Click for next page ( 101


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 100
Management of Carbon Monoxide Air Quality The Clean Air Act's mandate to "protect and enhance the quality of the Nations air resources so as to promote the public health and welfare" and subsequent scientific findings by the U.S. Environmental Protection Agency (EPA) served as the basis for the National Ambient Air Quality Standards (NAAQS) for carbon monoxide (CO). Chapter 1 discussed the CO standards, trends in ambient CO, and the studies that were influential in developing the health-based standards. Achieving and maintaining the NAAQS requires monitoring ambient CO, developing emissions invento- ries, implementing emissions regulations and related controls, and tracking progress. This chapter discusses the primary air quality management ele- ments needed to achieve those objectives the basic emissions control strategies used to reduce emissions and the monitoring and modeling tools used to characterize and assess the magnitude of the problem. EMISSIONS CONTROL PROGRAMS CO emissions control strategies have focused on controlling light-duty vehicle (LDV) emissions. The decline in concentrations noted in the earli- est stages of CO management in the 1 970s corresponded to the implementa- tion of a major enhancement in control of motor-vehicle emissions. There 100

OCR for page 100
Management of CO Air Quality 101 are four approaches for reducing vehicle emissions: ( 1 ) new-vehicle certifi- cation programs, (2) fleet-turnover incentives, (3) in-use vehicle control strategies, and (4) transportation control measures (TCMs) (Guensler 1 99S, 2000~. This section discusses vehicle emissions control strategies in more detail. As LDV emissions decrease, nonroad, area, and smaller stationary sources may become critical for controlling CO in some locations. This section concludes with a brief discussion on the regulation of these other sources of CO pollution. Federal New-Vehicle Emissions Standards Lowering emissions certification standards on new vehicles has been the largest source of reductions in CO emissions from LDVs. For example, the Alaska Department of Environmental Conservation in its most recent SIP for Fairbanks attributed over 70/O of total emissions reductions over the ~ 995-2001 time period to more stringent federal new-vehicle emissions standards (ADEC 20011. Table 3-1 shows emissions standards for passen- ger cars and light-duty trucks.2 CO emissions standards have dropped by over an order of magnitude since their inception emissions from new passenger cars have fallen from 84 grams per mile (g/mi) before emissions controls were instituted to below the current 3 .4 g/mi, which began in ~ 981 . New vehicle technologies offering much better environmental performance Vehicles are certified using the federal test procedure (FTP) and the supple- mental federal test procedure (SFTP), which specify the preconditioning a vehicle is to undergo before testing, the laboratory conditions the test is to occur in, and a specified driving cycle to be used. Testing is done at temperatures between 68F and 86F. Manufacturers are allowed to certify compliance to the 50,000- or 100,000-mile (ml) standards (11 years or 120,000 mi for heavier trucks weighing more than 5,750 lb) using low-mileage cars and an agreed-upon deterioration as- sumption. However, vehicles may be recalled if emissions control systems are found to be faulty. Tier 2 emissions standards, which will begin with model year 2004, are 120,000-mi standards. 2Light-duty trucks have been categorized for emissions certification purposes as light light-duty trucks having a gross vehicle weight rating (GVWR) <3,750 lb (LDT1) or from 3,750 to 5,750 lb (LDT2), and heavy light-duty trucks having a GVWR from 5,751 to 8,500 lb (LDT3 and LDT4~. Trucks with a GVWR greater than 8,500 lb are categorized as heavy-duty vehicles.

OCR for page 100
102 Managing CO in Meteorological and Topographical Problem Areas TABLE 3-1 Federal Passenger-Car and Light-Truck Exhaust Emissions Standards (g/mija Passenger Cars Light Trucksb ModelYear HC CO NOx HCb CO NOx PrecontrolC 10.6 84.0 4.1 1968-1971 4.1 34.0 8.0 102.0 3.6 1972-1974C 3.0 28.0 3.1 8.0 102.0 3.6 1975-1976 1.5 15.0 3.1 2.0 20.0 3.1 1977-1978 1.5 15.0 2.0 2.0 20.0 3.1 1979 1.5 15.0 3.1 1.7 18.0 2.3 1980 0.41 7.0 2.0 1.7 18.0 2.3 _ .. . . Tier O 1981-1983 0.41 3.4 1.0 1.7 18.0 2.3 1984-1986 0.41 3.4 1.0 0.8 10.0 2.3 1987-1993 0.41 3.4 1.0 0.8 10.0 2.3 1988-1993 0.41 3.4 1.0 0.8 10.0 1.2 Tier 1 (1994-) . . 1994 (100~000-mi 0.25 3.4 0.4 0.25 3.4 1.2 standards in (0~31) (4.2) (0.6) parentheses) 1995 (100,000-mi 0.25 3.4 0.4 0.25 3.4 0.4 standards in (0~31) (4.2) (0.6) parentheses) NLEVC (100,000-mi standards) 1999 0.09 4.2 0.3 0.09 4.2 0.3 aAll standards are for 50,000 mi unless otherwise noted. bStandards before 1988 are for all light-duty trucks. Beginning in 1988, light- duty trucks were separated into two weight classes (1988-1993) and then four weight classes (1994-present). The standards after 1988 are for LDT1, which have a 3,750 lb or less gross vehicle weight (GVW). The National Low Emissions Vehicle (NLEV) Program introduces California low-emissions cars and light-duty trucks into the Northeast in 1999 and the rest of the country in 2001. Sources: Davis 1997; Chrysler Corporation 1998. made these achievements possible. This is in contrast to in-use emissions controls, such as vehicle emissions inspection and maintenance (~/M) pro-

