NOAA-EMC / Pangu-Weather

An official implementation of Pangu-Weather
1 stars 1 forks source link

Add script to prepare input data from GDAS 0.25 product #3

Closed LinlinCui-NOAA closed 9 months ago

LinlinCui-NOAA commented 9 months ago

Description

This PR provides a script to prepare inputs from GDAS 0.25 degree data for Pangu-Weather model. The user can download data either from NOAA s3 bucket or NCEP nomads server (which can specify variables and levels). The script also provided two methods to process grib2 files. One is wgrib2, which needs to be installed from the sourcecode. The other one is pygrib, which is a python package, and can be installed with pip. Both methods will create two numpy's npy format files: input_surface.npy and input_upper.npy.

Linked Issues and Pull Requests

AssociatedIssue to close

Blocking Dependencies

Anticipated Changes

Input data

Needed libraries

SadeghTabas-NOAA commented 9 months ago

@LinlinCui-NOAA Did you run the Pangu-Weather model as well?

LinlinCui-NOAA commented 9 months ago

@SadeghTabas-NOAA Yes, I ran the example of CPU version with the 24-hour model. And it succeeded.

The inputs and outputs are on Hera: /scratch1/NCEPDEV/nems/Linlin.Cui/Tests/pangu-weather

input*.npy output*.npy

SadeghTabas-NOAA commented 9 months ago

@SadeghTabas-NOAA Yes, I ran the example of CPU version with the 24-hour model. And it succeeded.

Great! I see you took care of unit conversions as well. As the inputs are very similar, that makes everything straightforward. I hope we will have similar input in FourCastNet as well.