Variation of soil organic carbon and physical properties in relation to land uses in the Yellow River Delta, China

Soil physical properties and soil organic carbon (SOC) are considered as important factors of soil quality. Arable land, grassland, and forest land coexist in the saline-alkali reclamation area of the Yellow River Delta (YRD), China. Such different land uses strongly influence the services of ecosystem to induce soil degradation and carbon loss. The objective of this study is to evaluate the variation of soil texture, aggregates stability, and soil carbon affected by land uses. For each land use unit, we collected soil samples from five replicated plots from “S” shape soil profiles to the depth of 50 cm (0–5, 5–10, 10–20, 20–30, and 30–50 cm). The results showed that the grassland had the lowest overall sand content of 39.98–59.34% in the top 50 cm soil profile. The content of soil aggregates > 0.25 mm (R0.25), mean weight diameter and geometric mean diameter were significantly higher in grassland than those of the arable and forest land. R0.25, aggregate stability in arable land in the top 30 cm were higher than that of forest land, but lower in the soil profile below 20 cm, likely due to different root distribution and agricultural practices. The carbon management index (CMI) was considered as the most effective indicator of soil quality. The overall SOC content and CMI in arable land were almost the lowest among three land use types. In combination with SOC, CMI and soil physical properties, we argued that alfalfa grassland had the advantage to promote soil quality compared with arable land and forest land. This result shed light on the variations of soil properties influenced by land uses and the importance to conduct proper land use for the long-term sustainability of the saline-alkali reclamation region.

Soil physical properties and soil organic carbon (SOC) are considered as important factors of soil quality. Arable land, grassland, and forest land coexist in the saline-alkali reclamation area of the Yellow River Delta (YRD), China. Such different land uses strongly influence the services of ecosystem to induce soil degradation and carbon loss. The objective of this study is to evaluate the variation of soil texture, aggregates stability, and soil carbon affected by land uses. For each land use unit, we collected soil samples from five replicated plots from "S" shape soil profiles to the depth of 50 cm (0-5, 5-10, 10-20, 20-30, and 30-50 cm). The results showed that the grassland had the lowest overall sand content of 39.98-59.34% in the top 50 cm soil profile. The content of soil aggregates > 0.25 mm (R 0.25 ), mean weight diameter and geometric mean diameter were significantly higher in grassland than those of the arable and forest land. R 0.25, aggregate stability in arable land in the top 30 cm were higher than that of forest land, but lower in the soil profile below 20 cm, likely due to different root distribution and agricultural practices. The carbon management index (CMI) was considered as the most effective indicator of soil quality. The overall SOC content and CMI in arable land were almost the lowest among three land use types. In combination with SOC, CMI and soil physical properties, we argued that alfalfa grassland had the advantage to promote soil quality compared with arable land and forest land. This result shed light on the variations of soil properties influenced by land uses and the importance to conduct proper land use for the long-term sustainability of the saline-alkali reclamation region.
Land uses strongly influence the processes and capacity of ecosystem, causing the change of ecosystem functions and services 1,2 . Understanding ecological consequence of land use conversion is critical for maintaining ecosystem services and conserving biodiversity 3,4 . Conversion of land use from natural ecosystems to agricultural ecosystems may lead to the degradation of most of the ecosystem services, which may pose a direct threat to regional and global substantial environmental developments 5 . One important property of ecosystems that is likely to change with land uses is soil carbon stock, which is linked to carbon dioxide (CO 2 ) concentration in the atmosphere 6 and may have a significant feedback to the global carbon cycle, as the amount of carbon stored in soil is approximately twice more than that in the atmosphere 7,8 . Soil organic carbon (SOC) is the main component of soil carbon stock and plays a critical role in maintaining the services of ecosystems 9 , such as food production, water quality provision, soil fertilization, climate change abatement etc 10,11 . Land use changes have been largely expanded over last several decades due to the increase of population and food demand 12,13 , which may have caused increased CO 2 emission and intensified greenhouse effect 14 . Hence, variations of SOC as a result of land use changes have caught much attention worldwide as a critically important issue for agricultural management, ecosystem restoration and environmental conservation 5,9 . Soil physical factors are known to affect SOC. Both aggregate stability and soil particle size are among the most important soil physical properties 15 . Differences in human activities and vegetation types had a strong effect on the changes in soil biological properties 16,17 , and biomass and litter types also affected soil organic matter (SOM) 18 . The SOM is an effective indicator of soil resource condition that reflects functional traits such as soil aggregate stability, water holding capacity, and microbial activity 19 and had close relationship with aggregate Table 1. Soil water content (SWC), soil bulk density (BD) and total soil porosity (P t ) among different land uses (mean ± SD). Different letters at the same soil depth indicate significant difference (n = 5, p < 0.05) among different land uses by one-way ANOVA. www.nature.com/scientificreports/ Soil particle size among different land uses varied significantly. Soil texture was mainly composed of silt and sand, and the clay only occupied for less than 12% among three land uses in the YRD, and sand was dominant soil particle in arable land and forest land, accounting for nearly 80%. Silt and clay contents were generally the highest in grassland among different land uses in each layer of the top 50 cm soil profile (Fig. 1).
