1 Introduction

Soil phosphorus (P) is a curbing nutrient in agricultural ecosystems and a major cause of surface water eutrophication (Li et al. 2021b; Jin et al. 2023). Lower P-bioavailability in agricultural soils is generally a result of its strong adsorption to soil particles or natural colloids (Menezes-Blackburn et al. 2021; Eltohamy et al. 2023b). To maintain plant growth, organic P-fertilisers have continuously been added to soils both as a source of essential nutrients and as a soil conditioner to improve overall soil fertility and health (Wang et al. 2021; Khan et al. 2022a; Zhao et al. 2022). The overuse of P-fertiliser has caused a large pool of "legacy" soil-P (Hosseini et al. 2019; Menezes-Blackburn et al. 2021). When P sufficiency is achieved, organic fertilisers are still used to address nitrogen deficiency and may correspond to continuous legacy P build-up, especially in organic fractions (Menezes-Blackburn et al. 2016b; Li et al. 2021a). This legacy-P poses a significant risk of nutrient pollution of water bodies via leaching- and runoff-related losses that are directly related to the eutrophication of surrounding waters (Yan et al. 2016; Chen and Arai 2020; Eltohamy et al. 2021).

Natural colloids are considered important P carriers in agricultural ecosystems (Wang et al. 2020; Li et al. 2021b) and are thought to be a major cause of lower P-bioavailability to terrestrial plants (Montalvo et al. 2015). Owing to their higher surface charge and larger surface area, they remain in the soil solution for prolonged periods (Baalousha et al. 2005) and carry P to adjacent waterbodies upon irrigation or rainfall events via leaching or runoff (Jiang et al. 2015; Missong et al. 2016). These colloids contain the weathering products of minerals and the breakdown of organic matter (Missong et al. 2018; Wang et al. 2020; Eltohamy et al. 2023b). In soils amended with organic fertilisers, further colloids may originate from manure, slurry, or biochar applications (Chen and Arai 2020; Meng et al. 2021). Total colloidal P is outlined as P-containing soil particles within a size range of 1–1000 nm (Wang et al. 2020; Eltohamy et al. 2021; Khan et al. 2022b). Nonetheless, the soil particles that have better potential to bind P are < 450 nm in size (Henderson et al. 2012; Baken et al. 2016a; Eltohamy et al. 2023b). Water-dispersible colloid-bound P (Pcoll; in the size range of 1–450 nm) is characterised by properties such as small size, large specific surface area, increased reactivity, and high mobility that enable its movement in a soil solution (Konrad et al. 2021; Yang et al. 2021a; Eltohamy et al. 2023b). The Pcoll and the truly soluble P (TSP) in soil solution together represent the mobile P phase (i.e., total dissolved P; TDP) (Henderson et al. 2012; Li et al. 2021b). Nonetheless, Pcoll has distinct colloid-chemical properties that differ from those of TSP (Henderson et al. 2012). For instance, Pcoll may migrate more rapidly than TSP owing to the electrostatic repulsion and spatial exclusion of the soil matrix (Henderson et al. 2012; Li et al. 2021b). Several studies have documented considerably high TDP levels in soil solution and streams linked to nano- (< 100 nm) and colloidal particles (< 450 nm) (Gottselig et al. 2017, 2020; Adediran et al. 2021; Li et al. 2021b). It has been reported that Pcoll can be divided into three distinct size fractions: nano-sized colloids (NC; 1–20 nm), which are rich in organic carbon (OC), fine-sized colloids (FC; 20–220 nm), which are mainly composed of OC, iron (Fe) and aluminium (Al), and medium-sized colloids (MC; 220–450 nm), which contain a substantial quantity of phyllosilicates (Missong et al. 2018; Eltohamy et al. 2023b). Therefore, it is essential to understand how organic amendments affect the size distribution of these different Pcoll fractions (i.e., NC, FC, and MC) and their composition. This could help us figure out the role and relative contribution of these different-sized Pcoll fractions to P-bioavailability in these soils and their impact on soil P-cycling in terrestrial ecosystems.

