Data supporting a submitted journal paper: Multi-Level Earthquake Damage Assessment Using InSAR-Derived Displacement

This dataset supports machine learning–based classification of multi-level building damage using features derived from Synthetic Aperture Radar (SAR) data, including displacement, coherence, as well as building- and site-related attributes. The damage levels considered are Small, Moderate, and Severe. The <em>MultiDamageSAR-main</em> package includes four pre-trained models: one trained on damage data from the Amatrice earthquake (24 August 2016), one from the Norcia earthquake (30 October 2016), one from the Nepal earthquake (25 April 2015), and one trained on a combined dataset from all three events. These models enable both region-specific and generalised damage classification across different seismic contexts.In addition, the dataset includes the publicly available feature datasets used in this study, comprising site-related parameters such as VS30 and slope, earthquake-related parameters including Peak Ground Velocity (PGV), Peak Ground Acceleration (PGA), Pseudo-Spectral Accelerations (PSAs), and Modified Mercalli Intensity (MMI), as well as InSAR-derived products such as Line-of-Sight (LOS) displacement maps, zenith tropospheric delay maps for APS correction, and coherence maps for all events. The methodology underlying the models is described in the paper titled <em>“Multi-Level Earthquake Damage Assessment Using InSAR-Derived Displacement”</em>, which is currently under review.