Mountain unit extraction based on surface runoff simulation

Object-based terrain analysis represents a significant advancement in digital terrain analysis; however, prior research has predominantly focused on slope facets, while mountain-scale terrain analysis remains largely unexplored, primarily due to the lack of automated methods for extracting mountain units. This study presents a novel approach for mountain unit extraction based on surface runoff simulation. The fundamental rationale is as follows: terrain feature points and lines—including peaks, valleys, and ridges— are first derived from DEM data, and a topological ridge network is constructed to automatically delineate saddles along ridges connecting adjacent peaks. Subsequently, steepest descent paths are traced from each saddle perpendicular to both flanks of the ridge, using the D8 flow-routing algorithm. The resulting steepest descent paths, combined with valleys, collectively delineate enclosed mountain unit boundaries. Two governing parameters—saddle cut depth and flow steepness—are introduced to enable multi-scale mountain unit extraction at varying levels of granularity. An experimental site characterized by pronounced topographic gradients and well-developed mountainous terrain was selected in Xinlong County, Ganzi Tibetan Autonomous Prefecture, Sichuan Province. Results demonstrate that across the approximately 9,000 km² study area, the proposed method successfully identified 1,152 peaks and 1,100 valid saddles. By systematically varying screening parameter combinations, 272 (coarse-grained), 510 (medium-grained), and 801 (fine-grained) mountain units were delineated, yielding a hierarchical mountain partitioning framework. Terrain statistical analysis conducted at the mountain-unit scale revealed that as segmentation granularity increases, the standard deviations of mean elevation, slope, and aspect consistently rise, reflecting the scale-dependency effect intrinsic to geographical phenomena. A comparative assessment of three distinct spatial analysis units— mountain units, catchments, and regular rectangular grid cells—further indicates that mountain units exhibit the lowest dispersion and highest stability in terrain parameter statistics relative to catchments and grid cells. This study thus provides spatial analysis units that are more consistent with the intrinsic characteristics of natural geomorphic entities for applications in mountain geomorphology, ecosystem management, and geological hazard assessment.