1 NOVEMBER 2021

Bringing Radical
Transparency
to Global Emissions
How Climate TRACE is harnessing satellites and
artificial intelligence to advance emissions monitoring
through direct observation and open data
Climate TRACE (Tracking Real-time Atmospheric Carbon Emissions) is
a global coalition of nonprofits, tech companies, and universities created
to make meaningful climate action faster and easier by independently
tracking greenhouse gas (GHG) emissions with unprecedented detail and
speed. We harness satellite imagery and other forms of remote sensing,
artificial intelligence, and data science expertise to identify human-caused
GHG emissions when and where they happen. Coalition members
include Blue Sky Analytics, CarbonPlan, Earthrise Alliance, Hudson
Carbon, Hypervine, Johns Hopkins University Applied Physics Laboratory,
OceanMind, RMI, TransitionZero, WattTime, and climate leader and
former U.S. Vice President Al Gore.

For more information, please visit: ClimateTRACE.org

2
 TABLE OF CONTENTS

EXECUTIVE SUMMARY 1

INTRODUCTION 3

A CALL TO ACTION 4

AN URGENT NEED 6

HOW IT WORKS 8

TECHNOLOGY ROADMAP & EVOLUTION 11

DATA RECENCY 12

TIME-SERIES DATA GRANULARITY 16

ASSET-LEVEL SPATIAL GRANULARITY 18

ASSET DETECTION 20

ENDNOTES 22

Suggested Citation
Bringing Radical Transparency to Global Emissions: How Climate TRACE is harnessing satellites
and artificial intelligence to advance emissions monitoring through direct observation and
open data. Climate TRACE. 1 November 2021.

Contact: [email protected]

i
ii
 EXECUTIVE SUMMARY

Tracking greenhouse gas (GHG) emissions from Planned upcoming features and capabilities include:
nearly every major human-caused emitting activity
worldwide is an enormous undertaking. Climate • Data recency. We launched with data through
scientists now have a good understanding of how 2020, but for certain sectors, updated data will
much carbon dioxide is in the atmosphere. But integrate into the Climate TRACE dashboard to
that insight alone isn’t sufficient for informing show more recency, moving toward near-re-
climate action. al-time updates where possible. For example:

We need data to pinpoint when and where • Data from Climate TRACE member
human-caused emissions happen, and what WattTime show power sector emissions
sources are driving them. Data such as these rebounding in 2021 in places such as the
help pivot from accounting to action. Yet many European Union, United States, and India
countries must rely on general estimates, broad after 2020 declines due to COVID-related
assumptions, and emitters’ self-reported data. shutdowns.
Consequently, status quo GHG inventories are
often many years out of date, contain gaps, • Data from Climate TRACE member
have high levels of uncertainty, are high level TransitionZero show that in Q2 2021,
and not localised, and are fragmented and China’s steel-related emissions reached
not comprehensive. record levels, but Q3 saw a sharp drop
in the country’s steel-related emissions,
With recent advances in machine learning bringing it closer to meeting its target
and artificial intelligence (AI), and with greater of keeping 2021 emissions below
availability of remote sensing data such as satellite 2020 levels.
imagery from public and commercial sources, it
is now possible to harness these technological • Data on mining and quarrying from
advances to provide actionable emissions data to Climate TRACE member Hypervine show
all stakeholders and usher in a an era of emissions that China continues to be the world’s
monitoring that provides open, transparent, biggest emitter in the sector in 2021,
granular, timely data. with emissions 3.7x higher for Q1 than
for the EU.
The inaugural Climate TRACE dataset unveiled
in September 2021 covers annual emissions for • Data from Climate TRACE member Blue
every country on Earth across 10 sectors and 38 Sky Analytics show that Russia’s forest
subsectors for the years 2015–2020. From this fire emissions for 2021 YTD have already
baseline, Climate TRACE data will continue to get surpassed each of its record-setting 2019
better—rapidly—over time. Some of the biggest and 2020 totals. In fact, Russia’s total
developments coming down the Climate TRACE sector emissions for 2021 YTD are higher
tech pipeline include adding both temporal and than Brazil and the U.S. combined for
spatial granularity. that time period.

