Development of a city-wide physical property model of the greater Glasgow area using a voxel-based representation of lithology

The BGS PropBase project has sought to develop a physical property model
of a major UK city and to identify potential aquifer sequences in shallow
heterogeneous lithostratigraphic units. As a precursor to this, new
methodologies have been applied to develop a city-scale lithology model of
shallow, unconsolidated sediments in Glasgow, Scotland.
The model has been created by developing a stochastic voxel representation
of clastic geology, based on upscaling of observed borehole lithology,
independent of lithostratigraphy. Assessments of uncertainty are made using
multiple realisations of lithology and using these to compute the probability of
one of a limited set of lithologies existing in any voxel. Although similar to work
previously undertaken in the Netherlands by TNO, this model uses a borehole
dataset which has not been cleaned and refined, nor has it been substantially
reinterpreted to improve resolution.
Results have been compared with existing lithostratigraphic models of the
Greater Glasgow area created using the BGS GSI3D framework modelling
methodology. In a dense borehole field, this demonstrated that voxel
modelling may readily respect a larger proportion of the boreholes within the
modelling area.
Comparison of the voxel models with lithostratigraphic surfaces from the
previous GSI3D models has resulted in a quality assurance process that
allows assessment of the quality of the framework model in terms of
representing subsurface lithologies.
Subsequently, physical property data were upscaled into the voxel model on a
per-lithology basis. These parameterised grids may be integrated into BGS
ZOOM groundwater models; this project is thereby at the confluence of four
different modelling approaches:
(i) geological framework modelling,
(ii) voxel modelling,
(iii) property modelling, and
(iv)input grids for groundwater modelling.
The techniques applied in the course of the project have the potential to be
regularly used in developing future BGS model outputs and to be incorporated
as standard tools of the BGS modelling workflow.