Handout picture released on march 10, 2022 by the Parque Nacional Galapagos press office showing a ‘Chelonoidis chathamensis’ turtle at the San Cristobal island, Galapagos islands, Ecuador on August 2, 2019 Copyright AFP/File Aamir QURESHI
Scientists from University of Florida have used drones and smart modelling to accurately count over 41,000 endangered turtles nesting along the Amazon’s Guaporé River—revealing the world’s largest known turtle nesting site.
Their innovative technique, combining aerial imagery with statistical correction for turtle movement, exposes major flaws in traditional counting methods and opens doors to more precise wildlife monitoring worldwide.
By combining aerial imagery with statistical modeling, University of Florida researchers have documented more than 41,000 Giant South American River Turtles gathered along the Amazon’s Guaporé River. Their findings offer a new tool for conservationists seeking to monitor vulnerable animal populations with greater precision.
The science project began with Wildlife Conservation Society (WCS) researchers in Brazil, Colombia and Bolivia. One of the species the New Yorkbased organisation monitors is the Giant South American River Turtle, which is threatened by poachers who sell its meat and eggs. The turtles are exceptionally social creatures, and females congregate each year in July or August to nest in the Guaporé River sandbanks between Brazil and Bolivia.
Lead scientist Ismael Brack, a postdoctoral researcher within the UF Institute of Food and Agricultural Sciences, met WCS scientists at a conference, shares how the research group used drones to count the turtles. They create maps termed orthomosaics, which are highly detailed, highresolution composite images made by stitching together hundreds of overlapping aerial photographs.
Counting the animals shown in orthomosaics is a quicker, more accurate and lessinvasive approach than counting animals from the ground. The method alone, however, doesn’t account for the fact that animals sometimes move during observation.
The researchers used white paint to mark the shells of 1,187 turtles gathering on an island sandbank within the Guaporé River, according to the study. Over 12 days, a drone flew overhead on a meticulous, backandforth path four times a day and snapped 1,500 photos each time. Using software, scientists stitched the photos together, and researchers reviewed the composite images.
The scientists recorded each turtle, if its shell was marked and whether the animal was nesting or walking when photographed. Equipped with this data, they developed probability models that account for individuals entering and leaving the area, observed turtle behaviours and the likelihood of detecting an identifiable shell mark.
Together, researchers from UF and the WCS developed a method that improves counting accuracy by eliminating multiple sources of error, including double counts (the same individual counted multiple times) and missed individuals.
Error spotting
The models reveales several potential sources of error that could arise from traditional orthomosaicbased counts, according to the study. Only 35% of the turtles that used the sandbank, for example, were present during drone flights. And, on average, 20% of the those detected walking appeared multiple times in orthomosaics — some as many as seven times.
Researchers on the ground counted about 16,000 turtles, according to the study. Researchers who reviewed the orthomosaics (creating a precise, detailed overhead map) but this did not account for animal movement or shell markings counted about 79,000 turtles. When they applied their models, however, they estimated about 41,000 turtles.
According to Brack: “We describe a novel way to more efficiently monitor animal populations…And although the method is used to count turtles, it could also be applied to other species.”
Overall, the study describes ways to adapt and apply the approach to conservation efforts involving other species surveyed by dronederived orthomosaics. Past monitoring studies, for example, have involved clipping seals’ fur, attaching highvisibility collars to elk and marking mountain goats with paintball pellets to keep track of animal movement during counts.
The research appears in the Journal of Applied Ecology, it is titled “Estimating abundance of aggregated populations with drones while accounting for multiple sources of errors: A case study on the mass nesting of Giant South American River Turtles.”
