3D mapping and modeling of wildland fuels

Airborne laser scans (ALS; also referred to as airborne LiDAR) have been a multi-purpose tool used for over a decade in the world of forest research and management, most commonly for forest structure analysis and inventory. More recently, ground-based terrestrial laser scans (TLS; also referred to as terrestrial LiDAR) have become more prevalent in forest research, able to measure surrounding vegetation to millimeter accuracy. While ALS has proven itself to be extremely useful, it has been found to be somewhat less accurate for vertical tree metrics and when used in dense forest types with thick canopies. TLS, by comparison and nature, could potentially amend the accuracy of these vertical metrics and provide researchers with highly accurate vegetation and stand metric data, as well as individual tree metrics.  

Image: Three computer generated scans of a tree. Visual progression from scan with foliage, the foliage removed, to the quantitative structure model output from TreeQSM.

With the recent acquisition of a Leica BLK360 TLS, this project was developed to compare the output from TLS scans of two tree species commonly found in Montana, both with and without foliage, to lab measured data. The Leica BLK360 is a compact laser scanner capable of capturing 360,000 points per second, and while originally developed for use in architectural surveying, has found a foothold in forest research. While laser scanners are historically quite expensive and too large to easily transport into the field, the BLK360 is a relatively inexpensive option in a much more portable size, and therefore may be a better option for smaller budgets and field crews. TreeQSM modeling method reconstructs the point cloud datasets from the scanned trees into quantitative structure models, assigning cylinders to branches and the stem to provide an estimate of the volumetric and geometrical details of the individual trees.

By comparing the error of the quantitative structure models to the actual trees, as well as between the species, the objective of this project was to provide a dataset of potential error when sampling under optimal conditions, and to provide data on model accuracy for species commonly found within Northern Rockies forest types.  

Approach

For this project, 20 small diameter trees (10 lodgepole pine (Pinus contorta); 10 Douglas fir (Pseudotsuga menziesii)) were cut at ground height and brought to the lab. As individuals, they were placed upright in a tree stand and measured for height, basal diameter above the lip of the stand, and diameter-at-breast-height (if applicable). Three branches from each tree were further measured for length and diameter at base and tip. The trees were scanned from opposing sides using a Leica BLK360 TLS, and then stripped of their foliage, with the needles from the identified branches being placed in separate paper bags than the rest of the tree. The bare trees were once more scanned from the same locations. Following the second scan, the trees were dismantled and placed into aluminum pans for placing in the drying ovens. The 3 branches were placed in individual pans. Foliage and woody fuel were dried for 5 days at 105 degrees Celsius, removed, and weighed. The Leica BLK360 scans were uploaded using the Autodesk Recap Pro program, and the point cloud datasets clipped in the program CloudCompare. TreeQSM was then run on the revised point cloud datasets and used to reconstruct the trees through 3D quantitative structure models. TreeQSM outputs of volume and structure will then be compared to the lab sampled trees as well as for the individually sampled branches, and error compared between species as well.