1. Introduction
Air pollution is a serious threat to human health (Heft-Neal et al. 2018; Zhao et al. 2019; Dedoussi et al. 2020). The degree of health risk depends on the intensity of the air pollutant peaks as well as on the duration of exposure. Indeed, repeated and prolonged exposure to even low concentrations can lead to high respiratory and cardiovascular morbidity (Lim et al. 2012). In 2016, the World Health Organization (WHO) estimated that over 90% of the world’s population lives in places that do not meet air quality standards (WHO 2018). The WHO also considers that polluted air is responsible for 4.2 million premature deaths worldwide. Since the first harmful effects of air pollution appeared, efforts to control the sources of this hazard have multiplied. Among these, we distinguish between mobile (different types of transport) and stationary (industry, household waste treatment, domestic activities, etc.) anthropogenic sources. Many studies acknowledge that the tendency of air pollutants to accumulate and persist is governed by many interrelated factors. Some studies mention the determining role of weather: they show that pollution peaks are closely associated with thermal inversion situations (Raatikainen et al. 2014; Zheng et al. 2015; Yuan et al. 2020) and that severe episodes are specifically related to atmospheric stability (Miao et al. 2017; Sun et al. 2019; Wu et al. 2019; Q. Li et al. 2020). Under such conditions, the accumulation of primary pollutants [PM10, PM2.5, NO2, volatile organic compounds (VOCs), etc.] in the urban boundary layer is particularly favored in the absence of wind (Zhong et al. 2018; Miao and Liu 2019). Indeed, several studies have highlighted the role of wind direction and speed in the degradation of air quality (Toro et al. 2019; Abbassi et al. 2020; Kikaj et al. 2020; Reiminger et al. 2020). Clear skies favor ultraviolet radiation, which triggers photochemical reactions resulting in high ozone (O3) concentrations (Vukovich and Sherwell 2003; Camalier et al. 2007; Duché 2013; Uttamang et al. 2020). Martin (2008) shows that the air becomes more ozone laden during anticyclonic events and tends to improve when the weather is disturbed. Extensive research shows that knowledge of the relationship between pollution and weather type improves when the approach used examines meteorological variables as a whole rather than isolating them in an individual analysis (Kalkstein and Corrigan 1986; Tselepidaki et al. 1995; McGregor 1999). By aggregating several meteorological variables, this method can reveal whether the state of the atmosphere is favorable for the accumulation of pollutants (Hodgson and Phillips 2021). Carnerero et al. (2019), Yuval et al. (2020), and H. Li et al. (2020) highlight that it is particularly important to distinguish between primary and secondary pollutants in the context of air pollution studies. Also, several distinct meteorological conditions, when predominant, can generate very different physicochemical processes from one pollutant to another.
In the Paris region, PM10 and O3 remain problematic (AIRPARIF 2019) and largely exceed the thresholds defined by the WHO. PM10, called inhalable particles, is a primary pollutant emitted directly into the atmosphere (anthropogenic and natural origin) (H. Li et al. 2020). In contrast, O3 is a secondary pollutant originating from the photochemical reaction of precursors, such as VOCs, carbon monoxide (CO), and nitrogen dioxide (NO2). Because of their adverse effects on human health and the environment, they are identified as target molecules by the WHO (WHO 2001). This study aims to determine the weather types associated with high concentrations of PM10, in the first instance, and O3, in the second instance. The area chosen for this study covers the territory of the Cergy-Pontoise conurbation (CACP), located northwest of the Ile de France. This choice was based on the fact that the area has been undergoing changes in the urban environment for more than 25 years. This dynamic is reflected particularly in urban sprawl. In 2018, the CACP had 210 633 inhabitants and 88 428 dwellings, 61.4% of which were flats, that is, an increase of 1672 dwellings per year on average since 2013.
In this paper, based on data provided by the Interdepartmental Association for the Management of the Automatic Air Pollution Monitoring and Alert Network in the Île-de-France region (AIRPARIF) and Météo-France, we first identified pollution episodes and assessed their frequency on a monthly and seasonal scale as well as their persistence. Then, we identified the weather types associated with these episodes. Finally, the heterogeneity of pollution episodes according to weather types was studied using descriptive, bivariate, and multivariate statistics.
2. Study area
The conurbation of Cergy-Pontoise, 49°02′20″N; 2°04′37″E, is located 30 kilometers northwest of Paris on the road linking Paris and Rouen. Its surface area of 84.20 km2 is roughly equivalent to that of inner Paris. Located in the Val d’Oise department (95), the study area is at the gateway to the Parc Naturel Régional du Vexin Français (French Vexin Regional Natural Park). The Oise, before flowing into the Seine, crosses the area for about 18 km, passing through the heart of Cergy, where it irrigates a 250-ha leisure center (Fig. 1). Cergy-Pontoise benefits from three Société Nationale des Chemins de Fer (SNCF) railway lines, as well as lines A and C of the Réseau Express Régional (RER). It is directly linked to Paris by the A15 motorway and to the regional network by the Francilienne (Île-de-France expressways). The A86 motorway provides rapid access to the Yvelines and the Oise, two departments bordering the Val d’Oise. The surface area of this conurbation is undergoing a process of land artificialization to the detriment of areas used for agriculture. In 2012, 57% of the CACP was artificialized (+9% between 1990 and 2012). The most developed area is located in the center of the conurbation with traces of human occupation dating back to Gallo-Roman times. This indicates a centrifugal urban sprawl in which the periphery appears less artificialized in 2021. Analysis of the land-use map confirms these findings by showing that the area of the conurbation is characterized by (Fig. 1) a very dense center, with a concentration of buildings (residential and socioeconomic services); a more rural northern and western fringe; and a major business area in the northeast–Les Béthunes business park.
