Technical Support Document (TSD) Preparation of Emissions Inventories for the Version 7.1 2016 Hemispheric Emissions Modeling Platform


Technical Support Document (TSD)
Preparation of Emissions Inventories for the Version 7.1
2016 Hemispheric Emissions Modeling Platform

November 2019

Contacts:

Jeff Vukovich, Alison Eyth

U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Emissions Inventory and Analysis Group
Research Triangle Park, North Carolina

-------
TABLE OF CONTENTS

LIST OF TABLES	Ill

LIST OF FIGURES	Ill

ACRONYMS	IV

1	INTRODUCTION	1

2	HEMISPHERIC EMISSION INVENTORIES AND MODELING APPROACHES	4

2.1	Global HT AP version 2	10

2.1.1	Spatial Allocation	11

2.1.2	Chemical Speciation	12

2.1.3	Temporal Allocation	13

2.1.4	Vertical Allocation	14

2.1.5	Projection to 2014 via CEDS	15

2.2	THU China	17

2.2.1	Temporal Allocation	19

2.2.2	Chemical Speciation	21

2.2.3	Spatial Allocation	21

2.2.4	Vertical AI location	21

2.3	Global FINN fires (g_ptfire, g_ptagfire)	21

2.4	Other global emissions	23

2.4.1	Biogenic emissions	23

2.4.2	Lightning NOx emissions	25

2.4.3	Ocean chlorine emissions	26

2.5	Changes to North America sectors to support hemispheric modeling	26

2.5.1	Emissions inventory differences	26

2.5.2	Spatial Allocation	27

2.5.3	Vertical Allocation	28

2.6	Final CMAQ-Ready Hemispheric Emissions	28

3	SENSITIVITY RUNS	29

3.1 Sensitivity Run Specifications	30

3.1.1	ZANTH	30

3.1.2	ZROW.	32

3.1.3	ZUSA	33

3.1.4	ZSHIP	35

3.1.5	ZCHN	35

3.1.6	Z1ND	36

3.1.7	ZCANMEX	37

3.1.8	ZFIRE	38

3.1.9	EdgarCHN	38

4	EMISSION SUMMARIES	39

5	REFERENCES	42

11

-------
List of Tables

Table 1-1. Description of the hemispheric platform grid	1

Table 2-1. Sectors for the 2016 hemispheric modeling platform	4

Table 2-2. Summary of emissions differences between 2015 spinup and 2016 base case emissions by sector8

Table 2-3. HTAP sector mappings to platform sectors	10

Table 2-4. Emissions totals by HTAP sector, 108km hemispheric domain, tons/year for 2016	11

Table 2-5. Mapping of CEDS sectors to HTAP sectors	15

Table 2-6. List of THU China sectors and equivalent HTAP sectors	18

Table 2-7. Annual emissions totals by THU China sector, tons/year	18

Table 2-8. Emissions totals for global fire sectors, 108km hemispheric domain, year 2016, tons/year	23

Table 2-9. Annual biogenic emissions totals, 108km hemispheric domain, year 2016, tons/year	25

Table 2-10. Hemispheric spatial allocation approach for all North America sectors	27

Table 3-1. Hemispheric sensitivity run sector table summary	29

Table 3-2. Fraction of 2016 emissions from wildfire and prescribed burns in USA	31

Table 4-1. Domain total emissions by sector for 2016 hemispheric platform, tons/year	39

Table 4-2. Domain total emissions for 2016 hemispheric sensitivity cases, tons/year	41

List of Figures

Figure 1-1. Sample plot of Emissions on the 108km hemispheric modeling grid (log-scale tons/day)	2

Figure 2-1. Map of GEOCODE Country Codes	12

Figure 2-2. Day-of-week temporal profiles for HTAP sectors	13

Figure 2-3. Hour-of-day temporal profiles for HTAP sectors	14

Figure 2-4. Layer fractions for HTAP sectors	14

Figure 2-5. Projection factors from year 2010 to 2014 for POWER(left) and TRANSPORT(right) sectors for

NOx emissions derived from CEDS	16

Figure 2-6. Time allocation factors for the China Inventory. Averages of province-specific allocation

factors are shown	20

Figure 2-7. Diurnal temporal profiles for global wild fires and ag fires	22

Figure 2-8. Layer fractions for global wild fires (g_ptfire) and ag fires (g_ptagfire)	23

Figure 2-9. Spatial extent of 12US1 BEIS emissions on the 108km hemispheric domain	24

Figure 2-10. Lightning NOx Annual emissions	26

Figure 2-11. Annual NOx(left) and S02(right) emissions for base case run	28

Figure 3-1. Annual NOx(left) and S02(right) emissions for ZANTH sensitivity run	32

Figure 3-2. Annual NOx(left) and S02(right) emissions for ZROW sensitivity run	33

Figure 3-3. Annual NOx(left) and S02(right) emissions for ZUSA sensitivity run	34

Figure 3-4. Percent difference from basecase NOx(left) and S02(right) weekday-summer emissions for

ZSHIP sensitivity run	35

Figure 3-5. Annual NOx(left) and S02(right) emissions for ZCHN sensitivity run	36

Figure 3-6. Annual NOx(left) and S02(right) emissions for ZIND sensitivity run	36

Figure 3-7. Annual NOx(left) and S02(right) emissions for ZCANMEX sensitivity run	37

Figure 3-8. Annual NOx(left) and S02(right) emissions for ZFIRE sensitivity run	38

Figure 3-9. Percent difference NOx(left) and S02(right) weekday-summer emissions for EdgarCHN

sensitivity run	39

in

-------
Acronyms

AE5	CMAQ Aerosol Module, version 5, introduced in CMAQ v4.7

AE6	CMAQ Aerosol Module, version 6, introduced in CMAQ v5.0

NBAFM	Naphthalene, Benzene, Acetaldehyde, Formaldehyde and Methanol

BEIS	Biogenic Emissions Inventory System

BELD	Biogenic Emissions Land use Database

C1/C2	Category 1 and 2 commercial marine vessels

C3	Category 3 (commercial marine vessels)

CAP	Criteria Air Pollutant

CARB	California Air Resources Board

CB05	Carbon Bond 2005 chemical mechanism

CEMS	Continuous Emissions Monitoring System

CI	Chlorine

CMAQ	Community Multiscale Air Quality

CMV	Commercial Marine Vessel

CO	Carbon monoxide

EBAFM	Ethanol, Benzene, Acetaldehyde, Formaldehyde and Methanol

ECA	Emissions Control Area

EEZ	Exclusive Economic Zone

EGU	Electric Generating Units

EIS	Emissions Inventory System

EPA	Environmental Protection Agency

EMFAC	Emission Factor (California's onroad mobile model)

FF10	Flat File 2010

FIPS	Federal Information Processing Standards

FHWA	Federal Highway Administration

HAP	Hazardous Air Pollutant

HC1	Hydrochloric acid

Hg	Mercury

IMO	International Marine Organization

MCIP	Meteorology-Chemistry Interface Processor

MOVES	Motor Vehicle Emissions Simulator

NEI	National Emission Inventory

NESCAUM Northeast States for Coordinated Air Use Management

NESHAP	National Emission Standards for Hazardous Air Pollutants

NH3	Ammonia

NIF	NEI Input Format

NLCD	National Land Cover Database

nm	nautical mile

NOAA	National Oceanic and Atmospheric Administration

NOx	Nitrogen oxides

ORD	EPA's Office of Research and Development

PM2.5	Particulate matter less than or equal to 2.5 microns

PM10	Particulate matter less than or equal to 10 microns

ppb, ppm	Parts per billion, parts per million

RBT	Refinery to Bulk Terminal

RWC	Residential Wood Combustion

RVP	Reid Vapor Pressure

SCC	Source Classification Code

iv

-------
SMARTFIRE Satellite Mapping Automated Reanalysis Tool for Fire Incident Reconciliation

SMOKE

Sparse Matrix Operator Kernel Emissions

SOi

Sulfur dioxide

SOA

Secondary Organic Aerosol

TOG

Total Organic Gas

TSD

Technical support document

VOC

Volatile organic compounds

WRF

Weather Research and Forecasting Model

v

-------
1 Introduction

The U.S. Environmental Protection Agency (EPA) developed an air quality emissions modeling platform
for hemispheric modeling with the Community Multiscale Air Quality (CMAQ) for the year 2016. This
emission modeling platform represents an update to the 2011 platform described by Eyth et al. (2016).
Similar to that work, this platform uses publicly available emissions that cover the globe and integrates
these emissions with regional inventories to create a single consistent platform.

The hemispheric modeling platform includes the following components:

In the United States, Canada, and Mexico, emissions are based on the 2016 alpha platform for
regional scale air quality modeling. Details regarding the North America portion of the
hemispheric emissions platform are provided in the 2016 v7.1 (also known as alpha) platform
Technical Support Document).

