Ecological stoichiometry of plant leaves, litter and soils in a secondary forest on China’s Loess Plateau

Ecological stoichiometry can reveal nutrient cycles in soil and plant ecosystems and their interactions. However, the ecological stoichiometry characteristics of leaf-litter-soil system of dominant grasses, shrubs and trees are still unclear as are their intrinsic relationship during vegetation restoration. This study selected three dominant plant types of grasses (Imperata cylindrica (I. cylindrica) and Artemisiasacrorum (A.sacrorum)), shrubs (Sophora viciifolia (S. viciifolia) and Hippophae rhamnoides (H. rhamnoides)) and trees (Quercus liaotungensis (Q. liaotungensis) and Betula platyphylla (B. platyphylla)) in secondary forest areas of the Chinese Loess Plateau to investigate ecological stoichiometric characteristics and their intrinsic relationships in leaf-litter-soil systems. The results indicated that N concentration and N:P ratios in leaf and litter were highest in shrubland; leaf P concentration in grassland was highest and litter in forestland had the highest P concentration. Soil C, N and P concentrations were highest in forestland (P < 0.05) and declined with soil depth. Based on the theory that leaf N:P ratio indicates nutritional limitation of plant growth, this study concluded that grass and shrub growth was limited by N and P element, respectively, and forest growth was limited by both of N and P elements. The relationships between the N concentration in soil, leaf and litter was not significant (P >0.5), but the soil P concentration was significantly correlated with litter P concentration (P < 0.05). These finding enhance understanding of nutrient limitations in different plant communities during vegetation restoration and provide insights for better management of vegetation restoration.


INTRODUCTION
Soil erosion remains a major global environmental problem, accelerating soil nutrient losses and ecosystem degradation (Luque et al., 2013). Soil nutrient losses greatly decreased soil quality (Liu & Dang, 1993), which seriously threatens the stability of ecosystems. Vegetation restoration is a powerful approach for ecological restoration of degraded lands, as it can control soil erosion and improve ecosystem functions and services (Godefroid et al., 2003;Zheng, 2006;Jiao et al., 2012;Sauer et al., 2012;Zhao et al., 2015; one of the most severely eroded areas of China (Jiao et al., 2012;Zhao et al., 2015). In past centuries, the majority of forestlands were destroyed to satisfy the food needs of the growing population, which resulted in severe soil erosion and land degradation. The Grain to Green Program (GTGP) was implemented to control soil erosion and improve ecosystem degradation, with a main goal of converting low-yield steep-slope croplands into permanent vegetation cover (Jiao et al., 2012;Zhao et al., 2015). Vegetation restoration generated a diverse flora and reduced soil erosion, raising interest in the characterization of this recovering ecosystem. For example, An & Shangguan (2010) and Chai et al. (2015) studied leaf stoichiometric traits and concluded that the growth of vegetation was N-limited at each secondary successional stage, according to the leaf N:P threshold. Ai et al. (2017) observed that the slope aspect had various effects on plant and soil C:N:P stoichiometry. Variations in vegetation types influenced soil C:N:P ratios, which were higher in afforested lands than in slope croplands (Zhao et al., 2015;Deng et al., 2016;Zhao et al., 2017). Jiao et al. (2013) studied soil stoichiometry during vegetation successional changes and reported that soil N:P ratio increased with the vegetation restoration year. It was even reported that forest age had a significant effect on C, N, P and K concentrations and their ratios in plant tissues and soil (Li et al., 2013). Most previous studies addressed the stoichiometric characteristics of soil system and vegetation communities, including forests and grasslands, as well as litter individually or in both. However, the ecological stoichiometry of the plant-litter-soil system as a whole has been rarely described (Zeng et al., 2017;Cao & Chen, 2017), and the effects of dominant plant communities (tree, shrub, grass) during vegetation restoration on this ecological stoichiometry remains poorly understood. This will provide a better understanding of nutrient limitation in different plant communities during vegetation restoration and improve ecosystem management. In addition, the majority of previous studies have focused on topsoil (Jiao et al., 2013;Li et al., 2013;Zeng et al., 2016;Zeng et al., 2017), there is little information on stoichiometry change with the soil profile (Zhao et al., 2015;Deng et al., 2016). Due to the depth of thick loess on the Loess Plateau, the majority plant roots are distributed within the top 100 cm. Therefore, it is important to investigate change of the stoichiometry of C, N and P with soil profile depth.
Vegetation succession starts from annual grass, then to perennial grass, shrub, forest after farmlands are abandoned. So, three dominant plant communities of grasses (Imperata cylindrica and Artemisia sacrorum), shrubs (Sophora viciifolia and Hippophae rhamnoides) and trees (Quercus liaotungensis and Betula platyphylla) were selected in the Ziwuling secondary area of the Loess Plateau to investigate ecological stoichiometry in the plantlitter-soil system and their intrinsic relationships. The specific objectives of this study were to (1) determine leaf and litter C, N and P concentrations and their ecological stoichiometry characteristics in six dominant plant species; (2) investigate distributions of soil C, N, and P concentrations and ecological stoichiometry characteristics in soil profile; (3) examine the relationships of ecological stoichiometry in leaf-litter-soil system (C, N, and P); and (4) assess the limiting nutrient element for plant growth in the six plant species. The effort will provide information about ecological stoichiometry and theoretical support for enhancing vegetation and ecosystem restoration on the Loess Plateau.

