Using Artificial Intelligence to Improve Data Accuracy of Air Pollutants under the Clean Air Act

AuthorLauren Palley
PositionJ.D. Candidate, American University Washington College of Law 2019.
Pages13-14
13
Fall 2018
uSing artificial intelligence to improve Data
accuracy of air pollutantS unDer the clean
air act
Lauren Palley*
Over the last century, industrialization and the air pol-
lution that has come with it have put the planet and its
future stability at risk.1 Articial intelligence technol-
ogy (AI), part of a larger “Fourth Industrial Revolution,” has
the potential to mitigate these effects through widespread imple-
mentation by the Environmental Protection Agency (EPA).2
The EPA is the governmental agency responsible for regulating
air pollutants pursuant to the Clean Air Act (CAA).3 Congress
delegated the authority to the EPA to regulate greenhouse gases
(GHGs), such as carbon dioxide, that trap solar energy in the
atmosphere.4 Under the CAA and the Supreme Court’s decision
in Massachusetts v. U.S. Environmental Protection Agency,5 the
EPA has both the authority and the duty to regulate GHGs using
AI since it is the best available technology.
The CAA requires the EPA to set health-based standards for
ambient air quality, set deadlines as to when the achievement of
those standards must be met, and set national emission standards
for large sources of air pollution, including motor vehicles,
power plants, and other industrial sources.6 Section 109 of the
CAA requires the EPA to establish National Ambient Air Quality
Standards (NAAQS) for pollutants that endanger the public
health or welfare, in the EPA Administrator’s judgment, and
whose presence in ambient air results from numerous or diverse
sources.7 The NAAQS for “certain common and widespread
pollutants” must be based on the “latest science.”8 The latest
science is AI, and therefore, the EPA has the authority under the
CAA to use AI in relation to the NAAQS.
AI describes computer systems that simulate human intel-
ligence through their ability to think, learn, and sense their
environment.9 AI is the most advanced technology for analyz-
ing large amounts of data, reaching conclusions about that data,
nding patterns, and predicting future behavior.10 It has the
potential to be at the forefront of solving climate change issues
and creating a more sustainable future if it is implemented in
various key areas, especially in data collection and processing
of air pollutants pursuant to the EPA’s duty. AI can assist with
measuring harmful GHGs that have previously been invisible to
the naked eye, particularly methane, more effectively and thus
creating a larger, more accurate dataset to analyze.11 Acting
under its authority and duty to regulate GHGs, the EPA’s imple-
mentation of AI would allow pollution that was previously dif-
cult to observe, measure, and report to be visible by all parties
involved in real time.
Additionally, case law over the last decade has further
dened the EPAs authority and duty to regulate GHGs using the
newest technologies. In Massachusetts v. U.S. Environmental
Protection Agency, the Supreme Court relied heavily on sci-
entic data regarding global warming when it established that
the EPA not only has the ability, but also has a duty to regulate
GHGs as they fall under the CAA’s denition of “air pollutant.”12
The Court also emphasized in a later case, American Electric
Power Co. v. Connecticut,13 that due to Congress’s delegation
of authority to the EPA, it is the most equipped body to deal
with GHGs because agencies can utilize “scientic, economic,
and technological resources.”14 Additionally, the Court ruled in
Utility Air Regulatory Group v. U.S. Environmental Protection
Agency15 that the Agency has the authority to require the best
available control technology (BACT) from certain previously
regulated sources.16 Therefore, the EPA’s authority to regulate
GHGs using the best technology available, such as AI, is consis-
tent with relevant case law.17
The Ofce of Inspector General (OIG) of the EPA authored a
report and key example where AI could be benecial and should
be implemented. The report showed that of all major CAA
facilities that have had an evaluation in the last ve years, data
uploaded into the EPA’s Enforcement and Compliance History
Online (ECHO) system was inaccurate.18 Data was either not
reported or was inaccurately entered into the database.19 These
errors went undetected “because of a lack of data quality over-
sight that would identify facilities overdue for [Full Compliance
Evaluation].”20 This inaccurate data hindered the EPA’s over-
sight of compliance programs and allowed for numerous major
CAA facilities to potentially emit large amounts of undetected or
unreported air pollutants.
