*Article* **Source and Accumulation of Soil Carbon along Catena Toposequences over 12,000 Years in Three Semi-Natural** *Miscanthus sinensis* **Grasslands in Japan**

**David S. Howlett 1, J. Ryan Stewart 2, Jun Inoue 3, Masanori Saito 4, DoKyoung Lee 5, Hong Wang 6, Toshihiko Yamada 7, Aya Nishiwaki 8, Fabián G. Fernández <sup>9</sup> and Yo Toma 10,\***


**Abstract:** *Miscanthus*-dominated semi-natural grasslands in Japan appear to store considerable amounts of soil C. To estimate the long-term effect of *Miscanthus* vegetation on the accumulation of soil carbon by soil biota degradation in its native range, we measured total soil C from the surface to a 1.2 m depth along a catena toposequence in three annually burned grasslands in Japan: Kawatabi, Soni, and Aso. Soil C stock was estimated using a radiocarbon age and depth model, resulting in a net soil C accumulation rate in the soil. C4-plant contribution to soil C accumulation was further estimated by δ13C of soil C. The range of total soil C varied among the sites (i.e., Kawatabi: 379–638 Mg, Soni: 249–484, and Aso: 372–408 Mg C ha−1). Catena position was a significant factor at Kawatabi and Soni, where the toe slope soil C accumulation exceeded that of the summit. The soil C accumulation rate of the whole horizon in the grasslands, derived C mainly from C4 plant species, was 0.05 <sup>±</sup> 0.02 (Average <sup>±</sup> SE), 0.04 <sup>±</sup> 0.00, and 0.24 <sup>±</sup> 0.04 Mg C ha−<sup>1</sup> yr−<sup>1</sup> in Kawatabi, Soni, and Aso, respectively. Potential exists for long-term sequestration of C under *M. sinensis*, but the difference in the C accumulation rate can be influenced by the catena position and the amount of vegetation.

**Keywords:** *Miscanthus sinensis*; soil carbon; catena; radiocarbon dating; C4 grasses

#### **1. Introduction**

*Miscanthus*, a cold-tolerant perennial grass C4 native to East and Southeast Asia, exhibits potential as a feedstock for the production of biofuels and bio-based products [1–4]. As such, this genus may see a considerable increase in cultivation in the United States and Europe in the coming years. In order to estimate the potential effects on edaphic resources,

**Citation:** Howlett, D.S.; Stewart, J.R.; Inoue, J.; Saito, M.; Lee, D.; Wang, H.; Yamada, T.; Nishiwaki, A.; Fernández, F.G.; Toma, Y. Source and Accumulation of Soil Carbon along Catena Toposequences over 12,000 Years in Three Semi-Natural *Miscanthus sinensis* Grasslands in Japan. *Agriculture* **2022**, *12*, 88. https://doi.org/10.3390/ agriculture12010088

Academic Editor: Claudia Di Bene

Received: 21 November 2021 Accepted: 31 December 2021 Published: 10 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

several researchers have considered the impact this genus has on soil carbon, mostly in cultivated or fallow fields [5–9]. Observations, however, from these studies have been limited to less than 20 years. Semi-natural *Miscanthus sinensis* grasslands in Japan, some of which have been managed for hundreds of years [4], offer an opportunity to assess the effects of centuries of *Miscanthus* growth and management on soil C resources [10–15].

*Miscanthus* utilizes the efficient C4 photosynthetic pathway and, consequently, the origin of organic inputs to the soil from plants in this genus can be determined via their stable isotopic composition [9,16,17]. Differences in the relative abundance of the 13C/12C ratio in plants that utilize the C4 photosynthetic pathway allow for determining the relative contribution by *Miscanthus* to soil C stocks [18]. Using stable C isotopic analysis, Schneckenberger and Kuzyakov (2007) estimated *Miscanthus* C inputs ranging between 0.11 and 0.30 g C kg soil−<sup>1</sup> yr−<sup>1</sup> in a sandy versus loamy soil in Germany. Howlett et al. (2013) found that a majority of soil C, ranging between 52% and 85% at a depth up to 1.5 m, was derived from *Miscanthus* in a Typic Melanudans in a southern Japanese *Miscanthus*dominated grassland. To estimate soil C accumulation over time, the relationship between soil depth and age needs to be determined. Dating of soil C in soil profiles with low C content, however, is problematic. Bioturbation causes vertical mixing or movement of soluble C compounds, and additions of heterogeneous sources of C from land surface and groundwater reduce the integrity of age-to-depth models [19,20]. However, soils previously investigated in the same biome contained high C content and demonstrated highly correlated age-to-depth models (*R*<sup>2</sup> = 0.98–0.99) [11]. Using these age-to-depth models, Howlett et al. (2013) estimated *Miscanthus*-derived soil C accumulation at 0.62–0.85 Mg C ha−<sup>1</sup> yr−<sup>1</sup> down to a 1.5 m depth in a *Miscanthus*-dominated semi-natural grassland in southern Japan.

In volcanic regions with diverse topography where ash accumulation and other formative materials, such as humus, are subject to erosion and deposition processes along the continuum of different landscape positions [21], a soil catena study that encompasses hillside summits, mid-slopes, and toe slopes can shed valuable information on C accumulation phenomena. The dynamics of C associated with these topographic forms have previously been considered. Schimel et al. (1985) [22] reported double the surface soil C on lower foot slopes relative to that found in summit soils in Colorado, USA. While fully vegetated grasslands may not experience a considerable amount of erosion, the annual burning associated with traditionally managed *Miscanthus* grasslands in Japan removes the vegetative cover that protects soil from erosion [11,13]. Precipitation events following these traditional burnings redistributes C-containing sediments lower into the watershed, and soils developed on lower positions on a slope may demonstrate higher levels of soil C due to the fact of depositional processes. As soil from organic horizons erode and C-containing sediment accumulates below, the catena concept provides utility in characterizing the potential variability of soil C within the varied topography of many Japanese *Miscanthus* grasslands. Understanding the variability and accumulation of soil C underlying *Miscanthus* grasslands in Japan will help to determine whether there is greater benefit in growing *Miscanthus* as a soil C sequestration bioenergy crop or to serve alternative purposes such as a traditional landscape for ecotourism. Our efforts to identify the long-term impacts of *Miscanthus* on soil C deposition involved a three-pronged approach: (1) estimate the effect of catena position on the development of C stocks to 1.2 m underlying three semi-natural grassland catenas currently dominated by *M. sinensis* in Japan, (2) quantify the relative contribution of *M. sinensis* to these C stocks, and (3) calculate the rate of soil C accumulation contributed by *M. sinensis*.

#### **2. Materials and Methods**

#### *2.1. Site Descriptions*

Three grasslands sites were chosen based on their current dominance by *M. sinensis* and for a latitudinal gradient across Japan (Figure 1a).

#### 2.1.1. Kawatabi

The northernmost sampling site was at the Kawatabi Field Science Center of Tohoku University, located near the Kawatabi natural springs (KAW), Miyagi Prefecture, Japan (38◦46.25' N, 140◦45.16' E, 550 m a.s.l., approximately 14◦ of the slope), where the mean annual temperature is 11 ◦C, and there is a mean annual precipitation of 1460 mm [23] (Figure 1b). The site is dominated by *M. sinensis*, which is maintained by annual mechanical cutting in the fall with the grass left in place after cutting. Burning as a maintenance practice in this site ceased more than 40 years ago. Ito and Saigusa (1996) described several soil profiles from this site. One key aspect of this site is the documented non-allophanic chemistry of the Andisols. High contents of Al and Fe help to retain relatively large amounts of organic matter facilitating the formation of stable organo–mineral complexes, which likely also occur at the other two sites.

