ML Helps Eradicate Rats From Island Ecosystems

Off the coast of southern California sits five wonderful islands that have preserved the natural environment that the coastal region once was. Isolation over thousands of years has led to a unique ecosystem full of rare and endangered species. Visitors often come to the islands to hike, camp, or take in the scenic views. However, with each boat that transports visitors to the islands comes the opportunity for invasive species to make their way to the island as well. Specifically, rats can make their way to the island and disrupt the delicate ecosystem conservationists have tried to preserve for so long. Indeed, rats carry diseases and predatory behaviors that make them responsible for 40 to 60 percent of all bird and reptile extinctions on islands worldwide.

Knowing this information The Nature Conservancy (TNC) has put in a considerable effort to detect and monitor the amount of rats making their way onto the islands. Traditionally, invasive species have been detected and monitored using battery powered camera traps which would save data to an SD card. However, when trying to apply this method to detect rats many issues quickly became apparent. First, placing cameras in strategic locations where rats can be spotted is difficult and makes any camera maintenance needed difficult as well. Second, the cameras tended to fill up with unhelpful pictures due to wind blowing leaves and grass causing the camera’s motion sensor to trigger. Finally, there would be major gap times between when a rat was detected by a camera and when they were seen by the team on pictures. Making any potential response too slow to be effective.

The solution TNC eventually began to implement consisted of using thirty solar-powered trap cameras that had wireless technology for transmitting data. The solar panels would help alleviate battery draining issues. On the other hand, the wireless capabilities allowed for remote data collection and live streaming possibilities. The images captured by a camera would be transported through a wireless mesh network that the cameras form before being uploaded to an AWS-hosted data management platform. The platform, called Animl, is an open source platform developed by TNC that allows for the use of machine learning algorithms to predict what is in images. The process occurs in real-time and also includes filtering, analysis, and advanced querying capabilities. Overall, the system filters through vast amounts of visual data to detect which images capture rodents. A text or email is then sent to staff members notifying them of the event.

“The challenge from a computer vision perspective was that we knew we were going to see a lot of rodents. But we needed to be able to also distinguish between the mice and the rats. Because one is supposed to be there, while the other is not,” said Nathaniel Rindlaub, a software developer with TNC. This is where staff members will then check the images to determine the accuracy of the artificial intelligence.

The machine learning algorithms are helping to keep the ecosystem of the islands flourishing. The team is looking forward to seeing the technology adapted as a biosecurity application in other ecosystems as well. Currently, TNC has plans to implement similar systems on other islands including Catalina Island and Anacapa Island. Other potential future locations may include Puerto Rico, an island near Bermuda, and Palmyra Atoll.