四种检验方法在山东暴雪预报中的应用对比
作者:
作者单位:

作者简介:

李建明,男,高级工程师,主要从事信息技术与预报检验开发工作,350089739@qq.com。

通讯作者:

中图分类号:

P456;P457.6

基金项目:

山东省气象局科研项目(2021SDYD41)


请扫码阅读

Comparative analysis on application of 4 verification methods for snowstorm forecasting in Shandong
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    利用山东122个国家级地面气象观测站2017—2021年中1—3月和11—12月“24 h降雪量p≥10.0 mm”的暴雪实况资料,采用二分类法、邻域空间检验法(以下简称“邻域法”)、时间偏移法和量级模糊法等4种方法对山东暴雪预报进行检验与对比。结果表明:(1)山东暴雪具有明显的时空分布特征,暴雪主要出现在半岛北部地区,鲁东南和半岛南部产生暴雪的概率最小;暴雪出现次数的年际变化和月际变化较大,最多年份出现98次,最少年份仅有5次,2月是高发月,占全年暴雪总次数的38.5%。(2)现行业务中应用最广泛的二分类法检验的预报命中率较低,其中24 h预报命中率仅为12.08%,主要原因是该方法受到空间、时间和量级的多重影响,不能精细准确地反映预报能力。(3)邻域法、时间偏移法和量级模糊法对24 h的暴雪预报命中率分别为14.40%、14.69%和30.05%;相较于二分类法,这3种检验方法的预报命中率均有较大幅度提高,空报率和漏报率均有较大幅度下降。(4)融合邻域法、时间偏移法和量级模糊法的综合检验法,能从空间、时间和量级3个维度区分出预报差异,检验结果更加精细准确,有利于引导预报员放下“检验评分低”的思想包袱,做出更加科学客观的预报,进一步提升预报服务效果。

    Abstract:

    Using the actual snowstorm data (24-h snowfall p≥10.0 mn) at 122 national meteorological stations in Shandong from January to March and from November to December between 2017 and 2021, 4 methods are compared in the evaluation of snowstorm forecasting, including binary classification method, neighborhood spatial verification method, time migration method, and magnitude fuzzing method. The results are as follows. (1) Snowstorms in Shandong have obvious spatial and temporal distribution characteristics, and mainly appear in the north of Shandong Peninsula, while the probability of snowstorms occurring in the southeast of Shandong Province and the south of Shandong Peninsula is the lowest. The interannual and intermonthly variations of the occurrence of snowstorms are significant. The largest number of snowstorms that occur within one year is 98 and the smallest number is only 5; February is the peak month for the occurrence of snowstorms, accounting for 38.5% of the total number of snowstorms annually. (2) The probability of detection (POD) computed by the most widely used binary classification method at present is relatively low, with the POD of 24-h snowstorm forecasting only 12.08%. The main reason is that this method is affected by multiple factors, such as space, time, and magnitude, so it cannot accurately reflect the prediction ability. (3) The PODs of the neighborhood spatial verification method, time migration method, and magnitude fuzzing method for 24-h snowstorm forecasting are 14.40%, 14.69%, and 30.05%, respectively. Compared with the binary classification method, the PODs of these three methods significantly increase, while the false alarm rates and missing rates significantly decrease. (4) A comprehensive verification method that integrates the neighborhood spatial verification method, time migration method, and magnitude fuzzing method can distinguish the forecasting differences from three dimensions, namely space, time, and magnitude. The more precise and accurate verification results are conducive to guiding forecasters to put down the burden of low verification scores and make scientific and objective predictions, improving forecasting services.

    参考文献
    相似文献
    引证文献
引用本文

李建明,于怀征,韩永清,冀玉超,马晓睿.四种检验方法在山东暴雪预报中的应用对比[J].海洋气象学报,2024,(2):90-97.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
文章历史
  • 收稿日期:2023-12-18
  • 最后修改日期:2024-01-25
  • 录用日期:
  • 在线发布日期: 2024-06-07
  • 出版日期:

“中法海洋卫星”专刊征稿

关闭