GPS network simulations: How do I generate realistic covariance matrix data for GPS vectors?
There are, of course, many possible realistic covariance matrices for GPS vectors. An example coordinate difference observation group for GeoLab is as follows: 3DD DXYZ S001 S002 572.7240 -280.7980 -413.6370 CORR CT UPPR 32.50000 ELEM 1.0000000000000 -0.2032966000000 0.0045336000000 ELEM 1.0000000000000 -0.3644455000000 ELEM 1.0000000000000 ELEM 0.00056219690 0.00095490840 0.00115665900 This is from an actual network, and the standard deviations may be realistic given that the CORR record specifies a factor of 32.5 for the whole matrix. Normally, the covariance matrix (or the correlation matrix and standard deviations as used here) generated for GPS vectors by post-processing software is usually very optimistic and a scale factor is required to make it realistic. Experience in processing networks of these observations results in determining realistic scale factors for the covariance matrices. The standard deviations in the last ELEM record above range from 0.5 mm to about 1.1 mm (obvio
