img

官方微信

高级检索

中国沙漠 ›› 2025, Vol. 45 ›› Issue (1): 292-303.DOI: 10.7522/j.issn.1000-694X.2024.00184

• • 上一篇    

基于随机森林的局地起沙量预测评估模型

李彬1(), 孙小龙1(), 赵悦晨2, 江琪3, 卢士庆1, 唐家琦1   

  1. 1.内蒙古自治区生态与农业气象中心,内蒙古 呼和浩特 010051
    2.内蒙古自治区气候中心,内蒙古 呼和浩特 010051
    3.国家气象中心,北京 100000
  • 收稿日期:2024-11-20 修回日期:2024-12-27 出版日期:2025-01-20 发布日期:2025-01-13
  • 通讯作者: 孙小龙
  • 作者简介:孙小龙(E-mail: 15632470343@163.com
    李彬(1989—),男,河北承德人,硕士,工程师,主要研究方向为生态环境及天气遥感。E-mail: 983229508@qq.com
  • 基金资助:
    中国气象局风云卫星应用先行计划项目(FY-APP-ZX-2023.01);内蒙古自然科学基金项目(2024QN04019);高分气象行业应用示范系统(二期)课题“局地低能见度天气事件监测预警示范”;内蒙古自治区气象局科技创新项目(nmqxkjcx202320)

Prediction model of dust mass generation in dust source by random forest

Bin Li1(), Xiaolong Sun1(), Yuechen Zhao2, Qi Jiang3, Shiqing Lu1, Jiaqi Tang1   

  1. 1.Inner Mongolia Eco- and Agro-Meteorological Center,Hohhot 010051,China
    2.Inner Mongolia Meteorological Center,Hohhot 010051,China
    3.National Meteorological Center,Beijing 100000,China
  • Received:2024-11-20 Revised:2024-12-27 Online:2025-01-20 Published:2025-01-13
  • Contact: Xiaolong Sun

摘要:

沙尘天气发生源头一直是社会关注焦点,而蒙古国及中国境内沙尘源地起沙量及贡献尚不十分明确。利用2019—2024年的典型沙尘过程遥感监测产品,考虑沙尘发生阶段起沙量对应的气象要素、地表状况信息,采用随机森林机器学习方法确定起沙通量与气象要素、沙源遥感监测状况间的响应关系进而建立模型,并利用模式或实况数据等快速评估源区的起沙发展趋势和沙尘强度。在此基础上通过对多时相遥感沙尘判识结果精细化估计沙源地,分析了4次典型沙尘过程的起沙预测效果,并进行了模型的不确定性分析。结果表明:(1)利用监测数据对模型起沙量进行散点拟合验证,R2为0.84,平均绝对误差(MAD)为25.2 kg·km-2·min-1。(2)模型可预测局地起沙影响并对不同沙尘源的起沙量贡献提供定量化预测。

关键词: 起沙量, 沙尘监测, 随机森林, 遥感, 沙源地, FY-4

Abstract:

The source of dust has always been the focus of social attention, but the amount and contribution of dust between Mongolia and the China are not so clear. In this paper, typical dust monitoring product by remote sensing from 2019 to 2024 were used to establish the response relationship between dust flux, meteorological elements and sand sources, which considered the meteorological elements and surface condition corresponding to the mass of dust generated in the generation stage of dust weather. The trend of dust generation and intensity by the source are quickly evaluated by using model or real data. On this basis, the results by multi-temporal remote sensing are used to estimate the dust source, and the application effects of four typical dust processes were analyzed, as well as, the uncertainty of the model. The results showed that: (1) The scatter-point fitting of the monitoring data was verified with R2=0.84 and the mean absolute error (MAD)=25.2 kg·km-2·min-1. (2) The model could predict the impact of local dust release and provide quantitative assessment of the contribution from different dust sources, which could provide scientific basis for desertification control.

Key words: dust mass, dust monitoring, random forest, remote sensing, dust source, FY-4

中图分类号: