Coupling Hydrogeosphere and PEST for the inversion of steady and transient-state hydraulic tomography data sets

Eduardo Emilio Sánchez-León1, Claus Haslauer1, Carsten Leven1
1 Zentrum für Angewandte Geowissenschaften, Universität Tübingen

P 1.8 in Numerische Simulation von Strömungs- und Transportprozessen in Grundwasserleitern und angrenzenden Kompartimenten

Hydraulic head measurements acquired during consecutive cross-hole pumping tests with multiple pumping and observation locations contain non-redundant information about aquifer heterogeneity. Such hydraulic tomographic methods are therefore a promising technique to reconstruct an aquifer’s hydraulic conductivity (K) distribution from large datasets of head measurements by inverse parameter estimation schemes. However, the non-uniqueness of inverse problems requires solutions based typically on regularization schemes. For this work, a numerical groundwater flow model (HydroGeoSphere) and the parameter estimation software PEST were coupled. A pilot points based spatial regularization scheme was employed to invert hydraulic tomography data sets for reconstructing 3D K-fields. For this purpose, a synthetic flow model with known structure of spatially distributed K field was used to generate flow simulations with subsequent inversion of the simulated head measurements. These simulations were used to assess to which degree the synthetic K-field could be reproduced.

Both steady-state and transient-state hydraulic tomography data sets were analyzed, each with eight simulated sequential pumping tests for comparison with the reference field and between the inversion results. For better numerical stability Tikhonov regularization and truncated Singular value decomposition were included into the inversion process. Due to the computational burden – especially in transient conditions – so-called super-parameters were employed, which are linear combinations of the base parameters that are actually adjusted during the inversion process.

A sensitivity analysis was carried out to define the ideal number and spatial distribution of pilot points that would extract as much information as possible from the underlying dataset. The inversion was performed on hydraulic head measurements recorded during a field campaign at the Lauswiesen test site of the University of Tübingen. During the experiment, transient head data was collected during nine short term sequential pumping tests with a total of 40 observation locations.

The employed inversion approach allowed to recover major components of the spatial structure of the hydraulic conductivity distribution at the test site.

Letzte Änderung 31.10.2013