# \
\ d?
PRO*^
Guidance on the Preparation of Exceptional
Events Demonstrations for Wildfire Events that
May Influence Ozone Concentrations
-------
-------
EPA-457/B-16-001
September 2016
Guidance on the Preparation of Exceptional Events Demonstrations for Wildfire Events that
May Influence Ozone Concentrations
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Policy Division
Research Triangle Park, NC
-------
-------
SERA
Guidance on the Preparation of Exceptional Events Demonstrations
for Wildfire Events that May Influence Ozone Concentrations
Final
September 2016
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Policy Division
Geographic Strategies Group
Research Triangle Park, North Carolina
-------
Table of Contents
1. Highlights 1
2. Conceptual Model of Event 7
2.1 Overview and Exceptional Events Rule Provisions 7
2.2 Examples of Supporting Documentation 7
3. Clear Causal Relationship between the Specific Event and the Monitored Concentration.... 9
3.1 Overview and Exceptional Events Rule Provisions 9
3.2 Event-related Concentration in the Context of Historical Concentrations 9
3.2.1 Examples of Supporting Documentation 10
3.3 Concept of Different Tiers of Exceptional Events Demonstrations 12
3.4 Key Factor of and Suggested Evidence to Include in Tier 1 Analyses 13
3.4.1 Evidence the Event, Monitor(s), and Exceedance Meet the Key Factor for Tier 1
Clear Causal Analyses 13
3.4.2 Evidence that the Wildfire Emissions Were Transported to the Monitor(s) 14
3.5 Key Factors of and Suggested Evidence to Include in Tier 2 Analyses 15
3.5.1 Evidence that the Event, Monitor(s), and Exceedance Meet the Key Factors for
Tier 2 Clear Causal Analyses 16
3.5.2 Evidence that the Fire Emissions Affected the Monitor(s) 22
3.5.3 Evidence that the Fire Emissions were Transported to the Monitor(s) 23
3.5.4 Summary of Evidence that Could be Used to Meet the Exceptional Events Rule
Elements for Tier 1 and Tier 2 Demonstrations 24
3.6 Tier 3 Analyses to Support the Clear Causal Relationship 25
3.6.1 Relationship of the Event, Monitor(s), and Exceedance to the Key Factors for Tier
2 Analyses 26
3.6.2 Evidence that the Fire Emissions Affected the Monitor(s) 26
3.6.3 Evidence that the Fire Emissions were Transported to the Monitor(s) 26
3.6.4 Additional Evidence that the Fire Emissions Caused the O3 Exceedance 26
3.7 Example Conclusion Statement 30
4. Caused by Human Activity that is Unlikely to Recur at a Particular 30
Location or a Natural Event 30
4.1 Overview and Exceptional Events Rule Provisions 30
4.2 Examples of Supporting Documentation 31
4.3 Example Conclusion Statement 31
5. Not Reasonably Controllable or Preventable 31
5.1 Exceptional Events Rule Provisions 31
5.2 Examples of Supporting Documentation 32
5.3 Example Conclusion Statement 32
6. Public Comment 32
6.1 Exceptional Events Rule Provisions 32
6.2 Examples of Supporting Documentation 33
6.3 Example Conclusion Statement 33
Appendix Al. Example Conceptual Model/Event Summary 34
Appendix A2. Relating Fire Emissions and Downwind Impacts 38
Appendix A3. Interpreting HYSPLIT Results 54
Appendix A4. References for Guidance Document 57
11
-------
Acronyms
AGL
Above ground level
AQS
Air Quality System
CAA
Clean Air Act
CAMx
Comprehensive Air Quality Model with Extensions
CARB
California Air Resources Board
CFR
Code of Federal Regulations
CM
Conceptual model
CMAQ
Community multiscale air quality model
CO
Carbon monoxide
DDM
Direct decoupled method
EER
Exceptional Events Rule
EPA
Environmental Protection Agency
FINN
Fire inventory from the National Center for Atmospheric Research
FIPS
Federal Information Processing Standards
GDAS
Global data analysis system
HAURL
Human activity unlikely to recur at a particular location
HYSPLIT
Hybrid single particle lagrangian integrated trajectory model
K
Potassium
Km
Kilometers
Mb
Millibars
MDA8
Maximum daily 8-hour average for ozone
MODIS
Moderate Resolution Imaging Spectroradiometer
nRCP
not reasonably controllable or preventable
NAAQS
National Ambient Air Quality Standard or Standards
NAM
North American mesoscale forecast system
NCAR
National Center for Atmospheric Research
NDAS
North American mesoscale data analysis system
NEI
National Emission Inventory
NO
Nitric oxide
NOx
Nitrogen oxides
NO2
Nitrogen dioxide
NWS
National Weather Service
O3
Ozone
PM
Particulate matter
111
-------
PMio Particulate matter with a nominal mean aerodynamic diameter less than or equal
to 10 micrometers
PM2.5 Particulate matter with a nominal mean aerodynamic diameter less than or equal
to 2.5 micrometers
Ppb Parts per billion
Q/D 24-hour fire emissions, in tons per day, divided by the distance of the fire to the
monitor, in kilometers
ROG Reactive organic gases
rVOC Reactive volatile organic compounds
SIP State implementation plan
SMARTFIRE Satellite Mapping Automated Reanalysis Tool for Fire Incident Reconciliation
TOG Total organic gases including methane and other reactive volatile organic
compounds
VOC Volatile organic compounds
WRF-CHEM Weather research and forecasting model coupled with chemistry
iv
-------
1. Highlights
Statutory and Regulatory Requirements
The Environmental Protection Agency (EPA) promulgated the Exceptional Events Rule in 20071
to address Clean Air Act (CAA) section 319(b), which allows for the exclusion of air quality
monitoring data influenced by exceptional events from use in determinations of exceedances or
violations of the national ambient air quality standards (NAAQS). The EPA revised the 2007
Exceptional Events Rule in 20162 based on implementation experiences with the exceptional
events data exclusion process. The revised Exceptional Events Rule at 40 CFR 50.14(c)(3)
clarifies that an exceptional events demonstration must include the following elements:
1) A narrative conceptual model that describes the event(s) causing the exceedance or
violation and a discussion of how emissions from the event(s) led to the exceedance or
violation at the affected monitor(s);
2) A demonstration that the event affected air quality in such a way that there exists a clear
causal relationship between the specific event and the monitored exceedance or violation;
3) Analyses comparing the claimed event-influenced concentration(s) to concentrations at
the same monitoring site at other times. The Administrator shall not require a State to
prove a specific percentile point in the distribution of data;
4) A demonstration that the event was both not reasonably controllable and not reasonably
preventable;
5) A demonstration that the event was caused by human activity that is unlikely to recur at a
particular location or was a natural event; and
6) Documentation that the submitting air agency followed the public comment process
Demonstrations prepared by air agencies3 and submitted to the EPA must address each of these
rule elements. This document recommends example language and analyses that may be sufficient
to address these elements in demonstrations for wildfires that influence monitored ozone (O3)
concentrations.4 Air agencies are encouraged to contact their EPA Regional office as soon as the
agency identifies event-influenced data that potentially influence a regulatory decision or when
an agency wants the EPA's input on whether or not to prepare a demonstration.
1 "Treatment of Data Influenced by Exceptional Events; Final Rule" (72 FR 13560, March 22, 2007).
2 The EPA has prepared this guidance to align with the promulgated Exceptional Events Rule revisions signed on
September 16, 2016, and available on the EPA's exceptional events website at http://www2.epa.gov/air-quality-
analysis/treatment-data-influenced-exceptional-events.
3 References to "air agencies" include state, local, and tribal air agencies responsible for implementing the
Exceptional Events Rule. The regulatory text in the 2007 Exceptional Events Rule often uses "State" to apply to "air
agencies." In the context of flagging data and preparing and submitting demonstrations, the role of and options
available to air agencies may also apply to federal land managers of Class I areas and other federal agencies
managing federal land.
4 This guidance addresses wildfire events only, although many technical analyses described in Section 3 apply to
both wildfire and prescribed fires. The EPA intends to include additional detail for demonstrations for prescribed
fires on wildland in a future appendix to this guidance.
1
-------
Purpose of this Document
The EPA developed this document to assist air agencies preparing exceptional events
demonstrations for wildfire influences on O3 concentrations that meet the requirements of CAA
section 319(b) and the Exceptional Events Rule. This guidance document provides three
different tiers of analyses that apply to the "clear causal relationship" criterion within an air
agency's exceptional events demonstration.
The EPA recognizes the limited resources of the air agencies that prepare and submit exceptional
events demonstrations and of the EPA Regional offices that review these demonstrations. One of
the EPA's goals in developing this document is to establish clear expectations to enable affected
agencies to better manage resources as they prepare the documentation required under the
Exceptional Events Rule and to avoid the preparation and submission of extraneous information.
Submitters should prepare and submit the appropriate level of supporting documentation, which
will vary on a case-by-case basis depending on the nature and severity of the event, as
appropriate under a weight of evidence approach. This guidance identifies important analyses
and language to include within an exceptional events demonstration and promotes a common
understanding of these elements between the submitting air agency and the reviewing EPA
Regional office. As a result, this guidance is expected to improve the EPA's efficiency in
reviewing demonstrations prepared consistent with the guidance. While this guidance contains
example analyses that air agencies may use in their demonstrations, air agencies can also prepare
analyses or present documentation not listed or explained in this guidance provided the
information is well-documented, appropriately-applied, technically sound, and supports the
weight of evidence showing for the Exceptional Events Rule regulatory criteria.
The EPA acknowledges the complexity and intricacies of regional conditions prevalent across
the country. The EPA is committed to continuing to provide clarification and assistance to states
as the Exceptional Events Rule is implemented and through communications between the
Regions and the States to ensure that these regional conditions are adequately addressed.
Similarly, we intend to post new information and tools as they become available on the EPA's
exceptional events website at http://www2.epa.gov/air-quality-analysis/treatment-data-
influenced-exceptional-events.
Fire-related Definitions and Terminology
The Exceptional Events Rule at 40 CFR 50.1 (n) defines a wildfire as "... any fire started by an
unplanned ignition caused by lightning; volcanoes; other acts of nature; unauthorized activity; or
accidental, human-caused actions, or a prescribed fire that has developed into a wildfire. A
wildfire that predominantly occurs on wildland is a natural event." The Exceptional Events Rule
and this guidance document differentiate wildfires from prescribed fires in that a prescribed fire
is "any fire intentionally ignited by management actions in accordance with applicable laws,
policies, and regulations to meet specific land or resource management objectives." 40 CFR
50.1(m). An exceptional events demonstration must include a certification that a smoke
management plan or basic smoke management practices was employed. The 2016 Exceptional
Events Rule revisions also codified the following definition of wildland: "Wildland means an
area in which human activity and development are essentially non-existent, except for roads,
railroads, power lines, and similar transportation facilities. Structures, if any, are widely
2
-------
scattered." 40 CFR 50.1(o). This guidance document differentiates between wildfires on
wildland and wildfires on other lands, particularly in the "human activity unlikely to recur at a
particular location or a natural event" section of the document.
This guidance uses the following terminology:
Fire: While this document refers to "a fire" or "the fire," we recognize that there could be
multiple individual fires that, when aggregated, affect O3 concentrations at a given
monitoring site.
Event includes the fire (or fires), the fire's O3 precursor emissions, and the resulting O3
from the fire.
Exceptional event means an event(s) and its resulting emissions that affect air quality in
such a way that there exists a clear causal relationship between the specific event(s) and
the monitored exceedance(s) or violation(s), is not reasonably controllable or preventable,
is an event(s) caused by human activity that is unlikely to recur at a particular location or
a natural event(s), and is determined by the Administrator in accordance with 40 CFR
50.14 to be an exceptional event. It does not include air pollution relating to source
noncompliance. Stagnation of air masses and meteorological inversions do not directly
cause pollutant emissions and are not exceptional events. Meteorological events
involving high temperatures or lack of precipitation (i.e., severe, extreme or exceptional
drought) also do not directly cause pollutant emissions and are not considered exceptional
events. However, conditions involving high temperatures or lack of precipitation may
promote occurrences of particular types of exceptional events, such as wildfires or high
wind events, which do directly cause emissions. See promulgated definition at 40 CFR
50. l(j).
Episode refers to the period of elevated O3 concentrations in the affected area.
Plume means an air mass that contains pollutants emitted by a fire; it may be broad and
mixed into the surrounding air, or the more conventional long narrow plume with well-
defined edges.
Evidence includes, but is not limited to, measurements and analyses based on
measurements.
Tiered Approach for Determining the Level of Evidence Likely to be Necessary in
Demonstrations
Each event submitted by an air agency under the Exceptional Events Rule must meet certain
minimum criteria, as defined in the CAA and the implementing regulations. Some of the
minimum criteria involve a technical analysis that must be tailored to the specific event so as to
make the necessary demonstration. The EPA expects that the documentation and analyses that air
agencies should include in their demonstrations will vary consistent with the event
characteristics, the relationship to the monitor where the exceedance occurred, and the
complexity of the airshed, among other points. The EPA reviews exceptional events
demonstrations on a case-by-case basis using a weight of evidence approach considering the
specifics of the individual event. This means the EPA considers all relevant evidence submitted
with a demonstration or otherwise known to the EPA and qualitatively "weighs" this evidence
based on its relevance to the Exceptional Events Rule criterion being addressed, the degree of
certainty, the persuasiveness, and other considerations appropriate to the individual pollutant and
3
-------
the nature and type of event before acting to approve or disapprove an air agency's request to
exclude data.
This guidance outlines a tiered approach for addressing the clear causal relationship element
within a wildfire/ozone demonstration, recognizing that some wildfire events may be more clear
and/or extreme and, therefore, require relatively less evidence to satisfy the rule requirements.
Tier 1 clear causal analyses should be used for wildfire events that cause clear O3 impacts in
areas or during times of year that typically experience lower O3 concentrations, and are thus
simpler and less resource intensive than analyses for other events. Tier 2 clear causal analyses
are likely appropriate when the impacts of the wildfire on O3 levels are less clear and require
more supportive documentation than Tier 1 analyses. Tier 3 clear causal analyses should be used
for events in which the relationship between the wildfire and the O3 exceedance or violation is
more complicated than the relationship in a Tier 2 analysis, and thus would require more
supportive documentation than Tier 2 analyses. Tier 1 analyses are described in detail in Section
3.4, Tier 2 analyses are described in Section 3.5, and Tier 3 analyses are described in Section 3.6.
The Exceptional Events Rule at 40 CFR 50.14(c)(2) requires an air agency to provide an "Initial
Notification of Potential Exceptional Event" to the EPA Regional office after the air agency
identifies a potential exceptional event. During this process, the EPA expects to discuss potential
event-influenced exceedances with an affected air agency prior to the air agency preparing and
submitting a demonstration. For wildfire events, this "initial notification" is expected to focus, in
part, on observed ozone concentrations and how the wildfire event compares to the key factors
discussed in Sections 3.4 through 3.6 of this guidance. As a result of this discussion, the EPA
and the air agency will likely identify the appropriate tier (Tier 1, 2, or 3) for the event
demonstration. Figure 1 shows a flowchart summarizing the overall process for preparing,
submitting, and reviewing wildfire O3 demonstrations, which includes the Initial Notification
process and recommended review timelines.
Scope of This Guidance Document
Event types: This document focuses on the preparation of demonstrations for wildfires that cause
monitored O3 exceedances or violations. This document does not specifically address
demonstration components that may be necessary for showing prescribed fire impacts on O3
concentrations.5 However, many example technical analyses contained in the "clear causal
relationship" section of this document may also be appropriate for exceptional events
demonstrations for prescribed fires that cause O3 exceedances or violations. The "human activity
unlikely to recur" and "not reasonably controllable or preventable" elements require different
approaches for prescribed fires than those included in this guidance document because prescribed
fires are "human activities" under the Exceptional Events Rule. This guidance describes the
approach appropriate for wildfires, which are natural events.
5 The EPA is developing separate guidance on the preparation of demonstrations for prescribed fire impacts on 03
concentrations.
4
-------
Regulatory determinations: The Exceptional Events Rule clarifies that it applies to the treatment
of data showing exceedances or violations for the following types of regulatory actions:
An action to designate or redesignate an area as attainment, unclassifiable/ attainment,
nonattainment or unclassifiable for a particular NAAQS. Such designations rely on a
violation at a monitoring site in or near the area being designated;
The assignment or re-assignment of a classification category (marginal, moderate,
serious, etc.) to a nonattainment area to the extent this is based on a comparison of its
"design value" to the established framework for such classifications;
A determination regarding whether a nonattainment area has attained a NAAQS by its
CAA deadline. This type of determination includes "clean data determinations;
A determination that an area has data for the specific NAAQS, which qualify the area for
an attainment date extension under the CAA provisions for the applicable pollutant;
A finding of SIP inadequacy leading to a SIP call to the extent the finding hinges on a
determination that the area is violating a NAAQS; and
Other actions on a case-by-case basis if determined by the EPA to have regulatory
significance based on discussions between the air agency and the EPA Regional office
during the Initial Notification of Potential Exceptional Event process.
