This study presents the first application of a multidimensional conditional moment closure (CMC) combustion with a two-equation soot model to study the evolution of in-cylinder soot in diesel engines. Five high-fidelity reference cases from an optically accessible heavy-duty diesel engine were used to validate the model predictions; these include two high-temperature cases and two low-temperature cases with two different injection timings and one low-temperature case with a split injection strategy. Simulation results have been compared with experimental data by means of non-reactive fuel vapour distribution, apparent heat release rates (AHRR), temporal evolutions of quantitative in-cylinder soot mass and natural luminosity. The model was capable to reproduce semi-quantitative trends of soot mass for all five cases. In particular for low-temperature conditions the evolution of soot formation and oxidation is very well captured. A line-of-sight transformation of the computed 3D soot is proposed to ensure a consistent comparison with the measured 2D soot natural luminosity images. With this methodology additional effects related to radiation adsorption were observed. Evolutions of line-of-sight soot natural luminosity were qualitatively very well reproduced. Overall the findings suggest that the presented two-equation soot model implemented in the CMC framework is a highly promising candidate for predicting time evolutions of in-cylinder soot for diesel engines operating in different combustion modes. (C) 2013 Elsevier Ltd. All rights reserved.