- Email: [email protected]
Understanding particulate matter formation in the California San Joaquin Valley: conceptual model and data needs
Atmospheric Environment 33 (1999) 4865}4875
Understanding particulate matter formation in the California San Joaquin Valley: conceptual model and data needs Betty K. Pun*, Christian Seigneur Atmospheric and Environmental Research, Inc., 2682 Bishop Drive, Suite 120, San Ramon, CA 94583, USA Received 7 December 1998; received in revised form 3 May 1999; accepted 4 May 1999
Abstract Quantitative information from the 1995 Integrated Monitoring Study (IMS95) is used to develop a conceptual model, which describes the chemical characteristics and the physical processes responsible for the accumulation of PM in the San Joaquin Valley of California. One signi"cant "nding of the conceptual model is the sensitivity of ammonium nitrate (46% of winter PM ) and nitric acid to oxidants, which may be VOC-sensitive rather than NO -sensitive. Key gaps in V current knowledge are identi"ed using the conceptual model, e.g., the relative sensitivity of winter oxidants to VOC and NO , mechanistic details of secondary organic aerosol formation, mechanisms of dispersion under calm conditions, and V the importance of dry deposition. Some recommendations are also provided for the formulation of air quality models suitable to address the accumulation of PM in the San Joaquin Valley. 1999 Elsevier Science Ltd. All rights reserved. Keywords: Integrated Monitoring Study (1995); PM ; PM ; Conceptual model; Ammonium nitrate
1. Introduction Atmospheric particulate matter (PM) concentrations result from the emission of PM into the atmosphere (primary PM) and the formation of condensable material through atmospheric chemical transformations (secondary PM). Meeting the PM National Ambient Air Quality Standards (NAAQS) requires an understanding of the physical and chemical processes that lead to high PM levels. To that end, experimental "eld programs are conducted in the California San Joaquin Valley (SJV) to study PM episodes. The 1995 Integrated Monitoring Study (IMS95) was the planning study of the California Regional PM Air Quality Study (CRPAQS), a comprehensive data collection and modeling e!ort for central California. Under IMS95, a fall study (November 1995) and a winter study
* Corresponding author. E-mail addresses: [email protected] (B.K. Pun), seigneur@aer. com (C. Seigneur)
(December 1995}January 1996) were undertaken. Descriptions of the "eld program and the study domain are provided by Solomon and Magliano (1999). Limited PM data from central California indicate that the annual 15 lg m\ standard will probably be exceeded in several populated areas, especially in the SJV (Watson et al., 1998). High annual average concentrations are dominated by elevated concentrations during fall and winter. In addition, during IMS95 winter study, the 24-hour standard value for PM (65 lg m\) was exceeded "ve times in Fresno and once in Bakers"eld. The 24-hour standard value for PM (150 lg m\) was exceeded at many sites in the vicinity of Corcoran on November 8 and 14. A conceptual model is a qualitative compilation of the physical and chemical processes that govern the formation of PM, which, to the extent possible, is supported by quantitative information. The conceptual model may be used to guide data collection to characterize important processes and to "ll key knowledge gaps. In addition, model development opportunities can be identi"ed if the important processes of the conceptual model are not
1352-2310/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 2 - 2 3 1 0 ( 9 9 ) 0 0 2 6 6 - 6
4866
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
represented properly in computer models. The objectives of this work are to (1) formulate a conceptual model for PM in the California SJV, (2) identify major knowledge gaps and recommend "eld data and modeling approaches needed to address them.
2. Data available for constructing the conceptual model A comprehensive review of the IMS95 data reports forms the information base used to develop the concep-
tual model. Studies that we have assimilated are listed in Table 1. Elevated PM concentrations in the winter resulted from increased PM concentrations, because PM constituted on average 70}80% of PM concentrations. Hence, PM is the focus of the conceptual model of the winter episodes. Fig. 1a shows the PM mass concen trations observed at the core (community-representative) sites during the winter study. Di!erences in meteorology and emissions resulted in variations of the spatial and temporal patterns of PM during the episodes of De cember 9}10, 25}28, and January 4}6. In the fall, high
Table 1 Studies assimilated in the conceptual model Reference
Major "ndings
Magliano et al. (1999)
Chemical mass balance (CMB) analysis of PM (see Fig. 3). Midday peaks for nitrate and sulfate.
Kumar et al. (1998)
PM and gas concentrations of nitrogen and sulfur species. 75}80% total nitrate in particulate phase. 50}60% total ammonium in particulate phase. NH /HNO '1.05 in 93% of samples. TNO /NO "5.0 lg N m\/37.7 lg N m\ all episodes all sites. V NH abundant in urban and rural areas to convert oxidized NO to nitrate. Median ground level transport V scale 4 km during stagnation period of 3}5 h. Di!usive transport dominated during some stagnant periods.
Carr and Gray (1998) Schauer and Cass (1999)
Fingerprint-based CMB for PM, especially organic compounds. Vegetative burning highest in the December 26}28 episode. Unapportioned compounds interpreted as SOA.
Strader et al. (1999)
Range of SOA: 5}20 lg m\, determined by the OC/EC method and trajectory modeling.
Hoag et al. (1999)
Comparison of the O , H O , O pathways for oxidizing aqueous SO to sulfate. O was the dominant oxidizing agent in 98% of fog samples.
Collett et al. (1999)
Fog droplets are bu!ered at pH between 4 and 7 at urban sites. The bu!ering e!ect was not explained by carbonate, acetate, formate or the phenols analyzed.
Lillis et al. (1999)
Fog simulations. Nitrate and ammonium are scavenged, sulfate may be produced or scavenged. If wet deposition velocity reduced by factor of 2, removal of nitrate reduced by 25%.
Blanchard et al. (1999)
Spatial representativeness (fraction of the local saturation monitoring domain having concentrations $20% of the measurement) is 79% at the Kern Wildlife Refuge core site, 65% (Bakers"eld) and 44% (Fresno) at the urban core sites, and 87% at the Corcoran core site (fall). Secondary components more uniformly distributed than primary components Zones of in#uence of emission sources (approximate distances over which concentration peaks diminish to background levels): winter: 10 km (urban scale) and 20 km (regional scale); fall: 1 km (local scale) and 5}15 km (sub-regional scale).
Collins (1998)
Mean PM urban background of 40 lg m\. Mean PM regional background of 25 lg m\. Clean air, non-anthropogenic background not observed within the IMS95 domain. Fluxes across the northern edge of the domain poorly de"ned due to low and variable winds. Fall mean PM background of 100 lg m\. 3 layers in the SJV atmosphere: (1) surface layer ((1000 m), governed by local topography; (2) upper mixed layer (1000}2000 m), a!ected by mountain ranges and other regional terrain features; and (3) decoupled layer or free troposphere ('2000 m), governed by synoptic meteorology. Heights of layers varied with time and location.
Dye et al. (1997)
Lehrman et al. (1998)
Maximum daily mixing height ranged from 125}1100 m. Surface (0}300 m) wind speeds (2.5 m s\, aloft wind speeds up to 10 m s\.
Chow et al. (1999)
Fall PM : 134 lg m\ at industrial sites, 113 lg m\ at agricultural sites, 112 lg m\ at residential sites. Observed PM-generating activities did not correlate with 24-h data.
Coe and Chinkin (1998)
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
4867
Fig. 1. Temporal Variations of PM during IMS95. (a) PM mass concentrations during the winter study (Source: Magliano et al., 1996). Dotted line indicates the 24-hour average PM NAAQS value of 65 lg m\. (b) PM mass concentrations and 850 mB temperature during the fall study (Source: Solomon et al., 1996). Dotted line indicates the 24-hour average PM NAAQS value of 150 lg m\.
PM mass concentrations are of concern in the SJV. Fig. 1b shows the 24-hour average PM concentrations in the vicinity of Corcoran during the fall study of IMS95.
3. Conceptual model methodology The methodology used here is data-driven and works backwards in time from the observed ambient PM data to the precursors and emission sources. The two components of the methodology are (1) an analysis of the chem-
ical transformations, and (2) an analysis of the physical transport processes. The two analyses are then combined into the conceptual model. Fig. 2 shows the framework for analyzing PM by chemical composition. The categories used in this analysis were identi"ed by chemical mass balance (CMB) (Magliano et al., 1999). Several of these categories are primary emissions, including geological material (dust), primary mobile sources, and vegetative burning. Excess organic compounds (OC) consist of organic compounds that are not apportioned to speci"c sources. They may be
4868
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
Fig. 2. Breakdown of PM by chemical composition from CMB analysis (Source: Magliano et al., 1999). Precursors to PM components are also listed. Bold lines represent the in#uential variables for the formation of secondary PM, to the extent that su$cient information is available to determine which variables are the most in#uential. Gray bold lines indicate higher degrees of uncertainty associated with the determination of in#uential variables.
primary emissions from sources that are not used in the CMB analysis (e.g., meat cooking) or secondary organic aerosols (SOA) formed in the atmosphere. The formation of secondary compounds, nitrates, sulfates, and SOA, is analyzed further to identify the chemical reactions responsible for their production and the limiting reagents of the reactions (shown in bold lines in Fig. 2). The transport processes important on the regional scale include horizontal advection, vertical mixing, wet deposition, and dry deposition. Because there is a secondary component in PM, transport processes for PM precursors (such as HNO , NH , NO , volatile organic compounds (VOC), and oxidants) are also considered. Horizontal advection typically varies signi"cantly between the surface and aloft. Pollutants tend to be transported over long distances aloft, where the wind speeds are typically higher. The e!ect of upper-air transport on local air quality ultimately depends upon vertical mixing. As the mixed layer grows during the day, aloft air is entrained into the surface layer and an air parcel from a distant upwind location may then be mixed with the local air mass near the ground. Both wet and dry deposition can signi"cantly alter the atmospheric budget of some chemical species, particularly under calm conditions when there is little ventilation in the valley.
