Soil Chemistry and Clay Mineralogy of an Alluvial Chronosequence from the North Carolina Sandhills of the Upper Coastal Plain, USA

: Temporal changes in soil development were assessed on ﬂuvial terraces of the Little River in the upper Coastal Plain of North Carolina. We examined ﬁve proﬁles from each of six surfaces spanning about 100,000 years. Soil-age relationships were evaluated with inter-surface clay mineral comparisons and regression of chemical properties versus previously reported optically-stimulated luminescence ages using the most developed subsoil horizon per proﬁle. Bases to alumina (Bases/Al 2 O 3 ) ratios have negative correlations with age, whereas dithionite-Fe (Fe D ) concentrations are positively correlated with time and differentiate ﬂoodplain (<200 yr BP) from terrace ( ≥ 10 ± 2 ka) soils and T4 pedons (75 ± 10 ka) from younger (T1-T3b, 10 ± 2–55 ± 15 ka) and older (T5b, 94 ± 16 ka) proﬁles. Entisols develop into Ultisols with exponentially decreasing Bases/Al 2 O 3 ratios, reﬂecting rapid weatherable mineral depletion and alumina enrichment during argillic horizon development in the ﬁrst 13–21 kyr of pedogenesis. Increasing Fe D represents transformation and illuviation of free Fe inherited from parent sediments. Within ~80–110 kyr, a mixed clay mineral assemblage becomes dominated by kaolinite and gibbsite. Argillic horizons form by illuviation, secondary mineral transformations, and potentially, a bioturbation-translocation mechanism, in which clays distributed within generally sandy deposits are transported to surface horizons by ants and termites and later illuviated to subsoils. T5b proﬁles have Fe D concentrations similar to, and gibbsite abundances greater than, those of pedons on 0.6–1.6 Ma terraces along Coastal Plain rivers that also drain the Appalachian Piedmont. This is likely because the greater permeability and lower weatherable mineral contents of sandy, Coastal Plain-sourced Little River alluvium favor more rapid weathering, gibbsite formation, and Fe translocation than the ﬁner-grained, mineralogically mixed sediments of Piedmont-draining rivers. Therefore, recognizing provenance-related textural and mineralogical distinctions is crucial for evaluating regional chronosequences.


Introduction
Soil chronosequences are assemblages of soils of differing ages that have formed on similar parent materials and experienced comparable topographic, vegetative, and climatic conditions through time [1,2].While often the latter three factors are not entirely static over the duration of interest, in well-expressed chronosequences, time so heavily affects pedons that its influence on soil evolution may be assessed.A general working hypothesis for soil chronosequences is that certain mineral, chemical, and morphological components are non-randomly enhanced or depleted as a function of time.
Within soil science, chronosequences are the chief source of data used to evaluate rival theories of pedogenesis and ascertain rates of soil development [3].Chronosequence frameworks have improved understanding of the rates and pathways of primary mineral weathering [4,5], clay mineral transformations [6,7], and changes to bulk chemistry [8,9] in soils from a range of global settings.Furthermore, when the time required to develop certain pedogenic features is known, soil chemical and mineralogical properties are useful for establishing relative-age relationships among landforms [5,10], mapping and correlating Quaternary deposits [11], and aiding chronological interpretations in archaeological investigations [12,13].Given these attributes, soil chronosequences serve important functions in pedology, geomorphology, archaeology, and Quaternary geology [3].Globally, fertile soils are declining in areal extent due to residential, industrial, and transportation development, as well as improper land use practices [14,15].Because soils are a finite resource of fundamental importance to terrestrial ecosystems and human nutrition, understanding soil formation processes and the rates at which they occur is crucial [16,17].This is particularly true in the Coastal Plain of the southeastern United States, where improved knowledge of soil and landscape evolution is needed to better inform long-term, sustainable land management planning in the face of competing agricultural, suburban, and urban land uses and a changing climate [18].In this regional context, chronosequence studies potentially provide valuable insights into the time scales of pedogenesis [19], the time required for recovery of degraded soils [20], and soil response to climatic change [17].Despite these benefits, relative to other parts of North America, comparatively few chronosequences have characterized age-related changes in the chemistry and mineralogy of alluvial soils from the Coastal Plain of the southeastern United States [19,[21][22][23][24][25], and even fewer studies describe such changes for soils on fluvial terraces in the southeastern Appalachian Piedmont [26], Blue Ridge [12], and Valley and Ridge [27].Thus, there is a clear need to increase knowledge of temporal trends in soil chemical and mineralogical development in fluvial landscapes of the southeastern US.
The present study is unique relative to prior research because the source area for terrace alluvium is completely contained in the upper Coastal Plain, within the catchment of the Little River near Fayetteville, North Carolina.This setting produces soils that are almost exclusively derived from the deeply leached, siliceous, sandy surficial deposits of the Carolina Sandhills (Figure 1).In contrast, past regional fluvial chronosequences characterized pedons on terraces of Coastal Plain rivers whose basins extend headward into the neighboring Appalachian provinces [19,[21][22][23][24][25].The parent sediments of these chronosequences possess finer textures and greater concentrations of unstable minerals than those of deposits sourced chiefly from the Coastal Plain [28], and this contrast in source material likely affects rates of soil development.Also, relative to prior research, this study examines a shorter soil-forming interval (100 kyr).Fluvial chronosequences whose time frames collectively span from the Holocene to 13 Ma [19,[21][22][23][24][25] exist for the Coastal Plain yet offer only meager data for differentiating among pedons formed during the late Quaternary.
Additionally, previous studies relied on correlations between river terraces and dated marine deposits to estimate soil age, not numerically dated terrace alluvium.The present investigation is instead based on a previously reported optically-stimulated luminescence (OSL) terrace chronology [29].OSL is well-suited to dating fluvial deposits within the age range of several hundred to approximately 100,000 yr BP and provides an estimate of the time elapsed since sediment grains were last exposed to sunlight before being buried [30].The technique generates an age by using light to artificially stimulate the release of stored electrons, which accumulate in defects in the crystal lattice of quartz and feldspar grains when the electrons are evicted from their ground state by ionizing radiation from cosmic rays and U, Th, and 40 K decay in surrounding deposits [31].When trapped electrons are released and return to ground state, they emit a photon (luminescence) that is proportional to the total amount of ionizing radiation to which the sediment was exposed during burial [30].This radiation can be reconstructed in the laboratory and related to the amount of radiation annually received by the sample to estimate the time elapsed since its burial [31].OSL dating is particularly advantageous in the Little River valley, where older terrace sediments are beyond the range of radiocarbon dating (>45 ka), younger Soil Syst.2022, 6, 1 3 of 29 sediments lack radiocarbon-datable materials, and fluvial quartz sand is abundant.Thus, whereas prior Coastal Plain chronosequences used simple terrace correlation techniques to approximate the age of alluvial soils, ages in this investigation are based on direct dating of fluvial sediments obtained at or very near the studied pedons [29], providing a more reliable geochronologic framework.
The present investigation aims to improve understanding of pedogenesis with time in river valleys of the southeastern US Coastal Plain and identify soil properties that differentiate fluvial deposits of differing late Quaternary ages in the region.In support of these goals, the specific objectives of the study are to:

•
Assess the utility of soil chemical and mineralogical properties as indicators of terrace age; • Characterize temporal trends in the total chemistry and mineralogy of subsoils, including clay mineral transformations; identify pedogenic processes responsible for such changes; and relate these properties and processes to argillic horizon formation; • Compare soil developmental trends along the Little River with those from elsewhere in the Coastal Plain.
This research examines five representative pedons from each of six fluvial surfaces in the Little River valley and focuses on the clay mineralogy and total chemistry of the whole soil (<2 mm) and fine sand (0.125-0.250 mm) fractions of the most well-developed subsoil horizon per profile.Within this design, whole soil properties of interest consist of variables known to be effective chronosequence parameters in the region [12,19,21,23,24,26,27] and include the molar ratio of bases to alumina (Bases/Al 2 O 3 ), dithionite-extractable Fe (Fe D ), and the ratio of dithionite-extractable to total Fe (Fe D /Fe 2 O 3 ), among other properties.The elemental oxide concentrations of K 2 O, Na 2 O, and MgO comprise the variables measured for the fine sand fraction.
Among evaluated soil properties, we hypothesize that subsoil Fe D concentrations will be the most effective at discriminating between profiles of differing ages, as prior research indicates that iron chemistry is strongly correlated with time in Coastal Plain soils [23].With age, we anticipate that pedons will exhibit decreasing Bases/Al 2 O 3 ratios and increasing Fe D content in response to weatherable mineral depletion and the subsoil accumulation of clays and oxides from eluviation-illuviation and argillic horizon development.Over time, subsoils are expected to develop an increasingly stable clay mineral suite, consisting of kaolinite, hydroxy-interlayered vermiculite (HIV), and gibbsite on older, higher terraces.Finally, we hypothesize that Little River pedons will display a more advanced state of weathering than profiles of comparable age along Piedmont-draining Coastal Plain streams, owing to the more weathered, more permeable Coastal Plain-derived alluvium from which the sampled soils formed.

