Analysis of atmospheric air pollutants (CO, NO 2, and SO 2), through Sentinel-5P images in google earth engine, apurimac region, period 2020–2023

Atmospheric pollution associated with carbon monoxide (CO), nitrogen dioxide (NO 2), and sulfur dioxide (SO 2) represents an important environmental and public health concern in Andean regions influenced by urban growth, transportation, agricultural burning, mining-related activities, and complex topography. This study analyzed the spatiotemporal variability of CO, NO 2, and SO 2 in the Apurimac region, Peru, during 2020–2023 using Sentinel-5P/TROPOMI satellite products processed in Google Earth Engine. The original column-density data, expressed in mol/m 2, were converted into column-derived estimated concentrations using a simplified effective lower-atmospheric layer height of 2,000 m. These values were used only as relative indicators and were not interpreted as direct ground-level air quality measurements. The results showed heterogeneous spatial patterns across Apurimac. CO presented relatively higher values mainly in Abancay, Chincheros, and Andahuaylas, while NO 2 showed higher relative values in Cotabambas, Grau, Abancay, and Chincheros. SO 2 exhibited a more irregular and uncertain behavior, with localized positive values, negative retrievals, and high variability, indicating greater sensitivity to satellite retrieval noise. Therefore, SO 2 was interpreted as an exploratory indicator rather than robust evidence of province-level pollution hotspots. Temporal analysis showed seasonal fluctuations and isolated peaks, especially during dry-season months. Trend detection was performed using the Mann-Kendall test and Sen’s slope estimator. CO showed a statistically significant decreasing trend, NO 2 showed a weak but statistically significant increasing trend, and SO 2 did not show a statistically significant trend. Overall, Sentinel-5P/TROPOMI and Google Earth Engine proved useful for identifying relative pollutant patterns in a data-sparse Andean region. Future studies should integrate satellite observations with in situ measurements, meteorological data, boundary-layer information, emission inventories, and independent datasets to improve validation, source attribution, and regional air quality assessment.

Air pollution is one of the most important environmental and public health problems worldwide. According to studies reported by the World Health Organization, one out of every nine deaths is associated with atmospheric pollution (World Health Organization, 2016). Similarly, the Pan American Health Organization emphasizes that the quality of the air people breathe is a determining factor for human health, particularly because exposure to polluted air is closely related to respiratory diseases and premature mortality (Pan American Health Organization, 2026). Several studies have indicated that atmospheric pollution is a major environmental issue in industrialized and developing countries, producing impacts at local, regional, and national scales (Olmo et al., 2011; Lavaine, 2015). The World Health Organization has also reported that a large proportion of the global population is exposed to environments with pollutant levels that may affect human health (World Health Organization, 2016). Exposure to atmospheric pollutants such as nitrogen dioxide (NO 2), particulate matter, carbon monoxide (CO), ozone (O 3), methane (CH 4), and other gases has been associated with respiratory, cardiovascular, and reproductive health effects (Perez et al., 2013; Cesaroni et al., 2014; Slama et al., 2013). Among these pollutants, NO 2 is particularly relevant because it is linked to respiratory and cardiovascular diseases, reduced respiratory tract defense capacity, lung immune system weakening, and asthma (World Health Organization, 2021).

Nitrogen dioxide is mainly produced by combustion processes, including vehicle emissions, fossil fuel use, and industrial activities. Although NO 2 is not considered one of the main greenhouse gases, it plays an important role in atmospheric chemistry because it contributes to the formation of secondary aerosols, photochemical smog, and acid precipitation (Olmo et al., 2011; Mak and Ng, 2021). In urban and industrial areas, NO 2 is one of the most relevant pollutants for air quality assessment because its concentrations are strongly associated with traffic, energy use, and industrial emissions (Baldasano, 2020; Vîrghileanu et al., 2020). In addition, emissions from combustion-related activities contribute to changes in atmospheric composition and may interact with climate-related processes. Studies on environmental pollution have also shown that pollutants derived from deforestation, industrial processes, and combustion activities can affect air quality and increase the risk of respiratory diseases in exposed populations (Perez et al., 2013; Veettil and Kamp, 2021; Li et al., 2017).

Carbon monoxide is another important atmospheric pollutant, especially in areas influenced by incomplete combustion processes. According to Borsdorff et al. (Borsdorff et al., 2018), CO is a toxic gas that can have harmful effects on human health when concentrations are elevated. This pollutant is commonly associated with vehicle emissions, biomass burning, industrial processes, and domestic combustion. CO is also considered a relevant indicator of combustion sources because it can be transported over long distances and detected using satellite observations. During the COVID-19 pandemic, several studies reported reductions in atmospheric pollutants such as NO 2 and CO due to decreases in industrial activity, mobility restrictions, and reduced transportation emissions (Shami et al., 2022; Vellalassery et al., 2021). A global reduction in NO 2 levels during 2020 was also observed using satellite data, mainly associated with the restrictions implemented during the pandemic period (Bauwens et al., 2020).

Sulfur dioxide is mainly emitted from the combustion of fossil fuels and industrial activities, although natural sources such as volcanic eruptions can also contribute to its presence in the atmosphere. SO 2 has negative effects on human health, especially when exposure occurs at high concentrations, and it contributes to environmental degradation through the formation of acid rain (Jion et al., 2023). Previous studies have identified transportation, manufacturing, electricity generation, fossil fuel combustion, and other economic activities as relevant sources of SO 2 emissions (Opio et al., 2021; Wang et al., 2022). Domestic activities such as burning garbage, charcoal, or wood may also contribute to local air pollution, particularly in areas with limited environmental control (Langley DeWitt et al., 2019). According to information reported by the European Space Agency, SO 2 can originate from both natural and anthropogenic sources, although most emissions are associated with human activities.

The development of satellite-based Earth observation systems has improved the monitoring of atmospheric pollution at different spatial and temporal scales. The Sentinel-5 Precursor satellite, part of the Copernicus programme and the European Space Agency Sentinel missions, represents an important advance for atmospheric monitoring because it provides open-access observations for environmental applications (Paul et al., 2012; Veefkind et al., 2012). Sentinel-5P carries the TROPOspheric Monitoring Instrument (TROPOMI), which provides daily global measurements of atmospheric species relevant to air quality and climate studies (Tian et al., 2023). TROPOMI observations have been widely used to analyze atmospheric pollutants and have been evaluated, modeled, and validated in different environmental applications (Zeng et al., 2019; Beirle et al., 2019). These satellite products provide independent measurements for several pollutants and have been compared with other remote sensing and atmospheric monitoring systems (Zeng et al., 2019; Beirle et al., 2019). The instrument measures in the ultraviolet, visible, near-infrared, and shortwave infrared spectral ranges, allowing the retrieval of atmospheric trace gases such as NO 2, CO, CH 4, and O 3 (Paul et al., 2012; Veefkind et al., 2012).

