CIT source-oriented air quality model – Part III. Regional source apportionment of secondary and total airborne particulate matter

Atmospheric Environment 43 (2009) 419–430

Contents lists available at ScienceDirect

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

Modeling air quality during the California Regional PM10/PM2.5 Air Quality Study (CPRAQS) using the UCD/CIT source-oriented air quality model – Part III. Regional source apportionment of secondary and total airborne particulate matter Qi Ying a, b,1, Jin Lu a, b, Michael Kleeman a, * a b

Department of Civil and Environmental Engineering, University of California, Davis, CA 95616, USA Planning and Technical Support Division, California Air Resources Board, Sacramento, CA 95812, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 27 January 2008 Received in revised form 18 July 2008 Accepted 12 August 2008

A comprehensive air quality modeling project was carried out to simulate regional source contributions to secondary and total (¼primary þ secondary) airborne particle concentrations in California’s Central Valley. A three-week stagnation episode lasting from December 15, 2000 to January 7, 2001, was chosen for study using the air quality and meteorological data collected during the California Regional PM10/PM2.5 Air Quality Study (CRPAQS). The UCD/CIT mechanistic air quality model was used with explicit decomposition of the gas phase reaction chemistry to track source contributions to secondary PM. Inert artificial tracers were used with an internal mixture representation to track source contributions to primary PM. Both primary and secondary source apportionment calculations were performed for 15 size fractions ranging from 0.01 to 10 mm particle diameters. Primary and secondary source contributions were resolved for fugitive dust, road dust, diesel engines, catalyst equipped gasoline engines, non-catalyst equipped gasoline engines, wood burning, food cooking, high sulfur fuel combustion, and other anthropogenic sources. Diesel engines were identified as the largest source of secondary nitrate in central California during the study episode, accounting for approximately 40% of the total PM2.5 nitrate. Catalyst equipped gasoline engines were also significant, contributing approximately 20% of the total secondary PM2.5 nitrate. Agricultural sources were the dominant source of secondary ammonium ion. Sharp gradients of PM concentrations were predicted around major urban areas. The relative source contributions to PM2.5 from each source category in urban areas differ from those in rural areas, due to the dominance of primary OC in urban locations and secondary nitrate in the rural areas. The source contributions to ultra-fine particle mass PM0.1 also show clear urban/rural differences. Wood smoke was found to be the major source of PM0.1 in urban areas while motor vehicle sources were the major contributor of PM0.1 in rural areas, reflecting the influence from two major highways that transect the Valley. Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: Source apportionment Secondary particulate matter Ultra-fine particulate matter Central California CRPAQS

1. Introduction * Corresponding author. E-mail address: [email protected] (M. Kleeman). 1 Present address: Zachry Department of Civil Engineering, Texas A&M University, College Station, TX 77843, USA. 1352-2310/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2008.08.033

The San Joaquin Valley (SJV) experiences some of the worst wintertime particulate air quality pollution in the United States (American Lung Association, 2005). During the recent California Regional PM10/PM2.5 Air Quality Study

420

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

(CRPAQS), the fine PM concentration in the southern portion of the SJV reached a peak value of 200 mg m3 at Bakersfield (Chow et al., 2006a). Approximately 50% of the PM2.5 was secondary ammonium nitrate that formed in the atmosphere from gas phase precursors (Herner et al., 2005). The sources of this secondary PM cannot be determined using traditional statistical source apportionment methods and so new techniques must be used to identify the origin of the winter PM problem in the SJV. Mechanistic air quality models can predict changes in secondary PM concentrations in response to changes in precursor emissions (Stockwell et al., 2000; Pun and Seigneur, 2001). Mechanistic air quality models can also be used to identify source contributions to secondary PM (Mysliwiec and Kleeman, 2002; Kleeman et al., 2007). Previous modeling studies applied to a SJV winter air pollution episode that occurred in 1996 (Held et al., 2004) attempted to identify source contributions to the regional distribution of both the primary and secondary PM (Ying and Kleeman, 2006). This previous analysis was limited by the short duration of the study period (only three days) and the small size of the study domain (southern portion of the SJV only) leading to the conclusion that a large fraction of the secondary PM was transported from ‘‘upwind’’ sources or formed before the start of the three-day episode. The wintertime CRPAQS study was designed to provide a larger spatial and temporal coverage of air quality and meteorology in the SJV to help better understand the sources and formation mechanisms of PM (Chow et al., 2006a). The CRPAQS study covers the entire Central Valley of California including regions upwind of the SJV. Air quality and meteorology data were collected at over 100 stations throughout the Valley during the 13-month field campaign starting in December 1999. In addition, large amounts of aerosol chemical speciation and size distribution data were collected during four Intensive Operation Periods (IOPs) between December 2000 and February 2001 when PM2.5 concentrations reached their highest values. The purpose of the current paper is to determine the regional source contributions to secondary PM during CRPAQS. Source contributions to total (¼primary þ secondary) PM2.5 and ultra-fine PM (PM0.1) are also discussed. This work, together with the base case simulation (Ying et al., in press-b) and regional primary source apportionment study (Ying et al., in press-a), represents the first source-oriented air quality model application to study the regional PM formation and source apportionment of PM over a multi-week episode during the winter in central California. 2. Model description Regional source apportionment calculations for secondary PM are carried out using the source-oriented UCD/CIT air quality model. A comprehensive description of the UCD/CIT air quality model can be found in the base case paper (Ying et al., in press-b) and the references therein and so only the details related to source apportionment of secondary PM are discussed here. Source contributions to secondary PM are calculated with source-oriented gas phase chemistry and gas-to-particle partitioning (Mysliwiecand Kleeman, 2002). The source-oriented gas phase chemistry

