The Robustness and Restoration of a Network of Ecological Networks

See allHide authors and affiliations

Science  24 Feb 2012:
Vol. 335, Issue 6071, pp. 973-977
DOI: 10.1126/science.1214915

You are currently viewing the figures only.

View Full Text

Log in to view the full text

Log in through your institution

Log in through your institution

  1. Fig. 1

    Species’ interaction networks at Norwood Farm, Somerset, UK. The entire network of networks is shown at top left (in which each circle represents one species), and quantitative visualizations are shown for each of the seven quantified individual networks (in which each block is a species, and the width of blocks of each color represents relative abundance). Details of the networks are given in table S1 and (14). Dark green and light green circles and blocks indicate noncrop and crop plants, respectively, whereas other colors indicate animal groups. Scale bars indicate the abundance of animal taxa. Plants are scaled in proportion to their interactions with animals in each network.

  2. Fig. 2

    The average robustness of the 11 animal groups (median R ± 90% confidence interval). The colored bars, matching the colors in Fig. 1, show RS and the adjoining white bars show RQ. As Embedded Image →1, animal groups are increasingly robust to the simulated sequential loss of plant taxa, whereas for animal groups with low robustness, Embedded Image →0.5.

  3. Fig. 3

    Correlations between the robustness of animal groups and the simulated loss of plant taxa in networks of the farmland species’ interaction network. The robustness of flower visitors to one random sequence of plant loss is the area under the curve for (A) the qualitative case and (B) the quantitative case. The pairwise correlations in robustness varied in the 20,000 simulations of the sequential loss of plant taxa, as two examples (C and D) show. These pairwise correlations were summarized to show the connectivity between all animal groups, considering (E) RS and (F) RQ.

  4. Fig. 4

    The relative importance of the plants in the Norwood Farm network of quantified networks. (A) The relative importance of the plants varied by habitat with colors from white to red representing increasing abundance, as shown in (G), and was calculated as shown in this workflow (B to G). The importance of each species of plant (j) for each animal group (i) was the coefficient of determination (rij2); that is, the square of the correlation coefficient, between the calculated robustness with plants removed in random order and the position of the plant in that order, as exemplifed for (B) Rubus fruticosus and butterflies, (C) Anthriscus sylvestris and flower visitors, (D) Persicaria spp. and birds, and (E) Anthriscus sylvestris and leaf-miner parasitoids. (F) The weighted sums of these coefficients of determination across groups (g) gave the importance (I) of each plant taxon; in this case, the groups were weighted according to their uniqueness (SOM part 4). (G) Abundance [assessed as leaf area of the plants (14, 31)] was strongly related to importance for a subset of plant taxa, so the relative importance of each plant taxon (RI) was calculated as the residual from the steeper regression line (determined by a two-component mixture regression model), exemplifed by Cirsium vulgare (Cv), Anthriscus sylvestris (As), and Hordeum vulgare (Hv).

Stay Connected to Science