期刊文献+

三种聚类分析方法在中国温度区划分中的应用研究 被引量:28

Three Cluster Methods in Regionalization of Temperature Zones in China
下载PDF
导出
摘要 根据全国512个气象站1961~2010年的逐日气温观测资料,采用日平均气温稳定通过10 ℃的日数(≥10 ℃日数)作为划分我国温度分布的指标,经过旋转经验正交函数分析(REOF)方法处理,找出时间演变过程中变化相似的区域归为一类.在此基础上,采用快速样本聚类法(K-means法)、分层聚类法(Ward法)、聚类统计检验法(CAST法)3种聚类分析方法对全国范围的温度变化区域差异进行客观分区,结果分别将全国温度变化区划分为10个地区、9个地区、13个地区,且不同区域分界线与中国地形分布有很好的一致性.研究表明:K-means法运算简单快捷,结果会不断修正到最佳为止;Ward法聚类过程清晰明了,可以选取需要划分的类别数;CAST法使样本通过显著性检验,不但有助于气候变化的客观分区,而且在划分温度区时客观考虑气候变化对分区结果的影响也有很重要意义. A scheme for regionalization of temperature zones was established on the basis of daily surface air temperature observations from 512 stations in China during 1961-2010. Days with daily surface air temperature ≥10 ℃ were used as indicators for zoning the temperature distribution. After REOF (Rotated Empirical Orthogonal Function) analysis, areas with similar temperature changes in the time evolution were classified as one zone. The temperature zones were classified by three types of clustering analysis methods. The national area was divided into ten temperature zones by using the K-means method, nine temperature zones by using the Ward method, and 13 temperature zones by using the CAST (Cluster Analysis with Statistical Test) method. The boundaries of the various regions show a good consistency with the Chinese topography. Calculation by using the K-means method is considered to be simple and quick because, the results are revised until the best results are achieved. Moreover, the clustering process of the Ward method is clear; any number of categories may be selected. Finally, the results of the CAST method pass the significance test; therefore, this method is meaningful for zoning.
作者 韩微 翟盘茂
出处 《气候与环境研究》 CSCD 北大核心 2015年第1期111-118,共8页 Climatic and Environmental Research
基金 公益性行业(气象)科研专项GYHY201106018 国家自然科学基金项目41175080
关键词 聚类分析 日平均气温稳定通过10℃ 日数变化分区 Cluster analysis, Days with daily surface air temperature ≥10℃, Change, Zoning
  • 相关文献

参考文献14

二级参考文献113

共引文献1174

同被引文献443

引证文献28

二级引证文献173

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部