Environ. Sci. Technol. 2002, 36, 4099-4106
Modeling the Effect of Weekday-Weekend Differences in Motor Vehicle Emissions on Photochemical Air Pollution in Central California LINSEY C. MARR AND ROBERT A. HARLEY* Department of Civil and Environmental Engineering, University of California, Berkeley, California 94720-1710
Ambient ozone concentrations vary by day of week in some locations, often with higher concentrations observed on weekends in urban and downwind areas. Emissions of ozone precursors appear to be lower on weekends, so the behavior of ozone concentrations on weekends may indicate the outcome of particular ozone control strategies. To examine the influence of day-of-week differences in motor vehicle emissions on ambient ozone concentrations, we combine a fuel-based motor vehicle emission inventory containing weekend-specific activity with an Eulerian photochemical airshed model applied to central California. Emissions of NOx on weekends are ∼30% lower than on weekdays due to a large drop in heavy-duty diesel truck activity, and emissions of VOC are only slightly lower on weekends. In rural areas, passenger car traffic and the associated emissions are highest on Fridays and Sundays. The combination of VOC sensitivity and reduced emissions of NOx on weekends results in higher ozone concentrations on weekends. Changes in the timing of emissions also contribute to the weekend ozone effect, but sensitivity tests show that changes in emissions timing have a minor effect compared to changes in total mass of emissions on weekends. Even in situations where reductions in NOx emissions lead to higher ozone concentrations, NOx reductions may still be necessary for control of other air pollutants such as nitrogen dioxide, nitric acid, and aerosol nitrate.
Introduction Numerous studies have found that ambient ozone concentrations tend to be higher on weekends than on weekdays in many urban areas, a phenomenon known as the “weekend effect”. Ozone is a respiratory irritant that has been associated with decreased lung function, increased hospital admissions for respiratory ailments, and the aggravation of asthma (1). Higher weekend ozone concentrations are of added concern because people spend more of their time outdoors on weekends: 9.3 versus 6.9% on weekdays (2). Thus a disproportionate fraction of ozone exposure may occur on weekends. Additionally, if the ozone standard is exceeded more frequently on weekends, then this fact needs to be considered in designing ozone control strategies. Ambient ozone has been found to be higher on weekends in many areas. Day-of-week differences in ozone concentra* Corresponding author e-mail
[email protected], phone (510) 643-9168, fax (510) 642-7483. 10.1021/es020629x CCC: $22.00 Published on Web 08/29/2002
2002 American Chemical Society
tions were first reported in the northeastern United States (3-5) and Los Angeles (6,7) in the 1970s and have since been noticed in St. Louis (8), Vancouver (9), San Francisco (10), and Switzerland (11). Higher weekend ozone tends to be found in urban centers, while lower weekend ozone is found in outlying areas. Marr and Harley (12) reported that the weekend effect, with higher ozone on weekends, has become more prevalent throughout California between 1980 and 2000. Emissions of volatile organic compounds (VOC) and oxides of nitrogen (NOx) are typically lower on weekends due to reduced anthropogenic activity, so the weekend effect provides an opportunity to test the effects of emission reductions that would take years to achieve through emission control strategies. This also provides a useful test case for evaluating the ability of a major planning toolsphotochemical air quality simulation modelssto predict accurately the effects of emission control programs. Identifying the major causes of the weekend effect may aid efforts to reduce ambient ozone levels. The weekend effect may be a useful tool to help identify VOC or NOx as the limiting factor in ozone formation. However, ozone control should not be considered in isolation from other air quality problems. Even when reductions in NOx emissions are counterproductive for ozone abatement, NOx reductions may benefit other pollutants of concern such as nitrogen dioxide, nitric acid, and aerosol nitrate. Agencies such as the California Air Resources Board (CARB) have recognized the problems and opportunities posed by the weekend effect and are investigating its occurrence and causes. CARB and other researchers have proposed several hypotheses to explain the weekend effect: (1) it may indicate VOC sensitivity of ozone formation in an area because NOx emissions typically decrease