Enhancing Drought Forecasting Accuracy: A Wavelet-ARIMA Modelling Approach for Standardized Precipitation Index Data

Abstract

Droughts are periods of inadequate precipitation that have severe and diverse impacts on human societies and ecosystems. However, a reliable forecasting technique can improve drought monitoring and mitigate impacts. This research aimed to assess the predictive accuracy of a hybrid forecasting model, Wavelet-ARIMA (W-ARIMA), for drought forecasting, using the Standardized Precipitation Index (SPI) with a traditional ARIMA model as a benchmark. Monthly rainfall data from the Badeggi district in the north-central part of Nigeria, covering the period from January 1968 to December 2018, was utilized for the analysis. Subsequently, SPI values for various time scales (3, 6, 9, and 12) were computed. The Wavelet Transform was then employed to decompose the data series into L components, encompassing details and approximations from A to D (AnD1D2 DL 1) of the SPIs, respectively. Furthermore, an ARIMA model was fitted to each of these details and approximations, and their sum constituted the forecasted W-ARIMA values. Evaluation based on the performance metric, the RMSE of W-ARIMA (0.4889, 0.4326, 0.3566, and 0.2177) while that of ARIMA (0.7771, 0.6667, 0.4648, and 0.3212), revealed that the hybrid model (W-ARIMA) consistently outperformed the ARIMA model among the metrics for all the SPI 3, 6, 9, and 12, respectively.