Taking new technologies into consideration under its
authority and duty to regulate GHGs using the most advanced
technology, the EPA nalized a rule in 2016 establishing new
source performance standards for the oil and natural gas sector.21
In part, it mandates that “monitoring of the components must be
conducted using optical gas imaging,”22 in addition to adding a
provision for emerging technology such as continuous emissions
monitoring technologies.23 Optical gas imaging (OGI) is the use
of infrared cameras to detect invisible pollution such as methane
leaks and provides images of a leak depicted as black clouds.24
However, while these images are helpful, they cannot provide
quantitative information about the fugitive emissions they pho-
tograph.25 Quantitative data is crucial for GHG management
because “you can’t improve what you can’t measure.”26 The
*J.D. Candidate, American University Washington College of Law 2019.
14 Sustainable Development Law & Policy
EPA’s implementation of AI is necessary, in addition to OGI, to
measure several key GHG emissions. Without the use of AI, the
EPA is failing to meet its duty to regulate GHGs using the new-
est available technology.
Because the EPA has the authority to regulate signicant air
pollutants that are emitted from facilities, accurate data collec-
tion is crucial for effective reporting. Once the data is gathered,
the EPA must efciently analyze it to achieve accurate results
to view past, current, and future emissions. Gathering accurate
emissions data is only one step of the process, but it is essential
for determining NAAQS. As part of EPA’s Next Generation
Compliance Program, the EPA is “commit[ed]” to start using
outside sources for data to improve data accuracy.27 Once EPA
integrates AI within the Agency and employs outside sources, the
technology would benet the entities it regulates, decisionmak-
ers, and all communities impacted by air pollution.28 Pursuant
to the CAA and relevant case law, the EPA is required to use
the latest available technology and must strive to incorporate
articial intelligence more widely and with more urgency.
enDnoteS
1 Celine Herweijer, Dom inic Waughray, Fourth Indust rial Revolution
for the Earth: Harne ssing Articial Intellige nce for the Earth, pricewater-
houSecooperS (Jan. 2018), https://www.pwc.com/gx/en/sustainability/assets/
ai-for-the-earth-jan-2018.pdf (referring to a “ Fourth Industri al Revolution”
character ized by AI, an “establi shed digital economy . . . the I nternet of
Things, robo ts, autonomous vehicles, biotech nology, nanotechnology and
quantum c omputing” that will t ransform curre nt industries).
2 Id. (discussing how A I is only one character istic of the “Fourth Indu strial
Revolution” ).
3 42 U.S.C. § 7401 (2012).
4 Massachuset ts v. U.S. Envtl. Prot. Agency, 549 U.S. 497, 497 (2007)
(holding that t he EPA has the authority to regu late GHGs and explaining t he
danger of GHG as “when c arbon dioxide is released i nto the atmosphere, it
acts like the ce iling of a greenhouse , trapping solar energ y and retarding the
escape of reecte d heat”).
5 Id.
6 cong. reSearch Serv., rl 30853, clean air act: a Su mmary of the
act anD itS maJor r equirementS ( nov. 7, 2018) [hereinafter Clean Ai r Act:
A Summar y].
7 Id. at 3 (referring t o the Administrat or of the EPA, the head of the
agency).
8 Clean Air Act Require ments and History, u.S. envtl . prot. agency
(Jan. 10, 2017), https://www.epa.gov/clean-air-act-overview/clean-air-act-
requirements-and-history; see also Clean Ai r Act: A Summary, supra note 6,
at 3 (discussing that t he CAA requires the EPA to review t he scientic data
which the stand ards are based ever y ve years).
