Transport of unexploded ordnance compounds through soils
the kinematic viscosity of air, and Sc is given by /DA. The
reaction rate constant k1 was set to zero in all simulations. All
parameters were assumed to be constant during the simulation.
Wicking can be an important transport mechanism when
continuous replenishment of soil moisture occurs from beneath
the top layers of soil during surface drying. The laboratory
experiments described in this manuscript were performed with
a very thin layer of soil (2 cm) that had a low moisture content
Environ. Toxicol. Chem. 21, 2002
Table 4. Sediment properties for simulation
Water-
2023
Air-filled
Initial soil Air relative porosity porosity porosity
Experiment moisture humidity ⑀a ⑀w ⑀T
filled
Total
(
)
(
)
( )
Case 1
Case 2
Case 3
5%
20%
5%
Humid
Dry
Dry
0.08
0.16
0.31
0.23
0.15
0
0.31
0.31
0.31
(
5 and 20%) and no water replenishment from below.
The losses from any possible wicking can be estimated by
0
kadv
·
C
air
, where kadv is the advective velocity that may be in-
calculated to be 0.08 and was maintained at that value through-
out the simulation period. With humid air passing over the
surface, the soil was considered to be wet at that soil moisture
content, and therefore the KSA was estimated from KSW and
KAW as per Equation 4.
0
duced by wicking and
C is the equilibrium vapor phase con-
air
centration of the chemical in the soil. The advective velocity
of pore air saturated with water vapor was calculated from a
previous study [5] and corrected for the difference in airflow
rates used in both studies based on soil to air mass transfer
coefficients in the literature [9] and was found to be 0.034
cm /cm /h. The advective fluxes thus computed for the three
chemicals in cases 1 and 3 shown in this manuscript range
from 7.4
measured at times below 72 h (time period when steady state
water evaporation is expected to occur) range from 0.0083 to
7.2 ng/cm /h. The ratios of the computed advective fluxes to
the corresponding measured experimental fluxes range from
.00044 to 0.05. In the worst case, the advective fluxes are
% of the measured fluxes. This analysis shows that for chem-
icals that exhibit very low vapor pressures and very high par-
tition constants, wicking was not a significant mechanism for
vapor phase transport and was therefore neglected in the trans-
port model described in this manuscript.
Case 2. The initial moisture content of the soil was 20%,
and the relative humidity of the air passing over the soil surface
was 0%. The soil moisture filled almost all the soil pores
initially, but with moisture loss to air, the air-filled porosity
was expected to increase. Earlier reports of water evaporation
rates from similar experiments [5] showed that initially water
loss occurs very rapidly and that complete water loss takes
much longer. The air-filled porosity would increase as a func-
tion of time during this period, but in the absence of any
transient measured data, the air-filled porosity was set as an
average of the initial (zero) and the final expected value (total
porosity). For the duration of the experiment, the soil was
presumed to retain enough moisture to be considered wet, and
the KSA was estimated from KSW and KAW as per Equation 4.
Case 3. The initial moisture content of the soil was 5%,
and the relative humidity of the air passing over the soil surface
was 0%. The moisture from the pore-air space was expected
to decrease within a very short period of time, and therefore
the value of the air-filled porosity was set equal to that of the
total porosity throughout the simulation period. The rapid dry-
ing creates the possibility of dry-off of the soil surface. In this
3
2
Ϫ
6
2
ϫ 10 to 0.025 ng/cm /h. The experimental fluxes
2
1
0
5
Experimental soil–water adsorption constants
The aquifer soils from LAAP were higher in sand, ranging
from 27 to 77% sand and low in organic carbon (Table 1). Silt
and clay were present in all samples, although in lower
amounts. Cation exchange capacity (CEC) was also relatively
low, ranging from 6.6 to 15.5 meq/100 g. This is in marked
contrast to the Yokena clay surface soil that was high in clay,
organic carbon, and CEC (Table 1) compared to the aquifer
soils.
case, KSA cannot be approximated as KSW/HC and was not
directly measured either. As an adjustable parameter, KSA was
used in the model for this case. The porosity values used in
each case of the simulation are summarized in Table 4.
Figures 1 to 3 show the comparison of experimental fluxes
and the corresponding model simulations for 2,4-DNT, 2,6-
The experimentally measured soil–water partition constants
KSW) for 2,4-DNT, 2,6-DNT, and 1,3-DNB are listed in Table
. The KSW values for the Yokena clay surface soil were found
(
3
DNT, and 1,3-DNB, respectively, at 24ЊC from three different
to be larger than those for the aquifer soils. Previous work on
a UXO (2,4,6-trinitrotoluene) showed that the KSW for nitro-
compounds was strongly correlated to the CEC and clay con-
tent of the soil [10,11]. In the present work, we realized a
similar trend. The KSW for all four compounds increased with
CEC, clay content, and organic carbon fraction of the soils. It
is clear that sorption of these compounds is higher in the
surface soil (Yokena clay) than in aquifer soils (LAAP-C and
LAAP-D).
soils for cases 1 and 2. These three compounds were chosen
for comparison since the data set was most complete for these.
Figure 1a, b, and c shows the flux of 2,4-DNT from LAAP-
C, LAAP-D, and Yokena soils, respectively. The experimental
flux values are represented by discrete symbols, while the sim-
ulation curves are represented by lines. The experimental data
show the flux of 2,4-DNT decreasing gradually with time and
quickly reaching a steady state. Initially, the flux is air-phase-
resistance controlled. Very quickly, it becomes soil-side-re-
sistance controlled, and diffusion through the soil pores dom-
Experimental data and model simulations
0
inates. Thus, the initial flux can be given by NA
and as becomes large, the long-term flux is given by NA
R D ). The diffusion-controlled flux is proportional
. Any deviation from this behavior is indicative of pro-
(
t
)
ϭ
kaC
,
A
Experimental data analyzed using the model was classified
under three cases on the basis of the experimental conditions
t
ϭ
0
C
͙
t
(
½
/t
eff
A
f
(
initial soil moisture content and relative humidity of air pass-
to 1/
ing over soil surface).
cesses other than diffusion also being significant. Figures 2
and 3 show similar trends in the experimental fluxes of 26-
DNT and 13-DNB, respectively, for cases 1 and 2 for the three
soils.
The model curves in each case shown in Figures 1 to 3
were not a priori simulations but best fits of the experimental
Case 1. The initial moisture content of the soil was 5%,
and the relative humidity of the air passing over the soil surface
was 100%. The soil pore air was water saturated and hence
was expected to retain the initial moisture content since no
moisture loss is expected. The initial air-filled porosity was