How well do the S2S models predict intraseasonal wintertime surface air temperature over mid-high-latitude Eurasia?

The leading modes of intraseasonal surface air temperature (SAT) over mid-high-latitude Eurasia (MHLE) in boreal winter are captured with the observational 10–60-day SAT anomaly during 1979–2016 by an empirical orthogonal function (EOF) analysis. Each cycle of the southeastward-propagating intraseasonal oscillation (ISO) of SAT is divided into eight phases formed by the first two principal components (PC1 and PC2). Reforecast data from five operational models in the Sub-seasonal to Seasonal (S2S) Project are utilized to assess the prediction skill of the intraseasonal SAT over the MHLE based on the PC index. It is revealed that the upper limit of the useful forecast skill ranges from ~ 10 to ~ 20 days and the models display a better score for target strong cases, although they underestimate the amplitude of the intraseasonal signal. Generally, a higher prediction skill appears in target phase 3, while a lower score exists in target phase 6. The phase-dependent feature of prediction skill is found to be associated with the phase-dependence of amplitude bias and phase angle error, indicating the importance of skillfully predicting ISO amplitude and propagation speed. On the basis of PC index, the ECMWF model can skillfully predict the evolution of an ISO event up to ~ 20 days in advance. Further assessment of potential predictability reveals a 5-day skill gap for ECMWF model, which needs to be overcome by reducing forecast errors for the current prediction scheme. In a nutshell, the real-time PC index can be an efficient method to capture and predict the ISO signal over MHLE, providing possibilities for improving sub-seasonal forecast skill of wintertime intraseasonal SAT anomaly. Meanwhile, more efforts are still needed for the improvements of dynamical prediction systems, to facilitate more accurate and longer-lead predictions of intraseasonal SAT signal over MHLE during boreal winter.