Transcript Data Impact Experiments at the JCSDA and NCEP/EMC S. Lord (NCEP/EMC) L.P. Riishojgaard (JCSDA) Contributions by: L.

Data Impact Experiments
at the JCSDA and NCEP/EMC
S. Lord (NCEP/EMC)
L.P. Riishojgaard (JCSDA)
Contributions by:
L. Cucurull, J. Jung, L. Bi, D. Kleist, B. Yan, I.
Appel, D. Stokes
1
Overview
• Impact experiments for:
– COSMIC
– QuikSCAT & Windsat
– MODIS winds
– IASI
– SSM/IS
– ASCAT
• Summary and comments
2
COSMIC Data Impact
Tropics 200 hPa RMS Wind Error 48 h
•
•
•
•
•
0.1 m/s
improvement
No COSMIC
With COSMIC
1 Nov 2006 to 30 Nov 2006
Refractivity assimilated variable
~1000 profiles assimilated per day
NH 500 hPa Height Anom. Cor.
SH 500 hPa Height Anom. Cor.
L. Cucurull
3
QuikSCAT Impact on Standard
Verification Scores
• Operational GFS (T382/L64), GSI with FOTO
– Control (operational QuikSCAT)
– Q denial (no QuikSCAT)
– Q new (FY07 improved retrievals)
D. Kleist
• 5 July 2005 25 to October 2005 (~4 months)
NH 1000 hPa Height RMS Error
SH 1000 hPa Height RMS Error
4
Similar results obtained for 2006 cases
Windsat Impact on Standard
Verification Scores
• Operational GFS (T382/L64), GSI with FOTO
–
–
–
–
Control (operational QuikSCAT)
Ws included (with QuikSCAT)
Ws only (no QuikSCAT)
No Ws, no Qs
D. Kleist
• 25 April 2007 to 8 June 2007 (~1.5 months)
NH 1000 hPa Height RMS Error
SH 1000 hPa Height RMS Error
5
QuikSCAT Impact on Tropical
Cyclone Forecasts
2004 Study
Jung and
Zapotocny
Impact of Removing AMSU, HIRS, GOES Wind, Quikscat Surface Wind Data on
Hurricane Track Forecasts in the Atlantic Basin - 2003 (34 cases)
15.0
JCSDA
% Improvement
10.0
5.0
NOAMSU
0.0
NOHIRS
NOGOESW
NOQuikscat
-5.0
-10.0
-15.0
-20.0
12
24
36
48
72
96
120
Forecast Hour
6
QuikSCAT Impact on Tropical
Cyclone Forecasts (cont)
Track Error Atlantic 2005 00&12 UTC
Track Error Atlantic 2005 00&12 UTC
Climatology-persistence
No QuikSCAT
New QuikSCAT
D. Kleist
D. Stokes
65 cases
18 cases
7
Impact of MODIS winds on GFS
500 hPa Anomaly Correlation
NH July
Northern Hemisphere, February 2007
Northern Hemisphere, July 2007
1
1
0.95
0.95
0.9
0.9
Anomaly Correlation
Anomaly Correlation
NH February
0.85
0.8
0.75
0.7
0.65
0.85
0.8
0.75
0.7
0.65
0.6
0.6
0
1
2
3
4
5
6
7
0
1
2
Forecast [days]
Southern Hemisphere, February 2007
4
5
6
7
5
6
7
Southern Hemisphere, July 2007
SH February
SH July
1
1
0.95
0.95
0.9
0.9
Anomaly Correlation
Anomaly Correlation
3
Forecast [days]
0.85
0.8
0.75
0.7
0.65
0.85
0.8
0.75
0.7
0.65
0.6
0.6
0
1
2
3
4
Forecast [day]
5
6
7
0
1
2
3
4
Forecast [day]
I. Appel
8
IASI Impact Tests
• First attempt to use data at
JCSDA/NCEP(EMC)
• Channel selection: EUMETSAT longwave
only
• 30 day spinup for bias correction
• Experiment and control use same initial
bias corrections
• Scores averaged over last 30 days
– 1-31 August 2007
– 16 December 2007 -15 January 2008
9
IASI Impact on Standard
Verification Scores
1-31 August 2007
NH 500 hPa Height Anom. Cor.
