ATMOSPHERIC
SCIENCE
Attribution
of individual
methane
and carbon
dioxide
emission
sources using
EMIT observa
tions from space
Andr
ew K. Thorpe
1
*, Robert
O. Green
1
, David R. Thompson
1
, Philip
G. Brodrick
1
,
John
W. Chapman
1
, Clayton
D. Elder
1
, Itziar
Irakulis-Loitxa
te
2,3
, Daniel
H. Cusworth
4,5
,
Alana
K. Ayasse
4,5
, Riley M. Duren
1,4,5
, Chris
tian Frankenberg
6
, Luis Guanter
2,7
, John
R. Worden
1
,
Philip
E. Dennison
8
, Dar A. Roberts
9
, K. Dana
Chadwick
1
, Michael
L. Eastwood
1
, Jay E. Fahlen
1
,
Charles
E. Miller
1
Carbon
dioxide
and methane
emissions
are the two primary
anthropogenic
climate-forcing agents
and an im-
portant
source of uncertainty
in the global
carbon
budget.
Uncertainties
are further
magnified
when emissions
occur
at fine spatial scales
(<1 km), making
attribution
challenging.
We present
the first observa
tions from
NASA
’
s Earth Surface Mineral Dust Source Investigation (EMIT)
imaging
spectrometer
showing quantifica
tion
and attribution
of fine-scale
methane
(0.3 to 73 tonnes
CH
4
hour
−
1
) and carbon
dioxide
sources (1571
to
3511 tonnes
CO
2
hour
−
1
) spanning
the oil and gas, waste, and energy
sectors.
For selected
countries
observ
ed
during
the first 30 days of EMIT operations, methane
emissions
varied
at a regional
scale, with the largest total
emissions
observ
ed for Turkmenis
tan (731 ± 148 tonnes
CH
4
hour
−
1
). These
results
highlight
the contributions
of current and planned
point source imagers
in closing
global
carbon
budgets.
Copyright
©
2023
The
Authors,
some
rights
reserv
ed;
ex
clusiv
e licensee
American
Associa
tion
for
the
Advancement
of
Science.
No
claim
to
original
U.S.
Go
vernment
Works.
Dis
tributed
under
a Cr
ea
tiv
e
Commons
Attribution
NonC
ommer
cial
License
4.0
(CC
BY-NC).
INTRODUCTION
Carbon
dioxide
(CO
2
) and methane
(CH
4
) are the two primary
an-
thropogenic
clima
te-for
cing agents.
Fossil
carbon
dioxide
emis-
sions
are rapidly
becoming
the larges
t source of uncertainty
in
the global
carbon
budget
because
of increasing
emissions
in areas
with poor
reporting
requirements
(
1
). Global
inventories
of fossil
carbon
dioxide
emissions
rely on activity
data and have an estimated
uncertainty
in excess
of 0.5 billion
(metric)
tonnes
of C (Gt C)
year
−
1
(
2
). Uncertainties
vary consider
ably between those
countries
with mature reporting
requirements
and data transpar
ency
from a
few percent (
3
) to much
larger
uncertainties
(15 to 18%)
for coun-
tries such as China
(
4
,
5
). Carbon
dioxide
emissions
from stationary
sources (e.g.,
power plants,
cement
production,
and refineries)
make up approxima
tely one-third
of the anthr
opogenic
budget
(
6
), and the location and magnitude
of large
point
sources contrib-
ute to uncertainty
in gridded
inventories
(
7
). Reducing
uncertainty
from these
sources offers
the potential
to reconcile
global
fossil
carbon
dioxide
emissions.
For global
methane
emissions,
bottom-up
inventory
uncertain-
ties range
between 20 and 35% for the agricultur
e, waste, and fossil
fuel sectors;
50% for biomass
burning
and natural wetland
emis-
sions;
and 100%
or higher
for natural sources such
as geological
seeps
and inland
waters (
8
). Large
uncertainties
remain
regarding
partitioning
between these
sources and
the cause
of recent
changes
in the atmospheric
growth rate of methane
(
9
,
10
).
Current best estimates place anthr
opogenic
methane
emissions
at
around
half of total
methane
emissions
(
8
). Because
the methane
lifetime
in the atmospher
e is only about
10 years (mean
of 9.8 ±
1.6 years)
(
11
) and methane
is more efficient
at trapping
radiation
than
carbon
dioxide,
targeting
reductions
in anthr
opogenic
methane
emissions
offers
an effectiv
e approach to decrease overall
atmospheric
radiative forcing.
Satellite
instruments
measuring
atmospheric
methane
and
carbon
dioxide
in the shortw
ave infrared include
area flux
mappers
and point
source imagers
with complementary
attributes
(
12
).
With coarse
spatial resolution
(0.1 to 10 km),
area flux
mappers
such
as the Tropospheric
Monitoring
Instrument
(TROPOMI)
(
13
) and the Orbiting
Carbon
Observa
tory (OCO) in-
struments
(
14
) are best suited
to quantify
methane
and carbon
dioxide
emissions
on regional
to global
scales.
Because
of fine
spatial resolutions
(typically
≤
60 m), point
source imagers
can
resolve individual
point
sources down to the range
of 0.1 to 10
tonnes
CH
4
hour
−
1
and attribute
emissions
to specific
sectors.
The first observa
tion of a methane
point
source from space used
the Hyperion
imaging
spectr
ometer
for the 2015
Aliso
Cany
on
blowout
(
15
).
Since
2016,
the GHGSa
t constellation of Fabry-
Perot interfer
ometers
has identified
methane
point
sources at coal
mine
vents (
16
),
the oil and gas sector
(
17
),
and landfills
(
18
).
The
PRISMA
imaging
spectr
ometer,
focused
primarily
on measuring
Earth
surface properties,
has recently
been
used
to quantify
methane
point
source emissions
from a gas well blowout
(
19
),
as
well as oil and gas and coal
mine
emissions
(
20
).
