2010090212 PREDICT Forecasts

Ryan D. Torn, David Cook, Steven Cavallo

CURRENT FORECAST PRIOR 0000 UTC FORECAST PRIOR 1200 UTC FORECAST ANALYSIS SYSTEM



Description of the Ensemble System

This ensemble consists of 96 different ensemble members each with a different initial conditions (i.e., no model diversity).   The initial conditions from this ensemble are generated using a cycling ensemble Kalman filter that assimilates land and marine surface data, rawinsondes, dropsondes, commercial aircraft, CIMSS satellite winds, GPS refractivity, and TC latitude, longitude, and minimum SLP each 6 h.   The horizontal resolution of the large domain is 36 km.   Each time NHC declares an INVEST area, the assimilation system generates a 12 km horizontal resolution nest (1000 km x 1000 km) that follows the system as long as NHC continues to follow the system.   Observations are also assimilated on this domain each 6 h when observations are assimilated on the 36 km domain.   Due to time constraints, up to 1 vortex-following nest is used in the 72 h forecast. The nested domain is chosen depending on whether the system is of interest for PREDICT. As a consequence, nested domains are only considered for the 72 h forecast if the system is an INVEST and is south of 25N and east of 80W.   The vortex-following algorithm in WRF follows the minimum in geopotential height, so it is not possible to use vortex-following nests for sub-INVEST pouches.   All nested analyses are being retained, so it is possible to regenerate nested forecasts of all storms after PREDICT has concluded.

The analyses are integrated forward using WRF V3.1.1 and uses WRF 5-class microphysics scheme, Rapid Radiative Transfer Model (RRTM) longwave radiation, Goddard shortwave scheme, Kain-Fritsch cumulus parameterization, Yonsei University (YSU) boundary-layer scheme, and similarity theory land surface model that includes the updated enthalpy and momentum drag formulations described in Davis et al. 2008.   Lateral boundary conditions for each ensemble member are derived from the corresponding GFS forecast, with added perturbations consistent with the GFS background error covariance.

The ensemble is run in two phases. During Phase I, the first 48 members of the ensemble are integrated out to 72 h.   After Phase I is completed, members 49-96 are integrated forward, whereby the initial condition sensitivity calculations are performed. All figures are updated after each forecast hour is completed.