Results


It is recommended to perform a calculation with 0 iterations first. So without much effort, the target function can be controlled. The results are written in the output file (here: out.inv). The shares for each parameter/measurement group are listed individually.

It is important to balance the shares of the target function. After the first inverse calculation (0 iterations), the shares of the target function are compared. The best-known sizes (usually the potential heads) should have the largest share of the target function (ZF).

 

Possible sources of errors:

K-values are weighted too heavily. This forbids the algorithm to calculate the K-values far away from the estimated value, so there are almost no changes in the K-values.

Remedy: Increasing the standard deviation

Quantities are weighted too heavily. Thus the measured potential heads lose their weight and the algorithm tries (in vain!) to bring the volumes to match. In most cases, the rule applies: If the potentials and the K-values fit, the amounts are also set in.

Remedy: Weighting the shares of the target function balanced (using well-chosen standard deviations)

Transient hydrographs: The piezometer x is measured every day for over a year, in contrast, piezometer y is measured only once per month. When all measurements have the same (i.e. equally weighted), the well x is weighted about 30 times stronger than y (365 measurements compared to 12 measurements).

Remedy: Choosing a less for well y

 

If the shares of the target function are balanced, the number of iterations is set to 10 and the inverse calculation is started again.

 

Output files

out.inv

File with the protocol of the calibration (measured / calculated value / target function) and the K-values and leakage factors in the format of the model file. It is possible to assign the iterated K-values directly to the attribute KWER via the menu item Attributes Import data/computational results iterated K-values.

Invpar.csv:

File in the format of the parameter file (Note: semicolon as the column separator, point as decimal). In column D, the calculated K values are listed. This file can be renamed and then used as input file for a next calculation (e.g. with partially constant values).

Invpar_xy.csv:

Analogous to the file invpar.csv, but additionally with coordinates (not for LERA!). The file can be overlayed, for example, to visualize the results.

resan_pote.csv, resan_lkno.csv, resan_knot.csv (steady state only):

Contain all data of the measured values

 


Example of the file resan_pote.csv

 

 

name.csv (transient only):

For each measurement point (POTE, LKNO, KNOT) a *.csv file is stored with time series data. There will be no difference according to data types, so it is important that the names in the observation file (*.obs) are clearly defined (not a potential head called X1 and a quantity measurement called X1 (better: h_x1 and q_x1).
 


Example of a transient output file name.csv

 

Batch files