4. Conceptual model for the winter PM episodes 4.1. Chemical composition and transformations The average winter PM composition is reported in Magliano et al. (1999) (Fig. 3a). Signi"cant variability of
the source contributions was observed. PM at urban sites typically showed higher absolute and relative contributions from primary sources. Higher relative contributions from secondary compounds were observed at the rural sites (Table 2), although the absolute contributions of secondary NH NO were similar at urban and rural sites. 4.1.1. Ammonium nitrate (NH4NO3) NH NO was the most dominant compound in PM during IMS95 winter episodes. The ranges of concentrations observed were 9.6}24 lg m\ at rural sites and 10.6}29.5 lg m\ at urban sites. At the non-urban sites, NH NO can contribute 50% or more to the PM mass. As indicated in Fig. 1, NH NO is formed by the combination of NH and HNO . NH #HNO & NH NO . (1) NH is emitted mainly from dairies/livestock/poultry feedlots and natural soils (Haste et al., 1998), while HNO is typically formed in the atmosphere. Kumar et al. (1998) showed that 75}82% of the available nitrate is titrated by only 50}60% of the available NH , and the ratio of NH to HNO was greater than 1.05 in 93% of the winter samples. Hence, the formation of NH NO within the study area of IMS95 was limited by the availability of HNO in winter; NH sources were relatively abundant. Gas-phase reactions that form HNO are the follow ing. NO #OHPHNO N O #H OP2HNO
(2) (3)
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
4869
hand, Reactions (3)}(5) tend to take place after the sun sets. During the day, the lifetime of NO is short because the photolysis reaction of NO is fast. Consequently, the daytime concentration of N O is expected to be low. The presence of fog during the day may reduce NO photolysis and enhance HNO production via the N O pathway. This e!ect is uncertain. Magliano et al. (1999) noted an increase in the NH NO mass during the day (starting at 9 a.m.). Peak concentrations of total nitrate in the afternoon provided indirect evidence for the NO # OH reaction as a signi"cant HNO production route. HNO formation may be either NO -sensitive or V oxidant-sensitive (OH, NO , and O ). The concentration of total nitrate (HNO and particulate nitrate) was only a small fraction (13%) of that of NO during a typical V episode (Kumar et al., 1998). Furthermore, NO seemed abundant at all core sites (25 ppb in Bakers"eld, 29 ppb in Fresno, 17 ppb in Kern Wildlife Refuge). It is likely that the conversion of NO to HNO is limited by oxidants, particularly OH, since the daytime reaction is important. Carr and Gray (1998) also concluded from their model simulation that NO oxidation is sensitive to V the oxidation potential. Oxidants are produced from precursors, NO and VOC. Therefore, it is important to V understand which precursor is the most in#uential for the formation of oxidants in the system. This question is addressed further below. Fig. 3. Source contributions based on chemical mass balance analysis. (a) Winter PM , (b) Fall PM . (Source: Magliano et al., 1999).
Table 2 Di!erences in PM compositions at urban and rural sites (Magliano et al., 1999)
Average mass on episode days Nitrate Vegetative burning Mobile sources
Urban
Rural
57 lg m\
31 lg m\
19 lg m\(32%) 18 lg m\ (60%) 12 lg m\ (21%) 2.6 lg m\ (9%) 7.6 lg m\ (14%) 3.0 lg m\ (10%)
Reaction (3) also takes place in the aqueous phase, with increased reaction rate. N O is the combination prod uct of NO and nitrate radical (NO ). NO #NO & N O . (4) NO is formed when ozone (O ) oxidizes NO . NO #O PNO #O . (5) Reaction (2) takes place mostly during the day, as night-time OH concentrations are low. On the other
4.1.2. Organic compounds Organic compounds accounted for a substantial fraction (35%) of the PM mass at the urban sites of Bakers"eld and Fresno. Primary organic particles were attributed to mobile vehicles, vegetative burning, and (a fraction of) excess OC. The diurnal patterns of primary particles are driven by their emission patterns. Vegetative burning contributions were consistent with domestic "replace usage. The highest contributions were observed in the evening samples at the urban sites, especially during the holiday season, when they surpassed NH NO as the largest contributor of PM . On the other hand, in rural areas (e.g., Kern Wildlife Refuge), the percentage contribution remained small and essentially constant throughout the day, indicative of a location far from source areas. Motor vehicle particles also exhibited higher contributions at urban sites than at rural sites. The large amounts of excess, unapportioned OC in many CMB runs may result from inadequate organic source pro"les, which cause mis-allocation in the traditional CMB analysis. Missing source pro"les may also cause primary compounds to be apportioned to the &excess' category. The remaining fraction of excess OC corresponds to SOA. At present, no direct measurement technique allows the experimental discrimination of primary vs. secondary OC. Therefore, the fraction of excess OC that is primary cannot be independently evaluated. With detailed source pro"les and "ngerprint compounds,
4870
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
Schauer and Cass (1999) were able to increase the resolution of CMB and apportion a larger fraction of the primary PM. They reported that the unapportioned "ne organic mass averaged about 4.5 lg m\ at both urban and rural sites in the winter. The authors interpreted this unapportioned mass as a uniform regional background of SOA during PM episodes. SOA contributed to a maximum of 16% of PM to Fresno in the January 1996 episode. Similar high contributions were calculated at the Kern Wildlife Refuge, where SOA was the second largest contributor to PM behind NH NO . In smog chamber experiments, anthropogenic VOC (e.g., long chain alkanes, alkenes, and aromatics) and biogenic compounds (e.g., terpenes and oxygenates) form SOA. Strader et al. (1999) modeled SOA production from reactions of VOC with OH, O , and NO . Their results were generally consistent with Schauer and Cass (1999), indicating that SOA should not be ignored in the apportionment of the PM concentrations. This modeling approach could have been useful for the further development of the conceptual model to test the VOC/oxidant limitation of SOA formation, and to determine the VOC classes that form SOA. However, because of uncertainties in the emissions inventory and meteorological information, their "nding that aromatic compounds form much of the SOA in the SJV needs to be corroborated. 4.1.3. Ammonium sulfate Although ammonium sulfate (average 3.1 lg m\ at core sites) constituted a small fraction of the PM measured during IMS95, it was the third largest component at Kern Wildlife Refuge. The concentration of SO in the urban areas averaged about 3 ppb during the winter episodes, and accounted for almost 80% of the total sulfur. Therefore, it was likely that the conversion of SO to sulfate in urban areas was limited by the availability of oxidants. The concentration of SO was signi"cantly lower in the rural areas (0.1 ppb), and "ne and coarse sulfate accounted for 56 and 22% of the total sulfur, respectively. In situ sulfate production was probably limited in rural areas. Therefore, some particulate sulfate was likely formed during transport to rural locations. Sulfuric acid (H SO ) can be produced by the oxida tion of SO by OH in the gas phase, or inside fog drops in reactions involving O , H O , or O (#catalyst). Pre fog concentrations of sulfate (Lillis et al., 1999), especially midday peaks observed at the rural sites (Magliano et al., 1999) seemed to be consistent with day-time gas-phase production of sulfate rather than production in fog, which is more common at night. Nonetheless, much e!ort was invested during IMS95 to understand the aqueous-phase chemistry. There are three major oxidation pathways for dissolved SO : oxidation by O , cata lytic oxidation (dissolved Fe> or Mn> as catalysts) by O , and oxidation by H O . Since the H O oxidation reaction is insensitive to pH and the other two reactions
have kinetic rates that decrease as the pH decreases, di!erent reaction(s) may be responsible for sulfate production during the course of a fog episode due to pH changes. Hoag et al. (1999) concluded that O was the dominant oxidizing agent in 98% of the IMS95 fog samples. Sensitivity studies showed that assumptions about Fe> (which was not measured independently) and O (which sometimes fell below the instrument's detec tion limit) a!ect the determination of the dominant SO reaction at urban sites. Collett et al. (1999) found that at urban sites, fog bu!ering (at pH between 4 and 7) would allow the O reaction to remain competitive throughout a fog episode, despite the production of H SO . Coarse particles have higher pH, which would enhance sulfate production by O , relative to "ne particles (Gurciullo and Pandis, 1997). At this point, the higher concentration of sulfate in PM relative to PM and fog seems to be another indication that the gas-phase reaction and subsequent condensation of H SO may be an important pathway. Further con"rmation is needed. 4.1.4. Oxidant chemistry Oxidant chemistry, which a!ects the production of nitrate, sulfate and SOA, may be sensitive to VOC or NO . The sensitivity regime a!ects the response of the V system to control measures placed on the precursors. Fig. 4 depicts the major chemical pathways of the VOC/NO /O system. When NO is low, HO radicals V V preferentially recombine to form H O (Fig. 4, left-hand side). When VOC is limiting, OH reacts preferentially with NO to form HNO (right-hand side). Conse quently, the ratio of the termination products, H O and HNO (#PM nitrate), is a good indicator of the relative extent of NO and VOC sensitivity in the atmospheric V system (Sillman, 1995; Lu and Chang, 1998). Lu and Chang (1998) determined that summer O in much of the SJV was sensitive to NO , or both VOC and V NO . However, Jacob et al. (1995) reported a seasonal V transition from NO -sensitive conditions during summer V
Fig. 4. Simpli"ed description of photochemical smog formation showing the two major interacting cycles (NO/NO and OH/HO or OH/RO ) and the two major sinks for radical chain termination: H O (NO sensitive) and HNO (VOC sensitive).
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
4871
to VOC-sensitive conditions during winter for O pro duction over the eastern United States. Typical concentrations of HNO (1.5 ppb at urban sites, 2.4 ppb at rural sites) and H O (0.2 ppb at urban sites, 0.3 ppb at rural sites) suggest that the oxidant chemistry of the SJV may be VOC-sensitive during the winter season. This result is preliminary because the indicator approach has not been evaluated for the meteorological conditions prevalent in the SJV winter.
ber PM episode. The concentration of PM typically diminished after a major fog event, although the e!ect of wet deposition may be balanced by aqueous production of sulfate. The predicted wet deposition #uxes (Lillis et al., 1999) were sensitive to model assumptions. For example, when the wet deposition velocity was reduced by a factor of 2, the removal of nitrate was reduced by 25%. Dry deposition was not studied during IMS95.
