Carbon pools in China’s terrestrial ecosystems: New
estimates based on an intensive field survey
Xuli Tanga,1, Xia Zhaob,1, Yongfei Baib,1, Zhiyao Tangc,1, Wantong Wanga,d, Yongcun Zhaoe, Hongwei Wanb,
Zongqiang Xieb, Xuezheng Shie, Bingfang Wuf, Gengxu Wangg, Junhua Yana, Keping Mab, Sheng Duh, Shenggong Lii,
Shijie Hanj, Youxin Mak, Huifeng Hub, Nianpeng Hei, Yuanhe Yangb, Wenxuan Hanl, Hongling Hei, Guirui Yui,
Jingyun Fangb,c,2, and Guoyi Zhoua,2
South China Botanical Garden, Chinese Academy of Sciences, Guangzhou 510650, China; bInstitute of Botany, Chinese Academy of Sciences, Beijing
100093, China; cKey Laboratory for Earth Surface Processes of the Ministry of Education, College of Urban and Environmental Sciences, Peking University,
Beijing 100871, China; dCollege of Tourism, Henan Normal University, Xinxiang 453007, China; eState Key Laboratory of Soil and Sustainable Agriculture,
Institute of Soil Science, Chinese Academy of Sciences, Nanjing 210008, China; fInstitute of Remote Sensing and Digital Earth, Chinese Academy of Sciences,
Beijing 100094, China; gInstitute of Mountain Hazards and Environment, Chinese Academy of Sciences, Chengdu 610041, China; hState Key Laboratory of
Soil Erosion and Dryland Farming on Loess Plateau, Institute of Soil and Water Conservation, Chinese Academy of Sciences and Ministry of Water Resources,
Yangling 712100, China; iInstitute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China; jInstitute of
Applied Ecology, Chinese Academy of Sciences, Shenyang 110016, China; kXishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Mengla
666303, China; and lCollege of Resources and Environmental Sciences, China Agricultural University, Beijing 100193, China
carbon stock climatic influences
change of carbon pools has occurred in China’s croplands and
grasslands over the past three decades (8, 9).
Although there have been several studies of the carbon pools of
China’s terrestrial ecosystems, the estimates of these pools have
varied by more than 100 Pg C (SI Appendix, Table S1), suggesting
an inconsistency among these estimates. This inconsistency is likely
due to the limitation of sample size and data representativeness,
multiplicity of data sources, and inconsistency of methodologies. In
addition, previous estimates at both regional and national scales
were primarily obtained based on summarized data of the regional
or national censuses (e.g., China’s forest inventory and China’s
grassland resource survey) (7, 10, 11) and not from original observations (2). Our knowledge of the driving forces causing the changes
in terrestrial ecosystem carbon pools is also very limited and has
impeded the application of management measures.
To fill this knowledge gap, we conducted a nationwide field
campaign between 2011 and 2015 to investigate the carbon
stocks of terrestrial ecosystems in China. A reviewable, consistent inventory system, independent of the routine surveys
Previous estimations of carbon budgets in China’s terrestrial ecosystems varied greatly because of the multiplicity of data sources
and the inconsistency of methodologies. By conducting a methodologically consistent field campaign across the country, we estimated that the total carbon pool in China’s forests, shrublands,
grasslands, and croplands was 79.24 ± 2.42 Pg C. The carbon
density exhibited a strong dependence on climate regime: it decreased with temperature but increased with precipitation. The
country’s forests have a large potential of biomass carbon sequestration of 1.9–3.4 Pg C in the next 10 to 20 years assuming no
removals. Our findings provide a benchmark to identify the effectiveness of the government’s natural protection policies.
| human influences | spatial variations |
errestrial ecosystems are a significant carbon sink on Earth,
accounting for ∼20–30% of the total anthropogenic carbon
dioxide (CO2) emissions to the atmosphere. Compared with
oceans, terrestrial ecosystems can be readily managed to either
increase or decrease carbon sequestration by restoring or
degrading vegetation (1). China is a good example of this interaction between human-driven vegetation change and terrestrial
carbon exchange (2, 3). For example, China’s forest coverage decreased from 30 to 40% in the early 1950s to ∼14% in the early
1980s because of excessive exploitation of forest resources. However, since then, nationwide vegetation restoration practices, including several key ecological restoration programs, have been
implemented (4), resulting in a significant increase in forest coverage—from 13.9% in the early 1990s to 21% in the 2010s (5, 6).
