Environ. Sci. Technol. 1997, 31, 2331-2338
Nonequilibrium Transport of Reactive Solutes through Layered Soil Profiles with Depth-Dependent Adsorption L E I G U O , * ,† R O B E R T J . W A G E N E T , † J O H N L . H U T S O N , †,‡ A N D CHARLES W. BOAST§ Department of Soil, Crop, and Atmospheric Sciences, Cornell University, Ithaca, New York 14853, and Department of Natural Resources and Environmental Sciences, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801
The pesticide atrazine was used as a reactive tracer in a series of miscible displacement experiments that were conducted with soil columns containing the same amount of sludge or manure, but distributed either as a layer or uniformly in depth. Travel time probability density function (pdf) of Cl- and atrazine measured from both types of columns was analyzed based upon a transfer function model that assumes a two-site nonequilibrium advection-dispersion equation (ADE). Sorption sites contributed by both organic materials were primarily rate-limited. Travel time of atrazine, estimated by temporal moment analysis on the measured travel time pdf, was consistently increased in organic material-amended columns for both distributions. When flux concentration of atrazine was fitted to the nonequilibrium ADE, either a higher overall distribution coefficient (Kd) or a higher fraction of instantaneous adsorption sites (f) or both were found for layered columns, indicating an enhanced accessibility to sorption sites of organic amendments probably due to less shielding by soil minerals.
Introduction Movement of chemical pollutants through heterogeneous soil profiles is a prevalent process in the environment. One prominent feature of many heterogeneous soils is their change in sorption capacity for reactive solutes with depth due to changes in the content of organic matter, which is the predominant sorbent for organic pollutants in soils (1-3). In addition, organic materials from sources such as sludge, manure, and crop residues are often applied artificially to soil surfaces to increase, among many other beneficial purposes, the capacity of the soil to retain reactive chemicals such as pesticides and, hence, to reduce the potential of groundwater contamination (4, 5). Therefore, understanding chemical transport through heterogeneous soil profiles is essential for more reliable predictions of the fate of pollutants in the environment. Chemical transport in soil is most commonly described by the advection-dispersion equation (ADE), which is based on the mass-conservation law. Since this equation is most easily solved by assuming a homogeneous medium (6, 7), its application to describe chemical transport in layered and other heterogeneous soil profiles is often accomplished by * Corresponding author e-mail address:
[email protected]. † Cornell University. ‡ Present address: School of Earth Sciences, The Flinders University of South Australia, Adelaide, South Australia 5001, Australia. § University of Illinois at Urbana-Champaign.
S0013-936X(96)00944-3 CCC: $14.00
1997 American Chemical Society
treating the more complex system as an equivalent single layer (8-10). Using the method of time moment analysis, Barry and Parker (9) derived conditions under which the equivalent single layer model would accurately predict effluent concentrations through layered systems. The analytical description of transport in multilayered porous media often follows the transfer function approach, as presented by Jury and Roth (11). Using this approach, Jury and Utermann (12) developed analytical expressions for the flux and resident concentrations in layered soil for two limiting transport cases, i.e., the zero and perfect travel time correlation between individual layers. In the solutions developed by Moranville et al. (13), Bruch (14), and Shamir and Harleman (15), no particular correlation was assumed regarding transport among layers. Rather, each individual layer was treated mathematically as effectively semi-infinite, and the outflow from one layer was taken as the upper boundary condition for the next layer. Most of these analytical solutions, however, are only given in the Laplace domain and would frequently have to be numerically inverted with respect to time or space and thus may suffer numerical instability for complicated systems (9). In addition to analytical approaches, solute transport through layered soils can also be described using numerical simulations. An early attempt at such an approach was reported by Selim et al. (16), who simulated solute transport through a range of layer-stratification scenarios with both equilibrium and kinetic adsorption ADE. Recently, Hutson and Wagenet (17) have extended this effort into the widely used simulation model LEACHM by incorporating the twosite nonequilibrium transport parameters (18) into individual layers. One advantage of using numerical over analytical approaches is that the nonideal sorption behaviors (sorption nonlinearity and nonsingularity) can be evaluated. Although calculations based on the above methods can be used to predict transport through layered soils, very few experimental results have been obtained, especially for reactive solutes, under well-defined experimental conditions intended to test theoretical assumptions and predictions. In this study, we followed the transfer funtion approach to analyze the travel time probability density function (pdf) of the reactive solute atrazine through soil columns amended with organic material (OM). Our purposes were (i) to characterize transport parameters of atrazine, in terms of a two-site nonequilibrium ADE, through heterogeneous soil profiles amended with two forms of OM; (ii) to test whether a transfer function model, with parameters based on experiments performed independently on two individual layers, can represent transport through a two-layered system; and (iii) to test whether the retaining and nonequilibrium behavior of atrazine would be affected by the distribution of the heterogeneous sorbents in the soil.
