Extended Data Table 3 Explanatory variables for velocity in species richness changes
From: Accelerated increase in plant species richness on mountain summits is linked to warming
- Results of linear mixed effects models (Gaussian family error) showing the relationship of the average velocity in species richness changes with the change in potential explanatory variables (temperature, precipitation, nitrogen deposition). Initial species richness on the summits was added as a further independent variable and indicated that species-rich systems showed a larger net change. The implemented model formula was lmer(ΔSR/Δt ~ ΔT/Δt + ΔP/Δt + Naccum + richness + (1|mountain)). Model performance was compared using AICc, which also defines the order of models, with the best one on top. In addition, significant results from tests using F statistics are indicated by asterisks (***P < 0.001). All values indicate model coefficients ± s.e. Rerunning the analyses after centring (subtracting the means) and scaling (dividing by s.d.) the explanatory variables indicated a larger coefficient and thus stronger effect of temperature than that of precipitation (ΔSR/Δt = 0.00 (± 0.04) + 0.39 (± 0.05) × ΔT/Δt*** + 0.22 (± 0.04) × ΔP/Δt*** + 0.21 (± 0.05) × richness***; asterisks indicate significant effects with ***P < 0.001). As no nitrogen data were available for the seven summits on Svalbard, the analyses presented in the table were performed on a subset of 389 temporal comparisons (comparing surveys and resurveys resulting from 684 observations). To account for spatial autocorrelation, we further repeated the full model averaging over all summits sampled over the same time period and falling in the same grid cell of the original climate data. The results of this model were qualitatively similar (ΔSR/Δt = −0.004 (± 0.05) + 9.7 (± 1.1) × ΔT/Δt*** + 0.005 (± 0.001) × ΔP/Δt*** − 0.14 (± 0.09) × N accum + 0.005 (± 0.001) × richness***).
