Step 36: Run GPS_Postprocessor?

If you did NOT use a .gp2 GPS-velocity covariance matrix in your preferred solution, then:

· NeoKineMap provides a set of overlays:

            Overlay #8 :: geodetic benchmarks with velocities

with 5 sub-options that can be used singly, or in combination:

Which kind of geodetic velocities do you wish to plot?
   (1) observed geodetic velocities, used as input to NeoKinema
   (2) velocity adjustments (coseismic slip, frame change?) by NeoKinema
   (3) adjusted geodetic velocities which NeoKinema uses as targets
   (4) model long-term velocities at benchmarks as predicted by NeoKinema
   (5) unfit portion of GPS velocities (data - model residuals)
Please select 1 or 2 or 3 or 4 or 5? [1]:

and these [especially Option #5] should allow you to analyze remaining unfit portions of GPS velocities.

· BUT, this Step and the next Step do not apply to your model.  Please proceed to Step 38.

If you DID use a .gp2 GPS-velocity covariance matrix in your preferred solution, then
you may want to know:

Ø How much did this more-detailed model of uncertainty affect the results?

Ø What posterior estimates can be made for the magnitudes of assumed “noise” processes?

Ø How do these compare to the prior estimates that you entered into GPS_Covariance?

Ø Are any new “noise” sources apparent, once the effects of “known” sources are set aside?

The first of these questions can be answered by just re-running NeoKinema without using the .gp2 matrix,
and comparing the two solutions graphically (and also in terms of their different N₂ misfit norms).
To answer the latter three questions, you need a little bit more linear-algebraic analysis.

Assuming that you saved the log-file GPS_Covariance_log.txt, you can now perform this kind of analysis.
My utility program GPS_Postprocessor2 will examine the unfit portions of GPS velocities at each benchmark, and estimate the magnitude of each noise process (and any reference-frame uncertainty).

It re-loads the same .gps input data file used by NeoKinema.
Then, it looks at benchmarks still present in the g_{token}.nko file output from NeoKinema, and detects whether any benchmarks were left out of that solution (either because they fell outside the .feg area, or because the user was told to remove them, using Delete_Cracked_Benchmarks).

Then, it reads the GPS_Covariance_log.txt file (saved from the original run of GPS_Covariance that created the augmented .gp2 file), and uses this information to divide the unfit velocities into one of 3 “classes:”

(Class 1) Reference-frame loosening modeled in GPS_Covariance;

(Class 2) Anticipated noise sources (volcano inflation, reservoir depletion, very slow landslides, glacial isostatic adjustment, ...) that were modeled in GPS_Covariance;

(Class 3) Other velocity misfits whose origin is not understood (e.g., misfits caused by omitting an active fault from your modeling).

It provides small tables (in its output log-file, and on-screen) giving 3 measures of the amplitude of each noise process:

Ø m1 = mean velocity (in mm/a) across all benchmarks

Ø m2 = RMS velocity (in mm/a) across all benchmarks

Ø m3 = peak velocity (in mm/a) across all benchmarks

and also provides the same 3 measures of the sums of misfit velocities falling into each of the 3 classes of misfit.

It divides the geodetic velocity misfits reported in the g_{token}.nko file (from NeoKinema) into 3 vector components of misfit, at each benchmark, that sum to the total velocity misfit.

Finally, it uses methods similar to those of NeoKineMap to produce a set of 3 related Adobe Illustrator (.ai) files, showing these 3 components of the unfit residual geodetic velocities, relative to the relevant uncertainty ellipses.

NOTE that there is an easy way to give these new maps the same map-projection and virtual-paper size as the previous maps you created in Step 35.  Just re-copy the Map_Tools.ini file (created by NeoKineMap) that you copied and saved at the end of the last Step, and put a copy into the current folder where GPS_Postprocessor2 will be running (and reading its own input files, such as GPS_Covariance_log.txt and AI7frame.ai, and the geodetic input & output files [.gps & g*.nko] from the preferred NeoKinema model, etc.).