Three-dimensional reconstruction of aquifer heterogeneity for modeling the transport of Cr(VI) in an Alpine alluvial aquifer

This work presents a transport model of Cr(VI) considering the heterogeneity of the aquifer system. Three different settings of hydraulic parameters distribution were compared. The Alpine alluvial aquifer of the Aosta Plain (Aosta Valley Region, N-W Italy) was modelled using MODFLOW2000 and MT3DMS. The study area (~30 km2) is crossed from W to E by the Dora Baltea River that gains groundwater in the eastern part of the area. The aquifer system (average thickness of 85 m) forms an unique unconfined aquifer in the western part whereas is subdivided by a discontinuous silty layer (2-9 m thick) in a shallow unconfined aquifer (~20 m thick) and a semiconfined aquifer (25-12 m thick) in the eastern part. The two aquifer units in the eastern part are recharged by the groundwater coming from the unconfined aquifer in the western part. Groundwater flows from W to E and has a component of upward flow in the eastern part. This is confirmed by the hydrochemical data, e.g., the Cr(VI) plume only affects the unconfined aquifer. The influence of the distribution of the hydraulic parameters (e.g., hydraulic conductivity (K) and effective porosity (ne)) on the simulated heads and Cr(VI) concentrations was tested using three different settings of K and ne: S1 - a unique value of K and ne; S2 - a continuous distribution of K and ne obtained by kriging; S3 - a discrete distribution of K and ne classed in 10 different zones. The unique value used in S1 is the average K and ne value of the aquifer system. The continuous distribution of K and ne in S2 was reconstructed as follows. The lithologs in the Aosta Plain (176) stored in the TANGRAM database were coded and converted into numerical values of K and ne. This numerical coding of the lithological information was performed by TANGRAM through conversion tables. The K values assigned in the conversion process for each lithotype were calibrated using K values obtained from pumping tests performed in previous works. The K and ne values extracted by TANGRAM were then interpolated by kriging obtaining a 3D continuous distribution. The S3 followed the method applied in S2, but here the K and ne values were then classed into zones. The K and ne values of each zone were finally calibrated using PEST. The S3 gave the best fit between measured and simulated hydraulic heads and Cr(VI) concentrations