欢迎访问《农学学报》,

农学学报 ›› 2024, Vol. 14 ›› Issue (9): 62-68.doi: 10.11923/j.issn.2095-4050.cjas2023-0221

• 农业信息 农业气象 • 上一篇    下一篇

基于马尔可夫模型的油菜花期低温阴雨灾害预测

赵艺1(), 王鑫1, 郭翔1, 常俊2, 陈东东1, 杨德胜1()   

  1. 1 四川省农业气象中心,成都 610072
    2 中国气象局气象干部培训学院四川分院,成都 610072
  • 收稿日期:2023-10-12 修回日期:2024-04-18 出版日期:2024-09-18 发布日期:2024-09-18
  • 通讯作者:
    杨德胜,男,1974年出生,四川人,高工,本科,主要从事农业气象服务与业务应用方面的研究。通信地址:610072 四川成都青羊区光华村街20号 四川省气象局,Tel:028-87364604,E-mail:
  • 作者简介:

    赵艺,女,1990年出生,工程师,硕士研究生,主要从事农业气象灾害与产量预报方面的研究。通信地址:610072 四川成都青羊区光华村街20号 四川省气象局,Tel:028-87320665,E-mail:

  • 基金资助:
    高原与盆地暴雨旱涝灾害四川省重点实验室科技发展基金项目(SCQXKJQN202124); 高原与盆地暴雨旱涝灾害四川省重点实验室科技发展基金项目(省重实验室-2018-重点-05-03); 国家自然科学基金(42105153)

Prediction of Low Temperature and Overcast Rain Disaster in Rape Flowering Period Based on Markov Model

ZHAO Yi1(), WANG Xin1, GUO Xiang1, CHANG Jun2, CHEN Dongdong1, YANG Desheng1()   

  1. 1 Agro-meteorological Center of Sichuan Province, Chengdu 610072, Sichuan, China
    2 Sichuan Branch, China Meteorological Administration Training Centre, Chengdu 610072, Sichuan, China
  • Received:2023-10-12 Revised:2024-04-18 Online:2024-09-18 Published:2024-09-18

摘要:

花期低温阴雨是四川盆区油菜的主要气象灾害之一,本研究旨在通过预测四川盆区油菜花期低温阴雨灾害,为防灾减灾提供科学依据。本研究基于1961—2020年四川盆区101个站点油菜花期低温阴雨灾害损失评估结果,依据灾损率,将各站点60 a灾害序列划分为5个状态;利用马氏性检验筛选站点序列,选取满足预测条件的序列;对通过马氏性检验的站点序列,建立叠加马尔可夫链、加权马尔可夫链、改进叠加马尔可夫链、改进加权马尔可夫链模型预测油菜花期低温阴雨灾害,并对结果进行回代和预测检验。结果表明,4种模型均有一定的预测能力,其中改进后的模型预测总正确率较改进前有明显提高,且各等级灾害的预测正确率分布较改进前更加均匀,表明改进后的马尔可夫模型具有更好的预测效果。

关键词: 四川盆区, 油菜, 低温阴雨, 灾害预测, 马尔可夫模型

Abstract:

Low temperature and overcast rain weather during flowering period is one of the main meteorological disasters of rape in Sichuan Basin. Predicting and studying the low temperature and overcast rain weather can provide scientific basis for disaster prevention and reduction of rape. This study was based on the assessment results of the disaster losses of low temperature and overcast rain during rape flowering period at 101 stations in Sichuan Basin from 1961 to 2020. We divided the 60-year sequence into 5 states based on the disaster loss rate, and used Markov test to screen site sequences and select sequences which satisfied the prediction conditions. We set up four models of superposed Markov chain, weighted Markov chain, improved superposed Markov chain, and improved weighted Markov chain to predict the low temperature and overcast rain disaster during the flowering period of rape based on the station sequence passing the Markov test, and performed backtracking and verification on the predicted results. All four Markov models had certain predictive ability, and the improved model had a significant improvement in overall accuracy compared to before, and the distribution of prediction accuracy for various levels of disasters was more uniform than before. In conclusion, the improved Markov model has better predictive performance.

Key words: Sichuan Basin, rape, low temperature and overcast rain, disaster prediction, Markov model