Introduction

Atmospheric particulate matter (PM) is a complex mixture of solid and liquid particles suspended in the air (Seinfeld and Pandis 2006). In recent years, a growing body of research has established strong links between PM2.5, particles with a diameter of less than 2.5 micrometers, and a range of serious health conditions, including asthma, myocardial infarction, stroke, and even Alzheimer’s disease (WHO 2021). Epidemiological studies based predict a potential surge in global disease burden and mortality if current trends in rising PM emissions persist. While many of these studies use PM mass as a primary metric, some more advanced models incorporate key chemical species found in PM, such as metals and organic compounds (Yadav and Phuleria 2020). Nevertheless, investigations for linking PM toxicity with chemical composition have often shown that the toxicity of a mixed composition differs significantly from that of individual component or pollutant (Jia et al. 2017). Moreover, traditional toxicity characterization methods are labor-intensive and time-consuming, and therefore, only a limited number of samples can be analyzed for toxicity evaluation. This underscores the critical need to seek alternative metrics beyond PM mass to accurately represent PM toxicity in epidemiological models. Recently, oxidative potential (OP) has emerged as a proxy for PM toxicity, since inflammation and oxidative stress are the main biological pathways by which ambient air pollutants produce adverse effects on human health (Yu et al. 2021). OP has gained considerable attention in recent years (Fleck et al. 2022; Gao et al. 2020a; Robinson 2017; Weichenthal et al. 2019). OP describes the ability of PM mixture to produce oxidative stress and is affected by various factors, such as its chemical composition, size, and sources.

PM2.5 can carry a range of chemical species that can oxidize target molecules generating reactive oxygen species (ROS) (Visentin et al. 2016), which can cause cellular damage and promote oxidative stress in the body. The OP of PM2.5 is the capacity to induce oxidative stress in molecules in biological tissues or cells directly by the reactive oxygen species (ROS) of PM and indirectly by the ROS generated via PM interactions with the biological system (He and Zhang 2023; Mudway et al. 2020). Measuring the OP of PM2.5 provides relevant information of oxidative stress, which is the main mechanism of PM toxicity, that complements traditional measures of bulk PM mass.

There are different methods to measure the OP of PM2.5 with different sensitivity to the ROS generating compounds. These methods include consumption of dithiothreitol (DTT), consumption of ascorbate and glutathione using synthetic respiratory tract lining fluid (RTLF) assays, electron spin resonance (ESR) assays, and reductive acridinium triggering (CRAT) assays. The OP results are method-dependent, owing to their different sensitivities to particle components (Bates et al. 2019; Yang et al. 2014). Common PM oxidative constituents include redox active metals (i.e. Fe, Cu), organic compounds (water-soluble organic carbon – WSOC, photochemically-aged organics), peroxides, and humic-like substances (Gao et al. 2020b). OPAA has been responsive to soluble Cu, total Cu, soluble Fe, total Fe, total Pb, total Zn, soluble Mn and Organic Carbon, whereas OPGSH has shown evidence of total Al, total Pb and total Fe impact (Bates et al. 2019). Currently, there is no standardized protocol for OP measurement, and OP measurement techniques continue to evolve, as there is a desire to have a high-throughput method to screen PM with a toxicologically relevant metric. Some authors report measurements of OPDTT of PM2.5 only (Jedynska et al. 2017; Patel et al. 2018; Pietrogrande et al. 2019; Puthussery et al. 2020; Simmonetti et al. 2018; Styszko et al. 2017; Velali et al. 2016; Wang et al. 2019; Ma et al. 2021; Yang et al. 2014; Yang et al. 2021), while others (Gao et al. 2020a; Godri et al. 2011; Maikawa et al. 2016; Perrone et al. 2019; Pietrogrande et al. 2019; Ripley et al. 2022; Weichenthal et al. 2016; Weichenthal et al. 2019; Yu et al. 2021) have reported OPAA and/or OPGSH of PM2.5, some of them together with OPDTT. Gao et al. (2020a) reported Pearson’s r values of 0.45–0.7 between OPDTT and OPGSH, 0.53–0.66 between OPDTT and OPAA, and 0.62–0.78 between OPGSH and OPAA. These moderate-to-good correlations suggest that the tests are complementary.

