**Benzo[a]pyrene in the Ambient Air in the Czech Republic: Emission Sources, Current and Long-Term Monitoring Analysis and Human Exposure**

**Markéta Schreiberová 1,\* , Leona Vlasáková 1 , Ondˇrej Vlˇcek <sup>1</sup> , Jana Šmejdíˇrová 1 , Jan Horálek <sup>1</sup> and Johannes Bieser <sup>2</sup>**


Received: 3 July 2020; Accepted: 28 August 2020; Published: 7 September 2020

**Abstract:** This paper provides a detailed, thorough analysis of air pollution by benzo[a]pyrene (BaP) in the Czech Republic. The Czech residential sector is responsible for more than 98.8% of BaP, based on the national emission inventory. According to the data from 48 sites of the National Air Quality Monitoring Network, the range of annual average concentration of BaP ranges from 0.4 ng·m−<sup>3</sup> at a rural regional station to 7.7 ng·m−<sup>3</sup> at an industrial station. Additionally, short-term campaign measurements in small settlements have recorded high values of daily benzo[a]pyrene concentrations (0.1–13.6 ng·m−<sup>3</sup> ) in winter months linked to local heating of household heating. The transboundary contribution to the annual average concentrations of BaP was estimated by the CAMx model to range from 46% to 70% over most of the country. However, the contribution of Czech sources can exceed 80% in residential heating hot spots. It is likely that the transboundary contribution to BaP concentrations was overestimated by a factor of 1.5 due to limitations of the modeling approach used. During the period of 2012–2018, 35–58% of the urban population in the Czech Republic were exposed to BaP concentrations above target. A significant decreasing trend, estimated by the Mann-Kendall test, was found for annual and winter BaP concentrations between 2008 and 2018.

**Keywords:** benzo(a)pyrene; ambient air concentrations; spatial-temporal; long-term trends; population exposure; transboundary transport; source apportionment

#### **1. Introduction**

Polycyclic aromatic hydrocarbons (PAHs) are ubiquitously distributed in the environment [1]. They are common by-products of combustion processes of fossil fuels and wood. PAHs represent a group of substances, many of which have toxic teratogenic, mutagenic or carcinogenic properties [2,3]. They affect fetal growth. Prenatal exposure to PAHs is related to markedly lower birth weight [2] and probably also has negative effects on the cognitive development of young children [4]. Due to their physical and chemical properties, all these substances can be transported over long distances and deposited in remote areas [5–7]. PAHs can bioaccumulate, enter the food chain [1] and be toxic to the environment.

Benzo(a)pyrene, occurring in the atmosphere primarily bound to particulate matter, has been set as a suitable marker of ambient air pollution caused by PAHs. A European directive has set a target value of 1 ng·m−<sup>3</sup> for the total content of BaP in the PM<sup>10</sup> fraction, averaged over a calendar year [8], with the aim of avoiding, preventing or reducing harmful effects on human health and/or the environment as a whole. The WHO has not drafted a guideline for BaP, which is a potent carcinogen. The reference level of 0.12 ng·m−<sup>3</sup> was estimated assuming the WHO unit risk for lung cancer for PAH mixtures and an acceptable risk of additional lifetime cancer of approximately 1 × 10−<sup>5</sup> [9,10].

Residential combustion as an important source of PAHs and other air pollutants are responsible for the majority of anthropogenic emissions of BaP in Europe. Such emissions are linked to adverse health effects, especially in urban and suburban areas where emissions and population densities are higher [11]. A modeling study of Europe [11] stated that it is necessary to assess concentrations of BaP in Europe, as an indicator for PAHs, and quantify their health-related effects. The European Environmental Agency estimates that in 2017, 17% of the EU−28 urban population was exposed to above-target annual mean BaP concentrations; this is the lowest value since 2008. As in previous years, values above 1 ng·m−<sup>3</sup> are predominantly found in Central and Eastern Europe. The highest concentrations were recorded mainly at stations in Poland and the Czech Republic [12].

Air pollution by BaP is one of the main problems associated with ensuring air quality in the Czech Republic. BaP concentrations exhibit significant intra-annual variation with maxima in winter that are related to emissions from seasonal anthropogenic sources (local heating units) and generally worsened dispersion conditions.

The aim of this study is to assess the current levels and long-term trends of air pollution by benzo(a)pyrene in the Czech Republic, together with their causes. A transboundary contribution to the annual mean concentrations of BaP is also quantified via a modeling-based approach.

#### **2. Materials and Methods**

#### *2.1. Area Description*

The Czech Republic is located in Central Europe. The topography of the Czech Republic consists predominantly of hills and highlands. More than two thirds of the territory is located below 500 m altitude, with the large majority of the rest between 500 m and 1000 m and only 1% above 1000 m.

The climate in the Czech Republic is mild; it could be classified as somewhere between continental and maritime. It has 4 seasons. Local variations of weather and climate are influenced by ruggedness and altitude. The average annual air temperature in the Czech Republic usually varies from 5.5 to 9.5 ◦C. The coldest months of the year are December, January and February. The hottest are July and August. Usually, precipitation in the Czech Republic is at a maximum in July and a minimum in February. Currently, about 60% of the population lives in cities with more than 5000 inhabitants [13,14].

#### *2.2. Sampling and Analytical Methods*

In the Czech Republic, concentrations of BaP in the PM<sup>10</sup> fraction as measured at manual monitoring stations form the basis for the evaluation of air quality. The monitoring stations are placed mainly in cities and in areas with known high BaP concentrations (Figure 1). In 2018, BaP concentrations were measured at 48 monitoring sites (Table A1). The majority of stations are located in cities, with 28 urban and suburban stations. Transport and industrial contributions to BaP are monitored at six traffic and nine industrial stations. Background levels of BaP concentrations are monitored at 5 rural monitoring stations.

PM<sup>10</sup> samples are taken by low or high volume samplers on a quartz filter. The samples are processed in certified chemical laboratories and analyzed by high pressure liquid chromatography (HPCL) or gas chromatography with mass detection (GC/MS). The measured concentrations of BaP are daily averaged value and are collected with a minimum three-day frequency. The concentrations measured at the pollution monitoring stations are stored in the Czech national Air Quality Information System (AQIS) database. The lower detection limit is 0.04 ng·m−<sup>3</sup> for GC/MS and 0.10 ng·m−<sup>3</sup> for HPLC. The measurement uncertainty of BaP is up to 25%.

**Figure 1.** Monitoring stations of benzo(a)pyrene in the Czech Republic in 2018.

Short-term monitoring of BaP concentrations in small settlements has been made in campaigns within the Czech national project TITSMZP704—Measurement and Analysis of Air Pollution with Emphasis on the Evaluation of the Share of Individual Groups of Sources. These are case studies that monitor the variability of short-term BaP concentrations during the heating season under the specific conditions of particular small settlements. As this is an ongoing project finishing in 2021, here we will present only a sample of data from 3 small settlements (Figure 1)—Bochovice, Cern ˇ íny and Hˇrivice—to show the level of BaP concentrations in villages where they are not regularly monitored and where solid fuel heating is predominant. Bochovice is a small village with 143 inhabitants with 58 houses, which are heated by solid fuels. In Cern ˇ íny there are 370 permanent residents and 140 houses out of 167 are heated by coal or wood. Hˇrivice is a village with 631 inhabitants, with 251 houses heated by coal or wood. We also present the Kladno-Švermov locality as an example of an area surrounding a current monitoring station measuring very high levels of air pollution caused by local heating. Kladno-Švermov is a district to the north of Kladno city situated in a shallow valley, where almost 5000 people live. It has a high building density with both central and local heating.

#### *2.3. Emission Calculation*

BaP is mainly a product of incomplete combustion of organic substances at temperatures between 300 and 600 ◦C. Unsurprisingly, the main contribution to total BaP emission in the Czech Republic is from combustion of solid fuels in low-capacity boilers. This is mainly household fuel combustion for heating, cooking and water heating (residential sector).

BaP emissions from the residential sector are calculated on the basis of emission factors for various combinations of fuel type and installed combustion plant (see Tables A2–A5 in Appendix B). The combinations used are set from the annually updated distribution of solid fuel and type of heating equipment (Table A5 in Appendix B), resulting in country specific emission factors which reflect the particular circumstances for a given year. For the purpose of national emission inventory, the total fuel consumption in households is determined by the Czech statistical office [13]. The emissions in this

model are calculated as a sum for the whole country and comprise local heating, cooking and water heating. This national model uses emission factors at nominal heat output.

For territorial distribution of residential heating emissions, a bottom-up up approach is used. The emissions are calculated for each basic territorial unit—municipalities and city districts—and comprise only the local heating from permanently occupied households. The fuel consumption at a basic territorial unit is calculated based on the average annual heating amount and specific fuel type consumption per average housing type. The base data are obtained from the 2011 Population Census (number of households, their type of heating and average floor area) and the results of the ENERGO 2015 statistical survey (share of given fuel burned in a particular installation type, share of insulated/noninsulated flats). The year-by-year changes in fuel consumption are mostly influenced by the characteristics of the heating period, which is expressed as the number of heating degree days (the sum of the differences between the reference indoor temperature and the average daily outdoor temperature on heating days). Other regional annually updated parameters are the number of households and their type of heating. Solid fuel parameters and the share of solid fuel consumption according to the installation type of the combustion plant are annually updated at a national level, based on the statistics of boiler sales ascertained by the Ministry of Industry and Trade and data from the subsidy program for the boiler replacements. Solid fuel parameters are updated according to the results of the survey on supplies and quality of solid fuels in the Czech Republic carried out by the TEKO company [15]. The regional calculation model uses a 15/85 boiler operating mode, i.e., 15% of time at nominal heat output and 85% at lower heat output. This assumption is in accordance with the Ecodesign Directive 2009/125/EC. Calculation model for the territorial distribution of emissions is more sensitive to climatic conditions in a given year than model for the national emission inventory. The highest emission difference between these two models is in 2014 and 2018, when the heating period was mild and short (see Table A6 in Appendix B).

Emission factors of solid fuels for local heating were obtained from measurement results of the most common fuel and boiler type combinations used for household heating in the Czech Republic. These measurements were carried out at the Energy Research Center of VSB-TUO [16] during the years 2008–2016. Emission factors for liquid and gas fuel were taken from the Emission inventory guidebook [17].

The transport sector comprises emissions from road transport, railways, air and water transport, off-road transport used in agriculture, forestry, building construction and area transport within large industrial enterprises. Emissions are calculated at the Transport Research Center [18] in an up-to-date version of the COPERT program [19] based on nationwide fuel consumption.

BaP emissions from combustion in nonresidential stationary sources such as power and heat generation, combustion processes in industry and manufacturing, institutions and services and waste incineration are calculated from activity data and given emission factors; national emission factors are estimated for a particular sector or taken from the Guidebook [17]. Emissions from industrial sources are either reported by the operator or calculated from activity data and emission factors.

#### *2.4. Spatial Mapping and Population Exposure*

The methodology used for the creation of the BaP concentration maps is a linear regression model followed by an interpolation of its residuals; rural and urban areas are mapped separately and then merged by population density [20]. The methodology is referred to as the Regression—Interpolation—Merging Mapping (RIMM) method and is used for air quality mapping in the Czech Republic and elsewhere in Europe as well [11,21,22]. The estimate of concentrations is calculated using the relationship:

$$\hat{Z}(s\_0) = c + a\_1 \cdot X\_1(s\_0) + a\_2 \cdot X\_2(s\_0) + \dots + a\_p \cdot X\_p(s\_0) + \hat{R}(s\_0),\tag{1}$$

where *Z*ˆ is the estimated concentration value at point *s*0, *X<sup>i</sup>* are the various supplementary data, *c* and *a<sup>i</sup>* are the parameters of the linear regression model and *R*ˆ is the spatial interpolation of the residuals of the linear regression model at point *s*0, calculated on the basis of the residuals at the points of measurement.

The primary data for creating air pollution maps of BaP are concentrations measured at individual monitoring stations. Since there are only a limited number of monitoring stations and their spatial representativeness is variable, various supplementary (secondary) data are also used. These secondary data both provide comprehensive information about the entire territory and also exhibit regression dependence on the measured data. The main secondary sources of information are outputs of dispersion models, which combine data from emission inventories and meteorological data. In the Czech Republic, the secondary data mainly used are annual mean concentrations provided by EMEP/MSC-E [23] together with annual mean concentrations from the Czech Gaussian model SYMOS. Other supplementary data can be provided by maps of annual mean PM<sup>10</sup> and PM2.5 concentrations.

The kriging and Inverse Distance Weighting (IDW) techniques are used as interpolation methods [24]. Interpolation of residuals using IDW is calculated using the relationship:

$$\hat{\mathcal{R}}(s\_0) = \frac{\sum\_{i=1}^{N} \frac{\mathcal{R}(s\_i)}{d\_{00}^{\beta}}}{\sum\_{i=1}^{N} \frac{1}{d\_{00}^{\beta}}} \, \text{} \tag{2}$$

where *R*ˆ is the estimate of the field of residuals at point *s0*, *R*(*s<sup>i</sup>* ) is the residual of the linear regression model at the measuring site *s<sup>i</sup>* , *N* is the number of surrounding stations used in the interpolation, *d*0*<sup>i</sup>* is the distance between points *s*<sup>0</sup> and *s<sup>i</sup>* , and β is the weight.

