United States Air and Radiation EPA420-R-01-060
Environmental Protection November 2001
Agency M6.SPD.002
&EPA Final Facility Specific
Speed Correction Factors
yŁu Printed on Recycled
Paper
-------
EPA420-R-01-060
November 2001
Final Facility Specific Speed Correction Factors
M6.SPD.002
David Brzezinski
Constance Hart
Phil Enns
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
NOTICE
This technical report does not necessarily represent final EPA decisions or positions.
It is intended to present technical analysis of issues using data that are currently available.
The purpose in the release of such reports is to facilitate the exchange of
technical information and to inform the public of technical developments which
may form the basis for a final EPA decision, position, or regulatory action.
-------
Table Of Contents
1.0 Summary 2
2.0 BACKGROUND 4
3.0 VEHICLE TESTING 7
3.1 New Driving Cycles 7
3.2 Sample Selection 8
3.3 Vehicle Testing 9
4.0 STATISTICAL ANALYSES 9
4.1 Emitter Status 11
4.2 Roadway Type 12
4.3 Vehicle Class 13
4.4 Emission Standard 13
4.5 Convergence of Freeways and Arterial/Collectors 14
4.6 Summary 15
5.0 EMISSION LEVEL CALCULATION 16
5.1 Freeway Versus Arterial/Collector Effects 16
5.1.1 High Speeds 17
5.1.2 Intermediate Speed Freeways 17
5.1.3 Low Speed Freeways 17
5.1.4 Arterial/Collectors 18
5.1.5 Extremely Low Speeds and Idle 18
5.2 Local Roadways and Freeway Ramps 19
5.3 Special Cases 19
6.0 SPEED AND OFF-CYCLE CORRECTION FACTORS 20
6.1 Basic Modeling Approach 20
6.2 Off-Cycle Adjustment 21
6.2.1 CO Off-Cycle Emission Effects 23
6.2.3 HC Off-Cycle Emission Effects 24
6.2.3 NOx Off-Cycle Emission Effects 24
6.3 Calculating Speed Correction Factors 25
6.4 Test for Model Stability 26
7.0 BENEFITS OF THE SFTP REQUIREMENT 27
7.1 Methodology used to calculate SFTP benefit 27
7.2 Applying the SFTP Benefit in MOBILE6 32
7.3 Applicability of SCFs to SFTP-compliant vehicles 33
-------
8.0 Application in MOBILE6 35
8.1 Light Duty Diesel Vehicles 36
8.2 Heavy Duty Vehicles 37
8.3 Motorcycles 37
8.4 High Speeds 38
9.0 COMPARISON TO MOBILES 38
References 40
Tables 42
Figures 72
Appendices 95
Appendix A : Statistics 96
Appendix B : Example 108
Appendix C : Response to Comments 115
Appendix D : Peer Review of Speed Corrections 135
Appendix E : Peer Review of Off-Cycle Effects 146
-------
ABSTRACT
In MOBILES, adjustments were made to the basic exhaust emission estimates to account
for the effects of area wide average trip speeds using speed correction factors developed from a
number of driving cycles with varying average speeds. For MOBILE6 EPA has adjusted for
differences in driving behavior versus roadway (facility) type and aggressive driving effects as
well as average speed. EPA has developed new facility-specific inventory driving cycles, based
on "real world" representative driving studies, and tested vehicles using these cycles to address
these purposes. This report describes the analysis of the new driving cycle data and presents the
resulting speed correction factors used in MOBILE6 for the gasoline passenger car and light duty
gasoline truck classes.
Final M6.SPD.002 1 June 2001
-------
1.0 Summary
Although the adjustments described in this document are called "speed" correction
factors, the adjustments include all of the effects on emissions caused by differences in driving
behavior, of which average speed is the most obvious and easiest to measure. The speed
correction factors described in this document are used in MOBILE6 to replace the speed
correction factors now used in MOBILES for all light duty passenger cars and light duty trucks of
all model years and technologies for average speeds above 7.1 mph. Low speed adjustments,
below 7.1 mph, will still use the MOBILES estimates. The speed correction factors for heavy
duty vehicles, diesel fueled vehicles and motorcycles from MOBILES would be retained for use
in MOBILE6. This document also describes the method for applying the new speed correction
factors to future technology vehicles for which no data is yet available.
The new MOBILE6 speed correction factors specifically account for aggressive driving
behavior not represented in older driving cycles. The effect of aggressive driving behavior is
accounted for separately using an emission offset to allow for future control strategies, such as
the Supplemental Federal Test Procedure for vehicle certification, to be explicitly modeled. The
new speed correction factors also allow for evaluation of vehicle emissions by roadway type
(facility) and by roadway segments (links). There are four roadway types modeled in MOBILE6:
Freeways
Arterial/Collectors
Freeway Ramps
Local Roadways
EPA recognizes that many factors, such as the number of lanes and other roadway
geometry, are not explicitly accounted for in the development of the four roadway types.
However, each driving cycle used includes a representative amount of the driving behavior on a
variety of roadways of that roadway type. EPA is confident that these four roadway types will be
sufficient to allow for better modeling of the wide variety of roadways found in urban areas than
previous models.
The speed correction factors for freeways and arterial/collectors depend on both speed
and basic emission levels of the vehicles. The correction factors for freeway ramps and local
roadways depend only on emission level and cannot be adjusted for average speeds different than
the national average. All speed corrections are based on new driving cycles designed to reflect
"real world" representative driving behavior, including the effects of aggressive driving not
found in the standard vehicle FTP certification driving cycle (Urban Dynamometer Driving
Cycle) and most older driving cycles used in emission testing.
Since the data for this analysis were collected using the new, representative driving
cycles, an emission impact of aggressive driving is included in the effect of the new speed
correction factors on emissions. The introduction of the new Supplemental FTP (SFTP)1 into
Final M6.SPD.002 2 June 2001
-------
vehicle emission certification will require the reduction of the emission effects from aggressive
driving for future vehicle certification.
Table 16 contains the MOBILE6 speed correction factors for freeways. Table 17 contains
the MOBILE6 speed correction factors for arterial/collector roadways. For MOBILE6, the
correction factor for Local Roadways and Freeway Ramps assume a national average speed and
will not have an adjustment for local average speeds. The speed correction factors for freeways
and arterial/collectors converge below 7.1 mph and at higher speeds, depending on the pollutant
and emission level. At those points the freeway and arterial/collector speed correction factors
become identical. The speed correction factors for speeds below 7.1 mph will remain the same
as in MOBILES, but are adjusted to account for the difference between the old and new speed
correction factors at 7.1 mph.
MOBILES did not model average speeds above 65 miles per hour. The new driving
cycles also do not address average speeds above 65 miles per hour. EPA will consider whether
sufficient information is available to model average speeds above 65 miles per hour in MOBILE6
and will present any proposals for these higher speeds in a separate document. As in MOBILES,
MOBILE6 will not explicitly address average speeds less than 2.5 miles per hour. Idle emissions
will be assumed to be the same as the grams per hour emitted at an average speed of 2.5 miles
per hour. This "idle" emission rate will be available as an output from MOBILE6.
Table 13 shows the coefficients used to calculate the freeway ramp and local roadway
emissions from the basic emission rate. Table 14 and 15 show the additive offsets used to
calculate the adjusted basic emission rate which is adjusted by the speed correction factors.
Appendix B has an example calculation of the application of speed correction factors to the base
emissions calculated by MOBILE6.
Figures 4a through 4c show the effect of emission level in the sample of light duty
gasoline vehicles as a function of speed for freeways, estimated using the new MOBILE6 speed
correction factors. Figures 5a through 5d show the MOBILE6 speed correction factors for
freeways for the three emission levels. Figures 6a through 7i compare the new MOBILE6 speed
correction factors with selected MOBILES speed correction factors. Care should be taken in
interpreting these figures, since there are many differences in how these factors are applied in
MOBILE6 as compared to MOBILES. These figures are discussed in more detail in Section 9.
This report is organized into sections which address various aspects of the analysis.
Section 2 gives a brief background of the need for new, facility based, speed correction
factors.
Section 3 discusses the development of the facility cycles and the emission testing sample
used in the development of the speed correction factors.
Final M6.SPD.002 3 June 2001
-------
Section 4 discusses the statistical analysis of the data sample.
Section 5 describes the approach developed to summarize the emissions data.
Section 6 uses the emission levels developed in Section 5 to develop the speed correction
factors and off-cycle effects used for MOBILE6.
Section 7 discusses how the new Supplemental Federal Test Procedure (SFTP) will affect
the estimate for off-cycle emissions and speed correction factors.
Section 8 describes how the new speed correction factors will be used in MOBILE6 and
how MOBILE6 will estimate the speed correction for other vehicle classes.
Section 9 compares the speed correction factors developed in Section 6 to the existing
speed correction factors in MOBILES.
2.0 BACKGROUND
EPA's highway vehicle emission factor model, MOBILE, is used for inventory modeling.
MOBILE has historically been based on emission testing using the Federal Test Procedure (FTP)
used to certify all light duty vehicles sold in the United States. The FTP uses a driving cycle (the
Urban Dynamometer Driving Cycle, commonly referred to as the LA42) which simulates urban
driving on a laboratory dynamometer. Correction factors for various conditions (e.g., average
speed, temperature, fuels) are applied to emissions measured at the FTP "standard" conditions.
The speed correction factors were based on test results for vehicles tested on both the LA4
(Urban Dynamometer Driving Cycle) and several other cycles, each having a different average
speed. MOBILE6 will address two areas not adequately addressed in previous versions of the
model. These are "real world" representative driving behavior and the expanded use of
transportation models in determination of area-wide inventories.
"Real- World Driving "
The FTP has been used for emissions certification of all light duty vehicles sold in the
United States. The Clean Air Act Amendments of 1990 mandated a closer look at "real-world
driving" - that is, driving modes that are not covered by the FTP (and the Urban Dynamometer
Driving Cycle) and representative of actual observed driving behavior. EPA organized the
Federal Test Procedure Review Project to address this mandate. A new Supplemental FTP
(SFTP) rule was finalized in October 1996.3 This rule specifies the addition of a new
certification cycle with more aggressive driving and associated vehicle emission standards.
MOBILE6 must address both the emission impacts of more aggressive driving than is
covered in the driving cycles that were used to develop MOBILES and the effects of the new
Final M6.SPD.002 4 June 2001
-------
SFTP standards on future model year vehicles. A special EPA emission testing project was
initiated to address these concerns. The results of that testing are the basis for the analysis in this
document.
Transportation Models
The current and older versions of the MOBILE model were developed to estimate area-
wide emission inventories using trip-based emission estimates with trip-based adjustments for
average speed. Vehicle trips are defined as all driving from key-on to key-off These vehicle
trips may include a variety of roadways and speeds.
Local officials have begun to integrate transportation models into their regional air quality
planning processes. Most transportation models represent the roadway system as a network of
"nodes," which are usually intersections, connected by "links." Each link represents a particular
type of roadway or "facility." Transportation models generate link-specific estimates of speed
and traffic volume. Transportation planners have begun using MOBILE to generate link-specific
emissions estimates for planning purposes.
Recent data from instrumented vehicles and chase car studies show that some types of
facility-specific driving contain more frequent and more extreme acceleration and deceleration
than others.4 Different facilities may have similar average speeds, but may differ significantly in
the amount of steady cruise. These differences suggest that there is a need to quantify the
emission differences (if any) between facilities in order to evaluate facility-specific speed related
traffic control measures in inventory modeling.
For example, at an average speed of 25 mph, travel over surface streets is likely to have a
relatively low level of traffic congestion, but will include many stops for traffic signals. Travel
on a freeway at 25 mph may indicate a high congestion level, but may include fewer stops.
MOBILES's trip based emission estimates do not differentiate between roadway types. If
MOBILES is used to model roadways separately, it cannot account for any differences in
emissions at similar average speeds resulting from these differences in driving behavior. This
particularly affects the planning process, where plans that affect different roadways cannot be
modeled adequately.
Other Approaches
California is also updating it's highway emission factor model.5 However, California has
taken a different approach to modeling the effects of changes in vehicle speeds. Rather than
attempt to discern what the driving behavior is for various facilities at various average speeds,
they divide all observed driving into speed bins. Each bin contains "microtrips" with similar
average speeds, regardless of the roadway type where the driving was observed. By weighting
the results of the various speed bins, any area-wide average speed can be modeled. Changes in
driving behavior can be modeled by varying the distribution of speeds.
Final M6.SPD.002 5 June 2001
-------
This approach requires areas to evaluate their fleet activity as "trips", where individual
vehicles travel over a variety of roadway types at varying speeds to reach a destination. These
trips are then used to develop a distribution of average trip speeds. Transportation models
generally do not produce trip statistics and transportation planners would need to adjust their
models to generate these distributions. Any changes in the roadway system resulting in changes
in average speeds on specific roadways will require a change in the full area-wide distribution of
trip speeds. Evaluation of the emission impact of changes in the specific roadways will require
new estimates of the area-wide emission levels.
One important advantage of California's approach is the need for fewer driving cycles.
Given limited testing budgets, this allows more vehicles to be tested over each cycle, thus
increasing the statistical confidence in the emission test results. Development of the driving
cycles themselves requires fewer assumptions such as decisions about where and under what
conditions the observed driving occurred. The resulting trip-based California driving cycles are
also similar in concept to the trip-based Unified Driving Cycle (or LA92),6 which is used by
California as the basis for the highway vehicle emission factors. The approach for MOBILE6
requires more driving cycles with more detailed information about driving conditions and
location.
The most important disadvantage of California's approach is the dependence on vehicle
trip information. Since vehicle trips occur over a variety of roadways at a variety of average
speeds, evaluation of trips is most relevant for only area-wide (i.e., county-wide) emission
estimates, where all trips can be assumed to begin and end within the area. The confidence in the
estimate of emissions will decrease as the size of the area to be modeled is decreased or if only
specific roadways or links are to be modeled. In addition, many transportation planners do not
currently generate trip speed distributions and other trip information from their models. This will
mean that changes will need to be made in the transportation models in many cases in order to
effectively use the California emission factors. In comparison, the MOBILE6 approach is more
compatible with analysis by roadway type and link. Since most transportation models already
estimate speeds and miles traveled by link, MOBILE6 will not normally require major changes in
the output from existing transportation models. Using MOBILE6, the area-wide emissions are
still able to be estimated by compiling the results from the four roadway types.
A more detailed description of the California approach or a comparison of the two
approaches is beyond the scope of this document. Readers are encouraged to obtain information
directly from California7 for comparison with the results documented in this report.
Final M6.SPD.002 6 June 2001
-------
3.0 VEHICLE TESTING
3.1 New Driving Cycles
The basis for the analysis found in this report is a set of light-duty, gasoline fueled
vehicles recruited and tested in 1997. The testing included new driving cycles specifically
designed to address the effects of in-use driving behavior on emissions. Table 1 gives a brief
description of the new cycles that were used in the testing. Collectively, these new driving
schedules will be referred to in this document as "facility" cycles. The driving behavior in each
driving schedule is selected from data collected from a particular roadway type during periods of
various congestion levels. These congestion levels have been roughly grouped into "levels of
service" (LOS) using letters A through G, similar to congestion category designations used in
transportation models. Briefly, LOS "A" refers to "free flow" (uncongested) situations and the
subsequent letters indicate increasing levels of congestion. The data used and the definition of
these categories is discussed in more detail in a separate EPA report describing the development
of the new facility driving cycles.8 Although the new cycles are labeled using a letter system
similar to that used by transportation agencies, there will be additional uncertainty in matching
conditions in the field with these new cycles due to the mismatch between LOS defined by air
agencies and transportation agencies. This can be alleviated somewhat by careful selection of the
average speeds and roadway types used in MOBILE6 to be those that best represent the driving
behavior observed in the field.
Table 2 compares the new cycles' statistics to the target population statistics for each
cycle. The statistics for each driving cycle will differ somewhat from the observed speed and
acceleration statistics which the cycle is designed to simulate (or "target population"). For
example, the highest average speed of the new arterial/collector cycles is 24.8 mph. Driving on
specific arterial/collector roadways can have average speeds higher than that. The maximum
speed of the arterial/collector cycles is only 58.9 mph, while the maximum speed observed in the
target population is 74.9 mph. This is a result of the cycle development process which chooses
the best combination of microtrips to match the target population. It is likely that the particular
microtrip which contained the maximum observed speed in the targeted population over-
represent certain aspects of driving behavior and was, therefore, not able to be used within the
confines of a single driving cycle of limited duration.
Each cycle was designed to result in emission levels representative of the emissions that
are expected from the driving behavior observed in the target population. Characteristics which
were deemed important to the match were specific power, speed, and amount of acceleration,
deceleration and idle. The factor which most affects emissions, shown from previous experience
in development of the Supplemental FTP, is the power distribution. The average speed or
maximum speed of the resulting cycles may not exactly match the target population. More
importantly, however, the cycles approximate the power distribution of the target population.
EPA feels that the emissions generated from the new cycles are a good representation of the
expected emissions from the driving behavior that was observed in the target population from
Final M6.SPD.002 7 June 2001
-------
which the cycle was generated. The development of these cycles is discussed in much more
detail in a separate document.9
It must be pointed out that however well the resulting driving cycles match the targeted
population data, the use of fixed driving cycles, by it's nature, cannot precisely match a particular
roadway modeling situation. Any difference between the targeted population data and the
specific modeling scenario will add an additional uncertainty to the modeling results. This
uncertainty will be reduced the more the specific modeling scenario resembles the targeted
population data used in development of the driving cycles. This will be the case in larger, area
wide analysis, where average driving behavior will likely better match the population data that
was used.
In addition to the new cycles described in Table 1, each vehicle was also tested using the
following cycles:
- Federal Test Procedure (FTP), with an additional hot running 505 seconds of the LA4
(Urban Dynamometer Driving Cycle).
- California Air Resources Boards (CARB) area-wide Unified Cycle (LA92).
- New York City Cycle (NYCC), a low speed cycle which has previously been used for
speed correction factors in the MOBILE model.
- ST01, a cycle based on instrumented vehicle data representing the beginning of trips
which is the first 258 seconds of the vehicle certification air conditioning cycle (SC03).
Table 3 shows more information on these additional cycles.
Although the New York City cycle (NYCC) was not developed specifically for
MOBILE6, it had been developed specifically to address the effects of "real world"
representative driving behavior at low average speeds. Although it has a low average speed (7.1
mph), it has maximum accelerations that are twice those found in the LA4 cycle. It was included
in the set of results used to generate the speed corrections factors to provide a data point below
11.6 mph (Arterial/Collector LOS E-F) in the analysis and provides a data point common to
earlier emission testing samples for comparison.
3.2 Sample Selection
The vehicle sample for this analysis came from EPA Emission Factor testing performed
at both the Automotive Testing Laboratories, Inc. (ATL), in Ohio and EPA's National Vehicle
and Fuels Emission Laboratory (NVFEL), in Ann Arbor, Michigan, in the spring of 1997. All of
the vehicles at ATL were recruited at Inspection and Maintenance lanes run by the State of Ohio,
and were tested in an as-received condition (without repairs). At the time of this analysis, a total
of 62 1983 through 1996 model year vehicles had been recruited and had completed testing in
Ohio, and 23 1990 through 1996 model year vehicles recruited and tested in Ann Arbor. The
sample of 85 vehicles includes 22 light-duty trucks. Most of the 85 vehicles were fuel injection,
Final M6.SPD.002 8 June 2001
-------
with 3 carbureted passenger cars and 4 carbureted light duty trucks. Only 12 of the vehicles
tested were certified to the Tier 1 exhaust emission standards.10 The rest are certified to the
earlier Tier 0 emission standards.
The vehicles tested at the EPA laboratory were recruited randomly. The vehicles tested at
ATL were selected as a stratified random sample, with strata corresponding to EVI240 pass or fail
outcome determined at state run EVI240 inspection stations in Ohio. ATL used the final phase-in
cutpoints recommended by EPA for use in I/M programs using the EVI240 test procedure to
identify vehicles in need of maintenance. Twenty of the vehicles in the ATL sample failed the
EVI240 test. Proper analysis of the ATL data requires careful weighting of the passing and failing
vehicles if emitter status is not considered as a factor in the analysis.
Table 4 shows the mix of EPA vehicle emission certification standards and fuel delivery
technology in the sample used in this analysis. Table 5 lists all of the vehicles individually,
showing vehicle make and model, odometer mileage, engine size and whether the vehicle passed
or failed an EVI240 test procedure using final phase-in cutpoints. Table 6 shows the mix of
model years and vehicle class (car or truck) in the sample.
3.3 Vehicle Testing
All vehicles were tested using the driving cycles described in Section 3.1 above in the as-
received condition using vehicle certification test fuel. Testing of vehicles was done on the
cycles in random order to reduce any order bias. Vehicles were tested at FTP ambient conditions
(i.e., temperature and humidity). Emission results were measured both as composite "bags" and
in grams second by second. Only the bag results were used in this analysis.
4.0 STATISTICAL ANALYSES
The purpose of the testing using the new facility cycles was to determine the effects of
"real world" representative driving behavior on basic (LA4) emissions. Separate cycles were
developed for freeway and arterial/collector roadways to allow comparison of those two roadway
types. The testing program also "over sampled" high emitting vehicles in order to provide a
sufficient sample size to allow separate analysis of high emitting vehicles. Although vehicle
mileage (or vehicle age) is considered important for estimating emissions, it is not thought that
vehicle mileage is a factor in the effect of average speed on emissions. Together, the following
testing and vehicle parameters were considered as potentially important in determining the effect
of average speed on emissions:
1. Emitter status.
2. Roadway type.
3. Vehicle class.
4. Exhaust Emission Standard.
Final M6.SPD.002 9 June 2001
-------
These parameters subsume a host of other potentially important parameters, such as
technology differences. The statistics of each list item will reflect the variability in emissions
from a larger set of parameters that are subsumed within the list item.
The data from the vehicles tested using the new facility cycles were evaluated to
determine if the effect of the average speed on emissions differed significantly by these
parameters. The sections below (Sections 4.1 through 4.4) discuss the statistical results for each
parameter evaluated. Section 4.5 discusses the statistical support for the convergence of the
freeway and arterial/collector estimates. Section 4.6 summarizes the conclusions derived from
the statistical analysis. Section 5 describes the final methodology.
Although the vehicle pool contains a large range of model years (1983 through 1996
model years), the total sample size (85 vehicles) was not sufficient for analysis explicitly by
model year or age. For this analysis, any model year dependency is assumed to be captured by
through the emission standards.
Table 7(a-c) shows sample means and standard deviations for the combined dataset for
each cycle, stratified into high and normal emitter levels. A vehicle may be a Normal emitter for
one pollutant, but considered a High emitter for another. In some cases the sample sizes
(Normal and High) do not sum to 85 vehicles. This is because some test results on some vehicles
were voided due to errors in the testing or sampling and could not be used. No valid emission
test results were eliminated from the analysis.
Figure l(a-c) graphically shows the effects of average speed on emissions. Each point is
the ratio of the mean for the emissions of each of the 14 facility cycles versus mean emissions for
the LA4 (Urban Dynamometer Driving Cycle) for the same vehicles. The data show that the
high emitting vehicles do not exhibit as much sensitivity to speed, resulting in smaller ratios.
It was expected that as the average speed increases the difference between emissions from
cycles representing arterial/collector roadways and emissions from cycles representing freeways
would decrease. An analysis was done to confirm the observed convergence of freeway and
arterial/collector roadway emissions versus average speed. This analysis is discussed in Section
4.5 below.
The method of analysis of variance was used to judge the effect of the above parameters
on the relation between average speed and emissions. The dependent variables in these analyses
were chosen to be the logarithm of grams-per-hour emissions. The grams-per-hour measure is
more stable than grams-per-mile, particularly at lower speeds, where very little distance is
traveled over a long time. The log transformation yields values that better satisfy the ANOVA
test requirements of normally distributed constant-variance errors. In the actual fitting of speed
correction factor equations, described in Section 5, gram-per-hour units were used for analysis at
average speeds less than 30 mph. However, at high speeds (average speeds above 30 mph),
using a linear fit and grams per hour units, when converted back to grams per mile, forces a curve
Final M6.SPD.002 10 June 2001
-------
shape (tailing downward) which does not match the data trends (tailing upward). For speeds
above 30 mph, gram per mile units were used. The resulting equation has the following form:
EmissionsAverageSpeed = Exp(A + B * AverageSpeed)
Where Emissions are reported in units of grams-per-hour for segments covering speeds
below 30 mph and Emissions are reported in units of grams-per-mile for segments at speeds
above 30 mph. The coefficients A and B are determined from the linear regression of the log
transformed speed cycle results and the average speed of the speed cycles.
Table 8 reports the ANOVA results in terms of p-values associated with tests of the
various factors described above. The p-value gives a concise way of judging statistical
significance. The p-value of a test is the smallest level of significance at which the null
hypothesis can be rejected. In these models, the null hypothesis states that the levels of a given
factor, e.g., roadway type, have equal effect on emissions. The level of significance for this test is
the probability of rejecting the null hypothesis when it is true. That is, of falsely concluding that
a difference exists. This will be referred to as a Type I error. By convention, the level of
significance is chosen to be arbitrarily small, typically 0.05, in order to limit the occurrence of
Type I error. If p is smaller than the chosen level of significance, the null hypothesis is rejected in
favor of the claim that a difference exists.
For example, in comparing the normal and high emitter classes of total hydrocarbons,
Table 8 reveals a p-value of 0.000 for the main effect of the emitter class. In graphical terms, the
main effect captures the intercept of a line relating (the logarithm of the) emissions to speed.
Thus, the small p-value provides support for the rather obvious hypothesis that high emitters
have different average emissions than normal emitters. However, for the interaction of emitter
class with speed, the p-value is 0.1411, implying that the difference in the slopes (the relationship
between emissions versus average speed) of the normal and high emitter lines (regressions) is not
statistically significant.
Further, more detailed ANOVA results are shown in Appendix A at the end of this report.
Below are the statistical results for the individual factors.
4.1 Emitter Status
The sample was separated into "emitter status groups" based on their Hot Running LA4
exhaust emissions. Hot Running LA4 are emissions that would result from an FTP test which
does not include any engine starts. These emissions are intended to be the basic unit of exhaust
emissions for use in MOBILE6.11 The emitter status groups were defined by the following
pass/fail cutpoints:
0.8 g/mi Total Hydrocarbons (THC)
15 g/mi Carbon Monoxide (CO)
Final M6.SPD.002 11 June 2001
-------
2.0 g/mi Oxides of Nitrogen (NOx)
These are the final phase-in outpoints recommended by EPA for use in I/M programs
using the IM240 test procedure to identify vehicles in need of maintenance. A vehicle is
considered a Normal emitter if its emissions are less than or equal to the cutpoint level for that
pollutant. It is considered a High emitter if its emissions exceed the cutpoint level for that
pollutant. Once a vehicle is identified by emitter status for a pollutant using the Hot Running
LA4 emission results, it is always categorized that way in this analysis, regardless of its emission
results on another driving cycle. The cutpoints were not used in combination. A vehicle could
be considered as a Normal emitter for the CO analysis even if it were designated as a High
emitter for NOx or THC.
Table 8 confirms that the average emissions differ statistically by emitter class. The
speed variable also is significant, i.e., emissions vary with average speed. However, except for
CO, the emitter class-speed interaction is statistically non-significant.
While it is not always the case that the other factors available for analysis in the data
sample - roadway type, vehicle type, and emissions standard - interact statistically with emitter
class, engineering judgement warrant modeling these factors separately for normal and high
emitters. Statistical conclusions for these factors are presented next.
4.2 Roadway Type
For modeling in MOBILE6, four roadway types are considered: arterial/collectors,
freeways, freeway ramps and local roads. With arterial/collectors and freeways, the range of
average speeds in the facility cycles overlaps at speeds below 30 mph. At higher speeds, only
freeway cycles are available. The interaction between roadway type and vehicle type and
between roadway type and emission standard was examined.
Figure 2 (a-c) shows the effects of average speed on emissions in terms of the ratio of the
means for the emissions versus emissions for the LA4 (Urban Dynamometer Driving Cycle) for
normal emitting vehicles. The cycles representing freeway driving and arterial/collector driving
are connected with lines to show the difference in these road types versus average speed. The
Unified Cycle (LA92), the Area-wide Non-Freeway cycle, Local Roadway cycle and Freeway
Ramp cycle results are also shown in the figures. The same vehicles were tested on all cycles, so
differences between freeways and arterial/collectors are controlled for the vehicle effect.
The emissions data were compared statistically to determine if there is reason to model
arterial/collectors and freeways separately. The ANOVA results appear in Table 8. For all
pollutants in the normal emitter class, the main effects are statistically significant. The speed
interaction effects also are significant, albeit marginally so for hydrocarbons. Among high emitter
vehicles, only NOx exhibits a significant difference between the arterial/collector and freeway
road types.
Final M6.SPD.002 12 June 2001
-------
Since only freeway cycles are represented at speeds over 30 mph, no comparisons of
roadway type are required. Local roadways and freeway ramps are represented by only a single
cycle each and therefore cannot be analyzed for the effect of average speed.
4.3 Vehicle Class
Of the 85 vehicles in the facility cycle sample, 22 are light duty trucks. The high emitters
among the trucks number three for NOx, six for CO and 10 for THC/NMHC. In the freeway and
arterial/collector roadway categories, for the normal emitters, the ANOVA results for passenger
cars versus light duty trucks in Table 8 show significant main effects. However, the interaction
with speed effects all are non-significant. For the high emitters, none of the vehicle type
comparisons is statistically significant at the 0.05 level.
For the local and freeway ramp driving cycles, the results are mixed for normal emitter
vehicles. NOx emissions differ at the 0.05 level on both cycles and CO is significant for the
ramp cycle. Among high emitters, vehicle type is not significant for any of the pollutants on
either cycle.
4.4 Emission Standard
It was expected that vehicles certified to the new Tier 1 exhaust emission standards would
exhibit a different response to average speed than the Tier 0 vehicles. Since the facility cycle
sample contains only 12 Tier 1 vehicles, a method was developed for increasing the sample size
by reclassifying a portion of the Tier 0 vehicles in the sample. Vehicles were defined as "Clean"
Tier 0 vehicles if their emissions were less than 70% of both the NMHC and NOx Tier 1
certification standard as measured on the standard FTP test. The Tier 1 standards are:
o NMHC standard: 0.25 g/mi (< 50,000 miles), 0.31 g/mi (>50,000 miles).
o NOx standard: 0.4 g/mi (< 50,000 miles), 0.6 g/mi (>50,000 miles).
A total of eight clean Tier 0 vehicles were identified by this criterion. One Tier 0 vehicle
(number 5016) had low FTP Bag 1 and Bag 3 emissions and technically qualified for
reassignment. However, because it had large Bag 2 and EVI240 emissions, it was not considered
representative of Tier 1 emission behavior and thus retained Tier 0 status under the new
definition. The clean Tier 0 vehicles were used both in the analysis of both Tier 0 and Tier 1
emission levels. Table 9 shows the subset of 20 vehicles used to represent Tier 1 emission
behavior. Tables 11 (a-d) show the average emissions for each driving cycle in the sample of
normal emitting Tier 0 vehicles, high emitting Tier 0 vehicles and the expanded sample of
vehicles considered normal emitting Tier 1 vehicles. Figure 3 (a-c) compares the Tier 0 and the
expanded Tier 1 sample of vehicles for the difference in the effects of average speed on
emissions. Emissions are shown in terms of the ratio of the means for the emissions versus
emissions for the LA4 (Urban Dynamometer Driving Cycle).
Final M6.SPD.002 13 June 2001
-------
The ANOVA results in Table 8 compare emissions of the Tier 0 and Tier 1 vehicles for
the reallocated sample. On the arterial/collector and freeway cycles, for normal emitters the
emissions standard main effect is highly significant for all pollutants, and the interaction with
speed is significant for hydrocarbons. (The results are similar for the official emission standard
classification.) For the local and freeway ramp facility cycles, all main effects are also significant.
There were no high emitter Tier 1 vehicles for any of the pollutants, so no test of the
standard factor can be made for that emitter class.
4.5 Convergence of Freeways and Arterial/Collectors
The data show a statistical difference between the freeway and arterial/collector cycles
below 30 mph, where the data overlaps. However, there are no arterial/collector cycles above
24.8 mph and there are no freeway cycles below 13.1 mph. If the speed correction factors for
both of these roadway types are to cover the entire spectrum of average speeds available in the
MOBILE6 model (0 to 65 mph), then some assumptions about the effect of average speed on
emissions will need to be made for the speeds outside the typical range for these roadways.
Based on the facility cycle emission testing results, it appears that as average speed
increases there is a decrease in the difference between emission results for arterial/collector
cycles and freeway cycles at the same average speed. This suggests, that above a certain average
speed, the same relationship between average speed and emissions can be used for both freeways
and arterial/collector roadways.
Support for the hypothesis that mean gram-per-hour emissions of arterial and highway
driving converge in the neighborhood of 30 mph can be found in the data from tests on the cycles
that represent these two forms of driving. Consider the following model of emissions:
Y = b0 + bjX + b2*D + b3X*D
where Y is emissions (in grams/hour) of a given pollutant; X is average speed of the cycle tested;
and D is a dummy variable representing road type (D = 0 for arterial, D = 1 for highway). This
equation effectively models two lines. When D = 0, the function estimates emissions versus
speed for arterials, with slope bl and intercept b0. When D = 1, the line represents highway
emissions with slope (bx + b3) and intercept (b0 + b2).
This model is useful for examining differences between arterial and freeway emissions.
The basic question of whether the linear functions differ is answered by testing the coefficients of
terms involving variable D. If both these coefficients (b2 and b3) are zero, then the road types are
judged to be the same. For the 85 car sample, tests of this hypothesis are rejected for all
categories of emission standard and emission level.
Given that arterial and highway speed-emissions lines are significantly different, we now
Final M6.SPD.002 14 June 2001
-------
ask if they differ at a chosen speed, e.g., 30 miles per hour. This is answered by constructing an
appropriate function of the linear model described above. When X = 30, the function becomes:
Y= b + SO forarterials
Y= b0 + b1*30 + b2*l+b3*l*30
= (b0 + b2) + (bj + b3)*30 for highways
The two functions are identical when the linear combination b2 + b3*30 equals zero. This
hypothesis can be tested using the ESTIMATE feature of the SAS GLM procedure.
Table 10 presents results of these tests for Tier 0 normal and high emitters, and for Tier 1
normal emitters. At the five percent level, a significant difference is found in only in one case,
for Tier 0 normal CO emissions. This gives strong support for the claim that arterial/collector
roadway and freeway emissions are similar at speeds around 30 mph, even though their
relationship at average speeds below 30 mph is different. Based on this convergence, EPA has
concluded that the relationship between average speed and emissions for arterial/collector
roadways and freeways should be the same at average speeds above about 30 mph.
4.6 Summary
The statistical analysis of the important parameters resulted in the following decisions
about how the data would be grouped for the MOBILE6 analysis:
Roadway Type
There will be different equations for the two roadway types (freeways and
arterial/collectors) for CO and NOx emission at both High and Normal emitter groups. There
will be different equations for the two roadway types for THC and NMHC emissions only for
normal emitting vehicles. Since the equations converge, there will be only one equation for all
roadway types and pollutants at average speeds above about 30 mph. The exact average speed
where the equations converge varies. For high emitting Tier 0 vehicles there will be no
difference between the two roadway types for THC and NMHC emissions at any average speed.
Vehicle Class
There will not be different equations for vehicle class (car versus truck). The equations
used will depend on emission level (below), which will adequately cover any emission standard
differences between cars and trucks. Splitting the data by both emission standard (below) and
vehicle class would make sample sizes much too small for any meaningful results.
Final M6.SPD.002 15 June 2001
-------
Emission Standard
There will be separate equations for Tier 0 and Tier 1 emission standard vehicles for
normal emission levels. There are no high emitting Tier 1 vehicles in the sample.
Emission Levels
The Tier 0 emission standard data will be further separated by emitter status (Normal and
High) for all pollutants with separate speed equations for each. For the purpose of analysis, this
effectively results in three samples of vehicles representing three distinct emission levels:
Level 1 : Tier 1 (Normal emitter)
Level 2 : Tier 0 (Normal emitter)
Level 3 : Tier 0 (High emitter)
5.0 EMISSION LEVEL CALCULATION
Once the appropriate aggregations for the existing data were determined as described in
the previous section, least square linear regressions were fit to the emission results versus
average speed. This was done in a "multi-linear" fashion (piece-wise linear function,
continuous), rather than using a single line or using another non-linear curve shape. Attempts to
fit non-linear curves to the total data sample resulted in unacceptably high error coefficients. A
linear fit of smaller groupings of the data provided a closer fit to the data. A separate linear
regression was done for different groupings of cycles based on ranges of average speeds.
Together, these lines will define the change in emissions of the sample over the entire range of
average speeds.
5.1 Freeway Versus Arterial/Collector Effects
As discussed in the previous section, the data show a statistical difference between the
freeway and arterial/collector cycles below 30 mph, where the data overlap. However, there are
no arterial/collector cycles above 24.8 mph and there are no freeway cycles below 13.1 mph. If
the speed correction factors for both of these roadway types are to cover the entire spectrum of
average speeds available in the MOBILE6 model (0 to 65 mph), then some assumptions about
the effect of average speed on emissions will need to be made for the speeds outside the data
range.
Logically, both curves will converge at idle (zero mph). Idling emissions should not
depend on roadway type. Also, it is logical to assume that driving which has a high average
speed must consist almost entirely of cruise with little stopping or idle, regardless of roadway
type. This suggests a model where freeways and arterial/collector roadways have different
emissions at normal arterial/collector average speeds, but have the same emissions at extremely
Final M6.SPD.002 16 June 2001
-------
low speeds (and idle) and at higher speeds. Based on this model, EPA has defined the following
speed/facility segments:
o High Speeds (above about 30 mph) for both freeways and arterial/collectors.
o Intermediate Speed Freeways (from 13.1 to about 30 mph) for freeways.
o Low Speed Freeways (from 7.1 to 13.1 mph) for freeways.
o Arterial/Collectors (from 7.1 to about 30 mph) for arterial/collectors.
o Extremely Low Speed and Idle (less than 7.1 mph) for both freeways and
arteri al/coll ector s.
MOBILE6 will use a combined emission estimate for both arterial/collector and freeway
facilities for THC and NMHC at the highest emission level. This will mean that, at high emitting
THC and NMHC emission levels, that there will be no emission difference between the two
facility types. There are still separate freeway and arterial/collector estimates for CO and NOx
emissions at high emitting levels.
