Introduction

Human-induced climate change is expected to increase the severity and frequency of extreme weather events in the future (IPCC 2013). Some important consequences of this global phenomenon are that extreme weather events, such as prolonged summer droughts and severe floods due to heavy precipitation, become more intense (EEA-JRC-WHO 2008; IPCC 2013; Kerr 2007) and are predicted to occur more frequently (IPCC 2013; Jongman et al. 2014; Stott et al. 2004).

For the belowground compartment of terrestrial ecosystems, the negative ecological effects of floods are reduced oxygen conditions in the soil, which can lead to toxic metabolites due to anaerobic metabolism of microorganisms or plant roots (Plum 2005; Unger et al. 2009). This results in a strong decline in plant and animal biomass and changes in the community composition of soil organisms by favoring fast growing r-strategists over K-strategists that either recover more quickly or are the first to colonize again after local extinction (Plum 2005; Wagner et al. 2015; Wright et al. 2015). As a consequence, processes like nutrient cycling and decomposition can be affected as indicated by changes in the microbial community composition and microbial biomass shortly after flooding and two months later (Wagner et al. 2015). Although several studies indicate that increases of recalcitrant substrates, i.e., dead plant biomass, will benefit fungi over bacteria (Ruess and Ferris 2004), recent studies suggests that both microbial groups may co-dominate as fungi release labile compounds during decomposition, which can also be used by bacteria (de Vries and Caruso 2016). This is in line with findings of Wagner et al. (2015) reporting both fungal and bacterial biomasses to increase after a flooding event promoting nematode r-strategists and a shift to fungal decomposition.

Drought decreases soil moisture and nutrient mobility (Manzoni et al. 2012; Moyano et al. 2013; Schimel et al. 2007), consequently reducing plant biomass and leading to reduced resistance of plant biomass production and carbon sequestration (Tilman and Downing 1994; Pfisterer and Schmid 2002; Ciais et al. 2005; Reichstein et al. 2013). However, several drought experiments found only small or no effects of this disturbance on soil animals, microbial activity, and plant biomass (above- and belowground) arguing for a strong buffering effect of the soil–plant system (Jentsch et al. 2009; Kreyling et al. 2008; Vogel et al. 2013a). Although drought effects seem to be of minor importance, small disturbances like a drought can be more pronounced when directly succeeding another environmental disturbance (Kröel-Dulay et al. 2015).

Previous empirical studies in grasslands have reported both stabilizing (Hautier et al. 2014; Isbell et al. 2015; Tilman and Downing 1994) and destabilizing (Pfisterer and Schmid 2002; Wright et al. 2015) effects of plant species richness on plant biomass production during disturbances, suggesting complex interactions in biodiversity–stability patterns. Different responses depend on the specific types of extreme weather events involved, the focal stability metric, or the degree of disturbance (Isbell et al. 2015). While flooding is often the result of heavy rainfall, drought is usually the result of prolonged periods without rainfall, and different strategies may be necessary for species to cope with such disturbances in terrestrial ecosystems. A recent study investigating flooding in a grassland diversity experiment showed that different levels of flooding severity led to different responses of the plant community (Wright et al. 2015). Low flooding severity increased resource availability and led to increased plant biomass, especially in high diversity plots. In contrast, high flooding severity led to stress reactions due to reduced oxygen content with reduced plant biomass (−63%), and reduced stability (deviation in biomass production from baseline conditions) in higher diversity plots (Wright et al. 2015). Thus, the consequences of different levels of disturbance can vary from positive to negative effects, depending on plant species richness.

Most diversity–stability research has focused on aboveground primary production with little emphasis on responses of soil biota and functions (but see Milcu et al. 2010; Proulx et al. 2010; Wagner et al. 2015). However, soil biota are involved in many ecosystem functions, such as organic matter decomposition, nutrient mineralization, and carbon sequestration (Cragg and Bardgett 2001; Hättenschwiler and Gasser 2005; Wall et al. 2010), which may determine ecosystem responses to climate change (Kreyling et al. 2008). The underlying mechanisms of interactive effects of different co-occurring or consecutive global change agents on soil biota are particularly poorly understood (but see Butenschön et al. 2011; Eisenhauer et al. 2012; Vogel et al. 2013a).

Soil nematodes can be used as indicators of soil food web conditions and soil health (Bongers 1990; Ferris et al. 2001) as nematodes have evolved diverse feeding habits comprising bacterial-feeding, fungal-feeding, plant-feeding, predatory, and omnivorous species (Yeates et al. 1993). Furthermore, they vary substantially in life history strategies ranging from fast growing r-strategists to long-living K-strategists. Differences in life history strategies are accompanied by different sensitivities to disturbance. It is, therefore, possible to link specific components of the nematode community to obtain information about different aspects of community structure and functioning. For instance, the vulnerability to disturbances can be assessed by assuming that an increase of r-strategists and a decrease of K-strategists reflect disturbed conditions (Ferris et al. 2001; Voigt et al. 2007). Furthermore, linked processes like the relative importance of bacterial and fungal decomposition channels, as well as the nutrient status of the system can be assessed by the use of nematodes (Ruess and Ferris 2004).

Here, we took advantage of a large field experiment that manipulated plant species richness in grassland experimental plots (the Jena Experiment; Roscher et al. 2004), which experienced a natural flooding event (three weeks) and a subsequent simulated summer drought sub-experiment (2 months after the flood and lasting for 4.5 weeks).

  1. 1.

    We expect increasing flooding severity to reduce the total density and diversity of nematodes and to alter nematode community composition by favoring r-strategists over K-strategists. Specifically, we assume fungal and bacterial biomasses to increase after flooding with cascading effects to r-strategists and a dominance of fungal decomposition. Furthermore, dead plant and animal biomass is suggested to increase nutrient enrichment (Fig. 1a).

    Fig. 1
    figure 1

    Conceptual figure of the expected responses of the nematode community to plant species richness, flooding severity, and subsequent summer drought. a Increasing flooding severity has negative effects on total nematode diversity (DIV) and alters nematode community composition by favoring r-strategists (r) over K-strategists (K), thereby reducing soil nematode food web structure (SI). Fungal-feeding nematodes (FF) and bacterial-feeding nematodes (BF) are suggested to increase with a stronger dominance of the fungal decomposition channel (CI). In addition, we assume nematodes to indicate increased nutrient enrichment (EI). We further assume significant interactions between flooding severity and plant species richness: b At low flooding severity, we expect a positive relationship between plant species richness and the structure of the soil nematode community comparable to responses in a, where flooding is not yet imposing stress on the system. c At medium flooding severity, higher plant species richness is still able to buffer against negative flood effects, but these effects are less strong. d At high flooding severity, K-strategists decrease strongly (and consequently SI), especially at high plant species richness. In contrast, densities of r-strategists increase. A subsequent summer drought after the natural flood (dashed line) will not affect nematodes at low flooding severity b but in combination with increasing flooding severity, nematodes will be more strongly affected than at ambient conditions, especially at high flooding severity (c, d)

    Full size image
  2. 2.

    We hypothesize plant species richness and flooding severity to interactively affect soil nematodes: at low flooding severity, we expect a positive relationship between plant species richness and the structure of the soil nematode community (e.g., nematode density and diversity, particularly that of K-strategists), because moderately enhanced soil-water availability was shown to substantially increase plant biomass production at high plant species richness (Fig. 1b); at medium flooding severity, we hypothesize that relationships proposed for low flooding severity would be weakened and may become neutral as elevated soil-water availability starts to act as a stressor (Fig. 1c); at high flooding severity, we expect substantial shifts in the soil nematode community due to anoxic soil conditions, mortality, and increased availability of dead organic material (Fig. 1d). More specifically, we assume nematode diversity and K-strategists to decrease, whereas r-strategists would be less sensitive to disturbance and would benefit from increased nutrient enrichment due to mortality of plants and other organisms, especially at high plant species richness.

  3. 3.

    Finally, we hypothesize nematode communities in the highest diversity plots that experience the most severe flooding (Fig. 1d) to be the most vulnerable to the subsequent experimental drought. As plant biomass is reduced most strongly in the high diversity and high flood intensity plots, we expect the strongest reduction in K-strategists and an increase in r-strategists, and therefore, the greatest vulnerability to a subsequent disturbance.

Materials and methods

Study site and experimental design

The study was carried out on the field site of the Jena Experiment (50°55′N, 11°35′E, 130 m above sea level; mean annual temperature 9.9 °C, mean annual precipitation 610 mm; Hoffmann et al. 2014). The field site, formerly a highly fertilized arable field, is located on loamy Eutric Fluvisols in the floodplain of the Saale river, Jena (Thuringia, Germany). Since 2002, 82 grassland plots of different plant communities were established from a pool of 60 mesophilic grassland species of Molinio-Arrhenateretea meadows (typical for the regional alluvial plains). The gradient in plant species richness (1, 2, 4, 8, 16, and 60 species) is combined with a gradient in the number of functional groups (1, 2, 3, and 4 functional groups), namely grasses, small herbs, tall herbs, and legumes, with four replicates per species richness × functional group combination (for details, see Roscher et al. 2004). Plant mixtures were arranged in a randomized block design to account for changes in soil texture with increasing distance to the Saale river, e.g., decreasing sand content away from the river (Roscher et al. 2004). Sand content ranges from 6 to 48%, silt content from 38 to 71%, whereas clay content varies less with 13–25% (Fischer et al. 2015). Plant community compositions in the plots were maintained by a combination of biannual mowing (being a typical management in Central Europe) and weeding.