Outside North America, fire emissions are from the Fire Inventory from NCAR (FINN) for the
year 2016. This includes wildfires, prescribed burning, and agricultural fires.

In China, anthropogenic emissions excluding fires and carbon monoxide are from an emissions
dataset provided by Tsinghua University (THU) for year 2015.

Outside of North America and China, anthropogenic emissions excluding fires use the
Hemispheric Transport of Air Pollution (HTAP) emissions inventory, version 2. HTAP emissions
are for the year 2010, and are adjusted to year 2014 using factors derived from the Community
Emissions Data System (CEDS).

- Biogenic emissions in most of North America are equivalent to those from the regional alpha
platform and are generated by the BEIS model. Elsewhere in the modeling domain, biogenic
emissions are from the Model of Emissions and Gases and Aerosols from Nature (MEGAN).

Hemispheric air quality modeling is performed on a polar stereographic grid which is centered on the
North Pole, extends south to the equator, and has an approximate resolution of 108-km by 108-km per
grid cell. Table 1-1 includes the description (i.e., specification) of the hemispheric grid. Figure 1-1 shows
a visualization of NOx emissions on the 108km hemispheric modeling grid.

Table 1-1. Description of the hemispheric platform grid

Common
Name

Grid
Cell Size

Description

Grid name

Parameters listed in SMOKE grid
description (GRIDDESC) file:
projection name, xorig, yorig, xcell,
ycell, ncols, nrows, nthik*

Hemispheric
108km grid

108 km

Polar stereographic
grid of the Northern
Hemisphere

HEMI_108k

'POL HEMI', -10098000, -10098000,
108.D3, 108.D3, 187, 187, 1

* Corresponds to P

IOJ4 definition "+proj=stere +lat 0=90.0 +lat ts=45.0 +lon 0=-98.0

+y_0=l 0098000.0 +x_0=10098000.0 +a=6370000.0 +b=6370000.0 +to_meter= 108000 +no_defs"

-------
Figure 1-1. Sample plot of Emissions on the 108kin hemispheric modeling grid (log-scale tons/day)
2016fe hemi annual emissions total: mrqqrid NOX

tons/year
0 to 1.00
1.00 to 10.0
^ 10.0 to 100
¦¦ 100 to 1000
^ 1000 to 10000
^ 10000 +

The Community Multiscale Air Quality (CMAQ) model was used to model ozone (O3) and particulate
matter (PM) for this project. CMAQ includes support for hemispheric modeling, sometimes referred to as
H-CMAQ modeling, and can be used to develop boundary conditions (BCs) for regional scale air quality
modeling. CMAQ requires hourly and gridded emissions of the following inventory pollutants: carbon
monoxide (CO), nitrogen oxides (NOx), volatile organic compounds (VOC), sulfur dioxide (SO2),
ammonia (NH3), particulate matter less than or equal to 10 microns (PM10), and individual component
species for particulate matter less than or equal to 2.5 microns (PM2.5). In addition, the Carbon bond
version 6 (CB6) with chlorine chemistry used here within CMAQ allows for explicit treatment of the
YOC HAPs naphthalene, benzene, acetaldehyde, formaldehyde and methanol (NBAFM) and includes
anthropogenic HAP emissions of HC1 and CI. However, most emissions inventories in this platform
outside of the United States do not include any HAP emissions.

The primary emissions modeling tool used to create the air quality model-ready emissions was the Sparse
Matrix Operator Kernel Emissions (SMOKE) modeling system version 4.5 (SMOKE 4.5) with some
updates. Most regional scale CMAQ modeling is performed with "inline" plume rise, in which plume rise
calculations for point sources are performed by CMAQ using stack parameters and meteorology.

However, because the HTAP and Till.' China inventories do not include information on individual stacks,
it is not possible to run H-CMAQ with the inline plume rise option. Instead, we must create layered

2

-------
emissions within SMOKE by running the Laypoint program for sources with stack information, and by
running the Layalloc program to apply a fixed vertical allocation to sectors which do not have stack
information.

The hemispheric modeling platform includes multiple sets of model-ready emissions, or "cases":
An annual "base case" for year 2016, abbreviated 2016fe_hemi_cb6_16jh.

An annual "spinup case" for year 2015, abbreviated 2016fe_spinup_cb6_15jh. H-CMAQ
modeling is performed with an 8-month spinup period, and the emissions for the spinup period are
based on an entirely separate emissions case specific to the year 2015. For quality assurance and
for other modeling applications, the spinup case emissions were processed for all of 2015, not just
for the 8-month period required for spinup of 2016 modeling.

Several sensitivity runs in which one or more categories of emissions are zeroed out or modified,
as described in Section 3.

The gridded meteorological model used to provide input data for the emissions modeling was developed
using the Weather Research and Forecasting Model (WRF version 3.8, Advanced Research WRF core
(Skamarock, et al., 2008). The WRF Model is a mesoscale numerical weather prediction system
developed for both operational forecasting and atmospheric research applications. The WRF was run for
2015 and 2016 over the 108-km resolution hemispheric modeling domain with 44 vertical layers. The
WRF run for this platform is given the EPA meteorological case label "16jh." The full case name
includes this abbreviation following the emissions portion of the case name to fully specify the name of
the case as "2016fe_hemi_cb6_16jh."

The CMAQ model requires hourly emissions of specific gas and particle species for the horizontal and
vertical grid cells contained within the modeled region (i.e., modeling domain). To provide emissions in
the form and format required by the model, it is necessary to "pre-process" the "raw" emissions (i.e.,
emissions input to SMOKE). In brief, the process of emissions modeling transforms the emissions
inventories from their original temporal resolution, pollutant resolution, and spatial resolution into the
hourly, speciated, gridded resolution required by the air quality model. Emissions modeling includes
temporal allocation, spatial allocation, and pollutant speciation. For hemispheric modeling, emissions
modeling also includes the vertical allocation of point sources.

The temporal resolutions of the emissions inventories input to SMOKE vary across sectors and may be
hourly, daily, monthly, or annual total emissions. The spatial resolution may be individual point sources,
county/province/municipio totals, or pre-gridded emissions and varies by sector. This section provides
some basic information about the tools and data files used for emissions modeling as part of the modeling
platform.

For most sectors, SMOKE version 4.5 was used to process the raw emissions inventories into emissions
inputs for each modeling sector into a format compatible with CMAQ. For hemispheric modeling, plume
rise is not calculated "inline" during CMAQ modeling; instead, the emissions output from SMOKE and
input to CMAQ are 3-D emissions files and already have plume rise applied. For QA of the emissions
modeling steps, emissions totals by specie for the entire model domain are output as reports that are then
compared to reports generated by SMOKE on the input inventories to ensure that mass is not lost or
gained during the emissions modeling process. Preparation of China emissions from the THU dataset
(section 2.2), biogenic emissions from MEGAN (section 2.4.1), and lightning NOx emissions (section
2.4.2) was not performed in SMOKE, but with other utilities as described in the relevant sections below.

-------
The chemical mechanism used for the 2014 platform is the CB6 mechanism (Yarwood, 2010). We used a
particular version of CB6 that we refer to here as "CB6-CMAQ" that breaks out naphthalene (NAPH)
from XYL and PAR as an explicit model species, resulting in model species NAPH and XYLMN instead
of XYL, and revising PAR to remove the naphthalene portion (very small amount). CB6-CMAQ also
uses SOAALK (Pye and Pouliot, 2012), a species produced from TOG speciation that is not used in
CAMX. This platform generates the PM2.5 model species associated with the CMAQ Aerosol Module
version 6 (AE6).

Section 2 of this document describes the global inventories input to SMOKE and preparation of the
emissions for air quality modeling. Section 3 outlines several sensitivity cases in which new sets of
emissions were modeled with a portion of the emissions zeroed out. Data summaries are provided in
Section 4.

2 Hemispheric Emission Inventories and Modeling Approaches

This section describes the emissions data and processing that make up the 2016 hemispheric platform,
excluding the United States, Canada, and Mexico inventories that make up the corresponding regional
modeling platform. Documentation for the North America portion of the platform is in the 2016 alpha
platform TSD.

For the purposes of preparing the air quality model-ready emissions, the platform consists of several
sectors for emissions modeling. The significance of an emissions modeling or platform "sector" is that
the data are run through the SMOKE programs independently from the other sectors except for the final
merge (Mrggrid). The final merge program combines the sector-specific gridded, speciated, hourly
emissions together to create CMAQ-ready emission inputs.

Table 2-1 lists all sectors in the hemispheric platform and their data sources. For North America sectors
originating from the alpha platform for regional modeling, only a brief description of the sector is offered
here, with additional detail in the regional modeling TSD. The platform sector abbreviations are provided
in italics. These abbreviations are used in the SMOKE modeling scripts, inventory file names, and
throughout the remainder of this document. Most sectors which include global (e.g. HTAPv2) emissions
include a "g_" prefix.

