Modeling Wildfire Spread with an Irregular Graph Network

The wildfire prediction model is crucial for accurate rescue and rapid evacuation. Existing models mainly adopt regular grids or fire perimeters to describe the landscape. However, these have difficulty in explicitly demonstrating local spread details, especially a complex In this paper, we propose with an irregular graph network (IGN). This implemented IGN generation algorithm characterize wildland landscape variable scale, adaptively encoding regions dense nodes simple sparse nodes. Then, deep learning-based designed calculate duration of each edge under environmental conditions. Comparative experiments between widely used simulation were conducted on real Getty, California, USA. results show that can accurately spatiotemporal characteristics novel form while maintaining competitive refinement computational efficiency (Jaccard: 0.587, SM: 0.740, OA: 0.800).