The dynamic environment of natural river floodplains creates spatial heterogeneity that influences floodplain functions. Diverse human activities have homogenized natural floodplains and reduced their functions across many river networks in the temperate latitudes. Consequently, quantification of floodplain heterogeneity is needed to understand patterns of spatial heterogeneity on diverse floodplains and to inform floodplain restoration. We use a novel approach of spatially connecting field and remotely sensed data in order to interpret the output of, and build upon, a previous unsupervised classification workflow. We apply the method to three rivers in the US Pacific Northwest and the Altamaha River in the southeastern US and compare our results to a previous study. We find that field classifications, relative topography, and NDVI are useful for interpreting results from the unsupervised classification workflow. The interpretations are visually interesting, but we propose that it is the heterogeneity within the groups that is vital to floodplain functioning. Natural floodplains in the Pacific Northwest and coastal Southeast have moderate to high evenness, moderate to high intermixing, and moderate aggregation; and aggregation and evenness similar to rivers in Colorado and Oklahoma, USA, but lower intermixing. We attribute lower intermixing at the Altamaha River to slower rates of lateral channel migration, and lower intermixing at the Hoh River to the different hydrologic and sediment regimes and less stable braided planform. The results show that the larger rivers in this study (Altamaha, Hoh, and Sol Duc Rivers) have spatial heterogeneity similar to beaver-modified and shortgrass prairie rivers in Colorado, whereas the more inland and smaller river (Lookout Creek) has spatial heterogeneity similar to the tallgrass prairie site (Sand Creek). From the results of an ad hoc sensitivity analysis, we suggest using the highest spatial resolution topographic data available, using aerial imagery/mosaics from the same sensor, and removing largest patch index from the suite of comparable indices. The metrics reveal similarities and differences between rivers in the United States, and indicate that discernable trends may arise from a meta study comparing heterogeneity from more rivers across the country.
Read full abstract