Atmospheric Environment 92 (2014) 182e190

Contents lists available at ScienceDirect

Atmospheric Environment
journal homepage: www.elsevier.com/locate/atmosenv

Seasonal modeling of PM2.5 in California’s San Joaquin Valley

Jianjun Chen a, Jin Lu a, Jeremy C. Avise a, b, John A. DaMassa a, Michael J. Kleeman c,
Ajith P. Kaduwela a, d, *
a
Air Quality Planning and Science Division, California Air Resources Board, 1001 I Street, Sacramento, CA 95814, USA
b
Department of Civil and Environmental Engineering, Washington State University, 405 Spokane Street, Pullman, WA 99164, USA
c
Department of Civil and Environmental Engineering, University of California, One Shields Avenue, Davis, CA 95616, USA
d
Air Quality Research Center and the Department of Land, Air and Water Resources, University of California, One Shields Avenue, Davis, CA 95616, USA

h i g h l i g h t s

CMAQ was used to simulate winter PM2.5 formation in the San Joaquin Valley (SJV).

Primary PM2.5 and NOx controls are most effective in reducing winter PM2.5 in the SJV.

SOx control is as effective as NOx control, but SOx emissions are minimal in the SJV.

NH3 controls provide only a small benefit.

VOCs controls provide no benefit.

a r t i c l e i n f o a b s t r a c t

Article history: California’s San Joaquin Valley (SJV) is in non-attainment for the 2006 revised 24-h PM2.5 National
Received 18 September 2013 Ambient Air Quality Standard (NAAQS or standard) established by the United States Environmental
Received in revised form Protection Agency (U.S. EPA). As a part of the emissions control strategy development to bring the SJV
14 March 2014
into attainment of the standard, the Community Multi-scale Air Quality (CMAQ) model was used to
Accepted 16 April 2014
Available online 18 April 2014
simulate PM2.5 formation and its response to precursor emission reductions. Simulations were conducted
for the first (JanuaryeMarch) and fourth (OctobereDecember) quarters of 2007 since the 24-h PM2.5
violations typically occur during the winter months. The CMAQ model generally met the PM2.5 model
Keywords:
Community Multiscale Air Quality (CMAQ)
performance criteria and was suitable for the State Implementation Plan (SIP) applications. Sensitivities
model of the 24-h average PM2.5 concentrations to precursor emissions were investigated based on the baseline
Precursor equivalence ratio 2019 emissions forecast. Reductions in anthropogenic primary PM2.5 and oxides of nitrogen (NOx)
Carrying capacity diagrams showed the greatest impact on PM2.5 concentrations. Reducing anthropogenic ammonia and sulfur ox-
Particulate matter ides emissions only slightly lowered PM2.5 concentrations, and reducing volatile organic compounds
San Joaquin Valley (VOC) emissions provided no benefit in reducing PM2.5 concentrations. Particularly, PM2.5 nitrate, the
Seasonal modeling major PM2.5 component in the SJV showed no response to VOC reductions, which is also supported by
additional modeling results from the UCD/CIT airshed model. This study demonstrated that the current
strategy of controlling primary PM2.5 and NOx emissions in the SJV will continue to be effective for
further reducing PM2.5 in the SJV beyond 2019. It also demonstrated how the CMAQ model can be used to
design effective emissions control strategies for regulatory applications.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Epidemiological studies have shown that elevated ambient PM con-

Atmospheric particulate matter (PM) is associated with adverse
health effects, impaired visibility, and climate change.
* Corresponding author. Air Quality Planning and Science Division, California Air
Resources Board, 1001 I Street, Sacramento, CA 95814, USA.
E-mail address: akaduwel@arb.ca.gov (A.P. Kaduwela).
centrations are linked to increased morbidity and mortality (Pope III
and Dockery, 2006). The United States Environmental Protection
Agency (U.S. EPA) has established health-based National Ambient Air
Quality Standards (NAAQS or standards) for 24-h and annual average
PM2.5 (PM with aerodynamic diameter less than 2.5 mm)
concentrations.
The California San Joaquin Valley (SJV) is surrounded by
mountain ranges to the west, east, and south, which prevent the

http://dx.doi.org/10.1016/j.atmosenv.2014.04.030
1352-2310/Ó 2014 Elsevier Ltd. All rights reserved.
 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190 183

