Rapid loss of phosphorus during early pedogenesis along a glacier retreat choronosequence, Gongga Mountain (SW China)

Study area and samples collection

Hailuogou Glacier, located on the eastern slope of Gongga Mt., southwestern China (Fig. 1), has been retreating since 1890 (Li et al., 2010). A soil chronosequence has developed in the area exposed by the retreat that is approximately 2 km long, 50–200 m wide and has a 150 m altitude difference. The soil development on the chronosquence was described by He & Tang (2008). According to the World Reference Base for Soil Classification (2006), the soils are grouped as Regosols. The mineral composition of the parent materials is similar in the retreat area, including plagioclase (28.5%), quartz (24.5%), biotite (13.7%), hornblende (13.0%), K-feldspar (8.6%) and apatite (1.6%) (Yang et al., 2015).

A complete primary vegetation successional sequence has formed on the chronosequence, including (1) bare land, (2) Salix rehderiana C.K. Schneid.—Hippophae rhamnoides L.—Populus purdomii Rehder, (3) Populus purdomii Rehder, (4) Abies fabri (Mast.) Craib–Picea brachytyla (Franch.) E. Pritz., and (5) Picea brachytyla (Franch.) E. Pritz.—Abies fabri (Mast.) (Li & Xiong, 1995).

The climate on the Hailuogou chronosequence is controlled by the southeast monsoon with a mean annual temperature of 4.2 °C and a mean annual rainfall of 1,947 mm (Wu et al., 2013).

The soil samples were collected in December 2010 and July 2011 at the same plots in each site with different ages after the retreat of the glacier (Fig. 1). Three 2 × 2 m plots were created for each site, and the soil profiles were hand-dug at each plot. The soil profiles were divided into three horizons: the O, A and C horizons, except for the 12-year-old site, where the O and A horizon were absent.

At each site, the trunk, bark, latest-growth-year leaves and twig were collected and mixed to yield a composite sample from the dominant plants above the soil sampling sites. For each tree, a composite sample was collected from the upper and lower locations on the eastern, southern, western, and northern sides.

Chemical analysis

A modified Hedley fraction extraction technique (Tiessen & Moir, 1993) was used to separate the P into eight fractions (Fig. 2). Two strips of an anion exchange membrane (BDH 551642S, 9 × 62 mm) were put into the tube. The anion exchange membrane strips were converted to bicarbonate before they were used in the first step. The supernatant of Step 1–4 was shaken for 16 h at 25 °C. Thereafter, the supernatant was centrifuged at 25,000 × g for 10 min at 0 °C, and then passed through Millipore filters (pore size 0.45 µm). The concentrations of the organic P (Po) were calculated by the differences in the P concentrations between the undigested and the digested samples. The P concentrations were measured with a UV-V spectrophotometer (SHIMAZU UV 2450) using the phosphomolybdate blue method (Murphy & Riley, 1962). Blanks were mixed with the solvents to account for matrix interference during the extraction processes. The total P (Pt) concentration was measured using ICP-AES after subsamples were ground and digested with a microwave (HNO₃-HClO₄-HF).

The P fractions were defined as follows: (1) Resin-Pi (RPi, Resin-inorganic P) = exchangeable P, (2) NaHCO₃-P = labile Pi and Po, (3) NaOH-P = moderately bioavailable P, mainly Pi and Po tightly adsorbed and/or fixed by Al and Fe hydroxides, and P in humic and fulvic acids; (4) DHCl-Pi = Pi in primary minerals (mainly apatite); (5) CHCl-P = extractable P by concentrated HCl, refractory P (Tiessen & Moir, 1993).

Leaf, trunk, bark and twig were brushed clean and rinsed with deionized water before drying and grinding to a fine powder, respectively. The Pt concentrations of plant samples were measured by ICP-AES after microwave digestion (HNO₃-H₂O₂-HF). Values are expressed on a 60 °C dry-mass basis. The P concentration of plant samples was reported as the average concentration of leaf, trunk, bark and twig.

