Time series analysis of the air pollution around Ploiesti oil refining complex, one of the most polluted regions in Romania

Refineries and petrochemical industries are known to be the principal sources of emissions for a number of air pollutants, such as Volatile Organic Compounds (VOCs), greenhouse gases and particulate matter, which negatively affect the air quality. The primary goal of this research was the time series analysis of PM2.5, PM10, As, Cd, Ni, Pb, benzene, toluene, ethylbenzene, o-xylene, m-xylene, p-xylene, CO, NO, NO2, NOx, SO2 and O3 over an eleven-year period (2009–2019) and the connection between air pollution and meteorological parameters (air temperature, precipitation quantity and relative humidity). Regarding the pollution level of the major pollutants, the minimum pollution levels, except SO2 and O3, were recorded during warmer periods, meanwhile increased levels, were detected during the cold period (in winter). The air pollutants’ concentration and distribution are affected by meteorological parameters, such as wind speed and direction, rainfall or even relative humidity. Therefore, the highest concentrations in the winter season were 1.25 times higher than in autumn, 1.3 times higher than the average annual value, 1.57 times higher than in spring and 1.79 times higher than in summer. Monthly variation of O3 showed lower concentration during winter (27.62 µg/m3) and higher in summer (46.42 µg/m3). Based on the statistical analysis, a significant Spearman correlation was detected between the studied air pollutants and meteorological parameters, and according to the Principal Component Analysis (PCA) and cluster analysis, some common sources were also detected.

www.nature.com/scientificreports/ Fine particulate matter (PM 2.5 ) with an aerodynamic diameter of up to 2.5 µm, may pass via ingestion, inhalation and dermal absorption causing various health issues, such as respiratory and cardiovascular morbidity and lung cancer 2,10,11 . According to the World Health Organization (WHO) guideline, the PM 10 and PM 2.5 annual concentrations must be under 20 µg/m 3 and 10 µg/m 3 , respectively 12 . Based on the literature, strong correlations were revealed between elevated particulate matter and respiratory and cardiovascular mortality worldwide 13,14 .
The airborne particulate from industrial and traffic emissions may represent an increased health risk since different toxic trace elements and metalloids, such as arsenic, cadmium, molybdenum, nickel, sulfur, selenium, vanadium, zinc and lead are bound to PM [15][16][17] . Trace elements with important contributions in the development of carcinogenic diseases are Pb, Ni and Cd 18 .
Theozone as a secondary pollutant is formed during the reaction between VOCs andoxides of nitrogen in the presence of heat and sunlight. As a strong oxidizing agent, ozone may induce oxidative damage in the respiratory tract and lungs 19 . Nitrogen oxides may pass through the alveolar-cells (epithelium) and the adjacent capillary vessels of the lungs and destroy the alveolar structures and their function in lungs [20][21][22] . The petrochemical industry can be the main source of VOCs and NO x , which are involved in the ozone formation 23,24 . Several meteorological parameters, such as precipitation quantity, play an important role in pollutant concentration reduction through the washout effect, when air pollutants are washed out by rain, snow or fog 25 .
The main objective of this research study was to analyze the temporal distribution of the selected air pollutants, such as PM 2.5 , PM 10 , As, Cd, Ni, Pb, Benzene (C 6 H 6 ), Toluene (C 6 H 5 CH 3 ), Ethylbenzene (C 8 H 10 ), o-m-pxylene (C 8 H 10 ), CO, NO, NO 2 , NO x , SO 2 , and O 3 , and the variation of the meteorological parameters (precipitation quantity, air temperature and relative humidity) over an 11-year period (2009-2019) in Ploiesti, one of the most polluted regions in Romania.

Materials and methods
Study area. This study was conducted in the Ploiești city which is situated in the south-east of Romania with the following coordinates: latitude: 44° 56′ 24′′ N and longitude: 26° 01′ 24′′ E, at an altitude of 150 m (Fig. 1). The air pollution data were collected from 6 (PH1-PH6) monitoring stations and the average values were used for the statistical analysis. In order to compare the seasonal differences four seasons were defined: winter (December, January, February), spring (March, April, May), summer (June, July, August) and autumn (September, October, November). Ploiești has a temperate humid continental climate, with an average annual temperature of 10.5°C. The oil processing and refining industry is the main economic activity in this region. Regarding the emission of major air pollutants, several industries are present in Ploiești city, hence the emission is mainly related to different industrial activities, such as oil refining, oil extraction machinery and equipment, chemical and manufactured fiber and metalworking facilities as well 6 .  where CAGR -compound annual growth rate, V begin -beginning value, V final -final value, t-time in years.
Descriptive statistic calculations were carried out by using the monthly mean values of the 6 monitoring stations. The multiannual monthly variation of pollutant concentrations was presented using the box plot diagram, where the samples were divided into four quartiles. Spearman correlation analysis was carried out by the widely used R (Ri386 3.6.3.) statistical program in order to decipher the relationship between the studied pollutants and meteorological parameters. Furthermore, hierarchical cluster analysis is a commonly used multivariate statistical method in environmental sciences, therefore the classification was done by using the IBM SPSS Statistics 22 program's hierarchical cluster analysis method (Ward linkage, Euclidean distance) and the results were visualized in a dendrogram. Principal Component Analysis (PCA) was also implemented 38 , and before PCA analysis, the data were checked using the Keiser-Meyer-Olkin (KMO) test to verify how suited are the air pollution parameters for PCA (Table 1).

