plant functional diversity increases grassland
TRANSCRIPT
Milcu et al 2015 - main manuscript
1/29
Plant functional diversity increases grassland productivity-related water 1
vapor fluxes: an Ecotron and modeling approach 2
Alexandru Milcu1,2, Werner Eugster3, Dörte Bachmann3, Marcus Guderle4, Christiane Roscher6, 3
Annette Gockele5, Damien Landais1, Olivier Ravel1, Arthur Gessler8, Marcus Lange7, Anne 4
Ebeling8, Wolfgang W. Weisser9, Jacques Roy1, Anke Hildebrandt4 & Nina Buchmann3 5
6
1CNRS, Ecotron (UPS-3248), Campus Baillarguet, F-34980, Montferrier-sur-Lez, France 7
2Centre d’Ecologie Fonctionnelle et Evolutive, CEFE-CNRS, UMR 5175, Université de Montpellier – 8
Université Paul Valéry – EPHE, 1919 route de Mende, F-34293, Montpellier Cedex 5, France 9
3Institute of Agricultural Sciences, ETH Zurich, Universitaetsstrasse 2, 8092, Zurich, Switzerland 10
4 Friedrich Schiller University Jena, Institute of Geoscience, Burgweg 11, 07749 Jena, Germany 11
5Albert-Ludwigs-Universität Freiburg, Institut für Biologie II – Geobotanik, Schänzlestr. 1, D-79104 12
Freiburg, Germany 13
6 UFZ, Helmholtz Centre for Environmental Research, Department Community of Ecology, Theodor-14
Lieser-Str. 4, 06120 Halle, Germany 15
7Max Planck Institute for Biogeochemistry, Hans-Knoell-Str. 10, 07745, Jena, Germany 16
8 Swiss Federal Research Institute WSL, Zürcherstr. 111, 8903 Birmensdorf, Switzerland 17
9Friedrich Schiller University Jena, Institute of Ecology, Dornburger Str. 159, 07743, Jena, Germany. 18
10Terrestrial Ecology Research Group, Department of Ecology and Ecosystem Management, School of 19
Life Sciences Weihenstephan, Technische Universität München, Hans-Carl-von-Carlowitz-Platz 2, 85354 20
Freising, Germany. 21
22
Corresponding author: Alexandru Milcu, CNRS, Ecotron - UPS 3248, Campus Baillarguet, 34980, 23
Montferrier-sur-Lez, France, email: [email protected], phone: +33(0)467-613-243. 24
25
26
Milcu et al 2015 - main manuscript
2/29
Abstract 27
The impact of species richness and functional diversity of plants on ecosystem water 28
vapor fluxes has been little investigated. To address this knowledge gap, we combined a 29
lysimeter setup in a controlled environment facility (Ecotron) with large ecosystem samples/ 30
monoliths originating from a long-term biodiversity experiment (“The Jena Experiment”) and a 31
modelling approach. We aimed at (1) quantifying the impact of plant species richness (4 vs. 16 32
species) on day- and night-time ecosystem water vapor fluxes, (2) partitioning ecosystem 33
evapotranspiration into evaporation and plant transpiration using the Shuttleworth and Wallace 34
(SW) energy partitioning model, and (3) identifying the most parsimonious predictors of water 35
vapor fluxes using plant functional trait-based metrics such as functional diversity and 36
community weighted means. Day-time measured and modeled evapotranspiration were 37
significantly higher in the higher diversity treatment suggesting increased water acquisition. The 38
SW model suggests that at low plant species richness, a higher proportion of the available energy 39
was diverted to evaporation (a non-productive flux), while at higher species richness the 40
proportion of ecosystem transpiration (a productivity-related water flux) increased. While it is 41
well established that LAI controls ecosystem transpiration, here we also identified that the 42
diversity of leaf nitrogen concentration among species in a community is a consistent predictor 43
of ecosystem water vapor fluxes during day-time. The results provide evidence that, at the peak 44
of the growing season, higher LAI and lower percentage of bare ground at high plant diversity 45
diverts more of the available water to transpiration – a flux closely coupled with photosynthesis 46
and productivity. Higher rates of transpiration presumably contribute to the positive effect of 47
diversity on productivity. 48
Milcu et al 2015 - main manuscript
3/29
Keywords: biodiversity-ecosystem functioning, evapotranspiration, functional traits, 49
plant species richness, ecosystem evaporation, ecosystem transpiration, leaf area index, 50
Shuttleworth and Wallace model, lysimeter, The Jena Experiment, Ecotron. 51
52
INTRODUCTION 53
The evidence that biodiversity loss reduces the functioning of ecosystems has grown over the last 54
two decades (Cardinale et al. 2012, Hooper et al. 2012). Many experimental studies established 55
that increasing plant species richness in a community generally leads to higher biomass 56
production, an easily measurable end-product of ecosystem functioning. However, only little is 57
known about the ecosystem-level energy and mass fluxes underpinning the biomass production 58
(Duffy 2003, Cardinale et al. 2012, Milcu et al. 2014). To date, the selection (Huston 1997) and 59
complementarity (Tilman et al. 1997, Cardinale et al. 2007) effects are the main mechanisms put 60
forward to explain the positive biodiversity-ecosystem functioning relationship. While the 61
selection effect can lead to enhanced ecosystem functioning due to an increased probability of 62
including a highly-performant dominant species with increasing species richness, the 63
complementarity hypothesis is asserting that increasing plant diversity should lead to a more 64
complete exploitation of resources. Complementarity is expected to arise in more species rich 65
communities due to increased probability of including species that exhibit: (i) various forms of 66
niche partitioning leading to complementary resource capture in space and time; or (ii) 67
interspecific facilitative/ positive interactions that enhances the capture of resources. In 68
particular, ecosystem evapotranspiration (ET) and transpiration (T), two productivity-related 69
fluxes strongly coupled with photosynthesis, are expected to increase with plant diversity due to 70
Milcu et al 2015 - main manuscript
4/29
their well-documented positive effect on biomass production and leaf surface area per ground 71
area (the LAI) (Obrist et al. 2003, Hu et al. 2009). Belowground, complementarity in rooting 72
patterns (von Felten and Schmid 2008) could also potentially lead to increased water acquisition 73
and hence increased ET and T. 74
To our knowledge, the importance of plant species richness for water vapor fluxes has 75
only been investigated in a few studies. With some notable exceptions (Stocker et al. 1999, 76
Leimer et al. 2014), indications of more complete water use as well as increased water use 77
efficiency have been found at higher diversity levels (Caldeira et al. 2001, Van Peer et al. 2004, 78
Lemmens et al. 2006 and Verheyen et al. 2008). However, it is unclear whether these results can 79
be generalized since most of these studies included experimental constraints related to the use of 80
relatively small containers (Van Peer et al. 2004, De Boeck et al. 2006), short establishment 81
phase (Stocker et al. 1999, Van Peer et al. 2004, De Boeck et al. 2006) and indirect estimates of 82
evapotranspiration (ET) using water balance models (Verheyen et al. 2008, Leimer et al. 2014). 83
Furthermore, no attempt has been made to partition the ecosystem ET into evaporation (E), a 84
non-productivity related flux, and T. In addition to being controlled by different biological and 85
physical processes, by not partitioning the ecosystem water vapor fluxes in E and T, one cannot 86
disentangle the direct (water acquisition-related) effects of species richness on T from canopy 87
structure-mediated effects on ecosystem E. Consequently, we only have an incomplete 88
understanding of the underlying mechanisms influencing non-productive (E) and productive (T) 89
water vapor ecosystem fluxes, which could potentially play an important role in explaining the 90
biodiversity-ecosystem function relationship (Verheyen et al. 2008). 91
There is mounting evidence that consideration of plant morphological, biochemical, 92
behavioral, and phenological traits known to affect performance and fitness can improve our 93
Milcu et al 2015 - main manuscript
5/29
understanding of ecosystem functioning (Cadotte et al. 2009, Díaz et al. 2007, Milcu et al. 2014). 94
To date, the most common community-level functional predictors/ metrics derived from plant 95
functional traits include the community weighted mean-trait value (CWMs) and functional 96
diversity (FD). CWMs, computed as the average of the trait values weighted by the relative 97
abundances of each species, can be used to identify the relative importance of a functional trait in 98
driving ecosystem processes. Theoretically, this metric is related to the mass ratio hypothesis of 99
Grime (1998) stating that ecosystem functioning is primarily determined by trait values of the 100
dominant contributors to plant biomass (Díaz et al. 2007). In contrast, FD metrics quantifies the 101
variety, range and evenness of the functional traits present in communities, and have been 102