Comprehending the OP of PM2.5 is of paramount importance in elucidating the underlying mechanisms through which PM2.5 exerts its impact on human health. This knowledge is pivotal in formulating efficacious strategies to mitigate its deleterious effects, particularly in the context of Latin American nations. Predominantly, extant studies on this subject have been conducted within the regions of Canada (Ripley et al. 2022; Weichenthal et al. 2019), the United States (Yu et al. 2021), and Europe (Daellenbach et al. 2020; Jedynska et al. 2017; Pietrogrande et al. 2019; 2022; Simoneti et al. 2018), with a limited representation of one or two cities in Brazil (Serafeim et al. 2023) and Chile (Molina et al. 2023); respectively. Despite the highest average annual concentrations of PM2.5 being concentrated in low-middle income countries (HEI 2020) -especially Latin America-, a notable lacuna persists in research endeavors aimed at characterizing the OP of PM2.5 and discerning its spatial fluctuations within these contexts. Thus, this research aimed to analyze between-city, inter-city and site-type variations of OP of PM2.5 across the five largest urban centers in Colombia, South America.

Materials and methods

Study areas

The study was conducted in the urban areas of the five largest cities in Colombia: Barranquilla, Bucaramanga, Bogotá, Cali, and Medellín (Fig. 1). The population varies across these cities, with Bogotá being the most populated city, estimated to have a population of 7,823,334 inhabitants in 2021. The estimated total population in the same year was 1,294,390 for Barranquilla, 604,462 for Bucaramanga 2,271,293 for Cali, and 2,549,008 inhabitants for Medellín (DANE 2023). The altitude and average temperature also vary across cities, with Barranquilla being the warmer (Tmin = 24ºC; Tmax = 32ºC) and closest to sea level (25 masl) and Bogotá being the coldest (Tmin = 7ºC; Tmax = 19ºC) and highest (2600 masl) city.

Fig. 1
figure 1

Location of the study cities, and number of sites per site type in each city. Map generated using ArcGIS®

Full size image

PM2.5 sample collection

In 2021, comprehensive sampling campaigns were executed, covering a total of 120 sites spread across five cities, encompassing both dry and rainy seasons. Each city underwent meticulous sampling at 20 locations, with an exception in Bogota, where a more extensive 40-location sampling was conducted. The sampling schedule for each city was as follows: Barranquilla (April 19–21 and May 3–5; September 10–24, 18 sites), Bogota (March 17–29 and April 5–14; July 14–30 20 sites), Bucaramanga (February 24 – March 15; June 15–30, 17 sites), Cali (August 5–20; September 30- October 15, 18 sites) and Medellin (April 28–29 and May 12–14; August 20- September 3, 18 sites). For more information, please see Table S1.

We selected and categorized sampling sites according to the characteristics of the locations of each city, as background, residential, industrial, and traffic sites. This was performed by a panel of experts that included researchers and professionals at the environmental and health agencies of each city, considering qualitative criteria related with land use, land cover, proximity to primary roads, traffic and building characteristics. We included one background site and some sites co-located with official air quality monitoring stations, according to their local availability (Rodriguez-Villamizar et al. 2023). This selection, however, was limited by the fact that land use in Colombian cities is moderately to heavily mixed.

At each site, Ultrasonic Personal Air Samplers (UPAS, Access Sensor Technologies) (Volckens et al. 2017) were employed for the collection of 1-week integrated samples on 37-mm Teflon membrane disc filters with a 2 μm pore size (Pall Corporation) at a flow rate of 1 L/min. This process was repeated over two consecutive weeks, resulting in two filters per sampling period.

Quality control and limitations

Mass measurement was conducted three times, and the average measurement was reported for each filter. Prior to the oxidative potential analysis, we determined PM2.5 mass concentrations gravimetrically for all city samples in a single laboratory certified by IDEAM, complying with the general requirements for the competence of testing and calibration laboratories (ISO/IEC 17025:1999), utilizing a Sartorius Microbalance CPA26P. The samples were conditioned for 24-h under constant temperature (20–22 °C) and relative humidity (30–40%). For quality control, two blank filters were used for each city. The reported limit of detection was 1.36 µg/m3. The average PM2.5 mass concentration of the two-weekly filters from the same site and campaign was reported as the site concentration.

