The prediction of flowering date of Japanese cherry blossoms has an economic value on the local tourism industry, which could help the local government and tourists manage and arrange the tourism time. In this study, we investigated the prediction methods of flowering dates of Japanese cherry blossoms, and developed a new prediction method to improve accuracy. First, a 36-year-old (1981-2016) dataset of flowering dates of Japanese cherry blossoms at Wuhan University Campus associated with meteorological data was used for developing a method of forecasting the first-flowering date of cherry blossoms. The first-flowering date of Japanese cherry blossoms was determined by linear regression and the trend of flowering dates over the last 36 years. The annual variation in the first-flowering dates was large, making the trend non-significant. The average of flowering date, in days of the year, was 73.3, which corresponded to March 14-15 (March 13-14 in a leap year). A series of sensitive studies on active accumulated temperature were performed to investigate the method of improving the forecast equation of the first-flowering date. We calculated the active accumulated temperature from January 1 and February 1 to a series of possible flowering dates, such as February 25, February 28 (February 29 in a leap year), March 5, March 10, and March 15, and the results indicated that the correlation between the active accumulated temperature and the first-flowering dates was significantly negative, and thus can be used as a factor in the cherry blossom forecasting equation. The results also showed the forecast accuracy was significantly improved with active accumulated temperature rather than accumulated sunshine hours as a predictor in the prediction equation. Compared to the prediction using the average temperature, the active accumulated temperature method is more advantageous for prediction accuracy, and the better result is achieved when the prediction is made close to the real flowering date. We also discuss the prediction method that involved the accumulated sunshine hours, because the first-flowering dates were negatively correlated with the accumulated sunshine hours. This method, however, was not effective in improving the prediction accuracy and requires further investigation.