Modelling the relative risk of large fires across the informal settlements of Cape Town
Home to an estimated 1 billion people globally, informal settlements are urban environments that are subject to a high risk of extensive fire spread. Their dense layouts and light, combustible building materials often facilitate the spread of fire through tens or hundreds of homes at once, rendering the inhabitants homeless. Tackling this issue requires a sound understanding of the many spatial factors which can contribute to fire spread. The aim of this study was to quantify the relative risk of large fires across informal settlements in Cape Town, South Africa – a city which has a notable history of devastating informal settlement fires. This was conducted primarily by developing a risk-scoring model based on fundamental fire dynamics and a survey of expert opinion on informal settlements. The study included a review of past disaster risk studies to aid the establishment of solid principles for the risk modelling method. A ‘pairwise weighted’ risk model was developed, using GIS software to quantify the spatial environment. It showed a good degree of success in identifying settlements that have a history of severe fires, such as Masiphumelele, Imizamo Yethu and Kosovo, as being of very high fire risk. A particular advantage of the model is its ability to recognise three different categories of fire risk, imposed by infrastructural factors both within and external to a settlement, and environmental factors. However, the fire history data used as a metric to verify the accuracy of the model was unfortunately not of sufficient quality to facilitate a rigorous numerical validation of the model. Fire risk mapping for informal settlements is a relatively new field of research, therefore many potential developments to the model were also proposed. The relationship between climate and informal settlement fire spread is currently poorly understood so it must be studied and adapted accordingly within the risk model. This could further contribute to modelling of seasonally variable fire risk. Furthermore, future methods for modelling risk directly from estimates of settlement density should be developed, to allow for automatic satellite image processing. This would be of great benefit as it would speed up the GIS-based data collation process which proved time consuming for this study.