Physiography and Terrace Ages
This investigation was conducted at Fort Bragg Military Reservation along the Little River, which is part of the Cape Fear River drainage in North Carolina and has an 880 km 2 catchment that is completely contained within the Sandhills province of the upper Coastal Plain (Figure 1).The Sandhills bound the southern border of the Appalachian Piedmont and comprise a deeply dissected region with highly weathered, sandy soils that formed from Cretaceous to Tertiary fluvial and marine sediments and from eolian sands that cap the aforementioned deposits in some upland areas [28,32].Within the Little River basin, surficial upland sediments include nonmarine kaolinitic clays and clayey sands of late Cretaceous age that on summits and ridgelines are overlain by gravels and sands of the Pinehurst Formation [34,35].During the Late Pleistocene, the Pinehurst Formation experienced eolian sand deposition within the study area [36] and in similar settings in South Carolina [37].Elsewhere in the basin, sandy and clayey marine sediments, and minor outcroppings of slates, felsic tuffs, and Triassic sandstone and siltstone, are exposed at lower elevations and along tributaries [34].
The geomorphic framework for the Little River chronosequence is a 17 km long, 2 to 3 km wide valley segment with a contemporary floodplain and 5 alluvial terraces that Suther et al. [29] mapped and dated using OSL (T1-T5b, Figure 2).Table 1 provides terrace age estimates, relative elevations of fluvial surfaces, and classification of the sampled soils by Soil Taxonomy [38].Terrace mapping and dating procedures are discussed in Suther et al. [29] and thus not included herein.The morphological properties of soils in the present chronosequence were evaluated by Suther and Leigh [33].
Terrace OSL age estimates vary between 10 ± 2 (T1) and 94 ± 16 (T5b) ka (Table 1) [29].At the floodplain site, prehistoric overbank sediments returned a 1 ± 0.3 ka age.However, Suther et al. [29] regarded this estimate as possibly low-biased and inferred a middle to late-Holocene (≥1 ± 0.3 ka) to protohistoric age for the sediments.Historical alluvium is superimposed on these Holocene sediments in the uppermost ~100 cm in floodplain pe- Within the Little River basin, surficial upland sediments include nonmarine kaolinitic clays and clayey sands of late Cretaceous age that on summits and ridgelines are overlain by gravels and sands of the Pinehurst Formation [34,35].During the Late Pleistocene, the Pinehurst Formation experienced eolian sand deposition within the study area [36] and in similar settings in South Carolina [37].Elsewhere in the basin, sandy and clayey marine sediments, and minor outcroppings of slates, felsic tuffs, and Triassic sandstone and siltstone, are exposed at lower elevations and along tributaries [34].
The geomorphic framework for the Little River chronosequence is a 17 km long, 2 to 3 km wide valley segment with a contemporary floodplain and 5 alluvial terraces that Suther et al. [29] mapped and dated using OSL (T1-T5b, Figure 2).Table 1 provides terrace age estimates, relative elevations of fluvial surfaces, and classification of the sampled soils by Soil Taxonomy [38].Terrace mapping and dating procedures are discussed in Suther et al. [29] and thus not included herein.The morphological properties of soils in the present chronosequence were evaluated by Suther and Leigh [33].
Terrace OSL age estimates vary between 10 ± 2 (T1) and 94 ± 16 (T5b) ka (Table 1) [29].At the floodplain site, prehistoric overbank sediments returned a 1 ± 0.3 ka age.However, Suther et al. [29] regarded this estimate as possibly low-biased and inferred a middle to late-Holocene (≥1 ± 0.3 ka) to protohistoric age for the sediments.Historical alluvium is superimposed on these Holocene sediments in the uppermost ~100 cm in floodplain pedons and represents overbank deposits that accumulated following extensive immigration and erosive land use by non-indigenous people into the Little River watershed, circa 200 yr BP [29].
dons and represents overbank deposits that accumulated following extensive immigration and erosive land use by non-indigenous people into the Little River watershed, circa 200 yr BP [29].

Soil Classification and Climate
Sample pedons were described and classified by Suther and Leigh [33] and respectively range from Entisols on the floodplain and T1 (Typic Quartzipsamments, Typic Udipsamments) to Ultisols (Arenic Hapludults, Arenic, and Grossarenic Kandiudults) on T2-T5b (Tables 1 and 2).Land cover is dominated by longleaf pine (Pinus palustris) savanna but also includes scattered agricultural fields and residential areas.The study area features a humid subtropical climate [39] with an average annual temperature of 16.9 °C and average January and July temperatures of 6.0 °C and 27.5 °C, respectively.Average annual precipitation is 115.8 cm [40].The North Carolina Coastal Plain has a thermic soil temperature regime [28].

Field Sampling
Sample pedons were situated to limit variability in parent alluvium and topography among pedons of separate terrace levels so that the temporal aspect of pedogenesis could be narrowly isolated.Pedons were positioned on comparable geomorphic features within

Soil Classification and Climate
Sample pedons were described and classified by Suther and Leigh [33] and respectively range from Entisols on the floodplain and T1 (Typic Quartzipsamments, Typic Udipsamments) to Ultisols (Arenic Hapludults, Arenic, and Grossarenic Kandiudults) on T2-T5b (Tables 1 and 2).Land cover is dominated by longleaf pine (Pinus palustris) savanna but also includes scattered agricultural fields and residential areas.The study area features a humid subtropical climate [39] with an average annual temperature of 16.9 • C and average January and July temperatures of 6.0 • C and 27.5 • C, respectively.Average annual precipitation is 115.8 cm [40].The North Carolina Coastal Plain has a thermic soil temperature regime [28].

Field Sampling
Sample pedons were situated to limit variability in parent alluvium and topography among pedons of separate terrace levels so that the temporal aspect of pedogenesis could be narrowly isolated.Pedons were positioned on comparable geomorphic features within each surface (either sandy scroll bars, sandy flats, or sand ridges) that were likely characterized by alluvium with similar textural and stratigraphic attributes, based on soil-landscape relationships inferred from prior geoarchaeological [41] and geomorphic mapping [29] fieldwork that examined soils and sediments across a wide variety of landscape settings at Fort Bragg [33].[42,43].b Soil Taxonomy subgroup classification for soils on each terrace at pit sampling locations and the soil series that best fits their description, as previously reported by Suther and Leigh [33].Where one or more properties of the observed soil fall outside the range of characteristics of the most similar series, that soil is noted as a series taxadjunct, and the subgroup classification of the observed soil is given.For T3a and T5a, nd = no data.c Previously reported by Suther et al. [29]; OSL = optically-stimulated luminescence; nr = not reported.d Ages were obtained from 150-170 µm quartz.Equivalent dose was determined using the single-aliquot regenerative dose (SAR) protocol [44].e Average height above river level (HARL) of landform surface determined from 2 m pixel-edge Light Detection and Ranging (LIDAR) digital elevation model data.f Proportion of areal extent of the mapped segment of Little River valley (see Figure 2) that is occupied by each landform.
At locations of interest, sample pedons were sited on the highest, most stable, bestdrained parts of each landform to reduce drainage differences and toposequence effects among soils of the same terrace [12,23].Soils observed during the previous studies gave perspective on pedologic variability at the landscape positions targeted by the present investigation.Five profiles typical of the soils usually encountered at these locations within each surface were sampled, separated by at least ~5 to 40 m on each terrace.
One profile per terrace was examined in an excavated pit, and four pedons were sampled with a hand auger.The morphology of each soil profile was described with Natural Resources Conservation Service (NRCS) nomenclature [45], according to the procedures of Suther and Leigh [33].Samples were collected for chemical and mineralogical analysis from the most pedogenically well-expressed subsoil (typically B) horizon of each pedon.In floodplain profiles, which lack B horizons, the shallowest C2 horizon in historical alluvium was sampled.We also collected samples of "fresh" sediment for analysis from the modern river bed at five sites for use as a proxy for unweathered parent material.
Table 2. Abbreviated morphological descriptions for soils at sampling pits, adapted from Table 2 of Suther and Leigh [33] and Table 3 of Suther et al. [29].Color and textural characteristics of modern channel bed sediment are also shown.

Landform
Horizon   3 and 4, modern channel bed sediment (CB) samples are from: gravel point bar (1), sand point (2) or mid-channel (3,5) bars, and sand deposit on clay outcrop (4).e OSL sample TU5@90 reported by Suther et al. [29] was obtained from a depth of 90 cm in this C horizon, about 1 m away from the described profile, where the C horizon occurs at a slightly shallower depth.The lower boundary of Little River alluvium in this profile is at 210 cm.f Description is composited from two pedons for illustrative purposes.Ap-B t2 horizons are taken from T4, pedon 1, and B t3-B t5 horizons are taken from T4, pedon 2 [46].Auger refusal on gravel at 220 cm in pedon 1 prevented description of the entire profile.Lower boundary of Little River alluvium at the pedon 2 location is 307 cm.g Lower boundary of alluvium in this profile is at 318 cm, within the Btx6 horizon.na = not applicable; nd = no data.

