What is LiDAR terrain data?

Active measurement

LiDAR sends light out and measures what returns.

A LiDAR instrument emits short laser pulses towards a surface. The sensor records returned light and uses travel time to calculate range. Position and orientation measurements locate the instrument when each pulse was sent and received. Combining those observations places a return in three-dimensional space.

An airborne survey repeats this process across overlapping flight lines. The first result is not a smooth image or grid. It is a point cloud: a large set of irregularly distributed positions, each carrying coordinates and potentially other attributes from the measurement and processing chain.

LiDAR is the observing method. A terrain model is one product made from its observations.

What reflected the pulse

One survey can contain ground, canopy, roofs and water responses.

A pulse may meet the top of vegetation, travel through a gap and produce later returns from branches or ground. A roof, rock face or open field produces a different pattern. Survey systems and product specifications decide which returns are recorded and how they are described.

Processing then assigns classes such as ground, vegetation, building, water or noise. Algorithms use the shape and relationship of neighbouring points, but classification can also involve breaklines, reference data and manual correction. A point labelled ground is therefore an interpreted measurement, not an object the laser identified by name.

Build the ground

A DTM keeps ground evidence and fills the spaces between it.

To derive bare-earth terrain, processing excludes vegetation, buildings and other non-ground points. The remaining points do not usually form a perfect regular grid, so software interpolates heights between them. Breaklines can preserve sharp linear features such as banks or channel edges that a simple grid would soften.

A DSM uses a different selection intended to describe upper surfaces. Both outputs can be produced from the same broad survey, yet show sharply different heights in woodland or towns. DTM vs DSM explains how those surface choices affect a route.

The grid spacing of a delivered terrain raster is also a processing decision. It should reflect the survey specification and intended use, but a smaller cell does not by itself certify the point density, classification quality or vertical accuracy behind it.

Where the returns thin out

Laser coverage is detailed, not omniscient.

• Dense vegetationFew pulses may reach the ground, leaving wider spaces for interpolation and harder classification.
• WaterReflection and absorption can produce sparse or unsuitable topographic returns, so mapped water surfaces may receive separate treatment.
• OcclusionObjects and steep terrain can hide surfaces from a particular viewing angle. Overlapping passes help but do not make every geometry visible.
• Vertical facesA raster that stores one height per horizontal cell cannot preserve an overhang or two surfaces at the same position.
• TimeThe point cloud describes the collection period. Forestry, construction, erosion and landslip can change later.

Processing can also leave artefacts: fragments of removed objects, false pits, seams or over-smoothed features. Quality assurance reduces these problems, but users still need the product's metadata and stated limitations.

Read before rendering

Find the lineage between survey and surface.

1. Form

   Identify point cloud, DTM, DSM or derived visual.

   The word LiDAR can be attached to every stage, although the files contain different information.

2. Capture

   Check survey date, coverage and platform.

   A national composite may join work collected at different times.

3. Process

   Read classifications and surface treatments.

   Ground filtering, water handling, bridge removal and interpolation shape the terrain output.

4. Quality

   Keep point density, raster resolution and accuracy separate.

   They describe different parts of the dataset and cannot substitute for one another.

The beacen Library uses terrain to make route form legible.

Trace any precise claim through the delivered surface, its classifications and the survey metadata. A rendered planning surface is a processed derivative, not direct access to an unedited LiDAR point cloud.