Study site description
This study site is located at Fuxian County, Shanxi Province, China (35 • 5.4 N,109 • 8.9 E), in the center of Loess Plateau, south of the Yan'an city. The topography and landform belong to loess hilly-gully region with elevation ranging from 920 to 1,683 m (Zheng, 2006). The mean annual temperature ranges from 6 to 10 • C and mean annual precipitations is between 600 to 700 mm. The soil is mainly composed of loess, which can be classified as a Calcic Cambisol (USDA NRCS, 1999). The soil texture was 28.3% sand (>50 µm), 58.1% silt (50-2 µm) and 13.6% clay (<2 µm). Vegetation of the Loess Plateau was almost completely removed more than 100 years ago, and soil loss was 8,000 to 10,000 t km −2 yr −1 (Zheng et al., 1997;Kang et al., 2014). In 1866-1870, the inner war happened in this region (Zhang & Tang, 1992;Tang et al., 1993;Zheng et al., 1997) and as population moved out, and then secondary succession of vegetation began. Currently, forest canopy closure is more than 0.6 and dominant species for tree are Quercus liaotungensis (climax forest community) and Betula platyphylla (early forest community); dominant species for shrub are Sophora viciifolia and Hippophae rhamnoides, both does not concur in same places; and main grass species are Imperata cylindrica and Artemisia sacrorum (Zheng, 2006). The distribution area of the above mentioned six dominant species occupies more than 70% of total area in the study site.

