A B S T R A C T Quantifying carbon dioxide emissions from fossil fuel burning (FFCO 2 ) is a crucial task to assess continental carbon fluxes and to track anthropogenic emissions changes in the future. In the present study, we investigate potentials and challenges when combining observational data with simulations using high-resolution atmospheric transport and emission modelling. These challenges concern, for example, erroneous vertical mixing or uncertainties in the disaggregation of national total emissions to higher spatial and temporal resolution. In our study, the hourly regional fossil fuel CO 2 offset (DFFCO 2 ) is simulated by transporting emissions from a 5 min)5 min emission model (IER2005) that provides FFCO 2 emissions from different emission categories. Our Lagrangian particle dispersion model (STILT) is driven by 25 km)25 km meteorological data from the European Center for Medium-Range Weather Forecast (ECMWF). We evaluate this modelling framework (STILT/ECMWF ' IER2005) for the year 2005 using hourly DFFCO 2 estimates derived from 14 C, CO and 222 Radon ( 222 Rn) observations at an urban site in south-western Germany (Heidelberg). Analysing the mean diurnal cycles of DFFCO 2 for different seasons, we find that the large seasonal and diurnal variation of emission factors used in the bottom-up emission model (spanning one order of magnitude) are adequate. Furthermore, we show that the use of 222 Rn as an independent tracer helps to overcome problems in timing as well as strength of the vertical mixing in the transport model. By applying this variability correction, the model-observation agreement is significantly improved for simulated DFFCO 2 . We found a significant overestimation of DFFCO 2 concentrations during situations where the air masses predominantly originate from densely populated areas. This is most likely caused by the spatial disaggregation methodology for the residential emissions, which to some extent relies on a constant per capita-based distribution. In the case of domestic heating emissions, this does not appear to be sufficient.

No abstract