PM2.5 values obtained through UPAS samplers were contrasted with data acquired from the air quality station. During the dry season campaign, 11 monitoring stations were used to measure the differences in PM2.5 concentrations between the UPAS and local monitoring stations. The resulting average was − 1.5 µg/m3, with a 95% confidence interval (CI) of -6.8 to 3.9. The rainy season campaign, which utilized 13 monitoring stations, showed an average difference of -0.05 µg/m3, with a 95% CI of -11.5 to 11.4. These results indicate that the differences in PM2.5 concentrations between the UPAS and local monitoring stations were small during both seasons, with the largest difference occurring during the dry season campaign. This is also shown by Rodriguez-Villamizar et al., 2024 (supplementary material).

The sampling campaigns comprehensively addressed the climatic variations inherent to Colombia by encompassing both the rainy and dry seasons (Table S1). The two-week sampling period in the rainy and dry season covered the variability, as stated in several studies done in the world (Godri et al. 2011; Aileen et al. 2015). The sampling regimen incorporated a spectrum of urban locales (n = 120), comprising traffic, residential, industrial, and background areas within each city. This methodological breadth was pivotal in encompassing a diverse array of air pollution sources, thus facilitating a comprehensive and representative elucidation of air quality dynamics within these urban centers.

Oxidative potential analysis

Oxidative potential analysis was conducted on filter samples with a minimum mass loading of 40 µg on a single filter or on a combination of the two filters corresponding to the same site and measurement campaign. A total of 91 out of the 120 samples had sufficient mass for analysis: 17 from Bucaramanga, 18 from each of Barranquilla, Cali, and Medellín, and 20 from Bogotá. Oxidative potential was evaluated using the RTLF assay to determine ascorbate (AA) and glutathione (GSH) depletion. PM2.5 was extracted from filters using HPLC grade methanol through vortexing and sonication. Methanol was evaporated to dryness under a gentle flow of nitrogen at 37ºC. Samples were reconstituted to a unique mass concentration (75 µg/mL) with 5% HPLC grade methanol in ultrapure water solution. Samples were incubated in triplicate with the synthetic RTLF at 37ºC for 4 h in an ultraviolet-visible plate reader (Molecular Devices SpectraMax 190). The synthetic RTLF contained equimolar concentrations of AA, GSH, and urate (200 μm). Positive controls, consisting of a 1.0 µM Cu(NO3)2 solution, and a 0.02% H2O2 solution, and blanks were run together with each batch of samples. We calculated AA percent depletion based on the initial and final absorbance values at 260 nm, after subtracting blanks. To determine the percent depletion of GSH, we used the glutathione-reductase enzyme assay, and compared the GSH concentrations in samples and positive controls with those in blanks. Finally, we converted OPAA and OPGSH into four metrics: two for PM2.5-mass normalized oxidative potential (OPm, expressed as % depletion/µg and pmol/min µg), and two for depletion burden, OP (% depletion/m3 and pmol/min m3), as described by Maikawa et al. (2016) and Weichental et al. (2019).

Results and discussion

This study collected samples from various locations in the urban centers of five major cities in Colombia: Bucaramanga, Medellin, Barranquilla, Cali, and Bogota. The sampling sites included traffic, residential, industrial, and background areas in each city. This comprehensive approach ensured that the study captured a diverse range of air pollution sources and provided a representative understanding of air quality in these urban centers. Table 1 presents the descriptive statistics for PM2.5 (µg/m3), and oxidative potential burden (OPAAv and OPGSHv, expressed in % depletion/m3) in each city and in the overall data set, and Table S2 present the corresponding descriptive statistics for mass-specific oxidative potential (OPAAm and OPGSHm, expressed in % depletion/µg). On the other hand, Table S3 shows the Shapiro-Wilk normality test results: OPAAm, OPGSHm, and OPAAv data were normally distributed, while PM2.5 and OPGSHv data were not.

