Appendix C: Weight of Evidence AnalysisIn 1997, U.S. EPA adopted the first set of PM2.5 air quality...

Appendix C: Weight of Evidence Analysis

SAN JOAQUIN VALLEY

PM2.5 WEIGHT OF EVIDENCE ANALYSIS

TABLE OF CONTENTS

INTRODUCTION .......................................................................................... 3 PM2.5 STANDARDS AND HEALTH EFFECTS ............................................ 3 NATURE AND EXTENT OF THE PM2.5 PROBLEM..................................... 4

Established Monitoring Network .......................................................... 4 Field Studies ....................................................................................... 5 Current Air Quality............................................................................... 7 Chemical Composition and Secondary Aerosol Formation.................. 9

Secondary ammonium nitrate formation ..................................... 12 Secondary organic aerosol formation ......................................... 16 Secondary ammonium sulfate formation .................................... 19

Emission Sources in the San Joaquin Valley....................................... 20 Emission inventory ..................................................................... 20 Source apportionment using source receptor models................. 21

PM2.5 AIR QUALITY AND EMISSION PROGRESS ..................................... 26 PM2.5 Concentrations........................................................................... 26

Design value trends.................................................................... 26 Seasonal, daily, and hourly trends.............................................. 29

Chemical Composition ........................................................................ 34 Emission Inventory .............................................................................. 36 Effectiveness of Emission Controls ..................................................... 37

NOx controls .............................................................................. 37 PM2.5 controls ............................................................................. 40 SOx controls............................................................................... 41

MODELED ATTAINMENT DEMONSTRATION ........................................... 43 SUMMARY................................................................................................... 49 REFERENCES ............................................................................................. 50

1

APPENDIX C1. Analysis for the Exclusion of the April 11, 2010 PM2.5 Value Recorded at Bakersfield-Planz from the Modeling Analysis for the San Joaquin Valley 2018 PM2.5 Plan

APPENDIX C2. Analysis for the Exclusion of the May 5, 2013 PM2.5 Value Recorded at Bakersfield-Planz from the Modeling Analysis for the San Joaquin Valley 2018 PM2.5 Plan

APPENDIX C3. Source Apportionment of PM2.5 Measured at the Fresno and Bakersfield Chemical Speciation Network Sites in San Joaquin Valley Using the Positive Matrix Factorization

APPENDIX C4. Precursor Demonstration for Ammonia, SOx, and ROG

2

INTRODUCTION

The United States Environmental Protection Agency (U.S. EPA) recommends that states supplement required air quality modeling with additional analyses to enhance the assessment of whether emissions reductions outlined in the State Implementation Plan (SIP) will result in attainment (U.S. EPA, 2014). Employing multiple analytical methods in a Weight of Evidence (WOE) approach yields a better understanding of the overall air quality problem and the level and mix of emissions controls needed for attainment. It also provides a more broadly informed basis for the attainment strategy.

Following U.S. EPA guidance on how to deal with the uncertainty inherent in predicting absolute fine particulate matter (PM2.5) concentrations in the future, an attainment demonstration that shows modeled design values falling either just above or just below the standard in the attainment year should be accompanied by a WOE demonstration to support the attainment demonstration. U.S. EPA recognizes the importance of a comprehensive assessment of air quality data and modeling and encourages this type of broad assessment for all attainment demonstrations. Further, U.S. EPA notes that the results of supplementary analyses may be included in a WOE determination to show that attainment is likely despite modeled results which may be inconclusive.

U.S. EPA recommends the WOE supplement the air quality modeling by including: 1) analyses of trends in ambient air quality and emissions, 2) observational models and diagnostic analyses, and 3) additional modeling evaluations. The scope of the WOE analysis is different for each nonattainment area, depending on the complexity of the air quality problem, how far into the future the attainment deadline is, and the amount of data and modeling available. For example, less analysis is needed for an area that is projecting attainment near-term and by a wide margin, and for which recent air quality trends have demonstrated significant progress, than for areas like the San Joaquin Valley (SJV or Valley) with more severe air quality challenges.

The following sections present the WOE assessment that supports the attainment demonstration for the 2018 Plan for the 1997, 2006, and 2012 PM2.5 Standards (2018 PM2.5 Plan) for the Valley.

PM2.5 STANDARDS AND HEALTH EFFECTS

Fine particulate matter up to 2.5 micrometers in diameter—PM2.5—is made up of many constituent particles and liquid droplets that vary in size and chemical composition. PM2.5 contains a diverse set of substances including elements such as carbon and metals, compounds such as nitrates, sulfates, and organic materials, and complex mixtures such as diesel exhaust and soil or dust. Some of the particles (primary PM2.5) are directly emitted into the atmosphere while others (secondary PM2.5) result when gases are transformed into particles through physical and chemical processes in the atmosphere.

Numerous health effects studies have linked exposure to PM2.5 to increased severity of asthma attacks, development of chronic bronchitis, decreased lung function in children, increased respiratory and cardiovascular hospitalizations, and even premature death in

3

people with existing cardiac or respiratory disease. In addition, California has identified particulate exhaust from diesel engines as a toxic air contaminant suspected to cause cancer, other serious illnesses, and premature death. Those most sensitive to PM2.5 pollution include people with existing respiratory and cardiac problems, children, and older adults.

National Ambient Air Quality Standards (NAAQS or standards) establish the levels above which PM2.5 may cause adverse health effects. In 1997, U.S. EPA adopted the first set of PM2.5 air quality standards, a 24-hour standard of 65 micrograms per cubic meter (µg/m3) and an annual standard of 15 µg/m3. In 2006, the 24-hour standard was tightened to 35 µg/m3, and in 2012, the annual standard was lowered to 12 µg/m3.

NATURE AND EXTENT OF THE PM2.5 PROBLEM

Established Monitoring Network

An extensive network of PM2.5 monitors throughout the San Joaquin Valley, shown in Figure C4-1, provides data to understand the extent of the PM2.5 problem. The locations of monitoring sites are selected to capture population exposure; many sites operate multiple monitoring instruments running in parallel. Currently, eleven sites operate Federal Reference Method (FRM) monitors, which provide regulatory data that are used to assess compliance with the federal PM2.5 standards. An additional eleven Federal Equivalent Method (FEM) monitors provide hourly PM2.5 measurements which can also be used to assess compliance with the PM2.5 standards. In addition, data collected at these monitors, as well as other non-regulatory monitors, serve to report air quality conditions to the public and support forecasting for the District’s Smoke Management System and residential wood burning curtailment programs. Finally, monitors at four sites collect samples that are further analyzed in the laboratory to determine the chemical make-up, or speciation, of PM2.5.

4

Figure 1. San Joaquin Valley PM2.5 monitoring network (FRM, FEM, and CSN monitors).

Field Studies

The San Joaquin Valley is one of the most studied air basins in the world. Dozens of major reports and publications have appeared in peer-reviewed international scientific and technical journals. Since 1970, close to 20 major field studies have been conducted in the Valley and surrounding areas that have shed light on various aspects of the nature and causes of ozone and particulate matter pollution.

The first major study specifically focused on particulate matter was the Integrated Monitoring Study in 1995 (IMS-95). IMS-95 formed the technical basis for the SJV 2003 PM10 Plan (approved by U.S. EPA in 2004),1 and acted as the pilot study for the subsequent California Regional Particulate Air Quality Study (CRPAQS), conducted between December 1999 and February 2001.

1 69 FR 30006

5

CRPAQS was a public/private partnership study designed to advance the understanding of the nature of PM2.5 in the Valley and guide development of effective control strategies. The study included monitoring at over 100 sites (Figure 2).

Figure 2. CRPAQS monitoring locations and equipment for ground-based and upper air data collection.

Other relevant field studies include the California portion of the 2008 Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS-CARB) (Jacob et al., 2010) and the California Research at the Nexus of Air Quality and Climate Change (CalNex2010)2 study conducted in 2010. The monitoring operations for both studies occurred from early to mid-summer and extended over Southern California and the Central Valley. The final CalNex2010 report to CARB was a synthesis of policy relevant findings designed to help formulate scientifically sound policies (NOAA, 2016).

An additional field study, Deriving Information on Surface Conditions from Column and Vertically Resolved Observations Relevant to Air Quality (DISCOVER-AQ),3 gathered air quality data in the Valley with the objective to provide an integrated dataset of airborne and surface observations relevant to the diagnosis of surface air quality conditions from space. DISCOVER-AQ was conducted from mid-January through mid-February 2013; data results and implications for the San Joaquin Valley are still being evaluated.

Findings from CRPAQS, CalNex2010, DISCOVER-AQ, and other studies have been integrated into the conceptual model of PM2.5 in the San Joaquin Valley. This conceptual model provides the scientific foundation for the WOE analysis supporting the annual and 24-hour PM2.5 standards’ attainment demonstration. Specific findings are integrated into the various WOE analysis sections of this document.

2 National Oceanic and Atmospheric Administration (NOAA), www.esrl.noaa.gov/csd/calnex/ 3 National Aeronautics and Space Administration (NASA), www-air.larc.nasa.gov/missions/discover-aq/discover-aq.html

6

http://www.esrl.noaa.gov/csd/calnex/

http://www-air.larc.nasa.gov/missions/discover-aq/discover-aq.html

http://www-air.larc.nasa.gov/missions/discover-aq/discover-aq.html

Current Air Quality

Geography and large-scale regional and local weather patterns influence the accumulation, formation, and dispersion of air pollutants in the San Joaquin Valley Air Basin (Air Basin or Valley). Covering nearly 25,000 square miles, the Valley is a lowland area bordered by the Sierra Nevada Mountains to the east, the Pacific Coast range to the west, and the Tehachapi Mountains to the south. The mountains act as air flow barriers, with the resulting stagnant conditions favoring the accumulation of pollutants. To the north, the Valley borders the Sacramento Valley and Delta lowland, which allows for some level of pollutant dispersion. Because of geography and meteorology, PM2.5 concentrations are generally higher in the southern and central portions of the Valley.

To determine attainment for the annual and 24-hour PM2.5 standards, the corresponding design value at each monitoring site must be calculated following protocol in 40 Code of Federal Regulations (CFR) Appendix N to Part 50. A design value is a statistic that describes the air quality status of a given location relative to the level of the NAAQS. Design values presented in this section may not be identical to design values presented in the modeling attainment demonstration (Appendix K to the 2018 PM2.5 Plan) since two values that were not representative of air quality were removed from the modeling attainment demonstration.4

This adjustment is discussed in detail in Appendix C1 to this WOE.

The annual design value represents a three-year average of the annual average PM2.5 concentrations measured at the site. If the annual design value is equal to or below the 1997 15.0 µg/m3 standard or the 2012 12.0 μg/m3 standard, the site meets that standard.

Figure 3 shows the 2017 annual PM2.5 design values throughout the San Joaquin Valley.5 6

Sites are shown from north to south with sites above the standard generally found in the central and southern Valley.

4 PM2.5 data collected on May 5, 2013, and April 11, 2010, from Bakersfield-Planz (AQS ID 060290016) 5 Comparisons of PM2.5 concentrations recorded at collocated or closely located FRM and FEM monitors have shown that FEM monitors record higher concentrations (e.g. the monitors in Merced). 6 Figure 3 does not include a 2017 annual design value for Corcoran since only 11 data points were recorded at the site in 2015 (all in the same quarter), fewer than are typically used to calculate a representative annual design value.

7

Figure 3. 2017 annual PM2.5 design values in the San Joaquin Valley.

The 24-hour PM2.5 design value represents a three-year average of the 98th percentile of the measured PM2.5 concentrations. Depending on a site’s 24-hour PM2.5 data collection schedule, the 98th percentile usually corresponds to a value between the 2nd and the 8th

highest value. If the design value is equal to or below the 1997 65 µg/m3 standard or the 2006 35 μg/m3 standard, the site attains that standard. Based on 2017 24-hour design values, only one site in the Valley attains the 2006 35 µg/m3 standard (Figure 4).

Figure 4. 2017 24-hour PM2.5 design values in the San Joaquin Valley.

8

PM2.5 exhibits a distinctive seasonal pattern throughout the Valley. In general, PM2.5 concentrations are higher in the winter and lower in the summer (Figure 5). PM2.5 also exhibits a geographic pattern, with concentrations increasing in magnitude from north to south. These two patterns can be illustrated (Figure 5) with data from Modesto, Fresno, Visalia, and Bakersfield. These four sites were selected for this analysis to represent areas in the northern (Modesto), central (Fresno and Visalia), and southern (Bakersfield) geographic regions of the Valley.

Figure 5. Monthly average (2014-2017) PM2.5 concentrations at four sites in the San Joaquin Valley.

Chemical Composition and Secondary Aerosol Formation

Examination of the chemical make-up of PM2.5 using the Chemical Mass Balance (CMB) analytical model provides another important element in understanding the nature of PM2.5 in the Valley and identifying contributing sources. The pie charts in Figure 6 show the chemical components that contribute to PM2.5 levels on an annual average basis at urban sites in Modesto, Fresno, Visalia, and Bakersfield in the northern, central, and southern regions of the Valley. Figure 7 shows the chemical components contributing to peak day PM2.5 levels at the same sites. While the relative percentages vary, in all cases the major components are ammonium nitrate and carbonaceous aerosols (organic and elemental carbon).

Ammonium nitrate is a large contributor to PM2.5 levels, constituting about one third of annual PM2.5 levels (Figure 6). On peak PM2.5 days, the ammonium nitrate contribution is even higher, comprising about half of the PM2.5 mass (Figure 7). Ammonium nitrate is formed in the atmosphere through two distinct pathways (daytime and nighttime) that convert NO2 to HNO3, which then reacts with NH3 to form ammonium nitrate. The daytime pathway is initiated by the hydroxyl radical (OH), which is formed through complex photochemistry of VOCs and other trace gases in the atmosphere, while the nighttime pathway is initiated by

9

ozone. Sources emitting NOx include motor vehicles and stationary combustion sources, while the largest sources of ammonia are livestock operations and fertilizer application. The stagnant, cold, and damp conditions that occur during the winter promote the formation and accumulation of ammonium nitrate. Additional information on ammonium nitrate formation can be found in the secondary ammonium nitrate formation section below.

While the annual average and seasonal high day contributions of ammonium nitrate differ considerably, the average and high day contributions of carbonaceous aerosols (also known as carbon compounds) are fairly similar (Figure 6 and Figure 7). On an annual average basis, carbonaceous aerosols are responsible for 44 to 53 percent of the mass, and on a high PM2.5 day, the contribution is only slightly lower—39 to 47 percent. Carbonaceous aerosols include both organic matter, comprised of primary organic aerosols (POAs) and secondary organic aerosols (SOAs), and elemental carbon (EC). POAs are directly emitted into the atmosphere from activities such as residential wood combustion, cooking, biomass burning, and direct tailpipe emissions from mobile sources. SOAs are formed in the atmosphere through the oxidation of volatile organic compounds (VOCs) from numerous anthropogenic and biogenic emissions sources. EC is directly emitted PM2.5 and comes from mobile and stationary combustion sources, with significant contributions from diesel sources.

While ammonium nitrate and carbonaceous aerosols make up most of the mass of PM2.5, smaller contributions also come from ammonium sulfate, geological material, and elements. Ammonium sulfate contributes approximately 10 percent of annual PM2.5 levels at each of the four sites. On peak days, the ammonium sulfate contribution is about 5 percent. Ammonium sulfate forms in the atmosphere when oxides of sulfur (SOx) emitted from combustion sources reacts with ammonia from sources like livestock operations and fertilizer application.

Geological material or dust contributes approximately 10 percent to annual PM2.5 levels at Bakersfield, while at Fresno and Modesto it contributes about 6 percent. On a peak PM2.5 day, contribution from geological material is significantly lower and comprises only about 2 percent of the mass. Geological material is directly emitted PM2.5 and comes from dust suspended into the air by vehicle travel on roads, soil from agricultural activities, and other dust-producing activities such as construction.

Elements make up a small portion of PM2.5 mass, contributing about 2 to 4 percent on an annual basis and about 1 percent on a peak day. Elements found in the air include iron, silicon, aluminum, chlorine, and others, in trace amounts.

10

Figure 6. Annual average PM2.5 chemical composition (2015-2017).

11

a) Modesto b) Fresno

c) Visalia d) Bakersfield

Figure 7. Averaged peak day PM2.5 chemical composition (2015-2017).



Secondary ammonium nitrate formation

In January and February 2013, the DISCOVER-AQ field campaign, launched by the National Aeronautics and Space Administration (NASA), took place in the SJV. During t he field campaign, aircraft measurements of PM2.5 and its precursors were made within the planetary boundary layers over agricultural and urban regions in the SJV. These measurements included total nitric acids (total HNO3, gas + particle phases, or g + p), gaseous NH3, particulate ammonium (NH +4 ), and sulfate (SO 2-4 ), which allowed for the evaluation of precursor limitation for ammonium nitrate formation. Total nitric acid was measured by thermal dissociation laser induced fluorescene (TD-LIF) (Pusede et al., 2016; Womack et al., 2017). Ammonia was measured by both a proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS) and a cavity ring down spectrometer (CRDS) (Sun et al., 2015),

12

but for analyzing the precursor limitation, ammonia measurements based on PTR-ToF-MS were used, since the focus is within the planetary boundary layer, similar to Sun et al. (2015). Particulate ammonium and sulfate were measured by a particle-into-liquid sampler (PILS) and off-line ion chromatography (IC) analysis. The observational data were obtained from the NASA website.7

The excess NH3 is defined as the sum of gaseous NH3 and particulate ammonium minus 2x particulate sulfate and total nitric acids (g + p) (Equation 1 from Blanchard et al., 2000). Particle mass concentrations were converted to corresponding mixing ratios based on ambient air density and molecular weights of species. Excess NH3 is therefore expressed in the unit of mixing ratio. While the calculation of excess NH3 in Blanchard et al. (2000) also incorporated the impacts from other ions, such as sodium, calcium, magnesium, potassium and chloride, those ions were not included in this analysis because the measurements were sparse and including them would have significantly reduced the number of concurrent observational data points that could be used in the analysis. However, including the other ions would likely lead to even greater excess NH3 in the SJV, because sea salt is minimal in the SJV and other cations act to increase the excess NH3. In the calculation of excess NH3, if the value is greater than zero, this indicates that secondary particulate nitrate is in a NOx-limited regime. Conversely, a value less than zero demonstrates an ammonia-limited regime.

+ 2− 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝑁𝑁𝑁𝑁3 = 𝑁𝑁𝑁𝑁3(𝑔𝑔) + 𝑁𝑁𝑁𝑁4(𝑝𝑝) − 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝑁𝑁𝑁𝑁𝐻𝐻3(𝑔𝑔+𝑝𝑝) − 2 × 𝑆𝑆𝐻𝐻4(𝑝𝑝) (1)

Figure 8 shows the excess NH3 in the bottom 1km of the atmosphere in the SJV based on aircraft measurements on January 18 and 20, 2013, during which PM2.5 concentrations in the SJV were high. Each data point of excess NH3 was calculated based on the averaged 10 seconds’ observational data. Excess NH3 is clearly above zero and can be above 50 parts per billion (ppb) in many cases, which indicates that nitrate formation in the SJV is in a NOx-limited regime (Blanchard et al., 2000). Analysis using the inorganic aerosol thermodynamic model ISORROPIA (Fountoukis and Nenes, 2007) corresponding to conditions in the SJV observed on January 18 and 20, 2013 (i.e., temperature ~ 285 K, RH ~ 60 percent, sulfate concentration ~ 0.8 µg/m3, total nitric acid concentrations ~ 20 µg/m3) indicated that when excess NH3 is above ~4 ppb, more than 95 percent of nitric acids reside in the particulate phase and the sensitivity of ammonium nitrate to NH3 perturbations becomes small. Prabhakar et al. (2017) also showed that under typical conditions, more than 90 percent of nitric acid resides in the particle phase in Fresno, implying weak sensitivity of ammonium nitrate to ammonia changes.

7 NASA. https://www-air.larc.nasa.gov/cgi-bin/ArcView/discover-aq.ca-2013

13

https://www-air.larc.nasa.gov/cgi-bin/ArcView/discover-aq.ca-2013

Figure 8. Excess NH3 in the SJV on Jan 18 (Left) and Jan 20 (Right) based on NASA aircraft measurements in 2013.

