Small lakes and ponds are the unsung heroes of Arctic and Subarctic ecosystems! Despite their small size, they pack a big punch when it comes to regulating the climate, supporting biodiversity, and influencing greenhouse gas emissions. Yet, mapping these tiny but mighty water bodies—especially those under 0.01 km²—has always been a major challenge. These dynamic ponds and lakes, often created by permafrost thaw, can release significant amounts of methane and carbon dioxide as the frozen ground warms, affecting the global carbon cycle.
But here’s the breakthrough: researchers have developed HLWATER V1.0, an innovative image analysis tool powered by the advanced AI model Mask R-CNN. Trained on high-resolution PlanetScope satellite images, this AI can automatically identify and map even the smallest ponds, down to 166 m²! The research was conducted in Nunavik, Subarctic Canada, a stunning and complex landscape where tundra meets boreal forest, dotted with everything from glacial lakes to ponds in peatland areas.
And the results? Game-changing! The AI model successfully mapped diverse types of water bodies, even in the most challenging environments where traditional methods fail. This new approach opens up the possibility of monitoring vast Arctic and Subarctic regions more accurately and comprehensively. By tracking changes in these small water bodies, scientists can better understand the impact of climate change and how greenhouse gases are released from thawing permafrost.
This study showcases the incredible potential of artificial intelligence and high-resolution satellite imagery to revolutionize environmental monitoring. It’s a major step forward in the global mission to understand and adapt to climate change, and protect these vital, yet often overlooked, ecosystems!
Source: Freitas, P., Vieira, G., Canário, J., Vincent, W., Pina, P., & Mora, C. (2024). A trained Mask R-CNN model over PlanetScope imagery for very-high resolution surface water mapping in boreal forest-tundra. Remote Sensing of Environment. 304. 10.1016/j.rse.2024.114047.
Author: Diana Martins