Multiband
imagers,
such
as Sentinel-2
(
21
,
22
), WorldVie
w-3 (
23
),
and
Landsa
t-8 (
24
), have also been
able to identify
similar
source types.
The Earth
Surface Miner
al Dust Source Investigation
’
s
(EMIT)
(
25
,
26
) core technology
and potential
greenhouse
gas mapping
ca-
pability
were first demons
trated using
visible
to shortw
ave infrared
airborne
imaging
spectr
ometers
developed
at the Jet Propulsion
Labor
atory (JPL).
The Airborne
Visible/Infr
ared Imaging
Spec-
trometer
(AVIRIS)
(
27
)
mapped
the first remotely
sensed
methane
point
source plume
at the Coal Oil Point seeps
offshor
e
of Santa
Barbar
a, California
(
28
),
followed by emissions
at the
1
Jet
Pr
opulsion
Labor
atory
, California
Ins
titute
of
Technology
, Pasadena,
CA,
USA.
2
Univ
ersita
t Politècnica
de
València
(UPV),
Valencia,
Spain.
3
Interna
tional
Methane
Emissions
Observa
tory
, United
Na
tions
Envir
onment
Pr
ogr
amme,
Paris,
France.
4
Carbon
Mapper,
Pasadena,
CA,
USA.
5
Univ
ersity
of
Arizona,
Tucson,
AZ,
USA.
6
California
Ins
titute
of
Technology
, Pasadena,
CA,
USA.
7
Envir
onmental
Defense
Fund,
Ams
terdam,
1017,
Netherlands.
8
Univ
ersity
of
Utah,
Salt
Lak
e
City
, UT
,
USA.
9
Univ
ersity
of
California,
Santa
Barbar
a,
Santa
Barbar
a,
CA,
USA.
*C
orr
esponding
author.
Email:
andr
ew.k.thorpe@jpl.nasa.go
v
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
1 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
Inglewood
Oilfield
in Los Angeles
County
(
29
)
and the Aliso
Canyon blowout
(
30
).
The next-gener
ation instruments,
AVIRIS-
NG (
31
) and the Global
Airborne
Observa
tory (GAO) (
32
),
have
surveyed the oil and gas sector
in the Four Corners
Region
(
33
);
the Permian
(
34
);
the Uinta,
Denv
er-Julesburg,
and Marcellus
basins
(
35
); and the Gulf of Mexico (
36
) of the United
States. In ad-
dition,
methane
plumes
were identified
in Canada
(
37
) and India
(
38
).
Methane
emissions
have also been
characterized
for the
waste (
38
,
39
) and agricultur
al sectors
(
40
,
41
), as well as numer
ous
methane
emission
hotspots
identified
in thawing
perma
frost
regions
of Alaska
and Canada
(
42
,
43
). Results
from airborne
studies
using
AVIRIS-NG
and
GAO across multiple
source
sectors
(oil and gas, coal,
livestock,
and waste) indica
te that
methane
point
sources are often
consider
able contributors
to net
regional
emissions,
ranging
between 13 and 67% of the total
for
some
areas in the United
States (
35
).
Carbon
dioxide
emissions
have also
been
mapped
with
AVIRIS
(
44
),
AVIRIS-NG
(
45
),
GAO, and PRISMA
(
19
,
46
).
While
limited
in spatial coverage, this previous
work
demon-
strates the utility
of airborne
imaging
spectr
ometers
to identify
and quantify
methane
and carbon
dioxide
point
sources and to pri-
oritize
mitiga
tion efforts.
In addition
to the greenhouse
gas appli-
cation,
these
instruments
have been
used
for remote
sensing
of
Earth
’
s
surface, including
miner
al (
47
,
48
), ecosystems
(
49
,
50
),
water quality
(
51
,
52
), and snow and ice applica
tions
(
53
,
54
). Dem-
onstration
of these
varied
science
applica
tions
with
airborne
imaging
spectr
ometers
laid the groundwork
for the EMIT
mission.
EMIT
was launched
on 14 July 2022
and is currently
operating
on the Interna
tional
Space Station (ISS).
This imaging
spectr
ometer
measur
es reflected
solar
radiation from the visible
to shortw
ave in-
frared and was designed
to determine
the miner
al composition
of
arid,
dust source regions
and their
influence
on global
radiative
forcing.
EMIT
measur
es 285 distinct wavelengths
at 7.4-nm
spectr
al
sampling
between 381 and 2493
nm, including
prominent
methane
and carbon
dioxide
absorption
features in the shortw
ave infrared.
EMIT
features a 1242-element
swath at approxima
tely 60-m
image
pixel
resolution.
EMIT
has demons
trated exceptionally
high
performance,
with
a spectr
al uniformity
greater than
98%
and a signal-to-noise
ratio (SNR)
ranging
from 500 to 750 for
most wavelengths
at the US Geological
Survey (USGS)
Libya 4 cal-
ibration site (
55
).
Compar
ed to other
point
source imagers,
EMIT
provides
a unique
combina
tion of high SNR and large
80-km
swath,
which
enables
an average daily
coverage of 1.3 × 10
6
km
2
, equivalent
to the area of the Republic
of South
Africa.
This unique
combina-
tion of wide
coverage, fine spatial resolution,
high SNR,
and excel-
lent spectr
al uniformity
distinguishes
EMIT
from other
imaging
spectr
ometers
and makes it particularly
well suited
for identifying
methane
and carbon
dioxide
emissions.
While
focusing
on coverage
of arid lands,
EMIT
will map broad regions
of Earth
’
s
surface at lat-
itudes
from +51.6°
to
−
51.6°
(Fig.
1), which
is dictated by the in-
clined,
equatorial
orbit
of the ISS. EMIT
should
complete
its
primary
mission
by September
2023,
and an extended
mission
is
planned,
during
which
any regions
within
ISS latitude
constraints
could
be targeted.
