Importance of fossil fuel emission uncertainties over Europe for CO₂ modeling: model intercomparison

Inverse modeling techniques used to quantify surface carbon fluxes commonly assume that the uncertainty of fossil fuel CO(2) (FFCO(2)) emissions is negligible and that intra-annual variations can be neglected. To investigate these assumptions, we analyzed the differences between four fossil fuel emission inventories with spatial and temporal differences over Europe and their impact on the model simulated CO(2) concentration. Large temporal flux variations characterize the hourly fields (similar to 40% and similar to 80% for the seasonal and diurnal cycles, peak-to-peak) and annual country totals differ by 10% on average and up to 40% for some countries (i.e., the Netherlands). These emissions have been prescribed to seven different transport models, resulting in 28 different FFCO(2) concentrations fields. The modeled FFCO(2) concentration time series at surface sites using time-varying emissions show larger seasonal cycles (+2 ppm at the Hungarian tall tower (HUN)) and smaller diurnal cycles in summer (-1 ppm at HUN) than when using constant emissions. The concentration range spanned by all simulations varies between stations, and is generally larger in winter (up to similar to 10 ppm peak-to-peak at HUN) than in summer (similar to 5 ppm). The contribution of transport model differences to the simulated concentration std-dev is 2-3 times larger than the contribution of emission differences only, at typical European sites used in global inversions. These contributions to the hourly (monthly) std-dev's amount to similar to 1.2 (0.8) ppm and similar to 0.4 (0.3) ppm for transport and emissions, respectively. First comparisons of the modeled concentrations with (14)C-based fossil fuel CO(2) observations show that the large transport differences still hamper a quantitative evaluation/validation of the emission inventories. Changes in the estimated monthly biosphere flux (Fbio) over Europe, using two inverse modeling approaches, are relatively small (less that 5 %) while changes in annual Fbio (up to similar to 0.15% GtC yr(-1)) are only slightly smaller than the differences in annual emission totals and around 30% of the mean European ecosystem carbon sink. These results point to an urgent need to improve not only the transport models but also the assumed spatial and temporal distribution of fossil fuel emission inventories