PM2.5 chemical composition and spatiotemporal variability during the California Regional PM10/PM2.5 Air Quality Study (CRPAQS)

Authors


Abstract

[1] The 14-month-long (December 1999 to February 2001) Central California Regional PM10/PM2.5 Air Quality Study (CRPAQS) consisted of acquiring speciated PM2.5 measurements at 38 sites representing urban, rural, and boundary environments in the San Joaquin Valley air basin. The study's goal was to understand the development of widespread pollution episodes by examining the spatial variability of PM2.5, ammonium nitrate (NH4NO3), and carbonaceous material on annual, seasonal, and episodic timescales. It was found that PM2.5 and NH4NO3 concentrations decrease rapidly as altitude increases, confirming that topography influences the ventilation and transport of pollutants. High PM2.5 levels from November 2000 to January 2001 contributed to 50–75% of annual average concentrations. Contributions from organic matter differed substantially between urban and rural areas. Winter meteorology and intensive residential wood combustion are likely key factors for the winter-nonwinter and urban-rural contrasts that were observed. Short-duration measurements during the intensive operating periods confirm the role of upper air currents on valley-wide transport of NH4NO3. Zones of representation for PM2.5 varied from 5 to 10 km for the urban Fresno and Bakersfield sites, and increased to 15–20 km for the boundary and rural sites. Secondary NH4NO3 occurred region-wide during winter, spreading over a much wider geographical zone than carbonaceous aerosol.

1. Introduction

[2] The California Regional PM10/PM2.5 Air Quality Study (CRPAQS) was undertaken with an overall goal of understanding the causes of excessive PM (particulate matter) levels and to evaluate means to reduce them in central California and its major geographical feature, the San Joaquin Valley (SJV) [Watson et al., 1998]. The SJV represents one of the largest PM2.5 and PM10 nonattainment areas in the United States (PM2.5 and PM10 are particles with aerodynamic diameters less than 2.5 and 10 micrometers [μm], respectively). It was expected that considerable variability in emissions, meteorology, and terrain in the SJV would translate into substantial differences in PM concentration and composition across the region. Knowledge of these spatiotemporal distributions of PM and its chemical constituents is essential for understanding source-receptor relationships and chemical, physical, and meteorological processes that cause elevated PM levels in the SJV.

[3] The SJV air basin is bordered on the west by the coastal mountain ranges and on the east by the Sierra Nevada range. These ranges converge at the Tehachapi Mountains at the southern end of the basin, ∼200 km south of Fresno (the largest population center within ∼150 km along a north-south line of the basin). Weather changes seasonally. Spring often brings weak, fast moving frontal passages characterized by low moisture content and high wind speeds. Summer meteorology is driven by heating, which creates a thermal low-pressure system and a large onshore pressure gradient between the coast and the desert. Fall and winter are influenced by the Great Basin High, with prolonged periods of air mass stagnation and limited vertical mixing. Morning mixing depths are shallow and ventilation rates are low during all seasons. Wind speeds are low throughout the day during winter in the absence of storm systems. Relative humidity (RH) is highest in winter and lowest in summer and fall.

[4] Central California emission source categories include (1) small- to medium-sized point sources (e.g., power stations, natural gas boilers, steam generators, incinerators, and cement plants); (2) area sources (e.g., resuspended dust, petroleum extraction operations, cooking, wildfires, and residential wood combustion [RWC]); (3) mobile sources (e.g., cars, trucks, off-road heavy equipment, trains, and aircraft); (4) agricultural and ranching activities (e.g., tilling, fertilizers, herbicides, and livestock); and (5) biogenic sources (e.g., nitrogen oxides [NOx] from biological activity in soils and hydrocarbon emissions from plants). Agriculture is the main industry in the valley, where the major crops are cotton, alfalfa, corn, safflower, grapes, and tomatoes. Cattle feedlots and dairies constitute most of the animal husbandry in the region, along with chicken and turkey farms, which are major sources of ammonia (NH3) emissions.

[5] Past studies [Chow et al., 1992, 1993a, 1996, 1998] have shown that elevated PM concentrations frequently occur in winter, when PM10 concentrations are primarily in the PM2.5 size fraction. Chemical mass balance receptor models [Magliano et al., 1999; Schauer and Cass, 2000] have attributed winter PM episodes in urban areas to RWC emissions, motor vehicle exhaust, and secondary ammonium nitrate (NH4NO3). NH4NO3 generally accounted for 30–60% of PM2.5 during winter [Magliano et al., 1998a, 1998b, 1999; Chow et al., 1999]. Vehicular exhaust and RWC emissions are mostly in the PM2.5 fraction with abundant organic carbon (OC) and elemental carbon (EC).

[6] Watson and Chow [2002a] developed a conceptual model that describes the interplay of emissions and meteorology leading to transport of pollutants and formation of widespread PM2.5 episodes across the SJV in winter. The model begins with a shallow radiation surface inversion (100–200 m deep) which is decoupled from a valley-wide mixed layer aloft between ∼1700 local time (LT) and ∼1100 LT the next morning. At night, the cities experience a build up of primary pollutants emitted from traffic and RWC. Nitric acid (HNO3) can form in the upper layer during nighttime hours through a series of reactions [Atkinson et al., 1986; Stockwell et al., 2000; Pun and Seigneur, 2001]. Prevented from deposition by the surface inversion, this HNO3 would be made available over rural areas with high NH3 emissions to rapidly create NH4NO3. Limited upper air observations [Lehrman et al., 1998] indicate that winds within the valley-wide layer often reach speeds of 1–6 m s−1 while surface winds are <1 m s−1. This implies that secondary NH4NO3 can be mixed throughout the valley in one to two days. When radiative heating breaks the inversion after ∼1100 LT, turbulent mixing between the upper and surface layers intensifies, causing a net downward flux of NH4NO3, which escalates near the surface. In urban areas, this mixing also dilutes the concentrations of primary pollutants, creating a complex diurnal pattern of PM2.5 [Watson et al., 2002].

[7] This paper (1) statistically summarizes CRPAQS PM2.5 mass and chemical compositions, (2) investigates chemical closure for PM2.5 mass, (3) analyzes the spatiotemporal variability of PM2.5 and its chemical composition, (4) examines episodes of elevated PM2.5 during winter in the context of the conceptual model of Watson and Chow [2002a], and (5) evaluates the zones of representation for PM2.5 sampling sites and their implications for future air quality monitoring and research.

2. Ambient Network

[8] The CRPAQS set up a PM2.5 network consisting of 38 sites (Figure 1) where ambient measurements were acquired for 14 months. This network covered the SJV and surrounding air basins (i.e., San Francisco Bay, Sacramento Valley, Mountain Counties, Great Basin Valleys, and Mojave Desert), and sampled urban, suburban, regional, transport, and rural background environments. The entire network covered a region ∼600 km long by 200 km wide (Figure 1). Sampling took place from 2 December 1999 through 3 February 2001, including an annual program between 1 February 2000 and 31 January 2001. Sampling was also conducted during “Winter Intensive Operating Periods (IOPs),” which were selected on the basis of forecasts of high PM2.5 between 15 December 2000 and 3 February 2001. The annual program included every sixth day 24-hour sampling at three anchor sites (Fresno Supersite (FSF [Watson et al., 2000]), Angiola (ANGI), and Bakersfield (BAC)) and at 35 satellite sites (Table 1). Winter IOPs included five times/day 3–8 hour samples for 15 days at the five anchor sites (Bethel Island (BTI), Sierra Nevada Foothills (SNFH), FSF, ANGI, and BAC) and daily 24-hour sampling for 13 days at 25 satellite sites.

Figure 1.

The 24-hour average speciated ambient PM2.5 network at five anchor sites (denoted by asterisks) and 35 satellite sites during CRPAQS.

