Ambient air pollution is regarded as a significant threat to human health. According to the World Health Organization (1), the combined effects of ambient air pollution and household air pollution account for 6.7 million premature deaths globally. Fine particulate matter, with an aerodynamic diameter of 2.5 μm (PM_2.5), has been associated with many adverse health outcomes (2–5) because it can penetrate deeper into the alveolar regions of the lungs (6, 7). Human health risks of exposure to PM_2.5 extend beyond respiratory diseases to include cerebrovascular and cardiovascular diseases (8, 9), lung cancer, cardiopulmonary mortality, stroke, asthma, arrhythmia (10, 11). Exposure to PM_2.5 affects individuals differently based on their health status, age (12), gender and duration of exposure (13).

Vulnerable population sub-groups are likely to develop diseases due to lifetime PM_2.5 exposure to PM_2.5 (14). These groups include pregnant women, infants and children. Children breathe faster than adults because smaller lungs require more frequent breaths to meet oxygen demands (15, 16). Exposure to high PM_2.5 in these groups can affect how their lungs develop over time, increasing their chances of developing lung diseases (17). Exposure to PM_2.5 during pregnancy can impact fetal development and may have long-term health consequences (18, 19). The physiological development, high Inhalation Rate (IR), Body Weight (BW), Exposure Duration (ED) and Exposure Frequency (EF) are some factors that make individuals vulnerable to the adverse health risks of PM_2.5 exposure (20). Organs such as the lungs and heart undergo significant maturation over time (21). Exposure to PM_2.5 during these stages of organ development can lead to long-term health risks (22). The older adult are more at risk of developing adverse health risks due to weakened respiratory and cardiovascular systems (23). Furthermore, pre-existing medical conditions, including asthma, heart disease and chronic obstructive pulmonary disease (24), increase their vulnerability to the harmful effects of pollution (25).

Fine particulate matter (PM_2.5) is emitted from natural and anthropogenic sources (26). Natural sources of PM_2.5 include forest fires and volcanic eruptions. Forest fires emit huge amounts of PM_2.5 from the combustion of vegetation and biomass (27). Volcanic eruptions release ash and PM_2.5 into the atmosphere. Anthropogenic sources of PM_2.5 include industrial processes, domestic fuel burning, mining operations and agricultural activities. Burning of wood, coal and other solid fuels for residential heating and cooking also emits PM_2.5 (28). Gold mine tailings storage facilities (TSFs) are also a potential source of PM_2.5 emissions. The TSFs are designed to store waste generated during mineral extraction and processing. These tailings consist of crushed rocks, particles and residuals of chemicals left over from the gold extraction process (29).

PM_2.5 comprises a complex mixture of biological and chemical components (30). The chemical and biological components of PM_2.5 may originate from various sources (31). The biological components of PM_2.5 include bacteria, fungi, viruses and pollen (32). The chemical components of PM_2.5 include organic compounds (33), inorganic compounds (34) and trace elements (35). The physicochemical properties of fine particulate matter influence their toxicity and health impacts. Particles with larger surface areas and reactive chemical elements, can induce oxidative stress and inflammation, resulting in adverse respiratory and cardiovascular outcomes (36).

The assessment of exposure to PM_2.5 is a complex process. It requires information about the sources, site selection and data quality assurance (37). PM_2.5 concentrations vary spatially due to differences in emission sources, meteorological conditions, and topographical features. Furthermore, PM_2.5 concentrations can fluctuate over time due to diurnal patterns and seasonal variations (38). Human health risks can be assessed by applying the HHRA model (39), which estimates the likelihood of adverse human health risks associated with PM_2.5 exposure (40). Most studies have used deterministic HHRA models to assess potential human health risks (41–43). A deterministic HHRA model uses single-point estimates for input parameters to calculate specific risk values. One of the advantages of deterministic risk assessment is its simplicity. The disadvantage is its inability to account for the variability among individuals and the uncertainties in environmental data (44).

This study used a probabilistic HHRA model to estimate the potential human health risks of continuous exposure to PM_2.5 concentrations. The probabilistic HHRA model was chosen because it uses statistical techniques to account for variability and uncertainty in risk estimates (45). Furthermore, the model employs Monte Carlo simulations to generate a distribution of risk estimates based on random sampling from the probability distributions of input parameters (46). The advantage of probabilistic HHRA provides a more complete picture of potential health risks by estimating the range and likelihood of different outcomes. The disadvantage is that it requires detailed data on the distribution of input parameters (47).

