Model NetCDF data

For ease of comparing observations and models, it has been decided to create a common model format, based around NetCDF. has written a program that converts the ascii data from ECMWF, the Met Office and Meteo France into this new format, and preliminary output is shown on this page.

ECMWF

• NetCDF file: ECMWF_20020730_0000_Chilbolton.nc
• Corresponding CDL header: ECMWF_20020730_0000_Chilbolton_header.cdl.txt
• Plot of the fields: ECMWF_20020730_0000_Chilbolton.jpg

Met Office

• NetCDF file: MetOffice_20021014_0000_Chilbolton.nc
• Corresponding CDL header: MetOffice_20021014_0000_Chilbolton_header.cdl.txt
• Plot of the fields: MetOffice_20021014_0000_Chilbolton.jpg

Meteo France

• NetCDF file: MeteoFrance_20030227_0000_CHILBOLTON.nc
• Corresponding CDL header: MeteoFrance_20030227_0000_CHILBOLTON_header.cdl.txt
• Plot of the fields: MeteoFrance_20030227_0000_CHILBOLTON.jpg

Remaining problems

1. The following variables are not provided by all models so are not included in all of the netcdf files:
   • Screen temperature and surface altitude (not available from Meteo France)
   • Downwelling longwave and shortwave (not avaialble from Meteo France)
   • Latent and sensible heat fluxes (not available from Met Office)
   • Convective cloud fraction (Met Office only: included as a bonus!)
2. The following variables would be desirable but are only produced by one model so have not been included:
   • Surface clear-sky fluxes
   • Large-scale precipitation fraction
   • Albedo
3. The Meteo France specific humidity values go negative (although the corresponding relative humidities do not) which makes the calculated radar gas attenuations explode.
4. There is no indication yet in the files of which forecast each hour of data came from, especially important as some files are a concatenation of more than one forecast. I suggest a forecast_time(time) variable which gives the time in hours since the forecast was initialized, for each hour of the day.