ROI_PAC processing for Modern Geodetic Methods (Page 4 of 6)

Table of Contents

Preparing RAW data for ROI_PAC:

• If you download a .zip file that contains a .DAT file and no .raw file, you will need to run the following script to make the raw format.
  Note that depending on the satellite data you are using, a different program is needed to make the raw files. However, after this point all files should work similarly with the rest of the programs (but you still cannot combine data from satellites with different frequency-bands or orbits). More orbit information is available here.
  
  
• To convert a DAT file into a RAW file for ROI_PAC processing we need to run the following in each of the directories that contain radar image files.
  1. Unzip the files:

          % cd ~/MGM/HectorMine/T127_2907/E2_23027/  # get into image directory
          % unzip ERS2_127_2907_23027.zip 

     Repeat for other directory.
     
     
  2. Rename a couple of files (I use links). Note that you will need to know the calendar date of each image file. If you don't have this information elsewhere, you can get it from the WInSAR website that we used to download the data. For this example, the image directories correspond as:

                     .../E2_23027  --   99/09/15
     	 and    .../E2_23528  --   99/10/20 

     Such that we should create the links with names as follows for the first image:

          % ln -s LEA_01.001 990915.ldr
          % ln -s DAT_01.001 IMAGERY.dat

     Repeat for other directory.
     
     
  3. Run make raw on the data. Typing the command with no arguments gives the basic usage.

      make_raw.pl 'ORBIT' 'YYMMDD'.ldr 'YYMMDD'  # this can take a couple of minutes 
                 ORBIT  = ODR  Delft Orbit Determination for ERS satellites
                 YYMMDD = date of the image
     

          % make_raw.pl ODR 990915.ldr 990915    # in E2_23027

     Repeat for other directory.
     
     
  4. Check the file 990915.raw.rsc to see if DOPPLER_RANGE0 is between +-0.5. If it is not, add or subtract 1 to it and replace the field. Sometimes the doppler ambiguity is not correctly calculated, particularly for more recent ERS2 data (after the gyroscopes died), thus it needs to be updated manually. Look here for more information.

Prepare Running Directory

• We will now need to prepare our output directory:

  % cd ~/MGM/HectorMine/  # you may need to create this directory on tsunami
  # These will link to your DEM directory on terremoto
  % ln -s  /terremoto/home/$USER/MGM/HectorMine/DEM                        
  % mkdir output     
  % cd output
  # These will link to your raw directories on terremoto
  % ln -s /terremoto/home/$USER/MGM/HectorMine/T127_2907/E2_23027 990915  
  % ln -s /terremoto/home/$USER/MGM/HectorMine/T127_2907/E2_23528 991020  
  

• Create process file (990915_991020.proc) that the program process_2pass.pl reads. The contents of the file are as listed below. More information about default and optional parameters can be found here.

  SarDir1=990915 SarDir2=991020 IntDir=int_990915_991020 DEM=~/MGM/HectorMine/DEM/HectorMine.dem flattening=topo OrbitType=ODR

  # location of SAR data # directory to create interferograms in # location of your DEM # remove topographic fringes # Use Delft satellite orbit determinations

Process data to create interferograms

Here we will run the perl script 'process_2pass.pl' to do the complete processing from start to finish. While, for some datasets (including the class example) it is possible to run from start to end with no problems, this is not always the case. Thus, below are examples of processing either in its entirety or piece-wise. You can always look at the perl script to see precisely what it is doing.


• Perform processing from start to finish (good luck!)
  

   % time process_2pass.pl 990915_991020.proc &> 990915_991020.log& # this will take about an hour 
   % tail -f 990915_991020.log # to watch the progress

  
  You may see several "** ERROR **" and warning messages scroll across the log file. Unless the program unexpectedly stops, or the resulting image outputs do not look right, it has been our experience that many of these can be 'safely' ignored.
  
  If your processing stops at a point where it gives the message "That's all folks.", you were likely successful!
  
  :Or
  
  
• Run piece-wise
  By including additional variables after the .proc file, you can start and stop the processing at numerous places. Even if you have started from the fully automatic processing, it is possible to restart the processing from an intermediate point if it had unexpectedly crashed.
  
  Note: If you misspell a start/stop point the program may sit there perfectly happy doing nothing. Thus, it is a good idea to check that you are getting output either to screen or a log file. Some steps, however, may still go 15 or more minutes without any screen output, so you may also want to check what your processes is doing through a program like 'top'. More information about processing in stages can be found here.
  
  Options for DoItFrom and EndItAt are:

  raw roi_prep orbbase slcs offsets resamp flatorb full_res

  seismic begin_filt filtered make_mask unwrapped done_sim_off done_sim_removal redo_base

  use_sim_base synth_offset sim_removal sim_removal_bsim unwrapped_bsim done

  
  Example shown without sending to a log file:
  1. Go from raw data to make single look complexes:

      % process_2pass.pl 990915_991020.proc raw slcs # this will take a few seconds to minutes  

  2. Make flattened inteferogram (remove simulated topographic effects from image):

      % process_2pass.pl 990915_991020.proc slcs flatorb  &> 990915_991020.log&  # takes ~45 minues
      % tail -f 990915_991020.log # to watch the progress

  3. Unwrap inteferogram to get LOS displacements:

      % process_2pass.pl 990915_991020.proc flatorb unwrapped          # takes ~20 minutes

  4. Geocode the unwrapped interferogram to get LOS displacements relative to geographic coordinates (instead of satellite coords):

      % process_2pass.pl 990915_991020.proc unwrapped done             # takes ~5 minutes