Remote Sensing Regression Models to Estimate Water Quality Indicators in Continental Waters in North-East Argentina

To ensure water preservation is relevant to constantly monitor water quality. To facilitate this task, remote sensing techniques are applied combining water property measurements and spectral information. Two empirical regression models were developed to estimate chlorophyll-a concentration and turbidity from a water body located in North-East Ar-gentina, in the Metropolitan Area of Gran Resistencia, Chaco Province. The spectral platform used was Sentinel-2 MultiSpec-tral Instrument, and the physicochemical characterization was achieved by multiple field campaigns where water samples were collected. In the training step several candidate models were compared. Performance metrics (coefficient of determination, <tex>$\mathrm{R}^{2}$</tex>, and root mean squared error, RMSE) were calculated and used for the final models selection. The obtained algorithms were linear combinations of spectral bands. Chlorophyll-a algorithm validated performance metrics were <tex>$\mathrm{R}^{2}=0.819$</tex> and RMSE = 13.71 mg/m³. For the turbidity model, <tex>$\mathbf{R}^{2}=0.968$</tex> and RMSE = 1.62 NTU. Correlation between physicochemical parameters were asserted and maps were created applying the validated models. A correlation between chlorophyll-a and turbidity was observed. Due this, the spatial distribution in the maps of both parameters followed the same trend.