You are on page 1of 213

This PDF is available at http://nap.nationalacademies.

org/27730

Selection and Application of Manning s


Roughness Values in Two-Dimensional
Hydraulic Models (2024)

DETAILS
214 pages | 8.5 x 11 | PAPERBACK
ISBN 978-0-309-70944-6 | DOI 10.17226/27730

CONTRIBUTORS
Xiaofeng Liu, Ali Mahdavi Mazdeh, Lyle W. Zevenbergen, Casey M. Kramer; National
Cooperative Highway Research Program; Transportation Research Board; National
BUY THIS BOOK Academies of Sciences, Engineering, and Medicine

FIND RELATED TITLES SUGGESTED CITATION


National Academies of Sciences, Engineering, and Medicine. 2024. Selection and
Application of Manning s Roughness Values in Two-Dimensional Hydraulic Models.
Washington, DC: The National Academies Press. https://doi.org/10.17226/27730.

Visit the National Academies Press at nap.edu and login or register to get:
– Access to free PDF downloads of thousands of publications
– 10% off the price of print publications
– Email or social media notifications of new titles related to your interests
– Special offers and discounts

All downloadable National Academies titles are free to be used for personal and/or non-commercial
academic use. Users may also freely post links to our titles on this website; non-commercial academic
users are encouraged to link to the version on this website rather than distribute a downloaded PDF
to ensure that all users are accessing the latest authoritative version of the work. All other uses require
written permission. (Request Permission)

This PDF is protected by copyright and owned by the National Academy of Sciences; unless otherwise
indicated, the National Academy of Sciences retains copyright to all materials in this PDF with all rights
reserved.
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

N AT I O N A L C O O P E R AT I V E H I G H W AY R E S E A R C H P R O G R A M

NCHRP RESEARCH REPORT 1077


Selection and Application of Manning’s
Roughness Values in Two-Dimensional
Hydraulic Models

Xiaofeng Liu
Ali Mahdavi Mazdeh
The Pennsylvania State University
University Park, PA

Lyle W. Zevenbergen
Lyle W. Zevenbergen, LLC
Fort Collins, CO

Casey M. Kramer
Natural Waters, LLC
Olympia, WA

Subscriber Categories
Highways • Bridges and Other Structures • Hydraulics and Hydrology

Research sponsored by the American Association of State Highway and Transportation Officials
in cooperation with the Federal Highway Administration

2024

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

NATIONAL COOPERATIVE HIGHWAY NCHRP RESEARCH REPORT 1077


RESEARCH PROGRAM
Systematic, well-designed, and implementable research is the most Project 24-49
effective way to solve many problems facing state departments of ISSN 2572-3766 (Print)
transportation (DOTs) administrators and engineers. Often, highway ISSN 2572-3774 (Online)
problems are of local or regional interest and can best be studied by ISBN 978-0-309-70944-6
state DOTs individually or in cooperation with their state universities Library of Congress Control Number 2024931079
and others. However, the accelerating growth of highway transporta-
© 2024 by the National Academy of Sciences. National Academies of
tion results in increasingly complex problems of wide interest to high-
Sciences, Engineering, and Medicine and the graphical logo are trade-
way authorities. These problems are best studied through a coordinated
marks of the National Academy of Sciences. All rights reserved.
program of cooperative research.
Recognizing this need, the leadership of the American Association
of State Highway and Transportation Officials (AASHTO) in 1962 ini-
tiated an objective national highway research program using modern COPYRIGHT INFORMATION
scientific techniques—the National Cooperative Highway Research Authors herein are responsible for the authenticity of their materials and for obtaining
Program (NCHRP). NCHRP is supported on a continuing basis by written permissions from publishers or persons who own the copyright to any previously
funds from participating member states of AASHTO and receives the published or copyrighted material used herein.
full cooperation and support of the Federal Highway Administration Cooperative Research Programs (CRP) grants permission to reproduce material in this
(FHWA), United States Department of Transportation, under Agree- publication for classroom and not-for-profit purposes. Permission is given with the
ment No. 693JJ31950003. understanding that none of the material will be used to imply TRB, AASHTO, APTA, FAA,
FHWA, FTA, GHSA, or NHTSA endorsement of a particular product, method, or practice.
The Transportation Research Board (TRB) of the National Academies
It is expected that those reproducing the material in this document for educational and
of Sciences, Engineering, and Medicine was requested by AASHTO to not-for-profit uses will give appropriate acknowledgment of the source of any reprinted or
administer the research program because of TRB’s recognized objectivity reproduced material. For other uses of the material, request permission from CRP.
and understanding of modern research practices. TRB is uniquely suited
for this purpose for many reasons: TRB maintains an extensive com-
mittee structure from which authorities on any highway transportation
NOTICE
subject may be drawn; TRB possesses avenues of communications and
cooperation with federal, state, and local governmental agencies, univer- The research report was reviewed by the technical panel and accepted for publication
according to procedures established and overseen by the Transportation Research Board
sities, and industry; TRB’s relationship to the National Academies is an
and approved by the National Academies of Sciences, Engineering, and Medicine.
insurance of objectivity; and TRB maintains a full-time staff of special-
ists in highway transportation matters to bring the findings of research The opinions and conclusions expressed or implied in this report are those of the
researchers who performed the research and are not necessarily those of the Transportation
directly to those in a position to use them. Research Board; the National Academies of Sciences, Engineering, and Medicine; the
The program is developed on the basis of research needs iden- FHWA; or the program sponsors.
tified by chief administrators and other staff of the highway and
The Transportation Research Board does not develop, issue, or publish standards or spec-
transportation departments, by committees of AASHTO, and by ifications. The Transportation Research Board manages applied research projects which
the FHWA. Topics of the highest merit are selected by the AASHTO provide the scientific foundation that may be used by Transportation Research Board
Special Committee on Research and Innovation (R&I), and each year sponsors, industry associations, or other organizations as the basis for revised practices,
R&I’s recommendations are proposed to the AASHTO Board of Direc- procedures, or specifications.
tors and the National Academies. Research projects to address these The Transportation Research Board; the National Academies of Sciences, Engineering, and
topics are defined by NCHRP, and qualified research agencies are Medicine; and the sponsors of the National Cooperative Highway Research Program do not
endorse products or manufacturers. Trade or manufacturers’ names or logos appear herein
selected from submitted proposals. Administration and surveillance of
solely because they are considered essential to the object of the report.
research contracts are the responsibilities of the National Academies
and TRB.
The needs for highway research are many, and NCHRP can make
significant contributions to solving highway transportation problems
of mutual concern to many responsible groups. The program, however,
is intended to complement, rather than to substitute for or duplicate,
other highway research programs.

Published research reports of the

NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM


are available from

Transportation Research Board


Business Office
500 Fifth Street, NW
Washington, DC 20001

and can be ordered through the Internet by going to


https://www.mytrb.org/MyTRB/Store/default.aspx
Printed in the United States of America

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

The National Academy of Sciences was established in 1863 by an Act of Congress, signed by President Lincoln, as a private, non-
governmental institution to advise the nation on issues related to science and technology. Members are elected by their peers for
outstanding contributions to research. Dr. Marcia McNutt is president.

The National Academy of Engineering was established in 1964 under the charter of the National Academy of Sciences to bring the
practices of engineering to advising the nation. Members are elected by their peers for extraordinary contributions to engineering.
Dr. John L. Anderson is president.

The National Academy of Medicine (formerly the Institute of Medicine) was established in 1970 under the charter of the National
Academy of Sciences to advise the nation on medical and health issues. Members are elected by their peers for distinguished contributions
to medicine and health. Dr. Victor J. Dzau is president.

The three Academies work together as the National Academies of Sciences, Engineering, and Medicine to provide independent,
objective analysis and advice to the nation and conduct other activities to solve complex problems and inform public policy decisions.
The National Academies also encourage education and research, recognize outstanding contributions to knowledge, and increase
public understanding in matters of science, engineering, and medicine.

Learn more about the National Academies of Sciences, Engineering, and Medicine at www.nationalacademies.org.

The Transportation Research Board is one of seven major program divisions of the National Academies of Sciences, Engineering, and
Medicine. The mission of the Transportation Research Board is to mobilize expertise, experience, and knowledge to anticipate and solve
complex transportation-related challenges. The Board’s varied activities annually engage about 8,500 engineers, scientists, and other
transportation researchers and practitioners from the public and private sectors and academia, all of whom contribute their expertise in the
public interest. The program is supported by state transportation departments, federal agencies including the component administrations
of the U.S. Department of Transportation, and other organizations and individuals interested in the development of transportation.

Learn more about the Transportation Research Board at www.TRB.org.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

COOPERATIVE RESEARCH PROGRAMS

CRP STAFF FOR NCHRP RESEARCH REPORT 1077


Waseem Dekelbab, Deputy Director, Cooperative Research Programs
Ahmad Abu-Hawash, Senior Program Officer
Sheila A. Moore, Program Associate
Natalie Barnes, Director of Publications
Heather DiAngelis, Associate Director of Publications
Kristin C. Sawyer, Editor

NCHRP PROJECT 24-49 PANEL


Field of Soils and Geology—Area of Foundations and Scour
Solomon T. Woldeamlak, Minnesota Department of Transportation, Oakdale, MN (Chair)
Luke Assink, Washington State Department of Transportation, Olympia, WA
Jeffrey A. DeGraff, Vermont Agency of Transportation, Barre, VT
Kathryn L. Eagan, BSC Group, Inc., Boston, MA
Abiot Gemechu, Virginia Department of Transportation, Richmond, VA
Abderrahmane Maamar-Tayeb, Texas Department of Transportation, Austin, TX
Jose Vasconcelos, Auburn University, Auburn, AL
Scott Hogan, FHWA Liaison
Patricia Bush, AASHTO Liaison

AUTHOR ACKNOWLEDGMENTS
The research reported herein was performed under NCHRP Project 24-49, “Guidelines for Selection
and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models,” by the Depart-
ment of Civil and Environmental Engineering at The Pennsylvania State University (PSU). PSU was the
contractor for this study.
Dr. Xiaofeng Liu, P.E., an associate professor of civil engineering at PSU, was the project director and prin-
cipal investigator. The other authors of this report are Dr. Ali Mahdavi Mazdeh, a postdoctoral researcher at
PSU; Lyle W. Zevenbergen from Lyle W. Zevenbergen, LLC; and Casey M. Kramer from Natural Waters, LLC.
The work was done under the general supervision of Dr. Xiaofeng Liu. Mr. Zevenbergen and Mr. Kramer
worked as subcontractors to PSU on this project.
Scott Hogan from FHWA provided the Iowa River 2008 flooding case data. Scott Steele at Pennoni
Associates, Inc., provided the Delaware Bridge 1-242 case data. Aquaveo provided free licenses of SMS.
Reinaldo Garcia from Hydronia provided free licenses of RiverFlow2D. Dr. Yong Lai at the U.S. Bureau
of Reclamation and Gary Brunner (formerly at U.S. Army Corps of Engineers, now at HDR) provided
assistance and comments. Hassan Ismail, a former postdoctoral research associate in Dr. Liu’s group,
participated in the early stage of this project and proposed the research plan. The research team is grateful
to all of them.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

FOREWORD

By Ahmad Abu-Hawash
Staff Officer
Transportation Research Board

NCHRP Research Report 1077: Selection and Application of Manning’s Roughness Values in
Two-Dimensional Hydraulic Models presents guidelines for the selection and application of
Manning’s roughness values for two-dimensional (2D) hydraulic models. The development
of the guidelines involved extensive modeling and analysis. The research findings will be
of interest to hydraulic engineers at state departments of transportation (DOTs) and other
public agencies.

Measures to improve the accuracy, reliability, and consistency in selecting roughness values
for the application of 2D hydraulic models at highway crossings and in transportation cor-
ridors will represent a major step forward in the evolving state of practice. Some state DOTs
have readily embraced the application of 2D models for transportation facility planning,
analysis, and design; other states have held back and taken a more wait-and-see approach as
the tools and techniques have continued to advance. At present, whether a one-dimensional
(1D) or a 2D modeling approach is taken, inconsistencies can exist from state to state, and even
within states at the various district levels, on how to develop accurate and reliable hydraulic
models. The fundamental issue is the selection and assignment of Manning’s roughness values
for a variety of channel types and land uses. Manning’s roughness values, along with channel
and floodplain geometry (i.e., topographic and hydrographic surface data), are recognized
to be the two most important components for developing and calibrating hydraulic models.
Identifying feasible, cost-effective, and consistent guidelines to improve the state of the prac-
tice in selecting roughness values for hydraulic modeling is highly desirable. The potential
payoff to owners of transportation assets, in particular roadways, bridges, and culverts, is
significant if the accuracy and reliability of hydraulic models can be improved. Development
of detailed guidelines on hydraulic modeling was needed to improve the state of practice on
a national level and in the rapidly advancing field of 2D hydraulic models.
Under NCHRP Project 24-49, “Guidelines for Selection and Application of Manning’s
Roughness Values in Two-Dimensional Hydraulic Models,” Penn State University was asked
to (1) develop guidelines for the selection and application of Manning’s roughness values
in 2D hydraulic models for transportation-related riverine settings and (2) propose draft
language to the AASHTO Drainage Manual, the Federal Highway Administration (FHWA)
Hydraulic Design Series Number 7: Hydraulic Design of Safe Bridges, and the FHWA’s Every
Day Counts (EDC) Two-Dimensional Hydraulic Modeling for Highways in the River Environ-
ment: Reference Document.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CONTENTS

1 Summary

P A R T I  Literature Review and Identifying Knowledge


Gaps
11 Chapter 1 Literature Review and Knowledge Gaps
11 1.1 Introduction
11 1.2 Flow Resistance
14 1.3 Manning’s n
22 1.4 Current Guidance on Selection of Manning’s n
22 1.5 Synthesis of Literature Review and Knowledge Gaps

25 Chapter 2 Analytical Program to Achieve Research Objectives


25 2.1 Introduction
26 2.2 Analytical Program

P A R T I I Execution of Analytical Program and Results


33 Chapter 3 Verification of Existing n Values for Surface
Roughness and Spatial Effects of n
33 3.1 Comparison of Manning’s n in Different Models
37 3.2 Effect of Relative n Values in Composite Channels

44 Chapter 4 Flow Resistance Factors, Required Conditions


for Resolving a Feature, and Effect of Model Choice
44 4.1 Channel Obstruction
79 4.2 Bedforms
101 4.3 Channel Irregularities
103 4.4 Stage and Discharge Relations and Available Guidelines
104 4.5 Vegetation
107 4.6 Summary of Simulation Cases and Examples
114 4.7 How to Obtain Total Manning’s n

117 Chapter 5 Development of 2D Modeling Best Practices


117 5.1 Overview of 2D Modeling Best Practices
117 5.2 Best Practices in 2D Hydraulics Modeling

127 Chapter 6 Development of a Decision-Making Process


for Selecting Manning’s n for 2D Models
127 6.1 Overview
127 6.2 Decision-Making Process

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

135 Chapter 7 Verification and Demonstration


of Proposed Guidelines
135 7.1 Iowa River 2008 Flooding Case
152 7.2 Susitna River Case
168 7.3 Delaware Bridge 1-242 Case

179 References
183 Appendix A Evaluation of Solvers in HEC-RAS 2D
191 Appendix B Detailed Evaluation of Hole-in-Mesh Method
199 Appendix C Manning’s n Sensitivity Analysis

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

SUMMARY

Selection and Application


of Manning’s Roughness
Values in Two-Dimensional
Hydraulic Models
NCHRP Project 24-49, “Guidelines for Selection and Application of Manning’s Rough-
ness Values in Two-Dimensional Hydraulic Models,” provides detailed information on
Manning’s n value selection and specification for two-dimensional (2D) hydraulics models.
The objectives of this research were to (1) assess and document the state of the practice con-
cerning the selection of Manning’s n values for 2D hydraulic modeling applications and (2)
develop recommendations for selecting Manning’s n values for 2D modeling.
This report, NCHRP Research Report 1077: Selection and Application of Manning’s Rough-
ness Values in Two-Dimensional Hydraulic Models, documents the work completed in this
project and is organized following the structure of tasks and subtasks in the four phases of
the project’s work scope. The first phase was to conduct a complete literature review on flow
resistance, Manning’s n, and roughness factors. The information was synthesized to identify
knowledge gaps and create an analytical program for the second phase of the project. In the
second phase, the analytical program was used to assess the roughness values in 2D hydraulic
models, compare one-dimensional (1D) and 2D hydraulic models, and fill the knowledge
gaps identified in the first phase. In the third phase, all the information was integrated into
a step-by-step set of guidelines for selecting Manning’s n for 2D models (see Chapter 6). The
fourth phase was to summarize all work and findings into the final report.
The specification and calibration of Manning’s n should begin with a model that is appro-
priately developed using good modeling practices. If a model is not properly developed,
contains errors, or otherwise does not follow best practices, the simulation results will be
inaccurate or wrong. If this is the case, there is no point in wasting time and effort on deter-
mining Manning’s n. Best practices cover all aspects of 2D hydraulics modeling; because
of their critical importance to Manning’s n, a summary of best practices has been provided
(see Chapter 5).
This summary is included to illustrate the extensive work done on this project, present
key recommendations, and highlight general findings.

General Findings
• It is important to have clear definitions of Manning’s n in different contexts. Much confu-
sion in practice stems from the lack of a clear definition. In the physical world, Manning’s
n is a parameter to quantify the roughness and flow resistance induced by different factors,
for example, bedforms and vegetation. Hence, it is called the real-world total Manning’s n.
In the numerical modeling world, the total Manning’s n is composed of two parts: resolved

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

2   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Manning’s n and unresolved Manning’s n. In a 2D model, the terrain surface and mesh, if
appropriately developed, partially resolve the roughness, which can be quantified as the
resolved Manning’s n. The remaining part, the unresolved Manning’s n, is what a modeler
needs to specify in 2D models. How much is unresolved depends on factors like terrain
data and mesh resolutions, numerical schemes, and turbulence model. Collectively, the
resolved and unresolved Manning’s n values reflect the total Manning’s n in the physical
world. Thus, the Manning’s n specified in 2D models has a mixed meaning of physics
(unresolved flow resistance) and numerics (terrain and mesh resolution dependent).
• Based on flow physics, the relationship among total, resolved, and unresolved Manning’s n
should follow the law of the square root of the summation of squares, in other words,
vector addition:

n = n 2resolved + n 2unresolved (S-1)

where n is the total Manning’s coefficient.


This ensures the total resistance is correct. The flow resistance is proportional to the
square of Manning’s n.
For a particular area in a model domain, the total Manning’s n quantifies the total flow
resistance induced by all roughness factors. The law previously mentioned should also
apply when combining individual Manning’s n values into the total:

n = n 21 + n 22 + n 33 + g (S-2)

where n1, n2, and n3 are Manning’s n corresponding to different roughness factors, such as
background roughness, bedform, and vegetation. In the literature, a linear combination
is mostly used:

n = n1 + n2 + n3 + . . . (S-3)

This is a historical formula widely used in practice; however, as described in this report,
it lacks sound physical bases, and its use should be phased out.
• It is also necessary to emphasize the importance of clear definitions for geometrically and
hydraulically resolving hydraulic features in a domain.
“Geometrically resolved” means the geometry of a feature is resolved by terrain surface
and mesh. “Hydraulically resolved” means the hydraulic effects of a feature are accurately
captured by a computer model. Geometrically resolved is a necessary but insufficient
condition for hydraulically resolved. For example, a circular bridge pier can be captured
by a circular hole in a mesh. In this case, the pier is geometrically resolved. However, this
does not necessarily mean the hydraulic effects of the pier, such as the backwater effect
upstream and wake and shear layers downstream, are fully resolved. The full resolution
also depends on other factors, such as proper mesh resolution around the feature.
• Different 2D models may need different criteria for geometrically and hydraulically
resolving hydraulic features.
For example, in the Hydrologic Engineering Center-River Analysis System (HEC-RAS)
2D subgrid terrain treatment is implemented, so the model uses more terrain information
than those models that do not have such treatment. However, a modeler cannot blindly
rely on the subgrid terrain treatment with an extremely coarse mesh. The subgrid terrain
is good at constructing a volume–stage relationship. The cell faces still need to be aligned
around an obstruction or hydraulic control with breaklines to properly resolve them.
• Sedimentation and River Hydraulics-2D (SRH-2D) v3.3.0 and HEC-RAS 2D v6.1 were
compared with HEC-RAS 1D v6.1 and a simple backwater curve solver for some M1 and

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Summary  3

M2 profiles in a rectangular channel. The research team found they produce similar
results (with only some small differences due to numerics). Thus, SRH-2D and HEC-RAS
2D were implemented correctly (at least for these simple cases). The difference in their
treatment of subterrain and Manning’s n does not matter for these simple cases because
the bottom is flat.
• The research team performed informal sensitivity analysis on Manning’s n and found that
for some cases the results are not sensitive to Manning’s n but in other cases they are. The
sensitivity analysis shows that the lower the water depth, the more sensitive the results
are to Manning’s n. Moreover, the effect of Manning’s n takes distance to become evident,
especially for profiles like M1. For example, for a short modeling domain where the flow
is deep and approximately follows the M1 curve, the 2D results (velocity and water depth)
may not be sensitive to Manning’s n. The implications of these sensitivity analyses are
described and demonstrated in Section 3.1 and Appendix C in this report.
• During phase 2 of the project, HEC-RAS version 6.1 was released. By the time this report
was being written, several newer versions of HEC-RAS 6.x already had been released.
After careful consideration and consultation with NCHRP, the team decided to use
HEC-RAS 6.1 instead of HEC-RAS 5.0.7 (which was proposed originally). HEC-RAS 6.1
and later versions offer two options for solving the full dynamic equations [the Eulerian–
Lagrangian method (ELM) and the Eulerian method (EM)]. ELM is the same in both
HEC-RAS 5.0.7 and 6.1; EM is a new solver option introduced in HEC-RAS 6.1. A full
review and comparison have been performed on different solver options and versions of
HEC-RAS (see Appendix A). In the research team’s tests, the new EM method needed
an extremely small time step for local hydraulics, which might be too expensive for prac-
tical use. Based on this, ELM was recommended over EM in HEC-RAS 2D.

Flow Resistance Factors


Based on the proposed analytical program, the research team did extensive hydraulic
modeling to investigate different flow resistance factors, for example, bedforms, obstruc-
tions, and channel irregularities. The number of cases performed is much more than
what was originally planned. The team decided that a significantly larger number of cases
needed to be simulated to make the guidelines and recommendations complete and more
generalizable.
The team determined how each flow resistance factor affects Manning’s n value and when
they are geometrically and hydraulically resolved. For each roughness factor, they also com-
pared different 2D models. The team used SRH-2D as a base model for the whole project.
RiverFlow2D, HEC-RAS 2D, and HEC-RAS 1D were used for comparison. SMS and the
open-source Python package pyHMT2D (https://github.com/psu-efd/pyHMT2D) were
used to develop similar meshes and model domains to ensure a fair comparison. In other
words, for simple cases, the meshes were the same for different 2D models to the extent
possible.

Bedforms
• The research team ran several scenarios to study bedform’s effect on Manning’s n. Based
on the results, a methodology is proposed to define unresolved and resolved Manning’s n,
though this method is general and not limited to bedforms. Manning’s n specified in
2D models should account only for the unresolved roughness features. The new method-
ology provides a way to estimate the unresolved Manning’s n, which is the ultimate goal
of this project (see Section 4.2).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

4   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

• The team hypothesized that in HEC-RAS 2D the subgrid terrain treatment can capture
more roughness features; therefore, it resolves more roughness. Consequently, the unre-
solved Manning’s n should be smaller in comparison with other 2D models that do not
implement such a method. However, based on the research team’s results, that hypothesis
is not necessarily true. The reason is that the subgrid terrain treatment in HEC-RAS 2D is
applied to only some, not all, terms in the governing equations. In particular, the bed slope
in the gravity term, which is the driving force for open-channel flow, does not benefit from
the subgrid terrain treatment. Because of the nonlinear nature of the governing shallow
water equations, it is hard to draw a definitive conclusion on the effect of subgrid terrain
treatment on Manning’s n.
One significant conclusion, however, is that the subgrid terrain treatment in HEC-RAS
2D should not be considered as resolving features geometrically or hydraulically. This
should be done using mesh refinement.
• The research team also performed an analysis on the effect of mesh resolution on Manning’s n.
Theoretically, as the mesh gets coarser, it will increasingly miss more terrain features; there-
fore, the resolved Manning’s n should decrease. However, the team’s results do not consistently
show this trend—there are cases that contradict it. A given mesh samples the terrain data
and represents the underlying terrain with the sampled elevations. This sampled eleva-
tion field at mesh points may misrepresent the terrain, which could influence Manning’s n
in ways either increasing or decreasing the roughness. In general, if a coarse mesh is not
well designed, it is likely to reduce the height of elevated terrain features such as roads and
berms. On the other hand, it is likely to increase the height of low terrain features such as
channels (see Section 7.2).
• In practice, it may often be the case that small features like bedforms are not represented
in terrain data and mesh. Therefore, their flow resistance is not resolved at all.

Obstructions and Irregularities


• Obstructions (e.g., bridge piers) can be modeled with different approaches in different
2D models. Examples of approaches include hole-in-mesh, increased Manning’s n coeffi-
cient, added drag force, unassigned or disabled cells, and elevated terrain. The advantages
and applicability of these approaches were summarized and compared.
• For the hole-in-mesh approach, the team proposes a quantitative metric for geometri-
cally resolving an obstruction. This metric is the ratio of cell size to obstruction width
(CSO). Generally, the CSO should be at least less than 0.5 to geometrically resolve
the obstruction. This recommendation is in line with those in Robinson et al. (2019,
Figure A-29). A hexagon meets this criterion as representing a circular pier (π/6 is
approximately 0.5), and an octagon provides a better geometric resolution of a circular pier
(π/8 is approximately 0.4).
• Mesh refinement should also consider the width of obstructions compared to the over-
all channel width. This is required to capture the interactions between obstructions and
channel banks.
• Yarnell’s equation, based on the conservation laws and calibrated with measurement data,
was used to evaluate the hole-in-mesh method, which has been deemed the most accurate
method. The results showed that this method, using a geometrically and hydraulically
resolved mesh, can capture the hydrodynamics around piers and predict correctly the
amount of backwater effect.
• The unassigned and hole-in-mesh methods have the same results. The unassigned method
is a special method implemented in SMS for some 2D models in which the selected cells
do not participate in the simulation. Despite the similarity, their velocity results in SMS
appear different. This seems to be only a visual effect due to the postprocessing procedure

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Summary  5

for estimating velocity at nodes from the cell-centered solution. In SRH-2D, for the
hole-in-mesh method, SMS treats a wall with zero velocity. However, for the unassigned
method, a zero gradient is applied; in other words, the velocity of a node is assigned the
cell-centered velocity.
• In the elevated terrain method, in which an obstruction is incorporated directly into ter-
rain, a refined mesh is required around the obstruction. This method may cause near-
vertical cells around the edge of an obstruction, which may cause numerical stability
issues. This method is recommended only when the hole-in-mesh and unassigned mesh
methods are not possible.
• When it is not feasible to have a fine-enough mesh to resolve an obstruction and the
local hydrodynamics is not a priority for the modeling purpose, increasing Manning’s n
coefficient is a much less costly alternative. The increased Manning’s n will increase the
flow resistance, which is caused by the obstruction, and can be estimated by the equation
in Chapter 6 of the Hydraulic Design Series (HDS) No. 7 (Zevenbergen et al. 2012). The
increased Manning’s n method and the added drag-force method are similar; however,
their mathematical formulas are different.
The research team identified a drawback of the two methods in most 2D models regard-
ing obstructions. The drawback stems from the implementation of the two methods.
Details of the drawback are described in this report, and recommendations for correct
2D model implementation are provided (see Section 4.1.1.2).
With current 2D models, the following should be considered in using the increased
Manning’s n method (similarly for the added drag-force method):
– In preparing the material coverage for high Manning’s n at the obstruction location,
the modeler should consider which cells will be associated with it because the total area
of these elements is used later in calculating the needed Manning’s n.
– For pile or trestle piers that consist of several small piers over the channel width, the
calculated effective Manning’s ne using Equation 4-4 should spread over the width of
the channel. If the river cross section at the obstruction location has different n values
or the obstructions are in a limited area of the channel width, the team recommends
spreading their effect over the affected area only instead of the whole channel width.
• Some models (e.g., SRH-2D and RiverFlow2D) offer the additional drag-force method.
Because of the drawback described in Section 4.1.1.2, these models currently cannot correctly
reproduce the hydrodynamics around an obstruction when the mesh is fine and the footprint
of the obstruction covers more than one cell. In such cases, a much larger drag coefficient (CD)
needs to be used. On the other hand, when the mesh is coarse and the obstruction is much
smaller than the mesh cell that contains it, the problem is less severe.
• River channel irregularities cause head loss and flow resistance; however, 1D and 2D models
treat them differently. In 1D models, such as HEC-RAS 1D, contraction–expansion coeffi-
cients are used to model the head loss due to the variations in channel cross sections. In
2D models, a properly designed mesh and accurate terrain data should automatically capture
channel irregularities.
• One important question for this project was how to change Manning’s n when a 1D model is
replaced by a 2D model.
– For cases in which flow contraction and expansion are significant, the team compared
1D and 2D model results and found it impossible to provide a definitive answer. The
reason is that in 1D models the channel irregularity is captured by the combination of
contraction–expansion coefficients and Manning’s n. The team’s recommendation is
to use the same Manning’s n values in 1D models as a starting point for calibration (see
Section 4.3 for more information).
– For cases where there is significant shear between the main channel and its flood-
plains, for example, induced by drastically different roughness conditions, the value of

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

6   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Manning’s n in 1D models should be larger than that used in 2D models. The reason
is that 1D models cannot simulate the shear between the main channel and its flood-
plains. The shear, turbulence, and head loss all must be parameterized and added to
Manning’s n.

Vegetation
In practice, vegetation is typically not in terrain data. Therefore, its roughness must be
considered in the unresolved Manning’s n. The team synthesized and evaluated many for-
mulas for predicting the flow resistance of rigid and flexible vegetation.
For rigid vegetation, the method in Luhar and Nepf (2013) is recommended because it
captures the strong nonlinear behavior of Manning’s n near the transition between emerged
and submerged conditions (Shields et al. 2017).
For flexible vegetation, the team evaluated the methods in Järvelä (2004) and Whittaker et al.
(2015), which are formulated in more general forms compared with others (see Section 4.5).
When vegetation effects and surface roughness are evaluated separately, vector addition should
be used rather than scalar addition for the calculation of the total Manning’s n.

Best Practices for 2D Modeling and Guidelines


for Manning’s n
Two-dimensional modeling should follow best practices before considering how to specify
Manning’s n. If a 2D model is developed not following best practices or with errors, there is
no point in determining its Manning’s n values. Currently, there is limited documentation
on best practices for 2D hydraulics modeling. To fill this gap, the research team provided
detailed best practices covering the entire process of a typical 2D modeling project (see
Chapter 5).
The team then provided a step-by-step decision-making process for selecting Manning’s
n in 2D models based on the results obtained through the research for this project. Suffi-
cient, but not overwhelming, details are provided (see Chapter 6).

Proposed Modifications to AASHTO and FHWA Manuals


Technical Brief (TechBrief) is an FHWA publication series for the fast dissemination
of research results. Several of the key existing FHWA hydraulics-related publications are
near the end of updating. Because of time constraints, it is not possible to incorporate the
team’s research results in the next version of, for example, Hydraulic Engineering Circular
(HEC)-18 or HEC-20, before the end of this project. Thus, TechBrief is the appropriate
way for the hydraulic engineering community to benefit from the research results of this
project.
The TechBrief FHWA will publish about this project will provide background on Manning’s
roughness (Manning’s n) and information on selecting and applying values for hydraulic
models. This includes distinctions between 1D and 2D hydraulic models and practices
for developing 1D and 2D models. Sound modeling practices are an important aspect of
hydraulic modeling. Without sound modeling practices, which include accurately repre-
senting terrain, bathymetry, hydraulic controls, features, and obstructions and providing
sufficient model resolutions to hydraulically resolve the flow fields, the exercise of selecting
Manning’s n values is fraught with uncertainty and error.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Summary  7

The TechBrief will provide updated and improved information for the following:
• HEC No. 20, Stream Stability at Highway Structures, 4th edition (Legasse et al. 2012).
• HEC No. 18, Evaluating Scour at Bridges, 5th edition (Arneson et al. 2012).
• HDS No. 7, Hydraulic Design of Safe Bridges, 2nd edition (FHWA 2023).
• Two-Dimensional Hydraulic Modeling for Highways in the River Environment—Reference
Document (Robinson et al. 2019).
Following are the key takeaways from the forthcoming TechBrief:
• The total flow resistance of a hydraulic model includes resistance resolved by including
features in the model terrain and mesh, and resistance unresolved by the model terrain
and mesh. It is the unresolved portion of Manning’s n that is the correct input value for
the model.
• When components of Manning’s n are combined, scalar addition is not physically or
mathematically correct. The total Manning’s n is the square root of the sum of the squares
of the component values. This relationship also pertains to the resolved and unresolved
portions of Manning’s n that compose the total value.
• Although there are several reasons that 1D models are expected to have higher input
Manning’s n values than 2D models, there is no specific relationship for the relative
difference.
• Modelers should strive to geometrically resolve all important bathymetric and terrain fea-
tures and to hydraulically resolve the flow field around the features. This is done by devel-
oping models with adequate, often minutely detailed, terrain data and mesh resolutions.
• For bridge local hydraulics, piers should be included in 2D models as geometric features
using adequate mesh resolution.
• When SRH-2D and HEC-RAS 2D models are developed with similar adequately refined
meshes, similar terrain resolution, and the same Manning’s n values, the model results are
not expected to differ substantially, but they are not expected to be the same either. Both
models solve almost the same set of governing equations. However, there are several key
numerical scheme differences between the two models that may make direct comparison
of results difficult.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

PART I

Literature Review and


Identifying Knowledge Gaps

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CHAPTER 1

Literature Review
and Knowledge Gaps

1.1 Introduction
The celebrated Manning’s equation in open-channel hydraulics draws persistent criticism
because of ambiguity in the interpretation and quantification of the roughness coefficient n
(Manning 1891; Chow 1959; Yen 1992a, 2002; Zevenbergen et al. 2012; Liu 2014). This ambiguity
constantly manifests itself in literature and in practice when n values need to be assigned. The
problem boils down to what the Manning’s n physically represents and how to quantify it.
Often tables, charts, and empirical equations, in conjunction with the professional judgment
of hydraulic engineers, need to be judiciously used. There is never a definitive formula for
Manning’s n, which leaves vast room for subjectivity and, thus, large variations in roughness
values. In transportation-related fields, inconsistency often exists among departments of trans-
portation (DOTs) of different states and even within states.
Confusion and inconsistencies in selecting a roughness value have existed for many decades
when 1D hydraulics models [e.g., HEC-RAS 1D (Brunner 2016)] were the prevailing tools for
hydraulic calculations at highway crossings and in transportation corridors. Today, the FHWA
and many state DOTs have gradually transitioned to the use of more accurate 2D hydraulic
models, such as SRH-2D (Lai 2008). HEC-RAS 2D (Casulli 2009; Brunner 2016) and other
2D hydraulic models are also being used to complete a variety of hydraulic analyses that previ-
ously relied on 1D solutions. The newer 2D models are more reliable and accurate since they
provide more detailed representations of real-world terrain, rely on fewer assumptions, and
resolve flow features neglected by 1D models. However, the confusion and inconsistencies in the
selection of roughness values still exist. Indeed, they are exacerbated because of the inherent dif-
ferences between 1D and 2D models (Liu 2014) and the lack of literature and guidance on these
differences available to practitioners.
A thorough literature review of relevant research was performed to identify the state of the
practice related to 2D hydraulic models and best modeling practices for transportation-related
riverine settings.

1.2 Flow Resistance


In 1D and 2D models, Manning’s equation and its roughness coefficient n are used to quantify
flow resistance. Therefore, to get to the bottom of this Manning’s n issue, it is of primary impor-
tance to review the literature on flow resistance in open channels.
For steady, uniform, free-surface flows, it is assumed that the total flow resistance is in balance
with the driving force (i.e., gravity). Rouse (1965) described four sources of flow resistance: skin
friction due to surface roughness, form drag due to channel irregularities, resistance associated

11

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

12   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

with flow unsteadiness, and wave resistance due to free-surface irregularities. In contrast to the
other three sources, surface roughness is present in any wall-bounded flows. Form drag stems
from irregularities in bed elevation due to morphodynamic bedforms, such as dunes and ripples,
vegetation, in-stream obstructions, and channel alignment. Resistance from unsteadiness accounts
for energy or momentum loss due to local accelerations where the steady-flow assumption fails
to hold. Typically, wave resistance is small, as compared with other sources of resistance, and
often not analyzed.
A common formula used to relate flow velocity and resistance for open-channel flows is the
Manning equation (in U.S. customary units):

1.486 2 3 1
U= R S0 2
(1-1)
n

Here, U is the average channel velocity, R is the hydraulic radius, S0 is the bottom/bed slope,
and n is Manning’s roughness coefficient, which is a lumped parameter that must encompass all
sources of resistance. In theory, the Manning equation is only valid for uniform flows; however, it
is used in 2D models to calculate flow resistance for all types of flows (e.g., gradually and rapidly
varied flows).
In the following sections, the major sources of flow resistance (i.e., surface roughness and
bedforms) are reviewed separately, with the focus on how they are connected to Manning’s n.

1.2.1 Grain Roughness


The surface roughness component of total flow resistance is treated independently from others
and has direct ties to near-wall boundary-layer theories (de Saint-Venant 1843; Stokes 1880;
Prandtl et al. 1904; Nikuradse 1933). In the thin layer near a wall, resistance is generated based
on the vertical velocity gradient and is proportional to the fluid’s molecular dynamic viscosity.
Because of its independence from form drag, resistance associated with surface roughness has
historically been estimated from measurements of resistance of steady, uniform flow in straight
circular pipes. Moody (1944) recognized that with such simplifications the surface roughness
could be predicted based only on the Reynolds number and relative roughness of the boundary:

n J kN
= F KK Re, s OO (1-2)
R 1 6
L RP

where Re is the Reynolds number, F denotes functionality, and ks is the absolute roughness height
dependent on the bounding material.
This function was plotted in the Moody diagram. Several researchers have attempted to fit the
Moody diagram with approximate expressions; the most widely accepted of these is the implicit
Colebrook-White equation. It should be noted that the Colebrook-White equation predicts the
Darcy friction factor, f, which in U.S. customary units is related to Manning’s n by

f gn
= (1-3)
8 1.486R 1 6

When it comes to riverbed material, many researchers relate ks to bed-material size; values
of ks smaller than the viscous sublayer thickness indicate a smooth bed. Garcia (2008) provides a
summary of these relationships and includes a table showing ks = αDa, where a is the percentage
of bed-material size. The most common size in that table is 84% finer (a = 84), and the factor,
α, for this size ranges from 1.6 to 5.1, with an average value of 3.2. The range of values can be

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Literature Review and Knowledge Gaps   13

attributed to particle shape, size distribution of the sediment mixture, and other factors. There is
a large body of literature on how to connect Manning’s n with grain sizes, for example, Strickler
(1923) using D50, Keulegan (1938) using D90, and Limerinos (1970) using D84. All these empirical
formulas were from regression analysis and have their own range of applicability.
For 1D models, Hey (1979) showed that, in addition to relative roughness of particle size to
channel depth, channel shape contributes to the composite roughness of a gravel-bed river channel.
This also means that in 2D models, and as demonstrated by Equation 1-2, that Manning’s n
would vary within a channel of constant grain size and varying depths. This type of variation of
Manning’s n within a channel is less often considered in practice.
Chen et al. (2019) quantified flow resistance in gravel-bed rivers using high-fidelity computa-
tional fluid dynamics modeling in conjunction with high-resolution microtopography derived
from photogrammetry. In their cases, the bed material is coarse; therefore, the grains, and conse-
quently ks, protrude beyond the viscous sublayer thickness. It was determined that the standard
deviation (STD), σz, of the streambed elevation serves as the best vertical roughness scale, and
the friction factor follows the relationship in Equation 1-4:

f 1
= J H N (1-4)
8
5.75 log KK O
1.09v z O
L P
where H is water depth.

1.2.2 Form Drag


For typical engineering applications involving natural rivers, form drag can be the dominant
factor in total flow resistance. Form drag most commonly originates from sediment bedforms,
vegetation in channels and floodplains, and obstructions.
Common classifications of bedforms include ripples, dunes, and antidunes. The absence of
bedforms in a fluvial channel is referred to as plane bed. Plane-bed rivers have similar resistance
characteristics of rigid-boundary channels (i.e., dominance of surface friction); however, granu-
lar bed material allows flow to penetrate through void spaces between particles. Channels with
ripples will experience resistance due to surface roughness as well as form drag, and channels
with dunes or antidunes will experience resistance due to surface roughness, form drag, and
wave resistance (Yen 2002). For channels with dunes, the water surface is out of phase with the
change in bed elevation, and the opposite is true for antidunes. The associated changes of bed
elevation and water-surface elevation (WSE) result in local velocity and pressure gradients (and
nonhydrostatic pressure) that contribute to form drag (Parker et al. 2003).
Alluvial channels have further complexity when bedforms are present because the condition
requires active transport of bed sediments. For active channels with high levels of suspension,
velocity gradients differ from those of clear-water channels, thus the balance of flow resistance
and gravity is altered. Further, a portion of the flow energy and momentum is used in dislodging
and transporting grains of sediment, which further alters flow resistance estimates (Yen 2002).
This implies the need to adjust estimates of Manning’s n for active versus inactive sediment
transport.
Laboratory and field investigations of drag associated with vegetation date back to the early work
of Ree and Palmer (1949). Since then, a myriad of laboratory and field works have been conducted
(Petryk and Bosmajian 1975; Kouwen and Li 1980; Burke and Stolzenbach 1983; Eckman 1990;
Abdelsalam et al. 1992; Nepf 1999; Järvelä 2005; López and García 2001; Nikora and Nikora 2007;

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

14   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Tanino and Nepf 2008; Nepf 2012). Typically, investigations have sought generalized relationships
between vegetation density, submergence, and drag force. It has been argued that in deeply sub-
merged channels, vegetation can be treated as large-scale surface roughness. Conversely, emer-
gent or shallow submerged vegetation alters the flow dynamics such that resistance is significantly
affected (Nepf 2012). Recently, Shields et al. (2017) reviewed the state of knowledge and proposed
an adjustment to Manning’s n to correct for added resistance caused by vegetation. Vegetation that
is flexible may have greater flow resistance at low flow velocity than at high velocity when it bends.
In the context of engineered components within natural river systems, in-stream structures
and obstructions represent another potential source of flow resistance that should be counted.
Structures such as rock veins, woody debris, riprap, and bridge piers and abutments will all
induce additional drag forces and flow resistance (Zevenbergen et al. 2012). The drag associated
with bridge piers has been extensively studied, and typical values of drag coefficient are provided
in the HEC-RAS reference manual (Brunner 2016). Typically, equivalent-width approaches are
adopted for complex piers; this is a method that was further affirmed by Yang et al. (2019). More
natural in-stream structures (e.g., rock veins and woody debris) have received attention. Follett
et al. (2020) studied the flow resistance of wood accumulations using a backwater approach.
These and other components that add to flow resistance should be included in the total flow
resistance if not otherwise accounted for in any modeling effort.

1.3 Manning’s n
The brief review of flow resistance factors makes clear that proper assignment of the Manning’s
n value depends on what it represents. Tables listing typical values for different channel conditions
can be found in any open-channel hydraulics textbook. Indeed, the derivation of the Manning
equation was based on the Chezy equation, which predates the Manning equation:

U = CR 1 2 S 10 2
(1-5)

1.486 1 6 8g
with C = R = (1-6)
n f

In Equations 1-5 and 1-6, C is Chezy’s roughness coefficient. Open-channel hydraulics text-
books may state that Manning’s n considers all factors that exert a retarding effect on flow. Unfor-
tunately, this statement is not accurate in the context of 1D versus 2D hydraulic models. The
reasons are as follows:
• Some factors have been explicitly resolved in 2D models. For example, large bridge piers are
usually captured with holes in a mesh for 2D models.
• Numerical schemes used in computational hydraulics models might affect which factors
should be excluded from the Manning’s n.
• A less obvious and often overlooked aspect is the inherent assumption of uniform flow made
in the Manning equation and how valid the assumption is for gradually varied flows. This
assumption is not isolated to the Manning equation because it is inherent to all flow resistance
formulations (e.g., see Equation 1-5).
It is perhaps obvious, though not trivial, that skin friction or form drag from any source
should increase the value of Manning’s n. However, in application of the Manning equation to
2D hydraulic models, exactly what resistance sources should be included in Manning’s n is not
well studied.
In every case, surface roughness must be included as the baseline. To this baseline value of
Manning’s n, additional resistance from form drag components must be added. The components

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Literature Review and Knowledge Gaps   15

that should be included are not always obvious since those factors are site and scale specific. One
example of this complexity is simple seasonal changes, which will directly impact vegetation
growth and the resistance associated with vegetation. As another example, if bedforms are not
captured by the mesh for reach-scale models with coarse resolution, bedforms may be treated
as large-scale surface roughness elements (Yen 2002). On the other hand, section-scale models
would require more careful selection of Manning’s n associated with bedforms depending on
terrain data and mesh resolution. A value too large will result in potential interactions with the
water surface creating an artificially wavy free-surface. A value too low will underpredict resis-
tance; therefore, flow velocities will be overpredicted. Beyond the contributors to resistance,
the linear superposition approach (discussed in Chapter 4) can affect how the components are
included in a composite value of Manning’s n.

