How much data is enough? Sensor choice, scale, and point-density effects on terrain metrics in steep mountain forests

Abstract

ABSTRACT Accurate digital terrain models (DTMs) are essential for terrain analysis and planning in steep forest environments, yet their reliability depends on sensor type, ground-point density, and spatial resolution. This study assesses how the sensor type (airborne laser scanning (ALS), unmanned aerial vehicle LiDAR (UAV LiDAR), handheld LiDAR (HLS), and UAV photogrammetry via structure-from-motion (UAV-SfM)), along with ground-point density and grid resolution, affects (a) elevation accuracy, (b) slope-based terrain class aggregation, and (c) slope and terrain-variability metrics in steep closed-canopy forests. Using full-density UAV LiDAR as reference, vertical deviations on plot-scale (0.5–4 m grids) were quantified and linked to terrain/canopy factors such as slope, roughness, topographic wetness index (TWI) and composite terrain complexity index (TCI). Stand-scale analysis (0.5–30 m grids) evaluated sensor, density, and resolution effects on slope metrics, variability, and data volume. At plot scale, HLS and UAV LiDAR provided consistent terrain representations below dense canopy, outperforming ALS and UAV-SfM; roughness was strongest predictor of vertical error across sensors. At stand scale, coarser grids (2–4 m) with 5–10 ground points ${\rm m}^{-2}$m−2 achieved ${\sim } 90\percnt$∼90% agreement with reference, reducing volume and smoothing slope extremes and steep terrain. The terrain ruggedness index (TRI) correlated with slope across resolutions, indicating that it captures similar information.