Particle size-segregation and rheology of geophysical granular flows

Geophysical granular flows, such as snow avalanches, pyroclastic density currents, mudslides and debris flows, can be extremely hazardous to local populations, and understanding their complex behaviour remains an important challenge. This project aims to provide insight into these events by exploring different aspects in isolation, using a combination of mathematical theory, numerical simulations and small-scale experiments. Firstly, the effect of lateral confinement is examined by studying granular material moving in an inclined chute. This can have applications to natural releases flowing down confined valleys or conduits, and the relative simplicity of the geometry also provides a useful test case for new theoretical models. One such model is the recent depth-averaged μ(I)-rheology, which, because of the viscous terms introduced into the depth-averaged momentum balance, may be described as an intermediate approach between full constitutive laws and classical shallow-water-type equations for dense granular flows. Here, a generalisation of the new system to two spatial dimensions is described, and the resulting viscous equations are able to capture the cross-slope curvature of the downslope velocity profiles in steady uniform chute flows. This may be regarded as major progress compared to traditional hyperbolic models, which only admit constant velocity solutions. Particle size-segregation in geophysical granular flows is then investigated, which can cause important feedback on the overall bulk properties as it can lead to the development of regions with different frictional properties. A particularly striking example is segregation-induced 'finger' formation, where large particles are segregated to the flow surface and sheared to form a resistive coarse-rich front, which is unstable and spontaneously breaks down into a series of lobate structures. These travel both faster and further than one might anticipate. To model such segregation-mobility feedback effects, the depth-averaged μ(I)-rheology is extended to bidisperse flows by coupling with a depth-integrated model for size-segregation. The system of equations remains mathematically well-posed and is able to qualitatively capture finger formation, with the newly-introduced viscous terms controlling the characteristics of the leveed channels that develop. A more subtle segregation effect is studied in bidisperse roll waves, which form as small irregularities merge and coarsen as they move downslope, eventually growing into destructive large amplitude pulses. Experimental measurements show lateral, as well as vertical, segregation profiles, with the coarser grains accumulating at the fastest moving wave crests. The disturbances that form in mixtures with higher proportions of large particles grow more slowly, leading to smaller amplitude waves that travel at slower speeds, and the new coupled model predicts qualitatively similar behaviour. Finally, the influence of complex topography is investigated. A smooth two-dimensional bump is placed across the width of a chute, which, depending on the initial conditions, can lead to the formation of an airborne jet or granular shock at steady state. A simple depth-averaged model in a curvilinear coordinate system following the topography accurately captures both regimes, and represents a significant improvement on using an aligned Cartesian approach.