Access via: SPRING Calculation Inverse calibration…

A distinction is made according to the type of calculation:

 

 

Calibration takes place after the groundwater model has been created and all data has been assigned. SPRING supports calibration using the gradient method or the inverse modelling method.

Inverse modelling is an iterative process in which the system is explored step by step and must be reassessed time and again. The meaningfulness of the calculated values must be checked continuously by the user.

After correction of incorrect measured values or other incompatibilities in the input data, the following model parameters remain as uncertain variables:

 

• Hydraulic conductivities (K-values)

• Leakage coefficients (LERA, LEKN)

• Boundary inflows or outflows (VORF and LERA)

• Storage coefficients (for transient simulations, KSPE)

 

Measured potential heads at groundwater monitoring wells are generally used to determine the unknown model parameters p. For transient calculations, hydrographs of potential heads over the simulated time period should also be available for some measuring points. In addition, discharge measurements, for example, can provide information on the magnitude of leakage rates, and boundary inflow or outflow rates may be specified. These are measured values or observation data (m).

In the calculation of the direct problem, the potential heads and rates (measured variables m) are calculated from a data set of estimated but uncertain model parameters p. In the inverse method, on the other hand, the inverse problem must be solved, in which the unsecured model parameters p are determined from the secured measured variables m:.

 

 

In the inverse problem, the model parameters are the dependent variables that are to be determined from the independent observed variables. The aim of calibration is to estimate the unknown model parameters in such a way that a measured state is reproduced as accurately as possible by the numerical model.

When formulating the inverse problem, it must be taken into account that the hydrogeological concept may be subject to errors. A statistical formulation of the inverse problem can relativise errors in the subsequent interpretation of the measured data.

 

Model parameters as unknown variables