Table 2-2 provides a brief by-sector outline of the differences between the emissions used in the 2016
base case and the emissions used in the 2015 spinup case.

Table 2-1. Sectors for the 2016 hemispheric modeling platform

Platform Sector:

abbreviation

Data
Source

Description and resolution of the data input to SMOKE

U.S. EGUs:

ptegu

Alpha
Platform

Electric Generating Units (EGUs) from the US NEI. Annual
resolution, with hourly resolution for sources that match Continuous
Emission Monitoring (CEM) data.

U.S. point source oil
and gas:
pt oilgas

Alpha
Platform

Oil and gas production point sources from the US NEI. Includes
emissions from oil rigs in the Gulf of Mexico. Annual resolution.

U.S. remaining non-
EGU point:

ptnonipm

Alpha
Platform

All point sources from the US NEI not matched to the ptegu or
pt_oilgas sectors. Annual resolution.

-------
Platform Sector:

abbreviation

Data
Source

Description and resolution of the data input to SMOKE

U.S. agricultural:

ag

Alpha
Platform

Nonpoint livestock and fertilizer application emissions from the US
NEI. Livestock includes ammonia and other pollutants (except
PM2.5). County and annual resolution.

U.S. agricultural fires
with point resolution:

ptagfire

Alpha
Platform

Agricultural fire sources that were developed by EPA as point sources
for 2016. Daily resolution.

U.S. area fugitive
dust:

afdust

Alpha
Platform

PMio and PM2 5 fugitive dust sources from the 2014NEIv2 nonpoint
inventory; including building construction, road construction,
agricultural dust, and road dust. The NEI emissions are reduced
during modeling according to a transport fraction and a meteorology-
based (precipitation and snow/ice cover) zero-out. Does not include
onroad emissions in Alaska, Hawaii, Puerto Rico, or Virgin Islands
(see othafdust sector). County and annual resolution.

U.S. biogenic:

beis

Alpha
Platform

Year 2016, hour-specific, grid cell-specific emissions generated from
the BEIS3.61 model within SMOKE. Includes emissions in the
Continental U.S. and portions of southern Canada and northern
Mexico. Uses BELD v4.1 land use data.

Hemispheric
biogenic: g biog

MEGAN
model

Year 2016, hour-specific, grid cell-specific emissions generated from
the MEGAN model for the portion of the hemispheric modeling
domain not covered by the BEIS sector.

U.S. Category 1, 2
CMV:

cmv clc2

Alpha
Platform

Category 1 (CI) and category 2 (C2) commercial marine vessel (cmv)
emissions sources based on the 2014NEIv2 nonpoint inventory, with
SO2 emissions reduced by 90% for 2016. Includes emissions in U.S.
State and Federal Waters. County and annual resolution.

U.S. Category 3
CMV:

cmv c3

Alpha
Platform

Category 3 (C3) cmv emissions based on the 2014NEIv2 nonpoint
inventory, converted to point sources, and with SO2 emissions reduced
by 90% for 2016. Includes emissions in U.S. State and Federal Waters
only; CMV emissions beyond U.S. Federal Waters are in the g_ships
sector. Annual resolution.

U.S. locomotives:
rail

Alpha
Platform

Rail locomotives emissions from the 2014NEIv2. County and annual
resolution.

U.S. remaining
nonpoint:

nonpt

Alpha
Platform

2014NEIv2 nonpoint sources not included in other platform sectors.
County and annual resolution.

U.S. nonpoint source
oil and gas:
np oilgas

Alpha
Platform

2014NEIv2 nonpoint sources from oil and gas-related processes,
projected to 2016. County and annual resolution.

U.S. Residential
Wood Combustion:

rwc

Alpha
Platform

2014NEIv2 nonpoint sources from residential wood combustion
(RWC) processes. County and annual resolution.

U.S. Nonroad:

nonroad

Alpha
Platform

2016 nonroad equipment emissions developed with the MOVES2014a
model. MOVES was used for all states except California, which
submitted their own emissions. County and monthly resolution.

U.S. Onroad:

onroad

Alpha
Platform

2016 onroad mobile source gasoline and diesel vehicles from moving
and non-moving vehicles that drive on roads, along with vehicle
refueling. Includes the following modes: exhaust, extended idle,
auxiliary power units, evaporative, permeation, refueling, and brake
and tire wear. For all states except California, developed using winter
and summer MOVES emissions tables produced by MOVES2014a.
Does not include onroad emissions in Alaska, Hawaii, Puerto Rico, or
Virgin Islands (see onroad can sector).

5

-------
Platform Sector:

abbreviation

Data
Source

Description and resolution of the data input to SMOKE

U.S. Onroad
California:

onroad ca adj

Alpha
Platform

2016 California-provided CAP and metal HAP onroad mobile source
gasoline and diesel vehicles submitted to the NEI, gridded and
temporalized using MOVES2014a. Volatile organic compound
(VOC) HAP emissions derived from California-provided VOC
emissions and MOVES-based speciation.

U.S. point source
fires: ptflre

Alpha
Platform

Point source day-specific wildfires and prescribed fires for 2016
computed using SMARTFIRE2 and Blue Sky Framework for both
flaming and smoldering processes (i.e., SCCs 281XXXX002).
Smoldering is forced into layer 1 (by adjusting heat flux). Daily
resolution.

Other North America
fires:

ptfire othna

Alpha
Platform

Point source day-specific wildfires and prescribed fires for 2016.
Canada fires provided by Environment Canada, with data for missing
months filled in using fires from the Fire INventory (FINN) from
National Center for Atmospheric Research (NCAR) fires (NCAR,
2016 and Wiedinmyer, C., 2011). Fires in Mexico and Central
America are from FINN. Daily resolution.

Canada dust sources:

othafdust

Alpha
Platform

Fugitive dust sources from Canada's 2013 and 2025 inventories
(interpolated to 2016). A transport fraction adjustment is applied
along with a meteorology-based (precipitation and snow/ice cover)
zero-out. Also includes afdust emissions in Alaska, Hawaii, Puerto
Rico, and Virgin Islands from 2014NEIv2. County and annual
resolution.

Canada and Mexico
point sources:
othpt

Alpha
Platform

Point sources from Canada's 2013 and 2025 inventories (interpolated
to 2016) and Mexico's 2014 and 2018 inventories (interpolated to
2016), annual resolution.

Canada and Mexico
nonpoint and
nonroad:

othar

Alpha
Platform

Year 2016 Canada (province or sub-province resolution) emissions,
interpolated from 2013 and 2025: monthly for agricultural ammonia,
and nonroad sources; annual for rail, CMV and other nonpoint Canada
sectors. Year 2016 Mexico (municipio resolution), interpolated from
2014 and 2018: annual nonpoint and nonroad mobile inventories.

Other non-NEI
onroad sources:
onroad can

Alpha
Platform

Monthly year 2016 Canada (province resolution or sub-province
resolution, depending on the province) onroad mobile inventory,
interpolated from 2013 and 2025. Also includes onroad emissions in
Alaska, Hawaii, Puerto Rico, and Virgin Islands from 2014NEIv2.

Other non-NEI
onroad sources:

onroad mex

Alpha
Platform

Monthly year 2016 Mexico (municipio resolution) onroad mobile
inventory, interpolated from 2014 and 2018.

Point source wildland
fires outside North
America: g_ptfire

FINN

Point source day-specific wildfires and prescribed fires for 2016 from
the Fire INventory (FINN) from National Center for Atmospheric
Research (NCAR) fires (NCAR, 2016 and Wiedinmyer, C., 2011).
Includes all fires outside North America. Daily resolution.

Point source
agricultural fires
outside North
America: g ptagfire

FINN

Point source day-specific agricultural fires for 2016 from the Fire
INventory (FINN) from National Center for Atmospheric Research
(NCAR) fires (NCAR, 2016 and Wiedinmyer, C., 2011). Includes all
fires outside North America. Daily resolution.

Agricultural
emissions outside
North America and
China: g ag

HTAP v2
+ CEDS

Agricultural emissions outside North America and China, from the
gridded HTAP version 2 dataset, agriculture sector. Year 2010
projected to 2014 using factors from CEDS dataset. Monthly
resolution.

6

-------
Platform Sector:

abbreviation

Data
Source

Description and resolution of the data input to SMOKE

Aircraft
landing/takeoff
emissions: g air Ito

HTAP v2

Landing and takeoff emissions from aircraft, from the gridded HTAP
version 2 dataset, air sector. Annual resolution.

Aircraft climbing and
descent emissions:

g air cds

HTAP v2

Climbing and descent emissions from aircraft, from the gridded HTAP
version 2 dataset, air sector. Annual resolution.

