Rapid accumulation and turnover of soil carbon in a re-establishing forest

Present understanding of the global carbon cycle is limited by uncertainty over soil-carbon dynamics 1–6 . The clearing of the world’s forests, mainly for agricultural uses, releases large amounts of carbon to the atmosphere (up to 2 3 10 15 g yr 2 1 ), much of which arises from the cultivation driving an accelerated decomposition of soil organic matter 1–4 . Although the effects of cultivation on soil carbon are well studied, studies of soil-carbon recovery after cultivation are limited 4–11 . Here we present a four-decade-long ﬁeld study of carbon accumulation by pine ecosystems established on previously cultivated soils in South Carolina, USA 7 . Newly accumulated carbon is tracked by its distinctive 14 C signature, acquired around the onset of forest growth from thermonuclear bomb testing that nearly doubled atmospheric 14 CO 2 in the 1960s. Field data combined with model simulations indicate that the young aggrading forest rapidly incorporated bomb radiocarbon into the forest ﬂoor and the upper 60 cm of underlying mineral soil. By the 1990s, however, carbon accumulated only in forest biomass, forest ﬂoor, and the upper 7.5 cm of the mineral

1,300 g m -2 , or by ,40% of the carbon present at 0±30-cm depth before cultivation 7 . Here we examine recovery of soil carbon during four decades of reforestation, by using eight permanent plots that were last harvested of cotton in 1955 and were planted in 1957 with seedlings of loblolly pine (Pinus taeda L.). On these permanent plots, soil samples have been collected on seven occasions between 1962 and 1997 using the same sampling procedures. Nearly all soil samples are archived and available for analysis.
Since 1957, the aggrading forest accumulated ,3,925 g m -2 of new carbon in the soil pro®leÐthat is, the forest¯oor (O horizon) plus mineral soil: this represents carbon accumulation of nearly 100 g m -2 yr -1 for 40 years ( Table 1). Most of the new soil carbon, 3; 780 6 251 g m 2 2 (where 6251 indicates the standard error), is contained in the forest¯oor that now blankets the formerly cultivated mineral soil. Only 145 6 26 g m 2 2 of carbon accumulated in the underlying mineral soil (Table 1), about 4% of the new soil carbon. Although carbon in 0±7.5-cm mineral soil increased signi®cantly (P , 0:001) between 1962 and the 1990s, there were no signi®cant increases in soil carbon deeper than 7.5 cm (Fig. 1). In fact, carbon in the lowermost soil layer sampled, at 35±60 cm depth, signi®cantly decreased during the 40-year study, perhaps due to slow oxidation of organic-carbon input from crop roots during the period of farming.
Because the Calhoun forest was planted in 1957, new soil carbon derived from the forest has a distinctive isotopic composition because above-ground nuclear-bomb testing in the 1950s and 1960s greatly increased 14 CO 2 in the atmosphere. The fate of new forest carbon can be examined in the ecosystem because the Calhoun forest has grown entirely within the period when the concentration of 14 C has been elevated in both the global atmosphere and in inputs of carbon to the reforested soil.
Changes in soil radiocarbon ( 14 C) during forest development show that soil carbon has been more dynamic than might be suggested by the gradual reaccumulation of total carbon alone. A decomposition model 22 indicates that by 1965, D 14 C in the forest oor approached +700½ only one year after the 14 C peak in the atmosphere ( Fig. 2; D 14 C is de®ned in Methods). By the 1990s, forest-¯oor D 14 C declined to less than +300½, lagging the decrease in atmospheric 14 CO 2 owing to incorporation of 14 C into slow-todecompose, humic compounds of the acidic pine forest¯oor (Fig. 2).
By the 1990s, bomb-produced 14 C was most concentrated in the basal layers of the forest¯oor and the 0±7.5-cm mineral soil (Fig. 2   of forest¯oor) had D 14 C values of +247.3½ and +309.8½, respectively. These older forest litter materials are enriched in 14 C derived from plant biomass synthesized during the era of elevated 14 CO 2 (Fig. 2). The sur®cial Oi horizon (litterfall deposited within the past three to four years) had D 14 C of +152.2½ in 1992, closely comparable to that of atmospheric CO 2 during the late 1980s and early 1990s (Fig. 2).
