Introduction

Massive quantities of emissions due to rapid economic development, urbanization, energy consumption, transportation and motorization, result in elaboration of a variety of pollutants into the atmosphere causing air pollution with significant negative implications on ecosystem health. These pollutants include, but not limited to particulate matter (PM2.5, PM10), sulfur dioxide (SO2), nitrogen dioxide (NO2), ozone (O3), carbon mono and dioxide (CO, CO2)1. In this respect, it was estimated that about 67 million tons of pollution were emitted into the atmosphere in the United States in 20212. Worldwide, the harmful effects of air pollution on human health have been addressed including respiratory, cardiovascular and fertility diseases3,4,5,6 affecting all age groups particularly the vulnerable including children, women and elderly7. Furthermore, the World Health Organization (WHO) has been reported that air pollution represents the largest environmental health risk in the world where about 99% of the world’s population lives in places where the air pollutants exceed the WHO air guidelines levels8. Air pollution is also responsible for the greatest number of deaths, causing 6∙7 million deaths globally in 20199.

Additionally, Air pollution has significant adverse impacts on both crop productivity and building materials. For instance, the aggregated damage cost caused by emissions was estimated as a range from 277 to 433 billion € in 201710. Therefore, air pollution is considered as one of the most important and vital challenges, particularly in developing countries.

Addressing the challenge of CO emission into the atmosphere, through incomplete combustion of compounds such as gasoline, natural gas, oil, coal, and wood in addition to traffic emissions, industrial production, and biomass burning11, represents a prerequisite for evaluation of both air quality and potential risk to community health. Infants, elderly and people with anemia and/or a history of heart or respiratory diseases, are most likely to experience health effects upon exposure to elevated levels of CO12. CO can be considered also as a short-lived climate forcing agent since it contributes indirectly to climate change through participation in chemical reactions in the atmosphere that produce ozone; a climate change gas. Hence, prompting CO emission reductions is considered a possible strategy to mitigate effects of global warming.

On the other hand, PM represents a mixture of solid and/or liquid particles suspended in air resulting from smoke and dust and contains a higher number of harmful compounds. PM10 is considered as major component of air pollution and has threatened both environmental and human health whereas PM2.5 can cause cardiovascular, respiratory diseases and even early death13.

Deploying dense networks of instruments for air quality measurements are often too expensive and infeasible. Accordingly, there is a need for continuous spatio-temporal monitoring of air quality using In Situ measurements that are usually representative over a small spatial scale and provide only information at few selected points. In Situ measurements need a robust spatial and temporal modeling in order to detect the most affected areas in time (day, month, season)14,15.

Recently, remote sensing data can be used as a valuable alternative tool for monitoring air pollution and can support decision makers as they can be employed in retrieving and mapping of air quality parameters in a synoptic and multi-temporal coverage at regular intervals and dynamic scale16,17. The use of remote- sensed data in monitoring air pollution was promoted by the development of the space instruments designed to observe air pollutants, starting with the launch of the Total Ozone Monitoring Instrument (TOMS) in 1978, the Global Ozone Monitoring Experiment (GOME) in 1995, and Ozone Monitoring Instrument (OMI) in 200418.

Sentinel-5P TROPOMI is the latest satellite-based sensor that provided new opportunities for utilizing remote- sensed data in monitoring air quality19,20. Sentinel-5P TROPOMI covers wavelength ranges between ultraviolet and short-wave infrared, enabled observations of different species of air pollutants such as CO, NO2, O3, SO2, CH4, HCHO, and Absorbing or Ultra Violet Aerosol Index (AAI or UVAI), from which surface particulate matter (PM2.5 and PM10) may be inferred17. Sentinel-5P TROPOMI has a relatively high spatial resolution (3.5 * 7 km2 for all trace gases, except for CO and CH4 that is 7 * 7 km2)16,21.

The research in hand is intended to delineate vulnerability of densely populated areas to air pollution through retrieving CO, and PM2.5 from Sentinel-5P images and validating the retrieved data through simultaneous In Situ measurements.

Materials and methods

Study area

The study area includes two coastal governorates of the Nile Delta region; Kafr El Sheikh and El Behira Governorates covering a total area about 15,000 km2 and divided administratively into 32 districts, which are further subdivided into 765 Localities (Fig. 1). The total population of both governorates was estimated recently to be 10.2 million in 2021, which represents about 10% of the total Egyptian population22. Thus the study area represents a densely populated area with a variety of economic activities including agriculture, manufacturing and aquaculture.

Figure 1
figure 1

Source: developed by the authors using ArcGIS (Ver. 10.8).

Study area and administrative subdivisions.

Full size image

Methodology

In order to assess vulnerability of densely populated areas in the Nile Delta region to air pollution through Sentinel-5P remotely sensed data, a methodology of four main steps was applied including: acquiring and manipulating data, validating retrieved air quality data, and mapping vulnerability to air pollution (Fig. 2).

Figure 2
figure 2

Mythology of mapping vulnerability to air pollution through remote-sensed data.

