UAV Photogrammetry and LiDAR for Coastal Mapping

Unmanned aerial vehicles have become a standard platform for topographic and coastal mapping. Two sensor families dominate this work: camera-based photogrammetry and laser-based LiDAR. Each produces a three-dimensional representation of the terrain, but they do so through different physical processes, and each carries distinct strengths and limitations. In coastal environments, where the transition between land and water is gradual, dynamic, and often poorly textured, integrating both methods produces a more complete and more defensible terrain model than either method alone.

Photogrammetry: Structure from Motion and Dense Matching

Photogrammetric mapping with UAVs relies on structure-from-motion (SfM) processing. A camera captures a sequence of overlapping images along planned flight lines. SfM algorithms detect features that appear in multiple images, match them across overlapping frames, and estimate both camera positions and three-dimensional coordinates of those features through bundle adjustment. A dense matching step then generates a point cloud from the image set, and the point cloud is interpolated into a surface.

Key parameters include ground sample distance (GSD), which describes the ground footprint of one image pixel, and overlap, both along and across flight lines. GSD is governed by sensor resolution, focal length, and flying height. Smaller GSD improves detail but increases the number of images and processing load. Overlap of roughly 75 to 80 percent along track and 60 to 70 percent across track is commonly used for terrain mapping, because dense matching requires each ground point to appear in several images.

Structure from motion recovers three-dimensional geometry from two-dimensional image observations using feature correspondences and bundle adjustment. Its accuracy depends on image texture, overlap, camera calibration, and the distribution of ground control or precisely positioned camera stations.

Photogrammetry performs well over bare soil, rock, sand, asphalt, and low vegetation. It performs poorly over water, uniform surfaces, and areas with repetitive texture. It also produces a digital surface model rather than a bare-earth terrain model, because vegetation and structures are captured as part of the observed surface.

LiDAR: Direct Range Measurement

UAV LiDAR measures distance directly. A laser emits short pulses, and the receiver records the time of flight for each return. Because the pulse has a finite width and diverges slightly with distance, it can be partially reflected by leaves, branches, and the ground in sequence. A single pulse may therefore produce multiple returns.

• First return: Typically from the top of vegetation or the highest surface encountered.
• Intermediate returns: From branches, understory, or partially occluded surfaces.
• Last return: Often from the ground, provided the pulse is not fully intercepted by vegetation.
• Intensity: Records relative reflectance at the laser wavelength, useful for surface classification.

Most topographic UAV LiDAR operates in the near-infrared around 905 nanometres. Near-infrared is absorbed strongly by water, which makes standard topographic LiDAR unsuitable for bathymetric measurement but useful for detecting the water surface boundary. Bathymetric LiDAR uses a green wavelength near 532 nanometres, which penetrates water, often in combination with an infrared channel to detect the water surface. Depth penetration in bathymetric LiDAR is limited by water clarity and is generally a function of the diffuse attenuation coefficient in the green channel.

Why Integrate the Two Methods

Photogrammetry and LiDAR are complementary rather than competing technologies. The practical reasons for combining them are consistent across coastal and topographic projects.

• Vegetation penetration: LiDAR last returns and multiple returns support ground classification under canopy and scrub, which photogrammetry cannot achieve reliably.
• Texture independence: LiDAR does not require image texture, so it works over uniform sand, dry gravel, and shadowed terrain where dense matching is weak.
• Colour and interpretation: Photogrammetry provides true-colour orthomosaics and high-resolution imagery that support feature identification, shoreline interpretation, and surface classification.
• Redundancy: Independent surfaces derived from two physical principles allow cross-checking and detection of systematic errors in either dataset.
• Coverage efficiency: Imagery can be acquired over large areas at lower cost, while LiDAR can be concentrated on critical corridors such as shorelines, channels, and vegetated margins.

Integrating independent sensors improves reliability because agreement between datasets acquired by different physical principles provides stronger evidence of correctness than internal consistency within a single dataset.

Coastal Zone Challenges

The coastal zone presents conditions that are hostile to optical mapping. Water is a moving, specular, low-texture surface. Sun glint saturates image pixels. Wave action changes the surface between exposures, so matching fails or produces spurious points. The wet sand zone has reduced texture and variable moisture, which alters reflectance.

Shoreline extraction therefore requires care. The apparent waterline in an orthomosaic is not a fixed boundary; it is the instantaneous intersection of the water surface with the beach at the moment of capture. To produce a meaningful shoreline, the survey must record the tide level at the time of flight and reference the resulting boundary to a defined tidal datum.

Tidal coordination is a central planning requirement. Flights timed to low water expose more of the intertidal zone and reduce the area where optical methods fail. For projects requiring a consistent vertical reference, water level observations from a tide gauge or a nearby reference station should be logged during acquisition and used to reduce the survey to the project datum.

