OR/14/013 Model uncertainty

From Earthwise
Jump to navigation Jump to search
Monaghan A A, Arkley S L B, Whitbread K, McCormac M. 2014. Clyde superficial deposits and bedrock models released to the ASK Network 2014: a guide for users Version 3 . British Geological Survey Internal Report, OR/14/013.
Figure 9    a) Array of cross-section lines used to constrain the five superficial deposits models. b) Cross-section array with boreholes used in the modelling, and all boreholes shown.

The cross-sections and borehole data points constraining the Superficial Deposits Models shown in Figure 9 give an indication of the most certain areas of the model containing the most sections and borehole data (e.g. Central Glasgow superficial deposits model), and the least certain areas (southwest of Paisley model, South and North Glasgow superficial deposits models). Note also on Figure 9 the extent of the Clydebank and South Glasgow models, which do not cover the whole map sheet. A more quantitative approach (described below) is also used in BGS to indicate model uncertainty. In due course, uncertainty layers will be calculated for the whole area of the superficial deposits models.

Currently, ASK Network users are supplied with uncertainty layers from an earlier (2009) version of the 3D geological model for evaluation purposes. The uncertainty layers cover 3 quarter sheets of the Central Glasgow models (NS66SW, NS66NW and NS56SE). Comments on the utility of these uncertainty layers would be welcomed.

The uncertainty layers are supplied in two formats (1) .jpg with .jgw raster image files and (2) ASC files, which can be converted to ArcGIS® grids, for example, using ArcToolBox (‘conversion tools’ — ‘ASCII to raster’). When viewing the converted ASC files, the user needs to colour up the grids based on the ‘count’ field, which is the relative value of uncertainty for that layer of the model. The largest number represents relative high uncertainty and the smallest number represents relative low uncertainty.

Uncertainty layers were calculated from a combination of data density and geological complexity of the modelled surface. The BGS confidence calculator v1_2 — customised BGS software developed in Matlab — was used. This means the model will be most uncertain where there is little data and where the geological surface dip changes rapidly.

For the superficial deposits model, relative low uncertainty might be considered to have errors of the order of ± 10 m in XYZ, for example, those areas coloured green on Figure 10. In the most uncertain areas where the geology is complex and poorly constrained by borehole data, uncertainty may be approximately ± 70 m in XYZ for example, those areas coloured red on Figure 10.

Figure 10    Uncertainty layer for the Wilderness Till coloured up such that the relatively most uncertain areas are red, relatively least uncertain areas are green (covers NS66SW, NS66NW and NS56SE).

Figure 11    Example of combined uncertainty attribute for KILC draped on the geological surface in GOCAD®.

For the bedrock surfaces, lowest uncertainty (highest confidence) areas are those that are well constrained by geological data and where the geology is relatively simple. In these areas, the error on the model might be considered to be of the order of ± 10 m in XYZ; for example, those areas of the KILC uncertainty surface on Figure 11 that are purple or pink.

Highest uncertainty (lowest confidence) areas are areas that are not constrained by any geological data and where the geology is complex i.e. faulted or folded. In these areas, the error on the model might be considered to be of the order of ± 70 m in XYZ; for example those areas of the KILC uncertainty surface on Figure 11 that are orange or red.