Fig. 2: Computing the extinction area in ecological networks.

a–c show a toy example of the extinction area computation. Starting from the complete network (a), animals (matrix rows) are removed according to a certain ranking. In (b) the animal A₁ has been deleted leading to the extinction of the plant F₁ which remains without links. The extinction area is the integral of the curve (c) which shows the fraction of extinct plants at a given fraction of removed animals. d is the extinction area computed at different map exponents γ of the mutualistic system Robertson 1929. The blue continuous line correspond to animal removal, where the animal ranking is provided by the x-score at the given γ. The green dash-dot line is computed using the opposite procedure: the plants are progressively removed following the y ranking. e shows the number of matrices whose extinction area by animal removal is maximized at given exponents, γ_EA, among the 116 mutualistic networks of the Web-of-Life database, see ?? and table S1, having size (product between the number of rows and columns) larger that 500. f shows the correlation between matrix density and map exponent that maximizes the extinction area - Spearman correlation ρ = 0.61, p = 10⁻²⁴. g reports the ratio between the extinction areas obtained with the fitness-complexity map with our generalization. We consider both rows and columns removal procedures.