The forest, measured tree by tree.
Woodflock is building drone swarms that scan forests with LiDAR and cameras. Each survey becomes a record of every tree we can see: where it stands, how tall and thick it is, what species it is, and what changed since the last flight.
Forests change faster than anyone measures them.
Satellites see the canopy from far away. Field crews measure single plots by hand. The trees in between go years without a measurement.
Global forest area in 2025, about 32% of all land.
Deforested each year from 2015 to 2025, before counting regrowth.
Tree cover lost worldwide in 2025. Fire caused 42% of it.
How often the US national forest inventory remeasures a plot: every 7 years in the East, 10 in the West.
Many small aircraft, two altitudes, one tree list.
This is the system we are building. Each part has been shown to work on its own in published field research; our job is to make them work together, every day, at the scale of a real forest.
Sweep the canopy
A group of drones flies parallel lines above the trees with LiDAR. Splitting the area between aircraft shortens the time in the air, and the dense point cloud gives tree positions, heights, crown shapes and the ground beneath them.
Fly between the trunks
From above, trunks hide under the crown. Smaller drones go under the canopy without GPS, navigating by their own sensors, to measure stem diameter where it is actually visible.
Turn points into a tree list
Computer vision separates individual trees, labels species, flags dead and stressed crowns, and compares each survey with the previous one, so a change in land use shows up as a list of specific trees and plots.
Why “Woodflock”
A flock moves as one without a leader, each bird keeping clear of its neighbours. Our drones fly the same way, as a flock that lives and works in the woods: out between the trees, then back to a field base to recharge and offload data, which is where most of the time in a multi-drone survey is won or lost.
What published research already shows.
None of these results are ours. They are peer-reviewed measurements by university and government research groups, listed with the conditions they were made under so you can judge how far each one carries.
| Measurement | Result | Conditions |
|---|---|---|
| Sensing: what drone LiDAR and cameras can measure | ||
| Separating individual trees in drone LiDAR | 85.1% F-score | Deep learning on a benchmark of drone LiDAR from five countries. |
| Stem diameter, flying under the canopy | 0.60–0.92 cm RMSE | Two boreal plots in Finland, manually piloted LiDAR drone, checked against ground laser scans. |
| Stem diameter, flying above the canopy | about 6.0 cm RMSE | Spruce and pine stands in the Czech Republic. The gap between this row and the one above is why we fly both altitudes. |
| Tree height | −20 cm bias | Under-canopy drone LiDAR in a sparse boreal plot; −65 cm in an obstructed plot. |
| Stem volume per tree | about 10% std. error | Same Finnish study, from under-canopy LiDAR alone. |
| Tree species from drone photos | 0.73 mean F1 | Nine species plus other classes, German temperate forest, image detail finer than 2 cm. Falls to 0.26 at 32 cm. |
| Bark beetle damage in spruce | 0.76 / 0.88 / 0.93 F1 | Infested, dead and healthy trees, best result with hyperspectral drone images. |
| Autonomy: flying in and over forest without a pilot on the sticks | ||
| A swarm flying through forest | 10 drones | Palm-sized drones crossed a bamboo forest through gaps of about 30 cm, using onboard cameras and no GPS. |
| Speed between trees | 0 crashes at 3–5 m/s | Vision-only drone in natural environments; 8 of 10 runs succeeded at 7 m/s and 60% at 10 m/s. |
| Mapping with no GPS signal | 1,157 trees mapped | A 1.1 km loop under pine canopy in the US where GPS never locked. Tree-based mapping cut position drift from 10.1 m to 4.0 m. The flight was manually piloted. |
| Autonomous flights under canopy | 7 of 7, then 8 of 9 | Successful flights in spruce at 650 and 2,000 trees per hectare, with a camera-only drone under 1 kg. |
| Economics: what drone data replaces | ||
| Field plots needed for a timber valuation | 160 → 60 plots | A 40 ha Finnish property. Adding drone data kept the same precision and saved a two-person crew 8 to 10 workdays. |
More drones finish sooner, up to a point
Total time to survey 181 hectares, in minutes, by number of drones. This is a simulation, and it counts launch and battery changes as well as flight.
Nine drones did the job 4.7 times faster than one. Past nine, getting aircraft launched and swapped cost more time than the extra aircraft saved.
Tree detection depends on the forest
F-score for separating individual trees in drone LiDAR, by where the data was collected. Same method, same benchmark.
Tall, straight conifers are the easy case. Structurally complex forest is harder, which is why we plan to start in managed conifer stands.
The parts nobody has finished yet.
The pieces exist. The whole system does not, and we would rather say so than have you find out in the field.
No swarm has yet inventoried a whole forest
The largest forest swarm we can find in the literature is ten drones, and it navigated without mapping. The most recent review from Finland’s geospatial research institute says flying inside forests still relies mainly on manual piloting. Closing that gap is the company.
Small trees and summer leaves hide a lot
On one Norwegian plot the same method that found 83% of dominant trees found 39% of understory trees. In a US plantation, trunks could not be detected from above under full leaf cover. Under-canopy flights and leaf-off surveys are how we plan to recover them.
Ground handling limits a fleet, not flight
In the simulation charted above, 15 drones were slower than 9, and on a 22 hectare site the best fleet was only three aircraft. Automated charging and data offload at the field base matter as much as the aircraft.
Rules for one pilot, many drones are still forming
The US aviation regulator proposed a rule for routine flights beyond line of sight in August 2025; it was not final as of November 2026. In Europe an operator can declare how many aircraft one pilot supervises as part of a risk assessment, and the regulator’s own example is five. We plan around small pilot-to-drone ratios first.
Anyone who has to know what is standing on their land.
Timberland owners
Stand volume and growth from measured trees. About 1.2 billion hectares of forest worldwide are designated mainly for production.
Commodity supply chains
The EU deforestation regulation applies to large and medium companies from 30 December 2026. It asks for the location of each plot and proof it was not deforested after 2020.
Carbon and restoration projects
Repeat surveys of the same trees, so claims about growth and survival rest on measurements a buyer can inspect.
Public forest agencies
Faster revisits after fire, storm and pest outbreaks. Insects, disease and severe weather damaged about 41 million hectares of forest in 2020.
Have a forest that needs measuring?
We are looking for a small number of landowners and agencies to fly pilot surveys with us. Tell us where the forest is, how big it is, and what you need to know about it.
hello@woodflock.com