Characterization of phosphorus in a toposequence of subtropical perhumid forest soils facing a 2 subalpine lake 3

The productivity of forests is often considered to be limited by the availability of 22 phosphorus (P). Knowledge of the role of organic and inorganic P in humid subtropical forest soils is 23 lacking. In this study, we used chemical fractionation and 31P nuclear magnetic resonance (NMR) 24 spectroscopy to characterize the form of P and its distribution in undisturbed perhumid 25 Chamaecyparis forest soils. The toposequence of transects was investigated for the humic layer 26 from summit to footslope and lakeshore. The clay layer combined with a placic-like horizon in the 27 subsoil may affect the distribution of soil P because both total P and organic P (Po) contents in all 28 studied soils decreased with soil depth. In addition, Po content was negatively correlated with soil 29 crystalline Fe oxide content, whereas inorganic P (Pi) content was positively correlated with soil 30 crystalline Fe oxide content and slightly increased with soil depth. Thus, Pi may be mostly adsorbed 31 by soil crystalline Fe oxides in the soils. Among all extractable P fractions, the NaOH-Po fraction 32 appeared to be the major component, followed by NaHCO3-Po; the resin-P and HCl-Pi fractions 33 were lowest. In addition, we found no typical trend for Pi and Po contents in soils with topographical 34 change among the three sites. From the 31P-NMR spectra, the dominant Po form in soils from all 35 study sites was monoesters with similar spectra. The 31P-NMR findings were basically consistent 36 with those from chemical extraction. Soil formation processes may be the critical factor affecting the 37 distribution of soil P. High precipitation and year-round high humidity may be important in the 38 differentiation of the P species in this landscape. 39


Introduction
In terrestrial environments, mountainous forest is one of the canonical ecosystems that provide various ecosystem services such as maintaining biodiversity, water conservation [1] and carbon sequestration [2].Meanwhile, it is also a vulnerable ecosystem that can be influenced by different soil nutrients [3].Thus, understanding the soil nutrient distributions in a forest ecosystem is vital to maintain ecosystem functions and productivity [4].In such ecosystems, phosphorus (P) can be a limiting element because unlike nitrogen (N), which is mainly deposited from the atmosphere [5], P is mostly acquired from weathering soil parent material and is continuously lost due to soil erosion [6][7][8].The bioavailability of P in soils further relies on the chemical/physical conditions that fractionate the total P into different species [9,10].Understanding P availability and its transformation among each fraction will help assess the P supply capacity of the soil over the long term and to adapt management practices [11].
Some isolated P fractionation pools have key functions in the P cycle and plant nutrition [11][12][13][14][15].The organic and inorganic forms of P are usually separated and quantified by their plant availability.The mineralization of organic P is generally responsible for most of the P supply to plants [16,17], especially in mountain forest ecosystems [4,11,18].Actually, the sequential fraction can provide a general indicator of how biological and geochemical forms of P change during soil weathering in mountain forest ecosystems.In general, organic P was found to be the dominant fraction of total P (TP) in these forest ecosystems, and available P content was determined by the mineralization processes of organic P [11,19].Productivity, including plant growth and biomass production of trees in afforested mountain areas, is largely influenced by mineralization and microbial processes of organic P as well as soil organic matter.
However, identifying different fractions of P and evaluating their bioavailability are difficult because of spatial inconsistency and the complexity of geochemical properties.Several wet-chemical methods for determining P fractionations have been established and used in various ecosystems [20][21][22][23].Also, 31 P-nuclear magnetic resonance (NMR) spectroscopy has been used for determining the composition of soil P [24][25][26].De Feudis et al. [27] determined the P availability in subalpine forest soils in Italy and found organic P, bioavailable P contents and alkaline mono-phosphatase activity were all increased with altitude.Similarly, Doolette et al. [28] analyzed P composition in five alpine and subalpine forest soils and found that 54% to 66% of extractable P was contributed by organic P such as phosphomonoesters and inositol phosphonates, and the organic P composition was affected by temperature and soil moisture.
To our knowledge, studies of P fractionation in forests were mostly performed in temperate ecosystems [10,29,30], with relatively fewer studies from subtropical and tropical alpine forests.In this study, to evaluate the pedogenetic effects on the forms of P, we determined the composition of soil P with both chemical extraction and 31 P-NMR methods along a toposequence in a pristine subtropical subalpine forest.Because the changes in soil oxidation-reduction status affect the formation of iron (Fe) and consequently the P status in such humid forest soils, we hypothesized that the content of labile P associated with Fe oxides increases with changing topographic sequence from the summit to lakeshore because of leaching and soil erosion, whereas recalcitrant organic P, which is more complex-formed, will remain in the summit.

