Journal of Geochemical Exploration xxx (2014) xxx–xxx
GEXPLO-05307; No of Pages 9
Contents lists available at ScienceDirect
Journal of Geochemical Exploration
j ourna l homepage: www.e lsev ie r .com/ locate / jgeoexp
Biogeochemistry of soil inorganic and organic phosphorus: A compositional analysiswith balances
Léon E. Parent a,⁎, Serge-Étienne Parent a, Noura Ziadi b
a Department of Soils and Agri-Food Engineering, Paul-Comtois Building, Laval University, Quebec G1K 7P4, Canadab Soils and Crops Research and Development Centre, Agriculture and Agri-Food Canada, 2560 Hochelaga Blvd., Quebec City, QC G1V 2J3, Canada
Please cite this article as: Parent, L.E., et albalances, J. Geochem. Explor. (2014), http:
a b s t r a c t
a r t i c l e i n f o
Article history:Received 5 July 2013Accepted 29 January 2014Available online xxxx
Keywords:SoilPhosphorus cyclePhosphorus riskCompositional data analysisIsometric log ratio
Present soil P cyclingmodels and the related ecosystem services such as P retention deny the special properties ofcompositional data resulting from closure and leading tomethodological inconsistencies and pathological behav-ior. Our objective was to elaborate and interpret a hierarchy of supervised balances between components of thesoil P cycle computed as isometric log ratios (ilr) to avoid biaseswhen assessing P risk inmanaged terrestrial eco-systems. Fourty-one acid sandy soils, 56 acid to neutral loamy to clayey soils and 41 acid organic soils (10 fibric orhemic peat and 31 sapric ‘moorsh’ or ‘muck’ materials) were collected in Québec, Canada, analyzed for resin-P,NaHCO3-P, NaOH-P and residual P fractions, arranged into a hierarchy of sound balances and classified for therisk of P leaching. Pathological behavior was shown by spurious correlations between P fractions varying inmag-nitude, sign and significance depending on what was considered as the “whole” (e.g. total P or soil dry matter).Discriminant analysis across unbiased balances showed that the balances related to soil genesis dominated P bio-geochemistry. Soils at low or high P risk for P leaching showed contrasting degrees of Po and Pi loads. Low- andhigh-P risk loamy and clayey soils differed significantly for balances relating labile to slowly available Pi, aswell asthose relating Pi to Po. Low- and high-P risk sandy soils were contrasted by the balance relating oxalate-extractable Al and Fe. Peat and ‘moorsh’materials differed inmost ilrs. As soils loose ecosystem services providedby soil P geochemical and biological pools in terms of P-fractions balances, terrestrial ecosystems become lesssustainable.
The soil phosphorus cycle is related to soil genesis (Tiessen et al.,1984; Turner et al., 2007) and land management (Messiga et al.,2010). Ecosystem services provided by the soil P cycle are (1) plant nu-trition and productivity (Epstein and Bloom, 2005), (2) incorporation ofPi from the geochemical pool into the biological (Po) pool by soil biota(Stevenson, 1986), and (3) P sorption by Al and Fe oxi-hydroxides(van der Zee et al., 1987).
Phosphorus cycling is influenced by the nature of solid phases, soilpH and biological activity (Pierzynski et al., 2005). While the oxalate-extractable P, Fe and Al estimate P accumulation from fast and slow re-actions of phosphate ions with oxi-hydroxides (Lookman et al., 1995),soil pH determines the distribution of Pi species (Kovar and Claassen,2005). The joint P turnover in soils also depends on autotrophic and het-erotrophic activities as related to C and N cycling (Stewart and Tiessen,1987) and climatic conditions (Delgado-Baquerizo et al., 2013). Soil in-organic P (Pi) and organic P (Po) transformations are interrelated, be-cause Pi is the main source for P uptake by plants and soil organisms
fax: +1 418 656 3723.E. Parent).
ghts reserved.
., Biogeochemistry of soil in//dx.doi.org/10.1016/j.gexpl
and Po can replenish solution Pi through hydrolysis (Stewart andSharpley, 1987).
Hedley et al. (1982) proposed a sequential soil P extraction proce-dure to assess soil inorganic and organic P fractions. The Hedleymethodwas found to be useful to classify terrestrial ecosystems (Cross andSchlesinger, 1995; Litaor et al., 2004) and monitor long-term P changesin response to land management (Frossard et al., 2000). Because themethodmay overestimate Po and does not provide P speciation into dif-ferent P fractions, information about P forms should be interpretedwithcaution as ‘operational P fractions’ (Negassa and Leinweber, 2009) if notsupplemented by other techniques (Condron et al., 2005).
On the other hand, the P fractions are compositional data, i.e. strictlypositive data that convey relative information and have special proper-ties related to closure between 0 and the unit of measurement(Aitchison, 1986). Because redundancy, subcompositional incoherenceand inherent non-normal distribution lead to methodological bias inthe statistical analysis of compositional data (Filzmoser et al., 2009), sta-tistical models using raw concentration data have pathological behav-ior. Parent et al. (1992, 2000) and Duguet et al. (2006) thus appliedcompositional data analysis (CoDa) to C, N and P transformations insoils using the centered log ratio transformation of Aitchison (1986).Parent et al. (2009) and Abdi et al. (2011) suggested using the isometriclog ratio (ilr) transformation of Egozcue et al. (2003) and Egozcue and
organic and organic phosphorus: A compositional analysis witho.2014.01.030
Table 1Sequential binary partitions of (D= 11) parts made of C, N, Al and Fe oxi-hydroxides, inorganic P (Pi) and organic P (Po) pools for the formulation of balances as isometric logratio (n+ = number of components at numerator; n− = number of components at denominator; g(c+) = geometric mean of components at numerator; g(c−) = geometricmean of components at denominator).
ilr C N Alox Feox Resin-Pi NaHCO3-Pi NaOH-Pi NaHCO3-Po NaOH-Po Residual P Fv n+ n−
2 L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
Pawlowski-Glahn (2005, 2006) for the statistical analysis of soil P frac-tions as balances.
Our goals were to (1) demonstrate pathological phenomena in theclassical way to conduct statistical analyses on soil P biogeochemicalcompositions, (2) compare the P distribution in soils with contrastinggenesis, and (3) present ametaphoric pan balancemodel for diagnosingsoil P biogeochemical systems.
