Abstract
We constructed a simplified method to estimate first flowering dates of cherry tree (Prunus×yedoensis), considering temperature conditions in endodormancy process. We adopted the DTS method, which is an accumulation model of forcing effect of temperatures during developing process, as an estimation model of cherry blossom phenology in this study. In our previous studies, DTS method with fixed (pre-determined) starting dates for each site showed high accuracies for cold regions, whereas low accuracies for warm regions. Such accuracy drop was attributed to the volatility in endodormancy completion in warmer region of Japan, affected by inter-annual variation in chilling temperature during winter season. In order to reduce such error, it had been necessary to calculate chilling hours from hourly temperature data and evaluate the progress in endodormancy releasing process, with complicated conventional procedures. In our new model, an annual discrepancy in starting date of forcing effect from the pre-determined dates, calculated from a simplified procedure in our previous model, were calculated as the product of the correction coefficient, Ci, for each year and a winter temperature normal value above a threshold of chilling effect for endodormancy release, (TDJ-1.5), for each site. Annual correlation coefficient Ci was calculated from averaged winter temperatures at 7 secular observatories in warm region in Japan. Estimated first flowering dates with and without correction of starting dates were compared each other. The estimations by new method kept under about 3.2 days of RMSE. The introduction of new method also reduced RMSE within 3 days into approximately half of stations, applied to verification of new method with relatively long (50-year) period.
