* O \
KWJ
*1 PRO^^
Overview of Progress and Findings from the
Cross-EPA Coordination Effort for
Understanding and Evaluating NOx Emissions
Discrepancies
-------
-------
EPA-454/R-21-008
November 2021
Overview of Progress and Findings from the Cross-EPA Coordination Effort for Understanding
and Evaluating NOx Emissions Discrepancies
U.S. Environmental Protection Agency
Office of Transportation and Air Quality
Assessment and Standards Division
Ann Arbor, MI
U.S. Environmental Protection Agency
Office of Research and Development
Atmospheric and Environmental Systems Modeling Division
Research Triangle Park, NC
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Research Triangle Park, NC
-------
Contents
1. Executive Summary 2
2. Problem Statement 3
3. Motivation/Background 4
4. Cross-EPA Coordination Effort for Understanding and Evaluating NOx Emissions Discrepancies:
Structure and EPA Participants 7
5. EPA Assessment as of 2016 9
6. Key Hypotheses and Progress 11
6.1 Model biases related to uncertainties in specifying photochemical processes in the air quality
model or model evaluation methods 12
6.2 Model biases as related to methods used to process emissions for input to the photochemical
model. 21
6.3 Model biases related to overestimates of mobile source emissions from MOVES 23
7. Internal and external outreach activities 29
7.1 Cross-EPA coordination meetings 29
7.2 Technical discussions on Emissions and Atmospheric Modeling (TEAM) 29
7.3 Seminars, scientific conference presentations and special sessions 29
7.4 Journal publications 34
8. Conclusions 36
9. Ongoing Work/Next Steps 37
10. References 38
11. Appendix A: Acronyms 44
1
-------
1. Executive Summary
During 2013-2017, external groups published analyses suggesting that mobile source nitrogen oxide
(NOx) emission estimates developed by the U.S. Environmental Protection Agency (U.S. EPA) were
too high by a factor of up to two. From 2015 to 2020, a cross-EPA workgroup met to coordinate
complementary ongoing efforts in the Office of Air Quality Planning and Standards (OAQPS), the
Office of Transportation and Air Quality (OTAQ) and the Office of Research and Development (ORD)
to evaluate NOx emissions and modeling. The workgroup's aim was to understand discrepancies
between modeled estimates of atmospheric NOx and total reactive nitrogen (NOy) concentrations
and ambient measurements and determine whether these discrepancies were driven by mobile
source emissions estimates or other model processes. The workgroup identified key parameters
and processes impacting NOx and NOy model predictions including: MOtor Vehicle Emission
Simulator (MOVES) inputs and results, spatial and temporal allocation of emissions, meteorology,
mixing, and dispersion treatment in air quality models, air quality model chemistry, and air quality
model treatment of deposition. Smaller teams were formed to investigate key hypotheses
addressing each of the parameters listed above. In a related effort, the Technical discussions on
Emissions and Atmospheric Modeling (TEAM) team was formed as part of a cross-agency
coordination effort between the U.S. EPA, the National Oceanic and Atmospheric Administration
(NOAA) and the National Aeronautics and Space Administration (NASA).
Several major findings came out of the analyses undertaken by the workgroup. First, the
overestimates were most common in the summer, with distinct morning and evening peaks. EPA's
modeling system tended to underestimate NOx in winter. While the analyses did not find a unique
explanation for the summer over-prediction of NOx and NOy concentrations, they identified several
plausible hypotheses, while ruling out others, for the NOx positive biases seen at certain times and
locations in the modeling. Model over-predictions were likely due to multiple compounding factors
that each contributed to a portion of the bias.
Based on our review of the evidence, the most important contributing factor to the summer NOx
bias was:
Planetary boundary layer (PBL) and vertical mixing algorithms in the Community Multiscale
Air Quality (CMAQ) model (version 5.0.2 and earlier) led to too little vertical mixing at
certain times and in some locations. These algorithms have been improved in CMAQv5.1
and later versions of CMAQ. These changes substantially reduced the NOx bias, as well as
the NOx diurnal bias pattern in simulations run with more recent CMAQ versions.
We also demonstrated that there is important uncertainty in the model bias caused by NOx and NOy
measurement uncertainty, as well as chemical mechanism used. Caution should be taken in using
modeled NOx bias to constrain NOx emissions or processes incorporated into air quality modeling.
Through this effort, we identified aspects of the mobile source NOx emissions that were
overestimated in the evaluated air quality platforms. These could lead to important overestimation
of NOx emissions, but based on our analysis so far, only had a modest impact on the magnitude and
pattern of the bias in modeled NOx concentrations. We have identified and developed
improvements to the NOx mobile source emission inventory to address the summer overestimation
of mobile-source NOx, which include:
2
-------
MOVES light-duty NOx emissions rates.
o MOVES light-duty gasoline NOx emissions were reduced in MOVES3 compared to
MOVES2014b. The MOVES3 light-duty gasoline NOx rates are generally lower due to
updated modeling of high-load operation and updated deterioration trends.
MOVES inputs in the 2011 National Emissions Inventory (NEI) and EPA platforms
o MOVES inputs used in the 2011 NEI and EPA modeling platform were not consistent
with the ambient datasets to which they were compared when estimating bias. For
example, speed and acceleration assumptions were not consistent with vehicle
activity at remote sensing locations, long-haul truck hoteling activity was
overestimated nationally, and MOVES inputs did not accurately model local
variability in age distributions and car/truck splits. These inputs have been
improved in MOVES3, the 2016vl platform (U.S. EPA, 2021a) and the 2017 NEI.
National nonroad equipment populations.
o Nonroad equipment populations were overestimated in the 2015 and earlier
platforms. These estimates were updated in the 2016 platform using updated
nonroad population and activity data incorporated into MOVES2014b. These
changes had a noticeable, but relatively small, impact on the NOx bias.
EPA members of this workgroup have shared results and engaged with the scientific and regulated
communities through continued participation in scientific conferences, workshops, journal articles
and other outreach opportunities. Members of the workgroup chaired four special sessions focused
on this topic at conferences which included 14 EPA presentations and 26 relevant presentations
from outside groups (See Table 2). Findings from this work have also been shared in 24 conference
presentations at 11 conferences (See Table 3). Analyses from this workgroup have resulted in five
journal articles led by EPA authors (Referenced in Section 7.4).
Although the NOx evaluation workgroup is no longer active, work continues among EPA offices and
staff to update and improve aspects of the emissions and modeling systems. This document serves
to summarize our efforts and understanding of the NOx evaluation effort at the present time.
2. Problem Statement
This document describes the EPA exploration of a reported discrepancy between modeled estimates
of nitrogen oxides (NOx) and reactive nitrogen (NOy) concentrations and ambient measurements,
particularly summertime overestimates. NOx is defined as NO + N02. Reactive nitrogen includes
both NOx and oxidation products of NOx such as nitric acid (HN03). The mass of reactive nitrogen is
more conserved in the atmosphere than NOx or individual reactive nitrogen species (Seinfeld &
Pandis, 2006). The criticism and subsequent analysis focused on results from modeling systems
using 2011-based inventories and centered on NOx emissions from onroad mobile sources. Most of
this work was done in 2016-2020.
3
-------
3. Motivation/Background
Mobile source NOx emissions have received considerable attention from the scientific community
over the past 15 years since mobile sources are a major component of the inventory. For instance,
mobile sources account for 58% of the NOx in the U.S. EPA 2014 National Emissions Inventory (NEI)
(Toro et al. 2021). In addition, many major point sources are better characterized than mobile
sources since large point sources are generally equipped with Continuous Emission Measurement
Systems. Therefore, when discrepancies between measured and modeled NOx are identified,
mobile sources have been a major focus.
EPA estimates air pollution emissions from onroad mobile sources (cars, trucks, buses, and
motorcycles) using emission models that account for the turnover of the vehicle fleet to vehicles
meeting newer emission standards. Over time, these models have been updated to account for
changing vehicle and fuel regulations, improvements in vehicle technology, the impact of fuels and
ambient parameters, new assessments of real-world vehicle activity patterns, and improved
understanding of the various processes that contribute to vehicle emissions. The MOBILE series of
models was developed beginning in 1978 and culminated with MOBILE6.2 in 2004. In 2002, EPA
began work on the MOtor Vehicle Emission Simulator (MOVES) (https://www.epa.gov/moves),
which was first released for official use as MOVES2010. MOVES was first used to calculate mobile
source emissions for the NEI in 2008. Recent versions of MOVES also incorporated EPA's NONROAD
model that estimates emissions from nonroad mobile sources such as construction and lawn and
garden equipment. The data and detail in these models of onroad emission inventories has
improved substantially over the last few decades as summarized in Table 1.
4
-------
Table 6.1: MOVES Version History
Public Releases
Release Date
Key Features
M0BILE1-
MOBILE6.2
1978-2004
Predecessor to MOVES
Estimated g/mi onroad emissions
Increased scope and complexity over time
NONROAD
1998-2010
Predecessor to MOVES
Estimated emissions for nonroad sources
MOVES2010
2010
New model for onroad emissions
Incorporated vehicle activity
Designed to model at project, county, and national scales
MOVES2010a
2010
Modeled 2012+ Light-Duty (LD) Green House Gas (GHG) rule
MOVES2010b
2012
Performance improvements
Improved vapor venting calculations
MOVES2014
2014
Modeled Tier 3 and 2017+ LD GHG rules
Updated gasoline fuel effects
Improved evaporative emissions
Improved air toxics
Updated onroad activity, vehicle populations and fuels
Incorporated NONROAD model
MOVES2014a
2015
Added nonroad VOC and toxics
Updated default nonroad fuels
Added new options for user Vehicle Miles Travelled (VMT) input
MOVES2014b
2018
Improved emission estimates for nonroad mobile sources
Updated outputs used in air quality modeling
M0VES31
2020
Updated onroad exhaust emission rates, including Heavy-Duty
(HD) GHG Phase 2 and Safer Affordable Fuel Efficiency (SAFE)
rules
Updated onroad activity, vehicle populations and fuels
Added gliders and off-network idle
Revised inputs for hotelling and starts
MOVES3.0.1
2021
Fixed several small issues with processing and aggregation,
making it easier to use the model for variety of applications.
Included scripts to assist with checking MOVES3 submissions for
the 2020 National Emissions Inventory
1 M0VES3 and subsequent minor releases and "patches", are documented at
https://www.epa.gov/moves/moyes3-update-log and https://github.com/USEPA/EPA MOVES Model.
5
-------
Air quality models are often used to predict the fate and transport of atmospheric pollutants such as
NOx. Photochemical air quality models such as the Community Air Quality Model (CMAQ; U.S. EPA
2021b) and the Comprehensive Air quality Model with extensions (CAMx; Ramboll, 2020) simulate
the impacts of pollutant emissions, dispersion, chemical reactions, and deposition on air pollution
concentrations across a 3-dimensional grid. Air quality model output can be applied to project
compliance with National Ambient Air Quality Standards (NAAQS) and to estimate the human health
benefits of regulations that reduce air pollution emissions. Emissions are an important input into
these models and are often derived from the NEI in the U.S., which is released in full on a triennial
basis, although point source emissions are released annually. The NEI includes emissions estimates
from a variety of sources including onroad and nonroad mobile sources. Prior to 2008, the NEI
onroad emissions were estimated using the MOBILE series of models. MOVES2010b was used for
the 2008 NEI and the 2011 NEI, and versions of MOVES2014 were used for the 2014 and 2017 NEI
(U.S. EPA, 2013; U.S. EPA, 2015; U.S. EPA, 2018d; U.S. EPA, 2021e; Toro et al., 2021). NEI emissions
are generally reported at spatial and temporal resolution that is not sufficiently resolved for
photochemical air quality models. In order to address this, NEI emissions are processed through the
Sparse Matrix Operator Kernel Emissions (SMOKE) Modeling System to create gridded, speciated,
hourly emissions for input into a variety of air quality models such as CMAQ and CAMx. SMOKE
supports area, biogenic, mobile (both onroad and nonroad), and point source emissions processing
for criteria and toxic pollutants. Note that for 2008-and-later NEIs, hourly, gridded onroad emissions
are calculated using the SMOKE-MOVES tool and then totaled for the NEI.
Some studies in the literature have used comparisons between model predictions and ambient
measured pollutant concentrations to evaluate the accuracy of emissions inventories. Prior to the
release of EPA's 2011 NEI, published studies found varying model performance for NOx. Kota et al.
