Quantifying Uncertainty in Subsurface Systems

Céline Scheidt is senior research engineer at Stanford University with 10 years of experience in this field. She is known for her work on uncertainty quantification using machine learning methods and has published several impactful papers in that area. She will be the keynote speaker of the next international Geostatistics congress.Lewis Li is 3rd year PhD student at Stanford University. He has published three papers, with three more in the pipeline. With an Electrical Engineering degree from Stanford University, he has considerable expertise in software engineering and in addressing computational challenges.Jef Caers is a world-leading expert in quantifying uncertainty in the subsurface, has closely worked on 100+ projects with a variety of industries in this area and has been leading the Stanford Center for Reservoir Forecasting for 15 years, he has been Professor at Stanford University for 19 years.

• Preface viiAuthors xi1. The Earth Resources Challenge 12. Decision Making Under Uncertainty 293. Data Science for Uncertainty Quantification 454. Sensitivity Analysis 1075. Bayesianism 1296. Geological Priors and Inversion 1557. Bayesian Evidential Learning 1938. Quantifying Uncertainty in Subsurface Systems 2179. Software and Implementation 26310. Outlook 267Index 273

Reviews, The Leading Edge, SEG, May 2020The subsurface medium created by geologic processes is not always well understood. The data we collect in an attempt to characterize the subsurface can be incomplete and inaccurate. However, if we understand the uncertainty of our data and the models we generate from them, we can make better decisions regarding the management of subsurface resources. Modeling and managing subsurface resources, and properly characterizing and understanding the uncertainties, requires the integration of a variety of scientific and engineering disciplines.Five case studies are outlined in the introductory chapter, which are used to demonstrate various methods throughout the book. The second chapter introduces the basic notions in decision analysis. Uncertainty quantification is only relevant within the decision framework used. Models alone do not quantify uncertainty, but do allow the determination of key variables that influence models and decisions. Next, an overview of the various data science methods relevant to uncertainty quantification in the subsurface is provided. Sensitivity analysis is then covered, specifically Monte Carlo-based sensitivity analysis. The next three chapters develop the Bayesian approach to uncertainty quantification, and this is the focus of the book.All of this is brought together in Chapter 8, which describes a solution regarding quantifying the uncertainties for each of the problems presented in the first chapter. The authors admit that it is not the only solution. No single solution fits all problems of uncertainty quantification. The results in this chapter allow the reader to see the previously described methods applied and how choices influence models and decisions. The final two chapters discuss various software components necessary to implement the strategies presented in the book and challenges faced in the future of uncertainty quantification.The book uses a number of relevant subsurface problems to explore the various aspects of uncertainty quantification. Understanding uncertainty, and how it affects modeling and decision outcomes, is not always straightforward. However, it is necessary in order to make good, consistent decisions. The book is not an easy read. Some portions require good mathematical understanding of the underlying principles. However, the book is well documented and organized. I would say that is not a good book for a beginner, but it is a good resource for someone to get a grounding to go further into the subject. I appreciate the authors putting together this book on a complex problem that is important to our industry.—David Bartel, Houston, Texas