Effects of three plantation coniferous species on plant-soil feedbacks and soil physical and chemical properties in semi-arid mountain ecosystems

Background: Large-scale afforestation can signicantly change the ground cover and soil physicochemical properties, especially the soil fertility maintenance and water conservation functions of articial forests are very important in semi-arid mountain ecosystems. However, how different tree growth affect soil nutrient and soil physicochemical properties following afforestation and which is the best plantation tree species for improving soil fertility and water conservation functions remain largely unknown. Methods: This study investigated the soil nutrient contents of three plantations with different tree species (Larix principis-rupprechtii, Picea crassifolia, Pinus tabuliformis), soils and plant-soil feedbacks, as well as the interaction between soil physicochemical properties. Results: The results revealed that the leaf and litter layer strongly inuence soil nutrient availability through biogeochemical processes: P. tabuliformis has higher organic carbon, ratio of organic carbon to total nitrogen (C:N) and organic carbon to total phosphorus (C:P) in the leaves and litter layer than L. principis-rupprechtii or P. crassifolia, suggesting that higher C:N and C:P hinder litter decomposition. As a result, the L. principis-rupprechtii and P. crassifolia plantation forests signicantly improve soil nutrients and clay components than the P. tabuliformis plantation forest. Furthermore, the L. principis-rupprechtii and P. crassifolia plantation forests signicantly improved the soil capacity, soil total porosity, and capillary porosity, decreased soil bulk density, and enhanced water storage capacity than the P. tabuliformis plantation forest. In conclusion, the results of this study showed that the strong link between plants and soil is tightly coupled to C:N and C:P, and there had a close correlation between soil particle size distribution and soil physicochemical properties. Conclusions: Therefore, our results recommend planting the L. principis-rupprechtii and P. crassifolia as the preferred tree species to enhance the soil fertility and water conservation functions, especially in semi-arid regions mountain forest ecosystems.