In case of ordinary kriging, the interpolation of the residuals is calculated using the relationship:

$$\hat{\mathcal{R}}(\mathbf{s}\_0) = \sum\_{l=1}^{N} \lambda\_l \cdot \mathcal{R}(\mathbf{s}\_l), \text{ with } \sum\_{l=1}^{N} \lambda\_l = 1,\tag{3}$$

where *R*(*s<sup>i</sup>* ) is the residual of the linear regression model at the measuring site *s<sup>i</sup>* and λ*<sup>i</sup>* are estimated weights based on the theory of spatial statistics [24] derived from a variogram fitted to an empirical variogram 2γν(*h*) of the field of residuals. The variogram expresses the dependence of the variability between points on the mutual distance between the points and the empirical variogram is calculated as follows:

$$2\gamma\_{\boldsymbol{v}}(h) = \frac{1}{n} \sum\_{i,j:d\_{ij}=h\pm\delta} \left(\mathcal{R}(s\_i) - \mathcal{R}\{s\_j\}\right)^2,\tag{4}$$

where *R* are the residuals at measuring points *s<sup>i</sup>* and *s<sup>j</sup> , dij* is the distance between points *s<sup>i</sup>* and *s<sup>j</sup>* ,

*n* is the number of pairs of stations *s<sup>i</sup>* and *s<sup>j</sup>* whose mutual distance is *h* ± δ, and δ is the tolerance. The calculated urban and rural map layers are subsequently merged by a layer of population density α:

$$\hat{Z}(\mathbf{s}\_0) = \begin{cases} \hat{Z}\_r(\mathbf{s}\_0), for \, a(\mathbf{s}\_0) \le \alpha\_1 \\ \frac{a\_2 - a(\mathbf{s}\_0)}{a\_2 - a\_1} \cdot \hat{Z}\_r(\mathbf{s}\_0) + \frac{a(\mathbf{s}\_0) - a\_1}{a\_2 - a\_1} \cdot \hat{Z}\_u(\mathbf{s}\_0), for \, a\_1 < a(\mathbf{s}\_0) < \alpha\_2 \\ \hat{Z}\_u(\mathbf{s}\_0)\_\prime \, for \, a(\mathbf{s}\_0) \ge \alpha\_2 \end{cases} \tag{5}$$

where *Z*ˆ is the final estimate of the concentration at point *s*0, *Z*ˆ *r , Z*ˆ *<sup>u</sup>* is the concentration for the rural or urban map layer, and α1, α<sup>2</sup> are the classification intervals corresponding to the population density. For the BaP concentration maps α<sup>1</sup> was set to 200 inhabitants per km<sup>2</sup> and α<sup>2</sup> was set to 1000 inhabitants per km<sup>2</sup> .

The entire concept of separate mapping of rural and urban pollution is based on the assumption that *Z*ˆ *<sup>r</sup>*(*s*0) ≤ *Z*ˆ *<sup>u</sup>*(*s*0) for BaP. For areas where this assumption is not fulfilled, a third layer created in a similar fashion to the urban and rural layers is used; this third layer is created using all the background stations without distinguishing between urban and rural stations.

*Atmosphere* **2020**, *11*, 955

The maps are constructed with a spatial resolution of 1 × 1 km<sup>2</sup> . The uncertainty of the map was assessed using the cross-validation method: concentration at the location of a measuring site is always estimated from other station data only, thus providing an objective estimate of the map quality away from measurement site locations. In this article, the uncertainty of the maps is expressed by the relative root-mean-square error (RRMSE):

$$\text{RRMSE} = \frac{\sqrt{\frac{1}{N} \sum\_{i=1}^{N} \left(\hat{Z}(s\_i) - Z(s\_i)\right)^2} \text{RMSE}}{\frac{1}{N} \sum\_{i=1}^{N} Z(s\_i)} \cdot 100,\tag{6}$$

where *Z* is the measured value of the concentration at point *s<sup>i</sup>* , *Z*ˆ is its estimate using cross-validation and *N* is the number of measuring stations. The spatial distribution of the uncertainty was not estimated. It should be noted that the cross-validation is applied only during the interpolation of residuals; parameters of linear regression are always estimated using all the stations. Therefore, the overall uncertainty of the maps is somewhat underestimated. The uncertainties (RRMSE) were calculated for each map layer separately and were up to 30% for urban and over 60% for rural areas. The higher uncertainty of rural areas is due to lack of measurements at rural regional stations and the absence of more extensive measurements in smaller settlements in the Czech Republic.

The annual mean BaP concentration maps 2012–2018 were prepared at CHMI during the annual air quality assessments. Estimation of population exposure to above-target BaP concentrations were calculated based on maps of BaP and population density data with resolution 1 × 1 km [13].

#### *2.5. Trend Analysis*

Trends of annual average BaP concentrations were analyzed at six selected monitoring stations in the Czech Republic (Figure 1). Five stations were classified as urban or suburban, the remaining station was classified as a rural regional station. Station selection was based on their classification and the quantity of data availability for trend analysis. We focus on data from urban and suburban monitoring sites since one of the aims of this study is to assess human exposure to BaP concentrations. For a comprehensive overview, the data from the Košetice rural regional site are also presented.

Trends for annual, winter (October–March) and summer (April–September) average concentrations are analyzed. The authors annually prepare average monthly concentrations of BaP for the "Air Pollution in the Czech Republic" yearbooks [22] (the newest report) and have partitioned this data into winter and summer periods. From April to September, average monthly BaP concentrations are usually below or just above the target value while for the rest of the year they very often exceed the target value.

Temporal trends, i.e., annual averages of BaP concentrations and emission, were analyzed using the nonparametric Mann-Kendall trend test with a level of significance of 0.05 [25,26]. This test is among the most widely used statistical methods for this kind of data [27–31] and is particularly useful since it tolerates missing values and the data need not conform to any particular distribution. Moreover, as only relative rather than absolute magnitudes of the data are used, this test is less sensitive towards incomplete data capture and special meteorological conditions leading to extreme values [32] that often affect air quality data.

If a linear trend is significant, the slope and severity of the trend is estimated by Sen's test [32]. For all stations, annual average concentration and emission were analyzed for the same 2008−2018 time period. There were no missing averaged annual data.

We used R-Studio software for statistical analyses [33]. The maps were created using the geographic information system ArcGIS by ESRI [34].

#### *2.6. Source Apportionment*

During the update of the National air quality plans (NAQP) in 2018, a transboundary contribution to annual mean concentrations of BaP in 2015 had to be quantified. It is known that polycyclic aromatic hydrocarbons including BaP undergo gas-particle partitioning and degradation in the atmosphere, but these processes are not fully understood and other processes, e.g., secondary organic aerosol coating can protect PAHs from ozone degradation during long range transport [35]. Due to the limited time for the update of NAQP and unavailability of ready-to-use models for PAHs long-range transport, a chemical transport model CAMx v5.41 [36] was adopted to account for BaP as a passive tracer. The limitations of this approach are discussed further in the text.

The CAMx model was run in two nested domains d01 and d02 with 14.1-km and 4.7-km resolution respectively (Figure 2). The transboundary contribution was estimated with a brute-force method. Sources outside the Czech Republic were set to zero. The spatial distribution of concentrations originating from Czech sources within the 4.7-km CAMx grid was determined by the Gaussian model SYMOS [37] at 0.5-km resolution:

$$\mathbf{C}\_{\mathbf{C}\mathbf{Z}\_{\text{-}scaled}}(i) = \mathbf{C}\_{\mathbf{C}\mathbf{Z}} \cdot \frac{\mathbf{S}(i)}{\sum\_{j=1}^{n} \mathbf{S}(j)/n'} \tag{7}$$

where *CCZ* is the contribution of Czech sources calculated by the CAMx model in a 4.7-km grid, *S*(*i*) is total contribution of Czech sources in subgrid point *i* calculated by the SYMOS model and n is number of subgrid points. Next, a relative contribution of sector *C* of Czech sources was determined:

$$P\_{\mathcal{C}}(i) = \frac{S\_{\mathcal{C}}(i)}{S(i)} \cdot \frac{\mathbb{C}\_{\text{CZ\\_scaled}}(i)}{\mathbb{C}\_{\text{nonCZ}} + \mathbb{C}\_{\text{CZ\\_scaled}}(i)} \cdot 100,\tag{8}$$

where *CnonCZ* is the contribution of sources outside of the Czech Republic calculated by the CAMx model in a 4.7-km grid, and *S*c(*i*) is the contribution of the Czech sources (sector *C* only) in subgrid point *i* calculated by the SYMOS model.

**Figure 2.** Annual benzo[a]pyrene (BaP) emissions for 2015 [kg·km<sup>2</sup> ] used in the CAMx model.

Meteorological inputs with a 1-h time step were derived from the assimilation cycle of the numerical weather prediction model ALADIN/CE version ALARO [38] operated at the CHMI at 4.7-km resolution and with 87 vertical levels. In the assimilation cycle, the analysis at 0, 6, 12 and 18 UTC was followed by a 6-h forecast. Analysis of upper-air parameters combines the driving model ARPEGE with mesoscale structures of the ALADIN model through DFI blending complemented by 3DVAR assimilation of observations [39]. Analysis of surface temperature and relative humidity is based on optimal interpolation and serves as an input to the Interaction Soil Biosphere Atmosphere (ISBA) scheme describing exchanges between the atmosphere and the land surface [40]. The 68 lowest ALADIN levels were aggregated into 26 CAMx levels, with the top of the lowest level at approximately 50 m and the highest level at approx. 10 km above ground.

High-resolution BaP emissions for the Czech Republic were taken from the calculation model for territorial distribution. For the Polish Silesian and Lesser Poland Vovoidships, BaP emissions were estimated by the ATMOTERM company within the LIFE-IP MAŁOPOLSKA project (LIFE14 IPE/PL/000021). For Slovakia, emissions from the SNAP 2 sector were taken from the national emission inventory. For other sectors in Slovakia and the rest of the modeling domain, a top-down emission inventory based on the TNO emission inventory [41] was used: emissions for the year 2015 were estimated by linear interpolation and then distributed to ten sectors following the SNAP nomenclature. Industrial emissions were allocated to sources registered in the European point source emission register E-PRTR [42] and remaining industrial emissions as well as emissions from area sources were distributed using the "industrial area" land cover class from the Corine Land Classification (CLC) database [43]. For the spatial distribution of residential heating, a combination of population density and urban area was used, assuming different fuel mixes in metropolitan and rural areas. Typically, rural areas exhibit a larger per capita emission of BaP due to the increased usage of coal and wood. A map of annual emissions used is shown in Figure 2.

The time distribution of the residential heating emissions in the Czech Republic and Silesian and Lesser Poland Vovoidships was based on temperature profiles, otherwise factors for month, day of the week and hour of the day were used [44,45]. For the vertical distribution of point source emissions from top-down inventory, typical point source parameters based on the analysis of the data from the Czech database were used for each SNAP category.

#### **3. Results**

#### *3.1. Emissions of Benzo(a)pyrene*

In 2018, the residential sector accounted for more than 98.8% of total Czech BaP emissions (15.56 t out of total 15.74 t). The remaining percentage was produced mainly by the transport sector (0.78%, 122 kg), especially passenger cars (0.48%, 76 kg). The industrial share of total BaP emissions was 0.27% (43 kg) and the share of institutions and services 0.10% (16 kg). The emissions from industrial sources are year-round, in contrast with local heating, especially in the Moravian-Silesian region. There is also a higher proportion of coal combustion in households in this region, which is reflected in a higher BaP emission load. The contributions from the different sources of BaP emissions have not changed significantly in the last 10 years. Almost 58% of BaP emissions from the residential sector are produced by combustion of fuel wood, and the consumption of wood is still increasing (Table A2 in Appendix B). In contrast, coal consumption is decreasing in recent years, which is reflected in its decreasing share of emissions (Figure 3). In 2018, approximately 15% of households were using solid fuels for local heating, and these households are responsible for more than 98% of total BaP emissions in the Czech Republic.

Figure 4 presents BaP emissions from combustion of solid fuel in the residential sector in 2018 sorted by combustion installation and fuel type. More than 50% of total BaP emissions from the residential sector are estimated to be from fuel combustion in over-fire boilers and about 30% from fuel combustion in fireplaces and stoves.

**Figure 3.** The development of total BaP emissions and source share, 2010–2018.

**Figure 4.** BaP emissions from combustion of solid fuel in residential sector in 2018 sorted by combustion installation and fuel type.

#### *3.2. Ambient Air Concentrations of BaP and Population Exposure*

The map of annual average concentration of BaP in 2018 is shown in Figure 5. Areas where BaP concentrations were higher than the target value of 1 ng·m−<sup>3</sup> (above-target) are indicated in red or brown in the figure. The thresholds correspond to the upper and lower assessment thresholds of 0.6 ng m−<sup>3</sup> and 0.4 ng m−<sup>3</sup> , the target value of 1.0 ng m−<sup>3</sup> set by EU legislation [8], and to 2.0 ng m−<sup>3</sup> to distinguish the most polluted areas in the Czech Republic. High values of BaP concentrations were estimated in the North East of the Czech Republic, referred to as the Ostrava region. Other contaminated areas include the Kladno district (West of Prague), areas of Prague and a number of smaller towns. Areas with above-target concentrations comprised 12.6% of the Czech territory in 2018. The lowest annual average concentrations of BaP were estimated to be in locations distant from emission sources and therefore free from direct exposure (i.e., natural mountain areas). The range of measured concentration of BaP in 2018 was from 0.4 ng·m−<sup>3</sup> at the Košetice rural station to 7.7 ng·m−<sup>3</sup> at the Ostrava-Radvanice industrial station.

**Figure 5.** Field of annual average concentration of benzo[a]pyrene in the Czech Republic, 2018.

Further annual mean BaP concentration maps from 2012 to 2017, which were used in the population exposure estimation, are listed in Appendix C. Based on comparison of population numbers living in areas with the different BaP concentrations, it can be stated that there is no marked trend in the period between 2012 and 2018 (Figure 6). The average value of the percentage of the population living in the above-target areas was 49.6% during the years 2012 to 2018. The lowest number of inhabitants living in the above-target areas was estimated in 2018 (35.5%). The highest number of inhabitants (57.9%) exposed to above-target concentrations of BaP was estimated for the years 2012 and 2017. Thirty-three percent of inhabitants on average in 2012 to 2018 lived in places with concentrations of 0.7–1.0 ng·m−<sup>3</sup> . On average, between 2012 and 2018, 7.4% of the population lived in areas with BaP concentration lower than 0.4 ng·m−<sup>3</sup> , with values ranging from 3.5% in 2013 to 11% in 2014.

Figure 7 shows selected measured daily concentrations of BaP in the winter seasons of 2017 and 2018 at three project locations (Bochovice, Cern ˇ íny and Hˇrivice) together with the data from the CHMI Kladno-Švermov station. In Kladno-Švermov, Bochovice, Cern ˇ íny and Hˇrivice, the average concentrations of BaP were 8.0 ± 5.7, 2.1 ± 2.1 ng·m−<sup>3</sup> , 2.2 ± 2.0 ng·m−<sup>3</sup> and 5.4 ± 3.2 ng·m−<sup>3</sup> respectively. The highest daily average concentration of BaP over the sampling campaign—24.5 ng·m−3—was observed in the Kladno-Švermov station whereas the lowest daily average concentration of BaP (0.1 ng·m−<sup>3</sup> ) was recorded in Cern ˇ íny. The limited amount data obtained by these campaign measurements, which were only obtained in winter, does not allow for calculation of annual average concentrations. Nevertheless, in Bochovice and Cern ˇ íny the target limit value (1 ng·m−<sup>3</sup> ) established by European legislation was exceeded on 59% and 54% of measurement days, respectively. In contrast, the average daily BaP concentrations monitored in Hˇrivice and Kladno-Švermov were below the target limit in only one case.

