Emission Adjustments for Onroad Vehicles in MOVES5


Emission Adjustments for Onroad
Vehicles in MOVES5



£%	United States

Environmental Protect
Agency

-------
Emission Adjustments for Onroad

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.

Vehicles in MOVES5

Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency

NOTICE

4>EPA

United States
Environmental Protection
Agency

EPA-420-R-24-013
November 2024

-------
Table of Contents

1.	Introduction	7

2.	Temperature Adjustments	8

2.1.	Data Sources for Gasoline Temperature Effects	8

2.2.	Temperature Effects on Gasoline Start Emissions	10

2.2.1.	THC and CO Start Emissions for Gasoline-Fueled Vehicles	11

2.2.2.	Temperature Effects on Gasoline NOx Start Emissions	19

2.2.3.	Temperature Effects on Gasoline PM2.5 Start Emissions	21

2.3.	Temperature Effects on Running Exhaust Emissions from Gasoline Vehicles	31

2.3.1.	THC, CO, and NOx Running Exhaust Temperature Effects	31

2.3.2.	PM2.5 Running Exhaust Temperature Effects	31

2.4.	Temperature Effects on Diesel Vehicles	34

2.4.1.	THC, CO, and NOx Temperature Effects for pre-2027 Diesel Vehicles	34

2.4.2.	PM Temperature Effects for Diesel Vehicles	38

2.5.	Temperature Effects on Compressed Natural Gas Vehicles	39

2.6.	Temperature Effects on ICE Vehicle Energy Consumption	39

2.7.	Temperature Adjustments for Electric and Fuel-Cell Vehicles	41

2.8.	Conclusions and Future Research	42

3.	Humidity Adjustments	44

3.1.	Humidity Adjustment Equation	44

3.2.	Future Research	45

4.	Air Conditioning Adjustments	46

4.1.	Air Conditioning Effects Data	46

4.2.	Air Conditioning Effects on Emissions and Energy	48

4.2.1.	Full A/C Adjustments for THC, CO and NOx Emissions	48

4.2.2.	Full A/C Adjustments for Energy Consumption	49

4.3.	Adjustments to Air Conditioning Effects	49

4.4.	Conclusions and Future Research	50

5.	Inspection and Maintenance Programs	51

1

-------
5.1.	Overview of Exhaust Inspection & Maintenance in MOVES	51

5.2.	Development of MOVES l/M Factors	52

5.2.1. Inspection & Maintenance in MOBILE6	55

5.3.	I/M Compliance Factors	55

5.4.	Default l/M Program Descriptions (IMCoverage)	56

5.5.	Future Research	61

6.	Electric Vehicle Charging and Battery Efficiency	63

6.1.	MOVES Design and Implementation	63

6.2.	Data Analysis and Literature Review	64

6.2.1.	Charging Efficiency	64

6.2.2.	Battery Efficiency	66

6.2.3.	Conclusion	66

7.	Fleet Averaging Provisions	68

7.1.	Fleet Averaging for Criteria Pollutants	69

7.1.1.	Tier 3	69

7.1.2.	LMDV rule	70

7.2.	Fleet Averaging for Energy Consumption and C02	70

8.	References	73

Appendix A Derivation of Temperature, Humidity and Meteorology Calculations	78

Al. Data Sets and Quality Control	78

A2. County Temperature Assignment	79

A3. Temperature Recalculation	80

A4. Relative Humidity Recalculation	81

A5. Calculation of 10 Year Averages	81

A6. Calculation of Specific Humidity	82

A7. Calculation of Heat Index	83

Appendix B OTAQ Light-duty gasoline 2012 Cold Temperature Program	84

Appendix C Air Conditioning Analysis Vehicle Sample	85

Appendix D Consistency of MOVES EV Temperature Adjustment with Other Sources	87

Dl. North American Transit Bus Study	87

D2. Japanese Passenger Car Study	88

2

-------
D3. Canadian Passenger Car Study	88

D4. Conclusion	89

Appendix E Vehicles in the 2021 ORD Cold-temperature Program	90

Appendix F Model-Fitting Information for Analysis of Fuel-Injection Technology	91

3

-------
List of Acronyms

AAA	American Automobile Association

A/C	air conditioning

ABT	averaging, banking and trading

ACCF	air conditioning correction factor

ASM	Acceleration Simulation Mode

CO	Carbon Monoxide

CDB	county database

CF	critical flow factor coefficient

CFR	Code of Federal Regulations

CNG	Compressed Natural Gas

CV	coefficients of variation

DPF	diesel particulate filter

ECCC	Environment and Climate Change Canada

EPA	U.S. Environmental Protection Agency

EV	Electric Vehicle

E85	gasoline containing 70-85 percent ethanol by volume

F	Fahrenheit

FTP	Federal Test Procedure

GDI	Gasoline Direct Injection

GHG	Greenhouse Gases

GVWR	Gross Vehicle Weight Rating

HC	hydrocarbons

HP	horsepower

ICE	Internal Combustion Engine

l/M	Inspection and Maintenance program

IM240	Inspection and Maintenance roadside vehicle driving schedule

KCVES	Kansas City Light-Duty Vehicle Emissions Study

4

-------
kW

Kilowatt

LA-92

California dynamometer driving schedule for light-duty vehicles

LDT

Light-Duty Truck

LDV

Light-Duty Vehicle

LHDT

Light Heavy-Duty Truck

LLDT

Light Light-Duty Truck

MDPV

Medium-Duty Passenger Vehicle

M0BILE6

EPA Highway Vehicle Emission Factor Model, Version 6

MOVES

Motor Vehicle Emission Simulator Model

MPGe

Miles Per Gallon Equivalent

MSAT

Mobile Source Air Toxics rules

MSOD

Mobile Source Observation Database

NEI

National Emission Inventory

NMHC

Non-Methane Hydrocarbons

NMOG

Non-Methane Organic Gases

NMIM

National Mobile Inventory Model

NOx

Oxides of Nitrogen

OBD

On-Board Diagnostics

ORD

Office of Research and Development

OTAQ

Office of Transportation and Air Quality

PFI

Port Fuel Injection

PM

Particulate Matter

RIA

Regulatory Impact Analysis

SFTP

Supplemental Federal Test Procedure

SIP

state implementation plan

SRC

selective reduction catalysts

SwRI

Southwest Research Institute

THC

Total Hydrocarbons

5

-------
US06

A drive cycle that is part of the SFTP

VIN

Vehicle Identification Number

VOC

Volatile Organic Compound

VSP

vehicle specific power

6

-------
1. Introduction

The United States Environmental Protection Agency's Motor Vehicle Emission Simulator—commonly
referred to as MOVES—is a set of modeling tools for estimating air pollution emissions produced by
onroad (highway) and nonroad mobile sources. MOVES estimates the emissions of greenhouse gases
(GHGs), criteria pollutants, and selected air toxics. The MOVES model is currently the official model for
use for state implementation plan (SIP) submissions to EPA and for transportation conformity analyses
outside of California. The model is also the primary modeling tool to estimate the impact of mobile
source regulations on emission inventories.

MOVES calculates emission inventories by multiplying emission rates by the appropriate emission-
related activity, applying correction and adjustment factors as needed to simulate specific situations, and
then adding up the emissions from all sources and regions. The highway vehicle emission rates in the
MOVES model represent emissions under a single (base) scenario of conditions for temperature,
humidity, air conditioning load and fuel properties. MOVES is designed to adjust these base emission
rates to reflect the conditions for the location and time specified by the user. MOVES also includes the
flexibility to adjust the base emission rates to reflect the effects of local Inspection and Maintenance
(l/M) programs. In addition, adjustments are applied to account for electric vehicle charging and battery
efficiency, and to account for fleet-averaging provisions of EPA rules that make the emission limits for
internal combustion vehicles dependent on the fraction of electric vehicles sold. This report describes
how these adjustments were derived and how they are implemented in MOVES. Adjustments for fuel
properties are addressed in a separate report.1

This report describes MOVES adjustments that affect running exhaust, start exhaust, and extended idling
exhaust emissions for Total Hydrocarbons (THC), carbon monoxide (CO), nitrogen oxides (NOx), fine
particulate matter (PM25) and energy consumption. The temperature effects that impact these
pollutants, also affect the pollutants that are calculated from these pollutants in MOVES, such as volatile
organic compounds (VOC)2 and individual toxics such as benzene3 (chained to THC), N02 (chained to
NOx)45, elemental carbon (chained to PM25)2, and C02 emissions (chained to energy).6 The definitions of
these pollutants and the relationship to the primary pollutants are discussed in the cited MOVES reports.
The crankcase emission processes45 are chained to running exhaust, engine start, and extended idling
exhaust emissions, and thus are similarly affected by the temperature adjustments described in this
report. The impact of fuels, temperatures, and l/M programs on vapor venting, permeation, and liquid
leaks is addressed in a separate report on evaporative emissions.7

For MOVES5, we updated the algorithms for temperature effects related to HD diesel NOx (see Section
2.4.1). We updated Section 7 to account for the fleet averaging provisions of the recent Light- and
Medium-Duty Multi-Pollutant Rule (LMDV)& and Greenhouse Gas Emissions Standards for Heavy-Duty
Vehicles—Phase 3 (HDP3).9 We also updated information about which pollutants and processes are
covered by l/M programs in various counties and calendar years as described in Section 5.4. .

7

-------
2. Temperature Adjustments

Emission rates in MOVES are adjusted by the ambient temperature to account for temperature effects
that impact emissions such as inefficient oxidation of emissions at cool catalyst temperatures and
additional fuel needed to start an engine at cold temperatures. In MOVES, exhaust emissions are
adjusted relative to their base rates at 75 degrees Fahrenheit based on two considerations:

1. Ambient temperature

2. The latent engine heat from a previous trip, applied as an adjustment based on the length of time the
vehicle has parked since operating (soak time).

This report describes the adjustment based on ambient temperature. Soak time and start emissions are
addressed in the light-duty4 and heavy-duty5 emission rates reports.

This report addresses temperature sensitivity of emissions from gasoline vehicles in Sections 2.1.
through 2.3. Although the gasoline temperature effects are developed based on emissions data from
light-duty gasoline vehicles, they are applied to all gasoline vehicles in MOVES, including motorcycles,
heavy-duty gasoline vehicles, and light-duty vehicles fueled on ethanol-gasoline blends.

Section 2.4. discusses the temperature effects derived for diesel vehicles. The data used to derive diesel
temperature effects is based on light-duty diesel vehicles but are applied to all diesel vehicles in MOVES
due to a lack of temperature effect data on heavy-duty diesel vehicles. The diesel temperature effects
are also applied to CNG buses as discussed in Section 2.5.

Section 2.6. discusses the temperature effects for energy consumption for all non-electric vehicle types
in MOVES. These effects are applied only to vehicle starts.

Section 2.7. describes temperature effects on energy consumption from battery and fuel-cell electric
vehicles.

2.1. Data Sources for Gasoline Temperature Effects

To determine the impact of ambient temperature on running emissions, our analysis included the Bag 2
emissions of Federal Test Procedure (FTP) tests as well as US06 tests (without engine starts).

For start emissions, measurements from both the Federal FTP and California Unified Cycle (3-phase / 3-
bag tests) were used. Within each test cycle, the first and third phases are identical driving cycles, but
the first phase begins with a cold-start (cold engine and emission control equipment) while the third
phase begins with a hot-start (relatively warm engine and control equipment). The difference between
Bag 1 and Bag 3 (in grams) are the emissions attributed to the cold start of the vehicle.

The data used in these analyses are from the following sources:

-------
Table 2-1 Summary of Data Sources

Data Source

Test

Temperatures Tested (deg. F)

# of Vehicles

MY Range

MSOD

FTP +

15-110

Hundreds

Pre-2005

ORD (2002)

FTP, IM240

-20, 0, 20, 40, 75

5

1987-2001

MSAT

FTP

0, 20, 75

4

2005

OTAQ

FTP, US06

0, 20, 75

9

2006, 2010

ORD (2021)

FTP

20, 71

3

2014-2015

•	MSOD - EPA's Mobile Source Observation Database (MSOD) as of April 27, 2005. EPA has
acquired data representing emissions measurements over various cycles (often the FTP) on
tens of thousands of vehicles under various conditions. EPA has stored those test results in its
Mobile Source Observational Database (MSOD).10

For the data stored in MSOD, we limited our analysis to those tests for which vehicles were tested at two
or more temperatures. The subset of tests meeting this criterion covered a temperature range from 15
to 110°F. Note that the results acquired from MSOD were collected in aggregate or "bag" modes.

•	ORD (2002) - The EPA Office of Research and Development (ORD) contracted (through the
Clean Air Vehicle Technology Center; Inc.) the testing of five cars (model years 1987 through
2001). Those vehicles were tested using both the FTP and the IM240 cycles under controlled
conditions at temperatures of 75, 40, 20, 0 and -20°F.11

•	MSATProgram - Under a contract with EPA, the Southwest Research Institute (SwRI) tested
four Tier 2 vehicles (2005 model year car and light-duty trucks) over the FTP under controlled
conditions at temperatures of: 75, 20, and 0°F. This program was used in the Regulatory
Impact Analysis of Final Rule: Control of Hazardous Air Pollutants from Mobile Sources12,
which is referred to as MSAT-2 in this report to distinguish it from an earlier mobile source air
toxics (MSAT) rulemaking.13 The MSAT-2 rule required Tier 2 vehicles to meet a non-methane
hydrocarbon (NMHC) standard on the FTP cycle of 0.3 g/milefor light-duty vehicles (<6,000
lbs) beginning phase-in for model year 2010 vehicles.14

OTAQ Cold Temperature Program (2012) - EPA's Office of Transportation and Air Quality (OTAQ)
contracted the testing of nine Tier 2 vehicles (2006 and 2010 model year car and light-duty trucks). Eight
of the nine vehicles were Mobile Source Air Toxics (MSAT-2) rule compliant. Vehicles were tested on the
FTP and US06 under controlled conditions 75, 20, and 0°F. Information on the tested vehicles is
summarized in 0 . Note that for the estimation of the THC and CO cold start effects the two GDI vehicles
were excluded from the analysis.

9

-------
ORD (2021) - A recent program was conducted under the auspices of the Office of Research and
Development (ORD). In this project, emissions were measured on three vehicles equipped with gasoline
direct injection (GDI). All three vehicles are passenger cars, including a Ford Fusion, Honda Accord and
Volkswagen Jetta, all in model year 2015. One of the vehicles is naturally aspirated, and the others
turbocharged. Mileage at test ranged from 9,000 to 13,000 miles. Emissions were measured on chassis
dynamometers over two test cycles, the Federal Test Procedure (FTP) and the US06. For the FTP, results
are available by phase. Emissions were measured on a single fuel, a "winter E10" at two temperature
levels, 20 and 71°F. A variety of pollutants were measured, including the gaseous criteria pollutants and
particulate matter. Particulate matter, as PM2.5, was measured gravimetrically on three replicate filters
in a heated box and with sample flows drawn from a constant-volume sampler (CVS). Replicate
measurements were also collected from each filter holder.

2.2. Temperature Effects on Gasoline Start Emissions

When a vehicle engine is started, emissions can be higher than during normal operation due to the
relatively cold temperature of the emissions control system. As these systems warm up to their ideal
operating temperature, emissions from the vehicle can be dramatically reduced. The cold start effect can
vary by pollutant, temperature, and vehicle technology.

The effects of ambient temperature on THC, CO, and NOx start emissions were developed using the
following approach:

• No adjustment for temperatures higher than 75°F. 75°F is the midpoint of the allowable
temperature range (68°F-86°F) per the FTP.

• Additive adjustments for temperatures below 75°F. These adjustments are added to the
emissions that would occur at 75°F.

• Calculate the adjustments as either polynomial (Equation 2-1) or log-linear (Equation 2-2)
functions, depending on model year group and pollutant:

Additive Grams = A*(Temp-75) + B*(Temp-75)2 Equation 2-1

Additive Grams = Be A* + c Equation 2-2

This approach provides a value of zero change for the additive adjustment at 75°F (i.e., the temperature
of the federal FTP test). The coefficients, A and B, for the adjustment equations are stored in the
StartTempAdjustment table. This table contains temperature effect coefficients for each model year
group, operating mode, and pollutant.

In MOVES, the temperature effects for older model year groups use polynomial function (Equation 2-1)
and more recent model year vehicles use log-linear function (Equation 2-2). The data processing and the
model fitting process differed for the polynomial and log-linear fits, and each is described separately
below.

-------
2.2.1. THC and CO Start Emissions for Gasoline-Fueled Vehicles

In developing temperature adjustments for THC and CO start emissions, both polynomial and log-linear
regression models were used to fit the data. Data anomalies were resolved by combining two or more
model year groups to obtain a larger dataset, or by removing anomalous data points. We also distinguish
temperature effects between pre-MSAT-2 (Mobile Source Air Toxics) and MSAT-2 compliant vehicles,
which began phase-in starting in 2010. The MSAT-2 rule included the first regulation on low temperature
(20°F) non-methane hydrocarbon (NMHC) emissions for light-duty and some medium-duty gasoline
vehicles.14

Polynomial Fits

The coefficients for THC emissions for pre-2006 gasoline vehicles and CO emissions for pre-2001 gasoline
vehicles were calculated with polynomial fits to data processed in the following steps. First, the cold start
emissions (grams/start) were calculated as the difference between Bag 1 and Bag 3 emissions for each
relevant vehicle test in the MSOD, ORD and MSAT data. Next, the cold start emissions were stratified by
model year groups. The data was initially grouped according to the following model year groups:

» 1960 to 1980
» 1981 to 1982
» 1983 to 1985
» 1986 to 1989
» 1990 to 1993
» 1994 to 1999
» 2000 to 2005

Then, the mean emissions at 75°F were subtracted from the mean emissions at the other temperatures
to determine the change in emissions as functions of ambient temperature. Then, we modeled the
changes in cold-start emissions as a polynomial function of temperature minus 75°F. The additive
adjustments are set equal to zero for temperatures higher than 75°F. Thus, we did not use the changes in
emissions from temperature above the FTP temperature range (68° to 86°F). The model year groups
were aggregated to larger intervals when the less aggregated groups yielded non-intuitive results (e.g.,
older model year group had lower cold start emissions).

Table 2-2 summarizes the coefficients used with Equation 2-1 (polynomial) to estimate additive start
temperature adjustments for older model year gasoline vehicles.

-------
Table 2-2 Polynomial Model Coefficients for CO Temperature Effects for 2000 Model Year and Earlier Gasoline
Vehicles and THC Temperature Effects for 2005 and Earlier Gasoline Vehicles



CO

THC

Model Year Group

A

B

A

B

Pre-1981

-4.677



-0.631



1981-1982

-4.631



-0.414



1983-1985

-4.244



-0.361



1986-1989







0.002

1986-2000



0.023





1990-2005







0.003

The THC test data for the 1986-1989, and 1990-2005 model year groups included the ORD program
vehicles that were tested at an ambient temperature of -20°F. However, when this ultra-low temperature
data was included, the "best fit" THC regression curves (linear, quadratic, and cubic) all exhibited poor
fits for temperatures from zero through 20°F. We removed the five ORD vehicle tests conducted at -20°F,
which improved the estimate of the cold-start THC emissions in the more common 0° F to 20°F range.
Therefore, the coefficients in MOVES are based on the changes in cold-start emissions for temperatures
from zero through 75°. However, these coefficients are applied to all ambient temperatures below 75°F
in MOVES.

For CO, the temperature effect developed based on the 1994-2000 model year vehicles was applied to
all model years from 1986-2000, because including 1986-1993 model year vehicles in the analysis
resulted in cases where older model years were modeled with substantially lower CO emissions than
newer model years. Note that the base CO emission rates still vary across this model year range.

To adapt the additive ambient temperature adjustments to account for intermediate soak times, the A
and B coefficients for start operating modes other than cold starts were reduced by multiplying by a
factor equal to the ratio between emissions at the desired soak time and the cold start emissions for
catalyst equipped vehicles as used in MOBILE6.15 These factors are summarized in Table 2-3.

Table 2-3 Soak Time Multipliers for Additive Start Temperature Effects

Operating Mode ID

Nominal Soak Time (min)

THC

CO

NOx

108

720

1

1

1

107

540

0.908778

0.91377

1.053118

106

240

0.733962

0.79137

1.117624

105

105

0.64496

0.72996

1.128799

104

75

0.599625

0.6285

1.129778

103

45

0.444825

0.44136

1.02786

102

18

0.208548

0.199678

0.58398

101

3

0.037593

0.035422

0.20508

12

-------
Log-linear Fits

In estimating the THC temperature effect for model years 2006 and later and the CO temperature effect
for model years 2001 and later3, data from ORD, MSAT and OTAQ cold temperature programs'5 were used
to fit regression models. We used linear mixed models, with both continuous and categorical variables,
to fit to the logarithm of the start emissions. Second-order polynomial models exhibited non-intuitive
behaviors (e.g., negative values, non-monotonically increasing emissions). Thus, we chose to fit the data
with log-linear models because they provide monotonically increasing emissions at colder temperatures
and can model the strong curvature evident in the cold start data (See Figure 2-1 and Figure 2-2).