Previous studies also indicate an ammonia excess in the SJV. For example, Lurmann et al. (2006) showed that ammonia concentrations were an order of magnitude greater than nitric acid concentrations in Fresno and Angiola, based on denuder difference measurements made during CRPAQS in the wintertime 2000-2001. It is important to note that from 2000 to 2015, NOx emissions in the SJV have decreased by approximately 60 percent while NH3 emissions in the SJV have remained relatively constant, leading to an even greater excess of ammonia compared to 2000. Markovic, et al. (2014) showed that observed nitric acid mixing ratios were 2 orders of magnitude lower than observed NH3 mixing ratios in Bakersfield in May/June 2010 based on measurements using the Ambient Ion Monitor-Ion Chromatograph (AIM-IC) system during the CalNex2010 field campaign. Parworth et al. (2017) showed that on average, observed molar concentration of ammonia was 49 times greater than nitric acid, demonstrating that ammonium nitrate formation was limited by nitric acid availability at Fresno in winter 2013, based on measurements using a particle-into-liquid sampler with ion chromatography and annular denuders.

While ammonium nitrate formation is observed to be in a NOx limited regime, this does not conflict with the modeling results that showed some sensitivity of ammonium nitrate formation to ammonia emission reductions. At equilibrium state, the product of gaseous nitric acid and ammonia in the atmosphere is a constant, and the equilibrium constant depends on ambient conditions as well as particulate compositions (Seinfeld and Pandis, 2006). Even in a NOx-limited regime, the perturbation of ammonia mixing ratios influences the partitioning of nitric acids. As ammonia becomes more excessive, the partitioning of nitric acids shifts more towards the particulate phase. After the vast majority of nitric acids are in the particulate phase (e.g., > 95 percent of nitric acids in particulate phase), the formation of ammonium nitrate becomes far less sensitive to additional ammonia. Finally, the dry deposition velocity difference between gaseous nitric acid and particulate nitrate further adds to the complexity. When the partitioning of gaseous and particulate nitric acids is perturbed by changing ammonia, because of the different removal rate of gaseous and particulate nitric acids (Meng et al., 1997; Pusede et al., 2016), the mass of total nitric acids is perturbed as well, which

14

could somewhat amplify the response to ammonia emissions changes. In the SJV, because of the excessive ammonia, the formation of ammonium nitrate is much more sensitive to the reductions of NOx than to the reductions of ammonia, which has been widely documented in past modeling studies. Nevertheless, some sensitivity of ammonium nitrate formation to large ammonia reductions has been shown in previous modeling studies as well (Chen et al., 2014; Kleeman et al., 2005).

Role of ammonia in ammonium nitrate formation

As discussed in the previous section, the precursor in shortest supply limits the amount of ammonium nitrate formation. An evaluation of the magnitude of NOx and ammonia emissions provides a first-level assessment of the relative abundance of these two precursors.

Table 1 lists NOx and ammonia winter and annual average emissions in the current inventory for the base year and the three attainment years for all three standards (2013, 2020, 2024, and 2025). In simple terms, it takes one molecule of NOx and one molecule of ammonia to form one molecule of ammonium nitrate. Due to differing molecular weights, one ton of NOx emissions contains fewer molecules than one ton of ammonia emissions. An emissions inventory comparison should, therefore, be made after normalizing for molecular weight.

Due to emission source test procedures, most NOx emissions are expressed in terms of nitrogen dioxide (NO2). Since one NO2 molecule has a mass of 46 universal atomic units (u) and one ammonia (NH3) molecule has a mass of 17 u, one ton of NH3 has 2.7 times (46 u/17 u) the number of molecules as one ton of NO2. Dividing the NOx emissions by 2.7 therefore provides a common basis for comparison to ammonia emissions. On this normalized comparison basis, ammonia is significantly more abundant than NOx, particularly in the future year. In addition, as previously noted in the chemistry section, only a portion on the NOx is ultimately converted to ammonium nitrate.

Table 1. Comparison of NOx and ammonia emissions (tons per day [tpd]) in the base year and the three attainment years on a winter and annual basis.

2013 2020 2024 2025 Winter Annual Winter Annual Winter Annual Winter Annual

NH3 310 329 307 326 306 325 306 324 NOx 301 317 191 203 139 149 135 144 Normalized NOx 111 117 70 75 51 55 50 53

Role of VOC in ammonium nitrate formation

CARB used the integrated reaction rate (IRR) analysis in the Community Multiscale Air Quality Monitoring System (CMAQ),8 a numerical air quality model that predicts the concentration and deposition of airborne gases and particles, to understand the impact of VOC emission reductions on nitrate formation in the model. IRR gives production or loss rates for individual gas-phase chemical pathways in the model. The heterogeneous nitric

8 Specifically, CMAQv5.0.2

15

acid formation rate was obtained from the aerosol module. Two separate simulations using the January 2013 meteorological fields were conducted for two future year emission scenarios. One is the baseline future year and the other involved 25 percent reductions in VOC emissions from the baseline scenario. When VOC emissions were reduced by 25 percent, we found that both daytime and nighttime nitric acid formation rates were only slightly impacted by the VOC emission reductions.

Daytime homogeneous nitric acid formation is primarily through the gas-phase reaction of NO2 and the hydroxyl radical (OH), which is influenced by VOC levels in the atmosphere. When VOC emissions were reduced (particularly at urban locations such as Bakersfield), the daytime nitric acid formation rate was also reduced since lower VOC levels result in less OH through the photo-oxidation of VOCs emitted into the atmosphere (Pusede et al., 2016).

Nighttime heterogeneous nitric acid formation involves the heterogeneous reaction of nitrogen pentoxide (N2O5) on particles. N2O5 is formed from nitrogen dioxide (NO2) and nitrogen trioxide (NO3). NO3 is a reaction product between NO2 and ozone (O3) (Seinfeld and Pandis, 2006). In places like Visalia, the nighttime heterogeneous nitric acid formation rate above the surface was slightly increased when VOC emissions were reduced. Model output showed reduced peroxyacetyle nitrate (PAN) formation under reduced VOC emissions. Less PAN formation then leads to increased availability of NO2, enhanced N2O5 formation (Meng et al., 1997), and a slightly increased heterogeneous formation rate.

Overall, reducing VOCs emissions by 25 percent increased ammonium nitrate slightly (~1 percent) at PM2.5 design value sites, which is the net outcome from different competing chemical processes as well as the transport and mixing processes in the atmosphere.

Secondary organic aerosol formation

VOC emissions also have the potential to contribute to SOAs. While these components contribute to observed PM2.5 concentrations in the San Joaquin Valley to a small degree, the weight of evidence indicates that anthropogenic VOC is not a significant contributor to PM2.5.

SOAs form when intermediate VOCs, emitted by anthropogenic and biogenic sources, react and condense in the atmosphere to become aerosols. In addition, lighter VOCs participate in the formation of atmospheric oxidants which then participate in the formation of SOAs. The processes of SOA formation are complex and have not been fully characterized. The apportionment of PM2.5 organic carbon to primary and secondary components is a very active research area.

The UCD-CIT air quality model (Chen et al., 2010) was used to investigate the apportionment of PM2.5 organic carbon for the 2000/2001 CRPAQS episode. From the total predicted PM2.5 organic carbon in the urban Fresno and Bakersfield areas, 6 percent and 4 percent were SOAs, respectively, while in the rural Angiola area, just south of Corcoran, SOAs comprised 37 percent. The major precursors of SOAs were long-chain alkanes, followed by aromatic compounds, and the main sources of these precursors were solvents, catalytic gasoline engines, wood smoke, non-catalytic gasoline engines, and other anthropogenic sources, in that order.

16

In contrast, on an annual average basis, SOAs derived from anthropogenic VOC emissions (SA) account for only 1 to 2 percent of the annual total PM2.5 concentrations throughout the Valley, and are a small part of the organic aerosol concentrations (3.3 percent at Bakersfield and 3.1 percent at Fresno). CARB air quality modeling exercises conducted using the CMAQ model showed that primary PM2.5 emissions (PA) are the main contributor to organic aerosols. Furthermore, as illustrated in Figure 9, SOAs are mostly formed during the summertime, when total PM2.5 concentrations are low, and are mainly derived from biogenic emission sources (SB). SA and SB together comprise total SOAs.

17

Figure 9. Daily contributions to organic aerosol concentrations modeled with CMAQ (2013).

(a) Bakersfield

(b) Fresno

Note: SB = Secondary aerosols formed from biogenic VOC emissions, SA = Secondary aerosols formed from anthropogenic source VOC emissions, and PA = Primary organic aerosols.

18

2013 2020 2024 2025

As part of the CRPAQS study, simulations of a wintertime episode conducted using CMAQ-Madrid, a model with an enhanced SOA formation mechanism, also found that organic aerosol concentrations were dominated by direct (primary) emissions. The study found that, because of the dominance of primary PM2.5 organic matter, a 50 percent reduction in anthropogenic VOC emissions has limited effects on the modeled PM2.5 organic matter (Pun et al., 2009).

These study results show that for secondary organic aerosols, further VOC reductions would have very limited effectiveness in reducing PM2.5 concentrations. VOC reductions also result in small increases in PM2.5 overall, due to the fact that they increase nitrate.

Secondary ammonium sulfate formation

SOx emitted from stationary and mobile combustion sources, mostly as sulfur dioxide (SO2), are oxidized in the atmosphere to ultimately form sulfuric acid (H2SO4). Sulfuric acid then combines with ammonia to form ammonium sulfate:

𝑁𝑁2𝑆𝑆𝐻𝐻4 + 2𝑁𝑁𝑁𝑁3 → (𝑁𝑁𝑁𝑁4)2𝑆𝑆𝐻𝐻4

Table 2 lists SOx and ammonia winter and annual average emissions in the current inventory for four years (2013, 2020, 2024, and 2025), the base year and the three attainment years for the 2018 PM2.5 Plan. As shown in the above equation, in simple terms it takes one molecule of SOx and two molecules of ammonia to form one molecule of ammonium sulfate; however, due to differing molecular weights, one ton of SOx contains fewer molecules than one ton of ammonia. An emissions inventory comparison should, therefore, be made after normalizing for molecular weight.

Since one SO2 molecule weighs 64 u and one NH3 molecule weighs 17 u, one ton of NH3 has 3.8 times (64 u/17 u) the number of molecules as one ton of SO2. Since one molecule of SO2 reacts with 2 molecules of NH3, dividing the SO2 emissions by 1.9 provides a common basis for comparison to the ammonia emissions. On this normalized comparison basis, ammonia is approximately 75 times more abundant than SOx. Thus, SOx emissions are the limiting precursor for ammonium sulfate formation.

Table 2. Comparison of SOx and ammonia emissions (tpd) in the base year and the three attainment years on a winter and annual basis.

Winter Annual Winter Annual Winter Annual Winter Annual NH3 310 329 307 326 306 325 306 324 SOx 8.4 8.5 7.6 7.8 7.8 8.0 7.8 8.0 Normalized SOx 4.4 4.5 4.0 4.1 4.1 4.2 4.1 4.2

19

Emission Sources in the San Joaquin Valley

Emission inventory

Emission inventories provide emission estimates for sources of directly emitted (primary) PM2.5 and of each of the gaseous precursors of secondary PM2.5 (NOx, SOx, and ammonia).

Table 3 lists the main PM2.5 components and links them to their largest emission sources based on San Joaquin Valley emission inventory data for 2013, the base year for the 2018 PM2.5 Plan. VOC emissions are not listed, since, as previously discussed, VOC emission reductions have no effect on PM2.5 concentrations in the Valley. Emission sources are listed in descending order of magnitude.

Table 3. Main emission sources (2013) of PM2.5 components.

20

PM2.5 Component (percent of PM2.5)

Process Main Emission Sources

Ammonium nitrate (about 40 percent)

Formed in the atmosphere from the reactions of NOx

and ammonia emissions

NOx: Heavy-duty diesel vehicles account for

approximately 45 percent of annual NOx emissions.

Farm equipment; off-road equipment; light-,

medium-, and heavy-duty gas trucks; trains; light-duty passenger cars; and residential fuel

combustion account for an additional 40 percent. Ammonia: Livestock husbaccount for ove

emissions.

andry and r 90 perc

fertiliz ent of a

er application nnual ammonia

Ammonium sulfate

(about 5-15 percent)

Formed in the atmosphere from the reactions of SOx

and ammonia emissions

SOx: Manufacturing of chemicals, glass, and related

products; fuel combustion; and residential wood combustion account for about 80 percent of

annual SOx emissions. Ammonia: Livestock husbaccount for ove

emissions.

andry and f r 90 percen

ertilizer at of annu

pplication al ammonia

Organic carbon (about 20-35 percent)

Directly emitted from motor vehicles and combustion

processes

Combustion PM2.5: Residential fuel combustion, diesel trucks,

cooking, managed burning and disposal, farm equipment, oil and gas production, electrical

utilities, aircraft, and off-road equipment account for over 66 percent of the annual combustion

PM2.5 emissions.

Elemental carbon (about 2-5 percent)

Directly emitted from motor vehicles and combustion

processes

Geological matter Directly emitted from dust- Dust PM2.5: generating sources Farming operations, fugitive windblown dust, (about 5-15 percent) paved and unpaved road dust, construction and

demolition, and mineral processes account for 100 percent of the annual dust PM2.5 emissions.

While emission inventories provide a broad overview of Valley-wide and county-level sources, additional methods using ambient data and source apportionment modeling provide supplemental information on the sources directly impacting individual monitoring sites. The following sections describe these analyses.

Source apportionment using source receptor models

Positive Matrix Factorization

CARB staff applied the PMF2 model to the chemically speciated PM2.5 data collected at the Bakersfield-California and Fresno-Garland monitoring sites. Bakersfield data from 2011-2015 and Fresno-Garland data from 2012-2015 were used. The average source contributions to PM2.5 concentrations are illustrated in Figure 10. Similar to the results presented above from the chemical mass balance (CMB) analysis, secondary nitrate contributes the most at both sites—37 percent at Bakersfield and 39 percent at Fresno. Gasoline-fueled vehicles contribute about 20 percent at Bakersfield and 16 percent at Fresno. Biomass burning (which includes residential wood combustion, agricultural burning, and cooking) contributions differ significantly between the two sites, accounting for approximately 4 percent at Bakersfield, but 17 percent at Fresno. Secondary sulfate accounts for 15 percent at Bakersfield and 14 percent at Fresno. Similar to the biomass burning category, airborne soil contributions differ between the two sites with 12 percent at Bakersfield and 6 percent at Fresno. Other sources are minor contributors.

Figure 10. Average source contributions estimated using PMF.

a) 2011 - 2015 Average Source b) 2012 - 2015 Average Source Contribution in Bakersfield - California Contribution in Fresno - Garland

While the absolute magnitude of the contributions estimated by the two models varies to some extent, taken together, the CMB and PMF source apportionment studies confirm the importance of secondary nitrate contributions to PM2.5 levels both on an annual average basis and during the winter. In addition, exhaust from gasoline vehicle and biomass burning were

21

found to be significant contributors to PM2.5 levels. Appendix C3 describes the PMF analysis in greater detail.

In addition to the 2011-2015 Bakersfield and 2012-2015 Fresno data used previously, as described above, CARB staff also used PMF2 to analyze recent speciated PM2.5 data from 2016-2017 for the same Valley sites. During this time period, the contributions from secondary nitrate decreased the most at Bakersfield (2.4 µg/m3) and Fresno (2.9 µg/m3), which led to a decrease in PM2.5 mass concentrations at both sites. The trends of winter contributions from three significant sources are compared in Figure 11. As shown, secondary nitrate and biomass burning contributions decreased distinctly in the winter of 2016 at both Bakersfield and Fresno. This is discussed in more detail in Appendix C3.

22

Figure 11. Comparisons of median winter (November to February) contributions from significant sources at Bakersfield and Fresno.

b) Gasoline vehicles

c) Biomass burning

a) Secondary nitrate

23

San Joaquin Valley organic aerosol study (2016-2017)

Receptor-based source apportionment tools have been used extensively in the SJV and moreover globally to better understand ambient PM contributions; however, there is still significant uncertainty associated with the source contributions to carbonaceous aerosols. To address this uncertainty, a special project was conducted by the University of Wisconsin-Madison (UWM) investigating the sources of organic aerosols in the San Joaquin Valley (Skiles et al., 2018). The project involved fine particle sample collection on a three-day schedule at Fresno and Bakersfield from January 2015 through February 2016. These samples were analyzed at the UWM for organic molecular markers, water-soluble organic carbon (WSOC), and elemental and organic carbon (ECOC). Organic markers included n-alkanes, cycloalkanes, alkanoic acids, resin acids, aromatic diacids, alkanedioic acids, steranes, hopanes, PAHs, oxy-PAHs, phthalates, levoglucosan, and sterols. A database of these measurements was integrated with measurements from the fine particle mass FRM and CSN monitoring networks. The data were used in several source apportionment models to examine source contributions to fine particle organic carbon across seasons and for PM2.5 exceedance days. The trends in the monitoring data as well as the source apportionment results are discussed in more detail in the report, along with some meteorological analyses to help explain the observed trends in PM2.5 component concentrations and PM2.5 source contributions.

The UWM study coupled metrological transport data with receptor models to better understand the origin of carbonaceous aerosols. The integrated measurements were used for source apportionment and source mapping, including molecular marker chemical mass balance modeling (MM-CMB), molecular marker positive matrix factorization (MM-PMF) modeling, source mapping by potential source contribution function (PSCF), and Bayesian source apportionment model concentration field analysis (CFA). The project addresses the uncertainty of carbonaceous source contributions from both primary and secondary sources in the SJV, including motor vehicle contribution, seasonal secondary organic aerosols (SOA), meat cooking emissions, and controlled and uncontrolled biomass combustion emissions.

Overall, four primary sources were identified as contributing to PM2.5 OC in Bakersfield and Fresno: mobile sources, vegetative detritus, biomass burning, and meat cooking. Mobile sources are the total of diesel, gasoline, and smoking vehicles contributions. Mass not apportioned by the model is presented as “CMB Other” and represents secondary sources that are not accounted for by the primary source profiles used in the model.

In terms of both total OC and OC apportioned to primary sources, concentrations in Fresno were generally higher than those in Bakersfield. At both sites, the month with the lowest total and apportioned OC was May, and the month with the highest apportioned OC concentrations was December (although the highest total OC monthly average was November). There is a distinct seasonal pattern in biomass burning and meat cooking source contributions, while vegetative detritus and mobile sources remain relatively constant throughout the year. These source contribution trends parallel the trends in tracer compounds.

Of the four primary sources identified, the major source at both sites was biomass burning in the colder months (January-March and November-February in Fresno and January and

24

November-February in Bakersfield), and mobile sources were the major source for the remaining months; however, “CMB Other” exceeded apportioned mass during warmer months at both sites (March-October in Fresno and February-October in Bakersfield), indicating that secondary sources were important for more than half of the year at both sites. Vegetative detritus was a very minor source, with no monthly average source contributions exceeding 0.2 µg/m3 and no daily source contributions exceeding 0.5 µg/m3. Though the trends in sources are similar between the two sites, source contributions are not correlated between the two sites on a daily basis.

The PMF model further identified a forest fire contribution that could not be distinguished in the CMB analysis. The combination of two models (MM-CMB and MM-PMF) provides reasonable quantitative information on important sources of OC at the two sampling sites (Table 4). The “Derived SOA Total” is estimated as a difference between measured OC and primary sources identified by CMB along with Forest Fires identified by PMF. The accuracy of the “Derived SOA Total” estimate in turn depends on the accuracy of primary source apportionment. This is especially relevant with respect to meat cooking which could be underestimated during summer due to oxidation. Underestimation of meat cooking contribution could lead to overestimation of “Derived SOA Total.”

Table 4. OC source contribution at Bakersfield and Fresno.

OC Source Contribution (µg/m3) MM-CMB Biomass Burning

MM-CMB Mobile

Sources

Derived SOA Total

MM-CMB Vegetative

Detritus

MM-CMB Meat

Cooking

PMF Forest Fire

Bakersfield 2015 Annual Average 0.605 0.718 1.745 0.081 0.353 0.395 Exceedance Days Average a 1.833 0.947 2.169 0.112 0.959 0.306

Winter Average b 1.507 0.802 1.471 0.105 0.661 0.193 Fresno

2015 Annual Average 1.152 0.711 1.527 0.104 0.436 0.568 Exceedance Days Average a 4.022 1.086 1.321 0.258 1.199 0.695

Winter Average b 2.999 0.791 1.074 0.145 0.962 0.450 Summary of apportioned OC for Bakersfield and Fresno during the 2015/2016 project. Annual average, average exceedance days, and winter averages are reported, with MM-CMB apportioned results for vegetative detritus, biomass burning, meat cooking, mobile sources and PMF results for forest fires.

a Average when daily PM2.5 exceed the 24-hour regulatory over the project span. b Average of winter month November December, January, and February over the entire project period. c Derived SOA is estimated from “CMB Other” minus the PMF forest fire contribution. The PMF forest fire contribution includes some open burning of biomass.

