Upload
ngohuong
View
220
Download
2
Embed Size (px)
Citation preview
Review of NCEP GFS Forecast Skills in 2011 and Beyond
Fanglin Yang
IMSG - Environmental Modeling Center National Centers for Environmental Prediction
Camp Springs, Maryland
The 46th CMOS Congress and the AMS 21th NWP and 25th WAF conferences, Montréal (Canada), May 29th to June 1st, 2012.
Acknowledgments: many scientists at NCEP and in the community have made contributions to the development and application of the GFS. For this presentation, I would like to thank in particular Joseph Sela, Moorthi Shrinivas, Hualu Pan, Stephen Lord, Bill Lapenta, John Derber, Mark Iredell, Glenn White, Russ Treadon, Mike Ek, Henry Juang, Yu-Tai Hou, Suru Saha, Bob Kistler, Jordan Alpert, Daryl Kleist, Jongil Han, Peter Caplan, Yuejian Zhu, Jun Wang, Helin Wei, Sarah Lu, Hui-Ya Chuang and others I am not be able to include here.
Annual Mean 500-hPa HGT Day-5 Anomaly Correlation
0.55
0.65
0.75
0.85 19
84
1985
19
86
1987
19
88
1989
19
90
1991
19
92
1993
19
94
1995
19
96
1997
19
98
1999
20
00
2001
20
02
2003
20
04
2005
20
06
2007
20
08
2009
20
10
2011
GFS-NH CDAS-NH GFS-SH CDAS-SH
In past 25 years GFS anomaly correlations increased by about 0.3 in both hemispheres (0.12/decade)
Annual Mean 500-hPa HGT Day-5 Anomaly Correlation
-0.1 -0.08 -0.06 -0.04 -0.02
0 0.02 0.04 0.06 0.08
0.1 0.12 0.14 0.16 0.18
0.2 19
84
1985
19
86
1987
19
88
1989
19
90
1991
19
92
1993
19
94
1995
19
96
1997
19
98
1999
20
00
2001
20
02
2003
20
04
2005
20
06
2007
20
08
2009
20
10
2011
NH:GFS-CDAS
SH:GFS-CDAS
Best Year, NH
Annual Mean NH 500hPa HGT Day-5 AC
0.6
0.65
0.7
0.75
0.8
0.85
0.9 19
84
1985
19
86
1987
19
88
1989
19
90
1991
19
92
1993
19
94
1995
19
96
1997
19
98
1999
20
00
2001
20
02
2003
20
04
2005
20
06
2007
20
08
2009
20
10
2011
GFS-NH CDAS-NH ECMWF-NH UKM-NH CMC-NH FNOMC-NH
• GFS falls behind ECMWF all the time. • GFS and UKMO were comparable for most of the time, but GFS trailed
UKMO in recent years. • FNOMC has made significant improvement since 2009.
2011 Annual Mean 500hPa HGT AC
5
• GFS falls behind EC and UK, but is better than CMC, FNO and JMA. • GFS useful forecasts (>0.6) reached 8.0 days in the NH and 7.8 days
in the SH.
NH SH
0.6 – useful forecast
Day at which forecast loses useful skill (AC=0.6) N. Hemisphere 500hPa height calendar year means
Fore
cast
day
8 d
Credit:, Peter Caplan, Yujian Zhu, Fanglin Yang
7
Twenty bins were used to count for the frequency distribution, with the 1st bin centered at 0.025 and the last been centered at 0.975. The width of each bin is 0.05.
Look at the history of extremes in the distribution
–Poor Forecasts (AC < 0.7 )
–Excellent forecasts ( AC > 0.9 )
Reduced poor forecasts
0
10
20
30
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
%
8
Resolution: 1.2/2000:
T126L28 à T170L42 (70km)
2.11/2002: T170L42 à T254L64 (55km)
3.6/2005: T254L64 à T382L64 (38km)
4.7/2010: T382L64 à T574L64 (23km)
Percent of Poor Forecasts (AC <0.7) v.s. Model Changes
Physics and Data Assimilation:
