Chen Ma · Xiuqin Yin,2,3 · Huan Xu · Yan Tao

Abstract A total of 900 soil samples were collected from Tve habitats, including primary coniferous broad-leaved mixed forests, secondary coniferous broad-leaved mixed forests, secondary broad-leaved forests, secondary shrub forests,and cutover lands in spring,summer,and autumn to quantify responses of soil Collembolans(springtails)to the restoration of vegetation of temperate coniferous and broad-leaved mixed forests. The results reveal that the taxonomic composition of Collembolans varied in the different stages of vegetation restoration. Seasonal variations were in regard to their abundance and richness. High similarities existed in Collembola communities at different stages of vegetation restoration,and distribution patterns of Collembola taxa displayed an evenness throughout all habitats.Soil Collembolans tended to gather on litter layers and soil surface; the highest abundance was found in the upper 5 cm soil layer during the initial stages of vegetation recovery. Tomocerus, Proisotoma, and Folsomia genera responded positively to the restoration of vegetation.However, responses of Ceratophysella and Parisotoma genera were negative.In addition,the Onychiuridae family did not respond to the vegetation restoration process.It was concluded that restoration of vegetative cover can increase the abundance of soil Collembolans, but different genera respond differently.

Keywords Soil Collembolans (springtails) · Temperate coniferous and broad-leaved mixed forests · Vegetation restoration

Introduction

Collembolans (springtails) form the largest of the three lineages of modern hexapods considered important consumers and decomposers in the belowground ecosystem.Large numbers and great varieties of individual taxa exist in the terrestrial ecosystem (Hopkin 1997). At the same time, Collembolans, nematodes, and mites constitute the main components of soil fauna (Rusek 1998), and play crucial roles in maintaining soil fertility, releasing nutrients, and improving the soil environment (Petersen and Luxton 1982; Siddiky et al. 2012; Waagner et al. 2012).Due to the narrow geographic scope of their activities as well as their poor ability to migrate, Collembolans are sensitive to changes in soil environments (Ponge et al.2003). Previous research has shown that if changes to the soil or vegetation occur,and if these exceed the limitations of adaption and regulation,the survival and reproduction of the Collembolans could be negatively affected, and potentially cause the extinction of some species (Alatalo et al.2015; Silva et al.2016). Therefore,Collembolans,as indicator species, have become an important topic in the study of belowground systems (Bellinger et al. 2018).

Vegetation is a critical part of terrestrial ecosystems and the key link for material circulation and energy ?ow(Wardle 2013). Human activities have proven to be damaging to vegetation, not only changing natural landscapes,but also causing environmental problems such as soil erosion, climate changes, and decreased biodiversity.Therefore, the destruction of vegetation is a crucial environmental problem (Zirbel et al. 2017). Vegetation restoration is the practice of renewing and restoring degraded, damaged, or destroyed ecosystems and habitats.It may be a process of natural succession or active human intervention, and could possibly restore the diversity and function of ecosystems (Newmark et al. 2017). This process changes aboveground ecosystems and also in?uences belowground ecosystems(Klopf et al.2017).The effects of vegetation restoration on belowground ecosystems are an important area of research; previous studies have been primarily focused on nutrient cycling, environmental changes,and microbiology of the soil(B?on?ska et al.2016;Gellie et al. 2017). In contrast, there have been relatively fewer studies on soil fauna in this research area (Forey et al. 2018), and the research has been comparatively scarce in regards to soil Collembolans in particular.

The Changbai Mountains, situated on the eastern coast of Eurasia,are the main distribution of a coniferous broadleaved mixed forest. Over long periods of time, due to various disturbances and damaging events, including logging,Tres,and reclamation,areas of the primary coniferous broad-leaved mixed forests have been sharply reduced.This has led to secondary forests replacing the primary forests (Chen et al. 1994). As a result, the composition,structures, and functions of the plant communities have been considerably changed(Xu et al.2004).In light of this background, distribution patterns of soil Collembola communities and their responses to revegetation in typical temperate coniferous and broad-leaved mixed forests,were examined in this research study. There were two questions addressed: (1) What are the respective composition and distributions patterns of Collembolans in the different stages of vegetation restoration? (2) How do the Collembolans respond to vegetation restoration?