1.3.1 Effects of Unsteadiness and Nonuniformity


Manning’s equation, as all flow resistance formulations do, assumes a uniform flow condition,
under which the gravity-driving force and the flow resistance force are in balance (Chow 1959).
This is an important assumption to keep in mind when using the Manning’s roughness coeffi-
cient in 1D and 2D models. Uniform flows almost never happen. The farther the simulated flow
is from the uniform flow condition, the more potential error may exist in the use of the Manning’s
equation. The deviation of the simulated flow from the uniform flow condition and the asso-
ciated error will be lumped into the calibrated Manning’s n. In other words, this part of the
Manning’s n (and other flow resistance formulations) is purely because of the invalidity of the model
assumption rather than any physical factors.
Typically the depth-averaged shallow-water equations (SWE) are solved in 2D models. SWE
are the mathematical descriptions of conservation laws (mass and momentum):

2h 2hU 2hV
Mass: + + =0 (1-7)
2t 2x 2y

2hU 2hUU 2hVU 2hTxx 2hTxy 2z redx bx


x-momentum: + + = + - gh - + D xx + D xy (1-8)
2t 2x 2y 2x 2y 2x t

2hV 2hUV 2hVV 2hTxy 2hTyy 2z redx by


y-momentum: + + = + - gh - + D y x + D yy (1-9)
2t 2x 2y 2x 2y 2x t

In Equations 1-7 through 1-9, t is time; x and y are horizontal Cartesian coordinates; h is water
depth; U and V are depth-averaged velocity components in x and y directions, respectively; g is
gravitational acceleration; Txx, Txy, and Tyy are depth-averaged turbulent stresses; Dxx, Dxy, Dyx,
and Dyy are dispersion terms due to depth averaging; z = zb + h is WSE (zb is bed elevation); and
ρ is water density. The depth-averaged dispersion terms are implemented in only some 2D models.
For example, HEC-RAS 2D implements the dispersion terms directly. If a model does not imple-
ment such terms, such as in SRH-2D, their effects are implicitly lumped into model calibration
parameters, such as the turbulence model coefficient or Manning’s n. τbx and τby are the bed shear
stresses (flow resistance), which are the key to this project. Unfortunately, there is no general for-
mula for the calculation of bed shear stress under unsteady and varied flow conditions. The best
approximation is to use the shear stress under uniform flow condition, which can be calculated
with the Manning’s equation as follows:

Jx N J N
KK bx OO = tCf KK U OO U 2 + V 2 ; Cf = gn
2
(1-10)
LV P h 1 3
x
L by P
where Cf is a friction coefficient that depends on Manning’s n and water depth h.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

16   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

This same issue occurs in 1D models where energy slope is calculated using Manning’s equa-
tion. Theoretically, Equation 1-10 is valid only for uniform flows because of the use of the
Manning’s or any other flow resistance equation. In practice, it is used for gradually and rapidly
varied flows, although for gradually varied flows, hydrostatic pressure can be assumed. When
rapidly varied flows are simulated using standard 1D and 2D models, hydrostatic pressure is still
assumed but is likely inaccurate. Thus, the use of the Manning’s equation to estimate the flow
resistance (i.e., the bottom shear stress in 2D models) is only an approximation.
In summary, the proper assignment of the Manning’s n value depends mainly on what it rep-
resents physically (e.g., sediment size, shape, vegetation, channel irregularity, alignment, mean-
dering, obstructions). The model dimensionality (1D versus 2D) determines which factors are
considered explicitly in the model and as a result should be excluded from the roughness value.
Manning’s n also depends on the numerical treatment and schemes used in the computer models
(e.g., how the detailed bathymetric information is incorporated into the mesh). Lastly, the use of
the Manning’s equation to calculate the flow resistance in 2D models has an assumption of local
uniform flow, which is almost never valid in the real world. This invalidity and its associated error
are unphysically embedded in the Manning’s n equation (and all other flow resistance formulas)
because it is usually the only calibration parameter available in hydraulic models. All these impor-
tant factors need to be considered when roughness value guidelines are developed for 2D models.

1.3.2 Composition and Decomposition Methods


Natural and most manmade open channels are not straight and prismatic. So, for any channel
reach, care should be taken when a single value of Manning’s n is applied. Researchers need to
consider the uniformity of the channel’s characteristics in the cross-channel and streamwise
directions. Two issues arise: (1) The assumption that flow resistance along a channel’s wetted
perimeter at a cross section is constant does not always hold, and (2) linear superposition of dif-
ferent flow resistance components may not capture the total resistance correctly.
Often, flow resistance at a particular cross section is not constant along the wetted perimeter
(Figure 1-1). One example of this is when the channel bottom and sides of engineered canals are
made of different materials (Chaudhry 2008). Another commonly encountered example is when
the channel and floodplain are modeled in tandem. Out of necessity in 1D modeling and conve-
nience in 2D modeling, an equivalent Manning’s n, which was defined by Chow (1959) based on
the work of Horton (1933) and Einstein (1934), is applied as follows to these cases:

a / Pi n 3i 2 k
2 3

n= (1-11)
/ Pi
where P is the wetted perimeter, and the subscript i represents a value for the i-th subsection of
the wetted perimeter.
The following equation, another form of the composite roughness, was developed from the
work of Muhlhofer (1933) and Einstein and Banks (1950):

a / Pi n 2i k
1 2

n= (1-12)
a / Pi k
1 2

Both of these forms of equivalent Manning’s n were based on dimensional arguments, but
the underlying assumptions were slightly different. In the Horton (1933) and Einstein (1934)
methods, it was assumed that each part of the flow area has the same mean velocity. The
Muhlhofer (1933) and Einstein and Banks (1950) methods, on the other hand, assume only

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Literature Review and Knowledge Gaps 17

Figure 1-1. Compound channel cross section. Each subsection has an


independently defined Manning’s roughness coefficient, with sections
including the effect of grass, a mobile point bar, riprap, and a large
obstruction.

that the total flow resistance force is equal to the sum of individual resistive forces in each sec-
tion as in the following:

/ x i Pi L = xPL = cRS 0 (1-13)

where L is the streamwise length of the control volume, and γ is the specific weight of water.
Yen (1992b) proposed a new formula that assumes the total shear velocity, cRS 0 , is equal to
the weighted sum of subarea shear velocities such that the composite Manning’s n becomes the
following:

/ n i Pi R 1i 6

n= (1-14)
P R1 6

HEC-RAS 1D uses compositing Manning’s n for channels when roughness varies within the
channel but recommends subdivision of channel from floodplains and calculates the convey-
ance of each subarea separately. The total cross-section composite n is not used for hydraulic
calculations, rather it is an output. The energy equation is used to compute conveyance between
cross sections. For each overbank, conveyance is calculated by summing individual conveyances
in each subarea, and the main channel conveyance is calculated based on a composite n for the
main channel only. Finally, a composite Manning’s n for the entire cross section is calculated
based on the total conveyance, area, and wetted perimeter.
The equivalent Manning’s n method can be used to account for nonuniform resistance along
a channel’s cross section; furthermore, adding individual resistance contributors into each sec-
tion’s specified n should be done carefully. One common approach was initially proposed in
Cowan (1956) for channels and then explored in Arcement and Schneider (1989) for both chan-
nels and floodplains. For example, the composite Manning’s n value for a channel can be com-
puted by the following:

n = ` n b + n 1 + n 2 + n 3 + n 4 jm (1-15)

where nb is the base value, n1 is the addition for surface irregularity, n2 is the addition for channel
cross-section variation, n3 is the addition for obstruction, n4 is the addition for vegetation, and m is
a correction factor for channel meandering. Similar formulas are available for floodplains.
Caution should be taken when applying Equation 1-15 if the assumption is adopted that the
total flow resistance is equal to the sum of individual flow resistance as

x = xl + xm = cRSl + cRS m (1-16)

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

18   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

where τ is the total bed resistance, S is the friction slope, a single prime is the value associated
with surface roughness, and a double prime is the value associated with form drag. By substi-
tuting this decomposition into the Manning equation, the following can be obtained (in U.S.
customary units):

J N2 3
1.486 K xlO
nli = (1-17)
USl 1 6 K c O
L P

J N2 3
1.486 K xmO
nmi = (1-18)
US m 1 6 K c O
L P
Therefore, Equation 1-15 will not hold since to satisfy Equation 1-16 the following must be
true:

n 2i = nli 2 + nmi 2 (1-19)

This method assumes that the total shear stress can be linearly decomposed; this is opposed
to the linear decomposition of Manning’s n directly. Forces, which are vector quantities, can be
added as illustrated in Equations 1-8 and 1-9; therefore, Manning’s n values should not be addi-
tive. This was the approach used by Zevenbergen et al. (2012) to include pier drag as an increased
n value within an element of a 2D model. The bed shear force plus the pier form drag were
combined to calculate an equivalent Manning’s n value for the element. With this method, as
any other, it should be noted that the definition of τ′ should be carefully considered. Specifically,
if τ′ is estimated as previously discussed via the Moody diagram, adjustments should be made
to τ″ to account for differences between plane-bed alluvial channels and rigid beds (Yen 2002).
Equation 1-15, though it may be flawed, illustrates the value of Manning’s n as a lumped param-
eter. Equations that account only for grain resistance, such as Equation 1-2, cannot be easily
applied when additional processes contribute to total energy loss and total resisting forces. Many
channel and floodplain conditions include vegetation types, which cannot be represented by a
single roughness value.

1.3.3 Use in 1D and 2D Models


Many hydraulic engineers with previous experience developing 1D models may not realize
that the physical meaning of Manning’s n is different in 2D models. In 1D hydraulics models,
the cross-sectionally averaged mass, energy, and momentum conservation laws are solved (Fig-
ure 1-2). For example, the depth-averaged turbulent stresses in the SWE are not directly solved
by most models even though they affect the total flow resistance. The roughness coefficient is a
lumped quantity for all factors that cause flow resistance in a 1D channel (Chow 1959). These
factors include surface roughness, vegetation, channel irregularity, channel alignment, sedimenta-
tion and erosion, obstructions, stage and discharge, and more. On the other hand, in 2D hydraulic
models, the depth-averaged SWE are solved on a 2D mesh (Figure 1-2) (Lai 2008; Liu et al. 2008;
Brunner 2016). Instead of 1D cross sections, 2D meshes can explicitly represent some of the
factors influencing flow resistance, such as obstructions and channel alignment. As a result, the
Manning’s n used in 2D hydraulics models should exclude their effects and, thus, theoretically
has lower values than those used in 1D models for the same project domain. However, even in
2D models all effects of Manning’s n are attributed to surface stress even if the roughness value
includes other factors that are not actually associated with the surface.
In the river hydraulics community, there is some consensus on the reduction of roughness value
when transitioning the hydraulic models from 1D to 2D. The question is how much reduction it

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Literature Review and Knowledge Gaps 19

Source: Background adapted from a public domain image from openclipart.org. Link to the
background image: https://openclipart.org/detail/190920/riviere-river.

Figure 1-2. 1D versus 2D hydraulic model (left: 1D model with cross


sections; right: 2D model with cells covering the domain).

should be. Several previous case studies have reported reductions ranging from 0% to 50% (Deal
et al. 2017; Friend and McBroom 2018). The research team’s own experience shows Manning’s
n values are generally 10% to 20% lower in 2D models compared to 1D models. However, these
reductions are case specific and cannot be generalized. A more systematic approach backed by
theoretical analysis and broader case studies is needed.
Another major difference between 1D and 2D models is that discharge within a 1D cross
section is assumed to be spread based on the distribution of conveyance. This means that flow
can literally jump to different areas of consecutive cross sections in a 1D model without there
being an actual flow path and without energy cost. A consequence of this conveyance assump-
tion is to distribute most flow to the deepest section (main channel), which may not be true for
complex river systems. In a 2D model, however, flow can move only from one area to another
based on the hydraulic attributes of the intervening area. In practical terms, this would result in
higher Manning’s n values in a calibrated 1D model than a calibrated 2D model, which is con-
sistent with observations in practice.
Another difference between 1D and 2D models is that an entire cross section is assumed to
have the same energy slope throughout. This is an extremely limiting assumption, especially
during flood flows. 2D models resolve these terrain and hydraulic differences on the element scale
so they produce much more realistic hydraulic results. The simplifications of the 1D approach
lead to calibrating Manning’s n values that depend not only on physical conditions but also are
affected by the assumption built into the model algorithms.

1.3.4 Effect of Numerical Scheme and Mesh Resolution


Other potential contributors to the difference of roughness values between 1D and 2D models,
and even among different 2D models (e.g., SRH-2D versus HEC-RAS 2D), are the numerical
schemes, choice of variables, and approximations used. In 1D models, Manning’s n is used in the
calculations of energy slope, flow distribution (especially between channel and overbank areas),
energy correction factor (α), and kinetic energy. Typically for 2D models, finite difference, finite
volume (FV), and finite element methods are used. In theory, different numerical methods solv-
ing the same governing equations should give the same solution. However, this is not always the
case. Each hydraulic model has its special treatment for certain aspects of the code. For example,
how the bathymetry is represented in the model will affect the flow resistance calculation and,
subsequently, the Manning’s n. These treatments are typically designed to balance computational
speed, accuracy, and robustness. Because of these different treatments, results may be different

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

20   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

from model to model, even with the same selection of Manning’s n. Small differences in results
may be acceptable because all computer models are approximations. However, significant differ-
ences are not acceptable and should be adequately addressed.
In the context of this project, the different numerical treatments in different hydraulic
models may affect the roughness values. One such treatment is the subgrid terrain method in
HEC-RAS 2D (Figure 1-3). The intention is to use the high-resolution terrain data on a rela-
tively coarse mesh. The FV method is used to discretize the governing SWE on control surfaces
in a 2D mesh. Instead of using one representative elevation at an edge shared by two adjacent
faces (e.g., SRH-2D), it uses a terrain profile interpolated onto the cell face. Then, quantities
such as hydraulic radius and wetted perimeter are tabulated as a function of water depth. Con-
sequently, these quantities affect the calculation of force terms, including the flow resistance
involving Manning’s n, in the momentum equation. The subgrid terrain method in HEC-RAS 2D
also affects the volume–stage relationship of computational cells, which directly affect the solu-
tion of the mass conservation.
By comparing the two different terrain treatment methods in HEC-RAS 2D and SRH-2D, given
similar mesh resolution, HEC-RAS 2D sees more details of the terrain than SRH-2D does.
SRH-2D might also have a different terrain elevation along the face than HEC-RAS 2D because
the face elevation is interpolated from the cell-center elevations, which were originally inter­
polated from the cell-corner elevations. If a mesh is not well prepared, SRH-2D’s method may
remove crests and troughs along cell sides, resulting in a smoother surface and removal of hydraulic

Source: The background image is in the public domain created by the


United States Geological Survey (https://www.usgs.gov/media/images/
example-river-bathymetry).

Figure 1-3.   Different treatment of terrain data in


SRH-2D and HEC-RAS 2D. The background is an
example terrain dataset. In SRH-2D, the bed elevation
at the center of a cell edge (shown with red dots) is
interpolated from the elevations of two cells, P and N,
sharing the edge. In HEC-RAS 2D, the edge elevation
profile is cut through the original terrain (termed
subgrid terrain method).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Literature Review and Knowledge Gaps   21

controls. In other words, HEC-RAS 2D explicitly includes more of the flow resistance effects in the
mesh elevation information and consequently should exclude them from Manning’s n. Following
this logic, for the same site with comparable mesh resolution, the Manning’s n in HEC-RAS 2D
should be smaller than that used in SRH-2D. How this logic stands in reality needs extensive case
studies and detailed numerical analysis. This topic illustrates that mesh development of any model
must include careful attention to capture terrain details and hydraulic controls within the model
domain. This can be achieved with breaklines, high-resolution meshes, or a combination of both.
In addition to SRH-2D and HEC-RAS 2D, there are many other 2D hydraulics models avail-
able. Globally, different countries or agencies may prefer to use 2D models of their choice due
to various reasons. Each of these 2D models has its own way of numerically solving the govern-
ing SWE or their simplified versions. Consequently, the Manning’s n in each of these models
may have different values for a same case due to the convoluted nature of physical and numerical
representations of reality in these 2D models. Previously, benchmarking efforts by the UK
Environment Agency and the U.S. Army Corps of Engineers Hydrologic Engineering Center
(USACE HEC) have been performed to compare different 2D models (Néelz and Pender 2010;
Brunner 2018; Brunner et al. 2018). In these comprehensive benchmarking studies, the emphasis
was the model performance against either laboratory or field data. No special attention was paid
to the flow resistance coefficients, such as Manning’s n. For almost all their benchmarking cases,
the values of Manning’s n were listed with minimal justification.
Following the previous discussion, there is another aspect in 2D hydraulic modeling that will
affect the assignment of roughness values. This aspect applies to any 2D model, regardless of
what special numerical method is used. In computational hydraulics modeling, it is required
that the results should be independent of the mesh used. This is called a mesh independence
check. Typically, while keeping everything else fixed, successively refined meshes are used until
the changes in the results are below a predefined tolerance. This is also true for 1D models, where
adequate cross-section spacing is required. As previously discussed, meshes at different refinement
levels will see different details of the terrain (represented in the elevation information stored in the
mesh). The extra terrain details seen by a refined mesh should be excluded from the Manning’s n.
For hydraulic modelers, this means a mesh independence study should be performed concur-
rently with the Manning’s n calibration. Practically speaking, there is no mesh independence in
2D hydraulics modeling because of the interdependence of mesh resolution and Manning’s n
(and other model parameters, such as eddy viscosity).
In the literature, people may have reported mesh independence for 2D hydraulic modeling,
but their definition of “mesh independence” may be purely numerical. When all other model
parameters are kept constant, successively refined meshes will eventually exhaust all terrain data
information and reach a convergence in result; however, that converged result does not consider
the physics that the Manning’s n is dependent on mesh resolution.
A related topic is the quality of terrain data. Theoretically, researchers prefer high-quality
and high-resolution data. However, terrain data, especially those produced from light detection
and ranging (LiDAR) and other means, are usually noisy. Before terrain data can be used in
2D hydraulic modeling, it must go through quality control and a denoise process. Filtering and
cleaning may be necessary to remove artificial spikes and variability. This is especially true if
the HEC-RAS 2D model is used because of its subgrid terrain treatment. If the subgrid terrain
seen by a face in HEC-RAS 2D contains unreal variability, the wetted perimeter on that face
may be artificially increased and result in higher flow resistance than reality. Consequently, the
Manning’s n value must be lowered to match with reality.
The above conjecture on mesh–terrain resolution and its relation to Manning’s n is reason-
able from a theoretical point of view. However, there is literature that contradicts the general

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

22   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

statement in the previous paragraphs. For example, Lim and Brandt (2019) evaluated the com-
bined effects of mesh–terrain resolution and Manning’s n on 2D flood modeling performance.
They found that high-resolution mesh–terrain performed better with higher Manning’s n, while
lower resolution performs better with lower n value. The reason for their seemly counterintuitive
findings is not clear. It may be related to the 2D model used, the single case study, or the metrics
for the quantification of model performance.

1.4 Current Guidance on Selection of Manning’s n


Because of the popularity of the Manning’s equation, extensive research has been performed on
the numerical values of Manning’s n. There are many guidelines and tables in the literature, includ-
ing those shared by Bakhmeteff (1932), Cowan (1956), Chow (1959), Henderson (1966), Barnes
(1967), Arcement and Schneider (1989), Coon (1998), Yen (1992a), Yen (2002), Chaudhry (2008),
Sturm (2001), Zevenbergen et al. (2012), Soong et al. (2012), Yochum et al. (2014), and more.
Distinctions between channel classifications for the purpose of assigning a Manning’s n value
have depended traditionally on qualitative observation and subjective judgment. Early guide-
lines for selection of Manning’s n of natural channels relied on descriptions with accompanying
photographs, such as those from FHWA (Arcement and Schneider 1989) and the United States
Geological Society (USGS) (Barnes 1967). In those studies, field measurements were obtained
to calculate Manning’s n. Although indirect measurement of the total flow resistance is pos-
sible, data are often unavailable or unobtainable for every site. Therefore, recommendations
from Zevenbergen et al. (2012) and others include tabulated values of minimum, normal, and
maximum Manning’s n using qualitative descriptions. Examples of categories presented involves
wording like “clean, winding, some pools and shoals” as compared with another category described
as “sluggish reaches, weedy, deep pools.”
Almost all existing guidelines and tables are for the total Manning’s n in a river reach; in other
words, they are more suitable for 1D hydraulic models where nearly all the factors affecting the
flow resistance in a cross section have been lumped into the roughness value. Because of the
reasons discussed, these guidelines and values cannot be used directly in 2D models.
Equation 1-15 represents a feasible approach to properly account for the complexities described.
Explicit definitions of the individual resistance factors must be derived that rely less on qualita-
tive descriptions in assigning roughness coefficients. However, previous discussion of composite
Manning’s n suggests that a sum-of-squares method would correctly satisfy the condition that
the total flow resistance is equal to the sum of the flow resistance factors from individual con-
tributors (Equation 1-19). Some factors included in the Manning’s n for 1D models may need
to be excluded for 2D models depending on whether they have been captured by the 2D mesh
(Morvan et al. 2008). For example, if the 2D mesh already captures the surface irregularity and
obstructions (e.g., like bridges), then their corresponding n should be assigned a zero value.
In addition, river meandering is caused by the secondary current induced by streamline cur-
vature and centrifugal force in river bends. To the best of the authors’ knowledge, none of the
popular 2D models (e.g., SRH-2D and HEC-RAS 2D) fully considers this effect. As a result, the
meandering correction factor may need to, perhaps in part, stay in the roughness guidelines
for 2D models.

1.5 Synthesis of Literature Review and Knowledge Gaps


This section synthesizes a review of the vast literature on the topic of Manning’s n. The knowl-
edge gaps identified through the literature review are described.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Literature Review and Knowledge Gaps   23

1.5.1 Meaning of Manning’s n in 1D and 2D Models


The foremost gap in knowledge relates broadly to the meaning of Manning’s n in 2D hydraulic
models. In Manning’s formula, n is intended to account for all factors that exert a retarding
effect on flow. In hydraulic models, however, the meaning may be different. Certain modeling
approaches will explicitly include factors that contribute to overall flow resistance in the solution
technique, and thus should be excluded when specifying Manning’s n. This implies that the
meaning of n can change from one model to another. It depends on the selected model, the
computational grid resolution, and the geometric representation of the terrain as well as tradi-
tional flow resistance factors (e.g., friction, form drag). It remains unknown as to how each of
those components (model selection, grid resolution, and geometric representation) will change
the physical meaning of Manning’s n value to produce accurate and consistent model results.

1.5.2 Effect of Grid Resolution in Inclusion or Exclusion


of Flow Resistance Factors
The geometric representation of terrain in 1D and 2D hydraulic models is a major factor for
n values. Since sources of resistance include bed irregularities, for example, if the computational
mesh explicitly resolves those irregularities, they should be excluded from the specified, lumped
n value. A similar argument can be made for obstructions, vegetation, channel curvature, sedi-
mentation and erosion, and others. Manning’s n in 2D models should reflect only factors that
are not explicitly resolved. The effect of grid resolution relative to the length scales of resistive
features has not been addressed in the literature. The key question: At what resolution will
a feature be explicitly solved in the model? Further, how much of a reduction in Manning’s n is
required once a feature is partially resolved?

1.5.3 Method Used to Determine a Composite Manning’s n


Typically, contributing factors to flow resistance have been linearly superimposed to produce
a composite Manning’s n. This is the recommended method by some previous guidelines [e.g.,
Cowan (1956)]. However, since the overall goal of creating a lumped, composite Manning’s n is to
predict the total flow resistance, the resistance should be superimposed rather than Manning’s n
directly. The result is that a composite Manning’s n should be computed as a sum-of-squares
rather than a linear superimposition (Equation 1-19). Once an understanding of the individual
contributing factors for a given hydraulic model, terrain–mesh resolution, and geometric rep-
resentation are established, the method of combining factors into a single Manning’s n can be
addressed.
Further, there are no guidelines on how much area should be lumped into a single n. Often,
even in 1D models, the channel and floodplain at a study site are treated with independent rough-
ness values. However, there are no criteria used to distinguish areas that should be treated sepa-
rately. For example, an area with buildings is often represented as a high-resistance area rather than
a resolution of the building footprint itself. Alternatively, the buildings’ effect may be neglected
entirely based on engineering judgment. The judgment of the modeler should firstly reflect the
specific purpose of the model. In the building footprint example, whether the building should be
included as additional roughness or explicitly treated as an obstruction will differ if the purpose
of the study is to predict inundation versus understanding detailed hydraulics around a structure.
This can be further generalized as predetermination of the model’s scale by the modeler, which
will affect which terrain features must be explicitly modeled. To that same point, a modeler must
decide the spatial extent of areas that can be lumped together and treated with the same value of
Manning’s n. Criteria for spatial decomposition of flow resistance and Manning’s n have not been
established.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

24   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

1.5.4 Effect of Specific Models on Assigning Manning’s n


Since different models may have different solution strategies, guidance must be given in a
model-specific form. The most apparent example of this is in the geometric representations of
the bathymetry. SRH-2D and many other models represent the terrain with a single elevation at
each computational cell. This means that features of the terrain that are smaller than the cell size
will not be resolved and should be included in Manning’s n. Any resistive features much larger
than the grid size should be excluded from Manning’s n since their effect is explicitly resolved in
the model. HEC-RAS 2D, on the other hand, uses a unique subgrid representation of the terrain,
wherein features finer than the grid resolution do affect fluxes through computational cells by
more realistically representing cell volumes and terrain profile along faces. It should be stressed
that this does not fully account for these features hydraulically. A knowledge gap exists in how
different modeling techniques affect the selection of Manning’s n.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CHAPTER 2

Analytical Program to Achieve


Research Objectives

This chapter describes the analytical program that was developed and executed in the project
to achieve the research objectives. The design and rationale for the analytical program are pre-
sented. The execution and outcome of the program are presented in the following chapters.

2.1 Introduction
The analytical program was designed to assess roughness values in 2D hydraulic models, com-
pare 1D and 2D hydraulic models, and fill the identified knowledge gaps. The analytical program
is executed in Task 7 of the project and is reported in the following chapters.
The team developed, calibrated, and compared 1D and 2D hydraulic models: These models
started with simple hypothetical cases in which analytical solutions, or at least the bulk-flow
behaviors, are known (e.g., backwater curve in prismatic channels). Simple cases were informa-
tive on the correctness, responsiveness, and sensitivity of different models to roughness values.
Then, more complex cases were built with either small-scale laboratory experiments or field cases.
Finally, prototype cases in real engineering practice were built. Many benchmark studies have
been performed in the past for both 1D and 2D hydraulic models, such as the HEC benchmark
study for HEC-RAS (Brunner 2018; Brunner et al. 2018) and UK benchmark study for a suite
of 2D models (Néelz and Pender 2010). Several state DOTs have also performed some studies
on 2D modeling guidelines with examples, for instance Homan and Toniolo (2018). This project
and similar studies in the literature have a wide array of simulation cases. With the progression
of case complexity, it is possible to isolate different factors that affect the roughness and see the
response of models to cases at different spatial scales.
For the proposed analytical program, the SRH-2D program was used as the baseline model, and
comparisons were made by building comparable models using HEC-RAS 1D, HEC-RAS 2D, and
RiverFlow2D for further analysis. Mesh resolution compatibility is critical since coarser meshes will
capture fewer topographic features and, therefore, may require an adjusted Manning’s n. Similar
mesh resolutions were used when comparing cases using different models, and the effect of mesh
resolution on Manning’s n was evaluated independently.
The goal of the benchmarking and simulation cases is to provide some insights into the proper
assignment of the roughness values in 2D hydraulic models. The assignment of values should
be based on what roughness factors are at play, whether they have been captured in the model
explicitly, and by how much. As discussed previously, guidelines are available for 1D hydraulic
models. Guidelines for 2D models should be similarly constructed. The factors affecting the
roughness were examined separately in the designed cases. Table 2-1 shows some general guid-
ance on whether a factor should be included in the Manning’s n in 2D models.

25

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

26   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Table 2-1.   Factors affecting roughness values and whether they should be
included in the Manning’s n in 1D and 2D hydraulic models.
General Description on the Should Effect Be Included in
Characteristics Affecting Manning’s n?
Factors Manning’s n Value 1D 2D
Surface
Roughness Size and shape of sediment grains Yes Yes
Height, density, distribution,
Vegetation and types Yes Yes
Variations in cross section, size, and No (should be
Channel channel shape (e.g., bars, sand captured by high-
Irregularity waves, and bank-failure slumps) Yes resolution mesh)
No, if extra terms for
centrifugal force are
included in model
Channel equations; yes, if
Alignment Meanders and channel curvatures Yes otherwise.
Yes, if the digital
elevation model
(DEM) or mesh
is coarse; no, if DEM
Sedimentation, also called silting, and mesh are �ine and
Sedimentation can smooth out irregularities; scour capture the bed
and Scour holes increase irregularity. Yes features.
No, if structure is
No (HEC-RAS uses captured with 2D
other algorithms to mesh or model
Examples: bridges and model bridges and algorithm; yes,
Obstructions hydraulic structures culverts) otherwise.
Yes (the hydraulic
Yes (HEC-RAS 1D model should be able
has an optional to specify roughness
Manning’s n generally decreases calibration tool to coef�icient as a
Stage and with the increase of stage and generate �low versus function of water
Discharge discharge. roughness table) depth)
Yes (Manning’s n Yes (Manning’s n
Seasonal Mainly a concern for vegetation should be a function should be a function
Change growth of time.) of time.)
Yes (Currently no Yes (Currently no
Transport of sediment extracts hydraulic model hydraulic model
Sediment energy from �low and generally considers this considers this
Transport increases Manning’s n. explicitly.) explicitly.)
Mesh and
Terrain Data
(Resolution and Whether terrain and mesh capture Need to consider Need to consider
Quality) hydraulic features resolutions resolutions

2.2 Analytical Program


The analytical program is subdivided into six parts. In each part, the output consists of evalu-
ating a value or set of values of Manning’s n for the specific conditions. The roughness values
were tuned in each case to match existing validation data (if available). Each part produces out-
comes that may be used in later parts, as shown in Figure 2-1, toward the overall goal of filling
the knowledge gaps and producing guidelines for the selection of Manning’s n in 2D models.
During execution of the program, some changes were made because of new discoveries or unex-
pected outcomes. In the end, more simulation cases were performed than those in the original
program.
The following is a high-level overview of the six parts and their relationships. Detailed descrip-
tion of the cases and analysis is presented later.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Analytical Program to Achieve Research Objectives   27

• Part 1 consists of simple benchmark tests aimed at verifying existing values of Manning’s n
for surface roughness and to demonstrate the spatial effects of Manning’s n on flow velocity
and depth. These surface roughness values are used in later parts of the analytical program.
The sensitivity of Manning’s n in different flow regimes is demonstrated, and the effect of
relative n in a main channel and floodplain is examined.
• Part 2 systematically evaluates individual flow resistance factors (as listed in Table 2-1). First,
the factors will be unresolved by the computational mesh, then the cases will be rerun with
finer mesh resolution that explicitly resolves the features. The results will inform recommen-
dations on the value of Manning’s n associated with each component, the effect of inclusion
or exclusion of a component based on mesh resolution, and the method of computing a com-
posite value for Manning’s n.
• Part 3 specifically evaluates the mesh resolution required for a feature to be considered
resolved by the mesh and, therefore, eliminated from Manning’s n. A characteristic feature size
will be defined for which features are unresolved and must be represented in the prescribed
roughness value. The goal is to provide some quantifiable measure to define “resolved” versus
“unresolved.”
• Part 4 evaluates model-dependent effects of selecting Manning’s n. In this part, cases are built
using all proposed models (SRH-2D, HEC-RAS 2D, HEC-RAS 1D, and RiverFlow2D), and
direct comparisons are made.
• Part 5 develops a decision-making process for use by modelers based on results from Parts 1
through 4. The outcome of the decision-making process will be a recommended, model-specific,
mesh-specific, and site-specific value of Manning’s n.
• Finally, in Part 6, the proposed decision-making process is tested against prototype-scale data
using each of the models. Part 6 consists of cases having progressively higher complexity,
including multiple factors (Table 2-1), in-channel and floodplain flow, bridge crossings, and
multiple coverage areas requiring individual assignment of Manning’s n.

Figure 2-1.   Parts of the analytical program.


The decision-making process will be refined
during Part 6.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

28   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

2.2.1 Part 1: Verification of Existing n Values for Surface Roughness


and Spatial Effects of n
For these tests, Manning’s n is tuned to match laboratory and theoretical data. These results
are used in later parts.
• Case 01: Smooth-boundary prismatic channel (for plastic n 0.01).
• Case 02: Sand-bed, clean channel.
• Case 03: Short-length, prismatic channel. Two values of n tested independently.
• Case 04: Long-length, prismatic channel with the same n values as Case 03. This case is run
once with each of the four models to reveal any differences between the models. By comparing
Case 03 and Case 04 the length-dependent, cumulative effect of Manning’s n will be evaluated.
With longer channel length, the effect of n on the WSE will be amplified.
• Case 05: Idealized channel plus floodplain. The relative values of Manning’s n in the channel
versus in the floodplain are varied to determine the effect of relative n values on simulation
results. This case is run once with each of the four models to reveal any differences between
the models.

2.2.2 Part 2: Evaluation of Individual Flow Resistance Factors


Flow resistance factors, mainly those inducing form drags, will be assessed individually to
explicitly determine their effect on Manning’s n.
• Single obstruction in a smooth channel
– Case 06: Unresolved single bridge pier in prismatic channel. Increase n until it matches
measurements using Zevenbergen et al. (2012) as a guide. In this case, the resistance asso­
ciated with the bridge pier is lumped into Manning’s n, and the bridge pier is not included
as a feature of the computational mesh.
– Case 07: Resolved single bridge pier with n = n|Case 01 (meaning the n value from Case 01).
In this case, the bridge pier is included as a feature of the computational mesh and is
excluded from Manning’s n.
• Single obstruction in a rough-boundary channel
– Case 08: Same as Case 06 but with n = n|Case 02.
– Case 09: Same as Case 07 but with n = n|Case 02. The calibrated Manning’s n for Case 07
will reflect only the resistance associated with the bridge pier since the surface roughness
is near zero. Manning’s n for Case 09 will reflect the resistance associated with the bridge
pier and rough boundary. By comparing these and knowing n|Case 02, Equation 1-19 can
be verified.
• Bedforms (multiple, uniform channel irregularities)
– Case 10: Unresolved dunes or baffle blocks in a smooth, prismatic channel.
– Case 11: Unresolved dunes or baffle blocks in a rough, prismatic channel. By comparing
results from Case 10 and Case 11, the sum-of-squares composite method can be verified.
Channel irregularities will be generated with proper scaling with flow depth and channel
geometry.
– Case 12: Rerun Case 11 with resolved dunes or baffle blocks with n = n|Case 02.
• Other channel irregularities
– Case 13: Change in cross-sectional shape in 2D model.
– Case 14: Change in cross-sectional shape in 1D model. Increase n until the water-depth
profile matches that of Case 13.
• Stage/discharge relations
– No cases will be specifically run: n should be given as a function of water depth based on
previous studies.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Analytical Program to Achieve Research Objectives   29

• Vegetation
– No additional cases will be run regarding vegetation beyond its treatment as surface rough-
ness. However, a vegetation module based on an additional drag force acting on the flow
is in development for the SRH-2D model. Future releases of 2D models are expected to
include this treatment for vegetated areas. Since tabulated values of Manning’s n include
the effects of vegetation, if the drag due to vegetation is resolved in the equations, the speci-
fied Manning’s n should be reduced.

2.2.3 Part 3: Evaluation of Conditions Necessary


for a Feature to be Resolved
Part 3, run in tandem with Part 2, is a sensitivity study explicitly intended to define the condi-
tions for a feature to be considered resolved by the mesh and therefore excluded from Manning’s n.
The goal is to determine the number of cells required in each direction so that the prescribed
Manning’s n approaches n associated only with skin friction.
• Rerun Case 06 and Case 11 with progressively finer resolution meshes to quantify the threshold
at which the feature is resolved. The final resolved case is equivalent to Case 07 and Case 12.
• Additional cases (with progressively finer mesh resolution) in a prismatic channel with non-
uniform bed irregularities. A similar analysis will be performed as for Case 06 and Case 11.
From these results, a dominant, characteristic irregularity size will be defined. This will
become the basis for the criteria for whether the irregularities are resolved by the mesh since some
natural features are not regularly spaced in practical applications (e.g., riprap on a meander bend).

2.2.4 Part 4: Evaluation of Model Choice Effect


on Specifying Manning’s n
Part 4 directly compares results from each model given the same input parameters and com-
parable mesh resolutions. Cases from previous parts are rerun, and comparisons of the calibrated
Manning’s n are made.
• Rerun Cases 06, 07, 11, 12, 13, and 14 in HEC-RAS 2D using comparable grid resolution.
– Recalibrate n to quantify the difference between SRH-2D, HEC-RAS 2D, and RiverFlow2D.
Note: The dunes or baffle blocks (Case 11) will be represented in the subgrid terrain; the
bridge pier (Case 06 and Case 07) and resolved dunes or baffle blocks (Case 10) will not.
• Rerun Cases 06, 07, 11, 12, 13, and 14 in HEC-RAS 1D with sections spaced at reasonable
distances based on common practice.
– Evaluate the value of n that would match the results from Cases 06, 07, 11, 12, 13, and 14.
Note: These cases consider prismatic channels only. Additional 1D cases are run in Part 2
and Part 6.

2.2.5 Part 5: Development of Decision-making Process


Based on the findings from Parts 1 through 4, a decision-making process will be developed for
use by modelers. This decision process will be the main part of the final guidelines.

2.2.6 Part 6: Verification and Refinement of Proposed


Recommendations with Case Studies
The decision-making process will be demonstrated and validated in Part 6. The following is
a list of case studies: the Iowa River 2008 flood case; the DelDOT I-242 case; and the Susitna
River, Alaska, case.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

PART II

Execution of Analytical
Program and Results
The analytical program is designed to assess roughness values in 2D hydraulic models, com-
pare 1D and 2D hydraulic models, and fill the knowledge gaps identified in the previous phase
of the project.
The research team developed, calibrated, and compared 1D and 2D hydraulic models.
Researchers started with simple, hypothetical cases in which analytical solutions, or at least the
bulk-flow behaviors, are known (e.g., backwater curve in prismatic channels). Simple cases are
especially informative on the correctness, responsiveness, and sensitivity of different models to
roughness values. Part 1 of this program was devoted to these simple tasks. In Part 1, the team
also studied the effect of the ratio of Manning’s n in the floodplain to that in the main channel.
Parts 2, 3, and 4 were executed in tandem. The team considered different flow resistance factors
(e.g., obstructions, bedforms, and channel irregularities) through extensive hydraulic modeling
efforts (much more than initially proposed). The goal was to determine how each flow resistance
factor affects the Manning’s n value and when they are geometrically and hydraulically resolved.
For each factor, the effect of different models was studied. SRH-2D was used as a base model for
the whole project. RiverFlow2D, HEC-RAS 2D, and HEC-RAS 1D were used for comparison.
Same or similar meshes were used with different models for a fair comparison.
Based on the results in Parts 1 through 4, the team developed best practices for 2D modeling
and a step-by-step decision-making process as guidelines for determining Manning’s n. The
guidelines are mainly based on the “resolved and unresolved” concept that will be discussed in
more detail in Section 4.2.5. The best practices and guidelines are described in Chapters 5 and 6.
Both are demonstrated by three real-world cases.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CHAPTER 3

Verification of Existing n Values


for Surface Roughness and
Spatial Effects of n
3.1 Comparison of Manning’s n in Different Models
Four cases were designed to compare the models. SRH-2D and HEC-RAS were compared
with simple cases (mainly backwater curves). Because of the simplicity of these cases, researchers
know the solutions and can verify whether there is an error or a difference between models. For
most cases, the solutions from SRH-2D, HEC-RAS 2D, and HEC-RAS 1D are comparable (to
some percent difference due to numerical errors). All these cases were prepared with the help
of the open-source Python package pyHMT2D (https://github.com/psu-efd/pyHMT2D). The
meshes were generated in HEC-RAS v6.1 and then converted to the SRH-2D format. Therefore,
SRH-2D and HEC-RAS 2D used essentially the same meshes.
In addition, numerical parameters and options in SRH-2D and HEC-RAS 2D were set to be
the same (to the extent possible). Figure 3-1 shows the selected 2D options for HEC-RAS 6.1.
The important options include the solver option (equation set), turbulence model, and mixing
coefficients.
The following cases were originally proposed:
• Case 01: Smooth-boundary prismatic channel (for plastic n ≈ 0.01).
• Case 02: Sand-bed, clean channel.
• Case 03: Short-length, prismatic channel. Two values of n were tested independently.
• Case 04: Long-length, prismatic channel with the same n values as Case 03. This case was run
once with each of the four models to reveal any differences between the models.
Note: By comparing Cases 03 and 04, the length-dependent, cumulative effect of Manning’s
n was evaluated. With a longer channel length, the effect of n on the WSE is amplified.
• Case 05: Idealized channel plus floodplain. The relative values of Manning’s n in main channel
and floodplain were varied to determine their effects. This case was run once with each of the
four models to reveal any differences among the models.
Case 01 shows the comparison between SRH-2D and HEC-RAS 2D for a channel 150 m long
and 10 m wide with a slope of 0.0005. The Manning’s n was 0.01. The downstream boundary has
a specified water depth of 0.5 m and an upstream inflow of 6 m3/s, which results in an M1 curve
(it is not obvious in the figure because the domain is short). Both models were run for 15 h to
reach a steady state. The time step (dt) is 1 s, and the mesh resolution is dx = dy = 1 m.
The results for Case 01 are shown in Figure 3-2. It can be observed that the results are almost
the same (except for a small deviation at the upstream end, which may be because of the way
HEC-RAS 2D interpolates the results). Figure 3-2 also shows the absolute and relative errors,
defined based on the difference between SRH-2D and HEC-RAS 2D. These errors are extremely
small, which means both models produce similar results with the same governing equations.

33

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

34   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 3-1.   The selected options for HEC-RAS 2D.

Figure 3-2.   Comparison between SRH-2D and HEC-RAS 2D for a simple M1 backwater curve case (Case 01).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification of Existing n Values for Surface Roughness and Spatial Effects of n  35

Figure 3-3.   Comparison between SRH-2D and HEC-RAS 2D for a simple M2 backwater curve case (Case 02).

In Case 02, when the Manning’s n was increased from 0.01 to 0.03 and everything else remained
the same, the backwater curve type became M2. Figure 3-3 shows the simulated water-surface
profile by SRH-2D and HEC-RAS 2D. They are in nearly perfect agreement, and the absolute
error is less than 2 mm.

In Case 03, the channel length was increased from 150 m to 3,000 m, the Manning’s n was set
at 0.01, and the downstream water depth was set at 0.5 m. All other parameters remained the
same. Figure 3-4 shows the comparison of simulated WSEs from SRH-2D and HEC-RAS 2D.
The absolute and percentage errors in the WSE are also plotted. The difference is minimal.

In Case 04, when Manning’s n was increased from 0.01 to 0.03, the backwater curve type
became M2 (Case 04). Figure 3-5 shows the comparison between SRH-2D and HEC-RAS 2D.
The agreement between them is almost perfect, with a relative and absolute error of less than
0.4% and 6 mm, respectively. Thus, it can be concluded that both models produce similar results
for these simplified cases.

Figure 3-4.   Comparison between SRH-2D and HEC-RAS 2D for a simple M1 curve. Channel length is 3,000 m,
and the Manning’s n is 0.01 (Case 03).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

36 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 3-5. Comparison between SRH-2D and HEC-RAS 2D for a simple M2 curve. Channel length is 3,000 m,
and the Manning’s n is 0.03. (Case 04).

The 2D results of Cases 03 and 04 were compared with HEC-RAS 1D and a simple back-
water curve solver for wide channels. The simple backwater curve solver was implemented in
pyHMT2D and solved the 1D backwater curve equation with a finite difference scheme. The com-
parisons are shown in Figure 3-6 and Figure 3-7. The results show that HEC-RAS 1D produces a
slightly different result in comparison with all three other models. The three other models (SRH-2D,
HEC-RAS 2D, and the 1D backwater curve solver) produce similar results. HEC-RAS 1D includes

Figure 3-6. Comparison of a simple 1D backwater curve solver, HEC-RAS 1D, HEC-RAS 2D, and
SRH-2D for Case 03.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification of Existing n Values for Surface Roughness and Spatial Effects of n 37

Figure 3-7. Comparison of a simple 1D backwater curve solver, HEC-RAS 1D, HEC-RAS 2D, and
SRH-2D for Case 04.

the friction effect of sidewalls, which is not fully considered in all other models. This is clear from
the simulated normal depth at the upstream end. The inclusion of sidewall friction increases the flow
resistance and therefore the normal depth. This result is useful when comparing an HEC-RAS 1D
model with a 2D model.

3.2 Effect of Relative n Values in Composite Channels


The last section in Part 1 of the analytical program is Case 05, where the main channel connects
with a floodplain. This part aims to compare the effect of the ratio of Manning’s n in a floodplain
to that of the main channel. For this part, a composite channel (Figure 3-8) with a slope of 0.0005
and length of 1,000 m was considered. The main channel Manning’s n was fixed at 0.03, and the
Manning’s n of the floodplain varied from 0.03 to 0.1.
Table 3-1 shows all cases that were simulated in this section. The first two cases are to eval-
uate the effect of grid size. Two different grid sizes of 1 m and 2 m were considered. Breaklines

Figure 3-8. Cross section of the modeled


composite channel for Case 05.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

38   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Table 3-1.   List of different cases for composite channel.

were used along the edges of the main channel to accurately capture the terrain change. The
results are shown in Figure 3-9. Both grid sizes produced similar results.
The comparison of the results of other cases using HEC-RAS 2D and SRH-2D is shown in
Figures 3-10 through 3-17. These cases are designed to have different Manning’s n ratios between
the main channel and the floodplain. The downstream water depth was also varied to produce
different flow profiles. The absolute error and relative error of the models are also shown. Both
2D models produced nearly the same results.
Next, the 2D models were compared with HEC-RAS 1D. Figure 3-18 shows the comparison
of HEC-RAS 1D with 2D models for M1 profiles. Figure 3-19 shows the similar comparison for
M2 profiles. From both figures, when the difference between Manning’s n of the main channel
and the floodplain increases, the difference between 1D models and 2D models increases, too.
The 1D model needs higher Manning’s n values to match the 2D results. This is because the
shear between the floodplain and the main channel is not considered in 1D models. When the
ratio of Manning’s n between the floodplain and the main channel increases, the shear becomes
more prominent. This result implies that if a domain has been simulated with 1D models and
there is significant shear between the floodplain and the main channel, then Manning’s n in
2D models might need to be smaller than that in 1D models. Of course, this conclusion is not
general because there are other parameters in 1D and 2D models.

Figure 3-9.   Effect of grid size dx on SRH-2D and HEC-RAS 2D results.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification of Existing n Values for Surface Roughness and Spatial Effects of n  39

Figure 3-10.   Case 05_2 (nfloodplain 5 0.08, nmain channel 5 0.03).

Figure 3-11.   Case 05_3 (nfloodplain 5 0.08, nmain channel 5 0.03).

Figure 3-12.   Case 05_5 (nfloodplain 5 0.1, nmain channel 5 0.03).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

40   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 3-13.   Case 05_6 (nfloodplain 5 0.1, nmain channel 5 0.03).

