Remote sensing identification and model-based prediction of harmful algal blooms in inland waters: Current insights and future perspectives

Harmful algal blooms (HABs) in freshwater systems pose significant threats to water quality, ecological stability, and public health. Managing these blooms requires substantial resources, making early and accurate prediction essential. Remote sensing technologies have emerged as powerful tools for HAB identification and forecasting, providing critical data to support predictive modeling. However, forecasting HABs remains challenging due to inherent uncertainties in bloom dynamics. Recent advances in data science and computational methods have facilitated the widespread application of both data-driven (DD) and process-based (PB) models for HAB prediction. DD models, particularly machine learning techniques such as artificial neural networks (ANN), random forest (RF), and long short-term memory (LSTM), effectively capture relationships between environmental variables and bloom events from historical data, enabling accurate short-term predictions. In contrast, PB models simulate the biochemical processes driving algal growth, such as photosynthesis, nutrient uptake, and cell division, providing mechanistic insights and supporting targeted management strategies. Despite these advancements, challenges remain, including the selection of optimal input variables, model transferability across diverse water bodies, and the interpretability of complex machine learning models. Future research should focus on developing adaptive hybrid models, integrating interpretable artificial intelligence (XAI) techniques, and enhancing the synergy between remote sensing and predictive modeling. This comprehensive approach has the potential to provide robust early warning systems for HABs, contributing to sustainable freshwater management.