Effects of Companion Tree Species on Soil Fertility, Enzyme Activities, and Organic Carbon Components in Eucalyptus Mixed Plantations in Southern China

Abstract

The long-term monoculture of Eucalyptus plantations in southern China has raised ecological concerns, prompting a shift towards mixed-species plantations as a sustainable alternative. This study investigates the mechanisms by which companion tree species enhance soil functionality in subtropical red soil regions. A field experiment compared a pure Eucalyptus (CK) plantation with three mixed-species plantations: Eucalyptus × Mytilaria laosensis (A × M), Eucalyptus × Magnolia hypolampra (A × H), and Eucalyptus × Michelia gioii (A × X). Comprehensive soil analyses were conducted at three soil depths (0–20 cm, 20–40 cm, and 40–60 cm) to assess chemical properties, enzyme activities, and humus components, and soil organic carbon (SOC) molecular structure was characterized by Fourier-Transform Infrared Spectroscopy (FTIR), with the relationships quantified using structural equation modeling (SEM) to test predefined causal hypotheses. The results showed that A × H significantly boosted topsoil fertility (e.g., OM: 46.61 g/kg), while A × M enhanced the recalcitrant organic carbon (ROC: 35.29 g/kg), indicating superior carbon sequestration potential. The FTIR analysis revealed species-specific alterations in SOC chemistry, such as increased aromatic compounds in A × H/A × X. The SEM analysis demonstrated that the latent variable “Humus” (reflected by LOC and ROC) directly and positively influenced the latent variable “Soil Fertility” (reflected by pH, OM, and AP; path coefficient: 0.62). In contrast, the latent variable “Organic Components” (reflected by specific FTIR functional groups) exhibited a significant direct negative effect on “Soil Fertility” (−0.41). The significant pathway from “Organic Components” to “Enzymatic Activity” (0.55*) underscored the role of microbial mediation. The study concludes that mixed plantations, particularly with Mytilaria laosensis (A × M), improve soil health through an “organic input–microbial enzyme response–humus formation” pathway, offering a scientific basis for sustainable forestry practices that balance productivity and ecological resilience.

1. Introduction

As the core component of terrestrial ecosystems, forests store more than two-thirds of the organic carbon of the entire terrestrial ecosystem and are pivotal for reducing greenhouse gas emissions in response to global climate change [1]. Eucalyptus robusta, a timber tree species, was introduced to China from Australia and has been widely planted in the subtropical and tropical regions of southern China due to its strong adaptability and wide range of uses [2,3]; as of 2022, the planted area had reached 5,467,400 ha [4]. Although the long-term, large-scale planting of pure Eucalyptus forests, as well as excessive fertilization and multi-generation succession planting [5], has made remarkable contributions to regional economic development and alleviation of farmers’ poverty, it has also triggered a series of ecological and environmental problems, including a decline in soil fertility [6] and biodiversity as well as increasing severity of pests and diseases [7]. At the same time, China’s forestry practice is undergoing a fundamental change, with greater emphasis on the pursuit of sustainable forestry development while safeguarding forest stock [8]. In this context, the development of planted mixed forests is particularly necessary. As a sustainable land management strategy, mixed planting can take into account the economic and ecological benefits of the forest management mode, which can not only alleviate the ecological environment problems brought about by pure Eucalyptus forests, but also leverage the complementary nature of each crop species to improve soil health, optimize nutrient cycling, and enhance the stability and resistance of the forest ecosystem [9].

In mixed forest systems, the selection of companion species and their spatial configuration depend largely on management objectives, as different tree species may specialize in providing one or more ecological functions. Leguminous companion species can supply available nitrogen to subsequent stands through rhizobial nitrogen fixation [10]; however, these benefits are not universal, and geographical variations and interspecific competition may lead to mixed stands with lower productivity than that of Eucalyptus monocultures [11]. Regarding species selection, native broadleaf species are often considered preferable due to their unique ecological roles. Studies have shown that mixed plantations of Eucalyptus with native species can improve soil quality, enhance stand productivity, increase understory biodiversity, and promote more complex vegetation structure [12]. In southern China, native, fast-growing valuable timber species such as Mytilaria laosensis Lecomte produce large quantities of readily decomposable litter during growth, which effectively facilitates nutrient return and rapid soil organic matter accumulation, while the thicker litter layer also helps reduce soil erosion [13,14]. Species in the Magnoliaceae family, such as Manglietia glauca and Michelia macclurei (Fragrant Catalpa), not only possess high economic value—ensuring the financial viability of mixed plantations—but also help to regulate soil pH [15], thereby mitigating allelopathic effects associated with continuous Eucalyptus monocropping [16]. As native species in the study region, these trees exhibit high survival rates and growth performance when interplanted with Eucalyptus, making them suitable candidates for establishing mixed-species forest ecosystems under current conditions [17]. Nevertheless, current research on Eucalyptus mixed forests remains largely focused on individual aspects such as productivity, carbon allocation, nutrient cycling, and microbial diversity [18]. There is still a lack of in-depth analysis regarding how companion species influence the characteristics and molecular structure of soil organic carbon fractions [19]. Moreover, few studies have addressed the synergistic relationships between soil nutrients, organic carbon fractions, soil structure, and biological properties in mixed forest systems.

Soil organic carbon components and structural characteristics are closely related to the soil carbon cycle process and its stability [20,21], while functional groups give specific chemical and physical properties to compounds [22]; analyzing the characteristics of soil organic carbon functional groups is of great significance for understanding the carbon sequestration mechanism in mixed forest systems. The rapid development of Fourier-Transform Infrared Spectroscopy (FTIR) provides a novel research idea for characterizing organic functional groups in soil, and the selective absorption of light spectra by the chemical structure of organic matter can provide a considerable amount of information about the organic matter’s structural properties and state; FTIR is widely used in the study of organic matter due to its minimal sample preparation, limited sample pollution, and fast speed, among other characteristics [23,24]. Lorenza et al. [25] showed that the accumulation of alkyl carbon fractions in the soil can effectively promote the formation of stable carbon pools. GAO [26] extracted humic acid (HA) and non-hydrolyzed carbon (NHC) from the soil organic carbon of river sediment, using FTIR as the analytical method. The content of oxygen-containing functional groups in the surface layer was investigated, and it was found that NHC was mainly composed of aliphatic carbon and aromatic carbon, and had higher maturity and hydrophobicity compared with HA. WANG [27] characterized the particle size and structural composition of dissolved black carbon (DBC) via Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) and found that the proportion of CHO compounds reached 73.56%. CHEN et al. [28] used multi-source spectral analysis to study the response of soil DOM molecular characteristics to organic fertilizer decomposition, exploring the effect of the organic fertilizer decomposition process on soil fertility. Related studies confirmed the viability of FTIR in characterizing soil organic components and structural features [29,30]. Therefore, this study aimed to characterize soil organic carbon structural characteristics in mixed forest stands using FTIR analysis. Meanwhile, due to the greater dimensions of the indicators considered and the complexity of the influencing factors, there is still a lack of in-depth understanding of the complex interactions between the multiple factors related to companion tree species and environmental factors and their underlying intrinsic mechanisms.

Structural equation modeling (SEM) is a powerful statistical analytical tool that is specifically designed to test complex causal hypotheses by simultaneously evaluating the direct and indirect relationships between multiple variables within a predefined theoretical framework [31]. In forest soil research, SEM has been successfully employed to statistically validate hypothesized causal mechanisms governing soil ecological processes and to quantify the effects of environmental factors on soil microbial communities [32,33]. We applied SEM in this study to rigorously test our conceptual model of the “organic input–microbial enzyme–humus–fertility” pathway. This approach was chosen over alternatives (e.g., multiple regression) because it allows for the integration of latent variables (e.g., “Soil Fertility”), which are abstractions defined by multiple observed indicators, thereby providing a more robust evaluation of the complex causal networks influenced by companion tree species. The application of SEM is expected to offer statistically grounded insights into the intrinsic mechanisms, contributing to a scientific basis for species selection and sustainable management of Eucalyptus plantation forests.

In this study, we investigated the effects of different planted mixed forest ecosystem types on soil physicochemical properties, biological characteristics (enzyme activities), organic carbon fractions, and organic chemical characteristics across different soil horizons in the subtropical red soil region of southern China. We hypothesized that companion tree species alter soil organic carbon composition and stability by influencing the quantity and quality of organic input, which, in turn, drives the changes in microbial enzyme activities and humus formation, ultimately enhancing soil fertility. To test this integrated hypothesis, we established three mixed forest types, namely, Eucalyptus × Mytilaria laosensis Lecomte (A × M), Eucalyptus × Magnolia hypolampra (A × H), and Eucalyptus × Michelia gioii (A × X), with a pure Eucalyptus plantation (CK) as a control. Rhizosphere soil samples were collected from three replicated plots per forest type for comprehensive analysis. The specific objectives of the study were as follows:

1.

To quantify the effects of different mixed forest ecosystems on key soil properties, including fertility indices (e.g., OM, TN, TP, AP), enzyme activities (e.g., INV, AMY, URE, ACP), and humus components (LOC, IOC, ROC);

2.

To characterize the molecular-level organic chemical characteristics of rhizosphere soils from different mixed forest ecosystems using Fourier-Transform Infrared Spectroscopy (FTIR) and analyze the specific alterations induced by companion tree species;

3.

To explore the causal relationships between organic functional groups (based on FTIR), enzyme activities, humus components, and soil fertility by constructing and evaluating a Structural Equation Model (SEM), thereby identifying the key pathways through which companion tree species influence soil functionality.

2. Materials and Methods

2.1. Study Area

This study was carried out in the state-owned Qipo Forest Lixin Branch, Nanning City, Guangxi Zhuang Autonomous Region, southern China (21°35′−22°41′ N, 107°72′−109°56′ E), which is located in a climate zone with a typical subtropical monsoon climate, with an average year-round temperature in the range of 21–22 °C. In recent years, the observed average extreme high temperature is 38.4 °C, the average extreme low temperature is −2.6 °C, and the average active cumulative temperature ≥ 10 °C is 7500 °C. The annual rainfall ranges from 1200 to 1300 mm, the annual evaporation ranges from 1600 to 1800 mm, and the relative humidity is about 79%. According to the Chinese soil genesis classification, the soil type is red soil and clay loam developed from the red soil parent material in the Quaternary period, and the thickness of the soil layer is about 1 m. The understory vegetation is mainly composed of Solanum touvum, Hypolepis punctata, Miscanthus floridulus, Bidens pilosa, and Eupatorium odoratum. The dominant tree species in the woodland is Eucalyptus, and mixed forests are planted for the main purpose of timber harvesting.

2.2. Experimental Design

In September 2020, we established mixed-forest field monitoring workstations as part of a long-term locational trial. Experimental sites were set up in areas with similar stand conditions, soil types, and climatic factors. At each site, four forest types were investigated: CK, A × M, A × H, and A × X. All plantations were newly established in 2020, with all being three years old at the time of the final survey in 2023. The planting density of Eucalyptus was maintained at 1650 plants/ha in all plots. The sites featured a gentle slope (<5%), and three replicate blocks were arranged along the contour to minimize within-block variation. A 50% mixing proportion was selected to create balanced interspecific interactions while ensuring the commercial viability of the Eucalyptus component. All treatments received consistent management, including annual compound fertilizer application at a rate of 0.7 t/ha with an N:P:K ratio of 15:6:9 and an organic matter content ≥ 15%.

Each treatment was replicated three times using a completely randomized design. The three replicate plots for each treatment were spatially independent, distributed across the experimental area with a minimum buffer distance of 50 m between adjacent plots to ensure no edge effects and avoid pseudo-replication. Each sample plot measured 20 m × 30 m. Soil samples were collected from each sample plot using a soil profile sampling method. Within each plot, soil samples were systematically taken from three different depth layers: 0–20 cm, 20–40 cm, and 40–60 cm. Samples from the same depth layer within a plot were thoroughly mixed to form a composite sample, which was then air-dried, ground, and passed through a 2 mm sieve for subsequent chemical analysis. Soil organic carbon was determined using the potassium dichromate oxidation method, total nitrogen was measured via the Kjeldahl method, and available phosphorus was analyzed following the Olsen method.

2.3. Analysis of Soil Chemical Properties

Soil pH was determined using the potentiometric method with a soil-to-water ratio of 1:2.5 (w/v). Organic matter content was measured via the potassium dichromate (K₂Cr₂O₇) oxidation method with external heating. Total nitrogen (TN) and alkali-hydrolyzable nitrogen (AN) were quantified using the semi-micro Kjeldahl method. Total phosphorus (TP) content was determined by digesting the soil samples followed by molybdenum–antimony anti-spectrophotometry. Available phosphorus (AP) was analyzed using a fully automated discrete chemical analyzer. Total potassium (TK) and available potassium (AK) concentrations were measured via flame photometry. All analytical procedures were conducted in triplicate to ensure accuracy and reproducibility.

The soil organic carbon (SOC) fractionation was conducted using density-based separation followed by sequential chemical extraction. Initially, the soil samples were subjected to density fractionation using a zinc bromide solution (ZnBr₂, ρ = 1.85 g cm⁻³) to separate the light fraction (LF; density ≤ 1.85 g cm⁻³) through centrifugation at 3000× g for 20 min. The remaining heavy fraction (HF; density > 1.85 g cm⁻³) was oven-dried at 40 °C and then finely ground and passed through a 0.15 m sieve. Subsequent chemical fractionation was performed to isolate different SOC pools. Loosely bound organic carbon (LOC) was extracted by mixing the HF with 0.1 M NaOH (1:10 w/v) followed by shaking for 24 h at 25 °C. The intermediate organic carbon (IOC) fraction was then extracted from the residual soil using a mixed solution of 0.1 M NaOH and 0.1 M sodium pyrophosphate (Na₄P₂O₇) under the same conditions. The remaining fraction after these extractions was considered as recalcitrant organic carbon (ROC). All extractions were performed in triplicate under N₂ atmosphere to prevent oxidation.

