Assessment of Local Hydraulic Parameters by Enkf Data Assimilation in Real Aquifers: a Case Study in Downtown Padova (italy)

The calibration of natural aquifer model parameters in real world cases is a challenging goal that requires a careful tuning work, whose complexity increases with the number of available data. Paradoxically, instead of making easier the subsurface model numerical assessment, a high spatial density of data highlights the local scale heterogeneity effects, complicating the calibration procedure when an accurate reproduction of the local water table variations is requested. This is the case of the subsurface in the “Eremitani” area in downtown Padova (Italy), where an extensive monitoring was recently carried out to assess the water table alteration due to the possible realization of important underground works close to the “Scrovegni Chapel”, a renowned historical monument. Due to the implications for the foundation stability of the monumental building, the delicate structural equilibrium of the famous Giotto’s frescoes may be altered by the variations of the water table. For this reason, a relatively large number of piezometers and wells (16) were drilled in an area of approximately 8 ha, within two aquifers characterizing the subsurface medium down to a depth of 30 m. Measurements were collected for a relatively long period and some pumping tests, as well as a field experiment involving the controlled variation of the Piovego Canal – the watercourse crossing the study area –, were realized for monitoring the corresponding response of the water table in each observation well. To overcome the difficulties related to the calibration of a fully 3D finite element model solving the Richards’ equation, a data assimilation procedure was developed by integrating the groundwater model with the ensemble Kalman filter (EnKF) and the augmented state technique. The objective of this study is to retrieve the most relevant subsurface parameters (e.g., hydraulic conductivity and specific storage coefficient) by assimilating the piezometric data collected by the monitoring network, thus assessing the local scale heterogeneity due not only to the vertical stratification, but also to the horizontal spatial variability characterizing the area under investigation.