干旱气象

• 技术发展与应用 • 上一篇    下一篇

基于支持向量机的云量精细化预报研究


 赵文婧1赵中军2汪结华2尚可政1 王式功1柳志慧1 ,孔德兵3 苏俊礼4   

  1. 1.兰州大学大气科学学院,甘肃兰州730000;2. 中国人民解放军92493部队6分队,辽宁葫芦岛125000;
    3.中国人民解放军95606部队气象台,重庆402361;4.中国人民解放军95994部队气象台,甘肃酒泉735000
  • 出版日期:2016-06-30 发布日期:2016-06-30
  • 通讯作者: 尚可政(1960-),男,甘肃景泰人,博士,主要从事干旱气候和现代天气预报技术和方法研究. E-mail:shangkz@lzu.edu.cn
  • 作者简介:赵文婧(1990-),女, 汉族,甘肃天水人,硕士研究生,研究方向为现代天气预报技术和极端天气气候. E-mail:mmwenjing@163.com
  • 基金资助:

    国家公益性(气象)行业专项项目(GYHY201206004)、甘肃省国际科技合作计划项目(1204WCGA016)和兰州大学中央高校基本科研业务费专项(lzujbky-2013-m03)共同资助

A Study on Refined Forecast of Cloud Cover Based on Support Vector Machine

ZHAO Wenjing1, ZHAO Zhongjun2, WANG Jiehua2, SHANG Kezheng1,WANG Shigong1, LIU Zhihui1, KONG Debing3, SU Junli4   

  1. 1. College of Atmospheric Science, Lanzhou University, Lanzhou 730000, China;
    2. The Sixth Element of 92493 Unit of the Chinese People’s Liberation Army, Huludao 125000, China;
    3. Meteorological Observatory of 95606 Unit of the Chinese People’s Liberation Army,
     Chongqing 402361, China; 4. Meteorological Observatory of 95994 Unit of the
     Chinese People’s Liberation Army, Jiuquan  735000, China
  • Online:2016-06-30 Published:2016-06-30

摘要:

基于T639数值预报产品与地面气象观测资料,以环渤海地区兴城站为例,选取与云的形成密切相关的4类预报因子——水汽类、大气不稳定度类、大气上升运动类和天气系统强度类,以总云量、低云量为预报对象,运用支持向量机,选取最佳参数,建立兴城站云量的逐月、逐时次精细化预报模型。试预报结果表明:平均预报准确率总云量为71%,低云量为69%,预报准确率较逐步回归模型有所提高;在大部分月份、时次,试预报值的变化趋势与观测值一致,可以较好地反映实际阴晴变换和云量变化;基于支持向量机的回归模型对云量有较好的预报能力。

关键词: 支持向量机, 最佳参数, 云量, 预报

Abstract:

The refined forecast of cloud cover based on Support Vector Machine regression method  was studied by using the products of T639 model and the data of surface meteorological observation station at Xingcheng. Physical quantities about water vapor, atmospheric instability, ascending motion of atmosphere and intensity of weather system,  are closely related to  cloud formation, so they were selected as forecast factors of cloud cover. Then the refined forecast models of total cloud cover and low cloud cover were built by using Support Vector Machine with best parameters. The forecast results of  Support Vector Machine regression models showed that mean forecast accuracy of total cloud cover was about 71%, while that was  about 69% for low cloud cover, which were higher than those of stepwise regression models. And the  trend of  forecasted cloud cover was close to observation data at most times and in most months. So the model based on Support Vector Machine could forecast cloud cover well.

Key words: Support Vector Machine, best parameters, cloud cover, forecast