Chemical and Mineralogical Analyses
Elemental concentrations of the whole soil (<2 mm) and fine sand (0.125-0.250 mm) fractions of the best-developed subsoil horizon from each of the five profiles sampled per surface were determined by inductively-coupled plasma optical emission spectrometry (ICP-OES).Samples were digested by HF/HNO 3 -based microwave-assisted digestion (USEPA Method 3052) [47] and analyzed using a Perkin-Elmer Optima 4300-DV ICP-OES spectrometer according to the quality assurance-quality control standards for the procedure [47].Free Fe determinations were accomplished by citrate-dithionite and ammonium oxalate extractions using the respective NRCS procedures 4G1 and 4G2 [48], followed by analysis with a Perkin-Elmer AAnalyst 200 atomic absorption spectrometer.Internal replicates run every 10 samples by the University of Georgia Soil Characterization Laboratory ensured error associated with Fe analyses was within ±5%.Oxide concentrations were calculated from measured elemental concentrations according to standard ratios.
Clay mineralogy of the most well-expressed part of the B horizon of at least one pedon per terrace was determined by X-ray diffraction (XRD).In floodplain pedons, C horizons from both historical and prehistoric sediments were examined.The <2 µm fraction was isolated by siphoning at the depth and time specified by Gee and Bauder [49] following particle size analysis by hydrometer method.Clays that were not pretreated for Fe or organic matter removal were oriented under vacuum [50] and analyzed by XRD using the following treatments: Mg-saturation and glycol solvation at 25 • C, and K-saturation at 25 • C, 105 • C, 300 • C, and 550 • C. A Philips 3520 x-ray diffractometer with Cu-K-alpha x-radiation and a curved crystal monochromator was used to measure d-spacings.Kaolinite and gibbsite abundances were determined quantitatively from K-saturated clays using a Thermal Analyst Model 2970 differential scanning calorimeter (DSC), according to the procedures described by Tan et al. [51].Relative abundances of other clay minerals were estimated by their diffraction peak heights relative to those of kaolinite and gibbsite.In soils with substantial proportions of gibbsite, DSC was conducted on K-saturated clays before and after Fe oxide removal to account for interferences from goethite.

Chronosequence Variables, Soil Ages, and Statistical Analysis
Age-related trends in the chemistry of the whole soil (<2 mm) fraction were evaluated using the molar ratios of bases to alumina (Bases/Al 2 O 3 ) and bases to resistant oxides (Bases/R 2 O 3 ), the concentrations of K 2 O and dithionite-extractable Fe (Fe D ), and the ratios of oxalate-extractable to dithionite-extractable Fe (Fe O /Fe D ) and dithionite-extractable to total Fe (Fe D /Fe 2 O 3 ).Bases in both molar ratios included K 2 O, CaO, Na 2 O, and MgO.Resistant oxides in the Bases/R 2 O 3 ratio comprised Al 2 O 3 , Fe 2 O 3 , TiO 2 , and ZrO 2 for all samples.Chronosequence trends in fine sand (0.125-0.250 mm) fraction chemistry were characterized using the element oxide concentrations of K 2 O, Na 2 O, and MgO.
The T1 (10 ± 2 ka), T2 (17 ± 4 ka), T4 (75 ± 10 ka), and T5b (94 ± 16 ka) OSL ages utilized in this study were determined by Suther et al. [29] using lateral (T1, T2, T4) or lower vertical (T5b) accretion sands obtained at or near soil sampling sites (Table 1).Suther and Leigh [33] regarded these ages as reliable approximations of soil age at the sampling locations, noting that the interval between the deposition of dated sediment, termination of sedimentation, and inception of pedogenesis was probably small relative to the entire period of soil formation for each surface.
Suther et al. [29] reported two OSL dates from each of the T3b-1 and T3b-2 localities, which are situated within separate map units of T3b (Table 1, Figure 2).Of the two ages at each site, the most reliable estimates for the complete interval of pedogenesis for the T3b-1 and T3b-2 profiles are respectively provided by the 55 ± 15 ka (T3b-1, 190 cm) and 51 ± 12 ka (T3b-2, 180 cm) ages, which were both obtained from unweathered lateral accretion sands [33].These age estimates are therefore used in chronosequence analyses.The reader is referred to Suther and Leigh [33] for critical evaluation of the OSL ages and their relationships to stratigraphy and pedogenesis at each T3b site.
Temporal trends in soil chemical parameters were assessed by least squares regression.We used regression in this context to evaluate possible relationships between soil chemical properties and age, not to formulate predictive chronofunctions.In regressions, untransformed soil chemical parameters functioned as the dependent variables, and pedon age, represented by the mean OSL age from each surface (in years BP), was the independent variable.An age of 100 yr BP, which was assigned to the uppermost A-C horizon sequence by Suther and Leigh [33], was used for sediments from the shallowest C2 horizons in floodplain pedons.Regression models were selected according to (1) their goodness of fit across the entire duration of the chronosequence and (2) their consistency with theoretical expectations of the behavior of the soil system [52].Thus, among functions that could be justified by pedogenic theory for the variable of interest, the model with the highest R 2 value was typically selected.If two models produced similar R 2 values, the simpler of the two functions (i.e., the model with the fewest estimated parameters) was used.Using these criteria, nonlinear regression functions were adopted in some cases and linear models in others.
Chemical properties were additionally assessed using B or Bt horizon data exclusively [53] to ascertain whether including only subsoil parameters produces stronger soil-age relationships.We also conducted regression analyses, both including and excluding present-day alluvium, to evaluate if incorporation of modern sediment (given a 1 yr BP age) affects correlations of chemical properties with time.Temporal trends in clay mineralogy were qualitatively assessed by inter-terrace comparisons of clay mineral relative abundance estimates.

Soil Chemistry
From lower (younger) to higher (older) surfaces, subsoils exhibit systematic trends in both the major element oxide concentrations of their whole soil (<2 mm) and fine sand (0.125-0.250 mm) fractions (Tables S1 and S2), as well as in their concentrations of dithionite-and oxalate-extractable Fe (Table S3).Within the whole soil, Bases/Al 2 O 3 ratios, Bases/R 2 O 3 ratios, and K 2 O concentrations display strong, negative relationships with mean age estimates that are best modeled by single, two (Bases/R 2 O 3 , K 2 O), or three (Bases/Al 2 O 3 ) parameter exponential decay functions (R 2 ≥ 0.77, Figure 3A-C, Table S4).Among Fe parameters, both Fe D concentrations and Fe D /Fe 2 O 3 ratios increase with age, while Fe O /Fe D ratios decrease over time (Figure 3D-F).Relations between Fe variables and mean age are best represented by a single, two-parameter exponential growth function, with regard to Fe D concentrations (R 2 = 0.81); a single, three-parameter exponential decay function, in the case of Fe O /Fe D ratios (R 2 = 0.63); or a two-parameter power function, for Fe D /Fe 2 O 3 values (R 2 = 0.62) (Figure 3D-F; Table S4).In the context of these models, dithionite-Fe has the strongest correlation with age.
In the fine sand fraction, K 2 O, Na 2 O, and MgO exhibit the strongest correlations with age among individual element oxide concentrations [46].These parameters have moderate to strong, negative relationships with mean age that are best modeled by simple linear regression (R 2 ≥ 0.65, Figure 4A-C, Table S5).
For most chemical properties, using data from the best-developed subsoil horizon of every pedon (Figures 3 and 4; Tables S4 and S5) produces stronger correlations with age than using only B or Bt horizon properties (Table S6).Modern channel alluvium displays greater variability in chemical composition than do floodplain and terrace samples from within the same landform (Tables S1 and S2), probably because localized differences in hydraulic sorting between sample sites produced larger contrasts in the amount and type of sand-sized heavy minerals among sediments at modern channel sampling locations than among deposits at floodplain or terrace sites within the same surface.As a result, chemical properties exhibit stronger correlations with age if modern channel alluvium is excluded from regressions (Table S6).
The presented equations should not be interpreted as robust, predictive chronofunctions, owing to errors in OSL ages that produce some uncertainty in the exact form of temporal trends and regression assumption violations (e.g., non-normal distributions of residuals and dependent variables) that occur in both nonlinear (Table S4) and linear (Table S5) models.Nevertheless, the above findings broaden our understanding of temporal changes in Coastal Plain alluvial soils during the late Quaternary. of sand-sized heavy minerals among sediments at modern channel sampling locations than among deposits at floodplain or terrace sites within the same surface.As a result, chemical properties exhibit stronger correlations with age if modern channel alluvium is excluded from regressions (Table S6).The presented equations should not be interpreted as robust, predictive chronofunctions, owing to errors in OSL ages that produce some uncertainty in the exact form of temporal trends and regression assumption violations (e.g., non-normal distributions of residuals and dependent variables) that occur in both nonlinear (Table S4) and linear (Table S5) models.Nevertheless, the above findings broaden our understanding of temporal changes in Coastal Plain alluvial soils during the late Quaternary.