**Figure 6.** Population exposure to benzo[a]pyrene in the Czech Republic, 2012−2018.

**Figure 7.** Daily average BaP concentration in small settlements Cern ˇ íny, Bochovice and Hˇrivice and in the town of Kladno-Švermov in the Czech Republic, 2017–2018.

#### *3.3. BaP Concentration and Emission Trends*

Table 1 presents the BaP annual average concentrations during the study period 2008−2018. Annual average concentrations from 2008 to 2018 were analyzed at six selected sites (five urban and suburban stations, one rural regional station). The highest annual average value from all stations, which was 2.1 ng·m−<sup>3</sup> , was from 2008, with a range between 0.4 and 6 ng·m−<sup>3</sup> . When including only the five urban and suburban stations, the highest annual average value of 2.5 ng·m−<sup>3</sup> , with the same range, was also seen in 2008. The years with the lowest average BaP concentration of 1.6 ng·m−<sup>3</sup> were 2014−2018; the widest range was between 0.4 and 3.9 ng·m−<sup>3</sup> in 2015. When including only urban and suburban stations, the lowest annual average value of 1.8 ng·m−<sup>3</sup> was detected in 2016, 2017 and 2018; the widest range was between 0.7 and 3.7 ng·m−<sup>3</sup> in 2017. The average and median value for

all stations for the entire study period are 1.8 ng·m−<sup>3</sup> and 1.2 ng·m−<sup>3</sup> respectively. The average and median value for urban and suburban stations is 2 ng·m−<sup>3</sup> and 1.4 ng·m−<sup>3</sup> , respectively.


**Table 1.** Station characteristics, average concentrations of BaP (ng·m−<sup>3</sup> ) and emission of BaP for 2008−2018.

In contrast, the highest annual average value of 0.7 ng·m−<sup>3</sup> at the Košetice rural regional site, representing the wider area without the local emission, was measured in 2013. The lowest annual average value and the median of 0.4 ng·m−<sup>3</sup> was recorded for six years (2008, 2011, 2014−2016, 2018). The median value for the Košetice site is 0.5 ng·m−<sup>3</sup> .

With respect to the main emission source of BaP (Figure 3), i.e., local heating causing higher levels of BaP in ambient air, we assessed concentration trends specified for the winter (October–March) and summer (April–September) period and for the year as a whole (Figures 8–10). More detailed graphs presenting the course of annual, winter and summer concentrations using boxplots at each measuring station can be found in Appendix D, Figure A2.

Around 65% of annual average concentrations from suburban and urban stations during the study period were higher than the target value of 1 ng·m−<sup>3</sup> [8]. For the sake of completeness, this figure was around 96% for the winter periods and 7% for the summer period, respectively. Only one rural regional site registered no above-target concentration during the entire study period.

Comparing the annual average values for all stations between 2008 and 2018, there is a decrease of 27% in the BaP annual average concentration and a 24% decrease for winter and 50% decrease for summer average concentrations (Figures 8–10). For urban and suburban stations, the situation is very similar with a 28%, 25% and 49% decrease for the annual, winter and summer periods respectively (Figures 8–10).

**Figure 8.** Annual average concentrations of BaP at selected monitoring sites and emissions in the Czech Republic, 2008−2018.

**Figure 9.** Winter average concentrations of BaP at selected monitoring sites and emissions in the Czech Republic, 2008−2018.

**Figure 10.** Summer average concentrations of BaP at selected monitoring sites in the Czech Republic, 2008−2018.

The Mann–Kendall test was used to assess the monotonic trend of BaP at selected monitoring sites (Table 2). Concerning annual averages, significant decreasing trends were detected at two stations (Kladno-Švermov with *p* value = 0.01 and Sen's slope of −0.15 ng·m−<sup>3</sup> ·year−<sup>1</sup> and Ostrava-Poruba with *p* value < 0.005 and Sen's slope of −0.12 ng·m−<sup>3</sup> ·year−<sup>1</sup> ). Concerning winter averages, the significant decreasing trend was detected at the same stations—Kladno-Švermov with *p*–Value < 0.005 and Sen's slope of −0.28 ng·m−<sup>3</sup> ·year−<sup>1</sup> , Ostrava-Poruba with *p*–Value = 0.03 and Sen's slope of −0.21 ng·m−<sup>3</sup> ·year−<sup>1</sup> .

The average annual and average winter concentrations overall for all stations exhibited a significant decreasing trend. The decrease per year in the BaP concentration was equal to 0.05 ng·m−<sup>3</sup> for annual averages and 0.09 ng·m−<sup>3</sup> for winter averages, respectively. For urban and suburban stations, the annual and average winter concentrations showed a similar significant decrease is similar—0.06 ng·m−<sup>3</sup> for annual averages and 0.11 ng·m−<sup>3</sup> for winter averages, respectively. No clear trend was found for BaP summer concentrations.

In terms of total BaP emission for the Czech Republic, 2013 was the year with the highest annual mean value, which was 17.5 t. BaP emissions increased by more than 11% between 2009 and 2010. This increase was due to the implementation of new statistical data for hard coal consumption from 2010. In addition, the years 2010 and 2013 were characterized by long and cold heating season compared to other years. After 2013, BaP emissions had a decreasing trend supported by milder winter seasons (especially 2014 and 2018) but also by decreasing coal consumption and especially the replacement of high-emission boilers. Consequently, no significant trend (*p*–Value = 0.88) was found for BaP emission development (for more details, see Section 3.1). Moreover, no correlation (*p*–Value = 0.054) was found between emission and the concentrations from the Košetice rural regional site representing background levels in the Czech Republic.


**Table 2.** The Mann–Kendall test to assess the monotonic trend and Sen's slope assessment for BaP during 2008–2018.

\* Significant trend.

#### *3.4. Source Apportionment*

The transboundary contribution to annual average concentration, as estimated by the CAMx model in its 4.7-km resolution grid mode, is 46–70% for most (80%) of the territory of the Czech Republic, with a median value of 57% (Figure 11). The highest transboundary contribution is modeled in the relatively clean Western and Southern parts of the country and in the North-Eastern mountain regions (cf. Figure 5). When subgrid scaling is applied, the contribution of Czech sources can exceed 80% in residential heating hot spots (Figure 12). Three categories of Czech sources with a relative contribution to annual average concentration exceeding 10% were identified: residential heating—the absolutely dominant Czech source on most of the Czech territory; road transport—only in large cities Prague and Brno and in vicinity of major roads; and coke oven plants in the Ostrava agglomeration.

When mapped annual average BaP concentration is multiplied by the relative contribution of transboundary sources, regions in the North-East parts of the Czech Republic are still above 1 ng·m−<sup>3</sup> , which indicates that the target value cannot be reached without significant reduction of BaP emissions in Poland in hand with measures to mitigate Czech sources.

The reliability of source apportionment results depends of course heavily on the emission inventory and dispersion model used. As stated above, BaP was treated as passive pollutant in CAMx, which can probably lead to overestimation of long range transport due to neglect of its degradation. Nevertheless, emission inventory plays probably the more important role. At the beginning of our modeling effort in 2017, there was only a limited amount of European-scale BaP emission inventory available that could

be used in air quality models (authors were aware of [46]; gridded BaP emissions for 2015 were made available by EMEP in December 2017 and marked as unofficial data evaluation purposes only [47]). For this reason we used a top-down inventory based on [41].

**Figure 11.** Relative contribution of transboundary sources to annual average BaP concentration in 2015 (CAMx model, 4.7-km grid).

**Figure 12.** Relative contribution of Czech sources to annual average benzo[a]pyrene concentration in 2015 (CAMx model rescaled by the SYMOS model to 0.5-km grid).

Annual average BaP concentration modeled by the CAMx domain d02 was compared with measurements. Annual statistics at station locations were taken from Air Quality e-Reporting [48]. Statistics for the Czech stations were taken from the CHMI's Air Quality Information System, since there was an error in data provided to the AQ e-Reporting. Since annual statistic in AQ e-Reporting do not include information on station classification, all available data marked as valid and verified were used. As can be seen from Figures A3 and A4 and Table A7 in Appendix E, we get the best agreement with observations for the Czech stations (model/observation 0.3–3.6 with median 1.5). For Poland, values are generally underestimated (model/observation 0.1–1.5 with median 0.4), while for Germany and Austria the model largely overestimates measured values (model/observation 4–17.7 with median 5 for Germany and 1.9–26.1 with median 7.8 for Austria). This results in a clear South-West to North-East gradient in model bias. It seems reasonable to expect that the model bias will lead to overestimation of the transboundary contribution in the Southern part of the Czech Republic and to its underestimation in Northern parts (especially in the Ostrava agglomeration). To confirm this assumption, the transboundary contribution to annual average BaP concentrations was compared with data provided by EMEP/MSC-E [23] (transboundary contribution to annual mean concentration provided via personal communication with A. Gusev). For this purpose, CAMx results were aggregated to the EMEP 50-km grid. From Figure A5 we can see that over approximately one third of the Czech Republic the transboundary contribution estimated by this study and by EMEP does not differ by more than 5% (absolute difference). Compared to the EMEP results, the transboundary contribution in the South-central part of the Czech Republic is 10–20% higher, while in the Ostrava agglomeration it is 5–7% lower. Nevertheless it must be noted that EMEP results are based on EMEP/CEIP gridded emission for 2015 [47], where the total BaP emissions for the Czech Republic were estimated to be 8 t based on the Czech 2017 submission. This number was corrected to 16 t in resubmission in 2019. Therefore, the contribution of Czech sources must be underestimated approximately by a factor of two in EMEP results leading to overestimation of relative transboundary contributions by a factor of 1.1–1.6. From the text above it seems likely that modeling only the dispersion of BaP can lead to an overestimation of the transboundary contribution of BaP in the Czech Republic by factor of 1.5.

#### **4. Discussion**

Transport, industry and services combined do not contribute more than 2% to BaP emissions. The main source of BaP emissions in the Czech Republic is overwhelmingly local heating, especially the combustion of solid fuels in older types of boiler constructions with over-fire and under-fire type of burners (Figures 3 and 4). Compared to other European countries, the Czech Republic has the second highest average BaP emission per capita from the residential sector (1.5 g/(person·year)). The highest average emission is in Poland (1.6 g/(person·year)) whereas the EU average is 0.4 g/(person·year) [49].

The main reason for a high share of BaP emissions from local heating in the Czech Republic is specifically the combustion of solid fuel in older type of boilers (over-fire and under-fire). However, these types of boilers are being gradually replaced by low-emission boilers or by other types of heating. In 2018, the share of over-fire and under-fire boilers was estimated to be around 69%. These replacements are being accelerated by the legislative requirements of Act 201/2012 Coll. [50] that stipulates that after September 2022 only low-emission boilers meeting the parameters of at least 3rd boiler class (as defined in the EN 303-5:2012 [51]) can be in operation for solid fuel combustion in households. The replacement of old boilers was supported by a subsidy program between 2015 and 2019, with the aim of replacing up to 100,000 high-emission boilers. However, replacement of the boiler itself is not a guarantee of efficient emission reduction if the boiler is not operated properly in accordance with operating instructions.

The range of measured concentration of BaP was from 0.4 ng·m−<sup>3</sup> at a rural regional station to 7.7 ng·m−<sup>3</sup> at an industrial station in 2018. The area where BaP concentrations exceeded the target value was 12.6% in 2018 (Figure 5). The highest annual average concentrations of BaP have long been recorded throughout the entire Ostrava Region. In this region, there is the highest emission load due to a combination of local heating and the largest share of heavy industry in the Czech Republic. The transboundary contribution in the Ostrava region was estimated to be 40–60% but may be in fact somewhat lower due to model limitations. High BaP concentration values due to the effect of local heating systems have also been monitored in Kladno for a long time. Concentrations exceeding the target value for BaP also occur in Central Bohemia and in a number of municipalities. The values of

BaP concentration show that the ambient air concentrations of BaP are high in the Czech Republic in general, which is consistent with the model study of Europe [11], where the authors pointed out the ambient air concentrations of BaP to be substantially high in Central and Central Eastern Europe but also in some other European regions.

Relatively low levels of BaP have been recorded in large cities, like in the center of Helsinki [52] and in Porto [28]. In Porto, transport was identified as the main source of PAHs based on diagnostic ratios. In street canyons in Helsinki, the measured concentrations of BaP were at the same level as those in the urban background and clearly lower than those in suburban detached-house areas. These results indicate that local traffic has only a minor effect on BaP concentrations, compared with the corresponding effect of small-scale combustion. In the Czech Republic, transport is also a minor emission source of BaP (0.8%); nevertheless, levels of BaP concentrations in Prague were higher than in Porto and in street canyons in Helsinki. The higher BaP concentrations in Prague (Table 1) were caused by regional and long-range transport, especially in the center of the city in areas with a high proportion of remote central heating, where the main emission source of BaP is traffic. In suburban Prague, local heating was as important as it was in suburban Helsinki [52].

The lowest average annual concentrations are estimated at places distant from direct exposure to emission sources (natural mountain areas). The lowest measured BaP concentrations, ranging from 0.3 to 0.7 ng·m−<sup>3</sup> , have been recorded at the Košetice regional station. Nevertheless, these values are still relatively high (Table 1) and are above the WHO reference value (0.12 ng·m−<sup>3</sup> ). These relative high values of BaP concentrations at a regional background station pointed out the important role of regional and long range transport in the Czech Republic. The importance of regional transport is related to the high contribution of coal combustion to BaP emissions (about 30–40%) and the long lifetime of BaP derived from coal combustion in the atmosphere as found in this study [35].

Moreover, no correlation between background BaP concentration at Košetice and emission was found. The same (lowest) value of background BaP concentration in 2008, 2011, 2014 and 2016−2018 supports our conclusion on the combined effects of emission, meteorological and dispersion conditions and transboundary contributions. Based on the CHMI data, the last four years 2015−2018 of the study period are also years with the highest ratio of good meteorological and dispersion conditions and with BaP emissions decreasing after their peak in 2013. The year 2014 can be characterized by a milder winter season and the shortest heating season for many years [22]. In contrast, the beginning of the study period could be characterized by the greater presence of moderately poor and poor dispersion conditions [53] and the lowest BaP emissions in the study period (Table 1).

Since the higher BaP concentrations are a problem in urban and suburban areas due to local heating, we chose five urban and suburban sites in the Czech Republic for the current concentration level and trend assessment. The sixth rural regional site represents the background ambient air concentration in the Czech Republic.

