Table 3-5. Upgrades to the NOGAPS Model
Year
| Operational Upgrades to the NOGAPS |
Pre-1991
| -
NOGAPS spectral model resolution increased to T79L18 (~165 km horizontal resolution, 18 vertical levels). With this increase in resolution, it was found that NOGAPS had tropical cyclone track forecast skill (Hogan and Rosmond 1991).
-
Assimilation of synthetic tropical cyclone observations into NOGAPS (Goerss and Jeffries 1994).
|
1994
| -
Model resolution increased to T159L18 (~110 km horizontal resolution).
|
1996
| -
Assimilation of high-density multispectral feature-track winds from geostationary satellites (Goerss et al. 1998).
|
1997
| -
Assimilation of SSM/I precipitable water.
|
1998
| -
Model resolution increased to T159L24 (24 vertical levels).
|
2000
| -
Emanuel convective parameterization scheme replaces relaxed Arakawa-Schubert scheme (Peng et al. 2004).
|
2002
| -
Model resolution increased to T239L30 (~55 km horizontal resolution, 30 vertical levels) and improvement made to Emanuel convective parameterization scheme.
|
2003
| -
NRL Atmospheric Variational Data Assimilation System (NAVDAS), a 3D-VAR data assimilation system, replaced MVOI system (Daley and Barker 2001).
|
2004
| -
Direct assimilation of AMSU-A radiances replaces assimilation of NESDIS ATOVS retrievals.
-
Assimilation of Moderate Resolution Imaging Spectroradiometer (MODIS) polar winds from NASA satellites Aqua and Terra.
-
Assimilation of QuikSCAT and ERS-1 scatterometer winds.
|
2005
| -
Assimilation of synthetic tropical cyclone observations improved.
|
-
The improvements due to implementing the Emanuel convective parameterization were 16 percent at 24 hours, 10 percent at 48 hours, 6 percent at 72 hours, 12 percent at 96hours, and 16 percent at 120 hours. With the exception of 72-hour forecast length, these improvements were all statistically significant at the 95 percent confidence level.
-
While increasing the model resolution to T239L30 improved forecast performance in the extra-tropics (not shown), it resulted in degradations in tropical cyclone track forecasts (not statistically significant) at all forecast lengths except 24 hours.
Similar to the above NOGAPS experiments, NCEP/EMC performed tests of the new GFS after the substantial upgrades in 2005. Table 3-6 displays the results of the retrospective runs, using data from the 2004 Atlantic tropical cyclone season, compared with the results of the GFS version run operationally during the 2004 season. The 2005 version of the GFS had substantially reduced track forecast errors for the sample cases at all forecast verification times.
Table 3-6. Mean 2005 GFS Track Errors (in nautical miles) for a Sample of Cases from 2004
|
00 h
|
12 h
|
24 h
|
36 h
|
48 h
|
72 h
|
96 h
|
120 h
|
Operational GFS
|
12.4
|
35.1
|
52.1
|
72.2
|
88.0
|
140.2
|
204.3
|
275.5
|
New (T382) GFS
|
11.7
|
31.8
|
45.4
|
61/3
|
76.7
|
115.1
|
161.6
|
218.0
|
Reduction of Error With New GFS
|
5.7%
|
9.3%
|
12.9%
|
15.1%
|
12.8%
|
17.9%
|
20.9%
|
20.9%
|
# of Cases
|
61
|
59
|
57
|
55
|
53
|
50
|
43
|
35
|
Share with your friends: |
