The gdas_utility.py script downloads and prepares the initial states for the GraphCast model from GDAS data. Currently, the downloading process gets the GDAS data from the Amazon S3 bucket, however, the S3 bucket is 12-18 hours behind and it cannot be used for operational use. In this case, the gdas_utility.py needs to be updated to get the GDAS data from NOMADS which is up to date and can be used for generating real-time graphcast forecasts. In addition, we need the functionality to run the graphcast model on hera and it needs some updates on the scripts for loading required modules on the hera machine.
The
gdas_utility.pyscript downloads and prepares the initial states for the GraphCast model from GDAS data. Currently, the downloading process gets the GDAS data from the Amazon S3 bucket, however, the S3 bucket is 12-18 hours behind and it cannot be used for operational use. In this case, thegdas_utility.pyneeds to be updated to get the GDAS data from NOMADS which is up to date and can be used for generating real-time graphcast forecasts. In addition, we need the functionality to run the graphcast model on hera and it needs some updates on the scripts for loading required modules on the hera machine.