Libro Parte 2

Chapter 8
Plant-Herbivore Networks in the Tropics
Antonio López-Carretero, Ek del-Val, and Karina Boege
Abstract Understanding the patterns and processes behind the high biological
diversity of tropical ecosystems has been one of the most important issues in mod-
ern ecology. Plant-herbivore interactions constitute an important percentage of
biodiversity in the tropics, and their ecological and evolutionary importance has
been demonstrated in a large number of studies. However, it is only very recently
that plant-herbivore antagonistic interactions are being addressed from the
perspective of complex networks to evaluate how different factors influence their
interaction patterns. In this chapter, we provide a summary of the processes that
have been reported shaping the specialization and structuring of tropical plant-
herbivore networks. From the limited availability of studies in such habitats we
suggest that plant-herbivore networks are spatiotemporally dynamic and are the
result of multiple non-exclusive processes where seasonality, variation in
resource availability, habitat type, disturbance regime and species-specific
attributes contribute to structuring these highly diverse and specialized
antagonistic networks.
8.1 Introduction
Trophic interactions involving plants and their herbivores account for approxi-
mately 40% of global terrestrial biodiversity (Price 2002), which is mainly
concentrated in tropical ecosystems (Novotny and Basset 2005; Lewinsohn et al.
2006; Beck and Khen 2007), and also represent one of the major conduits of
energy flow to higher trophic levels (Futuyma and Agrawal 2009). The
diversification of
A. López-Carretero (*)
Facultad de Agronomía, Universidad de Buenos Aires,
Ciudad Autónoma de Buenos Aires, Argentina
e-mail: [email protected]
E. del-Val
Instituto de Investigaciones en Ecosistemas y Sustentabilidad, Universidad Nacional
Autónoma de México, Morelia, Michoacán, México
K. Boege
Instituto de Ecología, Universidad Nacional Autónoma de México,
Ciudad de México, México
© Springer International Publishing AG 2018

111
W. Dáttilo, V. Rico-Gray (eds.), Ecological Networks in the Tropics,
https://doi.org/10.1007/978-3-319-68228-0_8
112 A. López-Carretero et al.
herbivorous insects in the tropics is evolutionarily associated with plants (Futuyma

and Agrawal 2009; Forister et al. 2015), and the high diversity and specificity of
insect-plant interactions renders herbivory as one of the processes determinant in
the distribution and composition of plant species (Bagchi et al. 2014; Kempel
et al. 2015). Therefore, studying the different ecological and evolutionary pro-
cesses that influence the structure of plant-herbivore ecological networks is an
important step in understanding patterns of tropical biodiversity (Lewinsohn et al.
2005; Becerra 2015).
In biological communities, herbivorous species and plants interact with many
other species, originating an intricate network of trophic interactions. Network
analyses have been a useful tool to describe plant-animal interactions and to
analyze their role in the organization and dynamics of biodiversity in tropical
ecosystems (Bascompte and Jordano 2007). Furthermore, complex network
analyses can be informative to assess if the evolution of plants and their consumers
occurs through one-to-one coevolutionary processes or are the outcome of diffuse
coevolution, where the assemblage of multiple species play an important role. In
particular, different metrics used to describe the structure of plant-herbivore
networks can be useful to identify evolutionary interaction units, defined by the
species influencing fitness of target species under diffuse coevolution (Thompson
1982). In spite of the ecological and evolutionary importance of antagonistic plant-
herbivore interactions, just a few studies have examined these interactions from a
network perspective in the tropics.
In general, plant-herbivore networks are highly specialized mainly due to the
evolution of physical and chemical defenses of host plants and, in turn, the adapta-
tion of herbivores to such defenses (Wardhaugh et al. 2014; Becerra 2015; Forister
et al. 2015). Given this context, the preference of herbivores to feed on closely
related plants (Funk et al. 1995; Janz et al. 2001; Weiblen et al. 2006; Benítez-
Malvido and Dáttilo 2015) or those with similar leaf attributes (Becerra 1997;
Agrawal 2007) has been found to be one of the most important factors promoting
modularity in antagonistic networks (Thompson 1994; Prado and Lewinsohn
2004). Furthermore, the ecological and phylogenetic mechanisms that originate the
structural patterns of antagonistic networks have only recently begun to be
explored despite important prior advances in the identification and description of
the structure of different plant-herbivore networks. Research from temperate
ecosystems has shown how specialization and structural attributes of plant-
herbivore networks are influenced by phylogenetic signals (Volf et al. 2017),
intraspecific genetic variation of plants (Barbour et al. 2016; Lau et al. 2016),
habitat restrictions, and species attributes (Cagnolo et al. 2011) in addition to
habitat disturbances (Valladares et al. 2012).
In tropical ecosystems, patterns of plant-herbivore networks can vary with
respect to habitat disturbance (Villa-Galaviz et al. 2012; de Araújo et al. 2015;
Benítez-Malvido et al. 2014, 2016) and spatiotemporal variation in plant availabil-
ity (López-Carretero et al. 2016), host plant resistance (López-Carretero et al.
2016), herbivores specificity, and the degree of intimacy of interactions (Novotny
et al. 2010; Pires and Guimarães 2012). Even so, the study of processes
influencing
8 Plant-Herbivore Networks in the Tropics 113
the specificity and structural patterns of plant-herbivore networks is still under

development. Therefore, the objective of this chapter is to provide an update on
these processes, particularly in tropical ecosystems. We first characterize the
structural patterns of tropical plant-herbivore networks in relation to other types of
networks in non-tropical habitats. In addition, we review how antagonistic
networks vary spatially and temporarily, and the biotic and abiotic factors
influencing this variation. Then, we address how habitat disturbance and the
species-specific attributes of plants and their herbivores affect the structure and
resilience of antagonistic networks. Finally, we provide a conclusion with future
directions for the study of plant-herbivore networks in the tropics.
8.2 Structure of Plant-Herbivore Interaction Networks
In contrast with mutualistic networks that tend towards generalization,

antagonistic networks are more specialized, less nested, and have greater
modularity (Lewinsohn et al. 2006; Thébault and Fontaine 2010; Sauve et al.
2014). Nevertheless, within antagonistic networks, the degree of
compartmentalization depends on the interacting trophic guilds, the type of
ecosystem, and even latitude. In the case of plant- herbivore networks,
specialization and hence modularity can vary as a function of herbivore guilds,
being highly specialized in the case of herbivores living inside the plants (e.g.,
miners and galling insects), but less specialized in the case of free living
herbivores (e.g., suckers and chewers, Novotny et al. 2010, Pires and Guimarães
2012). Although there are not enough studies to perform formal analyses to com-
pare between temperate vs. tropical networks, a marked latitudinal gradient
towards a greater diversity (Novotny and Miller 2014), and specialization (Dyer et
al. 2007, Forister et al. 2015, but see Novotny et al. 2006, Stork 2007) of
herbivores in the tropics has been reported. From the available empirical evidence
cited in this chapter, we detected larger networks with greater specialization and
a higher number of rare hervibore species in tropical ecosystems. Other
parameters, such as connectance, modularity, and nestedness, seem to be
equivalent between tropical and temperate forests. However, more studies on
plant-herbivore networks are needed across large latitudinal gradients and among
different ecosystems to assess significant differences between temperate and
tropical ecosystems.
8.3 Plant-Herbivore Interaction Networks in the Tropics
Tropical plant-herbivore interactions have been studied since long time ago
(Janzen 1970; Coley 1983; Coley and Barone 1996); however, only recently the
interaction network approach has been considered in this area. The studies
comprise evaluations of different herbivore guilds, including leaf chewers
(lepidopteran, orthop- teran, and coleopteran), leaf miners (dipteran and
lepidopteran), and sapsuckers
Fig. 8.1 Distribution of studies of plant-herbivorous networks in the tropical regions of the
planet. Each symbol represents the locations of the studies reviewed in this chapter. Symbol
forms indicate different orders of herbivorous insects (circles = Lepidoptera larvae, triangles =
adult Coleoptera, diamonds = dipteran larvae, square = Hemiptera, and stars = studies that
include more than one insect order). Colors represent the different life forms of the herbivorous
host plants studied (blue = herbs, purple = vines, green = trees, red = more than one plant life
form)
(hemipteran), which are concentrated in the Neotropical region, mainly focused on

plant herb hosts (Fig. 8.1 and Table 8.1). There is a lack of studies that incorporate
different orders of herbivores associated with host plants of different life forms. In
addition, there are some regions were plant-herbivore networks have not been
explored, in particular studies in tropical Africa are unknown (Fig. 8.1 and Table
8.1). In general, studies have reported that tropical plant-herbivore networks have
high levels of specialization coinciding with other studies reporting diet breath of
herbivores being narrower in lower latitudes (Erwin 1991; Novotny et al. 2004,
2006; Forister et al. 2015). Network size (i.e., species richness and number of
interactions) is quite variable as a function of the taxonomical level of different
groups of herbivores. In terms of network parameters, most studies in the
tropics have found a modular structure (Meskens et al. 2011; Villa-Galaviz et al.
2012; López-Carretero et al. 2014; Wardhaugh et al. 2014; Bergamini et al.
2016) with the exception of work by Eben and Espinosa de los Monteros (2015),
in which they report a plant- Chrisomelid network without a modular structure,
in spite of being very specialized. A study in Australian rainforest reports
similar levels of specialization in plant-herbivore and flower visitor networks
although the herbivore network was modular and the flower visitor was nested
(Wardhaugh et al. 2014). Thus, in this case, the topology of interaction networks
appears to be independent of the level of specificity. Interestingly, some tropical
food webs, although are extremely complex,
seem to be dominated by relatively few interactions (Novotny and Miller 2014).
Table 8.1 Information of the plant-herbivore networks studies included in this review 8
Factor evaluated NP L/ Order P P H H Region Habitat type Source Pl
A F S F S ant
Phylogenetic signal M L CDL 1 2 6 5 Neo Cerrado Savanna Bergamini et al. (2016) -
2 5 Bra He
rbi
Phylogenetic signal CHM A C 8 N 1 4 Ame Several Eben and Espinosa de
vo
s 3 tropical, los Monteros (2015)
re
subtropical and Ne
temperate tw
habitats or
Host taxonomy C G H M N NO A C 5 1 1 3 Neo Rain forest Meskens et al. (2011) ks
and phenotype 7 0 Pan in
Forest canopy H A C 1 2 5 8 Aus Rain forest Wardhaugh et al. (2014) the
3 3 8 Aus Tr
Plant diversity G A H 5 1 8 3 Ori Chi Evergreen Staab et al. (2015) op
6 3 subtropical forest
Habitat type and C G H IE M V L L 2 5 1 1 Neo Different López-Carretero et al.
resource availability 0 6 2 7 Mex coastal (2014)
6 seasonal
habitats
Anti-herbivore d ss L L 1 2 9 1 Neo Different López-Carretero et al.
defenses 6 9 0 Mex coastal (2016)
4 seasonal
habitats
Habitat disturbance C D H LD VR A CDHL 1 2 n ns Neo Rain forest Benítez-Malvido et al.
L s Mex (2014)
Secondary succession C N M EC R H L L 3 1 2 4 Neo Dry forest Villa-Galaviz et al. (2012)
R 6 4 1 7 Mex
0 1
DNA barcodes use – A C 5 3 1 2 Neo Rain forest García-Robledo et al.
3 0 CR (2013)
Forest fragmentation DI N A CHL 1 1 9 2 Neo Rain forest Benítez-Malvido et al. 11
L 9 Mex 2016 5
Interaction intimacy C H N ID M A AHHyL n 2 1 9 Neo Cultivars and other Benitez-Malvido et al.
NO L s 4 Bra habitats (2015)
Herbivore guild ES HI A CDHL 3 8 4 3 Oce Rain forest Novotny et al. (2010)
L OP 8 8 4 9 PNG
9
NP network parameter evaluated, M modularity, c connectance, H network-level specialization, G generality, N nestedness, NO Niche overlap, IE interaction
evenness, V vulnerability, d species-level specialization, ss species strength, ES effective specialization, HI host plant isolation, LD linkage density, VR
variance ratio, EC extinction curve, R robustness, DI diversity of interactions, NO niche overlap. Herbivore developmental stage: L larval, A adult. Herbivore
order: C Coleoptera, D Diptera, H Hemiptera, Hy Hymenoptera, L Lepidoptera, O Orthoptera, P Phasmidae, A Acari. PF number of plant families, PH
number of plant species, HF number of herbivore families, HS number of herbivore species. Region: Ame several habitats types throughout the Americas,
Neo neotropical, Aus Australian, Ori oriental, Oce oceanian, Mex México, Bra Brazil, Chi China, CR Costa Rica, PNG Papua New Guinea. ns not specified
8.4 Spatiotemporal Variation of Plant-Herbivore Networks
In some tropical and non-tropical ecosystems where seasonality and climatic con-
ditions are not widely variable, specialization, and structure of plant-herbivore
networks remain stable over time and space (Wardhaugh et al. 2014; Kemp et al.
2016). However, the contrasting monthly and seasonal climatic variation (mainly
in precipitation and temperature) of some deciduous and sub-deciduous tropical
ecosystems affect the availability and quality of host plants (Janzen 1993; Coley
1998; Pearse and Hipp 2012) and therefore can promote important changes in
herbivore composition (Janzen 1993; Coley and Barone 1996), diet breath
(Scherrer et al. 2016), specialization, and structural patterns of plant-herbivore
networks (López-Carretero et al. 2014, 2016). For example, environments that
provide high richness and abundance of host plants (i.e., habitats rich in host spe-
cies with different life histories or rainy season) promote specialization, modular-
ity, and interaction evenness of plant-lepidopteran herbivore networks, as
observed in Mexican (López-Carretero et al. 2014) (Fig. 8.2b, c) and Brazilian
tropical forests (Scherrer et al. 2016). In contrast, in environments where resource
availability for herbivores is restricted (i.e., poorly structured habitats and/or
marked dry season), selectivity and modularity of plant-herbivore networks
decrease due to the dominance of generalist herbivores (López-Carretero et al.
2014) that are capable of expanding their host preferences in face of adverse
conditions (Scherrer et al. 2016, Fig. 8.2b, c).
The temporal increase in the availability and structural complexity of plant spe-
cies in the warmest and rainy months promote a great variety of microhabitats and
host plant richness, which in turn reduce niche overlap and promote a more equita-
ble distribution of food resources (i.e., interaction evenness) among herbivores in
the network (López-Carretero et al. unpublished data) (Fig. 8.2d). This pattern is
consistent with the notion that specialization of herbivores allows for the coexis-
tence of species through a fine distribution of trophic niches (Futuyma and
Moreno 1988; Dyer et al. 2007; Lewinsohn and Roslin 2008). In this context,
comparative studies of plant-herbivore interaction networks in tropical wet vs. dry
forests would be revealing, and comparisons between seasonal temperate and
tropical forests could help to disentangle the influence of species diversity vs.
seasonality on plant- herbivore interaction networks (Dirzo and Boege 2008).
In addition, tropical plant communities show a marked vertical stratification, as
does the assembly of associated herbivores. For example, herbivore diversity is
greater in the canopy than in the understory, and consequently the faunistic
similarity between the two strata is low (Basset et al. 2003; Ribeiro and Basset
2007). This is likely promoted by differences in abiotic conditions, the availability
of good quality foliage and predation risk between these strata (Van Bael et al.
2003; Boege and Marquis 2006; Neves et al. 2014). Nevertheless, the ways in
which the structure of herbivore networks varies with respect to vertical
stratification are still largely unknown.
Fig. 8.2 Spatiotemporal variation in selectiveness and structural patterns of plant-lepidoptera net-
work in a seasonal tropical ecosystem of México. (a) Weighted bipartite network between 176
caterpillar species (right nodes) and 56 plant species (left nodes) with different life history.
Linkage width indicates the frequency of each herbivory interaction. (b) Variation of
macroscopic network parameters in different habitats types that vary in their complexity (PIO
pioneer dune vegetation, DUN coastal dune scrub, FFW tropical lowland floodplain forest with a
wetland ecotone, SFY recently established tropical lowland sub-deciduous forest, SFO lowland
sub-deciduous forest in advanced stage of succession) (López-Carretero et al. 2014). (c)
Variation in network parameters across seasonality in year 2011: Dry1 (April-March), Dry2
(May-June), Wet1(July-August), Wet2 (September-October) (Dry = Dry season, Wet = Rainy
season). Different letters represent significant differences (P < 0.05); and (d) Significant negative
relation between monthly PC1 values (precipitation + temperature) and herbivore niche overlap
(P < 0.05). PC1 is a synthetic variable that describe the simultaneous variation of precipitation
and temperature, greater values of PC1 corresponded to greater precipitation and temperature
(López-Carretero et al., unpublished data)
8.5 Influence of Phylogeny, Species, and Genetic

Diversity on Plant-Herbivore Networks
Antagonistic plant-herbivore interactions are highly specialized as a result of

different ecological factors, including plant diversity, the biochemistry of host
plants, and historical factors (Ehrlich and Raven 1964; Weiblen et al. 2006;
Futuyma and
Agrawal 2009; Richards et al. 2015). In particular, evolutionary history often plays
an important role in determining both community assemblages and species interac-
tion networks (Peralta 2016, Volf et al. 2017 but see Novotny et al. 2010). This
influence can be detected through phylogenetic signals found in the properties of
plant-herbivore interaction networks. In general, studies have shown that phyloge-
netic distances influence interacting species, modularity, and network nestedness
(Peralta 2016). For example, in a leaf beetle-plant interaction network in Panamá,
Meskens et al. (2011) found that plant phylogeny constrains herbivore modules
and therefore determines network parameters. However, Bergamini et al. (2016)
reported a strong phylogenetic signal for flower-head herbivores but a mild signal
for interacting plants. Other studies on tropical herbivore communities in New
Guinea have found that host phylogenetic distance explains a significant fraction
of the variance in herbivore community similarity for some herbivore guilds
(Novotny et al. 2010). In particular, caterpillar species show higher phylogenetic
clustering than coleopter- ans and orthopteroids (Weiblen et al. 2006). However,
recent studies from temperate forest showed that not all levels of host plant
phylogeny are equal in their effect on structuring plant-herbivore food web, which
depends on the specialization of insect guild considered (Volf et al. 2017).
In some cases, plant defensive traits do not correlate with plant phylogeny
(Becerra 1997; Novotny et al. 2010) but are determined by local environmental
and ecological factors. For example, although the interactions between
Diabroticina beetles (Chrysomelidae) and their host plants (Cucurbitaceae) have
been considered to be the product of a coevolutionary process directed by the
secondary metabolites of hosts (Metcalf 1986); recently, Eben and Espinosa de los
Monteros (2015) showed that the structural patterns of plant-chrysomelid
networks do not correspond with plant phylogeny but are rather related to the
chemical and morphological similarity of the host plants.
8.5.1 Plant Diversity
Although plant diversity has been found to influence the stability of food webs
(Haddad et al. 2011), its effects on plant-herbivore network parameters has been
scarcely studied. Staab et al. (2015) investigated a plant-hemipteran network in a
gradient of plant diversity in subtropical China, and concluded that the number
and abundance of host species was not determinant for network structure as mod-
ularity and specialization remained constant across the gradient. This could be
due to the specific feeding behavior and particular morphological structures char-
acteristic of sap-sucker insects, which feed only on plant taxa to which they are
adapted (Walling 2008). In the case of caterpillars, for example, levels of specific
selectivity and species strength in plant-lepidopteran networks seem to be better
explained by foliar host plant traits than their relative abundance (López-Carretero
et al. 2016).
8.5.2 Genetic Diversity
Intraspecific genotypic variation in plant traits can also have an important

influence in herbivore communities (Whitham et al. 2006), food web complexity
(Barbour et al. 2016), and hence in plant-herbivore interaction networks in
general. For example, an analysis of genotype-species network has revealed that
genotypic variation in Populus angustifolia influences modularity, nestedness,
centralization, and species co-occurrence of the assemblages of associated
herbivores (Lau et al. 2016). The influence of genotypic variation in plant-
herbivore networks, however, has not been assessed in tropical systems and
requires further attention. Incorporating the evolutionary history and genetic
variation of species into the study of plant-herbivore interaction networks can
greatly contribute with our understanding of community assembly rules and
processes, community dynamics, and resilience.
8.6 How Host Plant and Herbivorous Traits Can

Affect Plant-Herbivore Networks
8.6.1 Host Defensive Traits
Foliar damage inflicted by herbivores strongly affects plant growth and reproduc-
tion (Rausher and Feeny 1980; Marquis 1984; Maron and Kauffman 2006).
Therefore, this antagonistic interaction results in continuous evolutionary
processes influenced by reciprocal selection. During these processes, plants
express defensive traits that reduce herbivore damage and in turn, herbivores adapt
to the defensive strategies of plants (Ehrlich and Raven 1964; Thompson 2005). In
the tropics, reciprocal selection can be so intense that dissimilarity in plant
defensive traits is positively associated with herbivore diversity and
specialization (Novotny et al. 2002; Weiblen et al. 2006; Forister et al. 2015). For
example, in tropical plant genera that maintain enormous herbivore richness, such
as Inga (Kursar et al. 2009), Bursera (Becerra 1997, 2007), 2015), and Piper
(Richards et al. 2015), the diversity of defensive metabolites is positively related
with the diversification and specialization of herbivores, both at ecological and
geological scales (Gentry 1982, 1989; Richards et al. 2015).
Although coevolutionary processes involving the continuous defense and
counter- defense of plants and their herbivores have been referenced to explain the
high specificity of plant-herbivore interactions (Thompson 1994; Coley and
Barone 1996; Prado and Lewinsohn 2004) the influence of specific physical and
chemical defensive plant traits on the specialization and structuring of plant-
herbivore networks has been prac- tically unexplored. A recent study in a Mexican
tropical forests showed that the simultaneous expression of foliar toughness,
trichome density, and phenolic content in different host species influences the
selectivity of caterpillars and the strength of plant
species (trophic importance; López-Carretero et al. 2016). In particular, host

species with high foliar toughness and low contents of foliar phenolic compounds
were related to high species strength and selectiveness, which suggest that plants
with this combination of attributes represent an important food resource for the
assemblage of specialist herbivores in the network. On the other hand, during the
dry season, host species with high trichome density and leaf phenolic content were
consumed by a small number of highly selective herbivorous species capable of
feeding on this combination of leaf attributes (López-Carretero et al. 2016).
Leaf morphology can also play an important role in the structure of plant-
herbivore networks. For example, in a Chrysomelinae-plant network in Barro
Colorado Island, Panama, the architecture of immature leaves forming standing
pools in some host families (Zingiberales) but not in others (Arecales and Poales)
have influenced the adaptive zones for different hispine beetles influencing as a
consequence the modularity within this interaction network (Meskens et al. 2011).
8.6.2 Herbivore Feeding Behavior
The specificity and intimacy (i.e., physiological dependence) of plant-herbivore

interactions, which transcend the general structure of herbivory networks, vary as
a function of the different herbivore guilds, plant developmental stage, and/or
organs consumed (Novotny et al. 2010). In the tropics, the degree of specialization
of different insect guilds that consume plants varies enormously, and the
percentages of specialization of different guilds is granivorous (99%), leaf
miners (96%), frugivores (83%), leaf chewers (56%), sap suckers (56%),
xylophages (24%), and root feeders (10%) (Novotny and Basset 2005). In general,
herbivorous guilds with low mobility that consume highly defended plants (e.g.,
lepidopteran leaf chewers) or those with high physiological dependence on their
hosts (e.g., leaf miners and gall-inducing insects) have highly specialized and
modular interaction networks, in comparison with more generalist herbivorous
guilds (e.g., phloem suckers, root eaters, borers). However, because most studies
on plant-herbivore networks have focused on exter- nal leaf eaters guilds
(particularly on caterpillars and chewing beetles) for practical reasons, these
studies may not be ecologically representative of all plant-herbivore networks
(Novotny et al. 2010). In fact, studies of herbivore networks including
endophagous herbivores or root feeders are nonexistent (Fig. 8.1).
8.7 Habitat Disturbance and Plant-Herbivore

Interaction Network Resilience
Ecosystems are in constant change due to natural and anthropogenic disturbances

and as a consequence, species interactions are modified after perturbation. Natural
disturbance in forests is generally associated with changes in tree canopy,
increased light availability, and drier conditions. These sudden changes imply the
disappearance
of some plant species and their associated specialist herbivores, but also facilitate
the production of fresh leaves on the remaining plants representing an increase in
resource availability for different herbivore species. Some studies have reported
that canopy openness is a determinant factor for lepidopteran richness and
abundance after disturbance (Barlow et al. 2007; Hawes et al. 2009) but
hemipterans have shown differential impacts depending on the species (Osborn
1935; Wolcott 1941). More recently, Grimmbacher and Stork (2009) investigated
changes in beetle assemblages in a fragmented tropical forest in Australia,
following the passage of tropical cyclone Larry finding that communities were
similar to their pre-hurricane condition since only 5% of the species was found to
have changed.
A logical prediction after the empirical evidence on the impacts of disturbance
on herbivore communities is that plant-herbivore network parameters should also
be affected by such perturbations. However, there are only a handful of studies
address- ing this prediction. Valladares et al. (2006, 2012) investigated changes in
a plant- herbivore network in the Argentinean Chaco forest after forest
fragmentation and found that species richness, network size, and connectance
diminished as the forest area decreased as a function of the area of forest
remnants. Benítez-Malvido et al. (2014, 2016) assessed anthropogenic
perturbation on a plant-arthropod network comparing forest gaps and forest edges
finding that the network properties were not affected by disturbance. Villa-Galaviz
et al. (2012) investigated if plant-lepidopteran network structure was recovered
during forest succession and found that indeed, after few years of land
abandonment all network parameters are indistinguishable from those found in
mature undisturbed forests. Therefore, with the available information, it appears
that plant-herbivore network attributes in the tropics are resilient to natural and
anthropogenic perturbations (as long as the disturbed area is sur- rounded by
sources of the original biodiversity), but further investigations are needed. In fact,
a recently published global analysis of the effect of land-use intensity on plant-
herbivore networks compared 72 individual networks and found that some
parameters are affected by land-use change. In particular, the most susceptible
parameter is network connectance, which increases under high levels of anthropo-
genic alteration due to a decrease in specialization. This suggests that specialist
herbivores are more susceptible to disturbance (de Araújo et al. 2015).
8.8 Concluding Remarks and Future Directions
Records of plant-herbivore interactions for used in network analyses have been

based on small geographic areas. Hence, recorded trophic interactions have only
included a small fraction of local floristic diversity and often considering only
common plant species or plants of the same taxonomic group or life story (Fig.
8.1). Because the estimation of herbivore specificity and the emergent network
structural patterns depend on the diversity, composition, and life history of the
plants considered (Novotny et al. 2006; Weiblen et al. 2006), estimates of
network-level and species-level specialization in tropical forests are still elusive.
This demonstrates the
importance of including greater ecological and taxonomic amplitude in studies on

tropical plant-herbivore networks (Novotny et al. 2010).
Furthermore, the vast majority of studies on herbivory networks have focused
on characterizing and evaluating variation of metrics that aim to capture the
overall structure of a network, considering macroscopic parameters such as
modularity, interaction evenness, robustness, and network-level specialization.
Very few studies have included species-level metrics (i.e., species-level
specialization, species strength, role of species, and food preferences). In solely
focusing on macroscopic network parameters, important information on the
dynamics of specific entities that constitute a network (of species) may be missing.
This is of obvious importance considering that plant-herbivore interactions are
highly specie-specific and are finely regulated according to the attributes of plant
and herbivore species. Additionally, given the importance of anti-herbivore
defenses in patterns of tropical plant-herbivore interactions, the role of different
chemical (i.e., saponins, alkaloids, and/or cyanogens compounds), and biotic
defenses (i.e., ants associated with extra- floral nectaries) should also be
considered in the study of process that influences the configuration of herbivory
networks.
Until recent years, a handicap for the study of plant-herbivore networks in the
tropics was the vast undescribed diversity, particularly for insect groups (Novotny
and Miller 2014). Accessible molecular tools such as DNA barcoding represent
today a viable opportunity to improve our knowledge on plant-herbivore networks
and make network comparisons across regions and ecosystems possible. For
example, García-Robledo et al. (2013) evaluated a plant-leaf-roller beetle
interaction network through DNA barcoding and demonstrated that host plant
identifications at the species level are accurate providing a sample size of four
individuals per herbivore and per plant. Because this methodology is a cost-
effective strategy for reconstruct- ing plant-herbivore interactions we expect to see
more studies in tropical systems using this approach in the coming years, which
will enhance our understanding of the ecology and evolution of species
assemblages in the tropics.
In addition, new statistical methods proposed to estimate diet breath may allow
to incorporate plant traits to the metrics and increase our understanding of the
factors influencing plant-herbivore interaction networks. For example, Fordyce
et al. (2016) have recently proposed an ordination methodology to quantify diet
breath, based not only in the associations of plants and herbivores, but also includ-
ing the multivariate distances among plants in ordination space.
Finally, incorporating the other trophic levels can be revealing as well. Some
authors are starting to test weather observed patterns in plant-herbivore networks
constructed from field data are related with herbivore preferences or other factors
such as predation and competition and have experimentally demonstrated that
herbivore preference can explain the modular structure in their leafhopper-plant
network (Augustyn et al. 2016).
Acknowledgements While writing the manuscript, EDV and KB were funded by PAPIIT-UNAM
IN211916, SEP CONACYT 2015-255544 and ALC by PDTS 2600/14 (CONICET). KB
acknowl-
edges logistic support by Rubén Pérez-Ishiwara.
References
Chapter 9
Host-Parasite Networks: An Integrative
Overview with Tropical Examples
Sybelle Bellay, Fabrício Hiroiuki Oda, Karla Magalhães Campião,

Fábio Hideki Yamada, Ricardo Massato Takemoto,
and Edson Fontes de Oliveira
Abstract The lack of ecological studies on parasite species is contradictory with

their high diversity. In this context, the ecological network approach has drawn
attention to patterns of tropical host-parasite interactions; however, it is still very
restricted to a few taxonomic groups. In tropical regions, studies on host-parasite
networks have encompassed specially fish, amphibian hosts, and malaria parasite
communities of birds. High specificity is peculiar to parasites, either considering
the host species or the site of infection. Variations in specificity indicate
differentiated structural patterns between ecto- and endoparasites networks, as
well as larval and adult stages of parasites, influencing the values of connectivity,
nestedness, and modularity. Host characteristics also influence the structure of
networks, including phylogenetic relationships and diet. Similarly to free-living
species in ecosystems suffering the influences of environmental change, parasite
species may also be extinct due to primary extinctions of hosts or possibly present
new patterns of interaction due to the arrival of invasive species in the
environment. New studies should
S. Bellay (*) • R.M. Takemoto

Programa de Pós-Graduação em Ecologia de Ambientes Aquáticos Continentais,
Núcleo de Pesquisas em Limnologia, Ictiologia e Aquicultura, Laboratório de
Ictioparasitologia, Universidade Estadual de Maringá, Maringá, Paraná, Brazil
F.H. Oda
Programa de Pós-Graduação em Promoção da Saúde,
Centro Universitário Cesumar, Maringá, Paraná, Brazil
K.M. Campião
Departamento de Zoologia, Universidade Federal do Paraná,
Curitiba, Paraná, Brazil
F.H. Yamada
Departamento de Ciências Biológicas, Universidade Regional do Cariri,
Crato, Ceará, Brazil
E.F. de Oliveira
Departamento de Engenharia Ambiental, Programa de Pós-Graduação em Engenharia
Ambiental, Universidade Tecnológica Federal do Paraná, Campus Londrina,
Londrina, Paraná, Brazil

127
https://doi.org/10.1007/978-3-319-68228-0_9
128 S. Bellay et al.
explore network patterns between tropic and non-tropic environments, in addition

to the use of molecular tools to identify especially cryptic parasites, which may
provide better-supported results.
9.1 Introduction
Parasites are hidden components of ecological communities, comprising approxi-

mately 40% of the species described worldwide (Dobson et al. 2008)—a scope in
which the tropics retain a large number of species yet to be described. These
“invis- ible” organisms play an important role in ecosystems (Holt and Boulinier
2005) for their ability to directly or indirectly affect host species at different levels
of organization (i.e., individual, population, and community levels) (Poulin 1999;
Wood and Johnson 2015). Lafferty and Kuris (1999) stated that the complexity
and ubiquity of host-parasite interactions are good arguments when considering
parasitism to assess anthropogenic impacts on aquatic systems. Nevertheless,
parasites have been neglected in the scope of conservation or management
strategies (Gómez and Nichols 2013; Dougherty et al. 2015).
Each parasite species exhibits a degree of specialization by its resource (e.g.,
host species or available microhabitat). Ecto- and endoparasites are exposed to
different selection pressures and outcomes (see Thompson 1994) when exploring
the host, which creates differentiated patterns in the organization of the
communities. In addition, a wide variation occurs in the distribution and intensity
of their interactions with the host species (Thompson 1994; Poulin 2007). These
variations are usually related to biological, ecological, and/or phylogenetic aspects
(Poulin et al. 2013).
Over the past few years, a remarkable number of studies have used network
analysis to assess the biodiversity, ecological processes, and relationships between
community structure and functioning in tropic and non-tropic environments
(Hagen et al. 2012; Heleno et al. 2014). In ecological parasitology, the network
approach has been used seeking to understand a wide variety of issues, for
example, (1) the role of parasitic biodiversity in ecosystem functioning, the
transmission of parasitic diseases via social or sexual contacts, (2) influence of
structural attributes of host species and their dynamics regarding food webs, (3)
effect of species interaction on the structure of networks, (4) coevolutionary
components within a larger system, and (5) community response to disturbances
such as migration, extinction, or bio- logic invasion (Vázquez et al. 2005; Chen et
al. 2008; Dallas and Cornelius 2015; Strona and Lafferty 2016). In a review
synthesizing the findings from network analyses of host-parasite interactions and
food webs, Poulin (2010) concluded that the benefits of this approach far outweigh
its disadvantages.
Even though many interactions between parasites and vertebrate hosts are
known, this chapter provides an approach of ecological networks on the
interactions between fish and amphibians along with their parasites, in addition to
malaria parasite communities of birds in the tropics. Furthermore, we also
demonstrate other examples of interactions, such as host-bat fly networks.
9 Host-Parasite Networks: An Integrative Overview with Tropical Examples 129
9.2 Environments and Host Characteristics with Possible

