Amidst the ongoing transition to renewable energy generation, coal remains a key source of energy with global coal consumption reaching a record high in 2023 (International Energy Agency (IEA), 2023). In the United States (U.S.), there exists expansive infrastructure for coal extraction via mining and transport via rail to coal terminals for handling, storage, and export. Coal mining activities are recognized contributors to particulate matter (PM) exposure, associated adverse health outcomes, and impacts to quality of life in both occupational (Shekarian et al. 2023; Petsonk et al. 2013; Liu and Liu 2020; Hendryx 2015; Hendryx et al. 2020b) and fenceline community (Kurth et al. 2014; Hendryx et al. 2020a; Hendryx and Entwhistle 2015; Cortes-Ramirez et al. 2018) contexts.

Compared to the state of knowledge about occupational exposure and health effects of coal mining, less is known about how coal transport, handling, and storage affect air quality, health and quality of life in neighboring fenceline communities. Recent research in a community at the fenceline of coal by rail infrastructure and an open-air coal terminal in Richmond, California linked coal transport, handling, and storage activities to local air pollution (Ostro et al. 2023a, b). This study demonstrated increases of PM_2.5 (particles of aerodynamic diameter ≤ 2.5 μm) attributable to the passage of uncovered coal trains (8.32 μg/m³ [95% CI = 6.37, 10.28; p < 0.01]) (Ostro et al. 2023b). Although these PM_2.5 increases may be transient, their repeated periodicity was estimated to contribute to increased morbidity and mortality due to asthma and cardiovascular disease (Ostro et al. 2024). This study also found coal and petcoke (petroleum coke) dust in the PM_2.5 and coarse (PM_{10 − 2.5}) size ranges at concentrations ranging from 0.1 to 0.6 μg/m³ and 3 to 18 μg/m³, respectively, via passive aerosol sampling in the community and computer-controlled scanning electron microscopy and energy-dispersive X-ray spectroscopy (SEM-EDX) (Ostro et al. 2023a). There remains a paucity of information about air pollution burdens in communities neighboring coal terminals, which have employed a wide spectrum of mitigation strategies to control the off-site migration of coal dust and other PM air pollution into proximal communities. Questions also remain about the temporal variability of air pollution in relation to weather conditions and activity patterns at coal terminals.

In Curtis Bay, South Baltimore, Maryland (hereafter Curtis Bay), an open-air coal terminal operated by CSX Corporation is a local pollution source of longstanding community concern (Fabricant 2023). In 2023, approximately 30 million metric tons of coal were exported from the Port of Baltimore. The CSX coal terminal in Curtis Bay–one of two operating coal piers in the Port of Baltimore–has an annual throughput capacity of over 12 million metric tons (U.S. Energy Information Administration (EIA), 2024; CSX Corporation). The coal terminal is ~ 300 m from nearby homes and the Curtis Bay recreation center with local schools and businesses less than a kilometer away. For decades, Curtis Bay residents have observed and reported a daily accumulation of black dust in community spaces, at business and residential properties, and inside people’s homes. In December 2021, there was an explosion at the coal terminal which showered the neighborhood with coal dust, shattered windows of nearby homes, and caused widespread panic (South Baltimore Community Land Trust 2023; Kazanjian 2022; Maryland Department of the Environment 2022).

A recent community-driven research study involving a partnership between Curtis Bay residents and several academic institutions observed coal dust (via SEM-EDX) in settled dust collected as far as 1,235 m from the coal terminal near a local high school (Aubourg et al. 2024). Curtis Bay community residents recently documented a fugitive dust emission event (about 30 min in duration) generated by the operation of an autonomous ballast cleaner along the rail lines at the coal terminal (Shen 2024; South Baltimore Community Land Trust (SBCLT), 2024).

In response to continued local concerns and questions about coal dust exposure in the Curtis Bay community, our collaborative team of community and academic scientists aimed to collect high-resolution 1-minute-level data on air pollutants, meteorological conditions, and coal terminal activities. This study aimed to determine how changes in wind direction (downwind of coal terminal vs. not) and camera-based observations of coal terminal activity (visible bulldozer activity on coal piles vs. none) were related to changes in air pollution burden in Curtis Bay.

This study demonstrated increased PM and BC air pollution burden in a community bordering a coal storage, handling, and export terminal which was related to time-varying patterns of wind direction and visible bulldozer activity at the terminal. Further, when both occurred jointly–i.e., during time periods when the monitoring site was downwind of the coal terminal and there was visible bulldozer activity–PM and BC air pollution burden was greatest. We used an effect measure modification analysis approach to understand evidence of statistical interaction, which demonstrated strong consistency in terms of the direction, order of magnitude, and precision of joint effect estimates across a range of transient time scales (1-, 5-, 15-, 60 min; Supplementary Results). The incorporation of NO as a control for local combustion and / or diesel exhaust from the bulldozers also enhanced the ability of the analysis to isolate the relationships of interest related to windblown dust.