Materials and methods

Study site

This study was conducted in the Changbai Mountains(38°44′D47°30′N, 121°8′D134°E) in the eastern section of northeastern China with the Sanjiang Plain and Liaodong Peninsula at their north and south ends, respectively. The nearly rhombus shaped area is 1300 km from north to south, 400 km from west to east, and the mountain ranges mainly includes Changbai Mountain, Laoye Mountain,Zhangguangcai Mountain, and Hada Mountain. It has a temperate continental monsoon climate with long cold winters, short cool summers, with mean annual temperatures ranging from 3 to 7 °C and mean annual precipitation of approximately 1000 mm.The soil is typical dark brown forest soil, classiTed as Udalfs (Soil Survey Staff 2001).The native zonal vegetation is Pinus koraiensis Siebold and Zucc. coniferous broad-leaved mixed forest. However, the P. koraiensis coniferous broad-leaved mixed forests have been seriously damaged since the early 20th century, and the current zonal vegetation is secondary forests (Wang et al. 2015).

Sampling design

In this study,in order to analyze the distribution patterns of soil Collembola communities and their responses to vegetation restoration in temperate coniferous and broad-leaved mixed forests, Tve habitats were selected. Based on the previous investigation of stand age and logging history,Tve habitats were selected from of 5-year-old (cutover land, COL), 15-year-old (secondary shrub forests, SSF),30-year-old (secondary broad-leaved forests, SBF),80-year-old (secondary coniferous broad-leaved mixed forests, SMF) and primary forest (primary coniferous broad-leaved mixed forests, PMF). The vegetation characteristics of each are shown in Table 1.

Samples were collected in May(spring),July(summer),and September (autumn) of 2015, corresponding to the periods of vegetative growth.Three separate stands of each habitat were randomly chosen on the Changbai Mountains at 5-km intervals. One 20 × 20 m plot was established in each stand and within each plot,Tve replicates of 1 × 1 m subplots were randomly established at 5-m intervals. All plots faced north and had a slope gradient less than 20°.Litter samples(10 × 10 cm)were taken from each subplot to extract litter-living Collembolans. In addition, soil core samples (10 × 10 cm), divided into three layers (0D5,5D10, and 10D15 cm), were collected to extract soil-living Collembolans. These samples were then taken back to the laboratory and extracted for 96 h via Tullgren funnelextractors (Burkard Mfg. Co. Ltd, Hertfordshire, UK). A total of 900 soil Collembolan samples were collected as follows: 5 habitats × 3 replicated stands × 1 plot × 5 subplots × 4 layers × 3 sampling periods. All of the samples were preserved in 75% alcohol. Collembolans were counted using an OLYMPUS SZX16 stereoscopic microscope and an OLYMPUS CX41 biomicroscope(Olympus Co.,Tokyo, Japan),and identiTed at the species level (Christiansen and Bellinger 1998; Dunger 2012;Bellinger et al. 2018). To determine soil temperature and moisture, an Em50 (Decagon Devices Inc., Pullman, WA,USA)was installed in each subplot while the samples were being collected.

Table 1 Vegetation characteristics of different vegetation restoration stages in temperate coniferous and broad-leaved mixed forests

For chemical analysis, additional soil cores(10 × 10 × 15 cm)were sampled with a soil auger beside the sample site for soil Collembolans in each subplot. A total of 225 samples were collected as follows: 5 habitats × 3 replicated stands × 1 plot × 5 subplots × 3 sampling periods. Leaves, roots, and gravel were removed from the soil samples,and each sample air-dried and stored at room temperature. The soil properties were determined by conventional methods (Bao 2000). The pH was determined in distilled-deionized water from the dried samples,and [a 1:2.5 through a pH Meter] PHS-3B (SOIF Inc.,Shanghai, China) was used. The organic matter was digested by K2Cr2O7-H2SO4, and examined using FeSO4titration. At this point, total nitrogen (N) levels were digested by H2SO4and K2SO4-CuSO4·5H2O-Se. Samples were digested by H2SO4and HClO4for total phosphorous(P), and total potassium (K), calcium (Ca), magnesium(Mg), and manganese (Mn) were determined by digesting with HNO3DHClO4DHF. As the pH of all samples was acidic, a Bray method was selected to extract available P.One mol L-1of NH4OAc was used for extracting available K.Following digestion and extraction,the solutions of total N, total P, and available P were determined using a SmartChem140 analyzer (Senesia SRL, Milan, Italy).Samples of total K, Ca, Mg, Mn, and available K were measured with a SpectrAA-220FS atomic absorption spectrophotometer (Varian Inc., Palo Alto, CA, USA).Available N was determined utilizing an alkaline hydrolysis diffusion method. The properties of the soil in the different habitats are detailed in Table 2.

