Introduction

Phosphorus is both an essential and a frequently limiting nutrient for crop growth (Raghothama 1999; Vance 2003). Farmers routinely apply P-containing fertilizers to increase plant-available soil P concentrations and enhance crop yield, but there are increasing concerns regarding its use efficiency (Hinsinger 2001; Vance 2003). Moreover, application of P fertilizers in highly-intensive agriculture can be problematic, because the soils have accumulated substantial amounts of P due to excessive historical P applications (Vance et al. 2003; Sattari et al. 2012). Phosphorus losses from these enriched soil can greatly increase the risk of pollution of aquatic environments (Carpenter et al. 1998). Hence, it is essential to develop P-management strategies to effectively use accumulated soil P, one of which is to increase soil P bioavailability through growing plant species with the ability to scavenge recalcitrant soil P

Richardson et al. (2011) recognize three strategies of plants exploiting soil P resources: (i) “root-foraging,” which is helpful for acquiring more available soil P through a greater exploration of the surface soil; (ii) “P-mining,” which improve the desorption, solubilization, or mineralization of sparingly available inorganic P (Pi) and organic P (Po) pools by using root exudates such as organic anions and phosphatases; and (iii) “improving internal P-utilization efficiency,” which increases plants yield per unit of P uptake. Although plants can use any of these strategies to acquire soil P, crop species differ in their ability to access P from soil P pools (Simpson et al. 2011). While many crops lack the ability, a limited number of species can acidify the rhizosphere, exude carboxylates, or release phosphatases to acquire sparingly-available P (Gerke 1994; Li et al. 2007; Li et al. 2014; Wang and Shen 2019).

Several legume crops such as white lupin (Lupinus albus L.), faba bean (Vicia faba L.), and chickpea (Cicer arietinum L.), exude large amounts of organic compounds that mobilize sparingly-available soil P (Veneklaas et al. 2003; Cu et al. 2005; Li et al. 2007). For example, faba bean releases protons, malate, and citrate into the rhizosphere to mobilize insoluble Al- and Fe-bound P in acid soils and Ca-bound P in calcareous soil (Li et al. 2007). Chickpea can effectively access Po through exuding acid phosphatases to hydrolyze Po into Pi (Li et al. 2003). Additionally, these legumes also potentially promote the growth of intercropped or subsequent cereal crops (Li et al. 2003, 2007; Nuruzzaman et al. 2005). The ability of legumes to secrete P-mobilizing substances to accesses more recalcitrant P sources indicates that they have a potential for enhancing P acquisition of P-inefficient cereal species when they share the rhizosphere in an intercropping system (Dissanayaka et al. 2015).

In several intercropping systems, interspecific facilitation increases P acquisition of cereal species when planted with P-mobilizing legumes. For example, white lupin mobilizes P from insoluble soil P pools to enhance P uptake of intercropped wheat (Triticum aestivum L.) (Gardner et al. 1983; Horst and Waschkies 1987; Cu et al. 2005). The same effect has been recorded for groundnut (Arachis hypogaea L.) and intercropped maize (Ae et al. 1996; Ae and Otani 1997; El Dessougi et al. 2003). Pigeon pea (Cajanus cajan L.) exudes piscidic acid and a p-O-methyl derivative that chelates Fe3+, effectively utilizing Fe-bound P and increasing P acquisition when intercropped with sorghum (Ae et al. 1990). In addition to the above interspecific facilitation, complementarity between intercrops also plays an important role in enhancing P uptake of both intercrops. Complementarity can occur in an intercropping system if two intercrops access different soil horizons (Li et al. 2006; Zhang et al. 2014) or different P pools (Cu et al. 2005; Li et al. 2008), and competition for nutrients is consequently mitigated.

