AbstractQuestions: 1. How do physiography, flooding regime, landscape pattern, land‐cover history, and local soil conditions influence the presence, community structure and abundance of overstorey trees? 2. Can broad‐scale factors explain variation in the floodplain forest community, or are locally measured soil conditions necessary?Location: Floodplain of the lower 370 km of the Wisconsin River, Wisconsin, USA.Methods: Floodplain forest was sampled in 10 m × 20 m plots [n= 405) during summers of 1999 and 2000 in six 12‐ to 15‐km reaches.Results: Species observed most frequently were Fraxinus pennsylvanica, Acer saccharinum and Ulmus americana. Physiography (e.g. geographic province) and indicators of flooding regime (e.g. relative elevation and distance from main channel) were consistently important in predicting occurrence, community composition, and abundance of trees. Correspondence analysis revealed that flood‐tolerant and intolerant species segregated along the primary axis, and late‐successional species segregated from flood‐tolerant species along the secondary axis. Current landscape configuration only influenced species presence or abundance in forests that developed during recent decades. Land‐cover history was important for tree species presence and for the abundance of late‐successional species. Comparison of statistical models developed with and without soils data suggested that broad‐scale factors such as geographic province generally performed well.Conclusions: Physiography and indicators of flood regime are particularly useful for explaining floodplain forest structure and composition in floodplains with a relatively high proportion of natural cover types.