proposed to be linked to niche complementarity hypothesis since a greater range of trait-values is 103
generally considered to indicate less niche overlap (Díaz et al. 2007). However, while functional 104
trait-based metrics have been extensively used to predict biomass production, the role of 105
functional traits for water vapor fluxes has not been yet explored despite evidence that this 106
approach may help to achieve a better predictive framework for ecosystem functioning (Wright 107
et al. 2004, Violle et al. 2007, Reiss et al. 2009). Based on our current understanding we would 108
expect the ecosystem water vapor fluxes to be best predicted by plant functional traits affecting 109
the LAI and the rooting patters (Reichstein et al. 2014). 110
To address the aforementioned knowledge gaps we took advantage of twelve large 111
lysimeters in a controlled environment facility for ecosystem research (Ecotron) hosting intact 112
vegetation-soil monoliths from a long-term biodiversity experiment (“The Jena Experiment”) 113
(Roscher et al. 2004). This was combined with a modeling approach, based on the Shuttleworth 114
and Wallace (1985) energy partitioning model (henceforth SW), which allowed to separately 115
quantify the proportion of energy that is dissipated in T, E and sensible heat flux (H). In order to 116
Milcu et al 2015 - main manuscript
6/29
derive a mechanistic and predictive understanding of plant diversity effects, alongside species 117
richness we also tested several predictors based on above- and belowground vegetation 118
properties and functional trait-based indices (Appendix A). Specifically, we aimed at (1) 119
quantifying the effect of plant diversity on day- and night-time ecosystem water fluxes during the 120
peak of the growing season, (2) partitioning the ecosystem ET in E and T using a modeling 121
approach to test the importance of plant species richness (4 vs. 16 species) on the partitioned 122
fluxes and (3) identifying the most parsimonious statistical models of water vapor fluxes using 123
above- and belowground vegetation properties and functional trait-based metrics [CWMs and 124
Rao’s quadratic entropy index of functional diversity (FDQ)]. 125
126
MATERIALS AND METHODS 127
The Ecotron facility 128
The experiment was conducted in the Montpellier European Ecotron 129
(http://www.ecotron.cnrs.fr), an experimental infrastructure developed by the Centre National de 130
la Recherche Scientifique (CNRS, France), to study the response of ecosystems to global 131
environmental changes. The lysimeters (2 m2, circular with a diameter of 1.6 m, weighing 7 to 8 132
tonnes) were located in 12 controlled environment units of the macrocosms platform. Each unit 133
consists of a 30 m3 dome-shaped chamber situated on top of a dedicated lysimeter room [see 134
Milcu et al. (2014) and Appendix B for more information on the macorocosms platform of the 135
Ecotron facility]. The soil surface and canopy of the lysimeters were exposed to natural sunlight 136
within each dome as a highly transparent material to light and UV radiation (250µm thick 137
Teflon-FEP film, DuPont, USA) was used as cover. Automatically controlled feedback loop 138
Milcu et al 2015 - main manuscript
7/29
algorithms based on industrial grade Proportional-Integral-Derivative (PID) controllers were 139
used to achieve the desired set-points for air temperature, air humidity and air CO2%. 140
Plant communities 141
The soil monoliths containing the plant communities originated from the long-term Jena 142
Experiment (50° 57.1 'N, 11° 37.5' E, 130 m above sea level; mean annual temperature 9.3 °C, 143
mean annual precipitation 587 mm). The site is located on the floodplain of the Saale River 144
(Jena, Germany), and was a former arable field until 2000, then kept fallow before 82 large plots 145
(20 ×20 m) varying in plant species richness (1 to 60 species), plant functional groups (1 to 4, 146
grasses, small herbs, tall herbs and legumes) and plant species composition were established in 147
May 2002 (Roscher et al. 2004). Twelve plant communities from two sown diversity levels (4 148
and 16 species) with six independent replicates per diversity level (Appendix C), were selected 149
according to the following criteria: (1) at least three functional groups (legumes, grasses and 150
herbs) were present, (2) realized species numbers were close to sown species richness, and (3) 151
plots were equally distributed across the experimental blocks of the field site to account for the 152
variability in soil texture. The monoliths were selected to be representative (as percentage 153
vegetation cover and standing biomass) of the plots they originate from. The monoliths were 154
sampled and placed in lysimeters in December 2011 following an established non-compacting 155
extraction method (see supplementary methods in Appendix D for more information on the 156
procedure used to extract the soil monoliths). At the end of March 2012 the lysimeters were 157
transported to the Ecotron facility in Montpellier. 158
Experimental conditions 159
Milcu et al 2015 - main manuscript
8/29
During the four months when the lysimeters where hosted in the Ecotron, we aimed to simulate 160
the average climatic conditions at the Jena Experiment field site since 2002. As the recorded 161
spring and summer climatic conditions of the year 2007 were very close to the average 162
temperature and precipitation regimes in Jena for the period 2002-2010, they were used as set-163
points (at 10-minute intervals). The average air temperature achieved was close to the set-point 164
(14.0 ºC vs. 14.9 ºC in Jena). The achieved average air humidity (RH) however, was somewhat 165
lower (58.9% RH vs.73.4% RH in Jena) because during night-time the humidifying system in the 166
Ecotron had to be stopped occasionally to prevent wetting the vegetation when the set-points 167
were higher than 80% RH. Because of this and also, as the monoliths were exposed to slightly 168
higher temperatures during transport and prior to installation in the Ecotron, we opted for 169
increasing the precipitation by +28% relative to 2007 (Appendix E). This allowed achieving 170
similar soil moisture conditions (Appendix F). The precipitation was applied by manual watering 171
using a hose equipped with a sprinkler and a flowmeter. As the incoming short-wave radiation 172
estimated from the HelioClim-1 database (Blanc et al. 2011) was on average 37% lower in Jena 173
than in Montpellier between April and July, a black shading mesh was added on the inside of 174
each dome, which reduced the incoming radiation by 44%. Unwanted plant species (weeds) were 175
removed every three weeks to maintain the targeted diversity levels. To recreate the mowing 176
management of the Jena Experiment, the aboveground biomass was mown at the end of April 177
and at the end of July. The final harvest took place at the time of the July mowing and included 178
destructive vegetation and soil sampling. As the modeling exercise relied on a period of six days 179
with undisturbed water vapor fluxes (28th of June to 3rd of July, 2012), the continuous 180
environmental conditions (air RH, air temperature, VPD and radiation) during this period are 181
presented in Fig. 1ab). 182
Milcu et al 2015 - main manuscript
9/29
Water vapor flux measurements 183
Ecosystem ET was measured as lysimeter weight changes over time. Four shear beam load cells 184
per lysimeter (CMI-C3, Precia-Molen, France), with an accuracy of ± 200 g, were used to 185
measure the changes in weight. The raw weight changes were corrected for temperature effects 186
on the sensitivity of the bean load cells, with correction factors provided by the manufacturer. 187
The resulting data were then smoothed using a symmetric loess smoothing function with a span 188
of 0.1 (as available in the R 3.1.2 software environment) to reduce the impact caused by minute 189
differential pressure changes between the dome and the lysimeter chamber where the bean load 190
cells are installed. Since not all weight changes can be associated with water vapor fluxes (e.g. 191
during activities that add or remove mass from the lysimeter such as sampling, weeding or when 192
a measuring device is placed on the lysimeter), we restricted our analysis to a time period with 193
the highest lysimeter data quality obtained during six consecutive undisturbed days (28th of June 194
to 3rd of July, 2012). This period was also close to the final harvest of the plant communities 195
(17th-20th of July) when all plant traits and vegetation properties were measured. The lysimeter 196
measured ET for this period is presented in Fig. 1c. These measurements were used to estimate 197
average day- (ETday) and night-time (ETnight) ecosystem evapotranspiration as well as the 198
evapotranspiration over 24h (ET24h). For more information on the measurement and sampling 199
