 Solution

Use airborne lidar with sufficient pulse density and ground-return classification. Laser travel times give ranges; the aircraft’s surveyed position and orientation place returns in three dimensions. Multiple returns include canopy and, where gaps permit, the ground. Classify ground points, remove vegetation returns and interpolate a bare-ground digital elevation model over the $20\mathrm{k}{\mathrm{m}}^2$ strip.

This specifically addresses the forest-obscured fault scarp, unlike ordinary optical stereo, which primarily models the visible canopy. Leaf-off acquisition can improve ground sampling where relevant. Dense understorey and a complete lack of ground returns still cause uncertainty; interpolation does not create observed ground heights. Inspect ground-point coverage and compare hillshades under several illumination directions.^[ref]