Retrieval of Turbidity and TDS of Deepor Beel Lake from Landsat 8 OLI Data by Regression and Artificial Neural Network

Landsat 8 Operational Land Imager (OLI) surface reflectance data corrected for sun angle was used to construct regression and artificial neural network (ANN) models of three parameters of pH, turbidity, and total dissolved solids (TDS) in Deepor Beel, Guwahati, India, where the water quality of the lake had shown a decline in recent past. The dates of collection of the field data and the satellite overpass were very near so as to preserve the temporal association between the two datasets. In the process, data collected from forty-five points were used as a training dataset to build the models. Another twenty points in the lake were used for testing the models. Regression models for turbidity and TDS showed a good R 2 fit while the same for pH behaved poorly. The results were also checked for mean absolute percentage error (MAPE) and root-mean-square error (RMSE) for the predicted dataset and found to exhibit errors within an acceptable range. Furthermore, ANN models showed better performance than the regression models. Thus, it was concluded that Landsat 8 OLI was useful in retrieving turbidity and TDS in the lake’s condition. The method had the scope to be used for continuous monitoring of the lake’s waters in the near future.