Location and land use of the Cergy-Pontoise conurbation; the reference pollution measurement station (AIRPARIF) is indicated on the map by a yellow circle.
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
In 2018, there were 210 633 inhabitants (National Institute of Statistics and Economic Studies 2018). The share of the population considered vulnerable to the health effects of pollution is significant. It represented 38.3% in 2018. It is composed of young children (under 14 years old) and people over 60 years old. The poverty rate, which is also a factor of vulnerability, was 16.3% in 2018 based on tax references, that is, a rate 0.7% higher than that of the region. In parallel with the demographic increase, the conurbation experienced an expansion of its car fleet. In 2018, Cergy-Pontoise had 64 755 households with at least one car; 52.5% of the population use the car to commute to work and 37.5% use public transport (National Institute of Statistics and Economic Studies 2018). This relatively high use of motorized transport affects air quality.
The conurbation has a mild oceanic temperate climate. Daily maximum temperatures average 25°C in the summer months, while minimum temperatures fall to 2°C in the winter months. In absolute values, extreme temperatures exceed 35°C in the summer and drop to below 0°C in the winter. Precipitation is distributed more or less evenly over the year, with an average of 137 rainy days and 221 dry days, with an average annual total of 650 mm. The prevailing winds are from the southwest quadrant (the westerlies); mild and wet (Fig. 2a). Southwest winds predominate in winter and autumn. However, in summer and spring, west and northeast winds dominate respectively. In terms of speed, the annual average is 3.4 m s−1. Besides the calm, which represents 15%, light to moderate winds dominate (Fig. 2b). Indeed, the range 2–5 m s−1 represents 52% of the annual frequency, followed by the range 5–8 m s−1 (20%) and 1–2 m s−1 (10%). Strong winds, above 8 m s−1, represent only 3%.
(a) Annual and seasonal wind roses; (b) annual frequencies of each speed range (2009–20 data from Cormeilles en Vexin meteorological station; NOAA site).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
The high frequency of light winds and calm can be explained by the high occurrence of anticyclonic situations. Based on the pressure data measured from 2009 to 2020, 78% of the days had an average pressure exceeding 1013 hPa (20% in winter, 14% in spring, 15% in autumn, and 29% in summer).
3. Data and methods
Hourly data on PM10 and O3 concentrations were provided by AIRPARIF from 2009 to 2020. They were recorded by the urban station “Cergy-Pontoise.” This station is a conventional urban background station located in the heart of the conurbation in Cergy. Hourly meteorological data were used to determine the state of the atmosphere associated with air pollution peaks. These are METAR data from the station located in Cormeilles en Vexin, 5.6 km from the AIRPARIF station, and downloaded from the National Oceanic and Atmospheric Administration (NOAA) website. The data were selected to cover the same period as that of AIRPARIF. We selected the main meteorological parameters: maximum temperature Tmax (°C), minimum temperature Tmin (°C), wind speed V (m s−1), flow, relative humidity RH (%), visibility VIS (m), dewpoint DEW (°C), and atmospheric pressure SLP (hPa). We calculated the daily thermal amplitude ΔT (°C), which can indirectly provide information on cloud cover.
A preliminary treatment was carried out to identify the concentrations of PM10 and O3, which exceed the thresholds decreed by the WHO, that is, 45 μg m−3 as a daily average for PM10 and 100 μg m−3 as an 8-h moving average, for O3. The PM10 and O3 pollution sequences were identified and classified according to their duration into isolated cases, short and long episodes. Then, the temporal variability of these sequences was analyzed to identify the air pollution episodes and their persistence. Once the days on which the WHO standard was exceeded were defined, we identified the weather types associated with air quality degradation. For this purpose, we refer to Vigneau (2000) who defines weather types as a combination of stationary meteorological data influenced by local geographical conditions and justified by general traffic. To define these characteristic situations, we used the pressure maps (surface and upper air) archived by the German meteorological office “Wetterzentrale.de” to group days with the same isobaric configuration together. We also use the definition of Carrega (2004), which specifies that the weather type cannot be reduced to synoptic movement alone because the same circulation does not have the same effect on dissimilar geographical environments. This is why we carried out a separate analysis using the daily information bulletins provided by Météo-France to verify the occurrence of particular local phenomena such as mists and fogs. The combination of large-scale and small-scale climatic factors has made it possible to identify the types of weather expressed in Cergy-Pontoise that are responsible for the pollution episodes. The days of pollution were then classified according to a manual grouping of sequences and isolated cases with similar weather conditions (airmass flow, temperature, humidity, visibility, sky conditions, and isobaric configuration at ground level and 500 hPa). Thus, to validate the results of the manual method, an automatic classification of weather types was conducted using principal component analysis (PCA) and hierarchical clustering (HCA). The objective and subjective clustering procedures often used in Europe are detailed in Philipp et al. (2010).
