Computational Fluid Dynamics (CFD) is currently viewed as one of the major tools for investigating aerodynamics of road vehicles. In spite of its well-documented limitations in predicting flows involving complex geometries and flow separations, the RANS (Reynolds Averaged Navier-Stokes) approach is still a widely used turbulence modeling methodology in motorsports/automotive industries due to its cost-effectiveness and fast turnaround time. In the past, automotive research and development efforts showed an intensive focus on the external vehicle aerodynamics for improved parasitic drag efficiency. However, as the opportunities of attaining aerodynamic advantage by pure manipulation vehicle?s external body-shape diminish, the industry turns to other areas of the vehicle for further aerodynamic efficiency improvement. One such area is the underhood airflow management, which is challenging for both experimental and analytical approaches due to the complexity and compactness of modern vehicle underhood compartments. Thereby, 3D CFD analyses appear to be a cost-effective alternative to investigate the underhood airflow characteristics during a vehicle's early design stage. Subsequently, one aspect of this thesis involves computational studies using the RANS approach to investigate the role of underhood airflow features on the radiator performance and cooling drag. Additionally, analysis of the impact of the front grille opening size and underhood passive aerodynamic devices on the cooling drag and radiator performance are presented. All of these simulations are based on full vehicle CFD simulations carried out using a detailed realistic model of Hyundai Veloster. It is demonstrated that by properly managing the cooling airflow, simultaneous improvement of radiator performance and total vehicle drag reduction can be achieved.However, existing literature suggests that, for automotive flows, RANS turbulence models often fail to capture the detailed flow features or even the integral aerodynamic quantities, and subsequently sometimes are used to assess the magnitude and direction of a trend. Moreover, even for such purposes, notable disagreements often exist between results predicted by different RANS models. Thanks to fast advances in computer technology, increasing popularity has been seen in the use of the hybrid LES/RANS (Large Eddy Simulation/Reynolds Averaged Navier-Stokes) approaches which have demonstrated the high potential of being more accurate and informative than the RANS approaches. Whilst evaluations of RANS and hybrid LES/RANS models, known as the DES (Detached Eddy Simulation), on various applications are abundant in literature, such evaluations on full-car models are rare. To further the investigations on-road vehicle aerodynamic simulation approaches, the prediction veracity of four RANS models which are widely used in industry, i.e., the realizable k - 𝜖, AKN k - 𝜖, SST k - 𝜔, and V2F models, are evaluated for a full-scale passenger vehicle with two different front-end configurations. The RANS simulated flow-fields are then compared against results from two hybrid LES/RANS approaches to highlight the predictive differences between different CFD simulation methodologies. It is found that the DES approach is superior in predicting the flow-field, but is not guaranteed to predict better correlated integral quantities. Consequently, this study explores the possibility of improving the prediction veracity of the SST k - 𝜔 model, the most promising variant of the RANS approach, by investigating the influence of a few selected model closure coefficients on the CFD predictions. This involves identifying the effect of each individual model parameter on the prediction first, and then formulating the optimized combination of the model closure coefficient values that yields the best correlation with the experiment. This procedure is applied to four different test objects: NACA 4412 airfoil at 12 degree angle of attack, the 25 and 35 degree slant angle Ahmed body, and a full-scale passenger road vehicle. Although some closure coefficients do not influence the CFD results much, the predictions are very sensitive to the choice of certain model constants irrespective of the test object geometry. The study also shows that it is possible to formulate a combination of closure model coefficients that can produce very well correlated CFD predictions of the overall flow-field.