National Academies Press: OpenBook

Modeling Mobile-Source Emissions (2000)

Chapter: 3 Technical Issues Associated with the MOBILE Model

« Previous: 2 Uses of MOBILE in Air-Quality Management
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 61
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 62
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 63
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 64
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 65
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 66
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 67
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 68
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 69
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 70
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 71
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 72
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 73
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 74
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 75
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 76
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 77
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 78
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 79
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 80
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 81
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 82
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 83
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 84
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 85
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 86
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 87
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 88
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 89
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 90
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 91
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 92
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 93
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 94
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 95
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 96
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 97
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 98
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 99
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 100
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 101
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 102
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 103
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 104
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 105
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 106
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 107
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 108
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 109
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 110
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 111
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 112
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 113
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 114
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 115
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 116
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 117
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 118
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 119
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 120
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 121
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 122
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 123
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 124
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 125
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 126
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 127
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 128
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 129
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 130
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 131
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 132
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 133
Suggested Citation:"3 Technical Issues Associated with the MOBILE Model." Transportation Research Board and National Research Council. 2000. Modeling Mobile-Source Emissions. Washington, DC: The National Academies Press. doi: 10.17226/9857.
×
Page 134

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Technical Issues Associated With the MOBILE Mocle' THE FOCUS OF CHAPTER 3 is the technical issues associated with the cur- rent (MOBILES) and upcoming (MOBILE6) versions of the MOBILE mod- el. The introductory portion of this chapter discusses the development of the model and the updates for MOBILE6. The chapter goes on to describe related models for estimating mobile-source emissions (PARTS, NON- ROAD, and others) and previous reviews of the model. A major portion of this chapter then focuses on the technical issues associated with the mod- el, such as how the model handles high emitters, driving cycle, start emis- sions, and many others details. The chapter concludes with a summary and recommendations related to technical aspects of the model. HISTORY AND STATUS OF THE MOBILE MODEL History of the MOBILE Model The MOBILE model for estimating on-road vehicle emissions factors (in grams per mile [g/mi]) was first developed by the U.S. Environmental Pro- tection Agency (EPA) in the late 1970s. Prior to that time, the agency pub- lished simple look-up tables for estimation of mobile-source emissions. The model, originally and still written using the Fortran scientific pro- gramming language, has had significant updates and new releases every few years as new data became available, new regulations were promul- 61

62 M ODEL/NG M OB![E-SOURCE EMISSIONS gated, emissions standards were established, and sources and processes of vehicle emissions were better understood. Each new version of the model has become more complex in the approach to modeling average in-use ve- hicle emissions, and has provided the user with additional options for tai- Toring emissions-factor estimates to local conditions. The model versions, release dates, and changes in each model update are summarized in Table 3-1 (EPA 1999e). Changes in the databases underlying the models and changes in model- ing methodology in each successive version result in changes to predicted total on-road vehicle emissions. From one model version to the next, these changes can be either increases or decreases in emissions factors, and the changes are not always in the same direction for all three pollutants (NOX, CO, and VOCs). Although these changes created somewhat of a moving target for air-quality planners and the regulated industries, the revised models should provide more accurate analyses of mobile source emissions and of the effects of mobile source control programs. As an example, Figure 3-1 shows emissions for the Baltimore area for calendar years 1988 and 1990 as predicted using three recent official re- lease versions of the model (MOBILE5a, MOBILE4.1, and MOBILES; these comparisons are unaffected by the inclusion of new emissions stan- dards and regulations. Carbon monoxide (CO) emissions and emissions of nitrogen oxides (NOx) increase from one model to the next, albeit in differ- ent proportions. Volatile organic compound (VOC) emissions, though, de- creased from MOBILE4 to MOBILE4.1, and then increased significantly from MOBILE4. 1 to MOBILE5a, while still remaining lower than MOBILE4 levels. MOBILES—The Current Moclel The MOBILES model, released in 1993, provides emission factors for on-road vehicles for the three regulated pollutants: VOCs, CO, and NOX. The model provides emission factors separately for the classes of vehicles listed in Table 3-2, and also for the average on-road fleet using a default national mix of vehicles; the user can optionally input a different fleet mix for the calculation of fleet average emissions. The vehicle classes are fur- ther subdivided into technology classes in MOBILE, to account for emis- sions differences between, for example, vehicles with carburetors and those with fuel injection. To estimate total on-road mobile emissions in a given area, either the vehicle class emissions factor is multiplied by esti- mates of vehicle miles traveled (VMT) by vehicle class for the area and summed, or the fleet average emissions factor is multiplied by total VMT (across vehicles classes) for the area. These VMT estimates are typically provided by local or regional transportation-planning agencies.

TECHNICAL ISSUES ASSOCIATED WITH THE M OBILE MODEL 63 TABLE 3-1 MOBILE Model Revisions Release Version Date Model Revisions MOBILE 1 1978 Included modeling of exhaust emissions rates as func- tions of vehicle age/mileage (zero-mile levels and deteriora- tion rates). MOBILE2 1981 Updated with substantial data (available for the first time) on emission-controlled vehicles (i.e., catalytic con- verters, model years 1975 and later) at higher ages/mileages. Provided additional user control of input options. MOBILES 1984 Updated with substantial new in-use data. Elimination of California vehicle emissions rates (contin- ued to model low- and high-altitude emissions). Added tampering (rates and associated emissions im- pacts) and anti-tampering program benefits. MOBILE4 1989 In-use emissions-factor estimates for nonexhaust emis- sions adjusted for real-world fuel volatility as measured by Reid vapor pressure (RVP). Updated with new in-use data. Added running losses as distinct emissions source from gasoline-powered vehicles. Modeled fuel volatility (RVP) effects on exhaust emis- sions rates. Continued expansion of user-controlled options for input data. MOBILE4.1 1991 Updated with new in-use data. Added numerous features allowing user control of more parameters affecting in-use emissions levels, including more inspection/maintenance (I/M) program designs. Included effects of various new emissions standards and related regulatory changes (e.g., test procedures). Included impact of oxygenated fuels (e.g., gasohol) on CO . . emlsslons. MOBILES 1993 Updated with new in-use data, including basing new ba- sic emissions-rate equations on much larger database de- rived from state-implemented IM240 test programs. Included effects of new evaporative emissions test proce- dure (impact on in-use nonexhaust emissions levels). Included effects of reformulated gasoline (RF G). Included effects of new NOX standard of 4.0 g/bhp-hr for heavy-duty engines. Included impact of oxygenated fuels on VOC emissions. Included Tier 1 emissions standards under 1990 Clean Air Act Amendments. Added July 1 evaluation option. (Corltirrued)

64 MODE1/NG MOB/[E-SOURCE EMISSIONS Table 3-1 Model Revisions (Continued) Included impact of low-emission vehicle (LEV) programs patterned after California regulations. Revised speed corrections used to model emissions fac- tors over range of traffic speeds. MOBILE5a 1993 Corrected a number of minor errors in MOBILES. MOBILE5b 1996 Included final on-board vapor-recovery regulations. Included final reformulated gasoline regulations. Added more user options for I/M programs. Source: EPA l999e All of the MOBILES emissions factors are estimated from existing test data, and engineering judgment in the absence of test data. Although there is a detailed User's Guide (EPA 1994) for the model, there is limited documentation from EPA describing the databases and analytical methods used in MOBILES to develop the emissions factors. The user provides inputs (some required and some optionaV to MOBILES that describe typical operating characteristics, fleet character- ization, and mobile-source control programs. These inputs (in addition to vehicle class and VMT mentioned above) are ambient temperature; average vehicle speeds by vehicle class; . fuel characteristics; (include fuel volatility and oxygen content, and if reformulated gasoline is in use); vehicle inspection and maintenance (I/M) program parameters, if such a program is in place; and vehicle age distributions (used to estimate composite emissions across all vehicle model years). Figure 3-2 shows emissions estimates developed using an updated ver- sion of MOBILES and updated travel-activity estimates. It shows the pre- dicted distribution of on-road mobile source emissions of VOCs and NOX in year 2007 by emissions category for New York and Chicago. These inven- tories were generated by EPA using the latest emissions model developed as part of the regulatory impact assessment for the recent Tier 2 vehicle emissions and fuel sulfur standards (EPA 19996~. This model, known as the Tier 2 Model, was developed from MOBILE5b and available elements of the upcoming version, MOBILES (EPA l999c). The Tier 2 model is ac- tually a spreadsheet program derived from MOBILE algorithms, outputs, supplemental test data, and assumptions. The MOBILES elements incor- porated into the Tier 2 Model include updated assessments of in-use vehi- cle deterioration, fuel sulfur impacts, and fleet characteristics. However,

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 65 60 - 50 - —- ~o - o ~ 30 o US In - 10 - o - 1E MOBILE5A 20 ~ ~ ll]~: · MOBILE4.1 O MOBILE4 HC HC NOX NOX CO/10 CO/10 1988 1990 1988 1990 1988 1990 FIGURE 3-1 Comparison of estimated emissions for Baltimore from three recent versions of the MOBILE model. Note that CO emissions are divided by 10. Also note that this report usually uses the term VOCs as opposed to hydrocarbons (HCs) to refer to the general class of gaseous organic com- pounds the Tier 2 Model does not use the MOBILES methodology and test data for estimating evaporative emissions. MOBILES will likely increase the frac- tion of evaporative to tailpipe emissions of VOCs compared with that ob- tained in MOBILE5b. Both cities depicted in Figure 3-2 show a fairly similar emissions profile, although important differences are clear. Chicago has much greater emis- sions from heavy-duty vehicles. This is especially apparent for NOX emis- sions; Chicago has 37% of emissions from HDDVs whereas New York only has 24%. Chicago also has greater emissions from motorcycles. For exam- ple, Chicago has 7% of their VOC emissions attributed to motorcycles, over twice the percent of emissions from motorcycles in New York. Generally speaking, the Tier 2 Model estimates that about 45% of the total on-road VOC emissions is from light-duty vehicle exhaust, about 30% is from light- duty evaporative emissions, and the remainder is primarily from heavy- duty vehicles. Note that the regulatory impact analysis was performed for four cities Atlanta, Charlotte, Chicago, and New York. However, the emissions profiles for Atlanta and Charlotte were similar to those for New York, and are not shown here.

66 MODELING MOB/LE-SOURCE EMISSIONS TABLE MOBILE b Vehicle Classes Vehicle Class Light-duty gasoline vehicles (passenger cars) Light-duty gasoline trucksa (pick-ups, minivans, passenger vans, and sport-utility vehicles) MOBILE Code LDGV Weight Description Up to 6000 lb gross vehicle weight (GV\iV) Up to 6000 lb GVVV LDGT1 LDGT2 Heavy-duty gasoline vehicles HDGV Light-duty diesel vehicles (passenger cars) Light-duty diesel trucks Heavy-duty diesel vehicles Motorcyclesb Emissions for light-duty trucks are modeled separately for two weight classes with different emissions standards in the Clean Air Act bHighway-certif fed motorcycles only are included in the model. OiT-road motor- cycles, such as dirt bikes, are modeled as a non-road mobile source in EPA's NON- ROAD model. LDDN7 LDDT HDDV MC 6001-8500 lb GVVV 8501 lb and higher GVW equipped with heavy-duty . gave .lne engines Up to 6000 lb GVVV Up to 8500 lb GVW 8501 lb and higher GVW MOBILE6- The Next Generation Model EPA's Office of Transportation and Air Quality (OTAQ) has for the last several years been working on the next generation of the MOBILE model, referred to as MOBILE6. This model will be significantly different from MOBILES in almost all model components, and will be based on an enor- mous amount of recent vehicle-emissions testing data from EPA, the Cali- fornia Air Resources Board (CARB), automobile manufacturers, and petro- leum refiners. At this point, MOBILE6 is expected to be released in the year 2000. However, EPA has already released substantial documentation and held workshops describing the model revisions, allowing the agency to gather feedback on its proposed modifications. This documentation and public outreach process will be discussed in the following section. The sig- nificant changes being incorporated into MOP;TT,F:Hi into t.h~ fallowing dramatically lower basic emissions rates, based on analyses of the Dayton, Ohio I/M program data; Reparation of start and running-exhaust emissions; addition of so-called off-cycle emissions (aggressive driving and air- conditioning operation, which are not included in the Federal Test Proce- dure [FTP]~;

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 67 2007 VOC New York 6% 3% 1%~ - ~...3 ~ l ~~ ~ 9% ~ l ~ ,.~ ~ Ft., ~ ~ ~ .. -.- :::- :.-. ~ ~ ~ ~ ~ 48% .:.:...: .-.:.:.:.:.:: :.: ::.:: :.::t ~ ~ ~ ~ (~ ! i ~ 33%~. 11 ~ i New York 24% ~ ~ I...*....*. I,-.--..--.. '* *-* ****,.*.* i.S**..SSSSSS... Fat——. -—*an. .~*~**~. ·****~-*-~*-**. ~0.2% 2007 NOx 4% Chicago 10% _ ` - ~ ~_ , _ I,,..,,_ ~SSSSSSS. 29% '>-an--. ~~..~., _,-----... ~SSS.SSS:SS ~ err ~—- ..—- ~—- - ~— `5 / /0 ~ a***-----*-* ——~——~—~—~ 1*~.—are- - - h. SSSS--SSS. S SSSSSSSSSSSSSS ·—+*~—- - - - - - -— .~-~. I---------- ~—- ~.—~— 22222222222-.* I-.-.-.-. . '. ''I. . ~ Chicago Legend LDV Exhaust ~ LDV Evap. ~HDGV Exhaust ~ ~ HDGV Evap. ~ HDDV Exhaust ~ MC Exhaust & I Evap. 58% FIGURE 3-2 VOC and NOx emissions inventories for New York and Chicago. Source: EPA 19994. control of off-cycle emissions with the Supplemental FTP (SFTP) in future years; . emissions factor estimates for different roadway types (e.g., highways arterials, locals); evaporative diurnal emissions factors estimated *om real-time diur- nal test data previously unavailable,

cards. 68 MODELING MOBILE-SOURCE EMISSIONS revised (lower) estimates of the effects of oxygenated fuels on CO . . emlsslons; revised (lower) effects of I/M programs on vehicle emissions; addition of off-cycle NOx emissions for heavy-duty diesel vehicles; effects of in-use fuel sulfur content on all emissions; and effects of national low-emissions vehicle (NLEV) and Tier 2 stan- Although the MOBILE6 documentation provides numerical results for changes in specific model components, overall changes to average in-use fleet emissions factors will not be known until the full model is released. Thus, it is not yet known whether regional emissions estimates from MOBILE6 will increase or decrease relative to MOBILES, though it is like- ly that they will increase at least for VOCs in order to be in better agree- ment with the findings of evaluation studies that are discussed in Chapter 4. FEDERAL ADVISORY COMMITTEE ACT PROCESS AND PUBLIC OUTREACH An important part of the developmental process for MOBILE6 has been public outreach. This includes input from EPA advisory committees, com- ments *om stakeholders and the interested public, and the release of tech- nical documentation describing model modifications. The Clean Air Act Amendments of 1990 (CAAA90) established the Clean Air Act Advisory Committee to advise EPA on issues of implementa- tion this law. One of the many subcommittees of the Clean Air Act Advi- sory Committee is the Mobile Sources Technical Review Subcommittee (MSTRS), often referred to as a FACA subcommittee because it is char- tered under the Federal Advisory Committee Act (FA CA). The MSTRS advises EPA's OTAQ on technical issues specific to the control of emissions from mobile-sources. It is composed of experts on mobile-source emissions from industry, academia, state agencies, and nongovernmental organiza- tions. Meetings are held quarterly and are open to the public. One of the MSTRS working groups is the Modeling Working Group, which provides on-going advice on the development and improvement of MOBILE and other emissions models. The specific charge for this work group includes helping to set priorities for developments to MOBILE6 and developing procedures for EPA to use when obtaining outside review for products used to support MOBILE6. This group also is producing a com- prehensive report on the MOBILE modeling process, problems, and oppor-

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 69 "unities for improvement on which this committee was briefed (this paper, "Big Picture Modeling Issues," is currently in draft form). Additionally, EPA has increased public feedback and public involvement in the development of MOBILE6 in several other ways. These steps were taken to make the model better understood by the user community and to counter criticism that the model was inadequately documented and peer reviewed. Although the draft MOBILE6 model is not expected to be avail- able until later in 2000, EPA has released detailed technical documenta- tion for most of the proposed changes in the Stakeholder Review Docu- ments on MOBILE6 web page (http://www.epa.gov/OMSWWW/m6.htm). EPA is to be commended for documenting the databases used and the de- velopment of revised emissions factors in the Stakeholder Review Docu- ments. This documentation, although not always complete in describing the full details of the analyses, is a major improvement from all previous versions of MOBILE. EPA provides a 60-day review period for each docu- ment as it is posted, and has stated its intention to provide all comments and responses to comments with each document in the final version. EPA has also held three workshops discussing the new version of the model and created an e-mail list server to update interested parties on new model developments. The workshops were open to the public. They were intended to update interested parties on EPA's plans for the model as well as solicit input and reaction to those plans. The workshops included both technical presentations describing changes to the model methodology and presentations oriented to model users describing changes to data in- put and output. The e-mail list server is used to announce the workshop agendas, the release of new documentation concerning the MOBILE6 mod- el, updates to the current version of MOBILE5b, and other information. RELATED MODELS There are several emissions models and databases related to EPA's MO- BILE model, which are used to estimate mobile-source emissions invento- ries and provide inputs for air-quality models. These are PARTS estimates particulate matter (PM) emissions factors for on- road vehicles; Complex Model estimates emissions impacts of reformulated fuel compositional changes on 1990 light-duty gasoline vehicles; MOBTOX estimates on-road mobile-source toxic emissions factors; and SPECIATE, and related databases and models provide VOC speciation profiles for complex photochemical grid modeling.