Land uses had strong effect on soil aggregates. The aggregate content at R 0.25 had significant difference among different land uses (Table 2), and this aggregate content was generally the highest in grassland among different land uses in the entire top 50 cm soil profile. Stability of soil aggregates, as measured by mean weight diameter (MWD) and geometric mean diameter (GMD), followed the order of grassland > arable land > forest land (Table 3).

Soil organic carbon fractions for different land uses.
The concentration of SOC in forest land was significantly higher than arable land and grassland in the top 5 cm soil layer (Fig. 2a). The overall SOC stock showed similar patterns with SOC concentration among different land uses, except for the grassland in 10-30 cm soil profile, which was significant higher than arable land and forest land (Fig. 2b).
The concentration of soil labile carbon (LOC) in arable land was significant lower than both grassland and forest land in the top 50 cm soil profile (Fig. 3a). The C LOC to C ORG ratio followed the order of forest land > grassland > arable land (Fig. 3b).
The concentration of soil microbial biomass carbon (MBC) decreased with the increase of soil depth among different land uses and followed the order of forest land > grassland > arable land (Fig. 4a). The C MBC to C ORG Figure 1. Soil particle size among different land uses at 0-50 cm depth. Different letters with the same particle size at the same soil depth indicate significant difference (n = 5, p < 0.05) among different land uses by one-way ANOVA. Table 2. Effect of land uses on soil dry-stable aggregates (%) at different depth (mean ± SD). Different letters with the same soil depth indicate significant difference (n = 5, p < 0.05) between different land uses by one-way ANOVA. R 0.25 is aggregates of diameter > 0.25 mm.
We calculated the carbon management index (CMI) for the soil of forest land and used it as the reference soil. The relationship between NL and CPI and between L and LI had the same patterns of variation ( Table 4). The CMI showed significant difference among different land uses, and changed according to the patterns of the LOC concentration. The CMI in arable land in the top 50 cm soil profile was the lowest, and decreased with the increase of soil depth. The CMI in grassland was more than 100 in the depths of 10-20 cm and 20-30 cm, suggesting that the CMI of grassland was higher than forest land (reference soil).

Relationship between soil carbon and physical properties.
Pearson's correlation revealed that the CPI was significantly (P < 0.01) and positively correlated with the SWC, R 0.25 , MWD, GMD, BD and silt content ( Table 5). The C MBC to C ORG ratio was strongly positively correlated with sand content, but negatively correlated with the contents of clay, silt and R 0.25 , MWD and GMD. A similar, negative correlation was also found between MBC and R 0.25 , MWD and GMD. Finally, we found that SWC was negatively correlated with LOC, MBC, C MBC to C ORG ratio and LI.  Table 4. Soil carbon management index (CMI) with different land uses (mean ± SD). Different letters with the same soil depth indicate significant difference (n = 5, p < 0.05) between different land uses by one-way ANOVA. NL non-labile carbon concentration (g kg −1 ), L Carbon pool lability, LI lability index, CPI Carbon pool index.