Table 1. Summary of CRPAQS Aerosol Measurements at the Anchor and Satellite Sites
Site CodeSite NameSite LocationSite Type/CharacteristicsLongitudeLatitudeElevation, mAnchor Sites,a Sampling PeriodSatellite Sitesb
Filter Pack, PM2.5Sampling Period
AnnualWinter IntensiveT/Cq/nAnnualWinter Intensive
  • a

    Anchor site annual sampling program used DRI medium-volume sequential filter samplers (SFS) equipped with Bendix 240 cyclone (Clearwater, Florida, USA) PM2.5 inlets and preceding nitric acid (HNO3) denuder, consisting of anodized aluminum tubes coated with aluminum oxide [Chow et al., 1993b]. The denuders remove HNO3 with an efficiency of >90% [Chow et al., 2005a]. Sampling was conducted daily, 24 hours/day (midnight to midnight) from 2 December 1999 to 3 February 2001 at a flow rate of 20 L min−1. Two filter packs were used for sampling. (1) Each Teflon/citric acid filter pack (FTC) consists of a front Teflon-membrane filter (#R2PJ047, Pall Corp, Putnam, Connecticut, USA) for mass, babs, and elemental analyses, backed up by a citric-acid-impregnated cellulose-fiber filter (31ET, Whatman, Brentford, Middlesex, U.K.) for ammonia (NH3), and (2) each quartz/sodium chloride (NaCl) filter pack (FQN) consists of a front quartz-fiber filter (#2500QAT-UP, Pall Corp, Putnam, Connecticut, USA) for ion and carbon analyses, backed up by an NaCl-impregnated cellulose-fiber filter for volatilized nitrate.

  • b

    Anchor site winter intensive sampling included both SFS for PM2.5 sampling and sequential gas samplers (SGS) for NH3 and HNO3 sampling by denuder difference on 15 forecast episode days (IOP_1: 15 December 2000 to 18 December 2000, IOP_2: 26 December 2000 to 28 December 2000, IOP_3: 4 January 2001 to 7 January 2001, and IOP_4: 31 January 2001 to 3 February 2001). The two SGS were equipped with (1) citric-acid-coated glass denuders and quartz-fiber filters backed up by citric-acid-impregnated cellulose-fiber filters for NH3 and (2) anodized aluminum denuders and quartz-fiber filters backed up by sodium-chloride-impregnated cellulose-fiber filters for HNO3.

  • c

    FTC filter pack: Teflon-membrane filter samples were analyzed for mass by gravimetry (for filters equilibrated at 21 ± 1.5°C and 35 ± 5%), filter light transmission (babs) by densitometry, and elements (Na, Mg, Al, Si, P, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Se, Br, Rb, Sr, Y, Zr, Mo, Pd, Ag, Cd, In, Sn, Sb, Ba, La, Au, Hg, Tl, Pb, and U) by x-ray fluorescence (XRF [Watson et al., 1998]); quartz-fiber filter samples were analyzed for anions (chloride [Cl], nitrate [NO3], sulfate [SO4=]) by ion chromatography [Chow and Watson, 1999], ammonium (NH4+) by automated colorimetry, water-soluble sodium (Na+) and potassium (K+) by atomic absorption spectrophotometry, and 7-fraction organic and elemental carbon (OC1 combusted at 120°C, OC2 at 250°C, OC3 at 450°C, OC4 at 550°C, EC1 at 550°C, EC2 at 700°C, and EC3 at 800°C with pyrolysis correction) by thermal/optical reflectance (TOR) following the Interagency Monitoring of PROtected Visual Environments (IMPROVE) protocol [Chow et al., 1993c, 2001, 2004, 2005a; Chen et al., 2004; Watson et al., 1994]; citric-acid-impregnated filter samples were analyzed for NH3 by automated colorimetry; and NaCl-impregnated filters were analyzed for volatilized nitrate by ion chromatography.

  • d

    Used battery-powered Minivol samplers (Airmetrics, Eugene, Oregon) equipped with PM10/PM2.5 (in tandem) or PM10 inlets at a flow rate of 5 L min−1. Filter pack assembly for PM2.5 followed the same FTC and FQN configurations of those at the anchor sites, with mass, elements, and NH3 acquired at all 35 sites. Carbon, ions, and volatilized NO3 were acquired at 29 sites.

  • e

    Parentheses indicate that the site includes the PM10 sites operated during the annual program.

ACPAngels Campelevated rural; 6850 Studhorse Flat Road, Sonoraintrabasin gradient−120.49138.006373  FTCcFQNdXX
ALT1Altamont Passelevated rural; Flynn Road exit, I-580interbasin transport−121.66037.718350  FTC XX
ANGIAngiola-ground levelrural; 36078 4th Avenue, Corcoranintrabasin gradient/transport, vertical gradient, visibility−119.53835.94860XX    
BACBakersfield-5558 California Streeturban; 5558 CA Ave. #430 (STI) #460 (ARB), Bakersfieldcommunity exposure, visibility−119.06335.357119XX    
BODGBodega Marine Labmarine; Bodega Marine Lab, 2099 Westside Road, Bodega Bayboundary/background−123.07338.31917  FTCFQNXX
BRESBAC-residentialurban; 7301 Remington Avenue, Bakersfieldsource: wood burning−119.08435.358117  FTCFQNXX
BTIBethel Islandrural; 5551 Bethel Island Road, Bethel Islandinterbasin transport−121.64238.0062 XFTCFQNX 
CARPCarrizo Plainelevated rural; Soda Springs Road, 0.5 mile south of California Valleyintrabasin gradient, visibility−119.99635.314598  FTC X 
CHLChina Lakeelevated rural; Baker sitevisibility−117.77635.774684  FTCFQNX 
CLOClovissuburban; 908 N. Villa, Cloviscommunity exposure−119.71636.819108  FTCFQNXX
COPCorcoran-Patterson Avenuerural; 1520 Patterson Ave., Corcorancommunity exposure−119.56636.10263  FTCFQN(X)eX
EDIEdisonurban; 4101 Kimber Avenue, Bakersfieldintrabasin gradient−118.95735.350118  FTC XX
EDWEdwards Air Force Baseelevated rural; north end of Rawinsonde Road, Edwards AFBintrabasin gradient, visibility−117.90434.929724  FTCFQNX 
FEDLfeedlot or dairyrural; 8555 S. Valentine, Fresno (near Raisin City)source: cattle−119.85536.61176  FTCFQNXX
FELFellowselevated rural;across from 25883 Hwy 33, Fellowssource: oilfields−119.54635.203359  FTCFQNXX
FELFfoothills above Fellowselevated rural; Texaco Pump Site 47-1, Fellowsintrabasin gradient−119.55735.171512  FTCFQNXX
FREMFresno MVurban; Pole #16629, 2253 E. Shields Ave., Fresnosource: motor vehicle−119.78336.78096  FTCFQNXX
FRESresidential area near FSF, with wood burningurban; Pole #16962, 3534 Virginia Lane, Fresnosource: wood burning−119.76836.78397  FTCFQN(X)X
FSFFresno-3425 First Streeturban; 3425 First Street, Fresnocommunity exposure, visibility−119.77336.78297XX    
HELMagricultural fields/Helm-central Fresno Countyrural; near Placer and Springfieldintrabasin gradient−120.17736.59155  FTCFQNXX
KCWKettleman Cityrural; Omaha Avenue 2 miles west of Hwy 41, Kettleman Cityintrabasin gradient−119.94836.09569  FTC XX
LVR1Livermore - new siterural; 793 Rincon Street, Livermoreinterbasin transport−121.78437.688138  FTCFQNXX
M14Modesto 14th St.urban; 814 14th Street, Modestocommunity exposure−120.99437.64228  FTCFQN(X)X
MOPMojave-Pooleelevated rural; 923 Poole Street, Mojavecommunity exposure−118.14835.051832  FTCFQNX 
MRMMerced-midtownsuburban; 2334 M Street, Mercedcommunity exposure−120.48137.30853  FTCFQNXX
OLDOildale-Manorsuburban; 3311 Manor Street, Oildalecommunity exposure−119.01735.438180  FTCFQN(X) 
OLWOlanchaelevated rural; just to east of Hwy 395background−117.99336.2681124  FTCFQNXX
PACPacheco Passelevated rural; Upper Cottonwood Wildlife Area, west of Los Banosinterbasin transport−121.22237.073452  FTC X 
PIXLPixley Wildlife Refugerural; Road 88, 1.5 miles north of Avenue 56, Alpaughrural, intrabasin gradient−119.37635.91469  FTCFQNXX
PLEPleasant Grove (north of Sacramento)rural; 7310 Pacific Avenue, Pleasant Groveintrabasin gradient−121.51938.76610  FTCFQNX 
S13Sacramento-1309 T Streeturban; 1309 T Street, Sacramentocommunity exposure−121.49338.5686  FTCFQNXX
SELMSelma (south Fresno area gradient site)rural; 7225 Huntsman Avenue, Selmacommunity exposure−119.66036.58394  FTCFQNXX
SFASan Francisco - Arkansasmarine/urban; 10 Arkansas St., San Franciscocommunity exposure−122.39937.7666  FTCFQNX 
SNFHSierra Nevada Foothillselevated rural; 31955 Auberry Road, Auberryvertical gradient, intrabasin gradient, visibility−119.49637.063589 XFTCFQNX 
SOHStockton-Hazeltonurban; 1601 E. Hazelton, Stocktonintrabasin gradient−121.26937.9508  FTCFQNXX
SWCSW Chowchillarural; 20513 Road 4, Chowchillainterbasin transport−120.47237.04843  FTCFQNXX
TEH2Tehachapi Passelevated rural; Near 19805 Dovetail Court, Tehachapiinterbasin transport, visibility−118.48235.1681229  FTC XX
VCSVisalia Church St.urban; 310 Church Street, Visaliacommunity exposure−119.29136.333102  FTCFQN(X)X
Total number of sites      3535293525

