The application of artificial neural networks for wildfire risk prediction

The City of Cape Town is declared the most fire-prone city in South Africa. This is attributed to its unique topographical, vegetative and climatic features. A novel data-driven intelligent system utilising artificial neural networks is proposed and developed for wildfire risk assessment for the City of Cape Town. The model uses vegetation, climate and location features to predict a rating corresponding to the risk of wildfire ignition for two different vegetation types. The system is trained and tested on historical fire incidence data from 2009-2015 and produces categorical outputs of low, moderate, high and extreme. Overall, the system is able to perform with an accuracy of up to 0.97 with a precision of 0.87 and a recall of 0.88.