Step 19: Run GPS_Covariance to augment GPS covariance?

As explained in Step 17, uncertainties provided with a GPS-velocity solution obtained from geodetic professionals [either as independent elliptical uncertainties of velocity vectors in a .gps file, and/or as velocity-component (co)variances in a .gp2 file] are not empirical uncertainties (or variances) obtained from a large number of independent velocity measurements over different decades.  Instead, they are model-based estimates that only consider the known noise sources in the GPS system (satellites, receivers, and software) and other inferred measurement-related noise such as ionospheric variations and near-antenna reflective interference.  They do NOT include the parts of total velocity variance that are due to actual natural variations in benchmark velocity over time scales longer than the measurement time-window (i.e., multiple decades, centuries, or thousands of years).

If this is not corrected, and if the uncertainties associated with the GPS velocities remain “incomplete” and “too small,” there is danger that this will distort your NeoKinema solutions by mapping some transient, time-dependent velocity features into your estimates of long-term-average tectonic flow.  This also complicates the task of balancing the fit of NeoKinema solutions to multiple classes of data (e.g., geologic offset rates, GPS velocities, and stress directions).

A possible solution to these problems was presented by Bird & Carafa [2016].  It consists of using outside information (e.g., geologic maps and seismic catalogs) to identify noise sources and then estimate the patterns of time-dependent benchmark motion that are most likely to appear in your study area, and using these to add to (“augment”) the GPS-velocity covariance matrix contained in your .gp2 file.  The first paper presented the motivation, theory, and two synthetic examples to illustrate the process and its benefits in NeoKinema modeling.  Then, Carafa & Bird [2016] presented a real case using actual GPS velocities (and noise sources) in Italy.

The risk of accidentally adding new artifacts to the NeoKinema solution is typically* small because we do not attempt to estimate and subtract the time-dependent parts of velocities from the .gps file.  In fact, the contents of the .gps file are not changed at all.  We merely make the NeoKinema solution less sensitive to suspected transient signals by adding terms to the covariance matrix in the .gp2 file.  Loosely speaking, we can “inoculate” our kinematic deformation models to be relatively insensitive to all non-tectonic or transient noise processes that we can identify in advance.  Bird & Carafa [2016] used their synthetic numerical examples to show that the estimates of local noise patterns did not need to be particularly accurate to be helpful.

The utility program that augments the GPS covariance matrix is GPS_Covariance.
It begins by reading your .gps file to obtain benchmark location and elliptical uncertainty in the horizontal velocity vector at each benchmark.
If you can supply a .gp2 file with GPS-velocity covariance matrix, it will read that as well.
(Otherwise, it builds an approximate, model measurement-covariance matrix from the information in the .gps file.)
Then, it prompts you to provide keyboard input (and probably also some small ASCII flat-file tables) that it will use to augment the .gp2 matrix.
The main product of GPS_Covariance is an improved (augmented) .gp2 file which you can send (later, in Step 27) to NeoKinema.
This work is done algebraically, without immediately producing any maps or other graphical output.
However, it also produces some small .gps and .dig files that (optionally) could be plotted with NeoKineMap to illustrate the spatial patterns of local noise sources, one-at-a-time.
Also, a log-file is saved for later use in another utility program, GPS_Postprocessor, that can be used to provide posterior maps of GPS-velocity misfits to a preferred NeoKinema model, in the later Step 37.)

If you wish to conduct this step, it would be good to begin by reading Bird & Carafa [2016] and Carafa & Bird [2016] for further details on how transient noise sources are identified and parameterized.

Then, run GPS_Covariance, which will prompt you for the specific input parameters needed in each case.

One of the output files produced with be GPS_Covariance_log.txt. SAVE THIS LOG-FILE, for later use in Step 36.

Finally, return to your NeoKinema-parameter file (introduced in Step #8; sample file is parameters_for_NeoKinema.nki.txt)
and open it with a plain-ASCII text editor such as NotePad or EditPad Pro.
In line #17, replace the name of your original .gp2 file [or, “none”] with the filename of the augmented .gp2 file produced by GPS_Covariance.