The highest annual average value in urban and suburban sites was detected in 2008, the lowest annual average value of 1.8 ng·m−<sup>3</sup> was detected in 2016, 2017 and 2018 with the widest range, from 0.7 to 3.7 ng·m−<sup>3</sup> in 2017. Nevertheless, the lack of correlation of BaP concentration in the five urban and suburban sites and emissions for the whole Czech Republic (mainly from local heating, Figure 8) highlights the possible domination of local influences and the influence of meteorological and dispersion conditions.

We found a significant decreasing trend for average concentrations and winter average concentrations. The development of concentrations from selected stations in the Czech Republic between 2008 and 2014 is comparable to the development of PAHs and BaP concentrations presented in the study of [28] who analyzed data from two suburban sites in Porto between 2004 and 2014. A significant decreasing trend in the framework of their study was also found. A downward trend for all types of stations and at two thirds of total stations over the period 2007−2014 was also presented by EEA [54]. A significant decreasing trend was found at 22% of European rural and urban stations [54].

The decrease in concentrations in the Czech Republic is especially noticeable since 2014 highlighting the influence of milder winter seasons in 2014 and 2018, the prevailing good dispersion conditions in 2015−2018 and decreasing coal consumption in the last few years. The significant decrease at two particular stations (Ostrava-Poruba and Kladno-Švermov) that are among those with the highest concentrations in the Czech Republic (and of course those above target value BaP concentration for the whole study period) point also to the effect of improvements in local heating.

Similarly to other studies [26,55–57], the typical seasonal variation for the BaP concentration has been shown. The BaP concentrations for October–March are more than eight times higher than for April–September. For instance, Albuquerque et al. (2016) [28] found a December–January/June–August ratio of 5 for PAHs for the eleven year study in Porto. The reasons for this are generally known—i.e., seasonal sources as local heating, higher emissions from motor vehicles and less mixing in the atmosphere due to inversions [28,57,58]. During the warmer season, on the other hand, concentrations decrease due to unstable atmospheric conditions favorable towards dispersion, increased chemical and photochemical decomposition of PAHs due to higher levels of solar radiation and higher temperatures and of course also due to decreased emissions from anthropogenic sources [59–61].

High values of daily BaP concentrations in winter months associated with local household heating were also recorded during the 2017–2018 campaign measurements in the small settlements of Bochovice, Cern ˇ íny and Hˇrivice (Figure 7), where concentrations of BaP are not regularly monitored and where solid fuel heating predominates. The range of measured BaP concentrations was 0.1–8.0 ng·m−<sup>3</sup> in Cern ˇ íny, 0.2–9.8 ng·m−<sup>3</sup> in Bochovice, and 1.0–13.6 ng·m−<sup>3</sup> in Hˇrivice. In particular, measured BaP concentrations in the small settlement of Hˇrivice were as high as and on some days even higher than in Kladno–Švermov, where some of the highest concentrations of BaP in the Czech Republic were recorded. Every year, concentrations there reach high values and exceed the target value by almost four times. Such high levels of BaP concentrations are caused on the one hand by an extremely high density of buildings, which leads to a higher BaP emission density near the surface, and on the other hand by the fact that the town is located in a shallow valley, which leads to a reduction in dispersion of pollutants during cold days. The limited amount of winter season data obtained by the campaign measurements does not allow for a calculation of annual average concentrations. The target limit value of 1 ng·m−<sup>3</sup> established by European legislation was exceeded in Bochovice, Cern ˇ íny and Hˇrivice on 59%, 54% and almost 100% of measurement days, respectively. The measurements in these three small settlements with low–density population clearly indicates that emissions of BaP by local heating influence the short-term BaP concentration in the surroundings. Local meteorological conditions, orography of the populated area and regional and transboundary long-range transport are further factors which influence the ambient air concentration of BaP. In the Czech Republic, due to the rugged terrain, a number of settlements are located in valleys, where there may be a frequent deterioration of dispersion conditions and thus an increase in pollutant concentrations.

The reference level established for BaP by the WHO of 0.12 ng·m−<sup>3</sup> was exceeded at all monitoring sites each year. The target value of 1 ng·m−<sup>3</sup> is set by a European directive (EU, 2004) with the aim of avoiding, preventing, or reducing harmful effects on human health and/or the environment as a whole. During the period of 2012–2018, 35–58% of the urban population in the Czech Republic was found to be exposed to BaP concentrations exceeding the above mentioned target value. The lowest number of inhabitants living in the above-target areas was estimated to be in 2018. The highest number of inhabitants (58%) exposed to above-target concentrations of BaP was estimated for the years 2012 and 2017. On average, only 7% of the population lived in the areas with the lowest concentration of BaP between 2012 and 2018. BaP is carcinogenic to humans and has been considered a good indicator for the assessment of risk to human health associated with exposure from PAHs found in the environment. The individual carcinogenic potencies of PAH in relation to BaP can be expressed through the BaP equivalent concentrations (BaP eq.) and evaluation of BaP alone will probably underestimate the carcinogenic potential of the PAHs mixtures [28,62]. The uncertainty of the map is a result of the inadequate number of measurements at rural regional stations and the absence of more

extensive measurements in smaller settlements in the Czech Republic, where the air pollution by BaP would demonstrate the fundamental effect of local heating units. In addition, the maps are prepared with a resolution of 1 × 1 km and therefore cannot take into account the local fragmentation of the terrain, which in the case of settlements located in valleys affects the levels of pollutants [63]. Thus, assessment of the interannual changes in the territory affected and population exposed to above-limit concentrations of BaP will also be accompanied by a greater margin of error.

#### **5. Conclusions**

A complex analysis of air pollution from BaP in the Czech Republic was carried out. Ambient air BaP concentrations and their long-term trends were studied to assess the level of BaP in the Czech Republic. The calculated emissions of BaP and modeling of the transboundary contribution to the annual mean concentrations of BaP were quantified to present the causes of BaP air pollution.

The measured concentrations of BaP are high due to high emission load from the combination of local heating and heavy industry in the Czech Republic. The residential sector is responsible for more than 98.8% of BaP emissions.

Many people (50% on average) in the Czech Republic live in the area where the BaP concentrations are above the target value set by a European directive with the aim of avoiding, preventing, or reducing harmful effects on human health and/or the environment as a whole.

On the basis of the observations in small settlements described above, where BaP concentrations are not regularly monitored and where solid fuel heating predominates, it can be assumed that in small settlements, carcinogenic BaP levels may reach high levels in the short-term. Above-target values where BaP is not routinely measured can also be expected in similar municipalities with a high proportion of local heating using solid fuels. Consequently, the fraction of people living in the above-target value areas will be higher than presented.

Due to the high number of small municipalities and particularly due to the high costs for laboratory analyses and limited capacity of the laboratories, the number of measurement locations will always be limited, and therefore it is desirable to specify in more detail data on emissions and to provide substantial support to modeling.

The transboundary contribution to annual average concentration was estimated to be between 46% and 70% for most (80%) of the territory of the Czech Republic. The contribution of Czech sources can exceed 80% in residential heating hot spots. Results are nevertheless subject to limitations of the modeling approach used—it is likely that the transboundary contribution to BaP concentrations in the Czech Republic was overestimated by a factor of 1.5. Apart from residential heating, two other categories of Czech sources with relative contribution to annual local average concentration exceeding 10% were identified: road transport (large cities and vicinity of major roads) and coke oven plants in the Ostrava agglomeration.

The typical seasonal variation for the BaP concentrations has been shown. The BaP concentrations for selected air quality monitoring stations for October–March were more than eight times higher than for April–September. This is in line with the composition of emission sources in the Czech Republic, with the dominance of the local heating sector and with the different influence of meteorological and dispersion conditions in the colder and warmer parts of the year connected with the atmospheric stability and chemical properties of BaP.

We found significant decreasing trends for average concentrations and winter average concentrations. No correlation between background BaP concentration and emission was found. BaP concentration development in the Czech Republic has been influenced by the combined effect of total emission, meteorological and dispersion conditions and transboundary contributions.

The significant decrease at two particular stations belonging to those with the highest concentrations in the Czech Republic also point out the effect of improvements in local heating infrastructure. To conclude, even assuming generally good dispersion conditions and milder winter seasons in future, a significant reduction of BaP emissions is needed to reach the target value for BaP in the Czech Republic.

**Author Contributions:** Conceptualization, M.S. and L.V.; formal analysis, M.S., L.V., O.V. and J.Š.; writing—original draft preparation, M.S., L.V., O.V. and J.Š.; writing—review and editing M.S., L.V., O.V., J.Š., J.H., J.B.; visualization, M.S., L.V., O.V. and J.Š.; supervision, M.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** The presented BaP concentrations in small settlements Bochovice, Cern ˇ íny and Hˇrivice were measured within the project "TITSMZP704—Measurement and Analysis of Air Pollution with Emphasis on the Evaluation of the Share of Individual Groups of Sources" funded with the state support of the Technology Agency of the Czech Republic under the BETA2 Programme. BaP emissions for regions outside the Czech Republic used for dispersion modeling were prepared within the project LIFE-IP MAŁOPOLSKA—Implementation of Air Quality Plan for Małopolska Region—Małopolska in Healthy Atmosphere (LIFE14 IPE/PL/000021).

**Acknowledgments:** The authors want to thank all colleagues from CHMI who participate in the measurement and processing of air quality data. We would also like to thank Alexey Gusev and Victor Shatalov from EMEP/MSC-E for providing BaP modeling results and for commenting on the text.

**Conflicts of Interest:** The authors declare that there is no conflict of interest regarding the publication of this paper.

#### **Appendix A**

**Table A1.** Specification of monitoring stations of BaP with annual average concentration in 2018.



**Table A1.** *Cont.*

NA—not available data for annual average calculation.

#### **Appendix B**

**Table A2.** Fuel consumption in residential sector 2008–2018.


Source: CZSO [13].

**Table A3.** Emission factors of BaP for local heating at nominal heat output.



**Table A4.** Emission factors of BaP for local heating at lower heat output.

**Table A5.** Distribution of solid fuel consumption according to the type of heating equipment in 2018.


**Table A6.** Comparison of BaP emissions from residential sector calculated in top-down and bottom-up model 2008–2018.


#### **Appendix C**

**Figure A1.** Field of annual average concentration of benzo[a]pyrene in the Czech Republic, 2012–2017.

#### **Appendix D**

**Figure A2.** Boxplots of BaP concentrations at selected monitoring sites, 2008–2018.

#### **Appendix E**

**Figure A3.** Annual average benzo[a]pyrene concentration in 2015 (CAMx model) and ratio of measured and modeled concentration at station locations.

**Figure A4.** Scatter plot of modeled and measured annual average benzo[a]pyrene concentrations in 2015 for CAMx domain d02. Solid line denotes ideal model equal to observation, dashed lines mark area, where model values are within a factor of two from observations.


**Table A7.** Ratio of modeled and measured annual average benzo[a]pyrene concentration.

**Figure A5.** Relative contribution of transboundary sources to annual average benzo[a]pyrene concentration in 2015: (**a**) CAMx model averaged on EMEP grid, (**b**) EMEP model, and (**c**) difference CAMx−EMEP.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **High NO<sup>2</sup> Concentrations Measured by Passive Samplers in Czech Cities: Unresolved Aftermath of Dieselgate?**

**Michal Vojtisek-Lom 1,2,\* , Miroslav Suta <sup>3</sup> , Jitka Sikorova <sup>1</sup> and Radim J. Sram <sup>1</sup>**


**Abstract:** This work examines the effects of two problematic trends in diesel passenger car emissions increasing NO2/NO<sup>x</sup> ratio by conversion of NO into NO<sup>2</sup> in catalysts and a disparity between the emission limit and the actual emissions in everyday driving—on ambient air quality in Prague. NO<sup>2</sup> concentrations were measured by 104 membrane-closed Palmes passive samplers at 65 locations in Prague in March–April and September–October of 2019. NO<sup>2</sup> concentrations measured by city stations during those periods were comparable with the average values during 2016–2019. The average measured NO<sup>2</sup> concentrations at the selected locations, after correcting for the 18.5% positive bias of samplers co-located with a monitoring station, were 36 µg/m<sup>3</sup> (range 16–69 µg/m<sup>3</sup> , median 35 µg/m<sup>3</sup> ), with the EU annual limit of 40 µg/m<sup>3</sup> exceeded at 32% of locations. The NO<sup>2</sup> concentrations have correlated well (R<sup>2</sup> = 0.76) with the 2019 average daily vehicle counts, corrected for additional emissions due to uphill travel and intersections. In addition to expected "hot-spots" at busy intersections in the city center, new ones were identified, i.e., along a six-lane road V Holešoviˇckách. Comparison of data from six monitoring stations during 15 March–30 April 2020 travel restrictions with the same period in 2016–2019 revealed an overall reduction of NO<sup>2</sup> and even a larger reduction of NO. The spatial analysis of data from passive samplers and time analysis of data during the travel restrictions both demonstrate a consistent positive correlation between traffic intensity and NO<sup>2</sup> concentrations along/near the travel path. The slow pace of NO<sup>2</sup> reductions in Prague suggests that stricter vehicle NOx emission limits, introduced in the last decade or two, have so far failed to sufficiently reduce the ambient NO<sup>2</sup> concentrations, and there is no clear sign of remedy of Dieselgate NOx excess emissions.

**Keywords:** NO<sup>2</sup> ; passive sampler; Dieselgate; Prague; traffic volume; citizen science; air quality; public policy; health effects

#### **Highlights**


#### **1. Introduction**

Mobile sources, including on-road vehicles, remain to be one of the largest contributors to the air pollution in most metropolitan areas in Europe, with particulate matter and

**Citation:** Vojtisek-Lom, M.; Suta, M.; Sikorova, J.; Sram, R.J. High NO<sup>2</sup> Concentrations Measured by Passive Samplers in Czech Cities: Unresolved Aftermath of Dieselgate? *Atmosphere* **2021**, *12*, 649. https://doi.org/ 10.3390/atmos12050649

Academic Editor: Iva H ˚unová

Received: 1 April 2021 Accepted: 16 May 2021 Published: 19 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

nitrogen oxides (NOx, defined as a sum of nitric oxide NO and nitrogen dioxide NO2) being of highest concern. Outdoor air pollution is now being considered one of the leading causes of premature death [1], with estimated tolls of approximately half a million premature deaths annually in the EU [2], and associated economic damage around 5% of HDP in Central Europe [3]. At the same time, the state-of-the art technology of the internal combustion engine has improved considerably over the last decades. Very low levels of sulfur and metals in the fuel have allowed the introduction of three-way catalysts on spark ignition engines, a common technology used throughout the U.S. over the last four decades with a somewhat delayed deployment in Europe, and the introduction of diesel particle filters on virtually all on-road diesel engines manufactured in the last decade. The emissions of nitrogen oxides, primarily NO, on engines operating with excess air remained a challenge, being ultimately resolved about a decade ago with selective catalytic reduction (SCR) systems on heavy-duty vehicles [4] and more recently also on light-duty vehicles.

