Dynamic ozone evaluation using a model intercomparison study for Germany.

Air quality remains a key topic for human wellbeing worldwide. Ozone (O₃) is still one of the most toxic and ecologically detrimental air pollutants in Europe and supplies a crucial impact factor for the air quality planning as millions of people are exposed to O₃ levels above the WHO guidelines. The chemical reaction processes leading to the formation of ozone are well documented in literature for long: O₃ is not emitted but rather formed through complex chemical reactions from precursor emissions such as nitrogen oxides or biogenic volatile organic compounds. Processes influencing ozone variability are highly sensitive to several meteorological parameters such as temperature, moisture or solar radiation. These processes can impact the emission rate of ozone precursors, the chemical production and destruction as well as the rate of ozone loss through dry deposition. The ambient air pollution for ozone is often assessed and forecasted using chemical transport models (CTMs). These CTMs aim to reproduce observed ozone variability as good as possible by comprehensively accounting the abovementioned processes.

To evaluate CTMs and identify directions for improvement multi-model intercomparison studies have proven very useful on the past. Often the ensemble mean or median shows a better model skill than the ensemble members. The quality of model (ensemble) results is normally assessed by calculating a number of statistical indicators in a paired comparison to measured timeseries. In addition, to assess the model quality and uncertainty one can use a dynamic evaluation. The dynamic evaluation relates the model error to input data such as the meteorology. The degree to which changes in ozone levels caused by varying meteorological conditions are then evaluated. This allows to assess whether numerical models can capture the chemical response to temperature, humidity or another meteorological parameter.

In this study we also make use of such an ensemble assessment to evaluate the multi-model performance and the skill for each ensemble member. We conducted air pollution simulations for four models (LOTOS-EUROS, REM-CALGRID, COSMO-MUSCAT and WRF-CHEM) across Germany for January 1^st to December 31^st, 2019. The models show a very consistent picture in the ranking of the model skill. Main differences between the four ensemble members we found for ozone episodes, the timing of daytime maxima or even the representation of the nighttime concentration. We further enhanced the understanding of the modelled ozone response to temperature and humidity and provided an in-depth understanding for differences occurring in the ozone production rates for all participating models separated by season and region. First results indicate main differences in the ozone productivity especially for warm and humid conditions during the ozone season. The COSMO-MUSCAT and REM-CALGRID models show largest variability for ozone production rates with respect to temperature and humidity. The overall best performance can be seen for LOTOS-EUROS and WRF-CHEM.