How to Test Climate Models Using GPS Radio Occultation

1
How to Test Climate Models Using GPS Radio
Occultation

• Stephen S. Leroy, James G. Anderson, John A.
  Dykema
• Harvard University

2
Outline

• Survey of climate model predictions for the
  atmosphere
• GPS radio occultation
• Optimal fingerprinting
• Climate predictability in GPS occultation
• Linear Inverse Modeling
• Conclusions

3
Climate Model Evaluation Project

• IPCC Fourth Assessment Report created CMEP (Susan
  Solomon, Gerald Meehl), distributed from
  Livermore
• Contributions from 22 climate models worldwide
• Multiple scenarios, standard output format. We
  will use SRES A1B (1 CO2/year until doubling).

4
Surface Air Temperature
5
Upper Air Temperature Trends
Pressure (hPa)
6
Radiosonde Observations
Difference of period 1961-1980 from period
1985-1995. Taken from Tett, Jones, Stott et al.,
2002.
7
Geopotential Height Trends
Pressure (hPa)
8
GPS Radio Occultation
Profiles refractive index vs. height 100 meter
vertical resolution 500 km horizontal
resolution 500 soundings per day per LEO (24 GPS
satellites) Directly traceable to NIST definition
of second. COSMIC Constellation Observing
System for Meteorology, Ionosphere and
Climate National Space Program Office (Taiwan)
UCAR COSMIC Project Launch March 2006 6
satellites, 3,000 soundings per day
9
GPS Occultation Dry Pressure

• Refractivity
• Dry Pressure
• Relationship to Geopotential Height

10
Occultation Dry Pressure
11
Optimal Fingerprinting

• Signals are uniquely identified by their
  normalized shapes
• An arbitrary signal will always look a little
  like natural variability, so detection is given
  in terms of confidence levels
• Detection is preferentially weighted toward
  components of the signal(s) where natural
  variability is small compared to the signal
• In absence of data, one can still project how
  long it should take for a signal to be detected.

12
Deriving Optimal Fingerprints

• I. Electrical Engineering Weight the data so as
  to minimize the error associated with the fitted
  coefficients (North and Kim 1995)
• II. Statistical Assemble the Bayesian evidence
  function given a model for the data (Leroy 1997)

fingerprint
13
Our Implementation
Determine detection times and optimal
fingerprints
14
Dry Pressure EOFs
ENSO Southern Annular Mode Northern Annular
Mode Symmetric Jet Migration (lagged response to
ENSO)
15
Open diamonds Natural variability
eigenvalues Filled squares Signals projections
onto EOFs 12 models for signal shape s 4 models
prescribe natural variability N The higher the
filled squares are with respect to the
eigenvalues, the more that mode will contribute
to detection (and increase the SNR).
16
Fingerprints
17
95 Detection Times
Model GFDL CM2.0 (yrs) ECHAM5/MPI-OM (yrs) UKMO-HadCM3 (yrs) MIROC3.2 (medres) (yrs) Tropospheric Expansion (m decade-1)
GFDL-CM2.0 8.67 9.05 8.29 6.63 11.02
GFDL-CM2.1 7.88 8.65 7.57 6.21 12.86
GISS-AOM 10.53 11.54 10.47 8.38 9.67
GISS-EH 10.41 11.74 10.77 8.50 9.12
GISS-ER 10.89 12.70 11.07 9.32 8.79
INM-CM3.0 9.98 11.23 9.79 8.15 10.71
IPSL-CM4 9.29 10.02 8.95 7.36 10.54
MIROC 3.2(medres) 7.09 7.47 6.83 5.39 13.04
ECHAM5/MPI-OM 7.78 8.16 7.45 5.87 12.34
MRI-CGCM2.3.2 9.95 11.70 9.92 8.35 10.68
CCSM3 8.87 9.62 8.68 6.80 11.97
PCM 12.69 12.32 11.95 8.45 7.27
18
ENSO and SJM
19
Linear Inverse Modeling
UKMO HadCM3
20
Conclusions

• GPS radio occultation will determine the
  sensitivity of the climate with 5 accuracy in 7
  to 13 years
• The sensitivity of the climate system is more
  sensitively measured by poleward jet migration
  not associated with NAM or SAM than by
  temperature
• Poleward jet migration does not seem to be a
  lagged response of an ENSO event
• Might be possible already with GPS/MET
  (1995-1997) and CHAMP (2001-), but preliminary
  analysis suggests GPS/MET has insufficient
  coverage
• Other data types will be necessary to constrain
  physical mechanisms responsible for climate
  change, most likely spectrally-resolved shortwave
  and longwave spectra