In the EU, the concentrations of NO2, deemed to be more detrimental to human health than NO, are limited and monitored in the ambient air. Overall, the concentrations of NO<sup>2</sup> have not been decreasing as fast as those of other key pollutants. In the Czech Republic, the concentrations of NO<sup>2</sup> at most air quality monitoring stations have been, according to the data in [5], decreasing by on the order of 1% a year over the last two decades. A gradual decrease of NO<sup>2</sup> concentrations in the overall atmosphere above the Czech Republic over the last decade has been also reported from remote sensing satellite measurements [6].

NO<sup>2</sup> in ambient air originates both from direct (primary) emissions and from gradual conversion of NO into NO<sup>2</sup> [7]. While the total emissions of NO<sup>x</sup> have been gradually decreasing, there is no apparent trend of a decrease in NO<sup>2</sup> primary emissions over the last 15 years [6]. One of the culprits of high primary NO<sup>2</sup> emissions are diesel vehicles, which have been, over the last two decades, equipped with oxidation catalysts, which convert a considerable portion of NO into NO2. In the U.S., average NO2/NO<sup>x</sup> ratio in vehicle exhaust (all vehicles, including predominantly gasoline cars and light trucks and predominantly diesel heavy trucks) was 5.3% [8], compared to approximately 15% in Europe [9].

This paper explores a hypothesis that the observed decrease in NO<sup>2</sup> concentrations falls short of that expected based on order-of-magnitude decrease in vehicle NO<sup>x</sup> emissions limits and that non-compliant diesel cars could substantially contribute to this shortfall. The underlying aspects of NO<sup>x</sup> emissions and the adverse health effects of NO<sup>2</sup> are summarized. The results of a monitoring NO<sup>2</sup> with passive samplers are reported and discussed in light of these findings. As an additional insight, the effects of coronavirus related restrictions on NO and NO<sup>2</sup> concentrations in Prague are reported and discussed.

#### **2. Review of Trends and Shortcomings in NO<sup>2</sup> and NO<sup>x</sup> Emissions from Vehicles**

Nitrogen oxide (NO) is formed in combustion processes from atmospheric nitrogen and oxygen at high temperatures [10,11], which are generally associated both with efficient combustion and with high thermal efficiency of the engine. Subsequent oxidation of NO in the atmosphere yields primarily nitrogen dioxide (NO2), a brownish irritant gas. Other oxides of nitrogen—N2O2, N2O3, N2O4, N2O5—are generated in small concentrations, are unstable and short-lived in the atmosphere. The oxides of nitrogen are summarily referred to as NOx, although there is no precise definition. Often, NO<sup>x</sup> is evaluated as the sum of NO and NO2. Technically, the sum of NO<sup>x</sup> also includes nitrous oxide (N2O), which is, however, not hazardous to human health, but is a potent greenhouse. NOx leads to the formation of nitrous acid (HNO2) [12,13], nitric acid (HNO3) and a variety of salts such as ammonium nitrate, present in the atmosphere as particulate matter [14]. Photodissociation of NO<sup>2</sup> under the presence of sunlight produces NO and atomic oxygen, which reacts with molecular oxygen to form ozone [15], a highly reactive compound generally harmful to human health, organisms and plants. NO<sup>x</sup> and ground-level (tropospheric) ozone are, together with particulate matter, the principal part of urban air pollution.

On spark ignition engines, CO and VOC, principally a product of incomplete oxidation of fuel and to a lesser extent engine lubricating oil, and NO<sup>x</sup> have been successfully abated by the combination of three-way catalysts [16] and by maintaining stoichiometric air–fuel ratio through closed-loop control of the quantity of fuel injected [17]. This technology has proven to be remarkably efficient.

On diesel engines, the emissions of NO<sup>x</sup> have been, at first, controlled through delayed combustion timing and exhaust gas recirculation, both associated with a slight fuel penalty, and at a later time, with NO<sup>x</sup> storage and reduction catalysts and selective reduction catalysts (SCR). The reduction of NO<sup>x</sup> has historically come at an expense of both capital and operating costs, with operating costs including either fuel (notably on older vehicles using delayed combustion, exhaust gas recirculation, NO<sup>x</sup> storage and reduction catalysts) or a reducing agent used in SCR (mostly aqueous solution of urea, known as diesel exhaust fluid or "AdBlue"). These costs have motivated, over the last few decades, many manufacturers and vehicle users to circumvent NO<sup>x</sup> reduction efforts, as the savings were realized by them directly, while considerably larger overall damage to human health was born by the society, a problem known as the Tragedy of the Commons [18]. A widespread practice of dual engine mapping in the U.S. in the 1990s [19,20] has led to the gradual extension of vehicle emissions limits to ordinary on-road operation first of heavy-duty and later of light-duty vehicles [21–23]. In the heavy-duty vehicle engine sector, many recent studies now show that on-road NO<sup>x</sup> emissions of newer heavy-duty vehicles have been successfully reduced by an order of magnitude except for low-load operation typical for congested urban areas. Quiros et al. [24] reports NO<sup>x</sup> emissions of 2013 and 2014 model year heavy trucks of 0.36 g/km during motorway operation in California. Jiang et al. [25] reports, for similar conditions, 0.3 g/km NO<sup>x</sup> during extraurban and motorway operation. Grigoratos et al. [26] reports NO<sup>x</sup> emissions during motorway operation in Europe of 0.07, 0.08, 0.17 and 0.24 g/kWh for four trucks and 0.80 g/kWh for a bus. Giechaskiel et al. [22] reports NO<sup>x</sup> emissions of a garbage collection truck of less than 0.4 g/kWh during extraurban operation (note: for heavy vehicles, emissions per kWh roughly correspond to emissions per km).

Unfortunately, this has not been the case with light-duty vehicles with diesel engines, highly prevalent in Europe, where they account for several tens of percent of vehicle registration and in Prague, for about two thirds of vehicles counted on the road [27]. Large portion of European automobile diesel engines produced over the last one to two decades have been reported to emit substantially, often by an order of magnitude, more NO<sup>x</sup> on the road than during the type approval test [28–32]. Weiss et al. [29] reports on-road NOx emissions factors 0.76 ± 0.12 g/km for Euro 4, 0.71 ± 0.30 g/km for Euro 5 and 0.21 ± 0.09 for Euro 6. In a more recent study by Suarez-Bertoa et al. [23], NO<sup>x</sup> emissions from Euro 6 diesel cars varied substantially from mid tens to mid hundreds of milligrams of NO<sup>x</sup> per kilometer, with a median value of about 0.2 g/km NO<sup>x</sup> during the city-motorway test.

At the same time, on nearly all light-vehicle diesel engines of the last decade or so, oxidation catalysts are used to convert NO into NO2, as higher concentrations of NO2, around 10%, are beneficial both for the combustion of soot in DPF and for the "fast" reduction of NO<sup>x</sup> in SCR catalysts. As a result, NO<sup>2</sup> from newer engines accounts for 10% of NO<sup>x</sup> [33,34]. On passenger cars and light-duty trucks, NO2/NO<sup>x</sup> ratios of around 10–15% up to Euro 3 and 25–30% for Euro 4 and 5 were found in a London remote sensing study [35]. In the U.S., NO2/NO<sup>x</sup> ratio from heavy duty diesel trucks have doubled from around 7% in 2010 (average of trucks passing on the road in a given year, not a model year of the vehicles) to around 15% in 2018 [36]. This increase, however, did not result in an absolute increase in NO<sup>2</sup> emissions, as total NOx emissions have decreased dramatically due to the widespread use of SCR catalysts. According to Preble [36], "Fleet-average NO<sup>2</sup> emission rates remained about the same, despite the intentional oxidation of engine-out NO to NO<sup>2</sup> in DPF systems, due to the effectiveness of SCR systems in reducing NO<sup>x</sup> emissions and mitigating the DPF-related increase in primary NO<sup>2</sup> emissions".

In Europe, NO<sup>x</sup> emissions from diesel cars have not, however, decreased in proportion to the decreasing emissions limits. A recent on-road study in Prague reports the mean emissions of Euro 5 and 6 diesel cars and vans of over 0.1 g/km NO<sup>2</sup> and over 0.5 g/km NO<sup>x</sup> [37], while a recent study of one of the most common diesel cars (Euro 6) reported about 0.15 g/km over WLTC cycle, and about 0.4 g/km over the Artemis driving cycle [38], which is more than the 0.08 g/km Euro 6 limit for total NO<sup>x</sup> (with which the vehicle reasonably complied over the NEDC cycle).

The presumption of the regulators that increased the NO2/NO<sup>x</sup> ratio after the oxidation catalyst and before the DPF, highly beneficial both for DPF and SCR operation, will be mitigated by the rather high efficiency of the NO<sup>x</sup> aftertreatment, envisioned in both U.S. EPA and EU emissions standards, which has been compromised by intentional acts resulting in diminished, or even zero, efficiency of the NO<sup>x</sup> aftertreatment. Examples of such acts include dual-mapping of the engines by the manufacturers (a prime example of which is "Dieselgate") and disabling of the SCR (and emulating its proper functioning to the on-board diagnostics by "SCR emulators") by vehicle operators. Under such conditions, relatively high amounts of NO2, intended to be reduced in NO<sup>x</sup> aftertreatment, are emitted out of the tailpipe. Logically, this results in very high, and much higher than intended, primary emissions of NO<sup>2</sup> in the streets. This finding is consistent with the rather slow decrease in NO<sup>2</sup> concentrations.

#### **3. Review of the Impact of NO<sup>2</sup> to Central Nervous System in Children and Adults**

The first experimental data were obtained several decades ago, indicating that air pollution may induce behavioral changes. Singh [39] studied the effect of NO<sup>2</sup> exposure on pregnant mice, exposed during gestation day 7–18. Prenatal exposure significantly altered the righting reflex and aerial righting score. These results suggest that maternal NO<sup>2</sup> exposure produce deficits in the functional capability of the offspring.

Wang et al. [40] was the first one, who studied the impact of NO<sup>2</sup> exposure to children's neurobehavioral changes. They studied this effect in the year 2005 on two groups of children (A *N* = 431, B *N* = 430) in the age of 8–10 years using neurobehavioral testing. Group A was exposed to 7 µg NO2/m<sup>3</sup> , group B to 36 µg NO2/m<sup>3</sup> . Children from the polluted area showed poor performance in all tests: visual simple reaction time, continuous performance, digit symbol, pursuit aiming and sign register, This study found a significant relationship between chronic low-level traffic related air pollution and neurobehavioral function in exposed children.

Guxens et al. [41] analyzed the association between prenatal exposure, diet and infant mental development in four regions in Spain, in 1889 children, who were exposed to 29.0 <sup>±</sup> 11.2 <sup>µ</sup>g NO2/m<sup>3</sup> (20.1–36.8). Infant mental development was evaluated at 14 months by Bailey Scales of Mental Development. Exposure to NO<sup>2</sup> did not show a significant association with mental development. Inverse association was observed in infants whose mothers reported low intake of fruit/vegetables during pregnancy (−4.13 (−7.06, −1.21)). This study suggests that antioxidants in fruits and vegetables during pregnancy may modulate an adverse effect of NO<sup>2</sup> on infants' mental development.

Kim et al. [42] investigated the association between maternal exposure to NO<sup>2</sup> of 49.4 µg/m<sup>3</sup> (25.9–84.8) and neurodevelopment in children in Korea (mental development index (MDI) and the psychomotor development index (PDI) by Bailey scales of mental development) at ages 6, 12 and 24 months. This study used 455–371 children. NO<sup>2</sup> exposure impaired psychomotor development (β = − 1.30; *p* = 0.05). At 6 months NO<sup>2</sup> affected MDI (β = − 3.12; *p* < 0.001) and PDI (β = − 3.01; *p* < 0.001). These data suggest that exposure to NO<sup>2</sup> may delay neurodevelopment in early childhood.

A similar study was organized in Spain on 438 mother-child pairs by Lertxundi et al. [43] at 15 months of age, using the Bailey scales of mental development. A 1 µg NO2/m<sup>3</sup> increase during pregnancy decreased the mental score (β = −0.29; 90% CI: −0.47; −0.11). Prenatal residential exposure to NO<sup>2</sup> adversely affects infant motor and cognitive development.

A prospective cohort study was conducted with 2715 children aged 7–10 years in Barcelona, Spain, as a part of the BREATHE project (brain development and air pollution ultrafine particles in school children [44]). Children were tested every 3 months with a computerized test. Cognitive development was assessed with the n-back and the attentional network test as working memory and inattentiveness. NO<sup>2</sup> exposure was completed in the outdoors in a low traffic region 40.5 <sup>±</sup> 9.6 <sup>µ</sup>g/m<sup>3</sup> and high traffic region 56.1 <sup>±</sup> 11.5 <sup>µ</sup>g/m<sup>3</sup> . Children attending schools with higher NO<sup>2</sup> pollution had an 11.5% (95% CI 8.9%–12.5%) slower working memory and slower growth in all cognitive measurements, which means a smaller improvement in cognitive development.

Pujol et al. [45] selected from this cohort 263 children, aged 8–12 years, for magnetic resonance investigation (MRI) to analyze brain volumes, tissue composition, myelination, cortical thickness, neural tract architecture, membrane metabolites and functional connectivity. Outdoor NO<sup>2</sup> exposure was 46.8 <sup>±</sup> 12.0 <sup>µ</sup>g/m3/year and indoor NO<sup>2</sup> exposure was 29.4 <sup>±</sup> 11.7 <sup>µ</sup>g/m3/year. Higher NO<sup>2</sup> exposure was associated with slower brain maturation with changes specifically concerning the functional domain.

Forns et al. [46] evaluated 2897 children from the Barcelona cohort within the BREATHE project. NO<sup>2</sup> exposure in schools was 29.82 µg/m<sup>3</sup> (11.47–65.65) and outdoor was 48.46 µg/m<sup>3</sup> (25.92–84.55). Behavioral development was assessed using the strengths and difficulties questionnaire (SDQ), which was filled out by parents. NO<sup>2</sup> exposure was positively associated with SDQ total difficulties scores, suggesting more frequent behavioral problems. This study was understood as the first one to evaluate the impact of air pollution on behavioral development in schoolchildren using both indoor and outdoor air pollution levels measured at schools. NO<sup>2</sup> outdoor levels (IQR = 22.26 µg/m<sup>3</sup> ) significantly increased total difficulties score (1.07, 95% CI: 1.01, 1.14, *p* < 0.05). NO<sup>2</sup> exposure at school is associated with worse general behavioral development in schoolchildren.