more than VOC emissions on weekends (10); (2) the different timing of emissions on weekends may affect ozone formation, either due to carryover of Friday and Saturday nighttime emissions or due to the lack of a morning rush hour on weekends (13); (3) reductions in emissions of fine particles, specifically soot, from heavy-duty diesel trucks may contribute to higher ultraviolet radiation levels and higher ozone concentrations on weekends (14); and (4) different spatial patterns of emissions on weekends, less concentrated near urban cores and enhanced in outlying areas, may also affect ozone formation. Existing gridded inventories of motor vehicle emissions used as input to photochemical models typically assume weekday patterns of driving and do not separate passenger cars from heavy-duty diesel trucks as sources of NOx emissions. Models that predict motor vehicle emissions often rely on travel-demand models to apportion emissions spatially and temporally. These models use population and employment data to predict commuter trips between home and the workplace, but such trips are greatly reduced on weekends. Furthermore, because heavy-duty diesel trucks comprise a small fraction of total vehicle counts, little effort has been devoted to representing spatial and temporal patterns of activity and emissions for these vehicles. Diesel trucks are responsible for a significant fraction of NOx emissions (15), and their travel patterns and emissions should be treated separately from those of passenger cars. Previous modeling studies of ozone formation have been limited by the lack of accurate estimates of motor vehicle emissions on weekends. For example, the 27-29 August 1987 episode of the Southern California Air Quality Study, which has been modeled extensively, spans a Thursday through Saturday, and modelers either avoid including the Saturday VOL. 36, NO. 19, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4099
FIGURE 1. Central California modeling domain, showing topography and the location of major cities and geographic features. in their studies (16, 17) or proceed using weekday traffic patterns on the Saturday of the episode (18). A study that modeled ozone concentrations over an entire year in the Los Angeles area relied on weekday traffic patterns to describe motor vehicle emissions on all days (19). The 3-6 August 1990 episode of the San Joaquin Valley Air Quality Study (SARMAP) spans a Friday through a Monday. The gridded emission inventory for this period (20) uses the diurnal patterns of passenger car traffic on weekdays to describe activity in all vehicle categories, including heavy-duty diesel trucks, on both weekdays and weekends. This gridded inventory has been widely used in photochemical modeling studies (21-23). Because motor vehicles account for a large portion of emissions of ozone precursors and have significantly altered activity patterns on weekends, vehicles are likely to play an important role in the weekend ozone effect. Lack of a weekend-specific emission inventory has hampered threedimensional photochemical modeling of the weekend effect. In previous work, we developed a motor vehicle emission inventory with hourly traffic patterns by day of week and vehicle class (24). The objective of this research is to identify factors in the motor vehicle emission inventory that contribute to higher weekend ozone. To address this question, we apply an Eulerian photochemical airshed model to central California for a four-day modeling episode that spans a weekend.
Methods Model Description. The modeling domain, shown in Figure 1, covers 240 km × 240 km in central California and includes the San Francisco Bay Area, Sacramento, and the northern San Joaquin Valley. The Sacramento and San Joaquin Valleys together comprise the Central Valley. We use an updated version of the CIT Eulerian photochemical airshed model (17) to predict concentrations of ozone and other species. The model solves equations describing transport and chemistry in each 4 km × 4 km grid cell, whose vertical thicknesses range from 25 m at the lowest layer to 455 m at the fourteenth and highest layer. The modeling domain reaches a height of 2173 m above ground level. The SAPRC99 mechanism (25) describes 227 chemical reactions involving 15 explicit and 15 lumped VOC species in the model. Model Application. We apply the air quality model to an intensive measurement period from a regional-scale air 4100
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 19, 2002
