区域数字土壤制图方法对比研究

发布时间：2018-04-14 11:33

本文选题：空间回归模型 + 土壤—景观模型　；参考：《郑州大学》2013年硕士论文

【摘要】：常规数字土壤制图技术不能准确揭示土壤在空间上的连续性和渐变特征,应用计量土壤学新理论和DSM新技术,在区域尺度上建立空间土壤信息推绎系统,完成数字化土壤制图对国内土壤地理学及其他相关学科的发展具有重要推动作用和示范意义。研究区位于黄河冲积平原区,受黄河改道、泛滥决口及人类活动影响,地貌类型复杂。本研究首先以模糊c一均值方法进行土壤连续分类,其次运用空间回归分析和土壤一景观定量模型将分类结果和环境协变量进行拟合,建立土壤空间预测模型,实现数字化土壤制图,并对比分析两种制图结果的优劣及在研究区的适用性,主要研究结果如下： (1)运用基于模糊分类算法的OSACA分类系统对土壤样点进行无监督分类,获取了研究区五种中心土壤类型以及采样点与各中心土壤类型间的分类距闵。 (2)空间回归分析结果表明：运用空间回归模型进行区域数字化土壤制图具有合理性和必要性；确定性趋势距离揭示了不同土壤之间自然成土因素的空间差异,非确定性残差表明了非环境因素导致的土壤空间变异；研究区中普通底锈干润雏形土所占面积最广,弱盐灌淤干润雏形土次之,普通底锈干润雏形土应举系所占面积最小；土壤空间变异较大的区域分布在黄河大堤、大功决口扇以及研究区东南部。 (3)土壤一景观模型结果表明：地形因子对平原区土壤形成影响较小,不适合单独作为环境要素组合进行推理制图。土壤类型图中普通简育干润土潘店系分布面积最大,其次为弱盐灌淤干润雏形土,再次为普通人为淤积新成土和普通简育干润雏形土应举系,分布面积最小的是普通底锈干润雏形土。 (4)对比空间回归模型和土壤—景观模型方法的制图结果,发现空间回归模型在进行预测制图时可以更细致地表达土壤在研究区的空间分布情况,对环境因素导致的土壤空间变异敏感度更高,可以很好地反映出区域尺度上自然成土因素对土壤发生发育的影响；土壤—景观模型方法得到的土壤类型单一且分布集中,不符合土壤发生学原理；研究区适合采用空间回归模型进行土壤预测制图。
[Abstract]:Conventional digital soil mapping technology can not accurately reveal the spatial continuity and gradual change characteristics of soil. Based on the new theory of soil metrology and the new technology of DSM, the spatial soil information inference system is established on the regional scale.The completion of digital soil mapping plays an important role in promoting and demonstrating the development of soil geography and other related disciplines in China.The study area is located in the alluvial plain area of the Yellow River, which is influenced by the diversion of the Yellow River, the flood crevasse and human activities, and the geomorphologic types are complex.In this study, the fuzzy c-means method was used to classify soil continuously, then spatial regression analysis and soil-landscape quantitative model were used to fit the classification results and environmental covariables to establish the soil spatial prediction model.Digital soil mapping is realized, and the advantages and disadvantages of the two mapping results and their applicability in the study area are analyzed. The main results are as follows:1) using the OSACA classification system based on fuzzy classification algorithm, the unsupervised classification of soil samples was carried out, and the classification distance between the sampling points and the central soil types was obtained.2) the results of spatial regression analysis show that it is reasonable and necessary to use spatial regression model for regional digital soil mapping, and the deterministic trend distance reveals the spatial differences of natural soil-forming factors among different soils.The non-deterministic residuals indicated the spatial variation of soil caused by non-environmental factors, the area of common rust-dry soil was the most extensive, that of weak salt irrigated silt soil was the second, and that of common bottom rust-dry soil was the smallest.The regions with large soil spatial variation are distributed in the Yellow River levee, the Dagong gully fan and the southeastern part of the study area.The results of soil-landscape model show that topographic factors have little influence on soil formation in plain area, so it is not suitable for inferential mapping as a combination of environmental factors alone.In the map of soil types, the distribution area of Pandian system was the largest, followed by weak salt irrigation and silt dry moistening soil, and again for the ordinary people to silt up newly formed soil and the ordinary simple breeding dry moisturizing soil should be used.The smallest distributed area is the rust-dry rust-dry embryonic soil.4) comparing the mapping results of spatial regression model and soil-landscape model, it is found that spatial regression model can more accurately express the spatial distribution of soil in the study area.The sensitivity of soil spatial variation caused by environmental factors is higher, which can well reflect the influence of natural soil-forming factors on the development of soil on a regional scale, and the soil types obtained by soil-landscape model method are single and concentrated.It does not accord with the principle of soil phylogeny, and the spatial regression model is suitable for soil prediction mapping in the study area.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P285.1

【参考文献】