Structural Equation Modeling of Phosphorus Transformations in Soils of Larix principis-rupprechtii Mayr. Plantations

: Understanding the soil phosphorus (P) cycle is a prerequisite for the sustainable management of land resources. The sequential-extraction method was used to determine P fractions in 513 soils of Larix principis-rupprechtii Mayr. plantations. With these data, this study applied structural equation modeling to evaluate the interaction between various soil P fractions. Quantitative analysis was conducted on the importance of different soil P pools and P transformation pathways on soil P availability in a larch plantation. Our study showed that soluble inorganic P (Pi) was directly positively affected by labile Pi, labile organic P (Po), secondary mineral P, and primary mineral P, and was directly negatively affected by moderately labile Po. Soluble Pi was not directly affected by occluded P. The primary mineral P ( β = 0.40) had the greatest total impact on soluble Pi, followed by secondary mineral P ( β = 0.32) and labile P (labile Pi and Po, β = 0.31), and then occluded P ( β = 0.11), with the total impact of moderately labile Po being relatively small ( β = − 0.06). In summary, this study reveals the important roles of soluble Pi in P transformations and in determining overall P availability in soils, as well as the extensive effects of weathering on soil P dynamics in L. principis-rupprechtii plantations.


Introduction
Afforestation, the conversion of nonforest lands to plantation forests, contributes to ecosystem restoration. It provides important ecosystem services and functions such as the production of wood and fiber, thereby affecting the storage of carbon (C), nitrogen (N), and phosphorus (P) [1,2]. Afforestation has expanded rapidly worldwide in recent decades [3]. From 1990 to 2015, global planted forest area increased from 168 to 278 million ha and continues to grow at a rate of 4.5 million ha per year [4]. Over the last two decades, soil C and soil N in plantations have been extensively studied at different scales [5,6]. Soil P is an essential nutrient element for plant growth [7]. The dynamics of soil P directly control its bioavailability [8], further affecting many key functions of forest ecosystems, such as biomass production and the soil C sequestration process [9][10][11]. Hence, an improved understanding of the dynamics and availability of P in soil is essential when evaluating current or potential soil productivity. Recommendations can be made for sustainable management of land resources, and awareness of climate change and its feedback can be raised.
A detailed investigation of P dynamics and bioavailability requires the separation and identification of different forms of P in soils [12]. Sequential fractionation is one of the most commonly used methods to study soil P dynamics owing to its advantages of low operating cost, small amount of samples needed, and practicality [13,14]. The P fractionation procedure of Hedley et al. [15] and its modification by Tiessen and Moir [16] are the most commonly used [17][18][19]. Based on the different solubility and mobility of soil inorganic P (Pi) and organic P (Po), the procedure can be used to extract each soil P fraction sequentially with a series of chemical reagents [15,20]. Each of these fractions plays an important role in the transformation of soil P. However, a separate soil P fractionation cannot evaluate the interaction among various P fractions.
The use of structural equation modeling (SEM) is common in the natural sciences, especially in ecology and evolutionary biology [21,22]. In soil science, SEM is used to study the degree of interaction between P fractions in soil. Tiessen et al. [23] constructed a path model to describe the transformations between Hedley P fractions in soils of different pedogeneses in the United States. The model has been adopted or modified in subsequent studies, primarily those involving agricultural soils [24,25]. With this model, researchers can infer whether one soil P fraction can be transformed into another soil P fraction directly or indirectly. Another fundamental feature of SEM is that the hypothetical model has two main groups of variables, namely, latent (unmeasured) and observed variables. The former represent the potential underlying causes, whereas the latter serve as indicators of the effects of the latent factors [26]. One of the advantages of SEM is its ability to rank the descriptive ability of different models to enable comparisons among models [27]. Accordingly, Gamarodrigues et al. [28] established an exploratory SEM with latent variables to study the P cycle in unfertilized tropical soils. Sales et al. [29] redefined this model. These authors evaluated a hypothetical model of the P cycle based on the relationship among five latent variables (P pools) with multiple observed variables (P fractions). They determined the interactions among the P pools and identified which pools act as P sinks or sources in unfertilized tropical soils. Owing to the lack of global research on the transformation between soil P pools, Hou et al. [30] utilized a validated and improved SEM of soil P transformation at the global scale (including Chinese data) to preliminarily reveal the core role of solid-phase labile Pi, rather than the traditional recognized soluble Pi fraction, in soil P transformation. They also aimed to demonstrate the importance of solid-phase P transformation.
The SEM method has great significance for verifying soil P transformation models and conducting in-depth research on soil P transformation [31]. However, differences in soil and vegetation types may result in global-scale assessments having limited value in extrapolating information regarding specific regions and tree species [32]. In almost every part of the P cycle, the direction and magnitude of P transformation in different regions or different tree species also differ [3,33]. Thus, a deeper understanding of the soil P transformation of specific tree species in specific areas is necessary to provide more general and practical recommendations for forest management based on local conditions and weather characteristics.
Larix principis-rupprechtii Mayr. is the main fast-growing timber species in North China. It plays an important role in timber production, water and soil conservation, ecological environment improvement, and other areas [34]. However, excessive initial planting density and untimely tending and thinning have led to an increase in canopy density and intensified competition, resulting in a poor forest environment and the slow development of understory vegetation, subsequently reducing soil fertility and forest productivity [35]. Studies have shown that soil microbial biomass P and labile organic P significantly decrease after natural secondary forests are converted into larch plantations. Natural forests are also reportedly more beneficial for maintaining soil organic P fertility than larch plantations [36,37]. Thus, a better understanding of soil P transformation in L. principis-rupprechtii plantations is needed to establish effective management strategies, resolve possible causes, and improve soil P availability.
Based on 513 soil samples from an L. principis-rupprechtii plantation in Taiyue Mountain, Shanxi Province, the present study aimed to (1) evaluate the interaction of soil P fractions; and (2) quantify the relative importance of different soil P pools and P transformation pathways in shaping soil P availability.