model tracks the precursor gases (NOx, NH3 and SO2) emissions from different sources through the complex nonlinear chemical reactions separately so that the source origin of the semi-volatile products can be explicitly B retained. For a simple example, NOA 2 and NO2 will be used to represent NO2 from diesel engines and gasoline engines. The chain-termination reaction that produces the semivolatile HNO3 will be expanded into two reactions: NOA2 þ OH / HNOA3

(1)

NOB2 þ OH / HNOB3

(2)

B Writing separate equations for NOA 2 and NO2 allows us to B separately quantify the buildup of HNOA 3 and HNO3. In reality, a large number of chemical reactions and intermediate species need to be expanded to properly retain the source information for semi-volatile products such as nitrate, ammonium ion, and sulfate from multiple emission categories. This is accomplished using automated software that expands chemical reaction mechanisms written in the State Air Pollution Research Center (SAPRC) format. The gasto-particle conversion routine that calculates the dynamic exchange of material between the gas and particle phases is also expanded to explicitly track semi-volatile products from each source category. Source-oriented concentrations for each semi-volatile species are aggregated during vapor pressure calculations so that the thermodynamics package that calculates the surface vapor pressure of semi-volatile species is not modified. Concentrations of secondary organic aerosol are expected to be small during the cold winter conditions experienced during the current study, and so source apportionment of SOA is not considered in the current analysis. The approach for the source apportionment of secondary PM described above is independent of the choice for primary particle representation in the model (internal mixture vs. source-oriented external mixture). When primary particles are tracked as a source-oriented external mixture, the secondary source apportionment calculations predict the amount of nitrate, sulfate, and ammonium ion originating for each source category that forms on primary particle cores released from each source category. As an example, it is possible that NOx emitted from diesel engines can form nitrate on primary particles originally released from wood combustion. When primary particles are tracked as an internal mixture, the secondary source apportionment calculations still predict source contributions to secondary PM, but the source origin of the primary particle core is not known. The internal vs. external mixture representation for primary particles may influence the overall aerosol chemistry, especially during periods of high relative humidity (Kleeman et al., 1997). The sensitivity of total nitrate formation to internal vs. source-oriented external mixture treatments for primary particles will be discussed in Section 4. Several components of the UCD/CIT air quality model were updated in the current study in an attempt to improve the prediction of secondary nitrate formation during the wintertime episode. Recent experimental studies show that the accommodation coefficient (a) for N2O5 hydrolysis

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

f ¼

½SðVIÞ ½SðVIÞ þ ½NðVÞ

a ¼ 0:02  f þ 0:002  ð1:0  f Þ

(3) (4)

The gas phase pollutant dry deposition scheme used in the UCD/CIT air quality model was also updated in an attempt to improve nitrate predictions using the deposition model described by Walmsley and Wesely (1996). In the original UCD/CIT model, the dry deposition of SO2 and O3 is directly calculated based on a table of surface resistance as a function of the solar radiation intensity, and does not consider the possible change in the surface resistance due to seasonal variations (Russell et al., 1993). The deposition velocity of other species is either set as constant or scaled based on the SO2 value. The Walmsley and Wesely scheme allows the direct determination of deposition velocities of 10 important gas species using the calculated solar intensity and the season-dependent surface resistance values. This modification allows the model to calculate dry deposition more accurately as a function of season and under different solar intensity conditions. The sensitivity of predicted nitrate formation and source apportionment results due to the changes to the model described above will be discussed in Section 5.2. 3. Model application The UCD/CIT source-oriented air quality model was configured to use the internally mixed particle representation with artificial tracers for primary source apportionment and expanded reaction chemistry for secondary source apportionment. Model calculations were carried out for the period December 15, 2000–January 7, 2001. This period includes three IOPs: IOP1 (December 15–18, 2000), IOP2 (December 26–28, 2000) and IOP3 (January 4–7, 2001). Regional source contributions to secondary PM2.5, total (¼primary þ secondary) PM2.5 and total PM0.1 were resolved for the entire central California region including the SJV. The source contributions to primary PM2.5 were calculated simultaneously with the secondary PM and the results are documented in a separate paper (Ying et al., in press-a). The simulation was carried out using 4 km horizontal grid resolution with 190  190 grid cells in the domain that covers the entire Central Valley of California. The computation domain covers land areas with surface elevation below 2000 m and ocean regions up to 100 km from the coastline (see Figure 1 of Ying et al., in press-a).