To date, there are conflicting statements in the literature on the effects of long-term organic amendments on the content of soil OC and mineral elements and their relationship with the release and loss of Pcoll from agricultural soils. Yu et al. (2017) found that organic amendments significantly increase the availability of Fe and Al minerals but decrease OC mineralisation. In contrast, Li et al. (2021b) reported a high concentration of OC in soil colloids, but a decrease in the concentrations of Fe and Al in soils amended with organic fertilisers. A study of a long-term application of manure to loam soil showed that P was associated with Ca rather than Fe or Al, leading to a transformation of relatively soluble calcium (Ca)–P into more colloidal crystal phases (Sato et al. 2005). Other studies have attributed the greater Pcoll release in soils amended with manure to the high co-occurrence of OCcoll and colloidal mineral elements (i.e., Fecoll, Alcoll, and Cacoll) in these soils (Liang et al. 2016; Khan et al. 2021). It has been suggested that the application of biochar to soils in appropriate quantities may be beneficial to avoid the risk of Pcoll losses to adjacent waterbodies (Hosseini et al. 2019; Li et al. 2021b). For instance, straw- and manure-derived biochar effectively reduced soil Pcoll loss by 55.2%–56.7% (Wang et al. 2023). Moreover, the addition of biochar can enhance soil phosphatase activity and promote the transformation of organic phosphate (Yang et al. 2021b; Wang et al. 2023). On the other hand, biochar contains a high content of dissolved P, a form of P that is easily lost from the soil, which can accelerate soil P loss when applied to the soil (Li et al. 2019; Eltohamy et al. 2021). Therefore, more systematic experiments on a broader scale are warranted to better investigate the associations of OC and mineral elements with Pcoll in different colloidal size fractions in organic fertiliser-amended soils.

Therefore, we collected soil samples from 12 field sites with different organic P fertiliser application practices (i.e., manure, biochar, and biogas slurry) throughout Zhejiang Province, China, to assess the effects of long-term application of organic fertilisers on the size and compositional characteristics of Pcoll. Our specific objectives were (a) to explore and understand the distribution of Pcoll concentration and its contribution to TDP content in the soil solution for soils amended with organic fertilisers; (b) to probe the relative importance of P-bearing colloids of different sizes (i.e., NC, FC, and MC) and compositions on the P mobilisation and transport from these soils; and (c) to identify potential linkages between Pcoll of different size fractions and other colloidal components at a regional-scale (Zhejiang Province, China). This study broadens our understanding of the loss behaviour of Pcoll of different size fractions in soils amended with organic fertilisers and will help us formulate appropriate fertilisation strategies in order to reduce potential Pcoll losses and improve its bioavailability.

2 Materials and methods

2.1 Study region and soil sampling

The soil samples used in this study were obtained from Zhejiang province in east-south China (Figure S1). This region experiences heavy rainfall (Table S1), causing soils to lose Pcoll via runoff (Li et al. 2021b). Bulk soils (0–20 cm) were assembled from 12 field sites with organic P fertiliser application practices between October 6 and 26, 2020 (Tables S2 and S3). Sites also represented different agricultural management systems (Table S1). Each site was assigned three typical fields (three replicates). Using the S-shaped sampling technique (Wang et al. 2017), a 10 kg topsoil composite sample (10–15 subsamples) was assembled from each field and carefully mixed to generate a representative sample. The sampled soils were then air-dried (25 ± 1 °C), coarsely crushed, and the roots and stones were eliminated. The fraction that passed through a 2-mm sieve was stored until use.

The soil samples’ general properties were characterised in order to control and understand other general drivers of soil solution Pcoll behaviour. Namely, the soil particle size distribution was determined using the hydrometric method (García-Gaines and Frankenstein 2015). The cation exchange capacity (CEC) was measured with 1 M ammonium acetate at pH 7 (Sumner and Miller 1996). Soil pH was measured at a soil-to-water ratio of 1:5 (w/v) after 30 min of shaking. The soil was digested with a nitric acid (HNO3)–perchloric acid (HClO4) solution plasma atomic emission and the contents of total P, Fe, Al, Ca, magnesium (Mg), and manganese (Mn) were detected using inductively coupled spectroscopy (Fresne et al. 2021). Soil total carbon was analysed by an elemental analyser (Vario MAX CNX, Elementar, Germany).

