Two proposals are presented on the estimation of maximum accelerations of strong ground motion near the causative fault. First, the crucial selection of the form fitted to observed accelerations as a function of distance from the source is discussed. In particular, evidence is given that the attenuation form adopted by Joyner et al. (1981) constrains the attenuation parameters so that inferences on near-fault motion and magnitude dependence are questionable. An alternative attenuation form is proposed which decouples far-field variations from near-field variations. Nonlinear least-squares regression to the same data set confirms the flattening of attenuation curves near to the source (<10 km) in the magnitude range 5.0 < M < 7.0 and shows no indication (from the small amount of data available) of any additional increase in acceleration for M > 7 earthquakes.
For 6.0 ≦ M ≦ 7.7, the regression yields
where y is the peak horizontal acceleration in g, and x (km) is the closest distance to the surface projection of the rupture.
Second, a robust and easily computed parameter is defined for significant peak acceleration that meets many engineering requirements. This “effective” peak acceleration is obtained by developing histograms for the number of peaks and troughs on the observed record and by choosing the acceleration value at about the 90 percentile level. This truncation excludes scattered outliers of high-amplitude peaks not representative of the general distribution of the ground motion amplitudes. Corresponding values for “effective peak acceleration” are tabulated for part of the basic data set, and it is demonstrated that the scatter about the attenuation regression line is reduced using the proposed parameter.