Ensemble, Corrected Consensus, and Weighted - PowerPoint PPT Presentation

Loading...

PPT – Ensemble, Corrected Consensus, and Weighted PowerPoint presentation | free to download - id: 250d1-NWE5Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Ensemble, Corrected Consensus, and Weighted

Description:

AVNI Global Forecast System (GFS) run at NCEP. GFDI GFDL model run at NCEP ... EGRI UK Met Office global model. EMXI ECMWF global model ... – PowerPoint PPT presentation

Number of Views:50
Avg rating:3.0/5.0
Slides: 18
Provided by: drjames66
Learn more at: http://www.ofcm.gov
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Ensemble, Corrected Consensus, and Weighted


1
Ensemble, Corrected Consensus, and Weighted
Consensus TC Track Forecasts James S.
Goerss NRL Monterey March 4, 2009
2
or How to Beat Consensus at least
for the Atlantic basin in 2008
3
Track Forecast Guidance
Interpolated Model Guidance AVNI Global
Forecast System (GFS) run at NCEP GFDI GFDL
model run at NCEP HWFI Hurricane WRF run at
NCEP NGPI Navy global model (NOGAPS) run at
FNMOC GFNI GFDL model run at FNMOC EGRI UK
Met Office global model EMXI ECMWF global
model TVCN Consensus of above models (at least
two) AEMI NCEP GFS ensemble mean
4
Corrected Consensus (TVCC)
  • Possible predictors are consensus model spread
    initial and forecast TC intensity initial TC
    location and forecast displacement of TC
    location (latitude and longitude) TC speed of
    motion and number of members available.
  • Regression models to predict TVCN east-west and
    north- south forecast error for all forecast
    lengths were derived using the aforementioned
    pool of predictors for the 2002-2007 seasons
  • The means of the TVCN east-west and north-south
    forecast errors for all forecast lengths were
    also found for the 2002- 2007 seasons to be used
    as bias correctors.
  • Optimal combinations of the statistical and
    bias correctors are used to produce corrected
    consensus (TVCC) forecasts for the Atlantic and
    eastern North Pacific basins.

5
2008 Atlantic Non-Homogeneous TC Forecast Error
(nm)
Best Non-Consensus Guidance





314
251
204
Number of Forecasts
160
131
6
2008 Atlantic Forecast Availability Percentage
7
2008 Atlantic TC Forecast Error Percent
Improvement over GFDI
8
Weighted Consensus Guidance
Based on the forecast performance for EMXI and
AEMI in the Atlantic for 2008, the following
weighted consensus forecast aids were
computed CNW6 Determined just like TVCN
except, that instead of equal weight for
all models, 6 times the weight was
applied to EMXI (roughly the same as taking
the average of TVCN and EMXI) CON3
Determined by taking the average of the TVCN,
EMXI, and AEMI forecasts (at least TVCN).
9
2008 Atlantic Homogeneous TC Forecast Error (nm)
306
249
202
Number of Forecasts
158
131
10
2008 Atlantic TC Forecast Error Percent
Improvement over TVCN
11
2008 Eastern North Pacific Non-Homogeneous TC
Forecast Error (nm)
Best Non-Consensus Guidance





238
178
130
Number of Forecasts
90
57
12
2008 Eastern North Pacific Forecast Availability
Percentage
13
2008 Eastern North Pacific TC Forecast
Error Percent Improvement over Best Model
14
2008 Eastern North Pacific Homogeneous TC
Forecast Error (nm)
236
178
129
Number of Forecasts
89
56
15
2008 Eastern North Pacific TC Forecast
Error Percent Improvement over TVCN
16
Summary
  • For both the Atlantic and eastern North Pacific
    basins in 2008, the corrected consensus (TVCC)
    forecast errors were not much better or worse
    than those for TVCN.
  • Due to the superior performance of EMXI and
    AEMI (at 96 h and 120 h) in the Atlantic, the
    forecast errors for CNW6 were 6-9 percent better
    than those for TVCN at all forecast lengths
    while those for CON3 were 14-16 percent better
    at 96 h and 120 h.
  • Due to the average performance of EMXI and
    inferior performance of AEMI in the eastern
    North Pacific, the forecast errors for CNW6 were
    5-12 percent worse than those for TVCN for
    forecast lengths greater than 72 h while those
    for CON3 were 14-26 percent worse for forecast
    lengths greater than 24 h.
  • The Atlantic results indicate that weighted
    consensus TC track forecast aids are capable of
    producing improved track forecasts if one or more
    of the models perform considerably better than
    the others. However, whether such improved
    model performance is consistent from season to
    season has yet to be determined. The eastern
    North Pacific results suggest that such
    performance and the utility of weighted consensus
    track forecasts are likely to be basin dependent.

17
Questions?
About PowerShow.com