Long-term variability of zooplankton populations in aquatic mesocosms.

The natural variability on a spatial and temporal scale was examined in the zooplankton community of mesocosms from Syngenta Crop Protection AG (Stein, Switzerland), with the focus on improving the experimental design and evaluation of mesocosm studies. Analysis was performed using zooplankton data collected during a three-year period in 3 (1996 and 1998) to 12 (1997) ponds. Interreplicate variability was measured as the variance among the 3 to 12 replicates at each sampling date. Temporal variation was examined as seasonal variability by comparing different sampling dates within a year and as year-to-year variation by comparing pooled data year by year. Univariate and multivariate methods were used for the evaluation of population and community data, respectively. Results from the present study indicate that because of the low interreplicate variability, only data from high-abundance species could be evaluated with a precision able to detect effects less than 20%. For the majority of the zooplankton populations, abundances were lower than 10 organisms/L, with frequent zero counts resulting in a weak evaluation of the data with a precision able to detect effects of greater than 20 and 110%. Ordination analysis of the community data from the three years revealed that approximately 29% of the total variance could be explained by year-to-year differences, whereas 11% could be attributed to seasonal variability within a year. The residual variance can be attributed to interreplicate variability and sampling error. These results were in line with findings for individual populations. The present analysis demonstrated that the inherent variability of a system should be investigated for a proper design and evaluation of mesocosm studies and promotes the use of multivariate tools for a more comprehensive interpretation of mesocosm data.