5.1.1 High Speeds
A regression was done of emissions versus average speed for the three emission
standard/emitter groups described above for the four freeway cycles with an average speed above
30 mph (Freeway at 30.5 mph, Freeway at 52.9 mph, Freeway at 59.7 mph and Freeway at 63.2
mph) in grams per mile for each pollutant. Tables 12a, 12b, 12c and 12d show the results of
those regressions. All of the slope coefficients of the regressions are statistically significant,
meaning that, with high probability, the increase or decrease in emissions versus average speed is
different than zero. These regressions will be used to estimate the emissions of vehicles on both
freeway and arterial/collector roadways at average speeds above the point where the equations
converge.
5.1.2 Intermediate Speed Freeways
A linear regression was done of emissions versus average speed for each of the emission
standard/emitter groups described above for the four freeway cycles representing freeway driving
in the most congested conditions (Freeway at 13.1 mph, Freeway at 18.6 mph and Freeway at
30.5 mph) in grams per hour for each pollutant. Tables 12a, 12b, 12c and 12d show the results of
those regressions. These regressions will be used to estimate the emissions of vehicles on
freeways between average speeds of 13.1 mph and about 30 mph. Note that the freeway cycle at
30.5 mph was included in both the intermediate speed freeway and high speed estimates. It is
expected that the two regressions should converge at about this average speed.
5.1.3 Low Speed Freeways
None of the existing facility cycles for freeway driving have an average speed below 13.1
mph. It will be assumed that at speeds lower than 7.1 mph (the average speed of the New York
Final M6.SPD.002 17 June 2001
-------
City Cycle) the effect of average speed on emissions will be the same for freeways and
arterial/collector roadways. The emissions of freeway driving for average speeds between 13.1
mph and 7.1 mph will be calculated by linear interpolation between these emission levels in
grams per hour. Tables 12a, 12b, 12c and 12d show the resulting equations representing this
interpolation. Most freeway travel will occur at average speeds well above this range.
5.1.4 Arterial/Collectors
The freeway cycle at 30.5 mph (already included in the freeway estimate) was included in
the arterial/collector roadway estimates as well. It was shown that the two regressions should
converge at about this average speed. The New York City Cycle was also included in the
arterial/collector roadway estimates. The New York City Cycle was not derived from the same
chase car or instrumented data used to develop the other facility cycles. However, the New York
City Cycle was originally developed as a speed correction cycle and, as shown in Table 3, does
contain acceleration rates higher than those contained in the LA4 (Urban Dynamometer Driving
Schedule). It was deemed that the New York City Cycle was representative of "real world"
representative driving and could be included in the analysis as another facility cycle.
A linear regression was done of emissions versus average speed for each of the emission
standard/emitter groups described above for the arterial/collector cycles (Arterial/Collector at
11.6 mph, Arterial/Collector at 19.2 mph, Arterial/Collector at 24.8 mph) in grams per hour for
each pollutant. Included in that regression was data from the New York City Cycle (with and
average speed of 7.1 mph) and the Freeway at 30.5 mph cycle for the same vehicles.
Tables 12a, 12b, 12c and 12d show the results of those regressions. These regressions
will be used to estimate the emissions of vehicles on arterial/collector roadways in this range of
average speeds.
5.1.5 Extremely Low Speeds and Idle
No data was collected for the vehicles in the sample at speeds lower than 7.1 mph (the
average speed of the New York City Cycle). In this range the model will assume that the effect
of average speed on emissions will be the same for freeways and arterial/collector roadways.
Since the MOBILES model already has estimates for the effect of average speed on vehicles at
speeds from 2.5 to 7.1 mph, and since there is no need to differentiate this effect by facility type,
the existing speed correction factors in MOBILES will be used for this range of average speeds
for both freeways and arterial/collectors.
The MOBILES speed correction factors do not match the new speed correction factors at
7.1 mph. This discontinuity will be resolved by adding the difference in the two estimates to
values calculated using the old MOBILES speed correction factors. As in MOBILES, emissions
at idle will be assumed to be the same (in grams per hour) as the emissions at 2.5 mph (the
lowest average speed modeled).
Final M6.SPD.002 18 June 2001
-------
5.2 Local Roadways and Freeway Ramps
There is only one cycle each to represent driving on local roadways and freeway ramps.
As a result, these cycles are not included in the analysis of emissions versus average speed.
However, the data from these cycles were separated using the same sample splits by emission
standard (Tier 0 versus Tier 1) and emitter status (Normal versus High) as are used for the
freeway and arterial/collectors. The average emission levels were analyzed as a quadratic
function of the base emission rate (hot running LA4 emissions). The local roadway regressions
do not include a constant (gram per hour) value, since it is assumed that driving on these
roadways does not include offcycle driving behavior. These regressions will be used to estimate
the emission levels for these roadway types as a function of the base emission rate calculated in
MOBILE6. The coefficients for these regressions are shown in Table 13.
5.3 Special Cases
Ideally, the equations above would define a rational, smooth relationship for emissions
versus average speed for the range of 0 to 65 mph for each pollutant based on the available data.
However due to vagaries of using observed driver behavior data and the use of a multi-linear
modeling approach, some of the equations resulting from the general approach will cause small
discontinuities in the overall relationship. For example, the intermediate speed freeway emission
level for NOx (computed in gram per hour) does not intersect with the high speed freeway
emission level estimate (computed in grams per mile) at any speed. These discontinuities will
require special handling to be coded mathematically. For MOBILE6, some basic "rules" will be
used to assure that there are no abrupt or counter-intuitive changes in emissions versus average
speed.
1) If at 30.5 mph, the emission estimate for the intermediate speed freeway equation is still
higher than the emissions for freeways calculated using the high speed equation, then the
emission value calculated for 30.5 mph using the intermediate speed freeway equation
will be used for speeds greater than 30.5 mph until the value for the high speed equation
for that speed exceeds the intermediate speed freeway value. This rule keeps the
intermediate speed freeway value from increasing beyond the emission level calculated at
30.5 mph, which is the highest average speed data point used in the regression (no
extrapolations).
2) When calculating the emissions of an arterial/collector roadway, the arterial/collector
estimate for emissions will be used unless the estimate for freeways at that same speed
are higher than the arterial/collector estimate. This rule defines at what average speed the
arterial/collector and freeway emission estimates will converge. Above that speed the
arterial/collector and freeway emission estimates will be assumed to be the same. All of
the MOBILE6 arterial/collector equations intersect with the freeway estimate between 24
and 34 mph.
Final M6.SPD.002 19 June 2001
-------
6.0 SPEED AND OFF-CYCLE CORRECTION FACTORS
Using the methods in the previous section, the emission data can be described as a series
of continuous, smooth functions for the two roadway types (freeways and arterial/collectors) by
emission levels for all pollutants over the entire range of average speeds in MOBILE6 (2.5 to 65
mph). This generalized relationship between emissions and average speed for any emission level
is referred to in the model as speed correction factors. For the freeway and arterial facility types,
these speed correction factors are the values which will be stored and used in MOBILE6 to adjust
exhaust emission estimates to account for average speed. As discussed in the following section,
an additional correction factor is applied for the freeway and arterial facility types to account for
"off-cycle" emissions, separately from average speed.
6.1 Basic Modeling Approach
The basic exhaust emission rate generated by MOBILE6 will be based on a hot running
LA4 emission estimate with an average speed of 19.6 mph. In MOBILE6, freeway ramp and
local roadway emissions do not depend on speed and can be determined directly from the basic
exhaust emission rate. For freeways and arterial/collector roadways, the adjustment to account
for the average speed and facility type is supplemented by an additional adjustment to account for
"off-cycle" emissions. The off-cycle adjustment is meant to capture the change in emissions
resulting from higher power operation not reflected in changes in average speed. The speed
correction factor (SCF) represents the change in emissions which results from a redistribution of
vehicle operating modes which occurs an different levels of roadway congestion. The
segregation of off-cycle and average speed effects was made in MOBILE6 primarily to assess the
benefit of the Supplemental Federal Test Procedure (SFTP) requirement, as discussed later in
Section 7.
In MOBILE6, the running basic emission rate is adjusted using the following general
method to account for off-cycle and average speed effects:
Local Roadways and Freeway Ramps:
Adjusted BER = BER * CF
Where:
BER = Basic Emission Rate (running emissions for the LA4 cycle).
CF = Multiplicative correction factor
Freeways and Arterials:
Adjusted BER = (BER + OC) * SCF
Final M6.SPD.002 20 June 2001
-------
Where:
BER = Basic Emission Rate (running emissions for the LA4 cycle).
OC = Off-Cycle Emissions Offset (a function of BER emissions)
SCF = Multiplicative Speed Correction Factor (a function of speed and emissions)
The calculation of the correction factor for freeway ramps and local roadways in
discussed in Section 5. The remainder of this section addressed the development of off-cycle
emissions adjustment and freeway/arterial speed correction factors.
Using the above equation, with a BER identical to the average hot running LA4
emissions of each sample of vehicles in each of the three emission level groupings described in
Section 4.6, the estimate of emissions at any speed for each facility will match the average
emission level predicted by the regression equations from the facility cycle data from that vehicle
sample. For cases where the BER is not identical to the average hot running LA4 emissions of
any of the facility cycle sample emission level groupings, the off-cycle adjustment OC will still
be calculated as a function of the BER, however the SCF will be interpolated using the three
emission level sample estimates. The interpolation would be determined by the emission level of
the sum of the BER and the OC. There are three cases, which cover the generation of SCFs for
any base emission level:
o If the emissions were equal to one the three predetermined emission level
thresholds, the SCF from that level would be used.
o If the emissions are between levels, the SCF would be interpolated between the
values for those levels.
o If the emissions were below the Level 1 level or the above the Level 3 level, Level
1 and Level 3 SCF would be used, respectively.
It is important to note that the latter case results in the Level 1 SCFs being applied for
most vehicles and trucks under the NLEV and Tier 2 emission standards. As discussed in
Section 7, the effects of the SFTP rule will reduce the overall adjusted emissions for vehicles
under these programs, but the relative magnitude of the SCF (i.e. the shape of the SCF curve)
will remain unchanged from Level 1.
6.2 Off-Cycle Adjustment
It has been long recognized that the FTP does not reflect vehicle operation at higher
speeds and accelerations, which contribute significantly to overall emissions. The maximum
speed of the FTP is 57 miles per hour (mph), and the maximum acceleration (determined by
limitations in chassis dynamometer technology in early 1970's) is 3.3 mph per second. A more
direct measure of cycle stringency is power, a combined measure of speed and acceleration
Final M6.SPD.002 21 June 2001
-------
which has been found to correlate well with emissions. One metric for power is "specific
power", calculated according to the following equation:
specific power = Vf2 - V;2, Vf > V;
Where V; and Vf are the initial and final velocity, respective to a one-second interval.12 Using this
metric, the average and maximum specific power for the FTP is 192 mph2/sec. In contrast, data
from Baltimore gathered in the early 1990's as part of the FTP review project's in-use driving
surveys found a maximum specific power level of 558 mph2/sec. More importantly, the distribution
of specific power for Baltimore showed a higher proportion of higher specific power events than the
FTP; for example, 2.6 percent of operation on the FTP is at specific power over 100 mph2/sec, while
6.4 percent of Baltimore events were above this threshold.13 This emissions implications of this are
significant given the contribution of these events to overall emissions.
In MOBILE6, an off-cycle correction factor was developed separately from the speed
correction factor in an attempt to isolate the effects of high power operation. An alternate
approach would have been to create a single correction factor which accounted for both "off-
cycle" and speed effects; this approach was suggested by Guensler in an independent peer review
of the off-cycle methodology, contained in Appendix E. We are maintaining these effects as
separate adjustments, however, because emissions resulting from aggressive driving behavior and
emissions resulting from changes in average cycle speed reflect two different processes which
should be accounted for separately in assessing the benefit of the off-cycle provision of the
SFTP requirement.
The off-cycle adjustment reflects the emissions change resulting from average power
levels higher than those on the FTP, independent of average cycle speed. In order to isolate the
high power effects separate from changes in average cycle speed, the off-cycle adjustment was
derived by comparing running LA4 emissions to an emission test cycle of comparable average
speed used in the development of the freeway speed correction factor, the Freeway Level Of
Service (LOS) F cycle. The average speed of this cycle is 18.6 mph, but the maximum
acceleration rate is 6.9 mph/s, as opposed to 3.3 mph/s over the LA4. The average specific
power of the Freeway F cycle is 44 mph2/s, 15 percent higher than the 38 mph2/s of the LA4. It
was thus assumed that significant emissions differences between this cycle could be attributed to
the difference in cycle stringency, or off-cycle emissions. The off-cycle correction factor is a
function of base running emissions, relating the difference in emissions between the freeway
LOS F cycle and the LA4 to base running (LA4) emissions, as shown in the following equation:
Off-Cycle (g/mi) = Freeway F(g/mi) - LA4(g/mi)
The data underlying the off-cycle emission effects is from EPA testing of 85 cars and trucks
with model years 1983 through 1996 (the same sample used for generating the SCFs). The sample
consisted of 63 cars and 22 light trucks that received emissions tests on both the LA4 and Freeway
F cycles. The additive approach of the equation above was used for all three pollutants. The overall
Final M6.SPD.002 22 June 2001
-------
model hypothesis was that ultimately only a vehicle's base LA4 emissions was significant in
determining the off-cycle emissions. Other variables potentially significant to emissions generation
such as vehicle type, model year, and fuel delivery system were accounted for in the base emission
rates, and were eliminated from the model through stepwise regression because of insufficient
statistical significance.
6.2.1 CO Off-Cycle Emission Effects
Prior to the analysis of the CO emission off-cycle effects, the data was examined in a
graphical form, and found to contain considerable scatter for vehicles with high CO emissions.
12 data points were between 10 and 100 grams/mile, in general an order of magnitude higher
than the remaining 73 data points. As a result of the scatter and the potential for these high
emission vehicles to have an unrepresentatively large influence on the final model, it was decided
to separate the sample into high and normal emitters based on a given vehicle's overall FTP
emission level. The threshold emission level dividing the 'normal' and 'high' emitters that was
chosen was 10.2 g/mi over the FTP, or three times certification standard. This level was chosen
to maintain consistency in the definition of 'high' emitter throughout MOBILE6.
Least squares regression was performed separately on both the normal and high emitter
samples. The CO Off-Cycle emissions as defined in Equation 4 were least squares regressed
versus LA4 CO emissions, LA4 CO emissions squared, fuel delivery type, vehicle type (car or
truck), and model year. For the 'normal' emitters, stepwise regression with a 5 percent
significance requirement (p-value < 0.05) eliminated all of the variables except the LA4 CO
emissions and LA4 CO emission squared terms. For the 'high' emitters, stepwise regression
eliminated all of the variables. A subsequent regular regression showed that none of the
variables were close to being statistically significant. All of the statistical results are shown in
Appendix A.
The fit which was chosen to represent the CO off-cycle emissions in normal emitters in
MOBILE6 was a quadratic linear fit of the CO off-cycle offset versus running LA4 CO emissions.
This regression fit was force through zero because it is believed that off-cycle emissions will be zero
on zero emitting LA4 vehicles. The regression equation proposed for use in MOBILE6 for normal
emitters is shown the the following equation
OCCO = 0.984 * LA4CO - 0.07638 * (LA4CO)2
Where:
OCCO is the CO emission increase due to off-cycle operation in g/mi.
LA4CO is the base CO emission level over the LA4.
A regression analysis of the high emitting vehicles resulted in no significant predictors of the
off-cycle offset, including LA4 CO emissions. The average results was also not significantly
Final M6.SPD.002 23 June 2001
-------
different from zero. Thus, we are establishing the off-cycle offset for CO high emitters at zero. A
likely explanation for this is that high emitting CO vehicles are likely already experiencing excess
enrichment over the LA4, which the increase in cycle stringency is not exacerbating.
The CO Off-Cycle effect as a function of LA4 CO emissions is shown in Figure 8a. The line
annotated with circles is CO off-cycle effect for the 'Normal' emitters. The zero line is the CO off-
cycle effect for 'High' emitters.
6.2.3 HC Off-Cycle Emission Effects
In the case of hydrocarbon emissions, the base LA4HC emission level and the squared value
of the LA4HC were found to be the only statistically significant variable using as stepwise
regression process. Segregation of the data into low and high emitters was not found to be necessary
in the case of HC offset. The full regression statistics are shown in Appendix A.
The equation predicting the NMHC off-cycle offset in units of g/mi is shown in the following
equation. The equation and the underlying data are also shown in graphical form in Figure 8b.
OCHC = 0.305 * LA4HC - 0.02492 * (LA4HC)2
Where:
OCHC is the HC emission increase due to off-cycle operation in g/mi.
LA4HC is the base HC emission level over the LA4.
Since the THC off-cycle emission fit is a quadratic in form, it produces a peak offset which
declines and eventually goes negative. The maximum THC off-cycle offset is 0.933 g/mi and occurs
when the base LA4 THC emission level equals 6.12 g/mi (See Figure 2). It was decided that the
value of the THC offset would be fixed at 0.933 g/mi for all LA4 emission levels exceeding 6.12
g/mi rather than let it decline. The rationale for this assumption is that it is counter-intuitive to
expect that the lightly loaded LA4 cycle would produce HC emission levels which are less than or
even declining relative to the Freeway F emission cycle. The HC offset function's maximum occurs
at a LA4 emission level which is virtually out of the range of the running LA4 HC emission levels
in the model. Only very high emitting older model year light vehicles will approach a value of 6.12
g/mi running LA4. Thus, the assumption of a maximum offset value of 0.933 g/mi THC will rarely
be invoked.
6.2.3 NOx Off-Cycle Emission Effects
Similar to HC, for NOx only the base LA4 emission level, its squared value, and an intercept
forced through zero were found to be statistically significant. Segregation of the data into low and
high emitters was also not found to be necessary in the case of NOx offset. The full regression
statistics are shown in Appendix A. The equation predicting the NOx off-cycle offset in units of
g/mi is shown in Equation 8.
Final M6.SPD.002 24 June 2001
-------
Since the NOx off-cycle emission fit is a quadratic form, it produces a peak offset which
declines and eventually goes negative. The maximum NOx off-cycle offset is 0.58 g/mi and occurs
when the base LA4 NOx emission level equals 3.50 g/mi (See Figure 3). At higher NOx levels,
the downward trend is dictated by limited data, and as a result we have decided to impose an
artificial cap of 0.58 g/mi for all LA4 NOx emission levels exceeding 3.50 g/mi rather than let it
decline. On late model light vehicles, only the very high emitter levels will approach a value of 3.50
g/mi running LA4 (for comparison the current FTP certification standard is 0.40 g/mi NOx). Thus,
the assumption of a maximum NOx off-cycle offset value of 0.581 g/mi will rarely be invoked.
OCNOX = 0.332 * LA4NOx - 0.04745 * (LA4NOx)2
Constraint: LA4NOx <= 3.50 g/mi
OCNOX = 0.58 g/mi
Constraint: LA4NOx > 3.50 g/mi
Where:
OCNOX is the NOx emission increase due to off-cycle operation in g/mi.
LA4NOx is the NOx emission level over the LA4.
6.3 Calculating Speed Correction Factors
As discussed in Section 6.2, MOBILE6 adjusts the basic exhaust emission rates (BER) by
first adding an emission value which accounts for off-cycle emissions (OC). From this point,
SCFs are generated to account for changes in facility type and average speed. For freeway and
arterial roadways, the SCF is applied as a multiplicative adjustment directly to the sum of the
BER and the off-cycle adjustment.
The SCF is defined as the ratio of the predicted emissions at any average speed to the
predicted emissions at 19.6 mph for freeways for the same vehicle. Using the emission level
equations described in Section 5, a set of SCFs will be determined for each speed in increments
of 5 mph beginning at 5 mph through 65 mph and at 2.5 mph for each of the three emission
levels within MOBILE6. These increments correspond to the increments of average speed for
the VMT average speed distribution for freeways and arterial/collector roadways in MOBILE6.
MOBILE6 calculates these speed correction factors directly from the emission levels, rather than
store the resultant speed correction factors themselves. Table 16 shows the freeway SCF sets for
the three emission levels. These SCF sets are shown graphically in Figures 5a, 5b, 5c and 5d.
Table 16 shows the freeway emissions at 19.6 mph for each emission level. Table 17 shows the
arterial/collector SCF sets for the three emission levels.
The vehicle miles traveled (VMT) average speed distribution for freeways and
arterial/collector roadways in MOBILE6 represent the distribution of speeds for the area to be
Final M6.SPD.002 25 June 2001
-------
modeled. Each average speed bin is the fraction of all miles traveled which occurs within a
given speed range. For example, the 30 mph bin will include the miles traveled on links where
the average speed is between 27.5 mph and 32.5 mph. The sum of the distribution fractions must
always be one. If only a single roadway is to be modeled, the average speed can be entered as
two adjacent bin values whose average speed matches the average speed for the roadway. In this
case, all other speed bin values would be zero. MOBILE6 is intended to model average roadway
speeds and cannot model the emissions of vehicles at an instantaneous (cruising) speed.
6.4 Test for Model Stability
Guensler's peer review comments (Appendix E) suggested a method to test for the
stability of the overall off-cycle and speed modeling methodology. This method was to develop
an alternative model in which the off-cycle adjustment was generated using a facility cycle other
than FWYF (LOS B or C was specifically suggested), and comparing the results of this alternate
model to that used in MOBILE6 on other LOS cycles, or an independent data set. Guensler
suggested that if the alternate model was shown to predict the same results as the MOBILE6
approach over the range of LOS cycles, the model estimation approach would be shown to be
stable.
Overall, we expect that the suggested approach would yield similar results no matter what
cycle was used as the basis for the off-cycle adjustment and SCFs. This is because the off-cycle
adjustment is simply the expression of the emission increase between the LA4 and a single cycle
(FWYF in this analysis), and the SCFs are in turn the expression of the emission increase from
any average speed relative to that same cycle. Changing the base cycle will not change this
overall relationship; if the relationship between the LA4 and base cycle were changed, the SCFs
would change accordingly, resulting in similar predictions of overall emissions.
In demonstrate this, we formulated an alternate model for off-cycle adjustments and SCFs
which relied on emission results from the Freeway AC cycle, one of the highest average speed
cycles tested for this analysis (average speed of 59.7 mph). We focused on the 22 vehicles used
to derive the "Level 1" SCFs discussed in Section 5. We then used this alternate model to predict
arterial emissions at 11.6 mph (average speed of the Arterial E cycle), representing the lowest
end of the speed range and relatively high SCFs. The hypothesis was that if stability were shown
at the low end of the speed range where SCFs are the highest, it could be inferred on the
remainder of the speed curve.
The base and alternative models are shown in Table 6-1, along with the prediction of
mean emissions for arterials at 11.6 MPH. The means show less than 3 percent difference
between the two models, and 95 percent confidence intervals confirm this difference is not
significant. This analysis suggests that the modeling approach used to estimate off-cycle
emissions and SCFs would result in stable emissions predictions regardless of which cycle is
used as the base cycle.
Final M6.SPD.002 26 June 2001
-------
Table 6-1: Test For Model Stability for Arterial at 11.6 mph
Pollutant
NOx
HC
CO
Model
Approach
MOBILE6
Alternate
MOBILE6
Alternate
MOBILE6
Alternate
Model Formulation
(LA4 + Off-Cycle Adjustment) *SCF
[LA4NOx+(0.332*LA4NOx-0.04745*LA4NOx2)]*(1.71)
[LA4NOx+(2.05*LA4NOx-4.42*LA4NOx2)]*(1.71/1.42)
[LA4HC+(0.305*LA4HC-0.02492*LA4HC2)]*(1.86)
[LA4HC+(0.775*LA4HC)]*(1. 86/1.24)
[LA4CO+(0.984*LA4CO-0.07638*LA4CO2)]*(1.34)
[LA4CO+(3.417*LA4CO-0.823*LA4CO2)]*(1.34/1.59)
Predicted Level 1
emissions (g/mi) ą
95% CI
0.43 ą0.13
0.42 ą0.09
0.093 ą0.023
0.095 ą0.023
2.23 ą0.91
2.26 ą0.7
7.0 BENEFITS OF THE SFTP REQUIREMENT
7.1 Methodology used to calculate SFTP benefit
Increasing attention to the importance of off-cycle emissions led to the development of a
new compliance procedure, known as the Supplemental Federal Test Procedure (SFTP). In
addition to "off-cycle" emissions, the SFTP addresses emissions which are generated with the air
conditioning on, which were also inadequately represented by the FTP. The SFTP requirements
grew out of the 1990 Clean Air Act Amendments, which instructed EPA to review the existing
procedures and revise them in whatever ways were necessary to make them more representative
of actual in-use conditions. Developed in conjunction with the California Air Resources Board
(ARB) and auto manufacturers, the SFTP requirement adds two additional certification cycles,
and tailpipe standards associated with those cycles, to impose control of off-cycle (US06 cycle)
and air conditioning emissions (SC03 cycle). The US06 is run with the vehicle in the hot
stabilized condition; that is, with the vehicle fully warmed up to insure that the engine and
catalytic converter have reached typical operating temperatures. The SC03 follows a 10-minute
soak and is run with vehicle air conditioning (A/C) in operation or with an appropriate simulation
of air-conditioning operation.
The assigned benefits of the SFTP rule will depend on whether a vehicle is a Tier 1
vehicle or a LEV. EPA and ARB promulgated separate requirements applying to these standard
levels, and hence the benefits resulting from the rule must take into account the relative
stringency of the EPA and ARB rules. Under NLEV, the Tier 1 rule will only apply to LDTs
above 6000 pounds (LDT3s and LDT4s), which phase in to the SFTP requirement at 40 percent
in 2002, 80 percent in 2003, and 100 percent in 2004.14 These trucks will be allowed to certify
to the Tier 1 SFTP standards until they begin phasing into the Tier 2 final standards in 2008, at
which point they will be required to comply with the SFTP provisions under the Tier 2 rule
discussed below.
Final M6.SPD.002
27
June 2001
-------
For Tier 1 and interim Tier 2 LDT3s and LDT4s, the benefits derived in EPA's SFTP
final rulemaking shown in Table 7-1 will be used directly in MOBILE6 (Post-SFTP CO air
conditioning emissions are a special case, as discussed below). The percent reductions shown for
the SFTP rule will be applied directly to the off-cycle adjustment to generate final off-cycle
adjustments for SFTP-compliant vehicles.
TABLE 7-1
TIER 1 SFTP BENEFITS FROM
OFF-CYCLE OPERATION and AIR CONDITIONING
POLLUTANT
HC
CO
NOx
OFF-CYCLE
88%
72%
78%
AIR CONDITIONING
100%
Fuel Consumption Increase
50%
*EPA rule estimated benefits of Tier 1 SFTP standards, in terms of percent reduction of uncontrolled "excess"
emissions.
A detailed derivation of these benefits are contained in the SFTP final rulemaking.15
They were derived by comparing the emission results over off-cycle driving from a sample of
light-duty vehicles and trucks tested in an uncontrolled condition (pre-SFTP), and with emission
control software modifications made by vehicle manufacturers to reduce off-cycle emissions (e.g.
eliminating commanded enrichment). This approach is consistent with that suggested by the
Guensler comments, which suggests "comparing the percentage reduction in emissions that will
occur for current vehicles as they move from their current emissions levels on the composite
SFTP to the compliance emission rates...". Because vehicles complying with the SFTP are just
starting to enter the market, an assessment of SFTP benefit on the in-use fleet is not yet possible.
We therefore consider the approach used in the EPA SFTP rule to be the best available.
Under NLEV, the ARE rule will apply to LEV LDVs and LDTs under 6,000 pounds
(LDT1/2). The ARB rule contains NOx and HC certification standards which differ from EPA's
both in terms of the relative stringency over the US06 and SC03 cycles, and the mileage at which
a vehicle is required to show compliance. The percent reductions derived for EPA's Tier 1
ruletherefore cannot be applied directly to vehicles complying with the ARB standards.
A sample of vehicles with emissions below the Tier 1 SFTP standards, based on
compliance strategies developed by the auto manufacturers, were available to develop the
benefits of EPA's Tier 1 rule. A comparable sample was not available in which to derive
benefits from the ARB standards under NLEV. For the purpose of MOBILE6, we therefore
developed a methodology which estimated the percent reductions for the ARB standards on
LEVs based on the EPA Tier 1 benefits presented in Table 1. This methodology required an
assessment of the relative stringency of the EPA and ARB SFTP standards compared to their
Final M6.SPD.002 28 June 2001
-------
respective FTP standard. Several factors added complication to this analysis: first, the ARB
standards are applicable at 4,000 miles whereas the EPA standards are applicable at 50,000 miles
and full useful life (100/120K miles); second, the SFTP standards are expressed at NMHC+NOx,
while MOBILE treats these pollutants separately. Third, the SFTP standards are based on
operation when the vehicle is warmed-up, necessitating that the warmed-up component of the
FTP be extracted in order to performing comparisons with the SFTP standards. An analytical
step was required to address each of these factors.
Reductions in off-cycle emissions due to ARB's LEV SFTP standards for NOx and HC
were estimated through a determination of the stringency of the ARB and EPA US06 standards.
The stringency of the ARB and EPA standards is characterized by how well they control off-
cycle emissions for LEVs and Tier 1 vehicles, respectively; this stringency is thus best
determined through a comparison. A direct comparison between these standards This
comparison was made in relation to emissions over the FTP The basis for this determination
was a comparison between the US06/SC03 standards and an estimation of "running certification
levels" (i.e. the running component of FTP certification levels) calculated for Tier 1 vehicles and
LEVs, according to the following steps, shown in Tables 7-2 and 7-3:
1) Average certification emissions for model year 1999 LDVs and LDTs were generated
from EPA's CFEIS database at 4,000 miles for LEVs and 50,000 miles for Tier 1 (Row
1). The certification database used to generate these averages are provided with this
report.
2) "Running certification levels" were estimated for Tier 1 and LEV by multiplying the
certification levels from Step 2 by the appropriate running BER fractions discussed in
Draft Final MOBILE6 Report M6.EXH.007 (December 1999); 0.90 for NOx and 0.23 for
HC. The FTP certification levels and the derived "running certification levels" are shown
in Row 2.
3) NMHC+NOx US06 and SC03 standards were split into separate NMHC and NOx
standards by applying a split of 0.14/0.86 for NMHC/NOx, derived from the development
of EPA's Tier 1 standards, and discussed in EPA's final SFTP rule (Rows 3 and 4).
4) A ratio of the resulting 50,000 mile SFTP NMHC and NOx "standards" from Step 3
and the running certification levels from Step 2 were calculated for both the Tier 1 (EPA)
and LEV(ARB) requirements for US06 (Row 5). The ratio (R) represents the magnitude
of increase allowed between the FTP and US06 cycles, and hence represents the
stringency of the SFTP standard relative to the FTP standards.
Final M6.SPD.002 29 June 2001
-------
5) The stringency of the ARB standards relative to the EPA standards were estimated by
comparing the value of R calculated in Step 4, according to the following equation (Row
6):
Additional Stringency of ARB Standards (%) = [(REPA -1) - (RARE-!)] / (REPA -1)
The additional stringency represents the additional off-cycle emissions which would be
eliminated above and beyond the reductions under the Tier 1 standards.
6) Benefits under the ARB rule were then derived by adjusting the Tier 1 benefits (Row 7)
from Table 7-1 according to the additional stringency contained in Step 5, according to the
following equation (Row 8):
ARB Benefit (%) = EPA Benefit + (Step 5) * (1 - EPA Benefit)
These steps are illustrated in Tables 7-2 and 7-3.
Final M6.SPD.002 30 June 2001
-------
Table 7-2: Worksheet for Developing LEV NOx Benefits
(1) Average FTP Certification Level
(2) Estimated "Running" Certification Level
(3) Estimated NOx 4K Standard (ARE)
(4) Estimated NOx 50K Standard (EPA)
(5) US06 Standard / Running Certification Level
(6) Additional Stringency of ARE Standard
(7) EPA SFTP Benefit (%)
(8) ARE Benefit (%)
Tier 1 (50K Miles)
L^ LDT2 LDT3 LDT4
0.17 0.19 0.24 0.30
0.15 0.17 0.21 0.27
0.50 0.78 0.78 1.15
3.26 4.57 3.69 4.19
78 78 78 78
LEV (4K Miles)
L^V/ LDT2 LDT3 LDT
0.07 0.11 0.13 0.16
0.06 0.10 0.11 0.14
0.12 0.22 0.34 0.52
1.87 2.15 3.03 3.68
62% 68% 25% 16%
92% 93% 83% 81%
Table 7-3: Worksheet for Developing LEV HC Benefits
(1) Average FTP Certification Level
(2) Estimated "Running" Certification Level
(3) Estimated NOx 4K Standard (ARE)
(4) Estimated NOx 50K Standard (EPA)
(5) US06 Standard / Running Certification Level
(6) Additional Stringency of ARE Standard:
(7) EPA SFTP Benefit (%)
(8) ARE Benefit (%)
Tier 1 (50K Miles)
LD,^7 LDT2 LDT3
0.10 0.10 0.10
0.02 0.02 0.02
0.08 0.13 0.13
3.62 5.34 5.50
88 88 88
LDT4
0.13
0.03
0.19
6.40
88
LEV (4K Miles)
LDV/
Tl
0.03
0.01
0.02
2.44
45%
93%
LDT2
0.05
0.01
0.04
3.17
50%
94%
LDT3
0.05
0.01
0.06
4.69
18%
90%
LDT4
0.07
0.02
0.08
5.40
19%
90%
Final M6.SPD.002
31
June 2001
-------
We consider the ARB CO standards over the US06 to be functionally equivalent to the
EPA standards. We expect that these standards (9.0 g/mi for Tier 1, 8.0 g/mi for LEV) will serve
as more of a cap on excess CO emissions, unlike the NOx and HC standards. As such, we are
proposing to apply the Tier 1 benefit (78 percent) to LEVs as well.
The resulting SFTP benefits for LEVs are presented in Tables 7-4.
TABLE 7-4
LEV SFTP BENEFITS OVER OFF-CYCLE OPERATION
HC
CO
NOx
LDV/T1
93%
78%
92%
LDT2
94%
78%
93%
LDT3
90%
78%
83%
LDT4
90%
78%
81%
7.2 Applying the SFTP Benefit in MOBILE6
The effect of the SFTP rule will be modeled in MOBILE6 by applying the percent
reductions derived in Section 7.1 to the off-cycle adjustment (OC) for freeways and arterial
roadways, and to the correction factor (CF) for freeway ramps. In equation form, this is
represented as follows:
Freeway Ramps: Adjusted BER w/ SFTP = BER * CF * (1-SFTP)_
Where:
BER = Basic Emission Rate (running emissions for the LA4 cycle).
CF = Multiplicative correction factor
SFTP = SFTP benefit from Section 7.1
Freeways and Arterials: Adjusted BER w/ SFTP = (BER + (OC*(1-SFTP)) * SCF
Where:
BER = Basic Emission Rate (running emissions for the LA4 cycle).
OC = Off-Cycle Emissions Offset (a function of BER emissions)
SCF = Multiplicative Speed Correction Factor (a function of speed and emissions)
SFTP = SFTP benefit from Section 7.1
This approach was generated based on the following assumptions about how "excess"
emissions are generated, defined as emissions higher than the running LA4 in terms of
grams/mile:
Final M6.SPD.002
32
June 2001
-------
"excess" emissions over the local roadway cycle are due solely to reduced travel
efficiency (e.g. less distance traveled per unit of engine work), rather than off-cycle
driving, and hence emissions will not be reduced from the SFTP.
"excess" emissions over the freeway ramp cycle are due solely to off-cycle driving, and
hence "excess" emissions will be reduced in direct proportion to the SFTP benefits;
"excess" emissions over the freeway and arterial cycles are due to both off-cycle
emissions and reduced driving efficiency, and hence emissions will be reduced only over
the off-cycle driving events on these cycles, in proportion to the SFTP benefits.
This latter situation for freeways and arterial is the primary reason why two separate
adjustments (Off-Cycle and SCFs) have been developed to address the overall issue of driving
behavior. Peer review comments from Guensler suggested that two separate adjustments were
unneccessary, instead recommending that a single adjustment be developed which encompassed
both corrections. However, the specific contribution of off-cycle emissions must be estimated in
order to account for the benefit of the SFTP. A single adjustment factor would combine "excess"
emissions due to both off-cycle emissions and reduced travel efficiency, making it difficult to
estimate the benefits of off-cycle control from SFTP.
Guensler's peer review comments infer that applying SFTP reductions only to the off-
cycle adjustment translates to emission reductions not being applied on all freeway and arterial
driving due to the SFTP. This is not the case; as indicated in the Freeway and Arterial equation
above, reductions in the off-cycle adjustment will result in reductions over the entire speed range
for freeway and arterial roadways. What will vary, however, is the relative magnitude of these
reductions. The SFTP benefits reported in Section 7.1 will only be realized in full over the
FWYF cycle, because the "excess" emissions over this cycle are assumed to be due exclusively
to off-cycle emissions. For cycles with average speeds lower than 19.6 mph, it is not appropriate
to treat all of the "excess" emissions as off-cycle emissions, since reduced travel efficiency
contributes to these emissions as well. For cycles with average speeds higher than 20 mph, it is
more reasonable to assume that excess emissions are caused predominately by off-cycle
emissions, and hence the assumption that the SFTP rule will reduce the calculated off-cycle
increment is likely conservative.
7.3 Applicability of SCFs to SFTP-compliant vehicles
The SCFs were developed on a sample of vehicles which were not certified to comply
with the SFTP. Given the advent of the SFTP rule, we felt it important to assess whether the
SCFs could reasonably be applied for vehicles which would comply with the off-cycle
requirements. We analyzed emissions results across EPA facility-specific test cycles for a subset
of test vehicles with low off-cycle emissions; our criteria for choosing these vehicles was
emission performance on the "Freeway Ramp" cycle, a short cycle which is meant to mimic
driving performance while entering a freeway. As such, this cycle as high acceleration rates,
comparable to those found on the US06 cycle. If a vehicle performs well on the Ramp cycle, we
Final M6.SPD.002 33 June 2001
-------
presume it would likewise perform well on the US06; in particular, it is likely that such a vehicle
would have adequate catalyst volume so as to not experience severe catalyst "breakthrough", an
important contributor to high off-cycle emissions.
Our criteria for choosing "clean" SFTP vehicles was whether the vehicle's emissions over
the ramp cycle were at or below EPA's implied NMHC+NOx US06 standard at 50,000 miles
(0.58 grams/mile). 20 vehicles met this criteria, and are listed in Table 7-5 along with g/mi
emissions over the Ramp cycle. CO emissions are also listed; the US06 standard for CO is 9
g/mi, which all of the vehicles meet, most by over 50 percent.
