Modelling Arctic tundra capillary networks with sub-meter resolution satellite imagery..

Modelling Arctic tundra capillary networks with sub-meter resolution satellite imagery, GeoAI, and graph theory by Michael Pimenta, University of Connecticut, PDG Webinar May 2026. Tundra Capillary Networks (TCNs) are emergent drainage systems that develop across degrading permafrost landscapes. Because TCNs structure surface hydrological connectivity, their expansion represents an important window into Arctic landscape reorganization. Existing TCN mapping approaches remain difficult to scale across the circumpolar Arctic. The increasing availability of very high spatial resolution Vantor (formerly Maxar) satellite imagery provides a unique opportunity to observe these features directly. In this work, we present the first-of-its-kind scalable and generalizable workflow as part of our High-resolution Arctic Built-infrastructure and Terrain Analysis Tool (HABITAT), integrating Vision Transformer-based semantic segmentation with graph-theoretic modeling to detect and quantify TCNs from Vantor satellite imagery. Applied across approximately 1 million km2 of Alaskan tundra, this framework mapped a total TCN length of more than 2.7 million km. By translating mapped TCNs into graph-theoretic metrics such as network length, node density, component structure, and connectivity, this approach reveals spatial patterns of landscape organization that are difficult to capture through pixel-based mapping alone.