Implications in Networks
By analyzing the data available in Bellay et al. (2015a), we observe that the
patterns of fish-parasite networks, such as connectivity and nestedness, differ
between tropical and temperate environments (Fig. 9.1), having resulted from
more specialized interactions in the tropics, reducing connectivity and
nestedness; even though both the modularity and species richness showed no
differences between environments. Specialization did not differ between the
tropical and temperate sites from the parasite perspective in multihost malaria
parasite communities of birds (Svensson- Coelho et al. 2014). Furthermore, studies
on lizards suggest that possible relationships between latitude and parasitic load
occur due to the host’s life history (Salkeld et al. 2008). Parasite-host interactions
are expected to present different patterns in relation to the taxonomic groups
involved, with greater evidence than between latitudes.
It is expected that the composition of the host community influences parasite
richness since a higher diversity of hosts represents a wider range of resources for
Fig. 9.1 Comparison between tropical and temperate environments regarding four parameters
from 44 fish-parasite networks. (a) Species richness corresponds to the total number of species in
the network (parasites + hosts); (b) Connectance; (c) Nestedness; (d) Modularity. Middle point:
Mean; Box: Standard Error; Whisker: Standard deviation. The significance of the Mann-Whitney
test was p < 0.05. Data available in Bellay et al. (2015a)
parasites (Hudson et al. 2006). Similarly, host and parasite abundance play an
important role at determining network structure. If trait matching underlies
potential interactions between species, abundance would determine their dynamic
(Canard et al. 2014). Empirical evidence indicates the occurrence of a few host
traits, such as abundance, body size, and diet, mostly contributing to the
maintenance of parasite diversity and network structure (Dallas and Cornelius
2015).
9.2.1 Host Specificity and Phylogenetic History
The combination of varying levels of host specificity and abundance contributes

to the high asymmetry observed in host-parasite networks. Specialist parasites
tend to interact with hosts presenting high parasite richness, whereas generalist
parasites tend to interact mainly with hosts containing low parasite richness
(Fortuna et al. 2010). This nested pattern, which is more evident in host-parasite
networks from temperate latitudes (Fig. 9.1c), may be related to host abun-
dance. High host abundance may contribute to harboring richer parasite faunas,
with higher proportion of rare specialists (Vázquez et al. 2005). Understanding
the mechanisms beneath the nested pattern in host-parasite networks might also
contribute to quantify the importance of neutrality in ecological interactions
(Krishna et al. 2008; Canard et al. 2014).
The interaction of hosts and parasites encompasses a high degree of intimacy
and adaptation between species (Fortuna et al. 2010). When different host spe-
cies share dietary, behavioral or habit preferences, the formation of modules in
the network is enhanced (Bellay et al. 2011; Lima-Junior et al. 2012; Bellay et al.
2013). Therefore, sharing ecological traits favors parasite sharing, which in turn
favors the formation of modules. Ecological and phylogenetic groups of related
host species can promote modularity in a way that the similarity (e.g., phyloge-
netic, ecological, or functional) is higher among species within the same module
(Fortuna et al. 2010; Bellay et al. 2011; Lima-Junior et al. 2012; Bellay et al.
2013). This has been observed in several networks, including fish (Fortuna et al.
2010; Bellay et al. 2011; Lima-Junior et al. 2012) and reptile (Brito et al. 2014)
hosts along with their parasites, highlighting the importance of historical and
ecological processes to network structure.
9.2.2 Host Body Size
Host body size has a positive relation to parasite species richness for a broad spec-
trum of host taxa (Kamiya et al. 2014; Campião et al. 2015b). Nonetheless, it is
expected that such influential trait played an important role in the architecture of
host-parasite networks. Indeed, it has been demonstrated that body size is an

important trait determining network structure and stability in ecological networks
(Woodward et al. 2005). Dallas and Cornelius (2015) found that fish biomass is
important predicting nestedness, which is influenced to host body size. These find-
ings suggest that species in antagonistic networks may be defined by host traits,
such as body mass.
9.2.3 Host Diet
For many parasite taxonomic groups of fish of a tropical floodplain, host diet is an
important factor affecting host-parasite interaction although weaker and less con-
served than phylogeny (Lima et al. 2016). For endoparasites, hosts that share food
items are also more likely to share parasites (Benesh et al. 2014). Consequently,
hosts that consume a wide variety of food items harbor a higher amount of parasite
species (Chen et al. 2008; Locke et al. 2014). An intriguing finding is a host with
the highest amount of shared parasites belonging to low trophic level species (i.e.,
they are preyed upon by other species of fish or birds) and with omnivorous diets
(Lima et al. 2016). A possible explanation would be associated with feeding
behavioral, exposing it to infection by free-living parasites, eggs, or cysts.
Many parasites trophically transmitted are able to furnish ecological
information on the links between their host and other organisms in a given
environment, where parasites may indicate long-term feeding information and
ontogenetic changes in the host’s diet (Marcogliese 2005). During ontogeny, the
host may undergo niche shifts and display feeding specialization; for instance,
adults often use different resources other than larvae or juveniles, thus affecting
the host-parasite network. Changes in the parasitic community of Prochilodus
lineatus (Characiformes, Prochilodontidae) during its ontogenetic shifts were also
supported by Lizama et al. (2005). Juvenile individuals of this host species live in
lagoons until 2 years, subse- quently living in river channels (Fig. 9.2). Such
changes in transition and adaptation expose the adult host to a higher degree of
parasitism.
Obviously, throughout all life stages, the resources are substitutable for the
most of host species; even though it may occur a certain degree of overlap in
resource use among stages. A host species with a broad diet can be less vulnerable
to secondary extinction; however, it could be an ontogenetic specialist (with
several stages, each consuming a different resource) (Rudolf and Lafferty 2011).
Seasonal shifts may also influence the characteristics of interaction networks by
changing the availability of food and host species density. Zarazúa-Carbajal et al.
(2016) showed the importance of biotic and abiotic factors in the dynamics of
host-bat fly interactions in temperate environments. It is necessary to assess the
approach of seasonality in tropical parasite-host networks.
Fig. 9.2 Schematic representation of the variation in the parasite species composition in the
Streaked prochilod, Prochilodus lineatus in a tropical floodplain. Ontogenetic shifts in the diet
and habitat transition during the host developmental stages (i.e., from juvenile stage to adult
stage) are factors driving the variation in the parasite species composition (circles). Colorful
circles represent the parasite species. Photos credits: Celso Ikedo (fish)
9.3 Parasite Characteristics Influencing Network Topology
The patterns observed in the topology of host-parasite networks result from the
match of several features of both interacting parts. Among these features, the
amount of host species that a parasite is able to interact with is undoubtedly a
crucial factor in network ecology. Host specificity is believed to be an intrinsic
species attribute, with some extent of phylogenetic signal as it tends to be more
similar among closely related taxa (Krasnov et al. 2011). It can be expressed as the
amount or the diversity of hosts a parasite is able to colonize. Moreover, host
specificity is not a fixed char- acter; it can vary within a species according to the
different localities of its geographic range or influences of the characteristics of
host populations (e.g., body size and life stage) (Fig. 9.3a); in addition, it may be
influenced especially by the composition of the host community (Fig. 9.3b),
stressing the importance of the whole ecosystem in the organization of each
network. This demonstrates the different scales of a study on host-parasite
interactions.
Hosts represent both the habitat and food resource to parasites and are
generally explored by parasite species from different lineages and with different
life strategies, such as endoparasites or ectoparasites as well as parasites with
simple or complex life
Fig. 9.3 Host-parasite networks. (a) Interaction network at the population level of hosts and their
parasites where each node is an individual, which may interact with a differentiated number of
parasites reflecting the influence of a particular host characteristics (e.g., sex or size) on the
interaction pattern; (b) Host-parasite interaction network at the community level where each node
is a species and the interactions reflect the total interactions observed in their populations.
Squares and circles represent the parasite species and size of the symbols, the number of
interactions. Hosts are illustrated in drawings of frogs. Data from parasite community available
in da Graça et al. (2017)
cycles. This variety in the ways that parasites exploit a resource may increase the
diversity of parasites associated with each host species (Bellay et al. 2013).
Parasites of different species may exhibit different levels of host specificity,
dependence, asymmetry, strength, and complementary specialization in such
networks (Bellay et al. 2015b). In turn, high parasite specificity favors the
formation of modules, while low parasite specificity contributes to an increased
amount of links in ecological networks. Studies conducted in the Neotropics with
amphibian hosts and their meta- zoan parasites demonstrated that the high
prevalence of parasite species associated with a broad range of host species
contributed to the high connectance and nestedness measures observed in this
system (Campião et al. 2015a, b).
The variation in the degree of specificity among parasites with different life
strategies and stages is also an important determinant of network structure. For
instance, fish ectoparasites tend to show higher host specificity than the
endoparasites associated with these hosts (Bellay et al. 2015a; Fig. 9.4).
However, due to the strong phylogenetic signal in parasitism strategy, it is still not
clear whether the influence is the mode of life itself or other intrinsic property of
the parasite clade. Networks including only fish ectoparasites have proved
significantly modular and weakly nested due to higher host specificity (Lima-
Junior et al. 2012; Braga et al. 2014).
Parasites at larval stage are normally generalists in fish hosts (Bellay et al.
2013), which is probably a strategy to enhance the chance of completing the
transmission cycle through different hosts. Moreover, parasites in larval stages are
important connectors in modular networks due to their low host specificity (Bellay
et al. 2013). In fact, the topology of fish-parasite networks is more closely related
to the parasite developmental stage than to their lineages. Therefore, the presence
of parasite larval stages may increase the strength of trophic links, contributing to
the stability of ecological networks.
In brief, the amount of interactions of a parasite species in the network might
reflect a trade-off between the benefit of having alternative resources (broader host
availability) and the cost of developing adaptations that allow such associations
Fig. 9.4 Fish-parasite interaction network of the upper Paraná River floodplain, Brazil. ( a)
Bipartite network considering ectoparasites (yellow squares) and endoparasites (green circles);
(b) unipartite host network, the links indicate that ectoparasite species sharing; (c) unipartite host
network, the links indicate that endoparasite species sharing. In (b) and (c), red triangles corre-
spond to the hosts and the thickness of the links indicates the intensity of parasite species are
sharing by host species. Data available in Takemoto et al. (2009)
(Poulin 2007). To completely understand why some parasite species are associated
with a broad spectrum of hosts, while some rely on a single host species, is still
something to be inquired. Notwithstanding, it is a key factor underlying the
topology of host-parasite networks. All of the varying patterns in host specificity
above men- tioned and discussed have profound influences on network topologies.
So far, the evidence we have reached indicates that the presence of generalist
parasites increases connectivity and nestedness measures, while the presence of
specialists contributes to the formation of modules in host-parasite networks
(Bellay et al. 2013; Campião et al. 2015b).
9.4 Robustness
Host extinctions or population decline may compromise the structure of parasite

communities, as observed in a study on the fish-parasite network of the Upper
Paraná River floodplain, Brazil (Dallas and Cornelius 2015). From a context of
host-parasite interactions viewpoint, Pascual and Dunne (2006) defined robustness
as an estimator of network structure maintenance when facing host species loss
considering the secondary extinctions of parasite species.
Environmental changes, as well as epidemiological outbreaks, may culminate
in the exclusion of host populations, destabilizing and significantly altering the
structure of parasite communities (see Lafferty 1997). In this context, the
intensity of the
removal effect of host species on a host-parasite network depends on the pattern of

distribution of species interactions, whether they are generalists (i.e., interact with
many species) or specialists (i.e., interact with few species), as well as their func-
tional roles in the network (Guimerà and Amaral 2005; Pascual and Dunne 2006;
Estrada 2007).
The influence of local extinctions of host species on the structure of host-
parasite networks may be distinct for communities of ecto- and endoparasites,
mainly due to different host specificities (Bellay et al. 2015a). Ecological networks
have complex structures that may also vary according to the environmental
conditions and their spatial and temporal variations, which govern the diversity of
species and their interactions. In this sense, the introduction of new competitors
and pathogens is a factor that may cause the decline of populations or even local
extinctions (Davis 2003).
9.5 Parasites and Species Introduction
Species introductions into new environments may radically alter native

communities, causing losses in local biodiversity (Michelan et al. 2010), especially
in communities with very specific interactions such as those observed in the
tropics. Invasions of potential hosts may alter the structural patterns of the
networks, promoting interaction ruptures and cascades of parasite extinctions due
to the competition between native and introduced hosts. Among the species of
native parasites, those with a higher degree of specificity regarding the use of
hosts would be the most sensitive to invasions due to a higher probability of
extinction of their hosts. Table 9.1 describes some scenarios and possible changes
in the structure of parasite-host networks as a function of species introduction.
Table 9.1 Possible scenarios of introduction in host-parasite networks

Event Scenario Structure
Introduced host 1. Parasite with high specificity Network structure will
species has being restricted the host present few changes with
parasites that species phylogenetically increasing interactions only
can infect close to the hosts in regions inside the module
native hosts native to the parasite
2. Parasite transmitted trophically Network structure will show
but restricted to hosts few changes with increased
phylogenetically close to interactions between some
hosts in native regions to the modules
parasite but with similar
eating habits
3. Simple-cycle (i.e., Network structure will
monoxenous) generalist present changes with
parasite (ectoparasite) increased connectivity and
increased interactions
between modules with
possible small grouping of
modules
Table 9.1 (continued)

Event Scenario Structure
4. Generalist parasite transmitted Network structure will change
trophically but restricted to with increasing interactions
hosts with similar eating inside modules and between
habits (endoparasite) modules
5. Complex cycle (i.e., Network structure will
heteroxenous) parasite present changes with
transmitted trophically but increased connectivity and
generally to intermediate increased interactions
hosts, which infects by between modules with
penetration into the skin possible small grouping of
modules
6. Strong competitor generalist Number of native parasites
parasite excluding native species may decrease;
parasite species network with greater
connectivity
7. Strong competitor parasite with Number of native parasite
high specificity and restricted species will decrease within
host species phylogenetically the modules
close to the host in native
regions of the parasite
8. Generalist parasite and Reduction in the number of
pathogenic to new hosts in host species and consequently
the region introduced with of native and specialist
possible elimination of parasites; high connectivity
populations of these hosts among species that remain in
the network and low or absent
modularity
The invasive 9. Invasive species of host acts as Increased connectivity and
host species reservoir for the parasite interactions between
acquires acquired in the introduced modules with possible
parasites from region and is expected clustering of small modules
the native increase of the parasite with related hosts
hosts abundance in the natural
environment in other species
of native hosts but restricted in
phylogenetically close hosts
10. Invasive species of hosts acts The increase of the
as a reservoir for the parasite connectivity and increase of
of the introduced region and the interactions between
with the increase of the modules with possible
parasite abundance in the grouping of small modules
natural environment other
species of native hosts are
gradually infected
Non- 11. Parasite is not established unlike Without obvious changes in
establishment host introduced the network; a decrease in
connectivity may occur due to
the greater number of host
species in the habitat
Events involving several introduced parasite species with generalist interactions may contribute to
increased network nestedness
The events and possible consequences of the scenarios described may occur
simultaneously and the effect on the structure and organization of the networks is
intensified according to the amount of species introduced in the environment. It is
necessary to understand the consequences of invasions in order to make manage-
ment decisions considering that even though at times the introduced host does not
present high abundance, its parasites may be increasingly dispersed in the local
host population. It is recommended to carry out further studies in this area in order
to better understand the effects that a introduced parasite species may promote in
different habitats.
9.6 Molecular Approach
The fast advance of molecular techniques has revolutionized the view of species
diversity and evolution (Morand et al. 2015). This development has important
implications to understand host-parasite network structure, since species
identification is a key milestone when studying biological diversity, especially in
the tropics, with so many species yet to be described. Taxonomic studies using
molecular tools have revealed that many parasite taxa considered generalists
actually encompass several cryptic species (Agosta et al. 2010; Braga et al. 2014).
Molecular techniques have also allowed us to construct dated phylogenies with a
satisfactory level of accuracy (da Graça et al. 2016). This expands even more the
limits of understanding how host-parasite networks evolve and have been
maintained throughout time. The advance of molecular tools with the completion
of host and parasite phylogenies will certainly provide a better understanding on
patterns and mechanisms underlying the structures of host-parasite networks.
9.7 Final Considerations
Over the past few years, an increasing number of network studies have greatly
contributed to the understanding of host-parasite interactions in tropical aquatic
ecosystems. However, it is still an incipient amount given the high species
diversity of parasites and hosts, in addition to the little information available to
establish comparisons between tropical and non-tropical networks, which may
generate biased generalizations. The patterns of host-parasite networks may vary
over time and the species development stages may contribute to this variation.
Extinctions or new host-parasite interactions can be expected with the
environmental changes caused by global changes and human activity in the
tropical ecosystems, which might alter the patterns known to the networks. Finally,
further ecological network studies with the use of molecular tools to taxonomic
identification of parasites species should assess these important issues to improve
the understanding on the host-parasite interactions in tropical ecosystems.
References
Chapter 10
Interaction Networks in Tropical Reefs
Mauricio Cantor, Guilherme O. Longo, Luisa Fontoura,

Juan P. Quimbayo, Sergio R. Floeter, and Mariana G. Bender
Abstract Tropical reefs are, figuratively, the underwater counterparts of tropical

rainforests. Both complex, three-dimensional natural systems harbour an impres-
sive diversity of species. The diversity of ecological interactions taking place
among these species is no less striking: their intricate webs add another level of
complexity to these natural systems. In this chapter, we dive into the ecological
networks of tropical reefs to present an overview of some of the negative, positive,
and neutral interactions among inhabitants of rocky and coral reef ecosystems. We
discuss trophic interactions among species as food webs; territorialism and
chasing behaviour as competitive networks of reef fish; cleaning behaviour
illustrating mutualistic networks, and following associations exemplifying
commensalistic networks among fish species. We close the chapter with a
biogeographical perspective of interaction networks in tropical reefs across the
globe to discuss how human activities have been threatening their plentiful life.
10.1 Introduction
Tropical reefs are one of the most productive and biologically diverse ecosystems
(Odum and Odum 1955). Paired with tropical rainforests, reefs are the archetypes
of natural ordered systems. These two three-dimensional systems host species
with a
M. Cantor (*) • L. Fontoura • J.P. Quimbayo • S.R. Floeter

Departamento de Ecologia e Zoologia, Universidade Federal de Santa Catarina,
Florianópolis, Brazil
e-mail: [email protected]; [email protected]; [email protected]; sergio.
[email protected]
G.O. Longo
Departamento de Oceanografia e Limnologia, Universidade Federal do Rio Grande do Norte,
Natal, Brazil
M.G. Bender
Departamento de Ecologia e Evolução, Universidade Federal de Santa Maria,
Santa Maria, Brazil