The model parameters were fit using linear mixed models using the function Ime within the R statistical
package nlme.16 Using random effects for vehicle, and the test temperature as a fixed effect, we
accounted for the paired test design of the data set, yielding robust temperature effect estimates for the
entire data set (e.g., not all vehicles were tested at the same set of temperatures which is evident at -
20°F in Figure 2-1).

The linear mixed model had the following form:

log(y) = oc + /?! ¦ Temp + Veh Equation 2-3

Where:

y = start emissions (grams)

Temp = temperature in Fahrenheit

Veh = random effect for each individual vehicle

The mean model simply removes the random vehicle effects:

log(y) = oc + /?i ¦ Temp Equation 2-4

We then converted the mean logarithmic model to real-space, yielding:

y = e
-------
We then normalized to degrees below 75°F, by setting T' = 73 — Temp , and substituting Temp =
75 — T' into the above equation and rearranging. This yields the equations:

y= e°c+(3i(75-T') Equation 2-6

y = /irT) Equation 2-7

y = e + c Equation 2-2

The initial estimated fixed effects (including p-values) for the linear model fit for CO are displayed in
Table 2-4. The model estimates that the Portable Fuel Injection (PFI) MSAT-2 compliant vehicles (Model
year 2010) tested in the OTAQ 2012 test program have consistently lower CO start emissions than the
pre-MSAT-2 vehicles (pre-2010), as shown by the positive pre-MSAT coefficient (a2). However, no
statistically significant difference in the log-linear impact of temperature (coefficient (B) was found
between the 2001-2009 and the 2010 model year groups for CO emissions, as shown in Table 2-4 (p-
value of the Temperature x pre-MSAT effect is >0.90).

Table 2-4 Fixed Effects for the Initial CO Model Fit to Data from 2001+ Model Year Vehicles from the ORD, MSAT
and Cold Temperature Programs (13 vehicles, 95 observations)

Value

Std.Error

t-value

p-value

Intercept (ai)

3.5502

0.1433

24.8

2.8E-39

Temperature (Pi)

-0.0380

0.0022

-17.5

4.3E-29

pre-MSAT (a2)

0.7378

0.2066

3.6

0.0044

Temperature ((3i) x pre-MSAT (012)

-0.0003

0.0032

-0.1

0.9225

Because there was not a significant temperature effect between the pre- and post-MSAT-2 vehicles, we
estimated the temperature effect (Pi) from a model fit where the pre-MSAT-2 and post-MSAT-2 vehicles
are pooled together as shown in Table 2-5.

-------
Table 2-5 Fixed Effects for the Final CO Model Fit to Data from 2001+ Model Year Vehicles from the ORD, MSAT
and Cold Temperature Programs (13 vehicles, 95 observations)

Value

Std.Error

t-value

p-value

Intercept (ai)

0.6914

0.1400

4.94

4.1E-06

Temperature (Pi)

-0.038

0.0016

-24.08

1.1E-38

pre-MSAT («;)

0.7284

0.1815

4.01

0.0020

The data along with the final model fits are displayed in Figure 2-1. The MSAT-2 compliant group (2010+)
has significantly lower base cold start (coefficient a), which causes the emissions to be lower across all
temperatures for the newer model year vehicles. The CO model coefficients in the form of Equation 2-2
for use in MOVES are provided in Table 2-8. The 2009 and 2013 model year B values are derived from the
linear mixed model for the pre-MSAT-2 and the MSAT-2 compliant groups, respectively. The 2010 through
2012 model year B values are derived by linearly interpolating the 2009 and 2013 values.

0 25 50 75

Degrees (F)

Figure 2-1 FTP CO Start Emissions with Log-linear Model Fit

For THC emissions, a statistically significant difference was detected in the log-linear temperature effect
((Bi) between the pre-MSAT-2 and MSAT-2 compliant vehicles as shown in Table 2-6 (p-value of the
Temperature x pre-MSAT term is much smaller than 0.05).

-------
Table 2-6. Fixed Effects for the Final THC Model Fit to Data from 2006+ Model Year Vehicles from the MSAT
Program and the Cold Temperature Program (11 vehicles, 69 observations)

Value

Std.Error

t-value

p-value

Intercept (ai)

1.8613

0.1321

14.1

4.6E-20

Temperature (|3i)

-0.0394

0.0011

-34.6

1.7E-39

pre-MSAT (a2)

0.7503

0.2254

3.3

0.0088

Temperature (|3i) x pre-MSAT (a2)

-0.0111

0.0021

-5.2

2.7E-06

The THC model fit to the cold start emissions data is graphed in Figure 2-2. As shown, the pre-MSAT-2
cold start emissions for THC are much more sensitive to cold temperature than the MSAT-2 compliant
vehicles.

Program

o MSAT Vehicle Data 2010+

• pre-MSAT Vehicle Data 2006-2009

Model

\ •

— Model Fit 2010+

E-Model Fit 2006-2009

v S
\ S
\ S

X. * v-

1 \

X. O s

•

ft s X ^

•

8 -—-

' — 3

Degrees (F)

Figure 2-2 FTP THC Start Emissions with Log-linear Model Fit

The differences in the THC cold start temperature effect represent the impact of the Mobile Source Air
Toxic (MSAT-2) rule. The MSAT-2 rule included a limit on low temperature (20°F) non-methane
hydrocarbon (NMHC) emissions for light-duty and some medium-duty gasoline-fueled vehicles.14
Specifically:

-------
• For passenger cars (LDVs) and for the light light-duty trucks (LLDTs) (i.e., those with GVWR up to 6,000
pounds), the composite (combined cold start and hot running) FTP NMHC emissions should not exceed 0.3
grams per mile.

• For light heavy-duty trucks (LHDTs) (those with GVWR from 6,001 up to 8,500 pounds) and for medium-
duty passenger vehicles (MDPVs), the composite FTP NMHC emissions should not exceed 0.5 grams per
mile.

These cold weather standards are phased-in beginning with the 2010 model year, as shown in Table 2-7.

Table 2-7 Phase-in of Vehicles Meeting Cold Weather THC Standard

Model Year

LDVs / LLDTs

LHDTs / MDPVs

2010

25%

2011

50%

2012

75%

25%

2013

100%

50%

2014

100%

75%

2015

100%

For the phase-in years, the coefficients for the THC temperature effect equation in the
startTempAdjustment table were adjusted linearly according to the light-duty vehicle phase-in. Equation
2-9 shows how the temperature effect is calculated for a model year 2010 LDV, where A2010 is the 2010
emissions rate:

•^2010 — -^2009(1 0.25)+ ^2013(0-25) Equation 2-9

With this approach, the log-linear temperature effect (coefficient A) for THC emissions is reduced from
2009 to 2013 while the base 75° F THC cold start (coefficient B) is relatively constant.

Within the current MOVES design, temperature effects are applied by fuel types and model year
vehicles, but not by regulatory class (e.g., LHDTs/MDPVs). As such, the light-duty rates, including the
light-duty MSAT-2 phase in are applied to all the gasoline-fueled vehicles in MOVES. No data on
LHDTs/MDPVs or heavy-duty temperature effects were available to assess this approach.

Table 2-8 summarizes the coefficients used with Equation 2-2 (log-linear) to estimate additive start
temperature adjustments for more recent model year gasoline vehicles.

-------
Table 2-8. Coefficients Used for Log-linear Temperature Effect Equation for All Gasoline Source Types

THC

Model Year Group

2001-2009

-0.038

4.136

-4.136

2006-2009

-0.051

0.308

-0.308

2010

-0.038

3.601

-3.601

-0.048

0.315

-0.315

2011

-0.038

3.066

-3.066

-0.045

0.322

-0.322

2012

-0.038

2.531

-2.531

-0.042

0.329

-0.329

2013 & later

-0.038

1.996

-1.996

-0.039

0.336

-0.336

Figure 2-3 and Figure 2-4 graphically compare all the cold start temperature effects for gasoline vehicles
by model year groups in MOVES for CO and THC, respectively. These include both the polynomial fits and
the log-linear curve fits to the data.

300

L—

-------
40-

O)30

120

20 40 60

Temperature (deg F)

Pre1981

— MY81 82
MY83_85

— MY86_89
MY90_05

— MY06_09
MY10
MY11
MY 12

— MY13 50

Note: in MOVES, "MY13_50" applies to all model years 2013-2060,

Figure 2-4 THC Additive Cold Start Temperature Effects for Gasoline Vehicles by Model Year Groups

To adapt the additive adjustments for intermediate soak times, the B and C coefficients for start
operating modes other than cold starts were reduced by multiplying by a factor equal to the ratio
between emissions at the desired soak time and the cold start emissions for catalyst equipped vehicles
as used in MOBILE61= as summarized in Table 2-3.

2.2.2. Temperature Effects on Gasoline IN Ox Start Emissions

Cold-start NOx emissions are not as sensitive to ambient temperature changes as THC and CO emissions,
because the fuel-rich conditions at engine start favor incomplete combustion of fuel, forming CO and
THC; NOx is favored under the lean burn, high temperature engine operation more typical of running
emissions. However, NOx emissions are impacted by the inefficiencies of the three-way catalyst at low
temperatures and a small cold start temperature sensitivity is expected.

Due to the small temperature effects and the variability of the data, the NOx temperature effect was
calculated in MOVES by averaging all the available NOx results (i.e., the 2005-and-earlier model year
data) together across model year groups and then performing regression. Table 2-9 lists the average
incremental cold start NOx emissions, compared to 76.3°F, from the MSOD, ORD, and MSAT programs.

-------
Table 2-9. Average Incremental Cold Start NOx Emissions by Temperature for Gasoline Vehicles Calculated from

the MSOD, ORD and MSAT Programs

Temp F

Delta

NOx (grams)

-20

1.201

1.227

19.4

0.202

20.7

0.089

22.4

-0.155

-0.007

0.876

48.8

0.127

49.8

0.333

0.325

54.2

0.438

76.3

95.3

0.225

97.1

0.37

105.8

0.543

Using the data above, we fit a linear regression to the emission averages for temperatures of 76.3°F and
lower and obtained the following fit:

NOx temperature additive adjustment = A * (Temp - 75) Equation 2-10

Where:

A = -0.009
Ft2 = 0.61

Although the value of R2 is not as high as for the THC and CO regression equations, the fit is statistically
significant.

Note that Equation 2-10 predicts a decrease in cold-start NOx emissions for temperatures greater than
75°F, while the data in Table 2-9 indicates an increase in cold-start NOx emissions as the ambient
temperature rises above 90°F. The increase is small and may be an artifact of how these data were
analyzed, since only a subset of vehicles was measured above 75°F. Therefore, as with the other
temperature adjustments, we have set the NOx additive adjustment to zero in MOVES for temperatures
higher than 75°F.

-------
In addition, we investigated whether different N0X temperature correction is needed for vehicles subject
to the MSAT-2 rule. Figure 2-5 shows a comparison between NOx start emissions data from OTAQ Cold
Temperature Program, including both the port-fuel injection (PFI) and gasoline-direct injection (GDI)
2006-2010 model year vehicles, and the emissions predicted using temperature effects calculated from
the MY2005-and-earlier vehicles. Because start emissions compose such a small percentage of total NOx
emissions, the differences between the MOVES temperature effects and the NOx data from the OTAQ
Cold Temperature Program were considered negligible. Thus, we applied the NOxtemperature
adjustment estimated in Equation 2-10 for all model years.

0 20 75

Temperature (degF)

Figure 2-5 FTP Start NOx Emissions, Bag 1 - Bag 3, Model Years 2006-2010

To adapt the additive adjustments for intermediate soak times, the A coefficients for start operating
modes other than cold starts were adjusted by multiplying by a factor equal to the ratio between
emissions at the desired soak time and the cold start emissions for catalyst equipped vehicles as used in
MOBILE6 and summarized in Table 2-3.

2.2.3. Temperature Effects on Gasoline PM2.5 Start Emissions

The temperature effects for particulate matter emissions from gasoline engines were obtained from the
Kansas City Light-Duty Vehicle Emissions Study (KCVES)17, conducted between 2004 and 2005. The KCVES
measured emissions from 496 vehicles collected in the full sample, with 42 vehicles sampled in both the

-------
winter and summer phases of the program. The EPA conducted an analysis of the temperature effects of
gasoline vehicles from the KCVES by estimating the temperature effect on PM emissions from 34 paired
vehicle tests that were sampled in both winter and summer ambient conditions (10 paired vehicle tests
were removed due to missing values and/or too small temperature differences between the phases) as
described in the EPA report17 and subsequent analysis.18

The analysis of the KCVES data indicated that ambient temperature affects for start PM emissions is best
modeled by (log-linear) multiplicative adjustments of the form:

Equation 2-11

Multiplicative Factor = cA"r2~Tem|;"

Where:

Temp = Temperature

A = log-linear temperature effect A = 0.0463 for cold starts from the KCVES analysis1516

The log-linear temperature effect of 0.0463 is used in MOVES for gasoline vehicles of model year 2009
and earlier (i.e., vehicles not affected by the MSAT-2 requirements).

The MSAT-2 rule (signed February 9, 2007) does not explicitly limit cold weather emissions of particulate
matter (PM). However, the Regulatory Impact Analysis (RIA) document that accompanied the rule12
noted there is a strong linear correlation between NMHC and PM2.5 emissions based on the MSAT
program discussed in Section 2.1. That correlation is illustrated in Figure 2-6 (reproduced from that RIA)
as the logarithm of the Bag-1 PM2.5 versus the logarithm of the Bag-1 NMHC (for various Tier-2 vehicles).

-------
CN _

^ _

CT)

CD _

V O A
V o

V
4- V

# ¥

™i

V v

A o

-3

-2

-1

Bag 1 NHMC - ln(g/mi)

Plot Icons are Vehicle-Specific

Figure 2-6 FTP Bag 1 PM and FTP Bag 1 NMHC for Tier 2 Vehicles

Therefore, the limitation on cold weather THC (or NMHC) emissions is expected to result in a
proportional reduction in cold weather PM2.5 emissions. In the MSAT-2 RIA (Table 2.1.-9), EPA estimated
that this requirement would result in a 30 percent reduction of VOC emissions at 202F. Applying the
same analytical approach that was used in the RIA means that a 30 percent reduction in VOC emissions
would correspond to a 30 percent reduction in PM emissions at 20° F (for Tier 2 cars and trucks).

Applying the 30 percent reduction for vehicles affected by the MSAT-2 requirements to the temperature
effects calculated for the fully phased-in (2015+) MSAT-2 vehicles implies a PM increase as the
temperature decreases from 72° to 20° F of:

Multiplicative Factor at 2G F for MSAT-2 Vehicles = o.7*e0 0463*(7 2 20) Equation 2-12

= 7.8

-------
Using Equation 2-12 with the MSAT-2 phase-in schedule from Table 2-7 leads to the following
(multiplicative) increases as the temperature decreases from 72° to 20° F:

Table 2-10 Multiplicative Increase in Cold Start PlVhsfrom 72° to 20° Fahrenheit for Gasoline Vehicles

Model Year

LDVs / LLDTs

LHDTs / MDPVs

2008

11.1

2009

11.1

2010

10.3

11.1

2011

9.4

11.1

2012

8.6

10.3

2013

7.8

9.4

2014

7.8

8.6

2015+

7.8

Solving for the corresponding log-linear terms gives us these "A" values:

Table 2-11 Log-linear Temperature Effect for Start PM2.5 Emissions (Coefficient A) for Gasoline Vehicles

Model Year

LDVs / LLDTs

LHDTs / MDPVs

2008

0.0463

2009

0.0463

2010

0.0448

0.0463

2011

0.0432

0.0463

2012

0.0414

0.0448

2013

0.0394

0.0432

2014

0.0394

0.0414

2015+

0.0394

We confirmed this theoretically derived temperature effect for MSAT-2 compliant vehicles by comparing
it to data from the OTAQ Cold Temp Study, which includes only the MY 2010 PFI vehicles(See Appendix B)
The temperature effect developed for MOVES fits this data well, as shown in Figure 2-7. Note, as
discussed in the light-duty report, we significantly updated the start PM2.5 emission rates to account for
GDI vehicles in MOVES3 and made additional minor updates in MOVES4,4 but we did not revisit the
temperature effects for start emissions.

-------
0 20 75

Temperature (degF)

Figure 2-7. FTP PM2.5 Start Emissions, MSAT-2 Compliant Vehicles (7 PFI Vehicles, 40 Tests with Nonzero PM
Measurements on E10 Fuel) from OTAQ Cold Temperature Program

Figure 2-8 presents the light-duty multiplicative temperature effects using the coefficient from Table
2-11, and the model form of Equation 2-11.

-------
if)

.2 20-

if> ^

c/> HT

£ ==
LLi .9-
±£
t =,

TO £

— Pre 2010
MY201 0
MY2011

— MY2012

MY201 3

2050

20 40

Temperature (deg F)

Note: In MOVES, "MY2013_2050" applies to all model years 2013-2060.

Figure 2-8. PM Start Exhaust Emissions Effect for Gasoline Light-Duty Vehicles in MOVES

Because the PM2.5 speciation profile for gasoline vehicles did not change significantly between the winter
and summer rounds of the KCVES,19 we apply the same temperature adjustment to each component of
the PM emissions, including elemental carbon, organic carbon, sulfate, and other species. The PM start
temperature adjustment does not vary with soak time since it is multiplicative.

Effect of Fuel-Injection Technology on Temperature Effects for PM Start Emissions
The adjustment for start emissions described above represents only vehicles equipped with fuel-
injection technologies prevalent in 2005, presumably port fuel injection (PFI). Since then, an alternate
technology, "gasoline direct injection" (GDI), has entered the market and come to represent a major
market share.

This development raises the question as to whether vehicles with GDI would respond differently to cold
ambient temperatures than those equipped with PFI. To investigate this question, we combined two
datasets, OTAQ (2012) and ORD (2021), which gives a vehicle sample that includes both technologies.
Our analysis, explained below, found that a single logarithmic slope term (or rate constant), as in
Equation 2-11 above, can be appropriately used as the basis for a temperature adjustment to represent
fleets including both PFI and GDI-equipped vehicles.

-------
As detailed in Appendix B , the ORD dataset includes three vehicles, all equipped with GDI. The OTAQ
data includes nine vehicles, of which two are GDI-equipped. Combining the two samples gives a total of
12 vehicles, with five GDI-equipped and seven PFI-equipped. This sample enables an analysis designed to
test the hypothesis that the trend in PM2.5 with ambient temperature might differ between GDI and PFI.

For this purpose, we used results from the cold-start phase of the FTP cycle (Bag 1). Figure 2-9 shows
logarithmically transformed PM, as mg/mi (InPM) vs. temperature for all 12 vehicles, with those from
the recent ORD project distinguished with the prefix "ORD_" and those from the older OTAQ program
identified with the prefix "OTAQ_". In this figure, the view is restricted to the temperatures between 20
and 75°F, despite the fact that some vehicles in the OTAQ program were measured at 0°F. This analysis
focused on the question of whether the temperature trend differs between PFI and GDI over this
temperature range. A linear trendline is imposed on each panel, which reflects an assumption that the
emissions trend is log-linear over this temperature range.

-------
ORD_Accord ORD_Fusion ORD_Jetta

S OTAQ_Lucerne OTAQ_Mazda6 OTAQ_Passat

temperature
o ln(PM Rate) Regression

Figure 2-9. InPM: logarithmically transformed FTP Phase-1 emissions (mg/mi) vs. temperature, by vehicle.

For a more focused comparison of the two fuel-injection technologies, Figure 2-10 shows the data
grouped by vehicle and paneled by fuel injection. As a body of data, the GDI data sits higher, with the
exception of the ORD Accord, with has the lowest emissions at warm temperature and an apparently
steeper trend. With the exception of this vehicle, the two bodies of data have similar slopes.

-------
injection = GDI injection = PFI

temperature

vehicle

ORD_Accord

0 ORD_Fusion

ORD_Jetta

OTAQ_Passat

0 OTAQ_STS

OTAQ_Accord

OTAQ_Forte

0 OTAQ_Gallant

OTAQ_Lucerne

OTAQ_Mazda6

OTAQ_Patriot

OTAQ_Santa Fe

Figure 2-10. InPM: logarithmically transformed FTP Phase-1 emissions (mg/mi) vs. temperature,

by vehicle and fuel-injection technology.

This body of data is sufficient to fit a model to test the hypothesis that the two fuel-injection
technologies could have different (logarithmic) trends with temperature over the range of 20-70°F. The
mixed-factor 'random coefficients' model includes 'fixed' effects for temperature and fuel injection, as
well as 'random' intercepts and slopes for each vehicle.

