Diagnosing Atmospheric Conditions Associated with Large Short-Term Intensity Forecast Errors in the Hurricane Weather Research and Forecasting (HWRF) Model

Wednesday, 20 April 2016: 9:00 AM
Ponce de Leon A (The Condado Hilton Plaza)
Daniel J. Halperin, SUNY, Albany, NY; and R. D. Torn

Understanding and forecasting tropical cyclone (TC) intensity change continues to be a paramount challenge in the research and operational communities alike. In fact, the National Hurricane Center and Joint Typhoon Warning Center list guidance on TC intensity change as their highest priority operational forecast improvement need. Numerous statistical, dynamical, and statistical-dynamical guidance products are available, and each contains inherent biases. This talk seeks to identify biases in the Hurricane Weather Research and Forecasting (HWRF) model that lead to large 24 h TC intensity errors. The results can provide important guidance to operational forecasters and inform model developers where deficiencies exist.

This study examines numerous 2015-version HWRF 24 h TC intensity forecasts from 2011-2014 over the North Atlantic and eastern North Pacific basins. Rapid intensification events, TCs near land, and recurving TCs are excluded. Forecasts with large intensity error are defined as errors in the 90th percentile of the sample distribution (mean of 21 kt). These are compared to analog forecasts with similar intensity and shear evolutions, yet small intensity errors. Results indicate that the large intensity error forecasts are associated with more compact TCs (i.e., stronger MSLP gradient, weaker wind speeds away from the TC center), less moisture at and below 850 hPa, but greater moisture above 850 hPa. Results also indicate that the large intensity error forecasts are associated with less convection upshear and downshear right of the TC. Finally, the potential role of the model initialization on the 24 h intensity errors will be discussed.

- Indicates paper has been withdrawn from meeting
- Submission entered in competition