25

PM2.5 AIR QUALITY AND EMISSION PROGRESS

PM2.5 Concentrations

Design value trends

On an annual average basis, PM2.5 air quality in the San Joaquin Valley has improved over the last dozen years. Figure 12 shows annual design value trends at sites in Modesto, Fresno-1st/Garland,9 Visalia, and Bakersfield in the northern, central, and southern regions of the Valley, respectively. As previously noted, design values presented in this section may not reflect design values presented in the modeling attainment demonstration.

Figure 12. Trend in annual PM2.5 design values (2001-2017) at the Modesto, Fresno, Visalia, and Bakersfield monitoring sites.

The Valley was nearing attainment of the 15 µg/m3 annual standard through 2012, with only a few sites recording design values over the standard. Extensive wildfires in 2008 impacted design values from 2008 through 2010. Meteorological conditions associated with a severe state-wide drought—including a persistent upper-level high pressure ridge that interrupted normal weather patterns, decreasing rainfall and causing longer-lasting stagnant conditions—

9 The Fresno-1st monitor operated from 1999 to 2011 and the Fresno-Garland site operated from 2012 to present. Data from the two sites, which were about a quarter mile apart, were combined to provide a continuous stream of data for trends analysis.

26

potentially increased PM2.5 concentrations during the 2013/2014 winter, with a subsequent rise in 2013 through 2015 design values.

Despite these increases, the Valley is still seeing overall progress. Between 2001 and 2017, annual design values declined between 24 and 44 percent. Approximately 70 percent of sites in the Valley attain the 15 µg/m3 standard in 2017 with around 25 percent attaining the 12 µg/m3 standard (see Figure 3). The highest remaining levels occur in the central and southern regions, where design values are about 9 to 31 percent over the 12 µg/m3 standard.

As illustrated in Figure 13, over the long term, the 24-hour PM2.5 design values also show a downward trend. The most pronounced progress occurred between 2001 and 2005. Extensive wildfires during the summer of 2008 in Northern California impacted the 2008, 2009, and 2010 design values throughout the Valley, with greater impacts in the northern Valley. Increases, potentially due to extreme drought conditions, were also noted from 2013 to 2015. Overall, between 2001 and 2017, the 24-hour PM2.5 design values in the Valley have decreased by 30 to almost 50 percent. In 2017, all sites in the Valley, with the exception of Corcoran, attained the older 65 µg/m3 standard and are well on the way to attaining the more recent 35 µg/m3 standard.

Figure 13. Trend in 24-hour PM2.5 design values (2001-2017) at the Modesto, Fresno, Visalia, and Bakersfield monitoring sites.

Looking at the number of days with measured PM2.5 concentrations over both the 65 µg/m3

and 35 µg/m3 standards provides another way to assess PM2.5 impacts. Over the long term, between 1999 and 2017, the number of days exceeding the 65 µg/m3 standard decreased 70 percent at both the combined Fresno sites and the Bakersfield-California site (Figure 14),

27

the only two of the four sites that collected daily data for the entire 18 year period. Within the same period, the number of days over the 35 µg/m3 standard saw a 50 percent decline at both the Bakersfield-California and Fresno sites (Figure 15).

The increase in the number of exceedance days in 2013 compared to 2012 was potentially due to severe drought-related conditions during the winter of 2013-2014. The Valley experienced similarly severe meteorological conditions during the 1999-2000 and 2000-2001 winters. The total number of exceedance days, however, was much higher during these earlier years, providing evidence that the emission reductions achieved in the Valley have resulted in PM2.5 air quality improvements.

Figure 14. Trend in measured days over the 24-hour standard of 65 µg/m3 (1999-2017) at the Modesto, Fresno, Visalia, and Bakersfield monitoring sites.

28

Figure 15. Trend in measured days over the 24-hour standard of 35 µg/m3 (1999-2017) at the Modesto, Fresno, Visalia, and Bakersfield monitoring sites.

Seasonal, daily, and hourly trends

Comparing the change in the frequency distribution of 24-hour PM2.5 concentrations over the last dozen years provides another means of looking at air quality changes over the years. As illustrated in Figure 16, the fraction of days recording PM2.5 over the 24-hour standard of 35 μg/m3 decreased between the three-year periods of 2005-2007 and 2015-2017 at the four monitoring sites shown. During the 2005-2007 period, the frequency of days over the 35 μg/m3 standard ranged from 9 percent at Modesto to 13 percent at Bakersfield. Ten years later, the frequency of days over the standard ranged from 6 percent at Modesto to a high of 8 percent at Bakersfield.

The frequency of days over the annual standard of 12 μg/m3 during the 2005-2007 period showed a range from 36 percent at Modesto to 63 percent at Visalia. This decreased approximately 10 percent by the 2015-2017 period, ranging from 26 percent at Modesto to 51 percent at Visalia.

29

Figure 16. Change in PM2.5 concentration frequency distribution between 2005-2007 and 2015-2017.



Focusing on the winters (November through February), when meteorological conditions are most conducive to PM2.5 formation and accumulation and when the highest PM2.5 levels generally occur, provides further insight into PM2.5 air quality progress. A clear downward trend from 1999 to 2017 is evident (Figure 17), with winter-averaged concentrations decreasing by 50 percent. Although drought conditions in 2013 potentially increased winter average PM2.5 concentrations, a downward trend is still evident, indicating that although some setbacks may arise, controls in place continue to improve air quality in the Valley.

30

Figure 17. Changes in winter (November to February) average PM2.5 concentrations.

Note: On the horizontal axis, “1999” refers to the winter from November 1999 through February 2000, “2000” refers to the winter from November 2000 through February 2001, etc.

Progress in reducing PM2.5 levels is further evidenced by comparing daily PM2.5 concentrations during two winters ten years apart. The graphs in Figure 18 compare PM2.5 concentrations measured at Modesto, Fresno, Visalia, and Bakersfield between November 2016 and February 2017 to PM2.5 concentrations between November 2006 and February 2007. Overall, the 2016/2017 air quality at these four sites showed improvements. Maximum 24-hour concentrations were approximately 20 to 30 percent lower, average concentrations during the three month period were approximately 40 to 50 percent lower, and the number of days over the 24-hour standard of 35 µg/m3 decreased by 40 to 80 percent.

31

Figure 18. Comparison of the 2016/2017 PM2.5 winter to the 2006/2007 PM2.5 winter in the San Joaquin Valley.



Progress in reducing PM2.5 levels is further corroborated by comparing changes in monthly average PM2.5 concentrations between 2005-2007 and 2015-2017 (Figure 19). All four sites, representing the northern (Modesto), central (Fresno) and southern (Visalia and Bakersfield) portions of the Valley, exhibit similar distinctive seasonal patterns of relatively higher PM2.5 concentrations in the winter that is consistent across time. Average monthly concentrations have decreased almost year-round, with minor declines outside of the winter months, particularly in the northern and central regions.

32

Figure 19. Changes in PM2.5 monthly concentrations (2005-2007 and 2015-2017).



Comparing changes in PM2.5 diurnal patterns offers further insight. Figure 20 illustrates changes in the three-year averages of hourly PM2.5 concentrations recorded during the winter months of November through February between the 20 05-2007 and 2015-2017 time periods at the four selected sites. The overall diurnal patterns have not changed, yet hourly concentrations have decreased throughout the day. Peak daytime concentrations decreased approximately 13 to 27 percent and peak nighttime concentrations decreased approximately 13 to 32 percent.

33

Figure 20. Changes in average winter (November-February) PM2.5 hourly concentrations (2005-2007 and 2015-2017).



Chemical Composition

Four monitoring sites in the SJV collect PM2.5 chemical composition data to support evaluation of long-term trends and to better quantify source impacts of PM2.5. Figure 21 illustrates the three-year average trends in individual PM2.5 components at Modesto, Fresno, Visalia, and Bakersfield. Between 2007 and 2009, the PM2.5 speciation network transitioned from one carbon analysis method to another (MetOne Total Optical Transmittance ( TOT) NIOSH 5040 carbon method to the URG 3000N/IMPROVE_A method). For trend analysis purposes, therefore, the total mass of carbon compounds (elemental carbon and organic material, also known as carbonaceous aerosols) was estimated as the difference between the measured PM2.5 mass and the inorganic components mass.

34

Figure 21. Trends in three-year average PM2.5 chemical components.



Ammonium nitrate, ammonium sulfate, and carbon compounds (carbonaceous aerosols) are the major constituents of PM2.5. On an annual average basis, concentrations of these key constituents have all shown significant decreases. Ammonium nitrate concentrations in the Valley declined about 40 to 51 percent between 2004 (using the 2002-2004 average) and 2017 (2015-2017 average). During the same time frame, concentrations of ammonium sulfate dec lined about 22 to 42 percent. Carbon compounds showed a wider range with declines ranging from 7 to 32 percent.

Since 2007, CARB has tracked concentrations of levoglucosan, a chemical marker of wood smoke, at the Modesto and Visalia monitoring sites. These data are useful for examining trends in PM2.5 mass from residential wood combustion. Figure 22 illustrates the trends in levoglucosan concentrations during the winter months of November through February. While concentrations fluctuate from year to year, there is a slight downward trend, suggesting that PM2.5 mass from residential wood smoke has decreased.

35

Figure 22. Trends in winter-averaged levoglucosan concentrations.

a) Modesto b) Visalia

The smoke from wood burning is made up of a complex mixture of gases and fine particles. Wood smoke also contains several harmful air pollutants including benzene, formaldehyde, acrolein, and PAHs. The health effects of smoke range from eye and respiratory tract irritation to more serious effects, including reduced lung function, bronchitis, exacerbation of asthma, adverse birth outcomes such as low birth weight, and some evidence for cardiovascular effects and premature death.

CARB published a study that examined the impact of District Rule 4901 (Yap and Garcia, 2015) which requires mandatory curtailment of residential wood burning when air quality is forecast to be poor. The study found that after the implementation of the wood burning regulation in the San Joaquin Valley in the winter, reductions were seen basin-wide in both fine (12 percent) and coarse (8 percent) particulate matter and the number of hospital admissions for cardiovascular disease in adults 65 and older dropped by 7 percent. In addition, hospitalizations for ischemic heart disease, a specific type of cardiovascular disease often known as coronary artery disease, dropped by 16 percent basin-wide. Reductions in rural areas were even higher for both categories of hospital admissions. The reductions were based on 2000 to 2006 data.

Emission Inventory

Reductions in direct PM2.5, NOx, and SOx emissions are key to effectively reducing PM2.5 concentrations. Figure 23 illustrates annual emission trends in the San Joaquin Valley Air Basin from 2000 through 201710 for PM2.5 and the two key precursors, NOx and SOx.

10 Historical 2000-2011 emissions are from the 2016 Ozone SIP baseline emission inventory

36

Figure 23. PM2.5 and PM2.5 precursor annual emission trends in the San Joaquin Valley.

NOx emissions have decreased by 400 tpd or 63 percent. Major reductions occurred in emissions from heavy-duty diesel trucks, stationary combustion sources, and other mobile sources (e.g., farm and off-road equipment and trains). On-road mobile emissions constitute over half of all NOx emissions in 2017, and has remained the dominant source category over this inventory period, down from 62 percent of NOx emissions in 2000. Emissions from both on-road mobile and stationary sources have declined over this period due to aggressive control programs by CARB and the District, respectively

Direct PM2.5 emissions decreased by 46 tpd or about 44 percent. Major reductions occurred in emissions from residential wood combustion, mobile sources, such as heavy-duty diesel trucks and off-road equipment, and entrained dust. The most significant decline occurred in on-road mobile sources with a 68 percent reduction. The largest contribution of PM2.5 emissions is made by areawide sources, which have been reduced by 44 percent from 2000 levels.

SOx decreased by 20 tpd or about 72 percent. Major reductions occurred in emissions from stationary fuel combustion sources and industrial processes, driven by reductions in the allowable sulfur content of mobile and stationary source fuel streams.

The combined downward trends in PM2.5 components and emissions of PM2.5, NOx, and SOx indicate that the ongoing control programs have had substantial benefits improving air quality in the SJV and that further emission reductions in the future are expected to provide continuing progress towards attaining the PM2.5 standards.

Effectiveness of Emission Controls

NOx controls

Programs aimed at reducing NOx emissions have played an important role in reducing nitrate concentrations and, consequently, overall PM2.5 concentrations in the Valley. As discussed above, studies have identified NOx as the limiting precursor for ammonium nitrate formation.

37

As a result, NOx emissions, ambient NOx concentrations, and PM2.5 nitrate levels track each other over the years. Figure 24 and Figure 25 illustrate the i nfluence of emission controls on ambient NOx concentrations and PM2.5 nitrate at Bakersfield and Fresno.

Figure 24. Comparison of trends in NOx emissions and ambient NOx concentrations.


Modesto NOx monitoring discontinued after 2005


38

Figure 25. Comparison of trends in NOx emissions and ambient PM2.5 nitrate concentrations.



A declining trend observed for ambient NOx and PM2.5 nitrate is consistent with reductions in NOx emissions, except for three years, 2014 through 2016, when meteorological conditions conducive to high PM2.5 persisted, extending PM lifetime and resulting in higher concentrations than expected based on emissions. Table 5 summarizes the ef fects of reductions in NOx emissions on ambient NOx and PM2.5 nitrate concentrations in Kern and Fresno Counties.

Table 5. Percent reduction in NOx emissions and ambient NOx and PM2.5 nitrate concentrations between 2004 and 2017.

Indicator Percent Reduction

Fresno County Kern County NOx Emissions 59 62

Ambient NOx Concentrations 51 53 PM2.5 Nitrate Concentrations 51 45

39

PM2.5 controls

Carbon compounds are a major component of PM2.5 annually and on peak PM2.5 days. They include POAs that are directly emitted into the atmosphere and SOAs that are formed in the atmosphere through the oxidation of gaseous precursors. Sources of POAs include combustion of fossil fuels, meat cooking, biomass burning, and mobile sources. SOAs are formed in the atmosphere by oxidation of VOCs.

The major sources of primary carbon compounds in the Valley are combustion sources such as residential fuel combustion, open burning, diesel and gasoline exhaust, and meat cooking. Emissions from combustion decreased about 55 percent between 2004 and 2017 (Figure 26(a)), while during the same time frame, PM2.5 concentrations of carbon compounds decreased 16 percent (Figure 26(b)).

Figure 26. Trends in San Joaquin Valley PM2.5 combustion emissions and carbon compounds concentrations.

40

a) PM2.5 combustion emissions

b) Carbon compound concentrations

The discrepancy between combustion emissions trends and concentrations of carbon compounds is due to several factors. First, emission trends are based on POAs while the concentration trends reflect both POAs and SOAs. SOAs include biogenic compounds which are not likely to change over the years. Second, concentrations of carbon compounds are heavily impacted by localized and episodic events and therefore do not closely match basin-wide emission estimates. Finally, most of the carbon reductions are achieved by implementing consumer-based programs which are influenced by consumer behavior.

SOx controls

Between 2004 and 2017, SOx emissions in the SJV decreased more than 50 percent. A smaller declining trend of 22 to 43 percent was observed for PM2.5 sulfate concentrations. Figure 27 shows examples of the effects of large reductions in SOx emissions on PM2.5 sulfate concentrations at Modesto, Fresno, Visalia, and Bakersfield.

41

Figure 27. Comparison of trends in SOx emissions and ambient PM2.5 sulfate concentrations at Modesto, Fresno, Visalia, and Bakersfield.

As SOx is emitted primarily during fuel combustion, emission control efforts have focused largely on reducing the content of sulfur in fuels. California has required the use of ultra-low sulfur diesel fuel for on-road vehicles since 2006. Off-road diesel fuel was required to transition to ultra-low sulfur by 2010. Railroad locomotive and marine diesel fuel was reduced to 500 ppm sulfur in 2007, and further reduced to ultra-low sulfur in 2012. By the end of 2014, all highway, off-road, locomotive, and marine diesel fuel produced was required to be ultra-low sulfur. SOx emissions from stationary sources have decreased due to improved industrial source controls and a switch from fuel oil to natural gas for electric generation and industrial boilers. Figure 28 illustrates trends in three-year average SOx emissions by category Valley-wide.

Figure 28. Trends in three-year average SOx emissions Valley-wide.

42

MODELED ATTAINMENT DEMONSTRATION

Photochemical modeling plays a crucial rule in demonstrating attainment of the national ambient air quality standards based on projected future year emissions. Currently, SJV is designated as a Serious nonattainment area for the 1997 annual 15 µg/m3 and 24-hour 65 µg/m3 PM2.5 standards with an attainment deadline 2020 for both standards. SJV is also designated as a Serious nonattainment area for the 2006 24-hour 35 µg/m3 PM2.5 standard with an attainment deadline of 2024. In addition, SJV is designated as a Moderate nonattainment area for the 2012 annual 12 µg/m3 PM2.5 standard; however, the District applied for a reclassification from a Moderate to Serious nonattainment area, which will extend the attainment deadline to 2025. Consistent with U.S. EPA guidance for model attainment demonstrations (U.S. EPA, 2014), photochemical modeling was used to project PM2.5 design values (DVs) to the future as follows:

1.) 2020 annual and 24-hour PM2.5 DVs at each monitoring site in the Valley to show attainment of the annual 15 µg/m3 and 24-hour 65 µg/m3 PM2.5 standards;

2.) 2024 24-hour PM2.5 DVs at each monitor in the Valley to demonstrate attainment of the 24-hour 35 µg/m3 PM2.5 standard; and

3.) 2025 annual PM2.5 DVs at each monitor in the Valley to demonstrate attainment of the annual 12 µg/m3 PM2.5 standard.

The findings from the model attainment demonstration are summarized below. A detailed description of the model inputs, modeling procedures, and attainment test can be found in Appendix K and Appendix L of the 2018 PM2.5 Plan.

The current modeling approach draws on the products of large-scale, scientific studies as well as past PM2.5 SIPs in the region, collaboration among technical staff at state and local regulatory agencies, and from participation in technical and policy groups in the region (see Appendix L of the 2018 PM2.5 Plan for further details). In this work, the Weather Research and Forecasting (WRF) model version 3.6 was utilized to generate the annual meteorological fields. The Community Multiscale Air Quality (CMAQ) Model version 5.0.2 with state-of-the-science aerosol treatment was used for modeling annual PM2.5 in the Valley. Other model inputs and configuration, including the modeling domain definition, chemical mechanism, initial and boundary conditions, and emission processing can be found in Appendix J and Appendix L in the 2018 PM2.5 Plan.

The U.S. EPA modeling guidance (U.S. EPA, 2014) recommends using modeling in a “relative” rather than “absolute” sense. Based on analysis of recent years’ ambient PM2.5 levels and meteorological conditions leading to elevated PM2.5 concentrations, the year 2013 was selected for baseline modeling calculations. In particular, in 2013, SJV experienced one of the worst years for PM2.5 pollution in the Valley within the last decade.

Specifying the baseline design value is a key consideration in the model attainment test, because this value is projected forward to the future and used to test for future attainment of the standard at each monitor. To minimize the influence of year-to-year variability in demonstrating attainment, the U.S. EPA modeling guidance recommends using the average of three DVs, where one of the DV years is the same as the baseline emissions inventory and modeling year. This average DV is referred to as the baseline (or reference) DV. Here, the

43

average DVs from 2012, 2013, and 2014 are used to calculate baseline DVs (see Tables 6-10 for the baseline DVs utilized in the attainment demonstration modeling).

In order to use the modeling in a relative sense, five simulations were conducted: 1) base year simulation for 2013, which demonstrated that the model reasonably reproduced the observed PM2.5 concentrations in the Valley; 2) reference (or baseline) year simulation for 2013, which was the same as the base year simulation, but excluded exceptional event emissions such as wildfires; and 3) future year simulations for 2020, 2024, and 2025. These simulations were the same as the reference year simulation, except projected anthropogenic emissions for 2020, 2024, and 2025 were used in lieu of the 2013 emissions.

Table 6 shows the 2013, 2020, 2024, and 2025 SJV annual anthropogenic emissions for the five PM2.5 precursors calculated from the model-ready emissions inventory. From 2013 to 2020, anthropogenic emissions in the SJV are estimated to drop approximately 35 percent, 8 percent, 6 percent, 8 percent, and 1 percent for NOx, reactive organic gases (ROG),11

primary PM2.5, SOx, and NH3, respectively. Compared to 2013, anthropogenic emissions in the SJV in 2024 will drop by 63 percent, 9 percent, 12 percent, 6 percent, and 1 percent for NOx, ROG, primary PM2.5, SOx, and ammonia, respectively. Relative to 2013, anthropogenic emissions in the SJV in 2025 will reduce by 64 percent, 9 percent, 11 percent, 6 percent, and 1 percent for NOx, ROG, primary PM2.5, SOx, and ammonia, respectively. Among these five precursors, anthropogenic NOx emissions show the largest relative reduction, dropping from 288.2 tpd in 2013 to 187.1 tpd in 2020,107.6 tpd in 2024, and 104.6 tpd in 2025. Note that the emission totals presented in Table 6 were calculated from the modeling inventory based on CEPAM version 1.05.

