A reactive transport model for mercury fate in contaminated soil--sensitivity analysis.

We present a sensitivity analysis of a reactive transport model of mercury (Hg) fate in contaminated soil systems. The one-dimensional model, presented in Leterme et al. (2014), couples water flow in variably saturated conditions with Hg physico-chemical reactions. The sensitivity of Hg leaching and volatilisation to parameter uncertainty is examined using the elementary effect method. A test case is built using a hypothetical 1-m depth sandy soil and a 50-year time series of daily precipitation and evapotranspiration. Hg anthropogenic contamination is simulated in the topsoil by separately considering three different sources: cinnabar, non-aqueous phase liquid and aqueous mercuric chloride. The model sensitivity to a set of 13 input parameters is assessed, using three different model outputs (volatilized Hg, leached Hg, Hg still present in the contaminated soil horizon). Results show that dissolved organic matter (DOM) concentration in soil solution and the binding constant to DOM thiol groups are critical parameters, as well as parameters related to Hg sorption to humic and fulvic acids in solid organic matter. Initial Hg concentration is also identified as a sensitive parameter. The sensitivity analysis also brings out non-monotonic model behaviour for certain parameters.

A reactive transport model for mercury fate in soil--application to different anthropogenic pollution sources.

Soil systems are a common receptor of anthropogenic mercury (Hg) contamination. Soils play an important role in the containment or dispersion of pollution to surface water, groundwater or the atmosphere. A one-dimensional model for simulating Hg fate and transport for variably saturated and transient flow conditions is presented. The model is developed using the HP1 code, which couples HYDRUS-1D for the water flow and solute transport to PHREEQC for geochemical reactions. The main processes included are Hg aqueous speciation and complexation, sorption to soil organic matter, dissolution of cinnabar and liquid Hg, and Hg reduction and volatilization. Processes such as atmospheric wet and dry deposition, vegetation litter fall and uptake are neglected because they are less relevant in the case of high Hg concentrations resulting from anthropogenic activities. A test case is presented, assuming a hypothetical sandy soil profile and a simulation time frame of 50 years of daily atmospheric inputs. Mercury fate and transport are simulated for three different sources of Hg (cinnabar, residual liquid mercury or aqueous mercuric chloride), as well as for combinations of these sources. Results are presented and discussed with focus on Hg volatilization to the atmosphere, Hg leaching at the bottom of the soil profile and the remaining Hg in or below the initially contaminated soil layer. In the test case, Hg volatilization was negligible because the reduction of Hg(2+) to Hg(0) was inhibited by the low concentration of dissolved Hg. Hg leaching was mainly caused by complexation of Hg(2+) with thiol groups of dissolved organic matter, because in the geochemical model used, this reaction only had a higher equilibrium constant than the sorption reactions. Immobilization of Hg in the initially polluted horizon was enhanced by Hg(2+) sorption onto humic and fulvic acids (which are more abundant than thiols). Potential benefits of the model for risk management and remediation of contaminated sites are discussed.

Including spatial variability in Monte Carlo simulations of pesticide leaching.

A methodology is developed to quantify the uncertainty in a pesticide leaching assessment arising from the spatial variability of non-georeferenced parameters. A Monte Carlo analysis of atrazine leaching is performed in the Dyle river catchment (Belgium) with pesticide half-life (DT50) and topsoil organic matter (OM) content as uncertain input parameters. Atrazine DT50 is taken as a non-georeferenced parameter, so that DT50 values sampled from the input distribution are randomly allocated in the study area for every simulation. Organic matter content is a georeferenced parameter, so that a fixed uncertainty distribution is given at each location. Spatially variable DT50 values are found to have a significant influence on the amount of simulated leaching. In the stochastic simulation, concentrations exist above the regulatory level of 0.1 microg L(-1), but virtually no leaching occurs in the deterministic simulation. It is axiomatic that substance parameters (DT50, sorption coefficient, etc.) are spatially variable, but pesticide registration procedures currently ignore this fact. Including this spatial variability in future registration policies would have significant consequences on the amount and pattern of leaching simulated, especially if risk assessments are implemented in a spatially distributed way.

Discriminating between point and non-point sources of atrazine contamination of a sandy aquifer.

This study analyses the sources of atrazine contamination in the Brusselian sandy aquifer of central Belgium. Atrazine has in the past been used for both agricultural and non-agricultural applications, but it is difficult to distinguish the contamination originating from these two sources. The spatial and temporal covariance of atrazine concentrations was studied by fitting semi-variogram models to monitoring data. Correlation ranges were found to be 600 m and 600-700 days, respectively. The results were used to apply a declustering algorithm before examining the distribution of atrazine concentrations measured in groundwater. Monitoring data appeared to follow a pseudo-lognormal distribution, as a lognormality test was negative. An inflexion point on the cumulative density function was thought to indicate the two different pollution processes, i.e., agricultural and non-agricultural contamination sources. A non-parametric one-way analysis of variance suggested that the vast majority of atrazine in groundwater was from non-agricultural, point sources. This was supported by the strong relationship between mean concentrations and land use, whilst other environmental variables, such as soil organic matter or groundwater depth, produced less meaningful results.