dispersion of pollutants. During the winter months, high pressure Community Multiscale Air Quality (CMAQ) model was used to
systems can cause the atmosphere to be stagnant for prolonged simulate PM2.5 formation in the SJV for the first and fourth calendar
periods of time leading to nighttime temperature inversions that quarters of 2007. Model performance was evaluated and validated.
exacerbate the buildup of PM2.5 concentrations in the Valley (Chow Sensitivities of PM2.5 formation to precursor emission reductions
et al., 2006, Turkiewicz et al., 2006). The SJV is home to almost four were assessed using the projected 2019 emissions as the base
million people and one of the most productive agricultural regions scenario. Compared to previous PM2.5 modeling studies, a much
in the world. Two of California’s major north-south highways pass longer period (i.e., 6 months) was simulated. PM2.5 carrying ca-
through the SJV. Consequently, a variety of emission sources pacity diagrams (also known as control isopleths) are presented
contribute to PM2.5 pollution in the SJV such as primary PM2.5 here for the first time for all precursors. To confirm CMAQ modeling
emissions from residential wood combustion, NOx emissions from results, an independent model (i.e., the UCD/CIT airshed model)
on- and off-road mobile sources, and ammonia emissions from was also used to study the sensitivity of PM2.5 nitrate to NOx and
agricultural sources. The SJV is in non-attainment for the 2006 VOCs emission reductions, which is of great interest to the stake-
revised 24-h PM2.5 standard (35 mg/m3), primarily due to high holders in the SJV.
concentrations occurring during the winter months (November
through February).
Ambient PM2.5 speciation measurements in the SJV show that 2. Methods
the wintertime PM2.5 composition averaged over high PM2.5 days in
2008e2010 period is dominated by ammonium nitrate (NH4NO3) 2.1. CMAQ model configurations and input preparations
and carbonaceous compounds (CARB, 2012). For example, in
Fresno, NH4NO3 and carbonaceous aerosol (including both Organic The CMAQ model version 4.7.1 (Foley et al., 2010) was used to
Carbon (OC) and Elemental Carbon (EC) which we will call Total simulate PM2.5 formation in the SJV during the years 2007 and
Carbon (TC)) constituted approximately 50% and 40% of PM2.5, 2019. The following options were used: the SAPRC99 gas phase
respectively (CARB, 2012). For the same period in Bakersfield, chemistry mechanism (Carter, 2000), the aero5 aerosol module, the
NH4NO3 and TC accounted for 67% and 21% of PM2.5 mass, unmodified piecewise parabolic method for horizontal and vertical
respectively. Together, NH4NO3 and TC account for w90% of the advection, and the eddy diffusivity algorithm for vertical diffusion.
total mass at both sites. Previous studies found that NH4NO3 in the Fig. 1 shows the CMAQ modeling domains that were used. The
SJV in winter is dominated by secondary formation from gas-phase larger domain covering all of California has a horizontal grid size of
ammonia (NH3) and nitric acid (HNO3), while TC in urban areas is 12 km and was used to provide boundary conditions for the smaller
mainly due to primary PM emissions (Chen et al., 2010; Chow et al., domain, which covers the SJV with a finer grid size of 4 km. Fifteen
2006; Watson and Chow, 2002; Ying et al., 2008a,b). vertical layers were employed extending telescopically from the
Prior control strategies for abating wintertime PM2.5 pollutions surface to 100 mb with the majority of layers in the planetary
in the SJV have focused heavily on nitrogen oxides (NOx) and pri- boundary layer. The boundary conditions for the larger 12 km
mary PM2.5 emissions in the SJV, particularly from mobile source domain were extracted from the global atmospheric chemical
and residential wood burning emissions (SJVAPCD, 2012). Ambient transport Model for Ozone and Related chemical Tracers (MOZART,
air quality measurements indicate that maximum 24-h PM2.5 Emmons et al., 2010; CARB, 2012; MOZART-4/GEOS-5 output
Design Values (DV, defined here as the annual 98th percentile of available at http://www.acd.ucar.edu/wrf-chem/mozart.shtml).
the observed 24-h PM2.5 concentrations averaged over three Meso-scale Model Version 5 (MM5, Grell et al., 1994) was used
consecutive years) in the SJV have dropped by 40% over the 1999e to generate the meteorological fields for the simulated period
2011 time periods (SJVAPCD, 2012). In addition, the number of days
exceeding the standard per year has also decreased by 46% from
1999 to 2011, which demonstrates the overall effectiveness of
control strategies implemented during the past decade (SJVAPCD,
2012).
A number of three-dimensional grid-based air quality model
simulations were previously conducted to study the formation of
PM2.5 and its precursor limitations in the SJV (Chen et al., 2009,
2010; Held et al., 2004; Kleeman et al., 2005; Livingstone et al.,
2009; Pun et al., 2009; Ying and Kleeman, 2006; Ying et al.,
2008a,b; Ying, 2011; Zhang et al., 2010). Generally, these
modeling studies have shown that reductions in NOx emissions
were most effective in reducing NH4NO3, the most prevalent
component of PM2.5 in the SJV in winter, while reductions in NH3
and volatile organic compounds (VOCs) emissions were less
effective (Kleeman et al., 2005; Pun et al., 2009). All these studies
were for 2000 or pre-2000 episodes. Emissions in the SJV have
decreased since then despite population growth, and the severity of
PM2.5 episodes have diminished accordingly. Thus, there is a need
for studies based on more recent or near-future emission scenarios
and a comparison with historical studies to determine if the
chemical regime is changing over time.
The purpose of this paper is to present key PM2.5 modeling re-
sults that were used to support the development of the 2012 SJV
Fig. 1. CMAQ modeling domains (12 km grid size for domain 1 and 4 km grid size for
PM2.5 State Implementation Plan (SIP or plan) (SJVAPCD, 2012). The domain 2) and PM2.5 monitoring sites (1 e Stockton; 2 e Modesto; 3 e Merced; 4 e
2012 SJV PM2.5 plan demonstrated SJV’s attainment of the 2006 Clovis; 5 e Fresno, 1st Street; 6 e Fresno, Hamilton & Winery; 7 e Visalia; 8 e Cor-
revised 24-h PM2.5 standard by the federal deadline of 2019. The coran; 9 e Bakersfield, California Avenue; 10 e Bakersfield, E. Planz Road).
184 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190

