Improving rainfall forecasting using deep learning data fusing model approach for observed and climate change data.

Area of Science:

• Environmental Science
• Data Science
• Climate Science

Background:

• Accurate rainfall forecasting is crucial for water resource management, flood prevention, agriculture, and disaster preparedness.
• Traditional forecasting methods struggle with the complex dynamics of rainfall patterns.
• Advanced machine learning offers potential for improved prediction accuracy.

Purpose of the Study:

• To enhance rainfall prediction precision using a fusion of observed data and climate projections.
• To evaluate the performance of various machine learning models for daily, 3-day, and weekly rainfall forecasting.
• To identify the most effective models for different forecasting intervals.

Main Methods:

• Fusion of observed rainfall data with climate change projections.
• Assessment of advanced machine learning models including Efficient Linear Support Vector Machine (ELSVM), Exponential Gaussian Process Regression (Exponential GPR), and Long Short-Term Memory (LSTM).
• Evaluation of model performance using metrics such as R², Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE).

Main Results:

• ELSVM achieved the highest accuracy for daily rainfall forecasting (R² = 0.3868).
• For 3-day forecasts, Exponential GPR slightly outperformed LSTM (MAE=15.84, MSE=547.04, RMSE=23.39).
• LSTM showed higher error rates for weekly forecasts (MAE=14.07, MSE=363.03, RMSE=19.05, R²=0.1662).

Conclusions:

• Combining machine learning with data fusion significantly improves rainfall prediction accuracy and reliability.
• Advanced models like ELSVM, Exponential GPR, and LSTM offer substantial potential for enhanced forecasting systems.
• These improved predictions contribute to better water resource management, climate adaptation, and disaster preparedness.