TESTING SPATIALLY FLEXIBLE BOTTOM BOUNDARY PARAMETER SCHEMES AND PRIORS FOR GEOTHERMAL RESERVOIR MODELS

Abstract

Geothermal reservoir models require suitable bottom boundary conditions to account for the influence of deep geothermal energy sources. When applying mass- and heat-flux boundary conditions to a model, it is standard practice to assign mass upflows to fixed locations on the bottom boundary. That is, during automatic model calibration, each boundary mass-flux region does not change its position, area, or shape. This rigidity in boundary source locations may limit 1) how well automatic calibration methods can match field data, and 2) how well model uncertainty is quantified. Here we consider more flexible parameterisation schemes and parameter priors to address those issues. The parameterisation schemes we use are based on sampling from multivariate Gaussian prior parameter distributions and applying thresholding to generate boundary regions that act as sources of geothermal fluid. Unlike the standard approach, the schemes proposed here allow upflow regions to appear on parts of the bottom boundary in keeping with field observations. This spatial flexibility not only enables upflow regions to appear or grow in order to better match observations but, in the context of uncertainty quantification, also allows the existence, geometry, and extent of upflow regions to vary between posterior sampled models. We demonstrate our parameterisation methods by applying them to uncertainty quantification of synthetic test models, including a model which is based on the Montserrat geothermal field.