Article pubs.acs.org/est
Effect of Subgrid Heterogeneity on Scaling Geochemical and Biogeochemical Reactions: A Case of U(VI) Desorption Chongxuan Liu,†,* Jianying Shang,† Huimei Shan,†,‡ and John M. Zachara† †
Pacific Northwest National Laboratory, Richland, Washington, 99352 United States China University of Geosciences, Wuhan, Hubei Province 430074, People’s Republic of China,
‡
S Supporting Information *
ABSTRACT: The effect of subgrid heterogeneity in sediment properties on the rate of uranyl[U(VI)] desorption was investigated using a sediment collected from the U.S. Department of Energy Hanford site. The sediment was sieved into 7 grain size fractions that each exhibited different U(VI) desorption properties. Six columns were assembled using the sediment with its grain size fractions arranged in different spatial configurations to mimic subgrid heterogeneity in reactive transport properties. The apparent rate of U(VI) desorption varied significantly in the columns. Those columns with sediment structures leading to preferential transport had much lower rates of U(VI) desorption than those with relatively homogeneous transport. Modeling analysis indicated that the U(VI) desorption model and parameters characterized from well-mixed reactors significantly overpredicted the measured U(VI) desorption in the columns with preferential transport. A dual domain model, which operationally separates reactive transport properties into two subgrid domains, improved the predictions significantly. A similar effect of subgrid heterogeneity, albeit to a lesser degree, was observed for denitrification, which also occurred in the columns. The results imply that subgrid heterogeneity is an important consideration in extrapolating reaction rates from the laboratory to field.
■
INTRODUCTION Heterogeneity in sediment properties controlling fluid transport and geochemical reactivity is a common phenomenon in subsurface environments. Heterogeneity affects transport, mixing, and interactions of reactants that affect local and overall reaction rates.1−10 Overall reaction rates can be orders of magnitude lower in heterogeneous porous media than those measured in well-mixed, homogeneous systems as a result of the coupling of local reactions with mass transport along reactive transport pathways.10−12 This rate change caused by heterogeneity poses a significant challenge to scale geochemical reaction parameters from the laboratory to field, as reaction rates or rate constants become scale-dependent.12−16 U(VI) sorption and desorption in sediments is primarily controlled by U(VI) surface complexation reactions (SCR) on mineral surfaces,17−20 and mass transfer that limits the rate of U(VI) SCR in intragranular and intergranular domains in sediments.21−25 As with other geochemical reactions, scaling is a major challenge to extrapolate U(VI) sorption and desorption models from one scale to others.26 A grain-size-based additivity model 24 was recently proposed to scale U(VI) SCR parameters from the laboratory to field. The approach described U(VI) sorption and desorption in sediment by summing U(VI) SCR contributions from individual grain size fractions comprising the whole sediment.24,25 An advantage of the approach is that the spatial distribution of U(VI) SCR properties in a field site © 2013 American Chemical Society
may be estimated from grain size distribution data once SCR models have been established for individual grain size fractions. This predictive strategy is, however, only applicable to wellmixed sediments.24,25 A sediment with the same grain size distribution can have different structures, such as layers, coarse grain channels, cross-bedding, and so forth, reflecting varying local sediment deposition environments. When these sediment structures are at the subgrid scale and below characterization resolution, their effect on U(VI) sorption and desorption is unclear and has not been investigated. One objective of this study was to determine how subgrid heterogeneity in fluid flow and geochemical reaction properties affects the rates of U(VI) desorption and other related biogeochemical reactions. Another objective is to evaluate approaches for scaling reaction rates from well-mixed sediments to sediments with heterogeneous structures. Six columns with different sediment structures were assembled to represent a range of field-scale subgrid heterogeneity in hydrogeochemical properties and were leached with a synthetic groundwater (SGW). The source sediment contained U(VI) that displayed kinetic desorption behavior.22,25 The effluent results from the Received: Revised: Accepted: Published: 1745
September 23, 2013 December 13, 2013 December 30, 2013 December 30, 2013 dx.doi.org/10.1021/es404224j | Environ. Sci. Technol. 2014, 48, 1745−1752
Environmental Science & Technology
Article
Figure 1. (A) Field-scale grain size distribution showing a numerical grid (white square of 30 × 30 cm2) and subgrid heterogeneity; (B) XCT images of structured sediments packed in columns to represent field-scale, subgrid features: well-mixed distribution (column 1), fine-grained aggregates randomly distributed in coarse grain matrix (column 2), coarse-grained channel surrounded by fine-grained material (column 3), layer structure parallel to flow direction (column 4), layer structure in an angle to flow direction (column 5), and layer structure perpendicular to flow direction (column 6). In Plot B, circles (5 cm diameter) are the XCT images of the columns perpendicular to flow direction, and rectangles (5 × 30 cm2) are the XCT images along the flow direction.
Table 1. Physical Properties in Columns column no.