Natural flooding

From 30 May, 2013 to 24 June, 2013, the Jena Experiment was flooded by a natural flood for 2–25 days. Variations in flooding severity among plots were due to natural differences in elevation across the field site. A flooding index was calculated to account for different levels of flooding severity (Wright et al. 2015). The index included the sum of days flooded and five different levels of water coverage (ratio flooded) ranging from 0 (no water) to 1 (fully submerged) in steps of 0.25 using the following formula:

$${\text{Flooding Index}}\;({\text{FI}}) = \mathop \sum \limits_{{{\text{day}} = 1}}^{{{\text{day}} = 25}} {\text{ratio flooded}}_{\text{day}}$$

The flooding index ranged from 1 to 22.75 (overall mean ± SD: 12.9 ± 4.6). In addition, we divided the flooding index in three categories to represent low, medium, and high flooding severity. The experimental plots were distributed as evenly as possible among the three flooding severity levels with the levels ranging from 1 to 11 (low), 11.25–15.75 (medium), and 16–22.75 (high), with 14, 18, and 12 plots, respectively.

Simulated summer drought

Summer drought was simulated in a nested sub-experiment every year since 2008 using transparent rain shelters excluding all rainfall from July through August (for ~6 weeks). Two rain shelters were used in each of the plots: a roofed subplot without water addition served as “drought” treatment, and a roofed subplot with ambient rain water addition served as “roofed control” treatment to exclude side effects of the roof itself (for more details, see Vogel et al. 2013b). In 2013, the summer drought experiment was delayed due to flooding. Therefore, the drought treatment was applied from 23 August until 23 September (4.5 weeks), three months after the onset of the flood and two months after water had completely retreated from the plots. From 2008 to 2013, overall mean precipitation exclusion during summer drought manipulation was 94 ± 56 mm, being a reduction of 44 ± 8% compared to ambient summer precipitation. In 2013, 50 mm precipitation was removed, which is a reduction by 41% compared to ambient conditions. Summer drought manipulation in 2013 reduced soil water content from 18 ± 3% in ambient subplots to 8 ± 2% in drought subplots in September 2013. In November 2013, i.e., two months after roof removal, differences in soil water content between ambient (21 ± 2%) and drought (20 ± 2%) subplots were no longer detectable.

Nematode sampling, extraction, identification, and indices

Soil sampling was carried out on 25 September, 2013 (2 days after roof removal and 93 days after the flood) and on 5 November, 2013 (43 days after roof removal and 134 days after the flood) in all experimental plots. Three soil cores of 2 cm in diameter were sampled down to a depth of 10 cm and were pooled to form one composite sample per plot. Nematode extraction was carried out using a modified Baermann method (Ruess 1995) from 25 g fresh soil for 72 h. Soil moisture content was estimated after drying at 45 °C for 72 h. All nematodes in the samples were counted to assess their densities. Killing and fixing of living nematodes were carried out with 4% formaldehyde at 65 °C. Nematode identification was carried out in plots with one (14 replicates), four (16 replicates), and 16 (14 replicates) plant species, resulting in a total of 44 plots sampled. A total of 176 samples resulted from 44 plots × 2 sampling campaigns (September and November) × 2 subplots (drought and ambient). A minimum of 100 individuals (if available) were identified per sample, or 10% when the sample exceeded 1000 individuals. Nematodes were identified following Bongers (1994); adults were identified to genus level, while juveniles were assigned to genus or family level (Hodda et al. 2009).

Nematode taxa were then assigned to trophic groups, i.e., plant feeders (PF), bacterial feeders (BF), fungal feeders (FF), predators (PR), and omnivores (OM) (Bongers and Bongers 1998; Yeates et al. 1993). In addition, nematodes were grouped according to their r vs. K life-history strategy using the colonizer (c)–persister (p) scale, which ranges from 1 to 5 (Bongers 1990). Not all trophic groups are represented in all c–p classes. C–p 1 nematodes are exclusively bacterial feeders. They have a short life cycle, an impermeable cuticle, a high reproduction rate, a high tolerance to environmental stress, and are generally assumed to be r-strategists. In contrast, c–p 5 nematodes are only plant-feeders, omnivores and predators, have large bodies, a long life cycle, a permeable cuticle, low reproduction rates, a greater sensitivity to soil disturbances, and are generally assumed to be K-strategists. C–p 2 (including bacterial and fungal feeders, and one genus of a predatory nematode) to c–p 4 nematodes have intermediate characteristics (Bongers 1990; Bongers and Ferris 1999). C–p 1 and c–p 2 nematodes were grouped as r-strategists, and c–p 3 to c–p 5 nematodes as K-strategists in the present study. Finally, nematodes were grouped into functional guilds (Ferris et al. 2001), by combining the feeding habit for each taxon (e.g., bacterial-feeding) with its corresponding c–p value (e.g., c–p 2) resulting in the functional guild BF2. Nematodes with similar biological attributes and inferred function in the soil food web were grouped in the same functional guild (Ferris et al. 2001), which therefore, provides more comprehensive information on the shifts in soil food webs and processes than the trophic structure alone (Cesarz et al. 2015). We calculated absolute (individuals of each taxa 100 g−1 dry soil) and relative (% individuals) densities of the nematode community. An increase in absolute densities reflected an increase in nematode biomass and overall activity, whereas relative changes reflected shifts in the community structure.

