Professional Documents
Culture Documents
2018 - Species-Habitat Networks - A Tool To Improve Landscape Management For Conservation
2018 - Species-Habitat Networks - A Tool To Improve Landscape Management For Conservation
DOI: 10.1111/1365-2664.13337
C O M M E N TA RY
1
DAFNAE, University of Padova, Padova,
Italy Abstract
2
Dpto. Ecologia Integrativa, Estacion 1. Land-use change is reshaping terrestrial ecosystems world-wide and is recognized
Biologica de Dõnana (EBD-CSIC), Sevilla,
as a key driver of biodiversity loss with negative consequences on ecosystem
Spain
3 functioning. Understanding how species use resources across landscapes is es-
Ecosystem Management, School of
Environment and Rural Sciences, University sential for the design of effective management strategies.
of New England, Armidale, NSW, Australia
2. Despite recent advances in network ecology, there is still a gap between theory
Correspondence and applied ecological science, and we lack the information to manage entire land-
Lorenzo Marini
scapes to maximize biodiversity conservation and ecosystem service delivery.
Email: [email protected]
While several pioneering approaches have tried to link ecological networks and
Funding information
conservation science, applied ecologists still struggle to incorporate these models
University of Padova STARS Consolidator,
Grant/Award Number: STARS-CoG; into research due to their inherent complexity.
Australian Research Council Discovery Early
3. We propose the application of bipartite networks principles to create species–habitat
Career Researcher Award, Grant/Award
Number: DE170101349 networks. This approach explicitly links multiple species and habitat resources, pro-
vides tools to estimate the importance of particular species or specific habitat in a
Handling Editor: Elizabeth Nichols
given landscape, and quantifies emerging properties of entire habitat networks.
Most existing metrics used to study properties of bipartite ecological networks can
easily be adapted to investigate species–habitat relationships. The tool use is rela-
tively simple and does not require advanced computational expertise.
4. Synthesis and applications. One of the biggest challenges in applied ecology is manag-
ing multiple habitats for the effective conservation of multiple species. One key ad-
vantage of this proposed approach is that the scale of the derived ecological
information could match the scale of landscape management interventions. The ver-
satility, visualization power and ease of interpretation of these networks will enable
application of the species–habitat network concept to a wide array of real-world
problems, such as multispecies conservation, habitat restoration, ecosystem services
management or invasion ecology. In particular, species–habitat networks could be
applied to identify optimal landscape compositions and configurations to design ef-
fective interventions at the landscape scale. This approach also enables the detec-
tion of emerging network properties that could also be used to test the effects of
large-scale drivers of global change upon ecosystem structure and stability.
KEYWORDS
bipartite networks, community stability, conservation prioritization, fragmentation, habitat
preference, landscape composition, landscape configuration, land-use change
J Appl Ecol. 2019;56:923–928. © 2019 The Authors. Journal of Applied Ecology | 923
wileyonlinelibrary.com/journal/jpe
© 2019 British Ecological Society
| Journal of Applied Ecology
924 MARINI et al.
F I G U R E 1 (a) Traditional approach to study species communities across heterogeneous landscapes. Most empirical research in landscape
ecology focuses on the local habitat versus surrounding landscape dichotomy, where the landscape is quantified in terms of composition
and/or configuration around a central point (circular buffer) where the community is sampled; (b) the species–habitat network whereby the
whole landscape is sampled and the species are sampled at multiple sites (line width proportional to species abundance). The landscape can
be classified in patches according to the functional role of the different habitats for the target species community
MARINI et al. Journal of Applied Ecology |
925
2 | B E YO N D TH E FO C A L H A B ITAT: provide examples of a few commonly used metrics, while several
B U I LD I N G S PEC I E S–H A B ITAT N E T WO R K S comprehensive reviews of different metrics are available (Blüthgen,
Menzel, & Blüthgen, 2006; Dormann et al., 2009). For a simple,
In the most simple case, habitat types and the species occurring technical application of the metrics described below we present a
within each habitat constitute the two types of nodes of a species– worked example in the Supporting Information using r.