#### 2.1.2. Soni Plateau

The middle latitude site was within the grasslands at the Soni Plateau (SONI) in Nara Prefecture, Japan (34◦31.07' N, 136◦09.80' E, 720 m a.s.l. approximately 30◦ of the slope), where the mean annual temperature is 12 ◦C, and it has a mean annual precipitation of 1720 mm (Figure 1c). The *M. sinensis*-dominated grasslands is maintained by annual burning in the spring, and the area is a tourist destination for recreation and ecotourism. Soils are also Typic Andisols with contents of volcanic glass [24]. Inoue et al. (2012) [24] and Okunaka et al. (2012) [25] reconstructed the vegetative history of the site, with *M. sinensis* becoming dominant on the site 1500 years ago.

#### 2.1.3. Aso-Kuju

The southern-most site is within the grasslands of the Aso-Kuju National Park (ASO) in Kumamoto Prefecture, Japan (32◦55.75' N, 131◦09.60' E, 843 m a.s.l., approximately 21◦ of the slope), where the mean annual temperature is 13 ◦C, and it has a mean annual precipitation of 3200 mm (Figure 1d). The *M. sinensis*-dominated site is considered a seminatural grassland, which has been annually burned for hundreds of years in early spring in order to maintain the *Miscanthus*-grassland ecosystem [4,13,26]. No additional management has taken place other than burning for at least 50 years [11]. The *Miscanthus* grasslands of

Aso-Kuju National Park are a tourist destination partly due to the rarity of grasslands in Japan. Moreover, ecotourism during the burning season significantly contributes to the local economy. Soils in this region are typical of Japan, derived from volcanic ash and characterized by the USDA soil classification system as Typic Melanudans [27]. A diagnostic general characteristic of these soils is the presence of a 2AB (K-Ah) volcanic deposit at approximately 60–70 cm depths in many parts of the caldera, which has been dated to a local volcanic eruption event by Mount Kikai approximately 7300 years ago [15,28].

#### *2.2. Field Collections*

At each of the three sites, soils were sampled at 10 cm increments down to 120 cm, following a hillside catena sequence transect, representing the summit, mid-, and toe slope. Transect lengths were 200–250 m from the toe slope to the summit. Replicate soil samples were taken 3 m on the left- and right-hand sides of each catena location (facing the summit). We selected the east and west slope aspects at each site to avoid known edaphic differences between north- and south-facing slopes. The slopes of the catena sequences varied among the sites, and the mean percent slope was calculated by dividing vertical distance from the summit to the toe slope by the transect distance. The mean percent slopes were 29% in KAW, 36% in SONI, and 18% in ASO. In total, 27 soil catenas were examined (three sites, three catena positions, and three replicates per catena position).

To estimate bulk density, a metal canister of a known volume (i.e., 100 cm3) was inserted into the soil profile and removed with an undisturbed soil core at four representative depths (15, 45, 75, and 105 cm, which were the midpoints for 0–30, 30–60, 60–90, and 90–120 cm soil depths). Bulk density samples were dried at 100 ◦C to a consistent weight.

#### *2.3. Laboratory Procedures*

Soil samples from each of the 10 cm depth increments for soil C content analysis were immediately stored in temperature-controlled conditions and then air-dried to a constant weight. Air-dried soils were passed through a 2 mm sieve, and subsamples were taken for determination of moisture content. Soil C content was determined by combustion of 2 mm sieved soil in an elemental analyzer (Leco CN Analyzer, St. Joseph, MI, USA).

Stable isotopic C composition and accelerator mass spectrometer (AMS) radiocarbon ages were determined for five soil depth increments: 0–10, 20–30, 50–60, 80–90, and 110–120 cm at each of the catena positions at all three sites. Soil samples with high organic C content were pretreated using the standard acid–base–acid (ABA) method as described by Brandt et al. (2012) [29]. The same pretreatment method was also applied to radiocarbonfree wood, IAEA (International Atomic Energy Agency) C5 wood, and FIRI-D (Fifth International Radiocarbon Inter-Comparison D) woodworking standards. Approximately 0.5 g of soil, 3–5 mg of working standards, and 200–300 mg of CuO granules were placed into preheated quartz tubes for sealed quartz tube combustion at 800 ◦C. Quartz tubes were preheated at 800 ◦C for 2 h, and CuO granules were preheated at 800 ◦C one day before usage. Combustion was set for 2 h at 800 ◦C. Samples were then cooled slowly from 800 to 600 ◦C for 6 h to allow Cu to reduce the NxO to nitrogen gas. Purified CO2 was collected cryogenically under vacuum conditions, which were less than 10 mTorr, and submitted to the Keck Carbon Cycle AMS Laboratory of the University of California-Irvine for AMS 14C analysis using the hydrogen–iron reduction method with δ13C values measured on prepared graphite [30]. All results were corrected for isotopic fractionation according to the conventions of Stuiver and Polach (1977) [31]. Sample preparation backgrounds were subtracted based on the measurements of radiocarbon-free wood blanks. The results indicated that after background subtraction, IAEA-C5 and FIRI-D wood reference materials yielded target values within 1σ deviations. Radiocarbon dating data greater than 100% of modern C were considered as present C, which was fixed from 1950 to 2012.

Soil texture was determined using the laser diffraction method [32]. Soil pH was measures in a 1:1 ratio of soil to water. Plant-available, exchangeable potassium, magnesium, and calcium were determined with the Bray-1 extraction method [33]. The content

of cations, cation exchange capacity, and percent base saturation of cation elements were calculated from extract results.

#### *2.4. Calculations*

Bulk density, estimated by dividing the oven-dried mass (g) by the canister volume, C content in <2 mm bulk soil (%), and the soil bulk density (*BD*, g cm−3), was used to estimate soil C stock per 10 cm depth increments (Mg C ha<sup>−</sup>1).

$$\text{Soil C stock} = \text{Soil C content} \times BD \times 10,\tag{1}$$

To estimate C4 plant contribution to soil C, *δ*13*C* of soil C (*δ*<sup>13</sup>*CSC*, ‰) were calculated as follows:

$$\text{J}^{13}\text{C}\_{\text{SC}} = \text{[}(\text{R}\_{\text{sample}}/\text{R}\_{\text{standard}} - 1)\text{]} \times 1000,\tag{2}$$

where *R* is the ratio of 13C/12C in bulk soil C. The standard was V-PeeDee Belemnite (V-PDB) carbonate. The measured *δ*13*C* values were converted to relative abundances of C3 and C4 plants using the mass balance equation:

$$\delta^{13}\mathcal{C}\_{\text{SC}} = |\langle \delta^{13}\mathcal{C}\_{\text{C4}}\rangle \times \mathfrak{x}\rangle + |\langle \delta^{13}\mathcal{C}\_{\text{C3}}\rangle \times (1-\mathfrak{x})\rangle,\tag{3}$$

where *x* indicates the ratios of C source derived from C4 and C3 plants, which were −13‰ and −27‰, respectively, and were used as the average values of *<sup>δ</sup>*<sup>13</sup>*CC*<sup>4</sup> and *<sup>δ</sup>*<sup>13</sup>*CC*<sup>3</sup> for calculation.