[9] The FSF and BAC sites represented the two major urban centers in the SJV. ANGI, located between these two urban centers, was chosen to represent regional transport and/or pollutant gradients. BTI and SNFH operated during winter IOPs were intended to represent interbasin gradient and transport boundary conditions. Both were also satellite sites during the annual program. BTI is located at the northwest corner of the SJV ∼50 km east of San Francisco. SNFH (589 m above mean sea level [MSL]) is located on the upslope of the western Sierra Nevada approximately at the same latitude as FSF. The 38 sites were categorized into eight site types depending on the type of land use and surrounding environs (Table 1). These included eighteen community exposure sites, eleven emissions source-dominated sites, nine visibility sites, eleven intrabasin gradient sites, two vertical gradient sites, one intrabasin transport site, six interbasin transport sites, and seven boundary/background sites. These were nominal classifications made during the study design, and it was later found that several sites represented different environments at different times of the year.

[10] At each of the anchor sites, a Desert Research Institute (DRI, Reno, NV, USA) sequential filter sampler (SFS) [Chow et al., 1994, 1996; Chen et al., 2002] collected PM2.5 through two sampling channels (20 liters per minute [L min−1]). Details of the sampling system and filter pack configuration are documented in Table 1 footnotes. The backup sodium chloride (NaCl)-impregnated cellulose-fiber filter collected nitrate (NO3) volatilized from the quartz-fiber filter to evaluate the negative bias for particulate NO3 measurements [Zhang and McMurry, 1992; Hering and Cass, 1999; Chow et al., 2005b]. The degree of NH4NO3 evaporation from the front quartz-fiber filter depends on the temperature, relative levels of gaseous NH3, HNO3, particulate NH4NO3 in the ambient air, and the fraction of gaseous species removed (denuded) in the sampling stream. Ashbaugh et al. [2004] reported that the inlet of a nondenuded IMPROVE sampler removed HNO3 as effectively as an IMPROVE sampler with a HNO3 denuder. In this paper, pNO3 (total particulate NO3) represents the sum of nonvolatilized NO3 from the front filter and volatilized NO3 from the backup filter. Two sequential gas samplers (SGSs [Chow et al., 1996; Chen et al., 2002]) at the five anchor sites during the winter IOPs quantified gaseous NH3 and HNO3 using the denuder difference method [Chow et al., 1993b].

[11] PM2.5 MiniVol samplers (AirMetrics, Springfield, Oregon, USA) that were used at the satellite sites yield mass concentrations comparable to PM2.5 Federal Reference Method (FRM) compliance samplers [Baldauf et al., 2001; Chow et al., 2005b]. Occasional malfunctions of batteries and pumps resulted in missing data. The satellite network had a valid data capture rate in excess of 80% over the study period with the exception of the dairy site (FEDL; 62%), the Edwards site (EDW; 77%), and the Bakersfield residential site (BRES; 66%) (Table 2). Since the missing data occurred randomly in time, they are not expected to bias the annual averages.

Table 2. Summary of PM2.5 Mass and Chemical Composition at 38 Sites During CRPAQSa
Site Code14-Monthb/Annualc Valid PM2.5 MeasurementsSpring Mean PM2.5,d μg m−3Summer Mean PM2.5,d μg m−3Fall Mean PM2.5,d μg m−3Winter Mean PM2.5,d μg m−3Annual Means From Quarters,e μg m−3Annual Mean PM2.5,c μg m−314-Month Mean PM2.5,b μg m−3Maximum PM2.5, μg m−3Maximum DateChigh,f μg m−3Clow,g μg m−3Fhigh,h %Annual NH4NO3 (Front),i μg m−3Annual NH4NO3 (Backup),j μg m−3Annual OM,k μg m−3Annual EC, μg m−3Annual (NH4)2SO4,l μg m−3Annual Crustal,m μg m−3Reconstructed PM2.5 Mass,n μg m−3PM2.5 Mass Closure,o %
  • a

    Anchor sites are coded in bold; refer to Table 1 for site codes.

  • b

    Consists of 72 every-6th-day sampling from 2 December 1999 to 3 February 2001.

  • c

    Consists of 61 every-6th-day sampling from 1 February 2000 to 31 January 2001 (CRPAQS annual period).

  • d

    Seasonal averages of spring (March to May), summer (June to August), fall (September to November), and winter (December to February) for the CRPAQS annual period.

  • e

    Arithmetic means of the four calendar quarters: January to March, April to June, July to September, and October to December during the CRPAQS annual period. January is from 2001 and the rest of the months are from 2000. Italics indicate <75% coverage in at least one quarter.

  • f

    Average of high PM2.5 period (1 November 2000 to 31 January 2001).

  • g

    Average of low PM2.5 period (1 February 2000 to 31 October 2000).

  • h

    Fhigh = [Chigh/(Chigh + 3 × Clow)] × 100.

  • i

    1.29 × ([NO3]FRONT).

  • j

    1.29 × ([NO3]BACKUP).

  • k

    1.4 × [OC].

  • l

    1.38 × [SO42−].

  • m

    2.2 × [Al] + 2.49 × [Si] + 1.63 × [Ca] + 2.42 × [Fe] + 1.94 × [Ti].

  • n

    NH4NO3 (front filter, nonvolatilized NH4NO3) + OM +EC + (NH4)2SO4 + crustal material (CM) + trace elements (other than geological material) and sea salt (Na+ + Cl).

  • o

    (Summed PM2.5 mass/annual mean PM2.5) × 100%.

