流域尺度土壤厚度的模糊聚类与预测制图研究  

Prediction and Mapping of Soil Depth at A Watershed Scale with Fuzzy-c-means Clustering Method

作　　者：王改粉[1,2] 赵玉国[1] 杨金玲[1,2] 张甘霖[1] 赵其国[1] 

机构地区：[1]土壤与农业可持续发展国家重点实验室(中国科学院南京土壤研究所),南京210008 [2]中国科学院研究生院,北京100049

出　　处：《土壤》2011年第5期835-841,共7页Soils

基　　金：中国科学院知识创新工程重要方向项目(KZCX2-YW-Q10-3);江苏省自然科学基金项目(BK2008058);国家自然科学基金项目(40771092)资助

摘　　要：基于土壤厚度与景观位置和特征之间的关系,运用模糊c均值聚类(FCM)方法对西苕溪流域的土壤厚度分布进行了空间预测。选取高程、坡度、平面曲率、剖面曲率、径流强度系数和地形湿度指数6个地形因子进行模糊聚类,根据相应的聚类参数将流域地形组合分为8类。利用部分调查获得的土壤剖面数据,结合样点属性和专家经验为典型区赋值,最后由加权平均得到流域土壤厚度预测图。验证结果表明,FCM方法可以对地形因子组合进行有效合理的分级,其预测精度较高,模型的稳定性较好,是一种低成本高效率的制图方法。该方法在土壤厚度预测方面具有一定的可靠性。Soil depth of the west Tiaoxi catchment was predicted using fuzzy c-means clustering（FCM） based on the relationships between soil depth and landscape parameters. Six terrain factors, i.e., elevation, slope, planform curvature, profile curvature, runoff intensity and topographic wetness index were clustered, then the whole catchment was classified into eight combinations of these factors. Typical soil depths from the training soil dataset, combined with attribute of samples and expert knowledge, were assigned to each cluster center. Soil depth map was predicted with weighted average model. Results showed that, FCM method could rationally and effectively classify the combination of terrain factors, and it is a low cost and high efficiency mapping method with satisfactory prediction precision and model stability and could be possible applied to areas with the similar landscape conditions.

关 键 词：模糊C均值聚类 聚类参数 土壤厚度 数字土壤制图 

分 类 号：P934[天文地球—自然地理学]