- Slow manual surveys
- Complex point cloud processing
- Dense, steep terrain
Manual surveys are time- and resource-intensive, processing UAV point clouds takes real expertise, and dense stands with overlapping crowns distort measurements.
Manual forestry surveys are slow and expensive, and tree detection in dense coastal BC stands is brutal — overlapping crowns and steep terrain distort every measurement.
We combined UAV imagery with GIS: digital surface, elevation and canopy height models, plus watershed analysis for crown isolation — validated against field survey data.


Part field work, part machine learning. Here’s how the forest got measured.
Manual surveys are time- and resource-intensive, processing UAV point clouds takes real expertise, and dense stands with overlapping crowns distort measurements.
A two-tiered methodology using Pix4D, Exelis E3D and ArcGIS: surface, elevation and canopy height models with raster math and watershed analysis, validated against field surveys.
I piloted the UAVs and designed low-altitude flight paths, trained AI models for tree detection and species differentiation, and reduced false positives by integrating field data.
Height accuracy within 5–20% — meeting or exceeding BC forestry standards — with less ground surveying, larger coverage, and a repeatable, partially automatable methodology.
Before the design systems and brand builds: flying drones over BC forests and teaching machines to count trees. Range is a feature.
MORE WORK ↗Have an interesting project? Feel free to reach out and talk it through. I’ll make it sharper.