Estimates of submicron particulate matter (${\mathrm{P}\mathrm{M}}_1$) concentrations for 1998–2022 across the contiguous United States

Abstract

Background

Excess health risk estimates of exposure per unit mass concentration of fine particulate matter (${\mathrm{P}\mathrm{M}}_{2.5}$) still exhibit a wide range, potentially due to variations in aerosol size and composition. Submicron particulate matter (${\mathrm{P}\mathrm{M}}_1$) was recently reported to exert stronger health impacts than ${\mathrm{P}\mathrm{M}}_{2.5}$ from studies in China, while a dearth of long-term ${\mathrm{P}\mathrm{M}}_1$ data in the United States has prohibited such investigations despite a wealth of cohorts.

Methods

We estimate biweekly gapless ambient ${\mathrm{P}\mathrm{M}}_1$ concentrations and their uncertainties at 1 km² resolution across the contiguous United States (CONUS) over 1998–2022, from hybrid estimates of ${\mathrm{P}\mathrm{M}}_{2.5}$ chemical composition that merged information from satellite retrievals, air quality modeling, and ground-based monitoring. The mass fractions of ${\mathrm{P}\mathrm{M}}_{2.5}$ components with diameters <1 μm are constrained from observations for four major components (sulfate: 91%; ammonium: 82%; nitrate 77%; organic matter: 86%), and from established scientific understanding for the other components (black carbon: 100%; dust: 12%; sea salt: 97%).

Findings

The biweekly ${\mathrm{P}\mathrm{M}}_1$ estimates are highly consistent with independent ground-based ${\mathrm{P}\mathrm{M}}_1$ measurements (slope: 0.97, R²: 0.78). The estimated $1-\mathrm{\sigma }$ uncertainties of annual mean ${\mathrm{P}\mathrm{M}}_1$ for 25 years over >8 million land pixels are <20% for 98% of data points, while 0.3% of the CONUS population is associated with uncertainties >30% due to wildfires. Population-weighted mean (PWM) ${\mathrm{P}\mathrm{M}}_1$ across the CONUS of 10.8 μg/ m³ in 1998 decreased significantly (p < 0.05) at −0.23 μg/ m³/yr during 1998–2022, accounting for 86% of the reduction of PWM ${\mathrm{P}\mathrm{M}}_{2.5}$. Over 2006–2022, Native Americans have persistently breathed air with the lowest PWM ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ of 0.78 and the slowest PWM ${\mathrm{P}\mathrm{M}}_1$ reduction of −0.11 μg/ m³/yr, while Black communities have been exposed to the highest PWM ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ of 0.82 and the fastest PWM ${\mathrm{P}\mathrm{M}}_1$ reduction of −0.25 μg/ m³/yr. At present (2018–2022), the two less mitigable components (dust and sea salt) contribute to >20% of ${\mathrm{P}\mathrm{M}}_{2.5}$ for 19% (10%) of the Native community (the other racial groups) that are exposed to > 5 μg/m³ ${\mathrm{P}\mathrm{M}}_{2.5}$, while cases with >20% in ${\mathrm{P}\mathrm{M}}_1$ only account for 1.8% (2.3%) of the same population.

Interpretation

The dominance of ${\mathrm{P}\mathrm{M}}_1$ in ${\mathrm{P}\mathrm{M}}_{2.5}$ reduction and the decreasing ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio reflect the strong association of ${\mathrm{P}\mathrm{M}}_1$ with fossil fuel and other combustion sources and their responses to air quality regulations during the last 25 years. The gradual coarsening of ${\mathrm{P}\mathrm{M}}_{2.5}$ and the racial disparity of ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ call upon increasing urgency to separately assess health impacts of ${\mathrm{P}\mathrm{M}}_1$ vs. ${\mathrm{P}\mathrm{M}}_{2.5}$, as supported by the quality of the derived ${\mathrm{P}\mathrm{M}}_1$ estimates. Future particulate matter monitoring programs, health studies, and regulatory deliberations should consider ${\mathrm{P}\mathrm{M}}_1$ in addition to ${\mathrm{P}\mathrm{M}}_{2.5}$.

Funding

National Institute of Environmental Health Sciences, National Institutes of Health.

Introduction

Exposure to fine particulate matter (${\mathrm{P}\mathrm{M}}_{2.5}$, particles less than 2.5 μm in aerodynamic diameter, $D_p$) air pollution is associated with mortality from all-causes, ischemic heart disease, stroke, respiratory infection, type-2 diabetes, chronic obstructive pulmonary disease, and lung cancer (1–11). Both long-term and episodic exposures to ${\mathrm{P}\mathrm{M}}_{2.5}$ air pollution are leading environmental mortality risks in the United States (US) and worldwide (12–16). In 2021, exposure to ambient ${\mathrm{P}\mathrm{M}}_{2.5}$ was associated with 50,064 annual premature deaths in the US (17), with a large 95% uncertainty interval (24,664–78,690) mainly caused by uncertainty in the relationship of mortality risk with ${\mathrm{P}\mathrm{M}}_{2.5}$ exposure (18). For example, the associated hazard ratio of all-cause mortality and a 10 μg/m³ increase in long-term ${\mathrm{P}\mathrm{M}}_{2.5}$ exposure was reported as in the range of 1.08–1.16 among different studies in the US (16,19,20). Variations in the source (21,22), composition (23–25) and size distribution (24–27) within ${\mathrm{P}\mathrm{M}}_{2.5}$ have been hypothesized as potential causes of this wide range. Emerging estimates of global and regional ${\mathrm{P}\mathrm{M}}_{2.5}$ concentrations parsed into sources and chemical constituents (3,28–34) have been developed based on various geophysical (3,22,34) or data-driven (29,35) approaches, as well as a hybrid of both (30,36). Such datasets have been crucial to support investigation of the above possible factors diverging ${\mathrm{P}\mathrm{M}}_{2.5}$ health impacts, especially in the US where a wealth of cohort epidemiologic studies have been performed (21–23,37,38).

