A LiDAR point cloud looks, at first glance, like a ghostly sculpture made of dots, and that impression is not far wrong: each dot is a single laser measurement, recording where a pulse of light struck a surface and returned. Airborne LiDAR, flown on a plane or drone, can record millions of such points over a landscape in a single survey, and together they describe the ground, the trees, the buildings, and everything else in three dimensions. Learning to read a point cloud means learning what each point knows: its position, its height, and its classification.

A LiDAR sensor fires rapid laser pulses toward the ground and times how long each pulse takes to return, converting that time into a distance with the speed of light. Combined with the sensor's position from GPS and its orientation from an inertial unit, each return becomes a point with precise three-dimensional coordinates. A single pulse can produce several returns, first from the top of a tree canopy, then from branches, then from the ground, which is why LiDAR sees both the forest and the forest floor in one pass, where a photograph would show only the canopy.

The density of points matters: a survey at a few points per square metre describes terrain well, while a dense survey at dozens of points per square metre can reveal individual trees, power lines, and building edges. File formats like LAS and LAZ store the points with attributes including intensity, the strength of the return, and classification codes that label each point as ground, vegetation, building, water, or noise. Those classifications are usually assigned by software and refined by hand, and they are what turn a raw cloud into a usable dataset.

The most common products derived from a point cloud are elevation models. A digital surface model takes the highest points and describes everything, including trees and buildings, while a digital terrain model filters down to ground points only and describes the bare earth beneath. The difference between the two is the height of everything standing on the ground, which is why subtracting one from the other gives a canopy height model, a map of tree heights across a forest that foresters use to estimate timber and habitat.

In QGIS, point clouds can be opened and visualised directly, with colouring by height, intensity, or classification, and profile tools let you slice through a cloud to inspect a hillside or a building in cross-section. For heavy processing, dedicated tools handle filtering, thinning, and classification more efficiently, but visualisation and basic measurement in QGIS are enough to learn the structure of a dataset. Start by colouring a cloud by classification and toggling ground points on and off: watching the trees vanish to reveal the bare earth is the fastest way to understand what LiDAR offers.

LiDAR excels wherever elevation and structure matter: flood modelling, drainage design, forestry inventory, archaeology under tree cover, and mapping terrain too steep or vegetated for traditional survey. Archaeologists have famously used it to find ancient earthworks and settlements hidden beneath forest canopy, because the ground model reveals shapes invisible in aerial photographs. Engineers use it to design roads and drainage on accurate terrain, and ecologists use canopy models to map habitat structure at landscape scale.

Its limits are cost, data volume, and the fact that it measures shape, not colour or chemistry: a point cloud will not tell you whether a crop is healthy or a roof is leaking, and a single survey is a snapshot that ages as the landscape changes. For many projects, a LiDAR survey every few years, combined with frequent satellite monitoring in between, is the practical balance. If you are new to the field, download a free sample dataset for an area you know, open it in QGIS, and explore: the moment the dots resolve into a landscape you recognise, the technology clicks.