GISCI 241 Chap.7 LiDAR Returns and Elevation Models
LiDAR Returns and Elevation Models
LiDAR measures distance by transmitting laser pulses and recording returned energy with direction and timing. One pulse may interact with canopy, branches and ground, producing several returns that form a three-dimensional point cloud. This chapter distinguishes raw returns, classified points, surface models and terrain models.
It computes object height by subtracting terrain from surface and explains why point density, scan geometry, classification and interpolation control whether a fine-looking grid represents real observed detail.
What this chapter covers
- 01
Laser ranging and pulse travel time
- 02
Single and multiple returns through vegetation
- 03
Three-dimensional point clouds and classifications
- 04
Ground filtering and non-ground objects
- 05
Digital surface and terrain elevation products
- 06
Interpolation, point density and height uncertainty
Worked example · free
Compute an object-height profile
- +1First height equals 31.6 minus 18.2, giving 13.4 metres.
- +1Second height equals 29.4 minus 18.7, giving 10.7 metres.
- +1Third height equals 25.1 minus 19.0, giving 6.1 metres.
- +1State that aligned support, correct return classification and interpolation are required for interpretation.
Key terms
- Laser pulse
- A short transmitted burst of laser energy used to measure direction and range.
- First return
- The earliest recorded return from a pulse, often associated with the highest intercepted surface.
- Ground return
- A return classified as originating from terrain rather than vegetation, buildings or noise.
- Point cloud
- A three-dimensional collection of sampled positions with attributes such as return order and class.
- Surface model
- A gridded elevation product representing upper visible surfaces including objects and vegetation.
- Terrain model
- A gridded elevation product intended to represent bare ground after filtering non-ground returns.
- Interpolation
- Estimating a continuous surface between irregular sampled points using an explicit method.
LiDAR Returns and Elevation Models FAQ
Can LiDAR see through a forest canopy?
Some pulses pass through canopy gaps and produce later returns from lower vegetation or ground. Dense canopy can still block ground observations, so terrain quality varies with canopy, footprint, scan angle and pulse density. A ground model under forest may depend heavily on sparse returns and interpolation.
What is the difference between a point cloud and a DEM?
A point cloud contains irregular three-dimensional samples and classifications. A DEM is a continuous grid estimated from selected points. Creating the grid requires ground filtering, interpolation and a chosen cell size. The smooth DEM can therefore imply continuity and detail that were not directly sampled everywhere.
How is object height calculated?
When compatible grids are available, subtract ground or terrain elevation from upper surface elevation at the same horizontal support. The result can represent canopy or building height. Check projection, vertical datum, cell alignment, classification and edge effects before treating the difference as accurate physical height.
Why inspect a point cloud beneath an elevation raster?
The raster may look continuous even where ground returns are sparse or absent. Viewing return classes, density and cross-sections reveals which cells are supported directly and which depend on interpolation. That distinction is essential near dense canopy, steep breaks and building edges, where a smooth grid can conceal substantial uncertainty.
Exam move
Trace one emitted pulse through canopy and ground and label first, intermediate and last returns. Then separate four products on a sheet: raw return, classified point, point cloud and interpolated surface. Practise surface minus terrain calculations and state the compatibility conditions before subtracting.
Inspect imagined cross-sections where ground points are dense, sparse or missing and predict how interpolation uncertainty changes. When a fine DEM is shown, ask for point density and checkpoints before accepting its apparent detail. Finish each answer by identifying whether a feature was directly returned, classified, interpolated or inferred from a surface difference.
Draw three vertical profiles: open ground, sparse canopy and dense canopy. Place first, intermediate and ground returns on each, then identify where terrain is directly supported and where a gap must be interpolated. Build a product ladder from emitted pulse to return, georeferenced point, classified point, point cloud, surface model and terrain model. At every transition, write the decision or assumption introduced.
Practise height calculations using surface minus terrain at matching horizontal locations, and add a compatibility checklist for projection, vertical datum, cell alignment, time and interpolation method. Compare two ground filters on a steep bank: an aggressive filter may remove real terrain, while a permissive one may retain vegetation. Explain how each error changes the final height.
For grid resolution questions, separate cell size from point spacing and vertical accuracy. A small output cell can only interpolate the available returns. Finish by designing a validation plan with independent checkpoints, density maps, class inspection and cross-sections through difficult terrain. State which locations carry direct evidence and which depend most heavily on model assumptions.