Spatio-Temporal Variation of Ozone Concentrations and Ozone Uptake Conditions in Forests in Western Germany

The study analyzes the long-term trends (1998–2019) of concentrations of the air pollutants ozone (O3) and nitrogen oxides (NOx) as well as meteorological conditions at forest sites in German midrange mountains to evaluate changes in O3 uptake conditions for trees over time at a plot scale. O3 concentrations did not show significant trends over the course of 22 years, unlike NO2 and NO, whose concentrations decreased significantly since the end of the 1990s. Temporal analyses of meteorological parameters found increasing global radiation at all sites and decreasing precipitation, vapor pressure deficit (VPD), and wind speed at most sites (temperature did not show any trend). A principal component analysis revealed strong correlations between O3 concentrations and global radiation, VPD, and temperature. Examination of the atmospheric water balance, a key parameter for O3 uptake, identified some unusually hot and dry years (2003, 2011, 2018, and 2019). With the help of a soil water model, periods of plant water stress were detected. These periods were often in synchrony with periods of elevated daytime O3 concentrations and usually occurred in mid and late summer, but occasionally also in spring and early summer. This suggests that drought protects forests against O3 uptake and that, in humid years with moderate O3 concentrations, the O3 flux was higher than in dry years with higher O3 concentrations.

Tropospheric O 3 is formed by interactions between precursors such as nitrogen oxides (NOx) and volatile organic compounds (VOCs) in the presence of sunlight [20][21][22]. These precursors are produced naturally or emitted from anthropogenic activities such as combustion of fossil fuels used for powering vehicles, power plants or industrial complexes, as well as burning of biomass (e.g., in agriculture).
Other meteorological parameters also influence the O 3 formation process, such as high temperature, wind speed and low air humidity [23]. Given the importance of NO 2 and NO as precursors for the

Measurements
Each ZIMEN station [60] is equipped with air pollution inlet devices that are mounted on the roof of a container at 3 m above the ground. Pollutants and meteorological parameters are measured continuously (reporting interval of 1 h) with devices listed in Table 2.
Precursor substances like non methane volatile organic compounds (NMVOCs) and methane were not included into the study, because they were only available for the entire time series at the site Hortenkopf. CO was not detectable at the forest plots listed in Table 1.
Plausibility studies [61,62] and comparisons of data with other surrounding meteorological sites revealed deviations especially in precipitation from data collected at ZIMEN stations. So Intermet Data, i.e., interpolated meteorological data from all available climate stations in a raster of 1 km around the Level II Forest Plots (1 km 2 raster data), were used to characterize the meteorological parameters T, rH, P and G [63] (Table 2). G, T, VPD, and precipitation were recorded at or calculated for standard height (3 m). Wind speed was measured at 10 m height at the ZIMEN station [60] (see Table 2).