Statistical analysis

To analyze the differences in the composition of soil Collembola communities within the Tve habitats, the endemic genera and species of each habitat were manually determined. A Venn diagram was created using a ?drawquintuple-Venn?function, available in a VennDiagram R package (Chen 2016). To test the effects of vegetation restoration and season on abundance(individuals per m-2)and richness (species numbers) of soil Collembolans, oneway ANOVAs were performed using SPSS 22(SPSS Inc.,Chicago, IL, USA).

The diversity of soil Collembolan communities was quantitatively analyzed by the ShannonDWiener index(H′)as follows (Weaver and Shannon 1949):where S is the number of species,and Pithe ratio of individuals to the total collected individuals in species i for each habitat. This index was calculated from the pooled data of four layers from each sampling site. Rarefaction curves were computed by EstimateS 9.1.0 to compare the respective Collembolan diversity of the Tve habitats during the sampling period(Colwell 2005).

)SE rests (Mean ±fo ixed ed m broad-leav iferous and perate con ration stages of tem n resto etatio ifferent veg roperties in d ble 2 Soil p Ta pH n Total M g-1)(g k g Total M(g kg-1)a Total C g-1)(g k ailable K Avg kg-1)(m ailable P Avg kg-1)(m ailable N Avg kg-1)(m Total K g-1)(g k Total P g-1)(g k tal N To kg-1)(g atter Organic m kg-1)(g Vegetation restoration stages c 0.01 0.01d 0.01e 0.01a 0.01b 0.02a 5.43 ±.32 ±4 ±1 ±0.01b 5.18 ±c 5.7 0.01 d 5 0.01 e 5.7 0.01 8 ±.0 1 ±8 ±c 1 0.05 b 0.7 0.03 0.03b 0.5 2c 0.55 ±0.02a 0.41 ±2 ±± 0.0 e 5.6 5 ±5.67.70 ±0.03.06b 5.8 4c 5.78 ±3d 0.08a 7 6.34 ±± 0 8.49± 0.0 8.26± 0.0 7.907 ±12.6 1.45c 2.07d a 2.19 b 0.26.13e 274.01 ±9.91 ±24 307.77 ±5.43 ±± 1 28 193.57 c 0.066e 8d 0.18a b 0.11.43 ±± 0.0± 0.0 3 ±5 ±137.14 8.94 16.8 14.4 0.86a 5.07b c 3.91 d 2.530e 02.79 ±± 3.6 7 ±c 680.7 3 ±7.03 ±48 211.48 0.06 0.13d 519.6 6d 5a 3 4.76 ±± 0.0± 0.06b 0.01a 1 d 13.97 ±± 0.0 0.01 c 13.77 0.01 c 18.11 0.01 0.01b 16.55 3 ±6a 1.2 0.64 ±.74 ±e 0.02 b 1c 0.68 ±± 0 0.03d 0.67 ±0.02 6.90 5.55± 0.0± 0.0 4.57 3.67 ±3.07 ±b 0.82 a 0.52 5.12 ±± 1.45c 0.35d 0.40e 5.57 ±.35 ±81 73 128.55.02 ±13 15 F PM F SM F SB SSF L CO L cutover CO rests,rests, SSF secondary shrubfo fo broad-leaved BF secondary ixed forests,S A OV AN ay s broad-leaved m ne-w y o e at p ＜0.05 b rest typ fo F secondary coniferou each SM forests,ixed ifferent between ad-leaved m o signiTcantly d F primeval coniferous bro e letter (s) indicate n PM lTahn de sam

An unweighted pair group method with a mathematical mean algorithm (UPGMA) was used to construct a cluster analysis of the similarities in the different Collembolan communities of each habitat.The UPGMA cluster used the ?hclust?function available in the stats R package (R Core Team 2016).A heat map of the hierarchical clustering was created to evaluate distribution patterns of different soil Collembola taxa using a pheatmap R package (Raivo 2015),along with the ?vegemite?function available in the vegan R package (Oksanen et al. 2016).

