Phases and rates of iron and magnetism changes during paddy soil development on calcareous marine sediment and acid Quaternary red-clay

Dynamic changes in Fe oxides and magnetic properties during natural pedogenesis are well documented, but variations and controls of Fe and magnetism changes during anthropedogenesis of paddy soils strongly affected by human activities remain poorly understood. We investigated temporal changes in different Fe pools and magnetic parameters in soil profiles from two contrasting paddy soil chronosequences developed on calcareous marine sediment and acid Quaternary red clay in Southern China to understand the directions, phases and rates of Fe and magnetism evolution in Anthrosols. Results showed that paddy soil evolution under the influence of artificial submergence and drainage caused changes in soil moisture regimes and redox conditions with both time and depth that controlled Fe transport and redistribution, leading to increasing profile differentiation of Fe oxides, rapid decrease of magnetic parameters, and formation of diagnostic horizons and features, irrespective of the different parent materials. However, the initial parent material characteristics (pH, Fe content and composition, weathering degree and landscape positions) exerted a strong influence on the rates and trajectories of Fe oxides evolution as well as the phases and rates of magnetism changes. This influence diminished with time as prolonged rice cultivation drove paddy soil evolving to common pedogenic features.

Selective extractions showed that the contributions of different Fe pools to total Fe varied markedly between two paddy soil chronosequences developed on different parent materials (Table 1). Silicate bound Fe was the dominant Fe pool in the calcareous paddy soil chronosequence developed on marine sediment, which corresponded to 52~91% of the total Fe and fluctuated with both soil depth and increasing paddy cultivation age (Table 1). In contrast, oxide bound Fe was the major Fe pool in the acid paddy soil chronosequence developed on Quaternary red clay, which represented 49~81% of the total Fe and tended to increase with soil depth and fluctuated with increasing paddy cultivation history ( Table 1). The weakly bound, poorly crystalline Fe pool was the smallest Fe pool in both chronosequences, contributing to 1~16% and 4~18% of the total Fe in the calcareous and acid paddy soil chronosequence, respectively (Table 1). There were strong correlations between the oxide bound Fe and total Fe in both chronosequences, with the correlation coefficient of 0.94 (n = 30, p < 0.01) and 0.92 (n = 22, p < 0.01) for the calcareous and acid paddy soil chronosequence (Fig. 1), respectively, suggesting that variations of Fe contents during paddy soil evolution were mainly caused by the oxide bound Fe.
Chronosequential changes in magnetic parameters. χ m , SIRM, IRM s , and ARM were much higher in the uncultivated soils (P0-MS, P0-RC) than the corresponding paddy soils, and these magnetic parameters decreased rapidly during the initial stage of paddy soil evolution (<100 years) and then declined gradually as paddy soils age (>100 years) in both chronosequences (Fig. 2). The consistent decrease of χ m , SIRM, IRM s , and ARM with paddy cultivation history irrespective of the different parent materials (Fig. 2) resulted in strong correlations between different magnetic parameters in both chronosequences (r > 0.90, p < 0.01, Table 2), and provided an opportunity to use these magnetic parameters for estimating the relative ages of paddy soils. Changes in the distribution of IRM h with paddy cultivation age were, however, much different from that of MS, SIRM, IRM s , and ARM in both chronosequences (Fig. 2). The weighted-mean value of IRM h within 120 cm profile in the SCIENtIFIC REPORtS | (2018) 8:444 | DOI:10.1038/s41598-017-18963-x calcareous paddy soil chronosequence developed on marine sediment decreased gradually from 3.58 × 10 −4 A m 2 kg −1 in the uncultivated pedon (P0-MS) to 3.14 × 10 −4 A m 2 kg −1 after 300 years of rice cultivation (P300-MS), and then declined rapidly to less than 1.30 × 10 −4 A m 2 kg −1 as paddy soils age (>700 years) (Fig. 2). IRM h in the acid paddy soil chronosequence developed on Quaternary red clay exhibited opposite trend in the upper (<50 cm) and lower soil layers (>50 cm), which was respectively lower (<50 cm) and higher (>50 cm) in the uncultivated pedon (P0-RC) than that of paddy soils (Fig. 2). χ d values showed variations in the vertical profiles in both calcareous and acid paddy soil chronosequence, and respectively fluctuated and decreased with paddy cultivation age (Fig. 2). S-ratio varied from 0.65 to 0.95 and from 0.77 to 1.00 in the calcareous and acid paddy soil chronosequence, respectively, which tended to decrease with paddy cultivation age in both chronosequences (Fig. 2). L-ratio varied from 0.29 to 0.56 and from 0.11 to 0.81 in the calcareous and acid paddy soil chronosequence, respectively, and there was no time-dependent changes for L-ratio in both chronosequences (Fig. 2).
The vertical distributions of χ m , SIRM, IRM s , IRM h , and ARM were relatively uniform in the calcareous paddy soil chronosequence developed on marine sediment, contrasting markedly with the larger profile fluctuations in the acid paddy soil chronosequence developed on Quaternary red clay (Fig. 2). The values of χ m , SIRM, IRM s , IRM h , and ARM in the calcareous paddy soil chronosequence were lower than those in the acid paddy soil chronosequence, and the discrepancies declined with paddy cultivation age (Fig. 2). Our results demonstrated the significant influence of parent material on paddy soil magnetism and this influence tended to diminish with time.

Discussion
Variations and controls of different Fe pools during paddy soil evolution. Our study using a chronosequence approach demonstrated that Fe was mobilized and translocated within profile during paddy soil evolution, which was confirmed by the increasing differentiation of Fe concentration and speciation within different selective extractions in both calcareous and acid paddy soil chronosequences (Table 1). Our results are consistent with prior observations that rice cultivation influences Fe differentiation within paddy soils, irrespective of different parent materials 36,39,44,46,47 . The alternating flooding and drying processes during rice cultivation are expected to cause changes in soil pH and Eh 48 , which would result in coupled reduction-oxidation and eluviation-illuviation processes of Fe in paddy soils 46,49 and thus lead to the formation of diagnostic horizons and features (Table S1) characterizing Fe distribution and redistribution as paddy soils age. In addition to the artificial flooding and drainage, seasonal fluctuations of groundwater level could also induce changes in soil redox potential 50,51 that favor Fe reduction and depletion in the lower horizons of paddy soils with shallow groundwater level ( Table 1). The comparison of calcareous and acid paddy soil chronosequences developed on marine sediment and Quaternary red clay showed significant influences of parent materials on the rates and trajectories of Fe evolution (Fig. 3). Total Fe and oxide bound Fe in the calcareous paddy soil chronosequence increased consistently from 47 and 5 kg m −2 , respectively, in the uncultivated soil (P0-MS) to 69 and 23 kg m −2 after 1000 years of rice cultivation (P1000-MS) (Fig. 3). The average increasing rate of total Fe (0.32 kg m −2 yr −1 ) and oxide bound Fe (0.19 kg m −2 yr −1 ) during the first 50 years of rice cultivation was, respectively, 36-and 28-fold greater than that between 50and 1000-yrs time period (Fig. 3). The silicate bound Fe in the calcareous paddy soil chronosequence increased gradually from 31 kg m −2 in the uncultivated soil (P0-MS) to 46 kg m −2 in the 50-yr paddy soil (P50-MS) and then remained relatively constant in the progressively older paddy soils (Fig. 3). The weakly bound Fe in the calcareous paddy soil chronosequence decreased at a rate of 0.12 kg m −2 yr −1 during the initial 50 years of rice cultivation while it showed minimal changes thereafter (Fig. 3). In a sharp contrast, total Fe and oxide bound Fe in the acid paddy soil chronosequence decreased consistently from 73 and 59 kg m −2 , respectively, in the uncultivated soil (P0-RC) to 42 and 31 kg m −2 after 300 years of rice cultivation (P300-RC) (Fig. 3). The average decreasing rate of total Fe (0.04 kg m −2 yr −1 ) and oxide bound Fe (0.20 kg m −2 yr −1 ) during the first 60 years of rice cultivation was, respectively, 0.36-and 2-fold of that between 60-and 300-year time period (Fig. 3). The silicate bound Fe and weakly bound Fe increased initially from 11 and 3 kg m −2 in the uncultivated soil (P0-RC) to 15 and 9 kg m −2 in the 60-year paddy soil (P60-RC), and then declined gradually to 7 and 4 kg m −2 in the 300-year paddy soil (P300-RC) (Fig. 3).