Linear regression, completed using PROC REG in SAS (version 9.2, Carey, NC, USA), was used to estimate soil C age at various soil depths with *p* < 0.05. The goodness of model fit and significance were estimated by *R*<sup>2</sup> and *p*-values. Profile summaries were calculated from the summed C stock for each treatment combination. Sampling sites were not compared, and only the catena position effect was assessed for total soil C and C4-source C within each soil depth at each site (ANOVA, PROC GLM in SAS).

Accumulation of C4-C (*Cflux*, Mg C ha−<sup>1</sup> yr−1) was calculated using soil C content (g C 100 g<sup>−</sup>1), sedimentation rates (SR, cm yr−1, from the surface down to 1.2 m), *BD*, and *x*, which is the C4-derived C content from 13C abundance in Equation (3) as follows:

$$\mathbf{C\_{flux}} = \text{Soil C content} \times \text{SR} \times BD \times \mathbf{x}\_{\prime} \tag{4}$$

Combined with the known depth of each of the soil profiles and C stock data, radiocarbon dating of the profiles was used to generate age–depth models to estimate the sedimentation rate to 1.2 m for total C and C from C4 plant sources as per the methods of Howlett et al. (2013). The risk exists for *δ*13*C* to become less negative due to the fact of isotopic fractionation, which could introduce uncertainty in determining the contribution of C from C3 and C4 plants. However, degradation-induced fractionation is essentially negligible, because new additions of organic C in the mesic *Miscanthus*-grassland ecosystem generally overwhelm the oxidation of the soil organic C pool. Moreover, decomposition only enriches less than 1–2‰ for soils in dry and/or hot environments, which was not the case in our study.

#### **3. Results**

Selected soil physical and chemical properties are shown in Table 1. Values represent the averages of whole soil samples from the surface to a 1.2 m depth, because soil samples were collected at 10 cm soil depth increments and could not be presented by soil horizons. Selected soil physical and chemical properties show that the soils from the three *Miscanthus* sinensis-dominated grassland catenas were low in pH, had a texture from silt to silt loam, and had low to moderate CEC (Table 1). Low BD is typical of volcanic ash-derived soils.


**Table 1.** Soil physical and chemical properties (Average ± SD) underlying three *Miscanthus sinensis* grasslands in Japan (i.e., Kawatabi Field Science Center, Miyagi Prefecture (KAW); Soni Plateau, Nara Prefecture (SONI); Aso-Kuju National Park, Kumamoto Prefecture (ASO)).

§ Cation exchange capacity; † exchangeable.

Whole-profile soil C stocks across all sites ranged from 249 to 640 Mg C ha−<sup>1</sup> for a 0–1.2 m depth (Figure 2). Across the study sites, the position along the catena sequence was a significant factor at KAW and SONI only. At KAW, soil C stock in the toe slope (640 Mg C ha−1) was greater than in the summit (379 Mg C ha−1) but statistically similar to that in the mid-slope (532 Mg C ha<sup>−</sup>1). At SONI, the pattern of the distribution of soil C stock was similar as in the mid-slope (483 Mg C ha<sup>−</sup>1) and in the toe slopes (358 Mg C ha−1) exceeded in the summit (249 Mg C ha<sup>−</sup>1). However, no differences in the soil C stock among catena positions were found at ASO.

**Figure 2.** Profile summary of mean soil carbon stock along three catenas at semi-natural *Miscanthus sinensis* grassland sites in Japan (i.e., Kawatabi Field Science Center, Miyagi Prefecture (KAW); Soni Plateau, Nara Prefecture (SONI); Aso-Kuju National Park, Kumamoto Prefecture (ASO)). Error bars represent the standard errors. Statistically different means within a site are noted by means separation letters (*p* < 0.05).

Catena position was also a significant factor for soil C stocks at certain depths at each site (Figure 3, Supplement Table S1). At KAW, soil C stock in the toe slope was greater than that in the summit slope from 50 to 120 cm (Figure 3a), while it was higher in surface soil at the summit. At SONI, soil C stock in the mid-slopes demonstrated higher levels only between 50 and 100 cm depths, but it was statistically indistinguishable from toe slopes at most of these depths (Figure 3b). At ASO, a relatively higher soil C stock at the summit was observed from the surface down to a 20 cm depth (Figure 3). It was lower at the summit compared to those at the mid- and toe slopes from 30 to 80 cm depths of soil. Below 80 cm of soil, however, soil C stock increased and was higher at the summit.

**Figure 3.** Soil carbon stock at 10 cm increments down to a 1.2 m depth for three positions along catenas at three semi-natural *Miscanthus sinensis* grassland sites in Japan: Kawatabi Field Science Center, Miyagi Prefecture (KAW) (**a**); Soni Plateau, Nara Prefecture (SONI) (**b**); Aso-Kuju National Park, Kumamoto Prefecture (ASO) (**c**). Error bars represent the standard errors.

The soil C accumulation rate for C4-based C across all sites within 10 cm soil depth increments ranged from 0.00 to 0.29 Mg C ha−<sup>1</sup> yr−<sup>1</sup> (Figure 4). Within each site, mean C4-C accumulation for the whole profile (0–120 cm) was 0.05 ± 0.02 (Average ± SE), 0.04 ± 0.00, and 0.24 ± 0.04 Mg C ha−<sup>1</sup> yr−<sup>1</sup> at KAW, SONI, and ASO, respectively. At ASO, C4-C constituted the vast majority of total C, especially from 80 to 120 cm (Figure 4c). To a lesser extent, KAW soil C was mostly C4-C (Figure 4a). In addition, C4-C closely followed the trend of total C, decreasing in content from 40 to 80 cm. At SONI, C4-C comprised the majority of total C (Figure 4b). ASO had the highest mean content of C4-C at 86.3% (57.0–100%) with KAW at 58.2% (28.6–99.1%) and SONI at 56.3% (37.0–76.9%).

**Figure 4.** Soil carbon accumulation rate and relative soil age (years before present) for C4-derived C (clear diamonds) and total C (black squares) to a 1.2 m in three semi-natural *Miscanthus sinensis* grassland sites in Japan: Aso-Kuju National Park, Kyushu Prefecture (ASO) (**a**); Kawatabi Field Science Center, Miyagi Prefecture (KAW) (**b**); and Soni Plateau, Nara Prefecture (SONI) (**c**). Error bars represent the standard errors.

One major difference between ASO and the two other sites was the age of the bottom soil depth at 1.2 m. At ASO, the age of the 110–120 cm depth was dated 1590 years before present, while the 110–120 cm depth at KAW and SONI was closer to 7836 and 6415 years before present, respectively (Figure 4). As such, the profiles at ASO represent a more recent portion of the age ranges found in the other sites (Figure 4). The age-to-depth models used to calibrate the soil ages throughout the profile were highly correlated with an *R*<sup>2</sup> in the range of 0.83–0.98 and *p* < 0.05 (Figure 5, Supplement Table S2). The only exception was the toe slope at SONI with an *R*<sup>2</sup> of 0.79 (Figure 5b).

**Figure 5.** Calibrated radiocarbon ages to a 1.2 m depth of soil carbon at three catena positions (i.e., toe slope, mid-slope, and summit) in three semi-natural *Miscanthus sinensis* grassland sites in Japan: Kawatabi Field Science Center, Miyagi Prefecture (KAW) (**a**); Soni Plateau, Nara Prefecture (SONI) (**b**); Aso-Kuju National Park, Kyushu Prefecture (ASO) (**c**).