quality study conducted in central California (26). During the study, numerous air quality and meteorological measurements were made at the surface and aloft to support photochemical modeling, and an emission inventory was developed for the period 3-6 August 1990, which spans a Friday through a Monday. The photochemical model runs over these four days, plus an 8-h spin-up period starting on the afternoon of 2 August. During the last two days of the episode, meteorological conditions were especially favorable for ozone formation, with higher than normal temperatures; ambient ozone concentrations as high as 160 ppb were measured on 6 August. Marr et al. (27) described in detail the modeling system and evaluation of its performance in predicting ozone and precursor concentrations for this episode. Revision of the original motor vehicle and biogenic emission inventories, which appeared to contain compensating errors in VOC, resulted in improved ozone and VOC predictions and an increase in model bias for NOx and its oxidation products (NOy). Average normalized biases over the modeling episode ranged from 1 to 11% for ozone, -12 to 7% for non-methane hydrocarbons (NMHC), and 16 to 35% for NOy. Emission Inventory. Estimates of emissions for the 3-6 August 1990 period were developed by Dickson et al. (20) using a customized emissions modeling system. We maintain the original estimates of area and point source emissions and replace the original biogenic and motor vehicle emission inventories with updated estimates (27). The inventory includes emissions of carbon monoxide (CO), organic gases, NOx, sulfur oxides, and particulate matter by hour on a 4 km × 4 km grid. In the original motor vehicle emission inventory, emissions of CO, NMOC, and NOx are all ∼20% lower on weekends, and the diurnal pattern of emissions is identical on weekends and weekdays. Also, the inventory lumps together all motor vehicle NOx emissions; it does not distinguish between passenger cars and heavy-duty diesel trucks. Marr and Harley (12) presented the original and revised emission inventories for the modeling domain for each day of the modeling episode. We use a revised motor vehicle inventory based on traffic counts, fuel sales data, and on-road measurements of vehicle emissions (24). Traffic counts by hour, day of week, and vehicle class were collected from three urban and four rural weigh-in-motion sensors located at highway sites throughout central California (24). These sensors provide separate counts for light-duty passenger cars (mostly gasoline-powered) and heavy-duty diesel trucks by hour and day of week. We designate each county in the modeling domain as rural or urban depending on population (>200 000 is urban) and assign the county’s running (but not resting or diurnal evaporative) vehicle emissions the appropriate weekly and diurnal patterns. Compared to the original motor vehicle emission inventory, these changes result in higher VOC emissions; a smoother, unimodal distribution of emissions on weekends; and lower NOx emissions on all days. When diesel trucks are treated separately from passenger cars, the resulting diurnal pattern of NOx emissions is smoother due to the increased contribution from heavy-duty diesel trucks during noncommute hours (24). Sensitivity Studies. The goal of this study is to quantify the impacts of changes in the mass and timing of on-road vehicle emissions on the formation of photochemical air pollution. We consider a base case where the total mass and timing of vehicle emissions are represented using typical weekday (Monday through Thursday) traffic data, and alternate scenarios where more realistic estimates of vehicle emissions specific to the days of the modeling episodes Friday, Saturday, Sunday, and Mondaysare used. We examine separately the effects of changes in the timing, i.e., diurnal distribution, of emissions and changes in the total
mass of emissions for typical weekday versus weekend conditions. We then consider the combined effect of both changes. In summary, we conduct four simulations: weekday mass with weekday timing, weekday mass with day-specific timing, day-specific mass with weekday timing, and dayspecific mass with day-specific timing. All air quality model simulations begin at 4 p.m. on Thursday, 2 August 1990 and continue through Monday, 6 August. All simulations use the same meteorological input fields, appropriate to the specific days being studied here. In all scenarios, we use the same day-specific estimates of stationary and biogenic emissions. We also identify VOCand NOx-sensitive regions of the domain by assessing the impact of a 25% reduction in NOx emissions. In areas where ozone formation is VOC-sensitive, we expect predicted ozone concentrations to increase when NOx emissions are reduced.
