Development of Emission Rates for
            Heavy-Duty Vehicles in the Motor
            Vehicle Emissions Simulator
            MOVES2010

            Final Report
&EPA
United States
Environmental Protection
Agency

-------
       Development of Emission Rates for
       Heavy-Duty Vehicles in the Motor
            Vehicle  Emissions Simulator
                      MOVES2010

                        Final Report
                    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.
United States
Environmental Protection
Agency
EPA-420-B-12-049
August 2012

-------
1 Heavy Duty Diesel Emissions	3
1.1 Running Exhaust Emissions	5
  1.1.1 Nitrogen Oxides (NOx)	6
     1.1.1.1 Data Sources	7
     1.1.1.2 Calculate STP from  1-hzdata	8
     1.1.1.3 Calculate emission rates	12
     1.1.1.4 Sample results	18
  1.1.2 Paniculate Matter (PM)	22
     1.1.2.1 Data Source	22
     1.1.2.2 Analysis	24
     1.1.2.3 Sample results	32
  1.1.3 Hydrocarbons (HC) and Carbon Monoxide (CO)	36
     1.1.3.1 Data Sources	36
     1.1.3.2 Analysis	37
     1.1.3.3 Sample results	38
  1.1.4 Energy	42
1.2 Start Exhaust Emissions	44
  1.2.1 HC, CO, and NOx	44
  1.2.2 Paniculate Matter	46
  1.2.3 Adjusting Start Rates for Soak Time	46
1.3 Extended Idling Exhaust Emissions	48
  1.3.1 Data Sources	48
  1.3.2 Analysis	49
  1.3.3 Results	51
2 Heavy-Duty Gasoline Truck emissions	52
2.1 Running Exhaust Emissions	52
  2.1.1HC, CO, and NOx	52
     2.1.1.1 Data and Analysis	52
     2.1.1.2 Sample Results	53
  2.1.2PM	55
     2.1.2.1 Data Source	55
                                                                                        1

-------
    2.1.2.2 Analysis	56
  2.1.3 Energy	57
2.2 Start Emissions	58
  2.2.1 Available Data	58
  2.2.2 Estimation of Mean Rates	59
  2.2.3 Estimation of Uncertainty	61
  2.2.4 Projecting Rates beyond the Available Data	64
    2.2.4.1 Regulatory  class LHD2b3	65
    2.2.4.2 Regulatory  classes LHD45 and MHD	66
    2.2.4.3ParticulateMatter	67
A. Appendices	68
A.I Calculation of Accessory Power Requirements	68
A.2 Tampering and Mai-maintenance	70
  A.2.1 Modeling Tampering and Mai-maintenance	70
  A.2.2 Data Sources	72
  A.2.3 Analysis	73
A.3 Extended Idle Data  Summary	81
A.4 Regression to develop PM emission rates for missing operating modes	86
A.5 Heavy-duty Diesel EC/OC Fraction Calculation	87
  A. 5.1 Introduction	87
  A.5.2 PERE for Heavy-duty Vehicles (PERE-HD) and Its Extensions	87
    A.5.2.1  PERE-HD  Fleet-wide Average Emission Rate Estimator	87
    A. 5.2.2 Prediction of Elemental Carbon and Organic Mass based on PERE-HD	93
  A.5.2 Comparison of Predicted Emissions with Independent Measurements	102
  A.5.3 Variability in Predicted EC and OC Emission Rates	104
  A.5.4 Calculating EC/OC fraction by MOVES operating mode	106
A.6 Heavy-duty Gasoline Start Emissions Analysis Figures	108

-------
Executive Summary
This report describes the analysis conducted to generate emission rate and energy rate inputs
representing exhaust emissions and energy consumption for heavy-duty vehicles in MOVES2010.
Exhaust emission rate inputs were developed for total hydrocarbons (THC), carbon monoxide
(CO), nitrogen oxides (NOx), and particulate matter (PM). Energy consumption rates were
developed based on measurements of carbon dioxide (€62), CO and THC. We developed inputs
for heavy-duty vehicles powered by both diesel and gasoline fuels, although emissions from the
heavy-duty sector predominantly come from  diesel vehicles. As a result, the majority of the data
analyzed were from  diesel vehicles.
Estimation of energy consumption rates for heavy-duty vehicles is covered in this report, but
emissions  of greenhouse gases other than CO2are not covered. Estimation of the emissions of
methane and nitrous oxide (N2O) are described in a separate report1.
Evaporative  emissions from heavy-duty gasoline vehicles are not covered in this report. Estimation
of evaporative hydrocarbon emissions from heavy-duty gasoline vehicles is described in a separate
document2.  Note that the methods described were developed for light-duty vehicles, but are also
applied to  heavy-duty gasoline vehicles.  The model does not estimate evaporative emissions for
diesel-powered vehicles.
Large volumes of continuous ("second-by-second") data from various sources were analyzed ,
including onboard emissions measurement systems, chassis dynamometer tests, and engine
dynamometer tests.  Data were collected by a number of entities, including EPA, West Virginia
University, and private parties under contract to EPA. For running exhaust emissions, data were
analyzed by  model year, regulatory class, and operating mode.
As with the development of emission rates for light-duty vehicles, operating modes for heavy-duty
vehicles are  defined  in terms of power output (with the exception of the idle and braking modes).
For light-duty vehicles, the parameter used is known  as vehicle-specific power (VSP), which is
calculated by normalizing the continuous power output for each vehicle to its own weight. For
heavy-duty vehicles, we have continued to related emissions to power output, but in a different
way. Rather  than normalize the tractive power for each vehicle to its own weight, we scale the
power by a fixed multiple designed to fit the resulting means into the existing operating mode
framework. We refer to this  parameter as "scaled-tractive power" (STP).  Because heavy-duty
vehicles are  primarily regulated on an engine work basis (g/kW-hr), we conclude that the use of
STP preserves the emission to power relationship, whereas the use of VSP confounds it, resulting
in unintended consequences in estimation of emissions in relation to vehicle size or weight.
Additionally, to address the question of deterioration, we estimated the effects of tampering and
mal-maintenance on emission rates as a function of age. We adopted this approach due to the lack
of data adequate to directly estimate deterioration for heavy-duty vehicles. Based on surveys and
studies, we developed estimates of frequencies and emission impacts of specific emission control
component malfunctions, and then aggregated these to estimate overall emissions effects for each
pollutant.
Final emission rates  in grams per hour were developed for inclusion in the emissionRateByAge
table in the MOVES database. The rates describe the effects of operating mode as well as model
year group, which serves as  a broad surrogate for changes in technology and emissions standards,

-------
especially for NOx and PM.  The MOVES framework and the emissionRateByAge table are
discussed in the report documenting the rates for light-duty vehicles3.

-------
1      Heavy Duty Diesel Emissions
This section details our analysis of data to develop emission rates for heavy-duty diesel vehicles.
Three emission processes (running, extended idling, and starts) are discussed.  The 'running'
process occurs as the vehicle is operating on the road either under load or in idle mode. This
process is further delineated by 23  operating modes which will be discussed below.  The 'extended
idle' process occurs during an extended period of idling operation such as when a vehicle is parked
for the night and left idling. Extended idle is generally a different mechanism (usually a higher
RPM engine idle to power truck accessories for operator comfort) than the regular 'curb' idle that a
vehicle experiences while it is operating on the road.

1.1    Running Exh aust Emissions
MOVES running-exhaust emissions analysis requires accurate second-by-second measurements of
emission rates and parameters that can be can used to estimate the tractive power exerted by a
vehicle. Compared to volumes of data available for light-duty vehicles, the amount of data
available for heavy-duty vehicles is small. Light-duty emissions were analyzed with respect to
vehicle-specific power (VSP), which represents vehicles' tractive power normalized by their
(individual) weights. The model approach used in MOVES was first developed for light-duty
vehicles, relying on the VSP concept, and later adapted for use with heavy-duty vehicles. For
practical reasons, it was thus desirable to retain the same operating mode structure for heavy-duty
emission rates.
While VSP is an effective way to characterize emissions from light-duty vehicles, the  range of
running weights, coarseness of the VSP bin structure, and work-based (rather than distance-based)
emissions standards make VSP-based emissions analysis for heavy-duty diesel vehicles an
untenable approach. This report describes how we analyzed continuous "second-by-second'
heavy-duty emissions data to develop emission rates applied within the predefined set of operating
modes.  As mentioned, the emission rates were using scaled-tractive power (STP), rather than VSP.
The development of STP is described in greater detail below.
MOVES source bins are groupings of parameters which distinguish differences in emission rates
according to physical differences in the source type or vehicle classification.  The source bins  are
differentiated by fuel type (gasoline or diesel), regulatory class (light heavy duty to heavy-heavy
duty) and model year group.  Stratification of the data sample and generation of the  final MOVES
emission factors were done according to the combination of regulatory class (shown in Table 1)
and the model year group.  The regulatory groups were determined based on gross vehicle weight
rating (GVWR) classifications. The model year groupings are designed to represent major changes
in EPA emission standards.

-------
                        Table 1. Regulatory Classes for Heavy-Duty Vehicles.
Regulatory Class Description
Light-heavy duty < 14,000 Ib
Light-heavy duty 4-5
Medium-heavy duty
Heavy-heavy duty
Urban Bus
regClassName
LHD<=14k
LHD45
MHD
HHD
Urban Bus1
regClassID
41
42
46
47
48
Gross Vehicle Weight Rating
(GVWR) [Ib]
8,501 - 14,000
14,001 - 19,500
19,501-33,000
> 33,000
N/A
1 see CFRง 86.091(2).
Heavy-duty diesel truck emission rates in MOVES are also stratified by age group.  Within a
particular model year group, these age groups are used to account for the effects of deterioration
over time.  The age groups are used in the model are shown in Table 2.
                             Table 2. MOVES Age Group Definitions
ageGroupID
3
405
607
809
1014
1519
2099
Lower bound
(years)
0
4
6
8
10
15
20
Upper bound
(years)
3
5
7
9
14
19
~
1.1.1 Nitrogen Oxides (NOx)
For NOx rates, we stratified heavy-duty vehicles into the model year groups listed in Table 3.
These groups were defined based on changes in NOx emissions standards and the outcome of the
Heavy Duty Diesel Consent Decree4, which required additional control of NOx emissions during

-------
highway driving for model years 1999 and later. This measure is referred to as the "Not-to-
Exceed" (NTE) limit.
              Table 3 - Model year groups for NOx analysis based on emissions standards
Model year group
Pre-1988
1988-1989
1990
1991-1997
1998
1999-2002
2003-2006
2007-2009
2010+
FTP standard
(g/bhp-hr)
None
10.7
6.0
5.0
4.0
4.0
2.4
1.2
0.2
NTE limit (g/bhp-hr)
None
None
None
None
None
7.0 HHD; 5.0 other reg. classes
1.25 times the family emission level
7.7.7.7
Data Sources
For NOx emissions from HHD, MHD, and urban buses, we relied on two data sources:
ROVER.  This dataset includes measurements collected during on-road operation using the
ROVER system, a portable emissions measurement system (PEMS) developed by the EPA. The
measurements were conducted by the U.S. Army Aberdeen Test Center on behalf of U.S. EPA5:
This ongoing program started in October 2000. Due to time constraints and data quality issues, we
used only data collected from October 2003 through September 2007. The data was compiled and
reformatted for MOVES analysis by Sierra Research6.  The process of analysis and rate
development was performed by EPA. The data we used represents approximately 1,400 hours of
operation by 124 trucks and buses in model years 1999 through 2007.
       The vehicles were driven mainly over two routes:
          •  "Marathon" from Aberdeen, MD to Colorado and back along Interstate 70
          •  Loop around Aberdeen Proving Grounds in Maryland
Consent Decree testing.  These data were conducted by West Virginia University  using the
Mobile Emissions Measurement System (MEMS)7'8 This program was initiated as  a result of the
consent decree between the several heavy-duty engine manufacturers and the US government,
requiring the manufacturers to test in-use trucks over the road.  Data was collected from 2001
through 2006. The data we used represented approximately 1,100 hours of operation by 188 trucks
in model years 1994 through 2003.  Trucks were heavily loaded and tested over numerous routes
                                                                                      7

-------
involving urban, suburban, and rural driving.  Several trucks were re-acquired and tested a second
time after 2-3 years. Data were collected 5-Hz frequency, which we averaged around each second
to convert the data to a 1.0-Hz basis.
From each data set, we used only tests we determined to be valid. For ROVER, due to time
constraints, we eliminated all tests that indicated any reported problems, including GPS
malfunctions, PEMS malfunctions, etc, whether or not they affected the actual  emissions results.
As our own high-level check on the quality of PEMS and ECU output,  we further eliminated any
trip where the pearson correlation coefficient between CC>2 (from PEMS) and engine power (from
ECU) was less than 0.6.  These filters led to a smaller and more conservative subset of the overall
ROVER data, than had we applied more detailed and selective criteria,  (i.e. not all eliminated tests
produced erroneous results). For the WVU MEMS data, WVU itself reported on test validity under
the consent decree procedure. No additional detailed quality checks were performed by EPA.
Table 4 shows the total distribution of vehicles by model year group from both of the emissions test
programs above.
Table 4. Numbers of vehicles by model year group from the ROVER and WVU MEMS programs used for
emission rate analysis.
Regulatory class
HHD
MHD
BUS
1991-1997 MY
19
0
2
1998 MY
12
0
0
1999-2002 MY
78
30
25
2003-2006 MY
91
32
19
7.7.7.2 Calculate STP from 1-Hz data
With on-road testing, using vehicle speed and acceleration to estimate tractive power is not
accurate given the effect of road grade and wind speed. As a result, we needed to find an alternate
approach. Therefore, we decided to tractive power from engine data collected during operation.
We first identified the seconds in the data that the truck was either idling or braking based on
acceleration and speed criteria shown in Table 9. For all other operation, engine speed coeng and
torque Teng from the ECU were used to determine engine power Peng, as shown in Equation 1. Only
torque values greater than zero were used so as to only include operation where the engine was
performing work.
                                       eng     eng  eng
                                                                             Equation 1
We then determined the relationship between the power required at the wheels of the vehicle and
the power required by the engine. We first had to account for the losses due to accessory loads
during operation.  These power loads are not subtracted in the engine torque values that are output
from the engine control unit. Heavy-duty trucks use accessories during operation.  Some
accessories are engine-based and are required for operation. These include the engine coolant
pump, alternator, fuel pump, engine oil pump, and power steering. Other accessories are required

-------
for vehicle operation, such as cooling fans to keep the powertrain cool and air compressors to
improve braking.  The third type of accessories is discretionary, such as air conditioning, lights,
and other electrical items used in the cab.  The calculation of the accessory load requirements is
derived below.
We grouped the accessories into five categories:  cooling fan, air conditioning, engine accessories,
alternator (to run electrical accessories), and air compressor. We identified where the accessories
were predominately used on a vehicle speed versus vehicle load map to properly allocate the loads.
For example, the cooling fan will be on at low vehicle speed where the forced vehicle cooling is
low and at high vehicle loads where the engine requires additional cooling. The air compressor is
used mostly during braking operations; therefore it will have minimal load requirements at
highway, or high, vehicle speeds. Table 5 identifies the predominant accessory use within each of
the vehicle speed and load areas.
At this point, we also translated the vehicle speed and engine load map into engine power levels.
The power levels were aggregated into low (green), medium (yellow) and high (red) as identified in
Table 5. Low power means the lowest third, medium is the middle third, and high is the highest
third, of the  engine's rated power.  For example, for an engine rated at 450 hp, the low power
category would include operation between 0 and 150 hp, medium between 150 and 300 hp, and
high between 300 and 450 hp.
           Table 5 - Accessory use as a function of speed and load ranges, coded by power level.
\. Speed
Load ^^\
Low
Mid
High
Low
Cooling Fan
Air cond.
Engine Access.
Alternator
Air Compress
Cooling Fan
Air cond.
Engine Access.
Alternator
Air Compress
Cooling Fan
Air cond.
Engine Access.
Alternator
Air Compress
Mid
Air cond.
Engine Access.
Alternator
Air Compress
Cooling Fan
Air cond.
Engine Access.
Alternator
Air Compress
Cooling Fan
Air cond.
Engine Access.
Alternator
Air Compress
High
Air cond.
Engine Access.
Alternator
Air cond.
Engine Access.
Alternator
Cooling Fan
Air cond.
Engine Access.
Alternator
We next estimated the power required when the accessory was "on" and percentage of time this
occurred.  The majority of the load information and usage rates are based on information from "The
Technology Roadmap for the 21st Century Truck"9
The total accessory load is  equal to the power required to operate the accessory multiplied by the
percent of time the accessory is in operation.  The total accessory load for a STP bin is equal to the
sum of each accessory load. The calculations  are  included in Appendix A.I Calculation  of
Accessory Power Requirements.

-------
The total accessory loads Pioss,acc listed below in Table 6 are subtracted from the engine power
determined from Equation 1 to get net engine power available at the engine flywheel.  For LHD
vehicles, we assumed negligible accessory losses.
                      Table 6 - Estimates of accessory load in kW by power range.
Engine power
Low
Mid
High
HDT
8.1
8.8
10.5
MHD
6.6
7.0
7.8
Urban Bus
21.9
22.4
24.0
We then accounted for the driveline efficiency.  The driveline efficiency accounts for losses in the
wheel bearings, differential, driveshaft, and transmission. The efficiency values were determined

through literature searches. Driveline efficiency rj^nveime varies with engine speed, vehicle speed,

and vehicle power requirements.  Using sources available in the literature, we estimated an average
value for driveline efficiency. 10>11>12>13>  4, is, ie, 17, is  ja^je 7 summarizes our findings.
                    Table 7 - Driveline efficiencies found through literature research.
                             General truck:
Barth (2005)
Lucic (2001)
80-85%
75-95%
HDT:
Rakha
NREL (1998)
Goodyear Tire Comp.
Ramsay (2003)
21st Century Truck (2000)
SAE J2188 Revised OCT2003:
Single Drive/direct
Single Drive/indirect
Single Drive/double indirect
Tandem Drive/direct
Tandem Drive/indiriect
Tandem Drive/double indirect
75-95%


91%
86%
91%
94%

94%
92%
91%
93%
91%
89%
Bus:
Pritchard (2004): Transmission Eff.
Hedrick (2004)
MIRA
96%
96%
80%
Based on this research, we used a driveline efficiency of 90% for all HD regulatory classes.

Equation 2 shows the translation from engine power Peng to axle power Paxie-
                                p    =77
                                 axle    I driv
                                           eline
o   _ p
 eng     loss,ace .
Equation 2
                                                                                            10

-------
Finally, we scaled the axle power by a multiplicative factor fscaie to fit light-duty operating-mode
ranges. The MHD, HHD, and Bus classes were scaled by 17.1, which is approximately the average
running weight for all heavy-duty vehicles, and the LHD trucks were scaled by 2.06, which is
equivalent to the fleet-average mass of light commercial trucks in MOVES. Table 8 shows the
values selected for the scaling factor.
                               Table 8 - Power scaling factor/sca/f.
Regulatory Class
MHD, HHD, Bus
LHD
Power scaling factor
17.1
2.06
Equation 3 shows the conversion of axle power to scaled tractive power using the method
explained above.
STP = -^-
                                                                     Equation 3
We then constructed operating mode bins defined by STP and vehicle speed according to the
methodology outlined earlier in MOVES development19 and described in Table 9. The
implementation of STP in MOVES for heavy-duty emission rates is the same as that of VSP for
light-duty emission rates. We will refer to the units of STP as scaled kW or skW.
        Table 9 - Definition of the Operating Mode Attribute for Heavy-Duty Vehicles (opModelD)
Operating
Mode
0
1
11
12
13
14
15
16
21
22
23
Operating Mode
Description
Deceleration/Braking
Idle
Coast
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Coast
Cruise/ Acceleration
Cruise/ Acceleration
Scaled Tractive Power
(STP,, skW)


STP,< 0
0 < STP/< 3
3 
-------
24
25
27
28
29
30
33
35
37
38
39
40
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
Cruise/ Acceleration
6 
-------
1.1.1.3.2      Statistics

Estimates of uncertainty were calculated for all the emission rates. Because the data represent
subsets of points "clustered" by vehicle, we calculated and combined two variance components,
representing "within-vehicle" and "between-vehicle" variances. First, we calculated the overall
within-vehicle variance swith.
                                       2   _ ;=1                                 Equation 5
                                     Swith ~


where

       sveh= the variance within each vehicle, and

       ntot = the total number of data points for all the vehicles.

Then we calculated the between-vehicle variance sletw (by source bin, age group, and operating

mode) using the mean emission rates for individual vehicles (7  ) as shown in Equation 6
                                            "f(r  -rj
                                            f-(\'p,i   pi
                                      2  _ ]=\   '                                   Equation 6
                                      betw             -,
Then, we estimated the total variance by combining the within-vehicle and between-vehicle
variances to get the standard error s-   (Equation 7) and dividing the standard error by the mean

emission rate to get the coefficient-of-variation of the mean cv _  (Equation 8).
                                                          'r
                                      S    =  hLs-+ ^                       Equation?
                                       rpol    \ "veh    "tot
                                                   Fpol
                                         C   l =^—                          Equations
                                          v,po

                                                   pot
                                                                                           13

-------
1.1.1.3.3     Hole Filling and forecasting
Since the data only covered model years 1994 through 2006, we needed to develop a method to
forecast emissions for future model years and back-cast emissions for past model years. For future
model years (2007-and-later), we decreased the emission rates for all operating mode bins by a
ratio proportional to the decrease in the applicable emissions standards.
While the NOx standard going into effect for MY 2007 is 0.2 g/bhp-hr, it was assigned to be
phased in over a three year period ending in 2010. Rather than phaseing in the after-treatment
technology needed to meet the new standard, most manufacturers decided to meet a 1.2 g/bhp-hr
standard for MY2007-2009, which did not require aftertreatment (down from 2.4 g/bhp-hr in
2006).  Therefore, to model this strategy, we estimated rates for MY2007-2009 by decreasing the
rates for MY2003-2006 emission rates by 50%.  Starting in MY2010, the NOx standard for all
heavy-duty trucks is 0.2 g/bhp-hr. We projected that almost all of these trucks will be using SCR
aftertreatment technology, which we assume to have a 90% NOx reduction efficiency from levels
for MY2006 levels, and estimated rates accordingly.
For model year 1990, we increased the 1991-1997 emission rates by 20% to account for the
reduction in NOx standard from 6.0 to 5.0 g/bhp-hr from 1990 to 1991.  For 1989 and earlier model
years, we increased the 1991-1997 model year group emission rates by 40%, which is proportional
to the increase of the certification levels from the 1991 model year to the 1989 model year. We
assumed that emission levels did not change by model year for 1989 and earlier.
For MHD and HHD trucks, the maximum operating mode represents a tractive power greater than
513 kW (STP= 30 skW x  17.1).  This value exceeds the capacity of most HHD vehicles, and MHD
vehicles and buses  exert even lower levels. As a result, data are very limited in these modes.
To estimate rates in the modes beyond the ranges of available data, we linearly extrapolated the
rates from the highest operating mode in each speed range where significant data were  collected for
each model year group.  In most cases, this modes was mode 16 for the lowest speed range, 27 or
28 for the middle speed range, and 37 or 38 for the highest speed range. For each of these
operating modes, work-specific emissions factors (g/kW-hr) were calculated using the midpoint
STP. Then, these emissions factors were multiplied by the midpoint STP of the higher operating
modes (e.g. modes 39 and 40 for speed>50mph) to inpute emission rates for the modes lacking
data. For the highest bins in each speed range, a "midpoint"  STP of 33 skW (564.3 kW) was used.
For certain  model years, such as 1998, data existed for HHD trucks, but not MHD or buses.  In
these cases, the ratio of standards between the missing regulatory class and HHD regulatory class
from the 1999-2002 model year group was used to determined missing class's rates by multiplying
that ratio by the existing HHD emission rates for the corresponding model year group.
1.1.1.3.3.1    LNT-equipped pickup trucks
To meet NOx emissions standards for the 2010 model year, the use of aftertreatment will probably
be needed.  For example, Cummins decided to use aftertreatment starting in 2007 in engines
designed to meet the 2010 standard and used in vehicles such as the Dodge Ram. The technology
adopted for this purpose was the "Lean NOx Trap" (LNT). This technology allows for the storage
of NOx during fuel-lean operation and conversion of stored NOx into N2 and H2O during brief
periods of fuel-rich operation.  In addition, to meet parti culate standards in MY 2007 and later,

                                                                                       14

-------
heavy-duty vehicles are equipped with diesel particulate filters (DPF).  At regular intervals, the
DPF must be regenerated to remove and combust accumulated PM to relieve backpressure and
ensure proper engine operation. This step requires high exhaust temperatures.  However, these
conditions adversely affect the LNT's NOx storage ability, resulting in elevated NOx emissions.
In 2007, EPA acquired a truck equipped with LNT and DPF and performed local on-road
measurements, using portable instrumentation.  We used the PEMS and ECU output to assign
operating modes and calculate emission rates by the same methods used to develop the heavy-
heavy-duty truck NOx rates.  While analyzing these data, we distinguished regimes of PM
regeneration from normal operation based on exhaust temperature, with temperatures exceeding
300ฐC assumed to indication  PM regeneration.  We performed the emission rate by operating mode
analysis separately for each regime, and weighted the two regimes together based on an assumed
PM regeneration frequency of 10% of VMT. This value is an assumption based on the limited data
available.  We will look for opportunities to update this assumption based on any additional
information that becomes available.
Because we assume that LNT-equipped trucks account for about 25% of the LHDDT market, we
again weighted the rates for the two LFtD regulatory  classes for  model years 2007 and later.  For
MY 2007-09, we assume that the remaining 75% of LFtD diesel trucks will not have aftertreatment
and will exhibit the 2007-2009 model year emission rates described earlier in this section. Starting
in MY2010, we assume that the remaining 75% of LHD diesel trucks will be equipped with SCR,
and will exhibit 90% NOx reductions from 2006 levels, also described in the hole filling section.
Table 10 summarizes this discussion and previous subsections regarding the methods used to
estimate emission rates for each regulatory-class/model-year-group combination.
Table 10. Summary of methods for heavy-duty diesel NOx emission rate development for each regulatory class
and model year group
Model year group
Pre-1988
1988-1989
1990
1991-1997
1998
1999-2002
2003-2006
2007-2009
HHD
Proportioned to
certification levels
Proportioned to
certification levels
Proportioned to
certification levels
Data analysis
Data analysis
Data analysis
Data analysis
Proportioned to
standards
MHD
Proportioned to
certification levels
Proportioned to
certification levels
Proportioned to
certification levels
Proportioned to HHD
Proportioned to HHD
Data analysis
Data analysis
Proportioned to
standards
Bus
Proportioned to
certification levels
Proportioned to
certification levels
Proportioned to
certification levels
Data analysis
Proportioned to HHD
Data analysis
Data analysis
Proportioned to
standards
LHD
Proportioned to
certification levels
Proportioned to
certification levels
Proportioned to
certification levels
Proportioned to HHD
Proportioned to HHD
MHD engine data with
LHD scale factor
MHD engine data with
LHD scale factor
Data (LNT), and
proportioned to
standards (non-LNT)
                                                                                       15

-------
2010 +
Proportioned to
standards
Proportioned to
standards
Proportioned to
standards
Proportioned to
standards
An important point to note is that we did not make a stratification based on age for vehicles not
equipped with NOx aftertreament technology (largely 2009 model year and earlier).  This is
because of a few reasons:
    •  The WVU MEMS data did not show an increase in NOx emissions with  odometer (and
consequently, age) during or following the regulatory useful life20. Since the trucks in this program
were collected from in-use  fleets, we do not believe that these trucks were necessarily biased
toward cleaner engines.
    •  Manufacturers often certify zero or low deterioration factors.
We estimated tampering and mal-maintenance effects on NOx emissions to be small compared to
other pollutants - around a 10% increase in NOx over the useful life of the engine. Our tampering
and mal-maintenance estimation methods are discussed below and detailed in Appendix A.2
Tampering and Mal-maintenance.