However, relatively less attention has been devoted to evaluating the possibly different health impacts of ${\mathrm{P}\mathrm{M}}_{2.5}$ in varying aerosol sizes. It has long been recognized that the fate and deposition of inhaled particles strongly depends on size, as the governing processes (e.g., Brownian diffusion, inertial impaction, interception, and gravitational settling) are each sensitive to different size regimes (39,40). Theoretical calculation indicates that particles with $D_p>2.5\mathrm{\mu }\mathrm{m}$ dominantly (>70%) deposit in the head/nose region of the respiratory tract, imposing less severe health impacts than smaller particles which are more likely to further penetrate and deposit in the tracheobronchial and alveolar regions (40). Stronger health impacts at smaller particle size are consistent with the scarcity of statistically significant association of ${\mathrm{P}\mathrm{M}}_{2.5–10}$ with mortality in epidemiological studies (41,42), and contributed to the US Environmental Protection Agency (EPA) legislation of a national ${\mathrm{P}\mathrm{M}}_{2.5}$ standard (on top of the existing ${\mathrm{P}\mathrm{M}}_{10}$ standard) in 1997. This size-cut of 2.5 μm, although a significant milestone, is nevertheless questionable as the optimal choice. One prevailing direction to further exploit PM impacts varied by size stems from the epidemiological evidence of higher hazard of ultrafine particles (i.e., PM_0.1) (43–46).

A related research topic is submicron PM (${\mathrm{P}\mathrm{M}}_1$) with $D_p<1\mathrm{\mu }\mathrm{m}$. ${\mathrm{P}\mathrm{M}}_1$ has a greater deposition fraction in the tracheobronchial and alveolar regions (40,46) and a larger surface area to mass ratio (47,48) than ${\mathrm{P}\mathrm{M}}_{2.5}$; both were hypothesized (with emerging evidences) to indicate stronger toxicity per unit mass. ${\mathrm{P}\mathrm{M}}_1$ is also more closely associated with the accumulation mode of the aerosol size distribution (49,50) and often dominates the observed signal in satellite aerosol optical depth and ground-based nephelometers due to greater mass extinction efficiency (51). Estimating ${\mathrm{P}\mathrm{M}}_1$ from a wealth of satellite and ground-based measurements is thus at least as promising as for the estimation of ${\mathrm{P}\mathrm{M}}_{2.5}$ (52). Furthermore, ${\mathrm{P}\mathrm{M}}_1$ is more modifiable than ${\mathrm{P}\mathrm{M}}_{2.5}$, as natural wind-blown dust (challenging to effectively mitigate) is a major source of ${\mathrm{P}\mathrm{M}}_{2.5}$ concentration in broad desert areas of the world, including the southwestern US (3,34,53). Emerging measurements indicate that >80% of dust ${\mathrm{P}\mathrm{M}}_{2.5}$ mass exists in the $D_p>1\mathrm{\mu }\mathrm{m}$ size regime (54–58), in contrast with other ${\mathrm{P}\mathrm{M}}_{2.5}$ components that are dominantly within $D_p<1\mathrm{\mu }\mathrm{m}$ (59–62). This reduced dust dominance and greater mitigation potential of ${\mathrm{P}\mathrm{M}}_1$ motivates further investigation of its own health impacts. Indeed, ${\mathrm{P}\mathrm{M}}_1$ has been reported to exhibit stronger association than ${\mathrm{P}\mathrm{M}}_{2.5}$ with lung cancer (63), cardiovascular diseases (64,65), respiratory diseases (66) and childhood pneumonia (67) in China. Notably, these findings are enabled collectively by emerging epidemiological studies, the nationwide ${\mathrm{P}\mathrm{M}}_1$ ground monitoring (>150 sites), and observationally validated gapless ${\mathrm{P}\mathrm{M}}_1$ estimates in China (52,61,62). However, information about long-term and nationwide ${\mathrm{P}\mathrm{M}}_1$ distribution in the US is not yet available due to a paucity of observational data, despite the affluence of available cohorts. This ${\mathrm{P}\mathrm{M}}_1$ data gap in the US severely limits the capability to evaluate the consistency and robustness of conclusions about ${\mathrm{P}\mathrm{M}}_1$ health impacts derived from other regions of the world.

Contrary to the scarcity of ${\mathrm{P}\mathrm{M}}_1$ observation data, long-term dedicated monitoring of ${\mathrm{P}\mathrm{M}}_{2.5}$ chemical composition across the contiguous US (CONUS) has been established for over two decades (68–70), which facilitated gapless estimates of major ${\mathrm{P}\mathrm{M}}_{2.5}$ components (30,35,71,72). As each component represents a unique size distribution, it is promising to derive ${\mathrm{P}\mathrm{M}}_1$ concentration maps across the CONUS based on these well-evaluated ${\mathrm{P}\mathrm{M}}_{2.5}$ component datasets. In this study, we develop biweekly estimates of gapless ambient ${\mathrm{P}\mathrm{M}}_1$ concentrations and its uncertainty at 1 km² (0.01° × 0.01°) resolution across the CONUS over 25 years, based on prior estimates of ${\mathrm{P}\mathrm{M}}_{2.5}$ chemical components, and robust constraints on size characteristics of each component. We also explore the long-term changes in population exposure to ${\mathrm{P}\mathrm{M}}_1$, in the ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass ratio, and in the associated racial disparity. The derived ${\mathrm{P}\mathrm{M}}_1$ estimates initialize possible assessment of health impacts of ${\mathrm{P}\mathrm{M}}_1$ vs. ${\mathrm{P}\mathrm{M}}_{2.5}$ in the US, and motivate future PM monitoring and regulation to consider ${\mathrm{P}\mathrm{M}}_1$ in addition to ${\mathrm{P}\mathrm{M}}_{2.5}$.