70 MODE[/NG MOBI[E-SOURCE EMISSIONS These models are each described briefly below. References are provided for readers who desire more detailed information on any of these models. PARTS EPA's PARTS mode! estimates PM emissions factors in g/mi for 12 vehi- cle classes. Emissions factor estimates are provided for particle diameter sizes from less than or equal to 1.0 to 10.0 ,um (micrometers). The model is referred to as PARTS to indicate consistency with MOBILES in fleet char- acterization data and in the general methods used to estimate basic emis- sions rates. PARTS estimates aD PM emissions associated with on-road travel: exhaust emissions, brake-wear emissions, tire-wear emissions, and fleet-average paved and unpaved road-dust emissions. For HDDVs, PARTS also provides estimates of idle emissions. The PARTS model has been updated in only very minor ways since the original model develop- ment in the mid-1980s. The emissions factors in the model are based on either engine certification data or on ratios from VOC emissions. Al- though there is a User's Guide for the model, there is no documentation that explains the derivation of the emissions factors in the model. The emissions-factor estimates in the PARTS model are seriously out of date. New PM test data have recently become available for both light- duty vehicles (LDVs) and heavy-duty vehicles (HDVs). Although on-road mobile-source emissions-factor estimates will be needed for state and local air-quality planning agencies to develop PM emissions inventories and air- quality management plans for the new PM standards, EPA's OTAQ has not focused as much effort on updates to the PARTS model as on MOBILE. However, many of the revisions developed for MOBILES can be easily in- corporated into PARTS, and OTAQ has done so for estimating PM emis- sions as part of the Tier 2 rulemaking (EPA 1998d). But the major im- provement required for the model is the inclusion of recent testing data for the revision of emissions factors. Because most of the methods for estimating PM emissions in PARTS are similar to the methods used in MOBILES, and because PARTS needs major revision, an updated version of MOBILES should incorporate re- vised PM emissions-factor estimates. For most users, it would be much more desirable to have one integrated model that provides emissions-fac- tor estimates for PM as well as VOC, NOX, and CO. Complex Moclel The Complex Model is used by petroleum refiners and other interested parties to estimate how gasoline composition affects vehicle emissions.

TECHN/CA~ ISSUES ASSOCIATED WITH THE M OBILE MODEL 7 7 The model was developed in a regulatory negotiation process between EPA and the affected industry. The model is fully described in EPA's reformu- lated gasoline Regulatory Impact Analysis dated December 13, 1993 (EPA 1993b). The Complex Model, which is a spreadsheet model downloadable from EPA's reformulated gasoline web page (http://www.epa.gov/ OMSWWW/rfg.htm), predicts percent change in 1990 technology vehicle emissions for a target reformulated gasoline (RFG) relative to U.S. 1990 baseline gasoline. Emissions are a function of the following input parame- ters: MTBE (methyl tertiary-butyl ether, weight percent oxygen twt%~), ETBE (ethyl tertiary-butyl ether, wt%), ethanol (wt%), TAME (tertiary-amyl methyl ether, wt%), sulfur (parts per million [ppm]), RVP (Reid vapor pressure, pounds per square inch [psi]), E200 (percent of fuel that evaporates at 200° F), E300 (percent of fuel that evaporates at 300° F), aromatics (percent by volume), olefins (percent by volume), and benzene (percent by volume). U.S. baseline emissions are calculated from MOBILES runs with U.S. industry-average gasoline. The model calculates changes from baseline emissions to emissions for the target fuel for exhaust and evaporative VOCs, air taxies (benzene, formaldehyde, acetaldehyde, and 1,3-buta- diene), and exhaust NOX. (The model does not estimate the effects of fuel reformulation on exhaust CO emissions.) The model is a statistical model based on testing data from several major programs measuring the emis- sions effects of the various fuels tested. The present and planned versions of MOBILE allow for the specification of limited fuel properties (e.g., whether or not RFG is used) but do not al- low for the specification of the detailed fuel properties that are available in the Complex Model. This means that if states or nonattainment areas choose to require a fuel with greater emissions reductions than required of federal RFG, the Complex Model must be run first to generate scaling fac- tors to apply to MOBILE output. Even then, there are questions as to how the results would be used given that COMPLEX was developed for 1990 vehicle technologies only. The fuel effects now calculated in the Complex Model should be updated so that they can appy to all model years and technology groups, not just 1990 technology vehicles, and added to future versions of MOBILE. CARB's Predictive Model estimates fuel effects for all on-road light-duty vehicles based on a broader database than that used in the development of the Complex Model; EPA should review and consider

72 M ODE1ING MOB![E-SOURCE EMISSIONS updating CARB's model. With this addition, local variations in gasoline properties can be more easily modeled for use in air-quality planning. It is important to also include CO emissions within the Complex Model. Historically, the RFG program was used for ozone reduction and the Com- plex Model was used as a tool for certifying the performance of RFGs with respect to VOCs, air tonics, and NOx. CO was not a component of the Complex Model because CO was not required for evaluating an RFG. However, there are some CO nonattainment areas that are developing RFGs. Additionally, CO is an ozone precursor. MOBTOX The MOBTOX mode] is used for estimating toxic-emissions factors for on-road motor vehicles. The mode! was originally developed as part of EPA's CAAA90 mandated study on "motor vehicle-related tonics", and is now being updated for use in both the regulatory impact analysis for the Tier 2 vehicle standards rulemaking and the development of regulations for controlling mobile-source toxic emissions (EPA l999f). The first ver- sion of the model was based on MOBILE4.1. The next version, called MOBTOX5b, was based on the modified version of MOBILE5b used in EPA's July 1998 Tier 2 study (EPA 19981~. The modified MOBILE5b model includes alternative basic emissions rates, and the effects of aggres- sive driving and air-conditioning usage. MOBTOX5b applies exhaust and evaporative toxic-adjustment factors for various vehicle classes and technologies to MOBILE5b VOC emissions factors. It estimates emissions factors for benzene, formaldehyde, acetaldehyde, 1,3-butadiene, and methyl tertiary-butyl ether GIBBET. For benzene and MTBE, the model estimates factors separately for exhaust, evaporative (diurnal and hot-soak), refueling, running-Ioss, and resting- loss emissions. The model also has the capability to account for differences in exhaust VOC toxic fractions between normal- and high-emitting vehi- cles. The toxic-adjustment factors for newer technology vehicles were de- veloped from the speciation data for 1990 technology LDVs developed for the Complex Model. The MOBILE revisions in MOBTOX5b, developed before many of the MOBILES revisions were proposed, are different from what is now pro- posed for MOBILES. Although the version of MOBTOX5b first developed for the Tier 2 rulemaking is publicly available, EPA is currently updating the air taxies model to more closely reflect MOBILES proposals. The up- dated model will be publicly released with the final Tier 2 rule. As with the PARTS model, MOBTOX must use the same vehicle-activity data as MOBILE, but has different emissions factors for toxic species. For both user convenience and model consistency, these toxic-emissions factors

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 73 should be incorporated into a future version of MOBILE and the use of the separate MOBTOX mode] should be discontinued. NONROAD EPA is currently developing a national emissions model for off-road mobile-sources. The current draft version of the model, called NONROAD, is available on EPA OTAQ web page (http://www.epa.gov/oms/ nonrdmdl.htm). The web page contains full documentation for the model, including a User's Guide and detailed technical documentation for all model estimates, inputs, and assumptions. Unlike the MOBILE model, NONROAD provides activity data as well as emissions factors. Thus, it can specifically estimate emissions inven- tories for off-road equipment. The model predicts emissions for vehicles and equipment types in the following categories: airport ground support, such as terminal tractors; agricultural equipment, such as tractors, combines, and balers; construction equipment, such as graders and back hoes; industrial and commercial equipment, such as fork lifts and sweep- ers; . cycles; residential and commercial lawn and garden equipment, such as lawn mowers and leaf and snow blowers; recreational vehicles, such as all-terrain vehicles and off-road motor- . . logging equipment, such as shredders and large chain saws; recreational marine vessels, such as power boats; underground mining equipment; and oil field equipment. The model includes more than 80 basic and 260 specific types of non- road equipment, and further stratifies equipment types by horsepower rating. Fuel types include diesel, gasoline, compressed natural gas, and liquefied petroleum gas. The model estimates six exhaust emissions (VOC, NOX, CO, carbon dioxide (CO2), sulfur dioxide (S Ox), and PM), and also estimates nonexhaust VOC emissions for six modes hot-soak, diur- nal, refueling, resting-loss, running-loss, and crankcase emissions. The model can estimate national total emissions, emissions by state, or for one or more counties or subcounties in one or more states. A more complete description of the NONROAD model's capabilities can be found in Pollack and Lindhjem (1998) and in the NONROAD User's Guide (ENVIRON 1998). The NONROAD model currently does not include emissions estimates

74 M ODE[`NG MOB/LE-SOURCE EMISSIONS from locomotives, aircraft, and commercial marine engines, and these will not be included in the official release of NONROAD. EPA is currently gathering the data necessary to estimate emissions for these important sources and plans initially to provide written guidance to states on devel- oping emissions inventory estimates for these sources based on these data. At present, EPA expects to have draft guidance documents available in 2000, with the final guidance documents available later that year. Release of NONROAD software modules for these applications is not expected be- fore 2001. EPA is currently addressing comments on the NONROAD draft model, and has not yet had the model peer-reviewed. The current draft is there- fore likely to undergo revision before the final version is released, which is expected to be sometime in 2000. As future Tier 2 vehicle standards and corresponding sulfur-reduction regulations reduce on-road mobile-source emissions, non-road emissions will become a larger fraction of the total emissions. The NONROAD model is extremely data driven, and there are many gaps in the available data. EPA should place more emphasis on im- proving both the emissions factors and activity data in this model. SPECIATE and Emissions Processing Systems Photochemical air-quality models require emissions of specific types of VOCs (e.g., formaldehyde, acetaldehyde, alkenes, aromatics, short-chained alkanes, and higher alkanes), as opposed to the total VOC mass provided by MOBILE. Providing such speciation information is done in two steps. First, the total VOC emissions, which are estimated using MOBILE and the activity data, are split into about 100 individual organic compounds and their emissions rates are determined (e.g., formaldehyde, ethane, and propane). Next, these compounds are grouped into a smaller number of lumped organic species used by air-quality models. (Using all of the indi- vidual compounds would be excessively burdensome in most photochemi- cal modeling applications, although it is possible.) Speciation is accomplished by using tables of profiles that have been developed from source testing. One such set of profiles for VOCs and PM has been developed by EPA, and is part of the SPECIATE system. The database and User's Guide are on EPA's web page at http://www.epa.gov/ttn/chie£/software.htm/#speciate. This database is now extremely out of date, especially for mobile-source emissions. On the web page, EPA itself has indicated serious concerns with the database. Re- cently, a number of emissions processing systems for producing spatially and temporally allocated speciated emissions rates for all emissions sources have been developed by contractors and universities. These in- clude the Emissions Processing System (EPS2) and the Emissions Model-

TECHNICAL ISSUES ASSOCIATED WrTH THE MOBILE MODEL 75 ing System (EMS951. The systems have a library of profiles, some of them originally based on SPECIATE, but with enhancements and additions. For specific air-quality modeling applications, users enhance the avail- able speciation profile databases and models for their own use. For exam- ple, the Auto/Oil Air Quality Improvement Research Program conducted a large test program to develop speciated automobile emissions rates, and used that information in follow-on air-quality modeling studies. CARD, using those measurements and others, has developed their own set of emissions profiles as well. PREVIOUS REVIEWS OF MOBILE Several technical reviews have been done for different versions of the MOBILE model. The reviews were inspired by field observations that in- dicated a disagreement between model predictions and actual emissions measurements. Industry groups and agencies outside EPA have spon- sored these reviews, primarily because the databases and methods under- Tying the models have not been well documented by EPA. The reviews in- cluded examination of the model's structure, assumptions, sensitivity to changes in model parameters, method of accounting for I/M emissions re- ductions, and the effects of model revisions on emissions inventories. Summary of Reviews The first review was performed for the Coordinating Research Council (CRC, which sponsors research for the automobile and oil industries) by Pollack et al. (1991~. This review was prompted by studies of field mea- surements (in particular, tunnel studies performed as part of the Southern California Air Quality Study) that indicated that the mobile-source emissions-factor models developed by EPA and CARB substantially underpredicted emissions levels (Ingalls et al. 19891. To assist in under- standing the potential sources of model underprediction, CRC sponsored a project to evaluate and compare the data and methodologies used in the then-current versions of the emissions-factor models EPA's MOBILE4 and CARB's EMFAC7E. A detailed comparative review was conducted of the databases and methods used to derive exhaust and evaporative emis- sions. The project report evaluated relative strengths and weaknesses of the models, and discussed potential causes of emissions underprediction. In addition, sensitivity analyses were performed to understand changes in model predictions in response to changes in model inputs and fixed model parameters. The American Petroleum Institute (API) sponsored reviews of

76 MODELING MOB`[E-SOURCE EMISSIONS MOBILE5a aimed at evaluating the basis and validity of updates from the previous version, with a focus on assumptions and extrapolations in the model. Heiken et al. (1994) critically reviewed the model's basic exhaust emissions rates for LDVs and HDVs, and all nonexhaust emissions rates (hot-soak, diurnal, running-loss, resting-loss, refueling). Vehicle test data, source code, and EPA internal memoranda were reviewed to replicate the methodology used to develop the model's algorithms and input data. Sta- tistical approaches for developing equations were also reviewed. In re- sponse to API interests, the report included a detailed review of the effects of fuel oxygenates on exhaust and evaporative emissions. Sensitivity anal- yses were performed to evaluate the effects of alternate assumptions and updated methodologies used for MOBILE5a. API also funded Sierra Research to review specific components of MOBILE5a. The first study evaluated and documented MOBILE5a meth- ods for estimating the benefits of I/M programs and adoption of the Cali- fornia Low Emission Vehicle (LEV) standards (Sierra Research 1994a). The evaluation included an analysis of the data that EPA used to develop I/M identification and repair effectiveness, and an assessment of the meth- odology used to estimate the benefits of evaporative system functional checks. The primary conclusion of the report was that the model likely overestimated the effects of enhanced I/M programs for both exhaust and evaporative emissions. The second Sierra Research study was focused on an important update in the MOBILE5a model-the use of I/M test data to develop LDV basic exhaust emissions rates (all previous versions of the model were based on EPA's FTP testing). Because most of the exhaust emissions-correction factors (i.e., temperature, speed) were based on the FTP, EPA developed a conversion *om the IM240 test data to FTP for cal- culating the basic emissions rates. The Sierra report critically reviewed the IM240-to-FTP conversion process, and checked the procedure with a second IM240 data set from Mesa, Arizona (Sierra Research 1994b). The U.S. Department of Transportation's (DOT) Federal Highway Ad- ministration (FHWA) also sponsored a review and evaluation of different versions of the MOBILE model. DOT's interest was prompted by require- ments in the CAAA90 and the 1991 Intermodal Surface Transportation Efficiency Act (ISTEA) for local transportation-planning and conformity determinations. Because MOBILE was the modeling tool to be used to estimate transportation emissions, DOT desired an understanding of the structure and operation of MOBILE, and a documentation of the changes that had occurred among model revisions. The report prepared for FHWA (Sierra Research 1994c) included an explanation of the basic parameters of the MOBILE model for individuals with little background in motor-vehicle emissions modeling. Detailed descriptions were then provided on the cal- culation of exhaust and evaporative emissions, methods used to determine I/M program effectiveness and other CAAA90 requirements (e.g., Tier 1

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 77 vehicle standards and reformulated and oxygenated gasoline). The report also included an evaluation of the changes in fleet average emissions fac- tors as predicted by MOBILE4, MOBILE4.1, and MOBILE5a. General Accounting Office Review of MOBILES The genesis of the 1997 report prepared by the General Accounting Of- fice (GAO) on the MOBILE model (GAO 1997, hereafter called the GAO report) was mentioned briefly in Chapter 1 of this report. The GAO report raised 14 specific concerns about the model. During the preparation of that report, EPA provided a response to an initial draft and noted the is- sues that they planned to address in the MOBILE6 update. Table 3-3 lists the issues raised about the MOBILE model in the GAO report and shows that EPA planned to address all of those issues except for road grade, heavy-duty I/M, and model uncertainty in the development of MOBILE6. In the following sections of this chapter, the GAO report issues listed in Table 3-3 are discussed and the improvements planned in MOBILE6 to address the issues are evaluated. HIGH EMITTERS Underrepresentation of emissions from high emitters in MOBILE, even in the emerging MOBILE6, is considered to be one of the chief reasons for MOBILE underpredicting real-world fleet emissions. In general, these ve- hicles are difficult to characterize statistically because the number of high emitters is relatively small and the range in their emissions is relatively large. Priority should be given to further improving this very important component of MOBILE. Exhaust High Emilters The problem of high emitters, and their correct representation in the databases for MOBILE, was identified as Issue 6 in the GAO report. Characterizing high emitters requires understanding of not only their level of emissions but also their population and activity. High emitters will generally represent a disproportionately high fraction of the total fleet emissions predicted by MOBILE. Concerns have been expressed that, be- cause of potential recruitment bias in FTP testing used as the foundation of MOBILE, the model underestimates emissions of the overall real-world fleet. Typically, recruitment acceptance rates are less than 25%. It is the- orized that owners of vehicles that are high emitters will be reluctant to