2. Material and methods
2.1. Soil collection
Mineral soils were sampled in the arable layer (0–15 to 0–30 cm) inthe Province of Quebec, Canada, and classified as clayey, loamy, andsandy. The 41 acid sandy soils under cranberry or potato productionwere classified (Soil Survey Staff, 1999) as sandy to loamy, coarse-loamy and coarse-silty Haplaquods, sandy to coarse-loamy, loamy,Haplaquepts, loamy Dystrocrepts, and loamy Endoaquents. The 56acid loamy to clayey soils under maize and soybean cropping systemswere classified as Haplaquepts (Soil Survey Staff, 1999). Organic soilswere Haplohemists and Haplofibrists (Soil Survey Staff, 1999): 10 peatmaterials were collected in natural mires (peatlands) and classified asfibric or hemic (Soil Survey Staff, 1999); 31 ‘moorsh’ (‘muck’) materials(Okruszko and Ilnicki, 2003) were collected in the arable layer(0–30 cm). ‘Moorshing’ (Okruszko and Ilnicki, 2003) is a soil transfor-mation process occurring in the aeratedprofile of organic soils followingdrainage and cultivation.
Table 2Properties of organic and mineral soils used in this study.
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
2.2. Soil analyses
Mineral soil materials were dried at 50 ºC for 24 h then sieved tob2 mm. The pH of mineral soils was measured in 0.01 M CaCl2 using asoil: solution ratio of 1:1 (v/v). Organic soil materials were air-dried toconstant weight and sieved to b1 mm. Soil pH was determined in 0.01M CaCl2 using a soil: solution weight: volumetric ratio of 5: 20 for‘moorsh’ materials and 1.2: 20 to 40 for peat materials, depending onliquid sorption capacity.
Soil texture was determined according to Gee and Bauder (1986).Organic C and Nwere quantified using CNS-Leco2000. Oxalate- extract-able P, Fe and Al were determined according to Ross and Wang (1993)as a measure of P sorption by oxi-hydroxides (van der Zee et al.,1987). The P, Al, and Fe were also determined according to Mehlich(1984), a method commonly used in North American routine laborato-ries to classify soils according to P fertility and the P risk of leaching(Guérin et al., 2007; Pellerin et al., 2006); approximately 10% of extract-ed P is Po (Khiari et al., 2000). Elements were quantified by inductivelycoupled argon plasma (ICP-OES).
The Hedley sequential P fractionationwas conducted using 0.5 g soilsamples according to Tiessen and Moir (2008). Briefly, resin-P (Pi) wasextracted with 25 mLwater and anionic membranes, shaken for 16 h at25°C, then centrifuged at 4°C for 15min at 38,000 g; the Pwas recoveredfrom membranes in 25 mL 0.5 M HCl. Thereafter NaHCO3-P (Pi and Po)was extracted using 25 mL 0.5 M NaHCO3, pH 8.5, shaken for 16 h at25°C, and centrifuged at 4°C for 15 min at 38,000 g. The NaOH-P (Piand Po) was obtained using 25 mL 0.1 M NaOH, shaken for 16 h at25°C; centrifuge at 4°C for 15 min at 38,000 g. The HCl-P (Pi): 25 mL
-textured mineral soils Fine-textured mineral soils
Fig. 1. Relationship between oxalate-extractable P and the sum of Hedley's slowly to rapidly available P (solid line is the 1:1 slope).
3L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
1.0MHCl, shaken for 16h at 25°C; centrifuge at 4°C for 15min at 38,000g. Residual P and total P in P fractions were extracted following potassi-um persulphate digestion in an autoclave (103.4 kPa, 121 °C) for 1.5 h(Zheng et al., 2001). The concentration of Pi was determined by themo-lybdenum blue colorimetric method (Murphy and Riley, 1962). TheNaHCO3-Po and NaOH-Po were calculated by difference between Piand total P.
The Hedley operational P pools were interpreted as follows (Crossand Schlesinger, 1995): (1) the resin-P fraction is a pool of easily ex-changeable and solution Pi; (2) the NaHCO3-P fraction is a labile poolof Pi or Po including microbial P; (3) the NaOH-P fraction comprises Piand Po of lower plant-availability associated with Al and Fe oxides andsoil organic matter; and (4) residual P that contains insoluble Pi associ-ated with Ca and Mg minerals and a highly resistant pool of inorganicoccluded P andPo included in stable humus. Operational P pools are het-erogeneous at molecular level (Turner et al., 2005).
2.3. Methodological limitations
Although the Hedley fractionation procedure is useful to investigatethe fate of native and applied P in terrestrial ecosystems, its value in pro-viding an insight of P biogeochemistry is limited (Hinsinger, 2001). Theallocation of sequential soil P extracts to operational P pools in terms ofenvironmental relevance and ecological functions has been criticized byTurner et al. (2005). To improve interpreting theHedley fractions, phos-phate molecules can be analyzed in chemical extracts using X-ray ab-sorption near edge structure [XANES], nuclear magnetic resonance[NMR], Fourier transform infrared [FTIR] and Raman spectroscopy(Kizewski et al., 2011; Negassa and Leinweber, 2009; Turner et al.,2007). However, due to complexity of environmental matrices, no sin-gle spectroscopic technique can characterize comprehensively the vari-ety of coexisting inorganic and organic P species in these systems(Kizewski et al., 2011).
Table 3Spurious correlations between P fractions expressed on soil dry mass basis (upper-right triang
ns, not significant.⁎ significant at 0.05 level.⁎⁎ significant at 0.01 level.⁎⁎⁎ significant at 0.001 level.
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
2.4. Statistical analysis
Statistical computationswere conducted in the R statistical environ-ment (R Development Core Team, 2013) and the “compositions” pack-age (van den Boogaart et al., 2013). The “robCompositions” (Templet al., 2013) package was used to robustly impute missing concentra-tions values. Levels of Po below detection limit were replaced by the de-tection limit of themethod multiplied by 0.65 (Martin-Fernandez et al.,2011). Discriminant analysis was performed using the “MASS” package(Venables and Ripley, 2002). Comparisons between balances were con-ducted using Tukey tests.
