Bar Plot from Scientific Research

A qualitative assessment of variable importance by showing results from a multimodel inference approach. Lists of the best generalized linear models out of all potential combinations of environmental stressors are given for Global #1a and Global #2. The R 2 s of the saturated models (those from which variable importance was calculated) were 0.3834 for Global #1 and 0.5905 for Global #2. If the variable (columns) was included in the model (is significant), this is shown with a coloured box (green for multistressors, red for climatic individual stressors, blue for soil-related stressors and yellow for other forms of human influence). Effect sizes of each variable can be seen in Supplementary Table 6 . The Bayesian information criterion (BIC) indicates the suitability of the model (the lower the better) and delta indicates the difference of BIC with respect to the best model (<4 indicates similar performance to the best model). The weight of the models is also represented in the table and indicates how each of the best models would contribute to an average model (results in Supplementary Table 7 ).

DOI