Introduction
The reforestation remains one of the most effective strategies for coping with climate change (Jean-Francois et al., 2019), which is also the most effective management method to solve the problems of soil erosion all over the world (Clemente et al. 2004;Kou et al. 2016). It is considered to be an effective strategy to prevent soil erosion and degradation and to promote the restoration of degraded ecosystems (Zhang et al. 2011). For the past three decades, to prevent soil erosion and deserti cation, improve water conservation capacity, the Grain to Green Program (GTGP) has been implemented by the Chinese government (Chang et al. 2012). Large-scale afforestation increased ground coverand caused changes in soil physical and chemical properties (Fu et al. 2010). Forests as an ecosystem engineer not only have species-speci c effects on soil physicochemical properties and soil communities (soil animal communities and soil microbial communities) (Prescott and Grayston, 2013;Vesterdal et al. 2008) but also regulate climate, mineral cycling and prevent soil erosion (Kozlowski 2002). Besides, arti cial forests could potentially lead to circulation and feedback effects of mineral nutrients between aboveground and below-ground ecosystems (Peichl et al. 2012;Wang et al. 2009). Therefore, the study of vegetation restoration processes and their impacts on nutrient cycling and soil properties will provide an important guide to forest management aimed at improving the ecological restoration of natural and arti cial forests, especially in semiarid mountain ecosystem regions.
It is well known that vegetation is an important factor affecting soil physical and chemical properties. Leaves of different tree species have generated species-speci c effects on litter layer decomposition and nutrient release into the soil (Norris et al., 2012;Aponte et al., 2013).
Tree species affect soil nutrient mineralization and availability through soil microorganisms, thus affecting soil fertility Aponte et al., 2013). Previous studies have shown that environmental factors in uence leaves and then affect many service functions of the ecosystems (Ayres et al. 2009;Aponte et al., 2013). Thus, leaf quality largely determines the decomposition of litter, as well as the release of nutrients and minerals into the soil (Norris et al. 2012;Aponte et al., 2013), indicating the relationship among leaves, litter, and soil (Lucas-Borja et al. 2019). However, the researches about the effects of different tree species leaf and litter on soil organic carbon, nitrogen cycling, and water conservation functions in semi-arid mountain forest ecosystems is lacking.
Soil plays an important fertility and stability function in the forest ecosystems (Lucas-Borja et al. 2019), and soil directly or indirectly regulates and in uences many biological processes in forest ecosystems . Soil properties are determined by chemical, physical and biological processes, which play a key role in determining plant growth, community composition, and individual productivity (Van der Putten et al. 2013). Besides, different plant species tend to have species-speci c effects on soil quality and quantity (Hobbie et al. 2006;Ayres et al. 2009), and they also change the physical, chemical, and biological properties of soil (Qiao et al. 2019). Thus, aboveground and belowground processes of forest ecosystems determine plant-soil feedbacks and in uence the composition of the plant community and nutrient cycling processes (Kardol et al. 2006;Van der Putten et al. 2013), potentially affecting ecosystem functioning, such as interactions between plants and other communities (Van der Putten et al. 2013), conserving water resource and preventing soil losses. Therefore, understanding the relationships between plantation types and soil physicochemical properties is of great signi cance for the soil and water conservation, nutrient cycling, and soil health assessment of forest stands.
Soil particle-size distribution (PSD) refers to the percentage of each particle size class in the soil, which can re ect the in uence of soil water movement, solute transport, and nutrient status, and vegetation types on soil texture (Sun et al., 2016). Soil texture is divided into clay, silt, and sand, which is one of the important physical parameters of soil (Mohammadi & Meskini-Vishikaee 2013;Hu et al., 2011;Xu et al., 2013). The change of soil particle-size distribution is the result of the combined effects of soil evolution, vegetation restoration, and environmental factors. Soil texture and organic matter are the key factors affecting soil particle size (Qi et al., 2018). Previous studies have shown that the aboveground part of plants can effectively increase the roughness of the surface, thus increasing the content of ne particles and nutrients in the soil, leading to the change of soil structure (Xiang et al., 2015). However, the relationship between soil physicochemical properties and soil particle-size distribution and their effects on water conservation functions are scarce.
Xinglong Mountain is an important water conservation area on semi-arid land in northwestern China. Since the implementation of China's Three-North Shelterbelt forest program in the 1980s, a large-scale arti cial afforestation project has been carried out on Xinglong mountain, and the planted forest species were Larix principis-rupprechtii, Picea crassifolia, Pinus tabuliformis. Although arti cial afforestation has been carried out many years, there is no systematic evaluation of the soil and water conservation capacity, ecological construction bene ts of the plantations. In this study, we hypothesize that there is a strong feedback effect of nutrients between plants and soil. Tree species may in uence the soil organic carbon (C) ,total nitrogen (TN), and total phosphorus (TP) stoichiometry of different afforestation and then will affect the soil physicochemical properties, structure, and texture. The s study aimed to: (1) investigate the in uence of three different tree species used for afforestation on the nutrient status of plants and soils and plant-soil feedbacks, (2) research the effects of three different tree species on soil physical and chemical properties, and (3) explore the impacts of soil physical and chemical properties of three different forest stands on soil particle-size distribution characteristics and its in uence factors. Therefore, the results of this study can provide theoretical guidance in the selection of forest species for afforestation and forest management, particularly in the semi-arid mountain forest ecosystem.

