Assessing Agricultural Drought Risk Under CMIP6 Scenarios Using Hybrid AI Models and Satellite-Derived TVDI

This study focuses on predicting the Temperature Vegetation Dryness Index (TVDI), an agricultural drought index, for a Mango orchard in Tamale Ghana using climate projections and satellite imagery. The nonlinear cross-correlation between the Standardized Precipitation Index (SPI) and TVDI was examined to comprehend the impact of meteorological drought on agricultural drought. Historical precipitation data and CMIP6 projected data (2015–2050) from 35 climate models across four Shared Socioeconomic Pathway (SSP) scenarios were subjected to bias correction and utilized to compute SPI. TVDI was obtained from Landsat 8/9 imagery and validated by UAV-based data, demonstrating good agreement. A nonlinear study indicated that SPI precedes TVDI by 2–4 months, facilitating lag-based forecasting. A hybrid Wavelet-ANFIS/FCM model predicted TVDI utilizing SPI as input, with significant accuracy. Mutual information (MI) analysis indicated SPI leads TVDI by 0–6 months, facilitating lag-based prediction. A hybrid model predicted TVDI using SPIs as inputs with high accuracy (RMSE < 0.08 in training and test phases). These findings underscore the escalating drought risk associated with high-emission SSP scenarios and emphasize the necessity for adaptive methods such as supplementary measures. The system provides AI-driven solution for predicting agricultural drought, thereby promoting sustainable development goals (SDGs) in the context of climate change.