Methods

Compositional ${\mathrm{P}\mathrm{M}}_{2.5}$ estimates and uncertainties

We use biweekly estimates of ${\mathrm{P}\mathrm{M}}_{2.5}$ chemical components over 1998–2022, including sulfate, ammonium, nitrate, organic matter (OM), black carbon (BC), mineral dust (DUST), and sea-salt (SS) that were developed from a combination of satellite remote sensing, air quality modeling, and ground-based measurements (30,72). For each component, the concentration includes the associated aerosol water at 35% relative humidity (RH) for consistency with the US federal reference method for gravimetric analysis. Details of the dataset and pixelwise uncertainty are provided in Supplementary Section S1.

Characterize component-specific ${\mathrm{P}\mathrm{M}}_1$ mass fraction in ${\mathrm{P}\mathrm{M}}_{2.5}$

We collect daily mean concentrations of ${\mathrm{P}\mathrm{M}}_{2.5}$ and ${\mathrm{P}\mathrm{M}}_1$ chemical composition in the CONUS to characterize the component-specific ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ relationship for sulfate, ammonium, nitrate and OM (Table 1). We include the two available measurement types: 1) ${\mathrm{P}\mathrm{M}}_1$ components from Aerosol Mass Spectrometer (AMS) or Aerosol Chemical Speciation Monitor (ACSM) measurements near ${\mathrm{P}\mathrm{M}}_{2.5}$ components from the IMPROVE/CSN networks (within 50 km), and 2) Measurements of size- and chemically- resolved PM samples from Microorifice Uniform Deposit Impactor (MOUDI). Details of these two data sources are provided in Supplementary Section S2. Based on the co-located daily ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_{2.5}$, we perform linear regression (with forced intercept of zero) to estimate the ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio (e.g., the slopes in Figure 1). We estimate the $1-\mathrm{\sigma }$ uncertainty range of these ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratios (e.g., the shades in Figure 1) by incrementally adjusting the fitted values by ±0.01 until including ≥68.3% data points.

Table 1.

Summary of mass fraction of PM_2.5 with diameters <1 μm for different components.

Component	Fraction	1–σ Uncertainty	Source
Sulfate	0.908	0.197	Constrained from observations (Figure 1)
Ammonium	0.815	0.217
Nitrate	0.769	0.409
Organic matter	0.856	0.056
Black carbon	1.000	0.150	State-of-science understanding of soot (75)
Dust	0.122	0.022	Observationally constrained ambient dust size distribution after emission (58)
Sea salt (accumulation mode)	0.967	0.032	1–σ range based on size distribution in chemical transport models (73,74).

Figure 1.

Constraints on the mass fraction of ${\mathrm{P}\mathrm{M}}_{2.5}$ with diameters <1 μm for sulfate, ammonium, nitrate and organic aerosols. The blue points show the MOUDI-measured daily mass concentrations from multiple campaigns. The orange crosses are the co-located daily mean measurements of AMS/ACSM and IMPROVE/CSN. Each panel shows the fitting equation, expected $1-\mathrm{\sigma }$ uncertainty range (shaded), and coefficient of determination (R²) using both data sources (except for organic aerosols for which only the MOUDI data are used). The sample number and fraction within the $1-\mathrm{\sigma }$ uncertainty range are listed on the bottom-right for each data source.

For the remaining three components (BC, DUST and SS), we did not obtain observational information and assumed a ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio based on state-of-science understanding of their size characteristics in literature (Table 1). They are smaller contributors to ${\mathrm{P}\mathrm{M}}_{2.5}$ mass concentrations and population exposure in the CONUS (35,68–71) relative to the above four main components. Specifically, BC and SS (in accumulation mode) mass exhibit mainly in the submicron size range, with >96% of ${\mathrm{P}\mathrm{M}}_{2.5}$ mass in ${\mathrm{P}\mathrm{M}}_1$ (73–75); while wind-blown dust mass exists dominantly in the > 1 μm regime, with a ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ of 12% commensurate with recent observations (54–58).

${\mathrm{P}\mathrm{M}}_1$ Estimates and Uncertainty

We estimate ${\mathrm{P}\mathrm{M}}_1$ concentrations using Equation 1:

$$\left[{PM}_1\right]=\sum _S{k}_S\times \left[S\right]$$ (1).

[$S$] and $k_S$ are the ${\mathrm{P}\mathrm{M}}_{2.5}$ concentrations (from the biweekly estimates) and ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio (Table 1), respectively, for component $S$. This approach yields biweekly and spatially gapless ${\mathrm{P}\mathrm{M}}_1$ concentrations at 1 km² resolution for 1998–2022 across the CONUS.

We estimate the $1-\mathrm{\sigma }$ uncertainty of the ${\mathrm{P}\mathrm{M}}_1$ estimates according to the chain rule:

$${\sigma }_{PM1}^2=\sum _S{\sigma }_{k,S}^2\times [S{]}^2+{\sigma }_{[S]}^2\times k_S^2$$ (2).

${\sigma }_{\left[S\right]}$ and ${\sigma }_{k,S}$ are the $1-\mathrm{\sigma }$ uncertainty range of the concentrations (constrained according to Section S1) and of the ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio (Table 1), respectively, for component $S$.

We quantify the contribution to the $1-\mathrm{\sigma }$ uncertainty from each component by considering the uncertainties in both the concentrations and the ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio ($k_S$):

$${\sigma }_S^2(\mathrm{\%})=\frac{{\sigma }_{k,S}^2\times [S{]}^2+{\sigma }_{[S]}^2\times k_S^2}{{\sigma }_{PM1}^2}\times 100\mathrm{\%}$$ (3).

Similarly, the $1-\mathrm{\sigma }$ uncertainty of annual mean ${\mathrm{P}\mathrm{M}}_1$ concentration is estimated as:

$${\sigma }_{ann}^2=\frac{1}{n^2}\sum _{i=1}^n{\sigma }_{bw,i}^2$$ (4).

Here $n$ is the number ($n=26$) of biweekly estimates (each with $1-\mathrm{\sigma }$ uncertainty ${\sigma }_{bw,i}$) used to derive the annual mean ${\mathrm{P}\mathrm{M}}_1$.

