Tree Species Identity Shapes Earthworm Communities

Earthworms are key organisms in forest ecosystems because they incorporate organic material into the soil and affect the activity of other soil organisms. Here, we investigated how tree species affect earthworm communities via litter and soil characteristics. In a 36-year old common garden experiment, replicated six times over Denmark, six tree species were planted in blocks: sycamore maple (Acer pseudoplatanus), beech (Fagus sylvatica), ash (Fraxinus excelsior), Norway spruce (Picea abies), pedunculate oak (Quercus robur) and lime (Tilia cordata). We studied the chemical characteristics of soil and foliar litter, and determined the forest floor turnover rate and the density and biomass of the earthworm species occurring in the stands. Tree species significantly affected earthworm communities via leaf litter and/or soil characteristics. Anecic earthworms were abundant under Fraxinus, Acer and Tilia, which is related to calcium-rich litter and low soil acidification. Epigeic earthworms were indifferent to calcium content in leaf litter and were shown to be mainly related to soil moisture content and litter C:P ratios. Almost no earthworms were found in Picea stands, likely because of the combined effects of recalcitrant litter, low pH and low soil moisture content.


Introduction
Earthworms have been studied for a long time (e.g., Darwin [1]), and it is known that soil and forest floor characteristics profoundly affect the composition and abundance of earthworm populations [2][3][4].On the other hand, earthworms are ecosystem engineers [5,6] that can physically, chemically and biologically modify their environment, impacting the habitat and the resources for other organisms [7][8][9] and consequently providing a wide diversity of ecosystem services including facilitation of nutrient cycling and formation of stable humic compounds [6] and mineral soil C sequestration [10,11].Earthworm species all have distinct feeding niches varying from fresh leaf litter to humus and even animal dung, fungal hyphae or soil [12].The simultaneous occurrence of species from different Forests 2017, 8, 85 2 of 20 ecological groups can have synergistic outcomes, and loss of species can lead to significant changes in the ecosystem services they provide [13].
Earthworm species can be classified according to their morphological features, habitat choice and feeding habits into three ecological groups: epigeic, endogeic and anecic earthworms (Figure 1, [14]).First, epigeic earthworms, or 'litter-dwellers', are detrivorous species feeding mainly on fresh or partially decomposed litter on or near the soil surface.They contribute to litter fragmentation, but have too little muscular power to enter the mineral soil.Given the important organic acid production in accumulating forest floors, most epigeic species are tolerant to acid conditions [15].For example, Dendrobaena octaedra can occur in Pinus, Picea or Fagus stands with soil pH H2O of less than 4.0 [16,17].
Secondly, endogeic earthworms, or 'soil-dwellers', are geophagous earthworms feeding on largely humified soil organic matter and dead roots.They are literally eating their way through the shallow soil and, thereby, ingesting large quantities of soil and mixing this with organic material, i.e., bioturbation [12].Most of these species are very sensitive to soil acidification [15,17], for example, the fitness of Aporrectodea caliginosa declined in acid soils (pH H2O < 4.8, [18]).
Thirdly, anecic earthworms, or 'deep-burrowers', are also detrivorous but differ from epigeic earthworms by their ability to create deep vertical burrows in the mineral soil.They can pull leaf litter from the soil surface into their burrows.The anecic species Lumbricus terrestris is known for removing significant quantities of litter from the forest floor [12] while casts are mainly deposited at the soil surface [19].This species can occur in a broad pH-range (pH H2O 4.0-7.2[17,20]) of soil acidity but main occurrence of the anecic species Aporrectodea longa and Lumbricus terrestris is at pH H2O above 4.6 [21].
Forests 2017, 8, 85 2 of 20 Earthworm species all have distinct feeding niches varying from fresh leaf litter to humus and even animal dung, fungal hyphae or soil [12].The simultaneous occurrence of species from different ecological groups can have synergistic outcomes, and loss of species can lead to significant changes in the ecosystem services they provide [13].
Earthworm species can be classified according to their morphological features, habitat choice and feeding habits into three ecological groups: epigeic, endogeic and anecic earthworms (Figure 1, [14]).First, epigeic earthworms, or 'litter-dwellers', are detrivorous species feeding mainly on fresh or partially decomposed litter on or near the soil surface.They contribute to litter fragmentation, but have too little muscular power to enter the mineral soil.Given the important organic acid production in accumulating forest floors, most epigeic species are tolerant to acid conditions [15].For example, Dendrobaena octaedra can occur in Pinus, Picea or Fagus stands with soil pHH2O of less than 4.0 [16,17].
Secondly, endogeic earthworms, or 'soil-dwellers', are geophagous earthworms feeding on largely humified soil organic matter and dead roots.They are literally eating their way through the shallow soil and, thereby, ingesting large quantities of soil and mixing this with organic material, i.e., bioturbation [12].Most of these species are very sensitive to soil acidification [15,17], for example, the fitness of Aporrectodea caliginosa declined in acid soils (pHH2O < 4.8, [18]).
Thirdly, anecic earthworms, or 'deep-burrowers', are also detrivorous but differ from epigeic earthworms by their ability to create deep vertical burrows in the mineral soil.They can pull leaf litter from the soil surface into their burrows.The anecic species Lumbricus terrestris is known for removing significant quantities of litter from the forest floor [12] while casts are mainly deposited at the soil surface [19].This species can occur in a broad pH-range (pHH2O 4.0-7.2[17,20]) of soil acidity but main occurrence of the anecic species Aporrectodea longa and Lumbricus terrestris is at pHH2O above 4.6 [21].In forests, the environmental factors that regulate earthworm communities include litter traits and soil characteristics, mainly clay content, pH, base saturation, soil moisture content, organic matter content and aluminium (Al) toxicity [2,23].Soil pH correlates strongly positively with soil calcium (Ca) and negatively with soil Al concentration, and is among the most important drivers for anecic and endogeic earthworms, i.e., the burrowing earthworm community.A study in multiple Fagus sylvatica stands on a gradient from acid to limestone soils showed that burrowing earthworm densities were most strongly positively linked with the presence of a limestone layer and almost absent from Ca-poor soils [24].
Tree species can significantly modify most of these soil characteristics [25][26][27][28].For instance, N2fixing tree species as Alnus glutinosa with high nitrification rates can acidify the topsoil, which hampers the activity of burrowing earthworms despite the high nutrient content of alder litter [27].It is known that the quality of litter as a substrate for earthworms increases with N, P, K, Ca, and Mg In forests, the environmental factors that regulate earthworm communities include litter traits and soil characteristics, mainly clay content, pH, base saturation, soil moisture content, organic matter content and aluminium (Al) toxicity [2,23].Soil pH correlates strongly positively with soil calcium (Ca) and negatively with soil Al concentration, and is among the most important drivers for anecic and endogeic earthworms, i.e., the burrowing earthworm community.A study in multiple Fagus sylvatica stands on a gradient from acid to limestone soils showed that burrowing earthworm densities were most strongly positively linked with the presence of a limestone layer and almost absent from Ca-poor soils [24].
Tree species can significantly modify most of these soil characteristics [25][26][27][28].For instance, N 2 -fixing tree species as Alnus glutinosa with high nitrification rates can acidify the topsoil, which hampers the activity of burrowing earthworms despite the high nutrient content of alder litter [27].It is known that the quality of litter as a substrate for earthworms increases with N, P, K, Ca, and Mg Forests 2017, 8, 85 3 of 20 concentrations and decreases with increasing lignin, lignin:N, and C:N ratios [29].Tree species can reinforce patterns of soil fertility through positive and/or negative litter feedbacks on earthworm activity and the rate of nutrient cycling [30].For example, several common garden experiments have shown that in just a few decades, tree species with slowly decomposing litter of poor quality for earthworms as Picea abies, Fagus sylvatica and Quercus robur acidify soils and create humus forms very different from those under species with fast decomposing nutrient-rich litter such as Fraxinus excelsior, Acer pseudoplatanus and Tilia cordata [27,31,32].The latter tree species thus lead to a higher earthworm species diversity and total earthworm biomass (4-6 species, 11-37 g•m −2 ) compared to species with more nutrient-poor litter (1-2 species, 0-4 g•m −2 [33]).
Epigeic and anecic earthworm species seem to be mainly affected by litter traits while endogeic earthworm species are affected by soil characteristics [23].However, it seems that these patterns depend on the soil context.For example, it was shown in acidic soils that endogeic earthworms were also significantly linked with several litter quality characteristics such as N, Mg and Ca concentration [34].
Here we investigate the effect of six common European tree species with diverging litter quality on earthworm communities.We used a 36-year-old common garden experiment with sites distributed across Denmark to verify how the identity of tree species results in very different earthworm community assemblages.This replicated common garden experiment has previously revealed tree species effects on C and N stocks in the forest floor and mineral soil [35], soil respiration and soil organic C turnover [36] and on N cycling and leaching and the water budget [37].We aim to broaden the understanding of plant-soil interactions in this common garden experiment by combining existing data on various litter quality characteristics and forest floor turnover rate with new data on soil biogeochemistry and earthworm communities.We expected the tree species to also affect the earthworm communities via altered soil and litter quality characteristics.We also assumed the need to look beyond the response of rough earthworm ecological categories, and rather focus on species-specific responses to varying litter and soil characteristics.We further discuss how our results can be used in the practice of forestry and in light of ongoing global changes.

Common Garden
This study was carried out in a 36-year-old common garden experiment replicated in six sites in Denmark (Appendix A).Two out of six study sites (Kragelund and Mattrup) were former agricultural land while the other sites were previously forested with Fagus sylvatica since the beginning of the 19 th century.Each site was planted with 0.25 ha adjacent unreplicated monoculture stands of six common European tree species: Acer pseudoplatanus L., Fagus sylvatica L., Fraxinus excelsior L., Picea abies (L.) H. Karst., Quercus robur L. and Tilia cordata Mill., except for Vallø, where the Fraxinus stand establishment failed due to deer browsing.For further details on the setup of this common garden experiment, we refer to Table 1 in [35].

Soil Sampling and Analyses
In each stand, we randomly selected three representative plots of 0.25 m 2 (total number of plots = 105) where we assessed earthworm density and biomass in October 2009 with a minimum distance of 10 m from the stand border.In each plot, three soil cores were combined into one composite soil sample for each of three depths below the forest floor (0-0.05m, 0.05-0.15m and 0.15-0.3m).Soil moisture content was measured gravimetrically by the weight difference before and after drying to constant weight at 40 • C.After sieving (2 mm sieve), soil pH-KCl was determined in a 1:5 soil/KCl solution (1 M) with a glass electrode (Ross Sure-flow 8172).The exchangeable concentrations of K + , Na + , Mg 2+ , Ca 2+ and Al 3+ were measured by flame atomic absorption spectrophotometry using BaCl 2 (0.1 M) as extractant (ISO 11260).Effective cation exchange capacity (CEC) was calculated as the sum of K + , Na + , Mg 2+ , Ca 2+ and Al 3+ , expressed in meq•kg −1 .The effective base saturation (BS) was calculated by dividing the sum of the base cations (K + , Na + , Mg 2+ and Ca 2+ ) by the CEC.

Sampling and Analyses of Litter and Forest Floor
Litterfall and forest floor sampling was described in detail in Vesterdal et al. [35] and Vesterdal et al. [36].Briefly, litterfall was collected monthly using ten circular littertraps with a diameter of 31 cm installed along two line transects.Litterfall was sampled in all species for one full year at two sites (Mattrup and Vallø) and at the remaining four sites the broadleaf species litterfall was sampled in the autumn to gauge the discrete foliar litterfall event.Annual litterfall amounts for the broadleaves were estimated by proportional upscaling using the annual litterfall amounts measured at Mattrup and Vallø.Litterfall was dried at 55 • C and hand-sorted and weighed in two fractions: foliar and non-foliar litter.Forest floor was sampled in 15 points along three line transects within each stand using a 25 cm × 25 cm wooden frame in September 2004 just before the onset of foliar litterfall for deciduous species, when forest floor mass was at a minimum.Foliar fractions of the forest floor were dried to constant weight at 55 • C before weighing.In the further analyses, we have used one value per stand for these litter and forest floor traits (n = 35).

Earthworm Sampling and Identification
In each plot, we used a combined sampling method, ensuring effective sampling of different ecological groups of earthworm species.First, the litter-dwelling specimen was collected by hand-sorting of a litter sample within a 0.25 m 2 frame; then, the deep-burrowing specimen was captured at the soil surface by application of a mustard solution (60 g mustard powder in 30 L water, [38,39]) within the same 0.25 m 2 frame, and 30 min after mustard extraction, the surface soil dwelling specimen was collected by hand-sorting of a soil core taken in the center of the frame (0.09 m 2 with a 0.2 m depth).The earthworms were, for each of the three methods separately, collected in pots containing ethanol (95%) and after a few hours transferred to a 5% formalin solution for fixation.After three days, they were transferred back to a 95% ethanol solution for further preservation and identification.All adult earthworms were identified with the key of Sims and Gerard [15] and were subsequently categorized into three ecological groups as defined by Bouché [40] (epigeic, endogeic or anecic species).In case of individuals with missing heads or juveniles, identification at species level was not possible.We identified these individuals to genus or ecological group level and assigned them pro rata to species.There are disadvantages to this prorating method [41] but since only about half of the individuals per plot were adult, it was important to include these juvenile and unidentifiable individuals.All worms were weighed individually, including gut contents, after briefly being dried on filter paper at room temperature.We did not correct earthworm biomass for potential bias caused by the gut contents.The density and biomass of earthworms sampled by the hand-sorting method was converted to an area of 0.25 m 2 .Then, for each of the plots, we calculated the area-weighted sum of the number and biomass of earthworms from the three sampling methods per m 2 .The number of earthworms per m 2 will hereafter be referred to as the earthworm density.

Handling of Missing Data
The foliar litterfall data for Picea stands in Kragelund, Odsherred, Viemose and Wedellsborg was missing.Instead, we used the mean values for the mass, C, N, P, Ca, Mn, lignin from Mattrup and Vallø based on evidence of limited site effects on litterfall mass across Denmark [42].

Forest Floor Turnover Rate
The forest floor turnover rate (k) for the broadleaved tree species was calculated according to the following equation for discrete litterfall events [43] because these tree species mainly shed their litter from September to November [35,42]: where k is the annual decay rate; DM litterfall is the average annual foliar litterfall measured, and DM Forest floor is the accumulated foliar litter layer on the forest floor.
For Picea, a tree species with continuous litterfall through the year [42], k was calculated by a different equation [43]:

Data Analyses
All statistical analyses were performed with R [44].

Tree Species Effect
The effect of tree species identity on the litter (n = 35) and soil variables (n = 105) was analyzed using linear multilevel models with site as a group-level effect to account for the spatial dependence of the measurements (e.g., in R syntax response variable~Treespecies-1, random= ~1|Site; lme; package nlme [45]).Differences between tree species were tested with Tukey post-hoc tests (p < 0.05) using the function glht from the multcomp package [46].Model fit was verified by examination of the residuals; we log-transformed the response variables in case it improved the fit ('soil K', 'soil Na', 'soil Mg', 'soil Ca', 'soil Al', 'litter Mn' and 'forest floor mass').
Due to the high frequency of zero's in the earthworm response variables (25%-35% for ecological earthworm groups and more for individual species), we used generalized linear multilevel models that allowed for zero-inflation (glmmADMB; package glmmADMB [47]).A negative binomial response distribution was assumed to allow for overdispersed earthworm count data [48][49][50].Model syntax was similar and differences between tree species were also tested with Tukey post-hoc tests in the method that was described previously.

Links between Litter Quality, Soil Quality and Earthworms
To explore the particular soil and litter variables underlying the overall tree species effect on earthworms (n = 105, models above), in models explaining various earthworm responses (the number of species, total, epigeic, endogeic and anecic earthworm density and biomass) we performed model selection with several soil and litter variables as predictors and site as a group-level effect (again glmmADMB allowing for zero-inflation).The included predictors, standardized by centering to the mean and rescaling by their standard deviation, were 'soil moisture content, soil pH 0-5 cm, soil Al 0-5 cm, soil Na 0-5 cm, litter Ca, litter N, litter Mn, litter P, litter lignin, litter C:N ratio, litter C:P ratio, forest floor turnover rate and the interaction between soil Al 0-5 cm and litter Ca'.We have also included a measure for the possible interaction between litter quality and soil quality affecting earthworm communities.Leaf litter Ca concentration appears to be an important litter quality trait affecting earthworms [26,27,31].Further, pH is an important soil quality trait for earthworm activity [23].Since exchangeable Al concentrations are strongly linked with pH-KCl and probably affect earthworm activity (indications of its toxicity for earthworms in [51]) we also included exchangeable Al to our models.Because exchangeable Al concentration seemed to be more important than pH-KCl in our models, we selected exchangeable Al concentration as a measure for soil quality in the interaction (Soil-Al:Litter-Ca).
To check for multicollinearity, we calculated the variance inflation factors of the standardized variables in a full model.None of the variables had a variance inflation factor (VIF) >3 and therefore, we could include these terms in our model selection [52].
We performed manual stepwise forward selection of predictor terms starting with null models where only site as a group-level effect was included.We added each of the predictors, calculated the Akaike information criterion corrected for small sample sizes (AIC c ) [53] using AICctab (package bbmle) and continued with the model with the smallest AIC c value.AIC c values, ∆-AIC c values and Akaike weights of the null, intermediate and final optimal models are shown in Appendix B. Predictor variables were added until the AIC c was minimized and, then, we evaluated the significance of the included variables by the Wald test (p < 0.05).If these models contained only significant variables, they were retained as the final optimal models.If these models contained non-significant variables, we went back one step in the selection process.We assessed goodness of fit of the optimal models by calculating a measure for R 2 describing the correlation between fitted and observed values by the r2.corr.merfunction, and we further evaluated plots of residuals and fitted versus observed data.We plotted the coefficients from these models with coefplot2 (coefplot2 package).

The Tree Species Effect on Soil Properties and Litter Quality
After 36 years of forestation, tree species consistently influenced all measured topsoil properties significantly across the six sites (Table 1).According to the pH (0-5 cm), the tree species could be ranked as: Fraxinus = Acer = Tilia >> Quercus = Fagus >> Picea.Picea had significantly lower soil moisture content, pH, exchangeable base cation concentrations and higher Al and Na concentrations.The tree species affected not only the topsoil (0-5 cm), as Picea stands also had significantly increased Na and Al concentrations in the deeper soil layers (5-15 cm and 15-30 cm; Appendix C) compared to the other tree species.
Table 1.Mean and standard deviation of topsoil (0-5 cm) properties for each tree species across all six common gardens.Significant differences between tree species are indicated with letters, means with the same letter are not significantly different (Tukey post-hoc tests on linear mixed-effects (LME) models, 1|Site).All measured foliar litter property variables were significantly influenced by the tree species (Table 2).Ca, Mg and K concentrations in leaf litter of Fraxinus and Acer were at least double the concentrations in leaf litter of Fagus and Picea.Lignin leaf litter concentrations were significantly higher in Picea, Tilia, Quercus and Fagus (ranging from 25% to 29%) compared to Fraxinus and Acer (on average 18%).Leaf litter of Fraxinus and Acer had significantly lower C:N ratios compared to Quercus, Picea and Fagus.Also, Mn and P in leaf litter varied significantly across tree species, with Fraxinus containing significantly lower Mn and higher P concentrations compared to the other species.
Also, the forest floor masses and turnover rates were significantly differing between the observed tree species, with Picea having the lowest forest floor turnover rate and the highest forest floor mass, and Fraxinus, Acer and Tilia stands having the highest forest floor turnover rates and lowest forest floor masses.Quercus and Fagus were found to be intermediate.Table 2. Mean and standard deviation of annual litterfall, foliar litter quality and forest floor accumulation for the tree species across all six common gardens.Significant differences between tree species are indicated with letters, means with the same letter are not significantly different (Tukey post-hoc tests on LME models, 1|Site).Litterfall and forest floor foliar mass and nutrient concentrations in litter were previously published by Vesterdal et al. [35,36].

Table 3.
Collected earthworm species, their ecological group according to Sims and Gerard [15], their incidence at the studied sites (K = Kragelund, M = Mattrup, O = Odsherred, V = Vallø, Vi = Viemose, W = Wedellsborg) and tree species (Fr = Fraxinus, Ac = Acer, Ti = Tilia, Qu = Quercus, Fa = Fagus, Pi = Picea).Total earthworm density in the studied plots ranged from zero to 428 individuals m −2 and total earthworm biomass amounted to maximally 202 g•m −2 .Figure 2 shows the tree species-specific responses in earthworm communities.In Picea stands, earthworm populations were absent from half of the plots, and across all sites, significantly, they contained the lowest total earthworm densities and biomasses of all tree species.Quercus and Fagus stands contained, on average, an intermediate total earthworm density and biomass.The highest earthworm densities and biomasses were found in Fraxinus and Acer stands and Fraxinus and Tilia stands, respectively.
Considering the three ecological groups, we found Fraxinus, Acer and Tilia stands to have higher densities and biomasses than Picea stands for all ecological groups (Figure 2).Epigeic earthworm biomass was significantly higher in Quercus stands compared to Picea and Fagus stands.Density and biomass of endogeic earthworms were significantly higher in Acer stands than in Picea stands, while anecic earthworm density and biomass were significantly higher in Fraxinus stands compared to Quercus, Fagus and Picea stands.

Links between Soil Quality, Litter Quality and Earthworm Communities
The earthworm groups were significantly linked with various soil and leaf litter characteristics (Tables 4 and 5; Figure 3).The best model fit was found for total and anecic earthworm density and biomass data.The models on endogeic and epigeic earthworm densities and biomasses showed a low increase in R 2 (6%-14%) compared to the null model.In case of endogeic earthworm densities and biomasses, the final R 2 was very low (21% and 28%) while these models contained seven selected predictor terms.
Table 4. Summary of the results identifying optimal models for total, anecic, endogeic and epigeic density and biomass.We used glmmADMB zero inflated models with Site as random effect, degrees of freedom (df), Akaike information criterion corrected for small sample sizes (AICc) and R 2 describing the correlation between fitted and observed values are shown.Soil predictor variables were measured in the topsoil (0-5 cm).Soil-Al:Litter-Ca is the interaction between exchangeable soil Al concentration and litter Ca concentration.The foliar litter nutrient concentrations were previously published by Vesterdal et al. [35,36].

Response Variable
Predictor  .The tree species were sorted according to decreasing soil pH-KCl (0-5 cm).Significant differences between tree species within each earthworm group are indicated with letters (Tukey post-hoc tests on generalized linear multilevel (glmmADMB) models, 1|Site).

Links between Soil Quality, Litter Quality and Earthworm Communities
The earthworm groups were significantly linked with various soil and leaf litter characteristics (Tables 4 and 5; Figure 3).The best model fit was found for total and anecic earthworm density and biomass data.The models on endogeic and epigeic earthworm densities and biomasses showed a low increase in R 2 (6%-14%) compared to the null model.In case of endogeic earthworm densities and biomasses, the final R 2 was very low (21% and 28%) while these models contained seven selected predictor terms.Table 4. Summary of the results identifying optimal models for total, anecic, endogeic and epigeic density and biomass.We used glmmADMB zero inflated models with Site as random effect, degrees of freedom (df), Akaike information criterion corrected for small sample sizes (AIC c ) and R 2 describing the correlation between fitted and observed values are shown.Soil predictor variables were measured in the topsoil (0-5 cm).Soil-Al:Litter-Ca is the interaction between exchangeable soil Al concentration and litter Ca concentration.The foliar litter nutrient concentrations were previously published by Vesterdal et al. [35,36].

Response Variable
Predictor  Table 5. Summary of the results identifying optimal models for total, anecic, endogeic and epigeic density and biomass.We used glmmADMB zero inflated models with Site as random effect.The foliar litter nutrient concentrations were previously published by Vesterdal et al. [35,36].Predictor variables were standardized to the mean and scales by the standard deviation, so that effects correspond to a change in the earthworm density/biomass (on the log-scale) for a one standard deviation change in the predictor.This is true for all predictors, so their relative effects are comparable.The models are shown for total (black), anecic (blue), endogeic (green) and epigeic (red) earthworms.Coefficients are shown with confidence intervals by thick lines for 50% credible intervals and by thin lines for 95% credible intervals.We used glmmADMB zero inflated models with Site as group-level effect.The foliar litter nutrient concentrations were previously published by Vesterdal et al. [35,36].
Although the explanatory variables differed between the ecological earthworm groups, exchangeable soil Al and leaf litter Ca concentrations were almost always present in the final models for burrowing earthworms (i.e., anecic and endogeic species).Since pH-KCl seems to be better known Predictor variables were standardized to the mean and scales by the standard deviation, so that effects correspond to a change in the earthworm density/biomass (on the log-scale) for a one standard deviation change in the predictor.This is true for all predictors, so their relative effects are comparable.The models are shown for total (black), anecic (blue), endogeic (green) and epigeic (red) earthworms.Coefficients are shown with confidence intervals by thick lines for 50% credible intervals and by thin lines for 95% credible intervals.We used glmmADMB zero inflated models with Site as group-level effect.The foliar litter nutrient concentrations were previously published by Vesterdal et al. [35,36].
Although the explanatory variables differed between the ecological earthworm groups, exchangeable soil Al and leaf litter Ca concentrations were almost always present in the final models for burrowing earthworms (i.e., anecic and endogeic species).Since pH-KCl seems to be better known as a soil variable than exchangeable Al concentrations, we provide more insight into exchangeable Al in Figure 4a.Exchangeable Al concentrations are strongly inversely related with pH-KCl.Burrowing earthworms related significantly negatively with soil Al concentrations, and were almost absent when soil pH-KCl decreased below a value of 4.0 (Figure 5a,b).Epigeic earthworms showed an optimum between pH-KCl 3.5 and 4.0 (Figure 5c).
Litter Ca concentrations were positively linked with forest floor turnover rates (Figure 4b) and burrowing earthworms (Figure 5d,e).Anecic and endogeic earthworm densities were generally not higher than 50 earthworms m −2 in stands with low litter Ca concentrations (<13 mg•g −1 ).Litter Ca concentration was not an explanatory variable for epigeic earthworms, which seems to be indifferent for this variable (Figure 5f).The interaction soil-Al:Litter-Ca represents a measure for the possible interaction between litter quality and soil quality affecting earthworm communities.The effect of litter Ca showed a more positive effect on anecic earthworm density when soil Al concentrations were low.In contrast, this Burrowing earthworms related significantly negatively with soil Al concentrations, and were almost absent when soil pH-KCl decreased below a value of 4.0 (Figure 5a,b).Epigeic earthworms showed an optimum between pH-KCl 3.5 and 4.0 (Figure 5c).
Litter Ca concentrations were positively linked with forest floor turnover rates (Figure 4b) and burrowing earthworms (Figure 5d,e).Anecic and endogeic earthworm densities were generally not higher than 50 earthworms m −2 in stands with low litter Ca concentrations (<13 mg•g −1 ).Litter Ca concentration was not an explanatory variable for epigeic earthworms, which seems to be indifferent for this variable (Figure 5f).Burrowing earthworms related significantly negatively with soil Al concentrations, and were almost absent when soil pH-KCl decreased below a value of 4.0 (Figure 5a,b).Epigeic earthworms showed an optimum between pH-KCl 3.5 and 4.0 (Figure 5c).
Litter Ca concentrations were positively linked with forest floor turnover rates (Figure 4b) and burrowing earthworms (Figure 5d,e).Anecic and endogeic earthworm densities were generally not higher than 50 earthworms m −2 in stands with low litter Ca concentrations (<13 mg•g −1 ).Litter Ca concentration was not an explanatory variable for epigeic earthworms, which seems to be indifferent for this variable (Figure 5f).The interaction soil-Al:Litter-Ca represents a measure for the possible interaction between litter quality and soil quality affecting earthworm communities.The effect of litter Ca showed a more positive effect on anecic earthworm density when soil Al concentrations were low.In contrast, this The interaction soil-Al:Litter-Ca represents a measure for the possible interaction between litter quality and soil quality affecting earthworm communities.The effect of litter Ca showed a more positive effect on anecic earthworm density when soil Al concentrations were low.In contrast, this interaction shows a different pattern for endogeic density and biomass: the effect of litter Ca was more positive when soil Al concentrations were high.However, this effect was minor, as the models for endogeic earthworms had a low goodness of fit with the data.
Anecic earthworms were further significantly positively related with other litter quality variables such as N concentrations and C:P ratios.Additionally, lignin concentrations and C:N ratios were explanatory factors for the endogeic earthworms, although the fits were not very good.Finally, litter Mn concentrations and forest floor turnover rates were significant predictors for epigeic earthworms.Furthermore, both endogeic and epigeic earthworms were significantly positively linked with the soil moisture content.
To test whether different earthworm species of an ecological group had the same sensitivity towards soil Al and litter Ca concentrations, we selected the two most common earthworm species of each group (Appendix D).For the anecic species, A. longa was only scarcely present when soil Al concentrations were higher than 50 mg•g −1 (and pH-KCl was below 4.2), while L. terrestris appeared to be abundantly present when litter with high Ca concentration was available.A. longa was the only earthworm species significantly associated with Fraxinus, Acer and Tilia (Appendix E).Similarly, but less pronounced was the difference between the two endogeic species: A. rosea seemed to be slightly more sensitive to high Al concentrations than A. caliginosa (Appendix D).The epigeic species, clearly present at a different pH-optimum (at pH-KCl 3.5-4.0),did not show a difference in sensitivity towards soil Al or litter Ca concentrations.

Discussion and Conclusions
Within less than four decades, the observed tree species established specific diverging soil conditions and significantly affected the earthworm communities.Topsoils under Picea, Fagus and Quercus appeared to be, on average, four times more acid than under Tilia, Acer and Fraxinus, and this was independent of the soil type, land use history and climate.From literature, we already knew that intrinsic differences in leaf litter quality among tree species fundamentally create different soil conditions and nutrient cycling, both directly through the chemical composition of the litter, and indirectly through its effects on the size and composition of especially burrowing (endogeic and anecic) earthworm communities [25,27,31].Tree species, such as Picea, Fagus and Quercus, with Ca-poor leaf litter, contribute to the absence of burrowing earthworm communities, which retards litter decomposition and results in forest floor build-up and high concentrations of exchangeable soil aluminium, which in turn negatively impacts on earthworm communities.Tree species with Ca-rich leaf litter such as Fraxinus, Acer and Tilia appeared to have abundant anecic and endogeic earthworm populations, lowest forest floor masses, highest forest floor turnover rates and highest pH values with lowest exchangeable Al concentrations.Other studies have reported good correlations between forest floor decomposition rates and earthworm populations, because of the positive influence of leaf litter quality on earthworm populations [26,27,31,33,54].The fact that exchangeable Al concentrations are negatively influencing earthworm populations is less published.
When soils acidify below pH H2O of 5.0, base cations, such as Ca, are removed from the cation exchange complex and are replaced by Al [55].The increase of exchangeable soil Al concentrations also takes place in the soil solution [55], which is toxic for earthworms [51].High exchangeable Al concentrations and low pH values were found to inhibit earthworm growth and cocoon production [51].Earthworms are negatively impacted by soil acidification and Ca-poor litter [27].The absence of an abundant burrowing earthworm population decreases bioturbation and increases the build-up of a forest floor, which in turn delays cations becoming available again for buffering the proton input.This chain reaction and the complex interactions were well explained in the conceptual model in Figure A1 by De Schrijver et al. [27].In our study, burrowing earthworm communities (endogeic and anecic species) appeared to be abundant when exchangeable soil Al concentrations were lower than 100 mg•Al•g −1 , and soil pH-KCl values were higher than about 4. Further soil acidification with exchangeable soil Al concentrations above 100 mg•Al•g −1 was associated with a complete absence of burrowing earthworms.However, favorable litter quality might compensate for unfavorable soil conditions.Cesarz et al. [34] showed that an acid soil (pH-H 2 O 3.7-4.5,and probably high exchangeable Al concentrations) combined with Ca-rich leaf litter of Tilia, Acer or Fraxinus resulted in viable endogeic earthworm populations, while the combination of an acid soil with Ca-poor leaf litter of Fagus resulted in earthworm mortality.
In our study, we found a significant interaction between exchangeable Al concentrations and leaf litter Ca concentrations for anecic earthworm densities.The effect of high leaf litter Ca was more positive at low exchangeable Al concentrations.For endogeic earthworm density and biomass, the interaction was the other way around, but very weak.The significance of the interaction in our models indicates context-specificity by these plant-soil interactions: the effects of leaf litter Ca concentration on burrowing earthworms cannot be extrapolated from any one site (e.g., [31]) to other sites with different soil properties.
According to our models, epigeic earthworms were not significantly affected by both high soil Al concentrations or leaf litter Ca concentrations.Epigeic earthworm biomass appeared to be negatively linked with the C:P ratio in litter, which is in accordance with the findings of De Wandeler et al. ( 2016), and positively to soil moisture content and leaf litter Mn concentrations, which is most bioavailable in the pH-KCl-range of 3.4-4.1 [56].
Picea stands proved to be exceptionally unfavorable for earthworms from all three ecological groups.Next to an acidified topsoil and recalcitrant litter, having negative impacts on burrowing earthworm populations, these stands were also characterized by significantly lower soil moisture contents compared to the other tree species.Also, Christiansen et al. [56] showed consistently decreased soil moisture content in Picea stands compared to Acer, Tilia, Fagus and Quercus stands.In these drier conditions, epigeic earthworms, which were positively linked with soil moisture content according to our model, were almost absent.Epigeic species can endure short drought periods by producing drought-resistant cocoons, but when drought periods take too long they can go extinct locally [57].The lower soil moisture content under Picea can contribute to low forest floor decomposition rates through its direct negative impact on the activity of soil biota [28].
Further, we found that soil salinization had taken place in the topsoil and deeper soil layers (15-30 cm) of Picea stands.Exchangeable soil Na concentrations were up to three times higher in the topsoil of Picea stands and were mainly increased in the sites enduring more westerly oceanic winds.These sites (Odsherred, Wedellsborg and Viemose) could be more influenced by marine sea salt deposition and Picea stands, known for their high atmospheric dry depositions [58] could, therefore, also capture more Na.Our models showed that anecic species were negatively related with soil Na concentration.In literature, the sensitivity of earthworms towards soil salinity was already reported for the endogeic A. caliginosa and the epigeic Eisenia fetida [59], but not for anecic species.
We combined existing data on various leaf litter quality characteristics and forest floor turnover rates with new data on soil biogeochemistry and earthworm communities to broaden the understanding of plant-soil interactions in this well-studied common garden experiment [35][36][37].We found that total earthworm biomass in Fraxinus, Acer and Tilia stands was, on average, two times higher than in Quercus and Fagus stands and eight to ten times higher than in Picea stands.A total of 50% of the earthworm biomass in Fraxinus, Acer and Tilia stands was made up of two anecic species, namely L. terrestris and A. longa, which might explain the higher carbon stocks at 15-30 cm soil depth found by Vesterdal et al. [35].Similar vertical distributions in soil C stock were found in stands of tree species from the same genera in other places, e.g., North America [10], suggesting specific tree species-soil interactions mediated by macrofauna species such as earthworms.
Within the endogeic and anecic ecological groups, a differentiation in sensitivity of earthworm species exists towards leaf litter quality (here illustrated by Ca concentration) and soil quality (here illustrated by exchangeable Al concentration).Because certain earthworm species can be bio-indicators for biological soil quality [21] and for forest site quality [17], studying the earthworm populations at species level can reveal greater detail in response to environmental conditions.The anecic A. longa and endogeic A. rosea appeared to be more sensitive to soil acidification (pH-KCl <4.0) than L. terrestris and A. caliginosa.Also, from our indicator species analyses, A. longa was associated with Fraxinus, Acer and Tilia stands.This is in accordance with the finding that A. longa is a sensitive species towards pH that is closely associated with crop-and grasslands [60] and eutrophic deciduous forests [21].A. rosea, however, was described previously as a species without a clear preference for a certain habitat type [21] and a tolerance for a broad pH-range [17,20].In accordance with literature [21], the anecic L. terrestris and endogeic A. caliginosa appear to be species with a broad ecological niche because they were also found in soils with pH-KCl between 3.5 and 4.0.However, they seemed to appear in higher numbers when leaf litter contained more Ca.So, it seemed that the effects of high Al concentrations were mediated by nutrient-rich litter [34] for these two species.These findings imply the need to look beyond the response of rough earthworm ecological groups, and also focus on species-specific responses to varying leaf litter and soil characteristics.
In conclusion, we have shown that ecological earthworm groups are highly influenced by the tree species via several leaf litter and/or soil characteristics, but not all groups and species are affected similarly.According to Millennium Ecosystem Assessment [61], climatic change and atmospheric deposition of reactive N, are two of the major drivers of biodiversity loss in forests.Future climate change may come in the form of higher summer temperatures and/or increased droughts in temperate European regions [62].On the other hand, atmospheric N deposition is expected to rise even further in temperate European regions and cause acidification [61].Climate change may even worsen the effects of acidification by air pollution [63].Future studies of plant-soil interactions should consider how the magnitude of litter impacts on soil organisms and soil processes might depend on how bedrock, climate, and atmospheric pollution have influenced soil acidification.Our results have shown that endogeic and epigeic earthworms were sensitive to drought, and endogeic and anecic earthworms were sensitive to acidification.In our study, we saw that planting Picea on these soils that are prone to acidification, resulted in an almost complete eradication of the earthworm population.To avoid the loss of earthworm biodiversity and functioning, foresters can mitigate these expected global changes by a substantiated choice of tree species.

Appendix B
Table A2.Details of the results identifying optimal models for total, anecic, endogeic and epigeic density and biomass.We used glmmADMB zero inflated models with Site as random effect.Soil predictor variables were measured in the topsoil (0-5 cm).Soil-Al:Litter-Ca is the interaction between exchangeable soil Al concentration and litter Ca concentration.The foliar litter nutrient concentrations were previously published by Vesterdal et al. [35,36].

Response Variable
Predictor

Figure 1 .
Figure 1.Illustration of the three ecological earthworm groups according to their food preference and behavior.The brown zone represents the soil.Reproduced from [22].

Figure 1 .
Figure 1.Illustration of the three ecological earthworm groups according to their food preference and behavior.The brown zone represents the soil.Reproduced from [22].

Figure 3 .
Figure 3.Estimated effects of litter and soil (0-5 cm) variables on the earthworm density (a) and biomass (b).Predictor variables were standardized to the mean and scales by the standard deviation, so that effects correspond to a change in the earthworm density/biomass (on the log-scale) for a one standard deviation change in the predictor.This is true for all predictors, so their relative effects are comparable.The models are shown for total (black), anecic (blue), endogeic (green) and epigeic (red) earthworms.Coefficients are shown with confidence intervals by thick lines for 50% credible intervals and by thin lines for 95% credible intervals.We used glmmADMB zero inflated models with Site as group-level effect.The foliar litter nutrient concentrations were previously published by Vesterdal et al.[35,36].

Figure 3 .
Figure 3.Estimated effects of litter and soil (0-5 cm) variables on the earthworm density (a) and biomass (b).Predictor variables were standardized to the mean and scales by the standard deviation, so that effects correspond to a change in the earthworm density/biomass (on the log-scale) for a one standard deviation change in the predictor.This is true for all predictors, so their relative effects are comparable.The models are shown for total (black), anecic (blue), endogeic (green) and epigeic (red) earthworms.Coefficients are shown with confidence intervals by thick lines for 50% credible intervals and by thin lines for 95% credible intervals.We used glmmADMB zero inflated models with Site as group-level effect.The foliar litter nutrient concentrations were previously published by Vesterdal et al.[35,36].
as a soil variable than exchangeable Al concentrations, we provide more insight into exchangeable Al in Figure4a.Exchangeable Al concentrations are strongly inversely related with pH-KCl.

Figure 4 .
Figure 4. Relation between pH-KCl and exchangeable Al concentration in the 0-5 cm soil layer ((a) n = 105).In (b), the relation between Ca concentration in foliar litter and the forest floor turnover rate (n = 35) is shown.The points are colored according to the tree species.The foliar litter Ca concentration was previously published by Vesterdal et al. [36].

Figure 5 .
Figure 5. Relation between exchangeable Al concentration in the 0-5 cm soil layer (a-c) or Ca concentration in foliar litter (d-f), and density of anecic (a,d), endogeic (b,e), and epigeic (c,f) earthworms.Plots where zero earthworms were found are indicated by a cross symbol.The foliar litter Ca concentration was previously published by Vesterdal et al. [36].

Figure 4 .
Figure 4. Relation between pH-KCl and exchangeable Al concentration in the 0-5 cm soil layer ((a) n = 105).In (b), the relation between Ca concentration in foliar litter and the forest floor turnover rate (n = 35) is shown.The points are colored according to the tree species.The foliar litter Ca concentration was previously published by Vesterdal et al. [36].

Forests
variable than exchangeable Al concentrations, we provide more insight into exchangeable Al in Figure4a.Exchangeable Al concentrations are strongly inversely related with pH-KCl.

Figure 4 .
Figure 4. Relation between pH-KCl and exchangeable Al concentration in the 0-5 cm soil layer ((a) n = 105).In (b), the relation between Ca concentration in foliar litter and the forest floor turnover rate (n = 35) is shown.The points are colored according to the tree species.The foliar litter Ca concentration was previously published by Vesterdal et al. [36].

Figure 5 .
Figure 5. Relation between exchangeable Al concentration in the 0-5 cm soil layer (a-c) or Ca concentration in foliar litter (d-f), and density of anecic (a,d), endogeic (b,e), and epigeic (c,f) earthworms.Plots where zero earthworms were found are indicated by a cross symbol.The foliar litter Ca concentration was previously published by Vesterdal et al. [36].

Figure 5 .
Figure 5. Relation between exchangeable Al concentration in the 0-5 cm soil layer (a-c) or Ca concentration in foliar litter (d-f), and density of anecic (a,d), endogeic (b,e), and epigeic (c,f) earthworms.Plots where zero earthworms were found are indicated by a cross symbol.The foliar litter Ca concentration was previously published by Vesterdal et al. [36].

Table 5 .
[35,36] of the results identifying optimal models for total, anecic, endogeic and epigeic density and biomass.We used glmmADMB zero inflated models with Site as random effect.The foliar litter nutrient concentrations were previously published by Vesterdal et al.[35,36].