Dust sources present particle size distributions that may be uni- or bi-modal, with tailing that extends into the finer particle sizes (Junge 1955; Brahney et al. 2024; Chalvatzaki et al. 2012; Seinfeld and Pandis 2006). Although the mode for the diameter of coal dust is ~ 10 μm, a substantial fraction of the particles fall in the respirable range under 2.5 μm. Coal dust, like wind-blown dust in general, follows a Gaussian, log-normal distribution (Junge 1955): (2) dNdlogr=Cr−β where N is the number of particles, r is their radius, C is a constant, and β is a constant ~ 3 in the classic Junge distribution (Junge 1955). This theory has been experimentally verified for coal and is described in textbooks (Barone et al. 2019; Seinfeld and Pandis 2006; Tessum and Raynor 2017). The large particles will fall out of the atmosphere rapidly, with a settling velocity corresponding to a transport distance of ~ 1 km, however, smaller particles will persist until rained out in ~ 1 week (Seinfeld and Pandis 2006). It is thus noteworthy that we observed consistency of associations not only for coarse PM₁₀ and TSP, but also for PM₁ and PM_2.5. Smaller size fraction PM, like PM₁ and PM_2.5, are pollutants typically associated with regional pollution, traffic, and combustion, while BC is also associated with traffic emissions. Consistency of associations with smaller PM and BC with downwind bulldozer activity may suggest that BC measurements alone or in combination with PM may have utility for identifying black PM air pollution, like coal dust, more broadly in communities bordering coal storage, handling, and export facilities. Black carbon monitors’ measurement methods select for fine mode aerosol while coal dust is generally coarse mode so future efforts should include speciated PM₁₀ or TSP.

We employed a fixed effects regression modeling approach which cannot be confounded by factors that do not vary at the same time scale as the temporal scale of the exposure-response being investigated. The results of this regression modeling approach for PM₁, PM_2.5, PM₁₀, TSP and BC support the hypothesis that mean concentrations of these air pollutants are higher when the winds blow toward monitoring locations downwind from the coal terminal. The statistical methodology adopted has certain limitations. When only considering downwind wind direction as the exposure, there is a chance of potential confounding from another source of pollution coming from the same direction as the coal terminal. This confounding risk is mitigated to a large extent when using the visible bulldozer activity data as the exposure as it is unlikely that there is another temporal confounder that is highly concurrent with bulldozer activity, particularly as we control for local combustion and / or diesel exhaust from the bulldozer activity. The assumption of linearity in the fixed effect models used can also be violated. The current analysis also only offers an average effect size based on pooled data from all ten air monitoring sites. Separate site-specific analysis can reveal variability in these effect sizes and how they relate to distance from the coal terminal.

Our focus on transient exposure-response time scales reflects the community lived experience involving reports of frequent short term dust events that recur frequently (analogous to our 1-min, 5-min, and 15-min time scales) as well as longer events that have been documented (analogous to our 60-min time scale). Curtis Bay community members recently documented an example of a longer high visibility fugitive dust event at the coal terminal which lasted for approximately 30 min (Shen 2024; South Baltimore Community Land Trust (SBCLT), 2024). The consistency of our exposure-response associations across all times scales considered (1-min, 5-min, 15-min, and 60-min) for both explanatory variables (downwind of the coal terminal and visible bulldozer activity) and all PM and BC air pollutants (Figs. 3, S4–S6; Table S8) is coherent with neighbors’ reports about the nature of fugitive dust events.

To our knowledge, this study is the first to investigate the relation between wind direction and coal terminal activity upon the air pollution burden in a proximal fenceline community. This study has some limitations including TSP measurements by moderate-cost sensors do not have a convenient reference comparator for validation and calibration, unlike PM_2.5, PM₁₀, and BC. Observed variation in TSP values with large standard deviations may be attributed to instrument error or natural variability due to the higher settling velocity of larger particles resulting in spatial heterogeneity on scales of ~ 100 m. In the coding of trail camera data, tree foliage inhibited the identification of bulldozer activity on certain coal piles during certain times of year. During these times, if there was no bulldozer activity at visible coal piles, the observation was coded as no activity “0” rather than missing “.”. We hypothesize that this could have led to an underestimate of the true extent of bulldozer activity that was occurring on the coal piles during these tree foliage-impacted times of year. However, this potential misclassification of bulldozer activity was likely non-differential with respect to the outcome (air pollution).

Our study also has several notable strengths. Our research questions were informed by the concerns, observations, and experiences of Curtis Bay community members who have lived proximal to the coal terminal for decades. This guided our analytic approach involving conditional fixed-effects regression modeling of multi-pollutant data. Thereby, we controlled for time-invariant characteristics of each location, such as proximity to pollution sources or local infrastructure, which could otherwise confound our observed association. By focusing on within-monitor variation in air pollutants over time, fixed-effects models isolated the effects of wind direction and visible bulldozer activity while removing the potential influence of location-specific non-time varying factors. Another strength is the use of the MODULAIR multi-pollutant air sensor which monitors dust in multiple particle size bins. This may be used for quality control to improve confidence in results– i.e., temporal patterns observed in multiple, independent channels are more realistic in mixed urban/industrial environments than a unique signal in a discrete particle size bin.

