Acquisition Systems

The first engineering decision on any project is which platform to fly, drive, or carry, and the wrong choice cannot be fixed in post-processing. An ALS survey of a heritage façade returns one or two points per square metre where the conservator needs a thousand, and no amount of densification will recover the stone joints that the beam never illuminated. A TLS campaign over a 50 km$^2$ watershed costs a small fortune in tripod setups and still misses the canopy interior that an airborne pulse would have penetrated through gaps in the leaves. A handheld SLAM device captures a building interior in twenty minutes that would take a static scanner two days, and then the trajectory drifts a centimetre or two over the long traverse and the absolute accuracy collapses just below the threshold the heritage office requires. This chapter walks through each acquisition system, what it measures well, and what its silent failure modes are: an ALS scanner sweeping the ground from AGL at 1 to 100 pts/m$^2$, a TLS instrument logging a million points per second at millimetre noise from a tripod, an MLS profiler streaming road corridors at traffic speed, a UAV LiDAR pod giving 100 to 500 pts/m$^2$ from above the canopy, and a handheld SLAM walking through interiors where GNSS cannot reach. I will also cover photogrammetric SfM/MVS clouds because the practical question is increasingly not LiDAR-or-not but which combination of sensors meets the brief at the lowest defensible cost.

In productionThis video is being recorded and will appear here when it ships.

This chapter covers the complete acquisition system: platform, navigation subsystem, and operational procedures for airborne, terrestrial, mobile, and bathymetric systems, followed by emerging technologies including solid-state sensors, multispectral LiDAR, and UAV-borne systems.

Airborne Laser Scanning (ALS)

Airborne Laser Scanning is the most widely used platform for large-area topographic mapping. National mapping agencies, from the US Geological Survey's 3D Elevation Program (3DEP) to the Netherlands' AHN programme, rely on ALS to generate country-wide elevation datasets. The scanner is mounted in the belly pod or cabin floor of a fixed-wing aircraft (for large-area surveys) or a helicopter (for corridor or high-density urban surveys), flying at altitudes typically between AGL.

System Components

A modern ALS system integrates five tightly coupled subsystems:

1. Laser scanner unit: produces laser pulses, directs them through the scanning mechanism, and detects reflected signals. Modern units achieve pulse rates of .

2. GNSS receiver: a multi-frequency, multi-constellation receiver (GPS, GLONASS, Galileo, BeiDou) determines the 3D position of the aircraft. Raw carrier-phase observations are typically logged at 1-10 Hz for post-processed kinematic (PPK) or real-time kinematic (RTK) solutions.

3. IMU: measures angular rates and accelerations at , from which the platform attitude (roll, pitch, heading) is derived.

4. Control and recording unit: synchronises all subsystems via a common GPS time base and records raw data for post-processing.

5. Camera (optional): acquires co-registered imagery for point cloud colourisation or orthophoto generation.

Figure pending

Typical Specifications

covers conventional linear-mode systems; single-photon and Geiger-mode systems (discussed in ) extend these parameters significantly.

Table pending

Single-Photon and Geiger-Mode ALS

The single-photon and Geiger-mode detector technologies introduced in have been deployed operationally in airborne systems. The Leica SPL100 (single-photon) splits each emitted pulse into $10 \times 10 = 100$ beamlets, operates from altitudes up to AGL, and achieves effective measurement rates exceeding 6 million points per second. The L3Harris Geiger-mode systems (formerly Harris IntelliEarth) operate at even higher altitudes (up to ${\sim}\SI{8500}{\metre}$ AGL) with wide swaths.

These systems offer dramatically higher area coverage rates than linear-mode ALS, making them attractive for national mapping programmes. The US Geological Survey's 3DEP programme has evaluated both technologies for nationwide elevation data acquisition. However, the data have characteristics that require adapted processing workflows: higher noise levels (especially from solar background), the absence of multiple returns per pulse, and intensity information that is binary (detected/not detected) rather than analogue. Post-processing pipelines for SPL and Geiger-mode data rely heavily on statistical filtering algorithms to separate signal from noise, typically exploiting the spatial coherence of surface returns versus the random distribution of noise photons.

Flight Planning

A successful ALS survey begins with flight planning: the process of determining the flying height, speed, strip layout, and overlap that will deliver the required point density and accuracy. Three parameters dominate this calculation. Adjacent flight strips typically overlap by 30-60 % of the swath width, partly to guarantee complete coverage despite cross-wind deviations during the flight and partly to provide redundant data for strip adjustment downstream. The required point density is set by the application: digital terrain model generation can be satisfied with 1-2 pts/m$^2$, whereas urban asset inventory typically demands 20 pts/m$^2$ or more. Finally the scan angle sets the trade-off between coverage and quality: wider angles increase swath width but degrade edge accuracy through longer slant ranges, larger incidence angles, and more frequent occlusion, so most surveys limit the half-angle to between $15^\circ$ and $20^\circ$.

The average point density at nadir can be estimated from the system parameters:

$$\rho_{\text{pts}} = \frac{f_{\text{pulse}}}{v \cdot W}$$

where $f_{\text{pulse}}$ is the effective pulse repetition rate (Hz), $v$ is the ground speed (m/s), and $W$ is the swath width (m).

Worked example: flight planning..

We want to survey a 50 km$^2$ area at a minimum point density of 8 pts/m$^2$. Our ALS system has $f_{\text{pulse}} = \SI{500}{\kilo\hertz}$, the aircraft flies at $v = \SI{70}{\metre\per\second}$ (about 135 knots), and the swath width at $h = \SI{800}{\metre}$ AGL with a $\pm 20^\circ$ scan angle is:

$$W = 2\,h\,\tan(20^\circ) = 2 \times 800 \times 0.364 \approx \SI{583}{\metre}.$$

The point density at nadir is:

$$\rho_{\text{pts}} = \frac{500{,}000}{70 \times 583} \approx 12.2\;\text{pts/m}^2.$$

This exceeds the 8 pts/m$^2$ requirement. To cover 50 km$^2$ with 50% strip overlap, we need roughly $50 \times 10^6 / (583 \times 0.5) \approx 172$ strip-kilometres of flying. At 70 m/s, this takes about $172{,}000 / 70 \approx 2457$ s $\approx$ 41 minutes of acquisition time (plus turns and transit).

Rule of thumb for ALS surveys. A rough planning check: the point density at nadir is $\rho \approx f_{\text{pulse}} / (v \cdot W)$. Higher density can be achieved by (1) flying slower, (2) flying lower (smaller $W$), or (3) using a higher-pulse-rate scanner. Each has trade-offs: slower flight increases cost and wind sensitivity; lower altitude reduces coverage per strip; faster pulse rates may reduce maximum range.

shows what such a plan looks like on a real survey area, with the parallel acquisition strips at the planned flying height, the swath width of each strip, the overlap zones where adjacent strips share coverage, the turning radii at each end of the block where the aircraft repositions for the next strip, and the cross strips flown perpendicular to the main pattern for strip-adjustment calibration overlaid on the orthophoto. Block design at this level of detail is what distinguishes a survey that delivers the specified density from one that requires a costly re-flight.

Real data pending

Real data

Terrestrial Laser Scanning (TLS)

Terrestrial Laser Scanning (TLS), also called static laser scanning, is the tool for mapping objects and structures. Mounted on a tripod at a fixed position, a TLS scanner captures a dense, millimetre-accurate 3D model of its surroundings: building façades, heritage monuments, industrial plants, tunnel walls, rock faces, or crime scenes.