Results and discussions
The main aim of this research was to decipher the relationship between the major air pollutants and meteorological parameters, and to describe the monthly and annually trends around the Ploiești petrochemical industries. According to the statistical analysis of the data recorded at the six automated stations in Ploiești during the studied period was evident that air pollutants with the highest annual average concentration were CO, NO x , O 3 , PM 10 Table 2). The petroleum industry requires a huge amount of fossil fuels for operation, and as a consequence the air pollutants release into the atmosphere is significant and in many cases are detrimental to human health 2 . In Ploiești city the multiannual mean PM 10 and PM 2.5 concentrations exceed the acceptable limit by 51.75% and 83.4%, respectively. The heavy metal content was determined from the PM 10 fraction; the rank of metals was Pb > Ni > As > Cd, with 0.73, 0.53, 0.95 and 14.2 ng/m 3 , respectively. The results revealed the most toxic metal (Pb) had a multiannual average of 14.2 ng/m 3 and mainly originated from the intense oil and petrol refinery activity and heavy traffic. The possible Ni complexes are often salts, nickel oxides, nickel sulfate, nickel sulfide, nickel silicate and nickel chloride 39 .
Regarding the organic fraction, the average concentration of benzene determined in the Ploiești region for the studied period (2009-2019) was 3.25 (± 1.56) μg/m 3 , that of toluene 3.09 (± 1.36) μg/m 3 , that of ethylbenzene 0.503 (± 0.24) μg/m 3 and that of xylene (calculated as the sum of para-, meta-and ortho-xylene) was 1.94 (± 1.28) μg/m 3 . According to the literature, high VOC concentrations were measured near the petrochemical industries around the world 40 .
The highest concentration in case of gaseous pollutants was recorded for CO with 0.28 (± 0.07) mg/L, followed by NO x , NO 2 and NO with 45.97 (± 12.12) μg/m 3 , 27.37 (± 6.82) μg/m 3 and 12.28 (± 4.6) μg/m 3 , respectively. Analyzing the variation of climatological parameters, the multiannual average daily precipitation quantity was 3.22 (± 3.12) mm, the temperature 13.06°C (± 8.73) and the relative humidity 70.88 (± 9.6).  Based on the compound annual growth rate, the highest decreasing trend was observed in the case of As, Cd and Pb with 12.48%, 7.34% and 6.1%, respectively. Between the meteorological parameters the precipitation quantity was reduced by 8.83%, on the other hand, the relative humidity and temperature increased by 0.54% and 0.24%, respectively.
During the studied period, the concentrations of air pollutants decreased continuously due to the implementation of the new ambient air quality standards.
Monthly multiannual variation of pollutants. The multiannual monthly variation of the studied parameters in Ploiesti city are presented in Fig. 3. All studied pollutants, except SO 2 and O 3 , show minimum concentrations during warmer periods, and the maximum concentrations were detected in winter, due to increased emission thanks to the biomass and combustibles consumption in the cold season (heating period). Another important factor is the atmospheric stability during this period, which allows and favors the accumulation of pollutants in the lower atmosphere.
As we expected, the monthly variation of O 3 showed an opposite trend, lower concentration during winter (27.62 µg/m 3 ), and higher concentration in summer (46.42 µg/m 3 ). The reduced rate of photochemical formation of O 3 , is in strong correlation with reduced sunshine, lower surface temperature, and higher primary pollutants level. On the other hand, in summer the petrochemical industry emissions of hydrocarbons and NO x are significantly higher, which can result in ozone (O 3 ) formation 24 .
The analysis underlines the differences between the seasons: the highest levels of air pollutants were identified in winter season and was 1.25 times higher than in autumn (the minimum ratio of 0.97 was recorded in the case of m-xylene while the maximum of 2.15 for CO), 1.3 times higher than the annual concentration (1.11 as a minimum for m-xylene and 2 as maximum for CO), 1.57 times higher than in spring (the lowest ratio of 1.21 was observed for NO 2− -meanwhile the highest 3.29 for CO) and 1.79 times higher than in summer (with 1.

