Marshland restoration benefits Collembola recruitment: a long-term chronosequence study in Sanjiang mire marshland, China

Site description and experimental design

The study was conducted at the Sanjiang Mire Marshland Experimental Station at the Chinese Academy of Sciences (47°35′N, 133°31′E) in the Sanjiang Plain, China (Fig. S1). The station is at an altitude representative of the natural freshwater marshland habitats of the Sanjiang Plain. The study site has a temperate continental monsoon climate with a mean annual air temperature of 2.52 °C, precipitation of 558 mm (more than 65% falls in July and August), and a frost-free period of 125 days (Zhang et al., 2014). The monthly mean temperature was −20 °C in January and 22 °C in July (Song et al., 2009).

Based on the marshland distribution and the present conditions of the farmland, four treatments, including one cultivation treatment, two restoration treatments and one reference, were established. The cultivation treatment (CU15) had been used for soybean planting since 1995 (CU15). The two restoration treatments (RE06 and RE12) were established on soybean fields that had been agriculturally abandoned since 2004 (RE06) and 1998 (RE12), respectively. Undisturbed natural marsh was chosen as intact marshland (IM). The dominant vegetation of the intact marshland and restored marshland both were Calamagrostis angustifolia.

Soil samples were taken on June 15, August 15, and October 15 yearly from 2010 to 2012. A total of 10 replicates were randomly selected from each sample site (50 × 50 m²), and all soil samples (10 × 10 × 10 cm³) were carried in polythene bags to the laboratory. We recognize that these 10 samples per site are not true replicates for the sites, but consider that the pooled sample is sufficiently large to be representative of each treatment. Collembola were then extracted using a Tullgren apparatus (Van Straalen & Rijninks, 1982). The extracted Collembola were preserved in 95% ethanol and identified to species or assigned a morphospecies according to the keys developed by Bellinger, Christiansen & Janssens (1996–2012) and Yin (1992, 1998) using a light microscope.

Analysis of soil physicochemical properties

The soil physicochemical properties were analyzed based on the methods described in Lu (2000). Soil moisture (SM) content was measured by oven drying (105 °C, 48 h). Soil pH was measured in a soil water suspension (1:2.5 w/v) with a pH meter. Soil organic carbon (SOC) was determined by dry combustion using a C/N analyzer (LECO Corporation, St. Joseph, MI, USA). The available nitrogen (AN), available phosphorus (AP), and available potassium (AK) were quantified using the alkaline hydrolysis diffusion method, Bray-1 method, and ammonium acetate extraction method, respectively.

Statistical analysis

We calculated the number of individuals for all identified species and morphospecies for each soil sample. We then used the number of individuals to account for the total amount of captured individuals to classify the quantitative degree of each taxon: the number of individuals that accounted for more than 10% of the total number of captured individuals was dominant species, the number of individuals that accounted for 1.0–10.0% were common species, and the number of individuals that accounted for less than 1% were rare species (Gao et al., 2016). We also explored several parameters to monitor the composition and biodiversity changes in the different wetlands. First, we determined the collembolan density (number of individuals per unit area) and the species richness (number of taxa) for each soil sample. Then we used Shannon-Wiener diversity (H′) (Shannon & Warren, 1949) to evaluate soil collembolan diversity in each habitat.

In order to determine whether marsh cultivation and restoration induced changes in the life-history traits of Collembola, we selected four traits that have previously explained shifts in Collembola species composition (Bokhorst et al., 2012; Makkonen et al., 2011; Vanhee et al., 2017; Widenfalk et al., 2015). The selected traits were body length of adults, life form, reproductive mode, and dispersal traits. Table S1 provides definitions and ecological significance (Berg et al., 1998; Ponge et al., 2006; Rusek, 2007). Trait values (Table S2) were obtained from various literature sources (Chen & Christiansen, 1996; Li, Hua & Chen, 2008; Palaciosvargas & Martinez, 2014; Vanhee et al., 2017; Yun & Gao, 2015). We calculated the community weighted mean (CWM) trait values for each of four traits according to Garnier et al. (2004), weighing species traits in each sample by the relative abundance. For each sample, the CWM trait values were calculated as follows: (1) $$\text{CWM}={\displaystyle \sum _{i=1}^{n}pi}\times \text{T}pi$$

where pi is the relative abundance of collembolan species i, Tpi is trait score of species i, and n is the number of species included in the calculation.

