Global patterns and predictors of trophic position, body size and jaw size in fishes

Abstract

Aim: The aim of this study was test whether maximum body mass and jaw length are reliable predictors of trophic position (TP) in fishes, and to compare linear and nonlinear machine-learning (ML) models incorporating biogeography, habitat and other morphological traits. Location: Global. Time period: Modern. Major taxa studied: Fishes. Methods: We compiled a global database of TP (2.0–4.5), maximum body mass, jaw length, order, ecoregion, habitat and other morphological traits of freshwater, estuarine and diadromous fishes (n = 1,991). We used Bayesian linear mixed effects and ML, with r2 analogues and 10-fold cross-validation, to explain and predict TP. Results: Random forest models outperformed Bayesian models in all comparisons. Jaw length was the most influential predictor of TP, but was weakly associated with body mass except in five orders of largely piscivorous fishes. Trophic position did not scale positively with body mass in global ecoregions, riverine fishes, or in 29/30 orders, but scaled positively in lacustrine fishes and Perciformes. Significant negative TP–body mass scaling was observed in Characiformes. Best models explained 55% of the global variation in TP, but over-estimated the position of herbivores-detritivores, and under-estimated the position of top predators. Main conclusions: Our study provides support for jaw length as an important mechanism constraining TP in one of the world’s largest groups of vertebrates. Jaw length and body mass were weakly correlated, and therefore body size was not a strong predictor of TP. The diversification of large-bodied herbivores-detritivores and omnivores in freshwater ecosystems, coupled with small predators in species-rich orders (e.g., Cypriniformes, Characiformes) in temperate and tropical rivers explains why TP globally shows a weak relationship with body size. Our model validation results underscore the importance of not assuming that explanatory power extends to predictive capacity in macroecology and machine-learning models.