Use of pattern recognition and Bayesian classification for earthquake intensity and damage estimation

Empirical methods, which correlate intensity with a damage index based on statistical observations of past events, have been conventionally used to forecast or estimate damage of structures that might result from future earthquakes. However, difficulties often arise in the quantitative precision of such estimates because intensity scales are usually not rigorously defined, particularly with respect to the damage distribution of modern structures.

This paper presents a consistent method for earthquake intensity classification based on the theory of statistical pattern recognition. A discriminative function is developed for such identifications based on the Bayesian criterion. All statistical data required is obtained from past earthquake investigations. The method developed can be used to identify the intensity levels of an earthquake as well as to verify the intensity classification of the past earthquakes.