Theory The flux-averaged concentration Cf(z,t) of a chemical exiting from a transport volume under steady-state water flow can be viewed as its travel time pdf when the solute is introduced in an infinitely small time interval. Based on the principle of linear superposition, the transfer function model of Cf(z,t) subject to an arbitrary inlet concentration Cf(0,t) can thus be written as a time convolution integral (19):
Cf(z,t) )
∫ C (0,t′)F (z,t - t′) dt′ t
0
f
(1)
f
where z is the distance; t is the time; t′ is the dummy time variable; Cf(0,t) is the inlet concentration of the solute as a function of time; Ff(z,t) is the travel time pdf of the solute
VOL. 31, NO. 8, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2331
injected as an infinitely narrow pulse. Parametric representation of Ff(z,t) can be formed based on mechanistic process models. For transport of reactive solutes such as atrazine in our column displacement experiments, Ff(z,t) can be described by a two-site nonequilibrium model. This nonequilibrium transport model, in essence, is a conventional ADE coupled with a term describing rate-limited adsorption (18, 20):
(
)
(2)
∂S2 ) R[(1 - f)KdC - S2] ∂t
(3)
Cf(L1,T) )
where F is the soil bulk density; θ is the volumetric water content; v is the pore water velocity; Kd is the distribution coefficient of the solute; D is the diffusion-dispersion coefficient; S2 is the solute concentration on kinetic sorption sites; f is the fraction of equilibrium sorption sites; and R is the first-order kinetic rate constant. Ff(z,t) can be expressed in dimensionless form (21):
∫
T
0
x
where T is the dimensionless time equal to vt/L; Z is the dimensionless distance equal to z/L; τ is a dummy variable; and P, R, β, and ω are dimensionless parameters defined by
FKd R)1+ θ ω)
(5)
R(1 - β)RL v
and
Γ(Z,τ) )
Z τ
[
x
]
[
(6)
][ x
]
(T - τ)τ β(1 - β)
ωτ ω(T - τ) 2ω I βR (1 - β)R 1 R
(7)
where L is the column length and I1 is the modified Bessel function of order one. The travel time pdf of solute transport through a twolayered soil system can be approximated by taking the outflow of the reactive solute through the first transport volume to be the inlet boundary condition for the second transport volume and assuming that there are no interactions with respect to solute transport at the interface of the two layers (9). Therefore, the travel time pdf for solute in the second layer can be written as:
Cf(Z,T) ) Cf(L1 + L2,T) )
)
∫
T
0
∫ C (L ,t)F (L ,T - t) dt T
0
f
1
f
2
Cf(L1,P1,R1,β1,ω1,τ)Ff(L2,P2,R2,β2,ω2,T τ) dτ (8)
where L1 is the thickness of the first layer; L2 is the thickness of the second layer; and Pi, Ri, βi, and ωi are the corresponding transport parameters for the first and second layers. Since we have applied a single step pulse to the inlet end of the first
2332
9
(9)
C0A(Z,T) C0A(Z,T) - C0A(Z,T - T0)
0 < T e T0 (10) T > T0
where
A(Z,T) )
∫
T
0
ΓJ(a,b) dτ
(11)
where Γ was given in eq 6, and
∫
J(a,b) ) 1 - exp(-b)
a
0
exp(-λ)I0[2xbλ] dλ
(12)
ωτ βR
(13)
ω(T - τ) (1 - β)R
(14)
and
b)
and I0 is the zero-order modified Bessel function. The parametric representation of Ff(L2,T) is the solution of the two-site nonequilibrium ADE for the second layer subject to a Dirac delta input function. The expression of this solution was given in eq 4, where Z is now replaced by L2. Equation 8 states that the concentration at L1 + L2 to the finite pulse input at L ) 0 can be described by the convolution of eqs 4 and 10.