Details about the model configuration and the preparation of the model input fields describing meteorology, emissions, initial and boundary conditions are described by (Ying et al., in press-b) and are not repeated here. Table 1 in Ying et al. (in press-a) summarizes the domain-average emission totals of major pollutants and precursors. Gasoline engines and diesel engines are the two dominant sources of NOx during the study episode. NOx and VOC emissions from diesel engines are approximately twice as high as gasoline engines during the current study. This contrasts with the SJV emissions inventories for January 1996 (Held et al., 2004) where emissions from gasoline engines were estimated to be approximately twice as high as those from diesel engines. Preliminary data analysis indicates a trend of decreasing VOC/NOx ratios over time at multiple sites within the SJV and the South Coast Air Basin in California consistent with a shift from emissions dominated by gasoline engines to emissions dominated by diesel engines. The implication of these emission trends for future emission control strategies will be discussed more comprehensively in a separate manuscript. 4. Results The base case predictions of gas and particulate pollutant concentrations have been compared extensively with observations and shown to be satisfactory at most sites (Ying et al., in press-b). The Fractional Bias (FB) values for PM2.5 mass, nitrate, ammonium ion, EC and OC are mostly within the range of 0.5 to 0.5 across all stations and all dates. This performance meets the FB criteria (0.6  FB  0.6) suggested by Boylan and Russel (2006).

10

External Mixture (µg m-3)

on wet particles is a function of particle composition (Riemer et al., 2003; Brown et al., 2006). The accommodation coefficient of N2O5 in the current model was revised from a constant value of 0.001 to be a function of aerosol sulfate and nitrate concentration, as shown in the Eqs. (3) and (4) below, based on Riemer et al. (2003). The [S(VI)] and [N(V)] represent the particle sulfate and nitrate concentrations, respectively. This parameterization allows a higher accommodation coefficient for particles that are mainly sulfate and a lower coefficient for nitrate dominant particles. For particles that do not have any nitrate or sulfate, an accommodation coefficient of 0.002 is used.

421

1

0.1 Dust Wood Smoke Diesel Engines Cat Engines Non-cat Engines Meat Cooking High Sulfur Fuel Other

0.01 0.01

0.1

1

10

Internal Mixture (µg m-3) Fig. 1. Source contribution to the 24-h average PM2.5 nitrate on December 26, 2000 calculated using the externally mixed aerosol model and the internal mixture with artificial tracer model. Different symbols represent different emission source categories. The data points included in the figures are predicted concentrations at Bethel Island, Sacramento, Fresno, Angiola and Bakersfield.

422

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

The UCD/CIT model predictions for source contributions to primary particulate matter were also found to be in good agreement with CMB predictions carried out using molecular markers (Ying et al., in press-a). Both of these comparisons are necessary quality control checks on the model simulation that build confidence in the results before secondary source apportionment calculations can be considered. Analysis shows that model calculations successfully capture the rapid regional buildup of secondary particulate nitrate during IOP2 (Ying et al., in press-a) of the current study. Model performance is best during the last week of December 2000 and there is no significant bias toward under or over-predictions across the entire episode. UCD/CIT model calculations in the current study used an internal mixture particle representation with inert source tracers for primary source apportionment. Previous work has shown that the predicted source contributions to primary PM using the internal mixture approach are in good agreement with predictions made using a sourceoriented external mixture particle representation (Ying et al., in press-a). Fig. 1 compares source contributions to secondary nitrate predictions made using internal vs. source-oriented external mixture particle representations. Different symbols on the figure correspond to different

source categories. For each source category, the predicted concentrations at five stations (Bethel Island, Sacramento, Fresno, Angiola and Bakersfield) are shown. The source contributions predicted by the internally mixed model with artificial tracers agree closely with the source-oriented externally mixed aerosol approach. This is no surprise since both models use essentially the same expanded reaction chemistry for the source apportionment of secondary PM components. The regional difference in predicted total PM2.5 nitrate concentration using the internal and external particle representation will be examined in Section 5.2.4. 4.1. Source apportionment of secondary PM at Fresno Fig. 2 shows the predicted hourly averaged relative source contributions to PM2.5 nitrate (N(V)), sulfate (S(VI)), and ammonium ion (N(-III)) for Fresno during the study period. Panel 2(a) shows the calculated source contribution to PM2.5 nitrate at Fresno. The initial concentration accounts for a major fraction of the nitrate in the beginning of the simulation but the initial conditions become negligible after two simulated days because most of these particles are advected out of the modeling domain. Previous studies have shown that dry deposition is not a significant fine

Fig. 2. Relative source contribution to PM2.5 nitrate N(V), sulfate S(VI), and ammonium ion N(-III) at Fresno from December 15, 2000 to January 7, 2001. Periods where boundary conditions dominate nitrate relative contributions correspond to low concentration events.