#### **4. Discussion**

#### *4.1. Soil Carbon Stock along Catena*

Soil C stocks in the *M. sinensis*-dominated grasslands of Japan appeared to be influenced by catena position along a toposequence. For total accumulated soil C stock for the 0–120 cm soil depths, toe and mid-slopes demonstrated a long-understood tendency to be the recipient of C-containing sediments from the summit (Figure 3) [21,22,34]. Jenny (1941) [34] indicated that erosion does not play a significant role in some well-vegetated catenas, given their minimal degree of erosion. However, the annual cultural practice of burning the *M. sinensis* grasslands in Japan reduces biotic control of erosion. Movement of sediment from organic horizons follows the course of gravity and increases C stock in mid- and toes slopes (Figure 3). This may partially help explain the pattern of soil

C accumulation found at KAW and SONI (Figure 3a,b). Upon further investigation of differences in soil C stocks at various depths within the catenas, many of the differences seen in KAW and SONI only occurred below 60 cm depths (Figure 3a,b), coinciding with soil ages in the range of approximately 4000 years before present. Differences in soil C in deeper layers across different topographic positions unlikely reflect differences in current vegetation, because *M. sinensis* rhizomes and roots mainly populated the surface layer down to a 20 cm depth [35]. Because the age of C in sediments at ASO was much younger than the other study sites, investigations of deeper soil profiles at ASO might be required to determine if catena position is a significant factor in determining soil C stocks at ASO, where no differences were found. While catena position does appear to affect soil C stock at KAW and SONI, the effect occurred between 4000 and 8000 years before present.

In all catena grasslands examined, very high total soil C stocks were found in upland (non-hydric) soils (up to 638 Mg C ha<sup>−</sup>1) (Figure 4). The likely presence of high contents of Al and Fe possibly contributed to large quantities of humus stabilization in the volcanic soils of this study [11,36,37]. Formation of recalcitrant organo–mineral complexes with Al and Fe reduces translocation and mineralization of C in the soil [38,39]. This may also help explain how the relatively high amounts of sequestered C [36,40] and low pH (~5), especially at KAW, contributed to the formation of these complexes [41]. In addition, the presence of these organo–mineral complexes might possibly explain the relatively high correlations of determination that provided confidence to the sedimentation rate calculations used to estimate C accumulation. Because we could not analyze the organo–mineral complexes with Al and Fe in this study, these analyses and evaluation need to be addressed in future research. Furthermore, regularly occurring fire events over hundreds of years at ASO and SONI may also have contributed to the stabilization of soil C. Burning has been shown to increase the stability of organic matter through the formation of highly condensed aromatic compounds [42]. Thus, the evaluation of soil humus characteristics could be important variables to include in future studies.

Toma et al. (2012) [13] and Howlett et al. (2013) [11] provided a broad review of work on C sequestration in soils where *Miscanthus* has been long established. Previous work at ASO demonstrated high total C stock levels down to a 1.5 m depth (515 and 559 Mg C ha<sup>−</sup>1) in two soil profiles dated, at a maximum, to 12,000 years before present [11]. The site, characterized by Howlett et al. (2013) [11], was relatively flat where erosion appeared to not be a significant factor. We considered nine soil profiles at ASO only to a depth of 1.2 m on younger soils where humus may not have had as much time to accumulate. As such, the results reported here appear to be consistent with previous work at ASO [11]. However, as with the site studied by Howlett et al. (2013) [11], there appears to be a buried organic soil horizon nearly 80 cm below the soil surface as reflected by the notable increase in soil C accumulation rates starting at that depth (Figure 4c). Basile-Doelsch et al. (2005) [43] reported high soil C stock levels 100 cm belowground of a volcanic–ash soil, which was located adjacent to the Piton des Neiges volcano on the island of La Reunion, where a burial event occurred sometime in the distant past.

#### *4.2. Source and Rate of Carbon Accumulation*

We assumed, for the purpose of this study, that all C4-C was derived from *Miscanthus*, as no other known species in the study areas utilize the C4 photosynthetic pathway. Miyabuchi and Sugiyama (2006) [44] detailed the dominance of *M. sinensis* via plant phytolith analysis in semi-natural grasslands located on the east side of the Aso caldera in Aso-Kuju National Park. While accumulation of C from C4 sources follows the trend of total soil C accumulation throughout the profiles examined here, the content of C4-C varied considerably among sites (Figure 4). Soil at ASO had the highest amount of C4 derived C accumulation, representing nearly 100% of C from 70 to 120 cm (Figure 4c). The C4-C and total-C accumulation trends at KAW and SONI underscore the importance of *Miscanthus*-derived C, but since not all C was from C4 sources, additions of C from non-*Miscanthus* sources were consistent with the presence of other plant species over time. Although currently dominated by *M. sinensis*, the composition of vegetative inputs in the plant community to soil C varied over the 12,000 year period [11,25]. At SONI, previous work identified charcoal remnants from anthropogenic fires that began around 7000 years before present with phytolith data indicating a vegetative shift [25]. If C4-C was mostly *Miscanthus*-derived, the general increase in C4-C accumulation at KAW (Figure 4a) may indicate a vegetative change from forest to grassland, which may have promoted more C storage in soil C pools relative to aboveground biomass [13,45,46].

Rates of *Miscanthus*-source soil C accumulation, highest at ASO, may be an indication of the greater net primary production that occurred under warmer and nearly double the precipitation than that observed at SONI and KAW (Figure 4). Howlett et al. (2013) [11] measured soil C accumulation at ASO on a site 14 km northwest of the current study site and found mean C4-C accumulation rates between 0.62 and 0.85 Mg C ha−<sup>1</sup> yr−<sup>1</sup> down to 1.5 m in the soil. These previously studied profiles likely represented several buried organic horizons dating to approximately 12,000 years before present, where humus accumulation occurred over an extensive period. As mentioned above, a similar phenomenon appears to have occurred at the current study site, where a buried organic horizon appears nearly 80 cm below the soil surface but is considerably younger (Figure 4c).

