Environmental predictors
We used a generalized joint attribute model for dynamic data (below) to assess how density-dependent and independent factors contribute to the observed changes in relative abundance of species groups over time and their steady-state predicted abundances across multiple global change drivers (Clark et al. 2020). We jointly estimated the influence of snow depth, nitrogen deposition and temperature on the density-independent growth rates of dominant, subdominant, moderate and rare species groups both in experimentally manipulated and control plots over time. We incorporated continuous annual environmental data as model predictors, following the approach of Farrer et al. (2014), as environmental data were not available at the plot level for the entirety of the study (See supplementary methods).