Accurate prediction of emissions, such as NOx or particulates, from combustion systems is a key tool to designing future engines. Accuracy of these simulations is limited by availability of detailed experimental data from inside realistic combustors and simulation techniques often present a compromise between fidelity and computational cost. Work at Loughborough University and other partners is aiming to enable detailed measurements of species, resolved in space and time, using gas spectroscopy techniques. This PhD will work alongside that project to simulate the combustor experimental set ups using CFD. The PhD will interact with the experimental work in two ways: 1) To understand and improve the gas spectroscopy method, Large Eddy Simulations (LES) of the experimental cases will be used to carry out numerical experiments. In these the expected outcome from spectroscopy measurements of known LES generated species concentration contours is calculated. This can then be used to understand errors in the data reconstruction process by comparing the processed 'synthetic spectroscopy' data with the original LES. Through this process the impact of factors such as reconstruction algorithm, spatial resolution and phenomena such as beam steering can be investigated. 2) Data from the experiments, and from existing emissions measurements techniques, can be used to validate simulations. This will be used in an attempt to identify sources of error and where limited computing resources can be best spent. For example if the mixing field and temperature field is shown to be simulated correctly then discrepancies in emissions can be attributed to the chemistry modelling. As part of this results from simulations using different methods (including some from related projects) will be compared. Baseline simulations using industry standard tabulated chemistry methods will be performed and compared with experimental data. It is intended that these will be compared with methods that allow for more detailed finite-rate chemistry but with lower spatial resolution such as reactor networks or methods such as Conditional Moment Closure.