Spearman correlation analysis.
To decipher the relationship between the studied parameters, Spearman's correlation rank analysis was carried out. In total 117 monthly average concentration values were used, and the correlation coefficients between two parameters were considered statistically significant at P < 0.05 and r ≥ + 0.182 and r ≤ − 0.182 (Fig. 4). www.nature.com/scientificreports/ Based on the Spearman rank correlation results, a very high positive correlation was observed between the fine (PM 2.5 ) and coarse particulate (PM 10 ) matter (r = 0.82). All trace elements (As, Cd, Ni and Pb) determined from the PM 10 fraction show significant correlation of 0.34, 0.44, 0.28 and 0.75, respectively. Furthermore, the BTEX components correlate with each other very well, and the nitrogen oxides show positive correlation with the PM's, trace elements and BTEX components. Significant negative correlation was found between SO 2 and BTEX components, and a negative correlation was detected between the O 3 and all studied parameters, except BTEX. www.nature.com/scientificreports/ Furthermore, between PMs and other pollutants, such as CO, NO, NO 2 and NO x significant strong correlations were identified, except the BTEX compounds where moderate negative correlations were detected between O 3 and other air pollutants. A moderate correlation was found between SO 2 and BTEX, r = (− 0.41) − (− 0.56). Regarding the environmental parameters, a strong negative correlation was between air pollutants and environmental parameters, such as temperature and precipitation, while positive correlation was detected with relative humidity. Due to washout effect of precipitation the negative correlation between the precipitation quantity and particulate matters, BTEX and gases is obvious 42 . Further investigations need to be carried out in order to decipher and better understand the underlying processes behind strong and negative correlations as well.
Hierarchical cluster analysis. The cluster analysis has shown that four main clusters can be identified (Fig. 5). Cluster 1 contains four different sub-clusters: 1.1 contains the PMs and nitrogen oxides, the 1.2 are represented by the trace elements and SO 2 . The 1.3 sub-cluster includes the BTEX components, and in 1.4 subcluster is the O 3 . The meteorological parameters-precipitation quantity, air temperature and relative humidity-represent separately the 2, 3 and 4 clusters.
According to the dendrogram, the results suggest that the pollutants share some common sources, hence air pollutants in the same sub-cluster may have common origin 16 . In environmental studies hierarchical clustering analysis methods are the most frequently used approaches to decipher if the pollutants share common sources 43 . In case of CO, NO 2 and particulate matter emissions, the main responsible sources include the petrol, diesel and other alternative-fuel engines. Furthermore, the BTEX origin are mainly related to the crude oil refinery 44 .

Principal component analysis (PCA). Principal Component (PC) loadings for all studied air pollutants
and for the studied period with corresponding eigenvalues and variances are given in Table 3 and loading factors of air pollutants in three-dimensional spaces are presented in Fig. 6.
Through the PCA the data were transformed into the new coordinate system, the greatest variance by scalar projection is the first component, the second greatest variance is defined as the second component, and the same order is following in the third and the fourth component. According to the Principal Component Analysis, four components were extracted from the component matrix, accounting for 75.1% of the overall variance. Factor 1 contains PM 2.5 , CO, PM 10 , NO x , NO, Pb, NO 2 and RH, and represent 33.54% of the total variance. The factor 2 was represented by O 3 , o-m-p-xylene, ethylbenzene, benzene and toluene. The third and fourth components   (33.54) suggest that these are the major sources of pollutants in this area. The PC 2 (24.89%) are represented by the following air pollutants parameters: O 3 , o-xylene, m-xylene, p-xylene, ethylbenzene, benzene and toluene. The PC 3 with 10.38% total variance shows highest loadings for Cd and Ni. According to Pandey et al. Cd and Ni have significant contributions from crude-oil combustion and metallurgical industrial areas and motor vehicle emissions 45 . As shown in Fig. 6, the studied parameters which are localized near to each other are in the same cluster.

Limitation
Greenhouse gases (CH 4 , N 2 O, CO 2 ) were not studied in this article because they were not monitored during the studied period.

Conclusions
The aim of the present study was to identify the level air pollution and the relationship between environmental parameters and pollutants in one of the most polluted regions in Romania, in Ploiești, during the studied period (2009-2019). The results revealed that the application of time series analysis can give information on the temporal trends of pollutants near the oil refinery. In concordance with many other industrial activities, petrochemical industry and refineries have a negative effect on air quality. Compared to the first reference year (2009), except ethylbenzene, o-xylene and temperature, all studied air pollutants showed a decreasing trend during the studied years. Seasonal analysis of the pollutant concentrations revealed that the highest values were measured in cold season, especially in winter. Variation of O 3 and SO 2 showed an opposite trend, lower concentration in winter and higher concentration in summer. Based on the cluster analysis some pollutants may share common sources, air pollutants in the same sub-cluster may have common origin.
In order to minimize the human health effects of the air pollutants it is essential to respect the Air Quality standards, especially near refineries.
The statistical interpretation of the air pollution distribution can contribute to the development of new and more sector-specific and regional air quality standards regarding oil refineries to reduce air pollution.
The main reason why pollution levels decreased in the last years was mainly attributed to the implementation of the new air quality standards for the industry and on the other hand to the changes in transportation, more sustainable and eco-friendly. www.nature.com/scientificreports/