78 MODELING MOB/1E-SOURCE EMISSIONS TABLE 3-3 Report Major Limitations in MOBILE5a Model Identified in GAO Areas of concern regarding MOBILE cited in GAO report 1. Emissions estimates for higher speeds, especially speeds in excess of 65 miles per hour (mph). 2. Representation of emissions from rapid acceleration and deceleration, including aggressive driving behaviors. Does EPA plan to ad- dress this . . Issue In MOBILE6? 3. Representation of emissions immediately after engine start-up, known as cold-start emissions. 4. Representation of emissions from air conditioner use. Representation of emissions from road grades, such as when a car climbs a hill. Representation of high-emitting vehicles in MOBILE's supporting database. Representation of emissions from lower-polluting fuels, especially fuels with lower volatility. 8. Representation of emissions-system deterioration for vehicles with 50,000 or more odometer miles. 9. Emissions estimates and assumptions for vehicle I/M programs. 10. Estimates and assumptions for nontailpipe evaporative Yes emissions when the vehicle is not operating. Emissions estimates and assumptions for the I/M of No HDVs those with a gross vehicle weight of 8,501 pounds or more. Yes Yes Yes Yes No Yes Yes Yes Yes 12. Data characterizing vehicle fleet. 13. Greater distinctions in roadway classifications. 14. Quantifying the uncertainty of the model's estimates. Yes Yes No Source: GAO 1997. submit their vehicles for intense emissions testing. Conversely, domestic automobile manufacturers, who have supplied much of the later model- year FTP data for use in developing MOBILES, have expressed the opinion that their recruitment program, which offered free repairs, would actually

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 79 encourage recruitment of higher emitters. Even if the latter is true, the data submitted by domestic automobile manufacturers are composed of vehicles roughly 2-3 years old where high emitting vehicles should be a very small fraction of these vehicles. The California Air Resources Board has used a high-emitter correction factor in its mobile-emissions model (EMFAC) for a number of years. This correction factor is based on random roadside emissions testing surveys of the real-world fleet and has histori- cally resulted in increasing the emissions estimates of EMFAC. EPA decided to use a high-emitter correction factor in development of MOBILE6. EPA selected first-year IM240 test data from Ohio for develop- ment of this correction factor (EPA l999g). The advantage of this approach is that it presents a relatively unbiased sample of the real-world fleet. EPA did recognize the disadvantages of using IM240 data compared to FTP testing, which are described in Chapter 4. EPA defined high emitters as those vehicles emitting more than twice the FTP standards for VOCs and NOX and more than three times the FTP standard for CO. Vehicles in these emissions categories were felt to have significant problems with their emissions control systems. The high-emitter correction factors devel- oped by EPA for MOBILE6 for running emissions result in a significant increase in projected emissions, ranging from +30% for CO to +150% for VOC for 1988-1993 port fuel-injection cars. EPA also used the high-emit- ter correction factor developed for running emissions to adjust the start- emissions component of MOBILE6 (EPA l999h) and also to address FTP recruitment bias. EPA's use of IM240 data to generate a high-emitter correction factor for MOBILE6 appears to be a step in the right direction toward improving the accuracy of MOBILE. Justifying this approach is the fact that numerous tests of MOBILE's validity tend to indicate MOBILE has been underesti- mating real-world fleet emissions (see discussions in Chapter 4), and EPA's new high-emitter correction factor result in increasing the base emissions rates in MOBILE6. As discussed in the evaluation section of Chapter 4, there are many rea- sons to suspect that even IM240 data fall short of truly reflecting the real- worId fleet emissions, particularly because it is not truly representing the high-emitter category. For example, remote-sensing data, discussed in Chapter 4, indicates the shortcomings of IM240 data representation of high emitters. Forthcoming roadside-pullover loaded-mode testing being conducted by the California Bureau of Automotive Repair might provide the greatest insight into how well MOBILE6 reflects the contribution of high emitters to total fleet emissions. Because start emissions are a major component of the total fleet emis- sions projected by MOBILE, the accurate characterization of start emis- sions *om high emitters is critical. EPA has acknowledged that confidence

80 MODELING MOB/LE-SOURCE EMISSIONS intervals around the average start-emissions levels of the high emitters are quite large due to high scatter and small sample sizes. Additionally, the representativeness of applying the IM240 high-emitter correction fac- tors developed for running emissions to start emissions must be ques- tioned. Although using the running-emissions correction factor to correct start emissions is undoubtedly better than applying no correction factor, running-emissions correction factors would be less likely to reflect the in- creased catalyst light-off temperatures associated with older, higher-emit- ting vehicles. Higher catalyst light-off temperatures, one of the common characteristics of high emitters, would cause the adjustment for cold-start emissions to be relatively higher than the adjustment for running emis- sions. This is because running emissions are measured for a fully warmed engine. Evaporative High Emitters As is the case for tailpipe emissions, the distribution of evaporative emissions among the in-use vehicle fleet is highly skewed towards high emitters. Skewness is characteristic of all three evaporative emissions types: hot-soak, running-loss, and diurnal. Because the measurement of evaporative emissions is difficult, available data are limited and come pri- marily from a series of studies sponsored by the CRC. EPA has used these data (as well as data obtained by EPA) in the formulation of MOBILES that has a special treatment for high evaporative emitters (EPA l999i). These relatively recent evaporative emissions measurements suggest that evaporative emissions of VOCs are greater than tailpipe emissions. A1- though older model, carbureted vehicles typically have higher evaporative emissions than newer model, fuel-injected vehicles, high emitters are found in the newer as well as older model vehicles. Running-Ioss emissions resulting from liquid leaks are probably the most important cause of evap- orative high emitters. DRIVING-CYCLE ISSUES Driving Cycles- Real-World Driving versus FTP Speeds, Accelerations, and Other Engine-Load Conditions The FTP for LDVs was designed as a certification test for measuring the emissions from new vehicles. As such, it is a compromise between the de- sire to have a representative sample of actual vehicle operating conditions

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL ~ ~ and the requirements of actual testing. The latter requirements include equipment limitations, test costs, and test time. The need to measure vehicle emissions over a range of conditions that simulate actual driving has been recognized since the start of vehicle emissions-control programs. However, the original certification test cycle used only steady-state operation and did not include start emissions. The data for the present FTP driving cycle was obtained from driving survey data taken in downtown Los Angeles during the 1960s. The basic test cy- cle, which was first used in 1972, measures transient vehicle emissions. including a cold start. A hot-start portion was added in 1975. The test cycle simulates driving a route of 7.5 miles with an average speed of 19.6 miles per hour (mph). The weighted sum of a cold-start (12- hr engine-off time or soak time at 75°F) and a hot-start (10-min soak time) trip over this route is used to compute the overall emissions. In practice, the trip is divided into two parts. The first part represents transient emis- sions after start. The test for this part lasts for 505 seconds and covers 3.59 miles. The second part represents stabilized emissions with a warmed engine and catalyst. This partlasts 867 seconds and covers 3.91 miles. The average speeds of the transient and stabilized parts are 25.6 mph and 16.2 mph. In the actual measurement, the stabilized part is measured only one time, immediately following the cold-start test, and the results of this mea- surement are assumed to apply to both the hot-start and the cold-start trips. The measured emissions consist of three parts: Bag 1, representing the cold-start transient emissions; Bag 2, representing the stabilized emis- sions after the engine is warm; and Bag 3, representing the hot-start tran- sient emissions. The certification emissions from the FTP are computed in the following manner: Average emissions = 43% (cold-start trip) + 57% (hot-start trip) Cold-start trip (g/mi) = t(3.59 mi) (cold-start phase g/mi) + (3.91 mi) (stabilized phase g/mi) ~ / (7.5 mi) Hot-start trip (g/mi) = [~3.59 mi) (hot-start phase g/mi) + (3.91 mi) (stabilized phase g/mi) ~ / (7.5 mi) Combining these three equations gives the final weights for each phase in the certification test procedure. Average emissions = 0.206 (cold-start phase) + 0.521 (stabilized phase) + 0.273 (hot-start phase) (3-1)

82 M ODE[!NG M OBILE-SOURCE EMISSIONS The certification results, with units of grams per mile, do not provide any spatial location for start emissions. The separation of start and run- ning emissions proposed for MOBILES is discussed later. At the time the cycle was established, limitations of existing dynamom- eters restricted the range of possible accelerations. Thus the maximum acceleration in the FTP was limited to 3.3 mph/e. The basic speed-time trace for the 7.~-~nile trip is shown in Figure 3-3. This cycle is sometimes referred to as the "LA4" cycle or the urban dynamometer driving schedule (UDDS). The selection of a downtown route limited the speeds in the FTP. As Figure 3-3 shows, there is a small portion of the cycle in which vehicle speeds exceed 50 mph. Otherwise, the preponderance of the vehicle speeds is below 30 mph. In addition, the limitation of the acceleration rate to 3.3 mph/s does not provide measurements over higher accelerations that are experienced in everyday driving. This implies that the FTP is not repre- sentative of modern urban driving. These limitations in the demand placed on the vehicle and its engine in the test cycle are important because emissions increase with engine load. The emissions of a particular pollutant from engine operation are equal to the product of the mass flow rate of the exhaust and the mass fraction of the pollutant species. As the load on the engine increases, more fuel and air are required, producing a higher mass flow rate of exhaust. In addi- tion, when optimizing power at high engine loads, the fuel metering sys- tem in a typical vehicle will supply the engine with a fuel-rich mixture causing a significant increase in the mass fraction of VOC species and CO. Thus higher engine loads especially those above the loads tested in the FTP can lead to very large emissions rates. The GAO report on the MOBILE model identified three issues related to the underrepresentation of high-load conditions in the model. The first two of these were the absence of high-speed driving conditions and the lack of "aggressive" driving operating conditions with high amounts of accel- eration and deceleration. Revisions to the MOBILE model should address these considerations as discussed below. A third issue, the absence of any accounting for road-grade effects, is discussed in a separate subsection below. These effects can be better understood in terms of the equations for engine load presented below. The load on the engine of a vehicle can be expressed in terms of its vari- ous components by the following equation (Bauer 1996~: P =—tPrf + Pad + Pa + Pg 4 + Pacc (3-2)

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 83 60 50 40 a, Q in ._ 30 20 10 o Bags 1 and 3 . ~ Bag 2 - Stabilized Emissions f) ~ 0 200 400 600 800 1000 1200 Elapsed time (seconds) 1400 FIGURE 3-3 FTP driving cycle for light-duty vehicles. Source: Sierra Research 1994b. where p = the engine power requirement, y<~ = efficiency of the drivetrain, Prf = power required to overcome rolling friction resistance, Pad = power required to overcome aerodynamic drag, Pa = power required to accelerate the vehicle, pg = power required for grade climbing, and Pacc = power required by vehicle accessories (such as air-conditioning). These individual power components are given by the following equations: P. f W V P. p a, CD Af V 3 p a W V P W V b (.3-3)

84 MODELING MOBI1E-SOURCE EMISSIONS where f = coefficient of rolling friction, W = weight of the vehicle, V = vehicle speed, Pa = density of air, CD Al = frontal area of the vehicle, = aerodynamic drag coefficient for the vehicle, a = vehicle acceleration, g = acceleration of gravity, and b = slope of the grade. These equations show that the power requirements for a given vehicle (and hence the emissions in grams per second) increase with increases in the speed, acceleration, and grade. In addition, the role of grade is seen to be similar to that of acceleration; the power required to climb a 10% grade is essentially the same as the power required to accelerate at 10% of the acceleration of gravity. The most significant accessory load is that of the vehicle-air conditioner. The increased exhaust flow with increased load is only one source of increased emissions. In addition, there is the potential that under such operating conditions the computer emissions-control systems will perform less effectively than during operation within the FTP speeds and loads. This raises two significant issues: (1) are vehicle emission controls as effec- tive as implied by the FTP-based emission standards, and (2) how realistic are the MOBILE emission estimates based on FTP data? The technical community has long known about the absence of high speeds and accelerations from the FTP. The CAAA90 required EPA to develop a new emissions test that accounted for this real-world driving. In response to this mandate, EPA sponsored various studies to determine the pattern of vehicle speeds and accelerations that are encountered in every- day driving. Based on these studies, they have developed a Supplemental Federal Test Procedure (SFTP). This SFTP procedure requires vehicle manufacturers to certify vehicles over an additional test cycle known as the US06 cycle. This cycle, whose speed-time trace is shown in Figure 3-4, has speeds as high as 80 mph and a maximum acceleration of 8.4 mph/e. The average speed of this cycle is 48.3 mph. Certification to this new cycle will be phased in between the 2000 and 2004 model years. This test proce- dure 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 emission control will come from reduced use of fuel-

TECHNICAL ISSUES ASSOCIATED WITH THE M OBILE MODEL 85 90 80 70 E 60 - ~ 50 i, a, 40 ._ s 30 20 10 o : A , ~ . fly 0 100 200 300 400 500 600 Elapsed time (seconds) FIGURE 3-4 The EPA supplemental driving schedule (USER. Source: Sierra Research 1994b. rich mixtures at higher loads. Although this cycle does not address the issue of simulating real-world driving in MOBILE, it does include some observed driving speeds and accelerations that are much higher than those used in the FTP. Accounting for Roar! Grade Equations 3-2 and 3-3 show that road grade plays a role that is as im- portant as vehicle acceleration. The effects of road grade are not included in the FTP or the US06 cycle. It is not a part of the MOBILE model, and EPA does not plan to add road grade effects to MOBILE6. From a stand- point of vehicle certification, the high-acceleration loads in the US06 cycle should ensure that vehicle emissions-control systems would be operative during grade-climbing operations in real-world driving. However, MOBILES will not be able to model the effects of road grade on emissions in a local area. This will be particularly important in urban areas, such as

86 M ODELING M OB![E-SOURCE EMISSIONS Denver and Spokane, which have a significant amount of vehicle operation at grade. In the development of the US06 cycle, EPA obtained some information on driving conditions at grade, but these were not incorporated in the final certification cycle. One possible approach for including grade operations in a future version of the MOBILE model would be the use of grade-correc- tion factors, similar to the facility-correction factors discussed below. However, the use of such factors would require the local planning agencies to specify the amount and type of driving on grades. EPA has stated that they plan to include grade operation in a future version of the MOBILE model. To make such a future model useful to local planning agencies, EPA should discuss with them the types of data that they require on grades so that the model is designed to best use available data. The im- portance of grade can be estimated from modal or instantaneous models currently under development (Guenstar et al. 1998; Barth et al. 1998) Speed-Correction Factors and Facility Driving Cycles The present version, MOBILES, as well as some earlier versions have used speed-correction factors (SCFs) to estimate emissions under operat- ing conditions that are different *om the FTP. These SCFs are based on a series of test cycles with different mean speeds. The FTP emissions (at the mean FTP speed of 19.6 mph) are multiplied by the SCF for a desired speed to give the emissions at the desired speed. SCF are a function of vehicle type, model year, and pollutant species. SCFs are applied to emissions rates expressed in terms of grams per mile. This emissions rate per unit distance, MD, is simply related to the mass flow rate of the pollutant per unit time, M, and the vehicle speed, V, by the following equation: MD = M (3-4) Although increased engine loads will cause the emissions rate, M, to increase with speed, the gram-per-mile emissions rate might increase or decrease depending on the relative changes in mass emissions rate and vehicle speed. This effect can be seen in the SCFs shown in Figure 3-5 (Sierra Research 1994c.) This figure shows the SCFs for 1990 model-year gasoline-powered passenger cars. In this figure, SCFs for all species is 1

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 87 5.0 4.5 4.0 3.5 o ``, 3.0 o o C' 1.0 0.0 I ~ I 15 ~ art_ 0 10 20 30 40 50 60 70 Speed (mph) CO NOx FIGURE 3-5 MOBILES speed-correction factors for 1990 model-year, gasoline-powered passenger cars. Source: Sierra Research 1994c. at the mean FTP speed of 19.6 mph. For low speeds the SCFs increase dramatically. At these low speeds, the emissions rate is approaching the idle mass emissions rate and the division by a small velocity causes a large emissions rate in grams per mile. Note that expressing emissions factors in units of g/mi creates a problem in describing vehicles at rest (idling). For vehicle speeds between 20 and 50 mph (for VOCs and CO) or 0 to 20 mph (for NOX), the SCF decreases, showing a decrease in gram-per-mile emissions with speed. In this region, the mass emissions rate is actually increasing slightly, but not as fast as the speed. At high speeds, the in- crease in emissions rate is larger than the increase in speed, so the SCF increases. The driving cycles for the various intermediate speeds that are used to determine the SCFS assume that all types of driving can be characterized by a single parameter, the vehicle speed. However, actual emissions also depend on the engine load, which is determined not only by the vehicle speed, but by a combination of vehicle speed and acceleration. The SCF approach, which assumes that all variation in operating conditions can be characterized by a single parameter, the vehicle speed, would be appropri- ate for emissions inventories if the speed correction factors in fact ac-

BB MODELING MOBILE-SOURCE EMISSIONS counted for all types of vehicle operation (see the description in Chapter 5 of an approach that relates emissions to vehicle-specific power). However, this approach is not useful for roadway modeling where the engine opera- tion will be significantly different, for a given vehicle speed, depending on the roadway type. For example, an average speed of 30 mph on a surface street would indicate free-flowing traffic. However, the same average speed of 30 mph on a freeway could indicate congested driving with signifi- cant amounts of acceleration and deceleration and significantly higher emissions. EPA will be using facility-correction factors in MOBILES (EPA l999j). These correction factors will be able to account for differences in overall vehicle operation, on a number of roadway types, under different levels of congestion. Sierra Research has developed specific cycles for these differ- ent facility types based on real-world driving data (Austin and Carlson 1997~. EPA has used these cycles to obtain data on the difference in emis- sions between FTP operations and the operations on the facility cycles. Various parameters, including maximum speed and acceleration for these cycles, are shown in Table 3-4. The application of facility-correction factors in MOBILES is similar to the use of SCFs. Instead of having a single set of SCFs, MOBILES will have factors that account for both speed and facility type. EPA has used emissions measurements on the various cycles to develop the ratio of emis- sions on a particular facility type to emissions on the FTP. Once the FTP emissions are computed by the usual MOBILE equations, the emissions for a particular facility type can be found by multiplying the FTP emis- sions by the ratio for that facility type, speed, and level of congestion. Overall emissions for a local area will require the user to develop a dis- tribution of VMT on various facility types in the region. Modeling to deter- mine the impact of new highway projects can use expected travel demand for a particular roadway type to get improved estimates of emissions for the existing and the proposed roadway configurations. The use of facility- specific correction factors should improve the calculations of emissions for different roadway types in MOBILES. This addresses Issue 13 raised in the GAO (1997) report, although further disaggregation might be needed. Other approaches, including the use of emissions rates in units of grams per second or fuel based emissions rates (e.g., in units of grams of emis- sions per gram of fuel used) could provide a better approach to emissions modeling, particularly at low speeds. START EMISSIONS Cold engines and cold catalysts have much higher emissions than those for normal operating temperatures. This makes cold starts a critical com-

