HAL Id: hal-03327523
https://hal.archives-ouvertes.fr/hal-03327523
Submitted on 27 Aug 2021
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Species richness and food-web structure jointly drive community biomass and its temporal stability in fish
communities
Alain Danet, Maud Mouchet, Willem Bonnaffé, Elisa Thébault, Colin Fontaine
To cite this version:
Alain Danet, Maud Mouchet, Willem Bonnaffé, Elisa Thébault, Colin Fontaine. Species richness and food-web structure jointly drive community biomass and its temporal stability in fish communities.
Ecology Letters, Wiley, In press, �10.1111/ele.13857�. �hal-03327523�
Species richness and food-web structure jointly drive
1
community biomass and its temporal stability in fish
2
communities
3
Alain Danet, Maud Mouchet, Willem Bonnaffé, Elisa Thébault, Colin Fontaine
4
July 20, 2021
5
• Author affiliation:
6
– Alain Danet: Centre d’Ecologie et des Sciences de la Conservation, UMR 7204 MNHN-
7
CNRS-Sorbonne Université, Muséum national d’Histoire naturelle de Paris, 43 rue
8
Buffon, 75005 Paris, France
9
– Maud Mouchet: Centre d’Ecologie et des Sciences de la Conservation, UMR 7204
10
MNHN-CNRS-Sorbonne Université, Muséum national d’Histoire naturelle de Paris, 43
11
rue Buffon, 75005 Paris, France
12
– Willem Bonnaffé: Ecological and Evolutionary Dynamics Lab. Zoology Research and
13
Administration Building, 11a Mansfield Rd, Oxford OX1 3SZ, United Kingdom.
14
– Elisa Thébault: Sorbonne Université, CNRS, IRD, INRAE, Université Paris Est Créteil,
15
Institute of Ecology and Environmental Sciences of Paris (iEES-Paris), Paris, France
16
17
– Colin Fontaine: Centre d’Ecologie et des Sciences de la Conservation, UMR 7204 MNHN-
18
CNRS-Sorbonne Université, Muséum national d’Histoire naturelle de Paris, 43 rue
19
Buffon, 75005 Paris, France
20
• Author contributions: AD, CF, MM and ET designed the study. WB designed the network
21
inference method and collected the food diet data. AD performed the data analysis and wrote
22
the first draft of the manuscript. All authors contributed substantially to revisions.
23
• Data accessibility statement: The raw data used for the analysis are available on zenodo:
24
10.5281/zenodo.5095656. The code used for the analysis and to generate the manuscript is
25
available on github: (alaindanet/fishcom.git). The raw fish database is public and available
26
upon request to [email protected] and [email protected]. The raw
27
data of water temperature, flow and Biological Oxygen Demand are available on http:
28
//www.naiades.eaufrance.fr/.
29
• Article type: Letters
30
• Running title: food-web structure, biomass and stability
31
– Abstract: 151
32
– Main text: 4789
33
• Number of references: 102
34
• Number of figures: 3
35
• Number of tables: 1
36
• Number of text boxes: 0
37
• Coresponding author:
38
– Alain Danet
39
– UMR 7204 CESCO, CP 135; 43, rue Buffon, 75005 Paris, France
40
41
– +33140798134
42
1 Abstract
43
Biodiversity-ecosystem functioning (BEF) and food-web complexity-stability relationships are
44
central to ecology. However, they remain largely untested in natural contexts. Here we estimated
45
the links among environmental conditions, richness, food-web structure, annual biomass and its
46
temporal stability using a standardised monitoring dataset of 99 stream fish communities spanning
47
from 1995 to 2018. We first revealed that both richness and average trophic level are positively
48
related to annual biomass, with effects of similar strength. Second, we found that community
49
stability is fostered by mean trophic level, while contrary to expectation, it is decreased by species
50
richness. Finally, we found that environmental conditions affect both biomass and its stability mainly
51
via effects on richness and network structure. Strikingly, the effect of species richness on community
52
stability was mediated by population stability rather than synchrony, which contrasts with results
53
from single-trophic communities. We discuss the hypothesis that it could be a characteristic of
54
multi-trophic communities.
55
2 Introduction
56
Current biodiversity losses and further global change have the potential to affect ecosystems in
57
complex ways (Loreau et al. 2001; Tilman et al. 2014). Biodiversity has indeed a major effect
58
on ecosystem functions and their stability in response to perturbations (Hector 1999; Loreau et
59
al. 2001; Hooper et al. 2005; Dunne 2006; Duffy 2009). Understanding the relationship between
60
biodiversity and ecosystem properties is thus crucial to better anticipate the impacts of global
61
change.
62
Species richness has been found to increase primary productivity, community biomass and its
63
temporal stability in various ecosystems and taxa both experimentally and in natural settings
64
(Tilman et al. 1996, 2006; Cardinale et al. 2002; Franssen et al. 2011; Grace et al. 2016; Houlahan
65
et al. 2018; Pennekampet al. 2018; Olivier et al. 2020). However, these studies mainly focused
66
on communities composed of a single trophic level, generally primary producers, thereby ignoring
67
potential effects of food-web structure (but see Scherber et al. 2010). Conversely, theoretical
68
works suggest that both species richness and food-web structure affect ecosystem properties such
69
as productivity, total biomass and its temporal stability. For example, the presence of species
70
at higher trophic levels may increase total primary production through top-down effects, which
71
result in the selection of primary producers that are larger and more complementary in their use of
72
resources (Wang & Brose 2018). In communities including producers and consumers, the food-web
73
connectance is predicted to modulate the diversity-productivity relationship (Thébault & Loreau
74
2003; Thébault et al. 2007). Similarly, theoretical studies generally show strong links between
75
food-web structure and stability (e.g. May 1972; Dunne 2006; Neutel et al. 2007; Duffy 2009;
76
Stouffer & Bascompte 2011). Overall, theoretical models suggest that both species richness and
77
food-web structure should affect community biomass and its stability, but empirical evidence of
78
such effects are lacking.