(2014) compared CMAQ modeling of Houston using inventories generated with MOBILE6.2 and
MOVES2010, finding that use of MOVES increased the overestimation of NOx, particularly in areas
where mobile sources represent more than 90% of the NOx. Brioude et al. (2011) performed inverse
modeling for NOy using flight data from the Texas Air Quality Study in 2006 (TEXAQS II) and found
good agreement with NEI 2005 (based on MOBILE6 for onroad emissions) for urban Houston as a
whole, but there were large overestimates for the Houston Ship Channel. Marr et al. (2013) found
good agreement in Norfolk between measurements and hourly inventory estimates based on EPA's
2008 NEI which used the MOVES model.
EPA's 2011 NEI and the associated 2011 emissions modeling platform were widely applied for
scientific and regulatory purposes. Scientific investigations compared emissions based on the 2011
NEI or model outputs from simulations using those emissions to measurements made in various
field campaigns in the summers of 2011 and 2013 as well as other available ambient measurement
and satellite data. Several external groups produced analyses suggesting that EPA's 2011 NEI
estimates of NOx emissions were too high. Anderson et al. (2014) compared NEI estimates of
emitted carbon monoxide (CO) to NOy ratios to measurements made on aircraft flights of the
Baltimore/Washington D.C. metro area in July 2011 and concluded that 2011 EPA mobile source NOx
emissions were overestimated by 51%-70%. They specifically recommended scaling EPA mobile
emissions by a factor of 0.5 to improve model performance. Canty et al. (2015) expanded on the
6
-------
work from Anderson et al. (2014) and reported that cutting mobile NOx emissions in half improved
modeled predictions of NOx concentrations across the eastern U.S. Souri et al. (2016) applied
satellite N02 estimates in an inverse modeling analysis over southeast Texas for the summer of 2013
to constrain NOx emissions from different source types. Their model results point to overestimates
of NOx emissions from area, point and mobile sources in urban areas and underestimates of soil and
area source NOx emissions in rural areas. Souri et al. found that reducing mobile source emissions by
30% improved agreement between modeled and measured ground-level N02 from the 2013
DISCOVER-AQ Texas Campaign. Travis et al. (2017), also investigated summer 2013 NOx emissions
estimates (scaled from 2011 emissions) in the Southeastern U.S. Travis et al. used a variety of data
sources including satellite N02, aircraft NOx measurements from the SEAC4RS field campaign, and
routine NADP nitrogen deposition measurements to conclude the 2011 NOx emissions were too
large. They found that reducing NOx by 63% from all sources other than power plants led to
improved model performance. One caution about using satellite N02 data to constrain emissions
came from Kemball-Cook et al. (2015) who found large differences between two Ozone Monitoring
Instrument (OMI) N02 retrievals. When applying these two different satellite retrievals to an inverse
modeling problem, one suggested that NOx emissions inventories in Texas were too high while the
other suggested that NOx emissions inventories in Texas were too low.
Separately, McDonald et al (2018) developed mobile source emissions rates per mass of fuel burned
derived from near-road remote sensing data. These emissions rates were multiplied by state-level
fuel sales to estimate total NOx emissions. When McDonald et al. (2018) compared their inventory
to the 2011 NEI-based emissions, they found that the fuel-based inventory had 50% lower NOx
emissions for onroad gasoline vehicles but 10% higher NOx emissions for onroad diesel sources than
the 2011 inventory. Overall, these results led to 10% lower total NOx emission (from all sources) in
rural areas and 20% lower total NOx emissions in urban areas than the 2011 NEI.
While individual studies have uncertainties and limitations, together they suggested that ambient
NOx and NOy estimates produced by various modeling systems using 2011-based inventories were
high and that additional investigations by EPA were warranted.
4. Cross-EPA Coordination Effort for Understanding and Evaluating NOx Emissions Discrepancies:
Structure and EPA Participants
In 2015, various U.S. EPA offices were involved in independent complementary efforts to
understand and characterize mobile source emissions. OTAQ simultaneously conducted MOVES
development, quality assurance, and validation activities. Within OAQPS, there were projects
evaluating CMAQ and CAMx NOx predictions using the 2011v2 emissions. Staff in EPA's former
National Exposure Research Laboratory had projects underway to evaluate CMAQv5.1 for 2011, as
well as projects analyzing satellite data and field measurements to understand NOx sources. Finally,
staff in EPA's former National Risk Management Research Laboratory (NRMRL)and OAQPS examined
the NOx measurement data from near-road field campaigns in Detroit and Las Vegas.
A cross-EPA coordination group was initiated to 1) increase communication and coordination
between the groups that were already working on various aspects of evaluating EPA NOx emissions
and model estimates and 2) allow for cross-office prioritization of analyses and research questions
that would advance our understanding of NOx emissions, atmospheric transformation and fate.
7
-------
The coordination group held conference calls 5-10 times per year. In addition, six smaller subsets of
the coordination group formed teams that focused on specific topics. Those teams evolved, and
new efforts and collaborations are continuing.
The foci of the initial six working teams were:
1. Use Detroit and Las Vegas near-road measurements to evaluate MOVES emissions factors
2. Use traffic counts from Detroit, Las Vegas, and other sources to develop/evaluate temporal
patterns to inform temporal allocation of mobile source emissions
3. Conduct in-depth and targeted photochemical model (CMAQ and CAMx) evaluation of NOy
species to diagnose important model processes contributing to bias in predictions of these
chemical constituents
4. Evaluate MOVES against roadside, tunnel, and on-road measurements from the literature
5. Conduct targeted MOVES sensitivity simulations to understand the range of MOVES results that
could be obtained different input assumptions such as vehicle speed and acceleration and fleet
age and composition.
6. Further evaluate CO:NOY to determine the usefulness of this metric for understanding NOx
emissions errors.
Cross-EPA coordination team members2 included:
Office of Air and Radiation
o Office of Air Quality Policy & Standards
ฆ Pat Dolwick
ฆ Alison Eyth
ฆ Barron Henderson
ฆ Shannon Koplitz
ฆ Chris Owen
ฆ Sharon Phillips
ฆ Norm Possiel
ฆ Venkatesh Rao
ฆ Heather Simon - Coordination Team Lead
ฆ Brian Timin
ฆ JeffVukovich
o Office of Transportation and Air Quality
ฆ Chad Bailey
ฆ Sudheer Ballare3
ฆ Megan Beardsley
ฆ David Choi
ฆ Jaehoon Han
ฆ Harvey Michaels
ฆ Sarah Roberts
ฆ Darrell Sonntag
2 Some people have retired or moved on to other roles since the start of the project
3 Former ORISE participant hosted by EPA, supported by an interagency agreement between EPA and DOE
8
-------
ฆ Claudia Toro3
ฆ James Warila
ฆ Margaret Zawacki
Office of Research and Development, Center for Environmental Measurement & Modeling
ฆ Wyat Appel
ฆ Jesse Bash
ฆ Kristen Foley
ฆ Deborah Luecken
ฆ George Pouliot
ฆ Havala Pye
ฆ LukeValin
ฆ Sue Kimbrough
5. EPA Assessment as of 2016
An initial evaluation of NO, N02, and NOY was conducted for EPA's 2011 emissions modeling
platform (including mobile emissions from MOVES2014b), and air quality modeling using CMAQ.
Figure 1 shows modeled NOx bias for summer months at all Air Quality System (AQS) monitors
across the U.S. for three consecutive public releases of CMAQ, shown by hour of day. The
evaluation found that NOx and NOv (not shown) were overpredicted at night and during morning
and evening rush-hour but were unbiased or even underpredicted at mid-day, forming a "bridge
pattern." NOx biases progressively decreased as CMAQ matured. CMAQv5.0.2 was the model that
was employed by Anderson et al. (2014). The largest improvement occurred from CMAQv5.0.2 to
CMAQvS.l because substantive updates were made to the model's treatment of vertical mixing in
the boundary layer. The increased vertical mixing in the CMAQvS.l version contributed to
decreasing the ground level NOx and NOY concentrations at night and during morning and evening
transitions, which reduced model overpredictions. Other updates in model processes and inputs
also had moderate impacts on predicted NOx concentrations.
CMAQv5.0.2
NOx bias - all AQS sites
May-Aug
CMAQvS.l
NOx bias - all AQS sites
May-Aug
CMAQv5.2
NOx bias - all AQS sites
May-Aug
--In -iiiilll
III
Figure 1: CMAQ 2011 summertime NOx bias shown by hour of day. Boxes show the interquartile range of bias for each
hour of day.
A comprehensive model evaluation was conducted for the CMAQv5.1 modeling system using a series of
2011 simulations (Appel et al. 2017). While NOx evaluation was not the primary focus of this paper,
9
-------
diurnal profiles of NOx performance were included for each season across all AQS N0X monitors (Figure
2), This figure shows a similar summertime pattern of NOx bias as Figure 1, with overpredictions
overnight, peaking during morning and evening rush-hours, but minimally biased or underpredicted NOx
values at mid-day. Appel et al (2017) provided additional insight into the seasonal nature of this bias by
showing that while the evening rush-hour overprediction was a common feature in all seasons, the
model tended to underestimate NOx overnight and during morning rush hour and during daytime hours
for fall, winter, and spring seasons.
40 -
AQS
#ol Sites 346
35 -
CMAQvS. 1_Base_NEIv1
CMAQvS.0.2_Base
30 -
25 -
20 -
/-/y x\
/A
A // '-"X _
15 -
10 -
AOS
CMAQv5.1_BaseJ\IEIv1
CMAQvS,0.2 Base
o( Sites: 350
22
-
20
-
18
-
16
3
14
Cl
o"
12
-
10
-
8
-
6
-
4
-
30
_
28
-
26
-
24
-
22
-
*
20
-
a.
18
-
O
16
-
14
-
12
-
10
-
8
6
_
AQS
CMAQvS. 1 _Base_NE
CMAQv5,0 2 Base
fr of Sites' 349
AQS
CMAQvS. 1 _Base_NEIv1
CMAQvS.0 2 Base
~iiiiiir~
10 11 12 13 14 15 16
iiiiiIT~
17 18 19 20 21 22 23
Figure 2 (adapted from Appel et al., 2017): Diurnal time series of NOx (ppb) from AQS observations (gray), CMAQv5.0.2
(blue) and CMAQv5.1 (red) for winter (top), spring (top middle), summer (bottom middle) and fall (bottom).
10
-------
Several EPA evaluations of the 2011 modeling system suggested the need to conduct additional
bottom-up comparisons, and targeted model evaluations of specific aspects of MOVES, SMOKE,
CMAQ, and CAMx. EPA staff recognized the complex and multifaceted nature of the modeling
system, which includes emissions from many different sources with varying spatial and temporal
patterns and model treatment of chemical and physical atmospheric processes that impact NOy
composition, transport, and lifetime. In addition, comparisons between models and measurements
are not always straightforward. Measurement artifacts and uncertainties must be considered as
well as differing spatial and temporal resolution of the models and the measurements as described
below. Figure 3 shows a schematic of the modeling systems and shows how different parties are
responsible for different portions of the system underscoring the need for a collaborative process to
evaluate NOx emissions inventories and model predictions.
Model Evaluation
Framework
OAQPS &
States
Spatial and Temporal
Allocation
ORD
& OAQPS
Onroad Mobile:
Cars, trucks, buses
Electric Generating
Units
Photochemical Model
Meteorology:
Mixing &
Transport
Chemistry
OTAQ & States
OAQPS & States
Nonroad Mobile:
Construction equip
Lawn & Garden equip
Ag equip etc.
Other Point Sources
Dry
Deposition
OAQPS a States
Area Sources
Matching Species Defns, Grid
Resolution, Kernel processing
OTAQ & States
Aircraft, rail, marine
OAQPS & ORD &
States
r
Fires
I
States, OAQPS & Outside
Groups
I
All Model
Users
Measurements
V
Measurement
Artifacts
Satellite
Retrieval
Methods
r
Figure 3: Schematic of NOx model evaluation framework and parties responsible for different aspects of the system.
6. Key Hypotheses and Progress
The NOx evaluation coordination group identified a list of potential hypotheses to explain discrepancies
between modeled and observed NOx concentrations. These hypotheses included errors or uncertainties
in the air quality models (i.e., treatment of meteorology, chemistry, dry deposition and grid-resolution),
issues with the measurement data (i.e., uncertainties in satellite retrieval algorithms, spatial
representativeness of measurements, and NOz (i.e. NOy-NOx) measurement artifacts for AQS NOx
monitors) and assumptions built into the emissions inventories (i.e., mobile source emissions rates and
vehicle fleet/activity assumptions, as well as emissions estimates from other sources of NOx).
11
-------
The details of U.S. EPA investigation of these and other hypotheses is provided below.
6.1 Model biases related to uncertainties in specifying photochemical processes in the air quality
model or model evaluation methods
Hypothesis 1.1: Model bias is caused by NOx and NOy measurement uncertainty
Status with regard to aircraft measurements taken from the DISCOVER-AQ Baltimore field
campaign: This hypothesis has been investigated and found to have an important impact on
air quality model NOx bias.
Comparison of model output to NOy measurements made from aircraft during the DISCOVER-
AQ Baltimore field campaign using different types of instruments results in very different
model biases (Figure 4). NOy component species were measured both using the
chemiluminescence and the thermal-dissociation laser-induced fluorescence (TD-LIF)
instruments. The chemiluminescence instrument measured total NOy and provided the data
used in Anderson et al. (2014). Summing the NOy component species CZNOyj) measured by
both the chemiluminescence and TD LIF instruments provides an independent estimate of
measured NOy. CMAQ model simulations using the CB05 chemical mechanism had
normalized mean NOy biases of 76% and 49% when comparing to the chemiluminescence NOy
and YjNOy i measured values, respectively. CMAQ model simulations using the CB6 chemical
mechanism had normalized mean NOy biases of 51% and 28% when comparing to the
chemiluminescence NOy and measured values, respectively. These results were
presented in Simon et al (2018a,b) and Toro et al (2021).
12
-------
_Q
Q_
a.
>,
O
c
ffl
a)
E
Observed
Observed ZNOyi
CB6
CB05
o o
10
15
"T"
20
25
30
Day, July 2011
Figure 4: Daily average NOy values of all DISCOVER-AQ Baltimore aircraft measurements taken within the planetary
boundary layer. Black and gray lines represent measurements made using chemiluminescence and TD-LIF instruments,
respectively. Blue and purple lines represent daily average of all modeled NOy values that were matched in space and time
to the DISCOVER-AQ Baltimore aircraft measurements.
Status with regard to around-based monitors: This hypothesis has been explored and suggests
likely NOx bias may be underestimated by comparisons against ground-based monitors.
