The effect of national protest in Ecuador on PM pollution

Particulate matter (PM) accounts for millions of premature deaths in the human population every year. Due to social and economic inequality, growing human dissatisfaction manifests in waves of strikes and protests all over the world, causing paralysis of institutions, services and circulation of transport. In this study, we aim to investigate air quality in Ecuador during the national protest of 2019, by studying the evolution of PM2.5 (PM ≤ 2.5 µm) concentrations in Ecuador and its capital city Quito using ground based and satellite data. Apart from analyzing the PM2.5 evolution over time to trace the pollution changes, we employ machine learning techniques to estimate these changes relative to the business-as-usual pollution scenario. In addition, we present a chemical analysis of plant samples from an urban park housing the strike. Positive impact on regional air quality was detected for Ecuador, and an overall − 10.75 ± 17.74% reduction of particulate pollution in the capital during the protest. However, barricade burning PM peaks may contribute to a release of harmful heavy metals (tire manufacture components such as Co, Cr, Zn, Al, Fe, Pb, Mg, Ba and Cu), which might be of short- and long-term health concerns.


Results and discussion
Satellite PM 2.5 concentration data. The national Ecuadorian strike of 2019 lasted 12 days during October 2-13, drawing people from all over the country to the capital city Quito. First, to study the impact of the strike on PM 2.5 levels in Ecuador, satellite images were analyzed for the whole country and the region (Fig. 1a-c). Comparing among the different 12-day periods (before ( Fig. 1c) the strike), it was observed that most of the polluting protest activities were present in the capital city Quito. Therefore, the focus of this study was narrowed down to Quito, the central place for the national strike. It was observed that Ecuadorian regional air quality during those dates was highly affected by the volcanic activity (active Sangay volcano ash plumes were detected in the Central part of the country, reaching over 70 µg m −3 ) (Fig. 1a-c). There were a few active-erupting volcanoes in the Andes cordillera at the time of the strike 38 . Also, a background PM pollution was spotted creeping in from the north (Colombia) along the eastern Andes cordillera (Fig. 1a-c). The mountains function as a barrier to push the pollution south along the mountain ridge, with the prevailing northeasterly winds.
Figure 1d-f show satellite images of PM 2.5 concentrations for the Pichincha region (red line) of Ecuador, containing the city of Quito (gray line), averaged over the three studied periods. The average PM 2.5 levels in the whole district of Pichincha during the strike were lower (6.99 µg m −3 ) than before (8.04 µg m −3 ) and after (7.8 µg m −3 ) the strike. The PM 2.5 levels in two 45 km × 45 km cells of Quito clearly reduce during the strike (11.13 µg m −3 ) if compared to before (12.92 µg m −3 ) and after (14.37 µg m −3 ) the strike as well. This shows a 13.85% reduction during the protests, if compared to the levels before the strike. Overall, regional changes from the reduced urban activities are clearly visible near Quito (Fig. 1). This is expected, as one of the biggest polluting sources in Quito (and Ecuador) is a poor-quality fuel used by older technology motorized fleet 39 . Therefore, our findings point to a positive regional impact of protests with paralysis of motor vehicular circulation on the environmental air quality in Ecuador, and most likely in other South American countries. PM 2.5 concentration change. While satellite data can provide an invaluable picture of the regional effects of strikes, they cannot possibly display the urban pollution in different parts of the city in greater detail. Hence, Table 1 shows statistics of surface level PM 2.5 data for seven monitoring sites of Quito for the three 12-day periods: before (9/20/2019-10/01/2019), during (10/02/2019-10/13/2019) and after (10/14/2019-10/25/2019) the strike. Most of the sites show the highest PM 2.5 levels before the strike, with an overall reduction of 13.9% during the strike. These results compare very well with the satellite data (Fig. 1). The same data can be visualized, in spatial interpolation graphs for each period in Fig. 2. It can be confirmed that most of the sites show a significant reduction of the PM 2.5 concentrations during the protests. There are two exceptions. S1-Carapungo, the site in the northern outskirts of the capital, indicates a small increase in PM 2.5 concentrations during the 12 days