CEEDS Faculty Fellows Present: Tracking Wildfire Fronts and Ice Melting with Luca Capogna

The dynamics underlying many natural phenomena—such as the expansion of wildfire fronts or the melting of ice—can be modeled mathematically using an algorithm called threshold dynamics, introduced by Bence, Merriman, and Osher in 1992. Similar mechanisms also appear in biology and other sciences under the name cellular automata. In collaboration with Smith students Abi Bowering ’24 and Almanzo Gao '25, as well as coauthors Elisa Negrini (UCLA) and Wei Zhu (Georgia Tech), we implemented a novel neural network ensemble architecture to learn the threshold dynamics of an evolving system from a limited number of photograms capturing its progression. We tested our model on both synthetic and real data, and found that it can accurately predict the future evolution of a moving front.