Table 1 Descriptive statistics for PM2.5 (µg/m3), OPAAv and OPGSHv (% depletion/m3) in each city and in the overall data set
Full size table

Between-city variation

Table S4 shows the Kruskal-Wallis test results for between-city differences. Differences were significant for PM2.5, OPAAm, OPGSHm, and OPAAv, but not for OPGSHv. In the examination of the five cities under consideration (Table 1), Cali emerged with the highest concentration of PM2.5 and OPAAv. Conversely, Medellin recorded the highest values for OPGSHv. In contrast, Barranquilla exhibited the lowest values for both OPAAm and OPGSm, and Bogotá displayed the lowest PM2.5 mass concentration. Noteworthy patterns include the similarity in mean PM2.5 mass concentrations between Bogotá and Bucaramanga, as well as the analogous levels observed between Medellin and Barranquilla.

In terms of maximum PM2.5 levels (Table 1), Cali had the highest concentration, albeit with considerable variation. On the other hand, Bucaramanga displayed the lowest maximum value. Regarding OPGSHv, Medellin recorded the highest maximum value, while Barranquilla had the lowest. Similarly, for OPAAm, Cali registered the highest maximum value, while Barranquilla had the lowest.

Table 2 and Figures S1 to S5 provide a comparative analysis of PM2.5 concentrations and oxidative potential (OP) metrics between Colombian cities and various international locations. To make comparisons possible, we converted our results to OPAAv and OPGSHv expressed in pmol/min/m3, by multiplying % depletion/m3 values by 9.3, as indicated by Weichental et al. (2019). The 9.3 factor was based on a 4-h incubation period, 200 µM of AA (or GSH) initially present, and 0.18 mL of PM solution added to the 0.2 mL well. Then, we obtained OPAAm and OPGSHm, expressed in pmol/min/µg, by dividing these values by the corresponding PM2.5 concentration. The comparison was not possible when authors did not report PM2.5 values. It is important to note that the studies included in Table 2 employed methods and techniques that may be different in various aspects, including sampling techniques (personal and ambient), extraction methods (RTLF and other simulated lung fluids), and OP assays (AA, and GSH). These methodological differences may introduce variability in the results and should be considered when interpreting the comparisons presented in this table.

Table 2 Comparative results between our study and around the world
Full size table

In comparison with other cities (Figure S1), average PM2.5 in Colombian cities exhibited elevated values surpassing the PM2.5 measurements in Italy’s Po Valley, as well as those in the United States (Illinois, Indiana, and Missouri) and Canada (Toronto and Ontario). The only sites of developed countries where PM2.5 values were higher than those in Colombian cities, except Cali, was the Salento Peninsula in Italy and Northern France. Moreover, PM2.5 values of developing countries (China and Brazil) were slightly higher than most Colombian Cities, except for Cali were PM2.5 values were higher than Korea (Chuncheon).

Our results show that OPAAv in Colombian cities (Table 2 and Figure S2) was, on average, higher than those observed in Canadian cities: OPAAv was 17% higher than cities across Ontario, Canada (Weichenthal et al. 2016), 14% higher than Toronto in summer, 56% higher than Montreal in summer, and 3 times as high as Montreal in winter (Ripley et al. 2022), 64% higher than Toronto, as reported by Weichenthal et al. (2019). However, it was lower than those in American and European cities: 30% of OPAAv observed in Atlanta (Gao et al. 2020a), Illinois, Indiana, and Missouri (Yu et al. 2021), 0.1–0.3 times OPAAv in Salento’s peninsula (Perrone et al. 2019), 0.05–0.1 times OPAAv in the Po Valley (Pietrogrande et al. 2019) and 0.06 times OPAAv in Northern France. It was also lower than OPAAv in developing cities such as Sao Paulo, Brazil (0.07 times) (Caumo et al., 2022), and Chuncheon, South Korea (0.3 times) (Kim et al., 2024).