methodology, see Milcu et al. (2014) and Appendix F. 200
The Shuttleworth and Wallace model 201
Probably the most widely used model for ecosystem-scale ET simulations is the Penman (1948) 202
model with Monteith (1981) extension to include stomatal resistance as a term that allows one 203
simulate the plant control of ET, today known as the Penman-Monteith equation (PM). 204
Shuttleworth and Wallace (1985) further developed a canopy model based on PM to allow 205
Milcu et al 2015 - main manuscript
10/29
partitioning of the radiative energy in sensible heat flux (H) and ecosystem evapotranspiration 206
(ET) which can be further partitioned in evaporation (E) and transpiration (T). This is done by 207
expressing PM (a) for the top of the vegetation canopy, (b) for the soil surface below the canopy, 208
and (c) for bare ground considering that in any canopy there is a certain fraction of projected 209
surface area that is not entirely covered by the plant community. PM uses available energy, air 210
temperature, atmospheric moisture and a series of transfer resistances, including the stomatal 211
resistance of vascular plants to predict ET (see Appendix F for equations and full model details). 212
213
Vegetation properties and functional trait-based metrics 214
We measured the following vegetation properties: shoot biomass (ShootBM), root biomass 215
(RootBM), root biomass and length by soil volume and depth (0-5, 5-10, 10-20, 20-30, 30-60 216
cm), total biomass (TotalBM), shoot biomass of legumes (LegBM), shoot biomass of grasses 217
(GrassBM), shoot biomass of herbs (HerbBM), leaf area index (LAI), leaf biomass (LeafBM) 218
and percentage bare ground (%bare-ground). 219
The functional–trait based metrics included functional diversity indices (FD) and community 220
weighted means (CWM) calculated from ten plant functional traits that have been previously 221
shown to be linked to plant transpiration, photosynthetic rates and light interception (see 222
Appendix A for an overview of tested predictors). Rao's quadratic entropy (FDQ) (Botta Dukát 223
2005) was preferred as an index of functional diversity as it incorporates information about 224
functional distance as well as functional evenness (abundance-weighted ) of a community. 225
Species-specific aboveground biomass was used for abundance-weighting. For each available 226
plant species the following traits were measured in-situ in each dome before the final destructive 227
harvest: stomatal conductance (gs; µmol m-2 s-1), specific leaf area (SLA; mm2 mg-1), leaf 228
Milcu et al 2015 - main manuscript
11/29
greenness (dimensionless measure of foliar chlorophyll content), leaf dry matter content (LDMC; 229
mg g-1), leaf N concentration (% N in leaf DW), species-specific plant height (cm) and specific 230
leaf nitrogen (SLN; g N m−2 leaf). See Appendix D for further details on the sampling and 231
procedure. Literature surveys were used for seasonality of foliage (ordinal, 1 = summer green, 2 232
= partly evergreen, 3 = evergreen), rooting type (ordinal, 1 = long-living primary root system, 2 233
= secondary fibrous roots in addition to the primary root system, 3 = short living primary root 234
system, extensive secondary root system) and rooting depth (cm) as used by Roscher et al. 235
(2004). FDQ was calculated for each of the ten functional traits separately, all available traits 236
simultaneously (FDQ-all) and only leaf-related traits (FDQ-leaf). FDQ and CWM were calculated 237
using the “FD” package (Laliberté and Shipley 2010) available through the R (Team 2013) 238
statistical package version 3.1.2. 239
Statistical analyses 240
Statistical analyses were performed in R (version 3.1.2.). T-tests were used to compare measured 241
and modeled water vapor fluxes. To test for plant species richness effects we used repeated 242
measures analysis of variance (ANOVA) on daily averaged data with dome as a random factor 243
with a temporal autocorrelation function (corAR1) for individual domes as available through the 244
“lme” function. To identify the most important predictors we fitted simple linear regression 245
models on fluxes averaged over six days for each predictor as well as available soil texture-246
related covariates (see Appendix A) on the averaged water vapor fluxes of the selected six days. 247
The resulting models were then simultaneously run through a model averaging procedure 248
(dredge function in MuMIn package) which computes Akaike weights (AICw), that represent the 249
probability that a particular model is the best fit to the observed data(Burnham and Anderson 250
2001). The predictors found within the 95% confidence interval (cumulative AICw ≤ 0.95) were 251
Milcu et al 2015 - main manuscript
12/29
selected and simultaneously included in a multivariate linear fitting regression model using the 252
lm function. The resulting regression models were then simplified to reach the most 253
parsimonious models by using the automatic model simplification “step” procedure based on 254
AICc (Venables and Ripley 2002). 255
256
RESULTS 257
Results of the Shuttleworth and Wallace model 258
Time traces of ecosystem soil evaporation (E) and plant transpiration (T) fluxes modeled with 259
the SW model are displayed in Fig. 2a, 2b as averages for the two plant species richness levels. 260
Although over 24h the modeled ET slightly underestimated the measured ET (Fig. 2c), when 261
restricted to only day-time, the model agreed well with the measured data (Fig. 2d), with average 262
day-time ET over six days within ± 10% of measured values for most of the lysimeters (r2modeled, 263
measured = 0.79) and a Root Mean Square Error of Approximation (RMSEA) of 1.45. T-tests based 264
on average daily ET showed that there was no significant difference between measured and 265
modeled day-time (t = -0.58, P = 0.568) and 24h ET (t = 1.33, P = 0.197). 266
As expected, modeled night-time evapotranspiration (ETnight) values were lower than day-267
time values (Fig. 2a). However, we found that the modeled ETnight values were on average 52% 268
lower (t = 6.37, P < 0.001) than the values measured with the lysimeters (Fig. 2b). Since we 269
argue that discrepancy between the measured and modeled ETnight is likely related to the 270
controlled environment in the Ecotron (see Appendix F), we restricted the further assessment of 271
the effect of species richness and functional diversity on the modeled T and E to day-time only. 272
273
Species richness effects 274
Milcu et al 2015 - main manuscript
13/29
Measured day-time ecosystem evapotranspiration (ETday) was significantly higher in the 275
lysimeters hosting monoliths from communities sown with 16 species relative to 4 species 276
(+16.6%, F1/10, P = 0.049, Fig. 2a). This is also visible in Fig. 1c with exception of the last day 277
after the rainfall/ irrigation of 20mm. However, no significant difference in ETday was found 278
between species richness during night-time (F1/20 = 0.25, P = 0.625), or when the ET was 279
integrated over 24h (F1/10 = 2.84, P = 0.122, Fig 2a). 280
Modeled ETday was also significantly higher in the treatment with 16 species relative to 4 281
species richness (+14.5%, F1/10 = 6.86, P = 0.026; Fig. 3b), a result in accordance with the 282
measured results. Averaged over 24h, modeled ET was also significantly higher in the treatment 283
16 sown species richness (+ 14.6%, F1/10 = 6.62, P = 0.028), while no diversity effect was found 284
for ETnight. 285
Modeled ecosystem transpiration during day-time (Tday), either expressed as fraction of 286
the available energy (+65.2%, F1/10 = 6.77, P = 0.026, Fig. 2c) or mean water vapor loss per day 287
(+68.4%, F1/10 = 6.79, P = 0.026, Fig. 2d) was significantly higher in the monoliths with higher 288
plant species richness (Table 1). Furthermore, Tday correlated well with the total plant biomass at 289
the final harvest (r2 = 0.72, P < 0.001; Fig. 3f). In contrast, modeled day-time evaporation (Eday) 290
was significantly lower at high species richness expressed as both fraction of available energy (-291
22.2%, F1/10 = 6.09, P = 0.033, Fig. 2d) and mean water vapor loss per day (-15.9%, F1/10 = 6.45, P 292
= 0.029, Fig. 2d). Consequently, the ratio of day-time ecosystem transpiration to evaporation 293
(Tday /Eday) was also significantly affected by sown plant species richness, with a significantly 294
higher ratio in the monoliths with 16 sown species (F1/10 =7.28, P = 0.022, Fig 2c). The 295
proportion of total energy dissipated as sensible heat flux (Hday) was significantly lower (-28.4%; 296
F1/10 = 7.30, P = 0.022) in the monoliths with 16 compared to 4 sown plant species (Table 1). 297
Milcu et al 2015 - main manuscript
14/29
Vegetation properties and functional trait-based metrics as predictors for water vapor 298
fluxes 299