Since the modeling inventory includes day-specific adjustments not included in the planning inventory, the planning and modeling inventories are expected to be comparable, but not identical. In addition, the 2024 and 2025 emission totals in Table 6 are from the attainment inventory, and so include additional emission reductions beyond the future baseline inventory for the respective year. Details about these additional emission reductions can be found in Appendix K of the 2018 PM2.5 Plan, while the actual emission commitments are outlined in Chapter 4 of the Plan.

11 ROG is similar, although not identical, to U.S. EPA’s term “VOC.” CARB’s inventory tracks ROG as a subset of total organic gases (TOG).

44

Table 6. SJV model-ready annual emissions for 2013, 2020, 2024, and 2025.

Category NOx ROG PM2.5 SOx NH3

2013 (tons/day) Stationary 38.5 90.8 8.5 7.2 13.9 Area 8.1 153.3 40.2 0.3 310.0 On-road Mobile 154.6 45.1 5.7 0.6 4.4 Other Mobile 87.1 35.8 6.2 0.3 6.0 Total 288.2 325.0 60.5 8.4 334.3

2020 (tons/day) Stationary 28.5 95.1 8.4 6.5 15.2 Area 7.8 151.8 40.0 0.3 306.9 On-road Mobile 81.0 22.4 3.2 0.6 3.6 Other Mobile 69.8 28.7 5.4 0.3 6.0 Total 187.1 298.0 57.0 7.7 331.7 Total change from 2013 to 2020 -35% -8% -6% -8% -1%



In this relative approach, the fractional change (or ratio) in PM2.5 concentration between the modeled future year (i.e., 2020, 2024, or 2025) and modeled baseline year (or reference year, 2013) are calculated. These ratios are called relative response factors (RRFs). Since PM2.5 is comprised of different chemical species, which respond differently to changes in emissions of various pollutants, separate RRFs were calculated for individual PM2.5 species. In addition, because of potential seasonal differences in PM2.5 formation mechanisms, RRFs for each species were also calculated separately for each quarter.

The RRF for a specific PM2.5 component j for each quarter is calculated using the following expression:

[C]j, future RRFj= (1) [C]j, reference

Where for the annual PM2.5 standard, [C]j, future is the modeled quarterly mean concentration for component j predicted for the future year averaged over the 3x3 array of grid cells surrounding the monitor, and [C]j,reference is the same, but for the reference year simulation. For the 24-hour PM2.5 standard, [C]j, future is the mean concentration for component j (for the

45

top 10 percent of modeled PM2.5 days in a quarter) predicted at the single grid cell which contains the monitor, and [C]j,reference is the same, but for the reference year simulation.

The measured FRM/FEM PM2.5 must be separated into its various chemical components. Species concentrations were obtained from the four PM2.5 chemical speciation sites in the Valley. These four speciation sites are located at: Bakersfield-California Avenue, Fresno-Garland, Visalia-North Church, and Modesto-14th Street. Since not all of the FRM/FEM PM2.5 sites in the Valley have collocated speciation monitors, the speciated PM2.5 measurements at one of the four speciation sites were scaled based on information learned during CRPAQS.

Since the FRM PM2.5 monitors do not retain all of the PM2.5 mass that is measured by the speciation samplers, the U.S. EPA modeling guidance recommends using the SANDWICH approach (Sulfate, Adjusted Nitrate, Derived Water, Inferred Carbon Hybrid material balance) described by Frank (2006) to apportion the FRM PM2.5 mass to individual PM2.5 species based on nearby chemical speciation measurements. Based on completeness of the data, PM2.5 speciation data from 2010-2013 were utilized. For each quarter, percent contributions from individual chemical species to FRM/FEM PM2.5 mass were calculated as the average of the corresponding quarter from 2010-2013 for the annual standard calculation. For the 24-hour standard calculation, only the top 10 percent of measured PM2.5 days from that quarter were utilized for percentage calculations.

Projected 2020 annual and 24-hour PM2.5 DVs for each site are given in Tables 7 and 8, respectively. For the annual standard, the Bakersfield-Planz site has the highest projected DV at 14.6 µg/m3, which is below the 15 µg/m3 annual PM2.5 standard. For the 24-hour standard, the Bakersfield-California Avenue site has the highest projected DV at 47.6 µg/m3, which is also below the 65 µg/m3 24-hour PM2.5 standard. Since projecting future year PM2.5 DVs is performed by projecting individual PM2.5 components and then summing those components to get the total PM2.5, it is useful to examine the RRFs associated with individual components to evaluate how the changes in each component contributes to the overall change in PM2.5. From 2013 to 2020, there are modest reductions projected for ammonium nitrate, EC, and OM, a slight reduction in sulfate, and a slight increase in crustal material. The reduction in ammonium nitrate is a direct result of NOx emission reductions from 2013 to 2020. EC and OM reductions are primarily tied to the reduction in primary PM2.5 emissions from 2013 to 2020. Detailed RRFs and base/future year concentrations for each individual species can be found in Appendix K of the 2018 PM2.5 Plan.

46

Table 7. Projected future year 2020 annual PM2.5 DVs at each monitor.

Site AQS ID Name Base DV (µg/m3)

2020 Annual DV (µg/m3)

60290016 Bakersfield - Planz 17.2 14.6 60392010 Madera 16.9 14.2 60311004 Hanford 16.5 13.3 61072002 Visalia 16.2 13.5 60195001 Clovis 16.1 13.4 60290014 Bakersfield - California 16.0 13.5 60190011 Fresno-Garland 15.0 12.4 60990006 Turlock 14.9 12.5 60195025 Fresno - Hamilton & Winery 14.2 11.9 60771002 Stockton 13.1 11.4 60470003 Merced - S Coffee 13.1 10.9 60990005 Modesto 13.0 11.0 60472510 Merced - Main Street 11.0 9.3 60772010 Manteca 10.1 8.7 60192009 Tranquility 7.7 6.4

Table 8. Projected future year 2020 24-hour PM2.5 DVs at each monitor.


2020 24-hour DV (µg/m3)

60290014 Bakersfield – California 64.1 47.6 60190011 Fresno – Garland 60.0 44.3 60311004 Hanford 60.0 43.7 60195025 Fresno – Hamilton & Winery 59.3 45.6 60195001 Clovis 55.8 41.1 61072002 Visalia 55.5 42.8 60290016 Bakersfield – Planz 55.5 41.2 60392010 Madera 51.0 38.9 60990006 Turlock 50.7 37.8 60990005 Modesto 47.9 35.8 60472510 Merced – Main Street 46.9 32.9 60771002 Stockton 42.0 33.5 60470003 Merced – S Coffee 41.1 30.0 60772010 Manteca 36.9 30.1 60192009 Tranquility 29.5 21.5

Projected 2024 24-hour PM2.5 DVs for each monitor are given in Table 9. The Fresno-Hamilton & Winery site has the highest projected DV at 35.2 µg/m3, which meets the 35 µg/m3 24-hour PM2.5 standard (technically, the form of the 24-hour PM2.5 standard means that a DV needs to be less than 35.5 µg/m3 to demonstrate attainment). The reduction in future year DVs are primarily attributed to significant reductions projected for ammonium nitrate and EC, with modest reductions in OM. Because of the large reduction in NOx emissions from 2013 to 2024, significant reduction is projected for ammonium nitrate. Reductions in EC and OM are primarily due to emission reductions associated with primary PM2.5 emission sources such as residential wood combustion and commercial cooking.

47

Table 9. Projected future year 2024 24-hour PM2.5 DVs at each monitor.


2024 24-hour DV (µg/m3)

60290014 Bakersfield – California 64.1 33.5 60190011 Fresno – Garland 60.0 32.9 60311004 Hanford 60.0 30.3 60195025 Fresno – Hamilton & Winery 59.3 35.2 60195001 Clovis 55.8 30.8 61072002 Visalia 55.5 31.3 60290016 Bakersfield – Planz 55.5 30.1 60392010 Madera 51.0 30.3 60990006 Turlock 50.7 30.2 60990005 Modesto 47.9 29.1 60472510 Merced – Main Street 46.9 27.5 60771002 Stockton 42.0 28.6 60470003 Merced – S Coffee 41.1 24.3 60772010 Manteca 36.9 25.8 60192009 Tranquility 29.5 16.2

Projected future year 2025 annual PM2.5 DVs for each monitor are given in Table 10. The Bakersfield-Planz and Madera sites have the highest projected DV at 12.0 µg/m3, which meets the 12 µg/m3 annual PM2.5 standard. Similar to 2024, the reduction in 2025 annual PM2.5 DVs is largely due to significant reduction in ammonium nitrate and EC, with modest reductions in OM. As discussed previously, reductions in ammonium nitrate are a direct result of dramatic NOx reductions from 2013 to 2025. Reductions in EC and OM are primarily due to emission reductions from primary PM2.5 sources, such as residential wood combustion and commercial cooking.

Table 10. Projected future year 2025 annual PM2.5 DVs at each monitor.


2025 Annual DV (µg/m3)

60290016 Bakersfield - Planz 17.2 12.0 60392010 Madera 16.9 12.0 60311004 Hanford 16.5 10.5 61072002 Visalia 16.2 11.5 60195001 Clovis 16.1 11.4 60290014 Bakersfield - California 16.0 11.0 60190011 Fresno-Garland 15.0 10.4 60990006 Turlock 14.9 11.1 60195025 Fresno - Hamilton & Winery 14.2 10.0 60771002 Stockton 13.1 10.6 60470003 Merced - S Coffee 13.1 9.6 60990005 Modesto 13.0 9.9 60472510 Merced - Main Street 11.0 8.6 60772010 Manteca 10.1 8.0 60192009 Tranquility 7.7 5.5

To evaluate the impact of reducing emissions of different PM2.5 precursors to PM2.5 DVs, a series of model sensitivity simulations were performed, for which anthropogenic emissions within the SJV were reduced by a certain percentage from the baseline emissions. Following

48

U.S. EPA precursor demonstration guidance (U.S. EPA, 2016) as well as considering SJV’s control strategies, sensitivity runs involving 30 percent emission reductions were performed for NOx and direct PM2.5. For other precursors (i.e., ammonia, VOCs, and SOx), both 30 percent and 70 percent emission reductions were performed. In addition, sensitivity simulations were performed for the years 2013, 2020, and 2024. The key conclusion from the sensitivity runs is that in 2024, reductions of direct PM2.5 and NOx emissions will continue to have a significant impact on annual and 24-hour PM2.5 DVs, while reductions of ammonia, ROG, and SOx have a much smaller impact compared to that of direct PM2.5 and NOx.

The U.S. EPA attainment modeling guidance also recommends conducting an unmonitored area analysis to ensure that there are n o regions outside of the existing monitoring network that could exceed the standard if a monitor was present at that location. Following the U.S. EPA recommended methodology, this unmonitored area analysis shows that in 2020, every modeling grid cell within the SJV meets the 15 µg/m3 annual and 65 µg/m3 24-hour PM2.5 standards except for a small area surrounding the Lemoore military facility due to emissions from the operations at that facility. In 2024, every modeling grid cell within the SJV meets the 35 µg/m3 24-hour PM2.5 standard except for a few cells surrounding the Lemoore military facility and a few grid cells located to the southeast of the Fresno metropolitan area. CARB plans to assess the elevated ammonium nitrate and organic carbon levels in the region and if appropriate, enhance PM2.5 monitoring in this area. In 2025, every modeling grid cell within the SJV meets the 12 µg/m3 annual PM2.5 standard except for a few cells surrounding the Lemoore military facility and Visalia. Per U.S. EPA guidance, an unmonitored peak does not negate the monitored attainment demonstration.

SUMMARY Following U.S. EPA guidance, CARB staff employed multiple analytical methods in a WOE approach to support the modeled attainment demonstration for the 2018 PM2.5 Plan that predicts attainment of multiple PM2.5 air quality standards in the Valley. This WOE contains data-driven trends in air quality, chemical compositions of PM2.5, and emissions of direct PM2.5 and precursors. Together, these analyses show that the control strategy is effective and focused on the appropriate emissions sources to achieve PM2.5 emissions reductions. In addition, the WOE analyses show that the modeled attainment demonstration is reasonable in concluding the San Joaquin Valley will attain multiple federal PM2.5 standards by the applicable attainment deadlines.

49

REFERENCES

Blanchard, C.L., Roth, P.M., Tanenbaum, S.J., Ziman, S.D., and Seinfeld, J.H., 2000, The use of ambient measurements to identify which precursor species limit aerosol nitrate formation, J. Air & Waste Management Association, 50: 2073-2084.

Chen, J., Ying, Q., and Kleeman, M.J., 2010, Source apportionment of wintertime secondary organic aerosol during the California regional PM10/PM2.5 air quality study, Atmospheric Environment, 44(10), 1331-1340.

Chen, J., Lu, J., Avise, J.C., DaMassa, J.A., Kleeman, M.J., and Kaduwela, A.P., 2014, Seasonal Modeling of PM2.5 in California’s San Joaquin Valley, Atmospheric Environment, 92, 182-190.

Frank, N.H., 2006, Retained nitrate, hydrated sulfates, and carbonaceous mass in federal reference method fine particulate matter for six eastern U.S. cities, Journal of Air & Waste Management Association, 56, 500-511.

Fountoukis, C. and Nenes, A., 2007, ISORROPIA II: A computationally efficient aerosol thermodynamic equilibrium model for K+, Ca2+, Mg2+, NH4+, Na+, SO42-, NO3-, Cl-, H2O aerosols, Atmospheric Chemistry Physics, 7, 4639-4659.

Jacob, D.J., Crawford, J.H., Maring, H., Clarke, A.D., Dibb, J.E., Emmons, L.K., Ferrare, R.A., Hostetler, C.A., Russell, P.B., Singh, H.B., Thompson, A.M., Shaw, G.E., McCauley, E., Pederson, J.R., and Fisher, J.A., 2010, The Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS) Mission: Design, Execution, and First Results, Atmospheric Chemistry and Physics, 10(11), 5191-5212.

Kleeman, M.J., Ying, Q., and Kaduwela, A., Control strategies for the reduction of airborne particulate nitrate in California's San Joaquin Valley, Atmospheric Environment, 2005, 39, 5325-5341.

Lurmann, F.W., Brown, S.G., McCarthy, M.C., and Roberts, P.T., December 2006, Processes Influencing Secondary Aerosol Formation in the San Joaquin Valley during Winter, Journal of Air and Waste Management Association, 56, 1679-1693.

Markovic, M. Z., VandenBoer, T. C., Baker, K. R., Kelly, J. T., and Murphy, J. G., 2014, Measurements and modeling of the inorganic chemical composition of fine particulate matter and associated precursor gases in California's San Joaquin Valley during CalNex 2010, Journal of Geophysical Research - Atmosphere,119, 6853–6866, doi:10.1002/2013JD021408.

Meng, Z., Dabdub, D., and Seinfeld, J.H., July 1997, Chemical coupling between atmospheric ozone and particulate matter, Science, DOI:10.1126/science.277.5322.116, 116-119.

NOAA, 2016, Synthesis of Policy Relevant Findings from the CalNex 2010 Field Study, available at www.esrl.noaa.gov/csd/projects/calnex/synthesisreport.pdf

50

http://www.esrl.noaa.gov/csd/projects/calnex/synthesisreport.pdf

Parworth, C.L., Young, D.E., Kim, H., Zhang, X., Cappa, C.D., Collier, S., and Zhang, Q., 2017, Wintertime water-soluble aerosol composition and particle water content in Fresno, California, Journal of Geophysical Research, Atmosphere., 122, 3155-3170. Doi: 10.1002/2016JD026173.

Prabhakar, G., Parworth, C.L., Zhang, X., Kim, H., Young, D.E., Beyersdorf, A.J., Ziemba, L.D., Nowak, J.B., Bertram, T.H., Faloona, I.C., Zhang, Q., and Cappa, C.D., 2017, Observational assessment of the role of nocturnal residual-layer chemistry in determining daytime surface particulate nitrate concentrations, Atmospheric Chemistry Physics, 17, 14747-14770.

Pun, B.K., Balmori R.T.F, and Seigneur, C., 2009, Modeling wintertime particulate matter formation in Central California, Atmospheric Environment, 43, 402-409.

Pusede, S.E., Duffey, K.C., Shusterman, A.A., Saleh, A., Laughner, J.L., Wooldridge, P.J., Zhang, Q., Parworth, C.L., Kim, H., Capps, S.L., Valin, L.C., Cappa, C.D., Fried, A., Walega, J., Nowak, J.B., Weinheimer, A.J., Hoff, R.M., Berkoff, T.A., Beyersdorf, A.J., Olson, J., Crawford, J.H., and Cohen, R.C., 2016, On the effectiveness of nitrogen oxide reductions as a control over ammonium nitrate aerosol, Atmospheric Chemistry Physics, 16, 2575-2596.

Seinfeld, J.H., and Pandis, S.N., 2006, Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, 2nd Edition, Wiley, Inc., New York.

Skiles, M.J., Lai, A.M., Olson, M.R., Schauer, J.J., DeFoy, B., 2018, Source apportionment of PM2.5 organic carbon in the San Joaquin Valley using monthly and daily observations and meteorological clustering, Environmental Pollution, 237, 366-376.

Sun, K., Cady-Pereira, K., Miller, D.J., Tao, L., Zondlo, M.A., Nowak, J.B., Neuman, J.A., Mikoviny, T., Müller, M., Wisthaler, A., Scarino, A.J., and Hostetler, C.A., 2015, Validation of TES ammonia observations at the single pixel scale in the San Joaquin Valley during DISCOVER-AQ, Journal of Geophysical Research, 120, 5140-5154, doi:10.1002/2014JD022846.

U.S. EPA, 2014, Draft Modeling Guidance for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5 and Regional Haze, available at https://www.epa.gov/ttn/scram/guidance/guide/Draft_O3-PM-RH_Modeling_Guidance-2014.pdf

U.S. EPA, 2016, PM2.5 Precursor Demonstration Guidance, available at https://www.epa.gov/sites/production/files/2016-11/documents/transmittal_memo_and_draft_pm25_precursor_demo_guidance_11_17_16.pd f

Womack, C.C., Neuman, J.A., Veres, P.R., Eilerman, S.J., Brock, C.A., Decker, Z.C.J., Zarzana, K.J., Dube, W.P., Wild, R.J., Wooldridge, P.J., Cohen, R.C., and Brown, S.S., 2017, Evaluation of the accuracy of thermal dissociation CRDS and LIF techniques for atmospheric measurement of reactive nitrogen species, Atmospheric Measurement Techniques, 10, 1911-1926.

51

https://www.epa.gov/ttn/scram/guidance/guide/Draft_O3-PM-RH_Modeling_Guidance-2014.pdf

https://www.epa.gov/ttn/scram/guidance/guide/Draft_O3-PM-RH_Modeling_Guidance-2014.pdf

https://www.epa.gov/sites/production/files/2016-11/documents/transmittal_memo_and_draft_pm25_precursor_demo_guidance_11_17_16.pdf



Yap, P.S., and Garcia, C., 2015, Effectiveness of residential wood-burning regulation on decreasing particulate matter levels and hospitalizations in the San Joaquin Valley Air Basin, American Journal of Public Health, 105(4), 772-778.

52

APPENDIX C1. Analysis for the Exclusion of the April 11, 2010 PM2.5 Value Recorded at Bakersfield-Planz from the Modeling Analysis for the San Joaquin Valley 2018 PM2.5 Plan

53

Analysis for the Exclusion of the April 11, 2010 PM2.5 Value Recorded at Bakersfield-Planz from the Modeling Analysis for the

San Joaquin Valley 2018 PM2.5 Plan

Overview

A fine particulate (PM2.5) 24-hour average concentration of 107.8 micrograms per cubic meter (µg/m3) was measured using a Federal Reference Method (FRM) sampler at the Bakersfield-Planz monitoring site on April 11, 2010.12 A concentration of this magnitude is considered unusual during the spring and summer months when PM2.5 concentrations in the San Joaquin Valley (Valley) are generally lowest. This document summarizes CARB staff’s assessment that the April 11, 2010 PM2.5 data point, collected during a month that typically records low PM2.5 levels, is anomalous and was likely the result of a high wind event.