InPM^,. = p0 + p±T + p2(}3 + P2P4T + b0v + blvT Equation 2-13

+ £v,r

Where:

InPM = natural-log transformed PM emissions (mg/mi), for a given replicate for a given vehicle,
T = soak temperature (°F), treated as a continuous variable,

60 = a fixed intercept term, reflecting averaging across all vehicles,

61 = a fixed slope term, reflecting averaging across all vehicles,

62 = a dummy variable indicating fuel- injection technology (0 if PFI, 1 = GDI),

63 = an fixed intercept increment representing the effect of fuel injection,

64 = a fixed slope increment representing the effect of fuel injection.

-------
bo,v = a "random" increment in the intercept with respect to 60, for vehicle v, e.g., the individual
intercept for vehicle v is 60 + b0,v.

bi/V = a "random" increment in the slope with respect to 61, for vehicle v, e.g., the individual
slope for vehicle v is 61 + bi/V.

sr = residual error variance for replicate r.

Accordingly, when (B2 = 0, the model for PFI vehicles = (B0 + (BiT, and when (B2 = 1, the model for GDI
vehicles = ((B0 + p3) + (Pi+ (B4)T

As Figure 2-9 and Figure 2-10 suggest, the model fits individual trends (intercepts and slopes) for each
vehicle and treats the trends for the vehicles as representing random variation around a mean "fleet"
trend.

For this dataset, the random component of the best-fit model contains 14 covariance parameters,
including two variances for the random intercepts and slopes, that describe the variance among vehicles,
plus individual error variances for each of the 12 vehicles.

The solution for the fixed-effects in the best-fit model is shown in Table 2-12. Additional model-fitting
information, including the solution for the random effects, is presented in Appendix F.

Table 2-12. Fixed-Effects Solution for the Best-fit Temperature-effects Model.

Effect

Fuel Injection

Estimate

Standard
Error

t value

Pr > M

Intercept (fin)

3.3669

0.1301

12.5

25.88

<.0001

Temperature T (fi1)

-0.03078

0.003357

14.6

-9.17

<.0001

Fuel injection (/;;,)

GDI (fi2 = 1)

1.1018

0.1833

9.78

6.01

0.0001

Fuel injection

PFI (fi2 = 0)

Temperature x Injection (/;4)

GDI (fi2 = 1)

-0.00563

0.004951

12.7

-1.14

0.2763

Temperature x Injection

PFI (fi2 = 0)

The initial question in model fitting is whether the interaction term for temperature and fuel injection
(64) is significant and improves model fit. If this term were significant, it would indicate that the
logarithmic slope for GDI-equipped vehicles differed from that for PFI-equipped vehicles. As the table
shows, the value for this coefficient is small relative to 61 and its own standard error, resulting in a small t
statistic and correspondingly large and insignificant p-value. The model fitting thus indicates that both
GDI and PFI equipped vehicles can be modeled with the same slope term.

However, the intercept increment for GDI is highly significant, indicating that two logarithmic trends exist
for GDI- and PFI-equipped vehicles. These trends have different intercepts but the same slope, i.e., they
are parallel, but with the GDI trend sitting higher. If the slope increment for GDI (1.1) is reverse
transformed, exp(l.l) = 3.00. This indicates that in this vehicle sample, the PM Phase-1 emissions are

-------
three times higher for GDI-equipped than for PFI-equipped vehicles over the measured temperature
range.

The overall conclusion from this analysis is that a single logarithmic slope term (or rate constant), as in
Equation 2-11 above, can be appropriately used as the basis for a temperature adjustment to represent
fleets including both PFI and GDI-equipped vehicles.

2.3. Temperature Effects on Running Exhaust Emissions from Gasoline
Vehicles

While MOVES is designed to model the impact of ambient temperature on running exhaust emissions,
current data suggests that there is little effect of temperature on THC, CO, NOx or PM. The sections
below discuss the relevant data and analysis for gaseous pollutants and for particulate matter.

2.3.1. THC, CO, and NOx Running Exhaust Temperature Effects

We examined the same data described above for starts to evaluate potential running temperature
effects. These test data suggest that there is very little effect of temperature on running emissions of
THC, CO, or NOx. Regression analyses found that the coefficients (slopes) were not statistically significant
(that is, the slopes were not distinguishable from zero). This contrasts with the significant temperature
effect in THC, CO, and NOx Bag 2 of the Kansas City Light-Duty Vehicle Emissions Study (KCVES) with
higher emissions at colder temperatures.17 As discussed for PM emissions in the next subsection, we
attribute the temperature effect on THC, CO, and NOx emissions observed in the KCVES to the short
duration and mild acceleration of Bag 1 of the LA-92 driving cycle, such that the vehicles had not fully
reached hot-stabilized condition by the beginning of Bag 2.

As an additional test, we examined a set of continuous data collected on the IM240 cycle in the Chicago
l/M program. To avoid potential confounding due to variable levels of conditioning vehicles experienced
in the queues at the l/M stations, we only used the second IM240s when back-to-back IM240s were
performed, and for single IM240s, we examined only the final 120 seconds of full duration IM240s.

Based on this analysis, we found no evidence of a temperature effect for THC, CO, and NOx between 5
and 95°F.

Because most of the data sets evaluated did not find a significant temperature effect, and the
temperature effect observed in the KCVES is attributed to the test conditions not achieving hot-stabilized
running conditions, we do not model temperature effects for THC, CO, and NOx in MOVES for running
exhaust for all gasoline vehicles. In MOVES, these effects are coded using polynomial functions as
multiplicative adjustments. Therefore, in MOVES, we set all of those adjustments equal to 1.0.

2.3.2. PM2.5 Running Exhaust Temperature Effects

The initial analysis of the Kansas City Light-Duty Vehicle Emissions Study (KCVES) data1516 indicated that
significant ambient temperature effects existed for both start (Bagl-Bag3) and running (Bag 2) PM
emissions on the LA-92 cyclec. Thus, MOVES2010 applied a temperature effect for running emissions for

cThe temperature effects in MOVES2010 and MOVES2014 for pre-2004 vehicles were substantial.
Emissions increased by a factor of 10 between ambient temperatures of 72°F and 0°F.

-------
all model year vehicles based on the Bag 2 measurements from paired vehicles tests conducted in the
winter and summer of the KCVES.

For MOVES2014, we updated the PM temperature effect for running emissions for Tier 2 and later model
year vehicles (2004+) based on data from the 2012 Cold Temperature Program (documented in Appendix
B). Experimental data collected in the 2012 OTAQ program involved measurement of PM emissions on
both the FTP (by phase) and the US06 cycles at temperatures of 0, 20 and 75°F of Tier 2 and MSAT-2-
compliant vehicles and PFI and GDI (See Appendix B). The results from these programs are plotted
against temperature in Figure 2-11. We also fit log-linear models to the data and found the effect of
temperature was not statistically significant on either cycle. Based on these results, we removed the
temperature effect for Tier 2 vehicles (model year 2004 and later) in MOVES2014.

0 .04 -

. .0.03 "

~ 0.02 -
£

0 01 -

Figure 2-11. Hot-running PM Emissions Measured on Two Cycles (FTP Bag 2, US06) on MSAT-2 Compliant MY

2010 Gasoline Vehicles, Reported as Grams/cycle

These results contrast with the significant PM running temperature effect detected for Bag 2 emissions
in the KCVES. Upon further analysis of the PM emissions from the KCVES study, we determined that
muchd of the temperature effect observed in the KCVES Bag 2 emissions was due to the short duration
and relatively mild accelerations of the cold-start phase of the LA92 cycle, which is only 310 sec (1.18 mi)
in length. We note that the PM temperature effect was much larger at the beginning of Bag 2 than at the

FTP

US06

4s ^ ^
~

75 0

Temperature (degF)

d We believe that the small, but statistically significant temperature effect that persists at the end of Bag
2, even after 1,025 seconds (17 minutes) of operation on the LA-92 in KCVES may be an artifact of this
particular study, because this persistent temperature effect on hot-stabilized running emissions was not
observed in other studies.

-------
end. In contrast, the cold-start phase of the FTP, used in the Cold Temperature Program is 505 seconds
(3.59 miles) in length.

For MOVES3, we conducted a literature review from other studies that measured particulate matter
emissions from gasoline vehicles including model years before 2004 at different ambient temperatures.
The results are summarized in Table 2-13.

Table 2-13. Literature Review of Temperature Effects on Running PM2.5 emissions from Gasoline Vehicles

Study

Vehicles and Test conditions

Findings on PM2 5 emissions

Measurements of
Exhaust Particulate
Matter Emissions from

In-Use Light-Duty
Motor Vehicles in the
Denver, Colorado
Area20 21

71 light-duty gasoline vehicles from model year
1970 to 1996 tested in the summer of 1996 and
winter of 1997 on a chassis dynamometer using
bag 2 of the FTP driving schedule.

Linear mixed model was fit and no
significant temperature effect was
observe.

Comprehensive particle
characterization of
modern gasoline and
diesel passenger cars at
low ambient
temperature22

Two Euro-3 (apply to 2000 -2004 model year
vehicles) port-injection gasoline vehicles
(Renault Megane and Alfa 406 TS)
Tested +23, -7 and -20 °C on a chassis
dynamometer on the common Artemis driving
cycle (CADC), after warmed up on 50-minute
IUFC15 driving cycle.

No temperature effect observed on
running emissions.

Characterization of
Metals Emitted from
Motor Vehicles23

Emission rates derived from PM2 5
concentrations measured at the entrance and
exit concentration of the Howell tunnel in
Milwaukee, WI in the summer of 2000 and the
winter of 2000-2001. Light-duty vehicles
constituted between 90.6 percent to 93.9
percent of the vehicle fleet, with 6.1 percent to

9.4 percent heavy-duty trucks.
Chemical mass balance methods were used to
estimate the contribution of tunnel emission
rates to gasoline tailpipe, diesel tailpipe, brake
wear, resuspended road dust, and tire wear
emissions.

Carbonaceous PM2 5 (EC+OC)
emission rates (mg/km) were
significantly lower (49-51 percent)
in the winter than the summer.
Gasoline tailpipe emissions are
estimated to be the largest
contributors to EC and OC
emissions3; more than diesel
tailpipe, brake wear, and
resuspended road dust.
The winter tests had comparable or
larger PM measurements of
inorganic ions and metals
(including Na and CI) presumably
due to road salt in the winter.

a Gasoline tailpipe and tire wear are combined because they have similar source profiles. However, for the
pre-2004 model years covered here, gasoline tailpipe emissions in MOVES contribute a much larger
share of PM2 5 emission rates (and thus EC and OC) than brake wear emissions.24

The result of the literature review (Table 2-13) suggested no temperature effects on PM exhaust
emissions, even for model year vehicles similar to the years measured in the KCVES. Thus, we now
believe the significant running PM temperature effect in KCVES was an artifact of the measurement
conditions of the study, including the short Bag 1 of the LA-92 cycle. Therefore, starting with MOVE3, we
have removed the running temperature effect for exhaust particulate matter emissions for all model year
light-duty gasoline vehicles.

-------
2.4. Temperature Effects on Diesel Vehicles

With the exception of projections for 2027 and later HD NOx effects (see Section 2.4.1.3), the data used
to evaluate and estimate temperature effects on diesel vehicles were limited to laboratory tests on pre-
2007 model year light-duty diesel vehicles. From this analysis, MOVES models a temperature effect only
for THC start emissions. The THC start temperature effect estimated from the light-duty diesel was
applied to all model year diesel vehicles in MOVES, including heavy-duty diesel vehicles. None of the
other pollutants in MOVES have temperature effects for diesel start emissions and MOVES has no
temperature adjustments for running emissions.

As described below, we reviewed more recent studies conducted on modern diesel and heavy-duty
diesel vehicles, but additional temperature effects data for US light-duty and heavy-duty diesel are
needed to fully evaluate the values now in MOVES.

2.4.1. THC, CO, and NOx Temperature Effects for pre-2027 Diesel Vehicles

For the development of the original diesel temperature effects in MOVES, we were able to identify only
12 diesel vehicles tested on FTP at multiple temperatures (9 passenger cars and 3 light-duty trucks).
However, only two of those 12 vehicles were tested at temperatures within the normal FTP range (68° to
86° F). None of these diesel trucks were equipped with aftertreatment devices.

2.4.1.1. Diesel Start Effects

The average start (Bag-1 minus Bag-3) emissions for those tests are shown in Table 2-14. We stratified
the test results into four temperature bands which yielded the following emission values (grams per
start) and average temperature value:

Table 2-14 Average Light-duty Diesel Vehicle Incremental Start Emissions (Bag 1- Bag3) by Temperature (grams

per start)

Temperature, F

Count

THC

NOx

34.6

2.55

2.44

2.6

43.4

2.68

2.03

0.32

61.5

1.69

0.67

69.2

1.2

1.91

0.36

Figure 2-12 shows the plot of mean THC start emissions versus temperature (where the vertical lines
represent 90 percent confidence intervals and the "dashed" line represents a linear regression through
the data).

-------
4

30 40 50 60 70

Temperature (degrees F)

Figure 2-12 Mean Light-duty Diesel Cold-start THC Emissions (in grams, shown on the y-axis) with 90 percent

Confidence Intervals vs Temperature

The dashed (blue) line in Figure 2-12 represents a linear regression line:

THC = (-0.04 * Temperature) + 4.22 R2 = 0.90 Equation 2-14

Transforming this equation into an equation that predicts the (additive) change/adjustment in the cold-
start THC emissions from light-duty diesel vehicles (in the MOVES format), we obtain:

THC additive temperature adjustment = A * (Temp. - 75)

Where: Equation 2-15

A = -0.04
Temp, is <75° F

The coefficient associated with this temperature adjustment term is statistically significant although its
coefficient of variation is relatively large (23 percent). We apply this adjustment to heavy-duty as well as
light-duty vehicles due to limited data on heavy-duty diesel starts.

The modified temperature adjustments for diesel THC emissions for starts with shorter soak times
(operating modes 101-107) are described in the MOVES heavy duty exhaust report.5

-------
On the other hand, the cold-start CO and NOx emissions did not exhibit a clear trend relative to the
ambient temperature. Plotting the mean CO and NOx cold-start emissions versus ambient temperature
(with 90 percent confidence intervals) produced the following two graphs:

1—~

30 40 50 60 70

Temperature (degrees F)

Figure 2-13 Mean Light-duty Diesel Cold-start CO Emissions (in grams) with 90 percent Confidence Intervals vs

Temperature

}

1—~—
1 ~

30 40 50 60 70

Temperature (degrees F)

Figure 2-14 Mean Light-duty Diesel Cold-start NO* Emissions (grams) with 90 percent Confidence Intervals vs

Temperature

-------
Statistical analyses of both the diesel cold-start CO and NOx emissions showed that the coefficients were
not significantly different from zero. Therefore, for both cold-start CO and NOx adjustments for diesel
vehicles, we set the temperature adjustment for start emissions to zero.

2.4.1.2. Diesel Running Effects

Since the diesel start temperature effects were either very small or zero, we did not evaluate the diesel
running temperature effect for THC, CO, and NOx for MOVES - we set temperature effects for diesel
running exhaust to zero, similar to the gasoline running exhaust adjustments. The exception is NOx
emissions for model year 2027 and later, as described below.

We are aware of studies suggesting that diesel NOx may be underestimated in current US emission
inventories during the wintertime25 and suggesting that there is an increase in heavy-duty diesel NOx
emissions at cold temperatures in the US.26,27,28,29 We will revisit the NOx temperature effects in MOVES
as more data on light-duty and heavy-duty diesels become available.

2.4.1.3. NOx Temperature Effects for III) Diesel Model Years 2027
and Later

Unlike earlier NOx standards, the HD2027 rule includes off-cycle standards that are a function of ambient
temperature; thus, MOVES incorporates cold temperature effects for NOxfrom heavy-duty diesel
vehicles of model year 2027 and later. This update was based on a 2022 testing program on a prototype
engine designed to meet the HD2027 emission standards.26 The testing was conducted using the CARB
Southern Route Cycle at laboratory temperatures (approximately 25°C) and with the ambient
temperature between 2 °C and 9 °C. The results from the testing showed that emissions were
approximately double at low ambient temperature versus standard laboratory temperature. The tests
showed that cold temperatures caused elevated NOx emissions at start and throughout the nearly 6-
hour test cycle.

These temperature effects were incorporated into the HD2027 off-cycle NOx standards as summarized in
Table 2-15.

Table 2-15 Temperature Adjustments to the Off-cycle NO* Standards in the HD2027 Rule
(§1036.104 Table 3 to Paragraph (a) (3) )27

Off-Cycle Bin

NOx Standard at 25 °C

Temperaturea-based Adjustment for NOx

Bin 1

10.0 g/hr

(25.0-Tamb)* 0.25

Bin 2

58 mg/hp-hr

(25.0-Tamb)* 2.2

aTamb is the mean temperature in °C over a shift day, or equivalent. The off-cycle NOx standard for Tamb below

25 °C is adjusted by adding the temperature adjustment to the specified NOx standard in g/hour for Bin 1 and

mg/hp-hr for Bin 2.

For MOVES, we used these values, combined with the temperature-independent duty-cycle standards, to
model effective NOx running and extended idle emission rates for each MOVES operating mode and all
relevant regulatory classes (42 thru 48) during in-use operations at both 25°C and 5°C. The details of the
HD2027 emission rate calculation process can be found in the MOVES heavy-duty exhaust emissions
report5

-------
Since MOVES applies the temperature adjustment after all operating mode detail has been aggregated
away, we calculated a nationally representative operating mode distribution for each regulatory class,
and derived a weighted average emission rate for each regulatory class at both 25°C and 5°C. From this,
we calculated a percent increase in NOx emissions per degree change in temperature. Since MOVES uses
the Fahrenheit scale, this was converted to a percent increase in grams of NOx per degree Fahrenheit
below 77°F.

The resulting multiplicative temperature adjustment varies by regulatory class and emissions process,
and is calculated for temperatures below 77°F as follows:

Adjustment = ((77.0 - temperature) x tempAdjustTermA) + 1 Equation 2-16

Table 2-16 shows the values of tempAdjustTermA used in the above equation, which are stored in the
TemperatureAdjustment table.

Table 2-16 NOxTemperature Adjustment Coefficients by Regulatory Class and Process

Process
(processID)

Regulatory Class
(regClassID)

tempAdjustTermA

Running (1)

LHD45 (42)

0.005139

MHD67 (46)

0.003957

HHD8 (47)

0.006352

Urban Bus (48)

0.008397

Extended Idle (90)

Doesn't matter (0)

0.01389

2.4.2. PM Temperature Effects for Diesel Vehicles

MOVES does not include any temperature effects for particulate matter emissions from diesel vehicles.
As presented in the previous section, hydrocarbon emissions from conventional diesel engines have
much lower temperature sensitivity than catalyst-controlled light-duty gasoline emissions. Limited data
exists on the ambient temperature effects of particulate matter emissions from diesel engines.

The EPA does not have data on PM start emissions on US-certified diesel vehicles tested across different
ambient temperatures. From a literature search, we were able to find two European test programs that
measured PM diesel start emissions from European light-duty diesel engines and vehicles at cold and
warm ambient temperatures.

Mathis et al. (2005)22 evaluated particle mass and number emissions from a conventional light-duty
diesel vehicle and a light-duty diesel equipped with a diesel particulate filter (DPF) at laboratory
conditions measured at +32, -7 and -2Cf C. Although the researchers observed an increasing trend in
particle mass emissions (g/start) from the conventional diesel vehicle at colder temperatures, over the
entire drive cycle, the particle number emission rates were not significantly impacted by the cold start
contribution. The particle mass emissions from the DPF-equipped vehicle were two orders of magnitude

-------
smaller than the conventional diesel engines, but the start contributed the majority of the particle
number emissions over the entire test cycle.

Sakunthalai et al. (201428) also reported significant increase in PM start emissions from a light-duty
diesel engine tested in a laboratory at +20 and -2Cf C. However, they only reported the PM mass
concentrations of the exhaust and not emission rates. Additionally, the engine was not equipped with an
emission control system. Other researchers have reported that PM emissions are larger at cold start than
hot start from diesel engines,3738 but have not investigated the relationship of cold starts with ambient
temperatures.

The reviewed studies suggest that temperature does influence cold start PM emissions from diesel
vehicles. However, at this time, MOVES does not include temperature adjustments to diesel start
emissions due to limited data on diesel engines and because diesel starts are a minor contributor to
particulate mass emissions in the mobile-source emission inventory. The diesel particulate matter
emission temperature effects in MOVES can be revisited in the future as additional data become
available.

2.5. Temperature Effects on Com pressed Natural Gas Vehicles

MOVES models emissions from heavy-duty vehicles running on compressed natural gas. However, at the
time the temperature corrections were developed, no data were available on temperature impacts for
compressed natural gas emissions. As discussed in the heavy-duty report,5 the start emissions for CNG
emissions for THC, CO, NOx and PM are set equal to diesel start emissions. Thus, we also applied the
diesel start temperature adjustments on THC emissions to CNG.