Table 7-5: "Clean SFTP" Vehicles
Vehicle
5021
5217
5240
5007
5229
5061
5063
5017
5018
5010
Ramp
NMHC+NOx/CO
(g/mi)
0.13/4.6
0.13/2.17
0.14/0.28
0.16/1.11
0.20/6.5
0.22/0.69
0.23/0.86
0.26/3.4
0.30/3.4
0.31/0.7
Vehicle
5213
5013
5062
5060
5038
5234
5231
5059
5223
5221
Ramp
NMHC+NOx/CO
(g/mi)
0.33/0.47
0.36/2.91
0.37/0.86
0.38/2.52
0.38/0.08
0.43/1.35
0.47/5.2
0.53/7.8
0.55/0.03
0.58/2.53
It is likely that most of these vehicles were included in the sample used to develop the
"Level 1" (low emission) MOBILE6 speed correction factors. Hence, the comparison is not of
two completely independent data sources. The main purpose of this exercise is to assess whether
the SCFs which were developed with pre-SFTP vehicles in mind can reasonably be applied to the
subset of vehicles we project would comply with the SFTP requirement. It is important to note
that, because of the SFTP benefits discussed in Section 4, speed-corrected emissions will be
lower for SFTP-compliance vehicles than for pre-SFTP vehicles, even if the same SCFs are
applied. At issue is whether the relative increase in emissions observed for pre-SFTP vehicles
across the speed range applies to post-SFTP vehicles as well.
Final M6.SPD.002
34
June 2001
-------
We computed average emissions from this sample for the freeway, arterial, and local
facility-specific test cycles used to develop the SCFs presented earlier in this paper. We then
divided these emissions by that of the Freeway F cycle, which is approximately the basis for
generating the MOBILE6 SCFs. In this way, we could develop SCFs specifically for the clean
SFTP vehicles, and compare them to the "Level 1" SCFs proposed for MOBILE6.
The results of this comparison are shown in Figures 9a-9c for NOx, HC and CO. As
shown by these charts, the SCFs for vehicles with low off-cycle emissions are consistent with
the MOBILE6 SCFs. From this we conclude that it is reasonable to apply the MOBILE6 SCFs to
vehicles which comply with the SFTP.
8.0 Application in MOBILE6
The speed corrections described in this document are applied to gasoline fueled, light-
duty vehicles (cars and light trucks) of all model years and technologies. The speed correction
factor would be applied to the basic exhaust hot running emission rates, adjusted to freeway
emission levels at 19.6 mph. Additional adjustment would be made to the freeway emission
estimate between 7.1 and about 30 mph to account for arterial/collector roadways. MOBILE6
would continue to use the existing speed correction factors and methodology found in MOBILES
for diesel vehicles, gasoline fueled heavy-duty vehicles and motorcycles. Heavy-duty diesel
vehicles will also be adjusted for NOx excess emissions separately from the MOBILE6 speed
correction factors.
In MOBILE6, the daily average emission rate will be calculated by VMT weighting an
emission estimate for each hour of the day. Within each hour of the day, there will be a
distribution of speeds (either a default national average or a user supplied distribution) for
freeways and arterial/collectors. The speed correction would be applied to the estimate of
Normal and High emitters within each model year separately. Older (pre-1981) model year
gasoline fueled, light-duty vehicles will have only composite (combined Normal and High) basic
exhaust emission rates. In these cases the speed correction will be applied to the composite basic
exhaust emission rates (including both Normal and High emitters). Speed correction factors will
not be applied to the effects of engine start on emissions estimated by MOBILE6.
The speed distribution in MOBILE6 will consist of average speed "bins" from 5 to 65
mph in 5 mph increments and for 2.5 mph (14 speed bins) representing the distribution of
average speeds within each hour.16 Each hour of the day will have an estimate of the distribution
of vehicle miles traveled (VMT) on freeways, ramps, local and arterial/collector roadways.
These distributions will be used to weight together the emission estimates in each speed bin to
give an hourly emission estimate. Freeway Ramps and Local Roadways will have hourly
emission estimates and VMT estimates, but will not have speed distributions. The hourly
emission estimates will be weighted by the hourly VMT distribution separately for each facility.
Finally, the VMT distribution between facilities will be used to combine the results into an area-
wide running exhaust emission estimate. Emissions due to engine start within each hour will be
calculated separately.
Final M6.SPD.002 3 5 June 2001
-------
In summary, MOBILE6 will:
o Determine the basic running exhaust emission rate (BER).
o For each hour, correct the BER for temperature and fuel effects.
o Using the corrected BER, calculate the emissions for Freeway Ramps, Local
Roadways and for the 14 speed bins for freeways and arterial/collectors using the
appropriate emission offsets and speed correction factors described in this
document.
o Using the speed distributions, weight the freeways, ramps, local and
arterial/collector speed bin results to get hourly emissions.
o Using the hourly VMT distributions, weight together the hourly facility results to
get daily emissions by facility.
o Using the facility VMT distribution, weight the daily facility emissions to get the
area-wide running exhaust emission estimate.
o Combine the running exhaust emission estimate with the engine start emissions to
get the composite exhaust emission rate.
Appendix B shows an example speed correction calculation.
The national average default factors used in MOBILE6 for VMT weighting the speed-
corrected, facility-type emissions into a single area-wide running emissions rate is described in a
report prepared for EPA by Systems Applications International.17 This report also contains the
default distributions of average speeds on each facility over the day. All of these default values
can be overridden by the user with local information using methods described in a separate
guidance document.18
The operating mode inputs used in MOBILES will not be needed for MOBILE6. Instead,
MOBILE6 uses values for the number of engine starts, the distribution of soak times between
engine starts, the mileage accumulation rates and the distribution of these factors over the day.19
These values are used to determine the weighting of the running exhaust emissions with the
effects of engine starts to calculate a composite exhaust emission factor. Although MOBILE6
contains default values, these default values will normally be overridden by user supplied local
information.
Similarly, once the composite running and engine start emissions are calculated, the
composite exhaust HC emissions can be combined with the calculated non-exhaust HC
emissions. The reader should refer to the reports regarding the non-exhaust emission estimates
and their associated activity for more details on how these values are calculated.20
8.1 Light Duty Diesel Vehicles
The speed correction factors (SCFs) for LDDV and LDDT in MOBILES were derived
from the SCFs calculated for HDDV. However, in MOBILES, an adjustment was added for
LDDV and LDDT to account for user supplied changes to the operating modes, which are cold
Final M6.SPD.002 36 June 2001
-------
start versus hot start VMT fractions applied to FTP bag emission rates, to account for the
different average speeds between the FTP bags. These fractions did not affect HDDV, since the
emissions from HDDV are not calculated from individual FTP bag emission results.
In MOBILE6, there will not be user inputs for operating modes. The emission estimates
for LDDV and LDDT in MOBILE6 have been split into the portion that occurs as a result of
engine starts separately from the hot running emissions. The intention of the operating mode
adjustment to the SCFs for LDDV and LDDT in MOBILES was to account for user supplied
changes in the mix of FTP bag results in the basic exhaust emission rates. However, the basic
exhaust emission rate for LDDV and LDDT in MOBILE6 is not affected by user inputs and is
based solely on the LA4 driving cycle, which is the basis for the full FTP. As a result, the
operating mode adjustment to the SCFs for LDDV and LDDT in MOBILE6 will be set to a
constant based on the standard FTP operating mode mix. The adjustment to account for changes
in speed that were caused by the operating mode will no longer be necessary in MOBILE6.
The MOBILES speed correction factor coefficients for LDDV and LDDT from
MOBILES21 are shown in Table 18. These factors will be used in MOBILE6 as well, but with
the SADJ value fixed at 19.6 mph, which is the speed of the basic emission rate cycle, the LA4.
8.2 Heavy Duty Vehicles
The speed correction factors for heavy duty vehicles are not changed from those used in
MOBILES.22 The coefficients used in MOBILE6 are shown in Table 19 and Table 20.
8.3 Motorcycles
The speed correction factors for motorcycles are not changed from those used in
MOBILES.23 However, the SADJ factor discussed in the light duty diesel vehicle section were
also applied to motorcycle speed correction factors in MOBILES. In MOBILE6, the SADJ factor
will be set to 19.6 mph. The coefficients used in MOBILE6 are shown in Table 21.
The speed correction factors for motorcycles were developed in earlier versions of the
model only for speeds up to 55 mph. In MOBILES, for average speeds above 55 mph for THC
and CO and 48 mph for NOx, the speed correction factor used for motorcycles was derived from
the speed correction factor used for light duty gasoline vehicles for 65 mph. The speed
correction factor (SCF) calculated for motorcycles is adjusted using the ratio of the difference
between the light duty vehicle SCF at 65 mph and the motorcycle SCF at the target speed divided
by the difference between 65 mph and the target speed.
RATIO = ( LDGVSCF(65) - MCSCF(s) ) / (65 - SPD )
MCSCF(s) = MCSCF(s) * (1.0 + ( RATIO*(s - SPD) ) )
Where : s = average speed (mph) target speed
Final M6.SPD.002 37 June 2001
-------
LDGVSCF(65) = light duty gasoline vehicle SCF at 65 mph
MCSCF(s) = motorcycle SCF at the target speed (s)
SPD = average speed where adjustment begins in mph
- 55 mph for THC and CO emissions
- 48 mph for NOx emissions
This adjustment was retained for the calculation of motorcycle speed correction factors in
MOBILE6.
8.4 High Speeds
The driving cycles developed for MOBILE6 were derived from data collected before
national speed limits were increased from 55 mph to 70 mph. The average speed of driving for
uncongested freeways in the data sample was 59.7 mph (73.1 mph maximum). Another driving
cycle, developed using a subset of vehicles driving over 55 mph, has an average speed of 63.2
mph. It is clear that using the existing driving behavior information collected before the increase
in the national speed limit cannot provide a credible estimate of the driving behavior at average
speeds above about 65 mph. Existing research24 has shown that even minor variations in driving
at high speeds can have a significant effect on driving emissions at those speeds. For this reason,
until new information about driving behavior at high speeds and their effect on emissions is
available, MOBILE6 will not directly predict emission impacts of average speeds over 65 mph.
9.0 COMPARISON TO MOBILES
Figures 6a, 6b, and 6c show the MOBILE6 speed correction factors (SCFs) for freeways
compared to selected speed correction factors used in MOBILES. This comparison cannot be
made clearly, since the two versions of the model use very different approaches.
The MOBILE6 SCFs depend on emission level and the MOBILES SCFs do not.
The MOBILES SCFs are applied to a composite exhaust emission rate, including
engine start emissions. The MOBILE6 SCFs will only be applied to the hot
running exhaust emissions, before the effects of engine start are added.
The MOBILE6 SCFs are intended to estimate the effects on freeways excluding
ramp activity, but the MOBILES SCFs are a composite of all roadway types.
The MOBILE6 SCFs include the effect of additional aggressive driving effects on
emissions missing from the MOBILES SCFs.
The overall shape of the MOBILES and MOBILE6 SCFs is similar. The MOBILE6
SCFs are flatter at speeds greater than 55 mph than in MOBILES, especially for CO and NOx.
This may be due largely to the fact that the old speed cycles above 48 mph all started from idle
(zero mph) and accelerated to a speed higher than the average speed of the cycle. This extra
acceleration, which is not generally found on cruising vehicles on limited access freeways, adds
to the power demand, therefore likely increasing emissions in the old high speed cycles relative
Final M6.SPD.002 3 8 June 2001
-------
to lower speed cycles. The acceleration to reach freeway speeds is now contained in the separate
ramp cycle. This additional ramp cycle will allow this effect to be weighted appropriately with
freeway driving. The effect from starting and ending at idle is less pronounced in the lower
speed cycles since they inherently have a higher percentage of driving at idle.
In Figure 6a (THC), the MOBILE6 SCF for the lowest emission level (based Tier 1
vehicles) has a positive slope beyond about 30 mph, indicating increasing THC emissions with
increasing average speed. However, as shown in Figure 4a, Tier 1 vehicles are much cleaner at
all speeds than the normal emitting Tier 0 vehicles. The shape of the THC MOBILE6 SCFs for
the higher emission levels (based on Tier 0 vehicles, Normal and High) is very close to the shape
of MOBILES SCFs.
For higher average speeds (above 19.6 mph) the MOBILE6 SCFs for CO emissions
(Figure 6b) have a strongly positive slope at lower emission levels (based on Tier 0 Normals and
Tier 1 vehicles). This is very different from the SCFs used in MOBILES. The MOBILE6 SCFs
for THC/NMHC emissions for Tier 0 Normal vehicles have a negative slope. However CO
emissions are more sensitive to aggressive driving than THC/NMHC emissions, which may
explain the difference in the trends.
The MOBILE6 SCFs for NOx emissions (Figure 6c) for the higher emission levels (based
on Tier 0 vehicles) have a slight upward trend at higher speeds, similar to the MOBILES trends.
The lowest emission level SCFs (based on Tier 1 vehicles) has a steep slope, similar to the oldest
MOBILES SCF. All of the MOBILE6 SCFs tend to rise as average speeds decrease, which is
expected with more accelerations and decelerations (stop and go driving) present in the driving
patterns. However, the MOBILE6 SCFs rise much more steeply and to higher levels than the
MOBILES SCFs.
Similar graphs comparing MOBILE6 speed correction factors for arterial/collector
roadways and freeways with MOBILES speed correction estimates are shown in Figures 7a
through 7i. Speed correction factors for arterial/collector roadways and freeways are the same
below 7.1 mph and above about 30 mph. In general, speed correction factors for arterial/collector
roadways are higher between those speeds.
Since the Freeway Ramp and Local Roadway emissions are estimated directly from the
basic exhaust emission rate (based on the hot running LA4 emissions), they cannot be compared
to the speed correction factor used in MOBILES.
Final M6.SPD.002 39 June 2001
-------
References
1 See the EPA web site http://www.epa.gov/otaq/sftp.htm for more information.
2 Federal Test Procedure, or FTP means the test procedure as described in the Combined Federal
Register § 86.130-00(a) through (d) and (f) which is designed to measure urban driving tail pipe
exhaust emissions and evaporative emissions over the Urban Dynamometer Driving Schedule as
described in Appendix I to this part. (See also http://www.epa.gov/otaq/labmthod.htm.)
3 See the EPA web site http://www.epa.gov/otaq/sftp.htm for more information.
4 Carlson, T.R., et al., "Development of Speed Correction Cycles," MOBILE6 Stakeholder
Review Document (M6.SPD.001). Prepared for EPA by Sierra Research, Inc., 1997.
(http://www.epa.gov/otaq/m6.htm)
5 See the California Air Resources Board web site, http://www.arb.ca.gov/msei/msei.htm.
6 Gammariello, R.T. and Long, J.R., "Development of Unified Correction Cycles," California Air
Resources Board, Sixth CRC On-Road Vehicle Emissions Workshop, March 1996. This report
is available at the CARB web site (http://www.arb.ca.gov/msei/pubs/ucc_crc5.pdf).
7 ibid
8 ibid
9 ibid
10 See the EPA web site, http://www.epa.gov/otaq/regs/ld-hwy/tier-l/ for more information.
11 Brzezinski, D.J., et al., "Coefficients for the Determination of Engine Start and Running
Emissions From FTP Bag Emissions," MOBILE6 Stakeholder Review Document (M6.STE.002),
1997. (http://www.epa.gov/otaq/m6.htm)
12 "Federal Test Procedure Review Project: Preliminary Technical Report," EPA Office of
Mobile Sources Certification Division, (EPA420-R-93-007), May 1993. This report is available
at the EPA web site (http://www.epa.gov/otaq/sftp.htm).
13 ibid
14 "Final Regulations for Revisions to the Federal Test Procedure for Emissions From Motor
Vehicles," Federal Register, 40 CFR Part 86 Volume 61, Number 205, Oct 22, 1996, Page 54869
(http ://www. epa. gov/otaq/ld-hwy. htm).
15 "Regulatory Impact Analysis (Final Rule) : Federal Test Procedure Revisions,"
EPA Office of Mobile Sources, August 15, 1996 (http://www.epa.gov/otaq/ld-hwy.htm).
16 M6.SPD.003, opcit
Final M6.SPD.002 40 June 2001
-------
17 "Development of Methodology for Estimating VMT Weighting by Facility Type," MOBILE6
Stakeholder Review Document (M6.SPD.003) EPA420-P-99-006. Prepared for EPA by Systems
Applications International, Inc., 1998. (http://www.epa.gov/otaq/m6.htm)
18 "Guidance for the Development of Facility Type VMT and Speed Distributions," MOBILE6
Stakeholder Review Document (M6.SPD.004) EPA420-P-99-004. Prepared for EPA by Systems
Applications International, Inc., 1998. (http://www.epa.gov/otaq/m6.htm)
19 Glover, E.L., et al., "Soak Length Activity Factors for Start Emissions," MOBILE6
Stakeholder Review Document (M6.FLT.003). (http://www.epa.gov/otaq/m6.htm)
20 See the EPA web site for MOBILE6 reports, http://www.epa.gov/otaq/m6.htm
21 "AP-42 Volume n, Compilation of Air Pollution Emission Factors, Mobile Sources,"
Appendix H, June 30, 1995. (http://www.epa.gov/otaq/models.htm)
22 ibid
23
ibid
24 Earth, M., et al., "Estimating Emissions and Fuel Consumption for Different Levels of
Freeway Congestion," University of California, Riverside. 78th Annual Transportation Research
Board Meeting, Washington, D.C., January 1999. This report is available from TRB
(http ://nationalacademies. org/trb/bookstore).
Final M6.SPD.002 41 June 2001
-------
Tables
Final M6.SPD.002 42 June 2001
-------
Table 1
New Facility-Specific/Area-Wide Speed Correction Cycle
Statistics
Cycle*
Freeway, High Speed
Freeway, LOS A-C
Freeway, LOS D
Freeway, LOS E
Freeway, LOS F
Freeway, LOS "G"
Freeway Ramps
Arterial/Collectors
LOS A-B
Arterial/Collectors
LOS C-D
Arterial/Collectors
LOS E-F
Local Roadways
Non-Freeway Area-
Wide Urban Travel
Average
Speed
(mph)
63.2
59.7
52.9
30.5
18.6
13.1
34.6
24.8
19.2
11.6
12.9
19.4
Maximum
Speed
(mph)
74.7
73.1
70.6
63.0
49.9
35.7
60.2
58.9
49.5
39.9
38.3
52.3
Maximum
Accel Rate
(mph/s)
2.7
3.4
2.3
5.3
6.9
3.8
5.7
5.0
5.7
5.8
3.7
6.4
Length
(seconds)
610
516
406
456
442
390
266
737
629
504
525
1,348
Length
(miles)
10.72
8.55
5.96
3.86
2.29
1.42
2.56
5.07
3.36
1.62
1.87
7.25
* LOS (level of service) refers to roadway congestion categories. See Section 4.6.
Final M6.SPD.002
43
June 2001
-------
Table 2
Comparison of Key Statistics
For Facility-Specific Cycle Schedules
Versus Total Vehicle Observations
Driving Cycle
Freeway High-Speed
Freeway LOS A-C
Freeway LOS D
Freeway LOS E
Freeway LOS F
Freeway LOS G
Freeway Ramp
Arterial LOS A-B
Arterial LOS C-D
Arterial LOS E-F
Local Roadways
Unified Cycle
Mean
Speed
(mph)
Cyc.
63.2
59.7
52.9
30.5
18.6
13.1
34.6
24.8
19.2
11.6
12.8
24.6
Obs.
62.7
59.2
52.0
32.1
19.9
14.4
35.4
25.2
18.9
12.0
14.6
26.3
Maximum
Speed
(mph)
Cyc.
74.7
73.1
70.6
63.0
49.9
35.7
60.2
58.9
49.5
39.9
38.3
67.2
Obs.
80.9
83.2
75.8
71.3
69.5
49.1
79.1
74.9
71.3
56.8
62.7
80.3
Maximum
Accel Rate
(mph/sec)
Cyc.
2.7
3.4
2.3
5.3
6.9
3.8
5.7
5.0
5.7
5.8
3.7
6.9
Obs.
5.8
6.8
6.1
8.5
9.6
5.7
9.3
14.9
10.4
10.2
12.5
10.4
Total
SAFD
Difference
* (%)
9.41
12.12
15.10
25.17
23.83
18.80
42.74
17.04
16.86
17.86
21.80
30.27
High-Power
Difference**
(%)
0.16
0.39
0.35
0.18
0.06
0.10
0.99
0.40
0.21
0.24
0.11
0.19
* The SAFD is the speed/acceleration frequency distribution based on time at each speed. Total SAFD Difference
is the sum of the differences between the final cycle distribution and the target population distribution from which
the cycle micro trips are chosen. (See M6.SPD.001)
** Specific power was calculated from the following equation:
P:
0
,ifSt>S,,
,ifSt
-------
Table 3
Statistics for Additional Tested Cycles
Cycle
LA4
(Urban Dynamometer
Driving Cycle)
Running 505
(First 505 seconds of the
Urban Dynamometer
Driving Cycle)
Unified Cycle (LA92)
ST01
(Engine Start Cycle)
New York City Cycle
(NYCC)
Average
Speed
(mph)
19.6
25.6
24.6
20.2
7.1
Maximum
Speed
(mph)
56.7
56.7
67.2
41.0
27.7
Maximum
Accel Rate
(mph/s)
3.3
3.3
6.9
5.1
6.0
Length
(seconds)
1368
505
1435
248
600
Length
(miles)
7.45
3.59
9.81
1.39
1.18
Table 4
Distribution of the Vehicle Sample
By Emission Standard and Technology
Fuel Delivery
Carburetor
Throttle Body FI
Multi-Port FI
Total Sample
TIER 0 Emission
Standard
7
27
39
73
TIER 1 Emission
Standard
1
11
12
Total
Sample
7
28
50
85
Final M6.SPD.002
45
June 2001
-------
Table 5
Vehicle Sample Description
Site
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
Veh.
No.
5001
5002
5003
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
Veh.
Class
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDT1
LDV
LDV
LDV
LDT1
LDV
LDV
LDV
LDT1
LDT1
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDT1
LDV
LDT1
VIN
1G4AH51R7J6401871
1G3NL54UXKM283722
2FACP74F3MX162914
1G1JC14GOM7126454
1G1JC111XK7150483
1G3HY54C9JW3 12653
1FACP57U5NG145893
1G2WP14T6KF307905
4T1SK12E9PU184046
2C1MS2468P6704533
1G2NV54D9JC821314
1G2NE5434PC795009
1G6CD53B7M4272204
1G2NE5438PC758996
1G4HR54C5KH488839
WVWEB5159MK012875
1B3ES27C9SD221573
1G1FP23TXLL1 11092
1FACP5245NG196687
1B3ES67C2SD188892
1G1JC1112KJ207455
1FAPP36X6JK249611
2FAPP36X8MB1 16542
1N6SD16S6MC351945
1MEBM50U3KG663746
JE3CU14A1NU003588
1YVGE22A8P5138202
1P4FH4430KX568849
1FABP29D9GA165884
JT2SV24E8J3 189405
1MEBP923XFA603099
1GCBS14E3H2170996
1GCBS14A3F2156946
1FABP37X6HK239681
JN1HM05S8HX081093
1P3BP49CXDF305484
2G1WL52M2T9212643
1HGED3554JA017137
2HGED6359KH534893
JT2EL32G3H0076681
1GCDM15NXFB180388
2HGED6349KH537915
1GTBS14E5J2520442
Mod.
Yr.
88
89
91
91
89
88
92
89
93
93
88
93
91
93
89
91
95
90
92
95
89
88
91
91
89
92
93
89
86
88
85
87
85
87
87
83
96
88
89
87
85
89
88
Make
BUICK
OLDSMOBILE
FORD
CHEVROLET
CHEVROLET
OLDSMOBILE
FORD
PONTIAC
TOYOTA
GEO
PONTIAC
PONTIAC
CADILLAC
PONTIAC
BUICK
vw
DODGE
CHEVROLET
FORD
DODGE
CHEVROLET
FORD
FORD
NISSAN
MERCURY
EAGLE
MAZDA
PLYMOUTH
FORD
TOYOTA
MERCURY
CHEVROLET
CHEVROLET
FORD
NISSAN
PLYMOUTH
CHEVROLET
HONDA
HONDA
TOYOTA
CHEVROLET
HONDA
CHEVROLET
Mod.
CENT
CUTL
CROW
CAVA
CAVA
DELT
TAUR
GRAN
CAMR
METR
GRAN
GRAN
SEDA
GRAN
LESA
CABR
NEON
CAMA
TAUR
NEON
CAVA
TEMP
TEMP
HARD
SABL
SUMM
626
VOYA
TAUR
CAMR
COUG
S10
S10
TEMP
STAN
RELI
LUMI
CIVI
CIVI
TERC
ASTR
CIVI
S15
Std.
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
Tier 1
TierO
TierO
Tierl
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
Tierl
TierO
TierO
TierO
TierO
TierO
TierO
Miles
129,698
61,956
53,003
54,658
107,611
101,534
74,078
155,181
29,392
105,445
89,764
72,348
51,707
58,538
65,212
67,496
20,855
71,258
84,148
28,525
110,929
107,979
97,522
103,346
107,075
129,457
103,171
118,586
50,755
197,090
113,584
128,681
89,435
118,148
58,173
94,399
16,557
184,457
161,598
136,654
179,855
122,821
115.693
Eng.
Size
2.5
2.5
5.0
2.2
2.0
3.8
3.0
3.1
2.2
1.0
2.3
2.3
4.9
2.3
3.8
1.8
2.0
3.1
3.8
2.0
2.0
2.3
2.3
2.4
3.0
1.5
12
3.0
2.5
2.0
14
2.5
1.9
2.5
2.0
2.2
3.1
1.5
12
1.5
4.3
1.5
2.5
Fuel
Inj.
TBI
TBI
PFI
TBI
TBI
PFI
PFI
PFI
TBI
TBI
TBI
PFI
TBI
PFI
TBI
TBI
TBI
PFI
TBI
PFI
TBI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
TBI
PFI
TBI
TBI
NO
TBI
PFI
NO
PFI
TBI
TBI
NO
NO
TBI
TBI
IM240
PASS
FAIL
PASS
PASS
PASS
PASS
PASS
FAIL
PASS
PASS
PASS
PASS
FAIL
PASS
FAIL
PASS
PASS
FAIL
PASS
PASS
PASS
PASS
FAIL
PASS
FAIL
FAIL
FAIL
PASS
FAIL
FAIL
FAIL
PASS
PASS
FAIL
PASS
FAIL
PASS
FAIL
PASS
PASS
FAIL
PASS
FAIL
Final M6.SPD.002
46
June 2001
-------
Table 5
Vehicle Sample Description
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5213
5217
5218
5219
5220
5221
5222
5223
5224
5225
5227
5228
5229
5230
5231
5232
5233
5234
5235
5237
5239
5240
5941
LDV
LDT1
LDT1
LDT1
LDV
LDT1
LDV
LDV
LDT2
LDV
LDT1
LDT1
LDV
LDT1
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDT2
LDT2
LDV
LDV
LDV
LDT1
LDT2
LDV
LDV
LDT1
LDV
LDV
LDT1
LDV
LDT1
LDV
LDT1
LDV
T DV
1G2WH54T6PF250844
1FTCR1056FUD20466
1FTDE14N8MHB05052
1FTCR10A2KUB93426
2G1AW19X5G1258479
1FDDE14F9FHA59240
1G1JF11W1K7156403
1G1JC14GXM7146551
1FDEE14NOMHB15171
1FAPP1282MW3 14230
2P4GH25K6MR240965
1GNDM15Z4MB190115
1G1LT53T9PY237873
1GCCS19Z5P0178401
4T1SK11E4PU252562
1HGCB7658PA075439
JN1HJ01POLT397615
JE3CA11A7PU098450
1G2WJ52M7TF204255
JT2AE94A5N0273089
1HGCD5632TA260884
1G8ZF5498NZ175489
1G1LW13T4NY109988
1FTEF14N3RLB27661
1FTEF1549TLB25543
JM1BG2263N0464490
2G1WL52M2T9212643
1G1JC5447N71 16728
1FTCR10A9TPB08548
1GNEV16K9LF1 16974
2C3ED56F7RH211101
1HGEJ8142TL073569
1GNDM19WXRB229457
1G8ZK5570RZ145840
KMHJF22M5RU669848
1GNDU06D3NT 126706
1FARP15J9RW262996
2P4FH5532LR534285
2G1WN54X7N91 17726
1GMDU06LXRT234029
4T1BF12K3TU871236
1B3XC56R3T D749334
93
85
91
89
86
85
89
91
91
91
91
91
93
93
93
93
90
93
96
92
96
92
92
94
96
92
96
92
96
90
94
96
94
94
94
92
94
90
92
94
96
90
PONTIAC
FORD
FORD
FORD
CHEVROLET
FORD
CHEVROLET
CHEVROLET
FORD
FORD
PLYMOUTH
CHEVROLET
CHEVROLET
CHEVROLET
TOYOTA
HONDA
NISSAN
EAGLE
PONTIAC
TOYOTA
HONDA
SATURN
CHEVROLET
FORD
FORD
MAZDA
CHEVROLET
CHEVROLET
FORD
CHEVROLET
CHRYSLER
HONDA
CHEVROLET
SATURN
HYUNDAI
CHEVROLET
FORD
PLYMOUTH
CHEVROLET
PONTIAC
TOYOTA
DODGF
GRAN
RANG
ECON
RANG
CELE
ECON
CAVA
CAVA
E150
ESCO
VOYA
ASTR
CORS
S10
CAMR
ACCO
MAXI
SUMM
GRAN
CORO
ACCO
SL
BERR
F150
F150
PROT
LUMI
CAVA
RANG
SURE
LHS
CIVI
ASTR
SL
ELAN
LUMI
ESCO
VOYA
LUMI
TRAN
CAMR
DYNA
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
TierO
Tier 1
TierO
Tier 1
TierO
TierO
TierO
Tier 1
TierO
Tier 1
TierO
Tier 1
TierO
TierO
Tier 1
TierO
TierO
TierO
TierO
Tier 1
TierO
TierO
Tier 1
Tier 1
TierO
85,789
56,488
79,573
123,419
131,601
86,203
123,581
90,945
97,531
105,861
72,032
90,880
41,766
48,578
67,344
61,163
120,786
52,447
20,451
77,310
7,573
89,995
94,316
97,629
12,877
10,727
17,233
90,196
10,064
97,658
59,937
9,433
77,178
25,930
57,960
33,872
51,168
98,530
16,133
68,305
18,992
6 813
3.4
2.8
5.8
2.3
2.8
5.8
3.1
2.2
5.8
1.8
2.5
4.3
3.4
4.3
2.2
2.2
3.0
1.5
3.1
1.6
2.2
1.9
3.1
5.8
4.9
1.8
3.1
2.2
2.3
5.7
3.5
1.6
4.3
1.9
1.8
3.1
1.9
3.0
3.4
3.8
3.0
3 3
PFI
NO
PFI
TBI
NO
NO
PFI
TBI
PFI
PFI
TBI
TBI
PFI
TBI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
TBI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
TBI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFT
FAIL
FAIL
FAIL
PASS
PASS
PASS
PASS
PASS
FAIL
FAIL
PASS
PASS
PASS
PASS
PASS
PASS
PASS
PASS
PASS
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NUT T
Final M6.SPD.002
47
June 2001
-------
Table 6
Distribution of Vehicle Sample
By Vehicle Class and Model Year
Model Year
1983
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
TOTAL
Passenger
Car
1
1
2
3
6
9
3
7
9
10
4
2
6
63
Light-Duty Truck
(0-6000 GVW)
4
1
1
2
1
4
1
1
2
1
18
Light-Duty Truck
(6000-8500 GVW)
1
1
1
1
4
Total
1
5
2
4
7
11
5
12
10
11
7
2
8
85
Final M6.SPD.002
48
June 2001
-------
Table 7a
Facility-Specific/Area-Wide Speed Correction Cycles Test Results
Total Hydrocarbons (THC)
Cycle
Freeway at 63.2 mph
Freeway at 59.7 mph
Freeway at 52.9 mph
Freeway at 30.5 mph
Freeway at 18.6 mph
Freeway at 13.1 mph
Freeway Ramps (34.6 mph)
Arterial/Collectors
at 24.8 mph
Arterial/Collectors
at 19.2 mph
Arterial/Collectors
at 11.6 mph
Local Roadways (12.9 mph)
Non-Freeway Area-Wide
Urban Travel (19.4 mph)
FTP (19.6 mph)
Running 505 (25.6 mph)
Unified Cycle (24.6 mph)
ST01(20.2 mph)
NYCC(7.1mph)
Normal Emitters
#of
veh.
61
61
61
61
61
61
61
61
61
61
61
60*
61
61
60*
61
61
Mean
(g/mile)
0.15
0.16
0.14
0.21
0.25
0.27
0.34
0.22
0.26
0.45
0.28
0.26
0.38
0.17
0.24
2.32
0.62
Std.
Dev.
0.19
0.17
0.17
0.26
0.30
0.33
0.46
0.26
0.32
0.84
0.34
0.31
0.27
0.23
0.27
2.29
1.09
High Emitters
#of
veh.
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
23*
24
Mean
(g/mile)
1.80
1.77
1.70
2.52
3.67
4.13
3.04
3.03
3.97
5.15
4.48
3.57
3.49
2.57
3.16
6.88
7.31
Std.
Dev.
1.66
1.69
1.38
2.12
3.75
4.06
2.21
3.07
4.79
5.63
5.07
3.06
2.77
2.51
3.33
5.36
7.82
* Test not done
Final M6.SPD.002
49
June 2001
-------
Table 7b
Facility-Specific/Area-Wide Speed Correction Cycles Test Results
Carbon Monoxide (CO)
Cycle
Freeway at 63.2 mph
Freeway at 59.7 mph
Freeway at 52.9 mph
Freeway at 30.5 mph
Freeway at 18.6 mph
Freeway at 13.1 mph
Freeway Ramps (34.6 mph)
Arterial/Collectors
at 24.8 mph
Arterial/Collectors
at 19.2 mph
Arterial/Collectors
at 11.6 mph
Local Roadways (12.9 mph)
Non-Freeway Area-Wide
Urban Travel (19.4 mph)
FTP (19.6 mph)
Running 505 (25.6 mph)
Unified Cycle (24.6 mph)
ST01 (20.2 mph)
NYCC(7.1 mph)
Normal Emitters
#of
veh.
70
70
70
70
70
70
70
70
70
70
70
69*
70
70
69*
70
70
Mean
(g/mile)
6.96
6.96
5.53
4.48
5.19
4.79
10.06
4.28
5.22
5.94
4.23
4.80
5.05
3.04
5.93
24.55
7.88
Std.
Dev.
7.71
6.12
5.33
4.01
4.90
4.45
10.79
3.87
5.01
5.65
4.14
4.62
3.70
2.75
5.34
16.54
8.12
High Emitters
#of
veh.
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
14*
15
Mean
(g/mile)
66.76
65.63
54.45
66.38
74.39
82.09
84.02
75.24
80.79
116.57
92.41
86.63
79.92
74.04
77.94
111.2
158.04
Std.
Dev.
52.09
54.63
41.82
43.18
63.48
77.01
57.32
59.12
62.65
94.9
87.81
62.32
56.89
57.5
58.19
70.30
136.34
* Test not done
Final M6.SPD.002
50
June 2001
-------
Table 7c
Facility-Specific/Area-Wide Speed Correction Cycles Test Results
Nitrogen Oxides (NOx)
Cycle
Freeway at 63.2 mph
Freeway at 59.7 mph
Freeway at 52.9 mph
Freeway at 30.5 mph
Freeway at 18.6 mph
Freeway at 13.1 mph
Freeway Ramps (34.6 mph)
Arterial/Collectors
at 24.8 mph
Arterial/Collectors
at 19.2 mph
Arterial/Collectors
at 11.6 mph
Local Roadways (12.9 mph)
Non-Freeway Area-Wide
Urban Travel (19.4 mph)
FTP (19.6 mph)
Running 505 (25.6 mph)
Unified Cycle (24.6 mph)
ST01 (20.2 mph)
NYCC(7.1mph)
Normal Emitters
#of
veh.
72
72
72
72
72
72
72
72
72
72
72
71*
72
72
71*
72
72
Mean
(g/mile)
0.77
0.74
0.70
0.63
0.72
0.51
0.98
0.68
0.79
0.96
0.73
0.71
0.70
0.59
0.84
1.85
0.95
Std.
Dev.
0.71
0.65
0.60
0.54
0.59
0.39
0.81
0.55
0.66
0.78
0.63
0.57
0.53
0.50
0.66
1.11
0.69
High Emitters
#of
veh.
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
12*
13
Mean
(g/mile)
3.35
3.27
3.20
3.15
3.73
2.81
4.00
3.47
3.77
4.44
3.74
3.56
3.25
3.67
3.83
3.78
4.07
Std.
Dev.
1.07
1.02
0.97
1.00
1.34
0.99
1.43
1.07
1.46
1.84
1.46
1.18
1.04
1.13
1.23
1.34
1.45
* Test not done
Final M6.SPD.002
51
June 2001
-------
Table 8
Analysis of Variance Results (ANOVA P- Values)
All Roadways
Factor*
Speed
Emitter Class
Speed*Emitter Class
THC
0.0000
0.0000
0.1411
CO
0.0000
0.0000
0.0152
NOx
0.0000
0.0000
0.9894
NMHC
0.0001
0.0000
0.1271
. .
Normal remitters
Arteri al/C ol 1 ector
and Freeway
Local Roadway
Freeway Ramp
Factor*
Roadway Type**
Speed*Roadway Type**
Vehicle Class
Speed*Vehicle Class
Standard***
Speed* Standard***
Vehicle Class
Standard***
Vehicle Class
Standard***
THC
0.0046
0.0354
0.0016
0.1754
0.0000
0.0002
0.0830
0.0000
0.2922
0.0003
CO
0.0006
0.0020
0.0031
0.8680
0.0000
0.0576
0.4038
0.0000
0.0443
0.0002
NOx
0.0000
0.0000
0.0012
0.5723
0.0000
0.6491
0.0124
0.0028
0.0018
0.0000
NMHC
0.0050
0.0440
0.0404
0.1802
0.0000
0.0001
0.5008
0.0000
0.7707
0.0007
. , .
rlign remitters
Arterial/Collector
and Freeway
Local Roadway
Freeway Ramp
Factor*
Roadway Type**
Speed*Roadway Type**
Vehicle Class
Speed*Vehicle Class
Vehicle Class
Vehicle Class
THC
0.1236
0.1176
0.5942
0.0641
0.8787
0.3701
CO
0.3307
0.6233
0.8984
0.0241
0.5511
0.1471
NOx
0.0000
0.0000
0.3961
0.9560
0.6093
0.6942
NMHC
0.1307
0.1203
0.5693
0.0699
0.8821
0.4075
* All emissions in Log (gram/hour) scale.
