28
S. C. Lessoff and P. Indelman
4 Results
Reactive properties of the herbicides were estimated by inverse simulation as
described above. The estimated infiltrating tracer mass was Mo = 70% which
agrees with the 75% mass recovery on the field surface in Petri dishes. The
estimated values of A, are in agreement with independent laboratory-measurements
(Table 1). The field estimated values for Kd of BRM and TBA are less than
laboratory-measured values, while for A TR the field and laboratory estimates are
very similar (Table 1). The relative behavior of the tracers is consistent with their
reported behavior K)TBA) > K)ATR) > K)BRM) (Flury 1996) . Concentration profiles
computed by the model using the best-fit properties compare favorably to
measured reactive tracer concentration profiles at all times and depths (see Fig. 4).
To properly understand the field results it is necessary to consider the
sensitivity of the identified values to measurement uncertainty (see Figure 5). This
includes both uncertainty concentration measurements as well as uncertainty in
flow parameters f.1w and CV w that were identified from conservative tracer
transport experiments.
TBA transport is sensitive to both Kd and A,. The confidence interval for the
KjTBA) is approximately 0.3 < Kd < 0.8. For the Bet Dagan soil (whose organic
carbon content in the top layers is about 1 %) this corresponds to 30 < Koc < 80,
where Koc is the organic carbon sorption coefficient. Indelman et al. (1998)
estimated Koc = 14 also from field data. These values are well below both
laboratory estimates of sorption coefficients at Bet Dagan Koc = 200 and those
reported in the literature Koc = 357 (Flury 1996). The field estimates are low even
when considering the full range of values of Koc in the literature from a minimum
of 93 by Walker and Blacklow (1994) to a maximum of 1195 by Rao and
Davidson (1980).
The estimated interval of confidence for the degradation rate of TBA is
om 5 < A, < 0.06. Indelman et al. (1998) estimated A, = 0.005 day-I which is less
than measured here. Note that the A, interval of confidence derived here is within
the range of values reported in the literature 0.007 to 0.023 day-I (Flury 1996).
For A TR, the model is reasonably sensitive to both Kd and A,. The confidence
interval for sorption is 0.1 < Kd < 0.6 ml g-I corresponding to 10 < Koc < 60 ml g-I.
This is in agreement with Kd = 0.4 ml g-I measured by Shapir and Mandelbaum
(1997) in laboratory cores from Bet Dagan, but it is less than laboratory batch
measurements reported in the literature, e.g., Koc=100 ml g.1 (Flury 1996). In
column studies, Chen and Wagenet (1997) have observed that Atrazine sorption
can be rate-limited with sorption coefficients ranging from Koc = 20 ml g-I in
short-term experiments to Koc = 60 ml g-I in long-term experiments. The
estimation range for the degradation rate of Atrazine in our study is
0.01 < A, < 0.055 day-I. This is in good agreement with laboratory assessment
A, = 0.024 dai l in the top soil layer (Shapir and Mandelbaum 1997), as well as
with literature values 0.01 day-I (Flury 1996).
S. C. Lessoff and P. Indelman
4 Results
Reactive properties of the herbicides were estimated by inverse simulation as
described above. The estimated infiltrating tracer mass was Mo = 70% which
agrees with the 75% mass recovery on the field surface in Petri dishes. The
estimated values of A, are in agreement with independent laboratory-measurements
(Table 1). The field estimated values for Kd of BRM and TBA are less than
laboratory-measured values, while for A TR the field and laboratory estimates are
very similar (Table 1). The relative behavior of the tracers is consistent with their
reported behavior K)TBA) > K)ATR) > K)BRM) (Flury 1996) . Concentration profiles
computed by the model using the best-fit properties compare favorably to
measured reactive tracer concentration profiles at all times and depths (see Fig. 4).
To properly understand the field results it is necessary to consider the
sensitivity of the identified values to measurement uncertainty (see Figure 5). This
includes both uncertainty concentration measurements as well as uncertainty in
flow parameters f.1w and CV w that were identified from conservative tracer
transport experiments.
TBA transport is sensitive to both Kd and A,. The confidence interval for the
KjTBA) is approximately 0.3 < Kd < 0.8. For the Bet Dagan soil (whose organic
carbon content in the top layers is about 1 %) this corresponds to 30 < Koc < 80,
where Koc is the organic carbon sorption coefficient. Indelman et al. (1998)
estimated Koc = 14 also from field data. These values are well below both
laboratory estimates of sorption coefficients at Bet Dagan Koc = 200 and those
reported in the literature Koc = 357 (Flury 1996). The field estimates are low even
when considering the full range of values of Koc in the literature from a minimum
of 93 by Walker and Blacklow (1994) to a maximum of 1195 by Rao and
Davidson (1980).
The estimated interval of confidence for the degradation rate of TBA is
om 5 < A, < 0.06. Indelman et al. (1998) estimated A, = 0.005 day-I which is less
than measured here. Note that the A, interval of confidence derived here is within
the range of values reported in the literature 0.007 to 0.023 day-I (Flury 1996).
For A TR, the model is reasonably sensitive to both Kd and A,. The confidence
interval for sorption is 0.1 < Kd < 0.6 ml g-I corresponding to 10 < Koc < 60 ml g-I.
This is in agreement with Kd = 0.4 ml g-I measured by Shapir and Mandelbaum
(1997) in laboratory cores from Bet Dagan, but it is less than laboratory batch
measurements reported in the literature, e.g., Koc=100 ml g.1 (Flury 1996). In
column studies, Chen and Wagenet (1997) have observed that Atrazine sorption
can be rate-limited with sorption coefficients ranging from Koc = 20 ml g-I in
short-term experiments to Koc = 60 ml g-I in long-term experiments. The
estimation range for the degradation rate of Atrazine in our study is
0.01 < A, < 0.055 day-I. This is in good agreement with laboratory assessment
A, = 0.024 dai l in the top soil layer (Shapir and Mandelbaum 1997), as well as
with literature values 0.01 day-I (Flury 1996).