OR/14/013 Model limitations
|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.|
- Best endeavours (quality checking procedures) were employed to minimise data entry errors but given the diversity and volume of data used, it is anticipated that occasional erroneous entries will still be present (e.g. borehole location).
- The model does not reflect the full complexity of the superficial deposits geology. In reality, surfaces have been subjected to more glacitectonic deformation than is represented in the model. It is also known that made, worked and artificial ground is more widespread than is shown by the model (e.g on NS56NE as proven by boreholes), and could be subdivided into more detail than the ‘made ground’ currently used.
- Smaller rivers, streams and water bodies have not been included in the models
- The model is attributed with geotechnical and hydrogeological properties — these can be seen in the Lithoframe Viewer. These are simply bulk attributions based on point data in boreholes from Central Glasgow and are provided for general guidance only.
- The cross-section density and therefore model certainty is variable across the model, and is based on complexity and type of geology, borehole density etc.
- The NEXTMap® Digital Elevation Model was subsampled from a 5 m resolution to 50 m or 25 m resolution which means the surface distribution and geometry of a geological unit does not reflect the highest resolution possible. This resolution was chosen based on the size of the area, the resolution of the original modelling and the software capability. Some minor mismatches between geomorphological features and modelled units (including water) and the DTM occur due to the coarse resolution of the DTM.
- The NEXTMap® Digital Elevation Model may contain artefacts such as trees or artificial structures such as pylons. The majority of these have been stripped out before modelling. If any of these artefacts were found during the modelling then the effects of these were minimised in the model as much as possible.
- The start heights of boreholes used might differ significantly from the NEXTMap® Digital Elevation Model. When modelling, these differences were taken account of by assessing the year the borehole was drilled and assessing the location of the borehole against other data such as historical maps. Therefore the modeller used their own judgment in some areas if the stratigraphy in the borehole did not match the modern day topography and changes in the subsurface (quarrying, landfill etc).
- A subset of the most reliable borehole data has been included to constrain the cross-sections within the model. However there is also a large subset of borehole data that has not been included within the model.
- In the absence of borehole information, the model is constrained by map data and the DTM. Modelling highlights areas where the geological map may need revision, but this was not undertaken as part of the modelling exercise. The most common areas requiring revision relate to anthropogenic activity post-mapping (e.g. worked/made ground).
- Small parts of unit extents (envelopes) constraining the modelled volumes are relatively straight-edged in a limited number of cases e.g. witi-dmtn in the northern part of Clydebank, pais-xczs in eastern parts of Paisley, which is not geologically realistic and should be improved by future work.
- The south-east central part of the Paisley model (northern part of sheet NS55NW) incorporates some inconsistencies of up to 20 m between 1:10 000 scale artificial ground and 1:50 000 scale superficial deposits DigMapGB map data. The made ground in this area has been selectively extended to match the superficial deposits based on 1:50 000 scale DigMapGB data (BGS, 2008).
Modelled surfaces and volumes
- The thin nature of made ground, and the thin draped form of some areas e.g. Paisley Clay Member deposits, means that these units are poorly shown in visualisations of the 3D model (e.g. in the Lithoframe Viewer 3D window). A substantial number of additional cross-sections (‘helper sections’) are needed to improve the calculation of thin deposits.
- A known limitation is that for some thin units close to DTM surface and over topographically variable ground (water, MGR, LDE, LAW, KELV, GOSA, KARN, PAIS, BILL1, WITI), the superficial deposits ASCII and ESRI grids contain small patches of no modelled surface within the unit envelope (outline, supplied as an ESRI shapefile) where it should exist. The limitation is greatest in the southern Paisley, South Glasgow and South-east Glasgow block models, where topography is greatest, and the largest area affected is a maximum of 600 by 400 m. This artefact of the modelling software/procedure can be rectified with additional interpreted cross-sections or improved meshing algorithms and will be addressed in future work.
- The modelled volumes (visible in the Lithoframe Viewer 3D window) representing some elongate units such as water, made ground along road or rail embankments and alluvium are in places spiky/angular due to a combination of steep edges, DTM resolution and limited constraining cross-sections. However, the size of the angularity is in proportion to the unit and is accepted as a known limitation. Quality control by visual inspection identified additional local high or low areas (lumps or spikes) in the modelled volume. The majority of these are intentional as they result from borehole interpretation in constraining cross-sections.
- Whilst effort has been put into making models consistent and edge matched there remain minor inconsistencies between models produced and calculated separately along 5 km grid square boundaries (e.g. NS56SE/NS66SW in the vicinity of the River Clyde) and small variations in interpretive style from different modelling geologists. These inconsistencies should not be apparent unless the models are being used at high resolution.
- In the South-east Glasgow model (SW 265000, 655000 to NW 275000, 665000) preference was given to published geological map data over lithological variability recorded in borehole data, such that the borehole data may indicate more lithological variation within some units than is indicated in the unit descriptions.