OCR for page 100
Management of CO Air Quality 103 grams and oxygenated fuels programs, which do not force the adoption of improved vehicle emissions control technologies and hence reduce vehicle emissions by a much smaller fraction (NRC 1999, 2001~. Recent New-Vehicle Emissions Standards Federal passenger car CO standards have remained at 3.4 g/mi (50,000- mi standard). However there are myriad regulations that have resulted in reductions in vehicle CO emissions. For example, though Tier 1 standards did not affect passenger-car CO emissions, they reduced CO standards for light-duty bucks. Tailpipe emissions of CO end hydrocarbon (HC)respond similarly to changes in air-fuel ratios, and CO is reduced by many of the same vehicle emissions control technologies as HC. Thus, the more strin- gent HC standards imposed since 1981 have resulted in concomitant reduc- tions in CO. In addition to these reduced HC standards that result in reduced CO emissions, there are a number of other changes that directly affect CO emissions. With the introduction of Tier 1 standards in 1994, the durability requirements increased from 50,000 mi to 100,000 mi. The supplemental federal test procedure (SFTP), which is discussed in a subsequent section, controls CO during non-FTP conditions of high acceleration and high speed. Cold-start standards, also discussed in a subsequent section, will limit CO emissions during cold-temperature starts. The recently finalized Tier 2 regulations will also impact CO emis- signs. Control of tropospheric (ground-level) ozone (03), which is caused principally by the interaction of nitrogen oxides (NOX), certain reactive volatile organic compounds (VOCs), and sunlight on hot summer days, has been a continuing need. On February 10, 2000, EPA promulgated a new series of vehicle emissions regulations, known as Tier 2, intended to par- tially address this problem by regulating passenger-car and light-duty truck NOX emissions. Tier 2 requires each manufacturer to meet a sales-weighted "corporate average NOX standard" of 0.07 g/mi. Lowering fuel sulfur con- tent, which is discussed in the section on in-use emissions controls, is also an integral part of the Tier 2 strategy. Table 3-2 lists new emissions limits for NOX, non-methane organic gases (NMOG), CO, formaldehyde (HCHO), and particulate matter (PM) by "bin." Manufacturers certify their vehicles in these bins, ensuring that these vehicles comply with all emissions levels associated with the bins.

OCR for page 100
104 Managing CO in Meteorological and Topographical Problem Areas TABLE 3-2 Tier 2 and Interim Non-Tier 2 Full-Useful-Life Exhaust Mass Emissions Standards (g/mi) Bin Number NO NMOG CO HCHO PM x 11ac 0 9 0.280 7.3 0.032 0.12 10a b d 0.6 0.156/0.230 4.2/6.4 0.018/0.027 0.08 ga,b.e 0.3 0.090/0.180 4.2 0.018 0.06 8 f 0.20 0.125/0.156 4.2 0.018 0.02 7 0.15 0.090 4.2 0.018 0.02 6 0.10 0.090 4.2 0.018 0.01 5 0.07 0.090 4.2 0.018 0.01 4 0.04 0.070 2.1 0.011 0.01 3 0.03 0.055 2.1 0.011 0.01 2 0.02 0.010 2.1 0.004 0.01 1 0.00 0.000 0.0 0.000 0.00 aThis bin and its corresponding intermediate life bin are deleted at end of 2006 model year (end of 2008 model year for HLDTs and MDPVs). Higher NMOG, CO, and HCHO values apply for HLDTs and MDPVs only. This bin is only for MDPVs. Optional NMOG standard of 0.280 glm: applies for qualifying LDT4s and qual- if~ing MDPVs only. Optional NMOG standard of 0.130 g/mi applies for qualifying LDT2s only. fHigher NMOG standard deleted at end of 2008 model year. Source: 65 Fed. Reg. 28 (2000), p. 6855. For example, a manufacturer might certify their sport utility vehicle (SW) in bin 7, their passenger car in bin 5, and an economy car in bin 3. All vehicles must meet the full useful life (which has been raised from 100,000 to 120,000 mi) certification limits for their respective bin. NOX emissions standards for the three bins would then be sales-weighted and compared with the average NOX standard of 0.07 g/mi. Tier 2 regulations also allow manufacturers to trade and bank credits. In years that a manufacturer's corporate average falls below 0.07 g/mi it can generate credits which it can bank and use in years when its corporate average exceeds 0.07 g/mi or it can sell these credits to manufacturers whose corporate average is above 0.07 g/mi. The technologies relevant to the Tier 2 standards will also have benefits for CO reduction. Since the mid-1980s, modern computer-controlled en-

