Divergent effects of climate change on future groundwater availability in key mid-latitude aquifers

The study investigates how anthropogenic climate change will alter groundwater storage (GWS) in major mid-latitude aquifers that are critical for water supply and food production, especially in arid and semi-arid regions. Climate change modifies hydrological processes including infiltration, percolation, recharge, and evapotranspiration (ET), while human activities, notably groundwater pumping and irrigation, directly deplete aquifers. Prior large-scale estimates have relied on offline hydrological models forced by GCM outputs, which carry uncertainties from climate projections, downscaling, and omission of land–atmosphere feedbacks. To better capture coupled hydroclimate processes, this study uses a fully coupled Earth System Model to assess climate-driven GWS changes through the 21st century under RCP8.5, explicitly excluding direct human water use to isolate natural climate effects. The goal is to identify mechanisms (precipitation, snowmelt, ET) governing future GWS trends and to contrast them with impacts of anthropogenic pumping.

Previous studies estimated global and regional groundwater availability using offline hydrological models driven by GCM outputs. These approaches face multiple uncertainties: internal climate variability, inter-model and scenario differences, and additional errors introduced by downscaling. Offline (one-way) coupling cannot capture essential land–atmosphere feedbacks, despite evidence that groundwater dynamics influence atmospheric processes. Groundwater parameterizations embedded in land models (e.g., CLM4.0) have improved representation of unconfined aquifers and have been evaluated against GRACE observations. However, local-scale limitations persist. The literature highlights the need for fully coupled modeling frameworks to better quantify hydroclimatic drivers of GWS changes and to separate climate-driven effects from anthropogenic pumping.

Study regions: Seven mid-latitude aquifers identified by GRACE as undergoing rapid depletion (2002–2013) were analyzed: (1) Central Valley (California), (2) Southern Plains (central U.S.), (3) Middle East (Turkey, Syria, Iraq, Iran; Tigris–Euphrates basins), (4) Northwestern India (upper Indus and Ganges), (5) North China Plain, (6) Guarani (South America), and (7) Canning (NW Australia). Aquifer boundaries follow prior studies. Model and experiments: Climate-driven projections use the Community Earth System Model Large Ensemble (CESM-LE) with 30 ensemble members at 0.9°×1.25° resolution, under RCP8.5 for 2006–2100. CESM components include CAM5 (atmosphere), CLM4.0 (land), POP2 (ocean), and CICE4 (sea ice). CLM4.0 includes a physically based representation of an unconfined aquifer, groundwater recharge/discharge, capillary rise, and a TOPMODEL-based runoff scheme. There is no lateral water exchange between land grid cells. Anthropogenic water use (e.g., pumping, irrigation withdrawals, dam operations, inter-basin transfers) is not included in the CESM-LE projections, isolating climate-driven effects. Diagnostics: GWS changes arise from the balance of recharge (sum of downward gravity drainage and upward capillary rise) and baseflow to rivers, modulated by precipitation partitioning (rainfall vs. snowfall), snowmelt, ET (evaporation and transpiration), and infiltration. Annual means are computed from monthly outputs; linear trends are estimated over 2006–2100. Trend significance is assessed via Student’s t-test (p<0.05). Uncertainty is quantified by the inter-member standard deviation (n=30). Seasonal changes are evaluated by comparing 2071–2100 vs. 2006–2035 means. Attribution of recharge drivers: Multiple linear regression relates standardized annual groundwater recharge (response) to standardized annual rainfall, snowmelt, and ET (predictors) at each grid cell. Statistically significant regions (90% confidence) are retained. Relative contributions (0–1) of each factor are derived from regression coefficients and visualized as RGB (red: snowmelt; green: ET; blue: rainfall). Aquifer-mean contributions are summarized. Comparisons with pumping: For context, twentieth-century simulations including groundwater pumping for industrial, irrigation, agricultural, and domestic use are taken from a prior CESM study with global abstraction datasets. These are compared to climate-only simulations to infer pumping impacts on historical GWS trends. Observational estimates: GRACE mascon total water storage combined with GLDAS (multi-model) estimates of canopy, snow, and soil moisture are used to infer GWS trends (2003–2014) after removing seasonality, providing observational context for recent depletion.

