updated top-level README and version_decl for V4.5 (#1847)

[WRF.git] / var / obsproc / Changes

Updated for Version 3.1 release.  Mar 17 2009

1, rename 3dvar_obs.F90 to obsproc.F90.
2, rename 3dvar_obs.exe to obsproc.exe.
3, rename namelist.3dvar_obs.wrfvar-tut to namelist.obsproc.wrfvar-tut.
4, remove some unnecessary files only for development team internal use only.
5, update README and README.namelist

Updated Makefile to support more compilers.  Mar 5 2009

1, to compile, add compiler choice in the make command line, for example,
   make FC="pgf90"
   make FC="ifort"     # for Linux only
   make FC="gfortran"  # for Linux only
   make FC="g95"       # for Mac only

Updated by Y.-R. Guo, 7 May 2008

1, To correct the "0 obs" ascii output file writing

   Modified:
       src/module_write.F90

Updated by Y.-R. Guo, 28 March 2008

1, Improve the printing, change "print *," to be "write(0,"

   Modified:
       src/module_map_utils.F90
       src/module_decoded.F90
       src/module_err_afwa.F90

2, Plot only one set of the obs distribution maps

   Modified
       MAP_plot/Dir_map/Map.F
       MAP_plot/Dir_map/plstn1.F

3, Added the function to write out the obs files for use of 
   3DVAR, FGAT, and 4DVAR

   For 3DVAR, the obs within slot [time_window_min, time_window_max)
              are outputted. no time duplicate obs;
.
   For FGAT,  the obs files for multiple time slots are outputted.
              each of the time slots has the closed low bound and 
              opened upper bound. Within the whole time window, no time-
              duplicate obs allowed;

   For 4DVAR, same as FGAT but there may be time-duplicate obs within 
              the whole time window. Within each of time slots,however,
              no time-duplicate obs allowed. 

   * The opened upper-bound is important for the cycling run to avoid the
     duplicated usage of the observations in the different cycles.

   * The number of time-slots before and after analysis time could be 
     different (i.e.asymetric). However, the number of the time-slots
     before the analysis time MUST BE >= 1 for FGAT or 4DVAR.

   * Currently the lengths for all of the time-slots are considered to be 
     same. 

   * Note: the current MAX_OB_RANGE in scripts is ONLY allowed to be integer 
     (hours), and the entir time window = 2 * MAX_OB_RANGE, must be the even 
     numbers (hours). So for 6-h cycling run (MAX_OB_RANGE=3), the time window 
     for obsproc is correct, but for a 3-h cycling run, the time window for 
     obsproc will not be correct. 

   Modified
       src/3dvar_obs.F90
       src/module_namelist.F90
       src/module_write.F90
       src/module_duplicate.F90

4, Introduced the switches to control the specific types of obs in
   the obs_gts (ASCII) files. This is mostly for testing purpose.

   Modified
       src/3dvar_obs.F90
       src/module_namelist.F90
   
Updated by Y.-R. Guo, 31 January 2008

1) Bug fix in module_recoverh.F90 -->
              SUBROUTINE recover_h_from_ptq(P,T,Q,HGT,KXS,K_START,K_TOP)

   it must be processed separately when k >= kxs under k_start == 0.

2) Add the limitation on iproj setting in module_namelist.F90

   modified codes:
  
       src/module_namelist.F90
       src/module_recoverh.F90
   and README
 
Updated by Y.-R. Guo, 10 May 2007

1) To make the namelist variables in record6 more understandable:

   ps0 --> base_pres
   ts0 --> base_temp
   tlp --> base_lapse

   and 

   pis0 --> base_tropo_pres
   tis0 --> base_start_temp.

2) To maintain the backward compatible, users could 

   either set 

     a set of variable names: ts0,  ps0, tlp, and tis0, pis0 

   or set 

     a set of variable names: base_temp, base_pres, base_lapse, and 
                              base_start_temp.base_tropo_pres 

   Both of the settings are giving the exactly same results.