Due to its physical location and surroundings, the sampler located at the Bakersfield-Planz site is subject to high-wind generated dust more than other urban sites in the Valley. Review of PM2.5 levels typically observed at the Bakersfield-Planz site in the spring and summer months, coupled with an analysis of wind speed data, leads CARB staff to conclude that the value of 107.8 µg/m3 is most likely the result of high-wind generated dust. Exclusion of this data point from use in the attainment determination, including air quality modeling, would be consistent with provisions in the Exceptional Events Rule.13

Temporal and Spatial Assessment

PM2.5 concentrations recorded at air monitoring stations throughout the San Joaquin Valley follow a distinct seasonal pattern and are, in general, of similar magnitudes. From April through September, PM2.5 values average 9.5 µg/m3 Valleywide with an average April concentration at the Bakersfield-Planz site of 12.0 µg/m3.14 Data presented in Figure C1-1 show the highest PM2.5 concentrations recorded during the month of April across the entire Valley over the last 17 years. The chart illustrates the consistency of relatively low PM2.5 values measured across the Valley during the month of April. With the exception of the April 11, 2010 value recorded at Bakersfield-Planz, the highest April PM2.5 concentrations average about 27 µg/m3 with a maximum of 40 µg/m3.

12 The Bakersfield-Planz (AQS ID: 060290016) sampler is sited at the Neighborhood Scale per 40 CFR Part 58, Appendix D. For this spatial scale of representativeness, the approximate extent of expected uniform pollutant concentrations around the sampler is 0.5 to 4 kilometers (0.31 to 2.5 miles) 13 81 FR 68216 (October 3, 2016) 14 Based on AQS data from 2010 through 2016

55

Figure C1-1. Peak 24-hour average PM2.5 concentrations recorded each April from 2000 through 2016 from PM2.5 monitoring sites in the San Joaquin Valley.

The anomalous nature of the April 10 value is noted both Valleywide as well as from the Planz site specifically (Figure C1-2). A plot of Bakersfield-Planz PM2.5 maximum average values collected each April from 2000 through 2016 indicates the April 10, 2011 value of 107.8 µg/m3 is more than three times the second highest observed April PM2.5 value of 32.3 µg/m3, which occurred in 2008.

Figure C1-2. Peak 24-hour average PM2.5 concentrations recorded each April from 2000 through 2016 at Bakersfield-Planz.

56

Spatially, the highest PM2.5 concentrations recorded on April 11, 2010 at either CARB or San Joaquin Valley Air Pollution Control District (District) air monitoring sites in the Valley, other than the Bakersfield-Planz site, ranged from 1.8 µg/m3 to 26.8 µg/m3 (Figure C1-3). The Bakersfield-California monitoring site is located 4.3 miles from, and is closest to, the Bakersfield-Planz site.

Figure C1-3. Peak 24-hour average PM2.5 concentrations recorded at San Joaquin Valley sites on April 11, 2010.

Analysis of Wind Speed Data and Additional PM data

As compared with other PM2.5 monitoring locations in the Valley, the location of the Planz site suggests that windblown dust may have been the critical factor contributing to high levels recorded on April 11. An analysis of meteorological data, coupled with an assessment of nearby dust sources close to the sampler, provides corroborating evidence that the April 11, 2010 value is anomalous and inconsistent with expected levels at the Planz monitoring station for the month of April.

Since the Bakersfield-Planz monitoring site is comprised a single, stand-alone FRM PM2.5 sampler, meteorological data are not available from the site for analysis. Data from the Bakersfield-California monitoring site were instead used to assess wind speed in the area near the sampler.15

The Bakersfield-California site recorded PM10 concentrations of 238 µg/m3 and PM2.5 concentrations of 26.8 µg/m3. Comparing these two concentrations reveals that approximately 89 percent of the PM10 sample was comprised of coarse particulate, while 11 percent was comprised of fine particulate. Figure C1-4 illustrates a relationship between the high winds and elevated concentrations. Highest PM concentrations were recorded

15 The Bakersfield Municipal Airport site (AQS ID: 060292012) is 0.5 miles from the Planz site and includes measurements of wind speed and direction; however, the site did not begin operation until June 2012.

57

during the high wind event, between 1000 hours Pacific Standard Time (PST) and 1800 hours PST, with 93 percent of PM mass captured during these hours. Documentation supporting this PM10 exceptional event was submitted to U.S. EPA on September 5, 2011 with a request that it be excluded from regulatory determinations. Elevated PM2.5 concentrations recorded at Bakersfield-Planz were flagged in the U.S. EPA Air Quality System (AQS) Database as “high winds” event but no additional documentation was submitted to U.S. EPA.

Figure C1-4. Hourly PM2.5 and PM coarse concentrations and wind speeds at Bakersfield-California on April 11, 2010.

Note: PM coarse is calculated as a difference between PM10 and PM2.5.

Evaluation of meteorological data collected during the month of April between 2010 and 2016 at the two Bakersfield sites revealed that April 11, 2010 was the windiest day over the course of seven years. Figure C1-5 illustrates wind speeds on ten windiest days during the month of April, between 2010 and 2016. As shown in the figure, April 11, 2010 was the only day in April with winds exceeding 15 mph for six consecutive hours.

58

Figure C1-5. Ten days with highest wind speeds during the month of April between 2010 and 2016.

No PM2.5 speciation data was available for April 11, 2010 for the Bakersfield area. Since filters are only stored for five years, it was no longer possible to conduct XRF analysis to confirm elevated levels of fugitive dust on the filter. Documentation for this exceptional event was submitted to U.S. EPA on October 5, 2011, with a request that the data be excluded for regulatory purposes.

Potential Fugitive Dust Sources Impacting the Bakersfield-Planz Site

In order to determine possible reasons for a significant difference in PM2.5 concentrations between Bakersfield-Planz and Bakersfield-California, we examined their surrounding and sampler placement. The Bakersfield-California monitor is located in the City of Bakersfield and is surrounded by the mix of commercial and residential buildings (Figure C1-6). The Bakersfield-Planz monitor is located on the grounds of the Bakersfield Municipal Airport, a city-owned airport used for private, civil aviation. The airport also includes a helicopter landing area and general aviation refueling station located near the monitor. As shown in Figure C1-7 below, the monitor is closely surrounded on several sides by open areas with the potential of emitting dust during high winds events. These open parcels of land are located to the east, west, and south of the monitor and include the airport infield areas between taxis and runways. Since dust particles require a substantial amount of energy to lift off the ground and don’t remain suspended for a long time, the immediate surroundings will have the greatest potential to contribute to a fugitive dust event. Furthermore, while Bakersfield-California sampler is placed on the rooftop (Figure C1-8), the Bakersfield-Planz sampler is located on the ground (Figure C1-9). Bakersfield-Planz sampler placement combined with its surroundings, makes it more susceptible to measuring elevated PM2.5 during a fugitive dust event.

59

Figure C1-6. PM2.5 FRM sampler at Bakersfield-California in relation to the surroundings.

Note: Green marker indicates monitor location.

60

Figure C1-7. PM2.5 FRM sampler at Bakersfield-Planz in relation to the surroundings.16

Note: Green marker indicates monitor location.

Figure C1-8. PM2.5 samplers placed on Figure C1-9. PM2.5 FRM sampler at the rooftop at Bakersfield-California. Bakersfield-Planz.

16 Areas to the west of the monitor location as well as other areas between runways and taxiways were covered with wood chips for dust control between March and November 2017.

61

CARB staff’s analysis indicates that elevated PM2.5 concentrations of 107.8 µg/m3 measured on April 11, 2010 at the Bakersfield-Planz monitoring site was potentially the result of windblown dust due to high winds. The 24-hour wind rose recorded at the Bakersfield-California station on April 11, 2010, is shown imposed over the Bakersfield-Planz monitor in Figure C1-10 to represent the wind speeds and direction that probably occurred at the latter station on this date. This figure indicates that the highest velocity and most prolonged winds on this date originated from the southeast and potentially carried windblown dust from disturbed infield soil areas over a distance of 1,400 feet upwind of the monitor.

Figure C1-10. April 11, 2010 hourly wind speeds (m/sec) and directions.

62

Summary

Comparison of the 107.8 µg/m3 concentration measured on April 11, 2010 to values typical for this season as well as comparison to values measured throughout the region on the same day, indicate that this concentration is not representative of ambient air quality in the Valley. Analysis of wind speeds, PM2.5, and PM10 data at Bakersfield-California suggest that the Bakersfield area was impacted by a fugitive dust event due to high winds. Since the Bakersfield-Planz PM2.5 FRM sampler is located on the grounds of the Bakersfield Municipal Airport in close proximity to airport runways and open fields, it is more likely to be affected by windblown dust than other samplers in the area, including Bakersfield-California samplers. Since there is a clear causal relationship between the high wind and elevated concentrations and the event is not reasonably controllable or preventable, it is appropriate to remove this concentration from the data set prior to modeling in order to calculate a design value that better reflects the actual PM2.5 concentrations in the area.

63

APPENDIX C2. Analysis for the Exclusion of the May 5, 2013 PM2.5 Value Recorded at Bakersfield-Planz from the Modeling Analysis for the San Joaquin Valley 2018 PM2.5 Plan

65

Analysis for the Exclusion of the May 5, 2013 PM2.5 Value Recorded at Bakersfield-Planz from the Modeling Analysis for the

San Joaquin Valley 2018 PM2.5 Plan

Overview

On May 5, 2013, a PM2.5 concentration of 167.3 µg/m3 was measured at the Bakersfield-Planz monitoring site. A concentration of this magnitude is extremely unusual, especially during the spring/summer when PM2.5 concentrations are typically low. Further, all other Bakersfield PM2.5 monitors recorded concentrations that were substantially lower, consistent with levels typical for the Valley during this time of year. Meteorological analysis shows that high winds on May 5, 2013 may have resulted in microscale PM2.5 impacts at Bakersfield-Planz that are atypical from measured concentrations at the Bakersfield-Planz site and other nearby sites during similar events. Elemental analysis of particulates collected on the filter indicated an extraordinarily high concentration of elements associated with windblown dust. This unusual measured concentration indicates that the sample collected on May 5, 2013 was not representative of the broader spatial scale the Bakersfield-Planz monitor is intended to capture. Based on the following analysis, CARB staff is therefore excluding the value of 167.3 µg/m3 from use in the modeling analysis for the SJV 2018 PM2.5 Plan.

Representativeness of Bakersfield-Planz PM2.5 Data

Air quality planning begins with evaluating pollutant concentrations measured at air monitoring stations and comparing those measurements to established air quality standards. In practice, monitors are only capable of sampling a relatively small portion of the atmosphere in the immediate vicinity around the inlet; however, the samples are intended to be representative of concentrations over a larger area as defined by the spatial scale of the monitoring site.

The Bakersfield-Planz monitoring site is identified as neighborhood scale, meaning that PM2.5 measurements are expected to be representative of air quality within an area that has relatively homogenous land use ranging from 0.5 to 4.0 kilometers around the monitor. If measurements at the site are overwhelmed by local dust sources and driven by unusual meteorological events atypical of the area, the measurements may no longer be considered representative of air quality within the broader area around the monitor.

San Joaquin Valley Seasonal PM2.5 Concentrations

PM2.5 concentrations throughout the Valley follow the same seasonal pattern. During the low concentration season (primarily April through September), concentrations are generally below 25 to 30 µg/m3 Valley-wide. A measured concentration of 167.3 µg/m3 in May is therefore extraordinarily unusual. Evaluating days where wind speeds were similar in magnitude shows that PM2.5 values measured during those days were much lower than the 167.3 μg/m3 measured on May 5, 2013.

67

Data presented in Table C2-1 illustrate the typical observed pattern and show the highest PM2.5 concentrations recorded between April and September in the Valley over the last 14 years. Apart from the May 5, 2013 value, flagged data, and an anomalous reading in April 2010 ( discussed in Appendix C1), other recorded P M2.5 values are consistently low during the April to September time period.

Table C2-1. Highest SJV PM2.5 concentrations April-September 2000-2013 (µg/m3).

* Bakersfield-California BAM recorded a 26.8 µg/m3 daily average; FRM value not available.

On May 5, 2013 all other monitoring sites in the Valley measured PM2.5 typical of the low concentration season. Measurements ranged from 9.9 μg/m3 to 24 μg/m3 (Table C2-2). The Bakersfield-California monitoring s ite recorded 24, 23, and 26 μg/m3 on the PM2.5 Federal Reference Method (FRM) monitor, and primary and collocated Beta Attenuation Monitors (BAM), respectively. As seen i n Table C2-2, the Bakersfield-Planz site recorded the highest 24-hour average PM2.5 concentration in the Valley on May 5, 2013, with levels an order of magnitude higher than any other site.

68

Table C2-2. PM2.5 FRM and FEM concentrations in the San Joaquin Valley on May 5, 2013.

Elemental Species Composition

To further evaluate the representativeness of the May 5, 2013 sample, CARB’s Monitoring and Laboratory Division analyzed the FRM filter using X-Ray Fluorescent Spectroscopy (XRF). The analysis revealed that the PM2.5 mass was heavily dominated by fugitive dust. In order to estimate the fugitive dust contribution to the total PM2.5 mass, CARB staff used the IMPROVE formula:

(2.2 x Al) + (2.49 x Si) + (1.63 x Ca) + (2.42 x Fe) + (1.94 x Ti)

The fugitive dust concentration, estimated at 107.7 μg/m3, far exceeded the values typically seen in the PM2.5 size fraction. The recorded value of 107.7 μg/m3 was over four times higher than the next highest value of 26.2 μg/m3 observed in the entire California network based on 14 years of available data. The PM2.5 fraction of fugitive dust is generally low, and PM2.5 concentrations during high wind events are thus typically not nearly as high as the May 5, 2013 reading.

Concentrations of total elemental species were also unusually high, about 6.6 μg/m3. Some of these species, such as cobalt, manganese, phosphorus, and rubidium, reached levels not previously measured in the State. These unusual concentration levels suggest that, along with fugitive dust, elemental species in the soil, combined with other chemical species, were deposited onto the filter. Figure C2-1 below compares average and maximum concentrations for select species historically measured at Bakersfield-California to what was measured at Bakersfield-Planz on May 5, 2013.

69

Figure C2-1. Comparison of concentrations of select soil elements, Bakersfield-California and Bakersfield-Planz.

Potential Fugitive Dust Sources Impacting the Bakersfield-Planz Site

To evaluate the potential influence of local fugitive dust sources on the Bakersfield-Planz monitor on May 5, 2013, the location of open soil areas, stationary sources, and known dust-generating activities were reviewed relative to the monitoring site. This information, coupled with observations of potential dust sources made by District enforcement staff on December 18, 2014, is summarized below.

The Bakersfield-Planz monitor is located on the grounds of the Bakersfield Municipal Airport, a city-owned airport used for private, civil aviation. The airport also includes a helicopter landing area near the monitor and helicopters are known to periodically use the airport. As shown in Figures C2-2 and C2-3 below, the monitor is closely surrounded on several sides by open areas with the potential of emitting dust during high wind events. These open parcels of land are located to the east, west, and south of the monitor and include the airport infield areas between taxiways and runways. Dust sources located nearest to the monitor have the greatest potential impact because dust particles do not remain suspended and deposit quickly. Additionally, as discussed above, the PM2.5 fraction of fugitive dust is generally low; therefore, the abnormally high value of 167.3 μg/m3 measured on May 5, 2013 is unusual.

70

Figure C2-2. Bakersfield-Planz PM2.5 FRM monitor.

Note: Photo taken looking west.

Figure C2-3. Aerial photo of Bakersfield Municipal Airport.

Note: Red marker indicates monitor location.

Additional potential sources of fugitive dust in the broader area surrounding the airport were also evaluated through field investigation by District enforcement staff. A review of aerial photos, combined with field investigations, indicate that additional potential dust emitting sources in the area are present to the east and southeast of the Bakersfield Municipal Airport (Figure C2-4). These sources are subject to District rules for controlling fugitive dust from construction and demolition activities; handling, storage and transport of storage of bulk materials; disturbed open areas; paved and unpaved roads; and off field agricultural sources.

71

Figure C2-4. Aerial photo of Bakersfield Municipal Airport - potential nearby fugitive dust sources.

Note: For reference, the distance from the monitor to the trucking yard is approximately 1 kilometer. Note: Red marker indicates monitor location.

No violations of District fugitive dust rules were documented at any nearby dust emitting facilities in Figure C2-4 on May 5, 2013. Based on this assessment of fugitive dust sources surrounding the monitor, the likely source of total particulate mass is from the open areas immediately adjacent to the monitor, reflecting a localized microscale impact; however, due to the significance difference in readings from nearby monitors, other factors may have contributed to the unusually high reading at Planz.

Meteorology at the Bakersfield-Planz Monitoring Site

An evaluation of Bakersfield area meteorology indicates that high winds measured at the airport are the expected cause of the localized dust impact on May 5. Wind speed data for the Bakersfield-Municipal Airport monitoring site was used to assess winds at Bakersfield-Planz. The Bakersfield-Municipal Airport meteorological site is located on the northern edge of the airport property, approximately one-half mile from the Bakersfield-Planz monitor. Strong winds on May 5, 2013 included 9 hours (including eight consecutive hours) exceeding U.S. EPA’s Exceptional Events high wind threshold of 25 mph, and the San Joaquin Valley’s Exceptional Events threshold of 17 mph as established in prior U.S. EPA-approved

72

Exceptional Event submissions. Figure C2-5 illustrates the difference between wind speeds on May 5, 2013 and a typical day in May of 2013. By contrast, wind speeds were notably lower at the Bakersfield-California monitoring station, located about 4 miles from Bakersfield-Planz.

Figure C2-5. May 2013 wind speeds at Bakersfield-Planz by hour.

To evaluate wind speeds on May 5, 2013, relative to other significant wind event days at Bakersfield-Planz, wind speed data were reviewed from the first day meteorological data were collected at the Bakersfield-Municipal Airport site on September 11, 2012, through December 31, 2014. During that 2 year and 3 month period, there were 3 days that included sustained winds over 25 mph (Figure C2-6). Among these high wind days, May 5, 2013, had over 8 hours with winds in excess of 25 mph, a significantly greater amount of time than the next highest day of December 11, 2014, with about 4 hours of sustained winds over 25 mph.

It should be noted that May 5, 2013 was the only high wind day during the dry season in the San Joaquin Valley. The other high wind days occurred during winter months, when moisture in the ground would minimize the potential for fugitive dust to become airborne. PM2.5 concentrations were measured only on one of these winter days, January 23, 2014, and reached 49.7 μg/m3, which is fairly typical for PM2.5 concentrations during winter in the Valley.

73

Figure C2-6. Days with high wind speeds at Bakersfield-Planz by hour.

The available meteorological data indicate that May 5, 2013 was highly unusual in terms of wind speed and the duration of high winds as compared with other days in which wind speed was measured at the airport.

Conclusion

In summary, comparison of the 167.3 μg/m3 concentration measured on May 5, 2013, to values typical for this season as well as comparison to values measured throughout the Valley on the same day, combined with the record high fugitive dust and elemental species concentrations, indicate that the monitor was impacted by microscale sources that are not representative of the neighborhood spatial scale the monitor is intended to represent. Therefore, this value is not included in modeling analysis for the San Joaquin Valley 2018 PM2.5 Plan.

74

APPENDIX C3. Source Apportionment of PM2.5 Measured at the Fresno and Bakersfield Chemical Speciation Network Sites in San Joaquin Valley Using the Positive Matrix Factorization

75

Source Apportionment of PM2.5 Measured at the Fresno and Bakersfield Chemical Speciation Network Sites in San Joaquin Valley Using the

Positive Matrix Factorization

Positive matrix factorization (PMF) is a multivariate source apportionment method that deduces source profiles as well as contributions from PM2.5 speciation data. PMF is one of several U.S. EPA recommended receptor modeling methods (US EPA, 2014). To identify major PM2.5 sources affecting the Fresno and Bakersfield monitoring site, PMF2 (bilinear PMF) was used in this analysis.

Sample Collection and Data Screening

The PM2.5 speciation samples were collected on a one-in-three day schedule at the Fresno-Garland and Bakersfield-California Ave. Chemical Speciation Network (CSN) monitoring sites located in the San Joaquin Valley (SJV). Because of relocation of the monitoring site, the PM2.5 speciation data at Fresno-Garland were available from January 2012. At Bakersfield, PM2.5 speciation samples were not collected between September 2013 and December 2014. The chemical analysis laboratory for the samples collected at both monitoring sites has been changed in November 2015. It resulted in the anomalous change of method detection limits (MDL) of chemical speciation. Therefore PM2.5 speciation data after November 20, 2015 at both monitoring sites were not included in this analysis.