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 89 TABLE 3-4 Parameters for New Facility-Specific Driving Cycles, with LA4 Cycle Used in FTP Included for Comparison Roadway Classification Average Maximum Maximum Cycle Cycle (LOS Indicates Levelof Speed Speed Acceleration Time Distance Serviced (mph) (mph) (mph/s) (s) (miles) LA4 Cycle Used in FTP Freewa77y, High Speed 63.2 74.7 2.7 610 10.72 Freeway, LOS A-C 59.7 73.1 3.4 516 8.55 Freeway, LOS D 52.9 70.6 2.3 406 6.96 Freeway, LOSE 30.5 63.0 5.3 456 3.86 Freeway, LOS F 18.6 49.9 6.9 442 2.29 Freeway, LOS "G" 13.1 35.7 3.8 390 1.42 Freeway Ramps 34.6 60.2 5.7 266 2.56 Arterials/Collectors LOS 24.8 58.9 5.0 737 5.07 A-B Arterials/Collectors LOS 19.2 49.5 5.7 629 3.36 C-D Arterials/Collectors LOS 11.6 39.9 5.8 504 1.62 E-F Local Roadways 12.9 38.3 3.7 525 1.87 Non-i~reeway area wide 19.4 52.3 6.4 1,348 7.25 aLevel of service (:LOS) is a measure of traffic congestion. According to the Transportation Research Board, LOS A has the least congestion and LOS F has the most congestion. LOS G was created to define a "subset of LOS F driving under the worst conditions routinely observed." Source: EPA l999j. ponent to account for in the model. As noted in Equation 3-1, FTP emis- sions in grams per mile are based on the weighting of the cold-start, hot- start, and stabilized portions of the FTP. MOBILES computes separate results for each phase and users can sup- ply weightings for each phase, which might be different from the FTP weightings shown above. User-supplied weightings can be used to account for differences in start activity in a local region. However, such accounting is not satisfactory in many applications, such as air-quality modeling, be- cause it allocates the start emissions in terms of grams per mile to an en- tire trip as opposed to the specific start location. Start emissions was the third issue identified in the GAO report (GAO 1997) on the MOBILE mod- el. Recall that MOBILE subdivides vehicle classes into technology groups. Each technology group is treated separately in the analysis of start emis-

Start emissions ratio 90 MODELING MOBl[E-SOURCE EMISSIONS signs in MOBILE6. The following discussion refers to obtaining sets of re- gression equations. Such equations are obtained for each technology group. In this way, MOBILE models the differences in the vehicles in the fleet. In addition, MOBILE6 divides the fleet into normal and high emitters for modeling the effects of inspection and maintenance. (This is a change from MOBILES where the fleet was divided into four emissions regimes: nor- mal, high, very high, and super.) Separate regression equations must be developed for each regime. Because of the large number of combinations of regimes and technology groups, a large number of vehicle tests is required to ensure that each group is accurately represented in the model. EPA plans to use a new start methodology for MOBILE6 (EPA l999h). In the new scheme, start emissions will be based on the difference between the emissions measured in the cold-start (or hot-start) phase and the emis- sions measured over the same 505 second driving cycle without a start. (The cold-start and hot-start emissions measured by this process are sub- sequently adjusted for soak time in MOBILE6, as described below.) To characterize these emissions, EPA obtained data for 77 vehicles tested over the 505 second cycle with no start (EPA l999k). EPA called this test the hot-running 505 cycle. These vehicles were also tested using the full FTP. EPA developed regression equations for this small vehicle data set, which related the hot-running 505 cycle emissions to individualcompo- nents of the FTP. These regression equations could then be used to com- pute the hot-running 505 cycle emissions for the large database EPA has on vehicles with conventional FTP tests. For 1981 to 1993 model-year ve- hicles, for example, there were 4,416 passenger cars and 1,205 trucks in the database for FTP emissions results. The cold-start emissions are then found, for each vehicle in the data- base, as the difference between the cold-start phase (Bag 1) emissions and the hot-running 505 cycle emissions. (The latter is found by regression based on the 77-vehicle data set.) The units of the start emissions are grams per start. The resulting data on emissions as a function of mileage are then used to develop regression equations. EPA did not provide any statistical results for the resulting regression equations. The regression equations give the start emissions for a vehicle, which has reached the ambient temperature. The FTP requires an engine off time (soak time) of 12 hr to reach this tempeture. MOBILE6 will use a modification of relations developed by the CARB to account for the effect of soak time on start emissions (CARB 1996a). CARB made measurements of start emissions for various soak times and developed a regression equa- tion for the start emissions ratio, defined below, as a function of soak time. Emissions at given soak time Emissions at 12 - hr soak time (3-5)

TECHNICAL ISSUES ASSOCIATED WITH THE M OBILE M ODES 9 ~ EPA developed a modified version of the CARE relationship for the start- emissions ratio to be consistent with EPA data on hot starts and cold starts. The approach to start emissions proposed for MOBILES allows for an improved modeling of these emissions. In MOBILES, such emissions were counted as part of the gram-per-mile emissions associated with overall vehicle operation. Having a separate model for start emissions provides a more accurate representation of these emissions. In addition, the separa- tion of start emissions allows such emissions to be identified with a partic- ular location where the emissions occur. This should provide more accu- rate results for air-quality models. Further improvements are possible. A recent review of EMFAC7G (Pol- lack et al. 1999a) has shown problems with the CARE start-emissions ra- tio used by EPA. These problems relate to activity data on the number of trips per day (and hence number of starts) and the small data set used to characterize start emissions as a function of the length of time the vehicle has been turned off. Future data collection efforts should include mea- surements with the hot-running 505 cycle. This will increase the database for measured start emissions. Additional studies should be done to im- prove the relationship between soak time and start emissions. Such stud- ies should be designed to develop an understanding of the effect of soak time on start emissions at non-FTP temperatures. It is known that VOC and CO emissions increase dramatically at temperatures below about 55°F and that the start mode accounts for much of this increase. IN-USE DETERIORATION Even with reasonable maintenance, vehicle emissions increase with mileage and age due to deterioration of engine and emissions-control com- ponents. In MOBILES and previous versions, EPA has projected major increases in in-use deterioration after 50,000 miles. For instance, MOBILES had a nearly 10-fold increase in VOC emissions over certifica- tion levels at 100,000 miles. The data analysis for this so called "dog leg" or "kink" effect has been questioned (Sierra Research 1994b). The GAO (1997) report identified deterioration above 50,000 miles as the eighth issue of concern for MOBILE. The substantial deterioration of emissions- control performance has lead to a tightening of certification standards, an increase in warranty requirements, and a requirement for local I/M pro- grams. EPA has made a major attempt to better characterize in-use dete- rioration and its effects on emissions in MOBILES. Yet, questions remain about the accuracy of these newly revised deterioration rates.

92 M ODE[`NG M OB![E-SOURCE EMISSIONS Issues with In-Use Deterioration Automobile manufacturers have made major strides in the last decade or so to reduce in-use deterioration of vehicle emissions-control equipment. Improvements in catalyst longevity, replacement of carbureted systems with fuel injection, conversion to platinum spark plugs, and the introduc- tion of closed loop air-fuel ratio control are among the advancements that clearly have improved the longevity of emissions-control performance. These events have motivated EPA to consider major revisions to in-use deterioration factors in MOBILE6. EPA primarily has relied on an FTP data set it collected from public vehicle-recruitment programs as well as an even larger FTP database sup- plied by automobile manufacturers to make adjustments to in-use deterio- ration in MOBILE6. Concerns about the representativeness of these data have been raised. As mentioned previously, EPA's public vehicle-recruit- ment programs had acceptance rates typically less than 25%, raising con- cerns about recruitment bias. The automobile manufacturer's database represented relatively young, high-mileage vehicles. This raised concern that more typical aging might result in higher deterioration rates. Other issues include effects of deterioration on start emissions and the overall concern that the FTP based cycle is not representative of more contempo- rary driving cycles. Effects of deterioration may not be evident under the relatively low-load of the FTP. Adjustments to In-Use Deterioration in MOBILE6 EPA has addressed issues regarding in-use deterioration in MOBILE6 by using all available new FTP data, splitting these data into start and running modes, and adjusting these data to IM240 tests (EPA 19991; EPA l999m). IM240 programs provide large samples of emissions test data using a test cycle that better emulates actual on-road driving compared to other I/M tests. Concerns about recruitment biases in the FTP data, dis- cussed previously, motivated EPA to use IM240 data to correct deteriora- tion estimates computed from FTP data. This adjustment required nu- merous assumptions and data conversions and introduced uncertainties in the estimates. However, it was felt that the end result was an improve- ment in estimating deterioration of the real-world fleet. MOBILE6 In-Use Deterioration Compared to MOBILES Figures 3-6, 3-7, and 3-8 present comparisons of MOBILES and pro- posed MOBILE6 VOC running-emissions factors for Tier 0 vehicles (EPA 199911. As can been seen, deterioration rates for MOBILE6 are substan-

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 93 tially reduced after 50,000 miles with the "kink" eliminated. Newer mode] years show greater reduction in in-use deterioration than older models. Conclusions on In-Use Deterioration EPA has made major changes to lower in-use deterioration in MOBILE6 in response to new FTP data and IM240 data that indicate newer model vehicle emissions-control systems are substantially more durable than older models. However, concerns remain about the new proposed MOBILE6 in-use deterioration rates. These include a small database for start emissions, representativeness of the FTP database (and even I/M databases, particularly for high emitters and typical aged vehicles), and overall accuracy of the FTP-based cycle to represent current real-world driving cycles. Because of these concerns, EPA should establish a long- term testing program to characterize the in-use deterioration of represen- tatively aged new-technology vehicles using a driving cycle more represen- tative of real-world driving conditions. It should also be noted that lower in-use exhaust deterioration rates might increase the discrepancy between MOBILE6 emissions estimates and emissions estimates obtained through tunnel and ambient studies. Chapter 4 describes these field observations in detail. HC (g/ml) 41 2 4 MOBILES 0 50 100 150 MILES (x1000) 2 1 o FIGURE 3-6 Comparison of MOBILES and MOBILE6 hydrocarbon (HC) emissions factors as a function of mileage for 1981 model-year passenger cars. Note that this report usually uses the term VOCs as opposed to HCs to refer to the general class of gaseous organic compounds. Source: EPA 19991.

94 M ODE[!NG M OB/1E-SOURCE EM!SS/ONS HO Carrel 4] 3- o ~ T 50 T —= tdOO~ 1 1W 150 1 - O FIGURE 3-7 Comparison of MOBILES and MOBILES hydrocarbon (HC) emissions factors as a function of mileage for 1987 model-year passenger cars. Source: EPA 19991. INSPECTION AND MAINTENANCE ISSUES Emissions Estimates ant] Assumptions for l/M Programs The use of the MOBILE model for the calculation of the benefits of vehi- cle TIM programs is one of the most controversial applications of the model. This was identified as Issue 9 in the GAO (1997) report on the MOBILE model. Individual nonattainment areas receive emissions-reduction cred- its in their State Implementation Plans (SIPs) based on the predictions of the MOBILE model. The current version of the model, MOBILES, gener- ally gives the most credit for a centralized inspection procedure (i.e., one in which the testing station does not do repairs) using a dynamometer test known as IM240. The IM240 test uses a driving cycle, which is a reduced portion of the FTP cycle, to better characterize real-world emissions com- pared with a 2-speed idle test of the vehicle. Various studies of I/M procedures have questioned the effectiveness of such programs, especially the benefits predicted by MOBILES. The Na- tional Highway System Designation Act, enacted in 1995 allowed states to provide alternative methods for determining the benefits of these pro- grams for their SIPs, provided they could adequately demonstrate the ben- efits of such programs.

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 95 HC ~m4 - O- o ~ A`: - 000, MOBILES MOBILE6 100 1SO - 4 2 FIGURE 3-8 Comparison of MOBILES and MOBILES hydrocarbon (HC) emissions factors as a function of mileage for 1992 model-year passenger cars. Source: EPA 19991. The effectiveness of I/M programs relies on a combination of technical and behavioral effects. The actual inspection program uses a short test that measures the vehicle's emissions. The short test is necessary because an FTP-type test would be too costly and too time consuming. The short test is designed to identify vehicles that have high emissions. The cut points for the tests the emissions results that identify vehicles as failing the test are set so that there is very little chance that a vehicle meeting the standards will fail the test. This means that there is a significant prob- ability that some vehicles that do not meet the standards will pass the test. Further, vehicles that fail are repaired only to the extent that they will pass the short test. Thus the cut points might affect the repair effec- tiveness. Other exhaust tests are possible in addition to the IM240 test. The sim- plest exhaust emissions test measures idle emissions. This is usually sup- plemented by a measurement with a no-load engine speed of 2,500 RPM. This test (or test combination) does not require a dynamometer. An alter- native test currently used is known as the acceleration-simulation mode or ASM. This test uses a single engine load to simulate a particular vehicle operating point. Because the dynamometer that is used for this test does not have to simulate the instantaneous changes required for the IM240 test, a less costly dynamometer can be used.

96 MODELING MOBI1E-SOURCE EMISSIONS In addition to the exhaust emissions test, tests are also done on the evaporative control system. It is also possible to conduct tests for tamper- ingi as part of an I/M program, which are also done in addition to the ex- haust emission test. The technical assessment of the effectiveness of I/M programs requires an assessment of the rate at which vehicles fail the short test (alternately called the failure rate or identification rate) and the effectiveness of their repair. The model must also account for the effect of a cost cap that gov- erns repairs made to vehicles failing inspection. If the vehicle's repairs cost more than a predetermined amount, additional repairs need not be performed on the vehicle. The CAAA90 set this cost cap in 1990 at $450 for late-model vehicles, which increases with inflation. The behavioral aspects of I/M programs can also have an impact on their effectiveness. Individuals who know that their vehicles will be in- spected might maintain their vehicles better. This would result in a Tower failure rate than predicted and in lower emissions that might not be cred- ited to the program. It is also possible, particularly in decentralized pro- grams where test and repair are combined, for some inspectors to fraudu- lently pass a vehicle that should otherwise fail. One common problem with older vehicles was that individuals could adjust a repaired vehicle so that its performance would improve, but its emissions would increase. This is less of a problem with modern, computer-controlled vehicles. Fi- nally, owners of vehicles that are subject to inspection might avoid inspec- tion completely or their vehicles might never be repaired after an initial failure. MOBILE attempts to capture some of these effects by using input data for such vehicles that have not been inspected. The analysis of I/M programs is based on the division of the fleet into emissions regimes. MOBILES uses two regimes, normal emitters and high emitters, whereas, previous versions of MOBILE used four regimes, nor- mal, high, very high, and super. Vehicles are assumed to move from a lower-emitting regime to a higher-emitting regime as a result of some fail- ure of an emissions component. Detection and repair of the failed compo- nent moves the vehicle from a higher-emitting regime to a lower-emitting one. This component failure is contrasted to the normal deterioration that is expected for wet/-maintained vehicles. The benefits of I/M have been computed in a separate model known as the TECH model. The results of this model are used as inputs to MOBILE. Starting with MOBILES, the TECH model will no longer be iTampering as defined in the MOBILE model is the malfunctioning of one or more emissions- control device due to either deliberate disablement or mechanical failure.