of strike. This increase might be due to the barricades and burning events near the entrances to the city in the outskirts. Meanwhile the central S3-Belisario site shows an increase in PM 2.5 concentrations after the strike, which agrees with satellite data post-strike. Post-strike weather became rainier and more humid ( Figure A1, Appendix 1), which might influence an increase in vehicle use and worsened combustion and thus increased anthro- www.nature.com/scientificreports/ pogenic PM 2.5 emissions in busy traffic areas 40 . While these data might be helpful to estimate urban population's exposure to air pollution, it might be skewed by the variation in meteorological conditions. PM 2.5 concentration change: machine learning modelling. To unbiasedly estimate the impact of the strike on PM 2.5 levels, the real PM 2.5 concentrations were compared to the business-as-usual PM 2.5 concentrations predicted by ANN models (Fig. 3). This approach has been proven more accurate rather than comparing the pollution changes during the protest to the real data from another period 41 , since the estimated data are pro-  www.nature.com/scientificreports/  www.nature.com/scientificreports/ vided by meteorology-normalized models 42 . Before the strike, the validation of the model can be verified as the modeled data mostly overlap with the actual PM 2.5 concentrations (the estimated values in Fig. 3 are within the standard deviation of the observed values). The predicted data (red line in Fig. 3) are obtained from the model getting the highest performance (see Table A1, Appendix 1, for the assessment of the modeling accuracy). Consequently, the business-as-usual air quality is determined by LSTM for S1-Carapungo, S3-Belisario, S5-Camal, S6-Guamani and S7-Chillos, and simple RNN for S2-Cotocollao and S4-Centro. The model accuracy is high for S2-Cotocollao, S3-Belisario, and S7-Chillos (RMSE < 10 µg m −3 ). The performance is good for S1-Carapungo, S4-Centro and S5-Camal (RMSE < 13 µg m −3 ), and only fair for S6-Guamani (RMSE = 15.6 µg m −3 ). Missing data in the training set can explain a lower accuracy for this latter model. Overall, the performance of the models is high enough to provide us with a reliable prediction of the business-as-usual air quality conditions. It can be observed that daily PM 2.5 concentrations decreased during the strike (− 10.75 ± 17.74%) over the whole Andean city if compared to business-as-usual (red shaded area in Fig. 3). This can be attributed to the limitations of traffic activities, due to the forced aggressive stagnation of any type of transportation. To confirm this, the sites that demonstrate the biggest reduction are S3-Belisario (− 47.84 ± 14.40%, Fig. 3c) and S4-Centro (− 12.98 ± 73.50%, Fig. 3d). These two sites are located in the central traffic-busy areas. However, S4-Centro is the site closest to the central protest activity at El Ejido park, thus showing a lower PM 2.5 reduction than in S3-Belisario. The paralysis of the use of buses and private cars can justify the greatly reduced particulate pollution, previously also seen in another study in Quito related to the strict regulations to slow down COVID-19 infections 43 . However, in the areas around the main road blockages ( Figure A2, Appendix 1), the overall elevated pollution levels and an increase in PM 2.5 concentration variations can be perceived. The former can be identified in the northern and southern city outskirts near S1-Carapungo (7.03 ± 63.72%, Fig. 3a) and S7-Chillos (1.26 ± 69.74%, Fig. 3g). Meanwhile the increased pollution variation (standard deviation) can be observed in S1-Carapungo  Figure A2, Appendix 1). Due to the complex geography of this Andean city, main roads often bottleneck at tunnels, bridges, serpentine-highways around numerous canyons and mountains. This last finding must be stressed, as most of the city had blocked private and highly polluting public transport circulation. Often, road blockages would have obstructions and active burning activity. As a result, during the strike, towers of smoke could be seen all over the city ( Figure A3, Appendix 1). The smoke dispersion and fire activity might have been enhanced due to the fact that during the strike the accumulation of precipitation was very low ( Figure A1, Appendix 1). This may help explain a high standard deviation of the PM concentrations in many sites of the city, as plumes of smoke would be transported to the monitoring sites. Our results indicate that even when most of the normal activities are seized in the urban area due to the protests, other types of unusual activities might keep worsening the air quality.