Materials and Methods
2
P(βRZ - τ) βRP exp 4πτ 4βRτ
and
H(τ,T) ) exp -
0 < T e T0 T > T0
a)
τ Γ(Z,τ)H(τ,T) dτ (4) β(1 - β)(T - τ)
θ + fFKd β) θ + FKd
C0 0
{
)
vL P) D
{
where
ωT + βR
Ff(Z,T) ) Γ(Z, T) exp ω R
Cf(0,T) )
This solution is given by (20):
fFKd ∂C F ∂S2 ∂C ∂ 2C 1+ + )D 2 +v θ ∂t θ ∂t ∂z ∂z
(
organic layer in our miscible displacement experiments, the Cf(L1,T) can thus be represented by the analytical solution of the two-site nonequilibrium ADE subject to the upper boundary condition:
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 8, 1997
Materials. All the experiments were conducted with a Niagara silt loam (fine-silty, mixed, mesic Aeric Ochraqualf) that was collected from Caldwell Field Farm at Cornell University, Ithaca, NY. The soil samples were taken from a depth of 0-15 cm and consisted of 1.77% organic carbon (OC), 18.3% clay, and 71.0% silt. The two organic materials, sludge and manure, contained 45.2% and 35.4% OC, respectively. The soil, manure, and sludge were air-dried and passed through a 2-mm sieve prior to use. The pulse solution for the miscible displacement experiments contained 5 µg/mL atrazine in 0.005 M CaSO4 and 0.025 M CaCl2. The pesticide solutions used in the batch experiments had an initial concentration of 1, 5, 10, 15, and 20 µg/mL and were all prepared in 0.005 M CaSO4. Batch Sorption Experiments. Batch adsorption studies with unamended and OM-amended soils were conducted to determine the distribution coefficient (Kd) of atrazine. The ratios of OM to soil used were equal to those in the soil columns of the miscible displacement experiments. Five gram of soil and 15 mL of pesticide solution at different concentrations were added to a 25-mL scintillation vial sealed with a Teflon-lined screw cap. The soil-water mixture was shaken on a reciprocating shaker for 24 h. The suspension was allowed to stand for 2 h before the supernatant was filtered through a disposable 0.45-µm nylon filter. The filtrate was then directly injected into a high-pressure liquid chromatography column (HPLC) for pesticide determination. All treatments were prepared in triplicates, including blanks (pesticide solution only) and controls (soil and pesticide-free
TABLE 1. Miscible Displacement Experimental Conditions flow regime
amendment treatment
length (cm)
density (g/cm3)
fast
sludge manure unamended sludge manure unamended
14.0 14.5 14.0 14.5 14.5 14.0
1.25 1.21 1.25 1.21 1.21 1.25
sludge manure sludge manure
14.5 14.0 14.5 14.5
1.21 1.25 1.21 1.21
sludge manure sludge manure
3.5 3.5 3.5 3.5
slow
fast slow
fast slow a
pore vol (cm3)
soil water content (cm3/cm3)
pore velocity (cm/h)
dispersion coeff (cm2/h)
Uniform Column 153.9 156.5 145.2 154.5 156.0 145.4
0.560 0.550 0.528 0.543 0.548 0.529
5.093 4.876 5.050 1.042 0.944 1.011
1.038 0.753 1.853 0.156 0.272 0.267
Layered Columna 151.8 145.2 156.7 155.0
0.533 0.528 0.533 0.528
4.869 4.931 0.949 0.961
1.183 1.068 0.198 0.201
Short Column (First OM Layer) 1.07 43.0 0.626 1.07 43.1 0.627 1.07 43.0 0.626 1.07 43.1 0.627
4.978 4.977 0.971 0.978
1.561 1.225 0.221 0.314
Values are overall for the entire column.
solution). The sorbed concentration was calculated as the difference between the initial concentration and the concentration at equilibrium. Miscible Displacement Experiments. The experimental conditions for the miscible displacement study are given in Table 1. Three groups of soil columns were studied. In one group, termed “layered columns”, the OM was amended only to the first 3 cm of the soil column. In another group, the OM was incorporated into the entire soil column, and the soil was macroscopically homogeneous in the vertical direction (uniform columns). The third group of columns contained only the first layer of the layered columns (short columns). The miscible displacement experiments were carried out under steady-state, saturated upward flow conditions. The amendment rate for both OM treatments was 9.8 g/column, equivalent to a surface application of 50 Mg/ha. At this rate, the OM content (dry soil basis) was 15.3% for the first OMamended layer and 2.9% for the columns homogenized with an OM amendment. For each experiment, the soil was mixed with the selected OM at the desired rate and was uniformly packed at incremental depths into a glass column, 5 cm in diameter, to a length that was adjustable using an adaptor at one end. The column was then saturated at a very slow flow rate (0.3 cm/h) with an isocratic LC pump (Perkin Elmer Model 250). For OM-amended columns, about 10 pore volumes of 0.005 M CaSO4 were first applied to leach out excessive amounts of soluble salts and organic matter contained in the amendment. After saturation and steadystate flow were established, a pulse of about 1 pore volume of aqueous atrazine and Cl- solution was applied. The chemicals were then subsequently displaced by tracer-free 0.005 M CaSO4. Two flow velocities, referred to as fast (pore water velocity ∼ 5 cm/h) and slow flow (pore water velocity ∼ 1 cm/h) were studied. The effluent samples from the column were collected in glass tubes using an automatic fraction collector. Samples were either analyzed immediately or stored at 4 °C until analysis. Chemical Analyses. The analysis of effluent samples for pesticide was performed by HPLC and for Cl- with a HaakeBuchler Digital Chloridometer. The HPLC system was equipped with a Perkin Elmer LC Pump Series 410, a LC-95 UV/Visible spectrophotometer detector (λ ) 220 nm), a Hewlett Packard ODS Hypersil column (5 mm packing and 125 × 4 mm), and a Rheodyne Model 7125 syringe injector with a 100-µL sample loop. The flow rate of the HPLC mobile phase, made of 60% methanol and 40% water, was set at 1