Description of the Study Area
The research area was located at the national positioning observation and research station (36 • 40 N, 112 • 04 E) of the forest ecosystem in Taiyue Mountain, Shanxi Province, China. The climate in this region belongs to the warm, temperate, semihumid continental monsoon climate, with annual sunshine of 2600 h. The average annual temperature is around 11 • C, with the highest temperature (around 26 • C) occurring in July and the lowest temperature (around −23 • C) occurring in January. The average annual precipitation is about 600 mm, with over 60% of the annual precipitation occurring in the three months of June, July, and August. The soil type in the study area is mainly brown soil and cinnamon soil [38], and the thickness of the hillside soil layer is 30-50 cm.
In the study area, the vegetation type is primarily L. principis-rupprechtii in a plantation [39], of which the near-mature stand (31-40 a) is the main type. Secondary forests such as Batula platyphylla and Salix caprea are scattered in low-altitude areas, often forming mixed coniferous and broad-leaved forests with younger L. principis-rupprechtii (<30 a). Shrubs and herbs primarily include Corylus mandshurica, Spiraea pubescens, Rosa xanthina, Carex lanceolata, Rubia chinensis, and Thalictrum petaloidium.

Experimental Design and Soil Sampling
In August 2020 and April 2021, sampling methods representing different vegetation types, terrain conditions, and uniform distributions were used to select sample plots (Table 1). A total of 33 (6 + 27) fixed plots of 20 m × 30 m were set up. Six soil samples were collected in August 2020; April, June, August, and October 2021; and August 2022. In each plot, nine sampling sites were selected along an S-shaped route, and soil samples were taken at three (0-10, 10-20, and 20-30 cm) different depths using an auger with a diameter of 5 cm. Nine soil samples at the same depth in each plot were mixed to form a composite soil sample. After removing stones, roots, and other impurities, approximately 1 kg of soil sample was taken by quartering and brought back to the laboratory for soil analysis.

Phosphorus Fractionation
The fractionation scheme for the different soil P fractions was conducted using the Tiessen and Moir [16] extraction method (Figure 1), that is, 0.5 g air-dried soil was sequentially extracted with resin strip, 0.5 M NaHCO 3 , 0.1 M NaOH, 1.0 M HCl, and hot concentrated HCl, and then concentrated H 2 SO 4 and H 2 O 2 were used to digest the soil residue. R-Pi represented the soluble Pi pool, which is immediately accessible to plants. Bic-Pi and Bic-Po were considered labile Pi and Po pools, respectively, which can be absorbed and utilized by plants (or mineralized). OH-Pi referred to the secondary mineral P pool, while OH-Po referred to the moderately labile Po pool. They were composed of amorphous and partially crystalline Al and Fe phosphates, with low plant utilization. Dil.HCl-Pi was a Pi fraction combined with Ca, which referred to the primary mineral P pool. To better analyze and compare with the results of the global-scale research [30], the sum of Conc.HCl-Pi, Conc.HCl-Po, and Res-Pi was classified as "Residue P fraction," and theyrepresented the occluded P pool.  [30], the sum of Conc.HCl-Pi, Conc.HCl-Po, and Res-Pi was classified as "Residue P fraction," and theyrepresented the occluded P pool.

