The paper presents a numerical study where a hybrid CFD-Chemical Reactor Network (CRN) approach is used to predict pollutant emissions in a tubular combustor for aero-engine applications. A fully-automated clustering of the simulated flow field with the generation of a reactor network representative of the main flow features is exploited. Similar cells are detected and grouped using a two step approach, the first one based only on aerodynamic criteria for turbulent flows followed by a chemical refinement based on mixture fraction. A formulation for turbulent diffusion fluxes is introduced in the reactor code to model species and energy exchanges between reactors. Three different operating conditions are studied for which measured NOx and CO are available. Results highlight the importance of including turbulent diffusion in the network solution. The accurate prediction of pollutant emissions at different load points confirms that CFD-CRN is a valid and flexible approach for preliminary assessment of aero-engine combustor emissions in the design phase.