mL/min. The retention time for atrazine under these conditions was 4.8 min. The effluent samples from soil columns were directly injected into the machine. The detection limit of this method was ∼5 µg/L. Evaluation of Transport Parameters. The unknown parameters for atrazine transport through the soil columns include the distribution coefficient (Kd), the dispersion coefficient (D), the fraction of equilibrium adsorption sites (f), and the first-order rate constant (R). For each material used in packing columns, the value of Kd was determined independently by batch experiments. For D, the breakthrough curve (BTC) of the nonreactive tracer Cl- was fitted to the solution of the equilibrium ADE (22) with R fixed to unity (Kd ) 0). The value of D was adjusted iteratively until the minimum sum of squared residuals between the solution and measured data was reached. After Kd and D were determined, the flux concentration of atrazine for each of the uniform and short columns was fitted to the solution of the nonequilibrium ADE (eq 10) to estimate f and R by optimizing the dimensionless parameters β and ω. The numerical procedure for the curve fitting was based on the LevenbergMarquardt method, and the optimization program used was CXTFIT 2.0 (21). The BTCs for the layered columns were analyzed in three ways. First, mean travel times were calculated using the measured data itself. Second, parameters determined on columns that consisted of material which was (as close as possible) identical to the material in the layers were used for direct prediction. The outflow from the first layer, calculated using the parameters determined on a short column, was taken as the input function for the second layer. The predicted BTC output from the second layer was then obtained using the parameters determined on the unamended column. A subroutine was added to the CXTFIT 2.0 to specify the input function from the first layer to the second layer. Finally, once the transport parameters for the individual layers were confirmed by this direct prediction, the parameters obtained by fitting the outflow flux concentration of atrazine from a layered column were compared to those obtained from a uniform column to evaluate the relationship between pesticide movement and the distribution of OM in the soil.
Results and Discussion Distribution Coefficient. Amendment with OM increased atrazine adsorption in the soil. The measured Kd was 0.93 cm3/g for the unamended soil; was 1.37 and 2.15 cm3/g for the soil amended with manure and sludge, respectively, at
VOL. 31, NO. 8, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2333
FIGURE 1. Distribution coefficient measured at various OM/soil ratios. the rate of the uniform columns; and was 3.56 and 6.79 cm3/g for the soil amended with manure and sludge in the first OM layer. The adsorption isotherms were highly linear with linear correlation coefficients ranging from 0.972 to 0.997, all significant at P ) 0.01. When these Kd values were normalized by % OC of the soil mixes, the resulting Koc ranged from 51.27 to 89.77 (cm3/g), which agrees well with the reported values in the literature (23). We conducted additional batch experiments to test if a linear relationship exists between Kd and the application rate of OM. The results are plotted in Figure 1. The highest OM/ soil ratio (0.152 g/g) in this figure corresponds to that of the OM-amended layer in the layered columns. The correlation coefficient obtained was 0.967 and 0.992 for sludge and manure, respectively, and both were significant at P ) 0.01. The excellent linear correlation between these two variables suggests that a weighted retardation factor of a layered column, calculated using the length of individual layers as a weighting factor, might be a good estimation of the retardation factor of a uniform column that contains the same amount of OM, but homogenized with the soil. Mean Travel Time. The measured BTCs of atrazine through unamended and OM-amended columns are shown in Figure 2. Compared to the unamended columns, the travel time of atrazine appears to have increased in all the columns with OM, as manifested by the suppressed and/or delayed peak, and the pronounced tail of the BTCs. A quantitative evaluation of atrazine travel time in these columns can be achieved through time moment analyses (24-26) of the flux concentration measured at a fixed depth, C(t). By definition, the ith order time moment of a finite quantity of solute traveling through a porous medium is (24, 25)
Mi )
∫
∞ i
0
t C(t) dt
(15)
The mean travel time of atrazine normalized with respect to the total mass applied to the column in the miscible experiments can be calculated by
µ′1 )
M1 1 - T M0 2 0
(16)
where M0, the zero-order time moment, equals the total mass
2334
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 8, 1997
