This project will construct, test and prepare for implementation an ensemble forecast system for the 1-35 days lead-time with the more advance ensemble methods, coupled with realistically evolving SST that outperforms current skill benchmarks, providing routine forecast outputs to CPC forecasters and contributing to the NMME-Phase 2 sub-seasonal project. The motivation of this project is the potential to implement a two-tiered GEFS forecast and hindcast system of “opportunity”, which can be setup and run routinely within a year. The two-tiered approach consists in prescribing bias-corrected predicted SSTs from the CFSv2 as the integration of the GEFS moves forward. The approach has been tested in the parallel version of the GEFS in a limited set of experiments resulting in skill gains in predicting the MJO signal and reducing the RMSE of forecasts of upper air circulations for weeks 3 and 4. The hindcast is being completed for the first 16-days forecast segment and an extension to the 35-days can be generated for the last 20 years and be setup to provide real-time updates. In parallel with this activity, a next version of the GEFS will be tested in which surface (SST and land) variables are stochastically perturbed to represent analysis uncertainty at initial time.