Modeling climate-driven groundwater decline and the effectiveness of progressive abstraction reduction using a CMIP6–NGBoost framework

Climate change and groundwater management framework using CMIP6 and NGBoost, showing projected warming, declining precipitation and groundwater levels, and the effects of 1–3% annual abstraction reductions on groundwater recovery. Groundwater in semi-arid regions is increasingly stressed by climate change and over-abstraction, yet the potential of management interventions to reverse depletion remains uncertain. This study integrates bias-corrected CMIP6 projections with a Natural Gradient Boosting (NGBoost) model to simulate groundwater dynamics and evaluate progressive abstraction reduction in the Shahrood Plain, Iran. Climate projections under SSP2-4.5 and SSP5-8.5 were downscaled using the delta change factor method and used to drive a model trained on hydro-climatic data and groundwater withdrawals (2000–2014). The model reproduced groundwater levels with strong performance (MAE = 0.242 m; KGE = 0.856; R2 = 0.845). Under constant abstraction, projections for 2015–2054 indicate cumulative storage deficits of 513–539 million m3, with greater decline under SSP5-8.5. Annual abstraction reductions of 1 and 2% mitigate deficits but do not restore aquifer balance, whereas a sustained 3% reduction reverses long-term decline under both scenarios. These findings showed that the proposed framework can be used to study climate-resilient groundwater management.