Evaluating the use of macroscale variables as proxies for local aquatic variables and to model stream fish distributions

Abstract

The geographical ranges of species are influenced by three components: spatial distribution of environmental conditions, biotic interactions and the dispersal capacity of species. The scarcity of distributional records in vast regions such as the Amazon impedes understanding of fish distribution. Predictive distribution models have emerged as a better alternative to surpass this problem, but the absence of large-scale maps for aquatic variables has been suggested as an important limitation. We aimed to evaluate the use of macroclimatic variables as surrogates for local limnological variables in the Brazilian Amazon. Ordinary least squares model were used to predict the local habitat variables from climatic and geomorphological information as macroscale variables. Models for six stream-dwelling fish were built in MaxEnt and validated using area under curve and true skill statistics (TSS). All local variables were predicted successfully (R2 > 0.39), and MaxEnt models had good suitability using the macroscale variables (TSS higher than 0.70). We conclude that macroscale variables can be effective surrogates for local habitat variables, at least for large-scale analyses on poorly sampled regions such as the Brazilian Amazon. © 2014 John Wiley & Sons Ltd.