Means and standard errors were calculated for soil parameters and the Collembolan indices of samples from each treatment. One-way analysis of variance (one-way ANOVA) was used to test the difference significance of the soil properties among treatments. In addition, repeated measures ANOVA was used to test for significant differences in density, species richness, Shannon-Wiener diversity, and CWM trait values among different treatments. The significance of post hoc pairwise comparisons were determined using Fisher’s least significant difference (LSD) tests under significance levels of 0.05. In addition, a Spearman correlation analysis was conducted to identify correlations between the soil properties and the response variables of the Collembola, including the abundance of taxa, species richness, and diversity. All statistical analyses were performed using the software package IBM SPSS statistics 22.0 for Windows (IBM SPSS Inc., Armonk, NY, USA).

The Collembola community was further analyzed using the software package CANOCO Version 5.0 for Windows. To test the relationships between Collembola community and recovery years, we performed analyses using redundancy analysis (RDA) based on the species abundance. The soil properties and species abundance were standardized via log transformation before analysis. The significance of the RDA results was determined using a permutation test (499 permutations). We also used an interactive forward-selection RDA (using the “manual selection of environmental variables” option in CANOCO 5.0) to test which variable among the soil properties significantly influenced the Collembola composition. The selection procedure was stopped when the next factor to be added was no longer significant (Xu et al., 2017).

Soil physicochemical properties

All soil properties were significantly different among treatments (Table 1). Generally, SOC, available N, P and K, and SM significantly increased from the cultivated treatment to the 6-year agricultural abandoned, and then 12-year agricultural abandoned treatment (Table 1). By contrast, soil pH value in the intact marshland was significantly lower than the cultivated and agricultural-abandoned treatments (Table 1).

Community composition of collembola

In total, 8,171 individuals and 28 species, with 4,669/25, 5,132/28, 8,539/28 of Collembola (Table S3) in the IM, CU15, RE06 and RE12 fields, respectively, were identified. Five Collembola species—Bourletiella sp. 2, Entomobrya sp. 2, Tomocerus nigrus, Desoria sp. 3, and Folsomia sp. 3—disappeared in the cultivated fields, and all of these species recovered in the restored wetlands (Table S3). Allonychiurus songi was the dominant species at all sites (Table S3). Desoria sp. 1, Desoria sp. 2, Desoria sp. 4, Hypogastrura sp. 1, and Hypogastrura sp. 2 were rare species in the cultivated fields but were common species in the marshlands (Table S3). Entomobrya sp. 1 and Folsomides sp. 1, which were absent in the IM soils, occurred in the cultivated and restored fields. Entomobrya sp. 3, Isotomodes sp. 1, and Tullbergia sp. 1 were absent in the restored fields whereas Hypogastrura sp. 3 only appeared in these fields (Table S3).

The first two RDA axes together explained 29.60% of the variation in the Collembola community (p = 0.032) (Fig. 1). The first axis explained 19.39% of the total variance in species composition and represented mainly SOC, AN, AK, and SM gradient. The second axis which accounted for 10.21% of the total variance was mainly represented by SM and soil pH. The interactive forward selection showed that SOC (p = 0.006) significantly affected soil Collembola structures, explaining 15.7% of the total variation.

Density, richness, and diversity of Collembola

The density and species richness showed a significantly increasing trend in the order of CU15, RE06, IM, and RE12 (Figs. 2A and 2B). The density of the hemi-edaphic Collembola species increased significantly with the length of agricultural abandonment (Table S4). The Shannon-Wiener diversity value was significantly higher in the RE12 than in the IM, RE06, and CU15 soils, whereas no significant variations were found between the IM and RE06 sites (Fig. 2C). The Spearman correlation analysis showed that the total abundance was positively correlated with soil SOC, AN, and AP (Table 2). The epi-edaphic species abundance was positively correlated with soil pH and SM (Table 2). Hemi- and eu-edaphic species abundance were positively correlated with soil SOC, AN, and SM (Table 2). Species richness was positively correlated with soil SOC and AN.