** Freeways versus Arterial/Collectors limited to speeds < 30 mph, including a
vehicle term.
*** There are no Tier 1 High emitters in sample. Some low emitting Tier 0 vehicles
are considered both as Tier 0 and as Tier 1 vehicles (see text).
Final M6.SPD.002
52
June 2001
-------
Table 8
Analyses of Variance Results (ANOVA p values)
Factor*
Emitter Level
Roadway type**
Vehicle Class
Standard***
Local/ Vehicle Class
Local/Standard
Ramp/ Vehicle Class
Ramp/Standard
Roadway type**
Vehicle Class
Standard***
THC
.0000
NMHC
.0000
CO
.0000
NOx
.0000
Normal Emitters Only
.0006
.0001
.0001
.0476
.0001
.0396
.0001
.0003
.0004
.0001
.1490
.0001
.0983
.0001
.0206
.0001
.0001
.0325
.0001
.0107
.0001
.0000
.0001
.0001
.2753
.0001
.0871
.0001
High Emitters Only
.3094
.067
NA
.3281
.067
NA
.0318
.0004
NA
.0000
.144
NA
* All emissions in Log (gram/hour) scale.
** Freeways versus Arterial/Collectors limited to speeds < 30 mph, including a
vehicle term.
*** There were no Tier 1 High emitters in sample. Some low emitting Tier 0 vehicles
were considered both as Tier 0 and as Tier 1 vehicles (see text).
Final M6.SPD.002
53
June 2001
-------
Table 9
Description of Sample Vehicles Used for Tier 1 Analysis
Veh
No.
5007
5010
5013
5015
5017
5018
5021
5038
5059
5060
5063
5217
5218
5221
5223
5225
5229
5234
5239
5240
Test
Site
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
E.LIB
AA
AA
AA
AA
AA
AA
AA
AA
AA
Tier
Std.
0
0
0
0
0
1
1
1
0
0
1
1
0
1
1
1
1
1
1
1
Mileage
101536
29392
72348
58538
67496
20855
28525
16557
6734
61163
20451
7573
89995
12877
17233
10064
9433
51168
68305
18992
FTP
NMHC
0.13
0.12
0.08
0.07
0.15
0.12
0.12
0.12
0.13
0.11
0.16
0.09
0.19
0.10
0.21
0.12
0.17
0.15
0.19
0.21
FTP
NOx
0.23
0.21
0.18
0.41
0.13
0.10
0.10
0.34
0.28
0.27
0.26
0.20
0.39
0.53
0.49
0.40
0.10
0.26
0.71
0.31
Veh
Class
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDV
LDT2
LDV
LDT1
LDV
LDV
LDT1
LDV
Model
Yr.
88
93
93
93
91
95
95
96
93
93
96
96
92
96
96
96
96
94
94
96
Eng.
Size
3.80
2.20
2.30
2.30
1.80
2.00
2.00
3.10
2.20
2.20
3.10
2.20
1.90
4.90
3.10
2.20
1.60
1.90
3.80
3.00
Fuel
Inj.
PFI
TBI
PFI
PFI
TBI
TBI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
PFI
IM240
Status
PASS
PASS
PASS
PASS
PASS
PASS
PASS
PASS
PASS
PASS
PASS
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
NULL
VIN
1G3HY5C9JW3 12653
4T1SK12E9PU18406
1G2NE5438PC758996
1G2NE5438PC758996
WVWEB5159MK012875
1B3ES27C9SD221573
1B3ES67C2SD188892
2GIWL52M2T92 12643
4T1SK11E4PU252562
1HGCB7658PA075439
1G2WJ52M7TF204255
1HGCD5632TA260884
1G8ZF5498NZI75489
1TEF1549TLB25543
2G1WL52M2T9212643
1FTCR10A9TPB08548
1HGEJ8142TL073569
1FARP15J9RW262996
1GMDU06LXRT234029
4T1BF12K3TU871236
Final M6.SPD.002
54
June 2001
-------
Table 10
Tests of Convergence in Arterial and Freeway Estimates at 30 mph
Tier 0 Normal Emitter Sample
Parameter
THC
NMHC
CO
NOx
Estimate
0.18089092
0.15532642
1.63652794
0.05946957
T for HO:
Parameter = 0
1.84
1.83
2.96
1.24
p value
Pr>|T
0.0670
0.0688
0.0033
0.2160
Standard Error
of the Estimate
0.09840365
0.08503405
0.55229111
0.04797825
Tier 0 High Emitter Sample
Parameter
THC
NMHC
CO
NOx
Estimate
0.95357931
0.84766279
24.7784634
-0.00945343
T for HO:
Parameter = 0
1.52
1.58
1.48
-0.04
p value
Pr> T|
0.1304
0.1161
0.1430
0.9705
Standard Error
of the Estimate
0.62676490
0.53612496
16.7645083
0.25464544
Tier 1 Normal Emitter Sample
Parameter
THC
NMHC
CO
NOx
Estimate
0.01509669
0.00615421
0.25453921
0.04101364
T for HO:
Parameter = 0
1.15
0.71
0.83
1.20
p value
Pr> T|
0.2534
0.4813
0.4114
0.2350
Standard Error
of the Estimate
0.01310665
0.00869272
0.30796933
0.03423678
Final M6.SPD.002
55
June 2001
-------
Table lla
Average Emissions by Emission Standard and Emission Level
Total Hydrocarbons (THC)
Cycle
Freeway at 63.2 mph
Freeway at 59.7 mph
Freeway at 52.9 mph
Freeway at 30.5 mph
Freeway at 18.6 mph
Freeway at 13.1 mph
Freeway Ramps
(34.6 mph)
Arterial/Collectors
at 24. 8 mph
Arterial/Collectors
at 19.2 mph
Arterial/Collectors
at 11. 6 mph
NYCC(7.1mph)
Local Roadways
(12.9 mph)
Non-Freeway Area-
wide Urban Travel
(19.4 mph)
Hot Running LA4
(19.6 mph)
Unified Cycle
(24.6 mph)
Tier 1*
#of
veh.
20
20
20
20
20
20
20
20
20
20
20
20
19
20
19
Mean
(g/mi)
0.050
0.066
0.035
0.038
0.044
0.046
0.083
0.044
0.060
0.063
0.122
0.053
0.057
0.036
0.060
Std.
Dev.
0.032
0.038
0.019
0.031
0.036
0.040
0.080
0.035
0.054
0.045
0.111
0.056
0.047
0.019
0.049
Tier 0 Normal
#of
veh.
49
49
49
49
49
49
49
49
49
49
49
49
49
49
48
Mean
(g/mi)
0.183
0.187
0.171
0.253
0.305
0.330
0.408
0.262
0.318
0.551
0.744
0.336
0.311
0.199
0.282
Std.
Dev.
0.200
0.180
0.178
0.272
0.318
0.341
0.488
0.278
0.341
0.917
1.183
0.360
0.325
0.201
0.287
Tier 0 High
#of
veh.
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
Mean
(g/mi)
1.798
1.771
1.702
2.523
3.672
4.127
3.036
3.028
3.970
5.155
7.306
4.478
3.571
3.175
3.158
Std.
Dev.
1.656
1.688
1.384
2.124
3.745
4.063
2.205
3.072
4.794
5.630
7.824
5.075
3.060
2.945
3.328
Final M6.SPD.002
56
June 2001
-------
Table lib
Average Emissions by Emission Standard and Emission Level
Carbon Monoxide (CO)
Cycle
Freeway at 63.2 mph
Freeway at 59.7 mph
Freeway at 52.9 mph
Freeway at 30.5 mph
Freeway at 18.6 mph
Freeway at 13.1 mph
Freeway Ramps
(34.6 mph)
Arteri al/C oil ectors
at 24. 8 mph
Arteri al/C oil ectors
at 19. 2 mph
Arteri al/C oil ectors
at 11. 6 mph
NYCC(7.1mph)
Local Roadways
(12.9 mph)
Non-Freeway Area-
wide Urban Travel
(19.4 mph)
Hot Running LA4
(19.6 mph)
Unified Cycle
(24.6 mph)
Tier 1
#of
veh.
20
20
20
20
20
20
20
20
20
20
20
20
19
20
19
Mean
(g/mi)
1.862
3.045
1.381
1.305
1.513
1.264
2.803
1.271
1.562
1.538
2.652
1.249
1.357
0.892
1.892
Std.
Dev.
1.765
1.446
1.179
1.636
1.570
1.564
2.651
1.215
1.638
1.699
3.068
1.727
1.580
0.846
2.104
Tier 0 Normal
#of
veh.
58
58
58
58
58
58
58
58
58
58
58
58
58
58
57
Mean
(g/mi)
8.157
7.755
6.449
5.218
5.978
5.596
11.665
4.934
6.052
6.902
9.061
4.924
5.497
3.569
6.855
Std.
Dev.
7.945
6.410
5.403
3.998
4.997
4.464
11.170
3.921
5.103
5.727
8.384
4.212
4.696
2.997
5.394
Tier 0 High
#of
veh.
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
Mean
(g/mi)
66.763
65.632
54.448
66.377
74.390
82.087
84.016
75.235
80.793
116.56
9
158.04
1
92.412
86.628
82.194
77.941
Std.
Dev.
52.094
54.628
41.822
43.185
63.484
77.005
57.322
59.118
62.646
94.897
136.341
87.806
62.322
64.114
58.194
Final M6.SPD.002
57
June 2001
-------
Table lie
Average Emissions by Emission Standard and Emission Level
Oxides of Nitrogen (NOx)
Cycle
Freeway at 63.2 mph
Freeway at 59.7 mph
Freeway at 52.9 mph
Freeway at 30.5 mph
Freeway at 18.6 mph
Freeway at 13.1 mph
Freeway Ramps
(34.6 mph)
Arterial/Collectors
at 24. 8 mph
Arterial/Collectors
at 19.2 mph
Arterial/Collectors
at 11. 6 mph
NYCC(7.1mph)
Local Roadways
(12.9 mph)
Non-Freeway Area-
wide Urban Travel
(19.4 mph)
Hot Running LA4
(19.6 mph)
Unified Cycle
(24.6 mph)
Tier 1
#of
veh.
20
20
20
20
20
20
20
20
20
20
20
20
19
20
19
Mean
(g/mi)
0.331
0.340
0.241
0.234
0.231
0.187
0.324
0.233
0.376
0.416
0.353
0.311
0.253
0.191
0.357
Std.
Dev.
0.353
0.287
0.164
0.158
0.168
0.143
0.222
0.163
0.476
0.605
0.292
0.426
0.159
0.123
0.255
Tier 0 Normal
#of
veh.
60
60
60
60
60
60
60
60
60
60
60
60
60
60
59
Mean
(g/mi)
0.840
0.806
0.789
0.709
0.817
0.585
1.106
0.769
0.905
1.093
1.093
0.830
0.796
0.591
0.943
Std.
Dev.
0.736
0.674
0.619
0.558
0.591
0.386
0.823
0.559
0.660
0.777
0.672
0.637
0.583
0.457
0.678
Tier 0 High
#of
veh.
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
Mean
(g/mi)
3.354
3.270
3.200
3.155
3.727
2.805
3.998
3.473
3.774
4.435
4.072
3.735
3.561
3.245
3.830
Std.
Dev.
1.069
1.021
0.970
0.996
1.339
0.995
1.435
1.068
1.461
1.841
1.455
1.463
1.179
1.045
1.230
Final M6.SPD.002
58
June 2001
-------
Table lid
Average Emissions by Emission Standard and Emission Level
Non-Methane Hydrocarbons (NMHC)
Cycle
Freeway at 63.2 mph
Freeway at 59.7 mph
Freeway at 52.9 mph
Freeway at 30.5 mph
Freeway at 18.6 mph
Freeway at 13.1 mph
Freeway Ramps
(34.6 mph)
Arterial/Collectors
at 24. 8 mph
Arterial/Collectors
at 19.2 mph
Arterial/Collectors
at 11. 6 mph
NYCC(7.1mph)
Local Roadways
(12.9 mph)
Non-Freeway Area-
wide Urban Travel
(19.4 mph)
Hot Running LA4
(19.6 mph)
Unified Cycle
(24.6 mph)
Tier 1
#of
veh.
19
20
19
19
16
17
18
20
18
20
19
17
18
20
19
Mean
(g/mi)
0.038
0.052
0.026
0.025
0.031
0.027
0.068
0.029
0.042
0.034
0.082
0.038
0.038
0.020
0.041
Std.
Dev.
0.023
0.035
0.015
0.020
0.031
0.022
0.069
0.028
0.041
0.022
0.089
0.045
0.033
0.009
0.039
Tier 0 Normal
#of
veh.
49
49
48
49
48
49
47
49
48
49
49
48
49
49
48
Mean
(g/mi)
0.148
0.154
0.140
0.207
0.250
0.259
0.357
0.214
0.264
0.458
0.622
0.280
0.257
0.157
0.232
Std.
Dev.
0.177
0.162
0.159
0.246
0.288
0.310
0.444
0.252
0.304
0.805
1.024
0.334
0.301
0.176
0.265
Tier 0 High
#of
veh.
24
24
24
24
24
24
24
24
24
24
24
24
24
24
24
Mean
(g/mi)
1.633
1.601
1.537
2.290
3.347
3.740
2.767
2.737
3.616
4.665
6.571
4.059
3.245
2.945
2.860
Std.
Dev.
1.524
1.518
1.231
1.847
3.295
3.463
1.957
2.672
4.291
4.888
6.609
4.426
2.635
2.770
2.930
Final M6.SPD.002
59
June 2001
-------
Table 12a
Regressions of Emissions Versus Average Speed
Total Hydrocarbons (THC)
Emissions = Constant + a*(Average Speed)
Roadway Type
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Arterial/
Collector
Arterial/
Collector
Arterial/
Collector
Emission Level
1
(Tier 1)
1
(Tier 1)
1
(Tier 1)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
3
(Tier 0 High)
O
(Tier 0 High)
3
(Tier 0 High)
1
(Tier 1)
2
(Tier 0 Normal)
O
(Tier 0 High)
Speed Data
Range (mph)
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 -24.8
7.1 -24.8
7.1 -24.8
Constant
(p value)
1.034*
0.202
(.4780)
0.019
(.2157)
6.672*
1.933
(.2284)
0.315
(.0000)
44.558**
(.0013)
44.558**
(.0013)
3.193
(.0000)
0.690
(.0009)
4.891
(.0001)
44.558**
(.0013)
a
(p value)
-0.032*
0.032
(.0175)
0.001
(.0533)
-0.170*
0.192
(.0094)
-0.00226
(.0570)
1.202**
(.0908)
1.202**
(.0908)
-0.024
(.0836)
0.017
(.0958)
0.081
(.1930)
1.202**
(.0908)
Emission
Units
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per hour
* The values are calculated based on the NYCC at 7.1 mph and Freeway at 13.1 mph cycles.
** Freeway and Arterial/Collector cycles were combined.
Final M6.SPD.002
60
June 2001
-------
Table 12b
Regressions of Emissions Versus Average Speed
Carbon Monoxide (CO)
Emissions = Constant + a*(Average Speed)
Roadway Type
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Arterial/
Collector
Arterial/
Collector
Arterial/
Collector
Emission Level
1
(Tier 1)
1
(Tier 1)
1
(Tier 1)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
3
(Tier 0 High)
3
(Tier 0 High)
3
(Tier 0 High)
1
(Tier 1)
2
(Tier 0 Normal)
3
(Tier 0 High)
Speed Data
Range (mph)
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 -24.8
7.1 -24.8
7.1 -24.8
Constant
(p value)
14.730*
1.655
(.9045)
0.246
(.7436)
46.679*
15.273
(.4824)
2.398
(.1526)
1206.641*
365.822
(.4888)
64.691
(.0147)
10.036
(.1950)
36.128
(.0054)
863.64
(.0114)
a
(p value)
0.280*
1.278
(.0454)
0.032
(.0263)
2.390*
4.788
(.0000)
0.0872
(.0060)
-9.747*
54.438
(.0275)
-0.0269
(.9559)
0.941
(.0138)
3.877
(.0000)
38.563
(.0202)
Emission
Units
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per hour
: The values are calculated based on the NYCC at 7.1 mph and Freeway at 13.1 mph cycles.
Final M6.SPD.002
61
June 2001
-------
Table 12c
Regressions of Emissions Versus Average Speed
Oxides of Nitrogen (NOx)
Emissions = Constant + a*(Average Speed)
Roadway Type
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Arterial/
Collector
Arterial/
Collector
Arterial/
Collector
Emission Level
1
(Tier 1)
1
(Tier 1)
1
(Tier 1)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
3
(Tier 0 High)
3
(Tier 0 High)
3
(Tier 0 High)
1
(Tier 1)
2
(Tier 0 Normal)
3
(Tier 0 High)
Speed Data
Range (mph)
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 -24.8
7.1 -24.8
7.1 -24.8
Constant
(p value)
4.625*
-0.855
(.5289)
0.126
(.2886)
8.291*
-0.957
(.7262)
0.594
(.0008)
24.889*
0.423
(.9717)
2.980
(.0000)
2.325
(.1066)
5.123
(.0027)
14.609
(.0471)
a
(p value)
-0.154*
0.264
(.0001)
0.0031
(.1667)
0.121*
0.761
(.0000)
0.00373
(.2575)
1.364*
3.232
(.0000)
0.00512
(.6389)
0.170
(.0167)
0.567
(.0000)
2.812
(.0000)
Emission
Units
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per hour
* The values are calculated based on the NYCC at 7.1 mph and Freeway at 13.1 mph cycles.
Final M6.SPD.002
62
June 2001
-------
Table 12d
Regressions of Emissions Versus Average Speed
Non-Methane Hydrocarbons (NMHC)
Emissions = Constant + a*(Average Speed)
Roadway Type
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Freeway
Arterial/
Collector
Arterial/
Collector
Arterial/
Collector
Emission Level
1
(Tier 1)
1
(Tier 1)
1
(Tier 1)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
2
(Tier 0 Normal)
3
(Tier 0 High)
3
(Tier 0 High)
3
(Tier 0 High)
1
(Tier 1)
2
(Tier 0 Normal)
3
(Tier 0 High)
Speed Data
Range (mph)
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 - 13.1
13.1 -30.5
30.5-63.2
7.1 -24.8
7.1 -24.8
7.1 -24.8
Constant
(p value)
0.685*
0.00266
(.9892)
0.00475
(.6971)
5.796*
1.328
(.3602)
0.259
(.0000)
40.178*
37.404
(.0580)
2.899
(.0000)
0.399
(.0082)
4.111
(.0003)
42.589
(.0023)
a
(p value)
-0.028*
0.0236
(.0105)
0.000592
(.0115)
-0.176*
0.165
(.0131)
-0.00189
(.0773)
1.103*
1.107
(.2142)
-0.022
(.0773)
0.0118
(.1048)
0.0617
(.2612)
1.017
(.1299)
Emission
Units
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per mile
grams
per hour
grams
per hour
grams
per hour
* The values are calculated based on the NYCC at 7.1 mph and Freeway at 13.1 mph cycles.
Final M6.SPD.002
63
June 2001
-------
Table 13
Freeway Ramp and Local Roadway Emissions
As a Function of Hot Running LA4 Emissions
In Grams/Hour
Emissions (g/hr) = Constant + a*(LA4) + b*(LA42)
where LA4 is the hot running LA4 emissions in g/hr
Roadway Type
Freeway Ramp
(34.6 mph)
Freeway Ramp
(34.6 mph)
Freeway Ramp
(34.6 mph)
Freeway Ramp
(34.6 mph)
Local Roadways
(12.9 mph)
Local Roadways
(12.9 mph)
Local Roadways
(12.9 mph)
Local Roadways
(12.9 mph)
Pollutant
THC
CO
NOx
NMHC
THC
CO
NOx
NMHC
Constant
(p value)
4.560
(.0302)
224.333
(.0010)
5.353
(.1103)
4.368
(.0193)
0.00
0.00
0.00
0.00
a
(p value)
2.046
(.0000)
2.040
(.0000)
2.863
(.0000)
2.014
(.0000)
1.0319
(.0000)
0.7405
(.0000)
0.8156
(.0000)
1.1097
(.0000)
b
(p value)
-0.00356
(.0000)
-0.000145
(.0074)
-.0101
(.0019)
-0.00387
(.0000)
-0.0007
(.2960)
0.000
(.9242)
-0.0005
(.4656)
-0.0015
(.0172)
R2
0.934
0.848
0.866
0.934
0.804
0.831
0.952
0.804
Final M6.SPD.002
64
June 2001
-------
Table 14
Emission Offset
(Predicted Freeway Emissions - Average Hot Running LA4 Emissions)
THC
CO
NOx
NMHC
Level 1 (Tier 1)
(grams per mile)
Fwy
0.042
1.363
0.220
0.024
LA4
0.036
0.892
0.191
0.020
Offset
0.006
0.471
0.029
0.004
Level 2 (Tier 0)
(grams per mile)
Fwy
0.290
5.567
0.712
0.233
LA4
0.199
3.569
0.591
0.157
Offset
0.091
1.998
0.121
0.076
Level 3 (High Emitters)
(grams per mile)
Fwy
3.476
73.102
3.253
3.153
LA4
3.175
82.194
3.245
2.945
Offset
0.301
-9.092
0.008
0.208
Final M6.SPD.002
65
June 2001
-------
Table 15
Arterial/Collector Emission Offsets (AEO)
(Predicted Arterial/Collector Emissions - Predicted Freeway Emissions)
Pollutant
THC
CO
NOx
NMHC
Average Speed*
(miles per hour)
10
15
20
25
30
10
15
20
25
30
10
15
20
25
30
10
15
20
25
30
Level 1
(grams per mile)
0.014
0.018
0.009
0.005
0.001
0.192
0.222
0.082
0
0
0.094
0.118
0.065
0.033
0.012
0.012
0.015
0.008
0.004
0.001
Level 2
(grams per mile)
0.073
0.086
0.037
0.007
0
0.431
0.479
0.131
0
0
0.171
0.211
0.110
0.049
0.009
0.069
0.082
0.035
0.008
0
Level 3
(grams per mile)
0
0
0
0
0
14.010
17.313
9.016
4.038
0.719
0.420
0.526
0.290
0.148
0.053
0
0
0
0
0
* Arterial/Collector Emission Offsets below 10 mph and over 30 mph are zero.
Final M6.SPD.002
66
June 2001
-------
Table 16
Speed Correction Factors
For Freeways
By Emission Level*
Avg.
Speed
(mph)
7.1
10
15
19.6
20
25
30
35
40
45
50
55
60
65
Total Hydrocarbons
(THC)
Level 1
2.71
1.71
1.08
1.00
1.00
0.95
0.91
0.91
0.97
1.04
1.10
1.17
1.24
1.30
Level 2
2.65
1.71
1.10
1.00
0.99
0.93
0.88
0.81
0.77
0.73
0.69
0.65
0.61
0.57
Level 3
2.15
1.63
1.20
1.00
0.99
0.86
0.77
0.68
0.64
0.61
0.57
0.54
0.50
0.47
Carbon Monoxide
(CO)
Level 1
1.73
1.29
1.02
1.00
1.00
0.99
0.98
1.00
1.12
1.24
1.36
1.47
1.59
1.71
Level 2
1.61
1.27
1.04
1.00
1.00
0.97
0.95
0.98
1.06
1.14
1.21
1.29
1.37
1.45
Level 3
2.19
1.52
1.08
1.00
0.99
0.94
0.91
0.87
0.87
0.87
0.87
0.86
0.86
0.86
Oxides of Nitrogen
(NOx)
Level 1
2.26
1.40
0.94
1.00
1.00
1.04
1.07
1.07
1.14
1.21
1.28
1.35
1.42
1.49
Level 2
1.81
1.28
0.98
1.00
1.00
1.01
1.02
1.02
1.04
1.07
1.09
1.12
1.15
1.17
Level 3
1.50
1.18
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.01
1.02
Non-Methane HC
(NMHC)
Level 1
2.87
1.69
1.00
1.00
1.00
1.00
1.00
1.07
1.20
1.32
1.45
1.57
1.70
1.82
Level 2
2.75
1.73
1.09
1.00
0.99
0.94
0.90
0.83
0.79
0.74
0.70
0.66
0.62
0.58
Level 3
2.14
1.62
1.20
1.00
0.99
0.86
0.77
0.68
0.64
0.61
0.57
0.54
0.50
0.47
* Emission levels shown as Fwy emissions in Table 14. See Section 4.6.
Final M6.SPD.002
67
June 2001
-------
Table 17
Speed Correction Factors
For Arterial/Collector Roadways
By Emission Level*
Avg.
Speed
(mph)
7.1
10
15
20
25
30
35
40
45
50
55
60
65
Total Hydrocarbons
(THC)
Level 1
2.71
2.04
1.49
1.22
1.05
0.95
0.91
0.97
1.04
1.10
1.17
1.24
1.30
Level 2
2.65
1.96
1.40
1.12
0.95
0.88
0.81
0.77
0.73
0.69
0.65
0.61
0.57
Level 3
2.15
1.63
1.20
0.99
0.86
0.77
0.68
0.64
0.61
0.57
0.54
0.50
0.47
Carbon Monoxide
(CO)
Level 1
1.73
1.43
1.18
1.06
0.99
0.98
1.00
1.12
1.24
1.36
1.47
1.59
1.71
Level 2
1.61
1.35
1.13
1.02
0.97
0.95
0.98
1.06
1.14
1.21
1.29
1.37
1.45
Level 3
2.19
1.71
1.32
1.12
1.00
0.92
0.87
0.87
0.87
0.87
0.86
0.86
0.86
Oxides of Nitrogen
(NOx)
Level 1
2.26
1.82
1.47
1.30
1.19
1.12
1.07
1.14
1.21
1.28
1.35
1.42
1.49
Level 2
1.81
1.52
1.28
1.16
1.08
1.04
1.02
1.04
1.07
1.09
1.12
1.15
1.17
Level 3
1.50
1.31
1.16
1.09
1.04
1.01
1.00
1.00
1.00
1.00
1.00
1.01
1.02
Non-Methane HC
(NMHC)
Level 1
2.87
2.18
1.62
1.34
1.17
1.06
1.07
1.20
1.32
1.45
1.57
1.70
1.82
Level 2
2.75
2.03
1.44
1.15
0.97
0.90
0.83
0.79
0.74
0.70
0.66
0.62
0.58
Level 3
2.14
1.62
1.20
0.99
0.86
0.77
0.68
0.64
0.61
0.57
0.54
0.50
0.47
* Emission levels shown as Fwy emissions in Table 14. See Section 4.6.
Final M6.SPD.002
68
June 2001
-------
Table 18
Speed Correction Factors for Light Duty Diesel Vehicles
SCF(s) = EXP(B*(s-sadj) + C*(s**2-sadj**2))
s = average speed (mph)
sadj = basic test procedure speed; adjusted for VMT fraction of cold start operation x (0.206)
and VMT fraction of hot start operation w (0.273), assuming FTP weighting.
1/sadj = (w+x)/26 + (l-w-x)/16
sadj = 19.6 mph
Pollutant
THC
CO
NOx
Model Years
All
All
All
Coefficient Values
(B)
-0.055
-0.088
-0.048
(C)
0.00044
0.00091
0.00071
From "AP-42 Volume II, Compilation of Air Pollution Emission Factors, Mobile Sources,"
Appendix H, Table 5.6 and Table 6.6 (June 30, 1995)
Final M6.SPD.002
69
June 2001
-------
Table 19
Speed Correction Factors for Heavy Duty Gasoline Vehicles
SCF(s) = EXP(A + B*s + C*s**2), for THC & CO
SCF(s) = A + B*s + C*s**2, for NOx
s = average speed (mph)
Pollutant
THC
CO
NOx
Model Years
All
All
All
Coefficient Values
(A)
1.608
1.520
0.824
(B)
-0.097
-0.098
0.0088
(C)
0.00083
0.0011
0.00
From "AP-42 Volume II, Compilation of Air Pollution Emission Factors, Mobile Sources,"
Appendix H, Table 4.6 (June 30, 1995)
Table 20
Speed Correction Factors for Heavy Duty Diesel Vehicles
SCF(s) = EXP(A + B*s + C*s**2)
s = average speed (mph)
Pollutant
THC
CO
NOx
Model Years
All
All
All
Coefficient Values
(A)
0.924
1.396
0.676
(B)
-0.055
-0.088
-0.048
(C)
0.00044
0.00091
0.00071
From "AP-42 Volume II, Compilation of Air Pollution Emission Factors, Mobile Sources,"
Appendix H, Table 7.6 (June 30, 1995)
Final M6.SPD.002
70
June 2001
-------
Table 21
Speed Correction Factors for Motorcycles
SCF( s) = SF( s)/ SF( sadj)
SF(s) = EXP( A + B* s + C* s** 2+ D* s** 3+ E* s** 4+ F* s** 5), for THC & CO
SF(s) = A + B* s + C* s**2+D* s** 3+ E* s** 4 + F* s** 5, forNOx
s = average speed (mph)
sadj = 19.6 mph
Pollutant
Model Years
Coefficient Values
(A)
(B)
(C)
(D)
(E)
(F)
Low Altitude
THC
CO
NOx
Pre-1978
1978-1979
1980+
Pre-1978
1978-1979
1980+
Pre-1978
1978+
2.31E+00
2.41E+00
2.25E+00
2.34E+00
2.78E+00
2.71E+00
1.69E+00
1.28E+00
-2.90E-01
-3.08E-01
-2.88E-01
-2.97E-01
-3.19E-01
-3.31E-01
-1.18E-01
-8.05E-02
1.53E-02
1.68E-02
1.57E-02
1.60E-02
1.53E-02
1.76E-02
6.55E-03
5.36E-03
-4.47E-04
-5.07E-04
-4.73E-04
-4.77E-04
-4.22E-04
-5.39E-04
-1.37E-04
-1.19E-04
6.48E-06
7.54E-06
7.08E-06
7.07E-06
5.85E-06
8.17E-06
1.01E-06
9.01E-07
-3.63E-08
-4.32E-08
-4.08E-08
-4.04E-08
-3.15E-08
-4.78E-08
O.OOE+00
O.OOE+00
High Altitudes
THC
CO
NOx
Pre-1978
1978-1979
1980+
Pre-1978
1978-1979
1980+
Pre-1978
1978+
2.25E+00
2.15E+00
2.12E+00
1.82E+00
1.82E+00
2.05E+00
2.44E+00
1.45E+00
-2.91E-01
-2.84E-01
-2.91E-01
-2.55E-01
-2.72E-01
-3.11E-01
-2.50E-01
-1.22E-01
1.59E-02
1.54E-02
1.69E-02
1.52E-02
1.70E-02
2.05E-02
1.38E-02
7.95E-03
-4.72E-04
-4.42E-04
-5.26E-04
-4.87E-04
-5.52E-04
-7.09E-04
-2.87E-04
-1.71E-04
6.94E-06
6.29E-06
8.03E-06
7.58E-06
8.63E-06
1.16E-05
2.08E-06
1.26E-06
-3.93E-08
-3.46E-08
-4.70E-08
-4.50E-08
-5.11E-08
-7.16E-08
O.OOE+00
O.OOE+00
From "AP-42 Volume H,
Appendix H, Table 8.6.1
Compilation of Air Pollution Emission Factors, Mobile Sources,"
and Table 8.6.2 (June 30, 1995)
Final M6.SPD.002
71
June 2001
-------
Figures
Final M6.SPD.002 72 June 2001
-------
Fig
ure 1a.
Facility Cycles Ratio of Means, HC by
C ^j
0) O)
*r C
O C
O 3
1 §
4
3
3
2
2
1
1
0
0
.000 -i
.500 -
.000 -
.500 -
.000 -
.500 -
.000 -
.500 -
.000 -
C
Emitter Level Groups
D
Q^p ^k
n * *
y fi *
Q
i
) 20
n
n <
Low Emitters
n High Emitters
^
n D n
I i
40
60
80
Average Speed (mph)
Figure 1b.
Facility Cycles Ratio of Means, CO by
Emitter Level Groups
-Ť 3.000 -
| 3 2.500 -
0) O)
^ .E 2.000 -
| I 1'50° "
| 0 1.000 -
~ 0.500 -
Onnn -
n * *
n
n JJ g n
D
I
ť Low Emitters
n High Emitters
n
D
i i
0 20 40 60
Average Speed (mph)
D
80
Final M6.SPD.002
73
June 2001
-------
Fig u re 1 c.
Facility Cycles Ratio of M
eans, NOx
by Emitter Level Groups
c
re
o>
^
"o
.0
"re
^
_i
0)
_c
'c
c
^
"x
O
2
1
1
1
1
1
0
0
0
0
0
.000 -i
.800 -
.600 -
.400 -
.200 -
.000 -
.800 -
.600 -
.400 -
.200 -
.000 -
* *
ť
nť C
- ° n ffl ° * °
T3 n
n
n
i i
0 20 40
Average Speed
ť
n an
ť Low Em itters
n H igh Emitters
i
60
(mph)
80
Final M6.SPD.002
74
June 2001
-------
.
<Ť 3
C _J
re '
0) O)
^ c
n- C
0 c
0 =
Ł S
Figure 2a.
Facility Cycle Data, HC
4.000 -i
3.500 -
3.000 -
2.500 -
2.000 -
1.500 -
1.000 -
0.500 -
0.000 -
-
- \
X
\
V +
\
*^^^Ť^
ť *-.
I I I
0 20 40 60 80
Average Speed
Legend for Figures 2a, b and c
.* -Arterial/Collectors
.^^Freeways
O LA92
A A re a-wide, non-free way
X Local
+ Ram p
Final M6.SPD.002
75
June 2001
-------
Figure 2b.
^ 3.000 -
n -1 2-500 -
0) O)
S .E 2.000 -
M- C
0 = 1.500 -
o ^
re Q 1 .000 -
0.500 -
0.000 -
C
Facility Cycle Data, CO
+
- \
X ^
m
I I I
) 20 40 60 8
0
Average Speed
^,
= 1
^~ ^
0 |
.2 ^
Ť ^
K 0
z
Fig u re 2c.
Facility Cycle Data, NOx
2.000 -i
1.800 -
1.600 -
1.400 -
1.200 -
1.000 -
0.800 -
0.600 -
0.400 -
0.200 -
0.000 -
C
- * % +
> 0
Ak
: ^
i i i
) 20 40 60 80
Average Speed
Final M6.SPD.002
76
June 2001
-------
23
8 g
^ .c
Ratio of r
HC/Runn
4.000 -
3.500 -
3.000 -
2.500 -
2.000 -
1.500 -
1.000 -
0.500 -
0.000 -
(
Figure 3a.
Facility Cycle Data, THC
* Tier 0 vs. Tier 1
X \
. V
V \
X ^^^
1
) 20
O
i i
40 60
Average Speed
i
80
Legend for Figures 3a, 3b and 3c
^^ - Arterial/Collectors, Tier 0
Freeways, Tier 0
o Local, Tier 0
n Ramp, Tier 0
o LA92, TierO
A Area-wide, non-freeway, TierO
- -*- -Arterial/Collectors, Tier 1
- Freeways, Tier!
x Local, Tier 1
Ramp, Tier 1
LA92, Tier 1
x Area-wide, non-freeway, Tier 1
Final M6.SPD.002
77
June 2001
-------
Figure 3b.
Facility Cycle Data, CO
6.000 j Tier o vs. Tier 1
5.000 -
w _
c ^*
(0 BŁ 4.000 -
G) ^
g^ť ^H ^^^
M- i < 3.000 -
js g 2.000 -
S ^
1.000 -
0.000 -
(
'!
* /
v n * 'ť
* * ť
V* * 7 ^
I I I
) 20 40 60
Average Speed
i
80
2.500 -
w < 2.000 -
rc o
|j c 1.500 -
0 ^ 1.000 -
ro "x
^ § 0.500 -
0.000 -
C
Figure 3c.