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

particle removal mechanism compared to advection (Herner et al., 2006). Diesel engines are the main contributors to the secondary nitrate concentrations at Fresno, with an approximate relative contribution of 40%. Gasoline engines contribute about half as much nitrate as diesel engines. This differs from the previous SJV simulation (Ying and Kleeman, 2006) in which the contribution from gasoline engines was higher than the contribution from diesel engines. These changes reflect differences in the emissions inventory over time. Panel 2(b) shows that 80% of the predicted sulfate concentrations originated from background non-sea-salt sulfate. Local sulfate production is low due to the low emissions of SO2 and slow SO2 oxidation rates in cold winter conditions. Aqueous phase sulfate production was not considered in this simulation and is likely insignificant due to low oxidants (O3 and H2O2) concentrations. Panel 2(c) shows that approximately

423

80% of the ammonium ion originates from ‘‘other’’ sources that includes dairy operations. Approximately 10% of the ammonium ion originates from wood smoke. Boundary conditions account for a further 10% of the ammonium ion, with slightly higher values during the day as both the wind speed and mixing height increase. Contributions from catalyst equipped gasoline engines account for only a small fraction of the ammonium ion. The relative source contributions illustrated in Fig. 2 can be converted to absolute concentration units (mg m3) by multiplication with the total concentration of each component summed across all sources. Figure 5 of Ying et al. (in press-b) illustrates that PM2.5 nitrate concentrations increase from w10 to 35–40 mg m3 and PM2.5 ammonium ion concentrations increase from w1 to w14 mg m3 during the study period. PM2.5 sulfate concentrations are relatively constant at w1–2 mg m3.

Fig. 3. Source contribution to PM2.5 nitrate concentrations on December 28, 2000. The scale on each panel is different. Units are mg m3.

424

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

4.2. Regional source contribution to secondary and total PM A unique feature of the source-oriented modeling technique applied in a 3D Eulerian air quality model is that source contributions to pollutant concentrations can be mechanistically determined over the entire model domain. Figs. 3–5 illustrate the capability of the CIT/UCD sourceoriented air quality model to determine the spatial distribution of secondary PM components as well as total PM from major sources. Companion regional source contribution plots for primary pollutants (OC and EC) are included in part two of the paper series (Ying et al., in press-a). Together, these figures illustrate the different spatial distribution patterns of primary and secondary pollutants. Quantitative knowledge about the spatial distribution of pollutant concentrations from different sources is important for the design of effective emission control strategies. Fig. 3(a) shows that predicted 24-h average PM2.5 nitrate concentrations on December 28, 2000 range from 10 to 20 mg m3 along the edges of the mountain boundaries to a maximum concentration of approximately 42 mg m3 in the areas southeast of Fresno. Contributions to PM2.5 nitrate from wood smoke are not significant (Panel 3(b)). Panels 3(c–e) show the contributions to PM2.5 nitrate from diesel engines, non-catalyst equipped gasoline engines and catalyst equipped gasoline engines, respectively. Diesel engines and catalyst equipped gasoline engines are the two most important sources that contribute to the elevated secondary nitrate concentrations in the Central Valley. The spatial distributions of nitrate from these three sources are similar, with high concentrations throughout most of the

SJV. The majority of the NOx associated with mobile sources is released in the urban areas of Fresno and Bakersfield as well as from the I-5 and CA-99 freeways that transect the valley. The NOx is not immediately transformed into nitric acid/ammonium nitrate due to slow photochemical reactions in the winter. Background ozone is the most important oxidant for reactive nitrogen with gradual conversion of NOx to particulate nitrate over several days. Wind speed and direction also vary significantly from day to day. These factors allow NOx emissions to be transported to remote regions of the Valley before being transformed into particulate nitrate and lead to the predicted high nitrate concentration throughout the entire Valley. Predicted PM2.5 nitrate concentrations are lower in the northern part of the Central Valley due to significant wind ventilation that moves the pollutants to the San Francisco Bay area. The maximum PM2.5 nitrate concentrations from diesel engines and catalyst equipped gasoline engines are approximately 19 and 11 mg m3, respectively. Non-catalyst gasoline engines have a maximum PM2.5 nitrate contribution of 1.8 mg m3 and are not significant sources of particulate nitrate in the current study. Panel 3(f) shows that high sulfur fuel combustion makes a peak contribution of approximately 3 mg m3 to PM2.5 nitrate in the area southeast of Fresno. Panel 3(g) shows that other anthropogenic sources contribute less than 6 mg m3 of PM2.5 nitrate in most portions of the Valley. At some isolated locations, the contribution from ‘‘other’’ sources can reach as high as 11 mg m3. Panel 3(h) shows uniform PM2.5 nitrate concentration of w4 mg m3 in the Valley from background NOx sources.

Fig. 4. Source contribution to PM2.5 ammonium ion concentrations on December 28, 2000. The scale on each panel is different. Units are mg m3.

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

425

Fig. 5. Source contribution to total PM2.5 mass concentrations on December 28, 2000. The scale on each panel is different. Units are mg m3.

Fig. 4 shows the regional ammonium ion concentrations and the major sources that contribute to the predicted ammonium ion concentrations. Panel 4(a) shows that the predicted maximum 24-h average PM2.5 ammonium ion concentration on December 28, 2000 is approximately 15 mg m3. Ammonium ion and nitrate have very similar spatial distributions since NH3 tends to condense together with HNO3 to neutralize the acidity of the particles. Panel 4(b) shows that the contribution of wood smoke to ammonium ion is most noticeable in urban areas with a maximum 24-h average contribution of 1.5 mg m3. Contributions from non-catalyst equipped gasoline engines are small based on Panel 4(c). As shown in Panel 4(d), the ammonium ion concentration associated with catalyst equipped gasoline engines reaches a maximum of 1.8 mg m3 in the San Francisco Bay Area. Panel 4(e) shows that majority of the ammonium ion is from ‘‘other’’ sources

that includes dairy emissions of NH3. The maximum ‘‘other’’ ammonium ion concentration coincides with the location of dairy operations in the central SJV between Fresno and Angiola. Panel 4(f) shows that the influence of background NH3 is small, with approximately 0.5 mg m3 of ammonium nitrate attributed to background sources near the computational boundary. Fig. 5 shows the major sources that contribute to the total (¼primary þ secondary) PM2.5 mass concentrations on December 28, 2000. The dust category (5(a)) includes fugitive dust and paved road dust and is mainly composed of primary particles. The model over-predicts the dust components by a significant amount. Based on a recent CMB analysis (Chow et al., 2006b), the averaged contributions of geological components to PM2.5 averaged over the three simulated IOPs are approximately 0.23  0.18 mg m3 at Fresno and 3.7  3.3 mg m3 at Angiola. The modeled