79
While community stability is measured in many ways in theoretical studies, most empirical studies
80
measure community stability as the inverse of the coefficient of variation of community productivity,
81
biomass or abundance, also called temporal stability (Kéfi et al. 2019). This stability metric is also
82
well grounded in theory (e.g. Thébault & Loreau 2005; Loreau & Mazancourt 2008; Thibaut &
83
Connolly 2013) and can be decomposed into two components: population stability and synchrony
84
(Thibaut & Connolly 2013). Studies considering a single trophic level (Yachi & Loreau 1999; Tilman
85
et al. 2006; Loreau & Mazancourt 2008; Olivier et al. 2020) as well as theoretical prediction
86
for multitrophic communities (Thébault & Loreau 2005) tend to highlight that asynchrony in
87
population dynamics is key for the stabilising effects of species richness on community biomass.
88
Theory predicts that higher food-web connectance may increase interaction strength and thereby
89
increase population variability, counterbalancing the positive effect of species richness on the
90
temporal stability of community biomass (Thébault & Loreau 2005). Predators can also have
91
cascading effects on both population variability and synchrony at lower trophic levels (Teng &
92
McCann 2004; Shanafelt & Loreau 2018). Overall, these theoretical studies show that the effects
93
of species richness and food-web structure, such as connectance and average trophic level, are
94
potentially conflicting or synergistic in such a way that they need to be simultaneously assessed.
95
The joint study of the effects of species richness and food-web structure on community biomass and
96
its stability is important in the context of environmental changes. Recent studies comparing natural
97
communities showed that environmental characteristics determine in part community biomass
98
(Graceet al. 2016; Woods et al. 2020) and its temporal stability (Franssen et al. 2011; Hansenet
99
al. 2013; Blüthgen et al. 2016; De Boeck et al. 2018; Fried-Petersen et al. 2020). The environment
100
can affect ecosystem function or stability directly. For instance, larger freshwater ecosystems are
101
expected to have more available resources and thus community biomass (Post et al. 2000; Doi et
102
al. 2009) while agricultural intensification destabilises bird communities by decreasing population
103
stability (Olivier et al. 2020). The effects of environment can also be mediated by changes in
104
species richness and food-web structure. Larger ecosystems host food-webs with higher diversity
105
and trophic levels (e.g., McHugh et al. 2010) and temperature is also known to affect food-web
106
diversity and structure (e.g., Edeline et al. 2013; O’Gorman et al. 2017; Gauzens et al. 2020;
107
Bonnaffé et al. 2021). Environmental characteristics have thus to be accounted for if we are to
108
tease apart the effects of species richness and food-web structure on community biomass and its
109
temporal stability.
110
Stream fish communities constitute a good model to study the effects of network structure on
111
community biomass and its temporal stability because their trophic interactions are well-known. A
112
substantial body of literature supports that trophic interactions can be inferred through body size
113
and diet information (Jennings et al. 2001; Beckerman et al. 2006; Gravel et al. 2013; Brose et al.
114
2019; Portalier et al. 2019). Using long-term monitoring of stream fish communities and inference
115
of trophic interactions (Bonnaffé et al. 2021), we provide one of the first empirical tests of the
116
expected relationship between BEF and food-web complexity-stability accounting for environmental
117
characteristics.
118
3 Methods
119
3.1 Data preparation and filtering
120
Fish communities were monitored across stream sections in metropolitan France over the period
121
1995-2018 by the French Office of Water and Aquatic Ecosystems (ONEMA) using electrofishing.
122
Two different standardised protocols were used for small and large streams (detailed in appendix
123
B, part 1.1). In the case where a stream was sampled alternatively with one of the two sampling
124
protocols, we only included data derived from the most frequently used sampling method. The
125
sampling season spanned from late spring to autumn. There was low heterogeneity in the sampling
126
month within stations, with the median sampling month being mid-August and the median standard
127
deviation of the sampling month was 0.7 month, i.e. three weeks.
128
We selected the stations that had been sampled ten years or more. As the concept of temporal
129
stability often referred to the variability around an equilibrium point (Donohue et al. 2016), we
130
selected the stations that did not show temporal trends in community biomass (detailed in appendix
131
B, part 2). However, the results were robust to the relaxation of this hypothesis (Fig. S7). These
132
selection steps resulted in 99 stations distributed over seven hydrographic basins (Fig. 1A).
133
3.2 Community biomass and its temporal stability
134
3.2.1 Biomass estimation
135
The fishes collected by electrofishing were grouped in batches according to their body size class
136
and species identity (Fig. S1, appendix B, part 1.1). When individuals were too numerous in a
137
given batch, the body size of a subsample of the individuals was measured. The body size of the
138
unmeasured individuals was inferred under the assumption that the body size of the individuals
139
in a batch follows a normal distribution (Fig. S1, appendix B, part 1.2). The body mass of each
140
individual was then estimated following the results of two meta-analyses on fishes (Lleonart et al.
141
2000; Froese 2006), using an allometric lawBi = 0.01×l3.03i , with B the body mass in grams, l
142
the body size in millimetres andi the individual. Population biomass and community biomass at
143
each sampling event were then calculated by summing individual body masses. To account for
144
the variation in sampling effort among sites, population biomass and community biomass were
145
corrected by the surface sampled (range 140 - 2700 square metres), and expressed in grams per
146
square metre (gm−2). In the analysis, the community biomass of a site was defined as the median
147
of the community biomass across sampled years.
148
3.2.2 Stability measures
149
Temporal biomass stability was defined as the inverse of biomass variability. Variability was
150
computed as Coefficient of Variation (CV) of the community biomass over time (CVi = σiµ−1i ,
151
with σi andµi being respectively the standard deviation and the temporal average of the annual
152
community biomass of the station i). Then, temporal stability (Si) was equal to CVi−1 =µiσ−1i .