Corresponding to the changes in forest area and the growth of
established forests, the carbon pools of China’s forest ecosystems
have significantly increased during these decades (7–9). Compared
with forests, biomass production of grasslands and croplands is
quite low, varying from 0.01 to 0.02 Pg C per year, and thus limited
Author contributions: J.F. and G.Z. designed research; X.T., Y.B., Z.T., Y.Z., H.W., Z.X., X.S.,
G.W., J.Y., K.M., S.D., S.L., S.H., Y.M., H. He, G.Y., and G.Z. performed research; X.T., X.Z.,
W.W., Y.Z., and B.W. analyzed data; and X.T., X.Z., W.W., H. Hu, N.H., Y.Y., W.H., J.F., and
G.Z. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Published under the PNAS license.
X.T., X.Z., Y.B., and Z.T. contributed equally to this work.
To whom correspondence may be addressed. Email: email@example.com or jyfang@urban.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
Published online April 16, 2018.
PNAS | April 17, 2018 | vol. 115 | no. 16 | 4021–4026
China’s terrestrial ecosystems have functioned as important carbon
sinks. However, previous estimates of carbon budgets have included
large uncertainties owing to the limitations of sample size, multiple data
sources, and inconsistent methodologies. In this study, we conducted an
intensive field campaign involving 14,371 field plots to investigate all
sectors of carbon stocks in China’s forests, shrublands, grasslands, and
croplands to better estimate the regional and national carbon pools and
to explore the biogeographical patterns and potential drivers of these
pools. The total carbon pool in these four ecosystems was 79.24 ± 2.42
Pg C, of which 82.9% was stored in soil (to a depth of 1 m), 16.5% in
biomass, and 0.60% in litter. Forests, shrublands, grasslands, and croplands contained 30.83 ± 1.57 Pg C, 6.69 ± 0.32 Pg C, 25.40 ± 1.49 Pg C,
and 16.32 ± 0.41 Pg C, respectively. When all terrestrial ecosystems are
taken into account, the country’s total carbon pool is 89.27 ± 1.05 Pg C.
The carbon density of the forests, shrublands, and grasslands exhibited
a strong correlation with climate: it decreased with increasing temperature but increased with increasing precipitation. Our analysis also
suggests a significant sequestration potential of 1.9–3.4 Pg C in forest
biomass in the next 10–20 years assuming no removals, mainly because
of forest growth. Our results update the estimates of carbon pools in
China’s terrestrial ecosystems based on direct field measurements, and
these estimates are essential to the validation and parameterization
of carbon models in China and globally.
Edited by Susan E. Trumbore, Max Planck Institute for Biogeochemistry, Jena, Germany, and approved November 21, 2017 (received for review February
conducted for forests and shrublands by the Chinese Ministry of
Forestry (6) and for grasslands by the Chinese Ministry of Agriculture (12) was developed based on the spatial distributions of
China’s terrestrial ecosystems (SI Appendix, Texts S1–S3). In total,
13,030 field plots were investigated across forests, shrublands, and
grasslands in mainland China using consistent methodology
(Materials and Methods and SI Appendix, Fig. S1). We also conducted a systematic field investigation for croplands, with 1,341 field
plots from 58 counties that represent typical cropping systems used
in China (SI Appendix, Text S4, and Fig. S1). Here we defined
“forest” as the land with an area of ≥0.067 ha dominated by trees
and with a tree crown coverage of ≥20%; “shrubland” as the land
dominated by shrubs with a canopy height of <5 m and canopy
coverage of >30–40%; “grassland” as the land dominated by herbaceous plants; and “cropland” as the land surface covered by crops
with a minimum area of 5,400 m2 that is seeded at least once per
year (SI Appendix, Texts S1–S4).