Outline of this Guidance
This guidance document is organized by Exceptional Events Rule-required elements in the
recommended order for inclusion within an exceptional events demonstration. Section 2 covers
the narrative conceptual model, Sections 3 through 5 discuss the Exceptional Events Rule
criteria, and Section 6 addresses the public comment process. Of particular note, Sections 3.4 -
3.6 discuss the three tiers of analyses to address the clear causal relationship criterion.
Role of this Guidance
The Exceptional Events Rule contains the regulatory requirements for exceptional events and
exceptional events demonstrations. This document provides guidance and applies the rule criteria
to the development of demonstrations for wildfire events that cause monitored ozone
exceedances or violations. It does not impose any new requirements and shall not be considered
binding on any party. If an air agency submits a demonstration using the approach in this
guidance and the EPA concurs with the request to exclude data,6 the EPA will also prepare
documentation to support the decision. The Exceptional Events Rule and the preamble to the rule
contain additional detail regarding those entities authorized to submit demonstrations; the timing
for demonstration preparation, submittal and review; the communications process between air
agencies and the reviewing EPA Regional office; regional consistency; dispute resolution; and
other concepts or rule provisions that apply generally to demonstrations for event types and
pollutant combinations that are not the specific focus of this wildfire/ozone guidance.
6 Submission of a demonstration containing technical analyses consistent with the guidance does not automatically
ensure the EPA's approval. The EPA will review each request under the Exceptional Events Rule on a case-by-case
basis using a weight of evidence approach.
5
-------
Figure 1: Flowchart for the recommended process for air agencies' preparation,
submission, and review of exceptional events demonstrations for wildfire influences on O3,
including communications with EPA Regional offices.
Wildfire-influenced ozone exceedance
Air agency provides the Initial Notification
of Potential Exceptional Event (letter,
email, meeting or documented phone
conversation) to the EPA Regional office
The EPA and air agency work collaboratively to determine
appropriate scope of demonstration based on regulatory
significance and approvability considerations.
The EPA reviews and communicates (by
email or letter and call):
1 - within 60 days (typically) and
2 - with prioritization for package review
based on regulatory significance.
After agreement on scope
(days and monitors) and
regulatory significance of
demonstration package, air
agency flags data requested for
exclusion in AQS.
Do the air agency and the EPA
agree that the exceedance,
monitor, and event qualify for
a Tier 1 or Tier 2 clear causal
analyses?
no
yes
Air agency prepares a
demonstration using Tier 1 or
Tier 2 clear causal analyses,
undergoes 30-day public
comment and submits
demonstration to the EPA with
public comments addressed.
Air agency prepares
demonstration with
Tier 3 clear causal
analyses, undergoes
30-day public
comment and
submits
demonstration to the
EPA with public
comments
addressed.
The EPA reviews and acts on the submitted demonstration:
The EPA generally intends to conduct its initial review of an exceptional events demonstration
with regulatory significance within 120 days of receipt at which point the EPA will respond to
the submitting air agency with a completeness determination and/or a request for additional
information, a date by which the supplemental information should be submitted (if applicable),
and an indicator of the timing of the EPA's final review.
o If the EPA identifies the need for additional information and if the information needed
is minor and a natural outgrowth of previously submitted information, the EPA will not
require the air agency to seek further public comment on the demonstration. However, if
the needed information is significant, the air agency may need to seek additional public
comment before resubmitting to the EPA.
o If the air agency does not submit the additional information within 12 months, then the
EPA will consider the submitted demonstration inactive, and will not continue the
review. If the air agency later decides to request exclusion again, it should submit a new
demonstration.
The EPA intends to make a decision regarding event concurrence as expeditiously as necessary
if required by a near-term regulatory action, but no later than 12 months following submittal of a
complete package.
-------
2. Conceptual Model of Event
2.1 Overview and Exceptional Events Rule Provisions
The Exceptional Events Rule at 40 CFR 50.14(c)(3)(iv)(A) requires that demonstrations include
a narrative conceptual model describing the event. This narrative conceptual model should also
discuss the interaction of emissions, meteorology, and chemistry of event and non-event O3
formation in the area, and, under 40 CFR 50.14(c)(3)(i), must describe the regulatory
significance of the proposed data exclusion. Because this narrative should appear at or near the
beginning of a demonstration, it will help readers and the reviewing EPA Regional office
understand the event formation and the event's influence on monitored pollutant concentrations
before the reader reaches the portion of the demonstration that contains the technical evidence to
support the requested data exclusion. The EPA expects that much of the information the air
agency discussed with or submitted to the EPA during the Initial Notification process would also
be useful in the narrative conceptual model section of a demonstration.
2.2 Examples of Supporting Documentation
The following sections describe the possible types of monitored evidence and technical analyses
that air agencies should include in their demonstration. To be meaningful and clearly interpreted,
air agencies should tie these analyses to a simple narrative describing how emissions from a
specific wildfire (or group of fires) caused O3 exceedances or violations at a particular location
and how these event-related emissions and resulting exceedances or violations differ from typical
high O3 episodes in the area. This narrative description of the cause of the exceedance and the
supporting data and technical analyses will provide a consistent framework by which the EPA
can evaluate the evidence in a demonstration. The interaction of the wildfire plume with non-
event emissions and meteorological conditions of the area will, in part, determine the relevant
evidence.
The narrative conceptual model should describe the principal features of the interaction of the
event and event emissions, transport (e.g., wind patterns such as strength, convergence,
subsidence, recirculation), and O3 chemistry that characterized the O3 episode. This narrative
should highlight key factors in O3 formation for the particular episode, and their relative
importance. A description of the typical urban plume direction (if present), hour of occurrence
for peak O3 concentration, distance downwind, typical wind flow patterns, expected influence of
major sources or emissions categories, relationship between O3 concentrations to diurnal
temperature and growth of mixing layer, the importance of O3 and precursors aloft, and multiple
day carry-over of pollutants are several items that could be used to discuss this conceptual
model. See Appendix A1 for an example of an event summary and conceptual model.
Finally, even if the monitored data and/or technical analyses may not unequivocally support the
clear causal relationship, agencies should submit available information regarding the event and
the monitored exceedances or violations. It may still be possible to explain, with a weight of
evidence approach, why the majority of the data or analyses are consistent with the event causing
elevated O3 concentrations (for example, that most of the meteorological parameters would have
indicated a lower O3 day under non-fire conditions, even though the temperature was high).
7
-------
Where a conceptual model that consistently explains non-event O3 exceedances in the area
already exists or can be formulated, highlighting the differences between the conceptual model
for the event day with the non-event conceptual model can significantly strengthen a
demonstration. For example, if the winds were from an urban center to the monitor of interest on
all non-event O3 exceedance days, but the winds are not from that direction on the event day, this
difference can form a theme in the overall demonstration if it is clearly noted in the conceptual
model discussion. Evidence substantiating the accuracy of the non-event conceptual model
would give this approach more "weight" in the weight of evidence determination. Section 3
discusses this type of evidence. Much of the evidence included in the conceptual model may
have also been included in the air agency's Initial Notification of Potential Exceptional Event.
To promote a shared understanding and interpretation of this information, the EPA recommends
that air agencies include the following information in the narrative conceptual model to the
extent available:
Maps and tables of the wildfire event information including location, size, and extent.
The maps should also include the location of the monitor(s) where data exclusion is
requested. This map and table should clearly identify the wildfire(s) believed by the air
agency to have caused the exceedance, not just a list of wildfires occurring within the
jurisdiction of the submitting air agency.7
Characteristics and description of the monitor with the request for data exclusion. Non-
event similarities and differences between this monitor and nearby monitors should be
explained.
A brief explanation and identification of the cause and point of origin for the event
wildfire(s) (to the extent known).
Examples of media coverage of the event, including special weather statements,
advisories, and news reports.
Smoke forecasts based on meteorology and burn conditions (often provided as part of the
Wildland Air Quality Response Program).
Description of meteorological data from or near the affected monitor and how this relates
to the transport of the wildfire emissions.
Description of the route of the wildfire emissions to the influenced monitor, including
meteorological information (e.g., general atmospheric circulation characteristics)
regarding the transport of wildfire emissions to the monitor.
Non-event O3 formation characteristics of the area normally influencing the monitor (i.e.,
the non-event conceptual model).
Discussion of the differences observed between the non-event conceptual model and
event related conditions causing high O3 concentrations at a particular location.
A summary of spatial and temporal O3 patterns on the day of interest, and days before
and after the event, relative to other, non-event days (either high O3 days, or days with
similar meteorology than the event day), including maps of affected and non-affected
monitors.
7 Burn scar areas by month, 2010-2014: http://activefiremaps.fs.fed.us/burnscar.phf, Federal Land Fires, 1980-
2013, with details (dates, acreage): http://wildfire.cr.usgs.gov/firehistory/viewer/viewer.htm.
8
-------
Description of the regulatory determination anticipated to be influenced by the
exceptional event, including a table of the monitor data requested for exclusion (e.g.,
date, hours, monitor values, and design value calculations with and without the
exceptional event).
NAAQS attainment and classification information, including O3 State Implementation
Plan (SIP) status.
3. Clear Causal Relationship between the Specific Event and the
Monitored Concentration
3.1 Overview and Exceptional Events Rule Provisions
The Exceptional Events Rule requires that demonstrations address the technical element that "the
event affected air quality in such a way that there exists a clear causal relationship between the
specific event and the monitored exceedance or violation" supported, in part, by the comparison
to historical concentrations and other analyses.8 Air agencies should support the clear causal
relationship with a comparison of the O3 data requested for exclusion with historical
concentrations at the air quality monitor. In addition to providing this information on the
historical context for the event-influenced data, air agencies should further support the clear
causal relationship criterion by demonstrating that the wildfire's emissions were transported to
the monitor, that the emissions from the wildfire influenced the monitored concentrations, and, in
some cases, quantifying the contribution of the wildfire's emissions to the monitored O3
exceedance or violation. Table 1 summarizes the tiered analyses for the clear causal relationship
criterion.
Table 1. Summary of Tiered Analyses.
Tier 1: Section 3.4
Tier 2: Section 3.5
Tier 3: Section 3.6
Wildfires that clearly influence
monitored O3 exceedances or
violations when they occur in an
area that typically experiences
lower O3 concentrations. This tier
is associated with an O3
concentration that is clearly
higher than non-event related
concentrations, or occur outside
of the area's normal O3 season.
The wildfire event's O3
influences are higher than
non-event related
concentrations, and fire
emissions compared to the
fire's distance from the
affected monitor indicate a
clear causal relationship.
The wildfire does not fall
into the specific scenarios
that qualify for Tier 1 or Tier
2, but the clear causal
relationship criterion can
still be satisfied by a weight
of evidence showing.
3.2 Event-related Concentration in the Context of Historical Concentrations
As noted above, part of demonstrating a clear causal relationship between the event and the
monitored O3 exceedance involves comparing the event-related exceedance with historical
concentrations measured at the affected monitor or at other monitors in the area during the same
season. Air agencies should compare the data requested for exclusion with the historical
8 See 40 CFR 50.14(c)(3)(iv)(B)-(C).
9
-------
concentrations at the monitor, including all other "high" values in the relevant historical record.
If other values in the historical record are alleged to have been affected by exceptional events,
the EPA recommends identifying those values and including event information to support that
the wildfire caused the monitored exceedance or violation, such as a list of previous wildfire
dates and locations, evidence of stratospheric intrusion, or evidence supporting other event types.
In addition to showing how the level of the event exceedance compares with historical data, air
agencies can also show how the diurnal or seasonal pattern differs, if such a deviation occurred,
due to the event. Effective statistical summaries that characterize non-event, high-concentration
day historical data and the differences seen on event days would carry more weight than
anecdotal or general assertions of when non-event behavior occurs, without evidence or
quantification.
The data used in the comparison of historical concentrations analysis should focus on
concentrations of O3 at the influenced monitor and nearby monitors if appropriate. Evidence of
additional impacts on air quality [carbon monoxide (CO), particulate matter (PM), nitrogen
oxides (NOx), etc.] can also be provided if they provide additional insight.
There is no pass or fail threshold for the historical concentrations data presentation. However,
these comparisons to historical concentrations may inform whether additional evidence is needed
to successfully establish the clear causal relationship element. For example, historical
comparisons conclusively showing that the event-influenced O3 concentration was outside the
range of historical concentrations will likely indicate less additional evidence may be needed to
demonstrate the clear causal relationship. The seasonality of the event-related exceedance versus
other exceedances may be used to determine the appropriateness of Tier 1 (Section 3.4) analyses
for the clear causal relationship criterion. Additionally, air agencies may be able to use the
percentile ranking of the event-influenced data against historical data to determine whether a
Tier 2 analysis (Section 3.5) is appropriate.
3.2.1 Examples of Supporting Documentation
Plot the maximum daily 8-hour O3 concentrations at the affected monitor(s) for the high
O3 seasons (April through October, or other months as appropriate) for at least 5 years.
Figure 2 provides an example of this approach. Alternatively, including separate plots for
each year (or season) may also be an informative approach to presenting this information.
Show time series plots of O3 concentrations at nearby monitors to demonstrate spatial
and/or temporal variability of O3 in the area.
Determine 5-year percentile of the data requested for exclusion on a per monitor basis.
Determine the annual ranking of the data requested for exclusion. This assessment may
show when the non-event O3 during the year with the exclusion request was lower than
surrounding years.
Identify the cause of other "peaks" - fires, other causes, or normal photochemical events,
and provide evidence to support the identification when possible.
Show a time series plot covering 12 months (or the months of the high O3 season)
overlaying all 5 years of data plotted to identify monitored concentrations that are
unusually high for a time of year, and/or that coincide with fire events. An example is
provided below in Figure 3.
10
-------
Discuss trends due to emission reductions from planning efforts, or other variability due
to meteorology or economics of an area, to explain the distribution of data over the
previous 5 years. For example, if a downward trend in O3 concentrations over the 5-year
historical data record obscures the uniqueness of the event-related concentration, the air
agency should use appropriate plots to explain this trend.
Figure 2. Example of an O3 time series plot from an event-influenced monitor to include in
a demonstration.
Chatfield Reservoir Monitor 2008-2014, Aprrl-Oetober Only
All Warm Mooch 03 of&KiedOira
0
0 Hi 423 W2 SSG 1070 12&4 W9S
Ozone Season DaysSiRce 1,2008
11
-------
Figure 3. Example of a seasonal O3 plot, overlaying multiple years of data from an event-
influenced monitor to include in a demonstration.
Chatfield Reservoir April-October Data, 2008-2014
8-hr 03 F fagged
.
f\|l* 3^*41/
t.f
ft # *
O. A* «% 97* *Z mjV*
V 7
,«l
-------
event is "close" to the cut-point for a particular tier, the air agency may choose to employ the
more complex analysis in order to ensure the submittal includes the appropriate level of
information to support the exceptional events demonstration.
3.4 Key Factor of and Suggested Evidence to Include in Tier 1 Analyses
The EPA expects that Tier 1 analyses supporting the clear causal relationship criterion may be
appropriate for wildfires that clearly influence monitored O3 exceedances or violations when
they occur in an area that typically experiences lower O3 concentrations (e.g., few or no O3
exceedances/violations), are associated with an O3 concentration that is clearly higher than non-
event related concentrations, or occur outside of the area's normal O3 season. Many "extreme"
wildfire events could employ Tier 1 analyses. In these situations, O3 impacts should be
accompanied by clear evidence that the wildfire's emissions were transported to the location of
the monitor.
3.4.1 Evidence the Event, Monitor(s), and Exceedance Meet the Key Factor for Tier 1 Clear
Causal Analyses
Key Factor - Seasonality and/or distinctive level of the monitored O3 concentration: The key
factor that delineates event-related monitored O3 concentrations for Tier 1 analyses is the
uniqueness of the concentration when compared to the typical seasonality and/or levels of O3
exceedances. For example, if an event-related exceedance occurs during a time of year that
typically has no exceedances, then that event-related exceedance may be more clearly
attributable to a wildfire than event-related concentrations that occur during the same month or
season as typical high O3 concentrations. If there are other exceedances during the same time of
the year as the wildfire-related exceedance, for example during the normal O3 season, they either
should also be attributable to wildfire (or other exceptional events) or if attributable to normal
emissions and photochemistry, they should be clearly lower in magnitude than the wildfire-
related concentrations. The EPA recommends that event-related exceedances should be at least
5-10 ppb higher than non-event related concentrations for them to be clearly distinguishable.
This key factor is based on the fact that if there are no similar-level non-event exceedances
occurring during the same timeframe as the event-related exceedance, then less evidence may be
necessary to demonstrate the clear causal relationship between the event and the monitored O3
concentration. Following are two types of analyses, either of which an air agency can provide for
this section of the demonstration.
1) Provide a time series plot covering 12 months (or the typical O3 season months
plus months with the event-related exceedance) overlaying at least 5 years or the
length of time data are available if less than 5 years, of O3 monitoring data. An
example is shown in Figure 3.
2) Provide a description of how the seasonality of the event-related exceedance
differs from the typical photochemical O3 season and how other exceedances, if
any, during the time of year of the wildfire-related exceedance are not attributable
to normal emissions and photochemistry, are attributable to wildfire (or other
exceptional events), or are clearly lower in magnitude than the wildfire-related
concentrations.
