Deterministic forecasts are initialized at 0000 UTC from the Climate Forecasting System Reanalysis (CFSR) on the 1st and 15th of the month over a 6-year period spanning April 2011 to March 2017 and integrated for 35 days. A fully coupled UGCS prototype experiment is performed along with a set of two-tiered uncoupled SST forcing experiments that relax the CFSR SST to (i) a bias corrected UGCS predicted SST (SST taken from the fully coupled experiment), (ii) a bias corrected Climate Forecasting System Version 2 (CFSv2) predicted SST, (iii) the predicted UGCS SST, and (iv) the CFSR SST analysis. All SSTs are updated daily during the model integration for the uncoupled experiments. Calibration climatologies are generated for all studied model forecast variables (SST, 2-m temperature, accumulated precipitation, 500-hPa geopotential heights, Outgoing Longwave Radiation, and 850- and 200-hPa zonal winds) for each experiment (including the 0000 UTC control member of the CFSv2 for comparison) as well as for the verification data sets to allow for consistent anomaly calculations and systematic error correction for all experiments.
Preliminary findings show that the prototype UGCS system performs as well or better than the control operational CFSv2 and all uncoupled UGCS experiments over the 6 year period over the NH for all variables studied when comparing anomaly correlations (ACs) over the weeks 3 & 4 period. Leading up to the subseasonal time scales, the fully coupled experiment does not hurt the 500-hPa geopotential height ACs and generally performs better than the uncoupled experiments. Over the tropics, the fully coupled prototype UGCS substantially improves the 850- and 200-hPa equatorial zonal ACs and the corresponding MJO RMM indices. Both the 2-m temperature and SST have a large systematic error correction for all experiments (accumulated precipitation and 500-hPa geopotential heights are much smaller), highlighting the need for hindcasts for subseasonal prediction.