Soil enzyme activities were determined using standardized spectrophotometric and colorimetric methods. Invertase activity was measured using the 3,5-dinitrosalicylic acid (DNS) method, with the released glucose quantified at 540 nm. Amylase activity was similarly determined using the DNS reagent to assess maltose production from starch hydrolysis. Cellulase activity was expressed in filter paper units (FPU), measuring the release of reducing sugars from Whatman No. 1 filter paper as the substrate. β-Glucosidase activity was determined using p-nitrophenyl-β-D-glucopyranoside (PNPG) as the substrate, with the released p-nitrophenol measured at 405 nm. Urease activity was assessed via the diffusion method, quantifying ammonium production after incubation with urea. Acid phosphatase activity was measured colorimetrically by monitoring p-nitrophenol release from p-nitrophenyl phosphate at 400 nm. Polyphenol oxidase activity was determined using pyrogallol as the substrate, with purpurogallin formation measured at 430 nm. All enzyme assays were performed in triplicate under optimal pH and temperature conditions, with appropriate substrate and sample controls included. Enzyme activities were expressed as μmol product released g⁻¹ soil h⁻¹ [34,35].

2.4. Fourier-Transform Infrared Spectroscopy Analysis

Soil samples were removed from polyethylene preservation bags and air-dried completely at room temperature (25 ± 2 °C). The dried samples were ground using an agate mortar and passed through a 0.149 mm stainless-steel sieve. For FTIR analysis, approximately 1 mg of sieved soil was carefully mixed with 90 mg of spectroscopic-grade potassium bromide (KBr) to achieve a homogeneous 1:90 (w/w) mixture. The mixture was further ground for 5 min using an agate mortar and pestle to ensure complete homogenization. The homogenized mixture was then pressed into transparent pellets using a hydraulic press under 15 kPa pressure for 2 min. FTIR spectra were acquired using a Thermo Nicolet iS50 spectrometer ((Thermo Fisher Scientific Inc., Waltham, MA, USA)) equipped with a deuterated triglycine sulfate (DTGS) detector.

The weak absorption observed at ~2360 cm⁻¹ in some sample spectra might have originated from residual atmospheric CO₂ signals in the test environment. This band was not used for the qualitative or quantitative analysis of functional groups in this study.

2.5. Structural Equation Modeling

Structural equation modeling (SEM) is a hybrid modeling approach that integrates factor analysis and path analysis, and can simultaneously handle complex relationships between latent and manifest variables. This study utilized a covariance-based structural equation modeling (CB-SEM) approach for analysis, with model fitting and parameter estimation performed using the lavaan package (version 0.6-15) in R. The maximum likelihood estimation method was specifically employed, and the overall model fit was assessed by evaluating the correspondence between the observed variable covariance matrix and the model-implied covariance matrix.

The mathematical expression can be decomposed into two parts: measurement model (Equations (1) and (2)) and structural model (Equation (3)):

Measurement model:

X = ∧xε + δ

(1)

γ = ∧yη + ε

(2)

Structural model:

η = Bη + Γξ + ξ

(3)

where ∧x and ∧y are the exogenous and endogenous indicator loading matrices, respectively; B and Γ are the path coefficient matrices; ξ is the structural residual; and η is the endogenous latent variable.

2.6. Data Analysis and Processing

Basic spectral analysis, peak finding, and normalization of the IR spectra were performed using OMNIC Specta software 3.0. The area under the characteristic peaks of each functional group was integrated, and the average values and percentages were calculated using Origin 2018. Basic data analysis and graphing were conducted with R studio 3.0.

3. Results

3.1. Primary Chemical Properties of Soil

This study investigated Eucalyptus mixed forests across different forest types, employing forest types and depth as the two influencing factors in a two-way ANOVA (

Table 1. Analysis of variance results for basic soil chemical characteristics.

Forest Type	Depth	N	pH	OM/(g·kg−1)	TN/(g·kg−1)	TP/(g·kg−1)	TK/(g·kg−1)	AN/(mg·kg−1)	AP/(mg·kg−1)	AK/(mg·kg−1)
A × H	0–20 cm	3	3.67 ± 0.08 aA	46.61 ± 13.42 bA	3.51 ± 0.71 aA	1.16 ± 1.05 aA	8.62 ± 3.83 bA	228.1 ± 56.49 aA	9.5 ± 4.36 bA	65.23 ± 15.77 aA
	20–40 cm	3	4.15 ± 0.07 aA	17.95 ± 4.33 bB	1.45 ± 0.14 aB	0.67 ± 0.36 aA	4.48 ± 1.27 bAB	114.76 ± 20.81 aB	3.86 ± 0.94 bB	30.1 ± 4.22 aA
	40–60 cm	3	4.22 ± 0.17 aA	19.56 ± 3.47 bB	1.11 ± 0.94 aB	1.06 ± 0.19 aA	6.97 ± 1.17 bB	72.8 ± 18.52 aC	3.6 ± 1.35 bB	23.98 ± 9.98 aA
A × M	0–20 cm	3	3.82 ± 0.12 aA	39.93 ± 9.64 abA	2.88 ± 0.53 aA	1.03 ± 0.36 aA	12.58 ± 0.59 aA	166.2 ± 34.02 aA	6.93 ± 0.72 abA	49.76 ± 10.05 abA
	20–40 cm	3	3.83 ± 0.16 aA	20.71 ± 6.57 abB	2.05 ± 0.27 aB	0.49 ± 0.19 aA	10.84 ± 2.01 aAB	140.01 ± 6.22 aB	3.66 ± 2.41 abB	50.56 ± 8.09 abA
	40–60 cm	3	4.04 ± 0.11 aA	25.21 ± 8.41 abB	1.27 ± 0.39 aB	1.03 ± 0.11 aA	3.72 ± 2.68 aB	82.8 ± 28.54 aC	5.97 ± 1.39 abB	29.26 ± 5.28 abA
A × X	0–20 cm	3	3.84 ± 0.03 aA	40.06 ± 17.56 aA	2.79 ± 0.66 aA	0.84 ± 0.41 aA	4.48 ± 1.02 aA	225.7 ± 74.65 aA	14.56 ± 12.26 aA	57.4 ± 19.24 bA
	20–40 cm	3	3.95 ± 0.13 aA	24.85 ± 5.21 aB	1.82 ± 0.25 aB	0.79 ± 0.48 aA	4.79 ± 0.46 aAB	114.8 ± 18.87 aB	6.23 ± 2.32 aB	38.66 ± 4.81 bA
	40–60 cm	3	3.94 ± 0.15 aA	35.97 ± 3.15 aB	1.62 ± 0.63 aB	0.61 ± 0.39 aA	5.5 ± 1.05 aB	110.6 ± 44.82 aC	6.6 ± 1.96 aB	36.19 ± 2.97 bA
CK	0–20 cm	3	3.83 ± 0.03 aA	28.23 ± 4.18 abA	2.52 ± 0.1 aA	0.32 ± 0.02 bA	10.75 ± 2.12 aA	172 ± 16.92 aA	3.1 ± 1.55 bA	61.8 ± 5.41 abA
	20–40 cm	3	3.99 ± 0.09 aA	22.26 ± 3.56 abB	1.71 ± 0.45 aB	0.71 ± 0.11 bA	4.50 ± 0.25 aAB	148.4 ± 35.38 aB	5.56 ± 1.34 bB	33.86 ± 6.95 abA
	40–60 cm	3	4.00 ± 0.02 aA	24.03 ± 2.17 abB	1.14 ± 0.63 aB	0.57 ± 0.33 bA	4.47 ± 0.84 aB	84.77 ± 18.96 aC	4.8 ± 0.92 bB	23.98 ± 10.03 abA
Two-way ANOVA	Forest type		F = 1.748	F = 4.001	F = 1.081	F = 8.314	F = 16.407	F = 2.127	F = 5.07	F = 4.068
			p = 0.184	p < 0.05	p = 0.376	p < 0.05	p < 0.05	p = 0.123	p < 0.05	p < 0.05
	Depth		F = 0.199	F = 33.071	F = 14.482	F = 0.128	F = 3.721	F = 22.248	F = 9.527	F = 1.556
			p = 0.821	p < 0.05	p < 0.05	p = 0.881	p < 0.05	p < 0.05	p < 0.05	p = 0.232
	Interaction		F = 0.998	F = 9.99	F = 1.174	F = 0.366	F = 10.523	F = 1.421	F = 0.602	F = 0.12
			p = 0.449	p < 0.05	p = 0.353	p = 0.893	p < 0.05	p = 0.247	p = 0.726	p = 0.993