Panoramic Scanning

Most modern TLS systems perform a panoramic (hemispherical or near-spherical) scan by sweeping the laser beam through two angular directions:

• Horizontal rotation: the scanning head or an internal mirror rotates a full $360^\circ$ about the vertical axis.

• Vertical deflection: a fast-spinning mirror deflects the beam vertically, typically covering $-45^\circ$ to $+90^\circ$ (or the full $300^\circ$ range in some systems).

The result is a point cloud organised in a spherical coordinate system $(r, \theta, \phi)$, where $r$ is range, $\theta$ is the horizontal angle, and $\phi$ is the vertical angle. Because of this regular angular sampling, the structured grid can be projected into a 2D panoramic (equirectangular) image for visualisation and analysis, a representation sometimes called a range image or depth panorama.

Figure pending

shows what a real TLS station looks like in operational setup, with the scanner on a survey-grade tripod and retro-reflective targets at known positions distributed around it so that the next station's scan can be bound back into the same coordinate frame. The number, geometry, and visibility of those targets directly drive the registration accuracy that the later chapter on registration formalises, and a station with too few or poorly distributed targets produces a scan that downstream algorithms cannot bind reliably to its neighbours.

Real data pending

Real data

Typical Specifications

Modern TLS instruments group into two categories by ranging principle:

• Phase-shift scanners (e.g., FARO Focus, Leica RTC360, Z+F Imager): very high measurement rates (up to 2 million pts/s) and sub-millimetre ranging noise at short to medium range (up to $\sim$ 100-350 m). These dominate for indoor surveys, forensics, heritage documentation, and construction.

• Time-of-flight (pulsed) scanners (e.g., Riegl VZ-6000, Leica ScanStation P50): longer maximum range (up to 6 km for the VZ-6000 on high-reflectivity targets; effective range on natural surfaces is typically 2-3 km) but typically lower point rates and slightly higher ranging noise. These are preferred for mining, topographic survey, and large-scale infrastructure.

A recent trend is the emergence of hybrid scanners that combine both ranging principles, using phase-shift for short-range precision and pulsed ToF for long-range capability, in a single instrument.

summarises typical TLS performance specifications.

Table pending

Multi-Station Surveys and Registration

A single TLS scan captures only what is visible from one viewpoint; anything behind an object, around a corner, or on the far side of a pillar remains unmeasured. Complete coverage of a building interior, for example, may require 10-50 scan positions. These individual scans must then be registered (aligned) into a common coordinate system. Three main approaches exist:

1. Target-based registration: reference targets (spheres, checkerboard planar targets, or tilt-and-turn targets) visible from adjacent scan positions define a rigid-body transformation that achieves sub-millimetre residuals but requires careful target placement.

2. Cloud-to-cloud registration: algorithms such as ICP (Iterative Closest Point, ) align overlapping point clouds without dedicated targets, but require sufficient geometric overlap and can fail in feature-poor environments.

3. Survey-control registration: scanner positions measured by total station or GNSS provide absolute georeferencing, typically combined with target-based or cloud-to-cloud methods for refinement.

How many targets per scan? A rigid-body transformation in 3D has six degrees of freedom (three rotations, three translations), so theoretically three non-collinear targets suffice. Always use at least four targets per scan pair to provide redundancy and enable error detection. Place targets at different heights and distances to improve the geometric rigour of the solution.

Mobile Laser Scanning (MLS)

Mobile Laser Scanning captures high-density point clouds (comparable to TLS) over extended corridors by mounting the scanner on a moving platform: usually a vehicle, but increasingly also backpacks, trolleys, UAVs, and boats.

Vehicle-Based MLS

The most common configuration mounts one or more 2D profile scanners on the roof of a survey vehicle. As the vehicle drives along a road at normal traffic speed (), each scanner sweeps a vertical plane perpendicular to the direction of travel. The forward motion of the vehicle "stretches" these 2D profiles into a 3D point cloud.

Vehicle-based MLS excels at mapping road corridors (road surfaces, lane markings, kerbs, signs, utility poles, and building façades) at densities of hundreds to thousands of points per square metre. The primary limitation is that the scanner can only see what is visible from the road: building rooftops, courtyards, and areas above the roofline are invisible.

Figure pending

shows what such a survey looks like on a real urban map: the trajectory is colour-coded by GNSS quality and IMU drift so that the parts of the cloud relying on dead reckoning under bridges or in canyons are visually distinct from the parts with a clean fix. Long sections benefit from a high-quality GNSS fix, but under bridges and in tunnels the IMU bridges the outage, the solution uncertainty grows visibly during that stretch and collapses again once GNSS reacquires lock, and the downstream point-cloud accuracy at any point along the trajectory tracks this uncertainty directly.

Real data pending

Real data

MLS Specifications

Table pending

A distance measurement indicator (DMI), or wheel encoder, is typically integrated with the GNSS/INS solution to provide along-track position updates. The DMI is particularly important in GNSS-degraded environments such as urban canyons and tunnels, where it constrains the inertial navigation drift.

Handheld and Backpack Systems

Scanners small and light enough to carry by hand or in a backpack have been available since the 2010s. These systems rely on Simultaneous Localisation and Mapping (SLAM), an algorithm that estimates the scanner's trajectory and builds the map at the same time, using only the scanner's own measurements and an IMU, with no external GNSS.

The SLAM algorithm works by continuously matching features (planes, edges, corners) between successive scans, so that as the operator walks through a building, the algorithm incrementally extends the trajectory and the point cloud. When the operator returns to a previously visited area, these loop closures allow the algorithm to detect and correct accumulated drift.

The trade-off these systems offer is sharp. Operating without GNSS makes them ideal for interiors, tunnels, mines, and urban canyons where airborne and vehicle-borne systems fail entirely; an operator can map a building interior in minutes rather than the hours needed for multiple TLS setups. The cost is positional accuracy: SLAM typically delivers compared to the millimetres reachable by static TLS, and drift accumulates in geometrically sparse environments such as long uniform corridors where the algorithm has few features to constrain itself against. In my own work I have found that the gap between the manufacturer's quoted figure and the field result depends almost entirely on whether the operator returns to the start position and forces a loop closure, and on whether the long sections of the traverse are anchored against control targets that a TLS or total station has measured beforehand.

UAV-Based LiDAR

Unmanned Aerial Vehicles (drones) equipped with lightweight LiDAR scanners combine the bird's-eye perspective of airborne scanning with the low altitude and flexibility of a small platform. Because UAVs typically fly at only AGL, the short range to the ground allows them to achieve very high point densities, often 100-500+ pts/m$^2$, along with centimetre-level accuracy.

UAV LiDAR has rapidly become the tool of choice for:

• Topographic surveys of small to medium areas (up to a few km$^2$),

• Forestry and vegetation structure analysis (the laser penetrates canopy from above),

• Corridor mapping (power lines, pipelines) where manned aircraft are too expensive,

• Archaeological site documentation,

• Post-disaster damage assessment (rapid deployment).

The main limitations are flight endurance (typically 15-40 minutes per battery), payload capacity, and regulatory constraints (flight permissions, visual line of sight, altitude limits).

Figure pending

summarises the typical operating parameters across all platforms.

Table pending

Bathymetric LiDAR

Bathymetric LiDAR extends the technology into the underwater domain by exploiting a key physical property: water is opaque to near-infrared light () but partially transparent to green light ().

