Hello,
I’m working with the ‘ERA5 hourly data on single levels from 1940 to present’ dataset.
I need to apply some timezone conversion based on longitudes (similar to the AgERA5 approach), but it is becoming complex to work on the GRIB files using valid_time as it is not a ‘sliceable’ dimension.
Applying this transformations to a NetCDF like structure seems to be easier. However, I am weary of the discouragement of using NetCDF for operational/downstream services.
Now, I wanted to ask if the discouragement still applies if I download the GRIB and then apply the new GRIB to netCDF conversion method (GRIB to netCDF conversion on new CDS and ADS systems - Copernicus Knowledge Base - ECMWF Confluence Wiki) but only until the ‘Open the GRIB as a dictionary of xarray datasets’ step. I notice that at this step I already have a dataset with a ‘sliceable’ valid_time and I can easier apply the timezone conversion logic that I need. After which I aggregate to daily values and extract a parquet. I do not need to save it as a NetCDF.
Is this approach ‘safer’ than downloading the NetCDF? And how robust is the new GRIB to NetCDF conversion method expected to be? Is the code in the link expected to change often/rarely?
Thanks for your help,
Gabriel Andrade