Figure 3-14.   Case 05_7 (nfloodplain 5 0.05, nmain channel 5 0.03).

Figure 3-15.   Case 05_8 (nfloodplain 5 0.05, nmain channel 5 0.03).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification of Existing n Values for Surface Roughness and Spatial Effects of n  41

Figure 3-16.   Case 05_9 (nfloodplain 5 0.03, nmain channel 5 0.03).

Figure 3-17.   Case 05_10 (nfloodplain 5 0.03, nmain channel 5 0.03).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

42   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 3-18.   Comparison of 2D models with HEC-RAS 1D for M1 profile.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification of Existing n Values for Surface Roughness and Spatial Effects of n  43

Figure 3-19.   Comparison of 2D models with HEC-RAS 1D for M2 profile.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CHAPTER 4

Flow Resistance Factors, Required


Conditions for Resolving a Feature,
and Effect of Model Choice
The research team considered different flow resistance factors such as obstructions, bedforms,
and channel irregularities. Through extensive hydraulic modeling, the team determined how
each flow resistance factor affects the Manning’s n value and when they are geometrically and
hydraulically resolved.
The definitions of geometrically and hydraulically resolved are important. Geometrically
resolved means the model replicates the shape of structures, hydraulic controls, and other terrain
features within certain tolerance. Hydraulically resolved means the model produces flow results
that fully capture the influence of flow resistance features in the domain. Geometrically resolved
is a necessary but not sufficient condition for hydraulically resolved. In general, to hydraulically
resolve features, mesh should be refined not only in the immediate vicinity of these features but
also in their influence zones. No model can hydraulically resolve every flow resistance feature.
For each flow resistance factor, the effect of model choice was studied. SRH-2D v3.3.0 in SMS
v13.1 was used as a base model for the whole project. RiverFlow2D v07.51.00, HEC-RAS 2D, and
1D v6.1 were used for comparison.

4.1 Channel Obstruction


Obstruction in channels is a common flow resistance factor. One of the most prominent exam-
ples of obstruction is bridge piers. In practice, obstructions can be modeled using the following
approaches:
• Implementing the hole-in-mesh method: The obstruction, which will be a hole in the mesh,
hence, the name, is fully resolved with mesh.
• Employing the unassigned element method: This is theoretically like the hole-in-mesh
method. However, in this method, the mesh should initially be generated for the pier footprint
and then the pier elements are disabled or unassigned.
• Elevating mesh and terrain: The mesh nodes and cells covered by the obstruction are elevated
by terrain. This approach effectively models the obstructions as islands with near-vertical sides.
• Increasing the Manning’s n for the area covered by the obstruction.
• Adding additional drag force for the mesh cells covered by the obstruction. This approach is
similar to increasing Manning’s n.
Although only four cases initially were considered for this part (Cases 06 through 09), the
research team found additional cases were required to appropriately investigate these methods
and simulated more than 50 cases, excluding several initial runs with the following objectives:
• To study the effect of mesh size on two pier shapes with two sizes to determine when an
obstruction can be considered geometrically resolved.

44

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   45

• To benchmark the hole-in-mesh method, which is deemed the most accurate method, against
Yarnell’s equation to see how well a 2D model can resolve the hydrodynamics of obstruction if it
is geometrically resolved by the mesh and terrain. Yarnell’s equation is a semianalytical equation
calibrated with flume data for obstruction’s effect on backwater curves (Yarnell 1934).
• To evaluate unassigned and elevated mesh methods as substitutes to the hole-in-mesh method.
• To investigate the increased Manning’s n method.
• To explore the additional drag method.

The major findings in this part follow:


• It is important to clearly define geometrically and hydraulically resolved obstruction.
For a model to predict the hydrodynamics around an obstruction correctly and accurately,
it needs to first geometrically resolve the obstruction. The mesh and terrain (depending on
how the obstruction is represented in the model) should have a fine-enough resolution to
describe the geometric details of an obstruction’s shape. On the other hand, other factors affect
whether the model can capture the real hydrodynamics around the obstruction. For example,
mesh and terrain resolutions in the influence zone of obstruction also need to be high enough.
• Based on the results, the team proposes a quantitative metric for geometrically resolved. This
metric is the CSO. Generally, the CSO should be at least less than 0.5 to geometrically resolve
an obstruction and capture the velocity field. Note that a hexagon just meets this criterion as
representing a circular pier (π/6 is approximately 0.5) and that an octagon provides a better
geometric resolution of a circular pier (π/8 is approximately 0.4).
• Mesh refinement should also consider the width of obstructions compared to the overall
channel width. This is required to capture the interactions among obstructions, flow fields,
and channel banks.
• The Yarnell’s equation, based on conservation laws and calibrated with measurement data, was
used to evaluate the hole-in-mesh method, which is deemed as the most accurate method of
all. The results show that this method, using a geometrically resolved mesh, can accurately cap-
ture the hydrodynamics around piers and predict correctly the amount of the backwater effect.
• The unassigned and hole-in-mesh methods have the same results. The unassigned method
is a special method implemented in SMS for some 2D models, in which the selected cells do
not participate in the simulation. Despite the similarity, their velocity results in SMS appear
different. It is perhaps only a visual effect due to the postprocessing procedure for estimating
velocity at nodes from the cell-centered solution. In SMS, for the hole-in-mesh method, the
perimeter is by default considered as a wall with a zero velocity assigned. However, for the
unassigned method, a zero gradient is applied (i.e., the velocity of a node is assigned as the cell-
centered velocity).
• In the elevated mesh method, in which an obstruction is incorporated directly into the terrain,
a refined mesh is required around the obstruction. This method may cause near-vertical cells
around the edge of an obstruction, which may cause numerical stability issues. This method
is recommended only when the hole-in-mesh and unassigned mesh methods are not feasible.
• When it is not feasible to have a fine-enough mesh to resolve an obstruction and the local
hydrodynamics is not a priority for the modeling purpose, increasing Manning’s n coefficient
is a much cheaper alternative. The increased Manning’s n will increase the flow resistance,
which is caused by the obstruction. The increased Manning’s n value can be estimated by the
equation in Chapter 5 of HDS 7 (Zevenbergen et al. 2012).
The research team also identified a drawback in most 2D models about the increased
Manning’s n or the additional drag-force method for obstructions when a refined mesh is used
(obstruction’s footprint covers more than one mesh cell). The drawback stems from the imple-
mentation of the two methods. Details of the drawback are described in Section 4.1.1.2. Recom-
mendations for correct 2D model implementation are provided.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

46   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

With existing 2D models, the following comments should be considered in using the increased
Manning’s n method:
– In preparing the material coverage for high Manning’s n at obstruction location, the modeler
should consider which cells will be associated with it; the total area of these elements is used
later in calculating the needed Manning’s n.
– For pile or trestle piers that consist of several small piers over the channel width, the cal-
culated effective Manning’s coefficient ne using Equation 4-4 should spread over the width
of the channel.
– If the river cross section at the obstruction location has different n values or the obstruc-
tions are in a limited area of the channel width, we recommend spreading their effect only
over the affected area instead of the whole channel width.
• Some models, e.g., SRH-2D and RiverFlow2D, offer the additional drag method. The addi-
tional drag method is like the increased Manning’s n method and thus it currently suffers from
the same drawback in most 2D models. When an obstruction’s footprint covers more than one
mesh cell, these models currently cannot correctly reproduce the hydrodynamics around the
obstruction unless an extremely high drag coefficient is used.

4.1.1 Methods for Modeling Obstructions in 2D Models


4.1.1.1 Brief Description of All Methods
Generally, if the mesh cell size is small enough when compared with the obstruction size, it
can be resolved using the hole-in-mesh method, the unassigned-material elements (disabled
elements) method, or the elevated mesh method. On the other hand, if the cell size is not small
enough to resolve the obstruction, then researchers can model obstruction with an additional
drag force or by adjusting Manning’s n in the equations of motion. These methods are listed in
Table 4-1 along with the requirements, suitable use scenarios, and advantages of each method.
• The hole-in-mesh method is numerically similar to the unassigned element method (also
called the disabled element approach). One difference though is how 2D models internally
deal with mesh topology and final-result visualization. For both methods, the mesh needs to
be refined around the obstruction to capture its geometry. With these two methods, there is
no need for any Manning’s n adjustment.
• In the elevated mesh method, the elements within the footprint of the obstruction are elevated
to at least the top of the obstruction or high enough to prevent overtopping. In this method, the
mesh needs to be refined near the obstruction to capture the terrain change (obstruction is part
of the terrain). This is important because of the rapid elevation changes around the obstruction.
Adjustment of Manning’s n coefficient is not needed in this method. Since this method is an
approximation for the hole-in-mesh method, it is recommended only when the hole-in-mesh
and unassigned mesh methods are not possible (e.g., when using HEC-RAS 2D).
In this method, the underlying terrain needs to include the obstruction. If the mesh is not
refined sufficiently around an obstruction, SRH-2D and RiverFlow2D cannot see or resolve
the obstruction. Figure 4-1 shows a schematic comparison between meshes resolving and
unresolving an obstruction. In HEC-RAS 2D, the effect of obstruction may be partially cap-
tured by an unresolved mesh because of its subgrid terrain treatment (the obstruction must
intersect with cell faces to be seen). However, coarse mesh like this is not capable of capturing
the true hydrodynamics.
• When the mesh size is not refined enough to resolve an obstruction, then either the increased
Manning’s n method or the additional drag-force method can be used. These methods can be
used when the required mesh resolution of other methods is not feasible and the local hydro-
dynamics is not important for the modeling purpose.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Table 4-1.   Different methods for modeling obstructions in 2D models.
Availability in 2D models
Terrain Data Manning’s n Suitable SRH- HEC- River
Methods Requirements Mesh Requirement Requirement Applications Comments 2D RAS 2D Flow 2D
Mesh needs to Most accurate
capture the High-resolution method.
obstruction’s hydrodynamics, No need to specify
No special footprint as a hole such as bridge Manning’s n. Suitable
Copyright National Academy of Sciences. All rights reserved.

treatment is and the perimeter hydraulics; scour for relatively large


needed; make as a wall. analysis; and obstructions; small
sure terrain Sufficiently refined habitat design obstructions need
does not mesh around the such as large refined mesh, which
Hole-in- include the obstruction is woody debris may come at a high
mesh obstruction. required. Not needed (LWD), groin, etc. computational cost. Yes No Yes
The perimeter High-resolution
should be hydrodynamics,
considered a wall. such as bridge
Sufficiently refined hydraulics; scour
Unassigned No special mesh around the analysis; habitat
mesh treatment is obstruction is design, such as This is numerically the
elements needed. needed. Not needed LWD, groin, etc. same as hole-in-mesh. Yes No No
Terrain at the
footprint of It is not as accurate as
obstruction the “hole-in-mesh”
needs to have method. If an
enough Mesh needs to be Low-resolution obstruction is smaller
resolution and refined near the Not relevant hydrodynamics, than the mesh cell, 2D
Elevated be high enough obstruction to (recommend to flood modeling, models cannot fully
mesh to prevent capture the terrain use background large capture the local
elements overtopping. change. Manning’s n) obstructions. hydrodynamics. Yes Yes Yes
Need to specify proper
The terrain Special Manning’s n. This
inside the roughness zone method is only an
footprint of
obstruction No specific for approximation and
should be leveled requirement obstruction’s Low-resolution may not fully capture
Increased with its (effective n is mesh artificial hydrodynamics, the obstruction’s
Manning’s n surroundings. dependent) Manning’s n flood modeling hydrodynamic effect. Yes Yes Yes
Need to specify proper
A proper drag drag coefficient or
The terrain coefficient is formula. This
inside the needed for the method is only an
footprint of the
obstruction element(s)/ approximation and
should be leveled cell(s) Low-resolution may not fully
Additional with its No specific containing the hydrodynamics, capture obstruction’s
drag force surroundings. requirement obstruction. flood modeling hydrodynamics. Yes No Yes
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

48 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-1. Schematic view of the resolved versus unresolved pier for
elevated mesh method.

4.1.1.2 Drawback in Some 2D Models


During this study, the team identified a drawback in the tested 2D models regarding the
specification of the increased Manning’s n or drag coefficient for modeling obstruction. This
drawback applies to the increased Manning’s n method and the additional drag-force method.
Both methods try to add the extra resistance force experienced by flow due to the presence of
obstructions (Figure 4-2). From Newton’s Third Law, this extra flow resistance force equals the
drag force (FD) on an obstruction, which can be written as

1
FD = CD tU 2i A frontal (4-1)
2

CD is the drag coefficient, ρ is water density, Ui is the incoming flow velocity, and Afrontal is the
frontal area of obstruction. Using the increased Manning’s n approach, the equation is

gn 2
FD = t U 2c A (4-2)
K n2 h 1 3

where g is the gravitational acceleration constant, Kn is the unit system conversion factor in
Manning’s equation, h is water depth, Uc is the velocity at the cell center covered by an obstruc-
tion, and A is the area of the cell.
The drawback stems from how 2D models currently implement these increased or extra flow
resistance terms. Based on researchers’ understanding and verified by the simulation results,
none of the tested 2D models implements it correctly. Or, their implementation is correct for
their intended purposes, but they are used incorrectly. Regardless, using the increased Manning’s n

Figure 4-2. A scheme for the increased Manning’s n


method and the additional drag-force method.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   49

or the drag coefficient in the literature, the effective drag force simulated is underestimated. The
reason is as follows:
• For the increased Manning’s n method, the specified Manning’s n value does not determine the
obstruction-imparted flow resistance. From Equation 4-2, the resistance force also depends
on the cell-center velocity Uc and water depth h, which themselves depend on the specified
Manning’s n for that cell. So, there is a circular dependence that makes the specification of
the increased Manning’s n difficult. Unfortunately, a modeler must use the trial-and-error
method to tune the increased Manning’s n such that the result can satisfactorily match with
calibration data.
To solve this problem, 2D models would have to internally adjust the increased Manning’s n
at each time step such that the resulted drag force matches with its intended value.
• For the additional drag-force method, it was found that all the tested 2D models use the cell-
center velocity Uc in place of the incoming velocity Ui. These two velocities are fundamentally
different. In addition, this implementation creates a circular dependence among the drag
coefficient, cell velocity Uc, and frontal area Afrontal.
To solve this problem, 2D models would have to sample the incoming velocity at some
upstream distance or ask the modeler to specify a reasonable incoming velocity for an
obstruction.
This drawback is more severe when a refined mesh is used and the footprint of an obstruc-
tion covers more than one mesh cell. On the other hand, when a coarse mesh is used and an
obstruction is much smaller than the mesh cell that contains it, the problem is less severe. This is
because the total momentum of flowing water in that mesh cell is large, and the presence of the
obstruction has relatively small impact.
Because of the drawback described above, the results show that currently the increased
Manning’s n based on existing formulas and the drag coefficient in textbooks and hydraulic
manuals are lower than what they need to be, especially when the footprint of an obstruction is
larger than mesh cells.
All five methods will be evaluated in this chapter. Since the hole-in-mesh method is the best
method in capturing the hydrodynamics around obstructions, it will be evaluated first. The effect
of mesh resolution was studied to determine when an obstruction is geometrically resolved by
mesh. Two pier shapes and two pier sizes were considered. After finding the necessary criterion
for geometrically resolved, the calculated backwater was compared to the result using Yarnell’s
equation for verification. Then, the results of the hole-in-mesh method were used as a reference
to evaluate other methods.

4.1.1.3 Hole-in-Mesh Method Versus Unassigned Mesh Elements Method


Although SRH-2D supports both methods, RiverFlow2D supports only the hole-in-mesh
method. Neither is available in the HEC-RAS 2D model. In both methods, the obstruction geometry
is incorporated directly into the mesh. This is illustrated in Figure 4-3. From the figure, obstruc-
tion is included in the hole-in-mesh method as a void in the mesh (left) and the unassigned mesh
elements method by disabling the elements (right) within its footprint. For the unassigned mesh ele-
ments method, the mesh is generated for the entire domain, including the obstruction footprint.
However, within the footprint, the mesh has an unassigned material or is disabled. The difference
between these methods is further discussed in Section 4.1.3.

4.1.2 Hole-in-Mesh Method


The research team first investigated when an obstruction is considered resolved using the
hole-in-mesh method by evaluating the water-surface profile and the velocity field in the vicinity
of the obstruction.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

50   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-3.   Illustration of hole-in-mesh (left) and unassigned mesh element (right)
methods.

For this study, a rectangular channel with a width of 10 m, a length of 150 m, and a slope of
0.0005 was used. The upstream boundary condition was a constant inflow of 6 m3/s, and the
downstream boundary condition was normal depth (yn = 0.88 m for n = 0.03). Two pier shapes
and sizes were considered:
• Simple circular pier
– 1-m diameter pier, with 0.05-, 0.1-, 0.2-, 0.3-, 0.4-, 0.5-, 0.7- and 1-m mesh cell size.
– 2-m diameter pier, with 0.1-, 0.3-, 0.5-, 1-, and 1.5-m mesh cell size.
• Rounded-nose elongated pier
– 1-m pier with both constant- and variable-resolution meshes. For variable-resolution
meshes, the number of cells on the body wall part of the pier is different from the nose of
the pier. For constant-resolution meshes, the resolution is the same around the obstruc-
tions. The constant-resolution meshes have mesh sizes of 0.1, 0.3, 0.5, and 1 m. The variable
meshes have mesh sizes of 0.5 to 1 m, 0.5 to 1.5 m, and 0.5 to 3 m.
– 2-m pier with both constant- and variable-resolution mesh generation. The constant-
resolution meshes have mesh sizes of 0.1, 0.3, 0.5, 1, and 1.5 m. The variable meshes have
mesh sizes of 0.5 to 1 m, 0.5 to 1.5 m, and 0.5 to 3 m.
Without loss of generality, only the results for the cases of the circular pier with a diameter of
2 m are reported in this report. Appendix B includes detailed results and discussions for all other
cases. The modeling domain is shown in Figure 4-4. To study the effect of mesh resolution, the
cell size in the area around the pier (7 m upstream and 20 m downstream) is varied. The influence
zones upstream and downstream of the pier were determined based on initial run results. Two
11-m-long areas were used as a transition from fine mesh near obstruction to the coarse mesh
areas upstream and downstream, which have a resolution of 2 m.
Figure 4-4 shows that the effective pier shape, the pier geometry represented in 2D models,
depends on mesh resolution. It is interesting to observe that for this case, the effective shape of
a circular pier could be a hexagon or a square if the mesh is coarse, which directly shows the
importance of mesh resolution. More examples on the same topic can be found in Robinson et al.
(2019, Figure A.29).
Figure 4-5 shows the water-surface profiles along the channel centerline in the streamwise
direction for different mesh sizes (0.1, 0.3, 0.5, 1, and 1.5 m). The absolute difference of the

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   51

Figure 4-4.   The domains for mesh generation and the effective pier shapes with
different mesh resolutions.

water-surface profiles for each mesh resolution compared to the 0.1-m mesh resolution is shown
on the right.
For these cases, it can be observed that the changes in mesh size mostly affect the water-surface
profile upstream with minor effects downstream for subcritical flows. It means that the mesh size
affects the simulated head loss due to the pier. The increased cell size decreases the number of
points around the pier and affects local flow patterns. This is mainly because the effective pier
shape is different when different mesh resolutions are used. In this case, decreasing the number
of points around the pier creates more abrupt changes in the flow field that leads to more head
loss. It becomes more significant for larger piers relative to the channel width (i.e., larger pier
diameter/channel width ratio D/W) (See Appendix B).
Figure 4-5 illustrates that the resulting water-surface profiles from 0.3- and 0.5-m mesh resolu-
tions are close to the ones from 0.1 m. However, the runs with 1- and 1.5-m resolution meshes
have much larger errors. In Figure 4-6, the absolute and relative changes of water depth ∆y are
plotted. Although the water-surface profile from the mesh with 0.5-m resolution is close to that
from the 0.1-m resolution mesh (Figure 4-5), its relative error is around 25%. Therefore, if the
prediction of obstruction-induced backwater effect is the main purpose of modeling, then the
finer mesh is necessary.

Figure 4-5.   Mesh size effect on the water-surface profile (pier diameter is 2 m with hole-in-mesh method).
Results are from SRH-2D.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

52   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-6.   Effect of mesh resolution on calculated absolute and relative errors for water
depth Dy. The relative error on the right uses the result from the mesh with 0.1-m resolution.

Based on the results reported here and in Appendix B, researchers proposed using CSO as the
metric to measure whether an obstruction is geometrically resolved. A modeler may consider
the following to make sure that an obstruction is geometrically resolved if the hole-in-mesh
method is used:
• The CSO should be at least less than 0.5 to geometrically resolve an obstruction and to capture
the velocity field around it.
• Mesh refinement should also consider the width of an obstruction compared to the over-
all channel width and the distance between obstructions to capture the interactions among
obstructions and channel banks.
• If the pier has an elongated wall, then at least several (e.g., 3, 4, or 5) points are needed along
the wall (see Appendix B).
Based on the results of cases in this section, the mesh cell size of 0.3 m was considered suf-
ficient for piers of 1- and 2-m diameters. The results from this mesh resolution were used as a
benchmark to evaluate other methods.

4.1.2.1 Validation with Yarnell’s Equation


The research team compared 2D model results using the hole-in-mesh method with backwa-
ter from laboratory experiments to determine whether the hole-in-mesh method can accurately
predict the backwater generated by piers. In 2D models, the velocity is depth-averaged and verti-
cal velocities are not considered. Therefore, the complete flow field is not represented, and the
simulated results may differ substantially from reality. Yarnell’s equation (Equation 4-3) is con-
sidered a reasonably predictive formula because it is based on physical laws and calibrated with
laboratory data (Yarnell 1934). Yarnell’s equation has the following form:

Dy
= K ` K + 5Fr 2 - 0.6 j` a + 15a 4 j Fr 2 (4-3)
y

where Fr is the Froude number downstream of the pier, α is the contraction ratio (area of the
pier divided by the total area of cross section), and K is an empirical coefficient that is a function
of the pier shape, which can be determined using Table 4-2.
Figure 4-7 shows the comparison between Yarnell’s equation and SRH-2D results for α = 0.1
and α = 0.2, which correspond to the cases with pier diameters of 1 m and 2 m in a 10-m-wide

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   53

Table 4-2.   Bridge pier backwater coefficients


[after Yarnell (1934)].

Pier Shape K
Semicircular nose and tail 0.90
Lens-shaped nose and tail 0.90
Twin-cylinder pier with connecting diaphragm (L/D = 4) 0.95
Twin-cylinder pier without diaphragm (L/D = 4) 1.05
90° triangular nose and tail 1.05
Square nose and tail 1.25

rectangular channel, and the K value of 0.9 based on Table 4-2. The figure also shows the experi-
mental data of Charbeneau and Holley (2001).
The comparison in Figure 4-7 shows that the SRH-2D model results agree well with Yarnell’s
equation. The estimated relative water-depth change ∆y/y by SRH-2D tends to be slightly higher
than Yarnell’s equations for Fr values around 0.2 to 0.3. A similar trend is observed in the experi-
mental data in Charbeneau and Holley (2001). These results confirm that SRH-2D with the
discussed mesh resolution can capture the hydrodynamics of the pier with reasonable accuracy.

4.1.3 Unassigned Mesh Elements Method


The unassigned mesh method is theoretically the same as the hole-in-mesh method. How-
ever, in this method, the mesh should initially be generated for the pier footprint, then the
pier elements are disabled. This method is available in SRH-2D but not in HEC-RAS 2D and
RiverFlow2D.
In Figures 4-8 and 4-9, the water-surface profiles and the velocity along the channel center­
line from the unassigned mesh are shown along with results from hole-in-mesh and elevated
mesh methods. The elevated mesh method will be discussed later. The unassigned and hole-in-mesh
methods have almost the same results. However, their velocity results in SMS are visually dif­
ferent. This should not cause any alarm. The difference is because of the postprocessing procedure

Yarnell, α = 0.2
Yarnell, α = 0.1
SRH2D, α = 0.2
SRH2D, α = 0.1
Charbeneau
and Holley (2001),
α = 0.108

Figure 4-7.   Comparison relative water-depth change Dy/y among


Yarnell’s equation, SRH-2D, and lab data.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

54   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-8.   Comparing the WSE of unassigned and elevated mesh methods with the hole-in-mesh method.

for calculating velocity results at nodes from cell centers. In SMS, for the hole-in-mesh method,
the perimeter is automatically considered as a wall with a zero velocity assigned. However, in an
unassigned mesh, the cell-centered velocity is assigned to the nodes (see Figure 4-10).

4.1.4 Elevated Mesh Elements Method


In this method, an obstruction is incorporated directly into the terrain. For this purpose, the
cells inside the footprint of the obstruction are elevated to be higher than the water surface. This
method also requires a detailed mesh. Because of the near-vertical surface created inside the
mesh at the obstruction location, the results may have additional errors (see Figure 4-8 and
Figure 4-9). It is clear from these figures that the WSE and velocity near the pier calculated by this
method are different from the unassigned mesh elements and hole-in-mesh methods.
To have a better understanding and comparison of the velocity field around the pier, Fig-
ure 4-11 shows velocity distribution calculated with hole-in-mesh and elevated mesh elements

Figure 4-9.   Comparing the velocity profiles of unassigned and elevated mesh methods with the hole-in-mesh
method.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   55

Figure 4-10.   Velocity magnitude contour around the pier for the hole-in-mesh
(left) and unassigned mesh elements (right) methods.

methods. Though comparable, the results are different. The wake area behind the pier from
elevated mesh is larger than the one from the hole-in-mesh method.
In this approach, the mesh should be sufficiently refined around the perimeter of the pier to
capture the large change of elevation across the edge (see Figure 4-12b). It is recommended
to consider two breaklines for resolving a pier, one inside and one outside the pier footprint
(named as the double-breakline approach). The distance between the two breaklines should
be neither too small nor too large. Figure 4-12 shows the effective terrains due to the use of a
single breakline or double breaklines; the pier is resolved better using the double-breakline
approach.
One commonly cited feature of HEC-RAS 2D is its subterrain treatment, which needs some
special considerations in its application of the elevated mesh approach. In HEC-RAS 2D, the
terrain data can be edited directly to raise the pixels of the GeoTiff, or the pier can be stamped
into the terrain. What HEC-RAS 2D really models, and consequently how much flow resistance
due to the bridge is resolved, depends on the mesh resolution relative to the pier size. To dem-
onstrate the point, a simple example follows.

Figure 4-11.   Velocity magnitude contours around the pier for hole-in-mesh
(left) and elevated mesh (right) methods.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

56 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-12. Comparing double


breaklines (a) and a single breakline
(b) for resolving pier.

A cylindrical pier is shown in the terrain in Figure 4-13a. To illustrate how this pier could be
modeled in HEC-RAS 2D, three different scenarios (out of many) are shown in Figure 4-13b
to 4-13d. The first scenario is that the mesh is coarse and no breakline is used so that the pier
is totally within one cell (Figure 4-13b). In this case, the faces and nodes of that cell will get no
information from the elevated pier. The pier’s effect is only in the cell’s volume–stage relationship.
In other words, the pier directly affects only the mass balance, not flow resistance force.
The second scenario is shown in Figure 4-13c. It only uses one breakline, which is within the
footprint of the pier. This scenario is better than the first. However, the resulted shape of the pier
in the mesh is not cylindrical, and the footprint is much larger than the real pier footprint.
The third scenario is shown in Figure 4-13d. It uses the double-breakline method with one
breakline within the footprint of the pier and the other outside. The topography represented by
the mesh is similar to the terrain shown in Figure 4-13a. If the detailed hydraulics and flow field
around the pier are important, this scenario is recommended in HEC-RAS 2D.
In conclusion, one should consider that the elevated mesh method approximates the hole-in-
mesh method, and the results near the pier are sensitive to the alignment of nodes to the edge of
the pier. Therefore, this method is recommended only when the hole-in-mesh and unassigned
mesh elements methods are not feasible. It is also noted that the elevated mesh method is not the
same as including the raised elevation in the terrain. The mesh needs certain refinement, and its
edges and nodes need to align with the obstruction to geometrically resolve it.

4.1.5 Increased Manning’s n Method


When it is not possible to have refined mesh to resolve obstructions or when the 2D model does
not support other methods, the increased Manning’s n method is an alternative. In this method,
the mesh cells covered by an obstruction’s footprint are assigned with a higher Manning’s n value.
This method can be used with SRH-2D, HEC-RAS 2D, and RiverFlow2D. In this section, this
method is evaluated by comparing it with the hole-in-mesh method.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 57

Figure 4-13. Simple pier in HEC-RAS 2D: (a) terrain, (b) coarse
mesh with no breakline, (c) single breakline, and (d) double
breaklines.

First, the most important question is what Manning’s n value to use. The research team studied
the effect of different n values on the results in comparison with the hole-in-mesh method results.
To make a fair comparison, the same mesh with a resolution of 0.3 m was used. The diameter
of the pier is 2 m.
The simulated water-surface profiles using Manning’s n values of 0.32, 0.5, 1, 10, and 100 are
shown in Figure 4-14. Again, these Manning’s n values are assigned only to the cells covered
by the pier footprint. The value of n = 0.32 resulted in a comparable water-surface profile with
the hole-in-mesh method. However, near the pier, the error increases. With larger n values, the
backwater curve deviates more from the hole-in-mesh method results and stabilizes after n = 10.
It means that n values of 10 and 100 produce similar results.
As it is depicted in the figure, the WSEs near the pier have some errors that become more
prominent with increasing Manning’s n. In the case of the calculated velocity (Figure 4-15),
this error becomes more significant, especially downstream of the pier. As shown in the figure,
the velocity error can go up to 600%. The reason for this large error is that for moderately low
Manning’s n, flow can still go through the pier, instead of going around the pier if using the hole-
in-mesh method. With the hole-in-mesh method, the low-velocity wake behind the pier can be
properly captured, which is not the case with increased Manning’s n unless the n value is quite
large (e.g., 100).
It is more straightforward to inspect the velocity contours around the pier for the hole-in-mesh
method along with the ones for increasing Manning’s n method with n values of 0.32, 1, and 100.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

58 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-14. Effect of n, applied at pier footprint, on WSE (D 5 2 m, mesh


resolution 5 0.3 m).

The comparison is shown in Figure 4-16. For all n values, the simulation results around the pier
are different than those predicted by the hole-in-mesh method. Extremely large n values effec-
tively block the flow within the pier footprint and produce flow fields bearing some resemblance
to those from the hole-in-mesh method. Nonetheless, the details of flow contraction and wake
dynamics are still quite different.
All these results are based on SRH-2D’s implementation on increased Manning’s n method,
which has the drawback described previously. Theoretically, to block the flow within the foot-
print of an obstruction, the n value should be infinity. In computer models, infinity cannot be
assigned. Instead, a finite but quite large value can be used.
Even when the drawback is addressed, there are still extremely severe limitations for the
increased Manning’s n method. Suppose a properly increased Manning’s n is found in a correctly
implemented 2D model that can capture the right drag force imparted by an obstruction. The
increased Manning’s n does not prevent flow from going through the obstruction. This through-
flow, as seen Figure 4-16 for n = 0.32 and 1, changes the hydrodynamics around the obstruction.
Because of the throughflow, there is less flow going around the obstruction, and subsequently
the velocity magnitude there is lower than it should be. Previously it was stated that the case
with n = 0.32 produces the best water-surface profile when compared with the hole-in-mesh
method. However, the velocity distribution with n = 0.32 is off because it cannot prevent flow

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 59

Figure 4-15. Effect of n, applied at pier footprint, on velocity (D 5 2 m, mesh


resolution 5 0.3 m).

Figure 4-16. Effect of n, applied at pier footprint, on velocity


distribution around the pier in comparison with hole-in-mesh
method.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

60 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

from going through the pier. On the other hand, the case with n = 100 produces the most reason-
able velocity field.
Based on the above, the increased Manning’s n method cannot fully capture the local hydro-
dynamics around an obstruction. Specifically, it cannot capture the dynamics of the water-surface
profile and velocity at the same time. Calibration to match the water-surface profile and velocity
separately will result in different n values. However, if used properly, this method is useful if local
hydrodynamics is not important.

4.1.5.1. Effect of Mesh Resolution on Increased Manning’s n Method


This section will address the effect of mesh resolution on the results of increased Manning’s n
method. The Manning’s n was fixed at 0.32, and the mesh resolutions of 0.3, 0.5, 2, and 3 × 4 m
(3 m and 4 m in different directions) were studied. The pier diameter is 2 m.
The water-surface and velocity profiles at the center of the channel for each mesh resolution
are shown in Figures 4-17 and 4-18. In these and subsequent figures, the stepped water surface
is due to the sampling algorithm. It does not reflect a stepped water surface in the model results.

Figure 4-17. Effect of mesh resolution on WSE with the increased Manning’s n
method.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 61

Figure 4-18. Effect of mesh resolution on velocity profile with the increased
Manning’s n method.

From the figures, the results of larger 3 × 4 m are far from the others, and it produces a much
higher backwater. This is because with this cell size an area is allocated to the high Manning’s n,
and, therefore, the resulting head loss will be higher. In other words, the real area AE is higher than
the assumed footprint area in calculating n = 0.32 using Equation 4-4. Therefore, for this mesh
resolution, a lower Manning’s n is needed, and the modeler needs to calculate a proper n value
based on cell size (see Section 4.1.5.2).
The modeler also should be aware of how a 2D model specifies cells with increased Manning’s n.
For example, in SMS a computational cell will be assigned to the material only if its center is
inside that coverage. Consider a mesh with an average cell size of 1 m, as shown in Figure 4-19.
If a pier of 1-m diameter (centered at one of the grid points) is defined as a new material with a
high Manning coefficient, then SMS will assign no cells to this material. For all four cells to be
considered with high Manning’s n, the radius of the pier should be at least 0.8 m.
In summary, the increased Manning’s n method is sensitive to cell size. The proper Manning’s n
value depends on the total area of cells covered by an obstruction. More discussions will be pro-
vided in Section 4.1.5.2.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

62 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-19. Schematic view of how SMS assigns cells to the high Manning’s n
material.

4.1.5.2 Comparing Hole-in-Mesh with the Method in HDS 7


This section examines an increased Manning’s n over the width of the channel instead of at
the pier footprint only. First outlined in HDS 7 Chapter 6 (Zevenbergen et al. 2012), the method
calculates an effective or equivalent to Manning’s n that is meant to produce a shear force that
reproduces the drag force from a pier (or group of piers) and spreads that force over an element
(or group of elements). The HDS 7 equation has the form of

K 2u y 1 3 / C D A P
ne = n2 + (4-4)
2g / A E

where ne is the effective Manning’s n of the elements or group of the elements. n is Manning’s
coefficient of the channel bed, which has a value of 0.03 in the test cases. CD is the pier drag
coefficient; its values can be found in textbooks and hydraulics manuals. Table 4-3 is one such
example. AP is the projected area of the pier in the direction perpendicular to the flow; AE is the
area of a select element or group of elements that are covered by an obstruction; y is the aver-
age flow depth; g is the gravitational constant; and Ku is 1.486 in U.S. customary units [1.0 in SI
(International Systems of Units)].
For the comparison, the following four different cases were considered:
1. D = 1 m, Manning’s n increased over the pier footprint (AE = 0.785 m2).
2. D = 2 m, Manning’s n increased over the pier footprint (AE = 3.14 m2).
3. D = 1 m, Manning’s n increased over the width of channel (AE = 10 m2).
4. D = 2 m, Manning’s n increased over the width of channel (AE = 20 m2).

Table 4-3. Drag coefficient values for different


pier shapes (from SMS v13.1 manual).
Type of Object Drag Coef�icient (CD)
Long streamlined body 0.10
Hollow semisphere facing stream 0.38
Solid hemisphere 0.42
Sphere 0.50
Cube 0.80
Thin disk 1.10
Square �lat plate at 90° 1.70
Circular cylinder 1.0–1.3
Hollow semicylinder opposite stream 1.20
Long �lat plate at 90° 1.98

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 63

Considering the normal depth y of 0.88 m and the drag coefficient of 1.2, the effective
Manning’s n values for the above four cases are calculated as 0.258, 0.184, 0.078, and 0.078,
respectively. A mesh of 1-m resolution was used.
During the study, researchers found the effect of how an obstruction is aligned with the mesh
cells. The original mesh is in a way that the center of the pier is on one of the grid points (see
Figure 4-19). The cells assigned the high Manning’s n by SMS are highlighted in yellow. For a pier
with a diameter of 2 m, four cells are selected. However, no cell is selected to the high Manning’s n
value for the pier with a 1-m diameter. To solve this problem, the mesh was adjusted in a way that
the center of the pier coincides with the center of one of the cells (adjusted mesh on the right). In
Figure 4-19, one cell is assigned the high Manning’s n for the pier with a 1-m diameter. However,
with this adjusted mesh only two cells are selected for the pier with a 2-m diameter. It is clear
here that the result will depend on how the obstruction is aligned with mesh. In the following,
the original mesh was used for the 2-m pier and the adjusted mesh was used for the 1-m pier.
In Figures 4-20 and 4-21, the results for the pier with the 2-m diameter are shown. As noted
previously, for the original and adjusted meshes, SMS assigned four and two cells to n = 0.184,
respectively. The result shows that the simulated backwater with the adjusted mesh is lower than
the original mesh. The n value of 0.184 is calculated by assuming that the area of selected mesh cells
is equal to the obstruction’s footprint area (AE = 3.14 m2). However, for the original and adjusted
meshes, the real AE is 4 m2 and 2 m2, respectively, which results in different flow resistance.
For accuracy, a modeler should always check which cells are selected and calculate their area.
Then, with that area, the correct Manning’s n can be calculated using Equation 4-4.
In Figures 4-22 and 4-23, the results for the case using HDS 7 method (increased Manning’s n
over the channel width) are shown for comparison, which shows that applying the increased n
either at the footprint or over the channel width leads to a similar backwater curve at the center
of the channel.
However, though it is not shown here, the cross-sectional water-surface profile is different
because the HDS 7 method spreads the flow resistance across the whole channel width.

Figure 4-20. Effect of how to distribute Manning’s n (only in obstruction’s


footprint or across channel) on WSE, D 5 2 m.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

64 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-21. Effect of how to distribute Manning’s n (only in obstruction’s footprint


or across channel) on velocity, D 5 2 m.

Figure 4-22. Effect of how to distribute Manning’s n (only in obstruction’s footprint


or across channel) on WSE, D 5 1 m.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 65

Figure 4-23. Effect of how to distribute Manning’s n (only in obstruction’s


footprint or across channel) on velocity, D 5 1 m.

Comparing the water-surface profiles with the hole-in-mesh method shows that for both piers
the estimated Manning’s n using Equation 4-4 is not high enough to reproduce the water-surface
profile. Based on the results of this section, the following recommendations are provided:
• If the bridge’s local hydraulics is the primary interest, the increased Manning’s n method is
not recommended because it cannot fully capture the hydrodynamics. Instead, the hole-in-
mesh, unassigned mesh, or elevated mesh methods should be used. On the other hand, if local
hydraulics is not the primary interest, the increased Manning’s n method may be used.
• For pile or trestle piers that consist of several small piers that cannot be resolved in the mesh,
calculate their combined ne using Equation 4-4, and spread their flow resistance over the
channel width (see Figure 4-24).
• If the river cross section at the bridge location has areas with different n values or the piers
are only in some portion of the channel width, then spread pier effects over the affected area
only (see Figure 4-24).

Figure 4-24. Schematic view of handling multiple small piers and large piers.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

66 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

4.1.6 Additional Drag-Force Method


The additional drag-force method is another method that can be used when the hole-in-mesh
method and unassigned mesh elements method are not feasible. In this method, an obstruction
is defined by its location, size, elevation, and drag coefficient. Then, the model adds an extra
drag term to the cells containing the obstruction. The mathematical formula for the additional
drag-force method has been described in Section 4.1.1.2. Currently, SRH-2D and RiverFlow2D
support this method. HEC-RAS 2D does not.
Again, as described previously, there is a drawback in the current implementation of the addi-
tional drag-force term in SRH-2D and RiverFlow2D. There is a circular dependence of the drag
coefficient and the velocity and water depth at a cell’s center. As a result, the drag coefficient in
the literature (see Table 4-3) is not enough to correctly produce the drag force. These values in
the table will always result in a lower drag force. Because of the similarity between this method
and the increased Manning’s n method, the limitations of both methods are the same.
Nonetheless, the effect of drag coefficient values on the results of the drag method is investi-
gated and compared with the hole-in-mesh method. The same mesh of 0.3 m was considered for
both methods. The diameter of the pier is 2 m.
The simulated water-surface profiles using drag coefficient values of 4, 10, 100, and 300, along
with the ones simulated with the hole-in-mesh method, are shown in Figure 4-25. The value
of CD = 4, which is higher than all the recommended values in Table 4-3, is needed to have the
closest water-surface profile to the hole-in-mesh method. As with the increased Manning’s n
method results, increasing the drag coefficient value will produce higher water-surface profiles.

Figure 4-25. Effect of drag coefficient on the water-surface profiles.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 67

Like the increased Manning’s n method, the additional drag-force method cannot fully capture
the local hydrodynamics around the pier. Indeed, it cannot capture the water-surface profiles and
velocity distribution at the same time. This is evident in Figures 4-26 and 4-27. Although the drag
coefficient value of 4 produces the best water-surface profile along the centerline, its velocity field is
way off. On the other hand, the drag coefficient value of 300 produces the worst water-surface pro-
file among all. However, its velocity field is the best in comparison with the hole-in-mesh method.
In conclusion, the additional drag-force method also cannot capture the local hydrodynamics
around an obstruction. Specifically, it cannot capture the dynamics of water-surface profile and
velocity at the same time. Calibration to match either water-surface profile or velocity will result
in different drag coefficient values. However, if used properly, this method is still useful if local
hydrodynamics is not of importance. This recommendation is in line with information in the
FHWA’s Two-Dimensional Hydraulic Modeling for Highways in the River Environment—
Reference Document (Robinson et al. 2019).

4.1.6.1 Effect of Mesh Resolution on Drag Coefficient Values


Mesh resolution has some effects on the actual drag force added. Four different meshes were used:
• Triangular, 0.3 m: The mesh is made of triangles with an average size of 0.3 m.
• Triangular, 0.5 m: The mesh is made of triangles with an average size of 0.5 m.
• Rectangular, 2 m: The mesh is made of rectangles with an average size of 2 m.
• Rectangular, 3 × 4 m: The mesh is made of rectangles with the size of 3 m by 4 m.
For comparison, the hole-in-mesh method with an average cell size of 0.3 m was used.

Figure 4-26. Effect of drag coefficient on the velocity profile.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

68   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-27:   Effect of drag coefficient on velocity distribution around


the pier.

The simulated WSE and velocity with different meshes are shown in Figures 4-28 and 4-29 for
CD = 4. Figures 4-30 and 4-31 show the results for CD = 100.
The results show that, like the increased Manning’s n method, the area of the selected cells
and the drag coefficient value are important. Different 2D models may select cells differently.
For example, in SRH-2D, if a cell’s center is within the boundary of obstruction, it is selected.
Thus, a modeler should be aware of how many cells are selected. Because of the similarity between
the increased Manning’s n method and the additional drag-force method, detailed discussion
on the results shown in Figures 4-28, 4-29, 4-30, and 4-31 are omitted.

4.1.7 Comparison Between SRH-2D and RiverFlow2D


In this section, the results of the hole-in-mesh method using RiverFlow2D and SRH-2D are
compared. Like the previous sections, a rectangular channel with a width of 10 m, a length
of 150 m, and a slope of 0.0005 was used. The upstream boundary condition was a constant
inflow of 6 m3/s, and the downstream boundary condition was fixed water depth at normal
depth (yn = 0.88 m for n = 0.03). A circular pier of 2 m in diameter was placed in the channel.
First, the channel without the pier using 0.5-m and 2-m resolution meshes was modeled. The
results of the water-surface profile are shown in Figure 4-32. The absolute error in comparison
with the 0.5-m resolution mesh is calculated for each model (the blue and black solid lines on
the right figure). The results show that SRH-2D is less sensitive to mesh resolution, and the abso-
lute error is less than 0.001 m. RiverFlow2D is slightly more sensitive, with a maximum error
of 0.005 m. With 0.5-m resolution meshes, the difference between RiverFlow2D and SRH-2D
is less than 0.002 m (the black dashed line in the right subfigure of Figure 4-32). To reduce the
difference, a slightly more refined mesh is needed for RiverFlow2D.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   69

Figure 4-28.   Effect of mesh resolution on WSE with the additional drag-force method, CD 5 4.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

70   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-29.   Effect of mesh resolution on velocity with the additional drag-force method, CD 5 4.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   71

Figure 4-30.   Effect of mesh resolution on WSE with the additional drag-force method, CD 5 100.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

72 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-31. Effect of mesh resolution on velocity with the additional drag-force method, CD 5 100.

Figure 4-32. Comparing RiverFlow2D and SRH-2D when there is no pier.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 73

Figure 4-33. WSE simulated by RiverFlow2D and SRH-2D in the presence of pier.

In the next step, the presence of the pier was simulated in both SRH-2D and RiverFlow2D.
The modeling domain is the same as shown in Figure 4-4. The mesh resolution around the pier
is 0.3 m to geometrically resolve the obstruction.
The water-surface and velocity profiles simulated by both models are shown in Figures 4-33
and 4-34 along with the absolute error of the RiverFlow2D results in comparison with the ones
from SRH-2D. The results show that there is some difference between the two models. RiverFlow2D
calculates a lower WSE upstream of the pier. It is important to note that RiverFlow2D does not
include a turbulence model. Consequently, turbulence dissipation and energy loss can be reflected
only in Manning’s n coefficient. In contrast, SRH-2D considers turbulence. In fact, there are two
options for turbulence in SRH-2D: the depth-averaged parabolic model and the two-equation k − ε
(Rodi 1993).
Researchers hypothesize that the difference in results is mainly caused by the fact that
there is no turbulence model in RiverFlow2D. To test that hypothesis, the turbulence model
in SRH-2D was disabled. The parabolic model was adopted, in which the eddy viscosity is
calculated as νt = CtU*h. Here, U* is the shear velocity, and h is water depth. The model con-
stant Ct ranges from 0.3 to 1.0, with a default value of 0.7. To fully turn off the turbulence, Ct
was set to 0 (in fact, the minimum allowable value in SMS is 0.01). The results are shown in
Figures 4-33 and 4-34.

Figure 4-34. Velocity profile simulated by RiverFlow2D and SRH-2D in the presence
of pier.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

74   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

The figures show that when the turbulence is turned off in SRH-2D, the simulated results are
similar to those from RiverFlow2D. A similar conclusion can be drawn from Figure 4-35, where
the velocity magnitude contours around the pier are plotted.