OCR for page 100
Management of CO Air Quality 105 gines have used electronic fuel injectors rather than carburetors to deliver fuel to cylinders in LDVs and most light-duty trucks. The engine computer system reads the signal from an O2 sensor in the exhaust system and contin- uously adjusts the air-fuel ratio. The continuous feedback adjustment of the air-fuel ratio is known as closed-Ioop control. The feedback provides enough air to burn the fuel while maintaining the optimal catalytic-con- verter efficiency (referred to as the stoichiometric ratio) for control of CO, HC, and NOX. Figure 3-1 shows the air-fuel ratio effects on catalyst con- version efficiency. During hard acceleration and high-speed operations, however, engine computers often use fuel-enrichment strategies to enhance engine perfor- mance for short time periods and to protect sensitive engine components from high-temperature damage. Likewise, fuel-enrichment strategies are often used during cold starts. Cold temperature CO standards and the SFTP, which are discussed in the following sections, are intended to further control CO for these conditions. Thus, in modern engines, CO as well as HC emissions are most prominent during enrichment associated with heavy loads, hard accelerations, and cold starts. Enrichment factors are much larger for CO compared with HC (see Figure 3-2) (M. Barth, University of California, Riverside, personal communication, October 30, 2002; Scora et al. 2000~. Conditions that produce a 10- to 1 00-time increase in CO emis- sions produce a 1- to 10-time increase in HC emissions. The primary methods for meeting the Tier 2 standardsensuring stoichiometric engine operation over a broader range of operation and promoting faster catalyst warm-u~will have benefits for CO reductions. As shown in Figure 3-3, the prototype Tier 2 vehicle maintains a stoichi- ometric air-fuel ratio more effectively than a 1996 vehicle certified to Cali- fo~nia's low emissions vehicle (LEV) standard.3 Although the current 3The CAAA90 authorized California, which has the nation's worst air pollution problems, to impose stricter vehicle emissions standards than those for the rest of the nation. California's low emissions vehicles (LEV) regulations require manufac- turers to meet fleet-weighted average emissions lower than those mandated by the federal Tier I regulations beginning with the 1994 model year. The California LEV program includes five progressively more stringent categories: transitional low emissions vehicles (TLEVs), LEV, ultra-low emissions vehicles (ULEVs), super ultra-low emissions vehicles (SULEVs), and zero emissions vehicles (ZEVs).

OCR for page 100
106 Managing CO in Meteorological and Topographical Problem Areas 100 he ~ 80 z z o Cl) o 60 40 20 WINDOW ~ NOX ~ RICH A/F MIXTURE \ \ BEST OPERATING AREA FOR 3-WAY CATALYST LEAN AJF MIXTURE O _ 13:1 14:1 14.7:1 15:1 16:1 AIR-FUEL MIXTURE RATIO \ 1 J FIGURE 3-1 Catalyst conversion efficiency as a fimction of airmail ratio. Source: Adapted from Canale et al. 1978. Reprinted with permission; copyright 1978, Society of Automotive Engineers. 3.4 g/mi federal new-vehicle standard for passenger cars dates to 1981, the use of advanced-technology three-way catalytic converters and continued improvements in stoichiometric ratio controls have had and will continue to have a collateral CO benefit. However, it will take years for Tier 2 regulations to be implemented (2007 for LDVs and 2009 for heavier light-duty trucks), and even longer for fleet turnover to occur and for the full benefits of the new technologies to be realized. An increase in vehicle durability has accompanied techno- logical improvements. According to Davis (2001) the rational average age of in-use passenger cars has increased from a mean of 5.6 years in 1970 to 8.9 years in 1999. The median lifetime of a 1990 model year passenger car is 4.6 years longer (16.1 years) than that of a 1970 model year car. This increase in vehicle durability will slow the penetration of vehicles with newer emissions control technologies into the fleet. In the meantime, the ongoing improvements resulting from HC standards under Tier 1 and NLEV, the cold-start CO standards, the increased durability required under Tier 1, and the introductions of the SFTP will continue to encourage the downward trend in CO emissions from light-duty vehicles in advance of Tier 2.

