Historic Look at the Tropical Cyclone Track

1
Historic Look at the Tropical Cyclone
Track Forecasting Skill of Operational NWP
Models and their Consensus James S. Goerss
and Charles R. Sampson NRL Monterey James M.
Gross TPC/NHC/NOAA/NWS Miami
2
NWP Model Availability at NHC and JTWC
NHC 1998-1999 NOGAPS, UK Met Office global
model, and the GFDL model 2000-2002 Addition
of NCEP global forecast system, GFDL model run
by FNMOC, and COAMPSTM. JTWC 1992-1995 NOGAPS,
UK Met Office global model, and JMA typhoon
model (JMA typhoon model out to 48h) 1996-1999
Addition of JMA global spectral model, GFDL
model run by FNMOC, and JMA typhoon model out to
72h 2000 Addition of COAMPSTM and MM5 run by
AFWA 2001-2002 Addition of NCEP global forecast
system
3
NOGAPS Upgrades
1990 Assimilation of synthetic TC observations
into NOGAPS T79L18, Arakawa-Schubert cumulus
parameterization 1994 NOGAPS T159L18 1995 TC
motion and proximity incorporated into synthetic
TC observations 1996 Assimilation of
feature-track winds from geostationary
satellites 1998 NOGAPS T159L24 2001 Emanuel
cumulus parameterization 2002 NOGAPS T239L30,
upgrades to Emanuel cumulus parameterization,
NOGAPS run 4 times per day
4
Jim Goerss 12 and 24 hour errors (km), NGPI
(blue), EGRI (red), and JTYI (green)
Western North Pacific Historic
Performance Interpolated Model Forecast Error (nm)
24h
12h
5
Jim Goerss 48 and 72 hour errors (km), NGPI
(blue), EGRI (red), and JTYI (green)
Western North Pacific Historic
Performance Interpolated Model Forecast Error (nm)
72h
48h
6
Jim Goerss 72 hour errors (km), JGSI (green),
GFNI (blue), COWI (red), AFWI (brown), and JAVI
(lavender)
Western North Pacific Historic
Performance Interpolated Model 72-h Forecast
Error (nm)
7
Jim Goerss 72 hour errors (km), JGSI (green),
GFNI (blue), COWI (red), AFWI (brown), and JAVI
(lavender)
Atlantic Historic Performance Interpolated
Model 24-h Forecast Error (nm)
8
Jim Goerss 72 hour errors (km), JGSI (green),
GFNI (blue), COWI (red), AFWI (brown), and JAVI
(lavender)
Atlantic Historic Performance Interpolated
Model 48-h Forecast Error (nm)
9
Jim Goerss 72 hour errors (km), JGSI (green),
GFNI (blue), COWI (red), AFWI (brown), and JAVI
(lavender)
Atlantic Historic Performance Interpolated
Model 72-h Forecast Error (nm)
10
Consensus Forecast Properties

• Consensus TC track forecast error is dependent
  upon two things (Goerss 2000)
• The TC track forecast error of the individual NWP
  models that make up the consensus
• The degree of independence (or effective degrees
  of freedom) of the TC track forecast errors of
  the individual NWP models that make up the
  consensus

