Ensembles
Home Introduction Methodology Testing Output Plots Validation Future Work Papers Comments

 

Track Forecasting of Atlantic Tropical Cyclones Using a Kilo-Member Ensemble

    Welcome to the web page of the Kilo-Member Ensemble. The purpose of this page is to provide information about the design and methodology of the ensemble system, as well as output from past and current storms, and validation.  Click on the links below for more detail in each sub area.

  • What is the Kilo-Member Ensemble?
    • A forecasting tool for tropical cyclone track prediction that samples a wide range of possible future atmospheric states with the goal of producing a more reliable forecast than possible using a single model realization.
    • Additional goals include obtaining reliable storm track envelopes (the range of all possible storm tracks) and estimates of forecast uncertainty (based on ensemble spread statistics).
    • Acknowledgements

 

  • How does it work?
    • Initial conditions and boundary conditions provided by NCEP Global Forecasting System (GFS) ensembles
    • Synthetic vortex inserted at location of operational storm position estimate with storm motion vector
    • Uncertainties relating to evolution of environmental flow, storm steering layer, vertical decomposition, vortex size/strength, and storm motion vector are simulated by ensemble perturbations
    • The MUltigriD BARotropic (MUDBAR) tropical cyclone track model is used to calculate track of each ensemble member
    • The resulting tracks from the 1980 ensembles are then processed and displayed for the forecast user

 

  • Real-time testing
    • System was run after the fact for the 2001 season
    • Ensemble was run in near real-time for 2002 Atlantic and Eastern Pacific storms
    • Ensemble also ran in near real-time for the 2003 Atlantic and Eastern Pacific storms
    • Current system configuration

 

 

  • Future Work
    • Change sequence so that post-processing and plots are done at the end of each forecast case, rather than after all the cases have been run
    • Fix lat/lon problem so that Eastern Hemisphere storms can be run
    • Create Java 'toolkit' to display ensemble data for forecasters
    • More research on post-processing statistics, possibly use error-recycling scheme or fuzzy logic/neural network to come up with optimal weights

 

  • Papers and Presentations about the Kilo-Member Ensemble
    • Conference presentation at the 25th Conference on Hurricanes and Tropical Meteorology in San Diego, 2002
    • Paper about the evaluation of MUDBAR by comparison with the operational LBAR track model, published in Monthly Weather Review
    • Masters Thesis, defended April 14, 2004
    • Conference presentation at the 26th Conference on Hurricanes and Tropical Meteorology in Miami, May 2004