Remote sensing inversion of water quality parameters in urban complex river networks based on the CPO-SVM algorithm

To address the issue of insufficient accuracy in retrieving water quality parameters of urban complex river networks using multispectral remote sensing technology, an inversion model based on the Crested Porcupine Optimizer-Support Vector Machine (CPO-SVM) is proposed. The model utilizes the CPO algorithm to automatically optimize the penalty factor and kernel parameters of the SVM, thereby addressing the difficulties in parameter selection and the slow convergence speed of SVM. Based on Gaofen-1 (GF-1) remote sensing images and in-situ water quality parameter data, the model establishes inversion models for ammonia nitrogen (NH₃-N), dissolved oxygen (DO), total phosphorus (TP), and permanganate index (CODMn). Experimental results demonstrate that the CPO-SVM model outperforms both statistical regression models and the SVM model in terms of the coefficient of determination (R²), root mean square error (RMSE), and mean absolute percentage error (MAPE). This indicates that the proposed model achieves higher inversion accuracy and better practicality, offering a novel approach for remote sensing monitoring of water bodies in urban complex river networks.