Previous studies have shown that the critical redox potentials for Fe reduction and consequent dissolution are between + 300 mV and + 100 mV at pH 6~7, and −100 mV at pH 8,while at pH 5 appreciable Fe reduction occurs at + 300 mV 52 . The pH value of paddy soils derived from calcareous marine sediment and acid Quaternary red clay ranged from 6.3 to 8.6 and from 4.8 to 6.4, respectively (Table S2). The alkaline environment at the initial stage (0~50 years) of calcareous paddy soil evolution (Table S2) would impede loss of Fe from the profile, and thus an initial period of Fe accumulation was observed in the calcareous paddy soil chronosequence (Fig. 3). As pedogenesis proceeded and CaCO 3 was gradually removed from the profile, the soil pH decreased (Table S2) and Fe accumulated at a lower rate in the later stages of calcareous paddy soil evolution (Fig. 3). In contrast, the acid  (Table S2) together with the relatively high leaching potential in the acid paddy soil chronosequence are expected to promote Fe mobilization and leaching loss after artificial flooding (Fig. 3). This was confirmed by the rapid decrease of Fe and clay content in the acid paddy soil chronosequence in the sloping upland area, which contrasted markedly with the gradual increase of Fe and clay content in the calcareous paddy soil chronosequence in the plain area (Fig. 3, Table S2). In addition to the reductive leaching 38,44,46,47 , particle-facilitated leaching and transport of Fe may also explain the rapid decrease Fe of in the acid paddy soil chronosequence developed on the Phases and rates of magnetism changes during paddy soil evolution. Previous studies have demonstrated that paddy soils exhibit lower magnetic susceptibility than their well-drained counterparts 40,42,44 , however, little is known about the rates of magnetism changes during long-term paddy soil evolution. Our study showed different phases and rates of magnetism changes during paddy soil development on calcareous marine sediment and acid Quaternary red clay (Fig. 4). The vertical distribution of magnetic parameters was uniform in the calcareous paddy soil chronosequence (Fig. 2) and we identified three periods/phases of magnetism changes based on the shifts in magnetic parameters (Fig. 4). The initial phase occurred within half a century and comprised rapid decreases in χ m , SIRM, IRM s , ARM and S-ratio, and a slow decline of IRM h (Fig. 4). The weighted-mean value of χ m , SIRM, IRM s , ARM, S-ratio and IRM h within the 120 cm profile in the 50-yr paddy soil (P50-MS) decreased by 78%, 73%, 80%, 72%, 26% and 12% respectively relative to the uncultivated soil (P0-MS). According to the physical meanings of the different magnetic parameters (Table S3), these changes suggest a rapid destruction of fine-grained maghemite and/or ultrafine magnetite during the initial stage of paddy soil evolution (0~50 years). The second phase lasted several centuries (50~300 years) comprising a relatively   constant IRM h and a slow rate of decline in χ m , SIRM, IRM s , ARM and S-ratio (Fig. 4). The rate of decrease in χ m , SIRM, IRM s , ARM and S-ratio within 120 cm profile at this stage (50~300 years) was less than 5% of that in the initial stage (0~50 years). These results suggest that ferrimagnetic minerals (magnetite and maghemite) decrease successively while the antiferromagnetic minerals (hematite and goethite) remain relatively constant within 50~300 years. In the third phase (700~1000 years), χ m , SIRM, IRM s , ARM and S-ratio showed minimal changes while IRM h declined rapidly (Fig. 4), suggesting significant depletion of antiferromagnetic minerals (hematite and goethite) occurred after 700 years of paddy cultivation. This resulted in the lowest content of magnetic minerals in the oldest paddy soil (Figs 2 and 4). The rapid decline of IRM h after 700 years coincided with the complete removal of CaCO 3 (Table S2). Higher soil pH due to the existence of CaCO 3 has been confirmed to retard the transformation of silicate Fe to secondary Fe oxides as well as the reduction and leaching loss of Fe oxides 4 . We thus hypothesize that the complete removal of CaCO 3 after 700 years of paddy cultivation would promote the reduction and leaching loss of Fe oxides (including the antiferromagnetic minerals) and consequently result in the rapid decrease of IRM h . Further work needs to be done to establish the link between CaCO 3 content and the formation and transformation of magnetic minerals. The acid paddy soil chronosequence showed two phases of magnetism changes, but the changes in the 0-50 cm soil layer were completely different from that in the 50-120 cm soil layer. In the first phase (0~60 years), χ m , SIRM, IRM s , and ARM declined but IRM h increased rapidly in the 0-50 cm soil layer, while all