A two-layer model for predicting the vertical distribution of stream-wise velocity in open channel flow with submerged flexible vegetation is proposed. Using the predicted deflection height of the flexible vegetation determined via the large-deflection cantilever beam theory, the flow is vertically separated into a bottom vegetation layer and an upper free water layer, and corresponding momentum equations for each layer are formulated. In the bottom vegetation layer, the resistance caused by the deflected plants is calculated accounting for plant bending rather than adopting the existing resistance formula for erect rigid vegetation. For the upper free water layer, a new type of polynomial velocity distribution is suggested instead of the traditional logarithmic velocity distribution to obtain a zero velocity gradient at the water surface. To validate the proposed model, the published experimental data are employed.

Surface flow redistribution on flat ground from crusted bare soil to vegetated patches following intense rainfall events elevates plant available water above that provided by rainfall. The significance of this surface water redistribution to sustaining vegetation in arid and semiarid regions is undisputed. What is disputed is the quantity and spatial distribution of the redistributed water. In ecohydrological models, such nonuniform flows are described using the Saint‐Venant equation (SVE) subject to a Manning roughness coefficient closure. To explore these assumptions in the most idealized setting, flume experiments were conducted using rigid cylinders representing rigid vegetation with varying density. Flow was induced along the streamwise x direction by adjusting the free water surface height H(x) between the upstream and downstream boundaries mimicking the nonuniformity encountered in nature. In natural settings, such H(x) variations arise due to contrasts in infiltration capacity and ponded depths during storms. The measured H(x) values in the flume were interpreted using the SVE augmented with progressively elaborate approximations to the roughness representation. The simplest approximation employs a friction factor derived from a drag coefficient (C_d) for isolated cylinders in a locally (but not globally) uniform flow and upscaled using the rod density that was varied across experiments. Comparison between measured and modeled H(x) suggested that such a “naive” approach overpredicts H(x). Blockage was then incorporated into the SVE model calculations but resulted in underestimation of H(x). Biases in modeled H(x) suggest that C_d must be varying in x beyond what a local or bulk Reynolds number predicts. Inferred C_d(x) from the flume experiments exhibited a near‐parabolic shape most peaked in the densest canopy cases. The outcome of such C_d(x) variations is then summarized in a bulk resistance formulation that may be beneficial to modeling runon‐runoff processes on shallow slopes using SVE.