1
Joining the dots: an automated method for constructing food webs
2
from compendia of published interactions
3
4
Clare Graya, b, David H Figueroac, Lawrence N. Hudsond, Athen Mae, Dan Perkinsb, Guy Woodwardb*
5
6
a
Queen Mary University of London, School of Biological and Chemical Sciences, London, E1 4NS,
7
UK
b
8
Department of Life Sciences, Imperial College London, Silwood Park, Buckhurst Road, Ascot,
9
Berkshire,SL5 7PY,UK
c
10
11
12
d
Universidad Católica de Temuco, Escuela de Ciencias Ambientales, Montt 56 Temuco Chile
Department of Life Sciences, Natural History Museum, Cromwell Road, London, SW7 5BD, UK
e
Queen Mary University of London, School of Electronic Engineering and Computer Science,
13
London, E1 4NS, UK
14
*Corresponding author
15
Email: guy.woodward@imperial.ac.uk (GW)
16
17
1
18
19
20
Abstract
Food webs are important tools for understanding how complex natural communities are
21
structured and how they respond to environmental change. However their full potential has yet to be
22
realised because of the huge amount of resources required to construct them de novo. Consequently,
23
the current catalogue of networks that are suitable for rigorous and comparative analyses and
24
theoretical development still suffers from a lack of standardisation and replication.
25
Here, we present a novel R function, WebBuilder, which automates the construction of
26
food webs from taxonomic lists, and a dataset of trophic interactions. This function works by
27
matching species against those within a dataset of trophic interactions, and ‘filling in’ missing trophic
28
interactions based on these matches. We also present a dataset of over 20,000 freshwater trophic
29
interactions, and use this and four well-characterised freshwater food webs to test the method.
30
The WebBuilder function facilitates the generation of food webs of comparable quality to
31
the most detailed published food webs, but at a fraction of the research effort or cost. Furthermore, it
32
matched and often outperformed a selection of predictive models, which are currently among the best,
33
in terms of capturing key properties of empirical food webs. The method is simple to use, systematic
34
and, perhaps most importantly, reproducible, which will facilitate (re-) analysis and data sharing.
35
Although developed and tested on a sample of freshwater food webs, this method could easily be
36
extended to cover other types of ecological interactions (such as mutualistic interactions).
37
38
Keywords – food web construction, gut contents analysis, food web modelling
39
1 - Introduction
40
Characterising food webs (networks representing trophic interactions between species) and
41
other ecological networks (networks which represent any type of ecological interaction, such as
42
pollination) can help us understand and, ultimately predict multispecies systems’ responses to changes
43
in environmental conditions (Thompson et al., 2012; Tylianakis et al., 2010). Food webs can reveal
44
subtle but important changes in the biotic interactions that underpin ecosystem functioning, stability,
45
and resilience to perturbations - higher-level phenomena that cannot be inferred from studying the
46
nodes (i.e., species or populations) alone (Gray et al., 2014; Thompson et al., 2012).
47
Despite the many advantages of a network-based approach to ecology, significant challenges
48
need to be overcome, particularly in terms of gathering interaction data. Interactions occur between
49
individuals and data are often collected at this level: for example, via collection, rearing and
2
50
identification of every leaf miner, and subsequent leaf miner parasitoid along a transect to build
51
herbivore-parasitoid networks (Macfadyen et al., 2011; as in Memmott et al., 2000), or through
52
dissecting and identifying consumer gut-contents via microscopy (as in Layer et al., 2013). Such
53
laborious methods require substantial investment of time and resources, and it can take many
54
thousands of lab hours to characterise just one food web, which even then may still be undersampled
55
for links between its rarer members (see Table 1; e.g. Woodward et al. 2005; Olito & Fox 2014).
56
Many hundreds or thousands of individuals of each species are often needed to fully characterise the
57
full set of feeding links within a food web (e.g. Ings et al., 2009), which is rarely practical given the
58
financial and time restraints of research funding. In addition, such comprehensive sampling is often
59
destructive and can impose undesirable disturbance on study systems. Consequently, empirical food
60
webs are often incompletely described and constructed from relatively small sample sizes (e.g.
61
Kaiser-Bunbury et al., 2010; Layer et al., 2013). This limits the conclusions that can be drawn and the
62
number of comparable food webs that are available both across and within studies (Bascompte et al.,
63
2003; e.g. Briand, 1983; Olesen et al., 2007) although exceptions to this exist (Bascompte et al., 2003;
64
Cohen and Mulder, 2014). Most studies still have patchy and differing levels of sampling effort and
65
taxonomic resolution, making meta-analyses difficult or even inappropriate: the ability to construct
66
large numbers of realistic, comparable food webs across multiple systems would, therefore, help
67
realise the true potential of network approaches (Gray et al., 2014).
68
69
Table 1. Methods for constructing food webs, with their advantages and disadvantages.
Method
Observation of
evidence of
interaction (e.g.
feeding trials or
gut contents
analysis)
Extrapolating
from previously
published
interactions
(e.g.
WebBuilder
function)
Advantages
High confidence in
links produced.
Disadvantages
Very slow and labour-intensive.
Rare interactions are often missed.
Interaction type is biased by the
method employed, e.g. the prey of
suctorial predators cannot be
determined through gut contents
analysis.
Fair confidence in links
produced.
Rare interactions can be
included.
Interactions from
multiple studies
determined through
different methods can be
easily incorporated.
Reliant on the quality of the data
contained within the reference
dataset.
Can only be used to construct
‘cumulative’ or ‘summary’ food
webs, i.e. temporal or spatial
changes in feeding behaviour
cannot be incorporated.
Low effort and quick.
3
Examples
Woodward et al. (2005)
Macfadyen et al. (2011)
Henson et al. (2009)
Ledger et al. (2012)
Hall & Raffaelli (1991)
Goldwasser et al. (1993)
Havens (1993)
Piechnik et al. (2008)
Pocock et al. (2012)
Layer et al.(2013)
Cohen & Mulder (2014)
Strong & Leroux (2014)
Predictive
models
Ecological rules and
theory can be
incorporated.
Require prior knowledge of the
structure of the food web in order
to optimize parameter values.
Low effort and quick.
Many perform poorly at predicting
individual interactions, even when
food web structure is predicted
well.
Cohen et al. (1985)
Williams & Martinez
(2000)
Petchey et al. (2008)
Allesina & Pascual
(2009)
Allesina (2011)
Olito & Fox (2014)
70
71
72
Ecological networks are often constructed by incorporating species interactions from the
73
published literature (Table 1) and many food webs are constructed entirely in this manner (Cohen and
74
Mulder, 2014; e.g. Goldwasser et al., 1993; Havens, 1993; Strong and Leroux, 2014), while other
75
food webs contain a blend of observational and extrapolated data (Layer et al., 2013; M. J. O. Pocock
76
et al., 2012). By filling in ‘missing’ trophic interactions to a given species list, the implicit assumption
77
is made that, if a given pair of species have been observed to interact at one site, they will interact in
78
the same way at other sites where they co-occur (at least in terms of a feeding link between the
79
species being realised, or not). Food webs built through this method are often referred to as
80
‘summary’ or ‘cumulative’ food webs as they represent all potential interactions (of a particular type,
81
for instance trophic interactions within a food web) between species of a particular community, rather
82
than a snapshot in time. As such, food webs build through this method are unsuitable for detecting
83
changes in species feeding behaviour across sites or over time, but are highly effective for detecting
84
broad macro-ecological trends such as changes in food web structure across environmental gradients
85
(K Layer et al., 2010; Mulder and Elser, 2009; e.g. Piechnik et al., 2008).
86
This approach can be taken further, by assigning interactions of species on the basis of
87
taxonomic similarity: i.e., species within the same genus are assumed to have identical links if a link
88
has been established through direct observation for at least one congener (e.g. Goldwasser et al., 1993;
89
K Layer et al., 2010). This process is often used when constructing summary food webs for species
90
the interactions of which have not been fully characterised (e.g., as revealed from yield-effort curves)
91
to minimise potential biases arising from under-sampling, i.e. including only observed links would
92
otherwise significantly underestimate food web complexity, especially among the rarer and/or more
93
obscure taxa (Woodward et al 2010). Recent work (Eklof et al., 2012) has provided justification for
94
this approach, by highlighting the strong influence that taxonomy has in determining the structure of
95
food webs. Thus, given the prevalence of undersampling in even relatively well-described food webs,
96
dietary data extrapolated from the literature and generalised taxonomically can potentially produce far
97
more complete and realistic summary food webs than those that rely solely on observations made in a
98
particular locale.
99
Despite the prevalence of these methods for constructing summary food webs in the literature
100
(Cohen and Mulder, 2014; e.g. Goldwasser et al., 1993; Havens, 1993; Katrin Layer et al., 2010; M. J.
4
101
O. O. Pocock et al., 2012; Strong and Leroux, 2014), there is still no standard method for inferring
102
feeding interactions, resulting in inconsistencies among studies, even within the same ecosystem type.
103
This is especially problematic because authors rarely state explicitly which links have been observed
104
or extrapolated, or the source from which they have been drawn, or how closely the previously
105
published interactions match those reported in their particular study, making replication impossible
106
and preventing other researchers from scrutinising published interactions fully (but see Strong and
107
Leroux, 2014).
108
Recent research has sought to develop predictive models of the structure of ecological
109
networks (Eklof et al., 2012; Gravel et al., 2013; Olito and Fox, 2014; e.g. Rohr et al., 2010). Simple
110
rules based on ecological theory have been used to model and predict the structure and topology of
111
food webs, the most successful of which include deterministic models based on information on
112
species’ body sizes, for example the ‘Difference’, ‘Ratio’, and ‘Difference/Ratio’ models (Allesina,
113
2011) and the Allometric Diet Breadth Model (ADBM; Petchey et al. 2008) which incorporates
114
allometric scaling and optimal foraging parameters. Whilst these models have been developed
115
primarily to advance ecological theory, they provide a possible tool through which food webs could
116
be built de novo in order to address questions about network structure across environmental gradients
117
or scales. However, to achieve their best performance (proportion of correctly predicted links) these
118
models require some prior knowledge about the number of links in the network. For instance, for the
119
models mentioned above a researcher is required to go through a parameter optimisation procedure,
120
by fixing the number of links, values of constants and exponents can be derived, by maximizing the
121
number of links correctly predicted. When constructing a network for the first time for a particular
122
system, a research would be required to fix the number of links to an expected value which would bias
123
the network structure towards that which the researcher expected to find.
