In this paper we consider an environment-dependent spatial model. Actually, this random model is closely related to some of the stochastic interacting system in Liggett  (1999). We shall show that rescaled processes of the models converge to a Dawson-Watanabe superprocess with suitable parameters. Our formulation of measure-valued branching Markov processes  is greatly due to a martingale problem formalism. The first step toward a transformation of spatial model into a superprocess is based upon construction of related empirical measures.