Plant diversity enhances production and downward transport of biodegradable dissolved organic matter

Plant diversity is an important driver of below‐ground ecosystem functions, such as root growth, soil organic matter (SOM) storage and microbial metabolism, mainly by influencing the interactions between plant roots and soil. Dissolved organic matter (DOM), as the most mobile form of SOM, plays a crucial role for a multitude of soil processes that are central for ecosystem functioning. Thus, DOM is likely to be an important mediator of plant diversity effects on soil processes. However, the relationships between plant diversity and DOM have not been studied so far. We investigated the mechanisms underlying plant diversity effects on concentrations of DOM using continuous soil water sampling across 6 years and 62 plant communities in a long‐term grassland biodiversity experiment in Jena, Germany. Furthermore, we investigated plant diversity effects on the molecular properties of DOM in a subset of the samples. Although DOM concentrations were highly variable over the course of the year with highest concentrations in summer and autumn, we found that DOM concentrations consistently increased with plant diversity across seasons. The positive plant diversity effect on DOM concentrations was mainly mediated by increased microbial activity and newly sequestered carbon in topsoil. However, the effect of soil microbial activity on DOM concentrations differed between seasons, indicating DOM consumption in winter and spring, and DOM production in summer and autumn. Furthermore, we found increased contents of small and easily decomposable DOM molecules reaching deeper soil layers with high plant diversity. Synthesis. Our findings suggest that plant diversity enhances the continuous downward transport of DOM in multiple ways. On the one hand, higher plant diversity results in higher DOM concentrations, on the other hand, this DOM is less degraded. This study indicates, for the first time, that higher plant diversity enhances the downward transport of dissolved molecules that likely stimulate soil development in deeper layers and therefore increase soil fertility.


| INTRODUC TI ON
The loss of biodiversity (Barnosky et al., 2011) has severe consequences for multiple ecosystem functions, such as reduced plant biomass production, soil organic matter (SOM) storage and soil nutrient provision (Balvanera et al., 2006;Hooper et al., 2012). Among other factors, the functioning of terrestrial ecosystems depends on the materials and energy supplied by plants, the retention of nutrients in the soil and the provision of nutrients to plants by microbial mineralisation of organic matter to a great extent (Bardgett & van der Putten, 2014;Wagg et al., 2014). Plant diversity has been demonstrated to impact plantsoil interactions via the release of increased quantities or more diverse rhizodeposits in the soil Lange et al., 2019) or by enhancing access of the microbial community to the rhizodeposits (Mellado-Vazquez et al., 2016). In turn, with higher plant diversity more active soil microbial communities mineralise higher quantities of organic matter (Eisenhauer et al., 2010;Zak et al., 2003), and thereby provide more plant-available nutrients (Hacker et al., 2015;Lange et al., 2019;Oelmann et al., 2011). These plant-soil interactions mostly occur in the liquid phase of soil (Schimel & Weintraub, 2003) and are mediated by dissolved organic molecules, which are detected in dissolved organic matter (DOM). However, despite this important function, the role of DOM in the relationship between plant diversity and soil functions has scarcely been addressed.
In a previous study in the Jena Experiment, a positive long-term effect of plant species richness on annual average concentrations of dissolved organic carbon (henceforth termed DOM concentrations) below the densest rooting zone of the top 30 cm was found . However, this increase due to plant diversity was only observed several years after the experiment had been established. The time lag of the plant diversity effect was mainly attributed to a gradual change in soil conditions, as plant diversity-dependent accumulation of organic matter in the topsoil takes time .
A similar trend of enhanced accumulation of SOM in high-diversity communities was also reported from other biodiversity experiments in grasslands (Cong et al., 2014;Fornara & Tilman, 2008) and forest (Li et al., 2019). However, the mechanisms underlying the plant diversity effect on DOM are not yet understood.
Generally, DOM is generated by decomposition of dead organic material or exudation of organic substances by roots (Evans et al., 2005). The quality and quantity of root inputs as well as the soil carbon stocks depend on plant community composition and diversity (De Deyn et al., 2011;Eisenhauer et al., 2017;Hooper et al., 2000;Lange et al., 2019;Steinbeiss, Bessler, et al., 2008).
Thus, changes in plant diversity are likely to impact DOM dynamics.
In the upper soil, DOM is mostly plant derived. However, plant-derived DOM compounds rapidly decline with depth (Klotzbücher et al., 2016;Scheibe et al., 2012), as DOM is subject to microbial uptake and transformation, as well as to sorption processes during its downward movement through the soil profile (Kaiser & Kalbitz, 2012;Roth et al., 2019). At the same time, soil microbial community composition and metabolic activity are strongly influenced by plant diversity (Eisenhauer et al., 2010;Lange et al., 2014;Schmid et al., 2019;Zak et al., 2003). However, the microbial impact on DOM may be contradictory. Besides the consumption of DOM and the accompanied decline in DOM concentration, increased microbial activity can also enhance mineralisation of SOM, thereby increasing DOM concentrations due to more residues of solubilised organic molecules (discussed in Kalbitz et al., 2000). Increases or decreases in DOM concentrations can indicate whether the microbial community acts as a sink or source of DOM (Neff & Asner, 2001).
Furthermore, microbial cycling of DOM impacts both its concentration and molecular properties (Roth et al., 2019). As a result of the microbial cycling, the more processed DOM is assumed to be both less bioavailable and less biodegradable for further microbial processing (Don et al., 2013;Marschner & Kalbitz, 2003). This has plant diversity effect on DOM concentrations was mainly mediated by increased microbial activity and newly sequestered carbon in topsoil. However, the effect of soil microbial activity on DOM concentrations differed between seasons, indicating DOM consumption in winter and spring, and DOM production in summer and autumn. Furthermore, we found increased contents of small and easily decomposable DOM molecules reaching deeper soil layers with high plant diversity.

4.
Synthesis. Our findings suggest that plant diversity enhances the continuous downward transport of DOM in multiple ways. On the one hand, higher plant diversity results in higher DOM concentrations, on the other hand, this DOM is less degraded. This study indicates, for the first time, that higher plant diversity enhances the downward transport of dissolved molecules that likely stimulate soil development in deeper layers and therefore increase soil fertility.

K E Y W O R D S
biodiversity, decomposition, dissolved organic carbon, ecosystem functions and services, plant-soil interactions, subsoil, vegetation consequences for the microbial communities subsequently exposed to the DOM during its downward transport in soil (Leinemann et al., 2018). However, the effect of plant diversity on the microbial DOM production and consumption patterns as well as on the molecular properties of DOM is unclear.
Furthermore, the interplay between soil microorganisms and DOM is highly sensitive to environmental conditions, such as seasonal fluctuations in temperature and soil moisture (Kalbitz et al., 2000).
In particular, DOM concentrations are subject to strong seasonal variations with the highest concentrations in late summer (Don & Schulze, 2008). Moreover, a shift in plant-soil interactions during the course of the growing season has been reported (Eisenhauer et al., 2018), suggesting that the first half of the growing season is presumably dominated by plant inputs of rhizodeposits, whereas the second half is dominated by decomposition of more slowly decomposable dead plant residues (Kuzyakov, 2002 Here, we investigated how plant diversity influences DOM concentrations in the Jena Experiment, and whether or not plant diversity impacts the molecular properties of DOM. To explore the mechanisms underlying potential plant diversity effects, we tested the impact of known drivers, such as below-ground productivity, microbial activity, SOM content, soil texture and hydrological conditions, on DOM concentrations. We hypothesise that plant diversity increases DOM concentrations, reflecting higher biological activity, and that this is mainly driven by greater below-ground root inputs and accumulated SOM that stimulate the microbial community activity. Therefore, during the growing season, DOM concentrations are highest and the effects of plant diversity on it are strongest. We further hypothesise that plant diversity effects will be reflected in molecular DOM data. In particular, we expect an increase in plant diversity to be reflected by an increase in the amount of plant-derived inputs with a low level of decomposition.

| Study site -The Jena Experiment
The study was carried out in the Jena Experiment, a large-scale grassland diversity experiment on the floodplain of the Saale River near the city of Jena (Thuringia, Germany; 50°57′N, 11°35′E; Roscher et al., 2004;Weisser et al., 2017). The soil of the field site is classified as Eutric Fluvisol. In spring 2002, 82 experimental grassland plots of 20 × 20 m were established. Plots are arranged in four blocks to account for changes in soil characteristics with increasing distance from the river. Soil texture in the upper 30 cm of the soil ranges from sandy loam to silty clay with increasing distance from the river. Sand content declines from 50% in the plots close to the river to 5% in plots that are furthest away from the river, while silt content increases from 35% to 70% respectively. Similar to silt, the clay content increased with increasing distance from the river from 15% to 25%. During the 40 years prior to establishing the experiment, the field site was an arable field with mineral fertiliser input.
The initial physico-chemical properties, such as pH (7.1-8.4), SOM content (5-33 g C/kg) and soil nitrogen concentrations (1.0-2.7 g N/ kg), varied across the field site and are considered in the block design of the Jena Experiment.
Plant communities were assembled along gradients of plant species richness (PSR; 1, 2, 4, 8, 16 and 60 species) and functional group richness (1, 2, 3, 4) with species randomly chosen from a pool of 60 Arrhenatherion grassland plant species and from the functional groups grasses, legumes, small herbs and tall herbs. The sown plant functional compositions along the PSR gradient can be found in Appendix S1. Functional group classification was based on morphological, phenological and physiological traits (Roscher et al., 2004).
Experimental plots are weeded manually two to three times a year to maintain the target plant community composition. The plots are mown and the mown plant material is removed twice a year in June and September, but not fertilised, which is typical for extensively used hay meadows in Central Europe.

| Soil water sampling and analysis
In April 2002, glass suction plates (pore size 1-1.6 μm, 1 cm thickness, 12 cm in diameter; from UMS GmbH), were installed in three of the four blocks (N = 62 plots) at a depth of 30 cm to collect soil water. In 2005, glass suction plates were additionally installed at a depth of 20 cm, so that soil water was sampled at two different depths on all 62 plots. The sampling bottles were continuously evacuated using a negative pressure of between 50 and 350 mbar, so that the suction pressure was approximately 50 mbar above the actual soil water tension. As a consequence, only the soil leachate was collected (Scheffer & Schachtschabel, 2002). Cumulative soil water was sampled fortnightly, and DOM concentrations were assessed by analysing the dissolved organic carbon concentration with a high TOC elemental analyser (Elementar Analysensysteme GmbH).
All samples were stored at 4°C until measurements and analysed as soon as possible, i.e. within 2 weeks after sampling.  . In both sampling campaigns, three soil samples were taken per plot (4.8 cm in diameter, 0-30 cm depth) using a split-tube sampler (Eijkelkamp Agrisearch Equipment). The cores were separated into 5 cm layers according to depth in the field and afterwards pooled per layer. The samples were dried (40°C), sieved (2 mm mesh) and milled (6 min, frequency of 30 s −1 ). The soil organic carbon content of the calcareous soil around Jena was assessed by first determining the total carbon concentration of ground samples with an elemental analyser after combustion at 1,150°C (varioMax CN elemental analyzer, Elementar Analysensysteme GmbH). Then, inorganic carbon concentration was measured by elemental analysis after removing organic carbon for 16 hr at 450°C in a muffle furnace by oxidation. Finally, soil organic carbon (henceforth termed SOM) concentration was calculated from the difference between total and inorganic carbon concentrations (Steinbeiss, Bessler, et al., 2008).

| Root and shoot standing biomass
Root standing biomass was sampled on all plots in 2011 and 2014 (for details see Ravenek et al., 2014). Three soil cores (3.5 cm) were taken per plot to a depth of 30 cm and pooled before root washing.
Root biomass was calculated as g dry mass per m 2 .