The Structure Index was used to assess the structure of the soil nematode food web. Weighted abundances of c–p 3 to c–p 5 nematodes are used to calculate the Structure Index with higher weight given to increasing c–p class. Higher scores indicate a structured and complex food web, suggesting environmental stability and homeostasis, whereas lower values are typical for systems with different degrees of environmental disturbance and a simplified food web (Ferris et al. 2001).

The Enrichment Index reflects the nutrient status of a system. It is based on the assumption that an increase in bacterial and fungal feeders, and especially BF1 nematodes, indicates food web enrichment (high Enrichment Index), mainly caused by disturbance, and therefore, mortality in other trophic levels (Ferris et al. 2001; Neher and Darby 2009).

The Channel Index indicates the dominant decomposition pathway of soil organic matter, based on the percentage of opportunistic fungal-feeding nematodes as a proportion of the total number of opportunistic fungal and bacterial-feeding nematodes in the community. Low values indicate bacterial-dominated decomposition pathway, whereas high values refer to a more fungal-dominated system (Ferris et al. 2001).

In addition, the Shannon diversity index of nematodes was calculated using the BioDiversity Pro software (McAleece et al. 1997) based on the following formula:

$$H^{\prime} = - \mathop \sum \limits_{i} p_{i} \ln (p_{i} )$$

with p representing the proportion of taxon i of the whole community. The term species richness is used for the number of plant species and nematode taxa.

Statistical analysis

We used sequential split–split plot ANOVA with type I sum of squares to analyze the effects of the flooding index [continuous (1–22.75) or categorical (low, medium, high flooding severity)], block (1, 2, 3, 4; accounting for differences in soil texture), plant species richness (1, 4, and 16), plot, subplot (roof treatment at subplot level), drought (ambient subplots and drought subplots), time (September and November), and interactions according to our hypotheses (Fig. 1). The order of the hierarchical analysis was performed as listed above, i.e., flooding index was always fitted first, followed by block, then by plant species richness and so forth. We used two different approaches to analyze the data. If the data had a linear relationship with plant species richness or flooding index, we used a linear regression model with plant species richness (log scale) and flooding index as continuous variables. This is indicated by the numerator degrees of freedom which is always 1 (e.g. F 1,17). In cases where no linear relationship was found, e.g., a humped-shape relationship or a saturating relationship, plant species richness and flooding severity were used as categorical variables indicated by a 2 in the numerator degrees of freedom as both factors have three levels (e.g. F 2,17). Due to reasons of uniformity, results are only displayed as bar graphs.

Summer drought manipulations were treated as subplots and time as sub-subplots in the split–split plot ANOVA (Scheiner and Gurevitch 2001). Treatments analyzed at the main plot level (flooding index, block, plant species richness, and interactions) were tested against the variance among plots to avoid pseudo-replication. Drought and interactions with drought were tested against the variance between subplots, whereas time and interactions with time were tested against the variance of time × subplot (Scheiner and Gurevitch 2001).

Nematode densities were log-transformed and percentage values were arcsine-transformed to achieve homogeneity of variance of the residuals. Indices representing ratios were not log-transformed. Response variables violating the respective assumptions (mainly due to low densities and many zeros) were not analyzed, i.e., single plant feeding groups as assigned by Yeates et al. (1993) and BA4 nematodes. Figures display non-transformed values. Statistical analyses were performed using SAS 9.4 (Statistical Analysis System Institute Inc, Cary, NC).

Non-metric multidimensional scaling (NMDS) was performed to inspect community changes in dependence of flooding severity and plant species richness for both sampling dates separately (September and November) using the whole set of nematode species. The community matrix was analyzed with PERMANOVA using the vegan package in R for two dimensions, and 10,000 iterations using Bray–Curtis distances (R version 3.3.1, R Development Core Team 2016).

Results

The nematode fauna of the Jena Experiment site included a total of 62 nematode taxa (56 genera, 6 taxa identified at the family level, Table S1). The highest nematode densities (mean ± SE: 1644 ± 146 Individuals 100 g−1 dry soil) and the most taxonomically and functionally diverse communities (total species richness: 13.2 ± 0.6; total Shannon Index: 1.90 ± 0.06) were distributed in plots with highest plant species richness. In contrast, the Enrichment Index and the Channel Index were not affected by plant species richness (Fig. S1, 2). Total nematode densities increased significantly from September to November by 91% (F 1,33 = 62.49, P < 0.0001) with the increase being higher for absolute densities of K-strategies (+116%; F 1,23 = 36.53, P < 0.0001) than for r-strategists (+82%; F 1,23 = 15.52, P = 0.0009). Furthermore, time or interactions with time did not have any significant effects, if not specified in the following.

Effects of flooding severity

Increasing flooding severity did not significantly reduce total nematode density (+0.14%; F 2,17 = 1.27, P = 0.3058; Fig. 2a) but nematode community composition was changed. In general, absolute densities of r-strategists (c–p 1 and c–p 2) increased with flooding severity, whereas more sensitive nematodes decreased (Fig. 2a). The increase was marginally significant for relative densities of c–p 2 nematodes (+21%; F 1,27 = 3.56, P = 0.0698). Absolute densities of c–p 3 nematodes (−52%; F 2,20 = 2.88, P = 0.0796; Fig. 2a) as well as relative densities (−42%; F 1,27 = 3.78, P = 0.0625) tended to decrease and absolute densities of c–p 4 (−57%; F 1,27 = 18.35, P = 0.0002) and c–p 5 nematodes (−34%; F 2,20 = 5.68, P = 0.011) decreased significantly with flooding severity (Fig. 2a). Summarizing c–p 3 to c–p 5 nematodes as K-strategists, we found densities of K-strategists to significantly decrease (−53%; F 1,27 = 16.30, P < 0.0001), whereas r-strategists (c–p 1 and c–p 2 nematodes) tended to increase, but this was not significant (+35%; F 2,20 = 1.73, P = 0.2068; Fig. 2b). Likewise, total nematode Shannon diversity was not significantly affected by increasing flooding severity (−9%; F 2,17 = 1.64, P = 0.2241), but Shannon diversity of K-strategists (−17%; F 2,20 = 4.32, P = 0.0304; Fig. 2c) and species richness of K-strategists (−20%; F 2,20 = 7.70, P = 0.0042) were significantly reduced at medium and high flooding severity compared to low flooding severity. In contrast, increasing flooding severity had no significant effect on Shannon diversity of r-strategists (−4%; F 2,20 = 0.17, P = 0.8433; Fig. 2c) and species richness of r-strategists (+4%; F 2,20 = 0.77, P = 0.4767).

Fig. 2
figure 2

Effect of low (l), medium (med./m), and high (h) flooding severity on a absolute nematode density and c–p classes 1-5, b density of nematode r- (i.e., c–p 1 and c–p 2 nematodes) and K-strategists (i.e., c–p 3 to c–p 5 nematodes), c Shannon diversity of r- and K-strategists, d Structure Index, e Enrichment Index, f Channel Index and g density of nematode trophic groups. Bars represent means and error bars standard errors of the mean. Asterisks indicate significant flood effects. The superscript “L” after the significance level indicates that the data was analyzed using flooding severity as linear variable, whereas data in the remaining panels were not linear, and categorical variables were tested with flooding severity as categorical variable. Due to reasons of uniformity, results are only displayed as bar graphs. dws dry weight soil. n in b indicates the number of plots entering the split–split ANOVA and is the same for all panels. ***P < 0.001; **P < 0.01; *P < 0.05; (*) P < 0.1; n.s. P > 0.1

Full size image

Similarly, we found the Structure Index (−38%; F 2,20 = 3.91, P = 0.0401; Fig. 2d) to decrease significantly with flooding severity. This decline in higher trophic levels with increasing flooding severity was neither accompanied by a significant change in nutrient enrichment (+13%; Enrichment Index: F 1,27 = 0.18, P = 0.8366; Fig. 2e) nor a significant change in the decomposition channel (−16%; Channel Index: F 2,20 = 1.63, P = 0.2215; Fig. 2f).

We also detected changes at the trophic group level in response to the flood (Fig. 2g): absolute densities of plant feeders tended to decline from low to high flooding severity by 51% (F 2,20 = 3.42, P = 0.0526), whereas the absolute density of bacterial feeders was highest at high flooding severity, and lower at medium (−40%) and low flooding severity (−42%), but without being significantly affected (F 2,20 = 0.70, P = 0.5103). We further found absolute densities of fungal feeding nematodes not being significantly affected by flooding severity (+42%; F 2,20 = 0.45, P = 0.6444; Fig. 2g). Similarly to bacterial feeders, high variation occurred in fungal-feeding nematodes, as opportunistic FF2 nematodes tended to increase from low to high flooding severity by +42% (F 1,27 = 1.55, P = 0.2243), whereas FF3 nematodes had a humped-shaped relationship with highest densities at medium flooding severity (F 2,17 = 6.54, P = 0.0078). In addition, absolute densities of predators (−72%; F 2,20 = 6.77, P = 0.0057) and omnivores (−25%; F 2,17 = 6.69, P = 0.0072) decreased significantly from low to high flooding severity (Fig. 2g).

Buffering effects of plant species richness

Absolute density (F 2,20 = 2.09, P = 0.1276), Shannon diversity (F 2,20 = 1.42, P = 0.2710), and species richness (F 2,20 = 1.92, P = 0.1534) of r-strategists were not significantly influenced by the interaction of plant species richness and flooding severity (Fig. 3a–c). In contrast, absolute density of K-strategists was highest in 16-species plots at all levels of flooding severity when compared to monocultures and four species plots, but patterns differed among flooding severity levels: in 4-species plots, the density decreased linearly with increasing flooding severity (−59%), whereas in monocultures, densities were lowest at medium flooding severity (−66% compared to low flooding severity); in 16-species plots, densities were highest at medium flooding severity (+113% compared to high flooding severity; F 2,20 = 9.16, P = 0.0004; Fig. 3a). Shannon diversity (F 2,20 = 1.24, P = 0.3302; Fig. 3b) and species richness (F 2,20 = 1.21, P = 0.3475; Fig. 3c) of K-strategists were not affected by the interaction of plant species richness and diversity as both response variables increased similarly with plant species richness at all levels of flooding severity. The Structure Index was marginally significantly affected by the interaction of flooding severity and plant species diversity: it increased with plant species richness at medium and high flooding severity. At low flooding severity, the Structure Index was highest in 4-species plots and lowest in monocultures (F 1,27 = 3.33, P = 0.0791; Fig. 3d). The Enrichment Index (F 2,20 = 0.48, P = 0.6261) and the Channel Index (F 2,20 = 0.76, P = 0.4799) were not significantly affected by the interaction of flooding severity and plant species richness, but the Enrichment Index was highest and the Channel Index lowest in monocultures at high flooding severity (Fig. 3e,f). Absolute densities of plant feeders increased significantly with plant species richness at low (+40%) and medium levels (+297%) of flooding severity compared to high flooding severity. At high flooding severity, however, densities were overall lower and did not differ among plant species richness levels (F 2,20 = 3.62, P = 0.0454; Fig. 3g). Absolute densities of total bacterial feeders (F 2,20 = 0.43, P = 0.6533; Fig. 3h) and fungal feeders (F 2,20 = 0.69, P = 0.5139, Fig. 3i) were not significantly affected by the interaction of plant species richness and flooding severity. However, total predator density tended to be lower in monocultures at all three levels of flooding severity when compared to 4- and 16-plant species plots (F 2,20 = 3.21, P = 0.0619, Fig. 3j). Total absolute omnivore density was highest in plots with 16 plant species at low and medium flooding severity as compared to monocultures and 4-species plots. At high flooding severity, total absolute omnivore densities were highest in plots with four plant species and lowest in monocultures (F 2,20 = 13.52, P = 0.0002; Fig. 3k). Overall, species community composition differed significantly among treatments of flooding severity and plant species richness in September, but no longer in November (Fig. S3). However, ordination did not reveal any clear pattern among treatments.

Fig. 3
figure 3

Effects of low, medium (med.), and high flooding severity (FS) and plant species richness (1, 4, 16 species) on a absolute nematode density, b nematode Shannon diversity, c nematode species richness, d Structure Index, e Enrichment Index, f Channel Index, absolute densities of g plant-feeding nematodes, h bacterial-feeding nematodes, i fungal-feeding nematodes, j predatory nematodes and k omnivorous nematodes. Bars represent means and error bars standard errors of the mean. Asterisks indicate significant interactions between plant species richness and flooding severity. The superscript “L” after the significance level indicates that the data was analyzed using plant species richness and flooding severity as linear variable, whereas data in the remaining panels was not linear, and categorical variables were tested with plant species richness and flooding severity as categorical variable. Due to reasons of uniformity, results are only displayed as bar graphs. dws dry weight soil. n in a indicates the number of plots entering the split–split ANOVA and is the same for all panels. ***P < 0.001; *P < 0.05; (*) P < 0.1; n.s. P > 0.1

Full size image

Interactions with summer drought

Interaction effects of flooding severity and drought affected some of the nematode variables: the Structure Index differed marginally significantly between dry and ambient subplots at low and medium flooding severity (F 2,23 = 2.62, P = 0.0947; Fig. 4a). Surprisingly, at medium flooding severity, the simulated drought increased soil nematode food-web structure slightly compared to ambient conditions. At high flooding severity, however, there were no differences in nematode community structure. The Channel Index did not differ between ambient and dry subplots at low flooding severity, but it was increased by simulated drought at medium and high flooding severity compared to ambient conditions (F 2,20 = 3.89, P = 0.0384; Fig. 4b). Only the species richness of r-strategists was affected by the three-way interaction of plant species richness × flooding severity × drought: generally, differences between ambient and dry subplots were low, but highest richness of r-strategists was found at high flooding severity in monocultures in ambient precipitation subplots, whereas the lowest richness was found in drought subplots (F 2,23 = 3.67, P = 0.0213; Fig. 4c), indicating higher response variability at low plant diversity.

Fig. 4
figure 4

Effects of summer drought (a or amb = ambient; d or dry = drought manipulation), flooding severity (FS) [low, medium (med.) and high] or plant species richness (PRS) on a Structure Index, b Channel Index, and c species richness of r-strategists. Bars represent means and error bars standard errors of the mean. Asterisks indicate significant interactions between all displayed factors in the respective panels. n in a and c indicates the number of plots entering the split–split ANOVA. a and b have the same  n. **P < 0.01; *P < 0.05; (*) P < 0.1; n.s. P > 0.1

Full size image

Discussion

Major flood events are known to have negative effects on plant communities, reducing plant growth and increasing plant mortality, particularly when flooding is severe (Garssen et al. 2015; Wright et al. 2015). Here, we show that flood effects on nematode communities depend on plant diversity and influence the way soil nematode communities respond to a subsequent drought. In particular, flooding decreased the abundance of plant feeders, predators, and omnivores, but increased bacterial feeders and fungal feeders. Flooding also increased the abundance of opportunistic r-strategists relative to K-strategists, mostly due to the loss of sensitive K-strategists in the most severely flooded plots. Overall, these patterns resulted in reduced nematode food-web structure in the most severely flooded plots. Importantly, several of these losses were buffered in higher diversity plant communities (plant feeders, predators, omnivores, and nematode food web structure). Furthermore, the co-occurring effects of a drought may somehow cancel out some of the negative effects of the most severely flooded plots (Structure Index). This may be due to the rapid reduction of physiologically stressful soil waterlogging in the drought conditions. Overall, results of the present study indicate for the first time complex interactions between different consecutive weather events in affecting the structure of soil nematode communities.

Flooding effects on nematode communities

In contrast to our hypothesis (1), we did not find flooding severity to significantly reduce total nematode diversity. However, the diversity, species richness, and density of K-strategists decreased strongly, whereas r-strategists were less affected by the flood. These results are in line with the notion that species should differ in their responses to a disturbance based on their traits and that higher trophic levels are more sensitive to environmental fluctuations than lower trophic levels (Hines et al. 2015; Voigt et al. 2007). Shifts in nematode community composition were likely caused by strong changes in abiotic conditions, loss of living plant biomass, increased availability of dead biomass, and elevated bacterial activity (Wright et al. 2015). In some regions, flooding is used as a management tool to control nematode populations (MacGuidwin 1993). To achieve a strong nematode reduction, however, agricultural fields are flooded for a very long time (several months up to two years) to kill the complete nematode community (Yepsen 1984). This indicates that nematodes must have strategies to overcome stressful conditions, and that some species differ enormously in their sensitivity to stress. One strategy is the formation of dauer larvae, i.e., a metabolically suppressed specialized survival stage. However, only c–p 1 nematodes are able to form dauer larvae, which only increased slightly at the time of sampling in the present study, probably no longer reflecting flooding conditions due to their short generation time. Nematodes of higher c–p classes were shown to enter different stages of dormancy during unfavorable conditions or avoid population reduction by delayed egg hatching (Evans 1987), which may explain why they did not experience stronger declines in this study.

Generally, shifts in soil community composition can alter the functioning of the soil food web (de Vries et al. 2013; Morriën et al. 2017). Regarding this flood event, it is, however, not clear whether these compositional changes are maintained after several years, though reduction of species numbers of K-strategists having generation times of several months points in this direction. By contrast, the fact that the Enrichment Index and the Channel Index did not change with increasing flooding severity suggests no strong changes in the functioning of the soil food web, possibly due to fast recovery of related processes as samples were taken three and five months after the flooding. Taken together, these findings indicate that shifts in nematode community composition in response to flooding events might not translate directly into shifts in soil community function.

Buffering effects of plant species richness on nematode communities after a flood

We found that plant species richness can buffer some variables of belowground nematode communities at each level of flooding severity (in contrast to hypothesis 2). High plant species richness, however, was not able to generally buffer negative effects with increasing flooding severity. For the most severely flooded plots, even high plant species richness could not buffer against the negative effects suggesting a threshold in disturbance severity that can be buffered by plant species richness. Despite negative flood effects, plant species richness was able to maintain higher trophic levels better than monocultures. The underlying mechanisms for these buffering effects of plant species richness depended on the focal nematode group and may be attributed to the living plant or other biotic and abiotic properties changing along the plant species richness gradient. Direct relationships are suggested for plant-feeding nematodes. Plant-feeding nematodes only benefitted at low and medium flooding severity from high plant species richness, whereas at high flooding severity, densities were low and did not differ between monocultures and plots with four plant species. At the study site, the amount of dead plant biomass increased significantly with increasing flooding severity and was highest at high plant species richness (Wright et al. 2015). Plant-feeding nematodes are closely linked to the vigor of their host plants (Ferris and Bongers 2009), and the high amount of dead biomass at high flooding severity and plant species richness suggests that reduced plant performance was accompanied by a decrease in the density of plant feeding nematodes in the present study. Cascading effects on higher trophic levels (predators and omnivores) may be related to the reduction in plant biomass and the resulting decrease in plant feeding nematodes or other organisms relying on plants.