Measurements
Each ZIMEN station [60] is equipped with air pollution inlet devices that are mounted on the roof of a container at 3 m above the ground. Pollutants and meteorological parameters are measured continuously (reporting interval of 1 h) with devices listed in Table 2.
Precursor substances like non methane volatile organic compounds (NMVOCs) and methane were not included into the study, because they were only available for the entire time series at the site Hortenkopf. CO was not detectable at the forest plots listed in Table 1.
Plausibility studies [61,62] and comparisons of data with other surrounding meteorological sites revealed deviations especially in precipitation from data collected at ZIMEN stations. So Intermet Data, i.e., interpolated meteorological data from all available climate stations in a raster of 1 km around the Level II Forest Plots (1 km 2 raster data), were used to characterize the meteorological parameters T, rH, P and G [63] (Table 2). G, T, VPD, and precipitation were recorded at or calculated for standard height (3 m). Wind speed was measured at 10 m height at the ZIMEN station [60] (see Table 2).  Missing data is a ubiquitous problem in evaluating long-term experimental measurements due to equipment failures, system maintenance, power-failure, and lightning strikes [64]. In the present study, the Intermet Data are free from gaps, because they are interpolated from different surrounding stations. O 3 and NOx-Data had data gaps of less than 8% because of data losses during calibration procedures. Data gaps of less than five hours were filled by calculating running averages, while gaps of more than five hours were filled using regression equations for hourly data between the five sites for each year [62,65,66].
The atmospheric water balance (AWB; mm) [55] was calculated to compare different sites in cases of different precipitation and potential evapotranspiration conditions, characterization of similarities (by macro weather situation) or differences in weather conditions (modified by altitude and spatial distribution of the sites) as well as climatic differences in the investigated time series. AWB refers to the difference between the sum of precipitation (P) and potential evapotranspiration (PET; see Equation (1)) [52] and is calculated from daily meteorological data according to the Penman-Monteith equation [67]. To attain daily PET values, we used the output of the DO 3 SE model Version 3.1.0 [68] (www.sei.org/projects-and-tools/tools/do3se-deposition-ozone-stomatal-exchange/) which is calculated based on Penman-Monteith [67].
The AWB only defines the atmospheric aridity and does not consider the whole water supply of plants. A dry atmosphere will lead to higher transpiration demand of plants, which could be balanced to some extent by increased soil water uptake particularly during the night. Only if the plant available soil water (PAW; mm) is limited too, a genuine drought for plants is manifested and can lead to water stress. Such drought is defined in the present study as a day with negative AWB and simultaneously low PAW stock of equal or less than 50% of usable field capacity (uFC (mm)) (see Equation (2)), which is the difference of the field capacity (FC) and permanent wilting point (PWP). Drought (mm) = Sum of daily negative AWB if PAW < 50% of uFC (2) A drought affects stomatal leaf conductivity. Hence it is important to estimate the plant available soil water. According to DO 3 SE [69] and FO 3 REST [70] models, leaf conductivity would be restricted if the PAW pool declines to less than 50% of uFC. Therefore, the uFC of the appropriate soils were assessed. The uFC assessment was based on soil profile descriptions provided by Landesforsten Rheinland-Pfalz [71], which consider soil texture, soil depth, soil density and stone volume, and then used for calculation of uFC according to the method of Bodenkundliche Kartieranleitung (2005) [72] for the five studied forest plots. To evaluate the daily amount of PAW, the soil water model of FO 3 REST [70] was applied to calculate the daily change of PAW stock depending on precipitation and evaporation. The model requires detailed (in daily resolution) meteorological parameters of global radiation, the amount of precipitation, temperature, air humidity, and wind speed [69,70]. If this coincides with a negative atmospheric water balance, plants will be exposed to water stress. Periods with restricted negative AWB and smaller than 50% uFC will be accumulated, still surplus precipitation will fill up PAW over a threshold of 50% of uFC. Such defined drought periods will have a length (in days), a frequency within a year and an amount (the accumulation of missing water within a period measured in mm). The amount, duration, day of the first appearance, and the frequency of drought during a vegetation period will be used to characterize yearly drought or water stress periods of a sites. This provides a way to compare different sites and characterize the weather conditions with regard to gas exchange conditions much better as the correlation to measured meteorological parameters, because the atmospheric water conditions are extended by available soil water conditions. Furthermore, mean daytime O 3 concentrations (mean daytime O 3 is the daily mean of hourly O 3 concentration if the global radiation is higher than 50 W/m 2 ; following the AOT40 definition) were calculated. The mean daily daytime O 3 concentration of the entire period from 1998-2019 was subtracted from each daily mean O 3 concentration to define elevated daily O 3 -concentrations if they were higher than zero (Table S1).
Lastly, the synchrony or asynchrony of elevated O 3 concentrations and drought periods was examined as this enables a qualitative view on the O 3 uptake risk, i.e., to identify days with high, medium, and low potential O 3 uptake rates. Asynchronous days will have a higher O 3 uptake risk, because leaf conductivity will not be restricted by water stress, whereas synchronous days of elevated O 3 concentration and drought will lead to reduced or entire stomata closure, which protects the trees against O 3 uptake.

Statistical Methods for Spatial and Temporal Data Analysis
Statistical analyses were done with Excel, SPSS, and R. The temporal trend in air pollution and meteorological data were calculated with regression analysis of annual means against the years. Scheffé's test was used as location test after the application of an ANOVA to check whether there were significant differences between sites measuring O 3 and NO 2 parameters over 12 months (Table S2).
Moreover, to identify correlations between all parameters, which will influence O 3 concentration and leaf conductivity, a principal component analysis (PCA) was applied to identify the importance of the variables in each factor [73][74][75][76][77][78][79][80]. This method is concerned with identifying linear components within datasets and with how a particular variable might contribute to that specific component. PCA interpretation depends on understanding eigenvalues and eigenvectors. Eigenvalues are the coefficients attached to eigenvectors, which give the axes magnitude [81][82][83][84][85].  (Table 3).

