ANALYSIS OF CHANNEL AND FLOODPLAIN HYDRODYNAMICS AND FLOODING ON
THE MISSISSIPPI RIVER

By

Muhammad Bilal Zafar

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

Civil Engineering—Master of Science

2023

ABSTRACT

Flooding poses a significant threat to communities, ecosystems, and infrastructure, necessitating a

comprehensive understanding of hydrological dynamics for effective mitigation and management.

Flood prediction and hydrological modeling stand at the forefront of addressing contemporary

challenges in water resources management. Against the backdrop of escalating climate variability

and landscape modifications, the accurate prediction of floods becomes paramount. Unprecedented

rainfall, changing river discharges, and evolving topographical features contribute to the challenges

of predicting and managing floods. This research delves into the complex dynamics of flood

modeling, presenting a comprehensive analysis of model performance, sensitivity to key parameters,

and the impact of various environmental factors with a focus on the Mississippi River. A high-

resolution hydrodynamic model was developed and tested using field observations and aerial images

of flood inundation extent for the Middle Mississippi River focusing on simulating significant

flooding events in 2019, 2018, and 2020. The research establishes the model’s temporal consistency,

showcasing its reliability across multiple years. A systematic analysis of model sensitivity to data

and mesh resolution, turbulence model choice, and Manning’s roughness on predictive accuracy

is reported. The study unravels the complex interplay of increased rainfall, river discharge, and

dynamic landscape changes, emphasizing their influence on the precision of flood prediction.

Model performance related to the choice of bottom roughness and the varying impacts of different

topo-bathy resolutions on location-specific hydrological dynamics are notable observations. The

investigation into the effects of levee construction at St. Louis, MO provides valuable insights into

the dynamic impact of levees on gauge height and discharge patterns. The research recommends

strategies for model optimization, including the meticulous selection of hydrological parameters and

enhanced spatial resolution. The need for continuous model refinement and validation against real-

world data is stressed. In addition to advancing hydrodynamic modeling and providing practical

guidelines for flood management and infrastructure development, the research underscores the

significance of collaborative decision-making with stakeholders and a comprehensive understanding

of the spatial and temporal dynamics of levee impacts on hydrological processes.

Copyright by
MUHAMMAD BILAL ZAFAR
2023

To my parents, whose unwavering support and encouragement have been my guiding light. To
Madiha, my constant source of strength and inspiration. To my beloved daughters, Shiza and Irha,
may you always strive for excellence and never give up on your dreams. To my family, friends and
my mentors, thank you for your endless love and encouragement. Your support has been the
cornerstone of my success. This achievement is as much yours as it is mine.

iv

ACKNOWLEDGEMENTS

I extend my deepest gratitude to my esteemed advisor, Dr. Mantha S. Phanikumar, whose unwa-

vering support, valuable suggestions, and continuous supervision have been instrumental in the

successful completion of my MS degree. His exceptional dedication went above and beyond, shap-

ing not only my academic pursuits but also fostering a profound growth in my research capabilities.

I am also indebted to Dr. Shu-Guang Li and Dr. Yadu Pokhrel for their influential support, which

significantly contributed to shaping my experimental methods. Their guidance, coupled with that

of Dr. Mantha, has made my study and life in the USA a truly enriching experience.

The faculty of the Military College of Engineering, Risalpur, Pakistan, played a crucial role in

grooming my professional knowledge, and I extend my sincere appreciation to them. Additionally,

I am grateful to the Ministry of Defence, Pakistan, for funding my studies in the Department of

Civil & Environmental Engineering, Michigan State University, East Lansing, USA.

My heartfelt thanks go to my parents for their unwavering prayers and unmatched support.

I

am equally grateful to my wife and daughters for their love, patience, and moral support. Their

understanding and encouragement have been the bedrock of my academic journey.

Special appreciation is extended to my friends, colleagues and my course mates for the cherished

moments spent together in the lab and social settings. A particular mention goes to my lab mate,

Saeed Memari, for his valuable feed backs specially during our brainstorming sessions.

v

TABLE OF CONTENTS

CHAPTER 1

1.1 Background .
.
1.2 The Mississippi River

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.

.

.

.

1
1
3

CHAPTER 2

.
2.1 Flood Overview . .
2.2 Flooding Impacts .
.
2.3 The Mississippi River
2.4 River Flood Modeling .

9
LITERATURE REVIEW . . . . . . . . . . . . . . . . . . . . . . . .
9
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. 10
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. 13
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

CHAPTER 3

3.1 Research Statement
.
.
3.2 Research Goals
.
3.3 Hypotheses
.
.
.
3.4 Site Description .

RESEARCH GOALS, HYPOTHESIS AND SITE DESCRIPTION . 24
. 24
. 24
. 25
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.

.
.
.
.

.

CHAPTER 4

. .
4.1 Data sets .
4.2 Theory . .
. .
4.3 Post Processing .

METHODOLOGY . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
. 39

.
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. .
. .

.
.
.

.
.
.

CHAPTER 5

RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . 42
5.1 Case 1 : 2019 River Model (Base Case Scenario)
. . . . . . . . . . . . . . . . 43
5.2 Case 2 : 2020 River Model
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.3 Case 3 : 2018 River Model
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
. . . . . . . . . . . . . . . . . . .
5.4 Case 4 : Effects of Manning’s Roughness
. 67
5.5 Case 5 : Effects of Interpolation Methods
. . . . . . . . . . . . . . . . . . .
. 72
. 78
5.6 Case 6: Impact of Topobathy Resolution . . . . . . . . . . . . . . . . . . . .
5.7 Case 7: Impact of Reduced Discharge . . . . . . . . . . . . . . . . . . . . . . 84
5.8 Case 8: Impact of Increased Discharge . . . . . . . . . . . . . . . . . . . . . . 88
5.9 Case 9: Effects of Levee Construction . . . . . . . . . . . . . . . . . . . . . . 91
5.10 Case 10: Turbulence Model Comparison (k-𝜀 vs. Parabolic)
. . . . . . . . . . 96

CHAPTER 6

. .
6.1 Conclusions . .
6.2 Recommendations .

CONCLUSIONS AND RECOMMENDATIONS . . . . . . . . . . . 101
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

.
.

BIBLIOGRAPHY .

.

.

.

.

.

.

.

.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

vi

CHAPTER 1

INTRODUCTION

1.1 Background

Floods have impacted both humans and the environment for millennia. The interplay of flooding,

land subsidence, and relative sea level rise constitutes a complex and interconnected challenge for

coastal regions worldwide. While relative sea level rise, driven by global climate change and the

melting of polar ice caps, poses a long-term threat, land subsidence further exacerbates the problem

in many coastal areas by causing the land to sink. Human activities, such as over-extraction of

groundwater and urban development, often contribute to land subsidence. The combined effects

of these factors increases the vulnerability of coastal areas to flooding. As sea levels rise, low-

lying regions become more susceptible to inundation, with potentially severe consequences for

communities, ecosystems, and infrastructure [1].

The frequency and intensity of flooding in inland areas has increased in many regions around

the world due to a combination of factors including heavy rainfall, snowmelt, deforestation and

urbanization, flash flooding due to sudden and intense rainfall in areas with poor drainage systems,

ice jams in rivers that obstruct the natural flow in colder climates causing downstream flooding,

storm surges in coastal areas associated with tropical storms or hurricanes pushing seawater inland,

and finally poor land use planning and extreme weather events in general. Addressing these

multifaceted issues requires a comprehensive approach that includes sustainable land use planning,

climate change mitigation efforts, and adaptive measures including the development of early flood

warning systems to enhance resilience in the face of a changing environment. Models are useful

tools for understanding the key factors that influence flooding and can inform the development of

early warning systems.

The impacts of flooding could be broadly characterized into social and economic domains.

Consequences have included the loss of human lives, displacement and damage to infrastructure.

Moreover, contamination of drinking water systems and the subsequent deterioration of public

health have posed long term challenges. The 1927 Mississippi flood led to substantial property

1

damage totaling over $41 million and significant livestock losses amounting to more than $3.5

million [2]. Floods have also significantly impacted regional and national economies, particularly

when public infrastructure such as roads, rail networks, and shipping ports are damaged [3].

Devastating impacts were observed including damage to infrastructure, contamination of drinking

water systems, and deterioration of health and infrastructure. Contamination of the Mississippi

River’s water systems and subsequent health deterioration have compounded the challenges faced

by the communities along its banks. The focus of the present research is on the Mississippi River.

A number of factors have contributed to the selection of the river including the length of the river

and the size of its drainage basin, the central role of the Mississippi in North American geography,

the economic importance of the river, the diverse ecosystems supported by the river, the cultural

and historical significance of the many cities along the river’s course such as New Orleans and St.

Louis and the continual challenges posed by flooding on the Mississipi and the complexity of flood

management along the river. Understanding the specific impacts of Mississippi River flooding is

pivotal in formulating tailored flood control and risk management strategies that can mitigate the

recurring devastation experienced by the communities reliant on its waters.

A number of factors contribute to flooding on the Mississippi including some key factors

described in the previous paragraphs. Although levees and other flood control structures are

intended to mitigate flooding, they can also compound problems by confining water within the river

channel, potentially leading to more severe flooding downstream if the structures are over-topped

or breached. Accumulation of sediment within the river channel can reduce the capacity to convey

water, increasing the risk of flooding during high-flow events.

In addition, the management of

reservoirs upstream can impact downstream flooding. Release of water from these reservoirs,

especially during periods of heavy rainfall, can contribute to increased water levels and flooding

in the Mississippi. Although a comprehensive analysis of all these factors is beyond the scope

of the present work, the primary focus is on the development, testing and the application of a

hydrodynamic model of the Mississippi. In particular, the model is applied to simulate the 2019

flood on the Mississippi River which was a significant event that brought widespread flooding

2

to various regions along the river and its tributaries. The 2019 flood was triggered by heavy

and sustained rainfall throughout the central United States (Annual 2019 National Climate Report,

National Centers for Environmental Information [4]). This led to swollen rivers, saturated soils, and

increased runoff, exacerbating flood conditions. The combination of melting snow from the winter

months, particularly in the northern parts of the Mississippi River Basin, added to the volume of

water flowing into the river during the spring months. The flood event featured multiple crests as

rain continued to fall and water moved downstream. This prolonged duration of high water levels

intensified the impact on communities along the river. Some levees along the Mississippi River

and its tributaries were breached, contributing to the extent of the flooding. Levees are typically

constructed to protect against high water levels, but they can be compromised under extreme and

sustained rainfall. The flooding had severe consequences for agriculture, as many fields in the

floodplain were submerged. This led to delays in planting, damaged crops, and economic losses for

farmers in the affected regions. The flood disrupted river navigation, affecting transportation and

commerce along the Mississippi River. High water levels and strong currents made it challenging

for barges and vessels to navigate the waterway. The flood prompted evacuations in some areas,

and numerous homes and businesses suffered damage. Communities along the river, particularly

in the Midwest and South, faced hardships as floodwaters persisted. The 2019 flood highlighted

the complexities of managing water resources in the Mississippi River Basin and the economic

losses associated with the 2019 flooding were estimated to be in excess of a staggering 20 billion

US dollars [5]. The specific research questions and tasks are highlighted in Chapter 3.

1.2 The Mississippi River

Mississippi River is one of the longest rivers in the world and stretches over 2,320 miles (3,730

km) from its source in Minnesota down to the Gulf of Mexico. With a catchment area of 1.2

million square miles (3.1 million square kilometers), the watershed is the fourth largest watershed

globally and its history is marked by the impact of numerous floods. It flows through or along

the borders of 10 U.S. states, from Minnesota in the north to Louisiana in the south, serving as a

major transportation route and influencing settlement patterns and economic development in the

3

region. The Mississippi River is a critical economic artery for the United States.

It facilitates

transportation of goods, with numerous cities and industries located along its banks. The river and

its tributaries are extensively used for shipping, supporting agriculture, industry, and commerce.

Its high susceptibility to periodic floods makes it one of the most flood-prone regions in the United

States, leaving a lasting impact on the communities and the environment in the region. In response to

these challenges, concerted efforts have been made to strengthen flood control measures, improve

river infrastructure, and develop early warning systems to reduce the risks to life and property.

Based on its location and characteristics, the Mississippi River is divided into three sections [6].

1. Upper Mississsppi River (UMR).

2. Middle Mississsppi River (MMR) (sometimes included as part of UMR).

3. Lower Mississsppi River (LMR).

1.2.1 Upper Mississippi River

The Upper Mississippi River, extends headwaters to its confluence with the Missouri River at

St. Louis. Its spans an impressive length of approximately 1,200 miles approximately (Figure 1.1).

1.2.2 Middle Mississippi River

The Middle Mississippi River is the section of the Mississippi River that is spanned from its

confluence with the Missouri River at St. Louis, Missouri, for 190 miles to its confluence with

the Ohio River at Cairo, Illinois (Figure 1.2). The Middle Mississippi River had been affected by

various significant floods over its history, among them the Great Mississippi Flood of 1927 and

the Great Mississippi and Missouri Rivers Flood of 1993. Flood control and risk management

strategies, including the implementation of levees and flood ways, had been employed to diminish

the impacts of flooding.

1.2.3 Lower Mississippi River

The Lower Mississippi River is the segment of the Mississippi River situated downstream of

Cairo, Illinois, and spans just under 1000 miles to the Gulf of Mexico (Figure 1.3). It is the most

4

heavily traveled component of the Mississippi River System, serving as a critical transportation

route for barges conveying goods such as grain, coal, and petroleum products.

1.2.4 Study Area of Interest

This research is focused on the Middle Mississippi River (MMR) flood modeling owing to

the factors briefly discussed in section 1.2.2 which are discussed in detail in chapter 2.3. The

Mississippi River is a long river and the application of multi-dimensional hydrodynamics models

involves significant computational resources in addition to large data requirements and processing.

Modeling the lower Mississippi River entails modeling the coastal ocean and the specification of

tidal boundary conditions. On the other hand, the Middle Mississippi River while still representing

a long river reach, has the advantage of having several USGS gages that provide data for model

forcing (boundary conditions) and model testing (e.g., using data from gages located between the

inlet and outlet sections). Therefore the MMR is selected for the present study.

5

Figure 1.1 Upper Mississippi River
(https://heartlandsconservancy.org)

6

Figure 1.2 Middle Mississippi River
(https://heartlandsconservancy.org)

7

Figure 1.3 Lower Mississippi River
(https://https://americaswatershed.org)

8

CHAPTER 2

LITERATURE REVIEW

The literature on flood modeling has seen significant expansion, focusing on diverse flood types,

including coastal and riverine floods, and investigating various components of flooding, notably

encompassing hydrological and hydrodynamic processes. Additionally, the literature underscores

the use of different modeling approaches, with a specific emphasis on the integration of physically

based and data-driven methods. Notably, recent studies have emphasized the significance of

accurately capturing the complex dynamics of flood propagation, highlighting the need for robust

and reliable models to aid in effective flood management and mitigation strategies. However, despite

the progress in flood modeling, there remains a persistent need for comprehensive assessments and

comparative analyses to validate the efficacy and accuracy of different modeling methodologies

under various scenarios. This review aims to synthesize the existing literature, identify critical

gaps, and provide insights into the current state of flood modeling, with a focus on understanding

the strengths and limitations of the predominant modeling approaches in addressing real-world

flood challenges.

2.1 Flood Overview

Floods have emerged as the most widespread catastrophe globally. Flooding has significantly

contributed towards fatalities caused by natural disasters and its effects are amplified due to a

number of factors including climate change, deforestation and the ever-increasing encroachment of

populous areas into coastal zones, river basins, and lakeside regions [7].

Over the past decades, there has been a significant surge in both the overall population and the

economic value of material assets situated in regions susceptible to flooding. This trend is projected

to persist and intensify, driven by two key factors. First, the continuous expansion of global

population and wealth has contributed to the increased development and habitation of flood-prone

areas. Second, the escalating impacts of climate change, including rises in sea levels and increased

frequencies of flooding events, are anticipated to further amplify the vulnerability of these regions

[8].

9

Throughout history, river floods have consistently posed a significant natural hazard, to which

our coping mechanisms have proven insufficient. Recent times have witnessed multiple instances

of river flood events, each resulting in material losses surpassing the staggering sum of US $10

billion [9]. Despite advancements, effective management of these recurring disasters remains a

formidable challenge, necessitating comprehensive strategies to mitigate their devastating impacts

on communities and infrastructure [9]. Anticipated projections indicate a substantial rise in flood

exposure. It is estimated to increase threefold by 2050, depending on the socioeconomic scenarios

[10]. Relative to the reference period of 1976–2005, the forecasted consequences include a potential

70–83% surge in human casualties resulting, along with a projected escalation of 160–240% in

direct flood-related damages. These projections are contingent upon a global temperature increase

of 1.5°C , as referenced in prior studies [10].

2.2 Flooding Impacts

Floods are regarded as the deadliest and most costly natural disaster in both the US and

worldwide [11]. Since 1980, global damages exceeding $1 trillion have been incurred by flooding

[11]. In addition to the extensive financial toll, floods have left a lasting imprint on communities,

which resonate well beyond its immediate aftermath.

