Riverine landscapes are characterised by recurrent flooding events and successional landscape mosaics with high habitat heterogeneity, providing species-specific patterns of suitable and unsuitable biotopes. Landscape characteristics, like distance, barriers and their specifications (e.g. cumulative barrier width, barrier number) and the spatial arrangement of suitable habitat areas, are expected to affect the dispersal of animals in landscapes. The distribution of voles, shrews and mice in a floodplain was monitored for 2 years using live traps. Recolonisation was found to be a slow process, resulting in a heterogeneous distribution of small mammals in the floodplains. Microtus arvalis was found just after the floods in low densities and on or near non-flooded areas only. From summer to autumn densities gradually increased, and specimens could be observed on larger distances from the non-flooded areas. The density development pattern of Crocidura russula was similar to that of M. arvalis, but densities increased faster. In contrast, Clethrionomys glareolus and Sorex araneus could immediately be observed in the former flooded areas just after the floods, but throughout the year these species were not trapped on distances further than 120 m from the non-flooded areas. Micromys minutus and Apodemus sylvaticus were trapped only occasionally in spring and summer after which those species could be found throughout the floodplain in larger densities in autumn. To analyse the influence of landscape characteristics on recolonisation, the floodplain was classified into suitable, marginal and unsuitable habitat landscape units for each of the small mammal species, based on the trapping results. Landscape characteristics relating to monitoring sites were measured from an aerial photograph using a geographic information system. After that, presence and recolonisation time of small mammals at monitoring sites could be described by multiple regression models based on these measured landscape characteristics. The predictive power of these models was tested in another floodplain by determining the species distribution after 35 weeks. Multiple regression models appeared to be useful in analysing recolonisation patterns and determining the importance of landscape characteristics for recolonisation by small mammals after flooding events. Available distribution data suggest three different types of recolonisers: (1) Gradual, density induced colonisers; (2) Active dispersers and (3) Long-distance dispersers after a lag. Results of regression models confirmed that M. arvalis could be characterised as a type 1 and A. sylvaticus as a type 3 species. The classification of the other species was not possible due to the relatively short time available for recolonisation.