TECHNICAL ISSUES ASSOCIATED WrrH THE M OBILE M ODES 97 used. The calculations previously done in this model are integrated into MOBILES. The I/M benefits for running emissions are computed as follows in MOBILES for a program using IM240 (EPA l999n): 1. All emissions benefits come from identifying and repairing high-emit- ting vehicles. 2. The identification (failure) rate for high emitters is determined by re- gression equations, which are functions of the cut points used in the IM240 tests. 3. High-emitter emissions are independent of vehicle age or mileage. Repaired emissions are computed as a multiplicative adjustment factor times the normal emissions level. The normal emissions, and hence the re- paired emissions, are a function of vehicle mileage. 4. The multiplicative adjustment factor used in Step 3 is a function of vehicle age. This factor is found *om regression equations based on data from the Arizona IM240 inspection program. 5. The after-repair emissions are adjusted by further multiplicative cor- rection factors to account for more stringent cut points and for technician- training effects. 6. The final emissions after the I/M process are the sum of the following components: the emissions of the normal fraction of the vehicle fleet, that is assumed not to be inspected or repaired; the emissions of the high-emitters fraction of the vehicle fleet that is not identified and repaired; the emissions of high emitters that have been identified and re- paired; the emissions of waivered vehicles, such as high emitters that have been identified but not fully repaired because of a cost limit, or older vehicles not required to be tested; and disappearing" vehicles that never show up for their emissions test or fail an initial test and never receive full repairs or a cost waiver. . A similar analysis is applied to start emissions as well. Because I/M exhaust measurements assume a fully warmed vehicle, start operations are not tested. The identification and repairs to reduce start emissions are assumed to occur through the identification and repair of failures in run- . . . mng emlsslons. The analysis outlined above is done on a year-by-year basis. MOBILE has fixed mileage accumulation rates (for a given vehicle class) so that there is a unique relationship between the average age and average mile-

98 MODE[/NG MOBILE-SOURCE EMISSIONS age accumulation in MOBILE. The year-by-year emissions benefits for a given starting year are then combined to account for deterioration between inspection cycles. The resulting process is called the sawtooth method be- cause of the appearance of the resulting graph. Emissions drop after in- spection and repair. They then are assumed to increase due to normal ve- hicle deterioration until the next inspection cycle. This process is repeated over the lifetime of the vehicle. Modifications to this process are made when the I/M program cans for off-cycle inspections. Such inspections might be required upon a change of ownership or by the use of remote-sensing detectors to identify high-emit- ting vehicles that can be called in for inspection. Several researchers have questioned the magnitude of the emissions benefits estimated for I/M programs in the MOBILE model. These studies are described below. Much of this criticism has been based on the data collected using remote-sensing devices (RSD). These devices measure the concentration of emissions in the exhaust as the vehicle is passing by a detector. This method has the potential advantage of making measure- ments with actual on-road driving. However, the vehicle is measured un- der a single operating mode and that mode might differ as the RSD is moved from location to location. Early remote-sensing studies (Stedman 1989; Lawson et al. 1990) and one study based on ambient measurements (Scherrer and Kittelson 1994) found very little change in on-road fleet emissions that could be attribut- able to I/M programs. Wenzel (1999) has analyzed extensive data from remote-sensing measurements of vehicles in the Arizona I/M program. Over 450,000 vehicles with both IM240 and remote-sensing measurements were available for his analysis. He found that the emissions from vehicles measured within 1 month after their inspection and maintenance had CO emissions2 reductions of 12%. This was less than the estimated IM240 emissions reduction of 14.5% and the MOBILE predictions of 16%. Fur- ther, Wenzel found that the improvement decreased over time. Vehicles measured within 3 to 6 months after I/M had only a 9% reduction; within 12 to 15 months the reduction was only 6%. This study shows a deteriora- tion of vehicles after repair is much greater than the deterioration of normal-emitting vehicles. This finding contradicts a fundamental assump- 2Wenzel's study focused on CO emissions because no NOX results were available from the remote-sensing readings and there was poor correlation between remote sensing and IM240 VOC data for 1991 and later model-year vehicles. All the data reported here are on the emissions improvement in the overall fleet. Improvement data are often reported as the percentage of emissions reduction in failed vehicles only. This percentage is much higher than the overall fleet reduction.

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 99 tion in MOBILE that both normal emitters and repaired vehicles have the same deterioration rate. Wenzel has not used the data on vehicle speed to select out vehicles, which might show high emissions due to normal enrich- ment operation even though the vehicle is operating properly. Jimenez (1999) has shown that there is a distribution of engine power conditions at remote sampling locations, which might cause this erroneous indication of a high-emitting vehicle. It is not clear how this would affect the emissions results percentages obtained by Wenzel. Wenzel also analyzed the effects of vehicle load on the IM240 emissions reduction. He noted that RSDs were placed in locations where the vehicle would be under moderate load to obtain a strong emissions signal. He de- termined the emissions reductions for the moderately loaded portions of the IM240 cycle and found that the measured CO reductions for the mod- erately loaded portions of the cycle were only 77% of the measured reduc- tions for the entire cycle. Thus, one possible discrepancy between the emissions reduction for the remote-sensing and the IM240 results might be due to differences in operating conditions. "Disappearing" Vehicles I/M programs maintain detailed records on vehicle emissions tests. From these records it is possible to determine the number of times a failed vehicle is retested and the change in the reinspection emissions resulting from repair of a failed vehicle. Recent studies have shown that a signifi- cant fraction of failed vehicles never appear for a retest (Wenzel 1999~. For example, an EPA study of the TIM program in Arizona (EPA 1997c) found that 15% of the failures has not been retested. In the draft plans for the MOBILES analysis of I/M programs, EPA staff stated that the default value for the noncompliance rate win be 15%. However, EPA has a stan- dard that requires enhanced I/M programs in certain nonattainment areas to achieve a noncompliance rate of 4% or less. The noncompliance rate includes both vehicles that disappear after an initial I/M failure and vehi- cles that never show up for an I/M inspection. According to EPA (1999n), this is a "generous default" because EPA staff analysis showed actual rates greater than 20%. Users can set higher rates, based on actual data, when they run MOBILE. The discussion of the noncompliance rate (EPA l999n) demonstrates the conflict that sometimes appears in the MOBILE model. Although the analysis of I/M programs shows a noncompliance rate greater than 20%, EPA staff has selected a default rate of only 15%. In addition, the discus- sion notes that this choice "does not constitute a policy by EPA to allow the

~ 00 MODELING M OBl1E-SOURCE EMISSIONS use of this value for SIP purposes." If MOBILE6 does not reflect the best information available, its value and accuracy will continue to be ques- tioned by the user community. Estimating l/M Repair Effects The repair effectiveness part of MOBILES has been criticized (Harring- ton et al. 1998) because it is based on the relatively small number of 266 vehicles. All of these vehicles were repaired in a laboratory setting instead of actual repair shops. In addition, it was not possible to repair all the vehicles to pass the IM240 test; 46% of the vehicles had emissions above the cut points even after repairs. In these cases, EPA extrapolated the repair results observed to determine what the emissions results would be, if additional repairs could be performed. For MOBILE6, EPA has based repair effectiveness on the results of the Arizona I/M program for 1981-1993 vehicles. The ratios were computed by combining data for cars and trucks and sorting the data by age. For each age group, the ratio of mean emissions from failed and repaired vehicles to the mean emissions from passed vehicles was calculated. These ratios were then regressed against age to get the regression equations for the ratio of after-repair emissions to the emissions of normal emitters used in MOBILE6. Because the Arizona data were only for one particular set of cut points, EPA then used data from the repair database that was used for MOBILES to determine the effect of cut points on after-repair emissions. These show significant effects for VOCs and NOX. Compared with the reference cut points of 1.2 g/mi for VOCs, reducing the VOC cut point to 0.4 g/mi is pre- dicted to reduce after-repair emissions by a factor of 0.59. Similarly, re- ducing the NOX cut point from its reference value of 3.0 g/mi to 1.0 g/mi is predicted to reduce after-repair emissions by a factor of 0.489. For the range of CO cut points reported (20 to 15 g/mi) the multiplicative factor is 0.87. Estimating Technician-Training Effects MOBILES has credits that increase the benefits from an I/M program for areas that have programs to train repair technicians. The repair-effec- tiveness data for MOBILE6 are assumed to be those achieved by master technicians. MOBILE6 wiB thus increase the estimated after-repair emis-

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 0 ~ signs for areas that have no training program. These emissions increases will be 78% for VOCs, 174% for CO, and 39% for NOx. These significant increases are based on a single study in which 11 mechanics participated, repairing three cars each. The emissions comparison is based on the dif- ference between the repair results of the students and those of the trainer. No data were available to directly compare the differences between the stu- dents' performance before and after the training, which is the factor to be quantified. EPA acknowledges that the training effect in MOBILES is based on "limited data." Users in areas that do not have a technician- training program can enter the expected increase in emissions due to the absence of a training program. Such an estimate might be based, for ex- ample, on engineering judgment or data from other programs. It is reasonable to assume that I/M programs that train mechanics should be able to obtain a greater emissions reduction from their programs compared with programs without technician training. However, the ac- counting for this training effect in MOBILE is based on minimal data of questionable applicability. Additional data are required to make a direct connection between a particular training program and the related im- provement in emissions reduction from repair under an I/M program. Tampering Effects MOBILES has data on effects that expresses tampering rates as a func- tion of vehicle mileage. Separate tampering rates are available for differ- ent components. These tampering data predict that the incidence of tam- pering will be much less in later- model vehicles. Harrington et al. (1998) examined the incidence of tampering in the Arizona I/M data. They based their evaluation on the amount of tampering in failed vehicles rather than the amount of tampering in the fleet. Consequently, they were unable to make a direct comparison with the tampering data used in MOBILE. For 1995 and 1996 model-year vehicles, the latest year in their study, they found that 6 of the 16 failing vehicles had been tampered with. Based on this very small sample, of failing vehicles only, the authors concluded that the tampering might account for a large fraction of failures in late-model vehicles. Additional data are required to determine the true extent of such tampering. The committee is aware that the California Bureau of Auto- motive Repair is preparing a report on a large data set from vehicles that were randomly pulled over and checked during 1997-1999. This report, which was not publicly released and therefore not available to the commit- tee, should provide useful data on the true extent of tampering in the on- road fleet.

702 MODEI/NG MOBI[E-SOURCE EMISSIONS On-Board Diagnostic Effects MOBILES will consider the effects of electronic diagnostic systems on- board the vehicle. On-board diagnostic (OBD) systems were required on all cars starting in the 1996 model year. Such systems have a malfunction indicator light (MIL) that is supposed to be illuminated when the diagnos- tic system detects a malfunction that would increase the exhaust emis- sions to 1.5 times the applicable standard or more. EPA assumes that a check of the OBD system will be a part of the I/M tests. In modeling OBD systems, EPA assumes that such systems will detect 85% of the high-emitting vehicles (EPA l999o; EPA 1999p). In other words, the MIL wiB illuminate in 85% of the vehicles that are high emit- ters. The response rate of drivers to the illuminated MIL is assumed to depend on the vehicle mileage and the presence of an I/M program. Table 3-5 shows the response rates. It is assumed that drivers will not respond to the illuminated MIL unless there is an I/M program in place or the vehi- cle repair is still under warranty. EPA believes that it might be possible to use an I/M procedure in some future year they suggest 2001 that has no exhaust emissions measure- ment, only an OBD check. However, they recognize that this is an unproven concept, and MOBILES will have calculation procedures that can handle OBD both as a stand-alone system and a system used in con- junction with an exhaust emissions measurement. Figure 3-9 displays the impact of the OBD only and OBD in conjunction with an I/M program on deterioration rates for Tier 1 LDVs and LDTs. The information presented in Table 3-5 illustrates two critical issues with MOBILE. One is the need to predict the life-cycle performance of new emissions control systems and the other is to develop greater knowI- edge of how motorists respond to factors such as illumination of MIL. The assumed response rates to MIL lights are reasonable guesses; however, model results developed with these response rates could be used to justifir the effectiveness of I/M programs. TABLE 3-5 Response Rate to Illuminated Malfunction Indication Light . Response Response Warranty Coverage for Mileage Rate in I/M Rate in non- Mileage Range Range Areas I/M Areas 0-36,000 Full warranty coverage 90% 90% 36,000-80,000 Only catalysts and electronic control 90% 10% module under warranty Over 80,000 No warranty coverage 90% 0% Source: EPA l999o; EPA 1999p.

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 03 1.0 ~ 0.9 ~ 0.7 ~ 0.6 0.4 - 0.3 0.2 - 0.1 No OBD/No l/M — OBD/No l/M - - - OBD/I/M o.o 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 Mileagel1 0000 FIGURE 3-9 Deterioration rates for nonmethane hydrocarbons (NMHCs) for Tier 1 light-duty passenger cars and class 1 light-duty trucks as a func- tion of OBD and I/M. Note that this report usually uses the term VOCs as opposed to NMHCs to refer to the general class of gaseous organic com- pounds. Source: EPA 1999p. Summary on Inspection and Maintenance The modeling of I/M programs has been one of the most controversial elements of the MOBILE models. The changes proposed for MOBILE6 are unlikely to resolve any of the significant issues in these controversies. Many vehicle owners are apparently able to avoid inspection or fail to get their vehicles repaired. Such vehicles must be accounted for properly in MOBILE. Remote-sensing measurements and IM240 data indicate higher deterioration rates for repairs to failed vehicles than those used in MO- BILE. This might be due to mechanical problems with the vehicles or to tampering by owners. It seems intuitively obvious that vehicle inspection and maintenance should result in cleaner, lower-emitting vehicles. This basic intuition has not been supported by unambiguous data on the emis- sions reductions and cost- effectiveness of I/M programs. Early indications are that MOBILE6 will substantially reduce the emissions-reduction bene- f~ts from I/M compared with MOBILES (Clean Air Report 1999~. EPA appears to be heading towards the use of OBD systems as an alter- native to current I/M programs. These systems are untried as an I/M tool and their treatment in MOBILE6 appears to be based on little more than

~ 04 MODE1/NG MOBJ1E-SOURCE EMISSIONS assumptions about system effectiveness and driver behavior. It is not clear that any reasonable model will be able to capture all the effects of I/M—including the assumption that vehicle owners in I/M areas will main- tain their vehicles more frequently- without a significant testing program. The testing program would be necessary to determine the model parame- ters and validate the results of the model. AIR-CONDITIONING EFFECTS Vehicle air conditioners place an additional Toad on the engine, as de- scribed in Equations 3-2 and 3-3. This load increases both the fuel con- sumption and the emissions for a given vehicle speed and road load. This emissions increase is an issue for both vehicle certification and emissions inventory. Emissions from vehicles with operating air conditioners was the fourth issue raised in the GAO (1997) report on MOBILE. Starting with model-year 2000 vehicles, EPA is phasing in a new certifi- cation procedure for vehicles with air conditioners. Previously, the load on the dynamometer was increased by 10% to account for the effect of air-con- ditioning during certification. The air conditioner is not actually operated during this certification test. The new certification procedure has a sepa- rate test that measures emissions from vehicles with the air conditioner operating. The results of this test wiD not be used in MOBILE, however. EPA has added a new set of calculations to determine emissions from air- conditioning operation in MOBILE6.3 There are two elements to these calculations: (1) the determination of emissions for full-Ioad air-condition- ing operation and (2) the estimation of the amount of actual air-condition- ing use. These effects were measured for actual driving cycles, not for the certification cycle to be used in the SFTP. The effect of full air-conditioning operation was determined from a sam- ple of 37 vehicles-23 passenger cars and 14 light-duty trucks-from model years 1990 to 1996 (EPA 1998e). The data were analyzed for both normal and high emitters, but only five of the vehicles were high emitters for at least one pollutant. All vehicles were tested over 15 different driving cy- cles. These included the speed-correction factors and facility driving cycles described for determining the effects of actual in-use operation on emis- sions. Additional cycles used in this study are shown in Table 3-6. 3MOBILE5 does contain an option for air-conditioning calculations. However, the data for these calculations were obtained in the early 1970s and users were advised not to use this calculation option.

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 05 TABLE 3-6 Parameters for Additional (Nonfacility-Specific) Driving Cycles Average Maximum Maximum Cycle Speed Speed Acceleration Distance Cycle Name (mph) (mph) (mphls) (miles) New York City Cycle 7.1 27.7 6.0 1.18 CARE "Unit Led" J`A92) Cycle 24.6 67.2 6.9 9.81 Start Cycle (STOl)a 1.39 20.2 5.1 1.39 aThe ST01 start cycle was an earlier version of the ST03 cycle used for air-conditioning tests in the SFTP. Source: EPA 1998e. The results of these tests will be represented in MOBILES as ratios of emissions with air-conditioning to emissions without air-conditioning. For running emissions these ratios are presented as regression equations that are quadratic functions of mean cycle speed. The input data to these re- gressions are the emissions ratio for each cycle and the mean cycle speed. These regression equations have r2 values between 0.25 and 0.80 depend- ing on the pollutant, vehicle type, and emitter class. For start emissions, a single value is used to give the ratio of emissions with air-conditioning to those without air-conditioning. The effect of air-conditioning ranges from almost no change to more than a doubling of emissions. The effect is most pronounced at lower speeds and is generally higher for passenger cars than for light-duty trucks. The emissions ratios discussed above apply only when the air-con- ditioning compressor is at maximum operation. MOBILES also accounts for the amount of air-conditioning operation. The activity data for air-conditioning operation were obtained on a fleet of 20 vehicles operating in Phoenix, Arizona, from August to October 1994 (EPA 1998f). Although the actual air- conditioning load depends on the torque generated by the compressor, no data were available for this vari- able. Consequently, all data on the extent of air-conditioning load were measured in terms of the *action of time that the air-conditioning com- pressor was actually operating. This was called the compressor-on frac- tion. The compressor-on fraction was modeled as a quadratic function of the heat index. The latter variable is intended to account for the combined effect of temperature and humidity on human comfort. Three separate regression equations were used to account for differences in solar insula- tion during different diurnal periods:

7 06 M ODEI/NG M OBI[E-SOURCE EMISSIONS nighttime, defined as sunset to sunrise; . peak sun, defined as noon to 4 p.m.; and morning or afternoon, defined as the other daytime periods. Because of small samples for high and low values of the heat index, the regression equations produced counterintuitive results in these regions. For example, the nighttime equation produced greater activity than the daytime equation. Because of this, additional regression equations were derived for all daytime data and for the entire data set. The MOBILES air-conditioning activity model selects the regression equations to be used based on the heat index. In general, the equations for different periods are used for the middle range of the heat index. When counterintuitive results occur from these regression equations, the combined equations are used. The compressor-on fraction is assumed to be zero for a heat index of 65°F and below and 1 for a heat index of 110°F and above. The MOBILES model will allow user input on cloud cover. Cloud cover is accounted for by using the nighttime equations during periods of cloud cover. The MOBILES model will also have data on the fraction of vehicles equipped with air conditioners and the fraction of air conditioners that are malfunctioning. These data will be expressed as a function of model year. The air-conditioning corrections will only be applied to the fraction of vehi- cles that have air conditioners that are installed and operating. During the development of the model for compressor-on fraction, consid- eration was given to the effects of vehicle speed, soak time, trip duration, and fraction of idle time during the drive. None of these effects was found to be significant within the limitations of the data available. The consideration of air-conditioning effects provides a good illustration of the uncertainty and resource issues in mobile-source modeling. Al- though the results of the proposed air-conditioning submodel are subject to inaccuracies and inconsistencies because of data limitations, they appear to provide a reasonable estimate of an important effect on emissions. Im- provements in the accuracy of the air-conditioning effect will require addi- tional data. Without some estimate of the uncertainty due to air-condi- tioning effects compared with other uncertainties in the model, it is diff~- cult to decide whether more resources should be spent on improving the air-conditioning results or on improving other parts of the model. Finally, the uncertainties in this submodel are of two kinds. First, there are the statistical uncertainties, which can be estimated quantitatively by the er- ror terms in the regression equations. Second, there are the unknown un- certainties which are due to factors such as the effect of vehicle speed on compressor-on fraction, the validity of the heat index as a measure of air- conditioner use, individual driver behavior in air-conditioner use, or the effect of the actual compressor torque.

