Aquatic invasive species exhibit contrasting seasonal detectability patterns based on environmental DNA: Implications for monitoring

Abstract

Abstract Aquatic invasive species (AIS) are a global threat to freshwater biodiversity and ecosystem services. Documenting AIS prevalence at broad spatial scales is critical to effective management and early detection. However, conventional monitoring for AIS is costly and is rarely applied at the resolution and scale required for effective management. Monitoring of AIS using environmental DNA (eDNA) has the potential to enable broadscale surveillance at a fraction of the cost of conventional methods, but key questions must first be addressed related to how eDNA detection probability varies among environments, seasons, and multiple species with different life histories. To quantify spatiotemporal variation in the detection probability of AIS using eDNA sampling, we surveyed 20 lakes with known populations of four aquatic invasive species: common carp (Cyprinus carpio), rusty crayfish (Faxonius rusticus), spiny waterflea (Bythotrephes longimanus), and zebra mussels (Dreissena polymorpha). We collected water samples at 10 locations per lake, five times throughout the open water season resulting in a total of 1,000 water samples. Quantitative polymerase chain reaction was used with species‐specific assays to determine the presence of each species' eDNA in water samples. With Bayesian occupancy models, we quantified the effects of lake and site characteristics and Julian date on eDNA detection probability. The probability of eDNA detection varied seasonally, and the seasonal variation was species‐specific and related to species life histories. Zebra mussel eDNA was generally the most detectable among the species we targeted, and detection probability peaked in midsummer when only six water samples were required to achieve a 95% probability of detection (80% Bayesian credible interval: 3–12 samples). Spiny waterflea eDNA detections also peaked in mid to late summer, but were overall the most difficult species to detect, requiring 160 samples for a 95% probability of detection (80% Bayesian credible interval: 67–1,616 samples). Common carp eDNA was most detectable in the spring and rusty crayfish eDNA was most detectable in the early autumn, corresponding to key life history events. Sampling for eDNA during the optimal time of the year for each species decreased the number of samples required to reach a 95% probability of detection by an order of magnitude or more. Our results are relevant for decision makers interested in using eDNA as a multi‐species monitoring tool and highlight the importance of life history in the efficacy of eDNA monitoring.