地形湿度指数与关键土壤属性的空间耦合关系研究

Concerning the interactions of topography and soil-landscape model, the spatial variation of soil properties was represented by terrain attributes in this item. Topographic wetness index (TWI) derived from gridded digital elevation model (DEM) is one of the most used predictors in digital soil mapping. Making use of topographic information, the spatial distribution of TWI could be used to represent the soil water storage below saturation and spatial variability of contributing area. Nevertheless, discussion of techniques for addressing scale issues and terrain, non-terrain variables considering of TWI has been limited. Methods for TWI calculation were improved by incorporating the terrain and non-terrain variables in this project. Multisource DEMs and different soil sampling data in small watersheds were utilized as basic experimental data. After a comprehensive discussion of multiple contradiction between quantitative and qualitative analysis, macroscopic and microcosmic perspective features, prototype and model, the dissimilarity matrixes of different TWI were established so as to analyze their similarity and otherness. Furthermore, the correlation analysis between TWI and the spatial variation of soil moisture was systematically carried out. According to the expression and dynamic characteristics of soil genesis, various TWIs were coupling with key soil properties by the implementation of dissimilarity matrix. Finally, the effect of different digital terrain analysis methods on the spatial variation of soil attributes was summarized aiming to deepen our knowledge of soil-landscape model. A new evaluation index was researched focusing on the overall accuracies of those maps generated by same terrain variables calculated by different algorithms, rather than only by the difference between the observations and predictions at validation sites. Our research was an innovation which could improve the efficiency and accuracy of digital soil mapping by using the digital terrain analysis and may replenish the fundamental theory of Pedometrics.

基于数字高程模型的地形湿度指数具有明确的物理意义。本课题紧紧围绕土壤-景观模型的核心——地形对土壤性质空间变异的表达，从同时考虑地形、非地形因子地形湿度指数算法的改进出发，以多源高精度DEM与多采样策略土壤属性数据为基本信息源，在综合分析定性与定量、微观与宏观、原型与模型多重矛盾的基础上，综合分析地形湿度指数的不同计算方式，系统构建多指数算法的相异度矩阵，以期揭示其空间分异特征及对实际土壤水分的表征能力；进而，根据土壤发生的表现和动态特征，确定相异湿度指数与土壤属性的空间耦合机制，阐述相异地形分析方法对土壤属性变异的影响，构建面向相异算法土壤制图的功能评价指标，深化地形对土壤发生过程作用的认识，丰富完善数字地形分析、土壤-景观定量模型理论基础。本课题是地形分析方法在数字土壤制图方面一次有意义创新，可望成为计量土壤学相关理论的重要补充，扩展地形分析在精准农业、环境管理等方面的应用范围。

围绕代表性地形因子地形湿度指数（TWI）与土壤-景观模型的构建，在项目执行期间，本研究共开展了以下几个方面的工作：（1）不同景观条件下土壤-景观定量模型构建；（2）基于不同地表曲面模型预测土壤有机碳含量；（3）表层土壤含水量时空预测模型；（4）TWI对于土壤厚度的表征。.研究发现：（1）地形湿度指数在全局/局部回归模型中对土壤有机碳含量具有显著的影响作用，是具有代表性的地形因子，但是其作用弱于高程、年均降雨与年均气温等变量。通过对土壤有机碳含量全局和局部回归中与环境因子的相关分析、全局与局部回归残差空间分异、以及GWR中的多重共线性分析，发现黑河流域土壤有机碳含量呈现出“全局趋势，区域高变异”的分异特征。.（2）合适的局部变量算法能够最大程度地捕捉土壤属性的空间变异特征。当一阶地形因子与二阶地形因子组合使用时，能够通过“黑盒测试”的方法遴选出目标研究区域最优的算法组合。交叉验证结果显示高阶地表曲面模型（Zevenbergen-Thorne与Florinsky）具有更高的预测精度。高阶曲面模型在山区获得了较高的预测精度，因此TWI的计算应该采用这些算法。.（3）TWI在预测模型中较低的重要性，说明在平缓地区TWI对于土壤含水量的表征作用较低。预测结果表明：基于深度信念网络的元胞自动机模型的预测精度较高，能够充分体现土壤含水量的空间异质性。.（4）海拔、剖面曲率和TWI3个地形因子指标能够有效地表征山地区土壤厚度的空间变异。模糊C均值聚类与决策树方法结合是一种获取土壤环境关键阈值和土壤-环境关系知识的有效途径，可为数字土壤制图提供一种有效的解决方案。.本项目的研究结果对于大尺度、山地区土壤制图环境变量的筛选与预测模型的构建具有重要的理论和现实意义。