Principal Investigator: Siegfried Schubert GMAO/GSFC/NASA Co-Investigators: Michele Rienecker , GMAO/GSFC/NASA

1. PROPOSALSubmitted to theNOAA Climate Program OfficeClimate Dynamics and Experimental Prediction (CDEP) ProgramCLIMATE TEST BEDDEVELOPMENT OF SUBSEASONAL ENSEMBLE FORECAST TECHNIQUESSeptember 24, 2007 Principal Investigator: Siegfried Schubert GMAO/GSFC/NASA Co-Investigators: Michele Rienecker, GMAO/GSFC/NASA Shu-Chih Yang, Dep Atmos Sci, National Central Univ., Taiwan Max Suarez, GMAO/GSFC/NASA Yehui Chang, GEST/UMBC at GMAO/GSFC/NASA Malaquías Peña, SAIC at EMC/NCEP/NOAA Zoltan Toth, EMC/NCEP/NOAA Augustin Vintzileos, UCAR at EMC/NCEP/NOAA Collaborators: Eugenia Kalnay, UMD David Unger, CPC/NCEP/NOAA Dan Collins, CPC/NCEP/NOAA

2. Focus is on developing and testing improved methods for initializing ensemble forecasts for subseasonal time scales • Key features: • Introduces initial perturbations that are optimized for subseasonal time scales (initial focus on ISO) • Allows for coupled (ocean-atmosphere) instabilities • Allows for rapid updates to the system (bias correction) • Addresses sensitivity to model physics (NCEP physics in GEOS-5)

3. Relevant GMAO Developments/Projects • Implemented a coupled bred vector approach for initializing seasonal-to-interannual (ENSO) forecasts • NASA-funded research projects to address predictability and prediction skill at subseasonal to S-I time scales • Implemented the NCEP physics in the GEOS-5 model • GEOS-5 is coupled to MOM4

4. Impact of Breeding on 9-Month SST Forecasts (1993-2005) Control Breeding Anomaly correlation between the 9-month lead SST anomaly (averaged from nine November initial conditions from 1993 to 2005) and the Reynolds SST anomaly using the GMAO CGCMv1 coupled forecast system. The left panel uses the single unperturbed initial condition and the right panel uses the mean of 8 bred ensemble members.

5. NASA Bred Vector vs. NCEP Bred Vector SST EOF1 Bred vectors obtained with an 8-year NCEP run are extremely similar to the NASA’s 20-year run!!! Z20 EOF1 Z20 EOF2 Yang et al., J. Clim. 2006

6. NASA GEOS-5 CGCM Coupled versus uncoupled 30-day Hindcasts of MJO NASA GEOS-5 AGCM

7. Relevant NCEP Developments and Plans Global ensemble forecast (GEFS) system ET introduced for atmospheric perturbations in May 2006 Upgrades – Dec. 2008 T126 L28 extension out to 35 days, 1 / day (00Z), 21 members Testing ground for CTB research results T190 L28 out to 16 days (replacing of T126 L28), 4 / day, 21 members Stochastic perturbations (& higher order diffusion) to increase spread CFSRR use for GEFS hindcast initialization use for GEFS forecast bias correction GFS coupling to ocean -Plan to operationally implement in 2009 Use same model version in GEFS for CTB research 7

8. GEME – current GEFS with extension GEFS – current GEFS operation with bias correction CFSOP – current operational CFS with bias correction Courtesy of Qin Zhang

9. Tasks • Develop and test various perturbation methods in both the NCEP and NASA coupled systems including • the generation of short term hindcasts for specific cases (focus on periods with active ISOs) • running coupled GEOS-5/MOM4 with NCEP physics • Test new GEFS over extended time periods • Set up software for parallel testing of new GEFS • Carry out a coordinated set of subseasonal GEFS and GEOS-5 hindcasts • Prepare new sub-seasonal ensemble system for operational implementation • Product generation: connect with NAEFS, CPC and other applications • Upgrade model/DA as needed • Make recommendations on multi-model approach, based on: • A comparison of the skill of GEFS and GEOS-5 at subseasonal time scales • An assessment of the skill of the GEOS-5 model run with NCEP physics