Agricultural drought forecasting using remote sensing: A hybrid modeling framework by integrating wavelet transformation and machine learning techniques

Climate change has intensified the frequency and severity of droughts, increasingly threatening environmental sustainability by reducing agricultural productivity and placing additional stress on water resources. Although accurate drought forecasting is vital for resource management and disaster mitigation, it remains a formidable challenge because of the intricate and non-stationary characteristics of climatic and ecological processes. This study proposes an innovative hybrid modeling framework that integrates wavelet transform preprocessing with ensemble-based machine learning (ML) models, XGBoost, AdaBoost, and Random Forest, to enhance agricultural drought prediction. The Sindh Province of Pakistan was selected as the case study area. CHIRPS precipitation data and MODIS-derived environmental indicators, including Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), Modified Normalized Difference Water Index (MNDWI), Vegetation Condition Index (VCI), and Vegetation Health Index (VHI), were introduced into models. Wavelet-based decomposition was first applied to the time-series data to capture multi-scale patterns and reduce noise. The resulting signals were then used to train the ML models to forecast drought, as represented by the VHI. Results show that wavelet preprocessing significantly improves prediction accuracy. Among the tested models, XGBoost achieved the highest performance (R² = 0.964), followed by AdaBoost (R² = 0.946) and Random Forest (R² = 0.9263). These findings highlight the effectiveness of combining wavelet analysis with ML techniques for drought assessment. The proposed framework offers a robust decision-support tool for identifying drought-prone areas, enhancing agricultural resilience, and informing policy responses to climate-related risks. • A hybrid drought prediction model combines wavelet decomposition with ensemble machine learning. • MODIS-derived environmental indicators (NDVI, LST, MNDWI, VCI, VHI) and CHIRPS data. • XGBoost delivers the highest accuracy, outperforming other models. • An early warning tool in drought-prone regions for sustainable water and agricultural management.