A bathymetric LiDAR system emits two wavelengths simultaneously:

• ** (infrared):} reflects off the water surface, which provides the surface elevation.

• ** (green):} penetrates the water column and reflects off the seabed or riverbed.

The water depth is the difference between the two ranges, corrected for the refractive index of water ($n_w \approx 1.333$, meaning $c_{\text{water}} \approx c/1.333 \approx 0.75c$):

$$d_{\text{water}} = \frac{c_{\text{water}} \cdot (t_{\text{bottom}} - t_{\text{surface}})}{2}$$

In addition, Snell's law must be applied to correct for the refraction of the green beam as it crosses the air-water interface. Without this correction, the apparent position of the seabed point is wrong both vertically and horizontally.

Figure pending

Operating bathymetric LiDAR involves four practical considerations that distinguish it from topographic acquisition. The most fundamental is the maximum reachable depth, which scales with water clarity, typically one to three Secchi depths. In clear oceanic water this allows depths of 40-50 m to be mapped, but in turbid coastal or estuarine waters the practical limit drops to 5-10 m. The point density obtainable is correspondingly lower than in topographic LiDAR because the green beam loses energy as it propagates through the water column. Refraction at the air-water interface introduces a second concern: the green beam bends as it crosses the interface, and without explicit correction (using Snell's law and a measured water-surface model) every below-surface point ends up displaced both vertically and horizontally. The reward for managing these constraints is unique: bathymetric LiDAR is the only technology that produces a continuous elevation model from dry land through the surf zone to the seabed in a single coherent acquisition.

The land-water transition. Traditional surveying struggles in the surf zone: airborne LiDAR cannot see through water, and boat-mounted sonar cannot operate in very shallow water. Bathymetric LiDAR fills this gap and is used for coastal erosion monitoring, flood modelling, and habitat mapping.

Photogrammetric Point Clouds (SfM/MVS)

While this book focuses on LiDAR, a large fraction of the 3D point clouds produced worldwide come from photogrammetry: extracting 3D geometry from overlapping photographs. Because readers will often encounter photogrammetric data alongside LiDAR data, or need to choose between the two, this section describes the photogrammetric pipeline and the characteristics of the resulting point clouds.

The SfM-MVS Pipeline

Photogrammetric 3D reconstruction follows a two-stage pipeline:

1. Structure from Motion (SfM) recovers camera poses and a sparse 3D point cloud from unordered photographs: - Detect 2D keypoints (SIFT, SuperPoint) in each image. - Match keypoints across image pairs using descriptor similarity. - Estimate relative camera geometry from matched pairs using the essential matrix and RANSAC. - Bundle adjustment: jointly optimise all camera poses and 3D point positions by minimising reprojection error across all images simultaneously. This is a large-scale nonlinear least-squares problem, typically solved with the Levenberg-Marquardt algorithm. The output is a set of calibrated camera poses (intrinsics and extrinsics) and a sparse point cloud (typically $10^4$-$10^5$ points).

2. Multi-View Stereo (MVS) densifies the sparse SfM reconstruction: - For each image, estimate a dense depth map by matching pixels across neighbouring views using patch-matching or learned stereo algorithms. - Fuse the per-image depth maps into a single consistent 3D point cloud, removing inconsistent measurements. - The result is a dense point cloud, typically $10^7$-$10^9$ points, with RGB colour for every point.

Figure pending

Photogrammetric vs. LiDAR Point Clouds

summarises the key differences. The choice between photogrammetry and LiDAR depends on the application: LiDAR excels where vegetation penetration, millimetre accuracy, or low-light operation is needed; photogrammetry is preferred when colour is essential, the budget is limited, or sensor weight is constrained. I find that the comparison is no longer the clean qualitative dichotomy that the textbooks of a decade ago could draw, and that for a heritage façade with strong texture and stable lighting a careful photogrammetric capture often delivers a cloud that is geometrically indistinguishable from a TLS scan once both are subsampled to a common resolution.

Table pending

Common Software

• COLMAP: Open-source SfM + MVS; the reference implementation used by most research papers. Provides the camera poses for NeRF () and Gaussian Splatting.

• Agisoft Metashape: Commercial; widely used in surveying and heritage documentation. Supports GCP integration and direct georeferencing.

• OpenDroneMap: Open-source drone mapping pipeline; produces orthomosaics, DSMs, and point clouds from UAV imagery.

• Pix4D: Commercial; optimised for drone surveys with RTK/PPK integration.

Many modern workflows combine LiDAR and photogrammetry: LiDAR provides the geometric backbone (accurate terrain, vegetation penetration), and photogrammetry adds colour and texture. In drone mapping, some platforms carry both a LiDAR sensor and a camera, producing a fused product that inherits the strengths of both modalities. When processing fused data, apply LiDAR-specific techniques (multi-return analysis, intensity-based classification) to the LiDAR component and colour-based methods to the photogrammetric component.

Platform Comparison

compares platform capabilities against the survey requirements that drive platform selection: coverage, density, accuracy, and budget.

Table pending

Figure pending

shows the five platform families in operational settings: an airborne sensor under the wing of a survey aircraft, a vehicle-mounted mobile mapping rig running a city corridor, a tripod-mounted terrestrial scanner on a heritage site, a UAV-borne sensor on a quadcopter, and a backpack or handheld walkaround scanner used for indoor and small-site capture. Each platform fills a distinct accuracy-density-coverage envelope, and the choice of platform is the first engineering decision in any project.

Real data pending

Real data

Coordinate Reference Systems

Before georeferencing can be discussed, the coordinate reference system (CRS) in which the point cloud will be delivered must be understood. Three levels of coordinate definition are relevant to LiDAR practice:

• Geographic coordinates (latitude, longitude, height): Defined on an ellipsoidal datum, most commonly WGS 84 (used by GNSS) or a regional realisation such as ETRS89 in Europe or NAD83 in North America. Heights above the ellipsoid (ellipsoidal height $h$) differ from the physically meaningful orthometric height $H$ (height above the geoid) by the geoid undulation $N$:

$$h = H + N$$

Geoid models (e.g., EGM2008, GEOID18) provide $N$ at any location. Failure to apply the geoid correction can shift the entire point cloud vertically by up to , depending on location.

• Projected coordinates (easting, northing): Map projections convert geographic coordinates to a 2D Cartesian system suitable for engineering work. UTM (Universal Transverse Mercator) is the most common: it divides the Earth into 60 zones of longitude, each with its own Transverse Mercator projection. Within a single zone, distortion is below 0.04%, making UTM suitable for most LiDAR projects. For projects spanning two UTM zones, a custom Transverse Mercator or Lambert Conformal Conic projection may be needed. Many countries define national projected systems (e.g., Lambert 93 in France, RD New in the Netherlands, State Plane in the USA).

• Local (project) coordinates: For engineering surveys and TLS projects confined to a small area, a local Cartesian system is often used, with an arbitrary origin and no map projection distortion. When the point cloud must later be integrated with GIS data, a transformation to a standard CRS is required.

Always record the CRS metadata (EPSG code or WKT definition) and the vertical datum used. A point cloud delivered in "UTM Zone 31N, ETRS89, orthometric heights (EGM2008)" is unambiguous; a cloud with no CRS metadata is a recurring source of errors when data are combined from multiple sources. In LAS files, the CRS is stored as a GeoTIFF VLR or a WKT VLR ().