SH 500 hPa Height Anom. Cor.
S. Hemisphere 500 hPa AC Z
20S - 80S Waves 1-20
1 Aug - 31 Aug 2007
1
1
0.95
0.95
Anomaly Correlation '
Anomaly Correlation '
N. Hemisphere 500 hPa AC Z
20N - 80N Waves 1-20
1 Aug - 31 Aug 2007
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0
1
2
3
4
5
6
7
0
Forecast [days]
Control
IASI
Control
1
2
3
4
5
6
Forecast [day]
IASI_EUMETSAT
Control
IASI_EUMETSAT
10
J. Jung
7
IASI Impact on Standard
Verification Scores
16 December 2007 - 15 December 2008
NH 500 hPa Height Anom. Cor.
SH 500 hPa Height Anom. Cor.
S. Hemisphere 1000 hPa AC Z
20S - 80S Waves 1-20
16 Dec 2007- 15 Jan 2008
1
1
0.95
0.95
Anomaly Correlation '
Anomaly Correlation '
N. Hemisphere 500 hPa AC Z
20N - 80N Waves 1-20
16 Dec 2007 - 15 Jan 2008
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0
1
2
3
4
5
6
7
Forecast [days]
Control
IASI
Control
0
1
2
3
4
5
6
Forecast [day]
IASI_EUMETSAT
Control
IASI_EUMETSAT
11
J. Jung
7
SSM/IS Impact Summary
 Positive impacts of SSMIS UPP data can be obtained
through improved cloud detection, surface snow
and sea ice emissivity simulations
 A positive impact of SSMIS UPP data is anticipated by
adding water vapor channels (not shown)
 The SSMIS UPP data displays some regionally
dependent biases at several sounding channels
which would reduce their assimilation impact (not
shown)
12
Improved SSM/IS forecast impact
due to science & processing upgrades
Improved cloud detection & QC
Improved snow and sea ice emissivity
Alternate
processing
CTL
EXP
Cloud detection & QC
EXP
CTL
13
B. Yan et al
ASCAT Impact Tests
•
•
•
•
First results at JCSDA/NCEP(EMC)
10 December 2007 – 19 January 2008
Thinned to 100 km
Quality Control:
Li Bi &
J. Jung
– Ocean only (from GDAS land-sea flag)
– Reject observation if O-B > 5m/s (U and V)
S. Hemisphere 1000 hPa AC Z
20S - 80S Waves 1-20
10 Dec 2007 - 19 Jan 2008
H. Hemisphere 1000 hPa AC Z
20N - 80N Waves 1-20
10 Dec 2007 - 19 Jan 2008
SH 1000 hPa Height Anom. Cor.
1
1
0.95
0.9
0.85
0.8
0.75
0.7
0.65
0.6
Anomaly Correlation '
Anomaly Correlation
'
NH 1000 hPa Height Anom. Cor.
0.95
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0
1
2
3
4
Forecast [day]
Control
ASCAT
5
6
7
ASCAT
Control
0
1
2
3
4
Forecast [day]
Control
ASCAT
5
6
14
7
ASCAT Impact Tests
•
•
•
•
First results at JCSDA/NCEP(EMC)
10 July 2007 – 17 August 200
Thinned to 100 km
Quality Control:
Li Bi &
J. Jung
– Ocean only (from GDAS land-sea flag)
– Reject observation if O-B > 5m/s (U and V)
S. Hemisphere 1000 hPa AC Z
20S - 80S Waves 1-20
10 July - 17 Aug 2007
H. Hemisphere 1000 hPa AC Z
20N - 80N Waves 1-20
10 July - 17 Aug 2007
SH 1000 hPa Height Anom. Cor.
1
1
0.95
0.9
0.85
0.8
0.75
0.7
0.65
0.6
Anomaly Correlation '
Anomaly Correlation '
NH 1000 hPa Height Anom. Cor.
