This paper aims to develop appropriate boundary conditions to be used in the numerical analysis (CFD) of smoke transport within a semi-enclosed space resembling a typical road tunnel. The study considers the following stages: (a) Model set up: the geometry of an existing tunnel, with available experimental data to validate numerical results, is recreated in 3D CAD model with physical properties and known boundary conditions; (b) Mesh verification: different aspects and techniques of mesh generation, such as local and global refinement and first/second-order discretization schemes were considered for the mesh sensitivity analysis. The aim of this stage is to demonstrate that numerical results are mesh-independent, i.e. use as large as possible element sizes to reduce the computational cost, yet refine grid in crucial locations to record important features of transport and thermal phenomena; (c) Model validation: This stage considers the inaccuracies associated to turbulence models, intensity of turbulence at inlet boundaries, buoyancy, turbulent diffusion and convection on the numerical results of the study. Different combinations of numerical aspects of the study were tested. Results of these combinations were compared with experimental data and the most accurate combination of variables was taken as benchmark. Obtained numerical model serves as a guideline tool capable of reproducing propagation of smoke in semi-enclosed spaces within minimal uncertainty than can be used in further studies as a predicting tool for engineering purposes.