In the second-year data, treatment accounted for a rather small fraction of total deviance as compared to block and topographic position. Only for unicity did treatment overall seem to have an appreciable effect (Table 2). When broken into individual treatment groups, species richness per plot (log transformed) was relatively high in winter-only and year-round grazing, while it was low in full exclosure. When treatment, location and block were considered, the model could account for 64 % of the total deviance in richness between quadrats. Treatment overall accounted for 10 % explained deviance, without being statistically significant.
A similar pattern was seen for unicity, for which winter-only grazing on average had a more than 50 % higher level than full exclosure. Here, the model could account for 61 % of the total deviance, with treatment overall being statistically significant and accounting for 17 % of total deviance.
No statistically significant effects of treatment on the forb:graminoid ratio was found. Nonetheless, we find the tendency for annual mowing to have the lowest ratio noteworthy.  For explanatory models based on calibrated biomass estimates, the patterns for forb:graminoid ratio and litter were very similar. The models showed lower explained deviance for litter and higher for the forb:graminoid ratio.