13
-------
3.4.2 Evidence that the Wildfire Emissions Were Transported to the Monitor(s)
In addition to the evidence suggested in Section 3.4.1, the air agency should supply at least one
piece of additional evidence to support the weight of evidence in a Tier 1 clear causal analysis
that the emissions from the wildfire were transported to the monitor location {i.e., the latitude
and longitude). Air agencies can use either a trajectory analysis or a combination of satellite and
surface measurements to show this transport. This evidence could include:
Trajectory analysis. Atmospheric trajectory models use meteorological data and
mathematical equations to simulate three-dimensional transport in the atmosphere.
Generally, these models calculate the position of particles or parcels of air with time
based on meteorological data such as wind speed and direction, temperature, humidity,
and pressure. Model results depend on the spatial and temporal resolution of the
atmospheric data used and also on the complexity of the model itself. The HYSPLIT
(Hybrid Single-Particle Lagrangian Integrated Trajectory) model is frequently used to
produce trajectories for assessments associated with air quality programs. HYSPLIT
contains models for trajectory, dispersion, and deposition. However, analyses applicable
to exceptional events demonstrations typically use the trajectory component. The
trajectory model, which uses existing meteorological forecast fields from regional or
global models to compute advection (i.e., the rate of change of an atmospheric property
caused by the horizontal movement of air) and stability, is designed to support a wide
range of simulations related to the atmospheric transport of pollutants.
Air agencies can produce HYSPLIT trajectories for various combinations of time,
locations and plume rise. HYSPLIT back-trajectories generated for specific monitor
locations for days of high O3 concentrations illustrate the potential source region for the
air parcel that affected the monitor on the day of the high concentration and provide a
useful tool for identifying meteorological patterns associated with monitored
exceedances. Forward-trajectories from specific wildfire events to specific monitors can
also be used to indicate potential receptors. HYSPLIT trajectories alone cannot
definitively conclude that a particular region contributed to high pollutant concentrations,
but a set of HYSPLIT trajectories that show no wind flow from a particular region on
days with high concentrations might support discounting that region as contributing to the
concentrations. Appendix A3 contains additional information on HYSPLIT trajectory
analyses.
Air agencies could use other trajectory models to demonstrate expected transport.
Exceptional events demonstrations using other trajectory models should contain enough
background information and detail supporting model application to allow reviewers to
thoroughly understand the model and to reproduce the results, if necessary.
Satellite Imagery of Plume with Evidence of the Plume Impacting the Ground. Because
plume elevation is not directly available from simple satellite imagery, plume imagery
alone does not conclusively show that wildfire emissions transported aloft reached a
ground-level monitor. If plume arrival at a given location coincides with elevation of
wildfire plume components (such as PM2.5, CO or organic and elemental carbon), those
14
-------
two pieces of evidence combined can show that smoke was transported from the event
location to the monitor with the elevated O3 concentration.
3.5 Key Factors of and Suggested Evidence to Include in Tier 2 Analyses
If a wildfire event influences O3 concentrations, but these influences are not clearly higher than
non-event related concentrations nor do the event influences occur outside of the affected area's
normal O3 season, then the event would not meet the Tier 1 key factor for seasonality and/or
distinctive level of the monitored O3 concentration and the air agency should not use Tier 1
analyses. The air agency should then determine whether Tier 2 analyses or Tier 3 analyses would
be appropriate. To identify key factors that could differentiate whether Tier 3 analyses or Tier 2
analyses are appropriate, the EPA reviewed previously approved exceptional events
demonstrations, conducted a literature review of case specific fire-03 impacts, and completed
photochemical modeling analyses. Section 3.6 discusses Tier 3 analyses. This section of the
guidance discusses the EPA's methodology for determining the key factors of a Tier 2 analysis.
Section 3.5.1 describes the results of this approach.
Literature review: Fires can impact O3 concentrations by emitting O3 precursors including NOx
and VOCs. These precursor emissions can generate O3 within the fire plume or can mix with
emissions from other sources to generate O3 (Jaffe and Wigder, 2012). Also, in some situations,
including near fires, reduced O3 concentrations have been observed and attributed to O3 titration
by enhanced NO concentrations and reduced solar radiation available to drive photochemical
reactions (Jaffe et al., 2008; Yokelson et al, 2003). The magnitude and ratios of emissions from
fires vary greatly depending on fire size, fuel characteristics, and meteorological conditions
(Akagi et al., 2012). As a result of variable emissions and non-linear O3 production chemistry,
the O3 production from fires is very complex, highly variable, and often difficult to predict (Jaffe
and Wigder, 2012).
Despite the complexities in predicting O3 formation from fire emissions, several studies have
found increases in O3 concentrations attributable to fire. For example, Pfister et al. analyzed
surface O3 data during a high wildfire year in California (2007) with modeled fire impacts and
found monitored 8-hour O3 concentrations were approximately 10 ppb higher when the modeled
fire impacts were high (Pfister et al., 2008). Jaffe et al. analyzed three wildfire periods in the
western U.S. during 2008 and 2012 and compared monitored surface O3 concentrations with two
different modeled estimates of fire contributions to O3 concentrations to find enhancements in O3
when fire impacts were predicted to be high (Jaffe et al., 2013). Many other publications have
found similar relationships between surface O3 concentrations and fire occurrences, using a
variety of technical approaches (Bytnerowicz et al., 2013). One literature study was used to
evaluate the relationship between O3 impact and fire characteristics (Jaffe et al., 2013).
Empirical Relationships between Fire Events and O3 Concentrations in Previous
Demonstrations: The EPA reviewed previous exceptional events demonstrations for specific fire
events to determine if general relationships exist between the magnitude of the fire emissions,
the distance of the fire to O3 monitors, and O3 impacts at those monitors. Between 2010 and
September 2015, the EPA approved two exceptional events demonstrations for fire-related
impacts on O3. In 2011, the EPA concurred on three exceedances of the 1-hour O3 NAAQS near
Sacramento, California in 2008 due to a series of lightning-initiated wildfires throughout
15
-------
northern California. In 2012, the EPA concurred with the exclusion of eight 8-hour daily
maximum O3 exceedances during April 2011 in Kansas caused by wildfires and prescribed fires.
Modeling Studies of O3 Impacts from Fires: To support the development of this guidance and to
assess the relationship between fire source strengths and resultant O3 concentrations at various
distances from the fire, the EPA conducted modeling analyses for fires identified in the EPA's
2011 National Emissions Inventory (NEI).9 See Appendix A2. Four fires of varying strengths
and locations were simulated with the Community Multiscale Air Quality Model (CMAQ)
model. The O3 impacts of these fires were estimated using a source apportionment technique
(Kwok et al., 2015). Consistent with previous literature studies, the EPA modeling suggests that
NOx and VOC emissions can lead to significant increases in O3 concentrations downwind of the
fire. The simulated O3 increases are related to distance downwind from the fire and the
magnitude of the fire emissions. Examination of this modeling and related studies suggests that it
is appropriate to use a simple Q/D (emissions/distance) metric to conduct a screening assessment
of potential fire impacts. This model application was evaluated against monitoring data and
appears to capture the ambient relationships between CO and O3 measured in the vicinity of
smoke plumes. The EPA acknowledges that the science continues to emerge in modeling the O3
impacts of fires (e.g., plume chemistry, plume rise). The 2011 modeling includes some limited
treatment of the sunlight-blocking impacts of smoke on O3 photochemistry.
The EPA used the general relationships between O3 impacts and fire characteristics from the
modeling study, in combination with the assessment of previously approved demonstrations and
fire case-studies from the peer-reviewed literature to develop two key factors (Section 3.5.1) for
a Tier 2 clear causal analysis. These two key factors act together to identify event and monitor
pairs that may be appropriate for a Tier 2 demonstration. Section 3.5.1 includes a recommended
value and guidance for determining Q/D.
3.5.1 Evidence that the Event, Monitor(s), and Exceedance Meet the Key Factors for Tier 2
Clear Causal Analyses
This section details the evidence to be included in a Tier 2 analysis for the clear causal
relationship rule element.
Key Factor #1 - Fire emissions and distance of fire(s) to affected monitoring site location(s): At
least one air quality related program (i.e., determining impacts at Class I areas) uses an emissions
divided by distance (Q/D) relationship as a key factor for determining the influence of emissions
on a downwind monitor. The EPA believes that it is appropriate to use a similar approach, along
with key factor #2 detailed below, to determine if a Tier 2 analysis provides sufficient evidence
to satisfy the clear causal relationship criteria for wildfire O3 demonstrations. To determine an
appropriate and conservative value for the Q/D threshold (below which the EPA recommends
9 The 2011 NEI is the most current publicly available version of the EPA's National Emissions Inventory at the time
of development of this guidance. We used the NEI rather than other emissions sources because it is nationally
consistent, quality-assured, and an inventory that is reviewed by state, local, and tribal air agencies and represents a
comprehensive and detailed estimate of air emissions of criteria pollutants, criteria precursors, and hazardous air
pollutants from air emissions sources. The EPA compiles the inventory from data provided by state, local and, tribal
air agencies for sources in their jurisdictions and then supplements these data with additional information developed
by the EPA.
16
-------
Tier 3 analyses for the clear causal relationship), the EPA conducted a review of approved
exceptional events demonstrations, a literature review of case specific fire-Cte impacts, and
photochemical modeling analyses, as described above. The three analyses generally showed that
larger O3 impacts occurred at higher Q/D values. The reviews and analyses did not conclude that
particular O3 impacts will always occur above a particular value for Q/D. For this reason, a Q/D
screening step alone is not sufficient to delineate conditions where sizable O3 impacts are likely
to occur. Given this, the EPA recommends, as the first of two key factors, that the Q/D (as
described below) should be > 100 tons per day/kilometers (tpd/km). The rationale for the
recommendation of > 100 tpd/km as a conservative indicator of O3 impacts is based on the Q/D
ratio for previously approved fire-related O3 exceptional events demonstrations and the modeling
results that showed the largest O3 impacts were often associated with high Q/D values. The O3
values within the approved demonstrations generally were associated with Q/D values above 50
tpd/km (Figure A2-1), though not all the concentrations shown were clear cases of causal
contribution from fires. The largest O3 impacts from the modeling studies of the two largest fires
(Wallow and Flint Hill fires) were associated with Q/D values above 100 tpd/km (Figure A2-5),
and large O3 impacts were not observed in the modeling of the two smaller fires (Big Hill and
Waterhole fires). Based on results from these analyses and reviews, if the Q/D (as defined and
calculated in Section 3.5.1) is > 100 (tpd/km), and key factor #2 is also met, then Tier 2 analyses
for the clear causal relationship criterion are likely appropriate. Following is a description of how
an air agency could develop a Q/D analysis.
Calculate Q/D for the event and monitor pairs:
Determine fire emissions (Q): For the purposes of exceptional events tiering, fire
emissions (Q in the Q/D expression) is defined as the daily sum of the NOx and reactive-
VOC emissions (in units of tons per day) from specific wildfire events impacting the O3
monitor on the day of the O3 exceedance. Air agencies should describe and characterize
in the conceptual model/event summary section of the demonstration all fires included in
the calculation of Q/D. Since a fire event can span several days and because fire
emissions may not impact a monitor on the day that they are generated, this guidance
suggests the following approach for assessing a range of days to determine the maximum
Q/D value to use for the screening test:
1) Determine the date of the 1st hour in the period of the 8-hour (or 1-hour) O3
average that is the subject of the demonstration. Example: August 15, 2014.
2) Determine the date of the 8th hour of that 8-hour period, which may be the same
as the first date or the following date. Example: August 16, 2014 if the 1st hour
occurred at 9 p.m.
3) Identify fires generating emissions on these one or two dates and identify the date
prior to the date of the 1st hour. Including the latter date allows for the possibility
that fire emissions on one day affected ozone on the next day. These are the two
or three dates that will be included in assessing the clear causal relationship.
Example: August 14, 15, and 16.
17
-------
The EPA recommends generating 24-hour back trajectories from the affected O3
monitoring site(s) beginning at each hour of these two or three dates. Identify fires
that are close to any of these back trajectories. Example: the air agency identifies
three fires: Fire A, Fire B and Fire C.
4) Identify the latitude/longitude of each fire for each day. Determine "D," the
distance in kilometers between the fire's latitude/longitude and the affected O3
monitor for each fire for each day. The reported latitude and longitude of the fire
from inventories is generally the centroid of the fire parcel. However, air agencies
are not limited to calculating distance based on the centroid of the fire parcel,
provided the latitude/longitude calculation is well-documented and supported.
5) For each fire and each day, identify the sum of NOx and reactive VOC (rVOC)
emissions in tons/day. If only total organic gas (TOG) emissions (versus rVOC)
are available, multiply the TOG emissions by 0.6 to represent the reactive fraction
that can contribute to O3 formation (see Appendix A2).10 Alternatively, sum the
specific rVOC emissions or use a multiplier other than 0.6 with appropriate
justification. This step is designed to account for the fact that some of the gases
included in the TOG emissions estimates do not contribute to ozone formation.
Day-specific emissions estimates should be readily available for wildfire (and
prescribed fire events) that occur during NEI years using the EPA methods.11 In
addition to the actual emissions estimates (NOx, VOC, CO, SO2, PM tons/day),
the NEI methods also result in many other data fields that will be made available
(date of fire occurrence, fire event name, state/county Federal Information
Processing Standard (FIPS) code, latitude, longitude, quality assurance flag, fire
type, acres burned). Detailed information about how the EPA develops
inventories for fires on wildlands is part of the latest NEI documentation available
on the EPA's Clearinghouse for Inventories and Emissions Factors (CHIEF) at
https://www.epa.gov/chief. In general, the EPA's approach for estimating fire
emissions relies on a combination of satellite detection of fires merged with on-
the-ground observational data (especially with activity data submitted by local air
regulatory and forestry agencies) and where available combined with models that
specify fuel loading, fuel consumption, and emission patterns/factors. These
emissions are based on the latest version of the Satellite Mapping Automated
Reanalysis Tool for Fire Incident Reconciliation (SMARTFIRE) system
10 rVOC is defined for the purposes of estimating wildland fire impacts as part of a screening level demonstration as
the sum of all VOC species excluding methane and those mapped to "nonreactive" or "unknown" when applying the
most up-to-date VOC speciation profile to total organic gases. VOC speciation profiles are available online as part
of EPA's SPECIATE database at https://www3.epa.gov/ttnchiel/software/speciate/.
11 An official version of the NEI is generally publicly available within 18 months of the end of the NEI year. States,
however, have 1 year to compile and submit inventory data at the end of an NEI year. In some cases, official NEI
data may not be available for use in an air agency demonstration, either because of the time lag between the end of
the inventory year and NEI availability or because the fire did not happen in a NEI year. In these scenarios, air
agencies may use any other well-documented and well-supported source of emissions and activity data.
18
-------
(ihttp://www.airfire.org/smartfire/). Air agencies can use sources other than the
EPA's NEI for their fire emissions and activity data as part of an exceptional
events demonstration if the air agency believes that another source of information
more accurately characterizes the event and its resulting emissions. Any
additional source of emissions and activity data must be well-documented and
supported.
To estimate fire-related emissions in non-NEI years, air agencies may use other
techniques to represent fire emissions, especially methods that have been agreed
upon by multiple public agencies (e.g., http://www.airfire.org/data/playground/)
or emission estimates that reside in the published literature. The fire activity data
and emissions estimation techniques should be well-documented and supported.
The EPA encourages the use of ground-based observations and local fuel
information whenever possible as these factors can significantly improve the
resulting estimates of fire emissions. As resources allow, to assist air agencies in
locating fire-related emissions in non-NEI years, the EPA anticipates providing
year and day-specific fire event emissions summaries using similar methodologies
to that used in the NEI.
6) Check the fires individually to see whether any one of them had Q/D >100 for
any of the days. If yes, evaluate key factor #2. If Q/D < 100, then the air agency
should consider the fires in aggregate, or use Tier 3 analyses to support the clear
causal relationship criterion.
7) If any of the individual fires do not have Q/D > 100, determine whether the fires
satisfy the Q/D test when aggregated. For each day of fire, weight the distances
between the fire locations and the O3 monitor by the NOx+rVOC emissions for
that day to get an emissions-weighted D. Sum the NOx+rVOC emissions of all
three fires (e.g., Fire A, Fire B and Fire C from the above example) from the day,
and calculate Q/D using the emissions sum and the distance.
For situations where only one fire parcel is thought to affect a monitored ozone
concentration:
The distance between the latitude and longitude of the monitor and the latitude
and longitude of the fire (accounting for the curvature of the Earth) should be
used. The reported latitude and longitude of the fire from inventories is generally
the centroid of the fire parcel. However, air agencies are not limited to calculating
distance based on the centroid of the fire parcel, provided the latitude/longitude
calculation is well-documented and supported.
For situations where multiple fires are thought to contribute to a monitored ozone
concentration:
The distance (D) between the "fire" and monitor should be determined using an
emissions weighted average distance between the fire and the monitor.
19
-------
For example 1, if two fire parcels A1 and B1 are found to contribute to an O3
concentration on a given day, find the locations and emissions (sum of NOx and
VOC) for each fire parcel. If parcel A1 is the closer fire, at 100 km from the
monitor and has relatively low emissions such as 1,000 tons while parcel B1 is
further from the monitor at 200 km and has larger emissions such as 10,000 tons,
then the air agency should calculate the emissions weighted distance as emissions
A1 times distance A1 plus emissions B1 times distance Bl. Then this weighted
sum is divided by the sum of the emissions A1 and Bl. Or, filling in the numbers,
100*1,000 + 200*10,000 divided by 11,000. The emissions weighted distance
would then be 190.9 km, and the Q/D would be 11,000/190.9 or 57.6 tons/km.