- Divergent climate-driven GWS responses: Future GWS changes are not governed by precipitation trends alone; increased ET (from higher atmospheric demand) and reduced snowmelt substantially modulate recharge, leading to heterogeneous aquifer responses. - Central Valley (California): Climate-only projections show no significant long-term GWS trend due to competing effects—more winter rainfall and less snowfall increase winter GWS, but reduced spring snowmelt lowers early-spring recharge; higher temperatures enhance ET and capillary upward fluxes. Seasonality shifts pose management challenges even without a significant long-term trend. - Southern Plains (central U.S.): Significant climate-driven GWS decline projected at approximately −23.3 ± 11.4 mm per decade, linked to reduced infiltration and diminished spring snowmelt, decreasing recharge and deepening the water table. Increased drought risk and reduced recharge are consistent with prior studies. - Middle East (Tigris–Euphrates–Western Iran): Projected continuous GWS decline of about −15.2 ± 3.4 mm per decade. Spring groundwater recharge is projected to drop markedly (by 77% in MAM) by century’s end due to reduced snowfall/snowmelt over the Iranian and Anatolian Plateaus and increased growing-season transpiration (~+13% in MAM) from CO2 fertilization and warming; long-term ET rises (~+1.5 ± 1.3 mm yr−1 per decade), further reducing recharge via capillary demand. - Northwestern India: Despite recent observed depletion (due to pumping), climate-only projections indicate increasing GWS driven predominantly by increased rainfall, with relative contributions to recharge change: rainfall 60%, snowmelt 21%, ET 19%. - North China Plain: Significant precipitation increases outweigh rising ET and decreasing snowmelt, boosting infiltration and recharge (recharge trend ~+3.7 mm yr−1 per decade) and leading to a shallower water table in the absence of pumping; seasonal recharge timing shifts. - Guarani (South America) and Canning (NW Australia): In these more humid systems (annual precipitation >900 mm), precipitation increases dominate over ET, raising P−ET, infiltration, and GWS. - Global attribution pattern: Rainfall changes dominate recharge in monsoon/humid regions; snowmelt controls recharge at higher latitudes/elevations; ET changes dominate in dry regions. Example aquifer-mean contributions: Canning (7): snowmelt 0.00, ET 0.36–0.40, rainfall 0.64–0.60; Southern Plains (2): snowmelt ~0.30, ET ~0.24, rainfall ~0.46. - Pumping vs. climate: In heavily pumped aquifers (Central Valley, Southern Plains, Middle East, Northwestern India, North China Plain), anthropogenic withdrawals can easily exceed natural climate-driven replenishment, producing net depletion. In lightly pumped areas (e.g., Canning), GWS trends largely reflect natural hydroclimatic variability. - Spatial heterogeneity: Strong intra-aquifer spatial variability in projected GWS change, especially in complex terrain (e.g., Central Valley, Northwestern India). Reduced snowmelt replenishment in snow-dominant regions is a robust negative driver of storage.

The study addresses how climate change will alter groundwater storage by isolating climate-driven mechanisms within a fully coupled Earth system model. Findings show that future GWS trends cannot be inferred from precipitation alone; increased ET and reduced snowmelt substantially affect recharge timing and magnitude. This mechanistic attribution explains divergent aquifer responses: declines in drier, snow-influenced systems (Southern Plains, Middle East) and potential increases where precipitation intensifies (North China Plain, Northwestern India, Guarani, Canning), with seasonality shifts that complicate management. Comparison with simulations including pumping indicates that, in many hotspots, anthropogenic withdrawals overwhelm natural replenishment, aligning with GRACE-era observations of rapid depletion. The results underscore the importance of land–atmosphere coupling and internal variability in shaping groundwater outcomes, highlight regional heterogeneity, and suggest that adaptive management—especially leveraging projected recharge increases and seasonal shifts—could help slow or reverse depletion in some basins.

Using a fully coupled Earth System Model with explicit unconfined aquifer dynamics, the study demonstrates that climate-driven groundwater storage changes through the 21st century will be highly heterogeneous and controlled by the interplay of precipitation, evapotranspiration, and snowmelt. Some key aquifers (Southern Plains, Middle East) face intensified climate-driven depletion, while others (Central Valley, Northwestern India, North China Plain, Guarani, Canning) may experience stable or increasing climate-driven recharge, albeit with altered seasonality. However, anthropogenic pumping can readily exceed natural replenishment, implying continued net depletion without changes in water use. Future research should: integrate socioeconomic scenarios of groundwater abstraction with coupled ESMs; evaluate alternative emission pathways; improve representation of mountainous snow processes, lateral subsurface flow, and land cover/soil property changes; and apply higher-resolution/downscaled modeling to better capture spatial heterogeneity and management-relevant seasonality.

- Model structure: CLM4.0 employs a simplified groundwater scheme without lateral subsurface flow and no inter-grid-cell water exchange, limiting local-scale realism. - Resolution: Coarse GCM resolution likely under-represents orographic precipitation and snow processes, potentially underestimating snowmelt contributions in mountainous regions. - Forcing uncertainties: Precipitation remains a major source of uncertainty; GCM biases (e.g., Northern Hemisphere winter meridional winds) can affect regional precipitation and thus recharge. - Scenario and model dependence: Results are contingent on CESM-LE physics and the RCP8.5 pathway; other ESMs or scenarios may yield different projections. - Anthropogenic water use: Future projections exclude direct human water management (pumping, irrigation withdrawals, dams, transfers), so total future GWS trajectories in managed basins may be more negative than climate-only results. - Process omissions: Potential impacts of changing soil properties, land cover/land use, freeze–thaw dynamics, and precipitation intensity are not fully resolved and could alter recharge and storage. - Observational comparisons: GRACE–GLDAS-derived GWS estimates carry uncertainties from mascon processing and land model component storage estimates.