Ground-based ${\mathrm{P}\mathrm{M}}_1$ data

We collect multiple sources of ground-based measurements of ${\mathrm{P}\mathrm{M}}_1$ concentrations (Table S3) to independently evaluate our estimates (a total of 202 biweekly measurements). We include data based on gravimetric analysis using two methods of particle size selection (MOUDI (76) and Sharp Cut Cyclone (77)), 4.5-year Tapered Element Oscillating Microbalance (TEOM) measurements in Spokane (78), and recent measurements in Indianapolis by the US EPA based on broadband spectroscopy using the Teledyne T640 PM Mass Monitor (79).

Population data

We use population count over the same 1 km² grids from the Gridded Population of the World (GPW v4) database (80), which is available every five years for 2000–2020. For each year in 1998–2022, we scale the GPW population distribution in the closest year by a constant factor to match the annual CONUS total population. For 2006–2022, we further collect five-year mean population divided for five racial groups (at census track level) from the National Historical Geographic Information System (81). We refer to the five racial categories in the data files with the following abbreviations throughout the paper: “Native” for “American Indian and Alaska Native”, “Black” for “Black or African American”, “White” for “White”, “Multiple” for “Two or More Races”, and “Asian&Other” for “Asian and Pacific Islander and Other Race”. For each race category, we allocate the five-year mean population count of each census track into 1 km² grids, based on the spatial distribution of the corresponding total population in GPW within the boundary of that census track.

Results

Observed ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ for PM components

Figure 1 shows the observationally constrained ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass ratio for four main PM components based on the daily measurements in Tables S1 and S2. The secondary inorganic components of sulfate, ammonium and nitrate PM exhibit a tight linear relationship (R² > 0.85) between their ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_{2.5}$ mass concentrations across ~2 orders of magnitudes. As expected, the data pairs of AMS/ACSM measurements versus nearby IMPROVE/CSN (orange) sites (within 50 km) are more scattered than the MOUDI data (blue) where ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_{2.5}$ are from the same MOUDI-collected aerosol samples. Stronger scatter is also observed at lower concentrations (e.g., < 1 μg/m³) for nitrate, reflecting the increasing uncertainty of the measurements at levels closer to the limits of detection. The fitted linear relationship (black line) crosses the center of scatter for both data sources, supporting the robustness of the derived relationships regardless of data acquisition methods. The fitted slopes indicate that 91% of sulfate ${\mathrm{P}\mathrm{M}}_{2.5}$ mass is within ${\mathrm{P}\mathrm{M}}_1$, while such mass ratio is 77% for nitrate. This difference is consistent with established observations and understanding of coarse-mode nitrate formation from heterogenous reaction of nitric acid or its precursors with DUST or SS (82,83). The coarse-mode nitrate is thermally more stable than, thus can compete with, the semi-volatile fine-mode nitrate, whereas for sulfate its fine-mode is thermally stable and ubiquitously dominant (84–86). Ammonium exhibits an intermediate ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio of 0.82, reflecting its association with both sulfate and nitrate. The MOUDI measurements of organic carbon (OC) sampled from the eastern and western US (Table S2) also exhibit a linear relationship (Figure 1, blue line), revealing a ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio of 0.86 for OM.

There are limited studies of component-specific ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ for the CONUS in existing literature. A recently emerging ACSM capability to measure ${\mathrm{P}\mathrm{M}}_{2.5}$ (87) enabled co-located observations of ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_{2.5}$ composition in the 2018 winter of Atlanta (88), revealing that 85% of nonrefractory ${\mathrm{P}\mathrm{M}}_{2.5}$ mass is from ${\mathrm{P}\mathrm{M}}_1$, and the ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ is relatively smaller for organics (78%) than for sulfate (97%). These values largely agree with our estimates (Table 1 and Figure 1). The regressed ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ for ammonium and nitrate in that study were somehow unphysical (>1.2), implying possible improvements needed for the characterization of capture efficiency of the instrument. We do not include any regional, seasonal, or RH-dependent variability in the component-specific ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ for the CONUS, as existing data samples (Figure 1) do not support such characterization. Such variability might exist, e.g., ACSM measurements in China indicated substantially enhanced (>40%) contribution from mass in the 1–2.5 μm size range to ${\mathrm{P}\mathrm{M}}_{2.5}$ during highly polluted and humid (RH > 90%) environments for these four main components (89). More measurements of PM size and composition in the future are warranted for a better characterization and constraints of ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ and ${\mathrm{P}\mathrm{M}}_1$ concentrations.

Evaluation using ground-based ${\mathrm{P}\mathrm{M}}_1$ measurements

Figure 2 shows the scatter plot of biweekly ${\mathrm{P}\mathrm{M}}_1$ estimates versus all co-located ground-based measurements (Table S3), for which at least five daily measurements are required to derive a biweekly average. Strong correlation (R² = 0.78) is observed between the 202 co-located biweekly ${\mathrm{P}\mathrm{M}}_1$ concentrations across a wide range of 4 to 30 μg/m³, with a linear slope of almost unity (0.97), a small negative intercept (−1.2 μg/m³), and 66% of the data pairs within an error envelope of ±21%±0.35 μg/m³. Of the four measurement methods, the two with gravimetric analysis (sharp cut cyclone and MOUDI) exhibit stronger correlations with the estimates (R² > 0.7) than the more indirect and less reliable methods (TEOM and broadband spectroscopy). Overall, a −15% normalized mean bias (NMB) is found for the estimates against all measurements, while the NMB against MOUDI measurements (−2%) is the smallest. These quantitative metrics indicate strong consistency, considering the differences in spatial and temporal sampling, and in various measurement protocols of the ground-based data. For example, the underestimation of TEOM measurements (Figure 2, green) may reflect biases in observations, since our ${\mathrm{P}\mathrm{M}}_{2.5}$ estimates exhibit opposite biases against the TEOM ${\mathrm{P}\mathrm{M}}_{2.5}$ at the same site (NMB = −14%, R² = 0.47, Figure S5) and against ${\mathrm{P}\mathrm{M}}_{2.5}$ measured by a nearby EPA monitor (~4.5 km to the south, NMB = 5.8%, R² = 0.93) for the same period. As the ${\mathrm{P}\mathrm{M}}_{2.5}$ component estimates have been comprehensively evaluated (e.g., Figure S2), this independent evaluation (i.e., the observation data is not used to derive the estimates) further confirms the reliability of the component-specific ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ (Table 1).

