Introduction

It is well known that plant species differentially affect the size, composition and activity of soil biological communities, and that such differential effects influence ecosystem functions such as soil carbon (C) and nitrogen (N) storage in soil and the retention of nutrients and water in soil (Hooper and Vitousek 1998; Bardgett and others 2008; De Deyn and others 2009; De Vries and others 2012; Zeng and others 2009). Moreover, effects of plants on soil microbes can feedback to the plant community through the stimulation of nutrient mineralization processes or by promoting mycorrhizal fungi which enhance plant nutrient uptake, resulting in qualitative differences in plant community composition (Bardgett 2005; Kardol and others 2006; van der Heijden and others 2008). For instance, changes in plant species and functional group richness have been shown to influence the storage and loss of both C and N in model grassland communities, although these responses are related to the presence and biomass of certain plant species, notably N-fixers and forbs (De Deyn and others 2009). As a result, plant–soil feedbacks are likely to play a key role in restoration processes in abandoned land, and as recently argued by Kardol and Wardle (2010), the integration of principles developed from research on aboveground–belowground linkages can greatly assist restoration ecology.

Cultivation of steep slopes using traditional technology including deep ploughing, harrowing, and rain-fed farming, has been ongoing for several centuries on Loess Plateau, China (Catt 2001; Liu and Ding 2004). Due to serious soil and water loss, the Chinese Government in 1999 proposed the Conversion of Cultivated Land to Grassland and Forest program which aims to reduce soil erosion and improve soil quality through the establishment of perennial grass cover on degraded arable land (Gan and Sang 2002). Through this program, 3.61 million hectares of sloping cultivated land are going to be abandoned gradually across the Loess Plateau. The current restoration program advocates the introduction of some legume species in the early abandoned fields to accelerate the rate of recovery of vegetation and soil properties (Xue and others 2002; Li and others 2007, 2008). However, our understanding of how different legume species affect soil properties, including the biomass and functioning of the soil microbial community, is limited.

Different legume species are often selected to conserve soil and water in arid and semi-arid regions of China for their strong capability to grow in poor conditions-drought, cold and poor soil. These include Medicago sativa Linn. (Fu 2001), a perennial herb, is cultivated almost worldwide. It can grow in a loess hill-gully region for 10 years, being most productive after 4 or 5 years since introduction, and then gradually declines because of the excessive depletion of soil water (Cheng and others 2005). Astragalus adsurgens Pall. (Fu 2001) is a perennial species, endemic to China. It grows for only 6–7 years in the loess hill-gully region, and its maximum biomass occurs after 3 or 4 years, and then its growth slows and biomass decreases due to reduced available soil moisture (Cheng and others 2004). Finally, Melilotus suaveolens Ledeb. (Ohashi and Tateishi 1984), an annual or biennial herb, is widely distributed in northern part of China and other countries. Although it is known that these species differentially affect soil nitrogenase activity, in the order of Medicago sativa > Melilotus suaveolens > Astragalus adsurgens (Wang and Hu 2001), little is known about how these species influence soil nutrient levels or microbial communities which, as mentioned above, play a key role in the restoration of degraded soils.

In the region of our study, the ultimate goal of restoration is to get abandoned land back to the native vegetation type, Stipa breviflora grasslands, as quickly as possible. Therefore, Stipa breviflora is the target species in the restored communities. The motivation for using legumes, however, is based on their ability to rapidly establish and facilitate the restoration of native vegetation (Zou and others 1998; Xue and others 2002; Li and others 2007, 2008), to aid soil stabilization, and to increase soil N availability (Rochon and others 2004; Hopkins and Wilkins 2006) and potentially soil C sequestration (Soussana and others 2004; Fornara and Tilman 2008; De Deyn and others 2009; De Deyn and others 2011). Also, Li and others (2008) reported weeds can be inhibited by the introduction of legume species, and Melilotus suaveolens can facilitate the growth and establishment of native grassland species, thereby accelerating the course of old-field succession. Despite these potential benefits, there have been few specific evaluations of the effects of introducing different legumes species on the recovery of soil properties and functions that underpin successful restoration programs.