Sentinel-5P/TROPOMI data are particularly useful for studying air pollution in regions where ground-based monitoring stations are limited or unavailable. The high temporal frequency of Sentinel-5P observations allows the analysis of daily and seasonal changes in atmospheric pollutants, supporting the identification of spatial patterns and temporal variability (Grzybowski et al., 2023; Dhankar et al., 2024). Previous studies have used Sentinel-5P data to evaluate changes in atmospheric pollution during the COVID-19 pandemic and to analyze the relationship between pollutant concentrations and public health conditions (Tian et al., 2023). In addition, satellite-based studies have shown that air pollution in urban areas can be reduced when appropriate policies are implemented to regulate traffic, vehicle circulation, industrial emissions, and other human activities (Shami et al., 2022).

In this context, the Apurimac region of Peru represents an important study area due to its complex Andean topography, urban growth, transportation routes, agricultural burning practices, and mining-related activities. However, continuous ground-based air quality monitoring remains limited in many Andean regions, making satellite-based approaches useful for regional environmental assessment. Therefore, Sentinel-5P/TROPOMI data processed in Google Earth Engine provide an opportunity to identify relative spatial and temporal patterns of atmospheric pollutants in Apurimac during recent years.

This study aims to analyze the spatiotemporal variability of satellite-derived CO, NO 2, and SO 2 patterns in the Apurimac region, Peru, from 2020 to 2023 using Sentinel-5P/TROPOMI data processed in the Google Earth Engine environment. Specifically, the study seeks to identify areas with relatively higher column-derived pollutant values, evaluate temporal variability using non-parametric trend analysis, and provide satellite-based information that may support regional air quality screening in an Andean region with limited ground-based observations. The results are interpreted as column-derived estimated indicators rather than direct ground-level air quality measurements.

Study area

The study area corresponds to the Apurimac region, located in the south-central Andes of Peru. Apurimac is characterized by complex mountainous terrain, marked altitudinal variability, and rugged topography, conditions that influence local climate, land-cover distribution, atmospheric circulation, and pollutant dispersion (Villacorta et al., 2016). Administratively, the region comprises seven provinces: Abancay, Andahuaylas, Chincheros, Aymaraes, Antabamba, Grau, and Cotabambas. It borders Ayacucho to the west, northwest, and southwest; Cusco to the north, northeast, east, and southeast; and Arequipa to the south (Figure 1). Official statistical information for Apurimac is compiled by the National Institute of Statistics and Informatics of Peru, which provides territorial, environmental, demographic, and socioeconomic data for the region (Instituto Nacional de Estadística e Informática, 2024).

The region covers approximately 2,111,415.20 ha and contains diverse land-cover types, including Andean grasslands, shrubland formations, agricultural areas, and natural and planted forests (Huacani et al., 2022; del Ambiente, 2015). According to the National Vegetation Cover Map of Peru, land-cover units are classified based on geographic, physiognomic, humidity, and floristic criteria, which supports the characterization of heterogeneous Andean landscapes such as Apurimac (del Ambiente, 2015). This environmental heterogeneity is relevant for atmospheric pollution studies because land use, seasonal agricultural burning, urban growth, transportation corridors, and mining activities may influence the spatial distribution of CO, NO 2, and SO 2.

Climatically, Apurimac presents a marked seasonal regime associated with its Andean setting. The wet season occurs mainly from December to March, while the dry season extends during the remaining months of the year (Huacani et al., 2022; Servicio Nacional de Meteorología e Hidrología del Perú, 2021). This seasonality is important for interpreting atmospheric pollutant behavior, since dry months may favor biomass burning, dust resuspension, and reduced atmospheric dispersion, whereas wet months may contribute to pollutant removal through precipitation. Therefore, the geographic, climatic, and land-cover characteristics of Apurimac make it a suitable region for analyzing relative spatiotemporal pollutant patterns using Sentinel-5P/TROPOMI satellite data and Google Earth Engine.

Sentinel-5p/TROPOMI datasets

For this study, atmospheric pollutant data were obtained from the Sentinel-5 Precursor satellite mission, which carries the TROPOspheric Monitoring Instrument (TROPOMI). Sentinel-5P is part of the Copernicus programme and provides global observations of atmospheric constituents related to air quality, climate, and atmospheric chemistry (Copernicus Data Space Ecosystem, 2026). TROPOMI measures several trace gases, including carbon monoxide (CO), nitrogen dioxide (NO 2), and sulfur dioxide (SO 2), which are associated with fossil fuel combustion, biomass burning, industrial activity, volcanic emissions, and other natural and anthropogenic sources (European Space Agency, 2026).

The Sentinel-5P Level-3 Offline products available in Google Earth Engine were selected for the period from 1 January 2020 to 31 December 2023. OFFL products were used because the study corresponds to a retrospective environmental analysis rather than a near-real-time monitoring application (Google Earth Engine Data Catalog, 2026a; Google Earth Engine Data Catalog, 2026b; Google Earth Engine Data Catalog, 2026c).

Three datasets were used: COPERNICUS/S5P/OFFL/L3_CO for CO, COPERNICUS/S5P/OFFL/L3_NO2 for NO 2, and COPERNICUS/S5P/OFFL/L3_SO2 for SO 2 (Google Earth Engine Data Catalog, 2026a; Google Earth Engine Data Catalog, 2026b; Google Earth Engine Data Catalog, 2026c). The selected bands were CO_column_number_density, tropospheric_NO2_column_number_density, and SO2_column_number_density, respectively. All variables are provided in mol/m 2 in the Google Earth Engine Data Catalog (Google Earth Engine Data Catalog, 2026a; Google Earth Engine Data Catalog, 2026b; Google Earth Engine Data Catalog, 2026c). The NO 2 tropospheric column was selected because it is more appropriate than the total NO 2 column for analyzing lower-atmospheric pollution patterns related to traffic, combustion, and industrial emissions. However, the selected Sentinel-5P/TROPOMI variables represent atmospheric column-density products expressed in mol/m 2 and not direct ground-level concentrations. Therefore, these satellite-derived column-density observations were used as input data to evaluate relative spatial patterns and temporal variability of CO, NO 2, and SO 2 in the Apurimac region.

