Soil carbon stocks

Many estimates of global SOC stocks have been published during the past seven decades (Figure 1, discussed in more detail below), most recently to support the calculation of potential CO₂ emissions from the soil under LULCC and climate change scenarios. Although most studies report a global SOC estimate of roughly 1500 petagrams of carbon (Pg C; 10¹⁵ g or billion tons of carbon), there is considerable variation (median across all estimates: 1460.5 Pg C, range 504–3000 Pg C, n = 27 studies).

The earliest estimate of global SOC that we found extrapolated the carbon content of nine soils in the USA (published in a 1951 textbook) to 710 Pg C globally ( cited in ). A more rigorous calculation was provided by Bohn who estimated a global SOC pool of 2946 ± 500 Pg C based on FAO soil maps for South and North America, Asia, Africa, Europe and Oceania separately . More recent studies incorporated larger numbers of soil profiles in their estimates, for example Bohn using 187 profiles to 1 m depth derived a total SOC pool of 2207 Pg C , whereas Batjes, using thousands of soil profiles in the World Inventory of Soil Emission Potentials (WISE), estimated global SOC stocks of 1462–1548 Pg C . Other inventory-based approaches to estimate soil carbon stocks involved area-weighted extrapolation either from carbon data for 11 ecosystems, which resulted in a global total of 1456–1515 Pg C , or from carbon estimates for 11 soil orders under five land use types, resulting in 1477 Pg C globally .

More recently, estimates of global SOC stocks have been based on more detailed mapped data (see also next section). The earliest map of SOC estimated total global stocks to 1 m depth to be 1457 Pg C under ‘primeval’ undisturbed conditions and 504 Pg C under hypothetical fully disturbed conditions . The latest, most detailed, global SOC map based on the Harmonized World Soil Database , reported total stocks of 1417 Pg C under partially disturbed conditions when bulk densities for organic soils are derived from a pedotransfer function .

The wide variation in estimates of global SOC stocks reflects the disparity in sampling period, intensity and spatial resolution of the soil profile databases, and differences in approaches to calculations (Figure 1). For example, the 10,253 georeferenced soil profiles available in the latest version of the WISE collected from 1925 onwards are unevenly distributed with most from Africa (41%, largely sub-Saharan), Asia (18%, few from China, Mongolia and Kazakhstan) and South America (18%), and very few profiles for North America (8%), Oceania (2%) and the north temperate regions (Canada, Fennoscandia and Russia) .

The SOC stock estimates can also vary because some studies included inorganic soil carbon, or varying levels of stone content or disturbance , in the carbon stock estimates. This is difficult to disentangle as most studies do not state explicitly which carbon pools, stone content and levels of disturbance were included.

In some cases, the same authors using the same basic data come up with different total SOC stock estimates (Figure 1). In a recent estimate of global SOC stocks (Table 18 in Hiederer and Köchy ) notable differences in global stock estimates were attributed to the values used for the bulk density of organic soils. Such large differences in global SOC stock estimates, greater than the total amount of carbon in the atmosphere (1053 Pg C vs about 816 Pg C as CO₂), highlight the need for caution to be applied during data processing and interpretation of differences in estimates, and for continued improvements in data collection and processing to derive better global SOC stock estimates.

Distribution of soil carbon

Because ecosystem carbon dynamics vary greatly depending on the balance between soil and phytomass carbon stocks , it is important to understand the distribution of soil carbon in relation to phytomass carbon. While phytomass carbon has been mapped by several studies , few have attempted to map the distribution of SOC globally.

Most maps of SOC have been coarse both in spatial resolution and in the number of soil parameter estimates considered. The first map at approximately 1° × 1° resolution (˜100 × 100 km at the equator) used multiple regression to extrapolate soil carbon globally to 1 m depth from 50 carbon measurements in north temperate regions based on evapotranspiration, soil moisture deficit and land use . Later, more detailed global maps combined soil profiles with versions of the FAO soil maps, for example, Batjes combined 4,353 soil profiles distributed globally from the WISE database with the 1:5,000,000 FAO/United Nations Educational, Scientific and Cultural Organization (UNESCO) Soil Map of the World to produce a 0.5° × 0.5° resolution (˜55 × 55 km) map, whereas the Global Soil Data Task Group compiled a map at 5 × 5 arc min (˜9 × 9 km) resolution also based on the FAO digital soil map . Another product based on the FAO-UNESCO soil data is the ‘Soil Organic Carbon Map’ from the US Department of Agriculture-Natural Resources Conservation Service from 2006, with six classes of SOC density mapped to a grid resolution of 2 arc min (˜4 × 4 km) .

