Determining the most suitable Sentinel-2 indices for turbidity and chlorophyll-a concentration for an oligotrophic to mesotrophic reservoir in Brazil

Water quality is a major issue for agencies responsible for the management and maintenance of reservoirs. In this article, Sentinel-2 (S-2) data is tested to produce reliable estimates of turbidity and chlorophyll-a concentrations in an oligotrophic to mesotrophic reservoir in Brazil. The spectral bands were all tested individually and jointly to determine the best and most stable features. Many indices are also tested. In situ data was acquired over a period of three years with synchronous S-2 data. The models were evaluated on their robustness and stability using a bundle of in situ dataset (all field campaigns). The results show that no model can be directly applied to other reservoirs or different dates without calibration. Conversely, calibrating the S-2 indices with two in situ measurements greatly reduces the errors. Average root mean square errors of 0.518 FNU and 1.1 mu g/l were obtained for turbidity and chlorophyll-a concentrations respectively.