1. The Relative Contribution of Atmospheric and Oceanic Uncertainty in TC Intensity Forecasts Ryan D. Torn University at Albany, SUNY World Weather Open Science Conference 20 August 2014 Montreal, Quebec

2. NHC Official Errors 3-4% reduction in TC track error per year at all lead times Relatively small reduction in intensity forecasts at short lead times, more so at longer lead times Need to understand why intensity errors are not decreasing National Hurricane Center

3. NHC Official Errors National Hurricane Center

4. Intensity Factors Courtesy National Hurricane Center TC Environment Surface Boundary

5. Large-Scale Sensitivity Kaplan & DeMaria 2003

6. Intensity Factors Courtesy National Hurricane Center TC Environment Surface Boundary Internal Processes

7. Van Sang et al. (2008) Zhang and Tao (2013) Moisture Exps. Without Shear With Shear

8. Motivation Fundamental question of how predictable TC intensity forecasts are given errors in model initial condition and formulation Not clear whether atmospheric and oceanic errors are complimentary to each other or orthogonal Has implications for designing tropical cyclone ensemble prediction systems Explore this using high-resolution ensemble forecasts that are characterized by different sources of variability

9. Data Use Advanced Hurricane WRF (AHW), 2012 HFIP retrospective configuration (Davis et al. 2008, 2010) –Initial conditions obtained from cycling ensemble Kalman filter (EnKF) using Data Assim. Research Testbed (DART) –Assimilates conventional data + dropsondes –Forecast: 36/12/4 km resolution, 30 ensemble members –1D column ocean model (Pollard 1973), NCEP SST, HYCOM MLD

10. Ensemble Experiments Atmosphere Only: –Use 30 analysis members from ensemble data assimilation system for atmosphere –Same ocean state for each member Ocean Only: –Use atmospheric analysis used in deterministic AHW forecast –Ocean perturbations from climatology Atmosphere + Ocean –Use atmospheric states from Atmosphere Only –Use ocean states from Ocean Only

11. Ocean Perturbation Ensemble ocean initial conditions hard to come by Ensemble ocean perturbations obtained by sampling from a 2006-2011 climatology of SST and MLD fields Remove mean, scale SST so that individual locations have a 0.5 K standard deviation MLD values are scaled by the same value

12. Cases Gustav (2008; 2)Erika (2009)Igor (2010; 3)Richard (2010)Maria (2011) Hanna (2008; 2)Fred (2009)Julia (2010; 2)Tomas (2010; 2)Ophelia (2011) Ike (2009; 3)Danielle (2010; 3)Karl (2010)Irene (2011; 2)Philippe (2011) Bill (2009; 2)Earl (2010; 2)Otto(2010)Katia (2011; 2)Rina (2011)

13. Overall Statistics Maximum Wind SpeedMinimum SLP

14. Will measure growth of initial condition uncertainty via amplification factor

15. Amplification Factors Atmosphere OnlyOcean Only

16. Amplification Factors Atmosphere Only

17. Case Comparison Atmosphere OnlyOcean Only

18. Case Comparison Atmosphere OnlyOcean Only

19. Case Comparison Sea Surface Temperature700 hPa Relative Humidity

20. Igor: 0000 UTC 9 Sept. 2010

21. Study process that lead to forecast error growth by comparing the 10 most intense members against the 10 least intense members

22. Igor 0520 UTC 10 Sept. Courtesy: NASA TRMM

23. IC Differences Axisymmetric Tangential WindAxisymmetric Water Vapor

24. Flux Differences

26. F018 F036

27. Left of Shear Water Vapor

28. Summary and Future Work Initial condition errors in the atmosphere translate into larger intensity variability on short time scales, while ocean initial condition errors are associated with comparable intensity variability over longer lead times (5 days) Largest error growth rate for weak storms with small 34 knot wind radii and/or storms in environment with higher RH Igor forecasts most sensitive to initial tangential wind errors, with little initial sensitivity to moisture Intensification initially slowed by greater downdrafts that are later mitigated by surface fluxes and moistening from above Suggestive that knowledge of 3D wind field for sheared TCs may benefit forecasts (i.e., Doppler Radar)