Conservation app’s censoring of observation data could hurt threatened species, scientists argue | Science

Rusby Contreras-Díaz was scrolling through a biodiversity database, looking at places where people had spotted margays and jaguars, when she noticed something odd: Every location had an uncertainty of 30 kilometers. Contreras-Díaz—a mammologist at the National Autonomous University of Mexico—had hoped to use the observations to map where the two carnivores were found in Mexico. But the uncertainty in the data gave her pause.

The observations were from iNaturalist, an app and website that allows members of the public to upload plant and animal photos to an online repository and report the locations of species they’ve identified. iNaturalist has amassed a tight-knit global community of users and the data it has generated have contributed to hundreds of scientific studies. “Many species didn’t have data before iNaturalist,” Contreras-Díaz says.

But for some species, the location data aren’t entirely accurate. Whenever a user reports seeing a species that is listed as threatened or endangered, iNaturalist automatically obscures the exact coordinates, creating a random location within a certain distance that varies depending on the latitude. “People have argued that handing out [exact locations] to the public is basically a road map for poaching,” says Flora Ihlow, a macroecologist at the Dresden University of Technology who has used iNaturalist data in her own research.

Contreras-Díaz acknowledges the benefit this data cloaking can have for species at risk of exploitation. But she worries the practice could harm conservation efforts for the very organisms that most need help—by distorting maps depicting where those species might be found.

In a study published last month in Biological Conservation, she tested that possibility using locations gathered by survey scientists for three amphibian species in Argentina. Using the same process as iNaturalist, Contreras-Díaz and colleagues modified where each individual was spotted, generating a random location within 30 kilometers of the original. They then modeled each species’ range using the modified data set and the one with true coordinates. After comparing the two maps, the team found that the overlap ranged from 45% to 56%.

Contreras-Díaz says that for species with broader ranges and less specific habitat requirements, the cloaking of location data might not interfere as much with the generation of accurate range maps. However, for some species even 1 kilometer of bias could lead to distortions—and without an accurate understanding of species’ ranges, it is incredibly difficult to protect them, she says.

Ihlow agrees. In her view, the locations of species that are heavily affected by poaching and the pet trade should stay obscured in the database, but the vast majority of other threatened species could benefit from exact location data.

Scott Loarie, co-director of iNaturalist, would like to see more of its data be uncensored. However, he emphasizes the difficulty of appeasing both those who want the data to be accurate and those who are afraid of them being misused. “This is one of these issues where there’s very strong voices on both sides,” he says.

He hopes iNaturalist can work with organizations such as the International Union for Conservation of Nature (IUCN), which maintains a list of threatened and endangered species, to find solutions. For instance, if IUCN added tags specifying the threats each species faces, iNaturalist could filter out those affected by trafficking and poaching and stop cloaking the locations of those that aren’t.

As for Contreras-Díaz’s carnivore study, she ended up having to contact individual iNaturalist users to ask for exact locations. “It is a titanic task.” She points out that eBird, another database that collects observations from the general public, has a process through which researchers can request exact data from threatened species by filling out a form explaining their work. “This is a better strategy,” she says. “You need to ask only one person.”

Rusby Contreras-Díaz was scrolling through a biodiversity database, looking at places where people had spotted margays and jaguars, when she noticed something odd: Every location had an uncertainty of 30 kilometers. Contreras-Díaz—a mammologist at the National Autonomous University of Mexico—had hoped to use the observations to map where the two carnivores were found in Mexico. But the uncertainty in the data gave her pause.

The observations were from iNaturalist, an app and website that allows members of the public to upload plant and animal photos to an online repository and report the locations of species they’ve identified. iNaturalist has amassed a tight-knit global community of users and the data it has generated have contributed to hundreds of scientific studies. “Many species didn’t have data before iNaturalist,” Contreras-Díaz says.

But for some species, the location data aren’t entirely accurate. Whenever a user reports seeing a species that is listed as threatened or endangered, iNaturalist automatically obscures the exact coordinates, creating a random location within a certain distance that varies depending on the latitude. “People have argued that handing out [exact locations] to the public is basically a road map for poaching,” says Flora Ihlow, a macroecologist at the Dresden University of Technology who has used iNaturalist data in her own research.

Contreras-Díaz acknowledges the benefit this data cloaking can have for species at risk of exploitation. But she worries the practice could harm conservation efforts for the very organisms that most need help—by distorting maps depicting where those species might be found.

In a study published last month in Biological Conservation, she tested that possibility using locations gathered by survey scientists for three amphibian species in Argentina. Using the same process as iNaturalist, Contreras-Díaz and colleagues modified where each individual was spotted, generating a random location within 30 kilometers of the original. They then modeled each species’ range using the modified data set and the one with true coordinates. After comparing the two maps, the team found that the overlap ranged from 45% to 56%.

Contreras-Díaz says that for species with broader ranges and less specific habitat requirements, the cloaking of location data might not interfere as much with the generation of accurate range maps. However, for some species even 1 kilometer of bias could lead to distortions—and without an accurate understanding of species’ ranges, it is incredibly difficult to protect them, she says.

Ihlow agrees. In her view, the locations of species that are heavily affected by poaching and the pet trade should stay obscured in the database, but the vast majority of other threatened species could benefit from exact location data.

Scott Loarie, co-director of iNaturalist, would like to see more of its data be uncensored. However, he emphasizes the difficulty of appeasing both those who want the data to be accurate and those who are afraid of them being misused. “This is one of these issues where there’s very strong voices on both sides,” he says.

He hopes iNaturalist can work with organizations such as the International Union for Conservation of Nature (IUCN), which maintains a list of threatened and endangered species, to find solutions. For instance, if IUCN added tags specifying the threats each species faces, iNaturalist could filter out those affected by trafficking and poaching and stop cloaking the locations of those that aren’t.

As for Contreras-Díaz’s carnivore study, she ended up having to contact individual iNaturalist users to ask for exact locations. “It is a titanic task.” She points out that eBird, another database that collects observations from the general public, has a process through which researchers can request exact data from threatened species by filling out a form explaining their work. “This is a better strategy,” she says. “You need to ask only one person.”