3. Theory
3.1. Compositional model
The constant sum assignment of a composition is computed asfollows (Aitchison, 1986):
C c1; c2;…; cDð Þ ¼ c1κXDi¼1
ci;
c2κXDi¼1
ci;…;
cDκXDi¼1
ci
0@
1A 1
Where C is the closure operator, κ is the unit of measurement and ciis the ith part of a D-parts composition. In this study, the parts consist ofsix P fractions, C, N, Al, Fe, and a filling value to the unit ofmeasurement.The filling value is an amalgamation of non-overlapping parts that al-lows back-transforming ilr values into the more familiar concentrationvalues.
A D-parts composition has D-1 dimensions (Aitchison andGreenacre, 2002). Egozcue et al. (2003) developed the isometric logratio (ilr) technique, that can generate D − 1 orthogonal balances orlog contrasts selected from a sequential binary partition (SBP)(Egozcue and Pawlowski-Glahn, 2005, 2006). Although a total of D! ×
le in italics) and total P basis (lower-left triangle).
Pi NaHCO3-Po NaOH-Po Residual P
0.18 ns 0.40⁎⁎ 0.53⁎⁎
0.29 ns 0.50⁎⁎⁎ 0.72⁎⁎⁎
0.54⁎⁎ 0.74⁎⁎⁎ 0.56⁎⁎
ns – 0.85⁎⁎⁎ 0.16 nsns 0.56⁎⁎ – 0.38⁎
ns −0.56⁎⁎ −0.41⁎⁎ –
organic and organic phosphorus: A compositional analysis witho.2014.01.030
Table 4Comparison between concentrations and ordinary log or ilr transformations showingmethodological biases in the computation of the means for sandy and organic soils.
4 L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
(D − 1)!/2D−1 schemes of orthogonal contrasts can be derived from aD-parts composition (Pawlowsky-Glahn et al., 2011), balances can besupervised by prior and expert knowledge (Aslam et al., 2013; Parentet al., 2012, in press, 2013a,b) assisted by exploratory bi-plot(Aitchison and Greenacre, 2002) and principal balances (Pawlowsky-Glahn et al., 2011).
A SBP is a (D− 1) × Dmatrix defining how the parts are partitionedinto balances, where parts labeled “+1” (group numerator) arecontrasted with parts labelled “−1” (group denominator) (e.g.Table 1). The ilr technique (Egozcue et al., 2003) projects the composi-tional vector into a Euclidean space of D − 1 non-overlapping ilrs(also called log contrasts or orthonormal balances) that are amenableto multivariate analysis without bias (Filzmoser and Hron, 2011). Theilrs are invariant upon change of units or scale. The possibility of usingilr for compositional data analyses is discussed in more details byMateu-Figueras et al. (2011).
Clay
1
810
3
6
9
4 2
5
7
Fig. 2. Discriminant analysis of six soil categories across ilr balances excepting ilr 1 (seeTable 1 for ilr identification). Categories are clayey, loamy and sandy soils, the latter soilsbeing under potato or cranberry production. Organic soil materials are either peat ormoorsh (‘muck’). Numbers in circles represent the ilr number in Table 1. Large semitrans-parent ellipses that enclose swarms of data points represent regions that include 95% ofthe theoretical distribution of canonical scores for each soil type. Smaller plainwhite ellip-ses represent confidence regions aboutmeans of canonical scores at 95% confidence level.
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
The ilrs are computed as follows (Egozcue and Pawlowski-Glahn,2005):
ilr j ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffinþj n
−j
nþj þ n−
j
vuut lng cþj� �
g c−j� � 2
Where nj+ and nj
− are the numbers of components in the “+1” orgroup and the “−1” or group, respectively, in the jth row of the SBP,g(cj+) is the geometric mean of components in the “+1” group andg(cj−) is the geometricmean of components in the “−1” group. The bal-anceswere expressed as [denominator | numerator]. A positive ilr valueindicates that the geometric mean of the numerator is higher than thegeometric mean of the denominator, and vice-versa.
3.2. Formulation of balances into coherent system and sub-systems
Complex yet organized systems such as ecosystems have some formof hierarchy (Jørgensen, 1992). The balance setup presented in Table 1to illustrate relationships between P fractions and soil properties follow-ed a hierarchy of high- to low-level balances among components in-volved in the P cycle.
The filling value was first contrasted with analyzed components tofacilitate classifying soils according to low (mineral soils) or high (or-ganic soils) C content (ilr 1). The C × N geometric mean that representsorganic materials was then balanced against other components (ilr 2).The C/N ratio, that decreases as organic matter is being decomposedby the biota and stabilizes at ≈10–12, is an index of organic matterquality (Stevenson, 1986). The relationship between C and N generallyexpressed as the C/N ratio is represented by the [N | C] balance (ilr 3),where N is positioned on the left side of the balance. As N loads morein the balance, the ilr becomesmore negative; in algebra, more negativenumbers are located to the left of an arraywhilemore positive numbersare located to the right.
Thereafter, residual P, Pi and Po were contrasted with amorphous(oxi-hydroxides or C-chelated) Al and Fe as P sorbents (Breeuwsmaand Silva, 1992; Breeuwsma et al., 1986; Celi et al., 1999; Ognalagaet al., 1994; Shang et al., 1990, 1992; Stevenson, 1986; van der Zeeet al., 1987) (ilr 4). The Al was then contrasted with Fe (ilr 5). The Ppools were balanced according to their bioavailability. Residual P is astable P fraction including insoluble Pi associated with Ca and Mg min-erals and a highly resistant pool of occluded Pi and Po included in stablehumus (Cross and Schlesinger, 1995): it was thus balanced against thegeochemical (Pi) and biological (Po) pools (ilr 6). The Pi and Po poolscontrasted in (ilr 7) were further partitioned into labile (resin-P andNaHCO3-P) and more slowly available (NaOH-P) pools (ilrs 8, 9 and10). The ilr1–5 and ilr6–10 balances are associated with soil genesisand land management, respectively.
