Estimation of the Key Water Quality Parameters in the Surface Water, Middle of Northeast China, Based on Gaussian Process Regression

Abstract

To estimate the key water quality parameters on a large scale, based on Pearson’s correlation analysis and band ratio, this study first obtains multiple sensitive band combinations (R ≥ 0.30, p < 0.01) for three key water quality parameters: dissolved oxygen (DO), total nitrogen (TN), and total phosphorus (TP). Then, principal component analysis is used to reduce the dimensions and analyze multiple optimal combinations, and the first three principal components (PCs) of the optimal combinations are selected to analyze the water quality parameters. Finally, the water quality parameter models of DO, TN, and TP are proposed and compared based on spectral analysis and field measured water quality data respectively using Gaussian process regression and PCs for each parameter. Through model verification and by comparing the performance of the three models, it is found that the TP model performed well (R = 0.9824, p < 0.01), and TP grade accuracy rate is up to 94.97%. Through the error analysis of TN and DO, it is found that 93.0% of error samples occurs when TP < 0.1 mg/L in the water quality. These results would provide a scientific basis for water quality monitoring and water environment management in the study area and could also be used as a reference for water quality monitoring in other basins.