153
The variability of community biomass (CVcom,i) can be decomposed into a synchrony (φ) component
154
and a weighted average population variability (CVsp,i) component according to Thibaut & Connolly
155
(2013):
156
CVcom,i=CVsp,i×qφi (1)
To assess the mechanisms driving stability of community biomass, we thus in addition measured
157
the synchrony and the weighted average population variability CVsp,i, the average CV at the
158
population-level in the station i, weighted by the relative biomass of the species populations, Bk,i.
159
Synchrony of the station i, φi, is the ratio between the variance of community biomass over time
160
(σx2
T ,i) and the squared sum of the standard deviation of population biomass over time (σxk,i), i
161
being the station iand k the species k (Loreau & Mazancourt 2008). Synchrony varies from 0 to 1;
162
0 indicating complete asynchrony and 1 perfect synchrony. Consequently, the lower is synchrony,
163
the higher are the compensatory dynamics.
164
CVsp,i=X
k
Bk,iCVk,i
P
kBk,i (2)
φi = σ2xT ,i (Pkqσx2
k,i)2 = σ2xT ,i
(Pkσxk,i)2 (3)
3.3 Species richness and food-web structure
165
3.3.1 Species richness
166
Species richness in a given site was defined as the total number of species present across all the
167
sampling events. We controlled species richness for sampling effort by dividing the total number of
168
species by the total surface sampled in the given station. Species richness was then expressed in
169
number of species per square metre sampled.
170
3.3.2 Food-web inference
171
We determined the food-web structure in two steps (Bonnaffé et al. 2021): (1) we built a metaweb
172
describing the trophic interactions among all possible fish size classes and species as well as with
173
non-fish resource classes (Fig. S2), then (2) we determined the food-web corresponding to each
174
sampling event by extracting from the metaweb the subset of fish species and size classes present
175
at a given site and sampling event (Fig. 1C and D).
176
The food-web was inferred from body size and ontogenic diet shift following previous studies (Gravel
177
et al. 2013; Poisotet al. 2016; Brose et al. 2019; Bonnafféet al. 2021). The diet of stream fishes,
178
as for many organisms, shifts throughout an individual’s life, and thereby with body size. For
179
example, larvae and small juveniles feed on plankton while adults may feed on algae and fishes, with
180
bigger individuals eating bigger fishes. Each fish species was divided into nine body size classes,
181
each corresponding to a trophic species (appendix B, Fig. S2). A trophic species constitutes a
182
group of individuals from the same species and size class, which are expected to share the same
183
sets of prey and predators. Additionally, we added seven resource nodes present in the ontogenic
184
food diet database: detritus, biofilm, phytoplankton, zooplankton, macrophages, phytobenthos and
185
zoobenthos (grey nodes in Fig. 1C & D). The trophic species and the resources constituted the 412
186
nodes of the metaweb, i.e. 45 fish species divided into nine body size classes plus seven resource
187
nodes. The inference of trophic interactions was shown to be more accurate when considering body
188
size classes, i.e. trophic species, than species (Jennings et al. 2001).
189
The feeding interactions of a trophic species were determined by both its species identity and its
190
body size. The ontogenic fish diet database was filled according to fishbase (Froese et al. 2021)
191
and literature (appendix A, Table S2). Each fish species had two or three life stages which are
192
delimited by its body size range (appendix B, Fig. S2). The life stage of a fish trophic species was
193
determined by the centre of its body size range, hereafter midpoint (appendix B, Fig. S2).
194
For a piscivorous trophic species, the fish-fish trophic links were defined in two steps and based on
195
the body size ratio between predators and prey. First of all, the predation window of a piscivorous
196
trophic species was defined as 3% to 45% of its midpoint (Mittelbach & Persson 1998; Claessen et
197
al. 2002; Hart & Reynolds 2008). Then, a trophic link was set between the piscivorous trophic
198
species and every trophic species whose midpoint in body size range was included in the predation
199
window (appendix B, Fig. S2). The fish-resource trophic links were set between the trophic species
200
and the resources nodes according to the food items of their size class. Lastly, the trophic links
201
among resource nodes were set according to the literature (Allan & Castillo 2007; Hart & Reynolds
202
2008).
203
3.3.3 Food-web structure
204
For each food-web, corresponding to a particular site and year, we computed the connectance
205
and mean trophic level. Food-web connectance describes the level of generalism in the network
206
and is calculated as the actual number of interactions divided by the number of interactions if
207
all species interact (C = L/N2, L and N being resp. the number of links and of nodes). The
208
trophic level of the nodes was computed as one added to the average trophic level of its food
209
items. The average trophic level T was computed as the average trophic level of the trophic species
210
weighted by their biomass (excluding resource nodes for which we had no biomass information), so
211
T =PNi BiTi×(PNi Bi)−1, withTi and Bi being respectively the trophic level and the biomass of
212
the node i. For each site, the median of connectance and average trophic level over time was used
213
in subsequent analysis (Fig. 1C & D).
214
3.4 Environmental variables
215
We characterised the sites by their altitude, slope, distance from source, stream width, depth
216
and strahler order, i.e. its position in the stream network hierarchy. We obtained data on Bi-
217
ological Oxygen Demand (BOD), which is often related to enrichment and eutrophication in
218
aquatic systems (Mallin et al. 2006), water temperature, flow and from the naiades database
219
(www.naiades.eaufrance.fr). As data from naiades are not collected on the same site as our fish
220
sampling sites, we performed an interpolation of the water temperature and flow based on the
221
spatial stream network models (Hoef et al. 2014; Isaak et al. 2014). The interpolation procedure is
222
detailed in appendix B (part 1.4).