Study site description
This research area is located in the Gansu Xinlongshan National Nature Reserve (35°44′20.12′′N, 104°1′3.07′′E, H: 2778m) located in the Loess Plateau, China (Fig.1). As a "green rock island" on the Loess Plateau, it is an important water conservation forest and biodiversity protection area in the upper reaches of the Yellow River. The climate in this region is classi ed as semi-arid continental monsoon climate, and the annual precipitation is about 450-622 mm, and the precipitation frequency is not uniform, mostly concentrated in July to September. The effective accumulated temperature was 1800-2800℃, and the average annual relative air humidity was 68%.
We selected three study site with different dominant tree species planted 30 years ago, the planting distance is about 4 m×4 m, and all the plantations are in semi-sunny slope. The distance between every study site is less than 10 km, and the environmental, meteorological, soil and the parent material among stands were homogeneous. The more basic information of the three plantation stands was summarized in Table 1. The type of land before plantations is a natural succession of grassland and no human disturbance and management to the forests and soils since the planted. During the growth and succession of different tree species, the soil physicochemical properties will change accordingly, thus potentially affecting ecosystem functioning. Therefore, these differences among stands can be attributed to tree species. There were a large amount of herbaceous vegetation (i.e., Carex rigescens, Fragaria orientalis, Aconitum sinomontanum and Potentilla bifurca) and shrubs (i.e., Sorbus koeheana, Berberis kansuensis, Rosa sweginzowii, Cotoneaster multi orus, Spiraea alpine and Lonicera hispida) growing on the forest oor and the litter thickness was about 10 cm. .2 Analysis methods of nutrient content in the leaf and litter layers Three sample plots (25 m × 25 m) in each forest stand were randomly selected to collect leaf (i.e., needle), soil, and litter samples. And two trees or points were selected randomly in each sample plot to collect leaf, soil, and litter samples. There are six replicates for each leaf, soil, or litter samples in each forest stand. The leaf samples were randomly collected from each site for three different forest stands in August 2018, and litter samples were sampled under the canopy of each selected tree (The distance between each sampling tree or point is greater than 10 m in each sample plot. The litter of we collected is mainly fallen needles of trees). Leaf and litter samples were processed in a grinder (dried to constant weight at 75 °C) and sieved through a 60-mesh sieve. The leaf and litter layer organic carbon values were determined using the K 2 Cr 2 O 7 -H 2 SO 4 oxidation method (Bao 2000;Wang 2009), the total nitrogen (TN) values were determined using the micro-Kjeldahl method (Bao 2000;Zhang et al. 2019a), whilst leaf and litter layer total phosphorus (TP) values were determined colorimetrically (ammonium molybdate method) after wet digestion with H 2 O 2 -H 2 SO 4 , and the total potassium (TK) values were determined using an atomic absorption spectrophotometer (detection limit is 0-1000mg·L -1 ) (Aurora, AI-1200, Canada) after wet digestion with H 2 O 2 -H 2 SO 4 (Bao 2000;Zhang et al. 2019a).

Analysis methods of soil physical and chemical properties
To determine soil nutrient content, the soil samples were collected from the 0-10 cm, 10-20 cm and 20-30 cm soil layers at each site (We randomly selected two pionts in each sample plot (25 m×25 m). For each forest stand, 18 soil samples, 6 leaf samples, and 6 litter samples were collected. Totally, 48 soil samples, 18 leaf samples, and 18 litter samples were collected.). Air-dried soil samples screened by 2 mm and 0.15 mm mesh sieve were used to determining soil physiochemical properties and particle-size distribution (PSD). Soil organic carbon (SOC) content was determined using the K 2 Cr 2 O 7 -H 2 SO 4 oxidation method (Bao 2000;Wang 2009). Soil total nitrogen (TN) values were determined using the micro-Kjeldahl method , whilst soil total phosphorus (TP) and total potassium (TK) values were determined colorimetrically (ammonium molybdate method) and ame photometer after wet digestion with HClO 4 -H 2 SO 4 (Bao 2000; Cao and Chen 2017), respectively. Inorganic nitrogen in the form of nitrate nitrogen ( -N) and ammonium nitrogen ( -N) were determined through colorimetry (Bao 2000), and available phosphorus (AP) was extracted with 0.5 mol/L NaHCO 3 then determined by molybdenum-antimony colorimetry (Bao 2000;Kou et al. 2016). Available potassium (AK) was extracted with 1 mol/L CH 3 COONH 4 then determined by ame photometry (Bao 2000;Zhou et al. 2015). TN, TP, -N and -N were measured using an automatic intermittent chemical analyzer (SmartChem140, France). The AK and TK in the soil were determined using ame atomic absorption spectrophotometric method (detection limit is 0-1000mg·L -1 ) (Aurora, AI-1200, Canada).
To determine soil physical properties, undisturbed samples were obtained from the 0-10 cm, 10-20 cm and 20-30 cm soil layers using a ring knife at each typically repeated plots for three different forest stands (six intact soil cores were obtained from each of the three soil layers for each forest stand). The bulk density and soil capacity of the soil samples were measured using the method exposed by Zhang et al. (2019b). The total porosity was determined by measuring soil moisture content at saturation (total volume of water-lled soil pores) and capillary porosity (capillary porosity is the percentage of soil voids in soil volume) was determined with the method exposed by Qiu et al. (2019).
2.4 Determination of the soil particle-size distribution of the soil samples The soil particle-size distribution was measured using a laser particle analyzer (Mastersizer 2000, Malvern Company, UK), samples were pretreated with 10% H 2 O 2 solution to each 0.25 g soil sample to remove organic matter, and add 10% HCl solution to remove carbonate salts. Add deionized water and soak for 12 h, and the liquid supernatant was then removed. The samples were chemically dispersed with 0.06 mol/L sodium hexametaphosphate and were mechanically dispersed in an ultrasonic bath for 10 min (Qi et al., 2018). The measurements were repeated three times for each sample, and the soil particle-size distribution (PSD) was classi ed into the clay ( 2 μm), silt (2-50 μm), and sand (50-2000 μm) according to United States Department of Agriculture classi cation (USDA)classi cation system (Xia et al., 2020;Zhai et al., 2020).

