Concentration , Chemical Composition and Origin of PM 1 : Results from the First Long-term Measurement Campaign in Warsaw ( Poland )

This paper presents a 120-day-long variability of chemical composition of submicron particulate matter (PM1) over Warsaw. The content of the following components was examined in the PM1 mass: primary (POM) and secondary (SOM) organic matter, secondary inorganic matter (SIM), elemental carbon (EC) as well as Na and Cl ions (primary inorganic matter). The 24-hour concentrations of PM1 were subject to seasonal fluctuations which are typical of urban areas in Poland; their values averaged 11 μg m in summer and 17 μg m in winter. Most of the PM1 components and gaseous pollutants (SO2, NO2 and NOx) revealed higher mean concentrations in winter than in summer. A statistical analysis of meteorological parameters and 24-h concentrations of PM1, PM10, SO2, NO2 and NOx confirmed a significant influence of air temperature and precipitation on the concentration patterns of these pollutants over Warsaw. The highest concentrations of PM1 occurred in winter for the following wind directions: S, SE, N and NE; in summer for NE, E and S. The analysis of back trajectories demonstrated that on days with the highest 24-h concentration of PM1 polluted air masses arrived from S and SE and affected the concentration of PM1 over Warsaw. The submicron particulate matter, in as much as 62%, comprises secondary matter (SOM and SIM). The primary sources of its precursors and – to a smaller extent – of the primary matter as well – are traffic and combustion of various fuels for the purpose of heat and power generation. Their average contribution to the development of PM1 was 15% and 51%, respectively, for the entire period of observations.


INTRODUCTION
Synoptic and meteorological conditions have a major role in the development of atmospheric particulate matter (PM) concentration.The most important parameters affecting PM concentration are: the wind speed -influencing the rate and scale of dispersion of PM, the humidity of air and the solar radiation leading to the formation of secondary PM, and precipitation influencing the rate of washing PM out of the atmosphere (Pueschel, 1996;Mues et al., 2012;Seinfeld and Pandis, 2012;Olszowski, 2016;He et al., 2017;Olszowski, 2017).Our earlier research shows that on windcalm days and without rain, the concentration of PM in Polish cities is the highest, whereas considerable wind speeds atmosphere (and thus deciding on what and in what quantities is introduced into the atmosphere and subject to the above-mentioned phenomena), is the emission.In other words, the chemical composition of PM at a given location, regardless of the prevailing synoptic and meteorological conditions, depends on the PM origin.(Chow, 1995;Rogula-Kozłowska, 2014;Chow et al., 2015;Błaszczak et al., 2016;Hong et al., 2017).
In Polish urban areas, main sources of PM and its gaseous precursor emissions are: the so-called low or communal emissions, which involves dust and gases from the combustion of coal and its derivatives in household stoves and local boiler plants.They tend to be occasionally fed with plastics, solid and organic waste.The main sources also involve energy sector and industries (such as coking plants, foundries, steel mills and cement plants), and -first of all -road traffic emissions in a broad sense (Rogula-Kozłowska et al., 2015).A dense network of traffic routes, increasing traffic volumes, growing numbers of vehicles, often with faulty catalytic converters or diesel particulate filters (DPFs), largely contribute to the increase of PM concentrations in urban areas in Poland (Rogula-Kozłowska et al., 2013;Rogula-Kozłowska, 2015).
The Warsaw conurbation is the largest metropolitan area in Poland with 1,735,442 inhabitants on a surface of 517 km 2 .Data on PM pollution over Warsaw conurbation comes directly from measurements carried out at six air quality monitoring stations, being part of the national network.Concentrations of fine particles (PM 2.5 ) are analysed at only four of these stations.As a matter of fact, there is no data on chemical composition of PM in that part of the country.At the same time, in order to establish the source of PM and implement measures to control PM concentrations in urban areas, comprehensive research has been conducted around the world already for years, regarding chemical composition of PM (Chow et al., 2015).
Submicron particulate matter (PM 1 ) is a significant element of PM-related research due to typically anthropogenic origin and prevalent mass fraction in PM, as well as insufficiently studied chemical composition (in Polish and other European conurbations).Correlation between exposure to submicron particles and health effects, as well as their influence on other environmental components, and even climate change, have been proven and are widely discussed in literature (Massolo et al., 2002;Pope and Dockery, 2006;Karlsson et al., 2009;Paasonen et al., 2013;Atkinson et al., 2015;Badyda et al., 2016;Majewski et al., 2018).
In this paper, concentrations and chemical composition of PM 1 were analysed on the basis of 120-day-long measurement campaign conducted at one measurement site within the Warsaw conurbation.We consider it to be the first long-term study where day-by-day differences in chemical composition of PM 1 are shown in typical urban area in Central-Eastern Europe.Correlations between PM 1 and its main components with meteorological parameters were examined and the sources of PM 1 in different measurement periods were identified.The PM 1 mass, which came either directly from emissions (primary PM) or from the transformations of PM gaseous precursors (secondary PM), was determined for each day of the measurement period.