The changes in soil P fractions after addition of fertilizer-P have been well documented to predict the response of crops to P fertilizers (Sharpley 1985; Beck and Sanchez 1994). Generally, at low soil P availability, mineralization of Po pools is the main source of P to crops (Tiessen et al. 1984; Beck and Sanchez 1996). With the addition of fertilizer-P, the amount of P that is not retained in the soil solution or removed by the crop, is transformed into less labile P pools (Schmidt et al. 1996). However, there are few studies on how crops use those less labile P pools. Phosphate rock, the source of P fertilizer, is gradually being depleted and becoming one of the national strategy materials. Hence, utilizing soil-accumulated P and then using less P fertilizer is one of the key issues to achieve sustainable agriculture. The present study used a long-term field experiment located in the North China Plain, where the calcareous soil at the site causes a strong retention of P on to the solid soil phases (Hu et al. 2012). The soil at the site was over-supplied with P fertilizer for four years until the end of the season in 2012. Since then, the soil did not receive nitrogen (N) and P fertilizer for four years. Soil samples were collected, and the P fractions were analyzed after one and four years of P depletion by the following three crop treatments: sole maize, sole faba bean, and intercropping of maize/faba. We hypothesized that (i) over-application of P fertilizer resulted in a greater P accumulation in the legume-only crop than in the cereal-only crop, because these crops differ substantially in their P demand and efficiency in acquiring labile P. We also hypothesized that, over the three-year depletion period, (ii) faba bean would primarily deplete the labile P and less labile P pools by mobilization, and that maize would rely largely on the labile P pools, without exploiting less labile P pools. We hypothesized that in the intercropped system (iii) the depletion of both P pools would be intermediate and expected that intercropped maize/faba bean would increase P utilization compared with that of sole cropping.

Materials and methods

Study area

A long-term field experiment was started in 2009 in Quzhou County (36.93 N, 115.17E; 40 m a.s.l.), Hebei Province, China. The study location has a typical monsoon climate, with an annual mean temperature of 13.1 °C and an annual precipitation of 556.2 mm. At the start of the study in 2009, the soil at the experimental location was a calcareous alluvial soil with a pH of 7.3 (2.5:1 water/soil) and contained 14 g kg−1 organic matter, 0.84 g kg−1 total N, 12.6 mg kg−1 Olsen-P, and 211 mg kg−1 exchangeable potassium (K). The methods used for the above analysis follow Bao (2000).

Since 2009, the field had been fertilized with superphosphate at 80 kg P ha−1, urea at 180 kg N ha−1, and potassium sulfate at 50 kg K ha−1. All the P and K fertilizer, as well as half of the total N, were broadcasted and incorporated into the upper 20 cm of the soil prior to planting by using a rotary tiller. The remaining N fertilizer was side-dressed only for maize at V12 stage. The soil Olsen-P reached 16.4 mg P kg−1 at the end of the season in 2012. In order to fully exploit the biological potential of crops to unlock the accumulated soil P and increasing the use of recycled P sources, both N and P fertilizer applications were stopped in 2013, while K fertilizer input was maintained.

Experimental design

The field experiment was a completely random design with three replicates. The treatment was cropping systems: sole maize (Zea mays L. cv. Zhengdan No. 958), sole faba bean (Vicia faba L.cv. Lincan No. 5), or intercropped maize/faba bean. The field was split into 9 plots, each measuring 4.2 m wide and 8 m long, with an area of 33.6 m2. One cropping system (sole maize, sole faba bean, or intercropped maize/faba bean) was assigned to each plot. Either maize or faba bean was used to maintain a consistent density in-between the plots. For maize, the row distance was 0.4 m, plant spacing was 0.3 m, and the density was 8.33 plants m−2. For faba bean, the row distance and plant spacing were both 0.2 m, and the plant density was 25 plants m−2. Thus, each plot of sole maize or faba bean contained 10 or 21 rows plants, respectively. The maize/faba bean intercropped plots were a replacement design, in which both row distance and plant spacing were identical to those of the corresponding sole crop. The row distance between maize and faba bean was 0.3 m. Each intercropped plot consisted of three strips, each containing two rows of maize grown alternately with three rows of faba bean. Maize strips were 0.8 m wide, and faba bean strips were 0.6 m wide, accounting for 57% and 43% of a whole strip area, respectively (Fig. 1a-c).

Fig. 1
figure 1

Layouts of the three cropping systems and soil sampling sites: (a) sole maize; (b) sole faba bean; (c) intercropped maize/faba bean; the soil sampling sites in (d) sole maize, (e) sole faba bean, or (f) in the strips of intercropped maize/faba bean. In a-c, each panel shows an area of 4.2 m width and 8 m length. There were 10 maize rows (0.4 m distance) over this width in panel a (sole maize) and 21 faba bean rows (0.2 m distance) in panel b (sole faba bean). In panel c (intercropped maize/faba bean), there were three strips of two maize rows (at 0.4 m row distance) and three faba bean (at 0.2 m row distance) each, and two paths between the strips, each measuring 0.3 m. There were three placements of soil samples in (d) sole maize from under the center of the maize row at 0.1 m intervals and (e) sole faba bean from under the center of the faba bean row at 0.1 m intervals. There were six placements of soil samples in (f) intercropped maize/faba bean from under the inter-row of maize at 0.1 m intervals