Ground-based NOx monitors commonly used in state and local monitoring networks reported
in the AQS have known measurement artifacts. Monitors using EPA's Federal Reference
Method (FRM) are known as chemiluminescence instruments. Within the instrument, NO is
oxidized into excited N02 which fluoresces at specific wavelengths. The FRM monitors
alternately sample ambient air to measure NO concentrations and air that has been routed
over a catalyst to reduce ambient N02 to NO before directing the sampled air to the
chemiluminescence instrument. In this way, the instrument is intended to measure both NO
and N02 concentrations. The catalyst, however, is non-specific and will reduce other NOz
compounds to NO. In monitors intended to measure NOx as opposed to NOy, the catalyst is
often located at the end of a sampling line such that a portion of reactive NOz compounds will
adsorb to the line and not reach the sensor. Despite some adsorption to the sample line,
these monitors often report NOy species as part of NOx measurements (Dunlea et al., 2007;
Dickerson et al., 2019). As a result, measurements are artificially inflated, suggesting that
model overpredictions from comparisons against ground-based monitors might be amplified,
if this impact were not considered.
13
-------
Hypothesis 1.2: Model bias is caused by comparing modeled grid-ceil average concentrations
to measurements made at a finer spatial resolution.
Status with regard to aircraft measurements: This hypothesis has been explored and is not
likely a driving cause of air quality model NOx bias
During DISCOVER-AQ (Baltimore/Washington, Houston, and Denver/Front Range) P-3 spirals
were approximately 4 km in diameter which is a finer resolution than model grid-cells for
simulations conducted at 12km resolution. Analysis of different ways to match modeled grid
concentrations to DISCQVER-AQ aircraft measurements showed that model bias is sensitive to
sampling error, but all tested sampling schemes generally still result in an NOY overprediction
(Figure 5). These results were presented by Simon et al. (2018b).
0 5 10 15 20 25 30
Day, July 2011
Figure 5: Daily average NOy values of all DISCOVER-AQ Baltimore aircraft measurements taken within the planetary
boundary layer. Black and gray solid lines represent measurements made using chemiluminescence and TD-LIF
instruments, respectively. Blue and purple solid lines represent daily average of all modeled NOy values that were
matched in space and time to the DISCOVER-AQ Baltimore aircraft measurements. For solid lines, model values represent
matching a measurement location to a grid cell (horizontal and vertical) and measurement time to the closest hour.
Shading represents the range of model values from sampling +/-1 grid cell in each direction and +/-1 hour.
Status with regard to ground-based monitors: Hypothesis has not yet been actively explored
This hypothesis has not yet been explored in-depth for ground-based monitors. A potential
follow-up would include analyzing modeled vs. measured concentrations from 2017 and 2018
field campaigns using forthcoming 4km resolution modeling simulations. Some insight may be
gained by comparing performance of multiple model simulations conducted using different
grid resolutions and examining within-cell variability in measured data by identifying locations
with multiple monitors within a single modeled grid cell.
Hypothesis 1.3: The planetary boundary layer (PBL) and vertical mixing algorithms in
photochemical models lead to too little vertical mixing at certain times and in some locations.
14
-------
Status: This hypothesis has been explored and found to have an important impact on air
quality model NOx bias. Updates to vertical mixing in the CMAQ model substantially reduced
the N0X overprediction, but did not completely resolve the discrepancy.
Evidence suggests that the PBL and vertical mixing algorithms in CMAQ versions 5.0.2 and
earlier led to too little vertical mixing at certain times and locations, especially at night and
during morning and evening transition periods when the PBL rises quickly (morning) and
collapses quickly (evening).
Extensive work has occurred to update vertical mixing schemes in CMAQ (starting with
CMAQv5.1). Specifically, updates were made both to the Pleim-Xiu land-surface model within
CMAQ and the asymmetric convective mixing version 2 PBL scheme within both the Weather
Research Forecasting (WRF) and CMAQ models. In addition, errors in the CMAQ calculation of
the Monin-Obukhov length (MOL) calculation were corrected (Appel et al., 2017). The impacts
of improved representation of vertical mixing were shown in Henderson et al. (2017a; 2017b)
and in Toro et al (2021). Figure 6 (right) (reproduced from Toro et al., 2021) shows the
dramatic improvement in 2011 NOx bias between model simulations conducted using
CMAQv5.0.2 and CMAQv5.1 at monitors in four urban areas. Additional testing showed that
most of the NOx changes between the simulations at these monitors were attributable to the
vertical mixing updates included in CMAQv5.1. The changes to the modeled PBL had the
largest impact on the air quality model NOx bias of all the hypotheses evaluated.
15
-------
Summer
Winter
ฎ Jo 0.5-
ro ,2>
> 0) T
r^O.O-
3 9?
-------
Hypothesis 1.4: Dry deposition velocities for NOy species are too low in CMAQ
Status: This issue has been investigated and is not likely a driving cause of air quality model
NOx bias
In CMAQ v5.3, the parameterization of NOx and PAN deposition from Pleim and Ran (2011)
and alkyl and peroxy nitrates from Nguyen et al. (2015) is used for the stomatal/mesophyll
resistance. Laboratory studies showed that approximately 80% vegetative sink of PAN is via
stomatal uptake (Sun et al., 2016) while micrometeorological field experiments typically
estimate a smaller stomatal sink (25-50%: Wolfe et al., 2009; Turnipseed et al., 2006).
However, an equal amount of the measured flux may be explained by thermal decomposition
resulting in similar (approximately 20%) non-stomatal fluxes as reported in laboratory
experiments (Wolfe et al., 2009). CMAQ v5.3 has a parameterization of PAN deposition (Pleim
and Ran, 2011) to non-stomatal surfaces consistent with Turnipseed et al. (2006) and higher
than many other contemporary models (Wu et al., 2012). In a sensitivity study, the mesophyll
resistance reactivity factor was decreased by an order of magnitude from the initial values of
Pleim and Ran, 2011 to match that of 03, near the values reported by Wolfe et al. (2009), and
similar changes were made to the mesophyll reactivity factor for alkyl and peroxy nitrates to
bound the impact that this parameter has on the overall modeled NOy budget. This was a
bounding experiment as PAN is the only organic nitrate documented to react with the enzyme
nitrate reductase in the mesophyll (Sparks, 2009). The non-stomatal CMAQ deposition velocity
was not changed as it is already at the high end of the observed values (Wu et al., 2012). The
results of model sensitivity simulations show little impact, less than 1% difference in the
modeled ambient concentration, of updates to surface resistance for alkyl nitrates and peroxy
nitrates. This is likely due to the high solubility of these species and the limiting resistance is
the stomatal resistance. These simulations and previous studies revealed that the dry
deposition velocity of NOz species is already at the high end of the observations (Wu et al.,
2012) and further increases would lead to unrealistic deposition rates and is unlikely to be a
driver of NOy overpredictions. Additional work to update land use (i.e., more accurately
specifying land vs. water cells) had both decreased and increased NOy concentrations near
coastal areas due to the parameterization of mixing and differences in surface areas between
vegetated land and water surfaces. There is ongoing work to investigate updates to VOC
deposition, which appears to have some impact on alkyl nitrate and peroxy nitrate
concentrations but does not appear to affect other NOy species.
Hypothesis 1.5: Model chemical mechanisms may not properly characterize individual organic
nitrogen species, particularly their solubility, formation, reaction rates and products, so it is
possible that more deposition and decay is occurring than predicted
Status: This issue has been explored and results indicated that including additional alkyl nitrate
species in the model substantially lowered NOY concentrations but had little impact on NOx
concentrations. Understanding the NOx lifecycle continues to be an active area of mechanism
development.
Evaluation of NOy species with different chemical mechanisms shows that the choice of
chemical mechanism impacts NOy model performance, with the newer CB6 mechanism
outperforming the older CB05 mechanism (Figure 4). However, further modifications to the
17
-------
chemistry cannot reduce total NOy without adding to the overpredictions of HN03, as shown
in Figure 7 reproduced from Toro et al. (2021). The results were first presented at the
Atmospheric Chemical Mechanisms Conference (Simon et al., 2018b) and are described in
more detail in Toro et al (2021).
r--i
S"
Q_
Q. CO-
CA
CD
"o LO-
CI)
Q.
-
c
CD
c
O co-
Q.
E
o
O CM"
~ NO ~ N02 ~ HN03 ~ ANs ~ PNs ~ NOy
34%
23%
23%
16%
32%
10%
35%
18%
23%
16%
39%
18%
23%
10%
34%
25%
CB05
CB05e51
CB6
Observations Observations
Figure 7 (reproduced from Toro et al., 2021) Aggregated model predictions and measurements of NOY species. All
measurements considered are within the boundary layer over all flight days part of the July DISCOVER-AQ 2011 field study
near Baltimore, MD. Bars representing observations are derived from both LIF (N02, HNOs, alkyl nitrates, peroxy nitrates)
and chemiluminescence (NO and NOY) instruments. NOY is shown as a gray bar. DISCOVER-AQ = Deriving Information on
Surface Conditions from Column and Vertically Resolved Observations Relevant to Air Quality.
Separately, EPA researchers led an effort to create a detailed chemical treatment of organic
nitrate species. Specifically, Zare et al. (2019) characterized their solubility, formation,
reaction rates, and products using an update to the RACM2 chemical mechanism. When
modeling a summer 2013 episode, Zare et al. (2019) showed that the updated chemical
treatment reduced the model prediction of alkyl nitrate species from a factor of 2
overestimate to a 32% underestimate at a rural measurement site in Alabama. Understanding
and representing the sinks of NOx in atmospheric models continues to be an area of active
research (e.g., Vasquez et al. 2020).
Hypothesis 1.6: Model bias is due to some unique feature of the 2011 modeling platform
Status for summertime: This hypothesis has been explored and is not likely a driving cause of
air quality model NOx bias.
NOx bias is not unique to 2011 and is seen throughout 2002-2012 model simulations using
CMAQv5.0.2. Model bias decreased over this time period as observed NOx decreased. The
NOx bias has decreased further in modeling of 2013-2017 that used more recent model
versions (e.g., CMAQv5.1, v5.2, and v5.3). The model overprediction in the most recent CMAQ
v5.3 simulations using 2016 emissions shows substantially lower bias than was seen in
simulations of 2011. These results are shown in Figure 8 and were presented at the 2019
CMAS and AGU conferences (Foley et al., 2019; Simon et al., 2019) and in Toro et al. (2021).
18
-------
West - NOx NMB [4am-9am]
Northwest - NOx NMB [4am-9am]
(v5.3) 2016
(v5.2.1)2016
(v5.2) 2015
(v5.2) 2014
(v5.1) 2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5.0.2) 2010
(V5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(V5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-54
-29
-14
-43
-54
-27
-11
-42
-54
-34
-8
-40
-56
-33
-13
-42
-57
-19
1
-35
-48
-20
6
-28
-54
-26
6
-30
-28
-4
38
-11
-51
-20
-3
-32
-41
-24
10
-30
-55
-30
-1
-34
-45
2
20
-11
-39
-11
9
-24
-51
-25
9
-28
-42
-10
6
-20
-45
-2
8
-21
(V5-3) 2016
(V5.2.1) 2016
(v5.2) 2015
(v5.2) 2014
(v5.1) 2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5.0.2) 2010
(V5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(V5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-41
-33
-27
-31
-41
-31
-22
-29
-39
-36
-27
-31
-43
-35
-32
-41
-35
-30
-16
-31
-23
-8
2
-17
-5
8
15
-5
-12
13
29
16
21
40
66
27
-19
-6
14
22
-21
-6
41
19
-61
-37
13
-37
-67
-21
35
-8
-78
-46
48
-12
-85
-45
17
-44
-57
-5
44
-54
Percent
ฆ >80
n 60 to 80
40 to 60
20 to 40
-20 to 20
-40 to -20
-60 to -40
-80 to -60
-100 to-80
winter spring summer
winter spring summer
UpperMidwest - NOx NMB [4am-9am]
OhioValley - NOx NMB [4am-9am]
Northeast - NOx NMB [4am-9am]
(V5.3) 2016
(v5.2.1) 2016
(v5.2) 2015
(v5.2) 2014
(v5.1)2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5-0.2) 2010
(v5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(V5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-24
-8
5
-16
-26
-14
-4
-23
-21
-6
34
-7
-31
-14
20
-18
-21
0
26
-12
1
40
76
17
18
30
88
22
17
42
118
41
-3
40
112
39
22
53
109
35
35
60
146
70
41
65
137
50
38
52
114
56
12
46
115
52
45
55
118
61
24
50
99
64
(v5.3) 2016
(V5.2.1) 2016
(v5.2) 2015
(v5.2) 2014
(v5.1) 2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5.0.2) 2010
(V5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(v5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-36
-14
9
-20
-39
-21
0
-26
-31
-14
12
-17
-31
-21
8
-20
-27
-16
17
-15
-12
14
35
-1
0
14
54
0
-5
-1
41
7
-6
6
35
26
-6
-2
31
6
-7
12
56
16
-4
9
56
12
4
22
86
25
-11
10
51
10
-3
22
72
14
-2
23
53
5
(v5.3) 2016
(v5.2.1) 2016
(v5.2) 2015
(v5.2) 2014
(v5.1) 2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5.0.2) 2010
(V5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(v5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-47
-35
-14
-34
-45
-33
-7
-32
-41
-38
-8
-33
-41
-32
0
-27
-36
-25
-6
-28
-20
-8
29
-2
-33
-11
3
-19
-15
-9
26
-2
-34
-22
18
-8
-25
-22
3
-17
-24
-26
8
-12
-19
-11
41
-3
-29
-19
8
-17
-35
-18
6
-20
-24
-17
19
-15
-30
-18
2
-12
winter spring summer
winter spring summer
winter spring summer
Southwest - NOx NMB [4am-9am]
South - NOx NMB [4am-9am]
Southeast - NOx NMB [4am-9am]
(V5.3) 2016
(v5.2.1) 2016
(v5.2) 2015
(v5.2) 2014
(v5.1) 2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5.0.2) 2010
(V5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(V5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-34
-9
-2
-4
-35
-8
4
-3
-33
-6
23
-4
-45
-27
-10
-30
-55
-30
-17
-37
-38
-23
7
-24
-41
-11
12
-21
-49
-30
1
-29
-48
-12
27
-20
-47
-29
5
-31
-50
-30
-2
-31
-46
-27
6
-23
-46
-7
21
-20
-56
-23
5
-24
-50
-18
-2
-29
-58
-29
-18
-35
winter spring summer
(v5.3) 2016
(V5.2.1) 2016
(v5.2) 2015
(v5.2) 2014
(v5.1) 2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5.0.2) 2010
(V5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(V5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-20
-5
31
5
-19
-4
38
9
-21
8
23
2
-22
-16
18
-13
-22
-6
37
2
0
31
83
15
-4
10
68
2
1
43
132
48
3
40
107
48
30
72
164
57
-1
27
98
29
-6
32
88
23
40
78
163
67
35
74
154
101 I
45
MEM
195
121
I 23
175
82 1
winter spring summer
(v5.3) 2016
(V5.2.1) 2016
(v5.2) 2015
(v5.2) 2014
(v5.1) 2013
(V5.0.2) 2012
(V5.0.2) 2011
(V5.0.2) 2010
(V5.0.2) 2009
(V5.0.2) 2008
(V5.0.2) 2007
(V5.0.2) 2006
(V5.0.2) 2005
(V5.0.2) 2004
(V5.0.2) 2003
(V5.0.2) 2002
-37
-20
16
-20
-39
-22
16
-23
-33
-15
12
-14
-33
-17
16
-13
-24
-9
27
-9
-9
27
95
20
-17
15
62
13
-7
20
108
33
-14
15
76
27
9
23
84
25
16
20
113
47
-9
10
12
-12
2
109
9
-30
-1
78
28
1
51
27
-17
15
93
44
winter spring summer
Figure 8 (from Toro et al., 2021) Normalized mean bias of morning modeled NOx minus observed NOx. Morning hours are
4-9 AM LST. Morning bias is aggregated by season for each annual simulation across monitors in multiple regions.