| Soil microbial activity
Annual measurements of basal respiration on soil from all plots in late May or early June (Strecker et al., 2016) were used as a proxy for soil microbial activity. For this, five soil samples per plot were taken at a depth of 5 cm, pooled and shortly stored at 5°C until basal respiration measurement. The samples were homogenised, sieved (2 mm) to remove larger roots, animals and stones and adjusted to a gravimetric soil water content of 25%. After adapting the soil to the measurement temperature of 22°C for 5 days, basal respiration was measured on c. 5 g of fresh soil (equivalent to c. 3.5 g soil dry weight) using an O 2 micro-compensation apparatus (Scheu, 1992).
The microbial respiratory response was measured at hourly intervals for 24 hr at 22°C. Basal respiration (µl O 2 h −1 g soil dry mass −1 ) was determined without any addition of substrate and measured as the mean of the O 2 consumption rates 14 to 24 hr after the start of the measurements (for details see Eisenhauer et al., 2010). In order to avoid measuring the Birch effect (Birch, 1964) i.e. a pulse in microbial activity following a disturbance and rewetting, we only considered the data after 14 hr when the respiration rates had stabilised.
The basal respiration data used in this study were previously published (Strecker et al., 2016), and data from 2011 were used in Lange based on a more extensive sampling infrastructure in block 2. Before measurement, soil water subsamples were acidified to pH 2 (HCl, p.a.) and stored at 2°C until DOM was concentrated and desalted by solid phase extraction (SPE; Dittmar et al., 2008) using Agilent Bond Elute PPL SPE cartridges (1 g). SPE-DOM is a subset of the dissolved DOM which includes the most apolar DOM species through to highly polar molecules, but not the smallest polar molecules such as short chain organic acids and free amino acids (Hawkes et al., 2016). Due all masses were excluded that were only detected in one measurement and had no reliable signal-to-noise ratio of the maximum of each m/z value (s/n Max,i ) (for details see Pohlabeln & Dittmar, 2015 and references therein). Only singly charged ions were considered (Koch & Dittmar, 2006). Consequently, the m/z values represent the molecular mass (in dalton) of the detected ions. To study specific formula-based characteristics, the weighted mean of m/z (m/z wm ) of hydrogen-to-carbon ratios (H/C wm ) and of oxygen-to-carbon ratios (O/C wm ) were calculated. Therefore, each measurement as the sum of the product of the individual information (m/z i , H/C i , O/C i ) and relative intensity I i was divided by the sum of all intensities i.e.: The m/z wm represents each samples' weighted mean of the detected molecular mass. The m/z, H/C (gives information on the saturation) and O/C (gives information on the oxygenation) can be used to describe the molecular properties of DOM (e.g. Roth et al., 2019). Furthermore, based on DOM compounds, an index of degradation (I deg ) was calculated.  (Flerus et al., 2012): where a lower I deg indicates less degradation of DOM. Using linear regressions, Flerus et al. (2012) identified single mass peaks that were positively and negatively correlated to 14 C of SPE-DOM measurements from 117 oceanic samples. Thereby, the 14 C gives the bulk age of the samples and is thus an indication of the degradation state of DOM. Although this index was developed using aquatic samples, Roth et al. (2019) showed that I deg works remarkably well on soil samples.

| Statistical analyses
For statistical analyses, the fortnightly sampling data of DOM concentration were averaged and soil water volume summed for each plot among seasons. Seasons are defined as calendric quarters of the year, namely the first quarter (Q1, Jan-Mar) is winter, Q2 (Apr-Jun) is spring, Q3 (Jul-Sep) is summer and Q4 (Oct-Dec) is autumn.
Aggregating data by seasons allowed to account for the seasonal dynamics in DOM concentrations and soil water volume as well as to reduce the high variability among sampling dates. All statistical analyses were conducted with the statistical software R (version 3.3.0; R Development Core Team, 2016).

