Comparison of drainage network extracted from global digital elevation models with simple-flow and multi-flow direction

In 2022, the FABDEM model was released. Its developers used machine learning techniques to remove the height of forest canopy and buildings from the COPDEM digital elevation model (DEM). In this context, the objective of this study was to analyze and compare the drainage networks extracted from the SRTM, COPDEM and FABDEM global models using single and multiple flow direction methods, focusing on a floodplain area in the central-eastern region of the State of Paraná. Elevation and slope data from the three models were compared, and vertical accuracy was calculated from the root mean square error (RMSE) between each model and the reference data obtained from an official 1:50000 cartographic base. Drainage networks were extracted using different methods to define the flow direction and parameters of the algorithms implemented in GRASS GIS. The quality of the extracted drainage networks was analyzed by comparing them with the official 1:50000 hydrographic network and with topographic and longitudinal profiles. The results showed the higher vertical accuracy of FABDEM, as well as its better performance in extracting the drainage network when combined with the multiple flow direction method. However, the FABDEM model still has limitations when it comes to extracting drainage networks in floodplains, mainly because it was not successful to completely filter out the influence of vegetation in its altimetric data.