全球变化知识资源中心

Over the past two decades, the terrestrial biosphere has acted as a sink for atmospheric CO₂, removing on average approximately 2.5 petagrams of carbon per year (PgC yr⁻¹): equivalent to 25% of fossil fuel emissions^{1, 2, 3, 4}. Additional emissions from land-use change reduce the global net land sink to approximately 1.5 PgC yr⁻¹, with forests playing a dominant role⁵. Monitoring C stock changes over time can be used to determine which ecosystems and processes drive changes in C fluxes and to develop strategies for climate change mitigation. The existing global ABC estimates based on field survey data, such as the recent global synthesis by Pan et al.⁵, provide snapshots in time and are limited to forest ecosystems only. Estimating ABC using satellite remote sensing can provide a more consistent methodology and global coverage⁶. Although offering high spatial resolution, current remote sensing products have limited temporal frequency and record length at the global scale^{6, 7, 8}.

Here, we derive global ABC estimates for all vegetation types for the past two decades using an entirely new approach that uses satellite-based passive microwave data rather than the optical or radar observations used previously. The intensity of natural microwave radiation from the Earth is a function of its temperature, soil moisture and the shielding effect of water in aboveground vegetation biomass, including canopy and woody components^{10, 11, 12}. The biomass signal is captured in the vegetation optical depth (VOD; refs 13, 14). A distinct advantage of passive microwave-derived VOD is that it remains sensitive to biomass variations at a relatively high biomass density (for example, rainforests), whereas optical-based remotely sensed vegetation products rapidly saturate^{14, 15}. ABC estimates can be derived for all vegetation types, not only forests, as a suitable harmonized global VOD record exists from the 1990s onwards¹⁴.

The main disadvantage of this technique is the relatively coarse spatial resolution (>10 km), which is a consequence of the low energy of the Earth’s natural microwave emissions. This means that individual plot measurements cannot be used directly to establish a relationship between VOD and ABC. Instead, we use an indirect calibration method based on spatial ‘snapshot’ ABC estimates from Saatchi et al.⁶, who combined three types of satellite observations with plot-based measurements to estimate ABC in tropical regions (see Methods and Supplementary Information).

We estimate total global ABC at 362 PgC with a 90% confidence interval (CI₉₀) of 310–422 PgC circa 2000 (that is, 1998–2002; Fig. 1a). ABC values per region and biome and annual ABC change rate are very close to values reported in other studies when the same categorization, definitions and assumptions are applied^{5, 6, 7} (see Supplementary Information for details). Our ABC estimates for boreal forests (CI₉₀ = 37–66 PgC) and temperate forests (CI₉₀ = 24–39 PgC) circa 2000 overlap with inventory-based estimates (44 and 36.4 PgC, respectively) by Pan and colleagues⁵. For tropical forests, our ABC estimates are comparable to Pan et al.⁵ (205 versus our 195; CI₉₀ = 180–208 PgC circa 2000), Saatchi et al.⁶ (173–212 versus our 211; CI₉₀ = 194–226 PgC for 10% tree cover threshold circa 2000) and Baccini et al.⁷ (159 versus our 157; CI₉₀ = 146–166 PgC circa 2007/8). For the savannahs and shrublands of the pan-tropics (excluding Australia), Baccini et al.⁷ reported 51 PgC ABC circa 2007/8, which is similar to our estimate of 49 (CI₉₀ = 42–56) PgC (see Supplementary Information).

a, Total ABC in eight biomes circa 2000 (mean estimate; error bars indicate the 90% confidence interval). ‘Tropical forests’ include those in Southeast Asia, Africa and the Americas (that is, South America, Caribbean countries and Mexico); remaining forests are considered as ‘boreal/temperate’. ‘Shrublands’ includes both open and closed shrublands; ‘Croplands’ includes both ‘croplands’ and ‘cropland/natural vegetation mosaic’ based on the MODIS IGBP land-cover map for 2001 (ref. 16). b, ABC density per unit area circa 2000. The bottom, middle and top band of the box represent the 25th, 50th (median) and 75th percentile, respectively, and the ends of the whiskers represent the 5th and 95th percentile for all corresponding grid cells. c, Annual trends in total biome ABC for 1993–2012 (mean estimate; error bars indicate the 90% confidence interval). The classification relates to year 2001 and grouping of different biomes to the same colour is mainly based on woody vegetation canopy cover, that is, 10–60% for shrublands, savannahs and woody savannahs and less than 10% for grasslands. Croplands with harvest and thus considerable variation in woody components are grouped with grasslands.

