Aerosol Thermodynamics: Nitrate Loss from Regulatory PM2.5 Filters in California

Fine particulate matter (PM2.5) mass concentrations reported by regulatory networks are declining across the United States. It is well established that ammonium nitrate contributes substantially to the PM2.5 mass in the western United States, and that Teflon filters commonly used by regulatory monitors are subject to negative mass artifacts due to ammonium nitrate volatilization. This study focuses on the San Joaquin Valley (SJV), an environmental justice (EJ) and agricultural region with persistently poor air quality. The SJV is a serious nonattainment area of PM2.5 National Ambient Air Quality Standards (NAAQS) with substantial nitrate mass concentrations. We explicitly model the chemical thermodynamic equilibrium of the ammonium nitrate–nitric acid systems and quantify volatilization across California as a function of the deliquescence point relative humidity (%DRH). Nitrate loss is estimated at all federal reference method (FRM) and federal equivalent method (FEM) monitors from 2001 to 2021. Nearly 20% of PM2.5 mass is lost from filters in the SJV area, especially during winter and fall when particulate nitrate mass is most abundant. All decadal PM2.5 trends calculated from reported measurements in Kern, Tulare, and Fresno counties in the SJV show greater decline in PM2.5 mass when nitrate loss is accounted for, up to a factor of 20 in Kern county. This suggests PM2.5 mass concentrations reported in regulatory networks are biased low relative to the actual atmospheric burden, notably in an EJ area that lags behind most of the country’s air quality improvements.


■ INTRODUCTION
−6 State and local entities measure surface mass concentrations of PM 2.5 as part of an EPA network with federal reference and federal equivalent methods (FRM and FEM).USEPA's chemical speciation network (CSN), implemented in 2001, chemically characterizes PM 2.5 and is a helpful tool to quantitatively understand sources and associations among ambient PM 2.5 chemical composition, health end points, and visibility degradation in urban areas. 7Since the introduction of the CSN in 2001 through 2021, PM 2.5 mass concentrations have declined by 37% across the United States. 8−11 These pollutants are also precursors of ammonium sulfate and ammonium nitrate and are substantial contributors to ambient PM 2.5 . 12−18 PM 2.5 chemical constituents have different physicochemical properties.Both ammonium sulfate and ammonium nitrate are hygroscopic and promote water uptake.−26 This is primarily due to changes in the ambient temperature and the pressure drop across the filter during sampling.−32 Nitrate loss is documented in the literature for PTFE Teflon filters, 7,23−29,33−36 glass fiber filter tape, 22,37,38 and Teflon-coated borosilicate glass fiber. 39,40Glass fiber filters used in the Met One BAM-1020 FEM are estimated to have greater ammonium nitrate volatilization loss than from Teflon filters. 22,37Studies demonstrate that this FEM monitor exhibits both positive and negative artifacts for PM 2.5 mass due to temperature, liquid water, and acidic gas absorption. 41The FEM tapered element oscillating microbalance with a filter dynamic measurement system (TEOM-FDMS) employs a Teflon-coated glass fiber filter and accelerates nitrate loss due to the heating of filters during sampling. 39,40ll measurements are operationally defined, and EPA defines PM 2.5 measurements in Title 40 of the Code of Federal Regulations (CFR 40) Part 50 Appendix L that specifies that the FRM employs polytetrafluoroethylene (PTFE) Teflon filters (46.2 mm diameter ± 0.25 mm, 2 μm pore size, 30−50 μm filter thickness).Ionic PM 2.5 constituents, such as sulfates and nitrates, are separately measured as part of the CSN with nylon filters and nitric acid denuders (Table S1). 27,29,42,43eflon filters are known to be susceptible to negative sampling artifacts for nitrate. 23,29,33,35Ammonium nitrate vapor pressure increases with temperature and drives evaporation, relative to the partial pressure in the atmosphere.Volatilization from Teflon filters can be quantified from the ambient temperature, amount of particulate nitrate, PM 2.5 mass, and understanding of relative humidity (RH).In comparison, nylon filters are less prone to ammonium nitrate volatilization and measure nitrate concentrations that more accurately describe ambient concentrations as nylon filters retain nitrate in the form of nitric acid well. 27,28,34−46 State of the art PM 2.5 measurement techniques such as the Time-of-Flight Aerosol Chemical Speciation Monitor (ToF-ACSM) are in good agreement with FRM/FEM, however, previous literature suggests that particles that undergo drying in inlets of aerosol mass spectrometers (AMS or Q-AMS) are also subject to losses during sampling. 