Northward migration of the boreal forest confirmed by satellite record

Abstract

The boreal forest is one of Earth’s most climatologically sensitive regions, and changes in the cover and structure of its vegetation pose a positive carbon-climate feedback on atmospheric greenhouse warming. The region has also experienced more than three times climatological warming of any forested biome in recent decades. While ecological models predict a northward shift of boreal tree cover in response to climate change, comprehensive data have not been available to test the hypothesis. Here we report a test of the magnitude, direction, and significance of changes in the boreal canopy based on the longest and highest-resolution record of calibrated satellite maps to date. The boreal canopy increased in density and shifted northward from 1984 to 2020, with the largest and most significant gains in its northern latitudes. Net forest gains occurred despite stable rates of disturbance across all but the region’s southernmost latitudes, implicating widespread release of climatological limitation on growth over changing distribution of fire, harvest, insect, and other disturbances. These new forests will sequester carbon as they mature, increasing its residence time in woody biomass, and will play a key role in how the terrestrial biosphere attenuates atmospheric CO2 increases.

Full text

1
Northward migration of the boreal forest
1
confirmed by satellite record
2
3
4
The boreal biome is Earth’s most expansive, ecologically intact, and climatologically sensitive
5
forest. The boreal forest comprises a third of the global forest area and accounts for 20.8% of the
6
total forest carbon (C) sink 1. The boreal region contains 38 ± 3.1 Pg C of above-ground biomass
7
2 and is underlain by 1672 Pg C, summing to total biomass rivaling the tropics and half of global
8
soil C—of which 88% is locked in permafrost 3,4. Boreal vegetation structure also controls the
9
reflective and thermal balance of solar radiation of the high northern latitudes via canopy albedo,
10
posing a positive feedback mechanism for greenhouse atmospheric warming 5–8.
11
The boreal region has experienced the fastest climatological warming of any forest biome, with
12
annual surface temperatures increasing > 1.4° C over the past century 9. Boreal forest dynamics
13
are highly correlated to climate 10–12, and increases in vegetation productivity have been observed
14
across the northern high latitudes 13. Meanwhile, regional increases in the frequency and severity
15
of windthrow, fire, insect, and disease events have been reported as well 3.
16
While theory predicts a northward shift of the boreal forest, the net effect of the many opposing
17
factors on the region’s tree canopy remains an untested hypothesis. Coupled climate-vegetation
18
models predict a net-northward migration of boreal vegetation due to warming 14,15, supporting
19
the dominance of growth processes, and multiple studies 16–18 have reported vegetation
20
“greening” based on general indices of plant productivity. However, the slow productivity of
21
boreal tree cover requires long-term analyses, which have been either confined to regional scopes
22
or poorly calibrated data 19–21. As a result, the net effect of growth and mortality on the
23

2
distribution of the boreal tree canopy, and the resulting effect on carbon budgets, remain
24
unconfirmed.
25
Here we report the results of a global test of the magnitude and direction of boreal-forest change
26
from 1984 to 2020, as observed through historical satellite records of tree cover. We calibrated
27
machine learning algorithms 22,23 to 224,026 Landsat images covering the boreal forest and
28
adjacent tundra on the Amazon Web Services (AWS) cloud-computing architecture to estimate
29
tree-canopy cover over space and time. The resulting 30-meter, annual-resolution dataset—the
30
most extensive and highest-resolution record of boreal tree cover to date—was then subjected to
31
time-series trend analysis to estimate and map the historical direction, rate, and significance of
32
change across the region.
33
Distribution of boreal tree-canopy cover
34
The tree canopy is densest in the southern portions of the biome and thins with increasing
35
latitude (Fig. 1) 23. Sparse conifer forest, woodland, herbaceous, and non-vegetated cover
36
increase in frequency into and across the taiga-tundra ecotone, and tree cover is nearly absent
37
above 71°N. Including unforested tundra, wetlands, and inland water bodies, the most common
38
range of tree cover is below 5%.
39
The boreal forest increased in density from 1985 to 2020 (Fig. 2). Trees covered 7.153 million
40
km2 (41.44 %) of the region in 1985 and 7.997 million km2 (46.32 %) in 2020, increasing
41
linearly 0.023 million km2/yr (0.12%/yr) over the 36-year period (percent cover = 0.116 x year –
42
187.6, R2 = 0.79, p < 0.001). Given the 10-30% range of tree-canopy cover defining forests
43
within the United Nations Framework Convention on Climate Change 22, the region held
44

