Formation

1
Damian Wilson, Met Office / Jean-Marcel Piriou,
Meteo-France
User requirements how to improve NWP models from
in-situ data?
CLOUDNET Workshop / Utrecht 21-22/10/2002
2
Model temporal integration

• Model prediction errors are to be found
• in the parameterizations (radiative
  simplifications of the RTE)
• in estimating sub-grid scale variability (e.g
  link between ltqlgt and ltdgt ? ltqlqlgt)
• in the prognostic variables from current time
  step (cum. errors or analysis errors)

3
CLOUDNET useful for modellers to provide

• RS, radar, lidar, ? variables u, v, T, qv, ql,
  qi, cloudiness,
• Sub-grid scale variability, correlations of these
  at ? space and time resolutions
• Fluxes radiation, precipitation,

4
What has already been done in other projects to
compare models to in-situ data a recent example
Evaluating mesoscale model predictions of clouds
and radiation with SGP ARM data over a seasonal
time scale, Françoise Guichard and al.,
submitted to MWR, revised version July 2002
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
Comparison model versus observations

• Difference between model and observations come
  from
• Model errors
• Observational errors
• Transform errors
• To have an idea whether a difference is a model
  error we need
• To assess observational and transform errors
• To compare model-observations from the different
  sites
• To compare model-observations from the different
  models and to relate their differences to model
  formulations
• ? Importance of having the same set of
  well-defined variables AND the same file formats
• Comparisons may be handled (i) qualitatively (ii)
  quantitatively (transforms)