A Billion Tons of Unaccounted for Carbon in the Southeastern United States
Abstract
Because Earth's soil contains more carbon than the atmosphere and all terrestrial vegetation combined, forecasting and managing the global carbon cycle in the face of natural and anthropogenic change requires accurate representations of this carbon. Here from regional geomorphic and soil databases, we characterize the mass, distribution, and cycling of previously unaccounted for soil carbon across the southeastern U.S. Coastal Plain, referred to as “deep-podzolized carbon.” We show that geomorphologic-hydrologic interactions stabilize approximately 1.1 × 10−9 t of deep-podzolized carbon (equivalent to roughly 18% of the soil organic carbon stored across the entire region from 0–30 cm), and that this potentially ancient carbon is predictably distributed coincident with Pleistocene marine transgressions. We not only redefine soil carbon storage in the region but we also introduce the Earth Sciences to a massive organic carbon pool that interacts with landscape evolution and hydrology, has essentially never been studied, and is ripe for interdisciplinary research.
Plain Language Summary
Soils can accumulate and store tremendous amounts of carbon from the atmosphere. Subsequently, Earth scientists have been keenly focused on soil carbon for decades with the aim of mitigating, forecasting, and managing climate change. Here for the first time, we characterize the mass, distribution, and cycling of a billion tons of soil carbon in the southeastern U.S. Coastal Plain that until now had evaded detection. We refer to this clandestine carbon as “deep-podzolized carbon.” We show that deep-podzolized carbon is predictably distributed across the region and that it is transported and stabilized belowground by hydrologic processes. While the Coastal Plain is already recognized as a “hot spot” for soil carbon storage, our work indicates that the region stabilizes appreciably more carbon than previously thought, and suggests that future quantification of soil carbon may require observations that extraordinary deep. We expect that further characterization of deep-podzolized carbon will help identify new strategies to promote soil carbon accumulation.
1 Introduction
Earth's soil contains more carbon (C) than the atmosphere and all terrestrial vegetation combined (Batjes, 2014; Houghton, 2007). Accordingly, policy makers, resource managers, and Earth scientists from a variety of disciplines depend upon accurate representations of soil C cycling, mass, and distribution to forecast and manage the global C cycle in the face of natural and anthropogenic change. Although more than half of Earth's soil C is stabilized in the “subsoil” (below approximately 30 cm), it is well documented that deep soil C investigations are rare compared to those of surface soil due to the fact that soil C concentrations and root abundances often decrease exponentially with soil depth (Jobbágy & Jackson, 2000; Kramer et al., 2017; Mobley et al., 2015; Rumpel & Kögel-Knabner, 2010; Tumbore, 2009). Subsequently, soil C cycling, mass, and distribution across much of Earth's terrestrial surface are poorly constrained or entirely unaccounted for, and ultimately limit ecosystem, regional, and global models (Harper & Tibbett, 2013; Jackson et al., 2017).
Very thick and very deep organic C rich subsoil horizons have been observed extensively across the Coastal Plain of the southeastern United States (Figure 1 and Table S1). This subsoil C, referred to here as deep-podzolized carbon (DPC), is most often observed and reported as a “B′h” horizon during two meter deep soil survey assessments (Soil Survey Staff, 2017a, 2017b). Many of Earth's soils, particularly those classified as Spodosols and/or Podzols (FAO, 2015; Soil Survey Staff, 2015), have organic C-rich subsoil layers called “Bh” horizons. Following conventions for describing soil (FAO, 2015; Soil Science Division Staff, 2017), the Bh designation identifies an accumulation of podzolized C (organic C that is hydrologically translocated with metals from the soil surface). In the case of the B′h, the “prime” denotes podzolized C that is separate from, distinct, and deeper than an overlying Bh. Soil survey activities identify 29 different soils (specifically soil series), covering 1.9 × 10−6 ha of the Coastal Plain, that potentially contain DPC as a B′h horizon (Figures 1 and S1–S6 and Tables S1, S2, and S4). Although Coastal Plain DPC has also been reported in peer-reviewed and technical literature (Bolivar, 2000; Daniels et al., 1975; Gaston et al., 1990; Holzhey et al., 1975; Leigh, 2007; Phillips et al., 1996; Stone et al., 1993), to date not a single estimate of the C mass stabilized as DPC exists at any scale. Accordingly, Coastal Plain DPC has yet to be accounted for in ecosystem, regional, or global C budgets. Here we couple and analyze regional geomorphic and soil databases in the southeastern U.S. Coastal Plain to provide first-order estimates of DPC mass, DPC distribution, and the mechanisms underpinning DPC stabilization.