Study Sites
This study was conducted in the Yuanyang Lake forest ecosystem (24 • 35 N, 121 • 24 E) in northeastern Taiwan.The study sites covered an elevation of 1700 to 2000 m a.s.l., with an average annual temperature of 12.5 • C and a mean annual precipitation of more than 4000 mm.The ecosystem consists of a primary forest dominated by Taiwan false cypress (Chamaecyparis obtusa var.formosana) and an evergreen broadleaf shrub (Rhododendron formosanum Hemsl.).The bedrock of the study sites is composed of interbedded Tertiary shale and sandstone [31].This locality can be described as a temperate, very wet and mountainous ecosystem.It has been established as a Nature Reserve and selected as one of the long-term ecological research sites in Taiwan.The forest soils are divided into three main groups, which are closely related to the topography.The soil at the summit, with a slope of about 15 • , is classified as Typic Hapludult (Ultisols) [32], which is relatively well-drained and develops clear eluvial and illuvial boundaries.The footslope, with a slope of about 28 • , is dominated by Typic Dystrochrept (Inceptisols), where poor drainage caused by the clay and silty clay mineral horizon beneath the organic layer limits the downward movement of soluble compounds and thus hampers the soil profile development.Lithic Medihemist (Histosols) stretches from the lakeshore to the toeslope, about 1.5 m above the lake and with a slope of about 10 • , which is inundated by occasional storms.Details of the environment of this ecosystem are described elsewhere [33].Briefly, the soils at the summit and footslope are derived from partially podzolized soils and pure peats, while the soils at the lakeshore are mostly derived from peat, and directly lie on the bedrock [33].

Soil Sampling
A pedon sample was collected from each site to the bedrock.However, the pedon sample at the lakeshore was limited to the O horizon because no mineral layers were developed on the bedrock.Each horizon in the pedon was collected separately to determine the basic soil physiochemical properties.
To further determine the P fractionation along the topography, soil samples were collected from three selected sites along a topographic sequence in the forest that covered the summit, footslope and lakeshore.At each sampling site, three composite samples, each containing five subsamples, were collected with a soil auger with 8 cm in diameter and 10 cm in depth (O e and O a horizons).
Visible coarse organic materials, such as roots and litter were manually removed before sieving.The remaining soil samples were air dried and sieved through a 2-mm sieve for chemical analysis.

General Soil Chemical Properties
Soil pH was measured at a soil:water ratio of 1:1.Total organic C (TOC) and total N (TN) contents in the soil were determined with an NCS Elemental Analyzer (Model NA1500 Fisons, Milan, Italy).Cation-exchange capacity was determined by the NH 4 /Na exchange method [34].Crystalline Fe (Fe d ) and Al (Al d ) oxide contents were calculated by difference between dithionite and oxalate contents with following the experiments of previous study [35].Amorphous Fe (Fe o ) and Al (Al o ) oxide contents were measured by an ammonium oxalate extraction method [36].

