# 温度変換日数法によるソメイヨシノの開花に関する気候学的研究 Climatological Studies on Blooming of Cherry Tree (Prunus yedoensis) by Means of DTS Method

## この論文にアクセスする

## この論文をさがす

## 著者

## 抄録

Blooming dates of Prunus yedoensis, the most popular species in cherry trees of Japan, have been recorded at many meteorological observatories since the late 19 th century. These records are useful for studies on evaluation of influence of climatic change to blooming at many places. Equations to estimate or predict blooming dates of P. yedoensis using temprature, were proposed by many workers. These equations gave accurate estimations and well expressed the relationship between both fluctuations in temperature and blooming dates. But, each of them were applicable only at or around the site for which equation was established. The primary objective of this study is to develop the generalized method for estimation of blooming date for P. yedoensis, which is applicable at any site of Japan. In order to estimate blooming date, the accumulation model of index refered as DTS (the number of days transformed to standard temperature) is used in the present study. This DTS accumulation model, or DTS method, has been introduced recently for the purpouse of prediction of blooming dates for some kinds of fruit species. In this study, a DTS accumulation model using air temperature data is developed applicable for many stations and to offer accurate estimation at each station. As the application of model, following are made: (1) A method to estimate blooming dates at places where phenological data are not available; (2) evaluation of the effect of urban warming to blooming dates; and (3) simulation of the impact of global warming on blooming date in Japan. Main results obtained in this study are summarized as follows. 1. Characteristics of distribution and fluctuation of blooming dates of P. yedoensis were examined. It was qualitatively suggested that interannual and geographical variations of blooming date were attributed to the temperature during developmental stage of flower buds, because the distribution and transition of springtime temperature considerably reflected in the distribution and density of isophen for blooming dates. Therefore, it was expected that blooming date of P. yedoensis might be well estimated by only taking account of temperature since rest break up to blooming. At warmer regions near the southern limit of range of P. yedoensis (i. e. Izu or Osumi Is.), however, it may be necessary to take account of interannual variation of transition of rest break, which makes blooming dates fluctuate largely. 2. The duration q, during which mean temperature affects the interannual variation of blooming date at eath site, and its mean strating point D_r, were determined preliminarily by means of linear regression analysis. In the warmer regions than Tohoku district, the temperature averaged over 35-50 days up to blooming date showed the highest correlation to blooming date, and those duration was determined as q. D_r, was determined from q as about 25 th day at the sites in Yamagata and Niigata prefectures, and the range of 40-55 th day at most of sites in other areas. 3. The DTS accumulation model to estimate blooming date of P. yedoensis was developed. DTS is the index which relatively expresses the velocities of growth or development of plants under a certain temperature condition, when that under a specific standard temperature (25℃ was used in this study, thouth it is found as not so proper value) is set as normal. Daily mean temperature data is transformed to DTS value by using Arrhenius' equation, and DTS distribution is expressed by exponential function of temperature. In order to estimate blooming date as accurate as possible, it is necessary to determine the adequate values of starting date of DTS