Vertical accuracy of open-source remote sensing data (AW3D30, TanDEM-X, ATLAS) for understory terrain estimation

Abstract

Understory terrain is an important fundamental indicator for monitoring changes in forest ecosystems. However, in-depth analysis of the accuracy of existing international open-source DEM products is still lacking for understory environments. The DEM data produced by Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) carries an advanced topographic LiDAR altimeter system (ATLAS) that contains elevation information of the ground surface as well as features such as the vegetation above. It can reflect the elevation of the ground surface, including the part obscured by vegetation or canopy, and can be better used to monitor changes in ground surface elevation. However, few existing studies have been analyzed the accuracy of open-source DEM products (include ATLAS, AW3D30 DEM, and TanDEM-X data) for estimating the ground in understory environments. Therefore, this study selected Aiken County, Cook Inlet, and Yucatan as study areas to evaluate ICESat-2/ATLAS data, AW3D30 DEM and TanDEM-X data using the airborne G-LiHT system’s post-processing DTM. The study also quantitatively evaluated the influence of slope, aspect, and land vegetation type on elevation accuracies of different open-source DEM products. The experimental results showed that the ATLAS data at the footprint scale produced more accurate evaluation indicators than the AW3D30 DEM and TANDEM-X data products. The average evaluation indicator R 2 was 0.98, RMSE was 1.44 m, and STD was 1.25 m. The results showed that ATLAS can be used as validation data to evaluate the accuracy of open-source DEM in estimating the vertical direction. At the regional scale, AW3D30 was the most robust and had the most stable performance in most of the tests (R 2 is 0.95, RMSE is 6.37 m and STD is 5.02 m) and is therefore the best choice for research on understory terrain. Slope had a greater impact on the accuracy of data estimation compared to aspect and surface vegetation type. The difference in RMSE for estimating forest understory terrain was 3.11 m for ATLAS products, 29.37 m for AW3D30 DEM data and 54.26 m for TanDEM-X data over the different slopes.