Dynamic niche partitioning in root water uptake facilitates efficient water use in more diverse grassland plant communities
Abstract
- Efficient extraction of soil water is essential for the productivity of plant communities. However, research on the complementary use of resources in mixed plant communities, and especially the impact of plant species richness on root water uptake, is limited. So far, these investigations have been hindered by a lack of methods allowing for the estimation of root water uptake profiles.
- The overarching aim of our study was to determine whether diverse grassland plant communities in general exploit soil water more deeply and whether this shift occurs all the time or only during times of enhanced water demand.
- Root water uptake was derived by analysing the diurnal decrease in soil water content separately at each measurement depth, thus yielding root water uptake profiles for 12 experimental grasslands communities with two different levels of species richness (4 and 16 sown species). Additional measurements of leaf water potential, stomatal conductance, and root traits were used to identify differences in water relations between plant functional groups.
- Although the vertical root distribution did not differ between diversity levels, root water uptake shifted towards deeper layers (30 and 60 cm) in more diverse plots during periods of high vapour pressure deficit. Our results indicate that the more diverse communities were able to adjust their root water uptake, resulting in increased water uptake per root area compared to less diverse communities (52% at 20 cm, 118% at 30 cm, and 570% at 60 cm depth) and a more even distribution of water uptake over depth. Tall herbs, which had lower leaf water potential and higher stomatal conductance in more diverse mixtures, contributed disproportionately to dynamic niche partitioning in root water uptake.
- This study underpins the role of diversity in stabilizing ecosystem function and mitigating drought stress effects during future climate change scenarios. Furthermore, the results provide evidence that root water uptake is not solely controlled by root length density distribution in communities with high plant diversity but also by spatial shifts in water acquisition.
A plain language summary is available for this article.
1 INTRODUCTION
Root water uptake and evapotranspiration of ecosystems link the water and carbon cycles between soil and atmosphere (Asbjornsen et al., 2011; Chapin, Matson, Chapin, & Mooney, 2002; Feddes & Raats, 2004; Feddes et al., 2001; Schneider, Attinger, Delfs, & Hildebrandt, 2009; Seneviratne et al., 2010; Teuling, Seneviratne, Williams, & Torch, 2006). Consequently, soil water access is a key factor for ecosystem functioning and ecosystem services (Cardinale et al., 2012; Hooper et al., 2005; Schwartz et al., 2000). Grasslands, which provide various ecosystem services, are strongly affected by climate change, especially due to increased drought stress (Barkaoui, Roumet, & Volaire, 2016). However, highly diverse grassland communities are able to mitigate drought stress effects and thus stabilize their ecosystem functions (Cardinale et al., 2012; Klaus et al., 2016; Knapp et al., 2002). For example, water stress of individual plant species alleviates if they are able to adjust their root systems to site-specific conditions, such as a shallow water table (Feddes et al., 2001). This also holds true for plant communities, where the diversity of plant species shape ecosystem evapotranspiration and root water uptake (Zea-Cabrera, Iwasa, Levin, & Rodríguez-Iturbe, 2006).
Until now, only a few studies have investigated patterns of root water uptake in plant species mixtures, and little is known about the effect of plant species richness and particular functional groups as well as their root traits on ecosystem evapotranspiration and root water uptake patterns. Among these, Van Peer, Nijs, Rehfeul, and de Cauwer (2004) and Milcu et al. (2016) found enhanced ecosystem transpiration with increased plant species richness. In general, enhanced transpiration with species richness could be explained by higher leaf and thus transpiring area (De Boeck et al., 2006; Li et al., 2005), which, in these studies, coincides with higher plant diversity and biomass production (Roscher et al., 2005; Van Peer et al., 2004). In addition, enhanced transpiration could be explained by more efficient water extraction (Kulmatiski & Beard, 2013), by which we mean the optimal acquisition of soil water within plant communities instead of the common ecophysiological concept of water use efficiency, which describes the productivity of plants depending on the volume of water that they use (Milcu et al., 2014; Stanhill, 1986).
Furthermore, niche partitioning has been suggested as the basis for species complementarity in diverse communities (Hooper, 1998; Loreau & Hector, 2001; Schenk, 2006) and for the positive relationship between biodiversity and ecosystem functioning (Cardinale et al., 2007; Hooper et al., 2005). For example, plant communities containing species with different root distribution patterns might be able to occupy a larger niche space and thus can acquire more unexploited soil resources compared to communities containing species with similar root distributions (Berendse, 1982; De Kroon et al., 2012; Mommer et al., 2010; Mueller, Tilman, Fornara, & Hobbie, 2013). In this context, diversity in root traits such as root length density (RLD), maximum and minimum root depth, root foraging behaviour, and root system architecture possibly plays a key role in ecosystem functioning (Schenk, 2006). Furthermore, research shows that the diversity of root traits is even more important than the taxonomic richness, as it allows accounting for different spatial below-ground niches in diverse plant communities (Cody, 1986; Schenk, 2006; Silvertown, 2004; Stubbs & Wilson, 2004). Such niche differentiation in rooting patterns may result in community-specific root water uptake profiles, soil water fluxes, and storage, which are probably key to achieving more efficient water acquisition in diverse plant communities (Schwendenmann, Pendall, Sanchez-Bragado, Kunert, & Hölscher, 2014; Zea-Cabrera et al., 2006).
