Bayesian Physics Informed Neural Networks for Data Assimilation and Spatio-Temporal Modelling of Wildfires

Using echo state networks to inform physical models for fire front propagation

“…An enhancement to traditional deep learning neural network approaches is to use so-called "physics-informed neural networks" that seek to encode physical relationships into the deep neural architecture (e.g., Raissi et al, 2019). These methods have recently been considered for level-set fire front propagation in Dabrowski et al (2022b). Their work is implemented in a Bayesian framework that allows for data assimilation and uncertainty quantification.…”