1
• Time-series data granularity. Typically, from Climate TRACE member CarbonPlan.
national emissions inventories are aggregated Knowing the locations of deforestation
annual emissions numbers. For many sectors, can highlight specific regions in a country
Climate TRACE will be able to document that are experiencing rapid deforestation
data on finer time scales, such as monthly or and the resulting CO2 released.
weekly emissions fluctuations from certain
sectors and individual assets. For example: • Asset detection. Measuring emissions from
specific assets is dependent on knowing those
• Data from Climate TRACE members assets exist in the first place. Yet in some
WattTime and TransitionZero for three instances, important sources of emissions
coal-fired power plants in India show remain undocumented—and therefore,
changes in emissions that resulted from uncounted in emissions inventories. Climate
the impact of major national events TRACE data will put these emissions sources
(such as India’s COVID-related lockdown on the map for the first time. For example:
and civil protests that disrupted power
plant operations).. • Concentrated animal feeding operations
(CAFOs) are a major source of methane
• Looking back at 2020, granular time-se- emissions. Attempts to regulate them
ries data from Climate TRACE member have largely failed, because knowing
OceanMind confirms a sudden, sharp how many of these facilities there are or
decline in maritime-related emissions as where they are located remains opaque.
pandemic lockdowns and travel restric- Climate TRACE modeling contributor
tions brought a temporary halt to the Synthetaic is using high-performance
cruise ship industry. computing, generative AI, and deep neural
networks to map CAFOs from satellite
• Asset-level spatial granularity. Although imagery.
accounting frameworks attribute emissions to
certain entities—whether countries or corpora- • Climate TRACE member Earthrise is
tions—in practice, all emissions come from supporting the Global Plastics Watch
specific sources: power plants, ships, factories, by the Minderoo Foundation to identify
etc. That’s why asset-level granularity and the collections of plastic waste from high-
ability to ‘zoom in’ on a specific source and resolution satellite imagery—documenting
its emissions are central to the tech roadmap. landfill area change over time and
For example: associated emissions..

• Data from Climate TRACE member RMI As a continually improving project developed by
show that two historic North Sea fields a global coalition, Climate TRACE welcomes feedback
in the widely traded Brent oil basket from experts and peer reviewers. In addition, we
have significantly different climate also invite new collaborators who have domain
footprints. The UK Brent field emits 25x expertise, access to high-quality ground truth
more kgCO2-equivalent per barrel oil data that could be used to train algorithms, or
equivalent than Norway Ekofisk. have expertise in sectors not yet covered by
Climate TRACE.
• As of 2020, Brazil leads the world in
forest-clearing emissions, based on data

2
 INTRODUCTION

Tracking greenhouse gas (GHG) emissions from sensing data such as satellite imagery from public
nearly every major human-caused emitting activity and commercial sources, it is now possible to
worldwide is an enormously difficult undertaking, harness these technological advances to provide
given the sheer diversity of emissions sources. actionable emissions data to all stakeholders.

Climate scientists have a good understanding of Climate TRACE—a global coalition of NGOs, tech
how much carbon dioxide is in the atmosphere. companies, and academia—was established to
But that insight alone isn’t sufficient for informing harness these advances in emissions monitoring,
climate action and emissions-reduction strategies. remote sensing, and AI / machine learning to bring
To address the climate crisis with the urgency radical transparency to global emissions. This
that’s needed, we need data to pinpoint when and brief provides an overview of how we estimate
where anthropogenic (human-caused) emissions emissions, who can use the data, and a preview
happen, and what sources are driving them. Data of our tech roadmap and upcoming data releases.
such as these help pivot from accounting to action.

Currently, the methods to answer those questions

in detail require the installation of expensive
sensors and extensive monitoring and oversight
capacity, which many governments, companies,
NGOs, and climate advocates lack. With recent
advances in machine learning and artificial intelli-
gence (AI), and with greater availability of remote

3
 A CALL TO ACTION
Why we need radical transparency
through open data

Parties to the United Nations Framework technology—a key source of timely, actionable,
Convention on Climate Change (UNFCCC) are trusted climate intelligence—has not kept pace with
required to submit national inventories.1 Countries climate ambition or action. Leveling the playing
designated as Annex 1 countries must submit field of emissions data is essential—especially
detailed descriptive and numerical information within the countries that are responsible for a
about their GHG emissions and removals. But higher rate of growth in emissions..
while most high-income countries have the
financial and capacity resources to produce annual
inventories, those inventories have shortcomings Challenges with status
and struggle with hard-to-measure sectors such
as land use change and landfills.
quo emissions inventories

Non-Annex 1 countries, meanwhile, have less-de- Many countries—Annex 1 and non-Annex 1 alike—
tailed reporting requirements in light of the must rely on general estimates, broad assumptions,
principle of ‘common but differentiated respon- and emitters’ self-reported data. Consequently,
sibilities’. This principle acknowledges that all status quo GHG inventories:
countries have a responsibility to reduce global
GHG emissions, but not all countries are equally • Are often many years out of date. At
responsible or equipped—financially or otherwise— present, more than 100 countries lack
to make large, swift emissions reductions or to updated emissions inventories for the years
produce the emissions inventories that document since 2015.3 These years-old inventories lack
them. Developing modern emissions invento- relevant information needed to inform new
ries can be challenging for many Non-Annex 1 policies or course-correct existing policies.
countries, as they often have limited resources and
systems in place for data collection.2 • Contain gaps. Many countries have a
large number of small- and medium-sized
Since the 2015 Paris Agreement, many high-emit- enterprises (SMEs), which are often excluded
ting countries have announced net-zero GHG from formal emissions-accounting and
targets. Outside of the international negotiation reporting processes. They may not have the
process, there has also been a groundswell of capacity to install monitoring sensors and so
action among non-state actors, as provincial / rely on self-reported estimates from industries
state and regional governments, cities, investors, with no mechanism to evaluate the veracity of
and businesses have all signaled their intention the claims.
to ramp up action. However, emissions monitoring