accumulation and temperature characteristic E_a which is a parameter in the exponential function of temperature. The value of E_a for flower buds development was determined using cut shoots of P. yedoensis forced in greenhouse. As the result of analysis by the use of Arrhenius plots, 16.8 kcal mol^<-1> was obtained as mean value of E_a. Then, estimations of blooming dates for P. yedoensis at 38 meteorological stations in Japan were made. DTS values are accumulated from a stating date, the estimated blooming date is the date when DTS accumulations reach at mean accumulations of DTS on observed blooming date for the site. The adequate starting date D_s at each station was determined from error analysis. At many stations, D_s lies between 40 and 55 th day from 1 Jan. It was found that 17 kcal mol^<-1> for E_a gives minimum error when averaged over all stations. Using this value of E_a (17 kcal mol^<-1>) and starting date D_s, computation was made to estimate blooming dates between 1961 and 1985. It was shown that, at stations in Hokkaido and Tohoku districts, RMSE between the observed and estimated blooming dates fell within 1-2 days, and at stations in other areas RMSE ranged between 2 and 3 days, except for Hachijo I. where it became 6.64 days. 4. Large RMSE in warm regions seemed to be due to the large interannual variation of the process of rest break of flower bud in late winter. An adjustment of estimation of blooming date at southern sites was made by applying the concept of chill-unit which is the weighted hour corresponding to the effectiveness chilling to rest completion. It was shown that this procedure reduced RMSE for blooming dates from 6.64 days to 2.32 days at Hachijo I. Errors were reduced to about 2 days at all stations in warm regions. 5. The procedure of estimation mentioned above was rather complex, because the optimum starting date and the mean DTS accumulations for each site were necessary to be determined. The method of estimation can be considerably simplified by using the common DTS accumulation value (8 days) and the starting date computed by the multiple regression for the corresponding DTS accumulation values. This simplified method was tested at 20 meteorological observatories which were not used to determine common DTS accumulations. RMSE were within 1-3 days except for a few sites in Kanto and Tohoku districts. The method was applied to estimate blooming dates at the parks in western Japan noted for the cherry blossoms. It was shown that accuracies of 2-3 days of RMSE were obtained to the parks where daily mean temperature data were available within 4-5 km. 6. The influences of the urban warming on blooming dates at three large cities in Japan, was evaluated by means of the DTS method. The evaluation was made by comparison of estimated blooming dates using temperature from which urban warming effect was eliminated, with that using observed temperature. It was clear that quickenings of blooming dates attributed to urban warming had appeared since 1920's at Tokyo and since 1960's at Osaka. In the recent 30 years, this quickenings of blooming had increased at the rates of 0.15 days year^<-1> for Tokyo and 0.14 days year^<-1> for Osaka. In early 1980's, mean quickenings were evaluated as 7.8, 4.6 and 4.8 days at Tokyo, Osaka and Kyoto respectively. 7. An attempt to simulate the impact of global warming to distribution of blooming date in Japan was made by using the DTS method. For the 2℃ uniform warming, estimated mean blooming dates will be earlier than those at present by 10-12 days in Hokkaido and Tohoku districts, by 5-6 days in Kinki and Chugoku districts, and by 0-2 days in Kyushu district. This suggests that global warming of the order of 2℃ will decrease the difference in blooming dates between northern and southern regions of Japan.