Soil and plant sampling
According to our field investigation, there are 38 species in the study site, including 18 artificial species and 20 natural species, which cover five tree species, six shrub species, nine grass species. Moreover, the six tree, shrub, and grass species, i.e., Quercus liaotungensis and Betula platyphylla (forest communities), Sophora viciifolia and Hippophae rhamnoides (shrub communities) and Imperata cylindrica and Artemisia sacrorum (grass communities) are dominant species and their distribution area occupies more than 70% of total area in the study site. Other studies also reported that these six species are dominant species of natural vegetation succession (Zheng, 2006;Wang, Shao & Shangguan, 2010;Zhang & Shangguan, 2016). Thus, these six species have been selected to investigate to ecological stoichiometry of plant leaf, litter and soil in a secondary forest on China's Loess Plateau. For each dominant species, three experimental sites (three replications) with a similar site condition including slope position (slope length, gradient and aspect), soil type and altitude were set up to collect samples. In addition, the distance within all experimental sites was within approximately 3 km, which reduced impacts of previous site condition. Plant leaves and soil samples were collected in late July 2016 when plants were in a vigorous growth period, and litter samples on the soil surface consisting of leaf fall over multiple years that were not decomposed were obtained in late October 2016. Table 1 shows the characteristics of these three plant types.
Two plots with 10 × 10 m size were established in each experimental site of forest type, and the plots sizes for shrub and grass types were 5 × 5 m and 1 × 1 m, respectively. Ten to twenty complete expanded living and sun-exposed leaves were randomly collected from five to ten healthy individual plants per plot from shrubs or trees, and a total of 80 to 100 leave samples were collected. For each grass plot, all stems and leaves were completely cut from three 0.25 m 2 sampling areas. Leaves from each plot were evenly mixed and then put into a paper bag. Litter samples were collected along the diagonal lines of three 1 × 1 m squares per plot, and mixed and stored in paper bags. All samples of leaves and litter were carried back to the indoor laboratory for analysis. The total of 256 soil samples from a 100 cm-depth profile were collected using a 5-cm diameter to collect soil samples along an S-shaped line in each plot. Before each soil sample was collected, soil sampler was sterilized with ethanol to avoid cross-infection. Moreover, the 100 cm soil profile was divided into six layers (0-10, 10-20, 20-40, 40-60, 60-80, 80-100 cm), and soil samples from each layer were obtained from five points. The five soil samples of each layer were mixed evenly and stored in plastic bags, and then all soil samples (6 plant species ×3 experimental sites ×2 sample plots ×6 soil sample layers) were transported to the indoor laboratory.

Sample analysis
Leaf and litter samples were oven dried at 70 • C for at least 48 h or more to reach a constant mass level, and then weighed. Dried plant samples were ground to a fine powder using a plant-sample mill (1,093 Sample Mill, Foss, Sweden). Soil samples were air-dried and sieved using a 0.25 mm mesh. To determine C concentration in plant and soil, the Walkley-Black modified acid-dichromate FeSO 4 titration method was used (Bao, 2000), and the Kjeldahl method (KJELTE2300, Sweden) was applied to measure the total N concentration in plant and soil. The total P concentration in plant was measured by using a Spectrophotometer UV-2300 (Techcomp Com, Shanghai, China) after digestion with H 2 SO 4 and H 2 O 2 , and the total P concentration in soil was determined by a spectrophotometer after wet digestion with H 2 SO 4 and HClO 4 (Bao, 2000). Leaf, litter and soil C, N, P concentrations were expressed as g/kg on dry weight basis. The C:N:P ratios in leaves, litter and soil were computed as mass ratios.

Statistical analysis
All data are presented as mean ± standard errors and tested for normality of distributions and homogeneity of variances before analysis. A one-way analysis of variance (ANOVA) was used to analyze the effects of the plant type on nutrients and stoichiometric characteristics in leaf, litter and soil. Two-way ANOVAs were computed to analyze the effects of plant type, soil depth and their interactions on soil C, N and P concentrations and their stoichiometry.

Notes.
Bars indicate the standard errors (n = 6). The lowercase letters above the bars indicate significant differences in leaf at different plant types and the capital letters represent significant differences in leaf at the same plant types of different species (P < 0.05).
The linear regression analysis was used to test the relationship between C, N and P concentrations in leaf, litter and soil. Pearson correlation was used to assess relationship between leaf, litter and soil stoichiometric characteristics. Differences were considered significant with a P < 0.05. All statistical analyses were determined with SPSS 19.0 software (SPSS, Inc., Chicago, IL, USA).