| Main effects of plant diversity over time
Linear mixed-effects models (LMM) applying the 'lme'-function in the R library 'nlme' (Pinheiro et al., 2016) were used to test for plant diversity effects on DOM concentrations (see Appendix S2 for the full analysis model). Starting from a constant null model, with plot as random intercept and year as random slope, the null model was extended stepwise. Thereby, the random-effects term 'plot' provided an error term for all fixed-effects terms that did not vary within plots among years and the random-effects term 'time' provided an error for fixed-effects time contrasts such as precipitation. We fitted block as fixed effects as their number of three is too low to reliably estimate a variance component and because they do not fulfil the requirement of a random sample with normally distributed effects since they are systematically arranged in a linear sequence (Schmid et al., 2017). The fitting sequence of fixed terms followed the a-priori hypotheses of the biodiversity experiment starting with block followed by PSR (log-linear term); by plant functional group richness (linear term); and in alternative models the presence of all individual plant functional groups, as they are not independent of each other.
Furthermore, sampling depth of the soil water (20 cm, 30 cm); the interactions between 'depth' and plant diversity variables; season; the interactions between 'season' and plant diversity variables; and the interaction between 'depth' and 'season' were tested. Plant functional group richness and the presence of grasses were never significant and were therefore removed from all analysis models. The response variable DOM concentration was log-transformed in order to obtain a normally distributed error structure and to stabilise variances. The maximum likelihood method was used and likelihood-ratio tests (L-Ratios) were applied to assess the statistical significance of stepwise model improvement. All variables were scaled to a range between 0 and 1 (Legendre & Legendre, 1998) to improve model convergence. To test plant community effects on molecular DOM properties (m/z, H/C, and O/C), similar LMM were used as described above, but only with plot as random effect as DOM properties were measured once in a single block.

| Identifying mechanisms underlying plant diversity effects on DOM
Using the R library 'piecewiseSEM' (Lefcheck, 2016)

| Impact of plant diversity on DOM concentrations
During our observation period from 2011 to 2016, the average DOM concentration was 9.6 mg/L (±4.4 SD, N = 9,563 of all fortnightly samples). However, DOM concentrations varied strongly with season (Figure 1a,b). In winter, the DOM concentration was lowest (7.7 ± 3.5 mg/L), while the concentrations increased towards spring (8.6 ± 4.4 mg/L), and reached their peak in summer (13.1 ± 3.6 mg/L). The concentrations decreased again towards autumn (11.6 ± 2.7 mg/L). Thus, the mean DOM concentrations in summer exceeded those in winter by almost 70%. These effects were consistent across all seasons and at both sampling depths (Figure 1b; Appendix S4). The presence of tall herbs had no significant consistent effect on DOM concentrations. Instead, the presence of tall herbs affected DOM concentration to varying extents at different sampling depths, as indicated by the significant interaction term 'tall herbs × depth' (L = 7.14; p = 0.008). At 30-cm sampling depth, DOM concentrations were higher in plots with tall herbs (10.5 ± 4.6 mg/L) than in plots without (9.5 ± 4.4 mg/L), while the concentrations did not differ at a depth of 20 cm (Figure 1c). This indicated that the strength of the PSR effect was the same at both depths and therefore smaller, less degraded, low O/C ratio DOM molecules can be found at both soil depths when there is a high PSR.

| Seasonal variations in DOM concentrations
The seasonal patterns of DOM concentrations in our study support our hypothesis and are in line with the general observations that DOM concentrations in soil solution are higher in summer than in winter (Kalbitz et al., 2000), but see Don and Schulze (2008) for inconsistent findings. A peak in late summer has been reported frequently and is called the 'rewetting peak', as it can be observed after dry periods when the soil gets humid again (Kalbitz et al., 2000). Our study is in line with these observations, with the highest DOM concentrations found in August (Appendix S9).
Previous studies suggested that over the dry period either microbial products (Don & Schulze, 2008) or plant-derived compounds (Karlowsky et al., 2018) accumulate in the dry soil. After rewetting of the soil, these accumulated compounds get leached and contribute to the high DOM concentrations. However, the DOM concentrations peak after such rewetting seems to depend on the soil type and its clay content, as Don and Schulze (2008) did not find a rewetting peak in a Vertisol that contains up to 70% clay.
At the Jena Experiment site, the soil clay content was much lower with an average of 21%. Besides this peak in DOM concentration, higher DOM concentrations were generally observed from July to November (Appendix S9) compared to the period from December to May. This can be most likely attributed to higher biological activity, resulting in higher plant productivity and microbial degradation (Kadereit et al., 2014;Pietikäinen et al., 2005). increased DOM concentrations at 30-cm soil depth but not above 20 cm indicates that deep-rooting plants enhance plant carbon allocation to deeper soil layers. Thereby, microbial activity can be enhanced with potential consequences for decomposition processes in deeper soil layers. However, the presence of tall herbs or its interaction with depth was not confirmed by the path model, and there was no effect of tall herbs on molecular properties of DOM at any depth (Appendix S3). This indicates that the microbial decomposition is not restricted to plant material, which is further supported by a decrease in the mean molecular DOM size (Roth et al., 2019), but also to the decomposition of stored SOM. However, as the molecular properties of DOM were only measured on a reduced dataset (see Section 4), future studies are needed to further investigate this relationship.

| Plant diversity effects on DOM concentrations and molecular properties of DOM
Beside the effect of plant diversity on DOM concentrations, our study demonstrates an impact of plant diversity on molecular properties of DOM, supporting our hypothesis. The molecular differences in DOM indicate a plant diversity-mediated shift in the plant-microbe interplay of carbon production and decomposition. The decrease in mean molecular weight of DOM molecules with plant diversity is caused by a higher abundance of small molecules that are likely to be microorganism-generated during the very early decomposition of plant-derived carbon (Roth et al., 2019). This is in line with previous findings that plant diversity impacts low molecular weight compounds in soil (El Moujahid et al., 2017). With increasing soil depth, the mean molecular weight of DOM increases due to preferential consumption of small DOM molecules and microbial production of larger molecules (Roth et al., 2019). Thus, molecules of microbial origin become more dominant in DOM with increasing soil depth (Kaiser et al., 2004;. Moreover, the lower O/C ratio that accompanies higher plant diversity shows that with higher plant diversity the molecules have a higher potential to be oxidised. In contrast, the H/C ratio, indicating the hydrogen saturation of carbon, was not related to plant diversity and composition but was strongly influenced by depth. This change in carbon saturation with soil depth (Roth et al., 2015) indicates independent mechanisms from the plant species richness induced changes in molecular DOM.
However, with higher plant diversity, less degraded and more easily decomposable small compounds were found in deeper soil layers in this study, indicating that the products of the early decomposition of plant inputs exceed the microbial consumption and reach deeper soil layers. Thus, with higher plant diversity, fresher and more reactive molecules are transported into deeper soil layers, which may foster a microbial life that initiates soil development, soil biodiversity and soil fertility (Klopf et al., 2017;Nielsen et al., 2015). This molecular DOM transformation is not only restricted to the top 30 cm of the soil presented in our study, but is a more general phenomenon observed along the soil profile (Roth et al., 2019).
In contrast to the clear difference in the mean molecular weight of DOM, there was no difference in DOM concentration between sampling depths in our study. This indicates, firstly, that the legacy of former land use as arable land with its homogenised plough horizon is still visible in the DOM concentrations and, secondly, that microbial processes cannot be fully understood without investigating molecular DOM properties. However, the molecular DOM analyses were limited to one time point and only 20 plots.
For DOM from forest soils, it was reported that its molecular composition correlates with DOM concentrations and season (Roth et al., 2015). This indicates that there is a constant input and recycling of organic matter. However, we suspect that plant inputs as well as the microbial consumption and production are constantly driven by plant diversity, so the effect of plant diversity on the molecular properties of DOM is likely to persist irrespective of the season. Moreover, the block from which the samples for molecular DOM analyses were derived was located between a block with higher silt and a block with higher sand content, i.e. it represents 'average soil conditions' of the experimental site. Nevertheless, we assume that the general plant diversity effect on molecular DOM metrics is not affected by these differences in soil conditions between blocks. The DOM transformation during its passage through soil in the Jena Experiment was confirmed at a different location with very sandy and acidic soil (Roth et al., 2019). This suggests a consistent influence of the DOM drivers, such as plant diversity, regardless of soil type and texture, although the strength of the drivers may be enhanced or attenuated. Moreover, Roth et al. (2019) showed that the explained variances in molecular DOM through plants and soil are independent of each other.
However, investigating the molecular DOM on a larger set of plots and throughout the year, together with more detailed information on soil microbial community composition, is likely to provide more detailed insights and greater mechanistic understanding of how plant diversity affects below-ground processes.

| Mediators of the plant diversity and composition effects
Structural equation modelling suggested that plant diversity effects on DOM concentration were mainly driven by increased microbial activity and SOM content in the topsoil. However, the effect of microbial activity and SOM content on DOM concentrations differed between seasons, indicating that the consistent plant diversity effect is mediated by different drivers in different seasons. Soil microorganisms have been found to act as both sinks and sources of DOM (Kalbitz et al., 2000;Neff & Asner, 2001). In our study, microbial DOM consumption exceeded microbial DOM production in winter and spring, as indicated by the negative relationship between soil microbial activity and DOM concentrations. The positive relation between microbial activity and DOM concentrations in summer and autumn indicate that the microbial activity changed its impact on DOM during the year from consumption to production of DOM.
The relationship between DOM concentrations in soil water and SOM contents in the soil, which are significantly higher in more diverse plant communities , also changed during the year. In contrast to soil microbial activity, the relationship was positive in winter and changed during the year to negative in autumn. These patterns indicate that in plant communities with high diversity, SOM is mostly a source of DOM in winter, but later in autumn DOM is consumed with higher SOM contents. However, these relationships were weaker than those of DOM with the soil microbial community and they likely depend on soil microbial activity. In winter, when the microbial community is less active and less abundant , SOM was positively related to DOM concentration, indicating that SOM leaching contributes to higher DOM concentrations. The tendency towards a negative relationship between SOM and DOM concentrations in autumn is likely to be related to increased consumption of the SOM leachates by the more abundant soil microbial community in highly diverse plant communities. In addition, the microbial community contributes to higher DOM concentrations by decomposition of dead plant residues in the second half of the growing season (Kuzyakov, 2002).
Root standing biomass only mediated the plant diversity effects on DOM concentrations in autumn. The negative impact of root standing biomass on DOM concentration was counter-intuitive at first glance.
Although the root standing biomass was shown to increase with plant diversity Ravenek et al., 2014) roots from plant communities with high diversity had lower decomposition rates  and longer life spans (Solly et al., 2013). Moreover, Chen et al. (2017) reported that root decomposition rates increase in the presence of legumes and small herbs due to an increase in substrate quality, i.e. a decrease in root C/N ratio. This likely further explains the positive effects of legumes and small herbs on DOM concentrations, as in their presence root decomposition is accelerated and more decomposition products are found in the soil water.
We are aware of the fact that we relate data of different temporal resolutions, and that annual measurements of predictors can partly be too coarse to fully explain the variability of DOM concentrations in different seasons of the year. However, most predictors do not change over the course of the year, including soil properties (silt and SOM content) and experimental design variables (sampling depth, plant species richness). In contrast, soil microbial activity and root standing biomass are likely to vary between seasons. Although it was shown in the Jena Experiment that the soil microbial community was more abundant in autumn than in spring, the plant diversity effect was consistently positive in both seasons . Root standing biomass is likely to be positively related to plant species richness throughout the growing seasons. Thus, these annual measurements may miss relevant variation between seasons as indicated by the direct paths of the experimental design variables in the structural equation models (basically reflecting unexplained variance), but they still turned out to be powerful proxies as mediators for the plant diversity effect over the course of the year. Therefore, our study gives first insights into the shifts in mediators of the plant diversity effects between seasons indicating that ecosystem functioning varies seasonally. Future studies should link temporally highly resolved measurements of predictors such as root production and decomposition as well as soil microbial community composition to DOM concentrations to further elucidate the drivers of DOM variability during the seasons.