Statistical analyses
The effect of different forest species on the physical and chemical properties of the soil, nutrient content in the vegetation and litter layers, and soil particle-size distribution (PSD) were evaluated using one-way ANOVA (the normal distribution and homogeneity of variance of the data had been checked), followed by least signi cant difference (LSD) tests for different soil layer (P<0.05). Pearson correlation analysis was undertaken to identify the relationships between SOC, TN, TP, TK, bulk density, soil capacity, total porosity, and capillary porosity. The relationship between soil physicochemical properties and soil particle-size distribution was analyzed by con rmatory factor analysis using the maximum likelihood method to build a path model. All statistical analyses were performed using SPSS 26.0 and AMOS 24.0(SPSS Inc. an IBM Company, Chicago, IL, USA), and all gures were prepared with Origin 2020 software (Origin Lab Inc., Northampton, MA, USA).

Soil nutrient content for the three plantation stands
Overall, SOC, TN, and TP showed a gradually decreasing trend from the litter layer to deep soil layers for the three plantation stands (Fig. 3).
The only exception was for the TP in the L. principis-rupprechtii stand, where there were no signi cant differences between different soil layers ( Fig. 3A-C). However, there was no signi cant difference in soil TK in the different soil layers for the three plantation species (Fig.   3D). Furthermore, the C:N ratio, C:P ratio, and N:P ratio exhibited a gradually decreasing trend from surface soil layers to deep soil layers; the exceptions were for the C:N ratio of the P. crassifolia stand and the C:P ratio and N:P ratio of the P. tabuliformis stand, where there were no signi cant differences between the different soil layers (Fig. 3E-G). On the whole, the SOC, TN, TP, C:N ratio, C:P ratio and N:P ratio of the L. principis-rupprechtii stand were higher than those of the P. crassifolia and P. tabuliformis stands; except for individual nutrient indexes (such as TN and N:P ratio), for which there was no signi cant difference between the surface and the deep soil layer.
Available nutrients ( -N, -N, AP, and AK) in the soil also exhibited a gradually decreasing trend from the topsoil to deep soil layers for the three forest stands (Fig. 4). The available nutrients were highest in the L. principis-rupprechtii stand ( -N is 26.28 mg·kg -1 , -N is 28.72 mg·kg -1 , AP is 4.62 mg·kg -1 and AK is 121.22 mg·kg -1 ), followed by the P. crassifolia stand ( -N is 24.77 mg·kg -1 , -N is 27.00 mg·kg -1 , AP is 2.61 mg·kg -1 and AK is 104.64 mg·kg -1 ), and lowest in the P. tabuliformis stand ( -N is 12.85 mg·kg -1 , -N is 15.29 mg·kg -1 , AP is 2.01 mg·kg -1 and AK is 86.78 mg·kg -1 ). The differences declined with depth in the soil pro le for the three forest stands, so there was no signi cant difference in AP and AK in the deepest layer.

Soil physical properties for the three plantation stands
The different tree species also had different effects on the soil physical properties of the different soil layers (Fig. 5). There was no signi cant difference in soil bulk density, soil capacity, soil total porosity, and soil capillary porosity in soil layers down to 30 cm under the P. tabuliformis stand (Fig. 5A-D). The soil capacity, soil total porosity, and soil capillary porosity exhibited a gradually decreasing trend from the topsoil to deep soil layers for L. principis-rupprechtii and P. crassifolia stands, while soil bulk density showed the opposite trend. The soil bulk density of P. crassifolia (0.92-1.26 g·cm -1 ) and P. tabuliformis stands (1.15-1.16 g·cm -1 ) was higher than that of the L. principisrupprechtii stand (0.76-0.94 g·cm -1 ) (Fig. 5A), while the soil capacity, soil total porosity, and soil capillary porosity of the L. principisrupprechtii stand were higher than those of the P. crassifolia and P. tabuliformis stands (Fig. 5B-D), except that there was no signi cant difference in the soil capillary porosity in the 20-30cm layer (Fig. 5D).

The correlation between soil nutrient content and physical properties
Pearson correlation analysis was performed to evaluate the correlation between soil nutrient content and physical properties ( Table 2). The results reveal that the SOC, TN, -N, -N, AP, and AK in the soil were signi cantly positively correlated with soil capacity, total porosity, and capillary porosity (P<0.05 or P<0.01), with the exceptions that there was no signi cant correlation between soil total porosity and -N, TP, and TK. However, there were signi cant negative correlation between soil bulk density and SOC, TN, -N, -N, AP, and AK (P<0.05 or P<0.01). In addition, the SOC, TN, -N, -N, AP, and AK contents of soil were signi cantly positively correlated with each other (P<0.05 or P<0.01), and the soil capacity, total porosity, and capillary porosity were also signi cantly positively correlated with each other (P<0.05 or P<0.01). These results indicate that the water permeability and water storage capacity of the soil signi cantly increased with increasing soil organic matter and available nutrients.

Soil particle-size distribution for the three plantation stands
The percentages of clay and silt in the topsoil of the three plantation stands were higher than those in the deep soil, which was gradually decreasing from topsoil to subsoil (Table 3). The different soil layers of L. principis-rupprechtii and P. crassifolia plantations had similar soil particle size composition. The sand content of subsoil was generally higher than topsoil, and the clay content of topsoil was generally higher than the subsoil. Moreover, the distribution of sand and clay contents of different soil layers had high heterogeneity for the Pinus tabuliformis plantation stand.
3.6 The relationship between soil particle-size distribution and soil physical and chemical properties Path analysis showed that soil organic carbon (SOC) had direct effects on clay (0.76), silt (0.66), and sand (-0.94), while total potassium (TK) had indirect effects on clay (0.75), silt (0.85), sand (-0.79). The SOC and TK hadnegative effects on sand, while SOC and TK hadpositive effects on clay and silt (Fig. 7). In addition, we can also nd that SOC had a negative effect on TN, while SOC had a positive effect on soil bulk density (BD). This was also con rms the results in Table 2. The different lowercase letters refer to significant differences among different soil layers in the same plantation stands (P 0.05).
-indicates no data.

