At Island Conservation, we pride ourselves on being strategic in our investments. One area that can return big dividends is our willingness to be early adopters of the latest and greatest technology to help us holistically restore islands for nature and people. With three decades of experience under our belt, we’re always looking for new ways to improve the scope and scale of our work around the world. Recently, we’ve been working with Conservation X Labs’ Sentinel AI to detect species on the remote islands where we work. Our GIS & Data Science Specialist, Sierra Moore, gives us the low-down on this cutting-edge tech, which, together with our partners, have been putting to the test on Floreana Island in the Galapagos and around the globe! 

However, given the remote environments we work in, we face challenges in effectively using this tool. Typically, our field staff will deploy a uniform grid of camera traps across the entire island, which means that, in some cases, hundreds of camera traps are deployed. The images taken on the cameras need to be downloaded from the device, which requires a team to physically access each camera somewhat frequently. In addition, the cameras may be trigger on moving vegetation, shadows, etc., meaning the team often must analyze thousands of images to determine if any of the photos contain our target species. We utilize AI models that are trained to detect animals in photos, which eliminates some of the processing time by filtering out many of the ‘empty’ images (or images without an animal detection). However, the time it takes to complete this entire process reduces a lot of the effectiveness of the tool. 

As an example, take Floreana in the Galapagos, where Island Conservation, amongst numerous partner organizations, implemented an eradication of invasive rodents in 2023 to restore the island’s biodiversity. At ~17,000 ha in size and much of the land covered in dense vegetation and steep terrain, the level of effort required to regularly check a grid of cameras across the whole island is extreme. This is where innovative technologies have given us the opportunity to increase the scale and efficiency of our work. On Floreana, we worked with the local NGO Fundacion Jocotoco to deploy Sentinel AI devices, created by Conservation X Labs, alongside a grid of camera traps. The Sentinels host an on-board custom AI detection model which was trained on images taken on Floreana to detect our target species. The model processes the images taken on the camera trap and outputs a text file that contains information about the species that was (or wasn’t) detected, the confidence the model had in that detection, and the latitude and longitude of the device on which the detection occurred, amongst many other key data. The Sentinel sends this information as well as a compressed image of the bounding box around a detection via satellite to an online dashboard that the team can access via internet from anywhere in the world. This means the tool provides the team with the potential to learn of a possible detection, check that the model was accurate, and implement a response to the detection without having to physically access a camera trap. 

Although the system didn’t work as efficiently as we hoped on Floreana, given issues with satellite communication, we still believe that the tool has the potential to transform the efficiency and scale at which we are able to monitor the presence/absence of species on the islands we work on. Some of the Island Conservation Innovation Team’s on-going work with CXL includes looking at ways to overcome the satellite communication issues which will supercharge work on Floreana and beyond. One method we are looking into is deploying Sentinels that communicate the detection information over LoRa instead of satellite. We’re also shifting to another provider next year.