2.6. Temperature Effects on ICE Vehicle Energy Consumption

The temperature effects on energy consumption for internal combustion engine (ICE) vehicles in MOVES
have not been updated since MOVES2004. No temperature correction is applied to energy consumption
from running activity because the analysis documented in the MOVES2004 energy report29 found no
significant temperature effects for warmed-up vehicles. The same report also details the analysis used
to derive temperature effects on start energy consumption in MOVES. As presented in heavy-duty
report,5 the energy consumption from starts is a small fraction compared to the total energy use of both
gasoline and diesel vehicles. As such, we have not updated the start energy rates or temperature
adjustments in subsequent versions of MOVES.

In this section, we provide a summary of the start temperature effects on energy consumption in
MOVES. MOVES applies temperature adjustments to the start energy consumption through a
multiplicative adjustment. The form of the multiplicative adjustments used in MOVES is shown in
Equation 2-17, which is applied to all ambient temperatures. Unlike the temperature adjustments for
criteria pollutants, MOVES does not limit the energy consumption adjustments to only cold
temperatures, but also adjusts the energy consumption for hot temperatures. This ambient temperature
adjustment is separate from the air conditioning adjustment described in Section 4, below.

The multiplicative temperature adjustments are applied to all start operating modes of varying soak
lengths. MOVES does have different baseline (75°F) start energy consumption rates for different soak
times, which are documented with the baseline energy start rates in the MOVES Greenhouse Gas and
Energy report6 for light-duty vehicles and heavy-duty exhaust report.5

-------
Multiplicative temperature adjustment

= 1.0 + tempAdjustTermA X (temperature — 75) Equation 2-17

+ tempAdjustTermB x (temperature — 75)2

Table 2-17 displays the coefficients used to adjust start energy consumption for gasoline, E85, diesel and
CNG-fueled vehicles. The temperature coefficients are stored in the MOVES temperatureAdjustment
table by pollutant, emission process, fuel type, regulatory class, and model year range. E85-fueled
vehicles use the same energy adjustments as gasoline vehicles, because they also use the same energy
rates as comparable gasoline-fueled vehicles.6 CNG vehicles use the same adjustments as diesel vehicles,
because they use the same energy start rates as comparable diesel vehicles. The start energy coefficients
do not vary by regulatory class, so regClassID 0 ("doesn't matter") is assigned to these rows.

Table 2-17. Multiplicative Temperature Coefficients for Start Emissions Used in MOVES

tempAdjustTermA

tempAdjustTermB

Fuel types

Model Years

-0.01971

0.000219

Gasoline, E85

1950-2060

-0.0086724

0.00009636

Diesel, CNG

1950-2060

Figure 2-15 displays the multiplicative temperature adjustments for starts as a function of temperature.
At 75°F, the multiplicative adjustment is one. Gasoline vehicles have a larger temperature effect than
diesel vehicles, increasing to 4.8 at -20°F, while decreasing to 0.64 at 100°F. Whereas, the adjustment for
diesel vehicles only increases to 2.7 at -20°F and decreases to 0.85 at 100°F.

Figure 2-15. Multiplicative Temperature Adjustments for Starts from Energy Consumption as a Function of

Ambient Temperature

-------
2.7. Temperature Adjustments for Electric and Fuel-Cell Vehicles

Electric vehicles (EV) do not have exhaust emissions like internal combustion engines, but ambient
temperature has a large impact on their energy consumption. Energy consumption can increase due to
increased resistance in the drive train and electrical components, but the largest cause for the increase is
the use of heating and air conditioning.

Heating is particularly important to consider because EVs cannot scavenge waste heat from the engine
like internal combustion engine (ICE) vehicles can. As noted in the sections above, while MOVES does
estimate a cold temperature effect on energy consumption from ICE vehicle starts, no direct or cold-
weather temperature correction is applied to energy consumption from ICE running activity.

Because MOVES does not estimate energy consumption from starts for electric vehicles, there is no start
temperature effect on EV energy consumption.

This sub-section describes how we used the limited available data, as well as existing assumptions on ICE
temperature and A/C corrections in MOVES, to develop the appropriate coefficients for the new EV
adjustments.

At a heat index above 67°F, MOVES3 and earlier versions of MOVES adjust energy consumption based on
ambient temperature via an air conditioning adjustment (see Section 4). The MOVES air conditioning
(A/C) adjustments are applied only to passenger cars, passenger trucks, and light commercial trucks. The
A/C adjustment algorithm is applied for these vehicle types regardless of fuel type. Therefore, the light-
duty EV source types only require a temperature adjustment for temperatures below 67°F. Because
heavy-duty EVs lack an A/C adjustment in MOVES, they require both a high and low temperature
adjustment for energy.

We use the temperatureAdjustment table in the MOVES default database to adjust EV energy
consumption based on ambient temperature. The adjustment is multiplicative, based on Equation 2-18.
This quadratic equation matches the basic form of many other MOVES temperature adjustments, such as
described in Equation 2-17, where temperature represents the ambient temperature in Fahrenheit.

Multiplicative temperature adjustment

= 1.0 + tempAdjustTermA X (temperature — 72) Equation 2-18

+ tempAdjustTermB x (temperature — 72)2

At Project Scale, the sign of the adjustment coefficients is flipped if the meanBaseRate value is negative;
this is a special case to ensure that regenerative braking on electric vehicles is not modeled as generating
more energy when electric heaters are running.

The primary data source for the EV temperature adjustments is an American Automobile Association
(AAA) study which tested several EV passenger cars on a chassis dynamometer at room temperature,
extreme cold (20°F), and extreme heat (95°F).30 Their testing included a 2018 BMW i3s, 2018 Chevrolet

-------
Bolt, 2018 Nissan Leaf, 2017 Tesla Model S, and a 2017 Volkswagen e-Golf. While all vehicles are
passenger cars, they cover a variety of heating and cooling technologies, including both heat pumps
(BMW i3s and Nissan Leaf) and resistive heaters (Chevrolet Bolt, Tesla Model S, and Volkswagen e-Golf).
All five vehicles were tested at all three temperatures, with the cabin temperature always set to maintain
72° F.

Unlike other potential data sources, the AAA study measures the influence of ambient temperature on
EV energy consumption directly through experimental design, rather than through real-world
observational data which can have several confounding factors. Therefore, we used the AAA study to
derive the exact temperature adjustment for EVs in MOVES. In Appendix D, we show that the
temperature adjustment calculated using the AAA study is broadly consistent with observational data.

Relative to room temperature, the AAA found a 39% reduction in miles per gallon equivalent (MPGe) at
20°F and a 17% reduction in MPGe at 95°F, corresponding to a 64% and 20% increase in energy
consumption, respectively. Using these changes in energy consumption, a set of linear equations can be
derived that allow us to calculate temperature adjustment term A and B for Equation 2-18. They are
0.00225 and 0.00028, respectively.

As noted above, passenger cars, passenger trucks, and light commercial trucks, are already subject to an
air conditioning adjustment in MOVES. Atypical A/C adjustment during a MOVES run is around 20%,
consistent with the AAA study results. To avoid double-counting, Equation 2-18 is applied only when the
air conditioning adjustment is not being used. The MOVES air conditioning activity demand function is
detailed in the MOVES Population and Activity Report.31 According to this function, 67°F is the minimum
heat index at which an A/C adjustment is applied. In MOVES, this value is hardcoded as the point above
which MOVES uses the A/C adjustment algorithm, and below which MOVES uses the temperature
adjustment algorithm to scale light-duty EV running energy consumption.

Aside from this exception for light-duty air conditioning, we assume the coefficients derived from the
AAA report are representative of all electric vehicles, including heavy-duty. Therefore, they are used for
every electric vehicle of every class and EV technology (fuel cell and battery electric). While the
adjustments were derived using only the AAA report, we analyzed the adjustments in relation to other
published studies and test programs to ensure that the temperature adjustment in MOVES is consistent
with many sources, including testing of heavy-duty vehicles. Appendix D evaluates this approach by
comparing the resulting energy consumption rates to data from independent studies and shows
reasonable agreement. As EV technologies continue to mature and as more temperature effect data
becomes available, we hope to revisit both the form and the coefficients for these adjustments.

2.8. Conclusions and Future Research

With improved calibration and temperature management, ambient temperatures generally have less
impact on emissions of newer vehicles than older ones but MOVES estimates temperature effects for
THC, CO, NOx and PM start emissions from gasoline vehicles, THC starts for diesel and CNG vehicles, NOx
running emissions for post-2027 heavy-duty diesel vehicles, and running energy consumption for electric
vehicles.

We recognize that additional data and analysis could improve the MOVES temperature effects.

Additional studies and analyses could include:

-------
• Evaluating the benefits of applying log-linear or other mathematical models for pre-MSAT2
gasoline vehicle THC & CO temperature effects and considering whether all temperature
effects could be multiplicative rather than using additive effects for THC/CO/NOx start
emissions.

• Investigating ambient temperature effects on cold start emissions at temperatures warmer
than IS F.

• Evaluating the interaction of ambient temperature effects and fuel effects.

• Evaluating the interaction of ambient temperature effects and emission control deterioration.

• Analyzing ambient temperature effects for modern (2007 and later) diesel vehicles from
recent studies, especially those equipped with emission control devices, including diesel
particulate filters (DPF) and selective reduction catalysts (SCR).

• Conducting studies of temperature effects in vehicles using alternative fuels such as
compressed natural gas and ethanol blends.

• Incorporating data on the impact of temperature effects on newer technology gasoline
vehicles, including Tier 3 gasoline direct injection, and dual port-fuel and direct injection,
stop-start technologies, battery electric vehicles and hybrid technologies.

• Analyzing the effect of temperature on other pollutants estimated in MOVES including
ammonia (NH3).

• Evaluating EV energy used to condition the battery at various ambient temperatures,
especially temperatures between 35 and 65°F which are most common in shoulder months.

• Evaluating different EV heating and cooling technologies (such as resistive heating and heat
pumps) and their efficiencies at various ambient temperatures.

• Evaluating energy used for EV heating and cooling in a wider range of vehicles, including
single-unit and combination trucks. For example, buses and cars need to maintain the climate
in close to the full volume of the vehicle, while combination trucks have a much smaller cabin
relative to their power requirements and may require a smaller multiplicative temperature
adjustment.

-------
3. Humidity Adjustments

Water in the ambient air cools the peak combustion temperature and lowers engine out NOx emissions.
We adjust for this when evaluating source data for MOVES. More specifically, the NOx exhaust emissions
data used to develop emission rates for MOVES are adjusted from actual measurement conditions to a
standard humidity; this includes the emissions data from the Evaluation Sample for the Denver
Metropolitan l/M Program used to develop NOx emission rates for MY 1990 and later gasoline vehicles4
and the emissions data from the Heavy-Duty Diesel In-Use Testing Dataset used to develop NOxemission
rates for MY 2010 and later heavy-duty diesel vehicles.5 At run time, these base NOx exhaust emission
rates are adjusted from the standard humidity level to the humidity conditions specified by the run spec
as described below.

3.1. Humidity Adjustment Equation

In MOVES, the base exhaust emission rates for NOx in all modes and all processes are multiplied by a
unitless humidity factor, K. This factor is calculated separately by fuel type, with diesel using one
equation and set of coefficients while gasoline, CNG, and E-85 use another equation and set of
coefficients.

The equations and coefficients for each fuel type are determined by the Code of Federal Regulations
(CFR). The diesel adjustment is based on Part 10 6 5 32 for heavy-duty in-use testing and the adjustment
for other fuel types is based on Part 86 33 for light-duty vehicle emissions testing. In each case, the
equation specified is the inverse of the adjustment specified in the CFR. This is because the CFR equation
is used to adjust emissions to a standard humidity level, while MOVES is taking base rates calculated at
the standard humidity level and adjusting them based on the humidity level in the run to calculate a real-
world emission rate. In MOVES4 and subsequent versions, the equations and coefficients were updated
to better represent this inverse relationship.

Table 3-1 shows the equation coefficients, bounding humidity levels, and humidity units used for each
adjustment, as represented in the noxHumidityAdjust table in the MOVES default database. If the
specific humidity input is outside the bounding humidity levels, the value of the limit is used to calculate
the adjustment. The adjustment for gasoline, CNG, and E-85 vehicles is shown in Equation 3-1 and the
adjustment for diesel vehicles is shown in Equation 3-2.

Table 3-1 NOx Humidity Adjustment Parameters for all Fuel Types

fuelTypelD

CFR Source

Adjustment
Equation Terms

Specific Humidity Bounds

Specific Humidity
Units

Lower
Bound

Upper
Bound

1 (Gasoline)

40 CFR
86.144-94

0.0329

3.00

17.71

grams of water / kg
of air

2 (Diesel)

40 CFR
1065.670

9.953

0.832

0.002

0.035

moles of water /
moles of air

3 (CNG)

40 CFR
86.144-94

0.0329

3.00

17.71

grams of water / kg
of air

5 (E85)

40 CFR
86.144-94

0.0329

3.00

17.71

grams of water / kg
of air

-------
K = 1 — humidityTermA * (specif icHumidity — 10.71)

Equation 3-1

K = Equation 3-2

(humidityTermA * xH20) + humidityTermB

MOVES only uses relative humidity as the input source for humidity, either by users or in the default
database via the zonemonthhour table. Appendix A6 shows how MOVES calculates specific humidity
based on relative humidity, ambient temperature, and barometric pressure.

3.2. Future Research

Future work could investigate whether the real-world emissions impact of humidity is similar to the
corrections developed from laboratory testing used in the Code of Federal Regulations. Additional work
could evaluate the emission impact of humidity on more recent gasoline, diesel and alternative-fueled
engines and consider whether modern engine calibration and emission control technologies impact the
humidity effect.

-------
4, Air Conditioning Adjustments

MOVES applies air conditioning adjustments to THC, CO, NOx and energy consumption from passenger
cars, passenger trucks and commercial light trucks. The air conditioning (A/C) effects described below
were originally derived for MOVES2010.

The air conditioning adjustment factors used in MOVES are based on data collected from light-duty
vehicles in a test procedure meant to simulate air conditioning emission response under extreme "real
world" ambient conditions. These factors predict emissions which would occur during full loading of the
air conditioning system and are then scaled down in MOVES according to the ambient conditions
specified in a modeling run. The second-by-second emission data were analyzed using the MOVES
methodology of binning the data according to vehicle characteristics (MOVES source bins) and vehicle
specific power bins (MOVES operating modes). The results of the analysis showed statistically significant
and consistent air conditioning effects for three types of operation (deceleration, idle and
cruise/acceleration) and the three primary exhaust pollutants (hydrocarbon, carbon monoxide and
nitrous oxides) and energy consumption. This section shows the results of the analysis for the air
conditioning adjustments used in MOVES for THC, CO, NOx and energy consumption. The impact of A/C
on particulate matter has not been evaluated for MOVES and therefore, MOVES currently has no air
conditioning effect for PM emissions.

The MOVES A/C adjustment varies by operating mode for total energy consumption and exhaust running
THC, CO and NOx emissions and applies only to passenger cars, passenger trucks and commercial light
trucks. The HD emission rates for conventional vehicles do not require explicit A/C adjustments because
they are based on real-world driving that includes A/C usage depending on ambient conditions when the
test was conducted. For example, the model year 2010 and later HD diesel energy rates are based on
manufacturer-run Heavy-Duty In-Use Testing (HDIUT) data.5 The impact of air conditioning usage on
energy consumption for heavy-duty electric and fuel-cell vehicles is handled as a temperature correction
as explained in Section 2.7.

4.1. Air Conditioning Effects Data

The data for the MOVES A/C Correction Factor (ACCF) was collected in 1997 and 1998 in specially
designed test programs. In the programs, the same set of vehicles were tested at standard FTP test
conditions (baseline) and at a nominal temperature of 95°F. Use of the same set of vehicles and test
cycles was intended to eliminate most of the vehicle and test procedure variabilities and highlight the
difference between a vehicle operating at extreme ambient conditions and at a baseline condition.

The data used to develop the MOVES ACCF consisted of emission results from 54 individual cars and light
trucks tested over a variety of test schedules. Overall, the database consisted of a total of 625 test cycles
and 1,440,571 seconds of emission, speed, and acceleration data. Because of the need to compute
vehicle specific power on a modal basis, only test results which consisted of second-by-second data were
used in the MOVES analysis. All second-by-second data were time-aligned and checked for errors.

The distribution of test vehicles by model year is shown in Table 4-1. Model years 1990 through 1999
were included. The data set consists of 30 cars and 24 light trucks. No test data were available on other
vehicle types (e.g., motorcycles or heavy-duty trucks). The individual test cycles on which the vehicles

-------
were run are shown with the test counts in Table 4-2. The data shows a balance between different test
cycles and cars and trucks. The individual vehicles are listed in Appendix C.

Only vehicles which were coded as having an emission test with the A/C system on were selected for this
analysis. The A/C On tests and the A/C Off (default for most EPA emission tests in general) were matched
by VIN, test schedule and EPA work assignment. The matching ensured that the same vehicles and test
schedules were contained in both the A/C On sample and the A/C Off sample.

Table 4-1 Distribution of test vehicles by Model Year

Model Year

Count

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

TOTAL

Table 4-2 summarizes the distribution of test-cycles analyzed. The test-cycles are defined in a MOBILE6
report.34

Table 4-2 Distribution of tests by test cycle

Schedule Name

Count

ART-AB

ART-CD

ART-EF

F505

FTP

FWY-AC

FWY-D

FWY-E

FWY-F

FWY-G

FWY-HI

LA4

LA92

LOCAL

NONFRW

NYCC

RAMP

ST01

TOTAL

625

-------
4.2. Air Conditioning Effects on Emissions and Energy

The data described above was then used to estimate factors to account for increases in emissions and
energy consumption with full loading of the air conditioning system. These factors are recorded for
running and extended idle emissions by sourcetype, pollutant and operating mode in the
fullACadjustment table of the MOVES database. Thus, the same effects are applied for all light-duty
fueltypes and model years.

4.2.1. Full A/C Adjustments for THC, CO and NOx Emissions

Average emissions for each pollutant (HC, CO and NOx) with and without A/C operation were computed
for each of the MOVES light-duty running operating modes as defined using vehicle specific power
(VSP).4 This resulted in 69 (23 VSP bins x 3 pollutants) pairs of emission averages. However, the trends
were erratic, and the results were generally not statistically significant. In addition, most of the high-
speed bins had little data. An analysis of cars versus light-duty trucks showed no statistical difference
between the two. To produce more consistent results, the individual VSP bins were consolidated to three
principal bins: Braking / Deceleration, Idle, and Cruise / Acceleration as defined in Table 4-3. These
consolidated operating mode bins are quite different in terms of engine operation and emissions
performance.

Full A/C adjustments were then generated by dividing the mean "With A/C" emission factor by the mean
"Without A/C" emission factor for each combination of consolidated operating mode and pollutant. Full
A/C adjustments are shown in Table 4-3. Measures of statistical uncertainty (coefficient of variation of
the mean) were also computed using the standard error of the mean. They are shown in Table 4-3 as
"Mean CV of CF."

A/C adjustments of less than or equal to one were found for the Braking / Deceleration mode for all
three pollutants. These were set to one for use in the MOVES model.

Table 4-3 Full air conditioning adjustments for THC, CO and NO*

Pollutant

Consolidated
Operating Mode

opModelDs

Full A/C CF

Mean CV ofCF

THC

Braking / Decel

1.0000

0.48582

THC

Idle

1.0796

0.74105

THC

Cruise / Accel

11-40

1.2316

0.33376

Braking / Decel

1.0000

0.31198

Idle

1.1337

0.77090

Cruise / Accel

11-40

2.1123

0.18849

NOx

Braking / Decel

1.0000

0.19366

NOx

Idle

6.2601

0.09108

NOx

Cruise / Accel

11-40

1.3808

0.10065

These adjustments are applied to passenger cars, passenger trucks and light commercial trucks only.

Note the higher air conditioning effect for NOx at idle. These results are consistent with those obtained
from Nam et al. (2000)35 who showed that at low load conditions, A/C greatly increased NOx emissions
due to reduced residual gas fractions in-cylinder.

-------
4.2.2. Full A/C Adjustments for Energy Consumption

The use of a vehicle's A/C system will often have a sizeable impact on the vehicle's energy consumption.
This was found statistically by analyzing the available second-by-second data on C02 and other gaseous
emissions and converting them to an energy basis using standard EPA vehicle fuel economy certification
equations. The vehicle emission data were binned by running operating mode and mean values were
computed. A separate analysis was done as a function of sourceBinID (combination of vehicle type, fuel
type and model year); however, the results were not statistically different across sourceBinID given the
relatively small sample sizes. As a result, the A/C adjustments for energy are a function only of running
operating mode. The resulting A/C adjustments are shown in Table 4-4.

Table 4-4 Full air conditioning adjustments for energy*

opModelD

A/C Factor

opModelD

A/C Factor

opModelD

A/C Factor

1.342

1.294

1.365

1.223

1.205

1.314

1.187

1.156

1.254

1.167

1.137

1.187

1.157

1.137

1.166

1.127

1.137

1.154

1.127

1.137

1.128

1.127

* These adjustments are applied to passenger cars, passenger trucks and light commercial trucks only.