Facility Cycle Data, NOx
Tier 0 vs. Tier 1 ^
_ *
^-^ n *
SI
o
^
^^v^-
I I I
) 20 40 60
Average Speed
i
80
Final M6.SPD.002
78
June 2001
-------
8.0
I 5'°
C/>
Ł
O 4.0
3.0 -
2.0
1.0
10
Figure 4a
Total Hydrocarbon (THC)
Freeway Emission Levels
15
-Level 1 (Tier 1 Normal)
-Level 2 (Tier 0 Normal)
-Level 3 (Tier 0 High)
20 25
Average Speed (mph)
180
160
140
100
80
60
40
20
10
Figure 4b
Carbon Monoxide (CO)
Freeway Emission Levels
15
-*- Level 1 (Tier 1 Normal)
--Level 2 (Tier 0 Normal)
-ť-Level 3 (Tier 0 High)
20 25
Average Speed (mph)
30
35
40
Final M6.SPD.002
79
June 2001
-------
Figure 4c
Oxides of Nitrogen (NOx)
5.0
4.5
4.0
3.5
°>
'§
Ł 3.0
0 2.5
=
tfi
.1 2.0
_tfl
1.5
1.0
0.5
0.0
I
Freeway Emission Levels
\
v^
^-
10 15 20 25
-*-Level 1 (Tier 1 Normal)
Level 2 (Tier 0 Normal)
--Level 3 (Tier 0 High)
30 35
40
Average Speed (mph)
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
Figure 4d
Non-Methane Hydrocarbons (NMHC)
Freeway Emission Levels
-*-Level 1 (Tier 1 Normal)
--Level 2 (Tier 0 Normal)
--Level 3 (TierO High)
10 15 20 25 30 35 40
Average Speed (mph)
Final M6.SPD.002
80
June 2001
-------
Figure 5a
Freeway Speed Correction Factors for
Total Hydrocarbons (THC)
-Emission Level 1
(0.04 g/mi)
-Emission Level 2
(0.29 g/mi)
-Emission Level 3
(3.48 g/mi)
30 40
Average Speed (mph)
Figure 5b
Freeway Speed Correction Factors for
Carbon Monoxide (CO)
--Emission Level 1
(1.4 g/mi)
-o-Emission Level 2
(5.6 g/mi)
-^-Emission Levels
(73.1 g/mi)
30 40
Average Speed (mph)
Final M6.SPD.002
81
June 2001
-------
Figure 5c
Freeway Speed Correction Factors for
Oxides of Nitrogen (NOx)
-Emission Level 1
(0.22 g/mi)
- Em ission Level 2
(0.71 g/mi)
- Em ission Level 3
(3.25 g/mi)
30 40
Average Speed (mph)
Figure 5d
Freeway Speed Correction Factors for
Non-Methane Hydrocarbons (NMHC)
o
O 1.5
-Emission Level 1
(0.02 g/mi)
-Emission Level 2
(0.23 g/mi)
-Emission Level 3
(3.15 g/mi)
30 40
Average Speed (mph)
Final M6.SPD.002
82
June 2001
-------
Figure 6a
Comparison to MOBILES
Speed Correction Factors for
Total Hydrocarbons (THC)
Emission Level 1 (0.04 g/mi)
Emission Level 2 (0.29 g/mi)
Emission Level 3 (3.48 g/mi)
M5 1975 MYR
M5 1981 MYR
M5 1991+ MYR
30 40
Average Speed (mph)
F ig u re 6 b
Comparison to MOBILES
Speed Correction Factors for
Carbon Monoxide (CO)
-Emission Level 1 (1.4 g/mi)
-Emission Level 2 (5.6 g/mi)
-Emission Level 3 (73.1 g/mi)
-M5 1975 MYR
-M5 1981 MYR
-M5 1993+ MYR
20 30 40 50
Average Speed (mph)
Final M6.SPD.002
83
June 2001
-------
Figure 6c
Comparison to MOBILES
Speed Correction Factors for
Oxides of Nitrogen (NOx)
-Emission Level 1 (0.22 g/mi)
-Emission Level 2 (0.71 g/mi)
-Emission Level 3 (3.25 g/mi)
-M5 1975 MYR
-M5 1981 MYR
-M5 1993+ MYR
30 40
Average Speed (mph)
Figure 7a
Arterial/Collector Speed Correction Factors
Total Hydrocarbons (THC)
Level 1 Emissions (0.04 g/mi)
3.0
1.0
0.0
-M5 1975 MYR
-M5 1981 MYR
-M5 1993+ MYR
-Freeway Level 1
-Arterial/Collector Level 1
10
15
20 25
Average Speed (mph)
30
35
40
Final M6.SPD.002
84
June 2001
-------
Figure 7b
Arterial/Collector Speed Correction Factors
Total Hydrocarbons (THC)
Level 2 Emissions (0.29 g/mi)
M5 1975 MYR
M5 1981 MYR
M5 1993+ MYR
Freeway Level 1
Arterial/Collector Level 1
20 25
Average Speed (mph)
Figure 7c
Arterial/Collector Speed Correction Factors
Total Hydrocarbons (THC)
Level 3 Emissions (3.48 g/mi)
-M5 1975 MYR
-M5 1981 MYR
-M5 1993+ MYR
-Freeway Level 3
-Arterial/Collector Level 3
20 25
Average Speed (mph)
Final M6.SPD.002
85
June 2001
-------
Figure 7d
Arterial/Collector Speed Correction Factors
Carbon Monoxide (CO)
Level 1 Emissions (1.36 g/mi)
-M5 1975 MYR
-M5 1981 MYR
-M5 1993+ MYR
-Freeway Level 1
-Arterial/Collector Level 1
20 25
Average Speed (mph)
2.5
1.0
0.5
0.0
Figure 7e
Arterial/Collector Speed Correction Factors
Carbon Monoxide (CO)
Level 2 Emissions (5.57 g/mi)
M5 1975 MYR
M5 1981 MYR
M5 1993+ MYR
Freeway Level 2
Arterial/Collector Level 2
20 25
Average Speed (mph)
30
35
40
Final M6.SPD.002
86
June 2001
-------
Figure 7f
Arterial/Collector Speed Correction Factors
Carbon Monoxide (CO)
Level 3 Emissions (73.1 g/mi)
M5 1975 MYR
M5 1981 MYR
M5 1993+ MYR
Freeway Level 3
Arterial/Collector Level 3
20 25
Average Speed (mph)
2.5
1-5
1.0
o.o
Figure 7g
Arterial/Collector Speed Correction Factors
Oxides of Nitrogen (NOx)
Level 1 Emissions (0.22 g/mi)
M5 1975 MYR
M5 1981 MYR
M5 1993+ MYR
Freeway Level 1
Arterial/Collector Level 1
20 25 30
Average Speed (mph)
35
40
Final M6.SPD.002
87
June 2001
-------
Z 1.0
8.
Figure 7h
Arterial/Collector Speed Correction Factors
Oxides of Nitrogen (NOx)
Level 2 Emissions (0.71 g/mi)
M5 1975 MYR
M5 1981 MYR
M5 1993+ MYR
Freeway Level 2
Arterial/Collector Level 2
20 25
Average Speed (mph)
2.5
Ł 1.5
1.0
0.0
Figure 7i
Arterial/Collector Speed Correction Factors
Oxides of Nitrogen (NOx)
Level 3 Emissions (3.25 g/mi)
-M5 1975 MYR
-M5 1981 MYR
-M5 1993+ MYR
-Freeway Level 3
-Arterial/Collector Level 3
10
15
20 25
Average Speed (mph)
30
35
40
Final M6.SPD.002
88
June 2001
-------
Figure 8a
CO Off-Cycle Emissions
-40
-50
10 20 30 40 50 60 70 80 90 100
Basic Exhaust Running Emissions (LA4 g/mi)
Final M6.SPD.002
89
June 2001
-------
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
-0.10
-0.20
-0.30
-0.40
-0.50
Figure 8b
THC Off-Cycle Emissions
o
I
u
>
O
2345
Basic Exhaust Running Emissions (LA4 g/mi)
Final M6.SPD.002
90
June 2001
-------
Figure 8c
NOx Off-Cycle Emissions
2.5
E 1.5
-ť <3.50 g/mi
-m- > 3.50 g/mi
o NOX DELT
X
O
0)
o
0.5
-0.5
Basic Exhaust Running Emissions (LA4 g/mi)
Final M6.SPD.002
91
June 2001
-------
Figure 9a
Comparison of MOBILE6 Level 1 NOx SCFs with Clean SFTP SCFs
2.4 1
2.0
c
o
I 1.6
o
O
"S 1.4
0
Q_
W
1.2
1.0
X
. X
10
20
30
40
50
Case
M6 Level 1 SCF
X Clean SFTP
60
70
Average Speed
Final M6.SPD.002
92
June 2001
-------
Figure 9b
Comparison of MOBILE6 Level 1 HC SCFs with Clean SFTP SCFs
3.0 1
2.0
c
o
'
0
0
Q.
W
1.0
X
10
20
30
40
Average Speed
50
60
Case
M6 Level 1 SCF
X Clean SFTP
70
Final M6.SPD.002
93
June 2001
-------
Figure 9c
Comparison of MOBILE6 Level 1 CO SCFs with Clean SFTP CO SCFs
2.5 T
2.0
u
0.
1.0
5 J
0
10 20
30 40
Average Speed
50 60
M6 Level 1 SCF
X Clean SFTP
70
Final M6.SPD.002
94
June 2001
-------
Appendices
Final M6.SPD.002 95 June 2001
-------
Appendix A : Statistics
MAIN EFFECTS & INTERACTIONS WITH SPEED
All Vehicles
FACTOR
S
EMIT_CLASS
S*EMIT_CLASS
THC
0.0000
0.0000
0.1411
NMHC
0.0001
0.0000
0.1271
CO
0.0000
0.0000
0.0152
NOX
0.0000
0.0000
0.9894
Final M6.SPD.002
96
June 2001
-------
EMIT NORMAL - ACTUAL TIER CLASS
ROAD
ART/FWY
LOCAL
RAMP
FACTOR
ROADTYPE
S*ROADTYPE
VEH TYPE
S*VEH TYPE
STANDARD
S* STANDARD
VEH TYPE
STANDARD
VEH TYPE
STANDARD
EMIT HIGH - A
THC
0.0046
0.0354
0.0016
0 . 1754
0.0000
0.0002
0.0830
0.0000
0.2922
0.0003
CTUAL TI
NMHC
0.0050
0 . 0440
0 . 0404
0.1802
0.0000
0.0001
0.5008
0.0000
0 .7707
0.0007
ER CLASS
CO
0.0006
0.0020
0.0031
0.8680
0.0000
0.0576
0.4038
0.0000
0.0443
0.0002
NOX
0.0000
0.0000
0.0012
0.5723
0.0000
0.6491
0.0124
0.0028
0.0018
0.0000
ROAD
ART/FWY
LOCAL
RAMP
FACTOR
ROADTYPE
S* ROADTYPE
VEH TYPE
S*VEH TYPE
STANDARD
S* STANDARD
VEH TYPE
STANDARD
VEH TYPE
STANDARD
THC
0.1236
0.1176
0.5942
0.0641
N/A
N/A
0 . 8787
N/A
0.3701
N/A
NMHC
0.1307
0.1203
0.5693
0.0699
N/A
N/A
0.8821
N/A
0.4075
N/A
CO
0.3307
0.6233
0.8984
0.0241
N/A
N/A
0.5511
N/A
0 . 1471
N/A
NOX
0.0000
0.0000
0.3961
0.9560
N/A
N/A
0.6093
N/A
0.6942
N/A
Final M6.SPD.002
97
June 2001
-------
EMIT NORMAL - CLEAN TIER 0 CLASS
THC NMHC CO NOX
ROAD FACTOR
ART/FWY
LOCAL
RAMP
ROADTYPE
0.0046
0.0050
0.0006
0.0000
S*ROADTYPE 0.0354 0.0440 0.0020 0.0000
VEH TYPE 0.0004 0.0243 0.0062 0.0026
S*VEH TYPE 0.1322 0.1476 0.8361 0.5608
CLEANTO 0.0000 0.0000 0.0000 0.0000
S*CLEANTO 0.0002 0.0001 0.0576 0.6491
VEH TYPE 0.0572 0.4049 0.1660 0.0184
CLEANTO 0.0000 0.0000 0.0000 0.0028
VEH TYPE 0.1570 0.5501 0.0201 0.0009
CLEANTO 0.0003 0.0007 0.0002 0.0000
EMIT HIGH - CLEAN TIER 0 CLASS
THC NMHC CO NOX
ROAD FACTOR
ART/FWY
LOCAL
RAMP
ROADTYPE
0.1236
0.1307
0.3307
0.0000
S*ROADTYPE 0.1176 0.1203 0.6233 0.0000
VEH TYPE 0.5942 0.5693 0.8984 0.3961
S*VEH TYPE 0.0641 0.0699 0.0241 0.9560
CLEANTO ....
S*CLEANTO ....
VEH TYPE 0.8787 0.8821 0.5511 0.6093
CLEANTO ....
VEH TYPE 0.3701 0.4075 0.1471 0.6942
CLEANTO ....
Final M6.SPD.002
98
June 2001
-------
EMIT NORMAL
ROAD
ART/FWY
LOCAL
RAMP
FACTOR
ROADWAY TYPE
VEHICLE CLASS
STANDARD
VEHICLE CLASS
STANDARD
VEHICLE CLASS
STANDARD
P
THC
PROB
0.0001
0.0000
0.0000
0 . 1017
0.0000
0.2047
0.0000
NMHC
PROB
0.0000
0.0640
0.0000
0.5022
0.0000
0.6109
0.0000
CO
PROB
0.0405
0.0000
0.0000
0.1380
0.0000
0.0213
0.0000
NOX
PROB
0.0000
0.0000
0.0000
0.0408
0.0000
0.0035
0.0000
EMIT HIGH
ROAD
ART/FWY
FACTOR
ROADWAY TYPE
VEHICLE CLASS
STANDARD
P
THC
PROB
0.9736
0.0667
NMHC
PROB
0.9570
0.0873
CO
PROB
0.0151
0.0004
NOX
PROB
0.0201
0 . 1444
Note: these probabilities are for tests of factor main effects, not
interactions with speed.
Final M6.SPD.002
99
June 2001
-------
EMIT NORMAL - CLEAN TIER 0 CLASS
THC NMHC CO NOX
ROAD FACTOR
ART/FWY
LOCAL
RAMP
ROADTYPE 0 . C
)001 0.0000
0.0405
0.0000
VEH TYPE 0.0000 0.0186 0.0000 0.0000
STANDARD 0.0000 0.0000 0.0000 0.0000
VEH TYPE 0.0572 0.4049 0.1660 0.0184
STANDARD 0.0000 0.0000 0.0000 0.0028
VEH TYPE 0.1570 0.5501 0.0201 0.0009
STANDARD 0.0003 0.0007
0.0002
0.0000
EMIT NORMAL - ACTUAL TIER CLASS
THC NMHC CO NOX
ROAD FACTOR
ART/FWY
LOCAL
RAMP
ROADTYPE 0 . C
)001 0.0000
0 .0405
0 .0000
VEH_TYPE 0.0000 0.0686 0.0001 0.0000
STANDARD 0.0000 0.0000 0.0000 0.0000
VEH_TYPE 0.0830 0.5008 0.4038 0.0124
STANDARD 0.0002 0.0001 0.0000 0.0024
VEH_TYPE 0.2922 0.7707 0.0443 0.0018
STANDARD 0.0013 0.0002
0 .0001
0 .0010
EMIT HIGH
THC NMHC
ROAD FACTOR
ART/FWY
LOCAL
RAMP
ROADTYPE 0 . S
3736 0.9570
CO
0 .0151
NOX
0 .0201
VEH_TYPE 0.0667 0.0873 0.0004 0.1444
STANDARD
VEH_TYPE 0.8787 0.8821 0.5511 0.6093
STANDARD
VEH_TYPE 0.3701 0.4075 0.1471 0.6942
STANDARD
GLM P-VALUES FOR MODELS WITH NO SPEED INTERACTIONS (FROM FACVEHA.SAS)
Final M6.SPD.002
100
June 2001
-------
EMIT NORMAL - CLEAN TIER 0 CLASS
ROAD
ART/FWY
FACTOR
S*ROADTYPE
P
THC
NMHC CO NOX
PROB PROB PROB PROB
0 .0354
0 .0440 0 .(
)020 0.0000
S*VEH_TYPE 0.1322 0.1476 0.8361 0.5608
S*STANDARD 0.0002
0.0001 0.0576 0.6491
EMIT HIGH - CLEAN TIER 0 CLASS
ROAD
ART/FWY
FACTOR
S*ROADTYPE
P
THC
NMHC CO NOX
PROB PROB PROB PROB
0 .1176
0.1203 0.Ť
5233 0.0000
S*VEH_TYPE 0.0641 0.0699 0.0241 0.9560
S* STANDARD ....
EMIT NORMAL - CLEAN ACTUAL TIER CLASS
ROAD
ART/FWY
FACTOR
S*ROADTYPE
P
THC
NMHC CO NOX
PROB PROB PROB PROB
0 .0354
0 .0440 0 .(
)020 0.0000
S*VEH_TYPE 0.1754 0.1802 0.8680 0.5723
S*STANDARD 0.0024
EMIT HIGH - CLEAN ACTUAL
ROAD
ART/FWY
FACTOR
S*ROADTYPE
0.0020 0.0560 0.0151
TIER CLASS
P
THC
NMHC CO NOX
PROB PROB PROB PROB
0 .1176
0.1203 0.Ť
5233 0.0000
S*VEH_TYPE 0.0641 0.0699 0.0241 0.9560
S* STANDARD ....
GLM P-VALUES FOR MODELS WITH NO SPEED INTERACTIONS (FROM FACVEHA.SAS)
Final M6.SPD.002
101
June 2001
-------
EMIT NORMAL - CLEAN TIER 0 CLASS
ROAD
ART/FWY
FACTOR
ROADTYPE
VEH_TYPE
STANDARD
P
THC
PROB
0 .0046
0.0004
0 .0000
NMHC
PROB
0 .0050
0.0243
0 .0000
CO
PROB
0 .0006
0.0062
0 .0000
NOX
PROB
0 .0000
0.0026
0 .0000
EMIT HIGH - CLEAN TIER 0 CLASS
ROAD
ART/FWY
FACTOR
ROADTYPE
VEH_TYPE
STANDARD
P
THC
PROB
0 .1236
0.5942
NMHC
PROB
0 .1307
0.5693
CO
PROB
0 .3307
0.8984
NOX
PROB
0 .0000
0.3961
EMIT NORMAL - CLEAN ACTUAL TIER CLASS
ROAD
ART/FWY
FACTOR
ROADTYPE
VEH_TYPE
STANDARD
P
THC
PROB
0 .0046
0.0016
0 .0000
NMHC
PROB
0 .0050
0.0404
0 .0000
CO
PROB
0 .0006
0.0031
0 .0000
NOX
PROB
0 .0000
0.0012
0 .0000
EMIT HIGH - CLEAN ACTUAL TIER CLASS
THC
PROB
I
NMHC
PROB
D
CO
PROB
NOX
PROB
ROAD
ART/FWY
FACTOR
ROADTYPE
VEH_TYPE
STANDARD
0 .1236
0.5942
0 .1307
0.5693
0 .3307
0.8984
0 .0000
0.3961
GLM P-VALUES FOR MODELS WITH NO SPEED INTERACTIONS (FROM FACVEHA.SAS)
Final M6.SPD.002
102
June 2001
-------
Regression statistics for CO Off-Cycle Emissions Analysis (Normal emitters)
Variables Entered/Removed(a,b)
Model
1
2
Variables
Entered
LA4COSQR
LA4CO
Variables
Removed
Method
Stepwise (Criteria: Probability-of-F-to-enter <= .050, Probability-of-F-to-remove >= .100).
Stepwise (Criteria: Probability-of-F-to-enter <= .050, Probability-of-F-to-remove >= .100).
a Dependent Variable: CO DELT
b Linear Regression through the Origin
Model Summary
Model
1
2
R
.678(b)
.889(c)
R Square(a)
.459
.791
Adjusted R Square
.450
.784
Std. Error of the
Estimate
3.0755
1.9279
a For regression through the origin (the no-intercept model), R Square measures the proportion of the variability in the dependent variable
about the origin explained by regression. This CANNOT be compared to R Square for models which include an intercept.
b Predictors: LA4COSQR
c Predictors: LA4COSQR, LA4CO
ANOVA(d,e)
Model
1
Regression
Residual
Total
Sum of Squares
497.694
586.440
1084.134(b)
df
1
62
63
Mean Square
497.694
9.459
F
52.618
Sig.
.OOO(a)
2
Regression
Residual
Total
857.399
226.735
1084.134(b)
2
61
63
428.700
3.717
115.336
.OOO(c)
a Predictors: LA4COSQR
b This total sum of squares is not corrected for the constant because the constant is zero for regression through the origin.
c Predictors: LA4COSQR, LA4CO
d Dependent Variable: CO DELT
e Linear Regression through the Origin
Coefflclents(a,b)
Model
1
2
LA4COSQR
LA4COSQR
LA4CO
Unstandardized Coefficients
B
-3.277E-02
-7.638E-02
.984
Std. Error
.005
.005
.100
Standardized Coefficients
Beta
-.678
-1.579
1.070
t
-7.254
-14.520
9.837
Sig.
.000
.000
.000
95% Confidence Interval for B
Lower Bound
-.042
-.087
.784
Upper Bound
-.024
-.066
1.184
a Dependent Variable: CO DELT
b Linear Regression through the Origin
Final M6.SPD.002
103
June 2001
-------
Regression statistics for CO Off-Cycle Emissions Analysis (Normal emitters)
Excluded Variables(c,d)
Model
1
LA4CO
FINJ
VTYP
MODEL YR
Beta In
1.070(a)
.500(a)
.416(a)
.470(a)
t
9.837
6.731
5.120
6.206
Sig.
.000
.000
.000
.000
Partial Correlation
.783
.653
.548
.622
Collinearity Statistics
Tolerance
.290
.923
.939
.946
2
FINJ
VTYP
MODEL YR
.064(b)
-.035(b)
.037(b)
.638
-.397
.395
.526
.693
.694
.082
-.051
.051
.341
.442
.388
a Predictors in the Model: LA4COSQR
b Predictors in the Model: LA4COSQR, LA4CO
c Dependent Variable: CO DELT
d Linear Regression through the Origin
Regression statistics for CO Off-Cycle Emissions Analysis (High Emitters)
Case Processing Summary
CO DELT
LA4CO
HIGHCO
Normal
High
Normal
High
Cases
Valid
N
63
22
63
22
Percent
100.0%
100.0%
100.0%
100.0%
Missing
N
0
0
0
0
Percent
.0%
.0%
.0%
.0%
Total
N
63
22
63
22
Percent
100.0%
100.0%
100.0%
100.0%
Final M6.SPD.002
104
June 2001
-------
Regression statistics for CO Off-Cycle Emissions Analysis (High Emitters)
Descriptlves
CO DELT
LA4CO
HIGHCO
Normal
High
Normal
High
Mean
95% Confidence Interval for Mean Lower Bound
Upper Bound
5% Trimmed Mean
Median
Variance
Std. Deviation
Minimum
Maximum
Range
Interquartile Range
Skewness
Kurtosis
Mean
95% Confidence Interval for Mean Lower Bound
Upper Bound
5% Trimmed Mean
Median
Variance
Std. Deviation
Minimum
Maximum
Range
Interquartile Range
Skewness
Kurtosis
Mean
95% Confidence Interval for Mean Lower Bound
Upper Bound
5% Trimmed Mean
Median
Variance
Std. Deviation
Minimum
Maximum
Range
Interquartile Range
Skewness
Kurtosis
Mean
95% Confidence Interval for Mean Lower Bound
Upper Bound
5% Trimmed Mean
Median
Variance
Std. Deviation
Minimum
Maximum
Range
Interquartile Range
Skewness
Kurtosis
Statistic
1.1968
.1884
2.2051
1.3587
1.0782
16.031
4.0038
-25.29
12.42
37.71
2.1735
-4.346
32.170
-1.9790
-17.1576
13.1996
1.7699
5.0273
1171.982
34.2342
-128.76
51.06
179.82
24.5018
-2.481
9.106
2.73210603
1.82125397
3.64295809
2.25774861
1.98924000
13.080
3.61669191
.019160
25.774990
25.755830
2.76831000
4.438
26.534
57.94719000
29.74081545
86.15356455
51.19508025
38.69987500
4047.173
63.61739899
1.729990
239.809000
238.079010
73.66766000
1.631
2.391
Std. Error
.5044
.302
.595
7.2988
.491
.953
.45566035
.302
.595
13.56327504
.491
.953
Final M6.SPD.002
105
June 2001
-------
Regression statistics for HC Off-Cycle Emissions Analysis
Variables Entered/Removed(b,c)
Model
1
Variables Entered
LA4HCSQR, LA4HC(a)
Variables Removed
Method
Enter
a All requested variables entered.
b Dependent Variable: THC DELT
c Linear Regression through the Origin
Model Summary
Model
1
R
.255(b)
R Square(a)
.065
Adjusted R Square
.043
Std. Error of the
Estimate
1.1960
a For regression through the origin (the no-intercept model), R Square measures the proportion of the variability in the dependent variable
about the origin explained by regression. This CANNOT be compared to R Square for models which include an intercept.
b Predictors: LA4HCSQR, LA4HC
ANOVA(c,d)
Model
1
Regression
Residual
Total
Sum of Squares
8.268
118.715
126.982(b)
df
2
83
85
Mean Square
4.134
1.430
F
2.890
Sig.
.061 (a)
a Predictors: LA4HCSQR, LA4HC
b This total sum of squares is not corrected for the constant because the constant is zero for regression through the origin.
c Dependent Variable: THC DELT
d Linear Regression through the Origin
Coefflclents(a,b)
Model
1
LA4HC
LA4HCSQR
Unstandardized Coefficients
B
.305
-2.492E-02
Std. Error
.134
.014
Standardized Coefficients
Beta
.571
-.437
t
2.283
-1.748
Sig.
.025
.084
95% Confidence Interval for B
Lower Bound
.039
-.053
Upper Bound
.570
.003
a Dependent Variable: THC DELT
b Linear Regression through the Origin
Final M6.SPD.002
106
June 2001
-------
Regression statistics for NOx Off-Cycle Emissions Analysis
Variables Entered/Removed(b,c)
Model
1
Variables Entered
LA4NOSQR, LA4NOX(a)
Variables Removed
Method
Enter
a All requested variables entered.
b Dependent Variable: NOX DELT
c Linear Regression through the Origin
Model Summary
Model
1
R
.615(b)
R Square(a)
.378
Adjusted R Square
.363
Std. Error of the Estimate
.3521
a For regression through the origin (the no-intercept model), R Square measures the proportion of the variability in the dependent variable
about the origin explained by regression. This CANNOT be compared to R Square for models which include an intercept.
b Predictors: LA4NOSQR, LA4NOX
ANOVA(c,d)
Model
1
Regression
Residual
Total
Sum of Squares
6.253
10.289
16.542(b)
df
9
83
85
Mean Square
3.126
.124
F
25.219
Sig.
.OOO(a)
a Predictors: LA4NOSQR, LA4NOX
b This total sum of squares is not corrected for the constant because the constant is zero for regression through the origin.
c Dependent Variable: NOX DELT
d Linear Regression through the Origin
Coefflclents(a,b)
Model
1
LA4NOX
LA4NOSQR
Unstandardized Coefficients
B
.332
-4.745E-02
Std. Error
.066
.018
Standardized
Coefficients
Beta
1.107
-.582
t
4.998
-2.627
Sig.
.000
.010
95% Confidence Interval for B
Lower Bound
.200
-.083
Upper Bound
.464
-.012
a Dependent Variable: NOX DELT
b Linear Regression through the Origin
Final M6.SPD.002
107
June 2001
-------
Appendix B : Example
Example Application of Speed Adjustment to Exhaust Emissions
The following description is meant as an example of how the basic exhaust emission rates
estimated by MOBILE6 will be adjusted for the effects of average speed and roadway type. The
example will show how the various parts of the overall emission estimate are weighted together.
It is beyond the scope of this document to explain fully the derivation of the basic exhaust
emission estimates or the weighting factors. The derivation of these distributions are described
in other documents. It is also not the intent of this example to reveal the values for emissions or
weighting factors that are used in MOBILE6. All of the values shown in this example should,
therefore, be considered as draft and may not match values shown in other documents. This
should not detract from the value of this example in showing the process of how the basic
emission rates are adjusted for speed.
Basic Emission Rates
For each scenario, MOBILE6 will calculate a basic exhaust emission rate (BER) for two
emission levels (high and normal) for each pollutant for each model year for each vehicle class.
The basic unit for the BER is the hot running LA4 (with an average speed of 19.6 mph) at
standard operating conditions (i.e., temperature, humidity, etc.). The effect of engine starts on
emissions is calculated separately and is not adjusted for the effects of average speed.
MOBILE6 calculates the emissions for each hour of the day, so the first step is to adjust
the BER for the conditions that affect exhaust emissions. For example, the temperature at 6 a.m.
will be different than the temperature at 1 p.m., so the BER at 6 a.m. will not be the same as the
BER at 1 p.m. after adjustment for temperature. Some adjustments (such as the effects of fuel
sulfur content) will not vary by time of day. Ultimately, there will be 24 values, one for each
hour of the day calculated from the same BER, adjusted for hourly conditions. There will be two
sets of adjusted BER values, one for normal emissions and one for high emitters.
Example Basic Emission Rates
For this example, we will follow the calculation of NOx emissions from a 1990 model
year passenger car. The calculation would be similar for the other pollutants and other vehicle
classes. This example will not fabricate values for all hours. The calculations will be similar in
all hours, so a single hour example is all that should be required. So, for a given hour, the NOx
emissions (BERs) for our vehicles will be assumed to be:
0.65 g/mi for normal emitters
2.10 g/mi for high emitters
Final M6.SPD.002 108 June 2001
-------
After adjustment, these values must be weighted together by their occurrence in the fleet.
The number of high emitters will depend on many things (i.e., age, I/M programs, OBD, etc.),
but for our example, we will assume that high emitters are 10% of 1990 model year passenger
cars in this scenario.
Freeway Ramps and Local Roadways
There are four basic roadway types; freeways, arterial/collectors, freeway ramps and local
roadways. The freeway ramps and local roadways can be determined directly from the BER,
since they do not vary with average speed. The freeway ramp and local roadway emissions are a
function of the BER (see Table 13). The NOx BERs we will use (described above) are in grams
per mile units and must be converted to grams per hour. The average speed of the hot running
LA4 is 19.6 miles per hour. For normal emitters, 0.65 grams per mile times 19.6 miles per hour
is 12.74 grams per hour. For high emitters, 2.10 grams per mile times 19.6 miles per hour is
41.16 grams per hour. Using the equation shown in Table 13, the freeway ramp and local
roadway emissions in grams per hour are:
Normal Ramp = 5.353 + 2.863*(12.74) - 0.0101*(12.74)2 = 40.19 g/hr
Normal Local = 1.870 + 0.701 *(12.74) + 0.000609*(12.74)2 = 10.90 g/hr
High Ramp = 5.353 + 2.863*(41.16) - 0.0101*(41.16)2 = 106.08 g/hr
High Local = 1.870 + 0.701 *(41.16) + 0.000609*(41.16)2 = 31.75 g/hr
The results will be weighted using VMT and must be converted to grams per mile units.
The freeway ramp cycle has an average speed of 34.6 miles per hour and the local roadway cycle
has an average speed of 12.9 miles per hour.
Normal Ramp = (40.19 g/hr) / 34.6 mph =1.16 g/mi
Normal Local = (10.90 g/hr) / 12.9 mph = 0.84 g/mi
High Ramp = (106.08 g/hr) / 34.6 mph = 3.07 g/mi
High Local = (31.75 g/hr) / 12.9 mph = 2.46 g/mi
Since we have assumed that 10% of the vehicles are high emitters, we can now weight the
normal and high emitter results to give a complete freeway ramp and local roadway estimate for
the 1990 model year in this hour.
Freeway Ramp = 1.16* 0.90 + 3.07 * 0.10 = 1.35 g/mi
Local Roadway = 0.84 * 0.90 + 2.46 * 0.10 = 1.01 g/mi
Each hour will have its own basic exhaust emission rate. Since the Freeway Ramp and
Local Roadway emission levels depend on the basic exhaust emission rate, a separate calculation
will be done for each hour of the day.
Final M6.SPD.002 109 June 2001
-------
Emission Offset
The emission offset (EO) represents the difference between the LA4-based BER and
freeway emissions at 19.6 miles per hour. The values for the EO are shown in Table 14. Since
the BER values lie between the LA4 values (0.591 and 3.245 g/mi) shown in Table 14, the EO
must be calculated using interpolation.
Normal EO = 0.121 + ((0.008-0.121)/(3.245-0.591))*(0.65-0.591) = 0.12 g/mi
High EO = 0.121 + ((0.008-0.121)/(3.245-0.591))*(2.10-0.591) = 0.06 g/mi
An additional emission offset is used for arterial/collector roadways, however this offset
depends on average speed and emissions. These are shown in Table 15. The ratio of the freeway
emission level at each speed plus the arterial/collector offset for that speed, divided by the
freeway emission level at 19.6 miles per hour is the arterial/collector speed correction factor.
These are shown in Table 17.
Freeway Emissions
Freeway emissions depend on average speed. For each hour of the day, MOBILE6 has a
default distribution of average speeds for freeways. Users will be able to enter local distributions
of freeway average speeds. This is not the same as a distribution of speeds on a particular
freeway.
The MOBILE6 default distribution of average speeds for freeways assumes that there are
many freeways in the area and the distribution represents the average speeds observed from the
different freeways at that hour. The cycles used to develop the speed correction factors each
contain the entire range of vehicle speeds on freeways grouped by ranges of observed congestion.
So, changing speed in the MOBILE6 model is changing the average speed of the combination of
all vehicles on freeways. MOBILE6 does not effectively model the effect of average speed on
individual vehicles or small groups of vehicles within a single freeway section. If you wish to
model a specific freeway, you would want to reduce the default distribution down to a single,
average speed for the freeway of interest.
In each hour, MOBILE6 will calculate values for each average speed "bin" from 5 to 65
mph in 5 mph increments and for 2.5 mph (14 speed bins) by applying the speed correction
factors from Table 16 to the base freeway emission level at 19.6 mph. The base freeway
emission level is simply the sum of the BER and the adjusted emission offset (EO).
Normal Base Freeway Emission at 19.6 mph = 0.65 + 0.12 = 0.77 g/mi
High Base Freeway Emission at 19.6 mph = 2.10 + 0.06 = 2.16 g/mi
There are three sets of speed correction factors in Table 16, one for each of three emission
levels. Both the Normal and High base freeway emission levels we have calculated lie between
Final M6.SPD.002 110 June 2001
-------
the Level 2 and Level 3 emission levels, shown in Table 16. So the speed correction factor will
be interpolated between the values for Level 2 and Level 3 in Table 16. However, these speed
correction factors do not apply below 7.1 mph. MOBILE6 will use the MOBILES speed
correction factors (See Table 1.6B in AP-42) for speeds below 7.1 mph. For our example, the
NOx speed correction factors for the 1990 model year have A and B coefficients of 1.456 and
0.926 respectively, where the form of the equation is A/speed + B, resulting in the following
speed correction factors:
SCF for 2.5 mph = (1.456/2.5) + 0.926 = 1.51
SCF for 5.0 mph = (1.456/5.0) + 0.926 = 1.22
SCF for 7.1 mph = (1.456/7.1) + 0.926 = 1.13
The MOBILES speed correction factor at 7.1 mph (1.13) was applied to all emission
levels in MOBILES. The MOBILES speed correction factors will be adjusted to match the speed
correction factors in Table 16 for NOx at 7.1 mph of 2.26, 1.81 and 1.50 for emission levels 1, 2
and 3 respectively by adding the difference to each value.
Level 1 SCF for 2.5 mph =1.51+ (2.26- 1.13) = 2.63
Level 1 SCF for 5.0 mph = 1.22 + (2.26 - 1.13) = 2.34
Level 2 SCF for 2.5 mph =1.51+ (1.81 - 1.13) = 2.19
Level 2 SCF for 5.0 mph =1.22+ (1.81 - 1.13)= 1.90
Levels SCF for 2.5 mph = 1.51 + (1.50 - 1.13)= 1.87
Levels SCF for 5.0 mph = 1.22 + (1.50 - 1.13)= 1.58
Using the average emissions for each speed correction factor emission level (from Table
14) of 0.712 and 3.253 g/mi NOx for Level 2 and Level 3 respectively and the predicted base
freeway emission rates of 0.77 and 2.16 g/mi for Normals and High categories, weighting factors
can be derived for interpolating between the speed correction factors. The sum of the two
weighting factors will equal 1.
Normal Level 2 Weighting = (3.253 - 0.77)/(3.253 - 0.712) = 0.978
Normal Level 3 Weighting = (1.0 - 0.978) = 0.022
High Level 2 Weighting = (3.253 - 2.16)/(3.253 - 0.712) = 0.431
High Level 3 Weighting = (1.0 - 0.431) = 0.569
These weighting factors are used to combine the Level 2 and Level 3 speed correction
factors for the calculated base freeway emission case. A new weighted speed correction factor is
calculated for each of the fourteen speed bins for Normals and Highs. For example, the 10 mph
speed bin speed correction factors (using values from Table 16) would be:
Final M6.SPD.002 111 June 2001
-------
Normal SCF for 10 mph = 0.978 * 1.28 + 0.022 * 1.18 = 1.28
High SCF for 10 mph = 0.431 * 1.28 + 0.569* 1.18= 1.22
These speed correction factors are applied to the predicted base freeway emission rates to
determine speed corrected emission rates for each speed bin. For example the speed corrected
emission rates for the 10 mph speed bin would be:
Normal emission level for 10 mph = 1.28 * 0.77 = 0.99 g/mi
High emission level for 10 mph = 1.22 * 2.16 = 2.64 g/mi
Each hour has a default VMT distribution of average freeway speeds that correspond to
these speed bins. The emission rates for each of the bins can be weighted, using this VMT
distribution, to give a composite freeway emission rate. This weighting is repeated for normal
and high emitters, and the two emitter groups can be combined to give an overall freeway NOx
emission rate for 1990 model year vehicles for that hour of the day.
Arterial/Collector Emissions
The arterial/collector speed correction factors shown in Table 17 are applied to the base
freeway emission rate calculated for the freeway emission levels. Since the three emission level
groups are identical for arterial/collector roadways and freeways, the same weighting factors are
used to interpolate between the speed correction factors. For example, the 10 mph speed bin
speed correction factors (using values from Table 17) would be:
Normal SCF for 10 mph = 0.978 * 1.52 + 0.022 * 1.31 = 1.52
High SCF for 10 mph = 0.431 * 1.52 + 0.569* 1.31 = 1.40
These speed correction factors are applied to the base freeway emission levels to
determine emission levels for each speed bin. For example the emission levels for the 10 mph
speed bin would be:
Normal emission level for 10 mph = 1.52 * 0.77 = 1.17 g/mi
High emission level for 10 mph = 1.40 * 2.16 = 3.02 g/mi
Since the speed correction factors for arterial/collectors (shown in Table 17) converge
with freeway speed correction factors (shown in Table 16) at higher speeds and below 7.1 mph,
the emission rate for arterial/collectors and freeways will be the same for some speed bins. All of
the speed bins are combined, weighted by the fraction of VMT in that speed bin for that hour.
The composite arterial/collector emissions for Normals and Highs are combined weighted by
their proportions in the fleet for that model year.
Final M6.SPD.002 112 June 2001
-------
Area-wide Emissions
Once a fleetwide (combined Normal and High), hourly (combined speed bins) estimate is
available for each roadway type (freeway, arterial/collector, freeway ramp and local roadway),
these estimates can be combined in a variety of ways, depending on the needs of the user. If an
area-wide, hourly result is needed, the results for the four roadway types can be combined,
weighted by the fraction of VMT for each roadway for that hour. An area-wide daily result can
be obtained by combining the hourly results weighted by the VMT fraction for each hour.
Although there are default values for the fraction of VMT for each roadway and the VMT
fraction for each hour, users may substitute their own values.
Composite Engine Start and Running Emissions
The emission rates addressed in this document do not contain the effects of engine starts.
The effect of engine start on emissions is calculated separately and is calculated in units of grams
per engine start. These emission effects resulting from engine starts are not determined by
roadway type and do not depend on average trip speed. They can, however, be combined with
the running emissions to give an overall exhaust emission estimate.
Since the MOBILE6 model does not include a distribution of the effects of engine start on
emissions by roadway type, the combination of the effects of engine start and running emissions
is best done on area-wide (combined roadway) emission results. This can be done on an hourly
or daily basis.
MOBILE6 has an estimate of the average daily vehicle miles traveled (VMT) for each
model year in a given calendar year and a distribution of that average VMT over the day by hour.
MOBILE6 also has an estimate for the number of engine starts per day and the distribution of
those starts over the day by hour. For a given hour, the grams due to engine starts in that hour are
calculated as:
Grams / Engine Start * Fraction of Starts in the Hour * Number of Starts / Day
This value can be converted to grams per mile by determination of the average number of
miles traveled by vehicles in that hour:
Hourly VMT = Daily VMT * Fraction of VMT in the Hour
Once the effect of engine start on emissions is converted to grams per mile, it can be
added directly to the area-wide emission estimate for that hour.
