General Soil Characterization
Despite their high total P contents (2076 and 1223 mg kg-1), sb2.1 and PV2A horizon samples had Mehlich III–extractable P (M3P) contents of less than 70 mg kg-1 (Table 1). According to previous Québec fertilizer recommendations (Conseil des Productions Végétales du Québec, 1996), the P fertility levels of sb2.1, Ma2, and PV2 soils would be considered respectively as low, high, and adequate. The lower available P content of the sb2.1 sample may be related to its high P sorption capacity as estimated by the Alox + Feox content (Table 1). For a given horizon, soils from the Beaurivage River watershed have greater P sorption capacities, as indicated by greater levels of Alox + Feox, than soils from the lowland area (Table 1; Ma2 vs. PV2, or Ma3 vs. AI2).
Sequential Phosphorus Fractionation
Chemical fractionation provided information on operationally defined P pools of varying solubilities (Table 2). For all samples, the largest P fraction was found in moderately labile (NaOH-P) or nonlabile fractions (HCl-P and H2SO4–P). Fractionation data for the A horizon samples suggested that P was mainly associated with Al or Fe oxides in the acidic sb2.1 and Ma2 soils (up to 55% of Pt as NaOH-P), whereas apatite-type minerals represented the major forms of P in the slightly alkaline PV2 sample (40% of Pt as HCl-P). In both B horizon samples, HCl-P was the main pool, suggesting that Ca-bound P minerals accounted for 71% of Pt in the calcareous AI2 sample and for 50% of Pt in the acidic Ma3 sample. Labile P (resin P + NaHCO3–Pt) accounted for 14 to 18% of total P in A horizon soils, whereas this pool was less than 10% for the B horizon samples. On average, 30% of the labile P in A horizons was organic (NaHCO3–Po). For B horizons, 40% of labile P was organic P. Moderately labile organic P (NaOH-Po), a fraction considered to be associated with humic compounds, represented between 13 and 21% of total P in most soils, except for the calcareous AI2 B horizon, which had only 3% of total P as NaOH-Po.
X-Ray Absorption Near-Edge Structure Spectroscopy Results
Details of P K-XANES spectra for a number of phosphate standards were discussed by Hesterberg et al. (1999). A subset of representative spectra are presented here to illustrate the range of spectral features for standards. As shown in Fig. 1A , spectra for Fe-phosphate minerals and PO4 adsorbed on Fe-oxides had a pre-edge feature between -5 and -2 eV [relative to the P(V) K-edge], which increased in intensity with increasing mineral crystallinity. The standards of PO4 adsorbed on ferrihydrite or goethite were further characterized by an intense white line peak. Their very similar XANES spectra indicate that these two adsorbed species may not be distinguishable from each other when fitting spectra of soils. Calcium-phosphate mineral standards all exhibited a shoulder on the high-energy side of the absorption edge, between 2 and 6 eV (relative energy; Fig. 1B). The structure of the shoulder varied among the different Ca-phosphate minerals. For example, hydroxyapatite and octacalcium phosphate had well-defined shoulders compared with monetite (Fig. 1B) and brushite (CaHPO4·2H2O; not shown). The XANES spectra for Al-phosphate minerals showed a weak pre-edge inflection at about -1 eV (relative energy). This feature was better defined on the first-derivative spectra (data not shown) and was more distinct for crystalline (variscite, berlinite) than noncrystalline Al-phosphate (see Fig. 1C for variscite). Similar to what was observed with PO4 adsorbed on Fe-oxides, PO4 species adsorbed on Al hydroxide or alumina had an intense white line peak but no pre-edge feature (Fig. 1C). The XANES spectra for ATP showed an inflection at 1 eV, similar to that of Al-phosphate minerals (data not shown). The spectrum for IHP was mostly featureless and had a broad white line peak (Fig. 1C).
Principal component analysis performed on the normalized K-XANES spectra of the five soil samples showed two significant orthogonal components at = 10%. Target transformation retained most of the standards as likely species except for strengite, variscite, and amorphous iron phosphate that had unacceptable SPOIL values of >8 and probabilities of F values of When using three orthogonal components, all standards came out as likely species. These results indicated that the first PCA step lacked sensitivity and that target transformation could not discriminate well the most likely species among our set of standards. Because target transformation is an oblique rotation, the selected targets may be correlated (Beauchemin et al. 2002), and the fact that most standards were potential targets suggested correlation among our standards. For this reason and because of lack of sensitivity observed with PCA results, LCF was used to achieve the best characterization of our soil samples using all standards. The least-squares fitting procedure was not restricted to two standards only (based on the number of orthogonal significant components identified in the first step via PCA), and a maximum of three standards was allowed in the fitting.