applied as a pulse, and T0 is the length of the pulse. The term 1/2T0 is to correct for a non-Dirac input function (27). The moment analysis for atrazine transport in the miscible experiments was performed on the measured flux concentrations using the trapezoidal rule (28). It should be noted that the mean travel time obtained through this numerical integration most likely underestimates the true mean to the extent that the mass eluted at concentrations below the method detection limit (C/C0 ) 0.001) or that mass remaining in the column at the end of the experiment was neglected. Deviation of this “mean” from the true mean would depend on the degree of nonequilibrium, becoming larger as BTC tailing increases. The results of time moment analyses are presented in Table 2. The travel time of atrazine is expressed as the number of pore volumes, instead of the absolute time, so that the effects of variability in the flow velocity and the length of packed soil among different columns are minimized. For the purpose of comparison, the travel time of Cl- is also reported. Inspection of Table 2 reveals that under similar flow conditions (fast or slow), the mean travel time of atrazine was ordered as sludge > manure > unamended soil, corresponding to the order of Kd measured in batch experiments. The mean travel time of nonsorbing tracer Cl-, however, was not affected or only negligibly affected by amendment with the OM. This kind of relationship between transport and Kd has been well documented for many organic chemicals (29-32). In fact, theoretically, the first normalized absolute moment of a chemical BTC (i.e., the mean travel time) is dependent solely on the retardation factor R (i.e., on Kd) (25). These approximated mean travel times probably represent a conservative estimation of the effects of organic amendments, since atrazine BTCs yielded from OM-amended columns tended to have longer tails than those of unamended columns (Figure 2). While mean travel times of atrazine were all increased in the OM-amended columns, the degree of retardation varied with the pattern in which the OM was distributed. Table 2 shows that transport of atrazine through the layered columns was retarded to a higher degree than that through the uniform columns, even though both the layered and the uniform columns contained the same amount of OM (9.8 g/column). Apparently the distribution of OM within the columns affects the transport of atrazine.
FIGURE 2. BTCs of atrazine measured in the miscible displacement experiments.
TABLE 2. Estimated Mean Travel Time (in Pore Volumes) for Atrazine and Clflow regime
amendment treatment
fast
sludge manure unamended sludge manure unamended
slow
uniform column atrazine Cl4.894 3.897 3.405 5.490 4.771 3.673
1.000 0.984 0.986 1.002 0.979 1.004
layered column atrazine Cl5.613 4.122
1.001 0.977
6.166 5.038
0.966 0.964
Direct Prediction by Transfer Function. In order to further evaluate this finding, we used the transfer function approach to test if the BTC from a layered column could be directly predicted, given the transport parameters known for the two individual layers. For this purpose, we conducted additional miscible displacement experiments using columns that contained only the first OM-amended layers of the layered columns (short columns). Using Kd values from the batch experiments, the atrazine flux concentrations from the short columns were fitted to eq 10. The fitted values of f and R, along with the batch-determined Kd values, are shown (in bold) in Table 3. The outflow from each OM-amended layer, calculated using the transport parameters fitted to the short column experiments, was taken as the input to the second unamended layer to directly predict the BTC from a layered column. The measured and predicted BTCs are plotted in
Figure 3, where the travel time in the x-axis for the layered column was scaled using the pore volume of the second layer. Overall, the transfer function approach was remarkably successful in predicting the transport of atrazine through the two-layer system for all the columns examined, given that the input to the transfer function calculations is based on batch and column experiments, which are independent of the layered column experiments. Barry and Parker (9) applied a similar approach, i.e., the convolution approximation, to calculate the BTC of Cl- through a two-layered column where the presence of mobile-immobile water was evident. They showed that the convolution approximation accurately predicted Cl- transport through the column. The agreement between the predicted and measured BTCs confirms that there was little interference in solute transport from the interface of the two layers and that the higher retardation of atrazine observed in the layered columns was, in fact, due to the difference in distribution of the OM. That is, the longer mean travel times observed for layered columns than for uniform columns are either totally or primarily due to the difference in OM distribution, not to some effect that is not included in the direct prediction process (e.g., to some effect of the interface itself). Nonequilibrium Transport Parameters. The estimated transport parameters for the uniform columns (R and f fitted, Kd from batch) and for the layered columns (Kd, R, and f fitted) are also listed in Table 3. The fitted curves for these columns are compared to the measured data in Figure 4. The
VOL. 31, NO. 8, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2335
TABLE 3. Estimated Nonequilibrium Transport Parameters of Atrazinea Kd b (cm3/g)
r (h-1)
f
r2
2.15 1.37 0.93 2.15 1.37 0.93
0.13 (0.12-0.14) 0.39 (0.37-0.42) 0.17 (0.16-0.19) 0.04 (0.04-0.04) 0.07 (0.07-0.08) 0.06 (0.05-0.07)
0.54 (0.49-0.58) 0.62 (0.60-0.64) 0.67 (0.64-0.71) 0.61 (0.59-0.63) 0.63 (0.54-0.72) 0.73 (0.70-0.77)
0.959 0.978 0.993 0.990 0.980 0.974
sludge manure sludge manure
Layered Column 1.97 2.20 (2.20-2.20) 1.28 1.21 (1.19-1.22) 1.97 2.36 (2.34-2.38) 1.37 1.80 (1.78-1.82)
