Intraurban and Longitudinal Variability of Classical Pollutants in Kraków, Poland, 2000–2010

In spite of a dramatic decrease in anthropogenic emissions, ambient concentrations of major pollutants have not changed within many urban locations. To clarify the relationship between ambient air quality trend and the population exposures, we compared the intraurban versus temporal variability of the collocated measurements of five major air pollutants including particulate matter (PM) with an aerodynamic diameter <10 µm (PM10), < 2.5 µm (PM2.5), tropospheric ozone (O3), sulfur dioxide (SO2), and nitrogen dioxide (NO2), in Kraków, Poland, during the 2000−2010 period. Strong seasonal trends and overall absence of spatial heterogeneity in PM10 and PM2.5, except in the traffic monitoring site, were observed across the monitoring network. The range of median PM2.5 concentrations during winter (54–64 µg/m3) was 3- to 4-times higher than the summer medians (15–26 µg/m3) across the sites during 2009−2010. Furthermore, large proportion of PM10 appears to be comprised of PM2.5 (PM2.5/PM10 concentration ratios range, 0.5–0.7). At each monitoring site, the Pearson’s correlation coefficients between PM2.5 and PM10 ranged between 0.944 and 0.963, suggesting a health-relevance of PM10 monitoring. One ln-unit increase in PM10 was associated with 92%–100% increase in PM2.5 concentrations in the same location. While PM10 did not demonstrate a clear temporal trend, SO2 concentrations steadily declined by 40% during the 2000–2010 period. Summertime median NO2 concentration was acutely elevated ‎(70 μg/m3 vs. 22 μg/m3) at the traffic oriented site compared to the city’s central monitoring site. The traffic and the industrial sites were associated with highest number of days during which 24-hour mean PM10 and PM2.5 concentrations exceeded the European Union standard. Steadily growing contributions by vehicular emissions appear to be associated with the absence of clear trend in PM10. Current practices of air quality control within Kraków may not be adequate for the protection of the public’s health.


Introduction
In spite of reduction in anthropogenic emission of major air pollutants within Europe during the last several decades, such a trend has not been matched by corresponding declines in childhood asthma and allergy prevalence [1,2]. Poland represents an example of such a contradiction. Staring around 1954 under the Communist regime [3], the country has emerged as one of the highest producers and consumers of coal within Europe [4][5][6][7]. For example, total annual emission of particulate matter within Kraków, a city with one of the highest historic levels of air pollution within eastern Europe, is estimated at 150,000 tons during the 1970s [3]. The associated mean ambient PM10 concentrations during the same period range between 180 μg/m 3 (the city center) and 109 μg/m 3 (the suburbs) [3]. Starting in 1980s, a number of semi-ecologic investigations in Krakow have shown an association between chronic exposures to airborne PM with cause-specific mortality [8,9]. In particular, exposures to particulate matter (PM) have demonstrated robust associations with wide number of health end-points [10][11][12][13][14][15][16][17][18]. At same time, concerns over the deterioration of the natural environment as well as the city's cultural heritage sites have also grown [19]. Around 1989, Poland's political transition to democracy following the collapse of communism has led to a substantial decrease in airborne concentrations of SO2, black carbon, PM, and airborne heavy metals [20]. Beginning around 1995, both regional and national government bodies have made concerted efforts to improve the air quality [3].
To deepen our understanding of the early-life environmental contributions on childhood asthma and neurocognitive impairments, we have been following prospective birth cohort in Krakow since 2000. Our exposure assessment analyses have shown that individual pregnant woman's personal exposure to particle-bound large PAHs is predominantly influenced by corresponding ambient concentrations [21][22][23]. In addition, there is an extremely high correlation between total sum of eight pro-carcinogenic PAHs and simultaneously monitored PM2.5 concentration [10,24,25]. Furthermore, between-person variability in personal exposure to PAHs at given 48-hour window are much smaller than within-person variability [23] or that of the mean ambient concentration [21]. We reported that time-activity pattern of the individual women was not a significant predictor of the personal exposure to particle-bound PAHs [21]. Contrary to our expectation, prenatal exposures to PM2.5 and PAHs pose significantly increased risks of intrauterine growth restriction, wheezing symptoms, and asthma during childhood, respectively, in spite of reduction in coal-burning related pollutant emissions [10][11][12][13][14][15][21][22][23][26][27][28][29].
Considering the impact of the ambient sources on the personal exposure, the overarching aim of this investigation is to characterize the intraurban trend of five major pollutants across the years 20002010. The time-period of our interest corresponds to prenatal and first seven to ten years of the cohort children's life. This analysis is expected to lay the groundwork for the clarification of the relationship between long-term intraurban trend and chronic exposure profile of each child in the cohort. Furthermore, we posit a priori that reduction in coal-burning related emissions is associated with temporally corresponding decline in PM10 concentration during the 11-year period of interest. In order to answer this postulate, we: (1) describe the overall trend in the five pollutant concentrations at the six monitors over an 11-year period; (2) compare the size of the season-dependent variability in the five pollutant concentrations according to the site; and (3) explore the influence of the known emission sources and the meteorological factors on PM10 and PM2.5 concentrations. Pollutants of interest include particulate matter (PM) with aerodynamic diameter < 10 µm (PM10), PM < 2.5 µm (PM2.5), tropospheric ozone (O3), sulfur dioxide (SO2), and nitrogen dioxide (NO2).

