Sampling designs for landscape-level eDNA monitoring programs using three-level occurrence models

Resource managers conduct landscape-level monitoring using environmental DNA (eDNA). These managers must contend with imperfect detection in samples and sub-samples (i.e., molecular analyses). This imperfect detection impacts their ability to both detect species and estimate occurrence. Although occurrence (synonymously occupancy) models can estimate these probabilities, most models and guidance for their application do not consider three levels. We studied this with three aims. First, we examined the number of samples required to detect a species at a site given imperfect detection. Second, we examined the ability of a three-level occurrence model to recover parameter estimates. Third, we examined the number of samples required to reliably recover parameter estimates. We found detecting eDNA in 1 sample at a site required 12 samples under most condition, but detection eDNA in situations that might be expected when looking for species at very low abundance required >50 samples. We found our occupancy model generally recovered known parameters unless detection and sample occurrence probabilities were <0.3. In these situations, >50 samples per site and 8 molecular replicates were required. Conversely, estimating and comparing occurrence and detection probabilities for species with moderate to high abundance may require 4 molecular replicates and 20-30 samples per site. More broadly, our findings illustrate the importance of study design, sample sizes, and molecular replicates for eDNA-based research, monitoring, and management.