Modelling and predicting freshwater mussel distributions in the Midwestern USA

Abstract

AbstractHistorically diverse and abundant freshwater mussel communities in the Midwestern USA, as elsewhere in the world, are in substantial decline. An understanding of mussel distributions, habitat requirements and biodiversity is required to develop and optimize mussel conservation efforts.Maxent was used to generate species distribution models (SDMs) based on a wide range of environmental variables. Binary predictions were generated using two approaches. Species richness (SR) was estimated through model stacking and evaluated with an independent dataset derived from intensive sampling. Two measures of species endemism were estimated. Factors important for mussel occurrence were identified and geospatial analyses characterized spatial patterns of predicted SR and endemism at multiple spatial scales.SDMs generally performed at ‘fair’ or better levels (0.7 < AUC < 1). Stream size, climate and land use were important to mussel occurrence. SR estimates were positively and significantly correlated with observed SR in a separate dataset. SR and endemism hot and cold spots varied across spatial scales.This study introduces a novel background selection process for aquatic SDMs that considers sampling biases, species records relative to stream size and species' geographical ranges. The results provide valuable insight into species status and distributions, important and actionable factors influencing mussel occurrence, and spatial patterns of highly diverse or endemic regions. This study demonstrates how synthesizing SDMs can inform regional, multi‐species efforts for conserving freshwater mussel diversity.