With one exception (ETnight, which was best predicted by the community weighted means of the 300
height of the species), among the vegetation properties, LAI proved to be the most consistent 301
predictor of water vapor fluxes (Table 1). Fig. 3e depicts the role of LAI in the modeled partition 302
of the available energy in Tday, Eday and Hday. Higher LAI also led to higher measured ET24h and 303
ETday (Table 1). Alongside LAI, the percentage of bare ground was retained in the most 304
parsimonious models for all of the modeled variables and influenced positively the Eday and Hday 305
and negatively the Tday and the Tday/Eday ratio. 306
The minimal adequate models for water vapor fluxes that directly focused on functional 307
trait based metrics had generally a weaker predictive power than those based on the LAI (Table 308
1). However, the diversity of leaf N concentration in the canopy (FDQ-leafN%) stood out as a 309
consistent predictor of ecosystem water vapor fluxes; increasing FDQ-leafN% was correlated 310
with an increase in measured and modeled fluxes. The relationships between FDQ-leafN%, LAI, 311
percentage bare ground and modeled daytime ecosystem transpiration (Tday) are depicted in 312
Figure 4. 313
314
DISCUSSION 315
Ecosystem ET is an important process for water and energy balance, and is closely linked to 316
ecosystem productivity (Hu et al. 2008, Verheyen et al. 2008). The Shuttleworth and Wallace’s 317
(SW) energy partitioning model has been widely used with good performance to partition the ET 318
in its component water vapor fluxes (T and E), in crops (Shuttleworth and Wallace 1985, Brisson 319
et al. 1998) and natural grasslands (Hu et al. 2009), but it has not been so far used in experiments 320
Milcu et al 2015 - main manuscript
15/29
addressing the role of biodiversity for ecosystem functioning. Here, we applied the SW model to 321
plant communities of contrasting species richness (4 vs. 16 species) in a controlled environment 322
facility (Ecotron) equipped with lysimeters. This setup allowed us to compare the performance of 323
the SW model with the lysimeter measured fluxes. Considering the large daily variability of 324
environmental variables, and particularities of the Ecotron environment (e.g. top down air flow), 325
the overall agreement of measured and modeled ET during the six days of our study was good (r2 326
= 0.79), and within the range found by previous studies (Brisson et al. 1998, Tourula and 327
Heikinheimo 1998, Kato et al. 2004). 328
Overall, our results show an increase of water acquisition in the high diversity treatment 329
(16 plant species) as indicated by higher ETday as well as higher modeled Tday. Although 330
speculative, since we provide no direct evidence for complementarity, increased water 331
acquisition can be interpreted as an indicator of complementarity, however, the jury is still out on 332
the importance of complementarity for grassland water acquisition (Verheyen et al. 2008, 333
Bachmann et al. 2015). In parallel with the higher modeled Tday—the productive fraction of the 334
ET— communities with 16 species also exhibited lower proportions of energy diverted into non-335
productive water and energy fluxes (Eday and Hday) and an increase in the ratio of energy diverted 336
to transpiration over evaporation during daytime (Tday/Eday; Table 1). This suggests that plant 337
diversity effects mediated by the structure of the canopy (by the LAI as indicated by Table 1) 338
occurred by reducing the propensity to dissipate the energy via the non-productive evaporative 339
process. Since Tday occurs only during the period of photosynthetic activity, a higher proportion 340
the T implies that more of the ET water was used to acquire carbon, which would increase the 341
water use efficiency (C gain/ET). Indeed, this is consistent with the results of Milcu et al. (2014) 342
from same experiment in the Ecotron showing a 37.6% increase in water use efficiency in 343
Milcu et al 2015 - main manuscript
16/29
communities with 16 plant species relative to 4 species. These findings are also in line with 344
several previous studies (Van Peer et al. 2004, De Boeck et al. 2006, Lemmens et al. 2006) 345
documenting an increase in ET and water use efficiency with increasing plant diversity. 346
However, our results are not in agreement with the findings of Leimer et al. (2014) which 347
indirectly estimated ecosystem ET using a simple soil water balance model in the same filed site 348
from where the soil monoliths of our study have been extracted (the Jena Experiment). By 349
modelling ecosystem ET from fluctuations in soil water content measured in two soil layers (0.0 350
– 0.3 m and 0.3 – 0.7 m), the study of Leimer et al. (2014) fund no plant diversity effects on 351
modeled ET with monthly resolution from 2005 to 2008. This is different from our results which 352
showed significantly higher modeled ET as well as a tendency of higher ET in measured ET. 353
While not fully comparable due to very different methodology and temporal resolution, it is 354
worth noting that the Leimer et al. (2014) study was only calibrated/ tested against the results of 355
a hydrological model and not against measured ET. Another possible explanation for the 356
discrepancy could be that, by using a controlled environment approach and direct measurements 357
of ET, we were able to reduce more confounding factors. Alternatively, we cannot discount the 358
possibility that the study of Leimer et al. (2014), which presented monthly aggregated data for 359
four years, also incorporated longer-term processes that are not included in our study. For 360
example, unaccounted potential discrepancies in the phenology of plant development throughout 361
the growing season might alter the differences in water use between the communities with low 362
and high diversity. 363
Of all tested predictors of water vapor fluxes based on vegetation properties such as 364
canopy structure and belowground rooting patters we found that the LAI and the percentage of 365
bare ground on the soil surface were the two predictors retained in the most parsimonious 366
Milcu et al 2015 - main manuscript
17/29
models. While it has been long established that the LAI is a consistent driver of water vapor 367
fluxes, and is actually factored into the SW energy partition model (Shuttleworth and Wallace 368
1985), here we show that plant diversity effects on ecosystem water vapor fluxes are mediated by 369
the leaf area index (LAI). Furthermore, to our knowledge, this is the first study attempting to link 370
ecosystem water vapor fluxes to plant functional trait-based metrics. By doing this we found that 371
of all functional trait-based metrics, the index of functional diversity based on leaf nitrogen 372
concentration in the canopy (FDQ-leafN%) predicted best the water vapor fluxes and was 373
strongly correlated to the LAI (Fig. 4). Moreover, it was a better predictor of measured day-time 374
ET than species richness. Previous studies from the same experiment showed that FDQ-leafN% 375
was an important predictor of C fluxes and community biomass (Roscher et al. 2012, 2013, 376
Milcu et al. 2014), and it was suggested it might capture the cumulative investment in light 377
acquisition by the community represented by the LAI, and hence, indirectly affect the 378
ecosystem-level plant transpiration. Alternatively, it has also been suggested that this index of 379
diversity measuring the unevenness of N allocation in the canopy could be linked to a more 380
optimal N distribution in the canopy (Field 1983, Hirose and Werger 1987, Milcu et al. 2014) 381
and not necessarily with the LAI. However, additional manipulative experiments are needed to 382
clarify the exact mechanism through which FDQ-leafN% affects the water vapor fluxes. 383
To conclude, our results show higher water acquisition (higher ET) in more diverse 384
communities during day-time, at least at the peak at the growing season. More novel, we 385
identified a plant functional diversity metric capturing the diversity of leaf N% in the canopy 386
(FDQ-leafN%) as a consistent predictor of several day-time ecosystem-level water and energy 387
fluxes (ET, T, E and H), presumably via its effect on LAI, a measure of the community 388
investment in light acquisition. Furthermore, since the SW model is showing that higher plant 389
Milcu et al 2015 - main manuscript
18/29
diversity leads to a higher T rates – a flux closely coupled with photosynthesis and hence 390
productivity-related – this presumably contributes to the well-documented positive effect of 391
diversity on productivity. This is also supported by the positive correlation between modeled Tday 392
and total plant biomass (Fig. 3f). Allocating more of the available water to the productivity-393
related flux is likely going to become more important in the future, since extreme climatic events 394