Study Site Characterization
Kraków, Poland, holds a unique position within Polish cultural and academic heritage. Located in southeastern Poland (see Figure 1), it encompasses 327 km 2 and supports 757,400 inhabitants as of 2005. The city has at least three well-recognized air pollution sources: industrial and coal-fired power plants [20], coal-burning domestic stoves with no or outdated abatement technologies [4,5], and automobile traffic [20]. Following Poland's annexation to Soviet Union around 1954, a coal-burning steel mill (i.e. Lenin Steelworks) and a power plant (i.e. Kraków-Łęg plant) were built [3]. To date, these industrial plants continue to provide electricity and heat for new sections of the city. However, collapse of communism in 1989 has reduced the heavy industrial activities within and around the city [3]. Such shift also introduced a gas-operated heating system within the city (covering approximately 30% of the homes) [3]. Accordingly, sulfur dioxide and particulate matter concentrations have steadily decreased.
Kraków is located in the Vistula river valley surrounded by Carpathian Foothills to the south and The Kraków-Częstochowa Upland to the north ( Figure 1). This geographic location has been associated with atmospheric inversions approximately 27% of the entire year, particularly during wintertime [3].

The Air Pollutant Sampling and Analysis
The ambient air quality monitoring network in Kraków is operated by Voivodship Inspectorate for Environmental Protection in Krakow (VIEP). Krakow air monitoring network was launched in 1991 in collaboration between the US Environmental Protection Agency and Voivodship Sanitary-Epidemiological Station in Kraków (19682001) and the Voivodship Inspectorate for Environmental Protection in Kraków (1992present) [30]. It has been providing automatic continuous measurement of air pollutants such as SO2, NO, NO2, NOx, CO, O3, particulate matter with an aerodynamic diameter less than 10 μm (PM10), and from 2009 also PM2.5. Laboratory of VIEP got accreditation (contract no AB 176) of Polish Centre for Accreditation (PCA) in 1998 for air quality monitoring testing as a first air monitoring network laboratory in Poland.
Primary automatic analytic methods for the pollutants during our investigation period include: UV fluorescence for SO2 and chemiluminescent method for NO and NO2 with gas analyzers produced by Thermo Environmental Instruments Inc, ( As a laboratory with accreditation (PCA no AB 176) it has implemented system of quality control and assessment in monitoring network according to PN-EN ISO/IEC 17025 norm. It is focused on internal quality control based on qualified staff, instruments calibrations, and completeness of measurements series. To guarantee accuracy and reliability of derived measurements, Krakow VIEP laboratory participate in inter-laboratories comparisons (both in Poland and other EU countries, e.g. Joint Research Center, Ispra, Italy) as well as in national and international proficiency tests. It is a member of AQUILA network. It is also responsible for setting up inter-calibration meeting on the national as well as EU level as a National Calibration Reference Laboratory.
PM10, SO2, and NO2 were monitored in all six stations year round as 24-hr mean concentrations. O3 was monitored in four stations (i.e., URBAN, CENTRAL, SUB1, and SUB2) during 20002010 period. In contrast, PM2.5 were monitored in four stations (i.e., URBAN, TRAFFIC, INDU, and SUB2) during 20092010 period only. Meteorological data were monitored in INDU, SUB1, and SUB2 sites for temperature and wind speed during January 2000December 2010. Figure 1 shows the approximate location of the six monitoring stations providing data for this analysis. The Rynek Główny (CENTRAL) station sits atop a bell tower in Old Town Square of Kraków of approximately 0.04 km 2 in size. Since construction during 13 th century, this largest central plaza in Europe has been a pedestrian square. In contrast, the Aleja Krasińskiego (TRAFFIC) station is located on a busy road in the commercial hub near the CENTRAL site. The Krowodrza (URBAN) station sits on the northern mixed residential and commercial zone as the urban background site. Nowa Huta (INDU) station represents a mixed suburban and industrial zone. Prokocim (SUB1) station represents the newly expanded southern district. Kurdwanów (SUB2), located in southern edge of Krakow, represents urban background site.  (2000)(2001)(2002)(2003), and Kurdwanów (SUB2)(2010) are 3.9 km, 1 km, 9.8 km, 7.5 km, and 7 km away, respectively.

Descriptive Analysis
Present analysis includes the data from four to six monitoring station during January 2000 and December 2010 period for PM10, SO2, O3, and NO2 and 20092010 period for PM2.5. Seasons were defined as summer (JuneAugust), transitional (April, May, September, and October), and winter (NovemberMarch). Considering large variability in sample size by site, year and season, extensive non-parametric analyses were conducted for each pollutant. The relevance of predictor variables were examined using Mann-Whitney U-test or the Kruskal-Wallis test depending on the number of categories for the independent variables at α = 0.05 level of significance. There were no pollutant concentrations below the detection limit. All extreme and outlying values were double-checked for accuracy in measurement. Upon positive verification, they were retained in the data. Descriptive analysis was conducted to identify monitoring sites, season, and year, which demonstrate significantly elevated concentrations.

Linear Regression Model
Pollution variables were natural-log (ln) transformed in order to achieve normal distributions (Komolgorov-Smirnov tests > 0.05) and homoscedasticity. To better understand the relative contribution of PM10 on PM2.5, a linear regression model of PM2.5 was run at four respective monitoring sites which simultaneously monitored PM10 on PM2.5 in 20092010 combined data. The outcome, PM2.5, was modeled as a linear function of PM10 as the main predictor variable, controlling for temperature, and wind speed, at the four collocated sites. Consistent with earlier investigations, the model regression coefficient was defined as a marker for model accuracy, and adjusted-R 2 as a marker for model precision [31].