| Environmental drivers of DOM concentrations independent of plant diversity
The negative effect of the silt content points to sorption processes that demobilise DOM and thereby decrease DOM concentrations (Kaiser & Guggenberger, 2000;Kalbitz et al., 2005). Higher silt content in soil is often accompanied by a higher free sorption capacity that decreases DOM concentrations. Soil clay content, which is considered to control sorption and adsorption processes, is relatively constant in the Fluvisol of the Jena Experiment. In contrast, there is a strong sand-silt-gradient at the field site (Roscher et al., 2004). The negative impact of silt on DOM concentration is likely to be driven by the fine silt portion, which has been demonstrated to increase the sorption capacities of soils (Schleuss et al., 2014). These sorption processes together with microbial mineralisation of DOM cause a strong decline in DOM concentrations with soil depth (Kaiser & Kalbitz, 2012). In our study, DOM concentrations did not differ between the soil depths we sampled, nor did the plant diversity effect on DOM concentration differ between sampling depths. This likely resulted from the small spatial distance of only 10 cm between the two sampling depths. In addition, the former land use caused a soil homogenisation in the upper 30 cm due to ploughing. However, the study of Roth et al. (2019), which took place in the same experiment, found a significant decrease in DOM concentrations towards 60 cm soil depth.
The changing effects of the sampled soil water volumes on DOM concentrations between seasons could reflect seasonally different interactions among the microbial community, SOM and soil water. In summer and autumn, when the DOM production is highest (Don & Schulze, 2008;Kalbitz et al., 2000), the negative effect of soil water on DOM concentrations indicates a dilution effect. In contrast, the positive effect of the sampled volumes of soil water on DOM in winter and spring indicates a faster vertical transport of soil water and its DOM to deeper soil layers, which is driven by higher SOM in the topsoil (Fischer et al., 2015(Fischer et al., , 2019Lange et al., 2019).
This suggests on the one hand that in winter and spring water transports more dissolved components of SOM to deeper layers, and on the other hand that this occurred in locations with substantial storage of SOM. Furthermore, the negative correlation between microbial activity and the sampled volume of soil water indicates that less microbial consumption and decomposition of DOM can take place when the soil water percolates faster through the soil.

| CON CLUS IONS
This study demonstrates that DOM may be an appropriate proxy to investigate ecosystem functions and functioning. DOM integrates the highly complex and interwoven processes of ecosystems, and reflects them in the concentration and molecular composition of DOM. Plant diversity affects a wide range of DOM drivers and their interactions , and thus DOM itself as an important mediator of above-below-ground coupling. DOM plays a crucial role in a multitude of processes, such as cycling and distributing of nutrients and carbon that are central to ecosystem functioning (Bolan et al., 2011;Jansen et al., 2014). With increasing plant diversity, soil water with elevated DOM concentrations, which has been less processed by soil microorganisms, reaches deeper soil layers. Thus, more of the less processed DOM can be utilised by the subsoil microbial community (Bolan et al., 2011). This less processed DOM might then foster a more fertile soil, that enables, for instance, higher rates of carbon storage (Fornara & Tilman, 2008;Lange et al., 2015) since DOM is, along with deep-rooting plants, the most important carbon source on which the formation of subsoil carbon is based (Rumpel & Koegel-Knabner, 2011). Thus, plant diversity stimulates above-below-ground interactions using DOM to 'activate' subsoils, which might in turn increase soil organic matter storage and soil fertility.

ACK N OWLED G EM ENTS
We gratefully thank U. Gerighausen for her great dedication in the soil water sampling and for initial data processing and also the service Group RoMA at the MPI for Biogeochemistry for measuring DOM concentrations. We further thank K. Klaproth for competent technical support with FT-ICR-MS. We thank Alice Orme for linguistic revision. Comments by two anonymous reviewers helped to improve the paper. We gratefully acknowledge Anne Ebeling,

PEER R E V I E W
The peer review history for this article is available at https://publo ns.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data on DOM concentration and soil organic carbon contents are available at https://doi.org/10.17617/ 3.52 (Lange, 2020