Comparison of Recurrent Neural Network Architectures for Wildfire Spread Modelling

TL;DR

This paper compares GRU and LSTM recurrent neural networks for wildfire spread modeling, focusing on predicting fire continuation and spread direction using limited time series data, highlighting GRU's superior performance for longer sequences.

Contribution

It provides a comparative analysis of GRU and LSTM architectures for wildfire modeling and evaluates their effectiveness in predicting fire spread and continuation.

Findings

GRU outperforms LSTM on longer time series.

Prediction of wildfire spread direction is reasonably accurate.

Determining if a wildfire continues to burn remains challenging due to data limitations.

Abstract

Wildfire modelling is an attempt to reproduce fire behaviour. Through active fire analysis, it is possible to reproduce a dynamical process, such as wildfires, with limited duration time series data. Recurrent neural networks (RNNs) can model dynamic temporal behaviour due to their ability to remember their internal input. In this paper, we compare the Gated Recurrent Unit (GRU) and the Long Short-Term Memory (LSTM) network. We try to determine whether a wildfire continues to burn and given that it does, we aim to predict which one of the 8 cardinal directions the wildfire will spread in. Overall the GRU performs better for longer time series than the LSTM. We have shown that although we are reasonable at predicting the direction in which the wildfire will spread, we are not able to asses if the wildfire continues to burn due to the lack of auxiliary data.