Assimilating Satellite Data into Antarctic Mesoscale Prediction System AMPS

1
Assimilating Satellite Data into Antarctic
Mesoscale Prediction System (AMPS)

• Hui Shao1 (huishao_at_ucar.edu)
• Zhiquan Liu2, Thomas Auligne2, Dale Barker1,2,
  Syed Rizvi2 and Jordan Powers2
• 1Data Assimilation Testbed Center (DATC)
• 2Mesoscale and Microscale Meteorology Division
  (MMM)
• National Center For Atmospheric Research
• The 3rd Meteorological Observations, Modeling
  Forecasting Workshop
• Madison, Wisconsin, June 9-12, 2008

2
Labs and Programs in NCAR
AMPS real-time
AMPS DA update
3
Data Assimilation in the Polar Areas
Sonde sites in AMPS 60km domain

• In-situ observations
• Topography
• Cloud cover
• Forecast models physical parameterizations

GPS
Monthly observation number
Sonde
4
AMPS Testbed at NCAR/DATC

• Operational Configurations
• Horizontal resolution 60km
• Vertical levels 31
• Model top 50mb
• Assimilation window ?2hr
• Testing period October, 2006

(From Dale Barker)
5
WRF-Var Observations (Used in AMPS)
Conventional Surface (SYNOP, METAR, SHIP,
BUOY). Upper air (TEMP, PIBAL, AIREP,
ACARS). Remotely sensed retrievals Atmospheric
Motion Vectors (geo/polar). Ground-based GPS
Total Precipitable Water. SSM/I oceanic surface
wind speed and TPW. Scatterometer oceanic surface
winds. Wind Profiler. Radar. Satellite
temperature/humidities. GPS refractivity (e.g.
COSMIC). Radiances SSM/I brightness
temperatures. Direct radiance assimilation
(RTTOVS, CRTM).
6
Impact Of WRF-Var Quality Control (QC)

• 1st (gross) QC performed by observation
  preprocessor.
• 2nd (difference between ob and forecast) QC
  performed in WRF-Var.
• Main impact of 2nd QC is on surface observations.
• Rejection rates will reduce with higher
  resolution, higher-order interpolation.

SYNOP
METAR
SHIP
SONDE
BUOY
AIREP
PILOT
COSMIC
SATOB
BEFORE 2nd QC
AFTER 2nd QC
AIRSRET
QSCAT
7
GPS Radio Occultation Techniques and Measurements

• The ray path of a transmitted radio signal during
  an occultation is bent due to the atmospheric
  refraction related to the atmospheric state (T, p
  and q) in neutral atmosphere.
• Constellation Observing System for Meteorology,
  Ionosphere Climate (COSMIC)
• Features of measurements
• high vertical resolution
• all-weather
• unbiased
• coarse horizontal resolution
• multi-path problem in lower levels

(From Bill Kuo)
Measurement time 0836 UTC 21 May 2002 Average
tangent point location (102.04 oW, 63.40 oS )
8
Impact of COSMIC (36hr Forecast Verification
Against Sondes)
U Bias
T Bias
Experiments Conventional Obs Only Conventional
COSMIC Conv. COSMIC Tuned BE
U RMSE
T RMSE

• Verified in the domain south of 60S

9
Sensitivity of Model Forecast to Model Top
Configuration (36hr Forecast Verification Against
Sondes)
U Bias
U RMSE
Experiments Conv. COSMIC Conv. COSMIC
below 250mb Conv. COSMIC 10mb top
T Bias
T RMSE
10
Impact of COSMIC with 10mb Model Top (36hr
Forecast Verification Against Sondes)

• Assimilation of COSMIC data
• Reduces the bias of T forecasts in the
  lower-middle troposphere and stratosphere
• Decreases the RMSE of T forecasts below 70mb

Conventional Obs Only Conventional COSMIC
11

• AMSU-A (Advanced Microwave Sounding Unit-A)

• Cross-track, line-scanned
• 15 discrete frequency channels
• Profiles from 3mb (45km) pressure height to the
  Earths surface
• 3.3 degrees antenna beamwidth
• 30 continuous scene resolution cells are sampled
  every 9 seconds
• 50km resolution at nadir
• 2343 km swath width from 833km
• nominal orbital altitude

(Courtesy from http//amsu.ssec.wisc.edu/explanati
on.html)
12
AMSU-A Observations

• Channel 4-9
• Oceanic data only

Statistical Errors for NOAA-15 AMSU-A
Scan angle index
13
Before and After Bias Correction
14
Impact of AMSU-A Radiance (36hr Forecast verified
against Sonde)
Experiments Conventional COSMIC Conv.
COSMIC AMSU-A
15
Summary and Conclusions

• AMPS testing of WRF-Var focusing on impact of
  satellite data.
• Majority of AMPS DA effort is in observation
  QC/bias-correction, testing and tuning.
• Assimilation of COSMIC data can improve
  temperature forecasts in both the troposphere and
  stratosphere. Its benefits are also prominent for
  fields (e.g., wind) indirectly related to T, P,
  and q through model physics.
• Assimilation of COSMIC data is sensitive to the
  forecast model top configuration.
• Direct assimilation of radiance data shows
  significant potential to improve the forecasts in
  the Antarctic areas, especially of the wind field.

16
Future Work

• Model top noise in AMPS/WRF forecasts
• Top boundary condition not optimal
• Lack of ozone in WRF
• Test/tune AMSU, AIRS, SSM/IS radiances in AMPS
• Thinning
• Bias correction VarBC
• Inland data
• Update AMPS testbed
• WRF 3.0 polar
• More recent testing period with updated
  observation network
• Increase horizontal resolution (20km).
• Test advanced DA techniques (4D-Var and EnKF) in
  AMPS.

17
Three- and five-day MSLP forecast scores during
1989 and 2006
Northern Hemisphere
Southern Hemisphere

5-Day
3-Day
The lower the curve, the more accurate the
forecast.
(From ECMWF 2006 Annual Report)
18
Observations available and assimilated
(From ECMWF 2006 Annual Report)