Inverse Groundwater Flow Modelling Via Conditional Simulation

Abstract:

The goal of inverse modelling is to estimate variables from their observations and observations of other variables which are coupled to the variable of interest via non-linear processes. The aim is to obtain fields:

1. with prescribed spatial variability
2. with observed values of the variable of interest at the observation locations (maybe at different spatial scales)
3. with observations coupled through the model.

For this purpose a new conditional MC simulation method was developed. This method uses a high dimensional geometric concept to generate conditional random fields as a weighted sum of unconditional fields. A key feature of this technique is that a connected domain of fields fulfilling the first and the second condition can be generated. This domain has a simple continuous parametrization and can easily be enlarged.
The idea of the inverse modelling approach is to generate fields that fulfill the first and the second conditions so that these fields form a connected domain which has a continuous parametrization. Then the third condition can be handled by optimization inside the above described connected domain. If no sufficient solution can be obtained the dimensionality of the problem is increased by enlarging the continuous domain and the optimization is continued.
To illustrate the performance a comparison to a published benchmark study is carried out.