Investigations of below-ground niche partitioning typically rely on assessing root length density at different depths, thus implying that root distribution reflects capacity for resource uptake. However, several studies show that root water uptake profiles may shift independently of root length distribution (Garrigues, Doussan, & Pierret, 2006; Hamblin & Tennant, 1987). This is because root water uptake is also related to root hydraulic properties (Kulmatiski & Beard, 2013; Kulmatiski, Beard, Verweij, & February, 2010; Nimah & Hanks, 1973), soil salinity, soil water distribution (Doussan, Pierret, Garrigues, & Pagès, 2006; Garrigues et al., 2006; Van der Ploeg, Gooren, Bakker, & de Rooij, 2008), the amount of fine roots, and the architecture of root systems (Javaux, Couvreur, Vanderborgth, & Vereecken, 2013; Nepstedt et al., 1994; Rosado et al., 2011; Schneider et al., 2009; Williams et al., 1998) as well as the transpiration of leaves (Jackson, Sperry, & Dawson, 2000; Quijano, Kumar, Drewry, Goldstein, & Misson, 2012; Steudle, 2000). For example, Rossatto, Sternberg, and Franco (2012) found that herbs are able to adjust their water uptake in dry and wet seasons in a Neotropical savanna. Similarly, Volkmann, Haberer, Gessler, and Weiler (2016) observed that tree species, i.e. oak (Quercus petraea Liebl.) and a mixed culture of oak and beech (Fagus sylvatica L.), had the ability to shift water uptake towards deeper soil layers when needed and proved that RLD alone does not control water uptake distribution. Thus, plant species identity (Hamblin & Tennant, 1987), root architecture (Bechmann et al., 2014), and temporal and spatial plasticity in below-ground resource use could be major drivers of niche partitioning or competition for soil resources (Ashton, Miller, Bowman, & Suding, 2010; Callaway, Pennings, & Richards, 2003; Hildebrandt & Eltahir, 2007; Kulmatiski & Beard, 2013) and independent approaches are warranted to assess differences in root water uptake between ecosystems. However, how species-specific strategies interact to affect the community-level soil water uptake in species rich communities remains poorly understood. Although root water uptake profiles provide detailed information on resource uptake, plant growth, nutrient cycling, and species coexistence (Kulmatiski, Adler, Stark, & Tredennick, 2017), the majority of studies measured only total water fluxes, and did not differentiate it according to depth. We address this knowledge gap by using a data-driven method to estimate root water uptake profiles, such as by combining measurements of short-term fluctuations of soil water content (Guderle & Hildebrandt, 2015) with lysimeter-measured evapotranspiration.
The aim of our study was to investigate characteristics of root water uptake profiles in grassland plant communities with different species richness and productivity to gain a deeper understanding of how plant species richness affects efficient use of available soil water. We hypothesized that increasing plant species richness leads to (1) increased evapotranspiration rates due to increased transpiring area, and (2) spatial and temporal complementarity in root water uptake, such that plants acquire more unexploited soil resources to cover the enhanced water demand. Specifically, we investigated how communities with 4 and 16 grassland species extracted from a long-term biodiversity experiment (Jena Experiment; Roscher et al., 2004) differed in their water uptake profiles as estimated in a controlled environment facility for ecosystem research (the CNRS European Ecotron of Montpellier; Milcu et al., 2014).
2 MATERIALS AND METHODS
2.1 Plant communities
The Jena Experiment is a long-term grassland biodiversity experiment located in the floodplain of the Saale River north of Jena, Germany (50°57′ 4.25″N, 11°37′ 28.52″E, 130 m a.s.l.; Roscher et al., 2004). The mean annual precipitation is 610 mm and the mean annual temperature is 9.9°C (Hoffmann, Bivour, Früh, Koßmann, & Voß, 2014). The experiment was established in 2002 and the site was previously used as arable land for growing wheat and vegetables since the early 1960s (Roscher et al., 2004). The experiment consists of 82 large plots (20 × 20 m), which are arranged in four blocks according to variation in the soil texture (i.e. from sandy loam near the river to silty clay far from the river). The 82 plots are comprised of grassland communities with different plant species richness (1–60 species) and functional group richness (1 to 4 functional groups: grasses, tall and small herbs, and legumes) assembled randomly from a species pool typical for the local Arrhenatherion grasslands (Roscher et al., 2004).
Twelve of the plots were selected for an Ecotron study according to the following criteria: (1) including the four functional groups, (2) realized species numbers were close to sown species richness, and (3) plots were equally distributed across the experimental field site to account for different soil textures. The selected plots met the aforementioned criteria with the exception of one plot where no grasses had been sown. The soil-vegetation monoliths (2 m2 surface area, diameter of 1.6 m, 2 m depth with a weight of 7 to 8 t) were cut with a steel cylinder, excavated, turned upside down for sealing at the bottom, and returned to an upright position in December 2011. The cylinders were buried to the top edge in the experimental field, before being transported and installed in the lysimeter system of the Macrocosms platform of the Montpellier Ecotron, France, in March 2012, before the start of vegetation growth. This allowed for recovery from the disturbance of excavation such that the soil could settle within the steel cylinder in the same environment as the experimental field site (Milcu et al., 2014). For more information on the Ecotron facility, please see Supplementary methods and Figure A1 in the Appendix.
2.2 Estimation of evapotranspiration and root water uptake profiles
Ecosystem evapotranspiration (ET) was measured from the lysimeter weight changes (later referred to as measured ET) to validate the ET estimated with a water balance method (later referred to as modelled ET). The weight measurements (6 min resolution) were smoothed using a moving average over 30 min to reduce noise due to the experimental setup (Milcu et al., 2016). The investigated period was constrained to those 5 days with reliable (noise-free) lysimeter weight records and without irrigation. A detailed description of these constraints is given in the Supplementary methods in the Appendix.