Georeferencing Strategies

Georeferencing transforms measurements from the sensor's own coordinate frame into a global reference frame (e.g., a national map projection or an Earth-centred Cartesian system) so that the resulting point cloud can be combined with other geospatial datasets. The fundamental georeferencing equation for a kinematic platform was introduced in and is reproduced here in the form most convenient for operational sensor orientation:

$$\vect{p}_{\text{global}}(t) = \vect{p}_{\text{GNSS}}(t) + \mat{R}_{\text{IMU}}(t) \bigl[ \vect{l}_{\text{lever}} + \mat{R}_{\text{bore}} \vect{p}_{\text{scan}} \bigr],$$

where $\vect{p}_{\text{GNSS}}(t)$ is the GNSS-derived position of the navigation centre at time $t$, $\mat{R}_{\text{IMU}}(t)$ is the rotation matrix from the platform body frame to the mapping frame (derived from the IMU), $\vect{l}_{\text{lever}}$ is the lever arm (the fixed offset vector from the GNSS antenna to the scanner origin, measured in the body frame), $\mat{R}_{\text{bore}}$ encodes the boresight angles between the scanner and the IMU, and $\vect{p}_{\text{scan}}$ is the range and angle measurement from the scanner itself . The quality of the final point cloud depends on how accurately every term in this equation is determined.

Three principal georeferencing strategies exist, each balancing onboard navigation hardware, ground control, and post-processing effort differently: direct georeferencing, indirect georeferencing, and hybrid (integrated) sensor orientation.

Direct Georeferencing

In direct georeferencing, every term in is determined from onboard measurements and system calibration alone, without any ground control points (GCPs). The GNSS receiver provides position, the IMU provides attitude, and the lever arm and boresight angles are established through a pre-survey calibration procedure . This is the standard approach for airborne LiDAR, where the cost and logistical difficulty of distributing GCPs over large areas would be prohibitive.

Four components must be in place for direct georeferencing to deliver on its accuracy promise. A multi-frequency GNSS receiver capable of carrier-phase observations provides the position trajectory of the aircraft, operating either against a single base station or against a network correction service. A high-grade inertial measurement unit (IMU) running at supplies the attitude trajectory, with the angular-rate and acceleration measurements that allow the platform's roll, pitch, and yaw to be reconstructed even during brief GNSS outages. These two trajectories must be related to the scanner through two physical parameters: a lever-arm vector $\vec{l}_{\text{LA}}$ that locates the scanner relative to the GNSS antenna in the body frame (typically surveyed to millimetre precision), and a set of boresight angles $\mat{R}_{\text{sensor}}^{\text{body}}$ that align the scanner frame with the IMU frame, determined by a dedicated calibration flight over a well-surveyed test site.

Under favourable GNSS conditions the accuracy achievable with direct georeferencing is typically (RMSE) in the vertical component and horizontally, depending on the IMU grade, the baseline length to the GNSS base station, and the quality of boresight calibration .

The principal advantages of direct georeferencing are operational speed (no ground control survey is needed), scalability to large areas, and the ability to work over terrain that is inaccessible for GCP placement, such as dense forests, wetlands, or disaster zones.

Real-time versus post-processed solutions..

GNSS positioning for direct georeferencing can be performed in two modes. Real-Time Kinematic (RTK) processing applies differential corrections during the flight and provides centimetre-level positions in real time; this is useful for quality monitoring but requires a reliable data link to a base station or network. Post-Processed Kinematic (PPK) processing applies corrections after the flight using the recorded base station data, which eliminates the data-link requirement and often yields slightly better accuracy because forward-backward Kalman filtering can be applied to the entire trajectory. For production-grade ALS, PPK is the more common choice.

Strip adjustment..

Even with high-quality direct georeferencing, systematic errors of a few centimetres may remain between overlapping flight strips due to residual GNSS biases, IMU drift, or imperfect boresight calibration. Strip adjustment is a post-processing step that estimates and removes these inter-strip discrepancies by minimising the misalignment between corresponding features (or surfaces) in the overlap zones . The corrections are typically parameterised as small shifts and rotations per strip. Strip adjustment is routinely applied in national mapping programmes to bring the internal consistency of the point cloud below the individual strip accuracy.

compares a GNSS-only solution with a tightly coupled GNSS+IMU solution on a real mobile-mapping run as the vehicle crosses an overpass: the GNSS-only solution loses lock as the antenna enters the bridge shadow and remains unusable for several seconds afterwards, while the integrated solution bridges the outage by integrating IMU accelerations and angular rates and recovers to centimetre-level positioning the moment GNSS reacquires lock. The cloud points captured during the outage are usable in the integrated solution and discarded in the GNSS-only one, which is why no production mobile mapping system relies on GNSS alone.

Real data pending

Real data

Indirect Georeferencing

Indirect georeferencing determines the sensor position and orientation at each exposure (or scan) time from observations of ground control points (GCPs) whose coordinates are known in the target reference frame. This is the classical approach in aerial photogrammetry, where a bundle adjustment simultaneously estimates camera exterior orientations and object-space coordinates by minimising the reprojection error at GCP locations and tie points.

Two subcategories can be distinguished:

• Full indirect orientation: no GNSS or IMU data are used; the sensor orientations are determined entirely from the GCPs and the observations themselves. This approach was standard in analogue photogrammetry and remains applicable in certain close-range or laboratory settings, but it is rare for airborne LiDAR because a scanning sensor does not produce the conjugate image points needed for a photogrammetric bundle adjustment.

• Integrated sensor orientation (ISO): GNSS/INS observations are introduced as weighted constraints in the adjustment together with the GCPs. The navigation data provide a strong initial approximation, while the GCPs absorb residual systematic errors and anchor the solution to the control network ().

When a sufficient number of well-distributed GCPs is available, indirect georeferencing can achieve sub-centimetre accuracy in all three coordinate components, which exceeds what direct georeferencing alone can deliver. The disadvantage is the operational cost and time required to establish, survey, and maintain the ground control network .

Hybrid Approach: Integrated Sensor Orientation

The hybrid approach, often referred to as integrated sensor orientation (ISO), combines the advantages of direct and indirect georeferencing. The GNSS/INS trajectory provides an accurate initial solution for every scan line or image exposure, and a set of strategically placed GCPs is then used in an adjustment to refine the trajectory, correct residual systematic biases, and verify the absolute accuracy of the result .

The raw GNSS/INS trajectory is computed using post-processed kinematic methods, and the resulting positions and attitudes are treated as observations with appropriate stochastic models in a combined adjustment that also includes GCP observations. The adjustment estimates corrections to the trajectory parameters and, optionally, to the boresight calibration and lever arm.

Integrated sensor orientation is the method of choice for high-accuracy applications such as cadastral surveying, engineering design, and calibration flights. It is also used routinely in production environments as a quality assurance step: even if the direct georeferencing solution meets the accuracy specification, a few check points (GCPs withheld from the adjustment) are measured to independently verify the result.

How many GCPs for integrated sensor orientation?

The answer depends on the project size and the accuracy target. For a single ALS block of moderate size (50-200 km$^2$), 4-6 well-distributed GCPs are typically sufficient to constrain the adjustment, supplemented by an equal number of independent check points. In photogrammetric block adjustments a commonly cited guideline is at least one GCP per 4-5 strips at each end of the block and at the block perimeter. Over-reliance on GCPs in the block interior adds cost without a proportional accuracy improvement once the systematic errors have been absorbed.

