Exploring the potential of well-upwelling for coral bleaching mitigation via integrated machine learning and particle swarm optimization algorithms

Coral reefs, critical marine ecosystems threatened by escalating sea surface temperatures, demand innovative solutions to mitigate thermal stress. This study introduces a novel Well-Upwelling (WU) system that strategically injects chilled seawater to form a three-dimensional cooling umbrella over coral habitats. Nozzle parameters and layouts are optimized by leveraging computational fluid dynamics (CFD), machine learning (ML), and a hybrid Particle Swarm Optimization (PSO-DPSO) framework. Categorical Boosting (CatBoost) was validated as superior for predicting jet dynamics, enabling efficient optimization of nozzle diameter (d₀), exit velocity (u₀), and discrete layouts. A case study near Nanshan Harbor, China—where severe heat stress occurred in 2024 (Degree Heating Weeks, DHW > 8)—revealed that the integrated PSO-DPSO algorithm improved cooling efficiency by 17.8–46.1 % compared to random layouts, identifying optimal parameters (u₀ = 0.20 m/s, d₀= 0.21 m) that balanced efficiency and thermal retention. Notably, the simulation-based results indicate that the optimized system reduced the DHW value during the study period from 1.83 to 0.45 in the coral region, corresponding to a 1.38±0.073℃ decrease in average temperature—below the threshold (DHW = 1) that induces visible coral stress. This work establishes a scalable, data-driven framework for hydrodynamic optimization, demonstrating the WU system’s potential to mitigate coral bleaching by effectively reducing thermal stress in dynamic marine environments.