Statistical descriptive and dispersion parameters for PM10 and O3 were calculated per time type to determine the distribution and dispersion characteristics within each time series. Each weather type would result in a different situation in terms of pollutant concentrations. To confirm the difference between the identified weather types, the Kruskal–Wallis method of analysis of independent samples was conducted. The null hypothesis (H0) of the test implies that these concentrations are identical, whereas the true hypothesis (H1) indicates the opposite. Using pairwise multiple comparison functions included in this test, a treatment was conducted to quantitatively compare the average air pollution levels between each pair of weather types. Then, a PCA and a bottom-up hierarchical clustering (BHC) were applied for the weather type that showed the greatest internal variation. A correlation test was used to determine whether the length of the sequences was disproportionately associated with certain types of air masses. All these statistical calculations were performed with SPSS (version 22) and XLSTAT (version 2021.3.1).
4. Results
a. Frequency study of exceedance days
1) The case of PM10
Processing the data recorded between 2009 and 2020 shows that the WHO standard was exceeded on 198 days, 29% of which occurred in March. The months of January and February account for an equal share of 18% of these days (Fig. 3a). In addition, we note 12.5% in April, 11% in December, and 6% in November. Values below 2% of exceedance days were recorded for the rest of the year. This is confirmed by the analysis of the seasonal distribution of exceedances (Fig. 3b). The largest number of polluted days is observed in spring and winter with a frequency of 47.3% and 42.2% respectively. In autumn, exceedances of the WHO standard are lower than in winter by a factor of 5. As for summer, their frequency is relatively negligible with only 0.5%.
(a) Monthly and (b) seasonal variability of the frequency of days on which the PM10 standard is exceeded (PM10 data from 2009 to 2020 from the AIRPARIF station).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
Intensity and persistence
Of the 198 exceedance days observed, some are isolated and others are part of a sequence of several consecutive days (Fig. 4a). The days that are not part of pollution episodes (isolated cases) correspond to 22% (43 days) of all exceedance days. Their frequency is higher from January to April, with an average of nearly 7 days month−1. They all have a concentration between 45 and 63 μg m−3. The total number of exceedance days that fall within the pollution sequences corresponds to 78% or 155 days. March accounts for almost 23% of consecutive exceedance days followed by January (15.5%) and February (15%). Sequences of 2–3 days represent 46% (91 days) and those of 4–7 days represent 32% (64 days) (Fig. 4b). Focusing exclusively on the length of the sequences, their monthly occurrence shows that the persistence of pollution depends on the month (Fig. 5). Long-duration sequences (from 4 to 7 days) occur mainly from November to April with a higher dominance in March and December. For the concentration, it appears that the sequences ranging from 63.7 to 162 μg m−3 (concentration higher than the 75% percentile) occur in winter and early spring; among them, 66% are part of the 4–7-day sequences and 34% are part of the 2–3-day ones. These results show that PM10 is a winter/spring pollutant with only three 2–3-day sequences during September and October, all with concentrations below 55 μg m−3. Four isolated cases in September and one in July were recorded, with a concentration slightly above 45 μg m−3.
(a) Monthly distribution of PM10 pollution sequences, (b) pollution persistence in number of days (PM10 data from 2009 to 2020 from the AIRPARIF station).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
Monthly distribution of the different sequence durations (PM10 data from 2009 to 2020 from the AIRPARIF station).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
2) The case of ozone
Ozone data also calculated over the period 2009–20 show that exceedances of the WHO standard occur mainly from April to September according to the scientific literature (Sun et al. 2018; Martin 2008). There is a peak in July (Fig. 6a). Ozone is more present in summer (284 days), spring (142 days), and autumn (32 days) and is absent in winter (Fig. 6b). This seasonal pattern of evolution is identical for the majority of years in which the exceedance level reached in summer is higher. In contrast, for the years 2011 and 2020, spring overtakes summer.
(a) Monthly and (b) seasonal variability calculated according to the number of exceedance days (O3 data from 2009 to 2020 from the AIRPARIF station).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
Intensity and persistence
The duration classes of the sequences were identified according to the discretization method of the natural thresholds. Of the 458 exceedance days (defined from the 8-h moving average,) 19.3% are isolated cases. Sequences consisting of several days represent 80.7% of all exceedance days. They are distributed over all the months of the photochemical season (Fig. 7a). The high percentage of these sequences as compared with the isolated cases within each month shows that the ozone pollution phenomenon appears mainly in the form of persistent episodes. The month of July alone accounts for 22% of the sequences. Within these sequences, 40% of the days have more than 10 hours of exceedance per day, which is also the case in June and August (25% and 19%). However, on average, the isolated cases have only 4 hours of exceedance per day. They are observed at 12.3% of total exceedances from April to June and at 7% for the other photochemical months.
(a) Monthly distribution of O3 pollution sequences, (b) pollution persistence in number of days (O3 data from 2009 to 2020 from the AIRPARIF station).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
According to the distribution of exceedances over consecutive days, three classes were defined: 2–3 days (short duration), 4–7 days (medium duration), and 8–13 days (long duration). The medium and short sequences are the most frequent with a percentage of 37% and 33%. On the other hand, long sequences do not exceed 10.7% of the total number of exceedance days (Fig. 7b). These occur in May and become increasingly frequent until September (Fig. 8). Medium sequences were recorded mostly in April and secondarily in August. Short sequences were recorded, on the other hand, throughout the photochemical period with dominance in March and October.
