- Author
- Robin Thibaut (UGent)
- Promoter
- Thomas Hermans (UGent) and Eric Laloy
- Organization
- Project
- Abstract
- Accurate modeling of the subsurface, a complex and heterogeneous environment that plays a crucial role in the Earth’s water cycle, is challenging due to sparse and incom- plete data. We can reduce the uncertainty associated with subsurface predictions, such as groundwater flow and contaminant transport, by conducting additional observations and measurements in the subsurface. However, practical and economic considerations frequently limit the number of measurements and their locations, such as land occupa- tion, which may limit the number of wells that can be drilled. In this dissertation, we propose simulation-driven methods to reduce uncertainty in subsurface predictions by identifying the most informative data sets to gather. Our method, which is based on Bayesian optimal experimental design and machine learning, determines the nature and location of these data sets, which can include measurements of groundwater levels, tem- perature, and other parameters collected through active or passive sensing methods such as pumping tests, tracer tests, and geophysical surveys. This dissertation is the first to use Bayesian Evidential Learning (BEL) for optimal experimental design, allowing for the optimization of data source locations and the comparison of the utility of different data sources. BEL is a framework for prediction that combines Monte Carlo sampling and machine learning in order to learn a direct relationship between predictor and target variables generated by a simulation model. We demonstrate the efficacy of our methods in three groundwater modeling case studies: (i) wellhead protection area delineation, (ii) an aquifer thermal energy storage monitoring system, and (iii) groundwater-surface water interaction. The case studies show that our approach can significantly reduce the uncertainty in subsurface predictions and guide further subsurface exploration. The first case study, in particular, uses the Traveling Salesman Problem to introduce a novel ap- proach to wellhead protection area delineation. The second case study, which compares well and geophysical data for temperature monitoring, introduces a new method for com- bining observations from multiple data sources in a latent space of the original data. The third case study introduces the Probabilistic Bayesian neural network (PBNN) method to BEL and transitions from a static experimental design framework to a sequential experimental design framework, which estimates groundwater-surface water interaction fluxes from temperature data. We have also developed a Python package, SKBEL, that implements our methods and can be used for a variety of Earth Science applications. Overall, this dissertation demonstrates the utility of BEL for optimal experimental de- sign in groundwater modeling, highlights the potential of BEL for predictive modeling in Earth Sciences, and opens up new avenues for data and simulation-driven subsurface modeling.
- "To consult the statistician after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment died of." Ronald Fisher. Het nauwkeurig modelleren van de ondergrond, een complexe en heterogene omgeving die een cruciale rol speelt in de watercyclus van de Aarde, is uitdagend vanwege de schaarse en onvolledige gegevens. Door extra waarnemingen en metingen in de ondergrond te verrichten, kunnen we de onzekerheid die verbonden is aan voorspellingen verminderen. Maar praktische en economische drijfveren beperken vaak het aantal metingen en de locaties van de metingen. In deze thesis stellen we simulatie-gedreven methoden voor om onzekerheid in voorspellingen van de ondergrond te verminderen door de meest informatieve datasets te identificeren. Onze aanpak, gebaseerd op Bayesiaanse optimale experimentele ontwerp machine learning, bepaalt de aard en locatie van deze datasets. We laten de effectiviteit van onze methoden zien door drie case studies in grondwatermodellering: het afbakenen van een beschermingsgebied rond putten, een monitoringsysteem voor thermische energieopslag in watervoerende lagen en de interactie tussen grondwater en oppervlaktewater. Onze resultaten laten zien dat deze methodologieën de nauwkeurigheid en precisie van voorspellingen met betrekking tot de ondergrond kunnen verbeteren. En we bespreken hun prestaties en de mogelijkheden voor toekomstig onderzoek.
- Keywords
- machine learning, hydrology, hydrogeology, experimental design
Downloads
-
(...).pdf
- full text (Published version)
- |
- UGent only
- |
- |
- 4.78 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-01H1XP5NTC07A6ZX3MNPMNJYQ4
- MLA
- Thibaut, Robin. Machine Learning for Bayesian Experimental Design in the Subsurface. Ghent University. Faculty of Sciences, 2023, doi:10.22541/essoar.168057575.58936022/v1.
