Probabilistic Modeling of Post-Earthquake Fire in Wellington, New Zealand

Abstract

Wellington, the capital of New Zealand, has both high seismic and high post-earthquake fire risk because it straddles the highly active Wellington Fault, has many closely spaced wooden buildings, and has a fragile water supply system. Repeated modeling of a Wellington Fault earthquake showed that the distribution of fire losses was much broader than that of the shaking losses, so that while fire losses were usually much smaller than the preceding shaking losses, they could occasionally be much greater than the shaking losses. Probabilistic modeling using a synthetic catalog of earthquakes gave estimates of post-earthquake fire losses in Wellington that were relatively minor for return periods up to 1,000 years, equal to the shaking losses at about a 1,400-year level, and that dominated the losses for 2,000-year and longer return periods.

Modeling and Estimating Post-Earthquake Fire Spread

“…The radiator is assumed to be rectangular with a width that is a multiple of 3 m reflecting cell sizes and a height of 4.5m to give a reasonable average approximation of the radiation heat flux produced by a burning house (see also pp. 15–17 of Cousins et al 2002). The radiator size, in conjunction with the distance to the target, determines the view factor.…”

“…In New Zealand, a database used to determine valuations for property is available. This database is described further in pages 7–9 of Cousins et al (2002), and includes the following information: Building floor area Building footprint (land coverage) area Date of construction Type of wall cladding Type of roof cladding Value of capital works (building value excluding land value) Building use (residential, commercial etc. ) …”

“…The dynamic fire-spread model uses a “cellular automaton” technique to model the spread of fire over time (Cousins et al 2002, pp. 7–9).…”

“…The emissivity and temperature of the radiator varies within a range of reasonable or possible values. A discussion on this, and a sensitivity study of the values used, is given in pages 18–19 of Cousins et al (2002). The value used for emissivity was 0.9 and for radiator temperature was 1000°C.…”

“…It has been used to predict expected fire losses for a very large synthetic catalogue of earthquakes and compares these with shaking losses. Previous publications include details on the initial model development (Cousins et al 2002, Thomas et al 2003), probabilistic modeling (Cousins and Smith 2004, Thomas et al 2005), comparison with the 1931 Hawke's Bay Earthquake (Thomas et al 2006), fire spread between windows and roofs and effects of vegetation (Heron et al 2003), and, in conjunction with the New Zealand Fire Service, an investigation into the model's usability as a firefighting tool (Heron et al 2007). This paper describes this work and expands on it with a comparison of the model and fire spread after the 1995 Kobe earthquake.…”