Applying this approach indicates that the weighted distance would be closer in
magnitude to the fire with the larger emissions {i.e., Fire Bl), but slightly smaller
than the actual distance to Fire Bl because of the contribution from the closer,
smaller fire.
For example 2 (involving Fires A2, B2, and C2), if an air agency determines that
3 fires contribute to the O3 exceedance or violation, the distance between the
center of fire parcel and the monitor would be calculated as follows:
. . 7 j j. (DA2QA2) + (DB2QB2) + (DC2DC2)
emissions weiqhtea averaqe distance =
y y qA2 + qB2 + qC2
Example 1:
Distance
between center
of fire parcel
and monitor
(km)
Sum of NOx
and VOC
emissions on
day being
investigated
Emissions
weighted
distance
between the
fire and
monitor
Q/D
Fire A1
100 km
1,000 tons
9.09
Fire Bl
200 km
10,000 tons
181.8
11,000 tons
190.9
57.6 tons/km
Example 2:
Distance
Sum of NOx
Weighted
Q/D
between center
and VOC
distance
of fire parcel
emissions on
between the
and monitor
(km)
day being
investigated
Fire A2
100 km
10,000 tons
62.5
Fire B2
150 km
5,000 tons
46.88
Fire C2
50 km
1,000 tons
3.125
16,000 tons
112.5
142.2 tons/km
20
-------
8) If the aggregate approach results in a Q/D >100 for the day, evaluate key factor
#2. Apply the same aggregated approach for the other identified days. If Q/D
under the aggregate approach is < 100, then the air agency would follow Tier 3
analyses for the clear causal relationship criterion. The demonstration should
show all calculations and values and clearly describe the result of the calculation,
and the emissions, distance, and any assumptions that the air agency made in
developing the Q/D ratio. The EPA acknowledges that some exceedances may be
caused by many small fires that when aggregated do not result in a Q/D >100 for
the day. When combined with satisfactory corroborating information, it is
possible that aggregated wildfires with a Q/D less than 100 could result in an
approved demonstration (following Tier 3 analyses). This corroborating
information is described in Section 3.6 of this guidance.
Key Factor #2 - Comparison of the event-related O3 concentration with non-event related high
O3 concentrations: The second key factor for a Tier 2 clear causal analysis considers the
characteristics of the event-related concentration versus the non-event O3 concentration
distribution at the monitor. Addressing key factor #2 involves showing that the exceedance due
to the exceptional event:
is in the 99th or higher percentile of the 5-year distribution of O3 monitoring data, OR
is one of the four highest O3 concentrations within 1 year (among those concentrations
that have not already been excluded under the Exceptional Events Rule, if any).
Applying this key factor recognizes that an air agency will likely need more detailed information
to establish a clear causal relationship between the event and the monitored exceedance in an
area or season with elevated non-event related O3 concentrations. Therefore, limiting the Tier 2
analysis to events in the 99th or higher percentile of 5 years of monitoring data will generally
ensure the event-influenced data are high compared to other data at the monitoring site. If event-
related concentrations have already been excluded for this year, then those values should not be
included when determining the ranking. However, if the non-event O3 concentrations at a
monitor in the year (or season) when the event-related O3 exceedance occurred are low when
compared with other surrounding years in the 5 year record, an exceedance in this "low" O3 year
could still affect design value calculations and determinations within the scope of the
Exceptional Events Rule. Therefore, if the data requested for exclusion are one of the four
highest within 1 year (among those concentrations that have not already been excluded under the
Exceptional Events Rule, if any), the key factor would be met. If both key factors (#1 and #2) are
met, then a Tier 2 demonstration will likely be sufficient.
Compare the event-related O3 concentration with non-event related high O3 concentrations:
1) Provide the percentile ranking of the data requested for exclusion when compared with
the most recent 5 years of monitoring data. Include the plot showing this result or
reference the generated plot in another section of the demonstration.
2) If data are in the 99th (or higher) percentile OR are one of the top four O3 maximums
within 1 year AND key factor #1 is satisfied AND the EPA Regional office and the
21
-------
affected air agency have discussed the potential event THEN the air agency should
prepare a Tier 2 analysis to support the clear causal relationship criterion. If the data are
not in the 99th (or higher) percentile and are one of the top four O3 maximums within 1
year, or if the EPA Regional office identified that a more complex analysis is needed,
then the air agency should prepare a Tier 3 analysis for the clear causal relationship
criterion.
3.5.2 Evidence that the Fire Emissions Affected the Monitor(s)
In addition to the evidence suggested in Section 3.5.1, the air agency should supply at least one
piece of additional evidence to support the weight of evidence that the emissions from the
wildfire affected the monitored O3 concentration. Air agencies can use the following example
evidence to demonstrate the wildfire emissions were present at the altitude of the monitor(s).
This evidence could include any of the following:
1) Evidence of changes in spatial/temporal patterns of O3 and/or NOx.
2) Photographic evidence of ground-level smoke at the monitor
3) Concentrations of supporting ground level measurements [CO, PM (mass or speciation),
VOCs, or altered pollutant ratios]
While fires typically generate emissions of CO, NO, NO2, VOCs, PM10, and PM2.5,
anthropogenic sources, such as industrial and vehicular combustion, also emit these pollutants.
Therefore, the air agency should distinguish the difference in the non-event pollutant behavior
(e.g., concentration, timing, ratios, and/or spatial patterns) from the behavior during the event
impact to more clearly show that the emissions from the fire(s) affected the monitor(s). Air
agencies can use evidence from regulatory and non-regulatory (e.g., special purpose, emergency)
monitors to support these analyses.
Specific analyses to support the above-identified evidence include the following:
Satellite evidence of smoke or precursors (NOx) at the monitoring site.
https://www.epa.gov/hesc/remote-sensing-information-gateway and
http://arset.gsfc.nasa.gov/airquality/applications/fires-and-smoke may be helpful
resources.
Photographic evidence of ground-level smoke at the monitor.
Plots of co-located or nearby CO, PM2.5, PM10, or O3 and PM2.5 precursor concentrations
in the same airshed (or nonattainment/near nonattainment area) that have increases or
differences in typical behavior that indicate the wildfire's emissions influenced the
monitor. Elevated levels of CO or PM (including pre-cursors) at an affected O3 monitor
upwind of urban centers or occurring at non-commute times at a monitor within an urban
area despite the lack of a surface inversion would be consistent with wildfire plume
impact. Include an explanation of the plots.
22
-------
Elevated light extinction measurements at or near the O3 monitoring site that cannot be
explained by emissions from other sources and are consistent with wildfire impact.
The timing and spatial distribution of NO, NO2, and O3, shown with data from multiple
monitoring sites. These pollutant concentrations may vary when influenced by a wildfire
plume. Elevated levels that are widespread throughout a region, or are upwind of the
urban area, may be due to impact of a fire plume. Peaks at locations and times different
than those normally seen in an O3 episode can indicate fire plume impact.
Differences in CO:NOx ratios: The ratio of CO and NOx emissions depends on their
source; for agricultural burning it is about 10-20, for wildfire and prescribed wildland
burning about 100 (Dennis et al., 2002), whereas for high-temperature fossil fuel
combustion sources it is more like 4 (Chin et al., 1994). Thus, an unusually high CO/NOx
ratio is consistent with fire impact. Similarly, the CO/PM10 emission ratio is 8-16 in fires,
but 200-2000 for vehicles (Phuleria et al., 2005). Changes in CO and CO ratios might be
difficult to discern in an area dominated by vehicular CO, however, as the fire signal may
be small in comparison.
PM speciation data: PM2.5 emissions from forest fires often contain elevated levels of
organic carbon (OC) and occasionally are enriched in water soluble potassium (K)
(Watson et al., 2001). Levoglucosan, a tracer molecule, is a constituent of smoke from
biomass burning that can serve as an indicator for fire; PM10 from wood smoke is 14% or
higher levoglucosan by mass (Jordan et al., 2006; Dennis et al., 2002). Co-located or
nearby particle speciation data (OC, K, and/or levoglucosan) can be used to indicate fire
impacts.
3.5.3 Evidence that the Fire Emissions were Transported to the Monitor(s)
In addition to the evidence suggested in Sections 3.5.1 and 3.5.2, an air agency should provide
evidence showing the emissions from the wildfire were transported to the monitor location {i.e.,
the latitude and longitude). Air agencies can use either a trajectory analysis or a combination of
satellite and surface measurements to show this transport. (These recommendations are the same
as for Tier 1 demonstrations in Section 3.4.2, but are explained here again for completeness).
Trajectory analysis. Atmospheric trajectory models use meteorological data and
mathematical equations to simulate three-dimensional transport in the atmosphere.
Generally, these models calculate the position of particles or parcels of air with time
based on meteorological data such as wind speed and direction, temperature, humidity,
and pressure. Model results depend on the spatial and temporal resolution of the
atmospheric data used and also on the complexity of the model itself. The HYSPLIT
(Hybrid Single-Particle Lagrangian Integrated Trajectory) model is frequently used to
produce trajectories for assessments associated with air quality programs. HYSPLIT
contains models for trajectory, dispersion and deposition. However, analyses applicable
to exceptional events demonstrations typically use the trajectory component. The
trajectory model, which uses existing meteorological forecast fields from regional or
global models to compute advection (i.e., the rate of change of an atmospheric property
23
-------
caused by the horizontal movement of air) and stability, is designed to support a wide
range of simulations related to the atmospheric transport of pollutants.
Air agencies can produce HYSPLIT trajectories for various combinations of time,
locations and plume rise. HYSPLIT back-trajectories generated for specific monitor
locations for days of high O3 concentrations illustrate the potential source region for the
air parcel that affected the monitor on the day of the high concentration and provide a
useful tool for identifying meteorological patterns associated with monitored
exceedances. Forward-trajectories from specific wildfire events to specific monitors can
also be used to indicate potential receptors. For purposes of assessing wildfire
exceptional events, HYSPLIT trajectories alone cannot definitively conclude that a
particular region contributed to high pollutant concentrations, but a set of HYSPLIT
trajectories that show no wind flow from a particular region on days with high
concentrations might support discounting that region as contributing to the
concentrations. Appendix A3 contains additional information on HYSPLIT trajectory
analyses.
Air agencies could use other trajectory models to demonstrate expected transport.
Exceptional events demonstrations using other trajectory models should contain enough
background information and detail supporting model application to allow reviewers to
thoroughly understand the model and to reproduce the results, if necessary.
Satellite Imagery of Plume with Evidence of the Plume Impacting the Ground. Because
plume elevation is not directly available from simple satellite imagery, plume imagery
alone does not conclusively show that wildfire emissions transported aloft reached a
ground-level monitor. If plume arrival at a given location coincides with elevation of
wildfire plume components (such as ground level measurements of PM2.5, CO or organic
and elemental carbon), those two pieces of evidence combined can show that smoke was
transported from the event location to the monitor with the elevated O3 concentration.
3.5.4 Summary of Evidence that Could be Used to Meet the Exceptional Events Rule Elements
for Tier 1 and Tier 2 Demonstrations
Table 2 summarizes the technical support that air agencies can use to support the clear causal
relationship in a Tier 2 demonstration, compared with a Tier 1 demonstration.
Table 2. Clear Causal Relationship Technical Demonstration Components Recommended
for Tier 1 and Tier 2 Demonstrations.
Tier 1 Analyses Should Include
Tier 2 Analyses Should Include
Comparison of the fire-influenced
exceedance with historical concentrations
Comparison of the fire-influenced
exceedance with historical concentrations
Evidence that the fire and monitor(s) meet
the key factor
Evidence that the fire and monitor(s) meet
the key factors (#1 and #2)
Evidence of transport of fire emissions
from fire to the monitor (one of these):
Evidence of transport of fire emissions
from fire to the monitor (one of these):
24
-------
Trajectories linking fire with the
monitor (forward and backward),
considering height of trajectories
Satellite evidence in combination
with surface measurements
Trajectories linking fire with the
monitor (forward and backward),
considering height of trajectories
Satellite evidence in combination
with surface measurements
Evidence that the fire emissions affected
the monitor (one of these):
Visibility impacts (satellite or
photo)
Changes in supporting ground level
measurements
Satellite NOx enhancements
Differences in spatial/temporal
patterns
3.6 Tier 3 Analyses to Support the Clear Causal Relationship
Although the EPA has identified specific wildfire/Cte scenarios that are appropriate for either
Tier 1 or Tier 2 analyses to demonstrate the clear causal relationship criterion, and we have
identified analyses and key factors associated with these tiers based on generally available data,
we do not intend to imply that demonstrations for all other wildfire/Cb events must include more
analyses with increasing complexity.12 Rather, this guidance is intended to indicate that if a
wildfire/ozone event satisfies the key factors for either Tier 1 or Tier 2 clear causal analyses,
then those analyses are the only analyses required to support the clear causal relationship
criterion within an air agency's demonstration for that particular event. Other wildfire/Cb events
will be considered based on Tier 3 analyses, but some Tier 3 clear causal analyses may also be
relatively straightforward and/or established with limited evidence. For example, a wildfire event
may cause an exceedance during an area's photochemical O3 season that is the fifth highest
concentration in a year and falls within the 98th percentile of the 5-year distribution. Because the
event occurs during the time of year as typically high O3 concentrations, it would not qualify for
Tier 1 analyses. Similarly, because the concentration in question is the fifth (versus fourth) high
value and falls within the 98th (versus 99th) percentile, the event would also not qualify for Tier 2
analyses. However, when addressing the (Tier 3) clear causal relationship criterion within its
demonstration, the affected air agency might complete a comparison to historical concentrations
(required for all event/pollutant combinations under 40 CFR 50.14(c)(3)(iv)(C)), prepare
backward and forward trajectories from the wildfire to the affected monitor, submit satellite
imagery showing the smoke plume over the affected monitor, and submit a vertical ozone profile
or model simulations. Together this information might satisfy the clear causal relationship
criterion under a weight of evidence approach. Other, more complicated relationships between
12 In developing the tiering approach, the EPA intended to base the key factors within Tier 1 and
Tier 2 on data or information that is generally available and accessible to all air agencies. We
recognize that other information may be equally (or more) convincing and carry more "weight"
under a weight of evidence assessment, but these data and/or tools may not be as widely
available. As noted in this guidance, it is not our intent to prevent air agencies from using any
relevant, well-documented, appropriately-applied and technically sound evidence.
25
-------
the wildfires and influenced O3 concentrations may require additional detail to satisfy the clear
causal relationship element.
Regardless, as indicated in the example above, the EPA anticipates that air agencies can build
upon the Tier 1 or Tier 2 analyses with the analyses described in this section (or other
appropriate analyses/tools). The EPA does not expect an air agency to prepare all identified
analyses but only those that add to their weight of evidence supporting the clear causal
relationship. As with all exceptional events demonstrations, the submitting air agency and the
EPA Regional office should discuss the appropriate level of evidence during the Initial
Notification process.
3.6.1 Relationship of the Event, Monitor(s), and Exceedance to the Key Factors for Tier 2
Analyses
As part of the weight of evidence showing for the clear causal relationship rule element, air
agencies should explain how the events, monitor and exceedance compare with the key factors
outlined in Section 3.5.1. The relationship of the event to the Tier 2 key factors may help inform
the amount of additional information that will be needed to support Tier 3 analyses.
3.6.2 Evidence that the Fire Emissions Affected the Monitor(s)
Because the relationship between the wildfire-related emissions and the monitored exceedance or
violation cannot clearly be shown using Tier 1 or Tier 2 analyses, air agencies will need
additional evidence to support the clear causal relationship criterion and show that the wildfire
emissions affected the monitor. The Tier 3 clear causal relationship analyses could include
multiple analyses from those examples listed in Section 3.6.4. Each additional piece of
information that supports the event's influence will strengthen the air agency's position.
3.6.3 Evidence that the Fire Emissions were Transported to the Monitor(s)
To demonstrate a clear causal relationship between the event's emissions and the monitored O3
exceedance, air agencies should show that the emissions from the wildfire were clearly
transported to the monitor. This will likely require a trajectory analysis or the satellite plume
analysis described in Section 3.5.3.
Because the uncertainty of trajectory analyses increases with transport distance, frontal passages,
and complex wind/terrain issues, additional information, such as analyses of surface meteorology
(wind speed and direction), could further support the clear causal relationship rule element.
3.6.4 Additional Evidence that the Fire Emissions Caused the O3 Exceedance
Depending on evidence supplied in other sections of the demonstration, an air agency may
further support the clear causal relationship between the wildfire and the O3 exceedance with
matching day analyses, statistical regression models, or photochemical models, all of which are
described in more detail below.
26
-------
Comparison of O3 Concentrations on Meteorologically Similar Days (Matching Day
Analysis). O3 formation and transport are highly dependent upon meteorology. Therefore,
a comparison between O3 on meteorologically similar days with and without fire impacts
could support a clear causal relationship between the fire and the monitored
concentration. Both O3 concentrations and diurnal behaviors on days with similar
meteorological conditions can be useful to compare with days believed to have been
influenced by fire. Since similar meteorological days are likely to have similar O3
concentrations, significant differences in O3 concentrations among days with similar
meteorology may indicate influences from non-typical sources.
