基于分类变量定量化的玉树市土壤全碳制图  

Soil total carbon mapping based on quantification of categorical variables in Yushu City

作　　者：李润祥 何林华 高小红 LI Runxiang;HE Linhua;GAO Xiaohong(School of Geographical Sciences,Qinghai Normal University,Xining 810008,China;MOE Key Laboratory of Tibetan Plateau Land Surface Processes and Ecological,Xining 810008,China;Qinghai Province Key Laboratory of Physical Geography and Environmental Process,Xining 810008,China;Academy of Plateau Science and Sustainability,Xining 810008,China;School of Geographical Sciences,China West Normal University,Nanchong 637000,China)

机构地区：[1]青海师范大学地理科学学院,西宁810008 [2]青藏高原地表过程与生态保育教育部重点实验室,西宁810008 [3]青海省自然地理与环境过程重点实验室,西宁810008 [4]高原科学与可持续发展研究院,西宁810008 [5]西华师范大学地理科学学院,南充637000

出　　处：《生态科学》2023年第6期51-62,共12页Ecological Science

基　　金：国家自然科学基金项目(41550003);青海省科技厅自然科学基金项目(2021-ZJ-913)。

摘　　要：基于数学模型的土壤属性制图,具有高效、快速及成本低等特点,弥补了空间插值法忽视与土壤属性密切相关的土壤类型、植被类型、环境因素等分类变量或定性变量的缺陷。为了提高土壤属性制图的准确率,降低不确定性因素的影响,研究分类变量在土壤属性制图中的应用。以青海省玉树市土壤全碳制图为例,引入由土壤类型、地貌类型等分类变量分别同DEM和NDVI共同构建加权变量,探索加权变量在土壤属性制图中的可行性和实用性。研究结果表明:(1)玉树市土壤全碳含量范围在19.60—120.55 g·kg^(–1),平均值为55.80 g·kg^(–1),标准差为19.22 g·kg^(–1),变异系数为34.44%,属于中等程度的空间变异;(2)加权变量建立的模型优于由数字高程模型、坡度、坡向等常规变量建立的同种模型,加权变量的整体重要性高于常规变量,由地质类型、土壤类型、植被类型与DEM构建的加权变量的重要性远高于DEM本身的重要性;(3)基于全变量多元回归模型是预测土壤全碳含量最佳模型,其中加权变量的累计重要性程度为0.55,预测结果符合研究区地学规律和实际情况。总之,加权变量是有效利用分类变量的一种新的方式,为获取土壤属性制图变量提供了新方法,其可行性与实用性得到了一定的验证。The mapping of soil properties based on mathematical model has the characteristics of high efficiency,high speed and low cost,which makes up for the deficiency of spatial interpolation which ignores the categorical or qualitative variables such as soil type,vegetation type and environmental factors closely related to soil properties.In order to improve the accuracy of soil properties mapping and reduce the influence of uncertain factors,it is a new idea to make rational use of categorical variables in continuous soil properties mapping.In this study,total carbon mapping of soil in Yushu City,Qinghai Province was taken as an example,and weighted variables were constructed by soil type,geomorphic type and other categorical variables together with DEM and NDVI respectively,to explore the feasibility and practicability of weighted variables in soil properties mapping.The results show as follows.(1)The total carbon content of the soil in Yushu City is 19.60—120.55 g·kg^(–1),the average is 55.80 g·kg^(–1),the standard deviation is 19.22 g·kg^(–1),and the coefficient of variation is 34.44%,which is a medium degree of spatial variability.(2)The weighted variable model is superior to the similar model established by the digital elevation model,slope,aspect and other conventional variables.The overall importance of weighted variable is higher than that of conventional variable.The importance of weighted variable constructed by geological type,soil type,vegetation type and DEM is much higher than that of DEM itself.(3)The all-variable multiple regression model is the best model to predict soil total carbon content.The cumulative importance of the weighted variables in the all-variable multiple regression model is 0.55,and the prediction results are in line with the geosciences law and actual situation in the study area.In conclusion,weighted variables are a new way to make effective use of classified variables,which provides a new method for obtaining soil properties mapping variables,and its feasibility and prac

关 键 词：土壤属性 制图 加权变量 全碳 分类变量 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]