Plant and Soil

, Volume 368, Issue 1, pp 641–648

The hidden organic carbon in deep mineral soils


    • School of Environmental ScienceMurdoch University
  • M. Tibbett
    • National Soil Resources Institute, Department of Environmental Science and TechnologyCranfield University
    • School of Earth and EnvironmentUniversity of Western Australia
Regular Article

DOI: 10.1007/s11104-013-1600-9

Cite this article as:
Harper, R.J. & Tibbett, M. Plant Soil (2013) 368: 641. doi:10.1007/s11104-013-1600-9



Current estimates of soil organic carbon (SOC) are based largely on surficial measurements to depths of 0.3 to 1 m. Many of the world’s soils greatly exceed 1 m depth and there are numerous reports of biological activity to depths of many metres. Although SOC storage to depths of up to 8 m has been previously reported, the extent to which SOC is stored at deeper depths in soil profiles is currently unknown. This paper aims to provide the first detailed analysis of these previously unreported stores of SOC.


Soils from five sites in the deeply weathered regolith in the Yilgarn Craton of south-western Australia were sampled and analysed for total organic carbon by combustion chromatography. These soils ranged between 5 and 38 m (mean 21 m) depth to bedrock and had been either recently reforested with Pinus pinaster or were under agriculture. Sites had a mean annual rainfall of between 399 and 583 mm yr−1.


The mean SOC concentration across all sites was 2.30 ± 0.26 % (s.e.), 0.41 ± 0.05 % and 0.23 ± 0.04 % in the surface 0.1, 0.1–0.5 and 0.5 to 1.0 m increments, respectively. The mean value between 1 and 5 m was 0.12 ± 0.01 %, whereas between 5 and 35 m the values decreased from 0.04 ± 0.002 % to 0.03 ± 0.003 %. Mean SOC mass densities for each of the five locations varied from 21.8–37.5 kg C m−2, and were in toto two to five times greater than would be reported with sampling to a depth of 0.5 m.


This finding may have major implications for estimates of global carbon storage and modelling of the potential global impacts of climate change and land-use change on carbon cycles. The paper demonstrates the need for a reassessment of the current arbitrary shallow soil sampling depths for assessing carbon stocks, a revision of global SOC estimates and elucidation of the composition and fate of deep carbon in response to land use and climate change.


Soil carbonRegolithGlobal change


Estimates of global soil organic carbon (SOC) stocks vary from 684 to 724 Pg in the surface 0.3 m (Batjes 1996), the Intergovernmental Panel of Climate Change (IPCC) standard sampling depth (Aalde et al. 2006), and 1462 to 1548 Pg in the top metre (Batjes 1996). Numerous studies report “deep” SOC in soils, but typically this is only considered to depths of 1–2 m at most (Batjes and Sombroek 1997; Harrison et al. 2011; Rumpel and Kögel-Knabner 2011; Sombroek et al. 1993). Fewer studies report SOC to 3 m (Davidson and Janssens 2006; Jobbágy and Jackson 2000) and exceptionally to 8 to 11 m (Jackson et al. 2002; Nepstad et al. 1994; Richter and Markewitz 1995). However, soils are often much deeper than any of these measurements, with many landscapes having soils that extend to depths of many metres (Anand and Paine 2002; ISSS Working Group RB 1998; Richter and Markewitz 1995). Here we define deep soils as those with profiles greater than 5 m deep.

Plant roots penetrate soils to considerable depths in a range of landscapes (Canadell et al. 1996; Jackson et al. 2000; Nepstad et al. 1994; Schenk and Jackson 2005). For example, Canadell et al. (1996) summarized reports of roots under different vegetation systems extending to 9.5 ± 2.4 m in deserts, 7.3 ± 2.8 m in tropical evergreen forest and 15.0 ± 5.4 m in tropical grassland and savannah, with maximum root depths of 68 m. Where deep soils coincide with deep rooting the biological deposition of carbon from roots (and their associated biota) is inevitable at depths at which SOC has rarely been measured.

Given the paucity of data on deep SOC contents we investigated the extent of SOC storage at depth, hypothesizing that SOC retained in soils below the top half metre (in highly weathered deep profiles) would account for the major proportion of SOC in the landscape. In order to test this we analysed deep soil samples taken from multiple bores in south-western Australia; a region that is characterized by deep weathering profiles formed on the Archaean granites and gneisses of the Yilgarn Craton (Gilkes et al. 1973) possibly under a relict tropical environment (Prescott and Pendleton 1952), with resultant weathering profiles up to 150 m thick (Anand and Paine 2002) but generally less than this (McArthur 1991). These profiles have subsequently been modified by localized erosion and deposition (Churchward and Gunn 1983). Prior to extensive contemporary land development for agriculture within the last 60 years (Australian Greenhouse Office 2001) this region was covered in a range of xerophytic plants, with root systems that extended to depths of 40 m, such as reported for a Eucalyptus marginata forest (Dell et al. 1983). In this study, we measured regolith depth and SOC concentrations and subsequently estimated total SOC stores and compared these to more commonly reported surficial values.


