摘要
荔枝病虫害易造成荔枝Litchi chinensis Sonn.产量少、品质劣、价格低,导致果农收益下跌,产业发展受制。因此,建立荔枝病虫害预测预报模型,开展病虫害发生程度预测,对荔枝产业高质量发展具有重要意义。基于2004—2020年广州市增城区第二代荔枝蒂蛀虫驻果率数据,以及前期气象要素数据,对影响蒂蛀虫驻果率的气象因子进行灰色关联分析,筛选出主要影响因子,采用多变量灰色预测模型GM(1,N)构建第二代蒂蛀虫驻果率预测模型。通过后验差以及小概率误差检验方法进行模型精度检验,并结合马尔科夫模型进行优化,以提高预测模型精度。研究结果表明,采用多变量灰色预测模型GM(1,N)进行荔枝蒂蛀虫驻果率预测,预测精度较高,可为荔枝病虫害预防提供决策参考。
Litchi diseases and insect pests easily cause litchi yield,poor quality,and low price,resulting in the decline of fruit farmers’income and industrial development.Therefore,it is of great significance for the high-quality development of the lychee industry to establish a prediction model of lychee pests and diseases and to predict the occurrence degree of lychee pests and diseases.Based on the data of the second generation of lychee borer infestations in the Zengcheng District of Guangzhou City from 2004 to 2020 and the data of previous meteorological factors,this paper conducted a gray correlation analysis on the meteorological factors affecting the infestation rate of lychee borer infestations and selected the main influencing factors.The multivariate gray prediction model GM(1,N)was used to construct the second-generation prediction model of lychee borer infestation rate.A posteriori error test and a small probability error test method were used to test the accuracy of the model,and the Markov model was optimized to improve the accuracy of the prediction model.The results showed that the multivariable gray prediction model GM(1,N)was used to predict the infestation rate of the lychee root borer,which had a high prediction accuracy and could provide a reference for decision-making regarding lychee pest prevention.
作者
欧善国
张桂香
OU Shanguo;ZHANG Guixiang(Guangzhou Meteorological Bureau,Guangzhou,Guangdong 511430,China)
出处
《热带农业科学》
2023年第11期79-86,共8页
Chinese Journal of Tropical Agriculture
基金
荔枝病虫害智能气象服务技术研究(No.GRMC2020M19)
广州智慧农业气象服务平台项目(No.[GZIT2020-B4-552])。
关键词
荔枝
蒂蛀虫
灰色关联度
多变量灰色预测模型
马尔科夫模型
Litchi chinensis Sonn.
root borer
gray correlation degree
multivariate gray prediction model
Markov model
