An empirical model for scaling Fourier Amplitude Spectra of strong earthquake ground acceleration in terms of magnitude, M, epicentral distance, R, and recording site conditions has been presented. The analysis based on this model implies that:
the Fourier amplitude spectra of strong-motion accelerations are characterized by greater energy content and relatively larger amplitudes for long-period waves corresponding to larger magnitudes M,
the shape of Fourier amplitude spectra does not vary appreciably for the distance range between about 10 and 100 km, and
long-period spectral amplitudes (T > 1 sec) recorded on alluvium are on the average 2.5 times greater than amplitudes recorded on basement rocks, whereas short-period (T < 0.2 sec) spectral amplitudes tend to be larger on basement rocks.
It has been shown that the uncertainties which are associated with the forecasting of Fourier amplitude spectra in terms of magnitude, epicentral distance, site conditions, and component direction are considerable and lead to the range of spectral amplitudes which for an 80 per cent confidence interval exceed one order of magnitude. A model has been presented which empirically approximates the distribution of Fourier spectrum amplitudes and enables one to estimate the spectral shapes which are not exceeded by the presently available data more than 100 (1 - p) per cent of time where p represents the desired confidence level (0 < p <1).