Gibbsite Determination
XRD data show that the Btx2 horizons of pedons 1 and 2 on T5b contain substantial amounts of gibbsite (Figure 5, Table 3).DSC analysis was performed on T5b soil clays both with and without Fe oxides removed to account for interferences from goethite, whose endotherm occurs in a similar temperature range as that of gibbsite.Pretreated T5b clays contain 44% gibbsite, approximately double the concentrations of their untreated counterpart samples (Table 3).It is therefore likely that other soils in the chronosequence with significant amounts of gibbsite (T3b-1, pedon 1; T3b-2, pedon 1; and T4, pedons 1 and 2, 4-10%) also contain slightly higher percentages of the mineral than those reported for untreated samples.Munsell hues of 10YR-7.5 YR [33] suggest that the most well-developed horizons of soils occurring on the floodplain and T1-T4 are dominated by goethite [54].However, other Fe oxides are present.A ~0.626 nm peak in the X-ray diffractograms of prehistoric floodplain and T2 pedons indicates the presence of lepidocrocite (Figure 5).The color of T2 Bt horizons (7.5YR 5/7) is also suggestive of this mineral [54].Munsell hues of 5YR-2.5YR[33] indicate that hematite occurs in T5b pedons [54], but because the color of goethite can be masked by small quantities of hematite, it is probable that T5b subsoils are also dominated by goethite.With age, poorly crystalline (oxalate-extractable) Fe oxides decrease in abundance relative to total free Fe (Figure 3E, Table S3) and are lowest in T5b pedons.
Munsell hues of 10YR-7.5 YR [33] suggest that the most well-developed horizons of soils occurring on the floodplain and T1-T4 are dominated by goethite [54].However, other Fe oxides are present.A ~0.626 nm peak in the X-ray diffractograms of prehistoric floodplain and T2 pedons indicates the presence of lepidocrocite (Figure 5).The color of T2 Bt horizons (7.5YR 5/7) is also suggestive of this mineral [54].Munsell hues of 5YR-2.5YR[33] indicate that hematite occurs in T5b pedons [54], but because the color of goethite can be masked by small quantities of hematite, it is probable that T5b subsoils are also dominated by goethite.With age, poorly crystalline (oxalate-extractable) Fe oxides decrease in abundance relative to total free Fe (Figure 3E, Table S3) and are lowest in T5b pedons.3A) and distinguish pedons on the historical floodplain and T1 (≤10 ± 2 ka) from older soils (≥17 ± 4 ka).Floodplain and T1 profiles have Bases/Al 2 O 3 ratios of >0.4,whereas older pedons yield lower values.

Fine Sand Chemistry
Fine sand fraction concentrations of K 2 O and Na 2 O are also useful age indicators (R 2 ≥ 0.88; Figure 4A,B).Fine sand fraction K 2 O has the strongest correlation with age of any parameter (R 2 = 0.92) and is the only property that clearly differentiates soils on terminal Pleistocene and younger surfaces (5718-9538 mg kg −1 ; Table S2) from those on older terraces (384-4714 mg kg −1 ).K 2 O and Na 2 O may also be used to separate T3b soils from those on T4 and T5b, as respective concentrations for these element oxides range from 732-4714 and 512-740 mg kg −1 in T3b pedons but are ≤453 and ≤480 mg kg −1 in T4 and T5b profiles.Because K 2 O concentrations are higher and display a greater range across terraces than do Na 2 O concentrations, K 2 O is the most useful fine sand fraction chronosequence parameter.MgO shows moderately strong correlation with age (R 2 = 0.65; Figure 4C), but very low concentrations (53-178 mg kg −1 ) and intra-terrace variation limit its utility for distinguishing soils by terrace level.

Within-Surface Variation in Soil Chemistry
The present investigation assesses the typical pedogenic expression of soils with comparable parent sediments and landscape positions within terraces; it does not attempt to evaluate the entire extent of soil variability on each surface.In this framework, profiles at the geomorphic settings of interest display a range of pedogenic development within individual terraces.On southeastern US Coastal Plain uplands, at the landscape scale, soil spatial variability has been linked to the effects of fluvial dissection, relief, and water table position on soil genesis [55][56][57][58].At local scales, more pronounced variation in morphological characteristics (e.g., surface horizon thickness and vertical clay distribution) has been observed over-printed on these broader soil-geomorphic relationships and attributed to the growth and expansion of disturbances from faunalturbation and tree uprooting [59,60].
In some instances, our soils exhibit intra-surface variation that is produced by localized differences in terrace alluvium, topography, and drainage.In other cases, variability occurs without obvious changes in these variables and may be attributable to past biotic activity or to locally differing conditions at the onset of pedogenesis that are now undetectable [59].Because clay mineral data are limited to one or two pedons per surface, direct assessment of intra-terrace variability is only possible for chemical properties.Among these parameters, within T1, T2, and T3b-2, variation derives in part from geochemical differences related to parent material texture.For these locations, the parent alluvium of several profiles is either more coarsely-or finely-grained than that of counterpart pedons within the same terrace.This localized sedimentological variation apparently promotes B horizon clay percentages that are either significantly lower (T2, pedon 5, 4.5%) or higher (T1 pedon 3, 7.9%; T3b-2, pedon 5, 21.5%) than those of other soils on the same surface [33,46].Such variability in parent material clay content also likely contributes to intra-surface variation in the Bases/Al 2 O 3 and Bases/R 2 O 3 ratios, as well as in Fe D concentrations, among T2 and T3b-2 pedons, because most Al and Fe in southeastern US Ultisols occurs in the <2 µm fraction within oxides [61] and clay minerals.Pedon 5 on T2, whose subsoil clay content is 9.3-14.3%lower than that of its counterpart profiles [33], registers the lowest Al 2 O 3 , Fe 2 O 3 , and Fe D concentrations (Tables S1 and S3), and the highest Bases/Al 2 O 3 (0.35), Bases/R 2 O 3 (0.24), and Fe O /Fe D (0.39) ratios (Figure 3) of any soil on the surface.In contrast, pedon 5 on T3b-2, which contains 8.2-15.0%more clay than other T3b-2 profiles, has the highest Al 2 O 3 and Fe 2 O 3 concentrations, the next to highest Fe D concentration, and the lowest Bases/Al 2 O 3 (0.12), Bases/R 2 O 3 (0.10), and Fe O /Fe D (0.03) ratios on the surface (Tables S1 and S3, Figure 3).Additionally, finely-grained samples on T1, T2, and T3b-2 exhibit higher whole soil K 2 O concentrations than their coarser counterparts.Muscovite mica, which contains K and is visible in minor quantities in samples, likely accounts for this.Because of its platy shape, mica has a lower settling velocity than detrital minerals of comparable density that occur as equant grains (e.g., quartz) [62].Greater abundance of sand-and silt-sized mica (and, in turn, higher K 2 O concentrations) would be expected in soils derived from alluvial sediments whose quartz-dominated sand fractions are finer than those of counterpart profiles within the same terrace.
Internal drainage effects caused by topographic variation probably also contribute to within-terrace variability, particularly with respect to Fe chemistry.Among T3b-1 soils, greater Fe D concentrations (9401-10,473 mg kg −1 ) occur in pedons at higher elevations within a sandy flat whose internal relief varies by ~40 cm.Across the other terraces, intra-surface variation in free Fe is likely influenced by the "edge effect," whereby profiles along terrace edges possess subsoils with greater rubification, clay contents, and Fe D concentrations than pedons at interior positions because of their increased depth to water table [55,63].For soils on these surfaces, the majority of profiles within 10-40 m of scarps and stream-cut edges of terraces exhibit higher subsoil Fe D concentrations than pedons farther inland.Among the soils whose landscape positions are associated with higher intra-terrace Fe D concentrations, most pedons also display greater subsoil rubification [33] than other profiles on the same surface.
It is often difficult to discern to what degree soil variation represents the outcome of earlier biotic perturbations, magnification of localized contrasts in initial ("time zero") conditions, or merely stochastic variability [64], and the same applies here.But the potential for such mechanisms to influence intra-terrace variation should be acknowledged, given that these phenomena have been linked to short-range variability in similar regional settings [59,60,65].Amplification of disturbances from past bioturbation, including pedoturbation by ants and termites, provides a reasonable explanation for within-terrace variability that occurs where evidence of site-to-site differences in fluvial sediments, relief, or drainage is lacking.For parameters influenced by subsoil clay content, this point is especially relevant, given that related characteristics in similar Coastal Plain Ultisols, like pedon clay distribution, have been attributed to floral-and faunalturbation [59,60].
Although chemical properties successfully discriminate pedons of particular surfaces in some cases (see Section 4.1), as a consequence of intra-terrace variation, no single parameter by itself effectively differentiates profiles by individual terrace level across the entire chronosequence.Nevertheless, distinct temporal trends are evident in the evolution of these soils.