141
https://doi.org/10.1007/978-3-319-68228-0_10
142 M. Cantor et al.
large variety of life histories in regulated populations within complex ecological

communities (Connell 1978).
Now imagine you could fly and see through these complex, diverse
ecosystems. While this may be difficult in a rainforest, it is not so much when you
dive in a colourful tropical reef. Reefs are oasis of life in the otherwise blue desert
of open oceans. Biogenic reefs are built by corals, sponges, and coralline algae,
and rocky reefs are geological formations that may result from rocks of different
kinds and sizes; moreover, reefs can even be artificial structures such as
shipwrecks. These habitats harbour an impressive collection of species: corals,
gorgonians, crustaceans, worms, echinoderms, seaweeds, not to mention the
most notable creatures, the abundant reef fish. None of these organisms are living
isolated. A key component of this underwater kaleidoscope of colour and forms
often escapes the eyes— ecological interactions. A more careful look into a
tropical reef system reveals the many intricate webs of ecological interactions
submersed in its plentiful live.
As in all ecological communities, dwellers and visitants of tropical reefs
interact with one another (Odum and Odum 1955). These interactions are intra-
and interspecific, and can be negative, positive, or neutral for the interacting
individuals. Interestingly, the effects of these ecological interactions can scale up
from individuals to populations (i.e. affecting fitness), to communities (i.e.
affecting structure), and even to ecosystems (i.e. affecting functioning). Several
types of ecological interactions can be observed in a tropical reef, and fish are a
conspicuous component of many of them. To name a few, reef fish interact
trophically among themselves and with benthic organisms (e.g. invertebrates and
seaweeds); some fish are territorial and aggressive towards other fish (and
sometimes towards divers too); while others may be helpful by removing parasites
from other animals. Such diversity of interaction types reflects the myriad of
ecological processes taking place in the fascinating reef environment (e.g. Moberg
and Folke 1999).
Assessing these interactions to understand the processes they mediate, however,
is not an easy task. Apart from the obvious fact that they occur underwater—indif-
ferent to our curiosity—recording the dynamic interactions among reef species is
laborious, demanding many hours at sea and a combination of methods. Data on
reef species interactions usually come from dedicated scuba or free dive
underwater surveys (e.g. Grutter 2005), remote video recordings (e.g. Longo et al.
2014), and indirect evidence such as stomach contents and literature review (e.g.
Arias- González et al. 2011). As data accumulate, one can employ the formalism
of complex networks to unravel patterns of species interactions occurring
underwater from the local (e.g. within 2 m 2 sampling quadrats) to the global scales
(e.g. across biogeographical provinces).
When ecology meets network theory, species depicted as nodes are connected
by links representing their biological interactions—be these negative, positive, or
neutral (Fig. 10.1). The nature of the biological interaction describes if the
network is directed or undirected (i.e., symmetric or asymmetric interactions
between i and j), binary or weighted (i.e., qualitative or quantitative interactions),
one- or two-mode (i.e., all species can interact, or there are two distinct sets of
interacting species) (Boccaletti et al. 2006). Food webs, for instance, are
traditionally represented by
10 Interaction Networks in Tropical Reefs 143
Fig. 10.1 Tropical reef interaction networks. (a) Direct binary one-mode network of a simplified
food web. Nodes representing functional groups are linked to those they predate upon off the
Virgin Islands (Opitz 1996). (b) Binary two-mode network of reef fish and benthos. Nodes repre-
senting fish genera are linked to the benthic functional groups they consume in Abrolhos, Brazil
(Longo, unpub. data). (c) Indirect weighted one-mode network of agonistic behaviour among reef
fish. Nodes representing genera are linked by the frequency they engage in territorial disputes at
Ascension Island (Fontoura, Bonaldo, Floeter, unpub. data). (d) Weighted two-mode network of
mutualism between cleaner and client reef species. Nodes representing genera (sizes proportional
to abundance, individual/m2) are linked by number of cleaning events in Ascension Island
(Morais et al. 2017). (e) Directed weighted one-mode network of following associations among
reef fish. Nodes representing genera (sizes proportional to abundance) are linked by weighted
directed links proportional to intensity of interactions from follower to nuclear species off
Ascension Island (Morais et al. 2017)
directed one-mode networks (e.g. Yen et al. 2016). Species are connected by
trophic interactions, either weighted arrows indicating the relative rate of energy
transfer (e.g. grams of carbon/day) or interaction strength between taxa, or binary
arrows indicating which resources a given species feeds on (Fig. 10.1a). If the
focus is on a subset of trophic interactions—e.g. between reef fish and benthic
community (Longo et al. 2014)—one can use two-mode networks to depict the
interactions between two trophic levels (Fig. 10.1b). The same approach can be
used to explore cleaning interactions (e.g. Guimarães et al. 2007). Although
essentially trophic, cleaning interactions are by contrast mutualistic (e.g. Côté
2000) between two sets of species—cleaners and clients—which communally
benefit from the interaction (Fig. 10.1d). Agonistic interactions among reef fish
(Robertson 1996), on the other hand, can be represented by undirected one-mode
networks in which species are
linked whenever they engage in disputes for resources, such as territory (Fig.
10.1c). Finally, commensalistic interactions such as following association among
reef fish (e.g. Sazima et al. 2007) can be depicted by directed one-mode networks
indicating which species follows and which is followed (Fig. 10.1e).
Networks give a panoramic snapshot of the liveliness of biological interactions.
These static depictions are very useful to describe structure (e.g. Bascompte et al.
2003), infer function (e.g. Yen et al. 2016), and predict changes (Memmott et al.
2004) in ecological communities. By summarizing a wealth of empirical data into
a network, we can immerse in its tangled structure in the search for emergent
regularities. Some non-random network properties are common across biological
systems. For instance, food webs can display similarities in connectance (i.e.
proportion of realized links), degree distributions (i.e. number of trophic
interactions across species; Dunne et al. 2002), and modularity (i.e. subsets of
highly connected species; Stouffer and Bascompte 2011); whereas nestedness (i.e.
hierarchical organization of interacting species into inclusive subsets) is a common
pattern of two-mode mutualistic networks (e.g. Bascompte et al. 2003). Even
though distinct processes may give rise to these structures, their implications for
the ecological and evolutionary dynamics of the systems may be similar. Nested
networks, for instance, may be robust against random species extinction regardless
of the interaction type (Memmott et al. 2004), potentially reducing competition
among sets of species and increasing the number of coexisting species (Bastolla et
al. 2009). Therefore, the network approach offers us a privileged viewpoint to the
biological processes operating on ecosystems, including tropical reefs.
Here, we explore interactions among tropical reef species to provide an
underwater perspective to ecological networks and equalize the focus on
terrestrial environments given so far in the previous chapters. Our goal is to offer
an overview of some of the negative, positive, and neutral interactions among
inhabitants of reef ecosystems. In the following sections, we briefly dive into
marine food webs to examine trophic interactions among reef species; into
networks of agonistic interactions (e.g. territorialism and chasing) to illustrate
competitive behaviour between reef fish; and into cleaning behaviour to illustrate
mutualism and following associations among fish to exemplify commensalism in
rocky and coral reefs. We return from this dive to close the chapter with a
biogeographical perspective of interaction networks in tropical reefs across the
globe and ponder upon how human activities are threatening them.
10.2 Trophic Networks: Marine Food Webs
Predation is perhaps the most notable of the interactions among species—

including marine ones. This notion and the interest on trophic interactions date
back from Darwin’s first descriptions of communities (Darwin 1859). The term
food web, however, emerged in the early 1900s when Elton (1927) defined a set
of “monopha- gous” consumers as a food chain, and a food web as a group
of chains that
incorporated “polyphagous” consumers (Fig. 10.1a). The first representations of

these trophic interactions consisted in descriptive diagrams with different trophic
levels, indicating a link connecting predator and prey with notes on the natural his-
tory of these organisms. These diagrams provided a general scheme of “who eats
whom” that were not necessarily taxonomically rigorous or comprised all the com-
ponents within a community (Paine 1980). However, only when the consequences
of the trophic interactions started to be revealed in the mid-1900s, Darwin’s inter-
ests on species coexistence started to benefit from our understanding of food webs.
The effects of trophic interactions in community dynamics were unravelled by
Paine’s (1966) seminal experiment removing the ochre starfish (Pisaster
ochraceus) from a rocky shore. This experiment demonstrated that predation by
the ochre starfish could regulate the abundance, diversity, and distribution of
benthic organisms in that rocky shore, enabling more species to coexist. This
study also gave rise to the concept of keystone species (i.e. those whose
interactions may have critical consequences for community dynamics) and had a
remarkable impact on how food webs were studied and interpreted since then. It
was no longer enough to describe who eats whom. Other metrics such as
interaction strength (i.e. a quantitative approach of the link between predator and
prey), connectance, and energy flow through trophic links (Lindeman 1942)
were necessary to move forward in the understanding
of food web dynamics.
Marine food webs are good examples of how dynamic interaction networks can
be. Here, interaction strength, connectance, and the structure and length of food
webs can be affected by multiple factors, for example, wave exposure, tides, cur-
rents, winds, upwelling regime, nutrient availability and, certainly, human interfer-
ence (McClanahan and Branch 2008). Despite being highly dynamic, some
general properties emerge from these webs. Most of the marine food webs
characterized in the literature present a relatively large richness of trophic
interactions, a high level of intermediate omnivore taxa, and an abundance of top
predators comparable to terrestrial ecosystems (Dunne et al. 2004). Chain length
may also vary among marine food webs depending on species richness, diversity,
and complexity of the system (Link 2002; Dunne et al. 2004).
The association of high species diversity in highly connected webs could result
in high structural robustness (Dunne et al. 2004). However, trophic interaction net-
works in reef ecosystems seem to be structured by “few strong and several weak”
interactions with keystone species and groups (e.g. Longo et al. 2014). Central
species can be identified in interaction networks by combining connectance and
interaction strength. A simplified network of trophic interactions between fish
and the benthos (Fig. 10.1b) of the Abrolhos Archipelago, north-eastern Brazil,
shows that herbivorous fish are central species in this reef habitat. Interestingly, in
tropical reef food webs, herbivores often play a central role or are keystone
species, determining the structure and complexity of reef communities by
controlling macroalgae abundance and diversity (e.g. Rasher et al. 2013). The
loss of these strong interactions could have profound impacts in the structure of
ecosystems, such as triggering phase shifts from a coral to an algae-dominated
community, regardless of high species diversity and connectance.
The role of trophic interactions in structuring reef ecosystems goes beyond

interaction strength. More than high abundance of herbivores, a proper
assemblage of these consumers is necessary to establish trophic links with a larger
array of primary producers (Rasher et al. 2013), generating redundancy (i.e.
species that have common trophic links) and complementarity (i.e. species whose
trophic links do not overlap but that contribute to high connectance when
combined). Another emergent property in reef food webs is modularity, which
may relate to redundancy and habitat partitioning. In the North Pacific Ocean, for
instance, sea otters connect an oceanic and a coastal module of a food web (Estes
et al. 2016). The emergence of these structures will also depend on reef
characteristics—e.g. productivity, temperature, and habitat complexity—not to
mention other kinds of interspecific interactions embedded in larger, multi-
interaction ecological networks (e.g. Pocock et al. 2012; Dáttilo et al. 2016). The
ecological consequences of trophic interactions are, therefore, inherently linked
to other ecological interactions.
10.3 Agonistic Networks: Territoriality and

Chasing Among Reef Fish
Living in a tropical reef may be costly. Resources are not always abundant, thus
tropical reef species often engage in agonistic interactions related to resource
partitioning, parental care, and territoriality (Robertson 1996; Bonin et al. 2015).
The outcomes of the agonistic behaviour are generally non-lethal, and subtler than
trophic interactions. Similarly to predation, parasitism, and competition,
agonistic interactions can be asymmetrical: while one individual benefits from
protecting a resource, the others may not only be deprived of it but also suffer
physical aggres- sion. These interactions can also be detrimental for both species
by reducing growth, reproduction, and survival rates (Robertson 1996). Therefore,
agonistic behaviour imposes trade-offs to the participants. Partitioning and/or
competing for limited resources is a daily task for the inhabitant of a tropical reef
—but not an obvious one for the outsider observer.
Direct and indirect competition are often not easy to observe, despite their clear
influence on the structure of ecological networks (e.g. Dáttilo et al. 2014). This
difficulty is particularly true in the heterogeneous, complex underwater reef
ecosystems. Alternatively, agonistic disputes for resources are much more
conspicuous and thus indicate resource-use overlap which essentially underpins
direct or indirect competition among species (Peiman and Robinson 2010). In
tropical reefs, fish species are constantly competing for shelter, food, and sexual
mates (Bonin et al. 2015). Consequently, several species defend territories and are
aggressive (e.g. Forrester 2015). Chasing events among individuals are common
indicators of such disputes for resources; and serve as measurable, and reliable,
proxies of interspecific competition (Robertson 1996).
The resultant network from these agonistic interactions can shed light into the
prevalence of intra and interspecific disputes, and how these interactions can mediate
resource partitioning and competition. Moreover, they give insights on the conse-
quences of agonistic behaviour and territoriality. For instance, engaging in these
behaviours demands considerable energy (potentially influencing individual
fitness) and can alter the density of fish populations in coral reefs, which in turn
can influence community structure (see Robertson 1996; Bonin et al. 2015).
A network of agonistic chasing events depict species linked according to the
frequency they chase each other (Fig. 10.1c). If these agonistic interactions are
organized into a modular network, it could indicate either spatial segregation or
association of taxonomically related species, or even reveal patterns of trait
association (e.g. smaller species would not chase after potential predators).
Modules could also emerge from neutral processes driven by species abundance
(see Vázquez et al. 2007). Interaction strength can also be affected by those
drivers. For example, in the agonistic network of fish in the reefs of Ascension
Island, South Atlantic Ocean, the interaction strength among pairs of species were
not driven by species abundance. Instead, species traits (e.g. the degree of diet
overlap and/or aggressive territorial behaviour) influenced the intensity of
agonism (Morais et al. 2017). Although agonistic behaviour is frequent in tropical
reefs, there is another way to deal with limited resources: cooperation.
10.4 Mutualistic and Commensalistic Networks: Cleaning

and Following Behaviour
Life in a tropical reef can be risky and competitive, especially if you are
someone’s resource or must compete for resources with someone else. Beyond
escaping from and chasing after a competitor, risk and competition sometimes can
be tempered by cooperative efforts. Interactions among reef species can also be
positive. For instance, one species may benefit from other species’ help on
removing parasites or dead tissues (cleaning mutualism) while a smaller fish may
benefit from following a larger one (following commensalism). In these
interactions, the payoffs for the involved parties can be bilateral or neutral—but
are rarely absent.
Cleaning mutualism is one of the most intriguing interactions among reef
species both from an ecological (i.e. what are the consequences of cleaning?) and
evolutionary (i.e. how these interactions were established or selected over time?)
perspective (Floeter et al. 2007). During these associations, a “cleaner” species
removes parasites, necrotic tissue, and mucus from the body surface of “clients”,
contributing to its health (Grutter 2005). Cleaning behaviour has been reported in
ca. 130 species of fish and crustaceans (Côté 2000), which can be facultative (i.e.
species cleaning sporadically or only during juvenile stages) or obligate cleaners
(i.e. species that clean throughout their lives; Grutter 2005). These interactions
often occur at specific sites known as “cleaning stations”, characterized by
prominent structures such as massive corals, sponges, and large rocks (Côté 2000),
forming two-mode networks between cleaners and clients (Fig. 10.1d).
Following association is a neutral interaction observed among reef fish (Sazima

et al. 2007) characterized as a temporary feeding association. Here, opportunistic
species (“followers”) benefit from following bottom-foraging species (“nuclears”)
by consuming the exposed food resources made available from the disturbed sub-
strata (Sazima et al. 2007). The nuclear species does not receive any payoff, and
also does not pay any costs or is harmed by this association. While the nuclear role
is mostly played by fish (less often by octopuses, sea-stars, and turtles) several
species may be followers (Sazima et al. 2007). Following interactions can be
complex, involving large groups of individuals (up to several hundred) from
different species: the “shoaling” associations (Lukoschek and McCormick 2000).
On the other hand, when following involves only up to four individuals, they are
called “attendant associations”. Attendant associations can be divided into four
groups (Ormond 1980): when species aggregates to feed from exposed or leftover
resources gained from the nuclear species (following and scavenging); when it
aggregates to hunt cooperatively with other species with similar feeding habits;
when it hunts by swimming alongside or above a nuclear species using it as a
cover to get access to the prey (hunting by riding); and when the species has an
easier access to the prey by mimicking harmless species (aggressive mimicry).
Following associations can be described by one-mode networks indicating, for
instance, the frequency of this commensalism in a given environment (Fig. 10.1e;
Morais et al. 2017).
Both of these positive interactions are essentially trophic—directly trophic in
the case of cleaning; indirectly in the following (through the exploration of a third
food source). Thus, these interaction networks can give insights on underlying
biological processes taking place in the community, such as niche partitioning. A
nested cleaning interaction network, for instance, indicates that the distribution
of mutualistic interactions is asymmetrical, with some species engaging in many
interactions and others in fewer interactions that constitute subsets of the most
connected species (Guimarães et al. 2007). Therefore, the community contains
species with different resource-use strategies: a core of more generalist cleaner
species (likely obligate mutualists) along with more specialist cleaners (likely
facultative cleaners).
If these positive interactions are specific among some species, the network
would be organized into modules. In this case, a given community would have
cleaner species that interact only with a group of clients; or follower species that
only asso- ciate with specific nuclear species. The modular structure suggests a
high level of specialization among species within these interactions. For instance,
fish species of the genus Elacatinus spp. are mainly obligate cleaners, and prefer
clients associated to the bottom, such as parrotfishes, eels, and groupers.
Alternatively, the network structure can be random, when only a few species
interact, or when most species are generalists. This case is more common in reefs
where the nuclear role is played by few species and the follower role is performed
by several abundant species (e.g. feeding frenzy; Quimbayo et al. 2014). Such
conditions seem to occur in some tropical reefs, particularly the more isolated ones
such as oceanic islands. Therefore, comparing networks depicting different
ecological interactions in tropical reefs around the globe can be profoundly
revealing.
10.5 A Biogeographic Panorama of Tropical Reef

Interaction Networks
The study of ecological networks at the community level provided several insights
on the processes operating at local scales and on how the structure and resilience
are related in ecosystems (Dunne et al. 2004; Bascompte et al. 2005). The
relevance of this approach is unequivocal. However, a biogeographic perspective
could shed light into the mechanisms operating at larger spatiotemporal scales and
potentially on how communities are assembled.
The diversity and composition of biological communities are shaped by
different processes across space and time. Large-scale processes (e.g. extinction,
dispersal) and environmental filters determine which species will successfully
colonize a site in a broader temporal scale (Mittelbach and Schemske 2015). Once
established in a local community, species engage in biotic interactions (e.g.
predation, competition, mutualism) that may affect the spatial distribution of
species at local and regional scales. This balance between evolutionary and
ecological processes, and the relative importance of biotic interactions on
community structure across different spatial scales can be assessed through studies
of species interactions in a macroecological context (Kissling and Schleuning
2015; Cantor et al. 2017).
Food webs, for instance, might conserve structural properties across a biogeo-
graphic scale (Fig. 10.2a). However, in a latitudinal comparison, food webs from
temperate reefs would differ from tropical reefs by exhibiting lower species diver-
sity, larger biomass, and for being more susceptible to changes in nutrient levels
and seasonal dynamics. These webs also differ in their major consumers
(McClanahan and Branch 2008). While fish are more central in tropical food
webs, invertebrates are more important in temperate reefs (e.g. Estes et al. 2011,
2016).
Environmental and ecological gradients are also crucial to the understanding of
large-scale patterns of interacting species (e.g. Trøjelsgaard et al. 2015).
Temperature and isolation are examples of abiotic factors that affect the spatial
distribution of species and, consequently, the diversity of species across spatial
scales (Schemske et al. 2009). Regional diversity patterns have a clear influence
over the topology of reef fish agonistic interactions networks, with a higher
number of nodes and lower values of connectance and centralization according to
the fish diversity gradient (Fig. 10.2b). Diversity may also promote stability and
increase resilience (Dunne et al. 2004). On the other hand, while high regional
diversity implies higher diversity of interacting species, we hypothesize that high
modularity and lower connectance in agonistic networks can suggest and indicate
spatial heterogeneity of species distribution or decreasing functional redundancy
due to spatial segregation of species.
At larger scales, ecological and mutualistic networks may display general struc-
tural patterns such as nestedness and asymmetry of interaction distributions
(Bascompte et al. 2003; Cantor et al. 2017). These resemblances among disparate
networked systems motivated studies testing whether neutrality could explain the
occurrence and strength of species interactions (e.g. Vázquez et al. 2009). In the
Fig. 10.2 Ecological networks from reefs across the globe. Positive (squares), negative (circles),
and trophic (triangles) indicate networks from 12 representative tropical reef communities. In
food webs (a), species (or functional groups when taxonomical level was unresolved) are
connected by trophic interactions (see Yen et al. 2016). In agonistic networks (b), fish species are
connected by territorial disputes (Fontoura, unpub. Data). In cleaning networks (c), cleaners
(coloured) are connected to the client species (Quimbayo, unpub. Data)
case of cleaning mutualism among reef fishes, species abundance is the major
driver of the frequency of species interactions, but it does not explain network
structure alone. Evolutionary signals can influence cleaning interactions (since
specialized cleaners from distinct biogeographic regions are taxonomically closer),
and more mobile species are more likely to interact (Floeter et al. 2007). Binary
cleaning networks from distinct reefs across the globe (Fig. 10.2c) vary in the
number of cleaner and client species. However, the constant presence of some
cleaners with a few links and others with many links suggests an evolutionary
pressure over specialization and a wide array of opportunistic species.
The current theoretical challenges are to assess whether the structure of local
ecological interactions scale up to larger spatial scales and whether these
structures contribute to shape species distributions and diversity patterns at
macroecological scales. A general overview of intrinsic properties of species
associated with
abundance, phylogeny, and their functional role within networks can reveal
evolutionary conservatism of functional roles and whether neutral processes (e.g.
density-dependence) predict the structure of biotic interactions (e.g. Vázquez et al.
2007). Comparing the structure of ecological networks along geographic gradients
can indicate how natural selection and evolutionary processes might vary
according to environmental conditions. This is particularly important given the
increasing human interference in the environment, especially large-scale
disturbances such as climate change and habitat loss.
10.6 The Undesirable Link Between Humans and Reef