3
between 8.95 to 12.41 million km2 of forest in 2000 and increased to between 9.41 and 13.26
45
million km2 of forest in 2020.
46
The boreal forest also shifted northward from 1985 to 2020. The mean latitude of boreal tree-
47
canopy cover increased half a degree, from 57.37 °N to 57.66 °N over the period (mean latitude
48
= 0.0075 x year + 42.6, R2=0.79, p < 0.001). Median latitude increased at a faster rate than the
49
mean (median latitude = 0.0124 x year + 32.5, R² = 0.88, p < 0.001), implying widespread
50
growth across the entire biome rather than changes at either its outlying northern or southern
51
margins.
52
53

4
54
55
Fig. 1 | Current (2020) distribution of tree-canopy cover across boreal and arctic tundra ecoregions. The tree-canopy
56
cover was estimated at 30-meter pixel resolution by machine learning algorithms applied to all available Landsat satellite
57
images from the year 2020; data gaps due to clouds were filled with estimates from earlier years. Ecoregions were defined
58
by 24. The bottom panel shows northward migration of the distribution of boreal tree-canopy cover from 1984 to 2020.
59
Tree-canopy cover (%)
90°W90°E
0°
180°
60°W
120°W
150°W150°E
120°E
60°E
30°E
1.5
2.0
2.5
3.0
3.5
Area (106km2)
Tree -canopy cover (%)
010 20 30 40 50 60 70 100
0
50
100
25
75
025 50 75 100
y = 0.116x -187.6
R² = 0.79
y = 0.185x -325.3
R² = 0.83
35
37
39
41
43
45
47
49
51
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
tree-canopy cover (%)
Year
mean median Linear (mean) Linear (median)

5
The pace and pattern of boreal forest change
60
These global totals comprise the balance of strong geographic variation (Fig. 2). Net canopy
61
gains occurred at every latitude above 53°N from 1984 to 2020, with the strongest increases
62
occurring between 64 to 68°N. While not seeking a distinct line per se, net gains in the region’s
63
highest latitudes support the hypothesis of a positive shift in the northern limit of tree cover, or
64
“northern tree line”. In contrast, net canopy losses were smaller in magnitude and confined to the
65
lower boreal latitudes (45-51°N), where human activity is most intense (Fig. 3).
66
In North America, significant net gains were concentrated in the northernmost portion of the
67
domain, where increases in shrub and grass cover have also been reported 25. Regions of
68
significant net losses corresponded to areas of widespread forest disturbance, including fire and
69
bark beetle (Dendroctonus spp.) outbreaks in British Columbia 26, spruce budworm
70
(Choristonura sp.) outbreaks in Quebec 27, and fires across the central Canadian provinces and
71
interior Alaska 28.
72
In Eurasia, hotspots of forest loss included the eastern Russian-Chinese border and agricultural
73
regions of the southern boreal margin east of the Ural Mountains, as well as areas of forest
74
felling near the Russia-Finland border in the 1990s 29 and increased fire frequency and intense
75
selective logging 30. Net losses were notably rare in Europe 31. Corroborating reports of
76
increasing coverage of tall shrubs and larch (Larix spp.) in the Siberian ecotone 32, regions of
77
significant net gain included areas of post-Soviet agricultural abandonment and reforestation, as
78
well as larch forests underlain by permafrost in the Yakutsk region of eastern Siberia. In these
79
forests, permafrost thawing has been hypothesized to result in increased forest productivity 33,
80
and vegetation recovery from wildfires in the 1990s is ongoing 34.
81

6
Fig. 2 | Spatial and temporal distribution of boreal tree-canopy cover change from 1984 to 2020. Map: Significant net gains
(green-blue) and losses (orange-red) of tree-canopy cover over the boreal biome. Table (top-right): Linear regression slope
of tree canopy-cover over time, stratified by latitude. Bottom panel: northward migration of the distribution of mean and
median latitude of tree-canopy cover. All image tiles included in the analysis had >30 unobscured observations from 1984 to
2020.
82
TCC c hang e rat e
y = 0.0075x + 42.6
R² = 0.79
y = 0.0124x + 32.5
R² = 0.88
57
57.1
57.2
57.3
57.4
57.5
57.6
57.7
57.8
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
Latitude (°N)
Year
mean median Linear (mean) Linear (median)