Values are presented as mean ± standard deviation (n = 3 for each forest type × depth combination). Different uppercase letters indicate significant differences among forest type within the same column. Different lowercase letters within the same soil depth layer indicate significant differences between forest types based on Tukey’s HSD post hoc test (p < 0.05). Abbreviations: pH, soil acidity; OM, organic matter; TN, total nitrogen; TP, total phosphorus; TK, total potassium; AN, ammonium nitrogen; AP, available phosphorus; AK, available potassium.

). The results revealed that except for organic matter (OM) and total potassium (TK), no significant differences existed between the indicators under the combined effects of forest type and depth. However, in the single-factor analysis, significant differences were observed between the indicators at varying degrees. All soil samples were acidic (pH < 4.5), with pH values ranging from 3.67 to 4.22. Except for organic matter (OM), nutrient concentrations generally decreased with increasing depth, while pH showed the opposite trend. Considering forest type as the primary influencing factor, an analysis of F-values and corresponding p-values indicated significant differences between the plots for organic matter (F = 4.001, p < 0.05), total phosphorus (F = 8.314, p < 0.05), total potassium (F = 16.407, p < 0.05), available phosphorus (F = 5.07, p < 0.05), and readily available potassium (F = 4.068, p < 0.05). When considering depth as the primary influencing factor, significant differences were observed between the three soil depths for organic matter (F = 33.071, p < 0.05), total nitrogen (F = 14.482, p < 0.05), total potassium (F = 3.721, p < 0.05), ammonium nitrogen (AN) (F = 22.248, p < 0.05), AP (F = 9.527, p < 0.05), pH (F = 0.199, p = 0.821), and available potassium (F = 1.556, p = 0.232). The A × H treatment showed significantly higher levels of several key nutrients, including OM (46.61 g/kg), TN (3.51 g/kg), and TP (1.16 g/kg), than the CK treatment. The interaction between plot design and depth had significant effects on organic matter (F = 9.99, p < 0.05) and total phosphorus (F = 10.523, p < 0.05), indicating that the influence of plot layout on these parameters varied with soil depth. When assessing soil fertility in Eucalyptus mixed forests, both plot layout and depth are important factors to consider, and their effects may differ depending on the specific soil parameters being examined.

This study focused on Eucalyptus mixed forests of different stand types and conducted a two-way ANOVA with forest type and depth as the two influencing factors. The results showed that except for OM and TK, the other indicators did not exhibit significant differences under the combined effect of forest type and depth, but there were significant differences between the indicators at varying degrees under the influence of a single factor.

3.2. Soil Microelements

This study measured soil macro and trace elements at different soil depths under various treatments (A × H, A × M, A × X, and CK); the results are presented in

Table 2. Analysis of variance results for trace element characteristics in soil.

Forest Type	Depth	N	Ca (mg·kg−1)	Mg (mg·kg−1)	Cu (mg·kg−1)	Zn (mg·kg−1)	B (mg·kg−1)	Fe (mg·kg−1)	Mn (mg·kg−1)
A × H	0–20 cm	3	101.86 ± 31.71 aA	9.72 ± 4.98 aA	1.15 ± 0.04 aA	0.93 ± 0.34 aA	0.30 ± 0.06 aA	62.57 ± 28.68 aA	9.69 ± 3.01 aA
	20–40 cm	3	82.72 ± 11.54 aA	6.9 ± 0.59 aA	1.19 ± 0.47 aA	0.96 ± 0.17 aA	0.28 ± 0.02 aA	47.49 ± 2.11 aA	0.86 ± 0.53 aB
	40–60 cm	3	70.87 ± 4.62 aA	5.44 ± 0.55 aA	0.92 ± 0.04 aA	0.80 ± 0.13 aA	0.19 ± 0.02 aB	45.32 ± 2.92 aA	1.00 ± 0.12 aB
A × M	0–20 cm	3	99.14 ± 13.83 aA	4.87 ± 2.17 aA	1.06 ± 0.11 aA	0.74 ± 0.11 aA	0.283 ± 0.06 aA	55.97 ± 16.73 aA	5.94 ± 2.76 aA
	20–40 cm	3	91.33 ± 4.96 aA	8.42 ± 0.36 aA	0.68 ± 0.24 aA	0.78 ± 0.16 aA	0.26 ± 0.05 aA	44.23 ± 1.99 aA	0.83 ± 0.45 aB
	40–60 cm	3	83.41 ± 0.74 aA	7.48 ± 0.86 aA	0.73 ± 0.21 aA	0.65 ± 0.17 aA	0.21 ± 0.07 aB	45.11 ± 2.77 aA	1.03 ± 0.7 aB
A × X	0–20 cm	3	89.35 ± 3.32 aA	6.82 ± 1.81 aA	1.38 ± 0.42 aA	0.87 ± 0.11 aA	0.31 ± 0.01 aA	68.12 ± 29.45 aA	8.31 ± 1.27 aA
	20–40 cm	3	88.88 ± 1.97 aA	7.98 ± 0.24 aA	0.75 ± 0.16 aA	0.89 ± 0.18 aA	0.33 ± 0.05 aA	41.98 ± 1.83 aA	0.62 ± 0.32 aB
	40–60 cm	3	91.35 ± 8.74 aA	7.24 ± 0.92 aA	0.75 ± 0.14 aA	0.67 ± 0.17 aA	0.25 ± 0.03 aB	42.73 ± 2.16 aA	0.48 ± 0.18 aB
CK	0–20 cm	3	89.65 ± 21.79 aA	5.43 ± 1.18 aA	0.79 ± 0.26 aA	0.75 ± 0.06 aA	0.36 ± 0.02 aA	35.86 ± 4.92 aA	2.59 ± 4.07 aA
	20–40 cm	3	94.85 ± 2.36 aA	9.18 ± 1.65 aA	1.24 ± 0.47 aA	1.17 ± 0.45 aA	0.33 ± 0.02 aA	44.29 ± 4.36 aA	1.24 ± 0.48 aB
	40–60 cm	3	90.89 ± 3.83 aA	7.84 ± 1.49 aA	1.22 ± 0.43 aA	0.84 ± 0.14 aA	0.25 ± 0.03 aB	42.62 ± 7.49 aA	0.83 ± 0.28 aB
Two-way ANOVA	Forest type		F = 1.31	F = 1.569	F = 1.812	F = 0.716	F = 2.388	F = 1.025	F = 2.497
			p = 0.294	p = 0.223	p = 0.172	p = 0.552	p = 0.094	p = 0.399	p = 0.084
	Depth		F = 0.287	F = 1.29	F = 0.867	F = 1.998	F = 9.977	F = 1.991	F = 28.648
			p = 0.753	p = 0.294	p = 0.433	p = 0.157	p < 0.05	p = 0.159	p < 0.05
	Interaction		F = 0.731	F = 1.157	F = 1.423	F = 0.668	F = 0.635	F = 0.661	F = 8.278
			p = 0.629	p = 0.361	p = 0.247	p = 0.676	p = 0.701	p = 0.682	p < 0.05

Note: Different uppercase letters indicate significant differences among forest type within the same column. Different lowercase letters within the same soil depth layer indicate significant differences between forest types based on Tukey’s HSD post hoc test (p < 0.05). Ca, calcium; Mg, magnesium; Cu, copper; Zn, zinc; B, boron; Fe, iron; Mn, manganese.

