This paper proposes a method of assessing the distribution of chlorinated solvents in soil and ground water using tree branches. Sampling branches is a potentially more cost‐effective and easier method than sampling tree cores, with less risk of damage to the tree. This approach was tested at Argonne National Laboratory, where phytoremediation is being used to remove tetrachloroethene (PCE), trichloroethene (TCE), and carbon tetrachloride (CCl4) from soil and ground water. The phytoremediation system consists of shallow‐rooted willows planted in an area with contaminated soil and deep‐rooted poplars planted in an area with clean soil and contaminated ground water. Branch samples were collected from 126 willows and 120 poplars. Contaminant concentrations from 31 soil borings and six monitoring wells were compared to those from branches of adjacent trees. Regression equations with correlation coefficients of at least 0.89 were obtained, which were found to be chemical specific. Kriged profiles of TCE concentration based on soil and willow branch data were developed and showed good agreement. Profiles based on ground water data could not be developed due to lack of sufficient monitoring wells for a meaningful statistical analysis. An analytical model was used to simulate TCE concentrations in tree branches from soil concentrations; the diffusion coefficient for TCE in the tree was used as the fitting parameter and the best‐fit value was two orders of magnitude greater than literature values. This work indicates that tree branch sampling is a useful approach to assess contaminant distribution and potentially to determine where to locate monitoring wells or perform detailed soil analysis. Further research is necessary prior to using this method as a quantitative monitoring tool for soil and ground water.
Pd-based catalysts provide efficient and selective reduction of several drinking water contaminants, but their long-term application requires effective treatments for catalyst regeneration following fouling by constituents in natural waters. This study tested alumina-supported Pd−Cu and Pd−In bimetallic catalysts for nitrate reduction with H2 after sulfide fouling and oxidative regeneration procedures. Both catalysts were severely deactivated after treatment with μM levels of sulfide. Regeneration was attempted with dissolved oxygen, hydrogen peroxide, sodium hypochlorite, and heated air. Only sodium hypochlorite and heated air were effective regenerants, specifically restoring nitrate reduction rates for a Pd−In/γ-Al2O3 catalyst from 20% to between 39 and 60% of original levels. Results from ICP−MS revealed that sodium hypochlorite caused dissolution of Cu from the Pd−Cu catalyst but that the Pd−In catalyst was chemically stable over a range of sulfide fouling and oxidative regenerative conditions. Analysis by XPS indicated that PdS and In2S3 complexes form during sulfide fouling, where sulfur is present as S2-, and that regeneration with sodium hypochlorite converts a portion of the S2- to S6+, with a corresponding increase in reduction rates. These results indicate that Pd−In catalysts show exceptional promise for being robust under fouling and regeneration conditions that may occur when treating natural waters.
The amount, location, and form of NAPL in contaminated vadose zones are controlled by the spatial distribution of water saturation and soil permeability, the NAPL spill scenario, water infiltration events, and vapor transport. To evaluate the effects of these processes, we used the three-phase flow simulator STOMP, which includes a new permeability–liquid saturation–capillary pressure (k–S–P) constitutive model. This new constitutive model considers three NAPL forms: free, residual, and trapped. A 2-D vertical cross-section with five stratigraphic layers was assumed, and simulations were performed for seven cases. The conceptual model of the soil heterogeneity was based upon the stratigraphy at the Hanford carbon tetrachloride (CT) spill site. Some cases considered co-disposal of NAPL with large volumes of wastewater, as also occurred at the Hanford CT site. In these cases, the form and location of NAPL were most strongly influenced by high water discharge rates and NAPL evaporation to the atmosphere. In order to investigate the impact of heterogeneity, the hydraulic conductivity within the lower permeability layer was modeled as a realization of a random field having three different classes. For six extreme cases of 100 realizations, the CT mass that reached the water table varied by a factor of two, and was primarily controlled by the degree of lateral connectivity of the low conductivity class within the lowest permeability layer. The grid size at the top boundary had a dramatic impact on NAPL diffusive flux just after the spill event when the NAPL was present near the ground surface. NAPL evaporation with a fine grid spacing at the top boundary decreased CT mass that reached the water table by 74%, compared to the case with a coarse grid spacing, while barometric pumping had a marginal effect for the case of a continuous NAPL spill scenario considered in this work. For low water infiltration rate scenarios, the distribution of water content prior to a NAPL spill event decreased CT mass that reached the water table by 98% and had a significant impact on the formation of trapped NAPL. For all cases simulated, use of the new constitutive model that allows the formation of residual NAPL increased the amount of NAPL retained in the vadose zone. Density-driven advective gas flow from the ground surface controlled vapor migration in strongly anisotropic layers, causing NAPL mass flux to the lower layer to be reduced. These simulations indicate that consideration of the formation of residual and trapped NAPLs and dynamic boundary conditions (e.g., areas, rates, and periods of different NAPL and water discharge and fluctuations of atmospheric pressure) in the context of full three-phase flow are needed, especially for NAPL spill events at the ground surface. In addition, NAPL evaporation, density-driven gas advection, and NAPL vertical movement enhanced by water flow must be considered in order to predict NAPL distribution and migration in the vadose zone.
