Ensemble Learning Model for El Niño Prediction and Global Warming Risk Assessment

Abstract

Climate change and the growing prevalence of extreme weather events have made El Niño prediction more precise and the evaluation of risks from global warming more dependable. With significant influence of El Niño on global temperature, rainfall distribution, droughts, floods and ecosystem stability, early prediction is crucial for effective climate adaptation and disaster preparedness. In this study, a novel Hybrid Ensemble Learning Model for El Niño Prediction and Global Warming Risk Assessment is presented by using state-of-the-art climate data preparation, feature engineering, adaptive feature selection, climate similarity clustering and an adaptive weighted stacking ensemble of Random Forest, Extra Trees, LightGBM and CatBoost. This framework converts raw oceanic and atmospheric observations into informative climate indicators that improve the learning capacity of the predictive models while removing redundant information. Multiple ensemble learners’ outputs are intelligently combined to improve prediction robustness, reduce over-fitting, and increase generalisation ability across different climate conditions. To improve the model transparency, SHAP-based explainable artificial intelligence is integrated to determine the contribution of the single climate variables towards the prediction outcomes, supporting the scientific interpretable decision-making. Furthermore, a Global Warming Risk Index is developed to assess future climate vulnerability and to provide useful information for environmental monitoring and early warning systems. The experimental evaluation reveals that the proposed hybrid framework outperforms the conventional machine learning approaches in terms of accuracy, precision, recall, F1-score, and error reduction, while it is robust under different climate scenarios. The proposed model offers a reliable, scalable, and interpretable solution for long-term climate forecasting, disaster risk reduction, environmental policy planning, and sustainable climate management, demonstrating the increasing potential of intelligent ensemble learning techniques to address complex global climate challenges.