Millions of people are exposed to enhanced levels of nitrogen dioxide in urbanized areas, leading to severe health effects. Moreover, nitrogen oxides contribute to the formation of ozone and particulate matter, and as such have wider health related impacts. A substantial reduction of nitrogen oxides may offer considerable health benefits for the human society. As a first step, this requires a detailed understanding of source sector contributions to nitrogen oxide levels. Whereas many regions have information on the local (traffic) contributions, the source contributions to the rural and urban background levels are commonly not available. In this study we compared and evaluated the results of two source attribution techniques to quantify the contribution of 5 source sectors to background nitrogen oxide levels across Germany. The results of a labelling technique were compared to brute force simulations with variable emission reduction percentages. The labelled NO2 source contributions of the main sectors averaged for all urban background stations are road transport (45 ± 5%), non-road transport (24 ± 6%), energy & industry (20 ± 3%), households (10 ± 6%), and the remaining source sectors (1 ± 1%). For the brute force technique, the explained mass differs from the unperturbed baseline concentration after scaling the impact of each sensitivity simulation to 100%. The attributed concentration of NO2 is lower in urban background areas (−3 ± 5%) and larger in the rural background (4 ± 6%) than that of the labelling. Largest deviations up to −15% are calculated for the major cities along the Rhine and Main. The annual average overestimation for NO is about 53 ± 24% for urban and 40 ± 26% for rural background sites based on a 20% reduction of emissions. On shorter time scales the differences are larger. These deviations are caused by (the lack of) regime changes in the titration of ozone, most notably present at ozone-limiting conditions during nocturnal winter periods. As a consequence, the differences between the methodologies are larger for smaller emission reduction percentages applied in the brute force technique. Similarly, for small-sized emission source sectors larger deviations were found compared to large-sized sector categories. Hence, applying the brute force technique for the source attribution for a single sector should be avoided as there is no way to verify for consistency and quantify the error for the sector and total explained contribution. We recommend applying the labelling approach to estimate sector contributions in forthcoming studies for nitrogen oxides.