- APA
- Thibaut, R. (2023). Machine learning for Bayesian experimental design in the subsurface (Ghent University. Faculty of Sciences). https://doi.org/10.22541/essoar.168057575.58936022/v1
- Chicago author-date
- Thibaut, Robin. 2023. “Machine Learning for Bayesian Experimental Design in the Subsurface.” Ghent, Belgium: Ghent University. Faculty of Sciences. https://doi.org/10.22541/essoar.168057575.58936022/v1.
- Chicago author-date (all authors)
- Thibaut, Robin. 2023. “Machine Learning for Bayesian Experimental Design in the Subsurface.” Ghent, Belgium: Ghent University. Faculty of Sciences. doi:10.22541/essoar.168057575.58936022/v1.
- Vancouver
- 1.Thibaut R. Machine learning for Bayesian experimental design in the subsurface. [Ghent, Belgium]: Ghent University. Faculty of Sciences; 2023.
- IEEE
- [1]R. Thibaut, “Machine learning for Bayesian experimental design in the subsurface,” Ghent University. Faculty of Sciences, Ghent, Belgium, 2023.
@phdthesis{01H1XP5NTC07A6ZX3MNPMNJYQ4,
abstract = {{Accurate modeling of the subsurface, a complex and heterogeneous environment that plays a crucial role in the Earth’s water cycle, is challenging due to sparse and incom- plete data. We can reduce the uncertainty associated with subsurface predictions, such as groundwater flow and contaminant transport, by conducting additional observations and measurements in the subsurface. However, practical and economic considerations frequently limit the number of measurements and their locations, such as land occupa- tion, which may limit the number of wells that can be drilled. In this dissertation, we propose simulation-driven methods to reduce uncertainty in subsurface predictions by identifying the most informative data sets to gather. Our method, which is based on Bayesian optimal experimental design and machine learning, determines the nature and location of these data sets, which can include measurements of groundwater levels, tem- perature, and other parameters collected through active or passive sensing methods such as pumping tests, tracer tests, and geophysical surveys. This dissertation is the first to use Bayesian Evidential Learning (BEL) for optimal experimental design, allowing for the optimization of data source locations and the comparison of the utility of different data sources. BEL is a framework for prediction that combines Monte Carlo sampling and machine learning in order to learn a direct relationship between predictor and target variables generated by a simulation model. We demonstrate the efficacy of our methods in three groundwater modeling case studies: (i) wellhead protection area delineation, (ii) an aquifer thermal energy storage monitoring system, and (iii) groundwater-surface water interaction. The case studies show that our approach can significantly reduce the uncertainty in subsurface predictions and guide further subsurface exploration. The first case study, in particular, uses the Traveling Salesman Problem to introduce a novel ap- proach to wellhead protection area delineation. The second case study, which compares well and geophysical data for temperature monitoring, introduces a new method for com- bining observations from multiple data sources in a latent space of the original data. The third case study introduces the Probabilistic Bayesian neural network (PBNN) method to BEL and transitions from a static experimental design framework to a sequential experimental design framework, which estimates groundwater-surface water interaction fluxes from temperature data. We have also developed a Python package, SKBEL, that implements our methods and can be used for a variety of Earth Science applications. Overall, this dissertation demonstrates the utility of BEL for optimal experimental de- sign in groundwater modeling, highlights the potential of BEL for predictive modeling in Earth Sciences, and opens up new avenues for data and simulation-driven subsurface modeling.}},
author = {{Thibaut, Robin}},
keywords = {{machine learning,hydrology,hydrogeology,experimental design}},
language = {{eng}},
pages = {{XII, 278}},
publisher = {{Ghent University. Faculty of Sciences}},
school = {{Ghent University}},
title = {{Machine learning for Bayesian experimental design in the subsurface}},
url = {{http://doi.org/10.22541/essoar.168057575.58936022/v1}},
year = {{2023}},
}
- Altmetric
- View in Altmetric