NCAR WRF/ARW REAL TIME HURRICANE FORECASTS 2005 The WRF ARW model was run at 4km resolution with no convective parameterization for five hurricanes.
Download ReportTranscript NCAR WRF/ARW REAL TIME HURRICANE FORECASTS 2005 The WRF ARW model was run at 4km resolution with no convective parameterization for five hurricanes.
NCAR WRF/ARW REAL TIME HURRICANE FORECASTS 2005 The WRF ARW model was run at 4km resolution with no convective parameterization for five hurricanes during the 2005 season. The storms were selected on the basis that they were threatening the United States and all were selected before their ultimate intensity was known. The storms were: Katrina, Maria, Ophelia, Rita and Wilma. The following statistics were prepared by Mark DeMaria and the intercomparisons are homogeneous, in that the same forecasts are used in all cases. I emphasize that this is a limited data set of only five cyclones. A much more comprehensive set is required for definitive results, however these are most encouraging. ESSL Background • The forecast numbers were: Period Number • • • 0 12 24 36 48 72 96 35 35 35 31 28 19 2 The periods out to 72 hours contained sufficient numbers of forecasts to provide reasonable statistics. The 96 hour statistics need to be treated with caution. The model was initialized mostly with GDFL, though there were a few occasions where the GFS was used as the GFDL was not available. Note that the GFDL uses a bogus cyclone and we consider that this had a marked effect on the first 24-36 hour forecasts with WRF-ARW. The horizontal boundary conditions were from the GFS. Note that this is the straight research model, with no attempt having been made at optimization or tuning of physical parameters ESSL Intensity Predictions 45 SHF5 40 OFCL 35 Error (kt) 30 DSHP GFDL FSSE 25 NWRF 20 NCHG 15 10 5 0 0 12 24 36 48 60 72 84 96 Forecast Tim e • • WRF is the black column. The other forecast technique names are indicated in the following slide. Early in the forecast period, WRF is in the pack, a result that we consider would be very much improved by use of a proper initialization. At longer time periods WRF is outstanding. ESSL Intensity Acronyms • • • • • • • SHF5 is a statistical forecast scheme OFCL is the official forecast DSHP is a statistical forecast scheme GFDL is the GFDL model run at 9km resolution FSSE is the Florida State Super Ensemble NWRF is the NCAR WRF-ARW model NCHG is straight persistence ESSL Track Predictions 600 500 CLP5 OFCL NWRF AVNO Error (nm) 400 GFDL NGPS 300 UKM FSSE 200 100 0 0 12 24 36 48 60 72 84 96 Forecast Tim e • • WRF is the black column. The other forecast technique names are indicated in the following slide. WRF-ARW has excellent forecast accuracy at all time periods, being generally very similar to the official forecasts. ESSL Track Acronyms • CLPR is a statistical climatological and persistence forecast scheme • OFCL is the official forecast • NWRF is the NCAR WRF-ARW model • AVNO is the global NCEP Aviation Model • GFDL is the GFDL model run at 9km resolution • NGPS is the Navy Global Prediction System • UKM is the UK Met office Global Model • FSSE is the Florida State Super Ensemble ESSL