This study focused on examining impacts of DEM resolution on stream delineation. Original and resampled DEMs constructed from three different types of source datasets – LiDAR, InSAR and cartographic data – were used to generate stream networks based on real-world data. Multiscale statistical characterizations and analyses were performed in conjunction with spatial mapping of stream networks to characterize and compare how the stream networks changed with respect to a.) DEM resolution and generation method; b.) terrain and c.) stream geometric complexity. Spatial and statistical results of generated streams were compared to real-world imaged and digitized stream data to evaluate statistical and spatial accuracy of the generated stream data. Statistical results showed that multiscale statistical methods, based on area-scale analyses and characterizations proved able to characterize and distinguish surface complexities between DEM resolutions and DEM-generated streams at specific scale ranges. Comparison with spatial results showed, however, that in stream delineation, these multiscale characterization methods may prove limited and problematic, because the statistical results generated from area-averaging characterizations may not indicate spatial sensitivity of the results. Therefore, in order to determine and validate spatial accuracy in stream delineation, spatial location analyses must also be performed in conjunction with statistical analyses in modeling contexts where spatial location matters.