Mission Summary: UgCS-Powered Forest Inventory in BC's Lakes TSA
Key Results: UAV Photogrammetry Accuracy and Forest Health Data
- 0.92 correlation between UAV-derived volume estimates and ground-verified measurements across 30 plots
- 1.0 to 1.5 cm pixel resolution orthomosaics generated for each one-hectare plot
- Approximately 200 images per hectare captured with automated terrain following
- Only 56% of lodgepole pine trees remained healthy after 24 years (down from 74%), highlighting the scale of biotic and abiotic damage in managed stands
- Growth model forecasts only matched UAV-measured volumes when both updated Site Index estimates and revised disease mortality scenarios were applied
The Challenge: Accurate Timber Volume Estimation Across 30 Hectares of Managed Forest Under Climate Stress
British Columbia's managed forests are under pressure. Over the past two decades, the province has experienced an unprecedented mountain pine beetle epidemic, a globally significant outbreak of Dothistroma needle blight, and the four most extensive wildfire seasons on record (2017, 2018, 2021, and 2023). These combined disturbances have killed large volumes of mature timber, placing heavy reliance on the productivity of younger, managed stands for future timber supply.
The problem: nobody had been closely monitoring those managed stands. Growth and yield models used for timber supply planning in BC were built on historical data from undisturbed forests. They don't account well for the complex interactions between climate change, disease, insect damage, and abiotic events that are now affecting stands throughout the province.
In 1997, researcher Alex Woods and colleagues established 30 one-hectare stem-mapped plots in lodgepole pine stands aged 15 to 25 years across the Lakes Timber Supply Area. Every tree taller than 1.5 m was GPS-located and assessed for hard pine rust infections. Nearly 50,000 trees were mapped across the 30 plots.
Twenty-four years later, the team needed to return to all 30 plots to answer a direct question: how do the stands look now, and did the growth models get it right?
Visiting 30 one-hectare plots spread across a remote timber supply area is a significant logistical effort. Ground-based measurement of every tree in every plot would have taken months. The team needed a method that could estimate stand volumes across all 30 hectares efficiently, while still providing enough accuracy to benchmark against growth model forecasts.
The Solution: UgCS Photogrammetry with Terrain Following
The research team used a DJI Mavic 2 Pro drone with UgCS flight planning software (by SPH Engineering) to collect high-resolution aerial imagery for UAV photogrammetry processing across all 30 plots in spring 2021.
The paper describes the approach directly: imagery was collected using UgCS software (SPH Engineering) "with the Terrain Follow feature to autonomously control flight path, flight height, and flightline overlap to achieve a desired pixel resolution (GSD). The software adjusted flight speed and shutter interval to ensure sufficient forward overlap, appropriate image exposure, and image focus."
Three UgCS capabilities were central to the data collection workflow.
UgCS Terrain Following for Consistent GSD Over Forested Terrain
The 30 plots are distributed across the Lakes TSA, which includes rolling terrain typical of BC's central interior. To produce drone orthomosaics and canopy height models accurate enough for individual tree identification and timber volume estimation, the drone needed to maintain a consistent altitude above ground level throughout every flight.
UgCS terrain following uses digital elevation model (DEM) data to calculate a flight path that adjusts altitude continuously based on the terrain below. This kept the sensor at a constant height above ground across each one-hectare plot, producing a consistent ground sampling distance (GSD) of 1.0 to 1.5 cm per pixel, a critical requirement for accurate UAV photogrammetry in forestry applications. Without terrain following, altitude variation over uneven ground would have produced inconsistent image resolution, compromising the accuracy of the canopy height model and individual tree detection.
Automated Photogrammetry Grid Planning with UgCS for Per-Plot Image Capture
Each one-hectare plot (100 x 100 m) required approximately 200 photos to achieve the target GSD with sufficient forward and side overlap for photogrammetric processing. UgCS calculated the flight lines, speed, and shutter intervals automatically based on the camera specifications and desired overlap parameters.
A 20 m buffer was added around each plot boundary to ensure adequate coverage and minimize distortion during image stitching. This buffer is a standard practice in drone photogrammetry, but it requires the flight planning software to extend the survey grid beyond the actual area of interest.
Repeatable UAV Mission Planning Across 30 Distributed Forest Sites with UgCS
The team flew 30 separate missions across a large geographic area. Each mission needed the same parameters: consistent GSD, consistent overlap, consistent altitude above ground. UgCS allowed the team to configure these parameters once and apply them to each new site, changing only the survey area polygon and letting the software automatically recalculate the UAV flight plan for local terrain conditions - a key advantage for multi-site forestry drone surveys.