Generalized Linear Mixed Effects Model
A linear mixed effects model was fit by entering year, month, and sites as indicators variables shown in equation [1] in order to detect a trend without imposing a structure on the relationship. The reference categories were set as Saturday, December, CENTRAL and Year 2010 for the variables, weekday, month, site, and year, respectively: Where α represent the y-intercept; βn, γm, θo, δp, ζ, and η, respectively, represent the slope of the independent variables. All independent variables were forward selected if the probability of given variable in the model showed F ≤ 0.05, and removed it if the probability of the model had F ≥ 0.10.
The pollutant, X, concentration was predicted by rewriting equation [1] as follows: In order to compare the relative importance of each predictor (e.g., year), we calculated concentration impact factor of given predictor variable as IFi = exp[∑βi × (variable)i], holding all other variables constant [32]. That is, the impact factor refers to concentration change associated with a given predictor variable apart from the baseline level (i.e. y-intercept). Accordingly, the intercept term of the equation [2] reflects the mean concentration at the baseline level for all predictors (e.g., Saturday, July, CENTRAL site, Year 2010, wind speed decrease by ≥ 1 m/s, and one °C reduction in ambient temperature from 14.60 °C). The impact factor at the reference level for a given variable equals 1, given that exp[0] = 1. Accordingly, impact factor > 1 indicates predicted concentration, which is greater than the baseline concentration. On the other hand, impact factor < 1 indicates a lower predicted concentration for a given predictor variable relative to the baseline level. We conducted all statistical analyses in SAS version 9.3 (SAS Institute Inc., Cary, NC, USA). All figures were generated using IBM ® SPSS © version 22.0 (SPSS Inc., Chicago, IL, USA). Table 1 and Figure 2 show site-specific central tendencies and the exceedance days (>25 μg/m 3 for PM2.5 and >50 μg/m 3 for PM10, based on the current EU standard), and PM2.5/PM10 ratios. During summer, TRAFFIC (21%) and INDU (16%) sites were respectively associated with the highest number of days during which 24-hour mean PM10 concentration exceeded 50 µg/m 3 (the current EU standard) compared to the CENTRAL site (3%). Two suburban sites (SUB1 and SUB2) were associated with even fewer number of exceedance days during the transition season (5 and 2%, respectively). Similar trend was seen in the number of exceedance days for PM10 during the transition season for TRAFFIC (31%) and INDU (37%) site, compared to the CENTRAL site (9%). Such spatial variability was particularly acute during winter, in which TRAFFIC and INDU had highest proportion of exceedance days (39% and 51%, respectively, vs. 13% in CENTRAL site). INDU was associated with a widest range for daily PM10 concentration (6.6592 g/m 3 ) during winter ( Figure 2). Due to the high mean summer PM10 concentration at TRAFFIC site, the mean winter/summer ratio for PM10 concentrations were lower for TRAFFIC (1.5), compared to the INDU (1.9) as well as URBAN (2.0). Examining the city-wide average PM10 annual levels of the 11-year period, we observe year to year variability but little evidence regarding improvement of PM10 air quality (43. 7  As indicated by Table 1 and Figure 3, site-specific daily mean PM2.5 concentrations showed a distinct seasonal trend. The median PM2.5 concentration during summer were highest at the TRAFFIC site (26 g/m 3 ) and uniform overall at other sites (18 g/m 3 for URBAN; 15 g/m 3 for INDU; and 16 g/m 3 for SUB2). In addition, the same site was also associated with higher number of exceedance days (6%) as well as highest mean PM2.5 concentration (25.2 ± 7.0 g/m 3 ), compared to all other sites during the summer.

Descriptive Analyses
During the transition season, TRAFFIC site was associated with the highest median (40 g/m 3 ) as well as a highest number of exceedance days (8%), whereas other sites demonstrated overall uniform median concentration (25 g/m 3 at URBAN; 27 g/m 3 at INDU; 26 g/m 3 at SUB2). In contrast, similar numbers of exceedance days as well as the mean were observed during winter for the URBAN, TRAFFIC, and INDU sites (11%, 12%, and 13%, respectively) compared to the SUB2 site. Accordingly, such seasonal pattern was associated with highest median winter/summer ratios for the INDU and SUB2 (3.7 and 3.4, respectively) and the lowest winter/summer ratio for the TRAFFIC (2.5).
Taking all four sites together, the combined annual mean concentrations of PM2.5 were 43.6 ± 31.6 g/m 3 during 2009 and 46.7 ± 43.1 g/m 3 during 2010. Such concentrations far exceeded the annual mean EU standard of 10 g/m 3 [34].