For SO 2, the selected product was interpreted with particular caution because weak SO 2 signals and retrieval uncertainty may produce negative column-density values, especially in areas without strong emission sources. These values were not considered direct physical negative concentrations, but indicators of retrieval uncertainty. The treatment of SO 2 negative values and quality screening is described in the preprocessing and methodological limitations sections.

Data preprocessing in google earth engine

All satellite data preprocessing was performed in Google Earth Engine using JavaScript scripts. This platform was selected because it enables large-scale geospatial analysis, direct access to Earth observation datasets, raster processing, temporal aggregation, and extraction of spatial statistics over user-defined regions (Gorelick et al., 2017). Previous studies have demonstrated the applicability of TROPOMI observations for monitoring CO, NO 2, and SO 2 at different spatial scales (Borsdorff et al., 2018; Theys et al., 2019; Ialongo et al., 2020).

The administrative boundaries of the Apurimac region and its seven provinces were imported into Google Earth Engine as vector layers and defined as the region of interest. These boundaries were used for spatial filtering, clipping, visualization, and extraction of regional and provincial statistics. The boundaries were consistent with the official provincial division of Apurimac reported by the National Institute of Statistics and Informatics of Peru (Instituto Nacional de Estadística e Informática, 2024). The Sentinel-5P/TROPOMI OFFL Level-3 image collections were then loaded from the Google Earth Engine Data Catalog and filtered for the 2020–2023 period (Google Earth Engine Data Catalog, 2026a; Google Earth Engine Data Catalog, 2026b; Google Earth Engine Data Catalog, 2026c).

For each pollutant, the corresponding column-density band was selected. The CO_column_number_density band was used for CO, the tropospheric_NO2_column_number_density band was used for NO 2, and the SO2_column_number_density band was used for SO 2. The Sentinel-5P products available in Google Earth Engine are provided as Level-3 gridded datasets derived from original Level-2 products through the harpconvert tool and spatial binning procedures. Therefore, this study used the catalog-defined Level-3 products and their product-specific quality and validity filtering rather than raw Level-2 retrievals. According to the Google Earth Engine Data Catalog, the Level-3 conversion removes low-quality pixels using product-specific QA or validity criteria during ingestion into Earth Engine (Google Earth Engine Data Catalog, 2026a; Google Earth Engine Data Catalog, 2026b; Google Earth Engine Data Catalog, 2026c).

For SO 2, particular caution was applied during interpretation because weak SO 2 signals and retrieval uncertainty may produce negative column-density values, especially over clean regions or areas without strong SO 2 emission sources. Negative SO 2 values were not replaced by zero, because the Sentinel-5P SO 2 product documentation indicates that negative vertical columns may occur due to retrieval noise and should not be filtered except for extreme outliers lower than −0.001 mol/m 2. Therefore, negative SO 2 values were retained in the time-series analysis as indicators of retrieval uncertainty and were not interpreted as physically meaningful negative concentrations.

After temporal filtering and band selection, each image collection was clipped to the Apurimac regional boundary. Annual mean composites were generated for each pollutant and year from 2020 to 2023 to analyze the annual spatial distribution of CO, NO 2, and SO 2. Monthly mean values were also calculated to construct temporal profiles and evaluate seasonal variability during the study period.

Spatial statistics were extracted at regional and provincial scales using the native nominal scale of each Sentinel-5P/TROPOMI Level-3 product in Google Earth Engine. For each pollutant, mean, minimum, and maximum values were calculated to compare pollutant behavior among provinces and years. The resulting raster layers and tabular outputs were exported from Google Earth Engine for map production, spatial classification, time-series visualization, and complementary GIS analysis. The main preprocessing parameters used in Google Earth Engine are summarized in Table 1.

The preprocessing workflow included dataset selection, temporal filtering, band selection, use of catalog-defined quality-filtered Level-3 products, clipping to the Apurimac region, annual mean composite generation, monthly time-series extraction, regional and provincial statistics, and export of raster and tabular outputs. This workflow is summarized in Figure 2.

Conversion of column density to column-derived estimated concentration

The Sentinel-5P/TROPOMI products used in this study provide atmospheric column-density information rather than direct surface-level concentrations. Therefore, the transformation from mol/m 2 to µg/m 3 was performed only as a simplified approximation to express the satellite-derived column values as column-derived estimated concentrations. Similar approaches have been used in satellite-based air pollution studies to obtain approximate concentration indicators from Sentinel-5P column-density data, while recognizing that more robust surface concentration estimation requires meteorological variables, boundary-layer height, ground-based observations, or atmospheric modelling (Wang et al., 2022; Grzybowski et al., 2023; Ialongo et al., 2020).

The column-derived estimated concentration was calculated using Equation 1: where C is the column-derived estimated concentration in µg/m 3, VCD is the Sentinel-5P/TROPOMI column density in mol/m 2, M is the molar mass of the pollutant in g/mol, H is the assumed effective lower-atmospheric layer height in meters, and 10 6 is the conversion factor from grams to micrograms.

The molar masses used in the conversion were 28.01 g/mol for CO, 46.01 g/mol for NO 2, and 64.06 g/mol for SO 2. The molar masses used in the conversion are summarized in Table 2. In this study, H was set to 2,000 m as an assumed effective lower-atmospheric layer height. This value was used to reduce the excessive dilution associated with a 10 km layer assumption and to provide a more conservative approximation for interpreting lower-atmospheric pollutant patterns in a high-altitude Andean region. However, H = 2,000 m should not be interpreted as a measured planetary boundary layer height. It is a simplified effective height used only for the normalization of satellite-derived column values. The use of a simplified vertical distribution approach follows the general methodological logic applied in previous Sentinel-5P-based pollution studies (Savenets, 2021), but the resulting values remain approximate.

Although this conversion allows the satellite-derived values to be expressed in µg/m 3, the resulting values should be interpreted as estimated indicators rather than direct ground-level measurements. Sentinel-5P/TROPOMI retrieves atmospheric column densities and not in situ surface concentrations. Therefore, the converted values are not equivalent to regulatory air quality measurements and should not be directly compared with WHO guidelines or national ground-level air quality standards.

Consequently, the converted values were used mainly to compare relative spatial patterns, temporal variability, and potential priority areas for further monitoring across the Apurimac region during the 2020–2023 period. This limitation is especially relevant for mountainous regions such as Apurimac, where topography, altitude, meteorological conditions, and vertical pollutant distribution may affect the relationship between satellite column values and near-surface concentrations.