habitat network. The focal species community would usually share In bipartite networks, nestedness is a central property that de-
a similar functional role, with examples being lichens, pollinators, scribes network structure. A network is said to be nested when the
ground-dwelling predatory arthropods or insectivorous mammals. communities of sites that have a few links (i.e. species) are a subset of
The flexibility of the proposed approach allows habitat nodes to the communities of sites with more links (Baselga, 2010; Figure 2a).
be further defined as individual sites where the community was In a nested species–habitat network, the entire system will likely be
sampled (Burns & Zotz, 2010), similar to ecological interaction net- affected if the most species-rich habitat or site is removed. In con-
works downscaled from species to individual levels (Koch, Dáttilo, trast, the removal of species-poor habitats that only interact with a
Camarota, & Vasconcelos, 2018; Tur, Vigalondo, Trøjelsgaard, few habitat generalists is unlikely to have significant ripple effects.
Olesen, & Traveset, 2014). That is, each individual site could affect In a bipartite network, it is also possible to quantify modularity that
network topology and stability depending on attributes such as hab- measures the strength of division of a network into modules. A mod-
itat quality, management or connectivity. Note that the links need to ule comprises a set of habitat sites and species that interact more
be carefully formulated as they can affect the ecological interpreta- with each other than with other sites and species outside the module
tion of the species–habitat network. Here, we use the operational (Figure 2b). Often, networks with a modular structure are expected
definition of a link as the occurrence/abundance of a particular spe- to have a lower risk of collapse due to their buffering capacity to
cies in a certain site. system perturbations (Dormann, Und, & Martin Schaefer, 2017;
The spatial extent in which the community is sampled should be Gilarranz et al., 2017). However, the loss of specific sites may also
selected according to species’ foraging ranges (e.g. for mobile organ- affect the associated species in the same module due to low redun-
isms) or propagule dispersal (e.g. for sessile organisms) and to the dancy. Hence, the degree of both nestedness and modularity can have
ecological hypotheses underpinning the study. This issue is similar profound conservation implications.
to the selection of buffer radii when adopting a traditional approach A common way to assess the implications of habitat loss is to look
to quantify landscape composition or configuration. It is important at network robustness. The robustness of a network can be a key met-
to stress that, as the spatial extent of the habitat mosaic used by the ric for conservation prioritization of high value sites and ecosystem
species is generally large (e.g. 1–10 km for mobile organisms), it is management (Sole & Montoya, 2001), as it is defined as the network
likely that most surveyed species–habitat networks would be sub- resilience to the loss of habitat patches or species. For instance, simple
units of much larger networks (Jordano, 2016). While the definition simulations that remove habitats randomly or in realistic sequences are
of species as nodes is usually straightforward, the way in which hab- one way to quantify community robustness to habitat loss (Figure 2c).
itats are defined as nodes can be more complex. Spatial grain and Understanding species selectivity is central to assess the extent
habitat classification can affect the topology (and hence interpreta- of habitat generalization. For instance, network specialization (H2′)
tion) of the network. In heavily modified landscapes, different habi- can be useful for comparisons across different networks. The more
tats are often organized in patches, which can be defined as discrete selective the species are the greater the value of H2′ for the net-
areas with a definite shape, size and configuration. The focal species work: 0 (no specialization) and 1 (complete specialization; Blüthgen
community may be used to guide the identification of habitat types et al., 2006). Unselective species used habitats proportionally to
that are functionally relevant. From an operational point of view, we their size (green line in Figure 2d), while selective species used sites
suggest that habitat nodes are defined according to the dominant irrespective of habitat area (blue line in Figure 2d). This metric can be
vegetation, accounting for differences in structure and function for used to test the effect of different drivers (e.g. landscape intensifica-
different communities. tion, climate change) on habitat preferences.
