DC2DInvRes - Inversion Menu

Starting model

The resistivities are reset to a homogeneous starting model with the median of the resistivities.

One inverse step

Only one inverse step including forward calculation is calculated. This is particularly usefull for testing options as regularization types and strengths.

Hotkey: 1

Full Inversion

The full inversion is carried out using the current options.
Inversion stops as soon as the chi-squared misfit stagnates or the data are fitted within standard deviation (chi^2<1).

Hotkey: Shift+I

Create Layered Model

Applying a one-dimensional inversion to the whole data set a starting model is constructed containing the main layering of the underground. The algorithm tries to equalize neighboring layers if the difference is not very large.

Set as Reference Model

The actual model is set as reference model, which is the homogeneous starting model. It can be successfully applied if an a-priori model is present and the inversion model is forced to keep close to is.

Back to old model

Sometimes it proves useful to play with some options using the one-step inversion. For that reason, it is possible to go back to the last model via this feature.

Clear Sensitivity

Sometimes it is useful to clear the sensitivity matrix (for recomputation or saving disk space).

Sensitivity Integration

Calculates the homogeneous half-space sensitivity matrix. Otherwise it is done automatically before the first inverse step. These sensitivities are saved for later use in a file with the root of the data file plus "-sens.mat". Once the sensitivities are calculated, they can be investigated using Show Sensitivities.

Sensitivity Forward

The sensitivity matrix is computed by forward calculation. This is a time-consuming task but (rarely) sometimes better than updating.

Depth of Investigation

The depth of investigation (DOI) index shows, following (Li and Oldenburg, 1999), the dependence on the starting model. One inversion step is carried out using two different homogeneous half-spaces. The difference between the two models can be treated as a reliability of the model cells. A well-resolved model cell is supposed to obtain nearly identical inversion results.

IP Inversion

If IP data are present, a single-step is carried out following (Beard, Hohman and Tripp, 1996). As a result the chargeability of the model cells can be displayed using the Show menu.

Time Lapse Inversion

If measurements are taken in varying time steps one is mainly interested in the changes. Since a pure comparison of inversion results often suffers from differently constructed inversion artifacts or blurs it is more comprising to constrain the differences.
In a single inversion step the model difference is calculated from the data differences using the identical inversion options (sometimes here 2nd derivative is better than 1st).
First the user specifies a file with the time step to be compared. Although the electrode positions must be identical the configurations may differ, since the common basis of both is used.
Note that this is only valid for small differences and thus does not guarantee the time step data to be fitted. See tutorial part 5 for details.