Collection and Preparation of Samples
Observations were conducted in Warsaw (λE = 21° 02′, φN = 52° 09′) at a measurement site located in the district of Ursynów, selected to satisfy the criteria of urban background location as specified in the Directive 2008/50/EC (Fig. 1).Topographical and meteorological conditions, the structure and amount of emissions from PM and gas pollution sources (the fraction of local emissions from boiler plants and rooms, stoves, industrial facilities and traffic at the sampling point) were typical of the whole conurbation (Majewski et al., 2014;Majewski and Rogula-Kozłowska, 2016).
From 24 June to 22 August 2014 (60 samples; summer) and from 8 January to 8 March 2015 (60 samples; winter), 120 diurnal samples of PM 1 were collected by means of an MVS6D PM sampler (provided by ATMOSERVICE; Poznań, Poland).The sampler was equipped with a sampling head (produced by TSI; PM inlet with PM 1 jet impactor) which aspired PM 1 at a flow of 2.3 m 3 h -1 (according to EN14907).The PM 1 was collected using quartz fibre filters (Quartz Air Sampling Filter, Grade QMA, 47 mm circle; GE Healthcare Life Sciences Corp.; Piscataway, NJ, USA).The mass of PM 1 was determined by weighing the filters before and after exposure; MYA 5.3Y.F micro balance (RADWAG; Radom, Poland) was used (1-µg resolution).Before each weighing, the filters were conditioned for 48-h in the weighing room (relative air humidity 45 ± 5%, air temperature 20 ± 2°C).
The adopted method of PM sampling and the subsequent gravimetric analysis of PM were consistent with EN 14907/2005 -Standard gravimetric measurement method for the determination of the PM 2.5 mass fraction of suspended particulate matter and EN 12341/1998: Air quality Determination of the PM 10 fraction of suspended particulate matter reference method and field test procedure to demonstrate reference equivalence of measurement methods.The weighing accuracy, determined as three standard deviations of the mean from ten weightings of a blank filter (conditioning performed every 48-h), was equal to 16 µg.
When the concentration of PM 1 was established, a 1 × 1 cm square was cut out of each diurnal sample in which the organic (OC) and elemental carbon (EC) content were determined.The remaining portion of each sample was then used to determine the content of ions from watersoluble compounds.

Chemical Analysis
The OC and EC contents of PM 1 were determined using a Lab OC-EC Aerosol Analyzer (Sunet Laboratories Inc.; Portland, OR, USA) and the EUSAAR protocol.The performance was controlled by systematic calibration of the analyser within the range proper for the determined concentrations, and by analysing standards with certified carbon content (RM 8785 and RM 8786, NIST, Gaithersburg, MD, USA) and the blank samples.The detection limit for total carbon (TC), computed after analysing the 10 blanks, was 0.91 µgC cm -2 (0.79 and 0.12 µgC cm -2 for OC and EC, respectively).The standard recovery was from 98% to 122% of the certified value for OC and from 95% to 116% for EC (the certified values were taken from the IMPROVE program).
Water extracts of PM 1 were made by ultrasonizing the filters containing PM 1 in 25 cm 3 of de-ionized water for 60 min at the temperature of 15°C and then shaking the extracts for about 12 h (18°C, 60 r min -1 ).The ion content of the extracts was determined by the ion chromatograph (Metrohm AG; Herisau, Switzerland).The method was validated against the CRM Fluka product nos.89316 and 89886; and standard recoveries were from 92% (Na + ) to 109% (Cl -) of the certified values, and detection limits were as follows: 10 ng cm -3 for NH 4 + , 18ng cm -3 for Cl - and SO 4 2-, and 27 ng cm -3 for NO 3 -and Na + .

