Forecasting East Asian winter temperature via subseasonal predictable mode analysis

The development of dynamical forecast systems has promoted seasonal forecast skills of winter climate; however, predicting subseasonal variations of East Asian (EA) winter temperature remains a challenging issue. To address this, this study presents a Subseasonal Predictable Mode Analysis (S-PMA) method that integrates the Season-reliant Empirical Orthogonal Function (S-EOF) analysis with the PMA approach and conducts retrospective predictions based on the physical interpretation of leading S-EOF modes. Three distinct S-EOF modes of EA winter temperature during the period of 1979-2021 have been identified: (1) the Consistent Mode (C-Mode) with a continuous in-phase temperature anomaly in winter (November-February); (2) the Progressive Reversal Mode (PR-Mode) with a slow-varying reversal of temperature anomaly between November-December and January-February; (3) the Rapid Reversal Mode (RR-Mode) with a fast-varying reversal of temperature anomaly between December and January-February. The C-Mode is linked to a simultaneous Arctic Oscillation anomaly and positive land-atmosphere feedback, while the PR-Mode and RR-Mode result from interactions between the preceding atmospheric variability and surface boundary conditions. The first three S-EOF modes could collectively explain over 40% of the total temperature variances over the mid-high latitudes of EA. A set of physically-based empirical prediction (PEP) models is established to hindcast each principal component (PC) of the S-EOF modes, with correlation coefficients of 0.62, 0.63, and 0.46, respectively. Compared to dynamical models, the PEP models show advantages in enhancing skills over extratropical land regions of EA and in late winter. A hybrid PEP-dynamical model is also introduced to complement their strengths. These results can facilitate seasonal forecast skills of winter temperature in not only EA, but also the entire Asian continent. Furthermore, the S-PMA method can be applied to a wide range of climate research in seasonal forecast and predictability.