Meteorological variables to include in a similar day (or "matching day") analysis should
be based on the parameters that are known to strongly affect O3 concentrations in the
vicinity of the monitor location. These variables could include: daily high temperature,
hourly temperature, surface wind speed and direction, upper air temperature and pressure
[such as 850 or 500 millibar (mb) height], relative or absolute humidity, atmospheric
stability, cloud cover, solar irradiance, and others as appropriate See e.g., Anderson and
Davis, 2004; Camalier et al, 2007; Eder et al, 1993; Eder et al, 1994. These parameters
should be matched within an appropriate tolerance. Since high O3 days may be relatively
rare, air agencies should examine several years of data for similar meteorology versus
restricting the analysis to high O3 days only. The complete range of normal expected O3
on similar meteorology days will have value in the demonstration. A similar day analysis
of this type, when combined with a comparison of the qualitative description of the
synoptic scale weather pattern (e.g., cold front location, high pressure system location),
can show that the fire contributed to the elevated O3 concentrations. Air agencies may
also want to consider non-meteorological factors such as choosing days with similar,
non-event emissions (possibly avoiding holidays and special public events, weekday
versus weekend mismatches, and other days with unusual emissions). In a recently
submitted demonstration,13 the state of Kansas included an analysis showing the
synoptic-scale weather pattern typing along with an evaluation of basic meteorological
parameters similar to the "Matching Days" analysis described here. Although this
demonstration preceded issuance of the instant guidance, the methods may be useful for
air agencies conducting Tier III analyses.
Statistical Regression Modeling
Air agencies can use O3 predictions from regression equations to assess the wildfire's
contribution to O3 concentrations. Regression is a statistical method for describing
relationships among variables. For estimating air quality concentrations, regression
equations are developed to describe the relationship between pollutant concentrations
(referred to as the prediction) and primarily meteorological variables (referred to as the
predictors). Because regression equations are developed with several years of data, they
represent the relationship between air quality and meteorology under typical emission
patterns; even if some historical exceptional events data are included in the development,
the influence of those days will likely be small on the developed model provided there
are far more typical days than event-related days. Therefore, the difference between the
predictions and observations can provide a reasonable estimate of the air pollution caused
13 Available at: http://www2.epa.gov/sites/production/files/2015-05/documents/kdhe_exeventsJinal_042011.pdf.
27
-------
by event-related emissions (e.g., emissions from wildfires) provided the analysis accounts
for the typical remaining variance of typical days (variability in monitored data not
predicted by the model).
Air agencies can develop the regression equation using the O3 data for the monitor(s)
under investigation and meteorology data from the closest nearby National Weather
Service station. A small subset of the data should be reserved for testing the regression
equation. Once a regression equation has been properly developed and tested, it can be
used to predict the daily maximum O3 values. The differences between the predicted
values and the measured values are analyzed, and the 95th percentile of those positive
differences (observed O3 is greater than predicted) is recorded. This 95 percent error
bound is added to the O3 value predicted by the regression equation for the flagged days,
and any difference between this sum and the observed O3 for the flagged day may be
considered an estimate of the O3 contribution from the fire if evaluation of the top 5th
percentile shows similar O3 days in the absence of smoke are rare or not observed.
Users of regression models should consider the uncertainties in the model's prediction
abilities, specifically at high concentrations, before making conclusions based on the
modeled results. A key question when considering model uncertainty is whether the
model predicts O3 both higher and lower than monitored values at high concentrations
(above 65 or 70 ppb) or whether the model displays systematic bias on these high
monitored days.
The limitations of the regression equation itself defines the limitations of this method.
This approach is more rigorous than a comparison to similar meteorological days in that
it considers the relationship between meteorological parameters, but regression is less
rigorous than air quality modeling, which employs more parameters and more physical
processes in its calculations. While statistical modeling does not resolve all the
complexities of the atmosphere, carefully crafted regression models can provide an
estimate of contribution to support the clear causal relationship portion of an exceptional
events demonstration. There are several methods for developing a regression equation to
estimate O3 concentrations from meteorological variables. See, e.g., Camalier et al., 2007;
STI, 2014.
Photochemical modeling
This section describes the air quality modeling tools best suited for estimating wildfire
emissions impacts in demonstrations needing a more refined assessment. Secondary
pollutant impacts, such as O3 and PM2.5, need to be assessed at various spatial scales
(near-source and long-range transport) for a variety of regulatory programs. Modeling
systems used for these assessments should be appropriate for this purpose and should be
evaluated for skill in replicating meteorology and atmospheric chemical and physical
processes that result in secondary pollutant formation and deposition. Photochemical grid
models treat emissions, atmospheric chemistry, and physical processes, such as
deposition and transport. These types of models are appropriate for assessment of near-
field and regional scale reactive pollutant impacts from specific industrial sources (Baker
and Foley, 2011; Bergin et al., 2008; Kelly et al., 2015; Zhou et al., 2012), specific fire
events (Kansas Department of Health and Environment, 2012), or all sources (Chen et al.,
28
-------
2014; Russell, 2008; Tesche et al., 2006). Photochemical transport models have been
used extensively to support State Implementation Plans and explore relationships
between inputs, such as emissions and meteorology, and air quality impacts in the United
States and elsewhere (Cai et al., 2011; Hogrefe et al., 2011; Russell, 2008; Tesche et al.,
2006). Several state-of-the-science photochemical grid models could be used to estimate
fire impacts, including (but not limited to) the CAMx (www.camx.com), CMAQ
(ihttps://www.cmascenter.org/cmaq/), and WRF-CHEM (https://www2.acd.ucar.edu/wrf-
chem) models. These models have been used to estimate fire contributions to O3 in the
past (Fann et al., 2013; Jiang et al., 2012; Kansas Department of Health and
Environment, 2012; Kwok et al., 2015; U.S. Environmental Protection Agency, 2014).
Predictions of fire impacts on air quality are complex due to uncertainties in emissions,
height of emissions, plume temperature, and plume chemistry (including radiative
impacts on chemistry). However, with proper set-up, application, and evaluation, air
quality models can be used to indicate fire impacts on O3 concentrations. Model
evaluation of predictive skill on both event days, both for concentration and spatial extent
of impacts, and for typical days with little or no exceptional precursor levels, is key to
using the model results in a demonstration.
Where set up appropriately, photochemical grid models could be used with a variety of
approaches to estimate and assess the contribution of single sources to primary and
secondarily formed pollutants. These approaches generally fall into the category of
source sensitivity (how air quality changes due to changes in emissions) and source
apportionment (what air quality impacts are related to certain emissions). The simplest
source sensitivity approach (brute-force change to emissions, described as the difference
between a model simulation with all sources and a subsequent model simulation where
the wildfire(s) being quantified for impact are removed) is to simulate two sets of
conditions, one with all emissions and one with the source of interest (e.g., a fire event)
removed from the simulation (Cohan and Napelenok, 2011). The difference between
these simulations provides an estimate of the air quality change related to the change in
emissions from the fire event (Kansas Department of Health and Environment, 2012).
Another source sensitivity approach to differentiate the impacts of fire events on changes
in model predicted air quality is the direct decoupled method (DDM), which tracks the
sensitivity of an emissions source through all chemical and physical processes in the
modeling system (Dunker et al., 2002). Sensitivity coefficients relating source emissions
to air quality are estimated during the model simulation and output at the resolution of the
host model.
Some photochemical models have been instrumented with source apportionment, which
tracks emissions from specific sources through chemical transformation, transport, and
deposition processes to estimate a contribution to predicted air quality at downwind
receptors (Kwok et al., 2015; Kwok et al., 2013). Source apportionment has been used to
differentiate the contribution from specific sources on model predicted O3 and PM2.5
concentrations (Baker and Foley, 2011; Baker and Kelly, 2014). The DDM has also been
used to estimate O3 and PM2.5 impacts from specific sources (Baker and Kelly, 2014;
Bergin et al., 2008; Kelly et al., 2015), as well as the simpler brute-force sensitivity
approach (Baker and Kelly, 2014; Bergin et al., 2008; Kelly et al., 2015; Zhou et al.,
2012). Limited comparison of specific source impacts between models and approaches to
29
-------
differentiate single source impacts (Baker and Kelly, 2014; Kelly et al., 2015) show
generally similar downwind spatial gradients and impacts.
Air agencies should corroborate the modeled estimates of wildfire events with other
sources of information, such as satellite products and ground-based measurements and
not use the model as the sole evidence supporting the wildfire event contribution.
Significant variation in the modeled result from other information sources may indicate
that the photochemical model predictions are unreliable for demonstration purposes.
3.7 Example Conclusion Statement
Air agencies should provide the supporting evidence and analyses identified in Sections 3.1-3.6
of this guidance to document the clear causal relationship between the wildfire event and the
monitored O3 exceedance or violation. In summarizing the clear causal relationship section of
their demonstration, the air agency should conclude with this type of statement, which states how
the demonstration should meet the relevant statutory and regulatory criteria:
"On [day/time] an [event type] occurred that generated pollutant X or its precursors resulting in
elevated concentrations at [monitoring location(s)]. The monitored [pollutant] concentrations of
[ZZ] were [describe the comparison to historical concentrations including the percentile rank
over an annual (seasonal) basis]. Meteorological conditions were not consistent with historically
high concentrations, etc." and "The comparisons and analyses, provided in [section X] of this
demonstration support Agency A's position that the wildfire event affected air quality in such a
way that there exists a clear causal relationship between the specific event and the monitored
exceedance or violation on [dates/time of data requested for exclusion, or reference to summary
table in demonstration] and thus satisfies the clear causal relationship criterion."
4. Caused by Human Activity that is Unlikely to Recur at a Particular
Location or a Natural Event
4.1 Overview and Exceptional Events Rule Provisions
According to the CAA and the Exceptional Events Rule, an exceptional event must be "an event
caused by human activity that is unlikely to recur at a particular location or a natural event"
(emphasis added). The definition of wildfire in the Exceptional Events Rule is: ".. .is any fire
started by an unplanned ignition caused by lightning; volcanoes; other acts of nature;
unauthorized activity; or accidental, human-caused actions, or a prescribed fire that has
developed into a wildfire. A wildfire that predominantly occurs on wildland is a natural event."
Prescribed fires can be treated as wildfires for purposes of identifying the applicable
demonstration requirements under the Exceptional Events Rule if the conditions of a prescribed
fire develop in a way that the project no longer meets the resource objectives (e.g., if the fire has
escaped secure containment lines along all or part of its boundary).
Natural factors are principally responsible for wildfires on wildland (defined as "an area in which
development is essentially non-existent, except for roads, railroads, powerlines, and similar
30
-------
transportation facilities. Structures, if any, are widely scattered.").14 Land within national parks,
national forests, wilderness areas, state forests, state parks, and state wilderness areas are
generally considered wildland. Land outside cantonment areas on military bases may also be
considered wildland. Therefore, the EPA believes that treating all wildfires on wildland as
natural events is consistent with the CAA and the Exceptional Events Rule. Since wildfires on
wildland are treated as natural events, it is expected that minimal documentation will be required
to meet the human activity that is unlikely to recur at a particular location or a natural event
element.
The EPA will address wildfires on other lands on a case-by-case basis.
4.2 Examples of Supporting Documentation
To support this rule element, the air agency should clearly identify the origin and evolution of
the wildfire event and describe how the burned area is a wildland according to the Exceptional
Events Rule definition.
4.3 Example Conclusion Statement
In addition to the supporting information suggested in Section 4.2, the air agency should include
a conclusion statement similar to the language below to demonstrate that the wildfire on wildland
was a natural event.
"Based on the documentation provided in [Section X] of this submittal, the event qualifies as a
wildfire because [lightning, arson, accidental campfire escape, etc. \ caused the unplanned
wildfire event. The EPA generally considers the emissions of O3 precursors from wildfires on
wildland to meet the regulatory definition of a natural event at 40 CFR 50.1(k), defined as one
'in which human activity plays little or no direct causal role.' This wildfire event occurred on
wildland [as documented in X, or because...] and accordingly, [Air Agency Name] has shown
that the event is a natural event and may be considered for treatment as an exceptional event."
[Note: if a prescribed fire develops into a wildfire, then the air agency should supplement the
language above with additional detail as to the conditions, which led to this evolution. For
example, the air agency should indicate that the prescribed fire escaped secure containment lines
and required suppression along all or part of its boundary or that the prescribed fire escaped as a
result of quickly changing weather and no longer meets the resource objectives (e.g., smoke
impact, flame height)].
5. Not Reasonably Controllable or Preventable
5.1 Exceptional Events Rule Provisions
According to the CAA and the Exceptional Events Rule, an exceptional event must be "not
reasonably controllable or preventable." The preamble to the Exceptional Events Rule clarifies
that the EPA interprets this requirement to contain two factors: the event must be both not
reasonably controllable and not reasonably preventable at the time the event occurred. This
14 40 CFR 50. l(o).
31
-------
requirement applies to both natural events and events caused by human activities, however it is
presumptively assumed that wildfires on wildland will satisfy both factors of the "not reasonably
controllable or preventable" element unless evidence in the record clearly demonstrates
otherwise. If a prescribed fire has developed into a wildfire, some of the basic smoke
management practices that were planned for use for the prescribed fire may continue to be
reasonable to apply during the wildfire period. In showing that a prescribed fire has developed
into a wildfire, air agencies should include the following documentation in their demonstrations:
(1) news reports or notifications to the public characterizing the nature of the fire and (2) the
demonstration submitters' explanation of the origin and evolution of the fire.
5.2 Examples of Supporting Documentation
The Exceptional Events Rule accepts that wildfire events on wildland are not generally
reasonable to control or prevent. Therefore, a statement that the wildfire event was caused by one
of the causes identified in the definition of wildfire (such as lightning), and thus by the terms of
the Exceptional Events Rule, was not reasonably controllable or preventable, should satisfy this
rule element. A report based on information from other agencies or from news reports may
potentially be sufficient for this statement. The air agencies should work with their EPA
Regional offices to ensure that their statements about the causes of the wildfire events are
sufficient.
5.3 Example Conclusion Statement
In addition to the supporting information suggested in Section 5.2, the air agency should include
a conclusion statement similar to the language below to demonstrate why the wildfire event was
not reasonably controllable or preventable
"Based on the documentation provided in [Section X] of this submittal, [lightning] caused the
wildfire event on wildland. The [air agency] is not aware of any evidence clearly demonstrating
that prevention or control efforts beyond those actually made would have been reasonable.
Therefore, emissions from this wildfire were not reasonably controllable or preventable."
6. Public Comment
6.1 Exceptional Events Rule Provisions
According to the provisions in 40 CFR 50.14(c)(l)(i), air agencies must "notify the public
promptly whenever an event occurs or is reasonably anticipated to occur which may result in the
exceedance of an applicable air quality standard." In addition, according to 40 CFR
50.14(c)(3)(v), air agencies must "document [in their exceptional events demonstration] that the
[air agency] followed the public comment process and that the comment period was open for a
minimum of 30 days...." Further, air agencies must submit any received public comments to the
EPA and address in their submission those comments disputing or contradicting the factual
evidence in the demonstration. Air agencies with recurring events may also be subject to the
mitigation requirements at 40 CFR 51.930. Air agencies subject to these requirements have
additional obligations regarding public notification and engagement.
32
-------
6.2 Examples of Supporting Documentation
Air agencies should include in their exceptional events demonstration the details of the public
comment process including newspaper listings, Web site postings, and/or places (library, agency
office) where the hardcopy was available. As noted in Section 6.1, the agency should also
include comments received and the agency's responses to those comments.
6.3 Example Conclusion Statement
In addition to the supporting information suggested in Section 6.2, the air agency should include
a conclusion statement similar to the language below to demonstrate that it followed the public
comment process.
"The [air agency] posted notice of this exceptional events demonstration on [date posted] in the
following counties/locations: [list counties affected and locations posted], [Number] public
comments were received and have been included in [Section X] of the demonstration, along with
[air agency's] responses to these comments.
33
-------
Appendix Al. Example Conceptual Model/Event Summary
The following example of a conceptual model/event summary is based on a demonstration
prepared by the California Air Resources Board (CARB), prior to promulgation of the
Exceptional Events Rule revisions, to demonstrate wildfire-influence O3 exceedances. The EPA
has modified the narrative to provide a clear example of the suggested content of a conceptual
model.
A. Area Description
The Sacramento federal 1-hour ozone nonattainment area (Sacramento region) consists of
Sacramento County, Yolo County, the eastern portion of Solano County, the western portion of
Placer County, the western portion of El Dorado County, and the southern portion of Sutter
County (see Figure 1). The region covers over 5,600 square miles, and has a population of over
1.8 million.
The Sacramento region is located in the Central Valley of northern California. The Central
Valley is a 500-mile long northwest-southeast oriented valley that is composed of the
Sacramento Valley and the San Joaquin Valley air basins. Elevations in the Central Valley
extend from a few feet above sea level to almost 500 feet (see Figure 2). This long valley is
surrounded by the Coast Range Mountains on the west, the Cascade Range on the northeast, the
Sierra Nevada Mountains on the east, and the Tehachapi Mountains on the south. The San
Francisco Bay Area separates the Coast Range Mountains into northern and southern ranges. The
Coast Range Mountains generally form a topographic barrier to air flow between the Pacific
Ocean and the Central Valley, with occasional breaks created by low elevation passes and the
small gap between the northern and southern ranges in the San Francisco Bay area known as the
Carquinez Strait.