On the other hand, OPGSHv (Table 2 and Figure S3) in Colombian cities was much higher than those in Canada: 30% higher than Toronto (Weichenthal et al. 2019), 85% higher than Ontario’s cities (Weichenthal et al. 2016), 5.4 times as high as Toronto in summer, 6.8 times as high as Montreal in summer and 13.5 times as high as Montreal in winter (Ripley et al. 2022). It was 5.2 times as high as OPGSHv in Shenzhen, China (Xing et al., 2023). In contrast, it was just half the OPGSHv observed in Atlanta (Gao et al. 2020a), 10% of OPGSHv in Sao Paulo (Caumo et al., 2022) and 20% of OPGSHv in Chuncheon (Kim et al., 2024).

When considering OPAAm and OPGSHm (Figure S4 and Figure S5), in terms of pmol/min/µg, they were, on average, lower in Colombian cities than in Canadian (Weichenthal et al. 2016; Weichenthal et al. 2019), American (Gao et al. 2020a; Yu et al. 2021), European (Perrone et al. 2019; Pietrogrande et al. 2019), Korean (Kim et al., 2024) and Brazilian (Caumo et al., 2022) sites, ranging from 0.04 to 0.6 times the OPAAm, and from 0.2 to 1.0 times the OPGSHm. Average OPGSHm in Colombian cities were higher than in Shenzhen (Xing et al., 2023), which is the city with the lowest reported OPGSHm. PM2.5 levels in Colombia were 1.3 to 3.5 times higher than those in the cities reported in many of the cited studies (Figure S1), with the exception of Salento’s Peninsula (0.9 times) (Perrone et al. 2019), Northern France (Moufarrej et al., 2020), Sao Paulo (0.6 times) (Caumo et al., 2022), and Shenzhen (0.5 times) (Xing et al., 2023). Therefore, higher OPAAv and OPGSHv in Colombian cities than in Canadian cities are associated with higher PM2.5 levels, rather than with OPAAm and OPGSHm. These results suggest that PM2.5 in Colombian cities have lower concentrations of chemical species that cause oxidation of AA and GSH than in many North American and European sites. Unfortunately, the main limitation of this study is that we could not afford chemical composition analysis of PM2.5, which precludes a more robust explanation for this observation. Also, previous particle composition studies in Colombian cities have been performed only for PM10, not for PM2.5. Therefore, further research on the association between OP and PM2.5 chemical composition in Colombia is needed.

As evident from our observations, there is no discernible relationship between OP values and PM2.5 levels. This reinforces the hypothesis that the OP of PM2.5 in Colombian cities is driven by relatively higher concentrations of PM2.5 rather than high mass-specific oxidative potential. These findings underscore the importance of analyzing this emerging health metric, emphasizing its significance beyond the conventional focus solely on PM2.5. They also suggest that higher levels of less oxidative PM2.5 may account for the greater impact of air pollution on the oxidative burden in Colombian cities than in Canadian ones.

The key factor in reducing OP burden would, therefore, be reducing PM2.5 concentrations, thus mitigating health risks associated with air pollution in Colombia. The first step to do this would be to conduct OP source apportionment studies, which would indicate which are the highest contributing sources to OP that would need to be more closely controlled.

Figure 2 shows the Spearman correlations between PM2.5 mass concentration and OP values per unit PM2.5 mass (OPm, % depletion/µg) across various cities. PM2.5 and OPm were negatively correlated across all sites. The highest negative correlation values were observed between OPAAm and PM2.5 in Medellín (rs = -0.73) as well as Barranquilla (rs = -0.63), while between OPGSHm and PM2.5, the strongest negative correlation was observed in Medellin (rs = -0.54) and Bucaramanga (rs = -0.37). This suggests that in Medellin and Barranquilla, an increase in PM2.5 concentration is more likely to be associated with a decrease in OPm than in the other cities, although the correlation is not significant. The correlation between PM2.5 and OPGSHm was very weak (|rs| < 0.2) in Bogotá and Cali. Thus, in general, our observations suggest that OPm in our samples does not depend on PM2.5 levels but should depend on PM2.5 composition. The higher the PM2.5 concentration, the lower the OPm, indicating that sources that produce higher PM2.5 concentrations appear to have lower concentrations of species with high oxidative potential. Furthermore, the negative association between PM2.5 and OPAAm is higher than with OPGSHm at all sites.