3) Note: In most cases, pis0, tis0 or base_start_temp, base_tropo_pres.
   do not need to set in namelist file. The default values are OK.

4) Modified subroutines

     src/module_mm5.F90
     src/module_namelist.F90
     src/module_write.F90

     README
     README.namelist
     namelist.3dvar_obs.sample1
     namelist.3dvar_obs.wrfvar-tut

Updated by Y.-R. Guo, 12 April 2007

Modified the functions Ref_height and Ref_pres in order to consider
the more accurate reference state in stratosphere. 

  The tropopause pressure (default=20000.0 Pa) and the isothermal 
  temperature above tropopause (default = 215 K) are introduced into 
  the "namelist.3dvar_obs" file.

As the model top level became high and high, 1000 Pa, even 100 Pa,
the accurate reference state must be provided in the OBSPROC to
do the gross quality control, recover the pressure or height, etc.

The codes are modified"

   3DVAR_OBSPROC/src/3dvar_obs.F90
   3DVAR_OBSPROC/src/module_mm5.F90
   3DVAR_OBSPROC/src/module_namelist.F90
   3DVAR_OBSPROC/src/module_write.F90

Updated by Y.-R. Guo, 22 March 2007

Jim Bresch found the variable, m_miss, not initiated to be zero in 

   3DVAR_OBSPROC/src/module_decoded.F90

Added "m_miss = 0" in this code.

Updated by Y.-R. Guo, 5 February 2007

To move the elevation missing check after the domain and time-window screen
in 3DVAR_OBSPROC/src/module_decoded.F90, and the printing stuff reduced.

Updated by Y.-R. Guo, 1 February 2007

1, Removed the line:

   $(RANLIB) $(LIBDIR)$(FDVARLIB)

   in MAP_plot/Dir_map/Makefile and MAP_plot/Dir_map/Skewt/Makefile 

   The Error in compiling Map.exe will disappeared.

2, Modified the MAP_plot/configure.user.ibm

   A -O2 option added to FCFLAGS. Otherwise, Map.exe does not work properly
   in bluevista. (This is happened in compiling the wrfvar code too). 

Updated by Y.-R. Guo, 20 January 2007

Hui-Chuan encountered the segmentation error with 2007010418 and
  2007010812 obs (Jim Brsch) data in check_duplicate_loc.

The code is rather complicated during the check duplication and data merging
together. I found that the logic is not correct when the "first" coming
report has a "0 level data record". Because the code is used the pointer/link
structure, it takes me 3 days to fix the module:

         module_obs_merge.F90

Updated by Y.-R. Guo, 28 November 2006

1, g95 compiler in Mac is added:

     mak mac_g95

2, Remove the redundant codes

   subroutine local in module_qc.F90

   AA = (PMAX - PMSL) / LOG (NTOP/NBOT) in module_err_afwa.F90

3, Added the AIRS plotting to MAP_plot

Updated by Y.-R. Guo, 10 November 2006

1, Added the GPS RO excess phase processing (GPSEP)

   GPS RO Refractivity observation error specification is updated (Chen and Kuo
   2005).

   The GPS RO little_r file is from the bufr decoder program. The data
   contain the height, lat, lon, impact parameter, azimuth angle, bending 
   angle, refractivity.

2, Reviewed the AIRS data processing

   p-h gross check should be done, not skept over by Rizvi's mod.

3, Note that AIRS and GPSEP are not includeed in prebufr output data yet.
   More work needed for this.

Modifications:

        M README
        M namelist.3dvar_obs.wrfvar-tut
        M src/3dvar_obs.F90
        M src/fm_decoder.F90
        M src/module_complete.F90
        M src/module_decoded.F90
        M src/module_diagnostics.F90
        M src/module_duplicate.F90
        M src/module_err_afwa.F90
        M src/module_err_ncep.F90
        M src/module_per_type.F90
        M src/module_qc.F90
        M src/module_recoverh.F90
        M src/module_recoverp.F90
        M src/module_sort.F90
        M src/module_write.F90
        M src/platform_interface.inc
        M src/sort_platform.F90

Updated by Y.=R. Guo, 29 September 2006

1, Added the system library SYSLIB=-lSystemStubs for Mac xlf.
2, Mercator map projection bug fix.

