We are very excited to announce the publication of a new paper by Amanda Hodgson and co-authors, David Peel and Natalie Kelly.
UAVs are an exciting new technology that have the potential to transform the way we do large scale marine fauna surveys. However, before we can convert to these new, safer and more environmentally friendly methods of conducting surveys, we need an understanding of how well you can detect marine fauna in UAV images. Our paper provides the results of our trial aerial surveys of humpback whales using UAVs. We consider ways you can account for the probability of detecting marine fauna in the images, and suggest data analysis methods that are adapted to UAV surveys.
This article is Open Access
Hodgson, A., D. Peel, and N. Kelly. 2017. Unmanned aerial vehicles for surveying marine fauna: assessing detection probability. Ecological Applications. DOI: 10.1002/eap.1519
Aerial surveys are conducted for various fauna to assess abundance, distribution, and habitat use over large spatial scales. They are traditionally conducted using light-aircraft with observers recording sightings in real time. Unmanned Aerial Vehicles (UAVs) offer an alternative with many potential advantages, including eliminating human-risk. To be effective, this emerging platform needs to provide detection rates of animals comparable to traditional methods. UAVs can also acquire new types of information, and this new data requires a re-evaluation of traditional analyses used in aerial surveys; including estimating the probability of detecting animals. We conducted 17 replicate UAV surveys of humpback whales (Megaptera novaeangliae) while simultaneously obtaining a ‘census’ of the population from land-based observations, to assess UAV detection probability. The ScanEagle UAV, carrying a digital SLR camera, continuously captured images (with 75% overlap) along transects covering the visual range of land-based observers. We also used ScanEagle to conduct focal follows of whale pods (n = 12, mean duration = 40 min), to assess a new method of estimating availability. A comparison of the whale detections from the UAV to the land-based census provided an estimated UAV detection probability of 0.33 (CV = 0.25) (incorporating both availability and perception biases), which was not affected by environmental covariates (Beaufort sea state, glare and cloud cover). According to our focal follows, the mean availability was 0.63 (CV = 0.37), with pods including mother/calf pairs having a higher availability (0.86, CV = 0.20) than those without (0.59, CV = 0.38). The follows also revealed (and provided a potential correction for) a downward bias in group size estimates from the UAV surveys, which resulted from asynchronous diving within whale pods, and a relatively short observation window of 9 s. We have shown that UAVs are an effective alternative to traditional methods, providing a detection probability that is within the range of previous studies for our target species. We also describe a method of assessing availability bias that represents: spatial and temporal characteristics of a survey, from the same perspective as the survey platform; is benign; and provides additional data on animal behavior.Figure_1_Map_survey_area_with_images_edited
Figure 1: Location of study site (inset a), location of ground control station, land-based station and UAV survey area (inset b), example of images captured during one survey and sighting locations mapped using VADAR (www.cyclops-tracker.com) (main map), and zoomed in showing images with whale sightings including duplicate sightings in overlap of images (inset c).
Hodgson, A. J., N. Kelly, and D. Peel. 2013. Unmanned Aerial Vehicles (UAVs) for surveying marine fauna: a dugong case study. PLoS ONE 8:e79556. DOI: https://doi.org/10.1371/journal.pone.0079556
Maire, F., L. Mejias, and A. Hodgson. 2015. Automating Marine Mammal Detection in Aerial Images Captured During Wildlife Surveys: A Deep Learning Approach. Pages 379-385 in B. Pfahringer and J. Renz, editors. AI 2015: Advances in Artificial Intelligence. Springer International Publishing. Download here