Temporal, Spatial Variance and Correlations of Pollution and Meteorological Data
At all sites, the annual mean of NO 2 concentrations ranged from 4.5 µg/m 3 to 16 µg/m 3 , with Neuhäusel and Herdorf showing the highest annual mean (16 µg/m 3 and 14.2 µg/m 3 , respectively) in every year. The hourly maximum of NO 2 was less than 175 µg/m 3 at all sites. However, there were significant spatial discrepancies among the sites (Table S2).
Boxplots of NO 2 concentrations will display unproportional long whiskers of the 4th quartiles and therefore the maximum values which limits the 4th quartile are displayed as points on the secondary y-axis in Figure 3. This unusual change highlights the temporal trend over the whole time series. The unproportional long 4th quartile indicates the skewed distribution of NO 2 concentrations with dominating low hourly concentrations (smaller than 15 µg/m 3 ; 1st to 3rd quartile) in comparison to the 4th quartile which are spread over a longer concentration range. This points out that extremely Atmosphere 2020, 11, 1261 7 of 28 high NO 2 concentrations of the 4th quartile are rare events, but they should not be marked as outliers, because they are plausible measurements attributed to presumably nearby combustion events or transport processes.  Table 3. The end of the upper and lower whiskers represent the hourly maximum or minimum values of the year, respectively. The line between the first and third quartile characterizes the annual median.   series. The unproportional long 4th quartile indicates the skewed distribution of NO2 concentrations with dominating low hourly concentrations (smaller than 15 µg/m 3 ; 1st to 3rd quartile) in comparison to the 4th quartile which are spread over a longer concentration range. This points out that extremely high NO2 concentrations of the 4th quartile are rare events, but they should not be marked as outliers, because they are plausible measurements attributed to presumably nearby combustion events or transport processes.   Table 3.
The annual mean concentration of NO 2 and NO (around 2 µg/m 3 ), despite high variances ( Figure S1), presented an overall significant decreasing trend since the end of the 1990s (Table 3).
In addition, the extreme values show the same trends with exception of NO, where the lowermost values are frequently smaller or equal than 1 µg/m 3 .
None of the sites (with exception of Herdorf) showed significant increasing trends in temperature, while global radiation increased at all sites inclusive the peak values (Table 3). At all sites, peak values of precipitation and wind speed (except Leisel) decreased with exception of Neuhäusel and Leisel ( Table 3). The temporal increase of global radiation included the peak values (Table 3). Only at Hortenkopf and Herdorf an increase of VPD peak values is significant.
Very important is that mean O 3 (24 h and daytime) concentrations do not show any temporal trend. They vary around the same level for the duration of the whole time series investigated. Only at Hortenkopf an increase of peak values was assessed, whereas in Neuhäusel and Herdorf an increase of the lowermost values was significant. These two effects will determine a steady mean value of O 3 during the time series on a site.

Spatiotemporal Trends
The typical yearly annual course of O 3 and its precursors NO 2 and NO is opposite. High O 3 concentration levels were present from April to September (Figure 4a). The highest monthly mean O 3 concentration value of 45.5 ppb was observed in June at Hortenkopf while the lowest value of 16.6 ppb was found in November at Neuhäusel. The curve progression from the sites Hortenkopf, Leisel and Wascheid (the highest altitude sites) showed nearly the same O 3 concentrations from April to September, and only outside the vegetation period in January, February, and March as well as in October, November, and December did they differ. A different curve shape was observed at Neuhäusel and Herdorf with significantly smaller concentrations throughout the whole year. Table S2 shows significant differences between months and sites in two groups with higher altitudes and smaller population density (Leisel, Wascheid, Hortenkopf) on the one hand and lower altitudes with higher urban density on the other hand (Neuhäusel, Herdorf). Herdorf showed the lowest O 3 concentrations during the vegetation period of all sites, with the Scheffe' test approving that there is a significant difference in some months (April and June) during the vegetation period between sites (Table S2).
An opposite temporal course as compared to monthly mean O 3 concentrations were found for nitrogen oxides (Figure 4b,c). NO 2 and NO concentrations were highest during the wintertime. At Neuhäusel and Herdorf the highest NO and NO 2 concentrations were recorded outside the vegetation period from October to March. The lowest NO 2 and NO concentrations were observed in the hottest months of June, July, and August. The highest monthly mean NO 2 concentration of 17.4 µg/m 3 occurred in February at Neuhäusel, whereas the lowest value of 4.6 µg/m 3 was found in July at Hortenkopf (Figure 4b). Furthermore, the highest monthly mean NO concentration of 2.67 µg/m 3 occurred in December at Neuhäusel, whereas the lowest value of 1.03 µg/m 3 was found at Hortenkopf in August (Figure 4c).
Neuhäusel and Herdorf possess the highest values of NO 2 while sites at higher altitude like Leisel, Wascheid, Hortenkopf exhibit lower values during the year. The higher NO 2 concentrations at the site Neuhäusel differ significantly from the other sites (Table S2).
Neuhäusel and Herdorf showed the highest annual means of NO concentration (2.2 µg/m 3 and 2 µg/m 3 , respectively) in every year and the hourly maximum of NO concentration did not exceed 170 µg/m 3 at all sites ( Figure S1). The levels of NO 2 and NO concentrations of Neuhäusel and Herdorf were higher than those of Hortenkopf, Leisel, and Wascheid.   (Table S2).
Neuhäusel and Herdorf showed the highest annual means of NO concentration (2.2 µg/m 3 and 2 µg/m 3 , respectively) in every year and the hourly maximum of NO concentration did not exceed 170 µg/m 3 at all sites ( Figure S1). The levels of NO2 and NO concentrations of Neuhäusel and Herdorf were higher than those of Hortenkopf, Leisel, and Wascheid.
The spatial distribution of meteorological parameters, as a function of altitude, is shown in (Table S1). For example, higher altitudes experienced lower temperatures and VPD but higher precipitations and O3 concentrations. On the other hand, in low altitude sites, temperature was higher whereas precipitation and O3 were lower than at high altitude sites (Table S1). Moreover, there are significant differences between sites for meteorological parameters during the time series ( Figure S2). For example, yearly mean temperatures were significantly lower in Wascheid than at all other sites for every year (Table S1). Wind speed showed ( Figure S2) significant differences between all sites (e.g., 1998, 2004, 2011, 2014, 2017, and 2019) and the values were significantly higher in Herdorf in some years.  The spatial distribution of meteorological parameters, as a function of altitude, is shown in (Table S1). For example, higher altitudes experienced lower temperatures and VPD but higher precipitations and O 3 concentrations. On the other hand, in low altitude sites, temperature was higher whereas precipitation and O 3 were lower than at high altitude sites (Table S1). Moreover, there are significant differences between sites for meteorological parameters during the time series ( Figure S2). For example, yearly mean temperatures were significantly lower in Wascheid than at all other sites for every year (Table S1). Wind speed showed ( Figure S2) significant differences between all sites (e.g., 1998,2004,2011,2014,2017, and 2019) and the values were significantly higher in Herdorf in some years.