Only very small amounts of data were available for operating modes 26 through 29 and 37 through 40.
As a result, the data from these bins was averaged together and binned into two groups. The resulting
group averages were used to fill the individual VSP bins. This averaging process has the effect of leveling
off the effect of A/C at higher power levels for an engine. This is an environmentally conservative
assumption since it is likely that the engine power devoted to an A/C compressor probably continues to
decline, sometimes to zero, as the overall power demand of the engine is increased.

Fuel economy and GHG regulations are expected to reduce energy consumption with air conditioning.
However, because, the MOVES A/C factors are multiplicative adjustments to the running energy rates, a
reduction in running energy rates also reduces energy consumption from air conditioning. In MOVES, we
project the light-duty A/C improvements of regulatory rules using the running energy rates as
documented in the MOVES Greenhouse Gas and Energy Consumption Rates Report.6

4.3. Adjustments to Air Conditioning Effects

In MOVES, the adjustments for each operating mode are weighted together by the operating mode
distribution calculated from the driving schedules used to represent the driving behavior of vehicles.
Average speed, road type and vehicle type will affect the operating mode distribution.

meanBaseRateACAdj = SUM (meanBaseRate*(fullACAdjustment-1.0)*opModeFraction)

Since not all vehicles are equipped with air conditioning and air conditioning is normally not on all the
time, the full air conditioning effect on emissions is adjusted before it is applied to the emission rate. The
adjustment account for (a) the fraction of vehicles in each model year that are equipped with air
conditioning, (b) the fraction of vehicles equipped with air conditioning of each age that have an

-------
operational air conditioning system and (c) the fraction of those vehicle owners who have air
conditioning available to them that will turn on the air conditioning based on the ambient temperature
and humidity (heat index36) of the air outside their vehicles. These MOVES defaults are documented in
the Population and Activity report.31 The fraction of vehicles equipped with air conditioning, the fraction
of operational air conditioning and the fraction of air conditioning use are used to adjust the amount of
"full" air conditioning that occurs in each hour of the day.

EmissionRate = (meanBaseRateACAdj *
ACPenetration*functioningACFraction*ACOnFraction) + meanBaseRate Equation 4-1

The air conditioning adjustment is applied to the emission rate after it has been adjusted for fuel effects.
At Project Scale, the sign of this adjustment is flipped if the meanBaseRate value is negative; this is a
special case to ensure that regenerative braking on electric vehicles is not modeled as generating more
energy when the air conditioning is running.

Air conditioners are also employed for defogging at all temperatures, particularly, at lower temperatures.
This secondary use of the A/C along with associated emission effects is not addressed in MOVES.

4.4. Conclusions and Future Research

MOVES applies air conditioning effects to emissions from passenger cars, passenger trucks and
commercial light trucks. The impact depends on pollutant, operating mode, ambient temperature and
humidity and the anticipated availability of air conditioning in the vehicle type, model year and age being
modeled.

There are a number of areas where our understanding of air conditioning impacts could be improved.
These include:

• Evaluation of the impact of air conditioning use on particulate matter emissions.

• Studies of air conditioning effects in a broader range of model years, particularly those with
the most recent emission control technologies.

• Evaluation of air conditioning effects in the highest operating mode bins.

• Updates to information on the fraction of vehicles equipped with air conditioning and their
malfunction rates.

-------
5. Inspection and Maintenance Programs

Inspection and Maintenance (l/M) programs are any state or locally mandated inspection of highway
motor vehicles intended to identify those vehicles most in need of emissions-related repair and require
repairs of those vehicles. MOVES3.1 and later MOVES versions model an l/M program reduction in
emissions of HC, CO and NOx for gasoline and flexible-fueled (E-85) vehicles less than 14,000 pounds
(regulatory classes 20, 30 & 41). MOVES does not model emission changes for programs that target
diesel or CNG vehicles, Class 4-or-higher heavy-duty vehicles. MOVES does not model an effect for
particulate matter.

There is great variation in how vehicles are selected for inclusion in l/M programs, how and when
vehicles are tested, and what happens when vehicles fail. MOVES is designed to take these variations
into the account when estimating the emission benefits of these programs.

This section describes the MOVES calculation of l/M benefits for exhaust emissions. The calculation of
l/M benefits for evaporative emissions is described in the MOVES Evaporative Emissions report.7

5.1. Overview of Exhaust Inspection & Maintenance in MOVES

MOVES uses a number of inputs to estimate the benefits of exhaust l/M programs.

The model starts with two sets of emission rates as a function of age, model year group and regulatory
class. The "mean base rate l/M" or "l/M rates" represent emissions for an area with a "reference l/M
program." The "mean base rate", or the "non-l/M rates" represent emissions in an area without l/M. The
reference l/M program is not the same as the l/M performance standard,37 but instead is a program
used as a data source in MOVES because it provides a large sample of consistent data covering many
years. The data analysis used to determine both the l/M and non-l/M rates is detailed in the MOVES
light-duty emission rate report.4 Both sets of rates are recorded in the emissionRateByAge table.

MOVES scales the emission rate between (or potentially beyond) the l/M and non-l/M rate using an "l/M
Factor" by source type that accounts for differences in l/M program design, including test type and
inspection frequency, as detailed in Section 5.2. The l/M Factor assumes full coverage and compliance.

The result is also modified by the l/M coverage table. For each county and calendar year, the table lists
the source types, pollutants and model years that are covered, and the compliance factor which adjusts
l/M benefit to account for covered vehicles that are not actually subject to the program, evade testing,
or have repairs waived. In MOVES, it is assumed that any repairs attempted on vehicles receiving
waivers are not effective and do not result in any reduced emissions.

Mathematically, the IM Factor for the program design and the Compliance Factor for the program
characteristics are combined into a single factor, "IMAdjustFract" as shown in Equation 5-1. The
Compliance Factor is entered in units of percent and is converted to a fraction.

IMAdjustFract = QMFactor * ComplianceFactor * 0.01) Equation 5-1

MOVES then estimates a net emission rate by weighing together the emission rate for the l/M reference
program and the non-l/M emission rate, using the IMAdjustFract.

-------
TargetRate = IMRate * IMAdjustFract
+NonIMRate * (1.0 — IMAdjustFract)

Equation 5-2

5.2. Development of MOVES I/M Factors

MOVES is designed to model the different effects of different l/M program designs, specifically different
test types and test frequencies. The relative effectiveness of the programs is input into MOVES as the
"l/M factor/' a value between zero and two, stored in the MOVES IMFactor table. It Is calculated with
Equation 5-3.

^ _ Ep ~ EnoniM Equation 5-3

Ejm Enon]M

Where:

Ep is the adjusted emission rate for a "target" l/M program,

Eim is the reference rate,

EnoniM is the non-l/M reference rate and

R is the l/M Factor; an aggregate adjustment representing the difference in average emission
rates between the target program and the reference program.

Depending on the value of R, Ep may be greater than EnoniM, fall between fnoniM and £|M, or be less than
£im- Thus, this framework can represent target programs as more effective or less effective than the
reference program. In MOVES, R is referred to as the "IMFactor."

For our initial version of MOVES (MOVES2010), EPA developed l/M factors based on the information
incorporated in MOBILE6.2.38 These factors have been carried into later versions of MOVES.
Mechanically, this step was achieved by running the MOBILE6.2 model about 10,000 times over a
complete range of pollutant-process combinations, inspection frequencies, calendar years, vehicle
types, test types, test standards, model year groups and ages. The mean emission results for each
combination were extracted from the output and used to compute estimated values for IMFactor.

The IMFactor table includes the following fields6:

• Pollutant / Process

e The IMFactor table also includes values for Test Standard "Heavy-Duty Diesel Vehicle Reflash", with
"continuous" frequency for other buses and long and short-haul combination trucks (sourcetypes 41, 61
and 62). These values were entered early in MOVES development but are never used. We intend to
delete them in a future MOVES version.

-------
» The IMFactor table has rows for HC, CO and NOx running and start emissions, as well as HC vapor
venting.

• Test Frequency

• Annual or biennial

• Test Standard

• See Table 5-1 below

• Source Type

» Passenger cars, passenger trucks, light commercial trucks, single-unit short-haul trucks and

motorhomes

» Fuel Type

» Only gasoline and gasoline/ethanol blend fuels are covered

• Model Year Group

• Age Group
» IMFactor

Table 5-1 MOVES l/M Test Standards

testStandardsID

testStandardsDesc

shortName

Unloaded Idle Test

Unloaded Idle

Two-mode, 2500 RPM/Idle Test

2500 RPM/Idle

Loaded / Idle Test

Loaded/Idle

ASM 2525 Phase-in Cutpoints

A2525 Phase

ASM 5015 Phase-in Cutpoints

A5015 Phase

ASM 2525/5015 Phase-in Cutpoints

A2525/5015 Phase

ASM 2525 Final Cutpoints

A2525 Final

ASM 5015 Final Cutpoints

A5015 Final

ASM 2525/5015 Final Cutpoints

A2525/5015 Final

IM240 Phase-in Cutpoints

IM240 Phase

IM240 Final Cutpoints

IM240 Final

Evaporative Gas Cap Check

Evp Cap

Evaporative System Pressure Check

Evp Pressure

Evaporative System OBD Check

Evp OBD

Evaporative Gas Cap and Pressure Check

Evp Cap, Prs

Evaporative Gas Cap and OBD Check

Evp Cap, OBD

Evaporative Pressure and OBD Check

Evp Prs, OBD

Evaporative Gas Cap, Pressure and OBD Check

Evp Cap, OBD, Prs

Exhaust OBD Check

Exhaust OBD

HDDV Engine Reflash Program

HDDV Reflash

The IMFactor value was computed for all reasonable combinations of the parameters listed in the
IMFactor table. MOBILE6.2 runs were done for each parameter combination (Target design, fp) and a set
of runs were done for the reference program (Reference design, EtM)- In these runs, the reference
program has inputs matching the Phoenix, Arizona l/M program during the time in which the data used
in the MOVES2010 emission rate development were collected (CY 1995-2005). The reference design
represents a biennial frequency with an exemption period for the four most recent model years. It uses
three different l/M test types (basic idle test for MY 1960-1980, transient tailpipe tests for MY 1981-1995

-------
(IM240, IM147) and OBD-II scans for MY 1996 and later). Each of these test types became the reference
for the respective model year groups.

The specific combinations of MOBILE6.2 runs performed are shown in Table 5-2 below. Each of these
runs represents a particular test type and test standard design. A set of these runs were done for each
calendar year 1990 through 2030, for cars, light trucks and heavy-duty gasoline vehicles and for
pollutants THC, CO and NOx.

The first four runs represent the Non-I/M reference and the three Arizona l/M references.

Table 5-2 MOBILE6.2 runs used to populate the MOVES l/M adjustment factor

RUN#

Description

Type

Non I/M Base

Non I/M Reference

IM240 Base (Biennial IM240/147)

I/M Reference

OBD Base (Biennial OBD Test)

I/M Reference

Basic Base (Loaded - Idle Test)

I/M Reference

Biennial - IM240 - Phase-in Cutpoints

Target I/M Design

Annual - IM240 - Phase-in Cutpoints

Target I/M Design

Biennial - IM240 - Final Cutpoints

Target I/M Design

Annual - IM240 - Final Cutpoints

Target I/M Design

Biennial - ASM 2525/5015 - Phase-in Cutpoints

Target I/M Design

Annual - ASM 2525/5015 - Phase-in Cutpoints

Target I/M Design

Biennial - ASM 2525/5015 - Final Cutpoints

Target I/M Design

Annual - ASM 2525/5015 - Final Cutpoints

Target I/M Design

Biennial - ASM 2525 - Phase-in Cutpoints

Target I/M Design

Annual - ASM 2525 - Phase-in Cutpoints

Target I/M Design

Biennial - ASM 2525 - Final Cutpoints

Target I/M Design

Annual - ASM 2525 - Final Cutpoints

Target I/M Design

Biennial - ASM 5015 - Phase-in Cutpoints

Target I/M Design

Annual - ASM 5015 - Phase-in Cutpoints

Target I/M Design

Biennial - ASM 5015 - Final Cutpoints

Target I/M Design

Annual - ASM 5015 - Final Cutpoints

Target I/M Design

Annual - OBD -

Target I/M Design

Annual - LOADED/IDLE

Target I/M Design

Biennial - IDLE

Target I/M Design

Annual - IDLE

Target I/M Design

Biennial - 2500/IDLE

Target I/M Design

Annual - 2500/IDLE

Target I/M Design

The MOBILE6.2 database output option was chosen for all runs. This step produced large sets of results
detailed by age, roadway type, and emission type. This output format necessitated additional processing
into composite running and start factors.

The IMFactor (R) was then calculated using the mean emission results from the target program, the l/M
reference program and the non-l/M reference using Equation 5-3.

-------
5.2.1. Inspection & Maintenance in MOBILE6

Because the IMFactors used in MOVES were generated with MOBILE6.2, it is useful to briefly review
MOBILE6 modeling of l/M. Readers interested in a more thorough treatment of the topic are
encouraged to review the relevant MOBILE6 documentation.38 39 40

The MOBILE6.2 model used a methodology that categorized vehicles according to emitter status (High
emitters and Normal emitters) and applied a linear growth model to project the fraction of the fleet that
progresses from the Normal emitter to the High emitter status as a function of age. Average emission
rates of High and Normal emitters were weighted using the High emitter fraction to produce an overall
average emission rate as a function of age, model year group and vehicle type. The emissions generated
represented the emissions of the fleet in the absence of l/M (the No l/M emission rate).

A similar approach was used to generate l/M emission rates. In this case, the initial starting point for the
function (where age=0) was the same as the No l/M case. However, the effects of l/M programs and
associated repairs were represented by reductions in the fraction of high emitters, which consequently
affected the average emission level of the fleet. We also modelled the re-introduction of high emitters in
the fleet due to deterioration of vehicle emission control systems after repairs. The underlying l/M and
non-l/M deterioration rates were assumed to be the same.

MOBILE6 modeled the non-l/M and l/M emission cases diverging from each other over time, with the
l/M rates being lower. The percentage difference between these two rates is often referred to as the
overall l/M reduction or l/M benefit.

The relative effectiveness of various l/M programs was modeled using "high emitter identification rates"
that varied by test type. Since we lacked new data for MOBILE6, the effectiveness of biennial programs
as compared to annual programs and the effectiveness of ASM tests relative to IM240 were calculated by
running MOBILE5. To determine the high emitter identification rates for the IM240 test, MOBILE6 relied
on a database of 910 results from 1981-and-later cars and trucks from EPA emission factor testing in Ann
Arbor, Indiana and Arizona in which vehicles were randomly recruited and tested on both a running LA4
test (derived from the FTP test) and the IM240 test. There was little data for OBD and the high emitter
identification rate for OBD testing was set at 85 percent.39

5.3. I/M Compliance Factors

While the IMFactor (R, Equation 5-3)) represents the theoretical effectiveness of a specific l/M program
design relative to the reference design, MOVES uses a "compliance factor" to account for l/M program
compliance rates, waiver rates, failure rates, and adjustments needed to account for the fraction of
vehicles within a source type that are covered by the l/M program (these last adjustments are referred
to as the "regulatory class coverage adjustment").

When modeling for state implementation plans or conformity determinations, EPA guidance
recommends that modelers review program descriptive parameters and enter compliance factors which
reflect current and expected future program operation.41

MOVES values of the l/M compliance factor (CF) are specific to individual programs. The compliance
factor is entered as a decimal number from 0 to 100 and represents the percentage of vehicles within a

-------
source type that actually receive the benefits of the program. The compliance factor is calculated as
shown in Equation 5-4.

CF = (CR * (1 -WR* FR) * RCCA) Equation 5-4

Where:

CF= Compliance factor
CR = Compliance rate
WR= Waiver rate
FR= Failure rate

RCCA = Regulatory class coverage adjustment

The MOVES Technical Guidance provides instructions for modelers on using l/M program data to
calculate each of these values and compute an appropriate compliance factor for use in MOVES.41

The default compliance rates in MOVES represent a mixture of state-submitted values and values carried
over from MOBILE6. State-submitted values may be based on historic information, including historic
regulatory class coverage. For values derived from MOBILE6, the MOBILE6 compliance rate, waiver rate
and effectiveness rate were used to determine the default MOVES l/M Compliance Factor. Equation 5-5
shows the relationship.

CF = M6ComplianceRate * M6EffectivenessRate Equation 5-5

* (1 — M6WaiverRate)

5.4. Default l/M Program Descriptions (IMC overage)

Information about which pollutant-processes are covered by l/M programs in various counties and
calendar years is listed in the MOVES database table IMCoverage. This coverage information may vary by
pollutant, process, county, year, sourcetype and fuel type. The table also lists the l/M compliance factors
described above.

The IMCoverage table includes the use of l/M program identifiers called IMProgramlDs. A particular
county will likely have several IMProgramlDs that reflect different test types, test standards or inspection
frequencies applied to different sourcetypes, model year groups or pollutant-process combinations. For
example, a county in calendar year 2007 may have an IMProgramlD=l that annually inspects pre-1981
model year cars using an Idle test and an IMProgramlD=2 that biennially inspects 1996 and later model
year light-trucks using an OBD-II test.

The IMCoverage table also shows other important l/M parameters for each IMProgramlD. These include
the relevant model year range (beginning and ending model year), the frequency of inspection (annual
or biennial), test type (Idle, IM240, ASM, OBD-II) and the test standard.

The structure of the IMCoverage table in the MOVES database is:

-------
• Pollutant / Process
» State / County

• Calendar Year

• Source Use Type

» Fuel Type (only gasoline and ethanol fuels)

• IMProgramID

» Beginning Model Year of Coverage
» Ending Model Year of Coverage
» Inspection Frequency (annual or biennial)

• l/M Test Standards (see Table 5-1)

» UselMyn

• Compliance Factor

The UselMyn toggle is a user feature that allows the user to completely disable the modeling of l/M for
one or more of the parameter combinations.

When modeling for regulatory purposes, it is expected that a state will enter their own set of program
descriptive parameters and compliance factors which reflect current and expected future program
operation. However, MOVES contains a set of l/M program descriptions for all calendar years intended to
reflect our best assessment of the programs in each state.

The data used to construct the default inputs for l/M programs before calendar year 2011 were taken
from MOBILE6.2 input files used in the National Mobile Inventory Model (NMIM) to compute the
National Emission Inventory of 2011. The MOBILE6 data fields listed in Table 5-3 were extracted and
processed into the various fields in the MOVES IMCoverage table for each state and county.

As seen in Table 5-3, MOBILE6.2 and MOVES do not have exactly compatible parameter definitions. The
MOBILE6.2 l/M Cutpoints data were used only to determine level of stringency of a state's IM240
program (if any). The MOBILE6.2 Test Type inputs provided a description of the specific l/M tests
performed by the state and test standards for the ASM and Basic l/M tests. The MOBILE6.2 inputs of
Grace Period and Model Year Range were used to determine the MOVES Beginning and Ending model
year data values for each l/M program. The MOBILE6.2 vehicle type input was mapped to the MOVES
sourcetype.

-------
Table 5-3 l/M Coverage Table Data from MOBILE6

MOBILE6 Data

MOVES I/M Coverage Parameter

Compliance Rate

Used in the MOVES Compliance Rate Calculation

I/M Cutpoints

Used to determine MOVES I/M Test Standards

Effectiveness Rate

Used in the MOVES Compliance Rate Calculation

Grace Period

Used in MOVES to Determine Beginning Model Year of Coverage

Model Year Range

Used in MOVES to Determine Ending Model Year of Coverage

Test Type

Used to determine MOVES I/M Test Type

Vehicle Type

Used to determine MOVES Sourcetype

Waiver Rate

Used in the MOVES Compliance Rate Calculation

For calendar year 2011 through 2013, the IMCoverage table default parameters were derived using the
IMCoverage tables from the county databases (CDBs) provided to EPA for the 2011 National Emission
Inventory (NEI) project42 (Versionl). These tables were available for review by states and updated as
needed. The l/M program descriptions were extracted from the CDBs and compiled in the default
IMCoverage table for calendar year 2011. The l/M descriptions for 2012 and 2013 calendar years were
derived from the 2011 l/M descriptions, assuming no changes in the basic l/M program design; however,
the model year coverage values were updated to properly account for the existing grace periods in the
future calendar years.

The calendar year 2014 and later values were initially derived from the 2014 NEI (Version l)43 CDBs
following review by the states, with the 2015 and later calendar year values computed assuming no
changes in the basic 2014 l/M program design but updating the model year coverage values to properly
account for the existing grace periods in the future calendar years. All of the l/M program descriptions
were checked using a script to look for cases where a model year coverage either conflicted with other
rows in the l/M description or where gaps without coverage were left between model years. This check
also looked for cases where the coverage beginning model year occurred later than the ending model
year coverage. Each problem identified was compared to the l/M program descriptions found in the
2013 EPA l/M Program Data, Cost and Design Information report44 to resolve conflicts. The county
coverage values in some states were also updated for some calendar years. In addition to the updates in
the l/M program descriptions, the table was updated to make sure each l/M program covered E85-
fueled vehicles in the same way as gasoline in all calendar years. Any program elements claiming
benefits for inspections to reduce liquid fuel leaks (pollutant process ID 113) were dropped from the
default l/M program descriptions. MOVES does not offer any benefits from inspection programs to
detect liquid fuel leaks.