Min and Min [47] studied in Korea 8936 children born in the year 2002 and followed them for the next 10 years, investigating the relationship between exposure to NO<sup>2</sup> and attention-deficit hyperactive disorder (ADHD). They diagnosed 313 children with ADHD. The hazard ratio (HR) associated with the increase in 1 µg of the NO2/m<sup>3</sup> was 1.03 (95% CI: 1.02–1.04). Comparing infants with lowest tertile of NO<sup>2</sup> exposure with the highest tertile of NO2, HR = 2.10 (95% CI: 1.54–2.85), exposure had a 2 fold increased risk of ADHD. The study showed a significant association between exposure to NO<sup>2</sup> and the incidence of ADHD in children.

Sentis et al. [48] evaluated prenatal and postnatal exposure to NO<sup>2</sup> and attentional function in children at 4–5 years of age in four regions of Spain (*N* = 1298). The attentional function was evaluated by the Conners kiddie continuous performance test (K-CPT). The prenatal NO<sup>2</sup> level was 31.1 µg/m<sup>3</sup> (18.4–37.9). Higher exposure to prenatal levels of NO<sup>2</sup> was associated with a 1.12 ms (95% CI; 0.22, 2.02) increase in hit reaction time and 6% increase in the number of emission errors (95% CI: 1.01, 1.11) per 10 ug/m<sup>3</sup> increase in prenatal NO2. Higher exposure to NO<sup>2</sup> during pregnancy is associated with impaired attentional function, especially increased inattentiveness in children aged 4–5 years. This reduced attentional function in population could lead to poor educational indicators. It seems to be important that this effect was observed with NO<sup>2</sup> concentrations lower than EU standard 40 µg/m<sup>3</sup> .

Sunyer et al. [49] followed in 2012–2013 2687 school children from Barcelona, assessing children´s attention process 4 times every three months, using the attention network test (ANT). NO<sup>2</sup> indoor pollution was 30.09 <sup>±</sup> 9.51 <sup>µ</sup>g/m<sup>3</sup> and ambient air pollution was 37.75 <sup>±</sup> 18.41 <sup>µ</sup>g/m<sup>3</sup> . Daily ambient levels were negatively associated with all attention processes (children in the bottom quartile of daily exposure to NO<sup>2</sup> had a 14.8 ms (95% CI: 11.2, 18.4) faster response time than those in the top quartile, which corresponds to a 1.1 month delay (95% CI: 0.84, 1.37) in natural development). Short-term exposure to NO<sup>2</sup> is associated with potential harmful effects on neurodevelopment.

Forns et al. [50] examined after 3.5 years the cohort of children from Barcelona (*N =* 1439), whose cognitive development was evaluated 4 times in the years 2012/2913 [43]. Working memory was estimated by a computerized n-back test. Exposure to NO<sup>2</sup> was related to the slower development of working memory (β = −4.22, 95% CI: −6.22, −2.22). These reductions corresponded to a −20% (95% CI: −30.1, −10.7) change in annual working memory development associated with one interquartile range increase in outdoor NO2. Forns et al. [50] observed a persistent negative association between NO<sup>2</sup> levels at school and cognitive development over a course of 3.5 years. Therefore, they suggested that highly exposed children might face obstacles to fully achieve their academic goals.

Vert et al. [51] analyzed association between exposure to NO<sup>2</sup> and mental disorders on 958 residents from Barcelona (45–74 years old). Long-term residential exposure (period 2009–2014) was related to patients' self-reported history of anxiety and depression disorders. NO<sup>2</sup> exposure corresponded to 57.3 µg/m<sup>3</sup> (50.7–62.7). NO<sup>2</sup> increased the odd ratio for depression of 2.00 (95% CI: 1.37, 2.93) for each 10 µg NO2/m<sup>3</sup> increase. The study shows that long-term exposure to NO<sup>2</sup> may increase the incidence of depression.

Alemany et al. [52] analyzed on the group of children from the BREATHE project (*N =* 1667 at the age of 11 years), if there is any association between traffic-related air pollution and the ε4 allele of the apolipoprotein E gene, which is understood as a genetic risk factor for Alzheimer´s disease. NO<sup>2</sup> exposure at the home address was 54.25 <sup>±</sup> 18.40 <sup>µ</sup>g/m <sup>µ</sup>g/m<sup>3</sup> and at schools was 47.74 <sup>±</sup> 12.95 <sup>µ</sup>g/m<sup>3</sup> . NO<sup>2</sup> exposure increased behavioral problems scores (characterized by SDQ) in ε4 carriers (*N =* 366) vs. non-carriers (*N =* 1223) 1.14 (95% CI: 1.04, 1.26) vs. 1.02 (95% CI: 0.95, 1.10, *p* = 0.04) and was associated with smaller caudate volume in ε4 carriers (*N =* 37) vs. non-carriers (*N =* 126) −737.9 (95% CI: −1201.3, −274.5) vs. −157.6 (95% CI: −388.8, 73.6, *p* = 0.03). Annual average NO<sup>2</sup> concentrations in children´s schools were associated with smaller caudate volume and higher behavior problem scores among APOE ε4 allele carriers. It is possible that ε4 carriers are more vulnerable to neuroinflammatory and oxidative stress induced by air pollution exposure.

Carey et al. [53] investigated the incidence of dementia to residential level of NO<sup>2</sup> in London. Among 130,978 adults aged 50–79 years was, in the period 2005–2013, 2181 subjects diagnosed with dementia (39% Alzheimer´s disease and 29% vascular dementia). The average annual concentration of NO<sup>2</sup> was 37.1 <sup>±</sup> 5.7 <sup>µ</sup>g/m<sup>3</sup> . Higher risk of Alzheimer´s disease was observed in subjects exposed to the highest concentrations of NO<sup>2</sup> (>41.5 µg/m<sup>3</sup> ) vs. subjects with the lowest concentrations of NO<sup>2</sup> (<31.9 µg/m<sup>3</sup> ) (HR = 1.40, 95% CI 1.12–1.74). These associations were more consistent for Alzheimer´s disease than vascular dementia. Study found evidence of a positive association between residential level of NO<sup>2</sup> across London and being diagnosed with dementia.

Roberts et al. [54] explored the effect of NO<sup>2</sup> exposure to mental health problems in children in London, U.K. (*N =* 284). Symptoms of anxiety, depression, conduct disorder and ADHD were assessed at ages 12 and 18. NO<sup>2</sup> concentration in the year 2007 was 37.9 <sup>±</sup> 5.5 <sup>µ</sup>g/m<sup>3</sup> (IQR 34.1–41.7). They did not observe any association between NO<sup>2</sup> exposure in childhood and mental health problems at age 12. However, they detected association between NO<sup>2</sup> exposure and subsequent development of symptoms and clinically diagnosable depression and conduct disorders at age 18. They demonstrated that NO<sup>2</sup> exposure at age 12 years was significantly associated with major depressive disorder at age 18.

Prenatal exposure to NO<sup>2</sup> and sex dependent infant cognitive and motor development was analyzed by Lertxundi et al. [55] in children at 4–6 years of age, in four regions in Spain (*N =* 1119). Infant neuropsychological development was assessed by McCarthy scales: verbal, perceptive-manipulative, numeric, general cognitive, memory and motor. NO<sup>2</sup> exposure during pregnancy was from 18.7 <sup>±</sup> 6.1 to 41.8 <sup>±</sup> 10.7 <sup>µ</sup>g/m<sup>3</sup> . The majority of cognitive domains were negative for NO2, associations were more negative for boys, statistically significant for memory, global cognition and verbal. These findings indicate a greater vulnerability of boys in domains related to memory, verbal and general cognition.

Jorcano et al. [56] assessed association between NO<sup>2</sup> and depressive and anxiety symptoms, and aggressive symptoms in children of 7–11 years, related to their prenatal

and postnatal exposure. Data were analyzed in 13,182 children from eight European population-based cohorts. Prenatal NO<sup>2</sup> levels ranged from 15.9 to 43.5 µg/m<sup>3</sup> , postnatal levels ranged from 14.0 to 43.5 µg/m<sup>3</sup> . A total of 1108 (8.4%) and 870 (6.6%) children were classified as having depressive and anxiety symptoms, and with aggressive symptoms. Obtained results suggest that prenatal and postnatal exposure to NO<sup>2</sup> is not associated with depressive and anxiety symptoms or aggressive symptoms in children of 7–11 years old.

Loftus et al. [57] used the mother–child cohort from the CANDLE study and analyzed the impact of prenatal NO<sup>2</sup> exposure (22.3 <sup>±</sup> 7.1 <sup>µ</sup>g/m<sup>3</sup> ) and postnatal exposure (16.2 <sup>±</sup> 4.7 <sup>µ</sup>g/m<sup>3</sup> ) on childhood behavior (*N =* 975). In the sample 64% were African American, 53% had a household annual income below USD 35,000 and the child's age was 4.3 years. Mothers completed the child behavior checklist, a measure of problem behaviors in the past two weeks. The 4 µg/m<sup>3</sup> higher prenatal NO<sup>2</sup> was positively associated with externalizing behavior (6%, 95% CI: 1, 11%) and the effect of postnatal exposure was stronger (8%, 95% CI: 0, 16%). Prenatal NO<sup>2</sup> exposure was also associated with significant internalizing and externalizing behaviors. NO<sup>2</sup> exposure is positively associated with child behavior problems and African American and low SES children may be more susceptible.

Kulick et al. [58] examined in 5330 participants from the Northern Manhattan area of New York City the effect of long-term exposure to NO<sup>2</sup> (annual estimates 57.4 <sup>±</sup> 22.1 <sup>µ</sup>g/m<sup>3</sup> ) and PM2.5 (annual estimates 13.1 <sup>±</sup> 4.8 <sup>µ</sup>g/m<sup>3</sup> ), predominantly in women, with a median age of 75.2 (±6.46) years. A + IQR increase of residential NO<sup>2</sup> was predictive of a 22.SD (95% CI, 0.30, −0.14) low global cognitive score at baseline and a more rapid decline (−0.06 SD; 95% CI −0.08, −0.04) in global cognitive function between biennial visits.

Erikson et al. [59] studied the association between NO<sup>2</sup> exposure and total gray matter and total white matter volumes in adults, using sample from UK Biobank. Participants were recruited from 2006 to 2010, a subset with magnetic-resonance brain imaging (MRI) included 18,292 participants, with an average age of 62 (44–80) and NO<sup>2</sup> levels were 25.61 <sup>±</sup> 6.86 <sup>µ</sup>g/m<sup>3</sup> . The mean total gray-matter volume was 708,111 mm<sup>3</sup> (±47,940), the mean total white-matter volume was 708,111 mm<sup>3</sup> (±40,696). The total gray-matter volume was inversely associated with NO<sup>2</sup> (b = −103, *p* < 0.01). The effect of NO<sup>2</sup> on gray-matter volume was more pronounced in females (b = 161, *p* < 0.05). Obtained findings suggest that NO<sup>2</sup> concentrations lower than EU standard could be associated with reduced total gray-matter.

All reviewed studies indicate a significant health risk of NO<sup>2</sup> exposure at concentrations lower than the EU annual limit of 40 µg/m<sup>3</sup> :


The overall evidence presented in the mentioned studies suggests that attainment of the current EU annual limit for NO<sup>2</sup> of 40 µg/m<sup>3</sup> may not be sufficient for the protection of human health and further reductions of NO<sup>2</sup> concentrations would be beneficial and should be considered. In Switzerland, the current limit for the annual average of NO<sup>2</sup> is 30 µg/m<sup>3</sup> .

#### **4. Measurement of NO<sup>2</sup> in Prague by Passive Samplers**

To build up on this hypothesis, the measurements of NO<sup>2</sup> concentrations at various locations by passive samplers are examined. Some of the results were presented by Deutsche Umwelthilfe [60] as preliminary data; in this study, the results from Prague were examined in a greater detail.

For passive monitoring, membrane-closed Palmes tube [61] passive samplers (Passam, Switzerland [62]) were used. Several hundreds of samplers were placed at selected locations

in the Czech Republic, out of which 65 were in Prague, during spring and fall of 2019 (46 and 58 samplers, respectively, a total of 104 samplers), each time for a period of approximately one month. The placement of the tubes generally followed the requirements set in the EU air quality directive (2008/50)—placement away from buildings at a breathing height 1.5–4 m, away from larger obstructions, and for traffic sites, within 10 m of curbside and, in most cases, over 25 m from intersections. In some cases, the samplers were placed closer to intersections, and in some cases, the samplers were placed in less conspicuous places such as behind a traffic sign (see photo in Figure 1), to reduce the chances of tampering. The expanded uncertainty (95% confidence) of the measurement given by the manufacturer is 18.3% for a concentration range 20–40 µg/m<sup>3</sup> [62]. The location of samplers is shown on an overview map in Figure 1. The same map also shows the locations of the national air quality monitoring stations referred to in this study.

**Figure 1.** Locations of the passive samplers and air quality monitoring stations used for comparison in this study. Photo of a sampler is shown in the upper right corner. (Map source: www.mapy.cz (accessed on 18 May 2021), © Seznam.cz, a.s., used with permission).

The measured concentrations are given in Table 1. For the spring campaign, the dates of the sampling are listed in the "spring measurement period" column, while for the fall campaign, a value is given when a measurement has taken place during the three sampling periods, as some locations were sampled twice. The spring, fall and overall average concentrations, divided by a correction factor of 1.185 (will be explained later in the manuscript) are given. For each location, the average daily vehicle traffic counts reported by the City of Prague Highway Department for 2019 [63] are reported. This table also reports vehicle counts adjusted for additional emissions due to inclines and intersections, these adjustments are discussed later in the manuscript.


