Estimation of total phosphorus concentration using a water classification method in inland water

Total phosphorus (TP) is an important factor of eutrophication. The accurate monitoring of TP concentration is conducive to the management of water environments. To improve the accuracy of TP remote sensing inversion, this paper improved and studied the following three aspects. First, a new optical classification method was constructed to classify water into two different types. Compared with other classification methods, the proposed method can be more effective for improving the accuracy of TP inversion. Second, data regression analysis and fitting (DRF), back propagation neural network (BP) and random forest (RF) were used to build TP inversion models based on the different water type. The validation results indicated that the DRF performed best, with the highest precision for all water types, and was suitable for TP inversion. Finally, the performance of the classification and inversion algorithms on Sentinel-3 data was evaluated and one image was acquired for mapping the TP concentration in Taihu Lake. All in all, this study improves the accuracy of TP inversion by combining the new classification algorithm with DRF. This is significant for the development of class-based water quality parameter inversion algorithms for water color remote sensing, and this approach can be applied in the effective management and control of the eutrophication of lake environments.