Multiple Point Statistics – A Case Study from Indian Wells Valley, California

Abstract:

The importance of handling uncertainty of data and being able to present the ambiguity of Geo-models of any kind have got more and more attention the last couple of years. In the same way as it is important for a Geo-modeler to understand the uncertainty and limitations of data to make an adequate geo-model, it is important for a decision maker to understand the uncertainty and limitations of a Geo-model to perform adequate and qualified decisions. Whether the geo-model is made to manage infrastructure projects or nature resources like oil, gas, or groundwater, being able to understand the uncertainty and limitations of a geo-model potentially have great economic value.

One way of presenting the uncertainty and ambiguity of a geo-model is to present a suite of geo-models instead of just one. Due to the underdetermined nature of the inverse problem to be solve (making a geo-model is an inverse problem), several solutions will all fit the available data and information about the problem. Multiple-Point Statistics (MPS) is an overall methodology representing a series of simulation algorithms all utilizing the available information (geophysics, boreholes, background knowledge, etc.) to produce a series of geo-model realizations all fitting the available information. Generating many of these realizations allow for any kind of statistical computations to be made: “What is the probability of having sand at a specific location and depth?”, “What is the probability that location A and B are placed in a fully connected clay layer?”, etc.

In this presentation we will present the workflow and results of a case study from the Indian Wells Valley in California. The case study is made as part of the Stanford Groundwater Architect Project (GAP) and the MPS-models are made by using a combination of SkyTem data, borehole information and a training image, i.e. a conceptual model, of the aquifer-aquitard system of the valley.