Species–habitat network analysis can also provide insights into
the roles of specific habitat sites or species in the network. First,
3 | TO O L S FO R A N A LYS I N G the influence of one site upon another site can be assessed using
S PEC I E S–H A B ITAT N E T WO R K S apparent influence metrics (Muller, Adriaanse, Belshaw, & Godfray,
1999). This index quantifies how much one habitat site contributes
Most existing metrics used to study properties of bipartite ecologi- to sustaining the species present in another site (Figure 2e). This a
cal networks can easily be adapted to the study of species–habitat typical problem in applied ecology when practitioners introduced
networks. These metrics can address key applied questions and novel habitats or interventions in a landscape mosaic.
can be broadly divided into two groups: emergent properties of Another useful metric is node strength (Bascompte, Jordano, &
the whole network and node-level metrics that measure the role Olesen, 2006). This metric captures a single site importance tak-
of single nodes (i.e. single habitats, sites or species) in the network ing into account how much the species depend on this site. A site
(Dormann, Frund, Bluthgen, & Gruber, 2009; Figure 2). Here, we has high strength if it supports a high number of species with high
| Journal of Applied Ecology
926 MARINI et al.
F I G U R E 2 An untapped network toolbox for assessing species–habitat links. Bipartite network analysis is a mature field able to identify
emerging properties of a system (a–d) as well as the roles that individual nodes (species or habitat sites) play in the network (e–h). Circles
and rectangles represent species and habitat sites, respectively. Here, we present only a few examples of the metrics that can be computed
(Blüthgen et al., 2006; Dormann et al., 2009). See text for details
dependency (i.e. specialists) on it (node a in Figure 2f). Conversely, the community is too large (e.g. large mammals or birds) compared
sites that only host a few generalist species (node b) have low with the feasibility of field sampling. Third, the species–habitat
strength playing a minor role in the landscape. networks may be limited in use when the landscape structure is
Finally, we can also assess the contribution of particular nodes characterized by very high habitat heterogeneity at a spatial scale
to network level metrics like modularity or nestedness. For exam- much smaller than the average species dispersal. However, human-
ple, using among-module connectivity (c) we can identify hub species impacted landscapes often present high contrast between habitat
connecting different modules (Figure 2g). This can help to identify types providing ideal conditions to apply the framework. Fourth,
key-stone sites or species that can affect the robustness of the whole in its present form species–habitat networks are not spatially ex-
network. To quantify habitat specialization at the species level, gen- plicit and therefore information on the spatial relationships be-
erality or preference can be also measured at the node level using tween patches (e.g. connectivity) cannot be derived. Finally, as for
several selectivity metrics (e.g. specialization d’; Figure 2h). most empirical interaction networks, species–habitat networks
would suffer to some extent from under-s ampling. This limitation
is particularly challenging in situations where the communities
4 | U S E A N D LI M ITATI O N S O F TH E host many rare species or when species are inherently difficult
FR A M E WO R K to sample or observe. Similar to ecological networks, this raises
the concern that rare species of high conservation concern may
There are several important conditions to consider when opera- not be appropriately represented in the data. Increasing sampling
tionalizing species–habitat networks. First, users must ensure that effort is, therefore, recommended to minimize these biases. At the
the data inputs are realistic and relevant to the sampled commu- data analysis stage, when interaction networks contain many sin-
nity to ensure meaningful results are obtained through network gletons, the role of these single observations needs to be carefully
analysis. For sessile organisms, such as lichen or plant species, explored with the available tools (Vizentin-B ugoni et al., 2016).
occurrence directly links to resource use and habitat preference
(Burns & Zotz, 2010). On the contrary, for mobile organisms that
use multiple resources, species occurrence can assume different 5 | I M PLI C ATI O N S FO R ECOS YS TE M
ecological meanings (Kremen et al., 2007). For instance, if we con- M A N AG E M E NT A N D P O LI C Y
sider a specific habitat patch, a species can be recorded at that site
because individuals can use multiple resources (e.g. host plants for An urgent question in conservation ecology is to understand how to
reproduction, prey, nesting site or structure for roosting or shel- manage whole landscapes to maximize biodiversity conservation or
ter) or simply because individuals are using that site as a stepping ecosystem services delivery (Albert et al., 2017; Harvey et al., 2017;
stone for dispersal. Second, not all taxa can be appropriately stud- Mendenhall, Shields-Estrada, Krishnaswami, Daily, & Levin, 2016).