223
We had one environmental variable value for each sampling event, except for water temperature for
224
which records began in 2006. Then, we computed the average and temporal CV values of water
225
temperature, flow, BOD, river width and depth. As habitat and environmental variables are often
226
collinear, a principal component analysis (PCA) was performed on the 14 scaled variables. We kept
227
in the analysis the two first principal components based on the elbow method on the eigenvalues
228
distribution (appendix B, Fig. S5). The two first principal components explained 46% of the
229
variance (Fig. 1B). A varimax axis rotation was performed to maximise the representation of the
230
environmental variables by the principal components (Kaiser 1958; Revelle 2019). The first axis
231
represented a gradient from small to big streams, characterised by their width and depth, but also
232
a gradient of distance from source. We reversed the values of the second axis, so that is represented
233
a gradient of increasing average water temperature, average BOD and decreasing altitude and slope.
234
Further description of the variables are provided in appendix B (part 1.4).
235
3.5 Statistical analysis
236
We analysed the relationships among environment, species richness, food-web structure, community
237
biomass and its temporal stability with structural equation models as they enable the assessment
238
of potential direct and indirect relationships among multiple variables (Grace 2008; Grace et al.
239
2014, 2016; Lefcheck 2016). The structural equation model explains the median annual community
240
biomass as a function of the environment, species richness and food-web structure (Maureaud et
241
al. 2019; Woods et al. 2020). We assumed that the environment, species richness and food-web
242
structure directly affect community biomass (Fig. 2A), that species richness affects food-web
243
structure (Winemiller 1989; Dunne 2006; Thébault & Fontaine 2010) and that environment affects
244
species richness and food-web structure.
245
We used the same model structure to assess the drivers of temporal stability of community biomass
246
(Fig. 3A). Instead of the annual community biomass, we modelled the synchrony and population
247
variability which in turn fully determined the temporal stability of community biomass. We
248
linearized the mathematical relationship linking temporal stability of community biomass Si to
249
CVsp,i and φ by taking the log of temporal stability such as
250
logSi =−logCVsp,i−1
2logφi (4)
The structural equation models were based on Gaussian linear mixed-effects models. All variables
251
except the connectance and average trophic level were log-transformed to fulfil model requirements
252
(see residual plots and data transformation in Fig. S5 and S6, Table S5, appendix B). We set the
253
seven hydrographic basins as a random effect on the intercept to account for their heterogeneity.
254
We checked for the absence of multicollinearity with Variance Inflation Factor (VIF, Table S6,
255
appendix B). The slopes estimated by the structural equation models and reported in the main
256
text were standardised in order to compare their magnitude (rδ =βσσx
y,β being the unstandardised
257
slope, σx andσy being respectively the standard deviation of the predictor variable xand of the
258
response variable y). Therefore, the standardised coefficients expressed the variation of x and
259
y in standard deviation units. According to the equation (4), the unstandardised coefficients
260
of synchrony and population variability determining temporal stability of community biomass
261
are respectively -0.5 and -1.0. When standardised, the relative comparison of those coefficients
262
comes down to a comparison of the standard deviation of these two variables. The deviation from
263
unstandardised effect values (i.e. -0.5 and -1.0) indicates that one component of community stability
264
displays more variations than the other, and that the determinants of the more variable component
265
will be those affecting community stability the most. The paths whose slope had a P value inferior
266
or equal to 0.05 were reported in the main text, all the path values are presented in Table S7 and
267
S8. Indirect effects on community biomass and its temporal stability were computed by multiplying
268
slopes along the path. Only significant direct slopes were included in the calculation of indirect
269
effects. Total effects were computed as the sum of direct and indirect effects.
270
Environmental variables were interpolated with the openStars and SSNR packages (Hoef et al.
271
2014; Kattwinkel & Szöcs 2018). Linear models and SEMs were computed with respectively the
272
nlme and piecewieseSEM R packages (Lefcheck 2016; Pinheiro et al. 2020). The code used for
273
the analysis and to generate the manuscript are accessible on github (alaindanet/fishcom). The
274
analyses were performed with R version 3.6.3 (R Core Team 2020).
275
4 Results
276
The structural equation models explained a large part of the variation of species richness, con-
277
nectance, synchrony, CVsp and community biomass (resp. R2 = .43, .38, .46, .47 and .32) but
278
less so for mean trophic level (R2 = .11, Fig. 2A and 3A). We found a negative relationship
279
between species richness and mean trophic level (rδ =−0.26, rδ meaning standardised slope) and
280
connectance (rδ =−0.7, Fig. 2A and 3A), meaning that a part of the effects of species richness on
281
community biomass and its temporal stability went through food-web structure. The relationships
282
resulting from the SEM are robust to different station selection criteria in the analysis, with different
283
selections ranging in 65 to 403 stations (see sensitivity analysis, Fig. S8).
284
4.1 Species richness and average trophic level are positively correlated
285
with community biomass
286
Species richness and mean trophic level had a significant positive direct effect on community biomass
287
(rδ= 0.32 and rδ = 0.3 resp., Fig. 2A, B, C, Table 1), while connectance had no direct effect. The
288
two PCA axes synthesising environmental variables had no significant direct effects on community
289
biomass, but indirect ones (Table 1). The first axis, related to the stream size, was positively linked
290
to community biomass (rδ = 0.18, Table 1), through both species richness and mean trophic level
291
(rδ = 0.08 and 0.1 resp., Fig. 2F, G, Table 1). The second axis, related to the average annual
292
temperature and BOD was positively related to community biomass (rδ = 0.21, Table 1) mainly
293
through species richness (rδ = 0.13, Fig. 2 G, Table 1). Warmer and enriched, i.e. with higher BOD,
294
streams contained more species, and thus indirectly increased the biomass of the communities.
295
The total effect of PCA1, PCA2 and species richness on community biomass were of the same
296
magnitude.
