8.1. Simulation and Prediction ofGroundwater Pollution
249
a. Geometric parameters of the aquifers (boundary shape, thickness, elevations of the roof and bottom);
b. Initial distributions of hydraulic head and concentration;
c. Locations and rates of the pumping wells (or injection wells), locations
and intensities of the pollution sources;
d. Hydraulic relationships with adjacent aquifers and surface water;
e. Estimations of various hydrogeological parameters, including porosity,
hydraulic conductivity, longitudinal and trans verse dispersivities of the
aquifer;
f. Historical observation data and the data obtained from field tests.
For so me of the data mentioned above, we can use their estimated values
first, and further determine more accurate values of them through model
calibration.
6. Model Calibration and Reliability Analysis
Model calibration and reliability analysis have already been illustrated in
Chapter 7. Hydraulic conductivity and porosity can be determined through
the calibration of the flow model. However, if observed water quality da ta
is available, better results may be obtained by simultaneously calibrating
the flow model and the water quality model. The local dispersivities can
be determined by fitting the observed results of tracer injection tests with
the output curves of corresponding analytic solutions or numerical solutions. Based on these local values, distributed parameters in the whole region
can be composed by kriging interpolation. Next, substitute these distributed
parameters into the water quality model to simulate the existing pollution
or tracer tests, and then check if the simulated results coincide with the
observations. If not, the trial-and-error method or other optimal methods
should be used to modify the parameters until a satisfactory fitting is
achieved.
Reliability analysis of the final model should be conducted to estimate
the statistical characteristics of model output (concentration distribution or
arrival time) and examine if the requirements of accuracy are satisfied. It is
better to verify the model by the data that has not been used for model
calibration.
7. Prediction and Control
With a reliable mathematical model, predictions can be made according to
the concrete tasks as mentioned above. The model can be used to predict
the developing tendency of contamination, to compare the results between
various remedia ti on and control schemes, to draw the concentration contours at different time. It can also be taken as a component in a water
resources management model.
249
a. Geometric parameters of the aquifers (boundary shape, thickness, elevations of the roof and bottom);
b. Initial distributions of hydraulic head and concentration;
c. Locations and rates of the pumping wells (or injection wells), locations
and intensities of the pollution sources;
d. Hydraulic relationships with adjacent aquifers and surface water;
e. Estimations of various hydrogeological parameters, including porosity,
hydraulic conductivity, longitudinal and trans verse dispersivities of the
aquifer;
f. Historical observation data and the data obtained from field tests.
For so me of the data mentioned above, we can use their estimated values
first, and further determine more accurate values of them through model
calibration.
6. Model Calibration and Reliability Analysis
Model calibration and reliability analysis have already been illustrated in
Chapter 7. Hydraulic conductivity and porosity can be determined through
the calibration of the flow model. However, if observed water quality da ta
is available, better results may be obtained by simultaneously calibrating
the flow model and the water quality model. The local dispersivities can
be determined by fitting the observed results of tracer injection tests with
the output curves of corresponding analytic solutions or numerical solutions. Based on these local values, distributed parameters in the whole region
can be composed by kriging interpolation. Next, substitute these distributed
parameters into the water quality model to simulate the existing pollution
or tracer tests, and then check if the simulated results coincide with the
observations. If not, the trial-and-error method or other optimal methods
should be used to modify the parameters until a satisfactory fitting is
achieved.
Reliability analysis of the final model should be conducted to estimate
the statistical characteristics of model output (concentration distribution or
arrival time) and examine if the requirements of accuracy are satisfied. It is
better to verify the model by the data that has not been used for model
calibration.
7. Prediction and Control
With a reliable mathematical model, predictions can be made according to
the concrete tasks as mentioned above. The model can be used to predict
the developing tendency of contamination, to compare the results between
various remedia ti on and control schemes, to draw the concentration contours at different time. It can also be taken as a component in a water
resources management model.