Total Exhaust = Engine Start / Hourly VMT + Area-wide Emissions for the Hour
Final M6.SPD.002 113 June 2001
-------
Similarly, a daily total exhaust emission rate can be calculated. Although there are
default values for the number of daily engine starts, the fraction of engine start in each hour, the
daily VMT and the fraction of VMT in each hour, users may substitute their own values.
A calculation is done for each model year of each vehicle class. These values are
weighted using travel fractions (as is done in MOBILES) to calculate area-wide, daily emission
rates for highway mobile sources.
FTP Emissions
The Federal Test Procedure (FTP) is a special case of vehicle driving. It can be simulated
in MOBILE6 by careful choice of weighting factors for engine start soak time, vehicle miles
traveled and roadway types. Since this case will be of special interest for comparison of
MOBILE6 emission rates to Federal certification standards, we plan to build in the appropriate
weighting factors so that calculation of FTP emission estimates using MOBILE6 can be done
simply and consistently.
Final M6.SPD.002 114 June 2001
-------
Appendix C : Response to Comments
Comment numbers refer to EPA indexing of comments received for easy reference. All
EPA responses are shown in italics.
Lance Freeman, October 12,1999:
I think there may be a fallacy, started by misinterpreting previous speed curves, that lower
speeds are correlated to higher emissions. But I'm guessing that those speed curves were heavily
weighted with emissions from the start mode.
All of the speed correction driving cycles were done *without* engine starts, so engine
starts are not a factor in the effect of low average speeds on emissions. However, since
the speed correction factors are applied to emission levels in grams per mile, the
corrections are very large for low speeds. This makes sense, since at low speeds you get
very few miles (the denominator) for each gram of emissions generated. For example, at
idle (average speed=0) the speed correction factor is infinite. So, at low speeds it often
makes more sense to examine the emissions in terms of grams per unit of time (usually
grams per hour). In those units, the effect of lower speeds on emissions is much different.
EPA uses units of grams per mile for travel inMOBILE6, since we consider the miles
traveled as the appropriate unit of work (i.e., purpose of travel) as opposed to travel as a
way to spend your time. Mathematically, however, they should be equivalent.
Sam Long. IL EPA (Comment #9) March 13.1997:
Under "Transportation Models" section, the paragraph is not clear. It should be made
clearer that 35 mph on a local street or collector (or even some arterials) would be a good speed,
but would be slow and represent very congested conditions on a freeway. (I presume this is what
was meant.)
Some additional text was added to clarify this issue.
HPMS facility types do not include specifically include on-and off-ramps, but quite often
metropolitan planning organizations (MPOs) list [some] on-and off-ramps on their network.
Data of this sort may or may not be readily available.
Guidance will be available to help areas determine appropriate inputs for MOBILE6.
Default values will be available for issues such as the fraction of freeway VMTthat
occurs on freeway ramps.
What is a "micro trip"? (One one thousandth of a mile is 5ft!)
Final M6.SPD.002 115 June 2001
-------
A "micro trip" is any portion of a trip between the time the engine is started and the
engine is shut down which begins and ends with a period of idling. One example would
be the driving from one signal to the next. Separating driving into micro trips allows an
unambiguous dividing of long trips into smaller parts that can be used in cycle
development. More information about micro trips are in the report M6.SPD. 001 listed in
the references.
What is meant by "Average Speed per %VMT" or indeed %VMT? In as much as area-
wide estimates (of emissions) are what emissions inventories and emission reduction strategies
are all about, I suggest that much effort go into [non-attainment] area-wide estimates. Similarly
for statewide inventories, which are needed for some circumstances.
Each roadway link has a distance (in miles) and an average speed (in miles per hour).
Average speed is the length of the link (distance) divided by the time (in hours) it takes for
vehicles to drive from one end of the link to the other. If a set of average speed bins are
created (i. e., every 5 mph) and the link distances are put into the bins depending on the
average speed, this will create a distribution of miles traveled (VMT) by average speed.
"Signal density"? What's that? The number of traffic signals in a given area? If that's the
case, I imagine each transportation model zone would have a different signal density. What about
stop signs? Are they traffic signals within the meaning of the act, or are we just talking about
traffic lights? Four-way stops are different from one- or two-way stops. How and where are
users to obtain such data? The MPOs presumably; but not all NAAs are completely covered by a
transportation network. Even where a comprehensive network exists, why, there are thousands
of zones in the Chicago area, for example.
EPA had originally discussed ways to account for the number of traffic signals on
roadways, but such plans were dropped from the final version of the model. Signal
density would allow users to better account for roadways with similar average speeds,
but with different driving behavior.
William Benjev. HPCC EPA (Comment #10) March 19.1997:
Moving to facility-type output with short-term (hourly instead of daily?) mobile emission
factor outputs for specific road types would in a sense be more consistent with hourly time scale
of most episodic air quality modeling. However, because the VMT data needed to use the
emission factors for specific road segments usually does not exist outside of a few urban areas, it
will be difficult to apply the Mobile 6 emission factors on a regional basis. Consequently think
that your efforts to provide a weighted running emission factor for all roadway types in addition
to the hourly facility-based factor information is crucial.
Users who do not have specific vehicle activity data will still be able to run MOBILE6
using the national average default values. Guidance will be needed to specify the local
Final M6.SPD.002 116 June 2001
-------
data that must be provided for specific modeling situations.
Given that the input and output file formats are likely to change appreciably, it would a
very significant help to regional modelers if the input and output data could be tagged with
geographic identifier information. In other words, since the input options affecting the emission
factors vary geographically, the output files vary geographically. Ideally, the mobile model could
be set to generate a set of geographically-specific emission factors for a region defined by the
user. Currently ,it is tedious and resource-intensive to sequentially run the mobile model
separately for all the different areas included in a regional air quality model run and then in turn
perform sequential air quality model runs or manually tag many mobile output files for different
geographic areas before a air quality model run. If the Mobile 6 input and output files were
tagged (or at least had the option of allowing the user to easily tag them) with geographic
identifiers, we could combine the output files and read them by identifier. Areas with only
county-level VMT data available would be identified with state and county-level FIPS codes and
would use the weighted average running emission factors. Areas with road link specific VMT
data would be identified by state and county FIPS codes plus latitude and longitude data for the
road link nodes (end points) and could use the facility-type emission factors.
MOBILE6 has an option for "database " output, where the emission values are written to
an ASCII file formatted for importing to database software. Since the output includes run
and scenario numbers, the database software can be used to easily "tag" the emission
results for linking with geographic information.
Harold Brazil. SEMCOG. MI (Comment #16) April 4.1997:
Is this "short time period emission factors" for a one hour period or peak
hour period? Would this be used for Photochemical Modeling purposes?
MOBILE6 will only provide information as hourly or daily, with all hours aggregated.
The hourly results can be aggregated by the user into other useful time periods, such as
peak hours, for use in photochemical modeling.
Celia Shih. NY PEP (Comment #17) April 7.1997:
Were there any new data collected under various speed cycles since MOBILES? Will
there be any update on the "regular" speed correction factors in MOBILE6?
This report represents the only new data collection since MOBILES specifically to
address speed correction factors. New driving cycles were developed from the new data
and these new driving cycles were used to develop the speed correction factors used in
MOBILE6.
John Walsh. EPA #2. NY (Comment #18) April 3.1997:
Final M6.SPD.002 117 June 2001
-------
How many facility types are there? Should there be a switch between facility-specific
emissions and more general area-specific emissions?
MOBILE6 will always calculate emissions for four facility types, but will automatically
aggregate the results for the descriptive output. These results are also available in the
database output using the AGGREGATED OUTPUT command. Facility specific
emission rates are only available using the database output.
Sam Long. Illinois EPA (Comment #21) April 10.1997:
The necessary [non-default] inputs may not be available to all users. Also, daily emission
rates will still be needed, and by no means everyone has link-based or hourly information. Link-
based information is not very adaptable to forecasting ROP or conformity. If you come up with
typical speeds for various facility types (at various levels of service), will these speeds be
published and be acceptable as MOBILE inputs for inventory and other purposes?
Although MOBILE6 will have national average speed distribution estimates, this is likely
one area where EPA guidance may require that local information be used, since driving
behavior has a significant effect on emissions and overall driving behavior (speed)
distributions will vary from area to area because of different roadway types available.
If link-based speeds (free-flow or congested) are available, it is possible to estimate an
average or representative speed for each functional class on the network. I did so in the '90
inventory, rounding off to the nearest 5 mph, and used the results in off-network areas. However,
the arithmetic average speed will differ from the average speed weighted by link-length, and both
will differ from the median and modes of the speeds. You should specify which of these speeds
is to be used as representative, and how they are to be calculated, if you want users to derive
them from link-based data.
Guidance on how to calculate average speed VMT distributions for MOBILE6 will be
needed. Briefly, each roadway link has a distance (in miles) and an average speed (in
miles per hour). Average speed is the length of the link (distance) divided by the time (in
hours) it takes for vehicles to drive from one end of the link to the other. If a set of
average speed bins are created (i.e., every 5 mph) and the link distances are put into the
bins depending on the average speed, this will create a distribution of miles traveled
(VMT) by average speed.
What facility types do you have in mind? The HPMS facility types loom large in USEPA
and FHWA planning, but those twelve types do not include such things as ramps and bridges.
On the other hand, some transportation model networks do have ramps and bridges as facility
types, but those and the other facility types in such networks may notoften do notmatch the
HPMS facility types. Do you have suggestions for equating various non-HPMS functional
classes to HPMS classes, apart from the methods appearing in Sections 2 and 3 of publication
Final M6.SPD.002 118 June 2001
-------
MD?
The need to distinguish between roadway types will be a new feature ofMOB!LE6.
However, there are only four categories of roadway types in MOBILE6. Guidance will
be needed to assist users in determining which of these four roadway types should be
used to model each particular roadway.
Congestion-level weighting factors: In-use level-of-service data are not, in my
experience, easy to come by. "Congested" and "free" speeds in transportation model outputs
from CATS represent two different levels of service, of course, but offhand I don't remember
just which ones they are; I don't think they were specified to me. I looked at some 1990 traffic-
by-hour data from IDOT for several continuous-traffic-count stations in Illinois, and estimated
that about 75% of VMT in the Chicago area occurs under more or less congested conditions, and
25% under free-flow conditions, and weighted total emissions accordingly, using the modeled
"congested" and "free" speeds as MOBILE inputs. My proportion above may be somewhat of an
overestimate; it may be closer to 60% congested/40% free, or even down to 50/50; but the
congested-free proportion, as long as it's within reasonable limits, doesn't affect the final
emission estimates all that much, as I noted in our '90 inventory document.
MOBILE6 will use average speed as a surrogate for roadway congestion. The driving
cycles were developed by grouping trips by level of congestion, but the emission results
are grouped by average speed. Lower speeds will correlate to higher congestion and
higher speeds to more free flow. There will be no need to specify the congestion levels
for roadways in MOBILE6.
The numbers in the speed-correction table (Table 1) in the Workshop handout, especially
average and maximum speeds, look reasonable and plausible.
Marion R. Poole. DOT. NC (Comment #25) April 16.1997:
Facility specific drive cycles are perhaps the most significant of the proposed
improvements to the MOBILE Model. North Carolina approves of the move in this direction.
The current model uses an average drive cycle to represent all possible driving conditions. This
leads to counterintuitive results in some cases. However, we have some concerns based on the
amount of aggregation and disaggregation in the supporting materials.
Is the variability of stop/delay time implicit in the drive cycle that will be used to develop
the basic emissions rates for each facility type? Our experience is that stop/delay time varies
across facility types. We believe that any future version of the MOBILE model should account
for this variation. An alternative method would be to allow the user to specify stop/delay time
for each facility type.
Final M6.SPD.002 119 June 2001
-------
InMOBILE6, the amount of stop/delay is implicit in the average speed. Average speed is
defined as the distance traveled (in miles) divided by the time (in hours). Stop/delay
would increase the amount of time, without changing the distance traveled, and decrease
the average speed. The stop/delay involved in each driving cycle is fixed, but when a
specific average speed is input to MOBILE6, an implicit amount of stop/delay will be
assumed based on the driving cycle at that speed. More detailed analysis of specific
roadways will likely require new emission modeling approaches that do not depend on
fixed driving cycles.
We also note that arterials and collectors will share a driving trace. As noted above our
experience indicates the existence of significant differences in stop/delay time and start mode
between facility types. Collectors resemble locals streets more than arterial streets.
It was not possible, using the available data, to make a finer distinction between roadway
types for MOBILE6. As discussed above, differences in stop/delay time are accounted for
in the average speed input. Although collectors may resemble local streets in terms of
stop/delay, there are important differences in the maximum speed and congestion levels
that, for purposes of emission modeling, make them more like arterials.
The proposed freeway drive cycles also provided some surprises. The proposed drive
cycles include: High Speed, LOS A-C, LOS D, LOS F, and LOS G. We recommend that the
High speed drive cycle and the LOS A drive cycle be combined, and that the drive cycle for LOS
B-C be kept together. Our understanding of the Highway Capacity Manual indicates that high
speed driving occurs under LOS A. We also propose that LOS F and LOS G be combined. To
the best of our knowledge, the Highway Capacity Manual does not recognize a LOS G. From the
associated driving trace, this drive cycle represents breakdown conditions and might best be
consolidated into the drive cycle for LOS F.
MOBILE6 will not use congestion levels as the method to associate driving with emission
levels. Instead, average speed will be used. The driving cycles were designed to give the
widest range of average speeds, adding a high speed cycle and a LOS G category.
Average speedwill be used as a surrogate for congestion levels.
Gary Flispart. Jefferson Ctv. KY (Comment #26) April 25.1997:
Many of the proposed changes suggest a movement away from the coarse focus of SIP
inventory modeling and toward the fine focus of transportation simulation. Accurately
approximating real-world behavior and associated emissions has clearly been the long-term goal
of both transportation evaluators and air quality regulators. The key difference, of course, has
been in the relative time focus: short-term (hour by hour) versus long-term (daily or annual
average). Traffic planners deal in peaks and valleys throughout a day, while the SIP focuses on
an annual inventory based on a typical summer day (for VOC). Both in the sharing of traffic--
related data and in mutual needs to comply with mandated SIP conformity, the relationship
Final M6.SPD.002 120 June 2001
-------
between transportation users and SIP users of MOBILE models is critical. MOBILE6
developments impact both the traffic-related data which must be gathered to prepare a SIP and
the eventual build-no-build evaluation of highway projects. It is essential that MOBILE6 not
support one type of user at the expense of the other, because the two are interdependent.
Since SIP inventories are currently based on daily average emissions, any change which
disfavors such daily estimates has profound effect on inventories and targets.
The new subclasses of facility type could either be ignored, force-fitted to existing
categories, or adopted by APCDJC. Because KDOT does not currently measure traffic for all the
proposed categories, if they were adopted the HPMS data gathering process would need to be
altered or supplemented, which could involve anywhere from a week's analysis of data from
other sources to a complete restructuring of transportation measurement in the area.
Coordination with KIPDA and KDOT would be essential.
Comment: SIP inventories are currently performed using daily averages and are
reasonably calculable when handled that way. In the District's experience, the people who used
MOBILE to compile SIP inventories and evaluate SIP strategies were not the same people who
compiled the link or trip models. The District is concerned that by shifting the emphasis toward
detailed transportation modeling, the primary efforts of SIP modeling may be undercut. Trying
to produce an annual inventory by summation of all trips or links in a simulation would add
tremendous complexity which APCDJC sees as unwarranted. The District strongly suggests the
need for MOBILE6 and other future models to continue to produce daily average emission
factors in a manner similar to that in MOBILES, to support SIPs and tracking.
The need for daily average emissions was recognized by the MOBILE6 team. MOBILE6
will continue to support daily average emission results in both the descriptive and
database output options. This will not completely eliminate the need for areas to produce
more detailed vehicle activity data, disaggregated by time of day, vehicle type and
roadway. Guidance will be needed to assist areas to determine what new data is needed
and generate the needed information.
Harold Nudelman. NYCDEP. NY (Comment #27) April 28.1997:
Have the new facility-specific cycles been reviewed by DOT/FHWA personnel? We are
especially concerned about in-City roadways (arterials/collectors and local) where there are speed
limits that may only allow 30-35mph. The maximum and average speeds for the bottom 3 cycles
on Table 1, for congested in-City arterials and local roads, may be too high for many congested
New York City streets during peak hours. New York City is likely to have a traffic control sign
and signal density which is at the extreme end of the range in the nation. Frequency of starts and
stops, and therefore of acceleration/deceleration, will not only affect average speeds but also the
emissions associated with a given speed. We support any efforts by EPA to develop operating
mode data that would allow us to project the impact on emissions of a high density of traffic
Final M6.SPD.002 121 June 2001
-------
signals on local streets as well as on arterials. Will speed corrections for arterials/collectors also
utilize data from the NYCC and FTP cycles? Will the speeds on local streets be adjusted? If
yes, what cycles will be used other than the NYCC?
The data available for MOBILE6 development did not allow for creation of new driving
cycles and data for extreme low speeds. However, each vehicle was tested using the New
York City Cycle to allow modeling of average speeds to that level (7.1 mph) and to allow
connection to the existing data from driving cycles below 7.1 mph. The speed correction
factors from these older driving cycles will be used in MOBlLE6for speeds below 7.1
mph.
Extreme conditions will always be difficult to model without data gathered to specifically
address those conditions. The driving cycle for local roadways has only a single average
speed, so changes in average speed on local roadways cannot be modeled. Guidance
will be needed to indicate the best methods to deal with specific situations where using
the default values would not be appropriate.
How will idle CO emissions be calculated? Will they be calculated from the 2.5 mph
emissions estimates? If yes, will the 2.5 mph emissions for local streets be the same as the 2.5
mph emissions for arterial/collectors? If they are not the same, how will they be calculated?
How different can we expect the low speed correction factors to be in the new model compared
to those used in MOBILE 5? Is there any reason why idle emissions data is not directly collected
to use in the model instead of adjusting the 2.5 mph emissions?
No new analysis of idling emissions has been performed since the release ofMOBlLESb.
InMOBILESb, idling emissions are calculated from the emission estimates for 2.5 mph,
as described in the MOBILES User Information Sheet #2. MOB1LE6 will not explicitly
estimate idling emissions at all. However, idling emissions will still be calculated using
the method described in the MOBILES User Information Sheet #2. The speed correction
factor (and thus the emission estimate) for emissions at 2.5 mph will be the same for both
roadway types inMOB!LE6 (freeway and arterial/collectors), since the speed correction
factor curves converge at extreme low (and high) speeds.
Collecting data for and analysis of idling emission data has not been a high priority for
EPA. Programs to develop idling emission factors have been proposed repeatedly, but
pushed aside by higher priority issues. An independently funded research project
specifically targeted at developing idling emission factors may be needed.
Dale Aspv. EPA #4. GA (Comment #30) April 30.1997:
A number of Region 4 states have requested the ability to model idle emissions for
project level analyses. Many of these same states have also requested the ability to conduct
facility specific modeling on an hourly basis to allow for peak use times. The Region supports
Final M6.SPD.002 122 June 2001
-------
the inclusion of non-FTP emission factors and the option of user supplied information regarding
non-FTP speed and acceleration activity factors.
Idling emissions were also discussed in the previous comment (#2 7). MOBILE6 will not
explicitly generate idling emission rates. However, idling emissions may still be
calculated using the method described in the MOBILES User Information Sheet #2.
Idling emissions were left out ofMOBILE6 since idling is a specific mode. Modal
emission rates (i.e., acceleration, deceleration, cruise and idling) were considered
outside the scope of the MOBILE6 project. Existing guidance (i.e., MOBILES User
Information Sheet #2) was considered sufficient for estimating idling emissions.
MOBILE6 does model emissions by hour and allows for user supplied speed VMT
distributions. MOBILE6 does not allow for adjustment of the amount of off-cycle driving
behavior, which is implicit in the driving cycles used to develop the speed correction
factors.
Michael Keenan. NYSDEC (Comment #31) April 30.1997:
Refocusing the Mobile model to the premise that most driving is non-FTP should prove
to be a most worthwhile development.
The attached tables contain the minimum and maximum average roadway type speeds
presently available for SIP modeling in New York State. The values shown are the estimates for
calendar year 1999. Comparison with EPA's New Facility-Specific Speed Correction Cycles
(i.e., Table 1 at March 1997 Workshop) indicates that the proposed Freeway Average Speeds of
13.1 to 63.2 mph would encompass New York's input range of 19.7 to 59.6 for Interstates,
Freeways and Expressways. However, for the various roadway types encompassing Arterials and
Collectors, New York's speed range of 7.2 to 55.9 mph is much broader than EPA's average
speed range of 11.6 to 24.8 mph. A large variance also exists within New York's Local average
speed range of 3.0 to 39.2 mph. Using a single local cycle with an average value of 12.9 mph
would appear to be most inappropriate for modeling such a variable speed range. Because the
range of possible average speeds for any given roadway type varies significantly, speed should
continue to be an input variable to the Mobile model.
However, by modifying the nomenclatures, the new facility specific cycles would prove
useful in better defining which speed correction factors should be applied. For instance, although
Freeways are limited access highways, the driving cycle traces may be applicable to certain rural
roadways as well. For example, the High Speed, Level of Service (LOS) A-C and LOS D
Freeway cycles appear to be suitable for any road that has unimpeded, nonstop free flow.
Freeway LOS E and perhaps even LOS F could be of use for modeling speed corrections on
roads which have a quick stop at a stop sign or slowing down to make a turn. Perhaps
differentiating roadways among nonstop, brief stops and many stops would be a more useful
approach for identifying which speed correction algorithm to use. This is perhaps what was
Final M6.SPD.002 123 June 2001
-------
meant by "signal density" in Issue #3.
The speed correction factors for freeways and arterial/collector roadways inMOBILE6
converge at low and high speeds. In this way, any speed from 2.5 mph to 65 mph can be
modeled on both freeways and arterial/collector roadways. Guidance will be needed to
assist users in choosing the appropriate roadway type in MOBILE6for the specific road
to be modeled.
Issue #1 points out that the "disaggregation ... by facility-type" is most appropriate for
short time periods. Shorter time periods generally have less variability of speed, temperature,
vehicle mix, etc. Further, four-step transportation models are now being developed for short
time frames (i.e., peak travel times). In addition, from the modeling perspective, combining
variables complicates input development. Thus, the Mobile model input at the scenario level
should reflect input appropriate to a discrete roadway type for a time period short enough to
minimize large differences in any input variable over that time period.
MOBILE6 allows for different VMT by facility type and vehicle type for each hour of the
day. However, guidance will be needed to choose the appropriate level of aggregation
for development ofMOBILE6 inputs for specific emission inventory analysis.
"Weighted running emission factors" and inputting %VMT by roadway type (ala Issue
#2's methodology) would complicate using the Mobile model and jeopardize input integrity.
While this concept may sound attractive and efficient, the model already suffers enough
uncertainty without introducing more. With today's desktop computers, multiple scenario runs
can be performed quite rapidly. Ample software and/or software packages are available for
preprocessing and postprocessing (e.g., G/mi times VMT). Therefore, the Mobile model does
not have to become its own postprocessor!
As was noted in other comments, daily average emissions continues to be a highly desired
output of the MOBILE model. As a result, MOBILE6 must be capable of producing
emission results for a given day that aggregates all hours, roadways and vehicle types.
However, it will be possible, with appropriate input commands, to get results specific to a
less aggregate scenario, such as for an individual roadway.
Shengxin Jin. NY DOT (Comment #36) May 6.1997:
Average vehicle speed is another important issue in CO intersection dispersion modeling.
Vehicle speeds differ from intersection to intersection. Even at the same intersection, vehicle
speeds can vary depending on the directions of traveling vehicles. Without a vehicle speed
option as a model input, the differences in vehicle emissions due to speeds can not be
determined. This will significantly affect CO dispersion modeling results.
Average speed will still be a user input for the MOBILE6 model. However, the
Final M6.SPD.002 124 June 2001
-------
complexity of the input will be increased due to the need for speed VMT distributions.
Marcel Halberstadt. AAMA. MI (Comment #37) June 5.1997:
The EPA approach is somewhat different than the ARB's approach, in which a single,
self-weighted inventory cycle was developed (the LA92, or Unified Cycle (UC)) and a significant
number of cars and trucks were tested on this cycle. Also, ARB developed Unified Correction
Cycles for developing speed correction factors for the UC. AAMA is unsure if EPA can devote
enough resources to make their approach more accurate than ARB. Concern stems from EPA's
desire to make one model fit all modeling purposes. The Unified Cycle approach is certainly
more simple, and has the advantage that only a single, self-weighted cycle needs to be run for
area-wide modeling. EPA's approach requires significantly more testing per vehicle,
consequently, fewer vehicles can be tested. There is also an issue with respect to whether
vehicles can be maintained at proper temperatures throughout the duration of the EPA cycle
testing. AAMA recommends that EPA also have all of the vehicles tested on ARB's Unified
Cycle as well as the other cycles, so the Unified Cycle can be compared to a weighted average of
EPA's cycles. AAMA will reserve further comments on both ARB's approach and EPA's
approach until it evaluates the data from EPA's test program, and particularly how EPA compares
the data on the Unified Cycle to the data from the EPA's test cycles. If EPA's approach of many
factor-specific correction cycles remains unchanged for MOBILE6, it is essential that the model
contain default (nationwide) statistics to develop average emission rates for a nationwide
inventory.
EPA understands and accepts the advantages of the California approach to emission
modeling. However, the EPA approach meets the important requirement that emission
estimates on smaller scales (i.e., individual roadways) be as accurate as possible.
California knows that transportation planners will use their model to estimate emissions
for individual roadways, despite the fact that this is not appropriate, because there is no
reasonable alternative. The EPA approach is more appropriate for modeling individual
roadways, which can then be aggregated to create area wide daily emission estimates.
EPA agrees that the current EPA approach will require significantly more testing per
vehicle and will result in fewer vehicle tests. However, EPA is confident that
improvements in instrumented vehicle technology will allow for the use of emission
testing results from in-use vehicles on roadways to create emission estimates for future
models. In this way, MOB1LE6 is a transition model which can be used to improve our
understanding of how emission inventories can be improved using a better designed
emission model.
EPA must also allow users to output emissions based solely on current FTP certification
test results, for ready comparison with the current and historical emission standards.
MOBILE6 is designed to estimate "real world" driving emissions and cannot easily
Final M6.SPD.002 125 June 2001
-------
replicate the exhaust emission certification procedure (FTP). Doing so would require
elimination of "real world" driving behavior effects on emissions and altering the default
vehicle activity distributions. This will be possible using diagnostic commands, but is not
an expected typical use of the model.
Marcel Halberstadt. AAMA (Comment #53) December 4.1997:
EPA is proposing to use data currently being developed from the testing of in-use
vehicles over a variety of driving cycles. This testing has been performed at ATL and EPA. Data
collected at the two different sites shows remarkably different sensitivity of emissions to average
speed. The ATL data generally showed lower emissions, and a lower emissions sensitivity to the
different test cycles. EPA indicated that because of the differences, it would run a correlation
program to try to determine the reason for the differences, but also indicated that EPA may base
the speed effects for current vehicles in MOBILE6 on the EPA data alone because the ATL data
"may be underloaded."
AAMA supports EPA's efforts to conduct a correlation program. AAMA believes the
differences between the ATL and EPA data must be thoroughly understood before EPA makes
significant decisions about which data to base the speed effects on. EPA did not indicate why it
thought the ATL data may be "underloaded". Another possible explanation, which was not
addressed at the workshop, is that the EPA data may be in error (or "overloaded"). If the reasons
for the differences are not thoroughly understood, EPA should combine the data, but not omit the
ATL data without very good reason.
EPA thoroughly investigated the differences between the results at the two testing sites
(including testing the same vehicle at both sites) and resolved that the differences are not
due to errors or differences in the testing procedures at the two sites. EPA has concluded
that the differences observed are vehicle to vehicle variance and all test results at both
sites have been used in the analysis.
Another issue relates to how EPA plans to use the freeway ramp driving cycle. AAMA
understands that EPA intends to develop national weighting factors for different types of
roadway operation, and allow users to input these fractions as well. The model would then
weight the emissions from the different cycles together. It is not clear, however, whether EPA
will also have speed correction factors, which will adjust emissions between the speeds of the
different cycles. If EPA plans to do this, then it should develop such speed correction factors
from the new data, but omit the data from the freeway ramp cycle. This cycle appears to result in
emissions that do not lie on the typical emissions/speed curve (see Figures la-lc of the above
report).
MOBILE6 separates driving into four roadways types, with freeway ramps as it's own
roadway, separate from other freeway driving. However, since there is only one ramp
driving cycle, ramp emissions will not be a function of average speed.
Final M6.SPD.002 126 June 2001
-------
Facility-type Speed Correction Factors - It is still not clear to AAMA how the facility-
type speed correction factors (SCFs) are being developed. Is EPA developing separate SCFs by
facility types, or a single SCF curve across the entire speed range? If SCFs are being developed
for different facility types, then how will EPA divide the facility types and levels of service?
How will high emitters be handled? Will the SCFs for low and high emitters be estimated
separately, and then combined by the estimated fraction of low and high emitters?
As this final report should make clear, MOBILE6 divides driving into different facility
types with a speed correction curve for freeways and arterial collectors. Levels of
service are not used directly. Instead, average speed is used as a surrogate for the
congestion level on roadways. Speed correction factors for high emitting vehicles were
determined separately from normal emitting vehicles.
Effect of SFTP Standards for Tier Is and LEV-Type Vehicles - EPA is proposing to
include the effects of off-cycle aggressive driving through the use of the facility-specific speed
correction factors. Thus, the SCFs will include the effect of speed as well as off-cycle effects.
How does EPA plan to incorporate the effects of the SFTP rules on Tier 1 and LEV vehicles,
using the facility cycle data on Tier 0 vehicles?
The development of the speed correction factors included an emissions offset which
attempts to capture the difference between the base exhaust emission factor, based on the
FTP driving cycle (LA4), and truly representative driving, which includes aggressive
driving behavior. A full discussion of the emission offset is in Section 6.0. This emission
offset is the portion of the overall adjustment to the base emission rate that will be
affected by the SFTP. The effects of the SFTP on emissions are discussed in the report,
"Determination of Off-Cycle Emissions and Supplemental FTP Control Modeling in
MOBILE6," (M6.SPD.005).
EPA estimated the benefits of SFTP rules in its support document to its supplemental
FTP final rule. However, in that analysis, EPA estimated emissions over ST01, REM01 and
REP05 from testing over the FTP and US06, along with some Tier 0 vehicle data. This
methodology contains a number of assumptions which have not been confirmed with data.
Therefore, AAMA does not recommend that EPA use this methodology in MOBILE6 without a
thorough review of it appropriateness.
Likewise, how does EPA plan to estimate these factors for LEVs with and without non-
FTP controls? ARB assumed that the impact of non-FTP driving on LEV emissions was the
same in relative terms as for Tier 1 vehicles in EMFAC7G. In its supporting analyses for its
proposed non-FTP standards, ARB also estimated the impact of non-FTP driving on LEV
emissions both with and without SFTP controls. However, as was the case above, this
methodology involves many unconfirmed assumptions. Also, the technology assumed by ARB to
enable compliance with the non-FTP standards (i.e., rich-bias) is not likely to be the technology
of choice for most manufacturers. Therefore, AAMA again recommends that EPA publish the
Final M6.SPD.002 127 June 2001
-------
details of any methodology which it plans to use to estimate LEV emission impacts for public
comment prior to its incorporation into MOBILE6.
The report, "Determination of Off-Cycle Emissions and Supplemental FTP Control
Modeling in MOBILE6, " (M6.SPD. 005) describes how MOBILE6 handles off-cycle
emissions and the effects of the SFTP. It should be of some comfort to know that off-cycle
emissions and the effects of the SFTP are added to the base exhaust emission rates. This
means that only the small increment of remaining off-cycle effects left after the
effectiveness of the SFTP has been applied is added to the base emissions of vehicles
affected by the SFTP. This should mitigate the effects of uncertainty in the estimate of
what off-cycle emissions might be for LEVs. The true effectiveness of the SFTP cannot be
known until vehicles certified under the new certification procedures can be tested.
Gary McVoy, Ph.D., Director, Environmental Analysis Bureau, New York State
Department of Transportation (Comment #86) August 2,1999:
Vehicle speeds on local roads are much different from project to project and from area to
area. For example, the local speed and driving pattern in New York City are much different from
those in the NY upstate cities. No speed adjustment for the local roadways affects our ability to
perform accurate and publicly defensible air quality analysis.
The limited data available on the driving behavior and their emission impacts on local
roadways make it impossible to accurately determine the effects of average speeds on
local roadways. However, it may be possible, with proper guidance, to account for the
differences in the average speeds on local roadways, using MOBILE6, in a manner
consistent with EPA policy. This issue will be addressed in guidance from EPA.
Gerry Kelpin, New York City Department of Environmental Protection (Comment
#93) October 5.1999:
We would like to raise a number of concerns with some of the assumptions proposed for
the model. The areas of our concern are as follows:
1) The use of the Mobiles low speed relation to estimate CO emissions for speeds below 7.1
mph, and the subsequent use of the estimated 2.5 mph emissions for estimating CO idle
emissions.
OMS views the Mobile Model as essentially a tool for developing emissions inventories.
However, the use of the model for providing link by link carbon monoxide emissions to be used
as input for intersection air quality modeling to determine compliance with the NAAQS has also
been an important function of the model. We believe that the proposed assumptions for adjusting
the extremely low speed emissions and using the resultant 2.5 mph emissions to estimate the idle
emissions overestimates CO emissions for these conditions and will result in overestimated
Final M6.SPD.002 128 June 2001
-------
modeled intersection impacts. If the Mobile6 CO emissions are not sufficiently lower than the
MobileS emissions then this could even result in erroneous determinations of non-compliance.
In order to explain our concern about the Mobile Model assumptions for low speeds and
idle, and their potential for generating incorrect results, we have to discuss how the emissions
input for the CAL3QHC(R) air quality model, that EPA designated as the reference model for
modeling CO at intersections in 1992, is generated. The CAL3QHC(R) model does not utilize
average speed emissions. The CAL3QHC(R) model utilizes free flow, or running speed
emissions, and idle emissions for each link. (In the air quality model, a link is usually the
distance on a roadway between traffic signals, or other types of intersection controls.) The
emissions on a given link are divided into idle emissions over the length of the queue for the
vehicles stopped in the queue, and the moving emissions from all the vehicles passing through
the link over the entire length of the link. The moving emissions for the CAL3QHC(R) model,
therefore, should reflect the running speed (speed when vehicles are in motion) and not the
average speed.
The running speed for a link is calculated by subtracting the stopped delay time from the
total travel time, and dividing that time into the link distance. The average speed includes the
stopped delay time in the total travel time. Running speeds will therefore always be higher than
the average speed on any link with a stopped delay. The running speed should contain no stopped
delay, and therefore no idle emissions. When we use the Mobile Model emissions for a given
speed that is equal to the running speed, we are using emissions for an average speed that
includes some percentage of idle emissions. In general the amount of time that is spent in idle
decreases with increasing average speed. If idle emissions are higher than the moving emissions,
the greater the percentage of idle time in the average speed, the more the running or moving
emissions will be overestimated.
It must be noted that although CAL3QHC(R) may request that running emissions
estimated from a free flow speed be used, this has never been the case. None of the
versions of the MOBILE model has ever been able to produce free flow emission
estimates. This is true no matter how the speed is calculated. EPA has allowed the use
of the MOBILE model as an input to the CAL3QHC(R) model only because there is no
credible alternative. In this respect, the MOBILE6 model in general will be no better, but
no worse, than current practice.
The truly appropriate model for use with CAL3QHC(R) is a modal model, which
estimates emissions base on modes instead of trips. A properly developed modal model
would be able to estimate free flow emission rates appropriate for input into models such
as CAL3QHC(R). EPA is currently working with researchers in California and Georgia
to develop such modeling tools.
The impact from the potential overestimation of moving CO emissions, however, is not
believed to be as significant as the impact from overestimating the idle emissions. The impact of
Final M6.SPD.002 129 June 2001
-------
idle emissions has generally been responsible for the major portion of the modeled local impact
at intersections that have been near the standard. This is related to the CAL3QHC model
assumption that all the cars in a queue on a link will idle for the entire red phase of the light
cycle. This means idle emissions from the entire queue on a link is modeled for a percentage of
the hour that is proportional to the red time divided by the total cycle time.
Given the major contribution of idle emissions to local CO predicted impacts, we do not
agree with the proposal to use the relationships between the 2.5 mph emissions and those at 7.1
mph from MobileS in Mobile6, and then use the 2.5 mph emissions to calculate idle emissions.
When this methodology for calculating idle emissions was adopted for MobileS, it was
recognized that it should be replaced. This is indicated by the statement on page 1-8 of the May
1994 User's Guide to MobileS - "EPA will continue to collect data and to work to develop a more
satisfactory approach to estimating idle emission factors." We believe that a more satisfactory
method is needed because as we understand it, the speed relationship for the low speeds is based
primarily on data from older technology, primarily carburetor, vehicles, and has not been
demonstrated to be appropriate for the current and future fuel injected vehicles. The relationships
that were included in MobileS for that speed range go back to Mobile4.1. The current and future
fuel injected vehicles, with air/fuel ratios controlled by computer chips, should be much more
efficient than the carburetor vehicles at controlling emissions at idle. Therefore, the relationship
of emissions at these low speeds, and the idle emissions themselves for the newer vehicles
should be different from what would be estimated utilizing the relationships developed for the
early technology vehicles.
Given the importance of the idle emissions to local CO impact prediction, we would
recommend that CO idle emissions for current and future conditions be estimated based on actual
measurements of idle. Is it possible to extract measured idle emissions data from the data
generated in "grams second by second" for the vehicle cycles that were tested and described in
the report ? If this is possible, we would recommend using relatively continuous periods of idle
conditions (over 10 seconds) rather than shorter periods to measure the idle emissions so that
they will be based on conditions that are similar to how they are modeled in the CAL3QHC
model. Other sources of idle emissions may also be available (I/M programs, certification tests,
etc.).
No new analysis of idling emissions has been performed since the release ofMOBILESb.
InMOBILESb, idling emissions are calculated from the emission estimates for 2.5 mph,
as described in the MOBILES User Information Sheet #2. MOBILE6 will not explicitly
estimate idling emissions at all. However, idling emissions will still be calculated using
the method described in the MOBILES User Information Sheet #2. The speed correction
factor (and thus the emission estimate) for emissions at 2.5 mph will be the same for both
roadway types inMOBILE6 (freeway and arterial/collectors), since the speed correction
factor curves converge at extreme low (and high) speeds.