Updated by Y.-R. Guo, 27 April 2006

  Added the directory: kmabufr_to_littler, in which the kma bufr format
  observations are converted to the little_r format observations.

Updated by Y.-R. Guo, 13 April 2006

  Incorporated the Jim Bresch's changes in preparing the release to KMA.

  3DVAR_OBSPROC/Makefile
  3DVAR_OBSPROC/lib/cwordsh/README_cwordsh
  3DVAR_OBSPROC/lib/cwordsh/Makefile
  3DVAR_OBSPROC/lib/cwordsh/cwordsh.f
  3DVAR_OBSPROC/lib/cwordsh/cwordsh.sh
  3DVAR_OBSPROC/lib/Makefile
  3DVAR_OBSPROC/lib/ncep/Makefile
  3DVAR_OBSPROC/src/module_decoded.F90

  3DVAR_OBSPROC/MAP_plot/Map.csh
  3DVAR_OBSPROC/README

Updated by Y.-R. Guo, 6 March 2006

  1, This version of 3DVAR_OBSPROC code included the prebufr OBS
     file output (see README.namelist), special for KMA 2005 project.

     What are changed by Jim Bresch and Jim Drake? see below:

     Updated by Jim Bresch, 1 February 2006
     --------------------------------------

     Completed prepbufr code. Added NCEP's bufrlib to the directory.
     Numerous changes to source code and Makefiles.

     Updated by Jim Drake, 20 Oct 2005
     ---------------------------------

     Modified to add PREPBUFR processing of surface observation types:

        3DVAR_OBSPROC/README
        3DVAR_OBSPROC/README.namelist
        3DVAR_OBSPROC/Makefile     (for faster builds)
        3DVAR_OBSPROC/little_r     (only FM-13 and FM-15 because that's
                                    all AFWA puts out right now. Tested
                                    with FM-12 and FM-18 too).
        3DVAR_OBSPROC/namelist.3dvar_obs
        3DVAR_OBSPROC/src/Makefile (for faster builds)
        3DVAR_OBSPROC/src/3dvar_obs.F90
        3DVAR_OBSPROC/src/module_write.F90

    For architectures other than IBM the macro LDFLAGS will need adjusting, and
    the macro LOCAL_LIBRARIES needs to be pointed at the BUFRLIB library, (cf.
    3DVAR_OBSPROC/Makefile).

    On AFWA IBM's, set the environment variable OBJECT_MODE to 64 in order to
    build.

    3DVAR_OBSPROC puts out a warning for some deficiency in some of the 
    observations as they are read in that I haven't pursued.

  2,In general, this version of 3DVAR_OBSPROC code can be applied to write out
    both the ASCII and prebufr 3dvar_OBS files. Only restriction is the number
    of vertical levels to 190. If users expect the number of vertical levels 
    > 190, such as GPS Ref from COSMIC/CDDAC, you should use a earlier version
    of 3DVAR_OBSPROC code obtained by

      cvs -d bay-mmm:/data3/mp/wrfvar co -r V2_11 3DVAR_OBSPROC

Updated by Y.-R. Guo, 20 February 2006

  AFWA found a "STOP" problem with SOUND data in Sub. reorder in 
  module module_recoverh.F90 because the pressure is missing at
  some levels. The problem is existed in module module_recoverp.F90
  that at certain bottom levels, the pressure was not retrieved.
  Two modules are modified.

Updated by Y.-R. Guo, 2 February 2006

  Jim Bresch found that all sea surface wind speed from SSMI became missing.
  However, I checked the old file: t1a_obs_ssmi_retrieval.02110418.3dvar,
  the wind speed is OK. This was caused by "meaningless wind check" added
  on 04/27/2004. Actually, for SSMI, this check should be skept over.

  Also I found the the quality index for SSMI PW became "4" --- error 
  specified. It may be better to keep it to be "0" as before.

