Satellite based monitoring of Arctic settlements on thawing ground

Infrastructure in the Arctic is expanding rapidly due to ongoing industrial development, yet it faces increasing challenges as climate warming accelerates permafrost degradation and leads to or increases ground instability. Satellite records support the identification of infrastructure developments. Remote sensing based ground surface deformation monitoring is another key component for managing permafrost-related infrastructure risks and supporting community planning.

To systematically map human-impacted Arctic coastal regions, the Sentinel-1/2 derived Arctic Coastal Human Impact (SACHI) dataset was developed. In this study, we updated this dataset for selected Arctic settlements in western Greenland using newly acquired imagery that captures recently constructed man-made features. Additionally, we investigated the potential of fully polarimetric PALSAR-3 L-band SAR data to complement the established Sentinel-1 (C-band)/ Sentinel-2 workflow, aiming to improve the detection and characterization of infrastructure. The scheme is based on fusion of two machine learning techniques, Gradient boosting machines (GBM) and a deep learning approach using convolutional neural networks. Specifically, the added value for building detection as part of the GBM analyses can been shown. Eventually we combined the settlement information with long-term vertical ground deformation for selected settlements.