Seasonal weather pattern prediction from enso indices using machine learning

Seasonal climate prediction in Bangladesh remains challenging due to the nonlinear nature of weather and climate interactions. This study investigates the correlation between nine El Niño–Southern Oscillation (ENSO) indices and seasonal temperature and rainfall patterns across Bangladesh, using monthly data from 29 meteorological stations (1977–2022). Six supervised machine-learning models, such as Random Forest (RF), XGBoost (XGB), Decision Tree (DT), Linear Regression (LR), K-Nearest Neighbors (KNN), and K-Fold Cross-Validation (KFCV) were evaluated using R2, MAE, and RMSE. XGB achieved the highest accuracy for temperature prediction (R2 = 0.8824 for Tmax, 0.9706 for Tmin, and 0.9559 for Tavg), with RF and KFCV performing comparably. Rainfall prediction accuracy was lower, with RF achieving the highest R2 (0.6273). Overall, the results confirm that multiple ENSO indices significantly influence Bangladesh’s seasonal climate and that advanced ML models, particularly XGB and RF, offer strong potential for improved prediction.