Collecting data for and analysis of idling emission data has not been a high priority for
Final M6.SPD.002 130 June 2001
-------
EPA. Programs to develop idling emission factors have been proposed repeatedly, but
pushed aside by higher priority issues. An independently funded research project
specifically targeted at developing idling emission factors may be needed to properly
resolve these issues.
However, all investigations on the existing data to date have supported the assumption
that emissions at 2.5 mph are similar to those during idle. There is no reason, based on
data, to be overly concerned about the idle emission estimates. One clear advantage of
basing idling emission on the emissions at 2.5 mph is that is allows the idling emissions
to be affected by all of the correction factors that are applied to running emissions, such
as fuels and temperature. It will be very difficult to replicate these corrections
specifically for idling emissions, even once base idle emission data becomes available.
The issue of estimating running CO emissions between 7.1 mph and 2.5 mph, without
using the old speed relationship from Mobiles, for intersection modeling must also be addressed.
We do not believe that these running speeds will occur very often. For example the running
speed for the low speed NYCC, whose average speed is 7.1 mph, is about 12 mph, after
subtracting out the 40 % of the time in the cycle that the vehicle is in idle. We have not evaluated
alternative solutions. However, one possible method that could be examined is to extrapolate
the curves down to 2.5 mph from the 7.1 mph emissions measured in the new cycles. In addition,
if new idle emissions data are available, it should be reasonable to use the idle emissions as an
approximation of the emissions at an average speed of 2.5 mph. This should not introduce much
error since an average speed of 2.5 mph will have a very high percentage of idle time. (The 7.1
mph average speed cycle has 40 % of its time in idle.) A best fit curve utilizing this additional
point, with the other points, could then be developed.
In the future, having an idle test done as part of the certification process or as part of a
mandated I/M program would seem to be a simple and inexpensive way of providing updated
information for future revisions to the models' idle emissions.
Although we have made our comments about the low speed adjustments with respect to
CO, there should probably be an examination of whether the low speed relationships for the old
technology vehicles is appropriate for estimating the emissions of HC and NOx from the new
technology vehicles. Unless the applicability of the adjustments can be demonstrated, an
approach similar to that mentioned above for CO might be worth evaluating.
There is no particular advantage to using the New York City Cycle (NYCC) results from
the new data set to estimate idling emissions in place of using the low speed (2.5 mph)
cycle. Extrapolating low speed emission corrections from higher speed results is very
difficult, since the low speed portion of the curve is quite steep. A very small error can
produce large differences. Basing the low speed portion of the curve on data (whatever
it's minor flaws) is a better choice than extrapolation.
Final M6.SPD.002 131 June 2001
-------
It is obvious that the only way to know what is happening with emissions at low speeds
and idle will be to properly analyze the existing data and collect new data specifically to
address these issues. There are new tools, such as the University of California modal
model (NCHRP Project 25-11) which can directly address the effects of driving behavior
on emissions at low speeds and idle and there is new data imbedded in any driving cycle
containing idling and low speeds which has been collected on a second by second basis.
However, much effort will be needed to review this information and propose a new set of
low speed and idling emission rates. This effort is beyond the scope of the MOBILE6
project and will have to be addressed as an update to the model.
2) The adjustment of low speed emissions for aggressive driving.
Since we do not think that the old speed adjustment factors should be utilized below 7.1
mph for CO, the issue of adjusting the old adjustments for the aggressive driving in the new
cycles becomes moot. We have reservations about correcting emissions for very low speeds,
characteristic of severely congested conditions, for aggressive driving. One would think that
under these type of conditions there would not be very much opportunity for this type of
behavior. This would be even more true for idle emissions, if they were estimated from a 2.5
mph emission.
The proposed adjustment for CO at 7.1 mph and lower speeds for level 1 and level 2
vehicles would result in a reduction of the old speed factors. This is because the new speed factor
for 7.1 mph is lower than the old factor. Could this reflect the reduced contribution of the idle
component to the total emissions in the NYCC for the vehicles recently tested as compared to the
tests of earlier technology vehicles?
The New York City Cycle (NYCC) is based on "real world" driving and includes
aggressive driving behavior not found in the FTP. Since the FTP, which was the base
emission rate in MOBILES, did not include enough aggressive driving behavior, the
speed adjustment in MOBILES from the base (FTP) to low speeds (NYCC) was large and
the speed adjustment included some effects from aggressive driving. InMOBILE6, the
speed correction factor is applied to a base emission rate after the effects of aggressive
driving have been added as an emission offset. This may be the reason the speed
correction is less moving from the base emission rate to the low speeds, since both now
contain the effects of aggressive driving.
Not much is really known about driving behavior at low speeds. The current
instrumented vehicle data does not allow us to easily separate roadway types, so that a
more thorough analysis of low speeds cannot be done. The chase car data, on which the
MOBILE6 freeway and arterial/collector roadway cycles were based, did not follow
vehicles onto local roadways, where most low speed driving occurs. New instrumented
vehicle data will include global positioning sensing (GPS) technology, which will allow
precise locating of the vehicle on the roadway system.
Final M6.SPD.002 132 June 2001
-------
3) The assumption that freeway and arterial emissions converge and have the same
emissions at, and below 7.1 mph.
We agree that the emissions for all types of roadways will converge as the average speed
approaches zero and the emissions become essentially idle. Given the different way arterials and
freeways function, however, we are not certain that the assumption that their emissions converge
at 7.1 mph and are the same below 7.1 mph is an accurate description of their behavior.
Generally freeways, because of the lack of traffic signals that create stopped delay, should have
the same running speeds and average speeds, while this is not the case at signalized arterials or
collectors. Would a low average speed of 7.1 mph on a freeway be characterized by the same or a
similar amount of idle time as the same speed on an arterial? It might be useful to compare how
the percentage of idle time changes for freeways and arterials as their average speed cycles
approach 7.1 mph (Freeways, LOS F to LOS G and arterials/collectors, LOS C-D to LOS E-F).
In addition, is the percentage of idle time in the lowest speed cycle (which had relatively close
average speeds) for each type of roadway consistent with their both converging to the 40 % idle
at 7.1 mph in the NYCC? If the above does not support the assumption that the emissions from
both roadway types converge at 7.1 mph, then it may be necessary to modify this assumption.
The lowest speed arterial/collector cycle has an average speed of 11.6 mph and 31.3% of
the cycle time is at idle. The lowest speed freeway cycle has an average speed of 13.1
mph and 3.3% of the cycle time is at idle. The New York City Cycle (NYCC) has an
average speed of 7.1 mph and 32.4% of the cycle time is at idle. Table lib shows the
average CO emissions from vehicles on these cycles. The NYCC has higher CO
emissions than the other two cycles in all cases. The arterial/collector cycle is higher
than the freeway cycle in all cases. Based on this information, CO emissions will
increase as average speed decreases and as the fraction of cycle time spent idling
increases.
Logically, EPA concluded that the speed correction factors for freeways and
arterial/collector roadways must be identical when 100% of cycle time was spent at idle.
However, based on the available information, it is not possible to determine precisely
where the two speed curves statistically converge. The odd cycle of the three is clearly
the freeway cycle, where very little of the driving time is at idle (3.3%). However,
freeway driving at these very low speeds is not typical, in terms of the daily VMT on
freeways. EPA concluded that it would be best to converge the speed correction curves
at the NYCC where an actual data point existed. This assumption should only affect the
limited amount of VMT that occurs on freeways at very low speeds.
Why isn't the estimation of local roadway emissions variation with speed based on the
relationships derived for arterials and collectors, since they would appear to be very similar to
these roads in the way they operate?
A detailed analysis of the emissions from the local driving cycle and the arterial/collector
Final M6.SPD.002 133 June 2001
-------
driving cycles has not been done. As you suggest, it is likely that use of the
arterial/collector speed corrections may be appropriate for modeling local roadways as
well. However, since local roads are not usually included in traffic demand models,
MOBILE6 includes an overall local roadway emission estimate. The issue of whether
local roadways can be modeled as arterial/collectors will be addressed in EPA guidance.
Final M6.SPD.002 134 June 2001
-------
Appendix D : Peer Review of Speed Corrections
Review of: Facility-Specific Speed Correction Factors, Draft
US EPA Report Number M6.SPD.002
By David J. Brzezinski, Phil Enns, and Constance J. Hart
Assessment and Modeling Division
Office of Mobile Sources, U.S. Environmental Protection Agency
Reviewed by: Simon Washington, Transportation Systems Group, School of Civil &
Environmental Engineering, Georgia Institute of Technology
October 20,1999
NOTE:
Most of the following comments have been addressed by making changes in the text of the
report to clarify or add information. However, some comments are addressed below the
comments. All EPA responses are shown in italics.
General Review Comments:
The comments below reflect three different types of comments. First are editorial comments that
I believe would improve the read of the document, or would make for more precise interpretation
of some of the statements made in the document. Second are short-term improvements, which I
believe could be addressed in the immediate future to improve the development of driving cycles
as proposed in MOBILE 6. The last section lists longer-term improvements, which could be
considered after MOBILE 6 has been released.
Overall the document is very thorough, well written, and concise. The authors have done an
excellent job documenting a difficult project, and should be commended for their professional
work. I would like to caveat all my comments by saying that they are intended merely to improve
the document, and not in any way to offend any of the highly qualified and experienced authors
who have prepared the document. The comments are my opinion only, and certainly can be over
ridden by consensus of the authors or other reviewers.
I have attached three asterisks (***) to those comments which I believe are the most
pressingthat is those areas of concern that I believe raise some serious questions as to the
validity of the currently proposed approach, and whose impact might be significant on the results
and conclusions of this work.
Editorial Comments:
Page 1, 4th paragraph: The first sentence should probably read, "The proposed basic
emissions levels of the vehicles."
Page 1, 4th paragraph: The fact that ramps and local roadways cannot be adjusted for average
Final M6.SPD.002 13 5 June 2001
-------
speeds other than the national average is a bit confusing. It suggests that the same average speed
be assumed for all local roads, regardless of the city that is being modeled. If this is the case, then
this is a problem when modelers may want to assess the effect of peak spreading (some TCM's)
or shrinking (some advanced ITS technologies), which would have an affect on local road traffic
volumes and therefore speeds as well. This statement should be clarified.
Page 2, 1st paragraph: The first line should read, "Since the data for this analysis were
collected....". The document should be searched for occurrences of the plural data to check verb
agreement, as this occurs elsewhere as well (e.g. see page 16, paragraph 4).
Page 2, 1st paragraph. The first sentence uses the word "realistic", which I think should not be
used. Any conceivable driving cycle obtained from real driving is realistic. The appropriate word
to use might be "representative", since what EPA is trying to do is bring into the fold a greater
number of driving cycles that represent collectively a greater number of driving conditions. Even
the "old" driving cycles are representative in and of themselves; however, there are fewer of
them, so driving cycle heterogeneity is not being captured. I think it is worth keeping in mind
(and perhaps in the text also) that there is a continuum of representation of real driving, ranging
from one assumed driving cycle (and its assumed emissions profile) to simulation, which derives
any speed-time profile of a fleet of vehicles given, roadway, traffic, and environmental
conditions. Of course the latter begs for an emissions model that can handle any feasible driving
cycle, which in fact can be accomplished by either UC Riverside's or Georgia Tech's Measure
model. The point is that the continuum of driving activity is getting further disaggregated, and
this is what MOBILE 6 is doing.
Page 3, Background: Last sentence refers again to real world driving behavior, which implies an
alternative to non-real world driving behavior (see previous comment).
Page 4, 5th paragraph: Last sentence refers to the difficulty of differentiating vehicle activity
across facilities. It might be worth emphasizing that this has particular consequences in the
planning process, whereby plans or program that might impact modal activity cannot be modeled
adequately.
Page 5, paragraph 4: The second sentence should read, "Readers are encouraged to obtain
information directly from California for comparison with the results documented in this report".
Page 6, paragraph 2: The last sentence is unclear, please clarify.
Page 6, paragraph 3: EPA's objective to match the power distribution is right on track, it is a
strong point of the current driving cycle project update.
Page 6, paragraph 3: I'm not sure that the authors want to use the word "We", and instead might
use "the US EPA". See also page 20, third paragraph. The authors should probably search and
replace entire document to be safe.
Final M6.SPD.002 136 June 2001
-------
Page 7, paragraph 4: The second sentence should begin, "Testing of vehicles was done "
Page 8, 5th full paragraph: The word "special" should be removed from the second sentence.
Page 9, 3rd paragraph: I suggest changing the phrase in the third sentence "agrees with" to
"provides support for".
Page 15, paragraph 2: The authors state, "All of the slope coefficients are statistically significant,
meaning that the increase or decrease in emissions versus average speed is different than zero."
This is not technically correct. The correct interpretation of a hypothesis test is as follows: If
repeated many times (i.e. many samples drawn from the population), the outcome (data)
observed by the analyst/engineer and reflected in a computed test statistic (e.g. t-statistic, F-ratio,
chi-square, etc.) would occur x percent of the time if the null hypothesis were true. In other
words, the probability of occurrence is conditional upon the null hypothesis being true. If x is less
then alpha, then the null hypothesis is rejected. When the null hypothesis is rejected, the
statistical evidence suggests that the null hypothesis is not true, and that some alternative
hypothesis provides a better account of the data. What is important to understand (and which is
commonly misinterpreted), is that the result does not provide the probability of the null
hypothesis being true, nor does it provide evidence that the particular alternative hypothesis is
true. In contrast, however, it provides the probability of observing the data if the null hypothesis
were true.
Page 38, Table 8: The first asterisk footnote should probably read "All emissions in Log
(gram/hour) scale", thus replace space with scale for this table and all similar tables.
Potential Short Term Improvements:
Page 1, bullets: Where do 2-lane highways fit into the picture? It is known from operational
aspects of traffic that 2-lane highways have very different emergent traffic than do interstates
with more than two lanes, due to a number of factors including limited access, truck activity and
restrictions, passing maneuvers, and weaving. The entire paper has not mentioned 2-lane
highways, and perhaps should address somewhere how this is being handled. In the longer term
perhaps some empirical data on two-lane highways could be collected to determine whether they
should server as a "separate facility type".
Page 5, paragraph 2: The second sentence states, "Given limited testing budgets,...., thus
increasing the statistical confidence in the emission test results". It puzzles this reviewer that this
is even a consideration, given that statistical confidence has never been considered in any of the
modeling in MOBILE or CARB to date, particularly with regard to confidence in model outputs.
So what if the confidence is better if it is not used in the modeling process? My point is that the
US EPA and CARB need to do a better job of carrying through the uncertainty in forecasted
outputs, not just in the internal decisions being made as to "how many" cycles should be
generated. It is actually a good point being made in the text, but is largely irrelevant because of
Final M6.SPD.002 137 June 2001
-------
its limited use in policy and practice.
Page 6, paragraph 3: Although the authors feel that the emissions generated from the new cycles
are representative of driving behavior "under specific conditions", it is not clear that the
forecasting of the particular "specific conditions" can be done accurately. The authors might
want to suggest that there is additional uncertainty in matching the specific conditions in the lab
with the same conditions in the field, due to the mismatch between LOS defined by air agencies
and DOT's.
*** Page 7: 1st paragraph: It seems an awful lot to ask from 85 vehicles to infer emissions for a
fleet of millions. It would be nice to add a table to show how representative (or not) these
vehicles are of the national fleet. Perhaps a cross-classification of model-year by technology
classes. Of course all the estimated means are dependent on a random sample of vehicles, and it
is not clear whether this is the case. More documentation needs to be provided here, with perhaps
some clues as to how the final estimates might be biased as a result of the biased sample. Also,
much more needs to be added about recruitment and acceptance, since these are two separate
issues. Specifically, what was the proportion of rejections, and how was recruitment performed.
Again, random sampling is fundamental to probabilistic methods, and without it properties of
probability can not be expected to hold. More information would inform the reader here.
It is understandable that the typical reader will not be aware of the EPA standard vehicle
recruitment practices. However, it is not reasonable to attempt to fully describe and
defend these practices in this report. It should be sufficient to understand the analysis to
know that EPA was attempting to recruit vehicles in either a random or stratified random
fashion. Certainly, it will be hard to evaluate the total uncertainty in the overall result
without some sense of the bias in the recruitment. However, there are plenty of
opportunities in a sample this small to be concerned about uncertainty without the
addition of recruitment bias.
Page 7, 4th paragraph: It would be helpful to add a table comparing the sum of second-by-second
emissions to bag emissions, or at least to discuss it in the text. There has been some concern
about the difference in some of these testing discrepancies, and it would be a nice addition to the
text.
The second by second emissions were not analyzed at all for purposes of this report,
since their results were not used. The issues related to second by second measurements
are important, but beyond the scope of this report. The data from this testing, including
the second by second results, are publically available in our Mobile Source Observation
Database.
Page 7, 5th paragraph: I suggest removing the words (throughout) "real world" with another
phrase, such as "additional" or "additional representative", or "further disaggregated", for
reasons explained previously.
Final M6.SPD.002 13 8 June 2001
-------
"Real world" is a term used in the National Research Council report on the MOBILE
model. It is used in this report to refer to data that addresses the concerns of the NRC. I
have changed all occurrences of "real world" to "real world" representative to address
this concern.
Page 8, 1st paragraph: The list of parameters thought to be important really includes of host of
parameters, especially list item 3, which includes catalytic converter type, fuel delivery system,
engine size, etc. The text should be amended to explain that list items reflect the variability in
emissions from a larger set of parameters that are subsumed in the list item.
Page 9, paragraph 1: Please show the plot of grams per mile versus speed that the authors claim
does not fit the data. It is always helpful for the reviewer to be able to concur with the authors
assessment of lack of fit, and this cannot be done here.
The real issue is not a true lack of fit, but an engineering judgement of the expected
trends. Linear fits in grams per hour units lead to curves, when converted to gram per
mile units, which "tail" downward as average speed increases. This is not what is
expected and is not suggested by the few data points at the higher speeds. Using gram
per mile units removes this artifact of the modeling approach (the tailing off of emissions
using grams per hour) from the model, without introducing more complex curve fits into
the model.
*** Page 9, paragraph 3: It is not clear that the homogeneity of variance assumption in ANOVA
was checked for reasonableness. This should be done and reported in the documentation. Also,
there are many ANOVA tests being conducted, thus, the expected number of type I and type II
errors is increased. For instance, with alpha = 0.05, and 20 rejections of the null hypothesis, one
would expect on average (20)(0.05)=1 error The authors should keep in mind the number of tests
they are conducting in concert with their selected alpha and beta levels. The authors might
consider giving the beta level associated with some of the tests conducted.
*** The authors have seemed to ignore (or simply not report) the type n error rate. Type II errors
often are ignored in the development of statistical modelswhich of course is embedded with
statistical tests of hypotheses. Analysts might set an alpha level associated with a t-statistic to
0.05, only to find that several variables in their models, which were thought to be important on
theoretical grounds, had p-values associated with t-statistics greater than 0.05and so they
subsequently removed them from consideration. In some cases these variables have suffered
from type n errors, and should still be included in the model, especially when there is theoretical
support for such variables. Systematically removing 'non-significant' variables from a model,
therefore, ignores the possibility of important variables and related t-tests that have suffered from
type II errors.
The determination of which statistical error is less desirable depends on the research question and
consequences of the errors. Because these errors are relatedsmaller alpha equals larger beta, all
else being equalcareful decisions need to be with regard to selection of alpha and beta, and
Final M6.SPD.002 139 June 2001
-------
attempts need to be made to quantify beta when appropriate. There are various software packages
available for calculating type n error rates, and many textbooks also provide the necessary tools.
The number of parameters examined were carefully selected on a theoretical basis to
reduce their number. Non-significant variables were removed from this pre-selected list
based on the type I error statistics and consideration of the theoretical implications.
EPA has not investigated the additional statistical parameters, but it is not likely that the
additional work would result in a different choice for the model.
*** Page 11, paragraph 3: The authors state "... .a method was developed for increasing sample
size by ". It seems a bit unorthodox to increase sample size using this technique (or any other
technique that does not simply collect additional appropriate data). I understand the motivation
for this action; however, it is not clear that this will lead to satisfactory results. My concern is
that vehicles 'substituted' for the Tier 1 vehicles (Tier 0 vehicles) will not perform similarly on
the range of driving cycles used to estimate the MOBILE 6 emission factors. Theoretically, it is
presumed that for a given vehicle manufacturer and model, a vehicle that meets Tier 0 standards
will be fundamentally different than the same vehicle meeting Tier 1 standardseither through
computer control of the engine, technology enhancements, or engine tuning. My skepticism also
stems from the fact that the FTP, which is a fairly 'tame' driving cycle, was used to 'classify'
Tier 0 vehicles as Tier 1. The ultimate manifestation of this action could be mild to fairly
extreme bias in the mean emission response over non-FTP driving cycles of the "Tier 1" group of
enhanced vehicles, most likely biased high. My recommendation is to not enhance the sample, or
provide a much stronger justification and demonstrate that the Tier 0 vehicles used also were
"clean" on non FTP cycles compared to their Tier 1 counterparts.
EPA shares the authors concerns in this area. However, when it became clear that
MOBILE6 would need a set of speed correction factors for emission levels below the
average for Tier 0 vehicles, a reasonable approach to estimate the factors was needed.
In the absence of more actual data from Tier 1 vehicles, some approach would need to be
used. The low emitting Tier 0 vehicles did have low emission rates on the more extreme
facility cycles as well as on the FTP. The actual technical differences between Tier 0 and
Tier 1 vehicles is not large andean be largely arbitrary based on the model year in which
the vehicle is certified for sale. It should be noted that because the speed correction
factors are applied by emission level, not emission standard, that low emitting Tier 0
vehicles will be using the factors as well.
Page 17, 2nd paragraph: I agree with the statement, "the equations above would define a rational,
smooth relationship for emissions versus average speed for...". The statement deserves some
further caveats, however; we might expect a piece-wise linear relation to occur when changing
roadway functional classes, vehicle classes, etc. In other words, a smooth line might not be
expected when comparing discretely different groups or traffic cases. However, we would (as the
authors assert) expect this to be the case for a homogenous vehicle class on a facility. The authors
then point out that this does not in fact happen in the case of freeways. It is problematic that the
Final M6.SPD.002 140 June 2001
-------
models predict two different values for the same average speed on the same facility. I
recommend truncating the portion of the curve where the relationship is in doubt, which is
essentially using the stated rule 1 to omit the portion of the curve that is not used.
This is in fact what was done for the final version of the model.
Page 17, 1st list item: An accompanying figure would be nice to show here, as the text is a bit
difficult to follow. The idea seems acceptable, however.
Page 19, Emission Offsets: The concept of emission offsets seems to be reasonable, however;
I'm not sure why the US EPA has replaced the use of the FTP as the BER with the FTP + EO.
This should be explained better in the text.
*** My bigger concern, however, is that Table 14 shows emission offsets computed by
differencing the means of Fwy and LA4 driving cycles without taking into account the variability
in the means of these tests. How many of the offsets are statistically significant, and how many
are spurious? The offsets for level 1 THC and NOx, and level 3 NOx seem to be rather small, but
one cannot tell 'how small' without knowing the variability in means. It is troubling to employ
emissions offsets when in fact the offset could be in fact could be in the reverse direction. The
authors should compute 95% confidence intervals on the offsets and only employ offsets whose
values are convincingly significant and that can be theoretically justified.
Whether the emission offsets are statistically significant or not will not affect the
calculations of the speed correction factors themselves, since they assume a base
emission level on the freeway speed correction curve. This fact is taken advantage of in
MOBILE6 by allowing the emission offset to change, due to factors such as the SFTP,
without significantly affecting the speed correction factors. The emission offsets should
be considered as part of the basic emission rate and any uncertainty included in the
uncertainty of the basic emission rate. The SFTP report mentioned above discusses the
emission offset in more detail.
Page 22, NLEV standards: The authors seem to be approaching the NLEV issue intelligently, and
I have no comments about the proposed method other than keep updating the method as new
information about the performance of the NLEV's is forthcoming. An average approach seems
reasonable at this point in time. I would suggest that what will become more critical in the future
(I believe) is the ability to predict and detect failures of the cleaner vehicles and subsequent
remediation of them. As vehicles become cleaner, the difference between 'failed' vehicles and
clean vehicles will become larger, and thus failed vehicles will provide even more bang for the
buck (in relative terms) then they do now. How failures are modeled will also become important,
and how various I & M strategies can be modeled by adjusting failure rates will also become
important.
Page 30, Table 2: Need to define the terms SAFD Difference and High-Power Difference. It is
Final M6.SPD.002 141 June 2001
-------
presumed that these are statistics of the difference between the cycle and the FTP.
Page 31, Table 4: As stated previously, a companion table showing the distribution of vehicles
for the national fleet would be helpful for determining whether the sample data is representative
nationally.
National average technology distributions used in MOBILE6 are described in the report,
"Emission Control Technology Distributions, " (M6.FLT. 008). These statistics can be
summarized numerous ways, depending on the concerns of the reader. Readers should
refer to the complete description of the technology distributions.
Page 32-33, Table 5: Again, it seems that a comparison by manufacturer would be useful. For
instance, I only see one German vehicle in the sample; does the German proportion of vehicles
really represent about l/85th of the national fleet? Again, a companion or enhanced table could
show whether the sample of vehicles used for testing was representative.
Manufacturer designation is almost never used in generating emission estimates for the
MOBILE model. Comparison by manufacturer is only an indirect way to compare
technologies, which are the real parameters that most affect emissions. It is nearly
always better in emission modeling to project technology trends rather than to predict
trends in sales by particular manufacturers.
Page 35-37, Table 7: It appears that the standard deviation is larger than the mean for some
cycles and emitter classes, see for instance FTP for normal emitters, or Running 505. This
suggests a non-normal distribution of emissions, since negative emissions cannot result. This
supports the notion that the ANOVA test assumption of normality and homogeneity of variance
needs to be tested and shown (a comment detailed previously).
EPA concedes, without additional statistical testing, that emission distributions are not
normal. As noted, this makes sense, since emission measurements cannot be less than
zero on any sample vehicle.
*** Page 38-39, Table 8: The results in the table may be very deceiving and/or misleading. Of
course one of the objectives here is to identify which factors are "important" for classifying
fundamentally different emission conditions, whether they be vehicle, roadway, or environmental
factors. But since all the ANOVA results are done individually, and the results were not obtained
from controlled experiments, there are many factors that are correlated, and will subsume all of
the variability in emissions simply because they are correlated with the "real" culprit. As a vivid
(and perhaps overly simplistic example), if one were to compare emissions up a grade for a
vehicle with a driver only and with three passengers, one might correctly conclude in an ANOVA
that the presence of additional load would result in significantly increased emissions. If one also
collected information on the number of seatbelts buckled, say 1 in one case and 4 in the other, the
ANOVA would give identical resultsit would suggest that buckling seatbelts increases
Final M6.SPD.002 142 June 2001
-------
emissions. Of course it is known that the load caused the excess emissions. My concern is that
the whole gamut of potential factors has been 'thrown' into the ANOVA without regard for
potentially overlapping or confounding effects. My suggestion is to provide a table with a
correlation matrix of the factors used in the table (using the correctly computed correlation
coefficients for nominal and ordinal data), and provide some discussion in the text to justify
theoretically the factors thought to be important in the emissions process.
EPA carefully selected the parameters to be tested based on engineering judgement of
which parameters were expected to be most important. This choice was based on the
extensive experience of the staff involved and, hopefully, did not include spurious factors
such as those in the above example. Admittedly, a more thorough investigation would
include many other potentially important factors. However, the sample size suggested
that a more rigorous search for significant factors would be fruitless.
*** Page 58-61: Figures 2 and 3:1 am very concerned about these figures, and what they
potentially represent (analytically), and consider this to be a potentially fatal flaw in the proposed
methodology. The following discussion assumes that the figures are regressions through the ratio
of means. There are several deep concerns I have about these graphs: 1) incorrectly specified
random variables; 2) data aggregation bias; and 2) biased estimators of emission ratios. Each of
these is now discussed in detail.
1. The random variable of interest is the ratio of emissions, such that:
vr*-r n emissionscydex
oCr' u pi
t- /
where 9 is the ratio of two random variables. To find the average of the random variable
9, one would compute:
y_\vfffri
-------
ratios is (3/2+1/5+5/3)73 = 1.12, whereas the ratio of the means is [(3+l+5)/3]/[(2+5+3)/3] = 0.9.
Thus, the use of the ratio of the means (as opposed to the means of the ratios) will not produce a
good estimate of the desired quantity, theta, the ratio of cycle emissions to FTP emissions.
2. In addition to the problem specified above, the data have been aggregated (incorrectly as
ratio of means) prior to regression. The data should be regressed using the original ratios from
the 85 vehicles (or appropriate subsets thereof). Aggregation problems have been identified in
previous research, and have been shown to result in incorrect relationships in the regression (the
classic example is aggregation of trip generation by traffic analysis zone, which results in an
incorrect sign of the relationship between trip generation and household auto ownership), and in
inflation of R-Square and regression model statistics. Aggregation of the data prior to regression
"throws away" the variability inherent in the data, and presents false confidence in the output.
3. Finally, the variable theta(hat), which represents the ratio of two random variables
(emission test results), is in fact a biased estimator of the true population parameter . An unbiased
estimator, obtained through the method of statistical differentials, is obtained as follows:
T\VAR[e,rr]\
For additional information on the method of statistical differentials consult Hauer, 1997,
"Before-After Studies in Road Safety, Pergamon.
Given these three important considerations, I recommend re-doing all regression equations that
have used Ratio of Means with regression on original ratio units using all data points. I would
omit the correction for bias (item 3), which should be neglible for large variance conditions, but
would comment on it in the text of the bias of the parameter theta. Again, I find this to be a
potentially large error in analysis, which could significantly alter the results of the analysis.
EPA agrees that regressions through the ratio of the means would present problems.
However, careful reading of the report should show that regressions used for the speed
correction factors were run on the emission levels of the individual vehicles themselves
and not on the ratios or the ratios of the means. These figures were included only as a
visual demonstration of the apparent effect of average speed (as defined by the driving
cycles) on emissions based on the hot running LA4 cycle as a first step in understanding
the analysis methodology. These figures are not a demonstration of the actual regression
results, which are shown in Figure 4.
Potential Long-Term Improvements:
Page 1, bullets: Where do 2-lane highways fit into the picture? It is known from operational
aspects of traffic that 2-lane highways have very different emergent traffic than do interstates
Final M6.SPD.002 144 June 2001
-------
with more than two lanes, due to a number of factors including limited access, truck activity and
restrictions, passing maneuvers, and weaving. The entire paper has not mentioned 2-lane
highways, and perhaps should address somewhere how this is being handled. In the longer term
perhaps some empirical data on two-lane highways could be collected to determine whether they
should server as a "separate facility type".
Page 6, paragraph 1: The paper states, "These congestion levels have been roughly grouped into
"levels of service" LOS using letters A through G, similar to congestion category designations
used in transportation models.
The definition of level-of-service (LOS) used to 'bin' vehicular activity used by EPA and CARB
is different than LOS used by traffic engineers and transportation planners. LOS has been
assumed (by air agencies) to be an attribute of an entire facility (as viewed from a platoon of
vehicles), not segments of a facility as defined by traffic engineers and planners. Traditional LOS
categories A through F have been used to bin vehicular activity, despite the fact that these
categories may not optimally separate characteristically different emissions-producing vehicular
behavior. Only density has been used to 'bin' vehicular activity, and it has been determined
through windshield observation. There are many factors that engineers used to compute level of
service such as number of lanes, speed, flow, percent truck volume, type of freeway section,
percent of weaving volume, etc. Future work should focus on improving the "gap" between the
two working definitions of level of service. A recent paper by Debbie Neimeyer at UC Davis has
quantified the vast difference between LOS as approximated by the CARB and EPA methods of
car-following cycle development, and has shown that the imprecision of this measure of LOS is
significant.
Page 7: 1st paragraph: It seems an awful lot to ask from 85 vehicles to infer emissions for a fleet
of millions. Perhaps future testing programs could incorporate a larger number of vehicles in the
testing programs.
The data available and the methods for collecting and analyzing new data have been
changing rapidly. EPA is seriously investigating methods that will not require fixed,
laboratory driving cycles to estimate the emissions from highway vehicles at a variety of
operating conditions, such as speed, load and acceleration. These new methods will be
able to better address the above issues in future versions of EPA models. We look
forward to working with everyone in the stakeholder community in developing these
methodologies.
Final M6.SPD.002 145 June 2001
-------
Appendix E : Peer Review of Off-Cycle Effects
MEMORANDUM
To: US Environmental Protection Agency
Office of Mobile Sources
From: Randall Guensler
Trans/AQ, Inc.
Date: June 10, 2000
Review of Off-Cycle Correction Factors for MOBILE6
This report constitutes a review of the EPA report entitled: Determination of Off-Cycle and
Supplemental FTP Control Modeling in MOBILE6 (Draft Report M6.SPD.005)." Hereinafter,
this EPA report is identified as the "Off Cycle Report." Randall Guensler of Trans/AQ, Inc.
conducted this review under contract with the USEPA. The Off-Cycles report content as well as
the data and statistical methods used to develop the documented relationships were the primary
components reviewed. However, the methods and assumptions employed in developing off-
cycle corrections inherently tie to the methods employed in developing EPA's new cycles-based
correction factors and engine start emissions rates. In performing a review of the Off-Cycle
Report, it was also necessary to review the methodologies outlined in the EPA report entitled:
Facility-Specific Correction Factors (EPA-420-P-99-002). Hereinafter, the second EPA report is
identified as the "Cycle Correction Factor Report."
Off-Cycle Emissions
Instrumented vehicle studies have indicated that the grams/second emissions of CO and HC from
well-maintained vehicles remain low and stable under conditions of nominal engine load.
Exhaust emissions under nominal engine load conditions do scale with engine rpm (exhaust gas
throughput) but remain at low levels. Given the inherent variability in second-by-second
emissions at these low levels, an average emission rate can represent the emission rates under
nominal load conditions very well. Once engine loads increase past certain threshold levels
(where engine load is a function of speed, acceleration, vehicle weight, grade, wind resistance,
tire rolling resistance, and accessory loads), engine computers respond by enriching the airfuel
mixture.1 The enriched mixture can increase emissions by orders of magnitude for short periods
of operation. Under enrichment conditions, engine power output increases, and peak combustion
temperatures and peak combustion and exhaust manifold gas temperatures drop. Hence,
manufacturers program engine computers to undergo enrichment to improve on-demand vehicle
1 Further, when engine loads begin to cycle rapidly from loaded to unloaded conditions (such as under conditions of
throttle dither) emissions also increase significantly.
Final M6.SPD.002 146 June 2001
-------
performance and to protecting valves, rings, and catalysts under high load conditions. Engine
computer programs may also use other input variables, such as rate of change of throttle position,
as an indicator that increased power output is requested and therefore send a signal to undergo
enrichment. Enrichment can also occur as the result of faulty sensor input. Under extreme
enrichment conditions, NOx emissions can drop significantly as a function of the drop in peak
combustion temperature. However, under most operating conditions, NOx emissions in
grams/second tend to scale well with engine load (NOx emission rates are a function of
temperature and pressure of combustion). Computer-programmed enleanment (and increased
NOx emissions) can occur for some vehicles under conditions of extended cruise when
manufacturers have programmed the vehicle to run lean to improve fuel economy. In addition,
the demands of air conditioning increase engine load and cause vehicles to experience increased
CO, HC, and NOx emissions.2
Modeling Paradigm
The key with any in-use vehicle emissions modeling approach is simultaneously accomplish two
goals: 1) adequately represent the stable baseline emission rates in the model, and 2) represent
the cause-effect relationships that result in significantly higher or lower emission rates under
environmental and onroad operating conditions that differ from those experienced during
collection of baseline emission rate data. These two components must be prepared
simultaneously so that the baseline emission rates and the emissions modifiers are appropriate for
the group of vehicles modeled. That is, if there is little variability baseline emission rates within
the group "fuel injected vehicles" and if all fuel-injected vehicles respond similarly to changes in
external operative variables (such as changes in acceleration rates), the baseline emission rates
and correction factors are likely to be appropriate. However, if there are subgroups of fuel-
injected vehicles that exhibit statistically significant differences in baseline emission rates or
responses to changes in environmental/operating parameters, then the subgroups need to be
modeled separately. Identifying these mutually exclusive technology groups requires a great deal
of up-front statistical analysis. In reviewing the EPA documents, the methodologies employed
are compared to this modeling paradigm. Deviations from the methods described above are
reviewed and the potential impact of each deviation is qualitatively assessed.
Baseline Emission Rates
The LA4 cycle is composed of the 505-second and 866-second dynamometer cycles employed in
the FTP. The full FTP test consists of an LA4 test (with the 505-second cycle being conducted in
cold start mode and the 866-second cycle conducted in hot stabilized mode), followed by a repeat
of the 505-second component of the LA4 cycle under hot start conditions. The report sometimes
uses LA4 and FTP composite emissions interchangeably. EPA staff should clarify the definitions
2 Some impact may result from cycling of the compressor and engagement/disengagement of the a/c clutch, and
transient changes in EGR or other parameters potentially affected by the cycling. More studies are needed in this
area to fully understand the cause-effect relationships at work..
Final M6.SPD.002 147 June 2001
-------
of the LA4, FTP, and the relationship between the two, throughout the Off-Cycle Report.
In testing the 85 vehicles in the Cycle Correction Factor study, EPA and EPA contractors also
tested all 85 hot-stabilized vehicles on the 505-second component of the LA4 (the FTP Bagl and
Bag3 cycle). Normally, certification laboratories would only test these vehicles on the 505-
second cycle under cold transient (FTP Bagl) and hot transient (FTP Bag3 conditions). EPA
performed the hot-stabilized test so that incremental engine start emission rates (grams/start)
could be developed for the 85 vehicles (see EPA's documentation associated with MOBILE6
engine start emission rate development). Because hot stabilized FTP Bagl and FTP Bag2 were
available for these 85 vehicles, EPA staff computed a Hot Running LA4 test result for each
vehicle by adding these two test results together (in grams) and dividing by the miles traveled on
the LA4 cycle. The exact methodology used to develop the hot stabilized LA4 base emissions
rates in MOBILE6 should be stated in both the Off-Cycle and Cycle Correction Factor reports as
they impact the application of the algorithms discussed in these reports. The report should
contain language similar to that provided by EPA staff below:
"When we refer to a hot running LA4, we mean the LA4 cycle run without any starts
at all. This should be equivalent to running all three bags of the FTP without a start,
but the results of Bagl and Bag3 should be the same. So, instead you can just add
a hot running 505 to Bag2 of the FTP to give a hot running LA4. The "trip" in each
case (FTP and HRLA4) would be the LA4 cycle (7.5 miles), but the FTP would
contain engine start effects (43% cold and 57% hot) and the HRLA4 would contain
no starts at all. So you could use the FTP weightings and substitute the FIR505 for
Bagl and Bag3, but this should be the same as just adding the grams in the HR505
and Bag2 (Brzezinski, 2000)."
