3D-Var assimilation of CHAMP measurements at the Met Office Sean Healy, Adrian Jupp and Christian Marquardt

1
3D-Var assimilation of CHAMP
measurements at the Met Office Sean Healy,
Adrian Jupp andChristian Marquardt

2
Acknowledgements GFZ Potsdam for providing CHAMP
measurements. The Met Office Satellite
Applications Section.
3

• Outline
• Theoretical 1D-Var information content
• Forward model description
• Implementation in 3D-Var
• Preliminary impact trial results
• Summary

4
RO - IASI information theoretical 1D-Var
information content comparison (Collard and
Healy, 2003)
Calculation with 1000 IASI channels. RO
temperature information content maximum in the
300-50hPa region.
RO humidity information near the surface probably
over-estimated. We are assuming 1
refractivity error, but recent (Kuo et al, 2003)
work suggests 3 is more reasonable.
5
Forward model used in trial
Our refractivity forward model has been written
to be consistent with the New Dynamics model,
which has a staggered height grid. The
refractivity forward model uses pressure and
specific humidity (or relative humidity) on
model height levels to simulate the observed
refractivity, N, values at the observation
heights.
6
Forward Model
Need to be able to calculate refractivity at
arbitrary geopotential heights. We have pressure
information on a levels and humidity information
on b levels.
Calculate the temperature on the b level using
the hydrostatic equation. Interpolate the
(Exner) pressure to the b level Calculate
refractivity on the b level. Interpolate the
refractivity to arbitrary observation height.
ln(refractivity) varies linearly with height.
b level
heights
a level
7
Implementation
(1)
(2)
(3)
Produce refractivity as a function of
height (Performed by GFZ)
Perform 1D-Var retrieval for QC.
Assimilate refractivity profiles.
8
Observation Processing System (OPS)
The 1D-Var is used for quality control. We assign
a probability of gross error (PGE) for each
refractivity value in the profile. PGE based
on the 1D-Var cost function at convergence and
the refractivity residuals. Refractivity
residuals - refractivity values calculated with
the 1D-Var solution minus the observed values.
If the residual is greater than 5 times the
observation error PGE 1.0, the value is not
used in the 3D-Var. Cost at convergence If
2J/m gt 20, the PGE of all profile 1.0 (m
size of observation vector 120).
9
Trial Period
Close to operational set-up, assimilating
ATOVS/sondes... We are assimilating CHAMP
refractivity profiles provided by GFZ, between
May 26, 2001 - June 11, 2001. 16, 24 hour
forecasts. Each profile contains 120
refractivity values (150 max), with a vertical
separation of 200m. Observation errors are
based on Kursinskis estimates, but we have
inflated them to 3 at the surface, falling
linearly to 0.25 at 10km and include vertical
error correlations. However, we do not
assimilate refractivity below 4km because of the
well known biases. Note, we only obtain 40
measurements per assimilation cycle (160 per
day).
10
Main results
We do not see any significant improvement in the
humidity fields. (Probably because of the 4km
lower limit. The amount of humidity information
falls with height.) Most significant
improvements in temperature between 250-50 hPa,
in the southern hemisphere. In the northern
hemisphere we can only see an impact around 50hPa
(22 km altitude).
11
Globally averaged 6 hour forecast temperature
differences against radiosonde at 250hPa
12
Globally averaged 6 hour forecast temperature
differences against radiosonde at 50hPa
13
(No Transcript)
14
(No Transcript)
15
NWP forecast fit to radiosondes at 50hPa (NH)
16
NWP forecast fit to radiosondes at 250hPa (Trop)
17
NWP forecast fit to radiosondes at 50hPa(Trop)
18
NWP forecast fit to radiosondes at 250hPa(SH)
19
NWP forecast fit to radiosonde at 50hPa(SH)
20
Tropical PMSL differences seem to be reduced
(T72,96)
21
Summary
We have incorporated a SAF refractivity forward
model into the Met Office 3D-Var system and
tested it with CHAMP measurements.
Successfully completed our first forecast
impact trial. The trial produced 16, 24 hour
forecasts, which have been validated
against observations. Humidity information
probably limited by the 4km cut-off. We see a
positive impact - particularly in the southern
hemisphere T250, T50 and H250. PSML in tropics.
Encouraging results! A constellation
(COSMIC/ACE) looks an exciting prospect.