Evaluating the effectiveness of joint species distribution modeling for riverine fish communities

Species Distribution Models (JSDMs) allow for accounting for both environmental predictors and species dependencies within a single statistical modeling frameworks. However, it is unclear under what conditions incorporating residual species dependencies (conditional predictions) might outperform predictions made using only environmental information (marginal predictions). Here, we evaluated how a JSDM approach could improve predictive strength for stream fishes in the Chesapeake Bay Watershed (CBW), USA, using a cross validation study fitted to electrofishing occurrence data for 54 species.