Initial impact of the Gabčíkovo hydroelectric scheme on the species richness and composition of 0+ fish assemblages in the Slovak flood plain,River Danube

Relatively little information exists on the effects of hydroelectric schemes on 0+ fish composition in large European rivers because few or no pre‐impact data exist. We compared 0+ fish species richness and composition, relative density, fish size as well as available and used habitat using data from 12 floodplain sites sampled just prior to (1992) and four years after (1996) the start of operations of the Gab?íkovo hydropower station on the River Danube (Slovakia). We also used modelling techniques to assess the change in species richness and habitat use and to predict 1996 occurrences from the 1992 data set. The floodplain was greatly modified by the hydroscheme. Only 12 of 27 sites sampled in August 1992 were extant in August 1996. Therefore, all four channel types identified (flowing, abandoned, weir, wing‐dam) were more lentic in 1996 than in 1992, with increased width, smaller‐sized sediment (silt, clay) and greater amounts of macrophytes. After the operations of the hydroscheme, the overall relative density of fishes (individuals per surface area) of all ages decreased, with the exception of 0+ fishes, despite a slight reduction in 0+ fish density in all channel types except weirs. Species number increased from 25 to 28, although in all channel types there was a change in the composition of the 0+ fish assemblages, with rheophiles generally replaced by limnophiles and migrants from the lower Danube. The two most important microhabitat variables were the proportion of macrophytes and gravel, the latter being the factor distinguishing 0+ fish microhabitat use in 1992 (preferences) and 1996 (indifference or avoidance). Species richness and 0+ fish density in 1996 could be predicted from the 1992 data using simple log‐linear models (density, richness, sample number). Species‐specific occurrence in 1996 could not be predicted using environmental/fish data from 1992 with multiple regression or generalized additive models (GAM). However, the overall GAM from 1992 could predict overall fish occurrence in 1996. Copyright © 2003 John Wiley & Sons, Ltd.