This study introduces a novel approach by applying a probabilistic Human Health Risk Assessment (HHRA) to evaluate the health risks associated with PM_2.5 exposure in eMbalenhle, a community near gold mine tailings storage facilities (TSFs). While most existing studies rely on deterministic risk assessments that often fail to capture the inherent uncertainties and variability in PM_2.5 exposure (39, 41, 43), this study addresses these limitations by adopting a probabilistic framework. Furthermore, the application of this method to communities affected by gold mine TSFs remains limited in the current literature. Importantly, this is the first study to assess human health risks in eMbalenhle using PM_2.5 data collected from low-cost monitors (LCMs), offering a cost-effective and practical strategy for air quality and risk assessment in resource-constrained settings. The findings of this study will contribute to Sustainable Development Goal 3 (SDG 3). SDG 3 aims at ensuring good health and well-being for all. Monitoring the ambient PM_2.5 contributes to achieving Target 3.9 of SDG 3, which addresses environmental pollution and its health consequences (48).

The study was approved by the University of the Witwatersrand Human Research Ethics Committee [HREC] (No: HRECNMW21/05/09).

This study reported an annual GM PM_2.5 concentration of 17 μg/m³. Various PM_2.5 sources, including traffic emissions, industrial activities, domestic cooking, space heating, waste-burning and dust from gold mine TSFs, could have contributed to the increased levels of PM_2.5 in eMbalenhle. This annual concentration level was below the South African NAAQS of 20 μg/m³. However, it exceeded the WHO air quality guidelines and the USEPA NAAQS of 5 μg/m³ and 9 μg/m³, respectively. This indicated that although the local air quality standards are met, the exposed population may still experience significant health risks when the reported PM_2.5 concentrations are compared with international reference standards. The exceedances of international standards from February 2022 to February 2023 indicated a potential public health concern, especially for vulnerable population sub-groups living close to sources of PM_2.5 and gold mine TSFs. Although annual PM_2.5 emissions reported in this study are slightly below the SA NAAQS, Millar et al. (65) that, for the past 10 years (2009–2019), the annual PM_2.5 concentrations in eMbalenhle exceeded the SA NAAQS each year. The study used high-resolution data from the South African Weather Service (SAWS) air quality monitoring station.

Different countries have varying ambient PM_2.5 NAAQSs. This is due to environmental, health, social, political and economic factors (66) affecting these countries. The WHO (51) provides guidelines for air quality, but countries adopt these recommendations differently based on their national circumstances. In most cases, developing countries may prioritize economic growth and industrialization over strict air quality standards to avoid limiting industrial expansion (67). In contrast, developed countries may impose more stringent regulations to protect public health and the environment. Moreover, countries with advanced air quality management and monitoring technologies may set and enforce stricter NAAQS (68). Countries with less technological capacity may set more lenient standards due to challenges in air quality monitoring (69).

For the probalistic HHRA, the PM_2.5 concentrations were adjusted to reflect an accurate estimation of an equivalent full-day exposure over 1 year in the eMbalenhle population. The results showed that the median (50th percentile) for non-carcinogenic risk was low (below 1) across all population subgroups based on SA NAAQS. However, compared to the more restrictive USEPA and WHO air quality guidelines, even the median HQ values demonstrated a substantially increased risk with HQ values of 1.5 and 2.8, respectively. Amnuaylojaroen and Parasin (70) computed higher HQ values than this study. This is partly because they used RfC of 35 μg/m³ to calculate the HQ and worst-case assumptions regarding the risk parameters. The mean HQs were 2.93, 2.59, 2.28, 1.88, and 1.26 for newborns, toddlers, young children, school-age children, and adolescents, respectively.

Although the WHO air quality guidelines are widely adopted and used in many countries, they are not absolute and can be overly simplistic. First, the guidelines fail to account for cumulative risks and variability in individual susceptibility. Second, the guidelines fail to account for risk tolerance, which can vary between populations and regulatory bodies. Given this, there is a greater need to develop more robust and advanced approaches that factor in these issues to reflect real-world exposures accurately and avoid underestimating or overestimating risk.

For the cancer risk (CR), results showed values greater than 1.0 × 10⁻⁴ (1 in 10,000) for all population subgroups, which is generally considered significant, calling for stricter regulatory action and interventions to prevent further exposure. For example, the 50th, 75th and 95th percentile values indicated significant CRs (all greater than 1.0 × 10⁻⁴) across all population sub-groups. The CR became more significant in male and female elders. This result is similar to Wu et al. (71), who found the CR more significant in male elders. In contrast, Stapelfeld et al. (72) found that females were more at risk of developing cancer than males. In our study, the slightly higher CR for men was due to the lower life expectancy than women, but this difference is not statistically significant. Furthermore, males have different hormonal profiles compared to females (73) and as they age, the risk of developing cancer increases (74).

The female children showed the least chance of developing CR because of PM_2.5 exposure compared to the other groups. Our estimations were determined by the ED/AT ratio, which was the lowest for toddlers and the highest for elders. Since toddlers and children are young, their risk for developing cancer is reduced because they have less accumulated genetic mutations (75). Contrary to this, Siegel et al. (76) observed that pediatric cancers were more pronounced in male children than in females.