Aircraft cruising
emissions: g air crs

HTAP v2

Cruising emissions from aircraft, from the gridded HTAP version 2
dataset, air sector. Annual resolution.

Energy sector
emissions outside
North America and
China: g energy

HTAP v2
+ CEDS

Energy sector emissions from the gridded HTAP version 2 dataset,
excluding North America and China. Year 2010 projected to 2014
using factors from CEDS dataset. Monthly resolution.

Industry sector
emissions outside
North America and
China: g industry

HTAP v2
+ CEDS

Industry sector emissions from the gridded HTAP version 2 dataset,
excluding North America and China. Year 2010 projected to 2014
using factors from CEDS dataset. Monthly resolution.

Residential sector
emissions outside
North America and
China: g residential

HTAP v2
+ CEDS

Residential sector emissions from the gridded HTAP version 2 dataset,
excluding North America and China. Year 2010 projected to 2014
using factors from CEDS dataset. Monthly resolution.

Transport sector
emissions outside
North America and
China: g transport

HTAP v2
+ CEDS

Transport sector emissions from the gridded HTAP version 2 dataset,
excluding North America and China. Year 2010 projected to 2014
using factors from CEDS dataset. Monthly resolution.

Shipping emissions
outside North
America: g ships

HTAP v2
+ CEDS

Shipping emissions from the gridded HTAP version 2 dataset, ships
sector. Excludes emissions in U.S. State and Federal Waters. Year
2010 projected to 2014 using factors from CEDS dataset. Annual
resolution.

China fertilizer
application:
china agrf

THU
China

Fertilizer application emissions in China from the THU dataset.
Annual resolution with hourly temporal profiles.

China livestock:
china agrl

THU
China

Livestock emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

China domestic
biofuel: china dobi

THU
China

Domestic biofuel emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

China domestic
combustion:
china docb

THU
China

Domestic combustion emissions in China from the THU dataset.
Annual resolution with hourly temporal profiles.

China domestic fossil
fuel: china dofu

THU
China

Domestic fossil fuel emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

China domestic
solvent use:
china doso

THU
China

Domestic solvent use emissions in China from the THU dataset.
Annual resolution with hourly temporal profiles.

China other domestic:
china doth

THU
China

Other domestic emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

China industry
combustion:
china inch

THU
China

Industrial combustion emissions in China from the THU dataset.
Annual resolution with hourly temporal profiles.

China power plant:
china ppcb

THU
China

Power plant emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

7

-------
Platform Sector:

abbreviation

Data
Source

Description and resolution of the data input to SMOKE

China cement:
china prce

THU
China

Cement process emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

China steel:
china prir

THU
China

Steel process emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

China industrial
solvent use:
china prso

THU
China

Industrial solvent use emissions in China from the THU dataset.
Annual resolution with hourly temporal profiles.

China other
industrial: china prot

THU
China

Other industrial process emissions in China from the THU dataset.
Annual resolution with hourly temporal profiles.

China transport off-
road: china trof

THU
China

Off-road mobile emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

China transport on-
road: china tron

THU
China

On-road mobile emissions in China from the THU dataset. Annual
resolution with hourly temporal profiles.

Lightning NOx
emissions: lightning

GEIA

Global NOx emissions from lightning strikes.

Table 2-2. Summary of emissions differences between 2015 spinup and 2016 base case emissions by

sector

Platform Sector

2016 base case

2015 spinup case

afdustadj

2014NEIv2, adjusted with 2016
meteorology

2014NEIv2, adjusted with 2015
meteorology

onroad

2016 activity / emission factors /
meteorology

2015 activity / emission factors /
meteorology

onroad ca adj

CARB inventory interpolated to 2016

CARB inventory interpolated to 2015

ag

2016 fertilizer + 2014NEIv2 livestock +
2016 met-based temporalization

2014NEIv2 fertilizer and livestock + 2015
met-based temporalization

cmv clc2

2014NEIv2 with SO2 reduced by 90%

Same as 2016

cmv c3

2014NEIv2 with SO2 reduced by 90%

Same as 2016

nonpt

2014NEIv2

Same as 2016

nonroad

2016 inventory from MOVES2014a

Interpolation of 2014 and 2016 inventories
from MOVES2014a

np oilgas

2014NEIv2 projected to 2016

2014NEIv2 projected to 2015

rail

2014NEIv2

Same as 2016

rwc

2014NEIv2 + 2016 met-based
temporalization

2014NEIv2 + 2015 met-based
temporalization

othafdust

2013 and 2025 Canada, interpolated to 2016

2013 and 2025 Canada, interpolated to 2015

onroad can

2013 and 2025 Canada, interpolated to 2016

2013 and 2025 Canada, interpolated to 2015

onroad mex

2014 and 2018 Mexico, interpolated to 2016

2014 and 2018 Mexico, interpolated to 2015

othar

2013	and 2025 Canada, interpolated to 2016

2014	and 2018 Mexico, interpolated to 2016

2013	and 2025 Canada, interpolated to 2015

2014	and 2018 Mexico, interpolated to 2015

othpt

2013	and 2025 Canada, interpolated to 2016

2014	and 2018 Mexico, interpolated to 2016

2013	and 2025 Canada, interpolated to 2015

2014	and 2018 Mexico, interpolated to 2015

ptagfire

2016 point ag fire dataset

2015 point ag fire dataset

ptegu

2016 point inventory

2015 point inventory

ptnonipm

2016 point inventory

2015 point inventory

pt_oilgas

2016 point inventory, with remaining 2014
sources projected to 2016

2015 point inventory, with remaining 2014
sources projected to 2015

ptfire

2016 SMARTFIRE2 dataset

2015 SMARTFIRE2 dataset

g ptfire

2016 FINN dataset

2015 FINN dataset

8

-------
Platform Sector

2016 base case

2015 spinup case

ptfireothna

2016 Environment Canada fires + FINN
fires

2015 STI fires

beis

2016 biogenics

2015 biogenics

g biog

2016 biogenics

2015 biogenics

B ag

2010 HTAPv2 + 2014 CEDS projection

Same as 2016

g ptagfire

2016 FINN dataset

2015 FINN dataset

g air cds

2010 HTAPv2

Same as 2016

g air crs

2010 HTAPv2

Same as 2016

g air lto

2010 HTAPv2

Same as 2016

g energy

2010 HTAPv2 + 2014 CEDS projection

Same as 2016

g industry

2010 HTAPv2 + 2014 CEDS projection

Same as 2016

g residential

2010 HTAPv2 + 2014 CEDS projection

Same as 2016

g ships

2010 HTAPv2 + 2014 CEDS projection

Same as 2016

g transport

2010 HTAPv2 + 2014 CEDS projection

Same as 2016

lightning

Gridded lightning NOx dataset

Same as 2016

All China sectors
(china *)

2015 China THU

Same as 2016

Documentation for the North America emissions sectors is provided in the 2016 alpha platform for
regional modeling TSD. The following sectors concern the emissions inventories for the portions of the
hemispheric platform that are not part of the regional modeling platform.

9

-------
2.1 Global HTAP version 2

The Hemispheric Transport of Air Pollution Version 2 inventory, or HTAPv2 inventory, is the basis for
most emissions in the platform outside of North America and China. HTAP inventories are available here:
http: //edgar. i rc. ec. europa. eu/htap v2/

Each HTAP sector is mapped to one or more platform sectors as outlined in Table 2-3. The representative
dates column indicates whether emissions are processed for every day of the year ("All"); seven days per
month, one for each day of the week ("Week"); four days per month, corresponding to Mondays, other
WeekDays, Saturdays, and Sundays ("MWDSS"); or for a single average day per month ("Aveday").

Table 2-3. HTAP sector mappings to platform sectors

HTAP sector

Platform sector(s)

Inventory
resolution

Pollutants

Representative
dates

Air

g_air_cds
g_air_crs
g air lto

Annual

CO, NOx, PMio, PM2.5,
BC (PEC), OC (POC),
S02, NMVOC

All

Ships

g_ships

Annual

CO, NOx, PM10, PM2.5,
BC (PEC), OC (POC),
S02, NMVOC

All

Energy

g_energy

Monthly

CO, NH3, NOx, PMio,
PM2.5, BC (PEC), OC
(POC), S02, NMVOC

All

Industry

g_industry

Monthly

CO, NH3, NOx, PMio,
PM2.5, BC (PEC), OC
(POC), S02, NMVOC

All

Transport

gtransport

Monthly

CO, NH3, NOx, PMio,
PM2.5, BC (PEC), OC
(POC), S02, NMVOC

Week

Residential

gresidential

Monthly

CO, NH3, NOx, PMio,
PM2.5, BC (PEC), OC
(POC), S02, NMVOC

MWDSS

Ag

g ag

Monthly

nh3

Aveday

Hemispheric emissions totals from each HTAP sector are provided in Table 2-4. These emissions totals
include CEDS projections (see section 2.1.5), do not include emissions in North America or China that
were zeroed out (see section Error! Reference source not found.), and are domain-wide for the 108km
hemispheric modeling domain.

