Bridging the resolution gap: Machine learning for local-scale groundwater drought monitoring in Punjab, Pakistan

Citation

Youssefi F, Ali S, Khorrami B, Khan MM, Jehanzaib M, Mahmood T, Han Z, Haotian W & Faisal M (2025) Bridging the resolution gap: Machine learning for local-scale groundwater drought monitoring in Punjab, Pakistan. Groundwater for Sustainable Development, 31, Art. No.: 101545. https://doi.org/10.1016/j.gsd.2025.101545

Abstract
The drought, resulting from insufficient precipitation, significantly impacts groundwater resources, thereby threatening water availability and human livelihood. Conventional approaches to evaluating groundwater shortages are influenced by limited observations across spatial and temporal dimensions. Previous studies on groundwater drought occurrences in the Punjab Doabs of Pakistan were explored at limited resolution (0.25°). Therefore, the Extreme Gradient Boosting (XGBoost) machine learning model was employed to downscale the Terrestrial Water Storage Anomaly (TWSA) data (0.25°) obtained from Gravity Recovery and Climate Experiment (GRACE) to 0.1° resolution and derived high-resolution Groundwater Storage Anomaly (GWSA) data (0.1°) to assess the Groundwater Drought Index (GGDI) at higher resolution. The XGBoost model had a Pearson correlation (R) of 0.99, an NSE of 0.99, an RMSE of 5.22 mm, and an MAE of 2.75 mm. The groundwater dynamics can be accurately simulated by the XGBoost model, which has demonstrated a validation with in situ groundwater level data showing an R of 0.77. The characteristics of the GGDI trend, along with its spatio-temporal variations, were analyzed in the Punjab Doabs from 2003 to 2020. Findings indicated severe drought events in 2010 and from 2018 to 2020 in the study area. Additionally, results indicated that GGDI exhibited similar drought patterns to the model-generated Groundwater Storage Percentile (GWSP) with a correlation coefficient of 0.66. This framework serves as an effective tool for monitoring drought in the Punjab Doabs and similar climatic zones.

Keywords
Punjab doabs; GRACE; Machine learning; Downscaling; GGDI; Drought

Journal
Groundwater for Sustainable Development: Volume 31