**Table 1.** Measured NO2concentrations and average daily vehicle counts.


**Table 1.** *Cont.*


**Table 1.** *Cont.*

#### *4.1. Validation by Comparison with the Air Quality Monitoring Network*

According to [64], passive diffusion tubes for measuring NO<sup>2</sup> concentrations in air were originally developed in the late 1970s for personal monitoring. They have been widely used in Europe for spatial and temporal measurement of NO<sup>2</sup> concentrations. The method has been found to be cheap, simple, and "provides concentration data in most circumstances that are sufficiently accurate for assessing exposure and compliance with Air Quality criteria" [64]. Reporting on a series of comparison tests, Buzica et al. [65] have concluded that "In the case of NO2, all the results of the laboratory and field experiments respected the requirements necessary for the demonstration of equivalence" and that the MCPT are equivalent to the reference methods for assessment of NO2. Passive diffusion tubes were reported to show a positive bias when sampling close to sources of NO, such as roadside or street canyons [64]. At the same time, prolonged (several weeks) sampling periods were reported to lead to negative bias [64]. A review done by the Joint Research Center of the European Commission [66], done in part to assess the feasibility of using the samplers for the long-term monitoring of nitrogen dioxide, with the particular aim of checking compliance with the European Union annual limit value of 40 µg/m<sup>3</sup> , citing a range of previous studies, reports that the "precision of the sampler showed that it is usually better than 5% when using a barrier or shelter to reduce effects of wind-induced turbulence" and that "the relative expanded uncertainty of individual results was estimated to be 32% for worst-case conditions", with lower values, generally <25%, obtained, for example, by parallel measurements with a reference method, by direct approaches, concluding that overall, "the Palmes tube is at least suitable for performing long-term measurements of NO<sup>2</sup> for indicative purposes, and possibly even for fixed measurements". Recent review of biases associated with Palmes tube type passive samplers by Heal et al. [67] suggests that "The effect of net bias can be reduced by application of a local "bias adjustment" factor derived from colocations of PDTs with a chemiluminescence analyzer. When this is carried out, the PDT is suitable as an indicative measure of NO<sup>2</sup> for air quality assessments".

To evaluate the bias, the data from passive samplers were compared to the data from selected relevant stations of the national air quality monitoring network, listed in Table 2. The national network uses chemiluminescence analyzers capable of measuring both NO and total NOx, with NO<sup>2</sup> calculated as the difference of total NO<sup>x</sup> and NO. The uncertainty of the measurements is periodically determined through analysis of reference samples, repeated measurements of the same sample, interlaboratory exercises, and for 2019, was reported to be a combination of absolute uncertainty of 2.3 µg/m<sup>3</sup> and a relative uncertainty of 12.3% [68].

The results of this comparison are given in Figure 2. In each case, the value reported by the passive sampler was compared to the average of hourly values from the monitoring station over the period during which the sampler was exposed. The three larger points (in red/orange) represent two samplers colocated with the Karlín monitoring station over two separate one-month periods and one sampler colocated with the Vysoˇcanská monitoring station, show a linear correlation with a slope of 1.185 (at zero intercept; standard error of slope 0.008; differences passive sampler vs. monitoring station of +20%, +17% and +18%). While it can be argued that a regression of three points has a limited meaning, in this case, it shows that three different samplers, each used in a different time period, has produced readings that are a consistent multiple of the monitoring station data. Additionally, two samplers placed at the city urban background reference station for particulate matter (Suchdol campus of the Czech Academy of Sciences, last two lines in Table 1) during the same time period show a relative difference of 6%. These findings are in line with the 5% precision of the Palmes tube samples reported in [66].


#### **Table 2.** Measured NO2concentrations and average daily vehicle counts—monitoring network.

**Figure 2.** Comparison of passive sampler reported NO<sup>2</sup> concentrations to the corresponding average values from corresponding monitoring stations. Larger points circled in red denote the colocation of the sampler at the monitoring station.

Smaller blue points in Figure 2 show additional locations. Two samplers were placed at an urban background monitoring station Suchdol, however, data from this station was not available, and the readings are compared with another background monitoring station in Kobylisy. Two samplers were placed near Námˇestí Republiky monitoring station, but a few dozen meters away and near an exit/entrance ramp to a large shopping center underground parking garage. Two samplers were placed on the corner of Legerova and Rumunská, near the monitoring station but at an intersection controlled by a traffic light. The readings from these four samplers were higher than from the monitoring station, which can be reasonably expected as they were near stopped and accelerating vehicles. The slope for the additional samplers was 1.17 with a standard error of 0.09; it should be noted that differences between actual NO<sup>2</sup> concentrations at the sampler and at the monitoring station are most likely the largest source of uncertainty.

Additional samplers close to the Legerova station (about 150 m from a large intersection) were closer to intersections and therefore exposed to additional cross-traffic, in addition to the increase in emissions rates in the vicinity of intersections. Two samplers were also placed at the Legerova monitoring station (urban hotspot) in the spring of 2019, but both were stolen. Additional samplers were placed near the Karlín monitoring station and near the Námˇestí Republiky monitoring stations, and in the general vicinity of the Legerova station. The NO<sup>2</sup> concentrations reported for the samplers were compared with the average NO<sup>2</sup> concentrations measured by the monitoring station, obtained by averaging data over the time the samplers were exposed on the site.

Additional samplers used in the comparison were at reasonably close locations with not overly dissimilar traffic, and were not too far from the 15% tolerance reported by the Defra report [64]. It should be noted that the tolerance is applicable to the deviation of the sampler-reported and reference value, and not to the differences due to the samplers being at different locations with different emissions characteristics.

For all subsequent data analysis, the concentrations from the passive samplers were divided by the regression slope of 1.185. It should be noted that while this correction represents the best judgment by the authors, it is based on limited data and could be viewed as arbitrary, as the difference could arise out of the 12.3% uncertainty of the reference measurement the manufacturer-reported 18% expanded uncertainty of the passive sampler.

#### *4.2. Comparison of NO<sup>2</sup> during Passive Samplers Deployment with Long-Term Averages*

The variation of climatic and weather conditions is an additional source of bias to consider when comparing passive samplers to annual mean values. Figure 3 shows that the average values of NO<sup>2</sup> recorded at the monitoring stations over sampling periods of individual samplers (different four-week periods in March–April 2019) did not dramatically differ from annual means during the last four years (2016–2019), although differences in trends were observed among the stations. For example, the Legerova urban hotspot station exhibited an annual average of 51 µg/m<sup>3</sup> (2016–2019), compared to 46 µg/m<sup>3</sup> during the period of 9 March–April 6 and 62 µg/m<sup>3</sup> during 19 March–24 April. The Námˇestí Republiky urban background station had a 2016–2019 average of 30 µg/m<sup>3</sup> , compared to 29 µg/m<sup>3</sup> during 9 March–6 April and 35 µg/m<sup>3</sup> during 19 March–24 April. It should be noted that the NO<sup>2</sup> concentrations were generally lower during mid-March and higher during mid-April. Overall, the NO<sup>2</sup> concentrations during the sampling periods are believed to be representative of the annual average concentrations.

**Figure 3.** Comparison of monitoring station NO<sup>2</sup> averages during sampling periods with fouryear average.

The consistency of the measurement by passive samplers during spring and fall periods is shown, along with data from the reference monitoring stations, in Figure 4. The slope of regression (with intercept forced through zero) was 0.91 ± 0.05 for the monitoring stations and 0.92 ± 0.02 for the passive samplers, showing that the monitoring stations and the passive samplers reported the same overall trends in NO<sup>2</sup> concentrations.

**Figure 4.** Comparison of spring and fall NO<sup>2</sup> concentrations.

#### *4.3. Effects of Traffic*

For further analysis, all passive sampler measurements were divided by a factor of 1.185 (the slope of regression of passive sampler vs. reference NO2, see Figure 1).

The relationship between the vehicular traffic intensity and the NO<sup>2</sup> concentrations measured by the passive samplers is given in Figure 5. As samplers were used over two different periods, they are plotted separately in two series, one for each period, along with the average values from Legerova and Námˇestí Republiky monitoring stations. It appears that there is a moderate positive trend of NO<sup>2</sup> increasing with traffic. Additionally, samplers located next to an uphill section of a divided highway (or a one-way street with the traffic going in the uphill direction) and next to an intersection tend to exhibit higher NO<sup>2</sup> concentrations. It also appears that the NO<sup>2</sup> concentrations are higher in urban canyons and congested streets of the city center and near intersections.

**Figure 5.** Relationship between traffic intensity and NO<sup>2</sup> concentrations measured by passive samplers in spring and fall of 2019 and by the national monitoring network (average of 2016–2019).

To assess whether high NO<sup>2</sup> are associated with truck traffic, samplers located in the area with limited access of vehicles over 6 tons gross weight (entry by permit only, restricted to local traffic) are plotted separately in Figure 6 (for locations where multiple samplers were used, average values are plotted). It is clear from the figure that the highest NO<sup>2</sup> were measured in areas where trucks over 6 tons are mostly excluded.

To account for additional emissions due to hills and intersections, the intensity of traffic traveling uphill was increased by 100% to account for additional fuel consumption, and for samplers located at intersections, the intensity of traffic was increased by 300% to account for fuel consumed at idle and when accelerating (where the intersection was without a major delay, such as time-synchronized signals at intersections of a larger oneway street with a side street or pedestrian crossing, the factor was reduced by one half). These adjustments factors were arbitrarily selected based on experience with vehicle emissions behavior (additional emissions due to climbing a hill, additional emissions due to idling at intersections and acceleration from intersections) and were independent of each other. (Note: as an example of rough calculation for a passenger car diesel engine, the acceleration of a 1500 kg car from 0 to 50 km/h requires a gain of kinetic energy of 145 kJ or 40 Wh, corresponding, at 250 g/kWh engine fuel consumption, to 10 g of fuel. The fuel consumption at idle is about 5 g/min. A one-minute stop and acceleration consumes 15 g of fuel. Driving at steady speed requires about 30 g of fuel per km, or 3 g per 100 m. If half of the cars stop and wait, the emissions in a 100 m segment around the intersection

are 9 g, compared to 3 g in the case of free-flowing traffic. For simplicity, NO<sup>x</sup> emissions are assumed to be proportional to the fuel consumption.) The relationship between the adjusted vehicle volume and NO<sup>2</sup> concentrations is plotted in Figure 7.

**Figure 6.** Relationship between traffic intensity and NO<sup>2</sup> concentrations measured by passive samplers (average of all measurement periods) and by the national monitoring network (average of 2016–2019).

**Figure 7.** Relationship between adjusted traffic intensity (traffic count × (1 + fraction of vehicles travelling uphill + 3 × fraction of vehicles stopping at an intersection)) and NO<sup>2</sup> concentrations measured by passive samplers (average of all measurement periods) and by the national monitoring network (average of 2016–2019).

The relatively strong correlation between the adjusted traffic volumes and NO<sup>2</sup> concentrations (R<sup>2</sup> = 0.78 for September-October data and 0.76 for spring-fall averages; slope 0.13 <sup>±</sup> 0.01; intercept 27 <sup>±</sup> <sup>1</sup> <sup>µ</sup>g/m<sup>3</sup> ) suggests that "local" NO2, comprising of primary NO<sup>2</sup> emitted from the tailpipe and NO<sup>2</sup> formed locally from NO by reaction with ozone (i.e., [69]), is a considerable and in many locations dominant source of NO2. There is no observable difference between the sampling locations where truck traffic over 6 tons was excluded and the locations where it was not excluded. Overall, there seems to be a very strong correlation between the estimated relative intensity of mobile source emissions and the measured NO<sup>2</sup> concentrations. It is likely that the correlation could be further improved by taking into the account distance from the traffic, traffic on adjacent streets, tunnel exits and other compounding factors.

A similar plot of the regression of the dependency of NO<sup>2</sup> on adjusted traffic volumes is plotted separately for the spring and fall campaigns in Figure 8, with red line denoting the legal annual NO<sup>2</sup> limit of 40 µg/m<sup>3</sup> and green line the Swiss federal limit of 30 µg/m<sup>3</sup> (shown for illustration in support of the health review). The regression shows that NO<sup>2</sup> concentrations, in all cases, increased by 0.13 µg/m<sup>3</sup> per 1000 vehicles daily traffic volume, adjusted for uphill and intersections, where adjusted traffic count is traffic count multiplied by a factor of (1 + fraction of vehicles travelling uphill + 3 × fraction of vehicles stopping at an intersection). It should be noted that the intercept of the regression (25–28 µg/m<sup>3</sup> in Figures 7 and 8; (standard error of slope is 0.01; standard error of intercept is 1 µg/m<sup>3</sup> ) is higher than the "urban background" concentrations of 15–20 µg/m<sup>3</sup> , most likely due to accounting only for traffic on major roads and not for parking garages, taxi waiting areas, and similar locations. Even the urban background concentrations cannot be considered as NO<sup>2</sup> concentrations that would be theoretically be expected if no motor vehicles were operated in Prague, due to the dispersion and transport of the pollutants.

**Figure 8.** Relationship between adjusted traffic intensity (traffic count × (1 + fraction of vehicles travelling uphill + 3 × fraction of vehicles stopping at an intersection)) and NO<sup>2</sup> concentrations measured by passive samplers (average of all measurement periods) and by the national monitoring network (average of 2016–2019). EU annual limit of 40 µg/m<sup>3</sup> NO<sup>2</sup> shown as a red line, Swiss federal limit of 30 µg/m<sup>3</sup> NO<sup>2</sup> shown as a dotted green line.

Even at a rather conservative adjustment of the passive sampler readings (according to the regression, the sampler readings were 18% higher, however, this was, to a large extent, due to many samplers being at locations where the concentrations would reasonably be

expected to be higher than at the corresponding monitoring station), it is clear from Figure 7 that the annual average limit of 40 µg/m<sup>3</sup> NO<sup>2</sup> is likely to be exceeded at numerous locations throughout Prague, generally, where the adjusted traffic volumes exceed the equivalent of 100 thousands of vehicles per day. This is, for example, the north-south passageway through the center city (Wilsonova, Sokolská and Legerova street) with many intersections, but also roads like V Holešoviˇckách (a six-lane road with 85–90 thousand vehicles per day, with a gradient of approximately 3%), a possible new hot-spot in Prague. In the worst case (intersection of two one-way streets with all vehicles traveling uphill), this limit could be reached already at 20 thousand vehicles per day, as also apparent from Figure 6.

#### **5. Effects of Travel Restrictions on Ambient NO and NO<sup>2</sup> Concentrations**

In order to assess the contribution of light and heavy vehicles to NO and NO<sup>2</sup> concentrations, hour-by-hour NO and NO<sup>2</sup> ambient air quality data from the national air quality monitoring network was analyzed for a period of 14 March–30 April 2020, during which travel restrictions were imposed, including the prohibition of all non-cargo international travel (truck traffic was exempted). For reference, the same period was assessed for four previous years.

A total of five stations in Prague were selected:


Arithmetic and geometric means and the NO2/NO<sup>x</sup> ratios are plotted, for each station and all years, in Table 3. A single-factor analysis of variance (ANOVA) was performed to compare the variances among the five data sets (one for the year 2020, four for each of the reference years 2016–2019) with the differences within the sets. The associated *p*-value (p1) was compared to the *p*-value (p2) associated with the difference between mean for the year 2020 and the grand mean for all five years. The higher of the p2/p1 ratio and the p2 (ensuring that the significance of the difference of the year 2020 is much higher than the difference among the years) is then considered the resulting *p*-value of the test.