0.29 (0.28-0.29) 0.54 (0.53-0.55) 0.10 (0.09-0.11) 0.05 (0.05-0.06)
0.50 (0.48-0.53) 0.64 (0.59-0.69) 0.52 (0.46-0.57) 0.66 (0.61-0.70)
0.991 0.985 0.988 0.986
sludge manure sludge manure
6.79 3.58 6.79 3.58
Short Column (OM Layer) 0.20 (0.20-0.21) 0.16 (0.15-0.17) 0.11 (0.10-0.12) 0.11 (0.11-0.12)
0.51 (0.46-0.56) 0.54 (0.48-0.60) 0.44 (0.33-0.56) 0.60 (0.55-0.65)
0.989 0.991 0.987 0.990
flow regime
amendment treatment
fast
sludge manure unamended sludge manure unamended
Uniform Column
slow
fast slow
fast slow
a Values in parentheses indicate 95% confidence limits in the fitting process. Values in bold are used in the “direct prediction by transfer function” calculations. b The Kd values for the uniform columns and the short columns are from batch sorption experiments. For the layered columns, the first Kd value is a length-weighted Kd, calculated from the retardation factors of individual layers. The second Kd value (and all the R and f values) are fitted from the flux concentrations.
FIGURE 3. Measured atrazine BTCs from layered soil columns in comparison to those directly predicted by transfer function using parameters determined for individual layers. The input flux concentration for the second (unamended) layer is the fitted outflow from the first (OMamended) layer. two-site nonequilibrium model describes atrazine BTCs well, with the coefficient of determination for all columns g 0.959. Amendment with OM changed the nonequilibrium transport behavior of atrazine. In general, a decreased f was obtained in OM-amended columns under both leaching velocities,
2336
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 8, 1997
implying that the majority of the sorption sites contributed by the OM were rate-limited sites. Except for the layered columns, the f and R values were obtained using distribution coefficients that were independently determined using batch experiments.
FIGURE 4. Comparison of atrazine BTCs through layered and uniform OM-amended soil columns. The optimized parameter values for the fitted BTCs are reported in Table 3. The fitted overall Kd was used for layered columns. The batch adsorption procedure we used for estimating Kd employed 24 h as the equilibration period for all soil treatments, chosen in part to minimize the effect of atrazine degradation on the results. This period may be inadequate to permit the sorption to fully reach equilibrium and, therefore, may underestimate the real distribution coefficient (33). This would be particularly true for soils amended with OM if many of the sorption sites associated with the OM are nonequilibrium sites. Since using a larger Kd would lower the f estimated by curve fitting to a given BTC, a larger difference in f values would be expected between the amended soil and soils amended with the OM if a longer equilibration time were allowed for the batch sorption experiments. Table 3 also reveals that f is related to Kd or, more precisely, % OC. Amendment with sludge, which increased Kd more than did manure, resulted in a lower f. We observed a similar trend for alachlor elution data in concurrent experiments with sterilized columns (data not shown). Such a relationship is consistent with the conclusion that the majority of OM sorption sites were nonequilibrium sites, as discussed previously. However, in the study reported by Brusseau (34), who conducted transport experiments on four hydrophobic organic compounds in three aquifer materials, the f values were not found to vary with % OC in a similar fashion. The nonequilibrium transport of these low-polarity compounds may have been dominated by physical processes, i.e., mass diffusion between mobile and immobile water regions (35). As noted by other researchers (36-38), the rate constant (R) was dependent on flow velocity, being higher for fast flow
than slow flow in all columns (Table 3). These results indicate that the rate-limited adsorption may have been caused by retarded mass transfer to intraparticle or intrasorbent adsorption sites (39-41), rather than by a slow adsorption reaction. The enhanced physical mixing of the solute at the fast flow, as suggested by a higher dispersion coefficient (Table 1), probably facilitated mass transfer to the interior sorption sites of the sorbents and thus resulted in a higher R. The values of the fitted overall Kd for layered columns are, except for the manure-amended column leached at fast flow, greater than both the weighted Kd calculated from the R values of the two individual layers and the Kd determined independently by batch experiments (uniform columns section of Table 3). For manure-amended columns subjected to the fast flow, the BTCs are virtually indentical for the two distributions (Figure 4). However, when the flow rate was reduced, we found again a greater retardation in the layered column. Experiments with these two columns were repeated, and the same results were observed. Overall, we found, when compared to the uniform column, the layered column either had a higher overall Kd, a higher f, or both, indicating an enhanced accessibility to the sorption sites of the organic amendments. Although transport of atrazine via nonequilibrium process was determined by the combined effects of all transport parameters, this conclusion is supported by the longer travel times of atrazine observed for the layered columns in contrast to those of the nonsorbing Cl-, which were essentially the same for both types of columns (Table 2).
