Enhanced soil aggregate stability limits colloidal phosphorus loss potentials in agricultural systems

Background: Colloid-facilitated phosphorus (P) transport is recognized as an important pathway for the loss of soil P in agricultural systems; however, information regarding soil aggregate-associated colloidal P (Pcoll) is lacking. To elucidate the effects of aggregate size on the potential loss of Pcoll in agricultural systems, soils (0–20 cm depth) from six land-use types were sampled in the Zhejiang Province in the Yangtze River Delta region, China. The aggregate size fractions (2–8 mm, 0.26–2 mm, 0.053–0.26 mm and < 0.053 mm) were separated using the wet sieving method. Colloidal P and other soil parameters in aggregates were analyzed. Results: Our study demonstrated that 0.26–2 mm small macroaggregates had the highest total P (TP) content. In acidic soils, the highest Pcoll content was observed in the 0.26to 2-mm-sized aggregates, while the lowest was reported in the < 0.053 mm (silt + clay)-sized particles, the opposite of that revealed in alkaline and neutral soils. Paddy soils contained less Pcoll than other land-use types. The proportion of Pcoll in total dissolved P (TDP) was dominated by < 0.053 mm (silt + clay)-sized particles. Aggregate size strongly influenced the loss potential of Pcoll in paddy soils, where Pcoll contributed up to 83% TDP in the silt + clay-sized particles. The Pcoll content was positively correlated with TP, Al, Fe, and the mean weight diameter. Aggregate-associated total carbon (TC), total nitrogen (TN), C/P, and C/N had significant negative effects on the contribution of Pcoll to potential soil P loss. The Pcoll content of the aggregates was controlled by the aggregate-associated TP and Al content, as well as the soil pH value. The potential loss of Pcoll from aggregates was controlled by its organic matter content. Conclusion: We concluded that management practices that increase soil aggregate stability or its organic carbon content will limit Pcoll loss in agricultural systems.


