Ground-return density and internal vertical differences in 1 m airborne lidar terrain models

Airborne lidar surveys are typically planned based on overall pulse density, yet digital terrain models (DTMs) rely exclusively on returns classified as ground. Because complex terrain and vegetation limit actual ground returns, high nominal pulse density does not guarantee adequate ground-return density. This study investigates how reducing the density of available ground returns impacts the internal consistency of 1-meter DTMs across diverse landscapes. Using data from two USGS 3D Elevation Program acquisitions in Utah and Idaho, four distinct sites were analyzed. At each location, 10% of the ground returns were withheld as a control set. The remaining 90% established a high-density reference DTM. This baseline dataset was systematically reduced using a spatially balanced thinning method with target densities ranging from 4 down to 0.03125 points per square meter. New DTMs were then generated via linear TIN interpolation. Results demonstrate that internal elevation differences increase steadily as point density declines. At 1 point per square meter, the Root Mean Square Error (RMSE) varied drastically by topography. It ranged from 0.010 meters on low-relief plains to 0.084 meters on steep escarpments. Errors peaked on steep slopes and in areas with dense non-ground features. Furthermore, purely random thinning increased the RMSE by 13% to 120% compared to spatially balanced thinning, proving that the spatial arrangement of points is as critical as the overall point count. Ultimately, these findings highlight DTM interpolation sensitivity and provide practical insights for evaluating surface reliability based on realized ground-return density.