Modeling the 3-D hydraulic conductivity distribution in the southern Po river plain by stationary and non-stationary geostatistics

In the last decades, numerical modeling has become one of the most important instruments to manage groundwater resources, in terms of sustainable exploitation, contamination fate assessment, and remediation scheme design and efficiency evaluation. Besides the assessment for the hydrogeological features (i.e. boundary conditions), the hydraulic conductivity (K) distribution assessment is an important issue to be faced, especially when the considered aquifer is complex. Generally, this parameter is assigned to each hydrogeological complex, relying on values obtained by point permeability tests (e.g. pumping tests, slug tests, etc.). Although the existing calibration methods allow a good model optimization, the natural variability of the hydraulic properties cannot be reproduced in detail using deterministic and point approaches to calibration. The main objective of the research is to draw a physically based 3­D hydraulic conductivity model by both stationary (i.e. Ordinary Kriging, OK) and non­stationary (Intrinsic Random Function theory, IRF­k) geostatistical methods, comparing them in terms of accuracy and precision. These techniques were applied to 182 Cone Penetration Test (CPT) profiles of the tip (qc) and shaft (fs) resistances, collected by the Emilia Romagna Regional Geological Survey. The obtained mechanical 3­D models, related to qc and fs variables, allowed to calculate the lithology index (Ic) 3­D model and the corresponding K 3­D model. The selected study area is located in the southern part of the Po plain and is characterized by mainly alluvial deposits made up of undifferentiated fine silty­sandy deposits, with coarser (i.e. alluvial fans and paleo­channels) and finer (i.e. lacustrine lenses) geological bodies. These inclusions consist of gravelly alluvial fans, prevalent nearby the Apennines reliefs, sandy paleo­channels, predominant northward, and lacustrine lenses, detectable all over the area. As a result, the OK and IRF­k method allowed to estimate the k values in a 3­D model, starting from CPT data. The obtained models reproduce as closely as possible the actual geological and hydrogeological features, not neglecting the multi­scale heterogeneity. The proposed methodological approach provides a physically based 3­D hydraulic conductivity model, with a noticeable degree of detail. Hence the models obtained can represent a useful starting point for hydrogeological modeling and/or a constraint for model calibration when the intrinsic deposit heterogeneity could affect substantially the groundwater flow and contaminant transport in the aquifer