Présentations Colloques

Oral Presentation
8.10
Session 8.10: Recent developments in groundwater modeling and mathematical tools in Hydrogeology
Jacopo Terrenghi
Should we collect more K data or more aquifer samples for effective subsurface characterization? A comparative study based on reproducibility of flow and transport modelling results.
Accurate numerical models of groundwater flow and solute transport behaviour require realistic conceptualizations of subsurface heterogeneity, and in particular of the hydraulic conductivity (K) field. In a stochastic framework, such conceptualizations can be generated with geostatistical methods, and depending on the type of conditioning data, two approaches can be taken to generate multiple realizations of the K field. One approach (here referred to as the hydrological approach) consists in an explicit reconstruction of the 3-D K field conditioned to available K measurements at selected points in the domain of interest. The second approach (here referred to as the geological approach) consists in an initial step in which realizations of the distribution of geological units with similar hydraulic properties are generated conditioned to aquifer samples along boreholes, and in a second step in which K values are mapped on the basis of distribution of these units. Because collecting observations of hydrological and geological properties is expensive and time consuming, it is important to evaluate the minimum amount of data that can provide accurate predictions of groundwater flow and solute transport. In this study, we analyse the sensitivity of results from numerical simulations of groundwater flow and advective transport with respect to the amount of conditioning information, here measured as the density of conditioning data points in the model domain. The domain considers the shallow aquifer underneath an industrial megasite (Porto Marghera, Italy). A series of stochastic models were implemented, each considering a different number of conditioning points in the generation of the K field, either with the hydrological or geological approach. Results were then compared to those from the model with the highest density. Comparisons show that the reproducibility of the results from models based on the geological approach is less affected by the loss of information caused by a lower number of conditioning points. This indicate that taking into account the geological structure in the K field is a cost-effective strategy for subsurface characterization.
Italy