Silicon pools in human impacted soils of temperate zones

Abstract

Besides well‐known effects of climate and parent material on silicate weathering the role of land use change as a driver in the global silicon cycle is not well known. Changes in vegetation cover have altered reservoirs of silicon and carbon in plants and soils. This has potential consequences for plant‐Si availability, agricultural yields, and coastal eutrophication, as Si is a beneficial element for many crop plants and an essential nutrient for diatom growth. We here examined the role of sustained and intensive land use and human disturbance on silicon (Si) pool distribution in soils with similar climatological and bulk mineralogical characteristics. We show that land use impacts both biogenic and nonbiogenic Si pools. While biogenic Si strongly decreases along the land use change gradient (from forest to croplands), pedogenic silica fractions (e.g. pedogenic clays) increase in topsoils with a long duration of cultivation and soil disturbance. Our results suggest that nonbiogenic Si pools might compensate for the loss of reactive biogenic silicon in temperate zones.

1 Introduction

The long‐term global biogeochemical Si cycle encompasses weathering and mobilization of dissolved silicate (DSi), its transport by rivers from land to ocean, and its eventual burial as diatom shells in ocean sediments [Tréguer and De La Rocha , 2013]. Both on land (i.e., by weathering of silicate rocks) and in the ocean (i.e., by coupled C:Si diatom pumping) the global Si cycle is tightly linked with buffering of atmospheric CO₂ concentrations over geologic time scales [Street‐Perrott and Barker , 2008; Goudie and Viles , 2012] and has thus received substantial scientific attention. The net result of continental silica (Si) cycling is that a secondary Si soil pool develops that is much larger than the annual release of Si from primary silicate minerals via weathering. Because of this, terrestrial ecosystems can buffer dissolved Si transport from soils to rivers and coastal zones [Derry et al. , 2005; Struyf and Conley , 2012]. Continental vegetation stores large quantities of biogenic Si (BSi) (~84 × 10¹² mol Si yr⁻¹ [Carey and Fulweier , 2012]), which can provide a rapidly available Si pool relative to mineral silicates and control the dissolved Si concentration of soil and/or river water in forests [Watteau and Villemin , 2001; Derry et al. , 2005; Gerard et al. , 2008; Sommer et al. , 2013] and grasslands [Blecker et al. , 2006; White et al. , 2012]. Storage and cycling of Si in ecosystem soils are functions of several parameters, including plant species and plant specific uptake mechanisms [Bartoli and Wilding , 1980; Kelly et al. , 1998], chemical weathering rate [Alexandre et al ., 1997], mineral composition of the soil [Henriet et al. , 2008], and climate conditions [Blecker et al. , 2006]. In addition, humans have significantly perturbed the natural Si cycle, by disrupting the balance between silicate weathering, internal (re)cycling, and release of Si from terrestrial ecosystems by land use changes, agricultural intensification, soil management, and climate change [Conley et al. , 2008; Struyf et al. , 2010; Clymans et al. , 2011; Guntzer et al. , 2012; Sommer et al. , 2013]. Agricultural harvest of Si‐accumulator crops or pasture grasses is responsible for a constant depletion of the Si reservoir in temperate zones [Vandevenne et al. , 2012], which can lead to a decreased bioavailability of Si [Guntzer et al. , 2012]. We currently lack knowledge of the full consequences of such alteration at a regional and global scale. While biogenic Si fractions have been put forward as determinant factors in controlling Si release rate and availability for plants, little is known about the role of other nonbiogenic Si fractions and how they are affected by cultivation. We here studied the impact of intensive cultivation and crop harvest on soil Si pools, Si availability, and potential soil mobilization along a temperate land use change (LUC) gradient. Multiple soil extractions were performed that allowed us to distinguish between and quantify biogenic from nonbiogenic Si pools and also assess Si availability. Our LUC gradient is represented by sites with different yet long duration (>240 years) of vegetation cover and land use management (i.e., forests, pastures, and croplands with row crops). All sites are situated in the Belgian Loam Belt [Van Ranst et al. , 1982] and are comparable in terms of climatological and bulk mineralogical properties of the soil. This study offers an opportunity to isolate the impact of land use and long‐term soil disturbance and can be seen as a reference for soil Si stocks in cultivated landscapes in temperate (European) weathering zones.