Five locations (Fig. 1) were selected as part of a program to monitor the effects of reforestation on farmland hydrology. Locations were named after the landholders Caldwell (CD), Curo (CU), Gillam (GL), Pratt (PT) and Scott (ST). This region has a Mediterranean climate with cool, wet winters and hot, dry summers (Table 1). Rainfall and potential evaporation were estimated using the SILO climate data base (Jeffrey et al. 2001). Land had been developed for farming in the last 50–80 years with this involving complete removal of former deep-rooted vegetation and replacement with cereal crops or annual pastures. Pinus pinaster (Maritime pine) was established on the cleared farmland in 2006 (Harper et al. 2009).
Fig. 1

Location of the five sampling sites within the south-western region of Western Australia

Table 1

Site details and carbon storage in deep profiles


No. of bores

Mean depth (m)

Rainfall (mm yr−1)

Potential evaporation (mm yr−1)

Depth interval

0–0.5 m

0–5 m


kg C m−2


kg C m−2


kg C m−2



22.8 ± 2.9



3.6 ± 0.6


13.8 ± 2.3


25.4 ± 2.7



12.9 ± 1.7



8.0 ± 1.2


16.3 ± 1.6


21.8 ± 2.7



27.8 ± 2.0



4.2 ± 1.0


13.8 ± 3.7


26.8 ± 4.0



20.6 ± 3.1



7.5 ± 1.4


22.8 ± 4.4


37.5 ± 4.7



27.5 ± 3.6



7.6 ± 1.9


20.0 ± 5.4


35.8 ± 6.7

All sites


21.4 ± 1.4



5.8 ± 0.6


16.3 ± 1.4


27.4 ± 1.8

Mean carbon mass density (kg C m−2) (±s.e.) for the surface 0.5 and 5.0 m and compared as a proportion (%) of the total carbon store to bedrock

Several bores were installed at each location in 2008 (38 in total) in a range of landscape positions. Soils were sampled from 0–0.1, 0.1–0.5 and 0.5–1.0 m using an auger and shovel, and below this by 10 cm diameter reverse circulation (RC) drilling extending to bedrock (at 5 to 38 m depth).

RC drilling is a procedure widely used for exploration and resource definition in the mineral and water resources industries, where uncontaminated samples are required for geochemical analysis (Marjoribanks 2010) and it is necessary to define both the depth distribution and concentration of the elements of interest. The technique is commonly used in Western Australia where much mining occurs in deep weathering profiles (Davis 1992) and thus in materials broadly similar to what were encountered in this study. Gold mining, for example, occurs across various geologies and estimates of mineralization are sought in amounts commonly between 0.5 and 8 μg g−1 (Davis 1992). This is an order of magnitude lower than the values of >0.01 % or 100 μg C g−1 reported in this study.

RC drilling proceeded in 1 m increments with compressed air used to carry the spoil directly from the cutting bit to the sample bin (cyclone) via the inside of the drill tube, which was thus separated from the walls of the hole (Marjoribanks 2010) removing contact with the surface horizons. The compressed air travelled down the centre of the drill tube and exited at the drill bit. It was thus also isolated from the walls of the hole. The drill was not removed from the hole and additional drill lengths were added from above.

Samples were taken from 1 m intervals for every 3 m of drill depth commencing from 1 m depth. Samples were taken as drilling proceeded, with the spoil being wasted until the required depth increment was reached, and a 1–2 kg field sample collected. With a diameter of 0.1 m, and length of 2 m, approx 25 kg of soil would travel through the sampling system between samples (under some force), and this would remove any residue from the previously sampled material. In this way horizontal (down-tube) and vertical contamination was prevented.

Samples were air dried and then sub-sampled and a 200 mg sample taken for carbon analysis. This sample was ground to 500 μm, oven dried over night (105 °C) and cooled in a dessicator. Samples returned from the drilling were quite disrupted and fine roots were not apparent.