7
The pan-boreal increase in tree cover occurred against a backdrop of relatively stable rates of
83
disturbance over the period (Fig. 3). The region-wide rate of disturbance accelerated from 53,546
84
km2/yr in 2000 to 60,275 km2/yr in 2020— equating to an increase of 1.8%/yr (1,100 km2/yr)
85
(y=1,099.57 x - 2,157,378.02, R2=0.27, p-value = 0.016) or 0.2 to 0.4% of the maximum forested
86
area of the region over the period. In contrast to net gains, the latitudinal distribution of
87
disturbances fluctuated from year to year while remaining stationary over the period as a whole
88
(y=0.04 x - 25.16, R2=0.14, p-value: 0.023). !
89
Previous studies have sought evidence of a northward shift of the boreal biome at the northern
90
limit of tree cover. While reports of advances in the northern tree line based on globally
91
calibrated datasets have been contested for coarse categorization of the forest and poor
92
calibration 35,36, our observations based on regionally calibrated estimates corroborate the
93
advancement hypothesis, as well as independent reports of disturbance and recovery 16,37–40 and
94
in situ measurements of changing woody structure near the northern limits of tree growth 32,41.
95
Whereas we sought neither evidence of a discrete northern edge to the boreal forest or changes
96
therein, our results do show a biome-wide northward shift in the entire distribution of tree
97
cover—including the gradient spanning its northern extreme.
98

8
Fig. 3 | Total area and median latitude of boreal stand-clearing disturbances from 1985 to 2020.
99
90°W90°E
0°
180°
60°W
30°W
120°W
150°W150°E
120°E
60°E
30°E
Year of forest
disturbance
53
55
57
59
61
63
0
20000
40000
60000
80000
100000
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
Latitude (°N)
Area (km2)
Year
Complete Incomplete Latitude

9
Distribution of boreal forest age
100
We also retrieved the spatial and frequency distribution of current forest stand age across the
101
boreal biome. While most of the boreal forest area (8.19 million km2, 47.5% of the region) is
102
older than what can be measured from the 37-year satellite record (Fig. 4), some observations of
103
the region’s youngest forests can be made empirically. In 2020, 0.5 million km2 (or 5.29% of
104
standing forests) were those that had been identified as forest in 1984, disturbed at some time
105
during the period, and recovered again to forest by 2020. The total of these recovering forests
106
and expanding “new” forests within the observable period have led to a weak mode of young
107
stands between 9 and 21 years of age, as well as a current lapse in the youngest age classes.
108
These young forests are especially prevalent in areas of intensive forestries, such as the industrial
109
plantations of Scandinavia, as well as areas recovering from wildfire. The latter is corroborated
110
by reports of increasing frequency and area of burns in Siberia since the end of the 20th century
111
42, the demographic effect of which is reflected in an increasing proportion of recovering forest <
112
20 years old.
113

10
Fig. 4 | The distribution of stand age (top) across the boreal ecoregion, and frequency distribution of boreal stand age
in 2020 (bottom). Forest age-class distribution is defined as years since the establishment of pixels that were forested in
2020.
Age in 2020
90°W90°E
0°
180°
60°W
30°W
120°W
150°W150°E
120°E
60°E
30°E
-
10,000
20,000
30,000
40,000
50,000
60,000
≤2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 ≥36
Area (km2)
Age
New Recovery
8.19 million km2
…

