Retrieving Eutrophic Water in Highly Urbanized Area Coupling UAV Multispectral Data and Machine Learning Algorithms

With the rapid development of urbanization and a population surge, drawback water pollution, especially eutrophication, poses severe threat to ecosystem as well human well-being. Timely monitoring variations quality is precedent preventing occurrence eutrophication. Traditional methods (station or satellite remote sensing), however, fail real-time obtain in an accurate economical way. In this study, unmanned aerial vehicle (UAV) with multispectral camera used acquire refined sensing data bodies. Meanwhile, situ measurement sampling in-lab testing are carried out observed values four parameters; subsequently, comprehensive trophic level index (TLI) calculated. Then three machine learning algorithms (i.e., Extreme Gradient Boosting (XGB), Random Forest (RF) Artificial Neural Network (ANN)) applied construct inversion model for estimation. The measured showed that status study area was mesotrophic light eutrophic, which consistent government’s water-control ambition. Among parameters, TN had highest correlation (r = 0.81, p 0.001) TLI, indicating variation TLI inextricably linked TN. performances models were satisfactory, among XGB considered optimal best accuracy validation metrics (R2 0.83, RMSE 0.52). spatial distribution map drawn by good agreement actual situation, manifesting applicability inversion. research helps guide effective timely warning