GEOSTATISTICALLY TUNED HYDROGEOLOGICAL 3D MODELLING OF COMPLEX GEOLOGY - BASED ON AEM, SEISMICS AND BOREHOLES

3D modelling is a very time-consuming process, especially when dealing with complex geological environments, in which a high degree of heterogeneity needs to be resolved. At the same time, the resolution of a complex model area requires large amounts of spatially distributed data.

Due to the marked geological heterogeneity of a 100 km² study area in western Denmark, it would be difficult and very time-consuming to produce an entirely manual 3D-model for the area. Instead an approach for 3D geological modelling incorporating geostatistical data assessment has been developed. The complex sedimentary setting was mapped with 600 line km of airborne transient electromagnetic data (AEM, SkyTEM) and 77 km of high-resolution seismic lines. Furthermore, an investigation well and around 600 existing borehole logs were available. All datasets confirm the known high degree of geological complexity in the area, especially in the Quaternary section.

At first, a geostatistical estimation concept (the SSV method) was applied to the SkyTEM and borehole data. The geostatistical concept utilizes non-linear inversion to estimate the clay content of each model-voxel by optimising a function that translates the geophysically measured electrical resistivities to clay content by the use of borehole information. However, since only few wells penetrate the deepest part of the model and the SkyTEM data loose sensitivity with depth, the geostatistical estimation is only applied to the upper section. The seismic data, on the other hand, has a relatively better resolution with depth and by correlating to the deep exploration well, the seismic data was used to interpret and model the deeper and more consistent layers. The results of the geostatistic calculations were then combined with the manually interpreted seismic data.

The approach generated a 3D model that could be completed significantly faster than a fully manual model while still providing the degree of detail required for the subsequent groundwater modelling. Uncertainties for the different data were evaluated and considered during modelling and due to the geostatistical approach the resulting model is very ‘true to the data’, which improves the transparency of the modelling process.