Utility of Occupancy Models with Environmental DNA (eDNA) fromOlympic Coast National Marine Sanctuary

Adjacent to the spectacular, rugged coastline bordering the Olympic Peninsula of Washington State, Olympic Coast National Marine Sanctuary (OCNMS) encompasses 8257 km2 of coastal waters that support one of North Americas most productive marine ecosystems. These rich waters support important recreational, commercial and subsistence fisheries for Washington state and four sovereign tribal governments: the Quinault Indian Nation and the Makah, Quileute and Hoh Tribes. Given its ecological, cultural and economic significance, managers from OCNMS are tasked with monitoring and management towards conserving the areas ecological integrity. The development of metabarcoding with environmental DNA (eDNA) as an effective freshwater monitoring tool has potential applications in marine systems, but the statistical analysis and the interpretation of those eDNA results are still under development. One promising strategy to analyze eDNA is through the use of occupancy models. Occupancy models enable us to calculate probabilities of detection from each taxon sequenced and are designed to work with presence-absence data. To inform our long term monitoring design of this coastal ecosystem, we collected 44 eDNA samples at nine of OCNMS long term mooring sites in 2019 and tested four different molecular markers for species reported and taxonomic richness. Additionally, we assessed occupancy models for use with eDNA to estimate the number of samples required to accurately predict the presence of a species. Occupancy models show great promise for use in eDNA studies provided there is sufficient replication; additionally, the choice of molecular marker strongly influences a taxas probability of detection. Author SummaryWe show the utility of using occupancy models with eDNA in Olympic Coast National Marine Sanctuary for monitoring near-shore marine biological communities. Based on probabilities of detection, long term coastal monitoring projects using eDNA would benefit from designs prioritizing sample number (10-20 samples per site) over sites sampled. With proper study design, occupancy models provide a statistical framework for comparisons between sites and over time. By accounting for simple non-detection vs true absence, occupancy models help to eliminate noise from eDNA studies, increasing detection of true shifts in community composition. We also demonstrate that marker choice is an important study design consideration not only for the types of taxa recovered, but also for consistency between samples and timepoints.