Full size image

All plots were irrigated during the growing season to prevent water stress and weeded manually. Each year from 2013 to 2016, faba bean was sown on 7–11 March and harvested on 3–6 July; maize was sown on 25–28 April and harvested on 7–11 September. The field was fallowed in winter. From 2013 to 2016, the annual mean temperature was 13.5 °C, 15.7 °C, 14.7 °C and 13.4 °C, respectively, and the annual precipitation was 555 mm, 364 mm, 407 mm and 424 mm, respectively.

Collection and measurement of plant and soil samples

At crop maturity each year, the aboveground biomass of four representative faba bean and two maize plants was harvested from each plot, and used for measurements of nutrient concentrations. The middle strip of each plot was used exclusively for grain yield monitoring when the crop was mature. In the yield-monitoring area, all cobs of two rows of maize with adjacent 10 plants in each row or all pods of three rows of faba bean with 20 adjacent plants in each row were collected, and then the grains from the cobs or pods were manually separated after air drying. All plant samples were oven dried at 65 °C for a minimum of 48 h and weighed. After digestion in a mixture of concentrated H2SO4 and H2O2, P concentrations were determined using the molybdovanadophosphate method by spectrophotometry (UV757T, Shanghai Instrument Co. Ltd., Shanghai, China) (Johnson and Ulrich 1959).

As with most P-related studies, we focused only on the topsoil (0–20 cm), since this layer is recognized to be the most relevant for crop growth (Blake et al. 2003; Wang et al. 2007; Soltangheisi et al. 2018; Sucunza et al. 2018; Jimenez et al. 2019). Before the crop was sown in March, 2013, soil samples at each replica plot were collected at depths of 0 to 20 cm to determine the chemical characteristics in the plot for sole maize, sole faba bean, or intercropped maize/faba bean (Table 1). In September, at the end of 2013 (after one year of no P fertilizer, Stage I) and 2016 (after four years of no P fertilizer, Stage II) growing seasons, soil samples at each replica plot from the uppermost 20 cm were collected using a 10 cm diameter soil auger. In order to collect representative soil samples, we took soil cores separately at the upper, middle, and lower site of a plot’s diagonal line to form composite samples. Each site included three or six cores at 10 cm intervals for the sole and intercropped system, respectively, as shown in Fig. 1d-f. All soil samples were air-dried and passed through a 2-mm sieve prior to analysis. The samples were used for soil P sequential fractionation (Fig. 2).

Table 1 Chemical characteristics of the top 20 cm soil in the different cropping systems in March 2013 before the crops were sown
Full size table
Fig. 2
figure 2

Hedley sequential soil phosphorus (P) fractionation method (Tiessen and Moir 1993). Pi and Po, represent inorganic and organic P, respectively; conc. HCl indicates concentrated HCl

Full size image

Soil P fractionation

The method reported by Tiessen and Moir (1993) has been widely used for quantifying various empirically-defined pools of soil P. Briefly, soil (0.5 g) in a 50 mL centrifuge tube was sequentially extracted as follows. (1) To the tube we added a 25 × 62.5 mm anionic resin strip and then 30 mL deionized water, and this was shaken for 16 h at 200 r/s; then the resin trip was placed in a clean 50 mL tube and shaken with 20 mL 0.5 M HCl for 16 h (Resin-P). (2) The water in soil-containing tube was discarded and 30 mL 0.5 M NaHCO3 at pH 8.5 was added, and the suspensions was then shaken for 16 h (NaHCO3-P). (3) 30 mL 0.1 M NaOH was added, and shaken for 16 h (NaOH-P). (4) 30 mL 1 M HCl was added, and shaken for 16 h (1 M HCl-P). (5) The soil residue with 10 mL concentrated HCl was heated at 80 °C in a water bath for 10 min, removed from the bath and then 5 mL of concentrated HCl was added, and the volume adjusted to 50 mL with deionized water (conc. HCl-P). (6) The soil residue was digested with 5 mL concentrated H2SO4-H2O2 at 350 °C for 3 h (Residual-P). In order to completely recover the soil sample, in between two consecutive steps during processing, the suspensions was centrifuged for 10 min at 25,000×g and 0 °C, and the supernatant passed through a 0.45 μm membrane filter afterwards, and then the filter was rinsed with the next step extractant. Inorganic P (Pi) concentrations for all the extracts were determined within 24 h using the ascorbic acid-molybdenum blue method (Murphy and Riley 1962). The total P (Pt) concentration in the different extracts (NaHCO3-P, NaOH-P and conc. HCl-P) was determined by ammonium persulfate digestion. Organic P (Po) concentrations were calculated as the difference between Pt and Pi. We duplicated soil testing once every five samples, but we did not test all samples twice, because of the heavy workload of soil P fractionations. In addition, each treatment in the field experiment included three replicates, and soil samples were taken in all plots, so soil sample had three replicates. Hence, we used those two different kinds of replicates to control data quality.