4
• Have high levels of uncertainty. Data on • Are fragmented and not comprehensive.
emissions-causing activity can be difficult Although there are many credible, non-gov-
to identify in countries that do not have a ernment sources of GHG data available, most
robust data repository system. For sectors are limited in scope, providing data for a
such as agriculture, activity data can be specific sector or subset of countries only.
particularly challenging, as agriculture surveys For example, difficult-to-measure sectors like
are conducted only sporadically in many land use change and agriculture are often
countries. Even in Annex 1 countries, the excluded from major inventories. Data that
agriculture, land use, and waste sectors have are disaggregated at the sub-sectoral level for
high levels of uncertainty due to the inherent all countries—such as parsing the manufac-
complexity of measuring their emissions. This turing sector into subsectors such as steel,
is because the emissions-causing activities in aluminum, cement, and pulp paper—would
these sectors are highly fragmented and total not only help policymakers meet transparency
activity data can only be based on sampled and verification requirements, but also provide
data rather than a comprehensive accounting. actionable insights.
In addition, there are many other natural
factors in play, such as the effect of carbon Despite these shortcomings, existing emissions
sinks and localised differences in emissions inventories have established a reliable baseline
(such as the difference in emissions between understanding of global GHG emissions. But
livestock of different breeds, or carbon stored current emissions data gaps underscore important
in different types of biomass), which can make opportunities to advance the state of the art and
emissions estimation complex. build upon the existing tapestry of GHG inventories
to usher in an era of radical transparency that
• Are high level and not localised. For ​​ provides open, granular, timely data that makes
many provinces / states and regions around bigger, better climate action easier.
the world, complete and up-to-date GHG
emissions data are low in quality or lacking
entirely, creating a barrier for many—
especially provinces / states, regions, and
cities in the Global South—to show leadership
in their climate action. Sub-national govern-
ments are now increasingly being recognised
as one of the most important stakeholders for
climate action, as many decisions are made at
the provincial / state or city level. Two of the
most significant barriers to provinces / states
and regions accessing regional GHG emissions
data are time and resource constraints.
Creating an annual GHG emissions inventory
via a bottom-up approach is time and
resource intensive, taking between six months
and two years to develop and often requiring
the involvement of many stakeholders across
different departments and organisations.

5
 AN URGENT NEED
How radical transparency can enable
actionable emissions insights

Meeting targets. Outdated GHG data can be a between global GHG emissions and the emissions
significant hurdle to policymakers, advocacy groups, reductions required to achieve a 1.5ºC climate
and corporations trying to reduce emissions. pathway.4
Without the latest available measurements,
emissions-reduction goals may not be sufficiently Enabling regional action. Provinces / states and
ambitious, emissions-reduction activities may regions are key players on the path to global
not be enough to meet stated targets, and/or net-zero emissions. They have the power to
emissions-reduction strategies may prioritize the implement sector-wide policy change within their
wrong sectors. For instance, the latest available jurisdictions, with large-scale emissions-reduc-
country data with the UNFCCC is based on data tion impacts. In fact, if provinces / states and
collected in the year 2019 or earlier (and this is regions from 10 major-emitting economies around
only available for a subset of countries). Also, many the world fully implemented all of their existing
data collection activities tend to be one-off efforts commitments, they could mitigate an additional 1.2
due to time or funding constraints. Climate TRACE GtCO2e/year in 2030 beyond national government
aims to address this by providing data that are policies, according to analysis from the NewClimate
available in real-time (or as nearly so as practi- Institute.5
cable for a given sector) and which will be updated
regularly. This could help facilitate cap-and-trade Supporting global equity. Empowering Global
markets, supply chain monitoring, carbon border South countries, provinces / states, and regions is
adjustments, and other market mechanisms to important to put them on equal footing, with open
place an economic value on carbon. access to high-quality emissions data. Enabling the
Global South to plan and implement high-quality
Setting new goals. Up-to-date and reliable data mitigation strategies can benefit their citizens and
can not only help stakeholders achieve their advance national and global climate ambition.
existing climate commitments, but also provide The granular and transparent data Climate TRACE
pathways to more ambitious action. In the provides will be important to encourage private-
international negotiation space, it may encourage sector support for climate solutions in developing
countries that have not yet announced enhanced countries.
Nationally Determined Contributions (NDCs) to
do so. This is crucially important, as the UNEP Supply chain decarbonisation. There is increasing
Emissions Gap Report from October 2021 showed pressure on companies with multinational supply
that with countries’ current NDCs, in 2030 there chains to decarbonize emissions from their supply
remains a substantial 28 GtCO2e annual gap chains and address Scope 3 emissions.6 However,
due to the opacity of these transactions and

6
the complexity of global supply chains, many
companies are unable to quantify emissions from
key sectors such as land use change, shipping, TYPES OF EMISSIONS
and mining—particularly if the emissions occur in
INVENTORIES
non-Annex 1 countries.