PM2.5/PM10 Concentration Ratio
The relative abundance of fine fraction to PM10 is shown according to site and season in Table 1. The PM2.5/PM10 ratio exhibited a distinct seasonality. However, there were no clear differences across the sites in the ratios. During the summer, the mean ratio ranged between 0.5 and 0.7 among the four sites. During the winter, the same ratio ranged between 0.7 and 0.8. Overall, these ratios are consistent with ratios observed in other urban and semi-urban locations within Europe, including Netherlands, Germany, and Spain at 0.8 [33]. However, considerably lower ratios have been observed more frequently in U.S. locations (range, 0.3-0.7) [35], in Sweden and in the Canary Islands (0.4), Santiago, Chile (range, 0.40.6) [32], or Athens, Greece (range, 0.40.6) [32].

Pearson's Correlation Coefficients
The ln-transformed PM2.5 and PM10 across the sites were associated with coefficients between 0.826 and 0.963 based on the collocated measurements on given day ( Table 2). In particular, at each site (i.e., URBAN, TRAFFIC, INDU, and SUB2, respectively), the correlations between PM2.5 and PM10 ranged between 0.944 and 0.963.
Within URBAN, INDU, and SUB2 sites, one ln-unit increase in PM10 was able to explain 91%, 93%, and 91% of total variability in PM2.5, respectively (Table 3). In addition, one ln-unit increase in PM10 concentration predicted 92%, 100%, and 99% increase in PM2.5 in the same sites. In contrast, the same model for the TRAFFIC site was associated with lower accuracy (β = 0.74) and precision (adjusted R 2 = 0.60) in predicting PM2.5 concentration per same unit increase in PM10.

SO2
Clear, yet, overall consistent seasonal variability in the median SO2 concentration was observed across the sites in the 20002010 combined data ( Figure 4). As shown in Table 4, the mean winter SO2 concentration was approximately 3-times higher than that during the summer, except in SUB2. Spatial variability in SO2 concentration was examined by standardizing the concentration at given site by that at the CENTRAL site, collocated measurement on given date (Table 4). During summer, SO2 concentration was highest at the TRAFFIC site relative to the CENTRAL site (1.8 vs. 1.1 in all remaining sites). A same pattern for SO2 was again observed during the transition season with highest ratio for the TRAFFIC site (1.8) compared to the remaining sites (range, 0.81.2). However, during the winter, little differences were observed among the URBAN, TRAFFIC, INDU, and SUB1 sites (range, 0.91.3).
During the years 19681973, mean daily SO2 concentration in Krakow was 119 µg/m 3 with various sub-sections of the city reporting even higher mean annual concentration of SO2 (80120 µg/m 3 ) [3]. In contrast, the median concentration during the winter over the years 20002010 period (Table 4 and Figure 4) at the CENTRAL (20 µg/m 3 ) and the INDU (14 µg/m 3 ) sites reflect a reduction in SO2 contribution to ambient air pollution in Kraków.
Spatiotemporal variability in relative abundance of SO2 against PM10 was compared as annual mean SO2/PM10 ratio based on collocated 24-hour measurements ( Figure 5). Wide variability in SO2/PM10 ratio was observed between the sites at the onset of the study period. The three sites in the northeastern portion of the city (CENTRAL, URBAN, and TRAFFIC) are associated with the highest ratios (range, 0.50.6), while SUB1 and INDU site have ratios < 0.4. However, the rates at all sites, except URBAN, were associated with a uniform decline to 0.