A. 3/1999: AMSU-A &
HIRS-3 data
B. 5/2001: prognostic cloud water, cumulus momentum transport
C. 6/2005: OSU 2-L LSM to 4-L NOHA LSM
D. 5/2007: SSI to GSI; Hybrid sigma-p; New observations
E. 2/2009: flow-dependent error covariance; Variational QC
F. 7/2010: New shallow convection; updated SAS and PBL; positive-definite tracer transport.
G. 5/2011: GFS/GSI upgrade: new satellite data and thermal roughness
A
1
B
2
3, C 4, F
year
NH
1996 2011
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
%
9
Percent of Poor Forecasts (AC <0.7) v.s. Model Changes
A
1
B
2 3, C
4, F
year
SH
D
Physics and Data Assimilation:
A. 3/1999: AMSU-A &
HIRS-3 data
B. 5/2001: prognostic cloud water, cumulus momentum transport
C. 6/2005: OSU 2-L LSM to 4-L NOHA LSM
D. 5/2007: SSI to GSI; Hybrid sigma-p; New observations
E. 2/2009: flow-dependent error covariance; Variational QC
F. 7/2010: New shallow convection; updated SAS and PBL; positive-definite tracer transport.
G. 5/2011: GFS/GSI upgrade: new satellite data and thermal roughness.
E
Resolution: 1.2/2000:
T126L28 à T170L42 (70km)
2.11/2002: T170L42 à T254L64 (55km)
3.6/2005: T254L64 à T382L64 (38km)
4.7/2010: T382L64 à T574L64 (23km)
1996 2011
G
2011: A Year with Record-Breaking Billion-Dollar Weather Disasters
Hurricane Irene: 10 billion August 20-29, Cat 3 56 deaths
11
Hurricane Irene
GFS track forecast was one of the best
GFS
track intensity
GFS intensity forecast was also better than other global NWP models
GFS 60-84hr Rainfall Forecast: Hurricane Irene
24-hr rainfall, valid at 12Z28Aug2011
24-hr rainfall, valid at 12Z29Aug2011
0
50
100
150
200
250
300
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Trac
k Er
ror (
nm)
Atlantic Hurricane Track Errors 0
12
24
36
48
72
Forecast within 3 days has been steadily improving, although the pace is slow. Beyond day 3, forecast still varies from year to year.
Hurricane Track and Intensity Errors , GFS: 2001-2011
0
50
100
150
200
250
300
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Trac
k Er
ror (
nm)
Eastern Pacific Hurricane Track Errors 0
12
24
36
48
72
96
120 fhr
Recent and Upcoming Changes Hybrid-Ensemble Data Assimilation. Implemented on May 22nd, 2012. Semi-Lagrangian dynamics, T1148L64, still in development
16
T574L64 Hybrid Ensemble GFS Parallel
Tropical Wind RMSE, verified against ROBS
Global Temp RMSE, verified against RAOBS
17
Most Recent T1148L64 Semi-Lag GFS Test
Promising, but still has issues. Still testing different package options and tunable parameters.
Parallel run by Fanglin Yang on ESRL Jet
18
Extra slides
Annual Mean SH 500hPa HGT Day-5 AC
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9 19
84
1985
19
86
1987
19
88
1989
19
90
1991
19
92
1993
19
94
1995
19
96
1997
19
98
1999
20
00
2001
20
02
2003
20
04
2005
20
06
2007
20
08
2009
20
10
2011
GFS-SH CDAS-SH ECMWF-SH UKM-SH CMC-SH FNOMC-SH
20
Tropical Wind RMSE, 850-hPa Day-3 Forecast
July2010 T574 GFS Implementation
GFS tied with EC and UK after July 2010 implementation.
21
Tropical Wind RMSE, 200-hPa Day-3 Forecast
• GFS has larger wind RMSE than EC and UK at the tropopause region.
• Improvement in the past 15 years is not significant.
22
Twenty bins were used to count for the frequency distribution, with the 1st bin centered at 0.025 and the last been centered at 0.975. The width of each bin is 0.05.
Look at the history of extremes in the distribution
–Poor Forecasts (AC < 0.7 )
–Excellent forecasts ( AC > 0.9 )
Hurricane Track and Intensity Errors
23 AVNO = GFS EMX = ECMWF 00Z and 12Z cycles
2011 Atlantic 2011 Atlantic
10
20
30
40
50
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Inte
nsity
Err
or (k
ts)
Atlantic Hurricane Intensity Errors 0
12
24
36
48
72
96
120 (fhr)
CONUS Precipitation, 0-72hr Total
25
2011 CONUS Precipitation Skill Scores
26
T574L64 Hybrid Ensemble GFS
The parallel outperformed operational GFS in both NH and SH.
NH SH
27
Percent of Excellent Forecasts (AC >0.9)
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
%
Year
Percent Anomaly Correlations Greater Than 0.9 GFS 00Z Cycle Day-5 500hPa Height
NH SH
1996 2001 2000 1998 1997 1999 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
2011 Hurricane Track and Intensity Forecast Errors
2011 Atlantic Hurricanes
2011 Eastern Pacific Hurricanes
http://www.wikipedia.org
2011 Eastern Pacific Hurricane Track and Intensity Errors
29 AVNO = GFS EMX = ECMWF 00Z and 12Z cycles
2011 Western Pacific Hurricane Track and Intensity Errors
30 AVNO = GFS EMX = ECMWF 00Z and 12Z cycles