Soot formation in 'Delft Flame III', a pilot stabilized turbulent diffusion flame burning natural gas/air, is investigated using ANSYS FLUENT by considering two different approaches for soot inception. In the first approach soot inception is based on the formation rate of acetylene, while the second approach considers the formation rate of two and three-ringed aromatics to describe the soot inception (Hall et al., 1997). Transport equations are solved for soot mass fraction and radical nuclei concentration to describe inception, coagulation, surface growth, and oxidation processes. The turbulent-chemistry interactions and soot precursors are described by the steady laminar flamelet model (SLFM). Two chemical mechanisms GRI 3.0 (Gregory et al.) and POLIMI (Ranzi et al., 2012) are used to represent the effect of species concentration on soot formation. The radiative properties of the medium are included based on the non-gray modeling approach by considering four factious gases; the weighted sum of gray gas (WSGGM) approach is used to model the absorption coefficient. The effect of soot on radiative transfer is modeled in terms of effective absorption coefficient of the medium. A beta probability density function (beta-PDF) in terms of normalized temperature is used to describe the effect of turbulence on soot formation. The results clearly elucidate the strong effect of radiation and species concentration on soot volume fraction predictions. Due to increase in radiative heat loss with soot, flame temperature decreases slightly. The inclusion of ethylene has less synergic effect than that of both benzene and ethylene. Both cases have less impact on the nucleation of soot. The increase in soot volume fraction with soot-turbulence interaction is in consistence with the DNS predictions. (C) 2015 Elsevier Ltd. All rights reserved.