面向多分辨率DEM的河网相似性测度与分析

河网是地形结构的核心要素,能够有效地反映DEM对地表形态的刻画能力。实现不同分辨率条件下DEM河网相似性测度对DEM地形综合、DEM质量评估及DEM不确定性分析等研究具有重要意义。基于此,本文以黄土高原典型样区为研究区,基于5 m高精度DEM建立的多分辨率DEM数据集,构建了地形特征自适应的DEM河网自动提取方法,建立了综合拓扑关系、方位关系、距离关系及属性特征四维信息的河网相似性度量模型,并分析了河网相似性变化特征及其与地形参数的相关性。实验结果显示,河网自动提取方法的制图精度和用户精度均在90%以上,总体精度优于Passalacqua等提出的GeoNet方法,能够有效实现不同分辨率下河网较高精度提取;河网相似性度能综合反映河网空间分布特征的差异性,河网相似度与分辨率变化率之间的幂函数关系R²达0.978,且河网相似度与高程、坡度中误差分别表现为幂函数和对数函数关系,后者对数关系显著性优于前者幂函数关系。研究表明,在以河网作为DEM构建重要数据源的背景下,本文的研究成果可为DEM质量及其应用适宜性的定量评价提供基础。

Quantifying Spatial Similarities of Dainage Networks on Multi-Resolution DEMs

Drainage network is a key element of topographic structure and can be applied to assess the DEM quality effectively. As the spatial resolution of DEM decrease, the topographic information represented by DEM will be more generalized. The generalized drainage networks can thus be derived from low-resolution DEM. The spatial variation of drainage networks at different DEM resolutions can be quantified by the metric of spatial similarity. This metric has been widely used in terrain generalization, quality assessment, and uncertainty analysis. Therefore, this paper aims to quantify the spatial similarity of drainage networks derived from multi-resolution DEM. Firstly, this paper presents an adaptive method to derive drainage networks from DEM at different resolutions. This adaptive approach combines the non-linear filtering of elevation data, a statistical analysis of geometric curvature, and geodesic minimization principles to automatically extract drainage networks. Secondly, the calculation of spatial similarity of drainage networks uses four indicators including topological relation, direction relation, distance relation, and attribute variation. The weighted sum is applied to sum up the four indicators, and the corresponding weights are 0.22, 0.25, 0.31 and 0.22, respectively. Finally, based on the spatial similarity of different drainage networks, the varying patterns and relationships with the errors of topographic parameters (e.g., elevation and slope) are analyzed via a sample area located in the hilly and gully regions in Loess Plateau. The results show that: (1) the adaptive method for extracting drainage networks can effectively capture the morphological features represented by DEM at different resolutions and automatically generate drainage networks with high accuracy; (2) there is a significant relationship between the spatial similarity of drainage networks and the corresponding resolution change, which can be fitted by a power function; and (3) the relationship of spatial similarity with the root mean square error of elevation and slope can be fitted by the power and logarithmic functions, respectively. The relationship with the latter is more significant than the former.