Introduction

Anthropogenic air pollution resulting from fossil fuel energy production and road traffic is one of the most important environmental problems in large cities and metropolitan areas in many countries (Manisalidis et al. 2020). The well-known greenhouse effect is a global phenomenon mainly caused by anthropogenic air pollution (Wuebbles et al. 2017). This phenomenon is the main trigger of variations in the Earth’s energy balance by blocking infrared radiation, thus changing the global temperature and causing global warming (Nave and Nave 1975; Kweku et al. 2018). In addition to their impacts on the global climate, most air pollutants directly threaten human health. The World Health Organization (WHO n.d.) has listed six major types, including particulate matter, near-surface ozone, carbon monoxide, sulfur oxides, nitrogen oxides, and lead. Among them, nitrogen dioxide (NO2) is one of the main air pollutants of serious concern (Goshua et al. 2022). As a short-lived gas, NO2 is a strong oxidant in the atmosphere (Correaa et al. 2022). It is mostly produced by fossil fuel consumption in urban transport (approximately 65%; Baldasano 2020) and industrial activities, power plants, and heating (Gheshlaghpoor et al. 2023). Moreover, NO2 is a precursor of near-surface ozone, which can affect human health (Ritz et al. 2019) and produce secondary pollutants (World Health Organization 2023). Approximately 91% of the world’s population is exposed to air pollution, leading to the death of seven million people every year (World Health Organization 2023). Many researchers have shown that the symptoms of lung cancer and diseases, premature death, cardiac arrhythmia and heart attacks, aggravated asthma, etc., are related to urban air pollution. In addition, air pollutants lead to the formation of smog and acid rain, as well as visibility reduction in megacities and urban areas. Apart from the impact on human health, certain pollutants (e.g. O3) damage vegetation, leading to reduced crop productivity and forest decline. Several research centres around the world are investigating this issue. Anthropogenic air pollution can be statistically approximated by recording the variability of the chemical constituents in air. According to the publication database, outstanding research on air pollution has been conducted in countries such as China and the USA (Kumar et al. 2023). Few collaborative international projects aim to continuously monitor and provide standardized information on air quality worldwide, including the World Air Quality Index Project (UNEP n.d.) and the European Environment Agency (EEA) European Air Quality Index Project (Eionet).

Following the COVID-19 outbreak in December 2019, several suspensions were implemented worldwide. As a result, significant reductions in air pollution have been reported. For instance, it has been estimated that the NO2 concentration in the Rajasthan region, one of the most polluted areas in India, has decreased by approximately 64% (Rivas 2020). According to the National Aeronautics and Space Administration (NASA) and the European Space Agency (ESA), the environmental quality has improved, and a 30% decrease in NO2 emissions has been observed. Gautam (2020) analysed the variation in NO2. Their results showed that the NO2 concentration in Asian and European countries significantly decreased due to COVID-19–related closures. Ogen (2020) estimated that NO2 levels decreased by 20–30% in European countries, mainly in Spain, Italy, and France (Bauwens et al. 2020) and by approximately 70% in Asian countries due to national closures. In the municipality of Quito, Ecuador, air pollution decreased by almost 50% (Vienneau et al. 2013; WAKI). It has thus been hypothesized that the COVID-19 nationwide closures, as well as the Iranian New Year holiday, have led to significant decreases in air pollution in the metropolitan area of Tehran.

With the use of different types of multispectral remote sensing data and more recently after the launch of the Sentinel-5P Precursor mission in October 2017, several researchers focused on the study of the air quality and atmospheric chemistry, for example the European Air Quality Index (AIRINDEX n.d.). Optical remote sensing data such as MODIS and MISR data have also been used to study atmospheric aerosol pollution, with a focus on determining the optical aerosol depth (Zweers 2021; Van Donkelaar et al. 2010). By analysing the ultraviolet aerosol index (UV-AI), Singh et al. (2023) demonstrated a rising tendency in aerosols in Ludhiana, India. This is particularly important when we consider that aerosols significantly influence the formation, life cycle, and properties of clouds (Singh et al. 2023). Gautam et al. (2022) highlighted the impacts of aerosols and fine particulate matter on rainfall patterns, temperature, and relative humidity at five Indian rural sites from 2015 to 2021 by using MODIS data (Gautam et al. 2022).