4.1.8 Comparing HEC-RAS 2D and SRH-2D in Modeling Pier


Using Elevated Mesh Method
In this part of the study, a similar example of a pier with a 2-m diameter in a simple channel
was simulated to compare the elevated mesh method results between SRH-2D and HEC-RAS
2D. The channel is 150 m long and 10 m wide. The discharge and downstream WSE are 6 m3/s
and 0.88 m, respectively.
The modeling domain is the same as shown in Figure 4-4. The cell size of the area around the
pier (7 m upstream and 20 m downstream) is changed from 0.3 m to 1 m and 1.67 m for both
models to study the effect of cell size. The influence zones upstream and downstream of the pier
were determined based on initial run results. Two 11-m-long areas were used as a transition
from fine mesh near the obstruction to the coarse mesh areas upstream and downstream, which
have a resolution of 2 m. The created meshes around the pier for HEC-RAS 2D and SRH-2D are
shown in Figure 4-36.
The modeled terrain by the fine meshes in HEC-RAS 2D and SRH-2D is plotted in Figure 4-37.
For the two specific meshes generated in this study, it seems the terrain in SRH-2D is slightly
smoother than that in HEC-RAS 2D. Perhaps because of this, when the 0.3-m resolution mesh is
used, HEC-RAS 2D results show more head loss and higher WSE upstream of the pier.
Figure 4-38 shows the resulting WSE in the middle of the channel for each case. SRH-2D is
more sensitive to the cell size, and as cell size increases, the predicted head loss due to the pier
increases. The head loss increase is because of the final represented topography to SRH-2D. As
demonstrated in Figure 4-39, the pier in SRH-2D is a square pyramid with a base size of 2.36 m
× 2.36 m, which is larger than the original pier size and leads to more head loss. The use of the
double breaklines method may improve the result (see Section 4.1.9). In comparison, HEC-
RAS 2D is less sensitive to the mesh resolution, and the results for 1-m and 1.67-m cell sizes are
approximately the same. It seems that for this example, the subgrid treatment of HEC-RAS 2D
compensates for the coarse mesh’s poor representation of the topography. However, it is impor-
tant to note that if the velocity distribution around the pier needs to be known, then a finer mesh
resolution is necessary.
This example shows that the subgrid terrain treatment in HEC-RAS makes its result less
sensitive to coarse mesh resolution; however, this does not necessarily mean its result is better.
The subgrid terrain treatment does not fully capture the hydrodynamic effects of obstructions.

Figure 4-35.   Velocity distribution around the pier using SRH-2D and RiverFlow2D.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   75

Figure 4-36.   Comparison of the generated meshes using


SRH-2D (left) and HEC-RAS 2D (right).

Figure 4-37.   Comparison of the generated topography with a


cell size of 0.3 m using SRH-2D and HEC-RAS 2D.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

76   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-38.   Comparison of the calculated water-surface profile using SRH-2D (left) and
HEC-RAS 2D (right) for different cell sizes.

4.1.9 Use of Double Breaklines in HEC-RAS 2D for Bridge Piers


Currently, HEC-RAS 2D does not support the hole-in-mesh method to model bridge piers.
The alternative is the elevated mesh method, where the bridge is embedded in the terrain. In this
method, it is critical to have a refined mesh around the piers even with HEC-RAS 2D’s subgrid
terrain treatment. Breaklines can be strategically placed around the piers to instruct the mesh
generator to put mesh nodes there.
Experience has shown that it is preferable to use two breaklines, one inside and one outside
the pier footprint (named as the double-breakline approach). The distance between the two
breaklines should be neither too small nor too large. The double-breakline approach is also
recommended for other 2D models when the elevated mesh method is used for obstructions.
To investigate the combined effect of breaklines and subgrid terrain treatment in HEC-RAS
2D, a simple case of flow around a circular bridge pier was studied. The pier has a diameter of 2 m
and is in the middle of a rectangular channel with a length of 150 m and a width of 10 m. The inlet
on the left has a discharge of 6 m3/s, and the outlet on the right has a specified WSE of 0.88 m.
Two background meshes were used: one with a cell size of 2 m (coarse) and the other with a
cell size of 0.5 m (fine). For each mesh, three different scenarios of breaklines were studied: no
breakline, single breakline, and double breaklines. The use of breaklines, especially when double
breaklines are used, makes the mesh around the pier refined and better defines the elevated

Figure 4-39.   The represented topography to SRH-2D using 1.67-m cell size.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   77

Figure 4-40.   HEC-RAS 2D: velocity around a circular pier with a coarse mesh
and breaklines.

terrain covered by the pier. For comparison, SRH-2D was also used to simulate the same case
with the hole-in-mesh method (the mesh resolution is much refined, so its result is treated as
the ground truth).
The HEC-RAS 2D results are shown in Figures 4-40 to 4-43. With a coarse mesh and no
breakline, the pier is missed by the model. This is because of the location of the pier, which
is totally within a mesh cell and HEC-RAS 2D’s subgrid terrain algorithm does not “see” it.
To improve, the pier has to intersect the cell faces. In this case, with no breakline in the coarse

Figure 4-41.   HEC-RAS 2D: WSE around a circular pier with a coarse mesh
and breaklines.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

78   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-42.   HEC-RAS 2D: velocity around a circular pier with a fine mesh
and breaklines.

Figure 4-43.   HEC-RAS 2D: WSE around a circular pier with a fine mesh
and breaklines.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   79

Figure 4-44.   Comparison of velocity results between hole-in-mesh (SRH-2D)


and elevated mesh methods with double breaklines (HEC-RAS 2D).

mesh, velocity and WSE look as if no pier exists (top figures in Figures 4-40 and 4-41). With one
breakline and double breaklines, the simulated flow field shows the blocking effect of the pier,
though the results do not look accurate. The result with double breaklines seems slightly better.
When a fine mesh is used, HEC-RAS 2D’s results are much better. Figures 4-42 and 4-43
show the simulated velocity and WSE, respectively. With a fine mesh, it seems HEC-RAS 2D’s
subgrid terrain algorithm does a good job in capturing the blocking effect of the pier. For the
velocity result, the flow contraction and the wake behind the pier are well captured, regardless
of whether a breakline is used or how many are used. More difference can be observed in WSE
results.
To further appraise the results, the velocity and WSE from the hole-in-mesh method (simulated
with SRH-2D) and the double-breakline approach with the elevated mesh method (simulated with
HEC-RAS 2D) are plotted in Figures 4-44 and 4-45. Although not exact matches, both give
comparable results. The most noticeable differences are the wake length behind the pier and the
WSE contours on the two sides of the pier. Nevertheless, with a fine mesh and double breaklines,
HEC-RAS 2D can produce reasonable local hydraulics results around bridge piers. In the future,
it will be valuable to have more rigorous benchmarking tests for the application of HEC-RAS 2D
in simulating local hydraulics around obstructions.

4.2 Bedforms
A simple and common flow resistance factor is bedforms. This section investigates the effects
of bedforms in 2D models. This part of the project was originally designed with three cases
(Cases 10, 11, and 12). However, to fully understand the dynamics, the team decided that it was
necessary to greatly increase the number of cases (to around 100).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

80   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-45.   Comparison of WSE results between hole-in-mesh (SRH-2D) and elevated
mesh methods with double breaklines (HEC-RAS 2D).

In the preparation of all cases, different artificial terrains with bedform were firstly generated
using pyHMT2D as follows:
• Bedform perpendicular to flow direction,
• Bedform perpendicular to flow direction + perturbation,
• Bedform rotated 30 degrees,
• Smooth slope channel (with no bedform), and
• Smooth slope channel + perturbation.
A channel with a background slope of 0.0005 was considered, and the bedform’s height and
length were calculated based on the empirical formulas as 0.5 m and 10 m, respectively. To
investigate natural irregularities and LiDAR errors, random perturbations in [−0.1 m, 0.1 m]
were added to the terrain for some cases. Figures 4-46, 4-47, and 4-48 show the contour map,
bed profile, and cross section of these terrains.
The river channel domain size was set as 500 m long and 10 m wide. The discharge and down-
stream WSE were set to 20 m3/s and 2 m, respectively. Based on a hypothetical sediment size
of 200 μm, the background Manning’s n corresponding to the (grain) skin friction was set as
0.013 (van Rijn 1984). To investigate the effect of mesh resolution, eight mesh resolutions of
20, 40, 60, 80, 100, 150, 300, and 1,000 cm were considered. As an example, Figure 4-49 shows
how different meshes represent the terrain.
Mesh resolution and how mesh faces and nodes are aligned with ground features influence
how the rough terrain is modeled. This has a direct impact on the model result and Manning’s n.
In addition, some 2D models, such as HEC-RAS 2D, have the subgrid terrain treatment. Theo-
retically, these types of models should show their advantages when coarse meshes are used. One
goal is to test whether this is the case.

4.2.1 Effect of Mesh Resolution


This section explores the effect of mesh resolution. The research team primarily compared
SRH-2D and HEC-RAS 2D; for a fair comparison, the meshes were made identically for both

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   81

Figure 4-46.   Created terrains for studying the effect of bedforms: bedform perpendicular
to flow direction, bedform perpendicular to flow direction with perturbation, and bedform
rotated 30 degrees.

models. To calculate the equivalent of Manning’s n in the channel, a 1D backwater curve solver
was used. The following is the basic methodology used:
1. Meshes with different resolutions were created in HEC-RAS 2D and translated into SRH-2D
format using pyHMT2D.
2. Next, simulations were run with HEC-RAS 2D and SRH-2D.
3. The water-surface profiles were then extracted from 2D results along the center of the channel.
4. Finally, the 1D backwater curve equation was solved in the 1D channel with the same overall
slope but without the bedforms. For each case, Manning’s n in the 1D equation was adjusted
so its water-surface profile matched those from 2D models. The calibrated 1D Manning’s n
is the total equivalent roughness from resolved and unresolved parts of Manning’s n. It is
noted that here the resolved Manning’s n in 2D models depends on the mesh resolution. The
unresolved Manning’s n is just the Manning’s n value of 0.013 specified for all cases.
As a first test, 2D models were run to simulate water-surface profiles over the bedforms with
no perturbation. The water-surface profiles simulated by HEC-RAS 2D and SRH-2D are shown
in Figure 4-50 for example mesh resolutions (20, 100, 300, and 1,000 cm). The calibrated total
equivalent Manning’s n values are also shown.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

82   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-47.   Created terrains for studying the effect of bedforms: smooth slope channel and
smooth slope channel with perturbation.

Figure 4-48.   The longitudinal and cross-sectional profiles of the created terrains.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   83

Figure 4-49.   Schematic view of how mesh represents the terrain.

Figure 4-50.   Comparison between HEC-RAS 2D and SRH-2D. The equivalent Manning’s n values backcalculated
from a 1D backwater curve solver are also shown.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

84   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-51.   The effect of mesh resolution on the total equivalent Manning’s n for HEC-RAS 2D and SRH-2D.

Figure 4-50 shows that the simulated WSE profiles using SRH-2D and HEC-RAS 2D are not
the same. For some mesh resolutions, HEC-RAS 2D has a higher WSE. In other cases, it is the
opposite. Consequently, the total equivalent of Manning’s n for each model with different mesh
resolutions is different. For example, the equivalent n value of the bedforms with a mesh resolu-
tion of 20 cm using SRH-2D and HEC-RAS 2D are 0.023 and 0.02, respectively.
By repeating this method for all mesh resolutions, the effect of mesh resolution on the total
equivalent of Manning’s n for each 2D model can be investigated. In Figure 4-51, the results are
shown: the graph on the right is for HEC-RAS 2D, and the graph on the left is for SRH-2D. For
one terrain, the equivalent Manning’s n is different for each mesh resolution. In addition, the
change of equivalent Manning’s n with the mesh resolution is not monotonic. These findings are
true for HEC-RAS 2D and SRH-2D.
To further illustrate the relationship between mesh resolution and equivalent Manning’s n,
the simulated results are collectively shown in Figure 4-52. This figure shows the equivalent
Manning’s n for each mesh resolution. These results and the implication for Manning’s n follow.
The research team originally hypothesized in the project proposal that when a mesh gets
coarser it will miss more terrain features and the resolved part of the Manning’s n will decrease.
As a result, the total equivalent Manning’s n from the 2D model results should decrease. Note
that the total equivalent Manning’s n referred to is assumed to be made of two parts: resolved
and unresolved. The unresolved Manning’s n, due to grain (skin) friction, was kept constant at
0.013 in these cases. From Figure 4-52, both SRH-2D and HEC-RAS 2D show the hypothesis
is inconclusive. The hypothesis is approximately true only when mesh resolution is exceedingly
coarse, coarser than 100 cm (except for HEC-RAS 2D at 300 cm).
The causes are discussed in the following:
• A coarser mesh may miss some features, but it can also misrepresent the terrain. The mis­
representation introduces error in the calculation of the bed slope and gravity term in the
governing equation. This misrepresentation is in all 2D models. The problem is more severe
when the background slope is smaller, in other words, when the erroneous slope because of
terrain misrepresentation dominates over the true bed slope and other terms. Slope calcula-
tion of the models and their effect will be discussed in more detail when the effect of perturba-
tion is discussed.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   85

Figure 4-52.   Equivalent Manning’s n for different mesh


resolutions based on HEC-RAS 2D and SRH-2D.

• 2D models cannot capture some flow features such as flow separation in the vertical direction
over bedforms, which is one major mechanism for head loss flow resistance. Thus, no matter
how refined the mesh is, the flow will not separate over the bedform in 2D models. As a result,
2D models cannot be expected to capture more flow resistance simply by refining the mesh. In
other words, even if the terrain is fully resolved geometrically by a mesh, it is not necessarily
fully resolved hydraulically.
Nonetheless, when the mesh resolution is not too coarse (e.g., cell size less than 150 cm), the
average equivalent of Manning’s n is around 0.03, which agrees with those from empirical flow
resistance formulas, for example, the formula in van Rijn (1984) gives a Manning’s n of 0.035.
This finding was used as a foundation for the research team’s proposed method to approximately
specify Manning’s n in 2D models.

4.2.2 Effect of Perturbation in the Terrain Data


Terrain data usually contain perturbations. The effect of perturbation in the terrain data was
examined by adding an artificial perturbation in [−0.1 m, 0.1 m] to the background terrain data
(see Figure 4-53). Random perturbations were added to mimic the noise and inaccuracy in
LiDAR data as well as natural fluctuations in terrain. Theoretically, they increase the roughness
and, thus, Manning’s n value.
The total equivalent Manning’s n was calculated for bedforms with perturbation and without
perturbation using the same method as in the previous section. The results are shown in Fig-
ure 4-54. For these cases, SRH-2D responds to the terrain perturbation slightly more than
HEC-RAS 2D. This is contrary to the expectation because HEC-RAS 2D implements the subgrid
terrain algorithm and is supposed to see more terrain variations (perturbation noises in this case).
Figure 4-53 shows clearly that the mesh resolution changes the shape of bed profiles sig-
nificantly. Each resolution produces a mesh that represents different versions of the terrain. It
seems the 100-cm resolution has slightly higher bedform peaks; therefore, it is rougher (a higher
equivalent Manning’s n). For the 300-cm resolution, the bedform peaks are much lower, and,
therefore, a lower equivalent Manning’s n. For the 1,000-cm resolution, the profile is close to a

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

86   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-53.   The longitudinal bed profiles from different


mesh resolutions and their calculated equivalent Manning’s
n values. The terrain is bedform with perturbations.

straight line. That is why its equivalent Manning’s n is 0.015, which is close to the background
value of 0.013.
In the next step, the effect of the perturbation in a channel without any bedform (a smooth
channel) was studied. The resulting bed profiles and equivalent Manning’s n values for different
mesh resolutions are shown in Figure 4-55. The figure shows that higher resolution results in
higher frequency oscillations, which are translated to a higher equivalent n value. For the coarsest
resolution of 1,000 cm, the equivalent Manning’s n is the specified Manning’s n value of 0.013,
which means the resolved Manning’s n using this mesh resolution is almost zero. All these results
agree with expectations.
The total equivalent Manning’s n values from both 2D models for the smooth channel with
and without perturbations are shown in Figure 4-56. Again, SRH-2D responds to the terrain
perturbation more than HEC-RAS 2D, which is contrary to expectation. However, the general-
ization of this result needs more validation cases in the future.

Figure 4-54.   Effect of perturbation on equivalent Manning’s n in HEC-RAS 2D and SRH-2D.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   87

Figure 4-55.   The shape of bed profiles from different


mesh resolutions for a channel without bedforms.

In general, the random perturbation in terrain with and without bedforms slightly increases
the equivalent Manning’s n (not all cases). Although it was expected that HEC-RAS subgrid ter-
rain treatment captures the bed variation better, this is not the case based on the tests that were
conducted. SRH-2D has a slightly stronger response to the perturbation. It may be explained by
the following:
• The flow resistance term is not the dominating term in the governing equations for these cases.
• The misrepresentation of terrain and the bed slope by the discrete meshes contaminated the
results (applies to all 2D models).
The effect of bed-slope calculation in 2D models and how the bed slope depends on mesh
resolutions and perturbations is explored next.
Bed-slope calculation is different in SRH-2D and HEC-RAS 2D, and it influences the results.
SRH-2D uses vertex elevations to get cell-center elevation, and then cell-center elevations are
interpolated to faces, which are in turn used to evaluate the bed slope at cell centers using the
Gauss theorem. The calculation stencil involves many vertexes (the current cell and its neighbors).

Figure 4-56.   The total equivalent Manning’s n for a smooth channel with and without perturbations.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

88   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

A larger stencil means less sensitivity to noise in elevation data. In contrast, the bed slope at a face
in HEC-RAS 2D uses the center elevations of only two adjacent cells, which can be extremely
sensitive to noise. The high sensitivity of the bed-slope calculation to noise in terrain data could
potentially contaminate the solution.
To demonstrate the difference, Figure 4-57 compares the slopes calculated by HEC-RAS 2D
and SRH-2D (a Python script was written based on the numerical methods described in their
references and user manuals). Figure 4-57 shows the results for bedforms with and without per­
turbation. The two models work similarly when there is no perturbation; however, HEC-RAS 2D
is more sensitive to noise when there is a perturbation in terrain data. This is evident from the
higher amplitude oscillations in its calculated bed slope.

4.2.2.1 Effect of Terrain Perturbation Noise Distribution


The effect of terrain perturbation noise distributions is addressed in this section. The previous
section used uniform distribution [noise is uniform within (−perturbation, perturbation)]. Pre-
vious studies in literature have shown the noise or error in LiDAR data may follow a normal distri-
bution. The research team decided to run additional cases to investigate the effect of terrain noise
distributions. In the new cases, the perturbation follows a normal distribution, in which average
and STDs are zero and perturbation/1.96 [at 95 percentiles, which means 95% of the noise added
is within (−perturbation, perturbation)], respectively. Therefore, there is a 5% chance that the

Figure 4-57.   Comparison of the calculated bed slope in SRH-2D and HEC-RAS 2D
for terrain data: (a) without perturbations and (b) with perturbations.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   89

noise amplitude exceeds perturbation, and the noise added is skewed toward the mean of zero.
Figure 4-58 shows the comparison between uniform and normal distributions.
Both channels with and without bedforms that have perturbations were simulated again. Their
bed profiles with normal and uniformly distributed perturbations are also shown in Figure 4-58.
From the longitudinal bed profiles, it seems normally distributed perturbation does have some
large excursions at some places. However, generally, it is hard to tell the difference.
The cases were rerun with the newly generated terrain that has normally distributed perturbations.
Only the mesh resolution of 60 cm is used because this resolution is reasonable to resolve bedforms.
The results are shown in Figures 4-59 and 4-60. The simulated profiles from HEC-RAS 2D and
SRH-2D are almost the same. The conclusion is that terrain perturbation distribution type does
not have a significant impact on these cases.

4.2.2.2 Effect of Mesh Boundary Shifting


The boundary of mesh, in conjunction with breaklines, grid spacing, and other meshing deci-
sions, determines where mesh grid points are located and what terrain elevations these grid
points will recognize. Therefore, it may affect the results. This issue is significant because two
modelers may come up with two different meshes for the same project.
This section examines this effect by shifting the mesh by 1 m while keeping everything else the
same. Only one case (bedform with perturbation) was rerun. A mesh resolution of 60 cm was
used. The comparison is shown in Figure 4-61. From the comparison, observations indicate that
the result does not change. This could be because of the high resolution of 0.6 m for this case. If
the resolution is low, the locations of the grid points might have more impact.

Figure 4-58.   Normal versus uniform distribution for terrain perturbation.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-59.   Effect of perturbation type on SRH-2D and HEC-RAS 2D results for terrain with bedforms.

Figure 4-60.   Effect of perturbation type on SRH-2D and HEC-RAS 2D results for terrain without bedforms.

Figure 4-61.   Effect of 1-m shift of mesh boundary for the terrain with bedforms and perturbation case. The red
and blue lines are the simulated WSEs.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   91

Figure 4-62.   Bed profiles from different mesh resolutions


for a channel covered with rotated bedforms.

4.2.3 Effect of Terrain Alignment


The research team studied the effect of the bedform rotation on the results in addition to the
cases where the bedforms are perpendicular to flow direction. The reason to do this is to check
the effect of HEC-RAS 2D’s subgrid terrain treatment. When the cell faces are not aligned with
bedforms, the face lines are curves in 3D that should be captured by HEC-RAS 2D’s subterrain
treatment; therefore, it should recognize more roughness. If this is true, HEC-RAS 2D should
have higher resolved Manning’s n and a higher total equivalent Manning’s n than SRH-2D. In
addition, the subgrid terrain treatment should be more important when the mesh gets coarser.
Next the team considered the cases of bedforms with no perturbation. The terrain was rotated
by 30 degrees. Figure 4-62 shows bed profiles with rotated bedforms and different mesh resolu-
tions. Contrary to expectations, the results from HEC-RAS and SRH-2D are comparable, which
means the subgrid terrain treatment does not play a dominant role in this case (see Figure 4-63).
From the figure, bedform rotation reduces the equivalent Manning’s roughness in HEC-RAS and
SRH-2D (except when the mesh is too coarse with a resolution of 1,000 cm). The reduction is for
this specific case only; it may not be generalized to other cases.

Figure 4-63.   The total equivalent Manning’s n for a channel covered with rotated bedforms.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

92   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

4.2.4 Dependence of Manning’s n on Relative Submergence


The roughness felt by the flow also depends on the relative submergence (water depth/roughness
height). In general, the lower the relative submergence, the higher the Manning’s n. To demonstrate
this point, the “bedform + perturbation” case was rerun using mesh resolution of 60 cm for dif-
ferent downstream WSEs of 1, 1.5, 2, and 3 m (approximately equal to water depth because the
bed elevation there is almost at 0 m). Everything else in the model domain remained the same as
it was before the rerun. The results are shown in Figure 4-64. When the water depth gets larger,
the simulated water-surface profile gets smoother (less influenced by the bottom roughness ele-
ments). In addition, the total equivalent of Manning’s n increases as the relative submergence
decreases, which is as expected.

For extremely low relative submergence (1 m over 0.4 m), the total equivalent to Manning’s n
goes as high as 0.045. However, for extremely low relative submergence, the 1D backwater curve
with a constant Manning’s n is not accurate, and it cannot capture the WSE profile curvature,
especially near the downstream boundary. This is because Manning’s n should be a function
of water depth, not a constant value. On the other hand, when the relative submergence is rea-
sonably high (in this case greater than 5 = 2 m/0.4 m), the total equivalent Manning’s n can be
treated as a constant (see Figure 4-65). Over the range of the other flows, Manning’s n varied by
only 0.002.

For more general cases, a threshold of relative submergence should be used to judge whether
roughness elements are fully submerged or not. For example, Kramer (2005) proposed the
threshold value of 3. When the relative submergence is less than 3, the protruding effect of the
bed roughness is important, and the form drag starts to dominate. For low relative submergence,
the Manning’s n cannot be approximated as a constant.

4.2.5 Preliminary Methodology for Specifying


Manning’s n in 2D Models
Previous sections addressed the effect of models, mesh resolution, perturbations, and bedforms
orientation. Many insights have been gained through these exercises; however, they still cannot be

Figure 4-64.   The effect of downstream water depth on WSE profile for bedforms with perturbation.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   93

Figure 4-65.   The total equivalent Manning’s n versus


downstream water depth.

used directly to determine what Manning’s n to use in 2D models. In practice, a modeler will face
the following two questions:
• What is the real, total Manning’s n for a zone in the simulation domain?
• What is the Manning’s n value that needs to be specified for that zone in a 2D model?

It is important to note that the total Manning’s n for an area or zone in a simulation domain
is different from what the modeler should assign to that zone in a 2D model. In the physical world,
different roughness features (e.g., bedforms and vegetation) determine the total Manning’s n.
In the numerical modeling world, terrain and mesh only partially resolve these features. The
resolved part of the roughness should be excluded from Manning’s n value specified in the model.
In other words, based on its resolution, a mesh can resolve some part of the real geometry and
represent it in a model. The remaining part is unresolved. The Manning’s n value specified in a
2D model should be for the unresolved part only.
The relationship among resolved and unresolved Manning’s n and the total n should be the
following:

n = n resolved
2
+ n un
2
resolved (4-5)
In this case, n is the total Manning’s roughness value. If there is an estimate of the total
Manning’s n and the resolved Manning’s n, then the unresolved Manning’s n to be specified in
2D models can be calculated as follows:

n unresolved = n 2 - n resolved
2 (4-6)
For real-world cases where there might be multiple roughness zones, it is not easy to use the
method in the previous section to estimate resolved parts from the simulation results since these
zones interact with each other. However, for simple cases, there is a way to calculate the resolved
part of Manning’s n, which will be explained as follows.
For a simple channel flow covered with bed features (e.g., bedforms), Manning’s n can be set as
zero (or a small number to make the model stable) in 2D models. Then, the equivalent Manning’s n

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

94 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

can be calculated using a simple 1D backwater curve solver. (The n value in the 1D backwater
solver is adjusted until the 1D WSE profile fits the 2D model result.) In this way, the calculated
equivalent Manning’s n is due to the resolved part. Now having nresolved and total n, nunresolved can
be estimated using Equation 4-6.
Equation 4-6 is further examined to find how a method for selecting appropriate Manning’s n
value for 2D models can be developed. The research team hypothesized that when the mesh
resolution increases (finer mesh), the mesh has a better representation of the terrain; therefore,
the resolved Manning’s n will be higher. As a result, the unresolved part decreases. This is sche-
matically shown in Figure 4-66.
To test this hypothesis, researchers reran the “bedform perpendicular to flow + perturbation”
cases with zero Manning’s n, then the resolved Manning’s n was calculated as described. Fig-
ure 4-67 shows the calculated resolved Manning’s n values for SRH-2D and HEC-RAS 2D models.
The resolved Manning’s n does not monotonically decrease when mesh gets coarser, as hypoth-
esized. However, for reasonably resolved meshes (e.g., for mesh resolution less than 150 cm and
greater than 20 cm), the resolved Manning’s n values are roughly constant.
If the total Manning’s n is estimated to be 0.035 (van Rijn 1984), then the unresolved part of
Manning’s n can be calculated using Equation 4-6. The results are shown in Figure 4-68.
The calculated unresolved Manning’s n values were then specified in 2D models, and the
cases were rerun. In theory, if these unresolved Manning’s n are used with their corresponding
meshes, the simulated result will recover the total Manning’s n of 0.035. The results are shown
in Figure 4-69 for both models. It is clear from the figure that the equivalent of Manning’s n is
indeed close to the target value of 0.035. This means the square root of summing Manning’s n
squares in Equation 4-5 works as expected.
Although this method works well for simple cases, for real-world cases where there might
be multiple roughness zones, the method of using the simple backwater curve in a single-zone
channel does not work. This is because researchers cannot simulate a backwater curve in a com-
plex real-world case and attribute the aggregated flow resistance to different roughness zones.
So how should a modeler know what the resolved Manning’s n is for a roughness zone in a
given mesh? In other words, how does a modeler know how much of the roughness is resolved by

Figure 4-66. The hypothesized


response of resolved and unresolved
Manning’s n values to mesh resolution.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-67.   Calculated resolved Manning’s n versus


mesh resolution.

Figure 4-68.   The resolved and unresolved Manning’s n for different mesh resolutions for HEC-RAS and SRH-2D.
The total Manning’s n is 0.035, and the unresolved part is calculated from Equation 4-6.

Figure 4-69.   The calculated total equivalent Manning’s n


when the unresolved Manning’s n is specified in 2D models.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

96   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 4-70.   The percentage of unresolved Manning’s n versus the percentage of the resolved Manning’s n.
The left figure shows the two zones on two ends of the quarter-unit circle. The right figure shows where the
simulated cases are on this quarter-unit circle.

the mesh and how much is not? This is a difficult question, and more research would be needed
to fully answer it. However, some insights can be gained from the analysis performed in this task.
Equation 4-7 shows how the relationship among resolved, unresolved, and total Manning’s n can
be rewritten:
J n resolved N2 J n unresolved N2
KK OO + KK OO = 1 (4-7)
L n P L n P
The plot of this equation shown in Figure 4-70 is a quarter-unit circle. For a refined mesh that
resolves a large portion of roughness (e.g., the red area where the ratio between the resolved and
total Manning’s n values is larger than 0.8), the unresolved Manning’s n drops quickly. On the
other hand, for a coarse mesh that does not resolve a significant portion of the roughness (e.g.,
the blue area where resolved/total Manning’s n is less than 0.6), the unresolved Manning’s n
remains in a narrow region (80%–100% of the total n). These observations are extremely infor-
mative in devising a method.
In Figure 4-70, the simulated cases in this part have also been shown. Except for the extremely
coarse mesh of 1,000-cm resolution (which probably should never be used in reality for the bed-
form feature in this task), the resolved Manning’s n is about 60%–80% of the total Manning’s n,
and the unresolved Manning’s n is about 50%–80% of the total Manning’s n. For the extremely
coarse mesh of 1,000-cm resolution, the unresolved Manning’s n is about 96% to 97% of the total
Manning’s n.
To put things into perspective, if unresolved/total Manning’s n is 0.7, modelers need to reduce
the total Manning’s n by 30% and specify that reduced value in 2D models.
As seen from the right subfigure in Figure 4-70, there is a weak correlation between resolved/
total Manning’s n and mesh resolution. However, in general (with exceptions), resolved/total
Manning’s n increases when the mesh resolution increases. Thus, modelers should decrease the
value of unresolved/total Manning’s n when the mesh resolution increases. Based on these find-
ings, the following general guidelines can be provided:

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   97

• In all simulated cases, the bedform has a wavelength of 10 m. It is reasonable to argue that
if there are more than 6–7 grid points per bedform (i.e., mesh cell size is at least less than
150 cm), the bedform is geometrically resolved by the mesh.
• For all the cases where the bedform is resolved (i.e., mesh cell size is 150 cm or less), the
average unresolved Manning’s n is approximately 70% of the total Manning’s n. A modeler
can use this as a starting point to calibrate Manning’s n. The coarser the mesh is when com-
pared to the resolved threshold, the higher the percentage of unresolved Manning’s n to the
total Manning’s n and vice versa.
• For all cases where the bedform is not well resolved [i.e., mesh cell size is 300 cm or more (less
than 6–7 points per bedform feature)], the unresolved Manning’s n is more than 80% of the
total Manning’s n. A modeler can use this value as a starting point to calibrate the Manning’s n.
The coarser the mesh, the higher the percentage.
• In 2D models, Manning’s n is not only a physical parameter but also a numerical one. There-
fore, the exact meaning of Manning’s n used in 2D models has blurred between a physical
parameter for flow resistance and a numerical parameter. The Manning’s n used in 2D models
is to compensate for the part of flow resistance unresolved by mesh and terrain, and it is a
function of mesh resolution and terrain accuracy. Thus, a modeler should keep this in mind
when calibrating Manning’s n.
• It is also important to be careful when using the calibrated Manning’s n, which is for the
unresolved part only. The total Manning’s n should be typically higher than the value used in
the model. For example, if sediment transport is also simulated, the wall shear (flow resistance)
is a key variable. A modeler should be aware that the reported wall shear in 2D models may only
contain the unresolved part. In addition, sediment transport is induced by skin friction, not form
drag. How to extract the required skin friction out of the simulated flow resistance is beyond the
scope of this project.
• The relationship between mesh resolution and terrain accuracy and the unresolved Manning’s n
seems to make 2D modeling uncertain and hard to evaluate. Two modelers using two different
combinations of mesh and terrain and unresolved Manning’s n may get similar flow results.
This is not surprising at all.
• When feasible and necessary, a finer mesh is better than a coarser mesh because of numeri-
cal accuracy considerations. However, one should also avoid excessively fine meshes. Results
from an extremely refined mesh do not necessarily mean they are better than those from a
reasonably less refined mesh.
• When assigning Manning’s n in a 2D model, keep in mind that in a sense it is only a tuning
parameter. What is more important is the combined effect of the assigned Manning’s n and
the mesh and terrain. The combined effect should reach the targeted total Manning’s n. Cali-
bration with measured data can help inform whether the targeted total of Manning’s n is
reached.
• The ratio between unresolved and total Manning’s n is theoretically between 0 and 1. In the
calibration process, if one finds that the unresolved Manning’s n specified in a model greatly
exceeds the targeted total Manning’s n, this is a signal that there may be other issues with the
model, such as inadequate terrain or the mesh is too coarse to capture important features.
• In a simulation domain with multiple roughness zones, the zones that are more engaged in the
flow, have large areas, and have more impact on the overall hydraulics need to be identified. The
calibration of Manning’s n should be focused on these zones, starting with the ones conveying
more flow. For example, roughness zones at the edge of flood inundation areas may have less
impact on the flow than the zones immediately in the flood path. Changing Manning’s n for
edge zones may not have a significant response at calibration locations. In this case, excessively
changing Manning’s n values in the edge zones may make them exceed the targeted total of
Manning’s n, which is not correct. The focus of calibration should be on the zones that are more
engaged with the flow.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

98   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

• Automatic calibration is implemented in many 2D models and their graphical user inter-
face software. The automatic calibration uses optimization algorithms to search for the best
Manning’s n values in their specified ranges. When possible, automatic Manning’s n calibra-
tion is recommended over manual calibration. If automatic calibration is used, a reasonable
range should be specified by the modeler such that the optimization algorithm gives a proper
Manning’s n solution.

4.2.6 Investigating Standard Deviation as a Measure


for Terrain Information Unresolved by Mesh
In the process of quantifying the unresolved Manning’s n, the research team also tried to
define a measure or index for the unresolved part of the terrain by a mesh. The STD of the eleva-
tion difference between terrain and mesh was one such measure.
First, the bed elevation difference between terrain and mesh was calculated using a Python
script, then the STD was calculated. The STD was used as a measure of how much terrain informa-
tion is missed by a mesh. Figure 4-71 shows an example for a mesh size of 300 cm. This example
uses the bedform with perturbations. This procedure is repeated for other mesh resolutions, and
the calculated STD is plotted against mesh resolution in Figure 4-72.
From the figure, it is clear the STD increases as the mesh gets coarser, which means the mesh
misses more terrain information. Comparing the STD with the unresolved Manning’s n, it is
also evident that STD is roughly proportional to unresolved Manning’s n, although a functional
relationship is not available at this point. There are methods for gravel beds that link Manning’s n
to STD, such as that in Chen et al. (2019). However, the elevation difference field in this example
does not behave like gravel beds. The use of the formula in Chen et al. (2019) gives unreasonable
Manning’s n values.

Figure 4-71.   The terrain, elevation represented with a 300-cm resolution


mesh, and their differences.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 99
Standard deviation of bed elevation difference

0.175

0.150
between terrain and mesh (m)

0.125

0.100

0.075

0.050

0.025

0.000
20 40 60 80 100 150 300 1000
Mesh resolution (cm)

Figure 4-72. Effect of mesh resolution on the STD of bed elevation differences (left) and Manning’s n (right).

The STD values for mesh resolutions 20, 40, 60, 80, 100, and 150 cm are comparable. This
explains the fact that the corresponding unresolved Manning’s n values are roughly compa-
rable, too.

4.2.7 Comparing Different Models in Modeling Bedforms


In this section, HEC-RAS 2D, SRH-2D, RiverFlow2D, and HEC-RAS 1D are compared in
modeling flow over bedforms. The previously created artificial terrain for perturbed bedforms
perpendicular to flow direction was used for this part. The channel domain size was set as
500 m × 10 m. The discharge and downstream WSE were set to 20 m3/s and 2 m, respectively.
The research team first studied the effect of mesh resolution on each 2D model. Then, the results
of the models are compared with 0.2- and 2.5-m mesh resolutions.

4.2.7.1 Effect of Mesh Resolution


Three different mesh resolutions (0.2, 2.5, and 10 m) were studied. However, RiverFlow2D
diverged with the 10-m mesh resolution; a 5-m mesh resolution was used instead.
Figure 4-73 shows how HEC-RAS 2D and 1D respond to different mesh resolutions. At first,
the default option of HEC-RAS 1D for contraction–expansion coefficients (0.1 and 0.3) was

Figure 4-73. Effect of mesh resolution on HEC-RAS 2D (left) and HEC-RAS 1D (right).
The effect of contraction–expansion coefficients in HEC-RAS 1D is also shown.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

100 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

used. The result shows that the 1D WSE results are much higher than the HEC-RAS 2D results,
which means the default contraction–expansion coefficients are too high for this case. When
both coefficients are set to zero, the resulted WSE profiles are more comparable with 2D results.
Figure 4-74 shows how SRH-2D and RiverFlow2D respond to different mesh resolutions.
From Figures 4-73 and 4-74, the following can be observed:
• For the cases tested here, RiverFlow2D is the least sensitive model to mesh size. It is because
triangular cells are used in RiverFlow2D (in comparison with the rectangular cells used in
SRH-2D and HEC-RAS 2D). To verify this, the SRH-2D model was run again using a 2.5-m
triangular mesh (see Figure 4-74). The sensitivity decreases significantly and the calculated
WSE using 2.5-m-resolution triangle mesh is closer to the one from the refined mesh with
0.2 m resolution. If everything else is equal, the triangle mesh is more refined when compared
to the rectangle mesh.
• HEC-RAS 2D is less sensitive to mesh size in comparison with SRH-2D, especially for mesh
resolutions less than 2.5 m. It may be due to the subgrid terrain treatment of HEC-RAS 2D.
However, in contrast to SRH-2D, decreasing the mesh resolution from 2.5 m to 0.2 m leads to
lower WSE in HEC-RAS 2D. This may be caused by how bed slope is calculated (as discussed
previously).

4.2.7.2 Comparison Among Models


The mesh resolutions of 0.2 m and 2.5 m were selected to compare the models. The results
are shown in Figure 4-75. The simulated WSE using RiverFlow2D is lower than that used in
other models. This is because currently RiverFlow2D does not include a turbulence model (see
Section 4.1.7).
For the mesh resolution of 2.5 m, HEC-RAS 2D, HEC-RAS 1D, and SRH-2D with a triangular
cell mesh produce similar results. SRH-2D using a rectangular cell mesh with the same mesh
resolution produces a slightly lower WSE. When the mesh cell size decreases to 0.2 m, the simu-
lated WSE by SRH-2D increases. HEC-RAS 1D does not show significant sensitivity to mesh
resolution when contraction–expansion coefficients are set to zero.
It is clear from the results that different models may produce different results. For one model,
different meshes will produce different results. These results are not surprising, and they should
be considered in assigning Manning’s n values. More discussion will be provided for real-world
examples.

Figure 4-74. Effect of mesh resolution on SRH-2D (left) and RiverFlow2D (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice 101

Figure 4-75. Comparing the WSE calculated by different models with mesh
resolutions of 2.5 m (left) and 0.2 m (right).

4.3 Channel Irregularities


The cross-sectional shape of streams varies. The variation may be induced by natural irreg-
ularities or by constructed structures, such as bridges and culverts. They cause contraction–
expansion head losses, which affect Manning’s n.
2D model equations consider 2D mass and momentum balance. If the terrain and mesh
are sufficiently refined to capture the cross-sectional changes (geometrically and hydraulically
resolved), then 2D models should be able to capture the effect of channel irregularities, and they
should not be explicitly considered in Manning’s n.
In 1D models, the head loss induced by channel irregularities is computed with the contraction–
expansion coefficients. The head loss is calculated as in Equation 4-8:

a 1 V 21 a 2 V 22
h ce = C - (4-8)
2g 2g

where hce is head loss, C is the contraction or expansion coefficient, V1 and V2 are the mean velocity
in two consecutive cross sections, and α1 and α2 are the energy coefficients.
In HEC-RAS 1D, the default values for contraction–expansion coefficients are 0.1 and 0.3,
respectively (Brunner 2021). In an extensive study by the HEC, using 3 real bridges and 76 idea-
lized bridge sites, the estimation of the head loss using 1D models was investigated, and the
results are published in USACE HEC Research Document 42 (Hunt 1995). The summary of this
research was also published in the HEC-RAS reference manual (Brunner 2021). Though their
research provided valuable insight into the expansion and contraction reach lengths, the regres-
sion analysis was not successful to estimate precisely the contraction–expansions coefficients.
Table 4-4 shows the contraction–expansion coefficients suggested in the HEC-RAS manual.
Here, a simple example of contraction and expansion is used to compare the results of 1D and
2D models. For this study, a rectangular channel with a width of 10 m, a length of 150 m, and a
slope of 0.0005 was used. The upstream boundary condition was a constant inflow of 6 m3/s, and
the downstream boundary condition was fixed at normal depth (yn = 0.88 m for n = 0.03). The
width of the channel at the center is decreased to 4 m and is increased back to 10 m using two
transitions with a length of 5 m. The modeled contraction and expansion in 1D and 2D models
using 0.5-m resolution is shown in Figure 4-76.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

102 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Table 4-4. Subcritical flow contraction–


expansion coefficients (Brunner 2021).

Type of transition Contraction Expansion


No transition loss computed 0.0 0.0
Gradual transition 0.1 0.3
Typical bridge section 0.3 0.5
Abrupt transition 0.6 0.8

Figure 4-76. Schematic view of 1D and 2D domain for contraction and expansion.

In Figure 4-77, the water-surface profile inside the channel in presence of the contraction
and expansion from HEC-RAS 2D is shown as a blue line. For comparison, the results from
HEC-RAS 1D with four contraction–expansion coefficient pairs of (0, 0), (0.05, 0.15), (0.1, 0.3),
and (0.3, 0.5) are also shown. The result shows that the contraction–expansion coefficients have
a significant effect on the simulated water-surface profile. Depending on what contraction–
expansion coefficients are used, HEC-RAS 1D may overestimate or underestimate the head loss
in comparison with HEC-RAS 2D.
In reality, a modeler must pay attention to defining contraction–expansion coefficients for
bridges and culverts. Default values are often used for natural irregularities in channel cross

Figure 4-77. Comparing HEC-RAS 1D and HEC-RAS 2D in modeling


contraction and expansion.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   103

sections. HEC-RAS 1D applies these default contraction–expansion coefficients (0.1 and 0.3,
respectively) whenever there are velocity head changes between two consecutive cross sections.
For most cases, it may need smaller coefficient values to model the irregularity. In other words,
the HEC-RAS 1D model with the default contraction–expansion coefficient may overestimate
the head loss. Consequently, HEC-RAS 1D may produce higher WSEs for most subcritical flows
in comparison with those from HEC-RAS 2D using the same Manning’s n.
Furthermore, the contraction–expansion coefficients in the 1D model are applied to the cross
sections only, and the irregularities between two cross sections are not considered.
It is impossible to provide a definitive guideline here regarding channel irregularities in 1D
and 2D models. The reason is because the contraction–expansion coefficients and Manning’s n
affect the result simultaneously.

4.4 Stage and Discharge Relations


and Available Guidelines
When the relative submergence with respect to roughness height decreases, flow resistance
generally becomes more pronounced. This was discussed in Section 4.2.4. As a result, in most
streams, the Manning’s n value increases with the decrease in discharge and stage (Chow 1959).
However, other factors may also come into play (e.g., land cover change across the channel and
flow regime change when stage changes).
For a channel with uniform roughness, the relation between depth and Manning’s n can
be derived as in Chow (1959). Based on Keulegan’s equation for rough channels, mean flow
velocity, V, can be written as follows:
J RN
V = Vf KK6.25 + 5.75 log OO (4-9)
ks
L P
where R is hydraulic radius, and ks is roughness height. The friction velocity, Vf , can be estimated
as gRS ,
where S is the channel slope for uniform flow or energy slope for gradually varied flow. Using the
Chezy equation, the following equation holds:
V C
= (4-10)
Vf g
where C is the Chezy coefficient. Substituting Equation 4-10 into Equation 4-9, the Chezy
coefficient for rough bed can be calculated as follows in Equation 4-11:
J 12.2R N
C = 5.75 g log KK O (4-11)
ks O
L P
On the other hand, with the Chezy and Manning equations, the following relation between C
and Manning’s n can be calculated as in Equation 4-12:
kn R 1 6
C= (4-12)
n
Finally, the Manning’s n can be written as a function of R, and ks is obtained as follows in
Equation 4-13:
kn R 1 6
n= J 12.2R N (4-13)
5.75 g log KK O
k O
L s P
This equation is only for the Manning’s n induced by surface grain roughness.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

104 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Equation 4-13 can be used to appreciate the effect of stage, discharge, and depth on Manning’s n.
Figure 4-78 (left) shows the n values versus R for different ks values of 0.002, 0.007, 0.015, 0.03,
and 0.1 m. Figure 4-78 (right) shows n versus R/ks. For all ks values, Manning’s n is not a constant.
However, for each ks, the n value is in a narrow range (except when the relative submergence is
small). Therefore, the recommendation is that the dependence of Manning’s n on stage and dis-
charge may be of secondary importance except for relatively shallow flows. From the two figures,
a rudimentary guideline is that when R/ks is larger than 5–10 (approximately), the Manning’s n
due to surface grain roughness can be treated as a constant.
The above recommendation is based on the classic equations for grain roughness. Thus, it is not
applicable when other roughness and flow resistance factors are engaged when the stage changes.
For example, in the presence of vegetation, usually on floodplains, the value of Manning’s n is a
function of the vegetation submergence. Another example is mobile bed rivers, where the bed may
transition to different regimes depending on flow and stage.

