Linking fire to climate in the Florida Everglades using cellular automata, evolutionary algorithms, and empirical likelihood

Projecting future climatic impacts on fire regimes and community dynamics is challenging due, in part, to complex feedbacks between vegetation and fire. We address this problem by linking fire occurrence and spread in the Everglades landscape to hydrology and climate in a simulation model that couples fire probability with vegetation dynamics. The simulation model requires parameters that govern interactions between community transitions and fire spread. Some parameters are estimable from independent empirical data, while others are not readily observable but can be identified from model behavior in landscape simulations. This presents two challenges: First, the simulation model generates data on ecosystem responses at multiple scales and identification of parameters from model dynamics is likely to be sensitive to the choice of metric(s) to measure model fit. Second, the link between model parameters and the predicted state of the Everglades landscape is stochastic, and this stochasticity is generated internally in the simulation model. We do not know the underlying form of the stochasticity and we can only generate distributions of results from multiple individual realizations. We address these challenges to parameterize a cellular-automata model of coupled fire-vegetation dynamics using an evolutionary algorithm that employs a fitness function based on empirical likelihood. Our parameterized model recreates historic fire patterns in the Everglades landscape in response to climate and hydrology, and our approach to parameter estimation from dynamic simulations should be generalizable to other complex systems.