Gzip reduces the size of the named  files  using  Lempel-Ziv
coding  (LZ77).

The source code is available on most platforms. Beside UNIX-Systems a
number of special targets like MSDOS, OS/2, VMS, Atari, Amiga, Primos
are supported. The source code including installation hints is available
for free from "The GNU Project" (http://www.delorie.com/gnu/).

In case of trouble on Windows, please try winzip (http://www.winzip.com).

The gzip'ed files are originally ASCII. The structure of the ASCII
files is like that

   header (1 record with 8 4-byte-integer)
   field  (1 record with n 4-byte-reals)
   header
   field
   and so on.
 
   the header consists of
       1. code number
       2. level
       3. date [yymmdd]
       4. hour
       5. nlon [or dim of spectral field, if encountered]
       6. nlat [or 1, if spectral field]
       7. DDC model ID
       8. Originating center ID
 
   so that the dimension of the data field is always given
   by header(5) * header(6).
 
A sample FORTRAN program to read a file of the above structure can be 
seen below. Let's assume that you want to read a data file containing
DKRZ's ECHAM4/OPYC3 datasets. Then (from the grid description) you will
know that one horizontal 2D field has 128*64 datapoints. They are stored
sequentially latitudewise. For instance the first line j in the DKRZ 
output represents the latitude 87.8638 deg. N. All the data for this
line would appear as the first 128 values in the ASCII file representing
all values of i from i = 1 (0 deg E) to i = 128 (357.1875 deg E).
The next 128 values would then represent the values from j = 2 (85.0965 deg N).

      program ipccread
      parameter (nheaddim=8,inputunit=66)
      parameter (nlon=128,nlat=64)
      character*80 inputfile
      integer      header(nheaddim)
      real         field(nlon,nlat)
C
      call getarg(1,inputfile)
      open (inputunit,file=inputfile,form='formatted')
C
  555 read (inputunit,1000,end=999) header
      read (inputunit,2000) field
      goto 555
C
  999 stop ' at eof'
 1000 format(8i10)
 2000 format(8e13.6)
      end

This sample program should work nearly unchanged for all datasets of
all models available at the DDC. The only thing to be changed for
different models is the line
      parameter (nlon=128,nlat=64)

Note too, that the orientation of latitudes can differ between models.
An orientation "North -> South" means that the datapoints of the
data line j = 1 lies on the northermost Latitude, "South -> North"
means that the datapoints of the data line j = 1 lies on the
southermost Latitude.

Here is the information on dimensions and orientation for the models:

Model     Dimension                      Latitude orientation
-----     ---------                      --------------------
HadCM2    parameter (nlon=96,nlat=73)    North -> South
GFDL      parameter (nlon=48,nlat=40)    South -> North
CCCma     parameter (nlon=96,nlat=48)    South -> North
DKRZ      parameter (nlon=128,nlat=64)   North -> South
 
Good luck
 
------------- Deutsches Klimarechenzentrum GmbH ---------------
 
Arno Hellbach                      email: hellbach@dkrz.de
Deutsches Klimarechenzentrum       Tel +49 40 41173363         
Bundesstr. 55                      Fax +49 40 41173298         
D-20146 Hamburg
 
---------------------------------------------------------------