Calculations

  1. (1)

    The Land Equivalent Ratio (LER) is often considered as an indicator of intercropping benefit (Willey 1979). The LER was calculated according to:

    $$ \mathrm{LER}=\frac{Y_{if}\ }{Y_{sf}}\times {P}_{if}+\frac{Y_{im}}{Y_{sm}}\times {P}_{im} $$

where Yif and Yim are the P uptake per unit area of intercropped faba bean and maize, respectively. Ysf and Ysm are the P uptake per unit area of sole faba bean and maize, respectively, and Pif (43%) and Pim (57%) are the relative density of faba bean and maize, respectively, under intercropping relative to sole cropping. If the LER is greater than 1, this indicates the effect of facilitation is greater than that of competition, and there is an intercropping P-uptake advantage.

  1. (2)

    Soil P balances were calculated for the experimental period as the difference between P inputs and outputs. For the calculation, mean annual values for all replicate plots within each treatment were summed. Phosphorus input was annually estimated from P-fertilizer application. Phosphorus output was P removed with the harvesting of the crop. Phosphorus removal with crops was calculated by multiplying the mean annual shoot biomass of the crop and the shoot P concentration of the crop according to Maltais-Landry et al. (2016). Phosphorus stocks in soil were calculated by multiplying the mean Pt concentrations at the particular soil layer of the treatment, soil bulk density (1.3 g cm−3), and thickness of the soil layer (20 cm).

  2. (3)

    Changes in soil P concentration over three years were calculated as the difference in P concentration after a four-year period of no P fertilizer application and after one year of no P fertilizer application.

  3. (4)

    The relative contribution of each P pool to total P reduction was calculated as (Hassan et al. 2012a).

    $$ \mathrm{Relative}\ \mathrm{contribution}\ \left(\%\right)=\frac{P_a-{P}_b}{\Delta P}\times 100 $$

where Pa and Pb are P concentration of the pool after 1-yr and 4-yr periods of no P fertilizer application, respectively. ∆P is the Pt reduction.

Statistics

Analysis of variance (ANOVA) was conducted using the SAS software package (SAS v.8.0). Shoot P-uptake data were analyzed using two-way ANOVA, with years (2013, 2014, 2015, and 2016) and cropping systems (sole crop vs intercrop) as the treatment effects. Soil P fractions were analyzed using two-way ANOVA, with years (Stage I and Stage II) and cropping systems (sole maize, sole faba bean and intercropped maize/faba bean) as the treatment effects. The chemical characteristics of soil and changes of P fractions were analyzed using one-way ANOVA (with the three cropping systems as the treatment effect).

Results

Shoot P content and P balance

From 2009 to 2013 (Phase I), the total fertilizer P input was 320 kg P ha−1, and the P output of sole maize, sole faba bean and intercropped maize/faba bean system was 222, 183, and 228 kg P ha−1, respectively, resulting in 98, 137, and 92 kg P ha−1 surplus, respectively (Table 2). The P surplus increased soil Olsen-P from the original 12.6 mg P kg−1 in 2009 to 16.4 mg P kg−1 (the mean of 17.1, 16.2 and 16.0 mg P kg−1 in the Table 1) in 2013. From 2014 to 2016 (Phase II), when there was no additional P fertilization, P removal was decreased by 61% (sole maize), 62% (sole faba bean), and 54% (intercropping), compared with Phase I, and leading to the P depletion of 87, 70, and 104 kg P ha−1 for sole maize, sole faba bean, and intercropped faba/maize, respectively.