Results and Discussion Emission Inventory. The key differences between weekday and weekend on-road vehicle emissions are in magnitude and timing. Emissions of both VOC and NOx are lower on the weekend in urban areas, but the weekday-weekend difference in NOx is larger. In rural areas, light-duty traffic is actually higher on weekends, but the contribution from these areas to total vehicle emissions is low. Integrated over the entire modeling domain, motor vehicle VOC and NOx emissions are ∼5 and ∼30% lower, respectively, on weekends compared to weekdays. The differences are due mainly to differences in urban traffic: a 5-10% decrease in light-duty gasolinepowered traffic on weekends and a much larger 70-80% decrease in heavy-duty diesel truck traffic. The temporal distribution of emissions relies on seven weigh-in-motion traffic sites, all on highways. Traffic patterns on surface streets and at finer spatial resolution than countywide would be useful to describe traffic patterns with greater accuracy. Nevertheless, these data represent a significant improvement over previous temporal descriptions of motor vehicle emissions, which assigned all emissions the diurnal pattern of light-duty traffic on weekdays. Diurnal patterns of motor vehicle emissions are driven by traffic patterns for light- and heavy-duty vehicles and the fact that while light-duty vehicles are responsible for almost all motor vehicle VOC emissions, heavy-duty diesel trucks are an important contributor to NOx emissions (15). Figure 2 shows traffic patterns for light- and heavy-duty vehicles in urban and rural areas. The figure presents the fraction of traffic counts at each hour to the weekly total for Monday through Thursday, Friday, Saturday, and Sunday. Motor vehicle VOC emissions on weekends versus weekdays differ more in their diurnal pattern than in their total mass. On weekdays, motor vehicle VOC emissions, which are dominated by light-duty vehicles, are bimodal due to morning and afternoon rush hour traffic; and on weekends, emissions plateau between late morning and early evening. The diurnal pattern of NOx emissions varies less by day of week, although emissions are shifted ∼2 h later on weekends. On weekdays, emissions of NOx are high during the morning rush hour and remain elevated during the middle of the day because heavy-duty diesel truck traffic is highest during these hours. On weekends, passenger cars dominate NOx emissions, so the diurnal pattern of emissions more closely follows weekend traffic patterns of these vehicles, with a single peak in the afternoon. For both VOC and NOx, there is a notable lack of emissions between 6 and 8 a.m. on weekends compared to weekdays. Figure 3 shows the emission inventories of VOC and NOx for the modeling episode using both the weekday and dayspecific traffic patterns (see also Table 1). Day-specific emission inventories for area, point, and biogenic sources were developed for the meteorology of the modeling episode.
FIGURE 2. Traffic patterns for light-duty passenger cars (solid line) and heavy-duty diesel truck (dashed line) traffic in urban and rural areas, shown as the ratio of traffic counts by hour to the weekly total traffic count. Off-road mobile sources are included with area-wide sources. Area and point source VOC emissions are 10-15% lower on Saturday and Sunday compared to Friday and Monday. Biogenic VOC emissions rise over the course of the four-day modeling episode due to rising ambient temperatures. Area and point source NOx emissions (no biogenic sources) are ∼30% lower on Saturday and Sunday than on Friday and Monday. With day-specific instead of weekday traffic patterns, on-road motor vehicle VOC emissions are higher on Friday (traffic is higher on Fridays than on Mondays through Thursdays) and slightly lower on Sunday. The two totals for Monday do not match exactly because of approximations made in designating counties as urban or rural. For NOx, the use of day-specific rather than weekday traffic patterns leads to higher vehicle emissions on Friday and 20-30% lower vehicle emissions on Saturday and Sunday. Modeling. Applying day-specific, rather than typical weekday, motor vehicle emissions to the Friday through Monday modeling episode results in changes in predicted pollutant concentrations. Figure 4 contrasts observed and modeled concentrations using the weekday and day-specific motor vehicle emission inventories. The time series show CO, NMHC, NOy* (NOx plus oxidation products of NOx except for nitric acid), and ozone concentrations in Redwood City, near the south end of the San Francisco Bay. We compare predicted NOy* concentrations to observed “NOx” concentrations because conventional NOx analyzers also measure other nitrogen-containing species as NOx. Because nitric acid deposits easily on sampling lines (28), it is excluded from NOy*. Also note that CO and NMHC concentrations are measured twice per day and hourly NOy* and ozone concentrations are reported to the nearest 10 ppb. VOL. 36, NO. 19, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4101
TABLE 1. Ratios of Traffic Counts to the Weekly Average for Light-Duty Passenger Cars and Heavy-Duty Diesel Trucks in Urban and Rural Areas traffic type light-dutya,b
urban urban heavy-duty rural light-dutyb,c rural heavy-duty
Mon-Thu
Fri
Sat
Sun
1.01 ( 0.02 1.29 ( 0.01 0.87 ( 0.07 1.24 ( 0.07
1.07 ( 0.03 1.25 ( 0.04 1.20 ( 0.09 1.13 ( 0.08
1.00 ( 0.04 0.37 ( 0.05 1.10 ( 0.06 0.53 ( 0.14
0.90 ( 0.07 0.23 ( 0.05 1.21 ( 0.15 0.39 ( 0.13
a Average ratio ( one standard deviation from three weigh-in-motion traffic count sites located in urban areas of central California. These ratios were used to scale vehicle emissions by day of week in the following counties: Alameda, Contra Costa, Fresno, Marin, Monterey, Sacramento, San Francisco, San Joaquin, San Mateo, Santa Clara, Santa Cruz, Solano, Sonoma, and Stanislaus. b “Light-duty” includes cars, pickup trucks, and sport utility vehicles. c Average ( one standard deviation from four weigh-in-motion traffic count sites located in rural areas of central California. These ratios were used to scale vehicle emissions by day of week in the following counties: Amador, Calaveras, Colusa, El Dorado, Lake, Madera, Mariposa, Mendocino, Merced, Napa, Nevada, Placer, San Benito, Sutter, Tuolumne, Yolo, and Yuba.
