Predicting diversity of juvenile neotropical fish communities:: patch dynamics versus habitat state in floodplain creeks

The species richness of communities should largely depend on habitat variability and/or on habitat state. We evaluated the ability of habitat variability and habitat state to predict the diversity of juvenile neotropical fish communities in creeks of a river floodplain. The young-fish fauna consisted of 73 taxa, and samples were well distributed over a wide range of relevant temporal and spatial habitat variability. We were unable to demonstrate clear patterns of richness in relation to temporal and spatial habitat variability (if habitat state variables were not included), regardless of the temporal variability scale, the grouping of sites (up- and downstream sites differed in temporal variability patterns), taxonomic units or life stages considered. Using stepwise multiple regression, 36% of the variance in species richness was explained for all data, and at best 47% was explained for all taxonomic units at upstream sites using temporal and spatial habitat variability and habitat state (bank length, mean width, mean water level before fishing and/or water turbidity). Using Monte Carlo simulations, we blindly predicted 31% (all data) and at best 37% (all upstream taxa) of the observed variance in species richness from these model types. This limited precision is probably because rare species produced most of the richness patterns in our creeks. The prediction of these rare species is generally difficult for various reasons, and may be a problem in many ecosystem types.