11
The interpolated models available to JTWC for the
western North Pacific are NOGAPS (NGPI), the NCEP
global forecast system (AVNI), the UK Met Office
global model (UKMI), the GFDL model run by FNMOC
(GFNI), the Japanese global model (JGSI), the
Japanese typhoon model (JTYI), COAMPS (COWI), and
the MM5 run by AFWA (AFWI). CONU is a consensus
forecast that requires that forecasts from at
least two of the aforementioned models are
available. As CONU has become JTWCs primary
guidance, the official JTWC forecast errors (not
shown) were quite close to those for CONU. The
forecast errors for AFWI and COWI are not shown
but are very similar to those for JTYI.
2002 Western North Pacific 72-h TC Forecast Error
(nm)
12
Jim Goerss 24 hour errors (km), 3 model
consensus (blue), 5 model consensus (red), 7
model consensus (green) and 8 model consensus
(lavender)
Western North Pacific Historic
Performance Model Consensus Forecast Error (nm)
24h
12h
13
Jim Goerss 24 hour errors (km), 3 model
consensus (blue), 5 model consensus (red), 7
model consensus (green) and 8 model consensus
(lavender)
Western North Pacific Historic
Performance Model Consensus Forecast Error (nm)
24h
12h
14
Jim Goerss 72 hour errors (km), 3 model
consensus (blue), 5 model consensus (red), 7
model consensus (green) and 8 model consensus
(lavender)
Western North Pacific Historic
Performance Model Consensus Forecast Error (nm)
72h
48h
15
Jim Goerss 72 hour errors (km), 3 model
consensus (blue), 5 model consensus (red), 7
model consensus (green) and 8 model consensus
(lavender)
Western North Pacific Historic
Performance Model Consensus Forecast Error (nm)
72h
48h
16
Jim Goerss 12 hour skill relative to best track
CLIPER, 3 model consensus (blue), 5 model
consensus (red), 7 model consensus (green) and 8
model consensus (lavender)
Western North Pacific Historic
Performance Model Consensus 12-h Forecast
Skill Percent Improvement Over Best Track CLIPER
(443) (397) (636) (359) (605)
(649) (281) (305) (559) (554) (485)
17
Jim Goerss 24 hour skill relative to best track
CLIPER, 3 model consensus (blue), 5 model
consensus (red), 7 model consensus (green) and 8
model consensus (lavender)
Western North Pacific Historic
Performance Model Consensus 24-h Forecast
Skill Percent Improvement Over Best Track CLIPER
(498) (386) (629) (348) (595)
(639) (261) (281) (509) (508) (451)
18
Jim Goerss 48 hour skill relative to best track
CLIPER, 3 model consensus (blue), 5 model
consensus (red), 7 model consensus (green) and 8
model consensus (lavender)
Western North Pacific Historic
Performance Model Consensus 48-h Forecast
Skill Percent Improvement Over Best Track CLIPER
(429) (314) (544) (267) (522)
(561) (206) (197) (405) (405)
(374)
19
Jim Goerss 72 hour skill relative to best track
CLIPER, 3 model consensus (blue), 5 model
consensus (red), 7 model consensus (green) and 8
model consensus (lavender)
Western North Pacific Historic
Performance Model Consensus 72-h Forecast
Skill Percent Improvement Over Best Track CLIPER
(177) (87) (236) (114) (322)
(385) (115) (96) (253) (283)
(270)
20
Percent Improvement over CLIPER 72-hour Forecast
(3-yr Weighted Mean)
21
Jim Goerss 72 hour errors (km), JGSI (green),
GFNI (blue), COWI (red), AFWI (brown), and JAVI
(lavender)
Atlantic Historic Performance Model Consensus
72-h Forecast Error (nm)
22
The interpolated models available to JTWC for the
western North Pacific are NOGAPS (NGPI), the NCEP
global forecast system (AVNI), the UK Met Office
global model (UKMI), the GFDL model run by FNMOC
(GFNI), the Japanese global model (JGSI), the
Japanese typhoon model (JTYI), COAMPS (COWI), and
the MM5 run by AFWA (AFWI). CONU is a consensus
forecast that requires that forecasts from at
least two of the aforementioned models are
available. As CONU has become JTWCs primary
guidance, the official JTWC forecast errors (not
shown) were quite close to those for CONU. The
forecast errors for AFWI and COWI are not shown
but are very similar to those for JTYI.
Atlantic Historic Perspective 72-h TC Forecast
Error (nm)
23
Availability Percentage for the 2002 Western
North Pacific season
2002 Western North Pacific Consensus Forecast
Availability Percentage
24
Availability Percentage for the 2002 Western
North Pacific season
Atlantic Historic Perspective Consensus
Forecast Availability Percentage
25
Summary
Over the past decade, there has been a steady
improvement in the TC track forecast performance
of the individual NWP models. NWP model 72-h TC
track forecasts today are as skillful as 48-h TC
track forecasts were in the early 1990s. Over
the past decade, there has also been a steady
increase in the number of NWP models capable of
making high-quality TC track forecasts. Today,
JTWC has eight NWP models available to make a
consensus, compared with three models in the
early 1990s. NHC has six NWP models available,
compared with three models in the late 1990s.
Both of these factors have resulted in a
dramatic improvement in consensus TC track
forecasts over the decade. Consensus 72-h TC
track forecasts today are more skillful than
consensus 48-h TC track forecasts were in the
early 1990s. Due to the increase in the number
of NWP models, consensus forecast availability
today is greater than 95 for all forecast
lengths out to 72 h.