Future directions for this line of research could extend the conditional fixed effects regression modeling approach to other time-resolved activity patterns at the coal terminal. While our analysis focused on visible bulldozer activity due to its association with general terminal operations, future analyses can be extended to code time-resolved activity patterns of coal transport by trains, coal ship / barge activity, coal pile height, and use of the dust wet suppression system. Additionally, the conditional fixed-effects regression modeling approach used assumes a linear relationship, and future studies could explore other non-linear approaches, assess and account for potential unmeasured temporal confounding and lags. Our preliminary analysis of how air quality burden varies with proximity of the measurement sites to the coal terminal could be expanded to investigate location-specific and other spatial patterns. Further, there are opportunities for source attribution analyses to identify combinations of air pollutants that are associated with coal terminal emissions.

Short-term exposure to PM has been increasingly linked to adverse health outcomes, with more research still needed (Gutiérrez-Avila et al. 2023; Orellano et al. 2020). Fine PM (PM₁ and PM_2.5) are of particular concern due to their ability to penetrate deep into the respiratory tract and enter the bloodstream, exacerbating cardiovascular and respiratory conditions. Acute exposure to PM₁₀ has been associated with increased hospital admissions for asthma, bronchitis, and chronic obstructive pulmonary disease (COPD), as well as heightened risks of stroke and cardiac events (Alessandrini et al., 2016, Orellano et al. 2020). These impacts are particularly pronounced in vulnerable populations such as children, the elderly, and individuals with pre-existing health conditions. Both long-term and acute exposure to BC have been linked to morbidity and mortality due to cardiovascular and respiratory disease (Song et al. 2022; Luben et al. 2017; Hvidtfeldt et al. 2019; Ostro et al. 2015; Janssen and Joint 2012), in addition to marked impacts upon fetal, pediatric, and maternal health outcomes (Paunescu et al. 2019; Goriainova et al. 2022). Residents in communities of color and low wealth are also disproportionately exposed to PM, BC, and other forms of air pollution, experience elevated impacts of this pollution, and are often overburdened by the cumulative impacts of other chemical and non-chemical stressors. Higher PM_2.5 exposure was also observed for Hispanic and Black residents in San Francisco, CA living near uncovered coal by rail transport infrastructure(Ostro et al. 2024).

Community observations, actions taken by state/local governments, and the results of this research suggest additional measures are warranted to end or mitigate fugitive dust emissions related to coal terminal operations and neighborhood health and quality of life impacts associated with such emissions. Currently, the facility maintains a wet dust suppression system and has a partial porous wind fence at the southern edge of the coal terminal. When restricting our analysis to the one location within the network on the opposite side of the wind fence, our results were generally consistent with the analysis that pooled data across all locations. This suggests that off-site community air pollution burden is not reduced by wind screen netting mitigation strategies. A range of additional strategies have been submitted by community members. These include, but are not limited to: (1) denial of the application for the renewal of its operating permit based on a record of complaints and violations; (2) full enclosure of the coal piles paired with enforceable limits on PM from the terminal; (3) the imposition of fees on a per-ton basis of coal transferred through the terminal to mitigate negative externalities; and (4) transitioning to the storage and handling of a safer material at the facility.

The findings of this study demonstrated higher PM and BC air pollution burden in Curtis Bay when downwind of the coal terminal and when there was visible bulldozer activity–highest air pollution burdens were observed when both occurred jointly. Results are consistent with the lived experiences of Curtis Bay community members who have documented ongoing, repeated patterns of black dust accumulation at and around their homes alongside recordings of transient fugitive black dust plumes at the coal terminal. Further context for our findings is provided by a recent study demonstrating the detection of coal dust in the black dust that accumulates in residential and community spaces of the Curtis Bay neighborhood (Aubourg et al. 2024). Overall, our findings in context with complementary research and community experiences contribute to a coherence of data across varied analytical and observational platforms that respond to residents’ questions and concerns with dark dust accumulation in Curtis Bay. Compared to both Baltimore City and the State of Maryland, the Curtis Bay community is overburdened by cumulative impacts of economic hardship, high vacancy rates, historical pollution burden, and health disparities such as higher mortality due to chronic obstructive pulmonary disease and lung cancer (Baltimore City Health Department 2017). Strategies to mitigate the observed air pollution burdens in the Curtis Bay community appear warranted, including those that could prevent windblown and mechanical dusts that might be associated with operating practices at the coal terminal.

The online version contains supplementary material available at https://doi.org/10.1007/s11869-025-01773-w.