The Sacramento Valley's usual summer daytime circulation pattern is characterized by onshore
flow through the Carquinez Strait (which flows from the Bay Area to Sacramento and is known
as the sea breeze). Once through the Strait, the wind flow divides. A portion of the wind flow
turns south, blowing into the San Joaquin Valley, a portion continues eastward, across the
southern Sacramento Valley, and a portion turns north, blowing into the upper Sacramento
Valley. At night, the sea breeze weakens, and the wind direction in the Sacramento Valley
changes. Typical downslope flow, known as nocturnal drainage, brings air from the Coast Range
and Sierra Nevada Mountains into the Sacramento Valley. With the weakened sea breeze, an
eddy circulation pattern forms in the southwest portion of the Sacramento Valley, which serves
as a mechanism to recirculate and trap air within the region.
Because of its inland location, the climate of the Sacramento region is more extreme than that of
more coastal regions, such as the San Francisco Bay Area. The winters are generally cool and
wet, while the summers are hot and dry. Both seasons can experience periods of high pressure
and stagnation, which are conducive to pollutant buildup. These climate conditions result in
seasonal patterns where ozone concentrations are highest during the summer, while PM2.5
concentrations are highest during the winter. The lack of summertime precipitation, coupled with
the extent of forested regions surrounding the Central Valley, also creates conditions conducive
to wildfires during the summer months.
34
-------
B. Characteristics of Non-Event Ozone Formation
Anthropogenic emissions contributing to ozone formation in the Sacramento Region comprise
reactive organic gases (ROG) and oxides of nitrogen (NOx). The main sources of these
emissions include mobile sources (cars, trucks, locomotives, off-road equipment) along with
stationary and area sources that include industrial processes, consumer products, and pesticides.
Mobile source emissions dominate the anthropogenic emissions, accounting for more than 85
percent of the total NOx inventory. ROG and NOx emissions have decreased significantly over
the past several decades. This reduction directly translates into fewer days above the former
federal 1-hour ozone standard. In 1990, ROG and NOx precursor emissions were estimated at
262 and 242 tons per day (tpd), respectively. In 2008, these emissions had decreased almost 50
percent, to 136 tpd of ROG and 167 tpd of NOx. These significant improvements occurred
despite increases in population, vehicle activity, and economic development.
The ozone season in the Sacramento region occurs from May through October. Although
exceedances of the 1-hour federal ozone standard are infrequent, they are most likely to occur
under certain meteorological conditions. By evaluating high ozone concentrations and associated
meteorological conditions in the Sacramento region we developed several rules of thumb to
predict when ozone concentrations will be elevated in Sacramento County (see Appendix Y for
details). In general, the synoptic (large-scale) weather conditions leading to elevated ozone
concentrations occur in the Sacramento region when a ridge of high pressure is located over
California, causing the air to subside, or sink. As the air sinks, it warms, which forms a
temperature inversion that stabilizes and dries the atmosphere. This process limits the vertical
mixing of boundary layer air, which traps pollutants near the ground. The process also limits
cloud production, which increases ozone photochemistry. In addition, surface wind flow patterns
conducive to high ozone concentrations occur when the thermal surface low is over or just west
of Sacramento. This results in a sea breeze that weakens or occurs late in the day. This prevents
the dispersion of pollutants and leads to high ozone concentrations.
Nighttime drainage flows can bring biogenic emissions from the Coast Range and Sierra Nevada
Mountains into the Sacramento Valley. During daytime wind flow patterns, anthropogenic
precursor emissions in the Bay Area and Sacramento combine with biogenic emissions to
undergo photochemical reactions generating ozone. Due to the general daytime flow pattern
from west to east, as well as the time needed for photochemical reactions to occur, the highest
concentrations in the Sacramento region generally occur in the afternoon in the downwind,
eastern portion of the region, such as Folsom.
C. Wildfire Description
From June 20 to June 22, 2008, over 6000 lightning strikes from a series of thunderstorms
ignited numerous wildfires throughout northern and central California. At its peak, what became
known as the Northern California Lightning Siege (or the Lightning Complex Fires) comprised
thousands of wildfires in 26 counties and sent smoke throughout the western United States.
California firefighters were assisted in their efforts to control these blazes by units from
throughout the U.S., as well as Australia, Canada, Greece, Mexico, and New Zealand. With
thousands of individual fires (subsequently grouped into fire complexes) in 26 counties, the
summer of 2008 was one of the most severe wildfire seasons in California history. Most of these
35
-------
fires were not contained until late-July or early-August, with some continuing to burn through
October. Vast areas experienced smoke impacts, especially areas in northern California. Table 3
summarizes the number of wildfires and acreage burned by county from mid-June to mid-July
2008, in the counties surrounding Sacramento. Figure 3, provides a map of fire locations. A
detailed table listing the fires, distance from Folsom, and acreage burned is included in Appendix
A. A summary report on these wildfires was prepared by an interagency team of investigators at
the request of California Department of Forestry and Fire Protection (CAL Fire), the U.S. Forest
Service, Office of Emergency Services, and the National Park Service.15 The following is an
excerpt from that report, "The 2008 Fire Siege": On June 20th and 21st a series of severe, dry
thunderstorms carpeted the state from Big Sur to Yreka with more than 5,000 lightning strikes,
and igniting over 2,000fires. During the following months, thirteen firefighters were killed and
many others were injured on fires in this siege. Over 350 structures were destroyed and
hundreds of millions of dollars ofproperty and natural resources were damaged. Thousands of
people were evacuated and smoke adversely effected air quality over much of the state for weeks.
Communications, power delivery, and transportation systems were disrupted. Despite the
intensive firefighting effort, some fires in remote areas continued to burn throughout the
summer. By fall, over J, 200,000 acres had burned.
Air quality in northern California deteriorated because of the smoke. From June 23 through
much of July, the Sacramento region was covered in a thick blanket of smoke. Many of the air
monitors recorded extremely high ozone concentrations, along with hazardous concentrations of
particulate matter. The hazardous air quality levels prompted air pollution control and air quality
management districts in the Sacramento region to issue air quality advisories and warnings. The
wildfires and smoke spread throughout the Sacramento region and were widely recognized by
residents in the region and the public media. Figures 4, 5, and 6 provide satellite maps
illustrating the extent of the smoke impacts on June 23, June 27, and July 10, 2008.
2. Conceptual Model of Ozone Formation from 2008 Wildfires
Substantial amounts of NOx and VOCs were generated from the 2008 wildfires during late June
and early July across a broad area surrounding the Sacramento Valley, corresponding to the 1-
hour ozone exceedances at Folsom on June 23, June 27, and July 10, 2008. Surface wind flow
conditions on these days were typical for the summertime, including nighttime drainage flow
from the Coast Range and Sierra Nevada Mountains, coupled with an eddy circulation in the
southern Sacramento Valley, followed by the daytime sea breeze. These wind flow patterns
transported, and subsequently trapped within the Sacramento region, wildfire precursor
emissions coming from multiple upwind locations. In addition to surface transport, due to the
buoyancy of fire plumes, substantial amounts of precursors were emitted aloft by the wildfires.
An increase in the mixed layer during the morning and early afternoon on each day allowed
additional wildfire precursors aloft to reach the surface.
Under typical daytime photochemistry, the increased levels of wildfire-related precursor
emissions in the Sacramento region resulted in enhanced levels of ozone throughout the region,
including Folsom. Although these surface windflow patterns would also have transported
15 California Department of Forestry and Fire Protection, "2008 Fire Siege" (retrieved April I, 2011) available at
http://www.fire.ca.gov/fire_protection/downloads/siege/2008/2008FireSiege^full-book_r6.pdf (Multiagency Fire
Investigation Report).
36
-------
anthropogenic emissions to Folsom, the meteorological conditions that existed on the three
exceedance days were not sufficient to have caused a 1-hour ozone exceedance without the
added burden of the additional wildfire-related precursor emissions. In addition, given the
lengthy duration of the fires, by June 27 and July 10 there were also substantial amounts of
wildfire-related ozone carried over from the day before the exceedance, further increasing ozone
concentrations.
Although, NO from fires can result in ozone titration very close to the source of a fire, Folsom
was sufficiently far enough downwind that a reduction in ozone concentrations due to this
phenomena was unlikely. In addition, while the increased smoke from the fires may have
reduced the amount of solar insolation, thereby potentially reducing photochemical activity, this
was compensated for by the substantially increased levels of ozone precursors generated by the
fires, resulting in a net ozone enhancement.
During this period, there were 15 monitoring sites operating in the Sacramento nonattainment
area, as shown in Figure 7, below. Ozone was dramatically elevated throughout the
nonattainment area and much of northern and central California during the fire period. In the
Sacramento nonattainment area, five monitoring sites recorded ozone concentrations above the
1-hour standard. More detailed information about the exceedances at these sites is shown in
Table 4. Section 3 provides a more detailed discussion of the day-specific meteorological
conditions that existed on each of the three 1-hour ozone exceedance days included in this
request to support the clear causal relationship between the wildfires and the ozone exceedances.
In addition, Section 4 provides information to demonstrate that the exceedances of the 1-hour
ozone NAAQS at Folsom on each of these days were directly due to the impacts of the wildfire
emissions.
The following figures and tables were included:
* Figure 1. Map of Sacramento Metropolitan non-Attainment Area
* Figure 2. Topographic map of Northern California
Table 3. Summary, by county, of wildfires that contributed to the exceedance
* Figure 3. Map of wildfires, colored and sized by geographic extent
Figure 4. MODIS image of June 23
Figure 5. MODIS image of June 27
Figure 6. MODIS image of July 10
* Figure 7. Map of air quality monitors in the Sacramento area
Table 4. 2008 Sacramento 1-hour ozone non-attainment days and concentrations
*These maps could be combined into one.
37
-------
Appendix A2. Relating Fire Emissions and Downwind Impacts
Summary
To understand general relationships between the magnitude of fire emissions and potential
downwind O3 impacts, the EPA conducted an assessment of fire case studies. These case studies
were drawn from peer-reviewed literature, EPA-approved exceptional events demonstrations for
fires that influenced O3 concentrations, and EPA-performed photochemical modeling studies.
The dependence of O3 impacts on fire emissions and distance from the fire across these case
studies has been compared to determine fire characteristics that are expected to lead to
meaningful O3 impacts.
Background
Fires can impact O3 concentrations by emitting known O3 precursors including NOx and VOCs.
These precursor emissions can generate O3 within the fire plume or can mix with emissions from
other sources to generate O3 (Jaffe and Wigder, 2012). Also, in some situations, including near
fires, reduced O3 concentrations have been observed and attributed to O3 titration by enhanced
NO concentrations and reduced solar radiation available to drive photochemical reactions (Jaffe
et al., 2008; Yokelson et al, 2003). The magnitude and ratios of emissions from fires vary greatly
depending on fire size, fuel characteristics, and meteorological conditions (Akagi et al., 2012).
As a result of variable emissions, radiative impacts, and non-linear O3 production chemistry, the
O3 production from fires is very complex, highly variable, and often difficult to predict (Jaffe
and Wigder, 2012). Understanding and predicting O3 formation from fires remains an active area
of research.
Despite the complexities in predicting O3 formation from fire emissions, several studies have
found enhancements in O3 concentrations attributable to fire impacts. For example, Pfister et al.
analyzed surface O3 data during a high fire year in California (2007) with modeled fire impacts
and found 8-hour O3 concentrations were approximately 10 ppb higher when the modeled
impacts were high (Pfister et al., 2008). Jaffe et al. analyzed three specific fire periods in the
western US during 2008 and 2012, and compared surface O3 concentrations with two different
modeled estimates of fire contributions to O3 concentrations to find enhancements in O3 when
fire impacts were predicted to be high (Jaffe et al., 2013).
Previously Approved Fire-Influenced O3 Exceptional Events Demonstrations
Between 2010 and August 2015, the EPA approved two exceptional events demonstrations that
linked monitored O3 exceedances to fire impacts. The first was approved in 2011. In this case,
the EPA concurred on three exceedances of the 1-hour O3 NAAQS near Sacramento, California
in 2008 due to a series of lightning-initiated wildfires throughout northern California. The second
demonstration for fire impact on O3 was approved in 2012. In this case, the EPA concurred with
the exclusion of eight MDA8 exceedances during April 2011 in Kansas due to impacts from
prescribed fires and wildfires. Additional information regarding these submissions is provided in
Section 3.5 of this document. Both of these demonstrations are available at
http://www2.epa.gov/air-quality-analysis/exceptional-events-submissions-table.
38
-------
Assessments of Q/D Relationships from Previously Approved Demonstrations and Relevant
Peer-Reviewed Literature
At least one air quality related program (i.e., determining impacts at Class I areas) uses an
emissions divided by distance (Q/D) key factor as a screening tool. The EPA believes that it is
appropriate to use a similar approach, along with additional information about the fire event, to
determine whether a simpler and less resource-consuming exceptional events demonstration
provides sufficient evidence to satisfy the clear causal relationship criteria of the Exceptional
Events Rule for fire O3 demonstrations.
To determine whether a relationship existed between approved demonstrations and Q/D values,
the EPA estimated Q/D values from previously approved, fire-related O3 exceptional events
demonstrations. The EPA also included in this comparison, the results from one peer-reviewed
publication, which included sufficient detail for a similar analysis (Jaffe et al., 2013). The EPA
used daily fire emissions estimates from the 2008 and 2011 NEIs (https://www.epa.gov/air-
emissions-inventories/national-emissions-inventory) to estimate Q from fires impacting the O3
monitors. For consistency, the EPA also used NEI-based estimates for the Jaffe et al. fires. In
determining the appropriate emissions to use in this assessment, the EPA summed NOx and
rVOC because both are precursors for O3 formation. The NEI reports total organic gas (TOG) so
the reactive fraction of these emissions (rVOC) was estimated by applying the fraction of
reactive gas to total organic gas based on speciation profiles for fires provided by the SPECIATE
database. A factor of 0.6 was selected based on the SPECIATE database profile used by CMAQ
for fires (speciation profile number 5560).16
Fire events included in the estimated Q values were based on the sum of emissions from only
some of the events listed by the relevant air agencies in the demonstrations because the
demonstrations included fires that may not directly impact the monitor. The CARB exceptional
events demonstration identified all wildfires burning in California during the time period of the
O3 exceedances, and a subset of those (within state of CA, with latitude north of 37N (-north of
Santa Cruz) and longitude west of-119W (-west of Mono Lake) were used. The Jaffe et al.,
article assessed the impact of the 2008 Northern California fires in Reno, NV (versus at
California monitors). The same fire subset was used for the Jaffe et al. analysis as for the CARB
demonstration. For the Kansas Department of Health and Environment demonstration, the EPA
included all fire events labeled as "Flint Hills" in the NEI emissions file. Emissions totals within
these bounds on the day of the O3 exceedances were used to calculate emissions totals, Q. The
uncertainty in Q was taken to be approximately ±25% and was taken from the differences
between the NOx estimates from the NEI and the NOx estimates from the Fire Inventory from
NCAR (FINN) emissions inventories of all fires (Wiedinmyer et al., 2011).
O3 impacts were determined differently by the CARB demonstration, the KDHE demonstration,
and the Jaffe et al. article. The CARB demonstration used a statistical regression model to
estimate fire contributions to O3 concentrations. The KDHE demonstration used both a matching
day analysis and photochemical modeling to estimate O3 impacts. The Jaffe et al. paper used
both photochemical and statistical residual modeling to estimate O3 impacts.
16 SPECIATE is the EPA's repository of volatile organic gas and particulate matter speciation profiles of air
pollution sources. Available at http://www3.epa.gov/ttnchiel/software/speciate/.
39
-------
A summary of the fire impacts on O3 compared with Q/D for the approved demonstrations and
the Jaffe et al. article is shown in Figure A2-1. Distance (km) between the fire and the O3
monitors was calculated based on an average fire location determined with an emissions-
weighted fire center. The uncertainty range in D was determined by using the maximum distance
between the monitor and a fire event (within the subset given above) on the day of the
exceedance. The range shown for the CARB O3 impacts reflects the uncertainty analysis
included in the demonstration. The ranges included for O3 impacts estimated by the KDHE
demonstration and the Jaffe et al. paper represent the range in estimates of O3 impacts
determined by the two different methods used in each case.
Modeling Studies of Wildfire Impacts on O3
Some uncertainty exists in the magnitude of emissions estimates, VOC and PM2.5 speciation of
emissions, downwind transport, chemical reactions in fire plumes, and representation of
important physical processes like reduced photolysis due to smoke. However, the emissions used
as input to air quality models can be paired with estimated downwind O3 contribution to assess
screening level relationships between precursor emissions and downwind impact. Constructing
these relationships is useful for planning purposes and making preliminary determinations about
whether fires with emissions of a certain amount and distance away may impact a monitor and
warrant further investigation for fire contribution using additional corroborative information.