2.2.1 Economic Impacts

Financial instability is witnessed in the wake of floods. Immediate consequences include dam-

age of infrastructure, residential properties, standing crops, roads, bridges. All these impede local

and regional connectivity, resulting in hampering of trade and economic activities. Businesses

encounter severe setbacks due to impaired inventories and interruption of supply chains. Further-

more, property damages result in displacement of inhabitants, which inflicts economic burden and

emotional distress to the affected communities. The scale of economic damage is further exacer-

bated by the long-term effects of the floods, including a decline in investment confidence and a

slow recovery process that continues to burden the affected regions.

The overall economic losses in 2022 floods in Pakistan were approximated at $15.2 billion [12].

The estimated requirements for the rehabilitation and reconstruction efforts, focusing on bolstering

10

resilience, were estimated at a minimum of $16.3 billion [12].

The direct impact of natural disasters are extended to the disruption or damage of crucial production

factors, notably labor [13] and physical capital [14]. This disruption can lead to a significant reduc-

tion in the workforce’s ability to participate in economic activities, thereby impeding the overall

economic output and recovery in the affected regions. In addition to the direct economic losses,

future remediation is an important aspect. Investments are also required for fortifying adaptation to

climate change and enhance the country’s overall resilience against potential future climate-related

adversities.

"The Great Flood of 2019," which primarily ravaged the Missouri River, had a widespread impact.

This catastrophic flood resulted in the loss of at least 12 lives and inflicted property and agricultural

damages exceeding $20 billions [5, 15] in 19 states. The disruption caused by the flood severely

impacted various industries. The agricultural production in the Southern Plains states bearing a

significant brunt due to extensive flooding and continuous heavy rainfall. Moreover, the prolonged

duration of the flooding event led to a substantial reduction in crop planting, spanning millions of

acres. The extensive infrastructure damage across multiple cities and towns in the region, cou-

pled with the disruption of barge and other transportation traffic due to high water levels, further

intensified the adverse impact on various industries [15].

2.2.2 Social Impacts

The aftermath of flooding entails more than just the destruction of infrastructure and property.

Flood fatalities are a major aftermath of flooding, with a large number of people losing their lives

to flooding worldwide (Figure 2.1). It also inflicts adverse effects on the citizens affected by the

disaster. Effects on physical and mental well-being can persist both in the short term and the long

term, leading to significant alterations in the livelihoods of the affected individuals.

Direct impacts include physical and mental health. Consequences arise from water borne diseases,

injuries due to floods and loss of medical care due to disrupted medical infrastructure [16], while

many people are left with stress, anxiety and emotional consequences.

The majority of physical health effects are manifested in the weeks and months following flooding,

11

while the psychological impacts could linger for years [17]

2022 floods in Pakistan had severe social impact.The floods affected 33 million people and more

than 1730 lost their lives. Stagnation of floodwaters in several areas led to the spread of water-borne

and vector-borne diseases. Moreover 8 million were displaced. Most of the affected are facing

acute health crises till date and the situation continues to evolve [12]. The crisis has resulted in

the loss of household incomes and assets, further amplified by soaring food prices and outbreaks

of diseases. Notably, those tied to agriculture and livestock have borne significant losses in their

livelihoods.

Figure 2.1 Flood Fatalities (1980-2016) according to EM-DAT

12

2.3 The Mississippi River

The Mississippi River serves as a vital economic lifeline for the United States. Its the fourth

largest river basin in the world [18]. Since the 19th century, federal initiatives have been in progress

to grasp, foresee, and regulate flooding along its path [19]. While floods occur across the USA,

their frequency is on the rise. This led to escalating losses, especially in the states astride the

Mississippi River. Notably, several recent records along major rivers in these regions have been

established (Figure 2.2 (source[18])).

Figure 2.2 Mississippi River flood records (1927-2016)

13

The recent flood of 2019 on the Mississippi River was one of the most severe flooding incidents

since the Great Flood of 1927. It resulted in a staggering $20 billion in damages [5]. Following

the Great Flood of 1927, the federal government began displaying a vested interest in establishing

sustainable flood control measures. The primary objective was to construct a flood-control system

capable of effectively managing a projected flood [20]. These included mainly construction of

levees, flood ways, reservoir and tributary basin improvements. However, these had both positive

and negative impacts including reduction of flood risks at the cost of narrowing the river and

increasing the river stage [18].

Native Americans had a belief that Mississippi river floods every 14 years. Subsequently each

wave of settlers in the region endeavored to manage the river by construction of levees. With

time, assistance from the federal government was also sought, which resulted in creation of the

Mississippi River Commission in 1879. The US Army Corps of Engineers was authorized to

participate in levee building on the river. In 1885, the "levees-only" policy was adopted based on

the theory that it would result in the scouring of the river floor. In reality this policy resulted in

the river to rise and with time, higher and higher levees were needed to contain the water. The

levees initially built to a height of 7 feet were to be raised as much as to 38 feet. As the levees

grew taller and stronger, river force, river volume, and consequences of levees break followed

the same trend. When the levee broke eventually, the poor construction was blamed rather than

blaming the levees-only policy. The fact that there were no effective modeling techniques available

to simulate the river flows could be used as an explanation to the erroneous judgments made in the

past. Those were the decisions made on the best available knowledge and experiences of that time.

In recent decades, the numerical modeling of flood events has undergone significant advancements,

primarily owing to the refinement of dependable numerical techniques, the increase in computing

capabilities, and the emergence of pioneering topographic survey methodologies [21]. Ironically,

despite the substantial advancements in flood simulation techniques, concerns have emerged within

the literature regarding an over reliance on these improvements. Specifically, a valid apprehension

has been raised that "sophisticated high-resolution models might be dangerous from this viewpoint

14

as the false sense of confidence derived from their spuriously precise results might lead to making

the wrong decisions [22]." A survey of recent literature on flood inundation modeling reveals certain

identifiable patterns. Especially in the context of managing urban flooding, a prevalent approach

involves adjusting the grid resolution based on the available topographic data. This strategy is

motivated by various factors. It is commonly accepted that refining the model resolution is crucial

to minimizing errors in representing physical processes, such as the impact of topographic forces

on flow dynamics. Additionally, the use of coarse meshes can introduce errors due to numerical

diffusion, which smoothens out inertial effects, a phenomenon that holds significance, particularly

in urban areas or during high-energy flow conditions [22].

2.4 River Flood Modeling

Given its capacity to predict and effectively mitigate the consequences of floods, flood modeling

stands as a crucial method in flood management [23]. Understanding and predicting flood patterns

and impacts are facilitated through the utilization of flood modeling, making it an indispensable tool

[24]. Anticipating the spatial and temporal dissemination of floodwaters, alongside the resulting

hazards and damages, demands the creation of mathematical models that imitate the hydrologic

and hydraulic processes accountable for flooding [25].

Among the various types of flood models, including 1D hydraulic models, 2D hydraulic models,

and hydrological models, 1D hydraulic models replicate river and canal water movement based on

hydraulic engineering principles. These models are commonly employed to simulate the effects

of flood control structures such as dams and levees and to forecast flooding. Despite their relative

simplicity and user-friendliness, 1D hydraulic models might not fully capture the complex inter-

play between floodwaters and the surrounding landscape [23]. Conversely, 2D (depth-averaged or

vertically-integrated) hydraulic models simulate water flow both longitudinally and in cross-section,

providing more comprehensive information on the distribution of floodwaters and their effects on

adjacent areas. Apart from representing the geographical features and land usage of the region,

these models can also demonstrate how floodwaters interact with their environment, including the

influences of urbanization and vegetation [26].

15

Although more intricate and computationally demanding compared to 1D hydraulic models, 2D

hydraulic models offer more precise and realistic flood simulations. Hydrologic models are uti-

lized to compute the volume and timing of catchment runoff, simulating the precipitation-runoff

mechanisms that lead to flooding [23]. Hydrologic models can aid in the estimation of the volume

and timing of catchment runoff, and they can also contribute to the development of flood forecasts

and warnings [27]. Irrespective of the model or methodology utilized, the precision and reliability

of flood models hinge on the quality and accessibility of data, encompassing hydrologic, hydraulic,

meteorological, and land use data [28]. To develop accurate flood models, the availability and re-

liability of hydrologic data, including precipitation and stream flow data, are crucial. Additionally,

the models themselves can be complex and computationally intensive, necessitating the allocation

of significant computer resources [23].

2.4.1 Common Flood Simulation Models

2.4.1.1 HEC-RAS (Hydrologic Engineering Center’s River Analysis System)

HEC-RAS was developed by the Hydrologic Engineering Center,U.S. Army Corps of Engi-

neers River Analysis System. It facilitates one-dimensional steady flow and one/two-dimensional

unsteady flow calculations, as well as sediment transport and water temperature/water quality mod-

eling [29].

The development of the HEC-RAS modeling system was a part of the Hydrologic Engineering Cen-

ter’s comprehensive NexGen project, which encompasses various facets of hydrologic engineering.

These include rainfall-runoff analysis (HEC-HMS), river hydraulics (HEC-RAS), reservoir system

simulation (HEC-ResSim), flood damage analysis (HEC-FDA and HEC-FIA), and real-time river

forecasting for reservoir operations (CWMS).

HEC-RAS can be used for unsteady flows simulations. The Unsteady Flow Simulation component

of the HEC-RAS modeling system is equipped to simulate one-dimensional, two-dimensional, and

combined one/two-dimensional unsteady flow across a complete network of open channels, flood-

plains, and alluvial fans. This module can handle subcritical, supercritical, and mixed flow regimes,

encompassing calculations for hydraulic jumps, drawdowns, and both subcritical and supercritical

16

flows.

Moreover, the unsteady flow module incorporates the hydraulic computations for cross-sections,

bridges, culverts, and other hydraulic structures from the steady flow component. Notable features

of the unsteady flow component include the capacity for dam break analysis, levee breaching and

over topping, pumping stations, navigation dam operations, pressurized pipe systems, automated

calibration features, user-defined rules, and combined one and two-dimensional unsteady flow

modeling. There are potential limitations for this model :

1. HEC-RAS 1D, operates solely in one dimension, disregarding the interaction between the

river and the floodplain. Consequently, this approach may result in i naccuracies in floodplain

mapping and flood hazard assessment [30].

2. HEC-RAS 1D requires a significant amount of data input, which can be time- consuming.

The accuracy of the results is also dependent on the quality of the input data [31].

3. HEC-RAS 1D uses simplified assumptions about the river system, such as uniform flow and

steady-state conditions. These assumptions may not always reflect the actual conditions of

the river system, leading to inaccuracies in the results [30].

4. HEC-RAS 1D has limited accuracy in predicting the extent and depth of flooding in a given

area. The accuracy of the results is dependent on the quality of the input data and the

assumptions made in the model [31].

5. While HEC-RAS 1D can be coupled with 2D models to simulate the interaction between the

river and the floodplain, this coupling can be complex and time-consuming [32].

When utilizing HEC-RAS 1D for flood simulations, it’s essential to account for the software’s

inherent limitations. These constraints encompass the one-dimensional model, specific data input

prerequisites, simplified assumptions, restricted precision, and intricate coupling with 2D models.

Although HEC-RAS 1D is extensively employed and delivers precise outcomes, it remains crucial

17

to take these limitations into consideration during software application.

In contrast, the SRH-2D (Sedimentation and River Hydraulics-2D) model offers a more comprehen-

sive approach, accounting for the interaction between rivers and floodplains. This two-dimensional

model is capable of addressing many of the limitations associated with the HEC-RAS 1D model,

providing a more accurate depiction of complex flood dynamics.

2.4.2

SRH-2D (Sedimentation and River Hydraulics-2D

SRH-2D is an evolving two-dimensional (2D) model for river systems developed by the Bureau

of Reclamation [33]. It is focused specifically on 2D river system modeling for flow hydraulics.

SRH-2D v2 is applicable to a variety of scenarios, including but not limited to the following:

1. Modeling flow in one or multiple streams, encompassing the main channel, side channels,

and floodplains.

2. Facilitating flood routing and generating inundation maps across diverse terrains.

3. Analyzing flow patterns around in-stream structures like weirs, diversion dams, release gates,

and cofferdams.

4. Assessing flow over-banks and levees.

5. Evaluating the interaction between flow and vegetated areas, particularly in relation to main

channel flows.

6. Studying flow dynamics in reservoirs with known flow release.

7. Assessing the potential for bed erosion through morphological analysis.

2.4.2.1 Flexibility

SRH-2D employs a zonal approach for the coupled simulation of the main channels, side

channels, and floodplains. This involves the division of the river system into distinct modeling

zones, each capable of being assigned various parameters, such as the Manning’s roughness

coefficient, and being meshed differently(Figure 2.3a).

18

(a) Zonal Partition and Mesh Layout

(b) Hybrid Mesh

Figure 2.3 Flexible layout of SRH2D

Also, in its modeling process, a notable aspect is utilization of the hybrid mesh, which is derived

from arbitrarily shaped element method for geometric representation(Figure 2.3b). This flexible

unstructured hybrid meshing strategy facilitates the integration of the zonal modeling concept.

SRH-2D accommodates various existing meshing methods, including the structured curvilinear

19

mesh (pure quadrilaterals), conventional finite element mesh (purely triangles), Cartesian mesh

(purely rectangular or square mesh), and the hybrid mixed element mesh.

2.4.2.2 Capabilities

1. Solves 2D depth-averaged dynamic wave equations using the finite-volume numerical method,

including steady-state or unsteady flows with robust and efficient time integration using an

implicit scheme.

2. Utilizes an unstructured arbitrarily-shaped mesh, combining the structured quadrilateral mesh

and the purely triangular mesh, or both, for increased efficiency and accuracy.

3. Simulates all flow regimes, such as subcritical, transcritical, and supercritical flows, simul-

taneously without requiring special treatments.

4. Implements a robust wetting-drying algorithm seamlessly for enhanced functionality.

5. Provides a range of solved and output variables, including water surface elevation, water

depth, depth-averaged velocity, Froude number, bed shear stress, critical sediment diameter,

and sediment transport capacity.

2.4.2.3 Boundary Conditions

SRH2D allows various boundary conditions types

1. INLET-Q: This represents an upstream inlet boundary with subcritical flow. A flow discharge

is specified with the unit, either a constant positive value for steady state simulation or a

negative integer for unsteady flow.

2. EXIT-H: This boundary serves as a downstream exit boundary with subcritical flow. A

specified water surface elevation is used, either a constant positive value for steady state

simulation or a negative integer for unsteady flow.

3. EXIT-Q: As a downstream exit boundary, this type denotes an exit with a known discharge.

Similar to INLET-Q, it requires the input of discharge and unit.

20

4. INLET-SC: This signifies an upstream inlet boundary with supercritical flow. Both discharge

and water surface elevation are specified, with the unit identified accordingly.

5. EXIT-EX: This boundary type represents a downstream exit with supercritical flow. No

specific boundary conditions are necessary at this type of exit.

6. WALL: This indicates a solid wall boundary where the velocity is zero.

It is typically

employed for river banks and domain edges, whether wet or dry.

7. SYMMETRY: This boundary type symbolizes a symmetry boundary, operating as a slip wall

boundary without any explicit boundary conditions required.

2.4.2.4 Monitoring Output

Outputs can be monitored by using monitor lines and monitor points.

1. MONITOR LINE: It is an internal polyline that can be utilized for monitoring the total flow

discharge passing through it.

2. MONITOR POINT: It is a node which is utilized for monitoring the water surface elevation

at any point within the model domain.

2.4.2.5

2D Mesh

SRH-2D does not include a mesh generation program. Instead, it relies on third-party mesh

generation software. The arbitrarily-shaped mesh system allows for the use of any 2D mesh

generation program. Typically, a combination of quadrilaterals and triangles is the most prevalent

type of mesh used by SRH-2D. There are several programs to generate 2D mesh for later use in

SRH-2D. Surface-Water Modeling System(SMS) is the mesh generator software which is supported

by SRH-2D (section 2.4.2.6).

2.4.2.6 Surface-Water Modeling System (SMS)

SMS provides essential pre-processing and post-processing functionalities for open channel

flow hydraulic modeling.

It integrates several GIS software attributes, enabling efficient mesh

21

generation. SMS offers a three-dimensional visualization of results and provides various tools for

data manipulation, enhancing its versatility and compatibility with multiple models [34].

2.4.3 Review of Past River Modeling Studies

A multitude of scientific studies have been conducted, emphasizing a range of flood types,

such as coastal and riverine floods. These studies have explored various aspects of flooding,

encompassing hydrological and hydrodynamic processes. Moreover, a diversity of modeling

approaches, including both physically based and data-driven methods, have been a focal point of

investigation [35]. There have been efforts in the past to check the best scenarios under which best

results can be drawn using different flood propagation models.

2.4.3.1 Mesh Sensitivity

Understanding the complexities of mesh sensitivity has been paramount in achieving accurate

and reliable results in flood modeling. The evolution of Root Mean Square Error (RMSE) concern-

ing the refinement mesh, provides valuable insights into the relationship between mesh resolution

and modeling precision. Optimal modeling accuracy and and assurance of robust modeling out-

comes are dependent on mesh refinement.Finer mesh lead to accurate results [34]. Nonetheless,

the scale of the area under consideration for modeling, along with the limitations imposed by the

model regarding the number of elements, govern the degree to which a mesh can be refined.

