8.1. Simulation and Prediction of Groundwater Pollution
251
deposits were absent, and areas which were unsaturated most of the time.
These areas formed internal barriers, which greatly afTected the pattern of
groundwater flows in the aquifers.
The study area was about 88 km 2 , and the iterative alternative direction
implicit (IADI) finite difTerence method was used to solve the groundwater
flow equation. The whole area was divided into 25 x 38 square meshes, with
the side of each square being 308 m long.
According to aerial survey photographs, the irrigation area occupied
about 111 squares. It is interesting that the sink/source term, including the
pure rech arge rate of irrigation, rainfall infiltration, the leakage of ditches,
and so forth, was determined by model calibration. Pumping rate from each
weIl was 50 m 3 /d on average. This procedure of calibration provided good
reference for the establishment of a regional flow model. The flow directions
in the whole domain, as calculated by the flow model, were not identical, and
the velocities ranged from 0.3 to 6.1 m/d.
The third phase of the study was the building of a solute transport model.
The two-dimensional advection-dispersion model was adopted and solved by
the method of characteristics introduced in Section 4.2. The distribution of
chloride concentration observed in 1956 was taken as the basis of model
calibration. The end of 1943 was chosen as the initial time. The simulated
results were compared with the observed data of 14 years. Through the
model calibration, the porosity obtained was 0.3 and the longitudinal and
trans verse dispersivities were both 30 m. Then, a verification of the water
quality model was made based on the data from 1956 to 1960, 1961 to
1968, and 1968 to 1972. The computed concentration distributions agreed
fairly weIl with the observed distributions.
The fourth phase was using the calibrated model to predict the development of contamination and evaluating the remediation strategy. According
to the prediction, chloride would be eventually drained from the aquifer of
the simulated domain, but the recovery of the water quality to its original
state would require at least several decades. In order to speed up the remediati on, researchers suggested digging a drainage ditch at the northern boundary
of the factory to extract the polluted water before it migrated from the
factory to downstream. The efTectiveness of the scheme may be pre-estimated
by the model. The model results showed that it was almost impossible to
remove all contaminants from most of the polluted aquifers at the Rocky
Mountain munitions site. However, some portions of the aquifers may be
afTected by artificial plans, which may speed up the improvement of water
quality in those aquifers. The feasibility of such plans largely depends on
the hydraulic features of the aquifers, the types of contamination, and the
pollution sources. To pre-estimate the efTectiveness of aremediation plan,
a reliable prediction of concentration changes must be made, which would
rely on the model output. Thus, the role of the water quality models was fully
affirmed.
251
deposits were absent, and areas which were unsaturated most of the time.
These areas formed internal barriers, which greatly afTected the pattern of
groundwater flows in the aquifers.
The study area was about 88 km 2 , and the iterative alternative direction
implicit (IADI) finite difTerence method was used to solve the groundwater
flow equation. The whole area was divided into 25 x 38 square meshes, with
the side of each square being 308 m long.
According to aerial survey photographs, the irrigation area occupied
about 111 squares. It is interesting that the sink/source term, including the
pure rech arge rate of irrigation, rainfall infiltration, the leakage of ditches,
and so forth, was determined by model calibration. Pumping rate from each
weIl was 50 m 3 /d on average. This procedure of calibration provided good
reference for the establishment of a regional flow model. The flow directions
in the whole domain, as calculated by the flow model, were not identical, and
the velocities ranged from 0.3 to 6.1 m/d.
The third phase of the study was the building of a solute transport model.
The two-dimensional advection-dispersion model was adopted and solved by
the method of characteristics introduced in Section 4.2. The distribution of
chloride concentration observed in 1956 was taken as the basis of model
calibration. The end of 1943 was chosen as the initial time. The simulated
results were compared with the observed data of 14 years. Through the
model calibration, the porosity obtained was 0.3 and the longitudinal and
trans verse dispersivities were both 30 m. Then, a verification of the water
quality model was made based on the data from 1956 to 1960, 1961 to
1968, and 1968 to 1972. The computed concentration distributions agreed
fairly weIl with the observed distributions.
The fourth phase was using the calibrated model to predict the development of contamination and evaluating the remediation strategy. According
to the prediction, chloride would be eventually drained from the aquifer of
the simulated domain, but the recovery of the water quality to its original
state would require at least several decades. In order to speed up the remediati on, researchers suggested digging a drainage ditch at the northern boundary
of the factory to extract the polluted water before it migrated from the
factory to downstream. The efTectiveness of the scheme may be pre-estimated
by the model. The model results showed that it was almost impossible to
remove all contaminants from most of the polluted aquifers at the Rocky
Mountain munitions site. However, some portions of the aquifers may be
afTected by artificial plans, which may speed up the improvement of water
quality in those aquifers. The feasibility of such plans largely depends on
the hydraulic features of the aquifers, the types of contamination, and the
pollution sources. To pre-estimate the efTectiveness of aremediation plan,
a reliable prediction of concentration changes must be made, which would
rely on the model output. Thus, the role of the water quality models was fully
affirmed.