Environments
It is no longer a mystery that ecological interactions are key components of life on

Earth. Species are never isolated, but are inserted in multilevel interaction
networks: from individuals to ecosystems, from local to biogeographic scales. The
complex, combined effect of positive, negative, and neutral biotic interactions can
shape the structure of biological communities (Paine 1980) and the dynamics of
the populations within them (e.g. Pires et al. 2011). A current motivation for
understanding the structure and dynamics of biological networks is to predict how
this ecological complexity will respond to the growing anthropogenic impact on
nature, and how one could alleviate its negative consequences (e.g. Pocock et al.
2012).
Tropical reefs suffer from a variety of anthropogenic impacts that operate at
multiple spatial and temporal scales. These include overfishing, over harvesting
of reef organisms, pollution, increasing sediment and nutrient output in the reefs,
uncon- trolled tourism, introduction of allien species and diseases, and climate
change (e.g. Moberg and Folke 1999). These activities have dramatically changed
interaction networks in reefs worldwide. Clear examples come from
destabilization of marine food webs due to overexploitation of resources (e.g.
Jackson et al. 2001; Estes et al. 2011, 2016). In theory, marine food webs should
be relatively robust to overfishing of random species; however, fisheries industries
are selective and often target key species, whose removal can subvert the food
web structure. For instance, the loss of top predators and primary consumers
caused dramatic changes in the biotic and abiotic conditions of ecosystems (i.e.
phase shifts; Jackson et al. 2001; Estes et al. 2011, 2016). In tropical reefs,
overfishing may cascade to decline herbivorous fish, contributing for a shift from
healthy and coral-rich ecosystems to impoverished and algae-dominated systems
(e.g. Bascompte et al. 2005).
Interaction networks can also be affected by the introduction of invasive
species. When the lionfish (Pterois volitans, a Pacific species introduced in the
Caribbean by the mid-1980s) was added to food web models of Caribbean reefs it
figured above the native top predators which are sharks, rays, and groupers (Arias-
González et al. 2011). These novel trophic interactions could lead to higher
mortality but could also modulate competition. Climate change could have similar
effects on food webs by facilitating tropical species to expand their ranges and
enter subtropical food webs
(tropicalization; Vergés et al. 2014). The negative impact of humans on marine

food webs (consequently, the functioning and structure of marine ecosystems) is
clear; there is no reason to doubt that networks of other biotic interactions are
disturbed too.
There are few clearer ways of capturing the urgency of understanding biotic
interactions than the famous quote by Janzen (1974) on tropical rainforests: “What
escapes the eyes (...) is a much more insidious kind of extinction: the extinction of
ecological interactions”. It goes without saying that the warning also fits perfectly
to tropical reefs. May we keep these underwater worlds complex, productive, and
diverse in shapes, colours, and functions. Only then the coming generations can
dive in their plentiful life to discover the importance of preserving their fragile
interaction networks.
References
Chapter 11
Ecological Networks in Changing Tropics
Eduardo Freitas Moreira, Patrícia Alves Ferreira, Luciano Elsinor Lopes,

Raimunda Gomes Silva Soares, and Danilo Boscolo
Abstract Earth has an extremely dynamic surface which changes naturally across
time. In the last century, however, vegetation cover underwent severe
modifications due to human demands for natural resources and food production.
These changes are deeply modifying the spatial distribution of native
environments, which exist today mostly in small patches embedded in human
dominated landscapes. This is even harsher in the tropics, where agricultural
expansion is more intense. Ecologically, this means that native species have to
cope with a heterogeneous set of new environments in which they did not evolve,
bringing difficulties for the movement of foraging individuals. This can impair the
encounters needed to establish biological interactions among individuals and
different species. In this chapter, we explore how landscape changes can lead to
variations in ecological networks structure and its consequences for biological and
ecosystem services conservation. Although there is a general lack of complete and
extensive studies regarding the effects of landscape changes on tropical ecological
networks, there is growing evidence that, given a certain native vegetation cover,
landscape heterogeneity may favor bigger and more complex networks across
scales or ecological hierarchical levels. The relationship between landscape
heterogeneity and the
E.F. Moreira
Departamento de Zoologia, Universidade Federal da Bahia, Salvador, Brazil
P.A. Ferreira
Departamento de Ciências Ambientais, Universidade Federal de São Carlos,
São Carlos, Brazil
Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto, Departamento de Biologia,
Universidade de São Paulo, Ribeirão Preto, Brazil
L.E. Lopes • R.G.S. Soares
Departamento de Ciências Ambientais, Universidade Federal de São Carlos,
São Carlos, Brazil
e-mail: [email protected]; [email protected]
D. Boscolo (*)
Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto, Departamento de Biologia,
Universidade de São Paulo, Ribeirão Preto, Brazil

155
https://doi.org/10.1007/978-3-319-68228-0_11
156 E.F. Moreira et al.
structure of ecological networks is however still an open field with many

challenges and opportunities and a huge potential for application for conservation
and environmental management.
11.1 Introduction
We live in an extremely dynamic world. The world surface changes naturally as a

condition of the planet. However, in the last 70 years the planet’s surface
underwent large sweeping changes in a short period of time. The growing demand
for food production, driven by the human population boom, agricultural
techniques developed during the green revolution, as well as the cultural changes
associated with the production of wealth and economy globalization, have led to
the conversion of large areas of native vegetation into agricultural lands (Foley et
al. 2005). This process is accompanied by a global biodiversity crisis which results
in accelerated species extinction rates. In addition, there is no evidence of
deceleration and the indicators of pressure over biodiversity increased in the last
decade (Ceballos et al. 2015). For example, Hansen et al. (2013) showed intense
transformations worldwide, with the loss of 2.3 million square kilometers of forest
cover in only 12 year and the gain of only 0.8 million square kilometers of new
forests at a global scale. This scenario becomes even more dramatic if we consider
that in the tropical regions, where most terrestrial biodiversity is found, these
transformations were even more intense. In the tropics, native vegetation has been
heavily replaced by agricultural production areas dominated by extensive pastures
or monocultures with intensive management (Ribeiro et al. 2009).
Changes in landscape structure such as the loss of native vegetation and reduc-
tions of environmental heterogeneity are among the main drivers of biodiversity
loss (Stein et al. 2014) causing, for instance, along with exotic plants invasion,
severe negative effects on flower visitors’ functional diversity (Grass et al. 2013).
Such changes in community structure may also be associated with changes at a
more systemic level (Fort et al. 2016). Biodiversity is more than a collection of
species. Biotic interactions involve many partners, are complex, dynamic and
play an important role in the organization and persistence of biodiversity
(Bascompte et al. 2003). These changes in community structure can have
important consequences for interspecific interactions and thus for ecological
processes depending on them (Valiente-Banuet et al. 2014). For example, Moreira
et al. (2015) observed that pollinators’ diversity was positively related to
landscape heterogeneity and that over- simplifying the landscape had negative
effects on the plant-pollinator networks structure, reducing its complexity (Fig.
11.1).
Much has been discussed about the principles behind the organization of eco-
logical networks structure, where trait matching and the abundance distribution in
the communities, are the major contributors to network formation (Blüthgen
2010). However, these two factors do not contribute in the same proportion since
the majority of the variation in networks structure can be explained by the
variation
11 Ecological Networks in Changing Tropics 157
Fig. 11.1 Adaptation from the scheme summarizing the results from Moreira et al. (2015), with 4
of 27 studied landscapes representing the positive relationship between the plant-pollinator net-
works complexity (A to D), the vegetation heterogeneity (left to right) and landscape
heterogeneity (top to bottom), including the number and proportion of the different land cover
types (grey levels) as well as the complexity of its arrangement
in the community structure, such as species richness and relative abundance

(Vázquez et al. 2009). The importance of those relationships is that network struc-
ture may be related with the ecological systems’ emerging properties such as
robustness and resilience, which are key factors to the maintenance of ecological
processes and biodiversity conservation (Tylianakis et al. 2010). This can be well
illustrated using pollination and predator-prey population dynamics, both good
examples of fundamental processes for biodiversity maintenance in terrestrial eco-
systems (Blüthgen 2010).
In this chapter, we will explore the most promising propositions for the
relationship between landscape structure, biodiversity, and interaction networks,
highlighting the latest evidences as well as the gaps and the implications of
landscape management applied to biological conservation in the tropics.
11.2 Landscape Ecology
Landscape, viewed as “a heterogeneous mosaic composed by interactive

landscape units, where heterogeneity exists for at least one parameter, one specific
observer and at a particular scale” (Metzger 2001), allows for a wide and useful
approach to understand environmental changes for many reasons. Most
importantly, this approach turns the landscape into a system composed by
elements that interact with each other. Those elements are classified into
landscape units according to each type of landscape component, i.e., types of land
cover, vegetation classes, or environments (Christian 1958). The criteria adopted
in the definition of these units are fundamental since they have great influence on
the meaning of landscape’s characterization. To be ecologically relevant, these
criteria must be biologically based and explicitly explained according to the
observer in question (Forman 1995).
Landscape ecology is, thus, not an organizational level or a scale of analysis. It
is, instead, an approach to study how spatial heterogeneities are generated and can
affect human society and biological processes across all organizational levels. The
landscape heterogeneity can, then, be decomposed in two components: the
configurational and compositional heterogeneities (Fig. 11.2). Composition
includes the characteristics associated with the types of environments included in
the landscape and their respective quantities (Forman 1995). Configuration deals
with the spatial arrangement of landscape units.
Another consequence of that landscape definition is the impossibility of a func-
tional characterization of its structure without a previous definition of the observer
(Moreira et al. 2017). This observing agent can be interpreted in two ways, more
literally as referring to organisms that directly observe, interpret, and react to the
landscapes, or alternatively in a more metaphoric sense where the term observer
may be best understood as the biological process of interest that may respond to
the landscape heterogeneity. In the last sense, the observer can refer to a whole
range of phenomena such as the reaction of individuals of a species, a certain
behavior of these individuals, the populational dynamics, or the network of
interspecific interactions between species (Boscolo et al. 2016).
For example, consider a frugivorous bird that has a seasonal migration. In each
season, these birds are doing different things, such as mating, nesting, and migrat-
ing. The availability of the conditions and resources necessary for each of these
activities varies in space/time at different scales. It is the same to say that a given
individual experience the spatial variability in different ways during its lifetime
depending on the specific activities that it does. Therefore, approaches considering
multiple scales can be relevant, depending on the question being asked (Kotliar
and Wiens 1990). If the question is relative to a specific behavior at an individual
level, such as nest building, it is possible that only a single aspect of the landscape
heterogeneity is relevant. However, if interest is on the reproductive success, mul-
tiple factors become relevant which may include the interaction among several
individuals or species, such as the availability of nesting material, food resources,
mates, predation and parasitism rate, and so on (Chittka and Thomson 2004;
Turner
Fig. 11.2 Adapted from Fahrig et al. (2011), represents the two axes of landscape heterogeneity:
compositional and configurational heterogeneity. The different color types represent different
land cover types, the landscape units, reflecting the heterogeneity of some environmental
characteristic relevant to the observer in question. The compositional heterogeneity increases
with the number and evenness of landscape units, what must reflect levels of variation in the
relevant characteristics. The configurational heterogeneity increases with the level of intricacy
between landscape units and is proportional to the increase in the contact surface between them
and Gardner 2015). The meaning of spatial heterogeneity then changes completely
from one question to another, while we may still be talking about the behavior of
one individual.
Landscape perception will also be related to other levels of organization such as
population dynamics, interspecific interactions, and ecological networks. The
reproductive success of the individuals are determinant factors for the
populational dynamics, which is determinant to the interspecific interactions that
in turn pro- duces the systems’ networks. This framework imposes a challenge. In
nature, all biological processes that respond to spatial variation are happening
simultaneously and, in general, are interdependent from each other. Therefore,
multiple levels of influence must be considered with regard to the relationship
between landscape heterogeneity and interaction networks, attempting to the
relevant spatial heterogeneity at each level as well as the scale of measurement
and the hierarchical nature of the causal relationships involved (Turner and
Gardner 2015).
11.3 Effects of Landscape Structure on Interaction Networks
In theory, any complex system can be represented and analyzed as interaction net-
works. In the last few decades, the developments of graph and system theories
were incorporated in ecology to help understand what are the consequences of the
community structure on the ecological interactions and emergent properties of
these systems (Bascompte et al. 2003; Tylianakis et al. 2010). In this section, we
will explore how landscape structure changes can lead to changes in ecological
networks. We will also discuss what those changes in ecological networks may
represent for ecosystem processes and biodiversity conservation. However, our
intention here is not to exhaust all possible intersections between landscape
ecology and network theory. We will focus mainly on mutualistic networks (e.g.
plant-pollinator and plant-disperser interactions networks) given its relevance in
the context presented before. In addition, interaction networks have many
features, and there are multiple ways to mathematically describe them. You can
find a full description of the network characteristics and their descriptors in Chaps.
3 and 13 of this book. Nonetheless, not all features are equally relevant regarding
network function and stability, especially in relation to a landscape approach
(Bascompte et al. 2003). Therefore, we will discuss mainly networks’
characteristics whose relationship with landscape and systems’
stability/productivity was previously proposed and tested (Tylianakis et al. 2010).
The study of mutualistic interaction networks provides an interesting
assessment of landscape change impacts on biodiversity and the functioning of
ecological processes, providing guidance to the conservation of both (Viana et al.
2012; Ferreira et al. 2013; Falcão et al. 2015; Moreira et al. 2015). However, there
is still theoretical controversy and few empirical data regarding the effects of the
loss of natural environments on network characteristics such as nestedness (how
much of the interactions with specialists species are a subset of the interactions
with generalists species) and complementary specialization—H2′ (the extent to
which specialist species interact with other specialist species) (Soares et al. 2017).
As suggested by Aizen et al. (2012), in the tropics, interactions between
specialized species can be more sensitive to forest loss, whereas asymmetric
interactions (specialist-generalist) or between generalized partners could be more
resistant to landscape changes. This variation in response can promote shifts in
interaction network characteristics, potentially affecting their robustness, leading
to secondary extinctions, loss of interactions and, consequently, disruption of
ecosystem functions (Soares et al. 2017).
To understand ecological systems as interactions networks and the effects of
changes in landscape structure, it should be clear that we are dealing with a spa-
tially/temporally oriented hierarchical complex system (Turner and Gardner 2015;
Ferreira et al. 2015; Moreira et al. 2015). Independently of the scales of observa-
tions, landscape structure is one of the most important factors that will influence
all levels. For example, at the lower levels landscape structure influences
individuals’ behavior (Cranmer et al. 2012), at intermediate levels it influences
population dynamics (Pulliam et al. 1992), and at the higher levels it influences
communities’ biodiversity (Slancarova et al. 2014).
The characterization of the network structure includes aspects of the interacting

partners and of the interactions themselves (Bascompte et al. 2003). First, to figure
out the relationship between the landscape and network structure we should estab-
lish the mechanisms behind that relationship. Networks are representations of sys-
tems, and the landscape can affect the components of these systems, and
consequently the system itself. The landscape spatial structure can directly affect
the movement of organisms. To be possible for organisms to find each other and
interact, individuals must move through space. For instance, regarding plant
pollination by animals, because pollinators move through a landscape to reach
flowering plants they have to face different environments and conditions, which
can change their routes, directly interfering on pollen transfer probabilities. In this
sense, landscape structural changes can modify the effectiveness of pollination
process of native vegetation and agricultural fields merely by hindering
pollinators’ access and movement among flowers (Moreira et al. 2015).
Another aspect of the effects of landscape changes on interactions networks
may be found in the effects of landscape heterogeneity over the organisms’
foraging behavior (Cranmer et al. 2012). Animals, for example, tend to optimize
foraging often preferring the nearest energy source because of its lower cost
compared with farther ones. However, when the energy gain is significantly
higher, individuals may assume higher costs, since the energy expended traveling
a greater distance is well compensated (Lihoreau et al. 2011). In the more
heterogeneous landscapes, the chances of having nearby alternative resources with
high value to an organism is higher, and it may increase the chances of that
organism to travel shorter distances from one landscape unit to another while
increasing fitness and reducing survival costs. In these cases, the organisms’
movement between landscape units with complementary resources is encouraged
by the low traveling cost since the distances necessary to cross between units tend
to be lower. Therefore, the interactions between species that depend upon
organisms’ movements across landscape units have higher probability to occur in
more heterogeneous landscapes than in homogeneous ones, where resources may
be spatially disaggregated (Cranmer et al. 2012). This can affect network size and
number of interactions observed in a given site of the landscape as well as have
effects on other important network characteristics since species will not meet their
network counterparts (Vázquez et al. 2009).
Another example are the adaptations of many species to the variation in
competitive pressure. Those mechanisms are activated depending on the context
in which the species are embedded. They can be triggered by competitors’ density
or resource scarcity and include behavioral changes, such as the broadening or
contraction of the diet breadth. Some animals tend to include more food items
when under the influence of increased competitive pressure, expanding their diet.
In more heterogeneous landscapes where the flux of organisms may be facilitated,
more competitive pressure is expected in places with high value resources as
more individuals of distinct species are drone to it from further places (Moreira et
al 2015). This can promote the diet expansion of the more plastic organisms that
reside nearby (Fontaine et al. 2008). Such changes can affect network structure
since the more plastic species will exhibit more interactions, increasing
connectance, network nestedness, and changing
the interaction strength distribution, as well as the overall network specialization