Soil microbial communities are crucial to the functioning of soils. This is because soil microbes are responsible for establishing biogeochemical cycles. Soil microbial properties such as microbial biomass may be used as early and sensitive indicators of soil quality (Bending and others 2004). Soil basal respiration, a measure of soil organic matter decomposition, has been reported to be influenced by soil temperature, moisture (Liu and others 2005), C input (Landi and others 2006) and microbial biomass. The metabolic quotient (qCO2) has been applied for the quantification of environmental effects on the microbial community in soils (Anderson and Domsch 1990; Brookes 1995; Wardle and Ghani 1995; Bååth 1998; Chodak and others 2009). In ecological terms, a high qCO2 reflects a high maintenance of carbon demand, and if the soil system cannot replenish the carbon which is lost through respiration, microbial biomass must decline (Anderson and Domsch 2010). The ratio of MBC:TOC has been used as sensitive index of soil quality for management practices (Saviozzi and others 2001) and for accumulation of organic matter in soil (Sparling 1992). Changes in the MBC:MBN ratio have often been attributed to shifts in the structure of the soil microbial community (Jenkinson 1976; Paul and Clark 1996; Kara and Bolat 2008). Increasing contents of organic carbon and microbial biomass may result in increased functional diversity of soil microbial communities (Yan and others 2000) and thus increased functionality and stability of soil ecosystems (Degens and others 2001).

The purpose of this study was to investigate the influence of different legume species on soil microbial communities and C and N contents in abandoned fields on the central Loess Plateau, China. Specifically, we assessed the effect of different treatments that are typically used to accelerate rates of grassland restoration, namely the introduction of the legume species Medicago sativa, Melilotus suaveolens, and Astragalus adsurgens on soil C and N content, and the biomass, activity, and functional diversity of the soil microbial community. Previous studies in this region have shown that the cover and density of Stipa breviflora, the target species of restoration, are 2.3 and 18.4 %, respectively, in near native grassland (Li and others 2007), and soil C and N contents are 10.01 ± 0.02 g kg−1 and 0.13 ± 0.03 g kg−1, respectively, and soil microbial C and N are 132.34 ± 2.77 mg kg−1 and 18.86 ± 1.22 mg kg−1 (Li and others 2007; Jia and others 2007); these values can hence be used as targets for the effective functioning of native grassland. The aim of this study, therefore, was to assess the effectiveness of the different legume species in the recovery of soil biological properties and C and N contents in abandoned fields, in order to provide a basis for future decisions on effective grassland restoration programs in this and other regions. We hypothesize that legume species will differentially impact on soil C and N contents, microbial biomass C and N, measured using chloroform-fumigation, microbial activity, measured as basal respiration and metabolic quotient (qCO2), and substrate utilization rates and functional diversity of soil microbial community, measured using the Biolog EcoPlate method.

Materials and Methods

The Experimental Site

The study was conducted at the Ecological Research Station of Lanzhou University in Yuzhong County, Gansu Province, China (43°26′–44°08′N, 116°04′–117°05′E, 2,400 m above sea level). The field site has a rusty dark loessial soil (Chinese soil taxonomy, Cooperative Research Group on Chinese Soil Taxonomy 1995) or Loess Orthic Entisol (FAO taxonomy, FAO-UNESCO 1988) (sand 12.3 %, silt 66.9 %, clay 20.8 %) with pH values ranging from 8.2 to 8.5 (Shi and others 2003; Wang and others 2008). The area has a semi-arid desert-grassland climate with mean annual precipitation of 360 mm, approximately 80 % of which falls during the 3-month period from June to August. Annual mean temperature is 4.4 °C, the maximum temperature is 19.0 °C (July), and the minimum temperature is −8.0 °C (January). Agriculture in this area is typically based on crops of spring wheat (Triticum aestivum), til (Linsum usitatissinum), pea (Pisum sativum), and potato (Solanum tuberosum).