Community Multiscale Air Quality model version used for each simulated year is shown in parentheses. Reds indicate
model overprediction, and blues show model underprediction. West = CA and NV; Northwest (combined with the
Northern Rockies and Plains) = OR, WA, ID, MT, NE, ND, SD, and WY; Upper Midwest = IA, Ml, MN, and Wi; Ohio Valley = IL,
IN, KY, MO, OH, IN, and WV; Northeast = CT, DE, ME, MD, MA, NH, MJ, NY, PA, Rl, and VI; Southwest = AZ, CO, NM, and
UT; South = AR, KS, LA, MS, OK, and TX; and Southeast = AL, FL, GA, NC, SC, and VA.
Status for seasons other than summer: This hypothesis has been explored and is not likely a
driving cause of air quality model NOx bias.
in winter months, NOx is underpredicted during most hours of the day except during evening
rush-hour (Figure 9). Similar to findings for summer for simulations from 2002-2016, modeled
winter NOx concentrations have decreased more than observed winter NOx concentrations.
Consequently, at times and locations with NOx underpredictions, these underpredictions have
worsened. Conversely, the evening hour NOx overpredictions have decreased over this time
period. These results were presented at the 2019 CMAS and AGU conferences (Foley et al.,
2019; Simon et al., 2019) and in Toro et al. (2021).
19
-------
Seasonal Mean NOx by Hour
Region = CONUS | Period = Winter 2002 - 2016
Seasonal Mean NOx by Hour
Region = CONUS | Period = Summer 2002 - 2016
Hour of Day (LST)
10 12
Hour of Day (LST)
Hour of Day (LST)
Seasonal Mean NOx by Hour
Region = CONUS | Period = Winter 2002 - 2016
Hour of Day (LST)
- .
2002 (v5.0.2)
2003 (V5.0.2)
2004 (V5.0.2)
2005 (V5.0.2)
-
2006 (V5.0.2)
2007 (V5.0.2)
-
2008 (V5.0.2)
2009 (V5.0.2)
2010 (V5.0.2)
2011 (V5.0.2)
2012 (V5.0.2)
2013 (V5.1)
-
2014 (v5.2)
2015 (v5.2)
-
2016 (V5.2.1)
2016 (v5.3)
- .
2002 (V5.0.2)
2003 (V5.0.2)
2004 (V5.0.2)
2005 (V5.0.2)
-
2006 (V5.0.2)
2007 (V5.0.2)
ซ
2008 (V5.0.2)
2009 (V5.0.2)
ซ
2010 (V5.0.2)
2011 (V5.0.2)
2012 (V5.0.2)
2013 (V5.1)
-
2014 (V5.2)
2015 (v5.2)
-
2016 (V5.2.1)
2016 (v5.3)
Seasonal Mean NOx Bias by Hour
Region = CONUS | Period = Winter 2002 - 2016
Seasonal Mean NOx Bias by Hour
Region = CONUS | Period = Summer 2002 - 2016
Figure 9 (reproduced from Toro et al (2021)) Diurnal mean NOx observations (top) and bias (bottom). Profiles for winter
(left) and summer (right) are shown by year. Modeled years include 2002-2016.
Hour of Day (LST)
Hour of Day (LST)
Seasonal NOx NMB by Hour
Region = CONUS | Period = Winter 2002 - 2016
Seasonal NOx NMB by Hour
Region = CONUS | Period = Summer 2002 - 2016
o
T ~
0 2 4 6 8 10 12 14 16
Hour of Day (LST)
0 2 4 6 8 10 12 14 16 18 20 22
Hour of Day (LST)
Seasonal Mean NOx by Hour
Region = CONUS | Period = Summer 2002 - 2016
20
-------
Hypothesis 1.7: Air quality model bias is unique to summer in the Eastern U.S.
Status: This hypothesis has been explored and confirmed to be an important characteristic of
the air quality model NOx bias. The model NOx overpredictions appear to be most prevalent in
the Upper Midwest, Ohio Valley, South, and Southeast regions during summer months. The
model often underpredicts NOx concentrations at other times of year and locations.
NOx bias appears to be a feature unique to summer in the model, with morning
overprediction eliminated in winter. These results are shown in Figure 8 and have been
presented in Henderson et al (2017a; 2017b) and documented in Toro et al. (2021). These
findings are also consistent with results presented in Appel et al. (2017). In addition, a recent
2015 wintertime field campaign in New England was conducted by external researchers. Two
studies from that campaign (Salmon et al., 2018; Jaegle et al., 2018) both concluded that
measurements were in agreement with EPA emissions inventories and resulting model
estimates for that time period. The cause of the seasonally changing NOx bias has not yet been
identified and is a topic that needs further research.
6.2 Model biases as related to methods used to process emissions for input to the photochemical
model.
Hypothesis 2.1: Spatial allocation (county to grid cell) is incorrect for onroad emissions
Status: This hypothesis has been investigated and is not likely a driving cause of air quality model
NOx bias
Improvements have been made to spatial allocation for onroad emissions in the 2014 and 2016
model simulations. Improvements included more fully resolved data for the road network
through the use of Annual Average Daily Traffic (AADT) data by road link in the development of
spatial surrogates, although allocation of off-network emissions such as starts and idling may
still need improvement. Air quality model-ready emissions were developed based on the AADT,
but an air quality model run was not performed because the emissions changes were small and
localized. Updates to the surrogates have been documented in the emissions modeling technical
support documents for the 2014v7.1 and 2016v7.2 platforms (U.S. EPA, 2018c; U.S. EPA, 2019).
Hypothesis 2.2: Heavy-Duty (HD) onroad vehicle running emissions are at the wrong time of day
Status: This hypothesis has been investigated and is not likely a driving cause of air quality model
NOx bias
Sensitivity model simulations in which HD diurnal profiles were altered to allocate more
emissions during daytime hours and fewer emissions during nighttime hours had little impact on
modeled NOx concentrations. Results from CAMx model simulations were presented at the 2017
CMAS conference (Timin et al., 2017). In addition, CMAQ model sensitivity simulations were
included in the Toro et al. (2021) journal article.
Hypothesis 2.3: Emissions from Electrical Generating Units (EGU) without Continuous Emission
Monitors (CEMS) may be inappropriately allocated from annual to hourly emissions
21
-------
Status: This hypothesis has been investigated and is not likely a driving cause of air quality model
N0X bias nationwide although it may have large impacts on modeled NOx bias over limited
spatial and temporal scales
An error was found in how temporal profiles were assigned to EGUs without CEMS. This error
led to some facilities such as municipal waste incinerators having temporal allocations that
resembled profiles that are applied to peaking units. While the municipal waste incinerators
and other non-CEMS EGUs did not account for a large fraction of the inventory, this error
resulted in annual emissions being predominantly allocated to only a few days in the 2011
modeling at some units. Model simulations were conducted to evaluate the impact of updating
the temporal profile to more realistically show these units emitting NOx throughout the entire
year. Fixing daily allocation for non-CEMs EGUs had a large impact for a few locations on a few
days. These days and locations happened to coincide with the DISCOVER-AQ Baltimore field
study which was used in the Anderson et al (2013) study, so it may have impacted their
conclusion that EPA NOx emissions were too high. This information was described in Simon et al
(2018a). The impact of updating this emissions temporalization was evaluated using CAMx
sensitivity simulations. The updated temporalization did not have a substantial national impact
on NOx concentrations or model performance, although impacts were important on several days
directly downwind for the EGU facilities. Results from CAMx model simulations were presented
at the 2017 CMAS conference (Timin et al., 2017). In addition, CMAQ model sensitivity
simulation results were included in Toro et al (2021).
Hypothesis 2.4: Monthly, day-of-week, and/or diurnal temporal profiles for nonroad
equipment activity are incorrect
Status: This hypothesis has been investigated and is not likely a driving cause of air quality model
NOx bias
Model sensitivities in which diurnal profiles for nonroad equipment activity were altered had
little impact on modeled NOx concentrations. Results from CAMx model simulations were
presented at the 2017 CMAS conference (Timin et al., 2017). In addition, updated CMAQ
model sensitivity simulation results were included Toro et al (2021).
Hypothesis 2.5: Spatial allocation (county to grid cell) of nonroad equipment is incorrect
Status: Hypothesis has not been actively investigated
This has not yet been evaluated but is likely less important than national to county equipment
allocation (see below).
Hypothesis 2.6: Spatial allocation of nonroad emissions (national to county) is not
representative of actual emissions
Status: This issue needs more analysis
The surrogate data used to allocate national populations of agricultural and construction
equipment to the county level was updated in the 2016vl emissions modeling platform. These
improvements in the spatial distribution of construction and agricultural nonroad emissions
have been implemented in the Inventory Collaborative's 2016vl emissions modeling platform
22
-------
(U.S. EPA, 2021a). The construction and agricultural equipment spatial allocation update
conserved the model's national base equipment population but re-allocated the population such
that some states saw an increase in population (and therefore activity and emissions), while
others saw a decrease. EPA staff have not conducted a sensitivity analyses of the platform to
investigate the individual impact of the updated spatial surrogates for agricultural and
construction equipment on the bias. Nationally, emissions didn't change significantly, and the
emissions and air quality impacts are anticipated to be localized. Further work is needed to
understand the impact of these spatial surrogates and estimated NOx emissions on local areas.
6.3 Model biases related to overestimates of mobile source emissions from MOVES
Hypothesis 3.1: Onroad light-duty (LD) gasoline vehicle emissions rates are too high.
Status: We invested significant effort to evaluate and update LD gasoline onroad vehicle
emission rates in MOVES. Our early work suggested significant reductions in NOx LD emission
rates in MOVES. In a sensitivity evaluation, these preliminary reductions had a noticeable, but
modest effect on the NOx bias evaluated in the 2016 air quality modeling platform. Additional
work showed that a smaller change in NOx rates was warranted (U.S. EPA, 2020b), so the final
MOVES3 light-duty emission rates are higher than the emission rates evaluated in the sensitivity
case. Thus, the sensitivity case serves as an upper bound to the changes in NOx ambient
concentrations in 2016. Despite the modest effect on the NOx bias, we are continuing to compile
data to improve LD base emission rates, adjustment factors and activity inputs for future
versions of MOVES.
There are many factors that impact LD emission rates estimated by MOVESthe NOx base
emission rates in the model vary with emission process, operating mode, and age. Emission
rates are then weighted together based on estimates of activity and are adjusted to account for
fuel effects, Inspection & Maintenance programs, ambient conditions, and air conditioning.
Exploring the hypothesis that the emission rates are too high requires analyzing these factors
individually. A description of the factors explored to date are listed below:
Emission rates under high-power driving conditions
Comparison of trends for NOx, HC, and CO emissions in relation to vehicle-specific power for
Tier 2 vehicles (model year 2004 - 2013) between MOVES2014 and real-world data suggests
that MOVES2014 emission rates for operating modes representing high speed and
acceleration are likely too high (Sonntag et al. 2018). Revised assumptions for high-power
emissions for vehicles in model year 2004 and later result in better agreement with real-
world measurements acquired using Portable Emissions Measurement Systems (PEMS).
Sensitivity analyses indicate that updating these assumptions has a minor impact (<5%
reduction) for pre-2016 calendar year NOx emissions from light-duty vehicles, but these
revisions become important for future years (>20% reduction in calendar year 2028) (Toro
et al. 2019a). These emission rates were updated in MOVES3 (U.S. EPA, 2020b). We are
continuing to gather and analyze data on emissions under high-power driving conditions and
expect to further refine these rates as data become available.
23
-------
Deterioration of start emissions
For emissions immediately following engine starts, comparisons to the In-Use Verification
Program (IUVP) suggest that for Tier 2 vehicles, NOx emissions from starts rise more slowly
with vehicle age (in proportional terms) than emissions during hot-running operation. This
result differs from the assumptions included in MOVES2014, where both emission processes
were assumed to deteriorate at the same proportional rate. Sensitivity analyses using IUVP-
based deterioration show a moderate (7-10%) decrease in NOx emissions for light-duty
vehicles for all calendar years evaluated (Toro et al. 2019a). Starts trends with age were
updated in MOVES3 (U.S. EPA, 2020b).
