基于多光谱和地理加权回归模型的石嘴山城市土壤有机碳空间分布研究  

Spatial distribution of soil organic carbon in Shizuishan based on multispectral and geographically weighted regression model

作　　者：夏子书 白一茹[1,2] 包维斌 钟艳霞 王幼奇[1,2] XIA Zi-shu;BAI Yi-ru;BAO Wei-bin;ZHONG Yan-xia;WANG You-qi(College of Resources and Environment,Ningxia University,Yinchuan 750021,Ningxia,China;Arid Area Characteristic Resources and Environmental Governance Department of Education International Cooperation Joint Laboratory,Yinchuan 750021,Ningxia,China)

机构地区：[1]宁夏大学资源环境学院,宁夏银川750021 [2]旱区特色资源与环境治理教育部国际合作联合实验室,宁夏银川750021

出　　处：《干旱区地理》2020年第5期1348-1357,共10页Arid Land Geography

基　　金：国家自然科学基金项目(41867003,41761049);宁夏高等学校项目(NGY2017015);宁夏自然科学基金项目(2018AAC03027);宁夏青年科技人才托举工程项目资助;宁夏重点研发计划重大项目(2018BFG02016);宁夏环境保护科学技术研究项目(2018-07)。

摘　　要：城市土壤有机碳(SOC)分布受城市建设、工业发展等人为因素的影响表现出明显的空间差异。为揭示石嘴山市SOC受城市化、工业化等人类活动的影响,分别利用普通克里格法(OK)、多元线性回归克里格法(RK)、遥感反演方法(RS)和遥感-地理加权回归克里格法(RGWRK)预测石嘴山市SOC空间分布。结果表明:石嘴山市SOC含量在1.31~66.92 g·kg^-1之间变化,其平均值为17.61 g·kg^-1。石嘴山市不同功能区SOC含量存在显著差异(p<0.05),具体表现为工业区>医疗区>商业区>道路>住宅区>公园>农田>科教区;SOC含量变异系数为66.27%,呈中等程度变异;其最佳拟合模型为高斯模型,C0/(C0+C)为0.02,属于强空间自相关。SOC与遥感影像波段DN值的差值(B1-B7、B3-B7、B4-B7)和地形因子(高程、坡度、起伏度)之间存在着极显著的相关性(p<0.01);通过对4种方法的结果进行对比可知以各波段DN值差值与地形因子为输入量,利用RGWRK预测的SOC精度最高,相比OK精度提高了10.05%,相比RK精度提高了8.79%,相比RS精度提高了8.92%;研究区SOC含量呈北高南低的趋势,工业区SOC含量是郊区农田SOC含量的1.92倍,表明城区SOC含量有富集的趋势。The distribution of soil organic carbon(SOC)in urban areas is influenced by human factors such as urban construction and industrialization,and thus shows obvious spatial differences.To reveal the impact of human activities such as urbanization and industrialization on SOC in Shizuishan City,the spatial distribution of SOC in Shizuishan was predicted using ordinary Kriging(OK),multiple linear regression Kriging(RK),remote sensing inversion(RS),and remote sensing-geographically weighted regression Kriging(RGWRK).The SOC content changed from 1.31 g/kg to 66.92 g·kg^-1,with an average of 17.61 g·kg^-1.There were significant differences in SOC content among different functional areas in Shizuishan(p<0.05).Specifically,the performance in decreasing order was,industrial areas,medical areas,commercial areas,roads,residential areas,parks,farmlands,scientific and educational areas.The coefficient of variation of SOC content was 66.27%and there was moderate variation.The best fitting model was the Gauss model,and C0/(C0+C)was 0.02,indicatingstrong spatial autocorrelation.There was a significant correlation(p<0.01)between SOC and differences of DN values in different bands of RS images(B1-B7,B3-B7,and B4-B7)and topographic factors(elevation,slope,and relief amplitude).Comparing the results of the four methods demonstrated that SOC predicted by RGWRK was the most accurate,being 10.05%higher than that of OK,8.79%higher than that of RK,and 8.92%higher than that of RS.SOC content in the northern part of Shizuishan was higher than the southern part.The SOC content in industrial areas was 1.92 times higher than farmlands,indicating that the SOC content in the urban area was enriched.

关 键 词：城市土壤有机碳 空间预测 遥感反演 地理加权回归克里格 

分 类 号：S153.6[农业科学—土壤学]