A water balance method was used to estimate daily root water uptake profiles and thus daily ecosystem ET (modelled ET) from diurnal fluctuation of soil water content measurements (Guderle & Hildebrandt, 2015). The method consists of applying a running regression over multiple time steps on soil water content time series at each measurement depth. Here we used measurements with a temporal resolution of 1 min from 10, 20, 30, and 60 cm depth (Figures A2 and A3). During the day, evapotranspiration leads to a decrease in volumetric soil water content. This extraction of soil water extends over the entire active rooting depth. Additionally, soil water flow occurs both at night as well as during the day (Chanzy, Gaudu, & Marloie, 2012; Khalil, Sakai, Mizoguchi, & Miyazaki, 2003; Verhoef, Fernandez-Galvez, Diaz-Espejo, Main, & El-Bishti, 2006), following water potential gradients in the soil profile. Thus, during dry weather conditions, the time series of soil water content shows a clear day-night signal. We split up the time series by fitting a linear function to each day and night branch of the time series in order to disentangle soil water flow and actual root water uptake. In a prior investigation, we found the main transpiration time lasted from 05.30 to 18.30 hr, so the onset of the day and night branches were fixed to these times. Nighttime transpiration was low (<23% of daytime transpiration) and therefore neglected (Milcu et al., 2016). Subsequently, the root water uptake profile was integrated over the entire soil profile to determine the modelled ET per 1 m2 surface and day. The modelled ET was furthermore multiplied by a factor of two in order to upscale the modelled ET to the surface of one lysimeter, which is two m2.
The water balance method was validated by estimating the coefficient of determination (R2) of the regression between the measured ET (predictor variable) and the modelled ET (response variable) on 25th, 28th, and 29th June 2012. Since the modelled ET is calculated from the integrated root water uptake profile, good fit between the measured and modelled ET indicates that the predicted root water uptake profile represents the true profile, which was shown in a prior study by Guderle and Hildebrandt (2015). A detailed description and evaluation of the method can be found in Guderle and Hildebrandt (2015) and Guderle (2015).
The predicted root water uptake profiles were used to investigate whether differences in depth and quantity of water uptake between 4-species and 16-species mixtures occurred. Furthermore, we compared the normalized root water uptake profiles and the normalized root length density profiles. This gives insight into dynamic niche partitioning in the examined ecosystems, which is indicated when root water uptake profiles differ from root length density profiles.
2.3 Soil water content measurements
Soil water content measurements used for estimating root water uptake (RWU) were taken with time domain reflectometry (TDR) sensors TRIME®-PICO 32 (IMKO Micromodultechnik GmbH, Germany), which were placed horizontally at six soil depths into the soil monoliths (one sensor at 10, 20, 30, 60, 100, and 140 cm respectively). The used sensors had a rod length of 110 mm and a rod diameter of 3.5 mm. The total length of the sensor was 328 mm. The measurements were taken every minute from mid-June to the end of July. The sensors were installed with a minimum distance of 22 cm to other sensor installations, and the edge of the lysimeter steel wall. Before estimating root water uptake profiles, outliers (defined as differing ±1.5 SD from the mean of 30 measurements) of the soil water content measurements (1 min resolution) were removed. To reduce the noise of the TDR sensors output, we applied a third-order Savitzky–Golay finite-impulse response (FIR) smoothing filter over 240 min for 10 and 20 cm depths and a moving average over 720 min for 30 and 60 cm depths (Figures A2 and A3). The larger window in deeper soil was necessary because daily soil moisture changes were small and the signal to noise ratio was correspondingly smaller, which required more smoothing to allow meaningful estimates of RWU at those depths (Guderle, 2015; Guderle & Hildebrandt, 2015).
2.4 Leaf water potential and stomatal conductance
Leaf water potential was measured with a pressure chamber (Pressure Chamber Instruments Model 600; PMS Instrument Company, Albany, OR, USA) at predawn (04.00–06.00 hr) and midday (13.30–16.00 hr) for the most abundant species on each plot. Two to three species were chosen in the 4-species mixtures, and five to eight species in the 16-species mixtures (Table A1). All measurements were performed within 3 days (17th to 19th July 2012). Measurements were carried out on young, but fully expanded leaves from four individuals per species per plot. Stomatal conductance (gs, mmol m−2 s−1) was measured with a portable leaf porometer (SC-1 Leaf porometer; Decagon Devices, Pullman, USA) on three leaves from different individuals of all available species per plot (Table A1). Since the instrument's reliable upper measurement limit is 1,000 mmol m−2 s−1, measurements above this threshold were replaced with 1,000 mmol m−2 s−1 as suggested in Caplan and Yeakley (2010). Measurements were done in the auto mode using the first 30 s of stomatal conductance data to predict the final stomatal conductance under true steady-state conditions. Values of leaf water potential and stomatal conductance of the individual measurements were averaged per species per plot.
2.5 Measurement of vegetation structural variables
2.5.1 Community leaf area index
Community leaf area index (lai) was measured with a portable LAI-2000 plant canopy analyzer (LI-COR, Lincoln, USA) on 15th July 2012. Leaf area index was measured under diffuse light (evening) by taking one reference measurement above the canopy and five measurements near ground level within the centre of each lysimeter. More detailed information on the LAI measurements can be found in Milcu et al. (2014).
2.5.2 Leaf area index per species

More information on the estimation of SLA can be found in the Supplementary methods in the Appendix.
