基于K均值聚类的中国夏季降水日变化空间分布模式

为了获取降水日变化的空间分布模式,本文采用K均值聚类算法对中国陆地区域的夏季逐时格网降水数据进行了聚类。首先,采用K均值聚类算法对每个格网上的逐时降水数据进行聚类。然后,根据每一聚类的降水日变化峰值,将具有相似峰值时间的聚类合并成为一个分类。合并后的分类对应一种降水日变化类型,分类中的格网边界则构成了该类降水日变化的空间分布模式。研究结果表明,中国大部分地区的降水量日变化由降水频率日变化主导。此外,一些盛行夜雨区域在空间分布上表现出从西向东的分布模式,且降水日变化峰值时间表现出了从西向东逐渐延迟的现象。结合地形分析,研究发现一些夜雨区的降水峰值延迟现象与MPS环流效应导致的雨带移动现象较为一致,得出MPS(Mountain-Plain Solenoid)环流效应是导致这些地区盛行夜雨的结论。本文研究结果可为探索降水日变化的形成机理提供线索,也可为研究其他地区降水日变化提供参考。

Spatial Distribution of Diurnal Rainfall Variation in Summer over China Using K-means Algorithm

In order to obtain an objective spatial distribution map of diurnal rainfall variation, this study uses the K means clustering algorithm to mine hourly gridded rainfall data over China. Firstly, a large enough number is used to cluster the hourly precipitation at each pixel. Then, according to the peak time of diurnal rainfall variation, the clusters with similar peak time are merged. Each merged cluster corresponds to one type of diurnal rainfall variation, and the locations of grid cells in each cluster form the map of diurnal rainfall variation over China. The results indicate that the characteristic of diurnal variation of rainfall amount is dominated by the diurnal variation of rainfall frequency. In addition, the regions prevailing nocturnal rainfall present a spatial pattern ranged from west to east, and the peak times of the nocturnal rainfall regions show a time delay propagation from west to east. Taking an overlay analysis between the regions prevailing nocturnal rainfall and terrain, the spatial pattern of the regions prevailing nocturnal rainfall can be well explained by the propagation of the rain belt driven by the MPS (Mountain Plain Solenoid) circulation. Overall, the results of this paper can provide clues for the mechanism study of diurnal rainfall variation and also provide a reference for studying the spatial distribution of diurnal rainfall variations in other regions.