ACP72/613.9 ± 2.13.6 ± 1.63.6 ± 2.25.0 ± 5.43.43.4 ± 2.14.2 ± 3.618.91/7/20003.53.425.51.00.13.80.91.10.47.6223
ALT168/614.2 ± 1.85.1 ± 2.66.1 ± 10.213.3 ± 18.77.27.3 ± 11.77.8 ± 12.271.71/7/200116.83.958.60.3
ANGI55/5010.9 ± 3.911.4 ± 5.620.6 ± 26.729.1 ± 31.018.719.1 ± 23.719.3 ± 23.1123.41/7/200141.411.355.08.11.55.30.82.03.019.6103
BAC66/5719.8 ± 12.813.3 ± 2.923.2 ± 22.043.5 ± 36.726.027.0 ± 27.528.1 ± 27.7132.71/1/200156.915.355.49.82.49.92.02.63.428.3105
BODG67/5710.5 ± 4.94.6 ± 4.15.5 ± 4.314.8 ± 9.88.99.3 ± 7.810.0 ± 8.135.31/19/200113.67.936.31.60.11.60.41.90.210.2110
BRES45/407.9 ± 3.17.6 ± 2.520.1 ± 27.653.6 ± 42.143.727.9 ± 36.530.6 ± 36.7158.91/1/200159.17.173.510.41.010.02.82.11.427.498
BTI72/614.1 ± 1.94.5 ± 2.97.0 ± 11.818.5 ± 19.28.88.9 ± 13.910.0 ± 14.276.61/7/200122.83.965.83.50.14.11.31.70.712.2137
CARP63/634.2 ± 2.33.4 ± 2.27.5 ± 8.18.3 ± 10.05.56.0 ± 7.06.2 ± 7.432.61/19/200111.83.950.40.9
CHL60/512.1 ± 1.33.3 ± 2.51.4 ± 1.45.6 ± 16.41.71.9 ± 1.83.4 ± 9.874.51/7/20000.82.49.90.40.24.10.81.00.47.1375
CLO66/567.8 ± 3.48.6 ± 2.521.4 ± 31.047.0 ± 38.220.620.8 ± 27.825.3 ± 32.0130.11/1/200155.69.167.07.80.89.62.51.91.223.5113
COP71/6010.0 ± 7.08.0 ± 1.720.0 ± 19.237.8 ± 35.717.918.2 ± 22.621.9 ± 26.5124.71/7/200142.39.459.98.70.77.21.72.01.621.8120
EDI64/5510.0 ± 5.710.5 ± 3.332.2 ± 43.838.3 ± 40.024.924.5 ± 34.524.9 ± 33.2160.81/1/200148.014.851.93.1
EDW50/474.5 ± 2.16.3 ± 1.35.0 ± 4.75.1 ± 4.95.05.3 ± 3.65.3 ± 3.716.99/21/20005.65.326.11.20.74.00.81.50.88.8165
FEDL38/3827.8 ± 12.325.3 ± 15.338.6 ± 32.728.829.9 ± 21.329.9 ± 21.3115.71/7/200138.423.735.010.60.29.71.92.33.429.097
FEL71/617.5 ± 6.95.4 ± 1.513.4 ± 19.220.1 ± 20.612.212.2 ± 16.412.7 ± 16.474.21/1/200130.05.963.05.40.85.51.12.11.215.9131
FELF70/605.1 ± 2.84.9 ± 1.713.2 ± 18.119.6 ± 18.311.611.6 ± 15.412.0 ± 15.169.41/1/200129.55.165.86.50.75.20.92.20.615.9137
FREM67/609.7 ± 3.69.1 ± 2.921.8 ± 27.256.5 ± 47.724.825.3 ± 35.227.6 ± 36.3176.01/1/200167.69.969.58.10.813.93.72.01.229.7118
FRES66/579.0 ± 3.97.8 ± 2.721.2 ± 27.055.9 ± 46.522.824.2 ± 33.928.2 ± 37.0169.41/1/200163.38.970.37.60.811.72.91.90.825.7106
FSF71/6011.2 ± 6.09.4 ± 2.920.1 ± 21.953.9 ± 41.523.323.7 ± 29.428.4 ± 33.4148.31/1/200160.010.665.45.22.610.42.01.81.521.590
HELM70/595.0 ± 2.05.5 ± 2.112.7 ± 15.525.9 ± 28.511.811.8 ± 16.314.4 ± 20.7114.812/26/199930.85.366.16.50.75.01.41.80.816.2138
KCW64/556.1 ± 3.36.1 ± 2.213.9 ± 26.732.6 ± 32.710.912.9 ± 19.316.8 ± 25.3112.71/7/200029.95.962.90.9
LVR172/616.2 ± 4.16.0 ± 3.68.7 ± 11.120.6 ± 22.310.510.6 ± 14.911.9 ± 15.895.41/7/200124.75.659.63.30.26.22.41.60.314.6137
M1471/606.1 ± 2.27.1 ± 5.116.0 ± 22.741.9 ± 34.417.317.3 ± 25.521.0 ± 27.7136.11/7/200147.86.272.16.20.69.22.42.00.621.1122
MOP69/585.6 ± 3.65.3 ± 1.94.8 ± 3.82.8 ± 3.34.44.3 ± 3.34.4 ± 3.415.611/14/20002.94.816.61.00.55.01.11.40.69.5219
MRM72/617.0 ± 3.56.8 ± 3.313.0 ± 13.236.6 ± 28.313.914.0 ± 14.418.9 ± 22.5115.912/20/199932.47.459.36.50.48.92.21.80.720.7148
OLD65/559.5 ± 5.77.6 ± 2.023.1 ± 26.840.9 ± 38.421.521.1 ± 29.023.5 ± 29.9140.61/1/200152.68.268.311.01.09.01.92.61.326.3125
OLW65/543.1 ± 4.45.5 ± 9.91.5 ± 1.33.0 ± 3.93.23.1 ± 5.83.2 ± 5.639.27/29/20001.93.615.30.30.14.00.70.80.77.0224
PAC71/613.5 ± 2.12.9 ± 1.64.6 ± 8.314.2 ± 17.66.16.1 ± 9.27.4 ± 12.164.312/26/199915.02.962.90.1
PIXL69/619.8 ± 6.710.0 ± 8.217.7 ± 14.538.6 ± 33.018.418.5 ± 21.521.2 ± 24.1106.61/7/200142.99.859.49.90.46.01.62.21.321.6117
PLE70/596.4 ± 2.56.4 ± 2.59.1 ± 9.818.0 ± 18.09.19.1 ± 9.111.1 ± 12.766.312/20/199918.65.652.43.00.26.81.91.40.514.2155
S1368/584.5 ± 1.64.9 ± 2.810.3 ± 10.824.4 ± 24.010.911.1 ± 14.813.2 ± 17.790.212/20/199927.84.766.23.80.27.22.31.50.716.3147
SELM71/6111.4 ± 7.38.9 ± 3.318.0 ± 16.541.2 ± 34.718.218.3 ± 19.422.8 ± 26.0110.412/26/199940.310.556.18.30.87.82.32.21.122.5122
SFA72/618.0 ± 4.75.1 ± 3.79.0 ± 7.916.2 ± 14.79.29.2 ± 8.610.5 ± 10.863.412/26/199918.36.050.53.10.14.51.82.00.513.8150
SNFH70/606.3 ± 3.75.6 ± 1.88.0 ± 4.418.1 ± 18.68.58.5 ± 8.110.7 ± 12.670.21/1/200015.65.946.82.80.46.41.31.60.613.4157
SOH70/595.4 ± 2.37.2 ± 5.29.4 ± 9.031.1 ± 27.312.712.8 ± 16.516.1 ± 20.7103.312/20/199932.86.064.74.80.67.22.21.90.617.9140
SWC70/597.4 ± 2.56.5 ± 2.213.7 ± 15.528.0 ± 28.713.112.9 ± 15.716.0 ± 20.997.412/26/199932.56.861.56.30.74.51.41.60.815.6122
TEH264/539.1 ± 3.76.5 ± 2.77.4 ± 5.15.6 ± 8.97.37.3 ± 6.26.8±6.235.412/8/20007.37.325.10.5
VCS72/6114.1 ± 8.79.5 ± 3.318.6 ± 18.146.9 ± 37.221.721.9 ± 24.725.9 ± 28.8123.71/1/200150.311.858.79.50.89.82.42.31.326.5121

[12] Uncertainty was determined for each measurement on the basis of (1) sample volume uncertainty, based on flow rate performance tests; (2) replicate precision from the chemical analyses; and (3) the uncertainty of the dynamic field blank, which is the larger of the standard deviation of the individual blank values or their root-mean-squared analytical uncertainty. The valley-wide average blank concentrations for PM2.5 mass, nonvolatilized NO3, volatilized NO3, OC, EC, and NH3 were 2.1 ± 1.2, 0.1 ± 0.1, 0.01 ± 0.04, 3.1 ± 1.3, 0.2 ± 0.3, and 0.9 ± 0.5 μg m−3, respectively. Ambient concentrations reported for CRPAQS are blank subtracted [e.g., Watson et al., 1995]. For mass, NO3, ammonium (NH4+), and total carbon (TC = OC + EC), the uncertainty was typically within ±10% for measured values exceeding 10 times the minimum detection level (MDL). Measured NO3 and sulfate (SO4=) were compared to NH4+ as part of the data validation process. The high correlation (R2 ∼ 0.99) between the anions and cations, and a nearly 1:1 molar ratio, indicates that the dominant form of NH4+ was NH4NO3. Only ∼9% of NH4+ was associated with other anions, mainly SO4=. Hereafter, the concentration of front quartz-filter NH4NO3 is estimated as 1.29 × [NO3] (and pNH4NO3 as 1.29 × [pNO3]). In addition to the CRPAQS network, PM2.5 and PM10 mass measurements were acquired with FRMs at compliance sites in cities and at IMPROVE sites in central California's Class I areas. Some of these were collocated with CRPAQS measurements [Chow et al., 2006a].