Full size image

Acquiring and manipulating data

This step involves acquiring ground-truth data that were gathered through four field trips undertaken during the period from October 2020 to June 2021. A number of ground truth georeferenced measurments were carried out at different locations within the study area (Tables 1, 2 and Figs. 3, 4). Measurement of CO concentrations was conducted using standard test method for Carbon Monoxide in the Atmosphere through CO automatic continuous monitor (Continuous Measurement by Nondispersive Infrared Spectrometry). Also, (EVM7) was employed to measure the particulate matter; PM10 and PM2.5 using both the gravimetric method and the laser scattering theory.

Table 1 CO concentration in air samples collected from different sites.
Full size table
Table 2 PM concentration in air samples collected from different sites.
Full size table
Figure 3
figure 3

Source: developed by the authors using ArcGIS (Ver. 10.8)25.

Samples sites of CO concentrations for In-Situ measurements on World Imagery ArcGIS Online Base map.

Full size image
Figure 4
figure 4

Source: developed by the authors using ArcGIS (Ver. 10.8)25.

Samples sites of PM concentrations for In-Situ measurements on world imagery ArcGIS Online base map.

Full size image

Additionally, this step includes acquiring satellite-based observations for CO concentration and UVAI. In this respect, two datasets for air quality parameters in the study area via Sentinel-5P imageries were collected. The first dataset involved simultaneous data on CO column number density measured in the unit of mol/m2 and UVAI, which is unitless, were collected from Sentinel Hub EO Browser23. EO Browser allows downloading satellite data from numerous satellites and data collections instantly. This dataset was used to calibrate the relationships between satellite-based observations and ground-truth data through regression analysis.

Meanwhile, the second dataset comprised average CO column number density and UVAI over the study area during the period (1 October 2020–30 June 2021). For this purpose, the Google Earth Engine; an Internet-based platform that provides satellite imagery and cloud-based computing algorithms24, was employed to download such data. These datasets was further employed to retrieve average CO and PM concentrations in the study area during the study period.

Regression analysis

To investigate the relationship between CO and UVAI concentrations retrieved from Sentinel- 5P images and In-Situ measurement, a regression analysis was applied. For that purpose, various georeferenced sampling sites were allocated and employed for extracting the retrieved CO, PM10 and PM2.5 at the corresponding locations of the measurements. This was carried out through Spatial Analyst in ArcGIS (Extract Values to Points Tool).

Thereafter, relationships between CO concentration retrieved from Sentinel- 5P images as a dependent variable and ground-truth data of CO concentration was estimated through linear regression analysis. Similarly, the relationship between Aerosol Index and the ground truth data of PM10 and PM2.5 was investigated.

Mapping vulnerability to air pollution

Finally, the slope and intercept values resulted from regression analysis were applied to air quality datasets downloaded from Google Earth Engine for estimating the average concentration of air quality parameters in various parts of the study. This is followed by mapping various air quality parameters and assessing vulnerability of air pollution. For this purpose, the average concentrations of CO and PM2.5 were estimated for each locality in the study area through Zonal Statistics Tool in ArcGIS.

Thereafter, two composite vulnerability indexes were developed for CO and PM2.5 considering two primary indicators namely; pollutant concentration prevailed in each locality and population size exposed to the pollutant. Therefore, a direct-relationship can be established since localities that have high pollutant concentration and large population size have been exposed to such pollutant compared to those having low concentration of pollutant and small population size.

The primary indicators were normalized through the following formula:

$${N}_{x}=\frac{X-Xmin }{Xmax- Xmin}$$

where \({N}_{x}\), Normalized value of primary indicator; X, Absolute value of primary indicator; \({X}_{min}\), Minimum absolute value of primary indicator; \({X}_{max}\), Maximum absolute value of primary indicator.

Thereafter, the composite vulnerability index was calculated through aggregating normalized primary indicators for each pollutant assuming equal weights of the three indicators, according to the following formula:

$$V = {\sum }_{ i=1}^{n} \left( {N}_{i} * {W}_{i}\right)$$

where V, Vulnerability index; Ni, Normalized value of indicator i; Wi, Weight of indicator Ni

It should be noted that the resulting composite index is a relative index ranging between 0 for least vulnerable localities and 1 for most vulnerable localities.

Results and discussion

As a result of regression analysis, strong relationships between ground-truth data of CO concentrations and CO retrieved from Sentinel-5P images with R2 = 0.91 (P value < 0.01) (Fig. 5) were recorded. Similarly, ground-truth data of PM2.5 concentration was found to have strong relationship with Aerosol index recording R2 of 0.79 (P value < 0.05) (Fig. 6). Meanwhile, the relationship between PM10 and Aerosol Index was found to be insignificant (Table 3).

Figure 5
figure 5

Linear relationship between CO concentration retrieved from Sentinel-5P and In-Situ CO concentration.

Full size image
Figure 6
figure 6

Linear relationship between Aerosol Index retrieved from Sentinel-5P and In-Situ PM2.5 concentration.