Several studies have demonstrated that the success of natural and engineered in situ remediation of groundwater pollutants relies on the transverse mixing of reactive chemicals or nutrients along plume margins. Efforts to predict reactions in groundwater generally rely on dispersion coefficients obtained from nonreactive tracer experiments to determine the amount of mixing, but these coefficients may be affected by spreading, which does not contribute to reaction. Mixing is controlled only by molecular diffusion in pore spaces, and the length scale of transverse mixing zones can be small, often on the order of millimeters to centimeters. We use 2D pore‐scale simulation to investigate whether classical transverse dispersion coefficients can be applied to model mixing‐controlled reactive transport in three different porous media geometries: periodic, random, and macroscopically trending. The lattice‐Boltzmann method is used to solve the steady flow field; a finite volume code is used to solve for reactive transport. Nonreactive dispersion coefficients are determined from the transverse spreading of a conservative tracer. Reactive dispersion coefficients are determined by fitting a continuum model which calculates the total product formation as a function of distance to the results from our pore scale simulation. Nonreactive and reactive dispersion coefficients from these simulations are compared. Results indicate that, regardless of the geometrical properties of the media, product formation can be predicted using transverse dispersion coefficients determined from a conservative tracer, provided dispersion coefficients are determined beyond some critical distance downgradient where the plume has spread over a sufficiently large transverse distance compared to the mean grain diameter. This result contrasts with other studies where reactant mixing was controlled by longitudinal hydrodynamic dispersion; in those studies longitudinal dispersion coefficients determined from nonreactive tracer experiments over‐estimated the extent of reaction and product formation. Additional work is called for in order to confirm that these findings hold for a wider variety of grain sizes and geometries.
Recent studies indicate that during in situ bioremediation of contaminated groundwater, degradation occurs primarily along transverse mixing zones. Classical reactive-transport models overpredict the amount of degradation because solute spreading and mixing are not distinguished. Efforts to correct this have focused on modifying both dispersion and reaction terms, but no consensus on the best approach has emerged. In this work, a pore-scale model was used to simulate degradation along a transverse mixing zone between two required nutrients, and a continuum model with fitted parameters was used to match degradation rates from the pore-scale model. The pore-scale model solves for the flow field, concentration field, and biomass development within pore spaces of porous medium. For the continuum model, the flow field and biomass distributions are assumed to be homogeneous, and the fitting parameters are the transverse dispersion coefficient (DT) and maximum substrate utilization rate (kS,c). Results from the pore-scale model show that degradation rates near the system inlet are limited by the reaction rate, while degradation rates downgradient are limited by transverse mixing. For the continuum model, the value of DT may be adjusted so that the degradation rate with distance matches that from the pore-scale model in the mixing-limited region. However, adjusting the value of kS only improves the fit to pore-scale results within the reaction-limited region. Comparison with field and laboratory experiments suggests that the length of the reaction rate-limited region is small compared to the length scale over which degradation occurs. This indicates that along transverse mixing zones in the field, values of kS are unimportant and only the value of DT must be accurately fit.