Evaluation of methods for identification of early detection monitoring sites based on habitat suitability for invasive European green crab in the Salish Sea, British Columbia

Abstract

European Green Crab (EGC; Carcinus maenas) was first detected in Sooke Basin in 2012, and in US waters of the Salish Sea in 2016. EGC is not yet known to be established in Canadian waters of the Salish Sea outside of Sooke, presenting an opportunity for early detection and management of newly invaded areas. AIS managers require information on where to target early detection monitoring programs. As such five existing models were evaluated based on habitat suitability using 447 sites in the Salish Sea. Each of the five individual models was informative at identifying sites suitable for EGC in Canadian waters of the Salish Sea but there was generally low agreement among models, likely because each model incorporates different aspects of EGC biology and habitat use. Without an independent validation dataset it was not possible to identify a single “best” model to identify early detection sites. An ensemble model approach can buffer uncertainty arising from individual models by combining the outputs of multiple individual models. To incorporate all of the models in an ensemble, the outputs for each model were rank-transformed into 20th percentiles (i.e., rescaled values 1-5), and the ‘mode’ (most frequent) value across all five models was determined for each site (n = 447). The 68 sites with a rank-transformed mode of 5 were prioritized for early detection monitoring (see Figure 1). While the 68 sites identified can serve as a starting point, there are other considerations beyond habitat suitability when selecting specific monitoring sites for EGC in the Salish Sea that were beyond the scope of this process. Managers may choose to add or remove sites as needed based on likelihood of arrival (e.g., through larval drift, human-mediated movements, immigration); presence of other important features (e.g., eelgrass); presence of available prey/absence of predators; ecologically, economically, or culturally important areas; site access; or partner interest. The ensemble model approach developed here may be applied elsewhere by using or deriving models specific for that region, based on the available data and management objectives. However, the input data does not capture all factors, including propagule pressure, that may contribute to invasion success.