Estimation of P loss

Because the largest thickness of A horizons was <12 cm in the Hailuogou chronosequence, P loss in the soils with a depth of 30 cm was evaluated. The P loss (P_LOSS) was calculated as following equation: (1)\begin{eqnarray} \rm P_{LOSS} = Stock P_{C30} - Stock P \end{eqnarray} where Stock-P_C30 (kg/ha) is total P stock in the C horizon with a thickness of 30 cm and Stock-P is the sum of P stock in the O horizon (Stock-P_O) (kg/ha)and the surface mineral (Stock-P_AC30, 0–30 cm) (kg/ha) soils. (2)\begin{eqnarray} \rm Stock P_{C30} = C_{Pc} \times BD_{c} \times 30 \div 10 \end{eqnarray} (3)\begin{eqnarray} \rm Stock P = Stock P_{O} + Stock P_{AC30} \end{eqnarray} (4)\begin{eqnarray} \rm Stock P_{O} = C_{Po} \times BD_{o} \times D_{o} \div 10 \end{eqnarray} where C_Po (mg/kg), BD_o (g/cm³) and D_o (cm) is the Pt concentration, bulk density and thickness in the O horizon, respectively. (5)\begin{eqnarray} \rm Stock P_{AC30} = (C_{PA} \times BD_{A} \times D_{A} + C_{PC} \times BD_{C} \times (30 - D_{A})) \div 10 \end{eqnarray} where C_PA (mg/kg), BD_A (g/cm³) and D_A (cm) is the Pt concentration, bulk density and thickness in the A horizon, respectively; C_PC (mg/kg) and BD_C (g/cm³) is the Pt concentration and bulk density in the C horizon, respectively.

Was P lost during the 120 years of pedogenesis?

The ratios of the Pt concentrations in the A horizon to that in the C horizon partly revealed the continuous loss of P with soil age. The ratio of Pt_A/Pt_C was 0.62 and 0.74 at the 80- and 120-year-old site, respectively. The ratios were much lower than those of the chronosequences of similar soil ages. There were no significant decreases of total P at the Morteratsch chronosequence (150 years) and the Damma chronosequence (120 years) in the Swiss Alps (Egli et al., 2012; Prietzel et al., 2013). Meanwhile, these ratios were also lower than or similar with those of several older chronosequences. The Pt_A/Pt_C ratio at the 5,000-year-old site of the Franz Josef Chronosequence was approximately 72.7% (Walker & Syers, 1976). In a dune chronosequence in New Zealand, the ratio at the 370-year-old site was 59.5% (Eger, Almond & Condron, 2011). There was no obvious change in the total P at a 7,800-year-old chronosequence in Northern Sweden (Vincent et al., 2013). Lower Pt_A/Pt_C ratios implied the severe loss of Pt in the soil under the climax community, and the loss rate was faster than other soil chronosequences.

The P stock in the O horizons and surface mineral soils (0–30 cm) directly showed the rapid loss of P on the Hailuogou chronosequence (Table 2). At the 30 and 40 year sites, P loss was not observed; while from the 52 to the 120 year site, the P loss increased with soil age. At the 120 year site, ∼17.6% Pt in the soils (thickness: 30 cm) was lost (Table 2). This is a relative high loss rate compared with other chronosequences with similar or greater ages (Table 2). The P loss rate at the 120 year site on the Hailuogou chronosequence was ∼7 times of that on the Rakata chronosequence, which was ∼110 years old (Table 2). Although the P loss rate on the Hailuogou chronosequence was lower than those on some ‘old’ chronosequences (Table 2), the P loss could not be observed until these ‘old’ chronosequences had been developed for thousands of years. For example, Galván-Tejada et al. (2014) found that P stocks started to decline after 2,185 years of soil development on the Transmexican Volcanic Belt chronosequence. Moreover, Turner et al. (2013) found that the P stock (TP in soils < 2 mm) was 115.0, 194.2, 169.6 and 58.0 g P/m² at the 5, 1,000, 12,000 and 120,000 year site on the Franz Josef chronosequence, respectively, suggesting P stock did not decrease sharply at the initial stage of pedogenesis.

