【作者】 李志斌；

【导师】 陈佑启；

【作者基本信息】 中国农业科学院 ， 农业遥感， 2010， 博士

【摘要】 土壤有机质是植物养分的主要来源,也是地球碳库中参与全球碳循环的一个重要组成部分。准确地估计土壤有机质空间变异性对于土壤质量评价和固碳潜力评估很重要,这对于农业管理和环境研究具有重要意义。土壤有机质具有空间变异性,它受土壤自身差异、地形和植被等的影响。近年来采用空间辅助信息来提高土壤属性预测的精度已成为计量土壤学的热点,而把遥感数据、地形数据等环境因素加入到土壤属性空间模拟中更是研究的一个发展方向。尽管借助辅助数据对于土壤有机质的空间变异性研究已有若干,但多是对于地形平坦的区域,而对于在区域尺度上地形较为复杂的土壤属性的空间分布特征的研究不多见,而对遥感数据和地形数据及其衍生数据进行数据挖掘,获取更多辅助信息从而提高对土壤属性模拟精度,也正是目前数字土壤研究的重点。因此本文结合3S技术,分别采用不同分辨率的DEM和较高分辨率的多光谱遥感影像,提取相关地形因子与遥感指数,分析土壤有机质与这些环境因子之间的关系,并选取协同克里金、回归克里金进行土壤有机质空间分布模拟,并与普通克里金进行比较研究,以寻求提高土壤属性空间变异预测精度的方法。论文先就DEM对土壤有机质空间模拟的影响进行了分析。选取分辨率为30m和90m的DEM,借助GIS生成坡度、坡向、地形湿度指数等衍生属性。然后提取采样点对应的地形属性值,并分析其与土壤有机质之间的相关性。经过分析发现地形湿度指数与有机质之间具有最高的相关系数,而采用多流向算法得到的地形湿度指数与有机质的相关性比采用单流向算法得到的地形湿度指数与有机质的相关性强。在此基础上,论文将基于不同算法的地形湿度指数分别作为辅助数据在回归克里金和协同克里金方法中加以应用,对研究区域的土壤有机质进行空间模拟,并将模拟结果与采用普通克里金方法得到结果进行对比研究。本文还分析了TM影像对土壤有机质空间模拟的影响研究。论文首先选取了研究区域2006年8月19日的TM影像,经过辐射校正、大气校正、投影转换后,进行穗帽变换,并提取植被指数。然后提取采样点对应的遥感信息,分析其与有机质含量之间的相关性。通过分析发现波段5反射率与有机质之间具有较强的相关性。论文采用协同克里金方法,以波段5反射率为辅助信息,模拟研究区土壤有机质空间分布,并采用验证数据集进行精度验证,最后与普通克里金的结果进行比较。本文的主要结论如下:(1)地形湿度指数与有机质之间具有较强的相关性,而地形湿度指数与有机质之间的相关性因DEM分辨率不同而不同,基于较高精度的DEM生成的辅助信息对提高预测精度效果明显。(2)当辅助变量与土壤属性之间相关性不强时,协同克里金的效果要比回归克里金效果要好。(3)以TM影像的波段信息,反映了地表植被的生长状况,在一定程度上能够间接地反映了土壤有机质状况。(4)以遥感信息为辅助变量的协同克里金方法对区域土壤有机质空间模拟效果要好于普通克里金,表明遥感数据作为辅助变量能够提高土壤有机质预测精度和可靠性。本文的主要创新点:(1)基于单流向算法和多流向算法的地形湿度指数在不同克里金插值方法中应用对土壤有机质空间模拟精度的影响研究对计量土壤学具有一定的理论价值。(2)在较大尺度上采用不同的栅格大小的DEM如何影响其对土壤特性的预测还并没有定论,而研究不同栅格单元的大小对土壤属性预测的影响的研究尚不多见。因此本文的研究对于土壤空间变异理论研究提供了案例,对丰富土壤空间变异理论具有一定意义。(3)借助影像分析NDVI从而间接反映出较大区域上土壤有机质空间分布特征。该研究是在较大尺度上一次方法的探索,对于土壤属性空间变异研究具有一定的理论指导。更多还原