such as droughts and heatwaves are predicted to increase in intensity and frequency in many 395
regions (Seneviratne et al. 2012). Under these conditions, plant communities containing more 396
functionally diverse species will likely use more of the available water for growth and cooling. 397
398
Acknowledgements 399
We thank DFG for funding the Jena-Ecotron experiment (FOR 456) and A. Milcu. A. 400
Hildebrandt and M. Guderle were supported by the ProExzellenz initiative of the German 401
Federal State of Thuringia. M. Guderle’s visit to the Ecotron was supported by the ExpeER 402
Transnational Access program (EU-FP7 I3). D. Bachmann and N. Buchmann acknowledge 403
funding of the SNF (315230E-131194). This study benefited from the CNRS human and 404
technical resources allocated to the ECOTRONS Research Infrastructures and the state allocation 405
'Investissement d'Avenir’ ANR-11-INBS-0001. The Languedoc-Roussillon region and the 406
Hérault Conseil Général are acknowledged for funding and supporting the Ecotron. 407
408
409
410
411
Milcu et al 2015 - main manuscript
19/29
References 412
Bachmann, D., et al. 2015. No evidence of complementary water use along a plant species 413
richness gradient in temperate experimental grasslands. Plos One 10:e0116367. 414
Blanc, P., B. Gschwind, M. Lefèvre, and L. Wald. 2011. The HelioClim project: Surface solar 415
irradiance data for climate applications. Remote Sensing 3:343–361. 416
De Boeck, H. J., et al. 2006. How do climate warming and plant species richness affect water use 417
in experimental grasslands? Plant and soil 288:249–261. 418
Botta Dukát, Z. 2005. Rao’s quadratic entropy as a measure of functional diversity based on 419
multiple traits. Journal of Vegetation Science 16:533–540. 420
Brisson, N., B. Itier, J. C. L’Hotel, and J. Y. Lorendeau. 1998. Parameterisation of the 421
Shuttleworth-Wallace model to estimate daily maximum transpiration for use in crop models. 422
Ecological Modelling 107:159–169. 423
Burnham, K., and D. R. Anderson. 2001. Model Selection and Multimodel Inference. 424
Cadotte, M., J. Cavender-Bares, D. Tilman, and T. Oakley. 2009. Using phylogenetic, functional 425
and trait diversity to understand patterns of plant community productivity. PLoS One 4:843–426
845. 427
Caldeira, M. C., R. J. Ryel, J. H. Lawton, and J. S. Pereira. 2001. Mechanisms of positive 428
biodiversity-production relationships: Insights provided by δ13C analysis in experimental 429
Mediterranean grassland plots. Ecology Letters 4:439–443. 430
Cardinale, B. J., et al. 2012. Biodiversity loss and its impact on humanity. Nature 486:59–67. 431
Cardinale, B. J., et al. 2007. Impacts of plant diversity on biomass production increase through 432
time because of species complementarity. Proceedings of the National Academy of Sciences 433
104:18123. 434
Milcu et al 2015 - main manuscript
20/29
Díaz, S., S. Lavorel, F. de Bello, F. Quétier, K. Grigulis, and T. M. Robson. 2007. Incorporating 435
plant functional diversity effects in ecosystem service assessments. Proceedings of the 436
National Academy of Sciences (USA) 104:20684–20689. 437
Duffy, J. E. 2003. Biodiversity loss, trophic skew and ecosystem functioning. Ecology Letters 438
6:680–687. 439
von Felten, S., and B. Schmid. 2008. Complementarity among species in horizontal versus 440
vertical rooting space. Journal of Plant Ecology 1:33–41. 441
Field, C. 1983. Allocating leaf nitrogen for the maximization of carbon gain: leaf age as a control 442
on the allocation program. Oecologia 56:341–347. 443
Grime, J. P. 1998. Benefits of plant diversity to ecosystems: immediate, filter and founder 444
effects. Journal of Ecology 86:891–899. 445
Hirose, T., and M. J. A. Werger. 1987. Maximizing daily canopy photosynthesis with respect to 446
the leaf nitrogen allocation pattern in the canopy. Oecologia 72:520–526. 447
Hooper, D. U., et al. 2012. A global synthesis reveals biodiversity loss as a major driver of 448
ecosystem change. Nature 486:105–8. 449
Hu, Z., et al. 2008. Effects of vegetation control on ecosystem water use efficiency within and 450
among four grassland ecosystems in China. Global Change Biology 14:1609–1619. 451
Hu, Z., et al. 2009. Partitioning of evapotranspiration and its controls in four grassland 452
ecosystems: Application of a two-source model. Agricultural and Forest Meteorology 453
149:1410–1420. 454
Hyer, E. J., and S. J. Goetz. 2004. Comparison and sensitivity analysis of instruments and 455
radiometric methods for LAI estimation: assessments from a boreal forest site. Agricultural 456
and Forest Meteorology 122:157–174. 457
Milcu et al 2015 - main manuscript
21/29
Kato, T., R. Kimura, and M. Kamichika. 2004. Estimation of evapotranspiration, transpiration 458
ratio and water-use efficiency from a sparse canopy using a compartment model. Agricultural 459
Water Management 65:173–191. 460
Laliberté, E., and B. Shipley. 2010. FD: Measuring functional diversity (FD) from multiple traits, 461
and other tools for functional ecology. R package version 1.0 9. 462
Leimer, S., et al. 2014. Plant diversity effects on the water balance of an experimental grassland. 463
Ecohydrology 1391:1378–1391. 464
Lemmens, et al. 2006. End-of-season effects of elevated temperature on ecophysiological 465
processes of grassland species at different species richness levels. Environmental and 466
Experimental Botany 56:245–254. 467
Milcu, A., et al. 2014. Functional diversity of leaf nitrogen concentrations drives grassland 468
carbon fluxes. Ecology Letters 17:435–444. 469
Monteith, J. 1981. Evaporation and surface temperature. Quarterly Journal of the Royal 470
Meteorological Society 107:1–27. 471
Obrist, D., P. S. J. Verburg, M. H. Young, J. S. Coleman, D. E. Schorran, and J. A. Arnone. 472
2003. Quantifying the effects of phenology on ecosystem evapotranspiration in planted 473
grassland mesocosms using EcoCELL technology. Agricultural and Forest Meteorology 474
118:173–183. 475
Osnas, J. L., J. W. Lichstein, P. B. Reich, and S. W. Pacala. 2013. Global leaf trait relationships: 476
mass, area, and the leaf economics spectrum. Science 340:741–744. 477
Van Peer, L., I. Nijs, D. Reheul, and B. De Cauwer. 2004. Species richness and susceptibility to 478
heat and drought extremes in synthesized grassland ecosystems: compositional vs 479
physiological effects. Functional Ecology 18:769–778. 480
Milcu et al 2015 - main manuscript
22/29
Penman, H. L. 1948. Natural evaporation from open water, bare soil, and grass. Proceedings of 481
the Royal Society. Series A, Mathematical and Physical Sciences 193:120–145. 482
Reichstein, M., M. Bahn, M. D. Mahecha, J. Kattge, and D. D. Baldocchi. 2014. Linking plant 483
and ecosystem functional biogeography. Proceedings of the National Academy of Sciences 484
(USA) 111:201216065. 485
Reiss, J., J. R. Bridle, J. M. Montoya, and G. Woodward. 2009. Emerging horizons in 486
biodiversity and ecosystem functioning research. Trends in Ecology & Evolution 24:505–514. 487
Roscher, C., et al. 2004. The role of biodiversity for element cycling and trophic interactions: an 488
experimental approach in a grassland community. Basic and Applied Ecology 5:107–121. 489
Roscher, C., et al. 2012. Using plant functional traits to explain diversity–productivity 490
relationships. PLoS One 7:e36760. 491
Roscher, C., et al. 2013. A functional trait-based approach to understand community assembly 492
and diversity–productivity relationships over 7 years in experimental grasslands. Perspectives 493
in Plant Ecology, Evolution and Systematics. 494
Seneviratne, S. I., et al. 2012. Changes in Climate Extremes and their Impacts on the Natural 495
Physical Environment. Managing the Risks of Extreme Events and Disasters to Advance 496
Climate Change Adaptation. 497
Shuttleworth, W. J., and J. S. Wallace. 1985. Evaporation from sparse crops-an energy 498
combination theory. Quarterly Journal of the Royal Meteorological Society 111:839–855. 499
Stocker, R., C. Körner, B. Schmid, P. Niklaus, and P. Leadley. 1999. A field study of the effects 500
of elevated CO2 and plant species diversity on ecosystem-level gas exchange in a planted 501
calcareous grassland. Global Change Biology 5:95–105. 502
Stull, R. B. 1988. An Introduction to Boundary Layer Meteorology. Kluwer. 503
Milcu et al 2015 - main manuscript
23/29
Team, R. 2013. R Development Core Team. R: A Language and Environment for Statistical 504
Computing. 505
Tilman, D., C. L. Lehman, and K. T. Thomson. 1997. Plant diversity and ecosystem 506
productivity: theoretical considerations. Proceedings of the National Academy of Sciences 507
(USA) 94:1857–61. 508
Tourula, T., and M. Heikinheimo. 1998. Modelling evapotranspiration from a barley field over 509
the growing season. Agricultural and Forest Meteorology 91:237–250. 510
Venables, W. N., and B. D. Ripley. 2002. Modern Applied Statistics with S. 511