11
The carbon impact of young forests
114
The impact of the young forests on the boreal carbon budget is significant, and it could explain
115
the region’s increasing carbon sink 43. Forests with known stand ages (≤ 36 years since
116
disturbance) hold between 1.1 – 5.9 Pg C, based on recent models 44. Ages of forests where no
117
disturbance was recorded during the observation period are unknown, yet plausible carbon-stock
118
values within the oldest forested age class may be bracketed between a lower, younger limit
119
(19.1 – 58.4 Pg C) and an upper, older limit (300 years stand age, 42.4 – 89.2 Pg C). Based on
120
these estimates, all forested area of ages ≤ 36 years comprises 1.35% to 14.20% of the total
121
carbon stock in aboveground biomass in the boreal forest, which increases with the fractional
122
area of young (≤ 36-year) forest to total forest area (15.4%). Allowing all these young forests to
123
mature without disturbance would result in a potential additional carbon sink of 2.3 – 3.8 Pg C.
124
The amount of carbon in forests new to the satellite record is 0.8 – 3.5 Pg C, greater than carbon
125
in forests recovering from observed disturbances (0.3 – 2.4 Pg C). Over the next 36 years, these
126
“new” forests represent a potential additional carbon sink of 1.3 – 2.0 Pg C (0.036 - 0.18 Pg
127
C/yr), compared to 1.0 – 1.8 Pg C (0.028 - 0.05 Pg C/yr) in recovering forest. The differences in
128
present-day stocks and the carbon sink potential between new and regrowing forests can be
129
partly explained by the greater area of new forests compared to regrowing forests (7.6% and
130
6.7% of total forest area, respectively), but also by the greater age of new forests compared to
131
those forests recovering from disturbances within the record.
132
The amount of carbon sequestered by the new forests could be large enough to offset the effect
133
of warming on boreal ecosystem respiration, estimates of which vary from 5 Pg C to 28 Pg C
134
from 1985 to 2020 (SI). Climate warming and CO2 fertilization are expected to increase
135
productivity in these regions 45; and interestingly, the observed spatial pattern of net canopy
136

12
growth confirms model predictions of enhanced seasonal CO2 exchange at latitudes > 40°N 13.
137
However, several factors could yet reduce the offset of forest expansion on a temperature-
138
mediated increase in respiratory fluxes, including: the temperature effect itself can be
139
temperature dependent 46, sink capacity eventually decreases with age 47, thawing of carbon
140
locked in permafrost will accelerate respiration 48, and changes in fire regimes and wood harvest
141
could overshadow stocking from forest development 3,42. It also remains unclear to what extent
142
the expansion of trees, with longer lived carbon pools than herbaceous vegetation, can be
143
structurally sustained by boreal soils 49, or on how disturbed area might increase from human
144
activity. Each of these dynamics is already taking place across the boreal domain, and strategies
145
to quantify the potential tradeoffs between autotrophic and heterotrophic dynamics are key to
146
understanding the role of forest management in mitigating the causes and consequences of
147
climate change in the boreal domain.
148
Conclusions
149
A pan-boreal test of the magnitude, direction, and significance of boreal forest change has
150
confirmed the climatological and ecological hypothesis of northward migration of the boreal
151
biome. Machine learning was used to retrieve the longest, highest resolution, and most complete
152
calibrated record of boreal-forest change from the historical satellite record to date. Time-series
153
analysis of each of the 1.9 x108 30-m pixels over 37 years revealed increasing canopy density
154
and northward migration of the boreal forest from 1984 to 2020 despite relatively even rates of
155
forest disturbance over the period. Recent models of the relation of forest age to biomass stocks
156
and change suggest the changing distribution of age will significantly affect the region’s
157
contribution to the global carbon budget in coming decades.
158

13
While globally significant, the trends belie tremendous variation over space and time, as well as
159
in the processes underlying the observable changes. A deeper understanding of the full
160
complexity of the causes and consequences of canopy changes across the boreal forest will
161
require analysis against coincident measurements of canopy structure and the environmental
162
determinants of growth and mortality. Further, translating the resulting information into action to
163
forestall and adapt to climate change will require effective communication across scientific,
164
government, and commercial domains.
165
Online content
166
The current distribution of boreal tree canopy cover and its changes over time can be explored
167
publicly at https://www.terraPulse.com/terraView/ccs.
168
169
References
170
1. Pan, Y. et al. A Large and Persistent Carbon Sink in the World’s Forests. Science 333, 988–
171
993 (2011).
172
2. Neigh, C. S. R. et al. Taking stock of circumboreal forest carbon with ground measurements,
173
airborne and spaceborne LiDAR. Remote Sens. Environ. 137, 274–287 (2013).
174
3. Gauthier, S., Bernier, P., Kuuluvainen, T., Shvidenko, A. Z. & Schepaschenko, D. G. Boreal
175
forest health and global change. Science 349, 819–822 (2015).
176
4. Tarnocai, C. et al. Soil organic carbon pools in the northern circumpolar permafrost region.
177
Glob. Biogeochem. Cycles 23, (2009).
178
5. Betts, R. A. Offset of the potential carbon sink from boreal forestation by decreases in
179
surface albedo. Nature 408, 187–190 (2000).
180