124
Additionally, recent work (Olito and Fox, 2014) has highlighted that while predictive models
125
might perform well at predicting metrics of network structure, they tend to perform poorly at
126
predicting pairwise interactions (Sáyago et al., 2013; e.g. Vázquez et al., 2009; Verdú and Valiente-
127
Banuet, 2011; Vizentin-Bugoni et al., 2014), so whilst they may predict network structural metrics
128
well, they are doing so for the wrong reason as the underlying biological mechanisms have not been
129
fully incorporated into the predictive models (Petchey et al., 2011). To the best of our knowledge, the
130
models used here have not, up until now, been used to predict network structure de novo, as this is not
131
the scenario for which they were developed.
132
Given the limitations of constructing food webs from observation of interactions or predictive
133
models, we need an automated, repeatable and reliable method of building local food webs that can be
134
applied across studies and, ultimately, different ecosystem and network types. Here, we introduce a
135
method, the WebBuilder function that assembles food webs by systematically assigning links for
136
taxa based upon a given set of user-defined rules applied to a dataset of known trophic interactions.
137
We provide an implementation of our method for the R statistical modelling language (R Core Team,
5
138
2013), building upon the methods and data structures provided by the Cheddar R package (Hudson et
139
al., 2013). We tested the method on four highly resolved freshwater food webs which have had their
140
interactions characterised through gut contents analysis, as these represent some of the most complete
141
food webs described to date, as a test case for our proof-of-concept. Specifically, our key aims were
142
to:
143
144
145
146
1. Collate a dataset of trophic interactions in a standard format to act as an example system in
which to test this method.
2. Automate the process of constructing food webs from this reference dataset in a repeatable
and reliable manner.
147
3. Compare the performance of this method with the structure of food webs with ‘known’
148
interactions, i.e. those which have been built through observation of the interactions.
149
150
4. Compare the performance of this method with another way of predicting food web structure;
the ADBM, Difference, Ratio and Difference/Ratio models.
151
2 - Methods
152
2.1 - Dataset of trophic interactions
153
We collated a dataset of 20,823 pairwise trophic interactions among species (or the next highest level
154
of resolution available, usually genus), from 51 different data sources, most of which were primary
155
literature (Table A.1, Appendix A). It contains trophic interactions between primarily UK freshwater
156
species, including 203 producer taxa, 593 invertebrate taxa, 24 fish taxa, 10,348 producer-animal
157
links, 9,531 animal-animal links and 944 detritus-animal links. When the necessary data were not
158
available in the original publications, we contacted the authors directly, where possible, to obtain the
159
raw data. The taxonomy of every resource and consumer has been standardised through the Global
160
Names Resolver (http://resolver.globalnames.biodinfo.org/) using the Global Biodiversity Information
161
Facility dataset. For every resource-consumer link the taxonomy (species, genus, subfamily, family,
162
order, class) of both is given, along with life-stage information, if relevant, and a literature reference
163
for the source of the link. This dataset builds upon the collection assembled by Brose et al (2005), and
164
to the best of our knowledge, represents the largest standardised collection of trophic links for
165
freshwater organisms. This dataset is available to download at
166
https://sites.google.com/site/foodwebsdataset/ (doi: 10.5281/zenodo.13751) and is designed to be
167
easily updated by the iterative addition of new data (details of how to submit new data to the dataset
168
are given on the website), allowing its content to improve over time, in an analogous manner to
169
molecular-based bioinformatics datasets. New data will be subjected to a quality assurance procedure
170
prior to inclusion in the dataset. Specifically, all taxa will be parsed through Global Names Resolver
171
(http://resolver.globalnames.biodinfo.org/) using the Global Biodiversity Information Facility dataset.
172
Additionally the data will be eyeballed for irregularities. It is anticipated that this data will exists as an
6
173
open access resource, and as such the community of researchers who access it will report any errors
174
they find so they can be double checked and removed. New iterations of the dataset can be produced,
175
hosted on the webpage alongside the original, and assigned a new DOI, allowing researchers to cite
176
exactly which version of the dataset they have used for their research, allowing analyses to be
177
repeated using identical versions to those cited in a given study, if required in the future.
178
179
2.2 - The WebBuilder function
180
The method of constructing ecological networks by extrapolating from previously published
181
interactions is implemented in a new R function - WebBuilder (see Appendix B for code). The user
182
is required to provide the following; firstly a list of taxa (i.e. nodes) in the community of interest (step
183
I, Figure 1), this data can be gathered from multiple sources and could be in the form of survey or
184
biomonitoring data. Secondly, for each node, the minimum level of taxonomic generalisation
185
(explained below; step II), and the taxonomic classification of each node (step III). Lastly a registry –
186
a dataset of known trophic interactions, including taxonomic classification (step IV), as example of
187
which is published here, but which can also be created by the user or obtained elsewhere. It is
188
recommended that the user resolve the taxonomy of their taxa list and registry using the same
189
procedure so as to ensure that taxa are matched correctly, if the user were using the registry provided
190
here they would need to parse their taxa list through Global Names Resolver
191
(http://resolver.globalnames.biodinfo.org/) using the Global Biodiversity Information Facility dataset.
192
The function searches the registry for every possible combination of resource-consumer interactions
193
(for N taxa there are N2 possible trophic interactions) which match the provided taxa list given the
194
specified level of taxonomic generalisation.
195
The minimum level of taxonomic generalisation determines the taxonomic rank at which
196
matches are made, thus generalising the resources or consumers of the candidate node to the species,
197
genus, subfamily, family, order etc level, as specified in the input (step II, Figure 1). For instance, a
198
researcher might decide to ascribe the level of taxonomic generalisation of ‘genus’ to the mayfly
199
Baetis fuscatus, allowing it to be matched with the more commonly studied species Baetis rhodani in
200
the dataset, and take on the appropriate feeding interactions of that species, i.e. those which include
201
taxa also present on the provided taxa list (see the first Scenario in Figure 2 ). This level of taxonomic
202
generalisation is selected based on knowledge of a candidate node’s trophic interactions in relation to
203
its sister taxa (i.e. if all members of the same taxonomic unit can be assumed to have the same trophic
204
interactions or not), and this can be tailored depending on the resource/consumer status of the node.
205
For example, consumers of the larvae of the non-biting midge subfamily Tanypodinae tend to be
206
trophic generalists and would likely consume other larvae of the family Chironomidae, while it is not
207
likely that the resources of Tanypodinae larvae (which are predominantly predatory) would be shared
208
by all Chironomidae larvae (many of which are grazers or filter feeders). Hence it would not be
7
209
appropriate to assign the trophic generalisation level ‘family’ to both resource and consumer
210
interactions of Chironomidae. Instead a researcher might ascribe the ‘resource method’ for
211
Tanypodinae as ‘family’, but the ‘consumer method’ as ‘subfamily’ (see the second Scenario in
212
Figure 2). The function output contains references to the original empirical links, the number of
213
matches that were found and the taxonomic level at which those matches were found, so links can be
214
additionally screened and scrutinised post hoc, and analysis can be repeated easily because the
215
function output contains the necessary information. Example R code is supplied (Appendix C).
216
2.3 - Comparing the WebBuilder function with empirical food webs
217
The WebBuilder function was validated on a collection of highly-resolved stream food webs which
218
have had their trophic interactions characterised through direct observation; Broadstone Stream
219
(Woodward et al., 2010), Afon Hirnant (Gilljam et al., 2011; Woodward et al., 2010), Tadnoll Brook
220
(Edwards et al., 2009) and the summary food web for the replicated four reference Mill Stream side-
221
channels (Ledger et al., 2012; Woodward et al., 2012). The replicates for the Mill Stream data were
222
pooled to aid comparison with the other food webs, which were all constructed as a single aggregate
223
food web. The Broadstone and Afon Hirnant food webs contained only trophic interactions between
224
macro-invertebrates, the Tadnoll food web contained interactions between macro-invertebrates and
225
fishes and the Mill Stream data contain interactions between macroinvertebrates, algae and detritus.
226
When the WebBuilder function was used to generate the empirical food webs, in turn each
227
respective local dataset was first removed from the global dataset, so each food web was generated in
228
the absence of its own link information (to remove circularities).
229
The performance of the WebBuilder function was evaluated by calculating the True Skill
230
Statistic (TSS; Allouche et al., 2006). This statistic was used as it can be digested into its component
231
parts to provide information on the types of differences between the empirical and generated food
232
webs, and builds upon the most commonly used metric which is simply the proportion of links
233
correctly generated (e.g. Allesina, 2011; Petchey et al., 2008; Woodward et al., 2010). This statistic
234
was chosen over likelihood based approaches because we were not interested so much in the
235
efficiency of these predictive models, more the biological realism of the generated food webs
236
(Petchey et al., 2011). The TSS is calculated from the following formula:
237
𝑇𝑆𝑆 = (𝑎𝑑 − 𝑏𝑐)/[(𝑎 + 𝑐)(𝑏 + 𝑑)]
238
where a is the number of links which were correctly generated by the function (the True Positives
239
Rate; TPR), b the number of links generated by the function but not observed empirically, c the
240
number of links not generated by the function but were observed empirically and d the number of
241
links neither generated by the function nor observed empirically. TSS score values range from -1 to 1,
242
where a score of -1 represents a generated food web that is the inverse of the empirically observed one
243
(no observed empirical links are seen in the generated food web, and every non-link in the empirical
8
244
food web is present in the generated food web), and 1 representing a generated food web having the
245
exact same links as the empirically observed one.