1.1.1.3.4 Tampering and Mal-maintenance
Table 11 shows the estimated aggregate NOx emissions increases due to T&M.  It also shows the
values that we actually used for MOVES emission rates.  As previously mentioned, we assumed
that in engines not equipped with aftertreatment, NOx does not increase due to T&M or
deterioration.
Table 11. Fleet-average NOx emissions increases from zero-mile levels over the useful life due tampering and
mal-maintenance
Model years
1994-1997
1999-2002
2003-2006
2007-2009
2010-2012 SCR
20 10-20 12 LNT
2013+
NOx increase from T&M
analysis [%]
10
14
9
11
77
64
58
NOx increase in MOVES
[%]
0
0
0
0
77
64
58
As described in Appendix A.2 Tampering and Mal-maintenance., these emissions increases are
combined with information in Table 37 to estimate the emissions increase for each age group prior
                                                                                        16

-------
to the end of the useful life for each regulatory class. With the introduction of aftertreatment
systems to meet regulatory requirements for MY 2010 and later, EPA expects tampering and mal-
maintenance to substantially increase emissions over time compared to the zero-mile level.
Though 77% may appear to be a large increase in fleet-average emissions over time, it should be
noted that the 2010 model year standard (0.2 g/bhp-hr) is about 83% lower than the 2009 model
year effective standard (1.2 g/bhp-hr). This still yields a substantial reduction of about 71% from
2009 zero-mile levels to 2010 fully deteriorated levels. As more data becomes available for future
model years, we hope to update these tampering and mal-maintenance and overall aging effects.

1.1.1.3.5 Defeat Device andLow-NOx Rebuilds
The default emission rates in MOVES for model years 1991 through 1998 are intended to include
the effects of defeat devices as well as the benefits of heavy-duty low-NOx rebuilds (commonly
called reflash) that occurred as the result of the heavy-duty diesel consent decree. Reflashes reduce
NOx emissions on these engines by reconfiguring certain engine calibrations, such as fuel injection
timing.  The MOVES database also includes a set of alternate emission rates for model years 1991
through 1998 assuming a hypothetical fully reflashed fleet. Users with questions about the use of
these alternate emission factors should contact EPA at mobile@epa.gov.
Since defeat devices were in effect mostly during highway or steady cruising operation, we assume
that NOX emissions were elevated for only the top two speed ranges in the running exhaust
operating modes (>25mph). To modify the relevant emission rates to represent reflash programs,
we first calculated the ratios emission rates in modes 27 and 37 to that for opMode 16, for model
year 1999 (the first model year with not-to-exceed emission limits).  We then multiplied the MY
1999 ratios by the emission rates in mode 16 for model years 1991 through 1998,  to get estimated
"reflashed" emission rates for operating modes 27 and 37. This step is described in Equation 9 and
Equation 11.  To  estimated "reflashed" rates in the remaining operating modes, we multiplied
"reflashed rates by ratios of the remaining operating modes to mode 27 for MY1991-98, as shown
in Equation 10 and Equation 12.
                                 ^re/M,91-98,27   ^91-98,16
 Operating modes
   (OM) 21-30
                                                      T
                                                      '1999,27
                                                     . ^1999,16  .
                                                                                  Equation 9
                             V            = T
                              reflash,9\-98,OMx    reflash, 91-98,27
                                                         91-98,27
                                                                                 Equation 10
 Operating modes
   (OM)31-40
^re/M,91-98,37   ^91-98,16
M999,37


^1999,16  .
Equation 11


Equation 12
                                                                                         17

-------
                                             = r
                            ' reflash,Affl99l-l99S,OMx   ' reflash, 91-98,371   _
                                                            ^91-98,37
The default emission rates were also slightly adjusted for age for the consent decree model years.
An EPA assessment shows that about 20% of all vehicles eligible for reflash had been reflashed by
the end of 2008.21 We assumed that vehicles were receiving the reflashes after the heavy-duty
diesel consent decree (post 1999/2000 calendar year) steadily, such that in 2008, about 20% had
been reflashed. We approximated a linear increase in reflash rate from age zero.

1.1.1.4        Sample results
The charts  in this sub-section show examples of the emission rates that resulted from the analysis.
Not all rates are shown; the intention is to illustrate the most common trends and hole-filling
results.  For brevity, the light-heavy duty regulatory classes are not shown, since the light-heavy
duty rates were based on medium-heavy data and follow similar trends.
In Figure 1, we see that NOx emission rates increase with STP for FfflD trucks. Figure 2 adds the
MHD and bus regulatory classes, with the error bars removed for clarity.  As expected, the
emissions increase with power, with the lowest emissions occurring in the idling/coasting/braking
bins.
Figure 1. Trends in NOx Emissions by operating mode from HHD trucks for model year 2002.

            4000


            3500
          s
          a, 2500  -
          +-ป
            2000  -


            1500  -
            1000
             500  -
                  0  1 11 12 13  14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
                                          Operating Mode
                                                                                          18

-------
The highest operating modes in each speed range will rarely be attained due to the power
limitations of heavy-duty vehicles, but are included in the figures (and in MOVES) for
completeness.  Nearly all of the activity occurs in modes 0, 1, 11-16, 21-28, and 33-38, with
activity for buses and MHD vehicles usually occurring over an even smaller range. In some model
year groups, the MHD and HHD classes use the same rates, based on lack of significant differences
between those two classes' emission rates.
 Figure 2. Trends in NOx emissions by operating mode from MHD, HHD, and bus regulatory classes for model
                                         year 2002.


L.
1
0)
E
|
^
ra
01
ง



6000
5000
4000

3000
2000

1000
n -

A
A A
• MHD
A Bus
*HHD •
^ * • t '
I
• s '
, I * * * S * " *
• JLAj A • !
0  1  11 12 13 14
                      Operating Mode
                                                                       38 39 40
The effects of model year, representing a rough surrogate for technology or standards, can be seen
in Figure 3, which shows decreasing NOx rates by model year group for a sample operating mode
(# 24) for HHD trucks. Other regulatory classes show similar trends.  The rates in this chart were
derived with a combination of data analysis (model years 1991 through 2006) and hole filling. The
trends in the data are expected, since the model year groups were formed on the basis of NOx
standards.
                                                                                        19

-------
Figure 3. Trends in NOx by model year for HHD trucks in operating mode 24. Increasingly stringent emissions
standards have caused NOx emissions to decrease significantly.
               2500
               2000
             01
             E
             x
             O
               1500
               1000
             ro
             01
                500 -
                  ^
                                           Model year group




Age effects were only implemented for aftertreatment-equipped trucks (mostly model year 2010
and later) based on an analysis of tampering and mal-maintenance effects. Due to faster mileage
accumulation, the heavy-heavy duty trucks reach their maximum emission at the youngest ages, as
shown in Figure 4. Coefficients of variation from previous model year groups were used to
estimate uncertainties for MY 2010.
                                                                                         20

-------
             Figure 4. Modeled NOx trends by age for model year 2010 for operating mode 24.

               140 -|

               120 -
            •^ 100
            1
            Si  80
            as
                60 -


                40 -


                20 -
ro
01
                 0
• HMD

*MHD

A Bus
                       0-3     4-5  6-7 8-9     10-14       15-19
                                           Age group [years]
                                                            20+
Figure 5 shows the mean emission rates for LHD trucks for model years 2007-2009.  As described
previously, this group of vehicles includes vehicles with LNTs (with NOX increases during PM
regeneration) and vehicles without any aftertreatment.  The estimated uncertainties are greater than
for the other heavy-duty regulatory classes, since there were fewer vehicles in our test data.
        Figure 5: Mean NOx rates by operating mode for model years 2007-2009 LHD trucks age 0-3

               300 -


               250 -
|B 200
01
E
x 150
O
I 100
50 -

0 -


i ,i
* ^ ^l ^
*
4
                     0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
                                          Operating mode
                                                                                           21

-------
1.1.2  Particulate Matter (PM)
In this section, particulate matter emissions refers to particles emitted from heavy-duty engines
which have a mean diameter less than 2.5 microns, known as PM2.5. Such particles consist of three
subtypes, including: (1) elemental carbon (EC), usually composed of black colored soot emitted
from combustion, (2) organic carbon (OC), consisting of particles of organic matter formed during
the combustion process or immediately after in the tailpipe.  It does not include particles formed in
secondary reactions in the atmosphere, and (3) sulfate particulate, which formed by agglomeration
of sulfur-containing compounds formed during combustion.  These subtypes are used to form the
inputs to MOVES.
As described above for NOx, the heavy-duty diesel PM emission rates in MOVES are a function
of: (1) source bin, (2) operating mode, and (3) age group.
We classified the data into the following model year groups for purposes of emission rate
development. These groups are generally based on the introduction of emissions standards for
heavy-duty diesel engines. They also serve as a surrogate for continually  advancing emission
control technology on heavy-duty engines.  Table 12 shows the model year group range and the
applicable brake-specific emissions standards.
            Table 12. Model year groups used for analysis based on the PM emissions standard
Model Year Group Range
1960-1987
1988-1990
1991-1993
1994-1997
1998-2006
2007+
PM Standard [g/bhp-hr]
No transient cycle standard
0.60
0.25
0.10
0.10
0.01
1.1.2.1 Data Sources
All of the data used to develop the MOVES PM2.5 emission rates was generated in the CRC E-
55/59 research program22.  The following description by Dr. Ying Hsu and Maureen Mullen of E.
H. Pechan, m \}\Q" Compilation of Diesel Emissions Speciation Data-Final Report" provides a
good summary of the program.  It is reproduced in the following paragraphs immediately below:
       The objective of the CRC E55/59 test program was to improve the understanding of the
       California heavy-duty vehicle emissions inventory by obtaining emissions from a
       representative vehicle fleet, and to include unregulated emissions measured for a subset of
                                                                                       22

-------
the tested fleet.  The sponsors of this project include CARB, EPA, Engine Manufacturers
Association, DOE/NREL, and SCAQMD.  The project consisted of four segments,
designated as Phases 1, 1.5, 2, and 3.  Seventy-five vehicles were recruited in total for the
program, and recruitment covered the model year range of 1974 through 2004. The number
and types of vehicles tested in each phase are as follows:

• Phase 1:     25 heavy heavy-duty (HHD) diesel trucks
• Phase 1.5:   13 HHD diesel trucks
• Phase 2:     10 HHD diesel trucks, 7 medium heavy-duty (MHD) diesel trucks,
               2 MHD gasoline trucks
• Phase 3:       9 MHD diesel, 8 HHD diesel, and 2 MHD gasoline
The vehicles tested in this study were procured in the Los Angeles area, based on model
years specified by the sponsors and by engine types determined from a survey. WVU
measured regulated emissions data from these vehicles and gathered emissions samples.
Emission samples from a subset of the vehicles were analyzed by Desert Research Institute
for chemical species detail. The California Trucking Association assisted in the selection of
vehicles to be included in this study. Speciation data were obtained from a total of nine
different vehicles.  Emissions were measured using WVU's Transportable Heavy-Duty
Vehicle Emissions Testing Laboratory. The laboratory employed a chassis dynamometer,
with flywheels and eddy-current power absorbers, a full-scale dilution tunnel, heated probes
and sample lines and research grade gas analyzers. PM was measured gravimetrically.
Additional sampling ports on the dilution tunnel supplied dilute exhaust for capturing
unregulated species and PM size fractions. Background data for gaseous emissions were
gathered for each vehicle test and separate tests were performed to capture background
samples of PM and unregulated species.  In addition, a sample of the vehicles received
Tapered Element Oscillating Microbalance (TEOM) measurement of real time particulate
emissions.

The HHDDTs were tested under unladen, 56,000  Ib, and 30,000 Ib truck load weights. The
driving cycles used for the HHDDT testing included:
• AC50/80;
•UDDS;
• Five  modes of an HHDDT test schedule proposed by CARB: Idle, Creep, Transient,
  Cruise, and HHDDT_S (a high speed cruise mode of shortened duration)
• The U.S. EPA transient test
The proposed CARB HHDDT test cycle is based  on California truck activity data, and was
developed to improve the accuracy of emissions inventories. It should be noted that the
transient portion of this proposed CARB test schedule is similar but not the same as the
EPA certification transient test.
                                                                                23

-------
The tables below provide a greater detail of the data used in the analysis.  Vehicles counts are
provided by number of vehicles, number of tests, model year group and regulatory class (46 =
MHD, 47=HHD) in Table 13.
              Table 13. Vehicle and Test Counts by Regulatory Class and Model Year Group
Regulatory Class
MHD
HHD
Model Year Group
1960-1987
1988-1990
1991 - 1993
1994 - 1997
1998-2006
2007 +
1960-1987
1988-1990
1991 - 1993
1994 - 1997
1998-2006
2007 +
Number of tests
82
39
22
39
43
0
31
7
14
22
171
0
Number of vehicles
7
5
2
4
5
0
6
2
2
5
18
0
Counts of tests are provided by test cycle in Table 14.
                            Table 14. Vehicle Test Counts by Test Cycle
Test Cycle
CARB-T
CARB-R
CARB-I
UDDS W
AC5080
CARB-C
CARBCL
MHDTCS
MHDTLO
MHDTHI
MHDTCR
Number of tests
71
66
42
65
42
24
34
63
23
24
29
1.1.2.2 Analysis

1.1.2.2.1 Calculate STP in 1-hz data
Within source bins, data was further sub-classified on the basis of operating mode. For motor
vehicles, 23 operating modes are defined in terms of scaled tractive power (STP), vehicle speed and
vehicle acceleration.  These modes are defined above in Table 9.
                                                                                         24

-------
The first step in assigning operating mode is to calculate scaled tractive power (STP) for each
emissions measurement. At a given time t, the instantaneous STP, represents the vehicle's tractive
power scaled by a constant factor. STP is calculated as a third-order polynomial in speed, with
additional terms describing  acceleration and road-grade effects. The coefficients for this
expression, often called road load coefficients, factor in the tire rolling resistance, aerodynamic
drag, and friction losses in the drivetrain. We calculated STP using the equation below:
                                                                             Equation 13
                                              J s
                                                scale
where

    A = the rolling resistance coefficient [kW-sec/m],
                                                9  9
    B = the rotational resistance coefficient [kW-sec /m ],

    C = the aerodynamic drag coefficient [kW-sec3/m3],

    m = mass of individual test vehicle [metric ton],

    fscaie = fixed mass factor (see Table 8),
    vt = instantaneous vehicle velocity at time t [m/s],

    at = instantaneous vehicle acceleration [m/s2]
The values of coefficients A, B, and C are the road load coefficients pertaining to the heavy-duty
vehicles23 as determined through previous analyses for EPA's Physical Emission Rate Estimator
(PERE).  This method of calculating STP calculates tractive power using the same equation used to
calculate vehicle-specific power (VSP) in the development of emission rates for light-duty vehicles
except that the scaling factor is used in the denominator, instead of the actual test weights of
individual vehicles24.
Note that this approach differs from that described above the NOx emission rate analysis since the
paniculate data was collected on a chassis dynamometer from vehicles lacking electronic control
units (ECU).  Grade effects are not explicitly included in either case because grade does not come
into play in chassis dynamometer tests, and it is already accounted for if STP is calculated through
engine speed  and torque from the engine control unit.
We have not formally compared the results of the two methods of calculating STP. However, on
average, we did find the operating-mode distributions to be similar between the two calculation
methods  for a given vehicle type. For example, we found that the maximum STP in each speed
range was approximately the same.
                                                                                        25

-------
1.1.2.2.2     Compute Normalized TEOM Readings
The TEOM readings were obtained for a subset of tests in the E-55/59 test program. Only 29
vehicles had a full complement of 1-hz TEOM measurements.  However, the continuous particulate
values were modeled for the remaining vehicles by Nigel Clark of West Virginia University, and
results provided to EPA. In the end, a total of 56 vehicles (out  of a total of 75) and 470 tests were
used in the analysis out of a possible 75 vehicles.  Vehicles and tests were excluded if the total
TEOM PM2.5 reading was negative or zero, or if corresponding full-cycle filter masses were not
available.  Table 15 provides vehicle and test counts by vehicle class and model year. The HDDV6
and HDDV7 groups were combined in the table because there were only seven HDDV6 vehicles in
the study.
Table 15 Vehicle and Test Counts by Heavy-Duty Class and Model Year
Model Year
1969
1974
1975
1978
1982
1983
1985
1986
1989
1990
1992
1993
1994
1995
1998
1999
2000
2001
2004
2005
HDDV6/7
No. Vehicles

1


1
1
1
1
2
1
1
1
1
2
2

2
1


No. Tests

10


5
10
28
3
11
12
11
11
9
24
20

18
5


HDDV8
No. Vehicles
1

2
1

1
1
1
1
1
1
1
3
3
3
3
5
2
4
1
No. Tests
6

10
5

6
10
4
4
3
11
3
15
13
28
43
44
21
29
6
Since the development of MOVES emission rates is cycle independent, all available cycles / tests
which met the above requirements were utilized. As a result, 488,881 seconds of TEOM data were
used.  The process required that each individual second by second TEOM rate be normalized to its
corresponding full-cycle filter mass, available for each combination of vehicle and test. This step
was necessary because individual TEOM measurements are highly uncertain and vary widely in
terms of magnitude (extreme positive and negative absolute readings can occur). The equation
below shows the normalization process for a particular one second TEOM measurement.
                                                                                       26

-------
                                           PM
                          PM         = - fllter-J  PM
                          A ^"normalized,;,;   ;,           TEOM' J ' i                 Equation 14
 Where
       /' = an individual 1-Hz measurement (g/sec),
      7 = an individual test on an individual vehicle,
            OMj,! = an individual TEOM measurement on vehicley at second /',
               = the Total PM2.5 filter mass on y,
       PMnormaiized,ij = an estimated continuous emission result (PM2.5) emission result on vehicley
       at second /'.

1.1.2.2.3      Compute Average Normalized TEOM measures by MO VES Bin
After normalization, the data were classified by regulatory class, model-year group and the 23
operating modes. Mean average results, sample sizes and standard deviation statistics for PM2.5
emission values were computed in terms of g/hour for each mode.  In cases where the vehicle and
TEOM samples were sufficient for a given mode, these mean values were adopted as the MOVES
emission rates for total PM2.5.  In cases of insufficient data for particular modes, a regression
technique was utilized to impute missing values.

1.1.2.2.4     Hole filling and Forecasting
1 . 1 .2.2.4. 1    Missing operating modes
Detailed in Appendix A. 4 Developing PM emission rates for missing operating modes , a log-linear
regression was performed on the existing PM data against STP to fill in emission rates  for missing
operating mode bins. Similar to the NOx rates, emission rates were extrapolated for the highest
STP operating modes.
1 . 1 .2.2.4.2    Other Regulatory Classes
The TEOM data was only available in quantity for MHD and HHD classes. There were no data
available for the LHD or bus classes. Thus,  rates  for these vehicle classes were computed using
simple multiplicative factors based either on engine work ratios or PM emission standards (i.e.,
buses versus heavy trucks).  The LHD classes' emission rates were set as a ratio of the  MHD
emission rates, and bus (class 48) emission rates were proportioned to HHD rates.
Because the certification standards in terms of brake horsepower-hour (bhp-hr)  are the same for all
of the heavy-duty engines, the emission rates of LHD2b3 are assumed to be equal to  0.46 *  MHD
emission rates.  The emission rate of LHD45 is assumed to be 0.60 *  MHD emission rate.
       LHD2b3 emission rate       =     0.46 * MHD emission rate
       LHD45 emission rate        =     0.60 * MHD emission rate
                                                                                        27

-------
	                                                                                  r\f
The values of 0.46 and 0.60 are the ratios of the MOBILE6.2 heavy-duty conversion factors  (bhp-
hr/mile) for the lighter trucks versus the MHD trucks. These are ratios of the relative amount of
work performed by a lighter truck versus a heavier truck for a given distance.

Urban Bus (Class48) emission rates are assumed to be either the same as the HHD emission rates,
or for some selected model year groups, to be a ratio of the EPA certification standards:

1991 - 1993 model years          Bus Emissions = (0.40) * HHD emissions

1994 - 1995 model years          Bus Emissions = 0.70 * HHD emissions

1996 - 2006 model years          Bus Emissions = 0.50 * HHD emissions
1.1.2.2.4.3 Model year 2007 and later trucks (with diesel particulate filters)
EPA heavy-duty diesel emission regulations were made considerably more stringent for total
PM2.5 emissions starting in model year 2007. Ignoring phase-ins and banking and trading issues,
the basic emission standard fell from 0.1 g/bhp-hr to 0.01 g/bhp-hr.  This increase by a factor often
in the level of regulatory stringency required the use of particulate trap systems on heavy-duty
diesels.  As a result, we expect the emission performance of diesel vehicles has changed
dramatically.
Unfortunately, no continuous TEOM data were available for analysis on the 2007 and later model-
year vehicles.  However, heavy and medium heavy-duty diesel PM2.5 data are available from the
EPA engine certification program on model years 2003 through 2007. These data provide a
snapshot of new engine emission performance before and after the introduction of particulate trap
technology in 2007. The existence of these data makes it possible to determine the relative
improvement in PM emissions from model years 2003 through 2006 to model year 2007. This
same relative improvement can then be applied to the existing, TEOM based, 1998-2006 model
year PM emission factors to estimate in-use factors for 2007 and later vehicles.
An analysis of the available certification data is shown in Table 16 below.  It suggests that the
actual ratio of improvement due to the particulate trap is reduction of a factor of 27.7. This factor
is considerably higher than the relative change in the certification standards, i.e., a factor of 10.
The reason for the change is that the new trap equipped vehicles certify at emission levels which
are much lower than the standard and thus create a much larger 'margin of safety' than previous
technologies could achieve.
As an additional check on the effectiveness of the trap technology EPA conducted some limited in-
house testing of a Dodge Ram truck, and  carefully reviewed the test results from the CRC
Advanced Collaborative Emission Study  (ACES) phase-one program, designed to  characterize
emissions from diesel engines meeting 2007 standards.  The limited results from these studies
demonstrate that the effectiveness of working particulate traps is very high.  The interested reader
can review the ACES report.26
                                                                                        28

-------
Table 16 shows average certification results for model years 2003-2007. Average ratio from MYs 2003-2006 to
MY 2007 is 27.7.
Certification Model Year
2003
2004
2005
2006
2007
Mean
(g/bhp-hr)
0.08369
0.08783
0.08543
0.08530
0.00308
St. Dev.
0.01385
0.01301
0.01440
0.01374
0.00228
n
91
59
60
60
21
1.1.2.2.5 Tampering and Mai-maintenance
The MOVES model contains assumptions for the frequency and emissions effect of tampering and
mal-maintenance on heavy-duty diesel trucks and buses. The assumption of tampering and mal-
maintenance (T&M) of heavy-duty diesel vehicles is a departure from the MOBILE6.2 model
which assumed such vehicles operated from build to final scrappage at a design emission level
which was lower than the prevailing EPA emission standards.  Both long term anecdotal data
sources and more comprehensive studies now suggest that the assumption of no natural
deterioration and/or no deliberate tampering of emission control components in the heavy-duty
diesel fleet was likely an unrealistic assumption, particularly with the transition to emission
aftertreatment devices with the 2007/2010 standards
The primary data set was collected during a limited calendar year period, yet MOVES requires data
from a complete range of model year/age combinations. As a result, the T&M factors shown below
in Table 17 were used to forecast or back-cast the basic PM emission rates to predict model year
group and age group combinations not covered by the primary data set.  For example, for the 1981
through 1983 model year group, the primary dataset contained data which was in either the 15 to 19
or the 20+ age groups.  However, for completeness, MOVES must have emission rates for these
model years for ageGroups 0-3, 4-5, 6-7, etc. As a result, unless we assume that the higher
emission rates which are were measured on the older model year vehicles have always prevailed -
even when they were young, a modeling approach such as T&M must be employed. Likewise,
more recent model years could only be tested at younger ages.  The T&M methodology used in the
MOVES analysis allows for the filling of age - model year group combinations for which no data
is available.
One criticism of the T&M approach is that it may double count the effect of T&M on the fleet
because the primary emission measurements,  and base emission rates, were made on in-use
vehicles that may have had some maintenance issues during the testing period. This issue would be
most acute for the 2007 and later model year vehicles where all of the deterioration is subject to
projection. However, for this model year group of vehicles, the base emission rates start at low
levels, and represent vehicles that are virtually free from T&M.
We followed the same tampering and mal-maintenance methodology and analysis for PM as we did
for NOx, as described in Appendix A.2 Tampering and Mal-maintenance.  The  overall MOVES
tampering and mal-maintenance  effects on PM emissions over the fleet's useful life are shown  in
                                                                                      29

-------
Table 17.  The value of 89 percent for 2010+ model years reflects the projected effect of heavy-
duty on-board diagnostic deterrence/early repair of Tampering and Mai-maintenance effects. It is
an eleven percent improvement from model years which do not have OBD (i.e., 2007-2009).
Table 17. Estimated increases in HC and CO emissions attributed to Tampering and mal-maintenance over the
useful life of Heavy-Duty Vehicles.
Model Year Group
Pre-1998
1998-2002
2003 - 2006
2007 - 2009
2010 +
Percent increase in PM due
toT&M
85
74
48
100
89
1.1.2.2.6     Computation of Elemental Carbon and Organic Carbon Emission Factors
The MOVES model reports Total PM2.5 emissions according to three species. These include
Elemental Carbon (EC), Organic Carbon (OC) and sulfate. In the results of a MOVES run, Total
PM2.5 is the sum of these three constituents. During rate development, the process is reversed, and
both EC and OC are computed directly from the total PM2.5 emissions using multiplicative factors.
Elemental Carbon is generally the black 'soot' that is often visible in engine exhaust. Organic
Carbon generally includes organic particles of large molecular compounds and metals.  Sulfate is
computed using a fuel sulfur balance (see the report "Development of Gasoline and Diesel Vehicle
Sulfate and SO 2 Emissions for the MOVES Model" for details).  Total PM2.5 as computed by
MOVES includes EC, OC and sulfate emissions.  Gaseous sulfur dioxide is also part of the fuel
sulfur balance, and is reported by the MOVES model.  It is not considered a paniculate in the
MOVES model, but can react in the atmosphere to form secondary particulate.
Since the fuel sulfur levels in  the underlying studies were not generally known, but believed to be
small (about 1 percent or less), sulfate emissions were ignored in the total PM2.5 emission levels.
As a result, total PM2.5 in this analysis was assumed to be comprised of only EC and OC.
                                    PM2  =EC+OC
Equation 15
Dividing both sides by PM2.5 and rearranging gives
                                                                                       30

-------
                                    oc  .=1.0- EC
                                   PM
                                       2.5
PM
                                                                                Equation 16
                                                     2.5
Thus, the OC fraction = 1.0 - EC fraction
The final EC fractions used in MOVES for pre-2007 model year trucks (i.e. before diesel
particulate filters (DPFs) were standard) are shown in Figure 6. These vary according to regulatory
class and MOVES operating mode. They typically range from 25 percent at low loads (low STP)
to over 90 percent at highly loaded modes. All of the EC fractions were developed in a separate
analysis and are documented in Appendix A.5.  The primary  dataset used in the analysis came from
Kweon et al. (2004) where particulate composition and mass rate data were collected on a
Cummins N14 series test engine over the CARB eight-mode engine test cycle.   The EPA PERE
model and a Monte Carlo approach were used to simulate and translate the primary PM emission
results into MOVES parameters (i.e., operating modes).
Figure 6. Elemental Carbon fraction by operating mode for pre-DPF-equipped trucks.






0.4



n
• • *
•


•
m

• • 1



* * 2 • • •
m
m


•

*






______



— *HHD
• MHD


        0   1  11  12 13 14  15  16  21  22  23  24  25  27 28 29 30 33  35  37  38  39  40
                                   Operating mode bin

A different methodology was used to compute EC factors for 2007 and later model year, DPF-
equipped vehicles.  For these vehicles it is believed that virtually all of the particulate that is
emitted from the tailpipe will be OC and that only a modest fraction will be EC. The traps are
designed to capture virtually all of the carbon. Potentially, small amounts of OC and sulfate may
escape.  This is essentially the opposite of the non-trap equipped heavy-duty diesel vehicles where
the total PM2.5 is dominated by EC. Unfortunately, only limited particulate data exists on trap
equipped vehicles.  These data are based on particulate-matter bound ionic species and EC/OC
emissions data from a few trap equipped buses and a heavy-duty tractor.  The data were extracted
and a simple average computed from a published source.27 Based on the  date of the paper, it is
likely that all of the diesel vehicle/trap systems were prototypes.  Extraction of data from the paper
                                                                                        ll

-------
yielded a single factor which will be applied to all regulatory types and operating modes for 2007
and later diesel trucks and buses. This factor is the elemental carbon fraction.  Table 18
summarizes the EC and OC fractions estimated from the paper.  These fractions were used for all
operating mode bins for model years 2007 and later heavy-duty vehicles.
          Table 18 shows EC and OC fractions used for DPF-equipped heavy-duty diesel vehicles.
EC fraction
OC fraction
0.0861
1 -ECfraction = 0.9139
As additional data become available, EPA will probably revise the EC Fraction used in MOVES
for these vehicles.
Temperature Correction Factors
The draft MOVES model released in March 2009 did not contain temperature correction factors for
PM2.5 emissions from heavy-duty diesel vehicles. This absence of temperature correction factors
does not imply that EPA believes that heavy-duty diesel vehicle PM emissions are insensitive to
temperature effects. In fact, it is quite likely that the reverse is true.  Both running and start PM
emissions from at least non-trap equipped vehicles are sensitive to temperature. However, EPA at
this time cannot adequately quantify such emission effects, and is currently using a multiplicative
placeholder value of 1.0 as the temperature correction factor.  EPA will update the MOVES model
when sufficient data on diesel temperature correction factors is available for analysis and inclusion
in the model.