Figure 2.

Evaluation of estimated biweekly ${\mathrm{P}\mathrm{M}}_1$ against various sources of ground-based measurements (e.g., Table S3, orange: sharp cut cyclone (SCC); red: MOUDI; green: TEOM; blue: broadband spectroscopy (BB_SPEC)). The vertical error bars represent the estimated $1-\mathrm{\sigma }$ uncertainty of ${\mathrm{P}\mathrm{M}}_1$. The expected $1-\mathrm{\sigma }$ error range (EE, grey shaded), fraction of data within that EE, linearly fitted equation, coefficient of determination (R²), and normalized mean bias (NMB) against all measurements are shown in black. Number of data pairs (in the brackets), R² and NME against each source of data are shown in the corresponding colors.

We also compare 4.5 years of TEOM measured and estimated ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_{2.5}$ at the same site of Spokane, WA (Figure S5). Our biweekly estimates consistently underestimate both TEOM ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_{2.5}$, thus exhibit smaller biases (NMB = −6.8%, R² = 0.25) versus the TEOM-derived ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass ratio (Figure S5, lowest panel). The estimated ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ well tracks the biweekly variation of TEOM observations, and successfully captures several dust storms when the mass ratio reduces to <0.7 (78). This agreement indirectly supports the rationale for and confidence of the adopted ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ (Table 1), especially the ratio for DUST that has distinctly lower values than the other components. At annual scale, the estimated ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratios more closely resemble the TEOM observations (Figure S5, red).

${\mathrm{P}\mathrm{M}}_1$ distribution and uncertainties across the CONUS

Figure 3a shows the spatial distribution of annual mean ${\mathrm{P}\mathrm{M}}_1$ in 2022 from our estimates. Enhancements of ${\mathrm{P}\mathrm{M}}_1$ (e.g., > 5 μg/m³) are observed broadly over the eastern US as expected from the well-known distribution and density of PM sources, and sporadically over the western US from a combination of strong anthropogenic sources and wildfires. Populous urban areas (e.g., Figure S8) usually co-locate with such enhanced ${\mathrm{P}\mathrm{M}}_1$, leading to a CONUS-wide population-weighted mean (PWM) ${\mathrm{P}\mathrm{M}}_1$ concentration of 6.2 μg/m³.

Figure 3.

Annual mean estimates of ${\mathrm{P}\mathrm{M}}_1$ (a), its $1-\mathrm{\sigma }$ uncertainty normalized by the ${\mathrm{P}\mathrm{M}}_1$ concentration (b, log-scale), contribution from dust and sea salt to ${\mathrm{P}\mathrm{M}}_1$ (c) and ${\mathrm{P}\mathrm{M}}_{2.5}$ (d), and ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass fraction (f) for 2022. Panel (e) shows the distribution of normalized $1-\mathrm{\sigma }$ uncertainty (x-Axis, in log-scale) of all 1-km annual mean ${\mathrm{P}\mathrm{M}}_1$ estimates across 1998–2022. The green shading represents the pixel number fractions in each bin, while the red line indicates the population fractions.

Figure 3b shows the spatial distribution of $1-\mathrm{\sigma }$ uncertainty of ${\mathrm{P}\mathrm{M}}_1$, featuring relatively smaller values (<6%) over the east and populated areas in the west (e.g., Los Angeles and Phoenix), and larger uncertainties (>30%) over the remote west. This sharp difference in the ${\mathrm{P}\mathrm{M}}_1$ uncertainties associated with enhanced ${\mathrm{P}\mathrm{M}}_1$ is due to the stronger dominance (>60%) of wildfire OM (e.g., Figure S6d) over the remote western US (13,90,91). According to the cross validation (Figure S2), OM has greater expected error (EE) at high concentrations (e.g., > 10 μg/m³) in the biweekly data records. Observed and estimated biweekly OM can often reach > 30 μg/m³ during wildfire episodes, leading to the large $1-\mathrm{\sigma }$ uncertainties for high ${\mathrm{P}\mathrm{M}}_1$ in the remote west (Figure 3b) that are dominantly from OM (Figures S6h). While for the other locations (e.g., the eastern US) with enhanced ${\mathrm{P}\mathrm{M}}_1$, secondary inorganics (sulfate, ammonium and nitrate) are increasingly dominant (68,71) as in Figure S6a–c. These components exhibit decreasing uncertainty with higher concentrations due to the linearity of the EE equations (Figure S2), where contribution from the absolute term decreases at higher concentrations. Even over the southeastern US with enhanced OM dominance in ${\mathrm{P}\mathrm{M}}_1$ and its uncertainty (e.g., Figures 6d and 6h), OM is mainly secondarily formed (68,92) in the warm season and less wildfire-relevant, with biweekly concentrations < 10 μg/m³ that lead to <10% overall uncertainties in annual ${\mathrm{P}\mathrm{M}}_1$ (Figure 3b). Over locations with low annual ${\mathrm{P}\mathrm{M}}_1$ (< 5 μg/m³), the $1-\mathrm{\sigma }$ uncertainty is usually 10–20% (Figure 3b) driven by absolute error terms in the EE equations (Figure S2). Figure S7 shows that in the biweekly estimates, high $1-\mathrm{\sigma }$ uncertainties (>50%) also occur with either low ${\mathrm{P}\mathrm{M}}_1$ (e.g., across the remote west in Day 141–154) or wildfire events (e.g., over the northwestern states in Day 211–224 and Day 281–294), which are reduced in Figure 3b after annual averaging (Equation 4).

Although the cross validations of DUST and SS reveal their greater uncertainties (>40%, e.g., Figure S2) than inorganic aerosols, Figure 3c shows that these two components together make <10% contribution to the estimated ${\mathrm{P}\mathrm{M}}_1$ except for the arid or coastal regions (see also Figure S6f–g). Therefore, $1-\mathrm{\sigma }$ uncertainties of annual ${\mathrm{P}\mathrm{M}}_1$ do not exhibit as significant enhancements over DUST- or SS-abundant regions, as over wildfire-impacted regions (Figure 3b). The PWM DUST (0.08 μg/m³) and SS (0.27 μg/m³) in ${\mathrm{P}\mathrm{M}}_1$ for 2022 are the lowest among the seven components. Such <6% contribution of these two dominantly natural components to PWM ${\mathrm{P}\mathrm{M}}_1$ is favorable with respect to mitigation. Nevertheless, for ${\mathrm{P}\mathrm{M}}_{2.5}$ the combined PWM fraction from DUST and SS (Figure 3d) more than doubles (12.5%), indicating the influence of these components on ${\mathrm{P}\mathrm{M}}_{2.5}$ exposure and the challenge of its regulation.

Figure 3e (green) reveals that 84% (98%) of all the estimated annual mean ${\mathrm{P}\mathrm{M}}_1$ for 1998–2022 and >8 million land pixels have $1-\mathrm{\sigma }$ uncertainties <10% (<20%). Wildfire-contaminated cases with >30% uncertainties account for 1% of all the annual records, and this contribution is reduced to 0.3% when considering the affected population (Figure 3e, red). In the biweekly estimates, the CONUS population with >30% (>50%) $1-\mathrm{\sigma }$ uncertainty of ${\mathrm{P}\mathrm{M}}_1$ increases from 9% (1%) in 1998 to 32% (4%) in 2022 (not shown), reflecting the reduced ${\mathrm{P}\mathrm{M}}_1$ levels as well as the increasing wildfire impacts.

Figure 3f shows the ratios of annual mean ${\mathrm{P}\mathrm{M}}_1$ vs. ${\mathrm{P}\mathrm{M}}_{2.5}$ for 2022, which reflect the dominance of different components with their characteristic ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ in Table 1. The highest ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratios (up to 0.95) appear over both coasts and the southeast, with strong contributions (Figures S6a and S6g) from sulfate (${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}=0.91$) and SS (${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}=0.97$). The ratios are reduced to 0.75–0.85 over the Midwest, Greater Los Angeles, and the Salt Lake City metropolitan area, where nitrate (${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}=0.77$) becomes more important in ${\mathrm{P}\mathrm{M}}_1$ (Figure S6c). Values of <0.7 are observed mainly in the arid southwest (down to 0.5), corresponding to enhanced fractions of DUST (${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}=0.12$) (Figures 3c and S6f).

Changes in ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$

Figure 4a shows the 25-year trends in ${\mathrm{P}\mathrm{M}}_1$ based on least-square linear fits of annual mean ${\mathrm{P}\mathrm{M}}_1$ for each CONUS pixel. We also tested the non-parametric Mann-Kendall fits (93) which yielded consistent results that did not alter any conclusions to be discussed in the next section. Significant reductions (p < 0.05) greater than −0.1 μg/m³/yr are observed over populous areas where the $1-\mathrm{\sigma }$ uncertainty of annual ${\mathrm{P}\mathrm{M}}_1$ concentrations are also lower (Figure 3b), benefiting 85% of the CONUS population. In a relative sense, 85% of the CONUS population also experienced significant ${\mathrm{P}\mathrm{M}}_1$ reductions greater than 36% (−1.5%/yr, normalized by 25-year mean ${\mathrm{P}\mathrm{M}}_1$) during this period; while this population fraction reduces to 60% for the second half of the period (2010–2022, not shown), driven by the increasing impacts of wildfire, as was similarly reported for ${\mathrm{P}\mathrm{M}}_{2.5}$ (13,94). Only 0.3% of the population experienced significant and positive trends for 1998–2022, which increases to 0.5% for 2010–2022.

Figure 4.

(a) Map of ${\mathrm{P}\mathrm{M}}_1$ trends over 1998–2022, with insignificant trends (p ≥ 0.05) indicated by more transparent colors. (b) Normalized population distribution (i.e., the area under each curve equals 1) as a function of these trends for five racial groups (color-coded). (c) Normalized population distribution as a function of ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio for five racial groups (consistently color-coded as in Panel b) for 2006–2010 (more transparent) and 2018–2022 (less transparent). The fractions of population with ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}<0.8$ for each group and five-year period are displayed on the upper-left.

Figure 4b reveals distinct disparity among racial groups regarding benefits of these 25-year ${\mathrm{P}\mathrm{M}}_1$ reductions. The Black, White, Asian&Other, Multiple and Native communities each have 96%, 84%, 82%, 81% and 58%, respectively, of its population (2006–2022 mean, Figure S8) that experienced significant (p < 0.05) ${\mathrm{P}\mathrm{M}}_1$ reductions greater than −0.1 μg/m³/yr. The two largest benefits for the Black and White communities are due largely to their greater population residing over the eastern US than the other racial groups (Figure S8), a broad region with stronger ${\mathrm{P}\mathrm{M}}_1$ reductions nationwide (Figure 4a), forming the population peaks with ${\mathrm{P}\mathrm{M}}_1$ trends <−0.3 μg/m³/yr in Figure 4b for both communities. The Black community has the lowest population fraction in the remote west, and less population in suburban/remote areas of the northeast relative to the White community (Figure S8), thus exhibits the narrowest distribution of population vs. ${\mathrm{P}\mathrm{M}}_1$ trends (Figure 4b). Oppositely, the Native community has least benefited from the ${\mathrm{P}\mathrm{M}}_1$ reductions, with a unique population peak around zero trends and 42% of its population experiencing trends > −0.1 μg/m³/yr or insignificant (Figure 4b), driven by the more population in the remote west (Figure S8) than the other communities, where such trends prevail (Figure 4a). The Asian&Other and Multiple communities also benefit slightly less than the White community, with relatively less population over the east while more population over large metropoles in the west coast (e.g., San Francisco and Seattle, Figure S8), where recent increases in wildfire diminished PM regulation efforts.

Population-weighted mean (PWM) ${\mathrm{P}\mathrm{M}}_1$ across the CONUS reduced (±95% confidence interval) significantly (p < 0.05) at −0.23±0.03 μg/m³/yr (−2.7±0.4%/yr) over 1998–2022, accounting for 86% of the −0.26±0.04 μg/m³/yr trend of PWM ${\mathrm{P}\mathrm{M}}_{2.5}$. For all CONUS population and for the five racial communities, the 5-year running PWM ${\mathrm{P}\mathrm{M}}_1$ since 2006 exhibit continuous reductions with a recent stalling since around 2015–2019 (Figure S9, circles). Based on the 5-year PWM ${\mathrm{P}\mathrm{M}}_1$, Native Americans experienced the slowest reduction of −0.11±0.05 μg/ m³/yr, while Black communities benefited from the fastest reduction of −0.25±0.05 μg/ m³/yr (Figure S9). The Asian&Other and Native communities suffered the most from the increasing wildfire-related stalling of reductions due to the previously mentioned characteristics of their population distributions (Figure S8). The Asian&Other community endure the greatest ${\mathrm{P}\mathrm{M}}_1$ exposure at present (2018–2022, Figure S9) among the five racial groups, with large gaps (>0.5 μg/m³) vs. the other communities.

The ${\mathrm{P}\mathrm{M}}_1$ data unveil a continuous coarsening of ${\mathrm{P}\mathrm{M}}_{2.5}$ that the CONUS population is exposed to, as depicted by Figure 4c showing the changes of five-year mean population distribution vs. ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ from 2006–2010 (lighter colors) to 2018–2022 (darker colors). Driven by the dominance of ${\mathrm{P}\mathrm{M}}_1$ in the historical ${\mathrm{P}\mathrm{M}}_{2.5}$ reductions across the CONUS (i.e., ${\mathrm{P}\mathrm{M}}_1$ reduced at faster paces than ${\mathrm{P}\mathrm{M}}_{2.5}$), the population vs. ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ histograms shifted leftward and widened for all racial groups. The fraction of population exposed to relatively coarser particles (e.g., ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}<0.8$) is the highest for the Native community (with PWM ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ of 0.78 for 2006–2022) and the lowest for the Black community (2006–2022 PWM ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ of 0.82), which has increased in all racial groups. Five-year running PWM ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass ratios also decreased (Figure S7, diamonds) continuously and exhibited high correlations (R² > 0.7) vs. the corresponding time series of PWM ${\mathrm{P}\mathrm{M}}_1$ concentrations for the five racial groups. For all population, annual PWM ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ decreased significantly (p < 0.05) at 0.11±0.03 percentage/yr, from 0.83 in 1998 to 0.81 in 2022.

Following such coarsening of ${\mathrm{P}\mathrm{M}}_{2.5}$ driven by ${\mathrm{P}\mathrm{M}}_1$ reductions, the challenge for ${\mathrm{P}\mathrm{M}}_{2.5}$ regulation from the DUST and SS components that were dominantly from natural sources has also increased, especially for the Native community. With the current mean ${\mathrm{P}\mathrm{M}}_{2.5}$ levels over 2018–2022, the majority (88% and 97%, respectively) of population in the Native and the other four communities are exposed to ${\mathrm{P}\mathrm{M}}_{2.5}$ higher than the WHO guideline of 5 μg/m³ (Figure S10, upper). For these population with nonattainment to such global ${\mathrm{P}\mathrm{M}}_{2.5}$ standard, 19% in the Native community and 10% in the other racial groups also associate with >20% of ${\mathrm{P}\mathrm{M}}_{2.5}$ mass as DUST or SS (Figure S10, lower), increasing from 14% and 6.5%, respectively, for 1998–2002. If ${\mathrm{P}\mathrm{M}}_1$ were instead the objective to mitigate, DUST and SS at present together contribute >20% of ${\mathrm{P}\mathrm{M}}_1$ for only 1.8% and 2.3% of the same population of the Native and other communities, respectively. Therefore, studies of independent health impacts from ${\mathrm{P}\mathrm{M}}_1$ (as supported by our new estimates) might imply significant changes in the effectiveness of future PM regulation policy (that will dominantly mitigate components other than DUST and SS) in the CONUS.

Discussion

Our biweekly and annual estimates provide the first long-term, high-resolution, gapless dataset of the spatial and temporal distribution of CONUS-wide ${\mathrm{P}\mathrm{M}}_1$ over 1998–2022. This dataset addresses the critical research needs to evaluate the possibly stronger toxicity of ${\mathrm{P}\mathrm{M}}_1$, which is more anthropogenically modifiable than ${\mathrm{P}\mathrm{M}}_{2.5}$. We used information about ${\mathrm{P}\mathrm{M}}_{2.5}$ chemical components, their corresponding ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass ratios, and uncertainties of these parameters, to obtain such estimates and constrain their uncertainties. Our component-driven estimates overcome the scarcity of direct ${\mathrm{P}\mathrm{M}}_1$ measurements to yield significant agreements with 202 biweekly ground-based ${\mathrm{P}\mathrm{M}}_1$ data (Figure 2) that span major urban areas across the CONUS (Tables S2 and S3). Additionally, our estimates exhibit spatiotemporal variability of ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ that is consistent with PM composition (e.g., Figures 3f and S5).

Our observationally constrained ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass ratios for four major PM components (Figure 1) are representative of co-located observations from multiple locations and observing times (Tables S1 and S2). Such representativeness is particularly strong for the secondary inorganic aerosols, while ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ information for organics are relatively sparse with MOUDI samples from two metropolitan areas (Table S2). An emerging opportunity to improve the characterization of these component-specific ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ (and consequently improve the ${\mathrm{P}\mathrm{M}}_1$ estimates) is the Atmospheric Science and Chemistry Measurement Network (95), which provide long-term and continuous monitoring of chemically speciated PM concentrations by ACSM and of aerosol size distribution by Scanning Mobility Particle Sizer, in major urban areas of the US.