For the modeling studies of wildfire impacts on O3, the entire year of 2011 was applied using the
CMAQ version 5.0.2 model (www.cmascenter.org). Meteorological input was generated using
version 3.4.1 of the WRF prognostic meteorological model (Skamarock et al., 2008). Both
modeling systems were applied using the same grid projection and model domain covering the
continental United States with 12 km sized grid cells. Contributions from four specific fire
events were tracked using source apportionment approaches. Source apportionment tracks
primarily emitted and precursor emissions from specific fire events through the model's
chemical and physical processes to track contribution to primary and secondarily formed
pollutants. The integrated source apportionment approach has been implemented in CMAQ for
O3 (Kwok et al., 2015) and PM2.5 (Kwok et al., 2013) and was used in this analysis to track the
contribution from each fire event. CMAQ with source apportionment was applied for four
different multi-day fire events in 2011: Wallow, Waterhole, Big Hill, and Flint Hills. The days
included in each model simulation for each fire event and the daily total fire event emission
estimates are shown in Table A2-1. Emissions-weighted fire event locations are shown in Table
A2-2. All the emissions from each multi-day fire were tracked as a single source, so it is not
possible to determine from the results how a single day of a particular multi-day fire event
emissions affects a single day of O3 concentrations. For example, O3 effects on the third day of a
fire may be a contribution of direct effects from a same day plume and effects from recirculated
VOC, NOx and O3 from earlier days.
Wildfire and prescribed fire emissions were included when and where these emissions occur
within the modeling domain. These emissions are based on the latest version of the
SMARTFIRE system (http://www.airfire.org/smartfire/). Detailed information about how the
EPA develops wildland fire inventories can be found in the 2011 NEI Technical Support
Document (U.S. Environmental Protection Agency, 2014). This approach relies on a
combination of satellite detection of fires merged with on-the-ground observational data where
40
-------
available. Ground-based observations and local fuel information are used whenever possible as
these factors can have a large impact on the emissions. CMAQ currently uses one single
speciation profile (5560; Table A2-3) to speciate TOG fire emissions into specific compounds
(e.g., toluene, benzene, etc.) that are subsequently used in the gas phase chemical mechanism
within CMAQ. Similarly, a single profile is used to map total PM2.5 emissions from fires to
specific compounds (e.g., elemental carbon, organic carbon, etc.). Daily total emissions for each
fire event tracked for O3 contribution are shown in Table A2-1. The EPA also conducted a
sensitivity analysis including reducing each fire's emissions to half the original emissions.
Figure A2-2 shows maximum hourly (across all modeled days of the event) source
apportionment based O3 impacts from the fire events tracked in this assessment. Fire NOx
emissions tend to contribute more to O3 formation than fire VOC emissions, on a per fire
comparison basis, for the fire events in the western United States where biogenic VOC is often
abundant (especially near these particular fire events). The stronger effect from NOx emissions
compared to VOC emissions on a per ton basis (not shown) is even more pronounced, given the
tonnage values in Table A2-1. The NOx contribution could be favored in the model if O3
formation was NOx limited even when the contributing VOC was also from the same fire event.
The fire event modeled in Kansas illustrates that VOC emissions from fires can also be
important, especially when other VOC sources are less abundant.
Figure A2-3 depicts downwind O3 and CO impacts. This figure also shows Q/D for these events
and forward HYSPLIT trajectory endpoints (from each day included in Table A2-1) from release
out to 48 hours. This figure clearly shows the importance of pairing information about the
trajectory of fire emissions in combination with simple metrics of impact such as Q/D. The
Wallow fire event had the most consistent trajectories across the days of the event. For the other
fire events, wind directions on different days differed considerably.
Maximum hourly fire impacts on O3 (that were greater than 1.0 ppb) and the corresponding
distance of the grid cell where the maximum impact occurred from the emissions-weighted
average location of the fire event are shown in Figure A2-4. The colored box represents the 25th-
75th percentiles of the distribution of O3 impacts larger than 1.0 ppb, and the solid line within the
colored box indicates the median of the distribution. Impacts only up to 1000 km (for Wallow,
Flint Hills, and Waterhole) and 550 km (for Waterhole and Big Hill) are shown since the
magnitude of the O3 impacts decrease at increased distances. The maximum O3 impacts tend to
be highest in closer proximity to the event and decrease as distance from the event increases
(Figure A2-4). When these impacts are normalized by the sum of NOx+rVOC emissions for the
event day with the highest emissions during the period modeled (Figure A2-5), the magnitude of
O3 impacts varies over the range of Q/D values, with larger O3 impacts occurring at higher Q/D
values. The truncation of distances used in Figure A2-4 leads to the absence of O3 impacts at low
Q/D values (e.g., -20 for the Wallow Fire) in Figure A2-5.
The results shown in Figure A2-5 help determine the appropriateness of using the Q/D approach
as one key factor in a simpler and less resource-consuming exceptional events demonstrations
for certain fire events (i.e., Tier 2). In the figure for each modeled fire event, modeled maximum
O3 impacts are shown for the first two days, except for the Big Hill Fire where the entire, three
day event is shown. Each data point represents the maximum, hourly O3 impact (over 1 ppb) that
occurred in a grid cell during the first 48 hours of the event. In general, higher O3 impacts are
41
-------
predicted at larger Q/D values. Comparisons across the four fire events modeled here indicate
more and larger O3 impacts at high Q/D values from the fires with the highest emissions
(Wallow and Flint Hills) versus the smaller, lower emissions fires (Big Hill and Waterhole).
When Q/D values from a fire event are paired with both elevated monitored O3 concentrations
{i.e., Tier 2 key factor #2) and evidence (e.g., HYSPLIT trajectory or other analyses identified in
Sections 3.5.2 and 3.5.3) linking the affected monitor to the location(s) of the subject fire(s), the
EPA believes that the Q/D relationship can be used to indicate when large O3 impacts are
expected to occur.
To examine the utility of the Q/D metric, Q/D was calculated for all fires in the National
Emission Inventory for the years 2008 through 2013 to provide an aggregate context for areas
and times where fires may be large contributors to elevated air quality. Figures A2-6 through A2-
8 show the count of days with NOx+rVOC Q/D values greater than 50, 100, and 200 for 2008
through 2013. These figures illustrate how the fire events modeled for this assessment from 2011
compare to other fires that year and to fires from other recent years where data are available.
These results can be used to investigate how many days and areas would meet various thresholds
for the Q/D key factor.
Conclusions
The fire event impacts estimated with the photochemical model CMAQ suggest both NOx and
VOC emissions from fire events can lead to downwind O3 formation and the importance of these
precursors varies among fires, most likely due to the surrounding environment's availability of
NOx and VOC emissions. Since information about the surrounding environment may not always
be practically available, the approach for estimating fire impacts should be inclusive of both NOx
and reactive VOC emissions.
The downwind O3 contribution from these fire events is greatest in the proximity of the fire and
tends to gradually decrease as distance from the source increases. The spatial plots of downwind
O3 impacts show that the impacts occur in the direction of air mass movement from the fire event
to specific places downwind. As indicated above, tiering approaches that do not explicitly
account for pollutant transport {e.g., Q/D) should be accompanied with information about
pollutant transport from another source such as HYSPLIT trajectories to better spatially represent
the downwind impacts.
42
-------
Acknowledgements
This Appendix was in part supported by contribution from Venkatesh Rao, Alison Eyth, Alexis
Zubrow, Allan Beidler, James Beidler, Chris Allen, Lara Reynolds, and Chris Misenis.
References for Appendix A2
Akagi, S., Craven, J., Taylor, J., McMeeking, G., Yokelson, R., Burling, I., Urbanski, S., Wold, C.,
Seinfeld, J., Coe, H., 2012. Evolution of trace gases and particles emitted by a chaparral fire in California.
Atmospheric Chemistry and Physics 12, 1397-1421.
Jaffe, D.A., Wigder, N., Downey, N., Pfister, G., Boynard, A., Reid, S.B., 2013. Impact of wildfires on
ozone exceptional events in the western US. Environmental science & technology 47, 11065-11072.
Jaffe, D.A., Wigder, N.L., 2012. Ozone production from wildfires: A critical review. Atmospheric
Environment 51, 1-10.
Kwok, R., Baker, K., Napelenok, S., Tonnesen, G., 2015. Photochemical grid model implementation of
VOC, NO x, and O 3 source apportionment. Geoscientific Model Development 8, 99-114.
Kwok, R., Napelenok, S., Baker, K., 2013. Implementation and evaluation of PM25 source contribution
analysis in a photochemical model. Atmospheric Environment 80, 398-407.
Pfister, G., Wiedinmyer, C., Emmons, L., 2008. Impacts of the fall 2007 California wildfires on surface
ozone: Integrating local observations with global model simulations. Geophysical Research Letters 35.
Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Barker, D.M., Duda, M.G., Huang, X., Wang, W.,
Powers, J.G., 2008. A description of the Advanced Reserch WRF version 3. NCAR Technical Note
NCAR/TN-475+STR.
U.S. Environmental Protection Agency, 2014. 2011 National Emissions Inventory, version 1 Technical
Support Document. http://www3. epa.gov/ttn/chief/net/201 lnei/2011 _nei Jsdv 1 _draft2june2014.pdf.
Wiedinmyer, C., Akagi, S., Yokelson, R.J., Emmons, L., Al-Saadi, J., Orlando, J., Soja, A., 2011. The
Fire INventory from NCAR (FINN): A high resolution global model to estimate the emissions from open
burning. Geoscientific Model Development 4, 625.
43
-------
Table A2-1. Daily emissions for each tracked fire event in 2011. rVOC is the sum of all
VOC excluding methane and non-reactive species.
Fire Event
Month-Day
CO
NOX
rVOC
NOX+rVOC
Waterhole
822
9,441
96
1,331
1,427
Waterhole
823
17,652
171
2,487
2,658
Waterhole
824
38,086
408
5,373
5,780
Waterhole
825
637
6
90
96
Waterhole
826
34
1
5
6
Big Hill
814
243
7
35
42
Big Hill
815
3,248
92
468
560
Big Hill
816
189
5
27
33
Flint Hills
401
30,675
867
4,417
5,285
Flint Hills
402
51,555
1,413
7,417
8,830
Flint Hills
403
14,526
383
2,087
2,470
Flint Hills
404
3,744
106
539
646
Flint Hills
405
20,233
564
2,912
3,477
Flint Hills
406
78,622
2,218
11,321
13,539
Flint Hills
407
9,719
263
1,398
1,661
Flint Hills
408
59,020
1,584
8,485
10,070
Flint Hills
409
60,294
1,656
8,675
10,331
Flint Hills
410
9,194
257
1,324
1,580
Flint Hills
411
57,428
1,540
8,256
9,796
Flint Hills
412
105,636
2,950
15,206
18,157
Flint Hills
413
60,484
1,670
8,704
10,373
Flint Hills
414
7,874
215
1,133
1,348
Flint Hills
415
95
3
14
16
Wallow
604
115,438
1,516
16,331
17,847
Wallow
605
49,951
697
7,074
7,771
Wallow
606
113,160
1,509
16,013
17,522
Wallow
607
53,030
705
7,504
8,209
Wallow
608
131,675
1,774
18,636
20,409
Wallow
609
59,155
839
8,379
9,218
Wallow
610
52,127
736
7,383
8,119
44
-------
Table A2-2. Emissions weighted fire event locations.
Fire Event
Waterhole
Big Hill
Flint Hills
Wallow
Latitude
45.6141
42.5673
37.9466
33.8174
Longitude
-106.7889
-115.8093
-96.3543
-109.3272
Table A2-3. Speciation profile (5560) used to map TOG emissions to specific lumped
compound groups for photochemical model application.
Profile Inventory Model Fraction
5560
TOG
UNR
0.22
5560
TOG
PAR
0.18
5560
TOG
CH4
0.18
5560
TOG
FORM
0.08
5560
TOG
MEOH
0.08
5560
TOG
OLE
0.07
5560
TOG
ALD2
0.05
5560
TOG
ETH
0.04
5560
TOG
TOL
0.03
5560
TOG
ALDX
0.02
5560
TOG
ETHA
0.02
5560
TOG
BENZENE
0.02
5560
TOG
TERP
0.01
5560
TOG
XYL
0.01
5560
TOG
IOLE
0.00
5560
TOG
ISOP
0.00
5560
TOG
ETOH
0.00
45
-------
Figure A2-1. Summary of O3 impacts versus Q/D relationships for approved
demonstrations (CARB_Folsom_2008 and KDHE_FlintHills_2011) and impacts reported
by Jaffe and Wigder (2012). No results from the EPA's photochemical modeling are shown
in this Figure.
90
CARB Foisorn 2008 KDHE FlintHills 2011 Jaffe Reno 2008
80
70
^60 !
| 50 !
m ;
° !
2 40 : h
-------
Figure A2-2. Event maximum 1-hour O3 (ppb) impacts (left panels). The percent
contribution from fire event NOx emissions to event maximum 1-hour O3 impacts shown at
right. The percent contribution plots show that both NOx and VOC emissions from fires
can contribute to downwind O3 formation.
Wallow - Ozone (ppb)
S
o
s
o
s
30
25
0
20
15
10
s -
1
5
-
0
0
8 -
Wallow - Ozone from NOX {%)
1
-2JOO -1EHW
1C0
- SO
60
40
20
~
-2HD -1DOO
47
-------
Figure A2-3. Event maximum 1-hour CO (left panels), O3 (second to left panels), Q/D
(second to right panels), and forward trajectories (right panels) shown for multiple fire
events. Q/D is based on daily maximum NOX+rVOC emissions from the fire event during
the period modeled. Forward trajectories are shaded by hours from release with warm
colors (red and orange) representing hours during the first day and cooler colors the 2nd
day (24 to 48 hours) from release.
Wallow - CO ISAM {ppb)
T
Wallow - Ozone ISAM (ppb)
Wallow - Q/D
I I I I I
-1500 -5D0 0 500
1 1 1 1 1
15 DO -SCO 0 500
Wallow - Forward Trajectories
l 1 1 i 1 r
-150Q -500 D 5G0
Flint Hills - CO ISAM (ppb)
1
Flint Hills - Ozone ISAM (ppb)
Flint Hills - Q/D
1 1 1
-1500 -5D0 0 500
V 1 1 1
150Q -5D0 0 500
Flint Hills - Forward Trajectories
1 1 1 1
-1503 -500 0 500
-150Q -500 0 50D
Watertiole - CO ISAM {ppb)
Watertiole - Ozone ISAM (ppb)
Waterttole - Q/D
Watertvole - Forward Trajectories
-1500 -5DO 0 500
1 1 1 1 1
-1500 -500 0 500
t i i i i r
1500 -500 0 500
Big Hill - CO ISAM (ppb)
Big Hill - Ozone ISAM (ppb)
Big Hill - QJD
Big Hill - Forward Trajectories
0
p 1QD 1
s r 1
- eo g
«o
- 60
0" \ L 1
- 40 0
° /
- 20 0
8 A /
8
\ / 1
- 0 1
li T 1- 1 1
-1300 -1400 -1000
0
r- 200
0
8
15
- 150
500
10
°"xZZ/^
- 100
0
5
- 50
0
£
0
L D
1
-1300 -14DQ -100CI
-1803 -14D3 -1D0Q
1 r"i r
-1B0Q -1400 -1DO0
48
-------
Figure A2-4. Distribution of hourly O3 impacts from fire events by distance from the
location of the fire event.
X3
a
Hourly maximum 03 impacts - Wallow fire
Hourly maximum 03 impacts - Half Wallow fire
n1111111111111i11111r
D 1Q0 2DC 3D0 400 500 600 700 800 S00 1000
Distance from the source (km)
Hourly maximum 03 impacts - Flint Hills fire
niiiiiiiiiiiiiiiiiiir
0 100 200 300 400 500 600 700 800 800 1000
Distance from the source (km)
Hourly maximum 03 impacts - Half Flint Hills fire
~iiiiiiiiiiiiiiiiiiiir
0 100 200 300 400 500 600 700 800 000 1000
Distance from the source (km)
tii\iiiiiiiiiiiiiiiir
D 100 2DO 300 400 500 600 700 800 600 1000
Distance from the source (km)
Hourly maximum 03 impacts - Watertiole fire
Hourly maximum 03 impacts - Half Waterhole fire
hiiiiiiiiiiiiiiir
D 100 200 300 400 500 600 700 800 600 1000
Distance from the source (km)
Hourly maximum 03 impacts - Big Hill fire
~i 1 i 1 i 1 1 1 1 r
50 100 150 200 250 300 350 400 450 500 550
Distance from the source (km)
X>
a
-O
a
"i r
0 50 1 00 150 2DO 250 300 35D 400 450 500 550
Distance from the source (km)
Hourly maximum 03 impacts - Half Big Hill fire
n 1 1 1 1 1 1 1 1 r
50 100 150 200 250 300 350 400 500 550
Distance from the source (km)
49
-------
Figure A2-5. Hourly maximum O3 impacts from the first two days of each fire event (Table
A2-1) shown by Q/D. O3 impacts only up to 1000 km from the fire have been included in
this analysis.
