NO2 Seasonal and Interannual Variability in Ukrainian Industrial Cities

Abstract The paper aims to define the main features and principles of seasonal and interannual NO2 variations in Ukrainian industrial cities. Using ground-based measurements for 15-year period, it shows weak NO2 seasonal variability that could intensify in case of three regularities. These regularities depend on impact of natural conditions during anthropogenic emissions growth and redistribution between emission sources. Most industrial cities are characterized by positive trends even if stationary industrial emissions fall. NO2 interannual changes forms under variety of fluctuations. However, 6.2- and 9.3-year periods have the biggest impact and might be explained by low-frequent lunar tidal forces through its influence on meteorological conditions.


INTRODUCTION
Nitrogen dioxide (NO2) is one of the most dangerous atmospheric compounds for human health [1]. Huge anthropogenic role in NO2 content increasing is the reason of great society attention to its temporal and spatial distribution [2,3]. NO2 seasonal variations do not coincide in different areas [4,5,6,7,8,9,10], thus analysis of seasonality formation helps to understand the main reasons of maximal values emergence. The biggest influence that drive NO2 seasonal dynamics has combustion of fossil fuels (industry and traffic) [11]. Biomass burning [12], lightning, microbiological activity [12,13] and other natural sources are not determinative. Temporal distribution and power of each emission source could cause different NO2 seasonal variations. The study presents analysis of seasonal NO2 dynamics in industrial cities with regularities that indicate seasonality smoothing or emergence.
Interannual variability forms under changes in meteorological conditions [14]. However, periodicity of nitrogen oxides changes still is controversial question, because of short life-time [15,16] and considerable spatial heterogeneity, especially in big industrial cities and regions with large anthropogenic emissions [6,17]. The study aimed to define spectrum of significant NO2 interannual fluctuations that observed in Ukrainian industrial cities. Knowledge about interannual periodicity allow to find regularities of NO2 distribution on global scale, its chemical feedbacks and interaction between atmospheric air composition and meteorological parameters.
The most investigated interannual changes connected with NO2 trends, which rapidly grow in urban areas [6,8,9,19]. Permanent changes of anthropogenic emission sources in industrial cities cause difference in NO2 tendency and have to be analyzed unceasingly. It will help to delimit areas with high pollution level or areas with high probability of natural changes which are favorable of nitrogen oxides accumulation.

DATA AND METHODOLOGY
Monthly ground-based NO2 data involved for analysis and were averaged in 14 industrial cities rather even distributed on the territory of Ukraine (Fig. 1). Denser distribution selected for regions with high pollution level. Total period in research cover 15 years from 2000 till 2014, which allowed to divide on three five-year' periods (2000 -2004, 2005 -2009, 2010 -2014) and make analysis of short-term variations within them.

Figure 1. Spatial distribution of cities involved in research
Seasonal variations were calculated as sum of first five significant harmonics obtained from Fourier analysis (1), where Fourier coefficients defined using least squares method. Trends and interannual variations were analyzed on residuals of seasonal variations after average values removal from time series. Interannual variations were also calculated using Fourier analysis for 130 equally distributed frequencies. The most significant separate variation had been calculated and its impact removed from the residuals. After this procedure Fourier analysis applied again while any significant fluctuation exists.

RESULTS AND DISCUSSION
Understanding the role of different emission sources in NO2 dynamics is not easy for Ukrainian industrial cities. The main problem lies in quality of anthropogenic sources inventory. For stationary sources, e.g. factories, fossil-fuel power stations etc., inventory reports in Ukraine rather well reflect air pollution [20], but there are no such quality inventories for traffic emission.
To cope with this problem, preliminary analysis for Kyiv city was made. Kyiv has the biggest air pollution measurement network and consist of 16 stations. It was found, that maximal values in winter -early spring are typical for all stations with high traffic pollution. Summer maximal values observed on two stations in rather clean areas (Hydropark and Nauky Avenue), which are typical for natural NO2 changes [9,10]. It means that meteorological conditions and land cover type on quite clean territories within cities could change seasonal variability. Six stations in Kyiv have both of these maximal. Therefore, distribution of anthropogenic sources within industrial cities cause different types of seasonality.

Seasonal variability
NO2 seasonal variations in cities are absent and determination coefficients usually do not exceed R 2 =0.10. Therefore, availability of seasonal variations in NO2 fluctuations evokes particular interest. For general period 2000 -2014 statistically significant seasonal variations observed only in Kharkiv with R 2 =0.47 and could be considered as weak (Fig.2). First harmonic (1-year period) has the biggest influence with 83% impact on total dispersion. The highest concentration observed in late spring, and phase falls on 149 th Julian day.