these magnetic parameters declined in the 50-120 cm soil layer (Fig. 4). The weighted-mean value of χ m , SIRM, IRM s , and ARM within 0-50 cm in the 60-yr paddy soil (P60-RC) decreased by 98%, 86%, 94%, and 82% respectively relative to the uncultivated soil (P0-RC). In the second phase (60~300 years), there were minimal changes of different magnetic parameters (χ m , SIRM, IRM s , IRM h and ARM) in the 0-50 cm soil layer, while χ m , SIRM, IRM s , and ARM decreased rapidly after 150 years of paddy cultivation in the 50-150 cm soil layer (Fig. 4). Previous studies have demonstrated the relations between Fe oxides, soil color and soil formation 4 . Hematite-containing soils (usually with associated goethite) have hues between 5YR and 10R, whereas goethite-containing soils with no hematite have hues between 7.5YR and 2.5Y. Soils with lepidocrocite and ferrihydrite covered the hues in-between-range of 5YR~7.5YR with values > 6 for lepidocrocite and <6 for ferrihydrite. Based on soil color (Table S1) and the magnetic properties (Fig. 2), the magnetic minerals were dominated by goethite within 0-50 cm and by hematite within 50-120 cm after 300 years of paddy cultivation.
The evident depletion of magnetism during the anthropedogenesis of paddy soils (Figs 2 and 4) contrasts markedly with the observations of magnetic enhancement during natural pedogenesis [17][18][19][20][21][22][23][24][25] . The increase of magnetic susceptibility during natural soil formation under predominantly oxic weathering conditions has been attributed to the formation of nano-sized magnetite and/or maghaemite, irrespective of the different parent materials 20,21,24 . The periodic submergence and drainage in paddy soils alters this trajectory of magnetism changes by destroying the ferrimagnetic minerals. Previous studies have shown that reducing conditions in paddy soils enhance the dissolution of ferrimagnetic minerals, leading to reduced magnetic properties in paddy soils relative to their well-drained counterparts 40,42,44 . Our results also showed that magnetic parameters (χ m , SIRM, IRM s , ARM and S-ratio) declined rapidly during the early stage of paddy soil development on different parent materials (Fig. 4). The overall magnetic depletion during anthropedogenesis of paddy soils over a millennium time scale (Fig. 4) provides an opportunity to use magnetic susceptibility for estimating the relative age of paddy soils. In addition, our study using a chronosequence approach demonstrated that the parent material and time-span influence the rates of magnetic depletion in different phases of magnetic property development (Fig. 4).
Conclusion remarks on the parent material effects. Paddy soil evolution under the influence of artificial submergence and drainage caused changes in soil moisture regimes and redox conditions with both time and depth that controlled Fe transport and redistribution, leading to increasing profile differentiation of Fe oxides (total Fe, oxide-bound Fe, silicate-bound Fe and weakly-bound Fe), rapid decrease of magnetic parameters (e.g., χ m , SIRM, IRM s , and ARM), and formation of diagnostic horizons (i.e., anthraquic epipedon and hydragric horizon) and features (i.e., anthraquic moisture regime), irrespective of the different parent materials. However, a comparison of the two contrasting paddy soil chronosequences developed on calcareous marine sediment and acid Quaternary red clay demonstrated significant influence of initial parent material characteristics (e.g., pH, Fe content and compositions, weathering degree and landscape positions) on the rates and trajectories of Fe oxides evolution as well as on the phases and rates of magnetism changes, but this influence diminished with time as prolonged rice cultivation drove paddy soil evolving to common pedogenic features. Yet, it remains to be evaluated whether the initial parent material affects the rate of chemical convergence or how long it takes for the parent material effects to be nullified. This is because some of the properties of parent materials persist following long-term paddy soil management over a millennial time scale 53,54 , which is known as the pedological memory and inheritance. Paddy soils may originate from many types of soils in pedological terms showing considerable differences in weathering degree and their initial constitutions 45 . We suggest extensively investigate iron and magnetism changes in paddy soil chronosequences developed on different parent materials and establish the linkage between the expected different evolutional patterns. This will not only provide basic data that are necessary to develop quantitative models of Fe and magnetism changes, but also offers an opportunity to use the established models for predicting the future evolution trends.