246
Each empirical food web was generated using the level of taxonomic generalisation
247
considered most appropriate (see Appendix D), this mostly consisted of exact and genus level matches
248
although some family and order matches were used. To test how the generated food webs compared to
249
their empirical counterparts, a series of network metrics were calculated; number of links (L), linkage
250
density (L/S; where S is the number of nodes), connectance (C, where C=L/S2), generality (the
251
average number of resources per consumer), vulnerability (the average number of consumers per
252
resource), and proportion of top, intermediate and basal nodes (with cannibalistic links removed). The
253
difference between the generated and empirical network metric was tested with paired Wilcoxon
254
signed rank tests.
255
To test how the quality of the generated food webs varied with dataset size, the dataset was
256
randomly subsampled, in sequential steps of 5% from 5-100%, of the original dataset size, and then
257
used to generate each food web. Each subsample size was repeated five times and each empirical food
258
web was generated in the absence of its own food web data as above, to remove circularities. For each
259
node within each network, the same level of taxonomic generalisation was used as above.
260
To test how the quality of the pairwise interactions generated by the WebBuilder function
261
varied with the level of taxonomic generalisation, each food web was built using exact, genus, family
262
or order taxonomic generalisation for all nodes. The degree (the number of links into or out of a
263
particular node), generality, and vulnerability for every node in the generated food web was compared
264
with that in the empirical network. The difference between the two for every node was recorded so
265
that a positive score represented interactions ‘missed’ by the WebBuilder function, and a negative
266
score represented ‘extra’ interactions not found in the empirical food web. The distribution of these
267
scores gives an indication of how well the WebBuilder function predicted pairwise interactions
268
across the whole network: i.e., if, on average, it tended to ‘miss’ more interactions, or tended to pick
269
up ‘extra’ interactions. To test if the mean was different from zero (indicating no difference in the
270
quality of pairwise interactions between the generated and empirical food webs) a one sampled t-test
271
was used.
272
273
2.4 - Comparing the WebBuilder function to theoretical food web models
274
The performance of the WebBuilder function was compared with examples of some of the best-
275
performing predictive models currently available: the ‘Difference’, ‘Ratio’ and ‘Difference/Ratio’
276
models (Allesina, 2011) as well as the Allometric Diet Breadth Model (ADBM; with “ratio” handling
277
time, Petchey et al., 2008). The ‘Difference’, ‘Ratio’ and ‘Difference/Ratio’ models all generate food
278
web links on the basis of body size, (either the difference between consumer body size and resource
279
body size, the ratio between the two, or the difference multiplied by the ratio). The ADBM builds on
9
280
this and incorporates allometries of body size and foraging behaviour of individual consumers to
281
model food web structure (Allesina, 2011 for more detailed explanations; see Petchey et al., 2008).
282
Detritus nodes were first excluded from the Mill Stream food web because these nodes had no body
283
size or abundance data. For the ‘Difference’, ‘Ratio’ and ‘Difference/Ratio’ models two parameters
284
required optimisation, a and b. For the ADBM we used parameter values for the mass to attack rate
285
constant (a), resource mass to attack rate exponent (ai) and consumer mass to attack rate exponent (aj)
286
from the literature (Rall et al., 2012) rather than through parameter optimisation as in Petchey et al.
287
(2008), so as to simulate a situation for which the WebBuilder function was designed, where food
288
webs are being generated for the first time with no prior knowledge of the system other than the
289
species richness. For two parameters (mass to handling time constant, h.a; mass to handling time
290
critical ratio, b) we were unable to find information in the literature with which to value these
291
parameters, so went through the process of optimisation. For all models this was achieved by
292
constructing food webs with a range of values for each parameter, and selecting those food webs
293
which had a number of links that was within the range set by the WebBuilder function, i.e. if the
294
WebBuilder function generated K links, and there were L empirical links and K-L=t we selected all
295
possible solutions within the range L-t:L+t, to make the comparison with the WebBuilder function
296
fair. Note that for some food webs the difference between L and K was large, leading to large
297
variation in the food web sizes generated by these models. Indeed for the Afon Hirnanlt food web this
298
range fell below zero, and so the range was arbitrarily set to be the same proportional size as that of
299
the Tadnoll food web, which had the next highest range. Parameter optimisation was conducted
300
without using the connectance of the empirical food webs, hence although the same data have been
301
used, results will vary from previous publications. Prior knowledge of the connectance of food webs
302
would not be possible if a food web were being built de novo, so here we are using these models in a
303
different way from their original application.
304
3 - Results
305
3.1 - Comparing the WebBuilder function with empirical food webs
306
When we constructed food webs from random subsets of the dataset, the quality (as measured by TSS
307
scores) of the generated food webs improved as the number of records in the dataset increased,
308
allowing more complete resource and consumer interactions to be ascribed to each taxa (Figure 3).
309
The strength of this relationship was food web specific, for instance Broadstone and Afon Hirnant did
310
not continue to improve beyond a dataset size of about 25%. These food webs are relatively simplistic
311
compared to Tadnoll and Millstream, and so the WebBuilder function reached its optimum
312
performance when generating these food webs with a fraction of the total dataset. Tadnoll and
10
313
Millstream did not reach their asymptotes suggesting that more data is needed to improve upon the
314
quality of their generated food webs.
315
The level of taxonomic generalisation for each node was important for the quality of the
316
generated food web; if the taxonomy of a given node list was generalised too far (typically beyond the
317
family level) then the ascribed links became unrepresentative and the food web become over-
318
connected resulting in an increased FPR and lower TSS score (Figure A.1, Appendix A). At the scale
319
of individual trophic interactions, the difference in degree, generality and vulnerability was generally
320
positive when matching taxa exactly or at the genus level, and becomes progressively more negative
321
as the taxonomic generalisation increased, indicating that the WebBuilder function was ‘missing’
322
links when matching nodes exactly or at the genus level, and including progressively more links the
323
further the taxonomy was generalised (Figure 4). For Afon Hirnant and Tadnoll there was no
324
significant difference in the generality of consumers between the generated and empirical food webs
325
when taxa were matched at the genus level, and there was no significant difference in vulnerability of
326
resources for the Tadnoll food web when match at the genus level. This suggests that matching taxa at
327
the genus level for these food webs produces the most ‘accurate’ pairwise interactions. For all other
328
food webs and levels of taxonomic generalisations the generated links were different from that of the
329
empirical food web (Figure 4).
330
The occurrence of each food web’s nodes in the dataset is given in Table 2. The coverage of
331
these within the reduced dataset (food web data were removed from the dataset when used to generate
332
the food web for that site) varied between 1,497 (Broadstone) and 6,704 occurrences (Mill Stream)
333
(Table 2). Even at the family level some nodes from Broadstone, Afon Hirnant and Mill Stream were
334
not represented in the dataset, meaning that those nodes needed to be generalised further still for the
335
WebBuilder function to generate their links. These nodes tended to be rare taxa which were poorly
336
represented in the dataset. The generated food webs (see Appendix S4 for the generated trophic links)
337
had similar network metrics to the empirical food webs (Table 3), although the proportion of top
338
nodes was consistently lower in the generated food webs, and the proportion of intermediate and basal
339
nodes was consistently higher. All generated food web metrics were found to be similar to that of
340
their empirical counterparts (paired Wilcoxon signed rank test, p=>0.05). The TSS ranged from 0.407
341
(Broadstone) to 0.557 (Tadnoll Brook). Two food webs (Broadstone and Afon Hirnant) contained
342
nodes that were found to have predatory links in the empirical food web but were not predicted to
343
have any by the WebBuilder function, and vice versa many nodes distributed across all the food
344
webs were predicted to have consumer links but were not found to have any empirically (Figure 5).
345
346
Table 2. The representation of the food web taxa within the full dataset and partial dataset (i.e., diet
347
data gathered from a food web were excluded from the generation of its own inferred food web).
11
Number of appearances in
Percentage of nodes appearing in partial dataset at
dataset
each taxonomic level
Food web
Full data
Partial data
set
set
exact
genus
family
Broadstone
2,196
1,497
81%
84%
94%
Afon Hirnant
2,945
2,266
72%
79%
92%
Tadnoll
12,405
4,314
84%
91%
100%
Mill Stream
11,545
6,704
87%
96%
97%
348
349
351
(C, where C=L/S2), Generality, Vulnerability, proportion of top, intermediate and basal species of the
352
empirical and generated food webs. The performance of the WebBuilder function (relative to the
353
original empirical food web) is summarised by the TSS statistic (which gives an overall measure of
354
performance), and TPR (the proportion of links correctly generated). All food web metrics for the
355
generated food webs were found to be similar to that of their empirical counterparts (paired Wilcoxon
356
signed rank test, p = >0.05)
4.43
6.93
2.82
7.58
2.91
4.91
9.19
8.64
0.158
0.247
0.085
0.23
0.05
0.085
0.124
0.117
13
9.84
7.67
15.87
9.33
12.45
16.98
16.63
iat e
Bas
al
124
194
93
250
169
285
680
639
Inte
rme
d
C
Top
357
358
L/S
Vul
nera
empirical
generated
empirical
Afon Hirnant
generated
empirical
Tadnoll
generated
empirical
Mill Stream
generated
Broadstone
L
Gen
Network
bili t
y
Table 3. The number of links (L), linkage density (L/S, where S=number of nodes), the connectance
eral
it y
350
4.33
7.48
4
8.5
3.23
5.59
14.15
11.7
0.68
0.04
0.58
0
0.69
0.07
0.46
0.19
0.29
0.64
0.12
0.45
0.21
0.31
0.19
0.32
0.04
0.25
0.24
0.39
0.1
0.53
0.35
0.41
TSS
0.407 0.376
0.418 0.24
0.557 0.372
0.547 0.531
359
3.2 - Comparing the WebBuilder function method to theoretical food web models
360
The percentage of links correctly predicted (TPR) by the ADBM ranged from 12-43%, for the
361
Difference model it was 0-46%, Ratio model it was 0-51% and the Difference/Ratio it was 0-54%
362
(Figure 6). All webs generated by the WebBuilder function had higher TPR and TSS scores than
363
the median values for the Difference, Ratio and Difference /Ratio models (Figure 6). In general the
364
WebBuilder function had higher TPR and TSS scores than the ADBM, however the TPR score for
365
Tadnoll and Afon Hirnant were similar to the median ADBM TPR score for that particular food web
12
TPR
366
(as opposed to the overall median). Additionally the TSS score for Tadnoll generated by the
367
WebBuilder function was similar to the median ADBM TSS score (Figure A.2, Appendix A).