   Updated by Y.-R. Guo, 17 January 2006

   1, with CWB 2003090312Z obs data: PILOT --> FM-32, ID=48698, 
      1.370, 103.980, the "dead-loop" occurred in module_recoverh.F90

      The bugs in re-assign the negative values to the HGT%qc in
      subroutine recover_h_from_ptq are found:

       HGT(k-1)%qc   = - HGT(k-1)%qc

      should be

       if (HGT(k-1)%qc>0) HGT(k-1)%qc   = - HGT(k-1)%qc

   Updated by Y.-R. Guo, 25 October 2005

   1, CWB (Hi-Chuan Lin) found the dead-loop with the data 
      pilot2005081800.problem.r in module_recoverh.F90.

      The bug corrected in module: pilot2005081800.problem.r

   2, Occasionally found 3DVAR_OBSPROC is not working properly 
      in MMM Dec alpha machine.


   3, For AMD Opteron Processor 848 (x86_64), the module: module_sort.F90
      must have a statement such as:

      write(1001,'("In  RECURSIVE SUBROUTINE merge_sort: high and low:",2i10)') high, low

  modified code:

    module_diagnostics.F90
    module_err_afwa.F90

    module_recoverp.F90 --- to guarantee the pressure is 
                            retrieved at all levels.
    module_recoverh.F90 --- the "STOP" statement commented, 
                            instead, if the "STOP" condition
                            happened, that SOUND will be discarded.    
    module_sort.F90     --- special for AMD Opteron


Updated by Y.-R. Guo, 23 September 2005

1, For Mercator projection (Hurrican Rita), 3DVAR_OBSPROC/src/map_utils 
   should be corrected to obtain the correct (x,y) from lat/lon;

2, When the projection is "LC" with TRUELAT1=TRUELAT2 and "ME", plotting 
   MAP is not working properly. 3DVAR_OBSPROC/MAP_plot/Dir_map/setup.F is 
   modified.

Modified:
        3DVAR_OBSPROC/src/map_utils.F90
        3DVAR_OBSPROC/MAP_plot/Dir_map/setup.F
 
Updated by Yong-Run Guo, 31 May 2005

To make the 3DVAR_OBSPROC working in Mac G5.

Modified:

        3DVAR_OBSPROC/README
        3DVAR_OBSPROC/Makefile
        3DVAR_OBSPROC/MAP_plot/Makefile

Added:

        3DVAR_OBSPROC/MAP_plot/configure.user.mac
        3DVAR_OBSPROC/MAP_plot/mac_Makefile1
        3DVAR_OBSPROC/MAP_plot/mac_Makefile2

Updated by Yong-Run Guo, 22 April 2005

Rizvi added code to module_decoded.F90 for MODIS obs: "FM-88 MODIS".
Also file: namelist.3dvar_obs.AMPS is updated.

Updated by Yong-Run Guo, 3 February, 2005

1, For application to the Polar Stereographic map projection, the
   calculation of (i,j) from latlon or latlon from (i,j) are different
   from the Lambert Conformal projection.

   In Lambert projection, it is always (1,1) from the (lat1,lon1),
      no matter what knowni, knowj given. So if (lat1,lon1) were
      assigned by the cen_lat and cen_lon, the computed (i,j) must
      have the correction with the knowi and knowj.

   In PolarStereographic projection, when (lat1,lon1)=(cen_lat,cen_lon),
      the (i,j) = (knowi,knowj).

          module_map_utils.F90

2, Conversion of the WRF domain configuration to MM5 domain configuration

          module_namelist.F90

3, For projection 'ST', it is always confac = 0.0, not need to call lccone.
   in MAP_plot/Dir_map/plots.F. Otherwise, there is a floating point error 
   when abs(fsplat)=90.0 

          MAP_plot/Dir_map/plots.F
 
Updated by Yong-Run Guo, January 11, 2005

1, For correctly computing the model coordinates, especially with 
   lat=-90.0 when truelat1>0 or 90.0 when truelat1<0.