3.3. Phosphorus risk classification
The P risk was assessed as follows:
IPS ¼ PMIII
AlMIII þ γFeMIII3
Where IPS is the Mehlich-III (MIII) index of P saturation and γ isequal to 1 for mineral soils (Pellerin et al., 2006) and 5 for organicsoils (Guérin et al., 2007). The amalgamation (AlMIII + γFeMIII) is anindex of P buffering capacity. The attentive reader will point out thatthe IPS and its γ parameter, computed across raw concentrations, donot appear to be compliant with compositional data properties. Howev-er, amalgamation (AlMIII + γFeMIII) is justified by the process being in-volved (Aitchison, 1986), here P sorption by Al and Fe. Suggestedcritical IPS values for P loss by leaching were found to be as follows:0.112 for mineral soils containing b20% clay, 0.079 for mineral soils
organic and organic phosphorus: A compositional analysis witho.2014.01.030
5L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
with 20–30% clay, 0.057 for mineral soils with 30–60% clay, 0.043 formineral soils containing N60% clay (Pellerin et al., 2006), and 0.050 fororganic soils (Guérin et al., 2007).
4. Results
4.1. Soil properties
The oxalate and Mehlich-III extractable Al, and Fe values presentedin Table 2 were within ranges for US and EU Spodosols, Inceptisolsand Entisols (Lookman et al., 1995; Sims et al., 2002; van der Zeeet al., 1987). Oxalate extractions ranged from 876 to 2990 mg Al kg−1
and from 4019 to 20815 mg Fe kg−1 in German organic soils(Schlichting et al., 2002) and were lower than in Karelian bogs(Russia) that contained 21 to 189 g FeOX kg−1 (Efimov et al., 1996).The C/N ratio was in the range of 12.9–114.2 (median = 25.2) insandy soils, 10.4–26.7 in loamy and clayey soils, and 10.3–66.8 in organ-ic soils, indicating large variation in soil organic matter quality withinsoils of the same category. Other concentration and ratio values variedwidely within soils as influenced by soil genesis and land management.
In loamy and clayey soils, the sumof resin-P, NaHCO3-P and NaOH-Pwas smaller than oxalate-P, indicating that the oxalate method extract-ed some residual P that contained insoluble inorganic P associated withCa and Mg minerals and a highly resistant pool of inorganic occluded Pand organic P included in stable humus (Fig. 1). The sum of labile andslowly available P forms was close to oxalate-Extractable P in sandyand organic soils.
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
4.2. Sample space definition and scale dependency
As shown in Tables 3 and 4, spurious correlations are generatedwhen data expressed on different scales, such as dry matter or total P,return different correlation coefficients and means. Correlation coeffi-cients varied in magnitude, sign and significance depending on samplespace and scale. As per example, the correlation coefficient betweenresin-P and residual P was 0.53 (p b 0.01) on dry weight basis and−0.42 (p b 0.01) on total P basis (Table 3). For raw data, the meanwas computed as the average of concentration values (Table 4). Forlog transformed data, themeanwas the geometric mean across concen-tration values. For ilr data, the means were back-transformations fromilr means. It is noteworthy that the methodological bias measured bydeparture of means from back-transformed ilr means was small orlarge depending on the data. For sandy soils, the bias was large forraw concentrations and small for their ordinary log transformation(Table 4). For organic soils, the bias was large for both the raw concen-trations and ordinary log expressions. Therefore, soil P biogeochemistryshould not be analyzed statistically (e.g. discriminant analysis) in thepathological concentration domain but in thewell-behaved balance do-main where orthogonality (linear independence) removes redundancyand scale invariance avoids spurious correlations.
4.3. Discriminant analysis
Discriminant analysis (DA) across ilr balances of six soil categories(loamy soils, clayey soils, sandy soils under potato or cranberry produc-tion, ‘peat’ and ‘moorsh’ organic soils) is shown in Fig. 2. Overall,mineral
organic and organic phosphorus: A compositional analysis witho.2014.01.030
Fig. 3. Discriminant scores for ilr balances, excepting ilr 1, and Tukey test (p b 0.05) com-paring ilr means of P-saturated (high P risk) and P-unsaturated (low P risk) in each soilcategory: (a) loamy and clayey soils, (b) sandy soils, and (c) organic soils. See Table 1for ilr identification.
6 L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
soils, moorsh and peat were largely discriminated while distances be-tween mineral soils were much smaller. The ilr 1, ilr 3 and ilr 5, thatare related to soil genesis (C/N, Alox and Feox), separated effectivelymin-eral from organic soils. The ilr 6 and ilr 9, related to land management,separated organic and mineral soils, on the one hand, and natural peatfrom cultivated moorsh materials, on the other. Indeed, peat materialswere much lower than moorsh in NaHCO3-Pi (lowest values inTable 2) compared to resin-Pi.
4.4. Correlation matrices
In Table 5, a positive correlation between two ilrs signifies that nu-merators or denominators load in the same direction; a negative corre-lation means that numerators or denominators load in the oppositedirection. Soil pH appeared to be a determinant factor in the P cycle.
In loamy–clayey soils, there were 17 significant correlation coef-ficients at the 0.01 level. The [N | C] balance (ilr 3) decreased signifi-cantly (r =−0.70, p b 0.001) with the balance relating the analyzedcomponents and the filling value (ilr 1). Residual P (denominator inilr 6) correlated with the filling value (denominator in ilr 1) (r= 0.71,p b 0.001), components of the P cycle (ilr 2) (r=−0.55, p b 0.001), the[N | C] balance (ilr 3) (r=−0.48, p b 0.001) and P sorption capacity rel-ative to P accumulation (ilr 4) (r =−0.55, p b 0.001). In contrast withother balances, ilr 6 increased with pH (r = 0.54, p b 0.001).
In sandy soils, there were 11 correlation coefficients significant at the0.01 level. The [N | C] balance decreasedwith ilr 1 (r=−0.62, pb 0.001),indicating stabilization of soil organic matter. The balance between Poand Pi fractions (ilr 7) increased with [N | C] (r = −0.60, p b 0.001)while [N | C] decreased as pH increased.
In organic soils, there were 32 correlation coefficients significantat the 0.01 level due to large contrast in land management betweenpeat and ‘moorsh’ materials. The ilr 9 and pH were closely correlated(r=−0.91, p b 0.001), indicating that, at low pH, resin-Pi predominat-ed over NaHCO3-Pi. Indeed, pH correlated negatively with most ilrs.
4.5. Phosphorus biogeochemistry and P risk
Robust Student-based discriminant analyses across balanceswere performed for each soil category (loamy and clayey soils,sandy soils, and organic soils) segmented by soil P risk (P-saturatedand P-unsaturated).