4.2. Transport processes
5. Conceptual model for the fall PM episodes
Typical transport scales may be inferred by the spatial distribution of PM. The zone of in#uence analysis (Blanchard et al., 1999) showed the urban-scale (10 km) and regional-scale dispersion (20 km) of emission sources. Table 2 shows that both urban-scale distribution of primary sources and regional-scale distribution of secondary compounds in#uence PM in the winter. High er spatial representativeness in the rural networks relative to urban ones (Blanchard et al., 1999) is consistent with the dominance of the secondary compounds. The contribution of primary PM is indicative of limited dispersion from urban centers where emissions occur (via surface advective transport or di!usion). The regional distribution of secondary compounds, e.g., NH NO , may be a result of widespread (area) sources and reactions, di!usive transport, and long-range aloft transport. Stagnation is a key reason contributing to the build up of PM. Surface advective transport is probably quite limited due to low wind speeds (generally (2.5 m s\). In fact, di!usion, with a length scale of 4 km in several hours, was the dominant mode of surface transport during some stagnation periods (Carr and Gray, 1998). Collins (1998) found that #uxes across boundary sites on the edge of the IMS95 monitoring domain were poorly de"ned because of low and variable winds. The low mixing heights during several episodes were likely conducive to pollutant build-up within the valley. During two of the winter episodes, the mixed layer extended beyond the surface layer (1000 m) at some locations in the afternoon. Since the wind speeds aloft were typically higher than those near the surface (Lehrman et al., 1998), transport of pollutants aloft and subsequent mixing into the surface layer may be a mechanism of distributing pollutants within the valley in a multi-day episode. With reduced mixing heights and reduced horizontal surface transport, much of the pollutants were trapped within the SJV. The dominant sink processes were, therefore, wet and dry deposition. Wet deposition was likely to be a signi"cant removal process for the soluble species, nitrate, ammonium, and sulfate. During fog episodes, a fraction of the particles is activated to grow by absorbing water. As fog droplets settle more quickly than smaller particles, PM is removed as fog. Measurements of wet deposition were taken at Fresno during the early Decem-
The chemical mass balance (Fig. 3b), with dominating contribution (57%) from dust and signi"cant contributions from nitrate (16%) and excess OC (10%) (Magliano et al., 1999), indicated the in#uence of both primary and secondary components on PM during the fall episodes (November 6}8, 11}14). The mean PM background in the vicinity of Corcoran was quite high at 100 lg m\. The key di!erences between the winter and the fall episodes are as follows: E PM concentrations were higher in the fall (max imum PM was 290 lg m\) than in the winter (max imum PM was 125 lg m\). E Geological material (dust) dominated the composition of the fall PM (57%), but was a relatively minor component of PM (10%) and PM (0.6%) in the winter. E Secondary NH NO was the dominating component of PM (40%) and PM (46%) in the winter but not in the fall (16%). E PM from vegetative burning contributed signi"cantly (15%) at many sites, especially urban ones (21%), in the winter, but not in the fall. E Ground-level temperature was higher in the fall (203C vs.(153C in the winter). E Fog was patchy and mild in the fall (relative humidity (RH)"50}80%). 5.1. Chemical composition and transformations 5.1.1. Geological material The largest component of PM was geological mater ial, i.e. dust (57%). Since dust is a primary component, it will be discussed further under transport processes. 5.1.2. Ammonium nitrate (NH4NO3) NH NO was the second most important component of PM in the fall, at the Corcoran core site. The nitrate sensitivity to HNO and NH should be con"rmed for fall conditions, which have di!erent reaction rates (due to di!erent temperature and solar radiation), mixing conditions, and emissions than winter. It is likely that the NH NO system is more sensitive to HNO during the rest of the year than during winter, especially since the NO -sensitive oxidant chemistry in the summer V
4872
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
(Lu and Chang, 1998) is not conducive to the formation of HNO . During fall, the atmosphere was relatively dry; therefore, it is postulated that the OH reaction (Reaction (2)) may be more important than the N O hydrolysis route (Reactions (3) and (4)) in the production of HNO . The sensitivities of HNO to oxidants and NO may also V be a function of season, as are the sensitivities of oxidants to precursors VOC and NO (Jacob et al., 1995). V 5.1.3. Excess OC Another PM component needing further analysis is the excess OC. The source of excess OC and the relative contributions of primary and secondary OC are unknown because tracer-based receptor modeling, similar to that of Schauer and Cass (1999), was not performed on the fall samples. 5.2. Transport processes During PM episodes in the fall, the spatial represen tativeness in the Corcoran vicinity tended to increase (Blanchard et al., 1999), indicating the sub-regional (5}15 km length scale), meteorology-driven nature of the PM build-up. As shown in Fig. 2b, PM concentrations correlate well with the 850 mb temperature (about 1500 m or 5000 ft) measured at Oakland, CA. PM concentrations were generally highest during periods of light and variable winds from the southeast (Collins, 1998). Lower PM concentrations were associated with stronger northeasterly winds. As in the winter, secondary components were more uniformly distributed than primary components. However, both primary (e.g., dust) and secondary components (e.g., nitrates) increased during the fall episodes, consistent with a scenario with sub-regional build-up of PM as the culprit of the pollution episodes. The dominance of geological material also supported the sub-regional (rather than regional) build-up hypothesis, as coarse materials are generally not transported over great distances. Spatial variability recorded by the network was attributed to local in#uences. Industrial sites tended to record higher concentrations than the residential sites (Chow et al., 1999). A neighborhood zone of in#uence of about 1 km was observed (Blanchard et al., 1999). However, because of the intermittent nature of the many PM-generating activities, observations of activities did not correlate well with 24-h data (Coe and Chinkin, 1998). Insu$cient meteorological data were available at the monitoring sites to evaluate the contribution of local wind-blown dust. Given low surface wind speeds ((2 m s\), mechanically generated geological material was more likely than wind blown dust. Fig. 1b indicates that a regional reduction in mixing heights was a major cause of elevated PM concentrations. No measurements were taken in the fall to assess the extent of vertical mixing within the IMS95 domain.
Long-range surface transport from upwind pollutant sources to Corcoran was unlikely, due to low surface wind speeds during the fall episodes. Although lower mixing heights were generally associated with PM buildup, the combination of aloft and surface air parcels may still be possible if the mixed layer extended beyond the surface layer in the afternoon. In that case, upper air winds and concentration data, together with the extent of vertical mixing, are necessary to evaluate the impacts of long-range transport of precursors and PM aloft. Wet deposition probably played a minor role in the dissipation of PM during the fall episodes. No rain was reported during the PM monitoring period in November, and only patchy evening fog was reported in Fresno. No information was available from IMS95 to determine the dry deposition #uxes of PM and PM precursors in the SJV in the fall. Depending on the size of the coarse material, dry deposition could play a role in the PM balance at the Corcoran location.
6. Knowledge gaps Knowledge gaps that were identi"ed by the conceptual model and measurements recommended to address them are discussed in this section. Secondary NH NO was an important component of winter PM . It was shown that HNO limited the yield of this secondary com pound. However, the abundance of NH needs to be con"rmed outside of the IMS95 study area and in the spring, summer, and fall seasons. Although the OH#NO reaction is important, it is also important to understand the extent of the N O pathway, especially under foggy conditions, either by experiments or by modeling. Photolysis rates are reduced within the fog layer and increased above due to back scattering of sunlight. In light of the e!ects of the photolysis reactions on the chemistry of NO and N O , measurements of the solar #ux (UV and broad radiation) are recommended to test a model's ability to treat photochemical processes under cloudy/foggy conditions. If warranted, the current empirical treatment in most air quality models can be improved by adding a radiative transfer module. To control HNO formation e!ectively, it is necessary to develop a better understanding of the NO /VOC/oxidant chemV istry for the SJV. In particular, the sensitivities of oxidant production to NO and VOC should be con"rmed V for the winter season. This basic understanding of the atmospheric system is important because any reduction of NO (e.g., for controlling summer O ) may increase the V abundance of HO radicals and H O (see Fig. 4); thus a!ecting the oxidation of SO . To that end, it is desirable to conduct an extensive gas-phase measurement program to characterize speciated VOC, NO, NO , O , H O , HNO , and, if feasible, OH, NO , HO , and RO during
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
winter. Other oxidized nitrogen species, including HNO , particulate nitrate, HNO , PAN, and other or ganic nitrates, may also be measured to improve the understanding of nitrogen chemistry. These data may also provide a suitable database to test current mechanisms, such as SAPRC (Carter, 1990) and CBM-IV (Gery et al., 1989), which have not been thoroughly tested under the wintertime conditions that are conducive to PM formation in the SJV. Numerous knowledge gaps are associated with SOA formation. Emissions inventories of SOA-forming VOC, especially biogenic VOC, should be a focus of the forthcoming "eld campaign. Information is lacking regarding the identity of the condensable compounds for many organic precursors and the reactions leading to their formation. In addition, organic compounds that partition into the aqueous phase should be studied because they may alter the hygroscopic behavior of inorganic particles and droplets (Saxena et al., 1995). It is prudent to examine the mechanisms for sulfate formation to assure that changes in precursor (NO or V VOC) emissions designed to reduce oxidant (or O ) and PM will not adversely a!ect sulfate. Collett et al. (1999) recommended investigations of the urban fog bu!ering capacity. The fog bu!ering phenomenon observed during IMS95 also needs to be represented in models, probably requiring additional species and reactions, even in the most comprehensive aqueous-phase mechanisms. Alternatively, a parametric modeling approach may be used with a simpler mechanism in the interest of computational resources needed to model fog droplets. Modern O instruments with sub-ppb detection limit and sensi tivity (Seigneur et al., 1999 and references therein) may aid in con"rming the importance of the O pathway in fog and allow a thorough evaluation of the aqueous and gaseous pathways. In addition, accurate RH measurements are needed to understand various aqueous-phase processes. RH is also an important factor in light extinction (Richards et al., 1999). Transport was not su$ciently addressed in IMS95, especially meteorological conditions associated with enhanced regional in#uence (Blanchard et al., 1999). To test transport of pollutants aloft and subsequent (afternoon) mixing into the surface layer as a mechanism for distributing pollutants within the valley in a multi-day episode, data on upper level wind, mixing height, and aloft concentrations are needed. Applications of models to study stagnant conditions may require alternative modeling techniques to reduce errors due to numerical di!usion, such as implementing a variable grid size system, activated to reduce the grid size during stagnant conditions to mediate the e!ects of numerical di!usion. The net e!ect of a fog event, wet deposition vs. PM production, which varies with location, fog duration, etc., needs to be better characterized. It may also be useful to understand if any recycling of PM and gaseous pollu-
4873
tants occurs through volatilization at the conclusion of the fog event. Dry deposition was not studied during IMS95, although it may be a signi"cant sink for both PM and precursor gases. For instance, Russell et al. (1985) determined that 75% of the atmospheric HNO could be removed in the California South Coast Air Basin in 24 hours via dry deposition. Direct measurements of dry deposition #uxes by the eddy correlation technique are needed over a variety of land-use types for both PM and precursor gases, including O , HNO , NH , and other nitrogen-containing compounds. Dry deposition measurements of PM and precursor gases may be used to re"ne estimates of areas of in#uence and to infer the mechanism of sub-regional buildup. Wind blown dust was inconsistent with the observed wind speeds of (2 m s\. Sources of mechanically generated aerosols, such as paved and unpaved roads and farming activities, may be addressed by tracer experiments. The primary components of excess OC (about 10% of PM) also need to be better characterized, especially in the fall. More detailed source pro"les (e.g., "ngerprint compounds) will improve the resolution of CMB. For example, generic pro"les allowed only the apportionment of one vegetative burning category (Magliano et al., 1999). Fingerprint compounds (e.g., levoglucosan, propionylsyringol, butyrylsyringol, and resin acids) allowed Schauer and Cass (1999) to distinguish between hardwood and softwood burning. More detailed geological material and organic pro"les are also useful for elucidating the importance of SOA using a CMB analysis.