47,48Moreover, the presence of organic compounds can interfere with nitrate detection using the Q-AMS and standard ACSM. 49High-resolution AMS and ACSM are not subject to these interferences. 50,51n this work, we focus on the fundamental equilibrium chemistry that drives the evaporative losses of nitrate in PM 2.5 from filter-based methods.For condensed phase ammonium nitrate in equilibrium with gaseous nitric acid and ammonia, the ratio between the gaseous and condensed phase compounds is defined by the dissociation constant, K amb,i .The dissociation constant is dependent on the deliquescence relative humidity (%DRH), which is a function of ambient temperature.%DRH for ammonium nitrate is 62% at 25°C. 45,52,53Previous research in arid locations to estimate nitrate loss from Teflon filters used in regulatory networks employs a constant factor and does not account for the change in K amb,i above the %DRH. 27,29Constant factors obscure accurate understanding of nitrate loss and the true ambient PM 2.5 burden, and subsequently, the effectiveness of control strategies in different communities and associations with health end points.Previous research demonstrates that PM 2.5 is a function of liquid water and that loss from filters has a RH dependence; therefore, we consider the aerosol phase state to account for different dissociation constant values at ambient temperature during each hour when above and below %DRH conditions. 35,36,38,45he top three most polluted cities in the U.S. for year-round particle pollution in 2022 are Bakersfield, Fresno, and Visalia, located in Kern, Fresno, and Tulare counties, respectively. 54All three counties are a part of the SJV in central California, where nitrate is a dominant constituent of PM 2.5 . 47,55The SJV is a critically important and highly productive agricultural region in the United States, especially for the dairy and cattle industries 56 that are major sources of the area's particulate nitrate.California defines most of the SJV as disadvantaged communities (SB 535). 57The annual average PM 2.5 concentrations for Kern, Fresno and Tulare counties between 2018− 2020 are 17, 15, and 16 μg m −3 respectively, and all exceed the annual PM 2.5 NAAQS that was 15 μg m −3 from 1997−2012 and 12 μg m −3 since 2012. 1,58PM 2.5 concentrations in counties of the SJV area consistently exceed NAAQS standards and filter measurements are prone to ammonium nitrate losses during measurement in this dry and hot climate. 59There is no clear strategy to improve air quality in this region.A recent EPA proposed rule plans to reject the California State Implementation Plan (SIP) for the SJV to attain the PM 2.5 NAAQS, in part due to uncertainty related to agricultural emissions that form particulate nitrate. 60There is currently no approved plan to bring SJV into attainment.Previous studies demonstrate that across the U.S., PM 2.5 ammonium nitrate volatilization is the greatest in California. 28Therefore, we investigate how reported PM 2.5 mass concentration may be biased low in the San Joaquin Valley due to substantial losses of a dominant chemical constituent.PM 2.5 is obtained from the FRM/FEM, while sulfate and nitrate mass concentrations are obtained from the CSN.We analyze annual average and annual seasonal average trends, where winter consists of January, February, and December months of the same year, spring consists of March, April, and May months, summer consists of June, July, and August months, and fall consists of September, October, and November months.Table 1 shows the types of filters used in FRM/FEM samplers and included in this analysis where nitrate loss is documented in the literature.
There is specific focus on data obtained from 3 CSN sites and 7 FRM/FEM sites in Kern County and Fresno County, 3 CSN and 2 FRM/FEM sites in Tulare county, and 3 CSN sites and 14 FRM/FEM sites in Los Angeles County.It should be noted that the CSN provides reconstructed PM 2.5 mass, which varies in comparison with FRM/FEM-measured PM 2.5 . 62Site locations analyzed here are provided in Figure S2.PM 2.5 and its chemical constituents are typically measured every 1-in-3 days; some sites sample with daily or every 1-in-6 days frequency.All reported negative concentration values are set to zero, and this may indicate concentrations measured near or below the limit of detection.Hourly surface temperature and relative humidity are obtained from USEPA meteorological data from weather stations across the contiguous United States (CONUS) and aggregated, by mean and median respectively, for each site monitor location.