Sequential Fractionation of P
Sequential fractionation was performed as described in [37].The sequential fractionation procedure removes progressively less available P with each subsequent soil extraction [38].The fractionation started with 0.5 g dried sieve soil.An anion exchange resin was used first to extract plant-available inorganic P (resin-P i ) [39].Then, the other P i content was determined directly in 0.5 M NaHCO 3 , 0.1 M NaOH, 1 M HCl and concentrated HCl extractions.The extracted solutions were then digested with H 2 SO 4 (97%) and H 2 O 2 (30%) at 300 • C to determine the total dissolved P (P d ) content of each fraction.The organic P (P o ) content was calculated by subtracting P i content from P d content in each fraction (P o = P d − P i ).The remaining soil was digested with H 2 SO 4 (97%) and H 2 O 2 (30%) at 300 • C to determine the residual P content.All extracts and digestions obtained were measured colourimetrically by the malachite green procedure [40].
Summed P i content was calculated as the sum of all analyzed P i fractions including resin-P i , NaHCO 3 -P i , NaOH-P i , HCl-P i and cHCl-P i .Summed P o content was calculated as the sum of all analyzed P o fractions including NaHCO 3 -P o , NaOH-P o and cHCl-P o .Summed P content was calculated as the sum of P i , P o and residual P content.Total P content of the soil samples was determined by digestion with H 2 SO 4 (97%) and H 2 O 2 (30%) at 300 • C.

31 P-NMR Measurements
Air-dried soil (5 g) was dispersed in 20 mL of 0.25 M NaOH-0.05M EDTA for 2 h, and the suspension was centrifuged at 12,100× g for 30 min.The extractant was then reacted with chelating resin for 6 h at room temperature to reduce the paramagnetic interference of iron and other metals in the NMR spectra.After being stirred, the resin was separated by filtration through Whatman 42 filter paper.The extract was freeze-dried for storage.A freeze-dried sample (0.1 g) of NaOH-EDTA extractant was dissolved in 0.5 mL of 0.5 M NaOH; then, 0.1 mL D 2 O was added, and the solution was transferred to a 5-mm NMR tube for 31 P-NMR analysis [41].The 31 P-NMR spectra were obtained at 242.86 MHz and 25 • C on a Bruker-600 NMR spectrometer with 60 • pulse, 3.5-s delay and 0.33-s acquisition time.The 31 P-NMR spectra were proton-decoupled by using an inverse-gated pulse sequence to overcome the nuclear Overhauser enhancement and for quantification [40,41].Depending on the P content in the alkaline extract, 500-2500 scans were used for an acceptable signal-to-noise ratio.Spectra were recorded with a line-broadening of 20 Hz.The chemical shift was measured relative to an external 85% H 3 PO 4 /D 2 O standard.The assignment of signals was based on Newman and Tate [42], Dai et al. [43], Condron et al. [24], and Robinson et al. [41].Contents of the various P components (phosphonate, inorganic orthophosphate, orthophosphate monoesters, orthophosphate diesters, pyrophosphate, polyphosphates) were determined according to relative resonance areas obtained by electronic integration.Inorganic orthophosphates and orthophosphate monoesters signals were separated by using a boundary determined from the valley between the two signals to the baseline [44].

Statistical Analyses
All extraction experiments were carried out in triplicate.Simple linear regression was used to compare the relation between soil P and soil Fe d , Fe o Al d and Al o concentrations.Differences in the P factions among the three sites were analyzed using one-way analysis of variance (One-Way ANOVA) and Tukey's honestly significant difference (HSD) test.JMP 11.0 (SAS Inc., Cary, NC, USA) was used for these statistical analyses.p < 0.05 was considered as statistically significant.