VOL. 31, NO. 8, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2337
Effects of Layer Ordering. In studying chemical transport in heterogeneous soil profiles, Selim et al. (16) conducted experiments on transport of 2,4-D through a two-layered soil column and found that the effluent concentration was not affected by the order of layers. Similar observations have also been reported by Panigatti for fluometuron (cited in ref 16). We checked the effect of layer ordering on atrazine transport using a layered, sludge-amended column. After completion of the first leaching, a second pulse was applied to the same column from the other end. No discernible difference between the two BTCs was observed (data not shown). These results demonstrate that it is the concentration of OM in the soil, not the placement of OM, that affects the transport. The enhanced accessibility to the adsorption sites of the OM in the layered columns probably resulted from less shielding on OM sorbents by soil minerals due to an increased OM concentration in the soil. A similar interaction between OM and clay has been described by Koskinen and Harper (42).
Acknowledgments This research was supported by funding from the U.S. Department of Agriculture, BARD Project US2268-93C. We would like to thank Dr. W. Chen for valuable suggestions and discussions during the preparation of this manuscript. We also thank the journal reviewers whose insights and comments greatly improved this manuscript.
Literature Cited (1) Karichoff, S. W. J. Hydraul. Eng. 1984, 110, 707-735. (2) Chiou, C. T.; Porter, P. E.; Schmedding, D. W. Environ. Sci. Technol. 1983, 17, 227-231. (3) Felsot, A. S.; Dahm, P. A. J. Agric. Food Chem. 1979, 27, 557-563. (4) Shea, P. J. Weed Sci. Suppl. 1985, 2, 33-41. (5) Clapp, C. E.; Larson, W. E.; Harding, S. A.; Titrul, G. O. Utilization of Sewage Wastes on Land, Research Progress Report; Agricultural Research Service, U.S. Department of Agriculture; University of Minnesota: St. Paul, MN, 1980; pp D1-D23. (6) van Genuchten, M. Th.; Wierenga, P. J. Methods of Soil Analysis, Part I. Physical and Mineralogical Methods; Agronomy Monograph 9 (2nd ed.); Soil Science Society of America: Madison, WI, 1986; pp 1025-1054. (7) Biggar, J. W.; Nielsen, D. R. Agronomy 1967, 11, 254-274. (8) Kreft, A.; Zuber, A. Chem. Eng. Soc. 1978, 33, 1471-1480. (9) Barry, D. A.; Parker, J. C. Transp. Porous Media 1987, 2, 65-82. (10) Barry, D. A.; Parlange, J.-Y.; Rand, R.; Parker, J. C. Transp. Porous Media 1988, 2, 421-423. (11) Jury, W. A.; Roth, K. Transfer functions and solute movement through soil: Theory and applications; Birkha¨user Verlag Basel: Birkha¨user, 1990; pp 85-104. (12) Jury, W. A.; Utermann, J. Transp. Porous Media 1992, 8, 277297. (13) Moranville, M. B.; Kessler, D. P.; Greenkorn, R. A. AIChE J. 1977, 23, 786-794. (14) Bruch, J. C. Water Resour. Res. 1970, 6, 791-800. (15) Shamir, U. Y.; Harleman, D. R. F. Proc. Am. Soc. Civ. Eng. Hydr. Div. 1967, 93, 236-260. (16) Selim, H. M.; Davidson, J. M.; Rao, P. S. C. Soil. Sci. Soc. Am. J. 1977, 41, 3-10. (17) Hutson, L. J.; Wagenet, R. J. LEACHM: Leaching Estimation and Chemistry Model, Version 3; New York State College of Agriculture and Life Sciences, Cornell University: Ithaca, NY, 1996.