For MOVES3, the table was further updated based on state supplied data through the OBD
Clearinghouse website45 and 2017 National Emissions Inventory (NEI).46 The updates include adding l/M
programs for Ascension Parish, Iberville Parish, and Livingston Parish in Louisiana; for Hamilton County,
Tennessee, and for Cache County, Utah. We also updated the program stop years for terminated l/M
programs. Terminated programs include programs in Anchorage Borough, Alaska; Grundy County,

Illinois; Clark County and Floyd County, Indiana; Shelby County, Tennessee; and seven counties in
Minnesota, 26 counties in North Carolina, and six counties in Ohio.

-------
We also deleted the l/M program for Harrison County, Indiana (for all the CY years), since it was
confirmed that Harrison County, IN has never been in nonattainment for any National Ambient Air
Quality Standards (NAAQS) and does not have a l/M program. We also updated the beginning model
year for North Carolina l/M counties to reflect changes to their program for 2020 and later.47 In
addition, to reflect the termination of l/M program in Washington state, l/M programs have been
removed from IMCoverage table for all counties in Washington state after CY2019.

California currently has three different l/M programs: an enhanced program, basic program, and
ownership change program. These may vary by zip code within a county; however, MOVES lacks this
specificity. We mapped California counties with l/M program types by checking all the zip codes in each
county. We use the basic program to represent a county if it has mixed programs. This methodology is
consistent with previous work. We updated l/M program details for ten counties in California based on
our research.

In MOVES3.0.4, we updated compliance factors using data from the 2020 National Emissions Inventory
for existing IM programs that match the description in the default database, for year 2020 and after. We
also used the 2020 NEI information to update Cache County, UT for calendar year 2020 and beyond. We
removed l/M program information for Montgomery County, OH for 2020 and beyond, and removed
programs for all counties in Tennessee starting with calendar year 2023. In MOVES4, we further updated
information for Montgomery Co, OH for historical years, to reflect that the county had an active l/M
program only between 1990 and 2008.

MOVES5 incorporated new l/M information obtained via EPA's 2022vl Emissions Modeling Platform or
through communications with states. New York and Colorado provided information to update all
counties within their l/M area starting in calendar year 2022. Georgia changed their evaporative testing
from gas cap evap test (testStandardID 45) to evaporative OBD check (testStandardID 43) starting in
2025, and Idaho ceased all l/M operations in 2023, which we reflected by removing all l/M information
from calendar year 2024 and forward. Finally, Delaware updated their l/M program and we have
updated our defaults to incorporate the information submitted to us by the State with modifications
starting on calendar year 2023.

Table 5-4 shows the states with l/M program descriptions in the MOVES5 l/M coverage table and shows
the number of counties covered by the programs by calendar year. For example Indiana has four
counties with l/M information; two counties were under a program that was active between 1990 and
2007, while the other two counties are under a program that covers 1990 to 2060.

-------
Table 5-4 States With l/M Programs as Listed in MOVES

Calendar Years

State

StatelD

Minimum

Maximu
m

Counties

Alaska

1990

2009

1990

2012

Arizona

1990

2060

California

1990

2060

1999

2060

Colorado

1990

2060

2015

2060

Connecticut

1999

2060

Delaware

1990

2060

District of
Columbia

1990

2060

Georgia

1999

2060

Idaho

1990

2023

2011

2023

Illinois

1990

2060

1990

2005

Indiana

1990

2007

1990

2060

Kentucky

1990

2005

Louisiana

2000

2060

Maine

1990

2060

Maryland

1990

2060

Massachusetts

1990

2060

Minnesota

1990

1999

Missouri

1990

2060

Nevada

1990

2060

New Hampshire

2002

2060

2011

2060

New Jersey

1990

2060

New Mexico

1990

2060

New York

1990

2060

2001

2060

North Carolina

2003

2060

2006

2060

2003

2005

2003

2018

2006

2018

-------

Calendar Years

State

StatelD

Minimum

Maximum

Counties

Ohio

1990

2008

1990

2060

Oregon

1990

2060

2001

2060

Pennsylvania

1990

2060

2001

2060

Rhode Island

2000

2060

Tennessee

1990

2016

1990

2022

Texas

1990

2060

2000

2060

2011

2060

Utah

1990

2060

2020

2060

Vermont

1990

2060

Virginia

1990

2060

Washington

1990

2019

Wisconsin

1999

2060

5.5. Future Research

For thoughts on potential improvements to the MOVES l/M and non-l/M rates, see the MOVES light-duty
report where the calculation of MOVES current rates is explained in detail.4

Values for IMFactor are generally based on analysis for MOBILE6 or earlier and should be updated to
reflect current vehicle technology and testing practices and to better correspond to the current l/M
reference program. An IMFactor update is particularly needed for OBD which is commonplace now but
was in its infancy when the current MOVES values were developed.

While county modelers should always review the MOVES default IMCoverage table to assure values are
up-to-date for a given county,41 the default values could be improved with a systematic comparison to

-------
state and local l/M program records to assure that all the default values reflect the best information
about historical, current and future l/M coverage and compliance data.

In addition, the MOVES algorithm could be improved to allow l/M Coverage by regulatory class to better
match program design and the underlying MOVES emission rates. This would eliminate the need for the
regulatory class coverage adjustment in computation of the Compliance Factor.

Furthermore, there are vehicle inspection programs not currently modelled in MOVES, including
programs to reduce tampering and deterioration of heavy-duty diesel trucks, programs based on remote
sensing, and programs intended to reduce emissions of particulate matter. Expanding the scope of
MOVES to estimate the benefits of such additional programs would be useful for those considering such
programs. However, such expansion would require a significant and long-lasting investment in research
and analysis, as illustrated by the difficulty in collecting and updating data to support MOVES current l/M
algorithms.

-------
6. le Charging and Battery Efficiency

MOVES base energy consumption rates include the power needed at the wheels for each operating
mode plus energy losses through the drivetrain,6 but this does not account for an electric vehicle's total
demand on the electric power grid. By calculating total energy demand of vehicles on the grid, MOVES
can better facilitate the modeling of emissions from power plants and associated air quality changes/
This section details how MOVES accounts for charging and battery efficiency when estimating energy
consumption for electric vehicles (EVs).

For MOVES purposes, charging efficiency captures the energy lost in the wall charger - essentially the
difference between energy drawn from the wall outlet and energy added to the battery. Battery
efficiency, meanwhile, captures the relative energy lost in the battery itself - the difference between
energy produced at the output terminal and energy added to the battery. Each of these can range from 0
to 1, with higher values being more efficient.

While these efficiencies are related, they depend on different physical components that are engineered
independently, so their baseline efficiency and deterioration are likely to be different. MOVES models
them individually to account for these differences, but in practice, they are difficult to measure
separately. Most studies and lab data report them together in a measure we call "wall-to-output"
efficiency.

6.1. MOVES Design and Implementation

The table evEfficiency contains the charging and battery efficiency for electric vehicles. Similar to
emissionRateAdjustment,5 the values in this table are applied once the base rates have been calculated,
at the same time as other adjustments and corrections like those for ambient temperature (see Section
2.7. ).

MOVES models fuel cell vehicles as vehicles of the "electric" fuelType (fuelTypelD=9), but with a separate
engine technology type (engTechID = 40). However, a limitation of this approach is that when charging
and battery efficiencies are applied during MOVES runtime, the different EV technologies have already
been aggregated together to produce an average EV base energy consumption rate. Therefore, the
evEfficiency values implicitly apply to all electric vehicles, including fuel cells. This is not desired, because
fuel cell vehicles get their power from the fuel cell rather than the grid. Therefore, the fuel cell base
energy consumption rates in emissionRate were scaled down by the appropriate values in evEfficiency.
This ensures that the final energy consumption for fuel cell vehicles represents their actual operation,
after all adjustments are incorporated.

The evEfficiency table contains separate columns for battery and charging efficiency, with dimensions for
pollutant and emission process, source type, regulatory class, model year range, and age range. This
design provides maximum flexibility to improve the modeling of chargers and batteries in future versions
of MOVES, including by specific vehicle types (regulatory class) and vocations (source type). This

f Similarly, estimating energy consumption of internal combustion engines is useful for estimating the
emissions associated with the production and distribution of gasoline, diesel, and other combustion
fuels.

-------
flexibility can be used in future MOVES versions to model the impact different driving behaviors,
charging behaviors, and drivetrain configurations have on overall EV efficiency.

The adjustments are applied using Equation 6-1.

finalRate = baseRate Equation 6-1

(batteryEfficiency * chargingEfficiency)

Consistent with MOVES design for electric vehicles, the only pollutant and process in the table is total
energy consumption while the vehicle is running. In MOVES, all electric vehicles use the same
efficiencies and deterioration trends, regardless of source type, regulatory class, or model year due to a
lack of specific data pertaining to these fields. MOVES design allows more granular efficiency values by
source type, regulatory class, and model year, provided sufficient data becomes available.

6.2. Data Analysis and Literature Review
6.2.1. Charging Efficiency

Data on EV charging efficiency is limited, and the technology is evolving rapidly. Our primary data source
for charging efficiency is from the Altoona Bus Research and Test Center in the Penn State College of
Engineering.48 They tested battery electric buses from a variety of manufacturers and reported the
energy consumption of the bus on various drive cycles as well as the power drawn from the charger for
each test. From these, an overall wall-to-output efficiency can be calculated, which represents the
combination of charging efficiency and battery efficiency.

The wall-to-output efficiencies vary from approximately 75% to 91% as shown in Figure 6-1. However,
most buses, including the newer model years with better technologies, range from 85% to 91%. Most
data reported by Altoona as well as other sources contains wall-to-output efficiency and is not separated
by battery and charger efficiency. Therefore, we had to combine the Altoona data with a literature
review and engineering judgement to separate the battery and charging efficiency values in MOVES. We
assign new EVs a battery efficiency of 95% and a charging efficiency of 94%, which results in a wall-to-
output efficiency of 89.3%.

-------
100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

BYD MOTORS GILUG KIEPE NEW FLYER NOVA BUS PROTERRA

ELECTRIC

Figure 6-1. Average of wall-to-output efficiencies of electric buses tested by the Altoona Bus Research and Test
Center, grouped by manufacturer. Only buses with test reports that included both battery energy levels and total

charging energy consumption were included.

Our use of a 94% charging efficiency is informed by a number of factors in combination with engineering
judgement based on conversations within the MOVES team and with external experts, including those at
the Altoona Bus Research and Test Center. The chosen value of 94% is broadly consistent with a variety
of sources for heavy and light-duty vehicles. Tan, et al. (2014)49 show values ranging from 97% to 98.5%
and Kreiger and Arnold (2012)50 show values ranging from 85% to 95%. Both studies are modeling
studies, so we feel they are good confirmation of our efficiency values, but we chose to use observed
real-world data to calculate our charging efficiency adjustment. Apostolaki-losifidou, et al. (2017)51 show
values ranging from 85% to 98% based in part on observed data. This study contains detailed data, but
only for a single charging system and two vehicles, which we feel is adequate to help confirm our
adjustment in MOVES but not to calculate the adjustment.

The literature cited above doesn't report that charging efficiency changes with age, and discussions with
experts in the field, including the Altoona Bus Research and Test Center, indicate no physical reason to
expect a deterioration with age. Therefore, we assume there is no age trend for EV charging efficiency.

-------
6.2.2. Battery Efficiency

Battery efficiency, however, does deteriorate with age. Loss of EV range as battery ages is well
documented, but most studies focus on a loss of capacity. In theory, a loss of capacity can explain
reduced range without a drop in efficiency. We could not find any real-world data on the change in
battery efficiency with age. However, Yang et al (2018)52 modeled battery aging in typical driving
conditions in each U.S. state to cover a wide range of operational conditions. While their battery model
is based on batteries used in most passenger car EVs, the fundamental battery cell technology and
specifications are also commonly used in heavy-duty BEVs.

Yang et al. show that internal resistance increases as batteries age, which means that the reduction in EV
range with age can be attributed, at least in part, to a change in energy consumption. Energy
consumption is related to resistance, as shown in Equation 6-2 where R is internal resistance and e is
energy consumption:

. _ 1 Equation 6-2

Yang et al show that the average increase in energy consumption related to increased internal resistance
is 17.29% over 10 years. Starting with no increase in resistance for new vehicles, we linearly interpolate
between ages 1 and 10, binned according to MOVES age groups in the evEfficiency table. Assigning new
EVs a battery efficiency of 95% based on engineering judgement and our literature review,5
calculate the average efficiency for each age group using Equation 6-3.

batteryEfficiencyageGroup = °'95 Equation 6-3

-L ~r ageGroup

Because electric vehicles are a relatively new technology, there is considerable uncertainty about how
batteries age beyond 10 years. While some will continue to deteriorate, other vehicles may get efficiency
improvements or battery replacements under warranty. Electric vehicles have an ability to manage
battery degradation through software improvements as well, which may also limit battery aging.
Therefore, we assume overall battery efficiency doesn't deteriorate beyond the first 10 years. This
approach is similar to how we model criteria pollutant emission rate deterioration for ICE vehicles.

6.2.3. Conclusion

The resulting charging efficiency and battery efficiency values used in MOVES are in Table 6-1. We use
the same charging and battery efficiency assumptions for all electric vehicles, regardless of vehicle class
and model year.

-------
Table 6-1 Battery and Charging Efficiency by Ageg

Age Group

Battery Efficiency

Charging Efficiency

0-3 years

0.95

0.94

4-5 years

0.903153

0.94

6-7 years

0.874407

0.94

8-9 years

0.847435

0.94

10-14 years

0.828273

0.94

15-20 years

0.828273

0.94

20+ years

0.828273

0.94

gAs noted in Section 6.1, the current MOVES code requires application of these values to both BEVs and
FCEVs. Thus, the FCEV base energy consumption rates were adjusted ("back-calculated") to generate the
correct net energy consumption.

-------
i^vr eraging Provisions

Some EPA regulations allow manufacturers to meet emissions standards through what are known as
"averaging, banking and trading" (ABT) provisions. These provisions allow higher emissions from some
vehicles in return for lower emissions from others. When EV market share was low, MOVES did not
account for these details because it is designed to estimate emissions of "fleet-average" vehicles, rather
than individual vehicles or vehicle families. However, with growing EV sales, it is possible for
manufactures to sell ICE vehicles with greater emissions due to ABT provisions, so MOVES4 and later
MOVES versions have been updated to better account for this impact.

MOVES explicitly accounts for expected increases in the emissions and energy consumption from
conventional vehicles when national EV sales increase within any given model year. MOVES does not
explicitly model other ABT provisions, such as those that allow credits to be carried across several model
years, so we refer to this algorithm as the fleet averaging adjustment instead of an ABT adjustment.

The fleet averaging adjustment is a multiplicative factor applied to the base emission rates in MOVES. It
is calculated with the following equation:

Adjustment — evFraction X evMultiplier Equation 7-1

(1 — evFraction) + (evFraction x evMultiplier)

In the above equation:

evFraction is the national fraction of electric vehicles for a given model year, grouped by
vehicles that may be averaged together. Except when running with national preaggregation, this
calculation does not use user-supplied EV fractions in the AVFT table because compliance with
fleet-wide averaging is based on national sales rather than the local fraction of EVs.

evMultiplier is a multiplying factor applied to EV vehicles. This multiplier increases the
apparent number of total vehicles sold for the purposes of the adjustment calculation. The
values vary with model years as determined from EPA regulations, including the LD Tier 3 rule,54
the LD GHG Phase 2 rule,55 The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model
Years 2021-2026 Passenger Cars and Light Trucks,56 the Revised 2023 and Later Model Year Light-
Duty Vehicle Greenhouse Gas Emissions Standards,57 the Light- and Medium-Duty Multi-
Pollutant Rule (LMDV),S and EPA's Greenhouse Gas Emissions Standards for Heavy-Duty
Vehicles—Phase 3 (HDP3).9

For regulations that do not include an EV multiplier, the adjustment equation reduces to the following
form, obtained by inserting a value of 1 for evMultiplier:

Adjustment = Equation 7-2

1 — evFraction

In addition, some rules limit how much the effective ICE emission rate may increase due to the presence
of EVs in the fleet. For example, LMDV rule limits the emission bins that manufacturers may certify
vehicles to, and therefore we limit the ratio of the effective ICE emission rate to the fleet average
emission rate in these cases.

-------
The inputs needed for these calculations are stored in the MOVES default database's
FleetAvgAdjustment table, which lists the pollutants, emission processes, model year ranges, and
fleetAvgGrouplDs affected by the fleet averaging algorithm. The fleetAvgGroupID column, which is
defined by regulatory class, is used to group vehicles that may be averaged together. For example, LD
vehicles may be averaged together under Tier 3 (among other rules), so regulatory classes 20 and 30 are
both assigned fleetAvgGroupID 1. This table contains the following additional fields:

• "evMultiplier", which is described with Equation 7-2. Some pollutants and processes,
including criteria pollutants, do not have an EV multiplier. The evMultiplier is set to 1 in these
cases.

• "adjustmentCap", which can be used to limit the impact of the fleet averaging adjustments.

Some rules do not set an explicit limit to how much more ICE vehicles may emit when there
are significant numbers of EVs in the fleet, and so a value of "NULL" in this table represents
no adjustment limit. If a value is present in this table, it represents the upper limit to the ratio
of the effective ICE emission rate to the fleet average emission rate.

7.1. Fleet Averaging for Criteria Pollutants

7.1.1. Tier 3

Under the Tier 3 rule, fleet averaging provisions are relevant for the NOx+NMOG (non-methane organic
gases) exhaust emission standard, but do not apply to PM or CO. The rule allows averaging electric
vehicles exhaust emissions with other light-duty vehicles with a one-to-one weighting. Manufacturers
may average across cars and light trucks. Thus, the sale of battery electric vehicles (BEV) in the U.S. light-
duty fleet effectively increases the Tier 3 NOx+NMOG limit for internal combustion LD vehicles. Similarly,
the sale of BEV medium-duty vehicles (class 2b and 3 trucks) increases the Tier 3 NOx+NMOG limit for
internal combustion medium-duty vehicles.54

We assume manufacturers will take full advantage of these higher effective standards for ICE vehicles
because this allows them to reduce costs by applying measures such as installing simpler after-treatment
technologies on hybrid vehicles or reducing precious metal loading in catalytic converters. Alternately,
they may sell more vehicles in higher Tier 3 emission bins or sell credits to another manufacturer.

Since the rule allows averaging with one-to-one weighting between electric vehicles and internal
combustion vehicles, Tier 3 model years appear in the FleetAvgAdjustment table with an evMultiplier
value of 1 for running and start exhaust emission rates for NOx and THC. While the Tier 3 standard and
the fleet averaging provisions are for NMOG, we follow the general MOVES practice of modeling relative
changes in THC as proportional to changes in the NMOG standard.4 Additionally, because the fleet
averaging applies to light-duty vehicles as well as medium-duty vehicles, entries appear in
FleetAvgAdjustment for both fleetAvgGrouplDs 1 (light-duty vehicles) and 2 (medium-duty vehicles). Tier
3 does not provide an upper limit to specifically internal combustion NOx+NMOG emissions, so Tier 3
entries in the FleetAvgAdjustment table have a value of NULL for the adjustment cap.

Pollutants, processes, fleet average groups, and model years not listed in the table are not adjusted.

-------
7.1.2. LMDV rule

In MOVES, we model LMDV fleet averaging like the Tier 3 averaging described above. However, the
LMDV rule provides an upper limit for ICE emissions that we model in MOVES with an adjustment cap.
The adjustment caps were calculated as the ratio between the highest allowable certification bin and the
model year specific fleet average requirements. Since there are differences between the definition of
regulatory classes in MOVES and the grouping of vehicle classes used to define fleet requirements in the
rule, for regulatory class 30, we weighted the calculated adjustment caps by the fractional population of
LLDT and HLDT based on a historic national vehicle registration dataset, similar to the process described
in Section 3.14 of the LD report58. Further, since the modeling of fleet averaging in MOVES is applied to
regulatory class 20 and 30 as a group, we weighted the relative projected population of each regulatory
class for each model year following the information in the default sourcetypepopulation table. The final
cap values applied to light-duty and medium-duty vehicles are shown in Table 7-1. These adjustments
apply to running and start processes for THC and NOx.

Table 7-1 Adjustment caps developed for NMOG+NOx

modelYearlD

fleetAvgGroupID

Adjustment
Cap

2027

1 (Light Duty)

2.69

2028

1 (Light Duty)

2.88

2029

1 (Light Duty)

3.09

2030

1 (Light Duty)

3.92

2031

1 (Light Duty)

4.25

2031+

2 (Medium Duty)

2.27

2032+

1 (Light Duty)

4.67

There are no fleet averaging provisions for criteria pollutant emissions for heavy-duty vehicles.

