Introduction
While uncrewed aerial vehicles are widely used for vegetation monitoring on land, comparable approaches for underwater vegetation monitoring remain less common but are becoming increasingly relevant.
In the canton of the Grisons, Isoëtes lacustris (lake quillwort), a vulnerable aquatic plant species, requires regular monitoring. Since the plant grows in shallow near-shore regions, monitoring is currently performed manually by biologists at selected locations. As a result, systematic large-scale monitoring of the species is currently not feasible. Autonomous robotic systems therefore offer significant potential for supporting and improving such ecological monitoring tasks.
The presented work aims to develop and evaluate an autonomous monitoring solution with a particular focus on deployment in shallow and remote alpine lakes.
Methodology
The monitoring concept is based on mapping the lakebed for later ecological evaluation. These maps, referred to as orthomosaics, are generated using the open-source photogrammetry software WebODM. To acquire the required data, a dedicated catamaran-style uncrewed surface vehicle (USV) was developed. The USV is equipped with an action camera, RTK-GNSS positioning, a low-cost sonar, onboard computing and autonomous navigation capabilities based on ArduPilot. Since deployment sites may only be accessible by foot, portability and low system weight were key design requirements. To avoid sediment disturbance in shallow near-shore areas and prevent the USV from becoming entangled in submerged vegetation, the system uses air propellers instead of underwater thrusters for propulsion and steering.
During operation, the USV autonomously follows predefined trajectories while continuously recording underwater video, water depth measurements and RTK-GNSS positioning data. In a post-processing step, video frames are extracted, synchronized with positional data and georeferenced before being processed by WebODM to reconstruct large-scale orthomosaics of the lake floor. In a separate processing step, sonar-based water depth measurements combined with RTK-GNSS positioning information are used to generate bathymetric maps of the lake.
The resulting orthomosaics can subsequently be analysed by biologists to assess the occurrence and density of Isoëtes lacustris.
Results
Pilot monitoring campaigns were conducted at Laghetto Moesola and Lag Bonaduz. During these deployments, the developed system successfully acquired georeferenced underwater imagery, which was processed into high-resolution orthomosaics of submerged vegetation regions. High image overlap obtained through cross-grid trajectories resulted in significantly improved orthomosaic quality and reconstruction consistency. However, orthomosaic quality remains dependent on environmental conditions, as direct sunlight can produce moving light patterns on the lake floor that interfere with photogrammetric reconstruction.
In addition to image-based monitoring, the system demonstrated the feasibility of integrating low-cost sonar measurements into the monitoring workflow. Initial bathymetric visualizations of lake sections were successfully generated and may support future depth-aware mission planning.
Discussion and outlook
The presented work demonstrates that the developed uncrewed surface vehicle system combined with underwater photogrammetry can provide an effective tool for ecological monitoring in alpine lakes. Feedback from participating biologists indicates that the generated orthomosaics are suitable for supporting monitoring tasks of Isoëtes lacustris. Compared to fully manual surveys, the system improves efficiency and enables systematic georeferenced documentation of underwater vegetation.
A first large-scale monitoring campaign covering multiple alpine lakes in the canton of the Grisons is currently ongoing.
Future work includes automated detection and density estimation of Isoëtes lacustris using computer vision approaches, improved obstacle avoidance capabilities and generation of more accurate bathymetric maps for adaptive path planning.
References
[1]
O. Putallaz, «Bestandessituation von Isoëtes lacustris (See-Brachsenkraut) im Mässersee (2115 m), Binn VS,» Valeco GmbH, Visp, 2019.