Furthermore, we investigated the particulate pollution evolution 2 weeks after the protests, in order to study the "return speed" of atmospheric pollution to the business-as-usual conditions. The overall change of pollution was − 6.8% averaged over 12 days after the protest. Interestingly, while some sites returned to "normal" levels (S1-Carapungo, S4-Centro and S5-Camal), the rest of the sites in this high elevation city show a continuous reduction in PM 2.5 pollution levels even 12 days after the national strike. This could be due to the time lag to get back to normal activities, especially, transportation of products and merchandize. We want to point out the social behavior during these unusual events like stocking up on food and water. This might have influenced the reduction of certain anthropogenic activities (e.g. grocery shopping or local and international tourism, etc.) for some time after the strike. In addition, when the strike was announced, the population might have increased their shopping a few days before the protests start, which might be observed as a slight increase from the businessas-usual levels ( Fig. 3b-f). A further study is required to investigate the time needed to go back to business-as usual pollution levels. However, an increase in PM 2.5 concentrations can already be seen in most sites at the end of the study period (see Fig. 3b-g).
The meteorology normalized approach is essential to get a fine and accurate analysis of the long-range impact of the strike on air pollution. The naïve observation of the raw data shows a 11.9% increase of the PM 2.5 concentration in Belisario after strike (Table 1). This value contradicts the 15.9% decrease estimated by the machine learning technique (Fig. 3). This difference can be explained by the fact that the meteorological conditions are taken into account in the ANN method but not in the observational study. Relative humidity (RH) and wind speed (WS) have a strong influence on the PM 2.5 concentration in Quito 44 . The higher the RH, the higher is the concentration. On the contrary, WS tends to reduce the level of fine particulate matter in the atmosphere. In Belisario before strike, the WS was stronger (1.42 m/s), and the RH was lower (65.5%) than after strike (WS = 1.14 m/s; RH = 75.5%). It is consequently not possible to reveal the pollution reduction effect of the strike, because it is masked by the high value of meteorological parameters that negatively affect air quality ( Figure A1, Appendix 1, for more details about the meteorological conditions). Since the machine learning modeling is meteorology normalized, the prediction of PM 2.5 concentration for business-as-usual is not impacted by this artifactual underestimation.
Plant sample chemical analysis. Distribution patterns of the concentrations, in μg g -1 DW, of the metals found in the Melaleuca armillaris leaves, collected at the urban park, housing the most intense protest activities, are presented in Fig. 4. Each of the metals is presented in a separate panel with the range of concentration from minimum to maximum values, depending on metal abundance in the sample. In general, the levels of all metals are higher at points E1 and E7 (see Fig. 4), where most of the visible burn signs were observed during the recollection of the leaf samples. In addition, in those sampling sites tree leaves were profusely covered in soot. Highest levels at those sites were especially observed for Co (Fig. 4a), Cr (Fig. 4b), Zn (Fig. 4c), Al (Fig. 4d), Fe (Fig. 4e) www.nature.com/scientificreports/ Pb (Fig. 4f), Mg (Fig. 4g) and Ba (Fig. 4h). Those metals are typically used in tire manufacture 27,28,45 . Cu is also known to make part of the composition of tires 27,28 , and its concentration was higher at point E4 (Fig. 4i). At point E4, visible burn signs were not observed at the moment of the sampling. However, it is located just in front of a street where there was a barricade (Fig. 5c) with an intense burning of materials during the protests, mainly tires. At E4 an elevated concentration of K, which is a tracer of biomass burning, was also observed (Fig. 4j). While K is often a good marker of biomass burning, it is also one of the chemical components naturally present in the plants. This might suggest that while at E1 and E7 mostly tires were burned, at E4 there was a mix of pollutants. As previously mentioned, protesters often burn wood to reduce the eye burning due to the tear gas, but also burn tires and other objects to create a dark smoke and block roads ( Figure A3, Appendix 1). Finally, site  www.nature.com/scientificreports/ E6 shows increased levels of Mn, Al and Fe, which are metals associated with both natural and road traffic emissions, because they are also components of steel and alloys widely used by the automotive industry 37,46 . Point E6 is near a major road with bus traffic, which could explain the high concentration of these metals accumulating over time 47,48 .
To be able to understand how the concentrations of metals found in an urban park after the protest weight against concentrations in urban areas with different levels of contamination 49 , we compared metal concentrations found in Melaleuca armillaris and Araucaria heterophylla. The analysis showed a higher capture capacity of the Araucaria heterophylla needles in relation to the Melaleuca armillaris leaves at point E4 (see Table A2). Table A2 shows that the concentration of Zn, Al, Pb, Fe, Cu and K in the Araucaria heterophylla needles collected at point E4 compared well to the areas with high and moderate vehicular traffic intensity, which are areas considered to have high concentration of pollutants 49 . This, together with the fact that the levels of most of the metals analyzed in the Melaleuca armillaris leaves were higher at points where visible burn signs were observed during the recollection of the leaf samples (E1 and E7), could indicate that the concentration of metals increased in this area during the protests due to the burning of materials.
