The Potential for Improved Tropical Cyclone Forecasts from

The Potential for Improved Tropical Cyclone Forecasts from High-Resolution Global Modeling – Results from the HFIP 2009 Summer Experiment Mike Fiorino Global Systems Division michael. fiorino@noaa. gov Jeff Whitaker and Tom Hamill Physical Sciences Division Earth System Research Laboratory Boulder, CO 3 March 2010 M. Fiorino : : 64 th IHC Savannah GA 20100304

Hi-Res Global Modeling for HFIP 2009 n TACC supercomputer – yes we can! run hires global model (> ECMWF) in near real time in the USA n FIM – Flow-following Icosahedral Model (fim. noaa. gov) (IC = GFS GSI) n n n (fim 8, f 8 em) G 8 – 30 km; 00 Z/12 (fim 9, f 9 em) G 9 – 15 km; 00/12 Z (f 0 em) G 9. 5 – 10 km; 00 Z only (f 8 XX) G 8 – 39 km; 21 -member ensemble GFS (T 382 ~ 45 km) n n 60 -member En. KF – highest-ever global En. KF 21 -member ensemble with En. KF M. Fiorino : : 64 th IHC Savannah GA 20100304

FIM G 5 grid (dx~240 km) – NHC LANT AOR M. Fiorino : : 64 th IHC Savannah GA 20100304

FIM G 5 (~240 km) JTWC/CPHC Local AOR M. Fiorino : : 64 th IHC Savannah GA 20100304

FIM G 6 (~120 km) JTWC/CPHC Local AOR M. Fiorino : : 64 th IHC Savannah GA 20100304

FIM G 7 (~60 km) JTWC/CPHC Local AOR M. Fiorino : : 64 th IHC Savannah GA 20100304

FIM G 8 (~30 km) JTWC/CPHC Local AOR M. Fiorino : : 64 th IHC Savannah GA 20100304

FIM G 9 (~15 km) JTWC/CPHC Local AOR M. Fiorino : : 64 th IHC Savannah GA 20100304

Mean Forecast Error for Hi-Res Models > 72 h : 8 -14 cases @ 48 h : En. KF lower error, no clear benefit from hi res M. Fiorino : : 64 th IHC Savannah GA 20100304

Intensity Bias/Error for Hi-Res Models improved Vmax error with hi-res assimilate central pressure in En. KF IC : lower bias (-30 -15 kt) M. Fiorino : : 64 th IHC Savannah GA 20100304

Ensembles for my : : what the 2 tell me… n ensembles attempt to ‘sample’ the model error space n how well ensembles sample/forecast error depends on the ratio of ‘spread’ to ‘error’; ideal is 1: 1 or spread ~ error n potential for accurate ‘guidance on the guidance’ or ‘forecast of forecast error’ n ‘error’ = track forecast error M. Fiorino : : 64 th IHC Savannah GA 20100304

21 -member GFS(T 382) Ensemble using ESRL En. KF 03 L. 2009 aka BILL 2009081700 “spread” = mean of the distance between each member and the ensemble mean position initial spread ~ 25 nm M. Fiorino : : 64 th IHC Savannah GA 20100304

The Bob Gall Plot – depict intensity in the ensembles… 03 L. 2009 aka BILL 2009081700 colorize track segment based on difference between member intensity and mean intensity – which members are intensifying v weakening M. Fiorino : : 64 th IHC Savannah GA 20100304

Spread v Error haloes… spread ~ error (good) 03 L. 2009 aka BILL 2009081700 add verifying track and the forecast error halo desired relationship – error ~ spread M. Fiorino : : 64 th IHC Savannah GA 20100304

Spread v Error for ECMWF EPS(T 399) 03 L. 2009 aka BILL 2009081700 EPS under dispersive – spread << error M. Fiorino : : 64 th IHC Savannah GA 20100304

Mean Spread v Mean Error M. Fiorino : : 64 th IHC Savannah GA 20100304

Some takeaways… n HFIP summer 2009 test demonstrated that models with res ≥ ECMWF can be run in near real-time on USA computers… n hi-res, however, is not a sufficient condition…for track/intensity skill, i. e. , we can’t compute our way to improved TC forecasting with more CPUs… modeling challenge is still physics as Δx 10 km n hi-res global ensembles showed both good track forecast skill and spread/error ratios ~ 1 potential forecasting the forecast error, but all systems under dispersive M. Fiorino : : 64 th IHC Savannah GA 20100304

resources: FIM: http: //fim. noaa. gov Current TC EPS graphics: http: //ruc. noaa. gov/hfip/tceps Current TC EPS tracks (google map): http: //ruc. noaa. gov/tracks Python S/W to analyze/display ATCF data: http: //sourceforge. net/projects/wxmap 2/ M. Fiorino : : 64 th IHC Savannah GA 20100304