2.2 Sample processing and soil solution extraction

Figure S2 shows the analytical workflow of the study. Soils were incubated at a field-like moisture content using the method outlined earlier (Menezes-Blackburn et al. 2016b; Eltohamy et al. 2023b). First, the soils were tested to establish their maximum water-holding capacity (MWHC) by adding Milli-Q water (18.2 MΩ cm, Millipore), followed by continuous mixing until the MWHC was attained. A visual appraisal of soil plasticity and the glistening of water on the soil surface was used to figure out MWHC (Menezes-Blackburn et al. 2016a). Soil samples (150 g) were wetted with Milli-Q water to reach 50% of their MWHC. Samples were incubated at 25 ± 1 °C for 48 h and then at 100% MWHC for an additional 24 h. After incubation of the soils, the soil solutions were extracted by centrifugation at 1957 × g for 15 min (Wu et al. 2018).

2.3 Colloidal size-fractionation

Differential centrifugation and ultrafiltration were used to separate colloids into three different size fractions: (1) NC (~ 1–20 nm, distinguishes between 300 and 3 kDa filtrates), (2) FC (~ 20–220 nm, distinguishes between  0.22 μm and 300 kDa filtrates), and (3) MC (220–450 nm, distinguishes between 0.45 µm and 0.22 µm filtrates), as previously described (Xia et al. 2018; Eltohamy et al. 2023b). Nitrocellulose membranes (Millipore, Billerica, Massachusetts, United States) with a pore size of 0.45 μm or 0.22 μm were used to filter out the < 450 nm or < 220 nm portions of the extracted soil solution. Filtration through 300- or 3-kDa ultrafiltration membranes (Millipore Amicon Ultra, United States) was used to separate the particle fraction < 20 nm at 3500 × g for 20 min (Xia et al. 2018) or < 1 nm at 4000 × g for 45 min (Li et al. 2021b). To eliminate the possibility of artifacts, ultrafiltration membranes were prewashed with 5% HNO3 (v/v) and then rinsed three times with Milli-Q water (Saeed et al. 2018). The association between the used molecular weight cut-offs and membrane particle size cut-offs has been previously validated (Henderson et al. 2012; Xia et al. 2018; Li et al. 2021b). The estimation was done based on the relationship between Stokes diameter and both molecular weight and molecular shape of organic test macromolecules (Xia et al. 2018).

Total P, Al, Fe, Ca, Mg, silicon (Si), and Mn concentrations were determined with an inductively coupled plasma-mass spectrometer (ICP-MS, Agilent 7800, Agilent Technologies, Santa Clara, CA, United States) in a subset of the filtrates (i.e., 0.45 µm, 0.22 µm, 300 kDa, and 3 kDa) acidified to pH 2 with 1 M HNO3. The internal standard was a Rhodium (103Rh) solution (GSS2; National Sharing Platform for Reference Material, China). After acidification (pH < 2) with 0.5 M hydrochloric acid (HCl), the total OC concentrations of the filtrates were measured using a total OC analyser (multi N/C2100, Analytik Jena, Germany). The quality control of these analytical analyses was validated in our prior study (Eltohamy et al. 2023b). To verify the effective size fractionation of the mobile colloids in the extracted soil solution (Figure S2), dynamic light scattering (DLS) measurements were performed on the non-acidified < 450 nm portions using the Zetasizer (Nano- ZS, Malvern Panalytical, Worcestershire, United Kingdom).

2.4 Determination of inorganic and organic phosphorus forms in dissolved and colloidal phases

The P concentrations in the 0.45 μm and 3 kDa filtrates measured by ICP-MS indicate TDP and TSP, respectively. The concentration of dissolved molybdate-reactive P (i.e., dissolved inorganic P; DPi) was measured colorimetrically (Murphy and Riley 1962) using part of the unacidified 0.45 μm filtrate. Dissolved molybdate-unreactive P (i.e., dissolved organic P; DPo) was estimated as the difference between TDP and DPi. Additionally, the colloidal molybdate-reactive Pcoll (inorganic Pcoll; Picoll) concentration was measured on the unacidified portions of all filtrates by reaction with molybdate (Murphy and Riley 1962) and the colloidal molybdate-unreactive P (organic Pcoll; Pocoll) was calculated by subtracting Picoll from Pcoll.

