COHERENT CHANGE DETECTION FOR REPEATED-PASS INTERFEROMETRIC SAR IMAGES: AN APPLICATION TO EARTHQUAKE DAMAGE ASSESSMENT ON BUILDINGS

Abstract. During disaster response, the availability of relevant information, delivered in a proper format enabling its use among the different actors involved in response efforts, is key to lessen the impact of the disaster itself. Focusing on the contribution of geospatial information, meaningful advances have been achieved through the adoption of satellite earth observations within emergency management practices. Among these technologies, the Synthetic Aperture Radar (SAR) imaging has been extensively employed for large-scale applications such as flood areas delineation and terrain deformation analysis after earthquakes. However, the emerging availability of higher spatial and temporal resolution data has uncovered the potential contribution of SAR to applications at a finer scale. This paper proposes an approach to enable pixel-wise earthquake damage assessments based on Coherent Change Detection methods applied to a stack of repeated-pass interferometric SAR images. A preliminary performance assessment of the procedure is provided by processing Sentinel-1 data stack related to the 2016 central Italy earthquake for the towns of Ametrine and Accumoli. Damage assessment maps from photo-interpretation of high-resolution airborne imagery, produced in the framework of Copernicus EMS (Emergency Management Service – European Commission) and cross-checked with field survey, is used as ground truth for the performance assessment. Results show the ability of the proposed approach to automatically identify changes at an almost individual building level, thus enabling the possibility to empower traditional damage assessment procedures from optical imagery with the centimetric change detection sensitivity characterizing SAR. The possibility of disseminating outputs in a GIS-like format represents an asset for an effective and cross-cutting information sharing among decision makers and analysts.