Soil Chemistry
Bases/Al 2 O 3 and Bases/R 2 O 3 ratios have relationships with mean age that resemble exponential decay functions (Figure 3A,B).The chemistry of floodplain through T3b subsoils indicates that both ratios decrease rapidly during the first 13-21 kyr of development, then decline more gradually thereafter.After 65-85 kyr of pedogenesis, the ratios stabilize, with little difference in either index among T4 (75 ± 10 ka) and T5b (94 ± 16 ka) soils.
Trends in the Bases/Al 2 O 3 and Bases/R 2 O 3 ratios result in part from the depletion of bases (CaO, K 2 O, MgO, and Na 2 O), which weather quickly from the whole soil during the first 40-70 kyr of pedogenesis but stabilize at similar low concentrations in T4 and T5b pedons (Table S1).Whole soil K 2 O concentrations exemplify this trend (Figure 3C).
Temporal trends in bases reflect weathering in all size fractions of soil, given that depletion of bases is evident in both whole soil and fine sand chemistry (Figures 3 and 4), as well as in soil clays, which display losses (or decreases in concentration) of base-bearing phyllosilicates, like mica, smectite, and HIV, with age (Table 4, Figure 5).However, because these species occur in such small quantities relative to other constituents of the whole soil, most bases lost must have been supplied by sand and silt fraction minerals.
The sand fractions of soils are overwhelmingly dominated by quartz.Although bases are clearly depleted with age, they together comprise ≤1% (1312-10,014 mg kg −1 ) of the <2 mm fraction of unweathered channel bed and historical floodplain sediments, indicating a paucity of weatherable minerals.This siliceous mineralogy is expected for floodplain soils like those of the present study that are sourced primarily from Coastal Plain deposits [28], and it is consistent with the general lack of weatherable minerals in the upper ~1-10 m of regional upland sediments on all but the youngest surfaces of the lower Coastal Plain [4,66].Among bases, K 2 O is most abundant and, in channel and floodplain alluvium, occurs at concentrations 4.1-11.4times higher than those of CaO, MgO, and Na 2 O combined (Table S1).Sources of K 2 O include muscovite mica (see Section 4.2) and possibly K-feldspars, which are found in Cretaceous Coastal Plain sediments in the area [35,67].The similar base concentrations among 75 ± 10-94 ± 16 ka pedons (Table S1) represent the exhaustion of unstable primary silicates from the sand and silt fractions after approximately 60-90 kyr.
Fine sand concentrations of K 2 O, Na 2 O, and MgO display negative, linear relationships with mean age, suggesting that bases in this size fraction decrease at more constant rates than they do in the whole soil (Figure 4A-C).The initially rapid decrease in bases in the <2 mm fraction in historical through 55 ± 15 ka soils likely represents the preferential weathering of silt-sized silicates, which weather more quickly than minerals in the fine sand fraction, owing to their larger surface area.Although fine sand fraction K 2 O steadily decreases during the first 65-85 kyr of pedogenesis, similar to K 2 O in the whole soil, it stabilizes in 75 ± 10-94 ± 16 ka pedons (Figure 4A).This implies unstable K-bearing detrital minerals have a residence time of about 60-90 kyr in the fine sand fraction in this environment.Although fine sand Na 2 O and MgO concentrations display slight decreases between T4 and T5b (Figure 4B,C), the extremely low values and inter-terrace overlap of T4 and T5b concentrations for these element oxides suggest their source minerals approach exhaustion after a similar duration.
Bases/Al 2 O 3 and Bases/R 2 O 3 ratios also reflect temporal trends in whole soil alumina concentrations.Subsoil alumina increases rapidly during the first 13-21 kyr of pedogenesis, is relatively stable in T2 and T3b soils (17 ± 4-55 ± 15 ka), and gradually increases with age in >55 ± 15 ka profiles (Table S1).These trends represent the enrichment of subsoils with alumina, attributable to the accumulation of Al-bearing clays, including kaolinite and gibbsite, in B horizons.Clay contents of Little River subsoils increase with age [33].The pronounced increase in alumina between T1 and T2 soils (Table S1) reflects the initiation of argillic horizon development in T2 pedons and their concomitant clay increase relative to T1 profiles (4.5-18.8versus 2.0-7.9% of <2 mm), whereas comparable alumina concentrations among 17 ± 4-55 ± 15 ka soils correspond to their similar respective Bt horizon clay contents (4.5-18.8 and 6.5-21.5%)[33].T4 and T5b soils (75 ± 10-94 ± 16 ka), on average, display a progressive increase in alumina relative to T2 and T3b profiles that result from their comparatively higher clay quantities (14.3-36.5%,[33]), as well as their mineralogy, which in T5b profiles includes a substantial proportion of gibbsite (Tables 3 and 4).Thus, alumina trends derive from the formation and continued development of argillic horizons and their constituent clay mineral assemblages with age.
Free Fe contents increase and probably change episodically over time.Fe D concentrations are lowest in floodplain pedons, have slightly higher values among T1-T3b profiles, and are greatest on older terraces, with the largest concentrations occurring in T5b subsoils (Figure 3D).Redness also increases with age [33] and has a strong correlation with extractable Fe [46].Increasing subsoil Fe D and rubification are believed to represent the illuviation and transformation of free Fe inherited from parent sediments, not Fe oxide formation via weathering of primary minerals.
Weatherable minerals have low concentrations in alluvium.Additionally, although respective subsoil Fe D concentrations increase from 886-3146 to 14,317-37,475 mg kg −1 from floodplain to T5b pedons, Fe within primary minerals (Fe 2 O 3 -Fe D ) is relatively consistent from lower to higher terraces, and small by comparison (974-11,345 mg kg −1 , Table S3).Considering the comparable parent material among surfaces [33], these data indicate that only minor differences in primary Fe content, and therefore only small variation in the amount of Fe attributable to detrital mineral weathering, exist between profiles on higher versus lower surfaces.
Daniels et al. [63] similarly attributed high subsoil Fe D concentrations to illuviation of inherited free Fe in Paleudults near upland terrace edges on the middle Coastal Plain of North Carolina.In those soils, high extractable Fe contents represented neither detrital mineral weathering nor site-to-site variation in parent material, but rather reflected locally deeper water tables and nearly continuous oxidizing conditions that inhibited free Fe losses and promoted the translocation of Fe oxides and clays to argillic horizons [63].Daniels et al. [63] suspected that most Fe within their soils originated from coatings on clay and sand grains in parent sediments.For the Little River chronosequence, Suther and Leigh [33] emphasized that the silt and clay fractions of overbank alluvium were a likely key source of extractable Fe, as the vertical accretion deposits that compose floodplain C horizons lack coated sand grains but have yellow-brown colors characteristic of goethite.
Iron contained at shallow depths within Little River terrace sediments were likely translocated to subsoils as E and B horizons developed.Increasing Fe D /Fe 2 O 3 ratios with age (Figure 3F) are consistent with this, as Fe D concentrations would be expected to comprise a progressively greater proportion of total Fe 2 O 3 if a small but stable quantity of Fe bound in primary minerals was being augmented by comparatively larger illuvial free Fe additions over time.The tendency for most free Fe to be clay-sized would facilitate this process.Subsoil Fe accumulation coincides with a general decrease in Fe O /Fe D ratios with increasing terrace elevation (Figure 3E).Declining Fe O /Fe D values represent an increase in crystalline Fe oxide species, particularly goethite, within the chronosequence, along with hematite in T5b Btx horizons.
B horizon Fe typically displays positive correlations with time in Ultisol chronosequences of the US Coastal Plain [19,23], suggesting that free Fe trends across Little River terraces are in part attributable to the duration of weathering.Scrutiny of Fe D , Fe O /Fe D , and Fe D /Fe 2 O 3 values against mean age estimates, however, suggests that Fe chemistry probably experienced periods of rapid, nonlinear development (Figure 3D-F).Such patterns may reflect the effects of improvements to internal drainage and/or paleoclimatic fluctuations from warm and moist to cold and dry conditions during the late Quaternary.
The increase in free Fe between floodplain and T1 soils (Figure 3D,F) is accompanied by an increase in redness that Suther and Leigh [33] attributed not only to a greater interval of soil formation, but possibly also to improvements in soil drainage resulting from deepening of the water table beneath T1 that likely followed floodplain incision and terrace formation.Such changes may have facilitated E horizon development and eluviation of Fe oxides to T1 subsoils.Relief-induced drainage effects may have also impacted free Fe contents on T5b, which is 29.0 m above the modern channel and >14 m higher than T4 (Table 1).In this context, T5b pedons occupied the best-drained landscape position in the chronosequence that, compared to other surfaces, likely better promoted the translocation of free Fe, the formation of crystalline Fe oxides from poorly crystalline forms, and hematite development [33].Also, exposure to more weathering-intensive paleoclimatic conditions provides a possible alternative (or additional) explanation for the high free Fe contents of T5b profiles relative to younger soils.Suther and Leigh [33] note that pedogenesis on T5b (94 ± 16 ka, 78-110 ka) may have overlapped with the later substages of the prior interglacial (Oxygen Isotope Stages 5d-5a, 115-75 ka), which in the Southeast were generally warmer and more moist than the following glacial [68,69] and probably featured more intensively leached conditions.