The measurements of CSN PM2.5 mass concentrations by the speciation samplers had been discontinued since October 2014 and September 2013 at Fresno and Bakersfield, respectively. Comparing available CSN PM2.5 data with PM2.5 data measured by the collocated Federal Reference Method (FRM) samplers show reasonable agreement at Fresno (slope = 1.00; r2 = 0.97).and Bakersfield (slope = 0.98; r2 = 0.85). Therefore, CSN PM2.5 data were combined with FRM PM2.5 data and used in this analysis.

The Thermal Optical Reflectance (TOR) protocol has been used to analyze carbon mass collected on the CSN quartz filters for eight temperature resolved carbon fractions. This protocol volatilizes OC by four temperature steps in a helium atmosphere: OC1 at 120 °C, OC2 at 250 °C, OC3 at 450 °C, and OC4 at 550 °C. After OC4 response returns to baseline or a constant value, the OP (pyrolyzed OC) is oxidized at 550 °C in a mixture of 2 percent oxygen and 98 percent helium atmosphere prior to the return of reflectance to its original value. Then three EC fractions are measured in oxidizing atmosphere: EC1 at 550 °C, EC2 at 700 °C, and EC3 at 850 °C.

In previous analyses, PMF2 could not clearly separate carbonaceous particle sources, especially traffic-related combustion sources due to their similar chemical profiles and emission patterns. To separate combustion sources into gasoline vehicles and diesel emissions, the TOR eight carbon fractions were used in source apportionment analyses.

For the source apportionment analysis, samples were excluded from the data set for which the PM2.5 or carbon fraction data were not available, or for which PM2.5 or carbon fraction data were flagged for errors. Samples for which the sum of all measured species were larger

77

than the PM2.5 concentrations or for which the sum of all measured species were less than 50 percent of PM2.5 concentrations were excluded. Since sulfate (SO42-) and sulfur (S) are collected on different filters with different flow channels, and analyzed by different instruments, the SO42--to-S ratio has been used as an indicator of PM2.5 speciation data quality. Samples for which the SO42--to-S ratio was higher than 4 were excluded. Samples that contain fireworks particles collected on Independence Day and New Year’s Day were excluded since they had unusually high concentrations of metals. Overall, 38 percent and 53 percent of the samples were excluded at Fresno and Bakersfield, respectively.

For the chemical species screening, X-Ray Fluorescence (XRF) S was excluded from the analyses to prevent double counting of mass concentrations since XRF S and Ion Chromatography (IC) SO42- were highly correlated at Fresno (slope = 2.5, r2 = 0.94) and Bakersfield (slope = 2.6, r2 = 0.94). Due to the higher analytical precision compared to XRF Na and XRF K, IC Na+ and IC K+ were included in the analyses. Chemical species that had missing data more than 50 percent or below MDL values more than 80 percent were excluded. The species that had Signal-to-Noise (S/N; U.S. EPA, 2014) ratios below 0.5 were excluded except carbon fractions. Thus, a total of 318 samples and 27 species including PM2.5 mass concentrations collected between January 2012 and November 2015 were analyzed at Fresno. At Bakersfield, a total of 184 samples and 27 species including PM2.5 mass concentrations collected between January 2011 and November 2015 were analyzed.

The application of PMF2 depends on the estimated uncertainties which are based on the analytical uncertainties for each of the measured data. Since TOR carbon fractions were not accompanied by detection limit and uncertainty values, a comprehensive set of uncertainty structure (i.e., 7 percent of measured concentration) estimated by Kim et al. (2005) and 0.1 µg/m3 of detection limit value estimated from the State and Local Air Monitoring Stations (SLAM) speciation data were used in this analyses. Since EC1 concentration reported in TOR protocol includes OP concentration, OP was subtracted from EC1 and used as an independent variable in this analysis. Summaries of PM2.5 speciation data are provided in Tables C3-1 and C3-2.

To assign input data for PMF2, the measurement values were used for the input concentration data, and the sum of the analytical uncertainty and one-third of the detection limit value was used as the input uncertainty assigned to each measured value. Concentration values below the detection limit were replaced by one half of the detection limit values, and their input uncertainties were set at five-sixths of the detection limit values. Missing values were replaced by the geometric mean of the measured values for each species, and to down-weight these replaced data and then to reduce their influence on the solution, their accompanying uncertainties were set at four times this geometric mean value.

To estimate the potential directions of the local sources impacting the monitoring sites, the conditional probability function (CPF; Kim and Hopke, 2004) was calculated for each source using the source contribution estimates from PMF2 coupled with the wind data measured at the monitoring sites. The same 24-hour averaged contribution was assigned to each hour of a given day to match to the hourly wind data. For each source the CPF estimates the probability of PM2.5 transport from each wind direction. The PM2.5 sources are likely to be

78

located in directions that have high CPF values. In this analysis, from tests with several values of percentiles of the contribution and different azimuths of wind sectors, the upper 20 percent of the source contributions and 12 wind sectors of 30 degrees each were chosen. And the CPF value was estimated for each wind sector independently. Calm winds (< 1 m/sec) were excluded from this analysis due to the isotropic behavior of wind vane under calm condition.

Results and Discussions

Eight-source models without matrix rotation (rotational parameter FPEAK = 0) provided the most physically interpretable source profiles for the Fresno and Bakersfield sites. It was found necessary to increase the input uncertainties of K+ by a factor of two and NO3-, Cl, and Fe by a factor of three at Fresno to obtain physically interpretable PMF2 results (Paatero and Hopke, 2003). Similarly, the input uncertainties of NO3- were increased by a factor of four at Bakersfield. At both sites, the estimated uncertainties of OC1 were increased by a factor of four to reduce the influence of the known positive artifact from the adsorption of gaseous OC and the estimated uncertainties of OP and EC1 were increased by a factor of four for the additional uncertainty from the subtraction of OP.

79

Table C3-1. Summary of PM2.5 species mass concentrations at Fresno between 2012 and 2015.

Species Arithmetic

mean (µg/m3)

Geometric mean

(µg/m3)

Minimum (µg/m3)

Maximum (µg/m3)

Number of below MDL1

values (%) S/N ratio2

PM2.5 16.3849 12.4385 2.9000 78.6000 0 NA3

OC1 0.4878 0.2430 0.0017 3.2100 23.3 4.1 OC2 0.8237 0.6977 0.2110 3.3900 0 7.3 OC3 1.1344 0.9614 0.2350 3.6800 0 8.2 OC4 0.6191 0.5304 0.1570 1.7600 0 6.5 OP 0.5322 0.4012 0.0342 2.6800 2.2 5.6 EC1 1.2532 0.9501 0.1720 5.4200 0 4.1 EC2 0.0693 0.0562 0.0097 0.2630 86.8 0.3 EC3 0.0070 0.0500 0.0001 0.0598 100.0 0 SO42- 1.0949 0.9756 0.1910 3.2300 0 8.9 NO3 - 4.5391 2.2507 0.2170 41.0000 0 9.7 NH4+ 1.5513 0.8284 0.0257 15.2000 0.3 9.3 Al 0.0753 0.0520 0.0008 0.4470 21.1 2.4 Br 0.0051 0.0041 0.0005 0.0208 11.6 1.9 Ca 0.0452 0.0361 0.0036 0.2010 3.1 5.7 Cl 0.0563 0.0174 0.0001 0.6520 34.9 3.1 Cr 0.0051 0.0022 0 0.1640 70.1 0.6 Cu 0.0061 0.0041 0.0001 0.0313 23.9 2.5 Fe 0.1206 0.1009 0.0006 0.6440 0.3 9.8 K+ 0.0822 0.0631 0.0097 0.3610 12.6 4.6 Mg 0.0208 0.0150 0.0001 0.1670 60.1 0.8 Mn 0.0022 0.0017 0.0000 0.0112 56.6 0.5 Na+ 0.1256 0.0905 0.0130 0.9930 0.9 5.3 Ni 0.0059 0.0019 0 0.1400 58.8 1.5 Si 0.1905 0.1339 0.0090 1.0900 1.6 6.6 Ti 0.0062 0.0046 0.0001 0.0349 56.6 0.7 Zn 0.0105 0.0063 0.0002 0.1320 17.0 3.2 1 Minimum detection level 2 Signal-to-noise ratio (U.S. EPA, 2014) 3 not available

80

Table C3-2. Summary of PM2.5 species mass concentrations at Bakersfield between 2011 and 2015.

Species Arithmetic

mean (µg/m3)

Geometric mean

(µg/m3)

Minimum (µg/m3)

Maximum (µg/m3)

Number of below MDL1

values (%) S/N ratio2

PM2.5 17.8929 14.5913 4.6000 74.4000 0 NA3

OC1 0.2984 0.1767 0.0174 1.8500 29.3 3.3 OC2 0.7906 0.7285 0.2920 2.0800 0 7.6 OC3 1.0928 1.0004 0.3880 2.8900 0 8.5 OC4 0.6180 0.5651 0.1900 1.5900 0 6.7 OP 0.5057 0.4063 0.0599 3.2000 5.4 5.5 EC1 1.3159 1.1236 0.2250 5.3100 0.0 5.1 EC2 0.0637 0.0560 0.0225 0.2050 87.5 0.2 EC3 0.0052 0.0500 0.0002 0.0319 100.0 0 SO42- 1.3825 1.2046 0.2300 6.2100 0 9.0 NO3 - 5.0889 2.8290 0.3250 32.0000 0 9.8 NH4+ 1.7389 1.0697 0.0518 10.3000 0 9.6 Al 0.1236 0.0814 0.0055 0.6080 12.0 3.8 Ba 0.0131 0.0122 0.0001 0.0605 75.0 0.3 Br 0.0055 0.0047 0.0006 0.0156 5.4 2.1 Ca 0.1083 0.0824 0.0075 0.4350 0 8.2 Cl 0.0354 0.0158 0.0001 0.4410 27.2 2.7 Cr 0.0022 0.0017 0 0.0169 71.2 0.4 Cu 0.0085 0.0054 0.0001 0.0972 20.7 3.0 Fe 0.1914 0.1530 0.0205 0.8120 0 9.9 K+ 0.0823 0.0559 0.0152 1.2700 15.2 4.3 Mg 0.0275 0.0202 0.0005 0.0925 52.2 1.1 Mn 0.0034 0.0024 0.0001 0.0138 38.6 1.2 Na+ 0.1301 0.0959 0.0102 0.7420 1.1 5.4 Si 0.3304 0.2267 0.0007 1.6200 1.1 7.9 Ti 0.0099 0.0069 0.0002 0.0428 34.8 1.4 Zn 0.0095 0.0067 0.0001 0.0526 13.0 3.4 1 Minimum detection level 2 Signal-to-noise ratio (U.S. EPA, 2014) 3 not available

As shown in Figures C3-1 and C3-2, and Tables C3-3 and C3-4 which present average source contributions, secondary nitrate contributed the most at both sits (39 percent at Fresno; 37 percent at Bakersfield). Four major sources (i.e., secondary nitrate, secondary sulfate, gasoline vehicle and biomass burning) contributed 86 percent of PM2.5 concentrations

81

at Fresno and 77 percent at Bakersfield. The pie charts also show the average source contributions when PM2.5 concentrations were higher than 35 µg/m3 in the high PM2.5 season (Nov. - Feb.) The contribution from secondary nitrate was increased from 39 percent to 65 percent when PM2.5 was higher than 35 µg/m3 at Fresno. The biomass burning contributions were decreased slightly from 17 percent to 15 percent at Fresno site. At Bakersfield, the contribution from secondary nitrate was increased from 37 percent to 65 percent and gasoline vehicle contributions were decreased from 20 percent to 13 percent when PM2.5 was higher than 35 µg/m3. Figures C3-3 and C3-4 show monthly patterns in source contributions.

Figure C3-1. Average source contributions at Fresno.

Figure C3-2. Average source contributions at Bakersfield.

82

Comparisons of the reconstructed PM2.5 contributions (sum of contributions from all sources) with measured PM2.5 concentrations shown in Figures C3-5 and C3-6 indicate that the resolved sources effectively reproduce the measured values and account for most of the variation in the PM2.5 concentrations (slope = 0.92, r2 = 0.96 for Fresno data; slope = 0.89, r2 = 0.95 for Bakersfield data). The source profiles, corresponding source contributions, monthly variations of source contributions, and potential source directions are presented in Figures C3-11 through C3-18.

Figure C3-3. Monthly pattern in source contributions at Fresno.

Figure C3-4. Monthly pattern in source contributions at Bakersfield.

Secondary nitrate has high concentrations of NO3- and NH4+. It consisted of NH4NO3 and several minor species such as secondary OC and EC that transport together. It contributed the most at both sites accounting for 39 percent and 37 percent of the PM2.5 mass concentrations at Fresno and Bakersfield, respectively. Bakersfield showed higher secondary nitrate mass concentrations than Fresno. Secondary nitrate particles had

83

November – February high trends at both sites. Secondary nitrate has source directionality of east at Fresno and northeast and southwest at Bakersfield.

Secondary sulfate is identified by its high concentration of SO42- and NH4+. It consists of (NH4)2SO4, secondary OC, EC, and several minor species. It contributed 14 percent and 15 percent of the PM2.5 mass concentration at Fresno and Bakersfield, respectively. Secondary sulfate showed strong seasonal variation with higher concentrations in summer when the photochemical activity was high. The CPF plots for secondary sulfate pointed northeast at Fresno and west and southwest at Bakersfield.

Figure C3-5. Measured versus PMF2 estimated PM2.5 mass concentrations at Fresno.

84

Figure C3-6. Measured versus PMF2 estimated PM2.5 mass concentrations at Bakersfield.

Gasoline vehicle and diesel emissions were separated by different abundances of carbon fractions. Gasoline vehicle emissions have large amounts of OC3. In contrast, diesel emissions contain high concentrations of EC1. In Figure C3-7, the fractional carbon profiles of gasoline vehicle and diesel emissions at Fresno are compared with those at Bakersfield. They show similar carbon profiles to those presented across U.S. (Kim and Hopke, 2006). Gasoline vehicle emissions account for 16 percent and 20 percent of the PM2.5 mass concentration at Fresno and Bakersfield, respectively. Diesel emissions account for 3 percent and 5 percent of the PM2.5 mass concentration at Fresno and Bakersfield, respectively. Gasoline vehicles and diesel emissions showed winter-high seasonal variations. The CPF plots suggest high contributions from southeast and southwest for gasoline vehicles and northeast and east for diesel emissions at Fresno. At Bakersfield, high contributions were from northeast and southwest for gasoline vehicles and diesel emissions.

85

Figure C3-7. Comparisons of fractional carbon profiles (estimation ± standard deviation): a) Gasoline vehicle, b) Diesel emissions.

86

Table C3-3. Average source contributions (µg/m3) to PM2.5 mass concentration between 2012 and 2015 at Fresno.

Sources Average source contribution (± 95 percent distribution)

Secondary nitrate 6.19 (1.14)

Biomass burning 2.65 (0.31)

Gasoline vehicle 2.56 (0.20)

Secondary sulfate 2.14 (0.15)

Airborne soil 0.89 (0.09)

Aged sea salt 0.59 (0.08)

Diesel emissions 0.50 (0.06)

Industrial 0.23 (0.04)

Estimated PM2.5 (µg/m3) 15.74 (1.48)

Measured PM2.5 (µg/m3) 16.38 (1.57)

Biomass burning was characterized by high carbon fractions and K+. Biomass burning contributed 17 percent of the PM2.5 concentration at Fresno and 4 percent at Bakersfield. Biomass burning category reflects contributions from field burning, wild fire, residential wood burning, and cooking. Biomass burning showed winter-high variations suggesting that it is mostly contributed by residential winter heating at both sites. The CPF plot for biomass burning suggests high contributions from east at Fresno and northeast and southwest at Bakersfield.

87

Table C3-4. Average source contributions (µg/m3) to PM2.5 mass concentration between 2011 and 2015 at Bakersfield.

Sources Average source contribution (± 95 percent distribution)

Secondary nitrate 6.38 (1.41)

Gasoline vehicle 3.43 (0.28)

Secondary sulfate 2.61 (0.32)

Airborne soil 2.01 (0.27)

Aged sea salt 0.93 (0.13)

Diesel emissions 0.79 (0.11)

Biomass burning 0.70 (0.15)

Industrial 0.29 (0.07)

Estimated PM2.5 (µg/m3) 17.25 (1.76)

Measured PM2.5 (µg/m3) 17.89 (1.92)

Aged sea salt was represented by its high concentrations of NO3-, SO42-, and Na+ accounting for 4 percent the PM2.5 mass concentration at Fresno and 5 percent at Bakersfield. Aged sea salt reflects particles in which Cl- in the fresh sea salt is partially displaced by acidic gases during the transport and collected along with NO3- and SO42-. Aged sea salt has a strong summer-high seasonal pattern. The CPF plots for aged sea salt suggested high contributions from northwest at both Fresno and Bakersfield.

Airborne soil has high concentrations of Si, Fe, Al and Ca. It contributed 6 percent of the PM2.5 mass concentration at Fresno and 12 percent at Bakersfield. The airborne soil category reflects wind-blown dust as well as re-suspended crustal materials by road traffic. Airborne soil contribution showed September-high variations at both sites. Airborne soil identified at Fresno was strongly associated with southwest wind. At Bakersfield, airborne soil mostly contributed from southwest and southeast.

Industrial sources characterized by high concentrations of carbon fractions and metals were identified. This source accounts for 2 percent of the PM2.5 mass concentrations at both sites. It shows a winter-high trend. The CPF plot for the industrial source suggests high contributions from east and south at Fresno and southeast and southwest at Bakersfield.

88

Conclusions

PM2.5 speciation data and related meteorological data collected at the Fresno monitoring site between 2012 and 2015 and Bakersfield monitoring site between 2011 and 2015 were analyzed. To separate combustion sources into gasoline vehicles and diesel engines, eight carbon fractions were used in source apportionment analyses. Using PMF2, the multivariate source apportionment tool, eight major PM2.5 sources were identified: Secondary nitrate, secondary sulfate, gasoline vehicle, diesel emissions, biomass burning, airborne soil, aged sea salt, and industrial sources. Average source contributions showed that secondary nitrate, secondary sulfate, gasoline vehicle, and biomass burning contributed 86 percent of PM2.5 concentrations at the Fresno monitoring site and 77 percent at the Bakersfield monitoring site.

Updates

Recent PM2.5 speciation data were analyzed by PMF2. Because of change of the chemical analysis laboratory and change of the method detection limits (MDL) of chemical speciation in November 2015, PM2.5 speciation data only after December, 2015 at both monitoring sites were included in the analyses. NH4+ was not included in the analyses because NH4+ did not show the temporal correlations with NO3- or SO42-. A total of 210 samples and 25 species collected at Fresno, and a total of 151 samples and 25 species collected at Bakersfield between December 2015 and November 2017 were analyzed.

89

Figure C3-8. Average source contributions at Fresno and Bakersfield.

Similar to previous 2011-2015 analyses, eight-source models without matrix rotation provided the most interpretable results at both Fresno and Bakersfield sites. Figure C3-8 presents average source contributions between December 2015 and November 2017. Secondary nitrate contributed the most at both sites (29 percent at Fresno; 32 percent at Bakersfield). When the source contributions in 2011-2015 were compared with those in 2016-2017 in Table C3-5, the contributions from secondary nitrate decreased the most during 2016 and 2017 at Bakersfield (2.4 µg/m3) and Fresno (2.9 µg/m3), which led to decrease of PM2.5 mass concentrations at both sites. The trend of winter (November - February) contributions from three significant sources are compared in Figure C3-9. Secondary nitrate and biomass burning contributions decreased distinctively in winter of 2016 at Bakersfield and Fresno. As shown in Figure C3-10 in which levoglucosan concentrations in winter (November - February) are presented, the decreasing trend of biomass burning contributions in 2016 in Fresno coincides with the decreasing trend of levoglucosan in Modesto.

90

Table C3-5. Comparison of average source contributions to PM2.5 concentrations.

Sources

Changes in source contributions (µg/m3):

[2011-2015] vs [2016-2017]

Bakersfield Fresno

Secondary nitrate -2.37 -2.85

Secondary sulfate -0.61 -0.37

Gasoline -1.65 -0.06

Diesel 0.17 0.19

Biomass burning 0.36 -1.16

Airborne soil -0.22 0.07

Aged sea salt -0.47 0.14

Industrial 0.33 -0.13

91

Figure C3-9. Comparisons of median winter (Nov.- Feb.) contributions. a) Secondary nitrate. b) Gasoline vehicle. c) Biomass burning.

92

Figure C3-10. Comparisons of median winter (Nov.- Feb.) contributions of levoglucosan.

References

Kim, E., Hopke, P.K. Comparison between conditional probability function and nonparametric regression for fine particle source directions, Atmospheric Environment, 38(11), 4667-4673, 2004.