Correlations between the Parameters
To characterize all correlations between all parameters and to identify the main factors contributing to the variance between data, a principal component analysis was performed ( Figure 5).
Atmosphere 2020, 11, x FOR PEER REVIEW 13 of 29 To characterize all correlations between all parameters and to identify the main factors contributing to the variance between data, a principal component analysis was performed ( Figure 5). The PCA was performed on meteorological and pollution data in hourly resolution to identify the variables with the most important impact on variance ( Figure 5, Tables 4 and 5). The first and second axes justify the most significant changes in the eigenvalues, respectively, at 3.05 and 2.30, corresponding to 21.8% and 38% of the accumulated variance, respectively (Tables 4). The PCA revealed that higher NO and NO2 values were associated with Herdorf and Neuhäusel ( Figure 5). NO2 and NO were negatively correlated with O3 concentration, wind speed, altitude, temperature, and global radiation. High altitude and high wind speed values were associated with the sites Hortenkopf, Leisel, and Wascheid ( Figure 5). The PCA was performed on meteorological and pollution data in hourly resolution to identify the variables with the most important impact on variance ( Figure 5, Tables 4 and 5). The first and second axes justify the most significant changes in the eigenvalues, respectively, at 3.05 and 2.30, corresponding to 21.8% and 38% of the accumulated variance, respectively ( Table 4). The PCA revealed that higher NO and NO 2 values were associated with Herdorf and Neuhäusel ( Figure 5). NO 2 and NO were negatively correlated with O 3 concentration, wind speed, altitude, temperature, and global radiation. High altitude and high wind speed values were associated with the sites Hortenkopf, Leisel, and Wascheid ( Figure 5).
Accordingly, VPD, T, G, and O 3 concentration vectors point in the same direction. The low angle is a sign of the high correlation between these parameters and the first axis. Moreover, there were very long vectors such as VPD, T, G, and O 3 that highlight the importance and high contribution of the first axis as compared to vectors with shorter length like P and W with wide angles to both axes. The Pearson correlation coefficients for O 3, NO 2, NO, and the meteorological parameters are shown in Table S3 for all sites. It shows that O 3 concentrations were negatively correlated with NO and NO 2 at all sites. Furthermore, there was a positive correlation between NO and NO 2 at all sites. O 3 features a positive correlation with global radiation, temperature, and VPD while NO 2 and NO show a negative correlation with temperature, global radiation, and VPD. Moreover, O 3 , NO 2 , and NO had week correlations with wind speed and precipitation at all sites.

