Natural and man-made porous structures are ubiquitous in rivers and streams. Flow partition, i.e., the split of flow through and around the porous structures plays an important role for many applications such as backwater rise, flood control, bed shear stress amplification, sediment transport, and habitat suitability. A simple algebraic model based on first principles, i.e., conservations of mass, momentum and energy, is developed to predict flow partition using three dimensionless parameters: the Froude number (Fr), the channel opening fraction ( \(\beta\) ), and the drag coefficient ( \(C_d\) ). The model is validated against a comprehensive dataset generated from flume experiments and numerical simulations using SRH-2D, a solver for depth-averaged shallow water equations. The simple algebraic model shows good performance and applicability in predicting the flow partition fraction ( \(\alpha\) ). Its simplicity makes it especially useful for preliminary engineering evaluations. A machine learning-based analysis further quantified the importance of the three dimensionless parameters and interpreted their control on the flow partition. The analysis revealed that \(\beta\) and \(C_d\) are the most influential parameters in determining flow partition, while Fr plays a less important role. Despite the good performance, the simple algebraic model’s prediction degrades in edge cases involving extreme parameter values.