Vegetation in coastal areas is often dense and salt-tolerant, forming low scrub and mangrove communities that obscure the ground. This is precisely where LiDAR multiple returns add value over photogrammetry, provided the pulse density is sufficient to achieve ground hits through gaps in the canopy.

Positioning, Control, and Geodetic Framework

UAV mapping accuracy depends less on the sensor than on positioning and control. Two approaches dominate.

1. Ground control points: Precisely surveyed targets distributed across the site and visible in the imagery. They define the transformation from image space to project coordinates.
2. Direct georeferencing with RTK or PPK: The UAV carries a GNSS receiver, and camera or laser positions are computed by real-time kinematic or post-processed kinematic methods. This reduces dependence on ground control but does not eliminate the need for independent check points.

For LiDAR, direct georeferencing is more demanding, because the laser scanner must be aligned to the GNSS and inertial reference through a boresight calibration. Angular misalignment of even a small amount produces a systematic offset in the point cloud that increases with range. Boresight calibration is normally performed over a known surface with distinct features at varying ranges and scan angles.

In Gulf coastal settings, the vertical reference requires explicit attention. GNSS-derived heights are ellipsoidal and must be converted to orthometric elevations using a geoid model, or reduced directly to a tidal datum such as mean sea level or chart datum for marine-related work. Mixing vertical references between datasets is one of the most common sources of error in coastal mapping.

Point Cloud Fusion and Surface Generation

Integration is performed at the point cloud level. Both datasets are transformed into a common coordinate reference frame, co-registered, and checked for systematic offsets. Where a common physical surface exists, the surfaces should agree within the stated uncertainty. Discrepancies indicate control error, boresight error, or a timing mismatch.

After co-registration, points are classified. Ground, vegetation, water, and structures are separated. Classification supports the generation of two distinct products:

• Digital surface model: The top visible surface, including vegetation and structures, suitable for volume and clearance analysis.
• Digital terrain model: A bare-earth representation after vegetation and structures are removed, used for earthworks, drainage, and terrain analysis.

In coastal applications, an additional boundary product is generated: a shoreline or waterline vector referenced to the survey epoch and tidal datum. Photogrammetric orthomosaics support interpretation of that boundary, while LiDAR supports its accurate horizontal placement.

Verification and Accuracy Reporting

Deliverables should be accompanied by an accuracy statement based on independent check points not used in processing. Check points should be distributed across terrain types, slopes, and vegetation conditions, and their residuals reported separately for horizontal and vertical components. Comparisons between photogrammetric and LiDAR surfaces over shared ground areas provide an additional cross-validation.

Optimization Strategies and Common Pitfalls

• Good practice: Plan flight lines with sufficient overlap and consistent altitude to achieve uniform GSD and point density.
• Good practice: Time flights to low water where possible and log tide level continuously during acquisition.
• Good practice: Acquire both datasets in the same campaign, under similar conditions, to minimise temporal mismatch.
• Good practice: Perform boresight calibration for LiDAR before production and verify on a known surface.
• Good practice: Use independent check points to report accuracy rather than relying on internal fit statistics.
• Good practice: Document the geodetic framework, vertical datum, and any tidal reduction applied.
• Poor practice: Using photogrammetry alone over vegetated coastal margins and presenting the resulting surface as a bare-earth terrain model.
• Poor practice: Treating the instantaneous waterline from an orthomosaic as a fixed shoreline without tidal reference.
• Poor practice: Mixing vertical references between LiDAR and photogrammetric datasets without conversion.
• Poor practice: Skipping boresight calibration or lever arm verification after sensor or mount changes.
• Poor practice: Flighting in high sun angle, which produces glint over water and deep shadow in vegetated terrain.
• Poor practice: Extrapolating derived depths beyond the optical or laser penetration limit and presenting them as measured bathymetry.
• Poor practice: Reporting accuracy from processing software defaults rather than field verification.

Practical Implications for Geomak Projects

Coastal and topographic mapping in the Gulf region involves wide intertidal flats, shallow nearshore water, sabkha surfaces, low scrub, and rapid coastal development. Photogrammetry provides efficient colour imagery and dense coverage across accessible terrain, while LiDAR delivers ground returns beneath vegetation and reliable surfaces over uniform sand. Combined, they support shoreline definition, coastal change monitoring, earthwork verification, and pre-construction terrain modelling.

Geomak applies UAV photogrammetry and LiDAR within a controlled geodetic framework, with tidal coordination for waterline definition, independent check points for accuracy reporting, and documented processing records. The objective is not to acquire the largest possible dataset, but to deliver terrain and shoreline products whose accuracy, reference frames, and limitations are clearly stated and defensible for engineering and regulatory use.