As a result, our study helps understand the importance of the toxicity of urban PM during such events. While overall concentrations of PM 2.5 seem to have dropped, there are peak concentrations generated by burning different objects that might produce high-level toxicity plumes. It is a quite different situation compared to COVID-19 quarantine regulation improvements on air quality and health seen worldwide 50 . It is known that even low levels of atmospheric particulates, loaded with heavy metals, may build-up in soils, plants and either right away or over time transport to the lungs of urban population and accumulate. It might be of a significant concern, as it has been shown before, that levels of pollution, specifically heavy metals in PM, in developing countries result in much higher concentrations of heavy metals in human matrices (e.g. blood, breast milk, etc.) 51 . Which might cause an extensive range of health conditions, including lung cancer, elevated blood pressure, organ failure (e.g. kidney), bone damage, and developmental and neurobehavioral disorders 52 . A number of studies consistently  (panels b,c). Geographical Information Software QGIS 3.18.2 was used ( available at https:// qgis. org/ downl oads/), employing ESPG 4326 (WGS 84) (panels a-c), Ecuador limits were retrieved from Geographic Military Institute (retrieved from https:// sni. gob. ec/ geose rvici os-ecuad or), Urban Growth of Quito was used for Quito limits (retrieved from http:// gobie rnoab ierto. quito. gob. ec/? page_ id= 1122), Quito's monitoring sites were retrieved from Secretaría de Ambiente from Quito (http:// www. quito ambie nte. gob. ec/ index. php/ gener alida des), Road blocks were retrieved from Rcuadorian newscasts (available at https:// www. primi cias. ec/ notic ias/ socie dad/ quito-manif estac ionesvias/ and ) and Satellite image was used from XYZ Tiles service from Google (available at http://mt0.google. com/vt/lyrs=t&hl=en&x={x}&y={y}&x={z}). www.nature.com/scientificreports/ suggest that the WHO may be under-estimating air pollution impacts in developing countries, especially due to short-term exposure to PM 53,54 . These findings raise a concern for immediate actions to control toxic pollutant peaks in densely populated urban areas.

Conclusions
In this paper the effect of national protest of 2019 in Ecuador on particulate matter concentrations was estimated. Satellite data indicates that the main changes were observed in capital Quito, where PM 2.5 levels reduced during the 12-day (October 2-13, 2019) political protest. While regional effect of PM 2.5 pollution from the Colombian cities and active volcano ash plumes can be detected in this Andean country, the protest had a positive regional impact on environmental quality. A simple comparison of PM 2.5 levels before, during and after the strike, confirms the findings of satellite data, these results might be skewed by the changes in meteorological conditions. Thus, the use of meteorologynormalized Machine Learning approach (Artificial Neural Network) estimated daily PM 2.5 concentrations reduction of − 10.75 ± 17.74% during the strike due to the limitations of traffic activities. The biggest reduction was found in the traffic-busy central area (− 47.84 ± 14.40%), in contrary to other barricade burning areas. It can be concluded that even when most of the anthropogenic activities are seized in an urban area due to the protests, other types of unusual activities might keep worsening the air quality.
Elemental PM analysis of the Melaleuca armillaris plant from the central protest action spot-an urban park-shows increased levels of most anthropogenic metals at points with visible burn signs. This is especially observed with Co, Cr, Zn, Al, Fe, Pb, Mg, Ba and Cu, which are metals typically used in tire (common object burned during protest) manufacture. In addition, at one of these points an elevated concentration of a biomass burning tracer K was also observed.
Our study contributes to the understanding of the importance of urban PM toxicity during such unusual events, however, escalating in occurrence. While at the city and regional levels, concentrations of PM decreased, there were peak concentrations generated by burning different objects that might produce high-level toxicity plumes. It might be of a significant concern, as even short-term exposure to toxic PM might cause a wide range of health problems, including lung cancer, elevated blood pressure, organ failure (e.g. kidney), bone damage, and developmental and neurobehavioral disorders. These findings call to not overlook the importance of toxic spikes produced by these events in the context of an overall reduction of air pollution.

