Spatiotemporal variation and key environmental drivers of Chlorophyll-a in Qionghai Lake: insights from Sentinel-2 satellite and machine learning models

As eutrophication in freshwater lakes continues to intensify, effective water quality monitoring is essential for sustaining aquatic ecosystems and ensuring resource security. Chlorophyll-a (Chl-a) is a widely recognized proxy for algal biomass and eutrophication status. This study integrates Sentinel-2 satellite imagery with in-situ measurements to dynamically monitor Chl-a concentrations in Qionghai Lake, China. Spearman correlation analysis was employed to identify sensitive spectral bands, and three models—Random Forest (RF), Back Propagation (BP) neural network, and multiple linear regression—were constructed for Chl-a retrieval. The RF model demonstrated superior performance, with a mean absolute percentage error of 25.52%, R2 of 0.81, and RMSE of 1.47 mg/m3. Utilizing this model, Chl-a distribution from 2016 to 2024 and across the 2023 seasons was successfully mapped. Higher Chl-a concentrations were consistently observed in the northwestern region of the lake, exhibiting clear seasonal fluctuations. Water quality improved during periods of reduced anthropogenic disturbance, notably during the government-led ecological restoration in 2017 and the COVID-19 pandemic. Key environmental drivers of Chl-a and algal cell density included total phosphorus, the TN/TP ratio, and suspended solids. The findings underscore the potential of remote sensing and machine learning for accurate, long-term eutrophication assessments in inland lakes. It offers insights into spatial-temporal algal dynamics and supports targeted lake management and restoration.