The 409 samples were analysed by dry combustion chromatography with an Elementar Vario Macro elemental analyzer (Hanau, Germany). Samples were combusted into a steam of He and CO2 and trapped using a temperature reversible trap with moisture removed using phosphorus pentoxide. The CO2 trap was heated driving off the trapped gas that is measured via thermal conductivity. In this way all the combustion products of the entire sample were measured, allowing accurate low SOC content determination, compared to other propriety instruments that subsample gases for analysis. Soil reference material (mean = 0.57 % C, s.d = 0.0192, n = 7) was diluted tenfold with acid washed sand and both were measured for carbon (net mean = 0.053 % C, s.d. = 0.00524, n = 12) and glutamic acid was employed as a reslope standard. In this way we ensured small SOC values were accurately reported. One outlier was removed. A subset of 10 samples across the dynamic range was analysed using the Walkley-Black (1934) procedure to ensure compliance with this alternative method of SOC determination.

The carbon mass density of each profile was estimated from the SOC content (%), the gravel (>2 mm) content, the depth interval of the sample (m), with assumptions of bulk density taken from an extensive coring study in this region (Robinson et al. 2006). For samples from the surface 1 m a bulk density of 1.5 g cm−3 was employed, for samples from 1 m to bedrock a value of 1.8 g cm−3 was used. Carbon mass density (CMD, kg C m−2) was calculated by meaning values for the depth intervals 0–0.1, 0.1–0.5, 0.5–1 and for each of the 5 m intervals to 35 m.


In this study across 38 deep bores at five locations, the soil-regolith profile varied between 5 to 38 m in depth with a mean of 21.4 ± 1.4 m (s.e., Table 1). There was a pronounced trend of decreasing mean SOC content with depth at each site, with no difference in the general trend between sites (Fig. 2). The mean SOC concentration across all sites was 2.30 ± 0.26 %, 0.41 ± 0.05 % and 0.23 ± 0.04 % in the surface 0.1, 0.1–0.5 and 0.5 to 1.0 m increments, respectively. The mean value between 1 and 5 m was 0.12 ± 0.01 %, whereas between 5 and 35 m the values decreased from 0.04 ± 0.002 % to 0.03 ± 0.003 %. A single sample at 37 m had a value of 0.02 %. All 408 samples contained some SOC, with this ranging between 0.01 and 6.32 %. The carbon is considered unlikely to be derived from carbonates, although these can occur in previously deeply weathered profiles (Milnes and Hutton 1983). All samples were analysed for pH with a mean value of 5.82 ± 0.04; similarly a comparison of the combustion-derived analyses with a Walkley-Black wet oxidation (Rayment and Higginson 1992) was in accord with an absence of carbonate.
Fig. 2

Mean soil organic carbon (SOC) concentrations (%) with depth for the five sampling sites. Error bars are s.e.

The total carbon mass density (CMD) of the individual profiles varied from 10.5 to 55.1 kg C m−2, with a mean value of 27.4 kg C m−2. The mean values for each of the five locations varied from 21.8–37.5 kg C m−2, and these values were generally associated with both soil depth and rainfall (Table 1). The site with the shallowest soils (CD) also had the smallest CMD of 21.8 kg C m−2. There were broad differences in CMD with differences in rainfall, such that the three sites with around 400 mm mean annual rainfall (CD, CU, PT) had a mean CMD of 24.6 kg C m−2, whereas those with a mean annual rainfall of around 570 mm (GL, ST) had a mean CMD of 36.6 kg C m−2.


When the SOC storage within the deep profiles was compared with what would have been reported from conventional sampling depths (Table 1), it is clear that considerably more SOC was stored in the soils than is normally reported. Across all samples, the surface 0.5 m, which is deeper than the standard IPCC sampling depth of 0.3 m (Aalde et al. 2006), contained 5.8 ± 0.57 kg C m−2 or 21 % of the total store to bedrock. If this is adjusted to 0.3 m depth, using an exponential function based on the samples in the surface metre, the value decreases to 5.6 kg C m−2. For the individual sites this ranged from 3.6 to 8.0 kg C m−2, or 14–37 % of the total store. Although several studies have reported deeper stores of SOC to depths of 5 to 8 m (Canadell et al. 1996; Jackson et al. 2000; Nepstad et al. 1994; Schenk and Jackson 2005), it is evident that more SOC is stored at even greater depths in the soil profiles. Indeed, in this study the surface 5 m contained 16.3 ± 1.38 kg C m−2 or 59 % of the total store to bedrock, with this proportion varying from 47 to 77 % across the five sampling locations. The amount of carbon stored in the soils can also be contrasted with the biomass carbon storage of 11.0–16.0 kg C m−2 expected at equilibrium following reforestation for these sites (Harper et al. 2007) and likely previously removed from the sites by deforestation in advance of agriculture. The increase in SOC storage with rainfall (Table 1) is consistent with studies reporting carbon accumulation in relation to water balance both within this region (Harper et al. 2007) and elsewhere (Jackson et al. 2002; Schenk and Jackson 2002, 2005).