Our field campaign investigated all carbon components of an
entire ecosystem in these four vegetation groups, including aboveand belowground biomass, understory plants, litter, and soils. The
major purposes of this study are to estimate the carbon pools of
these ecosystems and to elucidate the possible climatic and anthropogenic drivers of the spatial distributions of these carbon
pools by using direct field measurements collected in this study.
Note that we did not investigate the carbon pools in Taiwan, Hong
Kong, Macao, and the South China Sea Islands because of the
unavailability of fieldwork and the small land areas in these islands. Our study focused on forests, shrublands, and grasslands
when exploring the drivers shaping the distribution of carbon
stocks because croplands are intensively human managed.
Carbon Stocks and Their Spatial Variations. Ecosystem carbon
density (carbon stock per hectare) of forests, shrublands, and
grasslands exhibited large spatial variations at the national scale
(Fig. 1). Both biomass and litter carbon densities decreased from
the northeastern, southern, southeastern, and southwestern regions to the northern and northwestern regions and to the Tibetan
Fig. 1. Spatial distribution of ecosystem carbon density (Mg C ha−1) in
forests, shrublands, grasslands, and croplands in China. (A) biomass carbon.
(B) Soil organic carbon (up to 1 m in depth, where applicable). (C) Litter
carbon. (D) Total ecosystem carbon. The site-averaged carbon density of each
biome in each province was assigned to the corresponding polygons of the
ChinaCover map. (For details on the ChinaCover map and associated vegetation biomes, see ref. 5. Please note that we did not investigate the carbon
pools in Taiwan, Hong Kong, Macao, and the South China Sea Islands.)
4022 | www.pnas.org/cgi/doi/10.1073/pnas.1700291115
Fig. 2. Distribution of provincial-level total ecosystem carbon pools (Pg C) in
China’s forests, shrublands, grasslands, and croplands and their histograms
by region. In each histogram, the carbon pools of biomass, litterfall, and soil
in forests (F), shrublands (S), grasslands (G), and croplands (C) are shown for the
six regions (Northeast, North, Northwest, East, South Central, and Southwest).
Plateau (Fig. 1 A and C). However, the soil carbon density displayed complex variations: the maximum density occurred on
Mount Xing’an in the northeastern region, Mounts Qilian and
Bayan Har in Qinghai, and Mounts Tianshan and Alta in
northern Xinjiang, followed by the southern and southeastern
regions. The lowest soil carbon densities were in the lower basins
in Xinjiang, the Hexi Corridor in Gansu, and on part of the
Loess Plateau (Fig. 1B). The mean ecosystem carbon density
showed the highest value in forest ecosystems (163.8 ± 8.4 Mg C ha−1),
which is ∼1.8 times higher than that in shrublands (89.9 ± 4.4 Mg C
ha−1) and grasslands (90.3 ± 5.3 Mg C ha−1) (see SI Appendix,
Table S2 for details). Overall, the area-weighted average ecosystem carbon density of all three vegetation groups was 115.7 ± 6.2
Mg C ha−1, with 23.1 ± 5.7, 0.8 ± 0.9, and 91.8 ± 9.2 Mg C ha−1
stored in biomass, litter, and soil.