Soil P risk groups were discriminated by five balances in loamy–clayey soils, two balances in sandy soils and seven balances in organicsoils (Figs. 3a–c). In the loamy and clayey soil class, the ilr 8 [NaOH-Pi| NaHCO3-Pi, resin-Pi] showed the highest and positive score, indicatingthat high P availability as NaHCO3-Pi and resin-Pi compared to NaOH-Pi.In sandy soils, the ilr 4 and ilr 5 discriminatedmost, indicating that low-P risk soils contained more Alox than Feox and lower concentrations of Pfractions compared to high-P risk soils. In organic soils, low-risk soilscontained more NaHCO3-Po than NaOH-Po (ilr 10) and more P sorptioncapacity against P fractions (ilr 4), indicating effective biological cyclingand retention of Pi.
The metaphoric mobile-and-fulcrums pan balance setup(Figs. 4a–c) elaborated from SBP in Table 1 shows how thewhole pic-ture of soil P biogeochemistry changes as P fractions load differentlyin buckets. The setup coherently includes a domain of balances (bias-free results), where statistical analyses are conducted, and a domainof back-transformed concentrations, where raw concentrationvalues or proportions are appreciated relatively to each other (rela-tive values). In general, the labile to slowly available Pi pools ap-peared to be much larger where P risk for leaching was higherwhile Po did not vary markedly in mineral soils (Figs. 4a,b). In organ-ic soils, most components varied substantially between low- andhigh-P risk soils.
organic and organic phosphorus: A compositional analysis witho.2014.01.030
Fig. 4. Balance setup contrasting operational P pools and soil genetic attributes for (a) loamy and clayey soils, (b) sandy soils, and (c) organic soils. The setup comprises a balance domain,where statistical analyses are conducted and confidence intervals about balancemeans of soil low- and high-P risk are reported along a supervised hierarchy, and a concentration domain,where components of the sample space are appreciated in relation to each other. Univariate confidence intervals were computed from t distributions (p = 0.05). See Table 1 for ilridentification.
7L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
5. Discussion
5.1. Methodological biases in P fractionation data
It is common use to present and analyze compositional data relatedto the P cycle in terms of concentrations or proportions of P species,amalgamated compositions and dual ratios (e.g. Tiessen et al., 1984;Turner et al., 2007). As highlighted previously, this approach deniesthe intrinsic properties of compositional data that, if not adequately
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
handled, lead to methodological bias in their analysis (Aitchison,1986; Bacon-Shone, 2011).
Cade-Menun et al. (2010) warned that, if P recovery is low, careshould be taken when the sample space is changed from concentrationto percentage: in this case, scale change from soil mass to total P or thesum of P fractions produced sub-compositional incoherence that led toconflicting ANOVA results, hence pathological behavior, and possiblywrong interpretations and misleading decisions. Indeed, Buccianti(2013) also warned that, although differences are sometimes tenuous
organic and organic phosphorus: A compositional analysis witho.2014.01.030
8 L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
between the log ratio and classical approaches, most researchers stilladopt at fault the classical option referring to ‘common sense’, hence ob-scuring data analysis.
5.2. Balances between operational P pools
For loamy and clayey soils, resin-Pi and NaHCO3-Pi pools increasedwith P saturation (ilr 8) while Pi increase faster than Po (Fig. 4). Hence,those soils are at high P risk in presence of increasing amounts of labileP, especially as clay content increases (Pellerin et al., 2006).
In Spodosols,most organic P is located in the L-F-H(Ah)mor and acidmull horizons (Paré and Bernier, 1989). Weathering increasessesquioxide (Al, Fe, Mn oxi-hydroxides) activity (ilr 4 and ilr 5) thatwithdraw some P from rapid P cycling (Stewart and Tiessen, 1987). Inmanaged Spodosols, the genetic humus layers have been mixed withthe B horizons to form the Ap horizon. The P accumulation through in-tensive crop fertilization modified the P cycle profoundly thereafter assoil chemical P sorption capacity was gradually saturated with Padded as fertilizer. Indeed, the P inputs into terrestrial ecosystemsshould be tailored to soil's capacity to accumulate Pi and Po.
Although the ratios of C and N to P (corresponding to balances ilr 2and ilr 3) are unlikely to be unequivocal indicators of soil genesis(Stewart and Tiessen, 1987), ilr 3 clearly discriminated peat frommoorsh (Fig. 3c). Peat accumulation in mires is a geological C-buildupprocess (Grosse-Brauckmann, 1962), while moorsh formation is aman-made soil genetic process driven by drainage and cultivation(Jongerius and Pons, 1962; Okruszko and Ilnicki, 2003). Indeed, miresefficiently recycle P into easily available resin-Pi to sustain Nmineraliza-tion even at high C/N ratio (Damann, 1988), and thus accumulate slowlyavailable P minimally. In organic soils undergoing soil genesis where Prisk for leaching is high as Pi accumulates in the soil (Guérin et al.,2007), added P as well as pH increase through liming led to an accumu-lation of all P fractions but NaHCO3-Po (Parent et al., 1992).
5.3. The P risk
The predominance of Pi in high P risk soils as foundhere does not ex-clude the possibility that Po formsmay be at environmental risk. Indeed,Po can account for a large proportion of total P in soil solution, drainagewater, runoffwater and streamwater, and can be hydrolyzed by soil en-zymes (Condron et al., 2005). However, the identification of Po speciesrequires using NMR methods to ascertain soil's capacity to retain suchforms of Po.
5.4. Schemes of P transformation
The hierarchy of balances selected in this paper was based on priorknowledge of soil P fraction relationships derived from P extractability(Abdi et al., 2011), path analyses (Tiessen et al., 1984; Zheng et al.,2002) and state-space models (Shuai and Yost, 2004). In addition, thebalance designmay reflect the relationships between the C, N and P cy-cles at a higher hierarchical level Up till now, the coupling of C, N and Pcycles have been monitored using multiple ratios such as C:N:P:S(Stevenson, 1986), dual ratios such as the three N:P, C:P, and C:N ratiosfor three components, and rawconcentrations or proportions (Delgado-Baquerizo et al., 2013). By generating redundancy and spurious correla-tions, such pathological approachmay distort results and lead to wronginterpretations. To avoidmethodological inconsistencies, soil cycles andP biogeochemistry could be analyzed in future research using the com-positional balance concept.
