An innovative hybrid W-EEMD-ARIMA model for drought forecasting using the standardized precipitation index

Drought, a critical consequence of water scarcity and climate change, profoundly impacts human life. This study introduces a new W-EEMD-ARIMA hybrid model to forecast drought using Kabul's monthly precipitation data from 1970 to 2019. By integrating Ensemble Empirical Mode Decomposition (EEMD) and wavelet transform, we enhance the ARIMA/SARIMA model. Comparing the accuracy of our proposed method with ARIMA, Wavelet-ARIMA, and EEMD-ARIMA, using a training dataset (1970-2009) and validation data (2010-2019), we observed superior performance in our proposed W-EEMD-ARIMA across both datasets and all Standardized Precipitation Index (SPI) values. For SPI 12 validation, our model achieves an RMSE of 0.0736, MAE of 0.0575, MAPE of 18.9674, and R-squared of 0.9946, surpassing ARIMA (RMSE: 0.2561, MAE: 0.1874, MAPE: 60.0220, R-squared: 0.9361), Wavelet-ARIMA (RMSE: 0.1002, MAE: 0.0691, MAPE: 23.7122, R-squared: 0.9898), and EEMD-ARIMA (RMSE: 0.0858, MAE: 0.0660, MAPE: 24.5411, R-squared: 0.9925). Across SPI 3, 6, and 9, our hybrid model consistently outperforms others in both training and testing datasets, with lower RMSE, MAE, and MAPE, alongside higher R-squared values. These findings illustrate the superiority of our hybrid proposed model in enhancing drought prediction accuracy over the ARIMA, Wavelet-ARIMA, and EEMD-ARIMA approaches.