2.5 Statistical analysis

Significant differences in the independent variables between soils were assessed using a one-way analysis of variance, and the means were compared using Tukey’s multiple comparison test (p < 0.05). Pearson correlation coefficients were computed between the independent variables, and the significance of correlations was determined using Student’s t-test (significant at p ≤ 0.05 and highly significant at p ≤ 0.01). The correlation coefficients were calculated with log-transformed data to minimise heteroskedasticity. Principal component analysis (PCA) scores and model cross-validation (Q2) were calculated using partial least squares-discriminant multivariate analysis (PLS-DA). Data processing and analysis were performed with Microsoft Excel and SPSS statistical software package ver. 22 (SPSS Inc., Chicago, IL). Visualisation was conducted using OriginPro 2021 (OriginLab Corp., United States).

3 Results and discussion

3.1 General characteristics of the tested soils

Table 1 shows the general characteristics of the tested soils. The soil pH varied between acidic (4.58; S12) and alkaline (7.96; S9). Soil content of total carbon ranged from 5.62 g kg−1 in S3 soil to 27.27 g kg−1 in S11 soil. Soil total P content was the highest in the S2 soil (4.94 g kg −1) and the lowest in the S10 soil (0.38 g kg−1). The highest total Fe (32.14 g kg−1), total Al (41.08 g kg−1), and total Mg (9.92 g kg−1) contents were also found in S2 soil, whereas the lowest contents were 7.49 g kg−1 for total Fe in S12 and 9.61 g kg−1 for total Al and 0.94 g kg−1 for total Mg in S11 soil. The soils had much lower contents of total Mn (0.12 (S12)–0.97 (S7) g kg−1). The highest soil total Ca content occurred in S9 (8.22 g kg−1) and the lowest was in S12 (0.93 g kg−1). The CEC varied greatly among the soils, ranging from 11.22 cmol kg−1 (S3) to 166.71 cmol kg−1 (S10). The results showed a strong variation in the spatial distribution of the general characteristics of the studied soils (Table 1), indicating that the soils in central and southern Zhejiang are rich in Fe and Al and contain less carbon (Table 1 and Figure S1). The soils in the northern plain showed lower Al, distinct accumulation layers, as well as greater carbon storage and clay content (Table 1). Similar results were found by previous work from our group, (Li et al. 2021b; Eltohamy et al. 2023b), using the same study areas. These observed patterns, and heterogeneity of regional differences in the observed background properties of the soils, allow for contrasting different response parameters with the colloidal sizes and their compositional distribution patterns of these soils.

Table 1 Summary of the tested soil characteristics
Full size table

3.2 Partitioning of soil solution P between colloidal and dissolved phases of the studied soils

The total dissolved P in soil solution (i.e., TDP) consists of Pcoll (< 450 nm) and TSP (Henderson et al. 2012; Li et al. 2021b). The studied soils had a wide range of TDP contents (123.5 (S12)–2826.3 (S2) mg L−1; Table S4) and were dominated by DPi (mean ± SD = 68 ± 13% of TDP; Fig. 1A). Such a high presence of DPi in the soil solution on account of DPo implied a high P-bioavailability in the studied soils. This is because the DPi represents the soil solution labile P pool that compensates for the P depleted by plant uptake from the root zone (Eltohamy et al. 2023b). DPo, on the other hand, is not readily available for plant uptake as it has a high molecular weight and needs to be digested and made available in the form of orthophosphate, a soluble inorganic P form (Das et al. 2022).

