Building fire impact model

Abstract

The aim of the Building Fire Impact Model (BFIM) is to study the impact of human intervention in a bushfire situation, focusing on the peri-urban communities. The model resides within the Fire Impact and Risk Evaluation Decision Support Tool (FireDST) which simulates fire conditions (Bushfire CRC, 2013). The BFIM recognises that occupants can prepare their house to withstand a bushfire, and if such an event occurs, they can defend their house and may reduce total house loss in the bushfire. To take into account the large uncertainties associated with human intervention, the BFIM was designed to be a probabilistic model, i.e. it uses random variables to represent the main characteristics of occupant intervention. To deal with the complexity of such a model, a mathematical technique called Probabilistic Event Trees (PET) is utilised. In this technique events are represented by the nodes of a mathematical tree and the probability of these events occurring are the tree branches. The conditional dependency of variables is represented by their relative location in the tree. The BFIM consists of two PETs. The first PET simulates occupant intervention to defend their house from ember attack and possible help from neighbours and the rural fire brigade. The ability of occupants to defend a house depends on their preparation and the preparation of the house to withstand the fire. Occupant preparation comprises purchase of fire-fighting equipment, undergoing training, conducting regular drills, development of an evacuation plan, etc. House preparation involves activities such as cleaning gutters, cutting trees and grass surrounding the house, etc. A series of occupant characteristics are used to assess their efficiency in fighting the ember attack. This tree also calculates the impact of wind speed on the house. Houses damaged by wind/debris during the fire have a higher probability of being ignited by an ember attack. The second PET calculates the impact of radiation in 'house-to-house' fire spread. This tree examines the impact that houses ignited in the first PET have on nearby houses. The main variables to consider in this part of the model are wind damage to a house, proximity of a house to a neighbouring ignited house and the number of neighbouring houses burnt by the bushfire. An important problem also considered in this tree is 'smoldering embers'-embers that penetrate the house through the roof and slowly burn flammable material due to lack of oxygen-which can results in house loss long after the fire front has passed through the region. Wind and wind damage to the roof increase the danger from this type of ember attack process. To simulate the dynamic characteristics of these issues, the two PETs are solved sequentially: results of the first PET are used to set up the correct fire conditions for solution of the second PET. For this reason the FireDST simulation model is run three times (passes) for a region. In the first pass, the model calculates the fire conditions in the region of interest and returns the number and location of houses potentially impacted by the fire. In the second pass, the first tree of the BFIM is solved to calculate the impact of human intervention. This sets up different fire conditions and hence FireDST recalculates the number of houses burnt under the new conditions. In the third pass, the second tree of the BFIM is solved to assess house-to-house fire spread. Some houses with a high probability of being 'saved' in the second pass can be reclassified as having a high probability of being 'burnt' after the third pass has been completed. To illustrate the model, an example case based on the Kilmore bushfire that occurred on 'Black Saturday' (7 February 2009) is provided. Results show that human intervention to fight the fire threatening a house can make a substantial difference to the number of houses burnt.