Multiple factor analysis (MFA) seeks common structures among data matrices. Such a method was applied to the data sets of soil Collembolans,including nutrient levels(organic matter, total N, total P, and total K), available nutrients (available N, available P, and available K), minerals (total Ca, total Mg, and total Mn), and soil temperature and moisture levels to determine relationships among the data sets, and to evaluate the responses of soil Collembolans to vegetation restoration(Lamentowicz et al.2009). Prior to utilizing the MFA, a Hellinger-transformation was applied to the data sets of soil Collembolans and soil properties(Rao 1995;Abdi et al.2013).The MFA was performed using the ?MFA?function available in the FactoMineR R package (Le?et al. 2008). The similarity between the geometrical representations derived from each group of variables was measured by an RV coefTcient from 0 to 1 (Robert and EscouTer 1976). The RV coefTcients were tested using the ?coeffRV?function available in the FactoMineR R package (Josse et al. 2008; Le?et al. 2008).

Results

Taxonomic composition of soil Collembolan communities

A total of 24,225 individuals belonging to 30 species, 26 genera, 12 families, and 3 orders were collected from the Tve habitats during the study period.The dominant species were Proisotoma minima (18.75%), Isotomiella minor(12.42%), and Proisotoma sp.1 (0.98%), and the dominant genera were Proisotoma (29.73%) and Isotomiella(12.42%).Thirteen species were considered to be common and accounted for 51.98% of the total number. Twelve genera(52.94%)were common and in addition,another 14 genera were rare,accounting for 5.87%of the total number of individuals, with ten genera (4.90%) regarded as rare.

Fig. 1 Soil Collembolan community composition (%) in different vegetation restoration stages of temperate coniferous and broadleaved mixed forests. a relative abundance of species; b relative abundance of genera. PMF primary coniferous broad-leaved mixed forests, SMF secondary coniferous broad-leaved mixed forests, SBF secondary broad-leaved forests, SSF secondary shrub forests, COL cutover land

The taxonomic composition of soil Collembolans at the species level from the different periods of revegetation is summarized in Fig. 1a.P.minima was one of the dominant species in the PMF (25.93%), SMF (11.50%), SBF(21.35%), and SSF (16.77%). Proisotoma sp.1 was also one of the dominant in the PMF (11.14%), SMF (18.25%)and SBF (11.34%). Homidia quadrimaculata was a dominant species in the SMF (19.16%) and SBF (12.13%).Ceratophysella baichengensis was a dominant species in the SSF (26.20%) and COL (17.66%). Isotomiella minor(20.48%) was dominant only in the PMF. Parisotoma dichaeta(24.70%)and Protaphorura maoerensis(12.48%)were dominant species only in the COL. Figure 1b details the taxonomic compositions at the genus level.Proisotoma was considered to be the dominant genus in the PMF,SMF,SBF,and SSF,while it was the common genus in the COL.

Venn diagrams with unique and shared taxa are displayed in Fig. 2. All Tve habitats had 19 species and 17 genera in common, which contributed to between 70.37%and 79.17% of the full set of species, and 70.83% to 80.95%of the full set of genera in each habitat.Parisotoma sp.1 and Bourletiella sp.1 were found only in the PMF and SBF; and Pseudosinella sp.1 was recognized only in the SSF and COL. In addition, Friesea sp.1 were the unique taxa in the SBF.

Horizontal distribution characteristics of soil Collembolans

In this study, 27 species (23,153 individuals m-2) of Collembolans were collected in the PMF (primary coniferous broad-leaved mixed forests); 25 species (11,166 individuals m-2) in the SMF (secondary coniferous broadleaved mixed forests); 26 species (8466 individuals m-2)in the SBF (secondary broad-leaved forests); 25 species(6793 individuals m-2) in the SSF (secondary shrub forests); and 24 species (5020 individuals m-2) in the COL(cutover lands). Among these habitats, the PMF displayed the maximum abundance and the minimum was found in the COL.

