Iguanas from Above: Drones, citizen science and machine learning

Iguanas from Above essentially started with a pilot field season in early 2020. At the time, we were a team of just two — myself and now-PhD student Andrea Varela — flying drones from a boat at a handful of sites, hoping the novel method would work.

Fast-forward to early 2024, when we have now surveyed the entire range of the marine iguana, across all 13 main islands — the first time such an extensive survey of this emblematic species has been undertaken, providing data that will be vital in assessing the impact of the current El Niño conditions on marine iguana populations.

Now the focus is on counting iguanas from our enormous dataset of photographs. We currently have preliminary numbers for about half the islands, and our aim is to calculate population sizes for all 11 subspecies by early 2025. Our work has been greatly facilitated by the efforts of thousands of online volunteers who help count the iguanas via our project on the Citizen Science platform Zooniverse.org.

We found our way to using drones purely by necessity. When it became clear that traditional means (like walking surveys) would never deliver the data needed for effectively conserving marine iguanas, we decided to embrace new technologies; though this is not without its struggles. There are significant practical issues involved with flying drones from small boats in remote regions. These aircraft are not waterproof, and launching/ landing from small moving platforms — such as rubber dinghies on high seas — requires special protocols and a high degree of piloting skill.

Over the four field seasons of our project, we have expanded and massively improved our methods. Yet, we are still really pushing the limits of these tools, and using them as we do remains challenging. Nevertheless, drones have allowed us access to remote colonies where sea conditions and terrain make foot surveys impossible. Our ‘eye in the sky’ also protects us from dangerous boat landings and protects the Islands by minimising disturbance. Drone-based iguana field surveys are far faster than those done on foot, and they deliver more accurate counts. In addition, the images we collect are already being used by other researchers, and will be archived for later use — who knows what future questions they might help answer?

Our next frontier is to use machine learning (ML), and we are already collaborating with ML experts at the University of Bielefeld to use their platform ‘BIIGLE’ (Bio-Image Indexing and Graphical Labelling Environment) for this purpose. Our preliminary work indicates that ML algorithms will require significant training to reliably detect marine iguanas in their visually complex habitats. We plan to facilitate this learning phase by engaging some of our loyal Zooniverse volunteers to train and assess the ML efforts, and we hope to develop a system that may be capable of automated iguana counting in the years to come.

Our work shows that drones hold great promise for conservation, and seem particularly suitable for surveying coastal species. With sufficient training, drones can be safely and effectively used for boat-based surveys, and we hope they will continue to be used for this purpose long after our project is complete. Given the unprecedented challenge faced by conservationists today, we really must engage any and all tools to help us stem the biodiversity loss; drones are likely to prove a crucial weapon in this fight.