The daily temperature data at 599 stations during 1961-2007 are utilized to analyze the multispatial scale, multitime scale changes of seasonal start date based on wavelet analysis. The results show that the changes of seasonal start date are accordant with the changes of temperature over China. On national average, the seasonal start date all has a time period of nearly 20 years, and show a clear trend that spring and summer became earlier, autumn and winter became later in the late 1980s. After doing a wavelet transform to the major cycle over all regions, the trends are the same as the national average basically. However, some differences are still existing in varied regions. For example, in Northeast China there was a phenomenon of summer occurring later during 2005-2007. Winter occurring earlier in Northwest China and spring occurring later in South China during 2004-2007. In the Plateau spring also occurred later during 2003-2007， combined with the other seasonal changes, the changes of seasonal start date in these regions are also corresponding to the changes of temperature well.