2 Data and Methods
2.1 Observations of DPC in the Soil Profile
We compiled all known observations of Coastal Plain soil profiles with reported B′h horizons (n = 80 profiles; Text S1 and Tables S1 and S3). This soil profile database contains morphologic, chemical, and physical properties for the fine Earth fraction (<2 mm) of 627 individual soil horizons including soil organic C concentration following acid dichromate digestion, 0.1 M sodium-pyrophosphate extractable C, soil pH measured 1:1 with 1 N potassium chloride, and oven-dry soil bulk density (Soil Survey Staff, 2014; Walkley & Black, 1934). Because soil pH averages 4.2 across these records, exceeding 5.0 only 13 times, we assume that reported soil organic C concentrations equal total soil C. The raw database contains more than 90 different soil horizon designations (Table S3). To facilitate interhorizon comparisons and database wide estimates we recategorized horizon designations into A, E, Bh, E′, Bw, or B′h horizons (Text S1 and Table S3), and then vertically aggregated horizons within each profile to yield a total of 472 individual horizons for analysis. These 472 horizons have 427, 378, and 231 estimates of thickness, total C concentration, and bulk density, respectively.
For A, E, Bh, E′, and Bw horizons we calculated the mean and standard error of C concentration, bulk density, and thickness each horizon (Figure S7 and Text S2) across the entire database. For B′h horizons, we similarly calculated the mean and standard error of C concentration and bulk density across the entire database. However, because B′h horizons are deep subsoil horizons, observing their actual thickness requires unusually deep soil observations, as evident by the fact that the mean sampling depth across all 80 soil profiles is 189 cm (ranging from 83 to 390 cm), yet the bottom of the B′h horizon is encountered only 12 times (15%). Subsequently, mean B′h horizon thickness derived from our database will underestimate the true mean. A significant (unadjusted p < 0.0001) and positive linear relationship between the observed B′h thickness and the total depth of sampling indicates that deeper sampling will provide more accurate thickness estimates (Figure S8). Accordingly, while underestimating B′h thickness is presently unavoidable, we minimized this bias by calculating mean and standard error of B′h thickness (Figure S7c) from only the 12 observations where the bottom of the B′h horizon was observed and the 50 observations that sampled to ≥200 cm (n = 62; Figure S8).
2.2 Analysis of Soil Surveys
Soil surveys provide unique and invaluable opportunities to evaluate edaphic properties and processes across spatial scales where direct observations are unavailable. Therefore, for a more comprehensive regional analysis of DPC extending beyond point observations we analyzed soil survey map unit delineations (MUs) from the Soil Survey Geographical (SSURGO) database in ArcMAP (Soil Survey Staff, 2017b; Table S4). We first subset the entire database to include 1,564,908 MUs within our area of interest (Figures S1–S6). We used the taxonomic order, suborder, and subgroup reported for each MU to isolate Alfisol MUs (n = 96,404), Ultisol MUs (n = 854,640), Spodosol MUs (n = 193,693), and nonlithic Psamment MUs (n = 109,016). We further separated Spodosol MUs into three groups based on the MU component name and official taxonomic series descriptions (Soil Survey Staff, 2017a; Tables S2 and S4). The first group, referred to as B′h MUs, are Spodosols with descriptions that include a B′h horizon (n = 95,621). The second group, referred to as deep Bh MUs, are Spodosols that include a Bh horizon deeper than 130 cm (n = 20,006). The third group, referred to as shallow Bh MUs, contains all other Spodosols (n = 78,066). Of the 1,564,908 MUs (~20%), 311,155 were not analyzed because they identified water features or edaphic scenarios that are outside of the scope of this work.