Spatial classification and relative hotspot screening

Annual maps of column-derived estimated values of CO, NO 2, and SO 2 were generated to evaluate relative spatial patterns of atmospheric pollutants in the Apurimac region during the 2020–2023 period. For each pollutant, the annual mean raster layers were classified into five relative classes: very low, low, medium, high, and very high.

The classification thresholds were defined using fixed pollutant-specific ranges based on the value intervals used in the annual pollutant maps. The same classification scale was applied to all annual maps of each pollutant to ensure temporal comparability during the 2020–2023 period. The classification ranges used for CO, NO 2, and SO 2 are shown in Table 3.

Potential relative hotspots were screened by overlaying the classified pollutant maps with the provincial administrative boundaries of Apurimac. Areas classified as high or very high were considered potential priority zones when they appeared in one or more annual maps or when they were observed repeatedly during the study period. These areas were not interpreted as confirmed ground-level pollution hotspots, but as satellite-derived zones with relatively higher values that may require further monitoring.

For each province, zonal statistics were calculated to obtain mean, minimum, and maximum values of CO, NO 2, and SO 2. These values were used to compare relative pollutant patterns among provinces and to support the identification of potential priority areas for future environmental monitoring. Direct source attribution was not performed because the analysis did not include ground-based measurements, emission inventories, fire records, traffic data, or meteorological modelling.

The hotspot screening was interpreted as an exploratory regional approach aimed at identifying zones with relatively higher satellite-derived values. Therefore, the identified areas should be understood as potential priority zones for further monitoring rather than direct evidence of ground-level air quality exceedances. For SO 2, this interpretation was made with particular caution because weak signals, negative retrievals, and retrieval uncertainty may affect the robustness of the spatial pattern.

Temporal variability analysis

Temporal variability was analyzed to evaluate changes in the column-derived estimated values of CO, NO 2, and SO 2 over the Apurimac region during the 2020–2023 period. For each pollutant, the Sentinel-5P/TROPOMI image collections were filtered by date and spatially reduced within the Apurimac regional boundary to obtain regional mean values through time. These time-series values were used to describe temporal fluctuations, isolated peaks, and periods with relatively low satellite-derived values during the study period. Similar time-series approaches using Sentinel-5P/TROPOMI data have been applied by Shami et al. (2022), Tian et al. (2023), and Kazemi Garajeh et al. (2023) to evaluate temporal changes in atmospheric pollutants.

The temporal analysis considered the seasonal characteristics of the study area, particularly the distinction between the wet season and the dry season, as described by SENAMHI (Servicio Nacional de Meteorología e Hidrología del Perú, 2021). This distinction was relevant because dry months may be associated with biomass burning, agricultural activities, dust resuspension, and reduced atmospheric dispersion, whereas wet months may favor pollutant removal through precipitation. However, these associations were interpreted only as possible explanatory factors, because the analysis did not include meteorological modelling, fire records, traffic data, or emission inventories.

Annual mean values were also calculated for each pollutant to compare interannual variability between 2020, 2021, 2022, and 2023. This analysis allowed the evaluation of whether the satellite-derived values showed increases, decreases, or irregular behavior over time. Because Sentinel-5P/TROPOMI provides atmospheric column-density observations rather than direct ground-level measurements, temporal variations were interpreted as relative changes in satellite-derived values rather than direct evidence of surface-level concentration changes.

For SO 2, temporal variability was interpreted with particular caution because weak signals, negative retrievals, and high variability may indicate retrieval uncertainty in areas without strong SO 2 emission sources. Negative SO 2 values were retained in the time-series analysis as indicators of retrieval uncertainty and were not interpreted as physically meaningful negative concentrations.

The temporal profiles were plotted as line graphs to visualize pollutant behavior during the 2020–2023 period. These graphs were used to identify recurrent seasonal behavior, isolated increases, and irregular fluctuations. The temporal variability analysis complemented the spatial classification by showing when relatively high satellite-derived values occurred and whether these events were persistent, seasonal, or isolated.

Statistical trend analysis

Statistical trend analysis was performed using the time-series values of CO, NO 2, and SO 2 obtained from Sentinel-5P/TROPOMI data for the 2020–2023 period. For each pollutant, descriptive statistics, including mean, minimum, maximum, and standard deviation, were calculated to summarize temporal variability.

Because the pollutant time series may be affected by temporal fluctuations, isolated peaks, seasonality, and possible autocorrelation, ordinary least-squares linear regression was not used as the main method for trend inference. Instead, temporal trends were evaluated using the Mann-Kendall test and Sen’s slope estimator. The Mann-Kendall test was applied to identify the presence and direction of monotonic trends in the satellite-derived pollutant series (Mann, 1945; Kendall, 1970), while Sen’s slope estimator was used to quantify the magnitude of temporal change (Sen, 1968).

For each pollutant, the Mann-Kendall statistic, p-value, and Sen’s slope were calculated. Trends were considered statistically significant when p < 0.05. Positive Sen’s slope values indicated increasing tendencies, whereas negative values indicated decreasing tendencies. Since serial correlation may influence the statistical significance of the Mann-Kendall test, the results were interpreted cautiously (Yue and Wang, 2004).

The statistical results were interpreted as temporal tendencies of column-derived estimated values rather than validated ground-level concentration trends. This caution is necessary because Sentinel-5P/TROPOMI retrieves atmospheric column-density information, and the resulting values may be influenced by seasonal emissions, biomass burning, vehicular activity, mining-related sources, meteorological conditions, complex topography, and satellite retrieval uncertainty.

Methodological limitations

This study has several methodological limitations that should be considered when interpreting the results. Sentinel-5P/TROPOMI provides atmospheric column-density observations rather than direct ground-level measurements; therefore, the values converted to µg/m 3 should be interpreted as column-derived estimated indicators and not as in situ air-quality measurements (Borsdorff et al., 2018; Grzybowski et al., 2023; Ialongo et al., 2020).

The conversion from column density to estimated concentration was based on a simplified assumption of uniform vertical distribution within an effective lower-atmospheric layer. In this study, an assumed effective layer height of 2,000 m was used. This value should not be interpreted as a measured planetary boundary layer height for the Apurimac region, but as a methodological approximation used to normalize the satellite-derived column values. Actual pollutant vertical distribution may vary according to emissions, meteorological conditions, atmospheric stability, terrain, and boundary-layer dynamics. Therefore, the converted values are not equivalent to regulatory ground-level concentrations and should not be directly compared with WHO air-quality guidelines or national air-quality standards (Grzybowski et al., 2023; Savenets, 2021).

