampling periods separated by a decade. We analyzed a subset of 18 dominant groups of prey taxa based on their gravimetric contribution to the diet of 3122 yellowfin tuna whose stomachs contained food of certain taxa: three groups of cephalopods, three groups of crustaceans, and twelve groups of fishes. A spatial bootstrapping technique was utilized, to account for spatial dependence between samples and estimates of precision were provided accordingly. We quantified the importance of predictor variables and produced partial dependence plots to explore relationships between the predicted prey composition and covariates included in the model. Classification tree results and variable importance measures indicated that spatial covariates, latitude and longitude, are important covariates in the prediction of yellowfin tuna diet composition followed by quarter, the Pacific Decadal Oscillation index, year, and yellowfin size. Results revealed that yellowfin consumed a diversity of prey overall, but often high proportions of one or two dominant prey characterized the diet of fish at different regions. The classification tree highlighted differences in yellowfin diet composition north of 17°N versus south of that latitude. In the north, diet composition was diverse, consisting of crustaceans, squid, and fishes, whereas various fish taxa were the predominant prey in the south. A tree split on year revealed feeding differences between the two decadal sampling periods, with crustaceans and mesopelagic fishes the predominant prey during the 2003-2005 sampling period. In contrast, epipelagic fishes dominated the diet during the 1992-1994 sampling period. Classification tree methodology explores differences in complex feeding dynamics and examines their variation in relation to potential shifting environmental and fishing conditions. Our findings highlight the importance of conducting diet studies across temporal scales to elucidate change

Palabras clave:

Para obtener una copia del documento contacta la personal de la biblioteca a través del correo bibliocicimar{a}ipn.mx