- The model does not reflect the full complexity of the geology. In reality, geological surfaces will be cut with igneous intrusions/vents and cut by more faults than are modelled.
- The GOCAD® algorithm creates a triangular mesh to try and best fit all data points. For the Central Glasgow bedrock model a mesh size of around 120 m was optimal, as there are large areas without any data points. A smaller mesh size was tested but the resultant surfaces did not appear geologically realistic. In areas where there are abundant closely- spaced data (e.g. site investigations, or at a complex part of the outcrop line) a 100–200 m mesh size cannot represent the detail of the data density and complexity. In summary, the meshes are a representation of the geology, using all the data but not fitting all of it exactly. The maximum deviation between a surface and a data point is about 20 m. In the majority of cases the difference between any known data point and the modelled surface is less than 5 m.
- For the Clyde Catchment scale bedrock model, the model mesh size was between 100 to 500 m.
- All data points have been checked on data entry and for consistency in the model but there will be some errors that remain — for example boreholes whose site has been incorrectly located in the database, or whose recorded start height is wrong.
- Note that to load to ArcGIS®, the modelled TIN (triangular mesh) files have been converted to ASCII grids. A grid spacing of 25 m (Central Glasgow) or 50 m (Clyde Catchment scale bedrock model) was used so that some detail of fault gaps is preserved. This may give a false impression of the model resolution as the original TIN mesh spacing was 100–200 m. The ArcGIS® grids do not give a clean/fitted together 3D model (e.g. at fault-surface contacts) because of the TIN to grid conversion process.
- Faults with offsets less than 30 m have not been included in the Central Glasgow bedrock model because the data were insufficient to constrain them. There will be some areas where adjacent data points have significantly different Z values because they are offset by a fault that is not modelled. For the Clyde Catchment scale bedrock model, faults with offsets less than tens of metres were not included.
- Fault geometries which terminate against each other at low angles result in numerous thin ‘slivers’. These are difficult to model satisfactorily and are often poorly constrained by data (e.g. at junction of Blythswood, f38 and f39 in the Central Glasgow bedrock model). There is a small sliver of the Kiltongue Coal surface that penetrates through f38, and should not. Thus thin slivers should be treated as parts of the model with high uncertainty.
- In the Central Glasgow bedrock model f40 reverses throw along its length, as on the 10 000 scale map, but the fault dip has consistently been modelled at 60° to the north along the whole structure.
- In the Central Glasgow bedrock model the scale of faults modelled is inconsistent. Some smaller faults have been included (f14,15,6) where they constrain the outcrop of the modelled coals on NS66SW. In other areas of the model, only much larger fault structures have been included. Known, smaller faults have been excluded.
- A more recent and higher resolution model along a linear corridor covering the southwestern part of the Central Glasgow bedrock model (Monaghan, 2012b) has highlighted that the KILC, bULGS, KDG modelled surfaces are too high in the Central Glasgow Bedrock Model where they are interpreted in the hangingwall (eastern side) of the Dechmont fault.
- The two models supplied are suitable for use at different scales — Central Glasgow bedrock at 1:10 000 to 1:50 000 versus the Clyde Catchment scale bedrock model at 1:100 000 to 1:250 000. The models are not edge matched along their boundary.
- Smaller ‘holes’ in the modelled surface have not been cut in the Clyde Catchment scale bedrock model (e.g. noticeable at the edge join of the Central Glasgow and Clyde Catchment scale bedrock models on the Glasgow Ell surface).
- In the Central Glasgow bedrock model isopach maps and cross-sections through the stacked surfaces and faults highlight inconsistencies caused by lack of data, particularly on the more deeply buried parts of KDG and Base ULGS/ILS, or by patchy data coverage on one particular surface (e.g. circular inconsistencies caused by data on a particular coal seam from a particular, localised set of boreholes between GU and GE). For example, the variation in general dip between faults 37 and 38 on KILC-ILS-KDG surfaces is inconsistent. Generally the KDG and ILS modelled surfaces are much more uncertain away from their outcrop than the other modelled surfaces, due to lack of data.
- Very small areas of bedrock surfaces lie up to 2 metres higher than the rockhead model causing a local crossover. This happens in areas where there are local low points in the higher resolution mesh of the rockhead surface, and there are no TIN points within that area from the lower resolution bedrock mesh (crossovers at TIN points should have been removed).
- BRITISH GEOLOGICAL SURVEY. 2008. Digital Geological Map of Great Britain 1:50 000 scale (DiGMapGB-50) data. Version 5.18. Keyworth, Nottingham: British Geological Survey. Release date 20-05-2008.
- MONAGHAN A A. 2012b.Shieldhall bedrock model metadata report and GSI3D to GOCAD® faulted bedrock workflow. British Geological Survey Internal Report, IR/12/045. 35pp.