Results
The basic chemical properties of the studied soils are in Table 1.The soils were strongly acidic; pH values ranged from 3.3 to 4.5 in the three sampling sites.Both TOC and TN contents were high in the O horizon and decreased from the surface to the low horizons.Similarly, the cation-exchange capacity basically decreased from the surface to the low horizons.Total P content also decreased from the surface to the low horizons, but the difference was much less than for TOC and TN contents.
The Fe o , Fe d , Al o and Al d contents peaked in the Bt 2 horizon at the summit site and in the Bw 2 horizon at the footslope site.P i content was associated with amorphous (Fe o and Al o ) and crystalline (Fe d and Al d ) Al and Fe oxide contents and migrated vertically through the horizons with illuviation.Moreover, P i and Fe d contents were positively correlated, and P o and Fe d as well as P o and Fe o contents were negatively correlated in the soil samples (Figure 1a).However, the relation between P i and Fe o contents was not statistically significant (Figure 1b).In addition, only P i and Al o contents were positively correlated; P o and Al d contents were negatively correlated but not P i and Al d nor P o and Al o contents (Figure 2a,b).
Contents of total P and summed P and P i in the O horizons were greater at the footslope than the lakeshore, whereas the values in the summit site were in between those at the footslope and lakeshore.In addition, summed P o and residual P contents were similar among the three sites (Table 2).HCl-P i and cHCl-P i contents were similar among the three sites, whereas cHCl-P o content in the footslope soil was similar to that at the lakeshore but higher than that at the summit.NaOH-P i content was higher at the footslope than the summit and lakeshore.In addition, NaOH-P o content was higher at the summit than the footslope and lakeshore.NaHCO 3 extracted P i content was higher at the summit than the footslope and lakeshore, whereas NaHCO 3 -P o content was the highest at the footslope and was similar at the summit and lakeshore.Resin-P i content was higher at the lakeshore than the footslope, and resin-P i content at the summit was in between that at the other two sites.
Forests 2018, 9, x FOR PEER REVIEW 5 of 14 footslope and was similar at the summit and lakeshore.Resin-Pi content was higher at the lakeshore than the footslope, and resin-Pi content at the summit was in between that at the other two sites.Organic P was the dominant P fraction in the mountain forest soils of the three sites (Table 3).Moreover, NaOH-Po represented the major P fraction and contributed to more than 40% of the summed P content at the three sampling sites.NaHCO3-Po was the second most abundant P fraction among the three sites and contributed more than 20% of the summed P content.Summed Pi content (resin-Pi + NaHCO3-Pi + NaOH-Pi + HCl-Pi + cHCl-Pi) in surface soils of all study sites contained less than 18% of summed P, and NaOH-Pi was the major Pi fraction in total Pi.footslope and was similar at the summit and lakeshore.Resin-Pi content was higher at the lakeshore than the footslope, and resin-Pi content at the summit was in between that at the other two sites.Organic P was the dominant P fraction in the mountain forest soils of the three sites (Table 3).Moreover, NaOH-Po represented the major P fraction and contributed to more than 40% of the summed P content at the three sampling sites.NaHCO3-Po was the second most abundant P fraction among the three sites and contributed more than 20% of the summed P content.Summed Pi content (resin-Pi + NaHCO3-Pi + NaOH-Pi + HCl-Pi + cHCl-Pi) in surface soils of all study sites contained less than 18% of summed P, and NaOH-Pi was the major Pi fraction in total Pi.Organic P was the dominant P fraction in the mountain forest soils of the three sites (Table 3).Moreover, NaOH-P o represented the major P fraction and contributed to more than 40% of the summed P content at the three sampling sites.NaHCO 3 -P o was the second most abundant P fraction among the three sites and contributed more than 20% of the summed P content.Summed P i content (resin-P i + NaHCO 3 -P i + NaOH-P i + HCl-P i + cHCl-P i ) in surface soils of all study sites contained less than 18% of summed P, and NaOH-P i was the major P i fraction in total P i .Spectra obtained from 31 P-NMR analysis of NaOH-EDTA extracts revealed inorganic orthophosphate, orthophosphate monoesters, orthophosphate diesters, pyrophosphates, and phosphonates in the soil extracts (Figure 3).Spectra obtained from 31 P-NMR analysis of NaOH-EDTA extracts revealed inorganic orthophosphate, orthophosphate monoesters, orthophosphate diesters, pyrophosphates, and phosphonates in the soil extracts (Figure 3).Organic P compounds identified in the NaOH-EDTA extracts included orthophosphate monoesters, orthophosphate diesters, and phosphonates.Orthophosphate monoesters were the predominant species of extracted organic P in soil from all sites and contributed to more than 60% of the total P fractions (Figure 4).The proportion of orthophosphate diesters was much lower than that of orthophosphate monoesters and only contributed 15% to 20% of the total P pools.Content of phosphonates (18.7 ppm) ranged from only 2.2% to 4.0% of extracted P from the three sites, and the highest content was found at lakeshore, with waterlogged conditions.Inorganic P compounds identified in the NaOH-EDTA extracts included orthophosphate and pyrophosphates.Inorganic orthophosphate signals at 6.1-6.3 ppm ranged from 11.6% to 17.3% of the spectral area for all study sites (Table 3).The highest inorganic P content was found at the footslope.In addition, small additional pyrophosphate resonance (−4.3 ppm) was observed only at the summit site.Organic P compounds identified in the NaOH-EDTA extracts included orthophosphate monoesters, orthophosphate diesters, and phosphonates.Orthophosphate monoesters were the predominant species of extracted organic P in soil from all sites and contributed to more than 60% of the total P fractions (Figure 4).The proportion of orthophosphate diesters was much lower than that of orthophosphate monoesters and only contributed 15% to 20% of the total P pools.Content of phosphonates (18.7 ppm) ranged from only 2.2% to 4.0% of extracted P from the three sites, and the highest content was found at lakeshore, with waterlogged conditions.Inorganic P compounds identified in the NaOH-EDTA extracts included orthophosphate and pyrophosphates.Inorganic orthophosphate signals at 6.1-6.3 ppm ranged from 11.6% to 17.3% of the spectral area for all study sites (Table 3).The highest inorganic P content was found at the footslope.In addition, small additional pyrophosphate resonance (−4.3 ppm) was observed only at the summit site.

