Decoding lake water eutrophication using an innovative dynamic model pool framework in Erhai Lake

Abstract

Rapid global urbanization has led to water eutrophication, threatening the stability of aquatic ecosystems stability. Chlorophyll-a (Chla), a key indicator of algal biomass, is a widely recognized as a metric for eutrophication. However, existing remote sensing retrieval methods face limitations in addressing complex environmental variations. This study developed an innovative Dynamic Model Pool (DMP) framework to optimize water quality prediction performance dynamically. Using Sentinel-2 satellite imagery and monthly in-situ Chla measurement data from Erhai located in Southwest China spanning 2018 to 2020, this study tested the effectiveness of the DMP framework. The results demonstrated that: (1) The DMP framework dynamically selected the optimal model based on data-specific characteristics. In 2018, the CBR model achieved the highest accuracy, while in 2019, GBR and XGBR were the most accurate. In 2020, GBR outperformed other models. (2) Spatiotemporal Chla distribution maps recorded consistently higher concentrations in the south part of lake, while the central part showed minimal level and variation. (3) Seasonal precipitation and temperature variations and policy implementation were key drivers of Chla concentration changes. Seasonal variations in precipitation and temperature collectively influenced the nutrient input and dilution dynamics in Erhai. Meanwhile, policy interventions implemented between 2018 and 2022, such as pollution interception and wastewater treatment, substantially decreased nutrient inflows during flood seasons and effectively limited nutrient accumulation.