Geospatial Interpolation of Remotely Sensed Observations in the Chesapeake Bay: an Ecological Forecasting Application

In dynamic coastal systems such as the Chesapeake Bay, limited coverage and frequency of in situ measurements often makes generalizability of an ecological forecasting system difficult. Satellite-derived environmental variables have the potential to address this problem, but satellite datasets suffer from incomplete coverage as well: atmospheric conditions—most notably cloud cover—lead to data gaps that significantly hinder the broad application of satellite-informed ecological predictions. In this study, the Chesapeake Bay estuary was used as a model "test bed" to which we applied the power of near real-time satellite-derived observations to the issue of monitoring and ecological forecasting of environmental Vibrio spp. bacterium. To use remote sensing in support of spatially complete estimates of Vibrio spp. in the Bay, we tested geospatial interpolation techniques as a method for filling gaps and minimizing errors in the satellite record. These interpolated values were then compared to direct satellite retrievals and to ChesROMS hydrodynamic modeled fields in order to assess the relative value of each method for generating inputs to empirical Vibrio spp. forecasting models. Results show that satellite-derived salinity and temperature can be interpolated with acceptable accurately in the Bay, with a MAE of 1.88ppt and 0.59°C. These errors differed systematically from ChesROMS errors both spatially and seasonally, indicating that the two techniques offer complementary information that can be applied to Vibrio spp. monitoring, and to other ecological monitoring systems, in complex estuaries like Chesapeake Bay.