Experimental Design

The effects of sowing the three legume species, Medicago sativa, Astragalus adsurgens, and Melilotus suaveolens (hereafter referred to Med sat, Ast ads, and Mel sua, respectively) on recently abandoned fields, last cropped with til in 2002, were compared with a control site that was harrowed at the same time as the others, but left unseeded for a period after (Control). The recovery effect of legume species on soil properties were also compared with a target soil (Target), another successional control. The seeding treatments were carried out in late April 2003. Four plots for seeding treatments (Med sat, Ast ads, Mel sua, and Control) of 20 × 20 m were assigned in random order in each of three fields. All fields were located within an area of 3 km2, had the same south facing orientation, and a slope with 20°–25°. All fields were protected against large mammal herbivores (mainly sheep and cattle) with fencing. Each plot was plowed pulling by ox, and legume seeds were broadcasted by hand at the same time. Then the plot was harrowed to incorporate the seeds into soil. Due to differences in seed biomass between the species, standard seed planting densities of 22.5 kg ha−1, 22.5 kg ha−1, and 15 kg ha−1 of Medicago sativa, Melilotus suaveolens and Astragalus adsurgens, respectively, were used; therefore, the total number of seeds was same between species (Wang 2000). Before seeding legumes, all the plots were bare for weeds did not germinate before May in this region. The average density of the soil seed bank was (2.328 ± 0.12) × 107 seeds ha−1, and there was no significant difference among the plots (ANOVA P > 0.05). Soil organic C and N contents before sowing were 4.95 ± 0.11 g kg−1 and 0.08 ± 0.01 g kg−1, respectively, across all plots.

Plant Biomass and Soil Physical Characteristics Determination

Ten permanent quadrats of 50 × 50 cm each per plot were set up and randomly placed throughout the plot in spring 2003. The cover and number of plant species in each permanent quadrat were counted at peak standing biomass (5–15 August). Above-ground biomass was sampled by clipping all plants in ten quadrats of 50 × 50 cm2 situated adjacent to the permanent quadrats that were used for cover assessment. All living vascular plants were sorted into species, dried and weighed. The above-ground biomass was expressed as dry mass m−2.

Soil bulk density was determined on core samples which were taken by driving a metal corer into the soil at the 0–20 cm depth. The soil samples were then oven dried at 105 °C for 24 h and weighed. Bulk density was the weight of dry soil divided by the total soil volume. There were five replications in each plot at 0–20 cm soil depth. Soil moisture (gravimetric water content) in 20 cm was measured by weighting soil samples taken from fields and then oven drying at 105 °C for 24 h and weighting. Soil pH was measured by pH meter with a soil-to-solution ratio of 1:2.5. Plant above-ground biomass and soil physical characteristics are presented in Table 1.

Table 1 Plant biomass and soil physical characteristics under different treatments in 2005 and 2007
Full size table

Soil Organic Carbon and Total Nitrogen, Microbial C and N Determinations

After above-ground biomass sampling, three quadrats in each plot were randomly selected to collect soil samples in years of 2003, 2005, and 2007 at peak standing biomass (5–15 August). Five cores (5 cm in diameter and 20 cm in depth) of 0–20 cm surface soil were taken randomly in each quadrat and mixed completely to get a composite fresh sample for each quadrat, for a total of three samples per plot and nine samples per treatment each year. After removing large plant materials and fine roots, each sample was divided into two parts. One part of each soil sample was air-dried for the estimation of soil physicochemical parameters and the other part was used to assess biological properties. Specifically, soil C and N in microbial biomass was measured in 2005 and 2007. Basal respiration and metabolic quotient, as well as bacterial functional diversity of the microbial community were measured in 2007. This later part was sieved through a 2 mm screen and adjusted to 50 % of its water holding capacity and then incubated at 25 °C for 2 weeks to permit uniform rewetting and to stabilize the microbial activity after the initial disturbances.

For each sample, 0.5 g dry soil sample was used to measure soil organic C (TOC) using the dichromate oxidation method (Kalembasa and Jenkinson 1973) and 1 g dry soil sample was used to measure soil total N (TN) by the Kjeldahl method (Jackson 1958). Microbial C and N were determined by using the chloroform fumigation extraction method (Brookes and others 1985). Three 25 g sub-samples for each sample were fumigated with alcohol-free CHCl3 for 24 h at 25 °C. These samples were then extracted by adding 100 ml of 0.5 M K2SO4, shaking for 60 min and filtered through Whatman No. 2 paper. Three non-fumigated soil samples for each sample were processed in the same manner. Microbial biomass C (MBC) was estimated as the difference between fumigated and nonfumigated samples divided by the K2SO4 extract efficiency factor for microbial C (Kc = 0.38, Vance and others 1987). Microbial biomass nitrogen (MBN) was calculated by dividing the total N difference between fumigated and non-fumigated samples with K2SO4 extract efficiency factor for microbial N (K = 0.54, Brookes and others 1985).