Kim, E., Hopke, P.K., Qin, Y. Estimation of organic carbon blank values and error structures of the speciation trends network data for source apportionments, Journal of Air and Waste Management Association, 55, 1190–1199, 2005.

Kim, E., Hopke, P.K. Characterization of fine particle sources in the Great Smoky Mountains area. Science of the Total Environment, 368, 781–794, 2006.

U.S. Environmental Protection Agency. EPA Positive Matrix Factorization (PMF) 5.0 Fundamentals & User Guide (EPA/600/R-14/108). Research Triangle Park, NC. April 2014. Available on Internet at https://www.epa.gov/sites/production/files/2015-02/documents/pmf_5.0_user_guide.pdf.

93

https://www.epa.gov/sites/production/files/2015

Figure C3-11. Source profiles deduced from PM2.5 samples measured at Fresno.

94

Figure C3-12. Source profiles deduced from PM2.5 samples measured at Bakersfield.

95

Figure C3-13. Source contributions deduced from PM2.5 samples measured at Fresno.

96

Figure C3-14. Source contributions deduced from PM2.5 samples measured at Bakersfield.

97

Figure C3-15. Monthly variations of source contributions to PM2.5 mass concentration at Fresno (median ± 95 percent distribution).

98

Figure C3-16. Monthly variations of source contributions to PM2.5 mass concentration at Bakersfield (median ± 95 percent distribution).

99

Figure C3-17. Conditional probability function plots for the highest 20 percent of the source contributions at Fresno.

100

Figure C3-18. Conditional probability function plots for the highest 20 percent of the source contributions at Bakersfield.

101

APPENDIX C4. Precursor Demonstrations for Ammonia, SOx, and ROG

103

Precursor Demonstrations for Ammonia, SOx, and ROG

INTRODUCTION Fine particulate matter (PM2.5) is made up of many constituent particles that are either directly emitted, such as soot and dust, or formed through complex reactions of gases in the atmosphere. Oxides of nitrogen (NOx), sulfur dioxide (SO2), volatile organic compounds (VOCs), and ammonia (NH3) are gases that are precursors to PM2.5, transforming into particles through physical and chemical atmospheric processes.

The United States Environmental Protection Agency (U.S. EPA) finalized a PM2.5 State Implementation Plan (SIP) Requirements Rule17 (Rule) that identifies the four PM2.5 precursor pollutants—NOx, SO2, VOCs, and ammonia—that “must be evaluated for potential control measures in any PM2.5 attainment plan.”18 The Rule permits air agencies to “submit an optional precursor demonstration designed to show that for a specific PM2.5 nonattainment area, emissions of a particular precursor from sources within the nonattainment area do not or would not contribute significantly to PM2.5 levels that exceed” the National Ambient Air Quality Standards (NAAQS).19 If the agency’s demonstration is approved by U.S. EPA, the attainment plan “may exclude that precursor from certain control requirements under the Clean Air Act.”20

This appendix includes precursor demonstrations for three PM2.5 precursors: ammonia, oxides of sulfur (SOx), and reactive organic gases (ROG). The California Air Resources Board (CARB) inventory tracks SOx rather than SO2 specifically, but SOx consists mostly of SO2. ROG is similar, although not identical, to U.S. EPA’s term “VOC.”21 CARB’s inventory tracks ROG as a subset of total organic gases (TOG). This appendix does not include a precursor demonstration for NOx, since NOx is an important and significant precursor to PM2.5 and is controlled extensively in the SIP, and because reductions of NOx emissions are essential to the attainment strategy for the San Joaquin Valley (Valley).

Following U.S. EPA guidance, the three precursor demonstrations analyze “the relationship between precursor emissions and the formation of secondary PM2.5 components”22 using an air quality model, and take into consideration additional relevant factors.

17 81 FR 58010 (August 24, 2016) 18 United States Environmental Protection Agency. PM2.5 Precursor Demonstration Guidance: Draft for Public Review and Comment. 17 Nov. 2016. Web. 3 Oct. 2017. <www.U.S. EPA.gov/sites/production/files/2016-11/documents/transmittal_memo_and_draft_pm25_precursor_demo_guidance_11_17_16.pdf>. Page 7 19 Ibid. 7 20 Ibid. 7 21 See: California Air Resources Board. “FACT SHEET #1: Development of Organic Emission Estimates For California's Emission Inventory and Air Quality Models.” Aug. 2000. Web. 24 May 2018. <www.arb.ca.gov/ei/speciate/factsheetsmodeleispeciationtog082000.pdf> See also: California Air Resources Board. “Definitions of VOC and ROG.” Jan. 2009. Web. 24 May 2018. <www.arb.ca.gov/ei/speciate/voc_rog_dfn_1_09.pdf> 22 U.S. EPA. PM2.5 Precursor Demonstration Guidance: Draft for Public Review and Comment. Page 26

105

www.arb.ca.gov/ei/speciate/voc_rog_dfn_1_09.pdf

www.arb.ca.gov/ei/speciate/factsheetsmodeleispeciationtog082000.pdf

https://EPA.gov/sites/production/files/2016

U.S. EPA PM2.5 PRECURSOR DEMONSTRATION GUIDANCE In November 2016, U.S. EPA published a draft guidance document to “assist air agencies who may wish to submit PM2.5 precursor demonstrations.”23 The document provides recommendations or guidelines, as authorized under the Clean Air Act, “that will be useful to air agencies in developing the precursor demonstrations by which the EPA can ultimately determine whether sources of a particular precursor contribute significantly to PM2.5 levels that exceed the standard in a particular nonattainment area.”24 Recommendations include modeling procedures for conducting the required analysis and contribution thresholds to determine the impact of a precursor on PM2.5 levels.25 The guidance also describes an analytical process to perform the precursor demonstration, involving a concentration-based analysis followed by a sensitivity-based analysis and consideration of additional information.

Concentration-Based Analysis The evaluation of precursors begins with a concentration-based analysis using ambient data to determine whether precursor emissions contribute to total PM2.5 concentrations.26 Each precursor’s impact on total PM2.5 mass is compared to contribution thresholds. U.S. EPA recommends values for these thresholds, or air quality concentrations below which air quality impacts are not statistically significantly different from “the inherent variability in the measured atmospheric conditions,” and thus do not contribute to PM2.5 concentrations that exceed the NAAQS.27 These thresholds are 0.2 micrograms per cubic meter (µg/m3) for the annual PM2.5 standard and 1.3 µg/m3 for the 24-hour PM2.5 standard.28

As shown below in Table C4-1, based on this metric, ammonia, SO2, and VOCs contribute to total PM2.5 mass in the Valley in amounts that exceed U.S. EPA’s recommended thresholds.

Table C4-1. Contribution of Ammonia, SO2, and VOCs to Total PM2.5.

Species Relevant Precursor

Species Contribution (µg/m3) to PM2.5 Mass* Over Threshold?

Ammonium nitrate Ammonia 5.2 Yes Ammonium sulfate SO2 1.6 Yes Carbonaceous aerosols VOCs 6.2 Yes * 2015 annual average for Bakersfield

This concentration-based analysis, however, does not accurately capture the impact of reductions of precursor emissions on PM2.5 levels. Since the concentration-based analysis shows the precursors contribute to total PM2.5 mass in amounts over U.S. EPA’s recommended thresholds, CARB proceeded to conduct an optional sensitivity-based analysis to demonstrate that reductions of ammonia, SOx, and ROG will have negligible impact on PM2.5.

23 Ibid. 7 24 Ibid. 7-8 25 Ibid. 9 26 Ibid. 8 27 Ibid. 14, 15 28 Ibid. 15-16

106

Sensitivity-Based AnalysisThe SIP Requirements Rule allows for a sensitivity-based analysis to examine the degree to which PM2.5 levels are sensitive to precursor reductions. According to the guidance:

This modeling analysis examines the sensitivity of ambient PM2.5 concentrations in the nonattainment area to certain amounts of decreases in the precursor emissions in the area…. Where decreases in emissions of the precursor result in negligible air quality impacts (i.e., the area is “not sensitive” to decreases), such a small degree of impact is not significant and can be considered to not “contribute” to PM2.5 concentrations for the purposes of determining whether control requirements should apply.29

Generally, U.S. EPA recommends that the precursor demonstration “should be based on current conditions to demonstrate that precursor emissions do not contribute significantly to PM2.5 concentrations in the nonattainment area.”30 This means evaluating emissions in a selected base year, which may be the present or a previous year.

For each existing PM2.5 monitor location in the area,31 the first step for estimating PM2.5 impacts from ammonia, SOx, or ROG in the base year is to estimate the average PM2.5 concentration on an annual and 24-hour basis. The second step is to calculate the annual and 24-hour average PM2.5 concentration at each monitor with a specified percent reduction in precursor emissions, still in the base year.32 The difference between these two calculated PM2.5 values is the impact on PM2.5 levels from precursor emissions reductions.33 Note that “precursor demonstrations do not examine changes in emissions between a base year and a future year. Instead, the calculation of relative changes in PM2.5 concentrations occur between a modeled case with all emissions and a modeled case with reduced precursor emissions” (emphasis added).34 In addition, U.S. EPA recommends modeling reductions of between 30 and 70 percent of precursor emissions.35

The third step in the sensitivity-based analysis is to compare the modeled impact on PM2.5 levels from a decrease in ammonia, SOx, or ROG emissions to contribution thresholds for annual and 24-hour PM2.5.36 If the calculated PM2.5 impact is greater than 0.2 µg/m3 for the annual standard or greater than 1.3 µg/m3 for the 24-hour standard, then PM2.5 levels are sensitive to the modeled percent reduction in ammonia, SOx, or ROG emissions.

Consideration of Additional Information To supplement modeling analysis, U.S. EPA guidance also allows an air agency to consider additional information, assessing the significance of a precursor “‘based on the facts and circumstances of the area.’”37 The guidance states:

29 Ibid. 25 30 Ibid. 33 31 Ibid. 16 32 Ibid. 36 33 Ibid. 36 34 Ibid. 34 35 Ibid. 29 36 Ibid. 25 37 Ibid. 17

107

If the estimated air quality impact exceeds the recommended contribution thresholds…, this fact does not necessarily preclude approval of the precursor demonstration. There may be cases where it could be determined that precursor emissions have an impact above the recommended contribution thresholds, yet do not “significantly contribute” to levels that exceed the standard in the area.38

In these cases, an air agency may “provide the [U.S.] EPA with information related to other factors they believe should be considered in determining whether the contribution of emissions of a particular precursor to levels that exceed the NAAQS is ‘significant’ or not.”39

Such factors may include: trends in emissions of other precursors such as NOx,40 anticipated growth or loss of emissions sources,41 and the consequent appropriateness of modeling impacts in a future year instead of a base year;42 “available emissions controls,”43 and “the severity of nonattainment at relevant monitors.”44 These factors are discussed in the context of the precursor analyses for the Valley in the subsequent sections.

Other factors the agency may consider are: the amount by which a precursor’s contribution exceeds the recommended contribution thresholds; source characteristics (e.g., source type, stack height, location); analyses of speciation data and precursor emission inventories; chemical tracer studies; and special intensive measurement studies to evaluate specific atmospheric chemistry in an area. The agency may also provide other information not listed here.45

The following sections contain sensitivity-based analyses and supplemental information demonstrating that ammonia, SOx, and ROG are not significant precursors to PM2.5 in the Valley.

38 Ibid. 17 39 Ibid. 17 40 Ibid. 17 41 Ibid. 17 42 Ibid. 33 43 Ibid. 29 44 Ibid. 17 45 Ibid. 17

108

AMMONIA ANALYSIS Ammonium nitrate (NH4NO3) is a constituent of PM2.5, making up about 40 percent of fine particulate matter mass in the Valley. Ammonium nitrate forms when nitrogen dioxide (NO2) reacts with highly oxidizing species in the atmosphere to form nitric acid (HNO3). Nitric acid then reacts with ammonia (NH3) to yield ammonium nitrate as a particle. Since ammonia reacts chemically in this way to form a particle, ammonia is a precursor to PM2.5.

Lowering PM2.5 concentrations to levels that meet the NAAQS will rely upon an effective control strategy for ammonium nitrate. The amount of ammonium nitrate that can form in the atmosphere is limited by whichever precursor, either NOx or ammonia, is in least supply, and research studies confirm that there are relatively fewer NOx molecules in the air in the Valley than ammonia. This implies that reducing NOx, the limiting precursor in this case, is more effective for reducing ammonium nitrate concentrations and thus improving PM2.5 air quality.

Following the analytical process outlined in the U.S. EPA precursor demonstration guidance and summarized above, CARB has evaluated ammonia in the Valley. The results of the sensitivity-based analysis and consideration of additional information are presented below.

Sensitivity-Based Analysis CARB staff used an air quality model to estimate the PM2.5 design value for the annual and 24-hour standards in the base year of 2013 at each Valley monitor. Then, CARB staff applied the recommended lower bound of a 30 percent reduction to ammonia emissions and used the air quality model to estimate the PM2.5 design values, as shown in Table C4-2. The difference between the two design values represents the modeled impact on PM2.5 levels of a 30 percent reduction in ammonia emissions in 2013. This is the value that is compared to U.S. EPA’s recommended contribution thresholds of 0.2 µg/m3 for the annual standard and 1.3 µg/m3 for the 24-hour standard to establish if PM2.5 levels are sensitive to this level of ammonia reduction.

109

Table C4-2. Base Year 2013 PM2.5 – 30 Percent Ammonia Reduction. Annual 24-Hour

Site* 2013

Baseline DV

2013 DV with 30% Ammonia

Reduction+ Difference

2013 Baseline

DV


Reduction Difference

Bakersfield-Planz 17.19 16.76 0.43 55.5 53.3 2.2 Madera 16.93 16.29 0.64 51.0 49.2 1.7 Hanford 16.54 15.82 0.72 60.0 57.8 2.1 Visalia 16.20 15.82 0.38 55.5 53.5 2.0 Clovis 16.12 15.80 0.32 55.8 54.0 1.9 Bakersfield-California 16.02 15.58 0.44 64.1 60.8 3.3 Fresno-Garland 14.98 14.69 0.29 60.0 58.0 2.0 Turlock 14.88 14.46 0.42 50.7 49.3 1.5 Fresno-HW 14.22 13.95 0.27 59.3 57.4 2.0 Stockton 13.14 12.84 0.30 42.0 41.0 1.0 Merced-S Coffee 13.10 12.65 0.45 41.1 40.0 1.1 Modesto 13.03 12.66 0.37 47.9 46.5 1.5 Merced-M 10.97 10.77 0.20 46.9 45.9 1.0 Manteca 10.09 9.85 0.24 36.9 36.0 0.9 Tranquility 7.72 7.33 0.39 29.5 27.2 2.2

* The site at Corcoran does not have a valid design value because of missing data, and is thus excluded from all precursor analyses.

+ Numbers may not sum exactly due to rounding.

For completeness, CARB staff repeated this analysis, applying instead the U.S. EPA-recommended upper bound of a 70 percent reduction to ammonia emissions in the base year, as shown in Table C4-3.

Table C4-3. Base Year 2013 PM2.5 – 70 Percent Ammonia Reduction. Annual 24-Hour

Site 2013

Baseline DV



2013 Baseline

DV




110

From this analysis, the estimated air quality impact of reducing ammonia emissions by the lower bound of 30 percent in the base year exceeds U.S. EPA’s recommended thresholds at all but a few Valley monitors, for both the annual and 24-hour standards. Reducing emissions by the upper bound of 70 percent also shows impacts above the thresholds.

It is not possible, however, to conclude from this analysis that emissions of ammonia “significantly contribute.” In this case, ammonia emissions have an impact above the recommended contribution thresholds even at the lower bound, but, as the U.S. EPA guidance indicates, this does not necessarily mean the precursor contributes significantly to PM2.5 levels that exceed the NAAQS. Making the appropriate determination about the ammonia emission reduction impact requires further analysis of additional factors.

Consideration of Additional Information To supplement modeling analysis, U.S. EPA guidance also allows an air agency to consider additional information, assessing the significance of a precursor “‘based on the facts and circumstances of the area.’”46 CARB staff believes that there are several critical factors that must be considered in determining whether ammonia is a significant precursor to PM2.5 in the Valley.

Emissions Trends and Studies CARB has an extensive suite of measures in place to reduce NOx emissions from mobile sources that reduce ammonium nitrate. Between 2013 and 2020—the attainment year for the 1997 annual and 24-hour PM2.5 standards—total NOx emissions are expected to decline 36 percent, and between 2013 and 2024—the attainment year for the 2006 24-hour PM2.5 standard—total NOx emissions are projected to decline 53 percent. Meanwhile, total ammonia emissions are expected to remain flat, as shown in Figure C4-1. The San Joaquin Valley Air Pollution Control District (District) adopted four rules47 between 2004 and 2011 with measures that provided ammonia emissions reductions in the Valley of approximately 50 tons per day (tpd); however, reductions from these existing control measures are already accounted for in the inventory, prior to the base year of 2013. In the future, emissions from the main sources of ammonia—dairies, fertilizer, and non-dairy livestock operations—are not anticipated to either increase or decrease substantially.

46 Ibid. 17 47 District Rule 4550: Conservation Management Practices (adopted 2004); Rule 4565: Biosolids, Animal Manure, and Poultry Litter Operations (adopted 2007); Rule 4566: Organic Material Composting Operations (adopted 2011); and Rule 4570: Confined Animal Facilities (adopted 2006, amended 2010)

111

Figure C4-1. NOx and ammonia emission trends in the San Joaquin Valley between 2013 and 2024.

NOx and Ammonia Emission Trends in the San Joaquin Valley

317.2

203.3

148.9

329.2 325.9 324.6

0

50

100

150

200

250

300

350

2019 2020 2021 2022 2023 2024

Emis

sion

s (T

ons

Per D

ay)

2013 2014 2015 2016 2017 2018

NOx Ammonia

Source: CEPAM Inventory version 1.05

The steep downward trend of NOx emissions and the stability of ammonia emissions between 2013 and 2024 lead CARB staff to conclude that modeling the impact of ammonia emissions reductions in the future, rather than the base year, is appropriate and more representative of the Valley’s emissions conditions. U.S. EPA guidance states that, in some situations, it may be “more appropriate to model future conditions that provide a more representative sensitivity analysis.”48 This approach is applicable in the Valley. Although emissions of NOx and ammonia are of roughly similar magnitude in the base year, thereby leading to some modeled sensitivity of PM2.5 levels to a 30 percent reduction in ammonia emissions, these conditions do not persist and are not representative in the future. Recent research further supports the fact that ammonia emissions are already in excess in the Valley. Field study measurements conducted during the 2013 DISCOVER-AQ study indicate that ammonia is in excess of NOx on peak PM2.5 days in the Valley, as illustrated in Figure C4-2. These data imply that ammonium nitrate formation in the Valley is limited by the amount of NOx present in the air.

112

48 U.S. EPA. PM2.5 Precursor Demonstration Guidance: Draft for Public Review and Comment. Page 33

January 18, 2013 January 20, 2013

Figure C4-2. Excess ammonia (NH3) in the San Joaquin Valley on Jan 18 (Left) and Jan 20 (Right) based on NASA aircraft measurements in 2013.

This finding that nitrate formation in the Valley is in a NOx-limited regime is consistent with previous research. For instance, Lurmann et al. (2006) note that “[t]he consistent excess of NH3 over nitric acid levels indisputably shows that secondary ammonium nitrate formation is more limited by nitric acid availability than NH3 within the SJV and in the foothills.”49 Since ammonium nitrate formation is limited by NOx, reducing NOx emissions is the more effective strategy for reducing ammonium nitrate and PM2.5. Other research has found that ammonia concentrations in the San Joaquin Valley have increased, further confirming that NOx reductions are the most effective path to reducing PM2.5.

Future Year Modeling CARB staff therefore repeated the sensitivity-based analysis of ammonia for the future attainment years of 2020 and 2024.50 Staff used an air quality model to estimate the PM2.5 design value for the annual and 24-hour standards in 2020 and 2024 at each Valley monitor. Then, CARB staff applied a 30 percent reduction to ammonia emissions and used the air quality model to estimate the PM2.5 design values in 2020 and 2024, shown in Tables C4-4 and C4-5 respectively. The difference between the two design values represents the modeled impact on PM2.5 levels of a 30 percent reduction in ammonia emissions in each attainment year.

49 Lurmann et al. “Processes influencing secondary aerosol formation in the San Joaquin Valley during winter.” Journal of the Air & Waste Management Association. 2006. Web. 3 Oct. 2017. <http://www.tandfonline.com/doi/pdf/ 10.1080/10473289.2006.10464573>. Page 1688 50 CARB did not conduct sensitivity analysis for the 2025 attainment year for the 2012 annual PM2.5 standard due to the close proximity of the attainment years for the 2012 and 2006 standards. Precursor sensitivities in 2025 are assumed to be very similar to those modeled in 2024.

