A class of interacting superprocesses on $\mathbb{R}$, called
superprocesses with dependent spatial motion (SDSMs), were
introduced and studied in Wang [32] and Dawson et al. [9].
In the present paper, we extend this model to allow particles
moving in a bounded domain in $\mathbb{R}^{d}$ with killing
boundary. We show that under a proper re-scaling, a class
of discrete SPDEs for the empirical measure-valued processes
generated by branching particle systems subject to the same
white noise converge in $L^{2}(\Omega, \mathcal{F}, \mathbb{P})$
to the SPDE for an SDSM on a bounded domain and the corresponding
martingale problem for the SDSMs on a bounded domain is well-posed.