Coral Reef Bleaching Prediction: A Machine Learning Approach Using Environmental Factors

Coral reefs are important ecosystems that support marine biodiversity and human livelihoods. But they are increasingly under threat from climate change, with coral bleaching events becoming more extreme and common. Coral bleaching happens when corals eject their symbiotic algae in response to stress such as increased sea surface temperatures. This research utilizes machine learning to predict coral bleaching based on parameterization of environmental conditions, including Temperature Stress Anomaly (TSA), ClimSST, turbidity, and cyclone rates. We build predictive models using XGBoost, Random Forest, and Long Short-Term Memory (LSTM) networks, and we do that using a dataset that covers multiple regions and multiple timescales. SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) are used to interpret the model. The LIME analysis constructs TSA and ClimSST as the strongest predictors of bleaching events. The R-squared of the XGBoost model is 0.85, the highest among regression models in predictive accuracy. Training on data from the past to members of the deep sea coral community provides an incredible predictive knowledge base, pioneering machine learning for ecological forecasting with actionable insights to minimize coral reef decline. This research provides a comprehensive framework that combines sophisticated algorithms and interpretability methods to accurately predict coral bleaching events and gain insights into the complex interplay of environmental influences on coral health, with the potential for significant implications for coral conservation and management in a changing climate.