Statistical modelling of combined ozone-temperature events in Europe

Air pollution as well as high air temperature both pose a large risk to human health in Europe. High temperature levels are associated with an exceptionally high mortality rate, only representing the extreme end of a wide range of possible health effects. Tropospheric ozone, a secondary air pollutant, is primarily built by photochemical reactions under solar radiation with the involvement of precursor gases including nitrogen oxides, carbon monoxide, methane, and non-methane volatile organic compounds. Due to the specific characteristics of ozone formation, high levels of ozone and temperature often coincide, posing an even intensified threat to human health.The current scientific work focuses on the co-occurrence of these two health stressors as well as their underlying meteorological conditions. A subset of European ozone (AirBase_v8, EEA) and temperature (ECA&D) stations is selected for analysis based on individual station locations and data coverage. Taking into account different settings of air substances concentrations (urban, outer conurbation area, rural regions), these stations are classified and grouped by station type and area type resulting in five distinct station classes: urban traffic, urban background, suburban background, rural background and rural industrial.Maximum daily 8-hour average ozone values (MDA8O3, EEA), observed daily maximum air temperatures (TX, ECA&D) and meteorological variables (from ERA5, ECMWF) form the data basis for model building. Current thresholds and extreme definitions e.g. based on WHO air quality guidelines or high percentiles (75th and 90th) are examined and discussed to describe elevated levels of these variables and to finally define combined ozone-temperature events.Possible regional patterns as well as disparities between urban and rural areas regarding the specific settings for ozone formation as well as varying meteorological mechanisms for the occurrence of combined ozone-temperature events are closely examined. The methodological focus is primary on statistical modelling, the application and comparison of varying multivariate statistical approaches and different machine learning methods, e.g. various regression analyses using shrinkage methods or random forests. Consequently, statistical models are generated to analyse the influence of meteorological conditions on the occurrence of combined ground-level ozone and temperature events along with the identification of primary key factors (e.g. ozone persistence or larger-scale air temperature and wind conditions) at each specific location.Furthermore, frequency and intensity changes of combined ozone-temperate events in the scope of global warming are assessed. Thus, projections of these co-occurring events under the constraints of ongoing climate change until the end of the 21st century are analysed by integrating projections of general circulation models into the statistical modelling process.