2.5.3 RLD, BGBM and RAI
Three soil cores (diameter 3.5 cm) were sampled to 60 cm depth in each plot. Each core was divided into six layers (0–5, 5–10, 10–20, 20–30, 30–40, and 40–60 cm) such that the bottom of the soil cores from 5–10, 10–20, and 20–30 cm aligned with the depth of the TDR sensors. The respective layers were pooled per plot, and washed with tap water over a sieve (mesh size 200 μm) to separate roots from soil. Cleaned roots were weighed and a subsample of the fresh roots (c. 2 g) was stored in 70% ethanol before it was dyed (with neutral red solution) and scanned (Scanner Optical STD4800 Regent Instruments Inc. with top light unit and an image resolution of 600 dpi). Root length density and root surface area were estimated with WinRhizo (Reg 2009c; Regents Instruments Inc.). The smooth off area of WinRhizo was set to 0.0001 and the length-width ratio to 3.0. In order to obtain RLD values for soil layers represented by the water balance method, the RLD values of 0–5 and 5–10 cm depths were summed, as were RLD values from 30–40 and 40–60 cm. Below-ground biomass (BGBM) was estimated as described above but the samples were then dried at 65°C for 3 days.

where i is the index of the soil layer.
2.6 Statistical analysis
We used the statistical software r 3.0.2 (R Development Core Team, http://www.R-project.org) to perform linear mixed-effects model analysis, using the r package lme4 (Pinheiro, Bates, DebRoy, & Sarkar, 2014). First, we investigated the relationship between evapotranspiration as the response variable and the fixed effects of species richness, measurement day (Time), and their interactions (model 1; Table 1). Macrocosm (=plot) and replicates per day were defined as random effects. Second, we analysed the relationship between root water uptake (response variable) and the fixed effects of species richness, depth, measurement day (Time), and their interactions (model 2; Table 1). Again, plot and replicates per day were used as random effects. Third, we analysed the relationship between root water uptake in each single depth (response variable) and the fixed effects of species richness, measurement day, cover of tall herbs, and their interactions (plot and replicates per day as random effects; model 3; Table 2). The statistical analysis for the effects of species richness on evapotranspiration and root water uptake was done separately for June and July to account for differences in environmental conditions [e.g. temperature, vapour pressure deficit (VPD) and irrigation]. Furthermore, the separate analysis was conducted because the selected days in June coincided with periods selected by Milcu et al. (2014) and the days in July coincided with the measurements of leaf water potential, stomatal conductance, and root traits, which were done on these days. Fourth, we investigated the relationship between percentage of root water uptake and percentage of root length density (both as response variables) and depth and measurement day (Time; fixed effects; plot, replicates per day and depth as random effects; model 4; Table 3). Fifth, effects of species richness and functional group identity (FG) on leaf water potential and stomatal conductance (response variables, respectively) were tested using species richness (SR) and FG as fixed effects and plot and species identity as random effects (model 5; Table A4). Differences between functional groups were identified with Tukey's HSD tests, using the function glht in the r package multcomp (Hothorn, Bretz, & Westfall, 2008) based on models fitted with the restricted maximum likelihood method. To account for the dependency of evapotranspiration and root water uptake on measurement day and to exclude pseudoreplication due to measurements taken on the same day, the measurement day was considered as fixed effect and replicates per day as random effects. All fixed effects were added stepwise to a constant null model containing only the random effects (Roscher, Schumacher, Schmid, & Schulze, 2014). Significant improvement of the model after adding the fixed effects was determined, using likelihood ratio tests (χ2 ratio). Effects of species richness on RLD, BGBM, RAI and RAI:LAI ratio were tested using a one-way ANOVA, while a simple linear regression was used to analyse the relationship between LAI and evapotranspiration and the relationship between root water uptake and root length density over depth.
June | July | |||||
---|---|---|---|---|---|---|
df | χ2 | p | df | χ2 | p | |
ET | ||||||
Species richness (SR) | 5 | 10.93 | <.001↑ | 5 | 3.86 | .049↑ |
Time | 7 | 0.78 | .675 | 6 | 1.73 | .188 |
SR × Time | 9 | 0.53 | .766 | 7 | 0.02 | .885 |
RWU | ||||||
SR | 11 | 86.62 | <.001↑ | 11 | 60.55 | <.001↑ |
Depth | 12 | 8.83 | .003↑ | 12 | 20.48 | <.001↑ |
Time | 14 | 2.86 | .239 | 13 | 0.52 | .472 |
Depth × SR | 17 | 11.6 | .009↑ | 16 | 1.09 | .779 |
Depth × Time | 23 | 3.28 | .773 | 19 | 4.44 | .218 |
SR × Time | 25 | 0.44 | .802 | 20 | 1.24 | .266 |
Depth × SR × Time | 31 | 0.97 | .987 | 23 | 0.64 | .888 |
- Models were fitted by stepwise inclusion of fixed effects. Listed are the results of likelihood ratio tests (χ2) and the statistical significance of the fixed effects (p value). Arrows indicate increase (↑) or decrease (↓) of evapotranspiration or root water uptake with species richness. Bold values indicate significant effects on ET and/or RWU.