0.95
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0
1
2
3
4
Forecast [day]
Control
ASCAT
5
6
7
ASCAT
Control
0
1
2
3
4
Forecast [day]
Control
ASCAT
5
6
15
7
Summary & Comments
• Impacts measures by standard scores for
–
–
–
–
–
–
COSMIC
QuikSCAT & Windsat
MODIS winds
IASI
SSM/IS
ASCAT
• JCSDA and NCEP experience
– Spin up of global system takes ~6 weeks to produce reliable obs. sensitivity
signal
– Coarse resolution results not representative of those at resolutions higher
than ~50 km globally
– Use of downstream models and applications could be useful to determine
impacts
• Hurricanes
• Waves
• Impact depends on improvements to science (e.g. SSM/IS)
– Surface emissivity (MW and IR)
– Bias correction
– Cloud detection
16
Summary & Comments
(cont)
• Impact experiments are resource intensive
– Many take place in the course of the operational
implementation process and are necessary
– Many custom experiments generally beyond our means
• Respond to HQ requests for special programs (e.g. QuikSCAT)
– Could (and should) be done annually with proper
support (e.g. instrument programs)
• Would pay dividends in
– Focusing scientific development
– Providing feedback for future instrument programs
– Keeping entire community engaged in the total investment
• A complementary OSSE program would be a very
useful addition
17
Backup Slides
18
SATELLITE DATA STATUS – May 2008
Jason Altimeter
Implemented into NCEP GODAS
AIRS with All Fields of View
Implemented – 1 May
MODIS Winds
Implemented– 1 May
NOAA-18 AMSU-A
Implemented– 1 May
NOAA-18 MHS
Implemented– 1 May
NOAA-17 SBUV Total Ozone
4 December 2007
NOAA-17 SBUV Ozone Profile
Implemented– ???
SSMI/S Radiances
Preliminary forecast assessment completed
GOES 1x1 sounder radiances
Implemented 29 May 2007
METOP AMSU-A, MHS, HIRS
Implemented 29 May 2007
COSMIC/CHAMP
Implemented (COSMIC – 1 May) CHAMP Data in prep.
MODIS Winds v2.
Test and Development
WINDSAT
Preliminary forecast assessment completed
AMSR/E Radiances
Preliminary forecast assessment completed
AIRS/MODIS Sounding Channels Assim.
Data in Preparation
JMA high resolution winds
Implemented 4 December 2007
GOES Hourly Winds, SW Winds
To be Tested
GOES 11 and 12 Clear Sky Rad. Assim(6.7µm)
To be Tested
MTSAT 1R Wind Assim.
Data in Preparation
AURA OMI
Test and Development
TOPEX,ERS-2 ENVISAT ALTIMETER
Test and Development (Envisat) ERS-2 (dead)
19
TOPEX implemented in NCEP GODAS
FY – 2C
Data in Preparation
Impact of Removing AMSU, HIRS, GOES Wind, Quikscat Surface Wind Data on
Hurricane Track Forecasts in the Atlantic Basin - 2003 (34 cases)
Jung and
Zapotocny
15.0
% Improvement
10.0
5.0
NOAMSU
0.0
JCSDA
Funded by
NPOESS IPO
NOHIRS
NOGOESW
NOQuikscat
-5.0
-10.0
-15.0
-20.0
12
24
36
48
72
96
120
Forecast Hour
Impact of Removing AMSU, HIRS, GOES Wind, Quikscat Surface Wind Data on
Hurricane Track Forecasts in the East Pacific Basin - 2003 (24 cases)
Satellite data
~ 10-15% impact
30.0
20.0
% Improvement
10.0
0.0
NOAMSU
-10.0
NOHIRS
NOGOESW
NOQuikscat
-20.0
-30.0
-40.0
-50.0
-60.0
12
24
36
Forecast Hour
48
72
20
Operational data assimilation
at NCEP
Lidia Cucurull
Joint Center for Satellite Data
Assimilation
COSMIC IWG Meeting, New Orleans, LA, Jan 21 2008.
21
Achievements at the JCSDA
• The JCSDA developed, tested and incorporated into the new
generation of NCEP’s Global Data Assimilation System the
necessary components to assimilate two different type of
GPS RO observations (refractivity and bending angle).