1.1.2.3       Sample results
Figure 7 and Figure 8  show how PM rates increase with STP.  As with the NOX plots, the highest
operating modes in each speed range will rarely be attained due to the power limitations of heavy-
duty vehicles, but are included in the figures for completeness. At high speeds (greater than 50
mph; operating modes > 30), the overall PM rates are lower than the other speed ranges. For pre-
2007 model years the PM rates are dominated by EC. With the introduction of DPFs in model year
2007, we anticipate large reductions in overall PM rates and that the remaining PM will be
dominated by OC.
                                                                                        32

-------
Figure 7. Particulate Matter rates by operating mode representing Heavy heavy-duty vehicles (model year 2002
at age 0-3 years).

     160  -

     140  -

     120  -

   "SB 100
                           I EC

       80  H                IOC
   01
   E
   ro
   01
       40  -

       20  -

        0


           0  I  II 12  13  14 15 16 21 22 23 24 25 27 28 29  30  33 35 37 38 39 40
                                     Operating mode

Figure 8.  Particulate Matter rates by operating mode for Heavy heavy-duty vehicles (model year 2007 at age 0-
3 years).

     8  -t
     7  -
     6  -
     5
     3
                           I EC

                           IOC


        0   I  II 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
                                   Operating mode

Figure 9 shows an example of how tampering and mal-maintenance estimates increase PM with
age. The EC/OC proportion does not change by age, but the overall rate increases and levels off
after the end of useful life. This figure shows the age effect for MHD.  The rate at which emissions
increase toward their maximum depends on regulatory class.
                                                                                           33

-------
Figure 9.
mode 24).
Participate Matter rates by age group for Medium heavy-duty vehicles (model year 2002, operating
                     4-5
                     6-7       8-9       10-14     15-19      20+
                         Age group [years]
Figure 10 shows the effect of model year on emission rates. Emissions generally decrease with
new PM standards.  The EC fraction stays constant until model year 2007, when it is reduced to
nearly zero due to widespread DPF use.  The overall PM level is substantially lower starting in
model year 2007. The emission rates shown here for earlier model years are an extrapolation of
the T&M analysis since young-age engines from early model years could not be tested in the E-55
program.
                                                                                          34

-------
Figure 10. Participate Matter rates for Heavy heavy-duty vehicles by model year group (age 0-3 years,
operating mode 24).
                                                                          EC

                                                                          OC
         1960-1987 1988-1990 1991-1993 1994-1997 1998-2006 2007-2009 2010-2012 2013-2050
                                        Model year group
                                                                                                  35

-------
1.1.3 Hydrocarbons (HC) and Carbon Monoxide (CO)
Diesel engines account for a substantial portion of the mobile source HC or CO emission
inventories. Recent regulations on non-methane hydrocarbons (NMHC) (sometimes in conjunction
with NOx) combined with the common use of diesel oxidation catalysts will yield reductions in
both HC and CO emissions from heavy-duty diesel engines. As a result, data collection efforts do
not focus on HC or CO from heavy-duty engines.  In this report, hydrocarbons are sometimes
referred to as total hydrocarbons (THC).
We used certification levels combined with emissions standards to develop appropriate model year
groups. Since standards did not change frequently in the past for either HC or CO, we created
fewer model year groups than we did from NOx and PM. The HC/CO model year groups are:
    •  1960-1989
    •  1990-2006
    •  2007+

1.1.3.1 Data Sources
The heavy-duty diesel HC and CO emission rate development followed a methodology that
resembles the light-duty methodology, where emission rates were calculated from 1-hz data
produced from chassis dynamometer testing.  Data sources were all heavy-duty chassis test
programs:
   1.  CRC E-55/5922: Mentioned earlier, this program represents the largest volume of heavy-
       duty emissions data collected from chassis dynamometer tests.  All tests were used, not just
       those using the TEOM. Overall, 75 trucks were tested on a variety of drive cycles. Model
       years ranged  from 1969 to  2005, with testing conducted by West Virginia University from
       2001 to 2005.
   2.  Northern Front Range Air  Quality Study (NFRAQS)28:  This study was performed by
       the Colorado Institute for Fuels and High-Altitude Engine Research in 1997.  Twenty-one
       HD diesel vehicles from model years 1981 to 1995 selected to be representative of the in-
       use fleet in the Northern Front Range of Colorado were tested over three different transient
       drive cycles.
   3   New York Department of Environmental Conservation (NYSDEC)29:  NYSDEC
       sponsored this study to investigate the nature and extent of heavy-duty diesel vehicle
       emissions in the New York Metropolitan Area. West Virginia University tested 25 heavy-
       heavy and 12 medium-heavy duty diesel trucks under transient and steady-state drive
       cycles.
       4.  West Virginia University: Additional historical data collected on chassis
          dynamometers by WVU is available in the EPA Mobile Source Observation Database.
The on-road data used for the NOx analysis was not used since HC  and CO were not collected in
the MEMS program, and the ROVER program used the less accurate non-dispersive infrared
(NDIR) technology instead of flame-ionization detection (FID) to measure HC. To keep HC and
CO data sources consistent, we used chassis test programs exclusively for the analysis of these two
                                                                                      36

-------
pollutants.  Time-series alignment was performed using a method similar to that used for light-duty
chassis test data. The numbers of vehicles in the data sets are shown in Table 19.
Table 19 Numbers of vehicles by model year group, regulatory class, and age group .
Model year group
1960-2002
2003-2006
Regulatory class
HHD
MHD
Bus
LHD45
LHD2b3
HHD
Age group
0-3
58
9
26
2
6
6
4-5
19
6




6-7
16
5




8-9
9
4
1
1


10-14
16
12
3



15-19
6
15




20+
7
6




7.7.3.2
Analysis
As for PM, STP was calculated using an equation similar to the light-duty VSP equation, but
normalized with average regulatory class weight instead of test weight, as described by Equation
17.
                             STP =
                       Avt  + Bv2t + Cv] + mvtat
                                  f
                                 J scale
Equation 17
                                                                       23
The track road-load coefficients A, B, and C pertaining to heavy-duty vehicles  were estimated
through previous analyses for EPA's Physical Emission Rate Estimator (PERE). 21

Using a method similar to that used in the NOx analysis, we averaged emissions by vehicle and
operating mode.  We then averaged across all vehicles by model year group, age group, and
operating mode.  Estimates of uncertainty for each mean rate were calculated using the same
equations and methods used in development of the NOx rates.Instead of using our results to
directly populating all the emission rates, we directly populated only the age group that was most
prevalent in each regulatory class and model year group combination. These age groups are shown
in Table 20.
Table 20. Age groups used directly in MOVES emission rate inputs for each regulatory class and model year
group present in the data.
Regulatory class
HHD
HHD
MHD
BUS
LHD41
Model year group
1960-2002
2003-2006
1960-2002
1960-2002
1960-2002
Age group
0-3
0-3
15-19
0-3
0-3
                                                                                        37

-------
We then applied tampering and mal-maintenance effects through that age point, either lowering
emissions for younger ages or raising them for older ages, using the methodology described in
Appendix A.2 Tampering and Mal-maintenance.  The tampering and mal-maintenance effects for
HC and CO are shown in Table 21.
Table 21 Tampering and mal-maintenance effects for HC and CO over the useful life of trucks.
Model years
Pre-2003
2003-2006
2007-2009
20 10 and later
Increase in HC and CO
Emissions (%)
300
150
150
33
We multiplied these increases by the T&M adjustment factors in Table 37 in section A2.3
       Analysis to get the emissions by age group.
With the increased use of diesel oxidation catalysts (DOCs) in conjunction with DPFs, we assume
an 80% reduction in zero-mile emission rates for both HC and CO starting with model year 2007.

1.1.3.3 Sample results
The charts in this sub-section show examples of the emission rates that derived from the analysis
described above.  Not all rates are shown; the intent is to illustrate the most common trends and
hole-filling results. For simplicity, the light-heavy  duty regulatory classes are not shown, but since
the medium-heavy data were used for much of the light-heavy duty emission rate development, the
light-heavy duty rates follow similar trends.  Uncertainties were calculated as for NOx.
In Figure 11 and Figure 12, we see that HC and CO mean emission rates increase with STP, though
there is much higher uncertainty than for  the NOx rates.  This pattern could be due to the smaller
data set or may truly reflect a less direct correlation between HC,CO and STP.  In these figures, the
data for HHD and bus classes were combined to generate one set of rates for HHD and buses.
                                                                                        38

-------
Figure 11. THC emission rates [g/hr] by operating mode for model year 2002 and age group 0-3.
   100 -

    90 -

    80 -
 u
70 -

60 -

50 -
 8  40
    30 -

    20 -

    10 -
     0
                *MHD
                • HMD/Bus
         i  i  *
         0  I II 12 13 14  15 16 21 22 23  24 25 27 28 29 30 33 35 37 38 39 40
                                  Operating mode
Figure 12. CO emission rates [g/hr] by operating mode for model year 2002 and age group 0-3.
    600 -i
    500 -
  T 400
  I
  01
  ฃ 300
  8
  ro
  o>
    200
    100 -
             ปMHD
             • HMD/Bus
                ii
          •  •  *  *
                    ill
          0  I II 12 13 14 15 16 21 22 23 24 25 27 28 29  30  33  35  37 38 39 40
                                    Operating mode
                                                                                          39

-------
Figure 13 and Figure 14 show HC and CO emission rates by age group. Due to our projections of
T&M effects, there are large increases as a function of age. Additional data collection would be
valuable to determine if real-world deterioration effects are consistent with those in the model,
especially in model years where diesel oxidation catalysts are most prevalent (2007 and later).
Figure 13. THC emission rates [g/hr] by age group for model year 2002 and operating mode 24.
100 n
90 -
80 -
~ 70
.c
3ง 60
01
+j
2 50
u
E 40
ra
01
^ 30
20 -
10 -
n .






j
i
a




n "






-p j
II





-
n





j
n





-
11 BHHD
#MHD







A Bus
              0-3      4-5  6-7  8-9     10-14
                                     Age group [years]
15-19
20+
                                                                                           40

-------
Figure 14. CO emission rates [g/hr] by age group for model year 2002 and operating mode 24.

     600  -


     500  -
     400  -
   01
   2 300
  8
   ro
   01
     200  -
     100  -
                                         • HMD

                                         ปMHD

                                         A Bus
              0-3
4-5  6-7  8-9
   10-14        15-19
Age group [years]
20+
Figure 15 and Figure 16 show sample HC and CO emission rates by model year group. The two
earlier model year groups are relatively similar. The rates in the model year group reflect the use of
diesel oxidation catalysts.  Due to the sparseness of the data and the fact that HC and CO emission
do not correlate as well with STP (or power) as NOx and PM do, uncertainties are much greater.
Rates from HHD regulatory class were used for buses.  All regulatory classes have the same rates
for model years 2003 and later.
                                                                                         41

-------
Figure 15. THC emission rates by model year group for operating mode 24 and age group 0-3.
     30  -i
     25  -
     20
  ™  15
  u
  I  10
      5  -
                                            ปMHD
                                            • HMD/Bus
              1960-2002
                            2003-2006
                         Model year group
2007-2050
Figure 16. CO emission rates by model year group for operating mode 24 and age group 0-3.
    200 -
    180 -
    160 -
    140 -
 "53
 -120
  ฃ 100
  ra
80 -
60 -
40 -
20 -
 0
                                            ปMHD
                                            • HMD/Bus
              1960-2002
                            2003-2006
                         Model year group
2007-2050
1.1.4  Energy
The new data used to develop NOX rates also allowed us to develop new running-exhaust energy
rates. These were based on the same data, STP structure and calculation steps as in the NOx
analysis; however, unlike NOX, we did not classify the energy rates by model year or by age,
because neither variable had a significant impact on energy rates or CC>2.
                                                                                          42

-------
As for previous versions of MOVES, CC>2 emissions were used as the basis for calculating energy
rates. To calculate energy rates [kJ/hour] from CC>2 emissions, we used a heating value (HV) of
138,451 kJ/gallon and CC>2 fuel-specific emission factor (fcol) of 10,084 g/gallon for diesel fuel.
V    "=. V
'energy    CO2
                                              _   HV
                                                  fc
                                                                                   Equation 18
                                                   CO,
This analysis updates the running-exhaust energy rates estimated for MOVES2004 for diesel HHD,
MHD, and bus regulatory classes.19 The revised inputs are shown in Figure 17.
Figure 17 - Diesel running exhaust energy rates for MHD, HHD, and buses.
    V)

    O
       6  -I
       5  -
       4  -
    2
    01
    4-ป
    E
     ro  -,
     01  2
       1  -
          0  1  11 12 13 14 15  16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
                                   Operating mode

Compared to other emissions, the uncertainties in the energy rates are smaller in part because there
is no classification by age, model year, or regulatory class.  Thus, the number of vehicles used to
determine each rate is larger, providing for a greater certainty of the mean energy rate.
                                                                                           43

-------
1.2   Start Exh aust Emissions
The 'start' process occurs when the vehicle is started and is operating in some mode in which the
engine is not fully warmed up. For modeling purposes, we define start emissions as the increase in
emissions due to an engine start. That is, we use the difference in emissions between a test cycle
with a cold start and the same test cycle with a hot start.  There are also eight intermediate stages
which are differentiated by soak time length (time duration between engine key off and engine key
on), between a cold start (>  720 minutes of soak time) and a hot start FTP (< 6 minutes of soak
time).  More details on how  start emission rates are calculated as a function of soak time can be
found later in this section and in the MOVES light-duty emission rate counterpart document
Development of Emission Rates for Light-Duty Vehicles in the Motor Vehicle Emissions Simulator.
Start exhaust energy rates were not updated from previous MOVES analyses.

1.2.1  HC,CO,andNOx
For light-duty vehicles, start emissions are estimated by subtracting FTP bag 3 emissions from FTP
bag 1 emissions.  Bag 3 and  Bag 1 are the same dynamometer cycle, except that Bag 1 starts with a
cold start, and Bag 3 starts with a hot start.  A similar approach was performed for LFtD vehicles
tested on the FTP and ST01  cycles, which also have separate bags containing cold and hot start
emissions over identical drive cycles.  Data from 21 vehicles, ranging from model years 1988 to
2000, were analyzed. No classifications were made for model year or age due to the limited
number of vehicles. The results of this analysis for HC, CO, and NOx are shown in Table 22:
            Table 22 shows average start emissions increases for light-heavy duty vehicles (g).
HC
0.13
CO
1.38
NOx
1.68
For HHD and MHD trucks, data were unavailable. To provide at least a minimal amount of
information, we measured emissions from a 2007 Cummins ISB on an engine dynamometer at the
EPA National Vehicle and Fuel Emissions Laboratory in Ann Arbor, Michigan. Among other idle
tests, we performed a cold start idle test at 1,100 RPM lasting four hours, long enough for the
engine to warm up. Essentially, the "drive cycle" we used to compare cold start and warm
emissions was the idle cycle, analogous to the FTP and ST01 cycles used for LFtD vehicles.
Emissions and temperature stabilized about 25 minutes into the test. The emission rates through
time are shown in Figure 18.  The biggest drop in emission rate through the test is with CO,
whereas there is a slight increase in NOx (cold start NOx is lower than hot start NOx), and an
insignificant change in HC.
                                                                                      44

-------
Figure 18. Trends in the stabilization of idle emissions from a diesel engine following a cold start. Data were
collected from a 2007 Cummins ISB measured on an engine dynamometer.
                     0.00
                               1.00
                                          2.00        3.00
                                             time [hrs]
                                                               4.00
                                                                         5.00
We calculated the area under each trend for the first 25 minutes and divided by 25 minutes to get
the average emission rate during the cold start idle portion.  Then, we averaged the data for the
remaining portion of the test, or the warm idle portion. The difference between cold start and
warm start is in Table 23.  The NOx increment is negative since cold start emissions are lower than
warm start emissions.
               Table 23. Cold-start emissions increases in grams on the 2007 Cummins ISB.
HC
0.0
CO
16.0
NOx
-2.3
We also considered data from University of Tennessee30, which tested 24 trucks with PEMS at
different load levels during idling.  Each truck was tested with a cold start going into low-RPM idle
with air-conditioning on. We integrated the emissions over the warm-up period to get the total cold
start idling emissions. We calculated the hot-start idling emissions by multiplying the reported
warm idling rate by the  stabilization time. We used the stabilization period from our engine
dynamometer tests (25 minutes). Then we subtracted the cold-idle emissions from the warm idle
emissions to estimate the cold start increment. We found that several trucks produced lower NOx
emissions during cold start (similar to our own work), and several trucks produces higher NOx
emissions during cold start.  Due to these conflicting results, and the recognition that many factors
affect NOx emission during start (e.g. air-fuel ratio, injection timing, etc), we set the cold-start
increment to zero.  Table 24 shows our final MOVES inputs for HHD and MHD  diesel start
emissions increases.  The HC and CO estimates are from our 2007 MY in-house testing.
                                                                                          45

-------
               Table 24 MOVES inputs for HHD and MHD start emissions(grams/start).
HC
0.0
CO
16.0
NOx
0.0
1.2.2  Particulate Matter
Data for particulate start emissions from heavy-duty vehicles are rare.   Typically, heavy-duty
vehicle emission measurements are performed on fully warmed up vehicles. These procedures
bypass the engine crank and early operating periods when the vehicle is not fully warmed up.
Data from engine dynamometer testing performed on one heavy-heavy-duty engine, using the FTP
cycle with particulate mass collected on filters. The engine was manufactured in MY2004. The
cycle was repeated six times, under both hot and cold start conditions (two tests for cold start and
four replicate tests for hot start).  The average difference in PM2.5 emissions (filter measurement -
FTP cycle) was 0.10985 grams.  The data are shown here:
       Cold start FTP average       =      1.9314 g PM2.5
       Warm start FTP average     =      1.8215 g PM2.5

       Cold start - warm start       =      0.1099 g PM2.5

We applied this value to 1960 through 2006  model year vehicles.  A corresponding value of
0.01099 g was used for 2007 and later model year vehicles (90% reduction due to DPFs). We plan
to update this valuewhen more data becomes available.

1.2.3  Adjusting Start Rates for Soak Time

The discussion to this point has concerned the development of rates for cold-start emissions. In
addition, it was necessary to derive rates for additional operating modes that account for varying
(shorter) soak times.  As with light-duty vehicles, we accomplished this step by applying soak
fractions. As no data are available for heavy-duty vehicles, we applied the same fractions used for
light-duty emissions.   Table 25 describes the different start-related operating modes in MOVES as
a function of soak time.  The value at 720 min (12 hours) represents cold start. These modes are
not related to the operating modes defined in Table 9, which are for running exhaust emissions.
                                                                                       46

-------
                Table 25. Operating modes for start emissions (as a function of soak time)
Operating Mode
101
102
103
104
105
106
107
108
Description
Soak Time < 6 minutes
6 minutes <= Soak Time < 3
30 minutes <= Soak Time <
60 minutes <= Soak Time <
90 minutes <= Soak Time <
0 minutes
60 minutes
90 minutes
120 minutes
120 minutes <= Soak Time < 360 minutes
360 minutes <= Soak Time < 720 minutes
720 minutes <= Soak Time
The soak fractions we used for HC, CO, and NOx are illustrated in Figure 19 below. (Although,
since our current estimate for NOx starts is zero, the NOx fractions are currently irrelevant.)
Figure 19.  Soak Fractions Applied to Cold-Start Emissions (opModelD = 108) to Estimate Emissions for
shorter Soak Periods (operating modes 101-107).
             1.20
                                      240         360         480
                                           Soak Time (minutes)
600
720
The actual PM start rates by operating mode are given in Table 26 below.
                                                                                             47

-------
          Table 26. Participate Matter Start Emission Rates by Operating Mode (soak fraction).
Operating Mode
101
102
103
104
105
106
107
108
PM2.5 (grams per start)
1960-2006 MY
0.0000
0.0009
0.0046
0.0092
0.0138
0.0183
0.0549
0.1099
PM2.5 (grams per start)
2007+ MY
0.00000
0.00009
0.00046
0.00092
0.00138
0.00183
0.00549
0.01099
1.3    Ext en ded Idling Exh aust Emissions
In the MOVES model, extended idling is "discretionary" idle operation characterized by idle
periods more than an hour in duration, typically overnight, including higher engine speed settings
and extensive use of accessories by the vehicle operator. Extended idling most often occurs during
long layovers between trips by long-haul trucking operators where the truck is used as a residence,
and is sometimes referred to as "hotelling." The use of accessories such as air conditioning systems
or heating systems will affect emissions emitted by the engine during idling. Extended idling by
vehicles will also allow cool-down of the vehicle's catalytic converter system or other exhaust
emission after-treatments, when these controls are present. Extended idle is treated as a separate
emission process in MOVES.
Extended idling does not include vehicle idle operation which occurs during normal road operation,
such as the idle operation which a vehicle experiences while waiting at a traffic signal or during a
relatively short stop, such as idle operation during a delivery.  Although frequent stops and idling
can contribute to overall emissions, these modes are already included in the normal vehicle hours
of operation. Extended idling is characterized by idling periods that last hours, rather than minutes.
In the MOVES model, diesel long-haul combination trucks are the only sourceType assumed to
have any significant extended idling activity.  As a result, an estimate for the extended idling
emission rate has not been made for any of the other source use types modeled in MOVES.

1.3.1  Data Sources
The data used in the  analysis of extended idling emission rates includes idle emission results from
several test programs conducted by a variety of researchers at different times.  Not all of the studies
included all the pollutants of interest. The references contain more detailed descriptions of the data
and how the data was obtained.

   •   Testing was  conducted  on twelve heavy-duty diesel trucks and twelve transit buses in
       Colorado (McCormick)31.  Ten of the trucks  were Class 8 heavy-duty axle semi-tractors,
       one was a Class 7 truck, and one of the vehicles was a school bus. The model year ranged
                                                                                        48

-------
       from 1990 through  1998.  A  typical Denver area wintertime diesel fuel (NFRAQS) was
       used in all tests.  Idle measurements were collected during a 20 minute time period.  All
       testing was done at 1,609 meters above sea level (high altitude).

   •   Testing was conducted by EPA on five trucks in May 2002 (Lim)32.  T he model years
       ranged from 1985 through 2001. The vehicles were put through a battery of tests including
       a variety of discretionary and non-discretionary idling conditions.

   •   Testing was conducted on 42 diesel trucks in parallel with roadside smoke opacity testing in
       California (Lambert)33.  All tests  were conducted by the California Air Resources Board
       (CARB) at a rest area  near Tulare,  California in April 2002.  Data collected during  this
       study were included in the data provided by IdleAire Technologies (below) that was used in
       the analysis.

   •   A total of 63 trucks (nine in Tennessee, 12 in New York and 42 in California) were tested
       over a battery of idle test conditions including with and without air conditioning (Irick)34.
       Not all trucks were tested under all conditions.  Only results from the testing in Tennessee
       and New York are described in  the IdleAire report.  The Tulare, California, data are
       described in the Clean Air Study cited above. All analytical equipment for all testing at all
       locations was operated by Clean Air Technologies.

   •   Fourteen trucks were tested as part of a large Coordinating Research Council (CRC) study
       of heavy duty diesel trucks with idling times either 900 or 1,800 seconds long (Gautam)35.

   •   The  National Cooperative  Highway Research Program (NCHRP)36 obtained  the idling
       portion of continuous  sampling during transient testing was used to determine idling
       emission rates on two trucks.

   •   A total of 33 heavy-duty diesel trucks were tested in an internal study by the City of New
       York (Tang)37.  The model years ranged from 1984 through 1999.  One hundred seconds of
       idling were added at the end of the WVU five-mile transient test driving cycle.

   •   A  Class  8 F reightliner Century  with  a 1999 e ngine was  tested using EPA's  on-road
       emissions  testing  trailer based in Research Triangle Park, North  Carolina (Broderick)38.
       Both short (10 minute) and longer (five hour) measurements were made during idling.
       Some testing was also done  on three older trucks.

   •   Five heavy-duty trucks were tested for particulate and NOx  emissions under a variety of
       conditions at Oak Ridge Laboratories (Story)39.  These are the same trucks used in the EPA
       study (Lim).

   •   The University of Tennessee  tested  24 1992 t hrough 2006 model year heavy duty diesel
       trucks using a variety of idling conditions including variations of engine idle speed and load
       (air conditioning)30.

1.3.2  Analysis
EPA estimated mean emission rates during extended idling operation for particulate matter (PM),
oxides of nitrogen (NOx), hydrocarbons (HC), and carbon monoxide (CO).  This analysis used all
of the data sources referenced above.  This update reflects new data available since the initial
development of extended idle emissions for the MOVES model. The additions include the testing

                                                                                       49

-------
at Research Triangle Park (Broderick), the University of Tennessee study (Calcagno), and the
completed E-55/59 study conducted by WVU and CRC.  In addition, the data was separated by
truck and bus and by idle speed and accessory usage to develop an emission rate more
representative of extended idle rates.
The important conclusion from the 2003 analysis was that factors affecting engine load, such as
accessory use, and engine idle speed are the important parameters in estimating the emission rates
of extended idling.  The impacts of most other factors, such as engine size, altitude, model year
within MOVES groups, and test cycle are negligible. This makes the behavior of truck operators
very important in estimating the emission rates to assign to periods of extended idling.
The use of accessories (air conditioners, heaters, televisions, etc.) provides recreation and comfort
to the operator and increases load on the engine. There is also a tendency to increase idle speed
during long idle periods for engine durability. The emission rates estimated for the extended idle
pollutant process assume both accessory use and engine idle speeds set higher than used for "curb"
(non-discretionary) idling.
The studies focused on three types of idle conditions. The first is considered a curb idle, with low
engine speed (<1,000  rpm) and no air conditioning. The second is representative  of an extended
idle condition with higher engine speed (> 1,000 rpm) and no air conditioning.  The third represents
an extended idle condition with higher engine speed (> 1,000 rpm) and air conditioning.
The idle emission rates for heavy duty diesel trucks prior to the 1990 model year are  based on the
analysis of the 18 trucks from 1975-1990 model years used in the CRC E-55/59  study and one
1985 truck from the Lim study.  The only data available represents a curb idle condition. No data
was available to develop the elevated NOx emission rates characteristic of higher engine speed and
accessory loading, therefore, the percent increase developed from the 1991-2006 trucks was used.
Extended idle emission rates for 1991-2006 model year heavy duty diesel trucks are based on
several studies and 184 tests detailed in Appendix A. 3 Extended Idle Data Summary. The increase
in NOx emissions due to  higher idle speed and air conditioning was  estimated based on three
studies that included 26 tests. The average emissions from these trucks using the high idle engine
speed and with accessory loading was used for the emission rates for extended idling.
The expected effects of the 2007 heavy duty diesel vehicle emission standards on extended idling
emission rates are taken from the EPA guidance analysis (EPA 2003). The 2007 heavy duty diesel
emission standards are expected to result in the widespread use of PM filters  and exhaust gas
recirculation (EGR) and 2010 standards will result in after-treatment technologies. However, since
there is no requirement to address extended idling emissions in the emission certification
procedure, EPA expects that there will be little effect on HC, CO, and NOx emissions after hours
of idling due to cool-down effects on EGR and most aftertreatment systems.  However, we do not
expect DPFs to lose much effectiveness during extended idling. As a result, we project that idle
NOx emissions will be reduced 12% and HC and CO emissions will be reduced 9% from the
extended idle emission rates used for 1988-2006 model year trucks. The reduction estimates are
based on a ratio of the 2007 standard to the previous standard and assuming that the emission
control of the new standard will  only last for the first hour of an eight hour idle.  For PM, we
assume  an extended idling emission rate equal to the curb idling rate (operating mode 1 from the
running exhaust analysis). Detailed equations are included in the appendix.
                                                                                        50

-------
1.3.3  Results
Table 27 shows the resulting NOx, HC, and CO emission rates estimated for heavy-duty diesel
trucks. Extended idling measurements have large variability due to low engine loads, which is
reflected in the variation of the mean statistic.
                          Table 27. Extended idle emission rates (g/hour).
Model years
Pre-1990
1990-2006
2007 and later
NOx
112
227
201
HC
108
56
53
CO
84
91
91
PM
8.4
4.0
0.2
                                                                                         51

-------
2      Heavy-Duty Gasoline Vehicles
2.1   Running Exh aust Emissions

2.1.1 HC,CO,andNOx
2.7.7.7
Data and Analysis
As gasoline-fueled vehicles are a small percentage of the heavy-duty vehicle fleet, the amount of
data available for analysis was small.  We relied on four medium-heavy duty gasoline trucks from
the CRC E-55 program and historical  data from EPA's Mobile Source Observation Database
(MSOD), which has results from chassis tests performed by both EPA, contractors and outside
parties.  The heavy-duty gasoline data in the MSOD is mostly from pickup trucks which fall mainly
in the LHD2b3 regulatory class.  Table 28 shows the number of vehicles in cumulative data sets.
In the real world, most heavy-duty gasoline vehicles fall in either the LHD2b3 or LHD45 class,
with a smaller percentage in the MHD class. There are very few, if any, HHD gasoline trucks
remaining in use.
    Table 28. Distribution of vehicles in the data sets by model-year group, regulatory class and age group.
Model year group
1960-1989
1990-1997
1998-2002
Regulatory class
MHD
LHD2b3
MHD
LHD2b3
MHD
LHD2b3
Age group
0-5



33
1
1
6-9
2
10
1
19


Similar to the HD diesel PM, HC, and CO analysis, the chassis vehicle speed and acceleration,
coupled with the average weight for each regulatory class, were used to calculate STP (Equation
13).  To supplement the meager data available, we examined certification data as a guide to
developing model year groups for analysis.  Figure 20 shows averages of certification results by
model year.
                                                                                     52

-------
Figure 20. Brake-specific certification emission rates by model year for heavy-duty gasoline engines.
                o







-•
..
• CO
-.-I' A NOX

i • 1
, , ! * 1 1 [juf'iij
1 i siHr^11:
5lI**I*r, ,






i














- 1.2
- 1
- 0.8 F
Q.
ฃ

- 0.4 i
- 0.2
- 0
                   1980     1985      1990     1995     2000
                                          Model year
                                                             2005
                                                                     2010
Based on these certification results, we decided to classify the data into the coarse model year
groups listed below.

