|
| 基于分层多元复合模型的广东省农田土壤有机碳高精度制图 |
| High precision mapping of soil organic carbon based on multivariate composite model in Guangdong Province |
| 投稿时间：2021-08-10 |
| DOI：10.13254/j.jare.2021.0504 |
| 中文关键词: 农田，土壤有机碳，地理探测器，分层多元复合模型，高精度制图，广东省 |
| 英文关键词: farmland, soil organic carbon, Geodetector, hierarchical multivariate composite model, high precision mapping, Guangdong Province |
| 基金项目:国家重点研发计划课题（2020YFD1100205）；国家自然科学基金项目（U1901601） |
| 作者 | 单位 | E-mail | | 任向宁 | 华南农业大学资源环境学院, 广州 510642
广东省土地利用与整治重点实验室, 广州 510642 | | | 王璐 | 华南农业大学资源环境学院, 广州 510642
广东省土地信息工程技术研究中心, 广州 510642
自然资源部建设用地再开发重点实验室, 广州 510642 | selinapple@163.com | | 林赋英 | 华南农业大学资源环境学院, 广州 510642 | | | 陈淑莹 | 华南农业大学资源环境学院, 广州 510642 | | | 胡月明 | 华南农业大学资源环境学院, 广州 510642
广东省土地利用与整治重点实验室, 广州 510642
广东省土地信息工程技术研究中心, 广州 510642
自然资源部建设用地再开发重点实验室, 广州 510642
青海-广东自然资源监测与评价重点实验室, 西宁 810016 | |
|
| 摘要点击次数: 1067 |
| 全文下载次数: 955 |
| 中文摘要: |
| 农田有机碳库是唯一可在较短时间尺度上通过合理利用而进行适度调节的碳库，农田土壤有机碳高精度制图对进一步明析地理环境背景，提升区域土壤固碳潜力，促进碳交易、碳中和等具有重要的意义。本研究以广东省为研究区，在中大空间尺度区域综合特征分区的基础上，基于地理探测器确定农田土壤有机碳空间分异的变量结构，分区构建分层多元复合模型，根据208 503个土壤采样点数据编制研究区高精度农田土壤有机碳密度空间分布图。结果表明：耦合自然地理特征和社会经济特征，引入多距离空间聚类进行中大空间尺度综合特征分区，能够显著收敛样本离散程度，土壤有机碳样本标准偏差均值、方差均值较未分区前分别下降0.55、3.53，Moran's I指数上升0.08。受自然环境与人为扰动双重影响，农田土壤有机碳空间变异的变量众多，且不同综合特征分区内变量结构差异较大，年均降水量、海拔高度、地形坡度等变量在不同特征分区的影响力存在显著差异，土地利用方式及土壤理化性质等变量对不同特征分区均存在较大的影响力。基于地理探测器构建的分层多元复合模型，较好地解决了中大尺度和复杂情景下土壤有机碳空间分异规律与空间突变的同步表达矛盾，抑制了多变量插值噪声增加，其综合精度较地理加权回归模型（GWRK）、径向基函数神经网络（RBFNN）和普通克里格（OK）分别提升6.45%、10.45%和7.50%。在大密度样本集支持下，综合区域综合特征分区、地理探测器、分层多元复合模型等技术手段编制的广东省高精度农田土壤有机碳空间分布图，预测结果准确，空间细节表达清晰，为编制大空间尺度的土壤有机碳分布图探索了有效路径。 |
| 英文摘要: |
| The soil organic carbon pool of farmland is the only carbon pool that can be appropriately adjusted by rational utilization in a short time scale. High precision mapping of soil organic carbon is helpful to further enhance the potential of regional soil carbon sequestration, analyze the geographical environment background, and promote carbon trading and carbon neutralization. This study, The study took Guangdong Province as the study area, which was divided into 13 comprehensive characteristic zones on medium and large spatial scale. The variable structure of soil organic carbon spatial differentiation in farmland was determined by Geodetector, and hierarchical multivariate composite models(MCM) was constructed. According to the data of 208 503 soil sampling points, we chart a highprecision spatial distribution map of soil organic carbon density in the study area. The results show that comprehensive feature zoning on a medium and large spatial scale, which was carried out by coupling natural geographical characteristics with socio-economic characteristics, and introducing multidistance spatial clustering, can significantly converge the degree of sample dispersion.The mean standard deviation and mean variance of soil organic carbon samples decrease by 0.55 and 3.53 respectively, and Moran's I index increase by 0.08. Under the dual influence of natural environment and human disturbance, there are many variables of spatial variation of soil organic carbon in farmland, and the variable structure in different comprehensive characteristic zones is quite different. Average annual precipitation, altitude, terrain slope and other variables play a significant role in mountainous and hilly areas, but not in plain and hilly areas. However, variables such as land use modes and soil physical and chemical properties have extensive and significant influence on different characteristic zoning. The hierarchical multivariate composite model based on Geodetector better solves the contradiction between the synchronous expression of spatial differentiation law and spatial mutation of soil organic carbon in medium and large-scale and complex scenarios, and suppresses the increase of multivariable interpolation noise. Its comprehensive accuracy is 6.45%, 10.45% and 7.50% higher than geographically weighted regression model-Kriging(GWRK), radial basis function neural network(RBFNN), ordinary Kriging (OK)respectively. With the support of high-density sample set, the high precision soil organic carbon map of Guangdong Province integrates the methods of regional comprehensive feature zoning, Geodetector and hierarchical multivariate composite models. Its prediction results are accurate and the spatial details are clearly expressed, which explores an effective path for compiling the large-scale spatial soil organic carbon distribution map. |
| HTML
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
