Development of drone-based tools for the monitoring of submerged aquatic vegetation

The importance of submerged aquatic vegetation (SAV) as a coastal marine habitat and provider of a wide range of ecosystem services has for decades motivated researchers and managers to investigate abundance and growth dynamics to understand linkages to human perturbations, secure protection and allow restoration. In European coastal waters, SAV monitoring campaigns traditionally involve diver observations and/or video recordings. While these techniques provide very useful data, they are rather time consuming, labour intensive, and limited in their spatial coverage. Recent innovations in the field of drone technology have opened up for new approaches to SAV monitoring, allowing to shift from point specific measurements to area based monitoring, without compromising the spatial resolution needed.
To investigate the possibility of integrating drone technology into national SAV monitoring programmes, such as the environmental impact assessments (EIAs) within Natura 2000 areas or the 3rd generation water plans, the functionality of unoccupied aerial vehicles (UAVs) was tested in Danish water bodies of different characteristics. The results of the PhD project showed that accuracy and efficiency of traditional in-water SAV monitoring campaigns can be improved in many cases, if supported by UAV-based monitoring methods. After a literature review on remote sensing techniques used to map and monitor SAV in chapter 1 and conducted field studies that addressed different aspects of UAV-based monitoring in the chapters 2-7, a recommendation of a robust set-up for UAV-based SAV monitoring could be given.
Methods of UAV-based quantification of SAV coverage and its change over time, the area-specific estimation of SAV biomass and the identification of stressors affecting SAV communities were presented in chapter 2. A comparison of different sensors in chapter 4 showed that high spatial resolution of conventional red, green and blue (RGB) sensors is superior to more advanced narrow band but lower resolution multispectral sensors when monitoring SAV in turbid or deep waters. In chapter 5, a UAV-based underwater camera system was developed that can be added to the set-up, if ground truthing data needs to be collected. The analysis of the data obtained in this project was performed using an object- based image analysis (OBIA) approach. This approach proved to be useful when analysing UAV-based high-resolution imagery and the results from chapter 3 can be used as a guideline for the image analysis process and a support for the classification algorithm choice. In chapter 6, a convolutional neural network (CNN) was trained in order to test the possibility of providing an alternative SAV classification method of high robustness towards different environmental conditions and site characteristics. The results were promising, albeit showing potential for improvements. In relation to large-scale monitoring programmes, such as the EIAs in Natura 2000 areas, the results of chapter 7 showed that by focusing on deeper waters with the traditional video-based methods and instead use UAVs to monitor SAV in shallow areas, not only the survey time is reduced but also the quality of the produced information in terms of spatial accuracy and resolution can be significantly improved.
The PhD project concluded that UAV-based methods can and should be incorporated into future SAV monitoring programmes by applying a setup that facilitates repeatability, data comparability and easiness of its application while being robust and labour efficient. A light-weight multi‐rotor UAV equipped with a high-resolution RGB camera, RTK module and customized ground truthing device was found to be most suited setup for the task. However, in the future, other platforms, such as fixed-wing or hybrid UAVs, can be recommended for long-term reoccurring survey campaigns, when focusing solely on large-scale SAV monitoring.