Date

04 Jan 2023  

Attendees

• Matthew Griffith 
• Sebastien Villaume 
• Gabriele Arduini 
• Xabier Pedruzo Bagazgoitia 
• Gianpaolo Balsamo 

Apologies

• Richard Mladek 

Goals

• Meeting to discuss IFS tile variables in GRIB2 and considerations for designing a template that caters to the needs of IFS output on tiles.

Discussion items

• The IFS tile model (HTESSEL) currently has ten fixed tiles covering land and sea, with urban tiles and urban cover grib files in the pipeline.
• Each simulation is accompanied by 4 grib files – TVH (High Vegetation) and TVL (Low Vegetation) with corresponding cover CVH and CVL for the simulation grid. Within these, only the dominant tile type is represented within the tile grouping, e.g. if a given grid box has 40% needleleaf, 25% broadleaf, 25% grassland and 10% shrubland, this would be represented by 65% needleleaf in TVH, CVH and 35% grassland in TVL, CVL. This simplification allows easier inter comparability between different model resolutions.
  These are indicated in the grib files by integers in TVX with corresponding fractions in CVX for a given grid cell. The integer types correspond to the following table (from IFS documentation):
  
• Sum of CVH and CVL is not necessarily always 100% – it can be less. For example, bare ground is not included. New tiles such as urban tiles are added by subtracting the relevant fraction from the nature quota of the tile.
• The canopy cover fraction is currently assumed to be very general as the types of cover (e.g. deciduous) are difficult to specify.
• Tiles covered with snow/dew or flooded use a new tile type e.g. snow which replaces the fraction of the tile which was previously another type. This was discussed and it was suggested that one could instead just use the standard tile types (e.g. high vegetation cover) and then encode a variable on that tile such as snow-cover, dew-cover, inundation to indicate whether that tile is covered and to what extent.
• The current ECMWF tile output uses an Experimental Product GRIB code (paramId=100) on various model levels, which correspond to different tiles. See  for details and an example GRIB output using this format here. Output on tiles 1-9 is recorded for 6 variables on levels 1-9, 11-19, ..., 51-59.
• There is the possibility that ECMWF will move to SURFEX at some point in the future.

Action items

• Decide on the format and create tile template prototype to pass on for testing: Sebastien Villaume Matthew Griffith 
• Get sample tile output GRIB files.