Deterioration of running emissions
Comparison of MOVES2014 Tier 2 vehicle running emission trends with vehicle age and data
from the Denver Inspection & Maintenance (l/M) program suggest that the NOx
deterioration trend included in the MOVES2014 is more aggressive than that seen in the
Denver l/M data. Emissions impact sensitivity tests indicate that updates to the
deterioration trend result in moderate NOx emissions reductions for the LD sector, in the
range of 5-15% depending on the calendar year (Toro et al. 2019a). Updated deterioration
trends and baseline emission rates were estimated for light-duty NOx vehicles using the
Denver l/M data in MOVES3 (U.S. EPA, 2020b).
Preliminary sensitivity results
Preliminary sensitivity results from the three adjustments to base emission rates described
above indicated a cumulative reduction of LD NOx emissions on the order of 20% for calendar
year 2011 and 30% in 2016. An overview of the data and sensitivity analysis was presented at
the IEIC 2019 conference (Toro et al., 2019a). The sensitivity of these changes to modeled
ambient NOx concentrations was evaluated using the 2016 beta collaborative platform, with
mobile-source emissions estimated using MOVES2014b, and presented at the 2019 AGU Fall
Meeting (Toro et al. 2019b). The median NOx bias for the conterminous U.S. (CONUS) are shown
by season and hour in Figure 10. With the sensitivity case, the median modeled NOx
concentrations had a noticeable decrease. The NOx bias slightly improved for the winter evening
hours, and the summer morning, evening, and night-time hours. However, the morning winter
bias worsened with the change. While showing noticeable changes to the NOx concentrations
and bias, the overall NOx bias pattern is relatively unchanged. As discussed in Toro et al. 2019b,
results varied regionally. The Southeast showed the most model improvement in the summer
morning with the sensitivity change. However, all CONUS regions showed a worse model
performance in the winter morning periods.
24
-------
LST
Figure 10 (reproduced from Toro et al. 2019b) Hourly median NOx bias for the Conterminous U.S. (CONUS) by season and
hour (local standard time - LST), using a baseline case in 2016 and a sensitivity case with reductions to the light-duty NOx
emission rates due to revised high-power rates and deterioration trends.
Final MOVES3 light-duty emission rates
Since conducting the emissions and air quality model sensitivity analyses discussed above, we
developed updated light-duty vehicle emission rates for MOVES3 (US EPA, 2020b). Compared to
the sensitivity case discussed above, the new rates account for updated deterioration trend and
zero-mile rates estimated using data from the Denver l/M program. The MOVES3 start emission
rates also include revisions to the relationship between starts and soak time (time parked before
starting), which increases NOx start emissions during starts at intermediate soak times. The
combined updates to the MOVES3 light-duty emission rates still generally decrease the MOVES
light-duty emission rates compared to MOVES2014b by individual vehicle regulatory class,
operating mode, age, and model year, but to a smaller degree than evaluated in the preliminary
sensitivity case discussed above; in some cases the light-duty emission rates actually increase as
presented in the MOVES3 light-duty emission rate technical report (US EPA, 2020b).
The impacts of the final MOVES3 light-duty NOx emission rates on estimated NOx emissions and
NOx bias have not been evaluated separately. The updates to light-duty NOx emission rate were
made to MOVES3 in conjunction with other changes, including updates to light-duty population
and activity that also impact NOx emissions. Thus, the percent changes in the estimated NOx
light-duty emissions at national scale are different than the changes in the light-duty emission
rates. In fact, despite MOVES3 having generally lower LD NOx emission rates (e.g. gram/mile or
gram/start by vehicle regulatory class) than MOVES2014b, MOVES3 estimates higher LD NOx
total emissions (kilograms or tons) than MOVES2014b in some calendar years. For example,
calendar year 2023 shown in Figure 11 shows higher gasoline NOx in 2023 due to MOVES3
estimating a higher fraction of light-duty truck populations and activity compared to
MOVES2014b (Han, J. 2021). However, future gasoline MOVES3 light-duty NOx emissions are
significantly lower, which is consistent with the updates made to future model year light-duty
25
-------
emission rates based on our evaluation. A further discussion of uncertainty of population and
activity MOVES inputs on the NOx bias is discussed in Hypothesis 3.3.
~ Gasoline ~ Diesel ~ CNG ~ Ethanol (E-85)
if)
C
ฐ 1
3
M-
o
-------
Status: This hypothesis has been investigated and found to have an important impact on MOVES
NOx emissions. The isolated impact on NOx bias has not been evaluated.
Speed and acceleration assumptions, age distributions, and car/truck splits were
identified as influential MOVES inputs on NOx emissions (Choi et al. 2017).
Average speed distributions were updated for the 2017 NEI and MOVES3 based on
telematics data. However, the new distributions show more driving at higher speeds,
which can increase emissions. (U.S. EPA 2021b). When compared to four US road-side
remote sensing locations, MOVES default speed and acceleration assumptions for the
representative county were significantly more aggressive than was measured at the
three of the four sites (Choi et al. 2017), likely due to bias in site selection.
Age distributions were updated for the 2017 NEI and MOVES3. In general, the age
distributions show an older vehicle fleet, which can increase emissions. (U.S. EPA
2021c). When compared to four US road-side remote sensing locations, MOVES default
age distributions were significantly older than the age distributions at three of the sites.
Using the MOVES default age distributions to model these locations led to a significant
overestimation in NOx emissions (Choi et al. 2017).
The MOVES vehicle classifications do not map perfectly to vehicle activity and
registration data. Uncertainties in this classification can impact resulting emissions
since, for example, MOVES emission factors for passenger cars are lower than those for
passenger trucks and light commercial trucks. For the 2016vl platform, to ensure
consistency in "passenger car," "passenger truck," and "light commercial truck" splits
across the country, all state-submitted VMT for these vehicle types was summed and
then re-allocated using the splits obtained from county-level registration data from a
single data provider (U.S. EPA, 2019). EPA staff have not assessed the impact of this
change on NOx concentrations on the 2016vl compared to previous platforms, and this
approach was not incorporated into the 2016 NEI alpha and 2016 NEI beta platforms
evaluated by Toro et al. (2021). This approach is incorporated into the 2016vl, 2017 NEI,
and EQUATES emissions and air quality modeling platforms.
EPA staff identified and implemented improvements to make use of gridded hourly
humidity data instead of monthly average humidity and to use speed distributions
instead of individual speeds in the NEI and platform (Baek and Eyth, 2019). These
updates resulted in temporal and spatial differences in NOx emissions.
Based on the latest telematics data, EPA reduced the default hoteling hours for long-
haul combination trucks by over a factor of 3 in the 2016vl platform, compared to the
2014 NEI. (U.S. EPA (2019). This change had a significant reduction in total heavy-duty
NOx emissions. The updated hoteling activity decreased the total national heavy-duty
NOx emissions by 10% to 21% between 2010 and 2020 (Han et al. 2019). County-scale
reductions varied, with higher reductions seen in rural counties. Note that these
hoteling activity reductions were not included in the 2016 platform evaluated by Toro et
al. (2021), but are included in the 2016vl, 2017 NEI, EQUATES emissions and air quality
modeling platforms.
27
-------
Hypothesis 3.4: Nonroad emission rates are too high
Status: This hypothesis has been investigated and is not likely a driving cause of air quality model
NOx bias. Revisions made to Tier 4 nonroad engine emissions in MOVES2014b likely had a small
impact on the NOx bias in the 2016-and-later emissions and air quality modeling platforms. A
more comprehensive evaluation of nonroad NOx emissions has been postponed since there will
be future efforts to update nonroad equipment emissions rates.
In 2018, EPA released MOVES2014b, which incorporated updated NOx emission rates for Tier 4
(first introduced with model year 2008) nonroad engines based on updated certification data. In
addition, compliance provisions available to manufacturers were accounted for (U.S. EPA,
2018a). Based on MOVES national inventory tests, this change had very small impact in calendar
year 2011 but resulted in a 3% increase in nonroad NOx emissions in 2016, and a 1% decrease in
nonroad NOx emissions in 2025. These updates were incorporated in the 2016 platform
evaluated by Toro et al. (2021) which estimated mobile emissions using MOVES2014b. However,
changes in the NOx bias in the 2016 platform are not likely driven by this change due to the
small change in overall NOx emissions. We are continuing to evaluate nonroad equipment
emission rates as more data become available.
Hypothesis 3.5: National nonroad equipment population and activity are overestimated
Status: The uncertainty in nonroad population has been investigated and had modest impact on
air quality model NOx bias. An evaluation of nonroad activity is ongoing.
Updates to the growth of nonroad equipment populations were implemented in MOVES2014b
(U.S. EPA, 2018b). Nonroad equipment in MOVES includes off-highway mobile engines from
twelve broad economic sectors (including construction, agricultural, and lawn & garden) but
excludes locomotives, airplanes and commercial marine vessels which are handled by separate
emission models (U.S. EPA 2021c). Adjusting nonroad equipment populations has a moderate
impact on national NOx emissions, decreasing national nonroad NOx emissions by 7% in 2011.
EPA conducted a sensitivity test using the 2011 platform by adjusting the nonroad emissions to
reflect this change and it resulted in decreased NOx concentrations in the range of 0.5 ppb
across large urban areas in the northeast US for July 2011.
Nonroad NOx emissions from this update had a larger impact for more recent and future years.
Based on MOVES national runs, national nonroad NOx emissions decreased by about 13% in
2016 due to this update. The updated nonroad equipment population growth rates were
included in MOVES2014b and the 2016 emissions and air quality modeling platform evaluated
by Toro et al. (2021). The change in the mean NOx bias due to all changes in the 2015 and 2016
platforms varied across regions and seasons, increasing as much as 15 ppb in the summer in the
South and decreasing as much as 28 ppb in the summer in the Upper Midwest. Thus, while the
nonroad equipment population growth has been shown to have an important impact on NOx
concentrations (in the range of 0.5 ppb), it is only a minor contributing factor to the larger
changes observed in the 2016 platform NOx concentrations.
Representative and comprehensive nonroad equipment activity data are sparse and often non-
existent for certain sectors and types of equipment. EPA has initiated work to develop updated
28
-------
estimates of nonroad activity based on equipment activity loggers, auction house records, and
other data sources. It is anticipated that activity estimates likely have a similar impact on
nonroad emissions as nonroad equipment population and we are looking for data sources to
improve and evaluate nonroad activity.
7. Internal and external outreach activities
7.1 Cross-EPA coordination meetings
From 2015-2019, there were regularly scheduled NOx coordination calls which occurred 5-10 times
per year. On these calls, updates were provided on progress for specific projects and analyses. In
addition, calls included discussion of outstanding questions and uncertainties and prioritization of
new analyses.
7.2 Technical discussions on Emissions and Atmospheric Modeling (TEAM)
The Technical discussions on Emissions and Atmospheric Modeling (TEAM) team was formed as part
of a cross-agency coordination effort between national environmental agencies. The group was led
by representatives from EPA, NOAA and NASA and focused on communications between federal
agency staff on specific topics. The first topic area chosen for TEAM was the use of ambient
measurements and satellite to constrain NOx emissions. During this topic, there were four webinars
and three conference sessions. Members of the EPA NOx coordination group were substantial
participants in each webinar and conference session. The webinars included presentations by EPA
staff describing emission development techniques, NOAA staff describing fuel-based mobile NOx
inventories, NASA discussing satellite assets, and top-down constraint methodologies and results. All
three agencies presented results comparing their methodologies to field campaign data. The
conference sessions were held at the 2017 International Emissions Inventory Conference (IEIC), the
2017 Community Modeling and Analysis System (CMAS), and the 2017 American Geophysical Union
(AGU) Fall Meeting. These conference sessions included presentations from TEAM and EPA NOx
coordination group members, as well as contributions from others.
7.3 Seminars, scientific conference presentations and special sessions
EPA has shared results and engaged with the scientific and regulated communities through
continued participation in scientific conferences, workshops, and other outreach opportunities.
Members of the NOx evaluation group have organized and chaired four special sessions focused on
this topic at conferences which included 14 EPA presentations and 26 relevant presentations from
outside groups (Table 2). In addition, findings from this work have been presented in 24
presentations at 11 conferences (Tables 2 and 3). EPA staff have actively engaged in ongoing
discussions with researchers from varied institutions and with disparate points of view.
29
-------
Table 2: List of relevant conference presentations (talks and posters) from special sessions convened or
co-convened by members of the cross-EPA NOx evaluation group at four international scientific
conferences. Presentations given by EPA staff are shown in bold.
Conference Session
Title
Presenter
Presenter
Affiliation
22nd International
Emissions Inventory
Conference,
Baltimore, MD,
August 2017
Special Session:
Reconciling NOx
Emissions with
Ambient Observations
(Darrell Sonntag, U.S.
EPA & Greg Frost
NOAA)
Diurnal, Weekday and Long-term Patterns
in NOx Emissions Based on Decade-long
Time Series of Hourly AQS Data and
Comparison with Traffic Count Data
B. De Foy
St. Louis
University
Satellite N02 for the Evaluation of U.S. NOx
Emissions
M. Harkey
UW Madison
Evaluation of NOx Emissions and Modeling
B. Henderson
U.S. EPA-
OAQPS
MOVES-Based NOx Analyses for Urban Case
Studies in Texas
S. Bai
Sonoma
Technology
United States Light and Heavy-Duty Fuel
Specific On-Road NO and NOx Emission
Factor Trends and Their Importance in
Inventory Calculations
G. Bishop
University of
Denver
Modeling Ozone in the Eastern U.S. using a
Fuel-Based Mobile Source Emissions
Inventory
B. McDonald
NOAA
Comparison of Light-duty NOx Emission
Rates Estimated from MOVES with Real-
world Measurements
D.Sonntag
U.S. EPA-
OTAQ
Technical discussions on Emissions and
Atmospheric Modeling (TEAM)
B. Henderson
U.S. EPA-
OAQPS
16th Annual
Community Modeling
and Analysis System
Conference, Chapel
Hill, NC Oct 2017
Special Session:
Improving the
Characterization of
the Ambient NOy
Budget(Heather
Simon, U.S. EPA &
Darrell Sonntag U.S.
