yoon-etal_2025_naturewater

The persistence of groundwater depletion in the United States across widely ranging futures

Jim Yoon^1*, Stephen Ferencz¹, and Travis Thurber¹

¹ Pacific Northwest National Laboratory, Richland, WA, USA.

* corresponding author: [email protected]

Abstract

Groundwater resources are essential for crop production, enabling irrigation in many arid regions of the world. With a growing dependence on groundwater to serve increasing crop demands, concerns have been raised regarding alarming rates of groundwater depletion at the global scale. Modeling efforts focused on projecting future groundwater depletion have been limited by key factors: exogenous assumptions of human demand for groundwater that are unresponsive to changing groundwater availability and cost, simplistic treatment of the physics of groundwater depletion, narrow application to localized contexts, and limited exploration of future uncertainties. To address these gaps, we introduce a new coupled human-natural modeling approach for process-rich, broad-exploration of groundwater depletion futures, initially deployed for the continental United States. Our results indicate persistent groundwater depletion across a range of hydrologic-economic futures, with many of the key agricultural regions of the U.S. encountering substantial depletion even in the absence of adverse hydrologic or economic change.

Journal reference

TBD

Code reference

1. Jim Yoon, Stephen Ferencz, & Travis Thurber. https://github.com/jimyoon/farm_gw_sensitivity/

Data reference

Input data

2. Jim Yoon, Stephen Ferencz, & Travis Thurber. To be made available via MSDLIVE https://doi.org/10.57931/2565944

Output data

3. Jim Yoon, Stephen Ferencz, & Travis Thurber. To be made available via MSDLIVE https://doi.org/10.57931/2565944

Contributing modeling software

Model	Version	Repository Link	DOI
farm_gw_depletion	v1.0.0	https://github.com/jimyoon/farm_gw_sensitivity/	TBD

Reproduce my experiment

1. Run the farms_gw_future_3Dsurface_HPC.py from [1] in a high performance computing environment (current simulations were deployed on PNNL's Deception HPC)
2. Step 1 produces a consolidated csv file with results for all scenario/parameter combinations. These are made available in the raw_outputs folder of [3]

Reproduce my figures

3. Figure 1 - Conceptual Model Diagram: Manually generated in Powerpoint; Map of grid cells generated in QGIS |
4. Figures 2-4 - Depletion results: Process raw output from Step 2 above through grid_cell_depletion_post_process.py in [1]. Spatial joins to grid cells & aquifers and further visualization conducted in QGIS. Final touch-ups performed using Inkscape.