Saedi et al. (2026) A water-balance-guided hybrid machine-learning framework for quantifying recharge to assess groundwater depletion in the Southeastern USA

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-09-28
• Authors: Fatemeh Saedi, Mukesh Kumar, T. Prabhakar Clement
• DOI: 10.1016/j.ejrh.2026.104004

Research Groups

• University of Alabama, Department of Civil, Construction, and Environmental Engineering
• USGS (United States Geological Survey)

Short Summary

This study developed a water-balance-guided hybrid machine-learning framework to estimate net recharge in the Southeastern United States. The model used a multilayer perceptron (MLP) to predict WTF recharge from WB recharge data and other hydrological parameters, achieving an average annual depletion rate of about 2.5 km³/year.

Objective

• Estimate net groundwater recharge (WTF recharge) using a water-balance-guided hybrid machine-learning framework.
• Evaluate the spatiotemporal variability of groundwater depletion (GWD) levels in the Southeastern US.

Study Configuration

• Spatial Scale: Regional scale, covering three major USGS Hydrologic Unit Code level 2 (HUC-2) regions within the Southeastern region.
• Temporal Scale: Annual time step from 2008 to 2020.

Methodology and Data

• Models used: Multilayer Perceptron (MLP), Ridge Regression, Random Forest
• Data sources:
  • Recharge-per-specific-yield (RpSy) dataset developed by Malakar et al. (2024)
  • Water-balance derived recharge (WB recharge data) from Saedi et al. (2025)
  • Groundwater withdrawals for crop irrigation, public supply, and thermoelectric use from the USGS water-supply and thermoelectric water-use models

Main Results

• Net recharge for the region generally ranges from about 10–15% of average annual local rainfall for most of the catchments.
• Estimated GWD levels exhibited noticeable spatiotemporal variability, with cumulative depletion reaching approximately 31 km³ during 2008–2020.

Contributions

• This study highlights the importance of distinguishing different types of recharge estimates and demonstrates the need for using net recharge in GWD assessments.
• The developed framework provides a robust and physically consistent approach for estimating WTF recharge at regional scales.

Funding

• This research was funded by [insert funding agencies or projects, e.g., NSF, USGS, etc.].

Citation

@article{Saedi2026waterbalanceguided,
  author = {Saedi, Fatemeh and Kumar, Mukesh and Clement, T. Prabhakar},
  title = {A water-balance-guided hybrid machine-learning framework for quantifying recharge to assess groundwater depletion in the Southeastern USA},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.104004},
  url = {https://doi.org/10.1016/j.ejrh.2026.104004}
}