Fig. 2
figure 2

Spearman correlation plot between PM2.5 (µg/m3), OPAAm (% depletion/µg) and OPGSHm (% depletion/µg) in each city and in the overall data set. Negative correlations are depicted in blue, while positive correlations are represented in red

Full size image

On the other hand, a positive correlation was found between OPAAm and OPGSHm, weak (rs < 0.4) in Cali, moderate (0.4 < rs < 0.6) in Barranquilla and Bogotá, and strong (rs > 0.6) in Bucaramanga and Medellín. Ascorbic acid and glutathione are sensitive to trace metals and organic pollutants to different extents. Thus, PM2.5 emissions at some sites may amplify OPAAm rather than OPGSHm without affecting the bulk PM2.5, and vice versa. Consequently, we analyze the within-city spatial OPm variations for all five cities.

Within-city and site-type variation

Table S5 shows the Kruskal-Wallis test results for differences among site types, and Table S6 shows results for the Kruskal-Wallis test results for differences among site types in each city. No difference was significant in any case.

Figure 3 shows the different median PM2.5 and OPv levels for all cities and site-types. In terms of PM2.5 levels, Cali exhibited the highest overall concentration, although with notable variations. Conversely, Medellín had the lowest median concentration. Meanwhile, Barranquilla, Bogotá, and Bucaramanga had similar estimated median levels, but all cities experienced several outliers with PM2.5 concentrations surpassing 30 µg/m3. This situation differs from long-term measurements that indicate that Medellín and Bogotá usually have higher PM2.5 concentrations than the other cities. However, this is associated with the short-term nature of the campaign.

Fig. 3
figure 3

Comparison of PM2.5 (µg/m3), OPAAv (% depletion/m3) and OPGSHv (% depletion/m3) across cities (left column) and within-city site types (right column)

Full size image

When analyzing PM2.5 levels by site type, Bucaramanga had the lowest median concentration for industrial sites, while Barranquilla and Cali had the highest. In the case of residential sites, Bogota had the lowest PM2.5 levels, while Barranquilla and Cali had the highest. When it came to traffic sites, Barranquilla had the lowest median concentration, while the other cities had similar median values. Finally, for background values, we observed similar levels for all cities except for Medellin, which had the lowest concentration.

For OPAA burden (% depletion/m3), Cali had the highest levels, followed by Medellín and Bucaramanga. Barranquilla and Bogotá demonstrated comparable median values, lower than the other three cities. Across site types, we found similar median values for residential and traffic sites in Barranquilla, Bogotá, Bucaramanga, and Medellin, but not in Cali, where OPAA burden levels were higher in residential areas, likely due to their proximity to industrial sources, as we will explain later. Similar OP burden levels in traffic and residential sites were unexpected results, since other studies have found that traffic sites show much higher OP than residential areas. It is possible that the road configuration in Colombian cities makes traffic affect not only the sites located beside the busiest roads but also residential areas with smaller, yet also congested roads. It is also possible that the qualitative classification of the sites was insufficient to obtain clearer differences in OP, given the mixed land use and the influence of both traffic and small and mid-sized industrial facilities across the cities. On the other hand, the lack of PM2.5 composition data precludes a detailed analysis of the effect of chemical species concentrations in the sites on OP.

For industrial sites, Barranquilla, Bogotá, and Medellin had similar OPAA burden values, while Cali had the highest and Bucaramanga had the lowest. As for background sites, Barranquilla had the lowest levels, while Bucaramanga had the highest. Some sites presented background values higher than other impacted traffic and industrial sites.

OPGSH burden, on the other hand, was the highest in Medellín, while the remaining four cities exhibited similar median levels. Similarly, by site type, Medellin had the highest concentration in residential areas, while Bogota had the highest values in industrial sites. For traffic sites, Bogotá, Bucaramanga, and Medellin had similar median values, while Cali had the lowest. Barranquilla, Bogotá, Bucaramanga, and Medellin had similar OPGSH burden values for both residential and traffic sites, suggesting that vehicles are a possible source of emissions in residential areas as well. In the case of Cali, we observed higher OPGSH burden levels in residential areas, possibly due to the proximity of industrial sources and traffic emissions in those areas.