2.4.3.2 Time Step

Decrease in the time step results in a reduction of the error between observed and simulated

values [34]. Finer time steps could significantly increase the computational time required to com-

plete the modeling process, potentially leading to longer processing times and delays in obtaining

results. It also demands higher computational resources, such as increased memory, processing

power, large data storage and could also lead to increased susceptibility to errors resulting from

minor fluctuations in the input data.

22

2.4.3.3 Topobathy Data

The fusion of lidar and bathymetry datasets results in the creation of a topobathy dataset, rep-

resented as a raster, which serves as a vital tool in geospatial analysis (figure2.4). This combined

Figure 2.4 Data Sets Combined to develop Topobathy Coverage

topobathy single surface model holds significant importance in the study and evaluation of various

ecological processes occurring in aquatic and near-shore regions. By providing a comprehen-

sive representation of the integrated topographic and bathymetric features, this dataset enabled

researchers to conduct in-depth analyses of critical environmental factors, including habitat dy-

namics, sediment distribution, and the interaction between land and water systems. The utilization

of the topobathy single surface model plays a crucial role in facilitating a holistic understanding of

the complex interactions within these environments, thus aiding in the effective management and

conservation of aquatic ecosystems. Same data holds equal importance for modeling of rivers and

water bodies for hydrological processes [36].

23

CHAPTER 3

RESEARCH GOALS, HYPOTHESIS AND SITE DESCRIPTION

3.1 Research Statement

This research attempts to advance the understanding of flood dynamics in the Middle Mississippi

River, with an emphasis on the 2019 flooding. SRH-2D model will be used which has frequently

been used for small domains and seldom for long river sections such as the one in the present

study. Specifically, the study aims to examine the implications of varying discharge and effects

of construction of flood protection structures such as levees on the flow. SRH-2D outputs will be

examined by varying different model inputs and parameters.

By assessing the interplay between these key factors, the research seeks to provide insights into the

complex mechanisms governing flood propagation and to enhance the accuracy of flood modeling

techniques in this crucial geographical region. Through a comprehensive analysis of the Middle

Mississippi River’s hydrological dynamics, this study aspires to contribute to the development of

more effective flood management strategies and the mitigation of potential risks associated with

inundation events.

3.2 Research Goals

The overarching goal of this project was to gain a better understanding of flood modeling and

how different scenarios influence the forecast. Guiding questions include

1. Is SRH-2D suitable for extended river reaches such as the Middle Mississippi River?

2. To what extent changes in mesh resolution and time step and Manning’s coefficient(s) impact

modeled results?

3. How do changes in discharge influence the flood inundation extent?

4. What are the implications of constructing flood protection structures on the river parameters?

24

3.3 Hypotheses

The Initial hypotheses are the following.

1. SRH-2D can model the extended river reaches.

2. Finer mesh resolution, smaller time steps and suitable Manning’s coefficient values would

give better results but at the cost of computational time and resources.

3. Increased discharge in the river will increase flood inundation area and vise versa.

4. Construction of levees would increase the velocity and the water surface elevation and the

effects would dominate downstream areas.

3.4 Site Description

The portion of the Mississippi River under study was the MMR flood plain, bounded by St

Louis, MO in north and Thebes in south,IL (Figure 3.1). An approximate 230 river km of the

MMR floodplain was spanned. The width of the natural floodplain along the MMR varied from

2.0 km to over 3 km. Within the study reach, the width of the channel (bank to bank) ranged from

430 to 1,160 m. The primary tributaries that fed into the MMR were the Meramec, Kaskaskia,

and Big Muddy Rivers, which were not included. Historically, the flood season for the MMR

basin lasted from March through July. The study reach covers 900 km2 across the floodplain of

seven Illinois and six Missouri counties, including a portion of the city of St. Louis. This area

encompasses all or portions of 7,510 census blocks. Census blocks are the smallest unit of census

geography for which population data are reported and are generally bounded by physical features

such as roads, creeks, railroads, or political boundaries. For these census blocks, the Hazards U.S.

Multi-Hazard (Hazus-MH) aggregate database contains approximately 74,900 buildings, with a

total estimated value of $12.8 billion. The total population within these blocks is estimated to be

164,000 inhabitants, in over 61,000 households. The occupancy of these 74,900 building consists

of 69% residential, 17% commercial, 9% industrial and 5% other [37](Figure 3.2).

25

Figure 3.1 Area of Study : Middle Mississippi River
(Source:https://www.mvs.usace.army.mil, Date visited:2023-11-02)

26

Figure 3.2 Estimated exposure values (structure and content values) in thousands of dollars per
census block

27

CHAPTER 4

METHODOLOGY

4.1 Data sets

4.1.1 Topobathy Data

The US Army Corps of Engineers (USACE), Upper Mississippi River Restoration (UMRR)

program gathered separate data for floodplain elevation and bathymetry on the Upper Mississippi

River System (UMRS). Although these data served various purposes individually, the need for

a consistent elevation surface spanning the river and its floodplain was evident. To meet this

demand, a unified elevation surface was created by integrating lidar-derived floodplain elevation

and bathymetry data. This integration process entailed specialized handling at the transition zones

between the two data sets [38]. Site specific data was downloaded from USGS. Details of the data

are summarized in Table 4.1.

Table 4.1 Raw Raster Properties

Property

Details

Projected Coordinate System NAD 1983 UTM Zone 15N

Projection
Linear Unit
Datum
Spheroid
Cell size ΔX
Cell size ΔY

Transverse Mercator
Meters
D North American 1983
GRS 1980
2m
2m

Sr No
1
2
3
4
5
6
7

4.1.1.1 Spatial Resolution

The available data, with a resolution of 2m x 2m, was too fine and could not be directly processed

in SMS due to memory constraints. Various topobathy datasets were obtained using ARCGIS and

were verified for accuracy. Considering the precision, limitations of the model, and computational

resources, a resolution of 15m x 15m was selected for the modeling purpose (Figure 4.1).

28

Figure 4.1 Topobathy Data for MMR

29

4.1.2 River Forcing Data

River forcing data for the study was obtained from the U.S. Geological Survey (USGS) for three

key locations within the model domain (Figure 4.2).

1. St. Louis, Missouri (Northern Boundary).

2. Chester, Illinois (Central Location).

3. Thebes, Illinois (Southern Boundary).

The downloaded data include information on discharge and gauge height for each of these stations.

The data sets for the northern boundary, representing discharge, and the southern boundary, rep-

Figure 4.2 USGS Gauging stations in study reach

resenting water surface elevation, were employed as crucial boundary conditions for the model.

30

Additionally, data sets for discharge and water surface elevation from the central location at Chester,

IL, water surface elevation at St. Louis and discharge at Thebes were utilized to validate the results.

4.1.3 Gauge Height Datum Adjustment

The datum values provided by USGS for gauge height measurements need to be acknowledged.

At the Chester station, USGS used 103.85 meters as the datum for gauge height, while the corre-

sponding average river depth bathymetry data was measured at 104.2 meters. Similarly, at the St.

Louis station, a USGS datum of 115.69 meters was documented for gauge height, and the river

depth obtained from bathymetric data was 116.59 meters. Variations in datum values of gauge

heights and river depths could potentially impact and influence the outcomes of modeling efforts.

To accommodate these differences and avoid inaccuracy in results, the disparity was incorporated

by adding the difference of datum to the observed values making a common datum, i.e., average

river bed elevation.

4.1.4 Flood Inundation and Satellite Imagery

In this study, satellite imagery with a resolution of 3 meters per pixel is acquired from www.

planet.com. The imagery, selected for its temporal relevance to the flood event under investigation,

is processed using a straightforward approach. Utilizing Microsoft PowerPoint, flood inundation

areas are manually marked on the satellite images. This process involves visually inspecting the

imagery and identifying regions affected by flooding. The simplicity and accessibility of PowerPoint

make it an efficient tool for this task, allowing for the delineation of flood extents without the need

for complex image processing software. The marked areas are then used for further analysis and

comparison with hydrological model outputs. This methodology offers a practical and user-friendly

approach to visually identifying flood inundation, contributing to the overall assessment of the study

area’s hydrological dynamics.

4.1.5 River shape file

River Shape file was prepared using SMS map and GIS modules (Figure 4.3).

31

Figure 4.3 Shape file of the study reach

4.2 Theory

4.2.1 Governing Equations

In addition to the model theory discussed in section 2.4.2, SRH2D uses 2D Saint-Venant

equations 4.1, 4.2, 4.3, which are obtained after the Navier-Strokes is vertically averaged.

𝜕ℎ
𝜕𝑡

+

𝜕ℎ𝑈
𝜕𝑥

+

𝜕ℎ𝑉
𝜕𝑦

= 𝑒

(4.1)

𝜕ℎ𝑈
𝜕𝑡

+

𝜕ℎ𝑈𝑈
𝜕𝑥

+

𝜕ℎ𝑉𝑈
𝜕𝑦

𝜕ℎ𝑇𝑥𝑥
𝜕𝑥

𝜕ℎ𝑇𝑥𝑦
𝜕𝑦

+

=

− 𝑔ℎ

𝜕𝑧
𝜕𝑥

−

𝜏𝑏𝑥
𝜌

+ 𝐷𝑥𝑥 + 𝐷𝑥𝑦

(4.2)

𝜕ℎ𝑉
𝜕𝑡

+

𝜕ℎ𝑈𝑉
𝜕𝑥

+

𝜕ℎ𝑉𝑉
𝜕𝑦

𝜕ℎ𝑇𝑥𝑦
𝜕𝑥

𝜕ℎ𝑇𝑦𝑦
𝜕𝑦

+

=

− 𝑔ℎ

𝜕𝑧
𝜕𝑥

𝜏𝑏𝑦
𝜌

−

+ 𝐷 𝑦𝑥 + 𝐷 𝑦𝑦

(4.3)

32

The friction is determined using the Manning equation 4.4 and 4.5

(cid:0)𝜏𝑏𝑥, 𝜏𝑏𝑦(cid:1) = 𝜌𝐶 𝑓 (𝑈, 𝑉)

√︁𝑈2 + 𝑉 2

𝐶 𝑓 =

𝑔𝑛2
ℎ1/3

(4.4)

(4.5)

The Boussinesq equations 4.6, 4.7, 4.8 are used to compute the depth-averaged turbulence

stresses:

𝑇𝑥𝑥 = 2(𝜈 + 𝜈𝑡)

(cid:19)

(cid:18) 𝜕𝑈
𝜕𝑥

𝑘

−

2
3
(cid:18) 𝜕𝑉
𝜕𝑥

(cid:19)(cid:21)

−

𝑘

2
3

(4.6)

(4.7)

(4.8)

(cid:19)

(cid:20) (cid:18) 𝜕𝑈
𝜕𝑦
(cid:18) 𝜕𝑉
𝜕𝑦

+

(cid:19)

𝑇𝑥𝑦 = (𝜈 + 𝜈𝑡)

𝑇𝑦𝑦 = 2(𝜈 + 𝜈𝑡)

where 𝑡 is time, 𝑥 and 𝑦 are the horizontal Cartesian coordinates, ℎ is the water depth, 𝑈 and 𝑉

are depth-averaged velocity components in the 𝑥 and 𝑦 directions respectively, 𝑒 is excess rainfall

rate, 𝑔 is gravitational acceleration, 𝑇𝑥𝑥,𝑇𝑥𝑦 and 𝑇𝑦𝑦 are depth-averaged turbulent stresses, 𝐷𝑥𝑥 and

𝐷 𝑦𝑦 are dispersion terms due to depth averaging, 𝑧 = 𝑧𝑏 + ℎ is water surface elevation, 𝑧𝑏 is bed

elevation, 𝜌 is water density, 𝜈 is the kinematic viscosity of water, 𝜈𝑡 is the eddy viscosity computed

using the turbulence models described in the next section and 𝜏𝑏𝑥, 𝜏𝑏𝑦 are the bed shear stresses

(friction). Bed friction is calculated using Manning’s roughness equation as shown in equation

(4.5). The above equations ignore the contribution of wind stress at the top of the water column

and groundwater contributions are also assumed to be negligible.

4.2.2 Turbulence Models

Interaction of channel flows with complex bathymetry generates turbulence and it is important to

include the effects of turbulence in both hydrodynamic and transport simulations. SRH-2D includes

two turbulence models - an algebraic parabolic model and a two equation 𝑘-𝜀 turbulence model.

In the parabolic model, the eddy viscosity 𝜈𝑡 is directly calculated as:

𝜈𝑡 = 𝐶𝑡𝑈∗ℎ

(4.9)

33

where 𝐶𝑡 is a model constant with a range of values from 0.3 to 1.0 with a default value of 0.7 and

𝑈∗ = 𝐶1/2

𝑓

(cid:16)√︁𝑈2 + 𝑉 2(cid:17)

For the two-equation 𝑘-𝜀 model, the eddy viscosity is calculated as:

𝜈𝑡 = 𝐶𝜇

𝑘 2
𝜀

(4.10)

(4.11)

where the two additional variables for turbulent kinetic energy (𝑘) and turbulent energy dissipation

(𝜀) are computed using the two partial differential equations below.

𝜕 (ℎ𝑘)
𝜕𝑡

+

𝜕 (ℎ𝑈𝑘)
𝜕𝑥

+

𝜕 (ℎ𝑉 𝑘)
𝜕𝑦

=

𝜕
𝜕𝑥

(cid:19)

(cid:18) ℎ𝜈𝑡
𝜎𝑘

𝜕𝑘
𝜕𝑥

+

𝜕
𝜕𝑦

(cid:19)

(cid:18) ℎ𝜈𝑡
𝜎𝑘

𝜕𝑘
𝜕𝑦

+ 𝑃ℎ + 𝑃𝑘 𝑏 − ℎ𝜀

(4.12)

𝜕 (ℎ𝜀)
𝜕𝑡

+

𝜕 (ℎ𝑈𝜀)
𝜕𝑥

+

𝜕 (ℎ𝑉 𝜀)
𝜕𝑦

=

𝜕
𝜕𝑥

(cid:18) ℎ𝜈𝑡
𝜎𝜀

(cid:19)

𝜕𝜀
𝜕𝑥

+

𝜕
𝜕𝑦

(cid:18) ℎ𝜈𝑡
𝜎𝜀

(cid:19)

𝜕𝜀
𝜕𝑦

+

𝐶𝜀1𝜀
𝑘

𝑃ℎ + 𝑃𝜀𝑏 − 𝐶𝜀2ℎ

𝜀2
𝑘

(4.13)

Following the recommendations of Rodi [39], the following relations and constants are used in the

model.

𝑃ℎ = ℎ𝜈𝑡

(cid:19) 2

(cid:34)

(cid:18) 𝜕𝑈
𝜕𝑥

2

(cid:19) 2

(cid:18) 𝜕𝑉
𝜕𝑦

(cid:18) 𝜕𝑈
𝜕𝑦

+

𝜕𝑉
𝜕𝑥

+

+ 2

(cid:19) 2(cid:35)

, 𝑃𝑘 𝑏 = 𝐶−1/2

𝑓

(𝑈∗)3

𝑃𝜀𝑏 = 𝐶𝜀Γ𝐶𝜀2𝐶−1/2

𝜇 𝐶−3/4
𝑓

(𝑈∗)4

𝐶𝜇 = 0.09, 𝐶𝜀1 = 1.44, 𝐶𝜀2 = 1.92, 𝜎𝑘 = 1.0, 𝜎𝜀 = 1.3, 𝐶𝜀Γ = 1.8 − 3.6

A comparison of the parabolic and the 𝑘-𝜀 turbulence models is included in a later section.

(4.14)

(4.15)

(4.16)

4.2.3 Model Setup

For model setup and run, all the data sets projections were converted to WGS 1984 UTM Zone

15N, with NAVD 88 (meters) for vertical datum.

4.2.3.1 Mesh Generation

SMS was used to generate mesh using the shape file. Keeping in mind the limitations of SRH2D

and efficiency of model, a mesh of 120m to 80m node to node distance was generated (Figure 4.4).

34

Figure 4.4 Fine Hybrid Mesh

35

4.2.3.2 Mesh Quality

The computational mesh was checked for the mesh quality with element quality checks (table

4.2). To ensure smooth model run and to minimize the errors in calculations, all the errors found

Table 4.2 Mesh Element Quality Checks

Sr No
1
2
3
4
5
6
6

Check Type
Minimum interior angle
Maximum interior angle
Concave quadrilaterals
Minimum slope
Element area change
Connecting elements
Ambiguous gradient

Value
10
130
NA
0.1
0.5
8
NA

outside the range specified in table 4.2 were removed by using refinement tools in the SMS mesh

module.

4.2.3.3 Topobathy Interpolation

The 15m resolution topobathy raster (section 4.1.1) was converted to scatter points using the

GIS module in SMS. The scatter set was then interpolated to the mesh using linear interpolation

method and inverse distance weighted for extrapolation. Nearest eight points were assigned for

computation of interpolation weights (Figure 4.5).

4.2.3.4 Boundary Conditions

As discussed in detail in section 2.4.2.3, upper boundary condition and lower boundary condi-

tions were assigned to the model.

1. Upper Boundary Condition (INLET-Q): Upstream inlet boundary at St. Louis, MO with

subcritical flow in cubic meters per second was assigned. The input data covered an hourly

time series for one year . Hourly data for discharge was downloaded from USGS (https:

//waterdata.usgs.gov). The downloaded data was sifted through and discrepancies for time

interval and missing data were sorted out by interpolating the data using Matlab.