2.3 Relict Shoreline Features and Spatial Analyses
We compiled 1,410 previously published delineations of relict shoreline features (RSF) across the southeastern U.S. Coastal Plain (Figures 1 and S1–S6; Daniels & Kane, 2001; Lane, 1994; Rovere et al., 2015 [digital format]; Winker & Howard, 1977a, 1977b) and digitized them in ArcMAP (version 10.5.1) using the georeferencing menu and adding control points to hard copy maps and the reference layer. For each SSURGO MU and each B′h observation we calculated the distance to the nearest RSF delineation with the Near tool in ArcMAP. In R (version 3.2.3), we used the density() command to fit kernel density functions to the distributions of theses distances and then calculate the cumulative probability that a particular type of SSURGO MU or B′h observation occurs at various distances from a RSF delineation (Figure 3a).
3 Results and Discussion
3.1 DPC at the Soil Profile and Ecosystem Scale
We find that the average depth to DPC is 129 ± 4 cm (mean ± one standard error) and it is always reported below 61 cm (Figure 2a and Table S1). DPC in our profile database is reported below A, E, and Bh horizons, and it is separated from the overlying Bh by an E′ or Bw horizon. From mean C concentration (0.84 ± 0.08%), bulk density (1.61 ± 0.01 g cm−3), and thickness (61 ± 5 cm), we estimate that DPC contains on average 83 ± 10 Mg C ha−1 (Figures 2a and S5). This is the first published estimate of Coastal Plain B′h C contents, at any scale, and it demonstrates that when present in Coastal Plain soils DPC contains more C than any other mineral soil horizon including the overlying Bh (Figure 2a).
83 Mg C ha−1, although a conservative estimate, is on the same order of magnitude as average C contents in aboveground biomass of temperate forests (57 Mg C ha−1; Watson et al., 2000), the top 30 cm of wetland soil in the region (80 Mg C ha−1; Nahlik & Fennessy, 2016), and the top 30 cm of Arctic permafrost soil (91 Mg C ha−1; Hugelius et al., 2014). Our estimate is derived from soils that are sampled to an average depth of only 189 cm, and in 85% of these profiles, the depth to bottom of DPC is not known (Figure S8 and Table S1). Accordingly, any estimate of mean DPC thickness and C content from these observations will be an underestimate. A 9-m-thick B′h horizon was observed in North Carolina, and numerous B′h horizons thicker than 2 m were observed in Georgia without reaching their bottoms (Daniels et al., 1975; Leigh, 2007; Table S1). Our own fieldwork documents Florida B′h horizons thicker than 3.5 m without reaching their bottoms (Figure S9). Because thickness and C content vary directly, if average DPC thickness doubles or triples (to only 122 or 183 cm, respectively), so too would average C contents (to 166 and 249 Mg C ha−1, respectively).
Although the Coastal Plain is a depositional landscape and previous work either assumes (Bernal et al., 2016) or finds evidence (Pirkle et al., 2013) that regional C-rich subsoils result from burial, we find multiple lines of evidence indicating that DPC stabilization is unrelated to burial. First, not a single original horizon designation across our database identifies depositional discontinuities or burial in soils with DPC as professional soil scientists found no field or analytical evidence of burial (Table S3; Soil Science Division Staff, 2017). Second, burial and depositional are commonly identified in soils by abrupt changes to sand size distributions (Schaetzl & Anderson, 2005). However, fluctuating sand size distributions are rare within our compiled profiles demonstrating that DPC does not require depositional discontinuities (Figure S10). Finally, if DPC was a buried surface horizon, we would expect the sodium-pyrophosphate extractable C to total C ratio (Cpyro:Ctotal) to reflect this history. Because surface soil C is predominately decomposing plant material while podzolized C is predominately metal associated, these C pools are readily distinguished by Cpyro:Ctotal (a relatively low ratio identifies surface soil C; Banik et al., 2016). Across our compilation, however, this ratio indicates that DPC is not a buried surface as Cpyro:Ctotal of B′h horizons is nearly 3 times higher than A horizons and is indistinguishable from Bh horizons (Figure 2b). In aggregate we interpret these trends to emphasize metal-assisted hydrologic transport and stabilization (i.e., podzolization) as a central mechanism of DPC stabilization, thereby supporting Daniels et al. (1975) who conclude that regional C-rich subsoils do not necessarily result from burial.