Our findings revealed that background sites exhibited the lowest PM2.5 and OP concentrations. Notably, industrial sites displayed the highest OPAA and OPGSH burden, likely attributed to elevated levels of black carbon (BC), organics, and certain water-soluble metal species characteristic of industrial environments—factors that strongly correlate with OPAA. Conversely, OPGSH exhibited the higher values in industrial sites, too, probably to their exclusive sensitivity to water-soluble Cu and brown carbon ambient-humic like substances observed in some studies. Moreover, our results showed a higher correlation of PM2.5 with OPAA than OPGSH. This partial understanding potentially explains the observed similarity in OPAA values between traffic and residential areas, attributable to the characteristic traffic emissions in these regions. Likely, the comparable values observed in OPAA between residential and traffic sites may be attributable to traffic emissions. Furthermore, the spatial proximity of residential areas to industrial zones, as exemplified in the case of Cali, contributes to these trends for OPGSH, too, for example Medellin, where some residential sites are near industrial areas. This attribution, however, remains speculative and is limited by the fact that we don’t have PM2.5 compositional data.

Nevertheless, these insights advocate for refined land-use planning strategies in Colombian cities aimed at mitigating OP values, and consequently PM toxicity. Additionally, the importance of curtailing both industrial and traffic-related emissions across all urban centers is underscored. Collectively, these findings provide a comprehensive overview of air quality across diverse locations within each city, elucidating the intricate variability of air pollutant concentrations within urban landscapes.

Figure 4 shows the spatial variation in the five cities for both OP burden metrics. For Medellín, we observed high values of PM2.5 and OP burden in the south-west of the city (sites 9, 4, and 17), where some brick kilns and asphalt production companies are located. Some traffic sites (11, 13, 14) and near some hospitals (site 18) showed high OP burden values, too. In the north of the city, the highest OPAA burden value was potentially influenced by industrial emissions from a quarry and an aggregate processing plant located 500 m East from site 11, which, in addition, is surrounded by motorways. There are also a meat processing plant and a cosmetic products plant, 1 km South from the site. For OPGSH burden, the highest values were obtained in the south-west of the city. A recent study conducted in 2020 and 2021 in Medellín found that industrial sources and traffic are major contributors to the city’s poor air quality (AMVA 2021). The study found that these sources emit high levels of various heavy metals and polycyclic aromatic hydrocarbons, which are well recognised to be associated with health outcomes. The study found influence of industrial areas and brick kilns, diesel and road dust, where the receptor model showed that some species dominated such as Zn, Cu, Pb, Sb, As in industrial areas, Sn, Cu, Sb from diesel traffic sources and Fe, Zn, Cu, Sb and Mo from brake pads.

Fig. 4
figure 4

Within-city spatial variation of oxidative potential OPAAv and OPGSHv (% depletion/m3) of PM2.5 in five cities in Colombia

Full size image

For Bogotá, the highest values for PM2.5 were obtained at industrial (20) and traffic (18, 4) sites. Similarly, the highest OPGSH burden values were also obtained in industrial (20) and traffic sites (18, 16). For OPAA burden, the highest values were obtained in a traffic site (16), while some industrial (6 20) and residential sites showed medium levels. The spatiality from our measurements of PM2.5, OPAA or OPGSH burden differs from the pattern observed by the city’s air quality monitoring network, which show consistently higher PM2.5 levels in southern and southwestern areas. This difference suggests that our measurements are more strongly affected by local sources because of the instrument’s exposure level. Such sources also affect PM2.5 composition and, therefore, OPAA and OPGSH burden. Vargas et al. (2012) showed differences in PM2.5 composition and source contribution between the northwest (close to our site 8) and southwest (between our sites 3 and 16). The southwestern site showed higher PM10 as well as higher levels of oxidative species such as organic carbon and metals, which is consistent with higher PM2.5, OPAA, and OPGSH burden levels from our measurements. It also showed that the southwestern site had a stronger contribution of industrial facilities, whereas traffic was the main source affecting the northwestern site. In our measurements, industrial sites did show higher levels of OPAA and OPGSH burden in Bogotá.