(CARB, 2012). The initial and boundary conditions for MM5 were emission inventory and 2007 meteorological fields. CMAQ pre-
based on the 6-h interval data from the NCEP/NCAR Reanalysis dicted 24-h average concentrations for PM2.5 mass as well as major
Project (NNRP). In addition, surface wind measurement data were species (i.e., nitrate, ammonium, sulfate, OC, and EC) were
used for observational nudging. The physics options used for MM5 compared to observations at the four PM2.5 sites with speciation
include the Grell cumulus parameterization, the GaynoeSeaman data (Bakersfield, Fresno, Modesto, and Visalia) located within the
planetary boundary layer scheme, the Dudhia simple ice scheme, SJV. Model performance metrics such as mean fractional bias (MFB)
the rapid radiation transfer model, and the 5-layer soil slab model. and mean fractional error (MFE) were computed (shown in Table 1)
A similar Weather Research and Forecasting (WRF, Skamarock et al., and were compared to criteria established in Boylan and Russell
2005) simulation was also performed and the CMAQ model per- (2006) and U.S. EPA (2007). There metrics are shown in Section
formance based on the WRF generated meteorology was found to 3.1. While only PM2.5 related quantitative model performance was
be comparable to that of MM5. Therefore, only MM5-based CMAQ presented in the paper, an extensive evaluation of model perfor-
results are presented here. Meteorology model performance was mance was conducted. For example, the so-called “soccer plot”,
evaluated by comparing simulated fields with surface observations “bugle plot”, and “Taylor plot” (U.S. EPA, 2007) were generated for
of wind speed and direction, temperature, and relative humidity examining model performance. Time series and scatter plots of
and performance was found to be satisfactory. Temporal and sta- modeled and measured gases (e.g., ozone, nitrogen oxide, carbon
tistical analysis of the MM5 model performance can be found at monoxide, and sulfur dioxide) were also generated and gas phase
http://arb.ca.gov/eos/SIP_Modeling_PM25/24hr_PM25_ model performance was validated. These plots can be found at
ModelingPage.htm. http://arb.ca.gov/eos/SIP_Modeling_PM25/24hr_PM25_
Three sets of emission inventories were prepared by the Cali- ModelingPage.htm.
fornia Air Resources Board (CARB), which are a day-specific 2007
emission inventory, and two planning inventories for 2007 and
2.3. Calculation of the future-year PM2.5 design values
2019 featuring typical average day emission patterns (CARB, 2012).
The 2007 day-specific emission inventory was prepared for the
The future-year 2019 PM2.5 design values were calculated ac-
CMAQ model performance evaluation. The 2007 and 2019 planning
cording to the speciated modeled attainment test (SMAT) proce-
inventories were used for the purpose of calculating PM2.5 relative
dure described in the U.S. EPA air quality modeling guidance (CARB,
response factors (which is the ratio of averaged 2019 PM2.5 con-
2012; U.S. EPA, 2007, 2011). The design value refers here to the
centration to that in 2007).
annual 98th percentile of the 24-h PM2.5 concentrations averaged
The 2007 day-specific emission inventory was developed based
over three consecutive years. To calculate a future-year design
on day-specific information available for specific emission sources.
value, CMAQ was used to estimate the simulated response of PM2.5
For example, the day-specific emissions from prescribed and agri-
components to emissions changes from the base year (2007) to the
cultural burns were calculated based on the actual burning events.
future year (2019). This response (Relative Response Factor or RRF)
Emissions from mobile sources were also adjusted by using hour-
is then used to adjust the observed base-year design value to
specific temperature and humidity, as well as monthly, daily, and
calculate the future-year design value. Since peak PM2.5 days in the
hourly activity data. Using day-specific emissions when possible is
an improvement over a more average inventory since day-specific
emissions can have significant impacts on air quality.
Table 1
Unlike the day-specific inventory, the 2007 and 2019 planning
Key model performance metrics for the 24-h average PM2.5 mass and components
inventories used typical average-day emission patterns but still during the first and fourth quarters of 2007.
included differences in average emissions during weekdays and
Species Num. Mean Mean Mean Mean Mean Mean
weekends. The 2019 anthropogenic emissions were projected ac- of obs. mod. bias error fractional fractional
cording to forecasted economic conditions, population growth, and obs. (mg/m3) (mg/m3) (mg/m3) (mg/m3) bias error
implemented and planned emission controls. Different emissions
Bakersfield
growth factors were calculated and applied for different source PM2.5 51 27.97 28.19 0.23 11.20 0.13 0.40
categories. For example, area sources emission growth depended Ammonium 51 4.17 2.97 1.20 1.86 0.15 0.47
on future population projection, while mobile source emissions Nitrate 51 11.75 8.83 2.93 5.00 0.18 0.46
took into account the future growth of vehicle miles traveled. Sulfate 51 1.38 1.05 0.34 0.58 0.17 0.43
EC 51 1.49 2.85 1.36 1.37 0.68 0.68
Control factors of emissions incorporated adopted regulations at OC 51 6.21 5.41 0.80 2.56 0.09 0.41
the federal, state, and local levels. Fresno
Biogenic emissions were generated using a modified version of PM2.5 56 29.35 27.84 1.51 10.34 0.09 0.39
the MEGAN model version 2.04 (Guenther et al., 2006), which Ammonium 57 3.73 3.00 0.72 1.53 0.02 0.48
Nitrate 57 10.19 9.12 1.08 3.76 0.04 0.43
incorporated California specific emission factor and plant func-
Sulfate 57 1.19 0.92 0.27 0.47 0.19 0.38
tional type data, as well as 8-day MODIS LAI data (CARB, 2012). EC 56 1.46 2.34 0.89 0.95 0.51 0.54
Biogenic emissions were kept constant between 2007 and 2019 OC 56 9.09 5.05 4.05 4.14 0.49 0.51
because future inputs such as land use/land cover are highly un- Modesto
