Abstract
Contamination of soil and groundwater presents a widespread global problem, significantly impacting both human well-being and environmental stability. Conventional models employed for estimating pollutant concentrations under varying climatic conditions demand extensive computational power and high-performance computing resources. In response to this issue, we have devised an innovative method utilizing a physics-informed machine learning technique, known as the U-Net Enhanced Fourier Neural Operator (U-FNO), to generate rapid surrogate models for flow and transport. These models are capable of forecasting groundwater pollution levels under diverse climatic situations and subsurface characteristics without necessitating a supercomputer. In our research, we centered our attention on the Department of Energy's Savannah River Site (SRS) F-Area and established two time-dependent structures: U-FNOB and U-FNOB-R. Both frameworks incorporate a tailored loss function, including specific physical constraints of groundwater flow and transport such as spatial derivatives, and contaminant boundary conditions. The findings of our study indicate that the U-FNO models can consistently foresee spatialtemporal fluctuations in groundwater flow and pollutant transportation properties, such as contaminant concentration, hydraulic head, and Darcy's velocity. Our research reveals that the U-FNOB-R architecture is especially adept at predicting the effects of alterations in recharge rates on groundwater contamination sites, delivering superior time-dependent forecasts compared to the U-FNOB structure. Our novel approach holds the potential to revolutionize environmental monitoring and remediation efforts by providing rapid, precise, and cost-efficient estimations of groundwater pollution levels under uncertain climate conditions.
| Original language | English |
|---|---|
| Article number | 105508 |
| Journal | Computers and Geosciences |
| Volume | 183 |
| DOIs | |
| State | Published - Jan 2024 |
Funding
The present work was executed at the Frontier Development Laboratory (FDL) USA in 2022. FDL USA embodies a collaborative research initiative involving public and private organizations, including NASA, the SETI Institute, Trillium Technologies Inc, and a consortium of industry partners such as Google Cloud, Intel, IBM, Lockheed Martin, NVIDIA, and Pasteur Labs. These entities contribute critical data, expertise, training, and computational resources that facilitate rapid experimentation and iteration in data-intensive domains. The current material is based on research supported by the Department of Energy, United States of America ’s Artificial Intelligence & Technology Office under Award Number DE-AI0000001 . Furthermore, this investigation is backed by the U.S. Department of Energy’s Office of Environmental Management through the Advanced Long-term Monitoring Systems (ALTEMIS) project. Funding was granted under Award Number DE-AC02-05CH11231 to Lawrence Berkeley National Laboratory, United States of America . The research made use of the National Energy Research Scientific Computing Center (NERSC) resources, a U.S. Department of Energy Office of Science User Facility situated at Lawrence Berkeley National Laboratory, operating under contract DE-AC02-05CH11231. In addition, the study utilized the Lawrencium computational cluster resource provided by the IT Division at Lawrence Berkeley National Laboratory, which received support from the Director of the Office of Science, United States of America and the Office of Basic Energy Sciences within the U.S. Department of Energy under contract DE-AC02-05CH11231 .
Keywords
- Boundary condition loss function
- Groundwater contamination
- Physics situations
- Physics-informed machine learning
- Soil and groundwater contamination
- Supercomputer
- U-FNOB
- U-FNOB-R
- U-Net Enhanced Fourier Neural Operator (U-FNO)