Spatial Modeling and Uncertainty Analysis for Subsurface Feature Mapping - Integration of Geostatistical Concepts and Image-Based Machine Learning Model Validation - Thesis

Aug 4, 2026

Spatial modeling of subsurface features and uncertainty analysis is central to combining data analytics with machine learning in the petroleum industry. With energy systems evolving and new technologies emerging, accurate uncertainty assessment is essential for high-value decisions. Traditional uncertainty methods persist due to simplicity and limited pressure to update workflows, but improved procedures can better support uncertainty computation and machine-learning validation. The thesis develops multiscale geostatistical analytics to integrate datasets with different accuracies and support volumes, explicitly separating uncertainty sources and enabling more precise error estimation for cross-scale imputation. It also proposes image-model validation using minimum acceptance criteria and multi-scale SSIM to improve model checking beyond MSE and single-scale SSIM, supporting reliable deployment.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages61

Reading Time2 h 34 min

AccessFree to download

Chapter 1: Introduction

Chapter 2: Literature Review

2.1 Deconvolving Uncertainty

2.2 Image-Based Machine Learning Model Checking

Copyright by Blazej Ksiazek 2023 The Thesis Committee for Blazej Ksiazek Certifies that this is the approved version of the following Thesis: Spatial Modeling and Uncertainty Analysis for Subsurface Feature Mapping: Integration of Geostatistical Concepts and Image-Based Machine Learning Model Validation APPROVED BY SUPERVISING COMMITTEE: Michael J. Pyrcz, Supervisor Sahar Bakhshian Spatial Modeling and Uncertainty Analysis for Subsurface Feature Mapping: Integration of Geostatistical Concepts and Image-Based Machine Learning Model Validation by Blazej Ksiazek Thesis Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Master of Science in Engineering The University of Texas at Austin August 2023 Dedication This thesis is dedicated to my late friend, Ram Disabar, who was in pursuit of his Master’s Degree when he passed away, and to my family for their prayers and relentless support that made all my achievements in my master’s program possible. Acknowledgements I would like to thank my supervisor Dr. Michael Pyrcz for his immense support, contribution, and guidance throughout my academic journey; for his trust in me to join DiReCT and collaborate with all the amazing students working on groundbreaking projects. I would also like to thank Dr. Pyrcz for his feedbacks and insights in my writings, presentations, and meetings, and connecting me with a large network of academics and industry professionals. I would also like to thank Dr. Sahar Bakhshian from the Gulf Coast Carbon Center (GCCC) for her academic and financial support and for being my thesis reader. I am also thankful for the UT Austin PetroBowl team for their friendship and support over the years and their dedication to the team, allowing me to retire with a win. Abstract Spatial Modeling and Uncertainty Analysis for Subsurface Feature Mapping: Integration of Geostatistical Concepts and Image-Based Machine Learning Model Validation Blazej Ksiazek, M.S.E. The University of Texas at Austin, 2023 Supervisor: Michael Pyrcz Spatial modeling of subsurface features and uncertainty analysis plays a pivotal role in the integration of data analytics and machine learning techniques in the petroleum industry. As the energy landscape is always changing, and new technologies are emerging, the demand for accurate assessments of uncertainty to inform high-value decision-making is of utmost importance. Nonetheless, the same longstanding methods are used due to their simplicity and the lack of immediate necessity for change. However, with improvements and the implementation of proper workflows, the current methods for calculating uncertainties and validating machine learning models can be more effectively addressed. We developed multiscale methods for data analytics and machine learning. These approaches integrate geostatistical concepts to enhance the precision and reliability of subsurface modeling techniques. We address the challenge of integrating multiple datasets with varying accuracies and volume support sizes. We emphasize the importance of accounting for different sources of uncertainty in spatial modeling workflows. Leveraging geostatistical concepts, such as semivariograms, and dispersion variance, a novel approach is introduced to calculate a more precise measure of error when imputing smaller scale datasets with larger scale datasets. This refined measure of error allows for the direct integration of these datasets in spatial modeling workflows. Once all the uncertainty in our models is accounted for, we must check if our models are accurate. Therefore, we focus on the validation of machine learning models, particularly those tailored for image data. Image-based models often necessitate preprocessing steps, such as resizing and augmentation, to improve data quality for training. To ensure the performance and suitability

FAQs

What problem does the thesis address in subsurface feature mapping?

It addresses how to model subsurface features while quantifying uncertainty, then integrating this with machine learning workflows for more accurate and reliable mapping and decision-making.

How does the thesis incorporate geostatistical ideas into multiscale uncertainty computation?

It develops multiscale methods that integrate geostatistical concepts such as semivariograms and dispersion variance to compute a more precise error measure when imputing smaller-scale data using larger-scale datasets.

How are image-based machine learning models validated in the thesis?

The thesis proposes validation via minimum acceptance criteria combined with the multi-scale Structural Similarity Index (MS-SSIM), improving model checking of reconstructed and predicted images compared with approaches like MSE and single-scale SSIM.

Aug 4, 2026

Upload by Ava Thompson