Discussion
Soil physical properties are critical to soil quality in aspects of root growth, infiltration, water and nutrient holding capacity 32 . Land uses and vegetation types can significantly influence soil physical properties, particularly soil aggregates distribution 33,34 . In this study, the variation of SWC, BD and porosity occurred among three land uses, the higher SWC observed in arable land was likely related to the fact that farmland undergone the artificial irrigation. The lower SWC in forest land than that in alfalfa grassland was probably related to the lower surface cover and the characteristic of roots in forest land, which both lead to greater transpiration 35,36 . This result agreed with the study of Zhang et al. (2016) that SWC at the grass stage was a significantly higher than that at the forest stage 32 . The high soil BD of the arable land was likely the result of combined influence of the ploughing in tillage layer, roots distribution and decreased SOC and soil aggregation, as a result of repeated events of sowing and harvesting 20,37 . Soil particles and soil aggregate are the important physical properties for the process of soil physiochemical and biological properties, and soil particle size distribution is the fundamental physical factor affecting aggregate stability 38 . Results of our study indicated that sand is the primary soil particle among three land uses in the YRD, and the lowest sand content of 39.98-59.34% occurred in the grassland in the top 50 cm soil profile (Fig. 1). This result may be explained by different effects of soil erosion control under different land uses 39 . The vegetation types, coverage, and root system condition among different land uses are correlated to soil particle composition; higher root growth and litter input can improve soil physiochemical and biological properties, accelerate the formation of humus, reduce the surface wind erosion and facilitate the fixation of fine sand particles. In the grassland, high vegetation coverage and root activity can prevent soil erosion from rain splash therefore the loss of fine soil grains, because root activity of plant could greatly affect the distribution of soil particle size in newly formed wetlands in the YRD 34 . The soil aggregates were mainly non water-stable aggregates, and the number of water-stable aggregates was very small (Tables S1), which may be related with the special saline-alkali environment and local soil texture due to its new and fast formation of alluvial plain. To a certain extent, the situation of water-stable aggregates affects soil aeration and erosion resistance, and its small portion indicates the poor soil fertility and stability in this region. Averaged across three land uses, the contents of silt, clay and soil dry-stable aggregates (R 0.25 ) in grassland was highest among all three land uses studies, which is in agree with a previous study by Liu et al. 40 . It should be noted that tillage and harvesting practices in arable land and low vegetation coverage in forest land may promote soil erosion and cause the loss of silt and clay contents and the decrease of soil aggregate stability in topsoil 41,42 . The alfalfa plants generally have a well-developed root system, which produces more organic matter as the roots decompose. High content of soil organic matter will produce more soil aggregates and improve the soil structure 20,43 .
Soil acts as either a carbon source or a carbon sink, and land uses can change the function of source and sink 32,44 . For the top soil (0-5 cm), our study found that SOC content and stock in forest land were the highest, which was resulted from the input of litter on the surface soil. While the arable land had less litter, frequent disturbance and strong soil respiration in the surface, which accelerated the consumption of SOC in the top soil 45 . But at the deeper section of the soil profile (5-50 cm), SOC content and stock in arable land and grassland were higher than that in forest land due to continuous root production and decomposition. This result was probably related to the rich root system of arable land and grassland concentrated in the deeper soil profile 18,46 , while lower root production and poor soil permeability in forest land, because different vegetation type can regulate the distribution of SOC through plant growth and root distribution, and the lack of oxygen soil has a fundamental restriction on microbial decomposition 47 . Some studies confirmed that vegetation restoration and belowground biomass had a close relationship with SOC 48 , and played a critical role in improving SOC stock in a degraded salt land 47 . Chen et al. confirmed that soil carbon accumulation was strongly driven by the establishment of vegetation 49 . As a salt-tolerate plant, alfalfa has high biomass and root activity, which is the main reason why artificial alfalfa grassland has high carbon content and stock compared to arable land and forest land in the saline-alkali reclamation region. Xiao et al. 50 showed that conversion of natural system to other land uses decreased MBC, and the content of the LOC in TOC indicates soil quality 51,52 . As CMI is a good indicator of soil carbon quality 53 , we argued that alfalfa grassland, which has the highest CMI values, seems to provide better options for soil carbon management and soil quality. Table 5. Pearson's correlation between the characteristics of soil carbon and soil physical properties. *Correlation significant at the 0.05 level (two-tailed). **Correlation significant at the 0.01 level (two-tailed). www.nature.com/scientificreports/ Soil organic carbon, particularly active component of organic carbon, has been reported to act as important binding agents for soil aggregates and their stability 54,55 , and this assertation was demonstrated by the strong correlation between CPI and soil physical properties in our study. Moreover, we found that CPI were particularly sensitive to soil water content (SWC) and aggregate stability. According to Zhao et al., higher SWC could reduce the impact of soil salinity on soil carbon stock due to the variations of salt concentrations and O 2 diffusion of soil layers, because high salinity could influence solubility of SOM, inhibit microbial processes and the final soil carbon stock 47 , which could explain higher carbon content and stock in arable land and grassland than those in forest land in the deeper soil profile. Increased SOC could improve aggregate stability indirectly through increasing energy and nutrient availability for soil microbes 53 . Moreover, some studies found that the aggregate stability was related to SOM composition and had good correlations between carbohydrate content and soil aggregate stability 40 . Specifically, the occurrence of SOC and aggregate stability in arable land and grassland were higher than that in forestland in this region studies, but lower active component of organic carbon in arable land which resulted in lower carbon management. In brief, land use types had changed the vegetation types with different disturbance intensities, litter and roots inputs. The soil physical properties changed through biotic process regulated by plants and soil microbial communities; the inputs and accumulation of organic matter resulted from complex interactions between biotic processes and abiotic processes driven by anthropogenic disturbance and environmental factors (Fig. 5). Under different vegetation types, the soil texture and soil aggregate stability was improved; SOC and active carbon increased, followed by the increase of CMI in grassland. Therefore, combining soil physical properties and soil carbon index with land uses in comprehensive consideration, we believed that alfalfa grassland is the best land use type to improve soil physical properties and the quality of soil carbon in the YRD, which experienced frequent secondary salinization in the past decades. As a result, the proper land use, or the conversion of forest and grassland into arable land should be of concern in the context of soil quality and environmental degradation. In view of the arable land, to conduct proper agricultural practices, such as less impact of land tillage practices, may be a better option to improve soil quality for the long-term production and sustainability of the saline-alkali reclamation region.