During the different seasons, the horizontal distribution of soil Collembolan communities was different in the Tve habitats. Relative abundances of soil Collembolans are detailed in Fig. 3aDc and Supplementary material Table S1.During the spring, there were no signiTcant differences among the Tve habitats (p ＞0.05). Numbers of soil Collembolans in the PMF and SMF during the summer months were signiTcantly higher than in other habitats(p ＜0.05). In autumn, abundance levels were as follows:PMF ＞ SMF ＞ SBF ＞ SSF ＞ COL; the levels in the SSF and COL were signiTcantly lower than in other habitats(p ＜0.05). During the study period, with the exception of the PMF and SMF, the dynamics of soil Collembolans numbers were ?spring ＞ summer ＜ autumn?, and levels in autumn were signiTcantly higher than numbers during the summer months in all of the habitats (p ＜0.05).

Fig. 2 Venn diagram of the number of shared and unique Collembolan taxa in different vegetation restoration stages of temperate coniferous and broad-leaved mixed forests. a shared and unique species,b shared and unique Numbers in circles indicate either unique number of taxa or number of shared taxa in the overlap regions.PMF primary coniferous broad-leaved mixed forests, SMF secondary coniferous broad-leaved mixed forests, SBF secondary broad-leaved forests, SSF secondary shrub forests, COL cutover land

The richness of soil Collembolans is summarized in Fig. 3dDf. During the spring months, the richness of soil Collembolans in the SBF (secondary broad-leaved forests)was slightly higher than in the other habitats, whereas, in the PMF (primary coniferous broad-leaved mixed forests),it was slightly lower than the other habitats. During the summer months, the richness was as follows: PMF ＞SMF ＞ SBF ＞ SSF ＞ COL. Among these habitats, SSF(secondary scrub forests) and COL (cutover lands) were signiTcantly lower in Collembolans than in the others habitats (p ＜0.05).In autumn, the highest richness was in the PMF and the lowest in the COL. There were no signiTcantly differences in levels of Collembola among the SMF, SBF, and SSF.

The rarefaction curves indicate that the ShannonD Wiener indexes were different among the Tve habitats(Fig. 3gDi).The curves were plateaus for all of the habitats,which demonstrates that the majority of soil Collembolan taxa were detected. The diversity of soil Collembolans varied during the different stages of vegetation restoration.However, the lowest was consistently found in the COL(cutover lands). In addition, the diversity of soil Collembolans exhibited seasonal variations.With the exception of the PMF, the diversity during the summer months were slightly lower than during spring and summer.

A heat map showed that the 15 sampling sites could be divided into four clusters, indicating that major dissimilarities exist among the levels in the summer PMF,autumn PMF, and autumn SMF (Fig. 4). However, it was also revealed that a major similarity existed among the other sites.Soil Collembolan communities could also be divided into four clusters: Isotomiella minor, Proisotoma sp.1, and P. minima made up one cluster; Bionychiurus changbaiensis, Micronychiurus changbaiensis, and Folsomia candida were a second cluster; Parisotoma dichaeta,Protaphorura maoerensis, Ceratophysella baichengensis,and H. quadrimaculata composed the third; and other species made up the fourth cluster. A greater number of Isotomiella minor, Proisotoma sp.1, and P. minima were found in the PMF during autumn,and also a large count of Homidia quadrimaculata in SMF during the autumn months. In addition, most of soil Collembolan species displayed evident evenness throughout the habitats and seasons.

Vertical distribution characteristics of soil Collembolans

The vertical distribution of soil Collembolans is summarized in Fig. 5aDf.In the PMF,SMF,and SBF habitats,the abundance and richness of the Collembolans within the different soil depths were as:litter layer ＞upper 5 cm soil layer ＞ 5D10 cm soil layer ＞ 10D15 cm soil layer,whereas the trends in the SSF and COL were: upper 5 cm soil layer ＞ litter layer ＞ 5D10 cm soil layer ＞ 10D15 cm.In summary,Collembolans tend to gather on the surface of the soil, and 85.36% were found in the litter layer and the upper 5 cm soil layer.