    •   1960-1989

    •   1990-1997

    •   1998-2002

    •   2003-2006

    •   2007 and later
Although there was little data for 2007-and-later, we made a split at model year 2007 to account for
possible increases in three-way catalyst use and efficiency due to tighter NOx standards. We
assumed that these catalysts in gasoline vehicles will yield a reduction in HC and CO also. We
estimate that each of these three pollutants will decrease 70% from 2003-2006 MY levels.
Unlike the analysis for HD diesel vehicles, we used the age effects present in the data itself.  We
did not incorporate external tampering and mal-maintenance assumptions into the HD gasoline
rates. Due to sparseness of data we used only the two age groups listed in Table 28. We also did
not classify by regulatory class since there was only one regulatory class  (LHD2b3) predominantly
represented in the data.

2.1.1.2 Sample Results
Selected  results are shown graphically below. The first (Figure 21) shows all three pollutants vs.
operating mode for the LHD2b3 regulatory class.  In general, emissions follow the expected trend
with STP, though the trend is most pronounced for NOX. As expected, NOx emissions for heavy-
duty gasoline vehicles are much lower than for heavy-duty diesel vehicles.
                                                                                          53

-------
                Figure 21. Emission Rates by operating mode for MY 1994 at age 0-3 years.
450 -,
400 -
-. 350
"SB
"300
0)
2
|250
1 200
ra
U
^ 150
(0
Ol
2 100

50 -
n .

A
ANOx
ปHC
A
• CO

A
A
A


A
A
• A
A ป
• A •

• 1
5000
4500
4000
- 3500 „
.c
- 3000 3ง
0)
- 2500 2
8
- 2000 =
CD
- 1500

1000

- 500
- n
               1  11 12 13  14  15  16 21 22 23 24  25  27  28 29 30 33 35  37  38 39 40
                                        Operating mode
Figure 22 shows the emissions trends by age group.  Since we did not use the tampering and mal-
maintenance methodology as we did for diesels, the age trends reflect our coarse binning with age.
For each pollutant, only two distinct rates exist - one for ages 0-5 and another for age 6 and older.
              Figure 22. Emission ratess by age group for MY 1994 in operating mode bin 24.

       120 -i                                                                      r  1400
0-3      4-5    6-7   8-9       10-14
                      Age group [years]
                                                           15-19
20+
„ 100
"

_C
% 80
ro

ป
X
1 60 -





C
(0
(J
I 40
c
ra
Ol
S 20
n -










L i
i

h-ป— 1





|J

i \







^ I
i

. I








j

t \


n
•^




-L

j

1 1


II
-^




-L

j

^ k


n







j

1 1


n
-^




ANOx

j




n







#HC
• CO

1200
- 1000 -^
^
"SB

- 800 o.
^
E
O
- 600 <->
c
ra
01
400 2

- 200
- n
                                                                                            54

-------
Figure 23 shows emissions by model year group.  Emissions generally decrease with model year
group. Uncertainties are relatively high but not shown in this plot for clarity.
             Figure 23. Emission rates by model year group for age 0-3 in operating mode 24.
         120 -


         100
       Si  80
       E
       •D
       ra
       ra
       01
          60
          40
          20 -
                                                        HC

                                                        NOx

                                                        CO
                            -  600


                            -  500


                            -  400 I
                                  0)
                              300  E
                                 8
                              200  $
                                 ง

                            -  100
                                                                               0
                  1960-1989
                    1990-1997
1998-2006
2007-2050
Assumptions regarding the increased effectiveness of catalysts substantially reduce emissions
estimates for 2007 model year and later.
2.1.2  Particulate Matter
Unfortunately, the PM2.5 emission data from heavy-duty gasoline trucks are too sparse to develop
the detailed emission factors the MOVES model is designed for.  As a result, only a very limited
analysis could be done. EPA will likely revisit and update these emission rates when sufficient
additional data on PM2.5 emissions from heavy-duty gasoline vehicles become available.
ForMOVES2010, the heavy-duty gas PM2.5 emission rates will be calculated by multiplying the
light-duty gasoline truck PM2.5 emission rates by a factor of 1.40, as explained below^ Since the
MOVES light-duty gasoline PM2.5 emission rates comprise a complete set of factors - classified
by particulate sub-type (elemental and organic carbon), operating mode, model year and regulatory
class, the heavy-duty PM2.5 emission factors will also be a complete set.
2.7.2.7
Data Sources
This analysis is based on the PM2.5 emission test results from the four gasoline trucks tested in the
CRC E55-E59 test program. The specific data used were collected on the UDDS test cycle.  Each
of the four vehicles in the sample received two UDDS tests, conducted at different test weights.
Other emission tests using different cycles were also available on the same vehicles, but were not
used in the calculation.   The use of the UDDS data enabled the analysis to have a consistent
driving cycle. The trucks and tests are described in Table 29.
                                                                                         55

-------
               Table 29. Summary of data used in HD gasoline PM emission rate analysis.
Vehicle
1
2
3
4
MY
2001
2001
1983
1983
1993
1993
1987
1987
Age
3
3
21
21
12
12
18
18
Test cycle
UDDS
UDDS
UDDS
UDDS
UDDS
UDDS
UDDS
UDDS
GVWR
[Ib]
12,975
19,463
9,850
14,775
13,000
19,500
10,600
15,900
PM2.5 mg/mi
1.81
3.61
43.3
54.3
67.1
108.3
96.7
21.5
The  table shows only four vehicles,  two of which  are quite old and certified to fairly lenient
standards. A third truck is also fairly old at 12 years and certified to an intermediate standard. The
fourth is a relatively new truck at age three and certified to a more stringent standard. No trucks in
the sample are certified to the Tier2 or  equivalent standards.

2.1.2.2        Analysis
Examination of the heavy-duty data shows two distinct levels: vehicle #1 (MY 2001) and the other
three vehicles. Because of its lower age (3 years old)  and newer model year status, this vehicle has
substantially lower PM emission levels than the others, and was separated in the analysis.  The
emissions of the other three vehicles were averaged together to produce these mean results:
Mean for Vehicles 2 through 4:      65.22 mg/mi
Mean for Vehicle 1:                 2.71 mg/mi
Older Group
Newer Group
To compare these rates with rates from light-duty gasoline vehicles, we simulated UDDS cycle
emission rates based on MOVES light-duty gas PM2.5 emission rates (with normal deterioration
assumptions) for light-duty gasoline trucks.   The UDDS cycle represents standardized operation
for the heavy-duty vehicles.
To make the comparisons appropriate, the simulated light-duty UDDS results were matched to the
results from the four heavy-duty gas trucks in the sample.  This comparison meant that the
emission rates from the following MOVES model year groups and age groups for light-duty trucks
were used:
                                                                                        56

-------
   •   MY group 1983-1984, age 20+

   •   MY group 1986-1987, age 15-19

   •   MY group 1991-1993, age 10-14

   •   MY group 2001, age 0-3
The simulated HDDS emission factors for the older light-duty gas truck group are 36.2 mg/mi for
MOVES organic carbon PM2.5 emissions and 2.641 mg/mi for elemental carbon.  Ignoring sulfate
emissions (on the order of IxlO"4 mg/mile for low sulfur fuels), these values sum to 38.84 mg/mile.
                                           ฃ*.^ OO ง
This value leads to the computation of the ratio: —:	— = 1.679 .
                                           38.84^

The simulated HDDS emission rates for the newer light-duty gas truck group are 4.368 mg/mi for
MOVES organic carbon PM2.5 emissions and 0.3187 mg/mi for elemental carbon.  Ignoring
sulfate emissions (which are in the order of IxlO"5 mg/mile for low sulfur fuels), these values sum
to 4.687 mg/mi.
                                            2.71Jk
This value leads to the computation of the ratio: —:—^^- = 0.578.
                                           4.687^
The newer model year group produces a ratio which  is less than one and implies that large trucks
produce less PM2.5 emissions than smaller trucks. This result is intuitively inconsistent, and is the
likely result of a very small sample and a large natural variability in emission results.
All four data points were retained and averaged together by giving the older model year group a 75
percent weighting and the newer model year group (MY 2001) a 25 percent weighting.  This is
consistent with the underlying data sample. It produces a final ratio of:
                   Ratiofmal = Ratio MeWtFrac + Rationewer (1 - WtFrac]

                             = 1.679x0.75 + 0.578x0.25 = 1.40
We then multiplied this final ratio of 1.40 by the light-duty gasoline truck PM rates to calculate the
input emission rates for heavy-duty gasoline PM rates.

2.1.3  Energy Consumption
The data used to develop heavy-duty running exhaust gasoline rates were the same as those used
for HC, CO, and NOX. However, new energy rates were only developed for MHD, HHD, and bus
classes. Analyses performed for LHD vehicles were not updated in this analysis. Also, similarly to
the diesel running exhaust energy rates, classifications were not made based on model year group,
age, or regulatory class. To calculate energy rates (kJ/hour) from CO2 emissions, we used a
heating value (HV) of 122,893 kJ/gallon and CO2 fuel-specific emission factor (fco2) of 8,788


                                                                                       57

-------
g/gallon for gasoline (see Equation 18). STP was calculated using Equation 13. Figure 24
summarizes the gasoline running exhaust energy rates stored in MOVES.
Figure 24 . Gasoline running exhaust energy rates for MHD, HHD, and buses.
       6 -i
     c
     o
       5 -
       4 -
     01
     E
       3 -
     c
     ra
     01
       1 -
          0  1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40
                                  Operating mode

A linear extrapolation to determine rates at the highest operating modes in each speed range was
performed analogously to diesel energy and NOX rates (see Section 1.1.1.3.3     Hole Filling and
forecasting).

2.2   Start Emissions

2.2.1  Available Data
To develop start emission rates for heavy-duty gasoline-fueled vehicles, we extracted data available
in the USEPA Mobile-Source Observation Database (MSOD). These data represent aggregate test
results for heavy-duty spark-ignition (gasoline powered) engines measured on the Federal Test
Procedure (FTP) cycle. The GVWR for all trucks was between 8,500 and 14,000 Ib, placing all
trucks in the LHD2b3 regulatory class.
Table 30 shows the model-year by age classification for the data. The model year groups in the
table were assigned based on the progression in NOx standards between MY 1990 and 2004.
Standards for CO and HC are stable over this period,  until MY 2004, when a combined NMHC+
NOx standard was introduced. However, no measurements for trucks were available for MY2004
or later.
                                                                                        58

-------
    Table 30. Model-year Group by Age Group Structure for the Sample of Heavy-Duty Gasoline Engines
Model-year Group

1960-1989
1990
1991-1997
1998-2004
Total
Standards (g/hp-hr)
CO

14.4
14.4
14.4

HC

1.1
1.1
1.1

NOx

6.0
5.0
4.0

Age Group (Years)
0-3


73
8
81
4-5


59

59
6-7

1
32

33
8-9
19
29
4

52
10-14
22



22
Total

41
30
168
8
247
2.2.2  Estimation of Mean Rates
As with light-duty vehicles, we estimated the "cold-start" as the mass from the cold-start phase of
the FTP (bag 1) less the "hot-start" phase (Bag 3). As a preliminary exploration of the data, we
averaged by model year group and age group and produced the graphs shown in Appendix A. 6
Heavy-duty Gasoline Start Emissions Analysis Figures.
Sample sizes are small overall and very small in some cases (e.g.  1990, age 6-7) and the behavior
of the averages is somewhat erratic. In contrast to light-duty vehicle emissions,  strong model-year
effects are not apparent. This may not be surprising for CO or HC, given the uniformity of
standards throughout. This result is more surprising for NOx but model year trends are no more
evident for NOx than for the other two. Broadly speaking, it appears that an age trend may be
evident.
If we assume that the underlying population distributions are approximately log-normal, we can
visualize the data in ways that illustrate underlying relationships.  As a first step, we calculated
geometric mean emissions, for purposes of comparison to the arithmetic means  calculated by
simply averaging the data. Based on the assumption of log-normality, the geometric mean (jtg) was
calculated in terms of the logarithmic mean (xj)  as
                                               Inx,
                                                                              Equation 19
This measure is not appropriate for use as an emission rate, but is useful in that it represents the
"center" of the skewed parent distribution. As such, it is less strongly influenced by unusually high
or outlying measurements than the arithmetic means in Appendix A.6 Heavy-duty Gasoline Start
Emissions Analysis Figures.  In general, the small differences between geometric means and
arithmetic means suggest that the distributions represented by the data do not show strong skew in
most cases.  Assuming that emissions distributions should be strongly skewed suggests that these
data are not representative of "real-world" emissions  for these vehicles. This conclusion appears to
be reinforced by the values in Figure 30 which represent the "logarithmic standard deviation"
calculated by model-year and age groups. This measure (sj), is the standard deviation of natural
logarithm of emissions (xj) in . The values of s/ are highly variable, and generally less than 0.8,
showing that the degree of skew in the data is also highly  variable as well as generally low for
                                                                                        59

-------
emissions data; e.g., corresponding values for light-duty running emissions are generally 1.0 or
greater. Overall, review of the geometric means confirms the impression of age trends in the CO
and HC results, and the general lack of an age trend in the NOx results.
Given the conclusion that the data as such are probably unrepresentative, assuming the log-normal
parent distributions allows us to re-estimate the arithmetic mean after assuming reasonable values
for 5;. For this calculation we assumed values of 0.9 for CO and HC and 1.2 for NOx. These values
approximate the maxima seen in these data and are broadly comparable to rates observed for light-
duty vehicles.
The re-estimated arithmetic means are calculated from the geometric means, by adding a term that
represents the influence of the "dirtier"  or "higher-emitting" vehicles,  or the "upper tail of the
distribution," as shown in Figure 31 above.
_   „ 2
                                                                             Equation 20
For purposes of rate development using these data, we concluded that a model-year group effect
was not evident and re-averaged all data by age Group alone. Results of the coarser averaging are
presented in Figure 25  with the arithmetic  mean  (directly  calculated and  re-estimated)  and
geometric means shown separately.
We then addressed the question of the projection of age trends. As a general principle, we did not
allow emissions to decline with age. We implemented this assumption by stabilizing emissions at
the maximum level reached between the 6-7 and 10-14 age groups.
                                                                                        60

-------
Figure 25. Cold-start FTP Emissions for Heavy-Duty Gasoline Trucks, averaged by Age Group only (g:
geometric mean, a= arithmetic mean recalculated from xt and s/).
                                            Age (years)
2.2.3 Estimation of Uncertainty
We calculated standard errors for each mean in a manner consistent with the re-calculation of the
arithmetic means.  Because the (arithmetic) means were recalculated with assumed values of 5;, it
was  necessary to  re-estimate corresponding standard deviations for the parent  distribution s, as
shown in Equation 21.
                                                                                          61

-------
                                      -J,
x1 es (es  -1)                         Equation 21
After recalculating the standard deviations, the calculation of corresponding standard errors was
simple. Because each vehicle is represented by only one data point, there was no within-vehicle
variability  to  consider, and  the standard  error could be  calculated as s/4n.  We divided the
standard errors by their respective means to obtain CV-of-the-mean or "relative standard error."
Means, standard deviations and uncertainties are presented in Table 31 and in Figure 26. Note that
these results represent only "cold-start" rates (opModelD  108).
                                                                                          62

-------
Table 31.  Cold-Start Emission Rates (g) for Heavy-Duty Gasoline Trucks, by Age Group
(italicized values replicated from previous age Groups).
Age Group

n

Pollutant
CO
THC
NOx
Means
0-3
4-5
6-7
8-9
10-14
81
59
33
52
22
101.2
133.0
155.9
190.3
189.1
6.39
7.40
11.21
11.21
11.21
4.23
5.18
6.12
7.08
7.08
Standard Deviations
0-3
4-5
6-7
8-9
10-14





108.1
142.0
166.5
203.2
202.0
6.82
7.90
11.98
11.98
11.98
8.55

12.39
14.32
14.32
Standard Errors
0-3
4-5
6-7
8-9
10-14





12.01
18.49
28.98
28.18
43.06
0.758
1.03
2.08
2.08
2.08
0.951
1.18
2.16
1.99
1.99
                                                                                   63

-------
Figure 26. Cold-start Emission Rates for Heavy-Duty Gasoline Trucks, with 95% Confidence Intervals

250
.3200
•K
•5 i^n
y> 1OU
2
5inn
t c-n

(
(a) CO
V-v *^*^
T
I^^T
I

D 5


,x










10
AgeO






15
^ears)






20






2

14

10

5

2
n
/l_\ -TI l^\
(b)THC



•H^1
f J

>
/
-





















3

f i i i i



                                         10         15

                                          Age (years)
                                                            20
                                                                      25
                    12
                    ID-
                  'S 8-
                        (c)NQc
                                         10        15
                                           Age (years)
                                                             20
                                                                       25
2.2.4 Projecting Rates beyond the Available Data
The steps described so far involved reduction and analysis of the available emissions data. In the
next step, we describe approaches used to impute rates for model years not represented in these
data. For purposes of analysis we delineated three model year groups: 1960-2004, 2005-2007 and
2008 and later. We describe the derivation of rates in each group below.
                                                                                         64

-------
2.2.4.1 Regulatory class LHD2b3
For CO the approach was simple. We applied the values in Table 31 to all model-year groups. The
rationale  for this approach is that the CO standards do not change over the full range of model
years considered.
For HC and NOx we imputed values for the 2005-07 and 2008+ model-year groups by multiplying
the values in Table 31  by ratios expressed in terms of the applicable standards. Starting in 2005, a
combined HC+NOx standard was introduced. It was necessary for modeling purposed to partition
the standard into HC and NOx components. We assumed that the proportions of NMHC and NOx
would be similar to those in the 2008 standards, which separate NMHC and NOx while reducing
both.
We calculated the HC value by multiplying the 1960-2004 value by the fraction^c, where

                          (    0.14g/hp-hr    \0gfo-to)
                        _ ( (0.14 + 0.20) g/hp-hr f            _                 Equation 22
                     /TT/-1 	                                   	 \J . J /
                      HC               1.1 g/hp-hr
This ratio represents the component of the 2005 combined standard attributed to NMHC.
We calculated the corresponding value for NOx as
                               0.20 g/hp-hr     \
                                                1.0 g/hp-hr
                           (0.14 + 0.20) g/hp-hr [          -nn?            Equation 23
                                                           ~ 0.147
                     NOX               , n  ,,    ,
                                     4.0 g/hp-hr

For these rates we neglected the THC/NMHC conversions, to which we gave attention for light-
duty.
In 2008, separate HC and NOx standards were introduced.  To estimate values for this model-year
group, we calculated the values by multiplying the 1960-2004 value by the fractions/kc and/NOx
where
                                 -    0.14g/hp-hr
                                /HC = —-—;	r~ = 0.127                      Equation 24
                                      1.1 g/hp • hr
                                 ,.     0.20 g/hp-hr
                                /NOX =  . ..  ,.—— = ฐ-05                      Equation 25
                                       4.0 g/hp-hr
                                                                                      65

-------
2.2.4.2       Regulatory classes LHD45 andMHD
For LHD45 and MHD, we estimated values relative to the values calculated for LHD2b3.
For CO and HC,  we estimated values for the heavier vehicles by multiplying them by ratios of
standards for the heavier class to those for the lighter class.
The value for CO is
                                 /CO
37.1g/hp-hr
14.4g/hp-hr
= 2.58
Equation 26
and the corresponding value for HC is 1.73.
                                 /HC
 1.9g/hp-hr
 l.lg/hp-hr
= 1.73
Equation 27
We applied this ratio in all three model-year groups, as shown in Table 32.
Note that in Draft MOVES2009, the ratios in Equation 26 and Equation 27 were erroneously
applied to the 2005-2007 model-year groups for LHD45 and MHD vehicles. In MOVES2010,
values for these model-year groups were set equal to those for the LHD2b3 vehicles, with the
rationale that the standards converge for both groups.
For NOx, all values are equal to those for LHD2b3, because the same standards apply to both
classes throughout. The approaches for all three regulatory classes in all three model years are
shown in Table 32.
Table 32. Methods used to Calculate and Start Emission Rates for Heavy-Duty Spark-Ignition Engines
Regulatory Class

LHD2b3
LHD45, MHD
Model-year Group

1960-2004
2005-2007
2008 +
1960-2004
2005-2007
2008 +
Method
CO
Values from
Table 31
Values from
Table 31
Values from
Table 31
Increase in proportion
To standards
Increase in proportion
To standards
Increase in proportion
To standards
THC
Values from
Table 31
Reduce in
proportion
To standards
Reduce in
proportion
To standards
Increase
in proportion
To standards
Increase in proportion
To standards
Increase in proportion
To standards
NOx
Values from
Table 31
Reduce in proportion
To standards
Reduce in proportion
To standards
Same values as
LHD2b3
Same values as
LHD2b3
Same values as
LHD2b3
                                                                                       66

-------
As for heavy-duty diesel and light-duty vehicles we applied the curve in Figure 19 to adjust the
start emission rates for varying soak times.  The rates described in this section were for cold  starts
(soak time > 720 minutes).

2.2.4.3 Paniculate Matter
Data on PM start emissions from heavy-duty gasoline vehicles were unavailable.  As a result, we
used the multiplication factor from the running exhaust emissions analysis of 1.40 to scale up start
emission rates for light-duty trucks.
                                                                                          67

-------
A. Appendices



A.1 Calculation of Accessory Power Requirements





                      Table 33. Accessory load estimates for HHD trucks
VSP
Low
Power (kw)
% time on
Total (kW)
Mid
Power (kw)
% time on
Total (kW)
High
Power (kw)
% time on
Total (kW)
Cooling Fan

19.0
10%
1.9

19.0
20%
3.8

19.0
30%
5.7
Air cond

2.3
50%
1.2

2.3
50%
1.2

2.3
50%
1.2
Air comp
Off = 0.5 kW
3.0
60%
2.0
Off = 0.5 kW
2.3
20%
0.9
Off = 0.5 kW
2.3
10%
0.7
Alternator

1.5
100%
1.5

1.5
100%
1.5

1.5
100%
1.5
Engine
Accessories

1.5
100%
1.5

1.5
100%
1.5

1.5
100%
1.5
Total Accessory Load (kW)



8.1



8.8



10.5
                      Table 34. Accessory load estimates for MHD trucks
VSP
Low
Power (kw)
% time on
Total (kW)
Mid
Power (kw)
% time on
Total (kW)
High
Power (kw)
% time on
Total (kW)
Cooling Fan

10.0
10%
1.0

10.0
20%
2.0

10.0
30%
3.0
Air cond

2.3
50%
1.2

2.3
50%
1.2

2.3
50%
1.2
Air comp
Off = 0.5 kW
2.0
60%
1.4
Off = 0.5 kW
2.0
20%
0.8
Off = 0.5 kW
2.0
10%
0.7
Alternator

1.5
100%
1.5

1.5
100%
1.5

1.5
100%
1.5
Engine
Accessories

1.5
100%
1.5

1.5
100%
1.5

1.5
100%
1.5
Total Accessory Load (kW)



6.6



7.0



7.8
                                                                             68

-------
Table 35. Accessory load estimates for buses
VSP
Low
Power (kw)
% time on
Total (kW)
Mid
Power (kw)
% time on
Total (kW)
High
Power (kw)
% time on
Total (kW)
Cooling Fan

19.0
10%
1.9

19.0
20%
3.8

19.0
30%
5.7
Air cond

18.0
80%
14.4

18.0
80%
14.4

18.0
80%
14.4
Aircomp
Off = 0.5 kW
4.0
60%
2.6
Off = 0.5 kW
4.0
20%
1.2
Off = 0.5 kW
4.0
10%
0.9
Alternator

1.5
100%
1.5

1.5
100%
1.5

1.5
100%
1.5
Engine
Accessories

1.5
100%
1.5

1.5
100%
1.5

1.5
100%
1.5
Total Accessory Load (kW)



21.9



22.4


24.0
                                                                    69

-------
A.2 Tampering and Mai-maintenance
Tampering and mal-maintenance (T&M) effects represent the fleet-wide average increase in
emissions over the useful life of the engines. In laboratory testing, properly maintained engines
often yield very small rates of emissions deterioration through time. However, we assume that in
real-world use, tampering and mal-maintenance yield higher rates of emissions deterioration over
time. As a result, we feel it is important to model the amount of deterioration we expect from this
tampering and mal-maintenance. We estimated these fleet-wide emissions effects by multiplying
the frequencies of engine component failures by the emissions impacts related to those failures for
each pollutant. Details of this analysis appear later in this section.