In addition, the absence of continuous ground-based air-quality measurements limits the direct validation of the satellite-derived estimates. Sentinel-5P retrievals may also be affected by cloud cover, surface conditions, atmospheric variability, and product-specific uncertainty (Wang et al., 2022; Ialongo et al., 2020). For this reason, the results should be interpreted mainly as relative spatial and temporal patterns rather than direct evidence of ground-level air-quality exceedances. Future studies should validate or calibrate the satellite-derived estimates using ground-based measurements, atmospheric models, meteorological reanalysis data, boundary-layer information, emission inventories, fire records, traffic data, and mining activity records.

SO 2 results require particular caution. In areas without strong SO 2 emission sources, Sentinel-5P/TROPOMI SO 2 retrievals may be affected by retrieval noise, and negative column-density values may occur. In this study, negative SO 2 values were not interpreted as physically meaningful negative concentrations, but as indicators of retrieval uncertainty. Therefore, SO 2 patterns were treated as exploratory satellite-derived indicators rather than robust evidence of province-level SO 2 hotspots.

Finally, the Mann-Kendall test and Sen’s slope estimator were used to describe temporal tendencies in the satellite-derived time series. However, these statistical results should also be interpreted cautiously because the series may be affected by seasonality, isolated peaks, possible autocorrelation, retrieval uncertainty, and local atmospheric conditions. Therefore, the trend results describe temporal tendencies in column-derived estimated values, not validated causal trends in ground-level air pollution.

Spatial distribution of column-derived estimated CO, NO 2, and SO 2 values

The spatial distribution of column-derived estimated CO, NO 2, and SO 2 values was analyzed in the Apurimac region during the 2020–2023 period. The results showed heterogeneous spatial patterns among the three pollutants, with relatively higher values concentrated in specific sectors rather than uniformly distributed across the study area. These spatial differences were mainly observed in areas influenced by urban expansion, transportation routes, agricultural burning, and mining-related activities, as shown in Figures 3–5 However, the mapped values should be interpreted as satellite-derived indicators rather than direct ground-level air-quality measurements.

For CO, the annual maps showed relatively higher classes in urban-influenced areas, especially in the province of Abancay (Figure 3). According to the updated classification scale, high and very high CO classes ranged from 420 to 600 μg/m 3. Similar relatively high classes were also observed in Chincheros and Andahuaylas. This spatial pattern may be associated with combustion-related sources, including vehicular activity, domestic combustion, and seasonal biomass burning. CO has been widely used as a satellite-detectable indicator of incomplete combustion and regional pollution transport, as reported by Borsdorff et al. (2018) and Shami et al. (2022). Nevertheless, direct source attribution in the Apurimac region requires complementary ground-based measurements, fire records, traffic data, meteorological information, and emission inventories.

For NO 2, the highest relative classes were mainly observed in urban areas and zones with greater anthropogenic influence, including areas near mining-related activity such as Cotabambas province (Figure 4). According to the updated classification scale, higher NO 2 classes ranged from 1.14 to 1.50 μg/m 3. NO 2 is commonly associated with vehicular transport, fossil fuel combustion, and industrial activities, which may explain its relatively higher values in areas with stronger human activity (Li et al., 2017; Jamali et al., 2020). Previous studies have shown the usefulness of TROPOMI NO 2 observations for identifying urban, industrial, and background pollution patterns, while emphasizing the need for validation with surface observations (Ialongo et al., 2020; Bassani et al., 2021; Fioletov et al., 2025). Therefore, the NO 2 patterns identified in this study should be interpreted as relative satellite-derived indicators rather than direct surface-level NO 2 concentrations.

For SO 2, the annual maps showed a more localized and irregular spatial pattern compared with CO and NO 2 (Figure 5). According to the updated classification scale, higher SO 2 classes ranged from 18.00 to 30.00 μg/m 3. SO 2 is generally associated with fossil fuel combustion, industrial activities, and, in some regions, natural emissions such as volcanic sources (Jion et al., 2023; Opio et al., 2021; Wang et al., 2022). TROPOMI has shown potential for detecting SO 2 emissions, particularly where strong emission sources are present (Theys et al., 2019). However, in the Apurimac region, the SO 2 results should be interpreted with particular caution because weak signals, negative retrievals, and satellite retrieval noise may affect the robustness of the spatial pattern. Therefore, mapped SO 2 values were treated as exploratory satellite-derived indicators rather than robust evidence of confirmed province-level SO 2 pollution hotspots.

Overall, the spatial analysis indicated that CO, NO 2, and SO 2 presented different relative distribution patterns across Apurimac during the 2020–2023 period. CO showed relatively higher values mainly in Abancay, Chincheros, and Andahuaylas; NO 2 showed higher relative values in Cotabambas, Grau, Abancay, and Chincheros; and SO 2 showed localized but more uncertain patterns mainly in Cotabambas, Antabamba, Aymaraes, and Grau. These results provide a regional screening of satellite-derived pollutant patterns and support the identification of potential priority areas for future monitoring, rather than direct evidence of ground-level air-quality exceedances.

Relative hotspot screening by province

The spatial classification of column-derived estimated CO, NO 2, and SO 2 values allowed the screening of provinces with relatively higher satellite-derived values during the 2020–2023 period. The analysis showed that high and very high classes were not uniformly distributed across the Apurimac region, but appeared in specific provinces depending on the pollutant analyzed. These areas were interpreted as potential priority zones for future monitoring rather than confirmed ground-level pollution hotspots. The provinces with relatively higher column-derived estimated values are summarized in Table 4.

For CO, relatively higher classes were mainly observed in Abancay, Chincheros, and Andahuaylas. During the study period, these provinces showed more noticeable CO values compared with other areas of the region. This pattern may be related to urban activity, vehicular emissions, domestic combustion, and seasonal biomass burning, which are commonly associated with CO as an indicator of incomplete combustion (Li et al., 2017; Borsdorff et al., 2018; Shami et al., 2022). However, direct source attribution was not performed because the study did not include ground-based measurements, traffic records, fire occurrence data, meteorological modelling, or detailed emission inventories.

