Urban river water quality monitoring based on self-optimizing machine learning method using multi-source remote sensing data

Urban rivers are complex ecosystems that directly determine the living environment of human beings. Monitoring the urban river water quality indexes is a challenge in water quality evaluation. The purpose of this study was to propose a multi-source remote sensing water quality inversion method based on a small number of samples to solve the problem of scale inconsistency among multi-source remote sensing data, so as to achieve large-scale and efficient inversion of urban river water quality. Since there is a very important problem that the complex nonlinear relationships must be solved between simple ground point data and remote sensing data in water quality inversion, a novel self-optimizing machine learning monitoring method is proposed, which can automatically find the optimal parameters of the model from a small number of samples, and reduce the training time. Meanwhile, in order to strengthen the correlation between water quality parameters and remote sensing data, the feature enhancement method was used for generating the input data. Moreover, to solve the problem of the multi-source data quantity and quality, the spatial mapping method was used to achieve consistency in the water quality information since these data have different nonlinear characteristics. The experimental results show that for unmanned aerial vehicle (UAV) images, the R2 of chlorophyll a (Chla), turbidity (TUB), and ammonia nitrogen (NH3-N) can reached 0.917, 0.877 and 0.846, respectively. Using a satellite image, the R2 of Chla, TUB, and NH3-N can reach 0.827, 0.679 and 0.779, respectively. This method provides a new way to realize the integration of air-space-ground monitoring of urban inland rivers in the future.