Quality Assessment of Three Digital Elevation Models with 30 M Resolution by Taking 12 M TanDEM-X DEM as Reference

Digital elevation models (DEMs) are the basic data of science and engineering technology researches. The quality of DEMs affects the results of applications. Therefore the quality assessment of DEMs is valuable. DEMs' quality was reported by distribution agencies at a global scale, but they are not representative at a local scale. There are multiple DEMs products in the same local area. Researchers don't know the exact quality of these products and will be confused when choosing the one to use in the subsequent applications. SRTM and ASTER GDEM with 30 m spatial resolution are currently widely used global DEMs, which are freely available and accurate. As a global DEM product, TanDEM-X DEM with 12 m resolution, distributed in September 2016, has attracted users' attention due to its unprecedented accuracy. The geometric resolution, accuracy, and ability to depict complex topography of TanDEM-X DEM 12 m is excellent. It was then resampled to a 30 m pixel size. To help researchers choose the proper DEMs for their studies, we assessed the quality of these 30 m resolution DEMs on selected Chinese regions by taking TanDEM-X DEM 12 m as reference. The quality assessment of the datasets was carried out based on analyses of the differences in elevation between TanDEM-X DEM 12 m and each DEMs. Before calculating the height differences, co-registration is needed to eliminate the effect of systematic errors on quality assessment. This study proposes an improved Least Z-Difference (LZD) algorithm with nine parameters to complete co-registration. Quantile-quantile plots were then generated for visual inspection of normality of registered and non-outliers residuals' distribution between DEMs. The plots showed a residual non-normality. By statistical analysis of residuals, we found that TanDEM-X DEM 30 m had the highest vertical quality relative to the reference DEM, while ASTER GDEM had the worst one. SRTM had a relatively good quality.