Integrating wavelet transform and support vector machine for improved drought forecasting based on standardized precipitation index

ABSTRACT This study proposes a hybrid discrete wavelet transform (DWT) and support vector machine (SVM) model (W-SVM) to enhance drought prediction accuracy compared to traditional SVM and autoregressive integrated moving average (ARIMA) models using the Standardized Precipitation Index (SPI). Fifty years of historical average monthly precipitation data from the Kabul province, Afghanistan, are analysed. The data are split into training and testing sets at an 80/20 ratio. Daubechies order 2 (db2) is selected as the mother wavelet function with three decomposition levels for SPI-3, -6, -9, and -12. Each decomposed element, including detail components (D1, D2, D3) and approximation (A3), is forecasted with an SVM model, and the final forecast is obtained by summing all levels. Statistical metrics such as root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R2) are used for evaluation. Results demonstrate that W-SVM outperforms benchmark models, achieving lower RMSE and MAE values and higher R2 across all SPI values. These findings highlight DWT's effectiveness in enhancing SVM's capability for SPI pattern recognition. The proposed model provides accurate and timely drought predictions, making it a valuable tool for real-time drought monitoring and management, enabling informed decision-making to reduce drought impacts.