1. Periphyton chlorophyll a was measured at weekly or 2 weekly intervals from October 1992 to March 1994 at four sites in a Swiss prealpine gravel bed river that was frequently disturbed by unpredictable spates.

2. To evaluate the dominant processes that control periphyton biomass, measured data were compared with a set of simulations from an empirical dynamic periphyton model. Different combinations of process hypotheses were systematically activated and deactivated in order to assess their importance.

3. The simplest model leading to an acceptable agreement with measured data employs a biomass-dependent growth rate, a detachment rate directly proportional to discharge and biomass, and a catastrophic loss rate during bed moving spates. Terms describing light or temperature dependence had a minor effect on the model fit.

4. The model describes the temporal pattern of the periphyton biomass as a series of growth curves periodically truncated by spates. Within the uncertainties of the measurements, mainly caused by the spatial heterogeneity of periphyton, the biomass recovered along deterministic trajectories.

5. Sensitivity analyses with respect to model parameters and model structure showed that site-specific model parameters could not be unequivocally determined, and that the model yields similar results with slightly different formulations of processes. This indicates that the data base with respect to periphyton biomass was too small for a unique identification of model details but that the main conclusions on the significance of processes did not depend on arbitrary choices of the model formulation.