7. Conclusions We have presented a methodology for the development of conceptual models of air pollution. This methodology was applied to the California SJV to describe the key chemical and physical processes that govern PM formation and accumulation during fall and winter. Highlights of the conceptual models are summarized in Table 3. The conceptual model is not only an assimilation of the current state of knowledge, but also a valuable tool for "eld program design to guide data collection in areas that are key to improving our understanding. A major "nding of this work is that PM NH NO may be more sensitive to VOC than to NO , as previously V believed. A further implication of the conceptual model is that the sensitivities of oxidants to precursors VOC and NO in the winter may be di!erent from those in the V summer season. Therefore, a holistic approach must be taken to design control strategies for winter PM and summer O . The analysis performed in this work also highlighted data gaps in the current understanding of the PM phenomenon and model development and
4874
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
Table 3 Summary of the conceptual model for PM in the San Joaquin Valley, CA Major features of the conceptual model
Winter
Fall
Chemical composition and transformations
NH NO (46% of PM on average)
Geological material (57% of PM on average) NH NO (16% of PM on average)
E NH NO is more sensitive to HNO than NH E HNO formation appears to be limited by oxidants (OH, O , NO ) rather than by NO E Oxidants may be VOC-sensitive rather than NO -sensitive V Organic compounds (34% of PM on average) E Primary OC are mainly from vegetative burning and mobile sources E Secondary OC are from both anthropogenic and biogenic precursors Transport processes
Limited ground-level transport due to calm winds Aloft transport possible, but needs to be assessed Limited vertical mixing because of stable conditions Wet deposition due to fog droplet settling; removal of nitrate and ammonium Extent of dry deposition unknown, needs to be assessed In#uence of urban emissions on regional PM levels (secondary NH NO , SOA)
evaluation activities that should be implemented before PM model predictions are used to develop control strategies of these pollutants, in the SJV as well as at other locations.
Acknowledgements This work was supported by Paci"c Gas and Electric Company (PG&E)/California Energy Commission (CEC), under PG&E Contract No. 4600008160. The authors thank the Project Managers Gene Shelar (PG & E) and Guido Franco (CEC) for their support throughout the study. Sam Altshuler (PG & E), Karen Magliano (California Air Resources Board), and Paul Solomon (Environmental Protection Agency) provided valuable comments.
References Blanchard, C.L., Carr E.L., Collins, J.F., Smith, T.B., Lehrman, D.E., Michaels, H.M., 1999. Spatial representativeness and
Limited ground-level transport due to calm winds Aloft transport possible, but needs to be assessed PM build-up correlates with reduced mixing height Limited wet deposition (mild fog) Extent of dry deposition unknown, needs to be assessed In#uences of local sources (geological material) on sub-regional PM distribution (NH NO )
scales of transport during the 1995 integrated monitoring study in California's San Joaquin Valley. Atmospheric Environment, in press. Carr, E.L., Gray, H.A., 1998. 1995 Integrated monitoring study data analysis work element 4.6.3: time and length scales for mixing of secondary aerosols during stagnation periods. Final Report. California Air Resources Board, Sacramento, CA. Carter, W.P.L., 1990. A detailed mechanism for the gas-phase atmospheric reactions of organic compounds. Atmospheric Environment 24A, 481}518. Chow, J.C., Watson, J.G., Lowenthal, H.D., Egami, R.T., Hackney, R., Magliano, K.L., 1999. Temporal variations of PM , PM , and gaseous precursors during the Integrated Monitoring Study in Central California. Journal of the Air and Waste Management Association, in press. Coe, D.L., Chinkin, L.R., 1998. Emission source activity detection analysis of CMB model results and daily activity data for IMS95. In: Draft Final Report STI-9972121775-DFR. California Air Resources Board, Sacramento, CA. Collett, Jr. J.L., Hoag, K.J., Pandis, S.N., 1999. Internal acid bu!ering in San Joaquin valley fog drops and its in#uence on aerosol processing. Atmospheric Environment, in press.
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875 Collins, J.F., 1998. Characterization of boundaries and backgrounds in the San Joaquin Valley Integrated Monitoring Study 1995. In: Draft Final Report. California Air Resources Board, Sacramento, CA. Dye, T.S., Kwiatkowski, J.J., MacDonald, C.P., Ray, S.E., Chinkin, L.R., Lindsey, C.G., 1997. Measurement methods validation: adequacy and validation of meteorological measurements aloft during IMS95. In: Draft Final Report. STI997218-1766-DFR. California Air Resources Board, Sacramento, CA. Gery, M.W., Whitten, G.Z., Killus, J.P., Dodge, M.C., 1989. A photochemical kinetics mechanism for urban and regional scale computer modeling. Journal of Geophysical Research 94, 12925}12926. Gurciullo, C.S., Pandis, S.N., 1997. E!ects of composition in cloud droplet populations on aqueous-phase chemistry. Journal of Geophysical Research 102, 9375}9385. Haste, T.L., Chinkin, L.R., Kumar, N., Lurmann, F.W., Hurwitt, S.B., 1998. Use of ambient data collected during IMS95 to evaluate a regional emission inventory for the San Joaquin Valley. Draft Final Report, STI-997211-1800-DFR. California Air Resources Board, Sacramento, CA. Hoag, K.J., Collett, Jr. J.L., Pandis, S.N., 1999. The in#uence of drop size-dependent fog chemistry on aerosol processing by San Joaquin Valley fogs. Atmospheric Environment 33, 4817}4832. Jacob, D.J., Horowitz, L.W., Munger, J.W., Heikes, B.G., Dickerson, R.R., Artz, R.S., Keene, W.C., 1995. Seasonal transition from NO -to hydrocarbon-limited conditions for V ozone production over the eastern United States. Journal of Geophysical Research 100, 9315}9324. Kumar, K., Lurmann, F.W., Pandis, S., 1998. Analysis of atmospheric chemistry during 1995 Integrated Monitoring Study. In: Draft Final Report STI-997214-1791-DFR. California Air Resources Board, Sacramento, CA. Lehrman, D.E., Smith, T.B., Knuth, W.R., 1998. Final Report. Work Element 2.2.2 } Meteorological Representativeness. In: Work Element 2.2.3 } Fog and Low Cloud Characteristics. California Air Resources Board, Sacramento, CA. Lillis, D., Cruz, C.N., Collet, J.L., Jr. Richards, L.W., Pandis, S.N., 1999. Production and removal of aerosol in a polluted fog layer. Model evaluation and fog e!ects on PM. Atmospheric Environment 33, 4797}44816. Lu, C.H., Chang, J.S., 1998. On the indicator-based approach to assess ozone sensitivities and emissions features. Journal of Geophysical Research 103, 3453}3462. Magliano, K.L., Chow, J.C., Watson, J.G., Gillies, J., Egami, D., Roberts, P., Wright, D., Crow D., Newman, J., Berkowitz, C., Solomon, P.A., 1996. 1995 Integrated Monitoring Study: temporal variability of particulate matter } preliminary results. Proceedings of the 1996 U.S. EPA/A WMA International Symposium on Measurement of Toxic and Related Air Pollutants. Air and Waste Management Association, Pittsburgh, PA., pp. 451}464.
4875
Magliano, K.L, Hughes, V.M., Chinkin, L.R., Coe, D., Haste, T., Kumar, N., Lurmann, F., Blanchard, C.L., 1999. Spatial and temporal variations in PM10 and PM2.5 source contributions and comparison to emissions during the 1995 Integrated Monitoring Study. Atmospheric Environment 33, 4757}4773. Richards, L.W., McDade, C.E., Couture, T., Lowenthal, D., 1999. Optical properties of the San Joaquin Valley aerosol collected during the 1995 Integrated Monitoring Study. Atmospheric Environment 33, 4787}4795. Russell, A.G., McRae, G.J., Cass, G.R., 1985. The dynamics of nitric acid production and the fate of nitrogen oxides. Atmospheric Environment 19, 893}903. Saxena, P., Hildemann, L.M., McMurry, P.H., Seinfeld, J.H., 1995. Organics alter hygroscopic behavior of atmospheric particles. Journal of Geophysical Research 100, 18755}18770. Schauer, J.J., Cass, G.R., 1999. Source apportionment of wintertime gas-phase and particle-phase air pollutants using organic compounds as tracers. Environmental Science and Technology, submitted for publication. Seigneur, C., Pun, B., Pai, P., Louis, J.-F., Solomon, P., Emery, C., Morris, R., Zahniser, M., Worsnop, D., Koutrakis, P., White, W., Tombach, I., 1999. Guidance for the Performance Evaluation of Three-Dimensional Air Quality Modeling Systems for Particulate Matter and Visibility. Document Number CP033-98-1. American Petroleum Institute, Washington, DC. Sillman, S., 1995. The use of NO , H O , and HNO as W indicators for ozone-NO -hydrocarbon sensitivity in urban V locations. Journal of Geophysical Research 100, 14175}14188. Solomon, P.A., Chow, J.C., Watson, J.G., Gillies, J., Egami, D., Magliano, K.L., Roberts, P., Wright, D., Crow, D., Newman, J., Berkowitz, C., Solomon P.A., 1996. 1995 Integrated Monitoring Study: Spatial and Temporal Variability of PM During the Fall Saturation Study. In the Proceedings of the 1996 US EPA/A WMA International Symposium on Measurement of Toxic and Related Air Pollutants. Air and Waste Management Association, Pittsburgh, PA, pp. 434}440. Solomon, P.A., Magliano, K.L., 1999. The 1995 Integrated Monitoring Study (IMS95) of the California Regional PM /PM Air Quality Study (CRPAQS): study overview. Atmospheric Environment 33, 4747}4756. Strader, R., Pandis, S.N., Lurmann, F., 1999. Evaluation of secondary organic aerosol formation in winter. Atmospheric Environment 33, 4849}4863. Watson, J.G., Dubois, D., Roth, P.M., McDade, C., Hackney, R., Niccum, E., Magliano, K., Ranzieri, A., et al., 1998. Field Program Plan for the California Regional PM /PM Air Quality Study. DRI Document C9801.1D3. California Regional PM /PM Air Quality Technical Committee.