The interaction between leaf, litter, and soil nutrient content for the three different plantations
The plant-soil feedbacks as drivers of plant community composition and species coexistence is increasingly being recognized (Kulmatiski et al. 2008;Kardol et al. 2013;Aponte et al., 2013). Previous studies have shown that the C, N, and P contents of plants will signi cantly affect soil nutrient contents, and they are often species-speci c in that different species have different nutrient contents and deliver different elemental contributions to soil (Vesterdal et al. 2008). As a result, the C:N:P stoichiometry of the soil will inevitably occur due to different litter inputs and rhizodeposition (Peichl et al. 2012;Wang et al. 2009;Zhang et al., 2019b), which is very important to improve our understanding of the relationship between plants and soil nutrient contents (Cleveland and Liptzin 2007;Zhao et al. 2015). In this study, we found that the C:N, C:P and N:P of leaves and litter of P. tabuliformis were higher than those of L. principis-rupprechtii and P. crassifolia. The possible reason for these results is that the C:N and the nutrient contents of the litter are the most directly factors in uencing decomposition rate and nutrient release of litters (Prescott 2010;Ge et al. 2013), and higher C:N and C:P are important in hindering the decomposition of litter. Conversely, the decrease in C:N and C:P means that litter is converted into a decomposed state more readily (He et al. 2010). Moreover, the C:N and C:P of litter are negatively correlated with decomposition rate, and the litter with higher C:N and C:P needs to obtain a large amount of N and P from external sources to accelerate decomposition (Wang and Huang 2001;He et al. 2010). In addition, Because P. tabuliformis litter is richer in lignin than L. principis-rupprechtii and P. crassifolia litter, decomposition is hampered (He et al., 2010). This conclusion is also con rmed by previous studies indicating that the lignin/N or crude ber/N re ects the ease of litter decomposition: decomposition rate is negatively correlated with this ratio (Wang and Huang 2001;He et al. 2010). These reasons also explain why the soil nutrient contents of the P. tabuliformis plantation is lower than those of the L. principis-rupprechtii and P. crassifolia plantations.
Soil C:N, C:P, and N:P are important indexes for determining the mineralization and xation of soil nutrients during soil development (Tian et al. 2010). The N:P of soil not only re ect the availability of P and N in the forest ecosystem but also reveal nutrient movements between the plants and soils (Cao and Chen, 2017;Fan et al., 2015). Our research results showed that the soil nutrient contents of the three plantations gradually decreasing trend from the topsoil to deep soil layers. This trend was because C, N and P released by litter decomposition were mainly concentrated in the topsoil layer, with only a small percentage of nutrients reaching the deeper soil layers. Besides, the L. principisrupprechtii stand had the highest soil C:N, C:P, and N:P, and the C:N and C:P in leaves of L. principis-rupprechtii were higher than those of the other two tree species. Therefore, it can be speculated that the P. tabuliformis forest requires a larger amount of phosphorus than L. principis-rupprechtii and P. crassifolia stands, resulting in a decrease in phosphorus content in the soil. These also con rm the idea that the higher C:N and C:P values in plants usually represent higher N and P utilization (Wardle et al., 2004).

Effects of the three different plantations on soil nutrients
It is well known that the soil is critical to maintaining the productivity and sustainability of forest ecosystems, and the ability of forest soil to store and transform organic material is in uenced by the soil organic matter, which can be in uenced by forest vegetation types Xia et al. 2019). Therefore, knowledge about the soil nutrients in different forest soils is of great importance to understanding biogeochemical cycles (Yang et al. 2010). The results of this study show that the SOC, TN, and available nutrients ( -N, -N, AP, and AK) were highest in the L. principis-rupprechtii stand, followed by the P. crassifolia stand, and lowest in the P. tabuliformis stand. Moreover, all nutrient contents declined with depth in the soil pro le layer in the three different stands. This is because L. principis-rupprechtii is a deciduous coniferous specie, so the biomass of litter input is higher than that of P. crassifolia or P. tabuliformis. In addition, the soil organic carbon accumulation may be mainly driven by litter inputs (Zhao et al. 2017), and higher C:N and C:P hinder the decomposition of the litter layer (He et al. 2010). This suggests that the SOC, TN, and available nutrient contents in the L. principis-rupprechtii stand are higher than in the P. crassifolia and P. tabuliformis stands. Besides, the amount of potassium in the soil is directly related to the parent material (Mishra et al. 2017). Potassium in plants is involved many important biochemical processes, such as activation of biological enzymes, ion channels, synthesis of macromolecules, and regulation of transpiration, etc (Mishra et al. 2017), and the availability of potassium in the soil is maintained by the decomposition of organic matter (Basumatary and Bordoloi 1992). This is the reason that there is no signi cant difference in TK, although there was a signi cant difference in AK content between different soil layers in the three stands.

