Categorisation and harmonisation of probabilistic earthquake hazard assessments with respect to statistic representation of input data

Probabilistic hazard calculations were tested with respect to input data uncertainties. Statistical approximations of input data allow the regression coefficients and their standard deviations to be determined. Since inaccuracies of the regressions closely depend on the input data uncertainties, the standard deviations of the individual coefficients were used as a measure of input data parametrisation. Differences in the hazard outputs, caused only by various combinations of the regression coefficients and their standard deviations determined for the cumulative recurrence graph and the attenuation law, were studied. It was concluded that the hazard calculation conditions should be (a) strictly defined and always quoted with every hazard assessment, and (b) standardized in order to establish a common liability of earthquake hazard calculations. The established categorisation of the earthquake hazard assessments divides the probabilistic hazard calculations according to their reliability and helps users to apply them in earthquake engineering practice. The credible hazard assessments corresponding to 90% probability can be recommended for common structures whereas the maximum credible hazard assessments coinciding with 95% probability yield values close to hazard assessments of important structures, hospitals, bridges, etc.