Figure 2. Average NO2 seasonal variations in Kharkiv for 2000 -2014
Despite Kharkiv is big city with approximately 1.5 million inhabitants, the average NO2 concentration is only 0.03 mg/m 3 compared with 0.12 mg/m 3 in Donetsk or 0.09 mg/m 3 in Kyiv. On one hand, it is possible to make assumption that power of anthropogenic emissions is the main reason of seasonality smoothing in industrial cities. On the other hand, in Luhansk seasonal variations are absent with R 2 =0.04, but average NO2 concentrations are also low (0.03 mg/m 3 ). Therefore, NO2 seasonality formation in industrial cities is more complicated than only dependence from the power of anthropogenic emissions. Such abrupt seasonality emergence caused by three main regularities which depend on NO2 content. First regularity exists for cities with strong anthropogenic emissions. Seasonality appears if average concentrations are high and constantly increase. It is well observed in Kyiv where concentrations changed from 0.08 to 0.10 mg/m 3 , Lutsk -from 0.05 to 0.09 mg/m 3 and Donetsk -from 0.08 to 0.14 mg/m 3 . If NO2 content is not so high, any increasing does not entail changes in seasonality.
In case of low or moderate NO2 concentrations, an opposite effect is observed, and thus the second regularity acts -seasonality appears if concentrations continue to decrease constantly. In Odesa average concentrations fell from 0.07 mg/m 3  Both regularities represent the same idea: seasonality become stronger if prevailing anthropogenic emission source become more powerful: in highly polluted industrial cities with NO2 content increasing and in less polluted cities with NO2 concentrations decreasing.
The third regularity is connected with redistribution of emission sources. In this case, amplitudes or average values may not change. The best indicator of such type changes is phases shift. Of course, phases shift does not exist in polluted industrial cities with high NO2 content. In Kharkiv with concentrations less than 0.033 mg/m 3 phases constantly shift on 14 days every 5 years (from 121 to 149 Julian day) which coincide with seasonality emergence. In Lviv with concentrations less than 0.047 mg/m 3 phases shifted from 227 th day to 107 th day and determination coefficient decreased from R 2 =0.30 to R 2 =0.07. Three curves on Fig. 3 represent NO2 seasonal variations for 3 five-year' intervals based on first harmonic. Phase shift coincides with amplitude decreasing and seasonality smoothing, which relates to redistribution of NO2 emission. With average values increasing it is possible to make conclusion about intensification of definite anthropogenic emission and its preponderance in comparison to other sources. If anthropogenic emissions become more powerful, maximal values shifts from summer to winter-spring.

Figure 3. Phase shifts and seasonal NO2 variability in Lviv for 3 five-year' intervals
Redistribution between emissions usually observed in case of low concentrations. With NO2 content increasing such regularity lose its power. In case of high concentrations, caused by anthropogenic emission, phases shift connected only with adding or cessation of separate industrial sources, e.g. factory building in cities with high NO2 emission from transport. Direction of phase shifts show the intensification/ weakness of anthropogenic sources.

Trends
Almost all industrial cities in Ukraine have significant trends of NO2 concentrations. Fisher test showed tendency absence in Mariupol, Uzhgorod and Luhansk. Most trends are positive, even if emission from stationary sources decreased [20]. It means that traffic emissions have crucial impact on NO2 distribution in Ukrainian industrial cities.
There are two negative trends: in Odesa and Kharkiv with values -0.01 mg/m 3 per decade. Determination coefficients are R 2 =0.18 and R 2 =0.21 respectively (Fig. 4). The highest positive trend reached 0.07 mg/m 3 per decade in Donetsk city with R 2 =0.48 (Fig. 5)  In Kharkiv and Odesa monthly trends are negative as its general trends. Significant monthly trends in Odesa observed from November till April and vary from -0.01 mg/m 3 per decade to -0.02 mg/m 3 per decade. In Kharkiv monthly trends are significant in June, December and February -April with values -0.01 mg/m 3 per decade.
Despite most industrial cities have trends, there are no dependence and regularities in monthly trends. It means that tendency of NO2 emissions in industrial cities is unique for each city and depend on the processes within it.

Interannual variability and spectrum of fluctuations
According to slight seasonality, interannual variability is the only changes connected with natural forces. Significant interannual variations explain from 29 to 61% of residuals dispersion, however, there are no dependence withNO2 content. The highest determination coefficients of interannual dynamics reached 0.61 in Lviv, 0.60 in Dnipro (Fig. 6) and 0.55 in Armyansk. Less clear interannual variations observed in Zaporizhzhia, Uzhgorod and Krasnoperekopsk with R 2 =0.29…0.31. in Kyiv. Surely, these forces do not act directly on the NO2 concentrations. Interannual variability in wind and temperature distribution, connected with lunar nodal tidal cycle [21,22,23], caused changes in meteorological conditions and possibility to NO2 accumulation in the atmosphere.
Another big group of variations is situated within higher frequency spectrum with approximately 2.6 -3.4 years' period. Generally, 7 out of 14 researched cities have these fluctuations, but impact is lower with determination coefficients within R 2 =0.05…0.14. The highest amplitude is 0.01 mg/m 3 .
It was found another four groups of interannual fluctuations: with periods 2 -2.2 years (4 cities), 3. Other variations do not form any group and distribute separately in spectrum. For some cities these separate fluctuations have the biggest impact up to 45% in Armyansk, 43% in Kharkiv, 34% in Donetsk, 30% in Khmelnytskii and Lviv, 15% in Luhansk.

DISCUSSION
NO2 seasonal changes in industrial cities depend on predominant emission sources and clarity of fluctuations varies in time with emission changes. Maximal values are typical for winter in industrial regions and polluted areas, due to cold-started vehicles and residential heating [5,9,24], inversions and lower temperatures [5]. In more "clean" areas summer maximal is usual [9,10,18]. Our study finds three regularities of seasonality changes in industrial cities. However, maximal values on some territories may be observed in winter [6,9,17], summer [8,9,10,17] and spring [7,8].
NO2 in most industrial cities increase, even if pollution from stationary sources decrease. This fact showed that traffic emission has crucial impact on NO2 distribution. NO2 positive trends observed in other urban areas [6,8,9,18,19] and dependence from vehicles were proved [6,7,8,25]. NO2 interannual variability caused by natural forces. Changing meteorology (winds, temperatures, humidity and clouds) on a global scale drive interannual variability of NO2 [14]. As our research show, meteorological variations caused by low-frequent lunar tidal forces could be the reason of 6.3-and 9.2-year fluctuations, which have the biggest impact on interannual NO2 variability.