certain. Detailed descriptions of the emission inventory prepara- PM2.5 30 21.57 22.10 0.53 8.88 0.15 0.42
Ammonium 30 2.61 2.19 0.42 1.31 0.02 0.60
tion including temporal and spatial distributions as well as Nitrate 30 7.15 5.75 1.40 3.55 0.14 0.61
chemical speciation can be found in CARB (2012). In parallel, an Sulfate 30 1.14 1.37 0.23 0.75 0.22 0.51
extensive quality assurance/analysis was also performed to ensure EC 18 1.24 2.27 1.03 1.07 0.60 0.63
that gridding, temporal profiles, and speciation profiles were OC 18 5.71 5.20 0.51 2.52 0.00 0.41
Visalia
applied correctly to the inventory (CARB, 2012).
PM2.5 30 30.50 21.58 8.92 10.65 0.24 0.35
Ammonium 30 3.91 2.94 0.98 1.40 0.13 0.37
2.2. CMAQ model performance evaluation Nitrate 30 12.18 9.15 3.03 4.73 0.10 0.39
Sulfate 30 1.27 0.73 0.54 0.59 0.45 0.51
Model performance evaluation involved CMAQ simulations for EC 18 1.25 1.25 0.00 0.49 0.02 0.41
OC 18 6.26 2.95 3.31 3.35 0.64 0.65
the first and fourth quarters of 2007 using the day-specific 2007
 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190
SJV always occur in the wintertime (excluding exceptional events
such as wildfires), simulation of only the first and fourth quarters
was sufficient for calculating the future-year design value. Here,
CMAQ model simulations were conducted for the first and fourth
calendar quarters of 2007 (i.e., using the 2007 meteorological
conditions) using the 2007 and 2019 planning inventories to yield
RRFs for individual PM2.5 species in the SJV. The observed base-year
design values were based on 24-h average PM2.5 concentrations
measured by the federal reference method (FRM) at ten sites in the
SJV during the first and fourth quarters of 2005e2007. The FRM
measurement reflects the mass of the bulk sample, yet the relative
response factor for different PM2.5 species can differ. Therefore, the
bulk mass from FRM measurements needs to be apportioned to
major PM2.5 species. The sulfate, adjusted nitrate, derived water,
inferred carbonaceous material balance (SANDWICH) approach
developed by Frank (2006) was used for apportioning FRM PM2.5
mass to concentrations of individual species retained by the FRM
filter. The 24-h PM2.5 design values for 2019 were calculated for the
ten sites located in the SJV as shown in Fig. 1.
2.4. Future-year PM2.5 precursor sensitivity analysis
A series of CMAQ simulations were performed to investigate the
response of PM2.5 formation in the SJV to reductions of anthropo-
genic emissions of PM precursors (i.e., NOx, primary PM2.5, SOx,
VOCs, and NH3) in the SJV. The purpose was to investigate the
relative effectiveness of different PM precursor emission controls in
reducing PM2.5 concentrations in the SJV. The base scenario is the
CMAQ simulation for the 2007 modeling period (i.e., using the 2007
meteorological fields), but with the 2019 planning inventory. A set
of 16 emission scenarios were simulated by combining the scaling
of anthropogenic NOx and primary PM2.5 emissions in the SJV at 1.0,
0.75, 0.5, and 0.25 of base emissions (i.e., 16 combinations involving
four levels of NOx emissions and four levels of primary PM2.5
emissions). Similarly, a set of 16 CMAQ simulations were conducted
involving the scaling of anthropogenic NOx and VOCs emissions, the
scaling of anthropogenic NOx and SOx emissions, and the scaling of
anthropogenic NOx and NH3 emissions in the SJV. For each emission
scenario, the PM2.5 design values at the ten design sites were
calculated according to the SMAT procedure described in the pre-
vious section.
3. Results
3.1. Base year 2007 CMAQ model performance
Table 1 shows the key model performance metrics for PM2.5
mass and major species at the four sites with speciation data (i.e.,
Bakersfield, Fresno, Modesto, and Visalia) within the SJV. Boylan
and Russell (2006) proposed that a model performance goal is
met when both the MFE and MFB are less than or equal to 50% and
30%, respectively and that a model performance criteria is met
when both the MFE and MFB are less than or equal to 75% and
50%, respectively, when the observed concentrations are greater
than 2.25 mg/m3. According to Boylan and Russell (2006), model
performance goals are defined as the level of accuracy that is
considered to be close to the best a model can achieve while model
performance criteria are defined as the level of accuracy that is
considered to be acceptable for modeling applications. Based on
these metrics, the CMAQ model performance generally met the
model performance goal at Bakersfield while meeting performance
criteria at Fresno, Modesto, and Visalia. Thus, the modeling system
is robust enough for SIP applications. Table 1 also demonstrates that
nitrate and OC were the two major PM2.5 species in the SJV. The
185
modeling performance based on WRF meteorological fields is
shown in Table S-1.
Generally, the CMAQ model over predicted EC concentrations
while under predicting OC concentrations at these sites. The CMAQ
model predicted that the majority of OC at these four sites was
primary in nature (e.g., residential wood burning, meat cooking,
diesel equipment, etc.). Secondary OC was predicted to contribute
less than 5% of the total OC at each of these four sites. Given the
uncertainties in the atmospheric processing of primary and sec-
ondary OC (Chen et al., 2010; Robinson et al., 2007), poor model
performance for OC is not unexpected. Another confounding factor
might be due to the OC measurement technique. For example, it
was found that the CMAQ model performance for OC is much better
during OctobereDecember 2007 than during JanuaryeMarch 2007
at Bakersfield, Modesto, and Visalia. The OC measurement at
Bakersfield, Modesto, and Visalia was based on the old Speciation
Trends Network (STN)/Chemical Speciation Network (CSN) sam-
pling and analysis method during JanuaryeMarch 2007. But, it was
switched to the Interagency Monitoring of Protected Visual Envi-
ronments (IMPROVE) network sampling and analysis method