To develop hot stabilized LA4 MOBILE6 baseline emission rates, the EPA developed a model to
predict hot stabilized 505 emission rates from FTP test results (using the data collected from the
85 vehicles in the cycle correction factor database). The EPA engine starts report indicates that a
linear model was developed to predict hot stabilized Bagl emissions as a function of FTP Bagl
transient, Bag2 hot stabilized, and Bag3 transient test results. This way, EPA staff could
generate an artificial hot stabilized LA4 emission rate using the FTP Bag2 emissions and the
artificial Hot Stabilized Bagl for all vehicles in the comprehensive emissions testing database.
Again, the methods employed need to be clearly defined in the Off-Cycle and Cycle Correction
Factor reports.
There are two distinct advantages of using FTP emission rate data for developing baseline
emission rates. First, the EPA has maintained a continuous testing program (although spotty at
times) that has resulted in a database of more than 23,000 in-use vehicle tests for which FTP
emission rates are available. Second, the FTP test cycle is gentle enough to keep most vehicles
from undergoing enrichment or enleanment. Because the FTP cycle does not usually induce
enrichment, due to moderate speeds and low acceleration rates, researchers can more readily
identify and quantify differences in emissions behavior for vehicles across the FTP test and
Final M6.SPD.002 148 June 2001
-------
alternative test cycles. Of further importance, the FTP Bag2 cycle is even less likely to induce
enrichment than the LA4 cycle because the 505-second sub-cycle contains a significant initial hill
of acceleration that is more likely to induce enrichment. Hence, I believe that using Bag2 as an
emissions baseline is a sound modeling approach and better for noting systematic changes in
emissions behavior than the LA4 cycle. In fact, the MEASURE model even predicts higher
emissions on the FTP 505-second hot stabilized Bagl activity profile relative to the FTP Bag2
activity profile.
It is important to note at the outset that the FTP composite emission rate is composed of
contributions from a cold transient test, a hot stabilized test, and a hot transient test. Hence, the
engine and catalyst behavior during warm up stages of operation affect Bagl and Bag3 values
and therefore the composite emissions rate. In MOBILE6, start emissions are being modeled as
an increment (grams/start) and separated from the hot stabilized emissions component. Because
EPA was developing incremental engine start emission rates in MOBILE6, EPA staff made the
sound decision to avoid using the FTP composite emission rate as an emissions baseline. EPA
selected the hot stabilized LA4 as the emissions baseline because it would better reflect onroad
activity than other cycles. The LA4 emission rate can serve as an adequate baseline in modeling.
With LA4 as the baseline, the model needs to be able to predict vehicle emissions response when
vehicles experience nominal engine load conditions (such as those resulting on FTP Bag2 cycle).
Such nominal engine loads and lower emissions rates would be expected to occur when traffic
calming TCMs are implemented. The important consequence of using the LA4 as the emissions
baseline is that there is no comprehensive database of hot stabilized 505 emission rates. These
emission rates must be predicted for the comprehensive emission testing database as a function
of the composite emission test results. The basic problems with doing this are: 1) there is a
question as to whether the 85-vehicle data set used to develop these relationships is
representative, and 2) there is a large variability in vehicle-to-vehicle emissions response across
transient versus stabilized test conditions.
1. Representativeness of Sample Fleet - Based upon the statistical analyses associated
with MEASURE model development, it is apparent that the 85 vehicles employed in
MOBILE6 model development are not representative of the vehicle fleet. There is a good
distribution of model years and fuel-delivery technologies. However, the data set does
not control for other technology variables that Georgia Tech has found to be statistically
significant in terms of establishing baseline emission rates and emission responses.
Given the resource constraints that EPA has faced in collecting in-use emissions data for
use in MOBILE6 model development, the 85-vehicle is probably the best that could be
expected. However, when a vehicle sample set does not adequately control for the
variables suspected of being involved in cause-effect relationships, the real-world
statistical prediction intervals of the model cannot be determined. All of the statistical
tests associated with confidence bounds assume that a representative sample has been
analyzed. The impact of a non-representative sample fleet cannot be determined
statistically. Hence, the uncertainty associated with using the 85-vehicle data set cannot
be ascertained. It is impossible to determine whether the use of the 85-vehicle data will
Final M6.SPD.002 149 June 2001
-------
overestimate or underestimate baseline LA4 emission rates, or whether the emissions
response equations (i.e. cycle correction factors) will overestimate or underestimate the
real-world emissions response.
2. Large Variability in Vehicle-to-Vehicle Emissions Response Under Transient
Conditions - In conducting engine start emissions analyses for the MEASURE model,
Georgia Tech staff analyzed the contributions of Bagl, Bag2, and Bag3 emissions to the
composite emissions for normal and high emitting vehicles. GT analysts noted that the
major contribution to high emitter status often came from the cold and/or hot transient
tests, but for many vehicles, the contribution came from all three bags. There was a large
variability associated whether the transient tests or the hot stabilized portions of the tests
contributed to high composite emission rates and pushed the vehicle into high-emitter
status. Unfortunately, many of the suspected causal links could not be assessed because
second-by-second emissions were required to determine when these vehicles achieved
catalyst light-off. The issue that arises here is that the derived statistical relationship that
EPA is using to predict Hot 505 emissions from FTP Bagl, Bag2, and Bag3 is suspect.
All of the MOBILE6 modeling methods that follow the development of the LA4 emission
rates (cycle correction factors, off-cycle corrections, engine start algorithms, etc.) are
contingent upon the viability of the relationship between the FTP tests and LA4 tests.
It would have been more appropriate to use the FTP Bag2 (hot stabilized) emission rate as the
baseline emissions rate for all modeling work. True, emissions are lower on FTP Bag2 than they
would be under a Hot Stabilized 505 test, but this is not critical. As discussed earlier, statistical
analyses are designed to develop corrections from a baseline. It is advantageous to select a cycle
with driving conditions that do not induce enrichment/enleanment so analysts can identify the
factors that induce such enrichment/enleanment. The LA4 cycle contains higher load conditions
than FTP Bag2 and therefore provides less stable test results from which to determine correction
factors. Plus, hot-stabilized emission rates on the LA4 are simply not available for the vehicle
fleet and have to be derived from FTP Bagl, Bag2, and Bag3 data. Given the methodologies and
limited data employed, it is not possible to forecast whether the FTP baseline emissions are likely
to be biased high or biased low.
Off-Cycle Corrections
The Freeway Level of Service (LOS) F test cycle represents the activity likely to be experienced
by vehicles under congested freeway conditions (Off-Cycle Report, page 3). Both the LA4 test
and the LOS F cycle have approximately the same average speed of 19 mph. However, the
measured emission rates for almost every vehicle in the 85-vehicle data set under Freeway LOS F
conditions were significantly greater than their corresponding LA4 emission rates. In effect, the
off-cycle report indicates that the increase in emissions results from the fact that the congested
freeway test cycle conditions are more strenuous than the FTP or LA4 test conditions. That is,
the emissions increase results from the testing activity that is outside of the boundary of FTP or
Final M6.SPD.002 150 June 2001
-------
LA4 test cycles.3 The Off-Cycles Report states that a correction factor should be applied to the
baseline LA4 data to account for the higher level of emissions noted for the freeway test cycle
with the same average speed as the LA4 cycle. Hence, EPA staff developed an off-cycle
correction factor to elevate the FTP emissions to the levels noted at the desired baseline testing
condition. This way, cycle correction factor statistical analyses would generate a correction
passing through an emissions ratio of 1:1 for the Freeway LOS F cycle (average speed of
19mph).
In reviewing both the Off-Cycle Report and the Cycle Correction Factor Report, it was not
possible to determine the point in model development at which EPA staff determined an off-
cycle correction factor was required for developing MOBILE6. The Off-Cycle Report indicates
that there is a need to account for emissions that occur under operating conditions that differ
from those experienced in the LA4 test. The Off-Cycle Report also argues that there is a need to
correct the LA4 emissions to the emissions level observed on under the Freeway Level of Service
(LOS) F testing cycle. The Off-Cycle Report then states that the Cycle Correction Factor must
equal 1.0 at these testing conditions. In contrast, the Cycle Correction Factor report simply starts
with the premise that an off-cycle correction was to be applied to all FTP data for the 85 vehicles
before model development was to proceed (Cycles Report, page 18 and 19).
Although both reports argue that an off-cycle correction factor would be required, neither report
justifies the creation of such a correction. Given the nature of correction factor development,
there really is no compelling need to create such an off-cycle correction factor. Correction
factors are applied to a baseline emission rate, irrespective of whether the baseline emission rate
is an FTP Bag2 or a Hot Stabilized LA4. Introduction of an off-cycle correction factor to the test
results that are subsequently used in development of cycle correction factors is unnecessary. In
fact, such a correction is detrimental to the development of the cycle correction factors. This is
because the first correction factor algorithm yields a predicted value that must by its nature
reduce the variability in emissions response for a variable that is employed in the next algorithm
development process. It would have been more appropriate to model the net effect of the two
correction factors within a single cycle correction factor (statistical reasons are described in the
next section). True, the new cycles correction factor would not equal 1.0 at 19mph on a freeway,
but this is not a necessary condition for correction factor algorithms.
Assessment and Validation of Off-Cycle Analytical Methods
It was not possible to reproduce the analytical results presented in the Off-Cycle Report using the
emission rate data provided by EPA staff. Presumably, this is because the database provided
only contained test results for 84 of the 85 vehicles indicated in the report. The fact that the beta
coefficients were significantly different between the modeling runs for the 84 and 85 vehicle data
sets is a reflection of the inter-correlation of the LA4 and LA4-squared variables employed in the
3 The emissions increase may also result from other factors such as increased throttle dithering that affect computer-
controlled enrichment.
Final M6.SPD.002 151 June 2001
-------
model. The standard errors of the model beta coefficients are extremely high under conditions of
independent variable correlation. Hence, there can be large fluctuations in the beta coefficients
developed from model run to model run, as a function of the actual vehicle data employed in
model development. However, the predictions from two different models that use a 95% subset
of the original data can still be similar, even though the beta coefficients are wildly different.
This makes interpreting the meaning of the beta coefficients difficult. It is best to think of the
EPA off-cycle correction factor modeling approach as a curve fit rather than an explanatory
model.
The fact that the statistical analyses did not indicate statistically significant differences in vehicle
emissions as a function of model year, fuel delivery system, trucks vs. cars, etc. is a reflection of
the small sample size. Emissions variability within same-vehicle tests, let alone across vehicle
tests, is very large. The report should reflect that there are likely significant differences in
emission response across the various vehicle technology groups, but the database is of
insufficient size to distinguish and separate any systematic effects across these variables from the
highly variable emissions response.
EPA staff used engineering judgement to prevent the algorithms from predicting negative offset
increments. EPA staff clipped the quadratic form for high LA4 emissions levels. This way the
equation retains a maximum offset level, rather than predicting a decline in the offset for higher
baseline LA4 emissions levels. The report provides no scientific theory (i.e. cause-effect
relationships related to emissions production) to support the functional form of the model (hot
stabilized LA4 and LA4 squared) or to support the decision to truncate the model for high-
emitting vehicles. It is not clear that "off-cycle" emissions increments would not decline or even
go negative for certain high-emitting vehicles when onroad operations move from LA4 driving to
freeway LOS F driving (especially for NOx). Despite the mention in the report that there are few
onroad vehicles with sufficient emissions levels to invoke the clipped portion of the algorithm,
the validity concern that should be addressed through additional analyses.
Assessment of Corresponding Cycle Correction Factors
In developing the MOBILE6 mobile source emission rate model, the USEPA has attempted to
significantly improve the general modeling approach for predicting emissions as a function of
on-road vehicle operating conditions. Previous versions of MOBILE modeled emission rates as a
direct function of average speed, irrespective of actual on-road operating conditions. Hence,
previous model versions predicted the same emissions rates for vehicles at 25 mph, regardless of
whether the vehicles were operating under congested freeway conditions or free-flow arterial
conditions. Researchers have identified numerous theoretical and technical flaws with the
previous average speed modeling approach (see Guensler, 1993 and the CRC review of EMFAC
conducted by Environ). MOBILE6 attempts to dis-aggregate on-road emission rates by
integrating two new explanatory variables into the existing speed-related modeling framework:
facility type and level of service condition (reflective of congestion conditions). To develop the
new corrections, EPA staff contracted with Sierra Research to develop new testing cycles
Final M6.SPD.002 152 June 2001
-------
designed to be representative of the operating conditions on various facility classes (e.g.
freeways, ramps, arterial) under different level of service conditions. EPA and EPA contractors
used these new cycles to collect laboratory emissions data from 85 vehicles. Then, EPA
contractors derived cycle correction factors, or new relationships between emission rates and
operating conditions specifically associated with a facility class and the congestion conditions
noted to occur at various levels of service on these facility types. The cycle correction factor
approach disaggregates the MOBILESa relationship between emissions and average speed,
providing a significant theoretical improvement over the previous speed correction factor
modeling approach.
Cycle-based correction factors are multipliers to baseline emission rates. That is, MOBILE6
predicts emissions under onroad driving conditions as a multiple of the emissions occurring
under baseline testing conditions. To develop cycles-based correction factors, EPA developed
statistical relationships between measured emission rates on test cycles designed to represent
operating conditions on different facility types and level of service conditions and the calculated
(and then off-cycle-adjusted) LA4 baseline emission rates. If laboratory tests yielded significant
increases in emissions when vehicles changed from baseline-like operations to alternative
operating conditions, the model algorithms reflected the effect in the cycles-based correction
factors (by roadway class).4
In MOBILE6, the cycle correction factors were determined as a ratio of predicted average
emissions (average predicted emissions on a freeway test cycle divided by the average predicted
emissions on the Freeway level of service cycles). These predicted emissions are based upon the
results of a regression analysis for the 85 vehicles that simply predicted grams/second emissions
(by emitter class) as a function of average speed for the test cycles used to collect the data.
Before running these regressions, the emissions data were corrected by the off-cycle emissions
offset. The basic methodology employed in developing the cycle correction factors is a problem
unto itself. The prediction of gram/second emissions, calculation of the ratio of predicted
emissions for each cycle, and then the subsequent prediction of the relationships of the emissions
ratios to average speed is a fundamental flaw in the modeling approach. Such estimations toss
away all vehicle-to-vehicle emissions response variability. The best fit curve is forced through
the mean of the ratio of the average predictions for each cycle test result set, rather than being
allowed to provide a best overall fit to the entire range of vehicle emissions responses expressed
4 Strictly speaking, MOBILE6 models emissions as a function of the level of service conditions reflected in the test
cycles used to collect the laboratory data for cycles correction factors. MOBILE6 does not predict emissions as a
function of average speed. Instead, average speed plays a role in a best-fit interpolation between these test cycle
conditions. EPA staff simply plotted the average emissions from each testing cycle by average speed of testing
cycle. Then, staff derived a best-fit interpolation between averaged values of the test cycle results as a function of
average speed. Although the modeling approach provides significant improvement over the average speed regime
in MOBILESa, EPA should caution MOBILE6 users that the average speed relationships in MOBILE6 are still
highly uncertain. If bootstrap regression and Monte Carlo analysis were performed on the derived relationships,
both the average emissions response value for each LOS test cycle and the curve fits between the test cycle results
would exhibit wide confidence bounds (see Guensler, 1993 for more information on this modeling technique).
Final M6.SPD.002 153 June 2001
-------
across all of the cycles. Given the methodology employed, it is impossible to perform a standard
statistical assessment to determine if the modeling approach is more likely to be biased high or
biased low. The algorithms could be assessed using a combination bootstrap and Monte Carlo
assessment approach, but the task would be very time consuming.
With respect to the role of the off-cycle correction, one is to assume that if a cycle other than
freeway LOS F had been selected as the off-cycle offset baseline, the same cycle-correction
factor curves would have resulted. That is, the predicted incremental increase in baseline
emissions to account for off-cycle effects will not affect the derivation of cycle correction factors.
If appropriate statistical techniques were employed that retained response variability, this would
be highly unlikely, given the nature of the small data set and highly variable emissions responses
across these vehicles and test cycles. However, EPA staff can readily check this hypothesis. The
question is how significant the differences will be. EPA staff should replicate the model
estimation approach used to develop the off-cycle correction and the cycle correction factors
using a different assumed off-cycle correction baseline. The alternative model would use the
emission test results from freeway LOS B or LOS C cycle (instead of the LOS F cycle) as the
basis for developing the LA4 off-cycle increment. The same multi-step procedure would be used
to generate the alternative off-cycle correction (as a function of LA4 and LA4 squared) and the
subsequent cycle correction factors. If the model estimation approach is stable, this procedure
will yield the same net predictive algorithm for the range of LOS conditions. That is, when EPA
staff apply the original and alternative model to the 85-vehicle data set (or any validation data
set) both models will predict approximately the same emission rates for any specified operating
condition. The issue of whether the 85-vehicle fleet provides representative responses will still
be a major concern. However, even though standard statistical methods cannot be applied to
assess the adequacy of the current MPOBILE6 algorithms, at least the modeling approach taken
will be more defensible if the original and alternative modeling achieves the same correction
factors.
Correction of Correction Factors
One issue that is a major concern with the modeling methods outlined in MOBILE6 is the
interrelationships between the various algorithms and correction factors employed. The
presumption is that all of the relationships developed in a stepwise fashion from the 85-vehicle
test are applicable to the fleet.
o MOBILE6 hot stabilized baseline LA4 emission rates are predicted from FTP
transient data.
o Off-cycle correction factors (as a function of LA4 and LA4 squared) are applied to
the predicted LA4 test result to generate a corrected LA4 baseline.
o The various correction factors do not always account for high-emitter status, Tier
1 vs. Tier 0, and their interactions, meaning that the "technology groups"
Final M6.SPD.002 154 June 2001
-------
employed in the offset development are different than those used in the cycle-
correction factor development.
o The predicted corrected LA4 baseline emission rate is multiplied by an emission
ratio that is derived from predicted and adjusted gram/second emissions as a
function of average speed.
Every time a predicted variable is used as an explanatory variable in a subsequent algorithm, data
variability is reduced. That is, the error terms associated with the first algorithm, representing
random error and problems with model specification, do not carry forward into the development
of the subsequent algorithm. As such, each modeling step appears to provide a more systematic
response and better statistical fit. Using predicted variables as independent variables is not a
significant problem, provided that no important explanatory variables are omitted in the primary
steps. However, when a model has a large error element relative to the model output signal, such
as the case in emissions testing, it is difficult to argue that all of the important variables have
been included. This issue provides a compelling reason to derive all of the correction factor
relationships simultaneously.5
The complex step-wise modeling approach used in developing MOBILE6 results in a final model
structure that cannot be falsified by field experiments. Given the number of general assumptions,
and corrections to corrections employed in the modeling routines, it will not be possible to
pinpoint the source of error when the model is determined to over-estimate or under-estimate
emission rates under certain operating conditions. Simultaneous development of the modeled
relationships (which would have admittedly required a larger testing data set than EPA had
available) could have avoided the problems noted above. With simultaneous development of
baseline emissions rates and correction factors, the emissions from various subgroups of the fleet
under different operating conditions could be examined through ongoing roadway and laboratory
experiments.
There is an important bottom-line conclusion that arises from review of the various modeling
methodologies employed in developing MOBILE6. EPA management needs to be aware that
emission rate models cannot improve significantly until EPA has acquired the necessary
resources to test many more vehicles under testing conditions that differ from the FTP. As a side
note, it is not nearly as important that these laboratory tests be representative of real-world
driving as it is that these tests reflect the range of operating conditions that vehicles experience in
the real world. Statistical modeling approaches can generate appropriate baseline emission rates
and correction factors once second-by-second data under alternative testing conditions are
available.
Supplemental FTP Emissions Effects
5 Simultaneous development of baseline emission rates and modal correction algorithms was the basis of
MEASURE model development.
Final M6.SPD.002 155 June 2001
-------
By integrating two new emissions testing cycles into the FTP certification process, EPA is
effectively raising the bar for new vehicle certification. The first supplemental test, the US06
cycle, contains significantly higher speeds and acceleration conditions than the existing FTP.
The second supplemental test, the AC03 cycle, contains harder acceleration rates and must be
conducted while the vehicle's air conditioning is operating. The higher load test conditions and
the operation of air conditioning put significantly greater engine loads on the tested vehicles.
Consequently, the SFTP provides more opportunity for vehicles to undergo sustained high load
conditions (leading to higher NOx) and power-demand enrichment (greatly increasing CO and
HC emissions). The change in the test method, designed to capture more emissions from in-use
vehicles, results in a defacto change in emissions standards (even though the actual gram/mile
compliance limits remain unchanged. If manufacturers do not implement additional control
strategies, many of their vehicles will fail the SFTP certification test. The question that arises in
the Off-Cycle Report is what emission benefits will result from implementing the new SFTP
test?
In principle, the emissions benefit of the SFTP is the difference between in-use emissions before
implementation of the new test method and in-use emissions after the implementation of the new
test method and composite SFTP standard.6 If one assumes that manufacturers will comply with
the regulations, but will not provide significant reductions beyond compliance, the net reduction
of in-use emissions is represented by change in vehicle activity reflected in the new testing cycle.
There is a subset of onroad vehicle activity included in the SFTP and not included in the FTP that
leads to elevated in-use emissions with the current fleet. Under the SFTP, these activities will
likely yield normal stabilized operating emissions. Hence, the SFTP emission rate benefits are
associated with the change in capture of emissions by the new cycles, multiplied by the overall
fraction of onroad vehicle activity that is included within the operating boundaries of the new test
conditions.
Quantifying the emissions benefits of the SFTP requires comparative testing of vehicles on the
facility cycles as well as the US06 and AC03 cycles. Reduction in off-cycle emissions could be
estimated by comparing the percentage reduction in emissions that will occur for current vehicles
as they move from their current emissions levels on the composite SFTP to the compliance
emission rates for the 5year/50,000 mile US06 standard of 0.65 grams/mile HC+NOx and
(perhaps lower than this level to account for the benefit of compliance headroom). One could
then make the argument that if the US06 and AC03 cycles contain the conditions in the facility
cycles that currently lead to enrichment, the onroad emissions will be reduced by a similar
percentage (or by a percentage weighted by the 65% contribution of these cycles to the composite
SFTP). In plotting the cycle characteristics of the various freeway level of service cycles, the
US06, and the AC03, it is clear that then new cycles contains a significant fraction of high-speed
activity that appears in the Freeway LOS cycles. The acceleration rates associated with the high-
speed ranges in US06 are also on par with the acceleration ranges found in the Freeway LOS
Cycles and the US06 contains harder decelerations at higher speeds and harder accelerations at
6 Composite FTP is determined as a weighting of 35% FTP composite, 37% AC03, and 28% US06 contributions.
Final M6.SPD.002 156 June 2001
-------
higher speeds. Three-dimensional Watson plots of fraction of operation under given speed and
acceleration ranges could be used to determine whether the US06 and AC03 cycles cycle contain
a wider range of load-inducing activity than the various level of service cycles. EPA could then
effectively argue that because the SFTP cycles (which contribute 65% to the SFTP composite
emission rate calculation) are more likely to induce off-cycle emissions than the LOS cycles, and
because manufacturers will comply with the SFTP, off-cycle emissions under all LOS cycles will
decline. However, EPA SFTP benefit estimation methods did not take such an approach. As
discussed below, it is difficult to follow the logical progression of the benefit calculation and it is
not possible to verify many of the assumptions employed in the methods reported in the Off-
Cycle Report.
For Tier 1 vehicles, the Off-Cycle Report indicates that the off-cycle increment7 will be adjusted
downward by 88%, 72%, and 78% for HC, CO, and NOx respectively. The argument put
forward is that the implementation of the SFTP will reduce the emissions difference noted across
the LA4 test and the freeway LOS F test by these percentages. The Off-Cycle Report cites to
"rule-estimated benefits," presumably contained in the Tier 1 rule development report as the
source for the reduction claims. However, the basis for these claims is unverifiable in the Off-
Cycle Report. No scientific theory or empirical data are provided to justify the values.8 It would
be beneficial to test these hypotheses using new modal emissions models applied to the cycle
tests and weighted by fraction of onroad activity expected to occur under each operating
condition. New modal models could also be employed to assess the likelihood that these
estimates are reasonable.
The implementation of the SFTP stands to flatten the future cycle correction factor curves in
MOBILE6 for the future vehicle fleet. As discussed earlier, there was no compelling reason to
prepare a linear bump-up in emissions and deem it to be an off-cycle correction factor. Similarly,
there is no compelling reason to create a linear bump-down to adjust this off-cycle increment for
SFTP implementation. This is because the effects of operations that differ in engine load from
LA4 (or any other baseline) will be reflected in the final curve that result from the statistical
method (or combination of statistical methods) employed. To the extent that operating
conditions in the cycle correction factor test cycles are now reflected in the SFTP, there is reason
to believe that future vehicles will behave differently on the new facility cycles used to generate
the cycle correction factors. The Off-Cycle Report's discussion and re-derivation of cycle
correction factors using the test results from 10 vehicles that behaved well on the ramp cycle
does not provide adequate evidence to assess the validity of keeping the proposed cycle
correction factors when the SFTP is implemented. This seems especially true considering the
highly aggregated modeling approaches (and averaging of predicted values) employed in
7 Calculated as the difference between the hot stabilized LA4 test results and the freeway LOS F test results and
then predicted as a function of LA4 and LA4 squared.
8 There also appears to be no basis for the 50% decrease in off-cycle emission reduction effectiveness for SFTP
controls that are coded into the MOBILE6 model (noted in Appendix A, Table 3, and Table 4). This aspect needs
to be discussed and justified in the report.
Final M6.SPD.002 157 June 2001
-------
developing the cycle correction factors. Had the baseline emission rates and cycle correction
factors been modeled in a single statistical operation, as suggested earlier, significantly different
relationships would likely have resulted and the cycle correction factors may have taken on
different shapes both pre- and post-SFTP implementation. Unfortunately, it does not appear that
a scientific and statistical method that can be used to test the relationships presented in the report.
As such, EPA should propose a comprehensive testing program designed to collect data from
new SFTP certified vehicles when they become available. The goal of the testing program
should verify the onroad emissions response behavior reflected in MOBILE6 both pre-and post-
implementation of the SFTP.
The methodology used to develop the LEV benefits from implementation of the SFTP is not
clear and concise. Although the calculation methods have been conveyed in the text and tables,
it is impossible to follow the logical flow of the multi-step procedure outlined in the report. The
first step splits the effective SFTP standards (identified as US06 in Table 2) NHHC+NOx
compliance levels into separate components cannot be evaluated (theory, logical reasoning,
empirical evidence, and appropriate references should be provided to support the splits). The
second step of estimating average 1999 certification emissions levels from the Certification and
Fuel Economy Information System (CFEIS) should be clarified and additional detail on the
number of vehicles tested should be provided. Presumably, the CFEIS tests reported are test
results on the SFTP (indicated by the provision of 4K and 50K mileage accrual values
corresponding to SFTP standards) and not the FTP.9 The third step of estimating running
emissions levels is not clear and requires separate evaluation of the referenced MOBILE6
exhaust emissions report to ascertain why the adjustment was made (0.9 for NOx and 0.23 for
HC). The fourth step consists of calculating the ratio of the SFTP effective emissions standard
(Step 1 result) and the running certification level for 1999 model year vehicles (Step 3 result). A
discussion as to why the ratio of SFTP standard to current new vehicle certification emissions is
lacking. In using this value in emissions benefit assessment, it seems that EPA is asserting that
the certification standard is somehow responsible for the compliance headroom noted in the
CFEIS and that such headroom is systematic and will remain consistent over time.
The logical reasoning behind the next set of steps in the benefit assessment process is completely
lost in the text. The goal appears to be to determine the increased stringency of the ARB LEV
standards relative to the EPA standards and then to adjust the SFTP benefits accordingly.
Presumably, the calculation is designed to represent how much cleaner in use LEVs will be under
the California LEV standard (at 4,000 miles) relative to the 49-state standard (at 50,000 miles).
That is, the certification of LEVs under the California program is expected to garner additional
emissions reductions. Tables 2 and 3 illustrate the calculation method employed to represent the
increased stringency of the ARB standard relative to the EPA standard. However, the text does
9 The Tables indicate that "FTP Certification Levels" are reported. If this is the case, another major issue arises.
The use of a ratio of future SFTP certification limits to current FTP running certification emission rates would be
meaningless in the calculations employed. There is no reason to believe that current FTP certification emission
rates will correspond to future SFTP test results.
Final M6.SPD.002 158 June 2001
-------
not explain the logic behind the calculation. While a more stringent ARB requirement is likely
to capture emissions that the EPA requirement leaves in the in-use fleet, no compelling argument
is provided that indicates why the calculation methodology makes sense.
As mentioned above, it seems that the calculation methods are based upon the assertion that
compliance headroom (represented in the ratio surrogate) in the certification database is
systematic and the result of the differences in certification standards. The method also inherently
assumes that such headroom will remain consistent over time. Even if EPA staff provided the
logical flow of the equation derivation, it is unclear whether the logic behind the calculation
methodology would hold into the future. As manufacturers produce more reliable and durable
vehicles in response to the SFTP mandate, the compliance headroom or safety margin currently
experienced in the certification database may drop significantly. Plus, it is unclear how these
ratios will change for LEVs versus the 49-state fleet. Furthermore, there is no way of knowing
whether these respective vehicle fleets will age similarly and whether the effect predicted from
SFTP calculations based upon 1999 vehicle data will still be appropriate for these vehicles in 5 or
6 years. Changes in these factors would be evidenced as significant changes to the ratios
calculated in Step 4, significantly influencing the predicted CARB-related LEV benefit. Without
additional information and explanation, it is simply not possible to evaluate the algorithms
provided in the Tables. Since it is unlikely that EPA will be able to answer the question of
benefit estimation stability over time, it imperative that a comprehensive testing program be
implemented to check these significant MOBILE6 assumptions over time and make corrections
based upon observation.
Similar to the methods used to develop benefit estimates for the US06 cycle, the text discussion
of AC03 benefits also lack sufficient documentation to perform an assessment of the
methodologies employed. The first problem is failure to support the determination of the EPA
SFTP AC03 benefit of 50% for NOx. As mentioned earlier, the justification for using the ratio
of AC Standard to Running Certification Level (or even certification level in the CFEIS
database) as a measure of relative effectiveness across the EPA and ARB certification programs
is lacking. The fact that CARB has required the elimination of "commanded enrichment" is
unclear. Presumably, CARB has prescribed that vehicles cannot be pushed into commanded
enrichment simply whenever the air conditioning is turned on (i.e. the issue that apparently arose
in the Cadillac dispute). However, the text is unclear as to the extent to which CARB has
prescribed the elimination of A/C induced commended enrichment. When the air conditioning is
in operation, a vehicle will undergo enrichment more readily unless an A/C clutch causes the air
conditioning to disengage under these conditions (is thus what CARB actually prescribed?). If
so, the assumption of HC eradication does not seem unreasonable. Scaling CO emissions with
vehicle load also seems reasonable (this probably could have been done for HC as well) where
vehicle load is predicted as a function of predicted fuel consumption increases. However,
equations used to derive the increase in fuel consumption as a function of speed and speed
squared are not supported in the text of the report. A multitude of alternative fuel consumption
equations could have been employed. The concern here, given the discussions earlier regarding
the representativeness of vehicles tested, representativeness of test conditions, and treatment of
Final M6.SPD.002 159 June 2001
-------
data in developing regression models, is that these relationships may or may not be justified. The
report should provide more detail on how the equations were developed and how the predicted
changes in fuel consumption are used to estimate changes in CO. The derivation of the actual
values in Tables 4 and 5 must be supported with additional text material so that the reader can
follow the logical flow from calculation method to calculation method. As it stands, reasonable
readers could not be expected to understand the reasoning behind the development of the benefits
estimates.
Conclusions:
The methods used to develop the off-cycle corrections, cycle correction factors, and SFTP
benefits employ multiple assumptions that cannot be verified. Modeled relationships are derived
from a small fleet that is not representative of the on-road fleet. Data treatment is such that
averaged values and predicted values are often used as independent variables in the statistical
techniques from which MOBILE6 algorithms are based. A number of relationships are modeled
as independent effects when they are actually co-dependent (e.g. off-cycle correction and cycle-
correction factors). Given these problems, it is not possible to apply standard statistical
techniques to the model derivation process to determine confidence bounds around the
algorithms employed in MOBILE6. Further, given the model development discussed above, it is
not even possible to assess whether the predicted mean responses are likely to be biased high or
low.
Because confidence bounds around the MOBILE6 algorithms cannot be generated, a combined
bootstrap and Monte Carlo assessment could be undertaken to determine inherent model
uncertainty. In such an analysis, each algorithm discussed in this report would be derived 1000+
times using subsets of the original data set (with replacement). The modeled relationships would
then be represented as probability distributions in MOBILE. Then, MOBILE would be run in a
Monte Carlo fashion to develop a distribution of model output results. Guensler and Leonard did
this for the MOBILESa speed correction factors; however, a similar assessment for MOBILE6
would be extremely labor-intensive. One is left to conclude that there is little for EPA to do to
assess the adequacy of the algorithms in MOBILE6 than to assess the mean squared error and
bias of the model in its application to validation data sets. When the algorithms are fully
integrated in MOBILE6, EPA should use the model to predict the emissions of a validation data
set and examine the prediction errors of the model, comparing them to the prediction errors of
MOBILESa and other alternative modeling approaches.
Recommendations:
o The actual definitions of the LA4, FTP, and the relationship between the two should be
clarified in the report whenever the terms are employed.
o The exact methodology used to develop the hot stabilized LA4 base emissions rates in
MOBILE6 should be stated in both the Off-Cycle and the Cycle Correction Factor reports
Final M6.SPD.002 160 June 2001
-------
as they impact the application of the algorithms discussed in these reports.
o The relationships between the methods used to develop off-cycle corrections and other
correction factors should be delineated in a text and graphic (flowchart) format.
o All methods need to be better documented so that the reader can follow both the logical
reasoning behind the proposed analytical method and the estimation techniques
themselves.
o EPA would be better off to use FTP Bag 2 as the baseline standard (rather than LA4)
from which to determine the emissions effects that result from changes in vehicle
operating conditions represented in the various emissions testing cycles employed in
developing cycle correction factors and SFTP effects. At the very least, MOBILE6 LA4
baseline rates would not need to be predicted as a function of individual FTP Bag test
results.
o EPA presumably pursued the development of an independent "off-cycle" emissions offset
to provide an estimate of the benefits that the new emissions testing cycles were likely to
achieve. Alternative analytical methods could have, and should have, been applied to
estimate the SFTP effect. The process employed in developing the estimates for
MOBILE6 contains too many inter-correlated effects from testing on the FTP, SFTP, and
cycle correction factor test cycles. Hence, the "off-cycle" offset cannot be reliably
predicted from the analyses undertaken.
o Cycle correction factors should have been developed directly from the laboratory test
cycle data, rather than after artificially bumping up baseline emissions levels. The cycle
correction factor would be more appropriate if derived from a single statistical analysis,
rather than a staged analysis. Even if EPA retains the current method to estimate SFTP
effects (for use in policy analyses), a single-step cycle correction factor approach would
provide better estimates of the effect of changes in traffic conditions on emission rates.
Final M6.SPD.002 161 June 2001
-------
Response to "Review of Off-Cycle Correction Factors for MOBILE6"
The response to specific comments made in "Review of Off-Cycle Correction Factors for
MOBILE6" are contained throughout Section 6 of this report. A summary of the primary
comments and responses follows:
Comment: The comments suggested a lack of clarity in discussing the relationship between
the off-cycle adjustment and speed correction adjustments, originally documented in separate
reports.
Response: In response to this comment, the discussion of off-cycle effects has been
incorporated into M6.EXH.002 for improved cohesiveness.
Comment: The comments criticized the application of two correction factors to address off-
cycle and speed corrections, saying it artificially reduced statistical variability of the overall
emission effects. The comments suggested that one correction be developed to address both
issues.
Response: A separate correction factor is necessary to isolate the effects of increased
emissions due to "off-cycle" driving, for the purpose of applying benefit from the SFTP rule.
Another source of increased emissions (on a grams per distance basis) is due to the "reduced
travel efficiency" which results at lower speeds (i.e., less distance traveled). Emission
increases due to the latter are reflected in the speed correction factors, but will not be reduced
by the off-cycle provisions of the SFTP.
Comment: The comments suggested that alternate facility cycles be used to derive an off-
cycle correction, in order to demonstrate the stability of the modeling approach used in
MOBILE6.
Response: In response to this comment, an alternate model formulation was developed by
deriving off-cycle corrections and speed correction factors from the Arterial E cycle rather
than the Freeway F cycle, and the results compared to the MOBILE6 approach. The end
results were less than 3 percent different, and were not statistically significant.
Comment: The comments suggested that the derivation of SFTP benefits a) were not
presented in a clear manner, b) were not based on an analysis of US06 data with and without
SFTP control, and c) were based on reductions not clearly justified
Response: In response to this comment, the source and derivation methodology of the SFTP
reductions have been clarified. The Tier 1 reductions were in fact based on an analysis of
vehicles with and without control measures intended to allow compliance with the SFTP.
The dataset used for this analysis (performed as part of the SFTP rulemaking) is the only
database containing vehicles which have been modified by manufacturers to demonstrate
Final M6.SPD.002 162 June 2001
-------
compliance with the SFTP. No such data exists for obtaining LEV benefits, hence an
analytical approach was required. The derivation of LEV benefits have been clarified.
Comment: The comments suggest that speed correction curves will change for vehicles
complying with the SFTP requirement, and questions the EPA analysis provided to
demonstrate the adequacy of the speed correction curves for SFTP vehicles.
Response: Speed correction curves cannot be directly estimated for vehicles complying with
the SFTP, since these vehicles have not begun to penetrate the market in substantial numbers.
The analysis presented in M6.EXH.002 was intended to show that the magnitude of SCFs is
not inappropriate for vehicles which would likely comply with the SFTP. This issue cannot
be fully researched until a robust set of SFTP-compliant vehicles are exercised over the speed
correction cycles; until this point, the analysis presented does provide an initial support for
the assumption that Level 1 SCFs are applicable to vehicles with low off-cycle emissions.
Final M6.SPD.002 163 June 2001
------- |