As suggested by Chaopricha and Marin-Spiotta (2014) [47], soil burial is a globally important, yet largely underestimated, process involved in the storage and persistence of substantial C stocks in soils. Indeed, volcanic soils buried 3 m below the surface on the slopes of Mount Kilimanjaro were estimated to contain 820 Mg C ha−<sup>1</sup> [48]. In addition, Inoue et al. (2000) [49] found that high soil C levels, which were buried multiple times over for several thousand years in a volcanic basin 135 km south of Aso, had not substantially decreased since the initial burial events. Similarly, based on our data and that of Howlett et al. (2013) [11], we strongly suspect that large reservoirs of C are stored in buried soils throughout the Aso volcanic caldera. Indeed, several volcanic eruptions have occurred in the ASO area over the past several thousand years, including several that occurred in the early 1200s [50], which coincide with the putative burial event seen in the soil profile at the current study site. These events suggest that soil burial due to the soil sedimentation resulting in volcanic ash deposition and plant residue accumulation acts as an important process in maintaining soil C levels at deeper soil layers under the stable thermal environments and anaerobic conditions. Possibly due to the more recent eruption event, the current study site had much more ash deposition in the subsurface soil than the study site of Howlett et al. (2013) [11], which was likely due to the current site being 6.3 km closer to the volcano at Aso. Moreover, more ash deposition likely occurred given the westto-east prevailing wind direction in the region. The study site of Howlett et al. (2013) [11] was 15.9 km north of the volcano, whereas the current study site was 9.7 km east of the volcano. In this study, we report C4-C accumulation rates that were 3–4 times lower than that reported by Howlett et al. (2013) [11]. Given that the soil-C measurements of the current study were taken to only a 1.2 m depth in comparatively younger soils (to 1590 years before present) at ASO may explain the lower soil C accumulation rates. Zehetner (2010) [51] reported that soil C accumulation rates can range between 0.3 and 0.6 Mg C ha−<sup>1</sup> yr−<sup>1</sup> in relatively volcanic–ash soils. However, most studies on soil C accumulation in cultivated *Miscanthus* fields have reported considerably higher rates. Soils where *M. sinensis* had been established for 6 [7] and 14 years [52] under managed conditions in southeastern England accumulated C at approximately 0.80 Mg ha−<sup>1</sup> yr−1, which is similar to what Poeplau and Don (2014) [53] found in an analysis of six *Miscanthus* plantations ≥10 years old across Europe (0.78 Mg ha−<sup>1</sup> yr−1). In addition, based on 23 data sets, Agostini et al. (2015) and Qin et al. (2016) [54] both calculated global estimates of C accumulation under *Miscanthus* to be approximately 1.2 Mg ha−<sup>1</sup> yr<sup>−</sup>1. Differences in soil clay content, soil bulk density, and initial low C stocks between the semi-natural *Miscanthus* grassland site and the primarily managed fields in these other studies may have led to considerable differences in soil C sequestration [55–57]. In addition, given that the managed fields were amended with fertilizer, this undoubtedly contributed to the differences in soil C sequestration rates. KAW and SONI had mean C4-C accumulation rates roughly an order of magnitude less than ASO. These colder, more northern latitude sites, with half the precipitation of ASO, likely have lower net primary production. As such, potential C inputs to the soil would be expected to be lower.

#### **5. Conclusions**

As *Miscanthus* becomes more widely planted outside its native range, particularly in low soil C agronomic fields, the potential exists for sequestration of C over the long term. Moreover, anthropogenic fire events, which are used to maintain vegetation, may further increase soil C. Toposequence along a catena influence soil C stocks in *M. sinensis* grasslands in its native range of Japan. Consideration of C sequestration in cultivated *Miscanthus* fields should include characterization of topographic variability. A majority of soil C in the grasslands examined appears to have derived from C4-C. In addition, accumulation rates for C4-C were lower than previously demonstrated.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/agriculture12010088/s1, Table S1: Calibrated radiocarbon ages to 1.2 m depth of soil carbon at three catena positions (toe slope, mid slope, and summit) in three semi-natural Miscanthus sinensis grassland sites in Japan: Kawatabi Field Science Center, Miyagi Pre-fecture (KAW), and Soni Plateau, Nara Prefecture (SONI), and Aso-Kuju National Park, Kyushu Prefecture (ASO); Table S2: Calibrated radiocarbon ages to 1.2 m depth of soil carbon at three catena positions (toe slope, mid slope, and summit) in three semi-natural Miscanthus sinensis grassland sites in Japan: Kawatabi Field Science Center, Miyagi Pre-fecture (KAW), and Soni Plateau, Nara Prefecture (SONI), and Aso-Kuju National Park, Kyushu Prefecture (ASO).

**Author Contributions:** Conceptualization, D.S.H., J.R.S., T.Y., F.G.F. and Y.T.; methodology, D.S.H., J.R.S., J.I., M.S., A.N., F.G.F. and Y.T.; software, D.S.H.; validation, D.S.H. and Y.T.; formal analysis, D.S.H., H.W. and Y.T.; investigation, D.S.H., J.I., M.S. and Y.T.; resources, J.R.S., D.L., H.W., T.Y. and F.G.F.; data curation, D.S.H., J.R.S. and Y.T.; writing—original draft preparation, D.S.H.; writing review and editing, D.S.H., J.R.S., J.I., T.Y., F.G.F., M.S. and Y.T.; visualization, D.S.H. and Y.T.; supervision, J.R.S. and Y.T.; project administration, J.R.S. and T.Y.; funding acquisition, J.R.S. and T.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was funded by the Energy Biosciences Institute at the University of Illinois through a grant from the British Petroleum Corporation.

**Acknowledgments:** We would like to thank Makoto Nakaboh, and his staff at the Kyushu Biomass Forum for providing assistance to our field research. We would also like to express our appreciation to the Aso Environmental Office and owners of the study site in Aso, Kumamoto, Japan; student workers at Kawatabi Field Science Center; the Geology Department of Osaka City University; Carolina Bueno Wandscheer, who provided invaluable assistance to field collections.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript or in the decision to publish the results.

#### **References**


**Zhiyuan Wei 1,2, Quanchao Zeng <sup>1</sup> and Wenfeng Tan 1,\***


**Abstract:** Soil microbes play critical roles in nutrient cycling, net primary production, food safety, and climate change in terrestrial ecosystems, yet their responses to cover cropping in agroforestry ecosystems remain unknown. Here, we conducted a field experiment to assess how changes in cover cropping with sown grass strips affect the fruit yields and quality, community composition, and diversity of soil microbial taxa in a mango orchard. The results showed that two-year cover cropping increased mango fruit yields and the contents of soluble solids. Cover cropping enhanced soil fungal diversity rather than soil bacterial diversity. Although cover cropping had no significant effects on soil bacterial diversity, it significantly influenced soil bacterial community compositions. These variations in the structures of soil fungal and bacterial communities were largely driven by soil nitrogen, which positively or negatively affected the relative abundance of both bacterial and fungal taxa. Cover cropping also altered fungal guilds, which enhanced the proportion of pathotrophic fungi and decreased saprotrophic fungi. The increase in fungal diversity and alterations in fungal guilds might be the main factors to consider for increasing mango fruit yields and quality. Our results indicate that cover cropping affects mango fruit yields and quality via alterations in soil fungal diversity, which bridges a critical gap in our understanding of the linkages between soil biodiversity and fruit quality in response to cover cropping in orchard ecosystems.

**Keywords:** soil microbes; cover cropping; mango orchards; sown grass; fungal diversity

#### **1. Introduction**

Cover cropping (i.e., sown grass strips) has been used as an important and effective method to improve soil fertilizer and soil carbon stacks [1]. A 12-year field experiment indicated that cover cropping increased soil organic carbon (SOC) stocks [1]. A metaanalysis showed that cover cropping contributed to the changes in global cropland soil carbon, with an overall mean change of 15.5% [2]. Compared with monospecies cover crops, cover crop mixtures sequestered more SOC [2]. Elevated SOC is also associated with improved soil health and fertility; therefore, increasing SOC may help to enhance agricultural productivity [3]. However, not all studies found that cover cropping results in SOC accumulation. Some studies demonstrated that the introduction of cover crops resulted in losses of SOC due to the faster growth of cover crops [4]. In addition to an increased carbon input, cover crops have been shown to increase biodiversity [5].