368
369
4 - Discussion
370
4.1 - Strengths and weaknesses of the WebBuilder function
371
Here we have demonstrated a systematic and reproducible method for building ecological networks
372
from compilations of previously observed interactions. The WebBuilder function facilitates
373
comparability across studies, re-analysis and data sharing. Although developed in the context of
374
freshwater food webs, given its simplicity and generality the WebBuilder function could be easily
375
applied to other systems, such as terrestrial food webs or even mutualistic networks. Plenty of other
376
datasets already exist which could be exploited similarly to produce comparable, reproducible
377
networks from marine and terrestrial systems (e.g., Barnes, 2008; Database of Insects and their Food
378
Plants; http://www.brc.ac.uk/dbif/).
379
The WebBuilder function is an effective tool for constructing summary ecological networks for
380
the first time. The overall performance (TSS) of the WebBuilder function exceeded that of the
381
ADBM, Difference, Ratio and Difference /Ratio models. The proportion of correctly predicted links
382
(TPR) was similar to or exceeded that of the ADBM. The ADBM cannot predict links for nodes that
383
have no body-size information – either because it is not known or because the concept is meaningless
384
for the node, such as detrital resources. This problem does not apply to the WebBuilder function.
385
The ADBM, Difference, Ratio and Difference/Ratio models have been used to generate the food webs
386
presented here before, and have performed better than we have achieved here (Allesina, 2011;
387
Petchey et al., 2008; Woodward et al., 2010), however to achieve this accuracy the generated food
388
webs were constrained to have the same connectance as the empirical food webs, an approach not
389
available when building a food web for the first time. Indeed there were instances here that the TPR
390
and TSS of modelled food webs exceeded that of the WebBuilder function, but from the range of
391
possible food webs generated by these models, it is impossible to select the most ‘accurate’ one
392
without knowledge of the expected number of links. The WebBuilder function does not rely on
393
prior knowledge of the food web, only on the correct identification of the nodes, thus reducing biases
394
and restrictions.
395
It is perhaps unfair to compare the performance of the WebBuilder function to that of the
396
ADBM, Difference, Ratio and Difference/Ratio models due to the inherent differences in the
397
mechanisms through which they operate, and indeed it is not our intention for this exercise to be taken
398
as a criticism of these alternative approaches. Rather, we have compared them here in order to place
399
the WebBuilder method in the broader context of some of other more widely-used predictive methods
13
400
currently available. Comparing our approach with the performance of the ADBM essentially
401
represents a test of, and a means of improving, our understanding of the mechanistic theory behind
402
these trophic interactions. Comparing the food webs produced by our approach with empirical food
403
webs represents a test of the quality of the underlying data held within the dataset of trophic
404
interactions. The WebBuilder function should be used as tool with which to construct large
405
collections of food webs with which to test our understanding of food web structure across
406
environmental gradients. The WebBuilder function is particularly suited to constructing food webs
407
for data-poor systems, e.g., where there is no information available about the abundance or body size
408
of nodes, with the only information being a list of species present. Clearly the ADBM or other
409
predictive models would not be suited to these conditions, as they were never designed to work in this
410
way. The WebBuilder function, however, would be able to generate reasonably realistic food webs if
411
given a reference dataset of relevant trophic interactions. The WebBuilder function is adaptive, and
412
can be improved upon over time; for instance, by increasing the size and coverage of the dataset of
413
interactions. Hence it requires a substantial amount of data to perform well, unlike the predictive
414
models analysed here. These types of methods can be viewed as complementary: a researcher might
415
use both in conjunction in order to harness the advantages of both to better predict food web structure:
416
indeed we envisage combining the WebBuilder function and other predictive approaches in parallel
417
to build and understand food webs.
418
The four food webs presented here are among the most highly resolved and complete freshwater
419
food webs published to date, yet the links are still under-sampled for many nodes (Woodward et al.,
420
2010), due to methodological issues and logistical constraints on sampling effort. The WebBuilder
421
function can help to overcome these issues. Firstly, it can take many hundreds of individuals to
422
characterise a species diet (Ings et al., 2009) and thus the interactions between rare consumers and
423
rare resources are often under-sampled. The WebBuilder function helps to overcome this as rare
424
interactions need only be observed once in the dataset of previously published interactions in order to
425
be incorporated into applicable food webs as they are constructed: i.e., potentially the “global diet” of
426
a species is held within the dataset, and can be expanded in future data collections. Secondly, the
427
method of observing interactions often limits the types of interactions which can be characterised; for
428
instance, the prey of suctorial predators (which are especially common in terrestrial ecosystems)
429
cannot be identified through traditional gut contents analysis, but if characterised through other means
430
(e.g. laboratory trials or molecular sequencing) they can be included in the dataset and incorporated
431
into generated food webs. For instance, two suctorial predators in the Broadstone food web
432
(Platambus maculatus and Bezzia sp.) did not have their guts analysed for predatory links in the
433
original study (Woodward et al., 2005) and so had been previously excluded from the food web (e.g.
434
Petchey et al., 2008; Woodward et al., 2010), these nodes would have been predicted to prey upon
435
other species by the WebBuilder function. This is due to the WebBuilder function generalising
14
436
the taxonomy of these nodes, and their subsequent appearances in the dataset, as other studies have
437
characterised the diets of these taxa. Some links in the dataset of trophic interactions were known
438
from just a single data source, e.g. Cordulegaster boltonii as a consumer of Nemurella pictetii is
439
known only from the Broadstone food web. Therefore, when we excluded self-referrential diet data,
440
the WebBuilder function reconstruction of Broadstone did not predict a trophic link between C.
441
boltonii and N. pictetii. We have not quantified how often this effect occurred. As with other open-
442
source datasets, anomalies will be ironed out as the dataset is enriched with more observations as it
443
grows, and its coverage will improve over time.
444
Besides constructing food webs de novo, the WebBuilder function could be used to standardise
445
a collection of networks gathered from different sources prior to analysis. This would effectively
446
standardise the sampling effort for included interactions (although not for species richness or
447
taxonomic resolution) and would remove spatially or temporally explicit interactions (or lack thereof).
448
If the analysis was concerned with the structure of summary food webs from different locations and
449
habitat types then this might be an appropriate first step.
450
451
452
4.2 - Future Directions
The realism of links generated by the WebBuilder function could be addressed by assessing the
453
number of times a particular interaction appears in the dataset, as well as the number of times an
454
interaction could have occurred but did not (i.e. species found at the same site but not found to
455
interact). If a particular interaction has been observed many times across many systems, it is probably
456
reasonable to assume it also occurs at other sites where those species co-exist. However, if it has only
457
been observed rarely, or at a site with very different characteristics than the one in question (for
458
instance contrasting environmental conditions, or significantly different community assemblages) this
459
assumption might not be so reasonable. As the size of the dataset continues to grow, evaluation of
460
whether links are realised or not will improve over time.
461
The WebBuilder function is designed to construct summary food webs, and ignores potential
462
behavioural shifts of species, hence it is unsuitable for constructing temporally or spatially explicit
463
food webs. Additional data such as abundance information could be used to weight interactions, this
464
would, for instance, reduce the weight of interactions between rare species reducing their influence on
465
food web structure and increasing the realism of the resulting food web. There is an increasing body
466
of literature detailing the importance of weak and strong interactions within networks (Berlow et al.,
467
2004; e.g. de Ruiter et al., 1995; Vazquez et al., 2007) and a multitude of methods already exist for
468
determining interaction strengths in food webs (see Berlow et al., 2004) some of which can be
469
employed alongside the WebBuilder function. Thus, despite the ‘coarse’ nature of food webs built
470
in this way there is much potential for their use in ecological research, and by combining them with
15
471
models such as those presented here potential mismatches arising from behavioural shifts could be
472
highlighted.
473
It would be straightforward for the underlying code of the WebBuilder function to be extended
474
to incorporate a range of traits that could influence the realisation of potential trophic interactions,
475
other than phylogeny, such as life stage or body size. For instance, within freshwater food webs body
476
size is an important determinant of trophic interactions, and food web structure predicted using body
477
size alone may be more accurate than those predicted using phylogeny alone (Woodward et al., 2010).
478
This could further increase the realism of the constructed food webs and hence their wider
479
applicability and usefulness.
480
This dataset of trophic interactions was collated to test the performance of the WebBuilder
481
function when predicting the structure of the four empirical UK freshwater food webs used here. It
482
would be straightforward to extend the coverage of this dataset by augmenting it with data collected
483
from other geographic regions. If this dataset is used to construct food webs in the future researchers
484
will need to use their discretion to decide how applicable it is to their system. For instance, this initial
485
version of the dataset does not provide good coverage of lentic species, or species from across Europe
486
or other parts of the world. However, interaction data are being published at a rapidly accelerating rate
487
(Ings et al., 2009) and this can be used to form an iterative feedback process, improving data quality
488
over time; the presence of links predicted by the WebBuilder function can therefore be tested
489
evermore rigorously in the future. Identifying underrepresented nodes in the dataset will help target
490
further research more cost effectively: e.g., a great deal is known about the diet of a handful of often
491
economically valuable species in the dataset (for instance, brown trout, Salmo trutta appears >3,000
492
times), but very little is known about many others. Additionally, technologies such as those provided
493
by recent advances in molecular sequencing will improve the efficiency of trophic interaction
494
detection (Clare, 2014) and therefore the volume of data which can be incorporated into the dataset.