10

-------
Table 2-4. Emissions totals by HTAP sector, 108km hemispheric domain, tons/year for 2016

sector

NH3

NOX

PM10

PM2.5

S02

voc

g_ag

27,053,284

g_air_c;ds

134,140

1,089,169

18,439

92,194

31,552

g_air_c;rs

179,798

1,457,442

24,682

123,425

42,107

g_air_lt°

116,928

282,723

2,350

26,785

4,415

g_energy

8,597,034

71,485

14,168,795

4,951,791

2,077,552

26,978,678

884,267

g_industry

132,095,290

992,589

6,997,163

6,873,446

3,563,763

16,333,374

38,620,852

g_residential

232,141,408

4,546,822

4,344,817

15,661,354

11,074,589

3,478,108

33,272,813

g_ships

1,175,509

12,078,998

1,207,008

6,633,296

628,816

g_transport

99,498,142

189,347

24,530,368

1,827,155

1,678,990

1,621,702

23,083,023

2.1.1 Spatial Allocation

The HTAP dataset includes gridded annual and monthly datasets on a global 0.1° latitude by 0.1°
longitude grid. Spatial allocation of these inventories in SMOKE consists of two steps. First, in the
SMOKE program Smkinven, each 0.1° by 0.1° point on the global grid is mapped to a country code
(GEOCODE) and a time zone using a file called the GRIDMASK. Then, the SMOKE program Grdmat
spatially reallocates emissions from the global input grid to the 108km polar stereographic grid used for
air quality modeling.

The GRIDMASK file is a gridded file on the same 0.1° by 0.1° global grid and maps each point on the
grid to a GEOCODE and a time zone. Smkinven maps each point in the gridded inventory data to a
GEOCODE, enabling calculation of national emissions totals, and application of country-specific
projections or controls. GEOCODEs are defined not only for each country, but also for distinct bodies of
water, e.g. the North Atlantic Ocean, Gulf of Mexico, Caribbean Sea, and so on. The time zone
information is used by the SMOKE program Temporal to apply diurnal profiles with the correct time zone
offsets. Time zones with a 30-minute offset (e.g. Newfoundland, India) are rounded to the nearest hour in
SMOKE modeling.

A plot of the GEOCODEs in the GRIDMASK file is shown in Figure 2-1, in which distinct GEOCODEs
appear as different colors.

-------
Figure 2-1. Map of GEOCODE Country Codes

2.1.2 Chemical Speciation

HTAP inventories include the following pollutants: "no methane VOC" (NMVOC); black carbon (BC,
mapped to model species PEC); organic carbon (OC, mapped to model species POC); total PM2.5; and
other standard CAPs (CO, NH3, NOx, PM10, S02). HTAP inventories do not include any HAP
emissions.

Unlike regional modeling which uses emissions inventories with SCCs, HTAP inventories do not include
more detailed information beyond the sector total for each point on the grid by month and pollutant.
Therefore, we must apply sector-average speciation profiles to each HTAP sector. For VOC, average
speciation profiles for HTAP sectors for the CB6 chemical mechanism were developed from the 2011
emissions modeling platform. In that platform, U.S. SCCs were each mapped to one of the HTAP
sectors, and then speciated emissions were summed by HTAP sector group to develop an average VOC
profile for each HTAP sector. Since the HTAP inventories specify "NMVOC" instead of "VOC", these
average profiles do not include methane (CH4), but do include other species which may be considered
part of TOG but not VOC, such as ethane (ETHA). There is no VOC-to-TOG conversion prior to
speciation like there is in traditional emissions modeling, since these average profiles are computed on the
basis of (NM)VOC.

For PM2.5 speciation, the HTAP dataset includes emissions for black carbon, organic carbon, and total
PM2.5. In SMOKE modeling, we map black carbon to the model species PEC, and organic carbon to the
model species POC. Since PEC and POC are also part of total PM2.5, we must subtract PEC and POC
from total PM2.5 in order to prevent a double count. Prior to SMOKE modeling, an additional set of
gridded inventory files is generated for PM2 5 0TH (other PM2.5), which is equal to total PM2.5 minus
PEC minus POC. Then, PM2 5 0TH is speciated using speciation profiles which map to the remaining

-------
PM species needed for CMAQ modeling. It is still necessary for SMOKE to use total PM2.5 for the sole
purpose of calculating PMC (PM10 minus total PM2.5), but only PM2 5 0TH is mapped to PM model
species other than PEC, POC, and PMC. Similar to the VOC speciation profiles, speciation of
PM2 5 0TH uses sector-specific average profiles.

For all HTAP sectors, NOx is speciated to NO and N02 using a 90/10 split. Emissions for the HONO
species are not created outside of North America. The SULF species is calculated as a percentage of S02
in the HTAP energy, industry, and residential sectors.

2.1.3 Temporal Allocation

Similar to with chemical speciation, each HTAP sector has a single set of temporal profiles (weekly and
diurnal, and also monthly for sectors without monthly inventories) that is applied to the entire sector.
These temporal profiles were estimated using North American source-specific examples and best
judgement. Because different countries celebrate different holidays, there are no considerations for
holiday temporalization in the HTAP sectors.

Similar to some North America sectors, some HTAP sectors are processed with representative dates, as
specified in Table 2-3 (above). For example, the HTAP ag sector uses flat day-of-week temporalization,
and so for the g ag sector, we process emissions for one representative day per month. The aircraft,
energy, industry, and ships sectors are processed for every day to support plume rise, as discussed in the
next section.

The weekly and diurnal temporal profiles for HTAP sectors are shown in Figures 2-2 and 2-3. The HTAP
air and ships sectors each use a flat monthly temporal profile, while the other HTAP sectors have monthly
inventories and do not use monthly temporal profiles.

Figure 2-2. Day-of-week temporal profiles for HTAP sectors

M Tu W Th F

— ii3/ AiR/SHIF;- EI.ERGV —INDUSTRY/ RESIDENT

-------
Figure 2-3. Hour-of-day temporal profiles for HTAP sectors

o.i
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0

1 2 3 4 5 6 7 B 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

AG AIR CDS/LTO AIRCRS ^—ENERGY

^—INDUSTRY ^—RESIDENTIAL-^—SHIPS TRANSPORT

2.1.4 Vertical Allocation

Unlike in North America sectors, HTAP inventories do not include stack parameters or other information
which can be used to estimate plume rise for point sources. This means we cannot run CMAQ with inline
plume rise, and must apply vertical allocation in SMOKE to all sectors. In North America, we can use the
Laypoint program to calculate plume rise using stack parameters and meteorology. Without stack
parameters, we cannot do that for HTAP sectors; instead, we apply vertical profiles on a sector-wide
basis.

Figure 2-4 shows the vertical profiles applied to HTAP sectors. Among HTAP sectors, only the aircraft,
energy, industry, and ships sectors have a vertical profile applied. For ag, residential, and transport, all
emissions remain in Layer 1.

Figure 2-4. Layer fractions for HTAP sectors

0.35

layer

ships industry energy air LTO air CDS airCRS

Vertical profiles are applied using the SMOKE program Layalloc. First, the SMOKE program Smkmerge
generates daily emissions with everything in a single layer, and then the Layalloc program splits the
emissions from Smkmerge into vertical layers. Layer fractions are input to Layalloc in the following
format, where "bottom" and "top" are layer heights in meters:

1 ay er,b ottom,top, fracti on

-------
1,0,20,0.06667
2,20,44,0.06667
3,44,70,0.06667
4,70,101,0.06667

41,14388,15345,0
42,15345,16394,0
43,16394,17565,0
44,17565,18874,0

Layalloc requires that the bottom and top of each layer be specified in the layer fraction file, and applies
layer fractions according to those layer heights, not according to the layer number in the first column.
Layer heights are not constant, but vary according to meteorology. Our layer fraction files use layer
heights averaged across the year. Because layer heights vary temporally and spatially, the actual layer
fractions as applied by Layalloc are not constant. This means that occasionally, a small amount of
emissions may appear in a layer with a 0% layer fraction. For example, the layer fractions for aircraft
cruising emissions allocate 20% of the emissions to each of five layers from 34 to 38; but because of
varying layer heights, Layalloc does allocate a small amount of emissions to layers 33 and 39.

2.1.5 Projection to 2014 via CEDS

The HTAP version 2 inventory represents emissions for the year 2010, which is six years prior to the
model year for this project. A projection of the HTAP emissions to a year as close to 2016 as possible was
desired.

