Global Ensemble Forecast System (GEFS) Performance on Hurricanes

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Sunday, 2 February 2014
Hall C3 (The Georgia World Congress Center )
Ana P. Torres, University of Puerto Rico, Mayaguez, PR

The performance of the Global Ensemble Forecasting System (GEFS) reforecast data (GEFS-R) has been put to the test in an endeavor to examine its versatility and accuracy by forecasting three high-impact tropical cyclones. Initial observations indicate a 3-4 day window of predictability can be observed when predicting several parameters such as mean sea level pressure and precipitable water for the three storms. In terms of precipitation and the probability of exceeding specific threshold amounts, such as 25, 50 and 100 mm of precipitation, the GEFS-R provided about a 2-3 day window of predictability. The GEFS-R had biases in the forecast track of each storm and generally underestimated each storm's intensity. The CFS' resolution may have been too coarse to properly analyze the depth of the hurricanes. The IBTrACS data typically showed deeper cyclones and strong low-level winds about each cyclone. The limitations of the CFS data relative to the IBTrACS data may have contributed to some of the errors observed in the GEFS-R forecasts. This may have limited the skill of the GEFS-R over older forecasts from the operational GFS and GFDL models run in 1999. There was also some indication that initially weak storms tended to re-curve too early relative to observations, although this tendency decreased with the strengthening of the storms.