Leaf and litter nutrients and ecological stoichiometry in dominant plant communities
The leaf C, N and P concentrations were different among plant communities ( Table 2). The C concentration in leaf varied from 475 (grass) to 522 g/kg (forest), and was highest in B. platyphy lla and lowest in A. sacrorum. The leaf N concentration was 29.8 g/kg in shrub, and was significantly greater than in forest and grass (P < 0.05), while the leaf P concentration with 1.80 g/kg was highest in grass. The leaf C:N ratio varied from 16.9 (shrub) to 47.3 (grass), and was highest in I. cylindrica and lowest in H. rhamnoides. The leaf C:P ratio was significantly higher in Q. liaotungensis and S. viciifolia than other species (P < 0.05). The leaf N:P ratio varied from 6.12 (grass) to 22.6 (shrub) and was significantly higher in shrub than in grass and forest (P < 0.05).
The C, N and P concentrations in litter were significantly affected by plant types (Table 3). The litter C concentration varied from 360 (shrub) to 413 (forest), and was significantly higher in forest than in grass and shrub (P < 0.05). N concentrations showed a similar pattern between litter and leaf, and were significantly highest in shrub (P < 0.05). The litter P concentration varied from 0.51 (grass) to 0.97 g/kg (forest) and was highest in B. platyphy lla and lowest in I. cylindrica. The litter C:N and C:P ratios in grass were 52.9 and 735, respectively, and were significantly higher than in forest and shrub (P < 0.05). The litter N:P ratio varied from 12.5 (forest) to 24.2 (shrub), and was highest in H. rhamnoides and lowest in B. platyphy lla (P < 0.05).

Notes.
Bars indicate the standard errors (n = 6). The lowercase letters above the bars indicate significant differences in litter at different plant types and the capital letters represent significant differences in litter at the same plant types of different species (P < 0.05).

Soil nutrients and ecological stoichiometry in dominant plant communities and soil depths
Plant type and soil depth had significant effects on soil nutrients and their C:N:P ratios (Table 4). Soil C and N concentrations in forestland were greater than in shrubland and grassland at all soil depths and both were highest in Q. liaotungensis and lowest in A. sacrorum (P < 0.05). Soil P concentration in shrubland was lower than in grassland and forestland at every soil depth (P < 0.05), and there were no differences in B. platyphylla, Q. liaotungensis and A. sacrorum at 20-100 cm soil depths. Soil C:N ratio in forestland was significantly higher than in shrubland and grassland at both 0-10 and 10-20 cm soil depths (P < 0.05), but there were no significant differences at 20-100 cm soil depths (P > 0.05). Soil C:P and N:P ratios in forestland was significantly higher than in shrubland and grassland at both 0-10 and 10-20 cm soil depths (P < 0.05), but both were highest in shrubland at 20-100 cm soil depths (P < 0.05). Soil depth is a driving factor for soil nutrient concentrations and their ratios (Table 4 and Fig. 1). Soil C and N concentrations significantly decreased with soil sampling depth. Soil C and N concentrations decreased markedly from 10 to 40 cm of soil depth, and then slightly decreased from 40 to 100 cm. Soil P concentration tended to stable with the soil sampling depth. Soil C:N ratio fluctuated with depth, and soil C:P and N:P ratios had the same trend along the soil sampling depth and decreased markedly from 10 to 40 cm of soil depth, and then slightly decreased from 40 to 100 cm.
The results of the Two-way ANOVA analysis indicated that both plant type and soil depth significantly affected the soil C, N and P concentrations and their stoichiometry (C:N, C:P and N:P ratios). The interactions between plant type and soil depth significantly affected the soil C and N concentrations and C:N, C:P and N:P ratios but not soil P concentration (Table 5).

Relationships between C, N and P concentrations and their characteristics of ecological stoichiometry among leaf, litter and soil
There were significant correlations between leaf and litter for both N and P concentrations in three plant community types (P < 0.05) (Figs. 2B, 2C). The relationships between the plant C concentration and soil C concentration were significant in two soil layers (0-10

