1

I have a working program on an x86 linux machine that uses pynio to read grib2 files from NOAA's NBM product. A simplified version of the code looks like the following:

import Nio

f = Nio.open_file('foo.grib2', mode='r', options=None)
vars = list(f.variables)

When I print the list of variables, I get things like

APCP_P8_L1_acc1h
APCP_P9_L1_acc1h
APTMP_P0_L103
ASNOW_P8_L1_acc1h
CAPE_P0_L1
CAPE_P2_L1
CEIL_P0_L2
CEIL_P0_L215
CEIL_P5_L215
CEIL_probability2
DIST_P0_2L1_103
etc.

Now I have to port my program to an architecture (aarch64) that doesn't support the Nio module, so I apparently have to use the cfgrib module instead and began working with that today.

varlist = []

datasets = cfgrib.open_datasets('foo.grib2') # Opens a single file as a list of datasets
    
for ds in datasets:
    for v in ds.variables:
        varlist.append(v)
        
varlist = list(set(varlist)) # eliminate duplicate entries

This new list contains entries that look like:

aptmp
asnow
atmosphere
atmosphereSingleLayer
cape
ceil
cloudBase
cloudCeiling
cloudTop
d2m
dswrf
gust
h
heightAboveGround
heightAboveSea
etc.

The important points to note here are that

  • There is no direct correspondence between variable names obtained using the two different methods;
  • Multiple variables that I previously worked with (e.g., CEIL_P0_L2, CEIL_P0_L215, CEIL_P5_L215) seem to correspond to recurrences of the variable (e.g., ceil) in different sub-datasets when I read the file with cfgrib;
  • Other variable names show up (e.g., atmosphere, atmosphereSingleLayer) that seems to play a role other than as an actual numerical field.
  • A number of grid fields show up with the variable name unknown (not shown in partial listings above), and I have no idea how to figure out what these are, even printing all of the fields for the dataset.

My questions: Why are the variable names different, and, more importantly, how do I use cfgrib to reliably identify and extract the fields corresponding to the original variable names as seen (or constructed?) by Nio?

Caveat: The original Nio program was last run on NBM files from nearly a year ago, whereas I'm running cfgrib on files obtained from the same source in August 2021. I suppose the internal structure of these files could have changed in the interim, though that seems unlikely.

Grant Petty
  • 1,151
  • 1
  • 13
  • 27

1 Answers1

0

So first of all, this is not a question that belongs to stackoverflow, Please use e.g. https://earthscience.stackexchange.com/ in the future.

The reason why the names are different is that the software packages uses different default grib_tables. Within these grib_tables the variables and identifiers are defined and mapped. So if you want to have the same results, you have to synchronize the grib_tables. As cfgrib uses ECCODES as backend, you might take a look into ECCODES_DEFINTIONS_PATH

dl.meteo
  • 1,658
  • 15
  • 25