The AS is primarily designed to become an ICOS Level 1 station (required measurements are listed in the first six rows in Table 1). To fully exploit the infrastructure potential, selected air quality parameters were added to the list of measurement parameters (Table 1). Further long-term or campaign measurements may be added in future (e.g. halocarbons, isotopes in non-CO2 GHGs). Air quality and GHG monitoring is accompanied by meteorological measurements in order to support the interpretation of vertical and temporal concentration variabilities.

Air samples for GHG determination are transported in permanently flushed lines from different tower heights (Table 1) to analyzers placed in an air-conditioned ground based container. On the contrary, ozone and mercury analyzers are placed directly on the tower in an air-conditioned container installed at 230 m height directly on the mast body and in air-conditioned racks placed on platforms at 10, 50 and 125 m heights. This is because a long inlet system and analyzer on the ground may lead to large sample losses for these species (Galbally and Schultz, 2013). Aerosol instruments are placed in the ground-based container. Meteorological sensors are installed on 3 m long arms directed approximately south west (azimuth 215∘) which is the windward side of the prevailing air flow. To cover the eastern flow an installation of additional arms directed north east (azimuth 35∘) is planned.

Two pilot studies were conducted on two selected datasets. Organic and elemental carbon (OC and EC, respectively), as important climate change drivers (Penner et al., 1998), were measured from August to December 2013 by a field semi-online OC/EC analyser (Sunset Laboratory Inc., USA) using a PM2.5 cyclone inlet and a carbon parallel-plate diffusion denuder. Samples were taken at four-hour intervals, including the thermal-optical analysis, which lasts about 15 min. This is a much higher time resolution in comparison with commonly used 24 h sampling. The analysis was performed using the shortened EUSAAR2 protocol (Cavalli et al., 2010; Vodička et al., 2013). Data quality is ensured using an internal standard (CH4) during each analysis and by regular external calibration by sucrose solution.

EC data was used to investigate the occurrence of censored observations at the site, i.e. measurements under instrumental limits of detection and quantification (0.2 and 0.5 µgC m-3, respectively). Censored data may occur often at background sites and they have to be handled carefully in order not to negatively influence monitoring results (Helsel, 2006). Since two fixed detection and quantification limits exist, it is necessary to work with type I doubly left-censored samples (e.g. Fusek and Michálek, 2013, 2014a). For statistical modeling of EC concentrations, the method of maximum likelihood was used. The Weibull distribution was selected as a model distribution because of its flexibility. Moreover, using the asymptotic properties of maximum likelihood estimates, asymptotic tests with nuisance parameters (Fusek and Michálek, 2014b) were applied for comparison of two censored samples.

One minute averages (mean of 4 values, reading every 15 s) of temperature measured by the Vaisala HMP155 instrument at 10, 50, 125 and 240 m in the period 11 June–10 September 2014 (summer season) were investigated. Only those data were selected when records for all 4 measurement levels were available (92 % of the whole period length). Vertical temperature gradients – (ΔT/Δz) were then calculated from the finite differences for the layers 10–50, 50–125 and 125–240 m, and the data classified according to the gradient in the layer 10–50 m. For convenient plotting of the shapes of vertical profiles including uncertainties, differences of the temperature at 50, 125 and 240 m from the temperature at 10 m were also calculated.

EC and OC concentrations are slightly higher in autumn and winter 2013 (average of 0.76 and 0.70 µgC m-3 for EC, respectively and 2.93 and 2.86 µgC m-3 for OC, respectively) compared to summer 2013 (average of 0.39 µgC m-3 for EC and 2.33 µgC m-3 for OC). The results are similar to data obtained at the rural background site in Melpitz, Germany (Spindler et al., 2013) but lower in comparison with data from Diabla Góra, Poland (Rogula-Kozlowska et al., 2014). The time series of OC, EC, and EC/TC (where TC (total carbon) is the sum of EC and OC) are depicted in Fig. 1. The high variability of OC and EC concentrations is caused mainly by large changes in atmospheric mixing both due to variable boundary layer height and wind speed. A higher EC/TC ratio in autumn and winter usually indicates a higher proportion of combustion related sources (coal combustion and especially traffic, Lonati et al., 2005) and a lower proportion of secondary organic or primary biogenic aerosols among the EC and OC sources. Wood combustion emissions exhibit usually low EC/TC ratios (Hays et al., 2005).

The statistical methods based on censored samples were used for the analysis of EC measurements which contain a significant number of censored observations. A good example is the week of 21–27 October 2013, which contained approximately 25 % of concentration data lower than or equal to 0.5 µgC m-3 (limit of quantification). The suitability of doubly left-censored Weibull distribution for modelling and comparison of EC concentrations was verified using Pearson's χ2 goodness-of-fit test. The null hypothesis was not rejected at a significance level of 0.05 with a p value of 0.5445. The investigation of daily, weekly and seasonal variability of EC concentrations is planned in future.

Medians of the vertical temperature gradient in three layers (Fig. 2) reveal a typical summer daily course (Stull, 1988) although the plot is created from the whole summer dataset covering all weather conditions. Surface inversions dominate during nighttime, whereas the daytime is characterized by a convective boundary layer (CBL) with unstable stratification in the lower layer (10–50 m) and near-neutral stratification above. The lower layer has the largest diurnal amplitude of the temperature gradient, which crosses the adiabatic value early in the morning (preceded by the inversion destruction in this layer), reaches its maximum in the late morning, and again crosses the adiabatic value in the late afternoon.

An example of the vertical temperature profile is given in Fig. 3 for the class of significant surface inversions (between -0.05 and -0.015 K m-1) in the lowest layer. While the variability at 50 m is limited by the class definition, it increases with increasing height, reflecting different inversion depths and strengths. Relatively coarse vertical resolution and larger variability of the 1 min averages must also be kept in mind when analyzing the plot. In general, the medians in Fig. 3 suggest that in summer the tower top at 250 m is frequently above the nocturnal surface inversion, thus being decoupled from local influences.