3.2 Geomorphology Pinpoints the Distribution of DPC
Since first described in 1845 (Lyell, 1845), relict shoreline features (RSFs) from Pleistocene marine transgressions have been recognized and extensively mapped as the principal geomorphic components of southeastern U.S. Coastal Plain. We find that the regional distribution of DPC is tightly coupled to these prominent geomorphic features (Figures 1 and 3a). Observations in our profile database occur on average 8.2 km from the nearest RSF delineation (median = 2.8 km) with a 90% probability that they are within 26 km (solid orange in Figure 3a). Similarly, deep B′h MUs from SSURGO (Table S2), which cover more than 1.9 × 10−6 ha regionally, occur on average 7.8 km from a RSF (median = 3.5 km) with a 90% probability that they are within 21 km (dashed orange in Figure 3a).
Compared to most other soils across the Coastal Plain, the spatial coupling of DPC and RSFs is distinct. Regional analyses of Ultisol MUs (11.6 × 10−6 ha), Alfisol MUs (2.2 × 10−6 ha), and shallow Bh MUs (Bh <130 cm, 2.0 × 10−6 ha) reveals a 90% probability that they occur 35, 52, and 33 km from a RSF, respectively (gray in Figure 3), on average double that of B′h observations and MUs.
We attribute the distinct DPC-RSF spatial coupling to the fact that high-energy marine deposition concentrates coarse sediments (Winker & Howard, 1977a), making the historic nearshore landforms associated with RSFs extremely coarse-textured and ideal for postdepositional hydrologic C translocation deep into subsoil. Indeed, across our profile compilation total sand concentration averages 96 ± 3% and sand-sized distributions identify a dearth of fine sands (Figure S10). Moreover, as our spatial analysis indicates that DPC and fine-textured soil horizons (specifically the Bt horizon of Alfisols and Ultisols) exist on different landforms (Figure 3a), we suggest that such clay accumulations disrupt the hydrologic transport of C and prevent DPC formation.
3.3 Regional Extent and Mass of DPC
While B′h MUs cover 1.9 × 10−6 ha (Table S4), the regional extent of DPC is much larger. Deep subsoil horizons across the Coastal Plain are well known to evade detection by traditional 2-m deep soil survey assessments (Harris et al., 2005; Phillips et al., 1996). At the same time, B′h horizons are traditionally interpreted to be part of a larger, continuous subsoil stratigraphic unit that transgresses Coastal Plain landforms (Daniels, 1999; Gaston et al., 1990; Watts, 1998). Accordingly, while DPC might be detected and described as a B′h horizon during routine soil surveys, it might also be described as a deep Bh horizon, or be completely undetected by 2-m deep assessments (Figure 3b).
Considering deep Bh horizons (II in Figure 3b), deep Bh MUs cover 0.3 × 10−6 ha (Table S4). We find that these deep Bh MUs are even more tightly coupled to RSFs than B′h observations and MUs, as they occur on average 6.7 km from a RSF (median = 4.0 km) with a 90% probability that they are within 16 km (solid blue in Figure 3). Considering DPC that is not detected during 2-m-deep soil assessments (III in Figure 3b), regional delineations of Psamments without lithic modifiers cover another 2.4 × 10−6 ha (Table S4) and have potential to contain this unobserved DPC. Not only do these Psamments share DPC's distinct geomorphic setting (90% probability within 26 km of RSF, mean = 10.0 km, median = 4.7 km; dashed blue in Figure 3b), by definition they too are extremely coarse textured (Soil Survey Staff, 2015) and thus edaphically favorable to deep podzolization. Although regional MU-DPC correlations are not universal, we contend that the shared geomorphic settings of B′h horizons, deep Bh horizons, and nonlithic Psamments (Figures 1 and 3a) alongside long-standing soil-landscape interpretations (Daniels, 1999; Gaston et al., 1990; Harris et al., 2005; Watts, 1998; Figure 3b) identify more than 4.6 × 10−6 ha conducive to DPC in the Coastal Plain.
We estimate the regional mass of DPC to be 1.1 ± 0.9 × 10−9 t (or 200–2,000 × 106 Mg). This first order estimate is derived from mean C concentration and bulk density across our database (0.84% and 1.61 g cm−3, respectively; Figure S7) and by constraining mean thickness and total extent between 100 and 500 cm (Daniels et al., 1975; Leigh, 2007; Figure S9) and 1.5–3.0 × 106 ha, respectively (Figure 4a and Text S2). For comparison, 1.1 × 10−9 t C is roughly 18% of the organic C stored in surface soil (0–30 cm) across our entire 30 × 106-ha study area (Figures 4b, S11, and S12), and is comparable to soil C stored regionally from 0 to 30 (0.9 × 10−9 t) and 0–120 cm (2 × 10−9 t) in wetlands which cover a 3 to 7 times larger area than DPC (10.4 × 106 ha; Nahlik & Fennessy, 2016). Another way to conceptualize DPC mass is a typical temperate forest (57 Mg C ha−1; Watson et al., 2000) entirely covering the states of North and South Carolina (22 × 106 ha, with 1.3 × 106 Mg C).