As an alternative analysis, the statistical difference of data from each year from the combined data set for all five years was evaluated using a *t*-test, and the *p*-value associated with the test for the year 2020 was divided by the average of the four *p*-values associated with each of the four reference years.

It is apparent from the Table 2 that NO concentrations significantly decreased at all three traffic stations, with a highest mean decrease of 46% at Legerova and at the Košetice rural background station. The decrease in NO<sup>2</sup> concentrations was lower than for NO at all Prague stations, highest at Legerova (20%), and even higher (40%) at the Košetice rural background station. As vehicles emit primarily NO, the NO2/NO<sup>x</sup> ratio tends to increase with the age of the emissions, being lowest (around 60%) at Legerova street, 65–70% at Vysoˇcanská, Pr ˚umyslová and Námˇestí Republiky, 80% at the Kobylisy residential background station and around 90% at the rural station in Košetice. One possible interpretation of the increase in the NO2/NO<sup>x</sup> ratio at Legerova could be that the primary emissions of both NO and NO<sup>2</sup> were reduced, with lower reduction in "background" NO<sup>2</sup> originating from NO<sup>x</sup> emitted elsewhere. Another possible explanation is the reaction

of NO with ozone, yielding NO<sup>2</sup> [70]. Both March and April of 2020 were substantially sunnier than average—4 sunny days and 180 h of sunshine in March and 13 sunny days and 290 h of sunshine in April, compared to 1981–2010 average of about 3 sunny days and 120 h of sunshine for March and 3–4 sunny days and 180 h of sunshine for April [71].

**Table 3.** Comparison of NO and NO<sup>2</sup> concentrations at six monitoring stations during March–April 2020 travel restrictions with the same period during the prior four years.


\* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001, \*\*\*\* *p* < 0.0001.

It should be noted, however, that the interplay of different factors is rather complex. For example, diminished traffic volumes result in lower frequency of low-speed driving in congested areas, during which the efficiency of exhaust aftertreatment is reduced, resulting in higher overall exhaust temperatures (and thus higher production of NO<sup>2</sup> in oxidation catalysts), but also higher probability of SCR functionality (and thus lower NO<sup>x</sup> emissions)—however, due to Dieselgate, the reality of NO<sup>x</sup> aftertreatment efficiency is likely to be variable, questionable and poorly known.

Additionally, according to [72], it appears that on-road oxidation of NO by ambient O<sup>3</sup> is a significant, but so far ignored, contributor to curbside and near-road NO2. This is in agreement with on-road NO2/NO<sup>x</sup> ratios in U.S. being reported to be 25–35% and substantially higher than anticipated tailpipe emissions rates [73].

#### **6. Discussion**

A detailed analysis of NO<sup>2</sup> concentrations measured by the passive samplers shows a clear correlation of NO<sup>2</sup> concentrations with daily traffic counts, adjusted for additional emissions due to uphill travel and stopping at intersections. This finding is in good agreement with the data from the monitoring stations, which, by themselves, are too sparse to make such inference. The correlation of NO<sup>2</sup> concentrations with vehicular traffic intensity is also apparent from the comparison of the data from state air quality monitoring stations during the period of 14 March–30 April 2020, during which travel restrictions were imposed, including the prohibition of all non-cargo international travel, with comparable periods of four previous years. Overall, the findings confirm that vehicular traffic, through primary NO<sup>2</sup> emissions (and possibly through fast reaction of primary NO with ozone), directly affects the NO<sup>2</sup> concentrations in the immediate vicinity.

This correlation, along with correlation of passive sampler readings and air quality monitoring stations, and good consistency of reported NO<sup>2</sup> concentrations among samplers used within the same location at different time periods, all suggest that passive samplers appear to provide, at a reasonable cost and effort, a fairly good image of the distribution of NO<sup>2</sup> concentrations. Judging from limited data, the passive samplers were found to measure about 18.5% higher values than the monitoring stations. Repeated—and most likely deliberate—removals of passive samplers from the immediate vicinity of the monitoring stations have prevented a more quantitative comparison. A comparison of a broader set of data reveals a slightly smaller bias, contributed to, in several cases, by the passive samplers being at more exposed locations (i.e., near the exit of a large underground parking garage) than the monitoring stations. The true bias could therefore be possibly even lower.

Since the trends are comparable within and outside the heavy truck exclusion area, this seems to be primarily an effect of cars and other lighter vehicles (per city statistics, about 90% of traffic is passenger cars [63]). Additionally, there is no correlation between the measured NO<sup>2</sup> concentrations and the heavy vehicle traffic count or between the measured NO<sup>2</sup> concentration and the fraction of heavy vehicles. This is in line with the findings that truck NOx emissions have decreased to a considerably higher extent than those of diesel cars in Europe.

The samplers at the locations with highest fraction of heavy vehicles (10–15%, vs. average for all locations 4%) and with the highest absolute heavy vehicle counts (7–16 thousands/day, vs. average 1.7 thousands/day) have measured 25–35 µg/m<sup>3</sup> NO2, which is in the second lowest quartile (median concentration is 35 µg/m<sup>3</sup> ). This may also be, in part, due to a dependent factor that heavy vehicle traffic is limited in the high population density city center.

The monitoring station at Legerova street is most likely not the absolute hot-spot—it is expected that the emissions of NO<sup>x</sup> would be higher on the parallel street where the vehicles travel uphill (Legerova is one-way street downhill) and at nearby intersections. The street V Holešoviˇckách, a six-lane road, which is, unlike most other roads of similar size, immediately bordered by residential neighborhoods, with a traffic intensity approaching 100 thousand vehicles per day, a major increase after the opening of a new complex of tunnels providing an alternative route through congested areas, further complicated by a 3% grade, could easily be the next traffic hot-spot.

Considering the finding that about half of the vehicles traveling on the road are not older than 7 years [27], and the several-fold decrease in NO<sup>x</sup> emissions standards over the last decade and half, a much sharper decrease of NO<sup>2</sup> concentrations would be expected than the approximately 1% annually reported by H ˚unová [5]; a higher reduction of about 2.5% annually was observed in Western Europe, and about 4.7% annually in United States and Canada [74]. Given the decrease in the limit values of roughly two thirds from Euro 3 (0.50 g/km NOx, 2000) to Euro 5 (0.18 g/km, 2009–2010) and from Euro 4 (0.25 g/km NOx, 2005) to Euro 6 (0.08 g/km, 2014–2015), the introduction of Euro 5 in late 2009 and Euro 6 in late 2014 should have resulted in about a two thirds NO<sup>x</sup> reduction in at least half of the vehicles, or about one third reduction in NO<sup>x</sup> emissions in general. As learned from the analysis of the effects of traffic restrictions, the effect on NO<sup>2</sup> concentrations may be different, and possibly somewhat smaller than the reduction in NO<sup>x</sup> emissions, due to atmospheric chemistry. The effects of such a decrease could also have been diminished by an increase in traffic, however, in the center city, the intensity of automobile traffic has been stagnating, or even slightly decreasing.

The mediocre decrease in NO<sup>2</sup> concentrations, despite more dramatic reduction being expected from improving vehicle technology, is in line with earlier findings that the real NO<sup>x</sup> emissions of diesel vehicles did not decrease despite the decreasing emissions limits. The situation should have been, however, substantially remedied by "post-Dieselgate" vehicles and by repairs of vehicles affected by Dieselgate. Since it was not, a question therefore arises as to the possibility that Dieselgate relevant repairs were not done on a sufficient number of vehicles and/or were not sufficiently effective and/or were reversed to the "original factory conditions" by the vehicle owners. The authors do not have any reliable statistics on this matter. Furthermore, considering that all three mentioned situations could be associated with criminal offenses and/or considerable civil penalties, detailed investigation of the matter is likely to be considerably difficult.

If there is no assurance that the NO<sup>2</sup> concentrations will decrease dramatically due to a radical improvement in primary NO<sup>x</sup> emissions, the only other suitable strategy to improve the air quality is to reduce, to the extent required, the intensity of vehicular traffic. Contrary to the remote regions where automobiles are, in most cases, the only practical means of travel, Prague has an extensive network of public transit. According to the City of Prague statistics [63], only 29% of trips in Prague are done by automobile, 26% of trips are by walking and 42% of trips by public transit. Of the public transit, slightly over one third is done by subway, and another third by trams and commuter rail, which are, with the exception of a rather small number of diesel rail cars used on sparsely traveled rail lines, run on electric power, and therefore with very small effect on NO<sup>2</sup> emissions. The remaining third of trips is by diesel buses, the majority of which are equipped with SCR catalysts, and potentially reaching NO<sup>x</sup> emissions not much larger (and according to measurements possibly even smaller) levels, per kilometer and vehicle, than an average diesel car. It is therefore readily apparent that shifting from an average automobile to any other means of transport is likely to reduce the NO<sup>2</sup> concentrations. (Shift to electric power, compressed natural gas, or other "clean" propulsion is a gradual process and is unlikely to be done, within a few years, on a sufficiently large number of vehicles to make a difference throughout the city).

#### **7. Summary and Conclusions**

Despite massive reductions in diesel cars NO<sup>x</sup> emission limits, of about two thirds from Euro 3 to Euro 5 and from Euro 4 to Euro 6, NO<sup>2</sup> concentrations throughout the Czech Republic have been decreasing at a mediocre rate of 1% annually.

A review of the underlying engine emissions trends shows that the conversion of NO into NO<sup>2</sup> in diesel oxidation catalysts, beneficial for regeneration of diesel particle filters and for the functioning of the SCR systems for NO<sup>x</sup> reduction, did not, contrary to the intentions of the legislation, go hand in hand with a major reduction of NO<sup>x</sup> emissions in subsequent (downstream) NO<sup>x</sup> aftertreatment devices. As a result, primary NO<sup>2</sup> emissions from light duty diesel vehicles are in most cases considerably higher than intended in the emissions legislation due to non-adherence of many manufacturers to the primary intent of the legislation.

A review of the health effects on NO<sup>2</sup> on children shows that all reviewed studies indicate a significant effect of prenatal NO<sup>2</sup> exposure to children´s neurobehavioral development, in adults to dementia at concentrations lower than EU standards of 40 µg/m3/year. These results should be understood as a strong recommendation to reduce the NO<sup>2</sup> concentrations below the current EU standard. All presented studies prove that NO<sup>2</sup> can

significantly deteriorate CNS and therefore this knowledge should be used to improve the quality of our lives.

To elucidate the effects of motorized traffic on NO<sup>2</sup> concentrations, data from 104 passive NO<sup>2</sup> samplers deployed at 65 locations in Prague during March–April and September– October of 2019 were examined. Comparisons with the national monitoring network show a positive bias of 18.5% for colocated samplers and 17% for samplers nearby (or in similar settings as) the monitoring stations. There was a good correlation among repeated measurements at the same locations. The data from the national air quality monitoring network show that the average concentrations in both spring and fall sampling periods were consistent with 2016–2019 averages.

The average measured NO<sup>2</sup> concentrations at the selected locations, after correcting for the 18.5% bias, were in the range of 16–69 µg/m<sup>3</sup> , with a mean of 36 µg/m<sup>3</sup> and a median of 35 µg/m<sup>3</sup> , and were higher than the EU and national limit (annual average) of 40 µg/m<sup>3</sup> at 32% of locations. The NO<sup>2</sup> concentrations have correlated well with the intensity of traffic (average daily vehicle counts), corrected for additional emissions due to uphill travel and due to idling at, and accelerating from, intersections. Several additional "hot-spots" were identified, in addition to the "hot-spot" monitoring station at Legerova street (2016–2019 NO<sup>2</sup> average of 51 µg/m<sup>3</sup> ), where the vehicles travel on a slight decline on a one-way street: several intersections at Sokolská street, parallel with Legerova with uphill direction of travel, and emerging hot-spots along V Holešoviˇckách street, where the traffic intensity increased due to the opening of a new series of tunnels. Analysis of the effect of coronavirus related travel restrictions were evaluated by comparing the data from six monitoring stations (15 March–30 April 2020, relative to the same period during 2016–2019) reveal a reduction of NO, NO<sup>2</sup> and NO<sup>x</sup> (except for a small increase of NO<sup>2</sup> at one of the background stations), with NO reduction being, at high traffic locations, higher than that of NO2. The spatial analysis of data from passive samplers and time analysis of data during the travel restrictions both demonstrate a consistent positive correlation between traffic intensity and NO<sup>2</sup> concentrations along/near the travel path.

It appears that decreases in vehicle NO<sup>x</sup> emission limits, introduced in the last decade or two, have failed to sufficiently reduce the ambient NO<sup>2</sup> concentrations in exposed locations in Prague. This is in part due to increased fraction of NO<sup>2</sup> in NO<sup>x</sup> in newer vehicles, and in part due to "a major disparity between the numerical value of the emission limit and the actual emissions in everyday driving". Further, there is no apparent sign of, and it is far from clear that, the "excess emissions" of NOx, a problem known as Dieselgate, have been efficiently remedied.

**Author Contributions:** M.S. has organized the passive sampling campaign, selected locations, placed and removed samplers, and secured funding. R.J.S. has compiled the review of health effects. J.S. has participated in data analysis. M.V.-L. reviewed the engine emissions and did a large share of data analysis and manuscript writing. All authors have read and agreed to the published version of the manuscript.

**Funding:** The acquisition and analysis of the passive samplers was funded by Deutsche Umwelthilfe (Environmental Action Germany), Hackescher Markt 4, 10178 Berlin, Germany, www.duh.de (accessed on 18 May 2021). Evaluation of the passive samplers and some of the background research on emissions was done (M.V.) within the H2020 project no. 851002 uCARe. You can also reduce emissions. Review of the health effects and remaining work has been supported by the European Regional Development Fund under Grant Healthy Aging in Industrial Environment HAIE (CZ.02.1.01/0.0/0.0/16\_019/0000798).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Most of the relevant data is contained in the manuscript. Sampling and analytical protocols associated with passive samplers are available from Miroslav Šuta. Traffic volume data are publicly available, see the link in the reference list. Data from the national air quality monitoring network are a third-party data and must be requested directly from the Czech Hydrometeorological Institute.

**Acknowledgments:** The data from the national air quality monitoring network was provided by the Czech Hydrometeorological Institute. The authors thank Václav Novák, the Head of the Air Quality Information System Department, for providing the data and for helpful advice.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


*Article*
