Vaisala/University of Washington Real-observation Experiments

1
Vaisala/University of WashingtonReal-observation
Experiments

• Clifford Mass, Gregory Hakim,
• Phil Regulski, Ryan Torn,
• Jennifer Fletcher
• Department of Atmospheric Sciences
• University of Washington
• October 2006

2
Data Assimilation

• Fusion of models observations.
• Need error statistics!
• Spreads observational information.
• Analysis
• smaller error than observations.
• smaller error than model estimate of obs.

3
Data Assimilation in a Nutshell
observations model
4
Observation (green) Background (blue) PDFs
5
Analysis (red) PDF---higher density!
6
More-Accurate Observation
7
Less-Accurate Observation
8
Ensemble Kalman Filter
Crux use an ensemble of fully non-linear
forecasts to model the statistics of the
background (expected value and covariance matrix).
Advantages

• No à priori assumption about covariance
  state-dependent corrections.
• Ensemble forecasts proceed immediately without
  perturbations.

9
LR Lightning Real-Observation Experiment
10
Establish geographical domain for
Real-observation Experiment

• Dec 12-24, 2004
• Domain location that encompasses Pessi/Businger
  previously studied storm
• Pacific Ocean
• Low observation density location of important
  storm tracks errors propagate downwind to
  mainland United States
• North America
• High observation density forecast improvement
  interest area included to see the impact of
  regions of low and high observation densities

11
Real time observations

• Control case
• Observation locations from real data
• Radiosondes
• Surface stations (ASOS, ship, buoy)
• ACARS
• Cloud drift-winds (no sat radiances)
• Lightning experiment
• Assimilation of convective rain rate

12
A Traditional Observation Network2004100118
ACAR observations Soundings Surface observations
13
Experiment observationsACARS Obs.
14
Experiment observationsCloud Track Wind Obs.
15
Experiment observations

• Radiosonde Obs
• Surface Stations

16
Experiment observationsLTNG Obs
17
Experiment observationsLTNG Obs
18
Experiment observations

• Lightning assimilation
• Real LR LTNG strike is identified
• WRF-ENKF locates LTNG and feeds the experimental
  run the convective precipitation from the Pessi
  convective rain rate/LTNG rate relationship at
  the LTNG coordinates

19
Real-observation Experiments
20
2-Week Experiment

• 100 by 86 grid points
• 45-km resolution
• 33 vertical levels
• 48 ensemble members
• Assimilation every 6 hours
• Forecasts 6, 12, 18, 24, 30, 36, 42 and 48 hours

21
2-Week Experiment

• WRF ensemble Kalman filter settings
• Square root filter (Whitaker and Hamill, 2002)
• Horizontal localization Gaspari and Cohn 5th
  order piecewise
• Fixed covariance perturbations to lateral
  boundaries
• Constant uniform covariance inflation method
• Localization radius 2000km

22
Weather Pattern Sea level pressurePeriod
characterized by extratropical cyclone
23
Weather Pattern H500 Period with active weather
pattern Trough dominated
24
Control Experiments

• Control experiment 1
• Not enough variance
• Increase inflation factor
• Control experiment 2
• Still low variance
• Switching inflation method from constant
  inflation to Zhang method
• Control experiment 3
• Good variance

25
Control ExperimentsControl Experiment 1 Too
low variance
26
Control ExperimentsControl Experiment 1 Too
low variance
27
Control ExperimentsControl Experiment 2 More
variance but still too low
28
Control ExperimentsControl Experiment 3
Acceptable variance
29
Control ExperimentsControl Experiment 3
Acceptable variance
30
Control ExperimentsControl Experiment 3
Acceptable variance
31
Control ExperimentsControl Experiment 3
Acceptable variance
32
Control Experiments

• Control experiment 3
• Analysis variance
• H500mb
• T2m, T850mb, T300mb
• Y300mb
• SLP, REFL, RAINC etc
• Observation verification
• Rank histograms
• Profile
• Other

33
Control Experiments
34
Control Experiments
35
Control Experiments
36
Control Experiments
37
Control Experiments
38
Control Experiments
39
Test Experiments

• Coding LTNG assimilation into WRF-ENKF
• Assimilated LTNG rate
• Transformed LTNG rate into convective rain rate
• Final coding
• Testing
• Experiment Run (LTNG assimilation - 1.5 weeks)
• Comparisons

40
Test ExperimentsExample of comparison products

• Analysis fields
• H500, SLP, WINDS, RAINC
• Forecast fields
• All forecast hours

41
Test ExperimentsExample of comparison products
42
Test ExperimentsExample of comparison products
43
Test ExperimentsExample of comparison products
44
Summary

• Where we are at
• Data observations gathered from Dec 2002
• Cloud track winds
• ACARS
• Surface
• Radiosondes
• LTNG
• Performed control runs
• Final stages of coding LTNG assimilation code for
  real observation WRF-ENKF experiments
• Ongoing statistical analysis

45
Summary

• Future possibilities
• Alternative assimilation fields
• In-house rain rate/LTNG rate relationship
• Different domains
• Other sample storms

46
6-month goals

• Real-time lightning data feed into UW-ATMS
  WRF-ENKF system
• OSSE DE simulations
• Robust and flexible OSSE and real observation
  experiment systems
• Creation of flexible LTNG assimilation modules so
  new experiments can be quickly altered in
  parameter file
• Other suggestions and comments )