Fig. 3 Horizontal distribution of soil Collembolans in different vegetation restoration stages of temperate coniferous and broadleaved mixed forests. aDf BoxDWhisker plots illustrating medians(line in the box), 25th and 75th percentiles (box), 10th and 90th percentiles (outer lines), and counts outside the latter percentiles(dots) of abundance (aDc) and richness (dDf) of soil Collembolans in each season.Same letter(s)indicate no signiTcantly different between each forest type at p ＜0.05 by one-way ANOVA. gDi Rarefaction curves of soil Collembolan ShannonDWiener index assemblage in each season. PMF primary coniferous broad-leaved mixed forests,SMF secondary coniferous broad-leaved mixed forests, SBF secondary broad-leaved forests, SSF secondary shrub forests, COL cutover land

Responses of soil Collembolans to the restoration of vegetation

A two-dimensional scatter diagram of the MFA was completed to examine variations of the sampling sites(Fig. 6). The results show the distribution of the sampling sites within the space formed by the Trst two-dimensional axis of the MFA which explains 55.4% of the total variance. The Dimension-1 axis of the MFA explained 32.8%of the variances, and the Dimension-2 axis 22.6%. As illustrated by the scatter diagram, each sample from the PMF, SMF, SBF, SSF, and COL was located along the Dimension-1 axis and therefore the Dimension-1 axis indicated the process of re-vegetation. A positive direction is considered as the advanced stage of vegetation restoration, and a negative direction indicates the initial stage.

Fig. 4 Abundance heatmap of log(x + 1)-normalized soil Collembolans in different vegetation restoration stages of temperate coniferous and broad-leaved mixed forests. Dendrogram of sample sites based on similarity along right axis;Dendrogram of soil Collembolan species based on similarity along upper axis. Colors represent abundance of soil Collembolans. 5 indicate spring; 7 indicate summer; 9 indicate autumn. PMF primary coniferous broad-leaved mixed forests,SMF secondary coniferous broad-leaved mixed forests,SBF secondary broad-leaved forests, SSF secondary shrub forests,COL cutover land

A two-dimensional ordination plot of the MFA was constructed to determine the relationship between the abundance of soil Collembolans and environmental factors(Fig. 7). As illustrated by the ordination plot, Tomocerus,Proisotoma, Ceratophysella, Folsomia, and Parisotoma were positively related to the Dimension-1 axis,indicating that these genera were sensitive to vegetation restoration.Among these,Tomocerus,Proisotoma,and Ceratophysella were towards the positive direction of the Dimension-1 axis, indicating an ability to positively respond to vegetation restoration. However, Folsomia and Parisotoma pointed towards the negative direction were considered to be indicators of the initial stage of the vegetation restoration. In addition, Onychivridae (Bionychiurus, Micronychiurus, and Protaphorura) were located perpendicular to the Dimension-1 axis,indicating that the responses of these Collembolans to vegetation restoration were relatively slow.

There were diverse response patterns to changes in the soil caused by vegetation restoration(Fig. 7).For example,Tomocerus and Proisotoma responded positively to organic matter.Soil temperature changes caused positive responses by Homidia,Lepidiaphanus,Pseudachorutes,Tomocerina,and Protaphorura genera, and negative responses to soil moisture levels. There were positive responses by Folsomia and Parisotoma to the total calcium and magnesium changes, and positive responses by Bionychiurus, Micronychiurus, and Desoria to available phosphorus changes. Isotomiella and Ceratophysella genera responded positively to total phosphorus changes.

Patterns of relationships between Collembolans and the four levels of environment factors were further illustrated by the RV coefTcients of the MFA (Table 3). Available nutrients were most signiTcantly linked to Collembolan communities (RV = 0.527; p = 0.001). In contrast, the correlation coefTcient between all nutrients and soil Collembolans was RV = 0.507, p = 0.001. Additionally,minerals were signiTcantly linked to Collembolan communities (RV = 0.432; p = 0.007). Furthermore, the correlation coefTcient of soil temperature and moisture to Collembolan communities was the lowest (RV = 0.399;p = 0.035).

