中国农业气象 ›› 2021, Vol. 42 ›› Issue (11): 929-938.doi: 10.3969/j.issn.1000-6362.2021.11.004

• 农业生物气象栏目 • 上一篇    下一篇

基于花前物候利用灰色关联分析法建立油菜花期预报模型

冯敏玉, 孔萍, 胡萍, 陈晓磊, 吴风雨, 廖南京   

  1. 1.江西省南昌市气象局,南昌 330038;2.江西省生态气象中心,南昌 330096;3. 江西省南昌县气象局,南昌 330200;4.安义县气象局,安义 330500;5.进贤县气象局,进贤 331700
  • 收稿日期:2021-03-02 出版日期:2021-11-20 发布日期:2021-11-15
  • 通讯作者: 孔萍,高级工程师,主要从事气候与农业气象研究,E-mail:84193734@qq.com E-mail:84193734@qq.com
  • 作者简介:冯敏玉,E-mail: fmy3893@163. com
  • 基金资助:
    南昌市农业气象重点实验室开放研究基金项目(2019NNZS204)

Prediction Model of Flowering Date of Rape Established by Using Grey Relational Analysis Method Based on Pre-flowering Phenology

FENG Min-yu,KONG Ping, HU Ping, CHEN Xiao-lei, WU Feng-yu, LIAO Nan-jing   

  1. 1.Meteorological Bureau of Nanchang, Nanchang 330038, China;2. Jiangxi Eco-meteorological Center, Nanchang 330096;3.Meteorological Bureau of Nanchang County,Nanchang 330200;4. Meteorological Bureau of Anyi County, Anyi 330500; 5. Meteorological Bureau of Jinxian County, Jinxian 331700
  • Received:2021-03-02 Online:2021-11-20 Published:2021-11-15

摘要: 采用相关分析法确定与油菜始花期显著相关的冬季气候因子和用灰色关联分析法确定与始花期关联最大的花前物候期因子,分别建立多元回归线性方程,并进行回代检验,以探索简便易操作的油菜始花期预测方法。利用均方根误差(RMSE)和相对误差(RE)对模型的模拟效果进行评价。结果表明:(1)与油菜始花期显著相关的冬季气象因子为1月平均最低气温、2月平均最低气温和2月日照时数,相关系数分别为−0.404、−0.556和−0.478。三个自变量因子不存在共线性关系,建立的回归模型具有统计学意义且通过显著性检验。(2)油菜花前各物候期以抽薹期和现蕾期与始花期关联度大,相关系数分别为0.656和0.634。建立的回归模型同样具有统计学意义并通过显著性检验。(3)分别对两种方法建立的模型进行检验与评价,回代检验表明两种方法建立的模型拟合精度总体上较接近。基于气候因子的模型RMSE气候因子为7.16,RE气候因子为11.2%;基于物候因子的模型RMSE物候因子为6.50,RE物候因子为3.87%。皮尔逊相关分析表明,实际值与两种方法拟合值的相关系数R物候因子和R气候因子分别为0.738和0.658,均通过了0.01水平的显著性检验。R物候因子>R气候因子,综合各项指标分析认为,灰色关联分析法建立的模型预测油菜始花期比利用气候因子建立的模型更可靠。

关键词: 油菜, 花期预报模型, 相关分析法, 灰色关联分析法, 物候期

Abstract: In order to explore a simple and easy method to predict the initial flowering stage of rape, the correlation analysis method was used in this paper to determine the winter climate factors significantly related to the first flowering period, and the gray correlation analysis method was also used to determine the pre flowering phenology factors most related to the first flowering period. Then the multiple regression linear equations were established and back substitution test was carried out, and finally the root mean square error (RMSE) and relative error (RE) models were used to evaluate the simulated and measured values. The results showed that: (1) the winter meteorological factors significantly related to the first flowering period were the average minimum temperature in January, the average minimum temperature in February and the sunshine hours in February, and their correlation coefficients were −0.404, −0.556, −0.478, respectively. There was no collinearity between the three independent variables. The regression model was statistically significant and passed the significance test. (2) Among the pre-anthesis phenological stages, there is a high correlation between the sprouting stage, budding stage and the initial flowering stage; and their correlation coefficients were 0.656 and 0.634, respectively. The regression model also showed statistical significance and passed the significance test. (3) The models established by the two methods are tested and evaluated. The back substitution test shows that the fitting accuracy of the models established by the two methods is relatively close on the whole. The RMSE climate factor based on climate factor is 7.16, and the RE climate factor is 11.2%; The phenological factors based on RMSE and RE were 6.50% and 3.87%, respectively. Pearson correlation analysis showed that the correlation coefficients of R phenological factor and R climatic factor were 0.738 and 0.658 respectively, which passed the significance test of 0.01 level. Among them, R phenological factor is higher than R climatic factor. Based on comprehensive analysis of various indicators, the model established by grey correlation analysis is more reliable than the model established by climate factors can be drawn.

Key words: Rape, Flowering prediction model, Correlation analysis method, Grey correlation analysis method, Phenology