Sentinel-5 satellite images from the Copernicus programme of the ESA are freely available (Sentinel Online n.d.). TROPOspheric Monitoring Instrument (TROPOMI) technology provides useful data for measurements of complex tropospheric pollution (Veefkind et al. 2012; Kleipool et al. 2018). These data provide an unprecedented spatial resolution of 3.5 × 5.5 km, which is accompanied by an optimal temporal resolution (i.e. daily resolution). TROPOMI data are generally used for air quality monitoring (Eskes and Eichmann 2020) and climate monitoring and forecasting (Theys et al. 2017). The usability and efficiency of TROPOMI products for air quality monitoring have been widely demonstrated, e.g. Veefkind et al. (2012), Wang and Susumu (2015), Sharma et al. (2020), Lary et al. (2015), and Ogen (2020). Accordingly, these data have been analysed and validated by several researchers focusing on different topics in atmosphere sciences (Vîrghileanu et al. 2020).

In several studies, urban atmospheric pollution in major cities, megacities, and metropolitan areas around the world has been examined, for example in Indian cities based on GOME data (Zheng et al. 2016), East Asia based on the OMI sensor (Han 2019), and California using OMI data (Bechle et al. 2013). In a recent study, Gheshlaghpoor et al. (2023) calculated air pollutants from Sentinel-5P products and obtained urban land uses from the local climate zone (LCZ) scheme and Landsat-8 satellite imagery in summer 2020 and winter 2021 in Tehran. The relationship between the spatial composition and configuration of urban land uses and air pollutants was analysed. They showed and estimated that the land use spatial composition and configuration exhibited a positive correlation with NO2 and SO2. In contrast, the CO pollutant exhibited a significant negative correlation with green spaces.

Long-term investigation and detailed spatial analysis studies of air pollution in Tehran are missing. Moreover, the urban traffic system in the Teheran metropolitan area experienced significant development over the last two decades. In this research, we aimed to determine whether recent developments in the urban transportation systems in metropolitan areas affected urban air pollution. Data from two space-based sensors, including the Ozone Monitoring Instrument (OMI) and TROPOMI, were used in this study to investigate the spatial and temporal variations in major air pollutants over the metropolitan area of Tehran. In addition to long-term trends, we explored the subannual fluctuations in air pollutants. In regard to the considerable differences in atmospheric conditions between the cold and warm seasons, high seasonality in certain air pollutants was anticipated. The spatial distribution of pollutants was primarily determined by the urban traffic density. However, certain pollutants, such as CO1, depend on elevation. First, we investigated the determining factors of the long-term trends in specific air pollutants using OMI data. A detailed investigation of spatiotemporal fluctuations was conducted using TROPOMI data.

This research attempts to provide insights for researchers and professionals in the urban management and health sectors by providing up-to-date information and revealing the research gaps that should be addressed regarding air pollution. The main objectives of this study were to specify the spatiotemporal characteristics of air pollutants in the metropolitan area of Tehran. This includes the identification of the main air pollutants, the analysis of the long-term, annual, and short-term variations, and the determination of their spatial distribution and spatial gradients. By extracting the most dominant spatiotemporal patterns, we aimed to support urban management in reducing atmospheric pollution and improving the health of residents.

Material

Study area

The metropolitan area of the capital city Tehran is the most densely populated region in Iran, with a population density of approximately 13,100 people per km2. With a population of approximately 13 million people and an area of 974 km2, this metropolis is the 28th most populous city in the world as of 2023 (World Urbanization Prospects 2023). The metropolitan area is centred at approximately 35°36′N and 51°17′E. Currently, Tehran consists of 22 districts. As a result of rapid population growth and increasing urbanization, there are various types of urbanity in this region. Moreover, there are several industrial plants in the neighbourhood of Tehran. In contrast, the unique geographical location of the metropolis (it is surrounded by the Shemiran and Damavand Mountains in the north, west, and east) has led to an increased concentration of air pollutants (Zebardast and Riazi 2015). As a result, the metropolitan area of Tehran was ranked as the 26th most polluted city in the world in 2021 (https://www.iqair.com/iran/tehran). Currently, the metropolitan area includes contiguous urban and industrial areas within the two provinces of Tehran and Alborz.