OCR for page 100
Management of CO Air Quality 107 Engine Start Hot Stabilized Exhaust Engine On Acceleration Enrichment Grade F nri~hmPnt Space and Time (seconds) Engine Off FIGURE 3-2 Hypothetical carbon monoxide emissions rates for typical vehicle operation. In addition, some emissions control strategies for controlling cold-start emissions are particular to HCs and do not improve CO emissions. Some of the lowest-emitting vehicles in California (called super ultra-Iow emis- sions vehicles tSULEV]) use a carbon canister to store uncombusted HC emissions during cold starts. The HC emissions are then recirculated through the catalyst after light-off. Such a control strategy does not reduce CO emissions. ~ summary, the impact of Tier 2 requirements is complex. The CO limits for the higher-emissions vehicles, bins 5-8 (bins 9-1 1 disappear after 2009), is 4.2 g/mi based on a full useful life of 120,000 mi. On the surface, this limit is essentially the same as the Tier ~ and NLEV limits. However, Tier 2 standards are 120,000-mi standards, which should improve vehicle in-use performance. These bins will apply to passenger cars as well as all categories of light-duty bucks (LDT1-LDT4~. This wiTIrequirethatLDT2- LDT4, which under Tier 1 standards have 100,000-mi CO standards from 5.5 g/mi to 7.3 g/mi, meet the current CO standards for passenger cars (4.2 g/mi). At a national level, the result of these bins will be a reduction in CO. If bins 6-8 are used for any vehicle, then bins 14 must be used to average NOX below bin 5. Using the example of a manufacturer certifying their SUV in bin 7 (CO standard at 4.2 g/mi), their passenger car in bin 5 (CO standard at 4.2 g/mi), and their economy car in bin 3 (CO standard at

OCR for page 100
108 Managing CO in Meteorological and Topographical Problem Areas 1996 California Low Emissions Vehicle 15.0 14.9 14.8 14.7- o 14.6 ~ 14.5 2 14.4 14.3 14.2 14.1 14.0 15.0 ~ 14.9 ~ 14.8 ~ O 14.7- ._ 14.6- 14.5- 14.4- 14.3 ~ 14.2 ~ 14.1 ~ 14.0 ~ . . ~ ~~ . ~ . ~ . . . ~ ~ ~ %. ~ ~ ~1 ; -. . ~ - .~\ e. ~$.~,S^' , _,' - An_ ~ ~ _ ~ ~ a, ~ ~ ~ AIMS ~ . . A, ;. - : ~ ~ i. 0 100 200 300 400 500 Time (s) 2003 Tier 2 ~ ; 0 100 200 300 400 500 Time (s) FIGURE 3-3 Example of the improved control of air-fuel ratio resulting from new Tier 2 vehicle technologies. Source: Dana 2002. Reprinted with permis- sion from the author.

OCR for page 100
Management of CO Air Quality 109 2.1 g/mi), the resulting fleet sales-weighted CO will be Tower than 4.2 g/mi. However, this is a national fleet average. Local vehicle fleets may differ from the national average. Cold-Temperature CO Standards As described in Chapter 1, CO is predominantly a winter problem that occurs in regions known for extreme winter conditions (e.g., Fairbanks, Alaska). During cold starts the engine computer signals the fuel injectors to add excess fuel to the intake air to ensure that enough fuel evaporates to yield a flammable mixture in the engine cylinders. A typical engine-com- puter strategy injects several times the stoichiometric amount offue] during the first few engine revolutions, using a fixed fueling schedule to reach idling conditions. Excess fuel continues to be injected until the engine and O2 sensor are warmed up and the exhaust-catalyst inlet temperature reaches about 250-300C, sufficient for the catalyst to oxidize CO to CO2. This open-Ioop operation, before catalyst light-off (the time it takes the catalyst to reach peak efficiency after start), can continue for several minutes at Tow ambient temperatures. Cold-start enrichment is responsible for a signifi- cant fraction of CO, air tonics, and unburned HCs from properly operating vehicles. Once the engine and emissions control systems are warmed up, combustion becomes stoichiometric, and CO is converted to CO2 in the catalyst, keeping CO emissions very low under typical operating condi- tions. Warm up times under mild ambient conditions, at around 70-80F, can be around 1 min for modern catalysts and even as short as a few sec- onds for modern close-coupled catalysts (catalysts close to the engine). When ambient temperatures are -20F or Tower, however, catalyst and engine warm-up times can exceed 5 min (Sierra Research 1999~. In the case of Fairbanks, Alaska, this means that idling and cold-start emissions from LDVs are particularly high and make up a significant proportion of overall CO emissions. ADEC (2001) and NRC (2002) provide more dis- cussion of the role that cold-start emissions play in Fairbanks. Since 1994 new cars and the lightest category of light-duty trucks (LDT1) have been required to meet a CO limit of 10 g/mi on the federal test procedure (FTP) cycle conducted at 20F. For heavier light-duty trucks (trucks between 3,751 and 8,500 Ib gross vehicle weight tGVW]), the stan- dard is 12.5 g/mi. The cold-temperature CO emissions standard has been unchanged since it was promulgated in ~ 994, though certification data from