Statistical Analysis
Prior to statistical analysis, all data were subjected to normality testing and logarithmic transformation if necessary. Pearson's correlation analysis was conducted on the correlation between soil P fractions using SPSS 21.0 (IBM, Chicago, IL, USA). Path analysis was performed by building a full-path diagram in AMOS 22.0 (IBM, Chicago, IL, USA) with Bic-Pi and Po, OH-Pi and Po, Dil.HCl-Pi, and Residue-P as predictor variables and with R-Pi as the response variable.
An exploratory SEM was developed for the transformation of different soil P fractions using AMOS 22.0 (IBM; SPSS Inc., Chicago, IL, USA). The following assumptions were made: (1) Soluble Pi was primarily influenced and supplemented by labile P pools (inorganic and organic). Mineralization was also considered as a rate-limiting process in the soil P cycle. (2) The net transformation of soil P was from solid to liquid phase, from low-soluble to high-soluble P, from organic to inorganic P, and from primary mineral P to soluble or labile P. Firstly, a prior path model was constructed that included all theoretical causal relationships among soil P fractions ( Figure S1). By iteratively deleting the nonsignificant relationships among variables, the best-fit model was found, that is, p > 0.05, RMSEA ≤ 0.05, and the lowest AIC value. The best-fit model was further respecified by introducing latent variables.

Statistical Analysis
Prior to statistical analysis, all data were subjected to normality testing and logarithmic transformation if necessary. Pearson's correlation analysis was conducted on the correlation between soil P fractions using SPSS 21.0 (IBM, Chicago, IL, USA). Path analysis was performed by building a full-path diagram in AMOS 22.0 (IBM, Chicago, IL, USA) with Bic-Pi and Po, OH-Pi and Po, Dil.HCl-Pi, and Residue-P as predictor variables and with R-Pi as the response variable.
An exploratory SEM was developed for the transformation of different soil P fractions using AMOS 22.0 (IBM; SPSS Inc., Chicago, IL, USA). The following assumptions were made: (1) Soluble Pi was primarily influenced and supplemented by labile P pools (inorganic and organic). Mineralization was also considered as a rate-limiting process in the soil P cycle. (2) The net transformation of soil P was from solid to liquid phase, from low-soluble to high-soluble P, from organic to inorganic P, and from primary mineral P to soluble or labile P. Firstly, a prior path model was constructed that included all theoretical causal relationships among soil P fractions ( Figure S1). By iteratively deleting the nonsignificant relationships among variables, the best-fit model was found, that is, p > 0.05, RMSEA ≤ 0.05, and the lowest AIC value. The best-fit model was further respecified by introducing latent variables.

Subsection Relationships between Soil P Fractions
Total P fell in the range of 157.04-1148.92 mg kg −1 . Among all P fractions, Residue-P was the largest, followed by OH-Po. R-Pi and Bic-Pi were the smallest P fractions (Table 2). Bivariate correlations between total P and P fractions were all statistically significant (p < 0.01) except for the correlation between Dil.HCl-Pi and OH Po (Table 3). R-Pi was more closely correlated with Dil.HCl-Pi (r = 0.55, p < 0.01) than with other P fractions (r = 0.16-0.45, p < 0.01) ( Table 3).

Structural Equation Modeling
The fit between the path model of Tiessen et al. [23] (model A) and our data was not satisfactory (p < 0.001, RMSEA = 0.22; Figure 3a). The large χ 2 value (286.17) and χ 2 /df ratio (26.02) indicated that some important relationships between soil P fractions were not accounted for. The fit with the modified model of Gama-Rodrigues et al. [28] and Sales et al. [29] (model B) was also unsatisfactory (p < 0.001, RMSEA = 0.23; Figure 3b). Moreover, the fitting between the global-scale model of Hou et al. [30] (model C) and our L. principisrupprechtii plantations data was not satisfactory (p < 0.001, RMSEA = 0.13; Figure 3c).