3. Spatiotemporal Variations of PM2.5

[13] Annual-average PM2.5, based on four quarterly averages (i.e., calendar quarter [see U.S. Environmental Protection Agency (U.S. EPA), 1997]) at 14 of the 38 CRPAQS sites, exceeded the U.S. annual PM2.5 National Ambient Air Quality Standard (NAAQS) of 15 μg m−3 (Table 2). Most of these exceedances occurred in the southern SJV at urban sites such as FSF (23 μg m−3), Visalia (VCS; 22 μg m−3), and BAC (26 μg m−3), and also at the regional transport ANGI site (18.7 μg m−3). These averages differ from every sixth day arithmetic means from the annual program by <10%, except at BRES. This corroborates the limited influence of missing data on the annual averages. The annual average determined by sixth day sampling is used in subsequent analyses rather than the annual average of quarterly averages required to determine NAAQS attainment.

[14] The PM2.5 concentration decreased rapidly toward the higher-elevation valley boundary (Figure 2a). Three sites in Bakersfield (residential BRES site, urban BAC site, and interbasin gradient Edison [EDI] site, all ∼118 m above MSL) reported consistently high annual-average PM2.5 concentrations of 24–28 μg m−3, despite the fact that each site represents different microenvironments. Tehachapi (TEH2), an interbasin transport site, located ∼50 km to the southeast of EDI at 1229 m above MSL, recorded an annual-average PM2.5 concentration of 7.3 μg m−3. The annual-average PM2.5 concentration decreased further at the Mojave Desert (EDW; 724 m above MSL) and Mojave-Pool (MOP; 832 m above MSL) sites, averaging only 4.3–5.4 μg m−3. Similarly, annual-average PM2.5 concentration decreased from 24 μg m−3 at FSF, to 21 μg m−3 at Clovis (CLO; 108 m above MSL, a suburban site ∼10 km east of FSF), and to 8.5 μg m−3 at SNFH (589 m above MSL, 33 km east of CLO). This reflects the influence of topography and the generally low vertical mixing potential due to weak boundary layer turbulence on many of the high PM2.5 days. North of Fresno, the annual-average PM2.5 concentration was relatively low even at the urban centers of Sacramento (S13, 11.1 μg m−3) and San Francisco (SFA, 9.2 μg m−3), with the highest annual-average concentration of 17.3 μg m−3 observed at Modesto (M14).

Figure 2.

Spatial distribution of (a) annual PM2.5 concentration (1 February 2000 to 31 January 2001) during CRPAQS and geographical cross sections A, B, and C and (b) the sampling sites and cross section D. Contours are determined with a two-dimensional cubic-spline algorithm using only sites with >70% valid measurements. The stars indicate locations of the sampling sites.

[15] The stable atmosphere surrounding the Sierra Nevada and coastal mountains prevents precursor gases and PM released in the SJV from rapidly dispersing. To some extent, the valley is also isolated from the influences of outside sources. This is especially true for the southern SJV because the elevation of the valley floor generally increases from north to south as far as Fresno and descends south of Fresno. The five most northwestern sites in this network, located at Bodega Bay (BODG), BTI, SFA, S13, and Stockton (SOH), all have elevations less than 10 m above MSL. Marine air enters the SJV through the Carquinez Straight east of the San Francisco Bay area, leading to the lower PM2.5 in the northern valley. The highest annual-average water-soluble sodium (Na+; a sea salt marker) concentrations were found at coastal sites west of the valley (BODG, SFA, and Livermore (LVR1)) where annual-average PM2.5 and Na+ concentrations were 9.3 and 1.7, 9.2 and 0.58, and 10.6 and 0.32 μg m−3, respectively. Three sites in the northern valley, S13, BTI, and SOH, also experienced higher annual-average Na+ concentration (0.24–0.28 μg m−3) than FSF (0.11 μg m−3) and BAC (0.13 μg m−3). The lower PM2.5 concentrations at S13 and SOH compared with higher concentrations at down-valley urban sites demonstrate the influence of clean marine air in the northern valley.

[16] As shown in Figure 3, 24-hour PM2.5 concentrations at FSF were low from mid-February 2000 to late October 2000, but frequently exceeded 15 μg m−3 from November to January, and reached a maximum of 148 μg m−3 on 1 January 2001. A similar temporal pattern was found at BAC, which often reported higher PM2.5 concentrations than FSF. In addition to the increased RWC emissions during winter [Magliano et al., 1999; Schauer and Cass, 2000], the meteorological effect (i.e., prolonged Great Basin highs causing subsidence) on the ventilation of pollutants and formation of secondary aerosol also contributes to the seasonal cycle. The regional transport ANGI site, located in the ancient Tulare Lake bed, surrounded by farm fields and sparse residences, experienced wintertime PM2.5 concentrations similar to those at FSF and BAC. High winter concentrations at the other two interbasin anchor sites (BTI and SNFH) were much less pronounced. BODG in Figure 3 represents the northern boundary/background site of the SJV, while the ACP (373 m above MSL) and TEH2 (1229 m above MSL) sites represent the eastern and southern intra and interbasin transport sites. No appreciable seasonal variations were observed at these boundary and transport sites, especially at BODG and TEH2; the background PM2.5 level, which is often influenced by long-range (synoptic-scale) transport, appeared to be consistent year-round.

Figure 3.

Time series of 24-hour average PM2.5 concentration at selected sites during CRPAQS including northern boundary/background site (BODG); interbasin anchor sites (BTI, SNFH); eastern transport site (ACP); southern transport site (TEH2); regional transport anchor site (ANGI); and urban anchor sites (FSF and BAC). The Y axis is PM2.5 concentration (μg m−3). Vertical shaded areas indicate the Winter Intensive Operating Periods (IOPs).

[17] The patterns of temporal variations in Figure 3 are consistent with limited differences in PM2.5 spring (March to May) and summer (June to August) averages (Table 2). Urban-rural contrast in the northern SJV was minimal during spring and summer. For example, average spring PM2.5 concentration at BODG (10.5 μg m−3) compared well with those at FSF (11.2 μg m−3) and the Fresno residential site (FRES; 9.0 μg m−3). However, the source-dominated dairy site (FEDL) reported elevated average PM2.5 concentrations (25–28 μg m−3) during summer and fall, and reached a maximum of 39 μg m−3 in winter.

[18] CRPAQS annual measurements may be divided into high (Chigh; 1 November 2000 to 31 January 2001) and low (Clow; 1 February 2000 to 31 October 2000) PM2.5 periods. As shown in Table 2, PM2.5 approached 15 μg m−3 at Bakersfield (BAC, EDI), even during the low period. The maximum Clow occurred at the source-dominated FEDL site (24 μg m−3). The contributions of PM2.5 during Chigh to annual averages (i.e., Fhigh, defined in Table 2), ranged from 13% at China Lake (CHL) to 72% at M14. Chigh contributed more than 50% of annual-average PM2.5 concentrations at most sites inside the valley, 55% at the BAC, and 63% at the FSF urban centers. Fhigh was <25% only at three sites in the network: CHL (10%), MOP (17%), and Olancha (OLW; 15%), all of which are located in the Mojave Desert or Great Basin Valleys. The slightly higher Clow than Chigh at these desert sites is consistent with previously observed transport from the SJV and southern California during nonwinter months [Green et al., 1992].