Background
The loss of phosphorus (P) from agricultural soils has been identified as one of the main causes of eutrophication of lakes in the lower reaches of the Yangtze River in southern China [1]. Statistics have shown that major lakes and reservoirs in this area were eutrophic and mesotrophic [2][3][4], and the Yangtze River Delta region accounted for 17% of the 10 Tg annual increase in soil P around the world [2]. In soil, colloidal P (P coll ) is the P fraction bound to colloids [5]. Colloidal particles are highly mobile and are effective adsorbents of organic and inorganic contaminants and nutrient elements, such as P, owing to the high specific surface area and charge density [6,7]. Colloid-facilitated P transport is an important pathway for the migration of P into water bodies [5,8,9]. One study has reported that more than 75% of P in cultivated soil solution is combined with fine particles smaller Open Access *Correspondence: liang410@zju.edu.cn 1 College of Environmental and Resources Sciences, Zhejiang University, Hangzhou 310058, China Full list of author information is available at the end of the article than 240 nm [10]; similarly, 40-58% of molybdate-reactive P, with a size less than 450 nm in the water extract of grassland soil, demonstrates fine-grained P with a size of 25-450 nm [9]. Other studies have suggested that P coll reached up to 50% of total P (TP) in surface runoff, rivers, and lakes [11], which may lead to environmental risks.
Soil aggregate stability plays a key role in controlling the erosion processes and the loss of soil nutrients [12][13][14]. Water-dispersible colloids in the soil adhere to soil aggregates, forming a stable system [15]. Colloids either can bind to soil aggregates, or be physically strained from water flowing through pores between aggregates [16,17]. Some scholars determined the colloid content in 1-2 mm aggregates in 39 soils and revealed a significant positive correlation between the water-dispersible colloid content and clay content in soil aggregates [18]. Furthermore, they reported that water-dispersible colloid content was a function of total organic C and total clay. Other scholars modeled the release characteristics of colloids from soil aggregates, the attachment and detachment processes at the air-water interface, and flocculation and straining from interstitial water [19]. The release of colloids from aggregates may result in the disintegration of aggregates [20]. However, soil aggregation mainly depends on the availability of active mineral surfaces and the dispersion/ flocculation behaviors of the colloidal components [21]. Meanwhile, the degree of clay colloid dispersion can be reduced by increasing the aggregate stability [18,22,23]. Therefore, the stability of soil aggregates directly affects the migration of soil colloids.
The retention of P in soil aggregates depends on the aggregate sizes and chemical properties [24,25]. Notably, P has a relatively closed cycle, with most of the mineralized and dissolved P from microaggregates adsorbed onto unaggregated clay-sized particles (< 53 μm) or utilized by plants [26]. Reportedly, some studies have shown that the soil aggregate stability and size affect the soil P distribution [27][28][29]. Higher percentages of both waterextractable and Mehlich III-extractable P were observed in both the 0.50-0.25 and 0.25-0.125 mm aggregate fractions [30]. In contrast, reports have suggested that TP is the highest in small soil aggregates [31], or the TP content is uniform in soil aggregates of all sizes, whereas available P is higher in small soil aggregates [32]. Soil aggregation could reduce the loss of organic P in aggregates and increase the adsorption of inorganic P by silt and clay particles [26]. Meanwhile, the P forms in soil aggregates may vary with different particle sizes and land-use types [24,33]. For example, a study has indicated that aluminum oxide bond P (Al-P) is mainly dominated by soil aggregates < 1 mm; those of 2-8 mm were mainly iron oxide bond P (Fe-P) and calcium oxide bond P (Ca-P) [34]. Other investigators have claimed that the labile P in macroaggregates was higher under native land-use than other land uses, further confirming that soils under native use contained more Ca-bound P in macroaggregates than the disturbed soils [35]. These studies provide the first basis for the better understanding the relationship between soil aggregates and P.
To date, information on the P coll content and its loss potential from aggregates remains limited. The impact of the aggregation process on the P coll content in soils remains unclear. Moreover, the composition and structure of soil aggregates vary under different land-use management [36,37]. There are fewer large-sized aggregates present in rice soil than dryland due to long-term flooding and anaerobic conditions that cause the macroaggregates to be dispersed [36,38]. In addition, alternation between dry and wet conditions generally destroys macroaggregates and enhances the decomposition of organic carbon in paddy soils [39]. Therefore, we suspected that the content of P coll in the macroaggregates of paddy soils was less than dryland soils, mainly existing in microaggregates and small particles, with a higher loss potential.
This study mainly aimed to understand the effect of soil aggregate stability on soil P coll content and its loss potential, and to assess the core environmental factors affecting P coll in soil aggregates. Hence, we collected soil samples from 15 sites and 6 land-use types in the Yangtze River Delta region, Zhejiang Province for aggregates and P coll analysis. Firstly, we isolated the different-sized aggregates in the soil samples. Secondly, we determined the P coll , total carbon (TC), TP, total nitrogen (TN), Al, Fe, and Ca content in each aggregate size fractions. We hypothesized that (1) larger sized aggregates have higher TP and P coll content; (2) aggregates with a higher TC content have lower loss potential of P coll , and (3) land-use management with single rice has a higher loss potential of P coll .

Soil sampling and preparation
In total, soils with different land-use types ( Fig. 1) were collected from 15 sites, which were almost evenly distributed in the Zhejiang Province (an area of 1,055,000 km 2 ). Information on specific sampling points is presented in Table 1. The 15 sampling points covered six land-use types including orchards, single cropping rice, doublecropping rice, rice-rape rotation, rice-wheat rotation, and vegetables, generally established in the past 5 years.
Soil samples of 0-20 cm were collected from typical fields (long-term farmland with conventional fertilization by local farmers) in May 2018 during the second season of rotation systems and in other land-use types. Two samples, with three replicates, were obtained at intervals of 1000 m at each site with the same land-use type.
Next, the replicates were brought back to the laboratory and mixed. Then, the mixed soil samples from each site were divided into four equal parts by the diagonal quartering method and maintained for a follow-up test. All samples were air-dried and separated into two parts: one was finely milled and sieved through a 2-mm mesh to determine basic physical and chemical properties, and the other was carefully broken into small pieces manually and passed through an 8-mm sieve for aggregate separation and P coll determination.

