A national-scale concept for quantifying and utilizing interpretation uncertainties in hydrogeological layer models

3D hydrostratigraphic models provide the foundation for hydrological modelling. Traditionally, many hydrostratigraphic models have been constructed as layer models, created by interpreting available geoinformation (boreholes, geophysics, outcrops etc.). This procedure results in a single deterministic model that lacks information of uncertainties. The uncertainties in the hydrostratigraphic domain have therefore not been considered when using the models for subsequent hydrological modelling. To include the uncertainties in hydrological modelling, a series of realizations is required to express the uncertainty due to the non-linear and complex forward relationship between the geological and hydrological domain. In this study, we have assigned and modelled the uncertainties of the national hydrostratigraphic model of Denmark, which have been developed by several modelers over a long period of time. In order to do this, we have developed a concept for quantifying the uncertainties based on multiple factors, including; i) the availability of geoinformation (boreholes and geophysics) at the time of model creation, ii) the geological complexity and iii) the overall validity of the original modelling. The systematic concept offers a transparent solution for quantifying the interpretation uncertainty that can be carried out even on a national scale. The quantified uncertainties are assigned to a pointwise Gaussian distribution centered around the layer boundary. This creates a series of variance maps that makes it possible to define a Gaussian likelihood function to evaluate in a probabilistic framework. Geostatistical modelling using a localized kriging routine is used to form an ensemble of realizations of the national model, which is discretized in a 100 m x 100 m grid, with the purpose of propagating the hydrostratigraphic uncertainty into the hydrological domain. The proposed method provides a valuable tool for the modeler to quantify and visualize uncertainties of layer models and to revisit and assign uncertainties to existing models.