ied by species–habitat networks. One situation where the frame- The flexibility of the proposed approach is that it can easily be ap-
work might not be applicable is when average species dispersal in plied across a range of ecological fields such as species conservation,
MARINI et al. Journal of Applied Ecology |
927
habitat restoration, ecosystem services management, or invasion networks along relevant environmental gradients, these emerging
ecology (Memmott, Craze, Waser, & Price, 2007). Here, we provide properties can be used to test the effects of large-scale drivers of
as an example four important research directions that could be ad- global change upon ecosystem structure and stability (Pellissier et al.,
dressed by adopting species–habitat networks: 2018; Schleuning et al., 2012; Tylianakis & Morris, 2017). The ver-
satility, visualization power and ease of interpretation of these net-
works will enable the application of the species–habitat network
5.1 | Conservation prioritization
concept to a wide array of real-world problems concerning biodiver-
Conservation actions often face the trade-off between maximum pro- sity conservation and ecosystem service enhancement at different
tection of the environment and a limited budget. Site strength values in a spatial scales.
landscape or in a protected area network can be used to prioritize which
sites to conserve to maximize the biodiversity of any target taxon.
AC K N OW L E D G E M E N T S
5.2 | Land-use change and community stability L.M. was partially supported by the University of Padova STARS
Consolidator Grant (STARS-CoG)-2017. R.R. was supported by the
Conservationists often aim to achieve maximum biodiversity rep- Australian Research Council Discovery Early Career Researcher
resentation, without an explicit focus on the long-term stability. Award DE170101349. We thank Christopher N. Kaiser-Bunbury,
Recent studies (Gilarranz et al., 2017) on ecological networks have Kirsten Miller and one anonymous reviewer for the useful comments
tried to use architectural patterns such as modularity to understand that improved the manuscript.
the mechanisms underlying the stability of communities. Similarly,
we can investigate if certain species–habitat structures confer sta-
bility to the system in order to predict the robustness of species– AU T H O R S ’ C O N T R I B U T I O N S
habitat interactions to habitat perturbations.
L.M. and I.B. conceived the study and developed the concepts. All
authors contributed critically to the drafts and gave final approval
5.3 | Maximizing biodiversity-based ecosystem for publication.
services
Landscape interventions to support ecosystem services often require DATA AC C E S S I B I L I T Y
the introduction of new habitats (e.g. hedgerows, mass flowering
All code and data for creating the practical example included in the
crops) across a landscape. For instance, pollinators and pest control
supplementary material are available via the Zenodo Repository https://doi.
agents are known to be enhanced by the proximity to semi-natural
org/10.5281/zenodo.2525376 (Marini, Bartomeus, Rader, & Lami, 2018).
areas (Kennedy et al., 2013; Schellhorn et al., 2015). Simulations
using different landscape configurations of crop fields and green
infrastructures can be used to maximize the positive influence among ORCID
sites. For example, placing early mass flowering crops in the right
Lorenzo Marini https://orcid.org/0000-0001-7429-7685
configuration may maximize ecosystem service delivery, without
imposing negative effects on natural habitats (Magrach et al., 2017). Ignasi Bartomeus https://orcid.org/0000-0001-7893-4389
Blüthgen, N., Menzel, F., & Blüthgen, N. (2006). Measuring specialization Memmott, J. (2009). Food webs: A ladder for picking strawberries or a
in species interaction networks. BMC Ecology, 14, 6–9. https://doi. practical tool for practical problems? Proceedings of the Royal Society
org/10.1186/1472-6785-6-9 B: Biological Sciences, 364, 1693–1699. https://doi.org/10.1098/
Burkle, L. A., Marlin, J. C., & Knight, T. M. (2013). Plant-pollinator inter- rstb.2008.0255
actions over 120 years: Loss of species, co-occurrence, and function. Memmott, J., Craze, P. G., Waser, N. M., & Price, M. V. (2007). Global
Science, 339, 1611–1615. https://doi.org/10.1126/science.1232728 warming and the disruption of plant-p ollinator interactions.