For NO 2, relatively higher classes were mainly observed in Cotabambas, Grau, Abancay, and Chincheros. These areas may reflect the influence of anthropogenic activities such as vehicular transport, fossil fuel combustion, urban activity, and mining-related operations. Previous studies have shown that TROPOMI NO 2 observations are useful for identifying urban, industrial, and background pollution patterns, although validation with surface observations is required for ground-level interpretation (Ialongo et al., 2020; Jamali et al., 2020; Bassani et al., 2021; Fioletov et al., 2025). Therefore, the NO 2 patterns identified in this study were interpreted as relative satellite-derived indicators rather than direct evidence of surface-level NO 2 concentrations.

For SO 2, relatively higher classes were observed mainly in Cotabambas, Antabamba, Aymaraes, and Grau. However, this result was interpreted with greater caution because the SO 2 signal was more irregular and may be affected by weak retrieval sensitivity, negative values, and satellite retrieval noise in areas without strong SO 2 emission sources. Although SO 2 is generally associated with fossil fuel combustion, industrial activity, and natural sources in some regions (Jion et al., 2023; Opio et al., 2021; Wang et al., 2022; Theys et al., 2019), the present results should not be interpreted as confirmed province-level SO 2 pollution hotspots. Instead, they should be understood as exploratory satellite-derived indicators requiring validation with independent datasets or ground-based measurements.

Taken together, the relative hotspot screening showed different spatial patterns among pollutants. CO presented relatively higher values in Abancay, Chincheros, and Andahuaylas; NO 2 showed relatively higher values in Cotabambas, Grau, Abancay, and Chincheros; and SO 2 showed localized but more uncertain patterns in Cotabambas, Antabamba, Aymaraes, and Grau. These provinces may be considered potential priority areas for future air-quality monitoring, but the results should be interpreted as relative satellite-derived patterns rather than direct evidence of ground-level air-quality exceedances.

Temporal variability of column-derived estimated CO, NO 2, and SO 2 values

Temporal variability was evaluated using column-derived estimated values of CO, NO 2, and SO 2 for the 2020–2023 period, as shown in Figures 6–9. The results showed that the three pollutants presented temporal fluctuations and isolated increases rather than uniform behavior throughout the study period. These variations should be interpreted as relative satellite-derived temporal changes and not as direct evidence of ground-level concentration changes.

For CO, the time series showed recurrent increases mainly during dry-season months, particularly between August and October, with additional increases observed in some months of 2023. The highest CO value was observed in 2020, reaching approximately 743.70 μg/m 3 according to the recalculated values using H = 2,000 m. This temporal behavior may be associated with seasonal combustion-related activities, including biomass burning, vehicular emissions, and domestic combustion. Similar temporal approaches using Sentinel-5P/TROPOMI data have been applied to evaluate CO variability and changes related to human activity and combustion sources (Borsdorff et al., 2018; Shami et al., 2022). However, direct source attribution was not performed because fire records, traffic data, meteorological variables, and emission inventories were not included in the analysis.

For NO 2, the time series showed noticeable temporal variability, with relatively higher values during some dry-season months and a more evident increase during 2022. The maximum NO 2 value reached approximately 1.85 μg/m 3 in 2022. NO 2 is commonly associated with vehicular transport, fossil fuel combustion, and industrial or mining-related activity (Li et al., 2017; Jamali et al., 2020). Previous studies have shown that TROPOMI NO 2 observations are useful for identifying temporal and spatial pollution patterns, although interpretation as surface-level NO 2 requires validation with ground-based observations or modelling approaches (Ialongo et al., 2020; Bassani et al., 2021; Fioletov et al., 2025). Therefore, the NO 2 temporal pattern observed in this study was interpreted as a relative satellite-derived signal rather than as a validated ground-level concentration trend.

For SO 2, the temporal behavior was more irregular than that observed for CO and NO 2. The highest recalculated SO 2 value reached approximately 36.00 μg/m 3 in 2020; however, the SO 2 series also showed negative retrieval values and high variability. This behavior suggests that SO 2 retrievals in the Apurimac region may be strongly influenced by retrieval uncertainty, especially in the absence of strong SO 2 point sources. Although SO 2 can be associated with fossil fuel combustion, industrial activities, mining-related processes, and natural sources in some regions (Jion et al., 2023; Opio et al., 2021; Wang et al., 2022; Theys et al., 2019), the SO 2 results in this study should be interpreted as exploratory satellite-derived indicators rather than robust evidence of persistent SO 2 pollution events.

To complement the temporal analysis, annual descriptive statistics were calculated for each pollutant using the processed satellite-derived time-series data. Table 5 summarizes the annual mean, standard deviation, and maximum values of CO, NO 2, and SO 2 during the 2020–2023 period.

The annual statistics show that CO presented its highest annual mean in 2020, while NO 2 reached its highest annual mean in 2022. SO 2 showed relatively similar annual mean values across the study period, although isolated maximum values were observed in 2020 and 2022. The standard deviation of SO 2 was equal to or greater than its annual mean in several years, indicating high variability and supporting the cautious interpretation of SO 2 as a noise-sensitive and exploratory satellite-derived indicator.

To summarize the main temporal patterns observed in the three time series, Table 6 presents the main periods with higher values, maximum observed values, lower values, and general temporal behavior for CO, NO 2, and SO 2 during the 2020–2023 period.

Taken together, the temporal analysis showed that CO and NO 2 presented clearer seasonal fluctuations, particularly during dry-season months, whereas SO 2 showed a more irregular and uncertainty-sensitive temporal pattern, with isolated positive peaks and negative retrieval values. Therefore, the observed variations should be interpreted as relative temporal changes derived from Sentinel-5P/TROPOMI column observations, rather than as direct ground-level concentration measurements or validated air-quality trends.

Mann-Kendall and Sen’s slope trend analysis

The temporal trend analysis was performed using the Mann-Kendall test and Sen’s slope estimator for the column-derived estimated CO, NO 2, and SO 2 values during the 2020–2023 period. This approach was used to evaluate the direction, magnitude, and statistical significance of monotonic trends in the satellite-derived pollutant time series.

For CO, the Mann-Kendall test indicated a statistically significant decreasing trend during the study period. The Sen’s slope was negative, with an estimated value of −0.02540588 μg/m 3 /day, equivalent to an annual slope of approximately −9.2795 μg/m 3 /year. The Mann-Kendall tau value was −0.1457, with p = 1.943 X 10 –16, indicating that the decreasing tendency was statistically significant. This result suggests a gradual reduction in the column-derived estimated CO values over time, although the series still presented seasonal fluctuations and isolated peaks.