Soil Physiochemical Properties and Chemical Extractable P
The soil in this study site contains high moisture because of the year-round high precipitation [45].High soil moisture in mountainous forests retards decomposition of soil organic matter, and high precipitation increases the loss of cations, thereby resulting in decreased soil pH and Eh [38,46,47].This explains our observations of low soil pH in the studied sites.Moreover, our previous study at the same sites revealed a clay and silt-clay layer under the organic layer [45].This clay layer may retard the percolation [48] and therefore result in reduced soil TOC, TN and TP contents with increasing soil depth.
The high Feo, Fed, Alo and Ald contents in the B horizons implied that iron moved downwards and accumulated in the subsoil.Because high soil moisture and rich organic matter decreased soil redox potential in the surface layer, reduced Fe and Al ions moving from surface to the bottom layer were re-oxidized in the B horizons [49,50].This formation of accumulated Fe, which may due to the redoximorphic process [51,52], created a placic-like horizon and resulted in slow permeability of P in such perhumid forest soil [49].
The positive correlations between Pi and Fed but not Pi and Feo contents implied that most inorganic P may be adsorbed by crystalline Fe oxides in soils, whereas the negative correlation between Po and Fed as well as Po and Feo contents implied that organic P was in a complex formation in the soils.In addition, the vertical increase in Pi content, with an opposite trend to Po content, in each soil profile was significantly related to the content of amorphous and crystalline Fe oxides but not crystalline Al oxides.This observation implied that Fe oxides, rather than Al oxides, may chemically bind with Pi, and the accumulation of Pi in the subsoils could relate to the downward percolation and reoxidation of Fe.Sollins et al. [53] found that soil with high Fe hydrous oxides content tends to irreversibly fix polyvalent oxyanions such as phosphate because of chemosorption and occlusion.A similar trend was found in our previous study of subalpine forest soils [54], in which significant sorption of Pi to sesquioxides was via downward migration.
In addition, acidic soil conditions (pH < 4) typically facilitate Fe oxides reduction (i.e., Fe 2+ ), which may help the mobility of Fe in soil [55].The acidic soils at this study site may further affect the mobility of Pi, thereby resulting in low Pi content at the three study sites.
The depth of O horizons increased from the summit to the lakeshore, which suggests a process of erosion-deposition.In addition, because the mineral clay layer and placic-like horizon reduced vertical percolation, most of the soil organic matter in the water flow is transported laterally [33,45].This can help the downhill movement of soil nutrients in the O horizon [45,49] and may explain the