The XANES spectra and least-squares fits for each soil sample are illustrated in Fig. 2 . Table 3 reports the relative normalized proportions of each phosphate species in the soil as determined by fitting each soil spectrum as a linear combination of standard spectra. The goodness of fit indicated by 2 was typically (Table 3). The sum of fractions before normalization can also indicate, to some extent, the goodness of fit, as the individual component should ideally sum to 1 within the experimental error. Good fits can still be obtained with a sum as low as 0.6 to 0.7, as it is the case for AI2B (Table 3), but then the origin of the deviation should ideally be investigated (Manceau et al., 2000). For the data in Table 3, the greatest deviation (30%) from the ideal sum of 1 was obtained for samples sb2.1 and AI2B.
The P K-XANES fitting results for the PV2-A and AI2-B samples indicated that more than one best combination could be fitted for these samples (Table 3
). As was discussed earlier in reference to Fig. 1A
, the spectrum of PO4
adsorbed on ferrihydrite could not be easily distinguished from that of PO4
adsorbed on goethite. However, standards of PO4
adsorbed on Fe-oxides should be distinguishable from species adsorbed on Al-oxides, based on the weak but typical pre-edge feature in Fe-containing species (Fig. 1A
vs. 1C). The fitting results for PV2A and AI2B demonstrated that PO4
adsorbed on Fe-oxides could not be reliably distinguished from PO4
adsorbed on Al-oxides for the soil data. However, adsorbed species could be distinguished from other P minerals due to the characteristic intense white line peak in their spectra. Hence, we regrouped in Table 3
adsorbed species under the general term "adsorbed on Fe- or Al-oxides" and averaged the proportions obtained from the best reported combinations.
The K-XANES fitting results indicated that phosphate adsorbed on Fe- or Al-oxide minerals was present in all soil samples, but in greater proportion for the three acidic soil samples (>44%) than for the slightly alkaline PV2-A and AI2-B soils (Table 3). The greatest proportion of adsorbed P was found in sample sb2.1-A (88% of total P). This can be seen by an intense white line peak near 1 eV in the spectrum of this sample and a weak pre-edge feature near -3 eV that would reflect the presence of PO4 adsorbed on Fe-oxides (Fig. 2A). The XANES data suggested that the acidic Ma2 soil sample contained 22% of P as poorly crystalline iron phosphate (Table 3). Overall, the proportions of all P species associated with Fe or Al (adsorbed PO4 on either Fe- or Al-oxides + Fe-phosphate) determined from the XANES spectral fitting were significantly correlated with proportions of NaOH-extractable Pi (r = 0.99, p = 0.001, n = 5; Fig. 3B) , although proportions determined by XANES fitting tended to be greater than those obtained by chemical fractionation. This overestimation could be due to the fact that we restricted the XANES fitting to three standards, or to a lack of specificity in the chemical fractionation.
The XANES spectra of all soil samples exhibited a shoulder on the high-energy side of the white-line peak, indicating that some form of calcium phosphate was present in all samples (Fig. 2A–E
). Fitting results suggested that hydroxyapatite occurred in all soils but in lower proportions in acidic ( in neutral to slightly alkaline soil samples (>25%; Table 3
). This trend is consistent with the increased solubility of hydroxyapatite under acidic conditions (Lindsay, 1979
). The calcareous AI2-B sample contained the greatest proportion of hydroxyapatite (59%), as can be seen from the pronounced shoulder on the high-energy side of the white line in its spectrum (Fig. 2E
). For the two slightly alkaline soils (PV2-A and AI2-B) and for the acidic soil sample for which the Ca-phosphate appeared as a major sink for P (Ma3-B; Tables 2
), a significant proportion of octacalcium phosphate was also detected. The correlation between the proportion of HCl-extractable P ("apatite-like P") and the summed proportions of all Ca-phosphate species determined by XANES fitting was significant (r
= 0.87, p
= 0.05, n
= 5; Fig. 3A
), and showed nearly a 1:1 relationship between these two measurements.