- Entry
- 023
- Source scale
- 0.5 m grid
- Method
- Airborne laser, classified returns
- Section
- Canopy

Removing the trees from a model returns ground that has not been looked at directly in a very long time.
What the pulse finds that the photograph cannot
Dense woodland has always frustrated the map. Aerial photography, however good the camera, returns a picture of the canopy surface — a textured green roof that tells you almost nothing about what is underneath. Contours drawn from stereo photographs in forested terrain are, at best, interpolations across a surface the operator cannot actually see. For most of the twentieth century, that was the limit.
Airborne LiDAR changed the constraint. A laser pulse fired from a low-flying aircraft does not behave like a camera shutter; it is not blocked by the canopy as a whole. A proportion of pulses — enough pulses, when the flight is planned correctly — pass through gaps in the leaf layer, strike the ground, and return a last return echo that the instrument can separate from the canopy return above it. The classifier then strips canopy hits and produces a bare-earth model: a digital terrain surface representing the ground, not the trees.
What that surface reveals has surprised archaeologists, foresters, and surveyors in roughly equal measure.
- Airborne LiDARlaser pulses fired from low-flying aircraft; multiple returns from a single pulse separate canopy from ground
- Last returnthe final echo from a pulse, most likely to represent ground rather than vegetation
- Bare-earth modelpoint cloud with canopy hits removed; reveals ground surface beneath trees
- Pulse densityexpressed in pulses per square metre; determines how completely the ground is sampled under canopy
- Ground-point classificationautomated algorithm sorting returns into ground and non-ground; introduces its own errors
Relict systems that survived because of the trees
Under woodland that had been continuous for centuries, the laser found earthworks that open farmland would have erased. Ridge-and-furrow cultivation, deserted medieval settlements, hollow ways, field lynchets, prehistoric enclosure boundaries — all of them preserved as low earthworks because no plough had touched them since the land was taken out of cultivation and the trees grew back. The relief is real: measured in centimetres to a metre or two of vertical relief, sometimes less, but entirely visible in a bare-earth model rendered with oblique hillshading.
The paradox is a useful one. Precisely the land that was hardest to survey by any pre-LiDAR method turned out to hold the densest archive of how older landscapes were organised. Cartographers had mapped the wood; nobody had mapped what was inside it.
This is not a minor footnote. Survey programmes in Britain, France, and the Low Countries have collectively added thousands of previously unrecorded features to the heritage record from LiDAR data alone — features that lie under forest that was, on the printed map, simply a green wash or a woodland symbol with no interior detail at all.

Every sheet crosses a table like this one before it is printed, and picks up a decision at each pass.
What the model depends on
The quality of the bare-earth surface depends on several things that are worth holding in mind when interpreting results.
Pulse density matters. A flight planned for broad land-cover mapping — perhaps two or four pulses per square metre — may be too sparse to thread enough returns through a closed canopy to give confident ground coverage everywhere. Surveys intended for archaeological or detailed terrain use commonly specify ten pulses per square metre or higher, which increases the probability of ground hits in even dense stands.

A compiled sheet describes ground the compiler is not standing on.
Season matters, particularly for deciduous woodland. A flight in winter, when broadleaf canopy is open, produces far better ground penetration than one flown in July over the same stand. Conifer plantations — dense, year-round, with lower branches close to the ground — remain the hardest case: some areas under dense spruce or fir return very few ground hits at all, and the classifier is forced to interpolate rather than measure.
The classification algorithm itself introduces uncertainty. Automated ground-point classification uses assumptions about slope continuity and point-cloud geometry that can misread a sharp earthwork edge as noise, or smooth a genuine feature into the surrounding terrain model. Manual review of the point cloud, and comparison against the derived model, remains part of rigorous practice.
None of this undermines the basic achievement: bare earth under closed canopy is now, for the first time in the history of the map, something that can be measured rather than guessed. The contours that will eventually be drawn from that surface will describe ground that no surveyor has stood on, and no photograph has ever seen — produced by a counting of echoes from a pulse of light that threaded its way through the gaps in the leaves.
- Ridge-and-furrow, hollow ways, deserted settlements, lynchets, enclosure earthworksall found under continuous woodland cover
- Preserved because ploughing stopped when the trees took over; earthworks survived intact for centuries underneath
- Features were, on printed maps, simply absentshown only as generic woodland symbol