Structural Equation Modeling
The fit between the path model of Tiessen et al. [23] (model A) and our data was not satisfactory (p < 0.001, RMSEA = 0.22; Figure 3a). The large χ 2 value (286.17) and χ 2 /df ratio (26.02) indicated that some important relationships between soil P fractions were not accounted for. The fit with the modified model of Gama-Rodrigues et al. [28] and Sales et al. [29] (model B) was also unsatisfactory (p < 0.001, RMSEA = 0.23; Figure 3b). Moreover, the fitting between the global-scale model of Hou et al. [30] (model C) and our L. principis-rupprechtii plantations data was not satisfactory (p < 0.001, RMSEA = 0.13; Figure 3c).

Figure 2.
Path diagram illustrating the contributions of Bic-Pi, Bic-Po, OH-Pi, OH-Po, Dil.HCl Pi, and Residue-P fractions to R-Pi fraction in soils. All data are log10 transformed. The direct effects of these P fractions R-Pi fraction are represented by single-headed arrows. The covariances between P fractions besides R-Pi fraction are represented by double-headed arrows. The numbers on the single arrows are standardized regression weighted values, and italics indicate no significant difference. The bold value close to R-Pi fraction (0.52) indicates the percentage of variance in R-Pi fraction explained by the model (R 2 ). This model is constructed based on the hypothesis that all soil P fractions besides the R-Pi fraction covary with each other and can directly contribute to resin Pi fraction.

Structural Equation Modeling
The fit between the path model of Tiessen et al. [23] (model A) and our data was not satisfactory (p < 0.001, RMSEA = 0.22; Figure 3a). The large χ 2 value (286.17) and χ 2 /df ratio (26.02) indicated that some important relationships between soil P fractions were not accounted for. The fit with the modified model of Gama-Rodrigues et al. [28] and Sales et al. [29] (model B) was also unsatisfactory (p < 0.001, RMSEA = 0.23; Figure 3b). Moreover, the fitting between the global-scale model of Hou et al. [30] (model C) and our L. principisrupprechtii plantations data was not satisfactory (p < 0.001, RMSEA = 0.13; Figure 3c).  from Gama-Rodrigues et al. [28] and Sales et al. [29], and Model C (c) was constructed according to Hou et al. [30]. The data were converted logarithmically. The numbers on the single arrows are standardized regression weighted values. The bold numbers next to the endogenous variables are the square multiple correlations. In model B, error variance of HCl Pi and residue-P are set at 0.1 according to Sales et al. [29]. In model C, empirical regression weight and measurement error of each soil P fraction are set at 1.0 and 0 according to Hou et al. [30]. Overall fit of model A: Model D (p = 0.63, RMSEA < 0.001; Figure 4a) was the optimal model redetermined according to the prior path model ( Figure S1). On the basis of model D, model E further respecified the structural model formed by introducing the latent variable (soil P pool), that is, each soil P fraction was used as the index of a soil P pool (Figure 4b). This model had the same degree of fit as model D. Model E combined sequentially extracted soil P fractions with functional soil P pools and solved the connection problems between the fraction and the pool. Such problems included the measurement errors of soil P fractions and the incomplete extraction of soil P pool by reagents [40]. Figure 3. Test of three previous soil P transformation models using data from L. principis-rupprechtii plantations. Model A (a) was constructed according to Tiessen et al. [23], Model B (b) was modified from Gama-Rodrigues et al. [28] and Sales et al. [29], and Model C (c) was constructed according to Hou et al. [30]. The data were converted logarithmically. The numbers on the single arrows are standardized regression weighted values. The bold numbers next to the endogenous variables are the square multiple correlations. In model B, error variance of HCl Pi and residue-P are set at 0.1 according to Sales et al. [29]. In model C, empirical regression weight and measurement error of each soil P fraction are set at 1.0 and 0 according to Hou et al. [30]. Overall fit of model A: Model D (p = 0.63, RMSEA < 0.001; Figure 4a) was the optimal model redetermined according to the prior path model ( Figure S1). On the basis of model D, model E further respecified the structural model formed by introducing the latent variable (soil P pool), that is, each soil P fraction was used as the index of a soil P pool (Figure 4b). This model had the same degree of fit as model D. Model E combined sequentially extracted soil P fractions with functional soil P pools and solved the connection problems between the fraction and the pool. Such problems included the measurement errors of soil P fractions and the incomplete extraction of soil P pool by reagents [40].  The path coefficients between soil P pools in model E were the same as those between their corresponding indicators (i.e., soil P fractions in model D) (Figure 4). Similar to the path analysis (Figure 2), Model E showed that soluble Pi had significant direct positive effects on labile Pi (standardized β = 0.31, p < 0.001), labile Po (β = 0.31, p < 0.001), secondary mineral P (β = 0.19, p < 0.001), and primary mineral P (β = 0.36, p < 0.001), whereas moder- The path coefficients between soil P pools in model E were the same as those between their corresponding indicators (i.e., soil P fractions in model D) (Figure 4). Similar to the path analysis (Figure 2), Model E showed that soluble Pi had significant direct positive effects on labile Pi (standardized β = 0.31, p < 0.001), labile Po (β = 0.31, p < 0.001), secondary mineral P (β = 0.19, p < 0.001), and primary mineral P (β = 0.36, p < 0.001), whereas moderately labile Po (β = −0.19, p < 0.001) had significant direct negative effects (Figure 4b).

