Limits...
Strong impact of temporal resolution on the structure of an ecological network.

Rasmussen C, Dupont YL, Mosbacher JB, Trøjelsgaard K, Olesen JM - PLoS ONE (2013)

Bottom Line: The dynamical network had strong time delays in the transmission of direct and indirect effects among species.Twenty percent of all indirect links were impossible in the dynamical network.Consequently, properties and thus also robustness of ecological networks cannot be deduced from the static topology alone.

View Article: PubMed Central - PubMed

Affiliation: Department of Bioscience, Aarhus University, Aarhus C, Denmark.

ABSTRACT
Most ecological networks are analysed as static structures, where all observed species and links are present simultaneously. However, this is over-simplified, because networks are temporally dynamical. We resolved an arctic, entire-season plant-flower visitor network into a temporal series of 1-day networks and compared the properties with its static equivalent based on data pooled over the entire season. Several properties differed. The nested link pattern in the static network was blurred in the dynamical version, because the characteristic long nestedness tail of flower-visitor specialists got stunted in the dynamical networks. This tail comprised a small food web of pollinators, parasitoids and hyper-parasitoids. The dynamical network had strong time delays in the transmission of direct and indirect effects among species. Twenty percent of all indirect links were impossible in the dynamical network. Consequently, properties and thus also robustness of ecological networks cannot be deduced from the static topology alone.

Show MeSH

Related in: MedlinePlus

A five–species subnetwork of the Zackenberg pollination network, 2010, including three pollinator species and two plant species.The 2010–season lasted from day 165 to day 234. Numbers below a species gives its phenophase. A, static view: All species interact directly or indirectly, e.g. Cerastium arcticum interacts directly with Lasiopiophila pilosa and Drymeia segnis, and indirectly with Saxifraga cernua and Fucellia pictipennis. B, dynamical view: The season is cut into nine temporal windows as defined by the phenologies of the species, capturing the different linkage events. Bars above each time-window display the length of the time-frame. Within a given time window, highlighted species and links are active, whereas faded ones are inactive, i.e. they are either active earlier or later.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3852737&req=5

pone-0081694-g001: A five–species subnetwork of the Zackenberg pollination network, 2010, including three pollinator species and two plant species.The 2010–season lasted from day 165 to day 234. Numbers below a species gives its phenophase. A, static view: All species interact directly or indirectly, e.g. Cerastium arcticum interacts directly with Lasiopiophila pilosa and Drymeia segnis, and indirectly with Saxifraga cernua and Fucellia pictipennis. B, dynamical view: The season is cut into nine temporal windows as defined by the phenologies of the species, capturing the different linkage events. Bars above each time-window display the length of the time-frame. Within a given time window, highlighted species and links are active, whereas faded ones are inactive, i.e. they are either active earlier or later.

Mentions: In an ecological network all species are connected, either directly or indirectly (Figs. 1–2). If the latter is the case, the connection passes through a series of directly linked species. The time delay (or distance) dij, between two directly or indirectly linked species i and j is the time difference between the start of their phenophases. i and j may be a plant and pollinator species, two plants or two pollinators. In the latter two cases, the linkage is always indirectly through either a pollinator or a plant, respectively. Such species become temporal couplers of i and j (TC in Fig. 2B). In order to estimate dij, we used the phenophase data. Then the delay between any species pair has to be constrained by the relative position of their phenophases [20]. In [20], Tang et al. looked at 1–mode human–contact and brain–cortical networks, but the concept can easily be adopted by 2–mode network analysis. A 2–mode network is made up of two interacting communities, e.g. plants and their pollinators. Static ecological networks are most often small–worlds with short path length and high clustering, resulting in high connectivity, e.g. [36]. In such networks, species and their links are all assumed to have complete temporal overlap, i.e. their presence is simultaneous and dij,  = 0 (Figs. 1A, 2A). Consequently, disturbances spread immediately among species, whether they are directly or indirectly connected, because time delays are ignored. Thus static networks overestimate real connectivity because they do not catch these time–dependent properties [20]. This crucial difference between static and dynamic networks is illustrated in Figs. 1–2.


