Advanced Journal of Environmental Science and Technology

ISSN 2756-3251

Advanced Journal of Environmental Science and Technology | Vol. 17 (9), pp. 336-350, September, 2026
DOI: 10.46882/2026.AJEST.170910

Research Article

Real-Time Harmful Algal Bloom Forecasting in Eutrophic Freshwater Reservoirs Using Sentinel-2 Multispectral Imagery and Explainable Machine Learning

Hannah Lindholm¹*, Arjun Nair², Valeria Gomez³

¹Department of Environmental Science, Stockholm University, Stockholm, Sweden
²Centre for Water Resources Development and Management, Kozhikode, India
³Department of Limnology, National Autonomous University of Mexico, Mexico City, Mexico

Corresponding Author: Hannah Lindholm — [email protected]

Abstract

Cyanobacterial harmful algal blooms (HABs) degrade drinking water supplies and produce hepatotoxins that threaten aquatic ecosystems. We trained an eXtreme Gradient Boosting (XGBoost) model combining 10 m Sentinel-2 red-edge spectral indices with meteorological and hydrological telemetry across four eutrophic reservoirs. The model predicted phycocyanin concentrations seven days in advance with a root-mean-square error of 4.8 μg/L and a classification accuracy of 92.6% for bloom threshold exceedances. Shapley additive explanations (SHAP) identified water column thermal stratification and antecedent phosphorus loading as the dominant drivers of rapid bloom onset.

Keywords: harmful algal blooms; Sentinel-2 remote sensing; explainable machine learning; water quality forecasting; eutrophication; cyanobacterial dynamics

Manuscript Timeline: Received 22 August, 2026; Revised 20 September, 2026; Accepted 29 September, 2026; Published 30 September, 2026.

Citation: Lindholm, H., Nair, A., and Gomez, V. (2026). Real-Time Harmful Algal Bloom Forecasting in Eutrophic Freshwater Reservoirs Using Sentinel-2 Multispectral Imagery and Explainable Machine Learning. Advanced Journal of Environmental Science and Technology, 17(9), pp. 336-350.