|
The gradient method is a program-based calibration method. The permeabilities are based on predetermined start values and are iteratively changed element-wise so that a measured and interpolated groundwater table can be simulated as well as possible.
| The iterative change of the horizontal kf-values takes places via a comparison of the desired and actual potential gradient. The changes in the K-values from one iteration step to the next can be attenuated on different, by the user predetermined manners. As a result, oscillations based on improper starting values and extreme variations between iterations may be avoided. |
 |
1. Simple attenuation of the change of a kf-value of an element:
Instead of the newly calibrated kf-value kn+1 (see above) the value knew = kn + x * (kn+1 - kn) with 0 < x <= 1 is used.
2. Local attenuation of the relative change of the kf-value in an adjacent element:
With this attenuation the oscillations of the kf-values in adjacent elements based on the calibration iteration can be decreased. For each element, the calculated relative change (in %) of the kf-value is again smoothed in relation to the relative changes of the kf-values in the adjacent elements.
| Element |
kf-value of the last iteration step |
calculated new kf-values |
%-change of the kf-values |
|
1
|
0,40
|
0,01
|
2,50 %
|
|
2
|
0,30
|
0,45
|
150,00 %
|
|
3
|
0,35
|
0,70
|
200,00 %
|
|
4
|
0,3
|
0,15
|
50,00 %
|
|
 |
|
