GlobK Processing: refining and velocity determination for GPS (Page 3 of 3)

Previous Gamit processing: Determining loosely constrained GPS positions

Data Preparation: (assuming you have already run gamit)

• You will need to combine all solutions into a single directory. The easiest way to do this is to create a symbolic link for all the earlier data into the Socorro2004 directory in 'YEAR_DOY' format.

  % cd ~/MGM/Socorro2004/
  % for file in 3??   # this is for current directory 
     do 
        ln -s $file 2004_$file
     done
  % for file in ../Socorro2003/3??   # 2003 
     do 
        ln -s $file 2003_`basename $file`
     done
   
  % for file in ../Socorro2002/3??  # 2002 
     do  
        ln -s $file 2002_`basename $file`
     done
  

• We can use the shell script 'sh_glred' to take the data from multiple years of campaigns and calculate global velocities. The following script assumes that you have the daily solution output of gamit for each year and day of data of interest listed in local subdirectories with the following format 'YYYY_DOY'. The most necessary files for sh_glred are the daily H-files (e.g., hsmb1a.04319), which contains the full covarience matrix for all stations, position components, atmospheric/orbit/rotation/nutation/etc.. corrections.

  % sh_glred -s 2002 311 2004 327 -yrext -local -expt smb1 -opt H G E
  

Output:

• This run create two directories 'glbf' and 'gsoln', that will hold compact binary H-files and run output.
• In 'gsoln', files SUM.* and VAL.* contain individual station velocity and daily coordinate information that is tightly constrained.
• Using the 'E' option in 'sh_glred' will create a series of plot starting with 'psbase' and 'pshist'. Examine these using 'gs' to see what is used to constrain velocities. Note that all stations have significant horizontal slopes due to North American Plate Motion. This will need to be removed to constrain local motion.
  
• Note on sh_glred: Once this has been run once, a file is created within gsoln called 'globk_comb.cmd'. This file contains some default information about how to combine data in the total solution. In order to get an additional output about the precise geographic and UTM locations of the stations (daily), it would be good to modify the following line to include the variables 'GEOD' and 'UTM'. The line should now look like:

    prt_opt GDLF CMDS GEOD UTM
  

Cleaning up data:

Before we can adequately constrain velocities, we should remove any data that we know to be erroneous. We should check our current velocity plots to determine identify any significant outliers in camaign stations and first check station logs to see if anything wierd happened.
• Since these were data that I collected, you can check with me about logfiles. You will notice that stations BOWL and PRTC in the early part of 2004 has a signficant shift...what should we do?
• We see that we need to remove data from sites BOWL and PRTC for days 319 and 320 of 2004 because of a compromised spikemount setup and bad battery. Since I still have station YESO running on those days, I don't want to just get rid of the offending days. Instead, I will rerun those days using sh_gamit but "removing" the offending data from the rinex directory.
  
  • Before rerunning, I :

    % cd rinex
    % mkdir temp
    %  mv bowl3190.* temp
    %  mv BOWL3190.* temp
    %  mv BOWL3200.* temp
    %  mv bowl3200.* temp
    %  mv PRTC3190.* temp
    %  mv prtc3190.* temp
    %  mv prtc3200.* temp
    %  mv PRTC3200.* temp
    

  • rerun sh_gamit days 319 and 320.
  • rerun sh_glred on the entire dataset and check to see that BOWL and PRTC no longer have bad data and that YESO has 3 days in 2004.

Get velocities for individual years:

• Because we are looking at a magmatic source, there is a high likelyhood that deformation is not linear over time. Thus, if there is activity on timescales less than a year the transient information is mostly lost with this sort of campaign. While if activity is on timescales of several years data may appear nearly-linear. Without a near-by continuous network these sorts of determinations are difficult to assessr. Because of this I have added two new continuous sites to the Socorro Magma Body project in late 2005 (results will be discussed in class).
• Because of this potential non-linearity, it is necessary to breakdown the data into temporal components and create yearly velocity estimations (2002-2003 and 2003-2004).
  
  
• We could just recreate subsets of the necessary information in the gsoln directory this way, after running all the data, we just need to redo the velocity calculations and plotting using the 'E' option of sh_glred:

  # before running sh_glred again see above note on globk_comb.cmd
  # 2002-04 (all data)
  %  sh_glred -s 2002 311 2004 327 -yrext -local -expt smb1 -opt H G E
  %  cd gsoln
  %      mkdir 2002-2004
  %      mv SUM* VAL* ps* 2002-2004/
  %  cd ../
  # 2002-03
  %  sh_glred -s 2002 311 2003 365 -yrext -local -expt smb1 -opt E
  %  cd gsoln
  %      mkdir 2002-2003
  %      mv SUM* VAL* ps* 2002-2003/
  %   cd ../
  # 2003-04
  %  sh_glred -s 2003 322 2004 327 -yrext -local -expt smb1 -opt E
  %  cd gsoln
  %      mkdir 2003-2004
  %      mv SUM* VAL* ps* 2003-2004/
  %  cd ../
  

Removing North American Plate Motion:

• In order to look at local relative motion we should remove the North American Plate motion from the model.
• For those that have taken Geodynamics you may recall that we can describe all plate motions on the surface of the Earth by their Euler pole location and rotation. For the North American Plate, this motion can be described by (Lat=-4.59^o, Lon=-82.91^o, Rotation=0.1949^o/Ma) [Sella et al., 2002]
• I wrote a small program to calculate euler poles a while back that can be useful here. The program, called 'euler' is written in FORTRAN and has its executable and binary codes located in /usr/local/geophysics/bin. The program takes as input, the pole (lat, long, rot), and site location (lat, lon) and outputs the predicted motion (East, North) for the site in mm/yr. Please look at the source code for information about the input/output. As an example:

  % euler         
   -4.59 -82.91 0.1949 33.89 -106.97   # Approx location of site SC01 taken from O-file
   -12.439  -8.807                     # This line was output by program
  

• The program output should be removed from calculated velocities to see the remaining local motion. It is probably best to create a table that has each station name, location, velocity, velocity error, predicted motion, and period of reference.
  You can use the program 'clean_sln' in /usr/local/geophysics/bin/ (on tsunami) to create a velocity file in the below format.

  # NAME  LON            LAT           ELEV         VE       VN        VU       ErE    ErN    ErU      NA-E    NA-N      PERIOD
  BOWL    253.114570897  34.233055752  1424.6359   -10.84    -9.31      9.16    3.37   2.95   6.29   -12.538  -8.778    2002-2003
            ^             ^               ^          ^        ^         ^       ^      ^      ^          ^        ^
            |             |               |          |        |         |       |      |      |          |        |
            -------------------------------          ----------------------------------------            ----------
                             |                                            |                                   |
  	       from 'globk_smb1_?????.prt'               from  'SUMNEW.SUM.smb1'                 output from 'euler' 
  

  While velocity covarience would also be useful in this table, it is not included for simplicity. We would likely need to recalculate velocity covarience between components, to include in this table.
  

Plot Velocities on Map:

This will be done in class.

• References:
  Sella, G. F., T. H. Dixon, and A. Mao (2002), REVEL: A model for Recent plate velocities from space geodesy, J. Geophys. Res., 107(B4), 2081, doi:10.1029/2000JB000033.
  

Previous Gamit processing: Determining loosely constrained GPS positions