Meteorological Parameters and Air Pollution Data
In the course of observations, values of basic meteorological parameters were measured at Ursynów Meteorological Station (Ursynów-SGGW) of the Warsaw University of Life Sciences (λE = 21°02'; φN = 52°09') according to instructions for the network of state weather stations operated by the Institute of Meteorology and Water Management (IMWM).Following meteorological parameters were used: air temperature, solar radiation, relative humidity of air, atmospheric pressure, precipitation totals, speed and direction of wind.Hourly results of the measurements were averaged for 24-h intervals from 12:00 a.m. to 12:00 a.m.; diurnal samples of PM 1 were taken at the same time.
The methodology to evaluate the fraction of primary and secondary matter in the mass of PM 1 was based on following assumptions: -Primary matter comprises: elemental carbon (EC), primary organic matter (POM) and a sum of sodium and chloride (Na_Cl), -Secondary matter comprises: secondary inorganic matter (SIM) and secondary organic matter (SOM).The mass of EC was assumed to be the analytically determined mass [EC] A of elemental carbon: PM-bound organic compounds can be divided into secondary organic matter (SOM) and primary organic matter (POM) (Turpin and Lim, 2001).Their masses, [SOM] and [POM], in the sampled PM are computed from the mass [OC]A of the analytically determined OC.At first, by utilizing the known [OC]A, the mass of secondary [SOC] and primary organic carbon [POC] was calculated.The division of OC into secondary and primary was based on the following assumptions: where ([OC]/[EC])pri is the ratio of primary OC to EC, assumed to be relatively constant for a specific location, season and local meteorology (Castro et al., 1999;Wu and Yu, 2016) Minimum value of 24-hour OC/EC ratios for Warsaw, out of 60 achieved values, was equal to 1.3 for the summer season and 0.6 for winter one.Maximum values for both seasons amounted to about 6.However, the average ones for 24-h OC/EC ratios for Warsaw reached about 3 for both seasons.Previous research of the authors revealed that the traffic emission in Warsaw significantly contributed to fine PM concentration (Majewski and Rogula-Kozłowska, 2016).However, in the winter period, both the emission from energy sector and pollutants' inflow from highly contaminated region of southern Poland (Rogula-Kozłowska et al., 2013, 2014;Widziewicz et al., 2017a) may cause the increase of carbonaceous matter concentration in the air, including volatile organic compounds -the precursors of SOM (Błaszczak et al., 2016).The OC/EC ratio close to 1 is typical for traffic emission as proved by the measurements either in a traffic tunnel or in urbanized areas influenced mainly by traffic emission, showing also lower values for fumes from Diesel engines than for gasoline ones (Hildemann et al., 1991;Lee et al., 2006;Zhu et al., 2010;Ancelet et al., 2011;Khan et al., 2011).For the exhaust gases coming from coal and biomass combustion, that ratio is several times higher (Watson et al., 1994;Cao et al., 2005;Lonati et al., 2007;Keywood et al., 2011).Assuming that trafficrelated emission is the main source of PM 1 -bound OC and EC in the research area, it seems reasonable to use )min for the estimation of SOM and POM.Indeed, the achieved values of ([OC]/[EC])min for Warsaw were very low, close to 1. Minimum values of OC/EC for both seasons appeared on those days, when the conditions were unfavorable for SOM formation (e.g., in comparison to consecutive days): low solar radiation, high wind speed and low ozone concentration.The prevalence of traffic emission, that justifies the ([OC]/[EC])min assumption, is also supported by the fact of high OC/EC values occurring more rarely than lower ones in both seasons over the research area.This also results in a low average value of 24-h OC/EC ratios for both seasons (about 3) and the standard deviation reaching 1.In other regions of Poland, where the emission from biomass and coal combustion dominates, ([OC]/[EC])min is usually not lower than 3 (Rogula-Kozłowska et al., 2014;Błaszczak et al., 2016), and for this reason the estimation of organic matter fractioning into SOM and POM for Warsaw air seems to be more accurate with the use of OC/EC ratio than for the above mentioned regions.For the ratios of ([OC]/[EC])min higher than 2 it would be difficult to prove, that their values were influenced either by organic matter from fossil fuels and biomass combustion (reaching OC/EC ratios considerably higher than 1) or by photochemical reactions of SOM gaseous precursors (Schauer, 2003;Boogaard et al., 2011).
As a ratio for POM estimation, based on POC, the value of 1.4 was adopted for the whole observation period.Because the ratio of 1.4 can describe primary and less oxidized organic mass, typically observed in winter, the same conversion factor (1.4) was assumed to estimate SOM (basing on SOC) for the winter period.Higher values of the conversion factor are generally reported for summer season (up to 1.8 (Aiken et al., 2008)).Therefore, we assumed the summer conversion factor for SOM (based on SOC) to be equal to 1.8.The above methodological approach could guarantee to achieve in Warsaw similar results to those observed for other European countries (Crippa et al., 2014).Except for that, it would enable to minimize the underestimation of SOM in PM 1 mass, resulting from overestimation of POM in summer mass of PM 1 (the problem of a proper ratio of POC/EC).However it should be stressed, that in Poland there exists different structure of PM sources and precursors than in other European areas.The main energy source in Poland is hard and brown coal combustion.The emission from energy production is thought to be the prevailing one and can veil the influence of other emission sources of PM, depending on the country region.Besides, there has been no knowledge available to authors so far, neither on seasonal changes of SOM main precursors concentration, nor qualitative and quantitative SOM composition in Warsaw.That is why the above assumptions, basing only on available literature data, had to be made for both ([OC]/[EC])pri and the detailed composition of OM, and finally yield a conversion factor proper for a certain location, aerosol origin and research period.SIA consisted of SO 4 2-, NO 3 -, and NH 4 + and Further analysis of the data utilized the methodology described by Thurston and Spengler (Thurston and Spengler, 1985), which combines a factor analysis (FA) to identify the possible sources and a multi-linear regression analysis (MLRA) to quantify their contribution to PM concentrations.
The first step in this stage was the principal component analysis (PCA) to the 8 × 120 data matrix of the PM 1components concentrations representing the 24-h concentrations of PM 1 -bound EC, SOM, POM, Cl -, NO 3 -, SO 4 2-, Na + , NH 4 + .Subsequently, the values of new variables -principal components (only principal components with eigenvalues > 1.0 were considered according to Kaiser criterion) were used in MLRA.All the computations were performed using Statistica 12.0.