Fig. 5 Vertical distribution characteristics of soil Collembolans in different vegetation restoration stages of temperate coniferous and broad-leaved mixed forests. Area charts of soil Collembolan abundance (aDc) and richness (dDf) in different layers. PMF primary coniferous broad-leaved mixed forests, SMF secondary coniferous broad-leaved mixed forests,SBF secondary broad-leaved forests,SSF secondary shrub forests, COL cutover land

Discussion

Soil Collembolan taxonomic compositions and distribution patterns

Various Collembolan taxonomic compositions and distribution patterns were found in the different stages of vegetation restoration. The primary coniferous broad-leaved mixed forests (PMF) had the maximum number(23,153 individuals m-2) of Collembolan; the minimum number (5020 individuals m-2) was in the cutover lands(COL). This indicates that vegetation restoration can increase the abundance of soil Collembolans. Previous studies have shown that characteristics of the vegetation communities differ from each other at different stages of the vegetation restoration process (Table 1). Among them,the PMF had the thickest litter layer and herb coverage.Studies have revealed that litter is the crucial factor to increase the richness and abundance of soil Collembolans(Korboulewsky et al. 2016). Consequently, a large amount of litter which provided abundant food resources for Collembolans, promoted the high numbers of Collembolans in the PMF. Additionally, the PMF was characterized by a wetter environment.Nickerl et al.(2013)reported that the skin of Collembolans require respiration,making it difTcult to survive in arid environments. As a result, the wetter conditions in the PMF were a better survival environment for the different species of Collembolans. In contrast, the cutover lands (COL), in the initial stage of vegetation restoration, had low vegetation cover and dryer conditions. Consequently, the minimum levels of Collembolans in the COL was relatively lower than in the other habitats.

Fig. 6 Two-dimensional scatter diagram of the multiple factor analysis (MFA) performed for the different sampling sites in in different vegetation restoration stages of temperate coniferous and broad-leaved mixed forests. The percentages showed next to axis numbers are the total variation of the data explained by that axis. 5 indicate spring; 7 indicate summer; 9 indicate autumn. PMF primary coniferous broad-leaved mixed forests, SMF secondary coniferous broad-leaved mixed forests,SBF secondary broad-leaved forests,SSF secondary shrub forests, COL cutover land

Fig. 7 Two-dimensional ordination plot of the multiple factor analysis (MFA) performed for the whole data set, including environment factors and soil Collembolan data. ST soil temperature, SM soil moisture,SOM soil organic matter,TN Total N,TP Total P,TK Total K, AN available N, AP available P, AK available K, TCa Total Ca,TMg Total Mg, TMn Total Mn. The percentages shown next to the axis numbers are the total variations of the data explained by that axis

Twenty-seven species of soil Collembolans were collected in the PMF;25 species in the SMF;26 species in the SBF; 25 species in the SSF; and 24 species in the COL.Among the different stages of vegetation restoration, P.minima was found to be a dominant species with the exception of the COL. Previous studies reported that P.minima survive best in acidic soils (Ponge et al. 2002). In the COL (cutover lands), pH was signiTcantly less acidic than in the other habitats (Table 2). Therefore, few P.minima were collected. At the same time, similarities existed in the composition of soil Collembolans during the different stages of vegetation restoration and distribution patterns of taxa displayed evident evenness throughout all habitats and seasons. Earlier studies determined that the distribution of Collembolans in the soil depended on environmental factors and biological interactions (Ingimarsdo?ttir et al.2012;Sha et al.2015).The reason for this dependency was that all sites were in the Changbai Mountains, characterized by the same ?ora and fauna(Wang et al. 2015). Consequently, similarity in soil Collembolan distribution was observed in this study.Additionally, the abundance of Collembolan communities displayed some variations over the growing season.For all of the habitats, Collembolan numbers in autumn were signiTcantly higher than during summer (p ＜ 0.05).Yoshida and Hijii(2005)reported that the quality,quantity,and species of plant litter were factors which could affect the distribution of soil Collembolans. In the Changbai Mountains, September is the period of defoliation when a mass of leaf litter is deposited onto the soil surface. As a result, the levels of soil Collembolans increased.

Litter layers had the highest levels abundance and richness of Collembolans in the primary coniferous broadleaved mixed forests (PMF), the secondary coniferous broad-leaved mixed forests (SMF), and the secondary broad-leaved forests (SBF). In contrast, the vertical distribution in the secondary shrub forests (SSF) and cutover lands (COL) was: upper 5-cm layer ＞ litter layer ＞ 5D10 cm layer ＞ 10D15 cm during all seasons. Ponge et al.(2003) reported that most of the epedaphic Collembolans were dependent on surface environments.In this study,the SSF and COL habitats were in initial stages of vegetation restoration with thin, infertile soil layers along with harsh environmental conditions. Consequently, the abundance and richness of the epedaphic or surface-active Collembolans were relatively low. Additionally, 85.36% of soil Collembolans were found in the litter layer and the upper 5-cm soil layer. These results indicate that Collembolans tend to gather on the soil surface, which is in accordancewith results by Kwok and Greenslade (2016) and Rendos? et al. (2016).

