High-resolution global ensemble hurricane forecasts using an experimental ensemble Kalman filter based analysis and prediction system

As part of the NOAA Hurricane Forecast Improvement Forecast Project (HFIP), high-resolution (30 km) global ensemble forecasts were conducted in near-real time during the 2009 hurricane season. These ensemble forecasts were initialized with an experimental 60 member ensemble Kalman Filter (EnKF) data assimilation system. The data assimilation system was cycled every 6-hours from mid-July to the end of September 2009, using all of the observations used in the operational global data assimilation system at NCEP, plus the warning-center (NHC/JTWC) estimated tropical cyclone central pressure.

The operational NCEP global forecast system (GFS) model, run at T382 (~40 km) resolution, was used in the data assimilation cycle. 20 ensemble members were run out to 7 days once per day, using both the operational T382 GFS model, and a 30 km version of the FIM model (http://fim.noaa.gov) under development at ESRL. Additional deterministic forecasts were run from the ensemble mean EnKF analysis using the FIM model at 15 and 10 km resolution to assess the impact of resolution on hurricane track and intensity forecasts. The skill of ensemble-mean hurricane track forecasts from this system were significantly higher than those derived from the operational global ensembles run at NCEP, the U.K. Met Office and Environment Canada, but not quite as skillful as those produced operationally at the ECMWF. The relative roles of increased model resolution, data assimilation technique and assimilation of estimated TC central pressure in producing these forecast improvements will be discussed.