297
4.2 Temporal stability of annual community biomass
298
By its effects on synchrony and population variability (CVsp), species richness had a negative total
299
effect on biomass stability (rδ= −0.31, Table 1). This effect resulted from two opposite effects. On
300
the one hand, species richness had a direct negative effect on synchrony (rδ =−0.51, Fig. 3A, B),
301
thereby increasing stability (rδ = 0.4). On the other hand, species richness had a positive direct
302
effect on CVsp (rδ = 0.47, Fig. 3A, C), thereby decreasing biomass stability (rδ =−0.62, Table 1),
303
which outweighed the stabilising effect of species richness via decreased synchrony. The negative
304
effect of CVsp on temporal stability was twice higher than the negative effect of synchrony (resp.
305
rδ =−1.33 andrδ =−0.77, Fig. 3A), which is as expected from the mathematical decomposition
306
of the stability of community biomass. Population variability and synchrony indeed contribute
307
respectively to -1 and -0.5 when the coefficients are not standardised (Eq. 4). Our result thus
308
also means that the standard deviation of CVsp is of roughly the same magnitude as the one of
309
synchrony in this case.
310
The total effect of mean trophic level and species richness on temporal stability of community
311
biomass were of the same magnitude (resp. 0.35 and −0.31, Table 1). The mean trophic level had
312
only a direct negative effect on CVsp (rδ =−0.26, Fig. 3A, F) and thereby had a total positive
313
effect on temporal stability (rδ = 0.35, Table 1). Food-webs in which community biomass is located
314
on average at higher trophic levels thus tend to have higher stability of population and community
315
biomass. Connectance had no significant direct effect on eitherCVsp nor synchrony and thus no
316
effect on temporal stability of community biomass (Fig. 3A, Table 1).
317
The total effect of PCA2, related to average annual temperature and BOD, on temporal stability
318
was of higher magnitude than the effects of species richness and mean trophic level (Table 1). The
319
PCA2 had a total negative effect on the temporal stability of communities, meaning that higher
320
temperature and BOD results in lower stability (rδ = −0.41, Table 1). PCA2 had an indirect
321
positive effect on temporal stability through a direct positive effect on mean trophic level (rδ = 0.09,
322
Fig. 3 I, Table 1). However, PCA2 had a negative effect on stability of community biomass through
323
a direct positive effect on CVsp and species richness (resp. rδ =−0.33 andrδ = −0.17, Fig. 3G,
324
Fig. 2G, Table 1). PCA1, related to stream size, had in contrast a very slight positive effect on
325
stability of community biomass (rδ = 0.02, Table 1). Warmer and larger streams tend to have more
326
species, but the variability of their population biomass is respectively higher and lower, resulting in
327
respectively lower and higher temporal stability of community biomass.
328
5 Discussion
329
5.1 Species richness increased community biomass but decreased its
330
temporal stability
331
We found that species richness increased community biomass in line with the results in the plants
332
and tree communities (Hector 1999; Tilman et al. 2014; Wu et al. 2015; Li et al. 2018) as well as
333
those found for fish communities (Maureaud et al. 2019; Woods et al. 2020) and for theoretical
334
food-web models (Schneider et al. 2016). Theory suggests that the positive effect of species richness
335
on community biomass might be induced by resource use complementarity (Carey & Wahl 2011),
336
which is driven by niche differentiation (Loreau 1998; Loreau & Hector 2001) or driven by the
337
dominance of a highly productive species (i.e. positive selection effect, Loreau et al. 2001). Recent
338
findings suggest that such positive selection effect (Loreau & Hector 2001) is the main driver of the
339
observed positive relationship between biomass and species richness in fish communities (Maureaud
340
et al. 2019; Woods et al. 2020).
341
Contrary to most experimental and empirical results (Tilman et al. 2006; Franssen et al. 2011;
342
Pennekamp et al. 2018; Olivier et al. 2020; Valencia et al. 2020, but see, Declercket al. 2006),
343
we found a negative relationship between species richness and stability of community biomass.
344
This effect resulted from a combination of two effects, one on population synchrony and one
345
on population variability, both of which contribute to the stability of community biomass. The
346
negative effect of species richness on population synchrony we found, thereby increasing stability of
347
community biomass, is in accordance with most studies on experimental grasslands (Isbell et al.
348
2009; Roscher et al. 2011) as well as natural communities such as birds, grasses, bats or butterflies
349
(Blüthgen et al. 2016; Olivier et al. 2020, but see, Declercket al. 2006; Valencia et al. 2020). Our
350
result thus extend this relationship to multi-trophic communities.
351
The positive effect of species richness on population variability that we found, thereby destabilising
352
communities, is more intriguing and does not fit the findings of a meta-analysis reporting more
353
frequent negative effects of richness on population variability in multi-trophic communities (Jiang
354
& Pu 2009). Food-web theory predicts that the relationship between the temporal stability
355
of populations and species richness depends on the strength of interactions and its distribution
356
(McCann & Hastings 1997; Thébault & Loreau 2005; Downinget al. 2020), with stronger interactions
357
promoting a negative relationship between species richness and stability. This might be the case
358
in our study system but this hypothesis remains to be directly tested. Alternatively, a positive
359
richness-population variability relationship can come from an increased demographic stochasticity
360
with increasing species richness (Loreau & Mazancourt 2008).
361
In our study system, the destabilising effect of species richness via increased population variability
362
was stronger than the stabilising effect via decreased synchrony. Further, as the effects of species
363
richness on population variability and synchrony we found here were of similar amplitude, the
364
balance between both pathways is governed by the relative standard deviation of population
365
variability and synchrony, this because we used standardised coefficients (see methods). The fish
366
communities studied here therefore present similar level of variation in population variability and
367
synchrony, again contrasting with results on terrestrial animal communities that exhibited twice
368
as much variation in population synchrony than in population variability (Olivier et al. 2020).
369
This opens intriguing questions as to potential mechanisms constraining the variation range of
370
population variability and synchrony among communities. Lower variation in population synchrony
371
among the fish communities studied here might relates to their trophic structure as theory predicts
372
potential synchronising effect of generalist predators, such as piscivorous fishes, on prey dynamics
373
(Raimondo et al. 2004; Huber & Gaedke 2006). Trophic interactions might also raise synchrony
374
between predators and preys, (Fig. S9, Bulmer 1975), although predictions can vary depending
375
on the distribution of interaction strengths (Huber & Gaedke 2006; McCann & Rooney 2009).
