- Lead Authors:
- Kate Lajtha, Oregon State University
- Vanessa L. Bailey, Pacific Northwest National Laboratory
- Contributing Authors:
- Karis McFarlane, Lawrence Livermore National Laboratory
- Keith Paustian, Colorado State University
- Dominique Bachelet, Oregon State University
- Rose Abramoff, Lawrence Berkeley National Laboratory
- Denis Angers, Agriculture and Agri-Food Canada
- Sharon A. Billings, University of Kansas
- Darrel Cerkowniak, Agriculture and Agri-Food Canada
- Yannis G. Dialynas, University of Cyprus (formerly at Georgia Institute of Technology)
- Adrien Finzi, Boston University
- Nancy H. F. French, Michigan Technological University
- Serita Frey, University of New Hampshire
- Noel P. Gurwick, U.S. Agency for International Development
- Jennifer Harden, U.S. Geological Survey and Stanford University
- Jane M. F. Johnson, USDA Agricultural Research Service
- Kristofer Johnson, USDA Forest Service
- Johannes Lehmann, Cornell University
- Shuguang Liu, Central South University of Forestry and Technology
- Brian McConkey, Agriculture and AgriFood Canada
- Umakant Mishra, Argonne National Laboratory
- Scott Ollinger, University of New Hampshire
- David Paré, Natural Resources Canada, Canadian Forest Service
- Fernando Paz Pellat, Colegio de Postgraduados Montecillo
- Daniel deB. Richter, Duke University
- Sean M. Schaeffer, University of Tennessee
- Joshua Schimel, University of California, Santa Barbara
- Cindy Shaw, Natural Resources Canada, Canadian Forest Service
- Jim Tang, Marine Biological Laboratory
- Katherine Todd-Brown, Pacific Northwest National Laboratory
- Carl Trettin, USDA Forest Service
- Mark Waldrop, U.S. Geological Survey
- Thea Whitman, University of Wisconsin, Madison
- Kimberly Wickland, U.S. Geological Survey
Lajtha, K., V. L. Bailey, K. McFarlane, K. Paustian, D. Bachelet, R. Abramoff, D. Angers, S. A. Billings, D. Cerkowniak, Y. G. Dialynas, A. Finzi, N. H. F. French, S. Frey, N. P. Gurwick, J. Harden, J. M. F. Johnson, K. Johnson, J. Lehmann, S. Liu, B. McConkey, U. Mishra, S. Ollinger, D. Paré, F. Paz Pellat, D. deB. Richter, S. M. Schaeffer, J. Schimel, C. Shaw, J. Tang, K. Todd-Brown, C. Trettin, M. Waldrop, T. Whitman, and K. Wickland, 2018: Chapter 12: Soils. In Second State of the Carbon Cycle Report (SOCCR2): A Sustained Assessment Report [Cavallaro, N., G. Shrestha, R. Birdsey, M. A. Mayes, R. G. Najjar, S. C. Reed, P. Romero-Lankao, and Z. Zhu (eds.)]. U.S. Global Change Research Program, Washington, DC, USA, pp. 469-506, https://doi.org/10.7930/SOCCR2.2018.Ch12
Soils
SUPPORTING EVIDENCE
KEY FINDINGS
Key Finding 1
Estimates for soil carbon stocks in the conterminous United States plus Alaska range from 142 to 154 petagrams of carbon (Pg C) to 1 m in depth. Estimates for Canada average about 262 Pg C, but sampling is less extensive. Soil carbon for Mexico is calculated as 18 Pg C (1 m in depth), but there is some uncertainty in this value (medium confidence).
Description of evidence base
The value range of soil carbon to a depth of 1 m for the United States is based on several compilations: Alaska is estimated in Mishra and Riley (2012) as 77 Pg C, an increase from the value reported by Bliss and Maursetter (2010) of 48 Pg. The sampling for the Mishra and Riley (2012) estimate is quite extensive, and land types for areal weighting are well known and documented. Modern estimates for the conterminous United States (CONUS) span the range from the U.S. Geological Survey (USGS) estimate of Sundquist et al. (2009) at 77 Pg C and the Rapid Carbon Assessment (RaCA, initiated by the Soil Science Division of the U.S. Department of Agriculture’s National Resources Conservation Service in 2010) estimate (Soil Survey and Loecke 2016) at 65 Pg C (see Table 12.1). The RaCA estimate is based on 144,833 soil samples and extrapolation using detailed soil maps. The soil carbon value of 9 Pg C for Mexico is based on Paz Pellat et al. (2016), but that estimate is based on sampling to a depth of only 30 cm. Based on conversion factors in Jobbágy and Jackson (2000), a conservative extrapolation to 1 m yields a value of 18 Pg C. The estimates for Canada are from Tarnocai (1997, 2006). This assessment recognizes that 1 m is a very arbitrary depth to consider; Batjes (1996) reported a 60% increase in the global soil organic carbon (SOC) budget when the second meter of soil was included.
