Environmental DNA surveys can underestimate amphibian occupancy and overestimate detection probability: implications for practice

Abstract

Environmental DNA analysis of water samples has recently emerged as a powerful tool for studying the distribution, ecology, and conservation of many amphibian species. Validating efficacy through comparison with established methods of detection is important for any new method. We used multi‐method and single‐method occupancy models to explore the effectiveness of 3 techniques for detecting northern leopard frogs (Lithobates pipiens) and boreal chorus frogs (Pseudacris maculata). We conducted surveys using automated recording units, environmental DNA sampling, and visual encounter surveys in spring (30 Apr–3 Jun) and summer (23 Jul–27 Aug) of 2014 at 68 sites in southern Alberta, Canada. The multi‐method analysis incorporated data from all 3 survey methods while accounting for the lack of independence of detections within a sampling occasion, and we compared this to single‐method analyses to explore occupancy biases associated with each method and differences in method‐specific detection probability. Occupancy and detection probabilities estimated from environmental DNA analyses were biased for northern leopard frogs, overestimating probability of detection and underestimating occupancy. We could not assess bias for boreal chorus frogs because of overdispersion present in in the models. Although no single covariate explained this bias, it was ameliorated (albeit by reducing precision) by including a raw visual count of individuals as a proxy for abundance during the breeding but not the post‐metamorphic season. These results emphasize the need for careful consideration of temporal and spatial aspects of sampling design, conducting pilot studies, and external validation of eDNA‐only occupancy monitoring schemes prior to widespread implementation.