Soil Physiochemical Properties and Chemical Extractable P
The soil in this study site contains high moisture because of the year-round high precipitation [45].High soil moisture in mountainous forests retards decomposition of soil organic matter, and high precipitation increases the loss of cations, thereby resulting in decreased soil pH and Eh [38,46,47].This explains our observations of low soil pH in the studied sites.Moreover, our previous study at the same sites revealed a clay and silt-clay layer under the organic layer [45].This clay layer may retard the percolation [48] and therefore result in reduced soil TOC, TN and TP contents with increasing soil depth.
The high Fe o , Fe d , Al o and Al d contents in the B horizons implied that iron moved downwards and accumulated in the subsoil.Because high soil moisture and rich organic matter decreased soil redox potential in the surface layer, reduced Fe and Al ions moving from surface to the bottom layer were re-oxidized in the B horizons [49,50].This formation of accumulated Fe, which may due to the redoximorphic process [51,52], created a placic-like horizon and resulted in slow permeability of P in such perhumid forest soil [49].
The positive correlations between P i and Fe d but not P i and Fe o contents implied that most inorganic P may be adsorbed by crystalline Fe oxides in soils, whereas the negative correlation between P o and Fe d as well as P o and Fe o contents implied that organic P was in a complex formation in the soils.In addition, the vertical increase in P i content, with an opposite trend to P o content, in each soil profile was significantly related to the content of amorphous and crystalline Fe oxides but not crystalline Al oxides.This observation implied that Fe oxides, rather than Al oxides, may chemically bind with P i , and the accumulation of P i in the subsoils could relate to the downward percolation and reoxidation of Fe.Sollins et al. [53] found that soil with high Fe hydrous oxides content tends to irreversibly fix polyvalent oxyanions such as phosphate because of chemosorption and occlusion.A similar trend was found in our previous study of subalpine forest soils [54], in which significant sorption of P i to sesquioxides was via downward migration.
In addition, acidic soil conditions (pH < 4) typically facilitate Fe oxides reduction (i.e., Fe 2+ ), which may help the mobility of Fe in soil [55].The acidic soils at this study site may further affect the mobility of P i , thereby resulting in low P i content at the three study sites.
The depth of O horizons increased from the summit to the lakeshore, which suggests a process of erosion-deposition.In addition, because the mineral clay layer and placic-like horizon reduced vertical percolation, most of the soil organic matter in the water flow is transported laterally [33,45].This can help the downhill movement of soil nutrients in the O horizon [45,49] and may explain the increased soil TOC and TN contents from the summit to the footslope.In addition, because the soils at the summit and footslope were relatively well-drained while the soils at the lakeshore were poorly-drained [33], this may help the accumulation of organic matter and result in a thick O horizon at the lakeshore.

Chemical Extraction of Soil P
Because of the low overall P i concentrations in the three study sites, the different chemical extractable P i contents increased downhill but not significantly.In addition, the low labile P fractions, NaHCO 3 -P i and NaOH-P i , at the lakeshore may also be due to the vigorous fluctuation of the water level of the lake after showers or storms, which could remove the suspended particles or detritus of litter with the flooding and reduce the accumulation of P in the soil near the lakeshore [30].This was indirectly supported by the elevated TP concentration in the epilimnion of lake after medium rainfall events (256-620 mm) [56].
P o appeared to be the predominant fraction in the perhumid forest and was mostly non-acid extractable.Because the soil is acidic at the study sites, most P o fractions may not be labile and remained at higher values at the summit than at the lakeshore.The increase in total P o content in the surface horizon has been attributed to the input and accumulation of organic matter [57] and factors affected by topography can influence the availability of soil P [58].
Extracted soil P compounds showed that the contents of highly labile (NaHCO 3 -P o ), long-term P transformation (NaOH-P o ) and stable residual pool (cHCl-P o ) fractions changed between different sites, which showed that slope position affects the various P pools [59].As shown in Table 2, the sum of the highly labile P o fraction (NaHCO 3 -P o ) and long-term P transformation (NaOH-P o ) contributed more than 75% of the total extractable P in soils of all study sites, so organic P was the major P source in these soils.The P o accumulation in soil surfaces resulted from the biological cycling of P through the plant litter to the soil surface.In addition, we were aware that the density and height of understory Rhododendron were higher at the lakeshore than the footslope and summit, although we did not measure the abundance/density of the plants among the three sites.It could also be a potential factor causing the differences in P o accumulation.
The cHCl-P i extract has recalcitrant P forms associated with mainly Fe oxides and/or P derived from non-alkaline extractable debris, whereas cHCl-P o may include both stable, little and/or bioavailable (non-alkaline extractable) P forms.However, the proportions of cHCl-extractable P i and P o were only about 4.5% to 7.5% of sequentially extracted total P in soils of all study sites.The residual P is associated with highly organic materials such as lignin and organometallic complexes [60], but we have no information on the composition of organic matter in this fraction.

