Advances in forecasting of harmful algal blooms in freshwater ecosystems

Aquatic scientists are being increasingly challenged to better predict harmful algal blooms (HABs), given the accelerating rate of global eutrophication. Forecasting of HABs has undergone substantial transformations, given the need for better management practices and rising concern over the negative impacts from HABs. This review presents a chronological examination of HAB forecasting methods, tracing the field's evolution from early statistical models to the current application of convolutional neural networks, deep learning, and multi-model integration. Initially dominated by conventional statistical approaches, the field of predicting HABs has shifted towards models better able to capture the complex, nonlinear dynamics of HABs. With growing public concern and environmental awareness, recent application of models emphasise the importance of understanding the biological and environmental conditions that lead to HAB formation in their specific environments. Emerging trends include the hybridisation of models, increased use of unsupervised deep learning, and the incorporation of individual-based models to account for ecological heterogeneity. As HAB forecasting continues to evolve in response to climate change and technological progress, the integration of advanced computational methods, real-time data, and stakeholder collaboration stands out as essential for building more accurate, efficient, and adaptive forecasting systems.