EPA)
Technical discussions on Emissions and
Atmospheric Modeling (TEAM)
B. Henderson
U.S. EPA-
OAQPS
Evaluation of NOx Emissions and Modeling
B. Henderson
U.S. EPA-
OAQPS
Reconciling modeled and observed upper
tropospheric N02forthe interpretation of
satellite measurements
R. Silvern
Harvard
University
Evaluation of Emissions of Nitrogen Oxides
in Houston, Texas Using Three-Dimensional
Aircraft Observations during the
DISCOVER-AQ 2013 Mission
J. Smith
Texas
Commission
on
Environmental
Quality
Real world emissions of NOx and other
pollutants in the Ft. McHenry tunnel
A. Khlystov
Desert
Research
Institute
MOVES-Based NOx Analyses for Urban Case
Studies in Texas
K. Craig
Sonoma
Technology
30
-------
Conference Session
Title
Presenter
Presenter
Affiliation
Modeling Ozone in the Eastern U.S. using a
Fuel-Based Mobile Source Emissions
B. McDonald
NOAA
Inventory
Comparison of light-duty gasoline NOx
emission rates estimated from MOVES
D.Sonntag
U.S. EPA-
OTAQ
with real-world measurements
Updates on Production of NOx by Lightning
K. Pickering
University of
Maryland
Influence of different canopy reduction
Helmholtz-
functions on biogenic NO emission
J.A. Arndt
Zentrum
patterns in northern Europe
Geeshthacht
Sensitivity of MOVES emissions
specifications on modeled air quality using
traffic data and near-road ambient
C. Owen
U.S. EPA-
OAQPS
measurements from the Las Vegas and
Detroit field studies
Unconventional Constraints on Nitrogen
Chemistry using DC3 Observations and
Q. Shu
University of
Florida
Trajectory-based Chemical Modeling
Evaluating CO:NOx in a near-road
environment using ambient data from Las
Vegas
H. Simon
U.S. EPA-
OAQPS
Sensitivity of MOVES-estimated vehicle
emissions to inputs when comparing to
real-world measurements
D.Sonntag
U.S. EPA-
OTAQ
CAMx Model Sensitivity Analysis of
Emissions Temporal Profiles; Impacts on
2011 Modeled NOx/NOY Concentrations
B. Timin
U.S. EPA-
OAQPS
Exploring differences in nitrogen oxides
overestimation at the seasonal and day-
of-week levels to understand potential
C. Toro
U.S. EPA-
OTAQ
relationships with mobile source emission
inventories.
Investigating modeling platform emissions
for grid cells associated with a near-road
C. Toro
U.S. EPA-
study site during a field campaign in Las
OTAQ
Vegas
American Geophysical
Impacts of Aging Emission Control Systems
University of
Union Fall Meeting,
on In-Use Heavy-Duty Diesel Truck
C. Preble
California -
New Orleans LA, Dec
Emission Rates
Berkeley
2017
Multi-Year On-Road Emission Factor Trends
M. Haugen
University of
Session: Leveraging
of Two Heavy-Duty California Fleets
Denver
Inventories,
Observations, and
Comparisons of MOVES Light-duty
Gasoline NOx Emission Rates with Real-
D. Choi
U.S. EPA-
OTAQ
Models to Improve
world Measurements
31
-------
Conference Session
Title
Presenter
Presenter
Affiliation
the Scientific Basis of
Emissions (Greg Frost,
NOAA; Barron
Henderson, U.S. EPA;
Barry Lefer, NASA)
Evaluation of NOx Emissions and Modeling
B. Henderson
U.S. EPA-
OAQPS
Eddy Covariance Measurements Assessing
NOx Emission in London, UK
w. s.
Drysdale
University of
York
Evaluation of a Fuel-Based Oil and Gas
Inventory of Nitrogen Oxides with Top-
Down Emissions
B. McDonald
NOAA
Update of NOx emission temporal profiles
using CMAQ-HDDM
C. Bae
Ajou
University
Technical discussions on Emissions and
Atmospheric Modeling (TEAM)
G. Frost
NOAA
American Geophysical
Union Fall Meeting,
Washington DC, Dec
2018
Session: Improving
the Science of
Emissions Through
Inventories,
Observations, and
Models (Greg Frost,
NOAA; Barron
Henderson, U.S. EPA;
Barry Lefer, NASA)
Satellite and Surface Observations Confirm
Steady Decline in US NOx Emissions over
the 2004-2017 Period
R. Silvern
Harvard
University
Unexpected slowdown of US pollutant
emission reduction in the past decade
H.M. Worden
NCAR
A large decline of tropospheric N02 in
China since 2013 observed from space by
SNPPOMPS
Y. Wang
University of
Houston
Recent Advances in Deriving NOx Emission
Estimates from Satellite Data
D. Goldberg
Argonne
National
Laboratory
Truck Exhaust Plume Capture and
Quantification of Nitrogen-Species
Emission Rates: Impact of Diesel Particle
Filters and Selective Catalytic Reduction
Systems
T. Kirchstetter
Lawrence
Berkeley
National
Laboratory
Estimates of global biogenic soil HONO
emissions using a process-oriented model
H. Su
Max Planck
Institute for
Chemistry
Quantification of Global Reactive Nitrogen
Emissions from Biomass Burning using
Satellite Observations
C. Bray
North Carolina
State
University
32
-------
Table 3: List of conference presentations or seminars presented by members of the cross-EPA NOx
evaluation on this topic (excluding presentations already listed in Table 2)
Conference
Title
Presenter
15th Annual Community
Modeling and Analysis
System Conference, Chapel
Hill, NC,
Oct 2016
An analysis of sensitivity of MOVES emissions
estimates to traffic data and comparison to
grid-cell estimates and near-road
measurements from the Las Vegas field study
C. Owen
Modeled Source Contributions to CO and NOy
Concentrations during the DISCOVER-AQ
Baltimore Field Campaign
H. Simon
In-depth examination of emissions
inventories to support EPA evaluation of
modeled ambient nitrogen oxides (NOx and
NOy)
C. Toro
CRC Real World Emissions
Workshop, Long Beach, CA,
March 2017
Evaluation of NOx Emissions Projected by
MOVES2014 Using Dynamometer, Remote-
Sensing and Tunnel Data
J. Warila
MOVES Review Workgroup
meeting, Ann Arbor, Ml,
September 2017
Update on MOVES model evaluation: NOx
D.Sonntag
EPA/ORD NERL
Computational Exposure
Division seminar series, Sep
2017
Ongoing EPA efforts to evaluate modeled
NOy budgets
H. Simon
EPA Environmental
Modeling Community of
Practice webinar
Ongoing EPA efforts to evaluate modeled
NOy budgets
H. Simon
MD and Northeast states
Weekly Photochemical
Modeling Coordination Call,
November 2017
Ongoing EPA efforts to evaluate modeled
NOy budgets
H. Simon
Comparing light-duty gasoline NOx emission
rates estimated with MOVES to real-world
measurements
D.Sonntag
CRC Real World Emissions
Workshop, Garden Grove,
CA, March 2018
Updated Evaluation of MOVES Light-duty
Gasoline NOx Emission Rates with Real World
Measurements
D.Sonntag
Atmospheric Chemical
Mechanisms Conference,
Davis, CA, December 2018
Ongoing EPA efforts to evaluate modeled
NOy budgets
H. Simon
CRC Real-World Emissions
Workshop, Long Beach, CA,
March 10-13, 2019
Updates to EPA's Motor Vehicle Emission
Simulator (MOVES)
M. Beardsley
MOVES Review Work Group,
Ann Arbor, Ml, April 2019
Updates to "high-power" emission rates and
start deterioration for light-duty vehicles
C. Toro
33
-------
2019 International Emissions
Inventory Conference -
Collaborative Partnerships
to Advance Science and
Policy, Dallas, TX, July 2019
MOVES Light-Duty Emission Rate Evaluation
in the Context of Reconciling Modeled and
Ambient NOx
C. Toro
Planned Updates to EPA's MOVES Emission
Model for Heavy-Duty Onroad Vehicles
J. Han
SMOKE version 4.7 Recent Enhancements
B. H. Baek/A. Eyth
18th Annual community
Modeling and Analysis
System Conference, Chapel
Hill, NC,
Oct 2019
Evaluation of CMAQ Estimated NOxfrom
2002 to 2016
K. Foley
American Geophysical Union
Fall Meeting, San Francisco,
CA, Dec 2019
Evaluation of CMAQ Estimated NOx from
2002 to 2016
H. Simon/K. Baker
Comprehensive bottom
up analysis of the onroad mobile emission
sector: from NOx
emission rates to air quality impacts
C. Toro
7.4 Journal publications
Several externally peer-reviewed journal publications resulted from this coordination effort.
Simon et al. (2018a) re-examined the methods used by Anderson et al. applying new modeling to
understand source contributions of CO and NOy during the DISOCVER-AQ Baltimore field campaign
and showed that those results were impacted by the choice of metrics used. This analysis using the
2011 emissions modeling platform showed reasonably good agreement between modeled and
measured concentrations of N02 aloft in the boundary layer (NMB = 8%) but model over-predictions
of aged nitrogen species (NMB = 69%, 118% and 18% for alkyl nitrates, peroxy nitrates and nitric
acid respectively). These findings suggest that chemistry plays a key role in how well model
predictions of total NOy compare against measurements. This analysis also shows that two different
measurement methods used in that field campaign ( 1) total NOy measured by chemiluminescence
instrument and 2) sum of NOy component species measured by chemiluminescence and TD-LIF
instruments) produced widely diverging NOy values and that model NOx performance could vary
substantially depending on which measurement was used (NMB for NOy = 69% when compared
against chemiluminescence measurements versus 50% when compared against the sum of
measured NOy component species). When normalized to CO, the modeled NOy was not statistically
different from the measured normalized ratio indicating transport and dispersion processes may
play an important role in NOy model performance issues.
Day et al. (2019) reviewed progress on emissions inventories that has occurred since a 2005 NARSTO
report which included recommended areas for improvement. While this article did not focus
specifically on NOx emissions it included several relevant sections, including a detailed description of
updates to mobile source emissions modeling that have occurred through incorporation of new
vehicle testing data into the MOVES model.
34
-------
Zare et al. (2019) improved the characterization of organic nitrogen species in the CMAQ model,
particularly solubility, formation, reaction rates and products. At a rural Alabama monitoring site
these updates reduced the model prediction of alkyl nitrate species from a factor of 2 overestimate
to a 32% underestimate and brought the model into better agreement with measurements.
Simon et al. (2020) evaluated various methods for calculating CO:NOx ratios from measurements
taken in near-road environments. For this purpose, Simon et al. leverage an extensive near-road
dataset collected next to an interstate in Las Vegas, NV between December 2008 and February
2010. When measured ratios are compared to those derived from MOVES using traffic data from
the field location, MOVES values are generally unbiased but the full variability seen in observations
was not captured by the model.
Toro et al. (2021) examined model NOx performance against AQS monitors for 15 years of modeled
data. The CMAQ model simulations for more recent years in the 15-year timeseries apply updated
emissions inventory methods and model versions which incorporate more recent science and more
detailed data. The paper has several key findings. First, it demonstrates the seasonal dependence
of the modeled NOx performance with over-predictions mainly focused on summer months and
under-predictions in the winter months. The analysis also highlighted the spatial nature of NOx
model bias with most of the over-predictions occuring in the Upper Midwest, Ohio Valley, South,
and Southeast regions of the U.S. Comparisons of the model to measurements made in other
regions of the U.S. tended to show model underpredictions of NOx. In addition, the analysis shows
that NOx over-predictions have decreased substantially or been eliminated in some locations in
recent years. Updates to chemical and physical processes in CMAQv5.1 were identified as a major
driver in reducing NOx over-predictions in summer and increased the NOx underpredictions in
winter. In addition, the decrease in over-prediction may also be due in part to decreasing measured
ambient NOx concentrations accompanied by large decreases in modeled NOx emissions
concentrations in the emissions and air quality modeling platforms with more recent calendar years.
The paper then looks further into the 2011 calendar year with additional 2011 CMAQ simulations
meant to examine sensitivity to model chemistry and emissions assumptions. These sensitivities
show several things. First, while timing of NOx emissions from various source categories can be
important to model NOx predictions at specific times and locations, adjusting temporal assumptions
had little overall impact on model NOx performance. Updates to projected nonroad equipment
population did reduce predicted NOx emissions and improved model performance. In addition, the
model performance was sensitive to the chemical mechanism applied with newer chemistry
resulting in substantially lower total NOy concentrations. Finally, the paper compared model
predictions to aircraft measurements from the DISCOVER-AQ Baltimore field campaign that have
been the focus of several previous studies that evaluated EPA NOx emissions estimates. Consistent
with the findings in Simon et al. (2018a), this comparison showed that the magnitude of model bias
for aloft NOy within the boundary layer was sensitive to both the ambient measurements that were
used and the model chemical mechanism (model NMB could range from 27% to 77%, depending on
these factors). The findings that measurement uncertainty and model chemical and physical
processes play key roles in model performance suggest that caution should be taken in using
modeled NOx bias to constrain NOx emissions.