Burns, K. C., & Zotz, G. (2010). A hierarchical framework for investigat- Ecology Letters, 10, 710–717. https://doi.org/10.1111/j.1461
ing epiphyte assemblages: Networks, meta- communities, and scale. -0248.2007.01061.x
Ecology, 91, 377–385. https://doi.org/org/10.1890/08-2004.1 Mendenhall, C. D., Shields-Estrada, A., Krishnaswami, A. J., Daily, G. C.,
Dormann, C. F., Frund, J., Bluthgen, N., & Gruber, B. (2009). Indices, graphs and & Levin, S. A. (2016). Q uantifying and sustaining biodiversity in trop-
null models: Analyzing bipartite ecological networks. The Open Ecology ical agricultural landscapes. Proceedings of the National Academy of
Journal, 2, 7–24. https://doi.org/10.2174/1874213000902010007 Sciences of the United States of America, 113, 14545–14551. https://
Dormann, C. F., Und, J. F., & Martin Schaefer, H. (2017). Identifying doi.org/10.1073/pnas.1604981113
causes of patterns in ecological networks: Opportunities and limita- Muller, C. B., Adriaanse, I. C. T., Belshaw, R., & Godfray, H. C. J.
tions. Annual Review of Ecology, Evolution, and Systematics, 48, 559– (1999). The structure of an aphid-p arasitoid community. Journal
584. https://doi.org/10.1146/annurev-ecolsys-110316 of Animal Ecology, 68, 346–370. https://doi.org/10.1046/j.1365
Fischer, J., & Lindenmayer, D. B. (2006). Beyond fragmentation: -2656.1999.00288.x
The continuum model for fauna research and conservation in Pellissier, L., Albouy, C., Bascompte, J., Farwig, N., Graham, C., Loreau, M.,
human-modified landscapes. Oikos, 112, 473–480. https://doi. & Gravel, D. (2018). Comparing species interaction networks along en-
org/10.1111/j.0030-1299.2006.14148.x vironmental gradients. Biological Reviews of the Cambridge Philosophical
Gilarranz, L. J., Rayfield, B., Liñán-Cembrano, G., Bascompte, J., & Society, 93, 785–800. https://doi.org/10.1111/brv.12366
Gonzalez, A. (2017). Effects of network modularity on the spread of Schellhorn, N. A., Gagic, V., & Bommarco, R. (2015). Time will tell:
perturbation impact in experimental metapopulations. Science, 357, Resource continuity bolsters ecosystem services. Trends in Ecology &
199–201. https://doi.org/10.1126/science.aal4122 Evolution, 30, 524–530. https://doi.org/10.1016/j.tree.2015.06.007
Gonzalez, A., Rayfield, B., & Lindo, Z. (2011). The disentangled bank: How loss Schleuning, M., Fründ, J., Klein, A.-M., Abrahamczyk, S., Alarcón,
of habitat fragments and disassembles ecological networks. American R., Albrecht, M., & Blüthgen, N. (2012). Specialization of mu-
Journal of Botany, 98, 503–516. https://doi.org/10.3732/ajb.1000424 tualistic interaction networks decreases toward tropical lati-
Harvey, E., Gounand, I., Ward, C. L., & Altermatt, F. (2017). Bridging tudes. Current Biology, 22, 1925–1931. https://doi.org/10.1016/j.
