空间聚类分析是空间数据挖掘的主要方法之一,旨在发现海量数据中潜在的空间分布模式和异常特征.此处采用空间聚类分析和多尺度分析相结合的策略对海温进行多尺度气候分区.首先,针对气候时间序列特征和异常区域特征,提出一种顾及异常区域的多尺度分区方法；进而,分别采用传统方法和新方法对海温进行多尺度气候分区,通过比较分析发现新方法更优越；最后,利用气候指数对分区结果进行验证,说明新方法是有效的.%Spatial clustering is an important tool of spatial data mining and knowledge discovery. Spatial clustering aims to find the potential patterns of the spatial distribution and some abnormal patterns in a large spatial dataset. So spatial clustering is an effective approach for spatial regionalization based on the climate elements. For this purpose, a scale-space clustering method was proposed for multi-scale regionalization of sea surface temperature. Firstly, a modified multi-scale regionalization method with the consideration of abnormal region was developed based on scale-space theory for the climate time sequences. Secondly, a detailed comparison for the regionalization of the sea surface temperature was made between the traditional method and the modified method. It proved that the result of the modified method was more accurate than that of the traditional method. Finally, the ocean climate indices were employed to illustrate the results of the regionalization by the modified method.
展开▼
