Over the past 30 years there has been significant research to advance physics-based ground motion simulations via improvement of the simulation methodologies, and refinement of earthquake source, crustal velocity, and site effects modelling. Comprehensive validation of such simulations is essential to quantify improvements in their predictive capabilities, understand limitations, and identify pathways for further improvement. This poster presents the results of ground motion simulation validation using small-to-moderate magnitude () earthquake events across New Zealand with: (1) an improved 3D crustal velocity model, which includes additional sedimentary basins and better representation of the near-surface velocity structure; and (2) a modified version of the Graves and Pitarka (2015) hybrid broadband methodology, which uses an improved high-frequency path duration model and reduced low-frequency site amplification. Results are compared with simulations using a benchmark crustal velocity model and the unmodified Graves and Pitarka (2015) methodology, and using a range of ground motion intensity measures as summary statistics. Across the entire dataset (i.e. all earthquake events and recording stations), it was found that the modified simulations had less overprediction bias than the ‘unmodified’ methodology at both short and long response spectral periods. Interrogation of partitioned residuals, from mixed-effects regression, highlights the need for improved Vs30 estimates, consideration of topographic effects, and spatially variable stress parameter – all of which are being considered in an iterative cycle of model improvement.