Notes.
Bars indicates the standard errors (n = 6). The lowercase letters above the bars indicate significant differences in different plant species at the same four soil layers, and the capital letters represent significant differences in different soil layers at the same plant species (P < 0.05).  ). In the three plant community types, there were no significant correlations between leaf P concentration and soil P concentration (Figs. 3G, 3H, 3I), but the soil P concentration was significant correlated with litter P concentration in 0-10 cm soil depth (P < 0.05) (Fig. 3G).
For the three plant community types, leaf C:N and N:P ratios were positively correlated with litter C:N and N:P ratios, respectively (P < 0.05) (Figs. 2D, 2F), while leaf C:P ratio was negatively correlated with litter C:P ratio (P < 0.05) (Fig. 2E). Leaf C:P had a positive correlation with soil C:P ratio at the 0-10 cm soil layer and over 0-100 cm soil profile (P < 0.05) ( Table 6), Leaf N:P ratio had a positive correlation with soil N:P ratio at two soil layers (0-10 and 0-20 cm) and the profile (0-100 cm)(P < 0.05) ( Table 6); and there was significant correlation between leaf and soil C:N ratio at the 0-10 and 0-20 cm soil layers (P < 0.05) ( Table 6). At the 0-10 cm soil layer, there was a significant correlation between litter and soil C:N ratio (P < 0.05) ( Table 6), and litter C:P ratios were negatively correlated with C:P ratios at two soil layers (0-10 and 0-20 cm) and the profile (0-100 cm), and only in the profile (0-100 cm) did litter N:P ratio have a positive correlation with soil N:P ratio (P <0.05) ( Table 6).

Impacts of dominant plant communities on leaf and litter nutrients and ecological stoichiometry
As a key subsystem, plants have a vital function in governing the stability of terrestrial ecosystem. C, N and P are essential nutrients for plant (Han et al., 2005;John et al., 2007) and their interaction regulate plant growth (Güsewell, 2004). Litter is one main way for nutrients to return to the soil and is an important part of the forest ecosystem. The decomposition of plant litter replenishes soil nutrients to provide conditions for the adjustment and demand of the plant nutrients (Agren & Bosatta, 1998). There are differences in the types, quantity and utilization efficiency of absorbed nutrients in different plants types. In this study, the results indicated that leaf C, N and P concentrations differed across plant communities. The reason is that different plant communities has different adaptability to the environment, and possess different strategies of nutrient adaptation (Wright et al., 2004;Han et al., 2005;Zheng & Shangguan, 2007;He et al., 2008;Wu et al., 2012). In this study, leaf C concentration in forest species was significantly higher than in grass and shrub species while the leaf P in forest species was significantly lower than in grass species. An explanation may be that trees construct nutrient poor woody tissues while grasses do not. The results are consistent with those of Wright et al. (2004), which reported that the leaf P concentration in herbaceous plants is significantly higher than in woody plants. Moreover, in this study, the C, N and P concentrations in plant leaves were higher than in the corresponding litter, which was consistent with previous studies (Pan et al., 2011;Zeng et al., 2017). Pan et al. (2011) showed that the C, N and P concentrations in the leaves of trees, shrubs and grasses were significantly higher than in litter, likely due to the reabsorption processes. Previous studies have shown that nutrients present in leaves are transferred to flowers, fruits, branches, and roots before leaf falling, thereby preventing nutrients loss (Schreeg et al., 2014). The results showed that N and P concentrations in litter varied greatly in different plant communities, and were significantly higher in trees than in grasses. This is because tree and shrub are deep-rooted plants, and have the strong capability of absorbing nutrients from multiple sources in the environment; while grasses have shallow roots and rely more on the recycling of their own nutrients. N and P elements are major limiting factors for plant growth in terrestrial ecosystems, and the leaf N:P ratio could be used as an indicator to identify the limiting nutrient factors (Koerselman & Meuleman, 1996;Güsewell, 2004). However, the threshold of N:P ratio is affected by study area, plant growth stage and plant species (Güsewell, 2004). Güsewell (2004) reported that leaf N:P ratio can be used to reveal N-limitation (N:P ratio < 10) or P-limitation (N:P ratio > 20) in the ecosystem. In this study, based on the Güsewell's proposal that leaf N:P ratio indicates nutritional limitation for plant growth, we concluded that grass and shrub growth was limited by N and P element, respectively, whereas forest growth was co-limited by both of N and P elements together in the research area. In this study, the leaf N:P ratio in S. viciifolia and H. rhamnoides were 22.6 and 21.0, respectively,  suggesting that shrub growth was P-limited. The leaf N:P ratios in Q. liaotungensis and B. platyphylla were 14.3 and 12.9, respectively, indicating that their growths were co-limited by both N and P. The leaf N:P ratios in A. sacrorum and I. cylindrica were 6.12 and 9.93, respectively, indicating that grass growth was limited by N. The results indicated that different plant communicates had different nutrient limiting elements, which was consisted with previous studies (Han et al., 2005). The reason is that grass species (I. cylindrica and A. sacrorum) is a shallow-rooted plant with a strong ability to absorb soil surface nutrients, particularly it has a greater capacity of relocating its leaf P before leaf falling than forest and shrub species, and it can more effectively utilize leaf P concentration to meet growth demands. Moreover, the biochemistry of the grass organic structure determines that more nitrogen is needed for growth. Therefore, grass species were less limited by P element than by N element. In addition, the results indicated that the growth of shrub species was limited by P element, which was similar to results reported by Han et al. (2005). This is because S. viciifolia and H. rhamnoides are inherent species in vegetation restoration on the Loess Plateau and were nitrogen-fixing plants, and the absorption on of N element is far greater than that of P element, which results the shrub species to be limited by P element. Furthermore, the result showed that leaf C:N and C:P ratios were lower than in litter, which is consistent with results reported by McGroddy, Daufresne & Heedin (2004), indicating that the reabsorption capacity for C is lower than for N and P. Although leaf N:P ratio can effectively reflect N or P limitation, the importance of the N:P ratio is mainly in its function as an indicator (Güsewell, 2004). If the leaf N:P ratio is to be used as an index to evaluate both N and P nutrient supplies in the Loess Plateau, further diagnosis regarding nutrient limitations should be conducted.