Teflon filter-collected nitrate losses are calculated based on the thermodynamic equilibrium between the vapor pressure product of ammonium nitrate (particulate nitrate, pNO 3 ) and the product of gas-phase nitric acid (HNO 3 ) and ammonia (NH 3 ) partial pressures (eq 1): 27,28,53 s g g NH NO ( ) HNO ( ) NH ( ) Hering, Cass, and colleagues applied eq 2 to ambient samples to calculate the daily average nitrate losses (ΔNO 3 ) in micrograms per cubic meter (μg m −3 ) from the conversion of thermodynamic equilibrium constants (eq 2).Assumptions include that ammonia and nitric acid are formed in equal parts from the dissociation of ammonium nitrate, and conditions were below the deliquescence point of NH 4 NO 3 .
The reference temperature, T ref , is the daily mean temperature calculated from each ith hour value.K amb,i is the dissociation constant calculated at ambient temperature at each hour, T i , from eq 3. It is important to consider the changes in K amb,i throughout the day because volatilization is a function of temperature and RH, which also varies throughout the day (Figure S3).The partial pressure of gas-phase nitric acid in nanobars (nb) is converted to particle-phase nitrate concentrations (μg m −3 ) with the conversion factor of 745.7 for T ref . 27,53Below %DRH, the thermodynamic equilibrium dissociation constant, K amb,i (nb 2 ) is calculated from a Van't Hoff equation derived by Mozurkewich (eq 3): 53

K
T T ln 118.87 24084 6.025 ln The dissociation constant derived from eq 3 is applicable to the volatilization of ammonium nitrate below the deliquescence point.
The chemistry that describes the process of ammonium nitrate volatilization depends on whether conditions are above or below the deliquescence point.To determine if conditions are above or below the deliquescence point, we employ eq 4, which is taken from Tang and Munkelwitz: 52 The reference temperature (T*) in eq 4 is 298 K, and the thermodynamic and solubility coefficient values for NH 4 NO 3 are listed in Table 2. ΔH s is the integral heat of solution in joules per mole (J mol −1 ), R is the gas constant (R = 8.3145 J mol −1 K −1 ), and A, B, and C are the thermodynamic coefficients that relate solubility to temperature for a single salt solution.Ambient aerosol is a complex mixture, and this simple treatment introduces uncertainty that is difficult to explicitly quantify.When relative humidity is above the deliquescence point of ammonium nitrate, the equilibrium chemistry is more complex 45 because one must account for acid/base dissociation and the dissociation constant to describe partitioning.K amb,i in eq 3 is replaced with K amb,i * to reflect conditions above %DRH: 53 where a i is assumed to be the relative humidity as a fraction, 63 We evaluate and aggregate hourly ambient temperature (T i ) and RH for the nearest FRM/FEM and CSN monitors for the period investigated here to determine if ambient aerosol is above or below the %DRH (eq 4).Individual hours are analyzed for %DRH calculations, and we present the average prevalence for the period of each monitoring location for reference (Figure S4).For the purposes of this study, we assume that (1) particulate nitrate (pNO 3 ) measured on nylon filters is the most accurate, as is consistent with previous literature, 34 and (2) nitrate volatilizes from ammonium nitrate and does not include volatilization from other forms of semivolatile compounds, including organic nitrates.In cases where calculated ΔNO 3 exceeds the measured pNO 3 , it is assumed that ΔNO 3 = pNO 3 .This constraint ensures that calculations do not estimate more pNO 3 on Teflon filters than what would be measured by nylon filters. 29We exclude loss of semivolatile organic compounds 64 in these calculations, and our estimates of ΔNO 3 are likely a lower bound estimate of the total negative sampling artifacts for PM 2.5 collected with FRM and FEM monitors.The fraction of PM 2.5 mass lost by nitrate volatilization (PM 2.5,ΔNOd 3 ) is calculated with eq 9, where PM 2.5 mass concentration is obtained from the 24 h FRM: To test for significance in trends of PM 2.5 , nitrate, and sulfate mass concentrations, we use the Mann Kendall test, a nonparametric statistical test that provides an unbiased Sen slope and is robust against outliers and large data gaps (Table 3).
