Workbench Hydro Structural Modelling module (HSM)
Create hydrostratigraphic models with a semi-automated workflow that combines borehole lithologies and resistivity information derived from geophysical data.
Predictions made from groundwater models are very dependent on the uncertainty of the model inputs. Hence, the input variable uncertainties need to be carefully described to achieve predictions with relevant uncertainty spans.
One major cause of uncertainty in the predictions comes from uncertain knowledge about the subsurface structures, the hydrostratigraphy. The Hydro Structural Modelling (HSM) module is a transparent, objective, and data-driven workflow that creates exportable cluster models that can be used to create an ensemble of equally probable hydrostratigraphic models for groundwater modelling.
When used in conjunction with Aarhus Workbench, this tool creates a new data driven workflow of using geophysical resistivity models and lithology logs as input to hydrogeological modelling.
The HSM module creates hydrostratigraphic models with a 2-step, semi-automated workflow that combines resistivity information from geophysical data and borehole lithologies.
The first step uses Accumulated Clay Thickness (ACT) modelling to create a clay fraction model.
The second step is a clustering routine that creates a zonated model, which is the hydrostratigraphic model when assigning hydrological parameters to the zones.
The cluster model can be exported from the HSM tool and imported into third-party software, in which a third step can be performed using Multiple-Point Statistics (MPS) to create an ensemble of equally probable models, quantifying uncertainties of the hydrostratigraphic models.
Note that the HSM module in Workbench includes steps one and two, ACT modelling and clustering.
In sedimentary environments, a general assumption is that low resistivities derived from geophysical data mainly correspond to clay or clay-rich sediments (aquitards). High resistivities mainly correspond to potential aquifer lithologies such as sand, gravel, or chalk. This general link is utilized by the Accumulated Clay Thickness (ACT) concept, linking the geophysical data and the borehole information to build a combined clay thickness (or clay fraction) model.
First, the available lithological borehole logs are divided into aquifers and aquitards (for example, sand and clay). Then a 3D model grid covering the area of interest is defined. A translator function is defined on each node in this model grid, which links resistivities and clay fraction. The translator function is described by two parameters – an upper and a lower resistivity value. Resistivity layers below the upper value will get a weight of 1 which means that the entire length of the resistivity layer is presumed to be clay. In contrast, resistivity layers above the lower value will get a weight of 0, corresponding to no‐clay content (sand) for this resistivity layer. Mixed layers exist for values between the upper and lower value. By inversion, we find the set of parameters in the translator model (upper and lower) that best fit the borehole-derived clay fractions and the geophysical predicted clay fractions.
A vital aspect of this concept is that the translator function can change horizontally and vertically, adapting to the local conditions and borehole lithologies. Therefore, not one “global” translator function is used for an entire survey, but a translator function that is spatially varying on the 3D model grid. The result is a 3D clay fraction model.
The second step combines the clay fraction values from the ACT model and the geophysical resistivities in a k-mean clustering routine. As the clay fraction and resistivity models are correlated, the k-mean analysis is done on their principal components (PCA) to obtain uncorrelated variables. This produces a model reduced to several zonated clusters (typically 4-6), which can be used as hydrostratigraphic units in groundwater modelling when assigned relevant hydrological parameters.
Multiple-Point Statistics (MPS)
The cluster model can be exported from Workbench and used as a training image in an MPS tool, creating equally probable models. The uncertainty of the cluster model can be estimated from the model realisations and used in the groundwater modelling, or the entire ensemble of models can be used individually.
Bridging the Gap Between Geophysics and Hydrogeological Modelling
The certainty of groundwater predictions based on hydrogeological models is highly dependent on the available model input information, which is often sparse. This new method uses geophysical resistivity results, linked with borehole lithology, as model input. Through this data-driven workflow, uncertainty in hydrogeological structures can be more accurately estimated and groundwater prediction uncertainty can be minimized.