Spatial analyses of the influence of autocorrelation on seasonal diet composition of a marine fish species

Abstract

Spatial autocorrelation is common in environmental factors, species distributions, and predator-prey interactions. Spatially autocorrelated distribution of prey and strong prey selectivity by predators may cause spatial autocorrelation of the fish diet shown as higher similarity among diets of fish caught nearby or in the same trip. We explore the importance of spatial autocorrelation analysis in understanding the fish diet composition and predator-prey interaction in diet analysis based on one example fish whitespotted conger (Conger myriaster) in the Yellow Sea and its prey species that have been monitored across seasons. Among the spatial estimators, the cluster-based estimator is used to quantify fish diet similarity from the same trip, and the spatial Gaussian and exponential autocorrelated estimators are used to test the hypothesis that fish diet similarity decreases over distance. The two types of spatially autocorrelated estimators performed better than spatially-independent and cluster-based estimators, with lower mean square errors in a cross-validation procedure. Prey selectivity related to the prey availability, and the spatial overlap between predator and prey were likely important factors affecting seasonal and spatial variations in diet composition of the example fish species. Cross-variograms indicated that whitespotted conger and some of its dominant prey species were positively spatially autocorrelated, which suggested fishes generally aggregated in areas of high prey density. Our study provides a practical basis for considering the effect of spatial characteristics in quantifying the diet composition likely linked to the environmental driven distributions of both fishes and their prey.