Clay Mineralogy
Ascending from lower to higher terraces, the clay fraction of soils changes from a mixed assemblage of kaolinite, smectite, HIV, quartz, and mica to a mineralogy dominated by kaolinite and gibbsite.Modern channel alluvium, as well as historical (≤200 yr BP) and prehistoric (≥1 ± 0.3 ka) floodplain soils, are dominated by kaolinite but also contain smectite and HIV (Figure 5, Table 4).Similar relative abundances of smectite in these deposits suggest that the smectite in floodplain pedons is inherited.The lack of pedogenic alteration to floodplain soils (Table 2) [33], and the presence of clay fraction mica in only trace quantities in modern and historical sediments (Figure 5, Table 4), indicate that smectite is not forming from the weathering of mica or other silicates.Alternatively, smectite is likely sourced from Lower Cretaceous Coastal Plain sediments in the Little River basin that contain the mineral [34,35].
A shift in the diffraction peak from 1.77 nm in historical to 1.65 nm in prehistoric alluvium (Figure 5) indicates that smectite begins to degrade within several thousand years.The incomplete collapse of smectite in K-saturated floodplain specimens after heating to 105 • C (Figure S1) suggests the presence of hydroxy-interlayers within its structure [70,71].Hydroxy-interlayering in smectite may represent the beginning of its transformation to HIV; the more pronounced HIV (1.4 nm) diffraction peak in the prehistoric floodplain soil supports this.
The absence of smectite in the T1 soil (Figure 5) probably results from weathering, not a change in source material.Given the relatively thick, homogeneous sedimentary deposits in the Little River's headwaters [34,35], significant variability in source material through the late Quaternary is unlikely.Furthermore, the presence of smectite in both historical and prehistoric sediments eliminates the possibility that the mineral is a recent addition to alluvium exclusively attributable to historical gullies eroding into smectitebearing materials.The mineralogies and ages of T1-T5b pedons indicate that smectite is destroyed by weathering in <8-12 kyr.
Although it is usually regarded as a pedogenic mineral [25], in our soils, much of the HIV is likely inherited.HIV comprises a significant proportion of the <2 µm fraction of unweathered parent alluvium, based on relative abundance estimates from channel bed and historical and prehistoric floodplain deposits (Figure 5, Table 4).Upland-derived sediments provide the most plausible source for HIV in unweathered alluvium, as the mineral is known to be abundant in the epipedons of well-drained Coastal Plain Ultisols [25,[72][73][74][75] and sediments eroded from such soils on uplands and hillslopes in the Little River watershed comprise a meaningful proportion of the river's sediment load.In addition to inherited quantities, modest amounts of HIV may be attributable to weathering.Although base element concentrations of floodplain and channel bed sediments reveal that the weatherable mineral contents in alluvium are insufficient to yield substantial quantities of secondary clays (Tables S1 and S2), some HIV may have formed from smectite, as noted above.Also, whereas trace quantities of clay-sized mica (Figure 5, Table 4) are insufficient to produce substantial amounts of HIV, it is possible that some HIV in Little River pedons formed from the weathering of silt fraction muscovite, which is known to be an HIV precursor in similar soils elsewhere in the region [72,73,76].
T2-T4 subsoils contain much less HIV than kaolinite (Figure 5, Table 4).Numerous studies report high proportions of HIV in near-surface horizons and/or sandy epipedons and an increase in kaolinite relative to HIV beneath eluvial/illuvial boundaries and/or with depth [25].Declining HIV abundance with depth has been attributed to kinetics [74]; thermodynamics [77]; preferential eluviation of finer kaolinite particles [72,75]; and a lack of kaolinite source material in upper horizons [72].Here, greater subsoil concentrations of kaolinite than HIV partly reflect parent material influence.But preferential eluviation of kaolinite may also play a role, because further reduction of HIV relative to kaolinite occurs between floodplain and T1 versus T2-T4 soils, coincident with argillic horizon development.Kaolinite enrichment is expected in argillic horizons, given that fine clay (<0.2 µm) is more readily eluviated than coarse clay (0.2-2.0 µm), and that kaolinite is typically more abundant than HIV in the fine clay fraction, whereas HIV is more concentrated in the coarse clay and fine silt (2.0-5.0 µm) separates [25,72,73].
Most kaolinite now concentrated in T2-T4 subsoils probably was inherited from the parent alluvium, which is dominantly kaolinitic in the clay fraction and primarily sourced from Cretaceous fluvial and marine deposits high in kaolinite [34,35,67].This interpretation is consistent with evidence that soils on Coastal Plain uplands inherit high kaolinite contents from their parent material [78].Although limited kaolinite formation from weathering of sand and silt fraction silicates cannot be ruled out, low abundance of weatherable minerals suggests that the amount of kaolinite formed this way would be subordinate to inherited quantities.It is similarly unlikely that large amounts of kaolinite are produced from weathering of clay fraction phyllosilicates in T2-T4 soils, given the comparable HIV abundances and negligible mica contents of T2-T4 samples, and considering that smectite appears to weather to HIV early in the chronosequence.
Relative abundance estimates suggest kaolinite is weathering to gibbsite.Pedons with significant amounts of gibbsite (T3b-2, pedon 1; T4, pedons 1-2; T5b, pedons 1-2) display a relative increase in gibbsite and decrease in kaolinite from lower to higher surfaces (Tables 3 and 4; Figure 5).This is most evident in comparison of T5b clays, which have similar respective proportions of gibbsite and kaolinite (22-24% and 24-33% in untreated samples), with untreated T4 clays, which are dominated by kaolinite and contain only 9-10% gibbsite (Table 3, Figure 5).The fact that HIV, the only other clay mineral capable of supplying the Al required for gibbsite formation, occurs in comparable concentrations among T2-T5b soils suggests that gibbsite is not forming from HIV.
Direct formation of gibbsite from feldspars has been documented in the nearby Blue Ridge mountains [79].However, this mechanism cannot explain differences in gibbsite abundance between T4 and T5b pedons, as the chemistry of T4 profiles indicates most unstable sand-and silt-sized primary aluminosilicates are depleted from soils earlier in the chronosequence.This suggests that, at least for T5b soils, gibbsite is forming from kaolinite.Gibbsite formation from kaolinite in Coastal Plain Ultisols has been inferred by others [19,61,80,81] and is usually attributed to advanced weathering and/or desilicating pedoenvironments with enhanced drainage.Declining kaolinite concentrations, gibbsite formation, and the progressive loss of quartz in T3b through T5b subsoils imply significant silica losses from their clay fractions.
High fluvial terraces in the Coastal Plain often contain Paleudults and Kandiudults with clay mineralogies similar to those of upland soils, and these high terrace Ultisols reflect the influence of time, as well as the effects of landscape evolution and other factors [25].As with T5b Fe contents, the high proportion of gibbsite in T5b pedons may derive from the integrated influence of age, relief-induced drainage effects, and paleoclimate.