Understanding particulate matter formation in the California San Joaquin Valley: conceptual model and data needs Betty K. Pun*, Christian Seigneur Atmospheric and Environmental Research, Inc., 2682 Bishop Drive, Suite 120, San Ramon, CA 94583, USA Received 7 December 1998; received in revised form 3 May 1999; accepted 4 May 1999
Abstract Quantitative information from the 1995 Integrated Monitoring Study (IMS95) is used to develop a conceptual model, which describes the chemical characteristics and the physical processes responsible for the accumulation of PM in the San Joaquin Valley of California. One signi"cant "nding of the conceptual model is the sensitivity of ammonium nitrate (46% of winter PM ) and nitric acid to oxidants, which may be VOC-sensitive rather than NO -sensitive. Key gaps in V current knowledge are identi"ed using the conceptual model, e.g., the relative sensitivity of winter oxidants to VOC and NO , mechanistic details of secondary organic aerosol formation, mechanisms of dispersion under calm conditions, and V the importance of dry deposition. Some recommendations are also provided for the formulation of air quality models suitable to address the accumulation of PM in the San Joaquin Valley. 1999 Elsevier Science Ltd. All rights reserved. Keywords: Integrated Monitoring Study (1995); PM ; PM ; Conceptual model; Ammonium nitrate
1. Introduction Atmospheric particulate matter (PM) concentrations result from the emission of PM into the atmosphere (primary PM) and the formation of condensable material through atmospheric chemical transformations (secondary PM). Meeting the PM National Ambient Air Quality Standards (NAAQS) requires an understanding of the physical and chemical processes that lead to high PM levels. To that end, experimental "eld programs are conducted in the California San Joaquin Valley (SJV) to study PM episodes. The 1995 Integrated Monitoring Study (IMS95) was the planning study of the California Regional PM Air Quality Study (CRPAQS), a comprehensive data collection and modeling e!ort for central California. Under IMS95, a fall study (November 1995) and a winter study
* Corresponding author. E-mail addresses: [email protected] (B.K. Pun), seigneur@aer. com (C. Seigneur)
(December 1995}January 1996) were undertaken. Descriptions of the "eld program and the study domain are provided by Solomon and Magliano (1999). Limited PM data from central California indicate that the annual 15 lg m\ standard will probably be exceeded in several populated areas, especially in the SJV (Watson et al., 1998). High annual average concentrations are dominated by elevated concentrations during fall and winter. In addition, during IMS95 winter study, the 24-hour standard value for PM (65 lg m\) was exceeded "ve times in Fresno and once in Bakers"eld. The 24-hour standard value for PM (150 lg m\) was exceeded at many sites in the vicinity of Corcoran on November 8 and 14. A conceptual model is a qualitative compilation of the physical and chemical processes that govern the formation of PM, which, to the extent possible, is supported by quantitative information. The conceptual model may be used to guide data collection to characterize important processes and to "ll key knowledge gaps. In addition, model development opportunities can be identi"ed if the important processes of the conceptual model are not
1352-2310/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 2 - 2 3 1 0 ( 9 9 ) 0 0 2 6 6 - 6
4866
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
represented properly in computer models. The objectives of this work are to (1) formulate a conceptual model for PM in the California SJV, (2) identify major knowledge gaps and recommend "eld data and modeling approaches needed to address them.
2. Data available for constructing the conceptual model A comprehensive review of the IMS95 data reports forms the information base used to develop the concep-
tual model. Studies that we have assimilated are listed in Table 1. Elevated PM concentrations in the winter resulted from increased PM concentrations, because PM constituted on average 70}80% of PM concentrations. Hence, PM is the focus of the conceptual model of the winter episodes. Fig. 1a shows the PM mass concen trations observed at the core (community-representative) sites during the winter study. Di!erences in meteorology and emissions resulted in variations of the spatial and temporal patterns of PM during the episodes of De cember 9}10, 25}28, and January 4}6. In the fall, high
Table 1 Studies assimilated in the conceptual model Reference
Major "ndings
Magliano et al. (1999)
Chemical mass balance (CMB) analysis of PM (see Fig. 3). Midday peaks for nitrate and sulfate.
Kumar et al. (1998)
PM and gas concentrations of nitrogen and sulfur species. 75}80% total nitrate in particulate phase. 50}60% total ammonium in particulate phase. NH /HNO '1.05 in 93% of samples. TNO /NO "5.0 lg N m\/37.7 lg N m\ all episodes all sites. V NH abundant in urban and rural areas to convert oxidized NO to nitrate. Median ground level transport V scale 4 km during stagnation period of 3}5 h. Di!usive transport dominated during some stagnant periods.
Carr and Gray (1998) Schauer and Cass (1999)
Fingerprint-based CMB for PM, especially organic compounds. Vegetative burning highest in the December 26}28 episode. Unapportioned compounds interpreted as SOA.
Strader et al. (1999)
Range of SOA: 5}20 lg m\, determined by the OC/EC method and trajectory modeling.
Hoag et al. (1999)
Comparison of the O , H O , O pathways for oxidizing aqueous SO to sulfate. O was the dominant oxidizing agent in 98% of fog samples.
Collett et al. (1999)
Fog droplets are bu!ered at pH between 4 and 7 at urban sites. The bu!ering e!ect was not explained by carbonate, acetate, formate or the phenols analyzed.
Lillis et al. (1999)
Fog simulations. Nitrate and ammonium are scavenged, sulfate may be produced or scavenged. If wet deposition velocity reduced by factor of 2, removal of nitrate reduced by 25%.
Blanchard et al. (1999)
Spatial representativeness (fraction of the local saturation monitoring domain having concentrations $20% of the measurement) is 79% at the Kern Wildlife Refuge core site, 65% (Bakers"eld) and 44% (Fresno) at the urban core sites, and 87% at the Corcoran core site (fall). Secondary components more uniformly distributed than primary components Zones of in#uence of emission sources (approximate distances over which concentration peaks diminish to background levels): winter: 10 km (urban scale) and 20 km (regional scale); fall: 1 km (local scale) and 5}15 km (sub-regional scale).
Collins (1998)
Mean PM urban background of 40 lg m\. Mean PM regional background of 25 lg m\. Clean air, non-anthropogenic background not observed within the IMS95 domain. Fluxes across the northern edge of the domain poorly de"ned due to low and variable winds. Fall mean PM background of 100 lg m\. 3 layers in the SJV atmosphere: (1) surface layer ((1000 m), governed by local topography; (2) upper mixed layer (1000}2000 m), a!ected by mountain ranges and other regional terrain features; and (3) decoupled layer or free troposphere ('2000 m), governed by synoptic meteorology. Heights of layers varied with time and location.
Dye et al. (1997)
Lehrman et al. (1998)
Maximum daily mixing height ranged from 125}1100 m. Surface (0}300 m) wind speeds (2.5 m s\, aloft wind speeds up to 10 m s\.
Chow et al. (1999)
Fall PM : 134 lg m\ at industrial sites, 113 lg m\ at agricultural sites, 112 lg m\ at residential sites. Observed PM-generating activities did not correlate with 24-h data.
Coe and Chinkin (1998)
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
4867
Fig. 1. Temporal Variations of PM during IMS95. (a) PM mass concentrations during the winter study (Source: Magliano et al., 1996). Dotted line indicates the 24-hour average PM NAAQS value of 65 lg m\. (b) PM mass concentrations and 850 mB temperature during the fall study (Source: Solomon et al., 1996). Dotted line indicates the 24-hour average PM NAAQS value of 150 lg m\.
PM mass concentrations are of concern in the SJV. Fig. 1b shows the 24-hour average PM concentrations in the vicinity of Corcoran during the fall study of IMS95.
3. Conceptual model methodology The methodology used here is data-driven and works backwards in time from the observed ambient PM data to the precursors and emission sources. The two components of the methodology are (1) an analysis of the chem-
ical transformations, and (2) an analysis of the physical transport processes. The two analyses are then combined into the conceptual model. Fig. 2 shows the framework for analyzing PM by chemical composition. The categories used in this analysis were identi"ed by chemical mass balance (CMB) (Magliano et al., 1999). Several of these categories are primary emissions, including geological material (dust), primary mobile sources, and vegetative burning. Excess organic compounds (OC) consist of organic compounds that are not apportioned to speci"c sources. They may be
4868
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
Fig. 2. Breakdown of PM by chemical composition from CMB analysis (Source: Magliano et al., 1999). Precursors to PM components are also listed. Bold lines represent the in#uential variables for the formation of secondary PM, to the extent that su$cient information is available to determine which variables are the most in#uential. Gray bold lines indicate higher degrees of uncertainty associated with the determination of in#uential variables.
primary emissions from sources that are not used in the CMB analysis (e.g., meat cooking) or secondary organic aerosols (SOA) formed in the atmosphere. The formation of secondary compounds, nitrates, sulfates, and SOA, is analyzed further to identify the chemical reactions responsible for their production and the limiting reagents of the reactions (shown in bold lines in Fig. 2). The transport processes important on the regional scale include horizontal advection, vertical mixing, wet deposition, and dry deposition. Because there is a secondary component in PM, transport processes for PM precursors (such as HNO , NH , NO , volatile organic compounds (VOC), and oxidants) are also considered. Horizontal advection typically varies signi"cantly between the surface and aloft. Pollutants tend to be transported over long distances aloft, where the wind speeds are typically higher. The e!ect of upper-air transport on local air quality ultimately depends upon vertical mixing. As the mixed layer grows during the day, aloft air is entrained into the surface layer and an air parcel from a distant upwind location may then be mixed with the local air mass near the ground. Both wet and dry deposition can signi"cantly alter the atmospheric budget of some chemical species, particularly under calm conditions when there is little ventilation in the valley.
