The important target before full production archiving is to archive in MARS full UERRA samples from all already agreed
datasets by the June/July 2016.
To be able to do that there is a need for full samples from UERRA systems containing all agreed parameters, levels, steps etc. One full sample from one day (and all specified runs as e.g. output frequencies vary) per each available dataset is required at this stage. In the following the instructions and tools how to convert and archive UERRA data in MARS will be provided to each provider.
The UERRA data sample format
- Only the agreed parameters should be included (all which will be produced) and only on the defined levels, with the defined frequency etc.
- If the data is already in GRIB2 format it is should be encoded as UERRA test data (productionStatusOfProcessedData=9)
- Each sample dataset should be split for given day only by
- forecasting system type(deterministic/ensemble)
- forecast type (analysis/forecast)
- vertical level (pl/ml/hl/sfc/soil)
- model run(00/06/12/18UTC)
- Regarding vertical level type follow exactly the defined groups of parameters as defined in the UERRA parameters page.
- Follow this naming convention of the sample files:
<SYTEM TYPE>.<FORECAST TYPE>.<LEVEL TYPE>.<YYYYMMDDRR>.<GRIB ENCODING>
where
- SYTEM TYPE = det/ens for deterministic or ensemble system
- FORECAST TYPE = an/fc
- LEVEL TYPE = ml/pl/hl/sl/sl for model/pressure/height/surface or soil levels
- GRIB ENCODING = grib1/grib2 depending on GRIB edition used
e.g. det.an.sl.1962010106.grib1[grib2] for surface level analysis parameters (all steps) from deterministic system run on 1.1.1962, 6h.
Checking the data
It is up to every provider to provide only the required and correct data in their full samples and even more important during the later production phase . Here are examples of some common data problems:
- some parameters are missing compared to those specified as available from a model
- some parameters are redundant compared to those requested for UERRA
- some missing or redundant steps, levels, type of forecasts etc than requested
- wrong units
- wrong sign convention for some fluxes
- orography in m not gpm
- wrong intervals "since previous-postprocessing"
- zero fields
- some excessive unrealistic values (i.e. negative accumulated precipitation or grid point storms)
- completely wrong fields which could be revealed only by plotting and looking at maps (strange leftovers from interpolation; wavy patterns etc)
Please double check all your data using the provided tools but also by plotting some sample fields for each parameter or using your own data quality checks. |
GRIB conversion and encoding related pages
To track progress of the work on conversion to GRIB2 and test archiving in MARS the progress report pages have been created and are linked in the column "Progress log" of the top table in the progress status page.
The data providers should fill and keep up-to-date GRIB1 parameter numbers from their input data matching UERRA parameters in the conversion tables:
The providers who produce GRIB2 files directly should follow the required encoding definitions and rules described at the following pages