Ozone Monitoring Instrument

The Ozone Monitoring Instrument (OMI) is onboard NASA’s sun-synchronous polar-orbiting satellite Aura, which passes the equator at 13:38 local time since 2004. The OMI has a 2600-km viewing swath, with small gaps at the equator and overlaps at higher latitudes, and the pixel size decreases from 13 × 24 km at the nadir to 26 × 135 km at the edges of the visible range. The OMI collects data in the ultraviolet and visible (UV/Vis) spectral ranges from 270 to 500 nm (Kurosu et al. 2004). With these polar specifications, a nearly daily global coverage (Millet et al. 2008) but a daily coverage in the study region is possible. In previous work, the applicability of these data has been demonstrated, particularly for the study of short-lived air pollutants and the analysis of local relationships between air pollutants (Palmer et al. 2003, 2006). Despite the long time series, the OMI provides a limited spatial resolution. However, considering 19 years high temporal resolution and long-term data availability of the OMI, it has been adopted as a valuable tool for trend analysis of atmospheric components. Regarding the most variable urban pollutants, in this study, we analysed vertical NO2, SO2, and HCHO column density data over the metropolitan area of Tehran.

Sentinel-5 Precursor

Since 2017, freely accessible Sentinel-5 Precursor datasets have been commonly used to monitor the atmospheric chemical composition and air quality on various scales from regional to urban scales. Sentinel-5’s TROPOspheric Monitoring Instrument (TROPOMI) sensor collects radiation data in the ultraviolet and visible (UV/Vis) regions of the electromagnetic spectrum. The TROPOMI provides estimates of air quality parameters such as ozone, carbon monoxide, nitrogen dioxide, sulfur dioxide, methane, aerosols, and formaldehyde. The state-of-the-art multispectral imaging spectrometer provides a unique footprint of atmospheric gasses with high spatial and temporal resolutions. Its products represent vertically integrated estimates converted from the ratio of the initial satellite-measured oblique column density and total air mass factors. The data were accessed through the Google Earth Engine platform (GEE). The air quality parameters from the TROPOMI are summarized in Table 1.

Table 1 TROPOMI Sentinel-5P dataset
Full size table

Methodology

This study is based on the analysis of air quality parameters in different parts of the Tehran and Alborz provinces. First, seven sample regions are defined, representing the various land covers of the region (Fig. 1). This approach allows (1) analysing the spatial and temporal patterns of pollutants depending on different environmental factors, such as the distance from urban areas, elevation of the terrain and geographical situation (i.e. being surrounded by heights) and (2) comparing pollutants in regions with similar characteristics. This eventually facilitates the classification of air pollutants based on similarities in their spatiotemporal patterns. In this context, the cross-correlation between parameters represents their relationship. The selected sample regions are representative of the diversity of the land cover, climate, and topography of the study area. They are distributed across very different topographic settings, including urban lowlands in the south and sloping mountainous terrain in the north, with an elevation range of approximately 750 m (the elevation varies between 819 and 2572 m). Specifications of the sample regions are listed in Table 2. Due to the low spatial resolution of OMI data, the detailed spatiotemporal analysis of the air pollutants is based only on TROPOMI data.

Fig. 1
figure 1

Location and extents of seven sample regions having maximum differences, intended for time series analysis

Full size image
Table 2 Specification of sample regions
Full size table

Under clear sky conditions, the data provide information on the spatial distribution of air components, but they suffer the drawback of being disturbed by cloud cover, a major source of error (Boersma et al. 2004). To eliminate cloudy scenes, images with quality values below 0.75 are filtered out. Nevertheless, light cloud pixels or partially cloudy pixels may still exist in the process. Furthermore, remotely sensed data normally suffer from mismatching errors due to imperfect spatiotemporal colocation of satellite data sampling. Therefore, kernel averaging is applied to extract time series of the pollutants. This approach minimizes unfiltered cloud effects and mismatching errors.

