Integrated global assessment of the natural forest carbon potential

Abstract

Forests are a substantial terrestrial carbon sink, but anthropogenic changes in land use and climate have considerably reduced the scale of this system¹. Remote-sensing estimates to quantify carbon losses from global forests^2-5 are characterized by considerable uncertainty and we lack a comprehensive ground-sourced evaluation to benchmark these estimates. Here we combine several ground-sourced⁶ and satellite-derived approaches^2,7,8 to evaluate the scale of the global forest carbon potential outside agricultural and urban lands. Despite regional variation, the predictions demonstrated remarkable consistency at a global scale, with only a 12% difference between the ground-sourced and satellite-derived estimates. At present, global forest carbon storage is markedly under the natural potential, with a total deficit of 226 Gt (model range = 151-363 Gt) in areas with low human footprint. Most (61%, 139 Gt C) of this potential is in areas with existing forests, in which ecosystem protection can allow forests to recover to maturity. The remaining 39% (87 Gt C) of potential lies in regions in which forests have been removed or fragmented. Although forests cannot be a substitute for emissions reductions, our results support the idea^2,3,9 that the conservation, restoration and sustainable management of diverse forests offer valuable contributions to meeting global climate and biodiversity targets.

Fig. 1. The global distribution of tree carbon observations and the impact of human disturbances.

a, Map of ground-sourced aboveground tree carbon observations (GFBI data; aggregated to 30-arcsec (1-km²) resolution). b, Satellite-derived ESA-CCI map of current aboveground tree carbon stocks (1-km resolution). c,f, Observed biome-level tree carbon densities in existing forests based on ground-sourced (c) and satellite-derived (f) data. d,g, Principal component analysis (top two principal components shown) of the eight human-activity variables either directly or indirectly reflecting human-caused forest disturbances or the lack thereof, such as land-use change, human modification, cultivated and managed vegetation and wilderness area, to detect the effect of human disturbance on tree carbon densities for the ground-sourced (d) and satellite-derived data (g). e,h, Partial regression of the global variation in forest carbon density along the human-disturbance gradient (represented by the first principal component of the eight human-activity variables; see panels d and g) for the ground-sourced (e) and satellite-derived data (h), controlling for 40 environmental covariates. Relative carbon density is the observed carbon density divided by the global average.

Fig. 2. The natural tree carbon potential under current climate conditions in the absence of humans.

a,b, The total living tree carbon potential of 600 Gt C within the natural canopy cover area of 4.4 billion ha². c,d, The differences between current and potential tree carbon stocks, totalling 217 Gt C. e,f, The difference of tree carbon potential between the GS and SD models, subtracting the mean values of the six SD models from the mean values of the four GS models. Blue colours indicate that the GS models predict higher potential than the SD models, whereas red colours indicate the opposite. b,d,f, Latitudinal distributions (mean ± standard deviation) of the total tree carbon potential for the GS1, GS2, SD1 and SD2 models (b), the difference between current and potential tree carbon (d) and the difference of tree carbon potential between the GS and SD models (f). Maps represent the average estimates across all GS and SD models and are projected at 30-arcsec (about 1-km²) resolution. We show dryland and savannah biomes with stripes to denote that many of these areas are not appropriate for forest restoration. Where trees would naturally exist, they often exist far below 100% canopy cover, and restoration of forest cover should be limited to natural conditions.

Fig. 3. The living tree carbon potential estimated from the ground-sourced (GS1 and GS2) and satellite-derived (SD1 and SD2) models.

a, Total estimated living tree biomass potential of the GS1, GS2, SD1 and SD2 models. Error bars represent the lower and upper boundaries based on the 5% and 95% quantiles from a bootstrapping procedure. Colours represent the different input datasets, that is, upper or lower canopy cover boundaries (GS models) and ESA-CCI, Walker et al. or harmonized (SD models). Light colours above white lines indicate the difference between current and potential tree carbon stocks. b, Meta-analysis showing literature estimates of living tree carbon potential based on ensemble models^,,, inventory data^,– and mechanistic^– or data-driven models. The horizontal dashed line represents the average existing living tree carbon of 443 Gt C estimated in these publications. c, Differences between current and potential tree carbon stocks. d, Literature estimates for the difference between current and potential tree carbon stocks from ref.  (ensemble models), refs. ^,,, (inventory data), refs. ^, (mechanistic models) and ref.  (data-driven models).

Fig. 4. Sources of uncertainty in forest carbon potential for the GS and SD models.

a,b, Relative contribution of individual uncertainty sources to the overall uncertainty in carbon potential for the GS (a) and SD (b) models: (1) model approach (type 1 versus type 2 models); (2) input data (current aboveground tree carbon input, that is, upper and lower canopy cover boundaries for GS models and ESA-CCI, Walker et al. and harmonized for SD models); (3) aboveground biomass potential estimates (bootstrapping); (4) belowground biomass (accounting for uncertainties in both root mass fraction and aboveground biomass); (5) dead wood and litter (accounting for uncertainties in both dead wood and litter-to-tree biomass ratios and tree biomass); and (6) soil organic carbon potential. The maps show the top uncertainty source within each pixel. The pie charts show the relative contribution of uncertainties worldwide.

Fig. 5. Contribution of land-use types, forest types, carbon pools and countries to the difference between current and potential ecosystem-level carbon stocks.

a, Of the 328 Gt C discrepancy between current and potential carbon stocks, 226 Gt C is found outside urban and agricultural (cropland and pasture) areas, with 61% in forested regions in which the recovery of degraded ecosystems can promote carbon capture (conservation potential) and 39% in regions in which forests have been removed (restoration potential). b, Relative contribution of forest degradation (conservation potential; blue area) and land-cover change (orange colours) to the difference between current and potential ecosystem-level carbon stocks. The darker blue area represents the conservation potential of 10.5 Gt C in forest plantation regions. c, Relative contribution of tropical, temperate, boreal and dryland forests to the total forest conservation potential. d, Relative contribution of the three main carbon pools (living biomass, dead wood and litter, and soil) to the difference between current and potential carbon stocks. e, The nine countries contributing more than 50% to the difference between current and potential carbon stocks.