A skillful hybrid framework for seamless subseasonal PM2.5 prediction over North China

Severe PM2.5 pollution in North China persists despite substantial emission controls, underscoring the pivotal role of subseasonal meteorology and the urgent need for skillful prediction. Here, we propose a novel hybrid dynamical-statistical framework by integrating circulation evolution and overlapping time windows (ICEOTW) to bridge the critical 10–30-day prediction gap. We first establish that the dominant subseasonal PM2.5 variability is intrinsically linked to an eastward-propagating Rossby wave train associated with the Eurasian teleconnection mode. Building on this mechanism, key circulation variables are selected as predictors that collectively capture the coherent tropospheric structure of the Eurasian teleconnection mode and exhibit superior predictive skill in the European Centre for Medium-Range Weather Forecasts (ECMWF) Subseasonal-to-Seasonal (S2S) model. The ICEOTW framework shifts the focus from conventional day-to-day state prediction to continuous 30-day process prediction by modeling the relationship between the 30-day evolution of circulation and the 30-day evolution of PM2.5 concentrations. It strategically constructs predictors by merging recent observations with future ECMWF S2S model forecasts within overlapping time windows, thereby bridging initial conditions with predicted circulation evolution. This framework demonstrates statistically significant deterministic skill and reliable probabilistic forecasts of PM2.5 anomaly events up to 20 days ahead. This work provides a physically coherent and operationally viable approach for seamless subseasonal air-quality prediction, supporting early warning and proactive environmental management.