River Morphodynamics

River morphodynamics sits between two kinds of evidence. One is geometric: digital elevation, channel masks, centerlines, boundary regions, and grain-size rasters. The other is dynamic: water surface, momentum, bed shear, sediment concentration, erosion, and deposition. Torrentem's surface hydrodynamics work is where those two views begin to share a mesh.

The current workflow uses the Ostium surface-hydrodynamics stack to turn a river reach into a triangular shallow-water problem. A DEM becomes surface nodes and elements. The channel mask decides which cells begin wet. Fields such as bed_elevation, bed_change, bed_porosity, chezy_coefficient, grain_size_d50, active_layer_thickness, and erodible_thickness are initialized beside depth, eta, momentum, and velocity.

From Reach Data To Boundary Forcing

A river simulation is only as useful as its boundary story. Torrentem determines mouth inflow from the case bundle's outlet geometry, using the outlet tangent to orient the incoming flow and the outlet width to choose a practical inlet half-width. Nodes near the river mouth are selected from the mesh boundary, filtered against wetted depth when possible, and assigned a water-surface stage condition.

That river-mouth stage can be a simple ramp or a hydrograph with spinup, peak, hold, and recession phases. The detail matters because environmental flow is rarely a permanent laboratory inflow. A storm pulse, backwater event, or staged mouth boundary changes not just the depth, but the locations where the bed is exposed to shear long enough to move.

The Honest State Of The Coupling

Torrentem's sediment transport module is intentionally conservative today. It is a GPU-only suspended sediment shell for operator-split scalar advection-diffusion. It can accept velocity from an existing GPU field view, including the interleaved two-phase Navier-Stokes layout, and it includes an optional bed-exchange path. But the current design keeps bed geometry fixed, and full morphodynamic bed motion remains disabled.

That is a good boundary to state plainly. The scaffolding is there: bed fields, porosity, grain size, active layers, erodible thickness, bedload magnitude, Shields parameter, and sediment diagnostics. The moving-bed feedback loop is the next layer, not something to imply before the numerical contracts are ready.