A.2.1 Modeling Tampering and Mal-maintenance
As T&M affects emissions through age, we developed a simple function of emission deterioration
with age.  We applied the zero-age rates through the emissions warranty period (5 years/100,000
miles), then increased the rates linearly up to the useful life. Then we assumed that all the rates
level off beyond the useful life age. Figure 27  shows this relationship.
                 Figure 27. Qualitative Depiction of the implementation of age effects.
        Emission rate
     Final emission rate"
           Zero-milt
           emission
                   End of warranty
                                                       Age
The useful life refers to the length of time that engines are required to meet emissions standards.
We incorporated this age relationship by averaging emissions rates across the ages in each age
group. Mileage was converted to age with VIUS40 (Vehicle Inventory and Use Survey) data, which
contains data on how quickly trucks of different regulatory classes accumulate mileage. Table 36
shows the emissions warranty period and approximate useful life requirement period for each of the
regulatory classes.
                                                                                        70

-------
                  Table 36. Warranty and useful life requirements by regulatory class
Regulatory class
HHD
MHD
LHD45
LHD2b3
BUS
Warranty age
(Requirement:
100,000 miles or 5 years)
1
2
4
4
2
Useful life
mileage/age
requirement
435,000/10
185,000/10
110,000/10
110,000/10
435,000/10
Useful
life age
4
5
4
4
10
While both age mileage metrics are given for these periods, whichever comes first determines the
applicability of the warranty.  As a result, since MOVES deals with age and not mileage, we need
to convert all the mileage values to age equivalents, as the mileage limit is usually reached before
the age limit. The data show that on average, heavy-heavy-duty trucks accumulate mileage much
more quickly than other regulatory classes. Therefore, any deterioration in heavy-heavy-duty truck
emissions will presumably happen at at younger ages than for other regulatory classes. Buses, on
average, do not accumulate mileage quickly.  Therefore, their useful life period is governed by the
age requirement, not the mileage requirement.
Since MOVES  deals with age groups and not individual ages, the increase in emissions by age
must be calculated by age group. We assumed that there is an even age distribution within each
age group (e.g.  ages 0,  1,2, and 3  are equally represented in the 0-3 age group). This is important
since, for example, HHD trucks reach useful life at four years, which means they will increase
emissions through the 0-3 age group. As a result, the 0-3 age group emission rate will be higher
than the zero-mile emission rate for HHD trucks. Table 37 shows the multiplicative T&M
adjustment factor by age. We determined this factor using the mileage-age data from Table 36 and
the emissions-age relationship that we described in Figure 27. We multiplied this factor by the
emissions increase of each pollutant over the useful life of the engine, which we determined from
the analysis in the section A2.3     Analysis below and which is listed in the corresponding
running exhaust sections above.
                                                                                         71

-------
              Table 37 shows the T&M multiplicative adjustment factor by age (frM.age group)-
Age Group
0-3
4-5
6-7
8-9
10-14
15-19
20+
LHD
0
1
1
1
1
1
1
MHD
0.083
0.833
1
1
1
1
1
HHD
0.25
1
1
1
1
1
1
Bus
0.03125
0.3125
0.5625
0.8125
1
1
1
In this table,  a value of 0 indicates no deterioration, or zero-mile emissions level (ZML), and a
value of 1 indicates a fully deteriorated engine, or maximum emissions level, at or beyond useful
life (UL).  The calculation of emission rate by age group is described in the equation below. TMpoi
represents the estimated emissions rate increase through the useful life for a given pollutant.
                           pol.agegrp   pol,ZML
U "•" J
                                                          ol )
Equation 28
A.2.2 Data Sources
EPA used the following information to develop the tamper and mal-maintenance occurrence rates
used to develop emission rates used in MOVES:

       •   California's  ARE EMFAC2007 Modeling Change Technical  Memo41 (2006).  The
          basic EMFAC occurrence rates for tampering and mal-maintenance were developed
          from the Radian and EFEE reports and internal CARB engineering judgment.

       •   Radian Study (1988).  The report estimated the malfunction rates based on survey and
          observation.  T he data may  be questionable for current heavy-duty  trucks due  to
          advancements such as electronic controls, injection systems, and exhaust aftertreatment.

       •   EFEE report (1998) on PM emission deterioration rates for in-use vehicles. Their work
          included heavy-duty diesel vehicle chassis dynamometer testing at Southwest Research
          Institute.

       •   EMFAC2000 (2000) Tampering and Mal-maintenance Rates

       •   EMA's  comments on A RB's  Tampering,  Malfunction,   and  Mal-maintenance
          Assumptions for EMFAC 2007

       •   University of California -Riverside (UCR) "Incidence of Malfunctions and Tampering
          in Heavy-Duty Vehicles"
                                                                                      72

-------
       •  Air Improvement Resources, Inc.'s Comments on Heavy-Duty Tampering and Mai-
          maintenance Symposium

       •  EPA internal engineering judgment

A.2.3 Analysis

A.2.3.1      T&MCategories
EPA generally adopted the categories developed by CARB, with a few exceptions. The high fuel
pressure category was removed. We added a category for misfueling to represent the use of
nonroad diesel, not ULSD onroad diesel. We combined the injector categories into a single group.
We reorganized the EGR categories into "Stuck Open" and "Disabled/Low Flow." We included
the PM regeneration system, including the igniter, injector, and combustion air system in the PM
filter leak category.
EPA will group the LHDD, MHDD, HHDD, and Diesel bus groups together, except for 2010 and
beyond. We assumed that the LHDD group will primarily use Lean NOx Traps (LNT) for the NOx
control in 2010 and beyond. On the other hand, we also assumed that Selective Catalyst Reduction
(SCR)  systems will  be the primary NOx  aftertreatment system for  HHDD.  T herefore, the
occurrence rates and emission impacts will vary in 2010 and beyond depending on the regulatory
class of the vehicles.
A. 2.3.2       T&MModel Year Groups
   EPA developed the model year groups based on regulation and technology changes.

       •  Pre-1994 represents non-electronic fuel control.

       •  1998-2002 represents the time period with consent decree issues.

       •  2003 represents early use of EGR.
       •  2007 and 2010 contain significant PM and NOx regulation changes.

       •  EPA issued a rule to require OBD for heavy duty trucks, beginning in MY 2010 with
          complete phase-in by MY 2013.


A. 2.3.3       T &M Occurrence Rates

A. 2.3.3.1     EPA T &M Occurrence Rate Differences from EMFAC200 7
EPA adopted the CARB EMFAC2007 occurrence rates, except as noted below.
Clogged Air Filter: EPA reduced the frequency rate from EMFAC's 15% to 8%.  EPA reduced
this value based on the UCR results, the Radian study, and EMA's comments that air filters are a
maintenance item. Many trucks contain indicators to notify the driver of dirty air filters and the
drivers have incentive to replace the filters for other performance reasons.
                                                                                      73

-------
Other Air Problems:  EPA reduced the frequency rate from EMFAC's 8% to 6% based on the
UCR results.
Electronics Failed:  EPA will continue to use the 3% frequency rate for all model years beyond
2010. CARB increased the rate to 30% in 2010 due to system complexity. EPA does not agree
with CARB's assertion that the complexity of electronic systems will increase enough to justify a
ten-fold increase in malfunction occurrence rates. We believe that the hardware will evolve
through 2010, rather than be replaced with completely new systems that would justify a higher rate
of failure. EPA asserts that many of the 2010 changes will occur with the aftertreatment systems
which are accounted for separately.
EGR Stuck Open: EPA believes the failure frequency of this item is rare and therefore set the
level at 0.2%. This failure will lead to drivability issues that will be noticeable to the driver and
serve as an incentive to repair.
EGR Disabled/Low Flow:  EPA believes the EMFAC 20% EGR failure rate is too high and
reduced the rate to 10%. All but one major engine manufacturer had EGR previous to the 2007
model year and all have it after 2007. Therefore, EMFAC's frequency rate increase in 2010  due to
the increase truck population using EGR does not seem valid.  However, the Illinois EPA stated
that "EGR flow insufficient" is the top OBD issue found in their LDV I/M program42 so it cannot
be ignored.
NOX Aftertreatment malfunction:  EPA developed a NOx aftertreatment malfunction rate that is
dependent on the type of system used. We assumed that FIHDD will use primarily SCR systems
and LFIDD will primarily use LNT systems.  We estimated the failure rates of the various
components within each system to develop a  composite malfunction rate.
The individual failure rates were developed considering the experience in agriculture and stationary
industries of NOx aftertreatment systems and similar component applications. Details are included
in the chart below. We assumed that tank heaters had a 5% failure rate, but were only required in
one third of the country and  one fifth of the year. The injector failure rate is lower than fuel
injectors, even though they have similar technology, because there is only one required in each
system and it is operating in less severe environment of pressure and temperature.  We believe the
compressed air delivery system is very mature based on a similar use in air brakes. We also
believe that manufacturers will initiate engine power de-rate as incentive to keep the urea supply
sufficient.
                                                                                       74

-------

Occurrence Rate
SCR
Urea tank
Tank heaters
In-exhaust injectors
Compressed
air delivery to injector
Urea supply pump
Control system
Exhaust temperature sensor
Urea supply
0.5%
1%
2%
1%
1%
5%
1%
1%
                                             Overall
13%
LNT
Adsorber
In-exhaust injectors
Control system
Exhaust temperature sensor
7%
2%
5%
1%
                                             Overall
16%
NOx aftertreatment sensor: EPA believes the 53% occurrence rate in EMFAC2007 is too high
and will use 10%. CARB assumed a mix of SCR, which uses one sensor per vehicle, and NOx
adsorbers, which use two sensors per vehicle. They justified the failure rate based on the increased
number of sensors in the field beginning in 2010.
   We developed the occurrence rate based on the following assumptions:

   •   Population: HHDD: vast majority of heavy-duty applications will use SCR technology with
       a maximum of one NOx sensor. NOx sensors are not required for SCR - manufacturers can
       use  models or run  open loop.    Several engine manufacturers  representing  30%  of the
       market plan to delay the use  of NOx aftertreatment devices through the use of improved
       engine-out emissions and emission credits.

   •   Durability expectations:  SwRI completed 6000 hours of ESC cycling with NOx sensor.
       Internal testing supports longer life durability.   Discussions with OEMs in 2007 indicate
       longer life expected by 2010.

   •   Forward looking assumptions:  M anufacturers have a strong incentive to improve the
       reliability and durability of the sensors because of the high cost associated with  frequent
       replacements.
PM Filter Leak:  EPA will use 5% PM filter leak and system failure rate.  CARB used 14% failure
rate. They discounted high failure rates currently seen in the field.
PM Filter Disable:  EPA agrees with CARB's 2% tamper rate of the PM filter. The filter causes a
fuel economy penalty so the drivers have an incentive to remove it.
Oxidation Catalyst Malfunction/Remove: EPA believes most manufacturers will install
oxidation catalysts initially  in the 2007 model year and agrees with CARB's assessment of 5%
failure rate.  This rate consists of an approximate 2% tampering rate and 3% malfunction rate. The
catalysts are more robust than PM filters, but have the potential to experience degradation when
exposed to high temperatures.
                                                                                      75

-------
Misfuel: EPA estimated that operators will use the wrong type of fuel, such as agricultural diesel
fuel with higher sulfur levels, approximately 0.1% of the time.
A. 2.3.3.2      Tampering & Mai-maintenance Occurrence Rate Summary
 Tamper & Malmaintenance
 Frequency of Occurrence: Average rate over life of vehicle

Timing Advanced
Timing Retarded
Injector Problem (all)
Puff Limiter Mis-set
Puff Limiter Disabled
Max Fuel High
Clogged Air Filter - EPA
Wrong/Worn Turbo
Intercooler Clogged
Other Air Problem - EPA
Engine Mechanical Failure
Excessive Oil Consumption
Electronics Failed - EPA
Electronics Tampered
EGR Stuck Open
EGR Disabled/Low Flow - EPA
Nox Aftertreatment Sensor
Replacement Nox Aftertreatment Sensor
Nox Aftertreatment Malfunction - EPA
PM Filter Leak
PM Filter Disabled
Oxidation Catalyst Malfunction/Remove - EPA
Mis-fuel - EPA
Frequency Rates
1994-97
5%
3%
28%
4%
4%
3%
8%
5%
5%
6%
2%
5%
3%
10%
0%
0%
0%
0%
0%
0%
0%
0%
0.1%
1998-2002
2%
2%
28%
0%
0%
0%
8%
5%
5%
6%
2%
3%
3%
15%
0%
0%
0%
0%
0%
0%
0%
0%
0.1%
2003-2006
2%
2%
13%
0%
0%
0%
8%
5%
5%
6%
2%
3%
3%
5%
0.2%
10%
0%
0%
0%
0%
0%
0%
0.1%
2007-2009
2%
2%
13%
0%
0%
0%
8%
5%
5%
6%
2%
3%
3%
5%
0.2%
10%
0%
0%
0%
5%
2%
5%
0.1%
2010+HHDT
2%
2%
13%
0%
0%
0%
8%
5%
5%
6%
2%
3%
3%
5%
0.2%
10%
10%
1%
13%
5%
2%
5%
0.1%
2010+ LHDT
2%
2%
13%
0%
0%
0%
8%
5%
5%
6%
2%
3%
3%
5%
0.2%
10%
10%
1%
16%
5%
2%
5%
0.1%
A. 2.3.3.2     Emission Effects
   NOx Emission Effects
EPA developed the emission effect from each tampering and mal-maintenance incident from
CARB's EMFAC, Radian's dynamometer testing with and without the malfunction present, EFEE
results, and internal testing experience.
EPA estimated that the lean NOx traps (LNT) in LHDD are 80% efficient and the selective catalyst
reduction (SCR) systems in HHDD are 90% efficient at reducing NOx.
EPA developed the NOx emission factors of the NOx  sensors based on SCR systems' ability to run
in open-loop mode and still achieve NOx reductions.  The Manufacturers of Emission Controls
Association (MECA) has stated that 75-90% NOX reduction with open loop control and >95%
reduction with closed loop control.43  Visteon reports 60-80% NOX reduction with open loop
control.
44
The failure of the NOx aftertreatment system had a different impact on the NOx emissions
depending on the type of aftertreatment.  The HHDD vehicles with SCR systems would experience
a 1000% increase in NOx during a complete failure, therefore we estimated a 500% increase as a
midpoint between normal operation and a complete failure.  The LHDD vehicles with LNT
systems would experience a 500% increase in NOx during a complete failure. We estimated a
300% increase as a value between a complete failure and normal system operation.
                                                                                       76

-------
The values with 0% effect in shaded cells represent areas which have no occurrence rate.
 Tamper & Malmaintenance
 NOX Emission Effect

Federal Emission Standard

Timing Advanced
Timing Retarded
Injector Problem (all)
Puff Limiter Mis-set
Puff Limiter Disabled
Max Fuel High
Clogged Air Filter
Wrong/Worn Turbo
Intercooler Clogged
Other Air Problem
Engine Mechanical Failure
Excessive Oil Consumption
Electronics Failed
Electronics Tampered
EGR Stuck Open
EGR Disabled / Low Flow
Nox Aftertreatment Sensor
Replacement Nox Aftertreatment Sensor
Nox Aftertreatment Malfunction
PM Filter Leak
PM Filter Disabled
Oxidation Catalyst Malfunction/Remove
Mis-fuel
1 994-97
5.0

60%
-20%
-5%
0%
0%
10%
0%
0%
25%
0%
-10%
0%
0%
80%
0%
0%
0%
0%
0%
0%
0%
0%

1 998-2002
5.0

60%
-20%
-1%
0%
0%
0%
0%
0%
25%
0%
-10%
0%
0%
80%
0%
0%
0%
0%
0%
0%
0%
0%

2003-2006
4.0

60%
-20%
-1%
0%
0%
0%
0%
0%
25%
0%
-10%
0%
0%
80%
-20%
30%
0%
0%
0%
0%
0%
0%

2007-2009
2.0

60%
-20%
-1%
0%
0%
0%
0%
0%
25%
0%
-10%
0%
0%
80%
-20%
50%
0%
0%
0%
0%
0%
0%

2010+ HHDT
0.2

6%
-20%
-1%
0%
0%
0%
0%
0%
3%
0%
-10%
0%
0%
8%
-20%
5%
200%
200%
500%
0%
0%
0%

2010 LHDT
0.2

12%
-20%
-1%
0%
0%
0%
0%
0%
5%
0%
-10%
0%
0%
16%
-20%
10%
200%
200%
300%
0%
0%
0%

   PM Emission Effects
EPA developed the PM emission effects from each tampering and mal-maintenance incident from
CARB's EMFAC, Radian's dynamometer testing with and without the malfunction present, EFEE
results, and internal testing experience.
EPA estimates that the PM filter has 95% effectiveness.  Many of the tampering and mal-
maintenance items that impact PM also have a fuel efficiency and drivability impact.  Therefore,
operators will have an incentive to fix these issues.
EPA estimated that excessive oil consumption will have the same level of impact on PM as engine
mechanical failure. The failure of the oxidation catalyst is expected to cause a PM increase of
30%; however, this value is reduced by 95% due to the PM filter effectiveness. We also
considered a DOC failure will cause a secondary failure of PM filter regeneration.  We accounted
for this PM increase within the PM filter disabled and leak categories.
The values with 0% effect in shaded cells represent areas which have no occurrence rate.
                                                                                        77

-------
    Tamper & Malmaintenance
    PM Emission Effect

Federal Emission Standard

Timing Advanced
Timing Retarded
Injector Problem
Puff Limiter Mis-set
Puff Limiter Disabled
Max Fuel High
Clogged Air Filter
Wrong/Worn Turbo
Intercooler Clogged
Other Air Problem
Engine Mechanical Failure
Excessive Oil Consumption
Electronics Failed
Electronics Tampered
EGR Stuck Open/Low Flow
EGR Disabled
Nox Aftertreatment Sensor
Replacement Nox Aftertreatment Sensor
Nox Aftertreatment Malfunction
PM Filter Leak
PM Filter Disabled
Oxidation Catalyst Malfunction/Remove
Mis-Fuel
1994-97
0.1

-10%
25%
100%
20%
50%
20%
50%
50%
50%
40%
500%
500%
60%
50%
0%
0%
0%
0%
0%
0%
0%
0%
30%
1998-2002
0.1

-10%
25%
100%
0%
0%
0%
50%
50%
50%
40%
500%
500%
60%
50%
0%
0%
0%
0%
0%
0%
0%
0%
30%
2003-2006
0.1

-10%
25%
100%
0%
0%
0%
30%
50%
30%
30%
500%
500%
60%
50%
100%
-30%
0%
0%
0%
0%
0%
0%
30%
2007-2009
0.01

0%
1%
5%
0%
0%
0%
2%
3%
2%
2%
25%
25%
3%
3%
5%
-30%
0%
0%
0%
600%
1000%
2%
100%
2010
0.01

0%
1%
5%
0%
0%
0%
2%
3%
2%
2%
25%
25%
3%
3%
5%
-30%
0%
0%
0%
600%
1000%
2%
100%
   HC Emission Effects
EPA estimated oxidation catalysts are 80% effective at reducing hydrocarbons.  All manufacturers
will utilize oxidation catalysts in 2007, but only a negligible number were installed prior to the PM
regulation reduction in 2007.
We reduced CARB's HC emission effect for timing advanced because earlier timing should reduce
HC, not increase them. The effect of injector problems was reduced to 1000% based on internal
experience. We increased the HC emission effect of high fuel pressure to 10% because the higher
pressure will lead to extra fuel in early model years and therefore increased HC. Lastly, we used
the HC emission effect of advanced timing for the electronics tampering since this was the most
significant type of tampering that occurred.
The values with 0% effect in shaded cells represent areas which have no occurrence rate.
                                                                                        78

-------
  Tamper & Malmaintenance
  HC Emission Effect

Federal Emission Standard

Timing Advanced
Timing Retarded
Injector Problem (all)
Puff Limiter Mis-set
Puff Limiter Disabled
Max Fuel High
Clogged Air Filter
Wrong/Worn Turbo
Intercooler Clogged
Other Air Problem
Engine Mechanical Failure
Excessive Oil Consumption
Electronics Failed
Electronics Tampered
EGR Stuck Open
EGR Disabled / Low Flow
Nox Aftertreatment Sensor
Replacement Nox Aftertreatment Sensor
Nox Aftertreatment Malfunction
PM Filter Leak
PM Filter Disabled
Oxidation Catalyst Malfunction/Remove
Mis-fuel
1994-97
1.3

0%
50%
1000%
0%
0%
10%
0%
0%
0%
0%
500%
300%
50%
0%
0%
0%
0%
0%
0%
0%
0%
0%

1998-2002
1.3

0%
50%
1000%
0%
0%
0%
0%
0%
0%
0%
500%
300%
50%
0%
0%
0%
0%
0%
0%
0%
0%
0%

2003-2006
1.3

0%
50%
1000%
0%
0%
0%
0%
0%
0%
0%
500%
300%
50%
0%
1 00%
0%
0%
0%
0%
0%
0%
0%

2007-2009
0.2

0%
50%
1000%
0%
0%
0%
0%
0%
0%
0%
500%
300%
50%
0%
100%
0%
0%
0%
0%
0%
0%
50%

2010+ HHDT
0.14

0%
10%
200%
0%
0%
0%
0%
0%
0%
0%
100%
60%
10%
0%
20%
0%
0%
0%
0%
0%
0%
50%

2010 LHDT
0.14

0%
10%
200%
0%
0%
0%
0%
0%
0%
0%
100%
60%
10%
0%
20%
0%
0%
0%
0%
0%
0%
50%

A separate tampering analysis was not performed for CO; rather, the HC effects were assumed to
apply for CO.
Combining all of the emissions effects and failure frequencies discussed in this section, we
summarized the aggregate emissions impacts over the useful life of the fleet due to in the main
body of the document in Table 11 (NOx), Table 17 (PM), and Table 21 (HC and CO).
       HD OBD impacts
With the fmalization of the heavy-duty onboard diagnostics (HD OBD) rule, we made adjustments
to our draft 2010 and later model year to reflect the rule's implementation.
Specifically, we reduced our emissions increases for all pollutants due to tampering and mal-
maintenance by 33%. As data are not yet available for heavy-duty trucks equipped with OBD, this
number is probably a conservative estimate. Still, PM and NOX reductions from 2010 and later
model year vehicles will be substantial compared to prior model years regardless of the additional
incremental benefit from OBD. We assumed, since the rule phases in OBD implementation, that
33% of all engines will have OBD in  2010, 2011, and 2012 model years, and 100% will have OBD
by 2013 model year and later.  Equation 29 describes the calculation of TMpoi, the increase in
emission rate through useful life, where/OBD represents  the fraction of the fleet equipped with OBD
(0% for model years 2009 and earlier, 33% for model years 2010-2012, and 100% for model years
2013 and later). The result from this equation can be plugged into Equation 28 to determine the
emission rate for any age group.
                         = ™
                              pol,nonOBD '
                                                         l,nonOBDJ OBD
Jo.
Equation 29
                                                                                      79

-------
As data for current and future model years become available, we may consider refining these
estimates and methodology.
                                                                                      80

-------
A.3 Extended Idle Data Summary
 Idle HC Rates (gram/hour) Summary
Program [Condition
# Samples
Mean HC Emiss Rate
1991-2006 Low Speed Idle, A/C Off - HOT
McCormick, High Altitude, HOT
WVU- 1991 -2004
Storey

Low Idle, AC Off
Low Idle, AC Off
Low Idle, AC Off
12
48
4
10.2
9.5
28
Overall|| 64 || 10.8
1991-2006 High Speed Idle, A/C On - HOT
Broderick DC Davis
Storey

High Idle, AC On
High Idle, AC On
1
4
86
48
Overall|| 5 || 55.6
1975-1990 MY Low Speed Idle, A/C Off - HOT
Program
WVU -1975-1 990

Condition
Low Idle, AC Off
Samples
18
Mean
21
Overall!! 18 II 21.0
1991-2006 MY Low Speed Idle, A/C Off - Bus
Program
McCormick, High Altitude, Bus

Condition
Low Idle, AC Off
Samples
12
Mean
8.2
Overall)) H II O
                                                                       81

-------
Idle CO Rates (gram/hour) Summary
Program [Condition
# Samples
Mean CO Emiss Rate
1991-2006 Low Speed Idle, A/C Off - HOT
McCormick, High Altitude, HOT
Calcagno
WVU- 1991 -2004
Storey

Low Idle, AC Off
Low Idle, AC Off
Low Idle, AC Off
Low Idle, AC Off
12
27
48
4
71
37
23
25
Overall|| 91 II 33.6
1991-2006 High Speed Idle, A/C On - HOT
Calcagno
Broderick DC Davis
Storey

High Idle, AC On
High Idle, AC On
High Idle, AC On
21
1
4
99
190
73
Overall)) 0 II 91-2
1975-1990 MY Low Speed Idle, A/C Off - HOT
Program
WVU -1975-1 990

Condition
Low Idle, AC Off
Samples
18
Mean
31
Overall)) 18 || 31.0
1991-2006 MY Low Speed Idle, A/C Off - Bus
Program
McCormick, High Altitude, Bus

Condition
Low Idle, AC Off
Samples
12
Mean
79.6
Overall)) 12 || 79.6
                                                                                  82

-------
Idle PM Rates (gram/hour) Summary
Program [Condition
# Samples
Mean PM Emiss Rate
1991-2006 Low Speed Idle, A/C Off - HOT
McCormick, High Altitude, HOT
Calcagno
WVU- 1991 -2004
Storey

Low Idle, AC Off
Low Idle, AC Off
Low Idle, AC Off
Low Idle, AC Off
12
27
48
4
1.8
2.55
1.4
1.3
Overall|| 91 II 1-8
1991-2006 High Speed Idle, A/C On - HOT
Calcagno
Storey

High Idle, AC On
High Idle, AC On
21
4
4.11
3.2
Overall|| H II 4-ฐ
1975-1990 MY Low Speed Idle, A/C Off - HOT
Program
WVU -1975-1 990

Condition
Low Idle, AC Off
Samples
18
Mean
3.8
Overall)) 18 || 3.8
1991-2006 MY Low Speed Idle, A/C Off - Bus
Program
McCormick, High Altitude, Bus

Condition
Low Idle, AC Off
Samples
12
Mean
2.88
Overall)) 12 || 2.9
                                                                                  83

-------
 Idle Nox Rates (gram/hour) Summary
Program [Condition
# Samples
Mean NOX Emiss Rate
1991-2006 Low Speed Idle, A/C Off
McCormick, High Altitude, HOT
Lim, EPA
Irick, Clean Air Tech & IdleAire
WVU- 1991 -2004
WVU, NCHRP
Tang, Metro NY, 1984-1999
Calcagno
Broderick DC Davis
Storey

Low RPM, AC Off
Low RPM, No access

Low RPM, AC Off


Low RPM, AC Off
Low RPM, AC Off
Low RPM, AC Off
Overall
12
12
49
48
2
33
27
1
4
188
85
109
87
83
47
81
120
104
126
94
1991-2006 High Speed Idle, A/C Off
Lim, EPACCD
Calcagno

High RPM, No access
High RPM, AC Off
Overall
5
21
26
169
164
165

Lim, EPACCD
Broderick DC Davis
Calcagno
Storey

High RPM, AC On
High RPM, AC On
High RPM, AC On
High RPM, AC On
Overall
5
1
21
4
31
212
240
223
262
227

Program
WVU -1975-1 990
Lim, EPACCD, 1985 MY

Condition
Low RPM, AC Off
Low RPM, AC Off
Overall
Samples
18
1
19
Mean
48
20
47
1991-2006 MY Low Speed Idle, A/C Off - Bus
Program
McCormick, High Altitude, Bus

Condition
Low Idle, AC Off
Overall
Samples
12
12
Mean
121
121.0
2007 Extended Idle Emissions calculation:
           •   Assumed 8 hour idle period where the emissions controls, such as EGR, oxidation catalyst,
              and NOx aftertreatment, are still active for the first hour.
           •   HC emissions standards:
                  o  Pre-2007: 0.50 g/bhp-hr
                  o  2007: 0.14 g/bhp-hr
           •   NOx emissions standards:
                  o  Pre-2010: 5.0 g/bhp-hr
                                                                                           84

-------
                   o  2010:  0.2g/bhp-hr

Idle HC Rate Reduction = 1 - [(1/8 * 0.14 g/bhp-hr + 7/8 * 0.5 g/bhp-hr) / 0.5 g/bhp-hr] = 9%
Idle NOx Rate Reduction = 1 - [(1/8 * 0.2 g/bhp-hr + 7/8 * 5.0 g/bhp-hr) / 5.0 g/bhp-hr] = 12%
                                                                                                 85

-------
A.4 Developing PM emission rates for missing operating modes
In cases where an estimated rate could not be directly calculated from data, we imputed the missing
value using a log-linear least-squares regression procedure.  Regulatory class, model year group
and speed class (0-25 mph, 25-50 mph and 50+ mph ) were represented by dummy variables in the
regression. The natural logarithm of emissions was regressed versus scaled tractive power (STP) to
represent the operating mode bins.  The regression assumed a constant slope versus STP for each
regulatory class.  Logarithmic transformation factors (mean square error of the regression squared /
2) were used to transform the regression results from a log based form to a linear form. Due to the
huge number of individual  second-by-second data points, all of the regression relationships were
statistically significant at a high level (99% confident level).  The table below shows the regression
statistics, and the equation  shows the form of the resulting regression equation.
Regression Coefficients for PM Emission Factor Model
Model-year
group
1960-87
1988-90
1991-93
1994-97
1998-2006

Speed Class (mph)
1-25
25-50
50+
1-25
25-50
50+
1-25
25-50
50+
1-25
25-50
50+
1-25
25-50
50+
STP

Type
Intercept (JS0)
Slope 050
Transformation
Coefficient
(0.5a2)
Medium
Heavy-Duty
-5.419
-4.942
-4.765
-5.366
-4.929
-4.785
-5.936
-5.504
-5.574
-5.927
-5.708
-5.933
-6.608
-6.369
-6.305
0.02821
0.5864
Heavy Heavy-
Duty
-5.143
-4.564
-4.678
-5.847
-5.287
-5.480
-5.494
-5.269
-5.133
-6.242
-5.923
-6.368
-6.067
-5.754
-6.154
0.0968
0.84035
 ln(PM) = ft + ft STP + 0.5<72

Where :

/?o = an intercept term for a speed class within a model year group, as shown in the table above,

/?i = a slope term for STP, and
  r\
G = the mean-square error or residual error for the model fit,

STP = the midpoint value for each operating mode (kW/metric ton?, see Table 9, page 11).
                                                                                       86

-------
A. 5  Heavy-duty Diesel EC/OC Fraction Calculation

A. 5.1 Introduction
This memo describes the development and application of a "rough cut" emission model for
estimating elemental and organic carbonaceous material (EC and OM) emission rates (or EC/OM
ratios) from MOVES.  The memo describes the following steps involved in predicting EC/OM
ratios. The memo also briefly describes comparisons with independent emission data collected
using the "Mobile Emission Laboratory," Operated by the University of California Riverside.
The subsequent sections of the memo describe the following topics:

      •  the extension of Physical Emission Rate Simulator (PERE) to estimate heavy-duty
          fleet-average emission factors for any specified driving cycle;
      •  the acquisition of data used in estimating EC/OC rates as a function of engine operating
          mode and the fitting of simple empirical models to them;

      •  the application of PERE to estimate EC and OC emission rates for different test cycles;
          and,

      •  the comparison of PERE-based EC and OC emission rates to those measured by
          independent researchers in HD trucks.