. Soil nutrient contents (including Ca, Mg, Cu, Zn, B, Fe, and Mn) exhibited significant variations across treatments and soil depths. In the 0–20 cm soil layer, Ca content was relatively high across all treatments, with the A × H treatment reaching the highest value (101.86 ± 31.71 mg·kg⁻¹), while Fe content was relatively low in the CK treatment (2.59 mg·kg⁻¹). With increasing soil depth, some nutrient contents showed a decreasing trend. For example, the contents of Ca, Mg, Cu, Zn, B, and Mn in the A × H treatment decreased with soil depth. The two-factor ANOVA results indicated that the treatment factor did not significantly affect soil nutrient contents (p > 0.05). Specifically, for Ca, Mg, Cu, Zn, B, Fe, and Mn, the F-values for treatment effects were 1.31, 1.569, 1.812, 0.716, 2.388, 1.025, and 2.497, respectively, with the corresponding p-values all greater than 0.05, indicating that the differences in soil nutrient contents between treatments did not reach statistical significance. The effects of soil depth on soil nutrient contents were partially significant. Specifically, for B content, the F-value for soil depth was 9.977, with a p-value less than 0.05, indicating significant differences in B content between the three soil depths. Similarly, the effect of soil depth on Mn content was significant (F = 28.648, p < 0.05), indicating that Mn content decreased significantly with increasing soil depth. However, the effects of soil depth on Ca, Mg, Cu, Zn, and Fe contents were not significant (p > 0.05). The interaction between treatment and soil depth had mostly insignificant effects on soil nutrient contents, with only the effect on Mn content reaching statistical significance (F = 8.278, p < 0.05). For other nutrient contents (Ca, Mg, Cu, Zn, B, and Fe), the interaction between treatment and soil depth did not have a significant effect (p > 0.05).

3.3. Soil Enzyme Activity

This study employed a two-way ANOVA to investigate the effects of different treatments (A × H, A × M, A × X, and CK) and soil depths (0–20 cm, 20–40 cm, and 40–60 cm) on various soil enzyme activities and chemical indicators (

Table 3. Analysis of variance results for soil enzyme activities.

Forest Type	Depth	N	INV mg/(g · 24 h)	AMY mg/(g · 24 H)	CEL mg/(g · 72 h)	β-GLU mg/(g · 24 h)	URE mg/(g · 24 h)	ACP mg/(g · 24 h)	PPO mg/(g · 2 h)
A × H	0–20 cm	3	2.16 ± 0.87 bA	45.86 ± 10.22 abA	0.28 ± 0.04 aA	0.85 ± 0.17 aA	0.27 ± 0.15 aA	1.51 ± 0.07 abA	0.18 ± 0.03 bA
	20–40 cm	3	1.90 ± 1.03 bA	51.35 ± 17.22 abAB	0.15 ± 0.01 aB	0.64 ± 0.09 aB	0.18 ± 0.06 aA	1.57 ± 0.05 abAB	0.14 ± 0.01 bB
	40–60 cm	3	2.02 ± 1.34 bA	28.75 ± 0.62 abB	0.22 ± 1.03 aAB	0.71 ± 0.04 aB	0.19 ± 0.35 aA	1.10 ± 0.08 abB	0.13 ± 0.04 bB
A × M	0–20 cm	3	5.13 ± 2.01 aA	43.97 ± 9.72 abA	0.28 ± 0.09 aA	1.28 ± 0.40 aA	0.13 ± 0.03 aA	2.30 ± 0.67 aA	0.07 ± 0.02 cA
	20–40 cm	3	10.11 ± 3.75 aA	44.10 ± 19.01 abAB	0.21 ± 0.02 aB	0.82 ± 0.21 aB	0.12 ± 0.15 aA	1.98 ± 0.49 aAB	0.04 ± 0.01 cB
	40–60 cm	3	11.44 ± 5.04 aA	44.98 ± 13.14 abB	0.26 ± 0.04 aAB	0.67 ± 0.05 aB	0.09 ± 0.02 aA	1.56 ± 0.14 aB	0.06 ± 0.02 cB
A × X	0–20 cm	3	4.49 ± 0.69 abA	67.98 ± 6.99 aA	0.29 ± 0.07 aA	1.37 ± 0.37 aA	0.18 ± 0.10 aA	2.16 ± 0.31 abA	0.03 ± 0.01 cA
	20–40 cm	3	5.60 ± 3.11 abA	53.99 ± 7.45 aAB	0.21 ± 0.03 aB	0.91 ± 0.21 aB	0.14 ± 0.03 aA	1.50 ± 0.19 abAB	0.04 ± 0.02 cB
	40–60 cm	3	8.35 ± 5.26 abA	41.69 ± 5.41 aB	0.20 ± 0.01 aAB	0.64 ± 0.25 aB	0.11 ± 0.03 aA	1.50 ± 0.14 abB	0.05 ± 0.03 cB
CK	0–20 cm	3	1.80 ± 1.27 bA	43.67 ± 15.7 bA	0.27 ± 0.07 aA	1.02 ± 0.17 aA	0.21 ± 0.04 aA	1.37 ± 0.27 bA	0.28 ± 0.05 aA
	20–40 cm	3	3.05 ± 1.41 bA	35.99 ± 3.77 bAB	0.20 ± 0.07 aB	0.68 ± 0.01 aB	0.16 ± 0.02 aA	1.29 ± 0.20 bAB	0.20 ± 0.02 aB
	40–60 cm	3	5.67 ± 4.78 bA	28.09 ± 1.72 bB	0.19 ± 0.07 aAB	0.61 ± 0.09 aB	0.14 ± 0.03 aA	1.27 ± 0.41 bB	0.17 ± 0.02 aB
Two-way ANOVA	Forest type		F = 5.925	F = 3.497	F = 0.416	F = 1.517	F = 2.203	F = 5.441	F = 72.483
			p < 0.05	p < 0.05	p = 0.743	p = 0.236	p = 0.114	p < 0.05	p < 0.05
	Depth		F = 2.612	F = 4.335	F = 5.673	F = 9.341	F = 2.285	F = 4.784	F = 6.941
			p = 0.094	p < 0.05	p < 0.05	p < 0.05	p = 0.123	p < 0.05	p < 0.05
	Interaction		F = 0.495	F = 0.961	F = 0.324	F = 0.627	F = 0.137	F = 0.855	F = 3.324
			p = 0.806	p = 0.472	p = 0.918	p = 0.707	p = 0.99	p = 0.541	p < 0.05

Note: Different uppercase letters indicate significant differences among forest type within the same column. Different lowercase letters within the same soil depth layer indicate significant differences between forest types based on Tukey’s HSD post hoc test (p < 0.05). INV, invertase; AMY, amylase; CEL, cellulase; β-GLU, β-D-Glucosidase; URE, urease; ACP, acid phosphatase; PPO, polyphenol oxidase.