376
Restricted variations in population synchrony among communities can also be fostered by low
377
biomass evenness and thereby low portfolio effect (Doak et al. 1998; Thibaut & Connolly 2013),
378
but whether biomass evenness of stream fish communities is particularly low remains to be tested.
379
Alternatively, aquatic ecosystems have been suggested to show greater population variability than
380
terrestrial ones (Rip & McCann 2011), which could translate in greater variation range of population
381
variability among aquatic communities than among terrestrial ones.
382
5.2 Average trophic level increased community biomass and its tem-
383
poral stability while connectance had no effect
384
Our analysis further indicates that network structure, and in particular average trophic level was
385
strongly related to community biomass, as higher community biomass was found in communities
386
with higher average trophic level. These results are in accordance with the theoretical predictions
387
(Schneideret al. 2016; Wang & Brose 2018), proposing that predation selects for primary producers
388
with higher resource use efficiency and enhance complementarity between primary producers (Wang
389
& Brose 2018).
390
Further, our results show that communities with higher average trophic level had a lower population
391
variability, and were thereby more stable. This result is also in accordance with theoretical models
392
predicting a higher stability of populations of higher trophic levels (Shanafelt & Loreau 2018;
393
Barbier & Loreau 2019) or a damping effect of higher trophic levels on the variability of lower
394
trophic level populations (McCann 2000).
395
In contrast, we found that food-web connectance was not related to community biomass, which
396
contradicts theoretical models suggesting that higher connectance is expected to lower community
397
biomass by decreasing complementarity in resource use among species (Thébault & Loreau 2003;
398
Poisotet al. 2013). However, the effect of connectance on productivity remains poorly investigated.
399
Unexpectedly, we found no relationship between food-web connectance and stability of community
400
biomass which contrasts with many theoretical results highlighting a food-web complexity-stability
401
relationships (e.g., Angelis 1975; May 1972; Thébault & Fontaine 2010; but see Neutel et al. 2007),
402
although they generally consider different stability concepts and metrics (Arnoldi et al. 2016). The
403
lack of connectance effect on the temporal stability of community biomass might be related to the
404
connectance metric we used, in particular the fact that it did not account for interaction strengths.
405
Indeed, connectance stability relationship is expected to vary depending on the strength of trophic
406
interactions (Thébault & Loreau 2005). Further, in our dataset, resource nodes, i.e. primary
407
producers, decomposers and the benthic species, were highly aggregated which might blur potential
408
connectance differences among communities.
409
5.3 Environment had a strong effect on community biomass and its
410
temporal stability
411
We found that the larger as well as warmer and more enriched streams had higher community
412
biomass, and that this was mediated by higher species richness and average trophic level. Regarding
413
the effect of ecosystem size, our results thus support a classical theory of food-web ecology (Post
414
et al. 2000; Doi et al. 2009). Larger freshwater ecosystems provide higher diversity of resources
415
through habitat heterogeneity and then result in higher species richness and mean trophic level per
416
surface unit, through complementarity in resource use (Postet al. 2000; Doi et al. 2009; McHugh
417
et al. 2010). Regarding temperature and enrichment effect, our study joins the mounting body
418
of works reporting effect on the structure and biomass of aquatic systems (Allan & Castillo 2007;
419
Edeline et al. 2013; Emmrich et al. 2014; Hattab et al. 2016; Woods et al. 2020; Bonnaffé et al.
420
2021). However, current results do not show consensus (Yvon-Durocher et al. 2011; Dossena et al.
421
2012; Maureaudet al. 2019; Tabi et al. 2019), with potential antagonistic effects of temperature
422
and enrichment as well as non-monotonic effects (O’Connor 2009; Uszko et al. 2017; Tabi et al.
423
2019). Although we could not tease apart the respective effects of temperature and enrichment,
424
our results indicates that in the range of observed covariation in temperature and enrichment, their
425
joint effects result in a positive relationship with community biomass.
426
Our analysis further revealed that stream size had only a weak effect on the temporal stability of
427
community biomass, while warmer and enriched streams exhibit less stable community biomass
428
mainly via an increase in population variability. These results are in line with previous experiments
429
showing that warming or enrichment decrease the temporal stability of communities (Kratina et al.
430
2012; Tabiet al. 2019). However, several studies have highlighted that the effects of temperature
431
on stability can be complex, and interact with those of enrichment (Binzer et al. 2012; Uszkoet al.
432
2017).
433
5.4 Network reconstruction: limits and perspectives
434
The lack of experimental or empirical studies considering the effects of both species richness and
435
food-web structure on ecosystem functions can be explained by the challenges associated with
436
manipulating both species richness and food-web structure experimentally (e.g. Sander et al. 2017)
437
and by the fact that describing trophic interactions between species is highly labour-intensive.
438
These challenges are even more acute when studying the temporal stability of ecosystem properties,
439
which requires time series. In this context, recent developments to predict trophic interactions
440
among species (Beckerman et al. 2006; Gravelet al. 2013; Portalieret al. 2019), combined with
441
observational long-term monitoring of ecological communities, offer a promising opportunity. They
442
offer a way to maximise the amount of information that can be extracted from already collected
443
empirical data, but at the cost of their precision. Here, we built binary interaction networks
444
based on the ratio of body-size between predator and prey ignoring modulation of interactions
445
strength depending on preferences or environmental factors. Refining the construction of interaction
446
networks by, for example, integrating predator behaviour (Portalier et al. 2019) or developing
447
more statistically flexible framework such as machine learning (Pichleret al. 2020) will certainly
448
increase the accuracy of species interaction predictions. Although challenging, methods of food-web
449
reconstruction have a great potential to increase our understanding of network structure effects on
450
ecosystem function and dynamic in natural settings using available long-term empirical data.
