Author: Noah Daniel Athens

Publisher:

Published: 2021

Total Pages:

ISBN-13:

Development of new fault-controlled geothermal systems is stymied by high degrees of geologic uncertainty and the expense of drilling exploration wells. These systems, which rely on the circulation of fluids along normal faults, are often hidden by sedimentary overburden, the presence of a cold-water aquifer, or mineralization of fluid pathways. While exploration studies typically focus on data acquisition and geologic interpretation to predict locations with thermal fluids and high permeability, uncertainties in the geological modeling process are often neglected leading to an undercounting of development risk. This thesis addresses this deficit by providing practical methods to quantify uncertainty in the exploration stage for geothermal resources. Temperature wells in extensional geothermal basins show a staggering variability of geothermal gradient values, contributing to large development risk associated with well placement. In Chapter 1, a comprehensive Bayesian framework is proposed for predicting subsurface temperature in a geothermal basin given large uncertainties on geologic attributes such as the permeability field, basin geometry, and the presence of intra-basin faults. Using a synthetic case study problem based on the Dixie Valley geothermal field in central Nevada, the proposed method is applied to predict temperature in the basin constrained to a temperature well. Sensitivity analysis on prediction variables show that the bulk permeability and basal heat flux are the most important parameters for controlling the geothermal gradient at a proposed well location. Chapter 2 focuses on how to specify uncertainty of faults and basin geometry in a prior model constrained to geophysical data, an important challenge to properly implement the Bayesian framework proposed in Chapter 1. Using gravity data as an example geophysical method, 2D stochastic inversion is combined with kinematic structural models to directly incorporate structural uncertainty into the inversion process. The sampling approach uses Monte Carlo simulation to generate geologically realistic model realizations and the Gradual Deformation Method to further refine models to match observed data. The approach is first validated on a synthetic example before applying the method to field-observed gravity data from Dixie Valley, Nevada. Results of the inversion compare favorably with a previously published forward model but provide fault probability and density statistics derived from the ensemble of posterior models to quantify model uncertainty. Chapter 3 investigates the impact of data uncertainty on gravity inversion, which is often overlooked in conventional gravity modeling approaches that rely on a single interpolated gravity field. To address data uncertainty due to irregularly spaced gravity measurements, realizations of the gravity field in Dixie Valley are generated by geostatistical simulation and independently inverted to show how inversion results are affected by sparse data sampling and interpolation. Inversion is performed using a pseudo 3D approach in which subparallel profiles are inverted using the 2D inversion approach presented in Chapter 2, and model attributes are interpolated in depth using geostatistical co-simulation. To encourage continuity of structural features between neighboring profiles, model parameters are sequentially coupled during inversion. The results document a marked impact of both data and model uncertainty on depth-to-basement and distance-to-fault maps.