Methods
Study sites and data analysis. High elevation (2850 m above sea level (m.a.s.l.)) Ecuadorian capital Quito is located between the western and eastern branches of the Andes mountains (Fig. 5a). The Metropolitan District of Quito (DMQ), located on the equatorial line, stretches over a variety of elevations on the side of Pichincha volcano (elev. 4800 m.a.s.l.) and inner Andean valleys (approx. elev. 2300 m.a.s.l.), causing a formation of microclimates ranging in precipitation accumulation 40 . The DMQ has almost 3,000,000 inhabitants 55 and a rapid growth in motorized fleet running on poor-quality fuels, which considerably affects air quality 39,56 .
In the present investigation of air quality changes due to the political protests in Quito, PM 2.5 concentration data were taken from the environmental monitoring network of the Secretariat of the Environment of DMQ. Out of nine study sites, distributed across the city, data from seven sites were analyzed: S1-Carapungo (elev.  Fig. 5b). The air quality and meteorological data were collected through the methods in accordance with the guidelines of the United States Environmental Protection Agency 57 , previously described in detail 39 . The ground-based monitoring data were collected during September 20-October 25, 2019, to include the 12-day periods before, during and after the strike. Daily averages and accumulation (rain only) were processed. In addition, 12-day statistics for some parameters (PM 2.5 , wind speed and relative humidity) were calculated. For daily and 12-day data analysis and visualization, MS Excel and Igor Pro 6.0 (Wavemetrics Inc.) softwares were used.
Additionally, Melaleuca armillaris leaves were sampled in the El Ejido Park (Fig. 5c). This park was the hot spot of the protests, and where a large number of materials, such as wood and tires, were burned continuously throughout the strike. The plant samples were taken at seven different points distributed across the urban park (Fig. 5c, points E1-E7). In some of these points, there were obvious signs of burning at the moment of the sampling collection (Fig. 5c, red markers), while in others not (Fig. 5c, magenta markers). It is to be mentioned that, although at the point E4 there were no signs of burning, this point is located just in front of a blockage of the road, where tires were burned, thus we marked it as one of the intense burn sites (see Fig. 5b and c, red markers). www.nature.com/scientificreports/ earth-engine/ datas ets/ catal og/ ECMWF_ CAMS_ NRT? hl= en. The visualization was performed using ARCGis PRO software.

Machine learning modelling.
A machine learning (ML) approach was used to quantify the effect of the protests on fine particulate matter concentrations, in seven districts of Quito (see "Satellite PM 2.5 concentration da" section). Artificial Neural Networks (ANN) were trained to predict business-as-usual PM 2.5 from hourly measurements of meteorological and temporal variables. Since the models are built from meteorological parameters, the prediction keeps accurate even if unusual climatic conditions occur 41 . This is the reason why this approach is also known as meteorology-normalized modelling 42 . Seven meteorological features were selected: relative humidity, precipitation, temperature, solar radiation, pressure, wind speed and wind direction. The temporal attributes were year day, weekday, hour and trend (date index). Our dataset is composed of 41640 instances, which is suitable for training an ANN. The dataset was split into a training set (1/1/2015-4/1/2019) and a testing set (4/2/2019-10/1/2019). It means that the accuracy of the models was tested in the six months prior to the protests. The Root Mean Squared Error (RMSE) and the Coefficient of Determination (R 2 ) were the metrics chosen to assess the performance of the models. To validate a model, the RMSE must be as small as possible, and the R 2 the closest to 1.
The ANN models were used to estimate the business-as-usual PM 2.5 concentration. Then, the concentration change was computed from the difference between ML-based business-as-usual and actual value measured during the protests. A specific type of ANN, called Recurrent Neural Network (RNN), was chosen for being especially designed to capture information from sequences, such as time series data. The algorithm implements a recursive method, in which a new state of the model at time t is a function of its previous state (S t-1 ) and the input (meteorological and temporal features) at time t, as described in Eq. 1: where X t is the input a timestep t; S t is the state at timestep t; and F w is the recursive function. An adaptation of the simple RNN, called Long Short-Term Memory (LSTM), can also be applied for time series estimation of fine particulate matter, in which the prediction is long-term dependent (i.e., the accuracy is highly dependent on the number of states considered). This is the reason why these two types of RNN were tested and evaluated in this study. For each district, the change of concentration was quantified from the method (simple RNN vs LSTM) that provides the most accurate prediction (i.e., the lowest RMSE). All the deep ANN models were developed in Python, using the open-source library Keras.
