Site- and landscape-scale estimates of ecological suitability inform desert fish conservation

Abstract Objective Persistence of freshwater fishes native to dryland streams increasingly depends on ecologically informed conservation as environmental and anthropogenic pressures increase. Imperiled Rio Grande Chub Gila pandora and Rio Grande Sucker Pantosteus plebeius require management to slow their declining ­distribution and abundance, but insights into their ecological requirements have been restricted to site-scale inference over limited areas. To expand this understanding and broadly inform conservation efforts, we conducted two assessments of ecological suitability for each species using data collected across their overlapping ranges. The goal of our first assessment was to identify site-scale abiotic and biotic conditions associated with occupancy. Our second assessment sought to explore the importance of landscape-scale covariates to presence and predict probability of presence in unsampled streams. Inference gained from these approaches independently and together enabled us to understand factors that affect the distribution, ecology, and habitat needs of these species, which will improve conservation actions. Methods Fish community surveys, site-level habitat assessments, and remotely sensed landscape scale covariates were measured at 112 sites across the range of Rio Grande Chub and Rio Grande Sucker in the United States. To better understand site-scale physiochemical conditions and co-occurrences of White Sucker Catostomus commersonii and nonnative trout that are associated with occupancy, we used data from fish community and habitat surveys to fit multispecies occupancy models. We then used presence/absence information from a larger data set and a suite of remotely sensed covariates to develop random forest classification models that identified important landscape-scale predictors of presence. We used these models to predict the probability of presence for each Rio Grande Chub and Rio Grande Sucker in unsampled streams, and from those predictions we inferred distribution patterns of ecologically suitable streams across our study area. Results Site-scale ecological suitability assessments identified positive relationships between water temperature and occupancy of both Rio Grande Chub and Rio Grande Sucker. Rio Grande Chub occupancy was also positively associated with the quantity and depth of pools, and Rio Grande Sucker occupancy was negatively related to quantity of riffles. The probability of Rio Grande Chub occupancy was modestly reduced by nonnative trout occupancy, and Rio Grande Sucker occupancy was not associated with any nonnative species. Influenced primarily by modeled mean August stream temperature (NorWeST) and watershed position, defined as upstream drainage area and Strahler order, predictions from random forest models demonstrated higher probabilities of presence in warm, small- to medium-sized streams. While this produced higher proportions of ecologically suitable reaches toward the southern extent of their range in the United States, the greatest linear stream quantity predicted to support presence of both species occurred in north-central New Mexico. Conclusions Our first assessment, which was directed at understanding ecological suitability at the site scale, successfully described abiotic and biotic conditions associated with occupancy by Rio Grande Chub and Rio Grande Sucker, which is useful to inform habitat restoration, nonnative species management, and the selection of streams and reach conditions that are suitable for repatriation. Our second assessment revealed important landscape-scale predictors of presence and patterns of predicted suitable habitat, information that is required to prioritize conservation regions and select streams for future surveys and repatriation efforts. Together, inference from both assessments more broadly described the ecological needs of Rio Grande Chub and Rio Grande Sucker and will allow managers to direct conservation actions at multiple scales.