Climate-driven projections of cyanobacterial harmful algal bloom expansion in coastal waters

Cyanobacterial harmful algal blooms (CyanoHABs) in coastal waters are a growing ecological and environmental concern, especially in climate-vulnerable regions. While many studies have explored historical variations and short-term forecasting of CyanoHABs, this study extends projections into the coming decades, focusing on Oman's vulnerable coastal areas under future climate change scenarios. By integrating General Circulation Models (GCMs) outputs with machine learning and deep learning models, this research aims to enhance predictive accuracy and assess long-term CyanoHAB impacts. Historical satellite data (2003-2024) show elevated chlorophyll-a (Chl-a) levels along Oman's eastern coast, particularly near Muḥūt, Ad Duqm, and Al Jazīr. These areas are susceptible to bloom formation due to nutrient-rich coastal waters. MODIS satellite imagery and the Floating Algae Index (FAI) were used to estimate CyanoHAB areas, with Landsat 7 data providing validation (R² = 0.925-0.968). Analysis of 957 satellite images from 2000 to 2020 revealed an increase in CyanoHAB occurrences, especially in summer. Principal Component Analysis identified wind speed, temperature, and precipitation as key environmental drivers. To project future CyanoHAB areas, ML and DL models, including Random Forest, Extreme Gradient Boosting, Gated Recurrent Unit, and Long Short-Term Memory (LSTM), were fine-tuned using Particle Swarm Optimization. LSTM outperformed other models, and Sobol analysis revealed precipitation as the most sensitive parameter. Projections were made using three high-performing GCMs under SSP126, SSP245, and SSP585 scenarios for three periods (2030-2089). Results indicated a general increase in CyanoHAB areas, with the highest increase under high emission scenarios, requiring adaptive management and mitigation.