Monthly distribution of the different sequence durations (O3 data from 2009 to 2020 from the AIRPARIF station).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
b. Types of atmospheric circulation and weather associated with pollution episodes
The analysis of the ground pressure field and the 500-hPa geopotential height on the one hand, and of the state of the atmosphere from the different climatic variables recorded by the meteorological station on the other hand, allows the selection of three weather types associated with the PM10 peaks and four for O3 (Tables 1 and 2).
Number of days and average values of meteorological variables for the three weather types (2009–20) associated with PM10.
Number of days and averaged values of weather variables for the four weather types (2009–20) associated with O3.
1) The case of PM10
The first type (T1) is the most dominant type. It is observed in winter when the Azores High extends over central and Eastern Europe, topped by a ridge of high pressure over the whole of France at altitude (500 hPa geopotential height). France, on the western flank of this anticyclone, or south when it is centered on the British Isles, is affected by a mass of air coming from the southeast at the surface while a southwesterly flow dominates aloft. The sky is clear because the air is subsidence (ground pressure is around 1025 hPa). The wind is weak (2–3 m s−1) with temperatures above the seasonal norm (+3°C). The formation of fog is explained by the thermal inversion persisting for several hours.
The second type (T2) occurs when France is wedged between two troughs separated by a ridge of high pressure that rises to the Scandinavian countries at the surface as well as at altitude. At the surface, there is a dominant east/northeast flow. The extension of the Siberian anticyclone to the west (1022 hPa) favors the infiltration of very cold and dry air over the whole of Europe, resulting in a drop in temperature, which sometimes drops to −12°C. At the same time, the insolation is around 430 W m−2 h−1 and the wind is moderate (3–7 m s−1). Sunny but cold weather increases the massive use of heating.
The third type (T3), cold and cloudy weather, is observed mainly in early spring. It is characterized by a low geopotential height and low pressure configuration at the surface. France is affected by the trough that prolongs the subpolar low and carries a moderate flow (5–8 m s−1) from the northeast. As for T2, the temperature is low. It is on average 5°C below the seasonal norms. Contrary to the abovementioned radiative situations, the diurnal thermal amplitude is low (5.5°C).
2) The case of ozone
The first weather type (T1) accounts for almost 46% of the exceedance days. It corresponds to a situation of the barometric marsh (an area of the atmosphere between two weather systems where the pressure varies slightly but is slightly low) on the ground with a barometric ridge at 500 hPa. The relatively high atmospheric pressure and the widely spaced isobars favor clear skies, strong insolation (700–800 W h−1 m−2), and a weak wind (1–3 m s−1). The latter blows from the southeast at the surface; it is topped by a southwest flow aloft. It is often calm in the evening. For the temperature, the values are largely above the seasonal norm (+7°C). Large diurnal temperature ranges are observed (up to 17°C). In addition, visibility exceeds 7000 m.
The second type (T2) is observed on almost 20% of the exceedance days. Similar to the first type, it is characterized by a barometric marsh situation, but the upper ridge is absent. Nevertheless, there is still a high geopotential height with a high atmospheric pressure that can reach 1034 hPa (1018 hPa on average). The wind is weak (0–3 m s−1) with deviated directions (multidirectional). The latter is probably a thermal wind. The presence of fog, moderate visibility (2849 m), a high diurnal temperature range (12.5°C), and calm weather indicates a situation favorable to thermal inversion. This observation was validated by using data from the radiosondes of the Trappes station. This atmospheric stability favors the accumulation of pollutants near the surface.
The third type (T3), a ridge of high pressure over France, is present on 17% of the exceedance days. A 500-hPa geopotential height is observed. The wind changes from northeasterly when it flows over the eastern flank of the ridge at the beginning of the episode to southwesterly when the ridge migrates eastward and France is on its western flank. A moderate to relatively strong surface flow (4–8 m s−1) is then present. It is accompanied by high humidity brought by a maritime air mass. This results in moderate visibility (3636 m on average) and fog.
Fourth type (T4) has a frequency around 17% and is associated with the presence of a dynamic high pressure system over the Scandinavian countries and the British Isles or Germany. It is topped by a ridge of high pressure at 500 hPa. France, to the south of this high, is under the influence of an easterly flow. Atmospheric pressures vary from 1015 to 1029 hPa (1018 hPa on average). This position induces a northeasterly to easterly surface wind, topped by a southerly wind. The wind speed is moderate, varying from 3 to 7 m s−1. The air temperature is almost 8°C above normal, with a maximum of 25.4°C and an average temperature range of 11.7°C. A foggy phenomenon is observed at the beginning of the sequences. These conditions also coincide with strong insolation (up to 950 W m−2) and good visibility (4654 m on average).
c. Heterogeneity of pollution episodes according to weather types
1) The case of PM10
The central values (mean, median, and mode) indicate that the three weather types associated with exceedance days have an asymmetric distribution of PM10 concentrations. The T1 weather regime is characterized by a standard deviation (16.4 μg m−3) that is almost 2 times as high as the other two regimes (Table 3). The differences between the three weather types become more evident in Fig. 9; 25% of the concentrations between 68 and 162 μg m−3 are found in T1, and types T2 and T3 account for 25% of the concentrations between 48 and 53 μg m−3. The T2 and T3 weather regimes are similar in terms of their impact on air quality, unlike T1, which is an isolated type. Furthermore, the interquartile coefficient [(Q3 − Q1)/median; where QX is the quartile] shows that T1 (34%) has twice as many scattered concentrations as T2 (19%) and T3 (17%). The coefficient of variation (CV) calculated for T1 (27%) is roughly equivalent to twice that calculated for T2 and T3.