Materials and Methods
Study area and sampling sites. The studied paddy soil chronosequences were located respectively on a coastal plain in Cixi County, Zhejiang Province (between 121°2′-121°36′ E and 30°2′-30°19′ N) and on a slope upland in Jinxian County, Jiangxi Province (116°1′-116°32′ E and 28°2′-28°26′ N), in Southern China (Fig. S1). Cixi county has a mean annual air temperature of 16 °C, with yearly extremes ranging from −5 °C to 37 °C, and a mean annual precipitation of 1325 mm of which 73% is concentrated in the rice paddy flooding season (i.e., April SCIENtIFIC REPORtS | (2018) 8:444 | DOI:10.1038/s41598-017-18963-x to October). The coastal plain ranges from 2.6 m to 5.7 m above sea level, and slopes gently towards the northeast (Fig. S1). Soils in the studied area were developed on calcareous marine sediment from the East China Sea, which received large amounts of terrigenous materials from the nearby Qiantang and Yangtze Rivers 53 .
Step-by-step land reclamation of the tidal mudflat through successive dyke building 55 has resulted in a chronosequence with different stages of soil development 53 . Rice (Oryza sativa L.) cultivation in the lower areas where fresh water is readily available for irrigation generally began after five years of dyke building when the salt concentration decreased to agronomically tolerable levels. Sites with 50, 300, 700 and 1000 years of rice cultivation history (i.e., P50-MS, P300-MS, P700-MS and P1000-MS) were identified (Fig. S1) based on the chronology of dyke construction 55 . In addition, an uncultivated mud beach profile (P0-MS) was selected to represent the original soil (parent material, time zero) of the paddy soils (Fig. S1). Parent material homogeneity in the inter-and intra-profiles of the studied chronosequence (P0-MS, P50-MS, P300-MS, P700-MS and P1000-MS) has been evaluated by making use of various soil attribute parameters 56,57 . Details of these profiles (P0-MS, P50-MS, P300-MS, P700-MS and P1000-MS) and the soil chronosequence recognization have been given by Chen et al. 53,57 and Huang et al. 58,59 . Jinxian County has a mean annual air temperature of 17 °C, with yearly extremes from 5 °C to 40 °C, and a mean annual precipitation of 1587 mm, of which 79% is concentrated in the rice paddy flooding season (i.e., April to October). Terraced paddy fields are a common feature in this area. Soils at the bottom of slopes were generally the first to be converted to paddy field; as population pressure increased, lands upslope were progressively brought into paddy cultivation. Thus, these hillside terraces, with increasing cultivation age from the top to the bottom of the slopes, provide soil chronosequences. Soils were derived primarily from acid Quaternary red clays, which were highly weathered, clay rich and phosphorus deficient 60,61 . A paddy soil chronosequence (Fig. S1) consisting of three profiles with approximately 60 (P60-RC), 150 (P150-RC), and 300 years (P300-RC) of paddy cultivation history, and an uncultivated natural soil profile (P0-RC) representing the original soil (parent material, time zero) of the paddy soils were identified. The history of paddy cultivation in the older profile at the bottom of the slope was determined from local historical literature from when the nearby villages were settled. For the newer profile at the top of the slope, information was obtained by questioning the local farmers. The uncultivated soil profile at the highest slope position was treated as the original soil, i.e., time zero with respect to paddy cultivation. The relative ages of the collected paddy soils were confirmed by Han (2012) 60 using the profile development index (PDI) according to Harden (1982) 62 . Details of these profiles (P0-RC, P60-RC, P150-RC, and P300-RC) were given by Han 60 , Han et al. 61 , and Huang et al. 58,59 , who investigated pedogenic changes in basic soil properties, clay minerals and phosphorus fractions. The studied chronosequence was also a toposequence, which complicated the interpretation of the results 49 . In general, time of cultivation had the more significant role, since soil moisture regimes of paddy soils were maintained similarly. In hilly regions, terrace construction greatly reduced water loss and soil erosion and substantially weakened the influence of topography on pedogenesis. Additionally, all four sites were on geomorphically stable topographic positions with low slope gradient (<5°), minimizing the effect of local erosion and deposition. Thus, the role of topography was not separately analyzed for different sampling sites, as we ascribed differential pedogenesis to the difference in time available for pedogenesis.