4. Conceptual model for the winter PM episodes 4.1. Chemical composition and transformations The average winter PM composition is reported in Magliano et al. (1999) (Fig. 3a). Signi"cant variability of
the source contributions was observed. PM at urban sites typically showed higher absolute and relative contributions from primary sources. Higher relative contributions from secondary compounds were observed at the rural sites (Table 2), although the absolute contributions of secondary NH NO were similar at urban and rural sites. 4.1.1. Ammonium nitrate (NH4NO3) NH NO was the most dominant compound in PM during IMS95 winter episodes. The ranges of concentrations observed were 9.6}24 lg m\ at rural sites and 10.6}29.5 lg m\ at urban sites. At the non-urban sites, NH NO can contribute 50% or more to the PM mass. As indicated in Fig. 1, NH NO is formed by the combination of NH and HNO . NH #HNO & NH NO . (1) NH is emitted mainly from dairies/livestock/poultry feedlots and natural soils (Haste et al., 1998), while HNO is typically formed in the atmosphere. Kumar et al. (1998) showed that 75}82% of the available nitrate is titrated by only 50}60% of the available NH , and the ratio of NH to HNO was greater than 1.05 in 93% of the winter samples. Hence, the formation of NH NO within the study area of IMS95 was limited by the availability of HNO in winter; NH sources were relatively abundant. Gas-phase reactions that form HNO are the follow ing. NO #OHPHNO N O #H OP2HNO
(2) (3)
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
4869
hand, Reactions (3)}(5) tend to take place after the sun sets. During the day, the lifetime of NO is short because the photolysis reaction of NO is fast. Consequently, the daytime concentration of N O is expected to be low. The presence of fog during the day may reduce NO photolysis and enhance HNO production via the N O pathway. This e!ect is uncertain. Magliano et al. (1999) noted an increase in the NH NO mass during the day (starting at 9 a.m.). Peak concentrations of total nitrate in the afternoon provided indirect evidence for the NO # OH reaction as a signi"cant HNO production route. HNO formation may be either NO -sensitive or V oxidant-sensitive (OH, NO , and O ). The concentration of total nitrate (HNO and particulate nitrate) was only a small fraction (13%) of that of NO during a typical V episode (Kumar et al., 1998). Furthermore, NO seemed abundant at all core sites (25 ppb in Bakers"eld, 29 ppb in Fresno, 17 ppb in Kern Wildlife Refuge). It is likely that the conversion of NO to HNO is limited by oxidants, particularly OH, since the daytime reaction is important. Carr and Gray (1998) also concluded from their model simulation that NO oxidation is sensitive to V the oxidation potential. Oxidants are produced from precursors, NO and VOC. Therefore, it is important to V understand which precursor is the most in#uential for the formation of oxidants in the system. This question is addressed further below. Fig. 3. Source contributions based on chemical mass balance analysis. (a) Winter PM , (b) Fall PM . (Source: Magliano et al., 1999).
Table 2 Di!erences in PM compositions at urban and rural sites (Magliano et al., 1999)
Average mass on episode days Nitrate Vegetative burning Mobile sources
Urban
Rural
57 lg m\
31 lg m\
19 lg m\(32%) 18 lg m\ (60%) 12 lg m\ (21%) 2.6 lg m\ (9%) 7.6 lg m\ (14%) 3.0 lg m\ (10%)
Reaction (3) also takes place in the aqueous phase, with increased reaction rate. N O is the combination prod uct of NO and nitrate radical (NO ). NO #NO & N O . (4) NO is formed when ozone (O ) oxidizes NO . NO #O PNO #O . (5) Reaction (2) takes place mostly during the day, as night-time OH concentrations are low. On the other
4.1.2. Organic compounds Organic compounds accounted for a substantial fraction (35%) of the PM mass at the urban sites of Bakers"eld and Fresno. Primary organic particles were attributed to mobile vehicles, vegetative burning, and (a fraction of) excess OC. The diurnal patterns of primary particles are driven by their emission patterns. Vegetative burning contributions were consistent with domestic "replace usage. The highest contributions were observed in the evening samples at the urban sites, especially during the holiday season, when they surpassed NH NO as the largest contributor of PM . On the other hand, in rural areas (e.g., Kern Wildlife Refuge), the percentage contribution remained small and essentially constant throughout the day, indicative of a location far from source areas. Motor vehicle particles also exhibited higher contributions at urban sites than at rural sites. The large amounts of excess, unapportioned OC in many CMB runs may result from inadequate organic source pro"les, which cause mis-allocation in the traditional CMB analysis. Missing source pro"les may also cause primary compounds to be apportioned to the &excess' category. The remaining fraction of excess OC corresponds to SOA. At present, no direct measurement technique allows the experimental discrimination of primary vs. secondary OC. Therefore, the fraction of excess OC that is primary cannot be independently evaluated. With detailed source pro"les and "ngerprint compounds,
4870
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
Schauer and Cass (1999) were able to increase the resolution of CMB and apportion a larger fraction of the primary PM. They reported that the unapportioned "ne organic mass averaged about 4.5 lg m\ at both urban and rural sites in the winter. The authors interpreted this unapportioned mass as a uniform regional background of SOA during PM episodes. SOA contributed to a maximum of 16% of PM to Fresno in the January 1996 episode. Similar high contributions were calculated at the Kern Wildlife Refuge, where SOA was the second largest contributor to PM behind NH NO . In smog chamber experiments, anthropogenic VOC (e.g., long chain alkanes, alkenes, and aromatics) and biogenic compounds (e.g., terpenes and oxygenates) form SOA. Strader et al. (1999) modeled SOA production from reactions of VOC with OH, O , and NO . Their results were generally consistent with Schauer and Cass (1999), indicating that SOA should not be ignored in the apportionment of the PM concentrations. This modeling approach could have been useful for the further development of the conceptual model to test the VOC/oxidant limitation of SOA formation, and to determine the VOC classes that form SOA. However, because of uncertainties in the emissions inventory and meteorological information, their "nding that aromatic compounds form much of the SOA in the SJV needs to be corroborated. 4.1.3. Ammonium sulfate Although ammonium sulfate (average 3.1 lg m\ at core sites) constituted a small fraction of the PM measured during IMS95, it was the third largest component at Kern Wildlife Refuge. The concentration of SO in the urban areas averaged about 3 ppb during the winter episodes, and accounted for almost 80% of the total sulfur. Therefore, it was likely that the conversion of SO to sulfate in urban areas was limited by the availability of oxidants. The concentration of SO was signi"cantly lower in the rural areas (0.1 ppb), and "ne and coarse sulfate accounted for 56 and 22% of the total sulfur, respectively. In situ sulfate production was probably limited in rural areas. Therefore, some particulate sulfate was likely formed during transport to rural locations. Sulfuric acid (H SO ) can be produced by the oxida tion of SO by OH in the gas phase, or inside fog drops in reactions involving O , H O , or O (#catalyst). Pre fog concentrations of sulfate (Lillis et al., 1999), especially midday peaks observed at the rural sites (Magliano et al., 1999) seemed to be consistent with day-time gas-phase production of sulfate rather than production in fog, which is more common at night. Nonetheless, much e!ort was invested during IMS95 to understand the aqueous-phase chemistry. There are three major oxidation pathways for dissolved SO : oxidation by O , cata lytic oxidation (dissolved Fe> or Mn> as catalysts) by O , and oxidation by H O . Since the H O oxidation reaction is insensitive to pH and the other two reactions
have kinetic rates that decrease as the pH decreases, di!erent reaction(s) may be responsible for sulfate production during the course of a fog episode due to pH changes. Hoag et al. (1999) concluded that O was the dominant oxidizing agent in 98% of the IMS95 fog samples. Sensitivity studies showed that assumptions about Fe> (which was not measured independently) and O (which sometimes fell below the instrument's detec tion limit) a!ect the determination of the dominant SO reaction at urban sites. Collett et al. (1999) found that at urban sites, fog bu!ering (at pH between 4 and 7) would allow the O reaction to remain competitive throughout a fog episode, despite the production of H SO . Coarse particles have higher pH, which would enhance sulfate production by O , relative to "ne particles (Gurciullo and Pandis, 1997). At this point, the higher concentration of sulfate in PM relative to PM and fog seems to be another indication that the gas-phase reaction and subsequent condensation of H SO may be an important pathway. Further con"rmation is needed. 4.1.4. Oxidant chemistry Oxidant chemistry, which a!ects the production of nitrate, sulfate and SOA, may be sensitive to VOC or NO . The sensitivity regime a!ects the response of the V system to control measures placed on the precursors. Fig. 4 depicts the major chemical pathways of the VOC/NO /O system. When NO is low, HO radicals V V preferentially recombine to form H O (Fig. 4, left-hand side). When VOC is limiting, OH reacts preferentially with NO to form HNO (right-hand side). Conse quently, the ratio of the termination products, H O and HNO (#PM nitrate), is a good indicator of the relative extent of NO and VOC sensitivity in the atmospheric V system (Sillman, 1995; Lu and Chang, 1998). Lu and Chang (1998) determined that summer O in much of the SJV was sensitive to NO , or both VOC and V NO . However, Jacob et al. (1995) reported a seasonal V transition from NO -sensitive conditions during summer V
Fig. 4. Simpli"ed description of photochemical smog formation showing the two major interacting cycles (NO/NO and OH/HO or OH/RO ) and the two major sinks for radical chain termination: H O (NO sensitive) and HNO (VOC sensitive).
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
4871
to VOC-sensitive conditions during winter for O pro duction over the eastern United States. Typical concentrations of HNO (1.5 ppb at urban sites, 2.4 ppb at rural sites) and H O (0.2 ppb at urban sites, 0.3 ppb at rural sites) suggest that the oxidant chemistry of the SJV may be VOC-sensitive during the winter season. This result is preliminary because the indicator approach has not been evaluated for the meteorological conditions prevalent in the SJV winter.
ber PM episode. The concentration of PM typically diminished after a major fog event, although the e!ect of wet deposition may be balanced by aqueous production of sulfate. The predicted wet deposition #uxes (Lillis et al., 1999) were sensitive to model assumptions. For example, when the wet deposition velocity was reduced by a factor of 2, the removal of nitrate was reduced by 25%. Dry deposition was not studied during IMS95.
