Estimating the Data Reliability of Magnetotelluric Measurements
Author: D. Keith Smithson
Year: 2017
Supervisors: Hlynur Stefánsson, Egill Júlíusson & Samuel Perkin
Abstract
An overview of risk in geothermal drilling and value of information analysis is presented with a description of state-of-the art inversion methods used to process geophysical data. Assumptions about the treatment of error and correlation between fitness and likelihood are challenged. Iterative Complexity Addition (ICA) is a novel algorithm proposed to test the hypothesis of these assumptions and provide information about the data reliability of solutions returned from an underdetermined inverse problem. The algorithm is applied to the inversion of magnetotelluric (MT) data from four synthetic models and existing data from the Þeistareykir geothermal field in Northeast Iceland. The results indicate that there is not a strong correlation between fitness and likelihood. Taking the best-fit model as a solution yields an average likelihood of 48.49% while ICA's selection of most-likely solution yields an average likelihood of 63.59% when considering the total depth of the model. When limiting the scope of interest to a typically drilling range of 3km depth, the best-fit likelihood is shown to be 53.77% while ICA's most-likely solution has a data reliability of 68.71%. An improvement in data reliability can be manifested as improvement in drilling success rates. The algorithm design is described with a discussion of algorithm strengths, weaknesses, and potential improvements.