2. Lower Boundary Condition (EXIT-H): Downstream exit boundary at Thebes, IL water surface

36

Figure 4.5 Topobathy Interpolated to Mesh

elevation in meters was assigned. The input data covered an hourly time series vs water surface

elevation(m) for one year. Hourly data was downloaded from USGS (https://waterdata.usgs.

gov). The downloaded data was sifted through and discrepancies for time interval and missing

data were sorted out by interpolating the data using Matlab. Both boundary conditions data

were further converted to files with .𝑥𝑦𝑠 extension in order to upload them to SMS.

3. Boundary Condition For Small inlets: Only flow from the Kaskaskia River, which enters

Mississippi River near Chester was included in the model.

37

4.2.3.5 Monitor points

In order to monitor the outputs, monitor arcs were created to generate discharge data at Chester

and Thebes. Also, monitor points were inserted to output water surface elevation data at Chester

and St. Louis.

4.2.3.6 Material Properties

Material properties were assigned to river and flood plain areas. Different sets of Manning’s

roughness values were used in order to address the research question (Table 4.3). All these changes

were used in combination with finer mesh only. The Manning’s roughness values are based on the

Table 4.3 Materials and Manning’s Roughness Sets

Case Number Scenario Material

1

2

Medium

High

River
Flood Plain
River
Flood Plain

Value
(Manning’s Roughness) (n)
0.03
0.05
0.033
0.09

one-dimensional retro modelling study for the same study area reported in Remo et al. [37]. These

values served as the basis of the initial guess; however, the model was run for different values of 𝑛

to obtain an optimum value as these values are subject to spatial changes.

4.2.3.7 SHR2D simulations in SMS

SMS supports execution of SRH2D simulations. When all the required data for each scenario

was ready, new simulation was created in SMS for SRH2D model. Once that was done, following

were linked with the simulations :

1. Study area mesh.

2. Material properties.

3. Boundary conditions.

4. Monitor points and arcs.

38

4.2.3.8 Model Control

Model controls used for the simulations are summarized in Table 4.4.

Table 4.4 Model Controls

General

Flow Module

Result Outputs

Time Step
Initial Condition
Turbulence Model
Constant Value
Unit
Specified Frequency
Format

15s
Automatic
Parabolic Turbulence
0.7
SI
1 hr
XMDf

4.3 Post Processing

4.3.1 Data Analysis

The results obtained from the simulation were post-processed using Matlab. The performance

of the model was assessed by calculating several metrics (described below) to compare simulated

values against observed values.

4.3.1.1 Root Mean Squared Error (RMSE)

Root Mean Squared Error is a measure of the average magnitude of the errors between predicted

and observed values. It is calculated using the following equation:

𝑅𝑀𝑆𝐸 =

(cid:118)(cid:116)

1
𝑛

𝑛
∑︁

𝑖=1

(𝑂𝑖 − 𝑆𝑖)2

where 𝑛 is the number of data points (𝑖 = 1, 2, 3, · · · , 𝑛) and 𝑂𝑖 and 𝑆𝑖 denote the observed and

simulated values respectively. This metric provides an indication of the model’s accuracy, with

lower RMSE values indicating better performance.

4.3.1.2 Mean Absolute Error (MAE)

Mean Absolute Error represents the average absolute difference between predicted and observed

values. The equation for MAE is given by:

𝑀 𝐴𝐸 =

|𝑂𝑖 − 𝑆𝑖 |

1
𝑛

𝑛
∑︁

𝑖=1

39

Similar to the RMSE, lower MAE values signify better agreement between the model and observed

data.

4.3.1.3 R-squared (Coefficient of Determination)

R-squared measures the proportion of the variance in the dependent variable that is predictable

from the independent variable. The equation for R-squared is as follows:

𝑅2 = 1 −

(cid:205)𝑛
𝑖=1(𝑂𝑖 − 𝑆𝑖)2
(cid:205)𝑛
𝑖=1(𝑂𝑖 − ¯𝑂𝑖)2

where ¯𝑂𝑖 is the mean of the observed values. A higher R-squared value indicates a better fit of the

model to the observed data.

These metrics collectively provide a comprehensive evaluation of the model’s performance in

comparison to observed values.

4.3.2

Flood Severity Categorization

Flood severity was categorized into different stages based on observed water levels. These

categories serve as a framework for assessing the potential impacts on both the environment and

human activities (www.weather.gov/lot/hydrology_definitions).

4.3.2.1 Action Stage

The water levels may reach a point where they pose a potential risk, capable of causing minor

impacts and inconveniences. Local authorities may implement proactive measures to mitigate

damage and ensure public safety in response to the elevated water levels. It is a stage where caution

is advised, emphasizing the importance of monitoring and addressing the situation to minimize

adverse effects

4.3.2.2 Minor Flood Stage

A potential escalation where property flooding becomes a concern, posing a threat to public

safety. During this stage, there is a risk of flooding affecting various areas, including roadways,

trails, park lands, and private properties. It underscores the need for prompt action and heightened

awareness to address the emerging challenges and safeguard both public and private spaces.

40

4.3.2.3 Moderate Flood Stage

This Stage introduces a heightened level of concern as flooding extends to structures and main

roadways. This phase may necessitate evacuations to ensure the safety of residents and mitigate

potential risks. The resulting disruptions to daily life underscore the significance of strategic

planning and coordinated response efforts to address the increasing challenges posed by rising

water levels.

4.3.2.4 Major Flood Stage

Major Flood Stage represents a critical juncture, marked by extensive flooding affecting struc-

tures, main roadways, and essential infrastructure. The severity of the situation often warrants

evacuations, as there is a heightened risk to public safety. The ensuing disruptions to daily life are

expected to be significant, underscoring the urgent need for comprehensive emergency response

measures and community preparedness in the face of this formidable challenge

41

CHAPTER 5

RESULTS AND DISCUSSION

The model was executed under various conditions, each designated with a specific case number.

Detailed information on each run, including the model parameters and results, is described within

each case. Table 5.1 provides a brief overview of each case. As noted in an earlier chapter, the

inflow boundary condition was specified at St. Louis, MO in the form of 𝑄(𝑡) versus time while

the outflow boundary condition was specified at Thebes, IL in the form of water surface elevation

versus time. Hence independent data from the USGS gage at Chester, IL is used to test the model.

Case Number
1, 2, 3

4

5

6

7 , 8

9

10

Table 5.1 Overview of different model scenarios

Year(s)

Investigation or Topic

2019

2019

2019, 2020, 2018 Temporal consistency in model performance,
complex interplay of factors, impact of resolution
on model accuracy, gauge height and discharge
trends, model adaptability and refinement
Sensitivity of model results to Manning’s rough-
ness, Variable impact on discharge, Water Surface
Elevation accuracy
Effects of interpolation methods, Superiority of
IDW in discharge prediction, Sensitivity to in-
terpolation in water surface elevation, Location-
Specific impacts
Impact of topobathy resolution, Sensitivity to
topobathy resolution, trade-offs between resolu-
tion and accuracy, considerations for model se-
lection
Impact of reduced/increased discharge on hydro-
logical dynamics, gauge height responses, model-
simulated flood inundation extent
Effects of levee construction on hydrological be-
havior, impact on discharge and gauge Height,
model-simulated flood inundation extent
Effects of Turbulence Models

2019

2019

2019

2019

42

5.1 Case 1 : 2019 River Model (Base Case Scenario)

Table 5.2 provides an overview of the scenario and associated parameters.

Table 5.2 Case 1: Model Parameters

Detail
Topo-Bathy Resolution
Year
Mesh Size
Manning’s Roughness (𝑛)

Value
15m
2019
80m x 120m (approx)
River - 0.033
Flood Plain - 0.09

5.1.1 Comparative Analysis of Discharge and Gauge Height Data

To understand the dynamics of the flooding event, two primary datasets were considered: the

discharge-time relationship and the gauge height-time relationship. An analysis of flood categories,

including action, minor, moderate, and major flood stages, was conducted based on the gauge height

plot for each station based on the data provided by USGS (waterdata.usgs.gov). This assessment

was crucial as the observed discharge data relied on the rating curve, which, in turn, was derived

from the gauge height data.

5.1.1.1 Chester Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table 5.3

and Figure 5.1.

Table 5.3 Results for Chester Gauge Height (2019)

Metric
MAE
RMSE
R-squared

Value
0.3234 (m)
0.3844 (m)
0.9784

The obtained metrics indicate promising results for the model evaluation. With an MAE

of 0.3234 meters, an RMSE of 0.3844 meters, and an R-squared value of 0.9784, the model

demonstrates a high level of accuracy and a strong correlation between the predicted and observed

values. These values suggest that the model’s predictions are in close agreement with the actual

data, validating its efficacy in capturing the complex dynamics of the system under study. Further

43

Figure 5.1 Chester Gauge Height (2019)

examination of the Figure 5.1 reveals a close alignment between the model predictions and the

observed values from January until late March. During this period, the flood stage remained within

the range of normal to minor, indicating that the model accurately captures the nuances of the

system during relatively stable conditions. This alignment reinforces the model’s capability to

represent the intricate dynamics of the system under study, particularly during periods of relatively

lower flood intensity."

However, an evident deviation begins to manifest as the flood stage progresses into the moderate and

major categories, particularly during the first half of June, where the maximum deviation reached

0.6 meters. This discrepancy suggests that the model’s performance may be more variable during

periods of intensified flooding, indicating the need for further investigation into the underlying

factors influencing the model’s predictive capacity during such extreme events.

This phenomenon can be attributed to the increasing complexity of hydrological processes during

periods of heightened flood intensity, including the interplay of various factors such as increased

rainfall, river discharge, and potential dynamic changes in the landscape. Furthermore, the observed

44

deviation may be influenced by the simplification of certain hydrological parameters, such as the

assumption of a uniform Manning’s 𝑛 value, and the lack of consideration for small tributaries

that contribute to the river’s flow. Moreover, the model’s predictive accuracy might be affected by

the resolution of the mesh and the topobathymetric data, potentially leading to an underestimation

or overestimation of the flood extent and intensity, particularly in areas of complex topographical

variability.

After the first half of July, the flood stage fluctuates between minor action and normal stages,

exhibiting a notably strong correspondence between the observed and simulated data until the end

of December. This correspondence underscores the model’s robustness in accurately simulating

flood dynamics during periods of relatively stabilized water levels, reaffirming its efficacy in

capturing the system behavior during less extreme flood events.

5.1.1.2 Chester Discharge Data

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.4

and Figure 5.2. The analysis of the Chester discharge revealed an MAE of 752.5864 cubic meters

Table 5.4 Chester Discharge Results (2019)

Metric
MAE
RMSE
R-squared

Value
752.5864 (m3/s)
1088.9275 (m3/s)
0.9582

per second (cms) and an RMSE of 1088.9275 cms, within the context of the discharge ranging

between 10,000 and 30,000 units. While these absolute errors might appear relatively high, the

R-squared value of 0.9582 indicates a robust overall fit. This implies that the model effectively

captures the dynamics of the discharge, despite the minor discrepancies. The consistency exhibited

by the R-squared value suggests that the model adequately explains the majority of the variability

in the observed discharge data, reaffirming its reliability in representing the complex hydrological

processes within the system. Figure 5.2 illustrates the trend observed in the Chester discharge, which

closely mirrors the dynamics observed in the corresponding gauge height data. It has depicted a

consistent pattern over the recorded time frame, further reinforcing the model’s capability to

45

Figure 5.2 Chester Discharge Results (2019)

capture the intricate relationships between the fluctuating gauge height and the resulting discharge

dynamics. This close correspondence between the observed trends in both the gauge height and

discharge data provides additional evidence of the model’s efficacy in simulating the complex

hydrological processes within the studied system. Further investigations into the complexities of

the river’s flow regime, considering factors such as variations in channel geometry and the impact of

lateral inflows, are crucial to enhance the model’s accuracy in replicating the complex hydrological

behavior.

5.1.1.3 St. Louis Gauge Height

At the St. Louis station, where the discharge served as the boundary condition for the model

inputs, the measured gauge height serves as one of the primary model outputs. The analysis

yielded the following metrics for the St. Louis gauge height, as shown in Table 5.5 and Figure 5.3

The obtained metrics indicate a strong performance of the model at the St. Louis station. With

an MAE of 0.2923 meters, an RMSE of 0.4075 meters, and an R-squared value of 0.9793, the

model exhibits a robust ability to accurately represent the complex hydrological processes at this

46

Table 5.5 St. Louis Gauge Height Results (2019)

Metric
MAE
RMSE
R-squared

Value
0.2923 (m)
0.4075 (m)
0.9793

location. These values suggest that the model effectively captures the complex dynamics of the St.

Louis hydrological system, providing valuable insights into the reliability of the model’s predictive

capabilities. Figure 5.3 depicts the trends observed in the St. Louis gauge height, illustrating

Figure 5.3 St Louis Gauge Height (2019)

a consistent pattern in the normal, action, and moderate stages, with a maximum deviation of

0.3 meters observed during the moderate and major stages. Despite this disparity, the model

demonstrated a strong adherence to the anticipated trend, signifying the model’s ability to capture

the expected variations in the gauge height within the St. Louis area across different flood stages.

The observed variations in the St. Louis gauge height, particularly during the moderate and major

stages, can be attributed to similar factors identified during the analysis of the Chester gauge height.

47

5.1.1.4 Thebes Discharge Data

The examination produced the subsequent metrics for the Thebes discharge, as depicted in Table

5.6 and Figure 5.4. The examination of the Thebes discharge data revealed an MAE of 1015.4889

Table 5.6 Thebes Discharge Results (2019)

Metric
MAE
RMSE
R-squared

Value
1015.4889 (m3/s)
1276.4081 (m3/s)
0.9419

cubic meters per second (cms), an RMSE of 1276.4081 cms, and an R-squared value of 0.9419.

These metrics indicate a strong performance of the model in simulating the discharge dynamics

at the Thebes station, underscoring its capability to accurately represent the complex hydrological

processes within this particular location. Despite minor deviations, the high level of agreement

between the model predictions and the observed values suggests the model’s efficacy in capturing

the complexities of the discharge behavior at the Thebes station. Figure 5.4 showcases the trend

Figure 5.4 Thebes Discharge (2019)

48

observed in the Thebes discharge, closely resembling the dynamics identified in the corresponding

gauge height data.

It exhibits a consistent pattern throughout the observed time frame, further

affirming the model’s proficiency in capturing the complex associations between the fluctuating

gauge height and the resulting discharge dynamics. This notable alignment between the observed

trends in both the gauge height and discharge data provides additional support for the model’s

effectiveness in simulating the complex hydrological processes within the Thebes station.

5.1.2 Comparative Analysis of Model-Simulated Flood Inundation and Satellite Imagery

Two distinct representations of the flood inundation extent in St. Louis for 03 July 2019 were

evaluated, one from satellite imagery (Figure 5.5a) and the other generated by the model (Figure

5.5a). A comparison between these two datasets reveals a remarkable similarity, indicating the

model’s proficiency in simulating the flood dynamics at the specific time period. The visual

congruence between the observed aerial imagery and the model’s simulated inundation extent

provides substantial evidence of the model’s capability to accurately represent the complex interplay

of hydrological processes contributing to flood inundation events. These findings underscore the

model’s utility as a valuable tool for assessing and predicting flood risk in the St. Louis region,

thereby facilitating informed decision-making in flood management and disaster preparedness.

The comprehensive assessment and analysis of the model’s performance across multiple stations

and datasets have yielded valuable insights into its efficacy in simulating the complex dynamics of

flood inundation. The model’s consistent and close alignment with observed data at the Chester,

St. Louis, and Thebes stations, as indicated by the calculated MAE, RMSE, and R-squared values,

underscores its reliability in capturing the complex hydrological processes within each respective

region. Notably, the model’s proficient replication of varying flood stages and its ability to closely

match the observed trends in both gauge height and discharge data highlights its robustness in

representing the interactions within the hydrological systems.

Furthermore, the successful comparison between the model’s simulated flood inundation extent and

the observed aerial imagery reinforces its capability to accurately predict the spatial distribution

and extent of flood events. The congruence between the model’s predictions and the actual aerial

49

(a) Satellite Imagery of Flood Inundation

(b) Model Output of Flood Inundation

Figure 5.5 Flood Inundation Comparison at Arnold-St. Louis (03 July 2019)

observations signifies its potential as a valuable tool for assessing and mitigating flood risks in the

study areas. The identified minor deviations in certain instances emphasize the ongoing need for

continued refinement, incorporating finer-scale topographical data, and enhancing the calibration

techniques to further improve the model’s predictive accuracy.

50

5.2 Case 2 : 2020 River Model

Table 5.7 provides an overview of the case scenarios and associated parameters .

Table 5.7 Case 2: Model Parameters

Detail
Topo-Bathy Resolution
Year
Mesh Size
Manning’s Roughness (𝑛)

Value
15m
2020
80m x 120m (approx)
River - 0.033
Flood Plain - 0.09

5.2.1 Comparative Analysis of Discharge and Gauge Height Data

To understand the dynamics of the flooding event in Case 1, two primary datasets were consid-

ered: the discharge-time relationship and the gauge height-time relationship. An analysis of flood

categories, including action, minor, moderate, and major flood stages, was conducted based on the

gauge height plot. This assessment was crucial as the observed discharge data relied on the rating

curve, which, in turn, was derived from the gauge height data.

