A Preliminary Damage Assessment Using Dual Path Synthetic Aperture Radar Analysis for the M 6.4 Petrinja Earthquake (2020), Croatia

On 29 December 2020, an earthquake with a magnitude of M 6.4 hit the central part of Croatia. The earthquake resulted in casualties and damaged buildings in the town of Petrinja (similar to 6 km away from the epicenter) and surrounding areas. This study aims to characterize ground displacement and to estimate the location of damaged areas following the Petrinja earthquake using six synthetic aperture radar (SAR) images (C-band) acquired from both ascending and descending orbits of the Sentinel-1 mission. Phase information from both the ascending (Sentinel-1A) and descending (Sentinel-1B) datasets, acquired from SAR interferometry (InSAR), is used for estimation of ground displacement. For damage mapping, we use histogram information along with the RGB method to visualize the affected areas. In sparsely damaged areas, we also propose a method based on multivariate alteration detection (MAD) and naive Bayes (NB), in which pre-seismic and co-seismic coherence maps and geocoded intensity maps are the main independent variables, together with elevation and displacement maps. For training, approximately 70% of the data are employed and the rest of the data are used for validation. The results show that, despite the limitations of C-band SAR images in densely vegetated areas, the overall accuracy of MAD+NB is similar to 68% compared with the results from the Copernicus Emergency Management Service (CEMS).

Automated summary

This study utilized SAR images to characterize the ground displacement and estimate the damaged areas in Croatia following the earthquake. By using InSAR technique and MAD+NB method, the overall accuracy of approximately 68% was achieved, despite limitations in densely vegetated areas. Training and validating the data were crucial in the process.