ENSEMBLE FORECASTING OF TROPICAL CYCLONES
Principal Investigator:
Sim D. Aberson
Collaborating
scientist(s):
Morris Bender (NOAA/GFDL)
Robert Tuleya (NOAA/GFDL)
Objective:
Evaluation of methods to improve operational forecasts of tropical cyclone
motion and intensity.
Rationale:
Researchers and forecasters are increasingly turning toward methods of
ensemble forecasting in order to work toward improving operational
forecasters. Uncertainty in the initial conditions grows with time
until deterministic forecasts show no skill over forecasts from simply
climatology and/or persistence models. However, by optimally perturbing
model initial conditions using methods such as the breeding of growing
modes used operationally at the National Centers for Environmental
Prediction (NCEP), a set of forecasts which exemplifies the uncertainty
in the model forecast and the range of possibilities of the forecast is
presented. This allows for highly useful probabilistic forecasts
instead of single deterministic ones. Diagnostics, such as the ensemble
mean, ensemble spread, and bias information are clearly presented in
this framework.
Method:
Ensemble forecasting of hurricane tracks and intensity using the
Geophysical Fluid Dynamics Laboratory (GFDL) model will be used during
the 1997 hurricane season to further quantify the value of using the
bred growing modes from the NCEP ensemble forecasting program to assess
the variability of forecasts.
Accomplishment:
A set of almost 100 eleven-member ensembles have been
run for cases during the 1996 and 1997 hurricane seasons in both the Atlantic
and East Pacific basins. The control (GFCT) has been shown to provide
forecasts better than those provided by perturbations (Fig. 1), and that
the ensemble mean (GFMN) provides better forecasts than even the GFCT (Fig 2.).
Rank distributions show that the forecasts tend to have a westward bias,
and that about half of the forecasts do not fall within the envelope of
possibilities presented by the ensemble forecasts (Fig 3.). Finally, the
amount of ensemble spread seems to place an upper limit on the actual
error, with those forecasts with large spread having the largest errors
(Fig 4.). Results from the intensity study are similar, except that the
bias is toward weaker forecasts. Two papers will be presented at the AMS
Annual Meeting in Phoenix, AZ, in January, 1998, at the Numerical Weather
Prediction Conference and at the Special Session on Tropical Cyclone
Intensity Change.
Key references:
Toth, Z., and E. Kalnay, 1993: Ensemble forecasting at NMC: The
generation of perturbations. Bull. Amer. Meteor. Soc., 74,
2317-2330.
Bender, M. A., R. J. Ross, R. E. Tuleya, and Y. Kurihara: 1003:
Improvements in tropical cyclone track and intensity forecasts using the
GFDL initialization system. Mon. Wea. Rev., 121,
2046-2061.
Lord, S. J., 1993: Recent developments in troical cyclone track
forecasting with the NMC global analysis and forecasting system.
Preprints of the 20th Conference on Hurricanes and Tropical
Meteorology, San Antonio, Amer. Meteor. Soc., 290-291.
Sim Aberson administrative information
Project page scientist index
Project page subject index
Main intranet home page
Last modified: 6/26/97