Fig. 1
figure 1

Partitioning of total dissolved P (TDP) in soil solution between (A) dissolved inorganic P (DPi; in % of TDP) and dissolved organic P (DPo; in % of TDP) and between (B) colloidal P (Pcoll; in % of TDP) and truly soluble P (TSP; in % of TDP) and the distribution of Pcoll between inorganic Pcoll (Picoll; in % of Pcoll) and organic Pcoll (Pocoll; in % of Pcoll) in the size ranges (C) total (1 − 450 nm), (D) nano (NC; 1 − 20 nm), (E) fine (FC; 20 − 220 nm), and (F) medium colloids (MC; 220 − 450 nm)

Full size image

Pcoll represented a higher average proportion of TDP than TSP (Pcoll = 73 ± 14% vs. TSP = 27 ± 14%; Fig. 1B and Tables 2 and S5), indicating that the potential risk of TDP transport from organic fertiliser-amended soils to adjacent waters often takes place in the colloidal phase. The higher proportion of Pcoll relative to TSP also suggests that P-containing colloids may be resedimented during transport through soil macropores, which act as preferential flow pathways for TDP (Julich et al. 2017; Gu et al. 2020). This is consistent with previous studies which concluded that 40–58% of TDP transport from the topsoil profile is in the form of Pcoll (Hens and Merckx 2002; Wang et al. 2021).

Table 2 Mean concentrations of colloidal P (Pcoll; µg L−1), inorganic P (Picoll; µg L−1), and organic P (Pocoll; µg L−1) in the nano-sized (1 − 20 nm), fine-sized (20 − 220 nm), and medium-sized (220 − 450 nm) fractions in the soil solution of tested soils (n = 3)
Full size table

The results also showed that spatial distribution characteristics of the tested soils (Tables 1 and S1) probably influenced the Pcoll content in the soil solution (Table 2). The Pcoll in the soil solution can be categorised into the concentration ranges, low (103.6 (S12) to 142.0 μg L–1 (S10)), moderate (227.7 (S9) to 478.0 μg L–1 (S4)), and high (1053.6 (S3) to 2041.4 μg L–1 (S1)) (Table 2). Soils in central and southern Zhejiang had moderate to high Pcoll levels (S1–S9). This could be due to the fact that the soils in the central and southern plains are highly susceptible to desilication, are rich in Fe and Al, and contain less carbon (Table 1) (Zhejiang Soil Census Office 1994), resulting in high Pcoll release into the soil solution. While the northern plain soils have lower aluminium content, higher carbon storage and clay content (Table 1) (Zhejiang Soil Census Office 1994; Li et al. 2021b), result in low Pcoll release (S10–S12). This pattern causes regional differences in the distribution patterns of Pcoll as previously hypothesized (Li et al. 2021b).

3.3 Compositional characteristics of colloidal phosphorus of different size fractions

There were obvious differences in the contents of the three colloidal fractions between soils (Tables 2, 3 and S6). Note that the soils were ordered according to their soil solution Pcoll content, with S1 soil (2041.4 µg L–1) having the greatest content and S12 soil having the lowest (103.6 µg L–1) (Table 2). As shown by the DLS measurement (Fig. 2), the effective particle size of the 450-nm extracted soil solution ranged from 176.6 ± 47.7 (S1) to 352.3 ± 27.1 nm (S10). This indicates that most of the mobile colloids of the studied soils were distributed among the FC and MC fractions. On average, the size distribution of Pcoll was predominantly present as FC (322.4 µg L−1), compared to MC (168.4 µg L−1) and NC (61.7 µg L−1) (Table 2). Pcoll was enriched in inorganic P (Picoll = 56 ± 15% of Pcoll; Fig. 1C). However, the partitioning of Pcoll between inorganic and organic P (Picoll vs. Pocoll) appears to vary among the three colloidal fractions and shows different patterns in the speciation distribution of Pcoll (Figs. 1D–F). On average, NC- and FC-bound Pcoll made up 54 ± 20% and 63 ± 28% of Picoll, respectively (Figs. 1D and E). In contrast, MC-bound Pcoll was more enriched in Pocoll (58 ± 26%; Fig. 1F). This shows that NC- and FC-bound Pcoll fractions are more bioavailable than the MC-bound Pcoll fraction. This result supports the hypothesis of a recent study that NC- and FC-bound Pcoll can contribute to P uptake, whereas MC-bound Pcoll is involved in vertical P-transport from the soil solution and leads to P enrichment in receiving waters (Eltohamy et al. 2023b).

