The Atmospheric Data Assimilation Component

1
The Atmospheric Data Assimilation Component
NCEP CFSRR 1st Science Advisory Board Meeting
7-8 Nov 2007
2
GSI History

• The GSI system was initially developed as the
  next generation global analysis system
• Wan-Shu Wu, R. James Purser, David Parrish
• Three-Dimensional Variational Analysis with
  spatially Inhomogeneous Covariances. (MWR, 2002)
• Originated from SSI analysis system
• Replace spectral definition of background errors
  with grid point representation
• Allows for anisotropic, non-homogenous structures
• Allows for situation dependent variation in errors

3
Operational GSI applications
4
Global GSI upgrades

• 5/1/2007 - initial implementation
• 5/29/2007
• data upgrade
• Replace GOES 5x5 with 1x1 sensor based radiances
• Assimilate METOP-A HIRS, AMSU-A, MHS radiances
• 11/27/2007
• Data upgrade
• Replace Version 6 SBUV/2 ozone data with Version
  8 data
• Reduce high ozone bias in SH polar regions
• Assimilate high resolution JMA atmospheric motion
  winds
• Slight reduction in 200 hPa vector wind rms
  forecast error
• Code upgrade
• Addition of many new options to be turned on
  Spring 2008

5
Globally assimilated data types

• Conventional data
• Sondes, ship reports, surface stations, aircraft
  data, profilers, etc
• Satellite data
• Winds
• SSM/I and QuikSCAT near surface winds
• Atmospheric wind vectors
• Geostationary and POES (MODIS), IR and water
  vapor
• Brightness temperatures (Tb)
• Operational ATOVS, AQUA, GOES sounder,
• Experimental AMSRE, SSM/IS, IASI,
• New for CFSRR ? SSU

6
Globally assimilated data types

• Satellite data (continued)
• Ozone
• Operational SBUV/2 profile and total ozone
• Experimental OMI and MLS capabilities
• COSMIC GPS radio occulation
• Refractivity (operational) or bending angle
• Precipitation rates
• SSM/I and TMI products

7
Radiance (Tb) Assimilation

• GSI uses Community Radiative Transfer Model
  (CRTM) as its fast radiative transfer model
• CRTM developed/maintained by JCSDA
• Features
• Reflected and emitted radiation from surface
  (emissivity, temperature, polarization, etc.)
• Atmospheric transmittances dependent on moisture,
  temperature, ozone, clouds, aerosols, CO2,
  methane, ...
• Cosmic background radiation (important for
  microwave)
• View geometry (local zenith angle, view angle
  (polarization))
• Instrument characteristics (spectral response
  functions, etc.)
• Scattering from clouds, precipitation and
  aerosols

8
Tb Quality Control Issues

• Instrument problems
• Example Increasing noise in AQUA ASMU-A channel
  4
• Inability to properly simulate observations
• Example GSI/CRTM set up to simulate clear sky
  Tb
• IR and Microwave radiances
• IR radiances cannot see through clouds cloud
  heights difficult to determine
• Microwave impacted by thicker clouds and
  precipitation
• Less impacted by thin clouds (bias corrected)
• Surface emissivity and temperature not well known
  for land/snow/ice
• Complicates cloud and precipitation detection

9
Bias Correction

• Currently bias correct
• Radiosonde data (radiation correction)
• Brightness temperatures
• Biases can be much larger than signal ? crucial
  to bias correct the data
• NCEP uses a 2 step process for Tb
• Scan angle correction based on position
• Air Mass correction based on predictors

10
New GSI options (tested/ready)

• CFSRR will exercise several new GSI options
  pertaining to
• Time component
• FOTO (First-Order Time-extrapolation to
  Observations)
• QC
• Variational QC and tighter gross checks
• Tighter QC for COSMIC GPSRO data
• Background error
• Flow dependent variation in background error
  variances
• Change land and snow/ice skin temperature
  background error variances

11
FOTO First-Order Time-extrapolation to
Observations

• Many observation types are available throughout 6
  hour assimilation window
• 3D-VAR does not account for time aspect
• FOTO is a step in this direction
• Generalize operators in minimization to use time
  tendencies of state variables
• Improves fit to observations
• Some slowing of convergence
• compensated by adding additional iterations

Miodrag Rancic, John Derber, Dave Parrish, Daryl
Kleist
12
3D-VAR
Analysis
Obs - Background
13
FOTO
Analysis
Obs - Background
14
Variational QC

• Most conventional data quality control is
  currently performed outside GSI
• Optimal interpolation quality control (OIQC)
• Based on OI analysis along with very complicated
  decision making structure
• Variational QC (VarQC) pulls decision making
  process into GSI
• NCEP development based on Andersson and Järvinen
  (QJRMS,1999)
• Iteratively adjust influence of observations on
  analysis as part of the variational solution ?
  consistency

Xiujuan Su
15
Variational QC implementation

• Only applied to conventional data
• Slowly turned on in first outer loop to prevent
  shocks to the system
• Some slowing of convergence
• compensated by adding additional iterations
• In principle, VarQC allows removal of OIQC step
  
• This, however, has not been done (yet).
• When VarQC on, GSI ignores OIQC flags

16
Situation dependent B-1

• One motivation for GSI was to permit flow
  dependent variability in background error
• Background error variances modified based on 9-3
  hr forecast differences in Tv, and Ps
• Variance increased in regions of rapid change
• Variance decreased in calm regions
• Global mean variance preserved

Daryl Kleist, John Derber
17
As is 500 hPa streamfunction (1e6) background
error standard deviation Valid 2007110600
New flow-dependent adjusted background error
standard deviation
18
Land Snow/Ice variance change

• Operational global GSI has a uniform standard
  deviation of 1K for the skin temperature
• Modify GSI code to allow different values over
  ocean, land, and snow/ice
• Increase from 1 to 3K over land and snow/ice
• Results in
• More satellite data being assimilated
• More realistic skin temperature analysis (not
  used)
• Slight improvement in forecast skill

Daryl Kleist
19
CFSRR GSI

• Based on 11/27/2007 GSI with addition of
• SSU processing (requires updated CRTM)
• Possible adjustment to Tb QC for early satellites
• Includes GSI options targeted for Spring 2008
  global implementation
• FOTO
• VarQC
• Situation dependent rescaling of background error
• Tskin variance tweaks

20
Thanks!

• Questions?

21
(No Transcript)
22
Extra slides

• Bias, FOTO, flow dependent B-1, etc

23
Bias Correction (general)

• Simulated - observed differences can show
  significant biases
• Bias can come from
• Biased observations
• Deficiencies in the forward models
• Biases in the background
• Would like to remove bias except when it is due
  to the background

24

• Guess fields
• 500 hPa
• VT 2007110500

25

• 3D-VAR without FOTO
• Latitude-height cross section along 180E
• Shaded U-wind increment (m/s)
• Thick contour Temperature increment (K)

26

• 3D-VAR with FOTO
• Latitude-height cross section along 180E
• Shaded U-wind increment (m/s)
• Thick contour Temperature increment (K)
• Note asymmetry and smaller magnitude increments
  at off times

27
HPC Surface Analysis
a)
L
rescaled
b)

• Surface pressure background
• error standard deviation
• fields
• with flow dependent re-scaling
• without re-scaling
• Valid 2007110600

as is