Fire is a common consequence of large earthquakes. Sometimes, because fire suppression activities can be compromised after large earthquakes, it develops into conflagration that in turn leads to very serious loss of life and property. A GIS (Geographic Information System) model containing property and valuation data is shown to be a useful platform for modelling the spread of post-earthquake fire in the urban setting. We describe two approaches, one static and one dynamic. The static approach relies on a simple buffering technique to define potential “burn-zones” that are sampled randomly to give estimates of losses. From repeated sampling we are able to assess the probability of exceedance of various levels of loss as a function of the number of ignitions and the spacing between buildings. The dynamic approach uses a cellular automaton technique for determining both the rate and extent of fire spread in response to a wide range of factors including wind, radiation, sparking, branding, and individual separations of buildings. It is more realistic than the simple model but runs much more slowly.

The dynamic model utilises a set of fire spread “rules” that are based on fire physics modified by historical data. Calculations are made of the radiant heat flux from burning areas, and the ignition criteria for nearby buildings take into account cladding materiality. Spread by sparks and flying brands is dependent on wind speed and a user-defined probability. The model can be run in simulated real time for use in real events.

Neither model is yet fully developed. Two important features remaining to be incorporated are the effects of ground slope and vegetation. Bearing this in mind, some conclusions that may be derived from preliminary applications of the models are as follows:

• Losses due to fire following a major earthquake centred on Wellington City are likely to be smaller than losses due to shaking provided the wind at the time is no stronger than a “moderate breeze”. However, fire losses are likely to become severe for “near gale” and stronger winds.
• Firebreaks are not an effective way of minimising fire losses when there are multiple (30 or so) ignitions. A better approach is to make buildings resistant to fire-spread.
• For “calm” to “fresh-breeze” conditions the fire loss is roughly proportional to the number of ignitions. Thus minimising the number of ignitions is a good way of mitigating losses due to post-earthquake fires, as is the immediate containment of any fires that do start.
• The cost of not fighting fires is significant even under normal conditions when there is usually only one fire at a time. For a single ignition the 50th percentile loss varies from about $10 million in calm or light-wind conditions to more than $1 billion for severe wind conditions.