Argillic Horizon Development
Little River profiles have sandy epipedons and subsoils whose clay contents increase over time, with T4 and T5b argillic horizons displaying maximum respective clay percentages of 14-37% and 25-29% [33].Suther and Leigh [33] attributed increasing clay contents to illuviation of inherited clay, and potentially also to an integrated bioturbation-translocation mechanism, in which clays distributed throughout sandy alluvium are brought to the surface by bioturbative activity, then moved downward to accumulate in subsoils via eluviation-illuviation.This interpretation is consistent with the view of Phillips [82], who argued that a linked bioturbation-translocation process was the best explanation for sandy surface versus loamy subsoil textures in Ultisol profiles on the Pamlico and Talbot surfaces of the lower Coastal Plain in North Carolina.In this model, faunalturbation and floralturbation move clay and material of various particle sizes from pockets of fines spread throughout sandy sediments to upper horizons; lessivage translocates clay to the subsoil; and bioturbation promotes pedogenic structure that in turn encourages downward percolation that enables continued illuviation over time.
Suther and Leigh [33] state that an appreciable amount of subsoil clay in Little River pedons was probably supplied by the translocation of inherited fines, but they note proxies for parent alluvium have clay contents of only ~1-6%.Such low parent material clay content, along with A+E horizon thicknesses that average <50% of solum thickness among Ultisol profiles and clay maxima and cumulative argillic horizon thicknesses of 14-37% and 65-208 cm in T3b-1, T4, and T5b pedons [33], suggest that inherited fines alone were insufficient to produce observed subsoil clay quantities.Phyllosilicates and oxides formed by the weathering of sand-and silt-sized minerals may account for some clay, but the geochemistry of unweathered alluvium indicates that its weatherable mineral concentrations are inadequate to yield copious clay quantities via neoformation (see Sections 4.3.1 and 4.3.2).Dust deposition also fails to explain subsoil textures.Although modern dust sources exist [83,84] and the interior Coastal Plain experienced deposition of windblown sand during the Late Pleistocene [37,85,86], silt depth distributions in profiles are consistent with the sedimentology of terrace alluvium, rather than eolian influx [33].Furthermore, the lack of clay in Pleistocene dune sands elsewhere in the region [87,88], as well as in riverine [89], Sandhills upland [33], and relict coastal [90] landscapes in eastern North Carolina, suggests that late Quaternary dust inputs did not constitute a major source of silt and clay particles.
Given this context, a linked bioturbation-translocation mechanism offers a possible explanation for argillic horizon clay that cannot be attributed to alternative sources.In subsoils, argillans on sand grains and/or ped faces indicate illuviation from A and E to Bt horizons is occurring, and C horizon sands that contain thin clay beds, as well as sandy subjacent Coastal Plain sediments with large clay lenses, provide available fines sources [33].Additionally, bioturbation is commonplace in sandy Coastal Plain soils [60,[91][92][93][94], including biologic activity that brings material upward in profiles [95][96][97].Among these processes, Suther and Leigh [33] indicate that faunalturbation by ants and termites provides the most plausible clay source.Ground nesting ants are capable of burrowing to depths of >100-200 cm into C horizons and underlying sediments in sandy settings [82,98,99]; can transport large volumes of material (0.5-1.5 t yr −1 ) to the surface in some landscapes [100], including from depths in excess of Soil Syst.2022, 6, 1 22 of 29 200 cm [99]; and are capable of doing so without displacing much coarse sediment and without disrupting fining upward trends in the gravel and very coarse sand fractions that Little River profiles inherited from their parent alluvium [33].Considering these factors, Suther and Leigh [33] viewed bioturbative additions as a viable fines source in the present soils.
This study was not designed to assess the influence of bioturbation on pedogenesis, so explicit evaluation of bioturbative fines inputs into Little River subsoils is not possible.Nonetheless, chemical and mineralogical data are compatible with a pedogenetic model that incorporates this process, in which the formation and mineralogy of argillic horizons are driven mainly by the illuviation of inherited fines, possible clay additions from a combined bioturbation-translocation mechanism, and the transformations of secondary phyllosilicates and oxides.
Kaolinite is the most abundant <2 µm mineral in all but T5b soils and remains the predominant species as clay content increases from T1-T4 pedons.Illuviation of inherited kaolinite likely contributed to this trend, as the mineral is abundant in alluvium and readily eluviated (See Section 4.3.2).Kaolinite could also be supplied to subsoils via bioturbationtranslocation, if the process is active, given that the mineral occurs at depth in both C horizon alluvium (Figure 5, 210-220 cm floodplain sample) and in underlying Coastal Plain sediments [34,35].Increasing free Fe over time is compatible with the proposed model, as well.Most free Fe appears to have been inherited and concentrated in subsoils by eluviation-illuviation during E and B horizon formation (see Section 4.3.1).The predominantly fine particle sizes [54], tendency to illuviate [73,101], and presence in alluvium and subjacent sediments [34,35] of free Fe oxides indicate they also would have been subject to bioturbation-translocation, were it occurring.Additionally, observed increases in Al 2 O 3 and Fe 2 O 3 concentrations with age (Table S1, Figure 3A,B) would be expected in argillic horizons receiving illuvial inputs of kaolinite and Fe.
Low concentrations (<10%) of clay fraction HIV in T2-T5b subsoils are also consistent with argillic horizons whose clays were primarily sourced by illuvial processes.Although modest amounts of HIV may illuviate [75], because it is typically concentrated in the coarse clay and fine silt fractions, large quantities of the mineral would not be expected in Bt horizons that were formed mainly by the illuviation of inherited clays and/or fines delivered via bioturbation-translocation.The same is true for <2 µm quartz, which occurs in only low (<10%) or trace concentrations in most argillic horizons (Figure 5, Table 4) and is also less susceptible to illuviation than finer particles, owing to its coarse clay size [101].
In addition to eluvial-illuvial processes, the transformations of secondary clays and oxides also contribute to the mineralogies of argillic horizons.Crystalline Fe oxides become more abundant than poorly crystalline forms with increasing terrace height, culminating in hematite formation on T5b, and gibbsite forms within the solum, probably by weathering of kaolinite (see Sections 4.3.1 and 4.3.2).Gibbsite may form in argillic horizons from kaolinite that originally was either inherited or delivered to surface horizons by bioturbation, then translocated to subsoils prior to weathering.It is also possible that some gibbsite forms in surface horizons and is translocated to subsoils later, as gibbsite occurs in the A and E horizons of some Ultisols [72,73] and is subject to illuviation [101].
Our data do not allow for the definitive reconstruction of all clay sources and weathering pathways in the present soils.However, because "within solum" sedimentary inheritance does not fully explain Bt horizon clay content, clays sourced from neoformation and eolian deposition appear to be limited, and the requisite components of a combined bioturbation-translocation process are present, we view the illuviation of inherited clay, acting together with possible translocation of bioturbative fines additions and the transformations of secondary phyllosilicates and oxides, as a viable pedogenetic model for argillic horizon development.The morphological [33], chemical, and mineralogical properties of Little River soils are compatible with this interpretation.

Regional Variation in US Coastal Plain Chronosequences
Increases in Fe D concentrations and Fe D /Fe 2 O 3 ratios, decreases in Fe O /Fe D values, and development of an increasingly stable clay mineral suite with age are in general agreement with the results of prior research.Markewich et al. [21] reported Fe D and Fe 2 O 3 concentrations that exhibit logarithmic increases with age in soils on 28 ka to 1 Ma fluvial and marine deposits.Increasing Fe D , Fe 2 O 3 , and Fe D /Fe 2 O 3 values with logarithmic temporal trends were also documented in soils on 90 ka to 13 Ma terraces of Virginia's James River, along with increasing clay fraction gibbsite [19].
Our pedons, however, appear to develop more rapidly with respect to the above parameters than soils of previous research.Profiles on T5b (94 ± 16 ka) have respective Fe D concentrations and Fe D /Fe 2 O 3 ratios of 14,317-37,475 mg kg −1 (2.5-3.7%) and 0.75-0.91,as well as clay fraction gibbsite abundances of 22-24% (44% following Fe oxide removal) that are similar to those of kaolinite.In pedons of comparable age (60-125 ka) examined by Markewich et al. [21] and Howard et al. [19], argillic horizons possess Fe D concentrations and Fe D /Fe 2 O 3 ratios of 1.2-2.5% and 0.37-0.50,respectively, and contain less gibbsite than kaolinite.Respective ages of 0.6-1.0Ma [21] and at least 0.7-1.6Ma [19] are needed for these soils to reach the maximum Fe D concentrations of T5b profiles in the Little River valley, and no pedons attained Fe D /Fe 2 O 3 ratios or gibbsite abundances comparable to those of the oldest Little River soils.Even among James River profiles dating from 3.4-13.0Ma, maximum respective Fe D /Fe 2 O 3 values and gibbsite concentrations do not exceed 0.6-0.7 and 6.1-10.9%[19].
Suther and Leigh [33] indicate that the apparent expression of more rapid soil morphological development in Little River pedons probably results from their sandy, siliceous parent sediment, which is derived from the quartz-rich, coarse-grained deposits of the upper Coastal Plain Sandhills.This material probably led to more freely drained soils than did the more finely-grained, mineralogically mixed parent sediments on the terraces of Markewich et al. [21] and Howard et al. [19], which are sourced in both the Coastal Plain and neighboring Piedmont, Blue Ridge, and Valley and Ridge provinces [33].Apparent faster rates of soil chemical and mineralogical development are probably also explained by this parent material contrast, as the greater permeability and enhanced internal drainage likely associated with Little River alluvium would have favored more rapid chemical weathering, gibbsite formation, and illuviation of free Fe.The low weatherable mineral concentrations and consequent "pre-weathered" nature of Little River sediments would also favor development of chemical and clay mineralogical properties consistent with a more advanced state of weathering, including Fe D /Fe 2 O 3 ratios and gibbsite concentrations that are far greater than those of soils of comparable age along Piedmont-draining rivers.The differing temporal trends in these chronosequences show that there are substantial regional distinctions in rates of development for certain alluvial soil properties that are probably linked to provenance-related differences in the texture and mineralogy of parent sediments.
There are also noteworthy distinctions between the chronosequence equations for Little River chemical properties and the chronofunctions of prior research.In previous Coastal Plain chronosequences [19,21], Fe D concentrations, Fe 2 O 3 concentrations, and Fe D /Fe 2 O 3 ratios display positive, logarithmic trends with age, which suggest early, rapid pedogenic change that slows over time [2].In contrast, in Little River soils, exponential growth functions produce the strongest correlations with age for concentrations of Fe D (Figure 3D) and Fe 2 O 3 .Considering the relatively short interval of the present chronosequence (~100 kyr) in comparison to those of prior studies (1-13 Myr), the exponential increases of Fe D and Fe 2 O 3 concentrations in our soils may correspond to the early, rapid phases of soil development in much longer chronosequences elsewhere in the region.Similar to previous studies, Little River Fe D /Fe 2 O 3 ratios plateau with time, but do so faster (after 70-100 kyr) and at greater values (~0.6-1.0)than in soils along Piedmont-draining streams.This trend is consistent with pedons formed in freely drained sediments high in inherited free Fe, where over time B horizon Fe 2 O 3 becomes dominated by extractable Fe that accumulates in subsoils via translocation.

Geomorphic Mapping Applications
The best indicator of relative age for alluvium in this setting is terrace height above river level [46].However, soil chemical parameters also have strong relationships with time.Intra-surface variation in chemical properties may prevent discrimination of individual terraces in particular instances, but chemical parameters, used together with cross-cutting principles, soil morphology, and quantitative dating, can nonetheless greatly assist investigations in localities with discontinuous or unpaired fluvial surfaces or in areas where tectonics and base level change confound relations between terrace age and height.For geoarchaeological surveys, K 2 O concentration of the fine sand fraction is the most reliable property for differentiating Holocene and latest Pleistocene sediments, which in the southeastern US may contain artifacts buried by sedimentation, from older deposits.Similar age sensitivity has been documented for K 2 O concentrations in Coastal Plain-sourced, riverine eolian dune sands in Georgia [88], suggesting that fine sand fraction K 2 O may be effective for discriminating deposits in this age range along other Coastal Plain-draining rivers.