6. Conclusions
In this paper, statistical analyses were conducted using isometric logratio balances across P fractions rather than raw concentrations or dualratios that inherently generate methodological inconsistencies. We
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
showed that classical methods to handle soil P fractions were patholog-ical because they spurious correlations and biased means.
Soils differed markedly in terms of P-fractions distribution. Discrim-inant analysis of soil P balances separated soil classes according to soilgenesis and land management. Moreover, the geochemical or Pi poolappeared to be the one contributingmost to the P risk across soil classes,because the geochemical pool can accumulate P freely as ecosystem ser-vices to recycle the P biologically and sorb the P chemically are beinglost. Obviously, P inputs into terrestrial ecosystems should be tailoredto soil's capacity to buffer Pi and cycle Po.
Statistical results presented by Cade-Menun et al. (2010) andBuccianti (2013) and those reported in this paper call for an urgentchange from classical analysis using concentration data that conveypathological behavior by to well-behaved compositional data analysistechniques such as ilr in the study of P biogeochemical fractions andprocesses.
AbbreviationsCoDa compositional data analysisFv filling valueilr isometric log ratioIPS index of phosphorus saturationSBP sequential binary partition
Acknowledgments
This project was supported financially by the Natural Sciences andEngineering Council of Canada (DG-2254 and CRDPJ 385199-09) andCanadian farmpartners as follows: Cultures Dolbec Inc., St-Ubalde, Qué-bec, Canada; Groupe Gosselin FG Inc., Pont Rouge, Québec, Canada;Agriparmentier Inc. and Prochamps Inc., Notre-Dame-du-Bon-Conseil,Québec, Canada; Ferme Daniel Bolduc et Fils Inc., Péribonka, Québec,Canada.
References
Abdi, D., Ziadi, N., Parent, L.E., 2011. Compositional analysis of phosphorus pools in Cana-dianMollisols. Proceedings 4th International Workshop on Compositional Data Anal-ysis, San Feliu de Guixols, Spain, 9–13 May 2011 (http://congress.cimne.com/codawork11/Admin/Files/FilePaper/p57.pdf).
Aitchison, J., 1986. The statistical analysis of compositional data. Chapman and Hall,London.
Aitchison, J., Greenacre, M., 2002. Biplots of compositional data. J. R. Stat. Soc.: Ser. C: Appl.Stat. 51 (4), 375–392.
Aslam, S., Garnier, P., Rumpel, C., Parent, S.É., Benoit, P., 2013. Adsorption and desorptionbehavior of selected pesticides as influenced by decomposition of maize mulch.Chemosphere 91 (11), 1447–1455.
Bacon-Shone, J., 2011. A short history of compositional data analysis. In: Pawlowsky-Glahn, V., Buccianti, A. (Eds.), Compositional data analysis: Theory and Applications.John Wiley and Sons, New-York, pp. 3–11.
Breeuwsma, A., Silva, S., 1992. Phosphorus fertilization and environmental effects in theNetherlands and the Po region (Italy), Winand Staring Centre for Integrated Land,Soil, andWater Research, Agriculture Research Department Report #57.Wageningen,The Netherlands (39 pp.).
Breeuwsma, A., Wösten, J.H., Vleeshouwer, J.J., van Slobbe, A.M., Bouma, J., 1986. Deriva-tion of land qualities to assess environmental problems from soil surveys. Soil Sci.Soc. Am. J. 50, 186–190.
Buccianti, A., 2013. Is compositional data analysis a way to see beyond the illusion?Comput. Geosci. 50, 165–173.
Cade-Menun, B.J., Carter, M.R., James, D.C., Liu, C.W., 2010. Phosphorus forms and chem-istry in the soil profile under long-term conservation tillage: a phosphorus-31 nuclearmagnetic resonance study. J. Environ. Qual. 39, 1647–1656.
Celi, L., Lamacchia, S., Marsan, F.A., Barberis, E., 1999. Interaction of inositolhexaphosphate on clays: adsorption and charging phenomena. Soil Sci. 164, 574–585.
Condron, L.M., Turner, B.L., Cade-Menun, B.J., 2005. Chemistry and dynamics of soil organ-ic phosphorus. In: Sims, T.J., Sharpley, A.N. (Eds.), Phosphorus: agriculture and the en-vironment, Agronomy Series #46. Soil Science Society of America, Madison, WI,pp. 87–121.
Cross, A.F., Schlesinger, W.H., 1995. A literature review and evaluation of the Hedley frac-tionation: applications to the biogeochemical cycle of soil phosphorus in natural eco-systems. Geoderma 64, 197–214.
Damann, A.W.H., 1988. Regulation of nitrogen and retention in Sphagnum bogs and otherpeatlands. Oikos 51, 291–305.
organic and organic phosphorus: A compositional analysis witho.2014.01.030
9L.E. Parent et al. / Journal of Geochemical Exploration xxx (2014) xxx–xxx
Delgado-Baquerizo, M., Maestre, F.T., Gallardo, A., et al., 2013. Decoupling of soil nutrientcycles as a function of aridity in global drylands. Nature 502, 672–689.
Duguet, F., Parent, L.E., Ndayegamiye, A., 2006. Compositional indices of net nitrificationin organic soils. Soil Sci. 171, 886–901.
Efimov, V.N., Kornilova, L.I., Ryabtseva, M.E., 1996. Absorption capacity and kinetics ofsorption of phosphate ions in ferruginous peat soils of lowmoor bogs. Eurasian SoilSci. 928–933.
Egozcue, J.J., Pawlowski-Glahn, V., 2005. Groups of parts and their balances in composi-tional data analysis. Math. Geol. 37, 795–828.
Egozcue, J.J., Pawlowsky-Glahn, V., 2006. Simplicial geometry for compositional data. In:Buccianti, A., Mateu-Figueras, G., Pawlowsky-Glahn, V. (Eds.), Compositional DataAnalysis in the Geosciences: From Theory to Practice. Special Publications, 264. Geo-logical Society, London. ISBN: 1-86239-205-6, pp. 67–77.
Egozcue, J.J., Pawlowski-Glahn, V., Mateu-Figueras, G., Barcelo-Vidal, C., 2003. Isometriclog ratio transformations for compositional data analysis. Math. Geol. 35, 279–300.
