Asfaw et al. (2026) Explainable AI as a diagnostic tool for analyzing spatiotemporal variability in simulated groundwater recharge: Application to a semi-arid river basin

Identification

• Journal: Environmental Modelling & Software
• Year: 2026
• Date: 2026-09-17
• Authors: Dawit Asfaw, Ryan G. Smith, Michael J. Ronayne, Sayantan Majumdar, Salam A. Abbas, Ryan T. Bailey
• DOI: 10.1016/j.envsoft.2026.107178

Research Groups

• Department of Geosciences, Colorado State University
• Department of Civil and Environmental Engineering, Colorado State University
• Division of Hydrologic Sciences, Desert Research Institute

Short Summary

This study develops machine learning (ML) models to predict groundwater recharge using explainable AI (XAI) as a diagnostic tool. The research focuses on the semi-arid Lower Arkansas River Basin in Colorado, USA.

Objective

• Investigate the drivers of groundwater recharge and identify critical threshold values.
• Develop ML surrogate and stand-alone predictor models to analyze spatiotemporal variability in simulated groundwater recharge.

Study Configuration

• Spatial Scale: The study area is located in southeastern Colorado with an area of 64,000 km2.
• Temporal Scale: The period of study is from 2002 to 2015.

Methodology and Data

• Models used:
  • SWAT+ (a process-based model)
  • Random Forest Regressor (an ensemble machine learning algorithm)
• Data sources:
  • Simulated groundwater recharge values from SWAT+
  • Open-source data, including climate, soil physical properties, hydrogeological, land use and land cover, and topography-based hydrological factors.

Main Results

• The surrogate ML model shows high predictive accuracy with NSE values of 0.99 for training and 0.98 for testing.
• The stand-alone predictor model also demonstrates good performance with NSE values of 0.97 for training and 0.91 for testing.
• Explainable AI assessments reveal that precipitation, average soil water content, and snowmelt contribute positively to simulated recharge.

Contributions

• This study provides a valuable approach for investigating recharge dynamics in unconfined aquifers using ML surrogate and stand-alone predictor models with XAI as a diagnostic tool.

Funding

• The research was funded by the following projects/programs:
  • [Insert project/program names and reference codes]

Citation

@article{Asfaw2026Explainable,
  author = {Asfaw, Dawit and Smith, Ryan G. and Ronayne, Michael J. and Majumdar, Sayantan and Abbas, Salam A. and Bailey, Ryan T.},
  title = {Explainable AI as a diagnostic tool for analyzing spatiotemporal variability in simulated groundwater recharge: Application to a semi-arid river basin},
  journal = {Environmental Modelling & Software},
  year = {2026},
  doi = {10.1016/j.envsoft.2026.107178},
  url = {https://doi.org/10.1016/j.envsoft.2026.107178}
}