4. PM2.5 Chemical Composition

[19] Table 2 presents the annual-average concentrations of five main PM2.5 components (i.e., NH4NO3, ammonium sulfate [(NH4)2SO4], organic matter [OM = 1.4 × OC], EC, and crustal material), as well as the PM2.5 mass balance. CRPAQS confirms previous studies conducted in the SJV [e.g., Chow et al., 1992, 1993a, 1996, 1999] that PM2.5 consists mainly of NH4NO3 and carbonaceous material. Volatilized NH4NO3 from the backup filter is not included in the reconstructed mass (defined in Table 2) since PM2.5 mass determined from Teflon-membrane filters does not contain volatilized NO3 [Chow et al., 2005b]. The OC multiplier of 1.4, which accounts for unmeasured hydrogen, oxygen, and other elements in OM, was derived from the analysis of organic compounds in urban aerosols [Grosjean and Friedlander, 1975; White and Roberts, 1977]. This factor is environment-specific with lower values found in urban or source-dominated atmospheres, and with higher values in remote locations [Turpin and Lim, 2001; El-Zanan et al., 2005]. A value of 1.4, however, remains useful for cross-environment averages [Russell, 2003]. Because the CRPAQS network contained both urban and rural sites, the value of 1.4 is appropriate. This factor has also been adopted for mass- and light-extinction reconstruction in the IMPROVE network of U.S. national parks and wilderness areas [Malm et al., 1994].

[20] Besides the factors applied to OC and mineral oxides, potential biases in mass closure include sampling artifacts caused by volatile organic compounds (VOCs). Adsorption of VOCs onto quartz-fiber filters [Turpin et al., 1994; Chow and Watson, 2002; Chow et al., 2006b] is known to bias OM mass positively. This artifact may be partially compensated for by the evaporation of OM from the filters [Zhang and McMurry, 1987; Chen et al., 2002; Subramanian et al., 2004]. The relative importance of positive and negative sampling biases was examined at FSF using parallel denuded (organic vapor denuder) and nondenuded channels followed by quartz-fiber/quartz-fiber filter packs [Watson and Chow, 2002b; Chow et al., 2006b]. Average nondenuded and denuded front quartz-fiber filter OC concentrations were 11.8 ± 1.2 and 10.8 ± 1.1 μg m−3, respectively, during winter; and 4.8 ± 0.6 and 3.9 ± 0.5 μg m−3, respectively, during summer. Average nondenuded and denuded backup quartz-fiber filter OC concentrations were 2.1 ± 0.3 and 0.25 ± 0.41 μg m−3, respectively, during winter; and 1.84 ± 0.28 and 0.50 ± 0.42 μg m−3, respectively, during summer. On the basis of the nondenuded backup quartz-fiber filter concentrations, a seasonally constant sampling artifact of ∼3 μg m−3 OM could bias the mass closure for samples with low concentrations. The four lowest (<5 μg m−3) annual-average PM2.5 concentrations in Table 2 (i.e., CHL, OLW, ACP, and MOP) have mass closure >200%. In these cases, the positive VOC artifact appeared to dominate, which is consistent with other recent studies [e.g., Subramanian et al., 2004; Chow et al., 2006b].

[21] PM2.5 mass closure was <100% at FSF, FEDL, and BRES, where PM2.5 concentrations were relatively high. This may be due in part to water retention by hygroscopic species, such as NH4NO3 and (NH4)2SO4, and/or an underestimation of the OC multiplier [Andrews et al., 2000; Turpin and Lim, 2001; Rees et al., 2004; El-Zanan et al., 2005; Khlystov et al., 2005]. Nevertheless, measured and reconstructed mass were highly correlated, with an R2 of 0.94. NH4NO3 and OM are the most dominant components of PM2.5, accounting for 66% and 73% of PM2.5 mass at urban FSF and BAC, respectively.

[22] A triangle-based cubic interpolation algorithm [Sandwell, 1987; Barber et al., 1996] was used to visualize the spatial variability. Sites with more than 30% missing data (see Table 2) were excluded from the analysis. Average concentrations of PM2.5 mass and its chemical components during Clow and Chigh periods are compared along three geographic cross sections in Figure 4 (A, B, and C, defined in Figure 2a). These cross sections all intersect at FSF and cover major SJV geographical features. During the Clow period, urban sites experienced PM2.5 concentrations slightly higher than rural sites. A nearly uniform OM distribution was observed along cross section C that stretches between the Sierra Nevada and the coastal mountains (119.3–120.5° W longitude). Average OM concentrations at Helm (HELM), FSF, and SNFH during the Clow period were 5.3, 6.8, and 6.4 μg m−3, respectively. EC, which is not subject to sampling artifacts, showed a similar pattern, averaging 1.4, 1.1, and 1.2 μg m−3, respectively. Within the SJV, pNH4NO3 was generally <7 μg m−3, decreasing to ∼2 μg m−3 at the elevated mountain sites of ACP, SNFH, and CHL. Most (>80%) of the pNH4NO3 was found on the front filter, except at BAC (cross section B) where more than 50% was found on the backup filter. (NH4)2SO4 and crustal material were minor components of PM2.5.

Figure 4.

Chemical composition and mass closure of PM2.5 along the geographical cross sections A, B, and C (defined in Figure 2a) for the low (Clow) and high (Chigh) PM2.5 periods during CRPAQS. The Y axes represent concentration (μg m−3). Sampling sites located approximately on each cross section are noted. The dashed line represents measured PM2.5 mass.

[23] The pNH4NO3 in the southern SJV was much higher during the Chigh period than the Clow, with the highest average concentration of ∼30 μg m−3 observed at BAC. Elevated pNH4NO3 concentrations were not limited to urban areas. The rural HELM site in central Fresno County (55 m above MSL in cross section C), ∼41 km to the west of FSF, reported a pNH4NO3 concentration of ∼17.1 μg m−3, close to levels found in the Fresno area (19–22 μg m−3). The pNH4NO3 average was even higher at ANGI, approaching 22 μg m−3. However, pNH4NO3 concentrations decreased rapidly with site elevation and location outside the SJV. The MOP site (832 m above MSL in the Mojave Desert) reported 1.1 and 1.3 μg m−3 pNH4NO3 during the Clow and Chigh periods, respectively. Figure 4 demonstrates that widespread pNH4NO3 is the major contributor to wintertime basin-wide PM2.5 episodes.

[24] NOx is converted to HNO3 by photochemical processes that also involve VOCs [Stockwell et al., 2000; Pun and Seigneur, 2001]. HNO3 reacts reversibly with NH3 to form solid NH4NO3. If the RH is high enough, the solid deliquesces to form ions in solution. Winter's cold and humid weather favors NH4NO3 over gaseous NH3 and HNO3 [Seinfeld and Pandis, 1998; Moya et al., 2001; Takahama et al., 2004]. The seasonal cycle of the NH3(g)/NH4+(p) ratio at FSF (Figure 5) confirms the shift of this equilibrium toward NH3 in the spring-fall period. Surface wind speeds in the SJV are often <1 m s−1 during winter, with variable wind directions. Surface transport distances estimated from these wind speeds are insufficient to account for the mixing of nonurban NH3 emissions with urban NOx and VOC emissions for the formation of widespread NH4NO3 [Smith and Lehrman, 1994]. The spatial distribution of NH4NO3 corroborates the hypothesis of Watson and Chow [2002a] that transport/mixing is facilitated by a valley-wide mixed layer above the shallow (∼20 m) nighttime surface layer.

Figure 5.

Seasonal variation of total ammonia (NH3 + NH4+ as NH3 equivalent) concentration and NH3/NH4+ ratio at the Fresno site (FSF) during CRPAQS. Note that the Y axis on the right has a logarithmic scale.

[25] In contrast with pNH4NO3, no appreciable increases of EC or OM were detected at rural sites, such as BTI, HELM, and ANGI (Figure 4) during the Chigh period. While pNH4NO3 increased from 3.5 μg m−3 (Clow) to 17.1 μg m−3 (Chigh) at HELM, the EC concentration remained between 1.4 and 1.7 μg m−3. This is consistent with a weak source of primary PM emissions in the rural areas. EC concentrations of 4 μg m−3 or higher were found at the urban sites M14, S13, FSF, and BAC. OM generally tracked with EC, which exacerbated urban PM pollution already enhanced by NH4NO3. PM2.5 water-soluble potassium (K+), a prominent marker for RWC emissions [Chow et al., 1992], averaged 0.05 and 0.31 μg m−3 during the Clow and Chigh periods, respectively, at the FRES site. At BAC, K+ concentrations were 0.07 and 0.34 μg m−3 during the Clow and Chigh periods, respectively, while the corresponding average ratios of soluble to total K were 0.37 and 0.87, respectively. It is evident that RWC caused elevated winter EC and OM concentrations in the urban areas.

