An Investigation of the Spring Predictability Barrier Yehui Chang, Siegfried Schubert, Max Suarez, Michele Rienecker, Augustine Vintzileos* and Nicole Kurkowski Global Modeling and Assimilation Office, NASA/GSFC *Global Climate and Weather Modeling Branch, Environmental Modeling Center/NOAA NSIPP CGCM Introduction In this study, we use the NSIPP coupled model forecasting system to examine the nature of the spring predictability barrier. The focus is on the predictability of subsurface variability, and how that evolves to limit predictability in the sea surface temperatures and ultimately the atmosphere. The skill of both persistence and coupled model hindcasts of the Nino3.4 index show a clear signature of a spring barrier in predictability. The study demonstrates that there is no spring persistence barrier for upper ocean heat content consistent with McPhaden . Lagged correlations show the anomalous warm water volume (WWV) leads Nino3.4 SST by 2-3 seasons. The results of this study agree with other ENSO forecast model studies indicating that accurate initialization of the equatorial upper ocean heat content reduces the spring prediction barrier for SST. Data Sets вЂў вЂў вЂў вЂў AGCM: Finite-difference 2x2.5, dynamical core (Suarez and Takacs, 1995) PBL scheme (Louis et al., 1982) Radiative scheme (Chou and Suarez, 1996) Convection RAS (Moothi and Suarez, 1992) Gravity wave drag (Zhou et al., 1996) вЂў вЂў вЂў вЂў OGCM: Poseidon V4 quasi-isopycnal 1/3x5/8 (Schopf and Lough, 1995) Embedded surface mixed layer Kraus-Turner Vertical mixing and diffusion (Pacanowski and Philander, 1981) Data Assimilation (Keppene and Rienecker, 2003) вЂў LSM: вЂў Mosaic Land Surface Model (Koster and Suarez, 1996) Skill of persistence forecast for Nino3.4 index (1993-2005) Modeled Lag correlation of monthly anomalies in Nino3.4 SST as a function of start month and lead-time for coupled model forecasts. Summary and Conclusions пѓѕ A boreal spring predictability barrier exists in SST. In contrast, WWV does not show a spring predictability barrier. In fact, February-March WWV anomalies have the greatest persistence. пѓѕ The correlations show that the WWV leads Nino3.4 SST by 2-3 seasons. пѓѕ SST spring barrier and WWV persistence barrier developed in the boreal winter consistence with the observed results of McPhaden . пѓѕ Initialization of the equatorial subsurface helps break through the spring barrier. Skill of persistence forecast for Warm Water Volume (WWV) Nino3.4 vs WWV Crosscorrelations Observed Lag correlation of monthly anomalies in Nino3.4 SST as a function of start month and lead-time for observed results. Skill of CGCM forecast for Nino3.4 index (1993-2005) Forecasting skill of coupled model experiments. пѓ�3-member ensemble 12-month coupled model forecasts. пѓ�Forecasts initialed at 1st day of each month from 1993-2005. пѓ�The Reynolds SST for the same period. Difference of forecast skill between Baseline forecasts and model persistence forecasts. Lag correlation of monthly anomalies in WWV as a function of start month and lead-time. Crosscorrelation between monthly WWV and Nino3.4 SST anomalies as a function of start month and lead-time. Positive (negative) leads imply WWV leads (lags) SST. Monthly anomalies of WWV and Nino3.4 SST Monthly anomalies of WWV (5S-SN, 120E-90W above 20C isotherm) and Nino3.4 SST (5S-5N,170W-120W) for observed (red) and model forecasts (green) starting from February. Monthly anomalies of WWV (5S-SN, 120E-90W above 20C isotherm) and Nino3.4 SST (5S-5N,170W-120W) for observed (red) and model forecasts (green) starting from November.