The most recent and most detailed globally consistent and continuous map of SOC (Figure 2A), based on the Harmonized World Soil Database , combines information from 9,607 soil profiles in the WISE database version 2.1 and soil profiles from regional SOil and TERrain (SOTER) databases with 16,107 soil mapping units/polygons, converted to a 30 × 30 arc second (˜1 × 1 km) spatial resolution raster, derived from four spatially explicit soil databases: the European Soil Database , the soil map of China , regional SOTER databases for Central and Southern Africa, Latin America and the Caribbean, Central and Eastern Europe, and the FAO-UNESCO Soil Map of the World .

The spatial distribution of SOC differs substantially from that of carbon stored in above- and below-ground phytomass (Figure 2 A & B). Both SOC stocks and their contribution to total carbon stock vary with latitude and among IPCC climatic regions (Figure 3 & Table 1). Most of the SOC is stored at northern latitudes, particularly in the northern permafrost regions (included in ‘boreal moist’ in Figure 3), one of the regions most vulnerable to climate change . In contrast, the largest quantities of phytomass carbon can be found near the equator in wet and moist tropical forests (Figures 2A & 3).

However, these distributional data are not as accurate as they may appear to be. First, despite the most recent SOC map being notionally at 1 km spatial resolution , and some regions such as China and Europe have been mapped at that level of detail, it relies largely on the coarse resolution FAO soil maps from the 1970s (except in Central and South America, Caribbean, East Africa, Central Asia and Mongolia, which are derived from SOTER databases at medium spatial resolution). Furthermore, the existing distributional data most probably underestimate SOC stocks in some parts of the world, because they do not comprehensively address SOC stocks at depth. The current global SOC estimates and maps usually provide data to 1 m depth only, which may be sufficient in most mineral soils because organic carbon content declines with depth, however many organic soils are much deeper and these deep organic soils have high concentrations of SOC. For example, many peat soils in both the tundra and tropics have been reported to be, respectively, over 3 m deep containing 1672 Pg C and up to 11 m deep containing 89 Pg C , effectively doubling the global SOC stock estimates based on profiles to only 1 m depth. Jobbágy and Jackson report that measurements to 3 m depth yielded estimates of SOC stocks 1.5-times larger than those to 1 m depth , and these deep carbon pools might also be affected by LULCC . Overall there is a strong need to better understand the global distribution of SOC stocks in order to improve the understanding of the impacts of LULCC on SOC stocks and emissions to inform management decisions .

Modeling soil carbon

A wide variety of regional and global models have attempted to estimate soil carbon , often in conjunction with modeling other carbon pools – for example in vegetation and atmosphere – to assess fluxes of carbon between pools, and especially to investigate the effects on CO₂ in the atmosphere within Earth System Models. Uncertainty in modeled estimates of soil carbon is very large. For example, Todd-Brown et al. compared 11 models that gave global soil carbon stock estimates ranging from 510–3040 Pg C ; however the mean (±standard error) of the 11 model results was 1520 ± 770 Pg C, within 10% of the most recent global estimates (Figure 1), and three of the 11 models estimated global SOC stocks within 10% of the most recent mapping estimate of 1417 Pg C. These substantial discrepancies in modeled global totals and the distribution of soil carbon stocks are probably caused by uncertainties in the data, differences in the type of model and the inputs used (e.g., net primary production and temperature), and in the parameterization of turn-over times (e.g., decomposition rates), as well as the spatial resolution of the models . Modelers require more accurate SOC stock and flow data as inputs to their models, and for comparison with model outputs to benchmark and improve their models. As soil carbon stocks are relatively large compared with the atmospheric CO₂ carbon pool and are sensitive to climate change (recent reviews include Lal and Falloon et al.), more realistic representation of SOC stocks and soil carbon dynamics in Earth System Models are essential for more reliable predictions of future climate.

Impacts of LULCC on soil carbon & emissions

While the large biomass carbon changes resulting from LULCC are central to international policy agendas on reducing GHG emissions, the effects of LULCC on SOC stocks and consequent emissions, which may also be substantial, have not been fully considered in discussions of emissions reductions.

In some cases, converting native vegetation to cropland has resulted in losses of 25–50% of the SOC in the top 1 m , while conversion to pasture typically has resulted in smaller losses of SOC . Furthermore, the impact of LULCC and management on SOC is dramatically different in mineral versus organic soil types . On mineral soils, conversion of native grassland and forest to cropland can result in the loss of 20–40% of the original soil carbon stocks ; whereas on organic soils, the loss of SOC varies more, depending on depth of the drainage, climatic and hydrological conditions, as well as management and inputs of organic matter .

