Optimization of a 2d finite element unsaturated flow model in a deep DNAPL contaminated area in Rho (Milan, Italy)

Three different approaches have been considered for the optimization of a 2D finite element unsaturated flow model of a former chemical facility area in an industrial district nearby Milan, Italy. The groundwater resources are deeply affected by DNAPL contamination (especially by-products of chemical treatments, such as TCE and sub-products) to a depth of 45 meters. A small sector of the upper unconfined aquifer, divided from the lowest ones by a discontinuous 2-meter-thickness aquitard, has been supposed to contain the source bulk of the contamination, even though other sources seem to be possible. In the need of stopping the advancing front of the NAPL plume, hazardously headed toward some public wells, the upper part of the aquifer was partially encapsulated by means of subsurface engineering systems (i.e. vertical slurry cut-off walls). Besides, some wells have been drilled to form an hydraulic barrier system to remediate polluted water through a pump-and-treat plant. However, some evidences (anomalous contaminant concentrations and anomalous hydraulic heads in the observation boreholes) underline lacks of efficiency of the remediation system. Our purpose were to come up with the identification of the failures of the system considering: a) the water budget in the steady state of the encapsulated contaminated soil and b) the amount of inward-outward flow and the exchanges between the upper aquifer and the lower, semi-confined aquifer. An attempt to calculate the amount of NAPL released in the last years from the encapsulated zone have also been performed, even if no sufficient data were available to obtain a complete reliable value. The first step consisted in the design of the conceptual model of the system and in defining the recharge/discharge relationship. A small river and numerous irrigation channel might have to be concerned but their contribution can be considered negligible, and the unique recharge depend on precipitation. Thus, short term and seasonal changes of the water table have been constantly monitored, and analysed by means of statistical and geostatistical solutions. In addition, the comparison with the amounts of precipitation showed the degree of correlation between the two variables. Laboratory tests (indirect permeability tests, grain size, organic content, specific weight) and field tests (permeameter tests, double-ring infiltrometer test,) have been carried out. A monitoring systems, based on the Frequency Domain Reflectometry (FDR) technique, recording soil moisture content at different depth coupled with some tensiometer have been installed; a great number of data were thus obtained to analyse the soil behaviour in infiltration and draining conditions and to calibrate the soil moisture and hydraulic conductivity functions of soils. A 2D numerical model for unsaturated and saturated condition have been developed; its calibration process basically counted on three different steps: a) comparison and matching of the calculated and observed soil suction pressures vs. time and vs. depths in the vadose zone; b) comparison and matching of calculated and observed potential head vs. times curves, both inside and outside the confined source area; We found out that the original design of the vertical barrier suffered of lack of acknowledgments about the geological structure. In fact, a remarkable exchange between the inner and the outer part of the confined source area occurs through the aquitard horizon: the walls may have played a role as a funnel, and the plume been headed even more directly towards the lower layers. This indicates the need of a revision of the existing remedial action which can be effectively designed only with the support of a reliable knowledge of processes analysed and quantified, both in the vadose zone and in the aquifer, through a complete calibrated numerical model.