Table 2 The phosphorus (P) balances in the plow layer in the different cropping systems for the period 2009–2016
Full size table

From 2013 to 2016, when the LER ranged from 1.2 to 1.5, the maize/faba bean intercropping system had a P-uptake advantage compared with the sole cropping, and average P content of the shoots in the intercropped maize and faba bean plots was significantly increased (29% and 30%, respectively) compared with the corresponding sole crops (Table 3). After one (2013) and two years (2014) of no P fertilizer, the intercropped maize had significantly greater shoot P content (34% in both years), and intercropped faba bean significantly increased shoot P content by 15% and 67% in 2013 and 2014, respectively, compared with the corresponding sole crops. However, after three (2015) and four years (2016) of no P fertilizer, we did not observe a significant intercropping advantage, because the standard error of P content values increased with planting year.

Table 3 Shoot phosphorus (P) content (kg ha−1) of faba bean and maize under different cropping systems.
Full size table

Dynamics of soil P fractions

The dominant P fraction of the soil was 1 M HCl-Pi, which was more than 70% of the total P, followed by conc. HCl-Po, which was less than 10% of the total P. Resin-P, NaHCO3-Pi, and NaOH-Pi were the smallest fractions, making up 1–2% of the total P (Table 4). The impact of cropping system on P fractions depended greatly on the depletion duration. After one year of no P fertilizer input (Stage I), the primary P fractions in sole maize, sole faba bean, and intercropped maize/faba bean were 1 M HCl-Pi, followed by conc. HCl-Po, Residual-P and conc. HCl-Pi (Table 4). Among the three cropping systems, both sole maize and maize/faba bean intercropping had greater 1 M HCl-Pi and conc. HCl-Po than sole faba bean, but sole faba bean had a greater NaHCO3-Po.

Table 4 Phosphorus (P) sequential fractionation for soil after one and four years of no P fertilizer application following different cropping systems
Full size table

After four years without P fertilizer (Stage II), the dominant P fraction was 1 M HCl-Pi, which made up 70% or more of the total P. This was in accordance with Stage I. However, the second most dominant P fraction differed between crops (Table 4). Both sole maize and sole faba bean contained a Residual-P of up to 10% of the total P, but intercropped maize/faba bean had conc. HCl-Pi of up to 9% of the total P. Among the cropping systems, maize/faba bean intercropping had greater NaOH-Pi, 1 M HCl-Pi, and NaOH-Po, but a lower conc. HCl-Po than sole crop. Each cropping system generated larger changes to the soil P fractions at Stage II than in Stage I.

The Pt concentration at Stages I and II ranged from 968 to 1041 mg kg−1 and from 946 to 1008 mg kg−1, respectively (Table 4). At Stage I, the total P fractions across the three cropping systems averaged: 82% in Pi forms, 12% in Po forms, and 6% in Residual-P form. At Stage II, those P fractions averaged: 82% in Pi forms, 9.7% in Po forms, and 8.7% in Residual-P form. Over the three years of continuous P depletion, the total soil P declined by 26 mg kg−1 for sole maize, 22 mg kg−1 for sole faba bean, and 33 mg kg−1 for maize/faba bean intercropping (Table 4). These values converted, respectively, to 68, 57, and 86 kg P ha−1 (due to soil density of 1.3 g cm−3 and soil depth of 20 cm), accounting for nearly 78%, 82%, and 83% of the P balance (Table 2).

The changes in P fractions between Stages I and II (over the three-year continuous P depletion) revealed that continuously no P fertilizer input significantly depleted Resin-P, NaHCO3-Pi, 1 M HCl-Pi, NaOH-Po, conc. HCl-Po and Pt, but significantly increased conc. HCl-Pi and Residual-P (Table 4, S2 and Fig. 3). The changes in each P fraction mentioned above (except for Residual-P, NaOH-Po and conc. HCl-Po) had no significant difference between cropping systems, differing only in the amount (interactions between years (Stage I and Stage II) and different cropping systems: P > 0.05; Fig. 3 and Table S2). However, differences in other P fractions (NaOH-Pi, Residual-P, NaHCO3-Po, NaOH-Po and conc. HCl-Po) between Stages I and II depended on cropping system (interactions between years (Stage I and Stage II) and different cropping systems: P ≤ 0.05; Fig. 3 and Table S2). Sole maize significantly accumulated NaHCO3-Po and Residual-P, while significantly depleting NaOH-Pi and conc. HCl-Po, and sole faba bean significantly accumulated NaOH-Pi and Residual-P while significantly depleting NaHCO3-Po and NaOH-Po. Maize/faba bean intercropping significantly accumulated NaOH-Pi and NaHCO3-Po, and significantly depleted conc. HCl-Po, similar to sole maize (Fig. 3 and Table 4).