[26] Levoglucosan is also a marker for RWC emissions [Simoneit et al., 1999; Fine et al., 2002; Simoneit, 2002; Mochida and Kawamura, 2004]. Rinehart et al. [2006] measured annual-average organic compound concentrations at 20 satellite sites in the SJV and reported high concentrations of levoglucosan at FSF. Poore [2002] showed that levoglocosan concentrations at FSF were five times higher during winter than summer.

[27] The summed concentrations of nonvolatilized NH4NO3, OM, and EC exceeded measured PM2.5 mass during the Clow period (Figure 4). As discussed above, VOC adsorption on the front quartz-fiber filter is likely the major cause of mass closure >100% during Clow. During Chigh, the sum of chemical components did not explain PM2.5 mass at the urban sites (Figure 4). The amount of water retained by NH4NO3 on the Teflon-membrane filter (e.g., during weighing at 35 ± 5% RH) is not expected to be a substantial fraction of the PM2.5 concentration [Chan et al., 1992]. The VOC adsorption artifact appeared to be temporally uniform regardless of the PM2.5 concentration, as suggested by Chow et al. [2006b]. This implies that the artifact is more important for samples with PM2.5 <15 μg m−3. Note that 84% of the samples with PM2.5 concentration >20 μg m−3 showed mass closure at <110% regardless of the site; well within the ±10% measurement uncertainties of the PM2.5 mass.

5. Winter PM2.5 Episodes

[28] The winter IOPs were selected from boundary layer stability forecasts, which were based on meteorological characteristics in the SJV (mixing height, wind speed, and RH) normally associated with high PM2.5 concentrations. The selected four IOP periods are listed in Table 1. Five time-integrated measurements (0000–0500, 0500–1000, 1000–1300, 1300–1600, and 1600–2400 LT) were taken during each IOP day. Daily PM2.5 mass and chemical concentrations were calculated as time-weighted averages of the concentrations during these periods.

[29] Figure 6 presents the concentrations of PM2.5 mass, pNH4NO3, and EC along geographic cross section D (defined in Figure 2b), which represents the valley's primary axis. No apparent contrasts for PM2.5, or its chemical components, were found during IOP_1 (15–18 December 2000). The pNH4NO3 started accumulating during IOP_2 (26 December 2000) in the southern SJV and appeared to persist through early January 2001. High EC (>10 μg m−3) was also observed around urban centers such as M14, FSF, and BAC.

Figure 6.

Spatial and temporal variations of (a) PM2.5 mass, (b) particulate ammonium nitrate (pNH4NO3), and (c) elemental carbon (EC) across the four CRPAQS winter intensive operating periods (IOPs). The concentrations are those along the cross section D defined in Figure 2b (essentially a combination of the top portion of cross section A and bottom portion of B, shown in Figure 2a), calculated with a cubic-spline algorithm using all available measurements. Horizontal dashed lines indicate the latitude of Bethel Island (BTI), Fresno (FSF), and Bakersfield (BAC) sites.

[30] A major PM2.5 episode driven by pNH3NO3 occurred during IOP_3 (4–7 January 2001). From IOP_2 through IOP_3, the SJV was situated between a persistent high-pressure ridge over the Great Basin and a surface low off the southern California coast. On 4 January 2001, pNH3NO3 in the southern SJV was 75 μg m−3, and by 5 January 2001, this plume blanketed a broad region between BAC and FSF. The highest 24-hour average pNH4NO3 concentration of this episode (83 μg m−3) was measured at ANGI on 6 January 2001. For the first time during winter, M14 recorded a pNH4NO3 concentration approaching 60 μg m−3 (6–7 January 2001). While pNH4NO3 gradually dissipated in the southern SJV after 6 January 2001, the northern boundary BTI site reported a pNH4NO3 concentration of 40 μg m−3 on 7 January 2001 compared with ∼50 μg m−3 at FSF. The pNH4NO3 peak moved northward, consistent with the influence of regional transport.

[31] This episode is examined in greater detail in Figure 7 with subdaily pNH4NO3 concentrations at northern (FSF and BTI) and southern (BAC and ANGI) urban-rural pairs. All four sites showed a clear diurnal pattern with a midday peak caused by photochemical production and turbulent mixing, as suggested by Watson and Chow [2002a]. While there was virtually no increase in the 24-hour average pNH4NO3 concentration at either urban site, it increased throughout the IOP at both rural sites; occurring at ANGI on 6 January 2001 and at BTI on 6–7 January 2001. This again supports the conceptual model that urban precursors are transported aloft to rural locations, where they mix with nonurban NH3 emissions.

Figure 7.

Diurnal and 24-hour average total particulate ammonium nitrate (pNH4NO3) concentrations at (a) FSF and BTI and (b) ANGI and BAC during the 4–7 January 2001 IOP. Twenty-four-hour average concentrations are indicated by the dashed lines.

[32] Figure 8 presents the upper air wind structure measured with a wind profiler at ANGI. Wind speeds generally increased with altitude, and a sharp reversal of wind direction was observed during the evening of 5 January 2001 that explains the northward transport on 6–7 January 2001.

Figure 8.

Wind speed from the northeast to southwest as a function of altitude measured at the Angiola (ANGI) site during IOP_3, 4–6 January 2001. Positive component is flow from northwest to southeast.

[33] Twenty-four-hour average NOx and ozone (O3) concentrations, as well as temperature (T) and RH, at FSF and BAC during IOP_3 are compared with corresponding gas and PM concentrations in Table 3. The major difference between the two urban sites is higher pNH4NO3 concentrations at BAC, which is consistent with higher NOx and total ammonia (TNH3 = NH3 plus NH4+ as NH3 equivalent) at BAC. Low surface wind speeds minimize dispersion of primary PM2.5 components near their sources. Primary PM2.5 includes EC and OM from RWC, vehicle exhaust, and crustal material from resuspended dust. IOPs _2 and _3 experienced mean surface wind speeds of 0–2 m s−1 at FSF and a substantial accumulation of EC, especially on 1, 4, and 5 January 2001 (Figure 6). High EC concentrations also occurred at M14 and BAC. Soluble K+ concentrations were similar at FSF and BAC (Table 3). While EC and OM concentrations were about 50% higher at FSF, the EC/OM ratios at both FSF and BAC sites were nearly identical (0.23 ± 0.03 and 0.24 ± 0.02, respectively). By 8 January 2001, the surface low had advanced eastward, resulting in precipitation on 9 January 2001 that ended the episode. Both EC and pNH4NO3 concentrations decreased to less than 10 μg m−3 by 13 January 2001 at FSF and BAC.

Table 3. Twenty-Four-Hour Average Concentrations (μg m−3) at FSF and BAC During IOP3
SiteDateOMaECbK+cNH3dTNH3epNH4NO3fNOx,g ppbO3,h ppbT,i °CRH,j %
  • a

    Organic matter (OC × 1.4).

  • b

    Elemental carbon.

  • c

    Water-soluble potassium ion.

  • d

    Gaseous ammonia.

  • e

    Ammonia plus ammonium as ammonia equivalent.

  • f

    Total particulate ammonium nitrate (NH4NO3) = 1.29 × pNO3 (the sum of front-filter nonvolatilized nitrate [NO3] plus backup filter volatilized NO3).

  • g

    Nitrogen oxides.

  • h

    Ozone.

  • i

    Temperature.

  • j

    Relative humidity.

