Multi-Scale Modelling of Bed Load Transport

There are two main types of bed load transport models. Conventional bed load equations produce a bed load flux (which can be grain size specific) for a given set of flow conditions. They commonly require empirical parameters, which makes them unreliable and difficult to transfer between contrasting environments. The second model type is the mechanistic discrete element model (DEM). In these, each grain is individually modelled and its movement calculated by resolving the forces acting on it. However, the computational requirements of DEMs limits the number and size range of the modelled grains, therefore they have yet to be applied beyond the patch scale (cm to m). One alternative is a multi-scale approach whereby the DEM method is used to produce physically meaningful parameters for bed load transport equations. DEM beds of grains with a given grain size distribution can be created, and their surfaces interrogated to define distributions of parameters such as surface grain sizes, pivoting angles, grain exposure and roughness. A limited amount of field measurement to define the range of grain sizes would allow parameterisation of bed load transport equations, specific to an area. One limitation of many DEMs is that the grains are modelled as spheres. Grain shape is an important control on grain geometry, which affects grain entrainment. The grains in natural sediments are generally far from spherical. This could restrict the ability of a DEM to replicate natural sediment, suggesting that DEM derived parameters may also not be representative. In addition the way in which the DEM bed is created can also affect the surface properties. This paper presents research which illustrates the creation of DEM beds and measures their surface properties. The same properties are measured from gravel bar surfaces from the River Feshie, using techniques including photogrammetry, load cell measurements and stone counts. The different shape of the gravels grains, and their imbrication are found to affect the surface properties. The implications of this for DEM modelling are considered.