113

http://www.tandfonline.com/doi/pdf

Table C4-4. Future Year 2020 PM2.5 – 30 Percent Ammonia Reduction. Annual 24-Hour

Site 2020

Baseline DV



2020 Baseline

DV




In 2020, the modeled air quality impact of reducing ammonia emissions by 30 percent falls under U.S. EPA’s recommended threshold at all but four Valley monitors for the 24-hour standard. The air quality impact remains above U.S. EPA’s recommended annual threshold at most sites.


Site 2024

Baseline DV



2024 Baseline

DV




114

In 2024, the modeled air quality impact of reducing ammonia emissions by 30 percent falls under U.S. EPA’s recommended annual threshold at all but two Valley monitors, and falls under the 24-hour threshold at all sites.

For completeness, CARB staff repeated this analysis, applying instead the U.S. EPA-recommended upper bound of a 70 percent reduction to ammonia emissions in 2020 and 2024, as shown in Tables C4-6 and C4-7.


Site 2020

Baseline DV



2020 Baseline

DV





Site 2024

Baseline DV



2024 Baseline

DV




115

From this analysis, the estimated air quality impact of reducing ammonia emissions by the upper bound of 70 percent in 2020 and 2024 exceeds U.S. EPA’s recommended thresholds for both the annual and 24-hour standards at all sites except one.

Available Emissions Controls Available emissions controls on ammonia are also relevant to the decision-making process, influencing the extent of reasonable modeled reductions. While U.S. EPA recommends modeling emissions reductions of between 30 and 70 percent to estimate PM2.5 impacts,51

CARB staff, District staff, and the public process have not identified specific controls that are technologically and economically feasible to achieve reductions at the low end of the recommended sensitivity range (i.e. 30 percent), much less at the upper end of the range. Emissions of ammonia in the Valley are approximately 329 tpd, as shown in Figure C4-3, meaning reductions would need to be in the range of approximately 99 to 230 tpd (30 to 70 percent).

The District’s existing rules that provide ammonia emissions reductions reflect the best available control measures for ammonia sources in the Valley, and implementation of these measures cannot feasibly reduce emissions by 30 percent. Therefore, CARB staff determined that modeled emissions reductions of 30 percent were an upper bound for potential ammonia reductions. CARB nevertheless modeled 70 percent reductions (see Tables C4-6 and C4-7) for completeness. In addition, CARB continues to pursue research on the feasibility and effectiveness of further ammonia controls on Valley sources.

Figure C4-3. Sources of ammonia in the Valley, 2013.


51 U.S. EPA. PM2.5 Precursor Demonstration Guidance: Draft for Public Review and Comment. Page 29

116

Relevant Monitors

The impact of ammonia on PM2.5 at monitors that form the basis of the attainment finding for the Valley is the focus of this analysis. For purposes of demonstrating attainment of the PM2.5 standards, the design sites are Bakersfield and Fresno. U.S. EPA guidance permits consideration of “the severity of nonattainment at relevant monitors,”52 and in 2024, PM2.5 levels are not sensitive to ammonia reductions at these design sites.

The sites at Madera and Hanford show an impact over the recommended threshold for the annual standard. Based on CARB staff analysis, however, the Madera design value is biased high: measured PM2.5 values from the Madera site were substantially higher than historical trends would suggest for the area. In addition, the Madera monitor is already nearing the 12 µg/m3 PM2.5 standard. For Hanford, while the impact is over U.S. EPA’s recommended significance level, achieving the level of controls needed for a 30 percent reduction of ammonia is not feasible, as discussed above.

Conclusion CARB has followed U.S. EPA guidance to evaluate whether ammonia contributes significantly to PM2.5 levels that exceed the NAAQS. Considering relevant contextualizing information such as emissions, research, and available controls, along with performing sensitivity-based analysis in future years, CARB determined that emissions of ammonia do not contribute significantly to PM2.5 levels that exceed the 1997, 2006, or 2012 NAAQS in the area. Therefore, CARB has excluded ammonia from control requirements in the SIP.

52 Ibid. 17

117

SULFUR DIOXIDE ANALYSIS Ammonium sulfate ([NH4]2SO4) is a constituent of PM2.5, making up about 10 percent of fine particulate matter mass in the Valley. Sulfur oxides (SOx) emitted from stationary and mobile combustion sources, mostly as sulfur dioxide (SO2), are oxidized in the atmosphere to ultimately form sulfuric acid (H2SO4). Sulfuric acid then combines with ammonia to form ammonium sulfate. Since SOx reacts chemically in this way to form a particle, SOx is a precursor to PM2.5.

Following the analytical process outlined in the U.S. EPA precursor demonstration guidance and summarized above, CARB has evaluated SOx in the Valley. The results of the sensitivity-based analysis and consideration of additional information are presented below.

Sensitivity-Based Analysis CARB staff used an air quality model to estimate the PM2.5 design value for the annual and 24-hour standards in the base year of 2013 at each Valley monitor. Then, CARB staff applied the recommended lower bound of a 30 percent reduction to SOx emissions and used the air quality model to estimate the PM2.5 design values, as shown in Table C4-8. The difference between the two design values represents the modeled impact on PM2.5 levels of a 30 percent reduction in SOx emissions in 2013. This is the value that is compared to U.S. EPA’s recommended contribution thresholds of 0.2 µg/m3 for the annual standard and 1.3 µg/m3 for the 24-hour standard to establish if PM2.5 levels are sensitive to this level of SOx reduction.

Table C4-8. Base Year 2013 PM2.5 – 30 Percent SOx Reduction. Annual 24-Hour

Site 2013

Baseline DV

2013 DV with 30% SOx


2013 Baseline

DV



Bakersfield-Planz 17.19 17.15 0.04 55.5 55.9 -0.4 Madera 16.93 16.92 0.01 51.0 51.3 -0.3 Hanford 16.54 16.53 0.01 60.0 60.4 -0.4 Visalia 16.20 16.15 0.05 55.5 55.8 -0.3 Clovis 16.12 16.11 0.01 55.8 56.0 -0.2 Bakersfield-California 16.02 15.98 0.04 64.1 64.5 -0.4 Fresno-Garland 14.98 14.95 0.03 60.0 60.1 -0.1 Turlock 14.88 14.83 0.05 50.7 50.8 -0.1 Fresno-HW 14.22 14.18 0.04 59.3 59.4 -0.1 Stockton 13.14 13.07 0.07 42.0 41.8 0.2 Merced-S Coffee 13.10 13.08 0.02 41.1 41.2 -0.1 Modesto 13.03 12.97 0.06 47.9 47.9 0.1 Merced-M 10.97 10.95 0.02 46.9 47.0 -0.1 Manteca 10.09 10.02 0.07 36.9 36.6 0.2 Tranquility 7.72 7.73 -0.01 29.5 29.5 0.0

118

For completeness, CARB staff repeated this analysis, applying instead the recommended upper bound of a 70 percent reduction to the SOx emissions in the base year, as shown in Table C4-9.

Table C4-9. Base Year 2013 PM2.5 – 70 Percent SOx Reduction. Annual 24-Hour

Site 2013

Baseline DV



2013 Baseline

DV



Bakersfield-Planz 17.19 17.11 0.08 55.5 56.5 -1.0 Madera 16.93 16.95 -0.02 51.0 52.2 -1.2 Hanford 16.54 16.54 0.00 60.0 61.4 -1.4 Visalia 16.20 16.10 0.10 55.5 56.3 -0.8 Clovis 16.12 16.10 0.02 55.8 56.4 -0.6 Bakersfield-California 16.02 15.95 0.07 64.1 65.2 -1.1 Fresno-Garland 14.98 14.93 0.05 60.0 60.6 -0.6 Turlock 14.88 14.77 0.11 50.7 51.1 -0.4 Fresno-HW 14.22 14.15 0.07 59.3 59.8 -0.5 Stockton 13.14 12.99 0.15 42.0 41.9 0.2 Merced-S Coffee 13.10 13.08 0.02 41.1 41.4 -0.3 Modesto 13.03 12.90 0.13 47.9 48.0 -0.1 Merced-M 10.97 10.93 0.04 46.9 47.2 -0.3 Manteca 10.09 9.95 0.14 36.9 36.4 0.5 Tranquility 7.72 7.77 -0.05 29.5 29.7 -0.2

From this analysis, the estimated air quality impact of reducing SOx emissions in the base year by the lower bound of 30 percent is well under U.S. EPA’s recommended thresholds at all Valley monitors for both the annual and 24-hour standards. In fact, in some cases, the estimated air quality impact is negative, implying that a reduction in SOx emissions would in fact increase the modeled design value at certain sites. Reducing emissions by the upper bound of 70 percent also shows impacts below the recommended thresholds.

Consideration of Additional Information To supplement modeling analysis, U.S. EPA guidance also allows an air agency to consider additional information. Accordingly, CARB evaluated the trend of SOx emissions in the Valley to support the sensitivity-based analysis.

119

Emissions Trend

CARB’s SOx inventory indicates that emissions remain roughly constant between 2013 and 2024, as shown in Figure C4-4. Ammonia emissions also remain flat over the same time frame, as shown above in Figure C4-1. Thus, conditions for ammonium sulfate formation are similar in the base and future years, with relative levels of ammonia and SOx remaining the same. The sensitivity-based analysis performed for 2013 and r eflected in Tables 8 and 9 above is therefore representative into the future, and it is redundant to additionally model the sensitivity of PM2.5 formation to SOx emissions reductions in 2020 or 2024. Precursor sensitivities in the future years are assumed to be very close to those modeled in 2013 due to the similarity of emissions conditions over time, so 2020 and 2024 analyses are not included here.

Figure C4-4. SOx emission trend in the San Joaquin Valley between 2013 and 2024.

SOx Emission Trend in the San Joaquin Valley

Emis

sion

s (T

ons

Per D

ay)

6

5

4

3

2

1

0 2013 2014 2015 2016 2017 2024

8.5 7.8 8.0

7

8

9

2018 2019 2020 2021 2022 2023

SOx


Conclusion CARB has followed U.S. EPA guidance to evaluate whether SOx contributes significantly to PM2.5 levels that exceed the NAAQS. Using sensitivity-based analysis in the base year and considering that base year conditions are representative into the future, CARB determined that emissions of SOx do not contribute significantly to PM2.5 levels that exceed the 1997, 2006, or 2012 NAAQS in the ar ea. Therefore, CARB has excluded SOx from control requirements in the SIP.

120

ROG ANALYSIS Following the analytical process outlined in the U.S. EPA precursor demonstration guidance and summarized above, CARB has evaluated ROG in the San Joaquin Valley. The results of the sensitivity-based analysis and consideration of additional information are presented below.

Sensitivity-Based Analysis CARB staff used an air quality model to estimate the PM2.5 design value for the annual and 24-hour standards in the base year of 2013 at each Valley monitor. Then, CARB staff applied the recommended lower bound of a 30 percent reduction to ROG emissions and used the air quality model to estimate the PM2.5 design values, as shown in Table C4-10. The difference between the two design values represents the modeled impact on PM2.5 levels of a 30 percent reduction in ROG emissions in 2013. This is the value that is compared to U.S. EPA’s recommended contribution thresholds of 0.2 µg/m3 for the annual standard and 1.3 µg/m3 for the 24-hour standard to establish if PM2.5 levels are sensitive to this level of ROG reduction.

Table C4-10. Base Year 2013 PM2.5 – 30 Percent ROG Reduction. Annual 24-Hour

Site 2013

Baseline DV

2013 DV with 30% ROG Reduction

Difference 2013

Baseline DV


Difference


For completeness, CARB staff repeated this analysis, applying instead the U.S. EPA-recommended upper bound of a 70 percent reduction to ROG emissions in the base year, as shown in Table C4-11.

121

Table C4-11. Base Year 2013 PM2.5 – 70 Percent ROG Reduction. Annual 24-Hour

Site 2013

Baseline DV


Difference 2013

Baseline DV


Difference


From this analysis, the estimated air quality impact of reducing ROG emissions in the base year by the lower bound of 30 percent is under U.S. EPA’s recommended thresholds at all but two Valley monitors for the 24-hour standard, and falls below the recommended annual threshold at all sites. Reducing emissions by the upper bound of 70 percent shows impacts above the thresholds at about half the sites.

Consideration of Additional Information To supplement modeling analysis, U.S. EPA guidance also allows an air agency to consider additional information. Accordingly, CARB evaluated the trend of ROG emissions in the Valley to support the sensitivity-based analysis and conducted future year sensitivity modeling.

Emissions Trend CARB has an extensive suite of measures in place to reduce ROG emissions, particularly in the area of regulating consumer products. In addition, the District has numerous rules that provide ROG emissions reductions in the Valley. CARB’s ROG inventory indicates that these existing controls reduce emissions by approximately 30 tons, or nine percent, between 2013 and 2024, as shown in Figure C4-5. Thus, the role ROG plays in PM2.5 formation may differ in the base and future years, and the sensitivity-based analysis performed for 2013 is not representative into the future.

122

Figure C4-5. ROG emission trend in the San Joaquin Valley between 2013 and 2024.

ROG Emission Trend in the San Joaquin Valley

Emis

sion

s (T

ons

Per D

ay)

200

150

100

50

0 2019 2020 2021 2022 2023 2024

324.1 296.2 293.8

250

300

350

2013 2014 2015 2016 2017 2018

ROG


Future Year Modeling Even though the estimated air quality impact of reducing ROG emissions in the base year by 30 percent is under U.S. EPA’s recommended thresholds at all but two Valley monitors for the 24-hour standard, and falls below the recommended annual threshold at all sites, CARB staff repeated the sensitivity-based analysis of ROG for the future attainment years of 2020 and 2024 for completeness.53 Staff used an air quality model to estimate the PM2.5 design value for the annual and 24-hour standards in 2020 and 2024 at each Valley monitor. Then, CARB staff applied a 30 per cent reduction to ROG emissions and used the air quality model to estimate the PM2.5 design values in 2020 and 2024, shown in Tables C4-12 and C4-13 respectively. The difference between the two design values represents the modeled impact on PM2.5 levels of a 30 percent reduction in ROG emissions in each attainment year.

53 CARB did not conduct sensitivity analysis for the 2025 attainment year for the 2012 annual PM2.5 standard due to the close proximity of the attainment years for the 2012 and 2006 standards. Precursor sensitivities in 2025 are assumed to be very similar to those modeled in 2024.

123

Table C4-12. Future Year 2020 PM2.5 – 30 Percent ROG Reduction. Annual 24-Hour

Site 2020

Baseline DV


Difference 2020

Baseline DV


Difference

Bakersfield-Planz 14.58 14.55 0.03 41.2 40.9 0.3 Madera 14.15 14.12 0.03 38.9 38.6 0.2 Hanford 13.30 13.35 -0.50 43.7 43.7 0.0 Visalia 13.51 13.47 0.04 42.8 42.2 0.6 Clovis 13.43 13.37 0.06 41.1 40.9 0.3 Bakersfield-California 13.48 13.47 0.01 47.6 47.5 0.1 Fresno-Garland 12.42 12.37 0.05 44.3 44.0 0.3 Turlock 12.47 12.46 0.01 37.8 37.7 0.1 Fresno-HW 11.86 11.80 0.06 45.6 45.2 0.4 Stockton 11.43 11.42 0.01 33.5 33.4 0.1 Merced-S Coffee 10.86 10.86 0.00 30.0 29.9 0.0 Modesto 10.97 10.96 0.01 35.8 35.7 0.1 Merced-M 9.34 9.33 0.01 32.9 32.9 0.0 Manteca 8.67 8.66 0.01 30.1 30.0 0.1 Tranquility 6.40 6.41 -0.01 21.5 21.6 -0.1


Site 2024

Baseline DV


Difference 2024

Baseline DV


Difference

Bakersfield-Planz 12.03 11.92 -0.01 30.0 30.0 -0.2 Madera 11.98 11.99 -0.01 30.2 30.3 -0.1 Hanford 10.52 10.59 -0.07 30.1 30.5 -0.4 Visalia 11.09 11.1 -0.01 30.2 30.4 -0.3 Clovis 11.37 11.34 0.03 30.7 30.7 0.0 Bakersfield-California 11.01 10.91 -0.01 33.3 33.5 -0.4 Fresno-Garland 10.43 10.41 0.02 32.8 32.9 -0.1 Turlock 11.14 11.16 -0.02 30.2 30.3 -0.1 Fresno-HW 10.02 9.99 0.03 35.1 35.2 0.0 Stockton 10.66 10.67 -0.01 28.6 28.6 -0.1 Merced-S Coffee 9.65 9.67 -0.02 24.2 24.3 -0.1 Modesto 9.97 9.98 -0.01 29.1 29.2 -0.1 Merced-M 8.61 8.61 0.00 27.4 27.8 -0.1 Manteca 7.97 7.98 -0.01 25.8 25.8 0.0 Tranquility 5.54 5.55 -0.01 16.2 16.3 -0.1

In both 2020 and 2024, the modeled air quality impact of reducing ROG emissions by 30 percent falls under U.S. EPA’s recommended thresholds at all sites.

For completeness, CARB staff repeated this analysis, applying instead the recommended upper bound of a 70 percent reduction to ROG emissions in 2020 and 2024, as shown in Tables C4-14 and C4-15.

124


Site 2020

Baseline DV


Difference 2020

Baseline DV


Difference

Bakersfield-Planz 14.58 14.51 0.07 41.2 40.3 1.0 Madera 14.15 14.09 0.06 38.9 38.3 0.6 Hanford 13.30 13.40 -0.10 43.7 43.5 0.2 Visalia 13.51 13.40 0.11 42.8 41.3 1.5 Clovis 13.43 13.27 0.16 41.1 40.4 0.7 Bakersfield-California 13.48 13.44 0.04 47.6 47.2 0.5 Fresno-Garland 12.42 12.29 0.13 44.3 43.5 0.8 Turlock 12.47 12.43 0.04 37.8 37.5 0.2 Fresno-HW 11.86 11.71 0.15 45.6 44.6 1.0 Stockton 11.43 11.41 0.02 33.5 33.2 0.3 Merced-S Coffee 10.86 10.85 0.01 30.0 29.8 0.1 Modesto 10.97 10.95 0.02 35.8 35.6 0.2 Merced-M 9.34 9.30 0.04 32.9 32.9 0.1 Manteca 8.67 8.64 0.03 30.1 29.8 0.3 Tranquility 6.40 6.41 -0.01 21.5 21.7 -0.2

In 2020, the modeled air quality impact of reducing ROG emissions by 70 percent falls under U.S. EPA’s recommended annual threshold at all sites, and under the recommended 24-hour threshold at all sites but one.


Site 2024

Baseline DV


Difference 2024

Baseline DV


Difference

Bakersfield-Planz 12.03 11.94 -0.03 30.0 30.3 -0.5 Madera 11.98 12.01 -0.03 30.2 30.4 -0.3 Hanford 10.52 10.70 -0.18 30.1 31.1 -1.0 Visalia 11.09 11.11 -0.02 30.2 30.7 -0.5 Clovis 11.37 11.29 0.08 30.7 30.7 0.0 Bakersfield-California 11.01 10.94 -0.04 33.3 34.0 -0.9 Fresno-Garland 10.43 10.37 0.06 32.8 33.0 -0.2 Turlock 11.14 11.19 -0.05 30.2 30.5 -0.3 Fresno-HW 10.02 9.95 0.07 35.1 35.2 -0.1 Stockton 10.66 10.67 -0.01 28.6 28.7 -0.1 Merced-S Coffee 9.65 9.69 -0.04 24.2 24.5 -0.3 Modesto 9.97 9.99 -0.02 29.1 29.3 -0.2 Merced-M 8.61 8.60 0.01 27.4 27.7 -0.3 Manteca 7.97 7.98 -0.01 25.8 25.9 -0.1 Tranquility 5.54 5.57 -0.03 16.2 16.6 -0.4

In 2024, the modeled air quality impact of reducing ROG emissions by 70 percent falls under U.S. EPA’s recommended thresholds at all sites.

125

Conclusion CARB has followed U.S. EPA guidance to evaluate whether ROG contributes significantly to PM2.5 levels that exceed the NAAQS. Using sensitivity-based analysis in the base and future years, CARB determined that emissions of ROG do not contribute significantly to PM2.5 levels that exceed the 1997, 2006, or 2012 NAAQS in the area. Therefore, CARB has excluded ROG from control requirements in the SIP.

126

Appendix C: Weight of Evidence AnalysisIn 1997, U.S. EPA adopted the first set of PM2.5 air quality...

Documents

Transcript of Appendix C: Weight of Evidence AnalysisIn 1997, U.S. EPA adopted the first set of PM2.5 air quality...