35
-------
8. Conclusions
The evaluation of NOx model bias is complex and multi-faceted. The NOx bias varies significantly
geographically and temporally, including across hourly, seasonal, and yearly timescales. In addition,
updated emissions and air quality modeling platforms include an ensemble of changes, making it
difficult to attribute changes in the NOx bias to individual updates. As such, it was useful to
coordinate efforts across EPA offices to build upon expertise in emissions modeling, air quality
modeling, and ambient measurements. The coordinated effort to evaluate the NOx bias led to
improved methods to evaluate model bias, updated model inputs, as well as an increased
understanding of the impacts of model inputs and changes on NOx concentrations. Overestimation
of ambient concentrations of NOx has been substantially reduced or eliminated with the most
recent calendar year emissions and CMAQ air quality modeling platforms. One reason is that the
more recent modeling platforms correctly modeled a decrease in NOx emissions and ambient
concentrations which corresponded with real-world NOx concentration reductions. In addition, EPA
staff identified several key changes in the emissions and air quality modeling platform that reduced
the NOx model bias.
At this time, EPA staff have not found a unique solution for the overprediction of NOx
concentrations, but have identified several plausible hypotheses, while ruling out others, for the
NOx positive biases seen at certain times/locations in the modeling. Model overpredictions were
likely due to multiple compounding factors that each contribute to a portion of the bias. Based on
our review of the evidence, the most important was:
Planetary boundary layer (PBL) and vertical mixing algorithms in CMAQ led to too little
vertical mixing at certain times and in some locations. These algorithms have been
improved in CMAQv5.1 and later versions of CMAQ. These changes substantially
reduced the NOx bias, as well as the NOx diurnal bias pattern in simulations run with
more recent CMAQ versions.
We also demonstrated that there is important uncertainty in the model bias caused by NOx and NOy
measurement uncertainty, as well as the chemical mechanism used. Caution should be taken in
using modeled NOx bias to constrain NOx emissions or processes incorporated into air quality
modeling.
Through this effort, we identified aspects of the mobile source NOx emissions that were
overestimated in the evaluated air quality platforms. These could lead to important overestimation
of NOx emissions, but based on our analysis so far, only had a modest impact on the magnitude and
pattern of the bias in modeled NOx concentrations. We have identified and developed
improvements to the NOx mobile source emission inventory to address the summer overestimation
of mobile-source NOx, which include:
MOVES light-duty NOx emissions rates.
MOVES light-duty gasoline NOx emissions were reduced in MOVES3 compared
to MOVES2014b. The MOVES3 light-duty gasoline NOx rates are generally lower
due to updated modeling of high-load operation and updated deterioration
trends.
MOVES inputs in the 2011 National Emissions Inventory (NEI) and EPA platforms
36
-------
o MOVES inputs used in the 2011 NEI and EPA platform were not consistent with
the ambient datasets to which they were being compared. For example, speed
and acceleration assumptions were not consistent with vehicle activity at
remote sensing locations, long-haul truck hoteling activity was overestimated
nationally, and MOVES inputs did not accurately model local variability in age
distributions and car/truck splits. These inputs have been improved in
MOVES3, the 2016vl platform (U.S. EPA, 2021a) and the 2017 NEI.
National nonroad equipment populations were overestimated.
o Nonroad equipment populations were overestimated in the 2015 and earlier
platforms. These estimates were updated in the 2016 platform using updated
nonroad population and activity data incorporated into MOVES2014b. These
changes had a noticeable, but relatively small, impact on the NOx bias.
We believe continued collaboration, evaluation, and communication of the model bias within and
outside the EPA will lead to increased fidelity of future emissions and air quality modeling platforms.
9. Ongoing Work/Next Steps
EPA is continuing to update NOx emissions methods and input data and CMAQ model treatment of
physical and chemical processes. In addition, EPA is in the process of conducting an updated multi-
year modeling evaluation.
MOVES updates:
ฆ The EPA is continuing to evaluate and develop future MOVES versions using the latest
science and data. In November 2020, the EPA's Office of Transportation and Air Quality
released MOVES3 (U.S. EPA OTAQ, 2021c). MOVES3 contains significant updates that
impact NOx emissions, including:
ฆ Updated light-duty emissions rates based on new inspection and maintenance
(l/M) program data, remote sensing, and portable emission measurement
system data.
ฆ Updated start emissions as a function of parked time for light-duty and heavy-
duty vehicles
ฆ Updated heavy-duty emission rates for model year 2010 and later vehicles
based on manufacturer in-use testing data
ฆ Updated heavy duty emission rates for extended idling and auxiliary power units
ฆ Updated onroad activity, including idling activity, and vehicle populations
ฆ Accounted for glider vehicles (new heavy-duty vehicles with old engines)
ฆ Updated and increased resolution in heavy-duty vehicle masses using weigh-in-
motion and other datasets
ฆ Updated fuel properties based on latest fuel compliance data
ฆ Using MOVES3 code and national default inputs, national onroad vehicle NOx emissions
decreased by 8% in 2016 compared to MOVES2014b. Results will differ for different
calendar years, local areas, and for comparisons such at the NEI that use custom inputs
to represent local vehicle population and activity. EPA is evaluating the impact of using
37
-------
M0VES3 on developing national emission inventories and air quality as part of the
EQUATES project discussed below.
CMAQ chemistry updates (CRACMM): The chemical mechanism of an atmospheric chemical
transport model like the Community Multiscale Air Quality (CMAQ) system contains the
condensed set of reactions that describe the interactions between emitted hydrocarbons and
nitrogen oxides as well their reaction products. The mechanism affects the representation of
NOy including how it is partitioned among species and the gas and particle phase. Traditionally,
mechanisms in regional models were designed for the prediction of ozone under alkane-rich
urban atmospheres of the 1990s and are connected to independently developed modules and
metadata to compute fine particles and deposition among other endpoints. The Community
Regional Atmospheric Chemistry Multiphase Mechanism (CRACMM) is under development in
ORD and aims to couple gas- and particle-phase chemistry by treating the entire pool of
atmospheric reactive organic carbon (ROC) relevant to present-day emissions.
EQUATES: EPA's Air QUAIity TimE Series Project (EQUATES) is a cross-agency collaboration
between ORD, OAR/OAQPS, and OAR/OTAQto develop modeled meteorology, emissions, air
quality and pollutant deposition for 2002 through 2017. Modeled datasets cover the
Conterminous U.S. (CONUS) at a 12km horizontal grid spacing using WRFv4.1.1 for meteorology
and CMAQv5.3.2 for air quality modeling. New CONUS emissions inventories were developed
using (to the extent possible) consistent input data and methods across all years, including
MOVES3 modeling for onroad emissions. Evaluation of model estimated trends will be used to
inform model development and build confidence in the use of the modeling system to quantify
the impact of meteorological and emissions changes on air quality.
Disclaimer: The views expressed in this report are those of the authors and do not necessarily represent
the views or policies of the U.S. Environmental Protection Agency.
10. References
Anderson, D.C., Loughner, C.P., Diskin, G., Weinheimer, A., Canty, T.,P., Salawitch, R.J., Worden,
H.M., Fried, A., Mikoviny, T., Wisthaler, A., Dickerson, R.R. (2014). Measured and modeled CO and
NOy in DISCOVER-AQ: An evaluation of emissions and chemistry over the eastern US. Atmospheric
Environment, 96, 78-87.
Appel, K.W., Napelenok, S.L, Foley, K.M., Pye, H.O.T., Hogrefe, C., Luecken, D.J., Bash, J.O., Roselle,
S.J., Pleim, J.E., Foroutan, H., Hutzell, W.T., Pouliot, G.A., Sarwar, G., Fahey, K.M., Gantt, B., Gilliam,
R.C., Heath, N.K., Kang, D., Mathur, R., Schwede, D.B., Spero, T.L., Wong, D.C., Young, J.O. (2017)
Description and evaluation of the Community Multiscale Air Quality (CMAQ) modeling system
version 5.1. Geoscientific Model Development, 10, 1703-1732.
Beardsley, M., Brown, J., Han, J., Roberts, S., Sonntag, D. Updates to EPA's Motor Vehicle Emission
Simulator (MOVES), CRC Real-World Emissions Workshop, Long Beach, CA, March 10-13, 2019.
38
-------
Brioude, J., Kim, S.-W., Angevine, W.M., Frost, G.J., Lee, S.-H., McKeen, S.A., Trainer, M., Fehsenfeld,
F.C., Holloway, J.S., Ryerson, T.B., Williams, E.J., Petron, G., Fast, J.D. (2011), Top-down estimate of
anthropogenic emission inventories and interannual variability in Houston using a mesoscale inverse
modeling technique. Journal of Geophysical Res - Atmospheres, 116, D20305,
doi:10.1029/2011J DO 16215.
Canty, T.P., Hembeck, L, Vinciguerra, T.P., Anderson, D.C., Goldberge, D.L., Carpenter, S.F., Allen,
D.J., Loughner, C.P., Salawitch, R.J., Dickerson, R.R. (2015). Ozone and NOx chemistry in the eastern
US: evaluation of CMAQ/CB05 with satellite (OMI) data. Atmospheric Chemistry and Physics, 15,
10965-10982.
Choi, D., Sonntag, D., Warila, J., Toro, C., Beardsley, M., Brzezinski, D., Henderson, B., Timin, B., Eyth,
A., Driver, L. (2017). Comparisons of MOVES Light-duty Gasoline NOx Emission Rates with Real-world
Measurements. American Geophysical Union Fall Conference, New Orleans, LA, December 2017.
Day, M., Pouliot, G., Hunt, Sh., Baker, K.R., Beardsley, M., Frost, G., Mobley, D., Simon, H.,
Henderson, B.B., Yelverton, T., Rao, V. (2019) Reflecting on progress since the 2005 NARSTO
emissions inventory report. Journal of the Air & Waste Management Association, 9, 1025-1050.
Dickerson, RR, Anderson, DC, Ren, X. (2019) On the use of data from commercial NOx analyzers for
air pollution studies. Atmospheric Environment 214: 116873. DOI:
http://dx.doi.Org/10.1016/j.atmosenv.2019.116873.
Dunlea, EJ, Herndon, SC, Nelson, DD, Volkamer, RM, San Martini, F, Sheehy, PM, Zahniser, MS,
Shorter, JH, Wormhoudt, JC, Lamb, BK, Allwine, EJ. (2007) Evaluation of nitrogen dioxide
chemiluminescence monitors in a polluted urban environment. Atmospheric Chemistry and Physics
7(10): 2691-2704. DOI: http://dx.doi.org/10.5194/acp-7-2691-2007.
Foley, K.M., Simon, H., Toro, C., Baker, K., Appel, K.W., Henderson, B., Eyth, A., Luecken, D. (2019)
Evaluation of CMAQ Estimated NOx from 2002-2016. Community Modeling and Analysis System
18th Annual Conference, Chapel Hill, NC, October 2019.
Han, J., Sandhu, G., Sonntag, D., Bizer-Cox, D. (2019) Planned Updates to EPA MOVES Emission
Model for Heavy-Duty Onroad Vehicles. EPA 2019 International Emissions Inventory Conference:
Collaborative Partnerships to Advance Science and Policy, Dallas, TX, July 2019.
Han, J. (2021) MOVES3 Fuel Consumption Evaluation. MOVES Review Workgroup meeting, Ann
Arbor, Ml, September 2021. https://www.epa.gov/moves/moves-model-review-work-group
Henderson, B., Simon, H., Timin, B., Dolwick, P., Owen, C., Etyh, A., Foley, K., Toro, C., Baker, K.,
Appel, K.W. (2017a) Evaluation of NOx Emissions and Modeling, Community Modeling and Analysis
System (CMAS) Annual Conference, October 2017, Chapel Hill, NC
Henderson, B., Simon, H., Timin, B., Dolwick, P., Owen, C., Etyh, A., Foley, K., Toro, C., Baker, K.,
(2017b) Evaluation of NOx Emissions and Modeling, American Geophysical Union Fall Meeting,
December 2017, New Orleans, LA
Jaegle, L., Shah, V., Thornton, J.A., Lopez-Hilfiker, F.D., Lee, B.H., McDuffie, E.E., Fibiger, D., Brown,
S.S., Veres, P., Sparks, T.L., Ebben, C.J., Wooldridge, P.J., Kenagy, H.S., Cohen, R.C., Weinheimer, A.J.,
39
-------
Campos, T.L., Montzka, D.D., Digangi, J.P., Wolfe, G.M., Hanisco, T., Schroder, J.C., Campuzano-Jost,
P., Day, D.A., Jimenez, J.L., Sullivan, A.P., Guo, H., Weber, R.J. (2018) Nitrogen Oxides Emissions,
Chemistry, Deposition, and Export Over the Northeast United States During the WINTER Aircraft
Campaign. J. Geophys. Res. Atmos. 123, 12,368-12,393. doi:10.1029/2018JD029133
Kemball-Cook, S., Yarwood, G., Johnson, J., Dornblaser, B., Estes, M. (2015) Evaluating NOx emission
inventories for regulatory air quality modeling using satellite and air quality model data.
Atmospheric Environment, 117, 1-8.
Kota, S.H., Zhang, H.L., Chen, G., Schade, G.W., Ying, Q. (2014) Evaluation of on-road vehicle CO and
NOx National Emission Inventories using an urban-scale source-oriented air quality model.
Atmospheric Environment, 85,99-108.
Marr, L., Moore, T.O., Klapmeyer, M.E. Killar, M.B. (2013) Comparison of NOx fluxes measured by
eddy covariance to emission inventories and land use. Environmental Science & Technology, 47,
1800-1808.
McDonald, B.C., McKeen, S.A., Cui, Y.Y., Ahmadov, R., Kim, S.W., Frost, G.J., Pollack, I.B., Peischl, J.,
Ryerson, T.B., Holloway, J.S., Graus, M., Wameke, C., Gilman, J.B., de Gouw, J.A. Kaiser, J., Keutsch,
F.N., Hanisco, T.F., Wolfe, G.M., Trainer, M. (2018) Modeling ozone in the Eastern US using fuel-
based mobile source emissions inventory. Environmental Science & Technology, 52, 7360-7370.
