Mesoscale model experience with UM at met.no

1
Mesoscale model experience with UM at met.no

• HIRLAM-ALADIN-MF miniworkshop 12/12/05
• Dag Bjørge

2
NWP collaboration in Europe
3
Short version of UM history at met.no

• 2002 Met Officemet.no NWP collaboration
• 2003 UM 3km, S-Norway grid, forced by Met Office
  20km and Hirlam 10km
• 2004 UM 4km, Norway grid, forced by HL 10km
• 2005 UM4XL, Norway/Sweden grid
• Nov2005 UM 1km, 3 small grids nested into UM4XL

4
How does it perform ?
5
Temperature
6
(No Transcript)
7
(No Transcript)
8
How is the UM model set up ?

• Climate and parameter files from Met
  Office ancil generation software
• LBCs from Hirlam 10km, interpolation with
  hl2um ( a met.no mixture of subroutines from UM
  and Hirlam )
• Some surface fields from Hirlam
• Some fields from previous 12h forecast

9

• 300 500 38 gridpoints,
• 4km resolution, 100s timestep
• Hourly postprocessing of surface fields, 3
  hourly pressure levels, 6 hourly model levels
• Boundary fields written out every 20 minutes for
  8h to 21h, for nesting of 3 UM 1km grids
• Runs on 280 CPUs, at 00UTC and 12UTC
• 48h forecast in 1h45min
• SGI Origin 3800 ( to be replaced next year )

10
What does it look like ?

• Quasi operational runs
• Different experiments
• Temperature problem solved (last week)

11
The model chain at met.no

• The turbulence project is used to describe the
  different models we now have available for
  forecasting at global to local ( airport ) scale.

12
Why a turbulence project ?

• På grunn av variable og vanskelige
  vindforhold bestemte kapteinen seg for å avbryte
  landinga. Da kom det et kraftig vindkast, og
  flyet smalt i bakken.

13

• Orography and resolution ( Seven sisters )

14
LBCs (50km) for Hirlam20 and Hirlam10 at 3 hour
intervals ( usually 6 hours old )
15
Hirlam10

• Starts from interpolated Hirlam20-analysis
• Boundaries ( frames ) from ECMWF
• Writes model dumps / history files every hour
  for nesting of Met Office Unified Model ( UM4 )

16
Hirlam10 and UM4
17
UM 4km with three UM 1km grids
18
The lifetime of LBCs in a small grid

• 30min intervals

19
UM1

• Starts for all airports as soon as UM4 has
  reached its 21 hour forecast
• UM1 starts at UM4 8h forecast, spins up for one
  hour
• Writes full model dumps at 9, 12, 15, 18 and
  21
• The turbulence model called Simra is then run
  for each of these instances

20
Operational cycles

• The model system as described is run twice a day
  at 00UTC and 12UTC
• Results are continously updated through the run,
  and can be studied as the forecast is produced

21
UM1, Sandnessjøen
22
Example of product, inflight cone
23
Example of product, inflight turbulence in cross
section
24
Some examples of different treatment of
precipitation in the different models
25
Precipitation shaddow, Hirlam / UM
26
Tromsø-area, Hirlam10 and UM4
27
Structure of precipitation, HL20/10, UM4, radar
28
Ad hoc setup of UM for the Barents sea

• Example showing a polar low

29
Polar low in UM 4km
30
Hirlam10 og UM4
31
UM4 23h, mslp and wind
32
Temperature problem ( solved last week )

• um505_thetamono_no.mf77
• A fix for the valley cooling problem. It is an
  active area of investigation and we may come up
  with alternative/better solutions, but this will
  hopefully work in the short term.

33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
Comparing Hirlam and UM

• Hirlam has been set up on exactly the same (
  horizontal ) grid as UM4XL
• Hirlam version 6.4.1
• H04 runs started in November on the met.no Linux
  cluster
• Some technical difficulties
• Computer resources (SGI) UM 3 HL

37
Heavy precipitation and flooding at the western
coast of Norway in November
38
(No Transcript)
39
Hirlam 4km UM 4km
40
Hirlam 4km, 12h accumulated precip
41
UM 4km, 12h accumulated precip