全国极端气温事件的空间关联模式挖掘研究

The Spatial Association Rule Mining Research on the Extreme Temperature Events in China

根据全国739个气象台站1961年1月至2005年12月的逐日气象数据记录建立时空序列数据集,提取极端高温事件和极端低温事件。结合传统关联规则挖掘技术和地理空间数据分析方法,对极端气温事件数据集进行了空间关联模式的分析。实验结果显示,所得空间关联模式中涉及的区域在空间上具有明显的聚集性;在东北、华中两个局部地区的台站中,极端气温事件的发生存在较强的关联规则(支持度阈值6%,置信度阈值95%),而在其他区域的台站中,极端气温事件不存在类似的关联规则,且极端高温事件的关联规则数量要明显高于极端低温事件。对存在关联规则的台站进行空间分析发现,同一关联规则内的各台站具有空间邻近性,其邻近范围约为200 km。以上空间关联模式的挖掘分析,可以为我国极端气温事件的预警和防控提供有价值的参考。

The extreme high/low temperature events from Jan.,1961 to Dec., 2005 extracted from Chinese over 700 meteorological stations＇ daily temperature dataset. In order to identify the spatial association pattern that exists in the extreme temperature events dataset more effectively, the association rules of the original data set by combining the traditional spatial association rules mining method and the geography analysis method are discovered. When setting the support threshold equaled to 6%, and the confidence threshold equaled to 95%, the experimentation shows that the spatial extreme temperature events in adjacent areas have a higher occurrence probability in the same month, especially in Northeast and Central China. Within the study area, extreme high temperature event shows the higher frequency and stronger strength of association than extreme low temperature events. By analyzing those association rules, it is found that the meteorological stations in most of the rules were spatially adjacent, which means that the frequent extreme temperature events are happened in the nearby areas. The mining results and analyzing conclusions can provide a valuable reference for early warning and prevention on extreme temperature events in China.