Sustainable / climate finance. Reliable environ-

mental, social, and governance (ESG) data is critical
to mobilizing funds for climate action, particularly National.
from the private sector. Reliable, third-party data (e.g., countries’ Biennial Update
on whether portfolio companies are meeting Reports (BURs) to the UNFCCC)
emissions-reduction targets can help significantly
build trust in such market instruments. This is Sub-national.
particularly important in developing countries, (i.e., regional, provincial / state,
which are facing the biggest climate finance gap. or city GHG inventories)
Some of the major issues with ESG data are that:
company-level data are only available for purchase Industry-specific.
from data providers and represent a significant (e.g., World Steel, International
cost to investors that could otherwise be funding Maritime Organization, International
more mitigation action. Such data may also be Civil Aviation Organization)
unaffordable to some investors. In addition, many
currently-available datasets are overly reliant Company-specific.
on voluntarily self-reported data, resulting in (e.g., Carbon Disclosure Project)
incomplete data that does not cover all parts of
the world. Independent, asset-level data that is NGO / third-party.
freely available in all parts of the world could be a (e.g., EDGAR, PIK-Potsdam,
gamechanger for climate finance. Climate TRACE’s Carbon Monitor, GHG Protocol,
eventual goal is to provide global asset-level data Climate TRACE)
to help close this gap.

Support local action & advocacy. Transparency

and public availability of data in all parts of the
world can help local organisations and citizens
better identify where emissions come from
to advocate for stronger regulations. Currently,
grassroots activists in many parts of the world
are unable to hold governments and asset owners
accountable due to lack of reliable data; the
use of satellite imagery and other sources to
independently determine emissions can support
their efforts and level the playing field.

7
 HOW IT WORKS
Understanding how Climate TRACE harnesses
satellites and artificial intelligence

For decades, climate scientists have been able But tracing new GHG emissions to particular
to measure Earth’s total level of atmospheric sources has proven far more difficult, yet is
GHG concentrations. Since 1958, the Mauna Loa crucial for climate action.
Observatory has provided the measurements
behind the Scripps Institute of Oceanography’s In-situ surface measurements provide detailed
famous Keeling Curve and the UK Met Office’s and highly accurate GHG measurements and can
atmospheric CO2 forecasts.7 More recently, Japan’s be used to attribute emissions to particular
Greenhouse Gases Observing Satellite (GOSAT),8 sources. However, this level of detail is not feasible
the United States’ Orbiting Carbon Observatory-2 for every country, every sector, or every asset
(OCO-2),9 and the EU’s Copernicus Atmosphere and/or emissions source. Cost and installation
Monitoring Service (CAMS)10 have all provided ‘heat logistics​​are limiting factors. As a result, this
map’ visualizations of atmospheric CO2 concentra- approach is sparsely deployed globally. Sensors
tions around the world (Figure 1). tend to be located in easily accessible regions
of developed countries (leaving ‘blind spots’
elsewhere around the globe).

FIGURE 1
Understanding total atmospheric carbon dioxide concentrations

Historical ice-core data plus readings from the Mauna Loa observatory since 1958 document Earth’s atmospheric carbon dioxide
concentration (left). Modern visualizations based on satellite data further enable ‘heat maps’ of atmospheric carbon dioxide
concentrations. But identifying the sources of new human-caused GHG emissions and separating them from the background noise
of baseline carbon dioxide concentrations has been a challenge.

Image sources: Scripps Institution of Oceanography (left). OCO-2 (right).

8
The aforementioned GHG monitoring satellites • Collect satellite imagery. Climate TRACE
can measure atmospheric concentrations of GHGs doesn’t launch, own, or operate our own
with far better global coverage, but CO2 has a satellites. Instead, we use different types of
long lifespan in the atmosphere. With a baseline satellite imagery collected by various space
concentration now above 400 ppm, any new CO2 agencies and organizations from around the
emissions from human sources is difficult for these world—both public and commercial (Figure
satellites to pinpoint. 2). That satellite imagery, which observes
emissions-causing activities around the world,
Many companies and NGOs are now launching includes both visible spectrum and other
satellites dedicated to methane monitoring, spectrums (e.g., infrared).
but CO2—the largest source of global warming
pollution—is difficult for these satellites to measure • Acquire additional remote-sensed, ground
given the background concentration. This creates truth, and other data. We don’t stop with
challenges in identifying anthropogenic CO2 and satellite imagery. We also collect historical
other GHGs, which requires a high degree of and contemporaneous data from ground- and
accuracy in order to detect and identify specific sea-based physical emissions sensors, govern-
emissions sources and new or ongoing emissions. ment sources, corporate disclosure forms,
Thus, there is not yet any unified global GHG Satellite Automatic Identification System
monitoring system in place. (S-AIS) maritime vessel locations, and
other sources.
In recent years, the launch of numerous satellites
by both public and commercial entities and the • Train the AI algorithms. Next, the AI and
resultant falling costs for satellite imagery have machine learning software algorithms go to
made such data more accessible and actionable work. They ‘train’ on all that data—59 trillion
for GHG monitoring. This, along with Big Data plus bytes and counting—to reliably and accurately
advancements in artificial intelligence (AI) and spot indicators of human-caused GHG
machine learning, have made it more and more emissions from an ongoing stream of satellite
common for myriad organisations to develop the imagery. Applied over time and around the
essential building blocks of such a monitoring world across all major GHG-emitting sectors of
system. Thus, crucially, today many of the essential the economy, Climate TRACE’s approach gives
building blocks for global emissions monitoring all stakeholders unprecedented new insight
systems already exist, and are simply spread out into where assets are located, when emissions-
amongst dozens of different organizations. causing activities are happening, and how
much emissions result from those activities.
That’s why we created Climate TRACE: to provide
a more independent, empirical, and objective • Get continuously better over time—fast.
system that can trace emissions to their sources We’re constantly seeking new collaborators
in near-real time could be a gamechanger. Climate and data sources. As our AI and machine-
TRACE focuses on bringing together and making learning software trains on more data from
the best possible use of satellite-based as well as more sources, it gets better and better. This
remote-sensing and other data, then analysing it allows the Climate TRACE emissions inventory
with AI and machine learning to bring the world to update and improve faster than legacy
a new set of emissions intelligence. Here’s an approaches to emissions tracking.
overview of how we do it:

