GLOBK%20Velocity%20Solutions

1
GLOBK Velocity Solutions

• Thomas Herring

2
Overview

• Basic velocity solution
• Generating realistic velocity sigmas (eariler
  lecture)
• Accounting for earthquakes and site offsets
• Pre-stacking and speeding up results

3
Basic velocity solution

• Basic process is
• Make a file with list of binary h-files (usually
  .glx files). These hfiles span multiple years
  normally
• glist can be used to see list of stations and
  durations and check for large differences in
  coordinates.
• globk command file add
• apr_neu all 10 10 10 1 1 1
• Assigns 1 m/yr uncertainty to velocities at all
  sites
• glorg command file add
• rate_org xtran ytran ztran ltxrotgt ltyrotgt ltzrotgt
  ltscalegt
• Entries depend on if rotations and scale are
  estimated in globk

4
Basic solution Orbits

• Considerations in globk
• If Choice of Experiment RELAX in sestbl. then
  orbits needed to accounted for.
• Merge with global orbit files (available from MIT
  as combined binary h-files)
• Constrain or fix the orbits. Post-2000 IGS
  orbits can generally be fixed,
• There are advatages to generated updated h-files
  with orbits contsrained or fixed. (Speeds up
  velocity solution because apr_svs command not
  needed in velocity solution).
• If Choice of Experiment BASELINE then orbits
  are fixed so do not need to be treated specially.
• With older data (pre-2000) still useful to
  estimate orbits. Also depends on next size
  (lt1000km).

5
Basic solution EOP

• When RELAX mode, EOP parameters (apr_wob and
  apr_ut1) are estimated when BASELINE mode used,
  they are not estimated in gamit and therefore
  orientation of system is fixed.
• If apr_wob/apr_ut1 used, then mar_wob and mar_ut1
  should also be used to allow these to vary with
  time.
• For long baselines and BASELINE mode, these
  should be included in command file.
• For short baselines (lt1000km) and RELAX mode,
  they can be not included (not further
  constraint), or constrained (even F is OK).
• When tightly constrained (lt0.2 mas), xrot yrot
  zrot should be included in pos/rate_org commands

6
Basic Solution Scale

• Treatment of scale is unclear as to best approach
• For long baseline and long time series, apr_scale
  and mar_scale probably should be used until
  re-processed IGS products are available.
• Shorter baseline, probably dont need scale
  estimated.
• Suggestion is to try both ways and see impact on
  results (good use of OPTION string in globk and
  glorg command files).

7
Accounting for earthquakes and site offsets

• Earthquakes and offsets are accounted for with
  the eq_file.
• Generally, after and earthquake there is
  post-seismic motion which can be accounted for
  with the eq_post process noise or eq_log log
  estimates.
• Parameters for some earthquakes are given in
  distributed tables directory. Ideally for other
  earthquakes, model for coseismic motion can be
  used to determine parameters, Postseismic is
  often assumed to 10-20 of coseismic.
• The glorg commends equate and eq_dist can be used
  to make velocities the same before and after
  offsets. Apriori velocity be the same FIXA
  org_opt can bs used but should be checked.

8
Offsets

• The RNRP org_opt can be used to get an estimate
  of the offset magnitude to see if it is
  significant.
• The script sh_exeqs can be used to extract equate
  lines from glorg output. These can be used then
  in glorg to apply equates.

9
Pre-stacking and speeding up results

• With large numbers of days of data and stations,
  velocity solutions can take a long time. Some
  methods of speeding up runs
• The globk command use_num can specify the minimum
  number of times a site is used. Also glist
  output can be used for this (with grep and awk)
• The command decimate to process every nth file.
  With process noise, decimation make little
  difference to results. (decimate 7 or even 30
  works well with continuous data).

10
Pre-stacking

• Pre-stacking hfiles also works well
• Several approaches here
• Combination of local and global hfiles to get
  orbits
• Orbits loose in combination and then not used
  afterwards (I.e., no further constraints)
• Combined hfiles or individual hfiles can be
  combined in time to form n-day averages (7 to 30
  days work fine).
• With process noise used 7-30 day combination
  should estimate only position and not velocity.
  MIDP org_opt will refer result to center of data
  span.
• Same process noise model should be used when
  making 7-30 day combination and when combining
  the combined h-files. (Statistically, the
  results are same as direct solution)
• Care needed in duration of combinations during
  times of large post-seismic deformation (use
  shorter combinations)
• With time-stacks, the chi2 increments will be
  greater than 1 because random walk is too step a
  spectrum.

11
Extracting and displaying results

• There are shell scripts to extract results (e.g.,
  sh_exglk) and shell scripts to plot velocity
  field (e.g., sh_plotvel)
• velview and tsview can be used to display
  velocities and time series.
• tsview outputs rename files that can be used in
  globk. It can also read an eq_file from globk to
  apply breaks and earthquakes.

12
Iteration of results

• Common approach is to
• Run velocity solution and extract coordinates for
  new apriori coordinate file.
• Use this new apr_file and good stations to
  generate time series with glred.
• Use sh_gen_stats to get process noise on the
  sites and get lists fo stabile sites. Command
  sig_neu can used to account for occasional error
  such as campaign with bad height,
• Use these estimates to re-estimate velocity and
  iterate the system
• Time series can be used to detect outliers (use
  rename to ltnamegt_XCL or ltnamegt_XPS in eq_file)
  and breaks (can rename in eq_file to ltnamegt_?PS
  where ? Is letter or number)

13
Summary

• Basic velocity solution Useful when small number
  of days and stations.
• Accounting for earthquakes and site offsets
• Can be complex with large data sets
• Use source command to allow reading in of
  standard values. Runstring option can be used to
  have same command file for velocity and
  repeatability runs
• Pre-stacking and speeding up results Useful with
  large continuous data solutions. Multiple
  techniques can be used.