TECHNICAL ISSUES ASSOCIATED WITH THE M OBILE MODEL 7 07 EVAPORATIVE EMISSIONS Evaporative emissions have several categories. Four main physical mechanisms are used to account for evaporative emissions: diurnal emis- sions, resting losses, hot-soak emissions, and running losses. The combi- nation of diurnal emissions and resting losses are measured together as "real time diurnal" [RTD] emissions. The measurement and characteriza- tion of evaporative VOC emissions is based almost exclusively on certifica- tion test procedures with no real-world measurements. New vehicle evap- orative emissions certification to the 2 gram per test standard has been based on the sealed-housing for evaporative-determination (SHED) testing of a hot-soak and a 1-fur diurnal emissions test. The new evaporative test procedure introduced in 1996 includes real-time, multiday diurnal, and running-loss emissions measurement. Results of these tests provide a bet- ter picture of evaporative emissions, but the test procedure is lengthy and complex, and places a practical limit on the number of vehicles tested. The database of measurements for the in-use fleet is improving with MOBILES using data from approximately 300 vehicles. Hot-soak and running-loss evaporative emissions are highly skewed with substantial contributions from high emitters that apparently relate to liquid leaks. Recent studies (Gorse 1999) indicate that evaporative VOC emissions from the fleet studied exceed tailpipe emissions by a factor of more than 2. This is at variance with the results shown in Figure 3.2. This figure shows an evaporative to tailpipe ratio of about 0.7 for Chicago and New York, using the MOBILE5b approach to calculating evaporative emissions. Characterization of Multiday Diurnals Diurnal emissions come from natural cycling of the ambient tempera- ture and the resulting pressure-driven emissions of fuel vapors. The mag- nitude of these emissions depends on the ambient temperature variation, fuel vapor pressure, and the period of vehicle nonoperation. Resting-losses are gauged by the permeation of fuel through tanks, lines, and fittings, and liquid leaks that are not the result of temperature variation. These occur while the vehicle is not operated and are also captured in the multi- day diurnal tests. Figure 3-10 shows the temporal variation of these two types of evaporative emissions for a typical vehicle without large leaks. Real-time diurnal emissions tests of 270 vehicles provide the data for esti- mating diurnal and resting-loss emissions. The emissions *om the fleet tested are highly skewed, with liquid leak- ers dominating the high end of diurnal and resting-loss emissions. In MOBILES, the fleet is divided into normal and high emitters. The diffi-

~ 08 MODE[lNG MOBILE-SOURCE EM!SS/ONS - ~ 1.0 In o ._ on In ~ 0.5 I ~ 0.0 : : : : : : : : : : Pres. sure-Drive'' Vapor Leaks. . _ : . Resting . : : , Losses . ..~ 12 18 24 30 Time (hours) FIGURE 3-10 Illustration of real-time diurnal measurement for a typical vehicle without large leaks showing the combination of pressure-driven losses due to diurnal temperature variation (dark area) and resting losses (light area). Source: EPA l999q. culty of correctly determining the distribution of emissions in the vehicle fleet and the level of emissions from the high emitters remains a difficult problem. Characterization of Hot-Soaks Hot-soak emissions are evaporative emissions occurring, by definition, during the first hour following engine shutdown. Most of the hot-soak emissions occur during the first 10 min (EPA 1998g). Hot-soak emissions result from heating the fuel above ambient temperatures after vehicle operation, that is, during the hot-soak. They come primarily from the fuel tank and, in carburetor vehicles, from the carburetor bowl. They depend upon vehicle technology, ambient temperature, fraction of fuel in the tank, and fuel vapor pressure. With the replacement of carburetors by fuel injec- tors, the contribution of hot-soak emissions to evaporative emissions has fallen. Hot-soak emissions are not as skewed as diurnal and resting-loss emlsslons. Characterization of Running-Losses Evaporative emissions occurring during vehicle operation are termed running-losses. These emissions are measured while the vehicle is being

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 09 operated on a chassis dynamometer. Emissions depend upon driving cycle, fuel vapor pressure, and ambient temperature. The importance of these emissions was not well recognized at the time of the development of MOBILES. Measurement of running-loss emissions by the CRC from pre- 1992 vehicles (McClement et al. 1998) and from 1992 and later vehicles (McClement 1999) show emissions in reasonable agreement with those predicted by MOBILES for the newer vehicles but much greater than pre- dicted for the older vehicles. Running-loss emissions data developed for MOBILES indicate that MOBILES underestimates emissions by a factor of 2 to 4 (EPA l999r). Because of the skewness of the emissions and impor- tance of the gross liquid leakers, an accurate determination of the true in- use fleet running-Ioss emissions is difficult. FUEL EFFECTS Reformulated Gasoline Effects Use of RFGs can be specified in MOBILES input. The effects of RFG on vehicle emissions are determined mainly by the RVP and oxygenated-fuel parameters, discussed below. The effects of RFG depend upon the season. If RFG effects are modeled for summer conditions, any user-supplied RVP and oxygenated-fuel inputs are overridden by the provisions of the RFG rules. These rules require that for Phase I RFG (1995 through 1999), the maximum RVP be 7.1 and 8.0 pounds per square inch (psi) for fuel-volatil- ity Regions 1 (southern) and 2 (northern), respectively. Phase 2 RFG, which begins in year 2000, requires that aD regions have an RVP no high- er than 6.7 psi. In the winter, fuel RVP is not regulated. What is much more important in winter is the fuel oxygen content, and default values might be overrid- den by user-selected values as long as they are above 2.1 weight percent Wit%. The effect of the fuel sulfur level is fixed in MOBILES, because RFG is assumed to have a specific sulfur content. In MOBILES, there will be some changes in the way that the benefits of RFG are applied. Although MOBILES calculates the effect of RVP and oxygen before adjusting for RFG, MOBILES (as proposed) would adjust for RVP, oxygen, and sulfur before adjusting for RFG. Oxygenated-Fuel Effects Oxygenates are oxygen-containing organic compounds, alcohols, or ethers, that are added to gasoline to achieve enleanment in the combus-

7 7 O MODELING MOBILE-SOURCE EMISSIONS tion process. Enleanment typically results in a reduction of CO and VOC emissions and an increase in NOX emissions from the engine. The use of oxygenates (oxyfuels) was mandated by EPA under the CAAA90 in certain regions during winter months in response to violations of the CO air-qual- ity standard. Oxygenates are also a component of the RFG program, which has the objective of reducing ozone-precursor VOC emissions. The poten- tial increase in NOX emissions and a reduction in fuel economy are the main disadvantages of adding oxygenates to gasoline. In 1997 the Office of Science and Technology Policy (OSTP) published "Interagency Assessment of Oxygenated Fuels," which reviewed the state of the winter oxyfuel program (NSTC 1997~. In the assessment of the air- quality effects of the program, several issues were identified that are rele- vant to the crediting of such a program in MOBILE5a: The observable reduction in ambient CO levels that could be attrib- uted to the use of fuel oxygenates was lower by a factor of 2 or 3 than the amount predicted by the MOBILE5a model. The MOBILE5a model predicted CO emissions reductions that were about a factor of 3 larger than other EPA models, notably, a version of the Complex Model developed to represent fuel effects for the on road fleet. The emissions database was inadequate to accurately predict the ef- fects of fuel oxygenates on CO emissions at temperatures below about 50°F. The available data indicated that the emissions reduction was de- creased at low temperatures compared with the effect at 75°F. . Because of improvements in emissions-control technology new vehi- cles experienced relatively little CO emissions reduction from fuel oxygen- ates. According to EPA documentation for MOBILES (EPA 1998h), the model will have reduced oxygenate benefits "matching MOBILES predictions with ambient CO data." Unfortunately, adequate data do not exist from either ambient air analyses or from vehicle-emissions studies to develop an accurate prediction. For example, there are no accurate measurements of the ambient air effects of oxyfuels for any region in the eastern or mid- western states upon which to base a prediction. The vehicle-emissions data for oxygenate effects, upon which the MOBILES analysis is based, are about 10 years out of date, do not represent the current Beet, and are inadequate for temperatures below about 50°F. EPA documentation does not give a representative prediction for MOBILES to compare with the MOBILE5b or earlier model versions, but it does appear to reduce the oxyfuel effects on CO emissions. The OSTP assessment found that one of the reasons for MOBILE's overprediction of

TECHNICAL ISSUES ASSOCIATED WITH THE M OBILE MODEL 7 7 7 the oxyfuel effect was that the model incorporated large reductions for high CO emitting vehicles and used a vehicle distribution with a popula- tion of high emitters that was too large. It is not clear that the latter prob- lem has been corrected in MOBILES. The results from a recent emissions study conducted by the Colorado Department of Public Health and Environment should be noted. The study was conducted at 35°F on a fleet of 31 vehicles selected as represen- tative of the Colorado on-road fleet and used a 3.5 wt% ethanol fuel. The average CO emissions reduction was about 11%, which is about one-third of the MOBILE5a predicted benefit (Ragazzi and Nelson 1999~. Sulfur Effects Since the early 1970s, it has been known that sulfur in gasoline affects the conversion efficiency of automobile three-way catalysts. Studies of Tier 0 vehicles indicated that adverse effects of sulfur on catalysts could be sub- stantially reversed if they were subsequently refueled with low-sulfur gas- oline (AQIRP 1992~. New vehicles are certified on indolene fuel, which typically contains very low sulfur levels in the range of 30-50ppm. In con- trast the real-world sulfur levels in the United States average about 330 ppm with peaks in the range of 1,000 ppm. Figure 3-11 presents a distri- bution of sulfur levels in gasoline for select cities in the United States. Generally, premium grade gasoline is on the low end of the sulfur range because of the refining process used to produce this type of fuel. The emissions impact of sulfur in gasoline is reflected in EPA's Complex Model. This model, which applies strictly to 1990 technology vehicles, was used by EPA to adjust emissions factors in MOBILES to the national aver- age sulfur content of 330 PPM. Appropriate adjustment was also made for Phase I RFG fuel, which assumed an average sulfur content of 220 ppm. Use of MOBILES in areas that significantly deviated from the national average sulfur content would have some inaccuracy introduced by not ac- counting for the sulfur effect. The basic effects of sulfur in gasoline on Tier 0 vehicles reported by EPA are summarized in Table 3-7. A considerable amount of new data on the impact of sulfur in gasoline has been generated in the last few years, driven to some extent by sub- stantial concerns about larger impacts on new LEV and beyond vehicle technologies. EPA has analyzed this information and has developed a much more accurate estimate of fuel sulfur effects. EPA (1999s) proposes to incorporate this information in a sulfur-emissions correction in MOBILES, which will allow users to input the area-specific sulfur content

72 MODELING MOBIIE-SOURCE EMISSIONS 1200 1000- 800- ~ - 6001 400- 200- o- 111 . _. —l l i 1 17 1 1 1 1 1 1 1 1 1 1=1 1 1 1 1 1 ~ ¢ ~ ¢ ¢ ~ ~ ~ ~ ~ ~ ~ O ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ At ~ ~ ~ ~ E E. ~ ~ ° ¢ ~ ~ E in. a ~ ~ .~ FIGURE 3-11 1996 annual average and extreme sulfur levels in gasoline for selected U.S. cities. Source: Darlingtonet al. 1999. TTt I I I L I I 1 1 1~1l -I-~-t I TtTt I I rT of gasoline. This will greatly improve the simulation of gasoline sulfur factor impacts. Impacts will be segregated into start and running emis- sions. Figures 3-12 through 3-15 provide an indication of the composite correction factors for a range of vehicle technologies. TABLE 3-7 Estimated Emissions Reductions Due to Reductions in Fuel Sulfur for a Normal-Emitting Tier O Vehicle Reduction in Emissions, % (400 ppm to 50 ppm Sulfur) NOX VOC CO 6.3% 18.8% 21.7% Source: EPA l999s.

TECHNICAL ISSUES ASSOCIATED WrTH THE MOBILE MODEL 7 7 3 E Q 30 to v, 20 . o ._ to u) E till ~ 10 . _ a) u' ~ 0 - o ° 0 60 ~q W- ~ 0 0 ~ ~ 120 180 240 300 360 420 480 540 600 Sulfur (ppm) +HC ~ NMHC DECO ~ NOX FIGURE 3-12 The percent increase in exhaust emissions for Tier 0 normal-emitting vehicles as a function of fuel sulfur content. The base value for sulfur is 30 ppm. Note that this report usually uses the term VOCs as opposed to HCs or NMHCs to refer to the general class of gaseous organic compounds. Source: EPA l999r. The claim that the effects of fuel are reversible is the subject of some controversy and is not reflected in the proposed EPA MOBILES sulfur- emissions correction factor. Undoubtedly, additional testing for this effect will be done in the future, particularly on more advanced technology vehi- cles and vehicles meeting the SFTP. Compliance with the SFTP, begin- ning with model year 2000, is expected to result in much more precise con- trol of air/fuel ratios that will minimize the opportunities for deposited sulfur in catalysts to be burned off. EPA has estimated that about 50% of the effects of sulfur on catalysts for Tier 2 vehicles can be reversed by re- verting to Tow-sulfur fuel (EPA l999t). The effects of sulfur on catalysts for Tier 0 and Tier 1 vehicles are fully reversible. RVP Effects The Reid vapor pressure is defined as the fuel vapor pressure in pounds per square inch at 100°F. RVP affects vehicle emission primarily in two ways: the fuel evaporation rate and the exhaust emissions. The connec-

1 14 Q 40 lo In ,o us .m LL ._ ~ 10 - MODELING MOB`LE-SOURCE EMISSIONS _ ~- aK ~ ~ 1 1 1 0 60 120 180 240 300 360 420 480 540 600 Sulfur (ppm) HO ~ NMHC DECO ~ NOX FIGURE 3-13 The percent increase in exhaust emissions for Tier 1 normal-emitting vehicles as a function of sulfur content. The base value for sulfur is 30 ppm. Source: EPA l999r. tion between RVP and evaporative emissions is obvious, as one expects a greater evaporative emissions rate from a fuel with a higher vapor pres- sure. The link between RVP and exhaust emissions is less obvious but well documented. The effects of RVP on exhaust emissions are not signifi- cant for NOX but often are significant for CO and VOCs. Representation of emissions from Tower-polluting fuels, especially fuels with lower volatility, was the seventh issue noted in the GAO (1997) report. A CRC-AQIRP study (Reuter et al. 1992) found that decreasing the RVP of 9 psi fuels by 1 psi in FTP tests reduced the exhaust CO and VOC emis- sions by about 4.5% and 9.1%, respectively. An API study (Lax 1994) found that the effect of RVP on exhaust emissions for nominal 10 and 13 psi fuels varied with temperature. The effect was greatest at the highest temperature of the study, 80°F, was reduced at 55°F, and was insignifi- cant at 35°F. A plausible explanation of the RVP effect on exhaust emissions has been described by Hyde (1998~. He suggests that the enhanced CO and VOC emissions occur when a vehicle's evaporative-control system delivers a quantity of fuel to the engine that is so large that enrichment occurs. The evaporative-control system includes a canister packed with activated charcoal to capture excess fuel vapor from the vehicle's gas tank. When the canister becomes loaded, the vapor is released into the engine intake causing the enrichment. The enrichment condition means that the engine

TECHNICAL ISSUES ASSOCIATED WITH THE M OBILE M ODEL 7 7 5 Q Q 200 to co o ~= 150 to o ._ to A 1 00 UJ c ._ a) <`s 50 o me_ em ~_c~ '~ = .,_E3B— R~ , , 0 60 120 180 240 300 360 420 480 540 600 Sulfur (ppm) .. +HC ~ NMHC DECO KNOX FIGURE 3-14 The percent increase in exhaust emissions for LEV and ultra-Iow-emitting vehicles (ULEV) as a function of sulfur content. The base value for sulfur is 30 ppm. Source: EPA l999r. is supplied with fuel in excess of the amount that can be completely burned by the available oxygen. The fuel-rich condition and incomplete combustion results in an increase in CO and VOC emissions. High temper- atures and high RVP fuels will cause an increase in fuel evaporation, and thus the rate of loading and the frequency of purging fuel from the canis- ter. MOBILE6 exhaust RVP correction factors for RVP are linked to exhaust temperature correction factors (described in the next section). In MOBILES, RVP correction factors were restricted in that there was no emissions reduction for fuels less than 9 psi (but there was an increase for RVP greater than 9 psi). This was done primarily because there was little test data available at the time for low RVP fuels. At the time of MOBILES development, fuels with RVP less than 9 psi were not widely used. The GAO report expressed a concern about this limitation of the model. For MOBILE6, EPA has not updated the exhaust RVP correction fac- tors, even though there is a wealth of new test data available at low RVP levels, and most areas of the country are using fuels with RVP less than 9 psi in the summer. Although the RVP effects on exhaust emissions are relatively small, credit should still be supplied to states and local areas that are using low-RVP fuels. The exhaust RVP correction factors should be updated in the next version of MOBILE6.

