Characteristics and forecast of flowering duration of Cherry Blossoms in Wuhan University

Japanese cherry blossoms of Wuhan University is a city card of Wuhan. It can provide a reasonable reference for the management of tourism department and the arrangement of travel time for tourists to carry out the prediction of the length of cherry blossom period. A 40-year-old(1979-2018) dataset of flowering dates of Japanese cherry blossoms at Wuhan University Campus associated with the meteorological data was used for developing a method of forecasting the flowering duration of cherry blossoms.(1) The first-flowering and falling flowering date of cherry blossoms were obviously advanced in the 1980 s--1990 s. Since the end of 1990 s, the trend of initial flowering and falling flower period is not obvious, but the rate of variation is large, which coincides with the stagnation period of global climate change.In the past 40 years, the change rate of florescence length was very large, but there was no obvious trend of increase or decrease.(2) The average first-flowering date was March 14-15, the falling flowering date was March 31 to April 1, and the average flowering period was 18 days.(3) The length of flowering period was negatively correlated with the first-flowering date, and negatively correlated with the average temperature, the highest average and the lowest average temperature during flowering period. It′s also negatively correlated with the average daily temperature difference, and positively correlated with the total precipitation during florescence. There was no significant correlation with the average maximum wind speed, average precipitation and sunshine hours during flowering period.(4) Single factor fitting, multi-factor regression and principal component analysis model of cherry blossom period length were established based on the data from 1979 to 2015, which were tested by the data from 2016 to 2018. These models were used to predict the length of cherry blossom period in Wuhan University and good experimental results were obtained. Among them, principal component regression model, precipitation single factor fitting model, and multi-factor cherry blossom period length regression model are the best, with an average absolute error of about 1.5 days. In the future, the prediction model will be applied to the actual cherry blossom period prediction.