Abstract
Ensemble-based Data Assimilation (EDA) has been effectively applied to estimate model parameters through inverse modeling in subsurface flow and transport problems. To facilitate the management of EDA workflow and lower the barriers for adopting EDA-based parameter estimation in subsurface science, we develop a software framework linking the Data Assimilation Research Testbed (DART) with a massively parallel subsurface FLOw and TRANsport code PFLOTRAN. DART-PFLOTRAN enables an iterative EDA workflow based on the Ensemble Smoother for Multiple Data Assimilation method (ES-MDA) to improve estimation accuracy for nonlinear forward problems. We verify the implementation of ES-MDA in DART-PFLOTRAN using two synthetic cases designed to estimate static permeability and dynamic exchange fluxes across the riverbed from continuous temperature measurements. Both cases yield accurate estimations of the parameters compared to their synthetic truth. With a code base in Python and Fortran, DART-PFLOTRAN paves the way for large-scale inverse modeling using the sequential ES-MDA.
| Original language | English |
|---|---|
| Article number | 105074 |
| Journal | Environmental Modelling and Software |
| Volume | 142 |
| DOIs | |
| State | Published - Aug 2021 |
| Externally published | Yes |
Funding
Although iterative ES methods significantly reduce the number of forward simulation restarts required to conserve physical laws (Chen et al., 2013), the implementation of ES-MDA for large-scale inverse modeling is not trivial due to the complex workflow in launching multi-physics, parallel forward simulations, which is often required for managing the computational challenges (Chen et al., 2013; Song et al., 2019; Shuai et al., 2019). Therefore, a user-friendly software framework for performing EDA with computationally intensive forward models and heterogeneous observational data will significantly increase scientific productivity. There exist multiple community-supported data assimilation tools, including PEST++ and the Data Assimilation Research Testbed (DART). PEST++, developed by the U.S. Geological Survey for both parameter estimation and uncertainty analysis, adopts an iterative ensemble smoother for solving the Gauss-Levenberg-Marquardt algorithm in model calibration (White et al., 2020). DART, developed by the National Center for Atmospheric Research, provides a variety of EDA tools, including different filtering techniques as well as various localization and inflation options (Anderson et al., 2009). Here, we employ DART as the core assimilation engine due to its modular structure that allows integration with various forward simulators by customizing a model-specific interface while keeping the data assimilation engine of DART and the forward simulator intact. DART has been successfully linked with dozes of community codes for earth system research, such as the Weather Research and Forecasting Model (Kurzrock et al., 2019), the Community Atmosphere Model (Raeder et al., 2012), and the Community Land Model (Fox et al., 2018), to facilitate model-data integration and consequently improve model accuracy.The authors would like to thank Tim Hoar from National Center for Atmospheric Research for his insightful suggestions on improving the DART-PFLOTRAN software. This research was supported by the U.S. Department of Energy (DOE), Office of Biological and Environmental Research (BER), as part of BER's Subsurface Biogeochemical Research Program (SBR). This contribution originates from the SBR Scientific Focus Area (SFA) at the Pacific Northwest National Laboratory (PNNL) and was supported by the partnership with the IDEAS-Watersheds. This research used resources of the National Energy Research Scientific Computing Center, a DOE Office of Science User Facility supported by the Office of Science of the U.S. Department of Energy. PNNL is operated for the DOE by Battelle Memorial Institute under contract DE-AC05-76RL01830. This paper describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government. The authors would like to thank Tim Hoar from National Center for Atmospheric Research for his insightful suggestions on improving the DART-PFLOTRAN software. This research was supported by the U.S. Department of Energy (DOE), Office of Biological and Environmental Research (BER), as part of BER's Subsurface Biogeochemical Research Program (SBR). This contribution originates from the SBR Scientific Focus Area (SFA) at the Pacific Northwest National Laboratory ( PNNL ) and was supported by the partnership with the IDEAS-Watersheds. This research used resources of the National Energy Research Scientific Computing Center, a DOE Office of Science User Facility supported by the Office of Science of the U.S. Department of Energy. PNNL is operated for the DOE by Battelle Memorial Institute under contract DE-AC05-76RL01830. This paper describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government.
Keywords
- DART
- Data assimilation
- Ensemble smoother
- Inverse modeling
- PFLOTRAN
- Subsurface flow and transport