State-space reconstruction and spatio-temporal prediction of lattice dynamical systems

This paper addresses the problems of state space reconstruction and spatio-temporal prediction for lattice dynamical systems. It is shown that the state space of any finite lattice dynamical system can be embedded into a reconstruction space for almost every, in the sense of prevalence, smooth measurement mapping as long as the dimension of the reconstruction space is larger than twice the size of the lattice. Based on this result, an input-output spatio-temporal dynamical relation for each site within the lattice is derived and used for spatio-temporal prediction of the system. In the case of infinite lattice dynamical systems, an approach based on constructing local lattice dynamical systems is proposed. It is shown that the finite dimensional results can be directly applied to the local modelling and spatio-temporal prediction for infinite lattice dynamical systems. Two numerical examples are provided to demonstrate the proposed theory and approach.