7 7 6 M ODEL!NG M OBllE-SOURCE EMISSIONS Q 60 Q lo ~ 40 o . _ In In .E .= 20 a) a) it=_ / _ ~ ~ 1 1 1 1 1 1 1 1 1 ~~ O _ 0 60 120 180 240 300 360 420 480 540 600 Sulfur (ppm) + HC ~ NMHC DECO KNOX FIGURE 3-15 The percent increase in exhaust emissions for REV and ULEV trucks as a function of sulfur content. The base value for sulfur is 30 ppm. Source: EPA l999r. Hot-soak emissions are estimated as functions of RVP. These functions, derived from test data compiled by EPA, predict "full" soaks in grams per trip. The number of trips per day and fraction of days without trips are factored in to arrive at annual average rates. Diurnal emissions are handled using the Uncontrolled Diurnal Index (UDI). First, the Wade equation is used to calculate the total diurnal emis- sions for 9 psi, temperature rise from 60°F to 84°F, and 40% tank level. The UDI is the ratio of the Wade-based emissions of the model's parame- ter levels to the base conditions. The model then estimates the full, par- tial, and multiple diurnals as functions of the UDI and finally combines them based upon the relative occurrence of each phenomenon. Running-Ioss emissions are calculated using a simple algorithm. Vehi- cles are stratified by whether they pass a pressure-purge test. For each category, there exist running-loss emissions rates at four temperatures and four RVPs. The model interpolates these to obtain the correct passing and failing running-loss emissions rates and combines these by weighting them according to their expected occurrence in the fleet. EXHAUST EMISSIONS TEMPERATURE-CORRECTION FACTORS FTP tests are conducted at a nominal temperature of 75°F using a specified test fuel. Temperature-correction factors (TCFs) are used to ad-

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 7 7 just the measured FTP emissions to determine the emissions that are ob- tained at other operating temperatures. Temperature and RVP corrections are closely related. In MOBILES, TCFs for temperatures above 75°F are combined with RVP correction fac- tors. Below that temperature, the correction factors are simple multiplica- tive factors. TCFs are dependent on fuel-delivery technology and are calcu- lated separately for each FTP bag (the FTP and its separate components are discussed earlier in Chapter 3~. Technology mixes are then used to combine technology-specific TCFs to produce a TCF for a specific model- year and vehicle-class combination. For Bag 1 CO emissions at low tem- perature, the correction is in the form of an additive offset with the offset a function of temperature and fuel-delivery type. EPA (1999u) describes the methods proposed for MOBILES for predict- ing exhaust emissions for vehicle starts and running at non-FTP tempera- tures. The FTP test is conducted using a dynamometer at 75°F and em- ploys a driving cycle that is intended to be representative of vehicle use in an urban area. Many emissions tests using the FTP driving cycle have been conducted at temperatures above and below the FTP standard of 75°F. These "non-FTP temperature" tests show that in general exhaust emissions of CO and VOCs increase gradually, typically 10% to 30%, with decreasing temperatures from about 80°F down to about 50°F. Below 50°F the emissions increase dramatically in a nonlinear fashion. For ex- ample, an API study (Lax 1994) found an increase in VOC emissions of roughly 60% from 55°F to 35°F and an increase in CO emissions of roughly a factor of 2 from 55°F to 35°F. Because much of the driving in urban areas occurs at non-FTP tempera- tures, the methods used in MOBILE to predict the non-FTP emissions are critically important to the accuracy of the predictions. The method used in MOBILES to predict emissions at non-FTP temperatures involves a linear interpolation between the hot-start (10-min soak) and cold-start (12-hr soak) emissions test points of the FTP driving cycle. The soak time refers to the length of time the hot stabilized test vehicle is turned off and sits at the test temperature before it is restarted and the emission measurement begins. The 12-hour period is the standard for which it has been shown that the test vehicle will come to equilibrium with the ambient tempera- ture of the test chamber. There are two assumptions that appear to be im- plicit to EPA analysis: (1) the test vehicle changes temperature at a uni- form linear rate over a 12-hr period and (2) the emissions levels change uniformly with temperature. Neither of these assumptions is valid. The rate of cooling of a hot engine by convection and conduction is affected by the difference in temperature between the engine and its surroundings. The cooling rate is much faster at first when the engine is hot and becomes slower as the engine cools and its temperature approaches the ambient

7 78 MODELING MOBI[E-SOURCE EMISSIONS temperature. The engine's start emissions, which comprise a large fraction of the total driving cycle emissions, are very sensitive to the engine's tem- perature, so the assumption that the engine cools at a uniform rate over 12 hours is erroneous and arbitrary. As noted above, the effect of temper- ature on emissions is large and nonlinear below about 50°F. One expects errors in the MOBILE predicted emissions at non-FTP temperatures. The errors from using the linear interpolation method are expected to be larg- est at lower temperatures, typically below about 50°F. The MOBILES approach to start emissions at FTP temperatures, dis- cussed in the previous section on start emissions, does account for the non- linear temperature effects. In that approach, the soak temperature is con- stant at 75°F. The soak function, which depends only on time, accounts for the cooling of the engine and the catalyst as a function of time at the con- stant ambient temperature of 75°F. A proper approach to temperature corrections for start emissions would use an expanded soak function, which would depend on both ambient temperature and soak time. Such an expanded soak function would require additional data collection efforts. HEAVY-DUTY VEHICLE EMISSIONS Heavy-duty vehicles are those exceeding 8,500 lbs gross vehicle weight (GVW). Emissions from these vehicles have come under increased scru- tiny recently, particularly those from heavy-duty diesel vehicles (HDDVs). This is due to EPA's past emphasis on control of emissions from LDVs that has increased the relative significance of HDV emissions; public concern over the human health and environmental impacts from PM and NOX emissions, both of which are emitted in relatively large amounts from HDDVs; advances in emissions-control technologies that have increased the cost-effectiveness of regulating heavy-duty engines; and a recent en- forcement action over manufacturers use of"defeat devices" that allows engines to meet emissions standards during certification testing but al- lows excess emissions during highway driving (EPA l999v; EPA l999w). The GAO Report (1997) states the following as one of the major limitations (Issue 11) in the MOBILE Model: EPA's supporting data on the in-use emissions of this category of vehi- cles are about 20 years old; certification standards are higher for these vehicles than their light duty counterparts, and they are generally older and driven more miles annually than their light duty counterparts, al- though improved emissions technology has lessened the contributions of individual vehicles; some studies are under way, but agency officials

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 ~ 9 question whether sufficient ~ & M data will be obtained in time to change emissions estimates for heavy duty vehicles. The large majority of emissions testing for HDVs has been on an engine dynamometer. There are limited in-use data from I/M programs. Because of these factors, EPA plans to use test data required by EPA from engine manufacturers for new-engine certification as a surrogate for in-use emis- sions data (EPA l999v). Engine certification data involve zero-mile level (ZML) emissions plus rates of deterioration to the end of the useful life. Intended service classes and useful lives are shown in Table 3-8. A caveat to the use of these data in developing deterioration rates involves manu- facturers observing negative deterioration results. Although testing might indicate engine emissions rates that are lower than the ZML, a manufac- turer is not permitted to report a negative deterioration. In those cases, zero deterioration is reported. Therefore, the average deterioration rates calculated from the certification data are higher than the deterioration the manufacturers determined in their laboratory tests. It should be noted that, because all the engines tested for certification receive the proper maintenance and meet the manufacturers' specifications, the effects of inadequate maintenance of the engine and tampering on emissions are not included in the analysis (EPA l999v). Another problem using certification data concerns the undercounting of "excess" NOX emissions. These excess emissions were due to defeat devices used by manufacturers to allow engines to meet emissions levels during testing, but switches off emissions controls for improved fuel economy dur- ing highway operations. These defeat devices involve engines since 1988 having software that advances the injection timing under high-speed oper- ating conditions and increases the NOX emissions above the FTP transient cycle certification levels. Recently, EPA has entered into a consent decree TABLE 3-8 Intended Service Classes and Useful Lives for Heavy-Duty Engines 8,501- 19,500 19,501 - 33,000 33,001 - 60,000+ Engine Class All heavy-duty gasoline engines Light heavy-duty diesel engines Medium heavy-duty diesel engines Heavy heavy-duty diesel engines (incl. buses) Gross Vehicle Weight (GVW) (lb) ,501 - 60,000+ Useful Life (miles) 110,000 110,000 185,000 290,000 Source: EPA l999v.

120 MODEI/NG MOBI[E-SOURCE EMISSIONS with the diesel engine manufacturers requiring manufacturers to offer new software when 1988-1998 engines are rebuilt. These effects will be incorporated into MOBILES (EPA l999w). Further field data regarding the effect of the rebuild kits will need to be gathered for use in improving the modeling of NOx emissions in a revision of MOBILES. Although little is known to date about HDV high emitters, there is now starting to be some state effort to smoke test vehicles on the road. This will not likely give data about the regulated emissions of HO, CO, NOX and PM but will give some data about the state of maintenance of vehicles on the road. There is a need to determine data on high emitters, and the state programs could provide a source of vehicles that have high smoke emis- sions, which could then be tested for other emissions. Because of the lack of I/M data, MOBILES will not have accurate emis- sions rates that represent in-use emissions from HDVs. This is a serious shortcoming of the MOBILES model since the technology and the emis- sions levels have changed during the past 25 years. However, extensive data are now being gathered in a number of major research programs. These data should be analyzed and interpreted relative to determining better deterioration factors and improved in-use emissions factors from HDVs. Use of Engine Data ant] Conversion Factors To Estimate Vehicle Emissions For MOBILES, EPA (1998i) has updated the estimates of heavy-duty engine emissions factors currently contained in MOBILE5b. The method- ology involves the estimation of a gram per mile emissions factor by multi- plying a work-specific emissions level (in g/bhp-hr) by a conversion factor that converts work units into mileage units (bhp-hr/mi). In mathematical form, the conversion factors can be expressed as Conversion Factor (bhp - hr/mi) = Fuel Density (lb/gal) BSFAC (lb/bhp - hr) x Fuel Economy (ml/gal) (3-6) where BSFC is the brake-specific fuel consumption. EPA divides heavy-duty vehicles into several classes in MOBILE. Table 3-9 shows that the model classifies heavy-duty vehicles into those using

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 2 7 TABLE 3-9 Heavy-Duty Vehicle Classifications for MOBILE6 Designation Gasoline Vehicles HDGV (classes 2B-3) Heavy-duty gasoline vehicles HDGV (classes 4-8) Heavy-duty gasoline vehicles Diesel Vehicles HDDV (class 2B) HDDV (class 3) HDDV (classes 4-5) HDDV (classes 6-7) HDDV (class 8A) HDDV (class 8B) Urban Buses HDGB (school) HDGB (transit) HDDB (school) HDDB (transit) Description Gross Vehicle Weight (lb) 8,501-14,000 >14,000 Light he avy- duty die set trucks 8,501 - 10,000 Light he avy- duty die set trucks 10,001 - 14,000 Light heavy-duty diesel trucks 14,001-19,500 Medium heavy-duty diesel trucks 19,501-33,000 Heavy heavy-duty diesel trucks 33,001-60,000 Heavy heavy-duty diesel trucks >60,000 Heavy-duty gasoline school buses All Heavy-duty gasoline transit buses All Heavy-duty diesel school buses All Heavy-duty diesel transit buses All Source: EPA 1999x. gasoline or diesel fuels. These are further subdivided based on GVW. EPA uses the classifications in Table 3-9 to account for different characteristics and general uses of the engines included in each GVVV class. EPA decided to recompute the emissions levels and deterioration rates based on model-year groups that represent changes in EPA's emissions standards. To improve the flexibility of MOBILE6's emissions factors, EPA has chosen to use emissions rates for each service class as shown in Table 3-9 instead of a single rate as used in MOBILES. For heavy-duty gasoline engines, EPA will continue to have a total emissions rate in MOBILE6. These changes in MOBILE6 represent improvements over MOBILES. However, all of the different classes of vehicles shown in Table 3-9 likely have different driving cycles. At present this is not accounted for com- pletely in MOBILES (and for that matter, in MOBILE6), because the in- use emissions in grams per mile is based on FTP transient dynamometer emissions data multiplied by a conversion factor that accounts mainly for fuel economy differences and VMT as a function of vehicle age. There is a need to develop more modal emissions data and driving-cycle data by vehi- cle class so improved emissions in grams per mile can be made available in a revision of MOBILE6.

722 MODELING MOBILE-SOURCE EMISSIONS Adjustments to Heavy-Duty Vehicles Basic Emissions Rates The MOBILES model will calculate emissions factors at low and high altitudes. Low-altitude emissions factors are based on conditions repre- sentative of approximately 500 feet above mean sea level and high-altitude emissions factors represent conditions of approximately 5,500 feet above sea level. At the present time, EPA has been unable to obtain recent stud- ies as to the effects of varying altitude on exhaust emissions from heavy- duty gasoline vehicles. Therefore, MOBILES will use the same adjustment factors used in MOBILES. EPA was able to locate a small number of heavy-duty diesel vehicle studies evaluating the effects of altitude on changes in emissions. From these data, EPA developed new altitude-ad- justment factors for heavy-duty diesel vehicles in MOBILES (EPA l999v). At present, MOBILES does not account for the effects of temperature or humidity on HDDV emissions. Temperature and humidity vary in differ- ent parts of the United States and they will very likely affect NOX and also PM emissions. The Engine Manufacturers Association (E MA) is presently conducting tests at Southwest Research Institute (SwRI) to answer this question EPA should review these data when they become available in late 1999 and incorporate correlations into a revision of MOBILES. Recent tunnel data may show that HDVs emit more NOx then predicted by MOBILES (~.C. Sagebiel, Desert Research Institute, personal commun., May 19991. One of the explanations for this difference could be tempera- ture and humidity effects, along with the fact that the FTP NOx data might not be representative of in-use NOx. Finally, diesel fuel properties have been shown to have an effect on emissions. The properties of cetane number, aromatics, and sulfur content have been most often cited as the important variables. Sulfur content is presently in PARTS relative to SO2 and sulfate emissions. EPA should con- tinue to evaluate the available data to see if reasonable approximate corre- lations of these variables can be developed for MOBILE to relate fuel prop- erties to the NOx and PM emissions from HDDV. PARTICULATE EMISSIONS As noted in the introduction to this chapter, PM emissions from on-road mobile-sources are estimated using the PARTS model. PARTS is a sepa- rate Fortran program, but is compatible with MOBILE5a in format and fleet characterization. PARTS calculates exhaust PM as the sum of lead (for gasoline-powered vehicles), soluble organic fraction (SOF), remaining carbon portion (RCP), and directly emitted sulfate (Sit. The emissions *om tire and brake wear are also calculated in the model.

TECHNICAL ISSUES ASSOCIATED WrrH THE MOBILE MODEL 7 23 PARTS calculates PM factors in grams per mile for 12 vehicle classes: light-duty gas vehicles, light-duty gas truck 1, light-duty gas truck 2, heavy-duty gas vehicles, motorcycle, light-duty diesel vehicle, 2B heavy- duty diesel vehicles, light heavy-duty diesel vehicles, medium heavy-duty diesel vehicles, heavy heavy-duty diesel vehicles, and buses. It reports emissions as a function of particle sizes less than or equal to 1.0 to 10.0 Am. The fraction in each size range is hard-coded in the model. For exam- ple, PARTS assumes 97% of exhaust particulate is 10 mm or less for un- leaded gasoline-fueled vehicles. Although the documentation covering PARTS (EPA 1995) does not ex- plicitly address the source of the underlying data, communication with EPA staff confirm that most of the estimates for exhaust emission of HDVs and LDVs are based on either certification data and analyses dating back to early 198O, or on ratios to hydrocarbon emission factors. The woe- fully outdated emission factor estimates in PARTS make the accuracy of the model highly suspect; the emission factors must be updated using re- sults of recent test programs for both light-duty and heavy-duty vehicles (discussed below). Exhaust Particulate Emissions Exhaust PM have typically been a concern for diesel-powered vehicles, which are predominately used in heavy-duty vehicles. Concurrent with EPA's consideration of an ambient air-quality standard for particle sizes of 2.5 mm and less, there has been increased concern that gasoline-powered engines, the main propulsion source for the light-duty fleet, could make a significant contribution to ambient PM. Recent studies of PM emissions from LDVs, such as the CRC Projects E-24-1 (Cadre et al. 1998), E-24-2 (Norbeck et al. 1998), and E-46 (Cadre et al.1999), as well as the Northern Front Range Air-quality Study (Norton et al. 1998), show PARTS emis- sions rates to be Tower than those observed and that a significant fraction of emissions are coming from LDVs. Tables 3-10 and 3-11 show this underestimation of PARTS emissions rates for a range of LDVs. Table 3-10 contrasts the results of a test per- formed for CRC Project E-24-2 with estimates produced by PARTS for light-duty passenger cars and trucks. This underestimation of exhaust PM emissions in PARTS is also evident in the comparison of PARTS rates with the weighted FTP results from the CRC E-24-1 and E-46 projects. Table 3-11 displays those results. Although more attention has been given to particulate measurements from heavy-duty diesel engines, PARTS did not have sufficient data to de- termine a deterioration rate for PM emissions for these vehicles. A recent

Pre-1981 .315 .193 .368 .193 1981- 1985 .486 .025 .493 .025 1986- 1990 .172 .006 .122 .006 1991- 1997 .018 .004 .035 .004 724 MODELING MOB/LE-SOURCE EMISSIONS TABLE 3-10 PM Emissions Rates in g/mi from PARTS and from the CRC Project E-24-2 Passenger Cars Light-Duty Trucks Model Year CRC PARTS CRC PARTS _ . Source: Cadle et al. 1998. report prepared for EPA (Weaver et al. 1998) shows significant deteriora- tion for 1994 and later trucks and transit buses. In addition, work by Graboski et al. (1998) concluded that PARTS was "significantly underesti- mating" PM emissions from heavy-duty vehicles. It is important that EPA revise PARTS to reflect in-use PM emissions. This will likely require extensive field measurements. Data are also need- ed to assess differences between HDDV PM measurements obtained in the laboratory compared to in-use emissions; EPA should review work cur- rently being conducted by the CRC and the National Cooperative Highway Research Program in this area. In addition, data are needed relative to systems of the engine and the vehicle that reflect maintenance problems that affect emissions. Studies are also needed on the effectiveness of die- sel I/M programs and whether smoke I/M programs tend to increase NOX TABLE 3-11 Passenger Car PM Emissions Rates in g/mi from PARTS and from the CRC Projects E-24-1 and CRC E-46 Model Year CRC E-24-1 PARTS Pre-1981 .955 .193 1981- 1985 .474 .025 1986-1990 .444 .006 1991-1997 .028 .004 CRC E-46 PARTS Tier 0 .083 .025 Tier 1 .038 .006 l Source: Norbeck et al. 1998; Cadle et al. 1999.