Strong impact of temporal resolution on the structure of an ecological network.

Rasmussen C, Dupont YL, Mosbacher JB, Trøjelsgaard K, Olesen JM - PLoS ONE (2013)

A five–species subnetwork of the Zackenberg pollination network, 2010, including three pollinator species and two plant species.The 2010–season lasted from day 165 to day 234. Numbers below a species gives its phenophase. A, static view: All species interact directly or indirectly, e.g. Cerastium arcticum interacts directly with Lasiopiophila pilosa and Drymeia segnis, and indirectly with Saxifraga cernua and Fucellia pictipennis. B, dynamical view: The season is cut into nine temporal windows as defined by the phenologies of the species, capturing the different linkage events. Bars above each time-window display the length of the time-frame. Within a given time window, highlighted species and links are active, whereas faded ones are inactive, i.e. they are either active earlier or later.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3852737&req=5

pone-0081694-g001: A five–species subnetwork of the Zackenberg pollination network, 2010, including three pollinator species and two plant species.The 2010–season lasted from day 165 to day 234. Numbers below a species gives its phenophase. A, static view: All species interact directly or indirectly, e.g. Cerastium arcticum interacts directly with Lasiopiophila pilosa and Drymeia segnis, and indirectly with Saxifraga cernua and Fucellia pictipennis. B, dynamical view: The season is cut into nine temporal windows as defined by the phenologies of the species, capturing the different linkage events. Bars above each time-window display the length of the time-frame. Within a given time window, highlighted species and links are active, whereas faded ones are inactive, i.e. they are either active earlier or later.
Mentions: In an ecological network all species are connected, either directly or indirectly (Figs. 1–2). If the latter is the case, the connection passes through a series of directly linked species. The time delay (or distance) dij, between two directly or indirectly linked species i and j is the time difference between the start of their phenophases. i and j may be a plant and pollinator species, two plants or two pollinators. In the latter two cases, the linkage is always indirectly through either a pollinator or a plant, respectively. Such species become temporal couplers of i and j (TC in Fig. 2B). In order to estimate dij, we used the phenophase data. Then the delay between any species pair has to be constrained by the relative position of their phenophases [20]. In [20], Tang et al. looked at 1–mode human–contact and brain–cortical networks, but the concept can easily be adopted by 2–mode network analysis. A 2–mode network is made up of two interacting communities, e.g. plants and their pollinators. Static ecological networks are most often small–worlds with short path length and high clustering, resulting in high connectivity, e.g. [36]. In such networks, species and their links are all assumed to have complete temporal overlap, i.e. their presence is simultaneous and dij,  = 0 (Figs. 1A, 2A). Consequently, disturbances spread immediately among species, whether they are directly or indirectly connected, because time delays are ignored. Thus static networks overestimate real connectivity because they do not catch these time–dependent properties [20]. This crucial difference between static and dynamic networks is illustrated in Figs. 1–2.

Bottom Line: The dynamical network had strong time delays in the transmission of direct and indirect effects among species.Twenty percent of all indirect links were impossible in the dynamical network.Consequently, properties and thus also robustness of ecological networks cannot be deduced from the static topology alone.

View Article: PubMed Central - PubMed

Affiliation: Department of Bioscience, Aarhus University, Aarhus C, Denmark.

ABSTRACT
Most ecological networks are analysed as static structures, where all observed species and links are present simultaneously. However, this is over-simplified, because networks are temporally dynamical. We resolved an arctic, entire-season plant-flower visitor network into a temporal series of 1-day networks and compared the properties with its static equivalent based on data pooled over the entire season. Several properties differed. The nested link pattern in the static network was blurred in the dynamical version, because the characteristic long nestedness tail of flower-visitor specialists got stunted in the dynamical networks. This tail comprised a small food web of pollinators, parasitoids and hyper-parasitoids. The dynamical network had strong time delays in the transmission of direct and indirect effects among species. Twenty percent of all indirect links were impossible in the dynamical network. Consequently, properties and thus also robustness of ecological networks cannot be deduced from the static topology alone.

Show MeSH
Related in: MedlinePlus