In addition to large-scale analysis, the spatiotemporal fluctuations in air pollution at the local scale over the metropolitan area of Tehran may demonstrate particular patterns due to notable and spatially dense pollutants originating from transport and industry. The detailed spatiotemporal variations in the pollutants are studied in the next section.

Analysis of the air quality

NO2, SO2, and HCHO derived from the OMI

The spatial resolution of the obtained OMI data is not sufficient to investigate local variations in air quality (Fig. 2). However, these series are advantageous for inferring long-term trends not assessable from Sentinel-5 data beginning in 2017. The time series extracted for the metropolitan area of Tehran show clear seasonal cycles. NO2 exhibits higher seasonality than SO2 and HCHO. In addition, all three pollutants show slight declines over the last two decades, which are mainly the result of developments in urban transport systems in Tehran and its suburbs. Transport systems have been highly improved through the construction of subway lines and bus rapid transit (BRT) systems.

Fig. 2
figure 2

Spatial distribution map (top-left) and area-averaged time series (top-right, bottom row) for Tehran derived from OMI data

Full size image

NO2 derived from the TROPOMI

The map of the NO2 distribution (Fig. 3) shows that the maximum level over Tehran decreases radially and rapidly according to the urban borders. The mean levels of NO2 exceeded 0.0005 mm/m3 in Tehran and reached 0.001 mm/m3 in the western part of the city during the 3.5-year study period. The concentration of NO2 exceeds the commonly allowed value for residential areas. In terms of temporal patterns, the NO2 time series shows a seasonal pattern. However, while the seasonal cycle is obvious over the metropolitan areas (e.g. sample regions #1, #2, and #3), the low levels of NO2 in nonurban and mountainous areas lead to lower seasonality (e.g. sample region #4). The periods of the maximum and minimum NO2 concentrations in the metropolitan area are October–February and May–August, respectively.

Fig. 3
figure 3

NO2: time average map (top) and area average time series of the seven regions (bottom). The hatches in the map indicate the metropolitan area

Full size image

CO1 derived from the TROPOMI

Similar to NO2, the maximum rate of CO1 occurs over the city (Fig. 4). However, in contrast to NO2, the spatial distribution of CO1 also depends on the elevation. While CO1 decreases rapidly northwards in high mountainous areas, it shows a low gradient towards the south where the elevation decreases. Another difference is observed for the gradient slope of the pollutant; NO2 has a steep gradient on the order of 10, although the difference between the minimum and maximum CO1 levels is only approximately 50%. The time series of CO1 indicate a seasonality that coincides with the seasonality of NO2 in the urban areas, with the maximum during the cold season. However, the seasonal cycles are inverted in the mountainous areas (sample regions #4 and #5), with the maximum CO1 rates during the warm season. Generally, CO1 shows higher dispersion than NO2. The maximum rate of CO1 is 0.043 west of Tehran.

Fig. 4
figure 4

CO1: time average map (top) and area average time series of the seven regions (bottom). The hatches in the map indicate the metropolitan area

Full size image

SO2 derived from the TROPOMI

In contrast to NO2 and CO1, the spatial distribution of SO2 shows a particular pattern that does not merely follow the urban areas, including concentration islands. However, the general pattern is associated with the elevation (Fig. 5). In terms of temporal variation, SO2 shows seasonal cycles that coincide with those in NO2. However, the time series in all seven sample regions are in phase with the maximum rates during the cold season. The two time series of mountainous areas (samples #4 and #5) include data gaps due to cloudiness mainly during the cold season. Similar to its spatial heterogeneity, SO2 shows considerable temporal dispersion.