7.2. Fleet Averaging for Hnergy Consumption and CO:

Many EPA GHG standards use "electric vehicle multipliers" as a way to promote EVs.8955 56 They also
allow credit trading between light-duty cars and trucks, as well as credit trading between medium-duty
vehicles.

Similar to how we model the impact for criteria pollutants, we assume the manufactures will take full
advantage of these EV-related benefits with regard to energy and C02 emissions.

The FleetAvgAdjustment table contains entries for running energy consumption for light-duty vehicles
with the evMultiplier values as shown in Table 7-2. The values vary with model years as determined by
EPA regulations, including the LD GHG Phase 2 rule,55 SAFE,56 and the revised 2023 and later standards.57

-------
Table 7-2 Light-duty Electric Vehicle Energy Adjustment Weights

Model Years

EV Multiplier

2017-2019

2.0

2020

1.75

2021

1.5

2022

1.0

2023-2024

1.3h

2025+

1.0

Fleet averaging for C02 emissions from medium-duty vehicles is covered by HD GHG Phase 2 for model
years 2021 through 2026. HD GHG Phase 2 applies different advanced technology credit multipliers to
sales based on technology. Credits for plug-in hybrid electric vehicles are multiplied by 3.5, battery
electric vehicles by 4.5, and fuel cell electric vehicles by 5.5.59 Because MOVES cannot differentiate by
engine technology at the point where this adjustment is applied, we can only apply a single value for the
evMultiplier for these vehicles. We chose to use the battery electric value of 4.5 because MOVES models
plug-in hybrid vehicles as internal combustion vehicles, and battery electric vehicles in this class are
more common than fuel cell electric vehicles. For model years 2027 to 2060, the evMultiplier for
medium-duty vehicles is reduced to one based on the LMDV rule8.

We do not account for the fleet averaging provisions applying to heavy-duty vehicles in HD GHG Phase 2
because its impact on ICE emissions is captured in other ways, including updates to roadload coefficients
and base emission rate adjustments. We account for fleet averaging for heavy-duty vehicles beginning in
model year 2028 based on the HD GHG Phase 3 standards.9 The Phase 3 rule begins in MY 2027 but
retains the advanced technology credit multipliers from Phase 2 for MY 2027 only. Beginning in MY 2028,
credits may be traded across vocational and tractor categories so long as they are in the same weight
class. The HD fleet average groups are summarized in Table 7-3. There are no advanced technology credit
multipliers in HD GHG Phase 3 after MY 2027, so the evMultiplier is set to 1 for each fleet average group.

Table 7-3 Heavy-duty Electric Vehicle Energy Adjustment Weights

MOVES Regulatory Classes

MOVES Fleet Average Group

evMultiplier

LHD45 (regClassID 42)

Light Heavy-Duty (fleetAvgGroupID 3)

MHD67 (regClassID 46)

Medium Heavy-Duty (fleet AvgGroupID 4)

HHD8 & Urban Bus (regClassIDs 47 and 48)

Heavy Heavy-Duty (fleetAvgGroupID 5)

To illustrate the impact of the fleet average adjustment, imagine a MY2024 fleet with 10 percent light-
duty EVs and an evMultiplier of 1.3. To compensate for the flexibility allowed in current regulations, the
average energy consumption rate for the ICE vehicles would be increased as shown in Equation 7-3
below.

hThis value was adjusted from the 1.5 value listed in Table 14 of the Federal Register / Vol. 86, No. 248
[Error! Bookmark not defined.] to account for the cumulative cap described in Section 1. ii.b, of the rule.

-------
72

-------
inces

1 USEPA (2020). Fuel Effects on Exhaust Emissions from On-road Vehicles in MOVES3. EPA-420-R-20-116.
Assessment and Standards Division. Office of Transportation and Air Quality. US Environmental
Protection Agency. Ann Arbor, Ml. 2020. https://www.epa.gov/moves/moves-onroad-technical-reports

2 USEPA (2023). Speciation of Total Organic Gas and Particulate Matter Emissions from Onroad Vehicles
in MOVES4. EPA-420-R-23-006. Office of Transportation and Air Quality. US Environmental Protection
Agency. Ann Arbor, Ml. August 2023. https://www.epa.gov/moves/moves-onroad-technical-reports.

3 USEPA (2020). Air Toxic Emissions from Onroad Vehicles in MOVES3. EPA-420-R-20-022. Office of
Transportation and Air Quality. US Environmental Protection Agency. Ann Arbor, Ml. November 2020.
https://www.epa.gov/moves/moves-onroad-technical-reports.

4 USEPA (2024). Exhaust Emission Rates for Light-Duty Onroad Vehicles in M0VES5. EPA-420-R-24-016.
Office of Transportation and Air Quality, Ann Arbor, Ml. November 2024.
https://www.epa.gov/moves/moves-onroad-technical-reports

5 USEPA (2024). Exhaust Emission Rates for Heavy-Duty Onroad Vehicles in M0VES5. EPA-420-R-24-015.
Office of Transportation and Air Quality, Ann Arbor, Ml. November 2024.
https://www.epa.gov/moves/moves-onroad-technical-reports

6 USEPA (2024). Greenhouse Gas and Energy Consumption Rates for Onroad Vehicles in M0VES5. EPA-
420-R-24-018. Office of Transportation and Air Quality, Ann Arbor, Ml. November 2024.
https://www.epa.gov/moves/moves-onroad-technical-reports

7 USEPA (2024). Evaporative Emissions from Onroad Vehicles in M0VES5. EPA-420-R-24-014. Office of
Transportation and Air Quality, Ann Arbor, Ml. November 2024. https://www.epa.gov/moves/moves-
onroad-technical-reports

8USEPA (2024). Multi-Pollutant Emissions Standards for Model Years 2027 and Later Light-Duty and
Medium-Duty Vehicles, 89 Fed. Reg. 27842, April 18, 2024.

9 USEPA (2024). Greenhouse Gas Emissions Standards for Heavy-Duty Vehicles—Phase 3, 89 Fed. Reg.
29440., April 22, 2024.

10 USEPA (2008). Mobile Source Observation Database (MSOD): User Guide and Reference. EPA-420-B-
08-017. Office of Transportation and Air Quality. US Environmental Protection Agency. Ann Arbor, Ml.
December, 2008. https://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=P10028RC.TXT

11 Stump, F. D., D. L. Dropkin, S. B. Tejada, C. Loomis and C. Pack (2002). Characterization of Emissions
from Malfunctioning Vehicles Fueled with Oxygenated Gasoline-Ethanol (E-10) Fuel — Part III. US EPA's
National Exposure Research Laboratory (NERL), U. S. Environmental Protection Agency, Washington, D.C.,
EPA Report Number EPA/600/R-01/053 (NTIS PB2004-106735), July 2002.

-------
12 USEPA (2007). "Regulatory Impact Analysis for Final Rule: Control of Hazardous Air Pollutants from
Mobile Sources" EPA Report Number EPA420-R-07-002, February 2007, Chapter 2, pages 2-15 to 2-17.
Available at: https://nepis.epa.gov/Exe/ZvPdf.cgi?Dockev=P1004LNN.PDF

13 USEPA (1994). Regulation of Fuels and Fuel Additives: Standards for Reformulated and Conventional
Gasoline 59 Federal Register: Issue 32. February 16, 1994, Docket Number: AMS-FRL-4817-8.

14 USEPA (2007). Control of Hazardous Air Pollutants From Mobile Sources 72 Federal Register
8428.February 26, 2007 . https://www.govinfo.gov/content/pkg/FR-2007-Q2-26/pdf/E7-2667.pdf

15 E. Glover and D. Brzezinski, (2001) Exhaust Emission Temperature Correction Factors for MOBILE6:
Adjustments for Engine Start and Running LA4 Emissions for Gasoline Vehicles, EPA Report Number EPA-
420-R-01-029 (M6.STE.004), April 2001.

16 Jose Pinheiro, Douglas Bates, Saikat DebRoy, Deepayan Sarkar and the R Development Core Team
(2013). nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-109.

17 USEPA (2008). Analysis of Particulate Matter Emissions from Light-Duty Gasoline Vehicles in Kansas
City, EPA Report Number EPA420-R-08-010, April 2008. Available at:
https://www.epa.gov/moves/mobile-source-emission-factors-research

18 Nam, E.; Kishan, S.; Bauldauf, R.; Fulper, C. R.; Sabisch, M.; Warila, J. Temperature Effects on Particulate
Matter Emissions from Light-Duty, Gasoline-Powered Motor Vehicles. Environ. Sci. Technol. 2010, 44,
4672-4677.

19 Sonntag, D. B., R. W. Baldauf, C. A. Yanca and C. R. Fulper (2013). Particulate matter speciation profiles
for light-duty gasoline vehicles in the United States. Journal of the Air & Waste Management Association,
64 (5), 529-545. DOI: 10.1080/10962247.2013.870096.

20 Cadle, S. H., et al. (1998). Measurements of Exhaust Particulate Matter Emissions from In-Use Light-
Duty Motor Vehicles in the Denver, Colorado Area. CRC Project E-24-1 CRC Project E-24-1. Coordinating
Research Council. March 24, 1998.

21 Cadle, S. H., et al. (1999). Composition of Light-Duty Motor Vehicle Exhaust Particulate Matter in the
Denver, Colorado Area. Environ Sci Technol, 33 (14), 2328-2339. DOI: 10.1021/es9810843.

22 Mathis, U., M. Mohr and A.-M. Forss (2005). Comprehensive particle characterization of modern
gasoline and diesel passenger cars at low ambient temperatures. Atmospheric Environment, 39 (1), 107-
117. DOI: http://dx.doi.Org/10.1016/j.atmosenv.2004.09.029.

23 Schauer, J., et al. (2006). Characterization of Metals Emitted from Motor Vehicles. Health Effects
Institute Research Report Number 133 Health Effects Institute Research Report Number 133.
https://www.healtheffects.org/system/files/Schauer.pdf

-------
24 Sonntag, D., et al. (2018). Modeling Brake and Tire Wear Emissions in Regulatory Models in the United
States. 2018 ISES-ISEE Joint Annual Meeting. Ottawa, Canada.

25 McDonald, B., et al. (2019). Quantifying Urban Emissions Influencing Wintertime Ammonium Nitrate
Formation. AQUARIUS (Air Quality in the Western US). Salt Lake City, UT.

26 USEPA (2022). Chapter 3 of Regulatory Impact Analysis for Control of Air Pollution from New Motor
Vehicles: Heavy-Duty Engine and Vehicle Standards, Office of Transportation and Air Quality. US
Environmental Protection Agency, Ann Arbor, Ml., December 2022. EPA-420-R-22-035
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P1016A9N.pdf

27 40 CFR 1036.104

28 Sakunthalai, R. A., et al. (2014). Impact of Cold Ambient Conditions on Cold Start and Idle Emissions
from Diesel Engines. SAE Technical Paper.

29 USEPA (2005). Energy and Emissions Inputs. EPA-420-P-05-003. Office of Transportation and Air
Quality. US Environmental Protection Agency. Ann Arbor, Ml. March, 2005.
https://www.epa.gov/moves/moves-onroad-technical-reports

30 American Automobile Association, Inc. (2019). AAA Electric Vehicle Range Testing. AAA proprietary
research into the effect of ambient temperature and HVAC use on driving range and MPGe. American
Automobile Association, Inc. https://www.aaa.com/AAA/common/AAR/files/AAA-Electric-Vehicle-Range-
Testing-Report.pdf

31 USEPA (2024). Population and Activity of Onroad Vehicles in MOVES5. EPA-420-R-24-019. Office of
Transportation and Air Quality, Ann Arbor, Ml. November 2024. https://www.epa.gov/moves/moves-
onroad-technical-reports

32 40 Code of Federal Regulations 1065.670 (a)(Page 230). Available online:

https://www.govinfo.gov/content/pkg/CFR-2017-title40-vol37/pdf/CFR-2017-title40-vol37-seclQ65-
670.pdf

33 40 Code of Federal Regulations 86.144-94(c)(7)(iv-viii)(Page 706). Available online:
https://www.govinfo.gov/content/pkg/CFR-2015-title40-voll9/pdf/CFR-2015-title40-voll9-sec86-144-
94.pdf

34 USEPA (2001). Air Conditioning Correction Factors in MOBILE6. EPA-420-R-01-055. Assessment and
Standards Division. Office of Transportation and Air Quality. US Environmental Protection Agency. Ann
Arbor, Ml. November 2011.

35 Nam, E. K., Understanding and Modeling NOx Emissions from Air Conditioned Automobiles, SAE
Technical Paper Series 2000-01-0858, 2000.

-------
36 National Oceanic and Atmospheric Administration (2014). The Heat Index Equation. Weather
Prediction Center. National Weather Service
http://www.wpc.ncep.noaa.gov/html/heatindex equation.shtml

37 USEPA(2022) Performance Standard Modeling for New and Existing Vehicle Inspection and
Maintenance (l/M) Programs Using the MOVES Mobile Source Emissions Model, USEPA Office of
Transportation and Air Quality, Transportation and Climate Division. EPA Report Number EPA-420-B-22-
034 October 2022. Available at: https://www.epa.gov/state-and-local-transportation/vehicle-emissions-
inspection-and-maintenance-im-policv-and-technical

38 USEPA (2002) User's Guide to MOBILE6.1 and MOBILE6.2 Mobile Source Emission Factor Model,
USEPA Office of Transportation and Air Quality, Assessment and Standards Division. EPA Report Number
EPA-420-R-02-028 October 2002.

39USEPA (2002) MOBILE6 Inspection / Maintenance Benefit Methodology for 1981 through 1995 Model
Year Light Vehicles, USEPA Office of Transportation and Air Quality, Assessment and Standards Division.
EPA Report Number EPA420-R-02-014 (M6.IM.001) March 2002. Available at:
https://nepis.epa.gov/Exe/ZyPDF.cgi/P10022PN.PDF?Dockev=P10022PN.PDF

40 USEPA (2001) Determination of NOx and HC Basic Emission Rates, OBD and l/M Effects for Tier 1 and
later LDVs and LDTs Final Report M6.EXH.007, USEPA Office of Transportation and Air Quality,
Assessment and Standards Division. EPA Report Number EPA-420-R-01-056 (M6.EXH.007) November,
2001.

41 USEPA (2024). MOVES5 Technical Guidance: Using MOVES to Prepare Emission Inventories for State
Implementation Plans and Transportation Conformity, USEPA Office of Transportation and Air Quality,
Assessment and Standards Division. EPA Report Number EPA-420-B-24-043. November 2024.

42 USEPA (2011). 2011 National Emission Inventory - https://www.epa.gov/air-emissions-
inventories/national-emissions-inventorv-nei

43 USEPA (2016). 2014 National Emission Inventory, https://www.epa.gov/air-emissions-
inventories/national-emissions-inventorv-nei

44 USEPA (2013), l/M Program Data, Cost and Design Information, Final Report, Prepared by ERG for EPA,
Project No.: 0303.00.009.001, August 2, 2013.

45 OBD Clearinghouse (2019) "l/M Jurisdiction Report,"
https://www.obdclearinghouse.com/Jurisdiction/iurisdictionPDFs?

46 USEPA (2020). 2017 National Emission Inventory, https://www.epa.gov/air-emissions-
inventories/national-emissions-inventory-nei

-------
47 USEPA (2019), Air Plan Approval; NC: Revision to l/M Program & Update to Charlotte Maintenance
Plan for the 2008 8-Hour Ozone NAAQS, Federal Register/Vol. 84, No. 176/Wednesday, September 11,
2019/Rules and Regulations.

48 LTI Bus Research and Testing Center (2012-2020). Bus Testing Report.
https://www.altoonabustest.psu.edu/bus-list.aspx

49 Tan, K., Yong, J., and Ramachandaramurthy, V. (2014). Bidirectional battery charger for electric vehicle.
2014 IEEE Innovative Smart Grid Technologies. DOI: 10.1109/ISGT-Asia.2014.6873826.

50 Elena M. Kreiger and Craig B. Arnold. (2012). Effects of undercharge and internal loss on the rate
dependence of battery storage efficiency. Journal of Power Sources 210 (2012) 286-291. DOI:
10.1016/j.jpowsour.2012.03.029

51 Apostolaki-losifidou, E., Codani, P., and Kempton, W. (2017). Measurement of power loss during
electric vehicle charging and discharging. Energy 127 (2017) 730-742.
https://doi.Org/10.1016/i.energy.2017.03.015

52 Yang, F., Xie, Y., Deng, Y., and Yuan, C. (2018). Predictive modeling of battery degradation and
greenhouse gas emissons from U.S. state-level electric vehicle operation. Nature Communications. DOI:
10.1038/s41467-018-04826-0.

53 Kostopoulos, E., Spyropoulos, G., Kaldellis, J. (2020). Real-world study for the optimal charging of
electric vehicles. Energy Reports, Vol. 6, 418-426. DOI: https://doi.Org/10.1016/i.egyr.2019.12.008.

54 40 CFR§ 86.1811-17

55 USEPA(2012), 2017 and Later Model Year Light-Duty Vehicle Greenhouse Gas Emissions and Corporate
Average Fuel Economy Standards, Federal Register, 77 FR 199.

56 USEPA(2020), The Safer Affordable Fuel-Efficient (SAFE) Vehicles Rule for Model Years 2021-2026
Passenger Cars and Light Trucks, Federal Register, Vol.85, No.84.

57 USEPA (2021) Revised 2023 and Later Model Year Light-Duty Vehicle Greenhouse Gas Emissions
Standards, 86 FR 74434, December 30, 2021.

58 USEPA (2024). Exhaust Emission Rates for Light-Duty Onroad Vehicles in MOVES5. EPA-420-R-24-016.
Office of Transportation and Air Quality, Ann Arbor, Ml. November 2024.
https://www.epa.gov/moves/moves-onroad-technical-reports

59 40 CFR 1037.150(p)

-------
Appendix * 'Vox ation of Temperature, Humidity and
Meteorology Calculations

The MOVES default database includes default ambient temperature and humidity values for every
county, month, and hour. If modelers do not enter local data, MOVES will use these values to calculate
the temperature and humidity adjustments described in the main body of this report. These values
were derived from 10-year average temperature and relative humidity values from calendar years 2001
through 2011 by month and by hour (standard time) for each county in the United States for all calendar
years.

Due to the limited number of hourly observation stations (about 200 sites), interpolation of the
available data was required. This interpolation will not always produce accurate results, particularly in
areas where climate can vary significantly over distance, such as in mountainous terrain and near
coastlines or deserts. Moreover, it is important that the diurnal range of the average hourly
temperatures match those of the average monthly minimum and maximum values. This aspect arises
due to the averaging process and to the fact that daily maximum and minimum temperatures do not
always occur at the same hourly observation time.

To correct the diurnal range problem, EPA has developed a method to adjust the average hourly
temperatures so that the corresponding hourly-based maximum and minimum temperatures match
those of the true monthly maximum and minimum values. To correct the spatial problem, all of the daily
and monthly maximum and minimum temperature observations made by the National Weather Service
(NWS) and its Cooperative Observation branch (over 6000 sites), and the Federal Aviation
Administration (FAA) are used.

Note, temperature and humidity data are one of the many inputs that are averaged for simplified
national and state level onroad MOVES runs. The algorithms for this averaging ("aggregation") are
described in the MOVES code documentation at https://github.com/USEPA/EPA MOVES Model.

A1. Data Sets and Quality Control

The National Climatic Data Center (NCDC) is the national and international depository for weather
observations. As part of its many duties, the NCDC publishes and maintains many climatic data sets.
"Quality Controlled Local Climatological Data" (QCLCD) files were obtained for all locations across the
United States, Puerto Rico and the Virgin Islands from the NCDC for this analysis.

There can be significant problems with this information. Primary among these problems is that many
stations with daily data do not have corresponding monthly averages, and vice-versa. Further, some
stations may have the same identification numbers while others may have missing or incorrect latitude
and longitude coordinates. During the processing of the 2009 data, nearly 10 percent of the 1654
stations were found to have identification and/or location problems.

-------
Missing monthly temperatures can be calculated from the daily maximum and minimum observations
for these stations for the years of interest. To resolve the mislabeled station IDs and location data, it was
necessary to contact NCDC to obtain updated tables with corrected IDs before processing the data.

In addition to the hourly temperature and dew point data, the identification number and geographic
location (latitude and longitude) for all available weather stations across the United States, Puerto Rico,
and the Virgin Islands were obtained from the NCDC files. Using Geographical Information System (GIS)
software, the locations of the hourly weather observation stations were validated. To resolve duplicate
IDs and latitude/longitude issues, careful analysis of the station history files and conversations with state
climatologists and National Weather Service offices were made. Our contractor, Air Improvement
Resource Inc. (AIR), hand-edited the IDs and latitude/longitude data and supplied updates to our data
and to the NCDC.

For temperature disputes, such as maximum temperature less than minimum temperature (caused by
mistyped data), hourly and/or daily data from nearby sites were consulted and the data corrected
accordingly.

