Point of View: A Call to Revisit Roughness Representation in Hydraulic Models

The principle of energy conservation in one-dimensional (1D) open channel flow modeling asserts that the specific energy at an upstream section equals the specific energy at a downstream section plus the energy losses incurred along the flow path. This formulation is a direct expression of the universal law of energy conservation and remains a fundamental truth in hydraulic modeling.

In gradually varied, steady, unidirectional flows, the primary energy loss is typically attributed to boundary friction—that is, the resistance caused by shear stresses exerted by the channel bed and walls. This energy loss is classically represented as the product of the friction slope and the distance between two cross-sections. The friction slope is generally computed using empirical equations, most notably Manning’s equation, wherein the Manning roughness coefficient (n) is used to encapsulate resistance effects caused by bed material, vegetation, channel irregularities, planform curvature, and obstructions.

However, a key concern in current 1D modeling practice is the treatment of Manning’s n along river or channel reaches. It is common to assign a representative roughness to a reach by averaging the n-values at the upstream and downstream cross-sections—often using arithmetic or geometric means. While this simplifies computations, it lacks physical justification.

To illustrate the flaw, consider a metal rod with its ends held at different temperatures. Averaging the temperatures at the ends does not provide an accurate representation of the temperature profile along the rod—especially if thermal properties vary spatially. Likewise, assuming that end-point Manning’s n values can represent the roughness across an entire reach ignores the spatial variability of boundary conditions and flow resistance mechanisms.

In reality, roughness is a spatially distributed property that varies significantly over short distances, especially in natural rivers and urban channels, where vegetation, sediment texture, channel geometry, and human-made structures introduce local variations. Therefore, applying a single, averaged roughness coefficient over an entire reach misrepresents actual hydraulic resistance and can lead to erroneous water surface profiles, velocity distributions, and energy gradients.

This issue is even more critical in two- and three-dimensional (2D/3D) models, where roughness effects must be applied along flow paths—streamlines or stream tubes—rather than merely at grid points. Frictional resistance acts continuously along the path of flow, and averaging roughness values across cells or boundaries without accounting for local variation can compromise model accuracy.

On the Argument That Calibration Solves the Problem

A common rebuttal is that the calibration process—adjusting Manning’s n to match observed water surface elevations—can compensate for roughness averaging. However, this is a mathematical workaround, not a physical solution.

Here’s why:

• Calibration "hides" errors: When calibration is applied to a reach with averaged roughness, the model adjusts the Manning’s n to compensate for all unresolved internal variability. This means the calibrated value no longer reflects the true physical roughness—it becomes a "lumped" or "effective" parameter. While this may produce a good water surface match, it results in unphysical Manning coefficients that can't be transferred or interpreted meaningfully elsewhere.
• Averaging introduces artificial nonlinearity: Because Manning's equation involves the roughness coefficient in a nonlinear way (usually as n2n^2 or higher depending on formulation), arithmetic averaging is mathematically incorrect. For example, averaging two Manning’s n values of 0.025 and 0.040 gives 0.0325, but the corresponding energy losses are not halfway between those from 0.025 and 0.040. The frictional impact increases disproportionately with n.
• The result is location-specific and unstable: An unphysical Manning’s n derived from calibration is only valid for the specific flow condition and geometry it was fitted to. If geometry or flow rates change, or if the model is used in a different season, this "calibrated" value becomes unreliable.
• Misinterpretation risk: Engineers and decision-makers may assume the calibrated n represents actual bed and bank conditions. This misinterpretation can lead to design errors, especially in flood management, sediment transport estimation, or structure sizing.

Improved Alternatives Exist

Fortunately, models have been developed to address this issue. One such example is the work of Hafez and Kady (2001), titled “A Self-Calibrated Water Surface Profile Computer Model for the Nile River in Egypt”, presented at the World Water & Environmental Resources Congress. In this model, the Manning roughness is treated as a spatially distributed property, and energy losses are computed based on the actual variation of roughness along the flow path, providing a physically meaningful and more stable representation.

Although this work was published in a conference in 2001 and did not receive the publicity it deserved, it presents a valuable and modern perspective on roughness handling that remains highly relevant today.

Call to Action

As modelers, engineers, and researchers, we must challenge outdated simplifications that no longer reflect the complexities of real-world hydraulic systems. The hydraulic modeling community is urged to rethink how roughness is represented in 1D, 2D, and 3D models, and move toward practices that are:

• Physically defensible
• Spatially detailed
• Empirically supported

Let us not confuse computational ease with physical accuracy, and strive for models that inform decisions with integrity. Let us refine our tools to better serve both engineering design and water resource management in an era where accuracy, adaptability, and reliability are more critical than ever.

Prof. Dr. Youssef Hafez Ph.D. in Civil Engineering, Colorado State University Specialist in Hydraulics, Numerical Modeling, and River Engineering 📧 [email protected]

.