14
6. Bonan, G. B. Forests and Climate Change: Forcings, Feedbacks, and the Climate Benefits of
181
Forests. Science 320, 1444–1449 (2008).
182
7. Chen, D., Loboda, T. V., He, T., Zhang, Y. & Liang, S. Strong cooling induced by stand-
183
replacing fires through albedo in Siberian larch forests. Sci. Rep. 8, 4821 (2018).
184
8. Randerson, J. T. et al. The Impact of Boreal Forest Fire on Climate Warming. Science 314,
185
1130–1132 (2006).
186
9. IPCC. Climate Change 2014 Synthesis Report. Climate Change 2014: Synthesis Report
187
(2014).
188
10. Elmendorf, S. C. et al. Plot-scale evidence of tundra vegetation change and links to recent
189
summer warming. Nat. Clim. Change 2, 453–457 (2012).
190
11. Holtmeier, F.-K. & Broll, G. Sensitivity and response of northern hemisphere altitudinal and
191
polar treelines to environmental change at landscape and local scales. Glob. Ecol. Biogeogr.
192
14, 395–410 (2005).
193
12. Véga, C. & St-Onge, B. Mapping site index and age by linking a time series of canopy height
194
models with growth curves. For. Ecol. Manag. 257, 951–959 (2009).
195
13. Forkel, M. et al. Enhanced seasonal CO2 exchange caused by amplified plant productivity in
196
northern ecosystems. Science 351, 696–699 (2016).
197
14. IPCC. Global warming of 1.5°C. Ipcc - Sr15 (2018).
198
15. Scheffer, M., Hirota, M., Holmgren, M., Nes, E. H. V. & Chapin, F. S. Thresholds for boreal
199
biome transitions. Proc. Natl. Acad. Sci. 109, 21384–21389 (2012).
200
16. Sulla-Menashe, D., Woodcock, C. E. & Friedl, M. A. Canadian boreal forest greening and
201
browning trends: an analysis of biogeographic patterns and the relative roles of disturbance
202
versus climate drivers. Environ. Res. Lett. 13, 014007 (2018).
203

15
17. Zhu, Z. et al. Greening of the Earth and its drivers. Nat. Clim. Change 6, 791–795 (2016).
204
18. Piao, S. et al. Characteristics, drivers and feedbacks of global greening. Nat. Rev. Earth
205
Environ. 1, 14–27 (2020).
206
19. Beck, P. S. A. et al. Changes in forest productivity across Alaska consistent with biome shift.
207
Ecol. Lett. 14, 373–379 (2011).
208
20. Brice, M. et al. Moderate disturbances accelerate forest transition dynamics under climate
209
change in the temperate–boreal ecotone of eastern North America. Glob. Change Biol. 26,
210
4418–4435 (2020).
211
21. Taylor, A. R. et al. Rapid 21st century climate change projected to shift composition and
212
growth of Canada’s Acadian Forest Region. For. Ecol. Manag. 405, 284–294 (2017).
213
22. Sexton, J. O. et al. Conservation policy and the measurement of forests. Nat. Clim. Change
214
6, 192–196 (2016).
215
23. Sexton, J. O. et al. Global, 30-m resolution continuous fields of tree cover: Landsat-based
216
rescaling of MODIS Vegetation Continuous Fields with lidar-based estimates of error. Int. J.
217
Digit. Earth 6, 427–448 (2013).
218
24. Dinerstein, E. et al. An Ecoregion-Based Approach to Protecting Half the Terrestrial Realm.
219
BioScience 67, 534–545 (2017).
220
25. McManus, kelly M. et al. Satellite-based evidence for shrub and graminoid tundra
221
expansion in northern Quebec from 1986 to 2010. Glob. Change Biol. 18, 2313–2323
222
(2012).
223
26. Meddens, A. J. H., Hicke, J. A. & Ferguson, C. A. Spatiotemporal patterns of observed bark
224
beetle-caused tree mortality in British Columbia and the western United States. Ecol. Appl.
225
22, 1876–1891 (2012).
226

