Lessons learned from applying varying coefficient model to controlled simulation datasets

The development of site- and path-specific (i.e., non-ergodic) ground-motion models (GMMs) can drastically improve the accuracy probabilistic seismic hazard analyses (PSHAs). varying coefficient model (VCM) is a novel technique for developing non-ergodic GMMs, which puts epistemic uncertainty into spatially coefficients. coefficients at nearby locations are correlated by prior distribution imposed on Gaussian Process. correlation structure determined data, later used to predict their uncertainties new locations. It important carefully verify before its results be accepted engineering community. This study series simulation-based controlled datasets from CyberShake test modified VCM technique, partitions source, site, path terms. Because simulation parameters (inputs) known, verification what recovered simulations straightforward. We found that site effects in reliably estimated VCM. However, densely-located self-similar events along pre-defined faults violate isotropic assumption underlying VCM, thus preventing capturing genuine source effects. For effects, cell-specific attenuation approaches fail recover anelastic pattern 3D velocity model, most likely due other unmodeled inappropriate wave-propagation path. Instead, midpoint approach only considers aggregated better strong within basins fixing length Lessons learned this not provide guidance future applications both empirical datasets, but will also guide further with emphasis recovery