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL ~ 25 emissions and reduce PM emissions. Ways of screening and characterizing the in-use vehicle population for high-emitting vehicles are needed. EPA needs to determine the effectiveness of the catalytic oxidation devices used since 1994 to reduce PM emissions by reducing the SOF. There is no infor- mation on the long-term effectiveness, maintenance practices, or tamper- ing for these devices. Particulate Emissions from Tire and Brake-Wear PARTS estimates emissions from tire wear based on the assumption that the emissions rate of airborne particulate is 0.002 grams per mile per wheel (EPA 1998j). This reference, known as AP-42, lists two studies as the basis of the estimate for tire-wear emissions Williams and Cadle 1978; Brachaczek and Pierson 1974~. The single emissions rate is based on tests of LDVs and no estimate for the airborne particle size distribution for tire-wear is offered. Tire-wear emissions less than 10 ,um are based on interpolation. The dated references for the PARTS emissions factors sug- gests that these factors are based on tests of older biased-ply tires rather than longer-wearing tire technologies currently in use. PARTS reports brake wear as a separate emissions factor of 0.0128 grams per mile, based on a paper by Cha et al. (1983~. Brake-wear partic- ulate emissions are higher than for tire wear because a larger fraction is assumed to be less than 10 ,um in diameter. Brake-wear emissions factors in PARTS are assumed to be the same for all vehicle classes, although it could be assumed, as with tire wear, that the number of wheels, the weight of the vehicle, and the driving cycle would be significant contribut- ing factors related to the per mile emissions rate. As with tire wear, the dated reference suggests that the emissions factor is based on older mate- rials and needs to be updated. Issues for Model Revision EPA is planning to update PARTS after it completes the updates for MOBILES. As recommended above, such a revision should become part of MOBILE rather than being issued as a separate model. Issues that need to be addressed while updating PARTS are listed below: Data on particulate emissions from HDVs need to be updated to in- clude the effects of deterioration in emissions, adjustment for the benefits of I/M, and variations due to actual driving conditions.

726 MODELING MOBILE-SOURCE EMISSIONS Data from new studies on emissions from light-duty gasoline-powered and diesel-powered vehicles should be included in the updated particulate emissions model. Data collection efforts should be expanded to ensure that the effects of deterioration, I/M, and off-cycle driving conditions are included in the model. Data on modern tires and brake materials must be obtained for inclu- sion in future particulate emissions inventories. FLEET CHARACTERIZATION In MOBILE, the fleet is characterized by three parameters: age or regis- tration distribution, mileage-accumulation rate, and fleet or VMT mix. The age distribution gives the fraction of all vehicles in a particular class that are of a certain age. Because MOBILES accounts for 25 different ages (except for motorcycles), 25 fractions are required for each vehicle class, with the fractions summing to unity within each class. The mileage accu- mulation rate is the annual number of miles a vehicle is expected to be driven. It varies by vehicle age and class. The fleet mix gives the fraction of the fleet total VMT traveled by each of the eight vehicle classes. Again these fractions must sum to one. MOBILE calculates a vehicle class's emissions factor by computing the emissions factors for each of the model years, weights these by each model year's contribution to the vehicle class's total annual VMT, and then sums the weighted emissions factors. The weighting factors are termed travel fractions. The travel fraction, TFm, represents the fraction of the total VMT that is accounted for by a vehicle of age m years. It is calculated *om the fraction of vehicles registered that are m years old, REGm, and the annual mileage accumulation for these vehicles, MILESm, TF = REGm * MILES m MaxYears ~ (REGk M}~ESk k = ~ (3-7) where the summation is over all model years k. Once the vehicle-ciass- specific emissions factors are computed, the model then weights each of them by the corresponding fleet mix fraction and sums the results to pro- duce the fleet emissions factor. Although the user is allowed to enter custom registration distributions, mileage- accumulation rates, and fleet mixes, the model contains default (or for fleet mix, internally calculates) values for these parameters. The

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 727 default registration distributions are obtained from sales fractions through the 1980s. Mileage accumulation rates in MOBILES are based on 1990 National Purchase Diary (NPD) data. Fleet mix is internally calculated using registration distributions, mileage accumulation rates, diesel sales fractions, and total vehicle counts by class. MOBILE6 will see changes in all of the above parameters. These changes are documented in an EPA report (1999x). They are brought about by the availability of new data as well as shifts in methodology in some cases. The new model expands the previous 8 vehicle classes to 28, thus requiring much more detailed fleet characterization data. New mileage-accumulation rates for light-duty vehicles and light- and heavy- duty trucks are derived from the 1995 National Personal Travel Survey (NPTS) and the 1992 Truck Inventory and Use Survey (TIUS), respec- tively. Revised registration distributions are obtained from 1996 data compiled by the R.L. Polk company. Vehicle counts are based on data from various sources including the 1996 Polk data, 1998 Certification and Fuel Economy Information System (CFEIS) database, the Annual Energy Out- look (Energy Information Administration 1998), and a report by Ward's Communications (Pemberton 1996) which gives scrappage rates. Figures 3-16, 3-17, and 3-18 give sample comparisons of the MOBILES and MOBILE6 mileage-accumulation rates, registration distributions, and vehicle counts. Note that the new registration distribution is smoothed. This is a significant change from the former approach, which reflected ac- tual historical trends in sales and perpetuated them in all future-year cal- culations. SUMMARY AND RECOMMENDATIONS EPA is currently developing MOBILE6, the newest version of the MO- BILE model. This version is scheduled for release in the year 2000. In developing MOBILE6, EPA has used a more-open process, involving many stakeholders, and has been published much documentation for general review. In the development of MOBILE6, EPA has addressed many shortcom- ings of MOBILES, particularly those identified in the GAO (1997) report. The extent to which MOBILE6 has addressed those concerns is summa- rized in Table 3-12. The table also notes some improvements that can be made in future versions of MOBILE. In addition to the recommendations in Table 3-12, the committee offers the following recommendations for the improvement of MOBILE. These recommendations begin with changes to components in the existing

~ 28 M ODE1`NG M OB!LE-SOURCE EMISSIONS 20,000 1 8,000 1 6,O00 1 4,000 1 2,000 1 0,000 8,000 6,000 4,000 2,O00 + 1995 NPTS MOBILE5a (LDGV only) MOBI' E5a (LDDV only) Expon. (1995 NPTS) y = 1 s6g4e-o.oso6x O. ~ , 1 1 1 0 5 10 15 20 25 30 35 Age FIGURE 3-16 Light-duty vehicle annual mileage-accumulation rates. MOBILES uses the curve developed from the 1995 NPTS data. Source: EPA 1999x. MOBILE model. The next set of recommendations pertains to models that are closely related to MOBILE, such as those that estimate PM and air toxic emissions. The final recommendation deals with the need for long- range planning to guide the future development of the model. Obtain Better Data on High-Emitting Vehicles Establish a long-term testing program to characterize in-use deteriora- tion of representatively aged new-vehicle technology using a driving cycle more representative of actual driving conditions. This should focus on de- termining the nature of both exhaust and evaporative high emitters. Im- proved data on both the emissions rate and the *action of the vehicle pop- ulation that are high emitters are required. Inclusion of Road-Grade Effects in MOBILE The emissions increase from road grade is similar to that from accelera- tion and should be included in the model. This will be particularly impor-

TECHN/CAI ISSUES ASSOCIATED WITH THE M OBILE M ODEL ~ 2 9 0.1 - JO 0.08 a_ ~ 0~06 lo 0.04 - 0.02 Lo _~ , , , 'it O - 0 5 10 15 20 25 Age · MOBILES ~ MOBILE61 FIGURE 3-17 Comparison of MOBILE 5 and MOBILES light-duty vehicle registration distribution. Source: EPA 1999x. tent for areas where there is a significant amount of grade such as Denver and Spokane. Planning for this feature should include input from local regions that use MOBILE to ensure that grade information is available to potential users and that the model revisions are consistent with the avail- able formats of the grade data. Improve the Start-Emissions Database Routine tests of start emissions should be made as part of ongoing mea- surement programs unless there is confirmation that regression tech- niques, similar to those used for MOBILES, provide an effective estimation of start emissions. Additional measurements of the effects of ambient tem- perature, wind speed, and soak time on start emissions should be made to get a better representation of these important factors. Another factor that should be considered when estimating start emissions is the operating mode of a vehicle during the first minutes of operation. Modeling of Inspection and Maintenance Programs Benefits In particular, the treatment of vehicles that failed emission tests but never appeared for a retest, owners who never have their vehicles in-

30 MODEL/NO MOB!LE-SOURCE EMISSIONS 350,000,000- 300,000,000 250,000,000 200,000,000 150,000,000 100,000,000 50,000,000 7 - ,.~ O- . 1982 2002 2022 2042 Calendar Year —MOBILE 5 Linear Interpolation MOBILES FIGURE 3-18 Comparison of light-duty vehicle counts, 1982-2050. Source: EPA 1999x. spected, and the deterioration of vehicles after repair needs to be im- proved. The modeling of repaired vehicles' deterioration should be based on data from actual repaired vehicles. Improve the Emissions Factors for Heavy-Duty Vehicles Emissions factors for HDVs are woefully outdated and there are ques- tions about the conversion of engine dynamometer data into on-road gram- per-mile emissions. Appropriate chassis dynamometer cycles need to be developed for HDVs and data must be obtained on such cycles. Appropri- ate corrections for the effects of humidity and temperature, currently un- der development, should be incorporated into MOBILE. Data should to be generated for in-use conditions that might have significantly different emissions from those predicted based on engine certification tests. Updating of Fleet Characterization In recent years, there has been a significant increase in the use of light- duty trucks (especially sport-utility vehicles) instead of and in addition to passenger cars. EPA has updated fleet characterization data for MOBILES to reflect these current trends. EPA should at regular intervals (every 2 years or so) review the fleet characterization data, both current and pro-

TECHNICAL ISSUES ASSOCIATED WITH THE M OBILE MODEL 7 3 7 TABLE 3-12 Summary of Expected Changes to MOBILE6 That Respond to Problems Identified in GAO Report (1997) Area of Concern Regarding MOBILE Model Cited in GAO Report MOBILE6 Treatment of Issue 1. Emissions estimates for higher speeds, especially speeds in excess of 65 mph. Representation of emissions from rapid acceleration and deceleration, including aggressive clrlvlng behaviors. Representation of emissions immediately after engine start-up, known as cold-start emissions. 4. Representation of emissions *om air conditioner use. Representation of emissions from road grade, such as when a car climbs a hill. 6. Representation of high- emitting vehicles in the MOBILE's supporting database. 7. Representation of exhaust emissions from lower- polluting fuels, especially fuels with lower volatility (low RVP); representation of emissions from oxygenated fuels. MOBILE6 uses data obtained from recent studies on real-world driving conditions to develop facility-specific speed-correction cycles, which include higher speeds and aggressive driving behavior. The facility-specific speed-correction factors also provide greater distinction in roadway classifications. Start emissions have an improved treatment in MOBILE6; more study should be done to provide additional data for the approach proposed. MOBILE6 has an improved model of air-conditioner use. Additional data and model modifications could improve the estimates of this effect. Not addressed in MOBILE6. EPA used data from IM240 lanes to correct FTP data for recruitment bias in exhaust-emissions data. Special studies should be done to determine effect of high emitters. MOBILE6 exhaust emissions effects of low RVP fuels has not changed from MOBILES. MOBILE6 shows reduced benefits from oxygenated fuels, based on EPA analysis of more recent test data.

7 32 MODE1JNG M OBILE-SOURCE EMISSIONS TABLE 3-12 (Continued) Area of Concern Regarding MOBILE Model Cited in GAO Report MOBILE6 Treatment of Issue 8. Representation of emissions system deterioration for ve- hicles with 50,000 or more miles. 9. Emissions estimates and assumptions for vehicle I/M programs. 10. Estimates and assumptions for nontailpipe evaporative emissions when the vehicle is not operating. 11. Emissions estimates and assumptions for the inspec- tion and maintenance of HDVs—those with a gross vehicle weight of 8,501 pounds or more. Data characterizing vehicle fleet. Greater distinctions in road- way classifications. Quantifying the uncertainty of the model's estimates. New data have shown much lower emissions rates for such vehicles. These data have been used in MOBILE6. MOBILE6 shows reduced benefits from I/M programs. Questions re- main about the assumed benefits for OBD and the assumed deterioration of repaired vehicles. MOBILE6 includes updates to rest- ing loss emissions based on real- time (24-hr) test data. Not included in MOBILE6. EPA has updated fleet characteris- tics (fleet mix and age and mileage- accumulation distributions by vehi- cle classy for MOBILE6. See response to items one and two. Not included in MOBILE6. jected, to ensure that changes in the vehicle fleet are properly recognized in the emissions model. Complete Documentation of all Databases and Analyses EPA has done an excellent job of improving their documentation of the basic steps in the MOBILE model. Additional documentation should be provided to explain all the details of the analyses so that interested parties

TECHNICAL ISSUES ASSOCIATED WITH THE MOBILE MODEL 7 33 can readily check the analyses. Placing these detailed analyses and all the databases used on the internet would facilitate external review of the da- tabases and methods used in MOBILE. Integration of the PARTS and MOBILE Models PARTS and MOBILE do basically the same thing: compute actual emis- sions from on-road mobile sources. The separation of gaseous emissions in MOBILE and particulate emissions in PARTS is not necessary. It requires users to run two models instead of one and leads to the possibility that the on-road motor vehicle fleet and other important factors will not be treated consistently between the two models. When incorporating the PARTS model into MOBILE, several problems with PARTS need to be addressed. Updated emissions factors should be developed incorporating data on the effects of high-emitting vehicles (smoking vehicles), in-use deterioration, I/M programs, speed variations, and off-cycle emissions. Incorporation of the COMPLEX Model into MOBILE The effects of reformulated gasolines are currently estimated in a sepa- rate model called the COMPLEX model. Incorporating a COMPLEX-like model into MOBILE would allow states and regions to directly model gaso- line formulations with more stringent requirements than federal reformu- lated gasoline requirements. It would require the impacts of reformulated gasolines to be extended to include impacts on emissions from all vehicles and technology groups as well as the impacts on CO emissions. Incorporation of Toxic-Emissions Factors into MOBILE These emission factors are currently in a separate model, MOBTOX. The rationale for this recommendation is the same as that for the two pre- vious recommendations: convenience and consistency. More Emphasis on the NONROAD Model Although the EPA model, NONROAD, for off-road emissions sources is not part of MOBILE, the committee notes that this critical emissions model is lacking in data on emissions factors and activity levels. As more

~ 34 M ODE[/NG MOBILE-SOURCE EMISSIONS controls are placed on on-road sources, the off-road sources will become more important in the future, and the NONROAD model will play a larger role in estimating regional emissions inventories. Develop Long-Range Plans for the Evolution of the MOBILE Model The future implementation of new emissions and fuel standards, grow- ing concerns about PM and air tonics emissions, and the rising cost of con- trol strategies will increase the focus on MOBILE. Users will need im- proved accuracy and reliability from MOBILE, even as the regulatory set- ting, vehicle technologies, and fleet characteristics are changing. This poses a daunting challenge for EPA's Office of Transportation and Air Quality (OTAQ) to develop an accurate model that reflects uncertain regu- lations, unproven technologies, and shifting preferences among consumers. To address these demands, EPA must develop a long-range plan for ad- dressing critical modeling issues. EPA first should determine the most appropriate uses for the model and develop improvements to support these specific uses. This plan should then set priorities for model improvements that have the largest impact on emissions and develop a plan for collecting the necessary data to support these improvements. Most importantly, EPA must also address how close the modeling of mobile-source emissions should be to the development of regulations. The model must be seen as an accurate reflection of mobile-source emissions, not as a tool that is used to support proposed regulations.

Next: 4 Model Uncertainty and Evaluation »
Modeling Mobile-Source Emissions Get This Book
×
Buy Paperback | $75.00 Buy Ebook | $59.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

The Mobile Source Emissions Factor (MOBILE) model is a computer model developed by the U.S. Environmental Protection Agency (EPA) for estimating emissions from on-road motor vehicles. MOBILE is used in air-quality planning and regulation for estimating emissions of carbon monoxide (CO), volatile organic compounds (VOCs), and nitrogen oxides (NOx) and for predicting the effects of emissions-reduction programs. Because of its important role in air-quality management, the accuracy of MOBILE is critical. Possible consequences of inaccurately characterizing motor-vehicle emissions include the implementation of insufficient controls that endanger the environment and public health or the implementation of ineffective policies that impose excessive control costs. Billions of dollars per year in transportation funding are linked to air-quality attainment plans, which rely on estimates of mobile-source emissions. Transportation infrastructure decisions are also affected by emissions estimates from MOBILE. In response to a request from Congress, the National Research Council established the Committee to Review EPA's Mobile Source Emissions Factor (MOBILE) Model in October 1998. The committee was charged to evaluate MOBILE and to develop recommendations for improving the model.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!