Fig. 3
figure 3

Changes in soil phosphorus (P) fractions after one and four years of no P fertilizer application (after a four-year period of no P fertilizer application minus measurements collected at the end of one year of no P fertilizer application) following different cropping systems

Full size image

The relative contribution of each P fraction’s changes to the total reduction differed among cropping systems (Table 5). The moderately labile P pools (sum of NaOH-Pi, 1 M HCl-Pi, and NaOH-Po) were the main contributors, accounting for 45–69% in the reduction of Pt for the three cropping systems. Second, the labile P pools (sum of Resin-P, NaHCO3-Pi, and NaHCO3-Po) accounted for 30% in sole faba bean, and the sparingly labile P pools (sum of conc. HCl-Pi and conc. HCl-Po) accounted for 38% (sole maize) and 42% (maize/faba intercropping).

Table 5 The relative contribution of the phosphorus (P) pools as a percentage of the total reduction (%) over three years
Full size table

Discussions

The present study revealed that maize/faba bean intercropping had an advantage in P acquisition from the soil accumulated P, and the advantage existed across the experimental duration. Over three years continuous P depletion, sole faba bean mainly depleted labile and moderately labile P, in accordance with our expectation. Contrary to our hypothesis, sole maize showed a greater depletion of moderately-labile, and sparingly-labile P than sole faba bean. The intercropping system in this case had similar P utilization as sole maize which was distinct from a previous short-term greenhouse study (Li et al. 2008).

In the current study, the Resin-P, NaHCO3-Pi, 1 M HCl-Pi, NaOH-Po, and conc. HCl-Po fractions accounted for 95% of the Pt reduction (Table 5), which was close to the value in the literature (Blake et al. 2003). However, the decreases in Pt from Stage I to Stage II in the three cropping systems were less than the P balance of Phase II. As with all field experiments, some of these discrepancies might be attributed to soil sampling and analytical errors and the estimation of the P balances. However, we consider that some of the discrepancy is due to some of P loss from the 0–20 cm topsoil when the estimated P balance was large. Some of the lost P may be retained in the subsoil (Crews and Brookes 2014), but it may also likely be lost with runoff during the three years.

The Resin-P and NaHCO3-Pi are the most available fractions for plant uptake (Bowman and Cole 1978; Tiessen and Moir 1993), so they were significantly depleted with plant uptake, but their contribution to the Pt reduction was little. The HCl-P (the sum of 1 M HCl-Pi, conc. HCl-Pi, and conc. HCl-Po) that represents moderately labile or stable P compounds is the dominant P fraction in the calcareous soil (Gao et al. 2016), accounting for more than 83% of the Pt in this study (Table 4). The 1 M HCl-Pi was assumed to represent the fraction of calcium phosphate and P associated with Fe or Al oxides (Hedley et al. 1982a, 1983), but we acknowledge that this assumption is an oversimplification (Barrow et al. 2020). Calcium phosphate is the dominant fraction in calcareous soil based on previous studies and the present results. The contribution of 1 M HCl-Pi to the Pt depletion ranged from 33% to 44% (Table 5), and may have resulted from substantial H+ release by plant roots to dissolve calcium phosphate (Hinsinger 2001). The Po fractions comprised NaHCO3-Po, NaOH-Po, and conc. HCl-Po, and accounted for nearly 10% of the Pt in our study (Table 4). The moderately-labile and sparingly-labile Po fractions represented by NaOH-Po and conc. HCl-Po had similar contributions to the Pt depletion as 1 M HCl-Pi (Table 5), suggesting that the Po pools are important in providing plant-available P through mineralization (Stewart and Tiessen 1987; Randhawa et al. 2005). Declines in Po concentrations with the cessation of fertilizer-P additions have been noted, but are usually only reported with cropping rotations, where the action of tillage stimulates the mineralization of organic matter and associated Po (Hedley et al. 1982b; Linquist et al. 1997). Reductions in Po have also been noted in low-fertility grassland soils, where less Pi is available and production is more reliant on mineralization (Sharpley 1985). Thus, the P fractions mentioned above (Resin-P, NaHCO3-Pi, 1 M HCl-Pi, NaOH-Po and conc. HCl-Po) were significantly depleted at Stage II compared with Stage I. However, the conc. HCl-Pi and Residual-P proportions significantly increased, which may be the result of P movement from the labile pools into the less labile and even non-labile pools (Walker and Syers 1976; Smeck 1985; Stewart and Tiessen 1987; Negassa and Leinweber 2009). In summary, a negative P balance led to P depletion, consequently reducing labile P and accumulating non-labile P concentrations, indicating that a rational rate of P fertilizer input is necessary to maintain soil fertility while reducing the amount of legacy P and the risk of run-off of P.

