Parallel genetic Algorithms for calibrating Cellular Automata Models: Application to lava flows

Cellular Automata are highly nonlinear dynamical systems which are suitable for simulating natural phenomena whose behaviour may be specified in terms of local interactions. The Cellular Automata model SCIARA, developed for the simulation of lava flows, demonstrated to be able to reproduce the behaviour of Etnean events. However, in order to apply the model for the prediction of future scenarios, a thorough calibrating phase is required. This work presents the application of Genetic Algorithms, general-purpose search algorithms inspired to natural selection and genetics, for the parameters optimisation of the model SCIARA. Difficulties due to the elevated computational time suggested the adoption a Master-Slave Parallel Genetic Algorithm for the calibration of the model with respect to the 2001 Mt. Etna eruption. Results demonstrated the usefulness of the approach, both in terms of computing time and quality of performed simulations.