Plant description and sampling. In this study, Melaleuca armillaris plant was used to help understand the strike PM chemistry. Melaleuca is a shrub or small tree (see Figure A4, Appendix 1) genus of 290 species in the myrtle family, Myrtaceae, originating from Australasian region. Their flower generally forms a showy spike resembling a brush, up to 7 cm long, with many flowers generally white 61 . A specimen of the species is deposited at Herbarium QCA and HUTI in Ecuador (collection number Oleas # 1053). All collecting procedures followed the guidelines approved by the Ministerio del Ambiente in Ecuador, described in detail under permit number MAAE-ARSFC-2020-0778. The species was identified by Nora Oleas, using the key at https:// plant net. rbgsyd. nsw. gov. au/ cgi-bin/ NSWfl. pl? page= nswfl & lvl= gn& name= Melal euca# 11.
Melaleuca armillaris leaves were sampled in the El Ejido Park (Fig. 5c) in the morning of 15th of October 2019, less than two days after the strike ended and before the next rain event ( Figure A1, Appendix 1). We note that sample collection during the strike was impossible for personal safety due to the aggressive riots between the protesters and the police officers ( Figure A3, Appendix A1). The collection of the leaves was carried out following the procedure previously used by the research group 36,37 . Briefly, nearby 20 g of leaves were taken using gloves and kept in polyethylene bags. The samples were taken from the four directions of the trees at a height of about 2 m above the ground. The samples were transported to the laboratory and conserved at 4 °C in the dark until analysis. Part of the samples (about 2 g) were dried in triplicate at 70 ± 2 °C in order to report the metal concentrations on a dry weight.

Plant sample preparation and analysis.
The details about digestion process and quantification of metals in leaves have been fully explained elsewhere 36,37 . In brief, 7 mL of HNO 3 , 2 mL of H 2 O 2 and 1 mL of H 2 O were added to 0.5 g of fresh leaves and heated at 200 °C during 45 min in a microwave digester (MARS 6 -CEM Corporation). The extraction for each point was performed in triplicate. The samples were filtered followed by adjusting the volume to 25 mL with Milli-Q water, and the concentrations of metals were measured using an Inductively Coupled Plasma-Optical Emission Spectroscopy (ICP-OES, Thermo Scientific iCAP 7000 Series). The metals included in the composition of tires, with given certified values and recoveries in the range of 43.41-91.14%, were analyzed (i.e., Al, Fe, Zn, Pb, Cu, Mg, Co, Ba and Cr). K (% Recovery 78.89) was also quantified, since it is a metal commonly detected during biomass burning. Moreover, more natural metals, such as Ca and Mn (%Recovery 88.02 and 111.01, respectively), were also measured. For the recovery percentage, the certified reference material NIST SRM 1575a-Trace elements in Pine Needles was used. The limits of detection (LOD) were calculated as 3 times the standard deviation of 10 blanks measurements divided by the slope of the analytical curve, while the limits of quantification (LOQ) were calculated similarly by multiplying the standard deviation by 10. The range values for LOD and LOQ were 2.58 × 10 −7 -0.0088 μg g −1 and 8.59 × 10 −7 -0.029 μg g −1 , respectively.
In order to have an idea if the concentrations of metals found in this area were high-since there was no data from before the protest events -an extra sample of Araucaria heterophylla needles was collected at point E4 for the same chemical analysis. The Araucaria heterophylla concentrations of metals in the urban park were then (1) S t = F w (S t−1 , X t ) www.nature.com/scientificreports/ compared with those found in the streets of Quito with a range in vehicular traffic intensity 49 . It is important to compare the concentrations using the same plant species since it has been widely reported in literature that the accumulation capacity of plants for pollutants is species dependent 62,63 .
To present plant chemical analysis in an urban park, QGIS 3.18 software was used. EPSG.32717 (WGS84-UTM17 South) coordinate system was used. An Inverse Distance Weighting (distance coefficient p = 2; Grid resolution 1 m) was performed for data interpolation between the sampling points.
For the visualization of the ECMWF raster of PM 2.5 the ArcGis PRO software was used, with a resolution of 44 km /pixel, with a WGS84 projection system with ESPG 4326 coordinate system.