In this study, we present
the first results
using
EMIT
to charac-
terize
methane
and carbon
dioxide
point
sources from the first 30
days of data collection.
We gener
ated methane
and carbon
dioxide
retrievals
using
EMIT
radiance
data; estimated emission
rates using
the integr
ated mass
enhancement
(IME)
combined
with
wind
speeds;
and attributed
emissions
to the oil and gas, waste, and
energy
sectors.
We focused
our analysis on a number
of countries
in the Middle
East and Central Asia with consider
able production
of
oil, gas, and coal and limited
emissions
reporting.
We present
the
overall distribution
of observ
ed emissions
to assess
its potential
for
improving our unders
tanding
of global
greenhouse
gas budgets
and
to inform
mitiga
tion strategies.
RESUL
TS
EMIT
coverage,
revisit
frequency
, and implica
tions
for
emission
estimates
Given a primary
mission
to map the miner
al composition
of arid
lands,
EMIT
targets
the white
regions
shown in Fig. 1A, an area
of 2.8 × 10
7
km
2
and 19.3%
of Earth
’
s
land
surfaces. In the first
30 days of data collection,
EMIT
’
s total
coverage was 3.8 × 10
7
km
2
. This
includes
regions
with
methane
emissions
as predicted
by bottom-up
inventories
like those
of Scarpelli
et al.
(
56
), particu-
larly
for the Western United
States, North
and West Africa,
the
Middle
East, and Central and East Asia,
as well as a number
of
power plants
from Anne
x I and Non
–
Anne
x I countries
(Fig. 1A).
EMIT
’
s
wide
coverage offers
the potential
for identifying
emissions
not only in regions
where we might
anticipa
te emissions
but also in
places where there is poor
prior
knowledge
of emissions.
Figur
e 1B shows the expected
distribution
of the EMIT
revisit
rate over the course
of its 1-year primary
mission.
This simula
tion
does not include
all calibr
ation targets,
cloud
cover screened
out on
board,
or potential
downlink
interruptions.
However, the simula-
tion gives a sense
of the typical
revisit
capability
of the instrument,
with a median
of 10 revisits
per annum
and 90% of the target
mask
having between 3 and 15 revisits.
If revisits
were evenly spaced, then
this is equivalent
to a median
average revisit
rate of 36.5 days (90%
of the target
mask
with a frequency
between 24 and 122 days), al-
though
the revisit
interval
is rarely even due to a combina
tion of ISS
trajectory
and solar
angles.
Some
locations
near
the +51.6°
and
−
51.6°
turning
points
of the ISS orbit
will receive more than 20 re-
visits
over the year, with a minimum
average revisit
interval
of 4.9
days. The precessing
ISS orbit
allows for sampling
across the sunlit
interval,
thus allowing
assessment
of diurnal
variability
for some
emission
sources. This unique
combina
tion of large
area coverage
and repeat measur
ements
enables
tracking
of emissions
over time,
an assessment
of persis
tence,
and reduction
in uncertainties
associ-
ated with carbon
budgets.
Mapping
and attributing
carbon
dioxide
and methane
emissions
We demons
trate the capability
of EMIT
for quantifying
poorly
characterized
emissions
from carbon
dioxide
and methane
point
sources and attributing
colloca
ted emissions
to distinct
emission
sectors.
Both
of these
capabilities
are critical
to resolving
a class
of highly
uncertain
emissions
occurring
at fine spatial scales
that
constitute
an important
source of uncertainty
in the global
carbon
budget.
Cusworth
et al.
(
46
) demons
trated the quantifica-
tion of carbon
dioxide
plumes
using
both
the airborne
GAO and
spaceborne
PRISMA
imaging
spectr
ometers
with good
agreement
agains
t reported
emissions
for U.S. power plants.
This highlights
the
potential
for quantifying
carbon
dioxide
emissions
from global
sta-
tionary
sources using
instruments
such
as EMIT
, which
is
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
2 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
particularly
important
for non
–
Anne
x I countries
where emissions
are not reported
and uncertainties
remain
high.
Repeat observa
tions
of power plants
are critical
to constrain
annual
facility
emissions
in jurisdictions
that lack continuous
emis-
sion monitoring
and reporting.
Hill
et al.
(
57
) estimated that 18
OCO-2
observa
tions
are needed
to constrain annual
carbon
dioxide
emissions
of fossil
fuel power plants
to within
15%,
while
29 PRISMA
overpasses
are required to reduce
non
–
Anne
x I fossil
fuel power plant
emission
uncertainty
below 1 Gt CO
2
per year
(within
14%)
(
46
).
Figur
e 1A shows the locations
of power plants
with
≥
500-MW
capacity (
58
),
and the inlay shows the distribution
of those
power plants
that will be observ
ed with EMIT
during
the 1-
year primary
mission.
Measur
ements
range from 1 to 18 revisits
for
856 non
–
Anne
x I and 828 Anne
x I power plants
with approxima
tely
half observ
ed
≥
10
times.
Given this,
EMIT
observa
tions
will
contribute
appreciably
to the observa
tional
requir
ements
needed
to constrain stationary
source carbon
dioxide
emissions.
Carbon
dioxide
plumes
for two coal-fir
ed power plants
in the
Xinjiang
Uygur
Autonomous
Region
of the People
’
s
Republic
of
China
are shown in Fig. 2A. Carbon
dioxide
emissions
were esti-
mated at 1571 ± 229 tonnes
CO
2
hour
−
1
for the Shenhuo
Zhundong
(1400-MW
capacity)
and 3511
± 537 tonnes
CO
2
hour
−
1
for the
Xinjiang
Qiya Smelter
(2160-MW
capacity)
power stations
(
59
).
In both examples,
the observ
ed plumes
are colloca
ted with combus-
tion stacks. While
these
power plants
lack continuous
emission
monitoring
and reporting,
EMIT
will provide 13 repeat observa-
tions
during
the 1-year primary
mission,
offering
the potential
to
better
constrain emissions
at these
two facilities.
