欢迎访问《中国农学通报》,

中国农学通报 ›› 2025, Vol. 41 ›› Issue (14): 111-119.doi: 10.11924/j.issn.1000-6850.casb2024-0526

• 资源·环境·生态·土壤·气象 • 上一篇    下一篇

库尔勒香梨物候期特征及花期预测

霍锦1(), 昝永黎2, 刁鹏1, 吴新国2, 魏艳英1, 龚美玲1   

  1. 1 巴州气象局,新疆库尔勒 841000
    2 库尔勒市气象局,新疆库尔勒 841000
  • 收稿日期:2024-08-14 修回日期:2025-04-17 出版日期:2025-05-15 发布日期:2025-05-14
  • 作者简介:

    霍锦,女,1982年出生,高级工程师,本科,主要从事农业气象服务及应用气象方面的研究。通信地址:841000 新疆库尔勒市人民西路102号,Tel:0996-2161992,E-mail:

  • 基金资助:
    新疆巴音郭楞蒙古自治州气象局创新团队专项“巴州特色林果气象服务创新团队项目”(CX202402); 新疆维吾尔自治区气象局创新发展专项重点项目“中亚基础气象数据产品研发及其应用”(ZD202304)

Characteristics of Phenological Period and Prediction of Flowering Period in Korla Fragrant Pear

HUO Jin1(), ZAN Yongli2, DIAO Peng1, WU Xinguo2, WEI Yanying1, GONG Meiling1   

  1. 1 Bazhou Meteorological Bureau, Korla, Xinjiang 841000
    2 Korla Meteorological Bureau, Korla, Xinjiang 841000
  • Received:2024-08-14 Revised:2025-04-17 Published:2025-05-15 Online:2025-05-14

摘要:

为精准掌握库尔勒香梨生长周期,优化栽培管理,开展了库尔勒香梨物候期研究和预测。利用库尔勒市1981—2023年气象观测数据和库尔勒香梨物候观测资料,通过趋势分析、相关性分析、显著性检验等方法,研究库尔勒香梨物候期的演变趋势和主要气象因子对物候期的影响,并采用机器学习中随机森林(Random Forest, RF)和长短期记忆网络(Long Short-Term Memory, LSTM) 两种算法对开花期进行预测。结果表明,春季物候期中展叶盛期以0.29 d/a的速率推后,其余表现为提前趋势,提前速率为0.12~0.37 d/a,其中花芽开放期提前量最大,开花盛期最小。秋季物候期中果实成熟期以0.14 d/a的速率提前,叶变色期和落叶期以0.03~0.15 d/a的速率推后。生长季整体以0.14 d/a的速率延长,2012年以后延长较为明显。温度是影响物候期的主要气象因子,对春季物候期影响程度大于秋季物候期,日照时数、降水量也会对物候期有影响。RF预测始花期优于LSTM,LSTM预测盛花期好于RF,两种方法预测末花期表现一般。研究以期为库尔勒香梨的优化栽培管理、农产品提质增效、防灾减灾等提供科学依据。

关键词: 库尔勒香梨, 物候期, 相关分析, 随机森林, 长短期记忆网络, 花期预测

Abstract:

It is of great significance to study and forecast the phenological period of Korla fragrant pear in order to accurately grasp its growth cycle, optimize cultivation management, and improve yield and quality. The phenological evolution trend of Korla fragrant pear and the influence of main meteorological factors on the phenological period were analyzed by means of trend analysis, correlation analysis and significance test. Two machine learning algorithms, random forest (RF) and long short-term memory (LSTM), were used to predict the flowering period. The results showed that, in spring phenology, the peak stage of leaf development was delayed at a rate of 0.29 d/a, and the rest showed an advance trend, with an advance rate of 0.12-0.37 d/a, in which the advance of flower bud opening stage was the largest, and that of flowering peak stage was the least. In the autumn phenophase, the fruit ripening stage was advanced by 0.14 d/a, and the leaf discoloration stage and defoliation stage were delayed by 0.03-0.15 d/a. The growing season extended at a rate of 0.14 d/a, and the elongation was more obvious after 2012. Temperature is the main meteorological factor affecting the phenological period, and its influence on the spring phenological period is greater than that on the autumn phenological period. Sunshine hours and precipitation will also affect the phenological period. RF is better than LSTM in predicting the initial flowering period, and LSTM is better than RF in predicting the full flowering period. The research is expected to provide scientific basis for the optimization of cultivation management, quality and efficiency improvement, disaster prevention and mitigation of Korla fragrant pear.

Key words: Korla fragrant pear, phenophase, correlation analysis, random forest, long short-term memory, prediction of flowering period