A. 5.2 PERE for Heavy-duty Vehicles (PERE-HD) and Its Extensions
The Physical Emission Rate Estimator (PERE) is a model  employed by EPA in early development
of MOVES.45 In particular, the MOVES team employed it in development of MOVES2004 to
impute greenhouse gas emission rates for combinations of SourceBin and Operating Mode for
which data was unavailable or of insufficient quality.
The underlying theory behind PERE and its comparison with measured fuel consumption data is
described by Nam and Giannelli (2005).45 Briefly, PERE estimates fuel consumption and emission
rates on  the basis of fundamental physical and mathematical relationships describing the road load
that a vehicle meets when driving a particular speed trace. Accessory loads are handled by addition
of an accessory power term. In the heavy-duty version of PERE (hereafter, "PERE-HD"),
accessory loads were described by a single value.
For the current project, PERE was modified to incorporate several  "extensions" that allowed it to
estimate fleet-average emission rates, simulate a variety of accessory load conditions, and predict
EC and OC rates for any given driving cycle.

A.5.2.1 PERE-HD Fleet-wide Average Emission Rate Estimator
PERE-HD requires  a number of user-specified inputs, including:

      •  vehicle-level descriptors (model year, running weight, track road-load coefficients
          (A,B,C), transmission type, class [MDT/HDT/bus]);

      •  engine parameters (fuel type,  displacement); and

      •  driving cycle (expressed through a speed trace).


                                                                                    87

-------
The specification of these inputs allows PERE to model the engine operation, fuel consumption,
and GHG emissions for a HDV on a specified driving cycle.
However, the baseline PERE-HD provides output for only one combination of these parameters at
once. To estimate fleet- wide average a large number of PERE-HD runs would be required.
Furthermore, the specification of only fleet-wide average coefficients is likely to substantially
underestimate variability in fuel consumption and emissions. Emissions data from a large number
of laboratory and field studies suggest that a very large fraction of total emissions from all vehicles
derives from a small fraction of the study fleet.  Therefore, it is desirable to develop an approach
that comes closer to spanning the range of likely combinations of inputs than using a small
selection of "average" or "typical" values.
For the current application, PERE-HD (built within Microsoft Excel) was expanded to allow for a
representative sample of [running weight] x [engine displacement] x [model year] combinations.
A third-party add-on package to Excel, @Risk 4.5 (Palisade Corporation, 2004), allows users to
supplement deterministic inputs within spreadsheet models with selected continuous probability
distributions, sample input values from each input distribution,  and re-run the spreadsheet model
with sets of selected inputs over a specified number of iterations. This type of procedure is
commonly referred to as "Monte Carlo" simulation.

A. 5. 2. 1. 1 Monte Carlo Simulation in PERE-HD
To  illustrate how @Risk performs this process, we illustrate the application of a simple model,
employing both deterministic calculations and stochastic Monte Carlo simulation:
This equation defines the body mass index for humans, a simple surrogate indicating overweight
and underweight conditions. According to the Centers for Disease Control and Prevention (CDC),
the average U.S. woman weighed 164.3 Ib (74.5 kg) in 2002 and was 5 '4" (1.6 m) tall. This result
corresponds to a BMI of 28, suggesting that the average U.S. woman is overweight. While this is
useful information from a public health perspective, it does not provide any indication as to which
individuals are likely to experience the adverse effects of being overweight and obese. However, if
we were to assume (arbitrarily) that the range of weight and height within the U.S. population was
+/-50% of the mean, distributed uniformly,  and perform a Monte Carlo simulation (5,000
iterations) using @Risk, we would predict a probability distribution of BMI in the population as
follows:
                                                                                        88

-------
     Qstribution of BM in Simulated Population
        0.060
             10
                           BM
In contrast, here is the BMI distribution in the entire U.S. population, according to the CDC's
National Health and Nutrition Examination Survey (NHANES):
    201-
    15 -
                              NHANES 1976-1980
                                          NHANES 2005-2006
                                BMI
 SOURCE: CDC/NCHS, National Health and Nutrition Examination Survey (NHANES).

These graphs illustrate how Monte Carlo simulation can be used to provide meaningful information
about the variability in a population.  Although the model example is very simple, it illustrates the
point that a model with "typical" inputs provides much less information than does Monte Carlo
simulation with variable inputs.
For emission modeling purposes using PERE-HD, several key inputs were modeled as probability
distributions.

A.5.2.1.2Model Year
Model year is an important factor in PERE, as the frictional losses in the model, expressed as
"friction mean effective pressure" (FMEP), vary by model year, improving with later model years.
As such, model year was simulated as a probability distribution, based on data from the Census
Bureau's 1997 Vehicle Inventory and Use  Survey (VIUS), which reports "vehicle miles traveled"
(VMT) by model  year. Accordingly these data were normalized to total VMT to develop a

                                                                                         89

-------
probability distribution. Model year distributions in 1997 were normalized to the current calendar
year (2008).l For instance, the fraction of 1996 vehicles reported in the 1997 VIUS is treated as
the fraction of 2002 vehicles in the 2003 calendar year.  Although a 2002 VIUS is available,
previous analyses (unpublished) have shown the "relative" model year distribution of trucks to
have changed little between 1997 and 2002, though this assumption is one limitation of this
analysis.
The model year distribution for PERE-HD was represented as a discrete probability distribution, as
shown below:
                   Probability and Cumulative Probability Distributions of Model Years in PERE-HD
     0.12 T
                                                                  -Probability by Age
                                                                  -Cumulative Probability by Age
         01  2 3 4 5 6 7 8  9 1011 12131415161718 19202122232425
                    Relative Age (Calendar Year - Model Year)
A. 5.2.1.3 Vehicle Weight and Engine Displacement
Vehicle running weights and engine displacements were modeled as a two-way probability
distribution with engine displacement depending on running weight. These data were derived from
VIUS microdata obtained from the Census Bureau.46 A two-way table was constructed to estimate
VMT classified by combinations of [weight class] * [displacement class].  Analyses were restricted
to diesel-powered trucks only.
As a first step, @Risk selects a running weight from a probability distribution representing the
fraction of truck VMT occurring at a given running weight:
1 VIUS reports model years 11 years old and greater as a single number. For the current analysis, the fraction of
vehicles within each model year older than 10 years of age through 25 years was estimated using an exponential decay
of the form p(x) = A *exp[-B*(x-10)]. Coefficients representing the A and B parameters were estimated by minimizing
least squares of the residuals. The sum of probabilities for model years older than 10 years was constrained the fraction
of VMT driven by trucks older than 10 years in VIUS.
                                                                                              90

-------
                       Probability Distribution of Vehicle Running Weight based on VIUS
     0.4 -i
    0.35
     0.3
  ฃ• 0.25 -
  |  0.2 -
  ฃ 0.15
     0.1 -
    0.05
          t-   CO
                                 O   O

                                 O   O
                                                       O    O
                                          O   CM    CO

                                           Weight Range
t-    (N
Because VIUS reports classes defined as ranges in running weight, any value of weight within each
VlUS-specified class was considered equally likely and modeled as a uniform probability
distribution within the class.  For the upper and lower bounds of the distribution the minimum and
maximum running weights were assumed to be 7,000 and 240,000 Ib, respectively.
After @Risk selects a running weight, it selects an engine displacement based on a discrete
distribution assigned to every weight class in VIUS, represented below:
                       Distribution of Displacement (cu. in.) by Running Weight (Ib) in TIUS
   2
  Q_
   0)
   E
   ^
  O
                                       Running Weight
           D850+
           • 800-849
           • 750-799
           • 700-749
           • 650-699
           D 600-649
           D 550-599
           • 500-549
           • 470-499
           D 350-469
           • 430-449
           D 400-429
           • 370-399
           D 350-369
           D 300-349
           • 250-299
           D 1-249
Again, because VIUS describes ranges of values for displacement, all values within each range
were given uniform weight and assigned a uniform distribution. For the extreme classes, the
minimum and maximum engine displacements were assumed to be 100 in3 and 915 in3,
respectively.
                                                                                              91

-------
This procedure reflects the range in running weights present among HDV in operation, and
constrains the combinations of weight and displacement to plausible pairs of values based on
surveyed truck operator responses.  These steps allow for plausible variability in weight-engine
pairings, which translates into differences in engine parameters influencing EC and OC emissions.
       For use in PERE-HD, all units were converted to SI units (kg and L).

A. 5.2.1.4 Accessory Load
The original PERE-HD treats accessory load as a fixed value, which may be varied by the user. It
is set at 0.75, and used in calculating fuel rate and total power demand at each second of driving.
Following the development of PERE-HD, a more detailed set of accessory load estimates was
developed based on several accessories' power demand while in use and the fraction of time each
accessory is in use (see Table 6).47  High, medium, and  low accessory use categories were
estimated for three vehicle classes:  HOT, MDT, and buses. For the current version of the model,
only the HDT accessory load estimates were employed, though a sensitivity analysis indicated that
mean EC/OM ratios were most sensitive to accessory load  during idle and creep driving cycles. In
the "base case," a mean ratio of 0.54 was predicted, while in the sensitivity  case, a mean ratio of
0.50 was predicted. This issue may be revisited at some point, although the limited sensitivity of
total results limits the importance of the accessory terms within the current exercise.
Within @Risk, the variable in PERE-HD, Pacc for accessory use was substituted with a variable
representing the distribution (in time) of accessory loads as estimated as the sum of a number of
discrete probability distributions.
Depending on the assumption of high, medium or low use, the power demand for these accessories
is distributed in time  as follows:
Comparison of high / IVedum / Low Accessory
                   Load Cases
1.000 -

0.800

0.600 -

0.400 -

0.200 -

0.000
                                       10
                                                                      oo
                                                                      SI
                                                                      CO
                                                                      SI
                                                 30
A. 5.2.1.5 Driving Cycle
For purposes of this exercise, the four phases of the California Air Resources Board's Heavy
Heavy-Duty Diesel Truck (HHDDT) chassis dynamometer testing cycle were used to reflect
variability in vehicle operations for PERE-HD.
                                                                                        92

-------
A. 5.2.1.6 Other Factors
Some elements of variability were not examined as part of this study. Hybrid-electric
transmissions and fuel cell power plants were excluded from the analysis, due to their low
prevalence within the current truck fleet.
One important source of variability that was not examined in this analysis is the variation in
resistive forces among vehicles with identical running weights.  This exclusion is important, given
the potential role for aerodynamic improvements, low rolling resistance tires, and other
technologies in saving fuel for long-distance trucking firms and drivers.  Such considerations could
be incorporated into PERE-HD in the future as a means of estimating the emission benefits of fuel-
saving technologies.
A. 5.2.2 Prediction of Elemental Carbon and Organic Mass based on PERE-HD

A. 5.2.2.1     Definition of Elemental and Organic Carbon and Organic Mass
In motor vehicle exhaust, the terms "EC," "elemental carbon," and "black carbon" refer to the
fraction of total carbonaceous mass within a particle sample that consists of light-absorbing carbon.
Alternatively, they refer to the portion of carbonaceous mass that has a graphitic crystalline
structure. Further, one can define EC as the portion of carbonaceous mass that has been altered by
pyrolysis, that is, the chemical transformation that occurs in high temperature in the absence of
oxygen.
EC forms in diesel engines as a result of the stratified combustion process within a cylinder. Fuel
injectors spray aerosolized fuel into the cylinder during the compression stroke.  The high-pressure
and high temperature during the cylinder cause spontaneous ignition of the fuel vaporizing from the
injected droplets. Because temperature can rise more quickly than oxygen can diffuse to the fuel at
the center of each droplets, pyrolysis can occur as hydrogen and other atoms are removed from the
carbonaceous fuel, resulting in extensive C-C bond interlinking. As a result, pyrolyzed carbon is
produced in a crystalline form similar to graphite.
"Organic carbon" or "organic mass" (OC or OM) is used to denote the portion of carbonaceous
material in exhaust that is not graphitic.  Chemical analysis of this non-graphitic carbon mass
indicates that it is composed of an extensive mixture of different organic molecules, including CIS
to C44 alkanes, polycyclic aromatic hydrocarbons, lubricating oil constituents (hopanes, steranes,
and carpanes), and a sizeable fraction of uncharacterized material.  This component of exhaust can
derive from numerous processes inside the engine involving both fuel and oil. Because of the
complex chemical mixture that comprises this mass, its measurement is highly dependent on
sampling conditions. The wide range of organics that compose it undergo evaporation and
condensation at different temperatures, and the phase-partitioning behavior of each molecule is
dependent on other factors, such as the sorption of vapor-phase organics to available surface area in
a dilution tunnel or background aerosol.
                                                                                        93

-------
A. 5.2.2.2      EPA Carbon Analysis Techniques in Ambient Air
The definitions of EC and OM are critical, as different groups use different techniques for
quantifying their concentrations within a given medium. For purposes of this document, it is
assumed that EC, OC, and OM are operationally defined quantities, meaning that they are defined
by the measurement technique used to quantify their concentrations on a filter or in air.
The different types of commonly used approaches for carbon include:

       •  Thermal/optical techniques, where the evaporation and oxidation of carbon are used in
          conjunction with a laser to measure optical  properties of a particle sample. The major
          methods used for this type of analysis include:
              o  Thermal/optical reflectance (TOR).  EPA is adopting this technique for the
                 PM2.5 speciation monitoring network nationwide.  It is also employed by the
                 IMPROVE program (Interagency Monitoring of Protected Visual Environments)
                 in national parks.  This technique heats a punch from a quartz fiber filter
                 according to a certain schedule. A Helium gas atmosphere is first employed
                 within the oven, and the evolved carbon is measured with a FID as temperatures
                 are increased in steps up to 580ฐC.  All carbon evolved in this way is assumed to
                 be volatilized organic material. Next, 2% oxygen gas is added to the
                 atmosphere, and temperatures are stepped up a number of times to a maximum
                 of 840ฐC. All carbon evolved after  the introduction of oxygen is assumed to be
                 elemental carbon. The reflection of light from  a laser by the filter is employed
                 to account for the pyrolysis of organic carbon that occurs during the warm-up
                 process.
              o  Thermal/optical transmission (TOT). The National Institute of Occupational
                 Safety and Health (NIOSH) uses this technique for measuring EC concentrations
                 in occupational environments. It is  based on similar principles to TOR, but
                 employs a different heating schedule and transmission of light as opposed to
                 reflectance.

       •  Radiation absorption techniques
              o  Aethalometerฎ - This instrument reports "black carbon" (BC) concentrations
                 based the extent of light absorption  by a "filter tape," that allows for a time
                 series of BC concentrations to be estimated.  It has a time resolution of several
                 minutes.
              o  Photoacoustic Spectrometer (PAS) - This instrument irradiates an air sample
                 with a laser. The resulting heat that occurs from the absorption of the laser light
                 by light-absorbing carbon in the air  sample produces a pressure wave that is
                 measured by the device. The signal from this pressure wave is proportional to
                 the light-absorbing carbon content in exhaust.

       •  Thermogravimetric techniques, where the "volatile organic fraction" (VOF) is separated
          by heat from the non-volatile refractory component of a particle sample.

       •  Chemical extraction, where solvents are used to separate the soluble and insoluble
          components of exhaust.


                                                                                        94

-------
A number of additional techniques are also described in the published literature, but the above
techniques have been most commonly applied in emissions and routine ambient PM measurement.
Among the available techniques, it has been a point of controversy among academics as to which
method provides the "correct" carbon signal. Rather than addressing these arguments in detail, this
analysis adopts the technique employed by the EPA ambient speciation monitoring network, TOR.
Needless to say, different researchers employ different sampling, measurement and analysis
techniques. Desert Research Institute (DRI) employed TOR in analyzing the Kansas City gasoline
PM emission study samples [cite?], while other prominent academics employ TOT, notably the
University of California Riverside College of Engineering Center for Environmental Research and
Technology (CE-CERT) and the University of Wisconsin-Madison (UWM) State Hygiene
Laboratory.  As research results from these groups is employed throughout this analysis, an inter-
comparison of the methods of TOT/TOR is necessary to "recalibrate" various datasets with respect
to each other.
EPA defines measurement techniques for dynamometer-based sampling and analysis of particulate
matter, in addition to techniques for sampling and analyzing particles in ambient air.  Inventories
estimated for EC and OM can be considered to reflect both broad categories of measurement
techniques, depending on context.
The user community for MOVES is predominantly concerned with emissions that occur into
ambient air.  EPA regulations for demonstration of attainment of state implementation plans (SIPs)
are based on monitored ambient particulate matter using Federal Reference Methods (FRM) for
ambient air.  FRM monitors for particle speciation in ambient air undergo analysis for EC and OC
according to a  defined standard operating procedure.48  That standard operating procedure defines
thermal/optical reflectance (TOR) as the desired method for analysis of ambient carbon PM.

A. 5.2.2.3      TOR - TOR Calibration Curve
In the course of the Gasoline/Diesel PM Split Study funded by the Department of Energy (DOE),
researchers from DRI analyzed filter samples using both TOR and TOT methods [cite]. These data
were obtained  and analyzed in the SPSS 9.0 statistical package.
Briefly, the DOE study included emissions characterizations of 57 light-duty gasoline vehicles
(LDGV)  and 34 HD  diesel vehicles (HDDV).  The vehicles were operated on a number of different
test cycles including cold-start and warm-start cycles.  The data set employed in this study was
generated by DRI and obtained from the DOE study web site.49 Both EC and OC were analyzed
using the same approach.  All data from all vehicles were compiled.
First, EC and OC measured by TOR (denoted EC-TOR and OC-TOR) were regressed on EC-TOT
and OC-TOT.  Studentized residuals from these regressions were noted, and those with Studentized
residuals >3 were excluded from further analysis.
Second, each test in the reduced  data set was assigned a random number (RAND) on the range
[0,1].  Those cases with RAND > 0.95 were set aside as a cross-validation data set, and excluded
from additional regression analyses.
Third, those cases with RAND < 0.95 were regressed again, this time using an inverse uncertainty
weighting procedure for each data point. When DRI analyzes a filter sample, it reports an
analytical uncertainty associated with the  primary estimate of EC and OC. Accordingly, the quality
of each datum  depends on the level of analytical uncertainty reported. The inverse of the DRI-

                                                                                      95

-------
reported uncertainty (I/a) associated with the TOR-based measurement was used to weight each
point in the weighted regression.

It should be noted that for each regression, the intercept term was set to zero. Models including
intercepts did not have intercept terms that reached statistical significance. As such, R2 values are
not considered valid.
       Coefficients from the weighted regression for EC and OC are reported below:
Slope
EC-TOR
OC-TOR
Beta
1.047
1.014
Std. Error
0.011
0.007
t-value
91.331
153.923
Sig.
<0.0001
<0.0001
To evaluate the quality of predictions resulting from these statistically-based adjustment factors,
they were used to predict EC-TOR and OC-TOR values for the subset of data with RAND > 0.95.
Scatter plots of the statistical fits are illustrated below (note logarithmic scaling).
    CO
    CD
    CD
    Cฃ
    O
     I
    O
    LU
1000




 200

 100




  20

  10




   2

   1
  :s
  .2
  .1
             EC-TOR Predicted
                                                                                         96

-------
    T3
    CD
    CD
    CD
    O
    o
    O
1000


 200-

 100-


  20
  10


   2
   1
  .3
  .2
  .1
            OC TOR-Predicted

       When measured values are regressed against predicted values, the following statistical
estimates of fit are obtained:
Prediction
EC
OC
Slope
1.080
1.092
Std. Error
0.009
0.069
Intercept
3.737
-4.417
Std. Error
3.173
16.188
As shown, the prediction vs. observed comparison yields a slope near unity for both EC-TOR and
OC-TOR, with nonsignificant intercepts.  On this basis, the "calibration" factors for converting EC-
TOT and OC-TOT into their respective TOR-based metrics appear reasonable.
It remains an unverified assumption that the "calibration" factors derived from the emissions data
derived from DRI as part of the DOE Gasoline / Diesel PM Split Study are general enough to apply
to EC-TOT measurements obtained by other research groups.

A. 5.2.2.4 EC and OC Emission Rates

Selection of Engine Parameters for Predictive Modeling
PERE-HD produces estimates of engine operating conditions and fuel consumption for a given
driving cycle. Prediction of EC and OM emissions requires information on the composition of
particulate matter as a function of some factor that may be related back to MOVES' activity basis,
the time spent in a particular operating mode (opModelD).
                                                                                       97

-------
It should be noted that continuous ("second-by-second, or "real time") measurement of EC and
OM is an exceptionally complicated endeavor. While measurement techniques for EC have been
developed that produce apparently good correlation with traditional filter-based methods,
While numerous publications report the EC and OM (or OC) exhaust emission rates across an
entire driving cycle, it is not clear which parameter of a particular driving cycle, such as average
speed (or power), might be applicable to the extrapolation of the observed rates to other vehicles or
driving conditions. As a result, identifying one or more engine parameters that explain the
observed variation in driving cycle-based emission rates for EC and OM is desirable.  Such
parameter(s) will assist in estimating emission associated with short-term variations in driving.
One good candidate for establishing an engine-based emission model is mean effective pressure
(MEP).  MEP is defined as:
                                              VdN

Here, P is the power (in kW or hp), YIR is the number of crank revolutions per power stroke per
cylinder (2 for four-stroke engines, 1 for two-strokes), Vd is the engine displacement, and TV is the
engine speed.  In other words, MEP is the engine torque normalized by volume.
MEP can be broken into various components. "Indicated MEP" or IMEP refers to the sum of
BMEP (brake MEP) and FMEP (friction MEP). Heywood (1988) writes that maximum BMEP is
an indicator of good engine design and "essentially constant over a wide range of engine
sizes.[cite]" Nam and Giannelli (2004) note that it can be related to fuel MEP multiplied by the
indicated or thermal efficiency of an engine, and have developed trend lines in FMEP by model
year. As such, since maximum BMEP is comparable across well-designed engines and FMEP can
be well-predicted by Nam and Giannelli's trends within PERE, EVIEP should be an appropriate
metric for building an engine emission model that can be applied across vehicles with different
loads and  engine displacements.

Emission Data
Kweon et al. (2004) measured particle composition and mass emission rates from a single-cylinder
research engine based on  an in-line 2.333 liter turbo-charged direct-injection six cylinder Cummins
N14-series engine, with a quiescent, shallow dish piston chamber and a quiescent combustion
chamber.  Emission data were obtained from all eight modes of the CARB  8-mode engine test
cycle:

Speed
Load%
Equiv.
Ratio (cp)
IMEP
(MPa)
Mode 1
1800
100
0.69
1.083
Mode 2
1800
75
0.50
0.922
Mode 3
1800
50
0.34
0.671
Mode 4
1200
25
0.21
0.524
Mode 5
1200
100
0.82
1.491
Mode 6
1200
75
0.69
1.225
Mode 7
1200
50
0.41
0.878
ModeS
700
10 (idle)
0.09
0.150
                                                                                        98

-------
The study reports exhaust mass composition, including PM2.5, EC, and organic mass (OM,
estimated as 1.2 x OC) measured with TOT (denoted here as EC-TOT and OC-TOT). In the main
study, the authors report that EC and OC are highly sensitive to the equivalence ratio. However,
EVIEP is highly correlated with the measured equivalence ratio (R2 = 0.96). As such, it is
reasonable to report the data as a function of IMEP, expecting it to have approximately equal
explanatory power as has the equivalence ratio variable.  The figure below plots the emission data
from Kweon et al. (2002) as a function of EVIEP.
                       Particle Mass, EC-TOT, and OC-TOT vs. IMEP
                                                            1.4
                                                                    1.6
As shown in the figure, the EC-TOT work-specific emission rate is relatively insensitive to IMEP
except between EVIEP of approximately 0.85 and 1.1, where it undergoes a rapid increase. Overall,
the EC-TOR/EVIEP curve is S-shaped, similar to a logistic curve or growth curve. OC-TOT work-
specific emissions are highest at low EVIEP (i.e. idle) and are monotonically lower with higher
EVIEP. Total work-specific PM2.5 is not monotonic, but appears to be described by a single global
minimum around EVIEP ~ 0.9 and two local maxima around EVIEP of 0.2 and 1.2, respectively.
The oppositely signed slopes of the emission-EVIEP curves for EC-TOT and OC-TOT suggest that
there are different underlying physical processes. It is not the intent of this document to explicitly
describe the particle-formation mechanisms in a diesel engine. However, the use of two separate
functions to predict EC-TOT and OC-TOT separately is warranted.  This implies that the EC/OC
ratio will vary by engine operating mode.  The following figure depicts the EC/OC ratio as a
function of EVIEP.
                                                                                      99

-------
                       EC/OC Ratio vs. IMEP from Kweon etal. (2004)
                               Note Logarithmic Scale
     100 -,
     10 -
  o
  o
  o
  LU
     0.1 J
                0.2
0.4
0.6
 0.8       1
IMEP
1.2
1.4
1.6
Estimation of IMEP-based Emissions of EC andOC
To produce a relationship that generalizes the implied relationship between EC-TOT and OC-TOT
work-specific emissions and EVIEP in the data presented by Kweon et al. (2004), it is necessary to
specify some functional form of a relationship between the two.
A priori, on the basis of visual inspection of the data, a flexible logistic-type curve was fit to the
data by a least-squares minimization procedure using the Microsoft Excel "Solver" tool, which
employs the GRG2 optimization approach.
The functional form of the logistic-type curves fit to both the EC-TOT and OC-TOT data from
Kweon et al. (2004) is as follows:

                                       Y=^-
A least-squared error approach was implemented within Microsoft Excel to derive the coefficients
for the logistic curves for EC-TOT and OC-TOT.  The solutions to the fits are as follows:
Y
EC-TOT
OC-TOT
A
2.12x 10'5
0.155
B
-9.79
-2.275
C
4.67x 10'5
-0.859
Graphically, in comparison to observed values of EC-TOT and OC-TOT, the fitted curves result in
predictions reasonably close to the observed values. Furthermore, when compared to the observed
PM2.5 values, the sum of predicted EC-TOT and OC-TOT values predict the lack of monotonicity
and patterns of maxima and minimum seen in the PM2.5 data.
However, as a result of the values predicted by these sigmoid-type curves at high and low IMEP
values, extreme patterns in the EC-TOT/OC-TOT ratios predicted occur. These extreme values are
artifacts that result solely from the behavior of simplistic logistic curves at the bounds of IMEP in
                                                                                     100

-------
the observed data sets. As a result, for predictive purposes, the maximum and minimum observed
EC-TOT and OC-TOT values observed in the data set were set as the artificial limits of predicted
EC-TOT and OC-TOT, respectively. While this approach is arbitrary, it does ensure that extreme
predictions resulting from the selection of the logistic functional form do not occur.
The following graph (log-scale) depicts the behavior of the TOT-based EC/OC ratio as a function
of IMEP.  As demonstrated on the graph, without the max/min constraints on predicted EC-TOT
and OC-TOT, the predicted ratio assumes values with a much broader range than found in the data.
                            Comparison of EC/OC Ratio (TOT) by IMEP
                             With and Without Max/Min Constraints
     100
8
D
UJ
                                                               -EC/OC(TOT-Predicted)
                                                           - - - - EC/OC(TOT-Predicted) with
                                                                Constraints
                                                                EC/OC(TOT-Measured)
    0.01
                               IMEP
The approach of constraining predictions to the maximum and minimum values observed in the
measured data set is not grounded in any theoretical basis, but is a "brute force" approach. Future
revisions to this analysis may consider alternative approaches more grounded in accepted
theoretical or statistical methodology.
The logistic curves described above receive IMEP predictions from PERE to predict EC-TOT and
OC-TOT emission rates (g/bhp-hr) for every second of a driving cycle.  Combined with real-time
work estimates from PERE, emissions are expressed in g/s, the same units required for MOVES.
EC-TOT and OC-TOT emission rates  are converted to TOR-equivalent rates for use in MOVES,
using the TOT-TOR "calibration" relationships described above.  Alternatively, TOT-equivalent
rates can be used to  compare with data from studies employing TOT for carbon analysis.
It should be noted that these emission  estimates are based on a single engine. Therefore,
predictions of EC and OC emission rates based on these relationships are insensitive to model year,
although PERE-HD does vary frictional MEP as a function of model year.