OCR for page 100
110 Managing CO in Meteorological and Topographical Problem Areas EPA's certification database show that there have been continued improve- ment in cold-start emissions (Figure 3-4~. Reducing the 10 g/mi limit for the 20F cold-start test or reducing the test temperature might provide addi- tional CO emissions reductions for cold northern regions, such as Fair- banks. Indeed, CAAA90 mandated that if six or more areas were desig- nated as nonattainment as of July 1, ~ 997, EPA must require cars to meet a Phase I] cold-start emissions limit of 3.4 g/mi. In their presentations to the committee, representatives of the State of Alaska and the Fairbanks North Star Borough discussed how the adoption of the Phase II cold-start standard would aid in Fairbanks's effort toward long-term attainment ofthe CO NAAQS (Hargesheimer 2001; King 2001; Verrelli 2001~. However, EPA has yet to formally dete~ine the number of CO nonattainment areas that existed as of the deadline. Supplemental Federal Test Procedure An additional source of CO reductions is the SFTP. The technical community has long known about the absence of high speeds and accelera- tions from the FTP. The SFTP introduces speeds as high as 80 MPH and maximum accelerations of 8.4 MPH/s into the certification test (61 Fed. Reg. 54852 t199611. The FTP tests at a maximum speed of 57 MPH and a maximum acceleration of 3.3 MPH/s. Certification to this new cycle will be phased in during the 2000 and 2004 model years. This test procedure should ensure that vehicle emissions control systems will provide improved emissions control over a wider range of vehicle speeds and loads. Much of the improved emissions control will come from reduced use of fuel-rich mixtures at higher Toads. EPA estimates a CO emissions reduction of 1 1% from the LDV fleet in 2020 as a result ofthe SFTP (EPA 1996~. However, it should be noted that for some locations with severe winter Hiving condi- tions, such as Fairbanks, the high speed/high acceleration driving condi- lions within the SFTP are not considered representative. Thus, the benefits from certifying vehicles to the SFTP may be smaller there. Mobile-Source Compliance Programs Mobile-source compliance programs are intended to ensure that vehi- cles meet emissions standards throughout their useful life. There are three

OCR for page 100
138 Managing CO in Meteorological and Topographical Problem Areas Stationary Sources Stationary sources, especially power plants and large industries, may also have a large impact on local CO concentrations. As previously men- tioned, stationary-source emissions factors can be determined from AP42 (EPA 1995~. Stationary-source operations are usually more consistent then mobile-source operations, thus stationary emissions are easier to quantify. Activity, such as fuel usage (often in the form of BTUs generated or amount of fuel consumed per year), is multiplied by an emissions factor to estimate the total mass of CO emitted per year. However, CO exceedances in Birmingham, Alabama, demonstrate that an unregulated point source that experiences process upsets can become a large emissions source sufficient in itselfto create CO exceedances. Utilizing emissions factors from AP42 would underestimate the contribution from sources such as the one in Bir- mingham. In addition, estimating emissions from area sources, such as residential heating, is likely to be highly uncertain. During this study, the committee noted that the emissions inventory for Missoula, Montana, attributed ~ 8/0 of CO emissions to wood burning, whereas the inventory for Fairbanks, Alaska, attributed only 3/0 of CO emissions to that source. The disparity existed despite Missoula's fairly substantial effort to control emissions from wood stoves. The committee also questioned whether the increasing popularity of fuel-oi! stoves has resulted in the underestimation of this source in inventories. It is clear that emissions inventories for stationary sources need improvement. Air Quality Models Air quality modeling is an essential element of air quality management. Models can be used to evaluate plans for attainment of an NAAQS (also referred to as an attainment demonstration), to evaluate the effects of new construction projects, and to conduct further research into what causes pollution episodes and how they can be predicted. A number of modeling techniques requiring various levels of scientific expertise, input data, and computing resources are available for these purposes. The simplest mod- els, rollback models, assume a direct correlation between emissions and ambient pollutant concentrations; the most complicatedmodels, grid-based air quality models, resolve temporal and spatial variations in pollutant concentrations and the effects of meteorology, emissions, chemistry, and

OCR for page 100
Management of CO Air Quality 139 topography. Models are also characterized by the size of the problem they address: microscale models simulate pollution from a point source or intersection; mesoscale models simulate metropolitan or multistate pollu- tion; and large-scare models simulate continental or global pollution. In attainment demonstrations presented in SIPs, states are required by EPA to model how emissions reductions will lead to the desired air quality improvements. Three types of models have been used to demonstrate at- tainment ofthe CO NAAQS: rollback (also knows as statistical rollback), Gaussian dispersion, and numerical predictive models. Rollback Models The simplest of the three models used for attainment demonstrations is the statistical rollback mode] in which the needed reduction in emissions is assumed to be proportional to the required reduction in ambient CO concentrations (ADEC 2001~. CObaSeYear CONAAQS /0 reduct~orl= Cbaseyear Background where CObase year = the second highest 8-hour average in the base year; CONA 4QS = the NAAQS of 9 ppm (or sometimes 9.4 ppm); and Cobackground = an average regional background CO in the absence ~ . . 01 emissions. Although easy to implement, rollback models do not explicitly consider the role of meteorology or the spatial heterogeneity of CO emissions and con- centrations. EPA has allowed states to use rollback models rather than the more resource-intensive dispersion and urban-airshed models described below, to demonstrate attainment in smaller cities. An improvement on the simple rollback model is the probabilistic rollback model used in CO mod- eling for the Puget Sound area of the State of Washington (Ioy et al. 1995~. Gaussian Dispersion Models A second type of model that has been used for CO-attainment demon- strations is a Gaussian dispersion model, which is typically used to simulate CO concentrations for microscale analysis in the vicinity of intersections