One-way analysis of the variation of pollutant concentrations according to weather types for the PM10 case.
Dispersion of PM10 concentration values according to weather types (according to the Kruskal–Wallis test).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
To explain this variation between the exceedance days observed under the T1 regime, a PCA was performed with a Varimax rotation. The application of this type of analysis resulted in three principal factors (PC) explaining a cumulative variance of 88.7% (Fig. 10). PC1 (52.9% explained variance) shows a strong positive correlation with the thermal parameters (Tmin, deviation from normal temperature, and dewpoint). PC1 can thus be qualified as mild weather influenced by a westerly flow bringing a humid maritime mass. PC2, with an explained variance of 19.7%, is correlated to two meteorological variables (thermal amplitude and pressure). The presence of close values between two correlations indicates a dependence between the two variables. This is consistent with the fact that thermal amplitude is associated, in part, with the loss of heat by radiation from the earth at night and consequently increased by high pressure favoring low cloud cover. The PC2 factor thus presents mild anticyclonic weather during the day, cold and clear during the night. On the other hand, PC3 (16.1% explained variance) presents a single positive correlation with visibility. It is characterized by clear skies during the day and poor visibility in the evening, morning and at the end of the day due to the formation of radiation fog.
PCA score and factor matrix for the main meteorological parameters associated with PM10 pollution.
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
Pollution intensity and persistence by weather types
We investigated whether weather types influence the intensity and persistence of pollution. Accordingly, the total number of days, as well as the average PM10 concentrations observed during each pollution sequence, were carefully inspected for each weather type. As shown in Fig. 11, the first weather type (T1) is the most dominant during the pollution episodes. It appears to be the most conducive to the onset of persistent pollution episodes of types S2 and S3 (116 days in total). It shows relatively high concentrations for sequences of 4–7 days (70 μg m−3). It also triggers more isolated cases than the other two weather types (25 days) that are slightly less polluted (49 μg m−3). Weather type T2 only appears on 19 pollution days, which represents 10% of all pollution episodes. It has a minimal impact on the occurrence of 2–3-day sequences (S2) and isolated cases as compared with weather type T3, which presents 2 times as many S2 and 3 times as many isolated cases. On the other hand, long sequences (S3) are only minimally triggered under both T2 and T3 regimes.
Distribution according to weather types of the different pollution sequences expressed in number of days (crosses) and average concentrations (histograms).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
2) The case of ozone
The comparison of central values (mode; median; mean) indicates an asymmetric distribution of ozone concentrations for all weather regimes associated with pollution episodes (Table 4). The calculated coefficients of variation of ozone concentrations are close (12%–15%) for weather types T1, T3, and T4. However, CV reaches 19% for T2.
One-way analysis of the variation of pollutant concentrations according to weather types for the O3 case.
To confirm the influence of weather types on pollution levels, a bivariate statistical treatment was performed on ozone concentrations. The results are presented in Fig. 12a. The comparison of quartile values shows that type T2 is the worst in terms of air quality. It comprises 25% of the exceedance days with concentrations between 142 and 212 μg m−3, whereas type T1 accounts for only 25% of the days between 121 and 167 μg m−3. The concentrations of weather types T3 and T4 indicate that these are intermediate weather regimes in terms of impact on air quality (131–180 μg m−3). The calculated interquartile coefficients indicate that type T2 (27%) has 2 times as much dispersed ozone concentration as type T1 (14%). Weather types T3 and T4 account for 22% and 18% of this coefficient, respectively.
(a) Dispersion of O3 concentration values and (b) pairwise comparison of weather types (each node indicates the mean rank of each type according to the Kruskal–Wallis test).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
The multiple pairwise evaluations of the differences show that types T1 and T2 appear to be two isolated groups with regard to pollution, as they display the most pairwise differences (Fig. 12b). This is mainly because these weather types have substantial differences in temperature and wind direction. On the other hand, the two weather types T3 and T4 are the most similar groups, probably because the wind speeds are very close.
Pollution intensity and persistence by weather types
As for PM10, a cross-sectional analysis was performed to determine whether the duration of the sequences and the ozone concentrations vary according to weather types. As shown in Fig. 13, the longest pollution sequences (8–13 days) are more severe when induced by type T4 (131 μg m−3 on average). On the other hand, they are more often triggered under weather type T3 (21 days; 42% equivalent). This weather type favors higher ozone concentrations for all sequences shorter than 8 days. It presents extreme concentrations (138 μg m−3 on average) and is second (40 days) to T1 (61 days) in terms of the number of days for pollution sequences of 4–7 days. In contrast, the lowest concentrations (120 μg m−3 on average) are recorded for type T1. For sequences shorter than 3 days, 241 days are triggered only during type T1.