          module_namelist.F90
          module_decoded.F90

2, One more data type: Bogus soundings for tropical cyclone is added

   FM=135 and platform='BOGUS'

          module_per_type.F90
          module_complete.F90
          module_duplicate.F90
          module_qc.F90
          module_write.F90
          platform_interface.inc
          fm_decoder.F90
          module_decoded.F90
          3dvar_obs.F90
          sort_platform.F90

   U/V errors (m/s) are specified in the fields u%data.v%data in little_r file.
   SLP error (Pa) is specified in the field psfc%data in little_r file.

          module_err_afwa.F90

   In order to keep the data (U,V)=0.0 at the center of Typhoon, the 
   'diagnostics_wind' procedure is skept over.

          module_diagnostics.F90

------------------------------------------------------------------------------
Updated by Yong-Run Guo, November 11, 2004

1, Added the WRF application option (fg_format='WRF') and  the module:
      module_map_utils.F90
   for model coordinate (X,Y) calculation.

2, Removed those GPS related modules:
      module_gpspw.F90
      module_gpspw_caa.F9
      module_gpspw_gst.F90
   because they are never used in the program

3, The file namelist.3dvar_obs is changed (see README.namelist).

-----------------------------------------------------------------------------

Updated by Yong-Run Guo, September 2, 2004

1, Indian visitors found that time duplicate check is not working properly,
   not retaining the observation closest to the analysis time. Rizvi 
   worked with them to correct the "module_duplicate.F90". Y.-R. Guo 
   reveiwed it, and included the GPS Ref. in this module.

2, SHU-YA CHEN, Central University, Taiwan, found compiling failed in
   PC Linux with cpp flag -DBKG. Y.-R. Guo corrected the bug in setup.F90.
   There is no problem without cpp flag -DBKG, so not affect AFWA, KMA. 

3, Code changed
 
     src/module_duplicate.F90
     src/setup.F90
     src/3dvar_obs.F90

-----------------------------------------------------------------------------
Updated by Yong-Run Guo, July 29, 2004

Problem:  With AFWA obs data file:" 04072306.3DVAR", the job stopped.

Fix    : 1) To commented the STOP in 3DVAR_OBSPROC/src/error_handler.F90;
         2) To discarded the OBS with the fatal errors, if any, in
            3DVAR_OBSPROC/src/module_qc.F90;
         3) To add the hydrostatic check for the reported height in
            3DVAR_OBSPROC/src/module_recoverh.F90:

            If (abs(Hd - Ho) > 0.10*Hd), use Hd to replace the Ho.
            where Hd is the computed height from the (P,T,q) based
            on the hydrostatic equation. Ho is the reported height.

----------------------------------------------------------------------------
Updated by Yong-Run Guo, July 21, 2004

 I, Main points in this update

  1, The GPS Refeacrivity is included. The Refractivity N has the 
     WMO code = 116, and the field 'dew_point' is used to store the 
     refrectivity N. This will minimize the code changes. Currently,
     only the local OBS operator is used in 3DVAR, so N is the only
     observation data. 

     The observation errors are assigned as

        error = nbot*exp ( (pres-pmsl)/aa )

     where nbot = 10, ntop = 3, pmax = 1000, pmls = 100000, 
         aa = (pmax-pmls) / log(ntop/nbot), and pres is the pressure
         at the observed level.

     Reference
     
     Huang, C.-Y., Y.-H. Kuo, S.-H. Chen, and F. Vandenberghe, 2004:
       Improvements in Typhoon Forecasting with Assimilated GPS Occultation
       Refractivity. (Submitted to Mon. Wea. Rev.)

  2, For KMA Global unified 3DVAR system, one more namelist vaviable added:

     &record4
      domain_check_h           = .false.,

      domain_check_h=.false. -- ignore the outside domain check 
                                for global 3DVAR.
      domain_check_h=.true.  -- do the outside domain check for regional 3DVAR.

     Note: when running for Global OBS processing, you MUST delete the
           option  -DBKG in Makefile and use the namelist.3dvar_obs.template1
           because there is no Background file available.