Major uncertainties
There is medium high confidence in the estimates from CONUS due to new extensive and intensive sampling, although estimates for specific land-use classes still vary with different estimates. Confidence is relatively high for estimates in the agricultural areas of Canada but lower for forested areas. In Canada, uncertainty for the large peatlands areas in the boreal and Arctic regions is high due to low-sampling intensity and low-resolution mapping of peatland types. Uncertainty for estimates from Mexico are likely high due to low sampling coverage, and available data are only to a depth of 30 cm.
Assessment of confidence based on evidence and agreement, including short description of nature of evidence and level of agreement
Soil carbon was extensively sampled in three independent studies for CONUS, so the confidence for the range of values reported here is very high. Due to the complex nature of estimating soil carbon in boreal and peat regions, the uncertainty is greater surrounding values for Canada. There is low confidence in values reported for Mexico as sampling is not as extensive and the depth of sampling is not as great.
Summary sentence or paragraph that integrates the above information
The estimates of total soil carbon stores are reasonably accurate for CONUS and Canada but are less accurate for Mexico.
Key Finding 2
Most Earth System Models (ESMs) are highly variable in projecting the direction and magnitude of soil carbon change under future scenarios. Predictions of global soil carbon change through this century range from a loss of 72 Pg C to a gain of 253 Pg C with a multimodel mean gain of 65 Pg C. ESMs projecting large gains do so largely by projecting increases in high-latitude soil organic carbon (SOC) that are inconsistent with empirical studies that indicate significant losses of soil carbon with predicted climate change (high confidence).
Description of evidence base
A description of the scientific concerns with current ESMs is presented in He et al. (2016). They analyzed 14C data from 157 globally distributed soil profiles sampled to a depth of 1 m to demonstrate that ESMs currently overestimate the soil carbon sink potential. Todd-Brown et al. (2014) also pointed out major sources of error in current ESMs and suggested that most ESMs poorly represented permafrost dynamics and omitted potential constraints on SOC storage, such as priming effects, nutrient availability, mineral surface stabilization, and aggregate formation. For example, many ESMs simulated large changes in high-latitude SOC that ranged from losses of 37 Pg C to gains of 146 Pg C. The poor performance of current ESMs can result from biases in model structure, parameterization, initial values of carbon pools, and other variables (Luo et al., 2016).
There is currently a great deal of controversy over how to improve the representation of soil carbon in models (Chen et al., 2015); several authors suggest that microbial dynamics, including the priming effect, need better representation (Georgiou et al., 2015; Sulman et al., 2014; Wieder et al., 2014), as does soil carbon response to nitrogen enrichment (Janssens and Luyssaert 2009; Riggs and Hobbie 2016). However, there is no evidence that suggests how much detail is needed to adequately represent future soil carbon dynamics and soil carbon pools.
Deep carbon (>1 m in depth) generally has been found to be more stable and resistant to management or climate change than carbon in surface soils (Rumpel and Kögel-Knabner 2010; Schrumpf et al., 2013), but, given that subsurface horizons contain more than half the soil carbon (Jobbágy and Jackson 2000), small changes could significantly affect carbon budgets. Although less well studied, deep carbon has been shown to be sensitive to management practices (Alcantara et al., 2016; Ward et al., 2016).
Microbial dynamics, including the priming effect, are key controls on soil carbon turnover (Bernal et al., 2016; Guenet et al., 2012). Carbon-use efficiency of different substrates by microbes might be a key factor in soil carbon stabilization (Cotrufo et al., 2013).
Major uncertainties
How much detailed information on microbial physiology, coupled carbon-nitrogen cycles, or other processes is needed to improve soil carbon models is not well known.
Assessment of confidence based on evidence and agreement, including short description of nature of evidence and level of agreement
Models can be tested against empirical data, and they do not perform very well; thus, determining the accuracy of future projections is difficult.
Summary sentence or paragraph that integrates the above information
The poor performance of current ESMs can result from biases in model structure, parameterization, initial values of carbon pools, and other variables. Most ESMs poorly represent permafrost dynamics and omit potential constraints on SOC storage, such as priming effects, nutrient availability, mineral surface stabilization, and aggregate formation.