For Case 2, a similar approach was employed to analyze the discharge and gauge height data. The

methodology followed closely mirrored that of Case 1, providing consistency in the comparative

analysis. This ensures that the findings from both cases are directly comparable and facilitates a

comprehensive understanding of the flood dynamics over different time periods. For a detailed

discussion on the methodology, please refer to Section 5.1.1 for Case 1.

5.2.1.1 Chester Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table 5.8

and Figure 5.6.

Table 5.8 Results for Chester Gauge Height (2020)

Metric
MAE
RMSE
R-squared

Value
0.3657 (m)
0.4131 (m)
0.9781

The evaluation metrics for 2020 indicate continued promising results. The MAE of 0.3657

51

meters and RMSE of 0.4131 meters demonstrate the model’s consistency in accurately representing

the observed gauge heights. The R-squared value of 0.9781 reaffirms a strong correlation between

the model predictions and actual data.

Comparing these metrics with the results from the previous year (2019), we observe an increase

in MAE and RMSE. Despite this deviation, the high R-squared value signifies the model’s robustness

in capturing the complex hydrological dynamics at the Chester station during the year 2020. During

Figure 5.6 Chester Gauge Height (2020)

this period, the flood stage remained within the range of normal to minor, showcasing the model’s

persistent ability to accurately capture the nuances of the system under study during relatively stable

conditions. The simulated results begin to deviate from observed when water surface elevation is

approaching moderate flood stage.

This phenomenon in 2020 can be attributed to similar factors mentioned for 2019, including

the complex interplay of increased rainfall, river discharge, and potential dynamic changes in the

landscape. Additionally, the observed deviation may be influenced by the simplification of certain

hydrological parameters, such as the assumption of a uniform Manning’s 𝑛 value, and the exclusion

52

of small tributaries contributing to the river’s flow. The model’s predictive accuracy might also

be affected by the resolution of the mesh and the topobathymetric data, potentially leading to an

underestimation or overestimation of flood extent and intensity, especially in areas of complex

topographical variability.

5.2.1.2 Chester Discharge Data

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.9

and Figure 5.7. The evaluation metrics for Chester discharge data in 2020 indicate positive trends

Table 5.9 Chester Discharge Results (2020)

Metric
MAE
RMSE
R-squared

Value
563.3757 (m3/s)
693.1294 (m3/s)
0.9685

in comparison to the results obtained in the previous year (2019). The Mean Absolute Error

(MAE) of 563.3757 m3/s and Root Mean Squared Error (RMSE) of 693.1294 m3/s reflect [mention

any changes in the model’s accuracy or deviation]. The R-squared value of 0.9685 signifies a

[strong/adequate] correlation between the model’s predicted and observed discharge values.

The decrease in MAE and RMSE suggests improvement in the model’s performance in repre-

senting the Chester discharge dynamics for the year 2020. The high R-squared value indicates that

the model captures a substantial portion of the variability in the observed discharge data.

Further investigation into the positive trends observed in these metrics can provide valuable

insights into the model’s performance and guide potential refinements. Additionally, a comparative

analysis with the 2019 results, as presented in Figure 5.7, can offer a nuanced understanding of

the model’s behavior over different temporal conditions. Figure 5.7 illustrates the trend observed

in the Chester discharge for the year 2020, offering insights into the dynamics that closely mirror

the corresponding gauge height data. The figure depicts a consistent pattern over the recorded

time frame, reinforcing the model’s ability to capture the intricate relationships between fluctuating

gauge height and resulting discharge dynamics. This alignment provides substantial evidence of

the model’s efficacy in simulating the complex hydrological processes within the studied system.

53

Figure 5.7 Chester Discharge (2020)

The close correspondence between the observed trends in gauge height and discharge data

for 2020 builds upon the insights gained from the analysis of 2019 data (see Figure 5.7). As

observed, [mention any similarities or changes in trends]. This consistency across different temporal

conditions underscores the model’s robustness in representing the interconnected dynamics of river

systems.

To enhance the model’s accuracy in replicating complex hydrological behavior, continued

investigations into the river’s flow regime are essential. Factors such as variations in channel

geometry and the impact of lateral inflows should be further explored. This ongoing research will

contribute to refining the model and advancing its predictive capabilities, ensuring a comprehensive

understanding of the hydrological processes at play.

5.2.1.3 St. Louis Gauge Height

The analysis of St. Louis gauge height data for the year 2020, where discharge served as the

boundary condition for the model inputs, is summarized in Table 5.10 and Figure 5.8.

The obtained metrics for 2020 reaffirm the model’s strong performance at the St. Louis station.

54

Table 5.10 Results for St. Louis Gauge Height (2020)

Metric
MAE
RMSE
R-squared

Value
0.1436 meters
0.1641 meters
0.9967

With a significantly reduced MAE of 0.1436 meters, a smaller RMSE of 0.1641 meters, and

an exceptionally high R-squared value of 0.9967, the model demonstrates an enhanced ability to

accurately represent the hydrological processes at this location.

This notable improvement in performance, as compared to the metrics from 2019 (see Table

5.5 and Figure 5.3), underscores the model’s adaptability and continued refinement. The high

R-squared value suggests that the model effectively captures an overwhelming majority of the

variability in the observed gauge height data, providing valuable insights into the reliability and

robustness of the model’s predictive capabilities.

This comparative analysis between 2019 and 2020 metrics enriches our understanding of the

model’s performance across different temporal conditions, further supporting the credibility of its

predictive capabilities for the St. Louis hydrological system.

Figure 5.8 presents the trends observed in the St. Louis gauge height for the year 2020. This

visualization reveals a consistent pattern across normal, action, and moderate stages, maintaining

alignment with the model’s predictions. Notably, the model exhibits a strong adherence to the

expected variations in gauge height within the St. Louis area during different flood stages. The

trends observed in Figure 5.8 for the St. Louis gauge height in 2020 closely mirror the patterns

identified in observed gauge height. The model consistently captures the expected variations across

all flood stages maintaining a strong adherence to the anticipated trends. This continuity in the

model’s performance across different temporal conditions reaffirms its reliability in simulating the

complex hydrological dynamics of the St. Louis area. The similarities in trends between 2019

and 2020 further underscore the model’s consistency and effectiveness in representing the intricate

interplay of factors influencing gauge height variations.

55

Figure 5.8 St. Louis Gauge Height (2020)

5.2.1.4 Thebes Discharge Data

The examination of Thebes discharge data for the year 2020 resulted in the following metrics,

as shown in Table 5.11 and Figure 5.9.

Table 5.11 Thebes Discharge Results (2020)

Metric
MAE
RMSE
R-squared

Value
824.5657 (cms)
1102.3450 (cms)
0.9271

The metrics for 2020 indicate a notable performance of the model in simulating the discharge

dynamics at the Thebes station. Despite minor deviations, the high level of agreement between

the model predictions and the observed values suggests the model’s efficacy in capturing the

complexities of discharge behavior at the Thebes station, building upon the insights gained from

the analysis of 2019 data (see Figure 5.4).

Figure 5.9 showcases the trend observed in Thebes discharge for the year 2020, closely re-

sembling the dynamics identified in the corresponding 2019 simulations. It exhibits a consistent

56

Figure 5.9 Thebes Discharge (2020)

pattern throughout the observed time frame, further affirming the model’s proficiency in capturing

the complex associations between fluctuating gauge height and resulting discharge dynamics. Con-

tinued efforts to refine the model’s representation of hydrological processes at Thebes, including

investigations into factors influencing discharge variations, will contribute to further enhancing the

model’s accuracy in replicating the complex dynamics of this specific location.

5.2.2 Comparative Analysis of Model-Simulated Flood Inundation and Satellite Imagery

Two distinct representations of the flood inundation extent in St. Louis for 2020 were evaluated,

one from satellite imagery (Figure 5.10a) and the other generated by the model (Figure 5.10a).

A comparison between these two datasets reveals a remarkable similarity, indicating the model’s

proficiency in simulating the flood dynamics at the specific time period. The visual congruence be-

tween the observed aerial imagery and the model’s simulated inundation extent provides substantial

evidence of the model’s capability to accurately represent the complex interplay of hydrological

processes contributing to flood inundation events. These findings underscore the model’s utility as

a valuable tool for assessing and predicting flood risk in the St. Louis region, thereby facilitating

57

informed decision-making in flood management and disaster preparedness.

(a) Satellite Imagery of Flood Inundation

(b) Model Output of Flood Inundation

Figure 5.10 Flood Inundation Comparison at St. Louis (13 June 2020)

58

5.3 Case 3 : 2018 River Model

Table 5.12 provides an overview of the case scenarios and associated parameters. Upstream inlet

Table 5.12 Case 3: Model Parameters

Detail
Topo-Bathy Resolution
Year
Mesh Size
Manning’s Roughness (𝑛)

Value
15m
2018
80m x 120m (approx)
River - 0.033
Flood Plain - 0.09

boundary condition at St Louis, with subcritical flow in cubic meters per second(cms) was assigned

(Figure 5.11). Downstream exit boundary condition was assigned as water surface elevation at

Thebes ( (Figure 5.12).

Figure 5.11 Upstream Boundary Condition (2018)

59

Figure 5.12 Downstream Boundary Condition (2018)

5.3.1 Comparative Analysis of Discharge and Gauge Height Data

For Case 3, a similar approach as in Case 1 was employed to analyze the discharge and gauge

height data.

5.3.1.1 Chester Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table 5.13

and Figure 5.13.

Table 5.13 Results for Chester Gauge Height (2018)

Metric
MAE
RMSE
R-squared

Value
0.5232 (m)
0.5657 (m)
0.9356

The evaluation metrics for 2018 indicate continued promising results. The MAE of 0.5232

meters and RMSE of 0.5657 meters demonstrate the model’s consistency in accurately representing

60

the observed gauge heights. The R-squared value of 0.9356 reaffirms a strong correlation between

the model predictions and actual data.

Comparing these metrics with the results from the previous year (2019 and 2020), we observe

an increase in MAE and RMSE. Despite this deviation, the R-squared value signifies the model’s

robustness in capturing the complex hydrological dynamics at the Chester station during the year

2018. During this period, the flood stage remained within the range of normal to minor, showcasing

Figure 5.13 Chester Gauge Height (2018)

the model’s persistent ability to accurately capture the nuances of the system under study during

relatively stable conditions. The simulated results begin to deviate from observed when water

surface elevation is action and minor flood stage.

This phenomenon in 2018 can be attributed to similar factors mentioned for 2019 and 2020,

including the complex interplay of increased rainfall, river discharge, and potential dynamic changes

in the landscape. Additionally, the observed deviation may be influenced by the simplification of

certain hydrological parameters, such as the assumption of a uniform Manning’s n value, and the

exclusion of small tributaries contributing to the river’s flow. The model’s predictive accuracy

61

might also be affected by the resolution of the mesh and the topobathymetric data, potentially

leading to an underestimation or overestimation of flood extent and intensity, especially in areas of

complex topographical variability.

5.3.1.2 Chester Discharge Data

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.14

and Figure 5.14. The evaluation metrics for Chester discharge data in 2018 indicate positive trends

Table 5.14 Chester Discharge Results (2018)

Metric
MAE
RMSE
R-squared

Value
419.6917 (m3/s)
551.5659 (m3/s)
0.9666

in comparison to the results obtained in the previous years (2019 and 2020). The Mean Absolute

Error (MAE) of 419.6917 m3/s and Root Mean Squared Error (RMSE) of 551.5659 m3/s reflect

models accuracy. The R-squared value of 0.9666 signifies a strong correlation between the model’s

predicted and observed discharge values.

The decrease in MAE and RMSE suggests improvement in the model’s performance in repre-

senting the Chester discharge dynamics for the year 2018. The high R-squared value indicates that

the model captures a substantial portion of the variability in the observed discharge data.

Further investigation into the positive trends observed in these metrics can provide valuable

insights into the model’s performance and guide potential refinements. Additionally, a comparative

analysis with the 2019 results, as presented in Figure 5.14, can offer a nuanced understanding of

the model’s behavior over different temporal conditions. Figure 5.14 illustrates the trend observed

in the Chester discharge for the year 2020, offering insights into the dynamics that closely mirror

the corresponding gauge height data. The figure depicts a consistent pattern over the recorded

time frame, reinforcing the model’s ability to capture the intricate relationships between fluctuating

gauge height and resulting discharge dynamics. This alignment provides substantial evidence of

the model’s efficacy in simulating the complex hydrological processes within the studied system.

62

Figure 5.14 Chester Discharge Data (2018)

5.3.1.3 St Louis Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table 5.15

and Figure 5.15.

Table 5.15 Results for St Louis Gauge Height (2018)

Metric
MAE
RMSE
R-squared

Value
0.2140 (m)
0.2301 (m)
0.9900

The evaluation metrics for 2018 indicate continued promising results. The MAE of 0.5232

meters and RMSE of 0.2301 meters demonstrate the model’s consistency in accurately representing

the observed gauge heights. The R-squared value of 0.9900 reaffirms a strong correlation between

the model predictions and actual data. Figure 5.15 illustrates the trend observed in the St Louis gauge

height for the year 2020, offering insights into the dynamics that closely mirror the corresponding

gauge height data. The figure depicts a consistent pattern over the recorded time frame, reinforcing

the model’s ability to capture the intricate relationships between fluctuating gauge height and

63

Figure 5.15 Chester Gauge Height (2018)

resulting discharge dynamics. Furthermore, this alignment provides substantial evidence of the

model’s efficacy in simulating the complex hydrological processes within the studied system.

5.3.1.4 Thebes Discharge Data

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.16

and Figure 5.16. The evaluation metrics for Thebes discharge data in 2018 indicate positive trends

Table 5.16 Thebes Discharge Results (2018)

Metric
MAE
RMSE
R-squared

Value
678.3087 (m3/s)
932.8047 (m3/s)
0.9073

in comparison to the results obtained in the previous years (2019 and 2020). The Mean Absolute

Error (MAE) of 678.3087 m3/s and Root Mean Squared Error (RMSE) of 932.8047 m3/s reflect

models accuracy. The R-squared value of0.9073 signifies a strong correlation between the model’s

predicted and observed discharge values. The decrease in MAE and RMSE suggests improvement

64

in the model’s performance in representing the Thebes discharge dynamics for the year 2018. The

high R-squared value indicates that the model captures a substantial portion of the variability in the

observed discharge data.

Further investigation into the positive trends observed in these metrics can provide valuable

insights into the model’s performance and guide potential refinements. Additionally, a comparative

analysis with the 2019 results, as presented in Figure 5.16, can offer a nuanced understanding of the

model’s behavior over different temporal conditions. Figure 5.16 illustrates the trend observed in

Figure 5.16 Thebes Discharge Data (2018)

the Thebes discharge for the year 2020, offering insights into the dynamics that closely mirror the

corresponding discharge data. The figure depicts a consistent pattern over the recorded time frame,

reinforcing the model’s ability to capture the intricate relationships between fluctuating discharge

and resulting discharge dynamics. This alignment provides substantial evidence of the model’s

efficacy in simulating the complex hydrological processes within the studied system.

The close correspondence between the observed trends in gauge height and discharge data

for 2018 builds upon the insights gained from the analysis of 2020 data. This consistency across

65

different temporal conditions underscores the model’s robustness in representing the interconnected

dynamics of river systems. To enhance the model’s accuracy in replicating complex hydrological

behavior, continued investigations into the river’s flow regime are essential. Factors such as

variations in channel geometry and the impact of lateral inflows should be further explored. This

ongoing research will contribute to refining the model and advancing its predictive capabilities,

ensuring a comprehensive understanding of the hydrological processes at play.

5.3.2 Comparative Analysis of Model-Simulated Flood Inundation and Satellite Imagery

Two distinct representations of the flood inundation extent in St. Louis for 2018 were evaluated,

one from satellite imagery (Figure 5.17a) and the other generated by the model (Figure 5.17b).

A comparison between these two datasets reveals a remarkable similarity, indicating the model’s

proficiency in simulating the flood dynamics at the specific time period. The visual congruence be-

tween the observed aerial imagery and the model’s simulated inundation extent provides substantial

evidence of the model’s capability to accurately represent the complex interplay of hydrological

processes contributing to flood inundation events. These findings underscore the model’s utility as

a valuable tool for assessing and predicting flood risk in the St. Louis region, thereby facilitating

informed decision-making in flood management and disaster preparedness.

66

(a) Satellite Imagery of Flood Inundation

(b) Model Output of Flood Inundation

Figure 5.17 Flood Inundation Comparison at St. Louis (23 Oct 2018)

5.4 Case 4 : Effects of Manning’s Roughness

In this case, our focus turned to a meticulous exploration of the sensitivity of the hydrological

model to variations in Manning’s roughness values. Manning’s n, a key parameter characterizing

surface roughness, has a profound impact on the flow dynamics within river channels and over

floodplains. As illustrated in Table 5.17, this case involved a deliberate alteration of Manning’s

n values, specifically within the river and floodplain domains, while all other model parameters

remained same.

