Identifying errors in the math model

Although residual errors are not necessarily mistakes that need to be corrected, they may indicate problems with the math model. The following conditions may help you to identify such problems.

Outliers

A ground control point (GCP) or tie point with a very high residual error compared to the others in the Residual Errors report may indicate an error in the original GCP coordinate, a typographical mistake, or an error in the position of the GCP or tie point on the raw image. These points are called outliers.

To correct an outlier:

Verify that the feature that you picked in the raw image corresponds to the one from your source.
Verify that the typed ground coordinate matches the coordinate listed in your source.
Confirm that the ground coordinate that you collected in the raw image is consistent with the coordinate that you selected from the vector or the geocoded image.
Verify that the projection and datum for the ground coordinate are correct.

If all else fails, delete the point or change it to a Check Point.

Poor Math Model Solution

If the residual errors for all the GCPs and/or tie points in general are high, it may indicate a poor model solution. Poor model solutions can be the result of inaccurate GCPs, errors in the projection or datum, inadequate distribution of the ground control, or insufficient ground control.

Residual Errors Are All Zero

If all the residual errors for the GCPs and tie points read zero, it usually indicates that you have collected only the minimum number of ground control points or fewer. Collect more GCPs and tie points.

However, if you selected the Thin Plate Spline math model for your project, the residual errors will always indicate zero. Use Check Points to check the accuracy for the Thin Plate Spline math model.

Systematic Trends in Residual Errors

If you have high residual errors in one part of an image or project, it can indicate that you need more ground control in the problem area, or it may indicate that you have one or more bad points in the area that are skewing the math model. Some bad points are difficult to identify since some points may compensate for others.