2 Material and Methods

2.1 Sampling and Characteristics of the Land Use Change Gradient

The Belgium Loess Belt stretches from west to east across Central Belgium and is characterized by a drastic alteration of the landscape from dominant forest to highly cultivated landscapes during the last centuries. Loess thickness can be up to 20 m and mainly consists of quartz and feldspars and at some places mica and carbonates [Van Ranst et al. , 1982]. Six study sites were selected based on historical and contemporary land cover history. Two deciduous forest sites, forest 1 (Meerdaal: 50°48′1.92″N, 4°40′9.18″E) and forest 2 (Ronquières: 50°37′31.1196″N, 4°12′3.6678″E), are dominated (>80%) by oak (Quercus robur ) and beech (Fagus sylvatica ). Pasture 1 (Blégny: 50°40′22.3386″N, 5°45′52.2576″E) and pasture 2 (Herve: 50°38′52.93″N; 5°45′20.80″E) are used as permanent pastures (dominated by Lolium perenne ) with mowing activities in summer alternated with occasional grazing by cows. Cropland 1 (Ganspoel: 50°48′29.1882″N, 4°35′10.2762″E) and cropland 2 (Velm: 50°45′35.247″N, 5°7′25.2012″E) are cultivated with row crops. The former has only been managed as cropland since 1980 (with a monoculture of maize (Zea mays L. ) for the last 14 years); before 1980 periods of pasture and forest cover alternated [Van Rompaey et al. , 2002]. Cropland 2 has been managed as a rotation system with maize, wheat (Triticum aestivum ) and fodder beets (Beta vulgaris ). Based on cartographic evidence (e.g. Ferraris maps (1777)) we can rely on a similar land use management and vegetation cover 240 years ago in all sites. Climate in all sites is comparable, without a dry season and with a warm winter (Köpper‐Geiger climate map of Europe). Mean annual precipitation in all sites is estimated at ~820 mm with mean January and July temperature approximately 3 and 18°C, respectively [Vandevenne et al. , 2015].

Soil cores (2 m) were taken at the bottom of the topographical slope at all locations (Table 1). Cores were sliced according to horizons determined by color and texture in the field. After freeze drying of the soil, grain size, pH (H₂O), cation exchange capacity, base saturation, organic carbon (Table 1), and bulk and clay mineralogical content of soils (Tables S1 and S2 in the supporting information) were determined. Bulk mineralogical properties were very similar along the gradient, with quartz, feldspar, plagioclase, 2:1 clays, and kaolinite occurring as the most dominant minerals (Table S1 for detail). Clay mineralogy predominantly consisted of 2:1 clays (e.g. smectite and illite groups) and was similar within sites (Table S2 for detail).

Table 1. Description of the Depth Profile With Related Horizon Division Based on the Percentage of Clay, Silt and Sand, Texture, Soil pH (H₂O), CEC (meq/100 g), and Organic Carbon (g kg⁻¹) in All Sites of This Studya^a Forest 1 (Meerdaal), forest 2 (Ronquières), pasture 1 (Blégny), pasture 2 (Herve), cropland 1 (Ganspoel), and cropland 2 (Velm).

	Horizon	Depth (cm)	Clay (%)	Silt (%)	Sand (%)	Texture	pH	CEC (meq/100g)	Organic Carbon (g kg^−1)
FOREST 1	Oi	0–5	2	43	55	Sandy loam	3.8	29.0	97.1
	A	5–20	3	55	42	Silt loam	4.1	10.6	46.3
	E	20–50	5	61	34	Silt loam	4.5	3.2	11.7
	Bw	50–120	7	64	29	Silt loam	5.1	5.6	3.1
	Bw /C	120–140	7	57	36	Silt loam	5.8	7.6	‐
	C	140–145	3	19	78	Loamy sand	5.0	3.8	‐
FOREST 2	Oi	0–5	2	41	57	Sandy loam	‐	19.7	99.7
	Oe	5–10	5	71	25	Silt loam	‐	9.5	32.8
	AB	10–20	5	72	23	Silt loam	5.3	5.1	‐
	B1	20–60	7	72	21	Silt loam	5.6	6.8	6.4
	Bg	60–185	7	66	27	Silt loam	6.1	11.5	2.8
PASTURE 1	A	0–30	4	66	31	Silt loam	5.9	10.6	38.6
	B	30–60	5	74	21	Silt loam	6.7	5.5	5.2
	Bg	60–172	7	73	20	Silt loam	7.5	25.4	‐
PASTURE 2	A1	0– 2	4	61	35	Silt loam	‐	‐	‐
	A2	2–32	5	70	25	Silt loam	6.9	9.9	51.3
	E	32–52	8	71	21	Silt loam	7.0	5.4	8.6
	B	52–198	10	70	20	Silt loam	7.0	23.8	3.2
CROPLAND 1	Ap	0–35	5	57	24	Silt loam	7.2	9.8	12.8
	Bw	35–75	7	72	21	Silt loam	7.5	8.5	3.9
	B	75–157	8	73	19	Silt loam	7.2	10.7	‐
CROPLAND 2	Ap	0–50	7	75	18	Silt loam	7.8	15.0	14.9
	Bw	50–130	6	67	27	Silt loam	8.1	10.8	5.6
	BC	130–192	9	74	17	Silt loam	7.6	12.8	‐