This consistency matters for scientific validity. When comparing UAV-derived volumes across 30 plots to ground-verified measurements and growth model forecasts, any variation in data collection method introduces noise. Standardized flight planning eliminates that variable.
The Results: UAV-Derived Timber Volumes Match Ground Truth
The georeferenced images from each plot were processed in Agisoft Metashape Professional to produce orthomosaics, digital terrain models (DTM), and digital surface models (DSM). The DTM was subtracted from the DSM to create a Canopy Height Model (CHM) for each plot. Individual trees were identified using a combination of pixel color, CHM height values, and a watershed segmentation algorithm.
To validate the UAV-derived estimates, the team established five randomly located 3.99 m radius ground verification plots within each one-hectare area. In these plots, every tree was measured by hand for height, DBH, species, and health condition. The ground-measured volumes were compared to the UAV-derived volumes for the same locations.
The results: a correlation coefficient of 0.9173 between the UAV-based and ground-based volume estimates. The team then used this relationship to calibrate the full one-hectare UAV estimates through a Ratio of Means adjustment, producing stand-level volume figures with quantified standard errors.
Forest Health Findings: Lodgepole Pine Mortality, Site Index Change, and Growth Model Accuracy
The validated UAV data told a sobering story about the condition of BC's managed lodgepole pine stands.
At plot establishment in 1997, 74% of lodgepole pine trees were healthy. By 2021, only 56% remained so. Comandra blister rust was the leading cause of mortality, responsible for 42% of all tree deaths. But the combined effect of other damage agents (western gall rust, stem forks, snow damage, mountain pine beetle, competition) actually contributed more to the overall decline in healthy tree counts than rust alone.
Site Index, a measure of site productivity that forest managers in BC had long treated as stable, changed considerably over the 24-year period. One-third of stands showed a Site Index increase of three metres or more. The most productive stand jumped from 19.8 m to 25.3 m, a 28% increase.
When the team benchmarked the BC government's standard growth model (TASS III v. 4.1.3.7) against the UAV-derived volumes, the model only matched reality when both updated Site Index estimates and revised disease mortality scenarios were applied. Using the original 1997 inputs without accounting for rust produced the weakest correlation (0.5854). Adding the current Site Index and disease adjustments brought it to 0.7348.
The conclusion from the paper is clear: growth and yield forecasts based on outdated Site Index values and models that ignore disease impacts do not accurately predict what is happening in these stands. Long-term monitoring, combining UAV data collection with ground verification, is essential for adaptive forest management under climate change.
UgCS for Forestry Drone Operators: Scalable Inventory from Consumer-Grade UAV Hardware
This study demonstrates a practical UAV forest inventory workflow for large-scale applications using consumer-grade drone hardware (DJI Mavic 2 Pro) combined with professional flight planning software (UgCS). The 0.92 correlation with ground truth across 30 plots confirms that UgCS terrain following and photogrammetry planning produce data accurate enough for scientific benchmarking against growth models.
For forestry operators, the key takeaway is efficiency. The UAV approach allowed the team to assess 30 widely distributed one-hectare plots and produce stand-level volume estimates with quantified precision. Individual tree-level measurements (health status, stem defects, disease identification) still required boots on the ground, but the volume estimation, canopy structure analysis, and mortality mapping were handled from the air.
The combination of UAV-derived canopy height models with targeted ground verification plots is a scalable model for forest monitoring programs that need to cover large areas without the cost of comprehensive ground surveys.
Technical Specifications
- Platform: DJI Mavic 2 Pro
- Flight Planning: UgCS (SPH Engineering) with Terrain Follow
- GSD Achieved: 1.0 to 1.5 cm/pixel
- Images per Plot: Approximately 200
- Plot Size: 100 x 100 m (1 hectare) with 20 m buffer
- Number of Plots: 30
- Processing Software: Agisoft Metashape Professional
- GIS Analysis: ESRI ArcGIS with Spatial Analyst extension
- Ground Verification: Five 3.99 m radius plots per hectare, hand-measured
- Growth Model Benchmarked: TASS III v. 4.1.3.7
Citation
Woods, A., McCulloch, L., Watts, M. and Di Lucca, M. (2026). Bridging the gap between forecast growth and realized loss in managed forests: The value of large-scale, long-term monitoring. The Forestry Chronicle, 102(1): 44-58.