Epstein, E., Bloom, A.J., 2005. Mineral nutrition of plants: principles and perspectives, 3rded. Sinauer Associates Inc., Sunderland, MA.
Filzmoser, P., Hron, K., 2011. Robust statistical analysis. In: Pawlowsky-Glahn, V.,Buccianti, A. (Eds.), Compositional data analysis: Theory and Applications. JohnWiley and Sons, New-York, pp. 57–72.
Filzmoser, P., Hron, K., Reimann, C., 2009. Univariate statistical analysis of environmental(compositional) data: problems and possibilities. Sci. Total Environ. 407 (23),6100–6108.
Frossard, E., Condron, L.M., Oberson, A., Sinaj, S., Fardeau, J.C., 2000. Processes governingphosphorus availability in temperate soils. J. Environ. Qual. 29, 15–23.
Gee, G.W., Bauder, J.W., 1986. Particle-size analysis, In: Klute, A. (Ed.), Methods of soilanalysis, Part 1, Second ed. Soil Science Society of America & American Society ofAgronomy, Madison, WI, pp. 383–411.
Grosse-Brauckmann, G., 1962. Torfe und torfbildende Pflanzengesellschaften. Zeitschriftfür Kulturtechnik und Flurbereiningen 3, 205–225.
Guérin, J., Parent, L.E., Abdelhafid, R., 2007. Agri-environmental thresholds usingMehlich-III soil phosphorus saturation index for vegetables in Histosols. J. Environ. Qual. 36,975–982.
Hedley, M.J., Stewart, J.W.B., Chauhan, B.S.j, 1982. Changes in inorganic and organic soilphosphorus fractions by cultivation practices and by laboratory incubations. SoilSci. Soc. Am. J. 46, 970–976.
Hinsinger, P., 2001. Bioavailability of soil inorganic P in the rhizosphere as affected byroot-induced chemical change: a review. Plant Soil 237, 173–195.
Jongerius, A., Pons, L., 1962. Soil genesis in organic soils. Boor en Spade 12, 156–168.Jørgensen, S.E., 1992. Integration of ecosystem theories: a pattern. Kluwer Academic Publ,
An agroenvironmental phosphate saturation index for acid, coarse-textured, soils.J. Environ. Qual. 29, 1561–1567.
Kizewski, F., Liu, Y.T., Morris, A., Hesterberg, D., 2011. Spectroscopic approaches for phospho-rus speciation in soils and other environmental systems. J. Environ. Qual. 40, 751–766.
Kovar, J.L., Claassen, N., 2005. Soil–root interactions and phosphorus nutrition of plants.In: Sims, T.J., Sharpley, A.N. (Eds.), Phosphorus: agriculture and the environment,Agronomy Series #46. Soil Science Society of America, Madison, WI, pp. 379–414.
Litaor, M.I., Reichmann, O., Auerswald, K., Haim, A., Schenker, M., 2004. The geochemistryof phosphorus in peat soils of a semiarid altered wetland. Soil Sci. Soc. Am. J. 68,2078–2085.
Lookman, R., Freese, D., Merckx, R., Vlassak, K., Van Riemsdijk, W.H., 1995. Long-term ki-netics of phosphate release from soil. Environmental Science & Technology 29,1569–1575.
Martin-Fernandez, J.A., Palarea-Albaladejo, J., Olea, R., 2011. Dealing with zeroes. In:Pawlowsky-Glahn, V., Buccianti, A. (Eds.), Compositional Data Analysis: Theory andApplications. John Wiley and Sons, New-York, pp. 43–58.
Mateu-Figueras, G., Pawlowsky-Glahn, V., Egozcue, J.J., 2011. The principle of working oncoordinates. In: Pawlowsky-Glahn, V., Buccianti, A. (Eds.), Compositional Data Analy-sis: Theory and Applications. Wiley, Chichester UK (ISBN-10: 0-470-71135-3).
Mehlich, A., 1984. Mehlich-3 soil test extractant: a modification of Mehlich-2 extractant.Commun. Soil Sci. Plant Anal. 15, 1409–1416.
Messiga, A.J., Ziadi, N., Morel, C., Parent, L.E., 2010. Soil phosphorus availability in no-tillversus conventional tillage following freezing and thawing cycles. Can. J. Soil Sci.90, 419–428.
Murphy, J., Riley, J.P., 1962. A modified single solution method for the determination ofphosphate in natural waters. Anal. Chim. Acta. 27, 31–36.
Negassa, W., Leinweber, P., 2009. How does the Hedley sequential phosphorus fraction-ation reflect impacts of land use and management on soil phosphorus: A review.J. Plant Nutr. Soil Sci. 172, 305–325.
Ognalaga, M., Frossard, E., Thomas, F., 1994. Glucose-1-phosphate and myo-inositolhexaphosphate adsorption mechanisms on goethite. Soil Sci. Soc. Am. J. 58, 332–337.
Okruszko, H., Ilnicki, P., 2003. The moorsh horizons as quality indicators of reclaimed or-ganic soils. In: Parent, L.E., Ilnicki, P. (Eds.), Organic soil and peat material for sustain-able agriculture. CRC Press, Boca Raton, FL, pp. 1–14.
Paré, D., Bernier, B., 1989. Origin of the phosphorus deficiency observed in declining sugarmaple stands in the Quebec Appalachians. Canadian Journal of Forestry Research 19,24–34.
Parent, L.E., Sasseville, L., Ndayegamiye, A., Karam, A., 1992. The P status of cultivated or-ganic soils in Quebec. Proceedings of the 9th International Peat Congress, Uppsala(Sweden), 22–26 June 1992, 2. 3, pp. 400–410.
Please cite this article as: Parent, L.E., et al., Biogeochemistry of soil inbalances, J. Geochem. Explor. (2014), http://dx.doi.org/10.1016/j.gexpl
Parent, L.E., Viau, A.A., Anctil, F., 2000. Nitrogen and phosphorus fractions as indicators oforganic soil quality. Suo 51 (3), 71–81.
Parent, L.E., Bolinder, M.A., Gallichand, J., 2009. Contribution of Régis R. Simard to phos-phorus research in agroecosystems and future prospects. Canadian Journal of Soil Sci-ence 88, 145–155.