For Barranquilla, the highest PM2.5 concentrations were obtained at two residential sites (10, 15) and two industrial sites (11, 12) near the fluvial ports. For OPAA, the highest values were for these two industrial sites (11, 12), while it was the residential site (10) and the traffic site in the city center (3) for OPGSH burden. Probably, the industrial sites and the fluvial ports emit several transition metals (Gómez-Plata et al. 2022), thus enhancing this OP burden. For the residential site and traffic sites, we may observe that the OPGSH burden was higher, even though lower values compared with the other Colombian cities. That is, for low or high PM2.5 concentrations, the OPAA depletion value remains in a range of similar values, with the exception of the industrial sites. Recent studies showed that in this industrial zone, high concentrations of some metals, i.e. Cu and Cr were measured (Gómez-Plata et al. 2022). For OPGSH burden, we may observe high values in several traffic and residential sites, some of them near schools and hospitals.

For Bucaramanga, the highest PM2.5 values were obtained from industrial (15) and traffic sites (7, 12). For the OPAA values, the highest values were obtained in traffic sites (7, 12) and industrial sites (15; 16), including some residential sites (4; 6) near the industrial sites. The OPGSH sites with the highest values were for traffic (7; 12) in the city center, and a residential site (6). In industrial site 15, elevated levels of black carbon (BC) concentrations have been observed (Lloyd et al. 2021), a characteristic marker of diesel combustion. This finding corroborates the previously observed high OP values at the site, which are indicative of the production of reactive oxygen species (ROS).

For Cali, the highest PM2.5 values were observed in some industrial sites (17, 18). Although some traffic (4, 10) and residential (1 2) sites were affected, too, causing high PM concentrations there. We observed high OPAA burden but low OPGSH burden for some industrial sites (16, 18), likely due to the influence of trace elements that affect OPAA but not OPGSH. As a result, we observe high values of OPAA in several traffic and residential sites near these industrial sites.

Although our research did not include a detailed chemical characterization of PM2.5, the sampling sites were strategically selected and categorized with the valuable input of a panel of experts from the environmental and health agencies. These experts helped in the selection of the sites according to the characteristics of the locations of each city, as background, residential, industrial, and traffic sites. These experts provided valuable insights into the land use, land cover, proximity to primary roads, traffic patterns, and building characteristics, as well as the location and nature of potential emission sources in each city. This collaborative approach allowed for a comprehensive assessment of the relationship between PM2.5 and oxidative potential (OP) in diverse urban environments across Colombia.

Overall, our findings indicate that the OP burden in Colombian cities is higher than in Canadian cities and lower than in American and European cities. This burden is primarily influenced by PM2.5 mass concentrations rather than by the OP per unit mass of PM2.5. Therefore, reducing PM2.5 concentrations through emissions compliance enforcement will also reduce OP burden in Colombian cities. This suggests that reducing PM2.5 through stricter emissions controls and enforcement could effectively decrease the OP burden in Colombian cities, thereby mitigating the associated health risks.

Conclusions

We conducted a spatial monitoring study to identify differences between and within-city spatial variations of PM2.5 mass concentration and its associated OPv and OPm in the five largest cities in Colombia. Our findings showed moderate values for OPAAv and OPGSHv in Colombian cities compared to Canadian, American, European, Asian and Latin American cities. OPv varies between and within-cities. Differences between cities were significant, but those within cities were not, thus differences were not explained by the site types. The lack of PM2.5 composition precluded a more detailed analysis of the site type impact on OP. Moreover, we observed that OPm of PM2.5 is negatively, not significantly, correlated with PM2.5 mass concentrations. In addition, PM2.5, OPGSHv and OPAAv from our study didn’t show the patterns that are usually observed in PM2.5 maps built from official air quality monitoring data. These findings suggest that current air quality monitoring based only on PM2.5 concentration does not reflect OP exposure variations and OP burden in these urban settings. Future work is needed to assess the extent to which OP is associated with long-term adverse health effects in these populations and the source contribution to OP.