Fig. 5
figure 5

SO2: time average map (top) and area average time series of the seven regions (bottom). The hatches in the map indicate the metropolitan area

Full size image

HCHO derived from TROPOMI

The spatial distribution of HCHO is largely similar to that of NO2, decreasing radially from the centre of the metropolitan area (Fig. 6). However, HCHO shows a few differences in its spatial pattern: HCHO is more independent of the elevation, and in contrast to NO2, it has a low-sloped spatial gradient. In terms of temporal variation, a clearer seasonal cycle is revealed at the margins of the metropolitan area, i.e. in sample regions #2, #5, and #7, although the seasonal cycle is not dominant at the centre of the metropolitan area (sample region #1) or in distant areas (sample regions #4 and #6).

Fig. 6
figure 6

HCHO: time average map (top) and area average time series of the seven regions (bottom). The hatches in the map indicate the metropolitan area

Full size image

O3 derived from TROPOMI

Ground-level ozone is mainly formed by photochemical reactions of other pollutants such as CO1 and NO under strong sunlight and UV radiation. The spatial distribution of O3 is very different from that of the already mentioned pollutants, as it varies by only a few percent over the entire surface of the study area (Fig. 7). O3 follows a pattern that depends on the distribution of and distance to forests. Similarly, the land topography plays an important role in preventing or allowing O3 to spread from forested areas outward. The general rate of O3 decreases from north to south and from west to east. The highest level is recorded in the north, where large valleys are stretched towards the Mazandaran forests. The temporal variation in O3 is also unique among the considered pollutants. It shows the most regular seasonal cycle in all the sample regions. The times of maximum and minimum values are spring and autumn, respectively, which correspond to the times of maximum and minimum forest growth, respectively.

Fig. 7
figure 7

O3: time average map (top) and area average time series of the seven regions (bottom). The hatches in the map indicate the metropolitan area

Full size image

Aerosols derived from TROPOMI

Aerosols are a regional phenomenon in the Middle East. They can be considered an external phenomenon, entering the study region mainly from the southwest and sometimes from the west of Iran, with the main sources located in Iraq and Syria. Aerosol particles spread entirely throughout the atmosphere. Therefore, wherever the elevation is low, a larger column of spectral absorption occurs, which leads to higher index values. Conversely, lower index values are recorded in highly elevated areas. In a similar way, elevation affects the height and depth of aerosols. Generally, the height of aerosols is low, and their depth is large in low-elevation areas. Conversely, high-elevation aerosols with smaller depths occur in mountainous areas. The negative values in the map indicate nonabsorbing particles (Singh et al. 2023) due to the time averaging process of aerosol-free days. Aerosols have a seasonal cycle in the study area, with maximum and minimum values in June–August and December–February, respectively (Fig. 8). Dispersions depend on the year. For example, the maximum aerosol concentration occurred in late spring (March) 2020.

Fig. 8
figure 8

Aerosol: time average map (top) and area average time series of the seven regions (bottom). The hatches in the map indicate the metropolitan area

Full size image

Clouds derived from TROPOMI

Although cloudiness is not a factor of pollution, it may affect the rate and spatial distribution of certain pollutants, such as CO1. The TROPOMI delivers two principal specifications of clouds: cloud fraction and cloud depth. Cloudiness and precipitation rates in the study region are generally controlled by the geographical situation. The two sources of cloudiness are humid air masses originating from the west (the main source of precipitation in most parts of Iran) and the north (the humid Caspian Sea coastlines). In addition, elevation is always an important factor of cloudiness. Based on the geographical situation and elevation, the cloud fraction decreases from north to south in the study region. In the vicinity of the Caspian Sea, the Alborz Mountain Chains in the north exhibit a time-averaged cloud fraction of more than 50%, and the lowlands in the south next to the Dasht-e-Kaveir Desert exhibit the minimum cloud fraction (approximately 15%). Except for the urban pollution effect, the spatial pattern of the cloud fraction approximately matches that of CO1. A seasonal cycle in the time series of the cloud fraction is observed, with maximum and minimum values during the cold and warm seasons, respectively. The cloud fraction shows high temporal dispersion (Fig. 9). The cloud optical depth also follows a similar pattern to the spatial distribution of the cloud fraction. There is a weak relationship between the cloud fraction and cloud optical depth (average = 0.59). The correlation coefficients are higher in downhill areas than in mountainous areas and plains. The highest correlation (0.73) is in the metropolitan area sample region.