Aggregate separation and determination
Aggregate size distribution was determined for each soil sample using a modified wet sieving method [40]. Briefly, 50 g of unground soil, passed through an 8-mm sieve, was carefully placed above a nest of three sieves (2 mm, 0.26 mm, and 0.053 mm). Then, the sieves were submerged for 20 min in 2.5 L deionized water at room temperature and oscillated 300 times for 10 min with a 30-mm amplitude to separate aggregate fractions. Thus, four aggregate fractions were obtained on each sieve: large macroaggregates (2-8 mm), small macroaggregates (0.26-2 mm), microaggregates (0.053-0.26 mm), and (silt + clay)-sized particles (< 0.053 mm) [41]. Aggregates of each size were carefully removed from the sieve and placed into a beaker. The water used for wet sieving was left to rest for 48 h, silt and clay particles were collected, and the supernatant was used to determine total dissolved P (TDP), truly soluble P (TSP), and P coll content.
All aggregates were oven-dried at 65 °C for 48 h, weighed, and placed in a zip lock bag. To obtain water-stable aggregates, the sediment concentration was subtracted from that obtained by wet sieving as sand was not considered a component of water-dispersible aggregates [40]. The sand content was determined by the following process: 5 g of the dry aggregates obtained above were weighed, dispersed into 30 mL 5 g L −1 hexametaphosphate solution, placed into an ultrasonic cleaner, and dispersed for 30 min. The suspension was then poured through a nest of sieves. The residue left on the 0.053-mm sieve represented the sand content of each sized aggregate. After collection, sand was dried at 65 °C and weighed.
Soil pH was determined with a glass electrode pH meter (PHS-3C, Shanghai) using a soil-to-water ratio of 1:5. Soil cation exchange capacity was measured with ammonium acetate (12.5 mL 1 M NH 4 OAc, 2.5 g soil) [42]. Soil and aggregate-associated TP was determined by H 2 SO 4 -HClO 4 digestion and evaluated using the molybdenum-blue colorimetric method [43]. Soil particle size distribution was determined by the hydrometric method according to an international soil texture classification standard. Soil-and aggregate-associated TC and TN were determined using an elemental analyzer (dry combustion with Vario MAX CNS, Elementar, Germany). Soil-and aggregate-associated Al, Fe, and Ca were determined by inductively coupled plasma mass spectrometry (ICP-MS) after digestion with 5 mL HNO 3 (16 M), 1 mL HClO 4 (12.4 M), and 1 mL HF (23 M) for 12 h. All reagents were acquired from Sinopharm Chemical Reagent Co., Ltd.
Colloidal P was determined as described by Ilg [44]. Briefly, 10 g of unground soil was placed into a 250-mL flask, 80 mL deionized water was added. The sample was shaken at 160 rpm and 25 °C for 24 h. The supernatant was pre-centrifuged at 3000g for 10 min to remove coarse particles. After pre-centrifugation, the supernatant was filtered with a 1-μm microporous membrane, 5 mL of the primary filtrate was discarded, and the total filtrate was collected (sample I). This suspension included the colloidal and dissolved components. The filtrate was ultracentrifuged at 300,000g for 2 h to remove colloids (Optima TL, Beckman, USA; Sample II), and the residue at the bottom of the ultracentrifuge tube demonstrated the water-dispersible colloids. The TDP in Sample I and TSP in Sample II, in the solution, were determined after digestion with acidic potassium persulfate. The concentration of P coll indicated the difference between TDP in Sample I and TSP in Sample II. Previous studies have shown that soil P through leaching and surface runoff was usually in the soluble form, that can pass through the 0.45-1 μm filter [45,46]; therefore, in the present study, TDP including P coll and TSP in aggregates was defined as the potential loss P, and the P coll in TDP was defined as

Table 1 Basic chemical and physical properties of soil, and characterization of sites located in Zhejiang Province, China
TC total carbon, TN total nitrogen, TP total phosphorus, CEC cation exchange capacity Soils S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 pH 3.  P coll loss potential. TDP in the supernatant after 10 min wet sieving was considered as easy loss P.

Calculation of water-stable aggregate (WSA) size fractions
The proportion of WSA in each size fraction was obtained from Eqs. (1) and (2), as described by Alvaro-Fuentes [47]: where i is the ith size fraction (2-8, 0.25-2, and 0.053-0.25 mm); dry soil aggregate (DSA i ) is the oven-dried mass of total, non-dispersed aggregates collected on each sieve; sand is the oven-dried mass of sand collected after dispersal in the hexametaphosphate solution on the 0.053 mm sieve; Total Soil is the oven-dried mass of soil (50 g) for aggregate separation.