The rationale behind this investigation lied in the need to uncover the complex interplay between

surface roughness and simulated hydrological outcomes. By systematically varying Manning’s

n, we aimed to discern any noticeable changes in the modeled results, including alterations in

discharge patterns, and water surface elevations. This controlled experimentation provides a

better understanding of how adjustments in Manning’s roughness values contribute to the overall

67

predictive accuracy of the model.

Table 5.17 Case 4: Model Parameters

Detail
Topo-Bathy Resolution
Year
Mesh Size

Manning’s Roughness (𝑛)

Value
15m
2019
80m x 120m (approx)
River - 0.033
Flood Plain - 0.09
River - 0.03
Flood Plain - 0.05

5.4.1 Comparative Analysis of Discharge and Gauge Height Data

Aside from the variations in Manning’s values, the model settings remain consistent with those

detailed in Case 1. This allowed a targeted examination of how changes in Manning’s roughness

specifically influence the outcomes related to discharge and gauge height, offering valuable insights

into the model’s sensitivity to this key parameter.

5.4.1.1 Chester Discharge Data

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.18 and

Figure 5.18. while both high and low Manning’s roughness settings demonstrate a strong correla-

Table 5.18 Chester Discharge Results (Variable Manning’s Roughness)

Metric
MAE
RMSE
R-squared

High Manning Value Results Low Manning Value Results

732.8926 (m3/s)
998.9199 (m3/s)
0.9648

709.6547 (m3/s)
967.2168 (m3/s)
0.9670

tion with observed discharge, the lower MAE and RMSE values under low Manning’s roughness

imply a more accurate representation of the discharge dynamics in the model. When using the high

Manning’s value compared to the low Manning’s value, MAE and RMSE increased by approx-

imately 3.26% and 3.28%, while R-squared decreased approximately 0.23% respectively.These

results underscore the sensitivity of the model to Manning’s roughness and emphasize the impor-

tance of considering this parameter carefully for improved accuracy in simulating hydrological

processes.

68

Figure 5.18 Chester Discharge Results (Variable Manning’s Roughness)

5.4.1.2 Chester Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table

5.19 and Figure 5.19. while both high and low Manning’s roughness settings demonstrate a strong

Table 5.19 Chester Gauge Height (Variable Manning’s Roughness)

Metric
MAE
RMSE
R-squared

High Manning Value Results Low Manning Value Results

0.3235 m
0.3844 m
0.9784

0.3933 m
0.5501 m
0.9558

correlation with observed discharge, the lower MAE and RMSE values under HIGH Manning’s

roughness imply a more accurate representation of the water surface elevation in the model. Under

the high Manning’s value compared to the low Manning’s value, MAE and RMSE decreased by

approximately 17.75% and 30.1%, respectively, while R-squared increased approximately 2.37%.

These results highlight the significance of finding accurate Manning’s roughness for capturing

accurate the fluctuations in water surface elevation within the modeled system.

69

Figure 5.19 Chester Gauge Height (Variable Manning’s Roughness)

5.4.1.3 St Louis Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table 5.20

and Figure 5.20. Both high and low Manning’s roughness settings exhibit a strong correlation with

Table 5.20 St Louis Gauge Height (Variable Manning’s Roughness)

Metric
MAE
RMSE
R-squared

High Manning Value Results Low Manning Value Results

0.2924 m
0.4076 m
0.9793

0.7048 m
0.8712 m
0.9055

the observed water surface elevation, the lower MAE and RMSE values under High Manning’s

roughness suggest a more accurate representation of the water surface elevation in the model.

Comparing the high Manning’s value to the low Manning’s value, MAE and RMSE decreased by

approximately 141.17% and 113.99%, respectively, while R-squared decreased by approximately

7.54%. These findings underscore the importance of selecting an appropriate Manning’s roughness

value for accurately capturing the fluctuations in water surface elevation within the modeled system.

70

Figure 5.20 St Louis Gauge Height (Variable Manning’s Roughness)

5.4.1.4 Thebes Discharge Results

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.21

and Figure 5.21. While both high and low Manning’s roughness settings demonstrate a correla-

Table 5.21 Thebes Discharge Results (Variable Manning’s Roughness)

Metric
MAE
RMSE
R-squared

High Manning Value Results Low Manning Value Results

1015.4562 (m3/s)
1276.4613 (m3/s)
0.9419

959.3069 (m3/s)
1152.7524 (m3/s)
0.9527

tion with observed discharge at Thebes, the lower MAE and RMSE values under low Manning’s

roughness indicate a more accurate representation of the discharge dynamics in the model. Under

the low Manning’s value compared to the high Manning’s value, MAE and RMSE decreased by

approximately 5.55% and 9.80%, respectively, while R-squared increased by approximately 1.15%.

These results underscore the sensitivity of the model to Manning’s roughness and highlight the

importance of selecting an appropriate Manning’s roughness value for capturing accurate discharge

dynamics within the modeled system.

71

Figure 5.21 Thebes Discharge Results (Variable Manning’s Roughness)

The investigation into Manning’s roughness sensitivity during the 2019 flood revealed distinct

patterns in the model’s accuracy for discharge and water surface elevation, with lower Manning’s

values contributing to more precise discharge predictions and higher Manning’s values enhancing

water surface elevation accuracy. Varying degrees of sensitivity to Manning’s roughness, empha-

sizing the need for meticulous parameter selection in flood modeling, particularly when predicting

discharge and water surface elevation during flood events.

5.5 Case 5 : Effects of Interpolation Methods

In Case 5, our attention shifts towards a detailed examination of the impact of different topobathy

data interpolation methods on the hydrological model’s performance. Specifically, we explore two

interpolation techniques: Inverse Distance Weighted (IDW) and linear interpolation while keeping

all other model parameters constant (Table 5.22).

The motivation behind this investigation stems from the critical role that accurate interpolation

of topobathy data to the mesh plays in shaping the model’s predictive capabilities. Understanding

how different interpolation methods influence the model’s outcomes is paramount for improving

72

the precision of simulated hydrological processes. By systematically comparing IDW and linear

interpolation, we seek to uncover any noticeable differences in the model’s predictions, particularly

in terms of discharge patterns and water surface elevations. This controlled experimentation serves

as a valuable exercise to gauge the significance of selecting appropriate interpolation methods for

optimal model performance.

Table 5.22 Case 5: Model Parameters

Detail
Topo-Bathy Resolution
Year
Mesh Size
Manning’s Roughness (𝑛)

Value
15m
2019
80m x 120m (approx)
River - 0.03
Flood Plain - 0.05

5.5.1 Comparative Analysis of Discharge and Gauge Height Data

With consistent model settings given in Table 5.22, we focused on the evaluation of two

distinct methods employed for interpolating topobathy data onto the model mesh. This case aims

to understand the impact of interpolation techniques on simulated outcomes, including discharge

patterns and water surface elevations.

5.5.1.1 Chester Discharge Data

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.23

and Figure 5.22. The Mean Absolute Error (MAE) exhibited a notable improvement for IDW,

Table 5.23 Chester Discharge Results (variable Interpolation Methods)

Metric
MAE
RMSE
R-squared

IDW
510.6439 (m3/s)
748.9346 (m3/s)
0.9802

Linear
709.8542 (m3/s)
967.9925 (m3/s)
0.9670

with a percentage decrease of approximately 28% compared to Linear interpolation. Similarly,

the Root Mean Square Error (RMSE) demonstrated an approximately 22.64% reduction for IDW

compared to Linear interpolation. The R-squared value, indicating the goodness of fit, increased

by approximately 1.37% for IDW compared to Linear interpolation. These percentage differences

73

underscore the superior performance of IDW in terms of accuracy and predictive capability for

simulating Chester discharge within the studied hydrological model.

Figure 5.22 Chester Discharge Results (variable Interpolation Methods)

5.5.1.2 Chester Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table 5.24

and Figure 5.23.

Table 5.24 Chester Gauge Height (variable Interpolation Methods)

Metric
MAE
RMSE
R-squared

IDW

Linear

0.3761 m 0.3935 m
0.5255 m 0.5506 m

0.9597

0.9557

The Mean Absolute Error (MAE) demonstrates a decrease of approximately 4.41% when

employing IDW compared to Linear interpolation. Similarly, the Root Mean Square Error (RMSE)

shows a reduction of about 4.57% with IDW, indicating improved accuracy. Although the R-

squared values exhibit a marginal increase of approximately 0.42% for IDW, the overall trend

74

suggests that IDW outperforms Linear interpolation in capturing the dynamics of Chester gauge

height. These findings underscore the significance of the interpolation method in influencing

the accuracy of simulated water surface elevations, emphasizing the importance of choosing an

appropriate method for reliable hydrological predictions.

Figure 5.23 Chester Gauge Height (variable Interpolation Methods)

5.5.1.3 St Louis Gauge Height

The analysis yielded the following metrics for the St Louis gauge height, as shown in Table 5.25

and Figure 5.24.

Table 5.25 St Louis Gauge Height (variable Interpolation Methods)

Metric
MAE
RMSE
R-squared

IDW

Linear

0.6827 m 0.7054 m
0.8571 m 0.8721 m

0.9086

0.9053

The Mean Absolute Error (MAE) indicates a minimal decrease of approximately 3.22% when

utilizing IDW compared to Linear interpolation. Similarly, the Root Mean Square Error (RMSE)

75

reflects a slight reduction of about 1.72% with IDW, suggesting a modest improvement in accuracy.

The R-squared values exhibit a marginal increase of approximately 0.35% for IDW, suggesting an

obvious advantage in capturing the variations in St. Louis gauge height. While the differences

are relatively modest, these findings imply that IDW may offer a slightly enhanced performance in

simulating St. Louis gauge height compared to Linear interpolation, emphasizing the impact of

interpolation methods on hydrological modeling outcomes.

Figure 5.24 St Louis Gauge Height (variable Interpolation Methods)

5.5.1.4 Thebes Discharge Data

The analysis yielded the following metrics for the Thebes discharge, as shown in Table 5.26

and Figure 5.25. The Mean Absolute Error (MAE) values indicate a slight decrease of about

Table 5.26 Thebes Discharge Results (variable Interpolation Methods)

Metric
MAE
RMSE
R-squared

IDW
955.3412 (m3/s)
1148.2733 (m3/s)
0.9530

Linear
959.0168 (m3/s)
1153.0059 (m3/s)
0.9526

76

0.49% when utilizing IDW compared to Linear interpolation, implying a marginal improvement

in accuracy. Similarly, the Root Mean Square Error (RMSE) exhibits a modest reduction of

approximately 0.41% with IDW, suggesting a subtle enhancement in predictive capability. The

R-squared values demonstrate a negligible increase of about 0.04% for IDW, indicating a marginal

advantage in capturing the variations in Thebes discharge. While the differences are minimal,

these findings suggest that IDW may offer a slightly improved performance in simulating Thebes

discharge compared to Linear interpolation, emphasizing the impact of interpolation methods on

hydrological modeling outcomes. In this comprehensive evaluation of use of interpolation methods

Figure 5.25 Thebes Discharge Results (variable Interpolation Methods)

for interpolating the topobathy to the mesh, the results unveil differences in performance. For

Chester discharge, IDW demonstrates a favorable edge over Linear interpolation, showcasing lower

MAE and RMSE values and a higher R-squared value, indicative of enhanced accuracy. Conversely,

for Chester gauge height, the difference in performance between IDW and Linear is marginal, with

minimal variations in MAE, RMSE, and R-squared values. Transitioning to St. Louis gauge height,

IDW and Linear exhibit comparable performance, with negligible differences in MAE, RMSE, and

77

R-squared. In Thebes discharge, IDW showcases a subtle advantage over Linear, with minimal

improvements in MAE, RMSE, and R-squared values. These findings emphasize the importance of

carefully selecting interpolation methods tailored to specific modeling objectives, recognizing that

their impact varies across different hydrological parameters and locations. The nuanced variations

observed underscore the complex nature of hydrological modeling, where methodological choices

play a pivotal role in shaping the accuracy and reliability of simulation outcomes.

5.6 Case 6: Impact of Topobathy Resolution

In Case 6, our attention shifts to exploring the influence of topobathy resolution on the outcomes

of the hydrological model. The resolution of topobathy data, represented by the spatial dimensions

of the mesh, plays a crucial role in capturing intricate details of the terrain. For this investigation,

we compare two distinct resolutions: 10m x 10m and 15m x 15m. This deliberate alteration in

resolution aims to unravel any discernible differences in the model’s ability to simulate hydrological

processes, particularly focusing on river discharge and water surface elevation. All other model

parameters remain constant, allowing us to isolate the impact of topobathy resolution on model

outcomes (Table 5.27). Through this exploration, we seek to gain insights into the optimal resolution

for accurately representing the complex topography of the study area.

Table 5.27 Case 6: Model Parameters

Detail

Topo-Bathy Resolution

Year
Mesh Size
Manning’s Roughness (𝑛)

Value
15m
10m
2019
80m x 120m (approx)
River - 0.033
Flood Plain - 0.09

5.6.1 Comparative Analysis of Discharge and Gauge Height Data

With consistent model settings given in Table 5.27, we focused on the evaluation of two

distinct topobathy data resolution onto the model mesh. This case aims to understand the impact

of topobathy resolution on simulated outcomes, including discharge patterns and water surface

elevations.

78

5.6.1.1 Chester Discharge Data

The analysis yielded the following metrics for the Chester discharge, as shown in Table 5.28 and

Figure 5.26. In assessing the impact of different resolutions of topobathy data on Chester discharge

Table 5.28 Chester Discharge Results (variable Topobathy Resolution)

Metric
MAE
RMSE
R-squared

15m
757.1539 (m3/s)
1043.2965 (m3/s)
0.9616

10m
727.3129 (m3/s)
998.0488 (m3/s)
0.9649

simulation, the results reveal subtle distinctions. The lower resolution of 15m x 15m shows a

slightly higher Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) compared to the

higher resolution of 10m x 10m, with percentage differences of approximately 4.04% and 4.69%,

respectively. This suggests that the finer resolution contributes to a marginally better accuracy

in capturing the discharge dynamics. Additionally, the R-squared value experiences a minimal

increase of about 0.34%, implying a slightly improved goodness of fit with the higher resolution.

While the differences are modest, they underscore the sensitivity of the model to topobathy data

resolution, emphasizing the importance of considering finer data granularity for enhanced accuracy

in simulating hydrological processes.

79

Figure 5.26 Chester Discharge Results (variable Topobathy Resolution)

5.6.1.2 Chester Gauge Height

The analysis yielded the following metrics for the Chester gauge height, as shown in Table 5.29

and Figure 5.27.

Table 5.29 Chester Gauge Height (variable Topobathy Resolution)

Metric
MAE
RMSE
R-squared

15m

10m

0.3378 m 0.3533 m
0.3820 m 0.4000 m

0.9787

0.9766

For the Chester water surface elevation, transitioning from a resolution of 15m to 10m results in

a higher MAE and RMSE by approximately 4.59% and 4.71%, respectively, indicating a decrease in

accuracy. The R-squared value shows a slight decrease of approximately 0.21%. These percentage

differences highlight a trade-off between increased resolution and model performance. The finer

10m resolution might capture more detailed features but requires a compareable finer mesh. In the

same mesh size, it appears to introduce some variability, leading to higher errors and a slightly

80

reduced goodness of fit.

It’s crucial to weigh the benefits of increased resolution against the

associated uncertainties and potential noise in the model output.

Figure 5.27 Chester Gauge Height (variable Topobathy Resolution)

5.6.1.3 St Louis Gauge Height

The analysis yielded the following metrics for the St Louis gauge height, as shown in Table 5.30

and Figure 5.28.

Table 5.30 St Louis Gauge Height (variable Topobathy Resolution)

Metric
MAE
RMSE
R-squared

15m

10m

0.2871 m 0.2936 m
0.4112 m 0.4175 m

0.9789

0.9783

Transitioning from a resolution of 15m to 10m for St. Louis water surface elevation results in a

2.25% increase in MAE, a 1.54% increase in RMSE, and a negligible 0.06% decrease in R-squared.

These percentage differences suggest a slight decrease in accuracy, indicating a trade-off between

increased resolution and model performance. The finer 10m resolution may capture more detailed

81

features but appears to introduce some variability, leading to higher errors and a slightly reduced

goodness of fit.

It’s crucial to weigh the benefits of increased resolution against the associated

uncertainties and potential noise in the model output.

Figure 5.28 St Louis Gauge Height (variable Topobathy Resolution)

5.6.1.4 Thebes Discharge Data

The analysis yielded the following metrics for the Thebes discharge, as shown in Table 5.31 and

Figure 5.29. the transition from a resolution of 15m to 10m for Thebes discharge shows minimal

Table 5.31 Thebes Discharge Results (variable Topobathy Resolution)

Metric
MAE
RMSE
R-squared

15m
1034.5035 (m3/s)
1248.1966 (m3/s)
0.9445

10m
1034.0265 (m3/s)
1245.4360 (m3/s)
0.9447

changes. The MAE exhibits a negligible increase of approximately 0.046%, indicating a slight de-

crease in accuracy for the higher resolution. Similarly, the RMSE demonstrates a marginal increase

of about 0.22%, suggesting a minor escalation in the model’s predictive errors. The R-squared

82

value remains nearly unchanged, with a negligible decrease of approximately 0.021%. These

minute variations imply that altering the resolution has limited impact on the model’s performance

for simulating Thebes discharge. The results revealed slight variations in model performance,

Figure 5.29 Thebes Discharge Results (variable Topobathy Resolution)

with trade-offs observed between increased resolution and model accuracy. For Chester discharge,

transitioning from 15m to 10m resolution resulted in improved accuracy, while for water surface

elevation, the coarser 15m resolution showcased superior performance. Conversely, St. Louis

demonstrated minimal changes, and for Thebes discharge, the impact of resolution was marginal.