Comparison of Georeferencing Strategies

summarises the key characteristics of the three approaches.

Table pending

Direct georeferencing is the operational standard for most airborne and mobile LiDAR campaigns due to its efficiency and independence from ground access. Indirect georeferencing remains important in photogrammetric workflows where image-based bundle adjustment is applicable. The hybrid approach provides the highest accuracy and most robust quality assurance, and is recommended when project accuracy requirements are stringent or when the sensor system has not been recently calibrated.

Emerging LiDAR Technologies

The LiDAR industry is evolving rapidly, driven by demand from the automotive sector, the miniaturisation of optics and electronics, and the desire for richer spectral information.

Solid-State LiDAR

Traditional LiDAR scanners contain precision mechanical components (spinning mirrors, oscillating polygons, or rotating heads) that are expensive, bulky, and susceptible to vibration. Solid-state LiDAR replaces these moving parts with electronic beam steering:

• MEMS mirrors: micro-electromechanical mirrors (millimetre-scale) steer the beam at very high frequencies. Already dominant in compact automotive units.

• Optical Phased Arrays (OPA): an array of emitters controls the beam direction through constructive interference, analogous to phased-array radar. No moving parts at all.

• Flash LiDAR: the entire scene is illuminated by a single, wide-beam pulse, and a 2D detector array records the range at every pixel simultaneously. Very fast, but limited in range and resolution.

Solid-state systems are smaller, cheaper, and more mechanically reliable, making them the technology of choice for mass-market automotive LiDAR. Their narrower fields of view and shorter range, however, currently limit their use for topographic survey.

Multispectral LiDAR

Multispectral LiDAR systems emit laser pulses at multiple wavelengths simultaneously, for example at , , and . By analysing the relative reflectance at each wavelength, surface materials can be identified without a separate camera. This is conceptually analogous to multispectral remote sensing, but with the spatial precision and 3D geometry of LiDAR.

The Optech Titan (now Teledyne Optech) was the first operational multispectral airborne LiDAR, emitting at three wavelengths (532, 1064, and ) from a single platform. Each channel has independent beam optics and detector, and the three point clouds are co-registered using the shared GNSS/INS trajectory. Because all three wavelengths illuminate essentially the same footprint from the same trajectory, the resulting "spectral point cloud" provides per-point reflectance ratios that are free of the co-registration artefacts inherent in fusing separate LiDAR and camera datasets.

A key challenge in multispectral LiDAR is radiometric consistency across channels. Each wavelength has different atmospheric transmission, detector sensitivity, and beam divergence, so the raw intensities must be individually calibrated to physical reflectance before computing meaningful spectral indices. The normalised difference indices analogous to NDVI can then be formed directly from the LiDAR data:

$$\text{LiDAR-NDVI} = \frac{\rho_{1064} - \rho_{532}}{\rho_{1064} + \rho_{532}}$$

where $\rho_{\lambda}$ denotes the calibrated reflectance at wavelength $\lambda$.

Applications include:

• Land cover classification (vegetation, water, bare soil, asphalt),

• Vegetation species discrimination (different canopy types reflect differently in the green vs. near-IR),

• Road surface condition assessment (wet vs. dry, ice detection),

• Coastal and shallow-water mapping (combining bathymetric green with topographic IR),

• Wetland mapping (separating water, emergent vegetation, and upland using spectral and geometric features simultaneously).

Integration with Other Sensors

Modern survey systems increasingly combine LiDAR with complementary sensors, and the point cloud is no longer the sole product of an acquisition campaign but one layer in a multi-modal three-dimensional representation of the environment. The most common pairing remains LiDAR with a calibrated colour camera, which yields colourised point clouds and texture-mapped 3D models suitable for visualisation as well as photogrammetric cross-checks. Thermal imagery added alongside LiDAR supports building energy audits and infrastructure inspection where heat signatures reveal insulation defects or hidden mechanical activity that is geometrically invisible. In automotive systems, LiDAR is fused with radar so that the geometric measurements of the laser are complemented by the velocity measurements of the radar return. At the trajectory level, LiDAR is integrated with GNSS, INS, and wheel odometry to provide robust pose estimation in environments where any one of those sensors might fail.

LiDAR and Hyperspectral Fusion

While multispectral LiDAR provides 3-4 spectral channels from the active sensor itself, an alternative approach is to co-mount a hyperspectral imaging spectrometer alongside the LiDAR scanner. The spectrometer records hundreds of narrow spectral bands (typically 400-2500 nm), and the LiDAR provides the 3D geometry. After co-registration, each LiDAR point can be assigned a full reflectance spectrum, which enables detailed material identification (mineral mapping, crop species classification, invasive plant detection) that is beyond the capability of the few discrete wavelengths available from multispectral LiDAR.

Data Quality Control

Before any acquisition dataset enters the processing pipeline its quality must be verified, and quality control is performed at four successive stages from the cockpit to the desk.

The first stage is in-flight monitoring. Real-time checks on GNSS lock, IMU health, pulse rate, and scan coverage tell the operator whether the data being collected will meet specifications: a loss of GNSS lock or an IMU drift warning is treated as a potential re-fly trigger and acted on while the aircraft is still over the survey area, not after the data has been delivered.

The second stage is strip-level QC. Once the GNSS/INS trajectory has been post-processed, each flight strip is examined for systematic offsets by comparing it against its overlapping neighbours. Vertical discrepancies exceeding the project tolerance (commonly for QL1 data) indicate a calibration issue that must be resolved by strip adjustment or, in the worst case, by re-flying the affected block.

The third stage is ground-truth validation. Independent check points, surveyed by GNSS or total station on flat, hard surfaces, are compared against the LiDAR-derived elevations. Standard practice follows the ASPRS Positional Accuracy Standards for Digital Geospatial Data, which define accuracy classes by RMSE thresholds and require explicit reporting at the 95 % confidence level.

The fourth and final stage is noise and artefact detection. Point clouds are inspected for spurious returns from atmospheric scatter, bird strikes, and multipath reflections, for systematic artefacts such as banding from scanner timing errors or gaps from GNSS outages, and for intensity anomalies that may indicate a calibration drift across the acquisition window.

The 1% rule for check points. A useful guideline for production surveys is to allocate approximately 1% of the project budget to independent accuracy validation. This typically involves 20-40 ground check points distributed across the project area, stratified by land cover type (open terrain, vegetation, urban) and across multiple flight strips. Skimping on validation is a false economy: undetected systematic errors can invalidate entire datasets.

National LiDAR Acquisition Programmes

Several countries operate systematic, repeating LiDAR acquisition campaigns that produce wall-to-wall coverage as a public good. The resulting data underpin flood modelling, forestry inventory, infrastructure planning, heritage research, and the national 3D building model programmes discussed in . This section describes how those programmes are funded and specified, how acquisition is executed at country scale, how the data is processed and stored, and how it is disseminated to users. lists the major operational programmes as of 2026.