For NO 2, the Mann-Kendall test showed a weak but statistically significant increasing trend. The Sen’s slope was 0.00002186 μg/m 3 /day, equivalent to an annual slope of approximately 0.00798 μg/m 3 /year. The Mann-Kendall tau value was 0.049, with p = 0.005676. Although the magnitude of the increase was small, the result indicates a statistically significant upward tendency in the satellite-derived NO 2 values during the 2020–2023 period.

For SO 2, the Mann-Kendall test did not detect a statistically significant trend. Although the Sen’s slope was positive, with an estimated value of 0.00025948 μg/m 3 /day and an annual slope of approximately 0.09478 μg/m 3 /year, the p-value was 0.0669, which is greater than the significance threshold of 0.05. Therefore, the apparent increase in SO 2 should not be interpreted as a statistically significant trend. In addition, the SO 2 series showed irregular variability, isolated peaks, negative retrieval values, and retrieval uncertainty; therefore, its temporal behavior should be interpreted with particular caution.

Taken together, the Mann-Kendall and Sen’s slope analysis showed different temporal tendencies among pollutants. CO presented a statistically significant decreasing trend, NO 2 showed a weak but statistically significant increasing trend, and SO 2 did not show a statistically significant trend during the study period. These results should be interpreted as temporal tendencies in column-derived satellite indicators rather than validated ground-level air-quality trends.

Spatial behavior of column-derived estimated CO, NO 2, and SO 2 values

This study analyzed the spatial and temporal behavior of column-derived estimated carbon monoxide (CO), nitrogen dioxide (NO 2), and sulfur dioxide (SO 2) values in the Apurimac region during the 2020–2023 period using Sentinel-5P/TROPOMI data processed in Google Earth Engine. The results showed heterogeneous spatial patterns, with relatively higher satellite-derived values concentrated in specific provinces rather than uniformly distributed across the study area. This spatial variability suggests that pollutant behavior in Apurimac may be influenced by the interaction of urban expansion, transportation routes, agricultural burning, mining-related activities, seasonal conditions, and complex Andean topography. However, the mapped values should be interpreted as relative satellite-derived indicators rather than direct ground-level air-quality measurements.

For CO, relatively higher classes were mainly observed in Abancay, Chincheros, and Andahuaylas. These provinces include urban-influenced areas and zones where combustion-related activities, vehicular circulation, domestic combustion, and seasonal biomass burning may contribute to increased CO values. Borsdorff et al. (2018) demonstrated the usefulness of TROPOMI observations for mapping CO pollution from space at different spatial scales, supporting the application of satellite data for identifying regional CO patterns. Similarly, Shami et al. (2022) reported temporal changes in CO and NO 2 using Sentinel-5P data, showing that satellite observations can capture variations associated with changes in anthropogenic activity and emission intensity. Nevertheless, direct attribution of the CO spatial pattern to specific sources should be interpreted cautiously because this study did not include ground-based measurements, fire records, traffic data, meteorological variables, or detailed emission inventories.

For NO 2, relatively higher classes were mainly observed in Cotabambas, Grau, Abancay, and Chincheros. NO 2 is commonly associated with vehicular transport, fossil fuel combustion, and industrial activities; therefore, the observed spatial pattern may reflect the influence of localized anthropogenic sources, including transportation corridors, urban areas, and mining-related zones. Ialongo et al. (2020) compared TROPOMI/Sentinel-5P NO 2 observations with ground-based measurements and highlighted the value of satellite products for air-quality assessment, while also emphasizing the importance of validation with surface observations. Bassani et al. (2021) reported NO 2 variability using ground-based, satellite remote sensing, and in situ measurements, and Fioletov et al. (2025) showed that TROPOMI observations are useful for analyzing urban, industrial, and background NO 2 patterns at broad spatial scales. Therefore, the NO 2 patterns identified in Apurimac should be interpreted as relative satellite-derived spatial indicators rather than direct ground-level NO 2 concentrations.

For SO 2, the spatial distribution was more localized and irregular compared with CO and NO 2. Relatively higher SO 2 classes were mainly observed in Cotabambas, Antabamba, Aymaraes, and Grau. SO 2 is generally related to fossil fuel combustion, industrial activities, and, in some regions, natural emissions such as volcanic sources (Jion et al., 2023; Opio et al., 2021; Wang et al., 2022). Theys et al. (2019) demonstrated the potential of TROPOMI for monitoring SO 2 emissions, particularly where strong emission sources are present, while Wang et al. (2022) evaluated TROPOMI NO 2 and SO 2 products and reported their usefulness for detecting spatial and seasonal pollutant behavior. However, in the Apurimac region, the SO 2 pattern should be interpreted with particular caution because weak signals, negative retrieval values, and retrieval noise may affect the robustness of the spatial field. Therefore, the SO 2 results should be understood as exploratory satellite-derived indicators rather than confirmed province-level SO 2 pollution hotspots.

Temporal variability and Mann-Kendall trend results

The temporal analysis showed that CO and NO 2 presented clearer seasonal fluctuations, particularly during dry-season months, whereas SO 2 showed a more irregular and uncertainty-sensitive temporal pattern with isolated positive peaks and negative retrieval values. Higher CO values were observed mainly between August and October, with additional increases in 2023. This behavior may be associated with seasonal combustion-related activities, including biomass burning, vehicular emissions, and domestic combustion. Similar temporal approaches using Sentinel-5P/TROPOMI data have been applied to evaluate CO variability and changes related to human activity and combustion sources (Borsdorff et al., 2018; Shami et al., 2022). However, in the present study, these associations should be considered plausible explanations rather than direct causal evidence because fire records, traffic data, meteorological variables, and emission inventories were not included.

For NO 2, the time series showed noticeable temporal variability, with relatively higher values during some dry-season months and a more evident increase during 2022. This result is consistent with the annual statistics, where NO 2 reached its highest annual mean in 2022. Since NO 2 is commonly associated with vehicular transport, fossil fuel combustion, and industrial or mining-related activity (Li et al., 2017; Jamali et al., 2020), the temporal behavior observed in Apurimac may reflect changes in anthropogenic activity or atmospheric conditions. Similar applications of TROPOMI/Sentinel-5P NO2 observations have been reported by Jeong and Hong (2021). Nevertheless, interpretation as surface-level NO 2 requires validation with ground-based observations or modelling approaches, as emphasized in previous TROPOMI-based studies (Ialongo et al., 2020; Bassani et al., 2021; Fioletov et al., 2025).