Despite the relatively modest 40-year changes in mineral-soil carbon (Table 1), organic matter had incorporated 14 C throughout the entire 0±60-cm layer of mineral soil within 8 years after the atmosphere peaked in 14 CO 2 (Fig. 2). By 1968, D 14 C of mineral soil at 0±15-cm depth had increased to +200½, up from -10.4½ in 1962. By 1972, D 14 C of the entire 0±60-cm mineral soil averaged +125½, compared to less than -100½ in 1962.
Since the 1960s, however, only mineral soil at 0±7.5-cm depth has maintained its elevated 14 C (Fig. 2). This sur®cial layer of mineral soil is accumulating carbon in an incipient A horizon that is slowly reforming under the forest following long-term cultivation ( Fig. 1). But below 7.5-cm depth, decreases in 14 C generally parallel atmospheric decreases in 14 C, a pattern that indicates that forest inputs of carbon are being rapidly decomposed (Fig. 2). We suggest that the 0±7.5-cm mineral soil continues to receive 14 C from relatively recalcitrant and enriched 14 C compounds that mainly reside in the lowest layers of the forest¯oor (Fig. 2).
To understand better these dynamics, we estimated carbon inputs to soils in the 1990s from three main processes: canopy litterfall, rhizo-deposition (®ne-root sloughing and turnover), and hydrological leaching of dissolved organic carbon (DOC) from several sources.
Inputs of carbon to the forest¯oor in the mid-1990s totalled ,290 g m -2 yr -1 ( Table 1), most of which was from canopy litterfall, although smaller amounts were derived from turnover of ®ne roots 23 and the DOC in canopy throughfall. The forest's 40-year accretion of forest-¯oor carbon, 3,780 g m -2 , is about 13 times greater than current annual input ( Table 1). Accumulation of forest-¯oor carbon is attributed mainly to complex, acidic, and recalcitrant compounds derived from the coniferous leaf litter. The forest¯oor is classi®ed as an acidic mor, perched atop the mineral soil, with relatively minimal mixing by soil animals of solid organic material with mineral soil below 7 .
Inputs of carbon to the 0±15-cm layer of mineral soil totalled ,95 g m -2 yr -1 , much of which was from rhizo-deposition, although a substantial fraction was derived from DOC in leachates from the forest¯oor ( Table 1). Even with carbon input estimated conservatively (Table 1), long-term soil carbon accumulation (145 g m -2 ) is only 1.5 times that of estimated annual input in the 1990s. The short residence time for carbon input to mineral soils is notable 24 . Carbon inputs to sur®cial mineral soil in ®ne-root biomass and DOC (Table 1) are readily decomposed, as they have little protection from adsorption to organophilic clays in these surface soils with sandy loam textures 25,26 .
From an ecosystem perspective, this aggrading forest is a strong carbon sink. Accumulation of carbon is especially pronounced in tree biomass and the forest¯oor relative to that in mineral soil. During the growth of this forest, trees accumulated 14,060 g m -2 of carbon (between 1957 and 1990), compared with 3,780 g m -2 in the forest¯oor (1957±97) and 145 g m -2 in sur®cial mineral soil (1962± 97). Annual carbon accretions thus averaged about 426, 94 and 4 g m -2 in the three ecosystem components, respectively.
The overall pattern of carbon sequestration suggests a low carbon-storage potential for mineral soils compared to that in biomass and forest¯oor; a similar conclusion has been reached from a review of soil-carbon gains under primary vegetative successions 6 . Yet in comparison to other forested soils recovering from previous cultivation, the mineral soil at the Calhoun forest seems to be relatively slow to recover organic carbon. Previous estimates of mineral-soil carbon gains under forests 4,11 range from 21 to 55 g m -2 yr -1 , while soils at the Calhoun forest gained carbon at 4.1 g m -2 yr -1 (Table 1).
Although changes in soil carbon are not easy to estimate, the Calhoun experiment provides an approach for making these estimates: especially useful are its well replicated permanent plots, extensive within-plot composite sampling, and soil archive. The main factors driving relatively low carbon accumulations in Calhoun mineral soils are coarse soil texture, low soil surface area, lowactivity clay mineralogy 26 , and warm and humid climate. A network of long-term soil-ecosystem experiments similar to that at the Calhoun, at sites that encompass a range of controlling variables of soil and ecosystem carbon, would greatly facilitate future modelling and management of the carbon cycle 11,16 .  (in 1962,1968,1972,1977,1982 and 1990).   ³ Inputs of canopy litterfall to forest¯oor were estimated in 1991±92. Inputs of dissolved organic carbon (DOC) were estimated in bi-or tri-weekly collections over two years (1992± 94): into forest¯oor from DOC in canopy throughfall; into 0±15-cm mineral soil from DOC in water in®ltrating from the forest¯oor; and into 15±60-cm soil from DOC in water draining into soil at 15-cm depth. Inputs of carbon from rhizo-deposition were estimated in 1994±95 from 50% of the live ®ne-root (,2-mm) biomass in forest¯oor, or mineral soil at 0±15 and 15±30 cm depth. Fine roots were sampled in 1994±95, every three weeks over 18 months. The 50% factor is used as a conservative estimate of carbon inputs from rhizo-deposition.