Data from the Community Emissions Data System, or CEDS, includes emissions totals for the years 2010
and 2014, among other years, by country and emissions sector. Emissions totals for years after 2014 were
not available at the time of this study, so 2014 was used as a reasonable approximation for 2016.

Table 2-5 shows how CEDS sectors were mapped to HTAP sectors. CEDS data is available at a much
finer resolution than the aggregate sectors listed in this table, but because adjustment factors can only be
applied on a sector-wide basis in SMOKE, we could not apply projection factors at a finer resolution than
as shown in this table.

Table 2-5. Mapping of CEDS sectors to HTAP sectors

CEDS aggregate sector(s)

HTAP sector (SMOKE sector)

Agriculture

Agriculture (g ag)

Power

Energy (g energy)

Industry, Solvents

Industry (g industry)

Residential, Waste

Residential (g residential)

Ship

Ships (g ships)

Transportation

Transport (g transport)

HTAP projection factors for each country, HTAP/SMOKE sector, and pollutant were calculated using
CEDS emissions totals for 2010 and 2014 (examples in Figure 2-5), and then converted into GCNTL
format for application in SMOKE. Separate projection factors were calculated for PEC (BC), POC (OC),
and total PM2.5, as well as for CO, NH3, NOX, S02, and NMVOC. In China, we applied CEDS
projection factors to HTAP emissions for carbon monoxide (CO) only, because the THU dataset does not
include CO emissions. In the EdgarCHN sensitivity (see Section 23.1.9) CEDS factors were applied to

-------
HTAP emissions for all pollutants in China. For the ships sector, because most shipping emissions take
place over open water and not in a specific country, and because S02 is the primary pollutant of interest
with the ships sector, a global projection factor was applied to the entire domain for S02 only. The CEDS
dataset does not include emissions for aircraft, and so HTAP aircraft emissions were not projected.

Country-specific projection factors can be applied in SMOKE according to the GEOCODE for each
country, with GEOCODEs mapped to a set of points in the gridded HTAP inventories using the
GRIDMASK file. To limit the influence of unrealistically high or low projection factors from the CEDS
dataset, projection factors were limited to a range of 0.5 (50% decrease) to 1.5 (50% increase).

Figure 2-5. Projection factors from year 2010 to 2014 for POWER(left) and TRANSPORT(right)

sectors for NOx emissions derived from CEDS.

To prevent double counting of emissions with other datasets, a portion of the emissions from the HTAP
datasets must be zeroed out. In North America, all HTAP emissions except for in-flight aircraft are zeroed
out, to prevent a double count with the U.S., Canada, and Mexico inventories. In China, all HTAP
emissions except for CO (all sectors), and except for aircraft and shipping (all pollutants), are zeroed out,
to prevent a double count with the THU datasets. Zero-outs are implemented with a GCNTL file, which is
read by the SMOKE program Cntlmat and applied to model-ready emissions by pollutant and country
GEOCODE. The procedure for applying zero-outs is the same as the procedure for applying CEDS-based
projection factors, except in the case of zero-outs, the "projection factor" is zero.

In most HTAP sectors, including ag, residential, transport, energy, and industry, we zeroed out emissions
from the following GEOCODEs: United States, Canada, Mexico, and China (except CO). Puerto Rico,
Virgin Islands, Hong Kong, and Macau have separate GEOCODEs from the United States and China, and
were also zeroed out, as these areas were determined to overlap the U.S. and China datasets, respectively.
Taiwan is not covered by the THU dataset and is not zeroed out.

Bodies of water which border the United States, Canada, and Mexico were also zeroed out in most HTAP
sectors. This is because emissions from the HTAP dataset extend slightly into points that are assigned
water GEOCODEs instead of land GEOCODEs in the GRIDMASK file, and zeroing out bodies of water
adjacent to North America as well as the land areas prevents a "ring" of emissions from appearing along
the North America coast. In China, to prevent THU China and HTAP emissions from occuring in the

-------
same grid cells, the GRIDMASK file was modified to ensure that any point covered by the THU China
datasets is mapped to either the original China GEOCODE, or to a new GEOCODE called "China
Coastline". In SMOKE, we then zeroed out both China and China Coastline emissions (except CO).

The HTAP ships and air sectors are handled differently with respect to zero-outs. For shipping, the U.S.
CMV sectors (cmv_clc2, cmv_c3) include emissions in U.S. state and federal waters, extending 200
nautical miles off the U.S. coastline. The original GRIDMASK file did not distinguish U.S. federal waters
from non-federal waters, and so it was originally not possible to zero out HTAP ships emissions in the
exact area covered by the cmv_clc2 and cmv_c3 sectors. Previous hemispheric emissions modeling
applications resolved this by dropping U.S. federal water emissions from the cmv_clc2 and cmv_c3
sectors, but for this platform, use of U.S.-based emissions in federal waters was preferred over use of
HTAP ships emissions. To facilitate this, additional GEOCODEs were added to the GRIDMASK file in
U.S. state and federal waters. First, points in U.S. state waters that were originally assigned to a water
GEOCODE (e.g. 034000 for Gulf of Mexico) were reassigned to the code 3100XX, where XX = the U.S.
state FIPS code (e.g. 310012 for Florida). Second, points in U.S. federal waters that were originally
assigned to a water GEOCODE were reassigned to the code 85X000, where X = a number from 1 through
7 corresponding to the FIPS code associated with each offshore zone (1 = North Pacific, 2 = South
Pacific, 3 = Gulf Coast, 4 = East Coast, 5 = Alaska, 6 = Hawaii, 7 = Puerto Rico). When running the
HTAP ships sector, we zero out all emissions from the U.S., Canada, and Mexico GEOCODEs, and also
from all U.S. state and federal water GEOCODEs, but not from any bodies of water outside of U.S.
federal waters. We do not zero out shipping emissions in or around China because the THU dataset does
not include shipping.

The U.S. and Canada inventories include emissions from aircraft that are landing or taking off, which
overlaps the HTAP air LTO sector. Therefore, in the g air lto sector, we zero out emissions over land in
the United States and Canada. We do not zero out LTO emissions in Mexico, China, or over any bodies of
water including U.S. federal waters. We do not zero out any emissions in the HTAP g air cds
(climbing/descent) or g air crs (cruising) sectors, as the U.S. and Canada aircraft inventories do not
include those categories of aircraft emissions.

2.2 THU China

Gridded annual emissions for China for the year 2015 were provided by Tsinghua University (THU)

(Zhao et al., 2018,). The THU dataset includes emissions for 16 sectors, 15 of which are included in the
platform. The THU dataset includes annual gridded emissions on a Lambert projection grid with 27km
resolution; emissions forNH3, NOX, PM10, PM2.5, S02, and VOC, including partial speciation; and
temporal profiles which are used to generate hourly emissions.

The THU dataset emissions sectors, along with a mapping to the equivalent HTAP sectors, are listed in
Table 2-6. We are not using the open burning sector emissions from THU. Fire emissions from the FINN
dataset are preferred over the THU open burning emissions, primarily because the FINN data has better
daily resolution. Note that the THU dataset does not overlap the HTAP aircraft or ships sectors. Annual
emissions totals by sector for China are in Table 2-7.

The emissions files provided by Tsinghua University were annual files on a Lambert grid with 27km
resolution with partial speciation. Conversion of these emissions to the format needed for air quality
modeling was performed using standalone tools and not with the SMOKE model. First, the emissions
were temporalized from annual to hourly; then, additional speciation and unit conversion was performed

-------
as needed; then, emissions were reallocated to the 108km hemispheric grid; and finally, the emissions
were allocated vertically, depending on the sector.

Table 2-6. List of THU China sectors and equivalent HTAP sectors

THU China sector

Equivalent HTAP sector

AGRF (fertilizer application)

Agriculture

AGRL (livestock)

Agriculture

DOBI (domestic bio-fuel)

Residential

DOCB (domestic combustion)

Residential

DOFU (domestic fossil fuel)

Residential

DOSO (domestic solvent use)

Residential

DOTH (other domestic)

Residential

INCB (industry combustion)

Industry

PPCB (power plant)

Energy

PRCE (cement)

Industry

PRIR (steel)

Industry

PRSO (industry solvent use)

Industry

PROT (other industry process)

Industry

TROF (off-road transport)

Transport

TRON (on-road transport)

Transport

OPEN (open burning)

None; using FINN fires instead of THU fires

Table 2-7. Annual emissions totals by THU China sector, tons/year

China sector

NH3

NOX

PM10

PM2.5

S02

VOC1

AGRF (fertilizer application)

4,105,865











AGRL (livestock)

6,164,346











DOBI (domestic bio-fuel)



247,159

1,535,996

1,487,994

37,311

247,159

DOCB (domestic combustion)











2,254,033

DOFU (domestic fossil fuel)



684,444

2,030,505

977,311

3,894,725

684,444

DOSO (domestic solvent use)











1,863,873

DOTH (other domestic)

741,314









213,655

INCB (industry combustion)



2,520,744

1,453,406

965,475

5,920,805

2,653,154

PPCB (power plant)



4,390,020

1,175,701

702,486

4,992,558

4,390,020

PRCE (cement)



2,100,046

1,370,862

757,175

1,005,509

2,100,046

PRIR (steel)



603,922

1,300,269

939,834

836,222

603,922

PRSO (industry solvent use)











5,792,690

PROT (other industry process)

353,494

2,389,083

1,591,475

1,014,461

1,881,593

7,557,787

TROF (off-road transport)



2,684,613

219,988

208,409

159,214

3,354,114

TRON (on-road transport)



8,385,993

405,461

384,119

22,164

15,460,059

'VOC totals in this table are estimated from a sum of individual VOC species.

