In 2003, we chose 16 homogeneous grassland patches (ca. 5 m in diameter) that covered the variation in grassland productivity within one square-km area. This resulted in a relatively wide range of biomass values which made it easier to estimate relationship between biomass proxies and actual biomass. We sampled one, randomly located 0.5 m× 0.5 m plot in each patch in each year between 2003 and 2012, and in every even years since then. Patches are permanent, but plots within patches are not permanent to avoid the effect of disturbance due to repeated sampling.