Table 3 Mean concentration of colloidal organic carbon (OCcoll), aluminium (Alcoll), iron (Fecoll), magnesium (Mgcoll), calcium (Cacoll), silicon (Sicoll), and manganese (Mncoll) in nano-sized (1 − 20 nm), fine-sized (20 − 220 nm), and medium-sized (220 − 450 nm) fractions in the soil solution of tested soils (n = 3)
Full size table
Fig. 2
figure 2

Particle distribution of soil solution colloids (< 450 nm) in the tested soils

Full size image

Total OCcoll was the main constituent of the colloids suspended in the soil solution (Fig. 3), and its mean concentrations in FC (28.5 mg L−1) or MC (14.9 mg L−1) were far below those in NC (35.4 mg L−1) (Table 3). This suggests that NC-bound OCcoll is an important source of OC transfer in soils amended with organic fertilisers (Menezes-Blackburn et al. 2021; Eltohamy et al. 2023b). Other colloidal constituents were predominantly distributed in the FC or MC (Fig. 3 and Table 3), showing that these elements are important units for the formation of larger colloids (Wang et al. 2020). The size distributions of Alcoll and Fecoll were very similar (Fig. 3), revealing that Alcoll might be present as Al-oxyhydroxides associated with the Fe-oxyhydroxide colloids (Baken et al. 2016b). The OC-rich composition of the NC fraction (Fig. 3) agrees well with previous studies for P-bearing nanoparticles in an Albeluvisol and a Stagnosol (Jiang et al. 2017; Gu et al. 2020). According to Jiang et al. (2017), the reduction regime that characterises waterlogged soils (as in the present study, in which the soils were moistened to their MWHC) explains this difference in the composition of P-bearing colloids. These two factors would favor the formation of OC-dominated nanoparticles with a minor contribution of other colloidal components. The abundant presence of OC in both NC and FC relative to the MC could also be a reason for the better bioavailability of NC- and FC-bound Pcoll fractions than MC-bound Pcoll. This is because organic matter attached to the particle surface facilitates the dispersibility of these colloids in the soil solution by changing their surface charges to be more negative (Missong et al. 2018; Eltohamy et al. 2023a). This, in turn, leads to the dissolution or desorption of these NC- and FC-bound Pcoll fractions and thus causes diffusion near the plant root (Eltohamy et al. 2023b).

Fig. 3
figure 3

Concentrations of colloidal phosphorus (Pcoll) and other colloidal components in the nano- (NC; 1–20 nm), fine- (FC; 20 − 220 nm), and medium-sized (MC; 220–450 nm) fractions. Dots represent individual values (n = 36)

Full size image

3.4 Association of colloidal phosphorus of different size fractions with other colloidal components

The relationships between Pcoll in the three colloidal fractions, and the other colloidal components, are shown in Fig. 4A. The variances by multivariate principal component analysis showed significant discrimination (p < 0.05) among the concentrations of different-size colloids in the 12 tested soils based on the first two components, which accounted for 75.2% of the total variance (Fig. 4B), and the greater the range, the lower the bias in the correlation analysis. Additionally, the experimental PCA model showed high prediction based on the PLS-DA model indicators for the second PC (Q2 > 0.5) (Fig. 4C). That is, the model had a strong predictive ability based on the first two principal components (Szymańska et al. 2012).

Fig. 4
figure 4

Pearson correlation coefficients (A) principal component analysis (PCA) (B), and model performance (C) of colloidal P forms (Pcoll, Picoll, Pocoll) with other colloids (OCcoll, Fecoll, Alcoll, Mgcoll, Cacoll, Sicoll, and Mncoll) at nano- (NC; 1 − 20 nm), fine- (FC; 20 − 220 nm), and medium-sized (MC; 220 − 450 nm) fractions