Table pending

Institutional Models and Funding

National programmes sit in three institutional homes, each shaping budget cycles and specifications differently. National mapping agencies (IGN, Kadaster, swisstopo, Lantm"ateriet, Kartverket, GSI) operate the largest programmes; their multi-year master plans bind procurement (IGN's LiDAR HD first cycle covers 2021-2026 at roughly EUR 60 M total, with a second cycle now funded for 2027-2032). Water and environment authorities drive several programmes: Rijkswaterstaat and the regional waterboards co-fund AHN in the Netherlands; the UK Environment Agency, with DEFRA, drove the England-wide National LiDAR Programme to full coverage in 2022. Federal geological and topographic surveys include the United States Geological Survey's 3D Elevation Programme (3DEP), which completed first nationwide QL2 coverage of the contiguous USA in 2023 and is now in a QL1 upgrade phase, and Natural Resources Canada's CanElevation programme run with provincial partners.

International coordination has accelerated. EuroSDR's working group publishes harmonisation guidance for LiDAR specifications and topo-bathy practices, and EuroGeographics' Open Maps for Europe v3 (2025) ingests national elevation derivatives directly. The European Copernicus DEM 2.0 procurement (2024-2026) draws on national LiDAR rather than re-flying where coverage exists, a policy shift that makes consistent national specifications economically valuable beyond their original users.

Specifications

National tenders are unforgiving documents. The defining parameters are pulse density, vertical and horizontal accuracy, classification scheme, reference frame, strip overlap, and approved sensor models. compares the headline specifications of four recent programmes.

Table pending

The USGS Quality Levels are the most commonly cited ladder: QL0 ($\geq 8$ pts/m$^2$, RMSE$_z$, used for dam safety and coastal subsidence), QL1 ($\geq 8$ pts/m$^2$, ), QL2 ($\geq 2$ pts/m$^2$, , the long-time CONUS baseline), and QL3 (now being phased out).

LAS 1.4 with point data record format (PDRF) 6, 7, or 8 is mandatory in every current major tender; classification follows ASPRS standard classes 0-22 (ground, low/medium/high vegetation, building, water, rail, road surface, wires, towers, bridge decks, overhead structures), optionally extended with programme-specific codes. IGN LiDAR HD uses extended codes 64+ for synthetic classes (synthetic ground, hydro-flattened water surfaces); AHN reserves codes 26-27 for civil structures. Per-point attributes universally include adjusted-standard GPS time, intensity (16-bit, with documented normalisation), return number, scan angle, scan direction, and point source identifier.

Flight Planning at National Scale

Once the specifications are written, the operational question becomes when these blocks can actually be flown, and in what order. Five interlocking decisions answer it: how to tile the country, when in the year to fly, how to plan the GNSS situation, how to negotiate airspace and weather, and how to allocate aircraft to blocks across the season. Each is technically simple in isolation but unforgiving in combination.

The spatial unit of delivery has converged across programmes on a $1 \times 1$ km tile in the national projection (Lambert-93 for IGN, ETRS89/UTM in most of Europe, state plane or UTM in the USA), grouped into $10 \times 10$ km flight blocks. In mountainous regions the blocks follow watersheds rather than administrative boundaries, both for flight-line continuity and to align with how the data will later be used by hydrologists and ecologists.

Blocks are not flown when convenient; they are flown when the ground can be seen. Leaf-off acquisition over deciduous Europe and eastern North America runs roughly from 15 October to 15 April, yielding 90-120 usable days once weather is subtracted. Alpine snow-free acquisition compresses to a 60-90-day window between July and September, forcing swissSURFACE3D and the Norwegian Hydemodell to schedule their mountain blocks tightly within it; bathymetric and riparian work, by contrast, follows summer base-flow conditions. The sun adds its own constraint: solar elevation must exceed $20^\circ$ for adequate signal-to-noise on dark targets, so flights launch in the morning and avoid the afternoon turbulence rising over heated terrain.

The GNSS situation is checked before the fleet leaves the ground. Pre-flight PDOP and HDOP forecasts use multi-constellation availability (GPS, GLONASS, Galileo, and BeiDou have been the default since 2023), with a working minimum of six satellites and PDOP below three maintained throughout each strip. Reference stations come from national continuously-operating reference network (CORS) infrastructure such as Teria and Orpheon in France, NETPOS in the Netherlands, CORS in the United States, and SAPOS in Germany; PPP-AR post-processing has become the standard tool for trajectory recovery where local base stations are sparse.

Airspace and weather minima are negotiated next. Air traffic control coordination through filed flight plans with block clearance is required everywhere, and restricted zones around military and nuclear sites demand pre-arranged windows that remain a recurring schedule disruptor. Standard weather minima call for a cloud base above the flying altitude (1500-2500 m AGL for linear ALS, 3500-4500 m for SPL systems), no precipitation in the survey volume, and wind aloft below 35 knots. Cumulative cloud and precipitation losses leave Northwestern Europe with roughly 40-60 % of its seasonal window as effective flying days, a number that determines fleet sizing and contract duration far more than the headline density specification ever does.

The final decision is which aircraft flies which block on which day. National primes such as Hexagon, NV5/Quantum Spatial, Fugro, Geomaud, Sintegra, Aerodata, and (for France) IGN's own fleet operate 4-12 aircraft and shuttle them between blocks on a rolling five- to seven-day weather forecast horizon. At peak season IGN LiDAR HD typically has around six aircraft in the air simultaneously from its contracted primes, with overflow capacity reserved for weather re-planning.

Acquisition Execution

Production rates depend on sensor and flying altitude. A Riegl VQ-1560 II at AGL covers 100-180 km$^2$/h for QL2-equivalent density, or 600-1000 km$^2$ in a full mission day. A Leica CountryMapper at AGL reaches 250-400 km$^2$/h, enabling 1500-2500 km$^2$ daily but at lower per-pass density. Single-photon Leica SPL100, flown at 3500-4500 m, achieves 1000-1500 km$^2$/h, the only sensor that makes continental-scale acquisition tractable inside a single season.

Hybrid LiDAR + camera payloads have become the default rather than the exception. Vexcel UltraCam Eagle paired with a Riegl head, or Leica CountryMapper's integrated camera, capture 4-band orthoimagery (R, G, B, NIR) at 10-20 cm GSD concurrently with LiDAR; IGN LiDAR HD always pairs with a BD ORTHO HR (15 cm) acquisition. Concurrent imagery enables RGB/NIR colourisation of the point cloud (LAS 1.4 PDRF 7/8) and ortho-classification cross-checks.

Pilots and operators run flight-management software (Leica FlightPro, Riegl RiACQUIRE, Optech FMS) with live coverage map, point-density heat-map, and INS/GNSS quality indicators. Re-fly criteria are specified per tile rather than per block: cloud shadow, swath gaps, GNSS outages exceeding 60 s, IMU drift exceeding boresight tolerance, or systematic density loss above 5 % of cells trigger a re-acquisition obligation.

Strip Adjustment, Calibration, and Classification

Raw trajectories from POSPac MMS or equivalent are merged with calibrated point clouds in vendor pipelines: Riegl RiPROCESS (used heavily by IGN subcontractors), Leica HxMap (USGS 3DEP contractors), and TerraScan/ TerraMatch macros for boresight and lever-arm refinement. Cross-strip residuals are computed on flat asphalt patches and gable roofs; corrections are applied iteratively until median $\Delta z$ between overlapping strips falls below (3DEP QL1) or (IGN LiDAR HD). HxMap's GPU-accelerated bundle adjustment (2025) processes a typical 200 km$^2$ block in under 90 minutes on a single high-end GPU node, replacing the multi-hour CPU pipeline that defined the previous generation.

