Floods move across landscapes differently depending on how rough the land surface is—especially in areas next to rivers called floodplains. In many computer models used to predict river flow, these floodplain areas are treated in a very simple way, often without considering the effects of vegetation like trees or shrubs. This can lead to inaccurate predictions of how water spreads during storms. In this study, we tested whether using satellite measurements of vegetation structure could help improve flood predictions. We used a model that estimates how much vegetation slows down water in floodplains and applied it to a river basin in Oregon. We ran computer simulations of river flow using both the old method and the new vegetation-based method, and then compared the results to real streamflow data from dozens of sites. We found that the new approach improved accuracy at many locations, especially in areas that were steep, forested, and rainy. This shows that information about vegetation from satellites can help make flood prediction models more accurate in certain types of landscapes. The method could be used in other regions to help improve flood forecasting and water management.