PowerPointPrsentation

1
T2m Nowcasting Statistical vs. Physical
Adaptation T. HAIDEN, S. GREILBERGER, A.
SCHMALWIESER ZAMG, Vienna, Austria
Cloudiness

• Problem
• Hourly updated forecasts
• Error characteristics
• Evening cooling
• Process-oriented approach

T2m
Advection
Wind speed
Soil temperature
2
Problem
Customer request (power industry) High
precision, hourly, T2m point forecasts for the
next 4 hours (12 hours), during winter half
year, at location Vienna Hohe-Warte.
3
T2m nowcasting error
Adjusted LAM skill gt Climatology skill gt LAM DMO
skill
4
T2m error distribution during the first forecast
hours

• Error mostly between 2 and 2 K
• Occasional outliers with error of 3-6 K
  (non-Gaussian)

5
Error characteristics

• Air mass change (frontal passage) timing
  problem
• Amount/speed of evening cooling overestimated

6
Error characteristics (contd.)
Improvement by statistical adaptation (MOS,
Kalman, etc) limited by NWP forecast quality
7
Evening cooling
little/no cooling
cooling
cooling
8
Reduced leeside cooling
3-d high resolution (1 km) model
necessary? Statistical correction?
9
Low stratus
MODEL OBS

• Temperature inversion too smooth
• Inversion base too warm ? cloudiness
  underestimated
• Underestimated cloudiness ? PBL cooling too weak

10
Low stratus 1-d experiments
Experiment I Vertical diffusion subsidence
throughout PBL
00 UTC obs 12 UTC obs 12 UTC forecast
11
Low stratus 1-d experiments
Experiment II Vertical diffusion subsidence
above PBL
00 UTC obs 12 UTC obs 12 UTC forecast
12
Low stratus 1-d experiments
Experiment III Vertical diffusion subsidence
above PBL cloud-top cooling
00 UTC obs 12 UTC obs 12 UTC forecast
13
Process-oriented approach
Cloudiness
T2m
Advection
Wind speed
Soil

• 1-d model radiation fluxes, turbulent fluxes,
  surface exchange
• Run every hour, use adapted model sounding as
  initial condition
• Cloudiness extrapolate observed trend (
  trajectories)
• Advection apply trajectories to observed
  temperature distribution
• Wind speed weighted combination of model and
  observation
• Soil use observed near-surface temperatures,
  soil conditions
• ! perform separate verification of individual
  modules