495
We actively encourage researchers with suitable data to contribute them to this dataset. Exciting
496
initiatives such as Global Biotic Interactions (Poelen et al., 2014), by incorporating necessary
497
information such as the method through which an interaction was determined, could provide a global,
498
open source repository of interaction data which the WebBuilder function could access through R.
499
As more of these unknown links become known, nodes will not need to be generalised taxonomically
500
in order to find matches in the dataset, the links generated will more closely match the known links
501
for those species and therefore the quality of the ecological food webs generated by the
502
WebBuilder function will improve.
503
504
4.3 - Conclusions
505
We have demonstrated that the food webs generated here are comparable to empirically observed
506
food webs and exceeded the accuracy of other potential methods of predicting freshwater food webs.
16
507
This method could be used to build vast numbers of ecological networks from data that already exists,
508
such as routine biomonitoring data which is collected in huge volumes in many parts of the world
509
(e.g., Dutch soil biomonitoring data have recently been used to build a large collection of food webs;
510
Cohen & Mulder 2014). Producing collections of replicable networks is vital for advancing ecological
511
network research beyond the largely unreplicated case-study approach that has dominated to date: the
512
WebBuilder function approach presents a new robust and repeatable method that helps move us
513
considerably closer to that goal.
514
5 - Acknowledgements
515
This study is a contribution from the Imperial College Grand Challenges in Ecosystems and the
516
Environment initiative. CG was supported by Queen Mary University of London and the Freshwater
517
Biological Association. GW, DMP and LNH were supported by the Natural Environment Research
518
Council (Grants reference: NE/J015288/1 and NE/J011193/1). We thank Samraat Pawar and Uli
519
Brose for their helpful comments on an earlier draft that greatly improved the manuscript, as well as
520
Stefano Allesina who helped with the analysis of the ‘Difference’, ‘Ratio’ and ‘Difference/Ratio’
521
models.
522
6 - References
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
Allesina, S., 2011. Predicting trophic relations in ecological networks: A test of the Allometric Diet
Breadth Model. J. Theor. Biol. 279, 161–168. doi:DOI 10.1016/j.jtbi.2010.06.040
Allesina, S., Pascual, M., 2009. Food web models: A plea for groups. Ecol. Lett. 12, 652–662.
doi:10.1111/j.1461-0248.2009.01321.x
Allouche, O., Tsoar, A., Kadmon, R., 2006. Assessing the accuracy of species distribution models:
prevalence, kappa and the true skill statistic (TSS). J. Appl. Ecol. 43, 1223–1232.
doi:10.1111/j.1365-2664.2006.01214.x
Barnes, C., 2008. Predator and prey body sizes in marine food webs. Ecology 89, 881.
Bascompte, J., Jordano, P., Melian, C.J., Olesen, J.M., 2003. The nested assembly of plant-animal
mutualistic networks. Proc. Natl. Acad. Sci. U. S. A. 100, 9383–9387. doi:DOI
10.1073/pnas.1633576100
Berlow, E.L., Neutel, A.-M., Cohen, J.E., De Ruiter, P.C., Ebenman, B., Emmerson, M., Fox, J.W.,
Jansen, V.A.A., Iwan Jones, J., Kokkoris, G.D., Logofet, D.O., McKane, A.J., Montoya, J.M.,
Petchey, O., 2004. Interaction strengths in food webs: issues and opportunities. J. Anim. Ecol.
73, 585–598. doi:10.1111/j.0021-8790.2004.00833.x
Briand, F., 1983. Environmental Control of Food Web Structure. Ecology 64, 253–263.
Clare, E.L., 2014. Molecular detection of trophic interactions: emerging trends, distinct advantages,
significant considerations and conservation applications. Evol. Appl. 7, 1144–1157.
doi:10.1111/eva.12225
Cleveland, W., Grosse, E., Shyu, W., Chambers, J., Hastie, T., 1992. Local regression models. Stat.
Model. S.
Cohen, J.E., Mulder, C., 2014. Soil invertebrates, chemistry, weather, human management, and
edaphic food webs at 135 sites in The Netherlands: SIZEWEB. Ecology 95, 578.
Cohen, J.E., Newman, C.M., Briand, F., 1985. A Stochastic Theory of Community Food Webs: II.
Individual Webs. Proc. R. Soc. B Biol. Sci. doi:10.1098/rspb.1985.0043
17
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
De Ruiter, P.C., Neutel, A.M., Moore, J.C., 1995. Energetics, Patterns of Interaction Strengths, and
Stability in Real Ecosystems. Science (80-. ). 269, 1257–1260. doi:DOI
10.1126/science.269.5228.1257
Edwards, F.K., Lauridsen, R.B., Fernandes, W.P.A., Beaumont, W.R.C., Ibbotson, A.T., Scott, L.,
Davies, C.E., Jones, J.I., 2009. Re-introduction of Atlantic salmon, Salmo salar L., to the
Tadnoll Brook, Dorset. Proc. Dorset Nat. Hist. Archaeol. Soc. 130, 9–16.
Eklof, a., Helmus, M.R., Moore, M., Allesina, S., 2012. Relevance of evolutionary history for food
web structure. Proc. R. Soc. B Biol. Sci. 279, 1588–1596. doi:10.1098/rspb.2011.2149
Gilljam, D., Thierry, A., Edwards, F.K., Figueroa, D., Ibbotson, A.T., Jones, J.I., Lauridsen, R.B.,
Petchey, O.L., Woodward, G., Ebenman, B., 2011. Seeing Double: Size-Based and Taxonomic
Views of Food Web Structure. Adv. Ecol. Res. 45, 67–133. doi:Doi 10.1016/B978-0-12386475-8.00003-4
Goldwasser, L., Goldwasser, L., Roughgarden, J., Roughgarden, J., 1993. Construction and analysis
of a large Caribbean food web. Ecology 74, 1216–1233.
Gravel, D., Poisot, T., Albouy, C., Velez, L., Mouillot, D., 2013. Inferring food web structure from
predator–prey body size relationships. Methods Ecol. Evol. n/a–n/a. doi:10.1111/2041210X.12103
Gray, C., Baird, D.J., Baumgartner, S., Jacob, U., Jenkins, G.B., O’Gorman, E.J., Lu, X., Ma, A.,
Pocock, M.J.O., Schuwirth, N., Thompson, M., Woodward, G., 2014. Ecological networks: the
missing links in biomonitoring science. J. Appl. Ecol. 51, 1444–1449. doi:10.1111/13652664.12300
Hall, S.J., Raffaelli, D., 1991. Food-Web Patterns: Lessons from a Species-Rich Web. J. Anim. Ecol.
60, 823–841.
Havens, K.E., 1993. Predator-Prey Relationships in Natural Community Food Webs. Oikos 68, 117.
doi:10.2307/3545316
Henson, K.S.E., Craze, P.G., Memmott, J., 2009. The restoration of parasites, parasitoids, and
pathogens to heathland communities. Ecology 90, 1840–1851. doi:Doi 10.1890/07-2108.1
Hudson, L.N., Emerson, R., Jenkins, G.B., Layer, K., Ledger, M.E., Pichler, D.E., Thompson,
M.S.A., O’Gorman, E.J., Woodward, G., Reuman, D.C., 2013. Cheddar: analysis and
visualisation of ecological communities in R. Methods Ecol. Evol. 4, 99–104. doi:10.1111/2041210X.12005
Ings, T.C., Montoya, J.M., Bascompte, J., Bluthgen, N., Brown, L., Dormann, C.F., Edwards, F.,
Figueroa, D., Jacob, U., Jones, J.I., Lauridsen, R.B., Ledger, M.E., Lewis, H.M., Olesen, J.M.,
van Veen, F.J.F., Warren, P.H., Woodward, G., 2009. Ecological networks - beyond food webs.
J. Anim. Ecol. 78, 253–269. doi:DOI 10.1111/j.1365-2656.2008.01460.x
Kaiser-Bunbury, C.N., Muff, S., Memmott, J., Muller, C.B., Caflisch, A., Müller, C.B., 2010. The
robustness of pollination networks to the loss of species and interactions: a quantitative approach
incorporating pollinator behaviour. Ecol. Lett. 13, 442–452. doi:DOI 10.1111/j.14610248.2009.01437.x
Layer, K., Hildrew, A., Monteith, D., Woodward, G., 2010. Long-term variation in the littoral food
web of an acidified mountain lake. Glob. Chang. Biol. no–no. doi:10.1111/j.13652486.2010.02195.x
Layer, K., Hildrew, A.G., Woodward, G., 2013. Grazing and detritivory in 20 stream food webs
across a broad pH gradient. Oecologia 171, 459–471. doi:10.1007/s00442-012-2421-x
Layer, K., Riede, J.O., Hildrew, A.G., Woodward, G., 2010. Food Web Structure and Stability in 20
Streams Across a Wide pH Gradient. Adv. Ecol. Res. 42, 265–299. doi:Doi 10.1016/S00652504(10)42005-X
Ledger, M.E., Brown, L.E., Edwards, F.K., Milner, A.M., Woodward, G., 2012. Drought alters the
structure and functioning of complex food webs. Nat. Clim. Chang. 3, 223–227.
doi:10.1038/nclimate1684
Macfadyen, S., Craze, P.G., Polaszek, A., van Achterberg, K., Memmott, J., 2011. Parasitoid diversity
reduces the variability in pest control services across time on farms. Proc. R. Soc. B-Biological
Sci. 278, 3387–3394. doi:DOI 10.1098/rspb.2010.2673
18
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
Memmott, J., Martinez, N.D., Cohen, J.E., 2000. Predators, parasitoids and pathogens: species
richness, trophic generality and body sizes in a natural food web. J. Anim. Ecol. 69, 1–15.
doi:DOI 10.1046/j.1365-2656.2000.00367.x
Mulder, C., Elser, J.J., 2009. Soil acidity, ecological stoichiometry and allometric scaling in grassland
food webs. Glob. Chang. Biol. 15, 2730–2738. doi:10.1111/j.1365-2486.2009.01899.x
Olesen, J.M., Bascompte, J., Dupont, Y.L., Jordano, P., 2007. The modularity of pollination networks.