4.2. Transport processes
5. Conceptual model for the fall PM episodes
Typical transport scales may be inferred by the spatial distribution of PM. The zone of in#uence analysis (Blanchard et al., 1999) showed the urban-scale (10 km) and regional-scale dispersion (20 km) of emission sources. Table 2 shows that both urban-scale distribution of primary sources and regional-scale distribution of secondary compounds in#uence PM in the winter. High er spatial representativeness in the rural networks relative to urban ones (Blanchard et al., 1999) is consistent with the dominance of the secondary compounds. The contribution of primary PM is indicative of limited dispersion from urban centers where emissions occur (via surface advective transport or di!usion). The regional distribution of secondary compounds, e.g., NH NO , may be a result of widespread (area) sources and reactions, di!usive transport, and long-range aloft transport. Stagnation is a key reason contributing to the build up of PM. Surface advective transport is probably quite limited due to low wind speeds (generally (2.5 m s\). In fact, di!usion, with a length scale of 4 km in several hours, was the dominant mode of surface transport during some stagnation periods (Carr and Gray, 1998). Collins (1998) found that #uxes across boundary sites on the edge of the IMS95 monitoring domain were poorly de"ned because of low and variable winds. The low mixing heights during several episodes were likely conducive to pollutant build-up within the valley. During two of the winter episodes, the mixed layer extended beyond the surface layer (1000 m) at some locations in the afternoon. Since the wind speeds aloft were typically higher than those near the surface (Lehrman et al., 1998), transport of pollutants aloft and subsequent mixing into the surface layer may be a mechanism of distributing pollutants within the valley in a multi-day episode. With reduced mixing heights and reduced horizontal surface transport, much of the pollutants were trapped within the SJV. The dominant sink processes were, therefore, wet and dry deposition. Wet deposition was likely to be a signi"cant removal process for the soluble species, nitrate, ammonium, and sulfate. During fog episodes, a fraction of the particles is activated to grow by absorbing water. As fog droplets settle more quickly than smaller particles, PM is removed as fog. Measurements of wet deposition were taken at Fresno during the early Decem-
The chemical mass balance (Fig. 3b), with dominating contribution (57%) from dust and signi"cant contributions from nitrate (16%) and excess OC (10%) (Magliano et al., 1999), indicated the in#uence of both primary and secondary components on PM during the fall episodes (November 6}8, 11}14). The mean PM background in the vicinity of Corcoran was quite high at 100 lg m\. The key di!erences between the winter and the fall episodes are as follows: E PM concentrations were higher in the fall (max imum PM was 290 lg m\) than in the winter (max imum PM was 125 lg m\). E Geological material (dust) dominated the composition of the fall PM (57%), but was a relatively minor component of PM (10%) and PM (0.6%) in the winter. E Secondary NH NO was the dominating component of PM (40%) and PM (46%) in the winter but not in the fall (16%). E PM from vegetative burning contributed signi"cantly (15%) at many sites, especially urban ones (21%), in the winter, but not in the fall. E Ground-level temperature was higher in the fall (203C vs.(153C in the winter). E Fog was patchy and mild in the fall (relative humidity (RH)"50}80%). 5.1. Chemical composition and transformations 5.1.1. Geological material The largest component of PM was geological mater ial, i.e. dust (57%). Since dust is a primary component, it will be discussed further under transport processes. 5.1.2. Ammonium nitrate (NH4NO3) NH NO was the second most important component of PM in the fall, at the Corcoran core site. The nitrate sensitivity to HNO and NH should be con"rmed for fall conditions, which have di!erent reaction rates (due to di!erent temperature and solar radiation), mixing conditions, and emissions than winter. It is likely that the NH NO system is more sensitive to HNO during the rest of the year than during winter, especially since the NO -sensitive oxidant chemistry in the summer V
4872
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
(Lu and Chang, 1998) is not conducive to the formation of HNO . During fall, the atmosphere was relatively dry; therefore, it is postulated that the OH reaction (Reaction (2)) may be more important than the N O hydrolysis route (Reactions (3) and (4)) in the production of HNO . The sensitivities of HNO to oxidants and NO may also V be a function of season, as are the sensitivities of oxidants to precursors VOC and NO (Jacob et al., 1995). V 5.1.3. Excess OC Another PM component needing further analysis is the excess OC. The source of excess OC and the relative contributions of primary and secondary OC are unknown because tracer-based receptor modeling, similar to that of Schauer and Cass (1999), was not performed on the fall samples. 5.2. Transport processes During PM episodes in the fall, the spatial represen tativeness in the Corcoran vicinity tended to increase (Blanchard et al., 1999), indicating the sub-regional (5}15 km length scale), meteorology-driven nature of the PM build-up. As shown in Fig. 2b, PM concentrations correlate well with the 850 mb temperature (about 1500 m or 5000 ft) measured at Oakland, CA. PM concentrations were generally highest during periods of light and variable winds from the southeast (Collins, 1998). Lower PM concentrations were associated with stronger northeasterly winds. As in the winter, secondary components were more uniformly distributed than primary components. However, both primary (e.g., dust) and secondary components (e.g., nitrates) increased during the fall episodes, consistent with a scenario with sub-regional build-up of PM as the culprit of the pollution episodes. The dominance of geological material also supported the sub-regional (rather than regional) build-up hypothesis, as coarse materials are generally not transported over great distances. Spatial variability recorded by the network was attributed to local in#uences. Industrial sites tended to record higher concentrations than the residential sites (Chow et al., 1999). A neighborhood zone of in#uence of about 1 km was observed (Blanchard et al., 1999). However, because of the intermittent nature of the many PM-generating activities, observations of activities did not correlate well with 24-h data (Coe and Chinkin, 1998). Insu$cient meteorological data were available at the monitoring sites to evaluate the contribution of local wind-blown dust. Given low surface wind speeds ((2 m s\), mechanically generated geological material was more likely than wind blown dust. Fig. 1b indicates that a regional reduction in mixing heights was a major cause of elevated PM concentrations. No measurements were taken in the fall to assess the extent of vertical mixing within the IMS95 domain.
Long-range surface transport from upwind pollutant sources to Corcoran was unlikely, due to low surface wind speeds during the fall episodes. Although lower mixing heights were generally associated with PM buildup, the combination of aloft and surface air parcels may still be possible if the mixed layer extended beyond the surface layer in the afternoon. In that case, upper air winds and concentration data, together with the extent of vertical mixing, are necessary to evaluate the impacts of long-range transport of precursors and PM aloft. Wet deposition probably played a minor role in the dissipation of PM during the fall episodes. No rain was reported during the PM monitoring period in November, and only patchy evening fog was reported in Fresno. No information was available from IMS95 to determine the dry deposition #uxes of PM and PM precursors in the SJV in the fall. Depending on the size of the coarse material, dry deposition could play a role in the PM balance at the Corcoran location.
6. Knowledge gaps Knowledge gaps that were identi"ed by the conceptual model and measurements recommended to address them are discussed in this section. Secondary NH NO was an important component of winter PM . It was shown that HNO limited the yield of this secondary com pound. However, the abundance of NH needs to be con"rmed outside of the IMS95 study area and in the spring, summer, and fall seasons. Although the OH#NO reaction is important, it is also important to understand the extent of the N O pathway, especially under foggy conditions, either by experiments or by modeling. Photolysis rates are reduced within the fog layer and increased above due to back scattering of sunlight. In light of the e!ects of the photolysis reactions on the chemistry of NO and N O , measurements of the solar #ux (UV and broad radiation) are recommended to test a model's ability to treat photochemical processes under cloudy/foggy conditions. If warranted, the current empirical treatment in most air quality models can be improved by adding a radiative transfer module. To control HNO formation e!ectively, it is necessary to develop a better understanding of the NO /VOC/oxidant chemV istry for the SJV. In particular, the sensitivities of oxidant production to NO and VOC should be con"rmed V for the winter season. This basic understanding of the atmospheric system is important because any reduction of NO (e.g., for controlling summer O ) may increase the V abundance of HO radicals and H O (see Fig. 4); thus a!ecting the oxidation of SO . To that end, it is desirable to conduct an extensive gas-phase measurement program to characterize speciated VOC, NO, NO , O , H O , HNO , and, if feasible, OH, NO , HO , and RO during
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
winter. Other oxidized nitrogen species, including HNO , particulate nitrate, HNO , PAN, and other or ganic nitrates, may also be measured to improve the understanding of nitrogen chemistry. These data may also provide a suitable database to test current mechanisms, such as SAPRC (Carter, 1990) and CBM-IV (Gery et al., 1989), which have not been thoroughly tested under the wintertime conditions that are conducive to PM formation in the SJV. Numerous knowledge gaps are associated with SOA formation. Emissions inventories of SOA-forming VOC, especially biogenic VOC, should be a focus of the forthcoming "eld campaign. Information is lacking regarding the identity of the condensable compounds for many organic precursors and the reactions leading to their formation. In addition, organic compounds that partition into the aqueous phase should be studied because they may alter the hygroscopic behavior of inorganic particles and droplets (Saxena et al., 1995). It is prudent to examine the mechanisms for sulfate formation to assure that changes in precursor (NO or V VOC) emissions designed to reduce oxidant (or O ) and PM will not adversely a!ect sulfate. Collett et al. (1999) recommended investigations of the urban fog bu!ering capacity. The fog bu!ering phenomenon observed during IMS95 also needs to be represented in models, probably requiring additional species and reactions, even in the most comprehensive aqueous-phase mechanisms. Alternatively, a parametric modeling approach may be used with a simpler mechanism in the interest of computational resources needed to model fog droplets. Modern O instruments with sub-ppb detection limit and sensi tivity (Seigneur et al., 1999 and references therein) may aid in con"rming the importance of the O pathway in fog and allow a thorough evaluation of the aqueous and gaseous pathways. In addition, accurate RH measurements are needed to understand various aqueous-phase processes. RH is also an important factor in light extinction (Richards et al., 1999). Transport was not su$ciently addressed in IMS95, especially meteorological conditions associated with enhanced regional in#uence (Blanchard et al., 1999). To test transport of pollutants aloft and subsequent (afternoon) mixing into the surface layer as a mechanism for distributing pollutants within the valley in a multi-day episode, data on upper level wind, mixing height, and aloft concentrations are needed. Applications of models to study stagnant conditions may require alternative modeling techniques to reduce errors due to numerical di!usion, such as implementing a variable grid size system, activated to reduce the grid size during stagnant conditions to mediate the e!ects of numerical di!usion. The net e!ect of a fog event, wet deposition vs. PM production, which varies with location, fog duration, etc., needs to be better characterized. It may also be useful to understand if any recycling of PM and gaseous pollu-
4873
tants occurs through volatilization at the conclusion of the fog event. Dry deposition was not studied during IMS95, although it may be a signi"cant sink for both PM and precursor gases. For instance, Russell et al. (1985) determined that 75% of the atmospheric HNO could be removed in the California South Coast Air Basin in 24 hours via dry deposition. Direct measurements of dry deposition #uxes by the eddy correlation technique are needed over a variety of land-use types for both PM and precursor gases, including O , HNO , NH , and other nitrogen-containing compounds. Dry deposition measurements of PM and precursor gases may be used to re"ne estimates of areas of in#uence and to infer the mechanism of sub-regional buildup. Wind blown dust was inconsistent with the observed wind speeds of (2 m s\. Sources of mechanically generated aerosols, such as paved and unpaved roads and farming activities, may be addressed by tracer experiments. The primary components of excess OC (about 10% of PM) also need to be better characterized, especially in the fall. More detailed source pro"les (e.g., "ngerprint compounds) will improve the resolution of CMB. For example, generic pro"les allowed only the apportionment of one vegetative burning category (Magliano et al., 1999). Fingerprint compounds (e.g., levoglucosan, propionylsyringol, butyrylsyringol, and resin acids) allowed Schauer and Cass (1999) to distinguish between hardwood and softwood burning. More detailed geological material and organic pro"les are also useful for elucidating the importance of SOA using a CMB analysis.
