With the adoption of the Worldwide harmonized Light Vehicles Test Procedure (WLTP) and the Real Driving Emissions (RDE) regulations for testing and monitoring the vehicle pollutant emissions, as well as CO2 and fuel consumption, the gap between real world and type approval performances is expected to decrease to a large extent. With respect to CO2, however, WLTP is not expected to fully eliminate the reported 40% discrepancy between real world and type approval values. This is mainly attributed to the fact that laboratory tests take place under average controlled conditions that do not fully replicate the environmental and traffic conditions experienced over daily driving across Europe. In addition, any uncertainties of a pre-defined test protocol and the vehicle operation can be optimized to lower the CO2 emissions of the type approval test. Such issues can be minimized in principle with the adoption of a real-world test for fuel consumption. However, repeatability and an accuracy of a few gCO2/km is difficult to achieve due to the actual drag, the road surface effect on driving resistance, the road slope, the battery and auxiliaries use etc., which come naturally with on-road tests. Since a reference value for CO2 emissions should not depend on the testing circumstances, modelling can be deployed in order to introduce the necessary correction of measurement variations in a harmonized manner, based on individual vehicle simulation models that can produce smoothed results under the same driving conditions, ambient temperature, alternator operation etc. In this paper, such a simulation approach for the correction of RDE compliant measurements is introduced and demonstrated with a few real cases. The approach incorporates the encountered real-world effects for the prediction of a reference CO2 value. It is shown that with an expansion of the approach to cover variations for existing engine types, powertrains, vehicle types etc., a baseline accuracy for real world simulations can be established.