Markedly significant effects on all CWM trait values were affected by different periods of agricultural abandonment. The CWM_BL (F = 14.94, p < 0.001) of the cultivated treatment (CU15) was higher than that of the IM and restored treatment (Fig. 3A). The CWM_RM was highest for the CU15 field and lowest for the RE06 field (F = 7.53 p < 0.001) (Fig. 3B). The CWM_DI was significantly higher for RE06 than for the IM and CU15 fields (F = 6.05, p = 0.002) (Fig. 3C). The CWM_LF for RE06 was significantly higher than that for RE12, and for RE12, this index was higher than for the cultivated treatments and the IM (F = 27.54, p < 0.001) (Fig. 3D).

Recovery of soil properties

In our study, the soil nutrient levels, such as SOC, AN, and AP, decreased after marshland conversion to cultivation and increased after agricultural abandonment. These results were consistent with the findings of Xu et al. (2017). We speculated that two factors potentially contributed to the loss of nutrients after reclamation. One is that the grain and straw were harvested and removed off-site, greatly reducing the input of the organic matter which then decreased. Another reason is that wetland soil was water-saturated over long periods of time, and the organic matter decomposed more slowly and even gradually accumulated because of the anaerobic conditions (Compton & Boone, 2000). Once converted to agriculture, the removal of vegetation for cultivation and the exposure of the soil to the wind and sun increased the aeration and soil temperature, leading to increased soil organic matter decomposition. Besides, tillage generally decreased soil organic matter due to erosion and disruption of the physical, biochemical, and chemical mechanisms of SOM stabilization, and increase soil available C, K, and P leaching (Wei et al., 2014). In contrast, the increase of nutrients found in the restoration treatments could be ascribed to the increased litter input after agricultural abandonment, as the straw and grains were no longer removed from the marshland system. In addition, SOM and available nutrients can generally accumulate after the cessation of cultivation. In short, our findings strongly indicate that the cultivation of marshland led to soil degradation and that agricultural abandonment could improve soil quality.

Recovery of the Collembola community

In line with our first hypothesis, we found the density, species richness, and Shannon-Wiener diversity of Collembola sharply decreased in CU15 fields and increased in the restoration sites. Furthermore, the RDA result showed that the first axis separated the cultivated treatments from the IM and restored treatments, and the IM and restored treatment were separated by the second axis. They all indicated that marshland management significantly affect the community structure of Collembola. For instance, five species of Collembola which disappeared in the cultivated fields all recovered in the restored wetlands; however, the community structure of Collembola in restored marshland were also different from intact marshland. Folsomia bidendata and Tullbergia sp. 1 were more frequent identified in intact marshland soils while Entomobrya sp. 4, Folsomides sp. 2 and Folsomia sp. 2 were more frequent in restored marshland soils. Entomobrya sp. 3, Isotomodes sp. 1 and Tullbergia sp. 1 existed in the intact marshland soils while disappeared in the restored fields. Hypogastrura sp. 3 which was the new introduced species only existed in the restored marshland. We may infer that marshland restoration (after 12 years of agricultural abandonment) can significantly promote the recovery of the species diversity of Collembola; however, it might be difficult to restore to its original state once the composition of the marshland Collembola community has been altered by tillage. That is to say, natural restoration can recruit and reconstruct Collembola community composition.

