We studied the effects of inter-annual variability and serial correlation on the statistical power of monitoring schemes to detect trends in biomass of bream (Abramis brama) in Lake Veluwemeer (The Netherlands). In order to distinguish between 'true' system variability and sampling variability we simulated the development of the bream population, using estimates for population structure and growth, and compared the resulting inter-annual variabilities and serial correlations with those from field data. In all cases the inter-annual variability in the field data was larger than in simulated data (e.g. for total biomass of all assessed bream sigma = 0.45 in field data, and sigma = 0.03-0.14 in simulated data) indicating that sampling variability decreased statistical power for detecting trends. Moreover, sampling variability obscured the inter-annual dependency (and thus the serial correlation) of biomass, which was expected because in this long-lived population biomass changes are buffered by the many year classes present. We did find the expected serial correlation in our simulation results and concluded that good survey data of long-lived fish populations should show low sampling variability and considerable inter-annual serial correlation. Since serial correlation decreases the power for detecting trends, this means that even when sampling variability would be greatly reduced, the number of sampling years to detect a change of 15%.year(-1) in bream populations (corresponding to a halving or doubling in a six-year period) would in most cases be more than six. This would imply that the six-year reporting periods that are required by the Water Framework Directive of the European Union are too short for the existing fish monitoring schemes.