(Vázquez et al. 2009).
Invasive exotic species, for instance, are expected to present such behavioral
plasticity, and are likely to compete with and/or facilitate the maintenance of
native species and their interactions (Bartomeus et al. 2008). Taking pollination as
an example, the most reported effect is the usurpation of interactions by the exotic
species, which centralizes the interactions increasing the asymmetry of the
network and decreasing the interactions strength between the native species. This
is accompanied by the increase in abundance of species that benefits from the
invasion (Soares et al. 2017). On the other hand, there are reports of no effects of
invasive species on network metrics such as nestedness and connectance, what
may indicate some network robustness, with a reorganization of the interactions,
buffering the losses and behavioral changes caused by the introduction of super
generalists exotic species (Padrón et al. 2009; Vilà et al. 2009; Falcão et al. 2017).
Theoretical models suggest that the structures of mutualistic networks are in
general resilient to random extinctions, but the extinction of well-connected gener-
alists can result in quick network collapse due to interactions and species loss
(Kaiser-Bunbury et al. 2010). In tropical environments, where harsh landscape
structural changes occur, the loss of key generalists is probable (Aizen et al. 2012).
However, empirical data on the response of ecological network metrics to
landscape changes is scarce (Soares et al. 2017). The available data indicates that
after an environmental disturbance the network tends to reorganize towards a more
stable form, allowing for the coexistence of species and maintenance of
interactions (Nielsen and Totland 2014). However, in a recent review Soares et al.
(2017) showed that environmental degradation influenced network metrics but
degree, nestedness, and connectance (based on unweighted links) did not have a
simple, linear, or uni- directional pattern of response to changes in environmental
quality. Despite limited data, asymmetry and reciprocal specialization index (d′
and H2′ respectively) showed the most consistent responses to environmental
quality change.
In the tropics, landscape homogenization due to native vegetation loss can lead
to the reduction in nestedness and specialization mostly because of specialist spe-
cies loss (Moreira et al. 2015, Ferreira unpublished dataset). Another important
highlight is that the role of the species changes, ranging between generalists and
specialists under different conditions (Soares et al. 2017). There is a higher level
of specialization in better preserved landscapes, and the composition of the
network can vary even when its basic structure does not change (Nielsen and
Totland 2014). This shows how interactions networks have a particular importance
for ecological conservation. More resilient species may survive in a degraded
landscape; however, its role in the networks may change in ways that compromise
the ecological process underlined by those interactions (Aizen et al. 2012). For
example, as a specialist pollinator changes its role in the network, becoming more
generalist, may put pressure on other species, causing species loss and
disruptions of ecological processes (Nielsen and Totland 2014).
In the end, landscape heterogeneity may favor bigger, more connected, and
more nested networks across scales or hierarchy levels (Moreira et al. 2015). This
happens because the number, strength, and distribution of interactions are closely
related with the community structure (Vázquez et al. 2009). The landscape hetero-
geneity can also affect the density of individuals through two basic mechanisms.
First, it can increase recruitment of species that can benefit from the
supplementary or complementary resources available in different landscape units.
For example, multivoltine species may have more reproductive cycles during the
year in more heterogeneous landscapes when the different landscape units offer
complementary or supplementary resources in different seasons, also reducing
intraspecific competition (Burkle and Knight 2012). The second mechanism is
associated with the presence of attractor sites in the landscapes, increasing beta
diversity, for example (Veech and Crist 2007). More homogeneous landscapes may
thus promote the concentration of individuals in few edges between the
landscape units (Ferreira et al. 2015), whereas more heterogeneous landscapes
tend to have populations more evenly distributed promoting the dilution of
individuals (Moreira et al. 2015). The balance of these two mechanisms may
determine the concentration of individuals in a given point in the landscape and in
turn the probability of their interactions (Vázquez et al. 2009). On a broader level,
landscape heterogeneity can promote higher beta and gamma diversity, which can
function as source of diversity to the proximal level. The reduction of landscape
heterogeneity generally involves the substitution of natural vegetation by
agriculture or other land cover. In this context, species with more behavioral
plasticity may be less affected or even favored, exhibiting increased abundance
in landscapes with low levels of native vegetation cover, while species with less
behavioral plasticity may be locally extinct (Ferreira et al. 2015). This process
follows the regime-shift model that predicts a reduction in abundance and richness
of specialized species and an increase of generalized ones at different landscape
scales (Pardini et al. 2010). These changes tend to decrease the conservation state
of networks in degraded landscapes (Soares et al. 2017).
At a proximal scale, more heterogeneous landscapes can promote the mainte-
nance of higher species richness than the more homogeneous ones (Fahrig et al.
2011; Moreira et al. 2015). This happens because beyond the species that are
associated with each environment and the common ones, heterogeneous
landscapes also may sustain species that can survive only in regions where
multiple environments are present (Fig. 11.3). Such situations are expected in
cases where the species reproduce in one environment and forage in others or
where vital resources are sea- sonally available in each landscape unit along the
year. Regarding the configurational heterogeneity, landscapes with more
heterogeneous configurations will also favor the maintenance of the species that
depend or benefit from the presence of multiple environments, since it can lower
the costs of movement between landscape units, as explained above. Richer
communities will then tend to have networks that are larger more nested and with
more specialist interactions (Bascompte et al. 2003; Moreira et al. 2015).
Fig. 11.3 Adapted from Fahrig et al. (2011), represents the effect of complementation between
landscape units on biodiversity and network structure. Species diversity increases with the
number of landscape units because they maintain the species associated with each landscape unit
plus the species that require the multiple landscape units to survive. This process also affects the
probability of interactions and the network complexity as observed by Moreira et al. (2015)
11.4 Interactions Networks Conservation,

Management, and Future Perspectives
Populations’ densities and interactions are highly variable in space and time
(Petanidou et al. 2008). This natural intrinsic variation has to be considered in
environmental management based on single species or single interaction.
However, the structure of communities and interaction networks tend to be more
stable over time than its constituents (Blüthgen et al. 2016). For that reason, the
network structure could be the focus of management goals. However, not every
stable state of a system represents a desirable state from the point of view of
ecological process management. For example, reducing the diversity of plant
species in a given environment may lead to a more stable state, which may not
necessarily mean improvements in the ecological processes of that environment
(Morales and Aizen 2002). Environmental degradation may lead the system to
stable or resilient states, mainly by reducing the diversity of species and by
changes in species’ functions in these networks (Soares et al. 2017). Such stable
states are undesirable from any biodiversity conservation perspective and may
constitute a challenge to ecological restoration and management. Therefore,
initiatives for conservation and management of ecological processes should aim
not only on stability but also the complexity of ecological systems.
As interaction network structure reflects the general pattern and not the
intrinsic spatial/temporal variation of the system, it is important to discuss which
should be
the conservation target to networks in landscape management initiatives and how

land use management can help us approaching the desirable state. Some of the
mechanisms and tendencies involved in these relationships were discussed above.
However, those need to be viewed with caution and carefully criticized. For
instance, mutualistic networks tend to be nested. Additionally, network nestedness
is expected to increase the resilience of plant-pollinator interactions (Bascompte et
al. 2003). Should mutualistic network nestedness maintenance be a conservation
goal? Should management targets be likely “to increase nestedness”? This
questions have been raised and not answered yet, either theoretically or
empirically (see Soares et al. 2017; Tylianakis et al. 2010).
Empirical evidence has shown that a higher nestedness in degraded landscapes
can also indicate an increase in the resilience of a poorer undesirable state of the
network (Soares et al. 2017). Even very simplified and degraded networks can
maintain its nestedness and partially its functionality, without conserving the inter-
actions richness and network complexity. Therefore, we should not set increases in
nestedness as a management target. On the other hand, a management target such
as the conservation of all interactions is difficult to pursue by several reasons,
including the spatiotemporal dynamics of ecological systems.
Only the combined analysis of a set of metrics, including the network level and
the underlying mechanisms observed at lower hierarchical levels will be able to
indicate the conservation status of an interaction network and its relation to the
landscape structure. Despite still incipient, there are some candidate metrics for
that integrated analysis such as network size, interaction asymmetry, interaction
specialization or selectivity (H2′), plus additional information about changes in
species compositions and their roles in the interactions (Soares et al. 2017).
Additionally, the comparative study of networks in different environments,
associated with functional diversity and the measurement of ecological processes
in order to investigate if there are network characteristics capable of reflecting the
specificities of each ecosystem. Information of this nature may provide clues to the
relevance of specific network structural patterns for the ecosystem functions in
different environments and how they can be managed.
The network analysis also has great potential of application in monitoring of
ecological restorations since it allows to evaluate the restoration process covering
not only the development of the vegetal species but its interactions and increase of
the complexity of the community. In this sense, the networks also contribute to the
selection of the characteristics of the species that would be most important in the
process of community formation, thus enhancing the restoration action. Finally,
network analysis can serve as a parameter to evaluate the success of the restoration
(Forup et al. 2008).
In fact, we still know too little about the consequences of landscape changes on
the network structure to link diversity, stability, and function because of the lack
of empirical studies. In addition, the magnitude of spatial-temporal variations in
tropical landscapes is greater than in other regions because of the high diversity
and species turnover (Nielsen and Totland 2014). Therefore, such studies are
needed to sepa- rate the effects of natural variations from those resulting from
anthropic impacts.
In general, the number of empirical studies on networks is scarce and in the tropics
it is even smaller (Ferreira et al. 2013). Most of these studies do not relate the net-
work structure to environmental gradients or provide empirical evidence of its
impact on the ecological functions to which they are associated. Therefore, most
of the conclusions about the impacts of networks’ structure and dynamics over
ecosystems functions are based on conceptual and mathematical models, and/or
indirect relationships. Empirical studies could enable us to identify which network
interactions are most affected by anthropic impacts, as well as to analyze the
ecological significance of the behavior of these interactions at the community
level. However, it is important to bear in mind that “absence of evidence is not
evidence of absence.” You can find a current perspective about the empirical gaps
in this matter in the review by Soares et al. (2017).
Another interesting topic is the relationship between the dynamic aspects of
both landscapes and ecological networks. There are some propositions about the
consequences of the landscape dynamics for biological communities, biodiversity
stability, and conservation such as the spatial insurance hypothesis, considering
the interactions’ turnover (Loreau et al. 2003). However, such relationships were
not empirically evaluated yet. The same is true to the dynamics of interaction
networks and its relation to function and stability that were usually studied
through mathematical models (Vázquez et al. 2009). There is an interesting path
emerging from the network field that can help coping with these problems, the
multilayer approach to model ecological networks (Pilosof et al. 2017). This
approach allows us to consider both temporal and spatial dynamics of interaction
networks and can be associated with landscape hierarchical models. In sum, we
can safely assert that the relationship between landscape heterogeneity and the
structure of ecological networks is an open field with many challenges and
opportunities and an enormous potential of application for conservation and
environmental management in the tropics.
Acknowledgements We thank Rafaela Lorena da Silva Santos for all the support in challenging
times and the help with the references and spelling check.
References
Chapter 12
The Future of Ecological Networks
in the Tropics
Luciano Cagnolo
Abstract Ecological networks are one of the best approaches to describe interac-
tive communities of species. Accordingly, the development of network studies in
the tropics is imperative given the high rates of habitat loss and transformation. To
achieve this goal, we face the challenge of dealing with extreme complexity but
lacking complete taxonomic and natural history information. In this chapter, I ana-
lyze the trajectory of network studies in the tropics over time and describe some
promising avenues for the study of ecological networks in the next years. I built
keyword co-occurrence networks of network studies in the tropics for four periods
from 1970 to the present. The earliest network studies were concentrated on food
webs; in the following decades, network studies rose dramatically and diversified,
generating topic modules about different interaction types. The last period (2010–
2016) reflects a mix of different research areas, with food web studies being less
important and much more connected with other topics such as frugivory and myr-
mecophily. One of the major challenges of network research in the tropics is to
increase the level of network complexity. Here, I propose two ways: merging
different interaction types into single networks and disaggregating data into their
spatial, temporal, and individual-level layers. The multilayer approach requires
new conceptual and methodological frameworks that are starting to be
formalized. One of these tools is barcode sequencing directly from DNA extracted
from consumers, which provide strong physical evidence for the host association
and facilitates phylogenetic analysis.
12.1 Introduction
The megadiversity of the humid tropics has always amazed naturalists, as could be
noticed in the annotations and drawings of the nineteenth century (Bates 1864).
The fascinating abundance and diversity of forms and colors of tropical species
may
L. Cagnolo (*)
Instituto Multidisciplinario de Biología Vegetal (CONICET), Universidad Nacional de
Córdoba, Córdoba, Argentina

171
https://doi.org/10.1007/978-3-319-68228-0_12
172 L. Cagnolo
have served as inspiration for the cornerstone work of Darwin and Wallace in
1858. Since those days to the present, important advances have been made in
cataloging tropical biodiversity; however, most biologists agree that we still lack a
nearly complete description of tropical diversity, particularly regarding small, non-
conspicuous insects, fungi, and bacteria (Erwin 1991). Although this is
problematic— but affordable—it is undoubtedly an urgent task considering the
unprecedented rates of habitat loss and modification, with the consequent loss of
associated species (Pimm and Raven 2000).
The study of ecological networks increased in the last decade since researchers
have appreciated them as the most informative way to summarize and describe
biological communities. From the earliest food web studies to the complex
network approach in our days, ecological network science has evolved both in
theory and tools as a consequence of merging ideas coming from physics,
sociology, mathe- matics, and biology (Poisot et al. 2016). Network studies
applied to biological communities succeed in understanding the robustness and
fragility of ecosystems (Solé and Montoya 2006) and identifying key species and
interaction paths (Aizen et al. 2012), which are the foundation of conservation
management. In the tropics, the development of network studies is an urgent task
given the high rates of habitat loss and transformation (Barlow et al. 2016).
Nevertheless, developing network studies in the tropics requires facing the
challenge of dealing with extreme complexity but still lacking complete taxonomic
and natural history information.
In this chapter, I analyze the trajectory of tropical network studies over time,
focusing on the comparison of the type of interactions studied and the concepts
and tools applied. Finally, I describe some promising avenues for the study of
ecological networks in the next years, both in their theoretical and methodological
aspects.
12.2 The Evolution of Network Studies in the Tropics
The study of ecology through networks has a long history that can be traced back
to the 1800s, as cited by Dunne (2006). In this section, I provide a revision of
papers presenting network studies in the tropics by searching in Scopus and
Google Scholar using the keywords “network + tropics OR tropical” and “food
web + tropics OR tropical.” With this information, I built keyword networks for
each of four periods (1970–1989, 1990–1999, 2000–2009, and 2010–2016), by
establishing a link between words when they co-occurred in the title, abstract, or
materials and methods section. The idea behind this analysis was to evaluate how
concepts group together and to determine which ones are central nodes that
connect different areas of study.
The keywords search focused on concepts associated with hypotheses (e.g.,
“equilibrium,” “bottom-up control”), countries where the studies were conducted,
habitat type (e.g., “forest,” “freshwater”), statistics used to describe the networks
(e.g., “connectance,” “nestedness”), interaction type (e.g., “competition,” “pollina-
tion”), sign of interactions (e.g., “mutualism,” “antagonism”), and organisms
involved (e.g., “fishes,” “insects”). With this information, I constructed weighted
one-mode networks, in which each tie in the edge list was valued by a positive
num-
12 The Future of Ecological Networks in the Tropics 173
Fig. 12.1 Number of network studies of different interaction types in tropical ecosystems for four
periods of time
ber that indicates its frequency of co-occurrence. These networks were examined
to evaluate how words are grouped into highly connected groups (here referred to
as “modules”) and therefore identify the concepts, tools, or organisms that serve as
connectors between different areas of research (modules). By doing this, I expect
to identify which concepts helped to integrate and generalize the science of
ecology.
The number and diversity of interaction types in network studies conducted in
the tropics has risen dramatically since the 1970s (Fig. 12.1). Accordingly, the
concepts and tools applied in their study increased and diversified as well (Fig.
12.2). The 1970–1989 period presented only four studies in 20 years, with all of
them being descriptions of antagonistic interactions, such as predation and
competition, and three out of four being from aquatic habitats. The keyword
network was sharply divided into two blocks, connected by the word
“antagonism” (Fig. 12.2).
During the1990s, the number of studies more than doubled those of the
previous two decades. As in the previous period, antagonisms still prevailed but
ecologists started to focus on interactions other than predation, such as herbivory
and parasit- oidism (Fig. 12.1). In the 1990s, ecological studies were influenced by
research exploring regularities in food webs, which had been performed in
previous years (Briand and Cohen 1984; Pimm and Kitching 1987). Although food
web studies in aquatic systems were still important in the 1990s, there seemed to
be more interest in what happens with insects and plants in tropical forests.
Researchers incorporated new tools, such as the use of stable isotopes, to establish
feeding links, statistics such as modularity to describe global aspects of network
structure, and concepts of
174 L. Cagnolo
Fig. 12.2 Keyword networks of network studies in tropical ecosystems for four periods of time.
Font size of keywords reflects their frequency of appearance, and colors reflect the approximate
position of cohesive groups of keywords (modules)
scale-dependency of network properties; studies were also conducted in new