The pathway of P loss along the soil chronosequence

During the early stage of pedogenesis, P could be released from the parent rocks into the soil by weathering. On the Hailuogou chronosequence, the degree and rate of weathering changed with soil age (Zhou, 2014). The weathering rate at the 120-year-old site was 111 cmol_c/m² yr, which was higher than those in Alps and some tropical zones (Egli, Fitze & Mirabella, 2001; Taylor & Blum, 1995). The rapid weathering process could lead to the dissolution of apatite bound to carbonates, biotite, hornblende, plagioclase and other silicate minerals. The depletion factor of soil P on the Hailuogou chronosequence demonstrated that only approximately 22% of the original apatite was remained after 120 years of weathering and pedogenesis (Zhou, 2014). The rapid dissolution of apatite during the weathering process should be ascribed to the fine materials that resulted from strong glaciation and freeze-thaw processes, adequate moisture and sharp decrease of pH (Zhou et al., 2013) due to rapid vegetation succession.

P released from minerals as phosphate was available for plant assimilation. Along the Hailuogou chronosequence, the primary vegetation succession was quickly established with the weathering and pedogenesis. The increase of biomass, as well as P in biomass, was also prominent along the chronosequence (Table 2). The pool of P in biomass in the forest at the 120 year site was 303.9 kg/ha, which accounted for ∼36% of P loss (Table 2). Therefore, plant uptake was a significant pathway of P loss from the soil. Moreover, the lower concentrations of bio-available P in July (Fig. 4), the growing season, confirmed that plant uptake accelerated the loss of P from the soil.

Besides of the P assimilated by plant, about 540 kg P per ha was lost on the chronosequence. This suggested that in addition to plant uptake, there should be other paths for P loss from the soil on the Hailuogou chronosequence.

As P was released from parent rocks, Fe, Al, Ca and other metal ions were released as well during weathering and pedogenesis. Fe and Al ions were likely to form hydroxides. In addition to being assimilated by plants, the released phosphate tended to be adsorbed onto the surface of Fe and Al hydroxides. In the NaOH extracted solutions, the P concentrations were significantly positively related with Fe and Al concentrations (Fig. 5). This result implied that the P extracted by NaOH was from Fe and Al hydroxides that bound P, which accounts for more than 30% of the Pt at the 52, 80- and 120-year-old site in December, while it was a little lower in July. A previous study in dark coniferous forests showed that the P bound by Al and Fe was the major fraction of soil P (Wood, Bormann & Voigt, 1984). Kaňa & Kopáček (2005) confirmed that the adsorption capacity of Fe and Al hydroxides was the dominant factor controlling the transport of soil P to the water body. Šantrůčková et al. (2004) demonstrated that the lower acidity of the forest soil drove the transformation of P and changed the P loss rate. As the vegetation established and evolved on the Hailuogou chronosequence, the soil pH decreased with the soil age (Zhou et al., 2013). In the acidified soils, especially at the 80- and 120-year-old sites, the P bound by Fe and Al was potentially transported by runoff. The annual precipitation was concentrated in the summer (Wu et al., 2013). The P bound by Fe and Al could be more significantly discharged by runoff in the summer than in the winter, which could be interpreted as the reason for the lower contribution of NaOH-P to the Pt in July than in December. Meanwhile, the elevation difference of 150 m along less than 2 km of the Hailuogou chronosequence, led to the strong erosion and great loss of soil P. The mean concentration of Pt in the Hailuogou glacier waters was 0.049 ± 0.014 mg/l, according to the Alpine Ecosystem Observation and Experiment Station of Mt. Gongga, Chinese Academy of Sciences (2002–2005). The annual runoff was 307.1 m³/yr ha in the Hailuogou glacier watershed (Li et al., 2004). The P discharge rate was calculated as 1.5 kg P/yr ha.