Organic Carbon to Organic Mass Conversion
Carbon is only one component of the organic material found in PM emission samples. Hydrogen,
oxygen, and nitrogen are also components of organic molecules found in exhaust PM. For this
study, a simple set of OC/OM conversion ratios were employed.
                                                                                      101

-------
Heywood (1988) presents data on the chemical composition of diesel exhaust PM, presenting
characterization of both the "extractable composition" and "dry soot" components of PM measured
at idle and at 48 km/h.50  The composition data is as follows:

Atomic formula
OM/OC Ratio
Idle
C23H29O4.yNo.21
1.39
48 km/h
C24H3oO2.6No.18
1.26
The data for the "extractable composition" is assumed to represent  the organic mass of particles.
The total molar weight to carbon molar weight ratio was used to convert OC to OM. The idle data
from Heywood were used when engine IMEP was 0.15  or under, corresponding to the idle mode of
the cycle employed by Kweon et al. (2004). All other engine conditions employed the ratio based
on the 48 km/h sample in Heywood.
A.5.2 Comparison of Predicted Emissions with Independent Measurements
To ensure that predicted EC and OC emission rates from this approach are reasonable prior to any
application for MOVES, PERE-HD based EC and OC emission factors were compared with
measured emission factors from an independent study. Shah et al. (2004) report EC and OC
emission factor and rates for a series of heavy heavy-duty diesel trucks (HHDT) in California.51
Shah et al. report the results of emission testing using the CE-CERT Mobile Emissions Laboratory
(MEL), a 53-foot combination truck trailer containing a full-scale dilution tunnel designed to meet
Code of Federal Register (CFR) requirements.  The primary dilution tunnel is a full-flow constant
volume sampler, with a double-wall insulated stainless steel snorkel that connects the MEL directly
to the exhaust system of a diesel truck. PM collection systems were designed to meet 2007 CFR
specification, including a secondary dilution system (SDS).
The  11 trucks sampled in this study were all large HFIDDTs with engine model years 1996-2000,
odometers between approximately 9,000  and 547,000 miles, and rated powers from 360-475 hp.  It
should be noted that these trucks, on average, have larger engines and higher rated power than
"typical" trucks on the road.  Furthermore, they were loaded with only the MEL, which weighs
20,400 kg.  As a result, the emissions from these trucks do not reflect the expected variability in
truck running weight described above and used in the PERE-HD runs for this study.
Shah et al. (2004) report emission data for each of the four modes of the CARB HHDDT cycle,
including cold start/idle, creep, transient, and cruise. The test cycle represents a wide range of
driving patterns, as suggested in the table below. Note that these test cycles are trip-based,  so each
begins and ends with the vehicle at stop.
                                                                                     102

-------
Cycle
Cold start/idle
Creep
Transient
Cruise
Distance (mi)
0
0.124
2.85
23.1
Duration (s)
600
253
668
2083
Average
Speed (mph)
0
1.77
15.4
39.9
Maximum
Speed (mph)
0
8.24
47.5
59.3
Maximum
Acceleration
(mph/s)
0
2.3
3.0
2.3
The following table presents the EC-TOT and OC-TOT emission rates reported in Table 6 of the
study:
Rate
EC (mg/mi)
OC (mg/mi)
EC (mg/min)
OC (mg/min)
Idle


4.10ฑ2.38
20.9ฑ11.6
Creep
340ฑ140
607ฑ329
10.4ฑ4.8
17.0ฑ6.4
Transient
446ฑ115
182.9ฑ51.2
110.7ฑ27.0
45.5ฑ13.2
Cruise
175ฑ172
74.7ฑ56.3
93.0ฑ68.3
42.3ฑ26.8
The following graph illustrates the comparison between predicted EC-TOT and OC-TOT emission
factors predicted by PERE-HD and those reported by Shah et al. (2004).  The letters "H," "M," and
"L" refer to high, medium, and low accessory loads employed in the PERE-HD runs with IMEP-
based emission rates. As shown in the graph, it appears that for transient and cruise conditions,
PERE-HD predicts the general between-cycle trends in EC-TOT and OC-TOT emission factors. It
appears that for the low-speed "creep cycle," PERE-HD or the IMEP-based emission rates
underpredict total carbon (EC+OC) emission factors, but that the general trend in the EC/OC ratio
is directionally correct.
                                                                                   103

-------
             Predicted EC and OC Emission Factors(g/mi) vs. Measured Values in Shah et al. (2004)
                                                                  D Average of OC g/mi
                                                                  • Average of EC g/mi
A. 5.3 Variability in Predicted EC and OC Emission Rates
Through the modeling approach used here the influence of variability in vehicle weight and engine
displacement on heavy-duty EC and OC emission rates can be assessed. It should be noted that
these relationships are contingent on the particular algorithms employed in PERE-HD for
estimating power and EVIEP, as well as on the functional form of the IMEP-based emission
relationship described above.  As such, the analysis of variability in EC and OC emission rates is
constrained within the functional forms of all models employed.
The graph below depicts the TOR-specific ratios of the total amount of EC and OM emitted across
the transient driving cycle. As is apparent, increasing running weight per unit of engine
displacement is associated with an increased EC/OC ratio.  The highest EC/OM ratios, located in
the upper right-hand-quadrant of the graph, correspond to vehicles loaded with extreme weight
relative to the total available engine displacement.
                                                                                       104

-------
     25 -i
     20
     15 -
  O
  g  10-1
     5
                EC/CM Ratio (TOR-Spacific) versus Weight/Displacement Ratio for Individual Truck Samples
                              Transient Driving Cycle, High Accessory Load
                1000       2000       3000       4000       5000
                                  Running Weight/ Displacement (kg/I)
                                                                 6000
                                                                           7000
                                                                                    8000
In general, these results reflect the role that running weight has on IMEP in a truck. Since IMEP
correlates highly with the air/fuel ratio (or equivalence ratio (p), the data suggest that EC/OC
partitioning is driven by the pyrolysis that occurs in engines under load.
Very few weight/displacement pairings are greater than 3,300 kg/L.  The following graph depicts
the cumulative frequency distribution (CFD) of simulated weight/displacement ratios in PERE-HD.
                       Distribution of kg/I Ratios in Transient, High Accessory Load Simulation
    8000
    6000
    5000
    3000
       0  0.05 0.1  0.15  0.2  0.25 0.3  0.35  0.4  0.45  0.5  0.55  0.6  0.65  0.7  0.75  0.8  0.85  0.9 0.95
                                          Percentile
For a 12 L engine, 3,000 kg/L would correspond to a running weight of 39600 kg (87,302 Ib).
Such vehicle loadings are infrequent, as they exceed Federal and state limits for vehicle weights on
highways. The graph below presents the cumulative distribution of simulated weights, based on
                                                                                                105

-------
the VIUS microdata. Furthermore, the graph presents cumulative frequency distributions for
several broad weight categories reported by Ahanotu (1999) for trucks in the Atlanta metropolitan
     52
area.   Note that in the graph, the highest weight category reported by Ahanotu (1999) is
represented as 100%, although the actual maxima of observed trucks are unknown.
             Comparison of Simulated Weights with Atlanta Area Truck Measurements
                                                           Simulated Weight

                                                           GA Tech Class 9 Monday Midday
                                                      	GA Tech Class 9 Monday 3-7 PM

                                                      	GA Tech Class 8 Trucks M-F Midday

                                                      	GA Tech Class 10-13 Trucks Daytime
                                                      	GA Tech Class 10-13 Trucks
                                                           Nighttime
             20000    40000    60000    80000
                         Weight (Ib)
                                         100000
                                                120000
In general, the sensitivity of EC/OM ratios to the weight/displacement ratio suggest that properly
capturing the variability in both inputs is key to developing representative inputs for MOVES.
A.5.4 Calculating EC/OC fraction by OperatingMode
The modeling described in the previous sections has been employed to create second-by-second
estimates of EC-TOR and OC-TOR emission factors for use in the MOVES emissionRateByAge
table. The next step of consists of appropriately binning the outputs to fit the MOVES operating-
mode structure. EC and OC emission rates, as opposed to total PM, are the inputs to the MOVES
model for PM inventory calculations. To convert the total PM rates calculated from heavy-duty
emissions analysis into EC and OC rates, we must calculate EC and OC fractions by MOVES
operating mode. Then, the total PM rate can be multiplied by the EC  and OC fractions to obtain
EC and OC input emission rates.
One of PERE's outputs for heavy-duty vehicles is the track road-load coefficients.  For each
individual weight in the  distribution, PERE outputs a set ofA/B/C coefficients similar to the ones
used to calculate VSP in the HC, CO, and PM emission rate analysis.  We used these coefficients
and weights to calculate VSP  for each second using the equation below.

                                     Avt + Bv2t + Cv^ + mvtat
                                               m
This equation is implemented slightly differently than the one used for analysis of the chassis
dynamometer testing for PM, HC, and CO since the road load coefficients (A, B, and C) and weight
(or mass) m were specific to each individual vehicle, not general to the regulatory class. In the PM,
                                                                                        106

-------
HC, and CO equation, the road load coefficients and denominator mass were not specific to the
vehicle and the numerator mass was specific to the vehicle. We felt confident in using vehicle-
specific numbers because we performed the analysis using a full representative distribution of
weights and displacements. Also, since we are interested in the EC and OC fractions rather than
the actual rates themselves, normalizing by the actual weight provides a more accurate picture. For
example, a large engine operating at 90% of rated power (high VSP) would have a similar EC
fraction as a smaller engine operating at 90% of rated power, even though the large engine would
likely be hauling a proportionally greater amount of weight. This is also supported by the previous
research and analysis that relation EC fraction to EVIEP and not power itself. The large engine
would, however, emit a larger EC rate than the smaller engine, but this difference in rates is
captured by our PM emission rate analysis.
We separated vehicles into two different regulatory classes based on running weight (we did not
have GVWR information). The weight distribution used in the analysis is shown below.
                  Representative distribution of weights used in the EC/OC analysis.
                       5000 15000 25000 35OOO 45OOO 55OOO 65OOO 75OOO 85OOO 95OOO 1O5OOO 115OOO 125OOO 135OOO 145OOO
                                                              Weight_lb
Based on this weight distribution, we considered all vehicles weighing more than 40,000 Ib to be
HHD vehicles and all vehicles less than 40,000 to be MHD vehicles.  This was a very simple
approach to stratifying by regulatory class.
As EC and OC rates were also computed for each second during each cycle, we were able to
average the EC and OC rates by operating mode.  Then, we calculated the fractions of EC and OC
for each operating mode.  For the LHD classes, we used the MHD fractions, and for buses, we used
the FfflD fractions.
                     fEC =
                                 'EC
-  •> Joe
                                                       'OC
                                      OC
                                                                   I EC
                                                           OC
The resulting EC fractions by operating mode are shown in Figure 6 in the main body of this
report.
                                                                                       107

-------
A.6 Heavy-duty Gasoline Start Emissions Analysis Figures
Figure 28.  Cold-Start Emissions (FTP, g) for Heavy-Duty Gasoline Vehicles, averaged by Model-year and Age
Groups
                 FTP Cctd-Starts (g). HD SI (HD<= 14K)
                   CO slorts vs  kge b, MH
     (a) CO
                                      to   11
                       M-H-tป 19301990 1—1— 19911937
                 FTP Cold-Starts (g), HD SI (HD<=
     (b)THC
           iruroup S-e-e- 13C01989
                                 • 1991199?  -"- ^ & I9982O04
                 rrp cad-Starts (g), HD S' (HD< = -
      (c) NOx
                                      10   II
                 s 13601989 B-a-B 19301990 ' I > 19911997  a A ฑ t999?CO-l
                                                                                            108

-------
Figure 29.  Cold-Start FTP Emissions for Heavy-Duty Gasoline Vehicles, GEOMETRIC MEANS by Model-
year and Age Groups
                       FTP Cold—Starts CQ). MD SI (I-O< = 14K)
                      CO GEO-itBon starts is. Age by Hf
         (a) CO
         ซ.odelvear group  GK-ฉ~e 13601983  ^ '-'
                                   a gen id

                                  J 13301390
                                              I33M33/  a ฑ i 139B?001
                       FTP Cold-Starts (g), HD SI (hD< = UK)
                     THC  GEO-meon storls vs.  Age by M
         node1yeargr oup
                                                                                                                 109

-------
Figure 30.  Cold-start FTP Emissions for Heavy-Duty Gasoline Trucks: LOGARITHMIC STANDARD
DEVIATION by Model-year and Age Groups.
                       CO In.SB ปs.  Age by IIVS

                                               i -i: 19982004
                    FTP Cold-Starts 
-------
Figure 31.  Cold-Start Emissions for Heavy-Duty Gasoline Trucks:  RECALCULATED ARITHMETIC
MEANS by Model-year and Age Groups.
                    CO ARirH-rrieon starts ปs. ปge by HVG
         nodelyeargrOL*)  &--&-O 13601383  I  J-'-'f 13301390  1- ' * I331139?  .ซ 1 fi I33B20M
                      FTP Cold —Starts (g), HD SI (HD< = 14K)
                    THC ARITH-rteon storts ปs. Age b<
         node I year yr ou(j
                      FTP Cold-Starts (g). MD SI (HD< = 141^
                    NOx *R!TH-meon storts vs. Age by MYG
         (c) NOx
                        13601389  ^ Ct n 13901390  "•—*—<~ 1991 I
                                                                                                              111

-------
Table 38  - Emission Standards for Heavy-Duty Spark-Ignition On-road Engines
Regulatory Class

LHD2b3




LHD45, MHD




Model Year

1990
1991-1997
1998-2004
2005-2007
2008+
1990
1991-1997
1998-2004
2005-2007
2008+
Emissions Standards (g/hp-hr)
CO
14.4
14.4
14.4
14.4
14.4
37.1
37.1
37.1
37.1
14.4
THC
1.1
1.1
1.1


1.9
1.9
1.9


NMHC




0.14




0.14
NOx
6.0
5.0
4.0

0.20
6.0
5.0
4.0

0.20
NMHC + NOx



1.0




1.0

                                                                                 112

-------
A.7 Peer Review Comments and Responses

A. 7.1 Comments from Josias Zietsman, Ph.D., P.E.
Dr. Zietsman is the Head of the Environment and Air Quality Division at the Texas Transportation
Institute (TTI), as well as a faculty member at Texas A&M University.  He has over a decade of
experience in research in the areas of air quality, vehicle and engine emissions and transportation
planning. Dr. Zietsman has published widely and frequently addresses audiences in the United
States and internationally. He serves on the Transportation Research Board as Secretary of the
Performance Measurement Council and as a member of the Air Quality and Sustainability
Committees.
GENERAL COMMENTS

       •  MOVES will be a considerable improvement to MOBILE 6.2. The heavy duty
          vehicle component is a very important part of the overall MOVES program. The
          flexibility provided by the VSP approach will increase the flexibility and accuracy
          of emissions estimation. The EPA should be commended for taking on this
          important and ambitious task.

       •  The report is well written considering how difficult it is to convey the highly
          technical material. I did not notice any major flaws in the methodologies used and it
          is my opinion that the authors did a fine job in coming up with creative ways to
          produce emissions rates with limited data.

       •  It is clear that the MOVES model will have to be strengthened with a highly focused
          data collection effort that should involve emissions testing in areas where the data is
          lacking or non-existent. The MOVES team can also benefit from existing studies
          that were not included in this analysis.  For example, in my answers below I
          highlight a few studies performed by TTFs Center for Air Quality Studies that could
          be used in adding to the overall dataset.

       •  In my specific comments included in change tracker in the attached report I raise
          some questions and make some suggestions that could improve the report and  the
          analysis.
PEER REVIEW CHARGE QUESTIONS

1.      a) Does the presentation give a description of selected data sources sufficient to allow
       the reader to form a general view of the quantity, quality and representativeness of data
       used in the development of emission rates?

       The authors did a fine job in describing what data sources they used and what the
       limitations of the data were.

       b) Are you able to recommend alternate data sources might better allow the model to
       estimate national or regional default values?

                                                                                  113

-------
TTI's Center for Air Quality studies has performed quite a few studies using mostly
PEMS equipment that could enhance the database used for this analysis. We will be
happy to share any information gathered during these studies:
Identifying and Testing SmartWay Technologies for Drayage Trucks
Sponsor: Texas Commission on Environmental Quality
Budget: $220,000
Description: PEMS testing of before and after application of SmartWay technologies for 5
HHD Mexican dray age trucks.
Location: Study performed in El Paso Texas.

Emissions Testing of Carbon Chain Combustion Catalyst
Sponsor: Carbon Chain Technologies
Budget: $300,000
Description: PEMS testing of before and after application of 2CT Combustion Enhancer of
4 long haul HHD trucks and 2 LHD2b pickup trucks.
Location: Study performed at TTI's High Speed Test Track in Pecos, Texas

Expanding MOBILE6 Rates to Accommodate High Speeds
Sponsor: H ouston  Advanced Research Center  and Center for International Intelligent
Transportation Research
Budget: $150,000
Description: PEMS testing of 3 long haul HHD trucks and 3 LHD2b pickup trucks.
Location: Study performed at TTI's High Speed Test Track in Pecos, Texas

Emissions of Mexican-domiciled Heavy-Duty Diesel Trucks using Alternate Fuels
Sponsor: EPA Region 6 through AACOG
$160,000
Description: PEMS,  TEOM and filter testing of before and after application of ULSD and
biodiesel for 5 HHD Mexican drayage trucks and 5 HHD long haul Mexican trucks
Location: Study performed in Laredo, Texas

School Bus Biodiesel (B20) NOx Emissions Testing
Sponsors: CAPCOG and CAMPO
Budget: $35,000
                                                                               114

-------
       Description: PEMS testing of 5 school buses

       Study performed at TTI's Riverside Test Facility in Bryan, Texas


       Mexican Truck Emissions at Major Texas Border Locations

       Sponsors: SWUTC, EPA, and Border Environmental Cooperation Commission
       Budget: $100,000

       Description: PEMS, TEOM and filter testing of 5 HHD Mexican drayage trucks

       Location: Study performed in El Paso, Texas


       RESPONSE:
          EPA plans to analyze any new relevant data for future versions of MOVES. We are
          particularly interested infilling holes and also current and future model year data.  We
          have also conducted an extensive study at the Port of Houston on drayage trucks and hope
          to finalize the results and data from that study in the near future.
2.      a) Is the description of analytic methods and procedures clear and detailed enough to
       allow the reader to develop an adequate understanding of the steps taken and
       assumptions made by EPA to develop the model inputs?

       The descriptions are clear for the most part and the reader is able to develop an adequate
       understanding of the steps taken and assumptions made. In the attached document that is
       marked with track changing the specific places that need more clarity are shown.

       RESPONSE:
           EPA appreciates the comment.  We have incorporated many of the recommended edits
           directly into the final report.


       b) Are examples selected for tables and figures well chosen and designed to assist the
       reader in understanding approaches and methods?

       Yes, the examples for tables and figures seem to be representative to help the reader
       understand the approaches and methods.

3.      a) Are the methods and procedures employed technically appropriate and reasonable,
       with respect to the relevant disciplines, including physics, chemistry, engineering,
       mathematics and statistics?

       Yes, the methods and procedures employed seem technically appropriate and
       reasonable. The analyses are highly constrained by a lack of data and the MOVES team
       was able to use creativity  to overcome this burden
                                                                                     115

-------
      b) Are you able to suggest or recommend alternate approaches that might better achieve
      the goal of developing accurate and representative model inputs? In making
      recommendations please distinguish between cases involving reasonable disagreement
      in adoption of methods as opposed to cases where you conclude that current methods
      involve specific technical errors.

      In the attached document marked with change tracking I pointed out certain areas where
      I have some specific questions about the analyses or where I think it can be improved.
      The suggestions I made involve reasonable disagreements and I did not identify gross
      technical errors. Overall I think the methodologies are sound.

4.     In areas where EPA has concluded that applicable data is meager or unavailable, and
      consequently has made assumptions to frame approaches and arrive at solutions, do you
      agree that the assumptions made are appropriate and reasonable? If not, and you are so
      able, please suggest alternative sets of assumptions that might lead to more reasonable
      or accurate model inputs while allowing a reasonable margin of environmental
      protection.

      Yes, I think the assumptions made to deal with limited data are sound. However, in the
      attached document marked with change tracking I did point out certain areas where the
      assumptions can be improved. The key is to assemble and collect the lacking data to
      overcome the burden of making numerous assumptions and extrapolations.

      RESPONSE:
             Collecting additional data will certainly improve assumptions, but is usually a longer-
           term project.
5.     a) Are the resulting model inputs appropriate, and to the best of your knowledge and
      experience, reasonably consistent with physical and chemical processes involved in
      exhaust emissions formation and control?
      Yes, the model  inputs are appropriate are consistent with physical and chemical
      processes involved in exhaust emissions formation and control. The rates can be used as
      is but it could perhaps be improved by addressing some of the questions/comments
      highlighted in the attached report. The rates will benefit significantly from more data
      collection and assembly.
      b) Are the resulting model inputs empirically consistent with the body of data and
      literature that has come to your attention?
      Yes, the model  inputs are empirically consistent with the body of data and literature that
      has come to my attention. In my response to Question 1 above I listed 6 studies
      performed by TTFs Center for Air Quality  Studies that produced emissions testing data
      of heavy duty vehicles and that could be of use to expand  the existing  database. The
      following are some examples of areas where additional testing will greatly enhance the
      emission rates developed for heavy duty vehicles:
                 NOx emissions due to DPF regeneration


                                                                                   116

-------
           NOx during cold starts (varying soak times)
           PM emissions during starts (varying soak times)
           Aftertreatment effectiveness during extended idling
           PM2.5 emissions for heavy duty gasoline vehicles
           Emissions of 2007 and later heavy duty gasoline vehicles
           PM emissions at the higher operating mode bins
RESPONSE:
    EPA certainly is aware of our heavy-duty data needs as well as the value in additional
    testing.  We will look to incorporate any existing data that you and others have brought up
    to fill appropriate holes in the model.  Also, we are looking to improve our current and
    future model year data set with the manufacturer-run in-use heavy-duty diesel test
    program.
                                                                                  Ill

-------
A. 7.1 Comments from John Storey:
Dr. Storey is a senior researcher in the Fuels, Engines and Emissions Research Center at the Oak
Ridge National Laboratory.  He has 20 years of experience in research in emissions measurement
and characterization, including analysis of the composition of the hydrocarbon and particulate
components of vehicle exhaust. As a member of the Transportation Research Board, Dr. Storey
serves on the Committee on Transportation and Air Quality.
PEER REVIEW CHARGE QUESTIONS
    1.  The description of the data sources is adequate.  This reviewer was familiar with many of
       the studies cited, and in particular, the ROVER study, the WVU MEMS study and the
       Coordinating Research Council (CRC) E-55/59 study on heavy-duty vehicles. My
       comments are divided into sections corresponding to the report:
    NOx data The use of the ROVER and WVU MEMS data for NOx emission factor is
    appropriate and valuable in that it represents real-world operating emissions and can be directly
    related to the VSP from the weight of the vehicle and the engine data. However,  the on-board
    sensors used in the studies are typically NO sensors.  In looking at the referenced report by
    Jack, no detail on the actual sensor is given, except that it is a four gas analyzer for NOx, HC,
    CO, CO2. These units are most common in compliance inspection, not typically of research
    quality, and likely use an electrochemical sensor for NOx which is mostly sensitive to NO.
    Jack reports that intercomparison exercises were done with the EPA-NVFEL instruments, but
    nothing in the Draft MOVES2009 report explains how these results related to measurements
    with research quality instruments, nor is any assumed NO/NOx ratio reported. For instance, if
    the ROVER primarily measures NO, than the report should explain that an assumption was
    made that 95% of the NOx was NO.  This may have been done in the subsequent re-analysis of
    the data by EPA staff, but no mention is made. For the WVU MEMS testing, the referenced
    report shows that  a NO2 - NO converter is used before measurement by the ZrO2 sensor.  Also,
    a comparison was made with the WVU chassis dynamometer facility which employs research
    grade chemiluminescence analyzers for NOx measurement.  No mention of a NO2-NO
    converter prior to the ROVER is made in Jack's report.  Finally, none of the NOx data collected
    by WVU as part of the E55/E59 study was included, even though the study had trucks in  the
    age bins which were not represented by the ROVER study and the WVU-MEMS study.  The
    chassis  dynamometer data would likely be much more accurate because VSP is measured
    directly and the NOx emissions measurements are made with a research quality instrument.
           RESPONSE:
           In past testing, EPA has found good correlation in emissions results between ROVER and
           chassis and engine dynamometer testing. Further, we have conducted validation of the
           onboard in-use data (via MOVES outputs) with CRC E-55/59 results. These results are
           summarized in validation.
                                                                                     118

-------
Because of the sensitivity of tropospheric ozone predictions to NOx, and the relative
importance of HD vehicles to NOx budgets in non-attainment areas, it is my opinion that EPA
needs to use the NOx section  1.1.1 to carefully justify their decisions.  Furthermore, if there is a
future plan to include the E55/E59 NOx data, or use the data to validate MOVES, than it is
appropriate to state that in this section.
A new source of data for older vehicles which may be useful is an extensive study done by the
Texas Transportation Institute and Oak Ridge National Laboratory for the Alamo Area Council
of Governments under an EPA Region 6 grant. [Zietsman, J. et al. "Emissions of Mexican-
Domiciled Heavy-Duty Diesel Trucks Using Alternative Fuels.", October 2007] In this study,
emissions from ten trucks that operate on the Mexican border at the port of Laredo were
characterized with ultralow sulfur diesel, as well as biodiesel blends.  A wide range of model
years were represented.  Both idling and on-road drive cycle measurements  were made, the
latter with both a Sensors,  Inc. Semtech-D PEMS and a Clean Air Technologies Montana
system. Although the trucks were Mexican-registered, the engines were all  American made
(CAT, Cummins, and DDC), and the serial numbers recorded.  These engines were likely
certified to U.S. emissions standards at the time of purchase. A complete report is available by
contacting the author at tti.tamu.edu.
VSP  This section was  well-explained and the data and assumptions reasonable. The driveline
efficiencies do seem to vary considerably, but the assumption of 90% appears in line with the
data. For future consideration, there is a large set of data available to the public that includes
actual wheel torque measurements of a heavy-duty truck operating on a cross-country route,
along with engine bus and GPS data for speed, grade, etc.  The report describing the data
collection is titled "Class-8 Heavy Truck Duty Cycle Project Final Report", and is available at
http://cta.ornl.gov/cta/publications. shtml#2008 .
PM  This section does an excellent job of explaining the sources of the PM data and the
assumptions made.  The CRC E55/E59 dataset is large and very appropriate, and the use of
corrected TEOM data is as close to reality for transient emissions given the measurement
technology available at the time. Absent another huge study like E55/E59, this set of data
represents the only available comprehensive record of in-use trucks. Recent advances in
measurement technologies such as the DMM (Dekati), the Electrical Aerosol Analyzer (TSI),
and the quartz crystal microbalance (Sensors, Inc.) for real-time PM have been demonstrated,
but have mostly been applied to DPF-equipped vehicles to demonstrate their sensitivity, or
represent tiny sample sizes. It is recommended that these data, in particular for non DPF-
equipped vehicles be used in the future to validate the emissions factors in MOVES.
        RESPONSE:
        EPA agrees that it is important to test and analyze DPF-equipped vehicles for MOVES
        model validation and updates. Accurate determination of the in-use emission performance
        of these vehicles is critical in the development of future PM emission inventories.  When
        such in-use data become available in sufficient quantity EPA will analyze it and update
        the emission models as required and appropriate.
                                                                                    119

-------
HC and CO  The emissions data for HC and CO are also quite extensive and represent a wide
range of in-use vehicles. My major concern is that the altitude effects on the NFRAQ data was
considered to see if there was a systematic bias due to these effects.  Of course, the effect would
be more pronounced at low load conditions that have little or no turbocharging.  The decision to
exclude the NDIR based ROVER data; even though research grade instruments use NDIR for
CO, the instrument used in ROVER was meant for use with gasoline vehicles, and thus had
poor sensitivity at the lower levels of CO found in diesels.
Start emissions The data for diesel vehicles in this category were extremely limited. In
personal experience with these measurements, the trends shown by Figure 14 for gaseous
emissions are correct, with NOx emissions increasing after start, and CO and HC decreasing.
The difference in cold FTP vs hot FTP emissions is an appropriate measure of PM due to cold
start; I am surprised there isn't more data available since all engines must include a cold FTP in
their certification process.  Given the long daily duty-cycles of these vehicles, and their
relatively high emissions, the importance of cold start emissions for MHD HHD vehicles is
pretty limited in comparison to modern light-duty gasoline vehicles which have virtually zero
emissions after cold start and shorter daily duty-cycles.  The caveat is that the increased
emphasis on anti-idling policy may result in more start events occurring, in particular for MHD
vehicles. Thus more data is required, or at least a sensitivity analysis, to determine if multiple
starts in daily service will have an impact on a vehicle's overall contribution to the airshed.