Conclusions
Subsoil chemical properties are strongly correlated with time and constitute meaningful relative age indicators for Little River terraces.Bases/Al 2 O 3 ratios distinguish floodplain and T1 profiles from T2-T5b soils; Fe D concentrations differentiate floodplain from terrace soils and T4 pedons from those on younger (T1-T3b) and older (T5b) surfaces; and fine sand fraction K 2 O discriminates ≤17 ± 4 ka pedons from older ones.Although intra-terrace variation prevents any single parameter from by itself distinguishing soils by terrace level for each surface in the chronosequence, chemical properties, used in conjunction with soil morphology, geomorphic relationships, and numerical dating, should enable the best possible geomorphic mapping in similar settings in the southeastern US.
With age, pedons develop from Entisols to Ultisols, with exponentially decreasing Bases/Al 2 O 3 ratios, Bases/R 2 O 3 ratios, and K 2 O concentrations.These trends reflect the rapid depletion of bases and enrichment of subsoil alumina respectively produced by primary mineral weathering and the illuviation of phyllosilicate clays during initial argillic horizon formation in the first 13-21 kyr of pedogenesis.Thereafter, base concentrations decline more gradually, owing to the further depletion (and eventual exhaustion) of weatherable minerals in 17 ± 4-75 ± 10 ka profiles, while alumina progressively increases, driven by continuing clay illuviation and increasing gibbsite in >55 ± 15 ka pedons.Increasing subsoil Fe D concentrations over time represent the translocation and transformation of free Fe contained in parent sediments, not the formation of Fe oxides via weathering of detrital Fe-bearing minerals.
Clay fraction minerals change over time from a mixed assemblage of kaolinite, smectite, HIV, quartz, and mica to a mineralogy dominated by kaolinite and gibbsite.Significantly, gibbsite occurs in concentrations similar to (and likely forms from) kaolinite in T5b soils.The chemistry and mineralogy of sampled pedons are compatible with a pedogenetic model in which argillic horizon formation is driven by illuviation of inherited fines, the transformation of secondary phyllosilicates and oxides, and potentially a tandem bioturbation-translocation mechanism, in which fines found within generally sandy deposits are brought to upper horizons and the ground surface by ants and termites, and later illuviated to subsoils.
Pedons on T5b (94 ± 16 ka) have Fe D concentrations similar to, and Fe D /Fe 2 O 3 ratios and gibbsite abundances that are greater than, those of 0.6-1.6Ma profiles on Coastal Plain river terraces that are also sourced from Appalachian provinces.This is likely because sandy, alluvial sediments from the Coastal Plain have fewer weatherable minerals and greater permeability that promote more rapid weathering, gibbsite formation, and illuviation of free Fe relative to the finer, more mineralogically mixed alluvium derived from the Appalachian

Figure 1 .
Figure 1.Study area in the Sandhills province of the upper Coastal Plain of North Carolina (main map) and within the southeastern United States (inset), modified from Figure 1 of Suther and Leigh [33].The Little River and its major tributaries, as well as the location of Figure 2, are also shown.Reprinted from Geomorphology, 351, Suther, B.E., Leigh, D.S., Soil morphology of an alluvial chronosequence from the Little River, North Carolina Coastal Plain, USA (Article 106921), Pages 1-20, Copyright (2019), with permission from Elsevier.

Figure 1 .
Figure 1.Study area in the Sandhills province of the upper Coastal Plain of North Carolina (main map) and within the southeastern United States (inset), modified from Figure 1 of Suther and Leigh [33].The Little River and its major tributaries, as well as the location of Figure 2, are also shown.Reprinted from Geomorphology, 351, Suther, B.E., Leigh, D.S., Soil morphology of an alluvial chronosequence from the Little River, North Carolina Coastal Plain, USA (Article 106921), Pages 1-20, Copyright (2019), with permission from Elsevier.

Figure 2 .
Figure 2. Geomorphic map of the Little River valley, modified from Figure 2 of Suther et al. [29] and Figure 2 of Suther and Leigh [33].The inset depicts topographic cross-section A-A' in the vicinity of the Terrace 1 and 2 (T1 and T2) sampling locations.Map units in the inset are shown at 60 percent transparency so that hillshading is visible for terrace surfaces.Vertical exaggeration of shaded relief in both the main and inset maps is 3X.Reprinted from Geomorphology, 351, Suther, B.E., Leigh, D.S., Soil morphology of an alluvial chronosequence from the Little River, North Carolina Coastal Plain, USA (Article 106921), Pages 1-20, Copyright (2019), with permission from Elsevier.

Figure 2 .
Figure 2. Geomorphic map of the Little River valley, modified from Figure 2 of Suther et al. [29] and Figure 2 of Suther and Leigh [33].The inset depicts topographic cross-section A-A in the vicinity of the Terrace 1 and 2 (T1 and T2) sampling locations.Map units in the inset are shown at 60 percent transparency so that hillshading is visible for terrace surfaces.Vertical exaggeration of shaded relief in both the main and inset maps is 3X.Reprinted from Geomorphology, 351, Suther, B.E., Leigh, D.S., Soil morphology of an alluvial chronosequence from the Little River, North Carolina Coastal Plain, USA (Article 106921), Pages 1-20, Copyright (2019), with permission from Elsevier.

Figure 3 .
Figure 3.Chemical properties of the whole soil fraction (<2 mm) plotted as a function of mean age.(A) Bases to alumina ratio (Bases/Al2O3); (B) Bases to resistant oxides ratio (Bases/R2O3); (C) K2O; (D) Dithionite-extractable Fe (FeD); (E) Oxalate-extractable to dithionite-extractable Fe ratio (FeO/FeD); (F) Dithionite-extractable to total Fe ratio (FeD/Fe2O3).Data were obtained from the most pedogenically well-expressed part of the subsoil and exclude measurements from modern channel alluvium.Results of nonlinear regressions with the function used for each parameter are shown.pvalues apply to the significance level of the F-ratio for each regression, and gray dashed lines depict 95% confidence intervals.Error bars represent two standard deviations of the OSL age estimates.

Figure 3 .
Figure 3.Chemical properties of the whole soil fraction (<2 mm) plotted as a function of mean age.(A) Bases to alumina ratio (Bases/Al 2 O 3 ); (B) Bases to resistant oxides ratio (Bases/R 2 O 3 ); (C) K 2 O; (D) Dithionite-extractable Fe (Fe D ); (E) Oxalate-extractable to dithionite-extractable Fe ratio (Fe O /Fe D ); (F) Dithionite-extractable to total Fe ratio (Fe D /Fe 2 O 3 ).Data were obtained from the most pedogenically well-expressed part of the subsoil and exclude measurements from modern channel alluvium.Results of nonlinear regressions with the function used for each parameter are shown.p-values apply to the significance level of the F-ratio for each regression, and gray dashed lines depict 95% confidence intervals.Error bars represent two standard deviations of the OSL age estimates.

Figure 4 .
Figure 4.Chemical properties of the fine sand fraction (0.125-0.250 mm) plotted as a function of mean age.(A) K2O; (B) Na2O; (C) MgO.Data are from the most pedogenically well-expressed part of the subsoil and exclude measurements from modern channel alluvium.Results of linear regressions are shown for each parameter.p-values apply to the significance level of the F-ratio of regressions, and gray dashed lines depict 95% confidence intervals.Error bars represent two standard deviations of the OSL age estimates.

Figure 4 .
Figure 4.Chemical properties of the fine sand fraction (0.125-0.250 mm) plotted as a function of mean age.(A) K 2 O; (B) Na 2 O; (C) MgO.Data are from the most pedogenically well-expressed part of the subsoil and exclude measurements from modern channel alluvium.Results of linear regressions are shown for each parameter.p-values apply to the significance level of the F-ratio of regressions, and gray dashed lines depict 95% confidence intervals.Error bars represent two standard deviations of the OSL age estimates.

29 Figure 5 .
Figure 5. Mineralogy of the <2 µm fraction from the most well-developed subsoil horizon of at least one representative pedon per landform, as indicated by X-ray diffraction (XRD) patterns of Mgsaturated, glycol-solvated clays.HIV = hydroxy-interlayered vermiculite.

Figure 5 .
Figure 5. Mineralogy of the <2 µm fraction from the most well-developed subsoil horizon of at least one representative pedon per landform, as indicated by X-ray diffraction (XRD) patterns of Mg-saturated, glycol-solvated clays.HIV = hydroxy-interlayered vermiculite.

Table 3 .
Percentages of clay fraction (<2 µm) gibbsite and kaolinite determined by differential scanning calorimetry (DSC) for channel bed sediment and the most well-developed subsoil horizon of at least one representative pedon per landform from the Little River valley.

Table 3 .
Percentages of clay fraction (<2 µm) gibbsite and kaolinite determined by differential scanning calorimetry (DSC) for channel bed sediment and the most well-developed subsoil horizon of at least one representative pedon per landform from the Little River valley.
a T5b samples were analyzed with and without Fe oxides to account for interferences from goethite.bAbbreviations:nd = no data available; na = not applicable.cHereand in Table4, channel bed samples from sites 1, 2, 3, and 5 were composited to provide sufficient clay for analysis.d Measurement is considered a minimum value for this pedon.Soil color indicates the highest free Fe content in the profile occurs in the underlying Btx3 horizon.

Table 4 .
Relative abundance estimates of clay fraction (<2 µm) minerals for channel bed sediment and the most well-expressed part of the most well-developed subsoil horizon of at least one representative pedon per landform along the Little River.