OCR for page 100
140 Managing CO in Meteorological and Topographical Problem Areas or along major traffic corridors (EPA 1992~. One of the first effective Gaussian dispersion models for mobile sources was CALrNE3, which is still in use. Inputs for this model include meteorological data, such as wind speed and atmospheric inversion strength in the vicinity of the pollutant source, and temporally resolved emissions. Emission factors developed from other emissions models (MOBILE and EMFAC), along with traffic volumes, roadway geometries, and intersection information, are used to determine the emissions along a roadway. Dispersion modeling then in- cludes transport and mixing to calculate local concentrations. The model is Gaussian in nature, meaning it assumes that a plume of pollutant gas released from a point source can be described by a widening Gaussian function (a bell-shaped curve) as it travels downwind (Wayson 1999~. The model also makes the assumption that roadway segments can be cut into small sections with a point source approximation applied to each and their plume concentration contributions summed at a receptor site. This concept allows roadway curves or winds nearly parallel to the roadway to be mod- eled effectively. The shortcoming of CALINE3 is that it is only useful for vehicles that are moving at a constant rate of speed. At locations of high CO emissions (such as intersections), increased emissions due to vehicle delay and idling must be accounted for. To do that, two models are in use today: CAL3QHC and CALLNE4. Both use the same general approach to estimate dispersion as CALINE3 does. CAL3QHC is used in 49 states, and CALLNE4 is used in California. Gaussian dispersion models are typically used for local area (micro- scaTe) analysis and are used extensively in CO-related evaluations, includ- ing project-level conformity determinations. Modeling is done for the worst hour to compare with the 1-hour average CO NAAQS. Worst-case conditions (a windspeed of 1 MPH and a stable atmosphere) are often used. A persistence factor, whichis a multiplier ofthe peak 1-hour concentration that is based on changes in wind patterns and traffic, is used to estimate an 8-houraverageconcentrationforcomparisonwiththeS-hourNAAQS. The model results often determine whether a project can go forward. The American Meteorological Society (AMS) policy statement on dispersion modeling (Henna 1978) concluded that these models are accu- rate within a factor of 2 for reasonably steady horizontally homogeneous conditions; however, they will be less accurate, for example, when obstacle wakes flows (e.g., from buildings or vehicles) and extremely stable thermo- dynamic lapse rates occur. Dispersion accuracy will also be Tower, as listed

OCR for page 100
Management of CO Air Quality 141 in the AMS statement, for "dispersion over forests, cities, water and rough terrain." Grid-based Air Quality Models The most complicated models used for attainment demonstrations simulate how a pollutant concentration varies with time and space over an entire urban area. These numerical predictive models, generally intended for regional analysis, can simulate emissions from multiple sources end the dispersion, advection, and photochemical reactions of gaseous pollutants in the atmosphere. These models are integrated separately from meteoro- Togical models. Grid-based models, such as Models-3 and the urban air- shed mode] (UAM), have been used for many years to simulate 03, which is a region-wide or mesoscale pollutant. The UAM has been adapted to simulate CO in Denver (Colorado Department of Public Health and Envi- ronment 2000~. Because ofthe local nature of high-CO episodes, extensive modeling ofthe entire urban airshed may be unnecessary for CO-attainment demonstrations. Airshed modeling is resource-intensive, requiring detailed knowledge of an area's meteorology (usually based on the output of a mesoscaTe weather model constrained by observations), spatially and temporallyresolved emissions inventories, and measurements ofthe pollut- ant at several locations to allow model evaluation. Highly trained person- net are needed to conduct the simulations. More complicated models are not always appropriate for attainment demonstrations, but they can be valuable in improving the understanding of the interactions among atmospheric processes. Even better research tools than the numerical predictive models described above (such as Models-3 and the UAM) are process numerical models, which allow pro- cesses specific to air quality modeling and meteorology to be coupled within a single computational framework. Process numerical models typi- cally are formulated by adding pollutant emissions, chemistry, and trans- port into an existing meteorological model rather than simply using the meteorological data as a mode! input. The relatively nonreactive behavior of CO makes it an ideal chemical species for simulation in a weather model. Predictions of CO, for example, can be straightforward in the Na- tional Weather Service Eta model, Is which has a horizontal grid framework Resee NWS 2002 for information on the NWS Eta model.