A map of ABC trends over the period 1993–2012 (Fig. 2) reflects spatial changes in the underlying VOD data, which were attributed to their main natural and anthropogenic drivers in a previous study¹⁷. Prominent features of these trends include ABC changes due to widespread tropical forest clearing in several countries (for example, Brazil and Indonesia); rainfall variability in water-limited ecosystems (for example, savannahs and shrublands) causing an ABC increase in southern Africa and northern Australia; regrowth on abandoned farm land in former communist countries in boreal and temperate regions; and pest attacks and wildfires in the boreal forests of northeast Russia and Canada and the temperate forests of the USA.

Global carbon budget analyses^{1, 3, 25} and CO₂ atmospheric inversion studies⁹ suggest a terrestrial net carbon sink (measured as the net exchange between the atmosphere and all lands) of approximately 1.4 ± 0.4 PgC yr⁻¹. Using biome-specific ratios of aboveground and total biomass carbon (ABC/TBC; ref. 26) and ratios between TBC and total carbon stock (see Supplementary Table 3 in Pan et al.⁵), combined with the MODIS IGBP dynamic land-cover maps for 2003–2012, we estimated trends in mean forest ABC and total forest carbon of +0.10 and +1.19 PgC yr⁻¹, respectively, for this period (Table 1). Global total forest area was similar for 2003 and 2012, but there was a reduction in tropical forests and an increase in boreal and temperate forests. The very different carbon structure of tropical and boreal forests plays a crucial role in increasing the global total forest carbon stock: tropical forests store 44% of total carbon in aboveground biomass but boreal forests only 15%, with the remainder contained in living roots, litter and soil organic carbon⁵. Total biomass carbon in non-forest biomes increased by +0.63 PgC yr⁻¹ (Table 1), with an unknown change in non-living carbon. Combined, these numbers suggest a net terrestrial sink of at least +1.82 PgC yr⁻¹ for 2003–2012, still consistent with that estimated from the global carbon budget^{1, 3, 9}. The main uncertainty in the rate of increase is associated with the slow and passive carbon pools of soil organic matter, which are expected to take longer to respond than short-term biomass variations.

Full table

The ABC estimates were based on harmonized VOD data for 1993 onwards derived from a series of passive microwave satellite sensors¹⁴, including Special Sensor Microwave Imager (SSM/I), Advanced Microwave Scanning Radiometer for Earth Observation System (AMSR-E), FengYun-3B Microwave Radiometer Imager (MWRI) and Windsat. VOD values affected by the presence of inland water bodies were corrected by considering nearby grid cells with the same landscape type. Values missing owing to frost conditions were estimated using two alternative assumptions about ABC decreases in winter. An empirical relationship was established to convert VOD to ABC by calibrating against the aboveground biomass map for tropical regions from Saatchi et al.⁶ to predict the mean and 90% confidence interval (CI₉₀; see Section 2.1 in Supplementary Information). We combined these with the two frost corrections to produce six ABC estimates and report the mean and the CI₉₀ range of the resulting analyses. Global estimates of ABC stocks and changes were compared to ABC values estimated from Pan et al.⁵, Baccini et al.⁷ and Harris et al.⁸; taking care to match the reported biome area, measurement period, and definition of countries and regions. Pan et al.⁵ reported TBC only, but we could infer ABC from the description of methods. A global land-cover map based on the MODIS Collection 5 IGBP classification product (MCD12C1; ref. 16) was re-sampled to 0.25° by dominant land use to calculate ABC per biome. The Global Precipitation Climatology Centre (GPCC) precipitation data for 1993–2012 were used in interpretation³⁰. To estimate the TBC from our ABC values, we applied the conversion factors used by Pan et al.⁵ for different forests and used literature values for non-forest vegetation²⁶. Further details are provided in the Supplementary Information. The aboveground biomass carbon (ABC) data set derived and used in this study over the period 1993–2012 can be accessed at http://www.wenfo.org/wald/global-biomass and http://hydrology.unsw.edu.au/downloads/data/global-biomass

1. Le Quéré, C. et al. Global carbon budget 2013. Earth Syst. Sci. Data 6, 235–263 (2014).
   • Article
2. Le Quéré, C. et al. Trends in the sources and sinks of carbon dioxide. Nature Geosci. 2, 831–836 (2009).
   • CAS
   • ISI
   • Article
3. Canadell, J. G. et al. Contributions to accelerating atmospheric CO₂ growth from economic activity, carbon intensity, and efficiency of natural sinks. Proc. Natl Acad. Sci. USA 104, 18866–18870 (2007).
   • PubMed
   • Article
4. Peters, G. P. et al. The challenge to keep global warming below 2 °C. Nature Clim. Change 3, 4–6 (2013).
   • ISI
   • URL: