Sampling protocol
All NutNet sites followed standard sampling protocols. A 1×1m subplot within each 5×5 m plot was permanently marked and sampled annually at peak biomass. The number of species, species identity, and their covers were recorded. Species cover (%) was estimated visually for all species in the subplots. Aboveground biomass was measured adjacent to the permanent subplot, by clipping all aboveground biomass within two 1 × 0.1 m strips (in total 0.2 m2), which were moved each year to avoid resampling the same location. For shrubs and subshrubs occurring in strips, we collected all leaves and current year’s woody growth. Biomass was dried at 60 °C (to constant mass) before weighing to the nearest 0.01 g. Dried biomass was multiplied by 5 to estimate grams per square meter. At some sites with strong seasonality, cover and biomass were recorded twice per year to include a complete list of species and follow typical management procedures at those sites. For those sites, the maximum cover for each species and total biomass were used in the following analyses. The taxonomy was adjusted within sites to ensure consistent naming over time. Specifically, when individuals could not be identified as species, they were aggregated at the genus level but referred to as “species” for simplicity.