The observation data are used to judge the quality of the current parameter set p. Their analysis provides usually the main share of the current error f (p) after a calculation of the direct problem. Currently, the following observation data are possible:

 

Potential heads at single nodes of the model (attribute POTE)

Leakage rates (i.e. from discharge measurements) as sum of a defined section of a water course, which is defined by the attribute LERA (attribute LKNO, only inverse modelling),

Mass flow rates as sum for a defined number of nodes, which are assigned to the attribute POTE in the model file (attribute KNOT).

 

All data can be available at different times in transient calculations.

Which observation data are used, is regardless of the choice of model parameters to be optimized. For example, K-values can be optimized by observed potential and leakage rates, without optimizing leakage coefficients simultaneously. In contrast, for the optimization of leakage rates the existence of observed leakage data is not of necessity.

The number of available observation data has a significant influence on the success of the inverse modelling. The more observation data are available, the greater the prospect of finding an "optimum" parameter set. A successful optimization is achieved only if the zoning of the parameters to be optimized is reasonable. To define a reasonable zoning is the main task in the inverse modelling.

The observation data must be defined for an optimization in a separate file. This file is mentioned in the following as “observation file”. The structure of the observation file is described in chapter: "Realization in SPRING".

 

Zoning of the parameters