18

-------
2.2.1 Temporal Allocation

Included in the THU data release were annual emissions files, and temporal profiles on a month-of-year,
day-of-week, and hour-of-day basis. Temporal profiles vary by broad sector (AG/agriculture,
DO/domestic, IN/industry, PP/power plant, PR/industry process, and TR/transport) and by Chinese
province. Because daily temporalization depends only on the day of the week, emissions for seven
representative days per month, one for each day of the week, are sufficient.

A province-to-grid cell cross-reference for the China 27km grid was used in combination with the
temporal factors. Figure 2-6. shows the averages of province-level monthly, daily, and hourly temporal
profiles. These temporal profiles were gridded for each sector and combined to create an annual hour-of-
year basis file. Then, the gridded hourly profiles were applied to the annual gridded emissions for each
sector to create emissions files with 25 hourly timesteps for each representative day of the year (one file
for each day of the week for each month). Application of gridded hourly profiles was applied using a
Python utility called apply fracs.py which multiplies the annual China emissions files by the hourly
fraction matrix to generate the representative day, hourly emissions files.

For both 2016 modeling (base year) and 2015 modeling (spinup year), the same set of representative day
emissions are used in China.

19

-------
Figure 2-6. Time allocation factors for the China Inventory. Averages of province-specific

allocation factors are shown.

Agricultural Industry Power Plant Processing Transport

Domestic Open Burning

Agricultural Industry Power Plant Processing Transport

Domestic Open Burning

0.20 -

0.15

0.10 -

tu
>
<

'5i

£ 0.05

0.00 -

1 1 1 1 1 r~

1 3 5 7 9 11

month

0.16 -

u 0.14-

QJ
CD

0.12

a o.io -

-------
2.2.2 Chemical Speciation

The THU dataset included pre-speciated emissions for VOC and PM2.5 species, but additional processing
related to speciation was necessary to prepare these emissions for air quality modeling. The following
steps were all performed using the same Python utility (apply fracs.py) as the temporalization. For VOC
species, only a time conversion was necessary; emissions units from the previous step were in moles per
hour, which we converted to moles per second for CMAQ modeling. For PM2.5 species, emissions units
were in metric tons per hour (not short tons as in most North America emissions inventories), and were
converted to grams per second. The THU dataset included emissions for NOX, but not individual species
NO or N02; we split NOX into NO (90%) and N02 (10%), and also converted from metric tons per hour
to moles per second. NH3 and S02 only required a unit conversion from metric tons per hour to moles
per second.

Finally, the THU dataset included PM10 and individual PM2.5 species, but not the PMC species. We
added the PMC species to the emissions files by subtracting the sum of PM2.5 from PM10. Occasionally,
the sum of PM2.5 emissions slightly exceeded the PM10 value, resulting in a small negative value for
PMC. This was most common in the transport sector. In order to prevent errors in emissions processing or
air quality modeling, any instance of negative PMC was reset to a value of zero.

2.2.3 Spatial Allocation

The next step to prepate the THU China emissions for air quality modeling is to regrid the emissions from
the Lambert 27km grid to the 108km hemispheric grid. For this, the I/O API utility mtxcalc was run to
create a conversion matrix from the 27km China grid to the 108km hemispheric grid. Then, a Python
script was used to re-project the hourly speciated emissions from the 27km China grid to the 108km
hemispheric grid using the conversion matrix from mtxcalc. Early versions of I/O API v3.2 are known to
produce incorrect conversion matrices from mtxcalc, but this has been fixed in the latest build of v3.2; I/O
API v3.1 is also valid.

2.2.4 Vertical Allocation

Vertical allocation was applied to the hourly, speciated, and gridded emissions, consistent with how the
HTAP sectors are processed. China sectors that are mapped to the HTAP industry sector (see Table 2-6)
were vertically allocated using the same layer fractions that are used to allocate the g industry sector.
Similarly, the China power plant sector was vertically allocated using the layer fractions from the
g energy sector. All Chinese domestic, transport, and agriculture emissions were retained as single layer,
low-level emissions.

Layer fractions were applied using the SMOKE program Layalloc using a similar procedure as for the
HTAP sectors g industry and g energy. As with the HTAP sectors, because layer heights vary
temporally, layered emissions for the China industry and energy sectors are processed for each day of the
year, separately for 2016 (base year) and 2015 (spinup year).

The resulting China emissions files, which include hourly temporalization, full speciation, vertical
allocation, and spatial allocation on the hemispheric 108km grid, are read directly by the SMOKE
program Mrggrid and merged with the other sectors.

2.3 Global FINN fires (g_ptfire, g_ptagfire)

Annual 2015 and 2016 fire emissions outside of North America were developed from FINN (Fire
Inventory from NCAR) vl.5 daily fire emissions. For FINN fires, listed vegetation type codes of 1 and 9

-------
are defined as agricultural burning, all other fire detections and assumed to be wildfires. Wildfires are
processed in the g_ptfire sector separately from ag burning in the g_ptagfire sector. All wildland fires that
are not defined as agricultural are assumed to be wild fires rather than prescribed. FINN fire detects less
than 50 square meters (0.012 acres) are removed from the inventory because they are smaller than the
pixel size of the remote sensing data and are therefore assumed to be false detects. Fires in North
America, Central America, and the Caribbean were excluded from the g_ptfire and g_ptagfire sectors
because fires in these areas are included in North America portion of the hemispheric platform (ptfire,
ptagfire, and ptfire othna sectors). In China, we include fire emissions from the g_ptfire and g_ptagfire
sectors and do not include open burning emissions from the THU inventory.

Global fire processing uses a unique set of country codes that are similar to, but not the same as, standard
FIPS codes used in North America. The locations of FINN fires are geocoded from latitude and longitude
to a country code and a time zone code. The 6-digit FlPS-like codes used for global fires follow the
format 9TTCCC, where TT indicates the time zone (01 through 24), and CCC is the country-level
GEOCODE that is used for HTAP sector processing. This allows for the generation of emissions
summaries by country, and supports diurnal temporalization of fires in local time. Diurnal profiles for
global FINN wild fires and ag fires are shown in Figure 2-7.

Figure 2-7. Diurnal temporal profiles for global wild fires and ag fires

Global fire diurnal profiles

18.0056
16.00%

14.00%

12.00%

10.00%

8.00%

6.00%

4.00%

2.00%

0.00%

°_ptf r e g_ptagfire

The FINN fire inventory includes heat content and location information specific to individual fires, which
can be used to calculate fire-specific plume rise as is done for North America fires. However, given the
large number of fires that may exist around the globe at any given time, processing all global fires
individually through the SMOKE model is more computationally intensive than can be accomplished on
EPA computing systems in a reasonable amount of time. To reduce the computational complexity, we
reduce the number of unique fires in the global wildfire and ag fire datasets. After fires are geocoded to
time zone and country, latitude and longitude coordinates for each fire are rounded to the nearest tenth of
a degree (0.1°). Then, all fires that share the same time zone, country code, and rounded coordinates, are
summed to a single inventory source. This reduces the size of the annual fire inventory by a factor of
approximately 15, and the size of the daily fire inventory by a factor of approximately 10. While this does
reduce the spatial accuracy of the inventory, this new resolution is equivalent to the resolution of gridded
HTAP emissions inventories and is still more than 10 times more precise than the resolution of the 108km
hemispheric modeling domain, and so the impact of coordinate rounding on 108km air quality modeling
is minimal.

hour of day (local time)

-------
When summing fire sources that share a common time zone, country code, and rounded coordinates, we
lose the fire-specific heat content information that is needed to calculate plume rise. So for the g_ptfire
and g_ptagfire sectors, we apply sector-wide layer fractions with the SMOKE program Layalloc, similar
to how HTAP and THU China sectors are layered. Different layer fractions are applied to g_ptfire and
g_ptagfire; this is the main reason wild fires and ag fires are processed separately. The layer fractions for
the two global fire sectors are shown in Figure 2-8.