Basal Respiration and Metabolic Quotient

Basal respiration was determined by quantifying the carbon dioxide (CO2) released in the process of microbial respiration during incubation at 28 °C for 24 h. This was done by placing 25 g soil samples into 3.0 l capacity glass containers with hermetic lids, together with a smaller flask containing 20 ml 1 M NaOH to capture the released CO2. The container was tightly sealed. CO2 was determined by titration with 0.5 M HCl, after precipitation of the barium carbonate formed by adding barium chloride (BaCl2) aqueous solution to the NaOH solution, utilizing phenolphthalein diluted into 100 ml ethanol (60 %,v/v) as an indicator. The metabolic quotient (qCO2) was calculated as the ratio of basal respiration to microbial biomass C (Anderson and Domsch 1990).

Bacterial Functional Diversity in Biolog EcoPlate

EcoPlate was used to assess the functional diversity of the microbial community, measured by Biolog EcoPlate™ system (Garland and Mills 1991; Insam 1997) which is a rapid, community-level physiological profile (CLPPs) approach for assessing patterns of sole C source utilization by mixed microbial samples (Garland 1996; Gomez and others 2004; Nautiyal and others 2008). Each Biolog EcoPlate has 31 different carbon substrates with triplicates providing a metabolic pattern of the microbial community. Ninety-six well microplates were inoculated with 150 μl of the 10−2 suspension per well. This dilution was chosen in order to minimize the influence of soil-borne C sources and soil particles on the color development, which is indicative of carbon-source utilization. Plates were incubated at 28 °C for 168 h in the dark. The rate of utilization is indicated by the reduction of tetrazolium, a redox indicator dye, which changes from colorless to purple. Data were recorded at 12 h intervals for 7 days at 590 nm with Vamax reader (Microlog ReL 3.5). Microbial activity in each microplate, expressed as average well color development (AWCD), was determined as described by Garland (1996).

AWCD and Shannon–Wiener diversity index in Biolog EcoPlate were calculated using data of Biolog EcoPlate collected during 0–168 h and at 168 h, respectively, as described by Nautiyal (2009) and Staddon and others (1997).

AWCD = ∑(C–R)/31, C is sum of optical densities of 31 wells in a plate, R is optical density of control well in the same plate.

Shannon–Wiener diversity H = –ΣP ilnP i, of which P i is proportional color development of the ith well over total color development of all wells of a plate.

Statistical Analysis

Data were analyzed using SPSS 16.0 and Origin 7.0. Spearmen’s correlation coefficients (r) were calculated and ANOVAs were used for individual treatment comparisons at α = 0.05, with separation of means by LSD. Principal component analysis (PCA) was performed on data divided by the AWCD of 168 h as described by Merkl and Schultze-Kraft (2006), Chen and others (2011) to discriminate the carbon (C) utilization patterns of soil microbial communities.

Results

Soil TOC and TN, MBC, and MBN

The introduction of legumes significantly affected TOC (F 0.05 (3,107) = 5.273, P = 0.012) and TN (F 0.05 (3,107) = 2.843, P = 0.042), and as time progressed, both these measures increased across all treatments (Fig. 1). TOC was significantly greater in soil planted with Medicago sativa than in the control, while TOC in soil planted with Astragalus adsurgens was similar to that in the control, except in 2005 when this measure was greater than in the control. However, TOC in soil of Melilotus suaveolens was significantly lower than with the other two legumes treatments or the control, except that in 2005. Soil TN in the Medicago sativa treatment was significantly higher than in other two legumes treatments and also greater than in the control in each year. TN in soil of Melilotus suaveolens treatment was lower than that in the control in 2005 and 2007, but this difference was only significant in 2007. Compared to the target values, soil TOC under Medicago sativa in 2005, and Medicago sativa and Astragalus adsurgens in 2007, and TN under all three legumes treatments in 2005 and 2007, was higher (Fig. 1).