Key Finding 3
Soil carbon stocks are sensitive to agricultural and forestry practices and loss of carbon-rich soils such as wetlands. Soils in North America have lost, on average, 20% to 75% of their original top soil carbon (0 to 30 cm) with historical conversion to agriculture, with a mean estimate for Canada of 24% ± 6%. Current agricultural management practices can increase soil organic matter in many systems through reduced summer fallow, cover cropping, effective fertilization to increase plant production, and reduced tillage. Forest soil carbon loss with harvest is small under standard management practices and mostly reversible at the century scale. Afforestation of land in agriculture, industry, or wild grasslands in the United States and Canadian border provinces could increase SOC by 21% ± 9% (high confidence).
Description of evidence base
Converting native forests or pastures to cropland can reduce soil carbon by 42% to 59%, respectively (Guo and Gifford 2002). A meta-analysis for Canadian soils reported that, when native soil was converted to agricultural land, there was an average 24% loss of soil carbon (VandenBygaart et al., 2003). Estimates for Mexico also suggest that loss of soil carbon due to management remains significant (Huber-Sannwald et al., 2006).
Agricultural effects on soil carbon stocks, including effects of conservation measures, are reviewed and quantified in Angers and Eriksen-Hamel (2008), Hutchinson et al. (2007), Luo et al. (2010), Palm et al. (2014), Paustian et al. (2016), Powlson et al. (2014), and many others. Specific conservation measures for improved soil carbon retention have been shown to be effective in both Canada and the United States. In Canada, conservation measures, including reduced summer fallow and reduced tillage, have been widely adopted over many regions and have resulted in soil carbon increases and reduced erosion (Soil Conservation Council of Canada 2016). Agriculture and Agri-Food Canada (2016; AAFC) has 30 years of data showing that, in the Canadian Prairies, reduced tillage combined with reduced summer fallow have led to significant SOC increases. Improved residue management, including adding forage in crop rotations or adopting agroforestry, and practices that increase plant growth such as effective fertilization are effective in increasing soil carbon (Hutchinson et al., 2007; Palm et al., 2014). A meta-analysis by Angers and Eriksen-Hamel (2008) suggested that, although significant increases in surface soil carbon with reduced tillage are commonly observed, the slight decreases in soil below the plow layer also are common, thus making overall increases in total soil carbon profiles averaged across studies small but significant. In a more recent meta-analysis by Luo et al. (2010), increased soil carbon with reduced tillage was seen only for double-cropping systems, a finding which agrees with the AAFC result that reduced summer fallow and reduced tillage together caused significant increases in soil carbon.
Palm et al. (2014) point out serious methodological flaws with many tillage comparisons that include sampling by depth not equivalent soil mass, flaws which cause significant overestimates of soil carbon in no-till soils with higher bulk densities. In their 2014 meta-analysis, about half the paired comparisons showed small increases in soil carbon from reduced till but half did not, suggesting that increased residue retention is more significant than reducing tillage. A similar meta-analysis by Kopittke et al. (2017) that also corrected for changes in bulk density found an overall small positive (+9%) effect of conversion to no-till practices from conventional till. Powlson et al. (2014) point out that the gains in surface soil carbon with adoption of no-till methods can improve crop production and reduce erosion in many cases, but the reverse can be true in cool, wet climates or the wet tropics.
Several meta-analyses of afforestation effects on former croplands have been conducted, and there is general consensus that soil carbon gains may take more than 30 years to be seen (Barcena et al., 2014; Li et al., 2012; Nave et al., 2013) and can increase carbon stocks by 19% to 53% (Guo and Gifford 2002; Nave et al., 2013).
Data on forest harvest effects are from a comprehensive meta-analysis by Nave et al. (2010), who report variable and low changes in mineral soil carbon stocks with forest harvest but significant decreases in forest floor carbon. Several chronosequences support this meta-analysis. Dean et al. (2017) argue from a modeling standpoint that there are significant long-term losses of soil carbon with forest harvest of primary forests; however, much of this argument is based on assumptions about the relationship between plant inputs and soil carbon sequestration that are not necessarily supported by empirical studies.
Wetland estimates are based on information in this report’s (SOCCR2) two wetland chapters. All chapters showed findings of strong evidence that loss of wetlands is a significant factor for total soil carbon loss, given the very high carbon density of wetland soils.