Spectra of 31 P-NMR Analyses
Orthophosphate monoesters are the most common forms of organic P o in soils [61][62][63].Monoester P includes high proportions of inositol phosphate, sugar phosphate and choline phosphate primarily derived from plant, animal, and microbial residues [64].
Depending on the soil types, inositol phosphates are reported to be the predominant organic P forms in Podosols [28], whereas αand β-glycerophosphate are the predominant organic P forms in Vertosols [65].In addition, high inositol phosphate contents were reported from several studies with cold and wet climates [66][67][68].
Inositol phosphates are typically considered of limited bioavailability because of the complex structure with soil minerals, clays, and humic compounds [62,69].Although 31 P-NMR analysis in our study had limited resolution to identify the inositol phosphates content, the low temperature and high precipitation of the study site may likely result in high inositol phosphates content in the soil.
Orthophosphate diesters at about 0 ppm [42] can be further classified into nucleic acids (−1-0 ppm) and phospholipids (0-2 ppm) [62].Orthophosphate diesters, including nucleic and phospholipids, more frequently accumulate in cool and moist acidic forest soils with low microbial activities than in agricultural soils [70][71][72].In acidic or wet soils, diester P proportion is between 10% and 36% of extracted P [72][73][74].Our findings are consistent with previous studies because the diester P proportion was between 15% and 20% of extracted P in the study sites.
A higher proportion of diester P providing a labile source for available P [75] was found at the footslope site, with poor drainage, than at the other sites.The lower orthophosphate diesters than monoesters content at the three study sites may contribute to the complexity of the chemical compounds.Orthophosphate diesters are more rapidly mineralizable than monoesters because they are generally less adsorbed to soil colloids than monoesters and can be easily hydrolyzed [24,61,76].
The content of phosphonates in soil are due to bacteria such as Bacillus cereus, which has a phosphonatase enzyme that produces phosphonates, but the bacteria are less prevalent in acidic soil [44,74,77].This observation may explain the low phosphonates concentrations observed at the study sites.
A small amount of pyrophosphate resonance (−4.3 ppm) was observed in spectra of soils at the three sites.Pyrophosphate is believed to be involved in biological P cycling in the soils and may be present in relatively well-drained soil that provides a proper environment for microbial activity and fungus [78].In addition, pyrophosphate is contributed by fungal P compounds [79].Because the surface soil of the three sites contained high soil organic matter and high soil moisture, it may provide a less favorable environment for microbial and fungal activities, resulting in low pyrophosphate concentrations.
Organic P represented between 82-88% and 77-81% of total P extracted by NaOH-EDTA and by chemical extraction from all sites studied, respectively.This result is similar to research byCade-Menun and Preston [80], who found 77% to 83% of P o in a low-pH forest perhaps because of the low decomposition of P o compounds in acidic forest soils that reduced the P i concentrations [44].
The signal intensity of 31 P-NMR spectra caused by paramagnetic Al, Fe and Mn in soils may reduce the spectra quality [47,81,82].Moreover, chemical hydrolysis of P o to P i may occur during alkaline extractions [80,83].However, in general, the results from NMR analysis were consistent with those of chemical fractionation in this study and other alpine and subalpine forest soils [28,49,54,68].