during OctobereDecember 2007. The STN/CSN method typically
has higher sampling artifact than the IMPROVE method in urban
areas (Chow et al., 2010). Measured OC concentrations in Octobere
December 2007 were much lower than those measured in
JanuaryeMarch 2007 at Bakersfield, Modesto, and Visalia. The OC
measurement at Fresno was made by the STN/CSN sampling and
analysis method throughout the entire year of 2007. The difference
in model performance for OC between JanuaryeMarch and
OctobereDecember 2007 is not as drastic as that for Bakersfield,
Modesto, and Visalia.
3.2. 2019 PM2.5 precursor sensitivity results
Figs. 2e4 show the PM2.5 carrying capacity diagrams at three
representative sites in the southern, central and northern SJV:
BakersfieldeCalifornia Avenue, Fresno e 1st Street, and Stocktone
Hazelton Street. The 24-h PM2.5 design values calculated according
to the SMAT procedure are shown for each combination of scaling
of NOx vs. primary PM2.5, NH3, SOx, and VOCs emissions from
anthropogenic sources within the SJV. Emissions from biogenic
sources were not changed. Bakersfield and Fresno are the two
largest urban areas in the SJV. The PM2.5 monitoring site in
Bakersfield (California Avenue) has the highest observed (in 2007)
and predicted (in 2019) 24-h PM2.5 design value within the SJV.
Stockton is located in the northern portion of the SJV which does
not have as serious air quality problems as the central and southern
SJV. The PM2.5 carrying capacity diagrams for the other seven sites
are provided in Figs. S1eS7 of the supplemental material.
The 24-h average PM2.5 at Bakersfield showed strong sensitivity
to primary PM2.5 and NOx emissions. Reducing 50% primary PM2.5
emissions in the SJV decreased the Bakersfield design value from
35.7 mg/m3 to 26.2 mg/m3 (a 27% reduction), while reducing 50%
anthropogenic NOx emissions in the SJV decreased it to 27.1 mg/m3
(a 24% reduction). Simultaneous reductions of 50% anthropogenic
primary PM2.5 and NOx emissions decreased the design value to
19.5 mg/m3 (a 45% reduction). This reduction is slightly smaller than
what would be expected if the effects of reducing primary PM2.5
and NOx emissions were additive. This is because the effect of
reduced PM2.5 surface area on heterogeneous nitrate formation,
which is further described in a later paragraph, is smaller when NOx
emissions were also reduced. Reducing NH3 and SOx emissions has
a much smaller effect on PM2.5 concentrations. A 50% reduction of
NH3 emissions decreased the design value to 34.3 mg/m3 (a 4%
reduction) and a 50% reduction of SOx emissions decreased it to
35.4 mg/m3 (a 1% reduction). Reducing anthropogenic VOCs
186 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190
Fig. 2. 24-h average PM2.5 carrying capacity diagrams at BakersfieldeCalifornia Avenue in 2019.
emissions in the SJV has a slightly negative effect on PM2.5,
increasing PM2.5 design values from 35.7 to 35.9 mg/m3.
Given that the major PM2.5 component in the SJV in wintertime
is NH4NO3 and that it is generally limited by the availability of HNO3
(Kleeman et al., 2005; Lurmann et al., 2006; Pun et al., 2009), it is
not surprising that reducing NOx emissions has a significant effect
on PM2.5 formation. A 50% reduction of anthropogenic NOx emis-
sions in the SJV decreased nitrate concentration by approximately
42% at Bakersfield. This is comparable to Pun et al. (2009), which
showed a 30e45% decrease in nitrate concentrations at Bakersfield
when anthropogenic NOx emissions were reduced by 50%.
Another major PM2.5 component in the SJV in wintertime is OC
which is a part of primary PM2.5. The model predicted that almost
all of the OC is due to primary emissions with very little contri-
bution from secondary formation. Therefore, reducing primary
PM2.5 emissions can directly reduce OC concentrations in the SJV.
Primary PM2.5 emission reductions also decreased concentrations
of EC, sulfate from primary sources, as well as the soil portion of the
PM2.5. In addition, reduced particulate surface area from the
reduced primary PM2.5 also decreased the heterogeneous forma-
tion of HNO3 from dinitrogen pentoxide (N2O5), leading to a
decrease in nitrate formation. This finding is different from that of
Pun et al. (2009), who found no significant sensitivity of nitrate
concentrations to primary PM2.5 emissions. The difference could be
due to a larger contribution from primary PM2.5 to the particulate
surface area in this study compared to that in Pun et al. (2009) as
well as the different parameterizations of heterogeneous reaction
probability for N2O5 in CMAQv4.7 (Davis et al., 2008; Foley et al.,
2010) compared to CMAQv4.6 used by Pun et al. (2009).
Reduction of NH3 emissions has a small effect on PM2.5 con-
centrations at Bakersfield. A 50% reduction of NH3 emissions in the
SJV reduced the NH4NO3 concentration at Bakersfield by approxi-
mately 6%. Kleeman et al. (2005) reported that a 50% reduction in
NH3 emissions lowered nitrate formation by 10%. Pun et al. (2009)
also reported a very small sensitivity of nitrate formation at
Bakersfield to NH3 emission reductions in the SJV. While NH3 does
not affect chemical formation of HNO3, it can influence the total
nitrate (gaseous HNO3 plus particulate nitrate) in the atmosphere.
Because gaseous HNO3 is more efficiently removed by dry deposi-
tion than particulate nitrate, the total nitrate in the atmosphere can
be influenced because of enhanced removal of total nitrate as
gaseous HNO3 increases relative to particulate nitrate (Pun et al.,
2009).
Reduction of SOx emissions has a minor effect on PM2.5 con-
centrations at Bakersfield. Sulfate only constitutes a small fraction
of PM2.5 at Bakersfield in winter. In addition, sulfate concentrations
include contributions from both primary emissions of sulfate as
well as upwind sources outside of the SJV. Therefore, reduction of
SOx emissions in the SJV only slightly reduced PM2.5 concentrations
in the SJV.
