S. Lu, C. J. Liang, W. G. Song and H. P. Zhang (2013) Frequency-size distribution and time-scaling property of high-casualty fires in China: Analysis and comparison. Journal/Safety Science 51 209-216. [In English]
Web link: http://dx.doi.org/10.1016/j.ssci.2012.07.001
Keywords: Power-law distribution, Time-scaling property, Fatality statistics, Frequency-size distribution, Scaling behavior, SELF-ORGANIZED CRITICALITY, POWER-LAW DISTRIBUTION, FOREST-FIRE, CITY, FIRES, STATISTICS, SEQUENCES, FATALITY
Abstract: High-casualty fires (HCFs) involving 3 or more fatalities from the Fire Statistical Year Book of China and the Chinese Fire Services from 2002 to 2009 were analyzed. This study investigates the frequency-size power-law distribution and time-scaling property of HCFs. The results show that HCFs exhibit a frequency-size power-law distribution regardless of whether the fire size is represented by fatalities, direct loss or burned area. The frequency-fatality power-law distribution is a common phenomenon in fire accidents, even in international data. Six factors (place, cause, time of day, season, year and region of the fire) were analyzed to assess their effects on the frequency-fatality distribution and compared using the scaling exponent. Factors, such as a non-residential place, an electrical cause, the winter season and regions with strong economies, cause fire frequency to decrease slowly with increases in fatalities. That is, these factors are associated with higher fatalities. The time-scaling properties were detected using the Fano Factor and Allan Factor, and the results illustrate that HCFs exhibit obvious time-scaling behavior after the fractal onset times. The results of the fractal onset times imply that the time-scaling behaviors of HCFs exhibit daily and monthly periodicity. The 1-day onset time is consistent with previous studies, but the 1-month onset time is different from that reported previously. The time-scaling exponent decreases significantly with increasing fatalities, which indicates that fire sequences with more fatalities tend to behave as Poisson processes. (c) 2012 Elsevier Ltd. All rights reserved.