A drone cannot see through vegetation, but a drone-mounted LiDAR can measure the ground through gaps in it. Each laser pulse can return several echoes, and the last echo from a pulse that reached the ground gives a bare-earth point. Photogrammetry cannot do this: it models the canopy surface. The resulting ground model depends on canopy density, season and point density, and is verified with check points on the ground.

This guide expands the short answer on our LiDAR and laser scanning service page and supports the BNG survey and topographic survey services.

Why can’t photogrammetry see the ground under vegetation?

Photogrammetry reconstructs the surface the camera can see. Over a grassed embankment that is the top of the grass; over a hedgerow it is the top of the hedge; over woodland it is the canopy. The software has no way of knowing that the ground lies below. On a grassed or scrubbed embankment this produces an elevated false surface, which is why our reservoir survey guide treats photogrammetry alone as unsuitable for freeboard work.

How does LiDAR record the ground beneath a canopy?

A LiDAR pulse has a physical footprint on the ground. When part of the footprint strikes foliage and part passes through a gap, the sensor can record more than one echo from the same pulse: a first return from the canopy, intermediate returns from lower branches, and a last return from the ground. Full-waveform and multi-return sensors capture this. Ground classification then selects the returns that belong to the terrain and rejects the rest.

The important word is gaps. LiDAR does not pass through leaves or timber. The density of ground returns is governed by how much open sky the ground can “see” from the aircraft, which is why the factors below matter more than any headline sensor figure.

What determines how much ground is recovered?

FactorEffect on ground returnsWhat the surveyor controls
Canopy density and speciesDense, layered or evergreen canopy leaves fewer gapsChoice of season; multiple passes from different directions
SeasonLeaf-off deciduous canopy is far more open than leaf-onFlying window; winter and early spring preferred for woodland DTMs
Point densityMore pulses per square metre give more chances of a gapPulse rate, flying height, ground speed, overlapping passes
Flying height and footprintLarger footprint at height mixes canopy and ground in one pulseFlying height chosen for the specification
Scan angleOblique pulses meet more foliage than near-vertical onesOverlap so that most ground is seen near nadir from at least one strip
UnderstoreyBracken, bramble, gorse and reed form a second canopy close to the groundReported as a limitation; check points placed within it
Terrain slope and roughnessSteep aspects thin out returns on one sideCross passes on steep catchment boundaries
WaterThe laser is absorbed at the surfaceBathymetric survey where bed levels are needed

There is no fixed penetration percentage that can be quoted for a site in advance. What can be stated is the method, the season and the density flown, and then what the check points showed.

What can be expected from different vegetation types?

VegetationExpectation for a bare-earth DTM
Short grass, mown verge, stubbleGround effectively at the surface; DTM behaves like open ground
Long grass, crops, rushGround returns present but reduced; the DTM under standing crop should be treated as soft detail
Hedgerow lines, scattered treesGround recovered either side and normally beneath, subject to density; the hedge base and any ditch profile need checking
Deciduous woodland, leaf-offUsually the best woodland case; useful DTM for drainage, flood and archaeological work
Deciduous woodland, leaf-onReduced returns; the DTM is coarser and should be flagged
Conifer plantationClosed canopy year-round; returns limited to rides, gaps and thinned stands
Gorse, bramble, bracken, reedbedDense low cover close to the ground; the hardest case and often the widest residuals

Because these expectations vary so much, a rigorous survey places some independent check points on the soft-detail ground and reports them separately from the check points on hard detail. Our guide on how accurate a drone LiDAR survey is explains why.

How is the ground separated from the vegetation?

  1. Strip adjustment. Overlapping flight lines are matched so the whole cloud is internally consistent and tied to the ground control.
  2. Noise removal. Isolated high and low points (birds, multipath, reflections) are removed.
  3. Automated ground filtering. A progressive algorithm grows a ground surface from the lowest reliable points, accepting points within slope and distance thresholds tuned to the terrain.
  4. Manual editing. Automated results are reviewed in a hillshaded 3D view and corrected where they fail most often: hedge bases, ditch inverts, embankment toes, cut faces and stream banks.
  5. Classification to standard classes. Ground, low, medium and high vegetation, buildings, water and overhead lines are coded to the ASPRS LAS classification scheme so that downstream users can filter the cloud.
  6. Product generation. A DTM from the ground class, a DSM from the full cloud, and a canopy height model (CHM) as the difference between them.

The canopy height model is itself a deliverable. On BNG and habitat work it gives hedgerow heights, woodland structure and tree positions on a dated, georeferenced base; on landscape and visual work it drives screening assessments.

Where are the limits?

  • LiDAR does not penetrate water, closed evergreen canopy, or dense low cover with no gaps.
  • The DTM under cover is an interpolation between recovered ground points; where they are sparse, subtle features such as shallow ditches or ridge-and-furrow can be lost.
  • Leaf-on woodland surveyed for programme reasons should carry a stated limitation.
  • Check points under cover are harder to observe by GNSS because the same canopy degrades the satellite signal; our GNSS-degraded environments guide covers how control is carried in.

How Angell Surveys does this

Angell Surveys is regulated by RICS and flies UAV LiDAR from the Wingtra Ray fixed-wing VTOL alongside photogrammetry, so vegetated and open ground are captured on the same control in the same sortie. On a water company’s two reservoir sites in the Fens and south Lincolnshire, mature hedgerow boundaries and small woodland blocks completely obscured the ground from photogrammetry; first, intermediate and last-return classification produced the bare-earth model that the hydraulic modellers and archaeologists worked from, delivered as a tiled, ASPRS-classified point cloud with a canopy height model. For a rivers trust’s upland catchment in Stirlingshire, where the best public terrain data was 30 metre resolution, the LiDAR was specified with overlapping passes across the steep boundaries to keep return density on every aspect and delivered a sub-50 cm DTM. Every ground model we issue carries an accuracy statement against independent check points and states its season and limitations.

Sources