Large‐Scale eDNA Sampling and Hierarchical Modelling Elucidates the Importance of Stream Habitat for Eastern Hellbender (Cryptobranchus a. alleganiensis) Occupancy and eDNA Detection

ABSTRACT Accurate detection data are imperative to assess distributions and habitat‐associations for species of conservation need. Environmental DNA (eDNA) sampling is an effective tool to obtain detection data across large geographic scales; however, most eDNA studies do not account for environmental variation that could influence detection. Hierarchical modelling can be used to identify factors important to species occurrence while accounting for such factors. Local extirpations and significant population declines have been documented across the range of the Eastern Hellbender ( Cryptobranchus alleganiensis alleganiensis ) due to water quality and habitat degradation, but a paucity of information on the current distribution and status of hellbenders remains for certain regions. We conducted a state‐wide eDNA survey to (1) investigate the current distribution of hellbenders in Kentucky, a state which lacks extensive hellbender occurrence information, (2) evaluate habitat associations for hellbenders in this region and (3) identify environmental factors that influence eDNA detection. Environmental DNA samples, water chemistry and habitat data were collected from 90 sites state‐wide, 27 of which had historic records. We ran multiscale Bayesian occupancy models to determine occupancy and detection probabilities at each site, and to identify water chemistry, local habitat and landscape factors associated with hellbender occupancy and eDNA detection. Hellbender eDNA was detected at 22 sites total, including 12 (44%) historic locations. We found that total organic carbon in the stream significantly hindered eDNA detection and that local habitat quality was more important for hellbender presence than water chemistry or upstream catchment land cover. Hellbender occupancy was positively associated with the percent cobble, gravel and bedrock in the streambed and stream order, and negatively associated with the percent fine sediment in the streambed. Our results indicate that hellbender populations have significantly declined in Kentucky, and the quality of available stream substrate is critical for hellbender presence. This study demonstrates that by applying hierarchical modelling to large‐scale eDNA sampling, we were able to make robust inferences about factors associated with hellbender occurrence and eDNA detection.