Fig. 1
figure 1

Soil carbon (A) and nitrogen (B) properties under different treatments in different years. Med sat = Medicago sativa; Ast ads = Astragalus adsurgens; Mel sua = Melilotus suaveolens; TOC total organic carbon; TN total nitrogen. Values are means and SE (n = 9) and those with the same letter are not significantly different at the P < 0.05 levels

Full size image

Legume species significantly affected soil microbial biomass C (F 0.05 (3,71) = 4.420, P = 0.0008) and N (F 0.05 (3,71) = 2.979, P = 0.016). Soil MBC and MBN in the Medicago sativa and Astragalus adsurgens treatments were significantly greater than that in the control, although they were greatest in soil of Medicago sativa (Fig. 2). MBC and MBN in soil of Melilotus suaveolens was significantly greater than that in the control and Astragalus adsurgens treatment in 2005, but significantly lower than in the control and Astragalus adsurgens treatment in 2007. Soil MBC under the Medicago sativa and Astragalus adsurgens treatments in 2005 and 2007, and the Melilotus suaveolens treatment in 2005, were greater than the target soil values, while under the Medicago sativa treatment, in both years, MBN was greater than the target values (Fig. 2). Microbial biomass C and N were positively related to soil total N (r = 0.772, P < 0.01; r = 0.863, P < 0.05, respectively), but not TOC (Table 2).

Fig. 2
figure 2

Soil microbial biomass carbon (A) and nitrogen (B) and ratios of MBC:MBN (C) and MBC:TOC (D). Med sat = Medicago sativa; Ast ads = Astragalus adsurgens; Mel sua = Melilotus suaveolens; TOC total organic carbon; MBC microbial biomass carbon; MBN microbial biomass nitrogen; MBC:MBN ratio of microbial biomass carbon and nitrogen; MBC:TOC ratio of microbial biomass carbon and total organic carbon. Values are means and SE (n = 9) and those with the same letter are not significantly different at the P < 0.05 levels

Full size image
Table 2 Coefficients of correlation between the microbial and chemical properties of soils
Full size table

The ratio of MBC:MBN was greater in all legume treatments relative to the control, although this response was greatest for Medicago sativa and Melilotus suaveolens in 2005, and Astragalus adsurgens in 2007 (Fig. 2) (F 0.05 (3,71) = 3.535, P = 0.021). The value increased to 20.70 in Medicago sativa treatment compared to 11.86 in the control in 2005, while 10.92 in Astragalus adsurgens treatment compared to 7.48 in the control in 2007. The ratio of MBC:TOC was greatest in soil of the Medicago sativa and Astragalus adsurgens treatments relative to the control in 2005, but there was no difference in this measure for soil of the Astragalus adsurgens treatment and the control. However, the ratio of MBC:TOC decreased in 2007, and it was greatest in Astragalus adsurgens, followed by the Medicago sativa treatment. There was no difference in this measure for soil of the Melilotus suaveolens treatment and the control.

Basal Respiration and Metabolic Quotient

Basal respiration ranged from 8.46 to 11.76 mg C−CO2 kg−1 soil h−1 across legume treatments, which was significantly higher than that of the control (F 0.05 (3,35) = 7.315, P = 0.0004), although there was no significant difference among legumes treatments except Melilotus suaveolens and Astragalus adsurgens (Table 3). Basal respiration was highest in soils of the Medicago sativa and Astragalus adsurgens treatments. In general, qCO2 was significantly greater in soils of the legume treatments relative to the control (F 0.05(3,35) = 3.535, P = 0.021) (Table 3) and there were significant differences across all treatments except between Medicago sativa and Astragalus adsurgens.

Table 3 Basal respiration (C−CO2 mg kg−1 h−1) and metabolic quotient (qCO2) under different treatments
Full size table

AWCD and Functional Diversity of Soil Microbial Community

In general, during 168 h period of incubation, soil microbial communities under legumes had higher activities than the control (Fig. 3), although this measure was greatest under Astragalus adsurgens, followed by Melilotus suaveolens, Medicago sativa and then the control. At the early incubation time (60 h), there was no significant difference among all treatments, except Astragalus adsurgens and control. However, after 60 h incubation, there was significant difference across all treatments (F 0.05 (3,11) = 9.274, P = 0.009).