The sensitivity analysis indicated that C_air and ET had the most significant impact on non-cancer risks. Since C_air is measured, the estimates for ET are key for estimating the extent of non-cancer risk. These findings call for regulatory action and policies to reduce exposures, especially to vulnerable groups. Strategies could include promulgating stringent emission standards for petrochemical industries, power plants, coal mines and gold mines. Inside the community, there should be an advocation for the transition to cleaner fuels, energy-efficient cookstoves and renewable energy sources to mitigate combustion-related PM_2.5 emissions (77). Other strategies could include the spatial resolution of air quality monitoring in the area by adopting low-cost sensors (78), issuing early warnings for high pollution days (79), and raising public awareness to empower communities to take protective measures. On the other hand, the ED was the most sensitive parameter for the CR due to the assumed uniform distribution in the adult subgroup. For the older adult, where the ED/AT ratio is close to one, the variation in C_air becomes more significant for the estimates of the CR. In a similar study, Amoatey et al. (80) observed that long durations of exposure to ambient PM_2.5 increased adverse health outcomes amongst the Roman population.

Applying the probabilistic HHRA model may be more complex than the deterministic approach (81). The determinist HHRA approach uses fixed point estimates for input variables (ED, ET, EF, AT) (82). Often, the maximum values are used, resulting in worst-case assessments. In contrast, the probabilistic HHRA uses probability distributions for input variables to account for variability and uncertainty (83). Amnuaylojaroen and Parasin (70) applied a deterministic risk assessment model to assess the health impacts of exposure to PM_2.5 in different age groups of children in Northern Thailand. Compared to the probabilistic HHRA model that was applied in this study, their model considered the body weight and inhalation rate as variables, as also in Morakinyo et al. (84). The study also postulated that PM_2.5 exposure might affect children differently depending on gender, with males at a higher risk than females in adolescence (70).

The findings of the study demonstrated that the SA NAAQS may not adequately protect public health as it may underestimate the risks associated with PM_2.5. To protect public health, there is a need for the government to adopt more stringent standards and align them closely with international standards, including the WHO guidelines.

Using a probabilistic human health risk assessment (HHRA) approach, the study revealed that non-carcinogenic risks were low when benchmarked against the SA NAAQS, but substantially elevated when compared with international guidelines. Vulnerable groups, particularly newborns and toddlers, exhibited higher hazard quotients (HQs), underscoring their increased susceptibility. For cancer risks (CR), results exceeded the generally acceptable threshold of 1 in 10,000 across all population groups, particularly affecting the older adult.

PM_2.5 concentrations were the most significant factor in increased non-cancer health problems in all sub-population groups. This emphasizes the need for control measures to reduce PM_2.5 concentrations, including the maintenance of TSFs and increasing monitoring in hot spot areas. The ED was the most critical factor for the cancer risks, followed by the air concentration for all population sub-groups except for the older adult. Strategies to reduce the ED and EF, such as limiting outdoor activities during high pollution events could further minimize health risks associated with exposure to PM_2.5. The sensitivity analysis indicated that ambient PM_2.5 concentration (C_air) and exposure time (ET) had the most influence on non-cancer risk estimates, while exposure duration (ED) was key for cancer risks. These observations indicate the urgent need for targeted interventions to minimize exposure, especially among vulnerable populations. Policy actions could include tighter emissions controls, community-level interventions to promote dust control alternatives, TSF management strategies, and expanded air quality monitoring networks using low-cost sensors.

Considering the above, this study calls for an urgent review of the current South African NAAQS. However, policymakers should strike a good and functional balance between economic imperatives and public health outcomes. Even though the WHO provides global benchmarks, countries can adopt these to meet their socio-economic context, technological feasibility and regulatory frameworks.

The LCM measurement results over the year represent the temporal variations; however, spatial variations were not covered, limiting the representativeness of the PM_2.5 concentration data for the general population of eMbalenhle. In addition to this measured parameter (C_air), the HQ and CR risk equations, distribution and values for all other parameters, except life expectancy, were only estimates and not based on actual data or information from the community. In-depth population studies are needed to collect more relevant information, especially for the highest sensitivity risk parameters. Therefore, this study’s HQ and CR probability distributions can only be considered to flag potential health risks for the eMbalenhle residents. Lastly, this study made assumptions for the ED and EF for all population subgroups. This can limit the reliability and generalizability of the findings, as these parameters may not accurately reflect the diverse exposure patterns across the population subgroups. Uniform distribution patterns and assumptions of uniform exposure may overlook important activity data, including age-specific behaviours and varying susceptibility. This could lead to potential underestimation or overestimation of risks.