Decomposition
of organic
material
in anaer
obic
conditions
results
in consider
able
waste sector
methane
emissions,
Fig. 1. EMIT
coverage and total
observa
tions
for simula
ted 1-year mission.
(
A
)
EMIT
observa
tion
mask
sho
wn
in
white
rela
tiv
e to
bottom-up
methane
inv
entory
(
56
)
and
loca
tions
of
po
wer
plants
with
≥
500-MW
capa
city
sho
wn
in
blue
(
58
).
The
inla
y stack
ed
his
togr
am
sho
ws
≥
500-MW
po
wer
plants
that
will
be
observ
ed
with
EMIT
and
associa
ted
revisits
for
the
1-y
ear
primary
mission
for
non
–
Anne
x I (dark
blue)
and
Anne
x I (light
blue)
countries.
EMIT
should
complete
its
primary
mission
by
September
2023,
and
in
the
planned
extended
mission,
any
terr
es
trial
region
betw
een
the
dashed
lines
(+51.6°
and
−
51.6°
latitude)
could
be
targeted.
(
B
)
Total
observa
tions
and
average
revisit
interval
for
nominal
1-y
ear
mission.
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
3 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
representing
around
18% (for solid
waste and wastewater) of global
anthr
opogenic
methane
emissions
(
8
). Largely
driven by popula
tion
growth (
60
),
global
solid
waste sector
emissions
are expected
to
double
by 2050
with
uncertainties
remaining
high
(
61
).
Recent
studies
have identified
large
methane
emissions
at landfills
and
waste dumps
in the United
States (
35
,
39
), Argentina,
and India
(
18
).
While
a complete
global
database
of landfills
and dumps
cur-
rently
does not exist, satellite
observa
tions
have the potential
to play
an important
role in landfill
geoloca
tion and methane
emission
monitoring.
Methane
emissions
were observ
ed for 11 landfills
in this study.
One example
is shown in Fig. 2B, where a methane
plume
is clearly
visible
(5 ± 1 tonnes
CH
4
hour
−
1
) and is consis
tent with emissions
from the active face of the Arad Kouh
landfill
near
Tehran, Iran.
Operating since
1976,
this facility
uses sorting
and recycling
of
waste, as well as compos
ting (
62
),
but does
not use a landfill
gas
captur
e system.
Figur
e 3 highlights
EMIT
’
s
ability
to pinpoint
multiple
emission
sources that are in close
proximity
and to attribute
these
emissions
to differ
ent sectors.
Methane
plumes
appear
from multiple
sources
including
the Riyadh landfill
in Saudi
Arabia (8 ± 2 tonnes
CH
4
hour
−
1
). Opened
in 2006,
this landfill
receives unsorted
waste and
uses an open
collection
pit for leachate (
63
). A methane
plume
also
appears
from a sewage treatment
plant
(7 ± 1 tonnes
CH
4
hour
−
1
)
and natural gas
–
fir
ed power plant
(30 ± 2 tonnes
CH
4
hour
−
1
) (
64
),
where a carbon
dioxide
plume
was also visible
(2032
± 142 tonnes
CO
2
hour
−
1
). Emissions
of methane
likely associa
ted with
power
plant
startup
conditions
have been
quantified
in previous
studies
using
airborne
in situ (
65
,
66
) and imaging
spectr
ometer
measur
e-
ments
(
67
).
Given the overlap
between
methane
plumes,
Fig. 2. Carbon
dioxide
plumes
from power plants
and methane
from landfill
observ
ed from space.
(
A
)
Carbon
dioxide
plumes
from
emission
stacks
at two
po
wer
plants
in
China
(1571
±
229
tonnes
CO
2
hour
−
1
example
to
the
north;
3511
±
537
tonnes
CO
2
hour
−
1
example
to
the
south).
(
B
)
Methane
plume
from
activ
e face
of
a
landfill
in
Iran
(5
±
1 tonnes
CH
4
hour
−
1
). ppmm,
parts
per
million
meter.
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
4 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
interpr
etation of additional
sources is challenging;
however, a po-
tential
source (28 ± 2 tonnes
CH
4
hour
−
1
) was identified
in close
proximity
to the Shedgum-Y
anbu
natural gas liquids
pipeline
(
68
).
Multiple
prominent
methane
plumes
were observ
ed in Turk-
menis
tan near
the Caspian
Sea, including
a set of 12 sources
shown in Fig. 4A, with
total
emissions
of 163 ± 18 tonnes
CH
4
hour
−
1
. Clusters of plumes
are associa
ted with the Goturdepe
and
the Barsagelmez
fields,
both
producing
crude
oil, condensa
te, and
natural gas liquids
(
69
).
Oil and gas pipelines
are also shown;
however, all emission
sources appear
associa
ted with
gathering
lines
that are visible
in high-r
esolution
imagery
available
on
Google
Earth.
This location was imaged
twice
on 15 Augus
t 2022
at 4:28 UTC (54° solar
zenith)
and 10:58
UTC (38° solar
zenith),
and the orienta
tion of the methane
plumes
changes
with
shifting
wind
direction
(Fig.
4A, inset).
The ISS precessing
orbit
enables
this unique
capability
to assess
emission
variability
at two times
throughout
the day. In Fig. 4B, another
set of 15 methane
plumes
was observ
ed at the Dauletabad-Donmez
field (condensa
te, natural
gas liquids,
and gas),
with
sources that include
a gas compr
essor
station, flares, and gathering
lines.
For this example,
total emissions
were estimated at 64 ± 12 tonnes
CH
4
hour
−
1
. A number
of methane
sources shown in Fig. 4 were located in close
proximity
, which
re-
sulted
in overlap
for those
areas used
to estimate IME values
with
the prescribed
1000-m
maximum
fetch
(see the
“
Calcula
ting emis-
sions
”
section).
In these
examples,
overlapping
plumes
were
removed to ensur
e that emissions
were not overestimated.
