Estimating groundwater use with machine learning and remote sensing
Overview
As the increasing global population and climate change impacts are reducing surface water deliveries, there has been a substantial increase in groundwater use, which is projected to increase even further. However, proactive local-scale monitoring of groundwater withdrawals (extraction or pumping) is typically limited in most countries, including in the United States. Such monitoring is paramount to addressing the water security issues of the 21st century and beyond. Existing methods to estimate groundwater withdrawals rely on process-based models which are computationally intensive and require strict calibration procedures.
To overcome such limitations, our research group was the first to show that an integrated machine learning and remote sensing-based framework can effectively and accurately predict groundwater withdrawals at regional (2 km – 5 km) to field (1 km) scales (Majumdar et al., 2020, 2021, 2022, 2024). These models use open-source satellite products and in-situ groundwater pumping datasets to predict withdrawals. We successfully evaluated this framework over Kansas, Arizona, and the Mississippi Alluvial Plain (MAP). Currently, we are working toward improving and extending these methods to the national scale by ingesting newer and better remote sensing datasets, such as OpenET, IrrMapper, LANID, and others.
Funding
NASA, USGS
Study areas
Kansas, Arizona, Mississippi Alluvial Aquifer, western US
Figure Caption:
Right: Machine learning model predicted groundwater withdrawal estimates for (a) Kansas (Majumdar et al. 2020), (b) Mississippi Alluvial Plain (Majumdar et al. 2024), and (c) Arizona (Majumdar et al. 2022).
Bottom: Presentation on using ET to estimate groundwater use by PhD candidate Md Fahim Hasan
Papers
Majumdar, S., Smith, R., Butler, J. J., & Lakshmi, V. (2020). Groundwater withdrawal prediction using integrated multitemporal remote sensing data sets and machine learning. Water Resources Research, 56(11), e2020WR028059. https://doi.org/10.1029/2020WR028059
Majumdar, S., Smith, R., Conway, B. D., Butler, J. J., Lakshmi, V., & Dagli, C. H. (2021). Estimating Local-Scale Groundwater Withdrawals Using Integrated Remote Sensing Products and Deep Learning. 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS, 4304–4307. https://doi.org/10.1109/IGARSS47720.2021.9554784
Majumdar, S., Smith, R., Conway, B. D., & Lakshmi, V. (2022). Advancing Remote Sensing and Machine Learning‐Driven Frameworks for Groundwater Withdrawal Estimation in Arizona: Linking Land Subsidence to Groundwater Withdrawals. Hydrological Processes, 36(11), e14757. https://doi.org/10.1002/hyp.14757
Majumdar, S., Smith, R. G., Hasan, M. F., Wilson, J. L., White, V. E., Bristow, E. L., Rigby, J. R., Kress, W. H., & Painter, J. A. (2024). Improving crop-specific groundwater use estimation in the Mississippi Alluvial Plain: Implications for integrated remote sensing and machine learning approaches in data-scarce regions. Journal of Hydrology: Regional Studies, 52, 101674. https://doi.org/10.1016/j.ejrh.2024.101674