Concerns about large changes in carbon stocks have prompted estimation of carbon emissions. For example, estimation of emission/mobilization of SOC stocks from forests suggest that although soil carbon in forests has remained reasonably stable at around 380 Pg C from 1990 to 2007, there is substantial variation among regions . During 17 years, soil carbon stocks in boreal and temperate forests increased by 4.5 and 7.6% (from 167.0 to 174.5 Pg C and from 52.7 to 56.7 Pg C), respectively, as forest area expanded after agricultural abandonment and immature forests grew with reduced harvesting. However, soil carbon in tropical forests declined by 7.7% between 1990 and 2007 (from 164.0 to 151.3 Pg C), largely because of SOC loss caused by deforestation within intact tropical forest (16.7% loss, from 139.9 to 116.6 Pg C), compensated somewhat by increases within tropical regrowth forests (44% increase, from 24.1 to 34.7 Pg C) .

In addition to uncertainties around these estimates arising from variation in measurement approaches and methods, large uncertainties also arise from the differing responses of SOC stocks to LULCC and climate change. For example, a recent assessment doubled the estimate of emissions from disturbance to tropical peat . However, despite these large uncertainties in emissions estimations, there is evidence that SOC is a large source of atmospheric CO₂ and potentially an important sink.

The role of soil carbon management

For large portions of the world, SOC is the largest component of total carbon stock (Figure 2C). This is particularly true in regions that are not naturally forested, especially those with highly organic soils, and those that have already lost their natural vegetation. In these areas, where SOC may be an important source of emissions, there is both a need and an opportunity for policy and management to address the role of soil in global carbon emissions.

Supporting decisions on soil carbon management requires an improved understanding of the spatial variation in SOC stocks and their importance relative to phytomass carbon, as well as of the potential to use management approaches to reduce SOC losses and/or increase stocks. Identifying the potential for management targeted specifically at SOC and the options available is contingent on understanding where soil carbon plays a large role, especially relative to phytomass carbon. This is the case in over 60% of global land area, where SOC constitutes the vast majority of the total carbon stock (more than 75% of the combined soil and phytomass carbon stocks; areas shown in brown in Figure 2C). This may be because of low phytomass carbon, either naturally or as a result of disturbance that has removed vegetation, and/or because of the presence of highly organic soils. In such areas, climate change mitigation efforts may need to emphasize promoting land use practices that conserve soil carbon (and potentially restore native vegetation). This is in contrast to approximately one-fifth (18%) of global land area where SOC constitutes less than 50% of combined carbon stock (areas shown in dark green in Figure 2C), where promoting land uses that maintain phytomass carbon is more beneficial to overall carbon stocks and important for emissions reductions.

In some cases, SOC losses due to conversion can be limited or mitigated through management approaches specifically targeting soil carbon stocks in agricultural systems, and/or in SOC recovery associated with re-vegetation or succession . Management approaches can include manuring and fertilizing, no tillage, conservation tillage, reduced tillage, management of crop residues and cover cropping . For example, in a 160-year experiment adding farmyard manure to cropland increased soil organic matter, with greatest increases in the early years and limited change after 80 years . Application of inorganic fertilizers can also increase SOC, especially where SOC concentrations are low; in the North China Plain fertilization with nitrogen and phosphate together increased SOC by about 250 kg C ha^-1 year^-1 over 18 years . No-till and other forms of conservation agriculture can limit losses and even cause an increase in SOC. For example, under conservation agriculture SOC increased in Brazil by 800 kg C ha^-1 year^-1 (median of eight studies) and in France by 200 kg C ha^-1 year^-1. Increases in SOC can also be achieved through agroforestry, application of biochar, perennial crops, reforestation and afforestation, and other forms of management that affect land cover and land use. For example, a meta-analysis showed that SOC on cropland and pastures in tropical, subtropical and boreal zones increased 30 years after afforestation with hardwoods, but not with softwoods .

While maintaining and enhancing SOC stocks, soil carbon management is likely to not only improve climate regulation, but also have beneficial and/or detrimental effects on other ecosystem services . For example, increased water and nutrient storage, and greater resistance to erosion, are potential co-benefits of increasing the soil organic matter and thereby carbon content in soils . However, other soil carbon management actions, such as maintaining anaerobic conditions to increase soil carbon stocks in peatlands, can result in trade-offs, such as potentially increasing the risk of acidification, and increasing methane and N₂O emissions from peatlands . When deciding on soil carbon management actions, the co-benefits and trade-offs with other ecosystem services should be identified and considered to develop policies with multiple ‘wins’.