Modeling the Pathways of Soil P Transformations
Soil soluble Pi is a very dynamic P pool that can be greatly affected by the amount of soil solution and by the short-term changes in plant and soil microbial activities and leaching [41]. Soil solution is the most important transport medium of essential chemical elements for life [42]. Thus, the conceptual model of the soil P cycle generally assumes that soluble Pi mediates the transformation of most other P pools in soil [43]. Our results were consistent with this view, that is, soluble Pi was directly affected by labile Pi, labile Po, secondary mineral P, moderately labile Po, and primary mineral P. It was not directly affected by the occluded P, indicating that soluble Pi cannot be directly transformed from the low-soluble P pools, consistent with the conceptual model of Tiessen et al. [23] and Hou et al. [30].
Labile Pi was directly affected only by the secondary mineral P in the model of Tiessen et al. [23], and this P pool was not included in the final model of Gama-Rodrigues et al. [28] and Sales et al. [29]. However, in the global-scale SEM of Hou et al. [30], labile Pi was directly related to all other soil P pools. In this study, model E showed that labile Pi was significantly and directly associated with moderately labile Po and the major mineral P. This finding may indicate that the phosphate released by soil minerals during weathering and mineralization had a significant adsorption effect [44]. Given that weathering and mineralization occurred at the interface between the solution and the mineral [42], the released phosphate could enter the soil solution or be adsorbed by the soil minerals and become soluble Pi or labile Pi. The direct influence of occluded P on labile Pi was obtained in the model of Hou et al. [30]. The authors pointed out the importance of solid-phase P transformation, although the transformation mechanism remains unclear.

Modeling the Pathways of Soil P Transformations
Soil soluble Pi is a very dynamic P pool that can be greatly affected by the amount of soil solution and by the short-term changes in plant and soil microbial activities and leaching [41]. Soil solution is the most important transport medium of essential chemical elements for life [42]. Thus, the conceptual model of the soil P cycle generally assumes that soluble Pi mediates the transformation of most other P pools in soil [43]. Our results were consistent with this view, that is, soluble Pi was directly affected by labile Pi, labile Po, secondary mineral P, moderately labile Po, and primary mineral P. It was not directly affected by the occluded P, indicating that soluble Pi cannot be directly transformed from the low-soluble P pools, consistent with the conceptual model of Tiessen et al. [23] and Hou et al. [30].
Labile Pi was directly affected only by the secondary mineral P in the model of Tiessen et al. [23], and this P pool was not included in the final model of Gama-Rodrigues et al. [28] and Sales et al. [29]. However, in the global-scale SEM of Hou et al. [30], labile Pi was directly related to all other soil P pools. In this study, model E showed that labile Pi was significantly and directly associated with moderately labile Po and the major mineral P. This finding may indicate that the phosphate released by soil minerals during weathering and mineralization had a significant adsorption effect [44]. Given that weathering and mineralization occurred at the interface between the solution and the mineral [42], the released phosphate could enter the soil solution or be adsorbed by the soil minerals and become soluble Pi or labile Pi. The direct influence of occluded P on labile Pi was obtained in the model of Hou et al. [30]. The authors pointed out the importance of solid-phase P transformation, although the transformation mechanism remains unclear.

