Bayesian modeling-based analysis on the shared habitat and species association between four Gobiidae in a marine bay ecosystem

Abstract

In recent years, with the decline in marine fishery resources, ecosystem-based fisheries management (EBFM) has emerged as an important paradigm in fisheries management, emphasizing the need for species distribution information. Selecting appropriate habitat models is crucial in species distribution studies. Bayesian models could reduce the reliance of species distribution on the data and are particularly suitable for small datasets in marine surveys. In this study, we constructed three Bayesian models to analyze the spatial distribution and shared suitable habitats of four Gobiidae (Myersina filifer, Chaemrichthys stigmatias, Amblychaeturichthys hexanema, and Amoya pflaumi) in Haizhou Bay, China. The interspecific associations of these species were also evaluated using Pearson correlation coefficient (r). Our analyses found that Bayesian regularized neural network (BRNN) model performed better than the other two Bayesian models, and the four Gobiidae species mainly coexisted in the central and southern coastal areas of Haizhou Bay, prey, sea bottom temperature and sediment were the main correlated factors on the habitat of Gobiidae. Furthermore, although the four species exhibited similar feeding habits, intense interspecific competition might not occur due to their considerable dietary breadth, with species associations reflecting the similarity of habitat preferences. For example, M. filifer and A. hexanema inhabited similar areas in spring, and their species association was also relatively high (0.64). This study will help to enhance our understanding of the habitat preferences and interspecific associations of Gobiidae, and provide a framework of spatial based fisheries management at multispecies level in marine bay ecosystems.