This study shows a great effect of the cropping system on soil P fractions with increasing cultivation duration without additional P fertilizer. At Stage I of P depletion in our study, crop species showed slight differences in 1 M HCl-Pi, NaHCO3-Po, and conc. HCl-Po depletion (Table 4). This finding is in accordance with the short-term pot experiment conducted by Dissanayaka et al. (2015). However, at Stage II, when P had not being applied to soil for four years, the difference in P fractions among cropping systems was large (Table 4). It is not supprising that different cropping systems (sole maize, sole faba bean, and maize/faba bean intercropping) all depleted the most labile P (Resin-P and NaHCO3-Pi), and there was no difference between cropping systems (Fig. 3). This finding agrees with most studies (Kamh et al. 1999; Rose et al. 2010; Hassan et al. 2012a, b; Dissanayaka et al. 2015). However, the depletion in other P fractions varied largely with crop biological traits, similar to previous studies (Nuruzzaman et al. 2006; Li et al. 2008; Rose et al. 2010). Faba bean mainly used labile and moderately labile P (Table 5), because of its ability of releasing root exudates to enhance P availability (Lindsay and Moreno 1960; Nuruzzaman et al. 2006; Li et al. 2007, 2016b). Conversely, the ability of maize to modify the rhizosphere is relatively weak, and the crop is reportedly unable to use the acid-soluble Pi pool (Li et al. 2015; Cabeza et al. 2017). However, in the current study, sole maize depleted not only 1 M HCl-Pi and NaOH-Pi, but also Po, including NaOH-Po and conc. HCl-Po (Table 4 and Fig. 3). This might be due to the fact that maize was grown in infertile soil and P acquisition relied greatly on arbuscular mycorrhizas (Miller 2000; Zhu et al. 2005; Liu et al. 2018). Zhang et al. (2018) found that arbuscular mycorrhizal fungi and their hyphae microbiome can promote soil Po mineralization under field conditions, which was related to the function of the bacterial community on the hyphae surface. The difference in Po depletion between faba bean and maize might reveal the distinct contributions of released chemical and microbial activity to Po mineralization, which needs to be tested in future experiments. The NaOH-Pi fraction is readily depleted by wheat and maize (Vu et al. 2008; Cabeza et al. 2017), which was supported by the current results (Fig. 3 and Table 4). Thus, the labile, moderately-labile and sparingly-labile P all contributed to soil P depletion by sole maize (Table 5).Compared with sole maize, soil P fractions of the maize/faba bean intercropping system showed a similar contribution to the Pt reduction and a slightly greater depletion of the moderately-labile P pool (Table 5). In the North China Plains or Northwest China, faba bean is commonly sown in March, and harvested in July, while maize is sown in April and harvested in September (Li et al. 1999; Li et al. 2006; Xia et al. 2013; Li et al. 2016a). After the faba bean harvest, intercropped maize grows alone in the field for more than two months. Soil fractions modified by intercropped faba bean were probably eliminated gradually over the time, leading to a similar contribution of sole maize and the maize/faba bean intercropping system to soil P fraction depletion. The slight difference between the two systems can presumably be attributed to intercropped faba bean residue that changed the C/P ratio in the intercropping soil which is an important factor influencing Po mineralization (McGill and Cole 1981; Harrison 1982). Hence, the intercropping configuration needs to take into account, when considering P fraction changes in a long-term field experiment which might result in different results between short-term greenhouse experiments and long-term field trials. This area of study needs further investigation.