Distribution according to weather types of the different pollution sequences expressed in several days (crosses) and average ozone concentrations (histograms).
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
The seasonal distribution of O3 pollution days shows the highest frequency in the summer period. On the contrary, they are characterized by the lowest frequency in the autumn period. In spring, all weather types, except T2, favor air pollution episodes with an average frequency (Fig. 14).
Distribution of weather types by season.
Citation: Journal of Applied Meteorology and Climatology 62, 5; 10.1175/JAMC-D-22-0161.1
5. Discussion
For a long time, objective and subjective classifications have been considered the main procedures for determining weather regimes (Yarnal 1993). In the last decades, their application has become more widespread, especially in air quality analysis (Tang et al. 2009; Demuzere et al. 2009; Russo et al. 2014; Dahech 2007; Shahgedanova et al. 1998). The use of a hybrid approach to defining weather types according to Vigneau’s (2000) definition, as mentioned in this study, offers an advantage over purely subjective or objective studies. It shows that a subjective classification validated by an objective classification of air masses can be used to understand and confirm the role of the different weather types on pollutant concentrations. The principle of this approach is presented in the work of Philipp et al. (2010) but only at the synoptic scale. In Ile-de-France, and in particular, in the conurbation of Cergy-Pontoise, no study has been conducted on the impact of weather types on concentrations of PM10 and O3 pollutants. Only the categorization of the link between pollution and meteorological variables is known.
The use of bivariate and multivariate statistics allowed the identification of the role of weather types in the intensity and persistence of pollution episodes. The results for PM10 indicate that 72% of the exceedance days occur when the Azores high pressure overflows central and eastern Europe, with a ridge over France at 500-hPa geopotential height (weather type T1). The persistent sequences of higher concentrations are induced by a situation of high atmospheric pressure associated with low cloud cover. The resulting nighttime radiative cooling creates a thermal inversion preventing the vertical dispersion of pollutants. Nevertheless, there is a percentage of 19% of exceedance days attributable to low geopotential height and low pressure configurations, more favorable to an increase in residential heating emissions during the cold season (weather type T3).
For O3, the highest concentrations coincide with barometric marsh situations accompanied by a high geopotential height (weather type T2). This type is the most persistent, with pollution episodes lasting more than 8 days (up to 13 days), probably as a result of strong radiation combined with very weak wind favoring a thermal inversion and the triggering of thermal winds. Weather type T1 with 46% of exceedance days is characterized by ozone episodes with lower concentrations. The meteorological parameters present in this category are related to a barometric marsh situation on the ground and an anticyclonic ridge at 500 hPa. A total of 67% of the days belonging to this type (T1) last from 2 to 7 days.
The observed limitations are related to the nonrepresentativeness of the conventional pollutant measurement station. The results can be improved using fine-scale data because local conditions vary from one site to another, particularly in urban areas and with contrasting topography, which is the case in Cergy-Pontoise. This is why future research will focus on the study of the spatiotemporal variation of air quality through the installation of three meteorological stations and 16 fixed air pollutant measurement stations located at specific sites and equipped with sensors measuring PM10, PM2.5, PM1, NO2, and O3. The lack of radiosondes for the vertical structure in the conurbation is currently a disadvantage in the study of air quality. For this reason, a series of captive radiosondes will be conducted using thermal and pollution sensors aboard helium-filled balloons released in specific areas of the conurbation.
6. Conclusions
The seasonal variability of the frequency of exceedance days was studied on an hourly basis for PM10 and O3 concentrations collected from a conventional urban background station between 2009 and 2020. During this period, weather types associated with poor air quality in the Cergy-Pontoise conurbation were determined according to the definition of Vigneau (2000) using a subjective and objective classification approach. The work consisted of examining how this association intervenes in the persistence and intensity of pollution. It appears that the anticyclonic weather type is the most conducive to exceptionally high PM10 concentrations, while the presence of a barometric marsh on the ground and a ridge at 500 hPa corresponds to high O3 pollution. When these two weather types are accompanied by a thermal inversion, PM10 and O3 pollution last for several days. In conclusion, the results of this research suggest that the prediction of weather types permits the anticipation of the occurrence of episodes of high PM10 and O3 pollution.
Acknowledgments.
The authors thank the AIRPARIF and NOAA for providing the environmental and meteorological data that served as the basis for the drafting. The authors also acknowledge the support of the DIM Qi 2 program of the Île-de-France Region, the technical support of the Cergy-Pontoise conurbation, and the doctoral scholarship program of the Paris City University. All opinions, findings, and conclusions expressed in this article are those of the authors and do not necessarily reflect the views of the various funders.
Data availability statement.
The datasets analyzed during the current study are available online (pollution data—https://data-airparif-asso.opendata.arcgis.com; meteorological data—https://gis.ncdc.noaa.gov/maps/ncei).
REFERENCES
Abbassi, Y., H. Ahmadikia, and E. Baniasadi, 2020: Prediction of pollution dispersion under urban heat island circulation for different atmospheric stratification. Build. Environ., 168, 106374, https://doi.org/10.1016/j.buildenv.2019.106374.
AIRPARIF, 2019: Bilan de la qualité de l’air Année 2018: Surveillance et information en Île-de-France. AIRPARIF Rep., 92 pp., https://www.airparif.asso.fr/sites/default/files/pdf/bilan-2018.pdf.
Camalier, L., W. Cox, and P. Dolwick, 2007: The effects of meteorology on ozone in urban areas and their use in assessing ozone trends. Atmos. Environ., 41, 7127–7137, https://doi.org/10.1016/j.atmosenv.2007.04.061.
Carnerero, C., and Coauthors, 2019: Relating high ozone, ultrafine particles, and new particle formation episodes using cluster analysis. Atmos. Environ., 4, 100051, https://doi.org/10.1016/j.aeaoa.2019.100051.
Carrega, P., 2004: Avant-propos sur les « types de temps ». Norois, 191, 7–9, https://doi.org/10.4000/norois.982.
Dahech, S., 2007: Le vent à Sfax (Tunisie), impacts sur le climat et la pollution atmosphérique. Thèse de Doctorat, Université Paris, 309 pp.
Dedoussi, I. C., S. D. Eastham, E. Monier, and S. R. H. Barrett, 2020: Premature mortality related to United States cross-state air pollution. Nature, 578, 261–265, https://doi.org/10.1038/s41586-020-1983-8.
Demuzere, M., R. M. Trigo, J. Vila-Guerau de Arellano, and N. P. M. van Lipzig, 2009: The impact of weather and atmospheric circulation on O3 and PM10 levels at a rural mid-latitude site. Atmos. Chem. Phys., 9, 2695–2714, https://doi.org/10.5194/acp-9-2695-2009.
Duché, S., 2013: Atmospheric pollution in Paris area: Exposition and perception on tourist attractions. Thèse de doctorat, Université Paris-Diderot, 259 pp.
Heft-Neal, S., J. Burney, E. Bendavid, and M. Burke, 2018: Robust relationship between air quality and infant mortality in Africa. Nature, 559, 254–258, https://doi.org/10.1038/s41586-018-0263-3.
Hodgson, E. C., and I. D. Phillips, 2021: Seasonal variations in the synoptic climatology of air pollution in Birmingham, UK. Theor. Appl. Climatol., 146, 1421–1439, https://doi.org/10.1007/s00704-021-03779-7.
Kalkstein, L. S., and P. Corrigan, 1986: A synoptic climatological approach for geographical analysis: Assessment of sulfur-dioxide concentrations. Ann. Assoc. Amer. Geogr., 76, 381–395, https://doi.org/10.1111/j.1467-8306.1986.tb00126.x.
Kikaj, D., D. Scott, M. Kobal, J. Crawford, and J. Vaupotič, 2020: Characterizing atmospheric controls on winter urban pollution in a topographic basin setting using Radon-222. Atmos. Res., 237, 104838, https://doi.org/10.1016/j.atmosres.2019.104838.
Li, H., S. Sodoudi, J. Liu, and W. Tao, 2020: Temporal variation of urban aerosol pollution island and its relationship with urban heat island. Atmos. Res., 241, 104957, https://doi.org/10.1016/j.atmosres.2020.104957.
Li, Q., B. Wu, J. Liu, H. Zhang, X. Cai, and Y. Song, 2020: Characteristics of the atmospheric boundary layer and its relation with PM2.5 during haze episodes in winter in the North China Plain. Atmos. Environ., 223, 117265, https://doi.org/10.1016/j.atmosenv.2020.117265.
Lim, S. S., and Coauthors, 2012: A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet, 380, 2224–2260, https://doi.org/10.1016/S0140-6736(12)61766-8.
Martin, N., 2008: La pollutiion par l’ozone et la climatologie dasn un espace Méditerranéen: Les Alpes-Maritimes. Thèse de doctorat, Université Nice Sophia Antipolis, 291 pp.
McGregor, G. R., 1999: Winter ischaemic heart disease deaths in Birmingham, United Kingdom: A synoptic climatological analysis. Climate Res., 13, 17–31, https://doi.org/10.3354/cr013017.
Miao, Y., and S. Liu, 2019: Linkages between aerosol pollution and planetary boundary layer structure in China. Sci. Total Environ., 650, 288–296, https://doi.org/10.1016/j.scitotenv.2018.09.032.
Miao, Y., J. Guo, S. Liu, H. Liu, Z. Li, W. Zhang, and P. Zhai, 2017: Classification of summertime synoptic patterns in Beijing and their associations with boundary layer structure affecting aerosol pollution. Atmos. Chem. Phys., 17, 3097–3110, https://doi.org/10.5194/acp-17-3097-2017.
National Institute of Statistics and Economic Studies, 2018: Population census 2018, results on the territory of Cergy-Pontoise conurbation (in French). INSEE, accessed 15 October 2021, https://www.insee.fr/fr/accueil.
Philipp, A., and Coauthors, 2010: Cost733Cat—A database of weather and circulation type classifications. Phys. Chem. Earth, 35, 360–373, https://doi.org/10.1016/j.pce.2009.12.010.
Raatikainen, T., A.-P. Hyvärinen, J. Hatakka, T. S. Panwar, R. K. Hooda, V. P. Sharma, and H. Lihavainen, 2014: The effect of boundary layer dynamics on aerosol properties at the Indo-Gangetic plains and at the foothills of the Himalayas. Atmos. Environ., 89, 548–555, https://doi.org/10.1016/j.atmosenv.2014.02.058.