426

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

secondary nitrate (Fig. 3f). Panel 5(i) shows that background sources contribute approximately 9.5 mg m3 of total PM2.5 in the SJV. This material originates mainly from background NH3, NOx and PAN that is gradually transformed into secondary PM during the study episode. Fig. 6 shows the relative source contribution to nitrate and total PM2.5 mass for the entire SJV and at the Fresno site averaged over the duration of the model episode. The regional value was calculated as the average of the relative contributions for each source at each grid cell within the SJV. Uncertainty ranges were calculated as the standard deviation of the relative source contribution for all the grid cells. The average PM2.5 nitrate concentration for the SJV is 16.3 mg m3. Diesel engines are the largest source of nitrate in the SJV during the study episode (35.9  6.51%), followed by catalyst equipped gasoline engines (18.9  3.8%). The Fresno site shows lower averaged nitrate concentrations (11.2 mg m3) but very similar relative nitrate source contributions (34.9% and 20.4% for diesel and gasoline engines, respectively) compared to the regional average. A summary of NOx emissions for a typical weekday (Ying et al., in press-a) shows that diesel engines account for approximately 47% of the total NOx emissions in this region. Considering that 21.7  12.0% of nitrate is produced by reactive nitrogen from boundary conditions in the current study, the model prediction for nitrate production from diesel engines is in reasonable agreement with the emissions inventory totals. Fig. 6(b) shows that the relative source contributions to PM2.5 mass from each source category in urban areas differ from those in rural areas. Secondary nitrate dominates the total PM2.5 mass concentration in the rural areas. At Fresno, particles emitted from wood smoke (mainly primary particles) are the dominant source of total PM2.5 mass. Larger variations in the relative contributions were predicted for wood smoke particles than diesel and catalyst

Relative PM2.5 (%)

Relative PM2.5 NO3- (%)

results averaged over the entire modeling domain are 7.3 and 9.6 mg m3 at Fresno and Angiola, respectively. The over-prediction is mainly due to the overestimate of dust emissions in the emission inventory as discussed in Ying et al. (in press-a, in press-b). Meat cooking (5(b)) and wood smoke (5(c)) are also mainly composed of primary particles from urban centers. Wood smoke dominates the total PM2.5 mass concentrations with contributions as high as 60 mg m3. The contributions from diesel engines (5(d)) includes primary elemental carbon (EC), primary organic compounds (OC), and secondary nitrate with a maximum total concentration of approximately 16 mg m3. The primary contribution to the total diesel PM is higher in urban areas as most of the PM from diesel in rural areas of the Central Valley is secondary (Fig. 3). Non-catalyst equipped gasoline engines are not a major source of PM2.5, reflecting their small contribution in the total vehicle fleet. The total PM2.5 from catalyst equipped gasoline engines has a maximum concentration of 7.5 mg m3. The results show that the contribution from catalyst equipped gasoline engines to total PM2.5 is half that from diesel engines during the current study. This ratio reflects a change in the diesel/gasoline emission ratio in the emission inventory since 1996 (Held et al., 2004). Diesel engines emit more NOx but less VOC than gasoline engines, meaning that the ambient NOx to VOC ratio in the SJV has also increased. The consequence of this emissions trend for summer ozone concentrations will require further investigation since the SJV currently violates the ozone National Ambient Air Quality Standards (NAAQS). High sulfur fuel combustion (5(g)) contributes significantly to PM2.5 near the two air force bases in central California due to the use of high sulfur jet fuel. Approximately 30 mg m3 of PM2.5 in the SJV originates from sources that are not explicitly resolved in this study of which 15 mg m3 is due to secondary ammonium ion (mainly from dairy sources) (Fig. 4f) and 11 mg m3 is 70 60

a

Avg. PM2.5 Nitrate at Fresno = 11.2 µg m-3

Fresno Regional Avg.

Avg. PM2.5 Nitrate at SJV = 16.3 µg m-3

50 40 30 20 10 0 50 40

b

Avg. PM2.5 at Fresno = 64.2 µg m-3 Avg. PM2.5 at SJV = 38.5 µg m-3

30 20 10 0 O

S.

er

th

h ig

H

el Fu

e

ok Sm

e

in ng

.E

s ne gi

En

d oo

W

. at

C

ne

gi En

at C

el

on

N

s ei

D

C

t us

D

/B

IC

Fig. 6. Relative source contribution to (a) PM2.5 nitrate and (b) PM2.5 mass averaged from December 15, 2000 to January 7, 2001 for Fresno and the entire SJV region.

