Nature | Vol 633 | 26 September 2024 |
835
Article
Carbon emissions from the 2023 Canadian
wildfires
Brendan Byrne
1
✉
, Junjie
Liu
1,2
, Kevin W. Bowman
1,3
, Madeleine
Pascolini-Campbell
1
,
Abhishek Chatterjee
1
, Sudhanshu
Pandey
1
, Kazuyuki
Miyazaki
1
, Guido R. van der Werf
4
,
Debra Wunch
5
, Paul O. Wennberg
2,6
, Coleen M. Roehl
2
& Saptarshi
Sinha
7
The 2023 Canadian forest fires have been extreme in scale and intensity with more
than seven times the average annual area burned compared to the previous four
decades
1
. Here, we quantify the carbon emissions from these fires from May to
September 2023 on the basis of inverse modelling of satellite carbon monoxide
observations. We find that the magnitude of the carbon emissions is 647 TgC
(570–727 TgC), comparable to the annual fossil fuel emissions of large nations,
with only India, China and the USA releasing more carbon per year
2
. We find that
widespread hot–dry weather was a principal driver of fire spread, with 2023 being
the warmest and driest year since at least 1980
3
. Although temperatures were extreme
relative to the historical record, climate projections indicate that these temperatures
are likely to be typical during the 2050s, even under a moderate climate mitigation
scenario (shared socioeconomic pathway, SSP 2–4.5)
4
. Such conditions are likely to
drive increased fire activity and suppress carbon uptake by Canadian forests, adding
to concerns about the long-term durability of these forests as a carbon sink
5
–
8
.
Canadian forests cover a vast area of nearly 362 million ha (ref.
9
),
amounting to 8.5% of the global forested area
10
. These forests are an
important sink of carbon, absorbing fossil carbon dioxide (CO
2
) from
the atmosphere and slowing the pace of climate warming
11
,
12
. However,
climate change is increasing forest fire activity, acting to suppress the
carbon uptake capacity of these forests
13
. Although more frequent fires
have been widespread, 2023 has seen forest fires on an extreme scale.
With 15 million ha of Canadian forests burned (about 4% of forest area)
1
,
2023 saw more than seven times (8
σ
) the average burned area over the
preceding 40 years (1983–2022 mean, 2.2 million ha; range, 0.2–7.1 mil
-
lion ha)
1
. The adverse societal impacts of these fires are clear: 232,000
evacuations and poor air quality affecting millions
14
. However, the
carbon emissions from the fire events remain uncertain. In this study,
we quantify these emissions through inverse modelling of satellite
observations of carbon monoxide (CO). Then, we examine concurrent
climate anomalies and projected changes in the prevalence of hot–dry
weather under climate change. Finally, we discuss the implications of
fires for the Canadian carbon budget.
Fire emissions
Fire carbon emissions can be tracked from space using bottom-up
and top-down approaches. Bottom-up approaches use satellite
observations to track fire activity, such as burned area
15
or fire radia
-
tive power
16
. Emissions of CO
2
, CO and other trace gases are then
estimated by combining the estimates of fire activity with quantities
such as fuel loads and emission factors. Although these bottom-up
estimates are continually improving, inventories can vary significantly
in global and regional trace gas and aerosol emission estimates
15
,
17
.
Top-down approaches provide a method for refining bottom-up
trace gas emission estimates by optimally scaling emission estimates
to be consistent with the observed concentrations of trace gases in
fire plumes. A strength of this approach is that it integrates emis
-
sions from both flaming and smouldering combustion to capture net
emissions.
In this study, we perform top-down estimates of CO emissions from
the 2023 Canadian fires based on observational constraints from the
TROPOspheric monitoring instrument (TROPOMI) space-based CO
retrievals (Fig.
1a,b
). These estimates are performed using three differ
-
ent bottom-up fire emission inventories: the global fire emissions data
-
base (GFED4.1s)
15
, the global fire assimilation system v.1.2 (GFAS)
16
and
the quick fire emissions dataset v.2.6r1 (QFED)
18
. For each inversion,
the combined carbon emissions released as CO and CO
2
(CO
2
+ CO)
are then estimated using the CO
2
/CO emission factors from the same
bottom-up database. The CO
2
/CO emission ratios can be highly vari-
able, adding uncertainty to our analysis. We incorporate some of this
uncertainty here as each bottom-up database has different mean emis
-
sion ratios for Canadian forests (range, 7.7–10.8 gC of CO
2
per gC of
CO
2
). Details for these inversions are provided in the methods and
a description of the inversion results and evaluation of the perfor
-
mance of the top-down estimates are provided in Supplementary
Information sections 1 and 2). We find the top-down estimates are
relatively insensitive to choices about inversion configuration but do
show sensitivity to prescribed hydroxyl radical (OH) abundances
19
,
which determine the atmospheric lifetime of the CO emitted (Sup
-
plementary Information section 1 and Supplementary Fig. 1).
https://doi.org/10.1038/s41586-024-07878-z
Received: 29 November 2023
Accepted: 25 July 2024
Published online: 28 August 2024
Open access
Check for updates
1
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA.
2
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA, USA.
3
Joint Institute for Regional Earth System Science and Engineering, University of California, Los Angeles, CA, USA.
4
Meteorology & Air Quality Group, Wageningen University and Research,
Wageningen, The Netherlands.
5
Department of Physics, University of Toronto, Toronto, Ontario, Canada.
