Environ. Sci. Technol. 2003, 37, 5236-5246
Regional Modeling of the Atmospheric Fate and Transport of Benzene and Diesel Particles CHRISTIAN SEIGNEUR,* BETTY PUN, KRISTEN LOHMAN, AND SHIANG-YUH WU Atmospheric & Environmental Research, Inc., 2682 Bishop Drive, Suite 120, San Ramon, California 94583
The Community Multiscale Air Quality model (CMAQ) was modified to simulate the atmospheric fate and transport of benzene and diesel particles. We simulated the July 1115, 1995 period over a domain covering the eastern United States with a 12-km horizontal resolution and a finer (4 km) resolution over a part of the northeastern United States that includes Washington, DC and New York City. The meteorological fields were obtained from a simulation conducted earlier with the mesoscale model MM5. Gridded emission files for benzene and diesel particles were developed using the SMOKE modeling system. The results of the model simulations showed that benzene concentrations were commensurate with available measurements. Over the 4-km resolution domain, a comparison between simulated and measured 24-h average concentrations showed a fractional error of 0.46, a fractional bias of 0.14, and a coefficient of determination (r2) of 0.25. A comparison between simulated benzene hourly concentrations in New York City and in the Brigantine Wilderness Area, NJ, showed that urban concentrations were greater than the remote area concentrations by a factor of 2-5. The results of the diesel particle simulations showed spatial and temporal patterns that were similar to those obtained for benzene. However, because of the lesser contribution of on-road mobile sources to diesel particle emissions compared to benzene emissions, diesel particle concentrations showed stronger gradients between urban areas and remote areas. A comparison between diesel particle concentrations in New York City and in the Brigantine Wilderness Area, NJ, showed that the urban concentrations were greater than the remote area concentrations by a factor of 2-10. Assuming that diesel particles consist of 50% “elemental” carbon (EC), the simulated EC concentrations were in close agreement (within 10%) with the measured concentration in the urban area (Washington, DC) but were significantly lower than the measured EC concentrations in the remote area (Brigantine Wilderness Area). This result suggests that other sources beside diesel fuel engines contribute to atmospheric EC concentrations and that EC may not be a good surrogate for diesel particles. A comparison of both benzene and diesel particle simulated concentrations between an urban area (New York City) and a remote area (Brigantine Wilderness Area) shows that, at a spatial resolution of 4 km, the regional background may contribute from 10 to 20% * Corresponding author phone: (925)244-7121; fax: (925)244-7129; e-mail:
[email protected]. 5236
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 22, 2003
to the peak concentrations. These results suggest that the regional background may not be negligible and should be taken into account in urban air toxics studies.
1. Introduction Hazardous air pollutants (HAPs), also known as air toxics, are regulated under the Clean Air Act (Section 112) and are currently being studied by the U.S. Environmental Protection Agency (EPA) and several state agencies in the United States. Those studies focus primarily on the urban-scale impacts of air toxics. However, some of those air toxics, such as gaseous species with low reactivity and fine particles, have long atmospheric lifetimes (e.g., several days), and they may be transported over long distances. It is, therefore, of interest to investigate their atmospheric fate and transport at regional scales. Previous studies have investigated the atmospheric fate and transport of some air toxics. Mercury, a global pollutant, has been the subject of several atmospheric modeling studies at global (1, 2) and regional (2-4) scales. Dioxins, a product of the combustion of chlorinated compounds, have also been investigated with a regional scale model (5). Atrazine, a pesticide, has been modeled with a regional scale model (6, 7). Mobile sources emit a variety of air toxics including gaseous compounds (e.g., benzene, formaldehyde, and 1,3butadiene) and particulate compounds (e.g., particles from diesel exhausts). Those air toxics are also emitted by other sources. The impacts at urban scales are currently being investigated by EPA under the integrated urban air toxics strategy and by some states (e.g., the Arizona Department of Environmental Quality study) and local agencies (e.g., the South Coast Air Quality Management District’s MATES-II study in the Los Angeles basin). However, regional scale impacts have not yet been investigated although they may be significant for those air toxics that have a long atmospheric lifetime (e.g., benzene and particles from diesel combustion sources). We present here a case study for two air toxics that are emitted by motor vehicles among other sources. These two air toxics are benzene and particles from diesel combustion sources (hereafter referred to as diesel particles). EPA has listed benzene as a known human carcinogen that can lead to a variety of health effects including leukemia. It is a volatile organic compound (VOC) which has low reactivity. It