Power analysis results. The Power analysis relates the number of test areas to the actual effect size which a Pearson product-moment correlation test (a) (left panel) or a t test (b) (right panel) would identify as significant (at a p level of 0.05). Effect sizes are either correlations between species numbers and a numerical predictor variable (e.g. proportional area of GMP fields) or differences in species numbers between two sets of n test areas. We assumed that the required power of the test is 0.8 (i.e. a type II error probability of 0.2) and that significance is tested for the one-sided hypotheses of a correlation coefficient >0 (left panel) and a loss of species over time on the same plots (paired test of species number at time point 1 < species number at time point 2). The grey (standard deviation = 40 species) and black (standard deviation = 20 species) lines in the right panel represent different assumptions on the variance of species numbers across test areas as derived from precursor Austrian projects [e.g. ].