        RESPONSE:
       Manufacturers do not typically separate their cold FTP result from their hot FTP result
        when certifying their engines. Rather, they report a weighted average. Thus, publicly
        available certification data was not a viable source for cold start rates.
Extended idling emissions  Extended idling is critical to "hot spots" around truck stops and
understanding the effects of congestion on mobile source emissions. For this part of the model,
MOVES makes use of a large amount of available data and results presented in the appendix
show generally good agreement.  This reviewer's study on extended idling found that ambient
temperature played an important role due in particular to increased accessory loads such as
cooling fan and air conditioning. It wasn't clear if these effects were included.
        RESPONSE:
        EPA appreciates the comment. At this point, we do not attempt to model temperature
        effects on extended idling emissions.

Heavy-duty gasoline data This is a very small class of vehicles and the data available are
sparse.  The reported certification data are appropriate to use for HC. NOx, and CO.  The
analysis used to calculate the PM emissions factors and the cold start emissions will be
reviewed in the next section. The only other suggestion is that the LHD2B vehicles may be
represented in the NCHRP database from CE-CERT.

2.  The description of the analytical methods and procedures are clear and detailed, for the
    most part.  Detailed comments appear below in the order of the report headings. In general,

                                                                                     120

-------
    it would help the reader to have more detail in the figure captions, particularly for the
    examples. As an example, Figure 6 is typical; assumes a MY 2002 MHD vehicle operating
    in bin 24, then displays its emissions changes with time. It would be helpful if these
    example plots throughout the document were explained briefly; i.e.,to say  that emissions
    are projected to increase due to T&M, etc. Figure 10 and 11 are confusing in that no
    explanation is given for the huge departures of the MHD vehicles (Figure 10) and bus
    (Figure 11) from the curves for the other two classes of vehicles. The justification for
    geometric mean usage given in Appendix 6 is hard to understand because the figures are of
    such poor quality.
VSP calculations   This section was well-written and communicated a somewhat difficult
concept clearly. The hole-filling methodology is important and was described adequately.


        RESPONSE:
        EPA appreciates the comment. As described in the draft report and detailed in the final
        report, we modified the calculation of tractive power to normalize the axle power by fixed
        scaling factors for the purpose of fitting rates for heavy-duty vehicles into the same
        numeric ranges originally developed and applied to vehicle-specific power (VSP) for
        light-duty vehicles, which involved normalizing by weights of individual vehicles. Rather
        than keeping the term "vehicle-specific power" for this parameter, we renamed it as
        "scaled tractive power" (STP), to minimize confusion. This change in approaches
        preserves relationship between emissions and power for engines certified to brake-specific
        standards, which better characterizes heavy-duty emissions.
Emission Rate Calculations This section also was detailed and provided a good explanation of
how the emissions rates were calculated, although in Table 10 it illustrates how little NOx data
exists for the different classes and ages of vehicles. The data from E55/E59 could be used to
fill some of these missing table cells.
        RESPONSE:
        For a given pollutant, we did not want to mix the data sources, given the differences in
        calculating road load/STP and the data collection process (drive cycles, test equipment,
        etc).  Also, keeping CRC E55/59 separate helped use that as an independent validation of
        the onboard results.
PMdata analysis (1.1.2.2) The procedures to carry out the PM analysis were detailed enough
to understand.  It is critical that the TEOM correction be understood, so even including a figure
to illustrate a snapshot of data and corrected data would be great, but not necessary for the
report. The appendix A4 does not explain the hole-filling very well. An example in the main
text or the appendix would help.
                                                                                       121

-------
        RESPONSE:
        The design of the MOVES model was originally intended to be "data driven. " To
        implement this approach, large volumes of data were dis-aggregated and reclassified into
        ranges defined as "operating modes, "for which simple statistics were computed and used
        as emission rates in the model.  This approach, however, does not produce complete
        results across all vehicle classes, ages or operating modes due to lack of data in specific
        combinations of these categories. To compensate, a "hole-filling" method was developed
        where the entire dataset was regressed using a least-squared log-linear model.  Output
       from this model was used to fill in missing bins in the model matrix of operating mode,
        vehicle class and model year group.
Tampering and Mai-maintenance T&M calculations are detailed and good examples given.
OC/EC split An excellent and extensive explanation is given in the memo in the Appendix. My
only comment is that it seems excessively detailed and long (essentially a research paper is
reproduced in the appendix), thus giving the appearance of much more weight to this
calculation and the importance of OC/EC than the others.
        RESPONSE:
        It is not the intent of the authors to over-emphasize the importance of the OC/EC split
        topic in relation to other topics in the report.  However, it is a significant and exciting new
        area of future work that the authors felt deserved full treatment.
3.  The methods employed in this report are appropriate and reasonable from the viewpoint of
    the physical and mathematical sciences.  The VSP approach is well-suited to heavy-duty
    vehicles and off-road vehicles, which will be incorporated in the future.  Although
    alternatives to the VSP approach exist, a significant commitment has been made to this
    approach and it would be extremely costly to change.
An example of an alternative approach  would be to use vehicle emissions as a function of speed
and acceleration.  The advantage of this approach  is that it makes micro-scale simulations of
road projects, i.e. for road/signal changes, much easier to perform. In addition, data is
relatively easy to collect and exists for the light-duty fleet. The data sources for the heavy-duty
fleet could be the same as  used for the development of the emissions factors, just analyzed in a
different way.  The major  disadvantage is the inability of this approach to simulate weight and
grade changes, especially for HHD vehicles. Also off-road vehicles couldn't be simulated with
speed and acceleration maps. The VSP approach offers the future advantage of incorporating
GIS information about road grade so that adjustments can be made in VSP, and thus emissions,
as a function of terrain.
Specific recommendations and comments.  Figure 3 shows deterioration rates for vehicles in
MY2010. This appears unrealistic since emissions control systems are certified for 440,000
miles, and the deterioration factors shown are very high. With modern vehicles and the
extensive diagnostic systems, it may be important to revamp the deterioration assumptions
based on earlier, no-catalyst vehicles.  If there are valid reasons for Figure 3, they are not
clearly explained in the text.

                                                                                     122

-------
        RESPONSE:
        In Figure 3 (now Figure 4 in the final report), the age effect is modeled from the
        tampering and mal-maintenance analysis. This analysis and methodology are explained
        in Appendix A. 2. We will certainly consider revision of MY 2010+ age effects as more
        data becomes available.  We do not feel it is appropriate to use pre-catalyst assumptions
        for newer vehicles since the deterioration behaviors of the two types of technology are
        different.  Further, we reduced estimated T&M effects with the introduction ofOn-board
        Diagnostics (OBD).  Heavy heavy-duty Class 8 trucks accumulate mileage quickly; Table
        35 shows the warranty age and useful life age of the various heavy-duty regulatory
        classes.
In section 1.3.2, the justification for the extended idling PM factor of 11% makes no physical
sense. The MY 2007+ vehicles are equipped with a [diesel particulate filter ]DPF which
physically filters the exhaust.  This component can't be "turned off - it remains 95+%
effective until it fills up and the engine stalls.  No bypass exists for extended idling.  In extreme
cases, the heavy HC might accumulate and break through and be counted as PM in a filter
measurement, but no data or physical evidence exists  for such an occurrence.
        RESPONSE:
        We generally agree with this comment. Accordingly, we have revised the extended idle
        PM rates, making them equal to the curb idling rates. The resulting rates are shown in
        Table 26.
4. EPA has done an adequate job with hole-filling and data sparseness for the most part. I agree
with most of the assumptions, although I think speculating on the deterioration of MY 2007+
emissions controls is very challenging, and applying "old-style" deterioration factors may be
problematic. It will be critical to update these factors as more data becomes available. Specific
comments are covered in the previous sections of this review and below:
        RESPONSE:
        We agree that future updates will be required on 2007+ model year PM emission factors
        and 2010+ NOx emission factors.   When such data are available, EPA plans a full review
        and update as feasible and required.
NOx I believe there is a wider breadth of MY data available in the CRC E55/E59 program and
this should be taken advantage of. This was mentioned previously, and the rationale for
excluding the NOx data from the CRC study, but including the PM and HC/CO data is not
explained.
        RESPONSE:

                                                                                       123

-------
           We did not want to mix the data sources, given the differences in calculating road
           load/STP and the data collection process (drive cycles, test equipment, etc).  Also, keeping
           CRC E55/59 separate helped use that as an independent validation of the onboard results.
   PM  In section 1.1.2.2.4.2, the LHD and MHD factors are proportioned by engine work ratios.
   There is little justification for this approach given except that less work is done by the LHD.
   The LHD engines are certified to the same values as the MHD engines, and my understanding
   of VSP and the TEOM data taken from the CRC study is that the PM emissions rates were
   developed for VSP bins, so the LHD vehicles populate the higher VSP bins since their power to
   weight ratios are higher. I may be completely off-base with this comment, but at any rate, the
   explanation is not sufficient for me to understand.
     The bus data correction makes more sense to me since it is based on certification level and
   buses have similar hard accelerations to HHD vehicles.
   HD gasoline vehicles The reality of the newer HDGVs is that they are equipped with the same
   three-way catalyst technology as their LDV counterparts and likely have similar calibrations to
   maximize fuel economy. It seems that the data must exist in the NCHRP or similar efforts for a
   LHD2B like the Ford F-250 and a similar model year F-150 for comparison.  The larger heavier
   vehicle may not have different VSP emissions since they are likely similarly equipped.
           RESPONSE:
           We will look into this for future analyses.  Different regulations are the driving force for
           our separation of the LHD2b 's from LD trucks, for both criteria emissions and energy
           rates.
    5.
        The resulting model inputs are consistent with the physics and chemistry of exhaust
       emissions and emissions control. I have pointed out previous inconsistencies in the
       previous sections, the most important of which are the following:


PMfor extended idling The DPF for MY 2007+ will physically continue to work, so a factor of
11% effectiveness is not realistic.
           RESPONSE:
           We generally agree with this comment.  Accordingly, we have revised the extended idle
           PM rates, making them equal to the curb idling rates. The resulting rates are shown in
           Table 26.
Gasoline HDVs Because the vast majority of the gasoline HDVs are LHD2B's, and these trucks
use powertrains developed for their much more popular LDV cousins - e.g. Ford F series,
Chevrolet Silverado - it is reasonable to assume similar performance, on a VSP basis, for these
                                                                                        124

-------
vehicles to the LDVs, despite there being a different certification standard. I would say this would
apply at least to the last 8 model years. It is not clear to me how cooperative a manufacturer might
be with the EPA in confirming this assumption, but it is probably worth a try.
              RESPONSE:
           We will look into this for future analyses. Different regulations are the driving force for
           our separation of the LHD2b 's from LD trucks, for both criteria emissions and energy
           rates.
Summary thoughts
The MOVES report on heavy-duty diesel emissions demonstrates a breadth and depth of
knowledge of the pertinent aspects of emissions measurement and emissions control for HHDs.
The issues this reviewer had were fairly minor and can be resolved easily in the future.  The report
model appears ready for use and none of my comments should hinder its timely release. A separate
attachment will identify  misplaced words and typographical errors.
Future data
Of course, it will be important to keep updating the emissions factors as additional information is
made available. As an example, Oak Ridge National Laboratory is presently evaluating two
different classes of medium-duty diesel vehicles (three transit buses, three box vans) for the
purposes of developing duty-cycle simulations. As a disclaimer, the group doing the study is not in
the same center as this reviewer, and I have no funding relationship with their center.  No
emissions are being measured in this study, but extensive duty-cycle data is being collected from
each engine and vehicle  in its daily operation.  These vehicles represent an opportunity for EPA to
fill holes in the MHD  data.  Because the recruitment and instrumentation of the vehicles has been
done, the incremental  costs for obtaining emissions data are relatively small, and the DOE program
manager funding the project has stated his interest in collaborating with EPA and other agencies on
obtaining more data from the project.
It is recommended also that funding for these updates be maintained, perhaps with collaboration
from the Regions, which will be widely affected by the implementation of MOVES. Regions will
likely be motivated to have the information in MOVES that is critical for their area in order to help
their airsheds get out of non-attainment. By keeping MOVES in the forefront when new emissions
data is collected, the value of each new study is enhanced.


           RESPONSE:
           EPA welcomes any new relevant data to incorporate into MOVES, especially infilling
           holes in existing data as well as current and future model years.
                                                                                        125

-------
References
  USEPA Office of Transportation and Air Quality. Development of Greenhouse-gas Emission Rates for all Vehicles,
and Updates to Energy Consumption Rates for Light-Duty Vehicles in the Motor Vehicle Emissions Simulator
(MOVES2010a).  Assessment and Standards Division, Ann Arbor, MI.  April 2012.

2 USEPA Office of Transportation and Air Quality . Development of Evaporative Emissions Calculations for the Motor
Vehicle Emissions Simulator (MOVES2010): Final Report. Assessment and Standards Division, Ann Arbor, MI.  July
2012.
3 USEPA Office of Transportation and Air Quality. Development of Emission Rates for Light-Duty Vehicles in the
Motor Vehicle Emissions Simulator (MOVES2010): Final Report. EPA-420-R-11-011. Assessment and Standards
Division, Ann Arbor, MI.  August 2011. (http://www.epa.gov/otaq/models/moves/documents/420rl 1011 .pdf).
4 USEPA Office of Transportation and Air Quality. Heavy Duty Diesel Engine Settlement. U.S. District Court,
District of Columbia, July, 1999.  (http://www.epa.gov/compliance/resources/cases/civil/caa/diesel).
  Jack, Jason A.   U.S. Army Aberdeen Test Center Support of Heavy Duty Diesel Engine Emissions Testing. U.S.
Army Aberdeen Test Center CSTE-DTC-AT-SL-E, Aberdeen Proving Ground, Maryland.
(http://www.epa.gov/ttn/chief/conference/eil5/sessionl/jack.pdf).
6 McClement, Dennis. Reformatting On-RoadIn-Use Heavy-Duty Emissions Test Data.  Sierra Research, Sacramento,
CA. April 2008.

7 Gautam, Mridul, Nigel N. Clark, Gregory Thompson, Daniel K. Carder, and Donald W. Lyons. Evaluation of Mobile
Monitoring Technologies for Heavy-duty Diesel-Powered Vehicle Emissions. Dept. Mechanical and Aerospace
Engineering, College of Engineering and Mineral Resources, West Virginia University, Morgantown, WV.
Gautam, Mridul, Nigel N. Clark, Gregory Thompson, Daniel K. Carder, and Donald W. Lyons. Development ofln-use
Testing Procedures for Heavy-Duty Diesel-Powered Vehicle Emissions.  Dept. Mechanical and Aerospace
Engineering, College of Engineering and Mineral Resources, West Virginia University, Morgantown, WV.

Both reports available at: (Diesel Engine Consent Decree: In-Use Test Program:
http://www.epa.gov/compliance/resources/cases/civil/caa/diesel/test.html).

8 Gautam, M., et al. Evaluation of In-Use Heavy-Duty Vehicle Emissions Using the Mobile Emissions Measurement
System (MEMS) for Engine Model Years 2001 to 2003 :  Final Reports.  Present to engine manufacturers to fulfill
testing requirements  documented in Phases III and IV of the Heavy Duty Diesel Engine consent decree. Dept.
Mechanical and Aerospace Engineering, College of Engineering and Mineral Resources, West Virginia University,
Morgantown, WV. 2002 & 2007.

9 Bradley, Ron. "Technology Roadmap for the 21st Century Truck Program." U.S. Department of Energy: Energy
Efficiency and Renewable Energy, Washington, D.C., December 2000.
10 Rakha, Hesham and Ivana Lucic.  Variable Power Vehicle Dynamics Model for Estimating Truck Accelerations.
Page 6.
http://filebox.vt.edu/users/hraldia/Publications/Variable%20Powei%20Track%20Acceleration%20-%20Vei%202
11 National Renewable Energy Laboratory.  Development ofLNG-Powered heavy-Duty Trucks in Commercial
Hauling.  NREL/SR-540-25154, Golden, CO, December 1998.

   Goodyear. "Factors Affecting Track Fuel Economy - Section 9" Page 5.
http://www.goodvear.com/track/pdf/radialretserv/Retread S9V.pdf.

   Ramsay, Euan and Jonathan Bunker. Acceleration of Multi-Combination Vehicles in Urban Arterial Traffic
Corridors. PhD dissertation,  Queensland University of Technology. August 2003, Page 11.
http://eprints.qut.edu.au/archive/00002359/01/RS&ETechForum2003_Ramsay&Bunker_2.pdf

                                                                                                    126

-------
14 Society of Automotive Engineers.  Commercial Truck and Bus SAE Recommended Procedure for Vehicle
Performance Prediction and Charting. SAE J2188. Revised October 2003.

15 Bradley, Ron. "Technology Roadmap for the 21st Century Truck Program." U.S. Department of Energy: Energy
Efficiency and Renewable Energy, Washington, D.C., December 2000. Page 32.


16 Pritchard, Ewan G. D. and Richard R. Johnson. Hybrid Electric School Bus Preliminary Technical Feasibility
Report.  Advanced Energy Corporation and Department of Mechanical Engineering, North Carolina State University,
Raleigh, NC, September 14, 2004. Page 25.

17 Hedrick, J.K. and A. Ni. Vehicle Modeling and Verification ofCNG-Powered Transit Buses. California PATH
Working Paper UCB-ITS-PWP-2004-3. California Partners for Advanced Transit and Highways (PATH), Institute of
Transportation Studies, University of California, Berkeley. February 2004. Page 21.
http://repositories.cdlib.org/cgi/viewco ntent.cgi?article=1169&context=its/path
18 Motor Industry Research Association (MTRA). News from MIRA - Automotive Engineering Specialists. Hybrid
Theory: Hybrid Vehicle engineering for economy,  the environment, and customer delight.  Nuneaton, Warwickshire,
UK. Issue 2, Spring 2007. http://data.mira.co.uk/Newsletter/MIRA-Newsletter2.pdf.
19 USEPA Office of Transportation and Air Quality. MOVES2004 Energy and Emissions Inputs - Draft Report.
EPA420-P-05-003. Assessment and Standards Division, Ann Arbor, Michigan. February, 2005.

20 Darlington, T., Dennis Kahlbaum and Gregory Thompson. On-Road NOx Emission Rates from 1994-2003 Heavy-
Duty Diesel Trucks. SAE 2008-01-1299. Society of Automotive Engineers, April 2008.

21 USEPA Office of Transportation and Air Quality. Heavy Duty Diesel Engine Consent Decree Low NOx Rebuild
Program Summary. Washington, D.C. http://www.epa.gov/compliance/resources/cases/civil/caa/diesel/progress.pdf.
  Clark, Nigel et al.  California Heavy Heavy-Duty Diesel Truck Emissions Characterization for Program E-55/59.
West Virginia University Research Corporation. Morgantown, WV. November 2005.
23 Nam, Ed and Robert Giannelli. Fuel Consumption Modeling of Conventional and Advanced Technology Vehicles in
the Physical Emission Rate Estimator (PERE).  EPA420-P-05-001.  USEPA Office of Transportation and Air Quality,
Assessment and Standards Division, Ann Arbor, MI. (http://www.epa.gov/otaq/models/ngm/420p05001.pdf). February
2005.

24 USEPA Office of Transportation and Air Quality. Development of Emission Rates for Light-Duty Vehicles in the
Motor Vehicle Emissions Simulator (MOVES 2010): Final Report. EPA-420-R-11-011.  Assessment and Standards
Division, Office of Transportation and Air Quality, Ann Arbor, MI.  August, 2011
(http://www.epa.gov/otaq/models/moves/documents/420rl 1011.pdf).
25 USEPA Office of Transportation and Air Quality. Update Heavy-Duty Engine Emission Conversion Factors for
MOBILE6: Analysis ofBSFCs and Calculation of Heavy-Duty Engine Emission Conversion Factors. EPA420-R-02-
005, M6.HDE.004. Assessment and Standards Division, Ann Arbor, MI.
(http://www.epa.gov/otaq/models/mobile6/r02005.pdf).
26 Khalek, Imad, Thomas L Bougher and Patrick M. Merritt. Phase  1 of the Advanced Collaborative Emissions Study
(ACES).  SwRI Project No. 03.13062. Southwest Research Institute, San Antonio, TX;  Coordinating Research
Council (CRC), Alpharetta, GA; Health Effects Institute, Boston, MA.  June 2009.
http://www.crcao.org/reports/recentstudies2009/ACES%20Phase%201/ACES%20Phasel%20Final%20Report%2015J
UN2009.pdf
27 Lev-On, Miriam, et al. Chemical Speciation of Exhaust Emissions from Trucks and Buses Fueled on Ultra-Low
Sulfur Diesel and CNG.  SAE 2002-01-00432.  Society of Automotive Engineers. March 2002.
                                                                                                    127

-------
  Graboski, Michael S., Robert L. McCormick, Janet Yanowitz, and Lisa Ryan.  Heavy-Duty Diesel Vehicle Testing
for the Northern Front Range Air Quality Study. Colorado Institute for Fuels and High-Altitude Engine Research,
Colorado School of Mines, Golden, Colorado. Prepared for Colorado State University. February 1998.
29 Energy and Environmental Analysis, Inc.  Documentation and Analysis of Heavy-Duty Diesel Vehicle Emission Test
Data.  Prepared for New York Department of Environmental Conservation, December 2000.

30 Calcagno, James A.  Evaluation of Heavy-Duty Diesel Vehicle Emissions During Cold-Start and Steady-State Idling
Conditions and Reduction of Emissions from a Truck-Stop Electrification Program. PhD Dissertation, Department of
Civil and Environmental Engineering, University of Tennessee, Knoxville, December 2005.
31McCormick, Robert, M. S. Graboski, T. L. Alleman, J. Yanowitz. Idle Emissions from Heavy-Duty Diesel and
Natural Gas Vehicles at High Altitude. Journal of the Air and Waste Management Association, 50(11): 1992-8.
Revised May 3,2000.
32Lim, Han. Study of Exhaust Emissions from Idling Heavy-duty Diesel Trucks and Commercially Available Idle-
Reducing Devices. EPA420-R-02-025. US EPA Office of Transportation and Air Quality, Certification and
Compliance Division.  October 2002.
33Lambert,  Douglas, et al.  Roadside Emissions Study: Preliminary Results for Stationary and On-Road Testing of
Diesel Trucks in Tulare, California. California Environmental Protection Agency. Air Resources Board. Mobile
Source Operations Division, Clean Air Technologies International, Inc May 15, 2002.
34Irick, David K. and Bob Wilson.  NOx Emissions and Fuel Consumption ofHDDVs during Extended Idle.
University  of Tennessee, IdleAire Technologies Inc. In: Proceedings, Coordinated Research Council 12th Annual On-
Road Vehicle Emission Workshop, San Diego, California, April 15-17, 2002.
35Gautam, Mridual and Nigel N. Clark. Heavy-duty Vehicle Chassis Dynamometer Testing for Emissions Inventory,
Air Quality Modeling, Source Apportionment and Air Toxics Emissions Inventory.  Phase I Interim Report, CRC
Project No. E-55/E-59,  West Virginia University Research Corporation, Morgantown, July  2002.
36National Cooperative Highway Research Program (NCHRP). Heavy-duty Vehicle Emissions. NCHRP Project 25-14,
Cambridge Systematics, Inc., Battelle Laboratories, Sierra Research and West Virginia University. October 2002.
37Tang, Shida and John Munn.  Internal Report - Idle Emissions from Heavy-Duty Diesel Trucks in the New York
Metropolitan Area. New York State Dept of Environmental Conservation, November 9, 2001.

38Broderick, Dwyer. Potential Benefits of Utilizing Fuel Cell Auxiliary Power Units in Lieu of Heavy-Duty Truck
Engine Idling. Paper UCD-ITS-REP-01-01.  Institute of Transportation Studies, University  of California, Davis, 2001.

39Storey, JohnM.E., JohnF. Thomas, Samuel A. Lewis, Sr, Thang Q. Dam, K. Edwards, Dean, Gerald L. DeVault,
and Dominic J. Retrossa.  Particulate Matter and Aldehyde Emissions from Idling Heavy-Duty Diesel Trucks.  SAE
Paper 2003-01-0289. Society of Automotive Engineers, Warrendale, PA.
40 U.S. Census Bureau. Vehicle Inventory and Use Survey. Service Sector Statistics Division, Transportation
Characteristics Branch,  http://www.census.gov/svsd/www/vius/products.html.
41 Zhou, Lei.  Revision of Heavy Heavy-Duty Diesel Truck Emission Factors and Speed Correction Factors.
California Air Resources Board, Sacramento. October 2006.


42 Illinois Environmental Protection Agency. Effectiveness of On-Board Diagnostic I/M Testing: Report to the General
Assembly, Response to Public Act 92-0682.  Bureau of Air, Springfield, IL. September 2003. Page 21.
www.epa.state.il.us/air/publications/obd-report-final.pdf
43 Manufacturers of Emission Controls Association, www.meca.org
  1 Song, Qingwen, and Zhu, George.  Model-based Closed-loop Cot
for Diesel Engine.   SAE 2002-01-287. Society of Automotive Engineers, Warrendale, PA.
44 Song, Qingwen, and Zhu, George. Model-based Closed-loop Control of Urea SCR Exhaust Aftertreatment System
                                                                                                      128

-------
45 Nam, Edward K. and Robert Giannelli.  Fuel Consumption Modeling of Conventional and Advanced Technology
Vehicles in the Physical Emission Rate Estimator (PERE).  EPA420-P-05-001. EPA Office of Transportation and Air
Quality, Assessment and Standards Division, Ann Arbor, MI. February, 2005.
46  U.S. Census Bureau. Vehicle Inventory and Use Survey  1997Microdata File [CD-ROM].  U.S. Department of
Commerce, Economics and Statistics Administration.  2000.

47 Gururaja, Prashanth, and Angela Cullen.  Accessory requirement tab in Microsoft Excel spreadsheet, VSP Weights
2008-Feb.xls. Personal communication. USEPA Office of Transportation and Air Quality, Assessment and Standards
Division, Ann Arbor, MI.  February, 2008.

48 Desert Research Institute. DRI Standard Operating Procedure:  DRIModel 2001 Thermal/Optical Carbon Analysis
(TOR/TOT) of Aerosol Filter Samples-Method IMPROVE_A.  DRI SOP#2-216r2. Division of Atmospheric
Sciences, Environmental Analysis Facility, Reno, NV. July 2008. (http://www.epa.gov/ttn/amtic/specsop.html).
49 National Renewable Energy Laboratory. DOE's Gasoline/Diesel PM Split Study. Department of Energy, Office of
Energy Efficiency and Renewable Energy, Golden, CO.  October, 2009.
(http://www.nrel. gov/vehiclesandfuels/nfti/feat_split_study. html).
50Heywood, IB. Internal Combustion Engine Fundamentals. McGraw-Hill: New York, 1988.  (Data from Table
11.8, page 629).

51 Shah, S.D. D.R. Cocker,  J.W. Miller,  and J.M. Norbeck. Emission rates of paniculate matter and elemental and
organic carbon from in-use diesel engines. Environ Sci.  Technol. 2004.38(9):  2544-2550.

  Ahanotu, D.N. Heavy-duty vehicle weight and horsepower distributions:  measurement of class-specific temporal
and spatial variability. PhD Dissertation, Georgia Institute of Technology, Atlanta. July, 1999.
                                                                                                     129

-------