This article describes the peak accelerations observed on a network of strong-motion accelerographs above the subduction zone in Guerrero, Mexico, as a function of magnitude and hypocentral distance. Our empirical description differs from regression analysis, however, in that it produces a table of values of peak acceleration, for selected values of M and R, and an interpolation rule to obtain peak acceleration at intermediate values. With this approach, the existence and extent of saturation of peak acceleration as magnitude increases is dictated by the data, and unlimited flexibility is allowed in the distance dependence of the peak acceleration.
This nonparametric approach does not allow extrapolation beyond the range of the data, and is thus most useful when the data spans a large magnitude and distance range. After 7 yr of recording, the digital Guerrero accelerograph network has obtained enough data to test the concept. The records include events with magnitudes ranging from under 3 to 8.1, with all magnitudes recorded at distances ranging from nearly directly above the rupture to distances where acceleration has become too small to trigger the instruments. The data used in this study consist of 528 recordings (each three components) from 146 earthquakes that occurred between 1985 and 1990. All but one station are on rock sites. Since all the accelerograms are recorded by one of 30 stations, there is an average of about 18 records for each station.
The results suggest that at short distances (about 14 to 25 km from the fault), as magnitude increases the peak horizontal acceleration saturates at about 20% g. However, unlike small earthquakes, peak accelerations decrease very slowly with increased distance from the rupture in a large magnitude. Station corrections for peak acceleration reduced the rms misfit by about 15% for the horizontal component, by 12% for the vertical component, and by 21% for the ratio. For some stations, horizontal corrections are close to vertical corrections, but others show very different site effects on horizonal and vertical components. The ratios of peak vertical acceleration to peak horizontal acceleration show a very weak tendency to decrease as magnitude increases, and this tendency is further reduced at most distances after correction for average site effects. The average value of this ratio is 0.70 before the station correction.
After these corrections, for 55 events with known stress drop, the residual between the predicted peak acceleration and the observed peak acceleration is correlated with stress drop (Δσ). Although scattered, the observations are consistent with a theoretical prediction that Amax ∝ Δσ0.80.