Potential Sources of Predictability for the Surface Air Temperature over the CONUS During Boreal Winter

Abstract Subseasonal to seasonal (S2S) prediction skill of wintertime surface air temperature (SAT) over the contiguous United States (CONUS) remains limited by systematic model biases and poorly understood sources of predictability. Using deterministic reforecasts from the NOAA Unified Forecast System prototype 8 (UFS P8) for December–March 2011–18, we examine dominant SAT bias patterns and their physical origins associated with Arctic variability, tropical forcing, and land–atmosphere interactions. Empirical orthogonal function analysis applied to UFS P8 identifies two leading modes of winter SAT bias. The first mode is linked to Arctic surface temperature variability over the East Siberian–Chukchi Seas, where forecast skill decays beyond week 2. The second mode features a dipole pattern associated with an exaggerated upper-level circulation response and local surface processes. In boreal winter, UFS P8 exhibits an unrealistic upper-level circulation modulated by an erroneous, summer-like Rossby wave. Over the eastern CONUS, surface warming is primarily driven by excessive downward longwave heating. Despite a wet soil moisture bias, increased soil moisture does not enhance latent heat flux, implying inefficient surface energy partitioning and reduced evaporative cooling that amplify the warm bias. These combined deficiencies contribute to systematic warm biases in winter SAT over the eastern CONUS. They indicate that surface warming is primarily driven by cloud-induced downward longwave radiation, but its magnitude and variability are modulated by land–atmosphere interactions via surface energy partitioning. Thus, these coupled processes represent an unexploited source of predictability for S2S forecasts and highlight land surface processes as a barrier to advancing extended-range forecast skill.