Drought Forecasting using Palmer Drought Severity Index with Wavelet Transform Technique and Machine Learning Methods

Abstract

Drought is a climatic phenomenon characterised by persistent and anomalous water scarcity caused by weather anomalies. Drought is a slowly evolving phenomenon and affects a wide range of areas. The drought indices are one of the ways to monitor and measure drought. Therefore, the Palmer drought severity index (PDSI) was used as an operational and a valid model. This study calculated Palmer Drought Severity Index (PDSI) on a 12-month time scale with meteorological data for the Ceyhan Basin between January 1989 and July 2020. The calculated PDSI values were modelled with Discrete Wavelet Transform (DWT) technique using Artificial Neural Network (ANN) and Support Vector Machine (SVM) methods at different training rates and drought forecasts were made. The forecasting success rate with the models created using machine learning methods was statistically evaluated. This study determined that machine learning methods could be applied in drought forecasting in drought monitoring and mitigation.