Investigating SST Bias in the UFS Seasonal Forecast

We analyzed predicted sea surface temperature (SST) patterns using multiple seasonal forecast experiments with NOAA's coupled Unified Forecast System (UFS) model. The UFS model incorporates the FV3 atmospheric model with the Global Forecast System (GFS) physics package, as well as the MOM6 ocean model and CICE6 sea ice model. The baseline experiment utilized the UFS coupled model Prototype 8, with the exception of employing a coarser 1-degree mesh for the atmospheric, oceanic, and ice models. Additionally, the ocean's initial conditions were from the ECMWF Ocean Reanalysis System 5. Notably, the tropical SST exhibited an overall positive model bias. Discrepancies in global energy fluxes at both the top and bottom of the model's atmosphere suggest that more attention needs to be paid to conserving energy, and potentially mass field constituents, in the physics equations.

To investigate further, we conducted a second experiment mirroring the control experiment in all aspects, except for replacing the original NSST (near-surface SST) algorithm, initially designed for uncoupled applications. We replaced it by a skin temperature scheme that simplifies the information transfer from the mixed layer module in MOM6. The outcome demonstrates a reduction in bias in tropical SST in comparison to the control experiments. This highlights the sensitivity of the coupled system to differences in ocean skin temperature parameterization.