Speaker: Elias Manos, University of Connecticut
Date: Thursday April 30, 2026 at 9:00 am AKT
Abstract: Disaster risk assessments performed for heavily populated mid-to-lower latitude regions are often based on complete, high-resolution building stock information with detailed attributes. In the Arctic circumpolar permafrost region (ACPR), limited satellite data coverage, low commercial incentive, and a sparse population have all historically contributed to a lack of building stock mapping efforts. However, petabyte-scale high-resolution commercial satellite imagery and elevation data spanning the ACPR, along with artificial intelligence and high-performance computing, now provide the opportunity to fill critical infrastructure data gaps that will improve our understanding of Arctic climate risk and support adaptation planning. In this work, deep learning models were used to map buildings and their occupancy across the ACPR from Maxar satellite imagery. Residential building story counts were estimated using ArcticDEM elevation data and national construction standards. Combining this building map with projected ground instability, Monte Carlo-estimated mid-century circumpolar building damages amount to 76 B USD (2.88–259.48) under moderate emissions and 261 B USD (17.01–379.11) under high emissions (median and 5th–95th percentile range), with uncertainty primarily driven by climatic variability, engineering practices, and deep learning model errors. Results indicate that authoritative building stock information across the Arctic (e.g., U.S. FEMA, NRCan, Rosstat) and existing scientific literature comparatively underestimate infrastructure risk across degrading permafrost regions.
For additional PDG information: https://arcticdata.io/catalog/portals/permafrost/Stay-Connected
Provided by: Emily Longano
