By L. Wicks
•
05 Nov, 2019
Some species are more easy to study compared to others, but the challenges faced by ecologists and conservationists worldwide are the same when it comes to accurate data collection. In order to conserve and protect effectively, we must have the most up to date information on the species we are working with. Dr Claire Burke, an astro-ecologist from Liverpool John Moores University UK, and her team have been combining the disciplines, of astrophysics, computational analysis, ecology and machine learning utilising drones, resulting in some brilliant research on novel techniques to aid in monitoring and identifying species. Its one thing, undertaking research on fairly slow moving easy to spot species, however finding species such as orangutan within a rain forest canopy is another story, and until recently has been the sole unenviable task of conservation biologists and in-country experts tracking from the ground. The use of drones or UAVs (unmanned aerial vehicles) as they are sometimes known, has the benefit of covering large areas relatively quickly. In the thermal infra red, given the right conditions, animals are displayed as bright shapes effectively "glowing like stars". The left hand image below shows the visual image from the drone alongside the same frame shown in the thermal infra red - it is striking how the orangutan are clearly shown within the tree canopy. Thermal imaging's ability to 'see through' vegetation cover to a heat source partially concealed within it proves invaluable in detecting wildlife in tricky environmental conditions. Burke et al. (2019) research looked at the efficacy of using a drone equipped with thermal imaging device and found a high hit rate detecting 41 orangutans and a troop of proboscis monkeys, all of which were confirmed by ground observers. Dr Burke is also developing a machine learning classification system to help identify species from aerial thermal imaging by their unique thermal 'finger print' (right hand image below). This kind of system requires very large training data sets which need to be labelled and therefore massively time consuming. Her team are working to address this issue using a citizen science project called Zooniverse which provides access to vast numbers of human classifiers. They are also working on source detection software which could label species automatically prior to machine learning refinement. This work will be critical in effective conservation monitoring and could be applied to a range of wildlife applications. To find out more about Claire and her research please see the links below and for training opportunities,click on the Useful Resources tab: https://www.ljmu.ac.uk/about-us/staff-profiles/faculty-of-engineering-and-technology/astrophysics-research-institute/claire-burke https://www.researchgate.net/profile/Claire_Burke6 http://www.astro.ljmu.ac.uk/research/astro-ecology https://www.cburkesci.com/