Full size image

The elemental composition of these colloids showed that OCcoll, Alcoll, Fecoll, Mgcoll, Cacoll, Sicoll, and Mncoll were strongly correlated with Pcoll (r ≥ 0.76; Fig. 4A and Table S7). These correlations support the previous suggestion that mobile Pcoll tends to be associated with OC–elemental species containing Fe, Al, Ca, Mg, Mn, and Si (Konrad et al. 2021; Li et al. 2021b; Eltohamy et al. 2023b). Here, our results show that NC-bound Pcoll was correlated only with the OC–Ca colloids (r ≥ 55; Fig. 4A and Table S7) that presumably dominated by organic compounds stabilised by bridging Ca2+ ions (Fig. 5), which are stable enough to avoid hetero-aggregation with other colloidal soil particles (Li et al. 2021b; Eltohamy et al. 2023b). In contrast, MC-bound Pcoll was better associated with Mn/Fe/Al colloids (r = 0.82 to 0.86) complexed with OCcoll (r = 0.74) than OC–Ca colloids (r = 0.59 for Cacoll; Fig. 4A and Table S7). Previous investigations on natural and laboratory-extracted colloids in the range of 160 nm to 450 nm have also shown the importance of metal oxyhydroxides (e.g., Fe/Al-oxyhydroxides) and organic matter as effective P adsorption sinks (Jiang et al. 2015; Yan et al. 2016; Gu et al. 2020). Therefore, we can assume that the P-bearing colloidal vectors in the MC fraction are largely colloidal Fe/Al/Mn-oxyhydroxides complexed with OCcoll (Fig. 5). FC-bound Pcoll was highly correlated with all other colloidal components (r = 0.73 to 0.86; Fig. 4A and Table S7) suggesting that two phenomena (OC–Ca–P and OC–Mn/Fe/Al–P) were occurring in this size range (i.e., FC; 20–220 nm) (Fig. 5). Thus, we can conclude that the OC–Ca complex was the major carrier of NC-bound Pcoll, OC–Mn/Fe/Al colloids were the main vectors of MC-bound Pcoll, and these two phenomena (i.e., OC–Ca–P and OC–Mn/Fe/Al–P) co-occurred in the FC fraction.

Fig. 5
figure 5

Mechanisms of in-soil micro-scale release and migration of colloids at nano- (NC; 1 − 20 nm), fine- (FC; 20 − 220 nm), and medium-sized (MC; 220 − 450 nm) fractions in soils amended with organic fertilisers

Full size image

3.5 Effect of biochar amendment on colloidal phosphorus content and size distribution versus other organic amendments

The soils to which manure or biogas slurry was added, i.e., soils S1 to S6, had higher Pcoll concentrations (Tables 2 and S2), and this result was consistent with recent works on natural and laboratory-extracted soil colloids (Niyungeko et al. 2018; Eltohamy et al. 2021; Li et al. 2021b; Khan et al. 2022a). Li et al. (2021a, b) observed that the application of biogas slurry, instead of 30% chemical fertiliser, considerably enhanced soil Pcoll. They attributed this to the possibility that the biogas slurry itself contained a large amount of Pocoll bound to clay colloids or in the NC-bound Pcoll form once it entered the soil. Zang et al. (2013) and Makris et al. (2006) also pointed out the significance of manure input for the formation of soil colloids. These explanations, which are consistent with our results, indicate that further colloids may originate in soil due to manure or biogas slurry application (see Fig. 5) (Chen and Arai 2020; Meng et al. 2021). In our data, most soils amended with biochar together with, or instead of, manure or biogas slurry had lower Pcoll contents than those amended with manure or slurry alone (Tables 2 and S2). There are two possible explanations, namely that: biochar absorbs Pcoll; and that biochar enhances soil aggregate formation (see Fig. 5), as previously suggested (Li et al. 2019, 2021b; Zhou et al. 2019). Biochar can interact with soil OC and mineral elements (e.g., Fe/Al, Ca, Mg, Si, Mn), which in turn improves soil aggregation and thus reduces soil Pcoll loss (Li et al. 2019; Wang et al. 2021). This is because the formation of soil aggregates and their stabilisation are closely linked to OC, given that the OC distribution associated with the aggregates determines the OC storage and retention capacity of soils (Kumar et al. 2023). Biochar not only directly absorbs Pcoll to reduce its losses from the soil solution to the receiving water, but also creates hotspots for microbial communities that simultaneously affect the soil Pcoll content (Wang et al. 2023).