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

equipped gasoline engines because wood smoke particles are mainly primary and have significant spatial gradients around emission centers. This urban/rural difference also indicates that primary particles are mainly located in areas near their emission sources and are generally not transported over long distances during the current study. Gasoline and diesel engines contribute both primary PM mass and secondary PM nitrate. The uniform distribution of secondary nitrate reduces the variation in the total PM contributions from these sources among grid cells. Fig. 7 shows the calculated source contributions to total (¼primary þ secondary) PM0.1 and PM2.5 mass concentrations averaged over the entire modeling episode (December 15, 2000–January 7, 2001) along a transect line in the Valley that passes through Bakersfield and Sacramento. The Lambert Y positions for Bakersfield and Sacramento are 177.3 and 174.9 km, respectively. Road dust and fugitive dust sources are combined into a single dust source category in this figure. Panel 7(a) shows the average source contribution to PM0.1 mass. Two sharp concentration peaks can be seen around Bakersfield and Sacramento. The concentration gradient is most significant around Bakersfield, where concentrations decrease by a factor of 10 in approximately 25 km. PM0.1 concentrations between the two major urban areas are much lower (approximately 2.5 mg m3) and relatively uniform. Wood smoke accounts

427

for a majority of the ultra-fine particle mass (PM0.1) in the urban areas while particles from transportation related sources account for larger fractions of PM0.1 in rural areas. Model calculations predict that most of the PM0.1 mass is composed of primary EC and OC. Panel 7(b) shows the source contributions to total PM2.5 mass concentrations along the Bakersfield–Sacramento transect line averaged over the entire study period. The maximum concentrations occur in the region surrounding Bakersfield, with the highest predicted concentration approaching 80 mg m3. The urban peak of Sacramento can also be seen on the figure with highest episode-average concentration reaching approximately 55 mg m3. The largest sources of PM2.5 in these two urban areas are wood smoke and diesel engines. The PM2.5 concentrations in the rural areas are also high, with maximum episode-average concentrations of 60 mg m3 in areas between Fresno and Angiola. Secondary ammonium nitrate dominates the PM2.5 concentrations in the rural area, with most of the nitrate originating from diesel and gasoline engines. A significant contribution from the ‘‘other’’ source includes ammonium ion from animal sources. The contribution of dust particles to the average PM2.5 concentrations are likely over-estimated (see Section 4.1). The episode-average (24day) PM2.5 concentrations along this transect line in the Valley are higher than the newly proposed 24-h average PM2.5 NAAQS of 35 mg m3.

5. Discussion 5.1. Factors affecting source apportionment predictions

Fig. 7. Source contribution to PM0.1 and PM2.5 along a line passing through Sacramento and Bakersfield averaged from December 15, 2000 to January 7, 2001.

The CIT/UCD model directly determines the pollutant concentrations from different sources through a mechanistic simulation of the emission, transport, deposition and chemical transformation processes. The predicted absolute source contribution to PM or its components (in mg m3) for each source is affected by all these simulated processes. For example, the accuracy of the meteorology and emission fields, the mechanisms that describe the actual deposition, chemical reaction and gas-to-particle partition processes, and the numerical schemes to solve the underlying differential equations all affect the predicted absolute source concentrations. The relative source contribution (in percentage) is more likely to be affected by only a subset of the model processes: the physical/chemical properties of the pollutants, the spatial location of the emission sources and the wind speed/direction. Sources with same emission rates but different spatial distributions will produce different spatial patterns of downwind concentrations. Wind speed and wind direction, along with the physical/chemical properties of the pollutants, govern where the pollutants will be transported and how they will be transformed in the atmosphere. Chemical mechanisms and numerical solution techniques affect the predicted absolute concentration associated with each source in similar ways, and so they produce less of an effect on calculated relative source contributions. More detailed discussion of the transport and formation of pollutants in central California during this

428

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

air quality episode can be found in Ying and Kleeman (submitted for publication). A full sensitivity study to quantify the predicted source contributions due to the processes discussed above is beyond the scope of this study. Instead, particular attention was given to secondary nitrate. Several parameters that affect nitrate formation and predicted source contributions are discussed in the following section. 5.2. Sensitivity analysis for nitrate formation and source apportionment Previous work has shown that the UCD/CIT model captures regional nitrate formation mechanisms adequately but under-predicts local nitrate formation at Bakersfield during the latter portion of the current study. Sensitivity studies were carried out on several parameters that affect nitrate formation in an attempt to explain the nitrate under-prediction. 5.2.1. Dry deposition The original dry deposition scheme used by the UCD/CIT air quality model was developed for a summer smog simulation in Southern California (Russell et al., 1993) and so the estimated deposition rates are higher than those

calculated by the Walmsley and Wesely scheme for the winter season. It is expected that enhanced HNO3 deposition predicted by the original scheme will lead to lower nitrate concentrations. A sensitivity run using the original dry deposition scheme was used to quantify this effect. Fig. 8(a) shows that the Walmsley and Wesely scheme leads to an increase of approximately 6 mg m3 in the predicted 24-h averaged nitrate concentrations on December 28, 2000, due to lower dry deposition velocities. 5.2.2. Temperature variation Temperature is also a key factor that affects the nitrate formation. Gas/particle equilibrium of ammonium nitrate is highly temperature dependent (Aw and Kleeman, 2003). Increased temperature moves the gas/particle equilibrium toward gas phase. Another competing effect is that the chemical reaction rates are also temperature dependent. Higher temperatures lead to higher reaction rates and increased production of nitric acid. The input temperature for model calculations was uniformly decreased by 2  C to study the effect on nitrate formation during the current study. The relative humidity was held constant during this simulation. Fig. 8(b) shows that lowering the temperature by two degrees uniformly decreased 24-h averaged nitrate concentrations in the Central Valley on December 28, 2000,

