This dissertation presents novel numerical experiments and model developments to address challenges in coupling Numerical Weather Prediction models (NWP) with Large Eddy Simulation (LES) and diagnostic wind solvers. The benefit of coupling microscale models with large scale dynamics is to improve the characterization of the atmospheric boundary layer (ABL) in complex and realistic scenarios where canonical representations of the ABL are not valid, for instance when unsteadiness, advection, baroclinicity, subsidence and complex terrain phenomena are not negligible.We explore outstanding challenges in achieving NWP-LES coupling, such as the gray zone of turbulence in the convective boundary layer and the sensitivity of fine resolution LES driven by realistic surface fluxes, topography and gray zone tendencies. We introduce and validate an ODE-based 1D scheme to account for large scale dynamics in a fast-running and building-aware 3D urban diagnostic wind solver (QUIC). New insights, challenges and opportunities for multiscale modeling and its applications are discussed. We demonstrate that coupling large scale dynamics with turbulence-resolving and diagnostic simulations is a promising tool for atmospheric problems in complex scenarios.