2003 Volume 59 Issue 2 Pages 165-177
The generalization of model to estimate flowering dates for cherry trees (Prunus yedoensis) is attempted, considering processes both of endodormancy completion and development. As the basis of estimation, an accumulation model using an exponential function of daily mean temperature, called DTS (the number of days transformed to standard temperature) method, is used for expressing development of flower buds. The flowering date is estimated as the date on which the DTS, accumulated since a suitable starting date, reaches a predetermined DTS value. In this study, three models with different procedures to determine suitable starting data and DTS accumulations for flowering are constructed and verified, using data from meteorological stations in the warm region of Japan. The model A requires the complex procedure with analysis of previous temperature and phenological data to determine suitable starting date and DTS accumulations in each station, although this model gives accurate estimations. The generalized model B, without consideration of endodormancy completion process, includes the simplified procedure for determination of starting date and the constant DTS accumulations for flowering for all stations in Japan. However, in the warm region of Japan, interannual variation in advancement of the endodormancy completion process, which was influenced by winter temperature conditions, causes the change in the DTS accumulation required until the actual flowering date, and is attributed to the reduction in accuracy of estimations from models A and B. In generalized model C with consideration of endodormancy completion process, the suitable starting date is determined for each station, applying the equation, which includes some climatic and topographic factors, and the DTS accumulations for flowering are deduced as a function of the chill-unit accumulations, which represents the degree of endodormancy completion advancement in each year. On Hachijojima island, near the warm-side boundary for growing, model C reduced RMSE of estimates to 3.1 days from 6.4 days of that from model A. In the model verification using data from the warm region, the averages of RMSEs in estimations from models A, B and C were 2.7, 3.0 and 2.8 days, respectively. It means that the consideration of endodormancy completion process in model C compensated the decrease in model accuracy, which is attributed to introducing some simplified procedures into the model.
