Progress on the Development of a Coupled Forecast System for Sub-seasonal to Seasonal Prediction at NCEP/EMC

The NOAA Environmental Modeling System (NEMS) is a system infrastructure used for building a collection of modeling applications within a unified framework. It is based on the Earth System Modeling Framework (ESMF) and National Unified Operational Prediction Capability (NUOPC) Layer extensions to ESMF. The coupling of different model components is through a mediator component within NEMS. The roles of the mediator include field interpolation, field averaging, field merging, treatment of land-sea masks from component models, and custom air-sea flux calculations. NOAA is using this design architecture to build the Unified Forecast System (UFS) to provide guidance from weather to seasonal scales. At this moment, the FV3GFS atmospheric model, the MOM6 ocean model and the CICE5 sea ice model have been coupled using the NEMS mediator. A Community Mediator for Earth Prediction System (CMEPS) developed by NCAR and NOAA will be implemented in the near future once coupling behavior of the NEMS mediator is reproduced. Once completed, the coupled system will be used for sub-seasonal to seasonal (S2S) prediction at NCEP. During the initial development phase, the focus is on the sub-seasonal time scale.

We will present the current coupling setup using the NEMS mediator. Two sets of benchmark experiments have been run using a coupled FV3-MOM6-CICE5 model (UFS). For each benchmark set, the UFS is initialized on the 1^st and 15^th of each month between April 2011 and March 2018. The length of each model run is 35 days. These two benchmarks have shown that the UFS has better performance in reproducing the sub-seasonal sea-ice evolution over both hemispheres than the operational CFSv2. The UFS also has better skill than CFSv2 in terms of forecasting SST anomalies in the Niño 3.4 region and 2-m air temperature and precipitation in the continental United States. In the third set of experiments, which is ongoing, the UFS includes improved model physics and it uses improved initial conditions for the sea ice model from NCEP/CPC data assimilation system. Results from those three sets of experiments will be compared and discussed in terms of prediction skill.