From this long-term data, we can gain insight into answering critical questions. Is native vegetation recovering? Are seabirds returning to nest? Are invasive predators present? Which areas need additional action to meet restoration goals? Remote sensing helps guide Island Conservation’s decisions so we can make island recovery happen more efficiently or in new ways  

The Tools Behind the Scenes 

Our toolkit is diverse: 

• Camera traps capture images of ground birds and mammals, and even crabs (which require special timelapse settings because they don’t trigger heat vision or motion sensors). 

• Acoustic sensors record thousands of hours of island soundscapes, allowing us to detect calls from new and returning species. 

• NASA satellite imagery gives us a bird’s-eye view of vegetation health using indices like NDVI (Normalized Difference Vegetation Index), tree and water cover, biomass, and canopy moisture measurements. 

Each tool has its benefits and challenges. Remote islands often lack internet, so we still rely on SD cards instead of cloud uploads. Dense vegetation means batteries instead of solar panels, so we do have to send in real humans to change out SD cards and batteries. And then there’s the sheer volume of data—hundreds of thousands of images and thousands of hours of audio that need to be processed. 

Turning Data Into Action: A Case Study on Ulong Island 

One of my favorite examples of this work in action comes from Ulong Island in Palau. Over three years, we collected 528,310 camera trap images and 14,796 hours of acoustic recordings. That data traveled 6,666 miles from Ulong to our analysis hub in Santa Cruz, California.

We don’t analyze all this data alone. Partners like Conservation Metrics help us analyze acoustic data using a machine learning technique called a Convolutional Neural Network (CNN), which can detect species-specific calls—like the signature haunting cry of Ulong’s Tropical Shearwater—hidden in thousands of hours of recordings. 

For satellite imagery, we collaborate with terraPulse and academic partners to process NASA Landsat Earth observation data into metrics that describe vegetation cover, structure, and function, which give valuable landscape-scale insights to field teams. These ‘birds-eye-view’ datasets allow them to identify areas of interest that are difficult to access, saving them time and energy in the field. 

Artificial intelligence plays a huge role in making this work possible. Our AI-assisted workflow for processing camera trap imagery uses Megadetector, a computer vision model that flags images likely to contain animals. After the model identifies potential wildlife, our team reviews each flagged image to confirm the species. This human-in-the-loop approach ensures accuracy while saving countless hours of manual sorting.  

The Power of Reproducible Science

All of this effort—AI models, acoustic sensors, satellite imagery—feeds one overarching goal: reproducible science. By using standardized methods and transparent workflows, we make sure that our findings can be verified, repeated, and built upon. These tools allow us to provide meaningful evidence of the outcomes that can be realized through conservation action on islands.

Island Conservation prides itself on being a leader in conservation science. We’re always piloting new tools and methods to gather data, but it’s the stories that the data tells that make it valuable. And time and time again, it shows that holistic restoration on islands yields outsized benefits for nature and people. I’m glad to be part of a data journey that results in more resilient and biodiverse island ecosystems!