With regard to model complexity the two simplest

Similar to other studies in plant Naltrexone and hydrology, the strategy of using qualitative patterns (Wiegand et al., 2006 and Wiegand et al., 2008) or multiple model hypotheses (Clark et al., 2011) to select alternative model structures was effective in reducing uncertainty in selected model components. The coexistence pattern was the strongest filter in a cascade of three patterns (Fig. 1). This indicates that the coexistence pattern includes system information at a higher hierarchical ecosystem level (Wiegand et al., 2003) and can only be reproduced if models incorporate specific mechanisms (Section 4.2). In principle, only model structures that explicitly considered the functional relationship between flood events and biomass regulation were able to reproduce the coexistence pattern (Table 2), whereas model structures that considered flood disturbance as a random process failed passing the coexistence pattern. This minimises the risk of potentially misinterpreting the model structures that passed the coexistence pattern as plausible models (Wiegand et al., 2003). Note that model structure Mgs is an exception to this observation because it incorporates species-specific seasonal leaf shedding but intentionally ignores species-specific flood resistance. As a consequence interspecific competition becomes stronger than intraspecific competition, and coexistence of three species is not possible (Section 4.2).