Results
In control plots, there was no change over time in species richness (mean slope -0.08 species/year, -0.27 to 0.09 95% CI, Table S2) or aboveground biomass (4.97, 95% CI: -4.37 to 13.97, biomass g/m2/year) (Figure 2 a, b). Similar numbers of species were lost (-0.19, 95% CI: -0.28 to -0.11, species loss (s.loss)/year) and gained (0.12 95% CI: 0.04 to 0.21, species gained (s.gain)/ year) (Figure 3a, b). In control plots, biomass loss associated with species losses each year (-0.56, 95% CI: -0.97 to -0.26, SL g/m2 associated with species loss/year) was slightly less than the biomass gain associated with species gains each year (4.02, 95% CI: 2.6 to 5.86, SG g/m2 associated with species gain /year) (Figure 3c, d). Biomass change associated with persistent species showed considerable variation, but no directional change (-4.47, 95% CI: -10.76 to 1.84, PS g/m2associated with persistent species/year) (Figure 3e).
In plots treated with NPK, species richness declined (-0.47, 95% CI: -0.66 to -0.22, species/year, Fig 2a) and biomass increased (19.99, 95% CI: 7.14 to 31.11 g/m2/ year, Fig 2b) over time (see Figure S2 for site level means between year 0 and the most recent year of measurement). The response to NPK addition of species richness change over time and biomass change over time were not correlated (-0.29, 95% CI: -0.63 to 0.19; Fig 2c, Supplementary Information). Compared to controls, NPK treatments increased species loss over time (-0.38, 95% CI: -0.51 to -0.26 species/year Fig 3a), whereas the rate of species gained did not differ from zero (-0.01, 95% CI: -0.08 to 0.06 species/year, Fig 3b). In NPK plots, biomass loss was associated with species loss over time (-7.44, 95% CI: -10.18 to -4.92 g/m2/year, Fig 3c). Species that were gained in NPK plots were associated with positive biomass change over time (7.36, 95% CI: 5.27 to 9.77 g/m2/ year, Fig 3d). Finally, change in biomass over time associated with persistent species exhibited considerable variation in NPK treatments (3.05, 95% CI: -6.14 to 11.88 g/m2/year, Fig 3e). Combined, biomass gains associated with species gained, and biomass increases associated with persistent species over time contributed to overall biomass gained in NPK plots.
Effects of each component on species and biomass change through time between control plots and NPK plots can be compared directly (Figure 4). While species losses were greater than species gains under fertilization, species gains contributed positively overall to biomass, along with an increase in biomass of persistent species. Species losses and gains due to nutrient addition were largely uncorrelated (0.29, 95% CI: -0.03 to 0.58, Table S5), as was the net change in biomass from losses and gains (-0.07, 95% CI : -0.38 to 0.23). Biomass change associated with species losses and biomass change in persistent species responses to NPK were also uncorrelated (-0.24, 95% CI: -0.55 to 0.09), as was the relationship between biomass changes from species gains and persistent species (-0.06, 95% CI: -0.39 to 0.29).
Whilst the additive nature of the partition means that changes in species richness are the sum of the species gains and losses, and overall changes in biomass are the sum of biomass changes associated with species gains, losses, and persistent species, we found some differences between our estimates of the rate of change for the whole (e.g., species richness) versus the parts of compositional change that contribute to richness change (e.g., gains and losses). The overall model-estimated effect of NPK on species richness change per year (-0.47) (Figure 2c) was qualitatively similar to the sum of the overall model-estimated effect of NPK on species loss and species gains (-0.38 (s.loss) + -0.01 (s.gain) = -0.39) (Figure 4). However, the overall model-estimated effect of NPK on biomass (19.99, CI: 7.14 to 31.11) (Figure 2c) was much higher than the estimated overall effects of NPK on each partitioned component of biomass change (-7.44 (SL) + 7.36 (SG) + 3.05 (PS) = 2.97) (Figure 4). We attribute this difference largely to the uncertainty and variation in biomass change associated with three components, and statistical regularisation associated with partitioning plot-level biomass change into three separate parts across 59 sites, 3 blocks, and ~6 plots per site. This regularisation introduced on each partitioned component leads to a more conservative estimate of the rate of biomass gains overall when compared to the overall estimates of strip biomass (Figure 2c, Figure 4).
There are four combinations of richness and biomass responses to NPK addition for the experimental sites in this study. Most sites (41) experienced an overall decrease in species richness and an increase in biomass (-richness, + biomass), nine sites experienced a decrease in species richness and a decrease in biomass (-richness, - biomass), seven experienced an increase in species richness and an increase in biomass (+richness, +biomass), and two sites experienced an increase in richness and a decrease in biomass (+richness, -biomass), but these negative biomass responses were very close to zero (Figure 2c, Figure 5, see Supplementary Table S1, and Figure S7 for site level estimates). Sites with differing overall community responses in terms of species richness and biomass have notably different species gains and losses dynamics (Figure 5). Under fertilization we expect biomass to increase and species richness to decline (+ biomass, -species richness), but in ~15% of sites, biomass is decreasing and richness is increasing (-biomass, -richness) (Figure 5, Figure S7). This variation in overall biomass and richness responses was driven by a combination of biomass loss associated with species loss (SL) and a biomass decrease associated with persistent species (PS) (Figure 5). In contrast, in sites where biomass increased and species increased, few species were lost (s.loss) and there was little biomass loss associated with these losses (SL), biomass increases were associated with species gains, and biomass associated with persistent species was relatively stable overall (PS) (Figure 5).