Parent, S.-É., Parent, L.E., Rozane, D.E., Hernandes, A., Natale, W., 2012. Nutrient balance asparadigm of plant and soil chemometrics. In: Issaka, R.N. (Ed.), Soil Fertility. InTechPubl, New-York, pp. 83:114–83: (http://www.intechopen.com/books/soil-fertility).
Parent, S.É., Parent, L.E., Egozcue, J.J., Rozane, D.E., Hernandes, A., Lapointe, L., Hébert-Gentile, V., Naess, K., Marchand, S., Lafond, J., Mattos Jr., D., Barlow, P., Natale, W.2013a. The plant ionome revisited by the nutrient balance concept. Frontiers inPlant Science 4 (39), 1–10. http://dx.doi.org/10.3389/fpls.2013.00039.
Parent, L.E., Parent, S.É., Hébert-Gentile, V., Naess, K., Lapointe, L. 2013b. Mineral balanceplasticity of cloudberry (Rubus chamaemorus) in Quebec-Labrador. American Journalof Plant Sciences 4 (7), 1508–1520.
Pawlowsky-Glahn, V., Egozcue, J.J., Tolosana-Delgado, R., 2011. Principal balances. In:Egozcue, J.J., Tolosana-Delgado, R., Ortego, M.I. (Eds.), Proceedings of the 4th Interna-tional Workshop on Compositional Data Analysis (Codawork 2011), San Feliu deGuixols, Spain, 9–13 May 2011 (http://congress.cimne.com/codawork11/Admin/Files/FilePaper/p55.pdf).
Pellerin, A., Parent, L.É., Fortin, J., Tremblay, C., Khiari, L., Giroux, M., 2006. EnvironmentalMehlich-III soil phosphorus saturation indices for Quebec acid to near neutral mineralsoils varying in texture and genesis. Can. J. Soil Sci. 86, 711–723.
Pierzynski, G.M., McDowell, R.W., Sims, J.T., 2005. Chemistry, cycling, and potential move-ment of inorganic phosphorus in soils. In: Sims, T.J., Sharpley, A.N. (Eds.), Phosphorus:agriculture and the environment, Agronomy Series #46. Soil Science Society ofAmerica, Madison, WI, pp. 53–86.
R Development Core Team, 2013. R: a language and environment for statistical computing. RFoundation for Statistical Computing (Version 2.15.2. http://www.R-project.org).
Ross, G.L., Wang, C., 1993. Extractable Al, Fe, Mn, and Si. In: Carter, M.R. (Ed.), Soil sam-pling and methods of analysis. Lewis Publ, Boca Raton, FL, pp. 239–246.
Schlichting, A., Leinweber, P., Meissner, R., Altermann, M., 2002. Sequentially extractedphosphorus fractions in peat-derived soils. J. Plant Nutr. Soil Sci. 165, 290–298.
Shang, C., Huang, P.M., Stewart, J.W.B., 1990. Kinetics of adsorption of organic and inor-ganic phosphates by short-range ordered precipitate of aluminium. Can. J. Soil Sci.70, 461–470.
Shang, C., Stewart, J.W.B., Huang, P.M., 1992. pH effect on kinetics of adsorption of organicand inorganic phosphates by short-range ordered aluminium and iron precipitates.Geoderma 53, 1–14.
Shuai, X., Yost, R.S., 2004. State-space modeling to simplify soil phosphorus fractionation.Soil Sci. Soc. Am. J. 68, 1437–1444.
Sims, J.T., Maguire, R.O., Leytem, A.B., Gartley, K.L., Pautler, M.C., 2002. Evaluation ofMehlich-3 as an agri-environmental soil phosphorus test for the mid-AtlanticUnited States of America. Soil Sci. Soc. Am. J. 66, 2016–2032.
Stewart, J.W.B., Sharpley, A.N., 1987. Controls on dynamics of soil and fertilizer phospho-rus and sulfur. In: Mortdvedt, J.J., Buxton, D.R. (Eds.), Soil Fertility and organic matteras critical components of production systems. Soil Science Society of America &American Society of Agronomy Publ, #19, Madison, WI, pp. 101–121.
Soil Survey Staff, 1999. Soil taxonomy. A basic system of soil classification for making andinterpreting soil survey, Second ed. USDA–NRCS Agriculture Handbook #436, Wash-ington, DC
Templ, M., Hron, K., Filzmoser, P., 2013. Package ‘robCompositions’. Robust estimation forcompositional data. R package version 1.6.3. http://cran.r-project.org/web/packages/robCompositions/robCompositions.pdf.
Tiessen, H., Moir, J.O., 2008. Characterization of available P by sequential extraction, In:Carter, M.R., Gregorich, E.G. (Eds.), Soil sampling andmethods of analysis, 2nd ed. Ca-nadian Society of Soil Science and CRC Press, Taylor and Francis, Boca Raton, FL,pp. 293–306.
Tiessen, H., Stewart, J.W.B., Cole, C.V., 1984. Pathways of phosphorus transformations insoils of differing pedogenesis. Soil Sci. Soc. Am. J. 48, 853–858.
van den Boogaart, K.G., Tolosana-Delgado, R., Bren, M., 2013. Compositions: composition-al data analysis in R package version 1.30–1. http://CRAN.R-project.org/package=compositions.
van der Zee, S.E.A.T.M., Fokkink, L.G.J., van Riemsdijk, W.H., 1987. A new technique for as-sessment of reversibly adsorbed phosphate. Soil Sci. Soc. Am. J. 51, 599–604.
Venables, W.N., Ripley, B.D., 2002. Modern applied statistics with S: fourth edition.Springer, New York.
Zheng, Z., Simard, R.R., Lafond, J., Parent, L.E., 2001. Changes in phosphorus fractions of ahumic gleysol as influenced by cropping systems and nutrient sources. Can. J. Soil Sci.81, 175–183.
Zheng, Z., Simard, R.R., Lafond, J., Parent, L.E., 2002. Pathways of soil phosphorus transfor-mations after 8 years of cultivation under contrasting cropping practices. Soil Sci. Soc.Am. J. 66, 999–1007.
organic and organic phosphorus: A compositional analysis witho.2014.01.030
![Biogeochemistry Chapter - Reading Materials[1]](https://static.documents.pub/doc/80x56/563db8d4550346aa9a975901/biogeochemistry-chapter-reading-materials1.jpg)