Impacts of dominant plant communities on soil nutrient and ecological stoichiometry
Plants play an important role in improving soil fertility and contribute to the accumulation of soil nutrients. Fu et al. (2010) found that vegetation restoration could improve the net fixation of C and N and reduce their losses. However, the performance in soil quality recovery differed among plant communities (Jiao et al., 2012;Zeng et al., 2016;Deng et al., 2016;Zhao et al., 2017). In this study, soil C, N and P concentrations in forestland was greater than in grassland and shrubland which is consistent with the previous results of Jiao et al. (2012) and Qi et al. (2015). This result could be explained by a larger amounts of litter present in forestland, a more above-ground litter and a higher volume of root exudates reaching the soil, resulting in higher nutrient concentrations in the forestland than in other plant communities. Soil C and N concentrations decreased with increasing of soil depth, while soil P concentrations were relatively stable with depth, which was consistent with Wei et al. (2009). The reasons might be the influence of soil parent material, the amount nutrient content of returning litter, the rate of decomposition, and plant nitrogen fixation, absorption and utilization. With an increasing of soil depth, the input of organic matter gradually decreased (Nelson, Schoenau & Malhi, 2008). However, soil P is mainly derived from rock weathering and leaching, and its mobility is very low, which caused vertical variation of P along the soil profile to be relatively stable (Wei et al., 2009). Soil C:N:P ratios are important indicators of organic matter composition, soil quality and nutrient supply capacity (Bui & Henderson, 2013). In this study, soil C:N:P ratios among the three plant communities were 16.9:1.7:1, 25.0:2.3:1 and 28.6:2.5:1 at the topsoil (0-10 cm), respectively (Table 2), These values are substantially lower than the average global value of 186:13:1 (Clevel & Liptzin, 2007). Loess soils are naturally low in C, meanwhile, the Loess Plateau has undergone a serious soil erosion prior to recent efforts at vegetation restoration, resulting in a low C:N:P ratio. In this study, soil C:N ratio across different plant communities and soil depths was approximately 10.8 in the Loess Plateau, which was similar to the average level (11.9) in China (Tian et al., 2010), but lower than the world's average value of 13.3 (Clevel & Liptzin, 2007). Previous studies showed that soil C:N ratio is negatively correlated with the decomposition rate of organic matter, and low soil C:N ratio indicates that organic matter is well decomposed (Zhao et al., 2015;Deng et al., 2016). The soil C:N ratio in grassland, shrubland and forestland was 10.1, 10.8 and 11.7, respectively, implying that organic matter had been completely decomposed. The soil C:N ratio in each plant community maintained relative stability with increasing soil depth, which is consistent with previous studies (Tian et al., 2010). This may be due to the same change dynamics in C and N. Soil C:P and N:P ratios in each plant community decreased with increasing soil depth, which may be due to the difference in the source of soil C, N and P. Furthermore, this study showed that soil C:P and N:P ratios in forestland was higher than in shrubland and grassland in the topsoil depth, which may be due to the fact that forest had more above-ground biomass than shrubland and grassland (Qi et al., 2015).