FSF4 Jan. 20013910.30.521625411718.27.271
BAC4 Jan. 2001328.50.5626427720012.77.374
FSF5 Jan. 2001429.70.5715244216510.86.869
BAC5 Jan. 2001296.40.661634801718.46.780
FSF6 Jan. 20015210.30.6216264712415.97.764
BAC6 Jan. 2001316.70.5922387513812.89.567
FSF7 Jan. 2001388.20.542130428514.49.352
BAC7 Jan. 2001225.50.432137729313.99.562

[34] Similar synoptic meteorology and PM2.5 spatial distribution patterns were observed during IOP_2 and IOP_4. Valley-wide exceedances of 65 μg m−3 (the value for the 98th percentile averaged over three years for the 24-hour PM2.5 NAAQS) were found at 26 sites during IOPs_2, 3, and 4. In rural areas, these high concentrations consisted mainly of pNH4NO3, while urban PM2.5 contained larger carbonaceous fractions.

6. Zones of Representation

[35] In designing an ambient air quality monitoring network, it is essential to know the extent to which community exposure monitoring sites represent their surroundings. The CRPAQS network included sites representing regional- (100–1000 km) (U.S. EPA, 1997), urban- (4–100 km), neighborhood- (0.5–4 km), middle- (0.1–0.5 km), and micro- (0.01 km–0.1 km) scales. The median distance between two neighboring sites in the CRPAQS network was ∼25 km, but some sites were located less than 1 km apart in order to contrast and compare different microenvironments [Watson and Chow, 2002a]. For example, a roadside site (FREM) and a residential site (FRES) were located ∼1 km west and ∼0.5 km east of the FSF.

[36] Most previous studies of community exposure consisted of two or three sites, typically locating one in the city center, and others in surrounding neighborhoods [e.g., Louie et al., 2005; Chow et al., 2006c]. In contrast, the CRPAQS network covered a much larger area containing many more urban and rural sites. The zones of representation can therefore be quantified by spatial interpolation of measured concentrations through the use of contour plots. This interpolation relies on the assumption of a smooth transition between each pair of neighboring sites. The validity of this hypothesis improves as the number (or density) of monitoring sites increases, because finer features of the spatial variability will be captured.

[37] The first step was to estimate concentration fields and to map the concentration gradients (e.g., Figure 2a). The zone of representation is defined as the radius of a circular area in which a species concentration varies by no more than ±20% as it extends outward from the center monitoring site. The criterion of 20% is chosen since it translates to ∼10% or less of the difference between the center site concentration and the average over the entire circular area, where 10% is on the order of the measurement uncertainty. The approach provides a quantitative comparison of spatial distributions between different species, seasons, and sites. Uncertainties resulting from inadequate spatial resolution may be mitigated by appropriate site selection based on county-level emission inventories [e.g., California Air Resources Board, 2004]. Since increasing the number of sites would likely reduce the zone coverage, this current approach gives the upper limit of the “true” zone sizes.

[38] Table 4 summarizes the zones of representation of 11 community exposure sites and two interbasin transport sites (BTI and ANGI) for annual, seasonal, and IOP periods. FSF represents different zones for annual pNH4NO3 (19 km) and EC (<1 km), consistent with the sources and meteorology at FSF. The major contrast among the three sampling sites in Fresno (FSF, FRES, FREM) and its suburban site (CLO, 7 km east) is due to carbonaceous material, probably from mobile sources, rather than pNH4NO3. Neighborhood-scale monitoring is needed to resolve the spatial variability of EC, and perhaps OM, in this urban area. Because it is located in downtown Bakersfield, the BAC site also has a small zone of representation for EC (∼1 km). The difference between BAC and its satellite site (BRES, 2 km west) is ∼40%. BTI is influenced by nearby cities, including San Francisco (SFA, 70 km west), Sacramento (S13, 50 km northeast), and Stockton (SOH, 30 km east). It represents a consistent zone (10–15 km) for PM2.5 mass, pNH4NO3, and EC. In general, the zone size for pNH4NO3 is more consistent than for EC across different sites. For annual PM2.5, BTI and ANGI represent areas with radii of 15 and 19 km, respectively, larger than the urban sites of FSF and BAC (7–11 km).

Table 4. Zone of Representation for Eleven Community Exposure and Two Interbasin Transport Sites in the San Joaquin Valley on Different Temporal Scalesa
Site CodebAnnual: Zone of Representation, kmChigh: High_PM2.5 (Nov. to Jan.): Zone of Representation, kmClow: Low_PM2.5 (Feb. to Oct.): Zone of Representation, kmEpisode (7 Jan. 2001): Zone of Representation, km
PM2.5pNH4NO3ECPM2.5pNH4NO3ECPM2.5pNH4NO3ECPM2.5pNH4NO3EC
  • a

    Sites are arranged from north to south; anchor sites are noted in bold.

  • b

    See Table 1 for site description.

S1312121210111154313761015
BTI151310141091516105109
M1422202319201725203461812
MRM19141415141327151411211
CLO610361039113544
FSF1119<11018<1131211128
SELM132191219820231211012
VCS23192821182328213014916
COP2223621276161665186
ANGI19194191941616312144
OLD6214151233578143
BAC7121432586654
MOP83203171585514110

[39] The zones of representation during the Chigh period were similar to those of the annual-average zones except at BAC, where the zone for pNH4NO3 was much smaller (3 km). Such a small zone reflects the facts that BAC is located at the southern boundary of the NH4NO3 plume and that NH4NO3 concentration decreases sharply to the south. This also explains the small zone of representation of MOP, which is an elevated site south of BAC. For the Clow period, the zone of representation of EC at MOP was close to the background level in neighboring rural areas and this site represented a zone with a radius of up to 55 km. Larger EC zones were also found at other community exposure sites during the Clow period; for example, VCS (30 km), M14 (34 km), and S13 (37 km). To some extent, this reflects less urban-rural contrast in summer with regard to EC concentration. Sharp gradients in EC concentration near rural ANGI and COP, however, warrant further investigation.

[40] The pNH4NO3 zone appeared to be consistent between the Clow and Chigh periods for most of the sites in the valley (Table 4). Pollutants exhibit more spatial inhomogeneity for a shorter averaging period, as expected.

[41] Factors affecting the zone size are complex. Rural sites do not necessarily have larger zones of representation than urban sites, because many of the rural sites in the SJV are close to the valley boundary, or located between two urban centers where concentration gradients are steep. Future ambient network design efforts should consider findings from this study to determine more optimal site location and density. For instance, more sites could be located near ANGI and COP (mainly to the west) to verify a consistently high EC gradient. Temporary satellite sites equipped with portable monitors are a cost effective means of acquiring this information.

7. Discussion and Conclusions

[42] CRPAQS shows that PM2.5 mass and NH4NO3 concentrations vary as a function of elevation, and suggests that vertical mixing is limited in the SJV to about 600 m above MSL. While widespread PM2.5 exceedances (up to 30 μg m−3) occurred in the southern SJV, PM2.5 concentration was at near-background levels toward the mountainous boundary.

[43] For most of the sites within the SJV, 50–75% of annual-average PM2.5 concentrations occurred from November to January. Elevated PM2.5 at nonurban sites was caused by high concentrations of NH4NO3. During winter, temperature, RH, and the stable valley boundary layer were favorable for the formation of NH4NO3 from its NH3 and HNO3 precursors. High EC and OM levels exacerbate air quality most severely at urban sites. This is consistent with the use of wood fuel for home heating in winter, as well as motor vehicle emissions. PM2.5 mass closure was typically >100% for samples with concentrations <15 μg m−3, likely reflecting a positive VOC sampling artifact on quartz-fiber filters. For samples with PM2.5 >20 μg m−3, however, this artifact was relatively less significant and the mass closure was close to, or less than, 100%.

[44] Lateral dispersion of the pollution plume became evident during winter episodes, an observation that supports the hypothesis of the role of upper layer currents on the valley-wide formation and transport of NH4NO3. Gradients of spatially interpolated concentrations show that most sites in the network appeared to represent zone sizes on an urban scale (4–100 km) or a neighborhood scale (0.5–4 km). In general, the annual NH4NO3 zone sizes across different sites were not the same as those for EC, consistent with a more inhomogeneous EC distribution in the valley. This information can be used to refine long-term monitoring networks in the future.

Acknowledgments

[45] This work was supported by the California Regional PM10/PM2.5 Air Quality Study (CRPAQS) Agency under the management of the California Air Resources Board and by the U.S. Environmental Protection Agency under contract R-82805701 for the Fresno Supersite.

Ancillary