Regarding the size distribution of Pcoll, no clear patterns were observed among the three colloidal fractions (i.e., NC, FC, and MC) as a result of organic fertiliser application. This result is consistent with the hypothesis that differences in organic fertiliser application affect the concentration of compounds adsorbed by colloids (i.e., total Pcoll), while the colloidal composition affects their distribution and morphology (Li et al. 2021b).

3.6 Environmental implications and future perspectives

By jointly investigating the size and distribution of colloidal components in soils amended with organic fertilisers on a regional scale, the current study provides clear evidence that colloids are important carriers of P. The Pcoll has a different environmental fate and plays a more active role than other P-forms (particularly dissolved P, in which it is commonly considered) in the P-cycling in soils (Henderson et al. 2012; Lin et al. 2012). Additionally, Pcoll is often present in association with soil OC and mineral colloids. After its release, its small particles (i.e., NC and FC; < 220 nm) can become bioavailable by dissolution or desorption or by diffusion near the plant root (Montalvo et al. 2015; Eltohamy et al. 2023b). While its medium-sized fraction (i.e., MC) forms complexes with other colloids, causing P occlusion and contributing to vertical P-transport into receiving waters (i.e., leaching or runoff) (Konrad et al. 2021; Eltohamy et al. 2023b). Thus, the starting point for studying the colloidal loss potential of a particular soil is to understand the composition of its organic and mineral phases as well as its function (Li et al. 2021b). This is an important prerequisite for formulating an appropriate fertilisation strategy to reduce potential losses of Pcoll and make it more bioavailable to crops (Eltohamy et al. 2023a). Improving the bioavailability of Pcoll will help reduce overall organic P-fertilisers in soils to environmentally acceptable levels while maintaining optimal growing conditions for plants.

The results also showed that soils amended with biochar together with, or instead of, manure or biogas slurry had lower Pcoll levels than soils fertilised with manure or slurry alone. Biochar, as far as we know, is involved in the remediation of pollutants in soils via a variety of processes, either directly by immobilising pollutants or indirectly by affecting soil metabolic activities and degradation (Li et al. 2022; Xiang et al. 2022). Biochar also contributes to carbon neutrality due to its carbon-negative nature, CO2 sorption, and adverse priming effects (Wang et al. 2023). Moreover, biochar can afford microorganisms' habitats, reduce pollutants' toxicity to microorganisms, and mitigate various biotic and abiotic stresses on plants (Xiang et al. 2022). Overall, biochar offers positive attributes that make it a suitable means of improving soil health and promoting plant growth. In this context, we recommend further research to explore the efficacy of biochar amendments in decreasing the Pcoll release and loss from agricultural soils and to assess the economic and social viability of local manufacturing of biochar amendments.

4 Conclusion

The present study explored the abundance, size and compositional characteristics of Pcoll in 12 field sites subjected to long-term application of different organic amendments. The results showed that Pcoll was the major component of TDP in the tested soils, suggesting that the mobile colloidal phase acts as a P carrier, increasing the potential risk of P transport from these soils to receiving waters. It should also be highlighted that the soils amended with biochar showed lower Pcoll contents than those amended with manure or slurry alone, and this effect should be further explored in a controlled inductive research approach. In the three colloidal fractions, OCcoll was the major colloid component for P-binding in these soils, with varying contributions from the other colloidal elements. In the NC fraction, the OC–Ca associations were the major vectors for NC-bound Pcoll. In contrast, MC-bound Pcoll was better associated with OC–Mn/Fe/Al colloids than with OC–Ca colloids. In the FC fraction, both phenomena (i.e., OC–Ca–P and OC–Mn/Fe/Al–P) occurred together. It can be concluded that the compositions of Pcoll in different size fractions in soils amended with organic fertilisers are different. Therefore, the starting point for studying the colloidal loss potential of a particular soil enriched with organic fertiliser is to understand the composition of its colloidal organic and mineral phases as well as their function in the soil solution. This study adds to the understanding of how Pcoll of different size fractions behaves in soils enriched with different organic amendments. As such, it offers valuable insights for the formulation of future policies and the development of pertinent agronomic and environmental strategies.