Introduction of temperature observation of radio-sonde in place of geopotential height to the global three dimensional variational data assimilation system in JMA A.Narui Japan Meteorological Agency, Japan

1
Introduction of temperature observation of
radio-sonde in place of geopotential height to
the global three dimensional variational data
assimilation system in JMAA.NaruiJapan
Meteorological Agency, Japan
2
content

• ?. The reason to introduce temperature
  observation of radio-sonde in place of
  geopotential height
• ?. modification to assimilate temperature
• ?. Forecast experiments
• ?. summary

3
?.The reason to introduce temperature observation
of radio-sonde in place of geopotential height

• The global three dimensional variational data
  assimilation system (3D-Var) was implemented in
  JMA operation in September 2001.
• The direct assimilation of ATOVS radiances was
  introduced in May 2003.
• As the next step for our 3D-Var, we have a plan
  to introduce variational quality control (VarQC).
  Because VarQC is natural extension of variational
  method.
• Since it is favorable for VarQC that there is
  no correlation between observation data, we are
  going to assimilate temperature of radio-sonde in
  place of geopotential height, which has strong
  vertical correlation.

4
?.The reason to introduce temperature observation
of radio-sonde in place of geopotential height

• Though the main purpose of this work is to
  prepare for the introduction of VarQC, the use of
  temperature data itself showed some good impacts
  on the forecast score.

5
SYSTEM OF JMA/3D-VAR

• FORECAST MODEL GLOBAL MODEL
• RESOLUTION T213L40 (OUTER MODEL)
• T106L40 (INNER
  MODEL)
• INCREMENTAL METHOD
• ANALYSIS VARAIBLE ??D?T?Ps?ln q
• CONTROL VARIABLE ??Du?(T,Ps)u?ln q
• BACKGROUND ERROR NMC METHOD for 2000
• 
  HOMOGENEOUS
• MINIMIZATION LBFGS

6
?.modification to assimilate temperature

• 1. remove the vertical correlation of
  observational errors for all elements
• 2. use the significant level data in addition to
  the standard level data
• 3. recalculate all observational errors for all
  elements of radio-sonde from the statistics of
  observation departure from the background

7
?. Forecast experiments

• Exp.1 assimilation of geopotential height
• Exp.2 assimilation of temperature
• Forecast ModelGlobal Model T213L40
• 6hourly cycle
• period 1
  period 2
• assimilation 27Jun-30July 2002 27Nov-31Dec
  2002
• initial 12UTC 1-21 July 12UTC
  1-21 Dec
• Fcst range 216 hour 216
  hour

8
Analysis increment of sea level pressure00UTC
3rd July 2002 in N.H.lefttemperature
rightheight
9
Anomaly correlation of 500hPa height 2002
7/17/21 (against initial field)blueheightred
temperature Global NHTropics SH
10
Anomaly correlation of 500hPa height 2002
12/112/21 (against initial field)
blueheightredtemperature Global
NHTropics SH
11
RMSE and Bias of 500hPa height 2002 7/17/21
(against radio-sonde) greenheightredtemperat
ure leftrmse rightbiasNHTropics SH
12
vertical profile of temperature 2002
12/31(after one month anlalysis cyclemean of
Northern Hemisphere) greenheightblacktemperat
ure
13
?. Summary

• 1. We Introduced temperature of radio-sonde in
  place of geopotential height.
• 2. There is some good impact on forecast score.
• 3.Problem Bias against radio-sonde
• 4. VarQC is developed now.

14
Example of VarQC
Formulation (including gross error
probability)Cost function JOQC -l
n(?exp(?JON) )/(? 1)?JOQC ?JON (1-P)
JON normal cost function ? VarQC parameter
for each observationP gross error
probabilityOIQC oprerational QC in
JMAcompare each datum in turn against an
analysis based on surrounding data with OI
15
Example of VarQC rejected with VarQC(gross error
probability is gt75)
Analysed surface pressureupperincrement
leftVarQC rightOIQC Lower
leftdifference rightanalysed field
16
The End