Temperature-driven changes in interactions between populations are crucial to the estimation of the impact of global warming on aquatic food webs. We analysed inter-annual variability in two data sets from Bautzen reservoir, Germany. In a long-term data set (1981-1999) we examined the pelagic phenology of Daphnia galeata, a keystone species, the invertebrate predator Leptodora kindtii, phytoplankton and Secchi depth in relation to water temperature and the North Atlantic Oscillation index. In a short-term data set (1995-1998) we examined food web relations, particularly the consumption of D. galeata by young-of-the-year (YOY) percids and L. kindtii and rates of population change of D. galeata (abundance, recruitment pattern and non-consumptive mortality). The start of the clear-water stage (CWS) was correlated with winter temperatures. It started 5.8 days earlier per degree warming after warm winters (mean January-March temperature>or=2.5 degrees C) compared to cold winters (mean temperature<2.5 degrees C). However, the end of the CWS remained relatively constant. Predation by L. kindtii and YOY percids on D. galeata started distinctly earlier, i.e. by 13.0 and 6.5 days per degree warming, respectively, in years when the average May temperature was high (>or=14 degrees C) compared to years when it was low (<14 degrees C). Significant reductions of Daphnia abundance in midsummer occurred only in years in which the mean May temperature exceeded 14 degrees C. This temperature regime resulted in a match of over-exploitation of food resources by Daphnia during the CWS and strong predation by YOY percids and L. kindtii. Consumptive mortality increased at higher rates with a rise in temperature than net recruitment, resulting in lower Daphnia densities at the end of the CWS. Our data suggest that even low warming by 1.7 degrees C during a short, but critical seasonal period, resulting in the coincidence of two or more factors adversely affecting a keystone species, such as Daphnia, may induce changes in whole lake food webs and thus alter entire ecosystems.