OPTIMIZING THE OUTPUT OF 3-DIMENSIONAL SUBSURFACE MODELS USING A NEW ALGORITHM TO INTEGRATE DATA OF VARIABLE QUALITY

The accuracy and reliability of 3-dimensonal (3-D) subsurface geologic models is highly dependent upon the quality of the input data. Unfortunately, in many regions ‘high quality’ data providing accurate and reliable information about subsurface conditions are sparse and/or localized, therefore subsurface studies have to rely on large, regionally extensive digital databases (such as waterwell databases) to provide adequate data coverage. A significant drawback of using waterwell databases (WWD) is the extreme variation in both the quality and reliability of the data they contain. Algorithms used in conventional 3-D modeling programs do not discriminate between ‘high’ and ‘low’ quality data and model outputs can be severely compromised by the equal treatment of all available data in the modeling process.

A new algorithm for interpolating data of varying quality is presented here which allows data identified as being ‘high quality’ to take precedence in the spatial interpolation process. Prior to modelling, input data are classified according to their accuracy and reliability into ‘high quality’ and ‘low quality’ groupings. The new modeling algorithm then applies a ‘quality weighting’ factor to the input data, that allows the high quality data to have a greater influence over the interpolation of unit boundaries during the modeling process while still utilizing the enhanced spatial distribution of the lower quality data to expand the model regionally. The influence of the quality weighting factor on the data can be varied by the user according to their confidence and understanding of the available data.

The impact of input data quality on 3-D model outputs and the effectiveness of this new ‘quality weighted’ algorithm are illustrated with subsurface models of the 16 Mile Creek region of Georgetown, Ontario. New ‘high quality’ core data as well as WWD are available for this region. Initial results show that output models generated using the new algorithm are significantly more reliable than those generated with non-weighted algorithms. The new ‘quality weighted’ algorithm optimizes the strengths of both sparse and/or localized high quality data sources, as well as the more regionally extensive data lower quality data.