10 cm | 20 cm | 30 cm | 60 cm | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
df | χ2 | p | df | χ2 | p | df | χ2 | p | df | χ2 | p | |
June | ||||||||||||
Species richness (SR) | 5 | 0.01 | .939 | 5 | 9.11 | .003↑ | 5 | 11.92 | <.001↑ | 5 | 0.00 | .993 |
Time | 7 | 0.96 | .619 | 7 | 0.05 | .973 | 7 | 2.01 | .366 | 7 | 2.05 | .359 |
SR × Time | 9 | 0.19 | .909 | 9 | 0.19 | .912 | 9 | 0.31 | .856 | 9 | 0.84 | .658 |
Cover tall herbs (TH) | 10 | 3.63 | .057 | 10 | 0.19 | .657 | 10 | 0.06 | .811 | 10 | 0.02 | .883 |
Cover TH × SR | 11 | 0.25 | .614 | 11 | 0.90 | .341 | 11 | 7.98 | .005↑ | 11 | 1.20 | .274 |
July | ||||||||||||
SR | 5 | 0.00 | .936 | 5 | 4.46 | .035↑ | 5 | 6.89 | .009↑ | 5 | 6.85 | .009↑ |
Time | 6 | 1.41 | .235 | 6 | 0.01 | .908 | 6 | 0.44 | .508 | 6 | 0.92 | .337 |
SR × Time | 7 | 0.23 | .628 | 7 | 0.12 | .727 | 7 | 0.34 | .559 | 7 | 0.09 | .343 |
Cover TH | 8 | 2.33 | .127 | 8 | 0.21 | .649 | 8 | 9.45 | .002↑ | 8 | 0.84 | .358 |
Cover TH × SR | 9 | 0.01 | .919 | 9 | 0.50 | .479 | 9 | 12.87 | <.001↑ | 9 | 1.83 | .176 |
- Models were fitted by stepwise inclusion of fixed effects. Listed are the results of likelihood ratio tests (χ2) and the statistical significance of the fixed effects (p value). Arrows indicate increase (↑) or decrease (↓) of root water uptake with species richness. Bold values indicate significant effects on root water uptake.
4-Species mixtures | 16-Species mixtures | |||||
---|---|---|---|---|---|---|
df | χ2 | p | df | χ2 | p | |
RV | 9 | 2.79 | .094 | 9 | 11.72 | <.001 |
Depth | 12 | 185.13 | <.001 | 12 | 40.79 | <.001 |
Time | 13 | 0.09 | .761 | 13 | 0.13 | .717 |
Depth × RV | 16 | 24.14 | <.001 | 16 | 25.88 | <.001 |
Depth × Time | 19 | 6.19 | .102 | 19 | 0.39 | .942 |
RV × Time | 20 | 0.09 | .759 | 20 | 0.31 | .572 |
Depth × RV × Time | 23 | 7.14 | .068 | 23 | 0.57 | .903 |
- Models were fitted by stepwise inclusion of fixed effects. Listed are the results of likelihood ratio tests (χ2) and the statistical significance of the fixed effects (p value). Bold values indicate significant effects on RLD and/or RWU.
- RV, root variable, means root length density vs. root water uptake.
Inspection of the residual plots did not indicate any obvious deviations from homoscedasticity and normality. Exceptions were root water uptake as the response variable in June (model 2) and stomatal conductance (model 5), which were log transformed in order to obtain homoscedasticity and normality for the linear mixed-effects model analysis in Table 1 and Table A4, respectively.
3 RESULTS
3.1 Plant species richness effects on evapotranspiration
In general, the modelled community evapotranspiration explained measured evapotranspiration very well (R2 = 0.77, p < .001; Figure A4), thus giving confidence in the ability of the model to derive root water uptake profiles.
On all investigated days, the measured ET of the 16-species mixtures was significantly higher than the ET of the 4-species mixtures (Figure 1). On average, ET was 18% higher in June and 27% higher in July in the 16-species mixtures compared to the 4-species mixtures.

Leaf area index and ET (mean over all measurement days) were positively related (R2 = 0.73, p < .001; Figure 2), and the 16-species mixtures had significantly higher LAI than the 4-species mixtures (Figure 3b).


3.2 Plant species richness effects on root water uptake patterns
In both June and July RWU decreased with depth (Table 1). The interaction between species richness and depth was significant in June, but not in July, indicating that uptake profiles were different between the 4-species and the 16-species mixtures in June (see also Figure 4). Separate analyses per depth showed that 16-species mixtures had a higher root water uptake at 20 cm depth (+86%) and 30 cm depth (+124%) compared to the 4-species mixtures in June (Table 2). In July when the VPD increased (Figure A5), we found a higher root water uptake at 20 cm depth (+52%), 30 cm (+118%) and 60 cm depth (+570%) in the 16-species compared to the 4-species mixtures (Table 2).

3.3 Root water uptake distribution
Over 90% of the sampled root fractions in 4- and 16-species mixtures were fine roots (0–0.5 mm; Table A3). Depth distribution of root water uptake was clearly related to RLD in the 4-species mixtures on 17th and 18th of July. This is also supported by a significant positive relationship between mean RWU and mean RLD over depth (R2 = 0.907, p = .032; Table 3). Both RWU and RLD followed the same curve shape over the entire soil profile (Figure 4a,c). On the contrary, the RWU profiles of the 16-species mixtures deviated significantly from RLD profiles (Figure 4b,d; Table 3). Here, proportional RWU is substantially smaller than the proportion of RLD at 10 cm depth (about 32%), whereas it is considerably higher than the proportion of RLD in deeper layers (about 17% at 20 cm and 10% at 30 cm). This difference strengthened on 17th July 2012, the day with the highest evapotranspiration among the considered days. Furthermore, root water uptake in the 16-species mixtures varied more between individual ecosystems, as evident from the broad range interval shown in Figure 4b,d, which is also supported by the results of the fitted linear model (R² = 0.626, p = .133). Interestingly, we found no significant difference in root length density distribution and BGBM distribution between the two diversity levels (Figure A6, Table A2).
Root area index, a measure of root surface per area soil surface, was not significantly different between diversity levels (Figure 3a, Table A2). In contrast, the RAI:LAI ratio was marginally significantly higher in 4-species mixtures than in 16-species mixtures (p = .084, Table A2), suggesting that the transpiring surface areas were higher in 16-species mixtures while the relative surface for water uptake, the RAI, was similar at both diversity levels. Hence, uptake velocity (flux per unit root area) probably increased in 16-species mixtures.
