Integrating species distribution and occupancy modeling to study hellbender (Cryptobranchus alleganiensis) occurrence based on eDNA surveys

Managers often rely on species surveys and distribution models to evaluate species occurrence and develop management and conservation plans. However, these tools are rarely used in concert. We used a three-step framework to evaluate the distribution of a declining and elusive freshwater amphibian species, the hellbender salamander (Cryptobranchus alleganiensis). We used the Maximum Entropy (MaxEnt) algorithm to develop a habitat suitability model to guide sampling, environmental DNA (eDNA) surveys to ground truth the habitat suitability model, and multi-level occupancy modeling to assess species presence, while accounting for eDNA detection errors. Our suitability model (AUC = 0.941, True Skill Statistic = 0.7, sensitivity and specificity = 0.86) identified the greatest amount of high and very highly suitable habitat in the Interior Plateau and Blue Ridge ecoregions of the study area. We used eDNA survey results (n = 284 sites) to evaluate model fit and detected the species at 65 sites. Detection probability (p) was 0.692 (95% CRI: 0.547, 0.818) at the site level and 0.674 (95% CRI: 0.621, 0.721) at the quantitative PCR level. Ecoregion was the primary covariate that explained occupancy, with greatest estimates in the Blue Ridge ecoregion 0.695 (95% CRI, 0.390, 0.925). Although the MaxEnt output was not significantly correlated with occupancy probability, we established a relationship between habitat quality and the number of eDNA detections. This study highlights the use of a multi-level framework to optimize sampling, assess model fit, account for imperfect detection, and evaluate the distribution of rare species that have limited occurrence data available