These findings underscore the importance of careful consideration when selecting topobathy data

resolution keeping mesh size in mind. It recognised the intricate trade-offs between model accuracy

and computational efficiency. Such insights contribute to enhancing the reliability of hydrological

simulations, crucial for effective water resource management and flood prediction.

83

5.7 Case 7: Impact of Reduced Discharge

With consistent model settings detailed in Table 5.32 to study the impact of a reduced discharge,

the upper boundary condition flow 𝑄(𝑡) for 2019 at St Louis was reduced by 25% . This deliberate

alteration seeks to assess how changes in upstream discharge affect the hydrological dynamics

and water surface elevations at key locations such as Chester and St Louis. The choice of a 25%

reduction provides a controlled scenario to analyze the system’s sensitivity to variations in upstream

boundary conditions.

Table 5.32 Case 7: Model Parameters

Detail
Topo-Bathy Resolution
Year
Mesh Size
Manning’s Roughness (𝑛)

Value
15m
2019
80m x 120m (approx)
River - 0.033
Flood Plain - 0.09

5.7.1 Analysis of Gauge Height Data

Understanding the influence of reduced discharge on gauge height is essential for assessing

the downstream impacts within the hydrological system. This section scrutinizes the gauge height

responses at key locations, specifically Chester and St Louis, following a 25% reduction in upstream

discharge at St Louis. The examination of gauge height data provides valuable insights into how

changes in discharge conditions manifest in water surface elevations along the river system.

5.7.1.1 Chester Gauge Height

The analysis of gauge height data at the Chester location under the reduced discharge scenario

reveals an average percentage change of approximately 1.39% (Figure 5.30). This metric encapsu-

lates the overall impact of the 25% reduction in upstream discharge on water surface elevations at

Chester. The variations observed underscore the complex interactions between altered boundary

conditions and downstream gauge height responses. This insight enhances our understanding of

the system’s sensitivity to changes in discharge, providing a quantitative measure of the average

percentage shift in water surface elevations at the specified location.

84

Figure 5.30 Water Surface Elevation and Percentage Change Comparison at Chester

5.7.1.2 St Louis Gauge Height

The evaluation of gauge height data at the St. Louis location, considering the implemented

25% reduction in upstream discharge, reveals an average percentage change of approximately

1.36%(Figure 5.31. This quantification captures the alterations in water surface elevations at St.

Louis in response to the modified boundary conditions. The observed average percentage shift offers

valuable insights into the dynamic relationship between discharge adjustments and corresponding

gauge height variations. Such findings contribute to a comprehensive understanding of the system’s

responsiveness to changes in discharge.

5.7.1.3 Comparative Analysis of Model-Simulated Flood Inundation

Two distinct representations of the flood inundation extent in St. Louis for 2019 were evaluated,

one from with original discharge at St Louis as Boundary Condition (Figure 5.32a) and the other

with 25% reduced discharge as St. Louis boundary condition (Figure 5.32b). A comparison

between these two datasets reveals that changes in discharge effects the flood inundation area,

indicating the model’s proficiency in simulating the flood dynamics at the specific time period. The

85

Figure 5.31 Water Surface Elevation and Percentage Change Comparison at St Louis

visual congruence provides substantial evidence of the model’s capability to accurately represent

the complex interplay of hydrological processes contributing to flood inundation events. These

findings underscore the model’s utility as a valuable tool for assessing and predicting flood risk

in the St. Louis region, thereby facilitating informed decision-making in flood management and

disaster preparedness.

The impact of reduced discharge on gauge heights and flood inundation dynamics, has yielded

valuable insights into the model’s responsiveness to changes in boundary conditions. The analysis

of gauge heights in Chester and St. Louis revealed an average percentage change of approxi-

mately 1.375%, emphasizing the model’s sensitivity to variations in discharge. This indicates

the importance of accurate representation of boundary conditions for reliable hydrological simu-

lations. Furthermore, the comparative analysis of model-simulated flood inundation, considering

both original and reduced discharge scenarios, demonstrated the model’s proficiency in capturing

the dynamics of flood extent. The congruence between simulated and observed inundation areas

underscores the model’s reliability for flood risk assessment.

86

(a) Simulated Original Discharge extent

(b) Simulated Reduced Discharge extent

Figure 5.32 Flood Inundation Comparison at St. Louis (20 Jun 2019)

87

5.8 Case 8: Impact of Increased Discharge

With consistent model settings detailed in Table 5.33 to study the impact of a increased discharge,

the upper boundary condition flow 𝑄(𝑡) for 2019 at St Louis was increased by 25% . This deliberate

alteration seeks to assess how changes in upstream discharge affect the hydrological dynamics and

water surface elevations at key locations such as Chester and St Louis. The choice of a 25%

increase provides a controlled scenario to analyze the system’s sensitivity to variations in upstream

boundary conditions.

Table 5.33 Case 8: Model Parameters

Detail
Topo-Bathy Resolution
Year
Mesh Size
Manning’s Roughness (𝑛)

Value
15m
2019
80m x 120m (approx)
River - 0.033
Flood Plain - 0.09

5.8.1 Analysis of Gauge Height Data

Understanding the influence of increased discharge on gauge height is essential for assessing

the downstream impacts within the hydrological system. This section scrutinizes the gauge height

responses at key locations, specifically Chester and St Louis, following a 25% increase in upstream

discharge at St Louis. The examination of gauge height data provides valuable insights into how

changes in discharge conditions manifest in water surface elevations along the river system.

5.8.1.1 Chester Gauge Height

The analysis of gauge height data at the Chester location under the increased discharge scenario

reveals an average percentage change of approximately 1.16% (Figure 5.33). This metric encapsu-

lates the overall impact of the 25% increase in upstream discharge on water surface elevations at

Chester. The variations observed underscore the complex interactions between altered boundary

conditions and downstream gauge height responses. This insight enhances our understanding of

the system’s sensitivity to changes in discharge, providing a quantitative measure of the average

percentage shift in water surface elevations at the specified location.

88

Figure 5.33 Water Surface Elevation and Percentage Change Comparison at Chester

5.8.1.2 St Louis Gauge Height

The evaluation of gauge height data at the St. Louis location, considering the implemented

25% increase in upstream discharge, reveals an average percentage change of approximately 1.08%

(Figure 5.34. This quantification captures the alterations in water surface elevations at St. Louis

in response to the modified boundary conditions. The observed average percentage shift offers

valuable insights into the dynamic relationship between discharge adjustments and corresponding

gauge height variations. Such findings contribute to a comprehensive understanding of the system’s

responsiveness to changes in discharge.

89

Figure 5.34 Water Surface Elevation and Percentage Change Comparison at St Louis

5.8.1.3 Comparative Analysis of Model-Simulated Flood Inundation

Two distinct representations of the flood inundation extent in St. Louis for 2019 were evaluated,

one from with original discharge at St Louis as Boundary Condition (Figure 5.35a) and the other

with 25% increased discharge as St. Louis boundary condition (Figure 5.35b). A comparison

between these two datasets reveals that changes in discharge effects the flood inundation area,

indicating the model’s proficiency in simulating the flood dynamics at the specific time period. The

visual congruence provides substantial evidence of the model’s capability to accurately represent

the complex interplay of hydrological processes contributing to flood inundation events. These

findings underscore the model’s utility as a valuable tool for assessing and predicting flood risk

in the St. Louis region, thereby facilitating informed decision-making in flood management and

disaster preparedness.

The impact of increased discharge on gauge heights and flood inundation dynamics, has yielded

valuable insights into the model’s responsiveness to changes in boundary conditions. The analysis

90

(a) Simulated Original Discharge extent

(b) Simulated Increased Discharge extent

Figure 5.35 Flood Inundation Comparison at St. Louis (20 Jun 2019)

of gauge heights in Chester and St. Louis revealed an average percentage change of approximately

1.12%, emphasizing the model’s sensitivity to variations in discharge. This indicates the importance

of accurate representation of boundary conditions for reliable hydrological simulations. Further-

more, the comparative analysis of model-simulated flood inundation, considering both original and

increased discharge scenarios, demonstrated the model’s proficiency in capturing the dynamics of

flood extent. The congruence between simulated and observed inundation areas underscores the

model’s reliability for flood risk assessment.

5.9 Case 9: Effects of Levee Construction

Understanding the dynamic impact of levee construction on the hydrological behavior of the

modeled system is important. Levees are critical components in flood management, altering the

flow dynamics and influencing water surface elevations. The focus of this investigation is on the

91

introduction of a levee structure approximately 10 kilometers in length, implemented at St. Louis,

with a height of 4 meters (Figure 5.36). By comparing simulations with and without the levee, we

aim to discern the alterations in water surface elevations and discharge patterns. This case provides

a comprehensive understanding of how strategic levee construction can influence the hydrological

responses in the studied region.

Figure 5.36 Levee added at St Louis Location

5.9.1 Comparative Analysis of Discharge and Gauge Height Data

In the comparative analysis of discharge and gauge height data, the impact of levee construction

on upstream and downstream hydrological conditions was investigated. Notably, the simulation

results demonstrate consistent discharge values at Chester (Figure 5.37) and Thebes (Figure 5.40),

92

both downstream of the levee installation site, suggesting that the presence of the levee does not

significantly alter the discharge dynamics at these locations. Similarly, the water surface elevation

at Chester remains largely unchanged (Figure 5.38), indicating minimal influence from the levee

on this aspect of the hydrological system. However, St. Louis, located upstream of the levee,

shows a subtle increase in water surface elevation (Figure 5.39), particularly when the gauge

height enters Moderate and Major flood stages. This nuanced response at St. Louis suggests that

the levee construction may have a localized effect on water surface elevation, primarily evident

during specific flow conditions. These findings contribute valuable insights into the complex

interactions between levee infrastructure and riverine hydrodynamics, emphasizing the need for a

comprehensive understanding of the spatial and temporal effects of such interventions.

Figure 5.37 Chester Discharge Results (Levee Construction Upstream)

93

Figure 5.38 Chester Water Surface Elevation Results (Levee Construction Upstream)

Figure 5.39 St Louis Water Surface Elevation Results (Levee Construction Upstream)

94

Figure 5.40 Thebes Discharge results (Levee Construction Upstream)

5.9.1.1 Comparative Analysis of Model-Simulated Flood Inundation

In the comparative analysis of model-simulated flood inundation, the focus was on assessing

the effects of levee construction on the extent of inundation downstream of the intervention site.

The results reveal a discernible but localized impact, with a modest reduction in inundation area

observed immediately downstream of the levee. This reduction, however, diminishes progressively

as the analysis extends further downstream. The findings suggest that while the levee contributes to

a reduction in inundation within the immediate vicinity, its influence lessens as the distance from

the construction site increases. This nuanced spatial variation in inundation effects underscores the

importance of considering the specific geological and hydraulic characteristics of the study area

when evaluating the efficacy of levee interventions in mitigating flood risks.

Overall, the observed variations in discharge and gauge height at Chester and Thebes down-

stream of the levee construction site are minimal, indicating that the levee has limited influence on

these parameters at these locations. While a reduction in inundation extent is noted immediately

95

(a) Simulated Flood Inundation (Actual)

(b) Simulated Flood Inundation (Levee Added)

Figure 5.41 Flood Inundation Comparison at St. Louis (21 Nov 2019)

downstream of the levee, this effect diminishes with distance from the intervention site. These

findings contribute to a comprehensive understanding of the spatial dynamics of levee impacts on

hydrological processes.

5.10 Case 10: Turbulence Model Comparison (k-𝜀 vs. Parabolic)

The primary objective of this analysis is to evaluate and contrast the predictive capabilities of

the k-𝜀 turbulence model and the Parabolic model (0.7 Constant). The comparison focuses on

their performance across different hydrological conditions, providing a basis for understanding the

strengths and limitations of each model.

5.10.1 Chester Discharge Data Results

Figure 5.42 illustrates the comparative performance of the k-𝜀 and Parabolic (0.7 Constant)

models in simulating Chester discharge.

The comparison between the k-𝜀 and Parabolic (0.7 Constant) turbulence models for Chester

discharge data revealed minor differences in their performance. Analyzing the percentage changes

in key metrics provides a concise view of the models’ relative performance. The k-𝜀 turbulence

model exhibited a 0.31% decrease in MAE compared to the Parabolic (0.7 Constant) model. The

96

Figure 5.42 Comparison of k-𝜀 and Parabolic (0.7 Constant) Models for Chester Discharge

k-𝜀 turbulence model demonstrated a 0.35% decrease in RMSE relative to the Parabolic model.

A slight 0.04% decrease in R-squared value was observed with the k-𝜀 model compared to the

Parabolic (0.7 Constant) model.

5.10.2 Chester Gauge Height Data Results

Figure 5.43 illustrates the comparative performance of the k-𝜀 and Parabolic (0.7 Constant)

models in simulating Chester gauge height.

The k-𝜀 turbulence model exhibited a 2.19% increase in MAE compared to the Parabolic (0.7

Constant) model. The RMSE for the k-𝜀 model showed a 1.74% increase compared to the Parabolic

(0.7 Constant) model. However, the R-squared value demonstrated a slight improvement with the

k-𝜀 model, showing a 0.17% increase.

These small percentage differences in MAE, RMSE, and R-squared values indicate comparable

performance between the k-𝜀 and Parabolic (0.7 Constant) models for simulating gauge height at

the Chester location.

97

Figure 5.43 Comparison of k-𝜀 and Parabolic Models for Chester water surface elevation

5.10.3 St. Louis Water Surface Elevation Data Results

Figure 5.44 illustrates the comparative performance of the k-𝜀 and Parabolic (0.7 Constant)

models in simulating water surface elevation at St. Louis.

The comparison between the k-𝜀 and Parabolic (0.7 Constant) turbulence models for St. Louis

water surface elevation data revealed minor differences in their performance. Analyzing the

percentage changes in key metrics provides a concise view of the models’ relative performance.

The k-𝜀 turbulence model exhibited a 2.63% increase in MAE compared to the Parabolic (0.7

Constant) model. The RMSE for the k-𝜀 model showed a 1.60% increase compared to the Parabolic

(0.7 Constant) model. The R-squared value demonstrated a 0.20% increase with the k-𝜀 model.

5.10.4 Thebes Discharge Data Results

Figure 5.45 illustrates the comparative performance of the k-𝜀 and Parabolic (0.7 Constant)

models in simulating Thebes discharge.

The comparison between the k-𝜀 and Parabolic (0.7 Constant) turbulence models for Thebes

discharge data revealed minor differences in their performance. Analyzing the percentage changes

98

Figure 5.44 Comparison of k-𝜀 and Parabolic Models for St. Louis Water Surface Elevation

Figure 5.45 Comparison of k-𝜀 and Parabolic Models for Thebes Discharge

in key metrics provides a concise view of the models’ relative performance.

The k-𝜀 turbulence model exhibited a 0.05% decrease in MAE compared to the Parabolic (0.7

99

Constant) model. The RMSE for the k-𝜀 model showed a 0.07% decrease compared to the Parabolic

(0.7 Constant) model. The R-squared value demonstrated a negligible 0.01% increase with the k-𝜀

model.

These small percentage differences in MAE, RMSE, and R-squared values at all stations indicate

comparable performance between the k-𝜀 and Parabolic (0.7 Constant) models for simulating

discharge and water surface elevation. The comparative analysis of the k-𝜀 and Parabolic (0.7

Constant) turbulence models for discharge simulations at Chester, water surface elevation at Chester

and St. Louis, and discharge at Thebes revealed minor differences in their performance metrics. The

small percentage variations in MAE, RMSE, and R-squared values suggest comparable accuracy

between the two models for representing hydrological processes in the studied locations. While

both models demonstrate similar proficiency, considerations for the specific characteristics of the

study area are vital in selecting an appropriate model. Given the large spatial extent of the area

under investigation, factors such as computational efficiency and the ability to capture diverse flow

conditions become crucial. Therefore, a recommendation for the most suitable model depends on

the specific goals of the study, computational resources available, and the scale of the hydrological

processes being simulated. In general, the k-𝜀 turbulence model is known to be computationally

more demanding than simpler turbulence models, such as the Parabolic model. The k-𝜀 model

solves additional transport equations for turbulent kinetic energy and dissipation rate, introducing

more computational overhead.

100

CHAPTER 6

CONCLUSIONS AND RECOMMENDATIONS

6.1 Conclusions

The results presented for different scenarios in the previous chapter demonstrate the ability of

the SRH-2D model to describe the key aspects of flooding on the Mississippi River. Despite the

good model performance overall, several improvements can be made to the model. First the data

and mesh resolution can be further improved to address some of the discrepancies noted earlier,

especially while capturing peak discharges and river stages during periods of major flooding.

Second, instead of manually adjusting parameters such as the Manning’s roughness values in the

channel and the floodplains, the process can be automated by linking the model with a parameter

estimation software such as PEST [40] as has been done by others in the past [34]. Finally process

refinements are expected to further improve model performance.