Conclusions
Soil physical properties play an important role in the formation and transformation of soil carbon during the process of land use and land cover change. The study showed strong changes in soil physical properties and soil carbon among the arable land, grassland and forest land and these changes were not uniform along the soil profile to the depth of 50 cm. Overall, we found that alfalfa grassland had effectively improved soil physical properties and soil carbon, and the soil layer of 20-30 cm may be the turning point for soil physical properties change between the arable land and forest land. Land management practices, such as plowing and harvesting, had strong impact on the SOC, suggested by higher SOC in the arable land compared to forest land except 0-5 cm soil layer. Additionally, CMI of arable land was the lowest relative to grassland and forest land. Therefore, the long-term conventional cultivation of arable land is not favorable to soil carbon management and soil quality improvement, and more attention should be paid to improve the soil quality and ecosystem sustainability in the saline-alkali reclamation region.

Materials and methods
Study site. The study sites are located in the Hekou district of Dongying city, in the Yellow River Delta (YRD), Shandong Province, China (37°54′10.19″ N, 118°31′13.83″ E Fig. 6). It is a typical alluvial plain of the Yellow River and belongs to the semi-humid monsoon climate zone with warm temperate. Mean annual precipitation is about 692 mm occurring mainly during June, July and August, and mean annual temperature is 13.2 °C with seasonal variation. The soils are mainly of Calcaric Fluvisols (moisture soil) and Gleyic Solonchaks (coastal saline moisture soil) according to FAO 56 . The mixed forest is dominated by a Robinia pseudoacacia L. and an adjacent crop land situated side by side for the study. The cultivation vegetation after reclamation consists of wheat-maize and purple alfalfa predominantly, the arable land and artificial grassland sometimes were converted to each other due to the serious secondary salinization.
Experimental design. Three typical land use units (annual arable land, artificial grassland and artificial forestland) were selected according to the main land use types in the study area, and were investigated in detail for the history and current situation of land cultivation. The annual arable land was ploughed, fertilized, irrigated and planted with crop every year since reclamation of the 1950s, and rotationally planted with winter wheat (Triticum aestivum L.) and summer maize (Zea may L.) under conventional tillage for more than ten years. Winter wheat was sown in early October and harvested in early June next year, and summer maize was sown in mid-June and harvested in late September. The artificial grassland was transformed from the previous arable land in 2011, and then continuously planted with purple alfalfa (Medicago sativa L.) for five years. The alfalfa was harvested for four times each year as a source of livestock fodder. For the alfalfa field, base fertilizer was applied at sowing and no other fertilizer was applied during five-year of growth. The forestland is the result of artificial afforestation occurred in the 1960s on the saline-alkali land dominated with Robinia pseudoacacia Field work. A field survey was conducted in late September in 2016 (after the summer maize was harvested). In this study, we determined the sampling areas according to the size of the communities 30 , selecting five 2 m × 2 m plots from representative terrain in the herbaceous communities of the arable land and artificial grassland, five 5 m × 5 m plots in the forest land. The sampling plots were distributed according to a "S" shape in each land use unit, and 100 m apart between sampling plots. At each sampling plot, three soil sampling sites were located at the two diagonal corners and the center of the plot. Soils at each sampling site were collected from soil profiles to 50 cm depth (0-5, 5-10, 10-20, 20-30, and 30-50 cm) by using a drill. A composite soil sample at certain soil depth was obtained by mixed all these samples at each plot. The composite soil samples for soil water content (SWC) were stored in sealed aluminum cases to prevent potential moisture loss. Soil samples for SOC fractionation, particle sizes were stored in zip-top plastic bags. The undisturbed soil samples for aggregate analysis were wrapped up with paper to avoid destroying the aggregates. Soil bulk density (BD) was measured at the intermediate position of each soil layer using a cutting ring with inner diameter of 5.0 cm, and volume of 100 cm 3 . Overall, 75 composite soil samples were collected, representing three land uses, five depths and five replicates. The fresh soil samples were immediately taken back to the lab for the analysis preparation.