Relationships between C, N and P concentrations and their characteristics of ecological stoichiometry among leaf, litter and soil
Some previous studies have showed a strong correlation between leaf and soil nutrients (Parfitt, Yeates & Ross, 2005;Agren & Bosatta, 1998;Agren, 2008), while others found that there was no correlation between N and P concentrations in leaf and soil (Ladanai, Agren & Olsson, 2010;Yu et al., 2010). In this study, no significant correlation was found between soil N concentration with leaf N concentration for three plant community types. One possible reason is that through long-term adaptation to the habitat, the N concentration in plant leaves in this region may be more affected by the attributes of the species than the limitation of soil nutrients. In addition, Reich & Oleksyn (2004) showed that the mineral elements of plants are a combination of climate, soil nutrients and species composition. Other studies have suggested that soil temperature, soil water concentration, microbial activity and other factors have a greater impact on the mineral elements of plants (Chapin & Pastor, 1995;Güsewell, 2004). In this study, there was a significant correlation between litter N and P concentrations and their ratios with leaf N and P concentrations among the three plant types, indicating that the nutrients in litter were derived from plant leaves. In addition, a strong correlation between soil and litter for both C and P concentrations among the three plant types was observed. As a considerable portion of C and other nutrients elements in the litter could be released into the soil, such that litter was one of the main sources of soil nutrients (Agren et al., 2013). In general, this study showed that there is a close correlation between the concentrations of C, N and P and their ratios in leaf, litter and soil in three plant community types, which confirmed that C, N and P in the ecosystem were transported and transformed among plants, litter and soil (McGroddy, Daufresne & Heedin, 2004).

CONCLUSION
This study analyzed C, N and P concentrations and their stoichiometric characteristics in leaf, litter and soil of three dominant plant types: grass (I. cylindrica and A. sacrorum)), shrubs (S. viciifolia and H. rhamnoides) and tree (Q. liaotungensis and B. platyphylla)) during vegetation restoration on the Loess Plateau of China. The results indicated that plant community type had significant effects on leaf, litter and soil nutrient concentrations, and their stoichiometry characteristics. Grass species had highest leaf P concentration and forest species litter had highest P concentration. Leaf C, N and P concentrations were higher than in litter and soil (P < 0.05) and forest community type had highest soil nutrient concentrations at all soil layers and their ecological stoichiometries were highest in topsoil (P < 0.05). In addition, soil C:N:P ratios in all plant communities decreased with increasing soil depth. Soil P concentration and N:P ratio had significant positive correlations with litter P concentration and N:P ratio for the three plant community types (P < 0.05), respectively. However, there were no significant correlations between soil N, P concentrations and N:P ratio with leaf N and P concentrations and N:P ratio (P > 0.5), respectively. Based on the theory that leaf N:P ratio indicates nutritional limitation for plant growth, this study concluded that plant growth of the forest community type (Q. liaotungensis and B. platyphylla species) was co-limited by both of N and P elements, plant growth of shrub community type (H. rhamnoides and S. viciifolia species) was limited by P element and grass growth (I. cylindrica and A. sacrorum species) was limited by N element. These results can provide a scientific basis for the reconstruction of degraded ecosystem on the Loess Plateau of China.