• ^a Forest 1 (Meerdaal), forest 2 (Ronquières), pasture 1 (Blégny), pasture 2 (Herve), cropland 1 (Ganspoel), and cropland 2 (Velm).

2.2 Chemical Extractions

2.2.1 Continuous Extraction of Si and Al (Alkaline‐Extractable Si Fraction)

A 100 mg of soil sample (n  = 51) was constantly mixed with 175 mL of 0.5 M NaOH at 85°C in a stainless vessel for 30 min. A tube connecting to the vessel fed a continuous flow analyzer Skalar (Skalar®, Netherlands), where the spectrophotometric molybdate blue method continuously (every 15 s) measured monosilicic acid and the lumogallion fluorescence method measured Al [Koning et al. , 2002; Barão et al. , 2015]. Early dissolution of Si and Al is nonlinear, mainly because of dissolution of fractions with a high solubility, such as biogenic Si [DeMaster , 1981] or other reactive nonbiogenic Si phases, defined as the alkaline‐extractable Si fraction (AlkExSi) [Barão et al., 2015]. The number of AlkExSi fractions present in each sample is calculated with a mathematical model that fits Si and Al data measured during the extraction. A complete description of this method can be found in Barão et al. , [2014, 2015]. Replicates of each sample were not measured since the model has shown high reproducibility (<10%) in the past, and the parameters calculated are not significantly affected by small experimental variations (Saccone et al. , 2007).

Each AlkExSi fraction present in the sample is characterized by a Si:Al ratio which provide insight about the fractions' origin. If Si:Al is higher than 5 the fraction is attributed to a biogenic (BSi) source, as phytoliths and diatoms contain limited amounts of Al [Carnelli et al. , 2002; Koning et al. , 2002]. Fractions with Si:Al ratios between 1 and 5 are related to 1:1 or 2:1 clay minerals of the aluminosilicate families [Velde and Meunier , 2008]. This AlkExSi fraction reflects the presence of a mineral that initially dissolves faster in NaOH (nonlinear dissolution), and it has been shown that clay minerals can initially dissolve as fast as diatoms in 0.5 M NaOH [Koning et al. , 2002]. It differs from the clay mineral pool (Table S2) that accounts for all clay minerals present in the soil, as only the mostly reactive clays in these alkaline environments are quantified with the continuous analysis in NaOH. Here we define this AlkExSi fraction as the Pedogenic Clay Si (P_ClaySi), referring to clays formed or transformed during soil processes that present initial nonlinear dissolution in NaOH. Si:Al ratios can drop to values below 1: higher Al content suggests the presence of a more weathered product resulting from pedogenic processes, here defined as Pedogenic Si (PSi) [Parfitt , 2009; Barão et al. , 2014].

2.2.2 Extraction of Readily Soluble Si (Si_CaCl2)

Calcium chloride (CaCl₂) is a weak extracting agent that extracts the readily soluble Si fraction [Berthelsen et al. , 2001]. It can be used as an estimate of DSi availability for plants [Sauer et al. , 2006]. Dried soil samples (2.0 g), representative for the soil profile of each site, were shaken thoroughly for 16 h with 20 mL 0.01 M CaCl₂ solution in a tube (35 mL) at 20°C [Clymans et al. , 2011]. Extracted samples were then centrifuged at 4000 rpm for 30 min, and the supernatant was filtered over a filter with 0.45 µm pore size (Chromafil® A45‐25). Extracted Si (Si_CaCl2) was analyzed by inductively coupled plasma–atomic emission spectroscopy (ICP‐AES).