4 Conclusions
Although we demonstrate that DPC is hydrologically transported and stabilized in subsoils by metals (podzolization), and we constrain DPC's geomorphic distribution, DPC cycling remains enigmatic, particularly the timing and direction of source-sink dynamics. As the sole radiocarbon approximation of DPC age is 14 kyr (a 160–192-cm-deep B′h in an early Pleistocene RSF; Bolivar, 2000), and DPC exists in landforms that range from roughly 1 to 2 Ma old (Figure S13), the DPC cycle appears to have a pronounced geologic component. While compiled observations (Table S1) indicate that DPC content does not vary significantly with landform age (Figure S13), we view these observations as the “tip of the iceberg,” rather than an indication of geographic homogeneity, and reiterate that DPC has yet to be completely sampled in any systematic manner (Figures 3b and S8). Over shorter time scales we propose that the DPC stabilization may be dynamic in response to altered vadose zone hydrology. Specifically, we suspect that decadal and centurial alterations to soil water tables, in response to natural and anthropogenic processes, will both enhance (during water table drawdown) and diminish (during water table rise) aerobic respiration of DPC, thereby governing the rate at which DPC is respired and lost from the soil, although such dynamics are not yet studied. While multiple theories exist to resolve the nuances of podzolization (Bourgault et al., 2015; Lundström et al., 2000), shallow-podzolized C (SPC; i.e., shallow Bh horizon) across the region is well understood to accumulate in response to the frequency and duration of near-surface saturation (Banik et al., 2016; Harris et al., 2000). DPC however is likely responding to a different component of soil hydrology as its geomorphic distribution (Figure 3a), depth in the soil (Figure 2a), and radiocarbon age (DPC is estimated to be roughly 10 kyr older than overlying SPC; Bolivar, 2000) are distinct from SPC. As our work documents a dynamic C-landscape evolution-hydrologic relationship we anticipate that forthcoming and deliberate DPC characterizations will not only require and appeal to a diverse suite of Earth surface scientists but will also provide insights to better constrain the DPC cycle and ultimately identify strategies that improve subsoil C sequestration.
On one hand, given the extensive interdisciplinary attention paid to soil C and its relationship to climate change, it is remarkable that a billion tons can be unaccounted for in this intensively studied region. Although DPC has been routinely and extensively observed (Figure 1 and Table S1), our first-order estimates redefine the region's role in the global C cycle by quantifying previously unaccounted for and potentially ancient (Bolivar, 2000) terrestrial C. On the other hand, the tendency of the Earth Sciences to superficially characterize soil C is well documented and often recited (Harper & Tibbett, 2013; Jackson et al., 2017; Mobley et al., 2015; Rumpel & Kögel-Knabner, 2010). Most who champion this notion emphasize subsoil's thickness and volume but our work demonstrates that appreciable regions of Earth's deep terrestrial surface are characterized by high C concentrations too, and that assessments usually considered “deep” (i.e., 2 m) are not always able to comprehensively capture terrestrial C cycling. Subsequently, we maintain that accurate representation, forecasting, and management of the global C cycle requires unconventionally deep subsoil characterization; otherwise, tremendous amounts of terrestrial C will remain unaccounted for.
Acknowledgments
We acknowledge the thousands of individuals within the USDA-NRCS who conduct field, laboratory, and administrative work required to create, disseminate, and manage regional soil survey efforts. We thank T. Osborne for commenting on an early version of this manuscript and A. Hartshorn for the multiple thoughtful and detailed reviews. This research was supported by the University of Florida's Institute of Food and Agricultural Sciences and the Soil and Water Sciences Department. All analyzed soil data are freely available at https://www.nrcs.usda.gov/wps/portal/nrcs/main/soils/survey/, http://soils.ifas.ufl.edu/flsoils/ and https://soilgrids.org. Digitized relict shoreline features can be made available upon request (Allan R. Bacon, University of Florida, Gainesville, FL).