■ RESULTS AND DISCUSSION PM 2.5 mass volatilization from surface air quality monitors occurs ubiquitously across California (ΔNO 3 > 0; Figure 1).The percentage mass of PM 2.5 volatilized in Teflon filters is greatest in southern California, consistent with previous findings (Figure 1). 28In regions with the highest year-round PM 2.5 mass where nitrate is a dominant chemical constituent (Figure S5) and aerosol is most often under conditions below %DRH (Figure S4), total mass loss is greatest, most notably in winter.We estimate that the monitor in Rubidoux, Riverside County, CA that has an annual average PM 2.5 concentration of 16.8 μg m −3 , loses 20.6% of PM 2.5 mass from Teflon filter measurements used in FRM/FEM monitors, suggesting that the true average ambient burden in Rubidoux is closer to 21.1 μg m −3 (Figure 1).We estimate that for other monitor locations in southern California (such as in Los Angeles, Long Beach, and Orange), that measure between 8 and 15 μg m −3 annual PM 2.5 mass, losses are over 15%.The San Joaquin Valley (SJV) averages between 15 and 18 μg m −3 of PM 2.5 and loses between 10 and 15% of PM 2.5 mass in Teflon filter measurements annually.These findings suggest that the PM 2.5 mass is likely biased low relative to the actual ambient burden.This is a concern in central and southern California where reported PM 2.5 mass concentrations are already above the health-based PM 2.5 NAAQS.PM 2.5 mass concentrations in 2001 for Kern, Tulare, Fresno, and Los Angeles counties exceeded 15 μg m −3 (Figure 2) and were classified as nonattainment in 2005 using the three year rolling averages for PM 2.5 defined by the NAAQS.Decreasing mass concentrations for PM 2.5 are significant (p<0.05) for Fresno and Los Angeles counties.Tulare and Kern counties demonstrate no statistically discernible trend for PM 2.5 mass (Table 3), in sharp contrast to the national average.Air quality in the eastern U.S. is sufficiently improved to now attain PM 2.5 NAAQS, with one exception in Allegheny county, PA. 65 Decreasing PM 2.5 mass trends for the eastern U.S. are driven more heavily by sulfate mass compared to nitrate mass. 66ecline in annual average sulfate mass is noted for Los Angeles but not the other California locations analyzed here.The Spearman rank coefficients for PM 2.5 and nitrate mass are positive and highly correlated at all of the California monitoring sites (Figure S6), consistent with improving air quality due to PM 2.5 trends in the western U.S. driven by nitrate mass.The SJV and Los Angeles-South Coast Air Basin remain in nonattainment. 65Particulate nitrate is a major PM 2.5 chemical constituent in these locations.In the serious PM 2.5 nonattainment area of the agricultural SJV, in particular, Tulare and Kern counties, farm workers predominantly spend time in outdoor, ambient environments where the regulatory reported PM 2.5 mass concentrations are among the least reliable indicators of the true ambient burden.While nationally averaged air quality improved over the last 20 years, these locations lag behind the national average, including other environmental justice areas.We explore Kern, Tulare, and Fresno counties in detail because they are environmental justice areas in the SJV where pollution concentrations are regularly monitored and where policy is failing to attain air quality standards that safeguard human health.Reported annual FRM/FEM PM 2.5 trends show a decline between 2001 and 2021 in Kern, Tulare, and Fresno, with slopes of −0.012, −0.146, and −0.250 μg m −3 yr −1 , respectively (Figure 3).Accounting for nitrate losses, the trends of the ambient burden for Kern, Tulare, and Fresno counties show greater PM 2.5 decline than reported PM 2.5 , with slopes of −0.199, −0.286, and −0.408 μg m −3 yr −1 , respectively.Annual PM 2.5 trends in all counties studied are driven by high fall and wintertime particulate nitrate (Figure S7).Volatilized PM 2.5 from Teflon filters in Kern, Fresno, and Tulare counties are highest in the winter and fall, when concentrations are greatest.Declining coal use and implementation of low sulfur diesel in these regions contribute to decreasing PM 2.5 mass concentrations. 66In the eastern U.S., air quality policy is largely successful in attaining PM 2.5 NAAQS, where the dominant contributing species is well characterized by regulatory PM 2.5 monitoring.
