The change of the monthly mean temperature for March since the 11th century is estimated from the records of cherry blossom in old documents. The dates given as cherry blossom festivals in old diaries and chronicles are assumed as full flowering dates of Prunus jamasakura, one of the native cherry tree species of Japan. The temperatures are estimated by means of DTS method, which has been used to estimate flowering date from temperature data. It is shown that computations for the recent 40 years proved that the decadal average values of the March temperature may be obtained by the method for P. jamasakura with the accuracy of 0.1°C of the root mean square error.
It is shown that the entire study period can be roughly divided into three subsections. In the first period (the 11-13th centuries), the March appeared to be generally very warm compared with other periods, however, accuracy of the estimated values are rather poor because decadal averages are computed from relatively small number of years due to missing of records.
For the second period (the 14-16th centuries), there was a period of warming of 2°C from 1470's to 1610's, otherwise the decadal averages suggest large fluctuations, at least partly attributed to poor quality of full flowering data of P. jamasakura.
For the third period (the 17-20th centuries), temperatures generally tend to be estimated in low values, especially estimations in 1690-1710's and 1810-30's continued to the present is noticeable. The rise amounts to 3.4°C seems due both to urban warming and larger scale warming over the central Japan roughly the same degree.