The application of UAVs for monitoring environmental change on small carbonate islands

Abstract

Rising sea-levels and increased storm intensity are of particular concern to Large Ocean States like the Bahamas. The sustainability of development ambitions is also essential in order to not compromise ecosystem services vital to local communities. Unoccupied Aerial Vehicles (UAV) are rapidly deployable and are capable of producing high resolution datasets essential to monitoring changes to water-resources and habitat carbon budgets on low-lying carbonate islands. In the case of UAV-borne topographic surveys, the high quality represented by LiDAR is unmatched, but costs ranging between tens of thousands and hundreds of thousands of pounds limit widespread use. Photogrammetry surveys are significantly cheaper but have limited accuracy in heavily vegetated regions by canopy occlusion.
This study aimed to develop a rudimentary, low-cost methodology for converting Digital Surface Models (DSMs) to Digital Terrain Models (DTMs) in the varied vegetation of north Andros Island, Bahamas. The applications of this methodology were also explored. DSMs demonstrated the presence of vegetation occlusion in freely available and commercial satellite Digital Elevation Models. DSMs were classified using RBG-band ratios derived from inbuilt UAV sensors, achieving 77% to 89% accuracy limited to binary vegetation-bare-earth discrimination. Void models were statistically filtered for outliers before interpolating to produce DTMs with sub-meter 0.94 ± 0.04 m accuracy. DSMs and DTMs were used to develop Canopy Height Models, testing accuracy against LiDAR and manual survey techniques.
Though discrepancies in accuracy to LiDAR were present to nearly a meter, this methodology suggests that high quality data can still be achieved at a fraction of the price. Encouraging canopy modelling results may motivate future investigations towards a direct application to carbon budgeting. However, limited classification abilities would likely restrict success to discrete, uniform vegetation stands.

Date of Award	17 Mar 2026
Original language	English
Awarding Institution	University of Bristol
Supervisor	Kieran Wood (Supervisor) & Fiona F Whitaker (Supervisor)