A Geographically-Weighted Regression Model to Predict Oyster Density Hotspots in the Chesapeake Bay Region

Native oysters in Chesapeake Bay are nearing extinction due to a combination of disease, over-fishing, habitat loss, and declines in water quality.  The ability to predict hotspots in oyster densities and the drivers of these spatial patterns would be valuable to resource managers.  We used a novel spatially-explicit model called geographically-weighted regression (GWR) to relate known oyster densities with key habitat characteristics and to predict where oysters are most likely to occur in relatively high densities for two regions of Chesapeake Bay.  Live oyster density from a recent patent tong survey was used as the dependent variable, and bathymetry, water quality, and available oyster habitat were used as the independent variables.  GWR is similar to weighted least squares regression; however GWR applies a set of weights to the parameter estimates which depends upon the location of a sample point relative to other sample points in the data set. The results of this model predict oyster densities with a high degree of spatial resolution.  Bathymetry and the proportion of bottom that is cultch were found to be the best predictors of oyster density. This work provides insight into a new spatial statistical modeling approach for natural resource management.