FIGURE 3. Emissions of (a) VOC and (b) NOx in the modeling domain for each day of the modeling episode. In the weekday scenario, motor vehicle exhaust emissions are held at the Monday-Thursday average value for each day of the episode, with some variation in vehicle-related evaporative VOC emissions due to rising temperatures over the four-day episode. In the day-specific scenario, motor vehicle exhaust emissions are scaled by day of week according to the traffic data shown in Table 1. Day-specific emissions from area, point, and biogenic sources have been developed specifically for this episode and are unchanged in the two scenarios.
FIGURE 4. Concentration time series of CO, NMHC, NOy, and ozone at Redwood City using weekday and day-specific motor vehicle emission inventories. The modeling period begins on the afternoon of Thursday, 2 August 1990 and ends on Monday, 6 August 1990.
As shown in Figure 4, use of the day-specific emissions estimates results in slightly higher concentrations of the primary pollutants CO, NMHC, and NOy* on Friday, particularly after 6 p.m. In both urban and rural areas, lightduty traffic is higher on Friday than on the other weekdays, especially in the evening. On Saturday and Sunday, use of weekday emissions estimates results in a morning rush hour peak in predicted primary pollutant concentrations that is not reflected in the observed concentrations. In contrast, use of the more appropriate day-specific inventory estimates results in a closer match between predicted and observed concentrations. The overprediction of primary pollutant concentrations and underprediction of ozone concentrations on Friday and Saturday nights probably result from low nighttime mixing depths in the model and titration of ozone by nitric oxide in the model. Nighttime predictions may also be worse due to a higher degree of spatial heterogeneity in
the concentration field because of more stable atmospheric conditions. For ozone, concentrations predicted using day-specific emissions are a few ppb higher on weekends at Redwood City. In general, the model overpredicts daytime ozone concentrations at this site. Throughout the modeling domain, the use of day-specific rather than typical weekday emissions results in higher predicted peak ozone concentrations in the San Francisco Bay Area, unchanged concentrations in Sacramento, and lower concentrations in much of the San Joaquin Valley and Sierra Nevada foothills. For this period, ozone formation is generally VOC-sensitive in the San Francisco Bay Area and immediately downwind areas and NOx-sensitive in the remainder of the modeling domain. Figure 5 shows predicted ozone and NO2 concentrations on Friday, 3 August 1990 at 4 p.m. and the difference in predicted concentrations when emissions of NOx are
4102
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 19, 2002
FIGURE 5. (a) Predicted ozone concentrations (ppb) on Friday, 3 August 1990 at 4 p.m. and (b) percent change in predicted ozone concentrations when emissions of NOx are reduced by 25%. (c) Predicted NO2 concentrations (ppb) at the same time and (d) percent change in predicted NO2 concentrations when emissions of NOx are reduced by 25%. reduced by 25%, an amount in magnitude comparable to weekday-weekend differences in emissions. Figure 5a shows that ozone concentrations are lowest along the coast and increase eastward, where temperatures are warmer and both locally emitted and transported anthropogenic emissions are influential. The highest concentrations occur in the foothills east and downwind of Sacramento. A 25% reduction in NOx emissions results in predicted ozone concentrations, shown in Figure 5b, that are up to 25%, or ∼20 ppb, higher in areas downwind of the San Francisco Bay Area, near Livermore and San Jose, and suggests that ozone formation in these areas is VOC-sensitive. Predicted ozone levels in most of the remaining areas of the modeling domain, except for a small spot east of Sacramento, are lower, suggesting that ozone formation in these areas is NOx-sensitive on this day. Results in the northwest corner of the modeling domain are suspect due to a couple of grid squares containing extremely high emissions of NOx from motor vehicles in the original inventory.
Although reduced NOx emissions result in higher predicted ozone concentrations, the reduced emissions result in lower predicted concentrations of other pollutants, such as NO2. Figure 5d shows the change in predicted concentrations of NO2, relative to Figure 5c, when NOx emissions are reduced by 25%. Predicted NO2 concentrations decrease by 25% or more over most of the modeling domain. Increases in NOx emissions should not be considered as an ozone control strategy because this could exacerbate other air pollution problems including NO2, nitric acid, and aerosol nitrate, with adverse consequences for public health, visibility, and acid deposition. An observation-based study of the weekend ozone effect using spectral analysis of ambient ozone concentration time series (12) shows that the locations of monitoring sites with a significant weekend effect in 1990-1994 roughly correspond to the area showing VOC sensitivity in Figure 5b. All monitoring sites in the southeast San Francisco Bay Area show a significant weekend effect, while all but one in VOL. 36, NO. 19, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4103
FIGURE 6. (a) Predicted ozone (ppb) at 4 p.m. on Sunday, 5 August using motor vehicle emissions for average weekday (Monday-Thursday) conditions. Change in predicted ozone with (b) day-specific mass and weekday timing, (c) day-specific timing and weekday mass, and (d) day-specific mass and day-specific timing of motor vehicle emissions. Sacramento do not. A few sites in the northern San Francisco Bay Area and in the region between it and the Central Valley do not display a weekend effect, and two of three sites near Stockton show a significant weekend effect. In this case, a relationship exists between areas where weekend ozone concentrations are higher and ozone formation is VOCsensitive. The response of predicted ozone concentrations to dayof-week differences in the mass and timing of motor vehicle emissions varies by location, and differences in mass have a larger effect on ozone formation. Figure 6 depicts the results of sensitivity studies conducted to understand better the weekend effect. This figure shows predicted ozone concentrations using all possible combinations of typical weekday versus day-specific (Friday and weekend) mass and timing of motor vehicle emissions. Figure 6a shows predicted ozone concentrations on Sunday, 5 August at 4 p.m., the hour of peak ozone over the weekend, using typical weekday emissions. The remaining maps in Figure 6 show the differences in predicted ozone from the typical weekday base case. The case of motor vehicle emissions with weekend mass and 4104
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 19, 2002
weekday timing, shown in Figure 6b, results in higher ozone, up to a 10 ppb difference, in areas southeast and downwind of the San Francisco Bay Area and lower ozone in the Central Valley. Compared to the case using a weekday mass of motor vehicle emissions, total emissions of NOx and NMOC are 21 and 4% lower, respectively, on Sunday. Not surprisingly, therefore, these results are similar to those shown in Figure 5, which shows the response of predicted ozone to a 25% reduction in all NOx emissions. One notable difference between Figures 5b and 6b is the change in predicted ozone concentrations in the northern San Joaquin Valley, near Stockton. Figure 5b shows the changes in predicted peak ozone concentrations on Friday when all NOx emissions are reduced by 25%; predicted ozone concentrations in Stockton increase by over 15%, or 10 ppb. Figure 6b shows the changes in predicted peak ozone concentrations on Sunday when NOx emissions are reduced by ∼20% and VOC emissions by ∼5%, the difference between weekday and weekend emissions; predicted ozone concentrations in Stockton decrease, suggesting that ozone formation in the area is NOx-sensitive. We expect a similar response
FIGURE 7. Time series of predicted ozone on Friday, 3 August through Monday, 6 August at three sites for the following motor vehicle emission scenarios: day-specific mass and day-specific timing (DSMDST, solid line), weekday mass and day-specific timing (WDMDST, dot-dashed line), day-specific mass and weekday timing (DSMWDT, dashed line), and weekday mass and weekday timing (WDMWDT, dotted line). in both cases, but previous work has shown ozone sensitivity in the northern San Joaquin Valley to vary with the meteorological and emissions conditions of a particular day (27). On Sunday compared to Friday, area and point source emissions of NOx are lower because of reduced anthropogenic activity, and biogenic VOC emissions are higher because of higher temperatures. Therefore, because of the higher VOC/ NOx ratio, ozone formation on Sunday is more likely to be NOx-sensitive, and substitution of weekend-specific motor vehicle emissions results in lower predicted ozone concentrations in this area. Figure 6c shows the differences in predicted ozone using weekday mass and weekend timing of motor vehicle emissions. The differences are small, except for the area southeast of Stockton, where weekend timing leads to >6 ppb decreases in predicted ozone concentrations. Figure 6d shows the differences in predicted ozone using motor vehicle emissions with weekend mass and weekend timing. Predicted ozone concentrations are higher in the area downwind of the San Francisco Bay Area and lower in the Central Valley. Figure 6d is similar to the case using weekend mass and weekday timing (Figure 6b), implying that changes in the mass of vehicle emissions are more important than changes in timing in explaining weekday-weekend differences in ozone con-
centrations. The magnitude of the differences in predicted ozone concentrations, up to 10 ppb, is consistent with analyses of weekday-weekend differences in peak concentrations in the San Francisco Bay Area and Sacramento in the early 1990s (10, 29). Figure 7 shows how predicted ozone concentrations vary over the course of the modeling episode under the test of day-specific versus weekday mass and timing of motor vehicle emissions. The figure shows time series of observed and predicted concentrations for each of the four emission scenarios at San Jose, Sacramento, and Stockton. The locations of these sites, chosen to represent the variety of responses of ozone to day-of-week differences in motor vehicle emissions, are shown in Figure 6. On the first day of the modeling episode, a Friday, the four cases are virtually indistinguishable at all sites because weekday average (Monday through Thursday) patterns of motor vehicle emissions are not substantially different from those on Fridays. On Saturday and Sunday at San Jose, using dayspecific emissions results in peak ozone concentrations that are 5-10 ppb higher than the cases using weekday mass emissions. Day-of-week differences in the timing of emissionssboth the lack of a morning rush hour and higher latenight emissions on weekendssdo not have a large effect on VOL. 36, NO. 19, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4105
predicted ozone concentrations, as the time series for cases using the same mass but different timing are nearly coincident. Differences in the timing of emissions affects nighttime carryover of ground-level ozone on Sunday night to Monday morning, but predicted concentrations of ozone on Monday after 6 a.m. are not affected by the nighttime differences. At Sacramento on Saturday and Sunday, peak daytime ozone concentrations are similar for all cases, but night and morning time concentrations differ slightly. The largest differences occur on Sunday between midnight and 6 a.m. PST. During these hours, predicted ozone is highest using day-specific mass and weekday timing of emissions because this combination results in the lowest NO emissions in urban areas and therefore the least titration of ozone. Sunday lightand heavy-duty traffic is light, and nighttime traffic is lighter under weekday conditions. Although predicted ozone concentrations at 6 a.m. are several ppb higher under these conditions, predicted concentrations in the other cases quickly catch up with them. At Stockton on Saturday and Sunday, the day-specific mass and timing of emissions results in lower peak ozone concentrations compared to the weekday case, and again, the differences in mass have a larger effect than the differences in timing on predicted ozone concentrations. Larger decreases in NOx than in VOC emissions on weekends result in higher ozone levels on weekends in areas of central California where ozone formation is VOC-sensitive. The large weekday-weekend differences in NOx emissions are due mainly to the 70-80% drop in heavy-duty diesel truck traffic on weekends. Sensitivity tests, in which predicted ozone concentrations are compared under different emissions scenarios using a photochemical model, confirm this hypothesis. Results also show that day-of-week differences in the mass of motor vehicle emissions play a much larger role in influencing ozone formation on weekends than do differences in the timing of the emissions.
(3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20)
(21)
(22) (23) (24)
Acknowledgments
(25)
The authors thank Joe Avis of the California Department of Transportation for providing traffic count data, Vernon Hughes of the California Air Resources Board for providing emission inventory data, the California Air Resources Board’s weekend effect workgroup for helpful discussions, and Jana Milford of the University of Colorado for assistance with the chemical mechanism. This research was supported by a United States Environmental Protection Agency STAR Graduate Research Fellowship.
