Skip to content
Artificial intelligence. Clearly. RSS feed
AI NEWS AI

News. Evidence. Perspective.

ResearchExplainer

The young trees rising above Scotland’s heather

An aerial survey mapped 2.65 million young trees in the Cairngorms. A combination of laser mapping, geometric processing and machine learning reveals where woodland is returning—and why the first metre of growth is hardest to see.

A returning forest can be surprisingly hard to count. In the Cairngorms, small, scattered trees disappear into the heather that surrounds them, concealing the very regeneration land managers want to measure.[1]

Researchers have now mapped that emerging woodland across a 569-square-kilometre survey area. Their analysis identified more than six million trees, including 2.65 million below five metres tall. More than 5,200 hectares outside existing forest had estimated densities above 100 young trees per hectare—a substantial area where natural regeneration is taking hold.[2]

The study, published in the Journal of Applied Ecology on 23 September, analyses an aerial laser survey commissioned in July 2023. These figures describe the woodland visible in that survey, with field measurements used to check and calibrate the analysis.[1]

Making small trees look tall

LiDAR builds a three-dimensional picture from laser returns. But the usual processing can erase low vegetation by treating it as part of the ground. The team’s solution begins with a deliberately artificial landscape: it stretches the vertical structure of the point cloud, making the shrub layer resemble a forest of much taller objects.[3]

That mathematical stretching helps the software keep ground and vegetation separate. It can then recover a height map without flattening away the small features that matter. The open-source lidarSHM package also interpolates the terrain model to reduce this loss of low vegetation. Its documentation advises field calibration because laser-derived heights can still be underestimated.[3]

From the resulting map, the researchers located treetops and outlined crowns. Shape filters helped remove bracken. Machine learning had a specific supporting job: a random-forest classifier, trained on hand-labelled image pixels, rejected likely rocks. The complete pipeline combined geometry, aerial imagery and field ecology.[1]

The height at which a forest becomes visible

The results

Detection improved sharply as trees cleared the vegetation

Detection improved sharply as trees cleared the vegetation
Height of tree or tallest tree in clusterClusters detected
Below 1 metreLess than 1%Source 1
1.4–1.6 metres83%Source 1
Above 1.8 metresMore than 95%Source 1

This makes the map especially useful for following trees that have emerged above surrounding vegetation. The very smallest seedlings remain a job for closer inspection. Restoration has several stages; a method that records established young growth can reveal progress that a map of mature forest misses.

The geography of that growth was revealing. Young trees were concentrated near existing woodland, where seed supplies are close, though some established more than half a kilometre away. Drier heathlands supported more regeneration than bogs. A continuous map can show these scattered pioneers as well as the expanding edge of an established wood.[2]

Helping a landscape grow its own woodland

Cairngorms Connect’s restoration strategy gives those patterns a practical purpose. The partnership aims to double its existing forest, relying largely on natural regeneration. Deer management reduces browsing pressure so seedlings can survive and grow. The wider work includes native broadleaved trees and high-altitude woodland, as well as Scots pine.[4]

Where the right seed sources are missing, the partnership uses targeted planting, including plants raised from locally collected seed and cuttings. In this approach, planting and natural regeneration serve different conditions. One establishes a future source of seed; the other allows an existing source to spread.[4]

A detailed regeneration map could help distinguish those situations. A sparsely wooded hillside containing scattered young trees presents a different restoration question from one with no local seed source. Counting the small trees makes that distinction visible while there is still time to decide where help will matter.

Follow the evidence

Sources & further reading

  1. Tracking and modelling native woodland expansion in the Scottish Highlands using LiDAR Journal of Applied Ecology · 23 Sep 2026 · Accessed 18 Oct 2026
  2. Successful woodland expansion in Scotland’s Cairngorms revealed in aerial maps University of Cambridge / Conservation Research Institute · 28 Sep 2026 · Accessed 18 Oct 2026
  3. lidarSHM Aland Chan / GitHub · Accessed 18 Oct 2026
  4. Forest Restoration Cairngorms Connect · Accessed 18 Oct 2026