4.5 Vegetation
The contribution of individual plants or groups of plants to flow resistance has been exten-
sively studied. Vegetation flow resistance is determined by computing the exerted drag force and
then translating it into a coefficient of Manning or Chezy (Shields Jr. and Gippel 1995, Shields
et al. 2017). The drag force can be calculated as in Equation 4-14:
tCD A v U 2
F= (4-14)
2
where ρ is the density of water, CD is the drag coefficient, Av frontal vegetation area, and U is the
flow velocity. Vegetation can be categorized as rigid and flexible plants. Currently, it is still dif-
ficult to accurately calculate drag coefficient and frontal vegetation area. It is even more difficult
for flexible vegetation, which will bend and reconfigure under high-flow velocity. In general,
flow resistance due to flexible vegetation is higher under low flow velocity than higher velocity.
Laboratory and field investigations of drag associated with vegetation can be dated back to
the early work of Ree and Palmer (1949). Since then, many laboratory and field studies have
been conducted (Petryk and Bosmajian 1975; Kouwen and Li 1980; Burke and Stolzenbach
1983; Eckman 1990; Abdelsalam et al. 1992; Nepf 1999; Järvelä 2005; López and García 2001;
Nikora and Nikora 2007; Tanino and Nepf 2008; Nepf 2012). Most investigations have sought
generalized relationships between vegetation density, submergence, and drag force. It has been

Figure 4-78. Effect of hydraulic radius R and roughness height ks on Manning’s n.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   105

argued that when deeply submerged, vegetation can be treated as large-scale surface rough-
ness. Conversely, emergent or shallow submerged vegetation alters the flow dynamics such that
resistance is more significant (Nepf 2012).
Recently, Shields et al. (2017) reviewed the state of knowledge and proposed an adjustment to
Manning’s n for vegetation flow resistance. Among others, they compared the following methods
for rigid and flexible plants:
• Methods for rigid vegetation
– Arcement and Schneider (1989)
– Baptist et al. (2007)
– Luhar et al. (2008)
– Luhar and Nepf (2013)
– Katul et al. (2002)
• Methods for flexible vegetation
– Freeman et al. (2000, 2002)
– Järvelä (2004)
– Whittaker et al. (2015)
The results of their comparison for both flexible and rigid plants are shown in Figure 4-79. They
concluded it is hard to give clear guidelines for selecting one approach over another. However,
they did provide some suggestions for selecting Manning’s n for rigid and flexible vegetation.
Their suggestions are summarized in Sections 4.5.1 and 4.5.2.

4.5.1 Rigid Vegetation


For rigid vegetation, Shields et al. (2017) stated that the method in Arcement and Schneider
(1989) is limited to emergent vegetation with exceedingly low velocity and relatively shallow
channels. On the other hand, the method in Luhar et al. (2008) is strictly limited to situations
where flow through patches of vegetation is negligibly small relative to adjacent unvegetated

Figure 4-79.   Comparing different methods for estimating Manning’s n for rigid
and flexible vegetation; this figure clearly illustrates there is a wide range of
Manning’s n with different formulas.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

106   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

zones. The method in Luhar and Nepf (2013) is the best one for capturing the strong nonlinear
behavior of Manning’s n near the transition between emerged and submerged conditions (see
Figure 4-79). The formula in Luhar and Nepf (2013) is shown in Equation 4-15:
R V1 2
C ah
n = k n SS D WW h 1 6
For emergent vegetation ` H 2 h j
2g
T X
(4-15)
kn h 1 6
n= For submerged vegetation ` H 1 h j
J N1 2J
H N3 2 J 2g NO1 2 H
KK 2g OO KK1 - OO + KK
C L hP C aH O h
L vP L D P
where h is water depth, H is vegetation height, CD is drag coefficient, a is frontal area of vegeta-
tion per volume of vegetation region. Cv is the coefficient of friction at the top of the vegetation
layer, which is typically from 0.005 to 0.13 (Cheng 2011). The dimensionless vegetation density
called roughness density (aH) can be calculated as a function of leaf-area index (LAI). The LAI
is defined as the sum of plant surface area per unit bed (Finnigan 2000); see Equation 4-16:

JhN
ah = 05LAI KK OO For emergent vegetation
LH P
(4-16)
ah = 05LAI For submerged vegetation

4.5.2 Flexible Vegetation


Shields et al. (2017) compared the methods in Freeman et al. (2000, 2002), Järvelä (2004), and
Whittaker et al. (2015) for predicting Manning’s n as a function of flow depth to plant height
ratio (Figure 4-79). Their results show that the three methods selected for flexible vegetation
produce similar results when the water depth to plant height ratio is larger than 0.8. However,
these methods showed different trends for relatively shallow water. For lower flow depth to plant
height ratios, the method in Freeman et al. (2002) is only acceptable for the vegetation species
and experimental conditions reported in their work. Although the methods in Järvelä (2004)
and Whittaker et al. (2015) were also developed based on specific species, they were formulated
in a more general form. Moreover, the data used in Whittaker et al. (2015) are collected from a
large-scale flume and large plant specimens.
The two methods in Järvelä (2004) and Whittaker et al. (2015) will be described below.
For both methods, the friction coefficient f is calculated, which can be used to calculate the
Manning’s n:
f 1
n= R 6
(4-17)
8g
where R is hydraulic radius.
Järvelä (2004) proposed an equation for estimating friction coefficient f for woody, non­
submerged plants:
J U N| h
f = 4Cd| LAI KK OO (4-18)
U H
L |P
where Cdχ is the species-specific drag coefficient, Uχ is velocity, and χ is the Vogel exponent. The
Vogel exponent is a parameter for considering the reduction of drag due to the reconfiguration
of branches and leaves inflow, which depends on plant stiffness (Es). The values of χ and Cdχ for
some species are listed in Table 4-5 [Table 2 in Järvelä (2004)].

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   107

Table 4-5.   Example values of Cdx for different


plant species (Järvelä 2004).
Species Cdχ χ U χ (m/s) LAI
Cedar 0.56 −0.55 0.1 1.42
Spruce 0.57 −0.39 0.1 1.31
White Pine 0.69 −0.50 0.1 1.14
Austrian Pine 0.45 −0.38 0.1 1.61
Willow 0.43 −0.57 0.1 3.2

Table 4-6.   Example values of Cd0 and x


for different plant species from Whittaker
et al. (2015).

Species Foliated Defoliated


Cd0 χ Cd0 χ
A. glutinosa 0.99 −0.73 0.79 −0.57
P. nigra 0.76 −0.80 1.04 −0.81
S. alba 0.88 −0.81 0.94 −0.84

Whittaker et al. (2015) proposed the following equation for evaluating drag force of the indi-
vidual flexible plants:

1 J tU 2 VH N\ 2
F = tCd0 A p0 KK O U2 (4-19)
2 Es I O
L P
where Cd0 is species-specific rigid drag coefficient of undeformed vegetation. Table 4-6 shows the
values of Cd0 and χ for each plant. See Table 3 in Whittaker et al. (2015) for more information.
Ap0 is the projected area of undeformed vegetation, V is the submerged volume of vegetation, Es is
plant modulus of stiffness (Young’s modulus or bending modulus) (N/m2), and I is the second
moment of area. I for a circular cross section with diameter of Ds is I = πDs4/64. EsI is flexural
rigidity, and its values are reported in Table 1 of Whittaker et al. (2015).
After calculating the drag force due to one tree using Equation 4-19, the following equation
(Equation 4-20) for calculating the friction coefficient f for multiple trees with the longitudinal
and lateral spacing (sx and sy) was proposed:
F
f =8 (4-20)
s x s y tU 2
As discussed, flow resistance due to vegetation is complex. The HEC has a plan to add veg-
etation to HEC-RAS. SRH-2D is in the process of implementing the methods in Baptist et al.
(2007) and Järvelä (2004) (Dombroski 2014). Before these 2D models finish implementing
vegetation flow resistance, a modeler can use the formulas presented in this section to estimate
the Manning’s n or find a proper Manning’s n value from textbooks or hydraulics manuals. If the
formulas calculate a Manning’s n for vegetation only, the total Manning’s n should be calculated
using the square root of the sum-of-squares equation rather than scalar addition.

4.6 Summary of Simulation Cases and Examples


In this project, many cases and examples have been developed to help answer numerous ques-
tions in 2D modeling. To have a better view of these case examples, Table 4-7 lists all of them.
The data and case files are made as a companion to this report, which can be requested from
the authors.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Table 4-7.   List of all cases and examples used in this project.
Width Length Side n for Main n for Mesh Downstream
Name (m) (m) Slope Discharge Structure Channel Floodplain/Structure Cell Size (m) Water Depth (m) Notes
Metric unit, S0 = 0.0005. The downstream boundary is known as WSE. The floodplain width for Case 05 is 10 m.
The aim of Case 01 to Case 04 is to
compare HEC-RAS 1D, HEC-RAS 2D,
Case 01 10.0 150.0 0 6.0 0.01 1 0.5 and SRH-2D.
Copyright National Academy of Sciences. All rights reserved.

Case 02 10.0 150.0 0 6.0 0.03 1 0.5 Similar to Case 01 with different n
Case 03 10.0 3000.0 0 6.0 0.01 1 0.5 Similar to Case 01 with different length
Case 04 10.0 3000.0 0 6.0 0.03 1 0.5 Similar to Case 02 with different length
Case 05-1 6.0 1000.0 1 to 1 30.0 0.03 0.08 2 2.1
Case 05-2 6.0 1000.0 1 to 1 30.0 0.03 0.08 1 2.1
Case 05-3 6.0 1000.0 1 to 1 30.0 0.03 0.08 1 2.9
Case 05-5 6.0 1000.0 1 to 1 30.0 0.03 0.10 1 2.9
Case 05-6 6.0 1000.0 1 to 1 30.0 0.03 0.10 1 2.1
Case 05-7 6.0 1000.0 1 to 1 30.0 0.03 0.05 1 2.1
Case 05-8 6.0 1000.0 1 to 1 30.0 0.03 0.05 1 2.9
Case 05-9 6.0 1000.0 1 to 1 30.0 0.03 0.03 1 2.9 SRH-2D, HEC-RAS 2D, and HEC-RAS
Case 05-10 6.0 1000.0 1 to 1 30.0 0.03 0.03 1 2.1 1D are compared for compound channel.
Circular pier,
Case 06-01 10.0 150.0 0 6.0 width 2 m 0.03 0.1 0.88
Circular pier,
Case 06-02 10.0 150.0 0 6.0 width 2 m 0.03 0.3 0.88
Circular pier,
Case 06-03 10.0 150.0 0 6.0 width 2 m 0.03 0.5 0.88
Circular pier,
Case 06-04 10.0 150.0 0 6.0 width 2 m 0.03 1 0.88
Circular pier, Effect of mesh resolution for hole-in-
Case 06-05 10.0 150.0 0 6.0 width 2 m 0.03 1.5 0.88 mesh, 2-m pier, Cases 06, 07, 08, and 09
Circular pier,
Case 06-06 10.0 150.0 0 6.0 width 1 m 0.03 0.05 0.88
Circular pier,
Case 06-07 10.0 150.0 0 6.0 width 1 m 0.03 0.1 0.88
Circular pier,
Case 06-08 10.0 150.0 0 6.0 width 1 m 0.03 0.2 0.88
Circular pier,
Case 06-09 10.0 150.0 0 6.0 width 1 m 0.03 0.3 0.88
Circular pier,
Case 06-10 10.0 150.0 0 6.0 width 1 m 0.03 0.4 0.88
Circular pier,
Case 06-11 10.0 150.0 0 6.0 width 1 m 0.03 0.5 0.88
Circular pier,
Case 06-12 10.0 150.0 0 6.0 width 1 m 0.03 0.7 0.88
Circular pier, Effect of mesh resolution for hole-in-
Case 06-13 10.0 150.0 0 6.0 width 1 m 0.03 1 0.88 mesh, 1-m pier
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Width Length Side n for Main n for Mesh Downstream
Name (m) (m) Slope Discharge Structure Channel Floodplain/Structure Cell Size (m) Water Depth (m) Notes
Round-nose
Case 06-14 10.0 150.0 0 6.0 elongated, 1 m 0.03 0.1 0.88
Round-nose
Case 06-15 10.0 150.0 0 6.0 elongated, 1 m 0.03 0.3 0.88
Round-nose
Case 06-16 10.0 150.0 0 6.0 elongated, 1 m 0.03 0.5 0.88
Round-nose
Case 06-17 10.0 150.0 0 6.0 elongated, 1 m 0.03 1 0.88
Round-nose
Case 06-18 10.0 150.0 0 6.0 elongated, 1 m 0.03 0.5–1 0.88
Round-nose
Case 06-19 10.0 150.0 0 6.0 elongated, 1 m 0.03 0.5–1.5 0.88
Copyright National Academy of Sciences. All rights reserved.

Round-nose Effect of mesh resolution for hole-in-


Case 06-20 10.0 150.0 0 6.0 elongated, 1 m 0.03 0.5–3 0.88 mesh, 1-m pier
Round-nose
Case 06-21 10.0 150.0 0 6.0 elongated, 2 m 0.03 0.1 0.88
Round-nose -
Case 06-22 10.0 150.0 0 6.0 elongated, 2 m 0.03 0.3 0.88
Round-nose
Case 06-23 10.0 150.0 0 6.0 elongated, 2 m 0.03 0.5 0.88
Round-nose
Case 06-24 10.0 150.0 0 6.0 elongated, 2 m 0.03 1 0.88
Round-nose
Case 06-25 10.0 150.0 0 6.0 elongated, 2 m 0.03 1.5 0.88
Round-nose
Case 06-26 10.0 150.0 0 6.0 elongated, 2 m 0.03 0.5–1 0.88
Round-nose
Case 06-27 10.0 150.0 0 6.0 elongated, 2 m 0.03 0.5–1.5 0.88
Round-nose Effect of mesh resolution for hole-in
Case 06-28 10.0 150.0 0 6.0 elongated, 2 m 0.03 0.5–3 0.88 mesh, 2-m pier
Circular pier, Compare with unassigned mesh element
Case 06-29 10.0 150.0 0 6.0 width 2 m 0.03 0.32 0.3 0.88 method
Circular pier, Compare with elevated mesh element
Case 06-30 10.0 150.0 0 6.0 width 2 m 0.03 0.5 0.3 0.88 method
Circular pier,
Case 06-31 10.0 150.0 0 6.0 width 2 m 0.03 1 0.3 0.88
Circular pier,
Case 06-32 10.0 150.0 0 6.0 width 2 m 0.03 10 0.3 0.88
Circular pier,
Case 06-33 10.0 150.0 0 6.0 width 2 m 0.03 100 0.3 0.88
Circular pier, Increased Manning’s n, effect of n
Case 06-34 10.0 150.0 0 6.0 width 2 m 0.03 0.32 0.3 0.88 values
Circular pier, Effect of mesh resolution. n = 0.32 and
Case 06-35 10.0 150.0 0 6.0 width 2 m 0.03 0.32 0.5 0.88
cell size is varied.
(continued on next page)
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Table 4-7.  (Continued).
Width Length Side n for Main n for Mesh Downstream
Name (m) (m) Slope Discharge Structure Channel Floodplain/Structure Cell Size (m) Water Depth (m) Notes
Circular pier,
Case 06-36 10.0 150.0 0 6.0 width 2 m 0.03 0.32 2 0.88
Circular pier,
Case 06-37 10.0 150.0 0 6.0 width 2 m 0.03 0.32 3×4 0.88
Circular pier, n over footprint,
Copyright National Academy of Sciences. All rights reserved.

Case 06-38 10.0 150.0 0 6.0 width 1 m 0.03 0.258 1 0.88


Circular pier, n over footprint,
Case 06-39 10.0 150.0 0 6.0 width 2 m 0.03 0.184 1 0.88
Circular pier, n over canal,
Case 06-40 10.0 150.0 0 6.0 width 1 m 0.03 0.078 1 0.88
Circular pier, n over canal,
Case 06-41 10.0 150.0 0 6.0 width 2 m 0.03 0.078 1 0.88 The HDS 7 method is checked.
Circular pier,
Case 06-42 10.0 150.0 0 6.0 width 2 m 0.03 CD = 4 0.3 0.88
Circular pier,
Case 06-43 10.0 150.0 0 6.0 width 2 m 0.03 C D = 10 0.3 0.88
Circular pier,
Case 06-44 10.0 150.0 0 6.0 width 2 m 0.03 C D = 100 0.3 0.88
Circular pier, Effect of drag coefficient for
Case 06-45 10.0 150.0 0 6.0 width 2 m 0.03 C D = 300 0.3 0.88 additional drag force
Circular pier,
Case 06-46 10.0 150.0 0 6.0 width 2 m 0.03 CD = 4 0.3 0.88
Circular pier,
Case 06-47 10.0 150.0 0 6.0 width 2 m 0.03 CD = 4 2 0.88
Circular pier,
Case 06-48 10.0 150.0 0 6.0 width 2 m 0.03 CD = 4 0.5, triangular 0.88
Circular pier, Effect of mesh resolution on additional
Case 06-49 10.0 150.0 0 6.0 width 2 m 0.03 CD = 4 3×4 0.88 drag-force method
Circular pier,
Case 06-50 10.0 150.0 0 6.0 width 2 m 0.03 C D = 100 0.3 0.88
Circular pier,
Case 06-51 10.0 150.0 0 6.0 width 2 m 0.03 C D = 100 2 0.88
Circular pier,
Case 06-52 10.0 150.0 0 6.0 width 2 m 0.03 C D = 100 0.5, triangular 0.88
Circular pier, Effect of mesh resolution on additional
Case 06-53 10.0 150.0 0 6.0 width 2 m 0.03 C D = 100 3×4 0.88 drag-force method

Case 06-54 10.0 150.0 0 6.0 0.03 SRH, 2 0.88

Case 06-55 10.0 150.0 0 6.0 0.03 SRH, 0.5 0.88


RiverFlow2D,
Case 06-56 10.0 150.0 0 6.0 0.03 2 0.88
RiverFlow2D, Compare RiverFlow2D and SRH-2D for
Case 06-57 10.0 150.0 0 6.0 0.03 0.5 0.88 a simple channel.
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Width Length Side n for Main n for Mesh Downstream
Name (m) (m) Slope Discharge Structure Channel Floodplain/Structure Cell Size (m) Water Depth (m) Notes
Circular pier,
Case 06-58 10.0 150.0 0 6.0 width 2 m 0.03 Hole-in-mesh 0.3 0.88 SRH-2D with Ct = 0.7
Circular pier,
Case 06-59 10.0 150.0 0 6.0 width 2 m 0.03 Hole-in-mesh 0.3 0.88 RiverFlow2D
Circular pier,
Case 06-60 10.0 150.0 0 6.0 width 2 m 0.03 Hole-in-mesh 0.3 0.88 SRH-2D with Ct = 0
Circular pier, Elevated mesh
Case 06-61 10.0 150.0 0 6.0 width 2 m 0.03 element SRH, 0.3 0.88
Circular pier, Elevated mesh
Case 06-62 10.0 150.0 0 6.0 width 2 m 0.03 element RAS, 0.3 0.88
Circular pier, Elevated mesh
Case 06-63 10.0 150.0 0 6.0 width 2 m 0.03 element SRH, 1 0.88
Copyright National Academy of Sciences. All rights reserved.

Circular pier, Elevated mesh


Case 06-64 10.0 150.0 0 6.0 width 2 m 0.03 element RAS, 1 0.88
Circular pier, Elevated mesh
Case 06-65 10.0 150.0 0 6.0 width 2 m 0.03 element SRH, 1.67 0.88 Compare HEC-RAS 2D and SRH-2D for
Circular pier, Elevated mesh modeling pier using elevated mesh
Case 06-66 10.0 150.0 0 6.0 width 2 m 0.03 element RAS, 1.67 0.88 element.

Metric unit, S0 = 0.0005, Downstream boundary is known WSE, bedform height and length are 0.5 and 10 m.
Case 10-01 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-02 10.0 500.0 0 20.0 0.03 0.4 2
Case 10-03 10.0 500.0 0 20.0 0.03 0.6 2
Case 10-04 10.0 500.0 0 20.0 0.03 0.8 2
Case 10-05 10.0 500.0 0 20.0 0.03 1 2
Case 10-06 10.0 500.0 0 20.0 0.03 1.5 2
Case 10-07 10.0 500.0 0 20.0 0.03 3 2
Case 10-08 10.0 500.0 0 20.0 0.03 10 2 Bedform aligned with flow direction
Case 10-09 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-10 10.0 500.0 0 20.0 0.03 0.4 2
Case 10-11 10.0 500.0 0 20.0 0.03 0.6 2
Case 10-12 10.0 500.0 0 20.0 0.03 0.8 2
Case 10-13 10.0 500.0 0 20.0 0.03 1 2
Case 10-14 10.0 500.0 0 20.0 0.03 1.5 2
Case 10-15 10.0 500.0 0 20.0 0.03 3 2 Bedform aligned with flow direction +
Case 10-16 10.0 500.0 0 20.0 0.03 10 2 perturbation
Case 10-17 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-18 10.0 500.0 0 20.0 0.03 0.4 2
Case 10-19 10.0 500.0 0 20.0 0.03 0.6 2
Case 10-20 10.0 500.0 0 20.0 0.03 0.8 2
Case 10-21 10.0 500.0 0 20.0 0.03 1 2
Case 10-22 10.0 500.0 0 20.0 0.03 1.5 2
Case 10-23 10.0 500.0 0 20.0 0.03 3 2 Bedform rotated 30 degrees

(continued on next page)


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Table 4-7.  (Continued).
Width Length Side n for Main n for Mesh Downstream
Name (m) (m) Slope Discharge Structure Channel Floodplain/Structure Cell Size (m) Water Depth (m) Notes
Case 10-24 10.0 500.0 0 20.0 0.03 10 2
Case 10-25 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-26 10.0 500.0 0 20.0 0.03 0.4 2
Case 10-27 10.0 500.0 0 20.0 0.03 0.6 2
Copyright National Academy of Sciences. All rights reserved.

Case 10-28 10.0 500.0 0 20.0 0.03 0.8 2


Case 10-29 10.0 500.0 0 20.0 0.03 1 2
Case 10-30 10.0 500.0 0 20.0 0.03 1.5 2
Case 10-31 10.0 500.0 0 20.0 0.03 3 2
Case 10-32 10.0 500.0 0 20.0 0.03 10 2 Smooth slope channel (without bedform)
Case 10-33 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-34 10.0 500.0 0 20.0 0.03 0.4 2
Case 10-35 10.0 500.0 0 20.0 0.03 0.6 2
Case 10-36 10.0 500.0 0 20.0 0.03 0.8 2
Case 10-37 10.0 500.0 0 20.0 0.03 1 2
Case 10-38 10.0 500.0 0 20.0 0.03 1.5 2
Case 10-39 10.0 500.0 0 20.0 0.03 3 2 Smooth slope channel + perturbation
Case 10-40 10.0 500.0 0 20.0 0.03 10 2 [−0.1, 0.1]
Metric unit, S 0 = 0.0005, Comparing different models for flow over bedforms
Case 10-41 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-42 10.0 500.0 0 20.0 0.03 2.5 2
2.5
Case 10-43 10.0 500.0 0 20.0 0.03 triangular 2 Smooth slope channel + perturbation
Case 10-44 10.0 500.0 0 20.0 0.03 10 2 [−0.1, 0.1], SRH-2D
Case 10-45 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-46 10.0 500.0 0 20.0 0.03 2.5 2 Smooth slope channel + perturbation
Case 10-47 10.0 500.0 0 20.0 0.03 10 2 [−0.1, 0.1], HEC-RAS 2D
Case 10-48 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-49 10.0 500.0 0 20.0 0.03 2.5 2
0.2,
Case 10-50 10.0 500.0 0 20.0 0.03 no cont–exp 2
2.5,
Case 10-51 10.0 500.0 0 20.0 0.03 no cont–exp 2 Smooth slope channel + perturbation
Case 10-52 10.0 500.0 0 20.0 0.03 10 2 [−0.1, 0.1], HEC-RAS 1D
Case 10-53 10.0 500.0 0 20.0 0.03 0.2 2
Case 10-54 10.0 500.0 0 20.0 0.03 2.5 2 Smooth slope channel + perturbation
Case 10-55 10.0 500.0 0 20.0 0.03 5 2 [−0.1, 0.1], RiverFlow2D.
Metric unit, S 0 = 0.0005, Channel irregularities
0.88, normal
Case13 10.0 150.0 0 6.0 NA-2D 0.03 0.5 depth No contraction, HEC-RAS 2D
contraction– 0.88, normal The width decreased to 4 m and back
Case13-01 10.0 150.0 0 6.0 expansion 0.03 0.5 depth again over a length of 5 m.
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Width Length Side n for Main n for Mesh Downstream
Name (m) (m) Slope Discharge Structure Channel Floodplain/Structure Cell Size (m) Water Depth (m) Notes
0.88, normal
Case14 10.0 150.0 0 6.0 NA-1D 0.03 0.5 depth No contraction, HEC-RAS 1D
contraction– 0.88, normal
Case 14-01 10.0 150.0 0 6.0 expansion 0.03 0.5 depth 1D, contraction and exp. Coef. 0.1 and 0.3
contraction– 0.88, normal
Case 14-02 10.0 150.0 0 6.0 expansion 0.03 0.5 depth 1D, contraction and exp. Coef. 0.3 and 0.5
contraction– 0.88, normal
Case 14-03 10.0 150.0 0 6.0 expansion 0.03 0.5 depth 1D, contraction and exp. Coef. 0 and 0
contraction– 0.88, normal 1D, contraction and exp. Coef. 0.05 and
Case 14-04 10.0 150.0 0 6.0 expansion 0.03 0.5 depth 0.15
English unit, Real-world examples, Downstream boundary is known WSE
Susitna-01 25 kcfs 20 ft 564 Susitna River, SRH-2D
Copyright National Academy of Sciences. All rights reserved.

Susitna-02 100 kcfs 20 ft 570 Susitna River, SRH-2D


Susitna-03 25 kcfs 20 ft 564 Susitna River, HEC-RAS 2D
Susitna-04 100 kcfs 20 ft 570 Susitna River, HEC-RAS 2D
Susitna-05 25 kcfs 20 ft 564 Susitna River, RiverFlow2D
Susitna-06 100 kcfs 20 ft 570 Susitna River, RiverFlow2D
Susitna-07 100 kcfs 30 ft 570 Susitna River, SRH-2D
Susitna-08 100 kcfs 30 ft 570 Susitna River, HEC-RAS 2D
Susitna-09 100 kcfs 30 ft 570 Susitna River, RiverFlow2D
Susitna-10 100 kcfs 60 ft 570 Susitna River, SRH-2D
Susitna-11 100 kcfs 60 ft 570 Susitna River, HEC-RAS 2D
Susitna-12 100 kcfs 60 ft 570 Susitna River, RiverFlow2D
Iowa-01 39,900 cfs 14 bridges Hole-in-mesh 50 × 100 643.9 Near structure 20 × 30
Iowa-02 39,900 cfs 14 bridges CD = 4 50 × 100 643.9 Near structure 20 × 30
Iowa-03 39,900 cfs 14 bridges CD = 100 50 × 100 643.9 Near structure 20 × 30
Iowa-04 39,900 cfs 14 bridges CD = 100 80 × 120 643.9 Near structure 30 × 60
Iowa-05 39,900 cfs 14 bridges CD = 100 80 × 120 643.9 Near structure 30 × 60
Iowa-06 to
105 39,900 cfs 14 bridges Hole-in-mesh 80 × 120 643.9 100 runs for sensitivity analysis
DelDOT-
01 15,900 cfs Bridge Hole-in-mesh 7 × 13 40.0 Near structures 4 × 5
DelDOT-
02 to 101 15,900 cfs Bridge Hole-in-mesh 7 × 13 40.0 100 runs for sensitivity analysis
Note: cfs = cubic feet per second; kcfs = thousands of cubic feet per second.
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

114   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

4.7 How to Obtain Total Manning’s n


Based on the discussion of resolved, unresolved, and total Manning’s n values, it is important
to decide what total Manning’s n value to use for a particular zone in the modeling domain.
There are many different methods and tools currently used by practitioners. The research
team reviewed most of them and found the Excel spreadsheet developed by the U.S. Department
of Agriculture (USDA) to be one of the most comprehensive (Yochum 2018). This spreadsheet
contains three different approaches from which a modeler can choose: tabular, photographic,
and quantitative.

4.7.1 Tabular Method


In this method, tables for Manning’s n values for different land use and surface conditions
are provided [e.g., Aldridge and Garrett (1973), Arcement and Schneider (1989), and Brunner
(2021)]. These references are available in the Excel spreadsheet package. Table 4-8 shows an
example table from the HEC-RAS reference manual (Brunner 2021).

4.7.2 Photographic Method


In this method, Manning’s n can be determined based on representative field photos (Barnes
1967, Aldridge and Garrett 1973, Hicks and Mason 1998, Yochum et al. 2014). Moreover, USGS
developed a website (https://wwwrcamnl.wr.usgs.gov/sws/fieldmethods/Indirects/nvalues
/index.htm) that provides many photos and their corresponding Manning’s n values. Photo-based
guidance documents help modelers gain more experience in selecting proper Manning’s n by
looking at the appearance and geometry of the river channels. Figure 4-80 shows some examples
from the USGS website.

4.7.3 Quantitative Methods


The quantitative approach is probably the most studied among all methods. There are many
different formulas for different roughness conditions. They can be either quasi-quantitative or
fully quantitative (Yochum 2018):
• In fully quantitative methods, Manning’s n is typically estimated as a function of relative sub-
mergence based on bed-material size and bedform fluctuations. The drawbacks of these meth-
ods are their significant errors and ignoring other sources of resistance, including obstructions,
bank irregularity, vegetation, and sinuosity (Yochum 2018).

Table 4-8.   Example table guidance for selecting Manning’s n


from HEC-RAS user manual (Brunner 2021).

Type of Channel and Description Minimum Normal Maximum


A. Natural

1. Main channel
a. Clean, straight, full, no rift or deep pools 0.025 0.03 0.033
b. Same as above, but more stones and weeds 0.030 0.035 0.040
c. Clean, winding, some pools and shoals 0.033 0.040 0.045
d. Same as above, but some weeds and stones 0.035 0.045 0.050
e. Same as above, lower stages, more ineffective 0.040 0.048 0.055
slopes and sections
f. Same as "d" but more stones 0.045 0.050 0.060
g. Sluggish reaches, weedy, deep pools 0.050 0.070 0.080
h. Very weedy reaches, deep pools, or �loodways 0.070 0.100 0.150
with heavy stands of timber and brush

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Flow Resistance Factors, Required Conditions for Resolving a Feature, and Effect of Model Choice   115

Columbia River at Vernita, Washington n = 0.024

Clark Fork at St. Regis, Montana n = 0.028

Clear Creek near Golden, Colorado n = 0.050


Source: (https://wwwrcamnl.wr.usgs.gov/sws/fieldmethods/Indirects/nvalues/index.htm). Photo credit: USGS, public domain.

Figure 4-80.   Sample photos from the USGS online photo guide for Manning’s n.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

116   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Table 4-9.   Fully quantitative methods used in the USDA Excel spreadsheet tool
and their range of applicability.

Data
Method Points Slope Range Relative Submergence
Yochum et al. (2012) 78 0.02 to 0.2 hm/σz = 0.25 to 12
Rickenmann and Recking (2011) 2,890 0.00004 to 0.03 d/D84 = 0.18 to ∽100
Aberle and Smart (2003); in �lume 94 0.02 to 0.1 d/σz = 1.2 to 12
Lee and Ferguson (2002) 81 0.027 to 0.184 R/D84 (step) = 0.1 to 1.4
Bathurst (1985) 44 0.00429 to 0.0373 R/D84 = 0.71 to 11.4
Jarrett (1984) 75 0.002 to 0.039 n/a
Grif�iths (1981); rigid bed 84 0.000085 to 0.011 R/D50 = 1.8 to 181
Hey (1979); a = 12.72 30 0.00049 to ∽0.01 R/D84 = 0.8 to 25
Limerinos (1970) 50 0.00038 to 0.039 R/D84 = 1.1 to 69

• Quasi-quantitative methods are developed to overcome the weakness of fully quantitative


methods. In these methods, the total Manning’s n is considered as the sum of several inde-
pendent components affecting flow resistance. As discussed in this report, the linear sum of
Manning’s n values from different flow resistance factors is not a physically sound approach.
However, they do provide some historical value.
Nine fully quantitative methods and one quasi-quantitative method are included in the USDA
spreadsheet tool. The quasi-quantitative method is from Arcement and Schneider (1989), which
was developed based on Cowan (1956). It is important to note that these methods simply add
components of Manning’s n. They are not physically correct and are expected to overestimate
total Manning’s n.
In Table 4-9, the list of fully quantitative methods used in their tool and the range of their
applicability are shown. In this table, d is the mean depth, R is the hydraulic radius, hm is the
median thalweg depth, D84 and D50 are the 84th and 50th percentiles of sediment-size distribu-
tion, and σz is the STD of residuals of a thalweg longitudinal profile regression. There is a specific
tab in the USDA tool to calculate σz that contains sufficient guidance. As mentioned in the tool,
the method of Lee and Ferguson (2002) can substantially underestimate Manning’s n in streams
steeper than 0.03 where there are large woods that enhance the steps.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CHAPTER 5

Development of 2D Modeling
Best Practices

5.1 Overview of 2D Modeling Best Practices


The specification and calibration of Manning’s n should begin with a model that is appropri-
ately developed using good modeling practices. If a model is not properly developed, contains
errors, or otherwise does not follow best practices, the simulation results will not be accurate or
may even be wrong. If this is the case, there is no point in wasting time and effort determining
Manning’s n.

One obvious example is that if key ground controls or features in the domain are misrepre-
sented or not resolved, the simulated flow will significantly deviate from that in reality. During
calibration or parameter tuning, a modeler may have to use extraordinary or even absurd
Manning’s n values to compensate for the consequence of not following best practices. The abusive
use of Manning’s n as a turning parameter has dire consequences in practices. It will result in
wrong flow velocity, stage, wall shear, scour depth estimation, and unsafe bridges.

Best practices cover all aspects of 2D hydraulics modeling. Because of the critical importance
of these best practices to Manning’s n, a summary is provided. Complete and thorough best
practices for 2D hydraulics modeling are beyond the scope of this work. It is a good topic for
a future research project. Interested readers can find more detailed information in the FHWA
reference document Two-Dimensional Hydraulic Modeling for Highways in the River Environ-
ment (Robinson et al. 2019).

Computational hydraulics modeling belongs to the broader category of computational fluid


dynamics (CFD) modeling. CFD modeling has been used extensively in many scientific and
engineering disciplines. To ensure quality and fidelity, there are many best-practice guidelines
for CFD modeling. However, despite commonalities, these best practices are specific to cer-
tain disciplines and may not be directly applicable to hydraulics. The best practices outlined
in the next section will follow the general structure of existing guidelines. The uniqueness of
2D hydraulics models related to rivers and streams will be highlighted.

5.2 Best Practices in 2D Hydraulics Modeling


A typical hydraulics modeling project consists of four major steps: (1) project scoping and
problem definition, (2) preprocessing, (3) simulation, and (4) postprocessing. Figure 5-1 shows
these four steps and the subtasks within each step. This conceptual diagram will illustrate best
practices in the process of performing each subtask.

117

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

118 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 5-1. Steps and different aspects of a typical modeling project.

5.2.1 Step 1: Project Scoping and Problem Definition


Project scoping and problem definition are the first step in a typical modeling project. The
goal here is to properly scope the project and define the problem at hand, which includes the
following:
• Purpose of modeling: This is probably the most important question a modeler needs to answer.
Hydraulics modeling can be used for many purposes, for example, flood inundation mapping,
habitat study, bridge design, and scour analysis. Depending on the purpose, the decisions in
the following steps and the best practices may be different.
• Spatial domain: The spatial extent of the modeling domain is determined by the study area of
interest, influence zone of hydraulic structures, numerical considerations, and computational
cost. The spatial domain should be as large as can be reasonably achieved with the boundar-
ies far from the study area. However, a larger domain means higher computational cost, and
sometimes it is wasteful. At a minimum, the boundary should be far enough from the area of
interest to avoid unduly affecting the results.
For flood inundation studies, the domain must enclose all areas that can be inundated. To
save computational cost, high ground that cannot be inundated should be excluded. On the
other hand, the domain boundary should not stop in the middle of the flood zone. A modeler
should always check the simulated result to ensure there is no “glass wall” along the domain
boundary.
For local hydraulics studies such as bridges and other road crossing structures, the upstream
and downstream boundaries should be located outside of the influence zones of the modeled
structures (see Figure 5-2). For example, the existence of obstructions such as piers will create

Figure 5-2. Proper location of boundaries.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Development of 2D Modeling Best Practices   119

downstream wake zones. The downstream boundary should be located farther away from the
wake zone. For the upstream boundary, it will be ideal to locate it outside of the backwater
effect zone. At the minimum, it should be upstream of the contraction zone of the structure. If
there are upstream flow controls that could affect the structure hydraulics, then the boundary
should be set upstream of these controls. As a rule of thumb, the domain should include at
least several floodplain widths upstream and downstream of the structures.
Regardless of the modeling purpose, the upstream and downstream boundaries should be
placed in areas where flow is approximately fully developed and unidirectional. When the
flow at the boundary is developing or nonuniform, or even worse when there is recirculation,
the specified boundary conditions, such as fixed inflow discharge or WSE, cannot accurately
describe the hydrodynamics, and it will be difficult for 2D models to converge.
• Temporal domain: Temporal domain is relevant only for unsteady simulations, such as a flood
event or a tidal-influenced project. The period of simulation should include the whole tem-
poral domain of interest, such as the rising and falling limbs of the hydrograph. It is also rec-
ommended to include a warm-up period before the start of a real flood or use a presimulated
result as an initial condition. For steady-state simulations, pseudo time is typically used. The
pseudo period must be long enough such that all solved flow variables reach equilibrium,
which can be checked from plots of monitoring points or lines.
• Data availability and quality: Data is always the most important prerequisite for a success-
ful modeling project. Data should be collected and quality-checked before model building.
A modeler should always check the source and accuracy of data to ensure the model is not
built with wrong information. One way for quality control is to cross-check data from different
sources (if they are available). Another way is to do field reconnaissance and visit.
• Field reconnaissance and survey: If budget and time allow, a modeler should participate
in field reconnaissance and survey. There is nothing more valuable than seeing the river,
floodplain, landscape, hydraulic structures, and high watermarks in the physical world. Field
reconnaissance and survey also provide opportunities to identify inconsistencies in avail-
able data. For example, large wood may have accumulated, or bars may have migrated in the
channel since data were collected. Their exact locations may significantly alter the hydro-
dynamics. Collection of survey data is critical to ensure hydraulic controls are captured in
terrain data and mesh.
• Steady versus unsteady: This was discussed in temporal domain determination. A modeler
should decide whether the unsteady process is of interest, such as flooding in a broad flood-
plain or within a tidal zone. If yes, an unsteady simulation should be performed. Otherwise,
a steady simulation should be performed.
• Sediment transport: If sediment transport is part of the modeling purpose, information such
as sediment sizes, concentration, loading, and rating curves should be collected. Sediment
transport modeling is typically more difficult than flow modeling because of the complexity
of sediment movement, lack of high-quality data, and many other parameters. The best prac-
tice is to collect detailed information, such as the monitoring and postflood survey results of
bridge pier scour, sediment sizes in both channel and floodplain, and changes of prominent
geomorphic features.
• 2D model selection: Which 2D model to use should be determined by several factors. A modeler
should first consider whether a 2D model has the functionality necessary for the project. For
example, if bridge hydraulics is the focus, the hole-in-mesh approach is more accurate than
modeling bridge piers as obstructions with a drag coefficient. Currently, among the models
tested for this project, SRH-2D and RiverFlow2D provide the hole-in-mesh option. Other
considerations should include the familiarity and experiences of the modeler in using a par-
ticular model, regulatory requirements, computing speed, stability, technical support, and
cost. The best practice is to make a thorough investigation and comparison before adopting
a model.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

120   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

5.2.2 Step 2: Preprocessing


The preprocessing step is probably the most critical for the success of a modeling project.
Among all tasks in preprocessing, meshing is the most time-consuming one. Producing a high-
quality mesh will be rewarded by faster simulations, easier convergence, high-quality results, and
minimizing the possibility of going back and redoing everything.
Preprocessing includes the following aspects:
• Terrain data preparation: The underlying topography and bathymetry are fundamental for
model building. The processing of terrain data (e.g., cleaning, correction, filtering, smoothing,
resampling, merging, and cross checking) can be laborious and tedious. However, high-quality
and error-free terrain data are the foundation for all other steps.
• Meshing: The meshing process is to cover the whole simulation domain by nonoverlapping
cells (e.g., triangles and quadrilaterals). Meshing is one of the most important components of
2D modeling that directly affects Manning’s n. A valid and good-quality mesh, resulted from
proper placement of breaklines and appropriate mesh resolutions, is the foundation for accu-
rate results. On the other hand, an invalid or poor-quality mesh will produce garbage results
that are not usable. The calibration of Manning’s n with a bad mesh creates incorrect values
and can drive Manning’s n out of its reasonable range.
The definition of a good-quality mesh is difficult. Different modelers may come up with dif-
ferent, but equally reasonable, meshes for a particular modeling domain. The meshing process
is a mix of science and art. However, there are some general guidelines and best practices for
2D hydraulics model meshing. The following is an attempt to list some of the most impor-
tant ones:
– Ground control and hydraulic features. Because of the importance of these features in
steering and conveying flow, a mesh must resolve all features that affect flow patterns.
Examples include channel banks, roads, berms, levees, and other grade break features that
can impede or stop the flow. Excessively large elements in floodplains may miss important
hydraulic controls and result in too much flow outside the channel. On the other hand,
features like small channels and abandoned flood pathways need to be resolved so that
their contributions to conveyance are captured accurately.
There are different ways to check how well the mesh captures the terrain, and different
2D models provide different tools. For example, in SMS, one can calculate and plot the
difference between the elevations in the scatter data and mesh. Attention must be paid to
places with the largest difference so that it is guaranteed that these places are not important
grade break and hydraulic features. Figure 5-3 shows two examples. Figure 5-3a shows
(1) a proper mesh that captures the major terrain features and (2) the difference between
a digital elevation model (DEM) and mesh elevation is small in the area of interest.
Figure 5-3b shows an opposite example, where the mesh is too coarse to capture the terrain
feature and the difference is large.
Another way to check is to plot profiles along a line. Visual inspection can directly pin-
point where the mesh misrepresents the terrain the most. Figure 5-4 shows such an example.
The profiles at a river cross section from two meshes are compared with the original terrain
data. The coarse mesh misrepresents the deep part of the channel in the middle. As a result,
the conveyance of the channel with the coarse mesh is reduced and the backwater curve is
higher than it should be. This directly affects the calibrated Manning’s n value for this river.
With the coarse mesh, Manning’s n needs to be erroneously much smaller to reduce the flow
resistance and make up for the lost conveyance.
The importance of missed terrain features also depends on its relative contribution
to conveyance. For example, if WSE reaches only about 650 ft in the channel shown in
Figure 5-4, then the missed conveyance by the coarse mesh is important. However, if WSE

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Development of 2D Modeling Best Practices   121

Terrain - Mesh elevation Terrain - Mesh elevation

Figure 5-3.   Difference between terrain data and mesh elevation: (a) an appropriate mesh to
capture the major terrain features and (b) mesh is too coarse to capture the major terrain features.

is much higher and reaches, for example, 680 ft, then the lost conveyance by using the coarse
mesh is relatively smaller.
Another visual method for checking whether terrain features are sufficiently resolved
is to compare the mesh contours to the terrain contours. In SMS, this is done by display-
ing the mesh module’s elevation contours overlaying the scatter module’s terrain contours
at the same interval. Additional mesh refinement should be performed where these con-
tours deviate substantially.

Figure 5-4.   Difference between terrain data and mesh elevation.


The blue line shows an appropriate mesh to capture the major
terrain features, and the red line shows a mesh that is too coarse
to capture the major terrain features, e.g., the deep part of the
channel in the middle.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

122   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

In some 2D models, like HEC-RAS 2D, techniques such as the “subgrid terrain” treat-
ment have been implemented. These techniques consider the effect of ground features
smaller than mesh cell size, hence the name “subgrid.” It is a good idea to balance accuracy
and computing cost; however, caution must be exercised. One might think that with such
treatment, a mesh can be coarse and that geometrically resolving a ground control is not
that critical—this is not correct. A modeler should know that the subgrid terrain treatment
is only an approximation. This treatment is applied to some, not all, terms in the governing
equation. As a result, its use cannot fully compensate for the loss of terrain information due
to a coarse or poor mesh. The best practice here is to use breaklines and mesh refinement
along ground controls so that the mesh can better capture terrain change. Placing break-
lines along ground controls will align mesh cell sides with these features, and their effects
are better incorporated into the solution of the governing equations.
The added benefits of subgrid terrain treatment diminish when a mesh is reasonably
refined (see Figure 5-5). When the mesh resolution is reasonably high, the variation of
terrain along a mesh cell face is small; in other words, there is no significant subgrid ter-
rain change. On the other hand, when the mesh is coarse, terrain can change significantly
along a cell face. In this case, the subgrid terrain treatment in some 2D models can better
capture the wetted perimeter along the face and, therefore, better model the flow resistance.
However, a modeler should always keep in mind that the subgrid terrain treatment has its
limitations (as described previously).
– Other transition regions. There might be other regions in the domain where the flow con-
veyance has relatively rapid changes. Some examples are regions where there is a change
in land use (significant change in Manning’s n values) and shear layers around structures
and terrain features (e.g., bridge piers, barbs, groins, and engineered log jams). In other
words, these regions have high-velocity gradients. Velocity gradient is the major cause of
turbulence and controls overall hydrodynamics. If capturing these transition regions is
part of the modeling purpose or these regions greatly impact the hydraulics, then the mesh
should be refined in these regions to better capture their dynamics.
– What does it mean to resolve a feature in a mesh? It means the terrain changes, such as at
the edge of elevated roadway, bank lines, and other terrain features, are captured. Breaklines
and mesh cell edges must be placed along with the linear control features. In addition, a mesh
must be refined in areas covering these features. For example, several cells are needed across
the width of a channel to capture its conveyance. The same applies to features such as roads
and levees. In general, the mesh should have greater resolution wherever there are large
velocity gradients or large terrain changes.

Figure 5-5.   Subgrid terrain with refined and coarse meshes.


