Classification of Stream, Hyperconcentrated, and Debris Flow Using Dimensional Analysis and Machine Learning

Extreme rainfall events in mountainous environments usually induce significant sediment runoff or mass movements-debris flows, hyperconcentrated flows and stream flows-that pose substantial threats to human life and infrastructure. However, understanding of the sediment transport mechanisms that control these torrent processes remains incomplete due to the lack of comprehensive field data. This study uses a unique field data set to investigate the characteristics of the transport mechanisms of different channelized sediment-laden flows. Results confirm that sediments in hyperconcentrated flows and stream flows are mainly supported by viscous shear and turbulent stresses, while grain collisional stresses dominate debris-flow dynamics. Lahars, a unique sediment transport process in volcanic environments, exhibit a wide range of transport mechanisms similar to those in the three different flow types. Furthermore, the Einstein number (dimensionless sediment flux) exhibits a power-law relationship with the dimensionless flow discharge. Machine learning is then used to draw boundaries in the Einstein number-dimensionless discharge scheme to classify one flow from the other and thereby aid in developing appropriate hazard assessments for torrential processes in mountainous and volcanic environments based on measurable hydrologic and geomorphic parameters. The proposed scheme provides a universal criterion that improves existing classification methods that depend solely on the sediment concentration for quantifying the runoff-to-debris flow transition relevant to landscape evolution studies and hazard assessments.