A recently released voxel model quantifying aggregate resources of the Belgian part of the North Sea includes lithological properties of all Quaternary sediments and modelling-related uncertainty. As the underlying borehole data come from various sources and cover a long time-span, data-related uncertainties should be accounted for as well. Applying a tiered data-uncertainty assessment to a composite lithology dataset with uniform, standardized lithological descriptions and rigorously completed metadata fields, uncertainties were qualified and quantified for positioning, sampling and vintage. The uncertainty on horizontal positioning combines navigational errors, on-board and off-deck offsets and underwater drift. Sampling-gear uncertainty evaluates the suitability of each instrument in terms of its efficiency of sediment yield per lithological class. Vintage uncertainty provides a likelihood of temporal change since the moment of sampling, using the mobility of fine-scale bedforms as an indicator. For each uncertainty component, quality flags from 1 (very uncertain) to 5 (very certain) were defined and converted into corresponding uncertainty percentages meeting the input requirements of the voxel model. Obviously, an uncertainty-based data selection procedure, aimed at improving the confidence of data products, reduces data density. Whether or not this density reduction is detrimental to the spatial coverage of data products, will depend on their intended use. At the very least, demonstrable reductions in spatial coverage will help to highlight the need for future data acquisition and to optimize survey plans. By opening up our subsurface model with associated data uncertainties in a public decision support application, policy makers and other end users are better able to visualize overall confidence and identify areas with insufficient coverage meeting their needs. Having to work with a borehole dataset that is increasingly limited with depth below the seabed, engineering geologists and geospatial analysts in particular will profit from a better visualization of data-related uncertainty. |