NO2
In contrast to other pollutants, NO2 exhibited larger spatial heterogeneity in their median concentrations. Such heterogeneity was particularly apparent during summer (Tables 4 and 5, Figures 6 and 7). During the summer, the median NO2 at the TRAFFIC site was 3-times higher than that at the CENTRAL site (70 vs. 22 g/m 3 ). During all seasons, the median NO2 was lowest at the CENTRAL site compared to all other sites. This reflects the fact that the CENTRAL station sits within a square, protected from automobile traffic. In contrast, TRAFFIC was also the site in which the inverse trend was observed against the season. While all other sites were associated with an elevated median NO2 concentration during winter, the median NO2 concentration was highest during summer at the TRAFFIC site.
As shown in Figure 6, exceedance of the EU annual standard of 40 μg/m 3 was observed most frequently at the TRAFFIC site. Largest seasonal fluctuation for the median NO2 was observed at CENTRAL (Winter/summer = 1.5). During the winter, the median NO2 concentration in TRAFFIC was approximately 1.8-times higher than that at the CENTRAL (60 g/m 3 vs. 33 g/m 3 ) (Table 4).
Relative abundance of NO2 against PM10 was estimated as annual mean NO2/PM10 ratio based on collocated 24-hour measurements ( Figure 6). TRAFFIC was the only site for which NO2/PM10 ratio consistently remained greater than unity.  Furthermore, Table 5 shows the spatial variability in concentration ratios of PM10/PM10, considering the CENTRAL site in the denominator. Overall, there was little difference in PM10 concentration at URBAN, INDU and SUB1 sites, considering the PM10 concentration CENTRAL site as the reference. Regardless of season, the ratios of PM10 concentrations of given site, relative to CENTRAL site did not markedly differ from unity (range, 0.91.3). On the other hand, the median PM10 concentration at the TRAFFIC site was 50% (during summer and transition season) and 60% higher (during winter) than those at the CENTRAL site (range, 1.51.6). As shown in Table 5, spatial concentration ratios were highest at TRAFFIC for PM10, SO2, and NO2 regardless of season.

O3
Compared to the summer O3 concentration, the median level during transition and winter were 20% and 50%, respectively, of the summer level at the URBAN station. In SUB1 station, the median O3 decreased by 20% during transition, and by 40% during winter compared to the median during summer. Within SUB2 station, the median O3 decreased by 26% during the transition season, and by 67% during winter compared to the median during summer (Table 4 and Figure 8). Such levels remained well under the EU standard, 120 µg/m 3 , based on the daily 8-hour mean.  Table A1 show the effects of the site, year, month, season, day of the week, temperature, and wind speed on the pollutants. The mean predicted concentration of PM10, PM2.5, SO2, NO2, and O3 were 5.38 μg/m 3 , 5.61 μg/m 3 , 2.55 μg/m 3 , 4.31 μg/m 3 , and 3.19 μg/m 3 , respectively at the reference points (i.e., site CENTRAL, year 2010, summer, Saturday, wind speed ≥ 0.90 m/s, and temperature ≥ 14.60 °C).

Site Effect
Mean concentration difference at each site relative to CENTRAL is shown for the pollutants based on regression coefficients and standard error in Table A1. The mean concentrations of PM10 were 19% higher in TRAFFIC and 10% higher at INDU site, compared to the CENTRAL site. For PM2.5, CENTRAL, SUB1, and SUB2 represent the reference sites due to missing measurements in these sites. For PM2.5, the mean concentrations at URBAN, TRAFFIC, and INDU sites were not markedly different from the reference sites (impact points range, 1.001.14). Similarly, mean concentration at TRAFFIC was 13% higher than that at CENTRAL, while the SUB1 and SUB2 sites had mean SO2 were 9% and 7% lower than that at CENTRAL. Holding all other variables constant, the mean concentration of NO2 was 42% higher than that in CENTRAL.