9
 FIGURE 2
Training AI/ML to recognize emissions-
causing activities from satellite imagery

Visible plumes from Sentinel-2 optical

imagery indicate whether a coal-fired
power plant is generating at a given
time.

2017-10-05 2018-02-22

2018-04-03 2018-05-28

2019-08-11

10
TECHNOLOGY ROADMAP
& EVOLUTION
Climate TRACE’s capabilities will continue
improving rapidly over time

The inaugural Climate TRACE dataset unveiled corporations—in practice, all emissions come
in September 2021 covers annual emissions for from specific sources: power plants, ships,
every country on Earth in 10 sectors and 38 factories, airplanes, rice paddies, etc. That’s
subsectors for the years 2015–2020. From this why asset-level granularity and the ability
baseline, Climate TRACE data will continue to get to ‘zoom in’ on a specific source and its
better—rapidly—over time. Some of the biggest emissions are central to the Climate TRACE
developments coming down the Climate TRACE tech pipeline.
tech pipeline include adding both temporal and
spatial granularity. • Asset detection. Naturally, measuring
emissions from specific assets is dependent
Planned upcoming features and capabilities include: on knowing those assets exist in the first
place. Yet in some instances, important
• Data recency. We launched with data through sources of emissions remain undocumented—
2020, but for certain sectors, updated data will and therefore, uncounted in emissions
integrate into the Climate TRACE dashboard inventories. Climate TRACE data will put
with more and more recency, moving toward these emissions sources on the map for
near-real-time updates when possible. the first time.

• Time-series data granularity. Typically, These features and capabilities are coming to
national emissions inventories—as well as Climate TRACE in the near future—sooner than
our inaugural dataset—are aggregated annual later for certain sectors. In the meantime, explore
emissions numbers. For many sectors, Climate new insights and catch a glimpse of what’s
TRACE will be able to ‘zoom in’ to inspect the possible with satellites and AI in the remainder
data on finer time scales, such as monthly of this section.
or weekly GHG emissions fluctuations from
certain sectors and assets.

• Asset-level spatial granularity. Although

frameworks such as the Paris Agreement
and the GHG Protocol attribute emissions
to certain entities—whether countries or

11
DATA RECENCY
2021 YTD emissions through September for
the power sector (electricity)

Data from Climate TRACE member WattTime show

an overall trend of emissions rebounding in 2021
in places such as the EU (Figure 3), USA (Figure 4),
and India (Figure 5), after declining in 2020 due to
COVID-related shutdowns.

FIGURE 3
European Union power sector emissions 2021 YTD vs. 2020

Data for the EU show higher cumulative YTD emissions for 2021 vs. 2020, with year-over-year emissions for the month of April
30% higher than the same period in 2020. In fact, in 2021 monthly emissions from electricity generation have been higher in every
month except August, perhaps due in part to re-opening of many EU economies after the COVID-19 lockdowns of 2020.

12
FIGURE 4
United States power sector emissions 2021 YTD vs. 2020

Meanwhile in the United States, as expected, power sector emissions from electricity generation again peaked during July and
August. Monthly emissions YTD for 2021 have been consistently higher than 2020—again, partially attributable to pandemic-
related economic rebound. February 2021 is especially notable, where emissions were more than 50% higher than the same
month in 2020.

FIGURE 5
India power sector emissions 2021 YTD vs. 2020

Finally, monthly and cumulative YTD power sector emissions in India have also been higher in 2021 vs. 2020. March and April
especially stand out, with 2021 emissions ~30% and 50% higher than in 2020, respectively.