2.2.3 Oxalate Extractable Si (Si_ox)

Soil samples were analyzed for silica adsorbed to Fe and Al (hydr)oxides and for silica bound to poorly crystalline phases like allophane and imogolite using ammonium oxalate extraction technique [Saccone et al. , 2007; Cornelis et al. , 2011a]. One gram of dried soil was mixed with 50 mL of oxalate solution (17.56 g (COOH)₂ 2H₂O + 28.4 g (COONH₄)₂ H₂O L⁻¹) in a dark incubator for 1 h. The concentrations of Si, Fe, and Al in the oxalate extract were measured by ICP‐AES.

2.3 Physical Separation of Phytoliths From Soils

Physical separation and chemical extractions were performed on two subsamples from the same homogenized soil sample using a gravimetric separation method (adapted from Kelly [1990] and Piperno [2006]). Oven‐dried (105°C), sieved (<2 mm) soil was treated with 10% HCl to remove carbonates. Soils were then put in a hot water bath (85°C) with deionized water (DI) and a few drops of 30% H₂O₂ in order to remove organic matter. After rinsing with DI, samples were shaken overnight with 5% sodium hexametaphosphate, followed by wet sieving through 53 µm mesh size to separate the sand from the silt and clay fraction. Silt was further separated from clay by centrifugation and gravity sedimentation, and both fractions were rinsed with DI, dried overnight, and stored in plastic bottles. Phytoliths were obtained by placing a subsample (2.5 g) of either sand or silt in a 50 mL centrifuge tube along with a heavy liquid sodium polytungstate Na₆H₂W₁₂O₄₀·H₂O solution (e.g., 2.30 g cm⁻³). After centrifugation, floating phytoliths were decanted in a clean centrifuge tube. Additional stirring and centrifugation steps were repeated until negligible phytolith yield was obtained. The centrifuge tube was then rinsed with DI, and small amounts of clays, organic matter, and carbonates were removed by repeating the same cleaning steps. Sample solution was finally poured over a 0.2 µm polycarbonate filter (Whatmann), and phytoliths were dried on the filter overnight and stored in plastic petri dishes. A subsample of the dried phytoliths was prepared for microscopic examination on field emission scanning electron microscope (JEOL, JSM‐6500F). Subsamples were placed on Al‐stubs, fixed by adhesive tape, and coated with 20 nm gold particles. Samples were screened for contamination (esp. clay particles), and semiquantitative energy‐dispersive X‐ray analysis was performed to check elemental composition of certain fragments. Where possible phytoliths were described by their shape according to International Code for Phytolith Nomenclature [Madella et al. , 2005].

3 Results

3.1 Si Fractions From Chemical Extraction

3.1.1 AlkExSi in Soils

In forests, AlkExSi fractions with Si:Al > 5 dominated O and A horizons and run into the E (forest 1) and B horizon (forest 2) (Figure 1, top row). AlkExSi fractions with Si:Al < 5 peaked around 50 cm (5–6 mg g⁻¹) in the B and C horizons of forest soils. Fractions with Si:Al < 1 were predominantly localized in the E and B_w horizon (forest 1) or AB and B1 horizon (forest 2), while fractions with Si:Al between 1 and 5 were generally concentrated in the mineral horizons (B and C). In pastures, AlkExSi concentrations decreased through the A horizon until 50 cm (Figure 1, middle row). The majority of the AlkExSi fractions in this organic horizon were biogenic (Si:Al > 5). In the B horizon of pasture soils, AlkExSi fractions with Si:Al between 1 and 5 were observed followed by a strong increase in AlkExSi concentration (up to 60–80 mg g⁻¹) with Si:Al > 5. A small amount of AlkExSi with Si:Al < 1 was localized at the bottom of the A and/or E horizons and the top of the B horizon. In cropland soils AlkExSi concentrations were between 4 and 8 mg g⁻¹ throughout the profile (Figure 1, bottom row). Cropland 1 showed constant AlkExSi concentrations in the A horizon (~4 mg g⁻¹), with an increase to 8 mg g⁻¹ in B horizon. The top 20 cm of the soil contained AlkExSi fractions with Si:Al > 5 and Si:Al < 1. Only fractions with 1 > Si:Al < 5 were observed below this depth. Cropland 2 did not show clear AlkExSi variation along the profile. Fractions with Si:Al > 5 were almost absent and instead pedogenic clays (Si:Al ratios between 1 and 5) dominated. A detailed overview of the Si:Al ratios, k parameters and other parameters derived from the continuous extraction can be found in Table S3.