Figure 2-8. Layer fractions for global wild fires (g_ptfire) and ag fires (g_ptagfire)

Global fire layerfractions

g_ptagfire g_ptf"re

Annual emissions totals for the g_ptfire and g_ptagfire sectors are in Table 2-8.

Table 2-8. Emissions totals for global fire sectors, 108km hemispheric domain, year 2016, tons/year

Sector

NH3

NOX

PM10

PM2.5

S02

voc

g ptfire

121,285,767

2,005,131

6,058,488

21,767,734

13,499,258

780,250

34,654,568

g ptagfire

25,959,983

382,301

1,576,886

2,362,013

1,847,549

150,067

9,932,375

2.4 Other global emissions

In addition to emissions from the North America regional modeling platform, HTAP, THU (China), and
global fires from FINN, the hemispheric modeling platform also includes biogenic emissions, lightning
NOx emissions, and ocean chlorine emissions.

2.4.1 Biogenic emissions

Biogenic emissions for the 108km hemispheric modeling domain consist of two components: emissions
from the Biogenic Emission Inventory System (BEIS) version 3.61 in and near the United States, and
emissions from the Model of Emissions of Gases and Aerosols from Nature (MEGAN) elsewhere in the
domain. BEIS emissions from 12km regional modeling were elected for the region surrounding the
United States to be as consistent with the regional modeling platform as possible. However, the BEIS
model cannot be used globally because the land use data upon which BEIS relies, specifically the BELD
dataset, does not extend beyond North America. Therefore, we use biogenics from the MEGAN model to
fill in the rest of the hemispheric domain.

BEIS 3.61 was run using SMOKE for the 12US1 regional modeling domain. The gridded emissions from
12US1 were then spatially allocated to the 108km hemispheric domain using a Python re-gridding utility.

-------
Figure 2-9 shows the spatial extent of 12US1 BEIS emissions on the 108km hemispheric domain. This is
a plot of emissions on the hemispheric domain, zoomed in on North America; the red region represents
nonzero emissions from 12US1 BEIS. To simplify the process of preventing double-counted emissions
along the edges of the red area in Figure 9, rather than identify each individual cell as either a "BEIS" cell
or a "MEGAN" cell, we identified a rectangular region which has 100% 12US1 BEIS coverage -
specifically, (73,31) to (124,64), as shown in blue on Figure 2-9. For BEIS, we keep all emissions inside
the blue rectangle and zero out all emissions outside the blue rectangle. Conversely, for MEGAN, we
keep all emissions outside the blue rectangle and zero out all emissions inside the blue rectangle. All zero-
outs were performed using a Python-based netCDF utility. Sector names for these two sectors are "beis"
for the BEIS component and "g biog" (global biogenics) for the MEGAN component.

Figure 2-9. Spatial extent of 12US1 BEIS emissions on the 108km hemispheric domain

Biogenic emissions from the MEGAN model were processed for a 2° x 2.5° global GEOS-Chem domain.
The MEGAN emissions were reallocated to the 108km hemispheric domain. Model output mass flux
emissions values were converted to emissions rates using the area of each grid cell. Gas phase species
values were then converted from mass rates to mole emissions rates. Volatile organic species were
combined to match the CB6 species needed for CMAQ. To avoid the double counting of biogenic
emissions, the region in the blue rectangle zeroed out as described above. Biogenic emissions from
MEGAN are processed for a single representative day per month for both 2015 and 2016. Formaldehyde
emissions from MEGAN were lower than expected compared to alternative datasets, and so FORM
emissions were increased by 70% compared to MEGAN outputs.

Both BEIS and MEGAN include NOx emissions from soil. The total Co, NO and VOC emissions are
shown in Table 2-9.

-------
Table 2-9. Annual biogenic emissions totals, 108km hemispheric domain, year 2016, tons/year

Sector

VOC1

beis (US + vicinity)

9,956,463

1,961,185

58,179,072

megan (rest of domain)

34,386,775

16,252,008

260,825,447

1 VOC emissions are approximated using a sum of individual VOC species.

2.4.2 Lightning NOx emissions

NOx emissions from lightning strikes, known as Lightning NOx, were developed. Monthly lightning NOx
emissions including vertical allocation for all 44 layers were created from a global inventory (Price,
1997). Lightning NOx was speciated to model species NO and N02 using a standard 90/10 split.

Diurnal temporalization for lightning NOx was applied using lightning strike flash rate data from
Blakeslee, et al., 2014 (http://www.sciencedirect.eom/science/article/pii/S0169809512003250#t0010).
This lightning data includes diurnal variation in lightning strikes in UTC by season (summer, fall, winter,
spring) and by continent. Diurnal temporal profiles were calculated from this data for each season and
continent and then applied to the monthly total lightning NOx emissions, using the GRIDMASK file to
map cells to continents. Lightning strikes over open water use a flat diurnal profile.

There is no day-of-month temporalization, and so there is a single hourly lightning NOx emissions file for
each month. The same lightning NOx emissions are used for both 2015 (spinup year) and 2016 (base
year).

Figure 2-10 shows a plot of lightning NOx annual emissions for the 108km hemispheric domain. There
are approximately 30 million tons per year of lightning NOx emissions.

-------
Figure 2-10. Lightning NOx Annual emissions
Annual total lightning NOx

2.4.3 Ocean chlorine emissions

The ocean chlorine gas emission estimates are based on the build-up of molecular chlorine (C12)
concentrations in oceanic air masses (Bullock and Brehme, 2002). Ocean chlorine emissions for the
108km hemispheric domain were processed using an ocean mask file for the hemispheric domain,
representing the amount of salt water coverage in each grid cell in the domain.

2.5 Changes to North America sectors to support hemispheric modeling

Documentation regarding emissions from the United States, Canada, and Mexico, is available in the 2016
alpha regional platform TSD. This section covers differences between the regional platform and the
hemispheric platform concerning emissions in North America.

2.5.1 Emissions inventory differences

As discussed in the next section, hemispheric emissions in the onroad and afdust_adj sectors were
reallocated from 12US1 to the 108km hemispheric domain. Therefore, those two sectors exclude
emissions in Alaska, Hawaii, Puerto Rico, and Virgin Islands, collectively referred to as AK/HI/PR/VI. In
order to include AK/HI/PR/VI onroad and afdust emissions in the hemispheric modeling platform without
rerunning the onroad or afdust emissions for the entire U.S., the AK/HI/PR/VI onroad inventory from

-------
2014NEIv2 was included in the onroad can (Canada onroad) sector, and the AK/HI/PR/VI afdust
inventory from 2014NEIv2 was included in the othafdust (Canada afdust) sector.

In the regional modeling platform, the cmv_c3 sector includes emissions in U.S. state waters, U.S. federal
waters, and "non-US" emissions beyond federal waters. In the hemispheric modeling platform, the
cmv_c3 sector includes emissions in U.S. state and federal waters, but does not include non-US emissions
beyond federal waters. As discussed in the HTAP section of this document, non-US C3 emissions are
excluded from the hemispheric cmv_c3 sector to prevent a double count with emissions from the HTAP
ships sector.

Biogenics, as discussed in Section 2.4.1, are based on a combination of the BEIS and MEGAN models in
the hemispheric platform, compared to the regional platform which only uses BEIS.

Otherwise, the 2016 hemispheric modeling platform uses the same North America emissions inventories
as the 2016 alpha regional modeling platform.

2.5.2 Spatial Allocation

For most North America sectors, emissions were processed for the 108km hemispheric domain in
SMOKE using spatial surrogates for the hemispheric domain. For some North America sectors,
emissions were processed for the 12US1 regional modeling domain and then reallocated to the 108km
hemispheric domain, in order to ensure consistency between regional modeling and hemispheric modeling
for sectors that are sensitive to meteorology or other grid-specific inputs. Table 2-10 specifies the
approach for all North America emissions sectors.

Table 2-10. Hemispheric spatial allocation approach for all North America sectors

Platform Sector

Spatial allocation approach

afdust adj

Regridded from 12US1; excludes AK/HI/PR/VI

onroad

Regridded from 12US1; excludes AK/HI/PR/VI

onroad ca adj

Regridded from 12US1

Processed with 108km surrogates; temporalization based on 12US1 meteorology

cmv clc2

Processed with 108km surrogates

cmv c3

Point sources allocated to 108km grid; excludes non-US C3

nonpt

Processed with 108km surrogates

nonroad

Processed with 108km surrogates

np oilgas

Processed with 108km surrogates

rail

Processed with 108km surrogates

rwc

Processed with 108km surrogates; temporalization based on 12US1 meteorology

othafdust

Processed with 108km surrogates and adjusted with 108km meteorology; includes AK/HI/PR/VI

onroad can

Processed with 108km surrogates; includes AK/HI/PR/VI

onroad mex

Processed with 108km surrogates

othar

Processed with 108km surrogates

othpt