Explainable AI for Drought Forecasting: A Benchmark Dataset and XAI Framework for Andhra Pradesh and Telangana
Contributors
Dr. A Manjula
Dr. Shish Ahmad
Keywords
Proceeding
Track
General Track
License
Copyright (c) 2026 Sustainable Global Societies Initiative

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Abstract
Building on our IoT–AI crop-stress framework (Paper 1) and ConvLSTM–GNN spatio-temporal drought prediction for Telangana (Paper 2), this paper closes the adoption-critical gap of interpretability. We contribute (1) AP–TG DroughtBench, the first publicly available district-level drought benchmark for all 59 districts of Andhra Pradesh and Telangana (2001–2024), harmonising MODIS, Sentinel-2, CHIRPS, ERA5-Land, and IMD data into monthly feature tensors labelled across five drought-severity categories; and (2) an XAI-augmented ConvLSTM–GNN that integrates SHAP temporal attribution, GNN edge-importance saliency, and counterfactual drought scenarios to explain when, where, and why drought is forecast. The model retains ≥85% classification accuracy while producing expert-validated, human-interpretable alerts delivered through the existing multilingual mobile/voice DSS. This work supports SDG 2 (Zero Hunger) and SDG 13 (Climate Action).