Fig. 3
figure 3

Average well color development (AWCD) (means and SE (n = 3)) of soil microbial community in different treatments during incubation

Full size image

PCA of sole C source utilization patterns for soil microbial community, based on the OD of Biolog EcoPlate at 168 h incubation, showed that principal component 1 (PC1) explained 37.47 % of variance, which is mainly the carboxylic acid carbon source (except α-d-lactose and d-cellubiose). Principle component 2 (PC2) explained 17.86 % of variance, which is mainly the carbohydrate carbon source (except 1-phosphate glucose and glycine l-glutamate) (Table 4). The PCA results showed that the C substrate utilizing profiles were separated into two distinct groups: legume treatments and the control (Fig. 4). The main substrates utilized by microbes under legumes were l-asparagine, d-galactoic acid γ-lactone, d-gluconic acid, l-serine, and glycyl l-glutamine while those under control were Tween 80, Tween 40, pyruvic acid methyl ester, l-arginine, d-xylose, l-phenylalanine, α-ketobutyric acid, phenylethylamine, d, l-α-glycerophosphate, and 4-hydroxybenzoic acid. The metabolic Shannon diversity of soil microbial community in Biolog EcoPlate under legumes were significantly higher than that of control (F 0.05 (3, 11) = 6.969, P = 0.013), although there was no significant difference among legumes treatments. The highest Shannon diversity index was in the soil of Astragalus adsurgens (Fig. 5).

Table 4 Main carbon with high correlation coefficients for PC1 and PC2
Full size table
Fig. 4
figure 4

PCA of soil microbial communities under different treatments

Full size image
Fig. 5
figure 5

Diversity index of soil microbial communities (means and SE (n = 3)) calculated using data of Biolog EcoPlate after 168 h of incubation. Med sat = Medicago sativa; Ast ads = Astragalus adsurgens; Mel sua = Melilotus suaveolens. Values with the same letter are not significantly different at the P < 0.05 level

Full size image

Discussion

Our results show that the effects of legumes on soil microbial communities and soil C and N contents are both species and time dependant. After 3–5 years since their introduction, the perennial legumes Medicago sativa and Astragalus adsurgens positively affected soil microbial biomass and soil C content. In contrast, after 3 years, the annual or biennial legume Melilotus suaveolens had neutral or negative effects on soil C and N, and positively affected soil microbial biomass, while this species had negative effects on these measures after 5 years. Soil under Medicago sativa had the highest soil microbial biomass and C and N content (Figs. 1, 2). Such variation in the effect of legume species on soil microbial biomass and soil C and N contents could be due to differences in the productivity and nutrient content of the three legumes (Yang and others 2005), as well as differences in life history and other growth characteristics (Cheng and others 2004; Cheng and others 2005). For instance, Astragalus adsurgens had the highest above-ground biomass and highest total N and phosphorus (P) content of shoot and root, while Melilotus suaveolens had the lowest total N and P content (Yang and others 2005).

In both 2005 and 2007 in the Medicago sativa treatment, and in 2007 for the Astragalus adsurgens treatment, soil C contents were greater than the target value derived from native grassland, the desired end-point of restoration. However, soil N contents in 2005 exceeded the target for this measure in all three legume treatments, although soil N contents were greatest for Medicago sativa (Fig. 1). These results indicate that soil C contents could be restored in 3 years through the introduction of Medicago sativa, and in 5 years through the introduction of Astragalus adsurgens, whereas soil N can be restored within 3 years with all legumes. Higher soil MBC in the Medicago sativa and Astragalus adsurgens treatments in both 2005 and 2007 and higher MBN in the Medicago sativa treatment in 2005 and in the Medicago sativa and Astragalus adsurgens treatments in 2007 (Fig. 2) indicated MBC and MBN could be recovered after 3–5 years since their introduction. Overall, these results indicate that soil C and N could be effectively restored by introducing Medicago sativa and Astragalus adsurgens into early abandoned fields, whereas it could be retarded by introducing Melilotus suaveolens as evidenced by lower TOC, TN in this treatment relative to control in 3 years and lower MBC and MBN in 2007 (Figs. 1, 2).

The introduction of the legumes Medicago sativa and Astragalus adsurgens increased soil microbial biomass, the ratio of MBC:MBN, basal respiration and the metabolic quotient (qCO2), and the activity and functional diversity of the microbial community, measured using the Biolog EcoPlate method. In contrast, Melilotus suaveolens reduced soil microbial biomass C and N after 5 years (Figs. 2, 3, 4, 5). The positive effect of the two legumes on soil microbial communities is likely due to the promotion of plant biomass (Table 1) (Li and others 2007), and hence the amount of N-rich organic matter entering the soil, as well as nitrogen availability (Wang and Hu 2001), which could cause microbial populations to flourish. The high ratio of MBC:MBN generally, which ranged from 7.48 to 10.92, might suggest that the microbial biomass contains a higher proportion of fungi in soil of these abandoned-fields, and that all legume species increase the proportion of fungi biomass in soil microbial community (Jenkinson 1976; Pichtel and Hayes 1990; Paul and Clark 1996; Kara and Bolat 2008).

