Predicting Invasive Crayfish Distributions in the Southwest: Environmental Drivers and Model Transferability
The Lower Colorado River Basin of Arizona (LCRB) is invaluable to freshwater biodiversity, harboring several imperiled species. Over the last forty years, non-native virile crayfish (Faxonius virilis) have become established in waterbodies of the LCRB, outcompeting and preying upon native fishes. However, their potential distribution remains poorly understood. The objective of this study was to assess the predictive capability of species distributions models (SDMs) by addressing three questions: (1) What is the distribution of virile crayfish in the LCRB? (2) Are SDMs transferable in space? (3) How do proximal versus distal variables affect model performance? I conducted field surveys across 108 stream sites in the Little Colorado and Gila River subbasins of the LCRB in 2021 to assess site occupancy of virile crayfish, fit an ensemble of four SDMs, and project habitat suitability to all 52,639 streams in the two subbasins. I compared model performance and spatial projections for different (1) geographic subsets of sites and (2) environmental predictors to address my questions. The ensemble had high model performance (AUC = 0.844). Habitat suitability was greatest in perennial streams in high elevation forests. Percent volcanic rock and percent forest within watershed were the strongest predictors, whereas watershed area and hydrological category were relatively weak predictors. SDMs fit and projected within the same region performed better than those fit in one region and projected to another (AUC = 0.900 versus 0.703, respectively). When transferred, SDMs performed better with distal predictors than with proximal predictors (AUC = 0.804 versus 0.732, respectively).