Uncertainty Correction Of Ground Motion Attenuation In PSHA

PSHA has become an important basis of the seismic zoning mapping, however, dueto the lack of the earthquake and geological data，poor understanding of focal mechanism,seismic wave transmission and the local site conditions, there will be a lot ofuncertainties in the result, where the uncertainty of attenuation relationship is dominant.Corresponding correction naturally became an issue for people; especially the largeincrement of risk in low probability, which is always controversial. This paper researchesthe distribution characteristics of random error with PEER database, and put forward animproved correction calculation method for attenuation uncertainty.Summarization of currently widely used uncertainty correction method of groundmotion attenuation relationship shows that the increment of risk in correction is mainlythe contribution of the positive part of through specific examples, and the contributionof negative part to the correction result is very small. Truncated integration interval hasvery significant influence on the calibration results, and at present it‘s appropriate totruncate it at three times the standard deviation, which has little difference from that atfour times, if smaller, the result will be unreliable. Both standard deviation and meanhave an important impact on uncertainty correction.Analyze the distribution of random errors in attenuation relationship statistically,with expanded strong ground motion and attenuation relationship developed in NGA.Examine the correlation between characteristic of distribution and magnitude of M, R,and the ground motion parameter Y, point out that the correlation of magnitude is weak,distribution is not associated with the distance basically, and find that the mean ofrandom error distribution has stable variation tendency with the increase of PGA. Basedon the discover, put forward an improvement of probability density function f y()assegmentation normal distribution with mean of non-zero, and develop the uncertaintycorrection calculation method by segmentation, called segmentation correction method,with the mean and standard deviation in each section definition respectively.Three curves from different seismic risk levels sites are corrected with segmentationcorrection method based on the segmented statistical result of C-B NGA. It show that, incomparison with general correction method, PGA is slightly higher in the zonecorresponding to often encounter fortification level with high exceedence probability and low PGA, and PGA is significantly lower in the zone corresponding to occasional andrare encounter fortification level with low exceedence probability and high PGA. Thedifference is more significant at high seismic risk site.If the ground motion attenuation relationship‘s random error distribution with PGAindeed has significant change, it will result in overrating the seismic risk unreasonably ifwe neglect that and correct uncertainty with normal probability density function ofgeneral standard deviation and the zero mean. In comparison with general correctionmethod, segmentation correction presents a more reasonable reaction of differentinfluence that different section of the ground motion parameter Y has on uncertaintycorrection, as the former introduced the ground motion parameter Y as a parameterrelated to the distribution of random error in attenuation relationship in the way ofsegmentation.