The dominant production stack for classification remains TerraScan macros executing the canonical sequence: low/isolated noise removal $\to$ ground (TIN-densification) $\to$ below-surface filtering $\to$ low/medium/high vegetation $\to$ buildings $\to$ residual noise. Open alternatives have appeared in production through 2025-2026: PDAL with the LAStools-derived lasground/lasclassify, IGN's PyTorch-based Myria3D semantic segmentation (used in the LiDAR HD optimisé 2025 release, which reprocessed every tile of the 2021-2022 acquisition with the new classifier), and swisstopo's 3D U-Net variant trained on roughly 1.2 billion labelled points. Even with machine learning in the loop, dense urban cores, inland water bodies (rivers requiring breakline-driven classification), and conifer interiors still require human review; manual editing in IGN's production typically touches 8-15 % of points in urban tiles.

A trained classification technician handles 40-80 km$^2$/day in urban areas and 200-300 km$^2$/day in rural, mostly automated review. A modern automated cluster classifies a $10\,000$ km$^2$ block in roughly 36-50 hours wall-clock.

Quality Assurance at Scale

Every national programme publishes a quality assurance and quality control (QA/QC) regime as part of the contract. Three checks dominate.

First, density verification. Per-tile point density is rasterised at and compared against the contractual minimum; outliers are reviewed before tile acceptance.

Second, vertical accuracy against an independent network of ground control points (GCPs). GCPs are surveyed by static GNSS to $\leq 2$ cm 3D, distributed across terrain types and avoiding slopes above $10^\circ$. Reporting follows ASPRS Positional Accuracy Standards Edition 2 (2023), which superseded the 2014 edition; reports include NVA at the 95 % confidence level (RMSE$_z\times 1.96$ for non-vegetated surfaces) and the 95th-percentile linear error for vegetated surfaces.

Third, classification accuracy on a sampled 0.5-2 % of tiles, manually reviewed by senior operators. Confusion matrices are reported per tile-batch with target producer's accuracy of $\geq 95\,\%$ on ground and $\geq 90\,\%$ on buildings. USGS contracts an independent QA vendor (Dewberry, Woolpert, NV5/Quantum Spatial in recent rounds); IGN performs QA in-house at the Saint-Mandé production centre.

Cross-strip and block-boundary continuity round out the regime. Adjacent contractor blocks are checked along the seam; tolerances are (NL/CH/FR) and (USA). Classification continuity across block boundaries is enforced by reprocessing a 100 m buffer on either side of each seam. Metadata follows ISO 19115-2 and ISO 19157; INSPIRE compliance is mandatory for EU programmes.

Storage and Archival

National archives are firmly in the petabyte range. The complete AHN 1997-2025 archive is approximately 3.2 PB; IGN's LiDAR HD raw plus classified data is projected to exceed 6 PB by completion of the first cycle in late 2026; 3DEP holds approximately 12 PB as of early 2026.

Architectures are similar in kind across operators: a hot tier on object storage (Ceph at IGN, AWS S3 for 3DEP, Azure Blob at swisstopo, multi-region buckets being deployed at Lantm"ateriet) for classified LAZ, derivative rasters, and recent raw flight lines; a warm tier on spinning disk for raw flight-line LAS retained for one to three years; and a cold tier on tape (LTO-9 in most European archives, AWS Glacier Deep Archive for 3DEP) for raw IMU/GNSS observations and full-waveform data, with retention obligations of 25 years or longer. Catalogue infrastructure has converged on the SpatioTemporal Asset Catalog (STAC) 1.0.0 standard , with national instances at stac.usgs.gov, data.geopf.fr/stac, and the PDOK STAC launched in early 2025. CSW, the older OGC Catalogue Service for the Web, persists as an INSPIRE-compliance façade over STAC.

Dissemination Architectures

Every programme operates a public portal that serves both interactive discovery and bulk download. PDOK at pdok.nl hosts AHN; IGN Géoservices (geoservices.ign.fr/lidarhd) serves LiDAR HD; USGS LidarExplorer at apps.nationalmap.gov/lidar-explorer/ covers 3DEP; the Environment Agency's Survey Open Data portal serves the English National LiDAR products; swisstopo's map.geo.admin.ch provides the Swiss equivalent; PLATEAU at plateau.geospatial.jp hosts the Japanese 3D city model derivatives; and Lantm"ateriet's H"ojddata portal, fully open under CC0 since 2024, serves Sweden. Behind these portals, four parallel delivery channels coexist (bulk file download, raster web services, 3D streaming, and programmatic catalogue access), each suited to a different class of user.

For users who download data, tiled LAZ 1.4 is the universal format, and the Cloud-Optimised Point Cloud (COPC, ) variant, a LAZ 1.4 file with an internal octree organisation that supports HTTP range-request streaming, has become the default in IGN LiDAR HD optimisé (2025), in 3DEP (since 2024), and in swissSURFACE3D (since 2025). The older Entwine Point Tile (EPT) format remains on USGS S3 for legacy compatibility but is being deprecated. Derivative rasters ship as Cloud-Optimised GeoTIFF. Bulk transfer falls back on plain HTTPS with parallel range requests through aria2 or rclone; in the United States, federal users have additional Globus endpoints, and AWS S3 Transfer Acceleration is used on the 3DEP Requester-Pays buckets.

For users who consume rasters interactively rather than downloading, Web Map Service (WMS) and Web Map Tile Service (WMTS) endpoints provide hillshade, slope, aspect, and intensity for visualisation; the newer OGC API Features and OGC API Tiles, rolled out at IGN, swisstopo, and PDOK across 2024-2025, replace the older WFS for vector products. Programmatic catalogue access (find by date, by tile, by sensor) is provided by the STAC API and is increasingly the entry point used by data science workflows, accessed through the pystac-client or stackstac Python libraries.

For 3D users, 3D Tiles 1.1 with implicit tiling and metadata extensions is the streaming format of choice. PLATEAU streams nationwide LoD2 buildings and terrain; since 2025 swisstopo's digital-twin endpoint at twin.geo.admin.ch streams the entire country at LoD2 with classified point cloud overlays. Cesium ion hosts public mirrors of 3DEP and PLATEAU. Web visualisation defaults to Potree for raw point clouds (the AHN viewer is the canonical example), to CesiumJS for 3D Tiles, and to MapLibre or Mapbox GL for raster derivatives.

Standard Derived Products

National pipelines produce a small, stable set of derivative products automatically as part of the publication chain. These products fall into three families: elevation rasters, vector cartographic layers, and modelled hazard or environmental indicators. Together they cover the overwhelming majority of downstream uses, which is why programmes invest heavily in standardising their generation rather than leaving them to individual users.

The elevation raster family is anchored on three closely related products. The digital terrain model (DTM), produced from last-return ground points only, is delivered at resolution by IGN LiDAR HD, swissALTI3D, and AHN4/5; at by 3DEP and the Environment Agency; and at for 3DEP QL2 coverage. The digital surface model (DSM), produced from first-return canopy and structure points on the same grid, captures everything visible from above. The difference of the two (the normalised DSM or canopy height model) isolates above-ground objects and is the standard input to forestry and 3D-city pipelines; IGN publishes a national CHM that the EU Forest Carbon Monitoring programme uses directly. Cloud-Optimised GeoTIFF has been the default encoding for all three rasters since 2023 across every major programme. Two ancillary rasters complete the family: an 8-bit normalised intensity raster used for orthorectification quality control and shoreline mapping, and the hillshade, slope, and aspect layers, which are generally served only through WMTS rather than offered for bulk download.

