We explored patterns of land snail assemblages using 93 alluvial forest sites in six river floodplains of the Elbe drainage basin (northwestern Bohemia, Czech Republic). Differences in species richness and composition across the four floodplain forest types (i.e., alder carrs, ash-alder forests, willow-poplar softwood forests, and hardwood forests) were analysed using generalized linear models, multidimensional scaling and redundancy analysis with the Monte Carlo permutation test. The studied floodplain forest types did not differ in species richness, except for the alder carrs which were significantly poorer. The number of species expressed a significant unimodal response along with elevation and Ellenberg nutrients, and further significantly decreased towards the most humid sites. Contrary to species richness, the main forest types clearly differed based on land snail species composition, with the exception of the ash-alder and willow-poplar forest sites which became completely overlapped in the ordination space. The main changes in species composition were mostly associated with elevation and Ellenberg moisture on the first MDS axis: Ellenberg nutrients and light were fitted on the second and the third axes, respectively. These variables, along with calcium content estimated using Ellenberg indicator values for soil reaction, had significant effects on the variation and snail species composition in the final RDA model. No response of either species richness or compositional changes was found for the measured content of topsoil calcium, most likely due to the higher importance of other variables. On the basis of some recently published data we can conclude that historical development and long-term human activities on the succession of floodplain assemblages have resulted in a sharp impoverishment of strictly land snail species of several hardwood forest sites in the majority of lower river stretches. Whilst in most areas there are no exact palaeoecological data available, these historical influences were closely correlated with the site elevation in our dataset as the main difference in species composition was hard to explain solely using environmental predictors.