T262 Geologic Maps, Digital Geologic Maps, and Derivatives from Geologic and Geophysical Maps (Posters)

Wednesday, 12 October 2011
9:00 AM - 6:00 PM
Hall C

OPTIMIZING THE IMPACT OF LOCALIZED HIGH QUALITY DATA IN REGIONAL SCALE SUBSURFACE MODELLING

Paper number: T262-3
Program: Technical Sessions
Time: 9:00 AM – 6:00 PM

Authors

Abstract

The first step in creating a regional subsurface geology model is determining what pre-existing data are available for the study area. Typically a large proportion of the data comes from water well databases, which offer a significant amount of data with a good spatial distribution across the study area. However, these data are often considered to be unreliable and of variable quality and are commonly supplemented with more reliable data from localized or site specific studies. The quality of the input data has been shown to have a significant impact on the accuracy and reliability of the output model such that these localized datasets are considered highly valuable. However, the inclusion of multiple high quality data points from localized study areas often results in a clustered distribution of data points in the context of a regional study, which has been shown to negatively impact the quality of interpolated models. In order to avoid the negative impact of these clustered data, declustering algorithms are typically applied, thus potentially removing a substantial portion of the high quality data.

A new method will be presented that allows all high quality data from smaller scale studies to be incorporated into large scale regional models. The benefits of this method are; 1) all of the available high quality data are utilized in the model prediction, 2) the high resolution subsurface variability captured in smaller scale studies is incorporated into the regional model, 3) the negative impacts of clustered data are reduced, thus improving model accuracy.

The effectiveness of this method is illustrated using two regional subsurface geology models: one developed in the context of regional groundwater resource management in Georgetown, Ontario and the other developed in the context of a contamination migration study in Cottage Grove, Wisconsin. Both study areas contain small subsets of high-resolution, high quality clustered data. Significant differences are observed between the regional models produced by interpolating all the data together (typical procedure), and those models that apply the proposed new method that optimizes the use of localized, high quality, clustered data. In both cases the proposed new method yielded a model that enhanced the quality, accuracy and usefulness of the regional model.