Verheyen, K.,et al. 2008. Can complementarity in water use help to explain diversity–512
productivity relationships in experimental grassland plots? Oecologia 156:351–361. 513
Violle, C., et al. 2007. Let the concept of trait be functional! Oikos 116:882–892. 514
Wesely, M. L. 1989. Parameterization of surface resistances to gaseous dry deposition in 515
regional-scale numerical models. Atmospheric Environment 23:1293–1304. 516
Wilson, P. J., K. Thompson, and J. G. Hodgson. 1999. Specific leaf area and leaf dry matter 517
content as alternative predictors of plant strategies. New Phytologist 143:155–162. 518
Wright, I. J., et al. 2004. The worldwide leaf economics spectrum. Nature 428:821–827. 519
520
521
522
523
524
525
Milcu et al 2015 - main manuscript
24/29
Tables: Table 1. 526
ANOVA results for the effect of sown species richness (Sdiv, i.e. 4 vs. 16 species) alongside the 527
most parsimonious multiple regression models. ET24h, ETday and ETnight represent lysimeter 528
measured evapotranspiration values during 24h, day-time and night-time, respectively. Tday, Eday 529
and Hday represent the day-time transpiration, evaporation and sensible heat flux, respectively. 530
Vegetation properties (LAI = leaf area index, Bare = % bare ground, hrGR = biomass of grasses) 531
and functional trait-based predictors (FDQ-leafN% = diversity of leaf N concentration, CWM-532
height = community weighted means of species height) were separately analyzed in different 533
models. Non-significant effects are marked as “ns”. See Appendix A for the list of all predictors 534
included in the initial models before the AICc-based model simplification. 535
Response
variables
Sdiv
(ANOVA)
Best model based on canopy, root and
soil texture
Best model based on
functional-trait predictors
Measured
ET24h ns 2.788 + 0.342*LAI
P < 0.001, r2 = 0.78
ns
ETday P = 0.049
r2 = 0.34
2.138 + 0.344*LAI
P < 0.001, r2 = 0.93
2.46 + 0.67*FDQ-leafN%
P = 0.015, r2 = 0.46
ETnight ns 0.700 - 0.002*grBM
P = 0.094, r2=0.26
0.883 – 0.111*CWM-height
P = 0.015, r2 = 0.46
Modeled
Tday P = 0.026
r2 = 0.40
1.109 + 0.28*LAI – 0.017*Bare
P < 0.001, r2 = 0.99
0.893+ 0.918*FDQ-leafN%
P = 0.040, r2 = 0.36
Eday P = 0.029
r2 = 0.39
1.642 – 0.101*LAI + 0.006*Bare
P < 0.001; r2 = 0.97,
1.706 – 0.295*FDQ-leafN%
P = 0.076, r2 = 0.29
Tday /Eday P = 0.022
r2 = 0.42
0.597 + 0.276*LAI – 0.011*Bare
P < 0.001, r2 = 0.99
0.541 + 0.769*FDQ-leafN
P = 0.034, r2 = 0.37
Hday P = 0.022
r2 = 0.42
0.319 – 0.032*LAI+0.002*Bare
P < 0.001, r2 = 0.98
0.357– 0.111*FDQ-leafN%
P = 0.054, r2 = 0.32
Milcu et al 2015 - main manuscript
25/29
Figure legends 536
Figure 1. (a) Ecotron environmental conditions during the selected six days averaged for the two 537
plant diversity levels (4 vs. 16 species). RH represents the air relative humidity and VPD the 538
vapor pressure deficit. (b) Lysimeter measured evapotranspiration (ET) in the Ecotron. 539
Figure 2. Time trace of ecosystem evaporation (E) and transpiration (T) fluxes as modeled by 540
the SW model in (a) communities with 16 and (b) 4 plant species. T (grey area) and E (black 541
area) are plotted additively so that their sum represents total ET. Comparison of averaged 542
modeled and lysimeter ET with (c) and without (d) night-time values. 543
Figure 3. Plant diversity (4 vs. 16 species) effects on day-time, night-time and 24h 544
evapotranspiration fluxes (a) measured in lysimeters and (b) modeled by the SW model. (c) Plant 545
diversity effects on modeled partition of the available radiative energy in evaporation (E) 546
transpiration (T) and sensible heat flux (H). (d) Modeled E and T as affected by plant diversity. 547
(e) LAI and plant diversity effects on modeled E (squares), T (circles) and H (triangles). (f) 548
Relationship between modeled day-time T and plant biomass. 549
Figure 4. Scatterplot and correlation matrix depicting the relationships between the most 550
important predictors (FDQ-leafN% = diversity of leaf N concentration, LAI, % bare ground) of 551
water vapor fluxes and modeled daytime ecosystem transpiration (Tday). The line in the 552
scatterplots represents a locally weighted scatter-plot smoother (loess) fitting. 553
Milcu et al 2015 - main manuscript
26/29
Figure 1 554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
Milcu et al 2015 - main manuscript
27/29
Figure 2 570
571
572
573
574
575
576
Milcu et al 2015 - main manuscript
28/29
Figure 3 577
578
Milcu et al 2015 - main manuscript
29/29
Figure 4 579
580
581
582
583
584
585
Milcu et al. 2015 – appendices
1/12
Appendices (A to F) 586
Appendix A. Overview table presenting the mean ± 1*SD of the tested predictors and covariates for the two levels of sown species 587
richness. FDQ abbreviation indicates a functional diversity index estimated using Rao’s quadratic entropy index (Botta Dukát 2005). 588
CMW indicates a community weighted means estimated for the respective functional trait. 589
Functional
trait-based
metrics
Description Type of trait / source Unit 4 species
(mean ± SD)
16 species
(mean±SD)
FDQ-leafN%, based on leaf N% continuous / measured in situ dimensionless 0.40 ± 0.28 0.79 ± 0.42
CWM-leafN% based on leaf N% continuous / measured in situ mg N g-1 leaf
DW
2.55 ± 0.67 2.36 ± 0.35
FDQ-SLN based on N specific leaf area continuous / measured in situ dimensionless 0.17 ± 0.25 0.54 ± 0.34
CWM-SLN based on N specific leaf continuous / measured in situ g N m−2 leaf 1.50 ± 0.25 1.26 ± 0.34
FDQ-SLA, based on specific leaf area continuous / measured in situ dimensionless 0.14 ± 0.14 0.49 ± 0.22
CWM-SLA based on specific leaf area continuous / measured in situ mm2 mg-1 16.91 ± 2.25 19.77 ± 2.77
FDQ-GS, based on stomatal conductance continuous / measured in situ dimensionless 0.57 ± 0.72 0.77 ± 0.67
CWM-GS based on stomatal conductance continuous / measured in situ µmol m-2 s-1 469 ± 155 557 ± 167
FDQ-greenness, based on leaf greenness continuous / measured in situ dimensionless 0.16 ± 0.24 0.60 ± 0.29
CWM-greenness based on leaf greenness continuous / measured in situ dimensionless 37.56 ± 7.36 34.86 ± 3.41
FDQ-LDMC, based on leaf dry matter content continuous / measured in situ dimensionless 0.42 ± 0.30 0.92 ± 0.35
CWM-LDMC based on leaf dry matter content continuous / measured in situ mg g-1 231.55 ± 41.38 229.38 ± 31.47
FDQ-height, based on recorded species height continuous / measured in situ dimensionless 0.25 ± 0.35 0.58 ± 0.50
CWM-height based on recorded species height continuous / measured in situ cm 15.7 ± 10.48 16.6 ± 4.02
FDQ-rootdepth, based on rooting depth continuous / literature dimensionless 0.73 ± 0.88 0.67 ± 0.56
CWM-rootdepth based on rooting depth continuous / literature cm 3.51 ± 0.74 3.70 ± 0.35
FDQ-typeroot, based on rooting type ordinal / literature dimensionless 0.40 ± 0.17 0.74 ± 52
CWM-typeroot based on rooting type ordinal / literature NA 2.55 ± 0.30 2.37 ± 0.31
Milcu et al. 2015 – appendices
2/12
590
FDQ-leaf based on all six leaf related traits continuous / measured in situ dimensionless 2.05 ± 1.15 4.77 ± 1.76
FDQ-all based on all ten functional traits continuous and ordinal / in situ
and literature
dimensionless 3.44 ± 1.73 6.77 ± 2.41
Vegetation
properties
Description Type of variable / source
Sdiv sown species richness categorical (4 and 16 species) /
in situ
counts 4 ± 0 16 ± 0
Rdiv realized species richness at July
harvest
counts / measured in situ counts 3.16 ± 0.75 11.3 ± 1.5
LAI leaf area index continuous / measured in situ dimensionless 1.46 ± 0.97 2.78 ± 0.84
%bare percentage bare ground percentage / measured in situ % 26.83 ± 20.50 6.25 ± 9.23
ShootBM shoot biomass continuous / measured in situ g DW m-2 148.71 ± 61.07 193.01 ± 43.31
RootBM total root biomass continuous / measured in situ g DW m-2 100.52 ± 39.87 154.07 ± 37.77
RootBM/L root biomass per layer (0-5, 5-10,
10-20, 20-30, 30-40, 40-60cm)
continuous / measured in situ g DW m-2 layer dependent layer dependent
RootS/V root surface per volume soil (by
layer)
continuous / measured in situ cm2/cm3 layer dependent layer dependent
TotalBM total biomass (shoot + root biomass) continuous / measured in situ g DW m-2 249.24 ± 93.34 347.08 ± 55.75
LegBM biomass of legumes continuous / measured in situ g DW m-2 18.24 ± 24.65 42.86 ± 40.27
Leaf biomass biomass of leaves continuous / measured in situ g DW m-2 93.89 ± 50.93 124.21 ± 28.62
Leaf-to-
ShootBM-ratio
ratio of leaf to shoot biomass ratio / measured in situ dimensionless 0.63 ± 0.14 0.67 ± 0.09
HerBM biomass of herbs continuous / measured in situ g DW m-2 83.15 ± 71.47 95.58 ± 54.15
GraBM biomass of grasses continuous / measured in situ g DW m-2 34.12 ± 61.66 36.81 ± 52.14
Covariates Description Type of variable / source
Soil sand % soil sand content at 0-20 cm depth percentage / measured in situ % 21.16 ± 17.63 17.87 ± 15.89
Soil clay % soil clay content at 0-20 cm depth percentage / measured in situ % 26.09 ±7.09 24.42 ± 6.25
Soil silt % soil silt content at 0-20 cm depth percentage / measured in situ % 52.74 ± 10.94 57.70 ±12.28
