T217 Geological Mapping: Key to Successful Management of Water and Land Resources

Tuesday, 11 October 2011
1:30 PM - 5:30 PM
Room 101DE

QUANTIFYING UNCERTAINTY OF GEOLOGICAL AND GEOHYDROLOGICAL MODELS

Paper number: T217-9
Program: Technical Sessions
Time: 3:30 PM – 3:45 PM

Authors

Abstract

The Geological Survey of the Netherlands produces 3D geological and geohydrological models of the subsurface, as part of the national mapping program and for specific projects. There is an increasing demand among our user-community to quantify the uncertainty introduced by using data of multiple sources and varying data quality, in combination with interpolation uncertainty. The data include (amongst other data sources) borehole descriptions of varying quality, geophysical data and existing mapping products like 2D geological maps. To obtain sound geologically sound models, we think that including knowledge of the geological processes is a prerequisite and adds to the complexity of uncertainty quantification. The uncertainties of individual data sources and geological a-priori knowledge is often hard to quantify, and frequently includes subjective choices of the geologist. An example is the stratigraphic labeling of borehole descriptions.

We developed a method to quantify the model uncertainty of 3D layer models using a cross-validation technique, combined with the spatial correlation properties of the data. This results in a standard deviation around the modeled surface layer that can be used in further analysis and applied modeling. An example is the use of uncertainty in determining the heat storage capacity of the subsurface, which depends (a.o) on the thickness of the aquifer.

The uncertainty of 3D voxel models (characterized by discretisizing the subsurface into small cubes of e.g. 100x100x0.5m) is quantified using geostatistical simulation methods, employing Monte-Carlo type techniques. Realizations of equal-probable distributions of e.g. lithology can then be used in subsequent analysis. As an example, the analysis of highway roughness, as a consequence of lithological variation in the subsoil is presented, in which the uncertainty of the lithology is used to obtain confidence intervals around a pre-set roughness value.