中国区域TanDEM-X 90 m DEM高程精度评价及其适用性分析

数字高程模型(Digital Elevation Model,DEM)是地球表层系统科学相关研究的基础数据,DEM数据精度的定量评价对科学选择DEM数据源、量化数据误差的影响等具有重要意义.在目前全球尺度可免费获取的DEM数据中,2018年发布的TanDEM-X 90 m DEM(TanDEM-X 90)数据凭借其较好...

Vertical Accuracy Assessment and Applicability Analysis of TanDEM-X 90 m DEM in China

Digital Elevation Model (DEM) is the basic data of studies on monitoring earth surface status and processes. Accuracy assessment of DEM is of great importance in selecting the optimal dataset and estimating the influences caused by DEM errors. Among all the open-access DEM data at a global scale, TanDEM-X 90 m DEM (termed as TanDEM-X 90 hereafter), released in 2018, has attracted widespread attention due to its good performance demonstrated by recent studies. However, few studies focus on the accuracy assessment of TanDEM-X 90 at a large scale, lacking the knowledge of its overall accuracy and the spatial distribution of error influencing factors. In this study, ICESat/GLAS altimetry data was used as basic reference data while SRTM-3 DEM and AW3D30 DEM were used for comparison. Using statistical parameters of Mean Absolute Error (MAE), Root Mean Square Error (RMSE), skewness, and kurtosis, this study investigated the statistical characteristics and spatial distribution patterns of TanDEM-X 90 errors across mainland China. This study also explored the impacts of elevation, slope, geomorphic type, and land cover to DEM vertical accuracy and further analyzed the applicability of TanDEM-X 90. The results indicated that: ① MAE and RMSE of TanDEM-X 90 in China are 4.31 m and 7.87 m, respectively. The overall accuracy of TanDEM-X 90 is close to that of SRTM-3 (MAE=4.72 m, RMSE=7.71 m), but obviously poorer than AW3D30 (MAE=2.69 m, RMSE=4.17 m). ② TanDEM-X 90 achieves the highest accuracy among the three DEMs when the slope is smaller than four degrees. ③ TanDEM-X 90 has better performance than SRTM-3 in three types of landforms, plains, hills, and terraces. ④ Furthermore, the spatial distributions of vertical error of TanDEM-X 90 by watershed were also represented at a national scale, which can provide a beneficial reference for the data applications. In addition, TanDEM-X 90 has been proved to be better in depicting the recent elevation changes of land surface affected by human activities because of its advantages in data acquisition time. For areas with obvious artificial reconstruction such as mining districts, the appropriate time phase of the data according to the research objectives is more important than the difference in data accuracy. This study also found that there are obvious outliers in the TanDEM-X 90 data of the existing version, which restricts the regional availability of the data. Further improvement of this data is the focus of future researches.