Figure 1

AlkExSi concentration (mg g⁻¹) for the fractions identified in each sample along the soil profile of the forest 1 (Meerdaal), forest 2 (Ronquières), pasture 1 (Blégny), pasture 2 (Herve), cropland 1 (Ganspoel), and cropland 2 (Velm) using the continuous analysis of Si and Al in 0.5 M NaOH. The solid line represents the total AlkExSi concentration. Horizon transitions are marked using alternating grey and white bands (for details see Table 1). The black, grey, and white bars represent AlkExSi fractions with a Si:Al higher than 5, between 1 and 5 and lower than 1, respectively. Note that the total number of AlkExSi fractions (or bars) shown for each depth is a direct result from the model, while the attributed colors (grey, black, white) represent the classification adopted in this study. Raw data output of the model can be found in Table S3.

3.1.2 Si_CaCl2 and Si_oxalate Distribution in the Soil Profile

Forests showed distinct patterns of readily soluble silica (Si_CaCl2) and Si extracted with ammonium oxalate (Si_oxalate) (Figure 2). In the first 50 cm (horizons A, E, and B1) low Si_CaCl2 (<0.02 mg g⁻¹) and Si_oxalate (<0.25 mg g⁻¹) were present compared to deeper B and C horizons (0.02–0.05 mg g⁻¹ Si_CaCl2 and 0.2–0.5 mg g⁻¹). Pastures showed very little Si_CaCl2 in the A horizon (<0.02 mg g⁻¹), although an increase in depth was observed from 100 cm onward in pasture 1 (0.02–0.04 mg g⁻¹) and from 50 cm onward in pasture 2 (0.02–0.08 mg g⁻¹). Si_oxalate concentrations were rather low for the entire pasture soil profile (<0.25 mg g⁻¹). Croplands displayed distinct patterns in Si_CaCl2: while in the cropland 2 the top organic layers showed the highest concentration (0.03–0.06 mg g⁻¹) followed by a decrease with depth; cropland 1 did not show any clear variation along the soil profile (0.02–0.03 mg g⁻¹). For both croplands the Si_oxalate concentration remained constant with depth (~0.25 mg g⁻¹).

Figure 2

Concentration (mg g⁻¹) of (top row) readily soluble Si (Si_CaCl2) and (bottom row) silica extracted with ammonium oxalate (Si_oxalate) distribution along the soil profile of (left column) forests (filled diamonds = forest 1; open diamonds = forest 2), (middle column) pastures (filled triangles = pasture 1; open triangles = pasture 2), and (right column) croplands (filled circles = cropland 1; open circles = cropland 2).

3.2 Physical Separation of Phytoliths From Soils

Phytoliths could only be detected in the silt fraction of the top layers of all sites. Negligible phytoliths remained after physical separation of the sand fractions (data not shown). Indeed, no phytoliths (Figures 3a and 3b) could be visualized below the A horizon in pastures. Instead, another nonphytolithic material was detected, which consisted almost entirely of Si with negligible amounts of Al (semiquantitative energy dispersive X‐ray spectroscopy analysis). Further mineralogical analysis of the deeper soil layers in the pastures confirmed these fractions to be opal‐CT (Figure S1 in the supporting information), which is an amorphous material with a higher degree of crystal order in comparison to opal‐A. Opal‐CT can be formed by precipitation of silicic acid when the soil solution reaches concentrations above equilibrium [Chadwick et al. , 1987; Kastner et al. , 1977] or as an intermediate product resulting from the dissolution of opal‐A [Kastner et al. , 1977]. Spherular nonphytolithic Si specimens were mainly observed in forest soils (Figures 3c–3e) and sporadically in topsoils of pasture and cropland 1 (Figure 3j). These fractions could be pedogenic (opal‐A spherules) or siliceous resting stages (stomatocysts) from Chrysophycean algae [Duff et al. , 1994]. Some clay contamination (i.e., small flattened structures indicated with red arrows) was present in some samples (Figures 3m–3o). Phytoliths in forest topsoils were predominantly rod‐shaped (Figures 3h and 3f) or trapezoid‐shaped (Figures 3e and 3g). In some cases irregular polyhydral shapes were likely observed too (Figures 3e and 3f). Pasture topsoils principally consisted of typical grass trapezoid short cells (bilobate, polylobate, and sinuous based; Figure 3i) and bulliform cells. Cropland soils mainly showed trapezoid sinuate/polylobate (Figures 3k, 3m, and 3n) and bulliform morphotypes (Figure 3n). The scanning electron microscope (SEM) pictures of topsoils revealed a higher variability of phytolith morphologies in cropland 1, including rod‐shaped elongated shapes (Figures 3j–3m), as compared with cropland 2.

