|
Tropical Cyclones
|
Work & Research
|
Storm Chasing
|
About me
|
Weather Briefing
|
|
Last updated:
Fri 1 Aug 2003 4:45pm PDT
|
|
|
|
|
|
|
|
AMS 2004 Abstract - Submitted 1 Aug 2003 |
Correlation between Numerical Model Errors and Climatology in Tropical Cyclone Track Forecasting
Bradford S. Barrett
School of Meteorology, University of Oklahoma
Climatology and persistence have been shown to be good predictors of tropical cyclone motion (Neumann 1975). Dynamical models developed in the last decade have also led to significant improvement in track forecast errors. However, the rapid advance in skill of numerical models has resulted in a decline in development and improvement of statistical prediction methods that use environmental and model data (Bessafi et al 2002). This present study examines the relationship between climatology and dynamical model track, speed, and bearing errors. By correlating climatology with model error, the goal of this work is to provide forecasters with a confidence level for each dynamic model prediction. For given climatology values, forecasters will have at their disposal the mean model track, bearing, and speed errors and standard deviations, as well as the correlation between climatology and error. Knowing that a dynamical model typically has large errors for a given climatology would allow forecasters to place less weight on that particular model prediction. This discounting of certain model solutions is especially useful today because, as was recognized more than two decades ago, “without guidelines on model attributes … multiple guidance can be counterproductive” due to conflicting results (Neumann 1981).
The climatology was developed by first collecting the best track data of all Atlantic tropical cyclones from 1970 to 2001. This data was filtered to include only cyclones found within 250 km of a point. Then, motion probabilities were devised based on all cyclones found within this 250 km “radius of influence.” These probabilities were then separated into 30-degree bins and subdivided by distance. Finally, the climatology value (a fraction ranging from 0.0 to 1.0, where 0.0 implies no previous occurrences and 1.0 implies all storms within 250km of the chosen point have the same bearing and traveled the same distance) was calculated for each bin and assigned to each model run.
This work uses model output from operational NOGAPS and COAMPS for the tropical Atlantic 2001 and 2002 seasons. Track, bearing, and speed errors for 12, 24, 36, and 48 hrs are calculated for each model run. Then the climatology value that corresponds to each run’s forecast bearing and distance is assigned. Finally, the climatology values are correlated to model errors and results analyzed.
|
|
|
|