Milcu et al. 2015 – appendices
3/12
Appendix B. Top image: simplified schematic of one controlled environment unit (macrocosm) 591
of the CNRS Ecotron facility. Red arrows represent the direction of airflow (image by Marc 592
Saubion). Bottom image: closeup of the vegetation from dome no. 1 at the July harvest (photo by 593
Alexandru Milcu). 594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
Milcu et al. 2015 – appendices
4/12
Appendix C. Table with sown species richness (Sdiv) and functional group (G = grasses, H = 620
herbs and L = legumes) composition of the twelve selected plots from the Jena Experiment. The 621
species present at the final harvest are marked in bold. 622
Grasses (G): Ae = Arrhenatherum elatius L. (J. et C. PRESL), Ao= Anthoxantum odoratum L., 623
Ap= Alopecurus pratensis L., AVp = Avenula pubescens HUDS. (DUM.), Be= Bromus erectus 624
HUDS., Bh= Bromus hordeaceus L., Cc= Cynosurus cristatus L., Dg= Dactylis glomerata L., 625
Fp= Festuca pratensis HUDS., Fr= Festuca rubra L., Hl= Holcus lanatus L., LUc = Luzula 626
campestris L (DC.), PHp= Phleum pratense L., Pp= Poa pratensis L., Pt= Poa trivialis L., Tf= 627
Trisetum flavescens L. (P. BEAUV.); 628 629
Herbs (H): Am= Achillea millefolium L., Bp= Bellis perennis L., Cb= Crepis benis L., CAc= 630
Carum carvi L., CAp = Campanula patula L., Cj= Centaurea jacea L., Co= Cirsium oleraceum 631
L., Cp= Cardamine pratensis L., Dc= Daucus carota L., Gm= Galium mollugo L., Gh= 632
Glechoma hederacea L., Hs = Heracleum sphondylium L., Ka= Knautia arvensis L., La= 633
Leontodon autumnalis L., Lh= Leontodon hispidus L., Lv= Leucanthemum vulgare Lam., PIm = 634
Pimpinella major L. (HUDS.), Pl= Plantago lancelolata L., Pm= Plantago media L., PRv = 635
Primula veris L., Pv= Prunella vulgaris L., Ra= Rumex acetosa L., To= Taraxacum officinale 636
WEBER, TRp= Tragopogon pratensis L., Vc= Veronica chamaedrys L; 637
638
Legumes (L): Lp= Lathyrus pratensis L., Lc= Lotus corniculatus L., Ml= Medicago lupulina L., 639
Ms=Medicago x varia MARTYN, Ov= Onobrychis viciifolia SCOP., Td= Trifolium dubium 640
SIBTH., TRf = Trifolium fragiferum L., Th= Trifolium hybridum L., Tr= Trifolium repens L., 641
Tp= Trifolium pratense L., VIc= Vicia cracca L. 642
643
Plot
ID Sdiv G H L Species composition
Ecotron
dome
B2A22 16 5 5 6 Cc, Fp, Tf, Pp, Pt, Cj, Ra, So, Am, CAp , Th, Lc, VIc, Tr, Lp, Ov
1
B4A04 4 1 2 1 Ae, Pl, As, Tc
2
B1A01 16 4 8 4 AVp, Pp, Ao, Bh, Pl, To, Ar, Rr, As, Gp, TRp, CAc, Tc, VIc, Lp, Lc
3
B1A04 4 1 2 1 Fp, Pl, CAp, Ov
4
B3A23 4 1 2 1 Bh, Rr, Lv, TRf
5
B2A18 16 4 8 4 Ap, Bh, Pp, Cc, Rr, Pm, Ar, Pv, CAp, Gp, As, Cp, Ml, Tr, Td, Tc
6
B4A18 16 4 8 4 Cc, LUc, Ap, Bh, La, Pm, Vc, To, Cb, CAc, PIm, Hs, Th, Tc, Lp, Ov
7
B2A01 4 1 2 1 Ao, Pv, Ka, Tp
8
B3A22 16 4 8 4 PHp, Fr, Ao, Be, Rr, Ar, Bp, Vc, Gp, Cb, Ra, Gm, VIc, Ov, TRf, Td
9
B2A16 4 0 3 1 Pm, La, Ka, VIc
10
B3A24 16 6 5 5 Fp, Bh, Ap, Ao, Pt, Ae, To, Rr, Ar, Pv, Gh, Lc, Tp, Tr, VIc, Ms
11
B4A11 4 1 2 1 Tf, TRp, Hs, Ms
12
Milcu et al. 2015 – appendices
5/12
Appendix D. Comparison of cumulative precipitation (mm) recorded in the “Jena Experiment” 644
field site in 2007 and simulated in the Ecotron in 2012. 645
646
647
648
649
650
651
652
653
654
655
656
0
50
100
150
200
250
300
350
4002
6/0
3
09
/04
23
/04
07
/05
21
/05
04
/06
18
/06
02
/07
16
/07
30
/07
Pre
cip
itat
ion
(m
m)
Jena 2007
Ecotron
Time
Milcu et al. 2015 – appendices
6/12
Appendix E. Comparison of the soil moisture achieved in the Ecotron and the “Jena 657
Experiment” field site in 2007 at similar depths for the period with available Ecotron data form 658
2012. 659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
Milcu et al. 2015 – appendices
7/12
Appendix F - Supplementary Methods 691
Extraction of soil monoliths 692
Soil monoliths were extracted by UMS GMBH (München, Germany) following a proprietary 693
non-compacting extraction method. Using a hydraulic press, steel cylinders with cutting edges 694
were pressed down to 2 m depth (to include the average depth of the water table in summer 695
known from previous hydrological field investigations) while simultaneously digging out the soil 696
60 cm around the cylinder. The created space around the cylinders allowed to cut the bottom of 697
the soil monolith and to insert a sealing plate in order to be able to lift the monolith and to create 698
a water tight container/ lysimeter. The lysimeters were then extracted with a crane, and after 699
inspection, were buried to the surface level near the experimental field after extraction in order to 700
facilitate the recovery after the extraction disturbance, while being exposed to the same 701
environmental conditions as the field plots. At the end of March 2012 the lysimeters were 702
transported to the Ecotron facility in Montpellier. 703
704
Measurements of environmental variables 705
Air temperature and relative humidity were measured with the DT269 digital sensor (Michell 706
Instruments Ltd, Ely, UK). The soil moisture was measured with a Trime-Pico 32 time domain 707
reflectometry sensor (IMKO GmbH, Ettlingen, Germany). Solar radiation was measured with a 708
BF5 Sunshine Sensor (Delta-T Devices, Cambridge, UK). 709
710
Sampling methodology 711
Shoot biomass was estimated by clipping the vegetation at ground level in a rectangle of 0.8 x 712
1.0 m per plot and drying at 65°C for three days. Root biomass was estimated from three cores of 713
3.5 cm diameter and 60 cm depth. The soil cores were separated into six layers (0-5, 5-10, 10-20, 714
20-30 and 40-60cm) before they were pooled per layer, washed with tap water and dried 715
following the same procedure as for shoot biomass. Leaf area index (LAI) was estimated using a 716
portable LAI-2000 plant canopy analyzer (LI-COR, Lincoln, USA). LAI was measured in the 717
evening under diffused light conditions with one measurement above the canopy as a reference 718
Milcu et al. 2015 – appendices
8/12
and the average of five measurements near ground level positioned at different places in the 719
center of each lysimeter. The zenith angle was restricted to 0-43o in order to minimize the edge 720
effect (Hyer and Goetz 2004) inherent to a canopy with a surface of 2m2. In lysimeters with a 721
high proportion of one rosulate species (Plantago media) with leaves laying very close to the 722
ground, the values measured with the LAI-2000 were replaced with the average value of two leaf 723
area measurements using a LI-3100 leaf area meter (LI-COR, Lincoln, USA) leaf area meter) 724
from two rectangles of 0.2 × 0.5 m per plot. Percentage vegetation cover and bare ground as well 725
species specific percentage cover were visually estimated for the whole lysimeter (2 m2). 726
Stomatal conductance was measured at midday using a portable porometer (SC-1 Leaf 727
porometer, Decagon Devices, Pullman, USA). Leaf greenness was estimated with a hand-held 728
chlorophyll meter (SPAD-502, Konica-Minolta, Osaka, Japan), which enables non-destructive 729
assessment of leaf greenness by measuring the absorbance by the leaf of two different 730
wavelengths (650 nm and 940 nm). The SPAD values were calibrated (r2 = 0.69) against 731
spectrophotometrically determined chlorophyll concentrations from leaf extracts following the 732
method of Moran (1982). Leaf dry mater content (LDMC) is defined as the ratio of leaf dry mass 733
to saturated fresh mass according to Vile et al. (2005). LDMC was determined with the partial 734
rehydration method following the protocol of Wilson et al. (1999). Samples were put sealed 735
plastic bags promoting rehydration by storing leaves overnight between sheets of moistened 736
tissue paper. Then, samples were blotted dry using tissue paper to remove any surface water and 737
immediately weighed. Samples were oven-dried (65°C, 3 days) and reweighed to get values for 738
dry mass. This procedure gives a good approximation in comparison to the complete rehydration 739
method (Vaieretti et al. 2007). 740
741
The Shuttleworth and Wallace model 742
Evaporation and transpiration were modeled with the approach by Shuttleworth and Wallace 743
(1985). This approach is a further development of the Penman-Monteith model, 744
745
746
747
748
(1)
Milcu et al. 2015 – appendices
9/12
where x denotes the plant canopy compartment of interest. λEx is parameterised once for the 749
canopy top (λEc), once for the soil surface below canopy (λEs), and once for the fraction of bare 750
soil not covered by the canopy (λEb) total evapotranspiration λE is expressed as the sum of 751
transpiration from the plant canopy and evaporation from the soil weighted by their respective 752
surface cover fractions fc, fs and fb, λE = fc∙λEc + fs∙λEs + fb∙λEb, with fs = fc, and fc = 1–fb.. Thus, 753
transpiration T = λEc and evaporation E = λEs + λEb. Rn is net radiation (at canopy top Rnc or soil 754
