A comparative study of two stochastic mode reduction methods (Q2492255)

The paper provides a comparison of two methods for the reduction of the dimension of systems of nonlinear ordinary differential equations which exhibit two time-scale behaviour. One method, called asymptotic mode reduction strategy and proposed by Majda, Timofeyev and Vanden-Eijnden, consists in replacing the nonlinear self-interaction by a singularly perturbed stochastic term and using Kurtz's asymptotic theorem to arrive at an approximating equation describing the evolution of slow (resolved) variables. The other method is based on the Mori-Zwanzig projection formalism and the short-memory approximation arising from it. The author discusses the role of the specific projection operators used in both methods and notes that both methods lead to similar predictions provided the fast dynamics can be properly represented by a stochastic operator. Finally, the two methods are applied to some models and the earlier discussion is illustrated by numerical results.