Calculation of mean weight diameter (MWD) and geometric mean diameter (GMD)
The MWD and GMD of the aggregates were obtained by Eqs. (3) and (4)

Contribution of aggregate-associated P coll to TDP
The contribution rate (CR) was used to explore the impact of aggregate sizes on the P coll loss potential, calculated using Eq. (5): where Agg_CP is the concentration of aggregate-associated P coll (mg kg −1 ), TDP is the concentration of total dissolved P (mg kg −1 ), and i is the ith size fraction (2-8, 0.26-2, 0.053-0.26, and < 0.053 mm). (1)

Statistical analysis
Microsoft Excel 2016 and Origin 8.0 were used for data processing and cartography. Data were statistically analyzed using SPSS Statistics 22.0 (SPSS Inc. Chicago, USA) software. One-way ANOVA was performed using two samples from each site to examine differences of different variables as presented in Additional file 1: Tables S1, S2,  and Figures S1, S2 and Table 3. Pearson correlation analysis was used to identify the relationship between aggregate-associated P coll and other soil parameters. Stepwise linear regression was performed to evaluate the relationships between P indicators (content and loss potential of P coll ) and soil variables (pH, TP, TC, TN, C/N, Fe, Al, Ca, MWD, and GMD).

Soil and aggregate characteristics
Within the 15 soils collected, ten were acidic soils, two were neutral soils, and three were alkaline soils (  Figures S1 and S2, Table S2). Moreover, fractions of DSA and WSA (2-8 mm) in acidic soils ranged from 18.7 to 33.4% and from 10.5 to 47.1%, respectively. Conversely, fractions of DSA and WSA in alkaline soils ranged from 39.7 to 56.1% and from 38.1 to 55.6%, respectively. In contrast, size fractions of < 0.053 mm particles gradually decreased with increasing soil pH values, while the average contents of < 0.053 mm particles in acidic soils were 18.8 and 17.8%, compared with only 11.2% and 14.7% in alkaline soils, respectively.
The MWDs of acidic soils (pH < 5.5) were significantly smaller than those of alkaline soils (P < 0.05). The average MWD of acidic soils was 0.78 mm (Additional file 1: Figure S2), while that of neutral and alkaline soils was 1.36 mm. However, little difference in GMDs was observed between acidic (0.85 mm), and alkaline soils (0.91 mm).

Total and colloidal phosphorus content
Generally, the 0.26-2 mm aggregate fraction demonstrated the highest TP content, which accounted for 29.6% of the soil TP (Fig. 2a), and soil aggregates of 2-8 mm demonstrated the second-highest TP content. The TP content of (silt + clay)-sized particles was significantly lower than that of other fractions (P < 0.05, Additional file 1: Table S1), which only accounted for 19.7% of the soil TP (Fig. 2a). Moreover, TDP, TSP, and P coll contents were related to soil pH, and the highest TDP content was observed in 0.26-2 mm aggregates in most acidic soils and in (silt + clay)-sized particles in most alkaline soils (P < 0.05, Additional file 1: Table S1). In all soils, no significant difference was observed in the P coll fractions between different aggregate sizes (P > 0.05; Fig. 2b). However, the aggregate content associated with P coll was the highest in the 0.26-2 mm aggregates, and the lowest in the (silt + clay)-sized particles in acidic soils; in neutral and alkaline soils, (silt + clay)-sized particles demonstrated the highest TDP and P coll contents, followed by the 0.26-2 mm aggregates (Table 3).

Loss potential of colloidal phosphorus
After wet sieving, about 0.16-1.87% of the soil TP was lost in the supernatant as TDP, and P coll accounted for 8.5-84.1% of the TDP ( Table 2). The proportion of the easy loss P content in the various soils differed due to variations in soil physicochemical properties.
In general, the P coll loss potential gradually decreased as the size of the soil aggregates increased (Fig. 2c). The P coll loss potential was the lowest in 2-8 mm and 0.26-2 mm aggregates, with P coll accounting for 52.6% and 60.6% of TDP, respectively. However, the P coll loss potential of (silt + clay)-sized particles was the highest, with P coll accounting for 75.3% of TDP (Fig. 2a). The CR value of (silt + clay)-sized particles was mostly larger than that of the other aggregate sizes, except for S7, S11 and S15; the CR of larger macroaggregates was lowest in most soils (Table 3).
Considering different land-use types, the P coll content in the rice-dry land rotation and vegetable (VE) soils was significantly higher than that in paddy and orchard soils, regardless of different aggregate sizes (Fig. 3a). Higher P coll content also was found in macroaggregates (> 0.26 mm), except for the orchard soil (Fig. 3a). The colloidal P content in macro-and microaggregates in the paddy soils (including single cropping rice, and double-cropping rice) accounted for a relatively lower proportion of TDP (P < 0.05) than that in other land-use types (Fig. 3b). Overall, it accounted for only 39.3, 44.1, and 40.2% for the 2-8 mm, 0.26-2, and 0.053-0.26 mm aggregates in rice fields, respectively. Colloidal P in (silt + clay)-sized particles (< 0.053 mm) in paddy soils accounted for a TDP proportion as high as 83.0%, significantly higher than that observed in the orchard and rice-dryland rotation systems (Fig. 3b). This indicated that the loss potential of P coll was dominated by fine-grained and (silt + clay)-sized particles in paddy soils. However, the P coll carried by all sized aggregates in dryland and rice-dryland rotation systems (orchard, rice-rape rotation, vegetable, and rice-wheat rotation) exceeded 50% of the TDP.
The P coll loss potential (P coll in TDP) in total size of the soil aggregate negatively correlated with the TC, TN, pH, Ca, C/P, C/N ratios of aggregates (Table 4, P < 0.05). Considering different sizes, TC and TN significantly and negatively correlated with P coll loss potential in 0.26-to 2-mm-sized aggregates. Except for (silt + clay)-sized particles, P coll loss potential was significantly and positively correlated with Al (P < 0.05). P coll loss potential in 2-to 8-mm-sized aggregates was significantly and positively correlated with Fe (P < 0.05). However, P coll loss potential in 0.053-to 2-mm-sized aggregates negatively correlated with Ca (P < 0.05). The results of the stepwise regression demonstrated that the P coll loss potential can be described by TC, Fe, MWD, and GMD as follows: colloidal P/TDP = 0.478 − 0.028TC + 0.015Fe + 0.413MWD −0.363GMD , (R 2 = 0.539, P < 0.001, Table 4).

