DecTree v1.0-chemistry speedup in reactive transport simulations
- The computational costs associated with coupled reactive transport simulations are mostly due to the chemical subsystem: replacing it with a pre-trained statistical surrogate is a promising strategy to achieve decisive speedups at the price of small accuracy losses and thus to extend the scale of problems which can be handled. We introduce a hierarchical coupling scheme in which "full-physics" equation-based geochemical simulations are partially replaced by surrogates. Errors in mass balance resulting from multivariate surrogate predictions effectively assess the accuracy of multivariate regressions at runtime: inaccurate surrogate predictions are rejected and the more expensive equation-based simulations are run instead. Gradient boosting regressors such as XGBoost, not requiring data standardization and being able to handle Tweedie distributions, proved to be a suitable emulator. Finally, we devise a surrogate approach based on geochemical knowledge, which overcomes the issue of robustness when encountering previously unseen dataThe computational costs associated with coupled reactive transport simulations are mostly due to the chemical subsystem: replacing it with a pre-trained statistical surrogate is a promising strategy to achieve decisive speedups at the price of small accuracy losses and thus to extend the scale of problems which can be handled. We introduce a hierarchical coupling scheme in which "full-physics" equation-based geochemical simulations are partially replaced by surrogates. Errors in mass balance resulting from multivariate surrogate predictions effectively assess the accuracy of multivariate regressions at runtime: inaccurate surrogate predictions are rejected and the more expensive equation-based simulations are run instead. Gradient boosting regressors such as XGBoost, not requiring data standardization and being able to handle Tweedie distributions, proved to be a suitable emulator. Finally, we devise a surrogate approach based on geochemical knowledge, which overcomes the issue of robustness when encountering previously unseen data and which can serve as a basis for further development of hybrid physics-AI modelling.…
Author details: | Marco De LuciaORCiD, Michael KühnORCiDGND |
---|---|
DOI: | https://doi.org/10.5194/gmd-14-4713-2021 |
ISSN: | 1991-959X |
ISSN: | 1991-9603 |
Title of parent work (English): | Geoscientific model development : an interactive open access journal of the European Geosciences Union |
Subtitle (English): | purely data-driven and physics-based surrogates |
Publisher: | Copernicus |
Place of publishing: | Göttingen |
Publication type: | Article |
Language: | English |
Date of first publication: | 2021/07/29 |
Publication year: | 2021 |
Release date: | 2023/11/01 |
Volume: | 14 |
Issue: | 7 |
Number of pages: | 18 |
First page: | 4713 |
Last Page: | 4730 |
Funding institution: | Helmholtz-Gemeinschaft Helmholtz Association [ZT-I-0010] |
Organizational units: | Mathematisch-Naturwissenschaftliche Fakultät / Institut für Geowissenschaften |
DDC classification: | 5 Naturwissenschaften und Mathematik / 55 Geowissenschaften, Geologie / 550 Geowissenschaften |
9 Geschichte und Geografie / 91 Geografie, Reisen / 910 Geografie, Reisen | |
Peer review: | Referiert |
Publishing method: | Open Access / Gold Open-Access |
DOAJ gelistet | |
License (German): | CC-BY - Namensnennung 4.0 International |