Digital Elevation Models (DEM) consist of the digital representation of elevation values at different points in a specific geographic area. The use of a DEM requires an explicit definition of the physical surface to be modeled, as the term is generic and represents any surface, whether a digital surface model (DSM), which describes the forest canopy and other artificial or natural objects above ground, or a digital terrain model (DTM), which represents altimetric values at ground level. The main products available globally and free of charge, such as the SRTM, are DSM, while many geoscientific studies require ground surface recognition, which is only possible from an DTM. In plains, this problem is accentuated, since the vegetation height, the canopy cover and the errors of the elevation measurement technique can be greater than the real altimetric amplitude of the topographic surface, creating false reliefs and compromising a correct geomorphological interpretation. As an area with extensive vegetation cover, seasonally flooded, difficult to access and extensive periods of precipitation and cloud cover, the Amazon Coastal Zone is challenging for field studies and even for remote sensing analyses. In this context, this study aimed to evaluate and compare the performance of eight DEMs, one DTM obtained from airborne P-band radar and seven DSMs (AW3D30, ASTER GDEM, Copernicus DEM, NASADEM, SRTM, Topodata and an MDS obtained from X-band airborne radar) in the morphological characterization of a floodplain on the Amazon coast. A floodplain on the outskirts of the municipality of Mazagão, in the southern portion of the coast of Amapá, Brazil, was selected as a test area. All DEMs were resampled to the same 30 m mesh size and compared by visual control and statistical analysis based on slope elevation. The behavior of automated extraction from the hydrographic network was also analyzed. The comparison showed that the DTM obtained from P-band radar images was the most consistent with respect to terrain shapes, as it is less sensitive to vegetation. It was also found that even the DTM was not able to detect drainage lines or features corresponding to very small elevation variations. Rather than requiring more refined techniques or better spatial resolutions, which can result in unaffordable operational costs, we suggest that using external 2D data such as satellite imagery or existing databases can provide implicit 3D mapping for watershed modeling. hydrographic surveys in areas where elevations are not accurate enough. The approach was also applied in the detection and characterization of paleodrainages in lowland regions. These features are typically marked by the presence of vegetation on the banks and/or in the center of the paleochannels and record the evolution of river courses throughout the Quaternary. Although DSM are more suitable for this type of analysis, it was possible to verify, with the selection of altimetric points of the DTM, that they are likely to be recognized from the ground surface, although the product error is greater than the height variation. from the margins to the center of the paleochannels.
|