Anthropogenic VOCs emission reductions in the SJV were found
to slightly increase PM2.5 at Bakersfield. When anthropogenic VOCs
emissions in SJV were reduced, slightly higher NH4NO3 and slightly
lower secondary OC formation were predicted by CMAQ. The
 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190
Fig. 3. 24-h average PM2.5 carrying capacity diagrams at Fresno - 1st Street in 2019.
overall impact is very small. The PM2.5 design value at Bakersfield
was increased by 0.2 mg/m3 when anthropogenic VOCs emissions in
the SJV were reduced by half. Pun et al. (2009) also showed that in
the urban locations of the SJV, controls of anthropogenic VOCs
emissions only had small effects on nitrate concentrations.
The responses of PM2.5 concentrations at Fresno to reductions in
precursor emissions generally resembled those at Bakersfield. 50%
reductions of anthropogenic NOx and primary PM2.5 emissions in
the SJV lead to 20% and 28% reductions in PM2.5 at Fresno,
respectively. 50% reductions of anthropogenic NH3 and SOx emis-
sions in the SJV lowered PM2.5 by 5% and 1% at Fresno, respectively.
A 50% reduction of VOCs emissions slightly increased NH4NO3 and
slightly lowered secondary OC at Fresno, leading to a net decrease
of PM2.5 by 0.1 mg/m3. Compared to Bakersfield and Fresno, re-
ductions of anthropogenic NOx and primary PM2.5 emissions in the
SJV have smaller impacts at Stockton. 50% reductions in NOx and
primary PM2.5 emission reductions in the SJV lead to 14% and 18%
reductions in PM2.5 concentrations at Stockton, respectively, which
indicates larger impacts from sources outside the SJV at Stockton.
50% reductions of NH3 and SOx emissions in the SJV provided 6%
and 2% reductions in PM2.5 at Stockton, respectively. Reducing
anthropogenic VOCs emissions in the SJV by 50% produced no PM2.5
187
reduction at Stockton. The impacts from anthropogenic emission
reductions in the SJV showed similar patterns at the other seven
locations. On average, 50% emission reductions of primary PM2.5,
NOx, NH3, SOx, and VOCs lowered PM2.5 by approximately 24%, 23%,
5%, 1%, and 0%, respectively.
Table 2 shows the PM2.5 precursor equivalence ratios at the 10
locations in the SJV. These equivalence ratios show the ratio in the
reduction of 24-h PM2.5 concentrations at a particular location by
reducing a ton of other precursor (i.e., primary PM2.5, SOx, NH3 and
VOCs) emissions as compared to a ton of NOx emission reduction in
the SJV. The ratios shown in Table 2 were calculated from the
scenarios for which precursor emissions from anthropogenic
sources were reduced by 25% in the SJV. The equivalence ratios
based on 50% reductions of precursor emissions are somewhat
different given the nonlinearity in PM2.5 responses to precursor
reductions. Overall, on a per ton emission reduction basis, reducing
primary PM2.5 emissions is four times as effective in reducing PM2.5
levels as reducing NOx emissions. Reducing SOx emissions provided
similar efficiency as reducing NOx emissions, although there are far
less SOx emissions than NOx in the SJV. Reducing a ton of NH3
emissions provided approximately 10% of the benefit obtained from
a ton of reduction in NOx emissions. Reducing VOCs emissions
188 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190
Fig. 4. 24-h average PM2.5 carrying capacity diagrams at Stockton in 2019.
typically did not provide any benefit in terms of reducing PM2.5
concentrations at these sites.
3.3. Comparison to the year 2000-01 PM2.5 precursor sensitivity
analysis
PM2.5 precursor sensitivity analysis for the year 2000-01 was
performed using the UCD/CIT airshed model for the historical
period December 15, 2000eJanuary 7, 2001. A detailed description
of the modeling system and application to this period can be found
in Ying et al. (2008a). The emission inventory for this period was
prepared by CARB in a similar fashion as for the 2007 inventory
(Ying et al., 2008a). Numerous studies have evaluated the perfor-
mance of this modeling system, finding good agreement between
model predictions and measured PM2.5 mass, components of PM2.5
mass, and inferred source contributions to PM2.5 mass (Ying et al.,
2008a,b). Here, the response of PM2.5 nitrate, the key component
of PM2.5 in the SJV, to NOx and VOC emission controls are presented.
The effect of VOC emission controls on PM2.5 nitrate is of great
interest in the SJV (EPA, 2010). In addition to the CMAQ model,
using another independent model can corroborate model result
that supports policy development.
Fig. 5 shows the PM2.5 nitrate isopleths predicted at SJV loca-
tions using the UCD/CIT model for December 31, 2000. Daily-
average PM2.5 nitrate values are displayed for NOx and VOCs
scaling factors ranging from 40 to 100%. VOCs scaling was applied
proportionally to all sources, including biogenic sources, in the
simulation. The solid line in each sub-panel represents the
emissions trend for the SJV, which is represented as a fraction of
2000 emissions. Each dot on the emissions trend represents 5-
year intervals starting at the year 2000 (i.e., 2000, 2005, 2010,
and 2015). Future emissions projections are displayed as a dashed
line. Fig. 5 shows that Bakersfield, Fresno, and the SJV regional
average all achieve greatest the reduction in PM2.5 nitrate con-
centrations through emissions control programs that focus on NOx
control instead of VOCs control. In addition, the response is
generally consistent between past years (e.g., in 2000) and future
years (e.g., 2015). This result is logical since the major oxidant that
is active in the winter stagnation events is background ozone
transported into the SJV from the global background (Ying et al.,
2008a). Only a small amount of oxidant is produced by local
VOCs emissions and so VOCs controls have a negligible effect on
PM2.5 nitrate formation. The consistency of the precursor sensi-
tivities calculated by two different modeling systems for two
 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190 189

