Climatic changes in the onset of spring in northern China associated with changes in the annual cycle and with a recent warming trend were quantified using a recently developed adaptive data analysis tool, the Ensemble Empirical Mode Decomposition. The study was based on a homogenized daily surface air temperature (SAT) dataset for the period 1955-2003. The annual cycle here is referred to as a refined modulated annual cycle (MAC). The results show that spring at Beijing has arrived significantly earlier by about 2.98 d (10 yr)-1, of which about 1.85 d (10 yr)-1 is due to changes in the annual cycle and 1.13 d (10 yr)-1 due to the long-term warming trend. Variations in the MAC component explain about 92.5% of the total variance in the Beijing daily SAT series and could cause as much as a 20-day shift in the onset of spring from one year to another. The onset of spring has been advancing all over northern China, but more significant in the east than in the west part of the region. These differences are somehow unexplainable by the zonal pattern of the warming trend over the whole region, but can be explained by opposite changes in the spring phase of the MAC, i.e. advancing in the east while delaying in the west. In the east of northern China, the change in the spring phase of MAC explains 40%-60% of the spring onset trend and is attributable to a weakening Asian winter monsoon. The average sea level pressure in Siberia (55°-80°N, 50°-110°E), an index of the strength of the winter monsoon, could serve as a potential short-term predictor for the onset of spring in the east of northern China.