Proc. Natl. Acad. Sci. U. S. A. 104, 19891–19896. doi:DOI 10.1073/pnas.0706375104
Olito, C., Fox, J.W., 2014. Species traits and abundances predict metrics of plant – pollinator network
structure , but not pairwise interactions. Oikos 1–9. doi:10.1111/oik.01439
Petchey, O.L., Beckerman, A.P., Riede, J.O., Warren, P.H., 2011. Fit, efficiency, and biology: some
thoughts on judging food web models. J. Theor. Biol. 279, 169–71.
doi:10.1016/j.jtbi.2011.03.019
Petchey, O.L., Beckerman, A.P., Riede, J.O., Warren, P.H., 2008. Size, foraging, and food web
structure. Proc. Natl. Acad. Sci. U. S. A. 105, 4191–4196. doi:10.1073/pnas.0710672105
Piechnik, D. a., Lawler, S.P., Martinez, N.D., 2008. Food-web assembly during a classic
biogeographic study: Species’ “trophic breadth” corresponds to colonization order. Oikos 117,
665–674. doi:10.1111/j.0030-1299.2008.15915.x
Pocock, M.J.O., Evans, D.M., Memmott, J., 2012. The robustness and restoration of a network of
ecological networks. Science (80-. ). 335, 973–7. doi:10.1126/science.1214915
Pocock, M.J.O.O., Memmott, J., Evans, D.M., Memmott, J., 2012. Supporting Material - The
robustness and restoration of a network of ecological networks. Science (80-. ). 335, 973–7.
doi:10.1126/science.1214915
Poelen, J.H., Simons, J.D., Mungall, C.J., 2014. Global biotic interactions: An open infrastructure to
share and analyze species-interaction datasets. Ecol. Inform. 24, 148–159.
doi:10.1016/j.ecoinf.2014.08.005
R Core Team, 2013. R: A language and environment for statistical computing.
Rall, B.C., Brose, U., Hartvig, M., Kalinkat, G., Schwarzmüller, F., Vucic-Pestic, O., Petchey, O.L.,
2012. Universal temperature and body-mass scaling of feeding rates. Philos. Trans. R. Soc.
Lond. B. Biol. Sci. 367, 2923–34. doi:10.1098/rstb.2012.0242
Rohr, R.P., Scherer, H., Kehrli, P., Mazza, C., Bersier, L.-F., 2010. Modeling food webs: exploring
unexplained structure using latent traits. Am. Nat. 176, 170–177. doi:10.1086/653667
Sáyago, R., Lopezaraiza-Mikel, M., Quesada, M., Álvarez-Añorve, M.Y., Cascante-Marín, A.,
Bastida, J.M., 2013. Evaluating factors that predict the structure of a commensalistic epiphytephorophyte network. Proc. Biol. Sci. 280, 20122821. doi:10.1098/rspb.2012.2821
Strong, J.S., Leroux, S.J., 2014. Impact of Non-Native Terrestrial Mammals on the Structure of the
Terrestrial Mammal Food Web of Newfoundland, Canada. PLoS One 9, e106264.
doi:10.1371/journal.pone.0106264
Thompson, R.M., Brose, U., Dunne, J.A., Hall, R.O., Hladyz, S., Kitching, R.L., Martinez, N.D.,
Rantala, H., Romanuk, T.N., Stouffer, D.B., Tylianakis, J.M., 2012. Food webs: reconciling the
structure and function of biodiversity. Trends Ecol. Evol. 27, 689–697. doi:DOI
10.1016/j.tree.2012.08.005
Tylianakis, J.M., Laliberté, E., Nielsen, A., Bascompte, J., 2010. Conservation of species interaction
networks. Biol. Conserv. 143, 2270–2279. doi:http://dx.doi.org/10.1016/j.biocon.2009.12.004
Vázquez, D.P., Chacoff, N.P., Cagnolo, L., 2009. Evaluating multiple determinants of the structure of
plant-animal mutualistic networks. Ecology 90, 2039–2046. doi:10.1890/08-1837.1
Vazquez, D.P., Melian, C.J., Williams, N.M., Bluthgen, N., Krasnov, B.R., Poulin, R., 2007. Species
abundance and asymmetric interaction strength in ecological networks. Oikos 116, 1120–1127.
doi:DOI 10.1111/j.2007.0030-1299.15825.x
Verdú, M., Valiente-Banuet, A., 2011. The relative contribution of abundance and phylogeny to the
structure of plant facilitation networks. Oikos 120, 1351–1356. doi:10.1111/j.16000706.2011.19477.x
Vizentin-Bugoni, J., Maruyama, P.K., Sazima, M., 2014. Processes entangling interactions in
communities: forbidden links are more important than abundance in a hummingbird-plant
network. Proc. R. Soc. B Biol. Sci. 281, 20132397–20132397. doi:10.1098/rspb.2013.2397
Williams, R.J., Martinez, N.D., 2000. Simple rules yield complex food webs. Nature 404, 180–183.
19
656
657
658
659
660
661
662
663
664
665
Woodward, G., Blanchard, J., Lauridsen, R.B., Edwards, F.K., Jones, J.I., Figueroa, D., Warren, P.H.,
Petchey, O.L., 2010. Individual-Based Food Webs: Species Identity, Body Size and Sampling
Effects. Adv. Ecol. Res. 43, 211–266. doi:10.1016/b978-0-12-385005-8.00006-x
Woodward, G., Brown, L.E., Edwards, F.K., Hudson, L.N., Milner, A.M., Reuman, D.C., Ledger,
M.E., 2012. Climate change impacts in multispecies systems: drought alters food web size
structure in a field experiment. Philos. Trans. R. Soc. B-Biological Sci. 367, 2990–2997.
doi:DOI 10.1098/rstb.2012.0245
Woodward, G., Speirs, D.C., Hildrew, A.G., 2005. Quantification and resolution of a complex, sizestructured food web. Adv. Ecol. Res. 36, 85–135.
666
667
7 - Figure captions
668
669
Figure 1. A simplified workflow demonstrating the WebBuilder function. For a workable example
670
see Appendix C.
671
672
Figure 2. An example of inputs and outputs for the WebBuilder function. Two different scenarios are
673
highlighted. Firstly in blue the taxa Baetis fuscatus is generalised to the genus level for both its
674
consumer and resource links, this allows it to be matched with B. scambus in the registry and the
675
Navicula tripunctata – B. fuscatus, and Cocconeis placentula – B. fuscatus links to be included in the
676
output. Secondly in green, the taxa Tanypodinae are generalised to the family level for its resource
677
links and subfamily level for its consumer links, allowing it to matched with all entries in the registry
678
with the subfamily Tanypodinae and the Tanypodinae – Salmo trutta, and Baetis fuscatus –
679
Tanypodinae links to be included in the output.
680
681
Figure 3. The quality of the generated food web increases with the size of the dataset. Fluctuations in
682
the TSS score are caused by changes in the component parts of the TSS, i.e. while the TPR may
683
increase as the dataset size increases, other metrics such as the FPR might also increase, causing the
684
total TSS to fall (see methods). Lines are fitted using a LOESS smoother (Cleveland et al., 1992),
685
grey shading indicates the 95% confidence intervals.
686
687
688
Figure 4. Box plots showing the changes in generated trophic interactions as the level of taxonomic
689
generalisation is varied. The difference in degree (top), generality (middle) and vulnerability
690
(bottom) of individual nodes between the generated and empirical food web, thus the sample size
691
reflects the number of nodes in the empirical food web. Positive values represent links which were
692
‘missed’ by the WebBuilder function, while negative values represent additional links not found
20
693
empirically. Stars indicate if the mean is different from zero (one sampled t-test) and indicate if the
694
generated trophic interactions are different from that of the empirical food web, 0.05 >= p > 0.01 =
695
*, 0.01 >= p > 0.001 = **, p <= 0.001=***.
696
697
698
Figure 5. Predation matrices for the empirical food webs compared to those generated by the AR
699
method. Nodes are ordered by increasing body mass. A trophic link is represented by a point
700
indicating that the taxon in that column consumes the taxon in that row. Links generated by the
701
WebBuilder function are represented by empty circles, and those found empirically are
702
represented by smaller, filled circles.
703
704
Figure 6. Box plots showing the performance of the WebBuilder function compared to the ADBM,
705
Difference, Ratio and Difference/Ratio models. The performance of the WebBuilder function is
706
plotted as four vertical lines, one for each of the empirical food webs; Broadstone
707
, Tadnoll
and Mill Stream
, Afon Hirnant
. The TSS score (top panel) gives an overall measure of the
708
performance of the predictive method relative to empirical food webs, and varies between 1 (a
709
generated food web that is exactly the same as the empirical food web) and -1 (a generated food web
710
which is the exact inverse of the empirical food web). The TPR (True Positives Rate; bottom panel) is
711
the proportion of generated food web links that were also found empirically, and varies between 0 (no
712
links generated correctly) and 1 (all links generated correctly). A box plot of each set of values is
713
given, indicating the range, quartile ranges and median of each set of values. For the WebBuilder
714
function only the individual scores for the four food webs are shown, for all others there are too many
715
generated scores to be shown individually; ADBM (n=508), Difference (n=32,025), Ratio (n=43,638)
716
and Difference/Ratio (n=41,602).
717
718
21
719
Appendices:
720
Appendix A:
721
Table A.1 – Table of sources of trophic information contained in the dataset
722
Figure A.1 - The trade-off between taxonomic resolution and dataset coverage
723
Figure A.2 – The performance of the WebBuilder function compared to predictive models.
724
Appendix B: Function R code
725
Appendix C: Example R code
726
Appendix D: The trophic links predicted by the WebBuilder function.