Soil harbors a rich diversity of invertebrate and microbial life, which drives biogeochemical processes at local and global scales. Soil microbes play critical roles in the nutrient cycling, climate regulation, decomposition and turnover of soil organic matter [6]. Cover cropping alters soil quality and thereby influences soil microbial communities in agro-ecosystems. The potential of cover crops to increase soil biodiversity and specific microbial patterns has been highlighted in very few studies, especially in fruit orchards. For example, sowing plant seed mixtures promoted the growth of the bacterial community

**Citation:** Wei, Z.; Zeng, Q.; Tan, W. Cover Cropping Impacts Soil Microbial Communities and Functions in Mango Orchards. *Agriculture* **2021**, *11*, 343. https:// doi.org/10.3390/agriculture11040343

Academic Editor: Yinglong Chen, Masanori Saito and Etelvino Henrique Novotny

Received: 2 March 2021 Accepted: 27 March 2021 Published: 12 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and sarophytic fungi [7]. The long-term effects of green manure amendment are altered soil microbial properties, first found in a field experiment carried out in 1956 [8]. The soil microbiome plays important roles in fruit quality and production, as rhizosphere microbiome maintains plant health and primary productivity [9]. The plant-associated microbial community has been considered as the second genome of the plant, which is critical for plant health. Therefore, the cover cropping in orchards enhances the complexity and diversity of soil microbes in intensive agriculture soils, which in turn strongly influences plant health and net primary production.

The mango (*Mangifera indica* L.) is the most important fruit crop in the tropical zones, having socio-economic significance, originating from South East Asia and cultivated worldwide [10]. It is known as the king of fruits owing to its delicious taste and high vitamin C and mineral contents [11]. Mango fruit has become the second tropical crop in terms of production and cultivated acreage [12]. To steadily increase fruit yield, farmers have increased the use of chemical fertilizers, which has caused many environmental problems, such as soil degradation and water contamination [13]. Cover cropping began to be used in orchards to resolve degraded soils and maintain higher crop yields and quality [14]. This green manure method has been widely used in the fruit ecosystem worldwide. The positive effects of the long-term application of green manure have been reported by many studies [15,16]. For example, some studies found grass cover to affect SOC [17–19]. Soil carbon for plantings of switchgrass, no-till corn, and sweetgum with cover crops between the rows increased over the first 3 years [20]. However, the variations in the structures of the microbial community and diversity in soils caused by cover crops remain unknown, which have been considered as the main drivers of multiple soil functions and plant health. Therefore, the aims of this study are to identify (1) the effects of different sown grass types in a mango orchard on the diversity of soil bacteria and fungi; (2) the changes in the main taxa in response to sown grass trips; (3) the driving factors of the variations in soil microbial diversity and community structure; and (4) the link between soil microbial diversity, mango yield and fruit quality.

#### **2. Materials and Methods**

#### *2.1. Site Description and Sampling*

Mango is among the most important fruits in Hainan Province, Southern China. The planted area of this fruit is more than 25,000 ha, and the yields are more than 560,000 kg per year. The field experiment was conducted at the mango base in Tianya District, Sanya City, Hainan Province, China (109◦24 52.70 (E) and 18◦19 54.10 (N)). It belongs to the tropical maritime monsoon climate zone, with an annual average temperature of 25.7 ◦C. The highest temperature is in June, with an average of 28.7 ◦C. The lowest temperature is in January, with an average of 21.4 ◦C. The annual duration of sunshine is 2534 h. The annual average precipitation is 1347.5 mm. The terrain is a gentle area in the lower hills, and soil type is latosol [21].

A total of four treatments with different types of cover cropping were implemented, including control (no-till + herbicide, M), planting *Stylosanthes guianensis* (Z), planting *Brachiaria eruciformis* (B) between rows, and planting *Butterfly pea* (H). Each treatment was repeated three times, with a plot area of 120 m2 (10 × 12 m) (Figure 1). All the plots were arranged in a randomized block design and separated by five 5 m buffer zones. The cover cropping was carried out in 2017. All treatments were transplanted with seedlings. Two rows of grass were planted between each mango row. The tested mango was Hongjinlong, with an age of 13 years. The row space between mango trees was 5 m. Fertilizers were applied in accordance with the conventional fertilization of fruit farmers, and no fertilizer was applied between rows.

**Figure 1.** Experimental plot and soil sampling protocol: B, *Brachiaria eruciformis*; Z, *Stylosanthes guianensis*; H, *Butterfly pea*; M, no-till + herbicide. In the control, soils were collected between the rows (**M**) and the drip line of mango trees (**MS**). Other soils (**H**, **Z** and **B**) with cover crops were collected between the rows.

Soils were collected in July (summer) 2019. In the control, soils were collected between the rows (M) and the drip line of mango trees (MS). Other soils (H, Z and B) with cover crops were collected between the rows. In each experimental plot, we collected 20 surface soil cores (0–20 cm and 20–40 cm) randomly between rows (Figure 1). All the soil samples in one experimental plot were mixed. In total, 18 samples per soil layer were collected. After removing the roots, stones and litters by hand, we sieved them through a 2 mm sieve and separated them into two portions. One portion (approximately 5 g) was preserved in a 10 mL centrifuge tube (sterilized) and stored at −80 ◦C. The second portion was air-dried and used to measure soil properties.

#### *2.2. Mango Fruit Quality and Yield*

When mango fruits were nearly matured (80%), all the mango fruits in each experimental plot were harvested, and the masses were measured as yields. We randomly chose 20 mango fruits in each experimental plot and placed them into plastic bags for analysis of the fruit quality. Total soluble solids content of mango fruit was measured by a digital refractometer and presented in percent Brix (%TSS). Organic acid concentration was measured by an automatic titrator [22]. Vitamin C (Vc) content was determined via the method of 2,6-dichlorophenol indophenol (Shao et al. 2013).

#### *2.3. Analysis of Soil Properties*

Soil properties were measured by the standardized methods as described previously. Soil bulk density (BD) was measured by the volume–mass relationship via a cutting ring [23]. Soil pH was determined by a glass pH meter in a 1:2.5 soil–water suspension. Soil organic matter (SOM) was digested with 5 mL concentrated H2SO4 and 5 mL 0.8 M K2Cr2O7, and then determined by 0.2 M Ferrous Ammonium Sulfate [24]. Soil total nitrogen (TN) was determined as described previously [25]. Soil available nitrogen (AVN) was extracted by 1 mol/L KCl and then measured by a Seal Auto Analyzer3 [26]. Soil total phosphorus (TP) and soil available P (AVP) were determined by molybdenum, antimony and scandium colorimetry. Total potassium (TK) and available potassium (AVK) in soil were determined by a flame photometer.

#### *2.4. Molecular Analysis*

We used the PowerSoil kit to extract DNA from 0.5 g soils. The protocol was listed in the manufacturer's instructions. After extraction, the ratios of A260/A230 and A260/A280 were determined to assess the quality of extracted DNA.

The soil bacterial community was determined by sequencing hypervariable V3-V4 regions of 16S rRNA genes using primers 338F (ACTCCTACGGGAGGCAGCA) and 806R (GGACTACHVGGGTWTCTAAT) [27]. The PCR amplification conditions for 16S rRNA were 50 s at 94 ◦C, 30 s at 40 ◦C, 35 cycles of 60 s at 72 ◦C, followed by 5 min at 72 ◦C [28]. The ITS1 variable region was sequenced with primer sets ITS3 (5 -GCATCGATGAAGAACG-CAGC-3 ) and ITS4 (5 -TCCTCCGCTTATTGATATGC-3 ) [29,30] to assess the soil fungal community. As for the ITS1 variable region, the PCR program was 5 min at 94 ◦C, 32 cycles of 30 s at 94 ◦C, 30 s at 54 ◦C, and 1 min at 72 ◦C [31]. Sequencing was conducted on an Illumina MiSeq PE300 platform.