Table 2 criteria established in the literature. Consistent with observations,

PM2.5 precursor equivalence ratios at ten locations in the SJV.a CMAQ predicted that the two major PM2.5 components in the SJV
NOx Primary PM2.5 SOx NH3 VOCs in winter are NH4NO3 and OC. Generally, the CMAQ model under-
Bakersfield (California Avenue) 1.0 4.3 1.0 0.1 0.0
predicted OC at monitoring sites, likely due to a combination of
Bakersfield (E. Planz Road) 1.0 3.6 1.0 0.1 0.0 limitations of the emissions inventory, OC measurements, as well
Clovis 1.0 3.6 0.6 0.1 0.0 as model description of OC processing in the atmosphere. CMAQ
Corcoran 1.0 2.8 0.4 0.1 0.0 predicted a very small amount of secondary OC formation in the
Fresno (1st Street) 1.0 5.2 0.7 0.1 0.0
SJV in winter, an area worth further investigation considering the
Fresno (Hamilton & Winery) 1.0 4.7 0.8 0.1 0.0
Merced 1.0 3.8 1.0 0.1 0.0 potential importance of secondary OC formation in aerosol water
Modesto 1.0 4.8 2.0 0.2 0.0 and the existence of persistent wintertime fog in the SJV (Ge et al.,
Stockton 1.0 4.7 3.0 0.2 0.0 2012). Overall, the CMAQ model performance was acceptable for
Visalia 1.0 3.9 0.6 0.1 0.0
SIP applications. PM2.5 precursor sensitivities were evaluated
Average 1.0 4.1 1.1 0.1 0.0
based on projected 2019 emissions in the SJV. 24-hour PM2.5
a
Equivalence ratio shows the ratio in the reduction of 24-h PM2.5 concentrations concentrations showed strong responses to reductions of
at a particular location by reducing a ton of other precursor emissions as compared
anthropogenic primary PM2.5 and NOx emissions in the SJV. For
to a ton of NOx emission reduction in the SJV. They were calculated from the sce-
narios for which precursor emissions were reduced by 25% in the SJV. example, 50% reductions of primary PM2.5 and NOx emissions in
the SJV reduced PM2.5 at Bakersfield by 27% and 24%, respectively.
PM2.5 concentrations showed weak response to NH3 emission
different time periods confirms that PM2.5 nitrate in the SJV is reductions in the SJV. A 50% reduction of NH3 emissions in the SJV
most sensitive to NOx emissions control and not to VOCs emission only reduced PM2.5 concentrations at Bakersfield by approxi-
control. mately 4%. This demonstrated that NH4NO3 formation in the SJV
will be generally limited by the formation of HNO3 instead of the
4. Conclusions availability of NH3 in 2019. Given that sulfate only constitutes a
small fraction of PM2.5 in the SJV, reduction of SOx emissions in the
CMAQv4.7.1 was used for seasonal modeling (i.e., the first and SJV produced a very minor effect on PM2.5 concentrations. Emis-
fourth calendar quarters) of PM2.5 formation in the SJV. The CMAQ sion reductions of VOCs from anthropogenic sources in the SJV
model was found to generally meet PM2.5 model performance were not found to reduce PM2.5 concentrations. In the current