For each station, an inventory was made as to the number of hours with joint temperature and dew
point data. In order to be included in the analysis, each station had to have at least 50 percent data
recovery for each hour of each month.

The daily absolute maximum and minimum temperature data for all available stations were processed
into monthly averages. These stations covered all classifications, including First-Order (National Weather
Service), Second-Order (both Automated Surface Observing System (ASOS) and Automated Weather
Observing System (AWOS)) and cooperative (local). Following NCDC guidelines, a month's averages were
considered valid when no more than 5 days had missing data during that month. For each station, the
hourly temperature and dew point data was scanned for missing values. For missing data periods lasting
only 1 hour, the missing values was replaced with an interpolated value from the two adjacent valid
readings

After these filters were applied, the average monthly maximum and minimum temperature data were
adjusted to the common midnight-to-midnight observational period. This adjustment is necessary since
many of the cooperative stations take their observations either early in the morning or late in the
afternoon rather than at midnight. These observation times induce a bias into the monthly temperature
averages. Correction values were obtained from the NCDC and applied to the monthly averages.

A2. County Temperature Assignment

An octal search with inverse distance weighting was used to assign the monthly maximum and minimum
temperatures to each U.S. County. Population centroids (latitude and longitude) for each county were
obtained from the 2010 United States Census. Population, rather than geographic, centroids were used
to provide a reasonable estimate of where the county's vehicle miles traveled and nonroad activity
would be concentrated. From each county's centroid, the distance and direction to each weather station
was calculated. The shortest distance was computed using the standard great circle navigation method
and the constant course direction was computed using the standard rhumb line method. A rhumb line
is a line on a sphere that cuts all meridians at the same angle; for example, the path taken by a ship or

-------
plane that maintains a constant compass direction. Based on the computed directions, the stations were
assigned to an octant, as follows:

• Octant 1: 0°
-------
Temph is the hourly temperature in the profile,

AMin is the average monthly minimum temperature,

AMax is the average monthly maximum temperature,

PMin is the minimum temperature based on the averaged 24 hourly temperatures in the profile¦,

PMax is the maximum temperature based on the averaged 24 hourly temperatures in the
profile.

After this adjustment is applied, the maximum and minimum of the adjusted hourly temperatures will
exactly match the average monthly maximum and minimum temperatures.

A4. Relative Humidity Recalculation

Relative humidity depends on both temperature and dew point. Unfortunately, unlike daily maximum
and minimum temperatures, supplemental dew point data is not available. Consequently, an
investigation and literature search were made to determine a suitable estimation method. Surprisingly,
few were found. The scheme outlined below was suggested by the NCDC and was used in this analysis:

At any given time, the difference between the temperature and dew point is known as the dew point
depression (DPD). Since the dew point can never exceed the temperature, the minimum DPD is zero
(100 percent relative humidity) while the maximum can be several tens of degrees, depending on how
dry the air is. From the original data, the DPD was computed at each hour.

After the hourly temperatures were adjusted to be consistent with the county minimum and maximum
temperatures as described above, the DPDs were subtracted from the hourly temperatures to estimate
the corresponding dew point. The corresponding relative humidity was then computed from these two
values. In keeping with standard meteorological practices, the relative humidity is always computed
with respect to water, even if the temperature is below freezing. Comparative tests showed that the
new calculated relative humidity results were very close to the original values, which is the desired
outcome.

AS. Calculation x Averages

The monthly average hourly temperatures for each county from each calendar year from 2001 through
2011 were averaged to determine the default 10-year average temperatures stored in the MOVES
ZoneMonthHour table for each county. The relative humidity values were converted to specific humidity
(humidity ratio) for each hour before averaging and then converted back to relative humidity.

-------
A6. Calculation of Specific Humidity

While the MOVES default humidity is stored as relative humidity, the humidity adjustment uses specific
humidity. The adjustment for diesel fuel type uses specific humidity expressed as a molar fraction, while
the adjustment for other fuel types uses specific humidity expressed as grams of water per kilogram of
air.

MOVES uses the following equations to calculate specific humidity based on pressure, relative humidity,
and ambient temperature.

Inputs:

Tf is the temperature in degrees Fahrenheit, TK is the temperature in degrees Kelvin
PB is the barometric pressure, in inches of mercury
Hrei is the relative humidity

First, MOVES calculates the vapor pressure of water at the saturation temperature in kPa.

H20

10.79574-11

(l-27^16)-5.02800-iog10(27^16)+1.50475-10~4^l-10 8'2969'(273.16 ^

[4.76955.(1\

+0.42873-10 -1 10l V TK )\_1 _o.2138602

= 10

Next, MOVES calculates the molar fraction of water in the air. This is the molar fraction used to calculate
the NOx adjustment for diesel vehicles.

(Hrel \ . p
„ _ U00~J Fh2°

H2° PB * 3.38639

Finally, MOVES calculates specific humidity in grams of water per kilogram of air using the following two
equations (1 inHg = 3.38639 kPa).

/Hrel \

PV(kPa) = (—J ¦ (PH2„)
specif icHumidity =

Hrel \

621.1 * PV

(PB * 3.38639) - PV

-------
Calculation of Heat Index

MOVES air conditioning demand is calculated as a function of the heat index as described in the MOVES
Population and Activity report.31 In MOVES, the heat index is a function of temperature and relative
humidity. For temperatures below 78° Fahrenheit, the heat index is equal to the temperature. For
temperatures above 78, the following equation (which is a simplification of the National Weather
Service heat index equation1) is used,

Heat Index = min ( ( -42.379 + 2.04901523T + 10.14333127H

- 0.22475541™ - 0.00683783r2 - 0.05481717H2
+ 0.00122874r2H + 0.0008528277/2

- 0.00000199T2H2), 120)

Where:

T= temperature
H = relative humidity
T>=78°F

i National Weather Service, Weather Prediction Center, The Heat Index Equation, May 2014.
https://www.wpc.ncep.noaa.gov/html/heatindex equation.shtml

-------
Appendix B OTAQ Light-duty gasoline 2012 Cold Temperature
Program

EPA's Office of Transportation and Air Quality (OTAQ) contracted the testing of nine Tier 2 vehicles (2006
and 2010 model year car and light-duty trucks). Eight of the nine vehicles were Mobile Source Air Toxics
(MSAT-2) rule compliant. Vehicles were tested on the FTP and US06 under controlled conditions 75, 20,
and O^F. Note: we excluded the two GDI vehicles (Cadillac STS and the VW Passat) from the estimation
of the THC and CO cold starts'5 as mentioned in Section 0.

Information on the tested vehicles is summarized in Table B-l.

Table B-l Vehicles Tested in 2012 Cold Temperature Study

Vehicle Name

Model Year

Injection

Emissions Std

MSAT?

Odometer

Displ (L)

Cyl.

Buick Lucerne*

2010

PFI

Tier 2/Bin 4

MSAT-2

22000

3.9

V-6

Honda Accord*

2010

PFI

Tier 2/Bin 5

MSAT-2

24000

2.4

1-4

Hyundai Sante Fe

2010

PFI

Tier 2/Bin 5

MSAT-2

18000

2.4

1-4

Jeep Patriot*

2010

PFI

Tier 2/Bin 5

MSAT-2

22000

1-4

Kia Forte EX*

2010

PFI

Tier 2/Bin 5

MSAT-2

25000

1-4

Mazda 6*

2010

PFI

Tier 2/Bin 5

MSAT-2

24000

2.5

1-4

Mitsubishi Gallant*

2010

PFI

Tier 2/Bin 5

MSAT-2

38000

2.4

1-4

Cadillac STS

2010

GDI

Tier 2/Bin 5

MSAT-2

21000

3.6

V-6

VW Passat

2006

GDI

Tier 2/Bin 5

pre-MSAT

103000

1-4

*Tested at 0 F

-------
85

Appendix < iVo^n^ouv M\ >is Vefaiclr Sample

Table C-l lists the vehicles in the test program.

Table C-l Vehicle Sample for the Air Conditioning Analysis

Model Year

Make

Model

Vehicle Class

Weight

1990

DODGE

DYNA

CAR

3625

1990

NISSAN

MAXIO

CAR

3375

1991

CHEVROLET

CAVA 0

CAR

2750

1991

FORD

ESCO GT

CAR

2625

1992

CHEVROLET

CAVA

CAR

3000

1992

CHEVROLET

LUMI

CAR

3375

1992

MAZDA

PROT

CAR

2750

1992

SATURN

CAR

2625

1992

TOYOTA

CORO

CAR

2500

1993

CHEVROLET

CORS

CAR

3000

1993

EAGLE

SUMMO

CAR

2500

1993

HONDA

ACCOO

CAR

3250

1993

TOYOTA

CAMRO

CAR

3250

1994

CHRYSLER

LHS

CAR

3750

1994

FORD

ESCO

CAR

2875

1994

HYUNDAI

ELAN

CAR

3000

1994

SATURN

CAR

2750

1995

BUICK

CENT

CAR

3995

1995

BUICK

REGA LIMI

CAR

3658

1995

FORD

ESCO

CAR

2849

1995

SATURN

CAR

2610

1995

SATURN

CAR

2581

1996

CHEVROLET

LUMI 0

CAR

3625

1996

HONDA

ACCO

CAR

3500

1996

HONDA

CIVI

CAR

2750

1996

PONTIAC

GRAN PRIX

CAR

3625

1996

TOYOTA

CAMR

CAR

3625

1997

FORD

TAUR

CAR

3650

1998

MERCURY

GRAN MARQ

CAR

4250

1998

TOYOTA

CAMR LE

CAR

3628

1990

JEEP

CHER

LDT1

3750

1990

PLYMOUTH

VOYA

LDT1

3375

1991

CHEVROLET

ASTRO

LDT1

4250

1991

PLYMOUTH

VOYA

LDT1

3750

1992

CHEVROLET

LUMI

LDT1

3875

1993

CHEVROLET

S10

LDT1

2875

1994

CHEVROLET

ASTR

LDT1

4750

1994

PONTIAC

TRAN

LDT1

4250

1996

FORD

EXPL

LDT1

4500

-------
1996

1990

1991

1994

1996

1997

1998

1999

Make

Model

Vehicle Class

FORD

RANG

LDT1

CHEVROLET

SURB

LDT2

FORD

E150 0

LDT2

FORD

F150

LDT2

FORD

F150

LDT2

DODGE

DAKO PICK

TRUCK

DODGE

D250 RAM

TRUCK

DODGE

GRAN CARA

TRUCK

DODGE

CARA

TRUCK

FORD

F150 PICK

TRUCK

DODGE

GRAN CARA

TRUCK

DODGE

DAKOT

TRUCK

PONTIAC

TRANSSPOR

TRUCK

DODGE

CARA GRAN

TRUCK

FORD

WIND

TRUCK

-------
Appendix I ) Consistency of MOVES EV Temperature Adjustment
with Sources

As explained in Section 2.7., MOVES applies a temperature adjustment to energy consumption from
electric vehicles. While the adjustments were derived using only values from the AAA report30, we
analyzed the adjustments in relation to other published studies and test programs to ensure that the
temperature adjustment in MOVES is consistent with many sources.

i11 V rth American Tramn finely

Henning, Thomas, and Smyth published a paper which included observational data from both battery
electric and fuel cell urban busesj. The data was collected by eight transportation agencies in North
America, ranging from California to Minnesota, meaning they were able to collect data at a wide range of
ambient temperatures. The data was collected at the daily level, comparing daily energy consumption,
daily mileage, and daily temperature.

This means their observed temperature effects are approximate and not experimentally derived.
Attributing average change in energy consumption versus ambient temperature is difficult because of a
number of confounding factors, the most important of which is the uncertainty introduced by daily
averaging. Over the course of a day, temperature can change by as much as 20-30 degrees Fahrenheit
and this is not reflected in the data. However, the data is still precise enough to provide a general
comparison to the existing MOVES temperature adjustment and confirm that the adjustment is not
fundamentally different for HD EVs compared to the passenger cars measured in the AAA study.

Their data shows a similar temperature impact for both fuel cell and battery electric EVs, with fuel cells
possibly having a smaller impact. Despite the uncertainties in the data, it is possible to calculate a more
precise temperature effect, but we believe the difference is small enough that creating additional
complexity in MOVES to apply different temperature adjustments for each engine technology is
unwarranted.

Henning, Thomas, and Smyth note a drop in MPGe with decreasing temperature. For battery electric
buses, the average MPGe drops from 18.8 at 65°F to 14.4 at 32°F. This corresponds to a 27% increase in
energy consumption, while the MOVES temperature adjustment estimates a 29% increase. Henning,
Thomas, and Smyth report an average increase of about 6% at higher temperatures (80-95°F), which is
smaller than the MOVES' high temperature adjustment and the AAA findings of a 20% increase, but
directionally consistent with the AAA finding of less impact at warm than at cold temperatures (20 -
32°F).

Table D-l shows Henning, Thomas, and Smyth's observed temperature impacts on fuel economy of both
fuel cell and battery electric buses.

j Henning, Mark; Thomas, Andrew R.; and Smyth, Alison, "An Analysis of the Association between
Changes in Ambient Temperature, Fuel Economy, and Vehicle Range for Battery Electric and Fuel Cell
Electric Buses" (2019). Urban Publications. 0 12 3 1630.
https://engagedscholarship.csuohio.edu/urban facpub/1630.

-------
Table D-l. Fuel Economy Reductions for EV Buses observed by Henning, Thomas, and Smyth

Ambient Temperature Range

Fuel Cell Reduction

Battery Electric Reduction

22-32 F

28.6%

32.1%

50-60 F

70 - 80 F

6.6%

6.4%

I >2 Japanese Passenger Car Study

In 2018, Liu et al collected GPS and real-world energy consumption data from 68 passenger car EVs being
driven across Japan, at a wide range of ambient temperaturesk. They used the data to fit an EV energy
consumption model based on several factors, of which two key variables are the ambient temperature
and accessory load usage, which are related.

Because the energy consumption model is calibrated based on real-world data and temperature is a key
component of the model, it can be used to attribute an increase in energy consumption to a change in
temperature. First, they show that a quadratic equation similar to Equation 2-16 is a good fit for their
data. Second, they show that their quadratic fit is close to the MOVES adjustments, although it is a bit
steeper (a doubling of energy consumption, relative to about 65°F, at about 23°F instead of MOVES'
estimated 8°F).

We did not use this paper as a direct source for a MOVES temperature adjustment for three reasons.

First, the EVs in the study were owned and operated in Japan, and therefore may not be representative
of the American fleet or American driver behavior. Second, the attribution of a change in energy
consumption to temperature is done via a calibrated model, and not direct measurement. The AAA study
source is a more direct observation of the effect of temperature on EV efficiency via controlled
experimental design, which is a better input to MOVES. Third, the paper does not provide enough data
to calculate a temperature effect at the level of precision required by MOVES. Nonetheless, we should
expect the temperature effect modeled by Liu et al. to be broadly consistent with other sources such as
MOVES, and it is.

D3. Canadian Passenges > Mady

Environment and Climate Change Canada (ECCC) performed on-road, real-world testing of a 2018
Chevrolet Bolt in January and July of 2019, collecting energy consumption data at a frequency of 2 Hz
from the battery terminal1. Their instrumentation was able to collect energy consumption of various
components as well, and they show that HVAC is the dominant factor increasing energy consumption at

k Liu, K., et al. (2018). Exploring the interactive effects of ambient temperature and vehicle auxiliary
loads on electric vehicle energy consumption. Applied Energy, 227, 324-331. DOI:
https://doi.Org/10.1016/i.apenergy.2017.08.074.

I Emissions Research and Measurement Section (Environment and Climate Change Canada) and
ecoTechnology for Vehicles Program (Transport Canada), Government of Canada.

-------
extreme temperatures. The vehicle was driven on similar routes each day of testing, and therefore the
results of several trips are directly comparable.

There were only eight days' worth of testing, and the temperatures tended to be either extreme cold
(below 35°F) or at room temperature and above. Therefore, it is not appropriate to fit a quadratic curve
to the data and calculate an exact temperature effect. However, Figure D-l shows that, given the
expected variance that exists between individual tests, the ECCC data generally agrees with the MOVES
adjustments, including the A/C adjustment and light-duty EV cold temperature adjustment.

Energy consumption rates for all full-day tests

compared to MOVES EV ambient temperature adjustments

300-

0
03

C 200
o

"•4—1'

E
d

(j)

° 100
a>

c
LD

20 40

Temperature (F)

MOVES Adjustment EV Component: | | HVAC/Accessory | | Axle/Traction ~ DCDC

Motor

Figure D-l Comparison of ECCC test data and MOVES EV energy consumption with temperature adjustment.

D4. Conclusion

Overall, the MOVES EV temperature adjustment algorithm is generally consistent with the limited
available real-world data on changes in total energy consumption with temperature for electric vehicles
of all classes and technologies.

-------
Appendix E Vehicles in the 2021 ORD Cold-temperature Program

The vehicles measured in this program, designated as ORD (2021) in Section 0, are described in
-------
Appendix F Model-Fitting Information for Analysis of Fuel-
Injection Technology

The following tables include additional model-fitting information for the for the model presented in
Table 2-12, on page 30.

Table F-l. Dimensions for the best-fit Model

Covariance parameters

Columns in X

Columns in Z per Subject

Subjects

Maximum observations per subject

The 'subjects' are the 12 vehicles. The total number of observations was 112.

The table below presents the 14 covariances associated with the 'random' component of the best-fit
model. These include variances for the random intercepts and slopes for the vehicle subjects, as well as
individual residual error variances for each vehicle.

-------
Table F-2. Covariance Parameters for the best-fit model

Parameter

Subject

Group

Estimate

Intercept {a1 bo)

vehicle

0.1028

Temperature (T) (c^bi)

vehicle

0.000079

Residual (o2E)

(ORD) Accord

0.03175

Residual (o2E)

(ORD) Fusion

0.006631

Residual (o2E)

(ORD)_Jetta

0.01786

Residual (o2E)

(OTAQ) Accord

0.1275

Residual (o2E)

(OTAQ) Forte

0.007208

Residual (o2E)

(OTAQ) Gallant

0.02544

Residual (o2E)

(OTAQ) Lucerne

0.1377

Residual (o2E)

(OTAQ) Mazda6

0.4537

Residual (o2E)

(OTAQ) Passat

0.01273

Residual (o2E)

(OTAQ) Patriot

0.02139

Residual (o2E)

(OTAQ) STS

0.01867

Residual (o2E)

(OTAQ) Santa Fe

0.004494

The following table includes 'random' intercepts and slopes for the 12 vehicles included in the analysis.
See Equation 2-13 and discussion on page 29.

-------
Table F-3. Solution for the Random Effects for the Best-fit Model

Vehicle

Effect

Estimate

Std. Err. Pred.

d.f.

(value

Pr> |t|

(ORD) Accord

Intercept (b0)

-0.1100

0.1604

9.17

-0.69

0.5099

Slope (foi)

-0.01970

0.004186

10.7

-4.71

0.0007

(ORD) Fusion

Intercept (b0)

0.1084

0.1507

7.42

0.72

0.4939

Slope (foi)

0.005961

0.004069

9.65

1.47

0.1747

(ORD) Jetta

Intercept (b0)

-0.1434

0.1554

8.27

-0.92

0.3823

Slope (foi)

0.001323

0.004143

10.3

0.32

0.7558

(OTAQ) Accord

Intercept (b0)

-0.01107

0.2345

11.3

-0.05

0.9632

Slope (foi)

-0.000003.44

0.004952

16.1

-0.00

0.9995

(OTAQ) Forte

Intercept (b0)

-0.1180

0.1577

11.6

-0.75

0.4693

Slope (foi)

0.001859

0.003860

13.3

0.48

0.6380

(OTAQ)_Gallant

Intercept (b0)

-0.2306

0.1810

10.3

-1.27

0.2306

Slope (foi)

0.005038

0.004207

13.5

1.20

0.2517

(OTAQ)_Lucerne

Intercept (b0)

-0.3904

0.2127

16.2

-1.83

0.0849

Slope (foi)

0.006119

0.004741

18.7

1.29

0.2126

(OTAQ) Mazda6

Intercept (b0)

-0.07965

0.2717

9.79

-0.29

0.7755

Slope (foi)

-0.00304

0.006207

14.3

-0.49

0.6321

(OTAQ) Passat

Intercept (b0)

0.3832

0.1628

9.02

2.35

0.0429

Slope (foi)

0.002394

0.004266

11.1

0.56

0.5859

(OTAQ) Patriot

Intercept (b0)

0.3810

0.1766

10.5

2.16

0.0550

Slope (foi)

0.000341

0.004139

13.1

0.08

0.9356

(OTAQ) STS

Intercept (b0)

-0.2382

0.1681

9.7

-1.42

0.1878

Slope (foi)

0.01003

0.004297

11.5

2.33

0.0387

(OTAQ) Santa Fe

Intercept (b0)

0.4486

0.1533

11.7

2.93

0.0130

Slope (foi)

-0.01032

0.003798

-2.72

0.0176

-------