After sequencing, bioinformatics processing was performed using QIIME, USEARCH and UNIOISE3 [32]. After removing the short sequences (<20 nucleotides) and chimeric sequences, the remaining sequences were clustered into operational taxonomic units (OTUs) using 97% similarity [33] (version 7.1 http://drive5.com/uparse/). Taxonomy was assigned against the Greengenes (16S gene, Release 13.5 http://greengenes.secondgenome. com/) and UNITE (ITS gene, Release 7.2 http://unite.ut.ee/index.php) databases. The Shannon diversity index and richness (the numbers of OTUs) were calculated to express the soil bacterial and fungal diversity.

#### *2.5. Fungal Ecological Guilds Identification*

We used FUNGuild to identify the fungal ecological guilds using fungal OTU dataset with taxon assignments [34]. This prediction method has been widely used in gaining insights into the distributions of soil fungal ecological groups [35,36]. In this study, we used the data with confidence levels of "highly probable" and "probable" to perform further analysis.

#### *2.6. Statistical Analysis*

We first used ANOVA to compare the differences of soil properties, mango fruit yield and quality, and bacterial and fungal diversity among different treatments on SPSS 20.0 (IBM Corporation, Armonk, NY, USA). We then conducted Pearson correlation analyses between the microbial diversity (Shannon diversity index) and soil properties. Prior to ANOVA and Pearson correlation analyses, the data were used to conduct a log transformation to meet the normality and homogeneity. Multiple regression models were constructed to compare the effects of soil properties on microbial diversity in R 3.5. Anosim analysis was performed to compare the effects of sown grass strips on the soil microbial community structure in R 3.5. The bacterial and fungal community structure was calculated based on the Bray–Curtis dissimilarity and visualized by a nonmetric multidimensional scaling (NMDS) plot in R 3.5 with the vegan package [37]. The effects of soil properties on the bacterial and fungal community structure were analyzed by the Mantel test in R 3.5. The associations between the soil microbial community structure and soil properties were determined by the Mantel test in R 3.5 [38].

#### **3. Results**

#### *3.1. Soil Properties of the Tested Orchard*

The impacts of cover cropping with sown grass strips on soil properties depended on the types of sown grass and soil layers. The upper soils (0–20 cm) were more sensitive to the application of sown grass strips. B20 had the highest soil AVN, which was significantly higher than other sown grass types. The application of sown grass strips had no significant effects on soil pH, SOM and AVK at both soil layers (Table 1).


**Table 1.** The characteristics of soil properties of mango soils under different sown grass strips.

BD, soil bulk density; TK, total potassium; TN, total nitrogen; TP, total phosphorus; SOM, soil organic matter; AVP, available phosphorus; AVK, available potassium; AVN, available nitrogen.

#### *3.2. Yield and Fruit Quality of Mango under Different Sown Grass Trips*

The application of sown grass trips had significant effects on mango fruit yield, with the highest yield in B (Table 2). There were no significant differences observed between Z and M. The applications of B, H and Z significantly enhanced TSS content and decreased organic acid compared to M.

**Table 2.** The yield and fruit quality of mango under different sown grass trips.


TSS, total soluble solids content; Vc, vitamin C. Different letters indicate significant differences under different sown grass trips. B, *Brachiaria eruciformis*; Z, *Stylosanthes guianensis*; H, *Butterfly pea*; M, no-till + herbicide. Different letter indicates significant differences between different sown grass trips.

#### *3.3. Variations in Soil Microbial Diversity under Different Sown Grass Strips*

The observed bacterial Shannon diversity index ranged from 5.47 to 6.64, while phylotype richness (OTUs) varied from 1886 to 3509. Sown grass strips have no significant effects on soil bacterial α-diversity indices (Shannon diversity index and richness index). The soil fungal phylotype richness ranged from 419 to 1080, and the fungal Shannon diversity index ranged from 3.81 to 4.95. Among these soil properties, TN and AVN had significant associations with the bacterial Shannon diversity index (Figure 2). A multiple regression models indicated that soil TN and AVN were the best predictors of the soil bacterial Shannon diversity index (with a relative importance of 0.54 and 0.42), followed by AVP, with a relative importance of 0.04. Soil properties had no significant correlations with the fungal diversity index.

#### *3.4. Variations in Soil Microbial Community under Different Sown Grass Strips*

Across all soils, a total of 1,506,304 quality bacterial sequences, and an average of 50,210 sequences per sample, were obtained. Four of the 28 phyla detected were dominant, including Acidobacteria (15.51%), Actinobacteria (17.55%), Proteobacteria (24.26%) and Chloroflexi (24.77%) (with an average relative abundance of >5%, *n* = 30), accounting for more than 82% of the bacterial sequences (Figure 3A). The Proteobacteria taxa were dominated by Alphaproteobacteria (17.66%), followed by Deltaproteobacteria (4.12%) and Gammaproteobacteria (2.49%). The soil layer had no significant effects on the relative abundance of Acidobacteria and Proteobacteria. The upper soils (20.5%) had a higher relative abundance of Actinobacteria than that of the lower soils (14.6%).

**Figure 2.** The associations between soil bacterial diversity and soil properties. TK, total potassium; TN, total nitrogen; TP, total phosphorus; AVP, available phosphorus; AVK, available potassium; AVN, available nitrogen. B, *Brachiaria eruciformis*; Z, *Stylosanthes guianensis*; H, *Butterfly pea*; M and MS, no-till + herbicide.

**Figure 3.** The community compositions of soil bacteria (**A**) and fungi (**B**) at phylum level. B, *Brachiaria eruciformis*; Z, *Stylosanthes guianensis*; H, *Butterfly pea*; M and MS, no-till + herbicide.

Across all mango soils, a total of 1,777,037 quality fungal sequences, and an average of 59,234 sequences per sample, were obtained. The dominant fungal phyla in soils were Ascomycota and Basidiomycota, with average relative abundances of 79.38 and 11.28%, respectively (Figure 3B). Other minor phyla (Anthophyta, Cercozoa, Rozellomycota and Glomeromycota) were also found at a lower relative abundance (relative abundance < 1%). The soil layer and sown grass trip had no significant effects on the relative abundance of Ascomycota and Basidiomycota. Based on taxonomical classification at the class level, Sordariomycetes (36.69%), Eurotiomycetes (18.96%), Dothideomycetes (15.33%) and Agaricomycetes (9.50%) were more abundant than other groups (relative abundance > 1%), which accounted for 80.48% of the fungal sequences. Other fungal classes were less abundant in all the soils.

Anosim analysis indicated that the fungal community structure was strongly impacted by the application of sown grass strips (r2 = 0.825, *p* = 0.001 for 0–20 cm soil layer; r2 = 0.413, *p* = 0.001 for 20–40 cm soil layer). Soil bacterial community structure was not sensitive to sown grass strips (r<sup>2</sup> = 0.179, *p* = 0.095 for 0–20 cm soil layer; r2 = 0.092, *p* = 0.203 for 20–40 cm soil layer). Different soil samples were clearly separated by the sown grass types in the NMDS plot (Figure 4A,B). The Mantel test showed that SOM (r = 0.16, *p* = 0.036), TN (r = 0.45, *p* = 0.001), TP (r = 0.42, *p* = 0.001) and AVN (r = 0.23, *p* = 0.004) had significant effects on the bacterial community structure (Figure 4C). For the fungal community structure, soil BD (r = 0.17, *p* = 0.033), TN (r = 0.17, *p* = 0.016), TK (r = 0.21, *p* = 0.002) and AVN (r = 0.15, *p* = 0.032) were the main factors.