The schematic diagram assumes the underlying terrain is
not flat.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Development of 2D Modeling Best Practices   123

– Metrics for mesh quality. There are many metrics for mesh quality. Most 2D models and
their mesh generators have ways to calculate and visualize these metrics. For example, SMS
provides a way to plot metrics, such as minimum and maximum interior angles, number
of neighbor elements, and others. Most 2D models also provide some guidelines on the
recommended ranges of these metrics.
These guidelines and recommendations may be model specific; thus, a modeler should
always consult the user manuals.
As general guiding principles, a good mesh should use structured (patch) blocks, rather
than unstructured (paving) blocks, as much as possible. Some benefits of this are the
orthogonality of mesh cells, higher accuracy, and better numerical stability. With experi-
ence and some imagination, a modeler can divide the domain into many logically rectan-
gular subdomains and cover them with structured mesh. The remaining part of the domain
can be covered with an unstructured mesh. If the geometry of the domain is too compli-
cated to cover important areas with structured mesh, it is recommended to use relatively
refined unstructured mesh.
As a good practice, a modeler should start with a coarse mesh to ensure the layout of
mesh blocks is valid. Once the subdivision of the whole domain is validated, mesh refine-
ment can be performed where needed until model results are within a given tolerance.
– Mesh resolution. A mesh should have proper resolutions in different areas of the domain.
In areas of interest, mesh resolution should be as high as possible. In areas far away, mesh
resolution can be relaxed. The resolution required to resolve ground control and hydraulic
features has been discussed. Another requirement of mesh resolution is to limit numerical
errors. The larger the mesh cell, the higher the numerical error. Larger mesh cells also intro-
duce artificial numerical diffusion. It can be difficult to identify the level of numerical errors.
However, mesh refinement in areas with large ground surface gradients, large velocity
gradients, and large changes in Manning’s n helps minimize numerical errors.
– A good mesh should always have gradual changes and avoid sudden transitions. For example,
the mesh cell size should transition gradually from refinement areas to coarse areas. This
helps numerical stability. If necessary, add transition layers or areas.
– It is also a good practice to simplify mesh near model boundaries, especially at inflow and
outlet boundaries. The best practice is to use rectangular cells aligned with the flow direc-
tion near mesh boundaries.
• Boundary conditions: Without boundary conditions, the governing SWE have an infinite
number of solutions. The solution that a 2D (or 1D) model produces is the one that cor-
responds to the specified boundary conditions and initial conditions. Thus, wrong or inap-
propriate boundary conditions can produce unusable solutions. There are many different
boundary conditions, and each model may call them different names. Nonetheless, the most
important boundary conditions are for the inlet and outlet. Typically, inlet discharge and out-
let WSE are specified. A modeler should make sure the specified values are appropriate. If cali-
bration is performed, the calibration data, such as stage and velocity at calibration locations,
should correspond to the specified inlet discharge and outlet WSE. If they are not consistent,
the calibrated Manning’s n will be meaningless.
Boundary conditions are enforced strictly in 2D models. However, these boundary condi-
tions are a simplified mathematical description of what researchers think occurs at the bound-
ary. Mathematics may not exactly reflect reality. Models need some distance to adjust. In other
words, there is a transition region near inlet and outlet boundaries where the flow is develop-
ing. To avoid the contamination of model results from boundary effects, it is a best practice
to locate the boundaries far away from the area of interest. To reduce the computational cost,
mesh resolution near boundaries placed with sufficient distance can be reasonably reduced
without compromising the overall model accuracy. By the same token, inlet and outlet bound-
aries should also be placed away from calibration locations.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

124   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

• Initial conditions: An initial condition is like a boundary condition. Initial condition is the
boundary condition specified in the time domain. If the simulation starts cold, typically from
a uniform initially guessed flow field, a sufficient ramp-up or warm-up period should be given.
Even if the simulation starts from a presimulated restart file, a model may still need some time
to adjust if the restart file was generated with boundary conditions different from those in the
current simulation. For steady-state simulations, pseudo time is typically used by 2D models.
An appropriate initial condition can accelerate the convergence to a steady state.
• Solver selection: 2D hydraulics models may provide different solver options that solve different
versions of the governing equations. For example, HEC-RAS 2D 6.x has three solver options,
namely the diffusion wave, the ELM, and the newly introduced EM. For SRH-2D, there is one
solver option that solves the full governing SWE. In the future, more solver options will be
made available in SRH-2D. A modeler should make a well-informed decision based on the
applicability and limitations of each solver option. For example, the diffusion-wave solver
in HEC-RAS 2D is not suitable for bridge hydraulics because it ignores the inertia terms in
momentum equations and has no turbulence model. Inertia terms are important near hydraulic
structures where flow velocity has significant local changes. In general, a more simplified solver
option computes faster. However, as some flow physics have been simplified or ignored, the
result will not be as accurate.
• Turbulence model: Different 2D models may capture turbulence differently. Turbulence is
complex, and there are still many unsolved problems. However, many useful models have
been proposed, and they work relatively well if used properly. Many 2D hydraulics models
implement more than one turbulence model. For example, SRH-2D has the Laminar (turbu-
lence is off), Constant, Parabolic, and KE options. HEC-RAS 2D adopts the eddy-viscosity
approach, in which the effects of turbulence on mixing and momentum are quantified by a
turbulent eddy viscosity. HEC-RAS 2D considers turbulent shear in vertical and horizontal
directions. A modeler should consult the user manuals for 2D models on how to make the
proper selection. It is also important for a modeler to have some background in turbulence
and mixing processes. The blind use of default turbulence options and parameter values is
strongly discouraged.
• Other model parameters: In addition to the parameters discussed previously, there are num­
erous other model parameters that a modeler should specify. Each 2D model gives its users the
freedom to change certain parameters. For example, HEC-RAS 2D offers a lot of freedom to
change the model parameters, such as solver convergence tolerance, maximum iterations, and
even linear-equation solvers. Among all these model parameters, probably the most impor-
tant one is the time-step size, which has physical and numerical limits. The time-step size of
a 2D model is largely limited by cell size and flow velocity, which is constrained through the
Courant number. It is also affected by the numerical scheme used in a 2D model. Explicit
schemes usually require much smaller time steps than implicit schemes. Above all, time-step
size should be limited by model result accuracy. The larger the time-step size, the larger the
numerical error. The best practice is to specify a time-step size large enough to save computa-
tional cost and small enough to stabilize the model runs and reduce numerical error. Often a
model with stability issues can be improved with smaller time steps.

5.2.3 Step 3: Simulation


The model progress and status should be monitored during simulation runs. If any problems
are identified, they should be addressed. Model convergence on a meaningful solution is required.
For difficult or complex cases, some level of instability at the beginning can be acceptable if the
instability does not amplify, oscillate, or keep the model from converging.
• Numerical convergence: Numerical convergence has several layers of meanings in the con-
text of 2D modeling. Two-dimensional models discretize the governing equations on a mesh.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Development of 2D Modeling Best Practices 125

The governing differential equations are then transformed into an equivalent sparse linear-
equation system. An iterative method is typically used to solve the linear-equation system (with
some exceptions, such as HEC-RAS, which has the option of using direct-solution methods).
The iterations are often stopped once a preset convergence criterion has been achieved.
Another aspect of numerical convergence is for steady problems in which a pseudo time is
used and the simulation continues until the solution does not change (see Figure 5-6). The
best practice to check convergence is to check the time history of monitoring points and lines.
The monitored flow variables should asymptotically approach equilibrium values. If it is not
the case, the simulation diverges. There are many reasons for a model to diverge. The most
common reasons are poor mesh quality and excessively large time-step size.
• Stability: Stability refers to a model’s ability to remain stable during the simulation. The insta-
bility could be because of physical or numerical reasons. For physical reasons, instability does
not necessarily mean a model is wrong. The oscillation could be part of the solution. For
example, vortex shedding behind a bridge pier happens when the Reynolds number is high
enough. In other cases, the oscillation may be numerical, and a modeler should troubleshoot
and identify ways to remove the artificial oscillation. Common causes of numerical oscilla-
tion are poor mesh quality, large time-step size, turbulence model, and high-order numerical
schemes. If the order of the numerical scheme is adjustable, the best practice is to start with
low order. If the instability still exists, reduce time-step size. If a 2D model implements com-
plex turbulence models in which additional differential equations are solved, turn off the
turbulence model to check whether it is the cause.
• Mesh resolution effects: Simulation results should not depend on a particular mesh. In other
words, the result should be mesh independent. In 2D hydraulics models, because of the inter-
play among mesh and resolved and unresolved parts of the terrain, it is hard to strictly define
mesh independence. This problem is at the heart of how to define Manning’s n on different
meshes. Nonetheless, mesh resolution should be as reasonably high as possible, especially in
areas of interest, to increase numerical accuracy.
Currently, for most projects in practice, terrain data and mesh cannot resolve too many
roughness features. Thus, the Manning’s n specified in 2D models should be roughly the total
Manning’s n for different land-use types from textbooks and hydraulics manuals.
• Calibration and validation: Calibration and validation are two different steps that serve dif-
ferent purposes. For calibration, the purpose is to properly set the model parameter values
(e.g., Manning’s n) so the model prediction matches with measurement data. Manual and
automatic calibration methods are used in practice. Many 2D models and their graphic user
interfaces (GUIs) provide automatic Manning’s n calibrations with optimization algorithms.
It is recommended to use these automatic calibration tools. Once the model parameters have
been calibrated, they should be further validated with an independent set of data that has not

Figure 5-6. Model


convergence and stability
(lines and texts are color-
coded).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

126   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

been used in calibration. This is to ensure the generality of the calibrated model. There might
not be enough data to do both calibration and validation. In such cases, calibration should be
the priority.
• Uncertainty and sensitivity analysis: Because of the complexity of real-world rivers and their
floodplains, there are tremendous uncertainties in model parameters, especially in Manning’s
n. Textbooks and hydraulics manuals usually give a range of Manning’s n values for different
land-use types. It is often uncertain what exact value to use for a particular case. Even when
calibration data are available, there are uncertainties in them, too. On the other hand, it is often
of interest to check how sensitive the result is to Manning’s n. In general, for most subcritical
rivers, the sensitivity of the results to the selected Manning’s n values increases moving upstream.
This is because flow resistance is like skin friction in pipe flows. The head loss due to skin friction,
also known as global head loss, takes distance to accumulate.
For projects where measurement data are not available, Manning’s n can be set with the
best guess. It is strongly recommended that uncertainty and sensitivity analysis should be per-
formed, and decisions should be made based on acceptable risks and exceedance probabilities.
Uncertainty and sensitivity analysis should be performed on not only Manning’s n, but also
other variables that are uncertain or sensitive.

5.2.4 Step 4: Postprocessing


In the postprocessing step, the simulation results are processed, analyzed, and documented.
Everything done in the postprocessing step should serve the purpose of the modeling project.
• Result extraction and export: All 2D models and their GUIs provide tools to extract and export
results for further use, including the sampling of results at points, long lines, across a surface,
and more. For moderately sophisticated postprocessing, tools like integrated calculators pro-
vided by 2D models and their GUIs can be used. For more advanced processing, results can be
exported in a common scientific data format like the Visualization Toolkit, which can be read
by visualization software or scripts written in languages such as Python for further calculation.
Simulation results can also be exported using specialized tools in 2D models and their GUIs
for external tasks such as bridge scour analysis.
Some result-extraction operations should be specified before the simulation starts, for
example, monitoring points and lines. After the simulation, results still can be sampled at
points and along lines. However, only the results at saved time steps can be sampled. Results
at intermediate time steps are lost.
• Visualization: All 2D models and their GUIs provide some functionality to visualize the results,
such as contour and vector plots, streamline animations, and extracting inundation areas.
A modeler should consult the user manuals. Many 2D model software can make impressive
plots that have higher resolution than the computational mesh. Though visually pleasing, such
plots should be viewed with caution because the result has been numerically interpolated onto
a finer grid. Strictly speaking, it is not the real result. Sometimes nonphysical flow behavior may
be observed (e.g., the interpolated velocity vector goes into a wall).
• Data archiving: All data and model setups should be properly archived with protocols set by
the modeler’s organization or required by the client. Metadata and documentation should be
archived with the original data.
• Documentation and reporting: All modeling projects should have good documentation with
information on all aspects of the project shown in Figure 5-1.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CHAPTER 6

Development of a Decision-Making
Process for Selecting Manning’s n
for 2D Models
6.1 Overview
Manning’s n is a single variable to capture all factors contributing to flow resistance. Though
simple to use, the determination of its best value is difficult. The difficulty stems from the com-
plexity in the real world, the nonlinearity of the governing SWE, and the interdependence of
all variables involved. In addition, the definition of Manning’s n is different in various contexts.
In traditional open-channel hydraulics, Manning’s n is a physical parameter to quantify the
flow resistance force over an area—the rougher and the higher the flow resistance, the larger
the Manning’s n. In 2D models, Manning’s n is a half-physical, half-numerical parameter that
depends on terrain data, mesh, and even numerical schemes implemented in models.
Despite the difficulties, the findings from this project can be used to establish some gen-
eral guidelines. The guidelines are in the form of a step-by-step decision-making process for
modelers. These guidelines should be used in conjunction with the best practices described in
Chapter 5.

6.2 Decision-Making Process


The decision-making process is plotted as a flow chart in Figure 6-1 and described in the fol-
lowing steps:
1. Determine the purpose of the modeling project. Different modeling study purposes will affect
how flow resistance factors are treated or modeled. For example, obstructions can be modeled
with the following approaches: hole-in-mesh, disabled elements, elevated terrain, increased
Manning’s n, and additional drag force.
Hydraulic modeling purposes can be categorized into local and large-scale studies; expla-
nations follow:
(a) Local hydraulics, for example, hydrodynamics and sediment transport around bridges,
river training structures, and large woody material. Detailed hydraulics is important, and
hydraulic structures are better to be resolved in the mesh (e.g., using the hole-in-mesh
approach).
(b) Large-scale study, for example, flood inundation study and habitat study. The focus is
not on local hydraulics. Small-scale flow resistance factors such as bridge piers may not
be resolved in the mesh. Instead, other more economical approaches, like increased
Manning’s n, additional drag force, and elevated terrain, may be adopted to model their
effects to flow.
2. Select a 2D model. The selection of a 2D model should be based on its ability to provide output
to support the modeling purpose, strengths, weaknesses, availability, cost, modeler’s famil-
iarity and experience, acceptance and reputation of the model, and prior usage for similar

127

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

128 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 6-1. Flow chart for the decision-making process.

purposes. In the context of specifying Manning’s n, a modeler should consider the availability
of different approaches for roughness factors in different models. For example, SRH-2D can
model a bridge pier using the hole-in-mesh approach, while HEC-RAS 2D can only use other
approaches. Example 2D models include SRH-2D, HEC-RAS 2D, RiverFlow2D, and many
others.
3. List flow resistance factors in the domain and decide how to treat them in 2D models.
(a) Which flow resistance factors exist in each area? List all of them. Common factors and
whether they should be considered in Manning’s n are listed in Table 6-1.
(b) Determine how to model these identified flow resistance factors and whether they are
resolved. For some of the factors, how to model them is determined by the modeling
purpose. For example, for local bridge hydraulics, a bridge pier is best modeled with
the hole-in-mesh approach. For large-scale flood studies or when there are many small
bridge piers, other approaches can be used.
For a roughness factor or feature to be fully resolved, all of the following must be satisfied:
i. Terrain data should have sufficient resolution to represent them. For example, for
bedforms, large boulders, bars, and other features and obstructions in river channels,

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Development of a Decision-Making Process for Selecting Manning’s n for 2D Models   129

Table 6-1.   Factors affecting roughness values and whether they should be included
in the Manning’s n in 1D and 2D models.
General Description on the Should Effect be Included
Characteristics Affecting in Manning’s n?
Factors Manning’s n Value 1D 2D
Surface Yes (Most DEM cannot
Roughness Size and shape of sediment grains Yes resolve them)
Vegetation Height, density, distribution, and types Yes Yes
Variations in cross section, size, and No (should be captured
Channel channel shape (e.g., bars, sand waves, by high-resolution
Irregularity and bank-failure slumps) Yes mesh)
No, if extra terms for
centrifugal force are
Channel included in model
Alignment Meanders and channel curvatures Yes equations; yes otherwise
Yes, if terrain data or
mesh is coarse; no if
Sedimentation, also called silting, can terrain data and mesh
Sedimentation smooth out irregularities. Scour holes are fine and capture the
and Scour increase irregularity. Yes bed features.
No, if the obstruction is
captured by the 2D
model (either with
No (bridges and mesh or special
Examples include bridges and culverts are in 1D treatment); yes
Obstructions hydraulic structures. models) otherwise.
Yes (if n can be
specified as a function
Yes (HEC-RAS 1D has of water depth). For
Manning’s n generally decreases to an an optional calibration deep water, n can be
Stage and equilibrium value with the increase of tool to generate flow approximated as
Discharge stage and discharge. versus roughness table) constant.
Yes (Manning’s n Yes (Manning’s n
Seasonal should be a function of should be a function of
Change Mainly a concern for vegetation growth time) time)
Transport of sediment extracts Yes (currently no Yes (currently no
Sediment energy from flow and increases hydraulic model hydraulic model
Transport Manning’s n. considers this explicitly) considers this explicitly)
Mesh and
Terrain Data
(Resolution Determine how much of roughness Yes, if they are of low Yes, if they are of low
and Quality) features are resolved. resolution. resolution.

the bathymetry data resolution should be high enough to fully describe their exis-
tence. Always compare terrain data with aerial images and field photos to determine
whether any significant roughness factor or feature is missed.
ii. The mesh should have sufficient resolution to geometrically resolve the factor or
feature. At least several grid points (e.g., 6 or 7) are needed over one characteristic
size of the factor or feature. This characteristic size could be the wavelength of the
bedform, the diameter of a bridge pier, or the width of a channel.
iii. The mesh should have sufficient resolution in the vicinity or influence zone of a
roughness factor or feature to hydraulically resolve their effects. For example, the
mesh should be refined upstream and downstream of a bridge pier to capture the
backwater effect and wake.
(c) If a roughness factor or feature is fully resolved, it should not be considered in Manning’s n.
On the other hand, if a roughness factor or feature is not fully resolved or not in the terrain
or mesh at all, then its flow resistance effect should be considered in Manning’s n.
4. Determine the total Manning’s n for each roughness zone. For a roughness zone, the total
Manning’s n is defined as the n coefficient considering all roughness factors except local flow
resistance induced by, for example, bridge piers and other obstructions. Bridge piers and

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

130   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

obstructions should be accounted for with their selected modeling approach. Each rough-
ness zone should have a relatively uniform roughness condition. For example, it should be
covered by similarly sized sediment, bedform, vegetation, or land-use type.
There are at least three different methods to determine the total of Manning’s n:
(a) Tabular method: In this method, tables for Manning’s n values for different land use and
surface conditions are provided [see Chow (1959); Aldridge and Garrett (1973); Arcement
and Schneider (1989); and Brunner (2021)]. There are also tools that a modeler could use,
for example, the Excel spreadsheet package developed by USDA (Yochum 2018).
(b) Photographic method: In this method, Manning’s n can be determined based on repre-
sentative field photos (Barnes 1967; Aldridge and Garrett 1973; Hicks and Mason 1998;
Yochum et al. 2014). Moreover, USGS also developed a website (https://wwwrcamnl
.wr.usgs.gov/sws/fieldmethods/Indirects/nvalues/index.htm) that provides many photos
and their corresponding Manning’s n values. Photo-based guidance documents help
modelers gain more experience in selecting proper Manning’s n by looking at the appear-
ance and geometry of river channels and floodplains.
(c) Quantitative method: In this method, Manning’s n is calculated as a function, such as a
sediment size, water depth, or other relevant variables. There are many such formulas
for different roughness conditions, most of which are empirical. They can be either fully
quantitative or quasi-quantitative (Yochum 2018):
i. In fully quantitative methods, Manning’s n is typically estimated as a function of
relative submergence based on bed-material size and bedform fluctuations. The
drawbacks of these methods are their significant errors and ignoring other sources
of resistance, including obstructions, bank irregularity, vegetation, and sinuosity
(Yochum 2018). If vegetation is the main cause of flow resistance, a modeler can
select an appropriate formula for vegetation in Section 4-5.
ii. Quasi-quantitative methods are developed to overcome the weakness of fully quan-
titative methods. In these methods, the total Manning’s n is considered as the sum
of several independent components affecting flow resistance. As discussed in the
project report, the linear sum of Manning’s n values from different flow resistance
factors is not a physically sound approach. Its use is discouraged.
5. Determine resolved and unresolved Manning’s n for each roughness zone. The relationship
among total, resolved, and unresolved Manning’s n values is as follows:

n 2 = n 2resolved + n 2unresolved (6-1)

where n is the total Manning’s n determined from the previous step. This equation can be
reformulated as
J n resolved N2 J n unresolved N2
KK OO + KK OO = 1 (6-2)
L n P L n P
Graphically, it is a quarter of a unit circle (see Figure 6-2).
Notice there are three representative areas in Figure 6-2.
• One is circled with a blue solid line. This is when the resolved Manning’s n is between
0–60% of the total Manning’s n. This happens when terrain or mesh, or both, have low to
moderate resolutions. In this scenario, the unresolved Manning’s n is in a relatively narrow
range of 80%–100% of the total Manning’s n. For moderate resolution of terrain and mesh,
a modeler may choose, for example, nunresolved = 0.8 to 0.9 n as a starting point. For low-
resolution terrain and mesh, use nunresolved = n.
• The second area is circled with a blue dashed line. This is when the resolved Manning’s n
is 80%–90% of the total Manning’s n. This happens when terrain and mesh both have

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Development of a Decision-Making Process for Selecting Manning’s n for 2D Models   131

Figure 6-2.   The quarter-unit circle for the


relationship among the total, resolved, and
unresolved Manning’s n values.

extremely high resolutions. In this scenario, the unresolved Manning’s n changes quickly
in the range of 0–50% of the total Manning’s n. Fortunately, most of the flow resistance has
been resolved by terrain and mesh. The unresolved part plays only a minor role. A modeler
may choose nunresolved = 0.25 n, which is in the middle of the range circled by the red
dashed line.
• The third area is the quarter-unit circle outside the two circled areas. This illustrates the
resolved and unresolved portions of Manning’s n are comparable. This happens when the
terrain and mesh have moderate resolutions to resolve part of the roughness factors or
features. A modeler may choose nunresolved = 0.7 n. Note that because of the summation
of the square relationship shown in Equation 6-2, unresolved = 0.7 n means that about half
of the flow resistance is unresolved and the other half is resolved.
In the above guidelines, whether a terrain or mesh is of high or low resolution is relatively
speaking to the characteristic size of roughness factors or features. As already described, to fully
resolve a factor or feature, at least several grid points are necessary to cover one characteristic size.
6. Use the unresolved Manning’s n in 2D models. It is important to note that in 2D models,
Manning’s n is used only to parameterize the unresolved part of roughness or flow resistance
factors. The resolved part is already embedded in terrain and mesh.
7. Miscellaneous considerations.
– For bridge piers, if increased Manning’s n or additional drag-force approaches are used
and the pier covers more than one mesh cell, the Manning’s n and drag coefficient should
have a much larger value than those suggested in textbooks or 2D model user manuals.
This is because of the drawback in the current implementation of most 2D models. With
a coarse mesh, if the bridge pier is much smaller than the mesh cell that contains it, the
drawback is less of a problem.
– If converting a 1D model to a 2D model, the Manning’s n of river channels and their flood-
plains in the 1D model can be used as a starting point for the 2D model.
If the contraction–expansion is significant, then it is difficult to ascertain that Manning’s n
in a 1D model should be larger or smaller than that in a 2D model. This is because in
1D models, contraction–expansion coefficients, in conjunction with Manning’s n, are used
to account for the head loss from one cross section to another. On the other hand, in
2D models, the only tunable variable is the Manning’s n. The contraction–expansion

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

132   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

effect should already be considered in terrain and mesh. Depending on what contraction–
expansion coefficients are used in 1D models, the 1D Manning’s n could be higher or lower
than the value needed in 2D models.
If significant shear exists between the main channel and its floodplains, for example,
because of significantly different channel conditions and vegetation, then Manning’s n in
1D models should be higher than in 2D models. In other words, the 1D Manning’s n should
be reduced to be used in 2D models. This is because shear and the related head loss cannot
be directly modeled in 1D models. Their effects must be lumped into Manning’s n. How-
ever, in 2D models, the shear between the main channel and its floodplains is resolved, and
there is no need to consider it in the Manning’s n.
In addition, in 1D models flow is always assumed to be distributed within a cross sec-
tion based on the incremental conveyance. Therefore, two neighboring cross sections that
have differing conveyance in adjacent areas will redistribute the flow laterally without an
associated energy cost. A 2D model does not make this simplifying assumption, so any
flow redistribution must be physically realistic. The lateral redistribution in 1D models can
occur at large and small scales, so 1D Manning’s n values would need to be larger. Basically,
2D models hydraulically resolve variable conveyance and lateral movement of water that
1D models cannot hydraulically resolve fully.
– For 2D models that implement the subgrid terrain treatment, such as in HEC-RAS 2D, the
treatment may have some advantage when the mesh is exceedingly coarse. The advantage
is mainly in seeing the missed subgrid small channels and correcting the wetted perim-
eters at cell faces. It also tends to better approximate the volume–depth relationship of
computational cells.
Many modelers have assumed that because the treatment sees more terrain details,
more roughness factors or features are resolved, and, therefore, the Manning’s n should be
smaller. This assumption is not necessarily true. The reason is that the subgrid terrain treat-
ment is applied only to some, not all, terms in the governing equations. The gravity force
term involves the bed slope, and its calculation does not use subgrid terrain treatment. In
addition, the bed slope is sensitive to mesh resolution. Meshes with different resolutions
will represent the terrain differently. Consequently, the bed slope will be different. The
advantage of subgrid terrain treatment may be overshadowed by the bed-slope sensitivity.
When comparing different 2D models, such as SRH-2D and HEC-RAS 2D, many mod-
elers have assumed that because HEC-RAS 2D implements the subgrid terrain treatment
while SRH-2D does not, the Manning’s n in SRH-2D should be higher than that in HEC-
RAS 2D for the same roughness zone. This assumption is again not necessarily true. There
are at least two reasons for that. The first reason is again the bed-slope term. SRH-2D and
HEC-RAS 2D calculate the bed-slope term differently. SRH-2D calculates a much smoother
bed-slope term than HEC-RAS 2D. When terms in the governing equations are calculated
differently in different models, it is hard to draw a general conclusion on how the subgrid
terrain treatment affects Manning’s n. The second reason is that for cases with reasonable
mesh resolution and following the best practices described previously, the advantage of sub-
grid terrain treatment diminishes because the difference between terrain data and the mesh
representation of the ground is small. In these cases, the cell faces tend to be relatively
straight lines instead of curves.
– When comparing resolved versus unresolved Manning’s n, the underlying assumption is
that higher resolutions of terrain and mesh will resolve more roughness factors and fea-
tures, and therefore have higher resolved Manning’s n and lower unresolved Manning’s
n. This is generally true. However, there are always exceptions because of the nonlinear
nature of the governing equations, how all terms in the equations interact with each other,
and especially how the bed-slope term responds to mesh resolution.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Development of a Decision-Making Process for Selecting Manning’s n for 2D Models   133

– In many real-world cases, especially for large-scale flood studies, it can be hard to obtain
high-resolution terrain data and much harder to use high-resolution mesh. In these cases,
it is not uncommon that the terrain represented in the mesh can show only the gen-
eral trend without too many details on roughness. In these cases, the resolved part of
Manning’s n is almost zero.
– The sensitivity of 2D results to Manning’s n is dependent on specific cases. In general, the
sensitivity decreases as the water depth increases. This is because the flow resistance term
in 2D SWE is proportional to water depth to the –1∕3 power. A larger water depth reduces
the relative importance of the flow resistance term in the governing equations.
The sensitivity of the 2D result to Manning’s n for mild- and steep slope flow profiles
is as follows. For 2D cases, the flow profile types can be estimated with flow discharge,
mean channel slope, and total Manning’s n. For mild slopes, the sensitivity increases in
the order of M1, M2, and M3 curves. Similarly for steep slopes, the sensitivity increases
in the order of S1, S2, and S3 curves.
The sensitivity described above can be used by a modeler to better understand model
behavior and determine how critical it is to have an accurate Manning’s n. For example,
if the flow in a 2D case generally follows an M1 curve, the effect of changing Manning’s n
will take a long distance upstream to show. If the 2D simulation domain is short (e.g.,
for local bridge hydraulics), the overall result will not respond quickly to the change of
Manning’s n. On the other hand, if the flow generally follows an M2 curve, the result will
respond quickly to the change of Manning’s n in a much shorter distance.
For certain applications, the insensitivity of the result to Manning’s n does not mean
an accurate Manning’s n is not necessary. For example, if sediment transport is also simu-
lated, the bed shear stress depends on Manning’s n squared, and sediment transport can
be related to excess shear to the 1.5 or greater power. Therefore, sediment transport can
be extremely sensitive to Manning’s n. The bed shear might not be the dominant term in
the governing SWE. However, it is critical for the calculation of sediment transport. In this
case, an accurate Manning’s n is important.
– For 2D models that do not implement any turbulence model, such as in RiverFlow2D, the
Manning’s n should generally be higher than the calculated unresolved Manning’s nunresolved.
The extra flow resistance is used to partially model the effect of turbulence. Currently, it
is difficult and perhaps impossible to provide a guideline on what percentage increase to
nunresolved. In general, the more turbulent the flow, the higher the percentage.
– In computational modeling for fluid dynamics, it is required that the results are mesh
independent. If different meshes produce different results, there is no way to know which
result to use. In practice, a mesh independence study is performed with progressively
refined meshes until the results do not change significantly.
A mesh independence study should also be performed for 2D hydraulics modeling.
However, there is a complication due to the interdependence of mesh resolution and
Manning’s n (and other model parameters, such as eddy viscosity). Strictly speaking, there
is no mesh independence for 2D hydraulics modeling. Whenever the mesh resolution
changes, the resolved part of roughness changes, and the unresolved Manning’s n should
also change. In HEC-RAS 2D, its eddy viscosity is also mesh cell size dependent. To avoid
this dilemma, a mesh independence study and a calibration should be performed concur-
rently. In the end, the mesh resolution should be reasonable, and the Manning’s n values
should not be too far from the estimated nunresolved.
– If data are available, calibration should always be performed to fine-tune Manning’s n. The
previous guidelines can only provide a best guess on Manning’s n. The more accurate value
should be calibrated with real measurement data. If more data are available, it is beneficial
to perform validation for the calibrated model. Validation data should be different from
calibration data to check the generality of the calibrated model.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

134   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

– If calibration is not possible because of the lack of data, it is recommended to perform


sensitivity analysis on Manning’s n. In sensitivity analysis, different Manning’s n values are
used to simulate a case, and then the results are compared. The sensitivity (i.e., the change
of results) should be in a reasonable range.
– In a simulation domain with multiple roughness zones, researchers need to identify these
zones, which are more engaged in the flow, have large area, and have more impact on
the overall hydraulics. The calibration of Manning’s n should be focused on these zones,
starting with the ones conveying more flow. For example, roughness zones at the edge
of flood inundation areas may have less impact on the flow than the zones right in the
flood path. Changing Manning’s n for edge zones may not have a significant response
at calibration locations. In this case, excessively changing Manning’s n values in these
edge zones may make them exceed the targeted total Manning’s n, which is not correct.
– For real-world applications, there are abundant uncertainties in data and models. A modeler
can try to minimize uncertainties, but it is impossible to remove uncertainties entirely.
The best strategy is to quantify how many uncertainties are within the results and make
risk-based decisions. In the context of 2D hydraulics modeling, Manning’s n is probably
the most important source of uncertainties even when high-quality calibration data are
available. There are many methods and techniques for uncertainty quantification (UQ),
for example, the Monte Carlo simulations. It is beyond the scope of this project to go into
the details of UQ, but it is a good topic for a follow-up research project.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

CHAPTER 7

Verification and Demonstration


of Proposed Guidelines

This chapter demonstrates the use of proposed guidelines and recommendations in Chapters 5
and 6 by highlighting three real-world cases.

7.1 Iowa River 2008 Flooding Case


7.1.1 Background Information and Data
The dataset for this case—which contains comprehensive data for model building and
calibration—was provided by Scott Hogan from FHWA. The team thanks Scott for allowing the
use of this case.
The Iowa River, a tributary of the Mississippi River, is a sand-bed river approximately 323 mi
long located in the state of Iowa. The river meanders over the plains in the Midwest and has been
impounded by several dams and spillways along its course. The river is also prone to flooding.
One recent severe flooding, which led to evacuations and road closures, occurred in 2008. This
case study is for the modeling of that event.
The modeling domain in this case is from the Coralville Reservoir to the south of McCollister
Blvd. The main stem of the modeled Iowa River is about 10 mi long and includes 14 bridge cross-
ings and 2 dams (or weirs). High-water marks (HWMs) and satellite images during the 2008
flood were collected and used for model calibration. Land-use data, which contain 11 material
roughness types, were also provided. The Manning’s n values for different roughness types were
determined and calibrated.
Figure 7-1a shows the overview of the Iowa River modeling domain. Figure 7-1b shows the
flood inundation extent in a satellite image. Figure 7-1c shows the DEM data used; the DEM has
a 2-ft resolution.
To better appreciate the terrain and bathymetry, which is directly connected to the roughness
and flow resistance, Figure 7-2 shows the 3D view of the overall DEM and a zoom-in view around
the Iowa River Dam and footbridge. Overall, the DEM is smooth, especially within the channel
itself. Note the DEM has been processed, filtered, and resampled, all of which tend to smooth
out roughness; however, this is extremely common in practice. For this case, the 2-ft resolu-
tion means that any roughness features smaller than 2 ft will be lost. For a sand-bed river, 2 ft is
probably larger than or comparable with the size of most ripples and small bedforms. Therefore,
within the channel, the DEM can resolve only the Manning’s n caused by roughness elements
larger than 2 ft.
Because of the importance of resolved versus unresolved roughness in the domain, it helps
to compare the DEM with ground truth. Figure 7-3 shows the comparison of DEM and satellite

135

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

136   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-1.   Overview of the Iowa River 2008 flood case: (a) modeling domain
(shown within white boundary), the river centerline (shown in blue), and satellite
image, (b) flooding extent shown in a satellite image, and (c) DEM of the terrain
(2-ft resolution). The flow is from top to bottom. The star in (a) shows a location
that is addressed later in this chapter.

Figure 7-2.   Terrain and bathymetry of the modeling domain.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   137

Figure 7-3.   Comparison between DEM and satellite image to show the difference
of how much ground roughness is represented in the DEM.

imagery for the zoom-in region. Again, the in-channel DEM is quite smooth and doesn’t show
any small bed roughness elements. In the floodplain, the DEM is also smooth, for example, near
the inner bank of the meander. However, the satellite shows dense trees and vegetation, whose
flow resistance effect must be parameterized with the Manning’s n. The DEM does a good job
in capturing important ground controls, like the high elevation of roads and small hills, and
the spillway of the Iowa River Dam. Therefore, it would be redundant to add a weir structure in
2D hydraulic models.
For model calibration, 23 HWMs were provided. These HWMs are invaluable for the deter-
mination of Manning’s n. Figure 7-4 shows their locations. The numbering of 1 to 23 is in order
from upstream to downstream.

7.1.2 Demonstration of Proposed Guidelines


This section examines how to use the step-by-step decision-making process to determine
roughness factors and their corresponding Manning’s n.
1. Determine the purpose of the modeling project.
The purpose of this project is flood study and not bridge hydraulics. Therefore, the bridge
piers can be modeled as obstructions with one of the approximation methods, such as the added
drag-force method. Some of the bridges were pressurized during the flood. Therefore, they also
need to be modeled as pressure structures regardless of whether bridge piers are resolved by
mesh or not.
For the comparison of different approaches for modeling hydraulic structures, a case where
the bridge piers were modeled with the hole-in-mesh approach was created.
In addition to bridges, there is also a dam (or weir) within the Iowa River Dam and foot-
bridge. Dams and weirs are river control structures that greatly influence the backwater curve.
As mentioned previously, the DEM considers the dam to be terrain; therefore, it is redundant
to add a weir structure. However, to fully capture the effect of the dam, the mesh resolution
near the dam should be fine enough to capture this terrain feature.
There are no other significant hydraulic roughness features (e.g., large wood material) in
the domain.
2. Select a model based on its ability to provide output to support the purpose determined in the
previous step, strengths and weaknesses, availability, cost, modeler’s familiarity and experi-
ence, and prior usage for similar purposes.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

138   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-4.   Locations of the 23 HWMs for the Iowa River


flood case.

In this case, SRH-2D was selected for its familiarity. Other models may be used; however,
some of them may not have all the options for hydraulic structures (e.g., HEC-RAS 2D cannot
model bridge piers with hole-in-mesh). For the purpose of this project, comparisons need to be
made among different modeling approaches. That is an additional reason to choose SRH-2D.
3. Make decisions on how to treat different flow resistance factors and determine the total
Manning’s n value accordingly.
– Which flow resistance factors exist in each coverage area (as determined by the modeler)?
List all of them.
From the DEM, satellite images, land use, and other information sources, the following
flow resistance factors are present in the modeling domain: surface roughness, vegetation
(mostly in the floodplains), channel irregularity, channel alignment, sedimentation and
scour, obstructions, and ground controls. The land use of the modeled area is shown in
Figure 7-5.
Consider the resolution and quality of DEM and mesh data. The DEM has a resolution of
2 ft, and mesh resolutions are coarser than DEM (described in the next section). Therefore,
any roughness factor or feature smaller than the DEM resolution will not be fully captured
in the terrain. The coarse resolution of the mesh will further filter more roughness features.
Based on these considerations and the meshes used, the following can be stated:
▪ Surface roughness: This refers to relatively smaller roughness elements (e.g., sedi-
ment grains, ripples, and other small bed features). These features are typically much
smaller than 2 ft. Therefore, they are unresolved by the DEM, let alone the meshes.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   139

Material Types:
unassigned
Dense Trees
Channel Type 1
Moderate Density Trees/Rural Development
Low Density Trees/Veg
Roads
Overbank ponds
Urban Area
High Density Residential
Channel Type 3
Weirs
Channel Type 2
Feature Arc

Figure 7-5.   Land-use data for the Iowa River


flood case.

The Manning’s n for the unresolved surface roughness factor in 2D models should
be equal to the total Manning’s n. For a typical sand-bed river, like the Iowa River,
a value of 0.03 is used for the main channel throughout the domain. Although there
are three different land-use types for the upstream, middle stream, and downstream
of the main channel, there is no sufficient information to differentiate their roughness
characteristics. Therefore, a single value is used for all of them.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

140   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

▪ Vegetation: Vegetation and trees are primarily in the floodplains adjacent to the main
channel. Based on the land-use data, there are different areas with dense trees, moderate-
density trees, and low-density trees and vegetation. During the 2008 flood, overbank
flow and inundation occurred; therefore, these near-river vegetation zones played a role
in retarding the flood flow. The DEM data and the meshes do not have any representa-
tion of the trees and vegetation. Therefore, the resolved part of vegetation Manning’s n
is zero, and the unresolved part is equal to the total Manning’s n. The total Manning’s n
values for vegetation and forests with different densities can be found in hydraulics
tables. In this case, the values of 0.18, 0.08, and 0.05 were used for dense, moderate, and
low-density trees and vegetation, respectively.
▪ Channel irregularity: The river channel is relatively regular, and the channel width
and depth remain nearly uniform. Even if the channel shows significant irregularity,
the mesh in 2D models should capture these changes. Therefore, their contribution to
Manning’s n should be mostly resolved, and no additional consideration is needed in
2D models.
▪ Channel alignment: The river channel has significant meander. There are several high-
curvature bends in the modeling domain. The meshes in 2D models can capture mean-
der. However, none of the models evaluated in this work specifically considers the
hydrodynamics in meander bends. Therefore, the flow resistance effects of secondary
current and other flow dynamics within the bends must be lumped into the overall
Manning’s n value for the channel. Their effects also can be considered in the param-
eters for turbulence models. Overall, the meander bend hydrodynamics is partially
resolved by the 2D mesh. In this work, for simplicity, the river channel Manning’s n
for meander was not increased because its effect is local and of secondary importance.
▪ Sedimentation and scour: No sediment and scour information is available during the
flood; therefore, they are not considered.
▪ Obstructions: There are 14 bridges and a dam in the modeling domain. For flood study
in which the domain is typically large, bridge piers are recommended to be modeled as
obstructions, not with the hole-in-mesh approach. In this work, however, bridge piers
were modeled as obstructions and with the hole-in-mesh approach for comparison.
▪ When bridge piers are modeled as obstructions, additional drag force is added to the
flow. In addition to the parameters such as bridge pier shape and dimensions, the key
parameter is the drag coefficient. As discussed previously, because of the current imple-
mentation of obstruction drag in all 2D models evaluated, this drag coefficient value
should be much larger than what is recommended in hydraulics textbooks and manu-
als. Two drag coefficient values, 4 and 100, were tested.
▪ Hydraulic controls: There are some hydraulic controls, for example, roads, berms, and
in-channel features. Meshes generated in this work can capture these features well
through local refinement and breaklines. If they are not captured by the DEM and mesh,
water will flow over the high ground, or an artificially high Manning’s n is needed in
the adjacent area to prevent flow from spilling over. The results section will feature one
such example and explore why it is critically important to identify and properly model
these hydraulic controls.
The total Manning’s n values for all land-use types, which are also the values used in SRH-2D
because the resolved part for all land use is quite small, are listed in Figure 7-6. Only the Manning’s n
values for areas engaged in the flood water pathway are relevant. The more engaged with flow, the
more important the area’s Manning’s n is. Areas closer to the main channel are generally more
important. By comparing the land use in Figure 7-5 and the flood inundation satellite image in
Figure 7-1b, it is easy to see that land-use types Channel Type 1, Channel Type 2, Channel Type 3,
Dense Trees, Moderate Density Trees/Rural Development, Low-Density Trees/Veg, and perhaps

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   141

Figure 7-6.  Manning’s n values for different land uses for the Iowa River flood case.

also High Density Residential have greater influence on the 2D modeling results. In contrast, the
area labeled “Overbank ponds” in Figure 7-5 is probably less critical.
The guidelines proposed in this project were applied and demonstrated for this case. For most
land use and roughness factors, the resolved part of Manning’s n is almost zero because of the
coarse mesh and DEM resolution. Therefore, the unresolved Manning’s n (i.e., the value to be
specified in 2D models) should be approximately the total Manning’s n for each land-use area.
This is expected for flood modeling in an especially large domain.

7.1.3 Case Examples Setup


This section describes the setup of case examples, such as boundary conditions and mesh
generations.
For boundary conditions, there is one fixed-discharge inlet and one fixed-WSE outlet on the
main channel. There are also four tributaries. The SRH-2D boundary conditions are listed in
Table 7-1. Of the 14 bridges in the domain, 7 were pressurized during the flood. Therefore, pres-
sure structure boundary conditions were assigned to them accordingly.

Table 7-1.   Boundary conditions for the


Iowa River case example.
Boundary Type BC Value (cfs or ft)
Iowa River In�low INLET-Q 39,900
Rapid Creek In�low INLET-Q 475
Muddy Creek In�low INLET-Q 475
Clear Creek In�low INLET-Q 475
Ralston Creek In�low INLET-Q 475
Downstream Boundary EXIT-H 643.9
Note: cfs = cubic feet per second

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

142   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Table 7-2.   Meshes for the Iowa River case example.


Avg. Channel Avg. Near-Structure
Mesh No. of Nodes No. of Cells Cell Size (ft) Cell Size (ft)
Hole-in-mesh 30,371 47,732 50 by 100 20 by 30
Obstruction 30,306 47,776 50 by 100 20 by 30
Obstruction (coarse) 19,618 31,723 80 by 120 30 by 60
Obstruction (coarse, 2) 18,640 30,599 80 by 120 30 by 60

Four meshes were made. One mesh was made using the hole-in-mesh approach, where the
bridge piers were modeled as holes. Three other meshes were made where the bridge piers were
modeled as obstructions. For all meshes, the main channel was mainly covered with quadri-
lateral cells generated with the patch mesh generation method. At bridge crossings, the mesh
was greatly refined to accommodate the relatively small bridge piers. The rest of the domain
was covered with triangles generated with the paving mesh generation method. For the three
meshes where bridges were modeled as obstructions, the mesh near bridge crossings could have
been made coarser, and more quadrilateral cells could have been used. However, to have a fair
comparison between the hole-in-mesh and obstruction approaches, they were made comparable
with the hole-in-mesh mesh in this project.
The details of all the meshes (e.g., numbers of nodes and elements, approximate cell sizes), are
listed in Table 7-2. Among all meshes, the hole-in-mesh and obstruction mesh are comparable.
The obstruction (coarse) mesh is a coarser version of the obstruction mesh, with a reduction
of about 30% of cell numbers, and the cell size near structures is almost doubled. The obstruc-
tion (coarse, 2) mesh is similar to the obstruction (coarse) mesh and is used to demonstrate the
importance of resolving hydraulic controls. This mesh does not resolve an in-channel feature,
which resulted in a wrong backwater curve.
Three meshes were shown in Figure 7-7 [the fourth one is like the obstruction (coarse) mesh].
In Figure 7-8, the zoom-in view around some example hydraulic structures shows the difference
among different meshes.

Figure 7-7.   Meshes for the Iowa River flood case example.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   143

Figure 7-8.   Meshes for the Iowa River flood case example (zoom-in view around
hydraulic structures to show the differences).

For the obstruction cases, drag coefficient values of 4 and 100 were used to test their effects.
This resulted in a total of five cases as listed in Table 7-3. The computing time for all cases on a
desktop computer with Intel i7-8700 central processing unit (CPU) ranged from 1.5 to 2.7 h.

7.1.4 Modeling Results and Analysis


The modeling results and analysis are provided in this section.

7.1.4.1 Flood Inundation


Because the purpose of this study is for flooding, the flood inundation is one of the most
important results. Figure 7-9 shows the simulated inundation from Cases 1, 2, 3, and 4 with
different approaches for bridges. The simulated inundation is plotted on top of the inundation
satellite image. Overall, all four cases predict similar inundation patterns. There is no significant
difference if one looks at the full inundation maps between Case 1: Hole-in-mesh and the rest of
the cases using obstruction for bridge piers. This is probably not surprising because bridge piers,
although playing an important role in the local area near bridges, have relatively small impact
on the overall flood hydraulics. If the bridge piers are not choking the flow in narrow flood path-
ways, their impact is local.
However, this does not mean the bridge piers are not important. There are indeed some
differences in inundation at the local level among different cases. Figure 7-10 shows the local
inundation around the two central meander bends. Careful examination reveals some subtle
differences, especially near the water edge in the outer banks.
The inundation maps in Figures 7-9 and 7-10 still do not clearly and definitively show the
differences among the four cases. To better demonstrate the differences, Figure 7-11 shows the
zoom-in view of inundation around one of the bridges downstream. For Case 1: Hole-in-mesh,
the circle shows the flood water from upstream of the bridge spills over around that end of the
bridge and then joins the channel downstream.

Table 7-3.   Cases for the Iowa River example.