Methane
emissions
from the Middle
East and Central Asia
We analyzed
emissions
from energy-pr
oducing
countries
in the
Middle
East (Saudi
Arabia,
Iran, the United
Arab Emirates, Iraq,
Qatar, Kuwait, and Oman)
and Central Asia
(Kazakhs
tan, Turk-
menis
tan, and Uzbekis
tan).
These
countries
have large
disagr
ee-
ment
between global
inversion
results
(
70
,
71
) and discrepancies
relative to bottom-up
inventories
(
56
).
Figur
e 5 shows coverage
by country
for data acquir
ed in the first 30 days of the EMIT
mission,
as well as the location, magnitude,
and sector
for observ
ed
plumes.
Fig. 3. Source attribution
of greenhouse
gas plumes
in close
proximity
from differ
ent emission
sectors.
Emissions
fro
m
a landfill
in
Saudi
Ar
abia
(28
± 2 tonnes
CH
4
hour
−
1
), from
a was
tewater
trea
tment
facility
(7
± 1 tonnes
CH
4
hour
−
1
), from
a po
wer
plant
(30
± 2 tonnes
CH
4
hour
−
1
and
2032
± 142
tonnes
CO
2
hour
−
1
), and
potentially
from
a na
tura
l gas
pipeline
(28
±
2 tonnes
CH
4
hour
−
1
). Inla
y images
sho
w
a close
up
of
the
po
wer
plant
sour
ce,
with
coloca
ted
methane
and
carbon
dioxide
emissions.
NGL,
na
tur
al
gas
liquids.
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
5 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
The spatial distribution
of the oil and gas methane
emissions
shown in Fig. 5 clearly
indica
tes significant
regional
differ
ences.
For example,
Saudi
Arabia
is the world
’
s
larges
t oil exporter;
however, we identified
only
one methane
plume
from the Saudi
oil and gas sector
with
EMIT
despite
covering
65%
of the
country
. The total number
and magnitude
of observ
ed oil and gas
emissions
from Turkmenis
tan are far greater than
its neighboring
countries,
Iran and Uzbekis
tan, but the root causes
for these
differ-
ences
are unclear.
Uzbekis
tan and Turkmenis
tan production
is
mostly from natural gas (90 and 87%,
respectiv
ely) (
72
);
however,
their
oil and gas emissions
differ
ed markedly (87 ± 22 and 731 ±
148 tonnes
CH
4
hour
−
1
, respectiv
ely).
While
there are consider
able uncertainties
associa
ted with global
methane
flux
inversion
results
attributed
to these
countries,
Turkmenis
tan, Uzbekis
tan, and Iran have consis
tently
appear
ed as
the larges
t methane
emitters
in the region
(
70
,
71
).
Using
TROPOMI
results,
Lauvaux
et al.
(
73
) identified
Turkmenis
tan,
Iran, and Kazakhs
tan as the larges
t methane
emitters
in Central
Asia.
Consistent with
these
prior
studies,
EMIT
results
indica
te
that Turkmenis
tan (731
± 148 tonnes
CH
4
hour
−
1
), Kazakhs
tan
(207
± 11 tonnes
CH
4
hour
−
1
), Iran (87 ± 48 tonnes
CH
4
hour
−
1
), and Uzbekis
tan (86 ± 22 tonnes
CH
4
hour
−
1
) represent
the larges
t oil and gas emitters
of those
countries
analyzed
in this
study. While
these
initial
EMIT
results
are promising,
future anal-
yses will use additional
observa
tions
to ensur
e that these
countries
are mapped
in their
entirety and revisited
to determine
emission
persis
tence.
Fig. 4. Clusters of methane
plumes
in Turkmenis
tan with
distinct
sources.
(
A
)
Twelv
e methane
sour
ces
with
an
aggr
ega
te
emission
of
163
±
18
tonnes
CH
4
hour
−
1
acquir
ed
at 4:28
UT
C
on
15
Augus
t 2022.
The
inla
y sho
ws
a repea
t mapping
of
this
region
on
the
same
da
y (10:58
UT
C)
with
visible
differ
ences
in
the
plume
shapes
associa
ted
with
changing
wind
conditions.
Infr
as
tructur
e loca
tions
are sho
wn
for
the
Goturdepe
(pink),
Barsagelmez
(gr
een),
and
Burun
(or
ange)
fields,
as
well
as
gas
(light
gr
ay)
and
oil
(light
blue)
pipelines.
(
B
)
Fifteen
sour
ces
all
within
the
Dauletabad-Donmez
field,
with
an
aggr
ega
te
emission
of
64
±
12
tonnes
CH
4
hour
−
1
.
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
6 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
Previous
studies
identified
a handful
of large
methane
sources in
Turkmenis
tan using
GHGSa
t (
16
),
Sentinel
2 (
21
),
and Worldvie
w
(
23
).
Lauvaux
et al.
(
73
) identified
118 point
sources, mostly above
10 tonnes
CH
4
hour
−
1
and associa
ted with
oil and gas activity
in
Turkmenis
tan using
TROPOMI
and its 5.5 km
–
by
–
7
km spatial res-
olution.
While
Irakulis-Loitxa
te
et al.
(
74
) used PRISMA
to quantify
methane
emissions
for oil and gas fields
near the Caspian
Sea, EMIT
imaged
64% of Turkmenis
tan in just its first 30 days of data collec-
tion.
From this initial
survey, we identified
65 emission
sources
almos
t entirely from the oil and gas sector,
ranging
from 0.3 to 60
tonnes
CH
4
hour
−
1
, that were attributed
to specific
source locations
and organized
by upstream (well pad and gathering
line),
mid-
stream (gas compr
essor
stations
and gas processing
plant),
and
downstream (fertilizer
plant)
categories.