Relationship between 24-h Concentrations of PM 1 and 24-h Concentrations of Other Pollutants and Meteorological Factors
Twenty-four-hour (24-h) concentrations of PM 1 show a typical pattern of the study area: they are lower in warm months (July-August) and higher in the cold period, January-March (averaging 11.1 ± 3.3 µg m -3 and 17.4 ± 8.4 µg m -3 , respectively; Table 1).During summer the PM 1 mass is a major contributor to PM 10 , averaging about 60% of PM 10 mass.On the contrary, the coarse fraction prevails in winter when the PM 1 contribution to PM 10 tends to be lower (40% on average; Table 1).Mean annual and seasonal (winter and summer) concentrations of PM 1 in Warsaw were definitely lower than those observed over cities in the south of Poland (Katowice, Racibórz, Wrocław) but higher than observed in the north of the country (Gdynia) (Table 2).The concentrations of PM 1 in Warsaw did not differ much from the values given for other European cities and thus were considerably lower than the values obtained in Asia (Table 2).During Warsaw measurement campaign, a similar seasonal variation to that of the 24-h concentrations of PM 1 was also observed for gaseous pollutants subject to continuous monitoring: SO 2 , NO 2, NO x and O 3.An evident increase in the 24-h concentrations of SO 2 , NO 2 and NO x was observed in winter months (Jan-Mar) as compared with summer months (Jul-Aug).In the case of O 3 the trend was the opposite -in warm months the concentrations of O 3 were approx.44% higher on average than in cold months.This results from higher air temperatures and a greater solar radiation in summer, which increases the rate of ozone formation in the atmospheric boundary layer (Rozbicka et al., 2014).
Evident seasonal variability of pollutant concentrations is mainly correlated with the annual course of air temperature.Mean air temperature in Warsaw for a long-term period of 1960-2016 was 8.5°C with the lowest mean observed in January (-2.3°C) and the highest in July (19.1°C)(archive data of the Division of Meteorology and Climatology of the Warsaw University of Life Sciences).2015 was one of the warmest in years 1960-2015 when mean annual air temperature reached 10.6°C, the mean air temperature in summer (Jul-Aug) was 20.3°C, and 2.0°C in winter (Jun-Mar).
Air temperature affects the concentration of airborne pollutants in two ways: by participating in the development of conditions favouring their dispersion (which is connected with the height of the mixing layer) (Schäfer et al., 2006), and by influencing the scale of household heating practices (Rogula-Kozłowska et al., 2014).The high mean temperature in both measurement seasons resulted in rather low mean concentrations of PM 1 over Warsaw in 2015, which were similar to those observed in previous years over cities of Northern and Western Europe, rather than to the values measured in Southern Polish cities (Table 2).This conclusion applies to the winter season, when mean concentration of PM 1 observed over Warsaw in 2015 was a few times lower than the values determined for Zabrze, Katowice or Wrocław in the winter of that year.
Besides air temperature, precipitation also influences the concentration of pollutants in the air, especially of particulate matter (Olszowski, 2016;Widziewicz et al., 2016;Olszowski, 2017).During the above -mentioned period of measurements, which lasted 120 days altogether, on 60 days precipitation was greater than 0.1 mm.A clear decrease in pollutant concentrations was observed when daily precipitation exceeded 5 mm: the 24-h concentrations of PM 1 and PM 10 were on average 30% and 40% lower, respectively, than on dry days.The washing effect of precipitation in the period of measurements was also observed for NO 2 , NO x and SO 2 -their mean concentrations on days with precipitation were respectively 24%, 30% and 30% lower than on days without precipitation.In the case of ozone, its concentration on such days was 23% higher on average.
In the assumed 120 day period of observations, the 24-h concentrations of PM 1 , PM 10 , NO x and NO 2 decreased by 43%, 25%, 26% and 24%, respectively when a mean daily wind speed exceeded 5 m s -1 .However no significant difference was observed in the daily concentrations of other pollutants when the wind speed increased.At the time, the prevailing wind direction in Warsaw was SW (28.6%) and E (21%), followed by S and (13.4% in each case).The respective fractions of other wind directions: NE, SE, W and NW were 9.2%, 7.6%, 5.9% and 0.8%.
The highest concentrations of PM 1 in Warsaw coincided with the following wind directions: S, SE, N and NE in winter, and NE, E and S in summer (Fig. 2).To the SE and S of the measurement site there was a dense network of streets with a rather heavy road traffic.The minimum distance of such streets in those directions from the measurement site was about 500 m (SE) and 1000 m (S).In those directions, there were also small local house boilers and individual household furnaces and further to the south there was a polluted region of Silesia.In the east and south west, there were mainly residential buildings with individual furnaces and oil and gas boiler plants.The very centre of Warsaw, along with its developed traffic system, was north and north-east of the measurement site.Moreover, in the NE sector the largest Polish (and Europe's second in size) heat and power generating plant, Siekierki, was situated.
A statistical analysis of the relationship between meteorological parameters and the 24-h concentrations of PM 1 , PM 10 , SO 2 , NO 2 and NO x (for the whole period of measurements) confirmed the significant influence of air temperature and precipitation on the patterns of occurrence of those concentrations over Warsaw.The 24-h concentrations of PM 1 , PM 10 , SO 2 , NO 2 and NO x were generally closely correlated with a majority of meteorological parameters (the correlation was statistically significant), with Pearson correlation coefficients ranging from 0.35 to 0.74 (Table 3).The 24-h concentrations of the pollutants were also significantly correlated with one another, so that the greatest correlation was noted between PM 1 and PM 10 (r = 0.74), followed by PM 10 and SO 2 (r = 0.65) (Table 3).
An analysis of the correlation between ambient particles and temperature revealed different seasonal patterns for PM 10 and PM 1 .A positive correlation with temperature of higher coefficients appeared in summer for both PM 10 and PM 1 , which is associated with stronger convection and unstable atmospheric conditions.In summer, when temperatures and solar radiation are high, and when ozone concentration increases, reactions of gaseous precursors of PM occur in the atmosphere more intensively (Seinfeld and Pandis, 2012).A negative correlation between PM 1 and PM 10 in winter suggests an inverse relation with the temperature and seems to support the hypothesis of a reduced dispersion and stable atmospheric conditions, however the correlations are small (not statistically significant).However, the influence of temperature on PM concentrations is rather complicated.On one hand, vertical mixing, which causes dilution of particles in the boundary layer, is affected by the change in temperature and wind with height.Generally, vertical mixing becomes stronger in hot seasons as the temperature is also higher.On the other hand, temperature impacts the chemical formation of particles.Higher temperature accelerates the reactions to form more products that can partition into particulate phase, but meanwhile the products tend to remain in their gaseous phase when temperature is too high (Hu et al., 2008;Miao et al., 2015).
The concentrations of particulate matter correlated negatively with relative humidity, irrespective of the season of the year, but with a relatively greater coefficients in winter.Pollutants may be scavenged by fog or cloud droplets and deposited onto surfaces, leading to lower ambient concentrations (Olszowski, 2016(Olszowski, , 2017)).However, a far more effective process of pollutants' removal, contrary to precipitation, is condensation and cloud formation through forced and ascent airflow in a frontal zone In addition, the influx of fresh and clean air can be observed in such circumstances.Therefore, the concentration of PM does not change even if it rains, since there is no advection of cleaner air.For this reason, the observed decline in concentration cannot be attributed solely to precipitation, considering also the fact that PM's are pushed away by the rush of air before falling droplets.
Lastly, PM and wind speed are also negatively correlated, in both seasons.Higher correlation coefficients were observed in winter (Table 3).
The search for direct relationship between PM concentration and meteorological parameters has a complex character as suggested by the above mentioned results.For this reason, multivariate regression models were built at the next stage of research.Such models, in which all predictors are simultaneously applied, show much better agreement.Table 4 presents optimal empirical models for both the winter and the summer period as well as for the whole research period.
The model obtained for the whole research period proves, that air temperature and wind speed had the greatest effect on PM1 concentration.It was also found that PM 1 concentration was negatively correlated with both air temperature and wind speed.The multivariate regression coefficient for that model was equal to 0.601 indicating a moderate correlation.However, the model built for the summer period showed the PM 1 concentration to be positively correlated with air temperature only, while the multivariate R reached 0.675.Considerably higher value of that coefficient was found for the winter period (0.748) which incorporated two parameters: the wind speed and radiation as exerting the highest influence on PM concentrations.Like for the whole analysed period, the wind speed was also negatively correlated with PM concentration.Neither did the solar radiation.

Analysis of Meteorological Conditions on Days with High and Low 24-h Concentrations of PM 1 in Warsaw -Air Mass Trajectories Days of the Highest Concentrations of PM 1 in Winter and Summer
For two days: one in February 2015 (winter season) and one in August 2014 (summer season) 48-hour back trajectories were computed using the HYSPLIT_4 model (Hybrid Single Particle Lagrangian Integrated Trajectory) (Draxler and Rolph, 2012) (Fig. 3).They began at the sampling points, 50 m, 200 m and 500 m above the ground level, at 00:00 UTC.The two days were selected because of their highest 24-h concentrations of PM 1 in Warsaw which amounted to 39.0 µg m -3 in winter (5 February) and 22.6 µg m -3 (3 August).On 5 February 2015, Poland was being released from the influence of the Atlantic high pressure wedge: the maximum pressure was 1032 hPa, the average air temperature was -0.4°C and the wind speed did not exceed 1.5 m s -1 .The mixing layer height (MLH) ranged from 30 m to 350 m and a slight temperature inversion was noticeable [http://ready.arl.noaa.gov/HYSPLIT_traj.php].
Back trajectories, generated for 5 th February, indicated a period of dynamic weather conditions, with the flow of clean air from the west or north, the high pressure system is gradually extended.The wind is weakening and in the evening when low inversion is formed, impurities slowly undergo dispersion and the smog in a regional scale is created.The centre of the high pressure system is moving over Atlantic.Above the inversion level, the warmer air from the sector SW-W is flowing, which further strengthens the inversion.The inversion thickness is of major importance for the formation of PM concentrations, lower wind speed  and the precipitation are rather marginal.This is classic smog situation with the influx of contaminated air from Upper Silesia and Malopolska.These impurities can reach Warsaw within twenty hours.
Similar synoptic situation influences high concentrations in the summer period on 3 th August 2014.The high pressure system from NW Russia is gradually extended over the analysed area and the Eastern Poland, which form an obstacle for a formerly existing and weakening low pressure system over the Western Poland.A dry and hot air inflows from the territory of SE Romania and Ukraine, that contributes to the occurrence of high temperatures (25.6°C), a small wind speed and lack of cloudiness.At the same time, an inversion mixing layer of a considerable thickness is developed (up to 3000) during the day, but rapidly decreasing in the afternoon hours, reaching only 130 m. at night.The above synoptic situations contribute to high pollutants' concentration especially at night, which is the result of weak dispersion and finally pollutants cumulation in the ground layer of the air.

Days of the Lowest 24-h Concentrations of PM 1 in Winter and Summer
The days on which 24-h concentrations of PM 1 were the lowest were 12 January 2015 (winter) and 26 June 2014 (summer) (Fig. 4).On 12 th January, Poland was under the influence of a deep, low -pressure mass as a result of a dynamically developing western circulation from the Atlantic.This was related with high pressure gradients, stronger wind and precipitation, mainly rainfall.Through the inflow of a mild and humid air as well as increasing mixing layer height up to 3000 m. during the day, followed by only slight inversion at night, very good dispersion conditions were created, resulting in a low pollutants' concentration, amounting to 4.7 µg m -3 in the case of 24-h PM 1 concentration.
On 26th June 2014 there was a stagnant character of two pressure systems: the high pressure system from the Norwegian Sea that blocked the zonal, western circulation and weakly developing low pressure system over Russia.It caused the inflow of a clean and cold air from the territory of Finland and the Norhtern Baltic Sea.The presence of inversion during the day, and the increase in the mixing layer height to 2000 m. in the afternoon, contributed to good conditions for pollutants' dispersion which was manifested through low PM concentration, reaching (7.0 µg m -3 ) on that day.

Chemical Composition and Origin of PM 1
On average, 62.2% of the mass of PM 1 in Warsaw comprises secondary matter: inorganic (SIM) and organic (SOM) (Fig. 5).Particularly high contribution of secondary matter in the PM 1 mass can be observed in winter (heating season), when average ambient concentrations of SIM and SOM were 6.4 µg m -3 and 5.6 µg m -3 , respectively (Table 5), thus making up 36.8% and 32.3% of the total mass of PM 1 .During the summer period, SIM and SOM constituted 27.8% and 23.2% of PM 1 mass respectively, which is 10% lower in comparison to winter.There are few reasons for this situation, one of which is a probable underestimation of SOM mass for the summer season or its overestimation for winter, resulting from the assumptions made for SOC and SOM estimation (Eqs.( 2)-( 5)).It should be stressed here, that the conversion factor should be considerably higher for very warm seasons in comparison to the other ones (e.g., winter), even exceeding the value of 1.8, owing to photochemical reactions (Turpin and Lim, 2001;Aiken et al., 2008).In winter period, however, SOM formation becomes more difficult because of meteorological conditions, and the resultant conversion factor might be lower than 1.4 (Turpin and Lim, 2001;Chou et al., 2010).On the other hand, the estimation of SIM assumed that the whole mass of SO 4 2-, NO 3 -and NH 4 + originated from secondary inorganic compounds (mainly NH 4 NO 3 -and (NH 4 ) 2 SO 4

2-
).Such an assumption would be valid for one season only, neither for a second one (Hu et al., 2016;Li et al., 2016).However, the same nature of relationship, meaning higher concentrations in the air and higher fractions in PM 1 mass at the same time, referred to both: SOM and SIM.Moreover, the previous research realized for other regions of Poland made evident, that the organic matter fraction in the mass of fine PM and/or ambient concentration of PMbound SOC in winter could be higher than in summer ) of the main primary components (EC -elemental carbon, POM -primary organic matter, Na_Cl -sodium and chlorine concentration total), secondary components (SOM -secondary organic matter and SIM -secondary inorganic matter) and unidentified matter (Other) in Warsaw.
Table 5. Basic statistics of 24-h concentrations (µg m -3 ) of the main primary components (EC -elemental carbon, POMprimary organic matter, Na_Cl -sodium and chlorine concentration total), secondary components (SOM -secondary organic matter and SIM -secondary inorganic matter) and unidentified matter (Other) in Warsaw; three averaging periods.* The mean values are statistically different in both seasons, p < 0.05 (U Mann-Whitney test at p assump = 0.05).(Błaszczak et al., 2016;Witkowska et al., 2016;Pyta et al., 2017).Then it is possible to claim the achieved results to be valid, and attributable to the study site, despite being different from those reported for other parts of the world.More intensive energy production in Poland during winter is almost all based on hard and brown coal combustion, and becomes a major source of winter SOM and SIM gaseous precursors.Even if their changes could be less intensive in winter than in summer, it is likely the difference in the emission between those seasons, that caused average secondary matter fraction in fine PM over Warsaw to be considerably higher in winter .

Period
The fraction of primary matter (a sum of EC, POM and Na_Cl) in PM 1 over Warsaw was more or less stable throughout the year, averaging 22.6%.In summer POM prevails in the mass of primary matter related to PM 1 , whereas in winter the principal components are EC and Na_Cl (Fig. 5, Table 5).Similar conclusions were drawn in 2010 from a research conducted in cities in the south and north of Poland, where the considerable share of EC and Na_Cl in the mass of fine particulate matter in winter was attributed to the combustion of coal for heating, mainly in local boiler plants or rooms and household stoves (Rogula-Kozłowska et al., 2012, 2014).POM prevails in the mass of PM 1 -bound primary matter during the non-heating period probably due to significant contribution of road traffic emission to PM concentration and its components (Rogula-Kozłowska et al., 2012, 2014).POM in summer is prevailing in term of relative composition, but it is higher also in terms of absolute concentrations, with respect to winter.Because the traffic emission of POM in terms of absolute mass is probably not higher in summer than in winter, it seems that higher summer POM concentration over the research area is attributable to primary biological aerosol particles (PBAP) -bacteria and archaea, fungal spores and fragments, pollen, viruses, algae and cyanobacteria, biological crust sand lichens and others like plant or animal fragments and detritus, etc.Although plant spores and pollen are usually large particulates (> 20 µm), the mold spores, fungal spores, bacteria, viruses and some allergens have their aerodynamic diameter lower than 1 µm (Després et al., 2012;Clauß, 2015;Wolf et al., 2017).That fact may be supported by our observations, because during the summer period measurements the sampled PM 1 on filter surfaces was noticed to take a yellow or yellow-brown color, that was entirely different from the one of PM 1 sampled in winter.
The fraction of unidentified ('other') matter in the mass of PM 1 was very noticeable in summer but negligible in winter.This also supports previous conclusions from earlier research, where it was demonstrated, that in summer in Poland's urban areas considerable amounts of road dust (containing mainly certain metal oxides, not identified in the study; Rogula-Kozłowska, 2016) were present in both coarse and fine particles.On average, up to 19% of the mass of PM 1 over Warsaw, irrespective of the season, may comprise water built in the structure of other compounds (Rogula-Kozłowska et al., 2017).Therefore, the differences in the chemical composition of PM 1 in summer and winter are evident and have been ascertained by checking the hypothesis of the equality of 24-h concentrations of EC, SOM, POM, SIM and Na_Cl; using the U Mann-Whitney test.The hypothesis was rejected (p < 0.05 at p assump = 0.05), and thus statistically significant differences between 24-h concentrations of PM 1 components in summer and winter were confirmed (Table 5).Their presence is connected with different origins of PM 1 in both seasons.Therefore, the identification of PM 1 sources using principal components analysis (PCA) (Majewski and Rogula-Kozłowska, 2016) was carried out for both seasons separately.In either case (summer and winter), the PCA was performed for eight variables (PM 1 components) and 60 cases (60 daily measurements).For the analysis, in order to identify the sources of PM 1 , one of the principal components analysed -SIA -was divided into three water-soluble ions making it, i.e., SO 4 2-, NO 3 -and NH 4 + , considering, that although these compounds (or rather their precursors -sulphur and nitrogen oxides and ammonia) react with one another in atmospheric air and form parts of the main portion of SIM (Sainfield and Pandis, 2006;Chow et al., 2015), they come from different sources.The presence of nitrogen oxides is related to emissions from road traffic (exhaust emission) and, to a lesser extent, from the combustion of natural gas, whereas sulphur oxides in Poland's urban areas mainly come from the combustion of hard coal and lignite, and less often from industrial processes (Dębski et al., 2015).In general, the concentration of ammonia over the research area, as well as other European regions, tends to be related to agriculture and to a lesser extent to road transport (Werner et al., 2015;EEA, 2017).
Instead of 24-h concentrations of Na_Cl, for the PCA the concentrations of Na + and Cl -were used (Table 6).The obtained PCA models for summer and winter were evidently different.The PCA model for summer identified three main components with the following fractions in the total variance: 39%, 26% and 21%.Cl -, SO 4 2-and Na + demonstrated an evident and the strongest correlation with the first component PC1.These are the components of fine particles, occurring in the air over Poland as a result of the combustion of coal and lignite (Rogula-Kozłowska et al., 2015;2016;Majewski and Rogula-Kozłowska, 2016).In summer there is also considerable increase of Na + , Cl -and SO 4 2-concentrations, attributable to inflow of air masses from northern directions (Fig. 6).It should be mentioned, that except for the relationship of those directions with the inflow of oceanic air masses, those directions are also supposed to coincide with air masses polluted by coal combustion products.Both at the northern and the southern direction from the research area large heat and power plants are located, using coal and lignite (Siekierki combined heat and power plant in the NE and major plants of the Upper Silesia in the SE).In the south and south-west, on the other hand, residential estates are situated, where some houses may be heated using coal as well.Also the increasing share of Na + and Cl -in the mass of PM 1 (Fig. 5) and the greater average concentrations of Na + and Cl -in winter vs. summer (Fig. 6) suggest that the concentrations of Na + and Cl -are influenced by coal combustion emissions.The same follows from negative, statistically significant correlations 13 Total variance 84% Elements with factor loadings < 0.30 are not included.Elements presented in descending order of their factor loads, with factor loadings indicated as subscript.* The sets of measured PM 1 daily/24-h concentrations and the concentrations computed for each day from MLRAdetermined contributions were substantially correlated (R 2 = 0.93).
between 24-h concentrations of Na + and Cl -and the daily air temperature in the analysed period (Table 7).Therefore, it can be assumed that PC1 reflects emissions from energy production and in summer it has a huge influence on PM 1 concentrations in Warsaw.
The second component, PC2, is correlated the strongest with EC, POM and NH 4 + .The distribution of 24-h concentrations of these components follows distinctly different patterns in summer and winter (Fig. 6).In summer, their highest concentrations are observed for S, SW and E wind directions, whereas in winter for SE and northern directions (in the case of NH 4 + ).Considering that busy arterial roads are situated in SW and E (the very centre of Warsaw), in summer these components must be connected with traffic emissions and in winter with additional energy emissions from commercial sites and households, as it is with Na + , Cl -and SO 4 2-.This is also evident in the share of EC in PM 1 mass (Fig. 5), which increases slightly in winter, indicating the presence of an extra source of EC.Moreover, there is a positive correlation of 24-h concentrations of POM with temperature, which suggests that POM content in the air increases with temperature, as the fraction of traffic emissions increases; this kind of correlation is not observed for EC (Table 7).The third component, PC3, is the strongest correlated with NO 3 -and SOM, representing the fraction of PM 1 precursors in the air as the main source of PM 1 .
In the PCA model for winter, only two principal components were identified (Table 6).PC1 was strongly correlated with all analysed components, except for Na + , whereas PC2 was correlated the strongest with Na + and then with EC and POM.Taking into account, that salt is used as the ice-removing agent on the roads in cities during winter, traffic emissions could be attributed to PC2 (in the PCA model for winter), whereas a single source of PM is difficult to find for PC1.Undoubtedly, there are a few sources of PM 1 connected with the combustion of various fuels in PC1, however it is hard to define and identify them.Therefore, further analysis of the data -with the purpose to attribute percentage fractions for 24-h concentrations of PM 1 to sources identified as the principal components -used a PCA model developed for all 24-h data together, that is for the entire period of measurements (Table 6).Fig. 7 shows 120-day-long course of fractions of PC1 and PC2 in PM 1 concentrations throughout the period of measurements in Warsaw.What follows from this pattern, and from the interpretation of the variability and correlations of the concentrations of PM 1 components given above, is that PC1 can be described as energy emissions (commercial and municipal together) and PC2 as traffic emissions.Therefore, the average share of emissions from heat and power generation in PM 1 concentrations in Warsaw can be estimated at more than 50%, whereas traffic emissions account for 15% (Table 6).In summer, the share of traffic emissions is greater than in winter: 18.3% (summer) vs. 13.3%(winter), along with a smaller share of energy emissions: 43.6% (summer) vs. 57.9%(winter).It should be noted that the shares determined in this article result not only from primary particles of PM but also from its precursors, which are included in the compositions of SOM and SIA.Therefore, they are most probably estimated more accurately than what usually results from the analysis, for example, PM elemental compositions.Still, the results obtained from this research are consistent with those previously obtained during an analysis of a monthly PM 2.5 data series for winter (Majewski and Rogula-Kozłowska, 2016).
All the other -unidentified here -sources of PM 1 in Warsaw, which total average fraction in the concentration of PM 1 was approx.35%, include industrial emissions, soil and traffic dust, as well as foreign emissions.

