Integration of Multiple Field Methods in Characterizing a Field Site with Bayesian Inverse Modeling

A hydraulic property of a field can be expressed as a space random function (SRF), and the parameters of that SRF can be constrained by the Method of Anchored Distributions (MAD). MAD is a general Bayesian inverse modeling technique that quantifies the uncertainty of SRF parameters by integrating various direct local data along with indirect non-local data. An example is given with a high-resolution 3D aquifer analog with known hydraulic conductivity (K) and porosity (n) at every location. MAD is applied using different combinations of simulated measurements of K, n, and different scales of hydraulic head that represent different field methods. The ln(K) and n SRF parameters are characterized with each of the method combinations to assess the influence of the methods on the SRFs and their implications. The forward modeling equations are solved by the numerical modeling software OpenGeoSys (opengeosys.org) and MAD is applied with the software MAD# (mad.codeplex.com). The inverse modeling results are compared to the aquifer analog for success evaluation. The goal of the study is to show how integrating combinations of multi-scale and multi-type measurements from the field via MAD can be used to reduce the uncertainty in field-scale SRFs, as well as point values, of hydraulic properties.