【Abstract】 Soil organic matter is a major source of plant nutrients and a major reservoir in the global C-cycle. Accurate estimate of spatial variability of soil organic matter is essential to evaluate soil quality and assess the carbon sequestration potential from field to regional scales, which is important for agricultural management and environment research on terrestrial sequestration of atmospheric carbon. Soil organic matter, influenced by natural soil variability and topography, is often spatially variable. Recently utilizing spatially correlated secondary information to improve the accuracy of prediction of soil properties has received more attention in pedometrics. Although some studies have been conducted to identify the spatial pattern of soil properties distribution in low-relief areas, only little is known about spatial distribution of the soil properties in more complicated areas in Northeast China. In order to provide adequate soil information for the modeling of landscape process, such as soil erosion, soil quality evaluation and plant growth, this study investigates to what extent cheap and readily available ancillary information derived from digital elevation models and remote sensing data can be use to support soil mapping respectively, and to indicate soil characteristics on the regional scale. Therefore, the terrain attributes and remote sensing indices are derived from different resolution DEM and remote sensing image respectively by 3S technology, the correlations between soil organic matter and derived secondary variables are compared and mapping results using different kriging strategies are presented.In this paper, we firstly study the impact of DEM on SOM mapping by implementing different kriging methods in which DEM and derived attributes are employed as secondary variables. The DEM with the resolution of 30m and 90m are selected, and several corresponding terrain attributes are computed by GIS. And the correlations between the derived secondary variables and SOM are compared. It is found that TWI based on MFD algorithm showed a stronger correlation with SOM than TWI base on D8 algorithm. Then, to compare their performance in different SOM mapping strategies, we designated TWI based on different algorithm as exhaustive secondary variables and employed OK, RK and CK for spatial prediction.This study was conducted to evaluated and compare spatial estimation by kriging and cokriging with remotely sensed data to predict SOM. To predict SOM in the study area using remotely sensed data as auxiliary variables, we selected a Landsat Thematic Mapper (TM) image acquired on 18 Aug. 2006. Several vegetation indices and tasseled cap transformation were calculated after the image preprocessing. And the relationship between SOM and remotely sensed data was estimated. Strongest correlations were found between the Landsat spectral reflectance of B5 and SOM content. Then we predicted the SOM using the kriging and cokriging models based on the B5 as secondary variables. And the performance of kriging and cokriging was assessed.The main conclusions of this paper are as follows: (1) Stronger correlations were found between topographic wetness index and SOM, and the correlation varied with the DEM resolution. The secondary variables derived from higher resolution DEM improved the prediction accuracy obviously. (2) Cokriging performed better than regression kriging when secondary variables do not have strong enough correlations with the primary variable. (3) The vegetation condition revealde by TM images can reflect SOM distribution indirectly. (4) The result showed cokriging with remotedly sensed data was better than kriging in SOM prediction, and the cokriging approach indicated that remotely sensed data have the potential as auxiliary variables for improving the accuracy and reliability of SOM prediction.The main innovation of this study is as follows: (1) the study which incorporated TWIs base on SFD and MFD algorithms into different kriging methods to predict SOM has a theoretic value, few studies have characterized explicitly the impact of different TWI algorithms on SOM mapping. (2) How to influence the performance of SOM prediction with different resolution DEM as secondary variable is not conclusive in larger scale, and the research of comparing the SOM prediction performance with different resolution DEM is few. Thus it provided a case study on the theory of soil attributes spatial variability, and it is significant to the development of theory about soil characteristics spatial variability. (3) The NDVI can indirectly the spatial variability of soil organic matter. It is a method exploration in a large scale. The result provides guidance for soil spatial variability study.更多还原