An Enhanced Model for Wildfire Propagation Prediction Using GIS

Abstract

Wildfire modeling and simulation has been one of the major subjects under intense experimental research and theoretical work to address the ever-growing crisis of wildfire. Many researchers have benefitted from these work and have developed multiple wildfire propagation prediction systems for decision support. Despite the large-scale effort undertaken by the scientific community, it can be also observed that these advancements have become limited to the developed countries of the world. This can be attributed to the fact that a reliably accurate wildfire behavior model requires many input variables and acquiring these variables requires a great deal of infrastructure already in place. These infrastructures can be quite costly, making it infeasible for the developing countries to develop a wildfire propagation prediction system. The purpose of this research is to enhance an existing wildfire model in a manner that it requires less infrastructure at an acceptable accuracy level. The study was begun by analyzing the existing models for extensibility and enhanceability. It was discovered that the Rothermel’s Surface Fire behavior model can be enhanced by eliminating some of its many variables. Therefore a set of variables were selected through some rationale and were experimented upon using GIS platforms to observe the effect they have on the Rothermel’s model. The study was conducted using historical wildfire data and the primary measure used was the Jaccard Similarity Coefficient. To assess the practicality of the model, a novel framework named ‘MOD (Most Occurring Data) Sign’ analysis was proposed. The results of the study show that ‘fuel particle moisture’ and ‘live fuel load’ variables have significantly less effect on the Rothermel’s model. It was also discovered through the MOD Sign Analysis that ‘fuel particle moisture’ was the more practical variable to eliminate rather than ‘live fuel load’. Finally, it was concluded that a simplified model can be derived from the Rothermel’s model by eliminating ‘fuel particle moisture’ variable and while ‘live fuel load’ may also be eliminated, the resulting model will not be suitable for decision making.