Statistical methods for accounting and understanding ozone trends derived from observations

Emissions of ozone precursors have been regulated in Europe since around 1990 with air quality control measures, which resulted in reductions of nitrogen oxides and volatile organic compounds concentrations. In order to understand how these measures have affected tropospheric ozone, it is important to investigate its long-term temporal evolution in different types of environments and various geographic regions. Uncertainties in ozone long-term trends are associated to variations originating from meteorological influence on ozone. Also, ozone temporal evolution can vary significantly among different regions and types of environment. In this PhD thesis we used sophisticated statistical tools, and developed robust statistical approaches to study long-term trends of tropospheric ozone that reflect emissions reductions. First, we focus on a meteorological adjustment of ozone observations in order to derive long-term trends with lower uncertainties compared to common practices. In addition, a classification scheme for stations in Europe is needed to interpret response of ozone in site groups with similar spatio-temporal characteristics. A detailed long-term trend analysis was performed based on decomposition of the mean ozone observations in the time domain. The different time-dependent variations of ozone were extracted, namely the long-term trend, seasonal and short-term variability. This allows subtraction of the meteorologically driven seasonal variation from the observations and estimation of long-term trends on de-seasonalized concentrations. In addition, ozone peak concentrations were investigated using a localized regression approach, which corrects for meteorological influence. The meteorological adjustment of the mean and peak ozone allows estimation of long-term trends with lower uncertainty. A site grouping in Europe was developed using the long-term and seasonal variations of ozone. The implemented clustering approach based on the long-term variation resulted in a site type classification while a geographical classification was achieved on the seasonal variation. We observed that, despite the implementation of regulations, mean ozone has been increasing in most of the sites until mid-2000s, although, afterwards, a decline or a leveling off was observed. The point when the trend changes from increasing to decreasing depends on the site type; the closer a site locates to emission sources the later the change occurred. Also, it was concluded that with