Case No. Mesh CD Value Computing Time (h)
1 Hole-in-mesh N/A 2.73
2 Obstruction 4 2.67
3 Obstruction 100 2.64
4 Obstruction (coarse) 100 1.55
5 Obstruction (coarse, 2) 100 1.51

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

144   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-9.   Overall inundation results from different case examples of the Iowa
River flood.

Case 1: Hole-in-mesh Case 2: Obstruction (CD = 4)

Case 3: Obstruction (CD = 100) Case 4: Obstruction (coarse, CD = 100)

Figure 7-10.   Local inundation results around two meander bends from different
cases of the Iowa River flood.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   145

Figure 7-11.   Comparison between the hole-in-mesh and obstruction approaches for bridge
piers of the Iowa River flood.

However, Case 2: Obstruction (CD = 4) does not show that spill over. This is because the
drag coefficient value of 4 greatly underestimates the drag force imparted by bridge piers. With
less drag force, flood water can pass through the bridge without the need to flow around and
spill over the road. The hole-in-mesh approach is generally more accurate than the obstruction
approach. Therefore, if Case 1: Hole-in-mesh is used as a benchmark, the drag coefficient value
needs to be increased. For Case 3: Obstruction (CD = 100), where its value is increased to 100, the
flood water does spill over the road as shown in the figure. This again confirms the conclusion
that the drag coefficient in current 2D models should have an especially high value to mimic the
real bridge pier hydrodynamics.

7.1.4.2 HWMs
The 23 HWMs were used to calibrate the Manning’s n and the drag coefficient values. Fig-
ure 7-12 shows the comparison between simulated and observed HWM. Note that the diagonal
line is the perfect-match line. The mean absolute error and the max error are reported on the
figure for each case and listed in Table 7-4.
From the reported error values, the hole-in-mesh approach for bridge piers is most accurate
because it produces the smallest errors. A comparison of Cases 2 and 3 again shows the inappro­
priateness of the small drag coefficient value of 4. With a drag coefficient value of 100, the errors
are reduced, but are still larger than those from Case 1. Note that the meshes for Case 1 and Cases
2 and 3 are comparable in terms of resolution, especially near the bridges. In practice, when the
obstruction approach is used for bridge piers, the mesh resolution around bridges may tend to be
coarser. As a result, the accuracy of the results using the obstruction approach may be reduced.
For Case 4, where the mesh is coarser, the errors are the largest among all.
It is intriguing to observe that HWMs #4 and #6 are so off for Case 4: Obstruction (coarse,
CD = 100). The difference between the simulated and observed water elevations is around 5 ft.
One may wonder why this happens and how this affects the Manning’s n calibration in practice.
A large error like this will likely drive the calibration and optimization algorithms to search for
a Manning’s n solution, which puts more weight on the HWM location with large error. In other
words, if this large error is induced by numerics instead of physics, the calibration process of
Manning’s n values will be artificially and wrongfully biased.
Careful examination of the simulation results revealed the two root causes of this large dis-
crepancy: the cell containing the HWM location is dry in the simulation result, and the mesh is
extremely coarse around the HWM location.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Figure 7-12.   HWMs comparison for the Iowa River flood case.
Copyright National Academy of Sciences. All rights reserved.
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   147

Table 7-4.   HWMs simulation errors for the Iowa River case.
Case No. Mesh CD Value Mean Absolute Error (ft) Max Error (ft)
1 Hole-in-mesh N/A 0.61 1.19
2 Obstruction 4.0 0.93 2.63
3 Obstruction 100 0.71 2.63
4 Obstruction (coarse) 100 1.14 5.34

Taking HWM #4 as an example, Figure 7-13 shows its position in the mesh and the cells
around it. Cell #30993 is dry and is right at the edge of the inundation boundary. The three nodes
of the cell and their elevations are also displayed. Based on the nodal elevations, the cell-center
elevation is (682.88 + 651.42 + 645.77)/3 = 660.02 ft. In this case, Node 9063 has a much higher
elevation than the other two nodes. This is because the mesh is so coarse that Node 9063 is far
away and happens to land on a higher ground. If the containing cell is dry, then the WSE for the
monitoring point (i.e., HWM #4) is assigned the value of the containing cell’s center elevation,
which is 660.02 ft in this case. However, the observed water elevation is only 655.2 ft, which is
about 5 ft lower.

Figure 7-13.   Why is HWM #4’s WSE prediction so off? Dry cell and coarse mesh.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

148   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

To further show the cause of artificially large WSE errors at HWM locations when the cell is dry
and mesh is coarse, Figure 7-14a shows the channel cross section at HWM #4. In Figure 7-14b,
the location of HWM #4 in the cross-section profile is labeled. The WSE profile is flat and has
an approximate value of 655 ft, which is closer to the observed value of 655.2 ft.
Based on the results and analysis, following are some recommendations:
• Locally refine the mesh near HWM locations.
• Adjust model parameters and mesh to ensure the HWM is wet. A HWM means water reached
that location; therefore, it must be wet in the first place.
• Change the way that WSE is extracted at monitoring points. For dry cells, instead of using
the containing cell’s data, perhaps it is better to consider the cross-sectional water profile as
shown in Figure 7-14.
• As a last resort, remove the HWM from calibration if it is dry in the simulation result.

7.1.4.3 Long Profiles Along River Centerline


The comparison of simulated long profiles along the river centerline further shows the effect of
different aspects of 2D models. Figure 7-15a shows the simulated WSE profiles along the center-
line of the river on top of the riverbed. Overall, along the roughly 10-mi-long river, all four cases
(Cases 1, 2, 3, and 4) produce comparable WSE profiles. However, there are some differences,
which may be large in some local areas. The bulk of the differences is within the bound of −0.5 ft
to 0.5 ft. There are also some local oscillations at approximately –2 mi for Case 4. The oscillations
are mainly around pressurized bridges.
Figure 7-15b shows the WSE difference of all three obstruction cases (Cases 2, 3, and 4) in com-
parison with Case 1 (hole-in-mesh). Among all three cases, Case 3 (obstruction with CD = 100)
has the smallest difference throughout the length of the river. For Case 2 (obstruction with CD = 4),
the simulated WSE is systematically lower than that from Case 1: Hole-in-mesh. This is as expected
because a lower drag coefficient means lower flow resistance (like lower Manning’s n) and therefore
lower WSE in the backwater curve.
Comparing Cases 3 and 4, the only difference is the mesh resolution. Because mesh resolu-
tion determines how much the roughness is resolved, and the rest is unresolved, the com-
parison may shed some light on the effect of specified Manning’s n values. The Manning’s n

Figure 7-14.   Why is HWM #4’s WSE prediction so off? A cross-sectional profile shows the location
of HWM #4: (a) location of the cross section and HWM #4 and (b) cross-sectional profiles.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   149

Figure 7-15.   Simulated WSE profiles along the centerline of the main channel. For Case 4, there are some local
oscillations approximately 22 mi around pressurized bridges.

values specified in 2D models are expected to be lower when higher-resolution mesh is used. If
the same Manning’s n values are specified for cases with different mesh resolutions, the result
with the refined mesh should produce a higher WSE profile because the effective Manning’s n is
larger. From Figure 7-15b, it is hard to determine whether this is true. One possible explanation
is that all the meshes used in this case are too coarse and beyond the point of resolving any of
the relevant roughness that contributes to Manning’s n. The difference observed between Cases
3 and 4 is the result of other factors (other terms in the governing SWE).

7.1.4.4 River Bathymetry Seen by Different Meshes


Checking how meshes with different resolutions represent the bathymetry and terrain is inter-
esting. Figure 7-16 shows the riverbed along the centerline as seen by different meshes. None of
the meshes fully represents the real bathymetry. They tend to smooth out some of the features,
especially high peaks and low troughs. The coarser the mesh is, the more smoothing.
It is also interesting to notice the small oscillations on the real bathymetry. These oscillations
have a wavelength of about 5 ft and an amplitude of about 0.05 ft. With such a small amplitude,
it is believed these oscillations are numerical artifacts when SMS interpolates the DEM to the
centerline; they are not ripples or bedforms. In other words, the DEM does not have the surface
roughness element information.

7.1.5 Importance of Resolving Ground Controls


In 2D modeling, it is important to resolve ground controls. This section compares Cases 4
and 5. The only difference between the two cases is whether a critical in-channel feature is
resolved or not. In the upstream section of the main channel (at River Mile –6.5 and shown with
a star in Figure 7-1a), there seems to be an expanded abutment and a subchannel within the
main channel.
Figure 7-17 shows that ground control feature resolved by the two meshes in Cases 4 and 5.
Case 4: Obstruction (coarse) has some refinement near this in-channel feature and, therefore,
resolves it fairly well. In contrast, the Case 5: Obstruction (coarse, 2) mesh on the right does not
have refinement in this region. It misrepresents the real channel bathymetry and does not resolve

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

150   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-16.   How do different meshes see the riverbed


elevation along the centerline of the main channel? On the
insets, the oscillations on the real bathymetry line have a
wavelength of about 5 ft and an amplitude of 0.05 ft.

the deep channel in the middle. As a result, there is an artificial, broad-crest weir across the
channel in Case 5: Obstruction (coarse, 2), which blocks the flow and creates significant back-
water curve upstream. This elevated backwater curve can be seen in Figure 7-18 where the WSE
profiles along the river centerline are plotted.

7.1.6 Uncertainty and Sensitivity Analysis


There are substantial uncertainties in the Manning’s n values. Calibration can only partially
reduce the uncertainties. Another related question is how sensitive the 2D modeling result is

Figure 7-17.   Comparison between two meshes: Case 4: Obstruction (coarse) and
Case 5: Obstruction (coarse, 2). The Case 5: Obstruction (coarse, 2) mesh on the right
misrepresents the real channel bathymetry and does not see the deep channel in
the middle.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   151

Figure 7-18.   WSE profiles from different obstruction cases


along the river channel centerline. The Case 5: Obstruction
(coarse, 2) mesh does not resolve the deep channel in the middle
and, therefore, creates a nonphysical weir in the channel. The
corresponding backwater curve upstream is much higher than it
should be.

to Manning’s n. The joint evaluation of uncertainty and sensitivity is necessary to quantify how
uncertain the result is and how much it will change because of small changes in Manning’s n.
This section shows a demonstration of a sensitivity analysis. The main channel Manning’s n
was selected as the uncertain parameter because it is probably the most important one among all
other land-use types. Previously, the main channel Manning’s n was set at 0.03 based on textbook
tables and the guidelines proposed in this project regarding resolved versus unresolved parts
of Manning’s n. This Manning’s n value performed well when comparing model results with
observed HWMs. However, a modeler may still feel uncertain about this single value of 0.03.
So instead of a single value, a distribution of the main channel Manning’s n is assumed. This
distribution is assumed to follow a truncated normal distribution with the mean value at 0.03
and an STD of 0.005 (17% of the mean). The truncation makes the Manning’s n within the range
of 0.02 and 0.04 (±2 STDs). Random samples of Manning’s n values can be drawn from this dis-
tribution and run in SRH-2D with these values.
To evaluate the results, four monitor points were located along the channel to extract WSE
results for statistical analysis. Their locations are shown in Figure 7-19a. The numbering of the
four points from 1 to 4 is in the upstream direction. Point 1 is the most downstream point and
close to the outlet. This point should be the least sensitive to Manning’s n because it is closest to
the outlet boundary, which has a fixed water surface. Point 4 is the most upstream point close to
the Iowa River Dam.
For demonstration purpose, 100 Manning’s n values were drawn randomly from the distri-
bution; the histogram is shown in Figure 7-19b. The histogram shows that the distribution is
roughly centered at 0.03 and bounded by 0.02 and 0.04. If more values were drawn from the
distribution, the histogram would be smoother.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

152   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-19.   Sensitivity of 2D results to main channel Manning’s n: (a) locations of


four probing points and (b) histogram of sampled main channel Manning’s n values.

The 100 simulations were performed in parallel with eight processors on Amazon Web Services
cloud. The computer instance used was c5.xlarge. A Python script was written for parallel simula-
tions. Each case took about 1 to 2 h, and all 100 simulations took about 20 h to finish.
The simulated WSE at the four probing locations was used to calculate the exceedance of
probability. Figure 7-20 shows the curves for their respective exceedance of probability, and
Figure 7-21 lists the WSE values corresponding to 1%, 10%, 50%, 90%, and 99% exceedance
probability. The range of WSE between 1% and 99% exceedance probabilities is also reported.
It is observed that WSE at the four different locations changes within a range when the main
channel Manning’s n changes in the uncertainty distribution. These ranges and probability of
exceedance can be used by a modeler to assess the uncertainties and flood risks.
The range of WSE corresponding to exceedance probability between 1% and 99% is one indi-
cator to evaluate the sensitivity of the 2D modeling result. Apart from Point 2, the range increases
as the probing points are further upstream. This is as expected because flow resistance takes
distance to show its effect upstream for a subcritical flow. Point 1 is too close to the outlet, and its
WSE is more influenced by the fixed-WSE boundary condition at the outlet. Point 2 is an excep-
tion because it is at a location where there are several bridges. The bridge piers and the pressure
structures there are probably the primary hydraulic controls at this location, and Manning’s n is
a secondary effect.

7.2 Susitna River Case


7.2.1 Background Information and Data
The dataset for this case is available from the project database of the Susitna-Watana Hydro-
electric Project by the State of Alaska. The data are part of the Fluvial Geomorphology Model-
ing Studies below Watana Dam in Alaska. These studies were conducted to assess the potential
effects of the Susitna-Watana Hydroelectric Project on the dynamic behavior of the river to
understand how this project affects the in-stream changes and riparian habitat. This area is in the
Middle River reach of the Susitna River between the mouth of Skull Creek at project river mile
(PRM) 128.1 to PRM 130.5. The modeled area is approximately 15,600 ft long in the streamwise
direction with no bridge or any other structure. This reach contains an important salmon habitat

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines 153

Figure 7-20. WSE’s exceedance of probability curves at the four probing locations.

that includes several side channels. Figure 7-22a shows an overview of the Susitna River model-
ing domain. In this figure, the lines used for long and cross-river profiles are also shown.
The terrain data used for the model development are a product of two different surveys. First,
the bathymetric and topographic surveys were used to develop a triangulated irregular network
(TIN) of the channels. Then, 2011 LiDAR data were added to the TIN. These LiDAR data

Figure 7-21. WSE’s exceedance of probability and


range of WSE at the four probing locations.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

154 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-22. Overview of the Susitna River: (a) modeling domain and satellite image and
(b) terrain (resolution of 5 to 10 ft for small channels and up to 70 ft in floodplain). The flow
is from right to left. Profile lines are also shown.

include approximately 2.5 million triangles, most of which are in the overbank areas. The terrain
data’s horizontal datum is the North American Datum of 1983 (Alaska, Zone 4), and the vertical
datum is the North American Vertical Datum of 1988. Figure 7-22b shows the used terrain for
modeling (TetraTech 2014).
To better appreciate the terrain and bathymetry, which is directly connected to the roughness
and flow resistance, Figure 7-23 (parts b and c), shows the 2D and 3D bathymetry between two
cross sections shown in Figure 7-22. To help compare the terrain surface with ground truth, the
areal image is shown in Figure 7-23a.
Overall, the terrain surface is smooth, especially within the channel itself. For this case,
the terrain data resolution for small channels and the main channel are around 5–10 ft, but
can increase up to 70 ft in a floodplain. This means that any roughness element smaller than
these values in each area will be lost. This resolution is not sufficient to capture all details in the
Susitna River. This is clear by comparing the smooth bathymetry with the areal image shown in
Figure 7-23a (though the timing of survey and aerial photo may not exactly match). Therefore,
the terrain data can only resolve the Manning’s n caused by roughness elements larger than
5–10 ft within the channel. The satellite shows dense trees and vegetation, for which flow resis-
tance effect must be parameterized with Manning’s n. The terrain data do a good job in capturing
important ground controls (e.g., small streams). However, the terrain data seem to miss a small
channel cutting through a bar (highlighted by a dashed circle in the figure). This may be due to
the different times of the image and the terrain data.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   155

Figure 7-23.   Comparison between (a) a satellite image of the terrain and
(b and c) terrain data to show the difference in terms of how much ground
roughness is represented.

The model has been calibrated based on the WSEs and velocity (magnitude and direction) mea-
surements that were collected on September 10, 2013. The discharge in the river was 26,124 ft3/s.
The calibration is not repeated in this report. The model was validated using WSEs surveyed for
a range of flows up to approximately 50,000 ft3/s. More details can be found in TetraTech (2014).

7.2.2 Demonstration of the Proposed Guidelines


This section demonstrates how to use the proposed guidelines—the step-by-step decision-
making process—to determine roughness factors and their corresponding Manning’s n.

1. Determine the purpose of the modeling project.


The purpose of this project is to explore the effect of river hydraulics on aquatic and riparian
habitats. Therefore, in contrast to bridge hydraulic studies, it is not necessary to have a detailed
local velocity distribution around any specific feature. However, researchers do need to have
high quality in predicted water depth and velocities to understand how hydraulic parameters
may affect the habitats. In the modeled area, there is no bridge or any other structures. How-
ever, there are many small side channels and branches that should be resolved in the mesh
to produce good results. Meshes with different resolutions were used to see how it could
affect the results. Other than surface roughness and vegetation, there is no other significant
hydraulic roughness feature in the domain.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

156   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

2. Select a model based on its ability to provide output to support the purpose determined in the
previous step, strength and weakness, availability, cost, modeler’s familiarity and experience,
and prior usage for similar purposes.
In this example, SRH-2D, HEC-RAS 2D, and RiverFlow2D were compared for modeling
the Susitna River. For each model, three different mesh resolutions (20, 30, and 60 ft) were
tested.
3. Make decisions on how to treat different flow resistance factors and determine the total
Manning’s n value accordingly.
Which flow resistance factors exist in each coverage area (as determined by the modeler)?
From the terrain data, satellite images, site reconnaissance, and other information sources,
the following flow resistance factors are present in the modeling domain: surface rough-
ness, vegetation zones, channel irregularity, channel alignment, sedimentation and scour,
obstructions, and ground controls.
The resolution and quality of terrain data and mesh were also considered. The terrain
data’s best resolution for small channels is 5 ft, and the mesh resolution is coarser than the
terrain surface (described in the next section). Therefore, any roughness factor or feature
smaller than the terrain resolution will not be fully captured. On top of that, coarse mesh
resolutions will further filter out more roughness features. Based on these considerations and
the meshes used, the following can be stated:
(a) Surface roughness: This refers to relatively smaller roughness elements like sediment
grains, ripples, and other small bed features. These features are typically much smaller
than 5 ft; therefore, they are unresolved by the terrain data, let alone the meshes. The
Manning’s n for the unresolved surface roughness factor in 2D models should be equal
to the total Manning’s n. For the Susitna River, which is a sand, gravel, and cobble-bed
river, Manning’s n value of about 0.035 for the main channel would be reasonable. The
detailed calibration yielded a value of 0.032, which is quite close to what was expected
(TetraTech 2014).
(b) Vegetation: Vegetation and trees are mostly in the floodplains adjacent to the main
channel. Based on the land-use data, there are different areas from lightly vegetated to
areas with thick vegetation and trees. The terrain data and the meshes do not have any
representation of the trees and vegetation. Therefore, the resolved part of vegetation
Manning’s n is zero, and the unresolved part is equal to the total Manning’s n. The total
Manning’s n values for vegetation and forests with different densities can be found in
hydraulics tables. In this case, the values in the range of 0.08 to 0.17 were used in different
vegetation zones.
(c) Channel irregularity: The river channel is not regular, and the channel width and depth
change. The river has many branches in many places. It is expected that a reasonable
mesh in 2D models captures these changes. Therefore, their contribution to Manning’s n
should be fully resolved, and no additional consideration is needed in 2D models.
(d) Channel alignment: Although some of the branches are not straight and aligned through-
out the domain, the main channel is straight and has only some small to moderate changes
in the alignment. Therefore, the river channel’s Manning’s n was not specifically modified.
Furthermore, channel alignment has a relatively minor effect on flow resistance in com-
parison with other factors.
(e) Sedimentation and scour: No sediment and scour information is available during the
flood. Therefore, they are not considered.
(f) Obstructions: There are no obstructions in the domain.
(g) Hydraulic controls: There are some hydraulic controls, for example, branches and
in-channel bars, especially as there are many small channels. Meshes generated in
this work can capture these features well through local refinement and breaklines.
If these small branches are not captured well by terrain and mesh, the conveyance is

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines 157

not modeled correctly, and the Manning’s n value will be wrong. In the results section,
using three different mesh resolutions, it will be shown how a bad mesh can lead to
significant error.
In summary, six different roughness zones were used to represent the channel, islands, and
various overbank surfaces and vegetation zones. These roughness zones and their corresponding
Manning’s n values are shown in Figure 7-24.

7.2.3 Case Setup


This section describes the setup of cases, such as boundary conditions and mesh generations
for SRH-2D, RiverFlow2D, and HEC-RAS 2D. The upstream boundary condition is a constant
discharge. Two different discharges of 25,000 and 100,000 cubic feet per second (cfs) were used,
which represent low- and high-flow conditions. The downstream boundary condition was set
as a fixed WSE, which has a value of 564 ft and 570 ft for the low- and high-flow conditions,
respectively.
The meshes for the three 2D models were made comparable. At first, the mesh was generated
for SRH-2D using SMS. The average cell size for main channels and most of the region is 20 ft.
A smaller cell size of 10 ft is used for extra small channels. For each channel, breaklines were used
on the two banks to ensure it is captured by the mesh.
The mesh generation coverage for SRH-2D was then used for RiverFlow2D to create a com-
parable mesh for both models. Since RiverFlow2D uses the generic model feature of the SMS,
the material layer and its polygon should be defined inside the mesh generation coverage and
cannot be created separately. This led to more arcs and polygons inside the mesh generator for
RiverFlow2D. Figure 7-25 shows the mesh generator coverage of RiverFlow2D overlayed on the
one for SRH-2D in the region between the two cross-sectional profile lines.

Figure 7-24. Roughness zones and their Manning’s n values.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

158   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-25.   Mesh generation for SRH-2D and RiverFlow2D.


The black lines are for SRH-2D, and the blue lines are additional
polygons for the RiverFlow2D mesh generator to define the
material layer.

The same breaklines and domain were used for creating a mesh in HEC-RAS 2D. Since
HEC-RAS 2D does not use triangular mesh, the element size of 15 ft was considered
for HEC-RAS 2D to generate a comparable mesh with SRH-2D and RiverFlow2D. The generated
mesh for SRH-2D was later optimized using available tools in SMS to make it more like prac-
tical applications. These generated meshes have an average cell size of 20 ft for SRH-2D and
RiverFlow2D and 15 ft for HEC-RAS 2D. Two different discharges of 25,000 cfs and 100,000 cfs
were simulated. Table 7-5 shows the number of cells for generated meshes for each of the models.
One important aspect in this case is how to properly model the small channels. In other words,
how will the results be affected if the mesh cannot represent these small channels correctly? For
this purpose, two mesh resolutions of 30 ft and 60 ft were considered for whole domain and all
three models. Figure 7-26 compares the generated meshes for an area around the second cross-
sectional profile for all three meshes. From the figure, the mesh resolutions of 30 ft and 60 ft do
not capture the small river branches correctly.

7.2.4 Modeling Results and Analysis


7.2.4.1 Comparing Different Models
As explained previously, the average mesh resolution of 20 ft was used for comparing the
models in two discharges of 25,000 cfs and 100,000 cfs. Figure 7-27 shows the simulated WSE

Table 7-5.   Mesh resolutions and number


of mesh cells for the Susitna River case.
Average Mesh Size SRH 2D RiverFlow2D HEC-RAS 2D
20 ft 197,664 293,659 183,164
30 ft 109,046 109,771 45,557
60 ft 27,560 27,850 13,412

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   159

Figure 7-26.   Meshes for the Susitna River near the


second cross-sectional profile. The zoom-in is around
the small river branches.

Figure 7-27.   Simulated WSE by HEC-RAS 2D, SRH-2D, and RiverFlow2D for inflow
of 25,000 cfs.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

160 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

contours by HEC-RAS 2D, SRH-2D, and RiverFlow2D for the discharge of 25,000 cfs. With this
discharge, the flow is only in the main channel and connected small streams. While the general
trend of the simulated WSEs from different models are similar, there are some differences, espe-
cially in small channels. Based on this figure, the results of HEC-RAS 2D and SRH-2D seem to be
similar. Because RiverFlow2D does not model turbulence, its simulated WSE is lower than those
from the other two models and is consistently lower than the observed WSE used for calibration.
Similar WSE contours are created for the discharge of 100,000 cfs and are shown in Fig-
ure 7-28. Again, RiverFlow2D produces a lower WSE in comparison with HEC-RAS 2D and
SRH-2D, which leads to some dry area in RiverFlow2D’s result. The whole domain is inundated
in the results of SRH-2D and HEC-RAS 2D.
To have a more detailed view and comparison of the results, the long and cross-sectional
profiles of WSE are analyzed. The profile lines are shown in Figure 7-22. The long profiles for
all models for the discharges of 25,000 and 100,000 cfs are shown in Figures 7-29 and 7-30,
respectively. Moreover, the simulated WSE using SRH-2D is subtracted from the ones simulated
using HEC-RAS 2D and RiverFlow2D to calculate their differences. These results are shown,
too. These figures show that the difference between SRH-2D and HEC-RAS 2D for both flows
mostly changes in the range of −0.1 ft to 0.2 ft. The difference between SRH-2D and RiverFlow2D
is larger, and it reaches a maximum of −2.5 ft for the 100,000 cfs case.

Figure 7-28. Simulated WSE by HEC-RAS 2D, SRH-2D, and RiverFlow2D for inflow of 100,000 cfs.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines 161

Figure 7-29. Simulated long profile by different models for the discharge of 25,000 cfs (left)
and the difference of other models in comparison to SRH-2D (right).

This is also clear from cross-sectional profiles of WSE for both discharges (see Figures 7-31
and 7-32). When the discharges increase, the difference between RiverFlow2D and the other
two models increases. This is because turbulence becomes more important when the discharge
increases. For cross section 1 (see Figure 7-31), while the whole cross section is flooded based
on the results of SRH-2D and HEC-RAS 2D, there are two areas that are dry based on the result
of RiverFlow2D.
Therefore, a higher Manning’s n is needed to compensate the missing turbulence in RiverFlow2D
to produce comparable results with SRH-2D and HEC-RAS 2D.

7.2.4.2 Effect of Mesh Resolution


The effect of mesh resolution on the results of each model is discussed in this section. Only the
high-flow discharge of 100,000 cfs is considered.
Figures 7-33, 7-34, and 7-35 show the long profiles of SRH-2D, HEC-RAS 2D, and RiverFlow2D
for three mesh resolutions of 20 ft, 30 ft, and 60 ft. Furthermore, the simulated WSEs based on
20-ft mesh are subtracted from the ones simulated with 30-ft and 60-ft meshes to calculate their

Figure 7-30. Simulated long profile by different models for the discharge of 100,000 cfs
(left) and the difference of other models in comparison to SRH-2D (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

162 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-31. Simulated cross-sectional profile 1 by different models for the discharge
of 25,000 cfs (left) and 100,000 cfs (right).

Figure 7-32. Simulated cross-sectional profile 2 by different models for the discharge
of 25,000 cfs (left) and 100,000 cfs (right).

Figure 7-33. Effect of mesh resolution on long profile simulated by SRH-2D for the
discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison with 20 ft
mesh resolution (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines 163

Figure 7-34. Effect of mesh resolution on long profile simulated by HEC-RAS 2D for
the discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison
with 20-ft mesh resolution (right).

differences. For all models, the difference between 20-ft and 30-ft mesh resolutions is relatively
small and reasonable. In other words, there is some degree of mesh convergence between 20-ft
and 30-ft resolution. However, for the cell size of 60 ft, the difference is much higher. The same
observations can be made in the cross-sectional profiles, which are plotted in Figures 7-36, 7-37,
and 7-38 for cross section 1. Similar plots for cross section 2 are in Figures 7-39, 7-40, and 7-41.
When comparing the cross-sectional profiles from SRH-2D and HEC-RAS 2D, it seems they
have comparable mesh resolution sensitivity. For the 60-ft mesh resolution, the differences in WSE
in comparison with the refined 20-ft resolution are in the range of 0 ft to about 0.6 ft for both
models. The advantage of subgrid terrain treatment in HEC-RAS 2D does not show in this case.
Comparing all profiles and cross sections in Figures 7-35 to 7-41, it can be observed that when
the cell size increases, the WSE is also generally increasing. Therefore, a lower Manning’s n value
is needed for the coarse mesh so that its result will match the result of the refined mesh. For this
case, which may also apply to other real-world cases, the coarse mesh missed some of the small
channels and reduced their contribution to the conveyance. This point will be discussed in the
next section.

Figure 7-35. Effect of mesh resolution on long profile simulated by RiverFlow2D for the
discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison with 20-ft
mesh resolution (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

164 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-36. Effect of mesh resolution on cross-sectional profile 1 simulated by SRH-2D


for the discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison
with 20-ft mesh resolution (right).

Figure 7-37. Effect of mesh resolution on cross-sectional profile 1 simulated by HEC-RAS


2D for the discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison
with 20-ft mesh resolution (right).

Figure 7-38. Effect of mesh resolution on cross-sectional profile 1 simulated by RiverFlow2D


for the discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison
with 20-ft mesh resolution (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines 165

Figure 7-39. Effect of mesh resolution on cross-sectional profile 2 simulated by SRH-2D for
the discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison with
20-ft mesh resolution (right).

Figure 7-40. Effect of mesh resolution on cross-sectional profile 2 simulated by HEC-RAS 2D


for the discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison
with 20-ft mesh resolution (right).

Figure 7-41. Effect of mesh resolution on cross-sectional profile 2 simulated by RiverFlow2D


for the discharge of 100,000 cfs (left) and the differences of coarse meshes in comparison
with 20-ft mesh resolution (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

166 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

7.2.5 Importance of Resolving Key Features


As discussed, there are many small channels connected to the main river. These small side
channels are key features for this case. They must be properly captured by 2D models. To do that,
the terrain data and the mesh should at least geometrically resolve these small channels.
To investigate this, the long profiles of the bed with different meshes are shown in Figure 7-42.
The average cell sizes are 20 ft, 30 ft, and 60 ft for the three meshes. The profiles show that the
channel-bed elevation tends to be higher than the terrain in the deep portion of the river when
mesh cell size increases. Higher channel-bed elevation will result in higher WSE. During calibra-
tion, this translates to a lower Manning’s n with coarse mesh for this case.
The cross-sectional profiles of the channel bed with different meshes are shown in Figures 7-43
and 7-44. It is evident that the deep parts of the channels are misrepresented by coarse meshes.
Specifically, the channels are shallower with coarser meshes. On the other hand, the higher
ground in the floodplain gets lower with coarser meshes. This is because of the smoothing effect
of coarse mesh (missing peaks and troughs). For the case of Susitna River, the small channels are
important for the overall conveyance. The shallow channels in the coarse meshes reduce the con-
veyance; therefore, the simulated WSE is higher. The lowering of high ground in the floodplain
with coarse mesh will reduce WSE. However, for the Susitna River case, this effect is of secondary
importance because most of the conveyance is in the river channels.
To further show how mesh resolution affects the representation of terrain in 2D models, the
differences between terrain data and meshes are calculated for the area between the two cross
sections, and the contours are shown in Figure 7-45. The difference is defined as mesh elevation
minus terrain data. For the mesh resolutions of 20 ft and 30 ft, there is only some small eleva-
tion increase or decrease in the whole domain (between 0 to 1 ft). For the 60-ft mesh, there are
significant bed elevation differences, especially in small channels where 2 to 3 ft of difference is
pervasive. This again shows the importance of resolving key hydraulic features in the domain,
which are the small side channels in this case. At least several cells are needed to cover the width
of a small channel, and breaklines are strongly recommended along its two banks.

Figure 7-42. Effect of mesh resolution on the long profile of channel bed.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines 167

Figure 7-43. Channel-bed profiles with different meshes at cross section 1.

Figure 7-44. Channel-bed profiles with different meshes at cross section 2.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

168 Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-45. The calculated bed elevation differences between terrain data and meshes
with different resolutions. The difference is defined as mesh elevation minus terrain data.

7.3 Delaware Bridge 1-242 Case


7.3.1 Background Information and Data
The dataset for this case was provided by Scott Steele at Pennoni Associates, Inc., and contains
data and SRH-2D models.
This case was built for culvert pipe rehabilitation of bridges 1-242, 1-363, and 1-406 in Delaware.
Here, only bridge 1-242 is used for demonstration purposes. The case is simplified and modified
from its original version to highlight Manning’s n part. The bridge is on the crossing of Red Mill
Road over the White Clay Creek near Newark, Delaware. The White Clay Creek is about 18.5 mi
long, has a drainage area of about 107 square miles, and drains into the Christina River. The sedi-
ment in the White Clay Creek is mainly sand with some silt and clay (hence the creek’s name).
The modeling domain in this case is about 1,300 ft upstream and 1,000 ft downstream of the
bridge. The modeled floodplain is also about 1,000 ft. Figure 7-46a shows the satellite image and
the boundary of the modeled domain. Flow is from left to right. The DEM used is shown in Fig-
ure 7-46b, which shows the main channel (White Clay Creek Channel), the small side channel
(Mill Race Channel), and the Red Mill Road as prominent terrain features. The average width of
the main channel and the small side channel is about 80 ft and 30 ft, respectively. The research
team does not have access to the original DEM data. Instead, the DEM shown here is derived
from a mesh provided to the research team. Thus, the resolution of the DEM is the same as the
mesh resolution and is not uniform.
To better view the terrain and bathymetry, which is directly connected to the roughness and flow
resistance, Figure 7-47 shows the 3D zoom-in view around the road crossing. Overall, the terrain

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   169

Figure 7-46.   Overview of the White Clay River and the 1-242 bridge crossing: (a) modeling domain (shown
within white boundary) and satellite image and (b) terrain data (2-ft resolution). The flow is from left to right.

data, which have an especially coarse resolution, are smooth in the whole domain. The terrain data
have been resampled by the mesh provided and do not contain small-scale roughness factors, such
as surface roughness. This low resolution may occur in practice if high-resolution terrain data are
not available or if they are converted from 1D cross-sectional profiles. In this case, it is safe to say
that the terrain data capture only the bulk of the channel and the floodplain. No other roughness
factors are resolved. For example, a significant portion of the floodplain is dense woods. How-
ever, the terrain data are the bare earth, which is standard practice. In other words, the resolved
part of Manning’s n due to the trees and vegetation is zero.

7.3.2 Demonstration of the Proposed Guidelines


This section will demonstrate how to use the proposed guidelines (i.e., the step-by-step decision-
making process) to determine roughness factors and their corresponding Manning’s n.
• Determine the purpose of the modeling project.
The purpose of this project is to study bridge and culvert hydraulics. Therefore, the bridge
piers can be modeled with the hole-in-mesh approach. The alternative approach of obstruc-
tion produces less accurate local hydrodynamics. The 1-242 bridge becomes pressurized
during the flood. Therefore, a pressure structure boundary condition was also used on the
upstream and downstream of the bridge. There is also a culvert on the road crossing over the
smaller side channel. It is modeled with the HY-8 software (coupled with the SRH-2D model).
There are no other significant hydraulic roughness features in the domain. However, there
are ground controls and hydraulic features such as the high ground of the road, the channels,
and the culvert. These features greatly impact the river hydraulics and need to be accurately
represented in the model.
• Select a model based on its ability to provide output to support the purpose determined in the
previous step, strength and weakness, availability, cost, modeler’s familiarity and experience,
and prior usage for similar purposes.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

170   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-47.   Terrain and bathymetry of the modeling domain for the White Clay River and the 1-242 bridge
crossing case. Note the holes in the mesh to represent the blockages from buildings.

The research team selected SRH-2D because of familiarity with its use. Other models may
also be used; however, some of them may not have all the options for hydraulic structures. For
example, HEC-RAS 2D cannot model bridge piers with hole-in-mesh. Also, there are several
buildings in the floodway, which were modeled with hole-in-mesh.
• Make decisions on how to treat different flow resistance factors and determine the total
Manning’s n value accordingly.
– Which flow resistance factors exist in each coverage area (as determined by the modeler)?
List all of them.
From the terrain data, satellite images, land use, and other information sources, the follow-
ing flow resistance factors are present in the modeling domain: surface roughness, vegeta-
tion (mostly in the floodplains), channel irregularity, channel meandering, obstructions, and
ground controls. The land use of the modeled area is shown in Figure 7-48.
Based on the resolution and quality of terrain data and mesh (described in the next
section), the following can be stated:
▪ Surface roughness: This term refers to relatively smaller roughness features, like sedi-
ment grains, ripples, and other small bed features. They are unresolved by the terrain

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Figure 7-48.   Land-use data for the White Clay River and the 1-242 bridge crossing case.
Copyright National Academy of Sciences. All rights reserved.
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

172   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

data and mesh; this is almost universally the case. The Manning’s n for the unresolved
surface roughness factor in 2D models should be equal to the total Manning’s n. The
White Clay Creek is a sand-bed river. From the satellite image where some of the
rivers are exposed during low flow, the riverbed is especially rough with some debris.
The channel banks are lined with heavy vegetation and trees. Thus, an n value of 0.04
is used for the main channel throughout the domain. For the smaller side channel (i.e.,
the Mill Race Channel) the roughness is slightly smaller, and a value of 0.035 is used.
Because the Mill Race Channel is extremely small and short, its n value is probably less
important than the main channel. There is also a road in the domain, whose n value is
set to 0.02 because the road is relatively smooth. Like the small side channel, the road is
of secondary importance because its area is small within the whole domain.
▪ Vegetation: Vegetation and trees are mostly in the floodplains adjacent to the main
channel. Based on the land-use data, the main vegetated area, labeled “Woods,” is where
dense trees and vegetation can be observed from the satellite image. Manning’s n is set
as 0.18 in this case. The Woods land use covers more than half of the model domain, and
it is right in the main floodway. Thus, Manning’s n value is of the highest importance
in this case, probably as or more important than the value of the main channel. The
Manning’s n value of 0.18 for the Woods is the best estimate from open-channel text-
books and hydraulics manuals. However, there is significant uncertainty. To eval­
uate this uncertainty and see how sensitive the result is to Manning’s n of the Woods,
a sensitivity analysis will be demonstrated.
There is another land-use type labeled “Agriculture,” which is in the upstream inlet
area. This land use has a smaller area and is away from the main channel. Thus, it is
expected that its Manning’s n value is of secondary importance. From the satellite image,
it is well-maintained agricultural land, and a value of 0.04 is used.
▪ Channel irregularity: The river channel appears to be relatively regular, and the channel
width and depth remain uniform. Even if the channel shows significant irregularity, the
mesh in 2D models should capture these changes when the terrain data include the
irregularities and the mesh resolution is detailed enough. Therefore, it is assumed that
this contribution to Manning’s n has been fully resolved, and no additional consider-
ation is needed in this 2D model.
▪ Channel alignment: Like channel irregularity, the channel is not straight and aligned
throughout the domain. However, the river channel Manning’s n was not specifically
modified because the 2D model incorporates much of the resistance from redirecting
the flow down the meandering channel.
▪ Sedimentation and scour: No sediment and scour information is available during the
flood. Therefore, they are not considered.
▪ Obstructions: There is one bridge (1-242) and a culvert in the domain. Because the pur-
pose of the study is to examine bridge hydraulics at the road crossing, the bridge piers
are modeled with the hole-in-mesh approach; therefore, there is no need to specify
Manning’s n or a drag coefficient. The culvert (the road crossing with the Mill Race
Channel) is modeled with HY-8 (coupled with SRH-2D). The specification of the culvert
is provided in the case setup. Among all the culvert parameters, Manning’s n for the
culvert is of relevance to this project. The culvert is made of riveted corrugated metal,
and a typical Manning’s n value of 0.033 is set for this material. Because the culvert
is modeled within HY-8, this Manning’s n should be the total Manning’s n found in
textbooks and hydraulics manuals.
▪ Hydraulic controls: There are some hydraulic controls such as road, buildings,
in-channel features, and floodplain terrain variability. Meshes generated in this work
can capture these features sufficiently well through local refinement and breaklines.
Manning’s n for the road and channels has been described previously. For buildings,

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   173

there are some buildings just downstream of the road crossing right in the middle of
the floodway. These buildings partially block the flow, and their sizes are relatively large.
Hydraulically, they are important for the overall hydrodynamics. Thus, they are modeled
with the hole-in-mesh approach, and there is no need to specify any of Manning’s n
value. There are also houses in the domain. Within the bound of the “Subdivision” land
use, there are many houses located at the edge of the floodplain. Data show that most
of the houses are not flooded during the simulated flow. Thus, Manning’s n is probably
irrelevant. Anyway, these houses are modeled with a lumped Manning’s n value of 0.08,
which is appropriate for low to moderate density development. Because the floodplain
topography is well represented (i.e., resolved), the primary contributor to Manning’s n
is floodplain vegetation.
The total Manning’s n values for all land-use types, which are also the values used in SRH-2D
because the resolved part for all land use is extra small, are listed in Figure 7-48. Only the
Manning’s n values for areas engaged in the flood water pathway are relevant. The more an area
is engaged with flow, the more important the area’s Manning’s n is. And, the closer the area to
the main channel, the more important it is. By comparing the land use in Figure 7-48 and the
satellite image in Figure 7-46, it is easy to see that land-use types “Woods,” “White Clay Creek
Channel,” “Mill Race Channel,” and perhaps also “Road” have the greatest influence on the 2D
modeling results. In contrast, the areas labeled as “Subdivision” and “Agriculture” in Figure 7-48
are less critical.
In summary, the guidelines proposed in this project were applied and demonstrated for this
case. For most land use and roughness factors, the resolved part of Manning’s n is almost zero
because of the coarse DEM resolution. Therefore, the unresolved Manning’s n, i.e., the value to be
specified in 2D models, should use approximately the total Manning’s n for each land-use area.
This might be common in practice where no high-resolution DEM data are available or the DEM
is converted from a coarse 1D model.

7.3.3 Cases Setup


This section describes the setup of cases, such as boundary conditions and mesh generation.
The 100-year flood was simulated. For boundary conditions, there is one fixed-discharge inlet
and one fixed-WSE outlet on the main channel. The SRH-2D boundary conditions are listed in
Table 7-6. The 1-242 bridge should be pressurized during the simulated 100-year flood. There-
fore, the pressure structure boundary condition is assigned accordingly.
Only one mesh was made for this case. The bridge piers and buildings were represented with
the hole-in-mesh approach. The main channel, the small side channel, and the road are mainly
covered with quadrilateral cells generated with the patch mesh generation method. The rest of
the domain is covered with triangles generated with the paving mesh generation method.
The details of the mesh, such as numbers of nodes and elements, approximate cell sizes, etc.,
are listed in Table 7-7. The mesh is also shown in Figure 7-49, which shows the overview of the
mesh and the zoom-in view around the road crossing.

Table 7-6.   Boundary conditions for the Iowa


River case.
Boundary Type BC Value (cfs or ft)
Main channel in�low INLET-Q 15,900
Culvert Culvert HY-8 See case for details.
Downstream boundary EXIT-H 40.0

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

174   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Table 7-7.   Mesh for the Delaware 1-242 bridge case.


Avg. Near-
Avg. Channel Cell Structure Cell Size
No. of Nodes No. of Cells Size (ft) (ft)
10,657 16,693 7 by 13 4 by 5

The cases were run for 2 h to reach a steady state. The average computing time on a desktop
computer with an Intel i7-8700 CPU is about 0.5 h.

7.3.4 Modeled Flood Inundation


No calibration data were available to the research team, so no calibration was performed. The
Manning’s n values used here were best-educated guesses. No mesh convergence study was per-
formed either because the DEM data are especially coarse and further refinement of the mesh
does not help resolve anything.
For the simulated 100-year flood, the modeled flood inundation is plotted in Figure 7-50,
which shows that the whole floodplain is inundated and the road at the river crossing is flooded.
It also shows that the subdivision within the model domain is largely not flooded, which again
demonstrates that its Manning’s n is not relevant for this flow condition but could be for a higher
flow (e.g., 500-year flood).

7.3.5 Uncertainty and Sensitivity Analysis


Without calibration data, it is impossible to directly quantify the accuracy of the Manning’s n
values used in this case. An alternate and indirect method is to evaluate the uncertainty and
sensitivity of the result to Manning’s n.
There are substantial uncertainties in the Manning’s n values, especially the value for the Woods
in this case because it covers most of the modeling domain. A value of 0.18 was used in the previous
section; however, a modeler might be interested in how sensitive the result is to this value.
This section shows a demonstration of such analysis. The floodplain’s Manning’s n was selected as
the uncertain parameter because it is probably the most important one among all other land-use

Figure 7-49.   Mesh for the Delaware 1-242 bridge case: (a) an overview of the mesh
and (b) a zoom-in around the road crossing.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Verification and Demonstration of Proposed Guidelines   175

Figure 7-50.   Overall inundation result for the Delaware 1-242 bridge case.

types. It is assumed to follow a truncated normal distribution with the mean value at 0.18 and an
STD of 0.03 (17% of the mean). The truncation makes Manning’s n to be within the range of 0.12
and 0.24 (±2 STD); random samples can be drawn of Manning’s n values from this distribution
and run with SRH-2D with these values.
To evaluate the results, seven monitor points were located along the channel to extract WSE
results for statistical analysis. Their locations are shown in Figure 7-51. The numbering of the six
points from 1 to 6 is in the downstream direction. Point 1 is the most upstream and close to the
inlet. Point 6 is the most downstream close to the downstream outlet. Point 3 is just upstream of
the 1-242 bridge. The additional point 7 is just upstream of the culvert.
For demonstration purposes, 100 Manning’s n values were drawn from the distribution, and
the histogram is shown in Figure 7-52. The histogram shows that the distribution is roughly
centered at 0.18 and bounded by 0.12 and 0.24.
The 100 simulations were performed in parallel with six processors on a local desktop com-
puter with an Intel i7-8700 CPU. A Python script was written for parallel simulations. It took
about 2 h for all 100 simulations to finish.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

176   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure 7-51.   Locations of the monitoring


points.

The simulated WSE at the seven monitor locations was used to calculate the exceedance of
probability. Figure 7-53 shows the curves for their respective exceedance of probability, and
Table 7-8 lists the WSE values corresponding to 1%, 10%, 50%, 90%, and 99% exceedance prob-
ability. The range of WSE between 1% and 99% exceedance probabilities is also reported.
It is observed that WSE at the seven different monitor locations changes within a relatively
large range when the floodplain Manning’s n changes in the uncertainty distribution. These
ranges and probability of exceedance can be used by a modeler to assess the uncertainties and
flood risks.

