Do the S2S models have prediction skills beyond the weather timescale for winter snowfalls over eastern China?

During the winter of 2023/24, three distinct snowfall events occurred in eastern China, significantly impacting agriculture and transportation. The ability to provide subseasonal predictions with lead times beyond the weather timescale (longer than one week) is essential for effective disaster prevention and mitigation. Here, we assess the prediction skills of three subseasonal to seasonal (S2S) models from S2S prediction project regarding the three snowfall processes during the 2023/24 winter season, and identify the key sources of predictability for such snowfalls occurring over eastern China. The surface air temperature (SAT) and precipitation distribution for the three snowfall processes were successfully reproduced up to a lead time of 10–15-day and 10-day, respectively. Since the skill of predicting snowfalls is reliant on both SAT and precipitation predictions, and therefore all three S2S models failed to predict the three snowfall processes beyond the weather timescale. The model’s capacity in capturing Eurasian mid-latitude transient Rossby waves and tropical convection anomalies determines their ability to predict snowfalls, inaccuracies in modeling these circulation systems result in an underestimation of SAT and precipitation anomalies beyond 15 days and 10 days, respectively. The Singular Value Decomposition analysis based on winter seasons from 1991/92 to 2023/24 further identified the coupling modes that exist between Eurasian mid-latitude Rossby waves and SAT over eastern China, as well as between tropical convection and precipitation over the same region. These findings suggest that the configurations of tropical and extratropical signals provide universal subseasonal predictability sources for winter snowfall over eastern China.