Predicting wintertime windy days in Beijing with the preceding autumn sea surface temperature and large-scale circulation patterns

In this study, we investigate the role of sea surface temperature and large-scale circulation patterns in the preceding autumn for the prediction of wintertime windy days in Beijing during the period of 1979–2016, based on a time-scale decomposition approach. A windy day is defined according to the threshold of 2 m·s −1 , which can effectively remove or reduce air pollution of Beijing. Wintertime windy days in Beijing, preceding autumn sea surface temperature and large-scale climate indices are separated into the interannual and interdecadal components. For the interannual component, the preceding autumn sea surface temperature in central western Pacific Ocean, Polar/Eurasia pattern, and Scandinavia pattern are identified as predictors of wintertime windy days in Beijing. For the interdecadal component, the identified predictors include the preceding autumn sea surface temperature in northeastern part of North Atlantic Ocean, Arctic Oscillation, and North Atlantic Oscillation. We develop statistical models for the interannual and interdecadal components based on these preceding autumn predictors, and take them together to predict wintertime windy days in Beijing. The leave-one-out cross-validation for 1979–2016 time series shows that 76% of the total variance of observed wintertime windy days in Beijing can be predicted. Hindcast experiments for the period of 1999–2016 also yield the high prediction skills. The physical mechanisms responsible for links between these predictors and wintertime windy days in Beijing are also discussed. The statistical models proposed herein are expected to be applied to and aid to improve seasonal forecasting of wintertime windy days in Beijing.