Year Effect
As shown in Figure 9, yearly trend of PM10 and NO2 remained relatively constant over 20002010 period. The results of regression models (Table A1) and the concentration impact factor show that the mean concentrations of PM10 and NO2 remained overall constant throughout the monitoring period. Specifically, concentration impact factors for PM10 ranged between 0.88 and 1.03 over the period 2000-2010 or, differed from the reference point by 1% per year. The yearly effect of 2009 on PM2.5 shows that there was a 3% increase in mean concentration, after accounting for other variables, including temperature and wind speed. For NO2, the concentration impact factors remained near 0.99 throughout the monitoring period. In contrast, there was a dramatic decrease in annual mean SO2 concentration over the same period (Figure 8). Considering year 2010 as the reference point (impact point, 1), the impact factor of SO2 steadily decreased from 1.37 in 2000 to 0.90 in year 2009. The yearly trend of O3 was 20%23% lower than the reference year 2010. However, the impact factor increased by 7%25% in year 20022003 period. Subsequently, it leveled off towards unity in subsequent years. Such PM10 observation is consistent with a more recent analysis by Junninen et al. (2009), which has not shown a clear long-term PM trend in peak ambient levels for PM since our investigation during 20002002 period [4]. For example, during the winter of 2005, the peak ambient concentrations for PM with an aerodynamic diameter <10 µm (PM10) was 400 g/m 3 and peak ambient benzo[a]pyrene was 200 g/m 3 [4].

Month Effect
As shown in Figure 10, the effect of the month is strongly correlated with season for all pollutants of our interest. Considering December as the reference point (impact factor, 1) monthly concentration impact factors reach their lowest points during the May, June, July and August for PM2.5, PM10, SO2, and NO2. Specifically, July was associated with 14% decrease in mean NO2. When the temperature and wind speed variables were excluded from the regression models, the month of July was associated with 7% decrease in mean NO2. For O3, the effect of the month was in opposite direction. Between January and May period, the impact factor steadily increased from 1.18 to 1.57. Suring summer, the impact factor peaked between 1.52 and 1.61. It subsequently subsided from 1.35 to 0.98 during September to November period.

Weekday Effect
As shown in Figure 11, weekday played most visible role in NO2 and O3 concentrations, but not in PM2.5 and PM10 concentrations. During weekdays (MondayFriday), the mean concentration of NO2 increased approximately 6% compared to the reference day (Saturday). For O3, the same period was associated with impact factor decrease by 6% compared to the reference (Saturday).