Soil sampling and description. Within each area of identical paddy cultivation history, one representative profile was chosen for soil sampling based on soil landscape and geomorphological characteristics of that area. All soil samples were collected when the fields were drained after rice harvest. Soil profiles were described and sampled according to genetic horizons following standard field description guidelines 63,64 . The uncultivated soil profiles (P0-MS, P0-RC) in both chronosequences were generally homogeneous throughout its depth, with no visually discernible horizon differentiation (Fig. S2). In contrast, the paddy soil profiles showed complicated patterns with depth due to anthropedogenesis and consisted of an anthrostagnic epipedon, including the cultivated horizon (Ap1) and the plow pan (Ap2), and a hydragric horizon (Br or Bg) (Fig. S2, Table S1). Differences in morphological properties, including soil color, texture and redoximorphic features, were also evident between the relatively younger pedon and the older ones in both chronosequences (Fig. S2, Table S1). The original soils of the two paddy soil chronosequences were defined as Primosols (P0-MS) and Ferrosols (P0-RC), respectively. The paddy soils were defined as Hapi-Stagnic Anthrosols (P50-MS, P300-MS, P60-RC, P150-RC), Fe-leachi-Stagnic Anthrosols (P700-MS, P300-RC), and Fe-accumuli-Stagnic Anthrosol (P1000-MS) by referring to Chinese Soil Taxonomy 65 (Table S1). The detailed field descriptions and classifications of the soil profiles are given in Table S1.
Analysis of basic soil physicochemical properties. After collection, samples of each soil horizon were dried at room temperature and then gently crushed using a wooden pestle and mortar and passed through a 2-mm nylon sieve. Soil bulk density was measured on the 100 cm −3 undisturbed soil cores by drying the cores for 24 h at 105 °C. The particle size distribution was determined by the pipette method and the clay content was defined as the mass percentage of particles <2 μm in diameter for the whole soil. Soil pH was determined at a 1:2.5 soil/solution ratio using distilled water and the carbonate content was determined using a Dietrich Fruhling pressure calcimeter according to the Institute of Soil Science, Chinese Academy of Sciences (1978) 66 . Soil organic carbon (SOC) was measured by the Walkley-Black wet oxidation method 67 using the 149-μm fraction. Total nitrogen (N tot ) was measured by Kjeldahl method 68 and total phosphorus (P tot ) was determined by HClO 4 -HF digestion followed by colorimetric analysis 66 . For total elemental analysis, soil samples (<74 μm) were fused by a mixture of 1:1 lithium metaborate and lithium tetraborate for 30 min in a 1000 °C muffle furnace and then were dissolved in 10% HNO 3 + 1% HF solution. Total elemental concentrations including K, Na, Ca, Mg, Fe, Mn, Al, Si, Ti, and Zr were determined by inductively coupled plasma-optical emission spectrometry. We estimated the precision as 5~10% relative standard deviation based on replicates and standard samples (Geochemical Standard Reference Sample Soil, GSS-3). The measured data are listed in Table S2. The dynamic changes in basic soil physicochemical properties have been reported by Chen et al. 53 , Han et al. 61 , and Huang et al. 58,59 . Briefly, the calcareous paddy soil chronosequence developed on marine sediment over a millennium time scale showed three phases SCIENtIFIC REPORtS | (2018) 8:444 | DOI:10.1038/s41598-017-18963-x of pedogenesis: an initial phase during the first few decades (0~50 years) dominated by rapid desalinization, accumulation of topsoil organic matter and formation of a compacted plow pan (Table S1 and S2); the second phase lasted several centuries (50~700 years) comprising Fe and clay enrichment in the illuvial horizon, and the loss of phosphorus