Weather Centre Report: ECMWF

1
Weather Centre Report ECMWF

• Sean Healy
• GRAS SAF

2
Outline

• Status at ECMWF Highlight Introduction of
  COSMIC into ERA-Interim.
• Some recent GRAS versus COSMIC forecast impact
  experiments.
• Forecast sensitivity to the dry refractivity
  coefficient.
• Summary.

3
Status at ECMWF

• Operationally assimilate bending angle profiles
  from COSMIC and GRAS measurements, using a 1D
  bending angle operator.
• Since March 10 2009, we have assimilated bending
  angles up to 50 km and increased the weight given
  to the measurements in the upper stratosphere.
  (You may see improvements in your statistics
  against ECMWF near 40 km.)
• ECMWF believes that GPS radio occultation
  measurements are now an important component of
  the global observing system. Key characteristics
• Assimilation without bias correction
• Superior vertical resolution
• COSMIC and GRAS are now being assimilated in
  ERA-Interim.

4
Introduction of COSMIC measurements into
ERA-Interim (Dec 12, 2006)
5
Comparison of GRAS and COSMIC-4 bending angle
departures

• Both COSMIC and GRAS are now being assimilated
  operationally. Noise characteristics of GRAS
  measurements are slightly better.

(Jan 2009 Global, from operations )
6
Forecast impact

• We have compared the impact of GRAS and COSMIC,
  for experiments covering June-July, 2008.
• GRAS measurements are only assimilated above 8
  km, COSMIC used to the surface.
• 1800 COSMIC measurements, compared with 650
  GRAS.
• In general, COSMIC appears to do a better job
  constraining the stratospheric biases.

7
Verification against radiosondes
Standard deviation of the errors
Mean errors
More than one GPSRO instrument is needed to
constrain stratospheric biases.
8
Uncertainty in the k1 refractivity
coefficient(Lidia has also been looking at this)

• The refractivity equation can be written as

At ECMWF we use the Smith-Weintraub (1953) value
for k1
This value has been supported by more recent
work (77.600.05, Bevis et al, 1994).
9
Rüegers (2002) revised values

• Rüeger has proposed a new set of coefficients

The k1 has increased by 0.115 . I have performed
impact experiments comparing the Smith-Weintraub
and Rüeger expressions. The change k1 value does
have an impact, cooling the mean state in the
troposphere.
(See http//www.fig.net/pub/fig_2002/js28_rueger.p
df)
10
Difference in Mean Temperature analysis state
(Rüeger minus Smith-Weintraub, January 2009)
11
Fit to radiosonde height measurements (SH)Bias
worse with Rüeger!
Red SmithWeintraub Back Rüeger
12

• I have tried to find the error in the Rüeger
  work, but it seems to be solid. In fact, it has
  highlighted problems in the Bevis paper.
• Problem 1 The k1 is usually derived from a
  measurement of N, made at 0 degrees Celsuis and
  P1013.25 hPa. So you would expect
• But in fact the data quoted Bevis et al (1994)
  uses
• Problem 2 The Bevis data is for dry,
  carbon-dioxide free air!
• Correcting these two errors increases the Bevis
  k1 from 77.60 to 77.66.

13
Uncertainty in the refractivity coefficients

• There is more uncertainty in the refractivity
  coefficients than I had appreciated.
• Most of the experimental data is from the 1950s
  and 1960s. Given, the NWP, reanalysis and
  climate applications, I think we need to
  encourage new, state-of-the-art experiments to
  determine these values.
• Non-ideal gas effects need to be considered too.
  ECMWF is currently investigating Josep Aparicios
  idea of including compressibility in the GPSRO
  observation operators.

14
Summary

• ECMWF assimilates COSMIC and GRAS measurements up
  to 50 km. The measurements are a very important
  component of the global observing system, and we
  support the COSMIC follow-on.
• We obtained very good improvements in the
  stratospheric biases of ERA-Interim when COSMIC
  was assimilated.
• Assimilation experiments suggest that one
  instrument (GRAS) is not sufficient to constrain
  the stratospheric temperature biases.
• Uncertainty in the refractivity coefficients is
  larger than expected. More work required here
  new lab experiments?