Atmospheric Water Balance
To characterize varying conditions for the exchange of gases (CO 2 , O 3 , water vapor etc.) through the trees' stomata during the investigated time period, it is important to look at the AWB. Figure 6 shows the AWB of the sites during the growing season (1st April to 30th September) of the years 1998 to 2019. Black columns illustrate humid periods with a positive balance (more precipitation than evapotranspiration) while grey columns depict arid periods with more evapotranspiration than precipitation. Hortenkopf (11 periods) and Leisel (nine periods) offer the highest frequency of arid periods. Neuhäusel (five periods), Herdorf (six periods), and Wascheid (three periods) were more humid and showed fewer arid periods over 22

Drought Intensity and Elevated Daytime O 3 Concentrations
To characterize drought intensity in the present study, negative AWB values were accumulated during a year, if the water supply in the soil was below 50% of the useable Field Capacity (uFC) (lower part of Figure 7). Negative values of AWB are an indicator for an arid atmosphere. uFC values in the rooting zone of 100 cm at the five forest sites are listed in Table 6. Negative values of AWB, mainly predominantly attributed to evapotranspiration, reduce the supply with plant available water (PAW). According to Table 6, Neuhäusel had the highest while Herdorf and Leisel had the smallest uFC value.  When qualitatively identifying the synchrony or asynchrony of elevated O3 concentrations and drought periods for the whole time series from 1998 to 2019, Hortenkopf and Herdorf experienced the highest and lowest risk for O3 uptake with 32.8% and 15.8% of asynchronous days, respectively. Conversely, the highest and lowest synchrony between elevated O3 concentrations and drought was found for Leisel with 60.8% and for Hortenkopf with 45.6%, respectively. Hence, Hortenkopf will be the site with highest potential risk of O3 uptake. Periods with no synchrony between elevated O3 and drought, were particularly prevalent during the humid years 2000 and 2007 (see Table S4), having most likely led to high O3 uptake. These qualitative results suggest that O3 fluxes might have been on average higher in relatively humid years as compared to arid years and that there is a distinct correlation between AWB or drought intensity and O3 uptake.

Spatial and Temporal Trends of Meteorological and Pollution Data (O3, NO2 and NO)
The analysis of a 22-year long time series of weather and air pollution data recorded at five forest sites in Western Germany revealed differences in air quality and meteorological conditions between these sites, with some of them being significant. These differences can partly be explained by different altitude and distance to human settlements with increased traffic and industrial activity.
The study revealed a clear decreasing trend in NO2 and NO concentrations between 1998 and 2019, which is in agreement with findings by the German Environment Agency (UBA) [86], Junk et al. (2003) [87] and Georgoulias and Stammes (2019) [88]. The sites in the present study located above 500 m a.s.l. (Hortenkopf, Leisel, and Wascheid) are good examples for sites affected by long-range O3 transport, while the lower altitude sites are more affected by local O3 formation due to local precursor emissions [89]. Figure 4b,c show high monthly mean NO2 and NO concentrations during the winter period due to increased emissions from the use of fossil fuels for heating [90,91]. NO2 (17.4 µg/m 3 ) and NO (2.67 µg/m 3 ) concentrations were highest in winter months, possibly reflecting increased   These years can hence be classified as humid years. Figure 7 also presents elevated daytime O 3 concentrations (daily daytime O 3 mean with global radiation higher than 50 W/m 2 diminished by time series mean of daytime O 3 concentration). O 3 is also prevalent outside the vegetation period at all sites. However, this is usually of minor threat for forests, because only coniferous trees might take up some O 3 during those months (at a much-reduced rate only due to non-favorable gas exchange conditions). Nonetheless, Figure 7 Table S4).
When qualitatively identifying the synchrony or asynchrony of elevated O 3 concentrations and drought periods for the whole time series from 1998 to 2019, Hortenkopf and Herdorf experienced the highest and lowest risk for O 3 uptake with 32.8% and 15.8% of asynchronous days, respectively. Conversely, the highest and lowest synchrony between elevated O 3 concentrations and drought was found for Leisel with 60.8% and for Hortenkopf with 45.6%, respectively. Hence, Hortenkopf will be the site with highest potential risk of O 3 uptake. Periods with no synchrony between elevated O 3 and drought, were particularly prevalent during the humid years 2000 and 2007 (see Table S4), having most likely led to high O 3 uptake. These qualitative results suggest that O 3 fluxes might have been on average higher in relatively humid years as compared to arid years and that there is a distinct correlation between AWB or drought intensity and O 3 uptake.