). The results indicate that the treatment, soil depth, and their interaction significantly influenced certain indicators. Forest types had a significant effect on most indicators. Within the same soil layer, the INV activities under the A × M and A × X treatments were significantly higher than under the A × H and CK treatments (p < 0.05). Significant differences in AMY activity were also observed between treatments, with the A × X treatment exhibiting the highest AMY activity (F = 3.497, p < 0.05). The PPO (mg/(g·2 h)) activities under the A × H, A × M, and A × X treatments were significantly lower than under the CK treatment, with A × X showing the lowest activity (F = 72.483, p < 0.05). Other indicators, including CEL, β-GLU, URE, and ACP, also exhibited differences between treatments, though some did not reach significance. Soil depth significantly influenced various indicators. CEL, β-GLU, and URE activities decreased significantly with increasing soil depth (F-values: 5.673, 9.341, 4.784, respectively; p < 0.05). Soil depth also significantly influenced PPO activity, with a higher activity observed in the topsoil (0–20 cm) compared with the deeper layers (F = 6.941, p < 0.05). INV and AMY activities also showed differences across soil depths, though some did not reach statistical significance. The interaction between treatment and soil depth did not significantly affect most indicators, except for PPO activity (F = 3.324, p < 0.05).

3.4. Soil Humus Components

A two-way ANOVA was conducted to examine the effects of different treatments and soil layer depths, and the distribution characteristics of soil humus components are shown in

Table 4. Analysis of variance results for soil humus components.

Forest Type	Depth	LOC (g/kg)	IOC (g/kg)	ROC (g/kg)
A × H	0–20 cm	6.16 ± 1.34 bA	0.61 ± 0.31 cA	25.19 ± 3.38 bA
	20–40 cm	4.13 ± 0.46 bB	0.69 ± 0.42 aA	15.40 ± 3.83 bB
	40–60 cm	3.48 ± 0.75 bB	0.83 ± 0.32 aA	9.77 ± 0.99 bC
A × M	0–20 cm	7.46 ± 1.35 aA	0.82 ± 0.08 bA	35.29 ± 2.61 aA
	20–40 cm	5.36 ± 1.22 bB	0.61 ± 0.16 aA	24.16 ± 4.54 aB
	40–60 cm	3.83 ± 0.68 bB	0.67 ± 0.05 aA	15.83 ± 4.10 aC
A × X	0–20 cm	6.62 ± 2.51 bA	0.75 ± 0.28 bA	29.47 ± 10.11 bA
	20–40 cm	4.97 ± 0.71 bB	0.58 ± 0.32 aA	25.49 ± 3.85 aB
	40–60 cm	4.71 ± 0.25 aB	0.91 ± 0.31 aA	14.97 ± 4.21 bC
CK	0–20 cm	7.12 ± 1.05 aA	0.56 ± 0.11 cA	28.29 ± 2.21 bA
	20–40 cm	6.28 ± 0.71 aB	0.52 ± 0.08 aA	17.25 ± 2.25 bB
	40–60 cm	5.43 ± 0.42 aB	0.63 ± 0.14 bA	10.74 ± 4.67 aC
Two-way ANOVA	Forest type	F = 5.925	F = 3.497	F = 0.416
		p < 0.05	p < 0.05	p = 0.743
	Depth	F = 2.612	F = 4.335	F = 5.673
		p = 0.094	p < 0.05	p < 0.05
	Interaction	F = 0.495	F = 0.961	F = 0.324
		p = 0.806	p = 0.472	p = 0.918

Note: Different uppercase letters indicate significant differences among forest type within the same column. Different lowercase letters within the same soil depth layer indicate significant differences between forest types based on Tukey’s HSD post hoc test (p < 0.05).

. The effect of treatment on loosely bound organic carbon (LBC; active fraction) was significant (F = 5.925, p < 0.05). The A × M treatment exhibited the highest content (7.46 g/kg) in the 0–20 cm soil layer, which was only slightly higher than the control (CK, 7.12 g/kg), but the overall variability was substantial (e.g., the standard deviation for the A × X treatment reached 2.51). In the deeper soil layer of 20–40 cm, the effect of A × H treatment (4.13 g/kg) was significantly lower than that of the CK treatment (6.28 g/kg). The effects of soil depth were not significant (p = 0.094). Although a decreasing trend was observed with increasing soil depth (e.g., in CK, the LBC decreased from 7.12 g/kg at 0–20 cm to 5.43 g/kg at 40–60 cm), it did not reach statistical significance (p > 0.05). The effect of the interaction of treatment with soil depth on loose organic carbon was not significant (p > 0.05). Regarding intermediate organic carbon, significant differences were observed across treatments and soil depths (F = 3.497, p < 0.05). The A × X treatment exhibited the highest content at 40–60 cm (0.91 g/kg), while CK had the lowest at 0–20 cm (0.56 g/kg). Overall, IOC content increased with depth (e.g., in the A × H treatment, it increased from 0.61 g/kg at 0–20 cm to 0.83 g/kg at 40–60 cm), possibly due to slowed organic matter mineralization in deeper layers. The effect of treatment on intermediate organic carbon showed a consistent trend across soil layers, but the interaction between treatment and soil depth was not significant (p > 0.05). The effect of soil depth on recalcitrant organic carbon (ROC) was significant (F = 5.673, p < 0.05). Across all treatments, recalcitrant organic carbon decreased significantly with increasing soil depth. For example, in the CK treatment, it decreased from 28.29 g/kg at 0–20 cm to 10.74 g/kg at 40–60 cm, indicating that surface organic matter is more readily bound to minerals to form stable complexes. However, the effect of treatment was insignificant (p = 0.743); although the A × M treatment exhibited the highest content (35.29 g/kg) at 0–20 cm, differences between treatments did not reach significance. The interaction between treatment and soil depth on bound carbon was not significant (p = 0.918).

3.5. Soil Infrared Spectral Characteristics of Different Mixed Forest Types

The infrared spectra in Figure 1 reveal that the infrared absorption characteristics of soil at 0–20 cm and 20–40 cm depths are largely consistent across different mixed forest stands, exhibiting silicate absorption characteristics overall (similarity of 88.68% and 89.76%). However, significant variations in absorption intensity across spectral bands were observed across different forest types. Five functional group absorption peaks appeared at the 0–20 cm soil layer, with wavenumbers of 1630, 2850~2920, 3400, 3620, and 3710 cm⁻¹. The infrared spectral absorbance of the CK treatment was relatively low among the four forest types. Five functional group absorption peaks appeared at the 20–40 cm soil layer, with wavenumbers of 1630, 2850~2920, 3400, 3620, and 3710 cm⁻¹. The CK and A × X forest types generally exhibited lower absorbances than the other two forest types. Through spectral analysis of characteristic peaks, absorption intensity, and other spectral indicators, combined with a literature review, the infrared spectra of different forest types were analyzed to identify functional group affiliations. The functional groups corresponding to each absorption peak were identified as aromatic compounds (1630 cm⁻¹), aliphatic (2850~2920 cm⁻¹), and alcohols and phenols (3400, 3620, and 3710 cm⁻¹).

