Prediction of wildfire rate of spread in grasslands using machine learning methods. (October 2022)

• Record Type: Journal Article
• Authors: Khanmohammadi, Sadegh; Arashpour, Mehrdad; Golafshani, Emadaldin Mohammadi; Cruz, Miguel G.; Rajabifard, Abbas; Bai, Yu
• Is Part Of: Environmental modelling & software. Volume 156(2022)
• Publication Date: 2022-10
• Subjects: Artificial intelligence -- Bushfire and wildfire -- Machine learning -- Rate of fire spread -- Remote regions -- SHAP sensitivity analysis; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2022.105507
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Prediction of wildfire propagation plays a crucial role in reducing the impacts of such events. Various machine learning (ML) approaches, namely Support Vector Regression (SVR), Gaussian Process Regression (GPR), Regression Tree, and Neural Networks (NN), were used to understand their applicability in developing models to predict the rate of spread of grassfires. A dataset from both wildfires and experimental fires comprising 283 records with 7 features was compiled and utilized to develop and evaluate ML-based models. These models produced excellent fits to the model development dataset. Model fit against the evaluation dataset resulted in higher errors, with some of the models that yielded the lowest error against the model development dataset, producing the highest errors against the evaluation dataset. The predictive performance of the best ML-based models against that of operational models was evaluated. The SHAP visualization tool was used to determine the most influential variables in the best-performing models. Highlights: Predictive models were developed for grassfire rate of spread using machine learning. Performance of machine learning-based models against empirical approaches was compared. Sensitivity of machine learning models to input variables was explored using Shapely additive explanations.