(26)
Literature Cited
(29)
(1) Bates, D. V. In Air Pollution and Health; Holgate, S. T., Samet, J. M., Koren, H. S., Maynard, R. L., Eds.; Academic Press: San Diego, 1999; pp 1-4. (2) Klepeis, N. E.; Tsang, A. M.; Behar, J. V. Analysis of the National Human Activity Pattern Study (NHAPS) respondents from a standpoint of exposure assessment; U.S. Environmental Protection Agency, Office of Research and Development, National
4106
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 19, 2002
(27) (28)
Exposure Research Laboratory: Washington, DC, 1996; EPA/ 600/R-96/148. Lebron, F. Atmos. Environ. 1975, 9, 861-863. Cleveland, W. S.; McRae, J. E. Environ. Sci. Technol. 1978, 12, 558-563. Cleveland, W. S.; Graedel, T. E.; Kleiner, B.; Warner, J. L. Science 1974, 186, 1037-1038. Levitt, S. B.; Chock, D. P. J. Air Pollut. Control Assoc. 1976, 26, 1091-1092. Elkus, B.; Wilson, K. R. Atmos. Environ. 1977, 11, 509-515. Karl, T. R. Atmos. Environ. 1978, 12, 1657-1667. Pryor, S. C.; Steyn, D. G. Atmos. Environ. 1995, 29, 1007-1019. Altshuler, S. L.; Arcado, T. D.; Lawson, D. R. J. Air Waste Manage. Assoc. 1995, 45, 967-972. Bronnimann, S.; Neu, U. Atmos. Environ. 1997, 31, 1127-1135. Marr, L. C.; Harley, R. A. Atmos. Environ. 2002, 36, 2327-2335. Blanchard, C. L.; Fairley, D. Atmos. Environ. 2001, 35, 38613873. Dreher, D. B.; Harley, R. A. J. Air Waste Manage. Assoc. 1998, 48, 352-358. Sawyer, R. F.; Harley, R. A.; Cadle, S. H.; Norbeck, J. M.; Slott, R.; Bravo, H. A. Atmos. Environ. 2000, 34, 2161-2181. Lu, R.; Turco, R. P.; Jacobson, M. Z. J. Geophys. Res. 1997, 102, 6081-6098. Harley, R. A.; Sawyer, R. F.; Milford, J. B. Environ. Sci. Technol. 1997, 31, 2829-2839. Harley, R. A.; Russell, A. G.; McRae, G. J.; Cass, G. R.; Seinfeld, J. H. Environ. Sci. Technol. 1993, 27, 378-388. Winner, D. A.; Cass, G. R. Atmos. Environ. 1999, 33, 431-451. Dickson, R. J.; Wilkinson, J. G.; Bruckman, L.; Tesche, T. W. In Planning and Managing Regional Air Quality: Modeling and Measurement Studies; Solomon, P. A., Ed.; Lewis Publishers: Boca Raton, FL, 1994; pp 79-106. DaMassa, J.; Tanrikulu, S.; Magliano, K.; Ranzieri, A. J.; Niccum, L. Performance Evaluation of SAQM in Central California and Attainment Demonstration for the August 3-6, 1990 Ozone Episode; Technical Support Division, California Air Resources Board: Sacramento, CA, 1996. Dabdub, D.; DeHaan, L. L.; Seinfeld, J. H. Atmos. Environ. 1999, 33, 2501-2514. Jacobson, M. Z. J. Geophys. Res. 2001, 106, 5403-5420. Marr, L. C.; Black, D. R.; Harley, R. A. J. Geophys. Res. 2002, 107 (D6), ACH 5-1-5-9, 10.1029/2001JD000689. Carter, W. P. L. Documentation of the SAPRC-99 chemical mechanism for VOC reactivity assessment; Report to the California Air Resources Board; University of California, Riverside, College of Engineering, Center for Environmental Research and Technology: Sacramento, CA, 2000; Contracts 92-329 and 95308. Ranzieri, A. J.; Thuillier, R. H. In Planning and Managing Regional Air Quality: Modeling and Measurement Studies; Solomon, P. A., Ed.; Lewis Publishers: Boca Raton, FL, 1994; pp 1-28. Marr, L. C.; Noblet, G. S.; Harley, R. A. J. Geophys. Res. 2002, 107(D6), ACH 6-1-6-11, 10.1029/2001JD000690. Blanchard, C. L.; Lurmann, F. W.; Roth, P. M.; Jeffries, H. E.; Korc, M. Atmos. Environ. 1999, 33, 369-381. Austin, J.; Tran, H. In Proceedings of the Seventh International Conference on Air Pollution; Air Pollution VII; Wessex Institute of Technology Press: Southampton, U.K., 1999; pp 645-661.
Received for review March 4, 2002. Revised manuscript received July 11, 2002. Accepted July 17, 2002. ES020629X