(for network scientists) countries such as Costa Rica and New Guinea. The
keyword network of this period was composed of four modules, with the largest
ones being that of food web studies in aquatic systems. Consequently, the most
frequent keywords were those related to food web studies, such as “predation,”
“fishes,” and “antagonism” (Fig. 12.2).
In the following decade (2000–2009), the total number of studies rose, with the
peculiarity of an increase in the relative importance of non-food web studies,
which were less than half of the total (Fig. 12.1). Keystone papers evaluating
regularities
in the patterns of mutualistic interactions (Jordano 1987) and the first analyses of
pollination networks with food web tools and concepts (Memmott 1999) generated
new enthusiasm. However, the keyword network still reflected the dominance of
food web studies, showing a greater richness of concepts than in the previous
period and the emergence of interest in new interaction types. Hence, the food web
words formed a large module on the left side of the network graph, whereas the
remaining keywords were placed in three modules, each one representing different
lines of research associated with different interactions. The most frequent
keywords were those related to food webs, such as “trophic position,” “fresh
water,” and “stable isotopes” (Fig. 12.2).
The last period analyzed comprises only seven years but is sufficient to demon-
strate a dramatic increase in number and diversity of network studies in the
tropics. The decrease in the interest in food webs is remarkably opposite to the
increase of the remaining interaction types, particularly in myrmecophily (Fig.
12.1). The keyword network reflects these changes through a combination of
different areas of research; although the food web module can still be recognized,
it is relatively small and much more connected with other modules. Interestingly,
there is a change in the research topics related to the most frequent keywords, with
“specialization,” “plants,” and “pollination” being some of the most prominent
ones (Fig. 12.2).
The evolution of the study of ecological networks in the tropics is not different
from that in other latitudes (e.g., Dunne 2006), and the present of network studies
in the tropics shows an amazing diversity of topics and, particularly, of interaction
types. This diversity of studies requires the generalization—if possible—of the
conceptual frame and methodological approaches to reach a unified theory of
ecological networks. In the following sections, I outline what are considered the
most promising avenues to achieve this goal.
12.3 The Future of Network Studies: Dealing

with Complexity
Future studies in ecological networks have the main challenge of adding

complexity. To achieve this goal, we must look at the several forms of
oversimplification in the present status of network studies. For example, since
biological communities are composed of individuals, populations, and species
interacting in different ways, places, and times, incorporating complexity and
realism requires studies to be multilayered. Multilayer networks are composed of
different types of nodes and edges combined into a single representation (Pilosof et
al. 2015). Accordingly, a multilayer network should have intralayer edges (i.e.,
those linking intralayer nodes) and interlayer edges (i.e., those linking nodes
from different layers) (Fig. 12.3). Examples of multilayer networks in ecology are
those that combine different interaction types (e.g., Pocock et al. 2012), habitats
(e.g., Borthagaray et al. 2014), or points in times (e.g., Schoenly and Cohen 1991).
The integration of the multilayer network theory
176 L. Cagnolo
Fig. 12.3 Example of the

composition of a
multilayer network. Layers
can be different points of
time, sites, or interaction
types; each layer has
intralayer interactions
(colored lines) and
interlayer interactions
(dotted lines) composed of
shared nodes connecting
different layers
into ecology requires coming out of the comfort zone of, for example, our
organismal- taxonomic affinities, and provides an opportunity for multidisciplinary
collaboration. Unfortunately, ecology science still lacks a proper conceptual frame
and its appropri- ate tools (but see Pilosof et al. 2017 for a description of multilayer
networks and tools for ecological studies). Below, I present a list of ideas and
suggestions to incorporate complexity in research studies of ecological networks.
12.4 Combining Different Interaction Types
The study of ecological networks emerged as a consequence of understanding

that species do not occur alone in ecosystems, and their fitness is a conse-
quence of the interactions established with other species, among other factors.
Most research has focused on the study of networks of particular interactions
(e.g., herbivory, pollination) and involving particular guilds (e.g., leaf miner herbi-
vores, hummingbird pollinators). Although network science was (and is)
successful
in describing complex ecological systems, we still lack a comprehensive under-

standing of the biotic environment of species to draw more realistic conclusions.
Food web studies usually contain different types of interactions in a single
web (e.g., herbivores, detritivores, and predators), but in the last few years,
attempts have been made to explicitly merge different interaction types into
network studies (Lewinsohn and Cagnolo 2012). For example, in a large net-
work containing both mutualistic and antagonistic links from Spain, the ratio
between mutualistic and antagonistic links per plant species was very hetero-
geneously distributed in the plant community, which resulted in a strong influ-
ence on total species richness (Melián et al. 2009). Pocock et al. (2012)
assembled a “network of networks” from several studies conducted in an
organic farm in southwest England. They found that networks did not covary in
their robustness, suggesting that management actions that benefit one guild or
functional group will not necessarily benefit others. Recently, a plant-centered
mutualistic network including pollination, protection, and seed dispersal inter-
actions did not show to be more robust than the sum of their parts, but more
interestingly, some species that connected different sub-networks showed to
have a disproportionate effect on community cohesion (Dáttilo et al. 2016).
These examples show diverse evaluations of how different layers are connected
and influence each other in ecological communities. Unfortunately, ecology
has a deficiency of structured protocols to evaluate the properties of these tran-
sition zones between networks that share species and are composed of different
interaction types (Kéfi et al. 2016).
Merging different interaction types into a single network is a methodological
challenge because it requires consistent sampling protocols and standardized
statistics to make the effects of different interaction types comparable.
Accordingly, Palla et al. (2005) proposed several statistics to describe multilayer
networks where each node (species, individuals, etc.) i is a member of mi networks
ov
(e.g., layers in Fig. 12.3). a,b
The overlap size s between networks a and b (e.g.,
layers 1 and 2 in Fig. 12.3) is the number of shared nodes, which represents a link
between both networks (dotted lines in Fig. 12.3). The linkage level of aa network
com com
a is d , called community degree. Finally, a
they propose s as the number of
nodes in a network ov com com
a. The distributions of these four parameters mi, s , d , and s are then used
a,b a a
to describe multilayered networks (Palla et al. 2005). This approach is interesting
because it focuses on the transition zones of interactive communities composed of
species of different interaction types, and it allows us to evaluate the spread of
disturbances across a large ecosystem (Olesen et al. 2010).
12.5 Adding Trophic Levels
There are few examples of merging different interaction types into a single
network, but including several trophic levels into a single network is one of the
straightforward ways. The earliest food web studies usually included species
feeding in different
178 L. Cagnolo
modes, mixing producers with secondary and tertiary consumers, and even
scavengers (e.g., Polis 1991). Consequently, the concept of trophic levels became
useful to understand the mechanisms of energy transfer from producers to the
whole community (Cohen et al. 2009). However, the development of tools and
concepts for the study of bipartite networks in the last 15 years has focused on the
interaction between two contiguous trophic levels, usually plants and their
associated animals. Although the advances in this topic are amazing, as is reflected
by the fast growth of number and diversity of studies (Fig. 12.1), in the next
decade the challenge will be to identify the interface of different trophic levels and
understand how non- contiguous trophic levels influence each other.
The description of three-level networks is widespread in the ecological
literature, particularly regarding plant-herbivore-parasitoid networks (e.g.,
Memmott et al. 1994; Lewis et al. 2002). These “tri-trophic” networks allow us to
examine the reciprocal influence of extreme trophic levels mediated by the mid-
level, but neither the classical food web approach nor the bipartite network
analysis offers a proper conceptual and methodological framework to fully
understand a tri-trophic structure. The bipartite approach ignores bottom-up and
top-down effects of non- participant trophic levels. This is not irrelevant since, for
example, herbivores in tropical forests have to deal with plant defenses and higher
rates of predation and parasitism (Schemske et al. 2009), and it is likely that
herbivore specialization has evolved and is maintained in response to one or both
of these forces (Lewinsohn et al. 2005). Consequently, herbivores select host
plants by searching for enemy- free space (Heard et al. 2006), and parasitoids are
attracted differentially by plant volatiles (Boone et al. 2008), suggesting that non-
participant trophic levels affect species interactions in other trophic levels.
Fontaine et al. (2011) proposed six different schemes resulting from merging
two bipartite networks based on the combination of nested, modular, and random
topologies. A particular topology, in which modules of the basal network (e.g.,
plant- herbivore) translate into modules in the upper network or vice versa, is
interesting because it could be a clue of mutual influence. Accordingly,
overlapped modules in consecutive bipartite networks may represent a co-
evolutionary vortex of tri-trophic interactions; in addition, species and interactions
that connect different tripartite modules may be keystones for community
maintenance and species diversification (Leppänen et al. 2013). Moreover, it was
argued that cascades, i.e., the influence of biotic interactions on species in
contiguous upper or lower trophic levels (Polis et al. 2000) tend to be ameliorated
in tropical systems as a consequence of reticulate networks and high species
diversity (Dyer and Letourneau 1999); therefore, the extent of such cascades may
also depend on the structure of the tripartite networks and the degree of overlap
between successive modules.
The structural patterns of tripartite networks are unknown because there are
scarce tools and concepts for their study. Murata (2010) proposed a method for
detecting communities from tripartite networks composed of Internet users, URLs,
and tags. Murata’s approach offers a solution to evaluate tripartite network
modularity; this method employs spectral partitioning and can detect
communities from
networks that are composed of thousands of nodes and tens of thousands of hyper-
edges. This is an example of the usefulness of paying attention to areas other than
ecology and incorporate tools and concepts from other fields of study that could be
applied to solve ecological questions.
12.6 Disaggregating Data: Adding Space,

Time, and Individuals Information
Most network studies are the result of pooled interactions recorded at different
places (sites, transects, points, etc.) and moments (Fig. 12.3). How much of the
known network patterns is a consequence of the spatiotemporal accumulation of
interactions? For example, most networks constructed with aggregated data
overes- timate the connectivity of species (Trøjelsgaard and Olesen 2016),
affecting our predictions, for example, about the extent of secondary extinctions.
Aggregated data may mask the spatiotemporal roles of species. For example,
Pimm and Lawton (1980) proposed that modularity in food webs may emerge as a
consequence of habitat boundaries, as found in the Serengeti food web
(Baskerville et al. 2011) and in a bumble bee-flower network (Dupont et al. 2014).
Unfortunately, most studies actively select homogeneous plots for sampling and
pooling data from different plots or sites; thus, the opportunity to evaluate the
possibility that habitat generalists connect different modules composed of habitat
specialists is missed.
Taking space and time into account is a fundamental task since before
proposing any other explanation we should discard spatiotemporal overlap of
interaction partners (Vázquez et al. 2009). Studies comparing networks across
time and space found that macroscopic statistics (i.e., those that describe the global
structure of a network, such as connectance and modularity) tend to be more stable
than the microscopic structure (i.e., the identity and position of particular species
and interactions) (Trøjelsgaard and Olesen 2016). Recording changes in the
phenological intensity of biotic interactions through space would add another layer
to the microscopic features of interaction networks and could contribute with a
deeper understanding of fine-scale changes. Therefore, the inclusion of space and
time to our network studies will allow us to identify species with important roles,
which may be keystones for community persistence.
Another form of data aggregation is the accumulation of individual-level data
into a single node, usually referring to species (Ings et al. 2009). Nevertheless,
sometimes individuals display a variety of behaviors and morphological traits; for
example, the invasive ant Solenopsis invicta showed a highly variable trophic level
position between colonies in a 0.5-ha plot, with some colonies acting as primary
consumers whereas others were top predators (Roeder and Kaspari 2017). How
does this variability influence on the properties of ecological networks? As pro-
posed for space and time, the aggregation of data overestimates connectivity and
180 L. Cagnolo
underestimates specialization. For example, Tur et al. (2014) compared the modu-
larity of a plant-pollinator network resolved at the level of plant species and
pollinator individuals. Their main result reflects an increase in the modularity at
the individual’s level, mainly driven by phenology (Tur et al. 2014). These studies
reveal large variability that individual partners show in their interactions; conse-
quently, the individual-based networks do not seem to behave as the species-level
network.
While depending on the question under study, the patterns of networks com-
posed of aggregated data may lead to wrong conclusions (Poisot et al. 2014),
which may result in a misunderstanding of ecosystem functioning and in reducing
the prediction potential of ecology. Nevertheless, almost all the different kinds of
data aggregation in network studies can be solved if we consciously plan our sam-
pling design, incorporating the spatial and temporal scales that are relevant to the
studied system.
12.7 Using Barcoding Techniques
There are remarkable examples of combinations of theory and natural history for
the development of networks (e.g., Novotny and Basset 2005; Novotny et al.
2010), but the time necessary to acquire taxonomical and natural history
knowledge is usually scarce. Ecologists have incorporated different approaches
to solve the difficulty of cataloging interactions in complex environments. One of
these tools is sequencing barcodes directly from DNA extracted from consumers,
which provide strong physical evidence for the host association (Evans et al.
2016). As genes rather than morphology drive identifications, barcoding is less
dependent upon taxonomic knowledge, making it particularly appreciated for
studying poorly known taxa and ecosystems.
The use of barcoding in the construction of ecological networks is growing rap-
idly not only because it is an easy way of depicting interactions, but also because
it facilitates the discovery of rare ones. Most network studies that used barcoding
techniques coincide in showing a significant increase in the number of detected
interactions, leading to higher complexity and more connected networks (e.g.,
Smith et al. 2008). The utility of barcoding in resolving plant-herbivore-parasitoid
food webs is obvious, and recent studies have applied this approach creatively,
providing responses to interesting questions. For example, González-Varo et al.
(2014) extracted DNA from the surface of seeds in bird feces for the identification
of bird– plant interactions. The difference with traditional frugivory studies lies in
the possibility to evaluate not only “who eats who,” but the effectiveness of the
dispersion process. This example reflects how barcoding techniques are improving
network science and, hopefully, opening questions that may trigger a new wave of
network studies in ecology.
12.8 Final Considerations
The study of networks in the tropics had blossomed in the last 16 years, dramati-
cally increasing the diversity of studied interactions (Fig. 12.1). The evolution of
network studies reveals interesting trends supported by increasing
multidisciplinary. The studies conducted in the last 6 years have revealed that the
increase in the diversity of interaction types catalyzed the mixing of concepts and
tools from different areas of ecology and even from different areas of science (Fig.
12.2).
The possibility of incorporating new interactions into network science depends
on the knowledge of natural history and the development of new tools that solve
methodological difficulties. Although there are remarkable examples of combina-
tions of theory and natural history in network studies, the possibilities of develop-
ing research in tropical forests for periods of time long enough to incorporate
sufficient natural history knowledge is usually limited. Ecologists have incorpo-
rated different approaches to solve the difficulty of cataloging interactions in
complex environments. One of these tools is barcode sequencing directly from
DNA extracted from consumers, which provides strong physical evidence for
associations while facilitating phylogenetic analysis (Evans et al. 2016). The
future of networks in the tropics will necessarily rely on metabarcoding tech-
niques to make diversity traceable, particularly regarding metamorphosing
insects. Accordingly, there are promising examples showing differences in net-
works built using barcoding and classical techniques (Smith et al. 2008; Wirta
et al. 2014; Evans et al. 2016).
One of the major challenges of network research in the tropics is to increase
level of complexity. Here, I propose two ways: merging different interaction types
into single networks and disaggregating data into their spatial, temporal, and
individual layers. The multilayer approach requires new concepts and methods that
are starting to be formalized (Pilosof et al. 2017). Particular attention should be
paid to the tran- sitions zones between layers because they could provide clues
about keystone nodes and interactions that may be crucial for community stability
and cohesion (Olesen et al. 2010; Pocock et al. 2012). Besides using the theory
and tools developed by ecologists, we should pay attention to other areas of
science such as sociology and engineering for inspiration.
Unlike any other field of science, ecology deals with the rush of habitat destruc-
tion. The need of ecological studies is particularly urgent in the tropics because the
most diverse ecosystems are in developing economies, where nature conservation
and the necessity of resources exploitation are in conflict. Accordingly, network
science has the role of connecting species lists to ecosystem functions, a key
priority in conservation ecology.
Acknowledgements L.C. is a career researcher with Consejo Nacional de Investigaciones

Científicas y Técnicas of Argentina. This work was supported by a grant from FONCYT–
ANPCYT (PICT 2014-3168).
182 L. Cagnolo
References
Chapter 13
A Useful Guide of Main Indices and Software
Used for Ecological Networks Studies
Reuber Antoniazzi Jr, Wesley Dáttilo, and Victor Rico-Gray
Abstract Complex network analysis allows ecologists to implement interesting

and diverse approaches to study interactions among the most diverse life forms. In
the last decades, several tools and advances have been developed in software, ran-
domizations, and computer graphics; increasing the quantity of ecologists that lead
authorship when these analyses are used in their research. Thereby, some metrics
and indices have been improved and others appeared as novel approaches,
establishing a vast quantity of information in literature. In this chapter, you will
be able to find a compendium of the main descriptors currently used in the
literature, as well as the primary information to develop the statistical analysis and
graph visualization. It is important to have enough criteria when using these
metrics and indices, which must be complemented with both: knowledge
concerning natural history and the logic and limitations of the indices and analysis,
in order to avoid misleading conclusions.
R. Antoniazzi Jr (*) • W. Dáttilo

Red de Ecoetología, Instituto de Ecología A.C., Carretera antigua a Coatepec 351,
El Haya, CP 91070 Xalapa, Veracruz, Mexico
e-mail: [email protected]; [email protected]
V. Rico-Gray
Instituto de Neuroetología, Universidad Veracruzana,
Av. Dr. Luís Castelazo s/n, CP 91190 Xalapa, Veracruz, Mexico