  3, Modified the 'above_top` check by using 'ptop' instead of hlid

     &record4
      remove_above_lid         = .true.,

      remove_above_lid=.false. -- keep the data at all the levels. 
      remove_above_lid=.true.  -- discard the data at the levels above ptop.

  4, In order to plot a global coverage map background, one more value, 0,
     in addition to 1 (LAMBERT), 2 (POLAR), 3 (MECATOR), is effective for 
     namelist variable

     &record7
      IPROJ    =   0,

     This is Cylindrical Equidistance map projection for plotting the
     distribution of the Global observations.
 
     Note that for a global map, always set 

     &record8
      IDD      =   1,
      MAXNES   =   1,
      NESTIX   =      181,
      NESTJX   =      361,
      DIS      =    110.0,

II, Bug fix
 
  1, A bug fixed in src/module_map.F90 and MAP_plot/Dir_map/llxy.F.
 
     When usimg Mecator map projection and the domain across the date line,
     the (x,y) was computed increectly.

  2, A bug fixed in src/module_recoverh.F90

     When a multiple-level OBS with the height missing at the bottom
     levels, the heights recovered at those levels are incorrect. This
     bug sometime caused the program Preprocessor stopped, but not 
     affect the 3DVAR.

III, Code changed

     src/3dvar_obs.F90
     src/fm_decoder.F90
     src/module_complete.F90
     src/module_decoded.F90
     src/module_diagnostics.F90
     src/module_duplicate.F90
     src/module_err_afwa.F90
     src/module_err_ncep.F90
     src/module_map.F90
     src/module_mm5.F90
     src/module_namelist.F90
     src/module_per_type.F90
     src/module_qc.F90
     src/module_recoverh.F90
     src/module_write.F90
     src/platform_interface.inc
     src/setup.F90
     src/sort_platform.F

     MAP_plot/Dir_map/DA_Constants.f90
     MAP_plot/Dir_map/DA_Define_Structures.f90
     MAP_plot/Dir_map/DA_Read_Obs.inc
     MAP_plot/Dir_map/DA_Read_Obs_Info.inc
     MAP_plot/Dir_map/Map.F
     MAP_plot/Dir_map/llxy.F
     MAP_plot/Dir_map/plstn1.F
     MAP_plot/Dir_map/setup.F
----------------------------------------------------------------------------   

VERSION 1_3_0 3dvar_obs PROGRAM
(Updated by Yong-Run Guo, June 4, 2003)

This is the main directory of the "3dvar_obs" observation processing program 
of the MM5 3D-VAR data assimilation system (3dvar_1_3_0 and after)

Everything a user needs to run the 3dvar_obs program is contained in this 
directory.

For questions, please send emails to mesouser@ucar.edu

For detailed documentation about 3dvar_obs programs, please read below 
and file "README.namelist" in this directory.

For more information on the MM5 3D-VAR data assimilation system, please see

http://www.mmm.ucar.edu/mm53dvar/docs/

For details on the 3dvar_obs input data format, please see

http://www.mmm.ucar.edu/mm53dvar/3dvar_tutorial_presentations.htm

-------------------------------------------------------------------------------

What is in 3dvar_obs?

3dvar_obs process decoded observations in LITTLE_R format, such as the ouput 
of the MM5 "gts_decoder" or the "fetch.csh" utilities for inputing into
the MM5 3DVAR system. 

The basic operations performed by 3dvar_obs are:

1) Remove the observations outside the time range and domain (horizontal 
   and top).

2) Re-order and merge the duplicate (time and location) data reports. 

3) Retrieve the pressure or height based on the observed information with
   the hydrostatic assumption.

4) Check the vertical consistency and super adiabatic for multi-level 
   observations.

5) Estimate observational error based on the pre-specified error file.

6) Write out at the 3DVAR input format.

7) Plot the distribution for each type of observations.

-------------------------------------------------------------------------------

3dvar_obs is a Fortran 90 program, and requires Fortran 90 compiler.

