Monitoring salmonid populations using unmanned aerial vehicles: a case study of Arctic charr in Iceland

• Main Author: Garcia Martinez, Silvia, 1985-
• Other Authors: Háskóli Íslands
• Format: Thesis
• Language: English
• Published: 2019
• Subjects: Umhverfis- og auðlindafræði; Bleikja; Umhverfisvöktun; Drónar; Arctic; Arctic charr; Iceland; Salvelinus alpinus
• Online Access: http://hdl.handle.net/1946/33232

Salmonids are very susceptible to environmental changes and any shift in environmental conditions can lead to population decline or even population collapse. Monitoring the activities at their spawning grounds and more specifically their redds (salmonid nests) over time will help understand the population dynamics of salmonid species and subsequently, create strategies for species conservation. However, basic and costly methodologies like observations or manual counting to estimate the number of active redds are difficult and involve potential errors of interpretation. The goal of this study is to develop a reliable and efficient remote sensing protocol to aid future research and monitoring efforts with identifying changes in salmonid redd density. The study focuses on the spawning grounds of Arctic charr (Salvelinus Alpinus) in Thingvallavatn and Ellidarvatn (Iceland), two lakes with different environmental characteristics which allows for testing and comparing various classification algorithms. Data were collected by an aerial drone image survey during July and November of 2018, using a low-cost unmanned aerial vehicle (UAV) with standard visible-light images (red, green, blue spectral bands) from a digital camera sensor. We applied advanced algorithms of semi-automatic remote sensing image analysis and classification methods and we identified the spawning grounds of Arctic charr and compared the results of the two study areas to establish a procedure for future applications. The results indicate that UAV systems are suitable for environmental monitoring of shallow water in lakes in Iceland. However, several environmental conditions have to be kept in mind while obtaining the images such as sun angle, cloud cover, water clarity, and wind speed. The classifications in this study reached overall accuracies of up to over 90 per cent while detecting the actual Arctic charr redds reached accuracies of 72 to 76 per cent.