| Popis: |
Reliable information about the aquifer geometry and its spatial variability from geophysical methods plays a critical role for various hydrological research questions. Such information can be obtained with the transient electromagnetic (TEM) method that reaches a larger depth of investigation with a smaller survey layout compared to electrical or seismic methods. However, the quantitative interpretation of the resistivity model obtained from TEM data in terms of groundwater level and aquifer geometry might be biased by the non-uniqueness of the deterministic inversion. To overcome such limitations, we propose here to use Bayesian evidential learning (BEL1D) to evaluate the uncertainty of the groundwater level and aquifer thickness interpreted from TEM results using a classical deterministic inversion approach. Additionally, we investigate the effect of different prior model spaces on the uncertainty obtained from BEL1D and use the distance-based global sensitivity analysis (DGSA) to determine whether model parameters with a large uncertainty are actually non-influential on the model response. To test the uncertainty quantification, TEM data were measured at three sites in Austria representative of different hydrogeological settings and groundwater levels, namely: 1) a shallow (1 m - 10 m) aquifer located in the soda lakes of the Neusiedl-Seewinkel Basin in Burgenland, 2) an aquifer in intermediate (5 m – 15 m) depth located in the hydrological open-air laboratory (HOAL) in lower Austria and 3) a deep (> 35 m) aquifer in a farm land located in Upper Austria. We obtain TEM data with the TEM-FAST 48 instrument with a 6 m, 12.5 m and a 25 m square single-loop configuration to achieve sensitivities corresponding to the three different aquifer depths. Additionally, we use electrical resistivity tomography (ERT) and multi-channel analysis of surface waves (MASW) data to assess the TEM inversion results. The interpretation of the TEM inversion results is evaluated with BEL1D to obtain the uncertainty of the groundwater level from the cumulative uncertainty of all layers above the layer representing the groundwater level as well as the thickness of the aquifer. Additionally, we achieve a quantitative evaluation of the solved TEM model uncertainties with the DGSA method as well as with the ERT and MASW inversion results. Our results show a lower uncertainty for the electrical resistivity than for the layer thickness, while the DGSA reveals a decrease of sensitivity with depth. |
