The present paper is concerned with the enhancement of a hybrid LES-URANS method, which enables the simulation of turbulent flows with acceptable computational effort due to a decrease of the resolution requirements for the near-wall region. The unsteady Reynolds-averaged Navier-Stokes (URANS) mode within the hybrid approach is taken into account by an explicit algebraic Reynolds stress model (EARSM), which guarantees an appropriate representation of the anisotropic near-wall turbulence. In earlier stages of the hybrid method an enhancement of the production and the diffusion term was achieved based on the EARSM. In the present study the turbulent dissipation rate is improved by a proper formulation based on the splitting approach of Jakirlić and Jovanović (2010), where the homogeneous dissipation rate is expressed analytically by a Taylor series expansion of the homogeneous lateral Taylor microscale in the vicinity of the wall. Besides the correct asymptotic behavior for the near-wall region the formulation avoids any empirical constant. The interface location between the large-eddy simulation (LES) mode and the URANS mode is determined with the aid of the modeled turbulent kinetic energy and the distance to the wall and thus automatically switches to the correct mode. The enhanced near-wall formulation of the turbulent dissipation rate is evaluated in detail for the internal flow through two three-dimensional diffusers using different interface locations. Furthermore, the application area of this unique hybrid technique is extended towards external flows. For this purpose the flow past a SD7003 airfoil with a laminar separation bubble including transition onset is considered. An additional dynamic transition criterion is suggested which enables the determination of the laminar and the turbulent flow regime. The new interface criterion relies on the properties of the computational grid and the characteristic length scales of the resolved flow field and thus can be combined with the existing interface criterion. The performance of the hybrid method on two coarse grids is assessed based on a wall-resolved LES carried out on a fine grid. The results of the hybrid method are evaluated in terms of the influence of the mesh resolution, the interface type and the interface location and finally compared with experimental and numerical investigations of other research groups. © 2014 Elsevier Ltd.