In this study, the depletion of total Po (the sum of NaHCO3-Po, NaOH-Po and conc. HCl-Po) in all three cropping systems confirms that plants hydrolyze a small portion of these Po fractions into Pi forms, before it is taken up or reabsorbed/precipitated. This outcome is in line with results outlined by Li et al. (2015), who found a decrease in the NaOH-Po and total Po fractions for both faba bean and maize in the rhizosphere. Crews and Brookes (2014) also found a depletion in the NaOH-Po and conc. HCl-Po fractions in continuous unfertilized soil where wheat was grown from 1893 to 2009. Chen et al. (2002) showed that the depletion of NaOH-Po was related to high concentrations of water-soluble organic carbon, microbial biomass, and high phosphatase activities. Therefore, these results show that the mineralization of soil Po is related to microorganism and enzyme activities. Furthermore, crop residues may contribute to the Po pools. Many pot experiments show that the NaHCO3-Po pool in soil planted with legumes is significantly higher than that of soil planted with cereals, such as common bean and wheat (Li et al. 2008), soybean (Glycine max L.), maize (Rubio et al. 2012), and lupine and maize (Dissanayaka et al. 2015). Our findings with faba bean at Stage I also support this assertion, whereas no significant difference between faba bean and maize in this fraction was found at Stage II (Table 4). Over three years of P depletion, maize significantly increases the NaHCO3-Po fraction, indicating that NaHCO3-Po was not hydrolyzed or/and mineralized, but rather Pi was converted to Po by soil microorganisms, possibly due to an increase in the microbial P component of this fraction (Perrott et al. 1989; Guo and Yost 1998).

Besides P fertilizer input and crop physiological trait, soil sample processes, such as drying, storage and re-wetting, can alter the distribution of P among fractions. For example, Turner and Haygarth (2001) showed that drying increased labile forms of P in grassland soils, and especially Resin-P and NaHCO3-Po. These changes can be attributed to a combination of factors including the release of P from microbial biomass, dissolution of organic colloids and alterations in inorganic P adsorption and diffusion on mineral surfaces (Xu et al. 2011; Bünemann et al. 2013). In the current study, before fractioning P pools, we air-dried soil samples collected in 2016 to keep all collected samples in same condition, because the long-term stored soil samples collected in 2013 samples were air-dried. This might slightly change the value but wouldn’t alter the pattern of P fraction depletion. When comparing soil samples in 2013 and 2016, we were most concerned about mineralization of Po fractions during storage of soil samples in 2013, which would affect analysis of the change in Po fractions between 2013 and 2016. This concern was somewhat addressed with the findings that virtually all Po fractions of soil samples in 2013 were higher or equal to those in 2016 (described below in Fig. 3). This coincides with the finding by Blake et al. (2000), who found no evidence of net C or N mineralization in archived samples.

This study indicates that legume/cereal intercropping offers a promising practice to realize agriculture sustainable development, particularly in low-input systems or on nutrient-poor soils, where intercropping is becoming an increasingly attractive cropping system due to its advantages over sole cropping in terms of yield (Li et al. 1999; Li et al. 2010; Gao et al. 2019), nutrient use (Li et al. 2007; Xu et al. 2020), disease control (Zhang et al. 2019), and economic return (Huang et al. 2015). Crop production following the cessation of P-fertilizer additions to P-rich soils can be maintained by unlocking soil P reserves (Gillingham et al. 1990; Dodd et al. 2013; McDowell et al. 2016). Nonetheless, ensuring food security in the long run requires rational P input based on the characteristics of soil and cropping system.

Conclusions

Our study showed that the maize and faba bean intercropping system increased P utilization compared with sole cropping of maize or faba bean. The intercropped maize and faba bean showed a greater average shoot P content over four years than the corresponding sole crops, by 29% and 30%, respectively. Following the cessation of P fertilizer application, maize and faba bean showed different dynamics of various soil P fractions, from the labile fractions to more stable fractions. The largest difference was found for soil Po pools over three-year cultivation, in which maize depleted the conc. HCl-Po fraction and faba bean depleted the alkali-soluble Po fraction (extracted by NaHCO3 and NaOH). These Po fractions were depleted over the period, under both maize and faba bean, suggesting that it is an important source of plant-available P in non-fertilized conditions.

Changes in soil Pt and P fractions between the two periods we investigated without P fertilizer application suggest that a decline of Pt over the three-year cultivation was mainly related to changes in the moderately labile P pool. Under depletion conditions, all three cropping systems primarily depleted the 1 M HCl-Pi fraction, followed by a monocrop of maize depleting NaOH-Pi and conc. HCl-Po fractions, and a monocrop of faba bean depleting the alkali-soluble Po fraction (extracted by NaHCO3 and NaOH). Finally, the intercropped maize/faba bean depleted the conc. HCl-Po fraction in a similar manner as sole maize did.