The distribution
of methane
emission
rates observ
ed by EMIT
in
its first 30 days (0.3 to 73 tonnes
CH
4
hour
−
1
) is shown in Fig. 6A for
the regions
identified
in Fig. 5A. Compar
ed to airborne
results
(AVIRIS-NG
and GAO) across a number
of regions
in the United
States, the EMIT
distribution
is shifted
to larger
emissions.
This is
expected
both
because
EMIT
is observing
a class
of emissions
far
larger
than those
seen in U.S. airborne
surveys and because
EMIT
is less sensitiv
e to smaller
emission
rates due to the coarser
spatial
resolution.
However, there is substantial
overlap
in observ
ed emis-
sions.
For example,
the lowest emission
observ
ed with EMIT
was 0.3
tonnes
CH
4
hour
−
1
, and emissions
greater than this value
represent
60 to 85% of the total emissions
measur
ed with the airborne
surveys.
The total EMIT
detected
emissions
by sector
is shown in Fig. 6B,
with
88.7%
from oil and gas (63.5%
upstream,
24.3%
midstream,
and 0.9%
downstream).
Emissions
from the waste sector
represent
8.6%
of total observ
ed emissions,
including
examples
from 11 land-
fills and one wastewater treatment
facility
(Fig. 2C). For this study,
the landfill
emissions
ranged
from 2 to 28 tonnes
CH
4
hour
−
1
,
which
is broadly
consis
tent with previous
studies
that showed esti-
mates from four landfills
in Argentina
and India
that ranged
from 3
to 29 tonnes
CH
4
hour
−
1
. We note that the emission
sources detect-
ed were serendipitously
within
the EMIT
observa
tion mask.
More
complete
statistics for those
regions
will become
available
as more
data is evalua
ted and other
locations
can be targeted
during
the
EMIT
extended
mission.
DISCUSSION
EMIT
provides
important
contributions
through
the combina
tion
of large
area coverage and fine spatial resolution
imaging
spectr
o-
scopy
data, enabling
potential
future identifica
tion of both expected
(process-based)
and unexpected
(fugitiv
e) emissions.
Emissions
can
be quantified
and attributed
to specific
sectors,
which
is particularly
important
when
multiple
emissions
from differ
ent sectors
are
present
in close
proximity
(Fig. 3). We highlight
the first examples
of EMIT
imaging
spectr
ometer
observa
tions
of methane
and carbon
dioxide
emissions
from sources spanning
the oil and gas (upstream,
midstream,
and downstream),
waste (landfill
and wastewater treat-
ment),
and energy
sectors
(power plant).
We focused
our analy
sis on a number
of energy-pr
oducing
countries
in the Middle
East (Saudi
Arabia,
Iran, the United
Arab
Emirates, Iraq, Qatar, Kuwait, and Oman)
and Central Asia (Ka-
zakhs
tan, Turkmenis
tan, and Uzbekis
tan).
Oil and gas methane
emissions
represented
88.7%
of the total
(63.5%
upstream,
24.3%
Fig. 5. Location,
magnitude,
and emission
sector
for methane
plumes
observ
ed in the Middle
East
and Central Asia.
Dis
tribution
of
methane
emissions
for
se-
lected
countries
in
the
Middle
Eas
t and
Centr
al
Asia
with
consider
able
pr
oduction
of
oil,
gas,
and
coal.
The
magnitude
of
observ
ed
emissions
is
sho
wn
for
the
oil
and
gas,
was
te,
and
energy
sectors
(emission
magnitude
associa
ted
with
cir
cle
size
varies
by
sector).
EMIT
cloud-fr
ee
acquisitions
are sho
wn
in
white.
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
7 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
midstream,
and 0.9%
downstream),
followed by waste (8.6%)
and
energy
(2.2%).
Turkmenis
tan, Kazakhs
tan, Iran, and Uzbekis
tan
represent
the larges
t emitters
of those
countries
analyzed
in this
study, consis
tent with previous
studies
(
70
,
71
).
This study provides
a 30-da
y snapshot
of multina
tional
methane
emissions,
and the initial
set of EMIT
observa
tions
did not provide
100%
spatial coverage of individual
counties
or the repeat mapping
requir
ed to assess
persis
tence.
Despite
these
limita
tions,
initial
results
from EMIT
indica
te consider
able variability
in methane
emissions
at the regional
and country
scales
while
providing
in-
sights
into regions
with relatively large
emissions
and incomplete
activity
reporting.
Additional
observa
tions
with an average 50-da
y
repeat frequency
will permit
assessment
over time
to determine
emission
persis
tence
and reduce
uncertainty
in global
greenhouse
gas budgets.
During
the 1-year primary
mission,
more than
1600
power plants
(
≥
500-MW
capacity)
will be imaged
by EMIT
with
approxima
tely half observ
ed
≥
10 times.
In some
cases,
the ISS
orbit
enables
the additional
unique
capability
of assessing
emission
variability
two times
throughout
the day. In September
2023,
EMIT
should
complete
its primary
mission,
and an extended
mission
is
planned,
during
which
any regions
within
ISS latitude
constraints
could
be targeted.
Emissions
observ
ed by EMIT
from individual
plumes
ranged
between 0.3 and 73 tonnes
CH
4
hour
−
1
, and the 11 landfill
examples
presented
here had emissions
between 2 and 28 tonnes
CH
4
hour
−
1
.
The smalles
t emission
observ
ed in this study
was 0.3 tonnes
CH
4
hour
−
1
, and emissions
greater than this value
represent
60 to 85%
of the total
emissions
measur
ed with
the airborne
surveys in the
United
States. This
emphasizes
the potential
of EMIT
to map
large
regions
that are difficult
to access
with airborne
surveys. Anal-
ysis of additional
observa
tions
are required to determine
EMIT
’
s
detection
limit
for methane
and carbon
dioxide
point
sources,
which
has important
implica
tions
for unders
tanding
what emis-
sions
remain
undetected
relative to total
emissions.