CONCLUSIONS
For the first time a diurnal course of principal components  fraction in PM 1 has been presented in a long-term data series (120 days) for an urban area in Poland.The fraction of primary and secondary matter in the mass of PM 1 was analysed for each day of the observation period.It was found, that secondary aerosol accounts for approx.62% of the PM 1 mass over Warsaw.In summer, the average is 51% and in winter it is 69%.Secondary inorganic matter (SIM) constitutes nearly 57% of the secondary aerosol in winter, and 56% in summer.The respective fractions of secondary organic matter (SOM) are 43% and 44%.This means that almost 6 µg m -3 in summer and 12 µg m -3 in winter are submicron particles originating from transformations of volatile organic compounds, sulphur and nitrogen oxides, and ammonia.Therefore, limiting the emissions of primary PM (soot from diesel powered engines and household furnaces, dust from road surfaces wearing off, primary metallic particles from industrial plants, bacteria, spores and plant litter, soil matter etc.) will not result in a significant reduction of PM 1 concentrations over Warsaw.However, curbing emissions from some of the abovementioned sources will definitely limit PM 1 concentrations.On average, 15% of PM 1 in the air comes from traffic emissions, that is mainly from transformations of NO x and VOCs.Almost 51% comprises particulate matter connected with emissions from the combustion of various fuels to generate power and heat, both in households and in combined heat and power plants.Such a reduction will target gaseous precursors of PM and not the PM itself.So far, the attempts to determine the balance of emissions from various sources and to translate drops and rises in PM emissions into falling or growing concentrations of PM in the air over Poland, have assumed, that all the particulate matter is primary.Yet, this paper indicates, primary particles have a much smaller share in the concentration of submicron particulate matter than particles coming from transformations of gases, which have been overlooked in such balancing.Moreover, it should be also stressed that seasonality (winter and summer period) plays a significant role in the intensity of PM formation from its gaseous precursors.As a matter of fact, it determines the final share of SIM and SOM in PM.

Fig. 2 .
Fig. 2. Distribution of the mean concentrations (µg m -3 ) of PM 1 with regard to eight wind directions.

Fig. 6 .
Fig. 6.Distribution of the concentration of selected components of PM 1 (µg m -3 ) with reference to the eight wind directions.

Fig. 7 .
Fig. 7. 120-day-long course of share of A -energy emissions (commercial and municipal sources: boiler plants and household heating furnaces) and B -traffic emissions, in the 24-h concentrations of PM 1 in Warsaw.

Table 1 .
Basic statistics of the series of 24-h concentrations of PM 1 , PM 10 , gaseous pollutants and meteorological conditions for each season and the whole sampling period.
c Precipitation presented as the total precipitation during the reporting period.S.D. = standard deviation.Valid N = daily data.Min.-minimum value.Max.-maximum value.

Table 2 .
Comparison of mean concentrations of PM 1 in Warsaw with PM 1 concentrations in previous studies available in the literature for similar sampling sites.PM 1 in µg m -3 .

Table 3 .
Pearson correlation coefficients between 24-h concentrations of PM 1 and PM 10, gaseous pollutant and 24-h averaged meteorological factors in Warsaw.

Table 4 .
Regression result for the optimal empirical model of PM 1 in Warsaw.

Table 6 .
Results of principal components analysis (PCA) and a multi-linear regression analysis (MLRA) performed for the concentrations of PM 1 and PM 1 -bound elements.

Table 7 .
Pearson correlation coefficients between 24-h concentrations of PM 1 components and daily averaged meteorological factors in Warsaw.