基于降维处理密度图聚类解析的农田整治分区

为了增强多维空间大样本数据分类的直观性和可解释性,提高农田整治类型区划分结果的实用性,该文提出了一种基于降维处理的密度图法与定性解析相结合的分类方法。该方法首先对多维变量进行主成分分析以提取主变量,然后在第一和第二主变量构成的二维平面空间制作样品分布散点图,采用密度图分割法对样本进行初步分类;最后再结合定性变量特征,对初步分类结果进行定性解析续分。采用该方法对新郑市的315个行政村单元进行了农田整治分区研究,结果表明:1)第一主分量(F1)和第二主分量(F2)包含了土壤有机质、全氮、速效磷、速效钾、农田灌溉率等原始变量信息总量的85.55%,可由其替代原始变量信息;2)F1、F2主变量与粮食单产间的相关性总体优于原始变量,在F1-F2二维平面散点图上,样品聚类特征明显;3)新郑市的315个行政村可划分为6种类型的农田整治区,不同类型区的粮食单产及影响因素差异显著,农田整治工程建设的重点各不相同。结果说明,基于主成分分析的密度图法与定性解析相结合的分类方法适宜于多因素大样本情况下的农田整治类型区划分,研究结果可为新郑市的农田整治规划设计提供重要参考。

Farmland consolidation partitioning by clustering analysis of density graph based on dimension reduction processing

Abstract: For the factors affecting cultivated land quality vary with regions, classifying of land consolidation regions is the groundwork for farmland consolidation, which has great significance to improve the pertinence and effectiveness of the consolidation project. However, there exists some problems with the classification of large sample with multidimensional variable space, such as the classifying parameters being difficult to give and the result lacking visualization. In this paper, a new method called the graphical method of density diagram was put forward based on principal component analysis to improve the objectivity, visibility and interpretation of the classification. The processes were as follows: At the first, the method of principal component analysis was used to get the principal components from multiple factors, and then the values of first principal component (F1) and the second one (F2) of all the samples of Xinzheng City, Henan Province were calculated. Secondly, land consolidation regions were classified using cluster analysis method by making a scatter diagram of the samples in a two-dimensional plane of F1-F2 to create a scatter density figure. Finally, the types of land consolidation regions were subdivided by the method of parsing with the qualitative factors, such as landform type and soil texture. The study was performed in Xinzheng City, and the results showed that: 1) F1 and F2 contained 85.55% of the total information of the original factors including organic matter, total nitrogen, available phosphorus, available potassium of soil and irrigation rate, which could take the place of all of the original factors to classify land consolidation regions. 2) In the scatter density diagram of F1-F2, the obvious cluster feature was presented, by which different types of land consolidation regions could be classified effectively and objectively, and the types had clear connotations for F1 and F2 were well correlated with grain yield. 3) By cluster and analysis, the 315 administrative villages in Xinzheng City were divided into 6 types of land consolidation regions: the comprehensive factors limiting region, the available phosphorus-irrigation rate limiting region, and the available potassium and organic matter-irrigation rate limiting region with low grain yields; the potassium and organic matter-irrigation rate limiting region with medium grain yields; irrigation rate limiting region, and the unlimited region with a medium-high grain yield. The grain yield and factors features of the different types differed from each other. The conclusion can be given that the classification method put forward in this paper is suitable for farmland consolidation partition of large sample with multidimensional variable space, and the classification can be used in farmland consolidation in Xinzheng City.