Contrasting to predictions, roach (Rutilus rutilus) are more abundant than perch (PercaCluvi-atiIis) in a number of mesotrophic lakes in the Baltic lake region of northeastern Germany. To reveal underlying mechanisms, roach and perch habitat use were analysed on a diel and seasonal scale, and piscivorous fish predation was quantified in one of those lakes, Lake Groger V~tersee (Brandenburg). For the latter point, piscivorous and non-piscivorous fish population sizes were assessed by multiple mark-recapture experiments and piscivorous consumption was quantified by bioenergetics modelling. Piscivorous fish biomass accounted for about 30% of total fish biomass, and top-down control of third trophic level biomass was intense. Fish habitat shifts closely linked the littoral and pelagic habitat, and were an integral component of the fish species interactions. While pike predation was mainly restricted to the littoral zone, piscivorous perch foraged in the open water during daytime. Roach _> age-1 counteracted the pelagic predation risk by diel horizontal migrations. So they could prey on pelagic zooplankton, and at the same time kept their predation risk in the pelagic zone as low as possible. In the littoral zone, roach extensively used non-animal resources not accessible to perch. Non-piscivorous perch were almost exclusively restricted to the littoral habitat; only age-0 perch had an initial pelagic stage. They didn't experience a juvenile competitive bottleneck. Overall, perch were relatively more affected by piscivorous fish predation than roach. This might be decisive for roach dominance under mesotrophic conditions. Additionally, Orconectes limosus, an introduced crayfish species, played a decisive role both as alternative prey for piscivorous fish and as a potential structuring force in the littoral habitat.

SUMMARY 1. We used an individual based modelling approach for roach to (i) simulate observed diel habitat shifts between the pelagic and littoral zone of a mesotrophic lake; (ii) analyse the relevance of these habitat shifts for the diet, activity costs and growth of roach; and (iii) quantify the effects of a hypothetical piscivore‐mediated (presence of pikeperch) confinement of roach to the littoral zone on roach diet, activity costs and growth. 2. The model suggests that in the presence of pikeperch, roach shifts from zooplankton as the primary diet to increased consumption of less nutritious food items such as macrophytes, filamentous algae and detritus. 3. The growth of roach between May and October was predicted to be significantly higher in the absence of pikeperch, although the net activity costs were about 60% higher compared with the scenario where pikeperch were present. 4. These modelling results provide quantitative information for interpreting diel horizontal migrations of roach as a result from a trade‐off between food availability and predation risk in different habitats of a lake. 5. Altering the habitat selection mode of planktivorous roach by piscivore stocking has the potential to reduce zooplankton consumption by fish substantially, and could therefore be used as a biomanipulation technique complementing the reduction of zooplanktivorous fish.

In a small, 12 ha, mesotropic lake, roach Rutilus rutilus performed diel habitat shifts that clearly influenced the composition of their diet. During daytime, roach stayed in the littoral zone and concentrated on littoral prey. At night they were found in the pelagic zone, and pelagic prey items such as Daphnia spp. or Chaoborus flavicans dominated their food. On a seasonal scale, there were shifts in the importance of different food items and in the diel pattern of feeding intensity. Bioenergetics modelling in combination with an evacuation rate method for estimating daily rations allowed for changes in feeding modes to be taken into account, and so food item specific daily rations over the season could be determined. With the evacuation rate method applied on selected days, diel changes in diet compositions and feeding intensities could be quantitatively accounted for. When the 24 h integrated diet proportions were then used as an input parameter for bioenergetics modelling, food item specific consumption could be determined over the entire sampling season. The consideration of the diel diet shifts proved to be essential for the model output. If only the daytime or the night-time diet composition (derived from one single daily sampling) was taken into account for bioenergetics modelling, severe under-or overestimations of daily rations for specific food items resulted. 2002 The Fisheries Society of the British Isles