基于分类树方法的土壤有机质空间制图研究  

SOIL ORGANIC MATTER MAPPING BASED ON CLASSIFICATION TREE MODELING

作　　者：周斌[1] 许红卫[1] 王人潮[1] 

机构地区：[1]浙江大学农业遥感与信息技术应用研究所,杭州310029

出　　处：《土壤学报》2003年第6期801-808,共8页Acta Pedologica Sinica

基　　金：国家自然科学基金项目 (4 0 1 0 1 0 1 4和 40 0 0 1 0 0 8)资

摘　　要：以浙江省龙游县研究区为例 ,提供了一种推断和表达土壤有机质 (OM)含量空间分布信息的方法 ,通过一种数据挖掘方法———分类树建模方法将土壤OM含量与一些易于广泛观测的景观属性 ,包括地形、地质、土地利用和遥感影像建立联系 ,从而将有关土壤OM含量分布的知识转入一种清楚的、定量的、与景观因子相关联的规则系统中 ,并以此来预测研究区土壤OM水平的连续空间分布。树分析选取了高程、岩石类型、土属类型、PC4 、PC2 、土地利用类型、PC3、PC1、上坡贡献面积、坡度、坡向、平面曲率和剖面曲率来预测研究区土壤OM等级的分布。其中 ,高程、岩石类型、土属类型和反映植被覆盖度的PC4 、PC2 以及土地利用类型对于研究区土壤OM等级预测更为重要。从分析结果来看 。Based on the case study of Longyou County,Zhejiang Province,an approach was introduced to deducing and expressing spatial distribution of soil organic matter. This is a kind of data mining method or classification tree modeling method,which associates soil OM content with some extensive easily observable landscape attributes,such as landform,geology,landuse and remote sensing images,thus transferring the soil OM-related information into a clear,quantitative,landscape factor-associated regular system. This system can be used to predict continuous soil OM spatial distribution. By analyzing the factors such as elevation,type of the rock,type of the soil,PC 4,PC 2,land uses,PC 3,PC 1,upslope contributing area,slope,exposure,plane curvature and profile curvature,the classification tree can predict distribution of soil organic matter levels. Among the factors,elevation,type of rock,type of soil,landuse,PC 4 and PC 2 (two indexes of vegetation coverage) are considered as the most important variables for predicting soil OM. Results of the prediction show a quite close relationship between soil OM contents and types of the landscape sorted by the classification tree with an accuracy of 81.1%.

关 键 词：分类树方法 土壤有机质 空间制图 数据挖掘方法 建模方法 景观属性 

分 类 号：S153.621[农业科学—土壤学] S159.9[农业科学—农业基础科学]