The eutrophication of lakes causes serious problems, such as water-bloom co
nsisting of the harmful blue-green algae. To prevent an unusual degree of g
rowth of phytoplankton, various restoration methods have been tried. "Bioma
nipulation": which uses zooplankton to prey on the phytoplankton, is a prom
ising method because it can, In principle, create a desirable ecosystem. Ho
wever, there are many equivocal points in the long-term stability of a newl
y developed ecosystem after biomanipulations. The fate prediction of a wate
r ecosystem using a lake mathematical model is thus essential for implement
ing biomanipulations. There remain, however, many uncertain points in the d
ynamic parameters of the lake models, especially the parameters concerned w
ith zooplankton and fish. In this study, we established a lake mathematical
model based on the carbon, nitrogen and phosphorus flows among the biotic
and abiotic substances in a water ecosystem. In this model, two types of ph
ytoplankton (non blue-green algae and blue-green algae), three types of zoo
plankton (rotatoria. cladocera and copepoda), planktivorous fish (crucian c
arp), detritus and dissolved matters were considered. The dynamic parameter
s were estimated with a set membership estimation algorithm using random sc
anning based on mesocosm experiments, which can provide more rigid data tha
n actual lake observations can. The model calculations showed good agreemen
t with the observation data for both the non-filamentous blue-green algae d
ominant case and the filamentous blue-green algae dominant case. Using this
model, we calculated the long-term stability of the effects of biomanipula
tions in a hypothetical lake. The calculation results clarified the blue-gr
een algal species specificity in the longterm effects of biomanipulations.
In the non-filamentous blue-green algal dominant case, the Long-term stable
suppression of algal growth was predicted under a certain threshold fish b
iomass, while in the filamentous blue-green algal dominant case, the suppre
ssion was not sustainable even in a no-fish-coexisting series. Model simula
tion is very important to the design of biomanipulations, and the model pro
posed in this study is suitable for this purpose. (C) 2000 Elsevier Science
Ltd. All rights reserved.