ecology and conservation: From ecological networks to ecosys- cub.2012.08.015
tem function. Journal of Applied Ecology, 54, 371–379. https://doi. Sole, R. V., & Montoya, M. (2001). Complexity and fragility in ecologi-
org/10.1111/1365-2664.12769 cal networks. Proceedings of the Royal Society B: Biological Sciences,
Jordano, P. (2016). Sampling networks of ecological interactions. 268, 2039–2045. https://doi.org/10.1098/rspb.2001.1767
Functional Ecology, 30, 1883–1893. https://doi.org/10.1111 Tscharntke, T., Tylianakis, J. M., Rand, T. A., Didham, R. K.,
/1365-2435.12763 Fahrig, L., Batáry, P., & Westphal, C. (2012). Landscape mod-
Kaiser-Bunbury, C. N., & Blüthgen, N. (2015). Integrating network ecol- eration of biodiversity patterns and processes: Eight hy-
ogy with applied conservation: A synthesis and guide to implemen- potheses. Biological Reviews, 87, 661–685. https://doi.
tation. AoB Plants, 7, plv076. https://doi.org/10.1093/aobpla/plv076 org/10.1111/j.1469-185X.2011.00216.x
Kennedy, C. M., Lonsdorf, E., Neel, M. C., Williams, N. M., Ricketts, T. Tur, C., Vigalondo, B., Trøjelsgaard, K., Olesen, J. M., & Traveset, A.
H., Winfree, R., … Kremen, C. (2013). A global quantitative synthe- (2014). Downscaling pollen-transport networks to the level of
sis of local and landscape effects on wild bee pollinators in agro- individuals. Journal of Animal Ecology, 83, 306–317. https://doi.
ecosystems. Ecology Letters, 16, 584–599. https://doi.org/10.1111/ org/10.1111/1365-2656.12130
ele.12082 Tylianakis, J. M., & Morris, R. J. (2017). Ecological networks
Koch, E. B. A., Dáttilo, W., Camarota, F., & Vasconcelos, H. L. (2018). across environmental gradients. Annual Review of Ecology,
From species to individuals: Does the variation in ant–plant networks Evolution, and Systematics, 48, 25–48. https://doi.org/10.1146/
scale result in structural and functional changes? Population Ecology, annurev-ecolsys-110316-022821
60, 309–318. https://doi.org/10.1007/s10144-018-0634-5 Vizentin-Bugoni, J., Maruyama, P. K., Debastiani, V. J., Duarte, L. da S.,
Kremen, C., Williams, N. M., Aizen, M. A., Gemmill-Herren, B., LeBuhn, Dalsgaard, B., & Sazima, M. (2016). Influences of sampling effort
G., Minckley, R., & Ricketts, T. H. (2007). Pollination and other eco- on detected patterns and structuring processes of a Neotropical
system services produced by mobile organisms: A conceptual frame- plant-hummingbird network. Journal of Animal Ecology, 85, 262–272.
work for the effects of land-use change. Ecology Letters, 10, 299–314. https://doi.org/10.1111/1365-2656.12459
https://doi.org/10.1111/j.1461-0248.2007.01018.x
Magrach, A., Holzschuh, A., Bartomeus, I., Riedinger, V., Roberts, S. P. S U P P O R T I N G I N FO R M AT I O N
M., Rundlöf, M., … Vilà, M. (2017). Plant-pollinator networks in semi-
natural grasslands are resistant to the loss of pollinators during Additional supporting information may be found online in the
blooming of mass-flowering crops. Ecography, 41, 62–74. https://doi. Supporting Information section at the end of the article.
org/10.1111/ecog.02847
Marini, L., Bartomeus, I., Rader, R., & Lami, F. (2018). Data from:
Species–Habitat networks: A tool to improve landscape management
for conservation. Zenodo Digital Repository, https://doi.org/10.5281/ How to cite this article: Marini L, Bartomeus I, Rader R,
zenodo.2525376 Lami F. Species–Habitat networks: A tool to improve landscape
Mayer, A. L., Buma, B., Davis, A., Gagné, S. A., Loudermilk, E. L., Scheller, management for conservation. J Appl Ecol. 2019;56:923–928.
R. M., … Franklin, J. (2016). How landscape ecology informs global
https://doi.org/10.1111/1365-2664.13337
land-change science and policy. BioScience, 66, 458–469. https://doi.
org/10.1093/biosci/biw035