is oxidized in the atmosphere slowly by hydroxyl (OH) radicals; its atmospheric lifetime is on the order of 1 week. EPA has listed diesel exhaust as likely to be carcinogenic with the end point being lung cancer. They are emitted from diesel engines with a particle mass size distribution that is centered around 0.1-0.4 µm in diameter (8-10). These fine particles are subjected to coagulation processes and condensation of secondary aerosols (sulfate, nitrate, ammonium, organic compounds) and grow into the accumulation mode (i.e., in the range of 0.1-1 µm in diameter) as they age. A diesel particle initially consists of an agglomeration of “elemental” carbon (EC) spheres coated with organic and inorganic compounds that are adsorbed or absorbed at the surface of this agglomerate. The organic compounds include 3-ring to 6-ring polycyclic aromatic hydrocarbons (PAH); the inorganic compounds include sulfate and nitrate (11). The composition of diesel particles is variable. It consists primarily of carbonaceous compounds with EC typically accounting for 2560% (10, 12-16), with estimates ranging from 5 to 90% (17) 10.1021/es034433o CCC: $25.00
2003 American Chemical Society Published on Web 10/17/2003
TABLE 1. Emissions of Benzene by Major Source Categories (Source: 1996 NTI) source categorya point sources light-duty gasoline vehicles light-duty gasoline trucks heavy-duty gasoline vehicles light-duty diesel vehicles light-duty diesel trucks heavy-duty diesel vehicles motorcycles 2-stroke gasoline equipment 4-stroke gasoline equipment compression ignition equipment aircraft storage and transportc fuel combustion industrial processes waste disposal miscellaneous area sources total
FIGURE 1. Modeling domains for the NARSTO-Northeast simulation. and “organic” carbon (OC) typically accounting for 20-50% (15, 16) of total mass. (Some unpublished recent results suggest that the EC fraction may be higher, in the range of 61-76%.) Sulfate and nitrate may account for up to 12% and 4%, respectively, of total mass (16). We simulate a 5-day period of the NARSTO-Northeast study (18, www.aqd.nps.gov/ard/gas/nar_ne1.htm), Tuesday through Saturday, July 11-15, 1995, over the northeastern United States using a three-dimensional (3-D) air quality model, the EPA Community Multiscale Air Quality model (CMAQ) (19). This period was chosen because it included some high ozone pollution due to meteorological conditions that led to a low mixing height conducive to pollutant accumulation near the surface. First, we describe the simulation period and the modeling domain. Then, we discuss the development of the gridded emission files. Next, we present the modifications made to CMAQ to simulate benzene and diesel particles. Finally, we describe the model simulations and compare the simulation results with available data.
2. Description of the Simulation Period The July 11-15, 1995 period of the NARSTO/Northeast program has been used in earlier studies for meteorological modeling (20, 21) and air quality modeling (22-26). During the July 11-15 period, there was a synoptic scale mid-tropospheric flow with a westerly component across the Appalachian Mountains. In the leeward side of the mountains, air sank and warmed, creating the mesoscale Appalachian lee trough (20, 27). Winds ahead of the trough turned cyclonically and flowed in a south-southwesterly direction up the urban corridor. Having a marine origin, this air remained slightly cooler than the air to the west of the Appalachian Mountains. The result was a shallow boundary layer capped by hot midwestern air mass at middle levels. Nocturnal jets may have also played a role in long-range transport of pollution up the coast. Following earlier studies, the modeling domain consists of two nested grids as shown in Figure 1. The outer domain has 150 × 129 grid cells with a horizontal resolution of 12 km, and the inner domain has 99 × 135 grid cells with a horizontal resolution of 4 km. The vertical resolution consists of 13 layers with finer resolution near the ground (layer thickness, from the ground to the tropopause, is about 18, 33, 50, 160, 251, 258, 299, 447, 1416, 1465, 2390, 3049, and 5837 m).
emissions (tons/year)b % of total 4599 51580 40175 5573 212 109 1456 587 24458 19596 4942 565 3145 1563 936 792 12962 173250
2.7 29.8 23.2 3.2 0.1 0.1 0.8 0.3 14.1 11.3 2.9 0.3 1.8 0.9 0.5 0.5 7.5 100.0
a Emissions from rail and marine vessels are less than 1 ton/year. Emissions within the 12-km resolution domain. c Includes service stations. b
The full 5-day period (ending July 15, 00 GMT) was simulated in the 12-km domain, while the 4-km simulation commenced on July 12, using results from the 12-km simulation as boundary and initial conditions. The meteorological modeling conducted with MM5 by Seaman and Michelson (20) was used for the meteorological inputs. Emission files for nitrogen oxides (NOx), volatile organic compounds (VOC), and carbon monoxide (CO) needed to obtain oxidant concentrations for the benzene simulation were obtained from an earlier study (22). The first 2 days of the period were used as spin-up to clear the majority of the initial conditions from the modeling domain. Therefore, results from the last 3 simulation days were analyzed. All times listed hereafter in the figures are for Eastern Standard Time (EST).