The relationship between soil nutrient contents and soil physical properties for the three different plantations
The water storage capacity of the soil is in uenced by soil physical and chemical properties (Guzman et al. 2019). Soil bulk density and soil capacity play an important role in hydrological processes, which are essential to the supply and storage of water, nutrients, and oxygen in the soil (Krainovica et al. 2020;Wang et al. 2010). The size of soil pores porosity plays a key role in quantifying soil structure because it can affect soil hydraulic conductivity, solute convection and water retention (Zhang et al. 2019a). Therefore, these indicators can be used as indicators to evaluate the impact of vegetation restoration on soil properties (Gu et al. 2019). The results of this study indicated that the content of SOC, TN, soil available nutrients and soil capacity, soil total porosity, and soil capillary porosity exhibited the same changing trends for different plantation species, and the correlation analysis ( Table 2) also showed that the SOC, TN and available nutrient contents ( -N, -N, AP, and AK) were positively correlated with soil capacity, soil total porosity, and soil capillary porosity, while negatively correlated with soil bulk density. Furthermore, soil bulk density increased with the increase of soil depth, and the differences in soil bulk density between the different species stands were mainly related to the degree of decomposition and amounts of easily decomposable litters. Our results are in agreement with previous studies showing that increases in SOC are associated with an increase in soil total porosity (Abu, 2013) and decreases in soil bulk density (Koestel et al. 2013). Besides, there was also a signi cant positive correlation between SOC and available nutrients ( -N, -N, AP, and AK) ( Table 2). These results indicate that the soil physical characters and water conservation capacity are largely affected by soil nutrient contents after afforestation.
4.4 The relationship between soil particle-size distribution and soil physical and chemical properties for the three different plantations In general, the vegetation can not only improve soil fertility, increase carbon storage, enhance water conservation capacity, etc., but also improve soil particle composition, reduce the content of sand and silt, increase the content of clay, and thus improve soil structure (Su et al., 2018;Xia et al., 2020). The results of this study indicated that L. principis-rupprechtii and P. crassifolia plantations could signi cantly improve the nutrient contents of topsoil, make the topsoil particles ner, the clay content increased, and sand decreased than P. tabuliformis plantation. The main reason is that L. principis-rupprechtii and P. crassifolia plantations had higher soil nutrient returning capacity than P. tabuliformis plantation, which further increases the soil nutrient contents, improves the soil structure, and promotes the formation of soil clay. The soil particle-size distribution is closely related to soil organic carbon content and has a signi cant in uence on soil organic carbon conversion (Von Lützow & Kögel-Knabner, 2009). Generally, soil organic carbon is easy to combine with ner soil particles (silt and clay) to form organic-inorganic complexes. Meanwhile, the surface area is relatively large of silt and clay, which will expose more positive charges and combine with negatively charged humus (Zhao et al., 2014). On the other hand, the ner particles have poor permeability, and the organic carbon is more di cult to be decomposed by microorganisms once it combines with them. Compared with ner clay, sand particles are opposite to each other. Because sand have fewer positive charge sites and large particles, they have fewer opportunities to combine with organic carbon. Moreover, sand has strong permeability, the looser the soil structure and poorer soil water holding capacity, which can be easily decomposed by microorganisms (Zhao et al., 2014;Xia et al., 2020). Therefore, this is also the reason why the clay has a negative effect on soil bulk density.

Conclusions
In this study, we investigated the in uence of different tree species on the nutrient cycling of plants and soils and plant-soil feedbacks, as well as the interaction between soil physicochemical properties in semi-arid mountain forest ecosystems. Our study suggests that L.
principis-rupprechtii and P. crassifolia had higher TN, TP, and TK contents in their leaves and litter layer than P. tabuliformis, while P. tabuliformis had higher organic carbon, C:N and C:P in leaves and litter than. L. principis-rupprechtii and P. crassifolia. This suggests that higher C:N and C:P hinder the decomposition of litter. Thus, the leaves and litter layer strongly in uence soil nutrient availability through their biogeochemical processes, and L. principis-rupprechtii and P. crassifolia plantation forests were substantial improvement in soil nutrients and clay component than P. tabuliformis plantation forest. In addition, the L. principis-rupprechtii and P. crassifolia signi cantly improved the clay component, soil capacity, soil total porosity, and capillary porosity, decreased soil bulk density and sand component associated with a larger void ratio, and enhanced water storage capacity. In conclusion, we recommend planting the L. principis-rupprechtii and P. crassifolia as the preferred tree species to enhance the water conservation function, increasing soil fertility, which should be useful for ecological vegetation construction and management in semi-arid mountain forest ecosystems.