727
22
728
Supporting Information
729
Appendix A
730
Table A.1 The sources of trophic-interaction data used in the database.
Source
Number of
trophic links
6433
6,280
2,440
1,423
316
286
199
170
153
145
144
143
108
81
78
65
57
41
37
33
35
34
33
30
30
27
20
20
13
10
9
9
Gilljam et al. 2011
Layer et al. 2010
Ledger et al. 2012
Brose et al. 2005
Warren 1989
Becker 1990
Jones et al. 1951
Northcott 1981
Hynes 1950
Moore & Potter 1976
Iversen 1988
Spänhoff et al. 2003
Thomas 1962
Slack 1936
Clitherow et al. 2013
Maitland 1965
Lancaster et al. 2005
Rowan Dunn 1954
Radforth 1940
Woodward et al. 2008
Woodward et al. 2005
Woodward unpublished
Woodward et al. 2008
Badcock 1949
Mackereth 1957
Cook 1979
Perkins unpublished
Townsend & Hildrew 1979
Tikkanen et al. 1997
Harper-Smith et al. 2004
Englund 2005
N. Dewhurst & G. Woodward
unpublished data
23
System
Place
freshwater stream
UK
freshwater stream
experimental freshwater
channels
freshwater lake
experimental freshwater
stream
freshwater pond
UK
USA
freshwater stream
Europe
freshwater river
UK
freshwater lake
laboratory experimental
freshwater
freshwater stream
UK
freshwater stream
laboratory experimental
freshwater
freshwater stream
UK
Europe
freshwater stream
Europe
freshwater river
UK
freshwater river
UK
freshwater river
Europe
freshwater river
UK
freshwater stream
UK
freshwater lake
UK
freshwater river
UK
freshwater stream
UK
UK
UK
UK
UK
UK
UK
freshwater stream
UK
freshwater
unknown
freshwater stream
UK
freshwater stream
UK
freshwater river
UK
freshwater lake
UK
freshwater stream
UK
freshwater stream
UK
freshwater lake
Europe
7
4
4
3
3
2
2
2
2
1
Young & Procter 1986
Mann & Blackburn 1991
Warren, unpublished
Friday 1988
Gee & Young 1993
Elliott et al. 1988
Fox 1978
Harrison et al. 2005
Hall et al. 2000
Armitage & Young 1990
freshwater lake
USA
freshwater stream
experimental freshwater
stream
freshwater lake
UK
Europe
freshwater stream
UK
freshwater lake
UK
freshwater
UK
freshwater
UK
UK
freshwater stream
UK
freshwater stream
USA
731
732
733
References:
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
Armitage, M.J. & Young, J.O. (1990). The realized food niches of three species of stream-dwelling
triclads (Turbellaria). Freshwater Biology, 24, 93–100.
Badcock, R.M. (1949). Studies in Stream Life in Tributaries of the Welsh Dee. Journal of Animal
Ecology, 18, 193–208.
Becker, G. (1990). Comparison of the dietary composition of epilithic trichopteran species in a firstorder stream. Archiv für Hydrobiologie, 120, 13–40.
Brose, U., Cushing, L., Berlow, E.L., Jonsson, T., Banasek-Richter, C., Bersier, L.-F., Blanchard, J.L.,
Brey, T., Carpenter, S.R., Blandenier, M.-F.C., Cohen, J.E., Dawah, H.A., Dell, T., Edwards, F.,
Harper-Smith, S., Jacob, U., Knapp, R.A., Ledger, M.E., Memmott, J., Mintenbeck, K.,
Pinnegar, J.K., Rall, B.C., Rayner, T., Ruess, L., Ulrich, W., Warren, P., Williams, R.J.,
Woodward, G., Yodzis, P. & Martinez, N.D. (2005). Body sizes of consumers and their
resources. Ecology, 86, 2545.
Clitherow, L.R., Carrivick, J.L. & Brown, L.E. (2013). Food Web Structure in a Harsh Glacier-Fed
River. Plos One, 8, e60899.
Cook, M.P. (1979). The interrelationships of fish and zooplankton in gravel-pit lakes. City of London
Polytechnic.
Elliott, J.M., Humpesch, U.H. & Macan, T.T. (1988). Larvae of the British Ephemeroptera: a key
with ecological notes.
Englund, G. (2005). Scale dependent effects of predatory fish on stream benthos. Oikos, 1, 19–30.
Fox, P.J. (1978). The population dynamics of the bullhead (Cottus gobio L., Pisces) with special
reference to spawning, mortality of young fish and homoestatic mechanisms. Journal of Fish
Biology, 12.
Friday, L.E. (1988). A key to the adults of British water beetles. Dorset Press, Dorset, UK.
Gee, H. & Young, J.O. (1993). The food niches of the invasive Dugesia tigrina (Girard) and
indigenous Polycelis tennis Ijima and P. nigra (M�ller) (Turbellaria; Tricladida) in a Welsh
lake. Hydrobiologia, 254, 99–106.
Gilljam, D., Thierry, A., Edwards, F.K., Figueroa, D., Ibbotson, A.T., Jones, J.I., Lauridsen, R.B.,
Petchey, O.L., Woodward, G. & Ebenman, B. (2011). Seeing Double: Size-Based and
Taxonomic Views of Food Web Structure. Advances in Ecological Research, 45, 67–133.
Hall, R.O., Wallace, J.B. & Eggert, S.L. (2000). Organic matter flow in stream food webs with
reduced detrital resource base. Ecology, 81, 3445–3463.
Harper-Smith, S., Berlow, E.L., Roland, A., Williams, R.J. & Martinez, N.D. (2004). Ecology through
food webs: visualising and quantifying the effects of stocking alpine lakes with trout. Dynamic
Webs: Multispecies Assemblages, Ecosystem Development, and Environmental Change (eds P.
De Ruiter, J.C. Moore & V. Wolters), pp. 407–423. Elsevier Academic Press Inc.
Harrison, S.S.C., Bradley, D.C. & Harris, I.T. (2005). Uncoupling strong predator-prey interactions in
streams: the role of marginal macrophytes. Oikos, 108, 433–448.
24
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
Hynes, H.B.N. (1950). The Food of Fresh-Water Sticklebacks (Gasterosteus aculeatus and Pygosteus
pungitius), with a Review of Methods Used in Studies of the Food of Fishes. Journal of Animal
Ecology, 19, 36–58.
Iversen, T.M. (1988). Secondary production and trophic relationships in a spring invertebrate
community. Limnology and Oceanography, 33, 582–592.
Jones, J.R.E., Journal, S., May, N. & Jones, B.Y.J.R.E. (1951). An Ecological Study of the River
Towy. Journal of Animal Ecology, 20, 68–86.
Lancaster, J., Bradley, D.C., Hogan, A. & Waldron, S. (2005). Intraguild omnivory in predatory
stream insects. Journal of Animal Ecology, 74, 619–629.
Layer, K., Riede, J.O., Hildrew, A.G. & Woodward, G. (2010). Food Web Structure and Stability in
20 Streams Across a Wide pH Gradient. Advances in Ecological Research, 42, 265–299.
Ledger, M.E., Brown, L.E., Edwards, F.K., Milner, A.M. & Woodward, G. (2012). Drought alters the
structure and functioning of complex food webs. Nature Climate Change, 3, 223–227.
Mackereth, J.C. (1957). Notes on the Plecoptera from a Stony Stream. Journal of Animal Ecology, 26,
343–351.
Maitland, P.S. (1965). The Feeding Relationships of Salmon, Trout, Minnows, Stone Loach and
Three-Spined Stickle-Backs in the River Endrick, Scotland. Journal of Animal Ecology, 34,
109–133.
Mann, R.H.K. & Blackburn, J.H. (1991). The biology of the eel Anguilla anguilla (L.) in an English
chalk stream and interactions with juvenile trout Salmo trutta L. and salmon Salmo salar L.
Hydrobiologia, 218, 65–76.
Moore, J.W. & Potter, I.C. (1976). A Laboratory Study on the Feeding of Larvae of the Brook
Lamprey Lampetra planeri (Bloch). Journal of Animal Ecology, 45, 81–90.
Northcott, D.S. (1981). The role of aquatic macrophytes in the availabillty of food for young fish. City
of London Polytechnic.
Radforth, I. (1940). The Food of the Grayling (Thymallus thymallus), Flounder (Platichthys flesus),
Roach (Rutilus rutilus) and Gudgeon (Gobio fluviatilis), with Special Reference to the Tweed
Watershed. Journal of Animal Ecology, 9, 302–318.
Rowan Dunn, D. (1954). The Feeding Habits of some of the Fishes and some Members of the Bottom
Fauna of Llyn Tegid (Bala Lake), Merionethshire. Journal of Animal Ecology, 23, 224–233.
Slack, H.D. (1936). The Food of Caddis Fly (Trichoptera) Larvae. Journal of Animal Ecology, 5,
105–115.
Spänhoff, B., Schulte, U., Alecke, C., Kaschek, N. & Meyer, E.I. (2003). Mouthparts, gut contents,
and retreat-construction by the wood-dwelling larvae of Lype phaeopa (Trichoptera:
Pschomyiidae). European Jounal of Entomology, 100, 563–570.
Thomas, J.D. (1962). The Food and Growth of Brown Trout (Salmo trutta L.) and its Feeding
Relationships with the Salmon Parr (Salmo salar L.) and the Eel (Anguilla anguilla (L.)) in the
River Teify, West Wales. Journal of Animal Ecology, 31, 175–205.
Tikkanen, P., Muotka, T., Huhta, A. & Juntunen, A. (1997). The roles of active predator choice and
prey vulnerability in determining the diet of predatory stonefly (Plecoptera) nymphs. Journal of
Animal Ecology, 66, 36–48.
Townsend, C.R. & Hildrew, A.G. (1979). Resource Partitioning by Two Freshwater Invertebrate
Predators with Contrasting Foraging Strategies. Journal of Animal Ecology, 48, 909–920.
Warren, P.H. (1989). Spatial and temporal variation in the structure of a fresh-water food web. Oikos,
55, 299–311.