Fig. 5. 24-hr average PM2.5 nitrate concentrations response to NOx and VOCs controls on December 31, 2000 using the SAPRC90 chemical mechanism. (a) SJV average, (b) Fresno,
and (c) Bakersfield. Solid line with dots represents estimated emissions control trend since the year 2000 and dashed line with dots represents projected emissions controls towards
the year 2020 (dots represent 2000, 2005, 2010, and 2015 emissions. The projected 2020 NOx emissions are less than 40% of 2000 emissions. Therefore, the 2020 emission level is
not shown).
190 J. Chen et al. / Atmospheric Environment 92 (2014) 182e190
study, secondary OC was predicted to account for less than 5% of
OC. The effect of anthropogenic VOCs emission reductions on OC
concentrations is therefore minimal. In addition, the CMAQ and
the UCD/CIT modeling consistently showed that wintertime ni-
trate concentrations in the SJV respond strongly to NOx emission
controls but not to VOCs emission control.
Overall, this study showed the effectiveness of reducing primary
PM2.5 and NOx emissions for further reducing wintertime PM2.5
concentrations in the SJV in 2019. In the past, emission control
strategies have relied heavily on primary PM2.5 and NOx emissions
to reduce PM2.5 pollution in the SJV. The ambient air quality trend
of PM2.5 in the SJV demonstrates the effectiveness of those strate-
gies. Moving into the next decade, controls of primary PM2.5 and
NOx emissions should continue to be effective for the abatement of
PM2.5 pollution in the SJV.
Clearly, the response of PM2.5 to precursor emission reductions
in other regions of the world can be different from that in the SJV.
For example, sulfur and ammonia emission reductions can be
important for lowering PM2.5 in China (Wang et al., 2013). But,
overall, this study presents a framework that can be used to support
the development of PM2.5 control strategies with the CMAQ model
in other regions of the world.
5. Disclaimer
This paper has been reviewed by the staff of the California Air
Resources Board and has been approved for publication. Approval
does not signify that the contents necessarily reflect the views and
policies of the California Air Resources Board, nor does mention of
trade names or commercial products constitute endorsement or
recommendation for use.
Acknowledgments
We thank Daniel Chau and Vernon Hughes at CARB for
providing us with the meteorological fields and emission in-
ventories used in this study, and Karen Magliano of CARB for crit-
ically reviewing this manuscript, as well as staff at the San Joaquin
Valley Air Pollution Control District.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.atmosenv.2014.04.030.
References
Boylan, J.W., Russell, A.G., 2006. PM and light extinction model performance met-
rics, goals, and criteria for three-dimensional air quality models. Atmospheric
Environment 40, 4946e4959.
CARB, 2012. Modeling Protocol: Photochemical Modeling for the 24-hour PM2.5
State Implementation Plan in the San Joaquin Valley. Prepared by California Air
Resources Board & San Joaquin Valley Air Pollution Control District for the
United States Environmental Protection Agency Region IX. Available at: http://
www.valleyair.org/Workshops/postings/2012/12-20-12PM25/FinalVersion/16%
20Appendix%20F%20Modeling%20Protocol.pdf.
Carter, W.P.L., 2000. Implementation of the SAPRC-99 Chemical Mechanism into the
Models-3 Framework. Report to the United States Environmental Protection
Agency, Available at: http://www.engr.ucr.edu/wcarter/pubs/s99mod3.pdf.
Chen, J., Ying, Q., Kleeman, M.J., 2009. Source apportionment of visual impairment
during the California Regional PM10/PM2.5 Air Quality Study. Atmospheric
Environment 43, 6136e6144.
Chen, J., Ying, Q., Kleeman, M.J., 2010. Source apportionment of wintertime sec-
ondary organic aerosol during the California Regional PM10/PM2.5 Air Quality
Study. Atmospheric Environment 44, 1331e1340.
Chow, J.C., Chen, L.W.A., Watson, J.G., Loenthal, D.H., Magliano, K.A., Turkiewitx, K.,
Lehrman, D.E., 2006. PM2.5 chemical composition and spatiotemporal vari-
ability during the California Regional PM10/PM2.5 Air Quality Study (CRPAQS).
Journal of Geophysical Research. http://dx.doi.org/10.1029/2005JD006457.
Chow, J.C., Watson, J.G., Chen, L.W.A., Rice, J., Frank, N.H., 2010. Quantification of
organic carbon sampling artifacts in U.S. non-urban and urban networks. At-
mospheric Chemistry and Physics 10, 5223e5239.
Davis, J.M., Bhave, P.V., Foley, K.M., 2008. Parameterization of N2O5 reaction prob-
abilities on the surface of particles containing ammonium, sulfate, and nitrate.
Atmospheric Chemistry and Physics 8, 5295e5311.
Emmons, L.K., et al., 2010. Description and evaluation of the model for ozone and
related chemical tracers, version 4 (MOZART-4). Geoscientific Model Develop-
ment 3, 43e67.
Foley, K.M., et al., 2010. Incremental testing of the community multiscale air quality
(CMAQ) modeling system version 4.7. Geoscientific Model Development 3,
205e226.
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 the Air & Waste Management Association 56, 500e511.
Ge, X., Zhang, Q., Sun, Y., Ruehl, C.R., Setyan, A., 2012b. Effects of aqueous-phase
processing on aerosol chemistry and size distributions in Fresno, California,
during wintertime. Environmental Chemistry 9, 221e235.
Grell, G.A., Dudhia, J., Stauffer, D.R., 1994. A Description of the Fifth-Generation Penn
State/NCAR Mesoscale Model (MM5). NCAR Technical Note, TN-398þSTR,
Boulder, CO, U.S.A.
Guenther, A., Karl, T., Harley, P., Wiedinmyer, C., Palmer, P.I., Geron, C., 2006. Estimates of
global terrestrial isoprene emissions using MEGAN (model of emissions of gases
and aerosols from nature). Atmospheric Chemistry and Physics 6, 3181e3210.
Held, T., Ying, Q., Kaduwela, A., Kleeman, M.J., 2004. Modeling particulate matter in the
San Joaquin Valley with a source-oriented externally mixed three-dimensional
photochemical grid model. Atmospheric Environment 38, 3689e3711.
Kleeman, M.J., Ying, Q., Kaduwela, A., 2005. Control strategies for the reduction of
airborne particulate matter in California’s San Joaquin Valley. Atmospheric
Environment 39, 5325e5341.
Livingstone, P.L., et al., 2009. Simulating PM concentration during a winter episode
in a subtropical valley: sensitivity simulations and evaluation methods. Atmo-
spheric Environment 43, 6136e6144.
Lurmann, F.W., Brown, S.G., McCarthy, M.C., Roberts, P.T., 2006. Processes influ-
encing secondary aerosol formation in the San Joaquin Valley during winter.
Journal of the Air & Waste Management Association 56, 1679e1693.
Pope III, C.A., Dockery, D.W., 2006. Health effects of fine particulate air pollution: lines
that connect. Journal of the Air & Waste Management Association 56, 709e742.
Pun, B.K., Balmori, R.T.F., Seigneur, C., 2009. Modeling wintertime particulate matter
formation in central California. Atmospheric Environment 43, 402e409.
Robinson, A.L., et al., 2007. Rethinking organic aerosols: semivolatile emissions and
photochemical aging. Science 315, 1259e1262.
Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Barker, D.M., Wang, W.,
Powers, J.G., 2005. A Description of the Advanced Research WRF Version 2.
NCAR Technical Notes TN-468þSTR, Boulder, CO, U.S.A..
SJVAPCD, 2012. San Joaquin Valley Air Pollution Control District 2012 PM2.5 Plan.
Available at: http://www.valleyair.org/Air_Quality_Plans/PM25Plans2012.htm.
Turkiewicz, K., Magliano, K., Najita, T., 2006. Comparison of two winter air quality
episodes during the California Regional Particulate Air Quality Study. Journal of
the Air & Waste Management Association 56, 467e473.
U.S. EPA, 2007. Guidance on the Use of Models and Other Analyses for Demon-
strating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze.
EPA-454/B07-002.
U.S. EPA, 2010. Technical Support Document for the Proposed Action on the San
Joaquin Valley 2008 PM2.5 Plan and the San Joaquin Valley Portions of the
Revised 2007 State Strategy. Available at: http://www.epa.gov/region9/air/
actions/pdf/ca/SJV-TSD.pdf.
U.S. EPA, 2011. Memorandum on Update to the 24 hour PM2.5 NAAQS Modeled
Attainment Test. Available at: http://www.epa.gov/ttn/scram/guidance/guide/
Update_to_the_24-hour_PM25_Modeled_Attainment_Test.pdf.
Wang, Y., Zhang, Q., He, K., Zhang, Q., Chai, L., 2013. Sulfate-nitrate-ammonium
aerosols over China: response to 2000e2015 emission changes of sulfur diox-
ide, nitrogen oxides, and ammonia. Atmospheric Chemistry and Physics 13,
2635e2652.
Watson, J.G., Chow, J.C., 2002. A wintertime PM2.5 episode at the Fresno, CA
supersite. Atmospheric Environment 36, 465e475.
Ying, Q., Kleeman, M.J., 2006. Regional source contribution to secondary particulate
matter in California using a three-dimensional source-oriented air quality
model. Atmospheric Environment 40, 736e752.
Ying, Q., Lu, J., Allen, P., Livingstone, P., Kaduwela, A., Kleeman, M., 2008a. Modeling
air quality during the California Regional PM10/PM2.5 Air Quality Study
(CRPAQS) using the UCD/CIT source-oriented air quality model e part I. Base
case model results. Atmospheric Environment 42, 8954e8966.
Ying, Q., Lu, J., Allen, P., Kaduwela, A., Kleeman, M.J., 2008b. Modeling air quality
during the California Regional PM10/PM2.5 Air Quality Study (CRPAQS) using the
UCD/CIT source-oriented air quality model e Part II. Regional Source appor-
tionment of primary airborne particulate matter. Atmospheric Environment 42,
8967e8978.
Ying, Q., 2011. Physical and chemical processes of wintertime secondary nitrate
aerosol formation. Frontiers of Environmental Science and Engineering in China
5, 348e361.
Zhang, Y., Liu, P., Liu, X., Pun, B., Seigneur, C., Jacobson, M.Z., Wang, W., 2010. Finer
scale modeling of wintertime aerosol mass, number, and size distributions in
Central California. Journal of Geophysical Research 115, D15207. http://
dx.doi.org/10.1029/2009JD012950.