9 Using Articial I ntelligence in Law Depart ments, thomSon reuterS, 2018
https://1.next.westlaw.com/Document/Iddbbd1a8003511e89bf099c0ee06c731/
View/FullText.html?contextData=(sc.Default)&transitionType=Default (last
visited Dec. 20, 2018) (noting how AI improves t asks over time, but this is n ot
an ofcial legal de nition).
10 Id.
11 See Cynthia G iles, Next Generation Compliance, 30 envtl. forum 22, 22
(Sept./oct. 2013).
12 Massachusetts, 549 U.S. at 506 (citing to the CA A’s denition of “air
pollutant” t o include “any air pollution agent o r combination of such agents,
including any physic al, chemical, biological, rad ioactive . . . substance or
matter which is e mitted into or other wise enters the ambien t air”).
13 564 U.S. 410 (2011).
14 Id. at 428 (reinforcing t hat the EPA, as the expert agenc y on the matter, is
best suited “ to serve as primar y regulator of green house gases”).
15 573 U.S. 302 (2014).
16 Id. at 331 (holding the EPA’s decision to require best availa ble control
technology for gr eenhouse gases emit ted by certain sou rces is permissible
under the Clea n Air Act).
17 E.g., id. at 329; Am. Elec. Power C o., 564 U.S. at 410; Massachusetts, 549
U.S. at 497.
18 u.S. envtl. prot. agency, enforcemen t anD complianc e: clean air
act facility evaluationS are conDucteD, but inaccu rate Data hinDer epa
overSight anD p ublic awareneSS 9 (2016) (explaining how t he EPA conducts
facility evalu ations for all major CAA facilit ies in accordance wit h the Com-
pliance Monitor ing Strategy (CMS) to ensu re companies’ compliance w ith
EPA laws and regulations).
19 Id. (clarifyi ng that ECHO integrate s data from other EPA databas es to
provide the public wit h facility-specic co mpliance informatio n data from
other EPA databases t o provide the public with facil ity-specic compliance
information).
20 Id.
21 Oil and Natu ral Gas Sector: Emission St andards for New, Reconstr ucted,
and Modied Sou rces, Final Rule, 81 Fed. Reg. 107, 35824 (Jun. 3, 2016).
22 Id. at 35846 (asserting it s authority to regula te monitoring and rep air of
fugitive emi ssion components at well sites and co mpressor stations).
23 Id. at 35861 (allowing for the use of emerging t echnology, like continu-
ous emissions mon itoring technologies , and agreeing that th e continued
development of these cost- ef fective technologies should be e ncouraged); see
also David A. Hind in & Jon D. Silberman, Desig ning More Effective Rules
and Perm its, 7 geo. waSh. J. energy & envtl. l. 103, 123 (2016) (“[P]rovid-
ing regulate d entities with accur ate measures, in a st andardized form at, of
deviations ind ication that regulato ry requirements a re being, or may soon be,
viol at ed.” ).
24 Laith Amin , Disruptive Technolog y Meets the Intractable Chall enge of
Fugitive Gas Emissions, aDviSian, ht tps://www.worleyparsons.com/~/media/
Files/ w/WorleyParsons/docum ents/markets/fu gitivegasemissions .pdf (last
visited Nov. 12, 2018). (explaining how OGI utilizes i nfrared technolog y to
capture em issions otherwise i nvisible to the human eye); see also Rober t L.
Glicksman, Dav id L. Markell & Claire Monteleon i, Technological Innova-
tion, Data Analytics, and Environmental Enforcement, 44 ecology l . q. 41,
67 (2017) (discussi ng how agencies, including the EPA, are st arting to util ize
infrar ed camera technology).
25 Amin, supra note 2 4.
26 Id.
27 Glicksman, Ma rkell & Monteleoni, supra note 24, at 69; see also Giles ,
supra note 11 (describing the EPA’s Next Generation Compliance i nitiative
and its ve key component s: design regulations an d permits, advance s
emissions and pol lutant detection tech nology, electronic report ing, expanded
transpa rency, and innovative enforcem ent).
28 Glicksman, Ma rkell & Monteleoni, supra note 24, at 65.

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT