Across Chicago, 24-h average tract-level PM2.5 concentrations in the pre-peak, peak, and post-peak WFS periods were 14.5 μg/m3, 99.7 μg/m3, and 16.6 μg/m3, respectively. Averaged over the study period, the average PM2.5 concentration was 17.1 μg/m3. To visualize changes in PM2.5 over time, we plotted daily tract-level PM2.5 concentrations (Fig. 2). PM2.5 levels show an increasing, non-monotonic trend, from May 1 through June 26, 2023. On June 27, the average PM2.5 concentration in the City of Chicago was 96.2 μg/m3, representing a 62.8 μg/m3 increase over the day prior (188% change) and an 81.7 μg/m3 increase over the pre-peak period (563% change), overall. Average PM2.5 levels remained elevated on June 28 (103.3 μg/m3), then declined to 55.1 μg/m3 on June 29.

The red dashed line represents the start of the peak WFS period (June 27, 2023), the green dashed line represents the US EPA National Ambient Air Quality Standard (NAAQS) daily standard (35 μg/m3), and the blue dashed line represents average PM2.5 concentration across all non-peak WFS days (15.3 μg/m3) for reference.

We calculated pre-peak, peak, and post-peak census tract-level PM2.5 concentrations by region (Fig. 1 and Supplemental Material [SM] Table S1). In the pre-peak period, PM2.5 concentrations were highest in the Far Southwest Side, with a median (IQR) of 14.4 (10.9, 20.1) µg/m3, and the Southwest Side, with a median (IQR) of 13.2 (10.1, 19.4) µg/m3. The lowest concentrations in the pre-peak period were observed in the West Side (11.5 [8.9, 17.9] µg/m3). During the peak WFS period, the Central (103.7 [99.2, 107.5] µg/m3), Southwest Side (102.2 [97.8, 107.1]), and North Side (102.1 [101.3, 107.7]) had the highest PM2.5 concentrations, while the South Side (96.2 [89.7, 105.6] µg/m3) and Far Southeast Side (84.5 [82.2, 88.8] µg/m3) had the lowest concentrations. Finally, in the post-peak period the highest PM2.5 levels were observed in the Southwest Side (12.4 [9.9, 21.5] µg/m3) and Far Southwest Side (13.8 [10.5, 22.8] µg/m3), while the lowest levels were observed in the West Side (10.9 [8.5, 18.9] µg/m3) and Northwest Side (11.1 [8.7, 18.9] µg/m3).

Regional demographic, socio-economic, and health characteristics are provided in Table 1. The southern parts of the city have the lowest economic and educational levels and the highest health burdens. For example, the Southwest Side, which has a predominantly Hispanic population (61.9%), has the highest percentage of residents with diabetes, uninsured individuals, and those without a college degree. The Far Southwest Side has the highest rates of hypertension and rent burden among residents. The South Side reports the highest prevalence of asthma and the largest percentage living below the poverty line, while the Far Southeast Side has the shortest life expectancy and the highest rates of chronic obstructive pulmonary disorder (COPD). The South Side, Far Southwest Side, and Far Southeast Side are predominantly NHB communities (proportion of residents that are NHB ranges from 85.4 to 92.1%). The South Side, Southwest Side, and Far Southwest Side have the largest proportions of youth under 18 (range: 22.8–23.2%), whereas the Far Southwest Side and Far Southeast Side have the highest percentages of residents aged 65 and older (18.2% and 18.5%, respectively).

To compare regional differences in PM2.5 concentrations across pre-peak, peak, and post-peak time periods, we fit an interaction model. This model includes an intercept and an indicator variable representing the start of the peak WFS period; smooth terms for the pre-peak and post-peak WFS periods that we allowed to vary by region via a factor-smooth interaction; and a multiplicative interaction term between the peak WFS period and region. All smooth terms, which are summarized by the effective degrees of freedom, were significantly different from 1 for both pre-peak and post-peak WFS periods (SM Table S2, SM Table S3). For reference, effective degrees of freedom values equal or close to 1 indicate a linear relationship while effective degrees of freedom values greater than 1 suggests a non-linear relationship; thus, results are indicative of non-linear temporal trends in PM2.5 concentrations (SM Figure. S1). The adjusted R2 = 0.657.

Pairwise differences between region-level smooth interactions during the pre-peak (purple line) and post-peak (green line) WFS periods are plotted in Fig. 3, using the Southwest Side as the referent region. Recall, across the nine regions in Chicago, the Southwest Side had among the highest pre-peak PM2.5 levels and the highest proportion of population with diabetes, highest proportion uninsured, and lowest proportion with a college degree. Thus, the selection of the Southwest Side as a referent was not arbitrary, but other regions could be selected by applying alternative criteria. For plots of pre-peak and post-peak WFS concentrations relative to the NAAQS daily standard for PM2.5 (35 µg/m3), see SM Fig. S2.

Panel A shows daily estimated differences in pre-peak (purple) and post-peak (green) WFS concentrations relative to the Southwest Side. Panel B shows differences in PM2.5 on the two peak WFS days relative to the Southwest side.

Smoothed pre-peak temporal trends in PM2.5 concentrations compared to trends in the Southwest region (an area with chronically elevated PM2.5 from traffic concentrations and industry) are shown in Fig. 3 Panel A (purple line with shading to indicate 95% confidence band). In the interaction model, PM2.5 concentrations on the first day of the study period in the South, Far Southwest, and Far Southeast Side regions were not significantly different from PM2.5 concentrations in the Southwest Side (referent), where PM2.5 concentrations were 9.93 µg/m3 [95% CI: (9.30, 10.56)]. In contrast, PM2.5 concentrations in the Far North Side, Northwest Side, North Side, West Side, and Central regions were all significantly lower (i.e., differences of -1.16, -2.23, -1.61, -1.09, and -1.40 µg/m3, respectively) than PM2.5 in the Southwest Side. The Northwest Side, West Side, South Side, and Far Southeast Side had, for at least some portion of the pre-peak period, lower daily PM2.5 concentrations than the Southwest Side.

The Northwest Side and North Side regions had higher PM2.5 concentrations (difference of 2.29 and 1.81 µg/m3, respectively) than the Southwest Side in the peak period. In contrast, the West Side, South Side, and Far Southeast Side regions had peak-period PM2.5 concentrations that were significantly lower (differences of –1.91, –3.76, and –6.48 µg/m3, respectively) than the Southwest Side (Fig. 3 Panel B). The Far North Side, Central, and Far Southwest Side regions had peak-period PM2.5 concentrations that were not significantly different from the Southwest Side.

Smoothed post-peak temporal trends in PM2.5 concentrations are shown in Fig. 3 Panel A (green line with 95% confidence band). PM2.5 concentrations in the Southwest Side were higher than those observed in the Far North Side, Northwest Side, North Side, West Side regions in the first 10 to 12 days following the peak period (i.e., June 29, 2023–July 10, 2023). That is, following the peak WFS days—when PM2.5 concentrations were higher in Northwest Side and North Side regions—PM2.5 concentrations in the Southwest Side were slower to decrease compared to other regions (with the notable exception of the Far Southwest Side, where PM2.5 concentrations declined even more slowly following the peak WFS period).

Additionally, Fig. 4 shows, for each region, the number of days after the start of the peak WFS period until regional PM2.5 concentrations, measured as 3-day moving averages, returned to the study-wide pre-peak mean (14.5 μg/m3). We used 3-day moving averages rather than daily values because, across all regions, PM2.5 dropped briefly on a single day in the post-peak period but remained substantially elevated both before and after that one-day trough; that is, the decline was not sustained. Thus, the 3-day moving average better captures lasting recovery in air quality. The Southwest Side, Central, South Side, and Far Southeast Side returned to pre-peak PM2.5 levels 11 days after the start of the peak WFS period, and the Far Southwest Side took 12 days. In contrast, the Far North Side, Northwest Side, North Side, and West Side recovered more quickly, returning to pre-peak levels just 7 days after the peak period began. Additionally, because pre-peak PM2.5 concentrations varied by region, we assessed the time to return to pre-peak PM2.5 levels using region-specific pre-peak PM2.5 definitions rather than a study-area-wide PM2.5 definition. The Southwest Side, Central, South Side, Far Southeast Side, and Far Southwest Side took 11 days to return to region-specific pre-peak average PM2.5 levels; the Far North Side, Northwest Side, North Side, and West Side returned to pre-peak PM2.5 levels within 7 days (SM Fig. S3).

The horizontal black dashed line represents the average PM2.5 concentration during the pre-peak WFS period across Chicago (14.5 μg/m3). The vertical red dashed line represents the first day in which the 3-day moving average PM2.5 concentration in that region reached the pre-peak average of 14.5 μg/m3.

In summary, PM2.5 concentrations in the Southwest Side and Far Southwest Side were higher at baseline and, in some cases, throughout the pre-peak period, then declined more slowly in the first two weeks following the peak WFS period. As a result, PM2.5 concentrations were higher in these two southern regions for most of the study period, if not on the peak WFS days themselves. Moreover, differences in PM2.5 levels on peak WFS days were small relative to peak period levels.

As an alternative to assessing social and demographic characteristics at the region level, we also assessed trends in PM2.5 concentrations in pre-peak, peak, and post-peak periods using social and demographic characteristics summarized at the census tract level. The approach and findings are described in SM Appendix A and Table S4. Results were generally consistent with findings reported for the main analysis, i.e., higher PM2.5 concentrations in the Southwest Side and Far Southwest Side at baseline, and concentrations that were slower to decline following the peak WFS period. Additionally, we fit a segmented linear ITS model as an alternative to the GAM ITS model; again, findings were generally consistent with those reported in the main analysis. Details are provided in SM Appendix A, Table S5, and Fig. S4.