Important processes ignored

during the application of the current model include the role of groundwater in flooding.

It is

well-known that shallow groundwater tables and saturated conditions in subsurface and floodplains

can lead to longer and deeper inundation from relatively low intensity flood events. [41]. Therefore

modeling surface water - groundwater interactions in the floodplains can be expected to further

improve model descriptions of peak discharges and river stages. Additional conclusions and

recommendations are included below.

6.1.1 Temporal Consistency and Model Performance

The model consistently demonstrates reliability in accurately representing gauge heights and

discharge dynamics across various years (2019, 2020, and 2018), showcasing temporal consistency

in performance.

6.1.2

Impact of Data and Mesh Resolution

The resolution of mesh and topobathymetric data plays a crucial role in the model’s accuracy,

with potential underestimation or overestimation of flood extent and intensity in areas with com-

plex topography. The results can be further improved using higher resolution datasets and mesh

101

resolution.

6.1.3 Gauge Height and Discharge Trends

Consistent trends observed between gauge height and discharge data underscore the model’s

robustness in capturing interconnected river system dynamics.

6.1.4 Model Adaptability and Refinement

The model demonstrates adaptability and continued refinement, as seen in positive trends in

discharge accuracy and flood inundation simulations over different temporal conditions.

6.1.5 Sensitivity to Manning’s Roughness Value, 𝑛

Distinct patterns in model accuracy related to Manning’s roughness reveal its significant impact

on discharge predictions and water surface elevation accuracy. Lower Manning’s values contribute

to more precise discharge predictions, while higher values enhance elevation accuracy.

6.1.6

Impact of Interpolation Methods

Different topobathy data interpolation methods, specifically Inverse Distance Weighted (IDW)

and linear interpolation, significantly impact the accuracy of simulated hydrologic processes. IDW

outperforms linear interpolation in discharge predictions and water surface elevation simulations.

Other methods such as natural neighbor interpolation and manifold methods [42] offer addition

choices for interpolation.

6.1.7

Impact of Topobathy Resolution

The investigation into different topobathy resolutions (10m x 10m and 15m x 15m) reveals

nuanced variations in model performance. The hydrological model’s sensitivity to resolution is

location-specific and parameter-dependent.

6.1.8 Reduced Discharge on Hydrological Dynamics

A deliberate 25% reduction in upstream discharge at St. Louis provides valuable insights into

hydrological dynamics and water surface elevations at key locations, emphasizing the model’s

sensitivity to variations in boundary conditions.

102

6.1.9 Effects of Levee Construction on Hydrological Behavior

The investigation into the effects of levee construction at St. Louis provides valuable insights

into the dynamic impact of levees on hydrological behavior, with localized effects on gauge height

and discharge patterns.

6.2 Recommendations

6.2.1 Hydrological Parameter Sensitivity

Investigate the sensitivity of model performance to hydrological parameters, particularly Man-

ning’s roughness. Optimize Manning’s values based on modeling goals and conduct ongoing

adjustments to enhance accuracy during intensified flood events.

6.2.2 Enhanced Spatial Resolution

Evaluate and enhance the spatial resolution of the mesh and topobathymetric data to improve

the model’s accuracy, especially in regions with complex topographical features.

6.2.3 Continued Research on River Flow Regime

Further research into variations in channel geometry and the impact of lateral inflows is essential

for refining the model and advancing predictive capabilities. Consider location-specific sensitivities

and parameter-dependent resolution choices.

6.2.4 Temporal Dynamics Analysis

Conduct a detailed analysis of temporal dynamics, focusing on the model’s behavior during

different flood stages. Validate predictions against real-world data for various scenarios and consider

integrating additional data sources.

6.2.5 Optimized Manning’s Roughness Selection

Consider Manning’s roughness meticulously, optimizing values based on specific modeling

goals. Explore the possibility of dynamically adjusting Manning’s roughness during different

stages of flood events.

103

6.2.6 Dynamic Manning’s Adjustment

Explore the possibility of dynamically adjusting Manning’s roughness during different stages

of flood events. This adaptive approach could enhance the model’s ability to represent changing

flow conditions accurately.

6.2.7 Advanced Sensitivity Analysis

Conduct further sensitivity analysis on other key parameters to understand their interaction with

Manning’s roughness and contribute to refining the model for more robust flood predictions.

6.2.8 Validation with Real-world Data

Continuously validate the model’s predictions against real-world data for various scenarios and

Manning’s roughness values to ensure reliability and generalizability.

6.2.9 Tailored Interpolation Selection

Carefully select interpolation methods based on specific modeling objectives and locations.

Prioritize modeling objectives when selecting methods and consider validation across multiple

scenarios.

6.2.10 Consideration of Model Objectives

Prioritize modeling objectives when selecting interpolation methods.

If accurate discharge

predictions are critical, favor methods like IDW that demonstrate superior performance in this

aspect.

6.2.11 Validation Across Multiple Scenarios

Validate the model’s predictions using multiple scenarios and real-world data to ensure the

reliability and generalizability of the selected interpolation method. Continue exploring and refining

interpolation methodologies.

6.2.12 Continuous Exploration of Methodology

Given the complex nature of hydrological modeling, continue exploring and refining interpo-

lation methodologies. Stay updated on advancements in interpolation techniques to enhance the

model’s predictive capabilities over time.

104

6.2.13 Location-Specific Resolution Considerations

Consider location-specific sensitivities when choosing topobathy resolutions. Different areas

may respond differently to resolution changes, as observed in the varying impacts on Chester, St.

Louis, and Thebes.

6.2.14 Parameter-Dependent Resolution Choices

Recognize that the impact of resolution may vary for different hydrological parameters. Strive

for a balance between increased resolution and computational efficiency.

6.2.15 Balancing Accuracy and Efficiency

Strive for a balance between increased resolution and computational efficiency. While higher

resolutions may capture more details, weigh the benefits against potential uncertainties and com-

putational demands.

6.2.16 Continuous Model Refinement

Given the nuanced trade-offs observed, continue refining the hydrological model by exploring

different resolutions and their impacts. Stay attuned to advancements in computational techniques

and hydrological modeling to enhance the model’s predictive capabilities over time.

6.2.17 Boundary Condition Accuracy

Ensure the accurate representation of boundary conditions in hydrological simulations, partic-

ularly upstream discharge. Regularly update and verify discharge data to enhance the reliability of

model predictions.

6.2.18 Continuous Model Validation

Continuously validate the hydrological model with observed data to further enhance its accuracy

and reliability. This includes ongoing comparisons of gauge height responses and flood inundation

simulations under various scenarios.

6.2.19 Scenario-Based Analysis

Expand the analysis to include different scenarios of upstream discharge variations. Assess the

model’s response to a range of boundary conditions for a more comprehensive understanding of its

105

sensitivity and performance in different hydrological situations.

6.2.20 Collaboration with Stakeholders

Engage with stakeholders involved in flood management and disaster preparedness to incorpo-

rate their insights into the model refinement process. Collaborative efforts can lead to a more robust

and practical hydrological model that better serves decision-making processes.

6.2.21 Comprehensive Understanding of Levee Impacts

Given the nuanced and localized effects observed in levee construction, conduct further studies

to enhance the understanding of the spatial and temporal dynamics of levee impacts on hydrological

processes.

6.2.22 Continuous Monitoring and Evaluation

Implement a continuous monitoring and evaluation program to assess the long-term effective-

ness of levee construction. Regularly update and validate the hydrological model with observed

data to ensure the accuracy of predictions and adapt strategies based on evolving conditions.

6.2.23 Spatial Considerations in Levee Planning

When planning levee interventions, consider the specific geological and hydraulic characteristics

of the study area. Tailor interventions to local conditions, emphasizing the importance of spatial

variation in the effectiveness of levee constructions.

6.2.24 Collaborative Decision-Making

Engage in collaborative decision-making with stakeholders involved in flood management and

infrastructure development. Incorporating diverse perspectives can lead to more informed decisions

regarding levee construction and enhance the overall resilience of the hydrological system.

106

BIBLIOGRAPHY

[1]

[2]

Jeff Goodell. The Water Will Come: Rising Seas, Sinking Cities and the Remaking of the
Civilized World. Back Bay Books, New York, 2017.

John M Barry. Rising Tide: The Great Mississippi Flood of 1927 and How it Changed
America. Simon and Schuster, 2007.

[3] Gerald E Galloway Jr. Learning from the Mississippi flood of 1993: Impacts, management

issues, and areas for research. 1995.

[4] NOAA. Annual 2019 National Climate Report, National Centers for Environmental Informa-

tion. 2019. https://www.ncei.noaa.gov/access/monitoring/monthly-report/national/201913.

[5] Wikipedia. Mississippi river floods of 2019. 2023. https://en.wikipedia.org/wiki/Mississippi_

River_floods_of_2019.

[6]

James C Knox. The Mississippi River System. Large Rivers: Geomorphology and Manage-
ment, Second Edition, pages 205–252, 2022.

[7] Shannon Doocy, Amy Daniels, Sarah Murray, and Thomas D Kirsch. The human impact
of floods: a historical review of events 1980-2009 and systematic literature review. PLoS
currents, 5, 2013.

[8] Brenden Jongman, Philip J Ward, and Jeroen CJH Aerts. Global exposure to river and coastal
flooding: Long term trends and changes. Global Environmental Change, 22(4):823–835,
2012.

[9] Zbigniew W Kundzewicz, Yukiko Hirabayashi, and Shinjiro Kanae. River floods in the
changing climate—observations and projections. Water Resources Management, 24:2633–
2646, 2010.

[10] Bruno Merz, Günter Blöschl, Sergiy Vorogushyn, Francesco Dottori, Jeroen CJH Aerts, Paul
Bates, Miriam Bertola, Matthias Kemter, Heidi Kreibich, Upmanu Lall, et al. Causes, impacts
and patterns of disastrous river floods. Nature Reviews Earth & Environment, 2(9):592–609,
2021.

[11] Oliver EJ Wing, Nicholas Pinter, Paul D Bates, and Carolyn Kousky. New insights into us flood
vulnerability revealed from flood insurance big data. Nature communications, 11(1):1444,
2020.

[12] M Altaf. Pakistan: Flood Damages and Economic Losses Over USD 30 billion Reconstruction

Needs over USD 16 billion–New Assessment. World Bank, 2022.

[13] Jose-Miguel Albala-Bertrand et al. Political economy of large natural disasters: with special

reference to developing countries. OUP Catalogue, 1993.

[14] Nejat Anbarci, Monica Escaleras, and Charles A Register. Earthquake fatalities: the inter-
action of nature and political economy. Journal of public economics, 89(9-10):1907–1933,
2005.

107

[15] Quan Sun, John Mann, and Mark Skidmore. The impacts of flooding and business activity
and employment: a spatial perspective on small business. Water Economics and Policy,
8(03):2140003, 2022.

[16] World Health Organization. Global health risks: mortality and burden of disease attributable

to selected major risks. World Health Organization, 2009.

[17] Sue M Tapsell and Sylvia M Tunstall. “I wish i’d never heard of Banbury”: the relationship
between ‘place’and the health impacts from flooding. Health & place, 14(2):133–154, 2008.

[18] Robert E Criss and Mingming Luo.

Increasing risk and uncertainty of flooding in the

Mississippi River basin. Hydrological Processes, 31(6):1283–1292, 2017.

[19] Samuel E Munoz and Sylvia G Dee. El Niño increases the risk of lower Mississippi River

flooding. Scientific reports, 7(1):1772, 2017.

[20] Debbie M Chizewer and A Dan Tarlock. New challenges for urban areas facing flood risks.

Fordham Urb. LJ, 40:1739, 2012.

[21] Pierfranco Costabile and Francesco Macchione. Enhancing river model set-up for 2-d dynamic

flood modelling. Environmental Modelling & Software, 67:89–107, 2015.

[22] F Dottori, G Di Baldassarre, and E Todini. Detailed data is welcome, but with a pinch of
salt: Accuracy, precision, and uncertainty in flood inundation modeling. Water Resources
Research, 49(9):6079–6085, 2013.

[23] Vijendra Kumar, Kul Vaibhav Sharma, Tommaso Caloiero, Darshan J Mehta, and Karan
Singh. Comprehensive overview of flood modeling approaches: A review of recent advances.
Hydrology, 10(7):141, 2023.

[24] Mehdi Nikoo, Fatemeh Ramezani, Marijana Hadzima-Nyarko, Emmanuel Karlo Nyarko, and
Mohammad Nikoo. Flood-routing modeling with neural network optimized by social-based
algorithm. Natural hazards, 82:1–24, 2016.

[25] Avantika Gori, Russell Blessing, Andrew Juan, Samuel Brody, and Philip Bedient. Char-
acterizing urbanization impacts on floodplain through integrated land use, hydrologic, and
hydraulic modeling. Journal of hydrology, 568:82–95, 2019.

[26] Barry Hankin, Peter Metcalfe, Keith Beven, and Nick A Chappell. Integration of hillslope
hydrology and 2D hydraulic modelling for natural flood management. Hydrology Research,
50(6):1535–1548, 2019.

[27] Hapu Arachchige Prasantha Hapuarachchi, Mohammed Abdul Bari, Aynul Kabir, Moham-
mad Mahadi Hasan, Fitsum Markos Woldemeskel, Nilantha Gamage, Patrick Daniel Sunter,
Xiaoyong Sophie Zhang, David Ewen Robertson, James Clement Bennett, et al. Development
of a national 7-day ensemble streamflow forecasting service for Australia. Hydrology and
Earth System Sciences, 26(18):4801–4821, 2022.

108

[28] Kaustav Mondal, Soumya Bandyopadhyay, and Subhankar Karmakar. Framework for global
sensitivity analysis in a complex 1d-2d coupled hydrodynamic model: Highlighting its im-
portance on flood management over large data-scarce regions. Journal of Environmental
Management, 332:117312, 2023.

[29] Gary W Brunner. HEC-RAS River Analysis System: Hydraulic Reference Manual, version

5.0. US Army Corps of Engineers–Hydrologic Engineering Center, 547, 2016.

[30] Alexander Betsholtz and Beatrice Nordlöf. Potentials and limitations of 1d, 2d and coupled
1d-2d flood modelling in hec-ras: A case study on höje river. Lund University: Lund, Sweden,
page 128, 2017.

[31] Elena Huţanu, Alin Mihu-Pintilie, Andrei Urzica, Larisa Elena Paveluc, Cristian Constantin
Stoleriu, and Adrian Grozavu. Using 1d hec-ras modeling and lidar data to improve flood
hazard maps accuracy: A case study from jijia floodplain (ne romania). Water, 12(6):1624,
2020.

[32] Armas Andrei, Beilicci Robert, and Beilicci Erika. Numerical limitations of 1d hydraulic
models using mike11 or hec-ras software–case study of baraolt river, romania.
In IOP
Conference Series: Materials Science and Engineering, volume 245, page 072010. IOP
Publishing, 2017.

[33] YG Lai.

SRH-2D version 2: Theory and user’s manual. Sedimentation and river
hydraulics—Two-dimensional river flow modeling. US Department of the Interior, Bureau
of Reclamation, Technical Service Center, Denver, CO, 2008.

[34] Basile Lavoie and Tew-Fik Mahdi. Comparison of two-dimensional flood propagation models:

SRH-2D and hydro_AS-2D. Natural hazards, 86:1207–1222, 2017.

[35] Keighobad Jafarzadegan, Hamid Moradkhani, Florian Pappenberger, Hamed Moftakhari, Paul
Bates, Peyman Abbaszadeh, Reza Marsooli, Celso Ferreira, Hannah L Cloke, Fred Ogden,
et al. Recent advances and new frontiers in riverine and coastal flood modeling. Reviews of
Geophysics, page e2022RG000788, 2023.

[36] Jeffrey J Danielson, Sandra K Poppenga, John C Brock, Gayla A Evans, Dean J Tyler, Dean B
Gesch, Cindy A Thatcher, and John A Barras. Topobathymetric elevation model development
using a new methodology: Coastal national elevation database. Journal of Coastal Research,
(76):75–89, 2016.

[37] Jonathan WF Remo, Megan Carlson, and Nicholas Pinter. Hydraulic and flood-loss modeling
of levee, floodplain, and river management strategies, middle mississippi river, usa. Natural
hazards, 61:551–575, 2012.

[38] Jayme M Stone, James T Rogala, Stephanie R Sattler, and Jenny L Hanson. Upper mississippi
river restoration (umrr) project tier 3 topobathy data for mississippi river navigation pools 5a,
6, 10, 11, 12, 14, 15-28, 20, 22, and 24. 2016.

[39] Wolfgang Rodi. Turbulence Models and Their Application in Hydraulics. Routledge, London,

2000.

109

[40] John Doherty. PEST: Model-independent Parameter Estimation. User Manual, 5th edition.

Watermark Numerical Computing, Brisbane, 2008.

[41] Siddharth Saksena, Venkatesh Merwade, and Peter J. Singhofen. Flood inundation modeling
and mapping by integrating surface and subsurface hydrology with river hydrodynamics.
Journal of hydrology, 575:1155–1177, 2019.

[42] A. Safaie, C. Dang, H. Qiu, H. Radha, and M.S. Phanikumar. Manifold methods for assim-
ilating geophysical and meteorological data in earth system models and their components.
Journal of Hydrology, 544:383–396, 2017.

110