θ
va cm/h
ρb g/cm3
Db cm2/h
Dc cm2/h
θim/θm
ω hr−1
1 2 3 4 5 6
0.23 0.31 0.30 0.38 0.38 0.31
10.81 8.02 8.29 6.54 6.54 8.02
1.927 1.832 1.848 1.537 1.537 1.486
12.88(2.35) 23.67(6.61) 301(143) 317(125) 127(53.2) 6.39(1.18)
13.71(0.82) 13.35(1.53) 34.10(4.27) 27.50(2.99) 24.76(2.35) 6.39(1.18)
0.12(0.01) 0.19(0.01) 1.29(0.03) 1.43(0.02) 0.86(0.01) 0.00
1.00(0.01)x10−3 1.60(0.01)x10−3 1.46(0.64)x10−2 4.82(1.29)x10−3 1.80(0.01)x10−3 NA
Average pore velocity calculated from a constant flow rate of 45 cm3/h in all columns, column diameter of 4.8 cm, and porosity (θ). bEstimated from the single domain model (see text). cEstimated from dual domain model (see the text); NA denotes not applicable. Numbers in bracket are standard error values. a
where >SOH denotes the surface site and >SOUO2(CO3)23− is the adsorbed U(VI) species. Reaction 1 is the only SCR used in modeling U(VI) sorption and desorption in this study. Previous studies24,25 revealed that under well mixed conditions, U(VI) desorption in the sediment can be predicted by linearly summing U(VI) desorption in the individual grain size fractions weighted by their mass percentages in the composite. The surface area and labile U concentration were also found to be linearly additive with respect to grain size fraction.24,25 Column Experiments. Six columns with different sediment structures abstracted from field subgrid heterogeneity (Figure 1A) were designed to assess how the subgrid heterogeneity affects the rates of U(VI) desorption. The evaluated sediment structures included: (1) a random distribution of all grain size fractions (column 1, Figure 1B); (2) a random distribution of fine grain aggregates (made of 0.125 mm) (column 2, Figure 1B); (3) a coarse grain channel (made of 2−8 mm size materials) surrounded by finer grain materials (SOH, X (X denotes the cation exchange site), and 31 aqueous and ion exchange species were considered in modeling column results because they affected U(VI) SCR (eq 1) either directly or indirectly through aqueous speciation effects. All aqueous speciation and ion exchange reactions (SI Table S2) were modeled as equilibrium reactions. Aerobic denitrification, which was apparently driven by sorbed organic carbon as previously observed in the Hanford sediments,25,28,29 was also considered in the modeling because this process affects pH and carbonate concentration. The denitrification process was modeled as a first-order reaction with respect to both nitrate concentration and soil organic carbon.30 Sediment organic carbon was assumed to exist as CH2O for modeling purpose. Calcite dissolution was included to describe calcium and carbonate evolution during the experiments. Calcite dissolution was simulated as a kinetic reaction using the model developed by
(2)
i = 1, ..., N
∂t
where Ij is the rate for adsorbed U(VI) species j (mol/kg/h), qkj is the concentration of adsorbed U(VI) species j at adsorption site k (mol/kg), Ns is the total number of adsorbed species (only one sorbed U(VI) species (Reaction 1) was considered, i.e., Ns = 1, in this study), M is the total number of adsorption sites, αk is the rate constant at adsorption site k (h−1), Qkj is the adsorbed concentration of species j at sorption site k (mol/kg) in equilibrium with aqueous solution. The rate constants (αk) in eq 5 were assumed to follow a log-normal probability distribution,
∂Cim ∂C m ∂ 2Cim θm i = θD − θν − ωθim(Cim − Cii m) ∂t ∂x ∂x 2 θmR im
∑
∂qjk
(3)
where superscript/subscript “m” represents the mobile domain where transport in the intergrain pores is dominated by advection, and “im” represents the immobile domain where diffusion dominates in both intergrain and intragrain pores, Ci is the total (aqueous and equilibrium-sorbed) concentration of chemical component i (mol/L), Ri is the production rate of chemical component i (mol/L/h), ω is the mass transfer coefficient between the mobile and immobile domains (h−1), D (m2/h), and v (m/h) are the dispersion coefficient and averaged flow velocity in the column, θ is the total porosity (θ = θm + θim), and N is the total number of chemical components in the modeling. Equations 2 and 3 defined the dual domain model. By ignoring eq 3, and setting θim = 0 in eq 2, this model becomes the single domain model. Parameters v and θ in eq 2 were measured (Table 1). Parameter D is the only additional physical parameter to be determined from tracer data for the single domain model. For the dual domain model, D, θim/θm, and ω were estimated by fitting tracer Br data. Parameters θm 1747
dx.doi.org/10.1021/es404224j | Environ. Sci. Technol. 2014, 48, 1745−1752
Environmental Science & Technology
Article
fine grain regions in the columns would also contribute to the measured effluent U(VI), the slow flow rate in these regions diminished their effect on the flux-averaged effluent concentration. Column 4 had the lowest release rate, while column 5 had the highest among the slower U(VI) release group. The higher rate of U(VI) release in column 5 apparently resulted from the passage of SGW through finer grain materials near the column outlet (top, Figure 1B). The importance of fine material placement was further illustrated by the behavior of column 6, in which flow had to fully pass through all the fine grain materials before column exit. Desorption of U(VI) from column 6 was consequently much faster than that from column 5. The slowest U(VI) release rate from column 4 was caused by the placement of its finest layer (