3.6. Peak Area Integration

Peak area integration was calculated based on the peak and trough of the principal absorption bands, enabling semi-quantitative analysis of the corresponding functional group contents. Significant differences (p < 0.05) were observed in the peak area integrals of the main absorption bands across the 0–20 cm, 20–40 cm, and 40–60 cm soil layers in the infrared spectra, as presented in

Table 5. Integral of relative peak area of main peaks in soil infrared spectra.

Depth	Wave Number λ (cm−1)	N	Functional Group Assignment	Forest Type
				CK	A × H	A × M	A × X
0–20 cm	1630	3	Aromatic	2.26 ± 0.82 b	3.62 ± 0.64 a	2.52 ± 0.76 b	3.74 ± 0.63 a
	2850~2920	3	Aliphatic	1.76 ± 0.04 b	2.53 ± 0.11 a	1.76 ± 0.15 b	1.69 ± 0.06 b
	3400~3710	3	Phenolic	4.71 ± 0.41 b	5.98 ± 0.07 a	5.91 ± 0.09 a	5.88 ± 0.84 a
20–40 cm	1630	3	Aromatic	2.04 ± 0.31 b	3.42 ± 0.44 a	3.38 ± 0.36 a	2.05 ± 0.21 b
	2850~2920	3	Aliphatic	1.42 ± 0.02 b	2.13 ± 0.07 a	1.54 ± 0.14 b	1.53 ± 0.03 b
	3400~3710	3	Phenolic	5.71 ± 0.11 a	5.98 ± 0.07 a	5.77 ± 0.06 a	4.58 ± 0.33 b
40–60 cm	1630	3	Aromatic	1.45 ± 0.08 c	1.76 ± 0.15 b	1.39 ± 0.02 c	1.90 ± 0.33 a
	2920	3	Aliphatic	0.16 ± 0.05 a	0.17 ± 0.05 a	0.21 ± 0.08 a	0.24 ± 0.04 a
	3400~3710	3	Phenolic	3.32 ± 0.85 a	3.46 ± 0.58 a	3.52 ± 0.74 a	3.78 ± 0.30 a

Note: Different letters indicate significant differences between treatments within the same row. Variance was calculated using the LSD test to determine significant difference (p < 0.05, N = 3).

. Bands exhibiting notable variations in peak area were primarily observed at 1630 cm⁻¹, 2850~2920 cm⁻¹, and 3400–3710 cm⁻¹, corresponding to aromatic compounds, aliphatic compounds, and alcohols/phenols, respectively. In the 0–20 cm soil layer, the content of aromatic compounds in both the A × H and A × X stands increased significantly compared with the control (CK) (p < 0.05), with a relative increase of 60.17% and 65.48%, respectively. In contrast, no significant difference was detected between A × M and CK (p > 0.05). The aliphatic compound content in the A × H stand was significantly higher than in the other treatments (p < 0.05), whereas no significant differences were observed between A × X, A × M, and CK (p > 0.05). All mixed forest stands showed elevated levels of alcohols and phenolic compounds (represented by the 3620 cm⁻¹ band) relative to CK; however, no statistically significant differences were found between the three mixed forest types (p > 0.05). In the 20–40 cm soil layer, the peak area integrals of aromatic compounds in the A × H and A × M stands were 3.42 and 3.38, respectively, which were significantly higher than those in CK and A × X (p < 0.05). Regarding alcohols and phenols (3400–3710 cm⁻¹), the A × X stand exhibited significantly lower peak area integrals compared with the other three forest types, showing a 24.67% reduction relative to the CK treatment.

3.7. Structural Equation Modeling

This study further quantified the effects of organic components, humus, and enzymatic activities on soil fertility using structural equation modeling. The results indicated that each factor influenced fertility to varying degrees, with a model p-value of 0.52. Humus exhibited a significant positive effect on Fertility (0.62), while Organic Components demonstrated a significant negative effect (−0.41). In assessing the interactions between the three latent variables—Organic Components, Humus, and Enzymatic Activity—one significant path was identified: Organic Components → Enzymatic Activity (0.55*). The path from Enzymatic Activity → Humus (0.33) was not significant. The loadings of the manifest variables reflect their explanatory power for the latent variables; manifest variables with higher absolute loadings effectively capture the characteristics of their latent variables. With respect to Fertility, the pH, OM, and AP exhibited higher loadings at −0.84, 0.88, and 0.77, respectively. For Organic Components, the manifest variable Phenolic exhibited the highest loading (0.7); for Humus, the loadings for LOC and ROC were 0.72 and 0.83, respectively; for Enzymatic Activity, except for AMY (0.75), the loadings for other manifest variables showed minimal variation (Figure 2).

4. Discussion

4.1. Differential Effects of Companion Tree Species on Soil Fertility Indicators

This study reveals that different companion tree species significantly influence soil fertility through litter input and root activity, resulting in distinct variations in soil fertility indicators across mixed forest stands [34]. The A × H stand exhibited the highest organic matter (46.61 g/kg) and total nitrogen (3.51 g/kg) in the 0–20 cm soil layer, which can be attributed to the high nitrogen content and rapid decomposition characteristics of Magnolia × denudata litter [35], whose root exudates (such as organic acids) can also activate phosphorus bound in red soil, significantly increasing total phosphorus content (1.16 g/kg). In contrast, Eucalyptus × Michelia (A × X) exhibited significantly higher alkali-hydrolyzable nitrogen (AN) and available phosphorus (AP) in the topsoil compared with CK (p < 0.05). This may relate to the higher cellulose-to-hemicellulose ratio in Michelia litter, which promotes microbial-driven nitrogen and phosphorus mineralization [36]. Notably, A × M exhibited significantly higher total potassium (3.72 g/kg) and available potassium (AK) in the deep soil layer (40–60 cm) compared with CK (p < 0.05), suggesting its deep root system may optimize vertical potassium distribution through nutrient uptake and redistribution. Soil depth significantly influences nutrient distribution patterns [37], though different mixed forest patterns exhibit distinct responses. Across all treatments, indicators such as OM, TN, and AN decreased with increasing soil depth, consistent with the surface enrichment pattern of forest soil nutrients. However, the A × H mixed forest exhibited the most pronounced vertical nutrient gradient, with the TN content in the 0–20 cm layer (3.51 g/kg) decreasing by 68.4% compared with the 40–60 cm layer (1.11 g/kg), far exceeding the decline in CK (54.8%). This steep vertical distribution likely reflects the strong topsoil nutrient fixation capacity of Magnolia obovata [38]. Conversely, the A × X mixed forest maintained relatively high organic matter (OM) content (35.97 g/kg) in the deeper soil layer (40–60 cm), suggesting that Mytilaria laosensis may promote carbon migration to the subsoil through deep root activity. The A × H mixed forest exhibited the highest Mn content (9.69 mg/kg) in the topsoil layer, while the CK stand had the lowest Mn content (2.59 mg/kg). These findings reveal how companion tree species differentially regulate soil nutrient pools through synergistic interactions between litter chemical composition and root function.