Furthermore, we found a higher cover of tall herbs in 16-species mixtures than in 4-species mixtures as well as a significant effect of cover of tall herbs on root water uptake at 30 cm in June and in July (Table 2) for 16-species mixtures.
3.4 Leaf water potential and stomatal conductance of functional groups
Plant species richness (SR) and FG had no effect on measured predawn leaf water potential (ᴪp), but species richness effects varied among functional groups (significant interaction SR × FG, Table A4). Midday leaf water potential (ᴪm) differed among functional groups (Table A4). Grasses had lower ᴪm than the other three functional groups in both diversity levels (Figure 5b). Tall herb ᴪm was greater in 4-species mixtures than in 16-species mixtures (p = .003), but ᴪm was not affected by diversity in the other functional groups (Figure 5b).

Stomatal conductance (gs) differed between functional groups (Table A4). Tall herbs had higher and grasses much lower stomatal conductance than the other two functional groups. Tall herbs in more diverse mixtures had higher gs values than in less diverse ones (p = .002). Stomatal conductance of small herbs, grasses and legumes did not differ between diversity levels.
4 DISCUSSION
Briefly, our results indicate dynamic niche partitioning in diverse plots, likely involving primarily tall herbs, which decrease their leaf water potentials, thus maintaining high flow rates throughout the root system (Doussan et al., 2006), extract water from deeper soil layers, and contribute to enhanced transpiration.
4.1 Effects of species richness on evapotranspiration rates
For all studied days, we found an enhanced actual evapotranspiration for the 16-species mixtures compared to the 4-species mixtures (Table 4), similar to studies of Kreutziger (2006) and Van Peer et al. (2004), and in agreement with Milcu et al. (2016). An increase in transpiration with greater leaf area has been frequently reported (e.g. Eavis & Taylor, 1979; Iritz & Lindroth, 1996; Li et al., 2005, 2006; Milcu et al., 2016; Rosset, Riedo, Grub, Geissmann, & Fuhrer, 1997; Zhongmin et al., 2009) and our study confirms this finding at two levels of diversity (Figure 2). The strong correlation between LAI and evapotranspiration indicates that transpiration is a substantial portion of evapotranspiration, which corroborates the findings of Milcu et al. (2016), who used the Shuttleworth and Wallace energy-partitioning model to distinguish transpiration and soil evaporation in the same experiment. The increased leaf area in 16-species mixtures and thus the higher evapotranspiration rates could be due to an increased productivity of diverse plant communities, which has been shown repeatedly in recent studies on biodiversity in ecosystems (Cardinale et al., 2012; Hooper, 1998; Marquard et al., 2009; Naeem, Duffy, & Zavaleta, 2012; Roscher et al., 2012). In general, two main drivers are discussed: (1) complementarity, which is described as higher resource use efficiency in diverse mixtures, and (2) that diverse communities have an increased probability to contain particular species, which have an above-average effect on ecosystem processes (referred to as “sampling effect”; Marquard et al., 2009; Roscher et al., 2011). Our results support previous research that both processes can act together (Marquard et al., 2009).
Low diversity | High diversity | Ratio high/low | |
---|---|---|---|
Sown species number | 4 | 16 | 4 |
Number of small herb species | 1 (0–1) | 3 (0–4) | 3 |
Number of tall herb species | 1 (0–2) | 2 (0–4) | 2 |
Number of legume species | 1 (0–1) | 3 (2–5) | 3 |
Number of grass species | 1 (0–1) | 3 (1–4) | 3 |
Average ET (25./28./29.06.) (mm/day) | 2.7 (0.4–4.6) | 3.5 (1.2–5.2) | 1.3 |
Average ET (17./18.07.) (mm/day) | 3.4 (1.5–5.2) | 4.6 (1.9–8.9) | 1.4 |
Average LAI (m2/m2) | 1.1 (0.5–3.0) | 3.2 (1.3–3.5) | 2.9 |
Upscaled sLAI (m2/m2) | 2.1 (0.9–3.6) | 3.4 (2.4–4.0) | 1.6 |
sLAI small herbs (%) | 46 (0–90) | 28 (0–91) | |
sLAI tall herbs (%) | 13 (0–68) | 31 (0–74) | |
sLAI legumes (%) | 16 (0–60) | 25 (0–69) | |
sLAI grasses (%) | 25 (0–92) | 17 (0–63) | |
Stomatal conductance (mmol s−1 m−2 surface) | 870 (416–1455) | 1682 (813–2280) | 1.9 |
Small herbs (%) | 34 (0–99) | 29 (0–83) | |
Tall herbs (%) | 29 (0–85) | 43 (0–84) | |
Legumes (%) | 22 (0–79) | 22 (0–57) | |
Grasses (%) | 16 (0–74) | 6 (0–29) | |
Root area index (RAI) | 96 (58–132) | 121 (63–217) | 1.3 |
RAI:LAI | 90.6 (30.6–140) | 48 (20–93) | 0.5 |
- Given are the arithmetic mean, as well as the observed maximum and minimum in parenthesis.
4.2 Higher exploitation of deeper soil layers in more diverse mixtures due to dynamic niche partitioning
At higher levels of plant species diversity, we found significantly higher root water uptake at 20 and 30 cm depths in June and at 20, 30 and 60 cm depths in July. This shift of root water uptake to deeper layers can be explained by the increased VPD in July, which leads to a higher evapotranspiration demand during this time, especially for the 16-species mixtures with their higher LAI values compared to the 4-species mixtures. Thus, the 16-species mixtures need to exploit soil water resources more efficiently to reduce water stress, and may increase spatial niche differentiation for water use in diverse mixtures. These results are in line with those of Verheyen et al. (2008), who suggested a more exhaustive exploitation of soil water in more diverse grassland communities. However, Bachmann et al. (2015) found no indication of niche partitioning for water in the Jena Experiment. As will be discussed below, the different findings could be related to the time of measurement.
