UCLA

National Aeronautics and Space Administration Advanced Rapid Imaging and Analysis Damage Proxy Maps (DPMs) identify potentially damaged areas based on interferometric coherence loss in Synthetic Aperture Radar (SAR) data. I propose the first framework for quantitative validation of DPMs and apply it to a variety of damage types. Surface fault rupture data is considered from the 2019 Ridgecrest earthquake sequence. Quantitative analyses take two forms: (1) statistical distributions of DPM index (I_DPM) for fault displacement ranges as box and whisker plots, and (2) empirical fragility functions that relate I_DPM to displacement exceedance probabilities relative to certain thresholds. I develop these relationships for DPM1 (one pre-event pair and one co-event pair of SAR images) and DPM2 (multiple pre-event and co-event pairs of SAR images). Both DPM types perform similarly well for distinguishing between no surface displacement and some surface displacement. The predictive power of I_DPM metrics, as measured by fragility dispersion, shows the best performance for low threshold displacements and DPM2-based indices. Recall and precision error metrics show favorable performance of fragility models for identifying locations of fault displacement, but increasing rates of false positives as fault displacement increases. Results from similar analyses conducted on surface fault rupture data gathered following the 2016 Central Italy earthquake sequence are less favorable. Building damage data is from the 2020 Port of Beirut explosion and the 2016 Central Italy earthquake sequence. The ground failure data is from the 2019 Ridgecrest and 2016 Central Italy earthquake sequences for liquefaction and landslides, respectively. These data sets show that I_DPM metrics effectively identify damage in some cases (structural damage in Italy, liquefaction from Ridgecrest) and have limited effectiveness in other cases (Beirut structural damage, incidents of landslides). These results demonstrate that I_DPM can be effective as a rapid, post-event predictive tool in flat terrain with limited vegetation (Ridgecrest) but not as effective in sloping ground with vegetation (Italy). I also observe relatively poor results when the underlying reconnaissance information may be less reliable (Beirut buildings, Italy landslides). These results indicate DPMs can be used to rapidly identify damaged areas but not to distinguish between damage levels.