Mapping habitat suitability of invasive crayfish in aridland riverscapes: Virile crayfish (Faxonius virilis) in the Lower Colorado River Basin, USA.

Correlative species distribution modeling (SDM) is an important tool to predict distributions of invasive species. A unique challenge of implementing SDMs in aridland stream networks is identifying which streams are perennial and which of those offer suitable habitat for obligate aquatic species. Here, we map perennial streams and habitat suitability of invasive virile crayfish (Faxonius virilis) in the Lower Colorado River Basin (LCRB) in the southwestern USA, and quantify drivers of uncertainty in these projections. First, we surveyed F. virilis at 122 stream sites in May-July of 2021, 2022, and 2023. We then implemented an ensemble of SDMs fit using combinations of randomly drawn sites, categories of environmental covariates measured at different spatial scales, survey years, and statistical algorithms. Next, we used open-source street view images from May-July of 2022 and 2023 to assess dry-wet status at 326 road-stream intersections and the same ensemble framework to map perennial stream reaches. Lastly, we quantified drivers of variation in predictive accuracy and mapped habitat suitability across replicates. Median true skill statistic (TSS) across F. virilis replicates was 0.613 and habitat suitability was highest in mountain ecoregions and lowest in desert ecoregions. Of the 130,847 stream km in the LCRB, we estimate 29,078 km (22.2 %) have wetted channels during the May-June dry season and median 50.3 % of these perennial streams provide suitable habitat for F. virilis. The statistical algorithm was the strongest driver of TSS variation across replicates, whereas the spatial scale at which covariates were measured was the strongest driver of variation in mapped habitat suitability across replicates. We confirm the widespread invasion of F. virilis throughout the LCRB, particularly in perennial mountain streams. The modest predictive accuracy highlights the generalistic niche of F. virilis. Overall, we demonstrate that spatiotemporally comprehensive datasets combined with ensemble modeling can guide management at regional extents.