Quantifying floodplain heterogeneity with field observation, remote sensing, and landscape ecology: Methods and metrics

Floodplains provide numerous ecosystem services that depend on the spatial heterogeneity, or patchiness, of the floodplain. Direct and indirect human alterations of rivers can reduce floodplain heterogeneity and function, but relatively little is known of patterns of floodplain heterogeneity in natural, fully functional floodplains. We quantify floodplain heterogeneity at four sites in the United States with the objectives of (i) developing a method of combining field measurements and remote sensing data products to calculate integrative landscape-scale metrics of floodplain spatial heterogeneity and (ii) demonstrating which metrics from landscape ecology are likely to be useful for identifying qualities of natural floodplains, differentiating floodplains, and inferring processes, based on a case study of three prairie floodplains and one beaver-modified floodplain in the continental United States. We developed a new unsupervised classification workflow that combines field data, topography, and Sentinel-2A imagery to create classified floodplains for all four field sites that could be used to calculate heterogeneity metrics. We identified six heterogeneity metrics for characterizing natural floodplain heterogeneity: aggregation index, interspersion and juxtaposition index, largest patch index, patch density, percentage of like adjacencies, and Shannon's evenness index, and these metrics capture both intermetric (variation in spatial heterogeneity between the floodplains) and intrametric variation (variation in the patterns of the metrics). Results show that natural floodplains have high evenness and interspersion and juxtaposition of classes, and we attribute this to natural flow and sediment regimes driving channel migration, erosion, deposition, vegetation succession, and active beaver modifications. Colorado floodplains show higher aggregation and lower fragmentation than the Oklahoma floodplain. We attribute this to the greater incision and lower hydrologic variability at the Oklahoma site.