Remarkably, in our study RLD was similar between the two diversity levels throughout most of the soil depths, and differed (slightly) only in the topmost soil (0–5 cm). The same applies for BGBM. Thus, there is no evidence for a spatial below-ground niche differentiation with regard to root abundance within the two considered species-richness levels. This is in line with findings of Ravenek et al. (2014), who investigated the root biomass of the Jena Experiment (from which our lysimeter data originate) over 9 years. Standing root biomass was sampled with soil cores of 4.8 and 8.7 cm diameter down to 40 cm in all 1-, 2-, 4-, 8- and 16-species mixture plots (in total 76 plots per year) in 2003, 2004, 2006, 2008 and 2011. They found no difference in root distribution over depth dependent on particular functional groups, species, or functional group richness. It is commonly assumed that vertical root water uptake distribution is proportional to the distribution of root length density over soil depth (e.g. Feddes et al., 2001; Jarvis, 1989; Perrochet, 1987; Prasad, 1988). Niche differentiation could result from vertical differentiation of root systems between different species and/or functional groups (Berendse, 1982; Brassard, Chen, Bergeron, & Paré, 2011). However, root distribution may also be driven by other resources, in particular nutrient availability. Thus, one explanation for the similar RLD distribution between the two considered diversity levels could be the nutrient-rich soil of the Jena Experiment's site (Bessler et al., 2009; Ravenek et al., 2014).
However, although we found no spatial root niche differentiation, water uptake shifted towards deeper, less densely rooted soil layers. This finding suggests a dynamic (functional) niche partitioning (shift in uptake patterns) in more diverse plant communities. Similar dynamics were repeatedly observed on single plants (e.g. Doussan et al., 2006; Garrigues et al., 2006; Lai & Katul, 2000; Rosado et al., 2011) and is in accordance with models of root water uptake (Bechmann et al., 2014; Doussan et al., 2006; Schneider et al., 2009). For instance, Doussan et al. (2006) showed that the contribution of different parts of the root system to total water uptake changes dynamically with time and depends on hydraulic properties and arrangement of the root system. Moreover, changes in soil moisture over time affect the root water uptake depth of different species differently (Volkmann et al., 2016). Another mechanism could be the improvement of water uptake by arbuscular mycorrhizal fungi (AMF) due to direct water uptake and transport via the fungal hyphae from soil pores, which are not accessible to the plant roots themselves (Birhane, Sterck, Fetene, Bongers, & Kuyper, 2012; Khalvati, Hu, Mozafar, & Schmidhalter, 2005). Mellado-Vázquez et al. (2016) found significantly higher concentrations of the neutral lipid fatty acid (NLFA) 16:1ω5N, which acts as a biomarker for the presence of AMF, in 16-species mixtures than in 4-species in this lysimeter experiment. Unfortunately, soil samples were only taken down to 10 cm depth, whereas we found an enhanced root water uptake at 20, 30 and 60 cm depth. However, the decrease in this NLFA with depth was much lower than that of root biomass, which suggests that AMF were likely present in deeper soil layers as well.
On the other hand, our findings are in contrast with observations using stable isotopes, which found no evidence of enhanced niche partitioning in more diverse plant communities with regard to root water uptake in the Jena Experiment (Bachmann et al., 2015). This discrepancy may have several causes. First, our measurements were conducted during a period of high transpiration demand, whereas the tracer experiments of Bachmann et al. (2015) were performed at the beginning of the growing season and shortly after mowing in summer and autumn when leaf area was decreased and transpiration subdued. As will be discussed below, niche partitioning may be limited to periods of high transpiration only. Second, while providing a direct way of observing root uptake depth, isotopic signals could only be observed in the sap water of single plants and do not account for the abundances of species in a community, while our measurements were conducted at the whole-community level.
4.3 Which processes drive the dynamic niche partitioning in root water uptake?
Overall, the ratios between root and leaf area are high (Table 4) compared to values from literature (Sperry et al., 1998; Tyree, Velez, & Dalling, 1998), indicating that the root system poses little limitation to water uptake, rather reflecting requirements of nutrient uptake (Boot & Mensink, 1991; Sperry et al., 1998). Although root area was similar in 16-species compared to 4-species mixtures, LAI was greater in the more diverse mixtures. With increased water uptake per root area in 16-species mixtures, xylem potentials at the root base and in leaves need to be more negative to maintain flow (Hacke et al., 2000; Kalapos, 1994). Thus, the limiting factor for water uptake might be the higher water flux density in diverse mixtures rather than the absolute availability of soil water (Sperry et al., 1998).
