AG Laserfernerkundung Pr

1
Results of Working Group 4Data Assimilation
Predictability
Teilnehmer Marco Apagaus, Andrea Moltani,
Christian Bartlott, Claudia Wunram, Hannes
Thiemann, Raymond McCord, George Craig, Galina
Dick, Armin Mathes, Christian Koziar,
Hans-Stefan Bauer, Matthias Grzeschik, Florian
Zus, Volker Wulfmeyer, Andreas Behrendt
2
Since there were nothing really new to discuss
since the last meeting, the discussion focused on
practical issues concerning the following points

• Data Management
• How should the model and observed data be
  provided to the data base?
• How should the data be stored in the data base?
• Real-Time Assimilation
• Data format
• Best way to provide the data
• Visualization of COPS / D-PHASE model products
• Data for the COPS Operation Center

3
Data Management

• The D-Phase definition for model data (TIGGE)
  should be used to reduce the amount of data to be
  send to the data base.
• The model data for the data base has to be
  provided in GRIB1 format. Furthermore, metadata
  has to be provided to describe the single data
  files provided to the data base.
• A form will be sent around which collects the
  information necessary to properly store the
  data.
• Observations for the data base shall be provided
  in NetCDF format. Here, NetCDF cf is preferred,
  since it defines standard names for the stored
  parameters ? easier comparisons possible.
• Also for the provided observations the content of
  the data sets has to be described with meta
  data, since the NetCDFHeaders are not visible
  in the data base.

4
RealTime Assimilation

• It is essential that we know as soon as possible
  which data can be provided in real time (within a
  cut-off time of e.g. 2 hours).
• A direct communication with the observing groups
  is essential.
• The real-time efforts shall be independent of the
  data base
• Data used for real-time assimilation has to be
  provided within a cut-off time to ensure an
  operational assimilation cycle.
• The assimilating group (as all modeling groups in
  general) are responsible to store all data
  necessary to reproduce every single forecast.
• In the data base only a subset of the model
  output is stored.
• UHOH will provide an ftp Server, where data for
  the real-time assimilation (that is not
  automatically stored in the data base in Hamburg)
  can be stored.
• However, it is not possible that each observing
  group provides their own format ! ? Communication
  necessary
• The observing groups using this server have to
  ensure in addition that their observations are
  provided to the data base in the expected NetCDF
  format.

5
Operational Setup - Overview
Blue arrows Data transportRed Computing
(data processing, assimilation, forecast, )
Green Idle time (time window for data
retrieval)Blue Observation IntervalBlack
Technical Problems
6
Products provided to the Operation Center

• It is important to define as soon as possible the
  data links to define who is sending what.
• It has to be decided as soon as possible which
  model forecast products and observations are
  needed by the OC for mission planning.
• Is the same information as for D-PHASE necessary
  or only a subset.
• From this decision a list of needed plots has to
  be created.
• To ensure an easy comparison of the provided
  model forecasts, it is essential to use one
  graphics package and the same selected color
  tables for the vizualisation. Here, the following
  suggestion was made
• As visualization tool GrADS is suggested, since
  it is freely available and specialized to
  visualize model data.
• Every modeling group provides a sample GRIB1 file
  to UHOH.
• We try to read and process the data and log
  problems we run into.
• This log is send back to the modelling group and
  the problems are solved iteratively.
• Aim is that at the end a set of scripts are
  available so that every modeling group can create
  the necessary plots in quasi real-time.
• However, communication is necessary about how the
  work can be distributed within COPS.