Improving the Accuracy of Global DEM of Differences (DoD) in Google Earth Engine for 3-D Change Detection Analysis

Digital elevation models (DEMs) represent the geospatial dataset core needed to model 3-D changes. The optimal dataset must be selected according to the environmental phenomenon under investigation as the offered resolution strongly affects the information level. Nonetheless, high-resolution DEMs are not available for the whole earth and, when not at disposal, open-source, medium-resolution, global DEMs may be a relevant source of knowledge. Because of the large amount of data and the vertical accuracy inhomogeneity, their applicability in defining 3-D changes of large areas is not predictable. The aim of this article is to explore global DEMs feasibility in detecting 3-D changes at the global scale and to examine the impact of filtering propagated error on 3-D changes. To achieve these goals, a Javascript code in the Google Earth Engine environment was developed. After recognizing AW3D30 (version 3.2) and NASA SRTM DEM (version 3) as the optimal DEM combination, their DEM of Differences was computed. Such a product was affected by many of Tukey's outliers, subsequently cleaned out. Three different statistical approaches, i.e., limit of detection, uniformly distributed error, and probability map, were compared to avoid artifacts propagating further. All adopted filtering strategies improve the results reliability albeit the third one is the most effective in mountainous, urban, and rural areas. The proposed research shows that the combined use of the two above-mentioned DEMs and appropriate filtering methods allows an effective description of 3-D changes. Moreover, it outlines that such analyses are also possible by using time-saving cloud-computing platforms.