Total P in aggregates
In our study, we observed that the TP content was the highest in macroaggregates of the 15 soils, while the TP content in the (silt + clay) particles was the lowest.
These results indicated that soil P was mainly carried by larger aggregates, which confirmed our hypothesis and were in line with those previously reported [28,49,50]. For examples, some scholars have observed that P tends to concentrate in large WSAs in longterm fertilization experiments in a reddish paddy soil [49], and others claim that aggregate-associated total organic C, TN, and TP are preferentially enriched in large WSAs (4.76-2.0 mm) [50], reporting that the TP proportion increased with increasing aggregate size for native lands [35]. Higher P levels may be associated with higher levels of TC and TN in large aggregates [50,51]. Macroaggregates ([(Cl-P-OM)x]y) are usually formed by organic matter, clay (Cl) and multivalent ions of P, and other substances [52,53]. Organic matter (and associated P) is protected within stable aggregates against microbial degradation [54]. Moreover, it has been shown that the organic P forms that accumulate in soils are less available to enzymatic hydrolysis when bound to mineral surfaces [55,56]. On the other hand, the aggregation promoted by organic matter counteracts the dispersion of the small mineral particles (mostly Fe and Al-(hydr) oxides where P is retained [57,58]. Furthermore, this was validated by the positive correlation between aggregate-associated TP and TC, and the significant correlation between aggregate-associated TP and TN observed in this study (Additional file 1: Tables S2).

Table 2 The total dissolved P (TDP), and truly soluble P (TSP), colloidal P, colloidal P/TDP, and ratio of TDP to soil total P (TP) in the supernatant of different soils after wet sieving
Data represent the average of three replicates ± standard deviations TDP (mg kg −1 ) TSP (mg kg −1 ) Colloidal P (mg kg −1 ) Colloidal P/TDP (%) TDP/TP (%)  Table 3 The colloidal P content (mg kg −1

) of different-sized soil aggregates and the contribution rate of aggregates to colloidal P (CR) in the total 15 soils
Letters in different-sized aggregates indicated significant difference at P < 0.05 level Size of aggregates (mm) S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 Colloidal P

Colloidal P content in aggregates
In this study, although no significant difference was observed in the P coll fractions between different aggregate sizes in all soils, we observed that TDP, TSP, and P coll contents in soil aggregates were significantly related to soil pH. The TDP and P coll contents were high in macroaggregates of acidic soils, and low in microaggregates and silt + clay particles; however, contrasting results were observed in alkaline soil aggregates. This was not consistent with our hypothesis which could be attributed to soil clays being mostly negatively charged, but the aggregation of other soil colloids strongly depend on their surface charge, being favored when approaching their point of zero charge [59]. Under acidic conditions, the protonation of Fe-, Al-oxides, and organic matter in colloids results in positive charges, leading to their association with soil particles. However, the dissociation of Fe-and Al-hydroxyl and humic functional groups (R-COOH, R-CH 2 -OH, R-OH) under alkaline conditions results in a negative charge in the colloid [59], which promotes the release of fine particulate P and colloidal substances, thus increasing the TDP and P coll contents in small-sized aggregates and particles. On the other hand, the lower pH enhanced the adsorption of organic matter on the fine particles and masked the inorganic mineralogy, resulting in fewer chances of P carried by Fe-and Al-oxides [60][61][62]. In addition, we observed that the P coll content in soil aggregates positively correlated with the aggregate-associated Al and Fe content. This was attributed to the greater contents of Fe and Al in acidic soil than in alkaline soil, and that the presence of Al and Fe oxides may have enhanced the adsorption of P and stabilization of soil aggregates [15,31,63]. Al and Fe oxides have been recognized as important carriers of P coll [64][65][66][67].

