This is the same physical model as the 3D hillside — shown here as a side-on cross-section, which is actually the model's natural frame: the physics itself only ever depends on distance-down-slope and drop height, never on the depth (across-slope) axis the 3D scene uses for visual width.
Water gains speed falling down a slope the same way it would down a ramp: each vertical drop h converts to kinetic energy, giving a runoff velocity
V = √(2 · g · R · h)
where g is gravity and R is a hydraulic-radius friction factor (<1) from surface roughness. On a bare, unterraced slope the full height drops in one uninterrupted run, so h stays large the whole way down and V keeps climbing. Every terrace resets the drop to just its own riser height, then the flow crosses a near-flat platform where friction bleeds off most of the speed before the next riser — so final velocity at the toe falls roughly as 1/√(terraces).
Erosive power scales with kinetic energy, ∝ V², so halving velocity cuts transport-driven erosion to a quarter. Ground cover works differently: it intercepts raindrops before they strike bare soil, cutting splash detachment by an exponential canopy factor exp(-b·cover) (the same soil-loss-ratio form used in RUSLE), and it roughens the surface, nudging R down too.
Total modeled soil loss combines both channels: splash detachment (∝ effective rainfall reaching bare soil) plus transport by runoff (∝ V³, standard stream-power scaling). The readouts compare your combination of terraces, cover and rainfall against the same slope with zero terraces and zero cover — the bare-hillside baseline.
Drag the profile to pan and scroll/pinch to zoom — useful for inspecting a single riser closely when many terraces are stacked into a short slope.