Fine-resolution mapping of the circumpolar Arctic Man-made impervious areas (CAMI) using sentinels, OpenStreetMap and ArcticDEM

Abstract

Man-made impervious areas map is of great demand in environmental and urbanization studies since impervious areas are considered as a key indication of urbanization, especially for circumpolar Arctic. However, to date, finer resolution and spatially continuous impervious areas information remains scarce in the Arctic. In this study, we developed an accurate and complete circumpolar Arctic Man-made impervious areas (CAMI) map at a resolution of 10 m by combining Sentinel-1 C-band Synthetic Aperture Radar, Sentinel-2 multispectral images, OpenStreetMap, and ArcticDEM via Google Earth Engine platform. A random forest classifier model was trained and used to generate corresponding impervious areas map for the year 2020. The evaluation results suggested that CAMI was the most accurate with an overall accuracy of 86.36% and kappa coefficient of 70.73% as against the three existing impervious areas products. Based on the generated map and OpenStreetMap, we estimated that total impervious areas area in the Arctic has achieved 807.80 $$k{m^2}$$, of which roads, industrial and resident land were three major land use types, accounting for 54.08%, 17.85% and 10.34%, respectively. The CAMI map will support for new application and provide advanced insight into the infrastructure vulnerability evaluation and environmental sustainability in the Arctic.

Keywords

• artificial impervious areas
• 10-m spatial resolution map
• random forest
• northern arctic treeline
• google earth engine (gee)