Post-earthquake fires may be severe with substantial loss of life and property, such as those following the 1906 San Francisco, 1923 Kanto (Tokyo) and 1995 Great Hanshin-Awaji (Kobe) earthquakes. Two geographic information system (GIS) models have been developed to predict post earthquake fire losses. The simpler "static" model determines extent of spread based on possible spread distances between combustible cells, and the more complex, "dynamic" model determines extent and timing of spread based on fire spread "rules". These rules are based on seven distinct modes of fire spread; direct contact, spontaneous ignition of claddings, piloted ignition of claddings, spontaneous ignition through windows, piloted ignition through broken windows, fire spread via non fire rated roofs and branding. Aside from the first two, modes can be selected individually and given user defined or default probabilities. Fire spread modes can be added to the model or altered to suit available building information. Critical details of buildings are accessed from a buildings database or deduced using conditional probabilities. These rules are described in detail and the resulting models compared with fires after the 1931 Hawkes Bay, New Zealand and the 1995 Kobe earthquakes. A pilot study using the static model shows a significant increase in predicted fire losses if fire spread via combustible vegetation is included. The dynamic model shows some promise for predicting spread of wildfires, and combined with the urban model could be developed as a tool for modelling fire spread at the interface between wildland and the built Environment.
展开▼
