Statistical properties of peak ground accelerations recorded by the SMART 1 array

Abstract Some statistical properties of peak horizontal and peak vertical accelerations are estimated using 732 accelerograms from 25 earthquakes recorded by the SMART 1 array located in Taiwan. Two aspects of the statistical properties of peak ground accelerations (PGAs) are considered: (1) the shape of the distribution function of PGA for individual earthquakes, and (2) the magnitude dependence of the standard error of PGA. Power normal distributions with shape parameters in the range 0,1 are considered as possible models for the distribution function of PGA. Based on maximum-likelihood criteria, the preferred model has a shape parameter of 0, corresponding to a lognormal distribution; however, all power normal distribution functions in the range tested are occeptable at the 90 per cent confidence level. The standard error of PGA based on a lognormal distribution is estimated for each of the 25 events. Regression of the standard error of the largest horizontal and average horizontal accelerations versus earthquake magnitude show that the standard errors decrease as magnitude increases. The resulting empirical relation for the largest horizontal acceleration is σ = 0.658 − 0.0744 M ± 0.054 for 4 . 0 ≦ M ≦ 6 . 5 , σ = 0.174 for 6 . 5 M ≦ 7 . 8 , where σ is the standard error of the natural logarithm of PGA for an individual earthquake.