Spatial and Temporal Trends of Meteorological and Pollution Data (O 3 , NO 2 and NO)
The analysis of a 22-year long time series of weather and air pollution data recorded at five forest sites in Western Germany revealed differences in air quality and meteorological conditions between these sites, with some of them being significant. These differences can partly be explained by different altitude and distance to human settlements with increased traffic and industrial activity.
The study revealed a clear decreasing trend in NO 2 and NO concentrations between 1998 and 2019, which is in agreement with findings by the German Environment Agency (UBA) [86], Junk et al. (2003) [87] and Georgoulias and Stammes (2019) [88]. The sites in the present study located above 500 m a.s.l. (Hortenkopf, Leisel, and Wascheid) are good examples for sites affected by long-range O 3 transport, while the lower altitude sites are more affected by local O 3 formation due to local precursor emissions [89]. Figure 4b,c show high monthly mean NO 2 and NO concentrations during the winter period due to increased emissions from the use of fossil fuels for heating [90,91]. NO 2 (17.4 µg/m 3 ) and NO (2.67 µg/m 3 ) concentrations were highest in winter months, possibly reflecting increased residential heating. Junk et al. (2003) [87] stated high monthly NO 2 concentrations in winter and low monthly NO 2 concentrations in summer in Trier, Germany.
Surprisingly the present study revealed no temporal trend in mean 24 h or daytime O 3 concentrations during the time series. It seems that that decreasing O 3 extremes (95% Quantile) and increasing low O 3 concentrations (5% Quantile) held the high varying O 3 concentration on unmodified level (Table 3). High O 3 concentrations recorded during the 22 years at some high-altitude sites might be the result of long-range transport from nearby urban agglomerations, as the concentrations of O 3 precursors (e.g., nitrogen oxides) at these sites are too low to explain the high O 3 concentrations. These precursors are transported into clean air forest regions, where reduction agents in the air are either missing or are rare, hence O 3 has a higher life expectancy as in densely populated urban areas. Differences in O 3 concentrations depending on different altitudes are not as distinct as expected. That may have two reasons, at first the altitudinal gradient in the present study is small (200-300 m) compared to other studies and second the long-range transport of O 3 and the distances of the stations to urban agglomerations could be different to other studies. This is in agreement with Sicard et al. (2009) [27], who reported a slightly increasing trend in tropospheric O 3 between 1995 and 2003 at French rural monitoring network stations, which was strongly influenced by the altitude. Also, Klingberg et al. (2009) [92] in Sweden and Wehner and Wiedensohler (2003) [93] in Germany showed that high global radiation at high altitudes is contributing to high O 3 concentrations, which supports the findings of the present study. As shown in Figure 2 and Table S1, 2003 was exceptionally hot with the highest global radiation and the highest O 3 concentrations as compared to the other years. Baumgarten et al. (2000) [94] and Treffeisen and Hald (2000) [95] argued that the cumulative hourly O 3 concentrations were enhanced at higher altitude stations in Germany as compared to lower altitude stations most likely due to transport from urban agglomerations. Figure 4a shows a higher mean O 3 concentration observed in summer as compared to the winter due to high temperature, low humidity and high light intensity, all of which promote O 3 formation [96]. Similar annual trends in Germany were observed by Treffeisen and Halder (2000) [95] and Meleux et al. (2007) [97].
Surface O 3 showed negative correlations with NO and NO 2 at all stations (Table S3). Therefore, a rise in O 3 concentrations is associated with a reduction in NO and NO 2 concentrations. Correlations among the concentrations of different pollutants (O 3, NO 2 , and NO) were stronger at Herdorf than in other forest stations, because it was the lowest altitude station of this study with a higher density of traffic and human activity. Mavroidis and Ilia (2012) [98], Latif et al. (2014) [99], and   [86] found corroborating results, i.e., that there is are close correlations between concentrations of different pollutants (O 3, NO 2 and NO). Figure 5 and Table S3 show correlations of meteorological parameters (G, VPD, T, P and W) with O 3 concentrations, reflecting photochemical processes in the atmosphere which are responsible for O 3 formation. Figure 5 illustrates the highest O 3 concentrations during periods with low precipitation. Lower precipitation (lower cloud cover) usually corresponds with higher global radiation, higher temperatures, and higher O 3 formation rates as reported by Tarasova and Karpetchko (2003) [100] and Kovač-Andrić, et al. (2009) [79]. Dawson et al. (2007) [101] indicated that temperature exerted the largest influence on O 3 concentrations. Singla et al. (2012) [102] revealed a strong correlation between global radiation intensity and O 3 concentration. The analysis of Camalier et al. (2007) [103] confirmed that O 3 increases generally with increasing temperature while O 3 decreases with increasing relative humidity (because relative humidity usually decreases with increasing temperature). All these investigations confirm the results obtained by this study, which showed that in higher altitudes the air is cooler and more humid, and the vegetation periods are shorter, thereby representing a montane climate. Furthermore, the VPD is lower but wind speed is higher as compared to lower altitudes (Table S1).
The present investigations also display increasing temporal trends in global radiation at all sites probably as result of less cloud cover, decreasing temporal trend in precipitation and wind speed with exception of one site (Leisel), probably due to a decrease in thunderstorms during summer months. These meteorological trends could at least be partly linked to a changing climate, but might be partly modified by the altitude, slope, and aspect (orientation) of the forests sites and measuring stations.