and Mn coincident with the near complete removal of CaCO 3 (Table S2); in the third phase (>700 years), (trans-)formation and redistribution of metal oxides were accompanied by clearly visible hydromorphic patterns in paddy subsoils (Table S1, Fig. S2). The acid paddy soil chronosequence developed on Quaternary red clay over a centurial time scale showed rapid accumulation of SOC and increase of pH in surface paddy soils, loss of clay and Fe oxides with prolonged cultivation history (Table S1, Fig. S2), and shifts in phosphorus abundance and speciation as well as clay mineral compositions with both time and depth 58,61 . Extraction of Fe oxides and measurement of magnetic properties. Bulk soil samples were subjected to reducing agents with increasing strength to selectively extract major pools of Fe: (1) the Tamm's extraction 69 ; and (2) the citrate-bicarbonate-dithionite (CBD) extraction 70 . The Tamm's extraction is a mixture of oxalic acid and ammonium oxalate, which was performed by shaking the sample-solution mixture in the dark over 4 h at 20 °C with a solid/liquid ratio of 1.25 g/50 ml. The Tamm's method targets the extraction of weakly bound, short-range-ordered (SRO) and organic bound Fe 71 . For the extraction by CBD, soil samples were exposed to the reactant mixture at 80 °C for 30 min with a solid/liquid ratio of 0.5 g/25 ml. The CBD method extracts Fe in oxides and hydroxides (e.g., hematite, goethite, lepidocrocite) of all crystallinities-SRO and bulk crystalline 70 . In addition to the partial extractions, total Fe was dissolved in a HF-HClO 4 mixture after calcination of soil organic matter at 450 °C. Fe concentrations in the extracted solutions were analyzed using an Inductively Coupled Plasma-Atomic Emission Spectrometer (ICP-AES, LAS Arras). We calculated the oxide bound Fe concentration by subtracting oxalate-extractable Fe from the CBD-extractable Fe and the silicate bound Fe was calculated by subtracting CBD-extractable Fe from the total Fe concentration. Fe mass (kg m −2 ) in the soil pedon (0-120 cm) was calculated by multiplying Fe concentrations by bulk density and thickness of soil horizons using the following equation: where C Fe , D i , and E i is, respectively, the Fe concentration (g kg −1 ), bulk density (g cm −3 ) and depth (cm) in the i horizon. Magnetic parameters were measured and calculated according to Evans and Heller (2003) 21 and Lu (2003) 42 . Briefly, magnetic susceptibility (χ m ) was measured with a Bartington MS2 meter (Bartlington Instruments Ltd., Oxford, UK) at both low (0.47 kHz, χ lf ) and high frequencies (4.7 kHz, χ hf ). Frequency-dependent magnetic susceptibility (χ d ) was calculated as [(χ lf − χ hf )/χ lf ] × 100%. Isothermal remanent magnetization (IRM) was produced in progressively increasing magnetic fields (i.e., 20 mT, 30 mT, 50 mT, 100 mT, 300 mT, 1000 mT) and then was determined under reverse magnetic fields (−300 mT, −100 mT, −20 mT) using a Molspin pulse magnetizer (Molspin Ltd., Newcastle on Tyne, UK). The induced remanence after imposing each magnetic field was measured in a Molspin spinner magnetometer. The anhysteretic remanent magnetization (ARM) was acquired in a steady field of 0.04 mT imposed on an AC field with decreasing amplitude from a maximum of 100 mT to 0 mT 33 . The IRM at 1000 mT was defined as saturation isothermal remanent magnetization (SIRM) and was used to calculate the soft isothermal remanent magnetization (IRM s ), hard isothermal remanent magnetization (IRM h ), S-ratio and L-ratio using the following formulas: Detailed information and interpretations of the operationally defined Fe pools and measured magnetic parameters are given in Table S3.
Data availability. All data generated or analyzed during this study are included in the article and attached in the Supplementary Information files.