Simulating the effects of biomanipulation on the food web of Lake Ringsjön

A dynamic, process-oriented, deterministic and phosphorus-based model was developed to simulate the food web dynamics of Lake Ringsjön, in particular the long-term effects of biomanipulation in terms of reduction of omnivorous fish. The model contains 14 state variables, each with a differential equation describing sources and sinks of phosphorus. The state variables encompass piscivorous and omnivorous fish, zooplankton, phytoplankton, sediment and lake water. The model simulates densities of fish and phytoplankton adequately, both before and after biomanipulation, although the actual lake phytoplankton density varied more year-to-year compared to the model predictions. According to the model, a biomanipulation will cause an increase in zooplankton biomass. This prediction contradicts available field data from the lake which do not indicate any significant change in zooplankton biomass resulting from the performed biomanipulation. This discrepancy may partly be attributed to structural uncertainties in the model, related to the size structure of predators on zooplankton, i.e. the omnivorous fish community. The simulations suggest that phosphorus was routed along the pelagic food chain to a larger extent after omnivorous fish were removed, whereas the amount of phosphorus routed via the sediment and benthivorous fish decreased following fish removal. Accordingly, translocation of phosphorus from sediment to water by benthivorous fish is predicted to be substantially reduced by biomanipulation, resulting in an overall reduction in the release of new phosphorus to phytoplankton. Irrespective of simulated fishing effort (reduction of ≤0.5% d−1 for two years), the model predicts that P-release from the sediment and the external load will remain sufficiently high to force the system back to its previous state within a decade. Thus, recurrent biomanipulations and/or combined abatement strategies may be necessary to maintain low phytoplankton density. Known structural model uncertainties may however affect the robustness of such detailed predictions about the system resilience.