Fig. 8. Change in 24-h average PM2.5 nitrate concentrations on December 28, 2000 due to (a) updated deposition scheme, (b)2  C temperature change, (c) lower N2O5 accommodation coefficient and (d) external vs. internal particle representation. Units are in mg m3.

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

by approximately 1–2 mg m3. The results indicate that temperature effects on reaction rates are more significant than temperature effects on gas/particle partitioning under the current meteorological conditions. 5.2.3. N2O5 accommodation coefficient The N2O5 accommodation coefficient describes the probability that a N2O5 molecule that strikes a particle surface will stick. The heterogeneous reaction of N2O5 on particle surfaces is one of the most important pathways for the formation of secondary nitrate (Jacob, 2000). Accommodation coefficients used in previous modeling studies have varied significantly from 0.005 to 0.1 and may also be particle composition dependent (Evans and Jacob, 2005). An additional simulation was performed in the current study using a fixed N2O5 accommodation coefficient of 0.001 to test the lower limit of N2O5 hydrolysis during winter conditions in central California. Fig. 8(c) shows the change of 24-h average nitrate concentrations on December 28, 2000 when this change was made. Nitrate concentrations decreased by approximately 2 mg m3 in the northern portion of the Valley due to a lower N2O5 accommodation coefficient. In the southern part of the Valley, nitrate concentrations decreased by as much as 8 mg m3 in the region south of Bakersfield. These results indicate that N2O5 heterogeneous reaction is a significant pathway of wintertime particulate nitrate formation in the SJV and the amount of nitrate formed is quite sensitive to the selection of N2O5 accommodation coefficient. 5.2.4. Internal/external particle representation The internal mixture particle representation was used to generate the results presented in the previous sections. The source-oriented externally mixed particle representation is a more realistic way of representing particles in urban and regional airborne particles. Fig. 8(d) shows the change of the predicted 24-h average nitrate concentrations on December 28, 2000 using the source-oriented externally and internally mixed particle representations. Under the current modeling episode, little difference in the predicted nitrate concentration is noticed using the two different particle representations. The maximum difference associated with particle mixing state assumptions is 1.7 mg m3, with more typical values of 0.5 mg m3 at most locations. The close agreement of the internally mixed and externally mixed results shows that the internally mixed particle representation can be used as a base case calculation for future modeling studies. 5.2.5. Effect on nitrate source apportionment Bias in the dry deposition rates, temperature, N2O5 accommodation coefficients, and internal vs. sourceoriented external mixture particle representations are expected to affect the absolute source contributions but not greatly affect the relative source contribution results (see Section 5.1). Analysis of the source apportionment results from the current sensitivity simulations determines that the relative contributions of diesel engines to secondary nitrate averaged throughout the modeling domain for December 28, 2000 are 35.7%, 39.3%, 40.1% and 39.3%, respectively. These results are not significantly different from the base case source contribution result of

429

35.9  6.51% for diesel engines for that day. Changes to the dry deposition scheme have slightly larger impact than other factors because the updated dry deposition scheme reduces the amount of primary NOx in the surface layer, leading to an overall decrease in total nitrate concentrations and an increase in the relative contribution from background nitrate. 6. Conclusions Diesel engines are the largest source of secondary nitrate in central California during a severe winter air pollution episode, accounting for approximately 40% of the total PM2.5 nitrate. Catalyst equipped gasoline engines are also significant sources of secondary nitrate, contributing approximately 20% of the PM2.5 nitrate. Sharp gradients of total (¼primary þ secondary) PM concentrations were predicted around major urban areas. The relative source contributions to total PM2.5 from each source category in urban areas differ from those in rural areas, due to the dominance of primary OC in urban locations and secondary nitrate in the rural areas. The source contributions to ultrafine particles also show clear urban/rural differences. Wood smoke is the major source of PM0.1 in urban areas while motor vehicle sources are the major contributor of PM0.1 in rural areas, reflecting the influence from two major highways that transect the Valley. Acknowledgment This research is supported by the San Joaquin Valleywide Air Pollution Study Agency and the California Air Resources Board. The statements, opinions, findings, and conclusions of this paper are those of the authors and do not necessarily represent the views of the California Air Resources Board. The authors thank Dr. John Watson and Dr. Douglas Lowenthal at the Desert Research Institute for CMB source apportionment results. References American Lung Association, 2005. State of the air 2005. Tech. rep. Aw, J., Kleeman, M.J., 2003. Evaluating the first-order effect of intraannual temperature variability on urban air pollution. Journal of Geophysical Research – Atmospheres 108 (D12). Boylan, J., Russel, A., 2006. PM and light extinction model performance metrics, goals, and criteria for three-dimension air quality models. Atmospheric Environment 40 (26), 4946–4959. Brown, S.S., Ryerson, T.B., Wollny, A.G., Brock, C.A., Peltier, R., Sullivan, A.P., Weber, R.J., Dube, W.P., Trainer, M., Meagher, J.F., Fehsenfeld, F.C., Ravishankara, A.R., 2006. Variability in nocturnal nitrogen oxide processing and its role in regional air quality. Science 331, 67–70. Chow, J.C., Chen, L.W.A., Watson, J.G., Lowenthal, D.H., Magliano, K.A., Turkiewicz, K., Lehrman, D.E., 2006a. PM2.5 chemical composition and spatiotemporal variability during the California regional PM10/PM2.5 air quality study (CRPAQS). Journal of Geophysical Research – Atmospheres 111 (D10). Chow, J.C., Watson, J.G., Lowenthal, D.H., Chen, L.-W.A., Zielinska, B., Rinehart, L., Magliano, K.L., 2006b. Evaluation of organic markers for chemical mass balance source apportionment at the Fresno supersite. Discussion 6. Atmospheric Chemistry and Physics, 10341–10372. Evans, M.J., Jacob, D.J., 2005. Impact of new laboratory studies of N2O5 hydrolysis on global model budgets of tropospheric nitrogen oxides, ozone, and OH. Geophysical Research Letters 32 (L09813). Held, T., Ying, Q., Kaduwela, A., Kleeman, M., 2004. Modeling particulate matter in the San Joaquin Valley with a source-oriented externally