Conclusions
This study demonstrated that soil chemical extractable P i and P o can be vertically affected by the formation of Fe oxides in soils.Because of a clay layer combined with a placic-like horizon in the subsoil in our test site, both total P and P o contents were decreased with increasing soil depth, with P i content slightly increased in different soil horizons.The low permeable soil layer also favored downhill run-off, however, because P i contents were relatively low compared with P o contents; the contents did not significantly differ among the three study sites.Because most of the P was in organic forms, a negligible amount of P i may be released to the lake along the slope.Therefore, although topography and soil formation processes affect the distribution of soil P, high precipitation and year-round high humidity might be important for differentiation of the P species in this landscape.Moreover, the similarity of the 31 P-NMR spectra among the three sampling sites supports the alleviated differentiation of the P species in this landscape.Preserving plant abundance and plant litters should be able to maintain soil P cycles and prevent the P discharge from subtropical alpine forest ecosystems.

Figure 1 .
Figure 1.Correlations of contents of Pi and Po with Fed (a) and Feo (b) in the pedon samples collected from the three sampling sites.* Statistically significant at p < 0.05.

Figure 2 .
Figure 2. Correlations of contents of Pi and Po with Ald (a) and Alo (b) in the pedon samples collected from the three sampling sites.* Statistically significant at p < 0.05.

Figure 1 .
Figure 1.Correlations of contents of P i and P o with Fe d (a) and Fe o (b) in the pedon samples collected from the three sampling sites.* Statistically significant at p < 0.05.

Figure 1 .
Figure 1.Correlations of contents of Pi and Po with Fed (a) and Feo (b) in the pedon samples collected from the three sampling sites.* Statistically significant at p < 0.05.

Figure 2 .
Figure 2. Correlations of contents of Pi and Po with Ald (a) and Alo (b) in the pedon samples collected from the three sampling sites.* Statistically significant at p < 0.05.

Figure 2 .
Figure 2. Correlations of contents of P i and P o with Al d (a) and Al o (b) in the pedon samples collected from the three sampling sites.* Statistically significant at p < 0.05.

PFigure 4 .
Figure 4. Proportion of extracted P in various classes from humic samples at different sites determined by 31 P-NMR spectroscopy.Error bars indicate standard deviation.

Figure 4 .
Figure 4. Proportion of extracted P in various classes from humic samples at different sites determined by 31 P-NMR spectroscopy.Error bars indicate standard deviation.

Table 1 .
General chemical properties of studied soils.
CEC: cation-exchange capacity; TOC: total organic C; Fe o , Al o : iron and aluminum extracted by the ammonium oxalate method; Fe d , Al d : iron and aluminum extracted by the citrate-bicarbonate-dithionite method.

Table 2 .
The fractionation of P (mg kg −1 ) in humic soil samples (O horizon) in different topographic sites (sequential extraction).

in Extracts Summed Inorganic P (P i ) Organic P in Extracts Summed Organic P (P o ) Residual-P Summed P Total P Resin-P i NaHCO 3 -P i NaOH-P i HCl-P i cHCl-P i NaHCO 3 -P o NaOH-P o cHCl-P o
i : inorganic P extracted by 1.0 M HCl; cHCl-P i : inorganic P extracted by concentrated HCl; cHCl-P o : organic P extracted by concentrated HCl; Summed inorganic P: Resin-P i + NaHCO 3 -P i + NaOH-P i + HCl-P i + cHCl-P i ; Summed organic P: NaHCO 3 -P o + NaOH-P o + cHCl-P o ; Summed P: sum of P i + P o + residual P. Means followed by the same letters in the same column are not significantly different (p > 0.05) by Tukey's honestly significant difference test.

Table 3 .
Relative proportions of total P extracted for inorganic (P i ) and organic (P o ) forms in NaOH-EDTA extracts from humic soil samples (O horizon) determined by chemical extraction and 31 P-NMR spectroscopy.± 5.7 § P i : sum of resin-P i + NaHCO 3 -P i + NaOH-P i + HCl-P i + cHCl-P i .P o : sum of NaHCO 3 -P o + NaOH-P o + cHCl-P o .

Table 3 .
Relative proportions of total P extracted for inorganic (Pi) and organic (Po) forms in NaOH-EDTA extracts from humic soil samples (O horizon) determined by chemical extraction and 31 P-NMR spectroscopy.