Since root xylem and leaf water potentials are related, decreasing RAI:LAI ratio should affect leaf water potential. In general, the leaf water potential prescribes the potential gradients between leaves and soil, driving water uptake (Kalapos, 1994; Sala, Lauenroth, Parton, & Trlica, 1981; Sperry et al., 1998). It usually shows a daily pattern with a minimum in the afternoon where transpiration reaches its daily maximum (Sala et al., 1981). In our experiment, species of tall herbs showed particularly negative leaf water potentials and greater stomatal conductance in diverse plots (Table 4, Figure 5). Furthermore, the LAI of tall herbs in 16-species mixtures was higher than in 4-species mixtures (Figure A7), and the total cover of tall herbs in 16-species mixtures was higher than in 4-species mixtures. Additionally, we found a significant effect of cover of tall herbs on root water uptake at 30 cm. These results strongly suggest that tall herbs used their differentiated tap root systems (Figure A8) to obtain a greater proportion of soil water from deeper layers even if the root distribution between functional groups as well as between different species mixtures was similar (Ravenek et al., 2014) and thus likely caused the recorded shift of root water uptake in 16-species mixtures. This might explain the wide range in the root water uptake patterns in the 16-species mixtures, because these plants adjust their uptake depth according to their demand within the community. These findings are strongly supported by Leimer et al. (2014), who also found low water content in deep soil related to the presence of tall herbs in the Jena Experiment. Furthermore, Marquard et al. (2009) reported a strong positive effect of tall herbs on community productivity in the Jena Experiment, which was attributed to both complementarity as well as “sampling effects”. The latter refers to the higher biomass expected of tall herbs, and the higher likelihood of presence of those species diverse communities, which by itself increases the expected biomass. Complementarity refers to additional processes supported by tall herb presence, which increase the ecosystem productivity beyond effects expected from selection alone. As already discussed, increased biomass productivity leads to higher LAI and likely increases evapotranspiration and root water uptake. Thus, our results corroborate that both a sampling effect and complementarity act in concert.
In summary, the more diverse communities in this experiment consisted of more plants with the capacity to extend their root water uptake deeper into the soil, such that the water use of the entire community was more uniformly distributed over the depth profile (“sampling effect”). Presumably, plants with root systems structured in transport and uptake roots, like tall herbs, have an advantage because they can explore the soil water more dynamically. Furthermore, the comparison between June and July showed that some species in more diverse plant communities could escape possible competition for water by exploiting deeper water resources, especially if the evapotranspiration demand is increased (complementarity). The effect of AMF on water uptake in 16-species mixtures is likely secondary since we found a significant relationship only between the abundance of tall herbs and elevated root water uptake in deeper soil layers, but not for the other functional groups in the plant community.
4.4 Which ecological conclusions can be drawn from these findings?
Our results reveal that grassland communities with higher plant diversity have improved water uptake over the soil profile due to dynamic below-ground niche partitioning. This ability of higher diversity grassland communities to mitigate drought stress and level off the related C cycling by photosynthesis and above-ground biomass production could stabilize their ecosystem functions during future climate change scenarios (Cardinale et al., 2012; Klaus et al., 2016; Knapp et al., 2002). Thus, conservation of plant diversity and increasing the number of species in species-poor agricultural grasslands is an essential management strategy that needs to be put into practice (Klaus et al., 2016; Schwartz et al., 2000).
5 CONCLUSIONS
Our study suggests that the increased transpiration of diverse grassland plant communities, associated with higher productivity and LAI, was achieved by complementary root water uptake in the absence of corresponding complementarity of root distribution. This study underpins the idea that root water uptake is not solely characterized by root length distribution but is also affected by dynamic shifts in water uptake due to temporal changes in environmental conditions. A pre-requisite for such plasticity is the ability of specific plant species to explore deeper soil layers. This dynamic water uptake requires a low leaf water potential in combination with a particular plant root structure to transport the water from the soil to the atmosphere. In our experiment, such a shift in water uptake was enhanced in more diverse plant communities, likely evoked by specific plant species such as tall herbs, since their abundances were higher in the more diverse grassland plant communities. These facts corroborate the growing discussion advocating conserving plant diversity in natural and agricultural grasslands in order to maintain good ecosystem functioning and services, especially against the background of future climate change and related drought stress of grassland ecosystems. Haines-Young and Potschin (2009) highlighted that sustainable management strategies and policies are essential for maintaining ecosystem functions. However, a clear understanding of the influence of biodiversity on ecological processes is crucial.
ACKNOWLEDGEMENTS
Financial support through the ProExzellenz Initiative from the German federal state of Thuringia to the Friedrich Schiller University Jena within the research project AquaDiva@Jena for conducting the research is gratefully acknowledged. This work was also supported by the German Research Foundation (FOR1451). This study benefited from the CNRS human and technical resources allocated to the Ecotrons Research Infrastructure as well as from the state allocation “Investissement d'Avenir” ANR-11-INBS-0001. The ‘Transnational Access’ of M.G. to the Ecotron was funded by the ExpeER I3 project (7th Framework Programme of the EC). We are thankful to the Ecotron team for running the experiment and providing data. M.G. was also supported by the International Max Planck Research School for Global Biogeochemical Cycles (IMPRS-gBGC). We thank the Handling Editor Katie Field, the Senior Editor Alan Knapp and the Assistant Editor Jennifer Meyer, for handling the manuscript as well as the two anonymous referees for their thorough discussion of the manuscript and their very helpful comments. We also thank Andrew Durso for text editing.
AUTHORS’ CONTRIBUTIONS
M.G. analysed the data and wrote the manuscript. A.M. coordinated the experiment. C.R., D.B., A.G., N.B. and A.W. provided plant and root trait data. M.G., D.L. and A.H. provided soil, evapotranspiration and root water uptake data. J.R., D.L., S.D. and O.R. provided Ecotron-related data and methods. All authors contributed substantially to revisions. The grant was written by the Jena Forschergruppe including J.R., A.H., A.G. and N.B.
DATA ACCESSIBILITY
Environmental data, soil water content, root water uptake, evapotranspiration, plant trait and root trait data of the Jena-Ecotron experiment are published at the Open Access Repository PANGAEA https://doi.pangaea.de/10.1594/PANGAEA.877687 (Guderle et al., 2017).