Journal of Agricultural Science and Technology ›› 2020, Vol. 22 ›› Issue (10): 93-100.DOI: 10.13304/j.nykjdb.2019.0186

Previous Articles     Next Articles

Prediction of Apple Initial Flowering Period Based on Machine Learning

ZHANG Xingwei, CHEN Chao*, TIAN Shan, FU Lin   

  1. Chengdu University of Information Technology, Chengdu 610225, China
  • Received:2019-03-18 Online:2020-10-15 Published:2019-04-18

基于机器学习的苹果始花期预测

张兴伟,陈超*,田姗,付琳   

  1. 成都信息工程大学, 成都 610225
  • 通讯作者: *通信作者 陈超 E-mail:RoosterChen@163.com
  • 作者简介:张兴伟 E-mail:zxw-y@qq.com;
  • 基金资助:
    成都市科技局项目(2018-YF05-01217-SN)。

Abstract: The early and late of theflowering period is the cumulative effect of meteorological factors on fruit trees during growth. Dormancy of winter fruit tree is a process that requires cooling to heat, so this paper studied the effect of meteorological factors on the flowering period of apple trees and predicted the initial flowering period. Based on the phenological data of apples and the meteorological data in Ji County, Shanxi Province,  the effects of meteorological factors such as air temperature, humidity, ground temperature, precipitation and sunshine hours on the flowering period were determined to establish a prediction model for the initial flowering period of apple fruit trees at the three periods of whether frost damage occurs, whether it can overwinter normally, and heat and moisture requirements. by multiple linear regression methods and combined methods. The results showed that the coefficient of determination of the predicted value and the true value in three different time periods were 0.59, 0.71 and 0.48, respectively. Due to the different length of analysis days, the model based on the period of whether normally overwinter was the best with the coefficient of determination was 0.71, and the analysis with shortest days had the smallest coefficient of determination for the time of heat and moisture requirements, which indicated that the selection of analysis days should not be too small. The coefficient of determination of the combined method model was 0.78, which was a certain improvement over the 0.71 model based on whether it can overwinter normally. At the same time, the forecasting model could complete the accurate forecast of at least 24 d in advance on March 15.

Key words: prediction of initial flowering period, correlation analysis, coefficient of determination, combination method

摘要: 始花期的早晚是生长过程中气象因子累积对果树产生的影响。冬季果树休眠是一个需冷量到需热量的过程,研究了这一过程中气象因子对苹果果树始花期的影响并预测始花期。基于山西省吉县苹果物候数据和气象数据,研究三个时间段内(是否发生冻害、能否正常越冬和热量与水分需求)气温、湿度、地温、降水量和光照等气象因子对花期的影响程度,采用多元线性回归方法和组合方法,建立山西省吉县苹果果树始花期的预测模型。结果表明,在三个不同时间段内预测值与真实值的决定系数分别为:0.59、0.71和0.48。由于分析天数长短的不同,基于能否正常越冬建立的模型效果最好,其决定系数为0.71,分析天数最短的基于热量与水分需求时间段的模型决定系数最小,这表明预测过程中分析天数的选择不宜过小。使用组合方法模型的决定系数为0.78,比能否正常越冬时间段模型的0.71略有提升。同时预报模型可以在3月15日完成提前量至少为24 d的精准预测。

关键词: 始花期预测, 相关性分析, 决定系数, 组合方法