At its core, LiDAR is a distance technology. A sensor — mounted on an aircraft, a drone, a car, or a tripod — emits short pulses of laser light. Those pulses travel outward, strike objects in the environment, and reflect back toward the sensor. The system records how long each round trip takes, and because the speed of light is a known constant, that travel time converts directly into distance. Repeat this process hundreds of thousands of times per second, and you end up with a dense three-dimensional map of whatever the laser touched.

The name is a parallel construction to radar and sonar — Light Detection and Ranging — and the underlying physics is similar in spirit. Radar sends radio waves; sonar sends acoustic waves; LiDAR sends pulses of light, typically in the near-infrared or green visible bands. What sets LiDAR apart is the wavelength. Light pulses are orders of magnitude shorter than radio waves, which means LiDAR can resolve features at centimeter precision rather than meter precision. That resolution is why LiDAR now underpins everything from archaeological surveys in dense rainforest to obstacle detection in autonomous vehicles.

LiDAR is best understood as a sampling tool. A typical airborne system fires more than 160,000 pulses every second, and at standard flying altitudes each square meter of ground ends up receiving somewhere around 15 individual laser hits. Multiply that across a survey area measured in square kilometers and you quickly arrive at the defining output of any LiDAR job: the point cloud, a dataset containing millions — sometimes billions — of discrete three-dimensional points.

Because the sensor lives on a moving platform, accuracy depends on more than just the laser itself. Well-calibrated airborne systems typically achieve vertical error around 15 cm and horizontal error around 40 cm. As the aircraft flies, the sensor sweeps side to side, meaning most pulses travel at an angle rather than straight down. The processing software has to account for the off-nadir geometry of each shot, which is why inertial measurement and GPS are as critical to a LiDAR deployment as the laser head.

Coverage per flight line is governed by swath width — the ground-distance footprint the sensor can scan in a single pass. Traditional linear-mode LiDAR typically delivers a swath of around 3,300 feet. Newer Geiger-mode systems, which use single-photon detection, can push that to roughly 16,000 feet. Wider swaths mean fewer flight lines per survey, which translates directly into lower acquisition cost for large-area mapping projects like statewide elevation models.

A raw point cloud is interesting, but what makes LiDAR valuable is the catalogue of derivative products you can extract from it. The same flight can produce a bare-earth terrain model, a full vegetation canopy profile, a land-cover classification, and a building footprint layer — all from one dataset.

One of LiDAR’s most useful quirks is that it can effectively see the ground beneath dense vegetation. The sensor is not x-raying through leaves; it is exploiting the small gaps between them. If you stand in a forest and look up, you can see patches of sky — those same patches let laser pulses slip down to the forest floor. Some pulses strike the outer canopy and reflect immediately. Others slip past the first layer and bounce off mid-level branches. A fraction travels all the way to the ground and reflects back as the final return.

Modern systems record the order in which each echo arrives — the “return number.” A single outgoing pulse can produce a first, second, third, and final return, each corresponding to a different structural layer of the vegetation. For foresters, this is invaluable: the distribution of returns reveals canopy density, vertical structure, and even species-level clues. For topographers, only the last returns matter, because those are the ones that reached the ground.

Not all LiDAR systems are built for the same job. They differ along three main axes: the size of the ground footprint each pulse produces, the wavelength of light used, and the platform on which the sensor is mounted. A handful of distinct categories have emerged over the decades.

LiDAR is no longer a niche geospatial tool — it is embedded in dozens of industries, each leveraging a different subset of its capabilities. Foresters use it to measure tree height, canopy density, and biomass without ever entering the stand. Self-driving car programs rely on compact solid-state LiDAR to detect pedestrians, cyclists, and curb edges in real time. Archaeologists have used airborne LiDAR to reveal Maya settlements buried under rainforest canopy, including networks of roads and causeways that had been invisible from above for centuries. Hydrologists delineate streams and watersheds from high-resolution DEMs that LiDAR makes possible.

Urban planners use LiDAR-derived DSMs to run solar-potential studies across entire cities. Coastal scientists use bathymetric LiDAR to map near-shore seafloor without deploying a vessel. Emergency managers generate flood-inundation models from ultra-accurate bare-earth terrain data. The list keeps expanding as the hardware shrinks and the price per survey falls.

The two technologies are often lumped together because both bounce a signal off an object and time the return. In practice they serve different purposes. Radar uses radio waves, which are far longer than light waves, so they travel further and penetrate cloud cover with ease — but at the cost of spatial resolution. Synthetic Aperture Radar (SAR) has become the mainstream airborne and spaceborne radar modality, and its side-looking geometry is a fundamental design choice: the oblique view lets the platform’s motion simulate a much larger virtual antenna, which in turn sharpens image resolution.

LiDAR, by contrast, typically fires straight down (or close to it) and produces a true 3D point cloud rather than a 2D image. If you need a centimeter-accurate model of a bridge, a forest plot, or a city block, LiDAR is the right tool. If you need to image thousands of square kilometers in any weather, through clouds, at the cost of lower spatial detail, SAR is the right tool. Many serious remote-sensing workflows combine both.

Raw returns arrive unlabeled. Classification is the process of tagging each point with a category — ground, low vegetation, medium vegetation, high vegetation, building, water, noise — so the cloud can be sliced into useful subsets downstream. The American Society for Photogrammetry and Remote Sensing (ASPRS) maintains the standard classification codes used in the industry-standard LAS file format.

Classification is partly automated and partly manual. Ground filtering algorithms handle the easy cases, and software packages like TerraScan can take care of most of a standard classification pipeline. Harder cases — distinguishing a dense shrub from a small tree, for example — may require manual QA, and the scope of classification is almost always negotiated in the contract before the flight takes place. A cloud delivered as “ground + unclassified” is a very different product from one delivered with seven fully populated ASPRS classes.