13
Documenting the rise—and fall—of China’s emissions monitoring that shows changes more
steel industry emissions quickly and can encourage government policy
changes to rein in emissions.
China’s steel industry has been consistently
producing and emitting more every year since 2015, Emissions from mining and quarrying
according to data from Climate TRACE member
TransitionZero. In Q2 2021, steel-related emissions Data on mining and quarrying from Climate TRACE
in the country reached record levels and were on member Hypervine show that China continues
track for year-over-year emissions increases equal to be the world’s biggest emitter in the sector in
to the entire annual emissions of Singapore. But Q3 2021, with emissions for the first quarter of the
saw a sharp drop in China’s steel-related emissions year 3.7x higher than for the EU. In this sector
(Figure 6), bringing the country much closer to the EU emitted 10.33% more in the second quarter
meeting its target of keeping 2021 emissions compared to the first quarter of 2021. The United
below 2020 levels. This swing in emissions trends States also increased its mining and quarrying
from Q2 to Q3 highlights the need for real-time emissions by 8.40% in Q2 vs. Q1 2021.

FIGURE 6
China’s quarterly steel emissions 2015 through 2021 YTD

14
Tracking emissions from the 2021 Northern by satellites began (Figure 7).11 As in many other
Hemisphere fire season countries, due to increasing air temperatures, along
with increasing drought and heat-wave frequency,
Data from Climate TRACE member Blue Sky fire seasons are lasting longer than usual. May to
Analytics show that Russia’s forest fire emissions August is generally considered “peak” fire season
for 2021 YTD (through end of September but recent years indicate a longer fire season.
2021)—898 million tonnes CO2 equivalent (tCO2eq)— The United States’ fire emissions are abnormally
have already surpassed 2019 and 2020 totals by high this year as well, based on our estimates
nearly 50% each year. 2019 and 2020 were both for 2021 YTD. While initial U.S. fire emissions for
record-setting years for forest fire emissions, 2021 YTD have not yet passed 2020 totals, if fires
releasing 637 million and 615 million tonnes, persist on the country’s West Coast, emissions may
respectively. In fact, Russia’s total for 2021 YTD is end up higher than 2020. The 2021 YTD total is
higher than Brazil (439 million tCO2eq) and the also a significant increase from totals for the years
U.S. (344 million tCO2eq) combined for that time 2015–2019 (which ranged from 118 to 280 million
period. Climate TRACE’s findings are consistent with tCO2eq per year).
reports suggesting that 2021 may see the highest
fire emissions for Russia since record keeping

FIGURE 7
Documenting forest fire emissions in
Russia during 2021

An example of forest fires in the

Pyshminsky District of Russia. Top
image, Sentinel-2 natural-color image
of the smoke plumes. Bottom image,
Sentinel-2 false-color image incorpo-
rating the short-wave infrared measure-
ments, which reveal the fire beneath the
smoke plumes (bright red area in image).
Date of observation: 2021-08-22.

15
TIME-SERIES DATA 2020, there was another drop in generation when

GRANULARITY farmers’ protests disrupted rail freight and the

supply of electricity.12

Watching emissions fluctuate at three Indian Some plants were already experiencing reduced
power plants capacity factors from low demand before the
COVID lockdowns. The Indira Gandhi Super Thermal
Data from Climate TRACE members WattTime and Power Project, constructed in 2010, operated at 33%
TransitionZero for three coal-fired power plants in capacity in 2019 owing to low demand. According
India show the impact of major national events, to state dispatch records,13 at least one of the three
which are immediately apparent in the granular units at Indira Gandhi was shut down due to low
time-series emissions data (Figure 8). India demand for 206 days in 2019. When coal plants
declared a nationwide lockdown in response to the operate at low capacity, their cost of energy is
COVID-19 pandemic on March 24, 2020. Demand higher, making them less competitive in markets
for electricity dropped and the Talwandi Sabo and against other forms of energy, which may mean
Rajpura plants reduced their generation to almost they will be retired earlier than planned. Across
zero. These coal plants may have been some all of India, coal plant capacity factors are low,
of the first in line to reduce power generation, meaning that new renewable energy could begin
because Indian electricity dispatch rules mean that to replace existing coal fired power plants.
hydro and renewable energy are given priority in
supplying energy to the grid. Then in November

FIGURE 8
Watching emissions fluctuate at three coal-fired power plants in India

16
How pandemic travel restrictions saw cruise But as we look back at 2020, granular time-series
ship emissions plummet in 2020 data confirms a sudden, sharp decline in related
emissions as pandemic-related lockdowns and
For the maritime sector, after two straight years of travel restrictions brought a temporary halt to the
nearly 10% annual increase in emissions, through cruise ship industry (Figure 9).
three quarters 2021 we see about a 7% decline in
emissions compared to 2020, according to data
from Climate TRACE member OceanMind. This is a
broad-based trend, evident in small declines for all
of the top 10 countries with the highest shipping
emissions. This may be caused by widely noted
port congestion that has ships anchored outside
ports and facing substantial delays while they wait
for port service. If this is the reason for the decline
in emissions, the emissions reduction will not
persist, as the delays are creating pent up demand
for shipping so they are deferred emissions that
will occur later.

