Stochastic Data Integration to model Quaternary Aquifers: Application on
the Aare Valley, Switzerland
Abstract
Nowadays, the centralization of subsurface-related observation is more
and more common. Databases constituted of an ensemble of geophysical,
hydrological, or borehole measurement forms a unique source of
information for subsurface hydro-geological modeling. However, the
increasing amount of data and the importance of being able to update the
model when new data becomes available creates a huge need for automatic
workflow for such data integration, hopefully reeling on open-source
technology. Quaternary deposits are complex to model due to the high
heterogeneity they present. Over the last 2.5 Ma, multiple cycles of
glacial and inter-glacial phases deposited complex intertwining of
sedimentary pattern, with high contrast in parameters (e.g. :
permeability, porosity, nature of sediment) and space. Nevertheless,
these deposits are some of the most extensively used for water
resources, shallow geothermal exploitation, or construction material
exploitation. In this study, we propose a new method based on a flipped
stochastic joint inversion, and applied it on complex Quaternary
deposits. Geophysical, boreholes and hydrological observations are
inverted together. Boreholes are used to generate stochastic geological
models, populated with parameters. The inversion tune the geological
model in order to fit the field data. Our method allows not only to
integrate boreholes, geophysical and hydrological data, but also
conceptual models, with a robust uncertainty estimation. The method was
applied on some areas of the Upper Aare Valley (Switzerland), a valley
covered by more than 58’000 geophysical measurement points, 1’500
boreholes and 100 of hydrological observations. Our method showed
promising results in combining these data. Comparison with existing
geological models proves that our automated method not only show
realistic underground structures, but also significantly improved the
regional knowledge of the underground by combining all the existing
data, and will therefore lead to better decision making while being
based on open-source technology.