ABSTRACT Many freely available global or quasi-global digital elevation model (DEM) datasets, including Advanced Land Observing Satellite (ALOS) World 3 Dimensions (3D) 30 m DEM (AW3D30 DEM), Advanced Spaceborne Thermal Emission and Reflection Radiometer DEM (ASTER DEM), Shuttle Radar Topography Mission DEM (SRTM DEM), Multiple-Error-Removed-Improved-Terrain DEM (MERIT DEM), and TerraSAR-X add-on for Digital Elevation Measurement (TanDEM-X) DEM (TanDEM-X DEM), are extremely useful products for many applications, so it is necessary to make a reasonable analysis of the quality of these DEMs. The quality of these DEMs is affected by both horizontal and vertical errors. However, most related researches have paid little attention to the effects of horizontal errors on the vertical accuracy of DEMs. Herein, after analysing the horizontal errors caused in the data acquisition process from optical stereo pairs and interferometry, experiments were proposed to assess the horizontal errors in DEMs. Then, we used the high accuracy digital surface model (DSM) of the Changsha area generated from the ZiYuan-3 triplet stereos to investigate the impact of horizontal errors on the accuracy of the five freely available DEMs. The subpixel level horizontal errors of the five DEMs were calculated, showing that the TanDEM-X DEM has systematic offsets and the other four DEMs have strip-like geometric errors. All the DEMs, except ASTER DEM, attained an accuracy increase of approximately 10% after the horizontal errors were removed. In mountainous areas, correcting the horizontal errors could eliminate the elevation errors caused by the imaging geometry of the optical camera and radar system. Generally, AW3D30 DEM has the best performance in terms of overall elevation accuracy and geometric accuracy after eliminating horizontal errors, and so it can have wide-ranging applications. These findings demonstrate the importance of removing horizontal errors to improve the elevation accuracy in practical applications using these global freely available DEMs.

中文翻译：

---

水平误差对免费提供的数字高程模型 (DEM) 准确性的影响

摘要 许多免费提供的全球或准全球数字高程模型 (DEM) 数据集，包括高级陆地观测卫星 (ALOS) World 3 Dimensions (3D) 30 m DEM (AW3D30 DEM)、高级星载热发射和反射辐射计 DEM (ASTER DEM) )、航天飞机雷达地形任务 DEM (SRTM DEM)、多误差消除改进地形 DEM (MERIT DEM) 和 TerraSAR-X 数字高程测量附加组件 (TanDEM-X) DEM (TanDEM-X DEM) ，对于许多应用来说都是极其有用的产品，因此有必要对这些 DEM 的质量进行合理的分析。这些 DEM 的质量受水平和垂直误差的影响。然而，大多数相关研究很少关注水平误差对 DEM 垂直精度的影响。在此处，在分析了光学立体对和干涉测量数据采集过程中引起的水平误差后，提出了评估 DEM 中水平误差的实验。然后，我们使用由紫元三号三重立体声生成的长沙地区高精度数字表面模型 (DSM) 来研究水平误差对五个免费 DEM 精度的影响。计算了五个DEM的亚像素级水平误差，表明TanDEM-X DEM具有系统偏移，其他四个DEM具有条状几何误差。除 ASTER DEM 外，所有 DEM 在去除水平误差后都提高了约 10% 的精度。在山区，校正水平误差可以消除由光学相机和雷达系统的成像几何结构引起的高程误差。一般来说，AW3D30 DEM在消除水平误差后，在整体高程精度和几何精度方面的性能最好，应用范围很广。这些发现证明了在使用这些全球免费可用的 DEM 的实际应用中消除水平误差以提高高程精度的重要性。