The Rengen Grassland Experiment (RGE), set up on a Nardus grassland in 1941, consists of a control and five fertilizer treatments (Ca, CaN, CaNP, CaNP-KCl and CaNP-K 2 SO 4 ). In 2005, soil chemical properties were analyzed to investigate the effect of soil variables on biomass production, plant species composition and species richness of vascular plants. Further, the effect of sampling scale (from 0.02 to 5.76 m 2 ) on species richness was investigated. Soil properties (plant-available contents of K, P, C:N ratio, and pH) and biomass production were found to be strictly dependent on the fertilizers applied. Diversification of soil P content between treatments with and without P application is still in progress. Biomass production was most positively affected by P and K soil contents under N application. Furthermore, pH had a small positive effect on biomass production, and C:N ratio a moderately negative one. Two types of nutrient limitation were recognized: (1) limitation of total biomass production and (2) limitation of individual plant species. Long-term addition of a limiting nutrient affected the grassland ecosystem in three ways: (1) causing a change in plant species composition without significant increase in total biomass production, (2) causing no change in species composition but with significant increase in total biomass production, and (3) causing substantial change in plant species composition accompanied by significant increase in total biomass production. The explanatory power of all measured soil properties on plant species composition was almost the same as the power of the treatment effect (61.7% versus 62% of explained variability in RDA). The most powerful predictors of plant species composition were soil P, K and Mg contents, pH, and biomass production. The soil P content and biomass production were the only variables leading to a significant negative effect on species richness. An almost parallel increase in species richness with increasing sampling area was detected in all treatments. Constant differences among treatments were independent of sampling area.

No abstract

No abstract