Doppler radar wind data assimilation in HIRLAM 3DVar

1
Doppler radar wind data assimilation in HIRLAM
3D-Var

• SRNWP/COST-717 WG-3
• Session on assimilation of 'non-conventional
  data'
• 8.10.2003
• Kirsti Salonen (FMI),
• Heikki Järvinen (FMI),
• Magnus Lindskog (SMHI)

2
Contents

• Superobservations
• Variational data assimilation
• Observation operator for Doppler radar radial
  winds
• Fit of the observations with the model
  counterpart
• Future plans

3
Superobservations (1)

• Available raw radial wind data with high
  temporal and spatial density.
• Observations represent partly phenomena which are
  not resolved by the NWP model.
• Calculating spatial averages, Superobservations
  (SO), decreases the representativeness error.
• To minimize horizontal correlation, each piece of
  raw data is allowed to influence only one SO.

4
Superobservations (2)
5
3D Variational data assimilation

• Based on minimization of the cost function
• Observation operator H produces the model
  counterpart of the observed quantity.

6
Observation operator for Doppler radar radial
winds

• Interpolation of the NWP model wind (u,v) to
  the observation location.
• Projection of the interpolated NWP model wind
  towards the radar vhu sin?v cos??
• Projection of vh on the slanted direction of the
  radar beam vrvhcos(??).

7
Broadening of the radar beam

• Gaussian averaging kernel
• The effect of radar horizon is taken into accout.
• An empirical upper limit for the averaging
  kernel is set to 1.5 times the half-beamwidth.

8
Bending of the radar beam

• Taken into account by using Snell's law
• Beam path is accumulated until the radar beam
  reaches the observation location.
• Effective elevation angle is calculated and used
  in the projection of vh.

9
The fit of the observations with the model
counterpart
TEMP
Radar
10
Free parameters associated to SO

• Number of polar bins used in the SO generation
  (NPB).
• Measurement range.
• Variance of the raw radial wind values forming a
  SO (VRW).

11
Distribution of the observations

• Non-meteorological targets
• Birds
• Ships
• Remaining ground clutter
• Aliasing problems.

12
Applying quality criteria

• Measurement range less than or equal to 100km.
• NPB more than or equal to 5.
• VRW less than or equal to 10 m/s.

13
Future plans

• Parallel data assimilation experiments with
  Finnish and Swedish radar data.
• Make use of novel de-aliasing algorithm.
• Summer and winter months.
• Impact studies of using radar wind data on
  forecasting severe weather events.
• Studies with Luosto radar data.