Figure 3

Scanning electron microscope (SEM) pictures of the silt fraction of soils in this study. Blue arrows point at spherular specimens (pedogenic or biogenic origin; opal‐A or stomatocysts), green arrows show phytoliths, and red arrows show clay contamination. (a) opal‐CT in B horizon pasture 2, 200×; (b) opal‐CT in B horizon pasture 1, 16,000×; (c) spherical Si specimen with smooth surface (probably a stomatocyst), A horizon forest 1, 10,000×; (d) spherical Si specimen in Oi horizon of forest 2, 2200×; (e) spherical smooth spherule (blue arrow) and phytoliths in A horizon forest 2, 1300×, (f) phytolith in A horizon forest 1, 1500×; (g) phytolith in A horizon forest 1, 14,000×; (h) phytolith in A horizon forest 1, 1900×; (i) grass phytoliths in A horizon pasture 1, 1800×; (j) phytolith and spherical specimen in A_p horizon cropland 1, 1800×; (k) phytoliths A_p horizon cropland 1, 2700×; (l) phytoliths A_p horizon cropland 1, 300×; (m) phytolith and clay contamination in A_p horizon cropland 2, 5500×; (n) phytolith in A_p horizon cropland 2, 3700×; (o) phytolith in A_p horizon in cropland 2, 2700×.

4 Discussion

4.1 Effect of Cultivation on Secondary Si Pools in Soils

Our data show that human use of land impacts the concentration and the distribution of secondary Si fractions in soils. This becomes especially clear when the contribution of every Si fraction to the total AlkExSi pool (Figure 4) is compared between land uses (calculated for BSi (Si:Al > 5), PSi (Si:Al < 1), and Pedogenic Si Clays (Si:Al between 1 and 5). We focused on top 50 cm of soil as vegetation, soil (water) Si, and hydraulic properties are directly impacted by root activity and land use management. A linear interpolation was used between each two measurements of AlkExSi concentration (1) instead of using an average value per horizon. Although earlier results showed that both approaches are identical [Barão et al ., 2014], linear interpolation is likely more adequate here as concentrations may vary significantly along horizons.

(1)

Figure 4

Total AlkExSi stocks (t ha⁻¹) and Si_CaCl2 stocks (t ha⁻¹) along the land use change gradient and the distribution of BSi (Si:Al > 5), Pedogenic Si Clays (1 < Si:Al < 5), and PSi (1 < Si:Al) for the top 50 cm. Numbers inside bars are absolute values of the different fractions of Si in every 50 cm.

AlkExSi is the total pool (mg g⁻¹), ρ is the bulk density (kg m⁻³), h is the depth of each measurement, n is the total number of measurements per site, and i is the incremental step in the calculation.

“Biogenic Si (Si:Al > 5)” dominates in forests and pastures (50–60% of total AlkExSi) and drops in cropland 1 (30%) and cropland 2 (~10%) (Figure 4). Persistent harvest and ploughing activities in cropland 2 severely depleted soil BSi [Guntzer et al. , 2012; Vandevenne et al. , 2012] as compared to cropland 1, which still shows relicts of former pasture and forest biogenic Si pools (i.e., BSi makes up 55% of the AlkExSi stock in the organic horizon). Soil BSi depletion in our croplands is likely amplified by denudation losses of top layer BSi through erosion and runoff, a phenomenon which is well studied in these sloping catchments [Clymans et al. , 2015].