The soils examined in this study had been cleared of natural vegetation for several decades, and the change in SOC content in the period subsequent to clearing is unclear. Camargo et al. (1999) in the eastern Amazon found that fine roots decomposed to depths of up to 5 m following conversion of forest to pasture and it is likely that similar changes have occurred here. There are two possible sources for the deep carbon; that produced in situ by roots or dissolved carbon that has moved downward from nearer the surface. Fine roots were not readily apparent in the sampled material possibly due to the sampling technique. Although the particle size distribution of the deeper samples was not estimated, this region has a regional surface of deeply weathered material with granite as the most likely parent material (Anand and Paine 2002; Gilkes et al. 1973; McArthur 1991). Clay content will likely vary between 30–60 %. In this environment, however, water movement to depth is more controlled by the occurrence of macropores rather than properties such as particle size of the soil/regolith matrix (Johnston 1987).

What are the implications of these findings? It is clear that the somewhat arbitrary and surficial nature of soil sampling depths (Richter and Markewitz 1995) for SOC studies needs to be re-evaluated when terrestrial carbon stocks are being estimated. For our deep soil profiles, half to three quarters of the SOC is estimated to occur in the top 5 m (Table 1), below which there is little change in concentration to bedrock (Fig. 2). Combining sampling from this depth, an estimate of the depth to bedrock and an estimate of a particular site’s rainfall may provide a reasonable estimate of SOC stores in deep profiles. Further studies are now needed in other regions to substantiate these findings.

Questions also arise about the dynamics of the deep SOC with deforestation, reforestation and climate change. Carbon emissions from land-use change account for approximately 13 % of the global total (Nabuurs et al. 2007; Pan et al. 2011), with this based on estimates of losses from both biomass and soils (Camargo et al. 1999; Nepstad et al. 1994). For example, in Australia’s national carbon account deforestation losses comprise estimates of biomass removals with soil carbon losses estimated using the Roth-C model based on the surface 0.3 m (Skjemstad and Spouncer 2003). Many areas that are currently being deforested, such as the Amazon, central Africa, tropical Asia and Australia, have deep soil profiles (Schenk and Jackson 2005) and global deforestation emissions may thus be considerably underestimated depending on the dynamics of deep carbon decomposition. As far as we can ascertain, there are no estimates of carbon changes in deep profiles in response to reforestation, although estimates have been made for the near surface horizons (Guo and Gifford 2002; Harper et al. 2012; Paul et al. 2002).

There are conflicting conclusions as to how near-surface SOC will respond to increases in temperature (Davidson and Janssens 2006), and the same considerations will apply to deeper stores, with the added complication that profile hydrology may also be dramatically affected by both land-use (Peck and Hatton 2003) and climate change induced changes in the water balance and this in turn will affect carbon cycling. Koarashi et al. (2012) show that a large proportion of SOC turns over on the timescales of decades in forest and grassland subsurface (0.4–0.6 m) soils. They also suggest that the dynamics of subsurface SOC is largely controlled by interactions with soil minerals, which contrasts with stronger climate control in surface soil layers. Whereas some authors consider that even SOC at relatively shallow depths of 0.6 to 0.8 m is effectively inert (Batjes 1996; Fontaine et al. 2007), others (Fearnside and Barbosa 1998; Nepstad et al. 1994) have suggested that SOC to 8 m depth is relatively labile and up to 15 % of the total pool cycles at yearly to decadal timescales (Nepstad et al. 1994).

This work leads to numerous future avenues of enquiry, some of which have been posed for SOC horizons closer to the surface (Rumpel and Kögel-Knabner 2011). For example, what is the composition and what are the dynamics of deep SOC under different ecotypes and land-use systems? When is deep SOC being accumulated or mineralized in response to land use and environmental change? Does the SOC reported here participate in the global carbon cycle, or is it in effect outside of that cycle? How much SOC exists in other landscapes with soils that are many metres deep? Should this deep carbon be considered in carbon accounting at project and national scales? Perhaps most immediately we must consider how this deep carbon will affect global estimates of SOC storage and the models that estimate global carbon cycles.


We thank David Chittleborough, Jock Churchman, Guy Kirk, Stan Sochacki, Bernard Dell, Chris Mitchell, Rod Keenan, Keith Smettem and two anonymous reviewers for useful comments, Ian Truckell (Cranfield University) for preparing the Figures, Alex Winter (Forest Products Commission, Western Australia), Marianne Harkins (CSBP), Michael Smirk and Evonne Walker (UWA) for assistance in the field and laboratory and all landholders for access to sites. Drilling was undertaken by the Forest Products Commission (Western Australia) as part of the Australian National Action Plan for Salinity and Water Quality project “Strategic Tree Farming”.

Copyright information

© Springer Science+Business Media Dordrecht 2013