Systematic Error Characteristics of Weeks 3 and 4 2-m Temperature Forecasts in the NCEP GEFS Reforecast for SubX

The NCEP Environmental Modeling Center generated an 18-year sub-seasonal reforecast dataset to support the CPC’s operational mission. The GEFS-SubX version was run every 7-day initialized at 0000UTC with 11-members.The control initial conditions were generated by the Climate Forecast System Reanalysis (CFSR) and Grid-point Statistical Interpolation (GSI) system for the period of 1999-2010 and 2011-2016, respectively. The evaluation from a 2-year full retrospective runs indicates the GEFS-SubX significantly improves the forecast skill of MJO when compared to the current operational GEFS v11. The skill in predicting 500hPa geopotential height, precipitation, and 2-m temperature are also improved to some degree. The improvements are attributed to a combination of a new convection scheme, stochastic physical perturbations, and a two-tiered SST approach.

In this study, we analyze the bias characteristics of weeks 3 & 4 2-m temperature in the SubX reforecast dataset. The model bias attributed to inconsistent initial analyses will be investigated. The time-scale at which 2-m temperature bias reaches a saturation status will be identified to determine whether we could use week-2 bias to calibrate weeks 3 & 4 forecasts. Finally, we will explore the possible sources of 2-m temperature error.