In the present study, we found that the Collembola community was significantly affected by the SOC. This supports Pipan & Culver (2013) assumption that the SOC content is an important factor limiting the occurrence of organisms in shallow subterranean habitats, as the SOC could provide soil fauna with abundant food resources. In addition, the correlation analysis of the Collembola variables with environmental variables exhibited the significant positive correlation of AN and AP with total abundance and species richness of the Collembola. The reason for this relationship may be the fact that N and P can indirectly influence the habitat and diet of Collembola (Setälä, Marshall & Trofymow, 1995). Compared with RE06, the soil parameters of the RE12 were more similar to those of IM, demonstrating that 12 years was enough time for the species composition or diversity of Collembola to fully recover from a strong disturbance such as long-term soybean cultivation. Thus, in present study, we concluded that the soil Collembola communities were influenced by multiple soil parameters and were most affected by the SOC and nutrients that provide food and habitat for organisms (Birkhofer et al., 2012; Yin et al., 2018).

In support of our second hypothesis, we found higher CWM life forms of Collembola in the restoration sites, which is indicative of a community that has a greater contribution of surface-dwelling species. This is consistent with the observed increasing trend in the density of hemi-edaphic species after agricultural abandonment. This response of Collembola to marshland recovery is likely due to changes in the vegetation and soil properties after agricultural abandonment. Sabais, Scheu & Eisenhauer (2011) found that Collembola density and diversity increased with an increase in plant species and plant functional richness. Vegetation offers ecological links between the aboveground and belowground biota, affecting the diversity and quality of the litter that serves as a habitat and food resource for Collembola (Hopkin, 1997; Wardle, 2013). Particularly, the density of hemi-edaphic species living in upper soil and litter continue to increase and are likely influenced by the increased plant diversity and litter input after agricultural abandonment, as the straw and grains are no longer removed from the ecosystem. In the present study, we found the lowest CWM life form for Collembola in the intact marshland, which indicated the greater contribution of eu-edaphic species. This observation could be explained by the fact that the intact marshland habitat has thicker litter layers, allowing the maintenance of a higher resource quality and more stable conditions in terms of moisture and temperature, which is a preferable environment for soil-dwelling species (Berg & Bengtsson, 2007; Heiniger et al., 2015). Thus, the CWM trait of the life forms at the sites showed a response to the site gradient and soil properties, with a higher proportion of epi-edaphic and hemi-edaphic Collembola species at the recovery sites compared with the intact marshland and soybean fields. Similar results were reported by Da Silva et al. (2016).

In this study, the mean traits of the Collembola community differed remarkably between the cultivated and restoration treatments. We found CWM_BL increased at the farmland sites, which is indicative of a community that the decline in smaller size individuals of collembolan and the community turns to a K-strategy. However, the CWM_BL at the restoration sites declined to similar level as the control sites, it means the body length of collembolan go back to the original levels after marshland restoration. In general, parthenogenetic species were more ready to occupy new territory and became an invasive one, because parthenogenesis is an advantageous trait for colonization (Norton, 1994) which may facilitate the establishment of populations from very few individuals (Lindberg & Bengtsson, 2005). In our study, the peak CWM_RM was found in the CU15 fields and the minimum CWM_RM occurred in the RE06 fields. It may be caused by two reasons. First, the natural disturbance caused by flooding and dry-wet alternate action in marshland was greater than the human-caused disturbance in farmland due to the consistent state of dry field. Second, abundant food resources were conducive to the rapid reproduction of parthenogenetic groups. Compared with farmland, the food resources in marshland were relatively abundant because the litter always exit. So we considered the parthenogenetic taxa recovered faster than sexual individuals. Lindberg & Bengtsson (2006) found a tendency for the more mobile species of arthropods to recover faster than slow moving species. This was also found to be true in our study, where communities in areas undergoing restoration treatments were faster-moving species. Species with a high dispersal ability have an advantage in the colonization of habitats after disturbances (Burrows & Sutton, 2008) and the traits associated with dispersal were selected by land-use change in this prior study. In the present study, we use life history traits to reveal the recovery patterns in Collembola community after different periods of agricultural abandonment. Compared with taxonomic indices, the use of species traits is genetic and reduces environmental dependency, allowing for widely application of monitoring studies (Da Silva et al., 2016; Pey et al., 2014).