The overall tendency of increased densities of higher trophic levels of nematodes at high plant species richness at each level of flooding severity compared to monocultures may be due to altered habitat conditions in response to plant species richness. For instance, soil aggregate stability increases with increasing plant species richness at the study site and is generally associated with increased water infiltration and hydrophobic characteristics (Pérès et al. 2013). Furthermore, some plants are able to form aerenchyma, and others can increase shoot growth to escape submersion of leaves and may, therefore, increase oxygen content in the rhizosphere as photosynthesis is maintained (Blom and Voesenek 1996; Visser and Voesenek 2005; Wright et al. 2017). This may mitigate reduced oxygen content in soil. Moreover, late-season growing plant species were shown to be less detrimentally affected by the studied flood, as their main growing time was after the flood, and they may have profited most from elevated nutrient availability caused by increased dead plant biomass (Fischer et al. 2016). Those positive effects may be strongest at high plant species richness due to the increased probability of including flood-tolerant plant species in the community, and thereby may drive resistance of the whole system. However, plant-feeding nematodes were not able to profit from these positive conditions suggesting varying plant–nematode interactions among nematode functional groups.

Finally, we expected densities of r-strategists to rise with increased flooding severity and increased plant species richness due to elevated plant mortality leading to higher microbial activity (Wright et al. 2015). Following this, we found a non-significant trend towards increasing total bacterial feeders in higher diversity plots in the most severely flooded conditions, but this seemed to drop in 16-species plots compared to 4-species plots (Fig. 3h). Plant biomass recovered by September 2013 and recovery was highest at high plant species richness (Wright et al. 2015). This is in line with the responses of the Enrichment Index, which suggests no significantly increased nutrient availability due to dead biomass at any level of plant species richness (but a slight trend in this direction). Although the underlying mechanisms are unknown, we speculate that increased competition for nutrients between plants and microorganisms may cascade to microbial-feeding nematodes. Indeed, the recent study of Fischer et al. (2016) at the same study site showed that late-season plants with dense rooting systems and a high nutrient uptake capacity were able to grow right after the flood, thereby using nutrients which may have otherwise promoted bacteria and fungi with possible impacts on the Enrichment Index. Similarly, the Channel Index did not differ between plant species richness and levels of flooding severity suggesting that the shift to increased fungal decomposition observed shortly after the flood (Wagner et al. 2015) was only short-term.

Effects of consecutive environmental perturbations

Only few response variables were affected by interactions with the simulated summer drought. Negative effects of the flood on the soil nematode community were at least marginally reversed by the subsequent drought (Structure Index). This is in contrast to our hypothesis (3) that these consecutive perturbations would be synergistic. We only found richness of c–p 1 nematodes to be highest at high flooding severity in monocultures in ambient plots and lowest in drought plots (Fig. 4c). It is not clear if this reflects a long-term effect of the interaction of flooding severity × drought × plant species richness as c–p 1 nematodes have a very short life cycle of a few days. The lack of synergistic effects of consecutive disturbances is in direct opposition to the findings of Kröel-Dulay et al. (2015), who found that systems recovering from disturbances are more susceptible to additional disturbances caused by drought or warming. Importantly, the identity of the examined disturbances and the comparably short duration of the present natural experiment may be reasons for the differences among studies. While Kröel-Dulay et al. (2015) examined long-term effects (7–14 years) of fire and insect pest outbreaks, we examined fluctuations in water regime. As indicated above, we found interactions of flooding severity and drought suggesting that the two disturbances cancelled out each other at high flooding severity. Therefore, the outcome of consecutive disturbances may depend on the kind of disturbance (drought vs. flooding) and the life history traits of the focal taxa (r- vs. K-strategists).

It is essential to examine whether the present short-term findings also apply to long-term responses and a high frequency of extreme weather events. For instance, Jones et al. (2016) reported changes in the dominance structure of herbs and grasses only after ten years in the framework of a 15-year drought experiment manipulating the frequency and strength of rain events. We, therefore, cannot completely exclude negative effects on ecosystem functioning in the long-term when slow changes in dominance structure of plants and soil organisms or the subtle loss of species occur due to weather events. Disturbances may sometimes exacerbate each other and result in unexpectedly negative consequences. However, some perturbations may drive ecosystems in opposing directions that could potentially cancel each other out, such as in the present study, where the soil waterlogging effects of a flood may have been reversed by the drying effects of the subsequent drought. Thus, in addition to studying interaction effects of co-occurring environmental change effects (Eisenhauer et al. 2012; Tylianakis et al. 2008), future research should also investigate the mostly neglected role of multiple, consecutive disturbances for the composition and functioning of ecological communities.