Conclusion

The available data and analyses presented above suggest that in some parts of the world managing soil carbon may be of great importance to global carbon cycles and climate change mitigation. However, many uncertainties remain that have implications both for science and policy.

Better estimates of SOC stocks and SOC dynamics are needed for improved understanding of the carbon balance and potential for climate change mitigation, in particular to improve SOC representation in Earth System Models. Existing estimates could be improved by systematic collection of more data on soil profiles, particularly in the tropics and permafrost regions, and including sampling to greater depths. Using standardized sampling methods and making data available will enable the combination of data from different sources to provide better global and regional syntheses. The methods used to calculate SOC stocks also need to be standardized to consistently take appropriate account of the spatial variation in depth, bulk density and gravel content to permit comparisons among SOC stock estimates.

Current maps of soil carbon rely largely on the FAO soil maps, which are based on data collected prior to the 1970s and that need to be updated. A number of ongoing efforts aim to generate better mapped soil data, including on SOC, such as the GlobalSoilMap project at a global scale . Recent advances are particularly notable for Africa, where a multi-institutional collaboration has compiled an extensive library of soil-profile information and is generating mapped data, including on SOC . Improved maps and understanding of SOC distribution also depend on increased sampling intensity, and may draw on new sources of data including remote sensing. For example, remote sensing can be used to successfully estimate the depth and amount of organic matter in soils , detect variation in leaf nitrogen content , likely associated with low rates of litter decomposition and may therefore indicate higher SOC, and determine SOC of topsoil using mid-infrared and combined diffuse reflectance spectroscopy or soil color ; however, most remote sensing studies so far have been performed at relatively small scales, plot (<1 km²) to regional scales. These approaches may also improve monitoring of changes over time in SOC (and phytomass carbon) and factors that affect it.

Apart from better estimates of stocks and their distribution, we also need better data on the impacts of LULCC, and management on SOC and its dynamics in time and space . These should address the loss of SOC under different conversion/land uses and how it varies regionally, and should also provide information of which stocks are most vulnerable (e.g., to what depth).

Finally, the uncertainties surrounding carbon stock estimates, their distribution and impacts of management need to be quantified and explained clearly. Amongst these uncertainties is the question of whether and how climate change impacts may affect SOC stocks and fluxes , and how they may interact with management impacts on SOC.

Future perspective

A broader subset of the scientific community needs to interact to address issues relating to soil carbon stocks and their spatial distribution, as well as their responses to LULCC and to management. While soil scientists need to continue their involvement, there is also a need for interdisciplinary work incorporating input from remote sensing scientists, ecologists, agronomists, earth system modelers, spatial analysts, geographers, land-use planners and others. Integrating this diverse set of expertise is required to enhance our understanding of the issues discussed in this article.

For their part, policy- and decision-makers need to recognize the potential importance of SOC in the global carbon cycle and climate change mitigation. Some national policies already address soil carbon, for example, the ‘Low-Carbon Agriculture’ plan (the ABC Program) which Brazil launched at the 2009 UNFCCC Conference of Parties in Copenhagen . In pursuing such developments, policymakers should consider the following issues: First, SOC losses and gains are likely to make a large overall contribution to atmospheric CO₂ concentrations, and this may have implications for when and how SOC management should be recognized as a contribution to emissions reduction; second, the importance and appropriateness of SOC management is likely to vary depending on stocks, especially in relation to total organic carbon, their geographic variation, and the potential impacts of LULCC and management on those stocks. However, at present there are considerable uncertainties in our understanding of each of these factors. For example, current Earth System Models are unable to accurately represent SOC stocks and soil carbon dynamics under climate change and LULCC. While further research as outlined above will reduce uncertainties and should be supported, decision-making on land use and management of carbon will always be subject to some uncertainty, and, therefore, a precautionary approach should be considered.

An additional impetus for examining and upgrading existing knowledge and data on SOC stocks and flows is that stocks may have changed substantially since the sampling that generated some of the earlier data included in global datasets. In addition to, and potentially independently of, changes caused by LULCC, climate change will likely impact SOC stocks and flows (recent reviews include Lal, Falloon et al. and Ogle et al.).

Several initiatives to improve the data on soils are ongoing and may be given added impetus by the International Year of Soils ‘Healthy soils for a healthy life’ declared for 2015 by the FAO and partners. The International Year of Soils will serve as a platform for raising awareness on the importance of sustainable soil management as the basis for food systems, fuel and fiber production, essential ecosystem functions, and better adaptation to climate change for present and future generations . These efforts combined will strengthen the basis for appropriate management of SOC.