Fig. 9
figure 9

Cloud fraction: time average map (top) and area average time series of the seven regions (bottom). The hatches in the map indicate the metropolitan area

Full size image

Spatiotemporal outlines

Relationship between the air pollutants

The relationship between the studied parameters is separately explored for all sample regions. To identify the similarity between the temporal behaviours of the parameters, cross-correlation coefficients were calculated between all parameters. Two sets of correlation coefficients are provided: (1) correlation coefficients are calculated for 10-day smoothed data to reduce the dispersion but preserve the seasonal cycles in the data and (2) correlation coefficients are calculated based on cycle-free data. In the latter set, the parameters are analysed in terms of similarities in trends without the effect of seasonality, which highly impacts the correlation coefficients. The matrices of the cross-correlation coefficients for the 10-day smoothed data and cycle-free data are shown in Fig. 10.

Fig. 10
figure 10

Correlation coefficient between the seven parameters in the seven regions for the raw data (upper panel) and seasonal cycle-removed data (lower panel). The matrices are ordered based on the sample region from left to right: regions #1 to #4 in the first row and regions #5 to #7 in the second row

Full size image

As shown in Fig. 10, CO1, SO2, and NO2 exhibit the highest correlations, and aerosol and HCHO exhibit the lowest correlations within the urban sample regions (#1 to #3). However, in the cycle-free data, the order is changed. The cloud fraction exhibits the lowest values in most sample regions. To illustrate the correlation across all orders, the summation of the correlation coefficients within the sample regions is shown in Fig. 11. As depicted, SO2 and CO1 exhibit the highest correlation with all parameters. In the raw data correlation graph, clouds are highly correlated with the air pollutants (sum = 8.5), HCHO exhibits no correlation (sum = 0.15), and aerosols exhibit a negative correlation (−5.7) with the other parameters. However, the rankings of HCHO, aerosols, and clouds are corrected after cycle removal. In the cycle-free correlation graph, HCHO is highly correlated with the other parameters (sum = 7.9), aerosols also exhibit a correlation (sum = 5), and clouds exhibit no correlation (sum = 0) with the other parameters.

Fig. 11
figure 11

Summation of the correlation coefficients in each region and overall summation of the correlation coefficients sorted in terms of the air quality parameters for the raw data (upper panel) and for seasonal cycle removed data (lower panel)

Full size image

Spatial patterns

Based on the analysis of the correlation coefficients, the parameters with similar behaviours are identified. In this regard, the three pollutant factors of NO2, HCHO, and CO1 are specified as urban pollutants. These parameters are concentrated over the centre of the metropolitan area and decrease radially outward. Among the urban pollutants, CO1 is also affected by elevation. Due to the weight of CO1 particles, this pollutant rapidly decreases with increasing elevation. NO2 and CO1 show obvious seasonality, with a maximum at the time of air inversion (cold season) and a minimum at the time of air upwelling (warm season). HCHO has a seasonality that differs in amplitude and phase from those of NO2 and CO1. This is the reason why HCHO shows a nearly zero integrated correlation coefficient in the raw data. SO2 exhibits unique behaviour among the pollutants. It has the highest summation of the correlation coefficients after cycle removal. SO2 is an urban-industrial pollutant that is controlled by both vehicle and industrial activities in the south of the metropolitan area (refer to the time series of sample region #6). However, the dominant control on urban pollutants is provided by urban vehicle activities. Due to the weight of SO2 particles, it also rapidly decreases with increasing elevation. SO2 exhibits a seasonality similar to that in NO2 and CO1 with a maximum at the time of air inversion. O3, aerosols, and clouds are exterior parameters not controlled by metropolitan area activities. O3 shows the simplest and most regular patterns. It exhibits the smallest variations in all the sample regions in both amplitude and seasonality phase. Its seasonality coincides with the growth/decay in the northern forests (in Iran). The integrated correlation of O3 before and after cycle removal remains nearly constant (8.5–11). The high similarity between the time series of the study areas indicates that O3 is affected neither by urban and industrial activities nor elevation. Aerosol and clouds are produced and move on a regional scale. They are thus not significantly affected by local factors. The only relevant local factor is elevation, which can impact the column of aerosols or the optical depth sensed by the sensor.