Nevertheless,
EMIT
is almos
t two orders
of magnitude
more sensitiv
e than
TROPOMI
(minimum
detection
limit
of 10 tonnes
CH
4
hour
−
1
),
enabling
myriad
future studies.
In addition,
we demons
trated
EMIT
’
s ability
to identify
, geoloca
te, and quantify
multiple
emis-
sion sources (methane
and/or
carbon
dioxide)
that are in close
proximity
(<1 km) and to attribute
these
emissions
to differ
ent
sectors.
These
capabilities,
enabled
by the 60-m
spatial resolution,
are needed
to close
carbon
budgets.
Building
on the design
heritage
and capabilities
of the EMIT
in-
strument,
the JPL as part of the Carbon
Mapper
Coalition
is sup-
porting
the launch
of the first two satellites
equipped
with a JPL-
developed
imaging
spectr
ometer
in late 2023.
Those
instruments
will feature an improved 5-nm
spectr
al sampling
and finer
spatial
resolution
(30 m) and will be hosted on Planet
Labs
’
satellites,
of-
fering
target
tracking
(higher
effectiv
e SNR)
in noon-cr
ossing
sun-
synchr
onous
orbits,
resulting
in greater sensitivity
for methane
and
carbon
dioxide
point
source emissions.
Combining
measur
ements
obtained
from differ
ent instruments
improves global
coverage and
revisit
frequency
, which
is critical
to improving unders
tanding
of
global
emissions
and informing
mitiga
tion strategies.
To this end,
the EMIT
greenhouse
gas applica
tions
online
mapping
tool (https://
earth.jpl.nasa.go
v/data/data-portal/Gr
eenhouse-Gases)
will facili-
tate the distribution
of scientific
findings
in support
of NASA
’
s
Open
Source Science
Initia
tive.
MATERIALS
AND
METHODS
EMIT
imaging
spectr
ometer
The EMIT
is an imaging
spectr
ometer
that measur
es reflected
solar
radiation for 285 distinct wavelengths
from the visible
to shortw
ave
infrared (381
to 2493
nm) (
25
,
26
). This
pushbr
oom
instrument
builds
on the 30-year history of imaging
spectr
ometer
development
at the JPL and uses a large
conca
ve-shaped
blaze
grating fabrica
ted
at the JPL Microdevices
Labor
atory and a CHROMA
(Teledyne
Imaging
Sensors
Inc.)
HgCdT
e detector.
Its Dyson optical
design
enables
spatial uniformity
exceeding
98%,
and an F number
of
1.8. EMIT
has demons
trated exceptionally
high optical
throughput
performance,
with a spectr
al uniformity
greater than 98% (less than
0.2-nm
peak-to-peak
or 0.1-nm
peak-to-center
smile)
and an SNR
ranging
from 500 to 750 for most wavelengths
at the USGS
Libya 4
calibr
ation site (
55
). While
designed
to determine
the miner
al com-
position
of arid lands
responsible
for dust gener
ation,
the green-
house
gas mapping
capability
has been
demons
trated using
airborne
imaging
spectr
ometers
built at JPL (
28
–
30
,
33
–
40
,
42
–
46
).
EMIT
uses a 1242-pixel
element
swath at approxima
tely 60-m
image
pixel
resolution,
resulting
in an image
swath around
80 km
with an average daily
coverage of 1.3 × 10
6
km
2
. While
the primary
Fig. 6. Distribution
of methane
emissions
observ
ed with
EMIT
relative to previous
studies.
(
A
)
Hea
vy-tailed
dis
tribution
of
methane
emissions
for
results
sho
wn
in
Fig.
5 (Middle
Eas
t and
Centr
al
Asia)
fro
m
EMIT
compar
ed
to
pr
evious
airborne
campaign
results
made
available
thr
ough
the
Carbon
Mapper
open
da
ta
portal
(dat
a.
carbonmapper.org).
The
vertical
gr
ay line
indica
tes
the
smallest
EMIT
emission
(0.3
tonnes
CH
4
hour
−
1
) observ
ed
in
this
study
and
emissions
gr
ea
ter
than
this
value
repr
esent
60
to
85%
of
the
total
emissions
measur
ed
with
the
airborne
surv
eys.
(
B
)
EMIT
emissions
by
sector
as
determined
from
the
firs
t 30
da
ys of
da
ta
collection.
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
8 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025
mission
is focused
on coverage of arid lands,
EMIT
will map broad
regions
of Earth
’
s
surface between +51.6°
and
−
51.6°
latitude
(Fig.
1A), which
is dictated by the inclined,
equatorial
orbit
of the
ISS. The ISS orbit
coupled
with the large
daily
coverage, equivalent
to an area the size of the Republic
of South
Africa,
results
in a
median
revisit
frequency
of 36.5 days. This offers
the potential
to
track changes
in emissions
over time
(Fig.
1B), and in rare cases,
multiple
observa
tions
can
be obtained
within
the
same
day (Fig. 4A).
Methane
and carbon
dioxide
retrievals
Greenhouse
gas retrievals
using
airborne
imaging
spectr
ometers
were first demons
trated with
AVIRIS
(
75
) and AVIRIS-NG
(
45
).
These
algorithms
use absorption
spectr
oscopy
in the shortw
ave in-
frared to retriev
e gas concentr
ations
from solar
backsca
tter radiance
for each image
pixel
within
a scene.
At the 7.4-nm
EMIT
spectr
al
sampling,
methane
and carbon
dioxide
have distinct spectr
al finger-
prints,
as shown in fig. S1. In this study, we use a linearized
matched
filter
to calcula
te a mixing
ratio length
in units
of ppmm
(parts
per
million
meter),
representing
the thickness
and concentr
ation within
a volume
of equivalent
absorption
(
76
).
This
retrieval
has been
demons
trated on both
airborne
(
33
–
40
)
and spaceborne
data (
15
).