The pixelwise uncertainty estimates of the ${\mathrm{P}\mathrm{M}}_1$ products objectively reflect the various sources of uncertainties and their spatiotemporal distribution. The data quality (84% of all annual records with <10% $1-\mathrm{\sigma }$ uncertainties) well support applications of our estimates in epidemiology studies of health impacts of ${\mathrm{P}\mathrm{M}}_1$ exposure, which can be jointly considered during such assessments. High $1-\mathrm{\sigma }$ uncertainties (e.g., >50% for biweekly estimates and >30% for annual means) are mainly due to episodic high OM concentrations in wildfires, for which the uncertainty of OM estimates increases quadratically. We recommend screening pixels with such high uncertainties (verified to have limited relevance with population exposure, e.g., Figure 3e) before applications in health impact assessments. More measurements of ${\mathrm{P}\mathrm{M}}_1$ and its components during wildfire episodes are needed to additionally evaluate and constrain ${\mathrm{P}\mathrm{M}}_1$ estimates from wildfire smoke, that are increasingly pervasive in PM exposure for the US residents (13,90,91,94).

Over 1998–2022, we found significant reductions in both ${\mathrm{P}\mathrm{M}}_1$ concentrations and the ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ mass ratio, as well as distinctive racial disparities of these reductions. The dominance of ${\mathrm{P}\mathrm{M}}_1$ in ${\mathrm{P}\mathrm{M}}_{2.5}$ reduction and the decreasing ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio reflect the strong association of ${\mathrm{P}\mathrm{M}}_1$ with fossil fuel and other combustion sources and their responses to air quality regulations during the last 25 years. The revealed coarsening of particles led to enhanced relevance of the less mitigable DUST and SS in ${\mathrm{P}\mathrm{M}}_{2.5}$ exposure and regulation at present (especially for the Native community), while such unfavorable impacts are massively reduced for ${\mathrm{P}\mathrm{M}}_1$ (e.g., Figures 3c, 3d and S10). Effectiveness of PM reduction in the future thus might substantially differ by considering ${\mathrm{P}\mathrm{M}}_1$ or ${\mathrm{P}\mathrm{M}}_{2.5}$, calling upon increasing urgency to separately assess their health impacts. Such studies are well supported by the quality of our ${\mathrm{P}\mathrm{M}}_1$ estimates, and their outcomes about independent health impacts of ${\mathrm{P}\mathrm{M}}_1$ will attract more attention and efforts to broadly monitor ${\mathrm{P}\mathrm{M}}_1$ concentrations and properties in the US and worldwide, in addition to the existing monitoring and regulation program of ${\mathrm{P}\mathrm{M}}_{2.5}$. These enhanced monitoring and observational information will in turn improve and extend ${\mathrm{P}\mathrm{M}}_1$ estimation, e.g., facilitating an effective balancing between information from geophysics-driven modeling and from data-driven machine learning (96). Our estimates embark upon such a positive feedback loop of advancing ${\mathrm{P}\mathrm{M}}_1$ monitoring and estimates, health studies, and regulatory deliberations.

Research in context.

Evidence before this study

Data of ground-level ${\mathrm{P}\mathrm{M}}_1$ concentrations is fundamental to assessing how its health risk might differ from ${\mathrm{P}\mathrm{M}}_{2.5}$. However, such a long-term and gapless dataset for the CONUS does not exist. We searched PubMed and Google Scholar for measurement or estimates of ${\mathrm{P}\mathrm{M}}_1$ in the US, using the search terms “submicron aerosol”, “PM1”, “United States”, for articles published between Jan 1, 1995, and August 31, 2024, in English. We found that most previous studies reported measurements of ${\mathrm{P}\mathrm{M}}_1$ during field campaigns over limited locations and usually lasting < 2 months. The longest ${\mathrm{P}\mathrm{M}}_1$ mass concentration data record we found was in Spokane, WA (1995–2002). Few articles have reported perspective about nationwide and decadal ${\mathrm{P}\mathrm{M}}_1$ distribution.

Added value of this study

Our study for the first time provides the CONUS-wide and gapless perspective on the spatiotemporal distribution and uncertainties of ${\mathrm{P}\mathrm{M}}_1$ concentrations for 25 years (1998–2022), and assesses the trends and racial disparities in the 25-year ${\mathrm{P}\mathrm{M}}_1$ and ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$ ratio. Our approach takes advantage of a well-evaluated and long-term estimates of ${\mathrm{P}\mathrm{M}}_{2.5}$ chemical components, and knowledge of component-specific ${\mathrm{P}\mathrm{M}}_1/{\mathrm{P}\mathrm{M}}_{2.5}$. In addition, we developed an uncertainty characterization of the biweekly and annual ${\mathrm{P}\mathrm{M}}_1$ estimates at pixel (1 km²) level with careful consideration of errors in each source of our estimates.

Implications of all the available evidence

With strong agreement vs. observed ${\mathrm{P}\mathrm{M}}_1$ during independent evaluations, our ${\mathrm{P}\mathrm{M}}_1$ estimates uniquely enable independent assessments of ${\mathrm{P}\mathrm{M}}_1$ health impacts using heritage epidemiology studies. Our pixelwise uncertainty estimates further support such good data quality and are also valuable in statistical analyses during the health risk assessments. Our study also novelly reveals the significant racial disparity of ${\mathrm{P}\mathrm{M}}_1$ exposure and its 25-year reductions in the CONUS. We identify the enhanced relevance of dust and sea salt in PM air pollution at present when considering ${\mathrm{P}\mathrm{M}}_{2.5}$ while not for ${\mathrm{P}\mathrm{M}}_1$, indicating increasing urgency to evaluate their independent health risks.

Data sharing

The biweekly and annual mean ${\mathrm{P}\mathrm{M}}_1$ data and $1-\mathrm{\sigma }$ uncertainties will be openly available in the Atmospheric Composition Analysis Group (https://sites.wustl.edu/acag/datasets/) upon acceptance of the manuscript.