The total carbon pool of these three ecosystems was 62.93 ±
3.39 Pg C, of which biomass, litter, and soil organic carbon
[(SOC) at a 1-m depth, where applicable] were 12.55 ± 3.07
(20%), 0.46 ± 0.48 (0.7%), and 49.92 ± 4.98 Pg C (79.3%), respectively (SI Appendix, Table S3). The largest carbon pool was
in forests (30.83 ± 1.57 Pg C, 49%), followed by grasslands
(25.40 ± 1.49 Pg C, 40.4%), and shrublands (6.69 ± 0.32 Pg C,
10.6%). Geographically, 19.53 ± 0.54 Pg C (31%) was stored in
southwestern China (Fig. 2) because of its large area and high
carbon densities in vegetation biomass and soils. By contrast,
only 4.55 ± 0.11 Pg C (7%) was stored in eastern China (Fig. 2),
where carbon densities were quite low (Fig. 1D).
In addition, we used the Random Forest simulation (a machinelearning approach) to elucidate the detailed spatial patterns of
carbon density and then estimated the national total carbon pools
(for details, see SI Appendix, Text S2). The biome-scale mean
carbon densities based on the Random Forest simulation showed a
good coincidence with those based on the area-weighted average
approach (SI Appendix, Fig. S2). The overall carbon stock of forests, shrublands, and grasslands totaled 64.17 ± 1.92 Pg C, which is
highly consistent with our estimate using the area-weighted average approach (62.93 ± 3.39 Pg C) (SI Appendix, Table S3).
Compared with these three ecosystems, the cropland ecosystem had lower biomass carbon density (3.06 ± 0.87 Mg C ha−1),
Tang et al.
but similar soil carbon density (92.04 ± 4.06 Mg C ha−1) (SI
Appendix, Table S2). Higher values occurred in the northeastern
regions, followed by the southwestern regions, while lower values
were found in the dry areas in northern China. Overall, the total
carbon pool of China’s croplands was estimated as 16.32 ± 0.41
Pg C (SI Appendix, Table S3).
Carbon Allocation Between Below- and Aboveground Biomass and
Between Soil and Vegetation. Both above- and belowground bio-
Fig. 3. Frequency distribution of carbon densities of different carbon sectors in China’s forests, shrublands, and grasslands. (A, D, and G) Aboveground biomass (AGB). (B, E, and H) Belowground biomass (BGB). (C, F, and I)
Soil organic carbon (SOC). Line in A–C: forests; line in D–F: shrublands; line in
Tang et al.
Fig. 4. Relationships between carbon density and MAT and MAP in forests,
shrublands, and grasslands in China for two MAT groups (≤10 °C and >10 °C)
and two MAP groups (≤400 mm and >400 mm). (A and B) Vegetation biomass carbon. (C and D) Litter carbon. (E and F) Soil organic carbon. (G and H)
Whole-ecosystem carbon. Each dot shows the average carbon density within
each 1 °C MAT and 100 mm MAP.
and soil carbon) respond to climatic regimes under different
climatic conditions, as these two climatic thresholds are important
in China’s climatic classification (13). As a result, the spatial
pattern of the carbon density showed a strong correlation with the
climate variables (Fig. 4 and SI Appendix, Fig. S6). In general, the
total carbon density and all carbon sectors (biomass, litter, and
soil) decreased with increasing MAT but had a lower decreasing
rate in the regions where the MAP exceeded 400 mm. By contrast,
they increased with increasing MAP and showed a higher increasing rate in the regions in which MAT < 10 °C (Fig. 4).
Furthermore, we found a close relationship between ecosystem
carbon density and the wetness index (P/PET, a surrogate of the
moisture index that indicates the ratio of precipitation to potential
evapotranspiration) (r2 = 0.92, P < 0.0001) (SI Appendix, Fig. S7)
(14). Interestingly, an annual P/PET value of 1.0 strongly corresponded to the ecosystem carbon density value of 100 Mg C ha−1
and to the threshold to segment the linear relationship between
carbon density and the wetness index. Specifically, the carbon
density showed a strong correlation with P/PET when the density
was ≤1.0 (r2 = 0.96, P < 0.0001); otherwise, the correlation was
poor when the density was >1 (r2 = 0.16, P = 0.154). These results
suggest that carbon density exhibits various feedbacks to climate
under different moisture conditions.
