Degradação de redes de drenagem via upscaling das direções de fluxo a partir de MDT lidar de alta resolução espacial

Abstract

Digital Elevation Models (DEM) are a useful tool for hydrological studies, providing information about the terrain topography and allowing the extraction of hydrologic aspects of watersheds. The availability of free global spatial data with spatial resolution greater than 1 km, such as the SRTM DEM with resolutions of 30 and 90 meters, enables several application possibilities. However, in the case of large-scale hydrology, which works with basins larger than 10,000 km², there is a need to degrade the spatial resolution in order to reduce processing and storage costs. In this sense, there are scale-shifting procedures for degrading the spatial resolution, such as resampling the DEM and upscaling the flow directions. COTAT+ is an upscaling method that allows the degradation of spatial resolution from the adoption of two parameters: the minimum upstream flow path (MUFP) and the area threshold (AT) and was applied in this research to degrade a 1 m LiDAR DTM to 30 and 100 m spatial resolutions. The study area is a sub-basin (21.7 km²) inserted in the Riacho das Garças watershed (4,094 km²), in the semi-arid region of Pernambuco. A comparison was made between the performance of COTAT+ and resampling by the mean for extraction of drainage networks and watershed. For the application of COTAT+, two approaches were considered regarding the value of the parameters: (i) the adoption of the values indicated in a previous study, with another approach; and (ii) four variations for MUFP (1/10, 1/7, 1/4 and 1/3) and AT (-50%, -25%, 25% and 50%). The evaluation metrics used were visual inspection, mean distance (MD), percent within buffer (PWB), river lengths, and drainage area analysis. The results demonstrated the superior performance of upscaling over resampling in degrading the drainage networks, regardless of parameter variation, for the two resolutions considered. Furthermore, the drainage network plotting by resampling generated inconsistencies, such as displacements and spatial distance from the 1 m reference, as well as significant changes in river lengths.