Aktionsprogramm 2003

1
Early results with rainfall assimilation at
DWDRADVOR-OP, supported by LAWA
Stefan Klink Data Assimilation
Section stefan.klink_at_dwd.de
2
Early results with rainfall assimilation at DWD

• Radar data
• Real case study
• Humidity adjustment
• Filtering of vertical profiles
• Influence of convective parameterisation scheme
• Influence of mesh size
• Summary

3
Radar data

• Input data currently used
• radar reflectivity at the ground (int.
  Composite Germany and surrounding countries)
• pixel size 4km ? 4km
• time resolution 15 minutes (data from volume
  scan)
• conversion reflectivity to precipitation rate by
  simple Z-R-relation
• problems clutter, only 7 reflectivity classes

4
Radar data

• new radar data
• input data reflectivities from precipitation
  scan, mesh 1km ? 1, time resolution 5 minutes
• correction of orographic attenuation
• variable Z-R-relation
• compositing
• still incomplete clutter correction

5
Real case study experimental setup

• Convective event on 28. Aug. 2002 over western
  Germany
• LM runs on the operational domain
• mesh size 7 km
• LM runs on an experimental domain
• covering almost Germany
• mesh size 2.8 km
• convective parameterisation scheme switched off
• laterally driven by GME
• 6h Nudging LHN starting at 6 UTC
• adjustment of humidity during LHN
• 6h free forecast starting at 12 UTC

6
Real case study humidity adjustment

• Increased heating
• ? increase q to reach f100 over nudging
  timescale ?

• Reduced heating
• ? decrease q (maintain f)

7
Real case study humidity adjustment

• LM runs on the experimental domain
• mesh size 2.8 km
• without humidity adjustment during LHN
• with humidity adjustment during LHN

8
Real case study humidity adjustment

• Hourly sums of modeled and radar-observed
  precipitation in mm/h
• Nudging 08-09 UTC
• Nudging11-12 UTC

9
Real case study humidity adjustment

• free forecast14-15 UTC

10
Real case study filtering of vertical profiles
vertical cross section W-E, t 11 UTC, with
humidity adjustment specific cloud liquid
water content in g/kg
no filtering
filtering
11
Real case study influence of convective
parameterisation scheme

• LM runs on the operational domain
• mesh size 7 km
• no adjustment of humidity during LHN
• without convective parameterisation scheme
• with convective parameterisation scheme

12
Real case study influence of convective
parameterisation scheme

• Hourly sums of modeled and radar-observed
  precipitation in mm/h
• Nudging 08-09 UTC
• Nudging 09-10 UTC

13
Real case study influence of convective
parameterisation scheme

• Nudging
• 11-12 UTC
• free forecast
• 12-13 UTC

14
Real case study influence of convective
parameterisation scheme

• free forecast
• 14-15 UTC

15
Case Studies comparison mesh sizes 7 km and 2.8
km

• LM runs
• CTRL 7 km
• mesh size 7km (operational domain)
• LHN 7 km
• mesh size 7 km (operational domain)
• LHN 2.8 km
• mesh size 2.8 km (smaller model domain)

16
Case Studies comparison mesh sizes 7 km and 2.8
km

• Hourly sums of modeled and radar-observed
  precipitation in mm/h
• Nudging06- 07 UTC
• Nudging 08-09 UTC

17
Case Studies comparison mesh sizes 7 km and 2.8
km

• Nudging
• 11-12 UTC
• free forecast
• 12-13 UTC

18
Case Studies comparison mesh sizes 7 km and 2.8
km

• free forecast
• 14-15 UTC
• free forecast
• 17-18 UTC

19
Summary

• Results
• adjustment of humidity supports the assimilation
  of precipitation and leads to more exact free
  forecasts
• explicit simulation of convection produces more
  realistic patterns and amounts of precipitation
  than the corresponding model run with a
  parameterisation scheme
• model with 2.8 km mesh size shows potential for
  further improvements of precipitation forecast
• sporadical occurrence of spurious precipitation
• influence of the assimilation lasts for several
  hours

20
Summary

• Future work
• numerical problems with noisy profiles of latent
  heat (2 ?z-structures)
• quality of radar data
• improvements to extend the influence of
  assimilation during free model run
• prognostic precipitation