6
Division of Engineering and Applied Science, California Institute of Technology,
Pasadena, CA, USA.
7
Department of Energy, Environmental, and Chemical Engineering, Washington University in St. Louis, St. Louis, MO, USA.
✉
e-mail:
brendan.k.byrne@jpl.nasa.gov
836
| Nature | Vol 633 | 26 September 2024
Article
Figure
1c
shows the bottom-up and top-down CO
2
+ CO carbon emis
-
sions from fires during May–September 2023. The bottom-up datasets
show large differences, ranging from 234 to 735 TgC (mean of 469 TgC).
This range is reduced by 69% in the top-down estimates (570–727 TgC),
which also give a larger mean estimate of 647 TgC. Emissions during
2023 far exceed typical Canadian forest fire emissions, with 2010–2022
average emissions of 29–82 TgC for the bottom-up inventories and
121 TgC for top-down estimates (Supplementary Fig. 2). To contextu-
alize these numbers, we compare the top-down estimates to annual
national fossil fuel emissions for the ten largest emitters (Fig.
1d
). The
5 month 2023 emissions are more than four times larger than Canadian
annual fossil fuel emissions (149 TgC yr
−1
) and comparable to India’s
annual emissions (740 TgC yr
−1
).
Fire activity is affected by several complex drivers, including fuel
traits
20
and ignition probability
21
. However, fire weather—hot and dry
conditions—has been shown to be extremely important in driving
fire behaviour
22
. Climate data show an exceptionally hot and dry fire
season for Canadian forests during 2023 (Fig.
2
). This was the driest
January–September period for Canadian forests since at least 1980,
with about 86% of forested area having below-average precipitation
and about 52% being more than 1 s.d. below the 2003–2022 average
(Supplementary Fig. 4). May–September 2023 was the warmest since at
least 1980, with about 100% of the forest area above average and about
90% being more than 1 s.d. above the 2003–2022 average. Similarly,
the vapour pressure deficit (VPD), which is closely associated with
fire activity
22
–
24
, was the third highest since 1980, including 85% of the
forest area being above average and about 54% being more than 1 s.d.
above the 2003–2022 average.
Although hot–dry conditions were widespread across Canadian
forests, there are two notable regional patterns. Western Quebec
(49°–55° N, 72°–80° W), which is typically relatively wet (Supplementary
Fig. 5a), had exceptionally dry conditions during 2023, with precipita
-
tion through September being 23.7 cm (49%) below average. Coupled
with extreme heat and VPD during June–July, fire emissions in this
region contributed about 15% of the national total (Supplementary
Fig. 6). The other notable region was northwestern Canada near the
Great Slave Lake (57°–62° N, 110°−125° W). This region is drier than
western Quebec on average, with about half the annual precipitation.
However, 2023 was exceptional, with both a large precipitation deficit
of 8.1 cm (27% of January–September total) and exceptionally warm
conditions throughout May–September (+2.6 °C) (Supplementary
Fig. 6). This region contributed about 61% of the total Canadian forest
fire emissions.
Fires and climate
The relationship between climate variability and fire emissions for
Canadian forests is examined in Fig.
3
, which shows fire emissions as
a function of temperature and precipitation
Z
-scores over 2003–2023
for the 0.5° × 0.625° grid cells, in which
Z
-scores are the anomalies
divided by the standard deviation. May–September emissions are low
-
est for combined cool–wet conditions (5.2 gC m
−2
), whereas emissions
increase when either temperature is above average (19.5 gC m
−2
) or
precipitation is below average (9.2 gC m
−2
). However, emissions are
largest for combined warm–dry conditions (35.7 gC m
−2
). In particular,
fire emissions are much increased during exceptionally hot and dry
conditions (99.6 gC m
−2
, temperature
Z
> 1 and precipitation
Z
< −1).
These hot–dry conditions were much more prevalent in 2023 than in
preceding years, with a mean May–September T2M
Z
-score of 2.3 and a
precipitation
Z
-score of −1.1 across grid cells, explaining why fire emis
-
sions were extreme during 2023. Notably, the number of individual fires
during 2023 was not unusual, with 6,623 relative to a 10 yr average of
2019–2022
2023
b
70
80
90
100
110
120
130
140
X
CO
(ppb)
0
0.2
0.4
0.6
0.8
CO
2
+ CO May–September
re emissions (PgC)
c
Bottom-up
Top-down
Bottom-up
Top-down
2010–2022
2023
GFED
GFAS
QFED
0
1
2
3
Carbon emissions (PgC)
China
USA
India
2023 Canada res
Russia
Japan
Iran
Germany
Indonesia
South Korea
Canada
d
Fossil
Fire
a
Fig. 1 | CO enhancements and f ire emission estimates.
a
–
c
, May–September
TROPOMI dry-air mole fractions of CO (
X
CO
) averaged over 2019–2022 (
a
) and for
2023 (
b
) aggregated to a 2° × 2.5° grid.
c
, Canadian forest fire carbon emissions
(from CO and CO
2
) for the 2023 May–September fire season, compared with fire
emissions during 2010–2022 (distribution shown by box-and-whisker plots).
Top-down emissions over 2010–2022 are estimated from MOPITT (2010–2021)
and TROPOMI (2019–2022) CO retrievals.
d
, A comparison of May–September
Canadian fire emissions with 2022 territorial fossil carbon emissions for the ten
largest emitting countries, obtained from Global Carbon Budget 2022
2
.