The application of empirically derived site corrections was found to not significantly improve the predictions. This is due to the fact that the corrections are derived from observed rock/site ratios and the observed rock records are likely to be more complex than the rock Green’s functions we use due to regional structure. Slide 14 compares predicted-to-observed PGV ratios for the case in which the predicted rock PGV are modified by the QTM corrections utilized by the southern California TriNet ShakeMaps.  Significant under prediction is observed at stations in the Los Angeles basin.   It may be possible with 3D numerical approaches to develop a set of 3D-to-1D corrections.
	The application of both hard rock and soft rock Green’s functions to account for site variability was found to significantly improve the comparison between observed and predicted ground motions, and is shown in slide 14.  Note that the use of multiple 1D velocity models to account for site variability may not be necessary if a set of 3D-to-1D corrections can be obtained numerically.
	The application of an empirical attenuation relationship that incorporates directivity affects was found to greatly improve the ground motion simulations by providing more rapid preliminary estimates, and providing an empirical base level that is amenable to different levels of approximation such as strong motion centroid mapping (TriNet) and source specific conservative shakemaps.
	Alternative methods of simulating ground motions to high frequency were found to improve results somewhat but not enough to offset additional computational cost.  It may be worthwhile to investigate improving the computational efficiency of these methods.