A Neural Network Model for Wildfire Scale Prediction Using Meteorological Factors

A forest fire is a natural disaster that destroys forest resources, thus having a severe impact on humans and on the animals and plants that depend on the forest environment. This paper presents a model for predicting the scale of forest wildfires of Alberta, Canada. A fire's scale is determined by the combination of the fire's duration and the size of the area it burns. Our prediction model enables fire rescuers to take appropriate measures to minimize damage caused by a wildfire based on its predicted scale in the fire's early stages. The modeling data were collected from the Canada National Fire Database (CNFDB) published by Natural Resources Canada, which includes wildfire and meteorological data for Alberta, Canada. The size of the burned area and the fire's duration were used to estimate the scale of a wildfire. After multi-collinearity testing and feature normalization, the data were divided into training and testing sets. Taking the meteorological factors as input values, a backpropagation neural network (BPNN), a recurrent neural network (RNN), and long short-term memory (LSTM) were implemented to establish prediction models. Of these classification methods, LSTM exhibited the highest accuracy, 90.9%. The results indicate that it is feasible to predict the scale of a forest wildfire at the beginning of its occurrence using meteorological information.