Total soil carbon accretion*² (g m
The research area is located at 34.58 N, 828 W. Annual precipitation averages 1,170 mm (1950±87) and temperature 16 8C. In the early 1800s, primary deciduous forests at the site were cleared, mainly to grow cotton, and the site was managed for row crops, hay and pasture until the mid-twentieth century 7,16 . Soils are acidic Ultisols, classi®ed as the Appling series (®ne, kaolinitic, thermic Typic Kanhapludults). The Appling soil is a common soil of southeastern North America, and is formed from granitic gneiss, the bedrock from which about half the soils in the southern Piedmont region are derived.
In 1957, 16 permanent plots were installed on two cotton ®elds at the Calhoun Experimental Forest, eight of which are used in these carbon analyses. The mineral soils of these eight plots were sampled in 1962, 1968, 1972, 1978, 1982, 1990 and 1997. At each collection, each plot was sampled at least 20 times with a 2-cm-diameter punch tube in a systematic random fashion at four soil depths: 0±7.5, 7.5±15, 15±35 and 35±60 cm. Within each plot, the >20 samples per soil depth were composited, that is, in one sample per depth. Soil archive, radiation and total carbon. The Duke Soil Archive stores airdried soil samples at room temperature in sealed glass containers. Total soil carbon was analysed in powdered samples with a Perkin-Elmer CHN combustion instrument. Radiocarbon was measured by accelerator mass spectrometry (AMS) on graphite targets prepared from soil organic matter and is reported as D 14 C, the per mil deviation of 14 C/ 12 C compared with a decaycorrected oxalic acid standard 27 . Positive values of D 14 C indicate presence of bomb-produced 14 C, and negative values indicate predominance of old soil organic matter with 14 C that has experienced signi®cant radioactive decay (half-life is 5,730 yr). Radiocarbon was estimated using AMS of composite samples made from soil from the eight plots at each depth. Analysis of variance and means separation tests were used to test effects of time on soil carbon accumulation. Soil carbon inputs. To estimate carbon inputs to soils, carbon in litterfall, ®ne roots, and soil water were estimated. Litterfall was sampled in each of the eight permanent plots with ®ve collectors (each 0.72 m 2 in area) per plot. Canopy litterfall was collected monthly during 1991±92 28 .
Fine roots were sampled volumetrically using a 6-cm-diameter corer that collected undisturbed cores of O horizon and mineral soil from 0±15 and 15± 30 cm depths. Soil cores were taken every three weeks for 18 months in 1994± 95. Samples were returned to the laboratory where ®ne roots (,2-mm) were separated from soil by wet-sieving and hand picking. Live roots were separated from dead, and the former were ashed to estimate carbon contents (taken as half the loss on ignition). Only the live fraction of the ®ne roots is reported here, and carbon inputs from ®ne-root turnover were simply estimated as 50% of live ®ne-root carbon in forest¯oor, or mineral soil at 0±15 and 15±30 cm depth. This factor (50%) is taken to be a conservative estimate of carbon inputs from rhizo-deposition.
DOC was determined in atmospheric precipitation (wet deposition), canopy throughfall, and solutions draining forest¯oor and several depths of mineral soils to 60 cm (ref. 19). Wet-only precipitation was collected by an Aerochem Metric sampler. Throughfall was collected in three bulk precipitation collectors in each of 8 plots using 16-cm-diameter funnels that were continuously open. Gravitational lysimeters were used to collect water from below O horizons and at 15-cm depth. Tension lysimeters of porous Te¯onquartz design collected solutions at 60-cm depths. Solutions were collected every two or three weeks over two years, 1992±94, and solutions were estimated for DOC concentration by combustion and infrared analysis, after purging solutions of CO 2 and H 2 CO 3 by acidi®cation and sparging with N 2 gas.