| Coastal and inland waters represent a diverse set of resources that support natural habitat and provide numerous ecosystem services to the human population. Conventional techniques to monitor water quality using in situ sensors and laboratory analysis of water samples can be very time- and cost-intensive. Alternatively, remote sensing techniques offer better spatial coverage and temporal resolution to accurately characterize the dynamic and unique water quality parameters. However, bio and geo-optical models are required that relate the remotely sensed spectral data with color producing agents (CPAs) that define the water quality. These CPAs include chlorophyll-a, suspended sediments, and colored-dissolved organic matter. Developing these models may be challenging for coastal environments such as Long Bay, South Carolina, due to the presence of multiple optically interfering CPAs. In this work, a regionally tiered ocean color model was developed using band ratio techniques to specifically predict the variability of chlorophyll-a concentrations in the turbid Long Bay waters. This model produced higher accuracy results (r-squared = 0.62; RMSE = 0.87 micrograms per liter) compared to the existing models, which gave a highest r-squared value of 0.58 and RMSE = 0.99 micrograms per liter. To further enhance the retrievals of chlorophyll-a in these optically complex waters, a novel multivariate-based approach was developed using current generation hyperspectral data. This approach uses a partial least-squares regression model to identify wavelengths that are more sensitive to chlorophyll-a relative to other associated CPAs. This model was able to explain 80% of the observed chlorophyll-a variability in Long Bay with RMSE = 2.03 micrograms per liter. This approach capitalizes on the spectral advantage gained from hyperspectral sensors, thus providing a more robust predicting model. This enhanced mode of water quality monitoring in marine environments will provide insight to point-sources and problem areas that may contribute to a decline in water quality. Moreover, remote sensing applications such as this can be used as a tool for coastal and fisheries managers with regard to recreation, regulation, economic and public health purposes. |