The vector family is more heterogeneous because it depends on the underlying cadastre. Building footprints arrive either from a separate cadastral layer (BAG in the Netherlands, BD TOPO in France) or from an AI-derived global layer such as the Microsoft and Overture Maps releases that became continuous in 2024; in either case, building heights are derived by intersecting the footprint polygon with the national DSM. These footprints feed directly into 3D building models at LoD1 (block extrusions with a single eaves height) or LoD2 (parametric roof shapes), of which the Dutch 3DBAG (refreshed quarterly), Germany's federal LoD2-DE, and Japan's PLATEAU are the leading examples. Land cover and forest products sit alongside the vector family: Sweden's Nationella Marktäckedata combines LiDAR-derived height with a Sentinel-2 classification, and Finland's Multi-Source National Forest Inventory uses LiDAR-derived forest variables as core predictors.

The third family is modelled hazard and environmental indicators. Flood inundation maps, generated by HEC-RAS 2D and TUFLOW workflows fed from the national DTM, are the most prominent: they underpin FEMA Risk MAP in the United States, the Environment Agency's National Flood Risk Assessment 2 (2024) in England, and France's TRI flood mapping. Other indicators in this family (coastal erosion trajectories, landslide susceptibility, and forest fuel-load maps) are increasingly published by the same operators or by national mapping agencies that build on top of the elevation foundation.

Update Cycles, Versioning, and Change Detection

Programmes are versioned as discrete generations rather than by strict semantic versioning. AHN's cadence has been AHN1 (1997-2003), AHN2 (2007-2012), AHN3 (2014-2019), AHN4 (2020-2022, fully released 2023), and AHN5 in active acquisition with phased release through 2026. IGN LiDAR HD's first cycle runs 2021-2026; a second cycle is funded for 2027-2032 with explicit change-detection deliverables. USGS 3DEP runs continuously, with an eight-year reflight cadence adopted as policy in 2025.

The reprocessing of a published cycle with improved classifiers (the LiDAR HD optimisé 2025 release used Myria3D in place of the earlier rule-based stack) is now treated as a versioned reissue: the point cloud is unchanged, the classification is new, and the reissue is labelled v2 with full provenance.

Change detection between cycles is itself becoming a published product. AHN released $\Delta z$ rasters (AHN4-AHN3) at resolution in 2024. Norway's Endringsanalyse followed in 2025, focused on coastal erosion. USGS is piloting 3DEP change detection between QL2 epochs through the Coastal National Elevation Database. IGN has scheduled a national differential MNT for the second LiDAR HD cycle.

National data sets are large (a national point cloud at $10\,\text{pts/m}^2$ runs to tens of terabytes), but the formats and metadata are uniform, which makes them ideal for tile-aware and out-of-core processing using cloud-optimised formats (). Before commissioning a new ALS flight in a covered region, check whether existing public data already meets the specification. The marginal cost of using national data is essentially zero; the marginal cost of acquiring new data at equivalent density is typically several thousand euros per square kilometre.

The next chapter covers file formats and spatial indexing structures for storing and querying the point clouds these programmes produce.

Notes and Comments

What I want the reader to carry out of this chapter is a way of reading a platform specification rather than a list of platforms. Every sensor in this chapter is a compromise between coverage, density, accuracy, and cost, and there is no platform that wins on all four axes simultaneously. A national ALS programme is what happens when coverage is prioritised and the noise penalty paid for it, with IGN's LiDAR HD reaching 10 pts/m$^2$ over the entire territory of France at a non-vegetated vertical RMSE of and a vegetated 95 % error of . A heritage TLS campaign is what happens when density and accuracy on a single object are prioritised, with a Leica RTC360 or FARO Focus station logging close to a million points per second at sub-centimetre range noise inside the rated envelope. Read in isolation each is a good dataset, but a young surveyor who tries to use a national ALS acquisition for stone-by-stone documentation, or a heritage TLS campaign as an inventory of a 50 km$^2$ valley, has misread the specification.

I will state a view that the textbook consensus understates. The handheld SLAM revolution is real, the speed advantage in interiors and GNSS-denied environments is genuine, and yet the absolute-accuracy claims that come out of vendor brochures are systematically optimistic. Independent benchmarks of indoor handheld systems by found that the better backpack and handheld scanners reach relative accuracy in well-featured rooms but local drift can exceed on long corridors and stairwell traverses, an order of magnitude beyond what the same instruments are quoted to deliver. Apple's iPad and iPhone Pro LiDAR sensors, characterised against survey-grade benchmarks by , report a noise floor near at but their range falls off rapidly beyond and the resulting clouds sit two orders of magnitude noisier than a tripod-mounted RTC360. My own honest reading is that for any project that requires absolute georeferencing the SLAM cloud needs registration to a TLS or GNSS reference, and a project that does not specify how this anchoring will be done is a project that has not yet decided what its accuracy claim actually means.

The frontier I would point a younger researcher towards is not a faster ALS scanner or a denser TLS. It is the closing gap between LiDAR and photogrammetry in operational practice. showed that careful direct georeferencing of UAV photogrammetry with onboard RTK can deliver decimetre-class precision maps over hundreds of hectares for a fraction of a LiDAR campaign's hardware cost, with the noise three to five times larger than co-located LiDAR but correlated with surface texture rather than range and incidence angle. For many heritage projects, façade documentation, and small-area UAV mapping, the choice between TLS and SfM/MVS is no longer a question of quality at all. It is a question of capture economics, of what the project can tolerate in noise, and of what the deliverables will be cross-checked against. The integrated LiDAR-plus-camera payload now standard on national ALS contracts, and increasingly on UAV platforms, is a tacit recognition that no single sensor is the right answer any longer. The reader should expect the platform decision tree of 2030 to be richer, less deterministic, and more openly economic than the one this chapter sketches.

For comprehensive treatments of airborne laser scanning, including system design, flight planning, and data processing workflows, the reader is referred to and , both of which remain standard references in the field. in particular provides detailed discussions of the navigation subsystem (GNSS/INS integration) and strip adjustment procedures that are essential for production-quality ALS data.

Terrestrial laser scanning is covered extensively by (Part II) and in the technical documentation published by scanner manufacturers such as Leica Geosystems, FARO, and Riegl. The challenge of multi-station registration is addressed in of this book; for a deeper treatment of target-based and feature-based registration strategies, see the references therein.

Mobile mapping systems are reviewed by , who surveys the evolution from early vehicle-based systems to modern multi-sensor platforms. The SLAM paradigm for indoor and GNSS-denied mapping was pioneered in the robotics community; the seminal handheld LiDAR-SLAM system Zebedee is described by , and a comprehensive review of the broader SLAM field, covering visual, LiDAR, and multi-sensor approaches, is provided by .

Bathymetric LiDAR technology is discussed by and , who describe the SHOALS (Scanning Hydrographic Operational Airborne Lidar Survey) system and its applications in coastal mapping. The physics of green laser propagation in water, including attenuation, scattering, and the relationship between Secchi depth and maximum survey depth, is treated in these references and in the broader ocean optics literature.

For solid-state and automotive LiDAR, provides an up-to-date review of the principles, challenges, and market trends, including MEMS, OPA, and Flash LiDAR technologies. The automotive sector is driving much of the current innovation in LiDAR hardware, and the cost reductions it achieves will inevitably benefit geospatial applications as well.