Dynamic Hurricane Season Prediction Experiment with the NCEP CFS Jae-Kyung E. Schemm January 21, 2009 COLA CTB Seminar Acknowledgements: Lindsey Long Suru Saha Shrinivas Moorthi
Outline Description of the CFS experiment Datasets Used IRI Detection and Tracking Method Analysis of storm activity statistics Evaluation of CFS performance on environmental variables related to storm activity A look at CFS-based predictions for 2008 season Summary and future plan
CFS T382 hurricane season experiments 1.One of the FY07/08 CTB internal projects 2.AGCM operational NCEP GFS in T382/L64 resolution LSM - Noah LSM OGCM - GFDL MOM3 3. All runs initialized with NCEP/DOE R2 and NCEP GODAS. Initial conditions at 0Z, Apr. 19, 20, 21 and May 15(FY07) for Forecasts extended to December 1. 4.For May 15 cases, runs in T62 and T126 resolutions also performed.
Datasets CFS hindcasts at T382 May 15 th Initial Condition Output at every 3 hours , 27 years April 19 th, 20 th and 21 st Initial Condition Output at every 6 hours , 28 years More appropriate ICs for CPC Operational Hurricane Season Outlook Observations from the HURDAT and JTWC Best Track Dataset Tropical depressions and subtropical storms are not included in storm counts.
Tropical Cyclone Detection Method Based on method devised by Camargo and Zebiak (2002) at IRI Detection algorithm uses basin-dependent threshold criteria for three variables Vorticity, thresh thresh = 2 10-m Wind Speed, thresh thresh = gl + gl = wind speed averaged over all global basins Vertically integrated temp anomaly, T thresh T thresh = T /2 Calculated using only warm-core systems
Table of Basin Thresholds ATLENPWNPNI * T CFS T382 thresholds for vorticity, surface wind speed, and vertically integrated temperature anomaly for each ocean basin.
1.850-hPa relative vorticity > thresh 2.Maximum 10-m wind speed in a 7x7 box > thresh 3.SLP is the minimum in the centered 7x7 box 4.Temp anomaly averaged over the 7x7 box and three pressure levels (300, 500, & 700 mb) > T thresh 5. Temperature anomaly averaged over the 7x7 box is positive at all levels (300, 500, & 700 mb) * 6.Temperature anomaly averaged over the 7x7 box at 300mb > 850mb * 7.Wind speed averaged over the 7x7 box at 850mb > 300mb 8.The storm must last for at least 6 hours. * Criteria 5 & 6 define a warm core system. Eight conditions must be met for a point to be considered a tropical cyclone
Storm Tracking Once a point is designated as a tropical cyclone, the cyclone is tracked forward and backward in time to create a full storm track. The maximum vorticity in a 5x5 grid around the cyclone is found and a new 3x3 box is formed around it. At the next time step, if the maximum vorticity in this new box is greater than 3.5 x s -1, the procedure is repeated. This point has become part of the storm track. If two storm tracks are the same, they are considered the same cyclone and counted as one.
Four NH Ocean Basins
Western North PacificNorth Indian AtlanticEastern North Pacific Examples of Storm Tracks for 4 NH Basins
Atlantic Basin Atlantic Tropical Storms
Eastern Pacific Basin Eastern Pacific Tropical Storms
Western Pacific Basin Western Pacific Tropical Storms
Atlantic Basin Correlations Anomalies of Atlantic Storms N=27 N=13 N=13 CorrelationsTotal IC= IC= IC= April Ensm IC= Red = Statistically Significant at 0.95
N=27 N=13 N=13 ENP Correlations Anomalies of Eastern North Pacific Storms CorrelationsTotal IC= IC= IC= April Ensm IC= Red = Statistically Significant at 0.95
N=27 N=13 N=13 WNP Correlations Red = Statistically Significant at 0.95 Anomalies of Western North Pacific Storms CorrelationsTotal IC= IC= IC= April Ensm IC=
N=27 N=13 N=13 CorrelationsTotal IC= IC= IC= April Ensm IC= Atlantic Basin ACE Index Red = Statistically Significant at 0.95 % of Normal
ATL Detrended Correlations N=27 N=13 N= IC= April Ensm IC= IC= IC= TotalCorrelations Red = Statistically Significant at 0.95
ATL Detrended, Step Function Correlations N=27 N=13 N= IC= April Ensm IC= IC= IC= TotalCorrelations Red = Statistically Significant at 0.95 Climatology used for detrending is broken into two 13-year periods, the inactive period of and the active period
Evaluation of SST and Wind Shear prediction for storm season
JJA Nino 3.4 SST Index T382COR IC= IC= IC= April Ensm0.707 IC= Red = Statistically Significant at 0.95
ASO Nino 3.4 SST Index T382COR IC= IC= IC= April Ensm0.676 IC= Red = Statistically Significant at 0.95
JJA Atlantic MDR SST Index T382COR IC= IC= IC= April Ensm0.691 IC= Red = Statistically Significant at 0.95
ASO Atlantic MDR SST Index T382COR IC= IC= IC= April Ensm0.744 IC= Red = Statistically Significant at 0.95
JJA Atlantic MDR Shear Index T382COR IC= IC= IC= April Ensm0.603 IC= Red = Statistically Significant at 0.95 Wind Shear: U 200 – U 850
ASO Atlantic MDR Shear Index T382COR IC= IC= IC= Apr Ensm0.502 IC= Red = Statistically Significant at 0.95 Wind Shear: U 200 – U 850
TNA Non-Local Index, Jun-Nov T382TNA IC= IC= IC= April Ensm0.626 Red = Statistically Significant at 0.95 [60W-30W; 5N-20N] - [0-360, 0-15N] SSTA T382 ACE Obs/Fcst IC= /0.632 IC= /0.373 IC= /0.449 April Ensm0.662/0.648 Obs0.699
2008 Atlantic Season Summary Obs NOAA Outlook Outlook Update CFS-Stat May CFS-Stat Update Named Storm Hurricane Major Hurricane
Atlantic Basin Prediction for 2008 Two additional runs were made using July 15 th and 16 th initial conditions for 2008 only Used as a trial run for the CPC Hurricane Outlook Update
ENSM Obs TotalNovOctSepAugJulJunMay2008 Atlantic Basin CFS 2008 Monthly Storm Count 3
(Courtesy of Unisys) Large number of storms over the Gulf of Mexico this year 2008 Atlantic Storm Tracks
Resemble Bertha and Cristobal
. Summary CFS in T382 resolution exhibits robust climatological seasonal cycle of tropical cyclone over four NH basins. Warming trend and intensification of hurricane activity in the Atlantic basin captured in the CFS hindcasts. Fair level of skill in predicting interannual variability of seasonal storm activities based on the limited number of forecast runs. Model SST and wind shear bias explain part of less number of TCs in Atlantic and EN Pac. basins, in particular over the Gulf of Mexico. Continue to increase number of ensemble members for better climatology and storm statistics. Hope to provide input for 2009 Hurricane Season Outlook with real time prediction runs. Multi-model ensemble approach for dynamical hurricane season prediction is being explored.