FIGURE 9
Maritime emissions from cruise ships fell dramatically in 2020 in response to COVID-related travel restrictions

17
ASSET-LEVEL SPATIAL RMI for the North Sea, for example. Two of the

GRANULARITY historic fields in the widely traded Brent

oil basket—Norway Ekofisk and UK Brent—have
significantly different climate footprints (Figure 10).
Bringing climate differentiation to upstream The UK Brent field, which is in the process of
emissions in oil & gas production and refining being decommissioned, emits over 25 times more
kilograms CO2-equivalent per barrel oil equivalent
I​n 2021, the United States and Russia are once on a 100-year warming basis than Norway Ekofisk.
again projected to be the world’s largest oil and gas This climate fate has no bearing on today’s market
production emitters. But it also pays to look deeper because the Brent basket sells at a single price
into this country-level oil and gas emissions data. that does not reflect these emissions differences.
On a global scale, the emissions difference between
the lowest- and highest-emitting oil and gas fields Tracking deforestation in Brazil
(on a per-barrel oil equivalent basis) can differ by
an order of magnitude or more. Knowing which Over the period 2015–2020, Brazil led the world in
fields present the greatest climate challenges can forest-clearing emissions, 2.98BT of CO2 (excluding
help target countries’ climate stabilization efforts. carbon sinks), based on Climate TRACE data.
Even within a country, the climate intensity of oil Further, knowing the location of deforestation can
and gas can vary markedly. In the United States, for highlight specific regions in a country that are
example, a barrel of oil from California can be 10 experiencing rapid deforestation, something that is
times dirtier than a barrel of low-leak light oil achievable with Climate TRACE data. For example,
from Texas. Climate TRACE coalition member CarbonPlan is able
to identify locations experiencing forest losses and
Differentiating oil and gas assets offers essential quantifying the resulting carbon that was removed
intelligence for climate policymakers—as well as from the forest and thus is committed to being
the financial community, NGOs, industry, and the emitted as atmospheric CO2 (Figures 11 and 12).
public. Consider data from Climate TRACE member
Beyond this example, the forest clearing dataset
accounts for other non-fire stand-replacing distur-
FIGURE 10
Vastly different production and refining emissions associated bances like flooding/ghost forests, permafrost
with two fields in the Brent oil basket degradation, insect/drought-induced die-offs and
landslides. In the near future, CarbonPlan aims to
identify and attribute the causes of deforestation—
such as timber harvesting and the conversion of
forest to pasture or cropland—and the resulting
committed emissions from such activities.

18
 FIGURE 11
Tracking forest losses and associated
committed carbon emissions

The green box highlights a specific

deforested area in 2020 where approxi-
mately 1,000 ha were lost that released
~9 million tonnes of CO2.

Top left: Landsat-8 image (from

2020-08-04) of 2020 deforestation
occurring near the Serra do Pardo
National Park region, located in the state
of Pará, Brazil (red square in the inset).

Top right: the same Landsat-8 image

with CO2 emissions caused by deforesta-
tion (inset legend indicates tonnes of
committed CO2 emissions per hectare).

Bottom left: close-up of 2020 deforesta-

tion occurring near the Serra do Pardo
National Park region, highlighting
previous year’s and 2020 deforested
areas.

Bottom right: same close-up with CO2

emissions caused by deforestation (inset
legend indicates tonnes of committed
CO2 emissions per hectare).

FIGURE 12
Planet imagery highlighting the
progression of deforestation

Top left: date: 2020-01-01, intact forest

in green box; Top right: 2020-06-29,
significant amount of deforestation
in green box; Bottom: 2020-09-29,
deforestation expands northward in
green box.

19
ASSET DETECTION
Putting concentrated animal feeding
operations (CAFOs) on the map

There are tens of thousands of concentrated

animal feeding operation (CAFO) facilities in
the U.S. alone and tens of thousands more around
the world. Attempts to regulate them have largely
failed due to difficulties in determining how many
of these facilities there are or where they are
located. CAFOs are a major source of methane
emissions and as an industry, have not yet set
any targets for methane emissions reductions.
The global factory-farm system driven by
corporate meat and dairy giants has largely
evaded climate oversight.

AI company and Climate TRACE modeling

contributor Synthetaic is using high-performance
computing, generative AI, and deep neural networks
to map CAFOs from satellite imagery (Figure 13)—
to be used for regional and international
methane reporting.

FIGURE 13
Identifying CAFO locations using satellite
imagery and AI

Top: A heatmap showing CAFO locations

and density in the Texas panhandle,
based on AI processing of satellite
imagery. Bottom left: an individual CAFO
location in north Texas. Bottom right:
the same locations with heatmap results
of the AI processing trying to find the
extent of the CAFO.

20
Documenting new and expanded landfills

Climate TRACE member Earthrise is supporting the

Global Plastics Watch by the Minderoo Foundation to
identify collections of plastic waste from high-res-
olution satellite imagery (Figure 14). Earthrise used
Sentinel-2 multispectral data to build a profile of
known and geolocated landfills. The profiles are
then used to train a collection of three different
neural networks to classify a patch as waste versus
non-waste. The candidate detections are validated
by humans, using optical, high-resolution imagery,
Street View, and citizen science contributions. The
confirmed sites are then monitored on a monthly
basis to measure the changing footprint of the site
from June 2016 through the present. Finally, the sites
are cross-referenced with open data on infrastruc-
ture and geophysical attributes, like soil type, slope,
distance to waterways, and roads, to support policy
response to recycling and waste management.
Ultimately, these data make it possible to track
associated emissions.