7. Conclusions We have presented a methodology for the development of conceptual models of air pollution. This methodology was applied to the California SJV to describe the key chemical and physical processes that govern PM formation and accumulation during fall and winter. Highlights of the conceptual models are summarized in Table 3. The conceptual model is not only an assimilation of the current state of knowledge, but also a valuable tool for "eld program design to guide data collection in areas that are key to improving our understanding. A major "nding of this work is that PM NH NO may be more sensitive to VOC than to NO , as previously V believed. A further implication of the conceptual model is that the sensitivities of oxidants to precursors VOC and NO in the winter may be di!erent from those in the V summer season. Therefore, a holistic approach must be taken to design control strategies for winter PM and summer O . The analysis performed in this work also highlighted data gaps in the current understanding of the PM phenomenon and model development and
4874
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875
Table 3 Summary of the conceptual model for PM in the San Joaquin Valley, CA Major features of the conceptual model
Winter
Fall
Chemical composition and transformations
NH NO (46% of PM on average)
Geological material (57% of PM on average) NH NO (16% of PM on average)
E NH NO is more sensitive to HNO than NH E HNO formation appears to be limited by oxidants (OH, O , NO ) rather than by NO E Oxidants may be VOC-sensitive rather than NO -sensitive V Organic compounds (34% of PM on average) E Primary OC are mainly from vegetative burning and mobile sources E Secondary OC are from both anthropogenic and biogenic precursors Transport processes
Limited ground-level transport due to calm winds Aloft transport possible, but needs to be assessed Limited vertical mixing because of stable conditions Wet deposition due to fog droplet settling; removal of nitrate and ammonium Extent of dry deposition unknown, needs to be assessed In#uence of urban emissions on regional PM levels (secondary NH NO , SOA)
evaluation activities that should be implemented before PM model predictions are used to develop control strategies of these pollutants, in the SJV as well as at other locations.
Acknowledgements This work was supported by Paci"c Gas and Electric Company (PG&E)/California Energy Commission (CEC), under PG&E Contract No. 4600008160. The authors thank the Project Managers Gene Shelar (PG & E) and Guido Franco (CEC) for their support throughout the study. Sam Altshuler (PG & E), Karen Magliano (California Air Resources Board), and Paul Solomon (Environmental Protection Agency) provided valuable comments.
References Blanchard, C.L., Carr E.L., Collins, J.F., Smith, T.B., Lehrman, D.E., Michaels, H.M., 1999. Spatial representativeness and
Limited ground-level transport due to calm winds Aloft transport possible, but needs to be assessed PM build-up correlates with reduced mixing height Limited wet deposition (mild fog) Extent of dry deposition unknown, needs to be assessed In#uences of local sources (geological material) on sub-regional PM distribution (NH NO )
scales of transport during the 1995 integrated monitoring study in California's San Joaquin Valley. Atmospheric Environment, in press. Carr, E.L., Gray, H.A., 1998. 1995 Integrated monitoring study data analysis work element 4.6.3: time and length scales for mixing of secondary aerosols during stagnation periods. Final Report. California Air Resources Board, Sacramento, CA. Carter, W.P.L., 1990. A detailed mechanism for the gas-phase atmospheric reactions of organic compounds. Atmospheric Environment 24A, 481}518. Chow, J.C., Watson, J.G., Lowenthal, H.D., Egami, R.T., Hackney, R., Magliano, K.L., 1999. Temporal variations of PM , PM , and gaseous precursors during the Integrated Monitoring Study in Central California. Journal of the Air and Waste Management Association, in press. Coe, D.L., Chinkin, L.R., 1998. Emission source activity detection analysis of CMB model results and daily activity data for IMS95. In: Draft Final Report STI-9972121775-DFR. California Air Resources Board, Sacramento, CA. Collett, Jr. J.L., Hoag, K.J., Pandis, S.N., 1999. Internal acid bu!ering in San Joaquin valley fog drops and its in#uence on aerosol processing. Atmospheric Environment, in press.
B.K. Pun, C. Seigneur / Atmospheric Environment 33 (1999) 4865}4875 Collins, J.F., 1998. Characterization of boundaries and backgrounds in the San Joaquin Valley Integrated Monitoring Study 1995. In: Draft Final Report. California Air Resources Board, Sacramento, CA. Dye, T.S., Kwiatkowski, J.J., MacDonald, C.P., Ray, S.E., Chinkin, L.R., Lindsey, C.G., 1997. Measurement methods validation: adequacy and validation of meteorological measurements aloft during IMS95. In: Draft Final Report. STI997218-1766-DFR. California Air Resources Board, Sacramento, CA. Gery, M.W., Whitten, G.Z., Killus, J.P., Dodge, M.C., 1989. A photochemical kinetics mechanism for urban and regional scale computer modeling. Journal of Geophysical Research 94, 12925}12926. Gurciullo, C.S., Pandis, S.N., 1997. E!ects of composition in cloud droplet populations on aqueous-phase chemistry. Journal of Geophysical Research 102, 9375}9385. Haste, T.L., Chinkin, L.R., Kumar, N., Lurmann, F.W., Hurwitt, S.B., 1998. Use of ambient data collected during IMS95 to evaluate a regional emission inventory for the San Joaquin Valley. Draft Final Report, STI-997211-1800-DFR. California Air Resources Board, Sacramento, CA. Hoag, K.J., Collett, Jr. J.L., Pandis, S.N., 1999. The in#uence of drop size-dependent fog chemistry on aerosol processing by San Joaquin Valley fogs. Atmospheric Environment 33, 4817}4832. Jacob, D.J., Horowitz, L.W., Munger, J.W., Heikes, B.G., Dickerson, R.R., Artz, R.S., Keene, W.C., 1995. Seasonal transition from NO -to hydrocarbon-limited conditions for V ozone production over the eastern United States. Journal of Geophysical Research 100, 9315}9324. Kumar, K., Lurmann, F.W., Pandis, S., 1998. Analysis of atmospheric chemistry during 1995 Integrated Monitoring Study. In: Draft Final Report STI-997214-1791-DFR. California Air Resources Board, Sacramento, CA. Lehrman, D.E., Smith, T.B., Knuth, W.R., 1998. Final Report. Work Element 2.2.2 } Meteorological Representativeness. In: Work Element 2.2.3 } Fog and Low Cloud Characteristics. California Air Resources Board, Sacramento, CA. Lillis, D., Cruz, C.N., Collet, J.L., Jr. Richards, L.W., Pandis, S.N., 1999. Production and removal of aerosol in a polluted fog layer. Model evaluation and fog e!ects on PM. Atmospheric Environment 33, 4797}44816. Lu, C.H., Chang, J.S., 1998. On the indicator-based approach to assess ozone sensitivities and emissions features. Journal of Geophysical Research 103, 3453}3462. Magliano, K.L., Chow, J.C., Watson, J.G., Gillies, J., Egami, D., Roberts, P., Wright, D., Crow D., Newman, J., Berkowitz, C., Solomon, P.A., 1996. 1995 Integrated Monitoring Study: temporal variability of particulate matter } preliminary results. Proceedings of the 1996 U.S. EPA/A WMA International Symposium on Measurement of Toxic and Related Air Pollutants. Air and Waste Management Association, Pittsburgh, PA., pp. 451}464.
4875
Magliano, K.L, Hughes, V.M., Chinkin, L.R., Coe, D., Haste, T., Kumar, N., Lurmann, F., Blanchard, C.L., 1999. Spatial and temporal variations in PM10 and PM2.5 source contributions and comparison to emissions during the 1995 Integrated Monitoring Study. Atmospheric Environment 33, 4757}4773. Richards, L.W., McDade, C.E., Couture, T., Lowenthal, D., 1999. Optical properties of the San Joaquin Valley aerosol collected during the 1995 Integrated Monitoring Study. Atmospheric Environment 33, 4787}4795. Russell, A.G., McRae, G.J., Cass, G.R., 1985. The dynamics of nitric acid production and the fate of nitrogen oxides. Atmospheric Environment 19, 893}903. Saxena, P., Hildemann, L.M., McMurry, P.H., Seinfeld, J.H., 1995. Organics alter hygroscopic behavior of atmospheric particles. Journal of Geophysical Research 100, 18755}18770. Schauer, J.J., Cass, G.R., 1999. Source apportionment of wintertime gas-phase and particle-phase air pollutants using organic compounds as tracers. Environmental Science and Technology, submitted for publication. Seigneur, C., Pun, B., Pai, P., Louis, J.-F., Solomon, P., Emery, C., Morris, R., Zahniser, M., Worsnop, D., Koutrakis, P., White, W., Tombach, I., 1999. Guidance for the Performance Evaluation of Three-Dimensional Air Quality Modeling Systems for Particulate Matter and Visibility. Document Number CP033-98-1. American Petroleum Institute, Washington, DC. Sillman, S., 1995. The use of NO , H O , and HNO as W indicators for ozone-NO -hydrocarbon sensitivity in urban V locations. Journal of Geophysical Research 100, 14175}14188. Solomon, P.A., Chow, J.C., Watson, J.G., Gillies, J., Egami, D., Magliano, K.L., Roberts, P., Wright, D., Crow, D., Newman, J., Berkowitz, C., Solomon P.A., 1996. 1995 Integrated Monitoring Study: Spatial and Temporal Variability of PM During the Fall Saturation Study. In the Proceedings of the 1996 US EPA/A WMA International Symposium on Measurement of Toxic and Related Air Pollutants. Air and Waste Management Association, Pittsburgh, PA, pp. 434}440. Solomon, P.A., Magliano, K.L., 1999. The 1995 Integrated Monitoring Study (IMS95) of the California Regional PM /PM Air Quality Study (CRPAQS): study overview. Atmospheric Environment 33, 4747}4756. Strader, R., Pandis, S.N., Lurmann, F., 1999. Evaluation of secondary organic aerosol formation in winter. Atmospheric Environment 33, 4849}4863. Watson, J.G., Dubois, D., Roth, P.M., McDade, C., Hackney, R., Niccum, E., Magliano, K., Ranzieri, A., et al., 1998. Field Program Plan for the California Regional PM /PM Air Quality Study. DRI Document C9801.1D3. California Regional PM /PM Air Quality Technical Committee.