Earthquake losses and casualties: A statistical analysis
Statistical patterns governing the number of casualties and economic losses caused by earthquakes are examined for the period 1900--1996. It is shown that the trend of nonlinear growth of total loss with time, typical of natural disasters, is due to the "heavy tail" in the empirical density of loss distribution rather than to changes in the occurrence of disasters, the latter being of minor importance. The empirical distribution is modeled by the Pareto law to estimate parameters of the distribution, typical loss values, and confidence intervals. Since the maximum possible loss is obviously limited, the Pareto law is truncated to estimate the recurrence period and the size of the maximum possible disaster. We consider disaster occurrences in relation to social and economic factors. A method of predicting total loss values is suggested. The errors and limits of applicability for this method of prediction are discussed by taking into consideration socio-economic and demographic situations.