New publication presents a machine learning blueprint for monitoring lake water quality from space

Algal blooms fueled by nutrient pollution are a major threat to lake ecosystems, making timely and continuous monitoring essential for water management. While freely available satellite imagery offers a cost-effective way to track Chlorophyll-a (a key indicator of algal biomass), atmospheric interference and complex water optics make accurate readings difficult. Our new study evaluates various machine learning models and atmospheric correction tools using over two decades of satellite data (2000–2023) to optimize Chlorophyll-a retrieval in western Lake Ontario and Hamilton Harbour. We found that pairing the ACOLITE atmospheric correction processor with the XGBoost machine learning model consistently produced the most accurate results across multiple satellite platforms. Applying this optimal framework to map Hamilton Harbour revealed persistent midsummer algal hotspots near the shorelines, driven by high year-to-year variability rather than a simple long-term trend. Ultimately, this research provides a robust, highly scalable blueprint for managers to continuously monitor eutrophication and water quality in complex inland waters.

Link: https://doi.org/10.1109/TGRS.2026.3676135