An increase of “pedogenic clays” (1 < Si:Al < 5; P_claysSi) is observed along the LUC gradient, from ~20% in forest topsoils to over 80% in cropland 2 (Figure 4). This suggests that some of the clays in our soils (especially in the croplands) are either (i) intrinsically more reactive in NaOH or (ii) are a result of processes such as neoformation/transformation [Lucas et al. , 1993; Cornelis et al. , 2011b] and their degree of crystallinity is not completely reached. We suggest that the high proportion of P_claysSi in the AlkExSi fraction of cultivated topsoils might be related to the fact that part of the clays are impacted and transformed by persistent intensive agriculture and successive rounds of fertilizer use, plant Si uptake, and annual harvesting. Cropping method and fertilizer type have been shown to alter clay composition in the root interaction zone [Velde and Meunier , 2008; Velde and Barré , 2009; Cornu et al. , 2012; Velde and Peck , 2002], and the K input resulting from chemical fertilizers and/or manure corroborates the high illite values found in our cropland soils (19–26%; Table S2). Moreover, formation, transformation, and dissolution of clays can occur on short time scales (i.e., months to years), in particular in chemically reactive soil‐microenvironments impacted by plant roots and/or earthworms [Calvaruso et al. , 2009; Turpault et al. , 2008].

The second “pedogenic Si pool (Si:Al < 1)” observed in this study is much smaller than the other Si pools and does not show strong variation along the LUC gradient (Figure 4). Due to the low Si_oxalate concentrations (Figure 2) compared to other studies [Saccone et al. , 2007; Cornelis et al. , 2010] we exclude short‐order range minerals (allophane/imoglite [Sommer et al. , [2006]) and adsorbed Si on Al oxi/hydroxides as possible candidates for this Si pool. Instead, the presence of Si:Al < 1 can be (1) related to the weathering of feldspar and plagioclase, which includes the formation of a protective Al‐rich layer [Wollast , 1967; Busenberg and Clemency , 1976] or (2) can be due to hydroxy‐aluminosilicate formation, precursors for imogolite [White et al. , 2012]. Although detected in small quantities in our soils, these fractions could play an important role in the Si and Al cycle and should get more attention in the future, given their higher reactivity (i.e., in NaOH, k parameter; Table S3) and thus potential to contribute to short‐term Si cycling. Their presence/absence and abundance are likely controlled by pH [Doucet et al. , 2001] and thus the availability of Al which forms complexes with organic matter [Parfitt , 2009]. This might explain why PSi slightly increases in deeper layers of forest soils (Figure 1) as free Al becomes available due to the dissolution of organic matter complexed with Al in topsoils.

4.2 Plant‐Available Si in Soil Water

Despite long‐term depletion of soil BSi in cropland 2, we do not see a parallel drop in the CaCl₂‐ectractable Si pool (Figure 4) as is observed for monocultures of rice and sugarcane [Berthelsen et al. , 2001]. Croplands in this study show Si_CaCl2 pools (18–52 mg kg⁻¹) far above the critical value of 10 mg Si kg⁻¹ for sugarcane production [Haysom and Chapman , 1975]. Rotation of Si‐accumulating crops (maize and wheat) with non‐Si accumulator crops (beets and potatoes) could replenish plant‐available Si levels [Savant et al. , 1997] yet also high amounts of clay (i.e., 85% of AlkExSi stock in A_p horizon) could compensate for the absence of BSi in the readily soluble Si part of cropland 2. Differences in soil pH along the LUC gradient (Table 1 and Figure 5) are interfering with Si solubility and dissolution rates in a complex way [Ronchi et al. , 2013; Keller et al. , 2012; Haynes , 2014] and may induce a preferential dissolution of some Si fractions in situ. Indeed, our results show a positive correlation between clays and Si_CaCl2 while BSi and Si_CaCl2 are negatively correlated, suggesting that these non‐BSi pools are apparently contributing to the Si availability in the soil water (Figure 5).

Figure 5

Relation between (1) Si_CaCl2 and BSi, (2) Si_CaCl2 and PSi and pedogenic clays, and (3) Si_CaCl2 and pH (in H₂O). BSi and PSi/pedogenic clays are derived from continuous extraction of Si and Al in 0.5 M NaOH. The Spearman's correlation coefficient “ρ ” is indicated in each graph and only refers to the correlation of positive values for BSi and PSi/pedogenic clays. Different points represent different soil samples in this study (see Table 1): forests (black circles), pastures (grey circles), and croplands (white circles).