Atmospheric Water Balance and Drought Extent
O 3 is formed under hot and dry weather conditions, which coincide with water stress for plants. Under these conditions, plants will save water by reducing transpiration through lowering leaf conductivity, hence also limiting the flux of gaseous pollutants into the plant. Therefore, plants will Atmosphere 2020, 11, 1261 20 of 28 be less at risk from O 3 when water stress occurs. They protect themselves against water loss by closing stomata through transpiration control. Despite limiting the assimilation rate through stomatal closure, this behaviour can potentially benefit the plant, because drought stress as well as O 3 uptake can otherwise lead to even larger reductions in assimilation rate and therefore productivity, which is usually associated with hindered allocation of, assimilates. The net effect of stomatal closure for the plant's assimilation can only be quantified through the calculation of stomatal O 3 flux. This calculation also takes into account the frequency and intensity of dry periods and interacting O 3 concentrations under a changing climate (higher radiation, decrease in precipitation) as described above. Figure 7 documents high variance in drought intensity for humid and dry years. Dry periods (years) are characterized by insufficient summer precipitation and an above-average evaporation [52,54,55]. Years with a negative AWB are periods of high evapotranspiration and/or reduced water storage in the soil (water supply). Some studies in Germany [104][105][106][107], Romania [53], Italy [57] and Greece [58] demonstrated that evaporation and plant production are limited by reduced rainfall. If this deficit cannot be compensated by PAW, increased evapotranspiration is likely to lead to enhanced drought conditions [107,108]. According to Figure 6, Wascheid with the highest elevation (680 m a.s.l) and the highest mean of precipitation sum (Table S1)  Years with a negative AWB (e.g., 2003, 2011, 2018, and 2019) are expected to have posed a lower O 3 risk to forest trees, but at the same time these years will also have led to vitality loss of trees due to the experienced drought impacts. AWB alone is not a satisfactory tool to investigate the net drought effect, because plants can compensate atmospheric drought by higher water uptake from soil, provided it is available. To characterize real drought events, soil water influence and plant hydraulic conductance must be integrated. Therefore, a drought index was developed in the present study that includes atmospheric drought and the soil water supply up to a water stress level that is so high that leaf conductance will be reduced.

Synchrony of Drought Extent and Elevated Daytime O 3
Elevated daytime O 3 concentrations (concentrations above the time series mean of daytime O 3 ) and drought intensity appeared to be often synchronous during the study period ( Figure 7). Elevated daytime O 3 concentrations can lead to high O 3 flux rates in case of adequate water supply. This will be the case under asynchrony between elevated O 3 and drought. However, under water stress, leaf conductivity and therefore O 3 flux is reduced, as described above.
High O 3 concentrations can often be linked to water stress [109]. To investigate this phenomenon, the synchrony of O 3 and drought intensity during the entire study period was analysed and is shown in Figure 7 and Table S4. The coincidence of elevated daytime O 3 and high drought intensity was predominantly observed in mid and late summer, as also reported by Panek (2004) [110] who showed that stomatal conductance and O 3 uptake declined due to declining soil water content during the summer. Regarding the present results, the uptake of O 3 in spring and early summer will have been higher than during the rest of vegetation period, given the better water availability and hence better O 3 uptake conditions for the forest trees during the earlier parts of the growing season. This will have probably led to higher effects of O 3 uptake on growth and production of the trees in spring, early or midsummer than in later months; in contrast, O 3 uptake in late summer or early autumn often leads to premature senescence.
There were some years (1998, 2000, 2004, 2007, and 2008) in which the occurrence of O 3 events was not accompanied by pronounced droughts, which most likely led to high O 3 fluxes because leaf conductance was not reduced and the O 3 concentration was high. Table 6 and Figure 7 showed that drought intensity was more frequent and more intensive at sites with high and medium uFC (sandy soil at Hortenkopf and silty loam and strongly silty clay at Neuhäusel) and less frequent at stations with more silty soils with lower uFC (silty soil in Leisel, Wascheid and Herdorf). The latter stations also exhibited colder, rainier, and more humid conditions (montane climate).
Drought intensities in Germany are mainly driven by the European macro weather situation (North Atlantic Oscillation; NAO), but they do not occur regularly every year. However, if they occur, they often last until the end of the vegetation period in autumn. Hortenkopf experienced the highest O3 concentrations (Figure 7) and the longest drought duration (Figure 7 and Table S4), representing the highest O3 risk in this study. Jing et al. (2016) [111] in the USA, Matyssek et al. (2006) [49] in Germany and Lin et al. (2020) [44] in Europe found similar results; their demonstrated function mechanisms showed that that dry weather conditions likely lead to high O3 concentrations, but at the same time to unfavorable meteorologically driven O3 uptake conditions. It could hence be stated that the duration and intensity of drought will have had a high influence on O 3 uptake by the forest trees of this study. The quantification of the O 3 uptake of these forest trees under the influence of drought was not the aim of this study, but will be published in a separate article using an O 3 flux model (in preparation).

