STOCHASTIC MODELING OF INSITU BIOREMEDIATION IN HETEROGENEOUS AQUIFERS

The prolongation of in situ remediation due to aquifer heterogeneities is investigated in a series of Monte Carlo simulations. The remediation time in a homogeneous aquifer with an equal initial mass of organic pollutant serves as a reference. Non-uniform distributions of hydraulic conductivity and initial pollutant mass in the model are generated with a turning bands method assuming a log-normal distribution for both quantities. Deterministic reactive transport is computed in each aquifer realization with a coupled two-dimensional numerical model accounting for concentration changes due to advection, dispersion and aerobic microbial growth. A single-quarter five-spot serves as model area, where saturated groundwater flow is induced by an injection and a pumping well exclusively. The different model runs are compared on the basis of their respective time to achieve 90% pollutant removal. For the sake of comparability, the realizations are constructed such that they have equal initial pollutant mass and equal geometric mean of the hydraulic conductivity. Superposition of heterogeneous distributions of hydraulic conductivity and initial pollutant mass as well as an increase in standard deviations of both random variables from 0.92 to 1.84 had a pronounced unfavourable effect on remediation time, while doubling the auto-correlation length had only a smaller negative influence. In general, any type of heterogeneity tested in the numerical experiments led to a decreased overall pollutant removal rate compared to the homogeneous case. Furthermore, the crucial impact of local dispersivity as well as numerical dispersivity on removal time is shown. Due to the non-linear nature of the problem only a rough estimate can be derived to transfer results to situations with different initial masses. Finally, a method is presented to increase efficiency of in situ remediation in heterogeneous aquifers by operational means. It consists of changing locations of the active injection and pumping wells after a certain time in such a way that the mean flow direction in the single-quarter five-spot is changed by an angle of 90-degrees. Although this scheme demands twice the number of wells compared to the common scheme, it renders the remediation much more robust and thus trustworthy.