2338
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 8, 1997
(18) van Genuchten, M. Th.; Wagenet, R. J. Soil Sci. Soc. Am. J. 1989, 53, 1303-1310. (19) Jury, W. A.; Sposito, G.; White, R. E. Water Resour. Res. 1986, 22, 243-247. (20) Parker, J. C.; van Genuchten, M. Th. Determining transport parameters from laboratory and field tracer experiments; Bulletin 84-3; Virginia Agricultural Experiment Station: Blacksburg, VA, 1984. (21) Toride, N.; Leij, F. J.; van Genuchten, M. Th. The CXTFIT code for estimating transport parameters from laboratory or field tracer experiments, Version 2.0; U.S. Salinity Laboratory, Agricultural Research Service, U.S. Department of Agriculture: Riverside, CA, 1995. (22) van Genuchten, M. Th.; Alves, W. J. Analytical solutions of the one-dimensional convective-dispersive solute transport equation; Technical Bulletin 1661; U.S. Department of Agriculture: Riverside, CA, 1982. (23) Montgomery, J. H. Agrochemicals desk reference: Environmental data; Lewis Publishers: Chelsea, MI, 1993; pp 31-38. (24) Aris, R. Proc. R. Soc. London 1958, A245, 268-277. (25) Valocchi, A. J. Water Resour. Res. 1985, 21, 808-820. (26) Parker, J. C.; Valocchi, A. J. Water Resour. Res. 1986, 22, 399-407. (27) Brusseau, M. L.; Jessup, R. E.; Rao, P. S. C. Environ. Sci. Technol. 1990, 24, 727-735. (28) Jury, W. A.; Gardner, W. R.; Gardner, W. H. Soil Physics; John Wiley & Sons: New York, 1991; pp 243-254. (29) Chiou, C. T. Ecological Studies 73; Gerstl, Z., Chen, Y., Mingelgrin, U., Yaron, B., Eds.; Springer-Verlag: New York, 1989; pp 163175. (30) Rao, P. S. C.; Jessup, R. E. Chemical Mobility and Reactivity in Soil Systems; Nelson, D. W., Elrick, D. E., Tanji, K. K., Eds.; Soil Science Society of America: Madison, WI, 1983; pp 183-201. (31) Jury, W. A.; Focht, D. D.; Farmer, W. J. J. Environ. Qual. 1987, 16, 422-428. (32) O’Connor, G. A.; Wierenga, P. J.; Cheng, H. H.; Doxtader, K. G. Soil Sci. 1980, 130, 157-162. (33) Pignatello, J. J.; Xing, B. Environ. Sci. Technol. 1996, 30, 1-11. (34) Brusseau, M. L. J. Contam. Hydrol. 1992, 9, 353-368. (35) Brusseau, M. L.; Jessup, R. E.; Rao, P. S. C. Water Resour. Res. 1989, 25, 1971-1988. (36) Gaber, H. M.; Inskeep, W. P.; Comford, S. D.; Wraith, J. M. Soil Sci. Soc. Am. J. 1995, 59, 60-67. (37) Gamerdinger, A. P.; Lemley, A. T.; Wagenet, R. J. J. Environ. Qual. 1991, 20, 815-822. (38) Brusseau, M. L.; Jessup, R. E.; Rao, P. S. C. Environ. Sci. Technol. 1991, 26, 134-142. (39) Hamaker, J. W.; Goring, C. A. I.; Youngson, C. R. Organic Pesticides in the Environment; Gould, F. F., Ed.; Advances in Chemistry Series 60; American Chemical Society: Washington, DC, 1966; pp 23-37. (40) Rao, P. S. C.; Jessup, R. E.; Addiscott, T. M. Soil Sci. 1982, 133, 342-349. (41) Brusseau, M. L.; Larsen, T.; Christensen, T. H. Water Resour. Res. 1991, 27, 1137-1145. (42) Koskinen, W. C.; Harper, S. S. Pesticides in the Soil Environment: Processes, Impacts, and Modeling; Cheng, H. H., Bailey, G. W., Green, R. E., Spencer, W. F., Eds.; Soil Science Society of America Book Series 2; Soil Science Society of America: Madison, WI, 1990; pp 51-78.
Received for review November 6, 1996. Revised manuscript received March 25, 1997. Accepted March 28, 1997.X ES960944H X
Abstract published in Advance ACS Abstracts, June 1, 1997.