Contrary to our expectation, our analysis demonstrates overall poor ambient air quality in Kraków, with little improvement during the 11-year period. Such a pattern reflects the complex interplay of the sources, valley setting, and meteorological factors [4]. Clear seasonal trends of PM10, PM2.5, SO2, and NO2 suggest the importance of the both coal-burning as well as traffic sources. The levels of PM10 and PM2.5 seen in this study reflect vast improvement in air quality of Kraków, compared to that during the Communist regime. For example, average annual concentration of PM10 changed from 154 µg/m 3 in 1993 to 49 µg/m 3 in 2007 [36]. During 1992-1999 period, ambient concentration of lead ranged between 0.006 and 0.434 μg/m 3 (near residential area); 0.016-0.739 μg/m 3 (near the industrial area); and 0.021-1.147 μg/m 3 near roads [20]. Recent analysis estimated >50% of PM10 in Kraków are contributed by coal burning for residential heating, and rest to automobile traffic and industrial power plants [4]. As recently as 2005, 24-hour mean concentration of airborne benzo[a]pyrene (B[a]P) at 200 ng/m 3 has been observed during winter [4]. Small domestic stoves/boilers for heating represent the primary contributors of airborne PM and polycyclic aromatics hydrocarbons during winter [4,5]. Krakow also receives air pollution from the Upper Silesia coal region [37]. However, the ambient levels of PAH and heavy metals continue to be high in the central section of the city, because of increasing traffic (especially diesel) and continued coal burning by industrial and residential sources [4]. In addition, transition to a market economy expanded the vehicular fleet in Kraków. Traffic density in the city center is estimated at 2500-3000 cars/hour between 7 am to 5 pm, and subsequently decreases to 200-500 cars/hour during the night [38]. In residential area, mean traffic is estimated at 50 cars/hour [38].
Our analysis suggests that NO2 represents a dominant species in TRAFFIC site. TRAFFIC was the only site for which NO2/PM10 ratio consistently remained greater than unity. In addition, the spatial concentration variability ratios (PM10/PM10, SO2/SO2, and NO2/NO2, considering CENTRAL concentration as the denominator) suggest that NO2 and other vehicular emission factors are considerably higher only at the TRAFFIC site.
Three sites in northwestern portion of the city (CENTRAL, URBAN, and TRAFFIC) are associated with the highest SO2/PM10 ratios (range, 0.50.6), while SUB1 and INDU site have SO2/PM10 ratios < 0.4. While CENTRAL, URBAN, and TRAFFIC in northwestern portion of the city comprises high pollution and southeastern section of (comprised of INDU, SUB1, and SUB2) had overall lower pollution level and higher seasonal fluctuation in all of the pollutants. Our present observation is consistent with growing emissions from the mobile sources [20] as well as rising secondary particle formation since 1989 [4].
While domestic coal-burning boilers and local heating facilities without an abatement strategy have been replaced by gas-burning boilers [3], it remains unclear how effective they are as remediation strategy. Overall steady reduction in coal-burning related emission has failed to produce corresponding decrease in a number of childhood morbidity outcomes [3,36]. Furthermore, the average prevalence of childhood asthma has increased by 9% during 19932003 period in Poland [36]. More effective strategies for air quality improvement are needed for the protection of the health of the population in Krakow.