At the moment, 3dvar_obs is working only for Dec_alpha, IBM, SGI, NEC 
and Fujitsu with their native compilers, and PC running Linux using 
Portland Group Fortran 90 compiler. If your machine isn't one of the 
ones supported or the compiler isn't the native one from the vendor, 
you need to edit the Makefile. For CRAY, SUN, and HP, we did not do 
fully testing of this program although Makefile included the MACRO 
for these machines.

---------------------------------------------------------------

To compile 3dvar_obs:

- Type 'make' in the 3DVAR_OBS directory to complie;

- When make is successful, executable 3dvar_obs.exe will be created in this
  directory.

---------------------------------------------------------------

To run 3dvar_obs:

To create a namelist.3dvar_obs by edit the file namelist.3dvar_obs.sample
(README.namelist for details), and type 

    3dvar_obs.exe >& 3dvar_obs.out

to execute the program.

Required input files for 3dvar_obs are:

 1) First guess file (MMINPUT/MMOUT)**,

    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      ****** Note Note Note Note Note Note Note Note Note Note ****** 
             -------------------------------------------------
         The first guess MUST BE MM5V3 input format because the code 
       was origigally developed for MM5 3DVAR. If you want to apply 
       this OBS preprocessor for WRF 3DVAR, you must choose another 
       option "not usingthe first guess" described below.
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 2) A decoded observation file,
 3) An observational error file (obserr.txt provided in this directory)
 4) Namelist file: namelist.3dvar_obs

 ** Note that there is an option to run 3DVAR OBSPROC not using the 
    first guess (MMINPU/MMOUT) as input. In this case,

    1) you need to modify the 3DVAR_OBSPROC/Makefile by deleting -DBKG 
       from CPPFLAGS, and re-compile the executable with make;
    2) Use another sample namelist file: namelist.3dvar_obs.sample1 as 
       the namelist.3dvar_obs, and edit it to provide some MM5 parameter 
       related to the reference state and the domain configuration.  
       These parameters can be found in the MM5 INTERPF namelist and MM5 
       TERRAIN namelist.

    This option gives users a flexible choice to specify any domain
    settings, not necessary related to the MMINPUT/MMOUT. 

The output files from 3dvar_obs are:

1) File "obs_gts.3dvar" is the obsevation data file input to 3dvar.

2) File "obs_ssmi_retrieval.3dvar" is generated only when SSMI data 
   are present in the input file.

3) Several diagnostic files (*.diag), and with these files users can
   know the troubles happened at which observations. 

4) The observational error profiles (*.txt)

5) In the printing file: 3dvar_obs.out, the flow of executions is shown
   step by step. Users can look at this file to monitor the program
   execution.
-------------------------------------------------------------------------------

To get the plots of the observation distribution

    After the observation data files: obs_gts.3dvar and/or 
  obs_ssmi_retrieval.3dvar produced, there is a utility to plot 
  the observation distribution for each type of observations.

  ** NCAR GRAPHICS library should ne installed in computer system.

  1) cd MAP_plot;
  2) Modify the shell script Map.csh, set the time window and OBS filename:

     set TIME_WINDOW_MIN = '1999081911'
     set TIME_ANALYSIS   = '1999081912'
     set TIME_WINDOW_MAX = '1999081913'

     set OBSDATA  = ../obs_gts.3dvar

  3) Type

     Map.csh

  4) When job completed, you'll have

    a gmeta file         :  gmeta.ccyymmddhh, and 
    a printed output file:  Map.out.ccyymmddhh 
    
    This gmeta file displayed the OBS distribution for each type of
    observations within the analysis domain in the time window. A quick
    glance at these plots will make you more comfortable to go ahead 
    with 3DVAR.

  ** There are several plotting subroutines under the directory Dir_map.
     With minor modifications, it is easy to plot the observed parameters:
     pressure, temperature, wind, etc.
 
----------------------------------------------------------------------------

To clean the directory obsproc

     make clean

   will clean the directory obsproc and its sub-directory. Be sure
   to keep the files: obs_gts.3dvar, etc. before cleaning.