Abstract. Floodplain and terrace features can provide information about current and past fluvial processes, including channel response to varying discharge and sediment flux, sediment storage, and the climatic or tectonic history of a catchment. Previous methods of identifying floodplain and terraces from digital elevation models (DEMs) tend to be semi-automated, requiring the input of independent datasets or manual editing by the user. In this study we present a new method of identifying floodplain and terrace features based on two thresholds: local gradient, and elevation compared to the nearest channel. These thresholds are calculated statistically from the DEM using quantile–quantile plots and do not need to be set manually for each landscape in question. We test our method against field-mapped floodplain initiation points, published flood hazard maps, and digitised terrace surfaces from seven field sites from the US and one field site from the UK. For each site, we use high-resolution DEMs derived from light detection and ranging (lidar) where available, as well as coarser resolution national datasets to test the sensitivity of our method to grid resolution. We find that our method is successful in extracting floodplain and terrace features compared to the field-mapped data from the range of landscapes and grid resolutions tested. The method is most accurate in areas where there is a contrast in slope and elevation between the feature of interest and the surrounding landscape, such as confined valley settings. Our method provides a new tool for rapidly and objectively identifying floodplain and terrace features on a landscape scale, with applications including flood risk mapping, reconstruction of landscape evolution, and quantification of sediment storage and routing.