Blooming dates of Prunus yedoensis, the most popular species in cherry trees of Japan, have been recorded at many meteorological observatories since the late 19 th century. These records are useful for studies on evaluation of influence of climatic change to blooming at many places. Equations to estimate or predict blooming dates of P. yedoensis using temprature, were proposed by many workers. These equations gave accurate estimations and well expressed the relationship between both fluctuations in temperature and blooming dates. But, each of them were applicable only at or around the site for which equation was established. The primary objective of this study is to develop the generalized method for estimation of blooming date for P. yedoensis, which is applicable at any site of Japan. In order to estimate blooming date, the accumulation model of index refered as DTS (the number of days transformed to standard temperature) is used in the present study. This DTS accumulation model, or DTS method, has been introduced recently for the purpouse of prediction of blooming dates for some kinds of fruit species. In this study, a DTS accumulation model using air temperature data is developed applicable for many stations and to offer accurate estimation at each station. As the application of model, following are made: (1) A method to estimate blooming dates at places where phenological data are not available; (2) evaluation of the effect of urban warming to blooming dates; and (3) simulation of the impact of global warming on blooming date in Japan. Main results obtained in this study are summarized as follows. 1. Characteristics of distribution and fluctuation of blooming dates of P. yedoensis were examined. It was qualitatively suggested that interannual and geographical variations of blooming date were attributed to the temperature during developmental stage of flower buds, because the distribution and transition of springtime temperature considerably reflected in the distribution and density of isophen for blooming dates. Therefore, it was expected that blooming date of P. yedoensis might be well estimated by only taking account of temperature since rest break up to blooming. At warmer regions near the southern limit of range of P. yedoensis (i. e. Izu or Osumi Is.), however, it may be necessary to take account of interannual variation of transition of rest break, which makes blooming dates fluctuate largely. 2. The duration q, during which mean temperature affects the interannual variation of blooming date at eath site, and its mean strating point D_r, were determined preliminarily by means of linear regression analysis. In the warmer regions than Tohoku district, the temperature averaged over 35-50 days up to blooming date showed the highest correlation to blooming date, and those duration was determined as q. D_r, was determined from q as about 25 th day at the sites in Yamagata and Niigata prefectures, and the range of 40-55 th day at most of sites in other areas. 3. The DTS accumulation model to estimate blooming date of P. yedoensis was developed. DTS is the index which relatively expresses the velocities of growth or development of plants under a certain temperature condition, when that under a specific standard temperature (25℃ was used in this study, thouth it is found as not so proper value) is set as normal. Daily mean temperature data is transformed to DTS value by using Arrhenius' equation, and DTS distribution is expressed by exponential function of temperature. In order to estimate blooming date as accurate as possible, it is necessary to determine the adequate values of starting date of DTS accumulation and temperature characteristic E_a which is a parameter in the exponential function of temperature. The value of E_a for flower buds development was determined using cut shoots of P. yedoensis forced in greenhouse. As the result of analysis by the use of Arrhenius plots, 16.8 kcal mol^<-1> was obtained as mean value of E_a. Then, estimations of blooming dates for P. yedoensis at 38 meteorological stations in Japan were made. DTS values are accumulated from a stating date, the estimated blooming date is the date when DTS accumulations reach at mean accumulations of DTS on observed blooming date for the site. The adequate starting date D_s at each station was determined from error analysis. At many stations, D_s lies between 40 and 55 th day from 1 Jan. It was found that 17 kcal mol^<-1> for E_a gives minimum error when averaged over all stations. Using this value of E_a (17 kcal mol^<-1>) and starting date D_s, computation was made to estimate blooming dates between 1961 and 1985. It was shown that, at stations in Hokkaido and Tohoku districts, RMSE between the observed and estimated blooming dates fell within 1-2 days, and at stations in other areas RMSE ranged between 2 and 3 days, except for Hachijo I. where it became 6.64 days. 4. Large RMSE in warm regions seemed to be due to the large interannual variation of the process of rest break of flower bud in late winter. An adjustment of estimation of blooming date at southern sites was made by applying the concept of chill-unit which is the weighted hour corresponding to the effectiveness chilling to rest completion. It was shown that this procedure reduced RMSE for blooming dates from 6.64 days to 2.32 days at Hachijo I. Errors were reduced to about 2 days at all stations in warm regions. 5. The procedure of estimation mentioned above was rather complex, because the optimum starting date and the mean DTS accumulations for each site were necessary to be determined. The method of estimation can be considerably simplified by using the common DTS accumulation value (8 days) and the starting date computed by the multiple regression for the corresponding DTS accumulation values. This simplified method was tested at 20 meteorological observatories which were not used to determine common DTS accumulations. RMSE were within 1-3 days except for a few sites in Kanto and Tohoku districts. The method was applied to estimate blooming dates at the parks in western Japan noted for the cherry blossoms. It was shown that accuracies of 2-3 days of RMSE were obtained to the parks where daily mean temperature data were available within 4-5 km. 6. The influences of the urban warming on blooming dates at three large cities in Japan, was evaluated by means of the DTS method. The evaluation was made by comparison of estimated blooming dates using temperature from which urban warming effect was eliminated, with that using observed temperature. It was clear that quickenings of blooming dates attributed to urban warming had appeared since 1920's at Tokyo and since 1960's at Osaka. In the recent 30 years, this quickenings of blooming had increased at the rates of 0.15 days year^<-1> for Tokyo and 0.14 days year^<-1> for Osaka. In early 1980's, mean quickenings were evaluated as 7.8, 4.6 and 4.8 days at Tokyo, Osaka and Kyoto respectively. 7. An attempt to simulate the impact of global warming to distribution of blooming date in Japan was made by using the DTS method. For the 2℃ uniform warming, estimated mean blooming dates will be earlier than those at present by 10-12 days in Hokkaido and Tohoku districts, by 5-6 days in Kinki and Chugoku districts, and by 0-2 days in Kyushu district. This suggests that global warming of the order of 2℃ will decrease the difference in blooming dates between northern and southern regions of Japan.