How Can Drones and Lasers Reveal Lake Bottoms?

Lake bottoms and shoreline lines are key to understanding marine environments. Until recently, measuring these surfaces required echo‑sounders on boats, a slow and costly method. Today, drones (UAV) equipped with photogrammetric cameras and airborne LiDAR (ALS) systems can produce high‑resolution bathymetric maps quickly and efficiently.

In Lake Kłodno, researchers combined two techniques: drone photogrammetry to model the shallow floor and LiDAR to trace the shoreline. The data were processed with support vector regression (SVR), allowing depth estimation down to 1 m with a standard deviation of 0.11 m and an accuracy of 0.22 m at 95 % confidence.

However, only 44.5 % of the 1 × 1 m cells met the recommended minimum density (≥5 points/m²), while 43.7 % were empty. LiDAR, on the other hand, provided more uniform coastal coverage, with an average density of 63.26 points/m², though 27.6 % of cells remained empty. The shoreline extraction method applied to LiDAR achieved a mean positional accuracy of 1.24 m (95 % confidence).

The takeaway? Drones and LiDAR have distinct but complementary strengths: together they generate more precise and reliable maps of shallow waters and coastlines, supporting environmental management and infrastructure planning.

The combined UAV photogrammetry and ALS LiDAR approach exploits the distinct physical properties of light. Photogrammetry uses optical images to reconstruct surfaces through point matching, ideal in clear, shallow waters where light penetration is sufficient. LiDAR, on the other hand, emits green laser pulses (around 532 nm) that can penetrate water up to 30 m, providing direct distance measurements between the aircraft and the bottom.

In the case of Kłodno, photogrammetry was calibrated with ground control points (GCPs) to reduce systematic orientation errors. The rasterized data were then fed into a support vector regression (SVR) model, a machine‑learning technique that builds an optimal non‑linear function between spatial coordinates and measured depths. The SVR was trained on a set of depth samples obtained from traditional surveys, achieving a standard deviation of 0.11 m.

Regarding point density, NOAA guidelines require at least 5 points per square meter for adequate coverage. Photogrammetry reached only 44.5 % of this threshold due to shadows, reflections, and low water transparency. LiDAR far exceeded the requirement, with 63.26 points/m², yet 27.6 % of cells remained empty because of highly reflective surfaces or atmospheric interference.

Shoreline extraction from the LiDAR point cloud was performed using a gradient‑based elevation and point‑density filter, followed by an offset correction to compensate for light refraction in water. The resulting positional accuracy of 1.24 m (95 % confidence) is competitive with traditional methods but can be further improved by integrating multi‑temporal data and tide corrections.

In summary, the synergy between UAV and LiDAR overcomes the limitations of each technology: photogrammetry provides high‑resolution topographic detail in very shallow water, while LiDAR ensures broader, more consistent shoreline coverage and higher positional accuracy. Their joint use is especially promising for mapping dynamic coastal environments, managing water resources, and assessing climate‑induced impacts on vulnerable zones.