Evaluating observer bias and seasonal detection rates in amphibian pathogen eDNA collections by citizen scientists

We trained volunteers from conservation organizations to collect environmental DNA (eDNA) from 21 ponds with amphibian communities that had a history of Batrachochytrium dendrobatidis (Bd) and ranavirus (Rv) infections. Volunteers were given sampling kits to filter pond water and preserve eDNA on filter paper, as were the principal investigators (PIs), who made independent collections within 48 h of volunteer collections. Using multi-scale occupancy modeling, we found no evidence to suggest the observer who collected the water sample (volunteer or PI) influenced either the probability of capturing eDNA on a filter or the probability of detecting extracted eDNA in a quantitative PCR (qPCR) reaction. The cumulative detection probability of Bd eDNA at a pond decreased from May through July 2017 because there was a decrease in the probability of detecting eDNA in qPCR reactions. In contrast, cumulative detection probability increased from May to July for Rv due to a higher probability of capturing eDNA on filters later in the year. Our models estimate that both pathogens could be detected with 95% confidence in as few as 5 water samples taken in June or July tested with either 4 or 3 qPCR reactions, respectively. Our eDNA protocols appeared to detect pathogens with 95% confidence using considerably fewer samples than protocols which typically recommend sampling ≥30 individual animals. In addition, eDNA sampling could reduce some biosecurity concerns, jurisdictional and institutional permitting, and stress to biota at ponds.