430

Q. Ying et al. / Atmospheric Environment 43 (2009) 419–430

mixed three-dimensional photochemical grid model. Atmospheric Environment 38 (22), 3689–3711. Herner, J., Ying, Q., Aw, J., Gao, O., Chang, D., Kleeman, M., 2006. Dominant mechanisms that shape the airborne particle size and composition distribution in central California. Aerosol Science and Technology. Herner, J.D., Aw, J., Gao, O., Chang, D.P., Kleeman, M.J., 2005. Size and composition distribution of airborne particulate matter in northern California: I – particulate mass, carbon, and water-soluble ions. Journal of the Air & Waste Management Association 55 (1), 30–51. Jacob, D.J., 2000. Heterogeneous chemistry and tropospheric ozone. Atmospheric Environment 34 (12), 2131–2159. Kleeman, M.J., Cass, G.R., Eldering, A., 1997. Modeling the airborne particle complex as a source-oriented external mixture. Journal of Geophysical Research – Atmospheres 102 (D17), 21355–21372. Kleeman, M.J., Ying, Q., Lu, J., Mysliwiec, M.J., Griffin, R.J., Chen, J.J., Clegg, S., 2007. Source apportionment of secondary organic aerosol during a severe photochemical smog episode. Atmospheric Environment 41 (3), 576–591. Mysliwiec, M.J., Kleeman, M.J., 2002. Source apportionment of secondary airborne particulate matter in a polluted atmosphere. Environmental Science & Technology 36 (24), 5376–5384. Pun, B.K., Seigneur, C., 2001. Sensitivity of particulate matter nitrate formation to precursor emissions in the California San Joaquin Valley. Environmental Science & Technology 35 (14), 2979–2987. Riemer, N., Vogel, H., Vogel, B., Schell, B., Ackermann, I., Kessler, C., Hass, H., 2003. Impact of the heterogeneous hydrolysis of N2O5 on chemistry and nitrate aerosol formation in the lower troposphere

under photosmog conditions. Journal of Geophysical Research – Atmospheres 108 (D4). Russell, A.G., Winner, D.A., Harley, R.A., McCue, K.F., Cass, G.R., 1993. Mathematical-modeling and control of the dry deposition flux of nitrogen-containing air-pollutants. Environmental Science & Technology 27 (13), 2772–2782. Stockwell, W.R., Watson, J.G., Robinson, N.F., Steiner, W., Sylte, W.W., 2000. The ammonium nitrate particle equivalent of NOx emissions for wintertime conditions in central California’s San Joaquin Valley. Atmospheric Environment 34 (27), 4711–4717. Walmsley, J.L., Wesely, M.L., 1996. Modification of coded parametrizations of surface resistances to gaseous dry deposition. Atmospheric Environment 30 (7), 1181–1188. Ying, Q., Kleeman, M.J., 2006. Source contributions to the regional distribution of secondary particulate matter in California. Atmospheric Environment 40 (4), 736–752. Ying, Q., Kleeman, M.J. Regional transport of particulate matter in California with source contribution analysis. Atmospheric Environment, submitted for publication. Ying, Q., Lu, J., Kleeman, M.J. Modeling air quality during the California Regional PM10/PM2.5 Air Quality Study (CRPAQS) using the CIT/UCD source oriented air quality model – Part II. Source apportionment of primary particulate matter. Atmospheric Environment, in press-a. Ying, Q., Lu, J., Livingstone, P., Allen, P., Kleeman, M.J. Modeling air quality during the California Regional PM10/PM2.5 Air Quality Study (CRPAQS) using the CIT/UCD source oriented air quality model – Part I. Base case model results. Atmospheric Environment, in press-b.