Reiminger, N., X. Jurado, J. Vazquez, C. Wemmert, N. Blond, M. Dufresne, and J. Wertel, 2020: Effects of wind speed and atmospheric stability on the air pollution reduction rate induced by noise barriers. J. Wind Eng. Ind. Aerodyn., 200, 104160, https://doi.org/10.1016/j.jweia.2020.104160.
Russo, A., R. M. Trigo, H. Martins, and M. Mendes, 2014: NO2, PM10, and O3 urban concentrations and its association with circulation weather types in Portugal. Atmos. Environ., 89, 768–785, https://doi.org/10.1016/j.atmosenv.2014.02.010.
Shahgedanova, M., T. P. Burt, and T. D. Davies, 1998: Synoptic climatology of air pollution in Moscow. Theor. Appl. Climatol., 61, 85–102, https://doi.org/10.1007/s007040050054.
Sun, T., and Coauthors, 2019: Characterization of vertical distribution and radiative forcing of ambient aerosol over the Yangtze River Delta during 2013–2015. Sci. Total Environ., 650, 1846–1857, https://doi.org/10.1016/j.scitotenv.2018.09.262.
Sun, Y., and Coauthors, 2018: Ozone seasonal evolution and photochemical production regime in the polluted troposphere in eastern China derived from high-resolution Fourier transform spectrometry (FTS) observations. Atmos. Chem. Phys., 18, 14 569–14 583, https://doi.org/10.5194/acp-18-14569-2018.
Tang, L., D. Chen, P.-E. Karlsson, Y.-F. Gu, and T. Ou, 2009: Synoptic circulation and its influence on spring and summer surface ozone concentrations in southern Sweden. Boreal Environ. Res., 14, 889–902.
Toro, A. R., M. Kvakić, Z. B. Klaić, D. Koracin, R. G. E. Morales, and M. A. Leiva, 2019: Exploring atmospheric stagnation during a severe particulate matter air pollution episode over complex terrain in Santiago, Chile. Environ. Pollut., 244, 705–714, https://doi.org/10.1016/j.envpol.2018.10.067.
Tselepidaki, I. G., D. N. Asimakoupoulos, K. Katsouyanni, C. Moustris, G. Touloumi, and A. Pantazopoulou, 1995: The use of complex thermohygrometric index in predicting adverse health effects in Athens. Int. J. Biometeor., 38, 194–198, https://doi.org/10.1007/BF01245388.
Uttamang, P., P. C. Campbell, V. P. Aneja, and A. F. Hanna, 2020: A multi-scale model analysis of ozone formation in the Bangkok Metropolitan Region, Thailand. Atmos. Environ., 229, 117433, https://doi.org/10.1016/j.atmosenv.2020.117433.
Vigneau, J. P., 2000: Géoclimatologie. Ellipses, 334 pp.
Vukovich, F. M., and J. Sherwell, 2003: An examination of the relationship between certain meteorological parameters and surface ozone variations in the Baltimore–Washington corridor. Atmos. Environ., 37, 971–981, https://doi.org/10.1016/S1352-2310(02)00994-9.
WHO, 2001: Air Quality Guidelines for Europe. 2nd ed. European Series, Vol. 91, WHO Regional Publications, 273 pp.
WHO, 2018: Global ambient air quality database (update 2018). World Health Organization, accessed 20 May 2021, https://www.who.int/data/gho/data/themes/air-pollution/who-air-quality-database.
Wu, K., and Coauthors, 2019: Evolution and assessment of the atmospheric composition in Hangzhou and its surrounding areas during the G20 summit. Aerosol Air Qual. Res., 19, 2757–2769, https://doi.org/10.4209/aaqr.2018.12.0481.
Yarnal, B., 1993: Synoptic Climatology in Environmental Analysis: A Primer. Studies in Climatology Series. Belhaven Press, 195 pp.
Yuan, Z., J. Qin, X. Zheng, and Y. Mbululo, 2020: The relationship between atmospheric circulation, boundary layer, and near-surface turbulence in severe fog-haze pollution periods. J. Atmos. Sol.-Terr. Phys., 200, 105216, https://doi.org/10.1016/j.jastp.2020.105216.
Yuval, Y. Levi, U. Dayan, I. Levy, and D. M. Broday, 2020: On the association between characteristics of the atmospheric boundary layer and air pollution concentrations. Atmos. Res., 231, 104675, https://doi.org/10.1016/j.atmosres.2019.104675.
Zhao, H., and Coauthors, 2019: Inequality of household consumption and air pollution-related deaths in China. Nat. Commun., 10, 4337, https://doi.org/10.1038/s41467-019-12254-x.
Zheng, G. J., and Coauthors, 2015: Exploring the severe winter haze in Beijing: The impact of synoptic weather, regional transport and heterogeneous reactions. Atmos. Chem. Phys., 15, 2969–2983, https://doi.org/10.5194/acp-15-2969-2015.
Zhong, J., X. Zhang, Y. Dong, Y. Wang, C. Liu, J. Wang, Y. Zhang, and H. Che, 2018: Feedback effects of boundary-layer meteorological factors on the cumulative explosive growth of PM2.5 during winter heavy pollution episodes in Beijing from 2013 to 2016. Atmos. Chem. Phys., 18, 247–258, https://doi.org/10.5194/acp-18-247-2018.