surface Rns), G is ground heat flux, Δ is the mean rate of change of saturated vapor pressure with 755
temperature, γ is the psychrometric constant, cp is specific heat at constant pressure (1005.5 J kg–756
1 K–1), ρa is the density of air, e and esat are the actual and saturation vapor pressure, and ra and rs 757
are aerodynamic and surface (stomatal) resistances, respectively. 758
759
We used the mass flux of water vapor in our analysis, which is the simple conversion from 760
energy flux densities modeled by the Shuttleworth and Wallace (1985) model to mass flux 761
densities, with plant transpiration T = λEc/Lv, and soil evaporation E = λEs/Lv, with Lv being the 762
latent heat of vaporisation (Lv = 2.501∙106 – 2370∙Tair with air temperature Tair in °C and Lv in J 763
kg–1 (Stull 1988). 764
765
Saturation vaporpressure esat was determined via the Magnus equation for nonfreezing conditions 766
at air temperature Tair measured in °C in each dome (esat = 610.7⨉10[7.5⨉Tair/(235+Tair)], in Pa), and 767
actual vapor pressure e=RH/100∙esat, with relative humidity RH in %. Density of air was 768
computed from Tair and atmospheric pressure P [in Pa], ρa = P/(Tair+273.15)/R, with the gas 769
constant for dry air R = 287.05 J kg–1 K–1. Δ was approximated as Lv∙esat, /Rv/(Tair+273.15)2 770
with the gas constant for water vapor Rv = 461.5 J kg–1 K–1. 771
772
The key parameters used to model λEc, λEs and λEb are the short-wave albedo of canopy (αc) and 773
bare soil (αs= αb ), the fractions of vegetated (fc) and unvegetated (fb) ground surface, the wind 774
speed above the canopy (u), the fraction of that wind speed observed in mid-canopy (ξc) and at 775
the ground surface (ξb and ξs; where soil evaporation takes place; we assumed that ξb = ξc, but ξs 776
≪ ξc), and the soil surface resistance (rss). The albedo was used to compute Rn from available 777
global radiation (Rg) measurements as Rn = (1–α) Rg. It was assumed that 10% of Rn is 778
dissipated to G under a vegetation canopy, and 30% on the bare-soil component. Since 779
Milcu et al. 2015 – appendices
10/12
experiments were done inside a controlled environment facility, we assumed that the dome 780
temperature is not very different from the soil temperature and thus the net longwave component 781
of Rn is negligible. 782
783
The stomatal resistance of plants (rsc) was modeled according to Wesely (1989) and Erisman et 784
al. (1994), 785
786
where ri the minimal stomatal resistance, and Ts leaf surface temperature in °C in the range 0–787
40°C. The values for ri were determined from rsc measurements carried out for each relevant 788
species of each dome at midday (see sampling methodology). The weighted average of rsc was 789
used to parameterize ri. Weighting was by the proportion of leaf biomass of each plant species, 790
individually for each of the twelve plant communities. 791
792
Wind speed posed some challenges since the conditions in the domes of the Ecotron facility does 793
not represent the typical relationships between horizontal wind speed and mechanical turbulence 794
as the Penman-Monteith model implicitly assumed. The air condition systems of the Ecotron’s 795
domes created a turbulent environment, where wind speed varied between between 0.7–2.5 m s–1 796
in a fraction of a second, and with averaged (during a few seconds) anemometer readings 797
(Almemo 2890-9, Coalville, UK) of 0.9-1 m s-1. Video records however clearly showed that the 798
turbulent motion of the plants in the domes rather corresponded to the turbulence observed at a 799
higher wind speed. Hence, horizontal wind speed was set to 2 m s-1. The wind speed above bare 800
soil patches was set to the value specified above plant canopies. Wind speed inside the canopy at 801
mid-height was set to 35% of the above-canopy wind speed, and the wind speed at ground level 802
inside the plant canopy was set to 4%. These optimum values were determined iteratively via 803
minimization of deviations of the total ET from lysimeter measurements (nonlinear least-squares 804
fitting procedure). 805
While the parameters in the physically-based model by Shuttleworth and Wallace (1985) 806
are given, the search for the best model fit to experimental data obtained from the lysimeters in 807
each Ecotron unit was done with an iterative least squares fitting procedure to find best 808
parameter estimates for which no measurements were available. The free parameters for which 809
(2)
Milcu et al. 2015 – appendices
11/12
best estimates were searched for were: albedo of the plant canopy; albedo of the bare soil 810
fraction; relative proportion of available energy partitioned to ground heat flux under plant 811
canopy and on the bare soil part; fraction of above-canopy wind speed inside the plant canopy 812
and at the ground surface (a) below plants and (b) over the bare soil fraction; soil surface 813
resistance against evaporation. This procedure was however partially restricted to not overfit the 814
model in such a way that one parameter estimate was searched for all twelve Ecotron domes/ 815
experimental units to not introduce artificial differences between the domes for parameters that 816
conceptually should be the same in the Shuttleworth and Wallace (1985) model for all of the 817
Ecotron’s domes. To model E and T separately with the SW model, additional model parameters 818
were introduced to characterize the canopy structure and plant species specific contributions to 819
T, which were quantified using the best available empirical measurements (see below). Where no 820
direct measurements were available (soil resistance to evaporation, wind speed, albedo of plants 821
and of soil), a best estimate for the respective model parameter was made under the premise that 822
the Ecotron provides controlled conditions and hence the same parameter estimate can be used 823
for all replicates. Since the soil monoliths stem from the same soil extracted from the Jena 824
experiment, we argue that this this assumption is justified, but acknowledge that it does not 825
include potential indirect feedbacks to soil density and structure. 826
There are intrinsic limitations in the simulation of the night-time environmental 827
conditions in the Ecotron. The intensive air mixing of the atmosphere in the domes prevents 828
night-time stratification of air, which means that the air, soil surface and the soil will quickly 829
arrive at similar temperatures. In a natural settings however, the soil surface would be 830
considerably cooler than the soil below or the atmosphere above, which would reduce 831
evaporation from the cool nocturnal soil surface. In addition, due to the air-humidifying system, 832
the air relative humidity during night-time was limited at 65% which may increase evaporation 833
in the Ecotron. Presumably, these limitations explain the discrepancy between the modeled and 834
measured night-time ET. However, the day-time evaporative flux is the predominant component 835
of water vapor loss over 24h since the night-time fluxes typically represent only 10-15% (Caird 836
et al. 2007) of day-time fluxes. Therefore, we restricted our analyses of plant diversity effects on 837
the modeled E and T for the day-time period where the environmental conditions simulated in 838
the Ecotron match closely the field conditions at the “Jena Biodiversity Experiment” site. In 839
Milcu et al. 2015 – appendices
12/12
addition, there was good agreement between the day-time measured and modeled ET (see 840
results). 841
842
Supplementary References 843
Caird, M. A, J. H. Richards, and L. A. Donovan. 2007. Nighttime stomatal conductance and 844
transpiration in C3 and C4 plants. Plant physiology 143:4–10. 845
Erisman, J. W., a. Van Pul, and P. Wyers. 1994. Parametrization of surface resistance for the 846
quantification of atmospheric deposition of acidifying pollutants and ozone. Atmospheric 847
Environment 28:2595–2607. 848
Hyer, E. J., and S. J. Goetz. 2004. Comparison and sensitivity analysis of instruments and 849
radiometric methods for LAI estimation: assessments from a boreal forest site. Agricultural 850
and Forest Meteorology 122:157–174. 851
Stull, R. B. 1988. An Introduction to Boundary Layer Meteorology. Kluwer. 852
Vaieretti, M. V., S. Díaz, D. Vile, and E. Garnier. 2007. Two measurement methods of leaf dry 853
matter content produce similar results in a broad range of species. Annals of Botany 854
99:955–958. 855
Vile, D., et al. 2005. Specific leaf area and dry matter content estimate thickness in laminar 856
leaves. Annals of Botany 96:1129–1136. 857
Wesely, M. L. 1989. Parameterization of surface resistances to gaseous dry deposition in 858
regional-scale numerical models. Atmospheric Environment 23:1293–1304. 859
Wilson, P. J., K. Thompson, and J. G. Hodgson. 1999. Specific leaf area and leaf dry matter 860
content as alternative predictors of plant strategies. New Phytologist 143:155–162. 861
862