451
6 Conclusion
452
Our work contributes to bridging the gap between theoretical, experimental and empirical findings
453
on the functioning and stability of communities. Our study extends the assessment of the effect
454
of species richness on community biomass and its temporal stability to the case of freshwater
455
food-webs in natural settings. While our results are in accordance with numbers of studies showing
456
a positive effect of richness on community biomass, they provide a contrasting example of negative
457
relationship between richness and stability of community biomass. Our results highlight population
458
variability as key in transmitting effects of species richness, network structure and environmental
459
gradients on community stability, opening stimulating questions about how the food-web structure
460
and species richness drive population synchrony and variability.
461
References
462
Allan, J.D. & Castillo, M.M. (2007). Stream ecology: Structure and function of running waters.
463
Springer Science & Business Media.
464
Angelis, D.L.D. (1975). Stability and Connectance in Food Web Models. Ecology, 56, 238–243.
465
Arnoldi, J.-F., Loreau, M. & Haegeman, B. (2016). Resilience, reactivity and variability: A
466
mathematical comparison of ecological stability measures. Journal of Theoretical Biology, 389,
467
47–59.
468
Barbier, M. & Loreau, M. (2019). Pyramids and cascades: A synthesis of food chain functioning
469
and stability. Ecology Letters, 22, 405–419.
470
Beckerman, A.P., Petchey, O.L. & Warren, P.H. (2006). Foraging biology predicts food web
471
complexity. Proceedings of the National Academy of Sciences, 103, 13745–13749.
472
Binzer, A., Guill, C., Brose, U. & Rall, B.C. (2012). The dynamics of food chains under climate
473
change and nutrient enrichment. Philosophical Transactions of the Royal Society B: Biological
474
Sciences, 367, 2935–2944.
475
Blüthgen, N., Simons, N.K., Jung, K., Prati, D., Renner, S.C. & Boch, S.et al. (2016). Land use
476
imperils plant and animal community stability through changes in asynchrony rather than diversity.
477
Nature Communications, 7, 10697.
478
Bonnaffé, W., Danet, A., Legendre, S. & Edeline, E. (2021). Comparison of size-structured and
479
species-level trophic networks reveals antagonistic effects of temperature on vertical trophic diversity
480
at the population and species level. Oikos, n/a.
481
Brose, U., Archambault, P., Barnes, A.D., Bersier, L.-F., Boy, T. & Canning-Clode, J.et al. (2019).
482
Predator traits determine food-web architecture across ecosystems. Nature Ecology & Evolution, 3,
483
919–927.
484
Bulmer, M.G. (1975). Phase Relations in the Ten-Year Cycle. The Journal of Animal Ecology, 44,
485
609.
486
Cardinale, B.J., Palmer, M.A. & Collins, S.L. (2002). Species diversity enhances ecosystem
487
functioning through interspecific facilitation. Nature, 415, 426–429.
488
Carey, M.P. & Wahl, D.H. (2011). Determining the mechanism by which fish diversity influences
489
production. Oecologia, 167, 189–198.
490
Claessen, D., Oss, C.V., Roos, A.M. de & Persson, L. (2002). The Impact of Size-Dependent
491
Predation on Population Dynamics and Individual Life History. Ecology, 83, 1660–1675.
492
De Boeck, H.J., Bloor, J.M.G., Kreyling, J., Ransijn, J.C.G., Nijs, I. & Jentsch, A. et al. (2018).
493
Patterns and drivers of biodiversity–stability relationships under climate extremes. Journal of
494
Ecology, 106, 890–902.
495
DeClerck, F.A.J., Barbour, M.G. & Sawyer, J.O. (2006). Species Richness and Stand Stability in
496
Conifer Forests of the Sierra Nevada. Ecology, 87, 2787–2799.
497
Doak, D.F., Bigger, D., Harding, E.K., Marvier, M.A., O’Malley, R.E. & Thomson, D. (1998). The
498
Statistical Inevitability of Stability-Diversity Relationships in Community Ecology. The American
499
Naturalist, 151, 264–276.
500
Doi, H., Chang, K.-H., Ando, T., Ninomiya, I., Imai, H. & Nakano, S.-i. (2009). Resource
501
availability and ecosystem size predict food-chain length in pond ecosystems. Oikos, 118, 138–144.
502
Donohue, I., Hillebrand, H., Montoya, J.M., Petchey, O.L., Pimm, S.L. & Fowler, M.S. et al. (2016).
503
Navigating the complexity of ecological stability. Ecology Letters, 19, 1172–1185.
504
Dossena, M., Yvon-Durocher, G., Grey, J., Montoya, J.M., Perkins, D.M. & Trimmer, M. et al.
505
(2012). Warming alters community size structure and ecosystem functioning. Proceedings of the
506
Royal Society B: Biological Sciences, 279, 3011–3019.
507
Downing, A.L., Jackson, C., Plunkett, C., Lockhart, J.A., Schlater, S.M. & Leibold, M.A. (2020).
508
Temporal stability vs. Community matrix measures of stability and the role of weak interactions.
509
Ecology Letters, 23, 1468–1478.
510
Duffy, J.E. (2009). Why biodiversity is important to the functioning of real-world ecosystems.
511
Frontiers in Ecology and the Environment, 7, 437–444.
512
Dunne, J.A. (2006). The network structure of food webs. Ecological networks: linking structure to
513
dynamics in food webs, 27–86.
514
Edeline, E., Lacroix, G., Delire, C., Poulet, N. & Legendre, S. (2013). Ecological emergence of
515
thermal clines in body size. Global Change Biology, 19, 3062–3068.
516
Emmrich, M., Pédron, S., Brucet, S., Winfield, I.J., Jeppesen, E. & Volta, P. et al. (2014).
517
Geographical patterns in the body-size structure of European lake fish assemblages along abiotic
518
and biotic gradients. Journal of Biogeography, 41, 2221–2233.