Conclusions
Air quality in Krakow did not improve during the 20002010 period. Such a pattern is observed in spite of a dramatic decline in ambient SO2 concentrations over the 11-year period. The site-combined annual mean PM10 remained overall constant and considerably higher than the annual value reported for other urban background levels in other European cities. Both PM10 and PM2.5 exhibited clear season-dependent and site-specific variability in their mean concentration. Specifically, PM10 and PM2.5 concentrations due to vehicular emissions during a given season contributed to the largest spatial variability in their concentrations at the TRAFFIC site, relative to the reference site. Although the PM2.5 data were available only for years 20092010, annual mean concentrations of PM2.5 were approximately four-to five-times higher than the annual mean EU standard. Furthermore, PM2.5/PM10 ratio based on only 2-year long data suggest PM2.5 comprises major proportion PM10 concentration. This is of concern given the profound human health relevance of PM2.5 exposure. Furthermore, PM10 is associated with highly accurate (>92%) and precise (>91%) estimation of ambient PM2.5 concentration in all sites except the TRAFFIC site. In contrast, while no clear seasonal variability was seen for NO2, the median concentration was particularly elevated near the traffic sites. Median tropospheric ozone concentration remained well-below the EU standard value throughout the 20002010 period. During the years 20002010, the air quality of Kraków demonstrates an overall city-wide decline in ambient SO2 level, which is counterbalanced by the contributions of automobile traffic-related air pollution.

Acknowledgments
This work is supported by The National Institute of Environmental Health Sciences (NIEHS) (grant numbers 5 P01 ES009600, R01ES014939, 5 R01 ES008977, 5 R01ES11158, 5 R01 ES012468, 5 R01ES10165, and ES00002), the U.S. Environmental Protection Agency (EPA) (grant numbers R827027, 82860901, RD-832141), and the National Research Service Award (T32 ES 07069), and an anonymous Foundation. The authors thank Voivodship Inspectorate for Environmental Protection in Kraków for kindly sharing the data. The authors also thank Frederica Perera, Wiesław A. Jedrychowski, Renata Majewska, Elzbieta Mroz, Elzbieta Flak, Agata Sowa, and Ryszard Jacek for their respective contributions to study design, data collection, and management. Finally, the authors are indebted to the anonymous reviewers whose comments improved the quality of the analysis.

Author Contributions
Hyunok Choi conducted all statistical analyses. Steven Melly compiled multiple data, conducted geographic information analyses. John Spengler oversaw design, implementation, and statistical analysis.