We used
radiative transfer
simula
tions
to gener
ate unit absorption
spectr
a (
38
) for methane
and carbon
dioxide
parameterized
for each
EMIT
scene
(path length,
viewing
geometry
, and atmospheric
water
vapor
modeled
by the primary
EMIT
mission),
which
are used
to
convert changes
in radiance
associa
ted with
gas absorptions
in
the units
of parts
per million
meter.
Methane
and carbon
dioxide
retrieval
results
were used to gener
ate the plume
imagery
presented
here and
for emission
attribution
and
quantifica
tion
de-
scribed
below.
Calcula
ting emissions
Emission
estimates are gener
ated by combining
the IME,
plume
length,
and wind
speed
as described
in previous
studies
(
40
).
The
IME was calcula
ted by applying
a mixing
ratio threshold
to separate
the plumes
from background
(500 ppmm
for methane
and 25,000
ppmm
for carbon
dioxide).
The plume
was then delinea
ted using
a
merge
distance
defined
in meters
to isolate contiguous
plume
ele-
ments
in the presence
of gaps (pixels
with low methane
or carbon
dioxide
values)
and a maximum
plume
fetch
(radius
in meters
from
the plume
origin).
Higher
mixing
ratio thresholds
and lower merge
distances
typically
result in plumes
that are well defined
and contig-
uous
but have shorter
plume
lengths.
An iterative assessment
of
these
parameters,
balancing
the need
for clear
delinea
tion
of
plumes
but limiting
the amount
of overlap
between plumes
that
are located in close
proximity
(Fig.
4), resulted
in a 200-m
merge
distance
and 1000-m
maximum
fetch.
For the plumes
in this
study, the average plume
length
used
for the emission
calcula
tion
was 861 m (163-m
SD).
The IME is calcula
ted as follows
IME
¼
k
X
n
i
¼
0
α
ð
i
Þ
S
ð
i
Þ
ð
1
Þ
by summing
the product
of α (the mixing
ratio length
in units
of
ppmm
methane
or carbon
dioxide)
multiplied
by the pixel
area
S
(in m
2
) for the
n
pixels
in the plume
and then
converting
to
methane
or carbon
dioxide
mass
units
using
a constant
k
. The
IME
(tonnes
of methane
or carbon
dioxide)
is combined
with
L
,
the plume
length
(in m), and 10-m
above surface wind
speed
(in
m s
−
1
) to estimate the emission
rate (in tonnes
hour
−
1
) using
the
following
equation
Q
¼
IME
U
10
L
ð
2
Þ
Wind speed
informa
tion was obtained
using
the ECMWF
ERA5
hourly
data (
77
). Using
the UTC times
tamp
associa
ted with the ac-
quisition
for a given EMIT
scene
coupled
with the latitude
and lon-
gitude
of the plume
origin,
10-m
u
and
v
components
were used to
calcula
te a mean
(
U
10
) and SD (σ
U
) wind
speed
(in m s
−
1
) for nine
grid cells (3 × 3 box) center
ed on the plume
origin.
The mean
10-m
wind
speed
(
U
10
) was used in the equation above, while
the uncer-
tainty
in the emissions
were calcula
ted using
the SD (σ
U
) of the
winds.
For emissions
presented
in this study,
U
10
ranged
from 0.9
to 9.5 m s
−
1
(4.3 m s
−
1
mean;
1.9 m s
−
1
for 1σ) and winds
associa
ted
with the lowest and highes
t emission
estimates were 3.9 and 6.3 m
s
−
1
. Using
AVIRIS-NG,
emission
estimates derived from the IME,
plume
length,
and wind
speed
have been valida
ted through
compar-
ison with independent
estimates obtained
using
mass
balance
ap-
proaches from in situ gas sampling
(
30
,
40
) and contr
olled
release
experiments
(
78
).
The choice
of mixing
ratio thresholds
and maximum
plume
fetch
will affect
the IME/
L
component
of Eq. 2 and,
ultima
tely,
the emission
rate. In Fig. 7A, the IME/
L
values
shown on the
x
axis represent
those
presented
in this study
(500-ppmm
mixing
ratio threshold
and 1000-m
maximum
fetch)
and those
in
y
axis
use 500-m
maximum
fetch.
The
use of the smaller
500-m
maximum
fetch
typically
results
in higher
IME/
L
values.
In
Fig. 7B, results
that are shown on the
x
axis represent
those
present-
ed in this study and those
in the
y
axis use a 750-ppmm
mixing
ratio
threshold,
indica
ting that IME/
L
values
for the 750-ppmm
mixing
ratio threshold
are slightly
lower. IME/
L
values
calcula
ted for differ-
ent CH
4
mixing
ratio threshold
and maximum
plume
fetch
for CH
4
plumes
are also plotted
in Fig. 8. Outlier
values
tend to be associa
ted
with the smaller
500-m
maximum
fetch
(green and orange
circles),
and the IME/
L
values
using
the 500-ppmm
mixing
ratio threshold
Fig. 7. IME/
L
values
calcula
ted for differ
ent CH
4
mixing
ratio threshold
and
maximum
plume
fetch.
(
A
)
Results
sho
wn
on
the
x
axis
repr
esent
those
pr
esented
in
this
study
using
500-ppmm
mixing
ratio
thr
eshold
and
1000-m
maximum
fetch
and
those
on
the
y
axis
using
500-m
maximum
fetch.
The
use
of
the
smaller
500-m
maximum
fetch
typically
results
in
higher
IME/
L
values.
(
B
)
Results
sho
wn
on
the
x
axis
repr
esent
those
pr
esented
in
this
study
using
500-ppmm
mixing
ratio
thr
esh-
old
and
1000-m
maximum
fetch
and
those
on
the
y
axis
using
750-ppmm
mixing
ratio
thr
eshold.
SCIENCE
ADV
ANCES
|
RESEARCH
ARTICLE
Thorpe
et al.
,
Sci. Adv.
9
,
eadh2391
(2023)
17
No
vember
2023
9 of 13
Downloaded from https://www.science.org at California Institute of Technology on February 03, 2025