OCR for page 100
142 Managing CO in Meteorological and Topographical Problem Areas of 12 x 12 km over the contiguous United States. However, this resolution is insufficient for most CO problem areas. Initial work to simultaneously simulate atmospheric flow and diffusion of CO at high spatial and temporal resolution is described by Fullerton (20021. Box Models Box models are another tool available for microscale analysis of air pollution. The "box" is some volume of air into which emissions are in- jected. Box models may divide a region into cells of equal volume and use mass balances to treat the transfer of CO among cells. In their simplest application, they can consist of a single box. The cells may also be sepa- rated in the vertical direction. Air within each cell is assumed to be well mixed. Simplifications ofthis concept lead to the common single-cell box model. Though box models are not used in attainment demonstrations, they are particularly useful to understand how various emissions scenarios and meteorological conditions affect pollutant concentrations. For example, a box model for CO in Anchorage, Alaska, has been used to quantify how mechanical turbulence from roadway traffic might increase the mixing height and reduce CO concentrations on severe-stagnation days compared with concentrations observed in residential neighborhoods (Morris 2001~. Appendix C describes a single-cell box model, with and without recirculation. The committee's interim report on Fairbanks describes the application of such a model to Fairbanks, Alaska (NRC 2002~. Summary of Air Quality Models There is no single air quality model that is the best for CO for all Toca- tions. Typically the choice depends on the severity of the problem, the available data, and the resources available for modeling. It its interim report (NRC 2002), the committee recommended that Fairbanks, Alaska, use a simple box-model approach for air quality planning purposes in the near term. A box model simulates the effects of emissions end meteorology in a well-mixed controlled volume. The committee felt that such an ap- proach could provide greater insights into the effects of the timing of CO emissions and of meteorological variables, in this particular situation, given the limited vertical dispersion and available data. Box models could sup-

OCR for page 100
Management of CO Air Quality 143 plement Fairbanks's current approach of using a simple rollback model, which they used in their attainment demonstration (ADEC 20011. In the long-term, the committee recommended that more work be done to develop, apply, and evaluate more sophisticated, physically comprehen- sive models that would simulate how CO concentrations vary with time and space. Because CO is relatively conservative on time scales of hours, knowledge ofthe temporal and spatial distribution of CO emissions and of the observed CO concentration field provide an effective diagnosis of atmo- spheric dispersion patterns. For chemical species that are eliminated by reactions in the atmosphere, knowledge of the CO dispersion provides an observational constraint on the concentration fields ofthe reactive species. The committee concluded that more physically comprehensive models should be used for planning, forecasting, and assessing human exposures to high CO concentrations. It is important that mode] development and testing be specific to the extreme conditions that occur in CO problem areas such as Fairbanks. However, model development must occur in concert with improved monitoring to enable model evaluation. The committee believes that even in areas such as Fairbanks, which has experienced very few exceedances since 1996, and none since 2001, the development of comprehensive models is still worthwhile. The number of periods of ele- vated CO levels experienced in Fairbanks indicates that the city is still susceptible to exceedances. Furthermore, CO modeling can be used to better understand and characterize CO hot spots as well as other criteria pollutants and air tonics. The development of a better modeling approach today will benefit all problem areas in the future. Despite advances in air quality modeling capabilities over the last 30 years, many improvements are still possible and necessary. One problem is that the vertical and horizontal resolution of models is too coarse to capture the variability in pollutant concentrations, which is necessary to identify local hot spots and is important for determining local concentra- tions downwind of hot spots. In addition, the validity of mode] representa- tion becomes questionable when unusual meteorological conditions occur, and that could lead to errors in the prediction (Pielke 2002~. Models used for regulatory purposes can suffer a loss of realism as a result of such short- comings, leading to costly errors in planning. Models also need more real- istic three-dimensional dynamics (advection, pressure gradient forcing' turbulences and more realistic parameterizations of smaller-scale processes (e.g., turbulence fi om buildings, radiative flux divergence changes in the temperature profile associated with aerosols in the lower levels ofthe atmo-

OCR for page 100
144 Managing CO in Meteorological and Topographical Problem Areas sphere). The models also need higher spatial and temporal resolution. Ensemble runs of the models should be performed to provide a more realis- tic envelope of simulated dispersion patterns. However, the committee recognizes that this adds cost and time to the evaluation. Not only can these models be used for air quality applications, models with higher reso- lution can also assist in homeland defense because they can help understand the dispersion of accidental or deliberate releases of chemical, biological, and radiological materials. In 2003, a large dispersion research project will be undertaken to help define important dispersion parameters, primarily for homeland security purposes (DOE/DOD 2002~. The project will be a month-Ion" study con- ducted by a combination of federal and state governmental agencies with support from multiple universities. Research will include releases oftracer gases with careful measurements of meteorological parameters to determine dispersion trends for city-wide dispersion, dispersion in street canyons, infiltration to buildings, and effects of topography. Statistically Robust Methods to Assist in Tracking Progress The air quality models described above assess the effectiveness of emissions controls and the prospects for attaining the CO standard by repre- senting critical processes within a physically based model of the system. An alternative to those physical models is to take a statistical approach assessing the relationship among human activities, CO emissions, meteorol- ogy, and ambient air quality, as described below. Probability of an Exceedance Reddy (2000) carried out an analysis of the probability of a future CO exceedance in Denver that might be broadly applicable to other areas. The analysis took into account the historical variability in CO concentrations as a result of meteorology and unusual traffic events. The purpose of his analysis was to determine the risk of a CO exceedance associated with eliminating or altering the oxyfuels program during the first week in Febru- ary for the future years 2002-2013. He used CO monitoring data from the CAMP site (AIRS ID 08-031-002), which is the site in Denver that has historically shown the greatest number of exceedances. He used daily peak