Figure 7-52.   Histogram of the 100 Manning’s n values for


the Woods land-use type in the Delaware 1-242 bridge case.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models
Figure 7-53.   WSE’s exceedance of probability curves at the seven monitor locations.
Copyright National Academy of Sciences. All rights reserved.
Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

178   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Table 7-8.   WSE’s exceedance of probability and range of WSE at the seven
monitor locations

WSE (ft)
Exceedance Probability Point 1 Point 2 Point 3 Point 4 Point 5 Point 6 Point 7

The relatively large ranges of WSE at these locations can be explained by two facts. One is that
the “Woods” land use is important for this case. It covers most of the flood path. The second fact
is the range of Manning’s n value from 0.12 to 0.24 is especially large because of the uncertainty
of flow resistance in wooded areas. If there is some calibration data, the range of this Manning’s n
can be better constrained.
The range of WSE corresponding to exceedance probability between 1% and 99% is one indi-
cator to evaluate the sensitivity of 2D modeling results. For all six monitor locations in the main
channel, except Point 3, the range increases as the probing points are further upstream. This is as
expected because flow resistance takes distance to show its effect upstream for a subcritical flow.
Point 3 is an exception because it is near the pressure structure of the bridge. As found in other
cases, the solution near pressure structure can be sensitive to changes in model parameters. This
shows that, in addition to and in conjunction with Manning’s n, structure hydraulics is essential
to the modeling process and needs to have significant attention during model development.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

References

Abdelsalam, M., A. Khattab, A. Khalifa, and M. Bakry. 1992. Flow Capacity Through Wide and Submerged Vegetal
Channels. Journal of Irrigation and Drainage Engineering, 118(5), pp. 724–732.
Aberle, J., and G. Smart. 2003. The Influence of Roughness Structure on Flow Resistance on Steep Slopes. Journal
of Hydraulic Research, 41(3), pp. 259–269.
Aldridge, B. N., and J. M. Garrett. 1973. Roughness Coefficients for Stream Channels in Arizona. Technical Report,
U.S. Geological Survey.
Arcement, G. J., and V. R. Schneider. 1989. Guide for Selecting Manning’s Roughness Coefficients for Natural
Channels and Flood Plains. Technical Report, U.S. Geological Survey Water Supply Paper 2339.
Arneson, L. A., L. W. Zevenbergen, P. F. Lagasse, and P. E. Clopper. 2012. Evaluating Scour at Bridges, 5th edition.
Report No. HEC-18. FHWA, U.S. Department of Transportation.
Bakhmeteff, B. 1932. Hydraulics of Open Channels. McGraw-Hill, New York.
Baptist, M., V. Babovic, J. Rodríguez Uthurburu, M. Keijzer, R. Uittenbogaard, A. Mynett, and A. Verwey. 2007.
On Inducing Equations for Vegetation Resistance. Journal of Hydraulic Research, 45(4), 435–450.
Barnes, H. H. 1967. Roughness Characteristics of Natural Channels. U.S. Geological Survey Water-Supply Paper 1849.
U.S. Government Printing Office, Washington, DC.
Bathurst, J. C. 1985. Flow Resistance Estimation in Mountain Rivers. Journal of Hydraulic Engineering, 111(4),
pp. 625–643.
Brunner, G. 2016. HEC-RAS River Analysis System Hydraulic Reference Manual. Report No. CPD-69. U.S. Army
Corps of Engineers.
Brunner, G. 2018. Benchmarking of the HEC-RAS Two-Dimensional Hydraulic Modeling Capabilities. Report
No. RD-51. U.S. Army Corps of Engineers, Davis, Calif., pp. 1–137.
Brunner, G. 2021. HEC-RAS River Analysis System Hydraulic Reference Manual, Version 6.0. U.S. Army Corps
of Engineers, Hydrologic Engineering Center, CPD-69.
Brunner, G., A. Sanchez, T. Molls, D. Ford, and D. Parr. 2018. HEC-RAS Verification and Validation Tests.
Report RD-52. U.S. Army Corps of Engineers, Institute for Water Resources, Hydrologic Engineering
Center, 609 Second Street, Davis, CA, 95616–4687.
Burke, R. W., and K. D. Stolzenbach. 1983. Free Surface Flow Through Salt Marsh Grass. MIT-Sea Grant Technical
Report. MITSG 83-16, Cambridge, Mass.
Casulli, V. 2009. A High-Resolution Wetting and Drying Algorithm for Free-Surface Hydrodynamics. Inter-
national Journal for Numerical Methods in Fluids, 60(4), pp. 391–408.
Charbeneau, R. J., and E. R. Holley. 2001. Backwater Effects of Piers in Subcritical Flow. Technical Report,
University of Texas at Austin, Center for Transportation Research.
Chaudhry, M. H. 2008. Open-Channel Flow. Springer Science & Business Media.
Chen, Y., R. A. DiBiase, N. McCarroll, and X. Liu. 2019. Quantifying Flow Resistance in Mountain Streams
Using Computational Fluid Dynamics Modeling over Structure-from-Motion Photogrammetry-derived
Microtopography. Earth Surface Processes and Landforms, 44(10), pp. 1973–1987.
Cheng, N.-S. 2011. Representative Roughness Height of Submerged Vegetation. Water Resources Research, 47(8).
Chow, V. T. 1959. Open-Channel Hydraulics. McGraw-Hill Civil Engineering Series.
Coon, W. F. 1998. Estimation of Roughness Coefficients for Natural Stream Channels with Vegetated Banks,
Vol. 2441. U.S. Geological Survey.
Cowan, W. L. 1956. Estimating Hydraulic Roughness Coefficients. Agricultural Engineering, 37(7), pp. 473–475.
de Saint-Venant, B. 1843. Notea joindre au Mémoire sur la dynamique des fluides. Comptes Rendus, 17(1240).
Deal, E. C., A. D. Parr, and C. B. Young. 2017. A Comparison Study of One- and Two-Dimensional Hydraulic
Models for River Environments. Technical Report. Report No. KS-17-02, University of Kansas.

179

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

180   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Dombroski, D. 2014. A Deterministic Spatially-Distributed Ecohydraulic Model for Improved Riverine System
Management. U.S. Department of the Interior, Bureau of Reclamation, Technical Service Center.
Eckman, J. E. 1990. A Model of Passive Settlement by Planktonic Larvae onto Bottoms of Differing Roughness.
Limnology and Oceanography, 35(4), pp. 887–901.
Einstein, H. 1934. Der hydraulische oder Profil-Radius. Schweizerische Bauzeitung, 103(8), pp. 89–91.
Einstein, H., and R. Banks. 1950. Fluid Resistance of Composite Roughness. Eos, 31(4), Transactions American
Geophysical Union, pp. 603–610.
FHWA. 2023. Hydraulic Design of Safe Bridges, 2nd ed. Hydraulic Design Series No. 7. U.S. Department of
Transportation.
Finnigan, J. 2000. Turbulence in Plant Canopies. Annual Review of Fluid Mechanics, 32(1), 519–571.
Follett, E., I. Schalko, and H. Nepf. 2020. Momentum and Energy Predict the Backwater Rise Generated by a
Large Wood Jam. Geophysical Research Letters, 47(17), e2020GL089346.
Freeman, G., W. Rahmeyer, and R. Copeland. 2002. Development and Application of Methodology for Deter-
mination of Hydraulic Roughness for Vegetated Floodplains. In Hydraulic Measurements and Experimental
Methods 2002, pp. 1–11.
Freeman, G. E., W. H. Rahmeyer, R. R. Copeland. 2000. Determination of Resistance Due to Shrubs and Woody
Vegetation. Technical Report ERDC/CHLTR-00-25. U.S. Army Corp of Engineers.
Friend, A., and M. McBroom. 2018. How to Adjust Manning’s “n” Values for 2D Modeling. Presented at the
Association of Montana Floodplain Managers 19th Annual Conference, Great Falls, Mont., March 5–9.
Garcia, M. 2008. Sedimentation Engineering: Processes, Measurements, Modeling, and Practice. ASCE Manuals
and Reports on Engineering Practice No. 110.
Griffiths, G. A. 1981. Flow Resistance in Coarse Gravel Bed Rivers. Journal of the Hydraulics Division, 107(7),
pp. 899–918.
Henderson, F. 1966. Open-Channel Flow. Macmillan Publishing Co. Inc., New York, pp. 288–324.
Hey, R. D. 1979. Flow Resistance in Gravel-Bed Rivers. Journal of the Hydraulics Division, 105(4), pp. 365–379.
Hicks, D., and P. Mason. 1998. Roughness Characteristics of New Zealand Rivers: A Handbook for Assigning
Hydraulic Roughness Coefficients to River Reaches by the “Visual Comparison” Approach. National Institute
of Water and Atmospheric Research, Christchurch, New Zealand. Distributed by Water Resource Publica-
tions, LLC, Englewood, Colo.
Homan, J., and H. Toniolo. 2018. Developing Guidelines for Two-Dimensional Model Review and Acceptance.
Technical Report, University of Alaska Fairbanks, Center for Environmentally Sustainable Transportation
in Cold Climates.
Horton, R. E. 1933. Separate Roughness Coefficients for Channel Bottom and Sides. Engineering News Record,
111(22), pp. 652–653.
Hunt, J. H. 1995. Flow Transitions in Bridge Backwater Analysis. Research Document No. 42. U.S. Army Corps
of Engineers, Hydrologic Engineering Center.
Jarrett, R. D. 1984. Hydraulics of High-Gradient Streams. Journal of Hydraulic Engineering, 110(11), pp. 1519–1539.
Järvelä, J. 2004. Determination of Flow Resistance Caused by Non-submerged Woody Vegetation. International
Journal of River Basin Management, 2(1), pp. 61–70.
Järvelä, J. 2005. Effect of Submerged Flexible Vegetation on Flow Structure and Resistance. Journal of Hydrology,
307(1-4), pp. 233–241.
Katul, G., P. Wiberg, J. Albertson, and G. Hornberger. 2002. A Mixing Layer Theory for Flow Resistance in
Shallow Streams. Water Resources Research, 38(11), 32-1–32-8.
Keulegan, G. H. 1938. Laws of Turbulent Flow in Open Channels, Vol. 21. National Bureau of Standards,
Gaithersburg, Md.
Kouwen, N., and R.-M. Li. 1980. Biomechanics of Vegetative Channel Linings. Journal of the Hydraulics Division,
106(6), pp. 1085–1103.
Kramer, C. 2005. The Role of Relative Submergence on Cluster Microtopography and Bedload Predictions in
Mountain Streams. MS thesis. University of Iowa, Iowa City.
Lagasse, P. F., L. W. Zevenbergen, W. J. Spitz, and L. A. Arneson. 2012. Stream Stability at Highway Structures,
4th edition. Report No. HEC-20. FHWA, U.S. Department of Transportation.
Lai, Y. G. 2008. Srh-2d version 2: Theory and User’s Manual. Sedimentation and River Hydraulics–Two-Dimensional
River Flow Modeling, U.S. Department of Interior, Bureau of Reclamation, November.
Lee, A. J., and R. I. Ferguson. 2002. Velocity and Flow Resistance in Step-Pool Streams. Geomorphology, 46(1-2),
pp. 59–71.
Lim, N., and S. Brandt. 2019. Flood Map Boundary Sensitivity due to Combined Effects of DEM Resolution and
Roughness in Relation to Model Performance. Geomatics, Natural Hazards and Risk, 10(1), pp. 1613–1647.
Limerinos, J. T. 1970. Determination of the Manning Coefficient from Measured Bed Roughness in Natural
Channels. Geological Survey Water-Supply Paper 1898-B, U.S. Geological Society.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

References  181

Liu, X. 2014. Open Channel Hydraulics: From Then to Now and Beyond. In Modern Water Resources Engineer-
ing (L. K. Wang and C. T. Yang, eds.), Humana Press. https://www.springer.com/gp/book/9781627035941
Liu, X., B. Landry, and M. García. 2008. Coupled Two-Dimensional Model for Scour Based on Shallow Water
Equations with Unstructured Mesh. Coast Engineering, 55(10), pp. 800–810.
López, F., and M. H. García. 2001. Mean Flow and Turbulence Structure of Open-Channel Flow Through Non-
emergent Vegetation. Journal of Hydraulic Engineering, 127(5), pp. 392–402.
Luhar, M., and H. M. Nepf. 2013. From the Blade Scale to the Reach Scale: A Characterization of Aquatic Vegetative
Drag. Advances in Water Resources, 51, pp. 305–316.
Luhar, M., J. Rominger, and H. Nepf. 2008. Interaction Between Flow, Transport and Vegetation Spatial Structure.
Environmental Fluid Mechanics, 8(5), pp. 423–439.
Manning, R. 1891. On the Flow of Water in Open Channels and Pipes. Transactions of the Institute of Civil
Engineers of Ireland, Vol. 20, pp. 161–207.
Moody, L. 1944. Friction Factors for Pipe Flow. Transactions of the American Society of Mechanical Engineers, 66,
Princeton, N.J., pp. 671–684.
Morvan, H., D. Knight, N. Wright, X. Tang, and A. Crossley. 2008. The Concept of Roughness in Fluvial Hydraulics
and Its Formulation in 1D, 2D, and 3D Numerical Simulation Models. Journal of Hydraulic Research, 46(2),
pp. 191–208.
Muhlhofer, L. 1933. Rauhigkeitsuntersuchungen in einem stollen mit betonierter sohle und unverkleideten
wanden. Wasserkraft und Wasserwirtschaft, 28(8), pp. 85–88.
Néelz, S., and G. Pender. 2010. Benchmarking of 2D Hydraulic Modelling Packages. U.K. Environmental Agency,
Bristol, 169.
Nepf, H. M. 1999. Drag, Turbulence, and Diffusion in Flow Through Emergent Vegetation. Water Resources
Research, 35(2), pp. 479–489.
Nepf, H. M. 2012. Flow and Transport in Regions with Aquatic Vegetation. Annual Review of Fluid Mechanics,
44, pp. 123–142.
Nikora, N. F., and V. Nikora. 2007. A Viscous Drag Concept for Flow Resistance in Vegetated Channels. In
32nd IAHR World Congress.
Nikuradse, J. 1933. Stromungsgesetze in rauhen rohren. vdi-forschungsheft, 361, 1.
Parker, G., G. Seminara, and L. Solari. 2003. Bed Load at Low Shields Stress on Arbitrarily Sloping Beds: Alter-
native Entrainment Formulation. Water Resources Research, 39(7).
Petryk, S., and G. Bosmajian. 1975. Analysis of Flow Through Vegetation. Journal of the Hydraulics Division,
101(7), pp. 871–884.
Prandtl, L., J. Ackroyd, B. Axcell, and A. Ruban. 1904. On the Motion of Fluids with Very Little Friction. In
Proceedings of III International Mathematical Congress, Heidelberg.
Ree, W. O. and V. J. Palmer. 1949. Flow of Water in Channels Protected by Vegetative Linings, Technical Bulletin
No. 967. U.S. Department of Agriculture.
Rickenmann, D., and A. Recking. 2011. Evaluation of Flow Resistance in Gravel-Bed Rivers Through a Large
Field Data Set. Water Resources Research, 47(7).
Robinson, D., A. Zundel, C. Kramer, R. Nelson, W. deRosset, J. Hunt, S. Hogan, and Y. Lai (2019). Two-Dimensional
Hydraulic Modeling for Highways in the River Environment—Reference Document. Technical Report,
FHWA, U.S. Department of Transportation.
Rodi, W. 1993. Turbulence Models and Their Application in Hydraulics. CRC Press.
Rouse, H. 1965. Critical Analysis of Open-Channel Resistance. Journal of the Hydraulics Division, 91(4), pp. 1–23.
Shields, F. D., K. G. Coulton, and H. Nepf. 2017. Representation of Vegetation in Two-Dimensional Hydro­
dynamic Models. Journal of Hydraulic Engineering, 143(8), 02517002.
Shields Jr, F. D., and C. J. Gippel. 1995. Prediction of Effects of Woody Debris Removal on Flow Resistance.
Journal of Hydraulic Engineering, 121(4), pp. 341–354.
Soong, D. T., C. D. Prater, T. M. Halfar, and L. A. Wobig (2012). Manning’s Roughness Coefficients for Illinois
Streams. U.S. Geological Survey Data Series 668, 14.
Stokes, G. G. 1880. On the Theories of the Internal Friction of Fluids in Motion, and of the Equilibrium and
Motion of Elastic Solids. Transactions of the Cambridge Philosophical Society 8.
Strickler, A. 1923. Beitrage zur Frage der Geschwindigkeits-formel und der Rauhigkeitszahlen fur Strome,Kanale
und geschlossene Leitungen [Some Contributions to the Problems of the Velocity Formula and Rough-
ness Factor for Rivers, Canals, and Closed Conduits]: Bern, Switzerland, Mitteilungen des Eidgenossischer
Amtes fur Wasserwirtschaft, no. 16.
Sturm, T. W. 2001. Open Channel Hydraulics, Vol. 1. McGraw-Hill, New York.
Tanino, Y., and H. M. Nepf. 2008. Laboratory Investigation of Mean Drag in a Random Array of Rigid, Emergent
Cylinders. Journal of Hydraulic Engineering, 134(1), pp. 34–41.
TetraTech. 2014. Updated Fluvial Geomorphology Modeling Approach. Attachment A: FA-128 (Slough 8A)
Hydraulic Modeling for Proof of Concept. Technical Memorandum.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

182   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

van Rijn, L. C. 1984. Sediment Transport, Part I: Bed Load Transport. Journal of Hydraulic Engineering, 110,
pp. 1431–1456.
Whittaker, P., C. A. Wilson, and J. Aberle. 2015. An Improved Cauchy Number Approach for Predicting the Drag
and Reconfiguration of Flexible Vegetation. Advances in Water Resources, 83, pp. 28–35.
Yang, X., J. Qian, and S. Weng. 2019. Determination of Equivalent Roughness of Bridge Piers’ Flow Resistance.
Journal of Hydrologic Engineering, 24(8), 04019024.
Yarnell, D. L. 1934. Bridge Piers as Channel Obstructions. Technical Bulletin No. 442. U.S. Department of
Agriculture.
Yen, B. C. 1992a. Dimensionally Homogeneous Manning’s Formula. Journal of Hydraulic Engineering, 118(9),
pp. 1326–1332.
Yen, B. C. 1992b. Hydraulic Resistance in Open Channels. Channel Flow Resistance-Centennial of Manning’s
Formula. Water Resources Publications, LLC, Colo., pp. 1–135.
Yen, B. C. 2002. Open Channel Flow Resistance. Journal of Hydraulic Engineering, 128(1), pp. 20–39.
Yochum, S. E. 2018. Flow Resistance Coefficient Selection in Natural Channels: A Spreadsheet Tool. Technical
Summary 103. National Stream and Aquatic Ecology Center, United States Forest Service, U.S. Department
of Agriculture.
Yochum, S. E., B. P. Bledsoe, G. C. David, and E. Wohl. 2012. Velocity Prediction in High-Gradient Channels.
Journal of Hydrology, 424, pp. 84–98.
Yochum, S. E., F. Comiti, E. Wohl, G. C. David, and L. Mao. 2014. Photographic Guidance for Selecting Flow
Resistance Coefficients in High-Gradient Channels. General Technical Report RMRS-GTR-323, United
States Forest Service, U.S. Department of Agriculture, pp. 1–91.
Zevenbergen, L., L. Arneson, J. Hunt, and A. C. Miller. 2012. Hydraulic Design of Safe Bridges. Technical Report,
FHWA, U.S. Department of Transportation.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

APPENDIX A

Evaluation of Solvers in HEC-RAS 2D

HEC-RAS 2D can perform 2D flow simulations using the SWE and the diffusion-wave equa-
tions (DWE); DWE is a simplified version of SWE. In HEC-RAS 2D v5.0.7, this translates to
two options of DW or full dynamics to do the hydraulic modeling. However, starting from
HEC-RAS 2D v6.1, three solvers are available: DWE, the original SWE solver with the Eulerian–
Lagrangian method (ELM), and a new SWE solver with the Eulerian method (EM). All these
solvers can include turbulence (except the DW solver) and Coriolis effects. Chapter 2 of the
HEC-RAS Hydraulic Reference Manual describes the governing equations and numerical algo-
rithms of these solvers.

Different solvers in HEC-RAS 2D affect the simulation results and consequently the
Manning’s n values. Therefore, the research team decided to study these three different solvers to
make a more informed decision on which solver to use for this project. SRH-2D and HEC-RAS 2D
were also compared.

DW, ELM, and EM were compared with several generic cases. All methods use the full mass
balance equation. However, the difference is in the momentum equations. DW retains only the
pressure gradient and flow resistance terms; all other terms (unsteady, advection, turbulent dif-
fusion, wind, Coriolis, etc.) are neglected. In ELM and EM, all terms in the momentum equa-
tion are retained. ELM has been available in HEC-RAS v5, which uses a Lagrangian approach
for the material derivative (unsteady plus advection terms). The advantage of this method is
that it relaxes the stability criteria, which limits time-step size (the Courant-Friedrichs-Lewy
condition). The HEC-RAS Hydraulic Reference Manual states that ELM can create excessive
momentum diffusion; however, the results of the research team show the opposite. ELM has
less momentum diffusion than EM, at least for the test cases in this work. EM is a new method
introduced in HEC-RAS v6. It treats unsteady term and advection term separately, which is
designed to have better momentum conservation than ELM.

A.1 Flow Around Groin


The team started by modeling a groin in a straight channel using different methods. The channel
is 10 m long and 1 m wide (typical for lab-scale experiments). The groin is placed along one side
of the channel and has a width of 0.02 m and a length of 0.155 m. Figure A-1 shows the result of
different methods. With a resolution of 0.02 m, DW and ELM ran successfully. However, EM
diverged. The cell-reported problem is at the tip of the weir in the middle of the channel. The reason
is that EM requires a much smaller time-step size. For example, HEC-RAS documentation RD-52
reported a time-step size of 0.01 s. However, the lowest possible time step is 0.1 s in HEC-RAS.

183

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

184   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure A-1.   The results of different HEC-RAS 2D solvers for modeling


groin with high resolution. EM diverged.

It cannot be reduced to anything smaller than that. With a coarse mesh (resolution = 0.05 m),
EM runs with no problem.
From the simulated flow field in Figure A-1, the DW solver cannot capture the flow physics
behind the groin (shear layer and wake). It is expected because DW solver does not include
important momentum equation terms, which are responsible for shear. Based on these results,
the research team decided to run a case with a larger domain and not pursue the DW solver
further.

The larger domain is 100 m long and 10 m wide. The groin is 2 m wide and 2 m long. The team
ran HEC-RAS 2D (with ELM and EM) and SRH-2D with the same mesh. Figure A-2 shows the
results. The ELM result shows some instability and vortex shedding. The results from EM and
SRH-2D are steady, and no instability is observed (Note: the default depth-averaged parabolic
turbulence model was used. If the two-equation k-ε model is used, the shear layer and momentum
exchange behind the groin may be better captured). Momentum diffusion suppresses instability.
Thus, the results show that ELM has less momentum diffusion than EM, which is contrary to
HEC-RAS’s reference manual. In fact, the Lagrangian part of ELM is used to deal with the advec-
tion term to overcome the numerical diffusion if other numerical methods are used.

Table A-1 shows the computing time of different solvers with different time-step sizes. EM
converges only if the time step is less than 0.2 s. ELM and SRH-2D can use a larger time-step
size and less computing time. With the time-step size of 1 s, HEC-RAS 2D ELM uses about half
of the time used by SRH-2D. This may be because HEC-RAS 2D uses multiple cores to solve the

Figure A-2.   The results of different solvers for modeling a groin in a larger
domain.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Evaluation of Solvers in HEC-RAS 2D   185

Table A-1.   Computing time for each solver in


modeling a groin using different time-step size (total
cells: 22,282; min cell 5 0.04 m2; avg cell 5 0.04 m2;
grid size 5 0.2 m).
Solver Time-Step Size (s) Computing time
HEC-RAS DW 1.0 73 s
HEC-RAS ELM 1.0 24 m 25 s
HEC-RAS ELM 0.5 43 m 18 s
HEC-RAS EM 1.0 diverged
HEC-RAS EM 0.5 diverged
HEC-RAS EM 0.2 1 h 36 m 18 s
HEC-RAS EM 0.1 3 h 17 m 20 s
SRH-2D 1.0 49 m

linear-equation system. The future version of SRH-2D may also have parallel computing capabil-
ity, and its computing time will be greatly reduced.

A.2 Comparing Different HEC-RAS 2D Solvers


Using the Muncie Case
The Muncie case was also used to compare HEC-RAS 2D v5, v6.1, and different solver options.
The data for this case come with HEC-RAS tutorials. Its bathymetry data were used to build a
hypothetical 2D case for testing. The ELM methods in HEC-RAS v5 and v6.1 were found to
produce similar results (although HEC-RAS v6 reports some volume accounting error). As seen
in Figures A-3 and A-4, the simulated water depth and velocity for both versions are similar.
In addition, the computing time in v6.1 is less than in v5, probably because of some improved
solution algorithm.

In the next step, EM and ELM in HEC-RAS 2D v6.1 were compared. It is important to note
that EM can be used only with the new “conservative formulation” of the turbulence model (see
Figure A-5 for the error message if the nonconservative formulation is selected).

Another important observation is that in HEC-RAS v6.1, EM with conservative turbulence for-
mulation greatly reduces the volume accounting error. This is probably because of the difference

Figure A-3.   The comparison of simulated water depth by HEC-RAS 2D v5 and v6


using ELM.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

186   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure A-4.   The comparison of simulated velocity by HEC-RAS 2D v5 and v6


using ELM.

between ELM and EM. This volume accounting error is believed to be a false alarm. It may be
because of the way that velocity and flux are calculated at faces.

The results from ELM and EM seem qualitatively comparable as shown in Figure A-6. To have
a quantitative comparison, the differences in water depth simulated by the two solvers are calcu-
lated with pyHMT2D and shown in Figure A-7. There are substantial differences in some areas
where the magnitude of difference reaches about 1.5 ft.
To further investigate the differences between the two solvers, the water depth along the main
channel (longitudinal profile) and two cross sections was extracted and plotted. Figure A-8 shows
the locations of the profiles. The results are shown in Figures A-9 to A-11. From these plots, it is
clear that even within HEC-RAS 2D v6, the choice of solver method (ELM or EM) makes some
difference in simulation results. For this case, the maximum water-depth difference could be up
to 1 ft. This can impact the Manning’s n during calibration if different solvers are used.

Figure A-5.   The message that states EM works only with the new “conservative formulation.”

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Evaluation of Solvers in HEC-RAS 2D   187

Figure A-6.   The comparison of water depth simulated by ELM and EM in HEC-RAS
2D v6.1.

Figure A-7.   The differences in water depth simulated by ELM and EM in HEC-RAS 2D v6.1.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

188   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure A-8.   The location of longitudinal and cross-sectional profiles.

Figure A-9.   Longitudinal bed profiles using ELM and EM (left) and their difference (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Evaluation of Solvers in HEC-RAS 2D   189

Figure A-10.   The first cross-sectional profile using ELM and EM (left) and their difference (right).

Figure A-11.   The second cross-sectional profile using ELM and EM (left) and their difference (right).

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

190   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure A-12.   The comparison of water depth simulated by DW and EM in


HEC-RAS 2D v6.1.

Table A-2.   The computing time for different solvers (total cells: 19,407;
min cell 5 1,080 ft2; avg cell 5 2,464 ft2; background grid 5 50 ft 3 50 ft).
Solver Time-step size (s) Computing time
HEC-RAS DW 10 54 s
HEC-RAS ELM+nonconservative turb model 10 1 m 14 s
HEC-RAS ELM+conservative turb model 10 1 m 14 s
HEC-RAS EM 2 diverged
HEC-RAS EM 1 6 m 26 s
SRH-2D 10 1 m 14 s

Previously, it has been concluded that DW might be too simplified for engineering purposes,
at least for small-scale local hydrodynamics. The Munice case is a relatively large flood inunda-
tion case. It is of interest to see how DW compares with other solvers, such as EM. The case was
run with DW, and the comparison is shown in Figure A-12. DW produces visually different
results. Note that no turbulence model is available when DW is selected as the solver. Based on
these findings, DW is not recommended for a case like the Munice inundation study. Note that
in HEC-RAS 2D v5 and v6.1, DW is the default choice for the solver. A modeler should always
check to make sure the proper solver is selected.
The computing times of different solvers for this case were compared in Table A-2. The com-
puting time by SRH-2D is also shown. HEC-RAS 2D EM needs a small time-step size to run
(at least 1 s in this case). All other solvers can run with a time-step size of 10 s.
Based on the results in this section, The research team decided to use ELM in HEC-RAS v6.1
for most of the cases simulated in the project:
• EM is a new solver option in HEC-RAS v6.1 whose performance has yet to be evaluated; ELM
has been around for quite some time.
• EM is computationally more expensive. For the team’s tests, EM required much smaller time
steps, which is because it treats the momentum advection term explicitly. To make it implicit,
HEC-RAS would need to adopt a different numerical approach.
• EM does not show an obvious advantage over ELM for large real-world cases.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

APPENDIX B

Detailed Evaluation
of Hole-in-Mesh Method

The research team considers hole-in-mesh as the reference method because it produces a
reasonable result and resolves the velocity field around an obstruction. One key question in the
hole-in-mesh method is the mesh resolution and when an obstruction is resolved. This section
tries mainly to address that question.

B.1 Setup of Cases


For this study, a channel with a width of 10 m, length of 150 m, slope of 0.0005, and uniform
Manning’s n of 0.03 was considered. The upstream boundary condition has a constant inflow
of 6 m3/s. The downstream boundary has a fixed water depth at normal depth (yn = 0.88 m for
n = 0.03). Two pier shapes and two sizes were considered as follows:
• Simple circular pier (diameter D = 1 m and 2 m)
– For D = 1 m: Mesh resolutions around the pier are 0.05 m, 0.1 m, 0.2 m, 0.3 m, 0.4 m, 0.5 m,
– 0.7 m, and 1 m.
– For D = 2 m: Mesh resolutions around the pier are 0.1 m, 0.3 m, 0.5 m, 1 m, and 1.5 m.
• Rounded-nose pier shape (nose diameter D = 1 m and 2 m). Both constant and variable mesh
resolutions were used.
– For D = 1 m: The constant-resolution meshes have average cell size of 0.1 m, 0.3 m, 0.5 m,
and 1 m. The variable-resolution meshes have cell sizes of 0.5–1 m, 0.5–1.5 m, and 0.5–3 m,
where the first number is the mesh size for the round nose and the second number is the
mesh size for the rectangular body.
– For D = 2 m: The constant-resolution meshes have average cell size of 0.1 m, 0.3 m, 0.5 m,
1 m, and 1.5 m. The variable-resolution meshes have cell sizes of 0.5–1 m, 0.5–1.5 m, and
0.5–3 m.
The modeling domain around the piers is the same as shown in Figure 4-4. The whole domain
is divided into blocks, and a gradual transition in mesh resolution is made to control the quality
of the mesh. Figure B-1 shows the shapes and sizes of piers. Figure B-2 shows the mesh resolu-
tions and the resulted shapes in models for all cases simulated. To further demonstrate the effect
of resolution on geometrically resolving shapes, Figure B-3 shows the variable-resolution mesh
grid points for the round-nose piers and the resulted shape in meshes. It is obvious that the geo-
metric shape of obstruction in 2D models can be vastly different from its real shape depending
on mesh resolution. This appendix will examine its effect.

B.2 Results and Discussions


In Figure B-4, the calculated water-surface and velocity profiles of the circular pier of 1 m
diameter at the center of the channel for different meshes are shown. Using the result from the
0.1-m-resolution mesh as the base, the absolute errors of other meshes are also shown. The result

191

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

192   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure B-1.   Different pier shapes and sizes used in


the study.

Figure B-2.   Mesh resolution and its effect on pier shape in models. The piers are
aligned vertically based on the shapes represented by different mesh resolutions.

Figure B-3.   Geometries of the round-nose pier in the models as a function of mesh
resolution. The numbers in each figure are variable mesh resolutions, e.g., 0.5–1
means the mesh has a resolution of 0.5 m for the round nose and 1 m for the
rectangular body.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Detailed Evaluation of Hole-in-Mesh Method   193

Figure B-4.   Effect of mesh resolution on WSE (a) and velocity (c) around the circular pier (diameter D 5 1 m);
(b) and (d) are the error of calculated water surface and velocity in comparison with mesh resolution of 0.1 m.

shows that the 0.3-m-resolution mesh has the closest results to those from the 0.1-m-resolution
mesh. Coarser meshes with cell sizes larger than 0.3 m produce less accurate results.
Similar profiles of WSE and velocity are plotted in Figure B-5 for the circular pier with a diam-
eter of 2 m. The results show that the 0.3- and 0.5-m-resolution meshes produce results close to
those from the 0.1-m-resolution mesh. Coarser meshes produce less-ideal results.
For round-nose piers with diameters of 1 m and 2 m, the profiles are shown in Figures B-6
and B-7, respectively. For the case of 1 m diameter (D = 1 m), all meshes except the one with
1-m resolution produce comparable results. For the case of 2 m diameter, the 1-m and 1.5-m-
resolution meshes are far from the results of the 0.1-m-resolution mesh.
In general, for round-nose piers, mesh resolution near the nose has less effect on the results
than circular piers. This is because the main source of head loss around a round-nose pier is the
rectangular part of the pier. The two semicircles upstream and downstream (noses) have less effect.
To quantify the effect of mesh resolution, the research team calculated the average increase
in water depth in comparison with normal depth along the centerline of the channel (∆y) for
different mesh resolutions. Immediately upstream of the pier, there is an increase because of the

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

194   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure B-5.   Effect of mesh resolution on WSE (a) and velocity (c) around the circular pier (diameter D 5 2 m);
(b) and (d) are the error of calculated water surface and velocity in comparison with mesh resolution of 0.1 m.

pier local effects. Therefore, the average ∆y was calculated from the increase of the WSE further
upstream excluding this local increase. This water-depth change quantifies the backwater effect
of an obstruction.
The results of water-depth change for all cases are shown in Figure B-8 (left). Furthermore, the
relative error of each mesh resolution in comparison with the finest mesh is also shown (right).
The average water-depth change depends on mesh resolution. In general, it was found that lower
mesh resolution will produce larger water-depth change ∆y (with some exceptions). One pos-
sible explanation is that when the mesh is coarse, the pier geometry becomes edgier and the
corners become sharper. As a result, there is more head loss.
In the hole-in-mesh method, whether the cell size is coarse or refined is relative to the size of
the obstruction. For cases in this section, the width of a pier is one important metric for obstruc-
tion size. For example, the cell size of 1 m for a round-nose pier with a diameter of 1 m is too
coarse. In fact, the round-nose pier becomes a sharp-nose pier in the mesh (see Figure B-2).
To measure the relative mesh resolution with respect to obstruction size, use the CSO ratio.
The inverse of this ratio is approximately the number of grid points along an obstruction feature.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Detailed Evaluation of Hole-in-Mesh Method   195

Figure B-6.   Effect of mesh resolution on WSE (a) and velocity (c) around the round-nose pier (diameter D 5 1 m);
(b) and (d) are the error of calculated water surface and velocity in comparison with mesh resolution of 0.1 m.

From the results of this section, to geometrically resolve a feature, several grid points (e.g., 6 to 7)
are necessary. This number also should depend on the curvature of the geometry, in other words,
how quickly the shape changes along the perimeter. The quicker it changes, the more grid points
that are needed.
For obstructions with different sizes in different directions, the number of grid points in each
direction can differ. For example, in the round-nose pier, the body part has a difference in size
from the nose part. The variable-resolution meshes previously described were used to investigate
how many grid points are needed along the body of a round-nose pier. Figure B-3 shows how
they are represented in meshes with different resolutions.
In Figure B-9, the water-surface and velocity profiles along the center of the channel are shown
for the round-nose pier with a diameter of 1 m. The profiles with constant mesh resolutions
(0.5 m and 1 m) are also shown for comparison. The comparison shows that the water-surface
profiles with cell sizes of 0.5 m, 1 m, 0.5–1 m (0.5 m for the nose and 1 m for the straight body
part), and 0.5–1.5 m are close to each other. The results from the mesh resolution of 0.5–3 m
mesh are far from others. Thus, it is too coarse.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

196   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure B-7.   Effect of mesh resolution on WSE (a) and velocity (c) around the round-nose pier (diameter D 5 2 m);
(b) and (d) are the error of calculated water surface and velocity in comparison with mesh resolution of 0.1 m.

The similar plots for the round-nose pier with a diameter of 2 m are shown in Figure B-10. The
figure shows that the results with cell sizes of 0.5–1.5 m and 1–1.5 m are close to the results with
the constant cell size of 1 m. The results with a cell size of 1–3 m show it is too coarse.
Based on the above, for the cases studied in this section, at least three grid points are needed
for the rectangular body of the round-nose piers. It is much less than the six or seven grid points
needed for the round-nose part. This is expected because the geometry of the body part is just a
simple straight line, and the curvature is infinity.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Detailed Evaluation of Hole-in-Mesh Method   197

Figure B-8.   Effect of mesh resolution on simulated water-depth change from


background depth.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

198   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure B-9.   WSE and velocity profiles produced with different meshes for the round-nose pier with a diameter
of 1 m.

Figure B-10.   WSE and velocity profiles produced with different meshes for the round-nose pier with a diameter
of 2 m.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

APPENDIX C

Manning’s n Sensitivity Analysis

This part of sensitivity analysis is not in the originally proposed work scope. However, the
research team found it is important to explain some observations made in some of the simulation
cases. For example, the team found that generally, M1 profiles in a short domain are less sensitive
to Manning’s n changes than M2 profiles.
The sensitivity of the hydraulic model result to Manning’s n depends on several factors. It
is important for a modeler to understand these factors and expected model behaviors when
Manning’s n changes. For cases sensitive to Manning’s n, more care should be taken to specify
the best roughness coefficient. On the other hand, for cases less sensitive to Manning’s n, the
requirement of accurate Manning’s n can be slightly relaxed for certain applications.

C.1 Governing Equation


To have a comprehensive view of the sensitivity to Manning’s n and make the conclusions
more generalizable, a systematic and rigorous approach was adopted. The analysis is based on
the simplified version of the SWE, in other words, the backwater curve equation.
For a wide rectangular channel, the backwater curve equation has the form of the following:
q w2 n 2
s0 -
dy s 0 - sf y 10 3
= = (C-1)
dx 1 - Fr 2
q2
1 - w3
gy
where y is water depth, x is a streamwise coordinate, s0 is bed slope, sf is the friction slope induced by
flow resistance, Fr is the Froude number, and qw is the specific discharge (discharge per unit width).
At equilibrium (i.e., the normal flow condition), the gravity force and the flow resistance force
are in balance (s0 = sf). Disturbance in the channel will break the equilibrium. Water depth and
velocity will change along the channel. The change is governed by the backwater curve equation
shown in Equation C-1. Depending on the flow status, the variation of WSE in a channel can
be classified into different profiles (e.g., M1 and S2). Figure C-1 shows different water-surface
profiles. This report will not repeat the basic theories of open-channel flows. More information
can be found in many textbooks.

C.2 Some Analysis on Manning’s n Sensitivity


The research team noted dn as the change or error in Manning’s n. Two important WSE profiles
on mild (M) and steep (S) slopes, which are common in reality, will be discussed. The theoretical
derivation will give a valuable understanding of how Manning’s n error will affect different pro-
files and where to expect the maximum error in water depth will occur.

199

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

200   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Figure C-1.   Classification of open-channel flow profiles [after Chaudhry


(2008)].

From Equation C-1, when the depth y is larger, the sensitivity should be less because the water
depth is in the denominator of the friction slope term Sf , which involves Manning’s n:

q 2w n 2
Sf = (C-2)
y 10 3

Based on this, it is expected that the M3 profile is more sensitive than the M2 profile, which
itself is more sensitive than the M1 profile. The same is true for steep slope curves: S1, S2, and S3
curves are in the increasing order of sensitivity to Manning’s n if everything else is kept constant.
In reality, the flow profiles are more complicated, and there is probably no well-defined mild or
steep slope curves. To deal with that, some average needs to be done to capture the bulk behavior
of the open-channel flow. For example, average channel width, slope, Manning’s n, and down-
stream water depth can be used to determine the category of flow profile.
To demonstrate the sensitivity of flow to Manning’s n, the backwater curve equation (Equa-
tion C-1) was solved for some simple M1 and M2 flow profile cases. For all these cases, the spe-
cific discharge is qw = 6 m2/s, and the bed slope is S0 = 1E-4. The results are shown in Figures C-2
and C-3. For the M1 curve cases, the simulation domain length has a value of 2 km (short) and
20 km (long). The downstream water depth is 8 m. The Manning’s n values of 0.03 and 0.025 are
used. For the M2 curve cases, the simulation domain length is 2 km (no need to simulate in a long
domain for the research team’s purpose). However, the downstream water depth has two values
(4.5 m and 2 m). One water depth (4.5 m) is smaller but close to the normal depth (to ensure a
M2 curve). The other water depth (2 m) is much smaller than the water depth.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Manning’s n Sensitivity Analysis   201

Figure C-2.   Response of M1 curve to the change of Manning’s n in long and short domains.

Figure C-3.   Response of M2 curve to the change of Manning’s n and downstream boundary
condition.

For the M1 curve cases shown in Figure C-2, the effect of Manning’s n takes a long distance
to accumulate and show in the resulted flow profiles. In the short domain (2 km in this case),
the flow profiles for both Manning’s n values are almost identical. Only after about 20 km, there
starts to have some discernible differences. The insensitivity of the M1 curve in the short domain
is due to the relatively deep water.
For the M2 curve cases shown in Figure C-3, everything is the same except the downstream
boundary condition of water depth. When the water depth is far from the normal depth, the
M2 curve tends to recover to the normal depth upstream quickly. Thus, the sensitivity of the
Manning’s n is more significant over a short distance. This sensitivity is because the M2 curve has
relatively shallow water—the shallower, the more sensitive.
These examples are some specific cases to show the point that is made in this section. Similar
cases can be developed for other flow profiles. In the end, it all boils down to the relative impor-
tance of the flow resistance term among all terms in the governing equation. The following
guidelines and recommendations can be made:
• For M1 flow profiles,
– If the simulation domain is short, the result (velocity and WSE) is less sensitive to Manning’s n.
In this case, the gravity term is dominating over the flow resistance term.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

202   Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

– If the simulation domain is long, then the result, especially in areas way upstream, will be
sensitive to Manning’s n. The flow resistance eventually will show its cumulative effect after
a long distance.
• For M2 flow profiles, the result will be typically sensitive to Manning’s n if the downstream
boundary is not close to normal flow condition.
• Similar analysis can be done for other flow profiles.
• These results give the modeler some idea on what to expect in terms of Manning’s n value
calibration.
• It is important to point out that the sensitivity discussed here refers only to the overall flow
result (velocity and WSE). In some application scenarios, the insensitivity to Manning’s n does
not mean its accuracy is not important.
– For 2D simulations, Manning’s n also redistributes flow laterally. Flow will prefer the path
with the least resistance. For an M1 curve in a short domain, it was stated that the overall
result is not sensitive to Manning’s n. However, if there is flow splitting in the domain, then
Manning’s n in different zones across the channel will play an important role in directing
the flow.
– Manning’s n may also affect flow attachment and recirculation behind a structure. This
may impact the assessment of habitat suitability.
– Manning’s n is important regardless of the flow profiles if bed shear stress is of interest,
for example, for sediment transport and bridge scour analysis. The bed shear is calculated
as follows:

tgn 2 2
xb = U (C-3)
h1 3

So, doubling the Manning’s n value will quadruple the bed shear if everything else is kept
constant. In different 2D models, the way to calculate bed shear stress may be different. Further,
the total bed shear should be partitioned to skin friction and form drag, the former of which is
usually used for sediment transport. Considering the results reported in this project, the meaning
of the Manning’s n in 2D models is not exactly how it is defined in textbooks. The Manning’s n
specified in 2D models represents only the unresolved flow resistance. Therefore, the bed shear
calculated with Equation C-3 is theoretically only the unresolved bed shear. How to use that in
sediment transport and bridge scour analysis needs further study.

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Abbreviations and acronyms used without definitions in TRB publications:


A4A Airlines for America
AAAE American Association of Airport Executives
AASHO American Association of State Highway Officials
AASHTO American Association of State Highway and Transportation Officials
ACI–NA Airports Council International–North America
ACRP Airport Cooperative Research Program
ADA Americans with Disabilities Act
APTA American Public Transportation Association
ASCE American Society of Civil Engineers
ASME American Society of Mechanical Engineers
ASTM American Society for Testing and Materials
ATA American Trucking Associations
CTAA Community Transportation Association of America
CTBSSP Commercial Truck and Bus Safety Synthesis Program
DHS Department of Homeland Security
DOE Department of Energy
EPA Environmental Protection Agency
FAA Federal Aviation Administration
FAST Fixing America’s Surface Transportation Act (2015)
FHWA Federal Highway Administration
FMCSA Federal Motor Carrier Safety Administration
FRA Federal Railroad Administration
FTA Federal Transit Administration
GHSA Governors Highway Safety Association
HMCRP Hazardous Materials Cooperative Research Program
IEEE Institute of Electrical and Electronics Engineers
ISTEA Intermodal Surface Transportation Efficiency Act of 1991
ITE Institute of Transportation Engineers
MAP-21 Moving Ahead for Progress in the 21st Century Act (2012)
NASA National Aeronautics and Space Administration
NASAO National Association of State Aviation Officials
NCFRP National Cooperative Freight Research Program
NCHRP National Cooperative Highway Research Program
NHTSA National Highway Traffic Safety Administration
NTSB National Transportation Safety Board
PHMSA Pipeline and Hazardous Materials Safety Administration
RITA Research and Innovative Technology Administration
SAE Society of Automotive Engineers
SAFETEA-LU Safe, Accountable, Flexible, Efficient Transportation Equity Act:
A Legacy for Users (2005)
TCRP Transit Cooperative Research Program
TEA-21 Transportation Equity Act for the 21st Century (1998)
TRB Transportation Research Board
TSA Transportation Security Administration
U.S. DOT United States Department of Transportation

Copyright National Academy of Sciences. All rights reserved.


Selection and Application of Manning’s Roughness Values in Two-Dimensional Hydraulic Models

Transportation Research Board


500 Fifth Street, NW
Washington, DC 20001

ADDRESS SERVICE REQUESTED

ISBN 978-0-309-70944-6
90000

9 780309 709446

Copyright National Academy of Sciences. All rights reserved.

You might also like