基于决策树分类济宁市土壤有机碳遥感反演.docVIP

基于决策树分类济宁市土壤有机碳遥感反演 　　摘要：本研究利用Landsat 8遥感影像数据以及土壤有机碳实测数据，以研究区表层0～20 cm土壤的有机碳含量为研究对象，通过SPSS的多元线性回归分析，建立预测模型：Ysoc=23.448+65.958b1-67.703b4-21.778b7（R2=0.744，P0.01）；进而采用基于ENVI的波段运算及决策树分类法，获取了济宁市土壤的有机碳含量与分布状况。结果表明：（1）研究区的土壤有机碳含量平均约为12.45 g/kg，处于中等偏下水平，部分地区有机碳含量接近于零，微山县和任城区含量最高；（2）研究区有机碳含量主要集中在12～18 g/kg，面积约为6 165.13 km2，占研究区土地总面积的59.99%，分布最为广泛，在各个县区均有分布。 　　关键词：土壤有机碳；遥感反演；多元回归分析；决策树分类；济宁市 　　中图分类号：S127文献标识号：A文章编号：1001-4942（2018）04-0133-05 　　Abstract Taking 0～20 cm soil layer as research object， the remote sensing data of soil organic carbon were obtained from Landsat 8 and the actual values were measured in laboratory. Through the multiple linear SPSS regression analysis，the prediction model of Ysoc=23.448+65.958b1-67.703b4-21.778b7（R2=0.744，P0.01） was established. Then the soil organic carbon content and distribution in Jining City were obtained by the band math and decision tree classification based on ENVI. The results were as follows. （1） The average soil organic carbon content in the study area was about 12.45 g/kg， which was in the lower middle level. The organic carbon content in some areas was close to 0. The highest content was in Weishan and Rencheng districts. （2） The content of organic carbon in the study area was mainly concentrated at 12～18 g/kg，which was about 6 165.13 km2 and accounted for 59.99% of the total study land area. It had the most widely distribution area， even was in each county or district of Jining City. 　　Keywords Soil organic carbon； Remote sensing retrieval； Multiple regression analysis；Decision tree classification； Jining city 　　碳作?樽匀唤缰械囊恢衷?素，在各个圈层之间是不断循环的，尤其是大气圈CO2和岩石圈SOC的存在形式，与人类生存密切相关[1]。土壤有机碳（soil organic carbon，SOC）作为土壤碳库的一部分，直接或间接影响着全球的气候和土地生产力。土壤有机碳的研究至今已有200多年的历史，20 世纪 70 年代以来，随着统计学方法的普及，生命带法、土壤类型法、有机碳模型法等开始大量应用于有机碳含量及其垂直分布的研究[2-7]。近年来基于RS和GIS软件的遥感反演越来越成为研究土壤有机碳空间分布特征的重要方法，空间插值法、光谱反射率、DN值等遥感信息建模法[8，9]都是常用的遥感反演方法。代杰瑞等[10]利用土壤剖面资料和MapGIS 软件中的空间分析功能实现了山东省表层土壤有机碳密度空间分布的可视化；游浩辰[11]利用ArcGIS的空间插值法实现了遥感反演制图，进而研究了顺昌县林地土壤有机碳的空间分布特征；王