3. Emission Inventories and Emission Processing 3.1. Benzene. The 1996 EPA National Toxics Inventory (NTI) was used as the basis for the development of the gridded emission files. Canadian emissions were not included. The year 1996 was the year closest to 1995 for which an emission inventory was available for air toxics. NTI is an annual inventory with a county-based spatial resolution. To develop gridded emission files with 12-km and 4-km spatial resolution and a 1-h temporal resolution, we used the Sparse Matrix Operator Kernel Emissions model (SMOKE), version 1.3 (28). The emissions from 30 states and the District of Columbia were used for the modeling domain considered here (some states were only partially contained within the domain). The breakdown of this emission inventory by major source categories is summarized in Table 1. On-road motor vehicles represent a large percentage (∼60%) of the inventory. Point sources represent a small fraction (about 3%) of the total benzene emissions. Major point sources are distributed according to their exact location specified by latitude and longitude. Plume rise is calculated hourly within SMOKE to distribute the point source emissions in the appropriate layer. To develop a gridded emission file for area and mobile sources from the county-based NTI, it is necessary to use surrogates that allow the spatial distribution of the available emissions. On-road mobile sources were distributed as follows. Benzene emissions and vehicle-miles traveled (VMT) were available by county (source: http://www.ladco.org/ emis/guide/ems95.html). VMT were also available by road VOL. 37, NO. 22, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5237
FIGURE 2. Benzene emissions inventory at 3 p.m. EST on Saturday, July 15, 1995. type (e.g., rural interstate, rural major and minor arterial, rural major and minor collector, rural local, urban interstate, urban freeways, urban major and minor arterial, urban collector, and urban local). For each road type, we assumed that emissions were proportional to VMT and used separate spatial surrogates for nonurban highways and all other road types. For nonurban highways, we used the exact locations of these highways to distribute the emissions among the grid system. For all other road types, we assumed that VMT were proportional to population and used population distribution as the spatial surrogate. Benzene emissions are further distributed among eight vehicle types, including light duty gasoline vehicles, two categories of light duty gasoline trucks, heavy duty gasoline vehicles, light duty diesel vehicles, light duty diesel trucks, heavy duty diesel vehicles, and motorcycles. We assumed that the distribution of vehicle types was uniform across all roadways. This is a reasonable assumption if we consider that this assumption is applied at the county scale. Spatial surrogates for area and nonroad sources are available for the unified grid (http://www.ladco.org/emis/ unified/unified.html), which can be applied to the NARSTO/ Northeast domain with no change in the coordinate system. The surrogates used in this study include agriculture, airports, housing, major highways, ports, population, railroads, water, and land area. Each area or nonroad mobile source type is matched according to source category codes (SCC) to one of these surrogates within SMOKE for spatial allocation. SMOKE provides temporal emission profiles according to SCC. The temporal resolution includes a distribution of the annual emissions by season, a distribution among weekdays and week-end days and a diurnal profile. For the 5-day episode simulated here, there were 4 week days (July 11-14, Tuesday to Friday) and one weekend day (July 15, a Saturday). Benzene emissions are lower on weekends than weekdays by 7.8%. For comparison, Blanchard and Tanenbaum (29) reported a 14% decrease on average in early morning (5-8 a.m.) benzene concentrations in southern California. Figure 2 presents the benzene emission inventory for the 12-km domain at 3 p.m. on July 15, 1995. Areas with the highest emissions densities include primarily urban areas since these correspond to the highest traffic densities. Benzene emissions from marine vessels can be seen in Figure 5238
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 22, 2003
TABLE 2. Emissions of Diesel Particles by Major Source Categories (Source: 1996 NET) source category
emissions (tons/year)a
% of total
point sources light-duty diesel vehicles light-duty diesel trucks heavy-duty diesel vehicles heavy-duty buses construction equipment agricultural equipment commercial marine vessels railroad engines industrial equipment logging equipment recreational equipment lawn & garden equipment generator sets air compressors welders pumps pressure washers gas compressors airport ground support equipment marine pleasure craft total
12713 1208 756 101109 7961 66846 22330 19743 13071 18448 3792 343 5070 2361 1730 1670 1044 40 7 565 580 281387
4.5 0.4 0.3 35.9 2.8 23.8 7.9 7.0 4.6 6.6 1.3 0.1 1.8 0.8 0.6 0.6 0.4
