利用ASTER卫星立体像对提取喀斯特数字地形模型——以广西大化高峰丛深洼地喀斯特地貌为例

ASTER卫星立体像对已经被成功地用于提取全球数字地形模型（DEM）,其有效性已经在不同类型的非喀斯特地貌区域得到验证。和一般地貌相比,喀斯特地形,特别是中国南方喀斯特,具有自己独特的形态特征。能否从ASTER影像中精确地提取喀斯特地貌DEM目前尚未见到有任何报道。本文利用ASTER影像成功地提取了广西大化地区高峰丛深洼地的喀斯特地貌DEM,并且将提取结果与同一地区1∶5万地形图控制点的高程和4个剖面的高程变化进行了对比。研究区内绝大部分区域是喀斯特地貌,但在其西北部有一舌状砂岩地貌分布,从而提供了一个很好的喀斯特和非喀斯特地貌DEM精度对比的机会。本文的研究结果表明,从ASTER影像提取出来的砂岩地貌的DEM均方根误差要小于喀斯特地区,表明从ASTER影像中能更有效地提取非喀斯特地貌的DEM。相比之下,提取得到的喀斯特DEM精确度则稍差。究其原因主要是因为研究区内独特的高峰丛深洼地地形形态。由于峰丛地形的阴影在构成立体像对的两幅影像（nadir-looking和aft-loo-king）上的大小和形态差异很大,导致了DEM生产过程中两幅影像不能精确配准,从而产生比较大的误差。相比之下,砂岩地区地貌相对比较平坦,地形阴影大小和形态在两幅影像上差别不大,故而能得到精度比较高的数字地形模型。本文研究结果同时还表明适当增加地面控制点能在一定程度上改善提取出来的喀斯特地貌DEM的精度。尽管提取的数字地形模型精度稍低,其均方根误差仍然远远小于该地区峰丛和洼地个体体量。因此本研究认为从ASTER立体像对提取的数字地形模型可广泛用于喀斯特地貌形态特征研究。

Mapping cockpit karst in Southern China from ASTER stereo images: DEM validation and accuracy assessment

Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) stereo images were used to automatically produce a digital elevation model (DEM) of a rugged karst area in southern China. The accuracy of the DEM was assessed by comparing elevation data at check points and along four transects with reference data derived from 1:50 000 scale topographic maps. The quality of the DEM was significantly improved when the number of ground control points (GCPs) from the topographic maps was increased from 30 to 60 but only slightly improved when 90 points were used. The 90-point DEM had an elevation root mean square error (RMSE) of 42.7 m, which is higher than the accuracy specifications for the ASTER sensor, mainly due to the variable quality of GCPs and the rugged terrain of the study area. Topographic shadows also influenced the accuracy of the DEM, as indicated by check points in areas under shadow that had a higher RMS elevation error than those not in shadows. RMS elevation errors were also higher in areas of more rugged relief, possibly because of a larger area in shadow, and possibly also because of differences in the shadow footprint (shape and extent) on the ASTER nadir-looking and aft-looking (3n and 3b) images used to generate the DEM. Although the DEM has a slightly larger RMS error, it still provides a valuable source of topographic data for morphometric analysis of rugged cockpit karst in southern China, as the dimension of closed depressions is much bigger than the RMSE of the DEM.