Woodward, G., Papantoniou, G., Edwards, F. & Lauridsen, R.B. (2008). Trophic trickles and cascades
in a complex food web: impacts of a keystone predator on stream community structure and
ecosystem processes. Oikos, 117, 683–692.
Woodward, G., Speirs, D.C. & Hildrew, A.G. (2005). Quantification and resolution of a complex,
size-structured food web. Advances in Ecological Research, 36, 85–135.
Young, J.O. & Procter, R.M. (1986). Are the lake-dwelling leeches, Glossiphonia complanata (L.) and
Helobdella stagnalis (L.), opportunistic predators on molluscs and do they partition this food
resource? Freshwater Biology, 16, 561–566.
25
825
26
826
827
828
829
830
831
Figure A.1, The percentage of food web nodes that appear in the database, compared to the accuracy
of the resulting food web prediction. The taxonomic resolution of the food web nodes are reduced
systematically in order to increase their occurrence in the database.
27
832
833
834
835
836
837
838
839
840
Figure A.2. The performance of the WebBuilder function compared to the ADBM, Diff, Ratio and
Diff/Ratio models. The TSS score gives an overall measure of the performance of the predictive
method relative to empirical webs, TPR (True Positives Rate) is the proportion of generated food web
links which were also found empirically. A box plot of each set of values is given, indicating the
range, quartile ranges and median of each set of values. For the WebBuilder function only the
individual scores for the four food webs are shown, for all others the mean of each food web is shown.
28
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
Appendix B
# Functions for synthesising food webs from a registry of
# known trophic links
.StripWhitespace <- function(v)
{
# Returns v with leading and trailing whitespace removed
return (gsub('^\\s+|\\s+$', '', v, perl=TRUE))
}
.AddRegistryLinks <- function(synthesized, registry, res.col, con.col,
res.method, con.method, reference.cols,
debug)
{
stopifnot(res.col %in% colnames(synthesized))
stopifnot(res.col %in% colnames(registry))
stopifnot(con.col %in% colnames(synthesized))
stopifnot(con.col %in% colnames(registry))
if(any(!reference.cols %in% colnames(registry)))
{
missing <- setdiff(reference.cols, colnames(registry))
stop('Missing reference columns: ', paste(missing, collapse=','))
}
# Match on multiple columns by pasting
# http://www.r-bloggers.com/identifying-records-in-data-frame-a-thatare-not-contained-in-data-frame-b-%E2%80%93-a-comparison/
# Strip whitespace and convert to lowercase
a <- lapply(synthesized[,c(res.col,con.col)], .StripWhitespace)
a <- lapply(a, tolower)
a <- do.call('paste', a)
b <- lapply(registry[,c(res.col,con.col)], .StripWhitespace)
b <- lapply(b, tolower)
b <- do.call('paste', b)
Debug <- function(...) if(debug) cat(...)
Debug('Matching on resource', res.method, 'consumer', con.method, '\n')
# Match where both resource and consumer columns are non-empty and a
link
# does not already exist
to.match <- ''!=synthesized[,res.col] & ''!=synthesized[,con.col] &
is.na(synthesized$N.records)
for(row in which(to.match))
{
N <- sum(a[row]==b)
if(N>0)
{
synthesized$res.method[row] <- res.method
synthesized$con.method[row] <- con.method
synthesized$N.records[row] <- N
for(col in reference.cols)
{
29
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
synthesized[,col] <paste(sort(unique(registry[a[row]==b,col])), collapse=',')
}
}
}
return (synthesized)
}
.AllPossible <- function(nodes, exclude.cols=NULL)
{
# A data.frame of all possible links
resource <- consumer <- nodes[,setdiff(colnames(nodes),
exclude.cols),drop=FALSE]
colnames(resource) <- paste0('res.', tolower(colnames(resource)))
colnames(resource)['res.node'==colnames(resource)] <- 'resource'
colnames(consumer) <- paste0('con.', tolower(colnames(consumer)))
colnames(consumer)['con.node'==colnames(consumer)] <- 'consumer'
synthesized <- data.frame(resource[rep(1:nrow(resource),
each=nrow(resource)),,drop=FALSE],
consumer[rep(1:nrow(resource),
times=nrow(resource)),,drop=FALSE])
return (synthesized)
}
.ARWorker <- function(synthesized, registry, methods, reference.cols,
debug)
{
# Add links for each level in methods
for(outer in methods)
{
for(inner in methods[1:which(methods==outer)])
{
synthesized <- .AddRegistryLinks(synthesized, registry,
ifelse('exact'==outer, 'resource',
paste0('res.', outer)),
ifelse('exact'==inner, 'consumer',
paste0('con.', inner)),
outer, inner, reference.cols, debug)
if(inner!=outer)
{
synthesized <- .AddRegistryLinks(synthesized, registry,
ifelse('exact'==inner, 'resource',
paste0('res.', inner)),
ifelse('exact'==outer, 'consumer',
paste0('con.', outer)),
inner, outer, reference.cols, debug)
}
}
}
synthesized <- droplevels(synthesized[!is.na(synthesized$N.records),])
rownames(synthesized) <- NULL
return (synthesized)
}
WebBuilder <- function(nodes, registry,
methods=c('exact','genus','subfamily','family','order','class'),
reference.cols=c('linkevidence', 'source.id'), debug=FALSE)
30
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
{
# nodes - a data.frame containing columns 'node' (character - unique
node
#
names) and either 'minimum.method' the minumum method for
that
#
taxon or 'minimum.res.method' and 'minimum.con.method' - the
#
minimum methods for that taxon as a resource and a consumer
#
respectively. Values in 'minimum.method',
'minimum.res.method' and
#
'minimum.con.method' must be in the 'methods' argument.
# registry - a data.frame of known trophic interactions
# methods - character vector of methods, in order
# reference.cols - names of columns in registry that will be included
in the
#
returned data.frame
# Checks
stopifnot(0<nrow(nodes))
stopifnot(nrow(nodes)==length(unique(nodes$node)))
stopifnot(all(c('resource','consumer') %in% colnames(registry)))
error.msg <- paste('nodes should contain either "minimum.method" or',
'"minimum.res.method" and "minimum.con.method"')
if('minimum.method' %in% colnames(nodes))
{
if(any(c('minimum.res.method', 'minimum.con.method') %in%
colnames(nodes)))
{
stop(error.msg)
}
# Use minimum.method for both ends of the links
nodes$minimum.res.method <- nodes$minimum.con.method <nodes$minimum.method
}
else
{
if(!all(all(c('minimum.res.method', 'minimum.con.method') %in%
colnames(nodes))))
{
stop(error.msg)
}
# Use the provided minimum.res.method and minimum.con.method
}
stopifnot(all(nodes$minimum.res.method %in% methods))
stopifnot(all(nodes$minimum.con.method %in% methods))
synthesized <- .AllPossible(nodes,
c('minimum.res.method','minimum.con.method','minimum.method'))
synthesized[,c('res.method','con.method')] <- ''
synthesized$N.records <- NA
synthesized[,reference.cols] <- ''
links <- .ARWorker(synthesized, registry, methods, reference.cols,
debug)
# Set factor levels
# What client has specified
31
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
nodes$minimum.res.method <- factor(nodes$minimum.res.method,
levels=methods)
nodes$minimum.con.method <- factor(nodes$minimum.con.method,
levels=methods)
# What RegistryLinks found
links$res.method <- factor(links$res.method, levels=methods)
links$con.method <- factor(links$con.method, levels=methods)
links <links[as.integer(links$res.method)<=as.integer(nodes$minimum.res.method)[ma
tch(links$resource, nodes$node)] &
as.integer(links$con.method)<=as.integer(nodes$minimum.con.method)[match(li
nks$consumer, nodes$node)],]
return(droplevels(links))
}
32
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
Appendix C
#Gray et al Food Webs MS
#worked example of how to use the WebBuilder function within teh cheddar
package
library(cheddar)
source("WebBuilder.R")
#add in links
#read in without factor levels so NAs can be removed later
registry<-read.csv('Gray et al SuppMat data.csv')
data(BroadstoneStream)
BroadstoneStream <- OrderCommunity(BroadstoneStream, 'M')
IsolatedNodes(BroadstoneStream)
BroadstoneStream<-RemoveNodes(community=BroadstoneStream,
remove=c('Algae','CPOM','FPOM','Leptothrix spp.','Terrestrial
invertebrates','Diptera'))
#extract node information
#n.b. node names need to be resolved using the GBIF database on Global
Names Resolver
#in order to match the taxonomy used in the database
nps<-NPS(BroadstoneStream)
#supply a list of desired level of generalisation, either a single list
with one entry per node, or two
#with two entries per node, i.e. a 'resource method' and a 'consumer
method'
minimum.res.method<- c("class",
"family",
"family",
"family",
"family",
"family",
"family",
"family",
"family",
"genus",
"genus",
"family",
"family",
"genus",
"genus",
"family",
"family",
"genus",
"genus",
"order",
"genus",
"family",
"class",
"family",
"genus",
"family",
33
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
"family",
"genus",
"genus",
"family",
"genus")
minimum.con.method <- c("class",
"genus",
"genus",
"genus",
"genus",
"genus",
"genus",
"family",
"genus",
"genus",
"genus",
"family",
"family",
"genus",
"genus",
"genus",
"genus",
"genus",
"genus",
"order",
"genus",
"family",
"class",
"family",
"genus",
"family",
"family",
"genus",
"genus",
"family",
"genus")
nodes<-cbind(nps[,c(1,7:10)],minimum.res.method,minimum.con.method)
links<-WebBuilder(nodes, registry,
method=c('exact','genus','family','order','class'))
BroadstoneExample <- Community(properties = CPS(BroadstoneStream),
nodes = NPS(BroadstoneStream),
trophic.links = links)
table(links$con.method, links$res.method)
#visualise network
PlotWebByLevel(BroadstoneExample, level='ChainAveragedTrophicLevel')
34