Wear and Coulston (2015), using data from the National Greenhouse Gas Inventory (NGHGI), report annual forest carbon accumulation, including both sequestration and land-use transfers in the United States as 223 teragrams of carbon (Tg C) per year, roughly 0.5% of the stored forest carbon. This likely translates into increased soil carbon storage, although this distinction was not made in the analysis. Similar estimates have not been made for Canada or Mexico.
Major uncertainties
The certainty for forest harvest effects on soil carbon appears to be very robust and based on many studies across North America, although a recent modeling study suggests that these other studies, carried out over decades, miss a multicentury-scale slow loss of soil carbon with forest harvest. However, there are no data to support that model result. Uncertainty arises because there are few empirical studies that compare soil carbon stocks in true primary forests to forests that have undergone centuries-long harvest cycles.
Uncertainties for agricultural effects have to do with site-specific variation in management implementation and lack of knowledge of deep soil carbon dynamics. However, convergence of the different meta-analyses on similar figures and research in this field is quite extensive (Li et al., 2012).
The wetland estimate also is quite robust given the high sampling density of the National Wetland Condition Assessment (NWCA) of the National Aquatic Resource Surveys. The NGHGI estimate of forest cover increase is quite robust given the quality of input data.
Assessment of confidence based on evidence and agreement, including short description of nature of evidence and level of agreement
The meta-analyses of Nave et al. (2010, 2013) suggest very good agreement over forestry effects on soil carbon, although Dean et al. (2017) suggest that, over centuries, logging has had more significant effects on soil carbon. Given that the Dean et al. (2017) study is based on modeling with assumptions that are not supported in this analysis, such as that SOC is strongly related to biomass inputs, SOCCR2 is placing greater confidence in the Nave analyses (Nave et al., 2010, 2013).
The analysis by Paustian et al. (2016) suggests that there is some disagreement over agricultural management effects on SOC and that these effects are specific to local site and climatic conditions. The Li et al. (2012) meta-analysis suggests that afforestation of former croplands globally results in net SOC increases but that local results are so variable that local projection is difficult and results depend on soil type, management, and the type of tree species.
The wetland estimate is quite robust given the high sampling density of the NWCA.
Estimated likelihood of impact or consequence, including short description of basis of estimate
Conversion to agriculture is a significant source of greenhouse gases to the atmosphere and loss of soil carbon. However, across North America, mitigation strategies such as conversion to no-till or reduced-till methods, adoption of crop rotations that provide greater carbon inputs, increased residue retention, and the use of cover crops during fallow periods are reducing the impact of agriculture (Paustian et al., 2016). Similar results are seen in Canada (Soil Conservation Council of Canada 2016). Erosion of soil carbon from agricultural lands is still a significant concern (Montgomery 2007). Afforestation has caused increases in soil carbon across CONUS.
Summary sentence or paragraph that integrates the above information
Studies have shown that conversion of native land to agriculture significantly reduced soil carbon, although improved management of agricultural land has the potential to have significant positive effects on soil carbon reserves. While modeling exercises suggest that logging and management of primary forest cause a significant SOC loss, robust meta-analyses suggest that this loss is quite minimal with effective forestry management.
Key Finding 4
Large uncertainties remain regarding soil carbon budgets, particularly the impact of lateral movement and transport of carbon (via erosion and management) across the landscape and into waterways. By 2015, cumulative regeneration of soil carbon at eroded agricultural sites and the preservation of buried, eroded soil carbon may have represented an offset of 37 ± 10% of carbon returned to the atmosphere by human-caused land-use change (medium confidence).
Description of evidence base
Best estimates of the effects of erosion are summarized in Billings et al. (2010), Van Oost et al. (2007), and Wang et al. (2017). Erosion can significantly affect productivity in agricultural regions, and some authors have argued that loss of eroded carbon represents a true loss to the atmosphere (Lal and Pimentel 2008). However, work based on multiple eroding profiles indicates that approximately 26% of eroded SOC can be replaced at the eroding site, representing a small but significant carbon sink (Van Oost et al., 2007). Harden et al. (1999) suggest that U.S. cropping patterns before 1950 likely resulted in about a 20% to 30% reduction of original SOC but that on-site recovery of soil organic matter (SOM) levels occurred after the 1950s. In Canada, VandenBygaart et al. (2012) also note a net carbon sink for eroded agricultural soils. Van Oost et al. (2007) suggest that replacement of eroded SOC, along with damped SOC mineralization upon burial, may combine to generate a small net carbon sink up to about 0.1 Pg C per year. Wang et al. (2017) calculate that cumulative, agriculturally accelerated erosion prompted SOC replacement and buried SOC preservation, representing an offset of 70 ± 16% of carbon emissions by anthropogenic land-cover change up to AD 1600; after this period, the cumulative value represented a smaller offset (37 ± 10% in 2015).
