Abstract
Water quality monitoring is essential to assess and manage anthropogenic eutrophication, especially for coastal estuaries in heavily populated areas. Current monitoring techniques rely on in-situ sampling, which can be expensive and limited in spatial and temporal coverage. Satellite remote sensing, using platforms such as Landsat-8 (Operational Land Imager, OLI), Sentinel-2 (Multi-Spectral Imager, MSI), and Sentinel-3 (Ocean Land Color Imager), has the potential to provide more extensive coverage than traditional methods. Coastal waters are optically more complex and often shallower and more enclosed than the open ocean, presenting conditions that pose challenges to remote sensing approaches. Here we compared in-situ data from 16 stations around Pleasant Bay, Massachusetts from the years 2013-2021 to contemporaneous observations with the sensor onboard Landsat-8. Initial evaluations identified a subset of stations that were not suitable for satellite remote sensing, due to depth and proximity to land. Satellite-derived estimates of chlorophyll-a and Secchi depth were acquired using the “Case-2 Regional/Coast Color” (C2RCC) atmospheric corrections and for retrieval of water constituents. Based on our observations, Landsat-8 OLI provided the best performance when comparing satellite-derived estimates of chlorophyll concentrations(coefficient of determination (r²) =0.612, Root mean squared error (RMSE) =4.07 mg m³) and Secchi depth (r²=0.132, RMSE=2.43) to corresponding in-situ data. We also evaluated a machine-learning random-forest approach for satellite retrieval of water constituents using Landsat reflectances as input variables and comparing to in-situ data for chlorophyll-a, Secchi depth, and dissolved oxygen (DO). The Landsat-8-derivedresults indicate that predictions of water quality indices from both C2RCC and random-forest machine-learning techniques can be a useful addition to existing water quality monitoring efforts, potentially expanding both spatial and temporal coverage of monitoring efforts.