Effects of digital elevation model resolution on derived stream network positions

[1] This study examines the effect of digital elevation model (DEM) resolution on the positional accuracy of derived hydrologic networks and quantitatively confirms that DEM resolution should be greater than the average hillslope length when used for hydrologic modeling. Seven hundred kilometers of mapped stream networks are compared to stream networks derived from 17 DEMs with resolutions ranging from 30 m to 3 km. Comparison between predicted and mapped streams reveals that accuracy of predicted stream locations decays quickly beyond a DEM resolution of 180 m. A new application of Gyasi-Agyei et al. 's [1995] DEM resolution suitability test based on average slope and vertical resolution indicates an average hillslope length of between 150 and 180 m. Tarboton et al. 's [1991] method of determining average hillslope length based on slope and accumulation areas reconfirms the length to be 150 m and verifies the link between network accuracy and hillslope scale. Two algorithms are used to derive stream networks: the D-infinity algorithm [Tarboton, 1997], which allows for flow dispersion, and the D8 algorithm [O'Callaghan and Mark, 1984], which does not. A comparison between the D8 and D 1 algorithms shows that modeling flow dispersion is not necessary in steep terrain, as both algorithms performed equally well. However, the D-infinity algorithm is found to be less susceptible to modeling erroneous hillside flow convergence, making it preferable to the D8 algorithm when used by Tarboton et al. 's [1991] method of determining the average hillslope lengths. Finally, field observation indicates that predicted first- and second-order streams tend to exist when predicted; however, mapped streams have a higher positional accuracy.