**Figure 4.** The community structure of soil bacteria (**A**) and fungi (**B**), and the associations between community structure and soil properties (**C**). B, *Brachiaria eruciformis*; Z, *Stylosanthes guianensis*; H, *Butterfly pea*; M and MS, no-till + herbicide. In the control (M and MS), soils were collected between the rows (M) and the drip line of mango trees (MS).

#### *3.5. Variations in Soil Fungal Ecological Guilds under Different Sown Grass Strips*

Soil functional fungal groups significantly altered under different cover cropping with different sown grass strips. Cover cropping enhanced the proportion of pathotroph fungi and decreased saprotroph fungi. Cover cropping increased the proportion of Arbuscular Mycorrhizal fungi and wood saprotroph fungi, while cover cropping had no significant effects on the plant pathogen or animal pathogen (Figure 5).

**Figure 5.** The functional groups of soil fungi under different cover cropping treatments. B, *Brachiaria eruciformis*; Z, *Stylosanthes guianensis*; H, *Butterfly pea*; M, no-till + herbicide. In the control, soils were collected between the rows (M) and the drip line of mango trees (MS).

#### *3.6. The Associations between Soil Microbial Community and Diversity and Mango Fruit Yields and Quality*

The correlation between fungal diversity (richness) and mango fruit yields was significant (r = 0.75, *p* < 0.01) (Figure 6). The soil fungal community structure (repressed by NMDS1) was significantly correlated with mango fruit yields (r = −0.79, *p* < 0.01). These also showed significant correlations between fungal diversity and mango fruit TSS (r = 0.71, *p* = 0.01) and organic acid (r = −0.76, *p* < 0.01). However, the soil bacterial diversity had no significant associations with mango fruit yields and TSS (data not shown).

**Figure 6.** The associations between fungal community characteristics and mango fruit yield and quality under different cover cropping treatments. TSS, total soluble solids content; Vc, vitamin C; OC, organic acid. NMDS1 represented the community structure of soil fungi, and Richness indicated fungal diversity.

#### **4. Discussion**

Cover cropping has been widely used in agriculture systems to improve crop yield and quality as well as soil quality. In our study, the mango yield was significantly enhanced by 3–14% after the application of sown grass strips between rows. Similarly, winter cover cropping improved corn yields [39], and soybean yield significantly increased after the 3-year application with a multispecies mixture of legumes, grasses and Brassica spp [40]. These positive effects of cover cropping in crop yields suggested that cover cropping is an effective method to increase yields. In this study, we also found a significant increase in mango yields with B and H. However, Z had a slight increase in mango yields, suggesting that Z is not suitable to use in mango orchards. The positive effects of B and H on mango yields might be explained by the variations in nutrients and microbial community diversity in soils caused by cover cropping.

A meta-analysis found that global cropland soil carbon changes due to cover cropping, which increases SOC in near-surface soils by an average of 15.5% [2]. However, in our study, we found the sown grass did not enhance SOC in both soil layers. This difference might be explained by the duration of application of the sown grass. Sown grass strips caused the changes in available nutrients, such as AVN and AVP. The growth of sown grass could produce root exudates and litters, which were the main resources of soil nutrients, especially available nutrients. These variations in soil nutrients directly and indirectly influenced soil microbial diversity and community compositions, respectively.

In this study, we found that soil fungal diversity was sensitive to the application of sown grass strips. The introduction of sown grass between the rows enhanced the fungal diversity compared to the control (no sown grass). This might be explained by the higher decomposition ability of litters and roots than bacteria. Fungi are considered the primary decomposers of dead plant biomass in terrestrial ecosystems. The results from the litter-bag decomposition experiments in the field and the laboratory indicated the overwhelming advantage of fungi during the litter decomposition process [41–44]. Pascoal and Cássio (2004) showed that the contribution of fungi to litter decomposition greatly exceeded that of bacteria [45]. Despite the critic roles of fungi, the roles of bacteria could not be neglected [46], especially during the litter decomposition in which they mainly worked at the later decomposition stage. Therefore, the long-term introduction of sown grass strips might cause variations in soil bacterial diversity and community.

Different types of sown grasses also had different effects on soil fungal diversity. B and H soils had higher soil fungal diversity than Z. These variations might be explained by the quality of plant litter, roots and root exudates, which were the main factors impacting soil fungal diversity and community. For example, high quality litter decomposed faster than the low-quality litter [47,48]. The fast decomposition released a much higher amount of nutrients to the soil and resulted in the fast succession of soil fungi.

Nitrogen was considered as the main resource of soil microbes and strongly affected soil biodiversity and community structure [26,49]. Wang et al. (2018) found that tropical forest soil microbial community composition was shaped by N addition, with the increase in the proportion of arbuscular mycorrhizal fungi [50]. In the present study, we found that soil TN and AVN had significant effects on soil bacterial diversity, suggesting that soil N was the best predictor of the regulation of soil bacterial diversity in the mango orchards.

Soil microorganisms are considered as the main regulator of soil nutrient cycles and net primary production. In this study, we found that soil microbes enhanced mango fruit yields and quality via increasing fungal diversity and alterations in fungal community structure. Soil microbial diversity may improve crop yields through these mechanisms. First, soil biodiversity mediated nutrients available in the soil, which were the main resources of crop plant growth. Many previous studies have suggested that soil biodiversity enhances plant growth and drives crop yields [51,52]. Second, soil microorganisms help to maintain soil health and prevent the invasion of pathogens. Higher soil biodiversity provides higher ecosystem functions, such as net primary production and nutrient cycling. Therefore, soil microbial diversity, especially for fungal diversity, is vital in order to improve mango fruit yields.

#### **5. Conclusions**

This field experiment showed that cover cropping with sown grass had significant effects on the soil microbial community in a mango orchard. Fungal diversity was more sensitive than bacterial diversity in response to sown grass. The application of *Brachiaria eruciformis* (B) had the strongest effects on soil fungal diversity. The increase in fungal diversity suggested that sown grass had positive influences on soil biodiversity. Across soil properties, AVN and TN were the most important predictors affecting soil bacterial communities, while soil nutrients (TN and AVN) were the most important factors in mediating soil fungal communities. These results show that the sown grass strips in a mango orchard regulated the soil microbial community and diversity via soil organic matter and nitrogen, which might be of great significance in mango production and quality and the suitability of mango orchards. Mixtures of sown grasses might be more effective for soil biodiversity and soil function, and operations in mango orchards and future research should be focused on this aspect.

**Author Contributions:** Conceptualization, W.T. and Z.W.; methodology, Z.W.; data analysis, Q.Z. and Z.W.; investigation, Z.W.; writing—original draft preparation, Q.Z. and Z.W.; writing—review and editing, W.T.; visualization, W.T.; supervision, Z.W.; project administration, Z.W.; funding acquisition, W.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the National Natural Science Foundation of China (Nos. 41977023 and 32061123007).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data are contained within the article.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland www.mdpi.com

*Agriculture* Editorial Office E-mail: agriculture@mdpi.com www.mdpi.com/journal/agriculture

Disclaimer/Publisher's Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Academic Open Access Publishing

mdpi.com ISBN 978-3-0365-9725-6