Temporal patterns

In addition to seasonal cycles, air quality parameters show short-term fluctuations and long-term trends. To analyse the temporal variations, the parameters are compared in terms of influences due to three phenomena: (1) the Iranian New Year holiday, (2) outbreak of COVID-19, and (3) maximum and minimum air pollution levels. Time series with the mentioned events marked are shown for the various sample regions in Fig. 12.

Fig. 12
figure 12

Time series of the air quality parameters in the metropolitan area of Tehran with the period of significant changes marked

Full size image
  1. I.

    Starting on ~19 March and ending on ~4 April, the Iranian New Year holiday (called Norouz) lasts more than 2 weeks depending on the year. The termination of daily activities and the decrease in vehicle travel in the metropolitan area thus significantly affect air pollution. As a result of this holiday period, urban and industrial pollutants (NO2, CO1, HCHO, and SO2) experience significant decreases that start with a steep decline. The results show that except for HCHO, the other pollutants experience their minimum annual levels during the Norouz holiday period. HCHO recorded its minimum during the Norouz holiday periods in 2019 and 2021. However, in 2020, its minimum occurred in November.

  2. II.

    The COVID-19 outbreak occurred in Iran in February 2020. It was followed by a few temporary lockdowns and selective confinements for specific professions, which ended with a considerable decline in vehicle and industrial activities in the metropolitan area. The effect of COVID-19 lockdowns on air pollution can be detected in the time series of urban pollutants in the form of a declining trend after March 2020. In particular, NO2, CO1, and SO2 show negative trends.

  3. III.

    According to the Administration of Air Quality Control of Tehran, the maximum air pollution level in recent years occurred in October 2021. The time series shows that during this period, the air pollutants NO2 and CO1 recorded specific peaks, and SO2 recorded a high value. Accordingly, the minimum values of these pollutants were recorded in July 2021, which were particularly extreme for NO2. The NO2 extreme and mean values are shown in Fig. 13. As shown in the time series, exterior pollutants (O3 and aerosol) are not affected by the abovementioned phenomena.

Fig. 13
figure 13

Spatial distribution of NO2 over the two provinces (upper row) and over the metropolitan area (lower row)

Full size image

Discussion and conclusion

This study relies on remotely sensed air quality data to analyse the spatial and temporal variations in major pollutants in the metropolitan area of Tehran at scales ranging from a limited urban area to a large provincial-level region. Time series of OMI data for the major pollutants NO2, SO2, and HCHO are studied over the past two decades. Although the spatial distribution of the pollutants is not adequately represented in OMI-derived products, the long-term time series are advantageous for analysing trends in metropolitan areas. The results indicate slight negative trends in air pollutant concentrations over the last two decades. These trends are mainly due to the development of urban transportation systems in Tehran and suburbs through the construction of metro lines and bus rapid transit (BRT) systems. These results demonstrate the potential of OMI data for determining long-term variations in air quality in large metropolitan areas.

