Mobile laser scanning in forests: Mapping beneath the canopy

Laser scanning is a modern technique used in forest mapping and monitoring. Airborne laser scanning (ALS) has been extensively used in medium-resolution, wide-area mapping. Terrestrial laser scanning (TLS), which delivers high-resolution data, has increasingly been used to map precise forest structural detail in discrete forest plots. Handheld mobile laser scanning (HMLS), a medium-to-high-resolution technique, sits between ALS and TLS methods and was first assessed in forestry in 2015. This thesis builds on existing research in evaluating the utility of HMLS sensors in forest mapping.

GeoSLAM ZEB-1 and ZEB-REVO HMLS sensors were deployed in a variety of UK forests and woodlands. Study sites were chosen to reflect a mixture of forest type, tree density, height, structural complexity and topographic variety. Scans were acquired alongside reference measurements and TLS scans. Experiments were devised to assess relative performance of HMLS sensors in the measurement of DBH, height, stem position, crown extent and volume with novel analyses performed in 3D Forest and CloudCompare.

Results indicated DBH could be measured to accuracies of 0.016-0.026 m RMSE, agreeing with existing research. HMLS tree position compared with TLS was subject to 0.1-0.3 m error but superior to manual techniques. HMLS sensors could not resolve tree height or crown structure due to limited range, identified as 10-12 m from sensor. A combination of HMLS and ALS data fusion yielded more accurate results determining height, crown extent and crown volume. These findings indicate that HMLS are suitable for sub-canopy forest mapping. HMLS sensors with increased range are becoming widely available in tandem with increasingly lightweight portable mobile laser scanning (MLS) solutions attached to UAV platforms. This research contends that HMLS will play a major role in multi-sensor integrated forest mapping and is ideal in supporting remote and proximal airborne mapping by providing a rich and accurate ‘ground-up’ dataset.