Potential Impact of COSMIC GPS Radio

1
Potential Impact of COSMIC GPS Radio Occultation
Data on Regional Weather Analysis and Prediction
over the Antarctic Ying-Hwa Kuo and Tae-Kwon
Wee COSMIC Project University Corporation for
Atmospheric Research and David H. Bromwich Byrd
Polar Research Center The Ohio State University
2
GPS Meteorology

• UCAR established a GPS/MET program in 1993, with
  a goal to demonstrate the radio occultation
  sounding technique for Earths atmosphere.
• Results from GPS/MET experiment showed that GPS
  radio occultation soundings are of high accuracy
  and can have significant impact on weather
  prediction, climate, and ionospheric research.
• A single satellite does not produce sufficient
  data for NWP or climate analysis.

3
What is COSMIC?

• Constellation Observing System for Meteorology
  Ionosphere and Climate
  COSMIC
• 6 Satellites to be launched in 2005
• Three instruments
• GPS receiver, TIP, Tri-band beacon
• Weather Space Weather data
• Global observations of
• Pressure, Temperature, Humidity
• Refractivity
• Ionospheric Electron Density
• Ionospheric Scintillation
• Demonstrate quasi-operational GPS limb sounding
  with global coverage in near-real time
• Climate Monitoring

4
Topics Covered

• Introduction to Antarctica
• Motivation for RIME
• Climate Interactions Emphasizing the Ross Sea
  Sector
• Approach
• Process-based Studies
• Modeling Research
• RIME Activities and Timelines
• Proposed HIAPER Aircraft Program
• Conclusions

5
Northern Hemisphere
Southern Hemisphere
Palmer
Introduction to Antarctica Location and Size
South Pole
McMurdo
6
Traditional observing network
7
COSMIC soundings over Antarctica
COSMIC soundings
Current radiosonde stations
8
What is the potential impact of COSMIC data on
Antarctic forecasting?
9
OSSE

• COSMIC data will not be available until 2005.
  Therefore, we cannot do a real data impact
  study at this time.
• We can perform a series of observing systems
  simulation experiments (OSSE) to assess the
  potential impact of COSMIC, and to evaluate
  different strategies for the assimilation of
  COSMIC data.
• OSSE is a valuable tool to evaluate
• The potential impact of an upcoming observing
  system
• The relative importance of different observing
  systems
• Deployment strategies for observing systems (or
  network) for field experiment (e.g., RIME).

10
Key elements of OSSE

• Nature Run
• A high-resolution (in time and space) experiment
  which is assumed to be the truth.
• Forward observational operators are used to
  simulate possible observations from an observing
  system (e.g., COSMIC) using the results of the
  Nature Run.
• Simulated observations are assimilated into a
  lower-resolution model.
• The results of the data assimilation/forecast
  system are verified against the results of the
  Nature Run (the truth).
• It is important that the Nature Run produces
  realistic simulation of synoptic and mesoscale
  weather systems.
• 4DVAR Run and subsequent Forward Forecast Run
• Use lower grid resolution and less sophisticate
  to mimic what is possible in an operational
  setting.

11

• Perfect Initial Condition Run
• Best possible initial condition for a forecast
  model. This represents the upper-bound for the
  performance of a forecast model.
• No 4DVAR
• No data assimilation is performed. The model is
  initialized with typical operational analysis.
  This represents the lower-bound for the
  performance of a forecast model.
• Data assimilation experiments
• Assimilate data COSMIC using realistic orbit
  parameters.
• Simulated COSMIC radio occultation soundings are
  distributed irregularly in time and in space.
• Realistic measurement errors are added to the
  simulated radio occultation soundings
• Simulated data are assimilated into MM5 with
  four-dimensional variational data assimilation
  (4DVAR).

12
Experiment domain
13
OSSE design
14
OSSE design
15
Improvements to MM5 4DVAR

• Digital Filter
• High-frequency gravity waves can be excited by
  initial model imbalance and/or data insertion,
  measurement errors, terrain adjustment, etc.
• Gravity waves are permissible solutions of a
  nonlinear, nonhydrostatic, compressible model.
• 4DVAR can be fitting the optimal solution to
  gravity waves.
• It is desirable to filter high-frequency gravity
  waves during the 4DVAR minimization process. This
  forces the 4DVAR to fit its optimal solution to
  the meteorologically significant slow-mode
  component of the model.
• Model error correction
• Most 4DVAR system makes use of perfect model
  assumption in its search for optimal solution.
  This is known as the strong constraint.
• It is desirable to relax this requirement by
  including a correction term that accounts for
  model errors.

16
Formulation of digital filter constraint

• Total cost function

• Penalty cost function

• Initialized model state

• Digital filter

The filtering can be seen as a sort of time
average of model trajectory within the
assimilation window. This can be more effective
with the use of a well designed digital filter
which has optimal response
17
Weak constrain of model error

• To relax the perfect model assumption in 4DVAR
  which has clear deficiencies to be applied in
  Antarctic.
• The full implementation of model error weak
  constraint requires significant extra-cost of
  computational resources.
• Instead a computationally efficient method which
  corrects the model systematic error is used
  (Derber 1989, Zupanski 1993).

nonlinear model operator
model state vector
a predefined time dependent parameter
model error which is a spatially dependent
18
Simultaneous application of weak constraints

• Strong constraint (usually known as the 4DVAR)
  and the weakly constrained 4DVAR with digital
  filter only modify the initial condition
• The weakly constrained 4DVAR with the model
  error does not modify initial condition but only
  updates the systematic bias
• To receive the full benefits of two weak
  constraints, a simultaneous correction of both
  initial condition and model error is implemented.
  

19
COSMIC data distribution at 6-h intervals
20
Traces of P at the lowest level
21
(No Transcript)
22
Scale dependency of anal/fcst error
23
Analysis and forecast error
24
Comparison Perfect and No 4DVAR
PERFECT
No 4DVAR

• Precipitation (rainsnow) isosurface
• Sea level pressure
• Wind at the lowest model level

25
PERFECT
4DVAR (1 CYCLE )

• Precipitation (rainsnow) isosurface
• Sea level pressure
• Wind at the lowest model level

26
4DVAR (4 CYCLES )
No 4DVAR

• Precipitation (rainsnow) isosurface
• Sea level pressure
• Wind at the lowest model level

27
PERFECT
4DVAR (4 CYCLES )

• Precipitation (rainsnow) isosurface
• Sea level pressure
• Wind at the lowest model level

28
4DVAR (4 CYCLES )
4DVAR (1 CYCLE )

• Precipitation (rainsnow) isosurface
• Sea level pressure
• Wind at the lowest model level

29
Analysis and forecast increments (4DVAR No
4DVAR)

• Temperature -3K (blue) and 3K (Purple)
  isosurface
• Temperature at the lowest model level
  shade
• Winds at the lowest model level and 5km level

30
(No Transcript)
31
Summary

• COSMIC will be launched in 2005. The satellites
  are expected to operate for five years. Data will
  be freely available to all countries.
• COSMIC will provide much needed data over the
  Antarctic and Southern Ocean.
• Implementation of digital filter and model error
  correction terms significantly improves the
  performance of MM5 4DVAR.
• The assimilation of COSMIC GPS radio occultation
  data has the potential to significantly improve
  the accuracy of Antarctic weather analysis and
  prediction.
• The continuous 4DVAR cycles allows effective use
  of COSMIC data and has a much better performance
  than a single cycle (or cold start) data
  assimilation.