Table 3 RV coefTcients(below diagonal)and corresponding p values(above diagonal)among the Tve habitats of variables used in the multiple factor analysis

Responses of soil Collembolans to vegetation restoration

This study found that different genera of soil Collembolans responded to the restoration of vegetative cover in various ways. Tomocerus, Proisotoma, and Ceratophysella responded positively. Gisin (1943) considered that the life forms of soil Collembolans fell into three forms:epedaphic(surface-active),hemiedaphic(within the soil and partly on the surface), and euedaphic (subsurface active). Tomocerus, Proisotoma, and Ceratophysella are examples of epedaphic Collembolans. Previous studies have reported that soil surfaces become progressively wetter and litter layer thickness gradually increase in temperate coniferous and broad-leaved mixed forest habitats as vegetation restoration progressed (Perkins et al. 2015). The wetter conditions and abundant litter could establish a suitable environment for epedaphic Collembolans, thus Tomocerus, Proisotoma, and Ceratophysella respond positively to vegetation restoration. However, Folsomia and Parisotoma responded negatively as these are part of the hemiedaphic Collembolans, intermediate dwellers(Gisin 1943). Research has shown that litter input strongly stimulates rates of litter decomposition and CO2release from soil (Li et al. 2015). Thicker litter layers and greater concentrations of soil CO2may result in low oxygen levels at the intermediate layer of litter and soil. Thus hemiedaphic Collembolans responded to vegetation restoration at negatively. At the same time, this study found that responses of the Onychivridae family (Bionychiurus, Micronychiurus Protaphorura genera) to restoration of vegetation were relatively slow.The life forms of these genera are euedaphic,preferring to live in deep soils(Gisin 1943).Consequently, there were relatively sluggish responses by these Collembolans to vegetation restoration.

In this study, the MFA showed that the abundance of Collembolans correlated most signiTcantly to available soil nutrients.The growth of plants can change soil texture,and may result in considerable organic substances entering belowground ecosystems as root exudates and litter,sequentially improving the soil eco-environment and increasing available nutrients (Hector et al. 2000; Rutigliano et al. 2004). Available nutrients are of vital significance to plant growth and can provide sufTcient food for Collembolans(Cole et al.2005).Consequently, there were relatively stronger responses by soil Collembolans with regards to available nutrients in the study area.At the same time, different genera responded to soil environmental changes brought about by vegetation restoration. The feeding habits of Collembolans differ with each taxon,e.g.,Ceratophysella are predatory Collembolans which feed on nematodes and Isotomiella (Ferlian et al. 2015). Consequently, variations in feeding habits may have led to these different responses. Additionally, Collembolans have signiTcant differences in life history,life forms and nutritional requirements for each taxon(Hopkin 1997),and these may also have contributed to the different responses.

Conclusions

In summary, the restoration of vegetation increased the abundance of soil Collembolans. The taxonomic composition varied with the different stages of vegetation restoration in temperate coniferous and broad-leaved mixed forests,and similarities exist in the soil Collembolan communities. However, there were seasonal variations in the abundance and richness (diversity) levels. Soil Collembolans generally gathered in the litter layers and on the surface of the soil,yet the greatest number was found in the upper 5-cm soil layers in the initial stages of vegetation recovery. Additionally, soil Collembolans correlated most signiTcantly to available soil nutrients,and different genera responded to the restoration of vegetation in various ways.Epedaphic Collembolans responded positively to vegetation restoration; hemiedaphic Collembolans responded relatively negatively while euedaphic Collembolans responded sluggishly.

AcknowledgementsWe express our sincere thanks to Dr. Ernest Bernard(University of Tennessee,Knoxville,USA)for his kind help.At the same time, we would like to thank Dr. Xiaoqiang Li, Dr.Zhenghai Wang,Huiying Han,Hongyue Li,Wenli Xue,Yumei Guo,and Xinchang Kou for their help with Teld work and laboratory analyses.