These components include:
– complex forward models to simulate the observations (refractivity and
bending angles) from analysis variables and associated tangent
linear and adjoint models
– Quality control algorithms & error characterization models
– Data handling and decoding procedures
– Verification and impact evaluation algorithms
• Pre-operational implementation runs showed a positive
impact in model skill when COSMIC profiles were
assimilated on top of the conventional/satellite observations.
• As a result, COSMIC became operationally assimilated at
NCEP on May 1st 2007, along with the implementation of22
the new NCEP’s Global Data Assimilation System
Characteristics of COSMIC
observations
• Limb sounding geometry complementary to
ground and space nadir viewing instruments
– High vertical resolution (0.1 km surface 1km tropopause)
– Lower horizontal resolution (~300 km)
• All weather-minimally affected by aerosols,
clouds or precipitation
• High accuracy (equivalent to < 1 Kelvin from 525 km)
• Equivalent accuracy over ocean than over land
• Independent of radiosonde calibration
• No instrument drift
23
Characteristics of COSMIC
observations
• We assimilate rising and setting occultations, there is no
black-listing of the low-level observations (provided they
pass the quality control checks), and we do not assimilate
observations above 30 km (due to model limitations).
Average
at NCEPpoint
(2007) is considered.
• In an occultation,
theCOSMIC
drift ofcounts/day
the tangent
profiles received at NCEP in time for
operations
obs assimilated (%)
2000
1500
1000
500
0
October
November
December
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
October
November
December
The remaining ~30% received, but not assimilated, is due to:
– Preliminary quality control checks (bad data/format)
– Gross error check (obs very different from the model)
24
– Statistics quality control check (obs too different from the model-obs
Pre-operational
implementation run
• PRYnc (assimilation
of operational obs ),
• PRYc (PRYnc +
COSMIC
refractivity)
rms error
(wind)
• We assimilated around
Anomaly correlation as a function of forecast day (geopotential height)
1,000 COSMIC profiles
per day
25
Pre-operational implementation run (cont)
•Dashed lines: PRYnc
•Solid lines: PRYc (with COSMIC)
•Red: 6-hour forecast
•Black: analysis
26
Summary and future plans
• COSMIC (refractivity) became operationally
assimilated at NCEP on May 1st 2007, along with
the implementation of the new NCEP’s Global
Data Assimilation System (GSI/GFS).
• Several impact studies for selected periods show a
positive impact in model skill when COSMIC profiles
are assimilated on top of the conventional/satellite
observations. [We have recently improved the
assimilation of GPSRO profiles over complex
topography.]
what is next?
• Testing, tuning and assimilation of GSPRO from
CHAMP & GRACE (in pre-operational mode; March
27
2008) and MetOp/A GRAS (when available).
Time series of day-5 scores
Northern Hemisphere, July 2007
1
1
0.95
0.95
Anomaly Correlation
Anomaly Correlation
Northern Hemisphere, February 2007
0.9
0.85
0.8
0.75
0.7
0.65
0.9
0.85
0.8
0.75
0.7
0.65
0.6
1-Feb
6-Feb
11-Feb
16-Feb
21-Feb
0.6
1-Jul
26-Feb
1
1
0.95
0.95
0.9
0.85
0.8
0.75
0.7
0.65
0.6
1-Feb
6-Feb
11-Feb
16-Feb
21-Feb
11-Jul
16-Jul
21-Jul
26-Jul
Southern Hemisphere, July 2007
Anomaly Correlation
Anomaly Correlation
Southern Hemisphere, February 2007
6-Jul
26-Feb
0.9
0.85
0.8
0.75
0.7
0.65
0.6
1-Jul
6-Jul
11-Jul
16-Jul
21-Jul
26-Jul
28
Five-day forecast minus verifying
analysis. February 23, 2007, 00z
Control
Control + MODIS IR
29
Impact of Improved Snow and Sea Ice Emissivity at
SSMIS Channels on F16 UPP SSMIS Data Usage
More data is assimilated
Into GFS !
New
SNOW
EM
New
Ice
EM
Old EM
30
(July 1 ~ July 10, 2007)
A positive impact of SSMIS
UPP data at water vapor
sounding channels is detected
on GFS.
31