Major uncertainties
The fate of eroded agricultural soil can only be modeled, not directly measured, and the production of new soil carbon after exposure of new mineral surfaces also cannot be directly measured.
Assessment of confidence based on evidence and agreement, including short description of nature of evidence and level of agreement
Erosion of soil is known to occur, but the fate of the eroded SOC is less clear. Currently, findings conclude that the eroded SOM appears to represent a small sink of carbon but that not all material is accounted for, and the geographic extent of full carbon budget studies is quite limited. Although subsurface soil carbon appears to be relatively stable, the responses to future changes in management and climate are not well understood.
Estimated likelihood of impact or consequence, including short description of basis of estimate
In the United States, conservation measures introduced after the Dust Bowl of the 1930s suggest that the potential for massive erosional losses of soil carbon are unlikely, but similar measures are not used in Mexico. In Canada, conservation measures including zero-till have been widely adopted over many regions and have resulted in soil carbon increases and reduced erosion (Soil Conservation Council of Canada 2016). Estimates for Mexico suggest that loss of soil carbon due to management practices remains significant (Huber-Sannwald et al., 2006).
Summary sentence or paragraph that integrates the above information
Large uncertainties remain in specific key areas, including the impact of lateral movement and transport of carbon through erosion and management.
Key Finding 5
Evidence is strong for direct effects of increased temperature on loss of soil carbon, but warming and atmospheric carbon dioxide increases also may enhance plant production in many ecosystems, resulting in greater carbon inputs to soil. Globally, projected warming could cause the release of 55 ± 50 Pg C over the next 35 years from a soil pool of 1,400 ± 150 Pg C. In particular, an estimated 5% to 15% of the peatland carbon pool could become a significant carbon flux to the atmosphere under future anthropogenic disturbances (e.g., harvest, development, and peatland drainage) and change in disturbance regimes (e.g., wildfires and permafrost thaw) (medium confidence).
Description of evidence base
Although many laboratory experiments have shown that soils respond to increased temperature with increased respiration, there are many potential causes for this increase, including increased belowground inputs (Giardina et al., 2014) or increased plant production (Phillips et al., 2016). A global meta-analysis has shown that soil respiration increases with temperature (Bond-Lamberty and Thomson 2010), but how much of this is due to turnover of new, labile plant inputs is unclear (reviewed in Bradford et al., 2016). Empirical relationships developed by Crowther et al. (2016) suggest that global soil carbon stocks in the upper soil horizons will fall by 30 ± 30 Pg C under a temperature increase of 1°C, and 55 ± 50 Pg C with expected warming in the next 35 years, depending on the rate at which the effects of warming are realized.
Many studies have suggested that peatlands and boreal ecosystems are particularly vulnerable to warming (Bridgham et al., 2008; Dise 2009; Hicks Pries et al., 2015; Koven et al., 2015) because of factors such as permafrost thawing and drying effects on decomposition (Ise et al., 2008), increased fire from drying (Turetsky et al., 2014), and poleward expansion of low-carbon ecosystems (Koven 2013). Thawing of sporadic and discontinuous permafrost may release up to 24 Pg C currently stored in boreal peatlands over decades to centuries (Jones et al., 2017). Wildfire combustion of organic soils across permafrost-dominated landscapes can produce carbon losses ranging from 2.95 ± 0.12 to 6.15 ± 0.41 kilograms of carbon per m2, depending on the season (Turetsky et al. 2011).
Major uncertainties
Most laboratory experiments demonstrate that warming causes the loss of soil carbon, but how soils in natural ecosystems will respond to global warming is less predictable, given the different possible trajectories of plant production responses in different ecosystems and the possibility of increased plant production matching elevated soil respiration (Xu et al., 2016). Acclimation of soil microbes to warming could modulate the response of soils (Luo et al., 2001), although a meta-analysis (Wang et al., 2014) suggests that heterotrophic activity will not significantly acclimate to warming.
Assessment of confidence based on evidence and agreement, including short description of nature of evidence and level of agreement
At current rates of carbon dioxide and temperature increase, peatlands are highly likely to release a significant amount of stored soil carbon. Less certain is whether soils in other ecosystems, especially those subject to drought, will respond similarly to elevated temperature.
Summary sentence or paragraph that integrates the above information
The release of carbon from peatland soils could represent a major positive feedback loop to continued disturbance regimes related to climate change and human activities.
See Full Chapter & References