A novel multiscale coupling methodology based on a domain overlapping approach has been developed to couple a computational fluid dynamics code with a best-estimate thermal hydraulic code. The methodology has been implemented in the coupling infrastructure code Janus, developed at the University of Michigan, providing methods for the online data transfer between the commercial computational fluid dynamics code STAR-CCM+ and the US NRC best-estimate thermal hydraulic system code TRACE. Coupling between these two software packages is motivated by the desire to extend the range of applicability of TRACE to scenarios in which local momentum and energy transfer are important, such as three-dimensional mixing. These types of flows are relevant, for example, in the simulation of passive safety systems including large containment pools, or for flow mixing in the reactor pressure vessel downcomer of current light water reactors and integral small modular reactors. The intrafluid shear forces neglected by TRACE equations of motion are readily calculated from computational fluid dynamics solutions. Consequently, the coupling methods used in this study are built around correcting TRACE solutions with data from a corresponding STAR-CCM+ solution. Two coupling strategies are discussed in the paper: one based on a novel domain overlapping approach specifically designed for transient operation, and a second based on the well-known domain decomposition approach. In the present paper, we discuss the application of the two coupling methods to the simulation of open and closed loops in both steady state and transient operation. The objective of this study is to examine the performance of each coupling method in terms of convergence, consistency, and numerical stability. As expected, the results produced by the two methods were found to be identical once numerical convergence is achieved and consistent with the standalone STAR-CCM+ solution in both steady state and transient cases. However, the domain overlapping method was found to achieve convergence at larger integration time steps than the domain decomposition approach and exhibited superior convergence and numerical stability characteristics in both steady state and transient scenarios.