A multi-scale braided river substrate map – case study in the Rangitata

The accumulation of excess surficial and interstitial fine sediment in gravel-bed rivers may degrade ecosystem health and recreational value, alter bed mobility and change the natural character of a fluvial system. Our understanding of fine sediment in large rivers is limited by the effort and biases of conventional spot or visual monitoring methods. Here we present a new approach to mapping surficial river facies over 56 km of the Rangitata River, NZ through a fusion of dense airborne lidar and optical imagery, using semi-supervised machine learning. The hybrid dataset exploits multiscale information on the surface dimensionality, roughness and color. Detailed meter-resolution facies maps are generated for the exposed riverbed which covers >80% of the active area at low flow. The results are spatially integrated to generate longitudinal models of riverbed sediment cover at the system scale and offer a reproducible framework to monitor fine sediment and inform river process models.