Local site effects have an enormous influence on the character of ground motions. Currently, soil categories and site factors used in building codes for seismic design are generally based on, or at least correlated with, the seismic velocity of the surface layer. We note, however, that the upper 30 m (a typical depth of investigation) would almost never represent more than 1% of the distance from the source; 0.1% to 0.2% would be more typical of situations where motion is damaging. We investigate the influence of this thin skin on the high-frequency properties of seismograms. We examine properties of seismograms consisting of vertically propagating S waves through an arbitrarily complex stack of flat, solid, elastic layers, where the properties of the lowermost layer (taken at 5 km depth) and a surface layer (thickness 30 m) are constrained. Input at the bottom of the stack is an impulse. We find that the character of the seismograms, and the peak spectral frequencies, are strongly influenced by the properties of the intervening layers. However, for infinite Q, the integral of amplitude squared at the surface (which determines energy if the input and output are regarded as velocity, or Arias intensity if the input and output are regarded as acceleration) is independent of the intervening layers. Also, the peak amplitude of the seismogram at the surface is relatively independent of the intervening properties. For finite, frequency-independent Q, the integral of amplitude squared and peak amplitude decrease as t* increases. There is some scatter that depends on the intervening layers, but it is surprisingly small.
These calculations suggest that the surficial geology has a greater influence on ground motions than might be expected based on its thickness alone. They suggest that variable influences of Q along the entire path have a comparable importance for predictions of ground motions. Finally, they suggest that detailed characterization of deeper velocity structure in regions where a 1D model is appropriate gives only a limited amount of added information. Based on our 1D numerical results, we propose a new method to characterize these properties as site factors that could be used in building codes. Full three-dimensional synthetics are tested and give a similar conclusion.