OCR for page 100
Management of CO Air Quality 145 8-hour average CO concentrations for the first week in February for the 20- year period of 1975-1994. Because these values depended on the emissions during those years in addition to stochastic meteorology and occasional unusual traffic, Reddy corrected past CO concentrations for each year to what they would have been if the emissions for that year had been the same as those projected for 2002. The natural logarithms of the corrected peak 8-hour average CO con- centrations were normally distributed; the 8-hour averages themselves were not. By estimating future emissions inventories for the years 2002-2013, based on projected fleet composition and VMT, and assuming that the lognormal distribution would hold for future years, Reddy was able to calculate the probability of an exceedance on a single day of the first week in February (P~ ~ for the future years. He then used Equation 1 to compute the probability of one or more exceedance days during an entire week (P79 P76 = 1 - (1-P~ 97. (1) Reddy found a better than 5/O chance that an exceedance might occur if Denver immediately suspended the oxyfuels program for the period 2002- 2013. The study also found that Denver would likely not have an exceed- ance if 1.5% oxygen (which is less than the oxygen content used in the current oxyfuels program) was used in fuels for 2002 and 2003 before suspending the use of oxyfuels for 2004 through 20 ~ 3. Equation ~ assumes that exceedance events are independent over time (thus the probabilities can be multiplied, as in the second term on the right hand side of the equation). The assumption might not hold, for example, exceedance events might be positively associated over time. Given this possibility, Reddy' s method might overestimate P7 a. Alternatively, we can modify Reddy's equation as follows: Expected number of exceedances = N X Pi 4, (2) where N denotes the number of days in the time period being considered, under the assumption that the probability of exceedance is uniform over the time period. For Reddy's application, the time period considered is the first week in February, thus N= 7. Under more general conditions, Equation 2 can be modified as follows:

OCR for page 100
146 Managing CO in Meteorological and Topographical Problem Areas Expected number of exceedances = I. ~ Pi ~ (3) where Pi denotes the probability of exceedance on the i-th day. Equation 3 does not assume that the exceedance probability is uniform over time. For example, one might use a different exceedance probability for week- days versus weekends. The same procedure that Reddy used, or the modified one discussed above, could be applied to monitoring sites in other cities and for times other than the first week in February (e.g., a whole winter season), provided that there are enough historical data to establish the distribution of CO concentrations and to estimate emissions inventories for past and future years. Meteorological De-trending Ambient CO concentrations across the nation are going down. Un- doubtedly many of these reductions are due to emissions controls. Part of the kend, however, may also be meteorological. A warmer winter with less stagnation can lead to Tower winter CO levels. As noted by Neff (2001), Denver may be experiencing lower CO levels than would be expected from emissions reductions alone because of warmer winters with greater vertical mixing. How can the impact of meteorological trends on the observed concentrations be removed in order to assess the impact of emissions con- trols and to show true progress towards meeting air quality standards in the future, when meteorological conditions may not tee so favorable? One must "de-trend" the observations. Meteorological de-trending is accomplished by identifying how meteo- rological variables impact pollutant concentrations and removing the influ- ence of those variables. One way would be to create a physically realistic model that can simulate many years, developing emissions-to-air quality relationships and showing how they respond to meteorological influences. However, this approach would be cumbersome and would introduce signifi- cant uncertainties. The influence of meteorology is more typically identi- fied using an empirical approach. Many years worth of concentration data are analyzed, along with the corresponding meteorological data, to develop a statistically based model. That model is then used to remove meteorolog- ical impacts (Kuebler et al. 2001; Porter et al. 2001).

OCR for page 100
Management of CO Air Quality 147 Recent work by Flaum and colleagues used a multistep process to resolve the trends in ozone (03) into four components: a long-term trend, presumably due to emissions controls; a seasonal component; a component driven by meteorological fluctuations; and a noise component (Flaum et al. 19961. Kuebler et al. (2001) used a similar approach, not only for 03, but also for CO, NOX, and VOCs, and compared the meteorologically de- trended concentrations of the primary pollutants with the trends in emis- sions estimates. From that, a direct relationship between the emissions levels and pollutant concentrations could be established. The latter approach appears appropriate here given its prior use for CO, though the explanatory variables may depend on location. For example, in Fairbanks, a nonlinear response to temperature is expected because CO concentrations appear to be highest at about -20F to 20F, not at much lower or much higher temperatures. This approach is convenient for local air quality management organizations because it requires relatively little data (e.g., a long-term record of CO concentrations and meteorological variables such as temperature and windspeed would suffice, though more factors are useful) and nominal computational power. The de-trending analyses also can provide extra information for air quality planning. As noted above, de-trending can be used to help develop

OCR for page 100
148 Managing CO in Meteorological and Topographical Problem Areas probabilities of exceeding the NAAQS for CO at various emissions levels. From that, the necessary level of emissions can be identified in a more statistically robust fashion.