For SO 2, the temporal behavior was more irregular than that observed for CO and NO 2. Although isolated positive peaks were detected, the series also showed negative retrieval values and high variability. The annual statistics showed that SO 2 mean values were relatively similar across the study period, while the standard deviation was equal to or greater than the annual mean in several years. This supports the interpretation of SO 2 as a noise-sensitive and exploratory satellite-derived indicator in the Apurimac region. Therefore, the SO 2 temporal pattern should not be interpreted as evidence of persistent SO 2 pollution events, but as an uncertain signal requiring validation with independent datasets or ground-based measurements.

The Mann-Kendall test and Sen’s slope estimator provided a more appropriate assessment of temporal tendencies than ordinary least-squares regression for the pollutant time series. The results showed a statistically significant decreasing trend for CO, a weak but statistically significant increasing trend for NO 2, and no statistically significant trend for SO 2. This indicates that the temporal behavior differed among pollutants and should not be summarized as a uniform regional trend. CO showed a decreasing tendency despite seasonal peaks, NO 2 showed a small upward tendency, and SO 2 remained dominated by irregular variability, isolated peaks, negative retrieval values, and retrieval uncertainty.

These trend results should be interpreted as temporal tendencies in column-derived satellite indicators rather than validated ground-level air-quality trends. Although the Mann-Kendall test and Sen’s slope estimator are useful for detecting monotonic tendencies in environmental time series (Mann, 1945; Kendall, 1970; Sen, 1968), serial correlation and seasonal variability may influence trend interpretation (Yue and Wang, 2004). Therefore, the statistical results should be considered descriptive of satellite-derived temporal behavior and not as direct evidence of causal changes in surface-level air pollution.

Usefulness and limitations of sentinel-5p/TROPOMI and google earth engine

The use of Sentinel-5P/TROPOMI data processed in Google Earth Engine was useful for identifying relative spatial patterns and temporal variability of CO, NO 2, and SO 2 in the Apurimac region. Similar applications of TROPOMI and Google Earth Engine have shown their usefulness for evaluating spatiotemporal air-quality patterns and their relationship with urbanization and environmental conditions (Dhankar et al., 2024). Google Earth Engine has been widely recognized as a cloud-computing platform for planetary-scale geospatial analysis and environmental monitoring (Gorelick et al., 2017). In addition, Veefkind et al. (2012) described TROPOMI on Sentinel-5P as an instrument designed for global observations of atmospheric composition related to air quality, climate, and ozone-layer applications. This supports the use of Sentinel-5P/TROPOMI data for regional air-pollution screening, especially in areas where continuous ground-based monitoring is limited.

The results of this study are consistent with previous research showing that TROPOMI observations can support the detection and analysis of atmospheric pollutants at different spatial scales. Borsdorff et al. (2018) demonstrated the use of TROPOMI for CO mapping; Ialongo et al. (2020) compared TROPOMI NO 2 observations with ground-based measurements; Theys et al. (2019) showed the potential of TROPOMI for SO 2 monitoring; and Wang et al. (2022) evaluated the performance of TROPOMI NO 2 and SO 2 products. These studies support the usefulness of Sentinel-5P/TROPOMI for atmospheric monitoring, while also showing that interpretation improves when satellite observations are compared with ground-based or independent datasets.

Despite these advantages, the results should be interpreted with caution. Sentinel-5P/TROPOMI provides atmospheric column-density observations rather than direct ground-level measurements. Therefore, the converted values expressed in µg/m 3 should be considered column-derived estimated indicators and not equivalent to in situ air-quality measurements. This limitation is especially important in mountainous regions such as Apurimac, where altitude, slope, valley systems, local meteorological conditions, and vertical pollutant distribution can influence the relationship between satellite column values and near-surface concentrations.

Another relevant limitation is the uncertainty associated with the conversion from column density to estimated values. The conversion assumes a simplified effective lower-atmospheric layer height of 2,000 m, which was used as a methodological approximation and not as a measured planetary boundary layer height. In addition, cloud cover, surface reflectance, atmospheric conditions, and product-specific retrieval uncertainty may affect Sentinel-5P observations, especially for NO 2 and SO 2 products (Wang et al., 2022; Ialongo et al., 2020). For this reason, the present results should be understood as a regional screening of relative satellite-derived pollutant patterns rather than definitive evidence of ground-level air-quality exceedances.

Taken together, Sentinel-5P/TROPOMI and Google Earth Engine provide a useful approach for evaluating regional pollutant behavior in areas with limited monitoring infrastructure. The findings identify provinces with relatively higher column-derived estimated values and periods with greater temporal variability. Future studies should integrate Sentinel-5P/TROPOMI observations with ground-based air-quality measurements, meteorological data, boundary-layer information, fire occurrence records, traffic information, mining activity records, emission inventories, and independent datasets such as atmospheric reanalysis or cross-sensor satellite products. This integration would improve validation, source attribution, and the interpretation of atmospheric pollutant patterns in the Apurimac region.

This study analyzed the spatial distribution, temporal variability, and trend behavior of column-derived estimated CO, NO 2, and SO 2 values in the Apurimac region, Peru, during 2020–2023 using Sentinel-5P/TROPOMI data processed in Google Earth Engine. The results showed heterogeneous spatial patterns, with relatively higher satellite-derived values concentrated in specific provinces. CO showed higher relative values mainly in Abancay, Chincheros, and Andahuaylas, while NO 2 presented higher relative values in Cotabambas, Grau, Abancay, and Chincheros. SO 2 showed localized but more uncertain patterns, mainly in Cotabambas, Antabamba, Aymaraes, and Grau.

Temporal analysis showed that CO and NO 2 presented clearer seasonal fluctuations, especially during dry-season months, whereas SO 2 showed a more irregular pattern with isolated positive peaks, negative retrieval values, and high variability. Therefore, SO 2 results should be interpreted as exploratory satellite-derived indicators rather than robust evidence of persistent SO 2 pollution or confirmed province-level hotspots.

The Mann-Kendall test and Sen’s slope estimator showed a statistically significant decreasing trend for CO, a weak but statistically significant increasing trend for NO 2, and no statistically significant trend for SO 2 during the study period. These trends should be interpreted as temporal tendencies in column-derived satellite indicators rather than validated ground-level air-quality trends.

Sentinel-5P/TROPOMI and Google Earth Engine proved useful for identifying relative pollutant patterns in an Andean region with limited ground-based monitoring. However, the column-derived estimated values expressed in µg/m 3 should not be considered direct surface concentrations. Future studies should integrate satellite observations with in situ measurements, meteorological data, boundary-layer information, emission inventories, and independent datasets to improve validation and source attribution.