FIGURE 14
Documenting landfills and their associ-
ated emissions

Top: The Minderoo Foundation’s Global

Plastic Watch dashboard powered by
Earthrise identifies nearly 900 landfill
sites across Indonesia. Bottom: Users
can zoom in on individual sites and
view their change over time, including
associated emissions.

21
 ENDNOTES

1 U
 nited Nations Framework Convention 8 Greenhouse Gases Observing Satellite (GOSAT).
on Climate Change. “National Inventory http://www.gosat.nies.go.jp/en/
Submissions 2021.” https://unfccc.int/
ghg-inventories-annex-i-parties/2021 9 NASA Jet Propulsion Laboratory, California Institute
of Technology. Orbiting Carbon Observatory-2
2 S
 tephen M Ogle et al. “Advancing national (OCO-2). https://ocov2.jpl.nasa.gov/
greenhouse gas inventories for agriculture in
developing countries: improving activity data, 10 C
 opernicus Atmosphere Monitoring Service. “New
emission factors and software technology.” high-quality CAMS maps of carbon dioxide surface
Environmental Research Letters. 7 March fluxes obtained from satellite observations.” 10
2013. 8 015030. https://iopscience.iop.org/ December 2019. https://atmosphere.copernicus.eu/
article/10.1088/1748-9326/8/1/015030 new-high-quality-cams-maps-carbon-dioxide-sur-
face-fluxes-obtained-satellite-observations
3 C
 limate TRACE. Filling the Post-Paris Gap. 6 October
2021. https://medium.com/climate-trace-the-source/ 11  reenpeace Russia. “2021 Fire Record.” 17 September
G
filling-the-post-paris-gap-8cd9a1ee3eeb 2021. https://greenpeace.ru/blogs/2021/09/17/2021-
god-polnyj-rekordsmen-po-pozharam/
4 United
 Nations Environment Programme. Emissions
Gap Report 2021. 26 October 2021. https://www.unep. 12 Sravasti Dasgupta. “Why farmers’ protest against farm
org/resources/emissions-gap-report-2021 laws has led to a power crisis in power-surplus Pun-
jab.” The Print. 8 November 2020. https://theprint.in/
5 N
 ewClimate Institute, Data-Driven EnviroLab, Utrecht india/governance/why-farmers-protest-against-farm-
University, German Development Institute/Deutsches laws-has-led-to-a-power-crisis-in-power-surplus-
Institut für Entwicklungspolitik (DIE), CDP, Blavatnik punjab/539556/
School of Government, University of Oxford. Global
climate action from cities, regions and businesses. 13 Government of India, Ministry of Power. “MERIT.”
2021 edition. https://newclimate.org/2021/06/23/ https://meritindia.in/
global-climate-action-from-cities-regions-and-busi-
nesses-2021/

6 M
 cKinsey & Company. Making supply-
chain decarbonization happen. June
14, 2021. https://www.mckinsey.com/
business-functions/operations/our-insights/
making-supply-chain-decarbonization-happen

7 M
 et Office. “Atmospheric carbon dioxide
at record high levels despite reduced
emissions in 2020.” 7 April 2021. https://
www.metoffice.gov.uk/research/news/2021/
record-co2-levels-despite-lower-emissions-in-2020

22
 What makes Climate TRACE different?
With a diverse and growing body of emissions inventories, Climate TRACE
complements and enhances existing climate intelligence
in new and different ways:

UNIVERSAL ACTIONABLE
With global coverage and data harmonization Climate TRACE data is public, open,
across sectors and subsectors that enables and accessible—with ‘one-stop shopping’
‘apples-to-apples’ comparisons, Climate TRACE for emissions intelligence via a user-friendly
aims to remedy the issue of inconsistencies website UI and forthcoming public API. Climate
across sectors, methodologies, and geogra- TRACE aims to eventually move toward
phies. In addition, a global dataset levels the near-real-time updates for certain sectors,
playing field and equally supports and with increasing granularity for informed
Annex 1 and non-Annex 1 countries. emissions-reduction strategies
and climate action.

EMPIRICAL COLLABORATIVE
Climate TRACE focuses primarily on direct, Climate TRACE has partnered with numerous
objective observation of emissions-causing contributors, partners, and data providers.
activities at their source, which are not Further, dozens of peer reviewers from around
dependent upon legacy methods and their the world have reviewed our sector-spe-
potential shortcomings (e.g., self-reporting cific methodologies to bring additional
bias, higher levels of uncertainty, coverage scientific rigor to our approach. We continue
gaps). This acts as both a set of checks and to proactively invite further contributors
balances and enhances and adds confidence to expand the Climate TRACE network and
to the global body of emissions intelligence. strengthen our collective work.

START
START USING
USING THE
THE DATA
DATA BECOME A CONTRIBUTOR

Explore the emissions inventory Potential providers of satellite imagery,

at climatetrace.org/inventory ground-truth data, domain expertise,
peer reviewers, etc. are invited to contact
[email protected]
For more information, please visit
ClimateTRACE.org

24