Abstract
In order to odtain more reliable models for the estimation of the number of fire outbreaks following earthquakes, former estimation models have been reconsidered and the fire outbreaks in the 1923 Kanto earthquake have been analysed in detail. Kawasumi's model, the first trial for the estimation of fire outbreaks in earthquakes formulated by the linear regression of the logarithm of fire outbreak ratio on the logarithm of building collapse ratio using the data of the Kanto earthquake, has been found to be effective when the fire outbreak ratio is given excluding chemical fires. Mizuno's model, the same type of linear regression as the Kawasumi's and estimated by the thirteen earthquakes'data, has been shown to cause overesitmation on fire outbreak ratio because of the biased selection of data used for the regression. Analysing the data of fifteen wards of Tokyo city in the Kanto earthquake, a new model for the estimation of fire outbreaks has been obtained with higher correlation coefficient, which is the linear regression of fire outbreak ratio without chemical fires on the logarithm of building collapse ratio. In the Kanto earthquake, major uses of buildings of fire orign were restaurants, retail shops and manufacturing industries of dishes and processed food for general fires and schools, universities and drugstores for chemical fires. The number of fires of each kind has been increased when the number of facilities mentioned above has been increased. The ratio of fire extinguishment by citizen has been decreased when the ratio of buildings collapsed has been increased.