519
Franssen, N.R., Tobler, M. & Gido, K.B. (2011). Annual variation of community biomass is lower
520
in more diverse stream fish communities. Oikos, 120, 582–590.
521
Fried-Petersen, H.B., Araya-Ajoy, Y.G., Futter, M.N. & Angeler, D.G. (2020). Drivers of long-term
522
invertebrate community stability in changing Swedish lakes. Global Change Biology, gcb.14952.
523
Froese, R. (2006). Cube law, condition factor and weight–length relationships: History, meta-
524
analysis and recommendations. Journal of Applied Ichthyology, 22, 241–253.
525
Froese, R., Pauly, D. & others. (2021). FishBase. Fisheries Centre, University of British Columbia.
526
Gauzens, B., Rall, B.C., Mendonça, V., Vinagre, C. & Brose, U. (2020). Biodiversity of intertidal
527
food webs in response to warming across latitudes. Nature Climate Change, 10, 264–269.
528
Grace, J.B. (2008). Structural Equation Modeling for Observational Studies. The Journal of
529
Wildlife Management, 72, 14–22.
530
Grace, J.B., Adler, P.B., Stanley Harpole, W., Borer, E.T. & Seabloom, E.W. (2014). Causal
531
networks clarify productivity–richness interrelations, bivariate plots do not. Functional Ecology,
532
787–798.
533
Grace, J.B., Anderson, T.M., Seabloom, E.W., Borer, E.T., Adler, P.B. & Harpole, W.S. et al.
534
(2016). Integrative modelling reveals mechanisms linking productivity and plant species richness.
535
Nature, 529, 390–393.
536
Gravel, D., Poisot, T., Albouy, C., Velez, L. & Mouillot, D. (2013). Inferring food web structure
537
from predator–prey body size relationships. Methods in Ecology and Evolution, 4, 1083–1090.
538
Hansen, B.B., Grøtan, V., Aanes, R., Sæther, B.-E., Stien, A. & Fuglei, E. et al. (2013). Climate
539
Events Synchronize the Dynamics of a Resident Vertebrate Community in the High Arctic. Science,
540
339, 313–315.
541
Hart, P.J. & Reynolds, J.D. (2008). Handbook of Fish Biology and Fisheries: Fisheries. John Wiley
542
& Sons.
543
Hattab, T., Leprieur, F., Lasram, F.B.R., Gravel, D., Loc’h, F.L. & Albouy, C. (2016). Forecasting
544
fine-scale changes in the food-web structure of coastal marine communities under climate change.
545
Ecography, 39, 1227–1237.
546
Hector, A. (1999). Plant Diversity and Productivity Experiments in European Grasslands. Science,
547
286, 1123–1127.
548
Hoef, J.V., Peterson, E., Clifford, D. & Shah, R. (2014). SSN: An R Package for Spatial Statistical
549
Modeling on Stream Networks. Journal of Statistical Software, 56, 1–45.
550
Hooper, D.U., Chapin, F.S., Ewel, J.J., Hector, A., Inchausti, P. & Lavorel, S. et al. (2005). Effects
551
of biodiversity on ecosystem functioning: A consensus of current knowledge. Ecological monographs,
552
75, 3–35.
553
Houlahan, J.E., Currie, D.J., Cottenie, K., Cumming, G.S., Findlay, C.S. & Fuhlendorf, S.D. et al.
554
(2018). Negative relationships between species richness and temporal variability are common but
555
weak in natural systems. Ecology, 99, 2592–2604.
556
Huber, V. & Gaedke, U. (2006). The role of predation for seasonal variability patterns among
557
phytoplankton and ciliates. Oikos, 114, 265–276.
558
Isaak, D.J., Peterson, E.E., Ver Hoef, J.M., Wenger, S.J., Falke, J.A. & Torgersen, C.E. et al.
559
(2014). Applications of spatial statistical network models to stream data: Spatial statistical network
560
models for stream data. Wiley Interdisciplinary Reviews: Water, 1, 277–294.
561
Isbell, F.I., Polley, H.W. & Wilsey, B.J. (2009). Biodiversity, productivity and the temporal stability
562
of productivity: Patterns and processes. Ecology Letters, 12, 443–451.
563
Jennings, S., Pinnegar, J.K., Polunin, N.V.C. & Boon, T.W. (2001). Weak cross-species relationships
564
between body size and trophic level belie powerful size-based trophic structuring in fish communities.
565
Journal of Animal Ecology, 70, 934–944.
566
Jiang, L. & Pu, Z. (2009). Different Effects of Species Diversity on Temporal Stability in Single-
567
Trophic and Multitrophic Communities. The American Naturalist, 174, 651–659.
568
Kaiser, H.F. (1958). The varimax criterion for analytic rotation in factor analysis. Psychometrika,
569
23, 187–200.
570
Kattwinkel, M. & Szöcs, E. (2018). openSTARS: An Open Source Implementation of the ’ArcGIS’
571
Toolbox ’STARS’.
572
Kéfi, S., Domínguez-García, V., Donohue, I., Fontaine, C., Thébault, E. & Dakos, V. (2019).
573
Advancing our understanding of ecological stability. Ecology Letters, 22, 1349–1356.
574
Kratina, P., Greig, H.S., Thompson, P.L., Carvalho-Pereira, T.S.A. & Shurin, J.B. (2012). Warming
575
modifies trophic cascades and eutrophication in experimental freshwater communities. Ecology, 93,
576
1421–1430.
577
Lefcheck, J.S. (2016). piecewiseSEM: Piecewise structural equation modelling in r for ecology,
578
evolution, and systematics. Methods in Ecology and Evolution, 7, 573–579.
579
Li, S., Lang, X., Liu, W., Ou, G., Xu, H. & Su, J. (2018). The relationship between species richness
580
and aboveground biomass in a primary Pinus kesiya forest of Yunnan, southwestern China. PLOS
581