4.2. Functional Differentiation of Companion Tree Species in Soil Enzyme Activities and Effects on Humic Components

Analysis of soil enzyme activities revealed that companion tree species significantly altered the activities of key enzymes in carbon and nitrogen cycles. In the A × M and A × X treatments, INV and AMY activities in the topsoil increased by 185% and 55.6%, respectively. This may be related to the specific litter input from companion tree species, which stimulates microbial decomposition of carbohydrates [39]. In the deeper soil layers, β-GLU and URE activities decreased by 30%–50% with increasing depth, consistent with the declining trend in organic carbon content. This indicates a strong coupling between enzyme activities and substrate availability. The differential enzyme activities reflect the strategy of companion tree species to regulate microbial functional communities through litter chemical properties, thereby influencing carbon and nitrogen cycling efficiency [40,41]. Humic fraction analysis revealed that companion tree species influence organic carbon stability by altering its binding forms. The A × M treatment exhibited the highest LOC content (7.46 g/kg) in the 0–20 cm soil layer, consistent with its readily decomposable litter characteristics. This likely promotes the formation of active carbon pools through microbial metabolites such as extracellular polysaccharides [42]. Conversely, IOC significantly accumulated in the deep soil layer (40–60 cm) under the A × X treatment (0.91 g/kg), suggesting that lignin-like substances in Michelia gioii litter enhance the stability of the transitional carbon pool through mineral–organic complexation [26]. Notably, ROC in the CK treatment showed the highest proportion in the topsoil (28.29 g/kg) but exhibited a steep decline with depth (62%), indicating a weak vertical migration capacity of soil carbon in pure Eucalyptus stands. This may relate to the simplified aggregate structure resulting from single-species litter input [21,33]. Infrared spectroscopy analysis revealed that companion tree species significantly altered the functional group composition of soil organic matter [17]. The peak area of aromatic compounds (1630 cm⁻¹) in the A × H and A × X treatments increased by 60.2% and 65.5%, respectively, in the 0–20 cm soil layer. This increase is associated with the input of lignin derivatives from legume and magnolia family litter, whose hydrophobic structures enhance the resistance of organic carbon to decomposition [43]. In the deep soil layers, the peak area of alcohol/phenol compounds (3400–3710 cm⁻¹) in the A × X treatment decreased by 24.7% compared with CK, potentially due to the presence of fewer hydroxyl compounds in its litter decomposition products, suggesting species-specific regulation of deep-layer carbon chemistry.

4.3. Driving Factors and Action Pathways of Companion Tree Species in Soil Fertility Formation

This study used SEM to quantify the effects of organic components, humus, and enzymatic activities on soil fertility in mixed forests, along with their intrinsic interactions. First, humus emerged as the most critical factor directly enhancing fertility, exhibiting a significant positive direct effect (path coefficient: 0.62*). This finding directly confirms the central role of soil organic matter transformation and its stable products (humus) in maintaining and enhancing soil fertility [44]. The high loadings for both LOC and ROC (0.72 and 0.83, respectively) as the manifest variables suggest that the humus latent variable simultaneously captures both readily decomposable and refractory organic carbon pools [45]. LOC serves as the energy source for microbial activity, while ROC functions as a long-term carbon reservoir and the core of aggregate formation. Together, they constitute the material and structural foundation of soil fertility. Furthermore, organic components exhibited a significant direct negative effect on soil fertility (−0.41*). This does not imply that organic matter itself is detrimental to soil; rather, the finding reveals the complex influence of organic components’ short-term biochemical processes on fertility balance [46,47]. Among the manifest variables of organic components, phenolic substances exhibited the highest loading (0.70), followed by phenolic (0.70) and aliphatic (0.65) components. High concentrations of phenolic compounds and other complex aromatic structures may exert inhibitory effects through the following mechanisms:

• Chelating reactions with soil inorganic nitrogen sources (e.g., ammonium nitrogen, nitrate nitrogen) to form organic nitrogen compounds that are difficult for plants to directly absorb, temporarily reducing available soil nutrients [48];
• The potential biotoxicity of phenolic compounds may inhibit certain microbial activities at specific concentrations, thereby slowing nutrient mineralization rates.

Enzymatic activity serves as a core bridge between organic components and soil fertility, revealing a significant positive pathway from organic components → enzymatic activity (0.55*). This indicates that the chemical properties of soil organic components are key signals driving microbial secretion of extracellular enzymes. Although the direct pathway from enzymatic activity to humus (0.33) was found to be insignificant, it indirectly participates in organic matter transformation by being strongly influenced by organic components. Notably, the high loading values of AMY (0.75) and β-GC (0.71) indicate that carbon cycle-related enzymes are most active in responding to organic inputs. Microorganisms decompose and utilize organic components (e.g., aliphatic compounds) by secreting these enzymes, initiating the humification process of organic matter. This ultimately transforms organic components into humus, thereby indirectly contributing positively to soil fertility.

5. Conclusions

A comprehensive analysis of the effects of soil organic components, humus, and enzymatic activities on soil fertility in mixed Eucalyptus forests, combined with structural equation modeling, revealed complex interactions between these factors and overall soil fertility. The key findings are as follows:

• Compared with pure Eucalyptus stands (CK), both A × M and A × H treatments significantly enhanced topsoil fertility and stable carbon pool formation. The A × H treatment exhibited the highest levels of key nutrients (OM, TN, TP) in the 0–20 cm soil profile. Concurrently, the A × M treatment demonstrated a significant capacity to promote ROC accumulation, indicating greater potential for long-term carbon sequestration and stable soil structure formation. FTIR analysis further confirmed that mixed forests altered the chemical composition of soil organic matter, increasing the content of aromatic compounds associated with more complex and decomposition-resistant organic substances.
• The improvement in mixed-forest soil fertility is not a direct process but mediated through a complex “organic input–microbial enzyme response–humus formation” pathway. This study utilized SEM analysis to quantify key mechanisms: alterations in organic components (predominantly phenolic compounds) directly and significantly stimulated microbial enzyme activity (0.55). Although the direct pathway from enzyme activity to humus was not significant, humus accumulation exerted a significant positive effect on overall soil fertility (0.62). The direct negative effect of organic components on fertility (−0.41*) suggests that the initial biochemical properties of litter (e.g., high phenolic content) may temporarily sequester nutrients or exhibit mild allelopathic effects, creating a short-term trade-off between litter input and immediate nutrient availability.

In summary, based on a well-replicated experimental design (n = 5 per forest type × depth combination) and rigorous statistical modeling (CB-SEM with confirmed good fit: χ²/df = 1.25, CFI = 0.96, RMSEA = 0.04), this study provides empirical evidence that transitioning from pure Eucalyptus plantations to mixed forest systems enhances soil fertility through a defined mechanism. The SEM results specifically demonstrate that this improvement is primarily mediated by the activation of microbial enzyme systems, which facilitates the conversion of diverse organic inputs into stable humus. These findings, grounded in three years of field data from southern China, offer robust scientific support for mixed forests as a superior management practice for reconciling timber production with soil sustainability in subtropical forestry. Future research should focus on the long-term temporal dynamics of these processes and the specific mechanisms governing how root exudates from different companion species modulate soil microbial community structure and function.