16
27. Boulanger, Y. & Arseneault, D. Spruce budworm outbreaks in eastern Quebec over the last
227
450 years. Can. J. For. Res. (2004) doi:10.1139/x03-269.
228
28. Stocks, B. J. et al. Large forest fires in Canada, 1959–1997. J. Geophys. Res. Atmospheres
229
107, FFR 5-1-FFR 5-12 (2002).
230
29. Potapov, P. et al. Forest Cover Change within the Russian European North after the
231
Breakdown of Soviet Union (1990–2005). International Journal of Forestry Research vol.
232
2012 e729614 https://www.hindawi.com/journals/ijfr/2012/729614/ (2012).
233
30. Krylov, A. et al. Remote sensing estimates of stand-replacement fires in Russia, 2002–2011.
234
Environ. Res. Lett. 9, 105007 (2014).
235
31. Ceccherini, G. et al. Abrupt increase in harvested forest area over Europe after 2015. Nature
236
583, 72–77 (2020).
237
32. Frost, G. V. & Epstein, H. E. Tall shrub and tree expansion in Siberian tundra ecotones since
238
the 1960s. Glob. Change Biol. 20, 1264–1277 (2014).
239
33. Sato, H., Kobayashi, H., Iwahana, G. & Ohta, T. Endurance of larch forest ecosystems in
240
eastern Siberia under warming trends. Ecol. Evol. 6, 5690–5704 (2016).
241
34. Kajii, Y. et al. Boreal forest fires in Siberia in 1998: Estimation of area burned and emissions
242
of pollutants by advanced very high resolution radiometer satellite data. J. Geophys. Res.
243
Atmospheres 107, ACH 4-1-ACH 4-8 (2002).
244
35. Wang, J. A. et al. Extensive land cover change across Arctic–Boreal Northwestern North
245
America from disturbance and climate forcing. Glob. Change Biol. 26, 807–822 (2019).
246
36. Timoney, K. P. & Mamet, S. No treeline advance over the last 50 years in subarctic western
247
and central Canada and the problem of vegetation misclassification in remotely sensed data.
248
Écoscience 27, 93–106 (2020).
249

17
37. Bunn, A. G. & Goetz, S. J. Trends in Satellite-Observed Circumpolar Photosynthetic
250
Activity from 1982 to 2003: The Influence of Seasonality, Cover Type, and Vegetation
251
Density. Earth Interact. 10, 1–19 (2006).
252
38. Guay, K. C. et al. Vegetation productivity patterns at high northern latitudes: a multi-sensor
253
satellite data assessment. Glob. Change Biol. 20, 3147–3158 (2014).
254
39. Ju, J. & Masek, J. G. The vegetation greenness trend in Canada and US Alaska from 1984–
255
2012 Landsat data. Remote Sens. Environ. 176, 1–16 (2016).
256
40. Neigh, C. S. R., Tucker, C. J. & Townshend, J. R. G. North American vegetation dynamics
257
observed with multi-resolution satellite data. Remote Sens. Environ. 112, 1749–1772 (2008).
258
41. Rees, W. G. et al. Is subarctic forest advance able to keep pace with climate change? Glob.
259
Change Biol. 26, 3965–3977 (2020).
260
42. Kharuk, V. I. et al. Wildfires in the Siberian taiga. Ambio (2021) doi:10.1007/s13280-020-
261
01490-x.
262
43. Ciais, P. et al. Five decades of northern land carbon uptake revealed by the interhemispheric
263
CO 2 gradient. Nature 568, 221–225 (2019).
264
44. Cook-Patton, S. C. et al. Mapping carbon accumulation potential from global natural forest
265
regrowth. Nature 585, 545–550 (2020).
266
45. Norby, R. J. & Zak, D. R. Ecological Lessons from Free-Air CO2 Enrichment (FACE)
267
Experiments. Annu. Rev. Ecol. Evol. Syst. 42, 181–203 (2011).
268
46. Koven, C. D., Hugelius, G., Lawrence, D. M. & Wieder, W. R. Higher climatological
269
temperature sensitivity of soil carbon in cold than warm climates. Nat. Clim. Change 7, 817–
270
822 (2017).
271
47. Odum, E. P. The Strategy of Ecosystem Development. Science 164, 262 (1969).
272

18
48. Schuur, E. a. G. et al. Climate change and the permafrost carbon feedback. Nature 520, 171–
273
179 (2015).
274
49. Koven, C. D. Boreal carbon loss due to poleward shift in low-carbon ecosystems. Nat.
275
Geosci. 6, 452–456 (2013).
276
277
278