Hourly maximum 03 impacts - Wallow fire
200 400 600
Q/D (tons/dayvkm)
Hourly maximum 03 impacts - Flint Hills lire
R -
200 400 600
Q/D (tons/dayvkm)
Hourly maximum 03 impacts - Waterhole fire
Hourly maximum OS impacts - Half Wallow fire
200 400 000 (
Q/D (tons/dayvkm)
Hourly maximum 03 impacts - Half Flint Hills fire
200 400 000 e
Q/D (tons/dayvkm)
Hourly maximum 03 impacts - Half Waterhole fire
Q/D (tons/dayvkm)
Hourly maximum 03 impacts - Big Hill fire
Q'D (tons/dayvkm)
Hourly maximum 03 impacts - Half Big Hill fire
Q'D (tons/dayvkm)
Q'D (tons/dayvkm)
50
-------
Figure A2-6. Count of days with NOx+rVOC Q/D > 50 for 2008 through 2013. Note scale
has been capped at 10 to more easily distinguish the values below 10. Red may actually
indicate 10 or greater than 10.
2008 - Annual Count of Q"D > 50
2009 - Annual Count of Q'D > 50
2010 - AnnuaJ Count of Q' D > 50
21011 - Annual Count of Q'D > 50
-1000
I
1000
-1000
-1000
2012 - Annual Count of Q' D > 50
2013 - Annual Count of Q'D > 50
-1ooo
51
-------
Figure A2-7. Count of days with NOx+rVOC Q/D > 100 for 2008 through 2013. Note scale
has been capped at 10 to more easily distinguish the values below 10. Red may actually
indicate 10 or greater than 10.
52
-------
Figure A2-8. Count of days with NOx+rVOC Q/D > 200 for 2008 through 2013. Note scale
has been capped at 10 to more easily distinguish the values below 10. Red may actually
indicate 10 or greater than 10.
2008 - Annual Count of Q'D > 200
i n 1 1 1
-2000 -1DQD Q 1000 2000
20O9 - Annual Count of Q.'D > 200
i 1 1 1 r
-2Q00 10Q0 D 1000 2D00
i:
- e
I"
I " 2
o
i:
- 5
I"
l-*- 0
i::
- D
::
53
-------
Appendix A3. Interpreting HYSPLIT Results
A HYSPLIT backward trajectory, the most common trajectory used in assessments associated
with determining source areas, is usually depicted on a standard map as a single line extending in
two dimensional (x,y) space from a starting point, regressing backward in time as the line
extends from the starting point. An individual trajectory can have only one starting height;
HYSPLIT can plot trajectories of different starting heights at the same latitude/longitude starting
point on the same map, automatically using different colors for the different starting heights.
HYSPLIT will also include a vertical plot of the trajectories in time, with colors corresponding to
the same trajectory in the (x,y) plot. Diurnal mixing height data on flagged days should be
considered in setting up the starting point matrix. Caution is needed, because this display can be
easily misinterpreted as having finer accuracy than the underlying model and data.
It is important to observe the overall size of the plot, its width and length in kilometers, while
considering the size of an individual grid cell in the input meteorological data set. These input
grid cells are usually 40 km in width and length, so the total area of a trajectory plot may
sometimes represent only a few meteorological grid cells. It is also important to understand the
trajectory line itself. The line thickness is predetermined as a user option, so it does not imply
coverage other than to represent the centerline of an air parcel's motion calculated to arrive at the
starting location at the starting time. The range of the width and the height of plume can vary
significantly and are not normally part of the information output but clearly can lead to
uncertainty in source strength at the centerline. Uncertainties are clearly present in these results,
and these uncertainties can be thought to be a range on either side of the center line in which the
air parcel may be found. Further back in time along the trajectory path, that range may be
assumed to increase. In other words, one should avoid concluding a region is not along a
trajectory's path if that trajectory missed the region by a relatively small distance.
Operating HYSPLIT
Detailed information for downloading, installing, and operating HYSPLIT can be found at these
websites:
http://ready, arl. noaa.gov/HYSPLIT.php
http://www.arl.noaa.gov/documents/reports/hysplit user guide.pdf
http://www. arl. noaa.gov/documents/reports/arl-224.pdf
HYSPLIT's many setup options allow great flexibility and versatility. However, careful selection
and recording of these options is recommended to provide reviewers the ability to reproduce the
model results. The following paragraphs describe the options that should be recorded, at a
minimum, to reproduce a HYSPLIT model run.
Backward Versus Forward Trajectories. Forward and backward HYSPLIT trajectories use the
same scientific treatment and processing. These trajectories only differ in the location of the
discrete point of origin (forward) or destination (backward). For analyses to assess the potential
impact of a source area such as a wildfire on a discrete point of destination such as an air quality
monitor, a backward trajectory is more easily interpretable.
54
-------
Model Version. If the HYSPLIT trajectory is produced via the NOAA Air Resources Laboratory
(ARL) website (http://ready.arl.noaa.gov/HYSPLIT traj.php), note the "Modified:" date in the
lower-left corner of the webpage, as well as the date the trajectory was produced. If the trajectory
is produced using a stand-alone version of HYSPLIT, note the release date, which will be
displayed after exiting the main GUI screen.
Basic Trajectory Information. Note the starting time (YY MM DD HR), the duration of the
trajectory in hours, and whether the trajectory is backward or forward. Note the latitude and
longitude, as well as the starting height, for each starting location. Starting height is given by
default in meters above ground level (AGL) unless another option is selected. Starting heights
are typically no less than 100 meters AGL to avoid direct interference of terrain, and are
typically no greater than 1500 meters AGL to confine the air parcel within the mixed layer.
Some trajectories can escape the mixed layer, and this result would be considered in the
interpretation.
Starting height and starting location will identify the three-dimensional location of the
trajectory's latest endpoint in time if a backward trajectory is selected (i.e., the start of a
trajectory going backward in time).
Input Meteorological Data Set. Note the input meteorological data set used in the HYSPLIT
model run. The original file name provides sufficient information to identify the data set.
Meteorological data fields to run the model are already available for access through the
HYSPLIT menu system, or by direct FTP from ARL. The ARL web server contains several
meteorological model data sets already converted into a HYSPLIT compatible format in the
public directories. Direct access via FTP to these data files is built into HYSPLIT's graphical
user interface. The data files are automatically updated on the server with each new forecast
cycle. Only an email address is required for the password to access the server. The ARL analysis
data archive consists of output from the Global Data Analysis System (GDAS) and the NAM
Data Analysis System (NDAS - previously called EDAS) covering much of North America.
Both data archives are available from 1997 in semi-monthly files (SM). The EDAS was saved at
80 km resolution every 3-hours through 2003, and then at 40 km resolution starting in 2004.
Additionally, ARL has been archiving NAM hybrid sigma pressure coordinate data since March
2010. These data are in three domains: CONUS with 12 km, Alaska with 12 km and Hawaii with
2 km horizontal resolution. Air agencies can also use these meteorological datasets for the
applications described in the document.
Detailed information on all meteorological data available for use in HYSPLIT can be found in
the HYSPLIT4 Users Guide
(http://www. arl. noaa.gov/documents/reports/hysplit user guide.pdf).
If trajectories are used in areas of highly complex terrain and source-receptor relationships are
relatively close (10's - 100 km), the resolution of some of the routinely used meteorological
databases for HYSPLIT may not adequately capture the meteorological conditions that govern
source-receptor relationships for a particular event. Careful consideration should be used when
selecting meteorological databases, as these will largely determine the accuracy of the trajectory
for a given event. More information on meteorological databases and their applicability to
HYSPLIT can be found at https://ready.arl.noaa.gov/archives.php.
55
-------
Vertical Motion Options. HYSPLIT can employ one of 5 different methods for computing
vertical motion. A sixth method is to accept the vertical motion values contained within the input
meteorological data set, effectively using the vertical motion method used by the meteorological
model that created the data set. Note which method was selected as well as the value chosen for
the top of the model, in meters AGL.
Trajectory Display Options. The HYSPLIT trajectory model generates a text output file of end-
point positions. The end-point position file is processed by another HYSPLIT module to produce
a Postscript display file or output files in other display formats. Some parameters, such as map
projection and size, can be automatically computed based on the location and length of the
trajectory, or they can be manually set by the user. While these display options do not directly
affect the trajectory information itself, noting these options will eliminate possible
misinterpretation of identical trajectories because of differing display options. An important
display option is the choice of vertical coordinate, usually set to meters AGL for these
assessments.
56
-------
Appendix A4. References for Guidance Document
(References for Appendix A2 are separately identified within Appendix A2.)
Akagi, S., Craven, J., Taylor, J., McMeeking, G., Yokelson, R., Burling, I., Urbanski, S., Wold, C.,
Seinfeld, J., Coe, H., 2012. Evolution of trace gases and particles emitted by a chaparral fire in California.
Atmospheric Chemistry and Physics, 12, 1397-1421.
B. Anderson, and Davis, M., 2004. Analysis of April 12-13, 2003 Kansas City Ozone Exceedances. EPA
Region 7 Technical Report.
Baker, K.R., Foley, K.M., 2011. A nonlinear regression model estimating single source concentrations of
primary and secondarily formed PM2 5. Atmospheric Environment, 45, 3758-3767.
Baker, K.R., Kelly, J.T., 2014. Single source impacts estimated with photochemical model source
sensitivity and apportionment approaches. Atmospheric Environment, 96, 266-274.
Bergin, M.S., Russell, A.G., Odman, M.T., Cohan, D.S., Chameldes, W.L., 2008. Single-Source Impact
Analysis Using Three-Dimensional Air Quality Models. Journal of the Air & Waste Management
Association, 58, 1351-1359.
Bytnerowicz, A., Burley, J., Cisneros, R., Preisler, H., Schilling, S., Schweizer, D., Ray, J., Dulen, D.,
Beck, C., Auble, B., 2013. Surface ozone at the Devils Postpile National Monument receptor site during
low and high wildland fire years. Atmospheric Environment, 65, 129-141.
Cai, C., Kelly, J.T., Avise, J.C., Kaduwela, A.P., Stockwell, W.R., 2011. Photochemical modeling in
California with two chemical mechanisms: model intercomparison and response to emission reductions.
Journal of the Air & Waste Management Association 61, 559-572.
Camalier, L., Cox, W., Dolwick, P., 2007. The effects of meteorology on ozone in urban areas and their
use in assessing ozone trends. Atmospheric Environment, 41, 7127-7137.
Chen, J., Lu, J., Avise, J.C., DaMassa, J.A., Kleeman, M.J., Kaduwela, A.P., 2014. Seasonal modeling of
PM2 5 in California's San Joaquin Valley. Atmospheric Environment, 92, 182-190.
Chin, M., Jacob, D.J., Munger, J.W., Parrish, D.D., Doddridge, B.G., 1994. Relationship of ozone and
carbon monoxide over North America. Journal of Geophysical Research, 99, 13565-14573.
Cohan, D.S., Napelenok, S.L., 2011. Air quality response modeling for decision support. Atmosphere, 2,
407-425.
Dennis, A., Fraser, M., Anderson, S., Allen, D., 2002. Air pollutant emissions associated with forest,
grassland, and agricultural burning in Texas. Atmospheric Environment, 36, 3779-3792.
Dunker, A.M., Yarwood, G., Ortmann, J.P., Wilson, G.M., 2002. The decoupled direct method for
sensitivity analysis in a three-dimensional air quality model - Implementation, accuracy, and efficiency.
Environmental Science & Technology, 36, 2965-2976.
Eder, B.K., Davis, J.M., and Bloomfield, P., 1993. A characterization of the spatiotemporal variability of
non-urban ozone concentrations over the eastern United States. Atmospheric Environment, 27A, 2645-
2668.
Eder, B.K., Davis, J.M., and Bloomfield, P., 1994. An automated classification scheme designed to better
elucidate the dependence of ozone on meterology. Journal of Applied Meteorology, 33, 1182-1199.
Fann, N., Fulcher, C.M., Baker, K., 2013. The Recent and Future Health Burden of Air Pollution
Apportioned Across US Sectors. Environmental Science & Technology, 47, 3580-3589.
57
-------
Hogrefe, C., Hao, W., Zalewsky, E., Ku, J.-Y., Lynn, B., Rosenzweig, C., Schultz, M., Rast, S.,
Newchurch, M., Wang, L., 2011. An analysis of long-term regional-scale ozone simulations over the
Northeastern United States: variability and trends. Atmospheric Chemistry and Physics, 11, 567-582.
Jaffe, D., Chand, D., Hafher, W., Westerling, A., Spracklen, D., 2008. Influence of fires on O3
concentrations in the western US. Environmental Science and Technology, 42, 5885-5891.
Jaffe, D.A., Wigder, N.L., 2012. Ozone production from wildfires: A critical review. Atmospheric
Environment, 51, 1-10.
Jaffe, D.A., Wigder, N., Downey, N., Pfister, G., Boynard, A., Reid, S.B., 2013. Impact of wildfires on
ozone exceptional events in the western US. Environmental Science & Technology, 47, 11065-11072.
Jiang, X., Wiedinmyer, C., Carlton, A.G., 2012. Aerosols from fires: An examination of the effects on
ozone photochemistry in the Western United States. Environmental Science & Technology, 46, 11878-
11886.
Jordan, T., Seen, A., Jacobsen, G., 2006. Levoglucosan as an atmospheric tracer for woodsmoke.
Atmospheric Environment, 40, 5316-5321.Kansas Department of Health and Environment, 2012. State of
Kansas Exceptional Events Demonstration April 6, 12, 13, and 29, 2011. Department of Health and
Environment, Division of Environment, Bureau of Air. November 27, 2012.
http://www2. epa. gov/sites/production/files/2015-05/documents/kdhe exevents Jinal 042011 .pdf.
Kelly, J.T., Baker, K.R., Napelenok, S.L., Roselle, S.J., 2015. Examining single-source secondary
impacts estimated from brute-force, decoupled direct method, and advanced plume treatment approaches.
Atmospheric Environment, 111, 10-19.
Kwok, R., Baker, K., Napelenok, S., Tonnesen, G., 2015. Photochemical grid model implementation of
VOC, NOx, and O3 source apportionment. Geoscientific Model Development, 8, 99-114.
Kwok, R., Napelenok, S., Baker, K., 2013. Implementation and evaluation of PM25 source contribution
analysis in a photochemical model. Atmospheric Environment, 80, 398-407.
Miller D., DeWinter J., and Reid S. Documentation of data portal and case study to support analysis of
fire impacts on ground-level ozone concentrations. Technical memorandum prepared for the U.S.
Environmental Protection Agency, Research Triangle Park, NC by Sonoma Technology, Inc., Petaluma,
CA, STI-910507-6062, September 5, 2014.
Pfister, G., Wiedinmyer, C., Emmons, L., 2008. Impacts of the fall 2007 California wildfires on surface
ozone: Integrating local observations with global model simulations. Geophysical Research Letters, 35.
Phuleria, H., Fine, P., Zhu, Y., Sioutas, C., 2005. Air quality impacts of the October 2003 Southern
California wildfires, Journal of Geophysical Research-Atmospheres, 110.
Rolph, G.D., 2016. Real-time Environmental Applications and Display sYstem (READY).
http://www.ready.noaa.gov. NOAA Air Resources Laboratory, College Park, MD.
Russell, A.G., 2008. EPA Supersites program-related emissions-based particulate matter modeling: initial
applications and advances. Journal of the Air & Waste Management Association, 58, 289-302.
Stein, A.F., Draxler, R.R, Rolph, G.D., Stunder, B.J.B., Cohen, M.D., and Ngan, F., 2015. NOAA's
HYSPLIT atmospheric transport and dispersion modeling system, Bull. Amer. Meteor. Soc., 96, 2059-
2077, http://dx. doi. org/10.1175/BAMS-D-l 4-00110.1.
58
-------
Tesche, T., Morris, R., Tonnesen, G., McNally, D., Boylan, J., Brewer, P., 2006. CMAQ/CAMx annual
2002 performance evaluation over the eastern US. Atmospheric Environment, 40, 4906-4919.
U.S. Environmental Protection Agency, 2014. Health Risk and Exposure Assessment for Ozone, Final
Report. Office of Air Quality Planning & Standards, Research Triangle Park, NC. EPA-452/P-14-004a,
502pp.
Watson, J., Chow, J., Houck, J., 2001. PM2 5 chemical source profiles for vehicle exhaust, vegetative
burning, geological material, and coal burning in Northwestern Colorado during 1995. Chemosphere, 43,
1141-1151.
Wigder, N.L, Jaffe, D.A., Saketa, F.A., 2013. Ozone and particulate matter enhancements from regional
wildfires observed at Mount Bachelor during 2004-2011. Atmospheric Environment, 75, 24-31.
Yokelson, R., Bertschi, I., Christian, T., Hobbs, P., Ward, D., Hao, W., 2003. Trace gas measurements in
nascent, aged, and cloud-processed smoke from African savanna fires by airborne Fourier transform
infrared spectroscopy (AFTIR). Journal of Geophysical Research-Atmospheres, 108.
Zhou, W., Cohan, D.S., Pinder, R.W., Neuman, J.A., Holloway, J.S., Peischl, J., Ryerson, T.B., Nowak,
J.B., Flocke, F., Zheng, W.G., 2012. Observation and modeling of the evolution of Texas power plant
plumes. Atmospheric Chemistry and Physics 12, 455-468.
59
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-457/B-16-001
Environmental Protection Air Quality Policy Division September 2016
Agency Research Triangle Park, NC
------- |