Gas turbine pollutant emissions (especially NOx and CO) are limited to 25 ppmvd by the European legislation for a natural gas functioning. To meet this objective and the future legislations, it is crucial to have access to numerical tools that accurately predict NOx and CO formation when operating gas turbine (fuel flexibility, tuning of the fuel distribution between burners...). In this context EDF R&D has developed a 3D turbulent gas combustion model for several years. Nevertheless, the introduction of complex chemical kinetics in 3D turbulent combustion code is still too CPU time consuming to be performed for an industrial use. Thus, following previous works initiated by M.Falcitelli and L.Tognotti in 2002, a coupling between 3D CFD computations, using simple chemistry and a 0D Chemical Reactor Network (CRN), which includes a detailed chemistry mechanism, have been developed. The 3D simulations are used to provide information about the mixing, the aerodynamics and the turbulence effects. Using these, several splitting criteria are defined to build the CRN. The present study focuses on the effect of turbulent fluxes through reactor boundaries to correctly represent the flow field. This approach has been validated on a 2D turbulent academic diffusion flame (a lifted Masri-Bilger-Dibble methane/air flame). Finally, a simulation of a 3D tubular combustor of an industrial gas turbine combustion chamber has been performed.