Robb Kambic wrote:
Hiya,
As stated before, the NCEP data with multiple grids was split using LDM
pqact actions. Since that's not possible for the africa data, the grid_id
in the PDS section was used to split the data. But, the africa data had
grid_id = 255 for both grids so no split occurred. The data is organized
according to forecast time, first grid has forecasts 6-240 and second grid
has forecasts 246-360. The first grid has higher resolution then the second
grid. I can probably modify a current program to do the split.
yes, you have to split by reading the gds, not through pqact.
There is also a problem with the TOM_REALFORECASTS.20080112_v1_pf.grb grib
what is this dataset?
that has ~50 perturbations in the file. The variables look the same to the
our grib reader because ECMWF uses local additions to the PDS section to
distinguish the variables. Using the ECMWF decoder gribex, the following
local use information was displayed.
ECMWF local usage identifier. 30
Class. 1
Type. 11
Stream. 1035
Version number or Experiment identifier. 0001
I don't know what this means but it probably identifies the variables so
they aren't treated the same. According to the WMO documentation it states:
29-40 Reserved: need not be present
41-nn Reserved for originating center use
So ECMWF is complying with the Grib1 specfications.
The other 3 types of ECMWF files don't contain the perturbations so they
could be split according to forecast times.
My question, is where do I go from here?
Apparently this is the same problem as in BUFR - local tables. So I suppose
we have to get the local tables from ecmwf ?
Robb...
Section 1 - Product Definition Section.
---------------------------------------
Code Table 2 Version Number. 128
Originating centre identifier. 98
Model identification. 130
Grid definition. 255
Flag (Code Table 1) 10000000
Parameter identifier (Code Table 2). 167
Type of level (Code Table 3). 1
Value 1 of level (Code Table 3). 0
Value 2 of level (Code Table 3). 0
Year of reference time of data. 8 (2008)
Month of reference time of data. 1
Day of reference time of data. 12
Hour of reference time of data. 0
Minute of reference time of data. 0
Time unit (Code Table 4). 1
Time range one. 6
Time range two. 0
Time range indicator (Code Table 5) 0
Number averaged. 0
Number missing from average. 0
Century of reference time of data. 21
Sub-centre identifier. 0
Units decimal scaling factor. 0
ECMWF local usage identifier. 30
Class. 1
Type. 11
Stream. 1035
Version number or Experiment identifier. 0001
===============================================================================
Robb Kambic Unidata Program Center
Software Engineer III Univ. Corp for Atmospheric Research
address@hidden WWW: http://www.unidata.ucar.edu/
===============================================================================