Offshore snapper and shark distributions are predicted by prey and area of nearby estuarine environments in the Gulf of Mexico, USA

Abstract

Seascape ecology has demonstrated that marine fishes are associated with multiscale habitat characteristics; however, most species distribution models focus on only a few predictors (e.g. depth, temperature), and this limits knowledge of essential fish habitat characteristics. Our objectives were to (1) determine habitat associations of offshore predatory marine fishes using a comprehensive suite of predictors, including area of nearby estuarine environments, (2) assess variable influence, and (3) model the spatial distribution of selected fishes in the families Carcharhinidae and Lutjanidae. We hypothesized that the concept of coastal outwelling would be evidenced by species associations with areas of nearby estuarine environments, and prey abundance would correlate with predator distributions. Species distribution models were developed for 2 snapper and 3 shark species in the northern Gulf of Mexico, USA. We used 34 multiscale predictors to evaluate how fish probability of presence or catch per unit effort (CPUE) were associated with oceanography, geography, substrate, area of nearby wetlands and estuaries, and prey abundance. Boosted regression trees, a machine-learning technique, modeled the most influential variables and predicted distributions. Model validation showed an overall accuracy of 79-86%, and CPUE models explained >40% of model deviance. Oceanographic variables, particularly mixed layer depth, were most influential and most frequently selected. As hypothesized, predatory fish distributions were predicted by prey abundances, and shark distributions were predicted by area of nearby coastal wetlands and estuaries. Our findings suggest that spatial models can provide novel insights into prey associations and linkages of marine species with nearby wetlands and estuaries.