Benefiting from the comprehensive and high-resolution TROPOMI data products, spatiotemporal variations in seven air quality parameters, including NO2, CO1, SO2, HCHO, O3, aerosols, and clouds, are studied in detail. To determine the similarity between the temporal behaviours of the parameters, cross-correlations are analysed. Based on their spatiotemporal patterns, the parameters were categorized into three classes. The results show that the spatial distributions of CO1, NO2, and HCHO indicate their highest concentrations over a specific part of eastern Tehran, and the values decrease radially and rapidly along the city boundaries. Because of this notable characteristic, they can be classified as urban pollutants. The predominant control of urban pollutants is related to urban vehicle activity. However, in contrast to NO2 and HCHO, the spatial distribution of CO1 also depends on the elevation due to its specific molecular weight. While CO1 decreases rapidly northwards in the high mountainous regions, it exhibits a low gradient towards the south, where the elevation decreases relative to urban areas. SO2 shows a unique spatial distribution that does not follow only urban borders, indicating that SO2 is an urban-industrial pollutant. O3, aerosols, and clouds are not or are only slightly affected by activities inside the metropolitan area. Therefore, these parameters were evaluated as exterior parameters. The spatial distribution of O3 differs from that of the other six parameters with its low gradient. Notably, the mountain ranges and topography of the terrain play a role in preventing or allowing O3 to spread out from the forested areas in the north. Aerosols enter the study region mainly from the southwest, with their main origin in Iraq and Syria. The aerosol depth is generally higher in the southern lowlands than in the mountains in the north of the provinces. The gradients of the pollutants are highly different. NO2 and SO2 have steep gradients on the order of 10 times. In contrast, CO1 varies by approximately 50% and O3 only by approximately 2%.

The above classification could be helpful to provide insight into where a specific pollutant originates and how it may (or may not) be managed. For instance, the results demonstrate that CO1 tends to spread out through low-elevation routes, i.e. south of Tehran. The specific gradients of the pollutants determine the rate at which they may be controlled.

Urban (-industrial) pollutants include seasonality related to fossil fuel consumption in the metropolitan area on the one hand and related to air inversion/upwelling conditions (coinciding mainly with the season) on the other hand. NO2, CO1, and SO2 show obvious seasonality, with maximum values during the cold season and minimum values during the warm season. HCHO exhibits slight seasonal variations except at the margins of the metropolitan area where the seasonality is clearer. This may be due to the lifetime of HCHO or the influence of the other pollutants on HCHO. This question must be investigated further. The other pollutants also influence each other, e.g. CO1, NO1, and NO2 are precursors of O3. Although cloudiness is not a pollutant, it can affect certain pollutants, such as CO1. Exterior parameters also show seasonality associated with the movement of regional air masses (for cloudiness), the movement of regional dust masses (for aerosol), and the growth/decay in the northern forests (for O3).

The pollutants are finally examined in terms of short-term variations. The Iranian New Year holiday period (Norouz) significantly impacts air pollution. As a result of this event, NO2, CO1, HCHO, and SO2 experienced a significant decrease with a steep decline at the beginning of the holiday period. Except for HCHO, the pollutants reached their annual minimum levels during Norouz. The nationwide closures due to the COVID-19 outbreak resulted in a decline in pollution, especially NO2, CO1, and SO2, after March 2020. This trend was caused by the significant decrease in vehicle and industrial activities in the metropolitan area. Exterior parameters were not affected by the Norouz and COVID-19 events.

The results are compared to in situ measurements obtained from the Tehran Air Quality Control Company (AQCC; http://air.tehran.ir) affiliated with the municipality of Tehran. Consistent with the satellite measurements, the in situ records show that the highest and lowest pollution levels occur east of Tehran (Pirouzi Station) and west of Tehran (District #22 station), respectively. In addition, the in situ measurements show that the minimal level of air pollution (particularly for CO1 and NO2) occurs during the New Year holiday period.

The data may suffer drawbacks in terms of spatiotemporal mismatching and remaining cloud cover even after applying the cloud filter. We used area-averaged time series to minimize these effects. Moreover, the surface albedo is a source of uncertainty in estimating certain pollutants. Especially under low albedo values, some bias may occur. This study did not consider absolute estimations but only accounted for trends and fluctuations. If the total values of pollutants are the target, the impacts of the surface albedo should be considered. Regarding the 2-decade negative trend due to the developments in transportation, the medium-term declines due to COVID-19 closures, and the short-term decreases due to the Norouz holiday period; it could be concluded that the critical issue of air pollution in the metropolitan area of Tehran could be effectively managed.