Point Cloud Data in GIS

A point cloud is the closest thing GIS has to a direct measurement of reality. Instead of generalised lines and polygons drawn by a cartographer, it holds millions of individual points, and each one marks a place where a laser beam or a camera ray actually hit a surface. Raw point clouds are large and unstructured, so their value appears only after processing, when points are classified and turned into terrain models, building heights or volumes.

What Is a Point Cloud?

Every point stores X, Y and Z coordinates in a defined coordinate reference system. Survey data usually carries several more attributes:

Intensity records how strongly the surface reflected the laser signal.
Return number shows which echo of a single pulse the point represents, because one pulse can return several times as it passes through a tree canopy.
Classification marks the point as ground, vegetation, building or water.
RGB colour comes either from a camera on the same platform or from imagery draped afterwards.

Point density, measured in points per square metre, decides what can be detected. A national airborne survey delivers a few points per square metre, which is enough for terrain and building outlines. A mobile scan of a street delivers hundreds, enough to measure individual signs, cables and kerbs.

A colour coded LiDAR point cloud of an Sydney Opera

How Point Clouds Are Captured

LiDAR measures distance by timing a laser pulse, and combined with GNSS positioning it turns every measurement into a georeferenced point. Four platforms cover most projects. Airborne scanning from an aircraft suits regional terrain mapping, forestry and flood modelling, while a drone covers small areas such as construction sites and quarries at much higher density. Mobile scanning from a vehicle records roads, railways and urban corridors at traffic speed, and terrestrial scanning from a tripod handles buildings, industrial plants and heritage objects.

Photogrammetry builds a point cloud from overlapping photographs instead. It costs less and gives naturally coloured points, but it reconstructs only what the camera sees. It cannot measure the ground under a forest canopy and it handles water and glass poorly. That difference usually decides which method a project needs.

Point Cloud Formats

Format choice affects file size, compatibility and whether the data can be streamed to a web map at all.

Format	What it is	When to use it
LAS	The ASPRS binary standard for LiDAR, storing coordinates, intensity, returns and classification	Exchange and archiving when uncompressed size is acceptable
LAZ	Lossless compression of LAS, usually 5 to 10 times smaller with identical content	The default choice for storage and delivery
COPC	A valid LAZ file with an internal octree index	Streaming from object storage without downloading the whole file
E57	A vendor neutral format for terrestrial scanning that also stores images and metadata	TLS projects, BIM and as built documentation
PLY, XYZ	Simple point formats, often without coordinate system metadata	Quick exchange and 3D modelling software
3D Tiles, EPT	Tiled structures with levels of detail for web delivery	Publishing large clouds in a browser viewer

In practice the archive stays in LAZ, publication goes through COPC or tiles, and everyday GIS work runs on derived rasters and vectors.

Classification

Classification assigns a meaning code to every point, and this step turns a measurement into information. The ASPRS specification defines the standard codes. The most used are 2 for ground, 3 to 5 for low, medium and high vegetation, 6 for building, 9 for water, 11 for road surface and 13 to 15 for wires and transmission towers.

Software makes the first pass automatically, and an operator then corrects problem areas such as steep slopes, bridges and dense canopy. Errors made here travel into every later product, because a misclassified bridge deck becomes a wall across the terrain model.

From Points to GIS Products

Most GIS work happens on what is derived from the cloud rather than on the cloud itself:

A DTM interpolated from ground points gives the bare earth surface for hydrology and engineering design.
A DSM covers the top surface with buildings and vegetation included, which suits visibility and solar analysis.
Subtracting the DTM from the DSM gives object and canopy heights directly.
Building footprints and 3D models come from points classified as buildings.
Comparing two surveys measures excavation, stockpiles, erosion or subsidence.

These are ordinary rasters and vectors, so they can be styled, analysed and published like any other layer.

Where Point Cloud Data Is Used

Topography. High accuracy elevation models, slope analysis and flood modelling.
Forestry. Canopy height, timber volume estimation and detection of clear cuts.
Urban planning. Building heights for 3D city models, shadow studies and visibility analysis.
Infrastructure. Power line corridor inspection, road asset inventories and clearance checks under bridges.
Construction and mining. Progress monitoring, stockpile volumes and slope stability.
Archaeology. Earthworks hidden under vegetation, which LiDAR reveals where imagery cannot.

Point Clouds in a Web GIS

A browser is not the place to process raw point clouds, and that work stays in specialised software such as PDAL, LAStools, CloudCompare, QGIS or ArcGIS Pro. It is, however, where the results need to end up, because colleagues and clients look at data there.

The usual division of labour keeps filtering, classification and surface generation on the desktop, then exports the resulting DTM and DSM as GeoTIFF together with building footprints and canopy metrics as vector layers. Those layers are published in a web GIS such as GISCARTA, where they can be styled, combined with other data, viewed in 3D and shared as a link or an embedded map. The full cloud stays in object storage as COPC or tiles for the cases when someone needs the original detail.

A LiDAR derived surface model and building layer on the GISCARTA platform

Common Mistakes

Treating the raw cloud as a deliverable. Clients need derived surfaces and measurements, not a multi gigabyte LAZ file.
Ignoring the vertical datum. Horizontal coordinates are usually set correctly, and the height reference is where errors appear quietly.
Trusting automatic classification without inspection. Bridges, steep terrain and dense vegetation are the classic failure points.
Ordering density by ambition rather than by task. Higher density means longer surveys, slower processing and larger storage, with no gain if the deliverable is a one metre DTM.
Forgetting the capture date. A cloud recorded three years ago describes a city that has since changed.

FAQs

Can point clouds replace traditional surveying?

They complement it. Coverage and detail are exceptional, but control points and conventional measurements are still needed to establish and verify accuracy.

How are point clouds displayed on a web map?

Either as derived rasters and vectors, which any web GIS can show, or by streaming a tiled structure such as COPC or 3D Tiles into a dedicated 3D viewer. Loading a raw LAS file into a browser is not practical at production sizes.

Can point clouds be opened in QGIS?

Yes. QGIS loads LAS, LAZ and EPT data and renders it in both 2D and 3D, with colouring by classification, elevation, intensity or RGB. On first load a LAS or LAZ file is indexed into EPT, which takes time on large files but makes every later access fast.