Quantifying the Pathways of Soil P Transformations
The SEM for the transformation of soil P fractions in L. principis-rupprechtii plantations confirmed the soluble Pi supplementation mechanism proposed in the prior path P cycle, that is, soluble Pi was influenced more by the more labile P pools than by the stable P pools, such as occluded P. The direct influence of labile Pi on soluble Pi was consistent with the concept that labile Pi can be quickly exchanged with soluble Pi and can serve as a short-term plant-effective P pool [18]. The influences of labile Pi and labile Po on soluble Pi were consistent (β = 0.31), indicating that mineralization was not necessarily the rate-limiting process of the soil P cycle [45]. This finding was similar to the hypothesis model of Tiessen et al. [23], who also believed that labile Po has a direct positive influence only on soluble Pi.
Unlike labile P, secondary mineral P had a direct positive impact on soluble Pi and an indirect positive impact through labile Pi (β = 0.43 × 0.31 = 0.13). The direct influence of secondary mineral P on soluble Pi was smaller than that on labile Pi. This finding indicated that solid Pi was more likely to occur through the solid-phase P transformation transfer of low-soluble Pi to high-soluble Pi [46] than by dissolution and desorption transferred to the soil solution [47]. Similarly, moderately labile Po had direct and indirect influences on soluble Pi. However, its direct influence was negative, whereas the indirect influence through labile Po or labile Pi was positive (β = 0.31 × 0.31 + 0.11 × 0.31 = 0.13). The direct negative influence was due to organisms taking up phosphate from the soil solution and forming organic states in combination with amorphous and crystalline Fe/Al [48]. As soluble Pi became depleted, it was quickly replenished with labile P, which can be further replenished by moderately labile Po through mineralization or solid-phase P diffusion [49]. This finding may explain the indirect positive influence of moderately labile Po on soluble Pi through labile Po or labile Pi.
Like moderately labile Po, primary mineral P had direct and indirect influences (through labile Po or labile Pi) on soluble Pi. Compared with other P pools, the standardized direct effect and total effect of primary mineral P on soluble Pi was the largest, indicating that the weathering of the soil of L. principis-rupprechtii Mayr. plantations strongly affected soluble Pi. Wilson et al. [50] believed that weathering had a broader impact on the ecological availability of P than previously recognized. It affected the absolute content of P and involved a series of complex relationships such as soil mineralogy, chemical and physical weathering, and bioclimate zones. Occluded P did not directly influence soluble Pi, but it could indirectly influence soluble Pi through either secondary mineral P or moderately labile Po. This phenomenon can be due to two factors. First, the extremely low solubility of occluded P limits its desorption and dissolution, microbial immobilization, and mineralization processes [19]. Second, occluded P transformed steady-state P into low-solubility P through solid-phase P transformation.
Model E of this study was a comprehensive and exploratory analysis of the important pathways of P transformations in soils of L. principis-rupprechtii Mayr. plantations. Changes in soil P can be regulated by factors such as precipitation and temperature, local site properties, and even spatial variation in soil P content [51,52]. Thus, the significance and strength of P transformation pathways may vary with temporal and spatial scales, and the significance and intensity of P transformation pathways may vary with tree species and spatial scales. Our research findings suggested that soluble Pi played a central role in soil P transformation, contradicting the results of Hou et al. [30]. Therefore, compared with global-scale research, regional-scale investigations on soil P transformation may exhibit variations in direction and magnitude. This result confirmed the limited applicability of global-scale research in guiding more rational and scientifically informed management practices for L. principis-rupprechtii plantations.

Conclusions
SEM analysis based on the soil P fraction of L. principis-rupprechtii plantation showed that soluble Pi was directly related to most other soil P pools, which determined the availability of soil P. Labile Pi and labile Po had a direct effect on soluble Pi. The direct effect of secondary mineral P on labile Pi was greater than that on soluble Pi. Moderately labile Po and primary mineral P had a direct effect on soluble Pi and indirect effects through labile Pi and labile Po. Occluded P had no direct effect on soluble Pi. In terms of total influence, primary mineral P had the greatest influence on soluble Pi, followed by secondary mineral P and labile P, and then occluded P. Moderately labile Po had a weaker influence on soluble Pi. The SEM method was highly significant to the in-depth study of soil P transformation, revealing the core position of soluble Pi in the process of soil P transformation and the extensive influence of weathering on soil P dynamics. The proposed P transformation model of an L. principis-rupprechtii plantation can serve as a part of the soil P cycle model of the ecosystem in the study area. This work can guide future efforts concerning biogeochemical P cycling.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/f14091811/s1, Figure S1: A priori path model of transformations between soil P fractions.