Effects of Human Activities on Carbon Stocks. To examine the effects of human activities on different carbon sectors in forests,
shrublands, and grasslands, we divided all field sites into two
PNAS | April 17, 2018 | vol. 115 | no. 16 | 4023
Effects of Climatic Factors on Carbon Stocks. To illustrate relationships between ecosystem carbon stocks and climatic variables, we divided all of the field data into two groups according
to a mean annual precipitation (MAP) of 400 mm (i.e., the
threshold of an arid climate) and a mean annual temperature
(MAT) of 10 °C (i.e., the threshold of a warm temperate climate)
to detect how ecosystem carbon sectors (total, biomass, litter,
mass carbon densities varied among forests, shrublands, and
grasslands (Fig. 3). The site-averaged aboveground biomass
carbon densities were 42.5 ± 4.6 (mean ± 1 SD) Mg C ha−1 in
forests, 3.3 ± 4.6 Mg C ha−1 in shrublands, and 0.4 ± 0.6 Mg C
ha−1 in grasslands, respectively. Their site-averaged belowground
biomass carbon densities were 10.7 ± 7.1, 3.1 ± 4.6, and 3.5 ±
4.8 Mg C ha−1, respectively. The allocation of below- to aboveground biomass carbon (root to shoot ratio, or RS ratio) differed
markedly among forests and shrublands (SI Appendix, Fig. S4),
and the biomes in each vegetation group (SI Appendix, Fig. S5A).
The site-averaged soil carbon densities showed greater variations than did biomass carbon densities (Fig. 3). The mean SOC
densities were 126 ± 98.1 Mg C ha−1 in forests, 60.2 ± 83.2 Mg C
ha−1 in shrublands, and 58.4 ± 69.3 Mg C ha−1 in grasslands. The
ratio of soil to biomass carbon density showed large variation
across sites within vegetation groups (SI Appendix, Fig. S4).
Compared with forests, shrublands showed much larger ratios of
soil carbon to vegetation biomass carbon because of the relatively smaller vegetation biomass densities in shrublands (SI
Appendix, Figs. S4 and S5B).
groups based on the degree of human disturbance: sites with
intensive human influences, which included forest plantations
and intensively grazed grasslands, and other sites with fewer
human influences, which included natural forests, primary
shrublands, and natural or less-grazed grasslands (SI Appendix, Text
S5). Our results indicate that intensive human activities have reduced both the above- and belowground biomass of most vegetation types (SI Appendix, Fig. S5A) with overall reductions of 21%
(r2 = 0.96, P < 0.001) for aboveground biomass and 24% (r2 = 0.61,
P < 0.01) for belowground biomass (SI Appendix, Fig. S8). Interestingly, the reduction in belowground biomass was almost
proportional to the reduction in aboveground biomass, thus
resulting in insignificant changes in the RS ratio. In contrast to
forests and shrublands, human activities have significantly reduced
aboveground biomass in two of the four grassland types, but they
have not consistently decreased the belowground biomass, leading
to an elevated RS ratio in heavily influenced grassland sites.
However, human disturbance did not exert significant effects on
soil carbon stocks for all biome types; the overall SOC density of
14 vegetation types with intensive disturbances was approximately
equivalent to that with fewer intensive effects (slope of linear regression = 0.97, r2 = 0.61, P < 0.01) (SI Appendix, Fig. S8D).
Comparison of Carbon Pools with Previous Estimates. The extensive
field survey in the present study has provided a full picture of the
ecosystem carbon stocks in the forests, shrublands, grasslands,
and croplands of China. Our estimate of China’s forest biomass
carbon density was higher than that in previous studies (55.7 vs ...
Purchase answer to see full