Ngueyn, T.B., Crounse, J.D,. Teng, A.P., St. Clair, J.M., Poulot, F., Wolfe, G.M, Wennberg, P.O. (2015)
Rapid deposition of oxidized biogenic compounds to a temperate forest. Proceedings of the National
Academy of Sciences, 112(5), E392-E401, doi:10.1073/pnas,1418702112
Pleim, J., Ran, L. (2011) Surface flux modeling for air quality applications. Atmosphere, 2, 271-302,
doi:10.3390/atmos2030271
Ramboll, 2020, User's Guide Comprehensive Air quality Model with extensions Version 7.10,
Ramboll US Corporation, December ,2020. available at: https://camx-
wp.azurewebsites.net/Files/CAMxUsersGuide_v7.10.pdf
Salmon, O.E., Shepson, P.B., Ren, X., He, H., Hall, D.L., Dickerson, R.R., Stirm, B.H., Brown, S.S.,
Fibiger, D.L., McDuffie, E.E., Campos, T.L., Gurney, K.R., Thornton, J.A. (2018) Top-Down Estimates of
NO x and CO Emissions From Washington, D.C.-Baltimore During the WINTER Campaign. J. Geophys.
Res. Atmos. doi:10.1029/2018JD028539
Seinfeld, J.H. & Pandis, S. N. (2006). Atmospheric Chemistry and Physics: From Air Pollution to
Climate Change. Hoboken, N.J: J. Wiley.
Simon, H., Valin, L.C., Baker, K.R., Henderson, B.H., Crawford, J.H., Pusede, S.E., Kelly, J.T., Foley,
K.M., Owen, R. C., Cohen, R.C., Timin, B., Weinheimer, A.J., Possiel, N., Misenis, C., Diskin, G.S., Fried,
A. (2018a) Characterizing CO and NOy sources and relative ambient ratios in the Baltimore area
using ambient measurements and source attribution modeling. Journal of Geophysical Research -
Atmospheres, 123, 3304-3320, https://doi.org/10.1002/2017JD027688
Simon, H., Henderson, B., Luecken, D., Foley, K. (2018b) Ongoing EPA efforts to evaluate modeled
NOy budgets, Atmospheric Chemical Mechanisms Conference, Davis, California, December 2018.
40
-------
Simon, H., Foley, K.M. Toro, C., Baker, K., Appel, K.W., Henderson, B., Eyth, A., Luecken, D. (2019)
Evaluation of CMAQ Estimated NOxfrom 2002 - 2016. American Geophysical Union Fall Conference,
San Francisco, CA, December 2019.
Simon, H., Henderson, B.H., Owen, R.C., Foley, K., Snyder, M.G., Kimbrough, S (2020) Variability in
Observation-Based Onroad Emission Constraints from a Near-Road Environment, 2020 Atmosphere,
11 (11), 1243, https://doi.org/10.3390/atmoslllll243
Sonntag, D., Choi, D., Warila, J., Beardsley, M. (2017). Comparison of Light-Duty NOx Emission Rates
Estimated from MOVES with Real World Measurements. 22nd International Emissions Inventory
Conference, Baltimore, MD, August 2017.
Souri, A.H., Choi, Y.S., Jeon, W.B., Li, X.S., Pan, S., Diao, L.J., Westenbarger, D.A. (2016) Constraining
NOx emissions using satellite N02 measurements during 2013 DISOCVER-AQTexas campaign.
Atmospheric Environment, 131, 371-381.
Sparks, J.P. (2009) Ecological ramifications of the direct foliar uptake of nitrogen. Oecologia, 159, 1-
13, DOI 10.1007/s00442-008-l 188-6
Timin, B., Dolwick, P., Simon, H., Possiel, N., Eyth, A., Vukovich, J., Roberts, S., Toro, C. (2017) CAMx
Model Sensitivity Analysis of Emissions Temporal Profiles: Impacts on 2011 Modeled NOx/NOy
Concentrations. Community Modeling and Analysis System 16th Annual Conference, Chapel Hill, NC,
October 2017.
Toro, C., Sonntag, D., Warila, J., Choi, D., Beardsley, M. (2019a) MOVES Light-Duty Emission Rate
Evaluation in the Context of Reconciling Modeled and Ambient NOx. EPA 2019 International
Emissions Inventory Conference: Collaborative Partnerships to Advance Science and Policy, Dallas,
TX, July 2019.
Toro, C., Sonntag D., Warila, J. Choi, D., Beardsley, M., Simon, H., Foley, K., Appel, K.W., Eyth, A.,
Phillips, S., Timin, B., Possiel, N. (2019b). Comprehensive bottom-up analysis of the onroad mobile
emission sector: from NOx emission rates to air quality impacts. American Geophysical Union Fall
Conference, San Francisco, CA, December 2019.
Toro, C., Foley, K., Simon, H., Henderson, B., Baker, K.R., Eyth, A., Timin, B., Appel, K.W., Luecken, D.,
Beardsley, M., Sonntag, D., Possiel, N., Roberts, S. (2021) Evaluation of 15 years of modeled NOx
across the contiguous United States. Elementa: Science of the Anthropocene, 9(1). DOI:
https://doi.org/10.1525/elementa.2020.00158.
Travis, K.R., Jacob, D.J., Fisher, J.A., Kim, P.S., Marais, E.A., Zhu, L., Yu, K., Miller, C.C., Yantosca, R.M.,
Sulprizio, M.P., Thompson, A.M., Wennberg, P.O., Crounse, J.D., St Clair, J.M., Cohen, R.C., Laughner,
J.L., Dibb, J.E., Neuman, J.A., Zhou, X.L. (2017) Why do models overestimate surface ozone in the
Southeast United States? Atmospheric Chemistry and Physics, 16, 13561-13577.
Turnipseed, A.,A., Huey, L.G., Nemitz, E., Stickel, R., Higgs, J., Tanner, D.J., Slusher, D.L., Sparks, J.P.,
Flocke, F., Guenther, A. (2006) Eddy covariance fluxes of peroxyacetyl nitrates (PANs) and NOy to a
coniferous forest. J. Geophys. Res., Ill, D09304, doi:10.1029/2005JD006631
41
-------
U. S. EPA (2002) User's Guide to MOBILE6.1 and MOBILE6.2: Mobile Source Emission Factor Model.
EPA420-R-02-028. https://nepis.epa.gov/Exe/tiff2png.cgi/P1001DSE.PNG?-r+75+-
g+7+D%3A%5CZYFlLES%5ClNDEX%20DATA%5C00THRU05%5CTlFF%5C00001179%5CP1001DSE.TlF
U.S. EPA (2009) EPA Releases MOVES2010 Mobile Source Emissions Model: Questions and Answers.
EPA-420-F-09-073. December 2009, https://nepis.epa.gov/Exe/ZvPDF.cgi?Dockev=P1005ZAJ.pdf
U.S. EPA (2013) 2008 National Emissions Inventory, version 3, Technical Support Document DRAFT.
Air Quality Assessment Division, Office of Air Quality Planning & Standards, Research Triangle Park,
NC, September 2013, https://www.epa.gov/sites/default/files/2Q15-
)cuments/2QQ8 neiv3 tsd draft.pdf
U.S. EPA (2014) EPA Releases MOVES2014 Mobile Source Emissions Model. EPA-420-F-14-049. July
2014, https://nepis.epa.gov/Exe/ZvPDF.cgi?Dockev=P100JWJ5.pdf
U.S. EPA (2015) 2011 National Emissions Inventory, version 2 Technical Support Document. Air
Quality Assessment Division, Office of Air Quality Planning & Standards, Research Triangle Park, NC,
August 2015, https://www.epa.gov/sites/default/files/2015"
10/documents/nei2011v2. tsd 14aug2015.pdf
U.S. EPA (2018a). Exhaust and Crankcase Emission Factors for Nonroad Compression-Ignition
Engines. EPA-420-R-18-009. Assessment and Standards Division, Office of Transportation and Air
Quality, U.S. Environmental Protection Agency, Ann Arbor, Ml. July, 2018.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P100UXEN.pdf
U.S. EPA (2018b). Nonroad Engine Population Growth Estimates for MOVES2014b. EPA-420-R-18-
010. Assessment and Standards Division, Office of Transportation and Air Quality, Ann Arbor, Ml.
July, 2018. https://nepis.epa.gov/Exe/ZvPDF.cgi?Dockev=P100UXJK.pdf
U.S. EPA (2018c) Technical Support Document (TSD) Preparation of Emissions Inventories for the
Version 7.1 2014 Emissions Modeling Platform for the National Air Toxics Assessment.
https://www.epa.gov/sites/production/files/2018-08/documents/2014 114 emismod tsd.pdf
U.S. EPA (2018d) 2014 National Emissions Inventory, version 2 Technical Support Document. Air
Quality Assessment Division, Office of Air Quality Planning & Standards, Research Triangle Park, NC,
July 2018, https://www.epa.gov/sites/default/files/2018-
)cuments/nei2014v2 tsd 05jul2018.pdf
U.S. EPA (2019), Technical Support Document (TSD) Preparation of Emissions Inventories for the
Version 7.2 2016 North American Emissions Modeling Platform.
https://www.epa.gov/sites/production/files/2019-
09/documents/2016v7.2 regionalhaze emismod tsd 508.pdf
U.S. EPA (2020a). Exhaust Emission Rates of Heavy-Duty Onroad Vehicles in MOVES3. EPA-420-R-
20-018. Office of Transportation and Air Quality, Ann Arbor, Ml. November 2020.
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=P1010MC2.pdf
42
-------
U.S. EPA (2020b). Exhaust Emission Rates for Light-Duty Onroad Vehicles in MOVES3. EPA-420-R-
20-019. Office of Transportation and Air Quality, Ann Arbor, Ml. November 2020.
https://nepis.epa.gov/Exe/ZvPDF.cgi?Dockev=P1010MKM.pdf
U.S. EPA (2021a) Preparation of Emissions Inventories for 2016vl North American Emissions
Modeling Platform Technical Support Document. Office of Air Quality Planning and Standards, Air
Quality Assessment Division Emissions Inventory and Analysis Group, Research Triangle Park, NC,
March 2021. https ://www.epa.gov/csapr/preparation-emissions-inventories-2016vl-north-
american-emissions-modeling-platform-technical
U.S. EPA (2021b). CMAQ. Office of Research and Development, 2021. doi:10.5281/zenodo.107987
U.S. EPA (2021c). Overview of EPA's MOtor Vehicle Emission Simulator (MOVES3). EPA-420-R-21-
004. Office of Transportation and Air Quality, Ann Arbor, Ml, March 2021,
https://www.epa.gov/moves/moves-technical-reports.
U.S. EPA. (2021d) Population and Activity of Onroad Vehicles in MOVES3. EPA-420-R-21-012. Office
of Transportation and Air Quality, Ann Arbor, Ml, April 2021,
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockev=P1011TF8.pdf
U.S. EPA (2021e) 2017 National Emissions Inventory: January 2021 Updated Release, Technical
Support Document. EPA-454/R-21-001. Air Quality Assessment Division, Office of Air Quality
Planning & Standards, Research Triangle Park, NC, February 2021,
https://www.epa.gov/sites/defauIt/files/2021-02/documents/nei2( I jan2021.pdf
Vasquez, K.T., Crounse, J.D., Schulze, B.C., Bates, K.H., Teng, A.P., Xu, L., Allen, H.M., and Wennberg,
P.O. (2020) Rapid hydrolysis of tertiary isoprene nitrate efficiently removes NOxfrom the
atmosphere, PNAS, 117 (52), 33011-33016.
Wolfe, G.M., Thornton, J.A., Yatavelli, R.L.N., McKay, M., Goldstein, A.H., LaFranchi, B., Min, K.-E.,
Cohen, R.C. (2009) Eddy covariance fluxes of acyl peroxy nitrates (PAN, PPN, and MPAN) above a
Ponderosa pine forest. Atmos. Chem. Phys., 9, 615-635, doi:10.5194/acp-9-615-2009
Wu, Z., Wang, X., Turnipseed, A.A., Chen, F., Zhang, L., Guenther, A.B., Karl, T., Huey, L.G., Niyogi, D.,
Xia, B., Alapaty, K. (2012) Evaluation and improvements of two community models in simulating dry
deposition velocities of peroxyacetyl nitrate (PAN) over a coniferous forest. J. Geophys. Res., 117,
D04310, doi:10.1029/2011JD016751
Zare, A., Fahey, K.M., Sarwar, G., Cohen, R.C., Pye, H.O.T. (2019) Vapor-Pressure Pathways Initiate
but Hydrolysis Products Dominate the Aerosol Estimated from Organic Nitrates, ACS Earth Space
Chem, 3,8, 1426-1437.
43
-------
11. Appendix A: Acronyms
CMAQ - Community Multiscale Air Quality Model
CAMx - Comprehensive Air Quality Model with extensions
EQUATES - EPA's Air Quality Time Serie Project
MAR - Mileage accumulation rate
MOVES - Motor Vehicle Emission Simulator
NOx- nitrogen oxides: NO + N02
NOy- Reactive nitrogen compounds including: NO, N02, HN03, N03, HONO, N205, CIN02, PAN, other
organic nitrates, particulate nitrate
NOz- NOy-NOx
OAQPS - U.S. EPA Office of Air Quality Standards and Planning
OTAQ- U.S. EPA Office of Transportation and Air Quality
PBL- Planetary Boundary Layer
SMOKE - Sparse Matrix Operator Kernel Emissions modeling system
44
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-454/R-21-008
Environmental Protection Air Quality Assessment Division November 2021
Agency Research Triangle Park, NC
------- |