The negative effect of Melilotus suaveolens on soil microbial biomass after 5 years was likely due to reduced soil moisture (Table 1), given that this species is known to be very water demanding and its growth slows down and its biomass decreases due to deficiency of soil moisture on Astragalus adsurgens grasslands from the fifth year (Cheng and others 2004), as well as the previously shown inhibitive effect of this legume on natural colonizers (Li and others 2008). The notion that this legume species reduced soil moisture is supported by the fact that soil water contents were least under Melilotus suaveolens than under the other two legumes, and in the surface soil (0–20 cm) or the succession control during the growing season (Jia 2007). The inhibitive effect of the three legumes on weeds disappeared after 3 years, although the strongest effect in the first 2 years appeared in the Melilotus suaveolens-sown treatment (Li and others 2008). Although Melilotus suaveolens gradually disappeared from the community, its effect on soil desiccation remained for at least 3–5 years (Table 1). Combined, the decrease in soil water content and input of organic matter likely inhibited the growth of soil microorganisms; a view that is supported by the knowledge soil microbial biomass in semiarid temperate steppe is strongly regulated by these factors (Liu and others 2009). These legume species not only increased soil organic matter (except Melilotus suaveolens), but also affected its quality and availability of organic matter for microorganisms (Insam and Domsch 1988; Anderson and Domsch 1990; Wang and others 2003) as evidenced by the increase in the MBC:TOC (Fig. 2) and thereafter significantly affected the number and activity of soil microorganisms (Powlson and others 1987; VA’Squez_Murrieta and others 2007). More active microbial communities (Figs. 3, 5) of lower biomass in the Melilotus suaveolens treatment (Fig. 2) could explain the greater qCO2 in this treatment, although greater water limitation in this treatment might also play a role (Wardle and Ghani 1995).

Past studies have shown that there is a positive relationship between soil organic C and microbial biomass (Guggenberger and Zech 1999; Piao and others 2001; Liu and others 2010). However, in our study, MBC and MBN were positively related to soil total N, but not organic C, which was in accordance with a previous study of Jia and others (2007) in this region. Differences in AWCD could result from differences in the biomass and nutrient content of the legumes, given that Astragalus adsurgens had the highest biomass and total N and P content (Yang and others 2005). Previous research has shown that the differences in AWCD are in part due to lower microbial bio-mass (Garland 1996; Grayston and others 2004). In our study, AWCD values were correlated with microbial biomass (r = 0.771, P < 0.05, data not shown), which was con-sistent with other report (Bååth 1998), and might be due to variations in both the quantity and quality of organic matter in soil. PCA analysis of microbial communities revealed distinct different patterns of C utilization for the legumes treatments and the control. The pattern of C utilization in natural succession was clearly separated from those in legumes treatments (Fig. 4), which was generally consistent with AWCD (Fig. 3). Sugars and carboxylic acids were the main C sources utilized by the soil microbial community in the legumes treatments. This could result from differences in the composition of root exudates across legume species (Nannipieri and others 2008), and root exudates from legumes containing rapid utilization substrates by bacteria, such as sugars, amino acids and carboxylic acids (Gaworzewska and Carlile 1982). It could also be due to variations in the quality of litter inputs from roots and shoots among legume species.

In general, the impact on the soil microbial community and concentrations of C and N of introducing legumes into abandoned fields depended on time-scale and were species-specific. Soil C and N could be effectively restored by introducing Medicago sativa and Astragalus adsurgens into early abandoned fields. However soil C and N was retarded after 3–5 years, as was the biomass of microbes after 5 years following the introduction of Melilotus suaveolens. Overall, our results show that the restoration practice of adding legumes significantly affects soil C and N contents and the abundance, activity and functional diversity of the microbial community, albeit in a species specific way. Therefore, we propose that certain legume species could be used to accelerate ecological restoration of degraded soils, and hence assist in the protection and preservation of the environment.