Conclusions
The present study confirmed that (i) the temporal trends of the 22-year long time series of measured meteorological parameters and air pollutants corresponds with the global climatic change and human abatement strategies during that timeframe, (ii) the altitude of forest sites and their distance to urban agglomerations as well as the oncoming flow of pollutants from long-range transport and the existence of gaseous reducing agents determine the prevalent O 3 concentrations at the investigated sites, and (iii) the duration, sequence, and intensity of droughts as well as its synchrony to elevated daytime O 3 concentration determine the exposition of vegetation to O 3 and hence influence the plants' protection against O 3 exposure and toxicological offense. The imponderable and highly variable occurrence of drought under temperate climate conditions typical for Central Europe prearranges the risk of forest vegetation to O 3 .
The meteorological conditions in the investigated West German midrange belt are not constant and vary from year to year with arid and humid phases. The significant increase of global radiation and decrease of precipitation in this montane belt is a typical indicator for global climatic change. Human activity depending on population density and characterized by traffic density and industrial activities influence the pollutant concentration; these concentrations are further modified by pollutant transport processes and altitudinal influences in case of meteorological parameters.
The stomatal conductivity is highly dependent on water supply, so that, in dry years with high O 3 concentration, the O 3 uptake could be smaller than in humid years.
The described meteorological changes in the investigated time series from 1998 until 2019 based on real measurements most likely led to smaller gas exchange rates and hence assimilation rates caused by reduced leaf conductivity. This might have been at times an advantage for the vegetation, because a changing (i.e., hotter and drier) climate can protect vegetation against the attack of O 3 . However, the reduction in leaf conductivity has primarily its origin in water deficiency, which leads to less assimilation rates, less growth rates and therefore shortage in substance and energy for growth, competition, and stress defense.
As shown for the presented data series, climate change will lead in the targeted German midrange mountain region to higher radiation (that precedes in non-significant cases in higher air temperature and higher VPD), as well as to significant lower rainfall and wind speed, especially in the summer. These are conditions for increasing drought stress for plants, while at the same time higher precursor concentrations such as biogenic NMVOC occur due to increased temperatures and radiation. In the presence of NO concentrations, more O 3 will be formed, if enough energy in form of radiation is available [112,113]. The same meteorological conditions induce drought stress, which reduces leaf conductivity and therefore results in smaller assimilation rates and simultaneously smaller O 3 uptake.
The result would be a smaller growth rate (detectable e.g., as smaller tree rings and reduced carbon sequestration), but it would be very difficult to separate which part of the reduction in biomass production was caused by O 3 and which by drought.
The growth reduction has to be interpreted in the light of the complex linkage of carbon sequestration with growth, stress defence and competition [114]. The resources of a plant are limited. If the costs for the defence against increasing O 3 concentrations are rising, the plant emits reductants like carbohydrates (e.g., terpenes) and/or invest for detoxification of leaf internal O 3 by ascorbic acid, thereby reducing the energy available for growth or competition.
Losses in vitality and productivity of forest trees due to O 3 are seen as particularly worrying in the context of the current debate on the state of European and German forests, which have recently experienced an unprecedented attack from bark beetles (mainly on Norway spruce), favoured by unusually dry springs and summers. Forest trees already weakened by multi-year exposure to O 3 will be more susceptible to drought and bark beetle attacks, and vice versa. Hence, it is very important for the management of (state, communal or private) forests to quantify the effect O 3 has and will have under current and future physical and pollutant climates on the growth and productivity of forest trees. O 3 flux models are key to these quantifications; one of these models has also been used in the present project, as will be demonstrated in a separate publication (Eghdami et al., in preparation).

Supplementary Materials:
The following are available online at http://www.mdpi.com/2073-4433/11/11/1261/s1, Table S1: Yearly and total summary of meteorological parameters of the investigated stations, Table S2: Differences of monthly means of air pollutants in the time series from 1998 to 2019 at five stations, Figure S1: Boxplot of hourly NO concentrations (µg/m 3 ) with annual variance and trends at the stations (S3.a-S3.e), Figure S2: The boxplots present the variation of meteorological parameters for the 22 years (1998-2019) among five stations, Table S3: Pearson correlation coefficients between NO, NO 2 , O3 as well as to selected meteorological parameters for the all stations. Table S4