4.3 Outlook: Effect of Human Impacted Soils for Si Cycling in Temperate Zones

4.3.1 The Importance of Biogenic and Nonbiogenic Si Pools for Si Availability

Conceptually, three different stages could be identified during the conversion from pristine forest to long‐term cultivation of soils with row crops and Si harvest (three black circles in Figure 6). In the “reference” situation the temperate landscape is dominated by deciduous climax forests. These systems show tight recycling of Si in the soil‐plant system [Sommer et al. , 2013; Cornelis et al. , 2011b], and dissolved Si in soil water can be taken up by plants and/or incorporated in clays [Cornelis et al. , 2011a, 2011b, 2014; Vandevenne et al. , 2015]—“stage 1.” Biogenic Si (in litter) does not form a long‐term pool in our soils, and this is in line with other studies from temperate zones where a sharp drop in BSi and/or phytoliths is observed at the transition between organic and mineral horizons [Barão et al. , 2014; Fishkis et al. , 2010]. Deforestation and land use cultivation with moderate to complete removal of biomass will then decrease BSi stocks in soils, in particular when not compensated by Si fertilizers. With persistent harvest conditions (i.e., at least 240 years in cropland 2 in this study), soils enter “stage 2,” after passing the “turning point” (shown by the cross in Figure 6), i.e., when BSi drops below threshold values and no longer controls Si in soil water. Here rapid dissolution of BSi soil stocks could be partially compensated by DSi mobilization from nonbiogenic Si pools, which are shown to have high alkaline reactivity (Table S3) and higher field availability (i.e., CaCl₂). In cropland 2, δ³⁰ DSi soil water signatures are strongly depleted in light isotopes as compared to forest and cropland 1 in this study [Vandevenne et al. , 2015], showing that that contribution of BSi to Si availability in soil water is limited. We hypothesize that during this second stage, crop rotation with temporary cultures of non‐Si accumulators is crucial to allow soil regeneration and to prevent Si limitation for plants. Indeed, in the last “stage 3” after land use cultivation, soils are ultimately depleted in terms of plant‐available Si.

Figure 6

Conceptual summary figure of the change in AlkExSi pools in topsoils after land use conversion from forests to intensive cropland systems (adapted from Struyf et al. [2010]). Biogenic Si pool (BSi) and pedogenic Si Clay (P_ClaySi) are indicated with black lines. Land use types are indicated with numbers (1, 2,3) and filled circles. The cross sign indicates the threshold where BSi pools are not contributing to plant‐available Si and where plants may retrieve Si from other nonbiogenic Si soil pools.

4.3.2 Implications for Agriculture and Potential Mobilization of Si From Soils

Current models of deforestation [Struyf et al. , 2010] only consider rapid dissolution of biogenic Si stocks as the main driver for DSi control in soil and river water; here we show that lowered BSi stocks are compensated by and/or even trigger DSi release from biogeochemically active nonbiogenic Si pools. This study also suggests that other factors than land use may influence the formation and dissolution of pedogenic Si fractions, i.e., climate, parent material, weathering degree, and landscape topography. Knowledge of Si cycling in different soils is crucial to manage agricultural systems in view of global (climate) change and plant stress (e.g. drought), in particular given the stagnating crop yields of wheat in Europe during the last decades [Moore and Lobell , 2015] and the fact that Si fertilizers are not a common practice in many European croplands.

Whether or not landscape cultivation will change (lower) export fluxes of Si as compared with pristine forest conditions will not solely depend on biogenic Si availability in soils but on the complex and interacting effect of land use on reactive secondary Si, hydrological conditions (water availability and residence time), and also erosion processes [Clymans et al. , 2015]. The large reservoir of “reactive secondary Si” (biogenic + nonbiogenic) present in temperate weathered soils is likely concealing part of the land use imprint on Si cycling. In addition, local conditions that are independent of plant activity and land use can prevent Si transport from soil to river (e.g. opal‐CT precipitation in pasture soils). Our results using a multiextraction approach provide a first glimpse into the mechanisms that control Si availability in soils under different land uses.

Acknowledgments

F.V. and L.B. thank Special Research Funding of the University of Antwerp (BOF‐UA and NOI) for PhD fellowship funding, Flemish Hydraulics laboratory for using laboratory facilities, T. Van der Spiet and A. Cools for laboratory analysis, Wim Clymans and Lore Fondu for helping with the soil core collection expedition, and Daniel Conley for background data on Swedish soils. F.V. thanks Eugene Kelly and the Department for Soil and Crop Protection (Fort Collins, Colorado State University) for their laboratory facilities and expertise in phytolith extraction. E.S. thanks FWO (Research Foundation Flanders) for PhD fellowship, project, and postdoctoral research funding. We would like to thank BELSPO for funding the SOGLO project. B.R. acknowledges IWT for PhD fellowship funding. Data supporting this study are available to anyone from the authors by request to [e-mail] or [e-mail].