Laboratory analysis. In the laboratory, the moist soil samples were crushed to pass through 2 mm sieve, and removed the roots and other debris by tweezers. The sieved soil samples were divided into two sub-samples for air-dried and stored at low temperature, respectively. A part of air-dried samples was sieved through 0.18 mm screen to measure the soil SOC and labile organic carbon (LOC where P t is the total soil porosity, ρ b refers to the soil bulk density, and ρ p refers to the soil density (2.65 g cm −3 ). The pipette method was used to measure the soil particle size with Na hexametaphosphate after soil organic matter oxidation with H 2 O 2 by a Laser Grain-size Analyzer (Mastersizer 3000, Malvern Instruments Inc., Worcestershire, UK) with international classification 58 , then calculated the proportions of the clay (< 0.002 mm), silt (0.002-0.02 mm), and sand (> 0.02 mm) contents.
The soil aggregates were measured by the dry-sieving method and wet-sieving method using soil aggregate analyzer (TTF-100, Shunlong experimental instrument factory, Shangyu city, China). A set of five stacking sieves with openings of 5, 2, 1, 0.5 and 0.25 mm were selected to determine the dry-stable aggregates and wet-stable aggregates with air-dried soil samples of 100 g and 50 g with three replicates respectively. The aggregates were divided into aggregates sized > 5 mm, 2-5 mm, 1-2 mm, 0.5-1 mm, 0.25-0.5 mm. The contents of > 0.25 mm mechanically stable aggregates (R 0.25 ) were calculated using Eq. (2), for which the > 0.25 mm fraction was the most susceptible to changes in land use or management 40,59 , and soil aggregate fractions obtained by the dry-sieving method have been successfully used to analyze SOC pool 60 . The soil structural stability was characterized using the mean weight diameter (MWD), geometric mean diameter (GMD) of soil aggregates according to Eqs. (3) and (4) 61,62 . The > 0.25 mm percentage of aggregate disruption (PAD 0.25 ) was calculated using Eq. (5).
where R 0.25 is the content of soil aggregates > 0.25 mm, Mr >0.25 is weight of aggregate > 0.25 mm, M T is the total weight of soil tested. X i is the mean diameter of each size classes (< 0.25 mm, 0.25-0.5 mm, 0.5-1 mm, 1-2 mm, 2-5 mm and > 5 mm), and W i is the weight fraction of aggregates in size class i, and n is the number of size fractions. D 0.25 is the > 0.25 mm dry-sieved aggregate content, and W 0.25 is the > 0.25-mm water-stable aggregate content.
The total SOC concentrations were determined following the dry combustion method 63 using a CHN analyzer. The MBC was measured by chloroform-fumigation extraction method 51,64 . LOC was determined by using 333 mmol L −1 KMnO 4 Oxidation Method, and measured by the spectrophotometric of 565 nm wave length 65  www.nature.com/scientificreports/ total SOC was considered as equal to the total soil carbon because the measured inorganic carbon (carbonates) contents of the samples were almost nil 66 . Carbon Management Index (CMI) was calculated using the procedure outlined below 65,66 , using the no-reclamation forest soil as reference sample: Statistical analysis. One-way ANOVA was carried out using the SPSS software, ver. 16.0 (IBM, USA) to analyze the differences of soil physical properties and soil carbon among different land use types. Means of the main effect were compared using Duncan multiple-range procedure test at P ≤ 0.05 for significance. Pearson correlation coefficients were used for the correlation analysis between soil physical properties and soil carbon. All the figures were produced using Origin 10.0 (Originlab, Northampton, Massachusetts, USA).