Loss potential of colloidal P in aggregates
We observed that the CR value of (silt + clay)-sized particles was larger than that of aggregates of other sizes, while the CR of large macroaggregates was the lowest in most soils, indicating that the (silt + clay)sized particles contribute more to the P coll loss potential, while macroaggregates immobilized soil P coll . Colloidal P is highly bound to Fe and Al on the surface of macroaggregates [15,68], enabling the formation of a stable composite structure that can resist the shear force of pore flow [64,65]. Similarly, soil macroaggregates can increase the adsorption of P coll on the surface and reduce its mobility [26]. The ratio of P coll to TDP reflects the release potential of P coll in soil aggregates to soil solution. The high value indicates the high loss potential of P coll . We observed a negative correlation between the TC and TN content and P coll /TDP in aggregates, indicating that higher the TC and TN content, the less likely release of P coll from the soil aggregates, which was further confirmed by the negative correlation between TC and P coll /TDP in the regression model (Table 4). Studies had shown that C and N are important carriers of P coll (most organics act as organic colloidal complexes) [28,69,70], and the organic matter could stabilize Al/Fe colloids [71]. Therefore, increasing the carbon content in soil aggregates could OR orchard soils (n = 3), RICE rice fields including single-and double-cropping soils (n = 6), RDR rice and dryland rotation including rice-rape rotation and rice-wheat rotation soils (n = 3), VE vegetable soils (n = 3); capital letters indicated significant difference among different sizes of aggregate at P < 0.05 level. Lowercase indicated significant difference of the same size of aggregate among land-use types at P < 0.05 level be an important strategy to reduce the migration of P coll in the soil. The ratio of P coll to TDP for all macroaggregates in paddy soils was lower, which may be related to the long-term flooding of rice fields. Notably, the lower loss potential of P coll in macroaggregates of paddy soils does not imply that it is less likely to be lost to water bodies. However, this could indicate that P coll has been lost into water and discharged into the water body through the channels during rice seasons, particularly when under long-term flooding [72]. Moreover, flooding resulted in an anaerobic state in soil aggregates, and Fe bound to colloids and aggregates was reduced, resulting in excessive release and loss of P coll [15]. We observed that aggregates of all sizes in dryland and rice-dryland rotation systems carried higher P coll loss potential. Conversely, in paddy soils, the loss potential of P coll in macro-and microaggregates was lower than 50% of TDP. However, P coll in (silt + clay)-sized particles (< 0.053 mm) in paddy soils accounted for a TDP proportion as high as 83%, significantly higher than that observed in orchard and rice-dryland rotation systems. Therefore, we believe that P coll in paddy soil is mainly carried by small silt and clay particles, which may indicate that the loss of P coll is markedly severe in paddy soils.

Conclusions
This study discussed the relationship between P coll and soil aggregates, providing a potential solution for soil P loss in agricultural systems. The soil P coll content was affected by multiple factors including soil Al, Fe, and TP, and the distribution of P coll in aggregates can be regulated by altering aggregate sizes and soil pH. As small macroaggregates contributed the most to the immobilization of P coll , it is important to increase this fraction in the soil. To this end, soil organic matter must be improved owing to its crucial role in promoting soil aggregation and reducing the risk of P coll loss potential. Paddy soils, due to their high P coll /TDP ratio, are particularly at risk of high P coll loss through smaller particles and worthy of further attention.

Table 4 Results from correlation analyses and stepwise linear regressions of colloidal P and colloidal P/TDP in differentsized aggregates with soil aggregate-associated mean weight diameter (MWD), geometric mean diameter (GMD), pH, total carbon (TC), total nitrogen (TN), C/P, C/N, Al, Fe, and Ca
Total total soil aggregate sizes * P < 0.05. ** P < 0.01