Nitrate loss from regulatory monitors confounds interpretation of the true PM 2.5 burden especially in southern and central California, where frontline communities who typically work in outdoor, ambient environments experience the worst year round PM 2.5 .The discrepancy in reported versus actual annual PM 2.5 mass for counties located in the western U.S. (Figure S7) is a result of high winter nitrate measurements.Particulate nitrate as a percentage of reported PM 2.5 mass ( pNO PM ) is decreasing for the counties studied in California (Figure S8).The sole exception is Tulare county where wintertime pNO PM is increasing.Temperatures are highest in the summertime, approaching 30 °C on average, 20 °C in the fall and springtime, and 10 °C in the winter (Figure S9).The SJV has dry summers and more humid winters, with winters becoming drier over the studied period.Hotter and drier conditions promote evaporation in the PM 2.5 Teflon filters.Summertime PM 2.5 evaporative losses of nitrate are much lower than wintertime losses; however, this is because higher temperatures result in higher gas phase ammonia and nitric acid that do not condense onto the filter in the first place.It is important to note that ammonium nitrate precursors are abundant at this time, and they are difficult to measure with existing regulatory PM 2.5 monitors relative to nonvolatile species such as sulfate.As federal reference and equivalent methods of measuring particulate matter remain the standard, there continues to be a loss of PM 2.5 mass values that are unaccounted for in regulatory networks.
Conclusion.PTFE filters, as specified by CFR 40 Appendix L, hinder an accurate quantitative understanding of ambient PM 2.5 mass in some locations.Ammonium nitrate volatilization from PM 2.5 filters occurs at all EPA FRM/FEM sites across the state of California analyzed here.Reported PM 2.5 mass in the most serious nonattainment areas is biased low and fails to accurately describe the ambient burden due to volatilization of ammonium nitrate, most notably in southern and central California.While the PM 2.5 mass is declining on the national level, counties in the SJV do not experience the same level of improvement.Reduction of PM 2.5 mass is driven by different  indicating that national assessments may continue to neglect some areas, notably the environmental justice communities in the SJV agricultural region.The gap between regulatory definitions and actual burden may contribute to the fact that air quality improvement in the SJV lags behind the national average.In the context of rising ambient temperatures, the ability of FRMs and FEMs to accurately record the pollution burden that adversely impacts human health is an important and outstanding question.

Figure 1 .
Figure 1.Average percent mass of PM 2.5 volatilized from Teflon filters used in EPA's FRM/FEM monitors across California from 2001 to 2021.
chemical composition; east coast decadal trends are driven by sulfate concentrations and west coast trends are driven by nitrate concentrations.Reported PM 2.5 mass concentration in California with high nitrate, low sulfate, and high PM 2.5 concentrations are most biased due to ammonium nitrate losses in filter measurements.Accounting for nitrate losses, PM 2.5 mass trends show a greater fractional decline than what is reported by the FRM/FEM, while still remaining in nonattainment of the NAAQS.All studied sites in California continue to exceed PM 2.5 NAAQS during the time period,

Table 1 .
Filter Types of Designated Reference and Equivalent Methods Used in This Study PM 2.5 , sulfate, nitrate mass concentrations, and meteorology data between 2001 and 2021 are obtained from the EPA Air Quality System (AQS), which includes data from the FRM/FEM and CSN.
27,29They were able to characterize nitrate losses from Teflon filters observed in real-time samples collected in southern California accurately, under the assumption that and

Table 2 .
Thermodynamic and Solubility Data for NH 4 NO 3

Table 3 .
Mann Kendall Correlation Test in Order of PM 2.5 Species between 2001 and 2021 a Measurement starts in 2002.b Missing CSN data from 2014.