Full size image
Table 3 Regression analysis of ground-truth and retrieved data.
Full size table

This means that remote- sensed CO concentration and Aerosol Index are effective in predicting CO and PM2.5 concentrations in mg/m3. Based on the intercept and slope derived from regression analysis, the concentration of CO and PM2.5 can be estimated through the following formula:

$${\text{y}} = 18231{\text{x}} - 519.56$$

where y, CO Concentration (mg/m3); x, CO Concentration retrieved from Sentinel-5P (mol/m2).

Also, PM2.5 concentration can be estimated according to the following formula:

$${\text{y}} = 26.725{\text{x}} + 24.463$$

where: y, PM2.5 Concentration (mg/m3); x, Aerosol Index retrieved from Sentinel-5P.

The estimated average CO concentration during the period (October 2020–June 2021) revealed that different parts of the study area has experienced varied level of CO concentration ranging between 40–78 mg/m3. The highest CO concentrations were noted as some spots in the northern, western, and southern parts the study area. Meanwhile, the south-western parts of the study area have experienced lowest level of CO (Fig. 7).

Figure 7
figure 7

Source: developed by the authors using ArcGIS (Ver. 10.8).

Average CO concentration in the study area (October 2020–June 2021).

Full size image

It is worthy to mention that retrieved data concerning CO concentrations in the study area seem to be overestimated. However, such an overestimation can be overlooked as that the suggested methodology in this study for delineating vulnerability to air pollution is dependent on a composite index, which provides a relative measure, rather than absolute one, ranging between 0 and 1 reflecting for least and most level of vulnerability; respectively.

To explain such a spatial pattern of CO concentration, correlation coefficients between CO concentration in the study area and potential sources of pollution such as closeness to dumping sites, power plants, highways and urban centers were estimated using Multivariate Analysis Tools (Fig. 8). Our results exhibited that CO concentration has strong negative correlation with distance to highways (− 0.722), to power plants (− 0.684) and to urban centers (-0.689); respectively. Such findings clarified that the concentration of CO decreases notably with increasing distance to power plants, highways and urban centers as potential sources of CO emission. Meanwhile, CO concentration was found to have a moderate negative correlation with distance to dumping sites (− 0.412) (Table 4).

Figure 8
figure 8

Source: developed by the authors using ArcGIS (Ver. 10.8).

Distance to potentials sources of CO emission.

Full size image
Table 4 Correlation Coefficient between CO Concentration and key potential sources.
Full size table

Furthermore, it was found that concentration range of PM2.5 in the study area was < 2 to 32 mg/m3 and PM2.5 concentrations decreased gradually from north-east to south-west, where the south-western parts recorded the highest concentrations (Fig. 9).

Figure 9
figure 9

Source: developed by the authors using ArcGIS (Ver. 10.8).

Average PM2.5 concentrations in the study area (October 2020–June 2021).

Full size image

The results of the current study are consistent with previous studies in Egypt, which recorded highest levels of PM2.5 concentration in the western margin of the Nile Delta exceeding 1350 µg/m3 on average25 and association of highest levels of CO concentration with industrial areas, where CO concentration exceeded 3000 mg/m326.

Application of the composite index approach of vulnerability to air pollutants; mentioned earlier in the proposed methodology, revealed that different localities are susceptible to varied levels of CO and PM2.5. About 47.7% of total population in the study area was categorized as very highly to highly vulnerable to CO (average CO concentrations exceeded 65 mg/m3). Meanwhile, 31.1% of the population were moderately vulnerable (average CO concentrations were 63 mg/m3) and only 20.2% of the total population are living in low to very low vulnerability zones with an average CO concentration of 57 mg/m3 (Fig. 10).

Figure 10
figure 10

Source: developed by the authors using ArcGIS (Ver. 10.8).

Vulnerability to CO (October 2020–June 2021).

Full size image

Comparatively, the population of the study area was found to be less vulnerable to PM2.5, where about 19.6% of total population are living within high to very high vulnerability zones (average PM2.5 concentration recorded 13 mg/m3). Meanwhile, 22.6% of the population are moderately vulnerable to PM2.5 and 57.8% of the total population are living in low to very low vulnerability zone with an average PM2.5 concentration not exceeding 3 mg/m3 (Fig. 11).

Figure 11
figure 11

Source: developed by the authors using ArcGIS (Ver. 10.8).

Vulnerability to PM2.5 (October 2020–June 2021).

Full size image

Conclusion

Remote-sensing data can serve as a cost effectiveness tool for monitoring air quality. In this regard, strong correlations were established in the current work between Sentinel-5P data, CO and PM2.5 ground-truth data as two of the major air pollutants. Sentinel-5P data were further employed for retrieving and mapping of CO and PM2.5 concentrations in two governorates of the Nile Delta region. Distribution pattern of CO concentrations in the study area was found to be associated with distance to main sources of CO emission.

Furthermore, using the retrieved parameters accompanied with the population size facilitated the determination of a composite vulnerability index for the pollutants of concern. The developed vulnerability index highlighted different localities in the study area with varied levels of vulnerability and identified the main hotspots.

It’s worth noting that using Sentinel-5P imageries for monitoring air quality requires availability of simultaneous In-Situ measurements for validation purposes. According to our case study results, it can be concluded that the implemented study model can be tested and applied for other air pollutants and can be extrapolated to other areas with similar or different environmental settings.