Estimating Size Distributions of Building Areas for Natural Hazards Assessments

It is very helpful to be able to estimate the number of buildings in floor area size categories for various purposes in risk assessment, vulnerability reduction, emergency response, recovery and reconstruction planning. We compile a frequency distribution by floor area from a data set for 900,000 buildings using complete enumerations of buildings from several urban areas. Three models are fit to the data: the lognormal distribution and two regression models based on the negative exponential. A regression model estimating the percentage of buildings larger than a given size provides the best fit and is a useful technique for making predictions. Since estimates produced by regression techniques have known error distributions, the results will only be approximate. For places for which the building stock has considerably different characteristics from the data set, the parameters of the model will have to be recalculated. An actual example of an application of the model is presented.