Abstract: The growth of apple is greatly affected by meteorological factors. The increasing frequency of high and low temperature disasters in Shaanxi province can greatly influence the yield and quality of apple. Phenological period is helpful to guide fruit production and manage disaster risk. At present, phenological data are very scarce. The historical phenological period can be reconstructed by constructing phenological model. There were four main apple producing areas in Shaanxi province. Two representative stations with comprehensive phenological data were selected for each area. These representative stations were Liquan and Fengxiang (Guanzhong area), Xunyi and Changwu(Western area of Weibei), Tongchuan and Baishui (Eastern area of Weibei), Yanchang and Luochuan(Yan'an area). Firstly, the station with longer historical phenological data record among the two representative stations was selected. Then, the relationship between day of year(DOY) of phenological period and meteorological indexes was analyzed by using SPSS. Several single-index or multiple-index phenological prediction models were established through stepwise regression analysis. Finally, the optimal model was selected by back substitution test and prediction test. The test results were evaluated by the mean absolute error (MAE) and the relative accuracy (RA) with a difference of 0−3 days between the simulated value and the observed value. Considered both and chose the optimal model. The results showed that (1) the mean absolute errors (MAE) of sprouting time models were 0.8−2.4 days and the RA were 84.6%−100%. The MAE of flowering date models were 2.5−3.4 days and the RA were 55.6%−75%. The MAE of fruit development period models were 0.9−2.8 days and the RA were 63.2%−100%. The MAE of harvest period models were 2.2−3.2 days and the RA were 69.2%−72.2%. The sprouting time models had the best simulation effect. The fruit development period models were the second. The harvest period models were the third. The flowering date model had the worst simulation effect. (2) From 1981 to 2019, the reconstructed DOY of sprouting time in Yan'an area, Eastern area of Weibei, Western area of Weibei and Guanzhong area were between 72−98, 70−88, 73−98 and 71−85. And the reconstructed DOY of flowering date in these areas were between 102−116, 86−107, 100−125 and 84−115. The reconstructed DOY of fruit development period in Yan'an area, Eastern area of Weibei and Guanzhong area were between 114−122, 89−118 and 87−117. And the reconstructed DOY of harvest period in these three areas were between 260−301, 276−297 and 224−348. (3) The spatial distribution of reconstructed phenological periods showed that: from 1981 to 2019, the sprouting time of Yan'an area and Eastern area of Weibei was gradually delayed from southeast to northwest, while that of Guanzhong area and western area of Weibei was delayed from west to east. The flowering date and fruit development period were both later and later in north of whole study area. The harvest period of Yan'an area and eastern area of Weibei was delayed from east to west , while that of Guanzhong area and western area of Weibei was gradually delayed from west to east. The simulation effect of these models was generally good. The reconstructed apple phenological data can provide basic support for apple production management and disaster risk prevention. This study can be used as a reference for the development of fruit tree phenological period model.

Key words: Shaanxi apple, Sprouting time, Flowering date, Fruit development period, Harvest period, Phenological period model