185
https://doi.org/10.1007/978-3-319-68228-0_13
Table 13.1 Principal parameters, indices, functions, and metrics typically used in the analysis of ecological networks 18
6
Symbol Network descriptors Description Software
nx Number of nodes Species richness of any trophic level (e.g., plant n1, animal n2) or number of Spreadsheet
individuals
S Network order Total number of nodes, which S = n1 + n2 + ⋯ Spreadsheet
If n represents species richness the network is G (Delmas et al. 2017)
L Network size Number of edges present in the network, i.e., number of pairwise Spreadsheet
interactions (Delmas et al. 2017)
m Number of interactions Interaction frequency in a weighted web (Bascompte and Jordano 2013) Spreadsheet
aij Number of pairwise Pairwise interaction between partners i and j in a weighted web Spreadsheet
interactions between (Bascompte and Jordano 2013)
lower trophic level
species i and higher
trophic level species j
Ai, Aj Total number of Column or row sums in weighted networks (Bascompte and Jordano 2013) Spreadsheet,
interaction records for A = ∑ n1 a ; A =
n2
a “bipartite” package
lower trophic level (i) i j=1 ij j ∑ i=1 ij in R (Dormann et al.
or higher trophic level 2008), Pajek
(j) (Batagelj and Mrvar
1998)
k j, k i Degree of higher kj is the number of interactions between higher trophic level j and Spreadsheet,
trophic level (j) or lower trophic level i “bipartite” package
lower trophic level ki is the number of interactions between lower trophic level i and higher in R (Dormann et al.
(i) trophic level j (Bascompte and Jordano 2013) 2008), Pajek
(Batagelj and Mrvar R.
1998) An
〈km〉, 〈kn〉 Average degree of any 〈km〉 is average number of interactions for higher trophic level Spreadsheet, to
trophic level 〈kn〉 is average number of interactions for lower trophic level “bipartite” package nia
(Bascompte and Jordano 2013) in R (Dormann et al. zzi
2008), Pajek Jr
(Batagelj and Mrvar et
1998) al.
〈k〉 Average number of Spreadsheet,
interactions per any two L “bipartite” package
trophic levels ‹k› = in R (Dormann et al.
n1 + n2 2008), Pajek
(Batagelj and Mrvar
1998)
P(k) Degree distribution N(k ) “bipartite” package in 13
Probability that a node has k edges within the network, P(k ) = , in which R (Dormann et al. A
S of
N(k) is the number of nodes with k edges, and S is the number total 2008), Pajek (Batagelj Us
nodes in the network (Delmas et al. 2017) and Mrvar 1998) ef
ul
CD Degree centrality Vertex: Simple count of the number of interactions established by a node “bipartite” package in Gu
R (Dormann et al. ide
CD(i) = ki, in Freeman (1977) or normalized by the maximum degree 2008), Pajek (Batagelj
C of
C= D and Mrvar 1998) M
D
kmax ain
In
dic
Whole graph:′ ∑ n a ( p ,p ) , in Freeman (1978) es
C (D p )k = i=1
i k
n −1 an
CC Closeness centrality Measures the proximity of a node to all other nodes in the network “bipartite” package d
d in R (Dormann et al. So
ij
CC(i ) = ∑ , in Freeman (1978) and Freeman et al. (1979) 2008) ft
j≠i
n −1 wa
CB Betweenness centrality Gives a measure of times a node is between two others nodes “bipartite” package re
g jk(i ) / g jk in R (Dormann et al. Us
CB (i ) = 2 × ∑ , in Freeman (1977) 2008) ed
j<k;i≠ j
(n −1 )( n − 2) for
Ec
ol
CE Eigenvector centrality Provides a measure how influences all other nodes “igraph” package in R og
1 (Csardi and Nepusz ica
C (i ) = A C ( j ) , where A is 1 if i interacts with j and 0 otherwise,
2006) l
and
E
λ j
∑ ij E ij Ne
tw
λ is a constant, in Bonacich (1987) or
C Connectance Proportion of established interactions relative to all possible interactions Spreadsheet, ks
L “bipartite” package St
(Jordano 1987), C = for bipartite networks, see more information in in R (Dormann et al. ud
n1 ⋅ n2 2008) ies
Delmas et al. (2017)
(continued)
18
Symbol Network descriptors Description Software 8
LDq Linkage density The average number of interactions per node in the web, “bipartite” package
n
1 1 a
n2
a  in R (Dormann et al.
2008)
LD =  ∑ k• m +∑  in Bersier et al. (2002), and the
•k
m
weighted
q n1 ,k n2 ,k
2 k=1 a ••  k=1 a ••
LDq
connectance is calculated as , where S is the total number of all
possible
S
interactions in the web
CC Clustering coefficient 2Ni “bipartite” package
For each node i: CC i
= where ki is i’s degree and Ni is the total
k ( k −1) in R (Dormann et al.
i i
2008)
number of interactions between i’s neighbors and the average local
clustering
coefficient: CC =
∑ ci , in Watts and Strogatz
(1998)
i
1
S
For the entire graph: CC = 3Nt , where Nt is the number of triangles in graph
2
N
c
G and Nc is the number of 3-nodes subgraphs, in Soffer and Vázquez (2005)

and Saramäki et al. (2007)
MB Modularity Nm
 l  d Ad B  MODULAR
M = ∑  m −  m m   where Nm is the number of modules in the (Marquitti et al. R.
Bnetwork, l 2
2014) An
 I I
m=1
 to
is the total number of links, lm is the number of links between nodes in nia
module A B
zzi
m, and dm and dm are the sum of the links of all nodes in module s which Jr
belong to A-set and B-set, respectively, in Barber (2007) et
ci Among- A measure of how connected node i is to all modules: “bipartite” package al.
module NM
 k 2 in R (Dormann et al.
connectivity c = 1 − ∑ is  ,c ∈[0,1] 2008)
i
S=1 ki 
zi Inside- Kis − Ks
Zi = , for more details see Guimerà and Nunes Amaral (2005)
module SD
ks
connectivit
y
T Matrix temperature T = kU, where ANINHADO 13
1  d 2 (Guimarães and A
U= ∑ uij , where u = ij . T = 0° is defined for maximum Guimarães 2006), Us
ij  
m.n  Dij  Nestedness ef
Temperature ul
nestedness in Atmar and Patterson (1993) Calculator (NTC) Gu
(Atmar and Patterson ide
1995) of
N Nestedness 100 − T ANINHADO M
N= ; which values range from 0 to 1 (maximum nestedness)—See ain
(matrix 100 (Guimarães and
temperature Guimarães 2006) In
above about T (Matrix temperature)
based) dic
NODF Nestedness metric ANINHADO es
based on overlap and NODF =
∑ n2 Mi< j +ij∑ n1i< j Mij ; where the first sum is across all pairs (Guimarães and an
decreasing fill of Guimarães 2006), d
 n2 ( n2 −1)   n1 ( n1 −1)  So
  +  “bipartite” package in
 2   2  R (Dormann et al. ft
lower trophic level, the second sum is across all pairs of higher trophic 2008) wa
level, n2 and n1 are the total number of lower and higher trophic level, re
respectively. Mij = 0 if ki = kj and Us
nij ed
M = otherwise (Almeida-Neto et al. 2007).
for
min (ki ,k j)
ij
Ec
η Nestedness without Nestedness ol
n(
n2 )
the rule of (n2 ) ∑ i< j ij Temperature og
η = ,
decreasing fill
∑ i< j
( ( n2 )
min n i ,nj
( n2 )
) Calculator (NTC)
(Atmar and Patterson
ica
l
here defined for lower trophic level set n2, in Bastolla et al. (2009) 1995) Ne
cn Contributio This metric is a z-score relative to null models: “bipartite” package tw
 N − Ni  or
n to cn = , where N is the observed nestedness (NODF) of the in R (Dormann et al.
nestedness 2008) ks
network, St
i 
σ N i
 ud
Ni is the average of nestedness when randomizing just the interactions of ies
the species i, and σNi is the standard deviation of nestedness when
randomizing just the interactions of the species i (n = 100 randomizations).
Positive values of cni indicate a higher contribution of species (i) to the
nested structure
(Saavedra et al. 2011)
(continued)
19
0

Symbol Network descriptors Description Software
W Web asymmetry n2 − n1 “bipartite” package
Balance between species richness of any two trophic levels, W = ;
n1 + n2 in R (Dormann et al.
2008)
positive numbers indicate more lower-trophic level nodes; negative,
more higher-trophic level species; rescaled to [−1, 1] (Blüthgen et al.
2007)
AS Mutual Difference between the interaction strengths of partners: “bipartite” package
dependence ( bij − bji ) ij a ij a in R (Dormann et al.
asymmetry n1S ij = ; where b ij= and b = jiand b values are the 2008)
max (b ,b ) n1i
n1j
ij ji
proportion of interactions between two partners, aij, relative to the totals of

each partner (n1i), (n1j)
1  aijaij 
n
H i, H j Diversity of “bipartite” package
interactions (Shannon Hi = −∑  .ln  (Bascompte and Jordano 2013) in R (Dormann et al.
j=1 n 1i n1i 
interactions) per 2008)
higher trophic level (i)
or lower trophic level
(j)
Es Evenness (Shannon −∑ i ∑ j pij ln pij “bipartite” package
entropy) of Es = in R (Dormann et al. R.
(Bascompte and Jordano 2013) An
interactions across ln ( nn
12 ) 2008)
the matrix to
Gqw Generality, or mean
n1
n1 j H “bipartite” package nia
G = ∑ 2 j , adapted from Bersier et al. (2002) by Bascompte zzi
number of links per in R (Dormann et al.
qw Jr
any higher trophic j=1 m 2008)
level mutualist et
and Jordano (2013) al.
Vulnerability, or Replace j by i and n1 by n2 in the equation for Gqw (Bascompte and Jordano
mean number of 2013)
links per any lower
trophic level
Lq Weighted interaction ∑
Lq = 0.5 n1
n1 j2 + ∑n2
n 2 (Bascompte and Jordano 2013) “bipartite” package 13
Hj 1i Hi
density in R (Dormann et al. A
 j=1 m i=1 m  2008) Us
ef
H2 Interaction diversity
i j
 aij aij  “bipartite” package ul
H
Network-level measure of diversity of interactions. 2 = − ∑∑  . ln m
i=1 j=1  m
in R (Dormann et al. Gu
(Blüthgen et al. 2006) 2008) ide
Measure of realized A standardized H2 to account for the total number of records each node “bipartite” package in of
H 2′ Eltonian has (Ai, Aj), (Blüthgen et al. 2006; Devictor et al. 2010) M
R (Dormann et al.
specialization ain
2008) or in:
In
http://rxc.sys-bio.net/
dic
by Blüthgen and
es
Blüthgen
c 
an
(d − d ) p  ′
′
d′ Species-level measure ′ “bipartite” package d
d = i min , where d = ∑  p ln ij 
of specificity  with c the in R (Dormann et al. So
dmaxof− resources,
number dmin j=1  ij qj  2008)
i ft
′ wa
p the proportion of interactions divided by the sum of performances of re
node
ij Us
i, and qj the sum of interactions of resource j divided by the total number of ed
interactions in the matrix (Blüthgen et al. 2006, 2007) for
AS′ Scaled mutual Provides the average difference in the interaction strengths of partners: “bipartite” package Ec
′ ∑ ( b ij − b )ji
dependence ′ ′
in R (Dormann et al. ol
asymmetry n′ S = , where k is the number of interactions for species i, and 2008) og
b
1 ij
k
ij
i ica
′
i
l
and b are scaled interaction-strength values (Bascompte and Jordano 2013)
ji Ne
C-score C-score Represents the average number of checkerboard units for each unique “bipartite” package tw
species in R (Dormann et al. or
M (M −1) 2008) ks
pair. For M species, there are P = species-pairs, and hence
2 St
C
C = ∑∑
scoreij
(Stone and Roberts 1990) ud
score ies
i< j P
(continued)
19
Symbol Network descriptors Description Software 2
Bw Network dissimilarity T T  ST 
B = T − C ( T −1) −1 , when CS =
T 1 −  , in which T is sites
and
w S
T −1
 ∑ n
 i i 
ST is the number of interactions and n is local richness, for more details see
Poisot et al. (2012)
BWN Additive partition of βWN = βST + βOS , i.e., the additive partition of the beta diversity of “betalink” package
the beta diversity of interactions, since βST (species turnover) and βOS (interaction rewiring) are a in R (Poisot 2016)
interactions a+ b+
subset of βWN (whittaker beta diversity of networks). Then β =
c
W
−1 , where b
(2a + b + c) / 2
is the number of unique species of the first network, c is the number of
unique species of the second network, and a is the number of shared
species between networks.
Gc Network core-peripheral ( Ki − K mean )
G = k = mean number of links for higher or lower trophic
level
c
σK
i
i
in a network, kmean = mean number of links for all higher or lower trophic
level in a network and σk = standard deviation of the number of links for all
higher or lower trophic level in a network. Species with Gc > 1 are the node
with more interactions in relation to others in the same trophic level,
namely the central core of highly node of the network (Dáttilo et al. 2013) R.
“Symbol” refers to common used on literature, “Network Descriptors” to name, a brief description appears in “Description” and when applicable “Software” An
refers to the most common to use although others can do the same analysis. See below Table 13.2 for more details about “Software.” Higher trophic level, to
e.g., refers to animals and lower trophic levels to plants, but it can be parasitoid, inquiline, predator, prey, etc. Hereafter, we chose to use n1 to the higher nia
trophic level (e.g., A) and n2 to lower trophic level (e.g., P) because then we are not being restricted. For more details about network analysis, see Bersier et zzi
al. (2002), Bascompte and Jordano (2007), Dormann et al. (2008), Ulrich et al. (2009), Vázquez et al. (2009), Rayfield et al. (2011), Bascompte and Jordano Jr
(2013), Dáttilo et al. (2016) and Delmas et al. (2017) et
al.
12 A Useful Guide of Main Indices and Software Used for Ecological Networks Studies 193
Table 13.2 Some useful software for network analysis and draw graphs
Software Brief description Source
ANINHADO Provides calculation https://www.guimara
of NODF and es. bio.br/soft.html
Temperature,
commonly used
measures of
nestedness
(Guimarães and
Guimarães 2006)
Cytoscape Useful for visualizing http://www.cytoscape
networks and allows . org/
to implement these
with attribute data.
This is an open source
software (Shannon et
al. 2003)
Gephi Affords to https://gephi.org/
visualizing and nice
exploration for
networks. This is an
open-source and
free software
(Bastian et al. 2009)
MODULAR Calculates the http://sourceforge.
modularity in net/ projects/
networks using two programmodular/
different modularity
metrics and also
includes two null
models (Marquitti et
al. 2014)
Pajek It is a profitable http://mrvar.fdv.uni-
program for lj.si/ pajek/
analyzing network
using various
metrics, as well as
offering good
drawing tools
(Batagelj and Mrvar
1998)
R software “betalink” package Calculates https://CRAN.R-
the beta- project.
diversity of org/package=betalink
interactions, besides
providing
visualization of the
networks (Poisot
2016)
“bipartite” package Provides helpful https://CRAN.R-
tools for project.
visualizing webs org/package=bipartit
and to calculates e
the most used
indices in network
ecology (Dormann
et al. 2008)
“foodweb” package Measures of food https://CRAN.R-
web network project.
structure besides org/package=foodwe
tools for b
visualizing graphs
(Perdomo et al.
2012)
(continued)
194 R. Antoniazzi Jr et al.
Software Brief description Source

“igraph” package Provides a good https://CRAN.R-
networks analysis project.
and graphs org/package=igraph,
visualization, http://igraph.org/
besides calculates
commonly used
indices (Csardi and
Nepusz 2006)
“mangal” package Offers tools to manage http://mangal.io/
data on ecological
interactions (Poisot
et al. 2016)
“network” package Allows to edit https://CRAN.R-
network objects (Butts project.
et al. org/package=networ
2008) k
“statnet” package Allows to perform https://CRAN.R-
network data, be project.
the representation, org/package=statnet
visualization,
analysis and
simulation,
(Handcock et al.
2008)
“econulln Useful to detects https://CRAN.R-
etr” resource preferences project.
package by ecological org/package=econull
networks analysis netr
using null models
(Vaughan et al.
2017)
Acknowledgements We thank P. Luna, R. Almeida, and T. Suarez for their comments on the manuscript.
References

Libro Parte 2

Uploaded by

Copyright:

Available Formats

Libro Parte 2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Libro Parte 2

Uploaded by

Copyright:

Available Formats

Chapter 8

Plant-Herbivore Networks in the Tropics

Antonio López-Carretero, Ek del-Val, and Karina Boege

© Springer International Publishing AG 2018

herbivorous insects in the tropics is evolutionarily associated with plants (Futuyma

the specificity and structural patterns of plant-herbivore networks is still under

8.2 Structure of Plant-Herbivore Interaction Networks

In contrast with mutualistic networks that tend towards generalization,

8.3 Plant-Herbivore Interaction Networks in the Tropics

(hemipteran), which are concentrated in the Neotropical region, mainly focused on

8.4 Spatiotemporal Variation of Plant-Herbivore Networks

8.5 Influence of Phylogeny, Species, and Genetic

Antagonistic plant-herbivore interactions are highly specialized as a result of

8.5.1 Plant Diversity

8.5.2 Genetic Diversity

Intraspecific genotypic variation in plant traits can also have an important

8.6 How Host Plant and Herbivorous Traits Can

8.6.1 Host Defensive Traits

species (trophic importance; López-Carretero et al. 2016). In particular, host

8.6.2 Herbivore Feeding Behavior

The specificity and intimacy (i.e., physiological dependence) of plant-herbivore

8.7 Habitat Disturbance and Plant-Herbivore

Ecosystems are in constant change due to natural and anthropogenic disturbances

8.8 Concluding Remarks and Future Directions

Records of plant-herbivore interactions for used in network analyses have been

importance of including greater ecological and taxonomic amplitude in studies on

Sybelle Bellay, Fabrício Hiroiuki Oda, Karla Magalhães Campião,

Abstract The lack of ecological studies on parasite species is contradictory with

S. Bellay (*) • R.M. Takemoto

© Springer International Publishing AG 2018

explore network patterns between tropic and non-tropic environments, in addition

Parasites are hidden components of ecological communities, comprising approxi-

9.2 Environments and Host Characteristics with Possible

9.2.1 Host Specificity and Phylogenetic History

The combination of varying levels of host specificity and abundance contributes

9.2.2 Host Body Size

host-parasite networks. Indeed, it has been demonstrated that body size is an

9.2.3 Host Diet

9.3 Parasite Characteristics Influencing Network Topology

Host extinctions or population decline may compromise the structure of parasite

removal effect of host species on a host-parasite network depends on the pattern of

9.5 Parasites and Species Introduction

Species introductions into new environments may radically alter native

Table 9.1 Possible scenarios of introduction in host-parasite networks

Table 9.1 (continued)

9.6 Molecular Approach

9.7 Final Considerations

Mauricio Cantor, Guilherme O. Longo, Luisa Fontoura,

Abstract Tropical reefs are, figuratively, the underwater counterparts of tropical

M. Cantor (*) • L. Fontoura • J.P. Quimbayo • S.R. Floeter

© Springer International Publishing AG 2018

large variety of life histories in regulated populations within complex ecological

10.2 Trophic Networks: Marine Food Webs

Predation is perhaps the most notable of the interactions among species—

incorporated “polyphagous” consumers (Fig. 10.1a). The first representations of

The role of trophic interactions in structuring reef ecosystems goes beyond

10.3 Agonistic Networks: Territoriality and

10.4 Mutualistic and Commensalistic Networks: Cleaning

Following association is a neutral interaction observed among reef fish (Sazima

10.5 A Biogeographic Panorama of Tropical Reef