High-Resolution Water Quality Monitoring of Small Reservoirs Using UAV-Based Multispectral Imaging

Multispectral satellite imagery has been widely applied in water quality monitoring, but limitations in spatial–temporal resolution and acquisition delays often hinder accurate assessments in small water bodies. In this study, a DJI M600PRO UAV equipped with a Sequoia multispectral sensor was used to assess the water quality in Zhangshan Reservoir, a small inland reservoir in Chuzhou, Anhui, China. Two regression approaches—the Window Averaging Method (WAM) and the Matching Pixel-by-Pixel Method (MPP)—were used to link UAV-derived spectral indices with in situ measurements of total nitrogen (TN), total phosphorus (TP), and chemical oxygen demand (COD). Despite a limited sample size (n = 60) and single-day sampling, MPP outperformed WAM, achieving higher predictive accuracy (R2 = 0.970 for TN, 0.902 for TP, and 0.695 for COD). The findings demonstrate that UAV-based MPP effectively captures fine-scale spatial heterogeneity and offers a promising solution for monitoring water quality in small and turbid reservoirs, overcoming key limitations of satellite-based remote sensing. However, the study is constrained by the temporal coverage and sample density, and future work should integrate multi-temporal UAV observations and expand the dataset to improve the model robustness and generalizability.