University of Southern California Dissertations and Theses

Spatiotemporal analysis of the SLOSH and ADCIRC storm surge models: a case study of hurricane Ida

(USC Thesis Other)

Spatiotemporal analysis of the SLOSH and ADCIRC storm surge models: a case study of hurricane Ida

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright © 2022 John Carlson Bowers Jr.
Spatiotemporal Analysis of the SLOSH and ADCIRC Storm Surge Models
A Case Study of Hurricane Ida
by
John Carlson Bowers Jr.
A Thesis Presented to the
FACULTY OF THE USC DORNSIFE COLLEGE OF LETTERS, ARTS AND SCIENCES
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)
December 2022
ii

To my wife and boys. I love you.
iii

Acknowledgements
I would like to eternally thank every person at USC that has helped me through this journey.
From the acceptance phone call by Mr. Sabey, to the Catalina fieldwork with Dr. Marx, to my
thesis advisor Dr. Sedano, thesis committee members Dr. Huang and Dr. Ruddell, and to Drs.
Bernstein, Loyola, and Swift, thank you for challenging me both mentally and academically.
These years I will never forget.

iv

Table of Contents
Dedication……… …………………………………………………………………………………ii
Acknowledgements...…………………………………………………………………………….iii
List of Tables ................................................................................................................................ vii
List of Figures .............................................................................................................................. viii
Abbreviations ................................................................................................................................. xi
Abstract ........................................................................................................................................ xiii
Chapter 1 Introduction .................................................................................................................... 1
1.1. Hurricane Ida ......................................................................................................................2
1.2. Study Area ..........................................................................................................................7
1.3. Storm Surge Modeling ........................................................................................................9
1.4. Hypothesis.........................................................................................................................10
1.5. Thesis Overview ...............................................................................................................11
Chapter 2 Related Work................................................................................................................ 12
2.1. Natural Disasters and Climate Change .............................................................................12
2.1.1. Temperature Increase ...............................................................................................12
2.1.2. Hurricane Season .....................................................................................................14
2.1.3. Precipitation .............................................................................................................16
2.1.4. Damage ....................................................................................................................17
2.2. Numerical Storm Surge Modeling ....................................................................................17
2.2.1. SLOSH .....................................................................................................................19
2.2.2. ADCIRC ..................................................................................................................22
2.2.3. Data-Driven Models.................................................................................................24
2.2.4. Artificial Intelligence Models ..................................................................................25
2.3. Spatial Analysis and Validation ........................................................................................26
v

2.3.1. Hindcast ...................................................................................................................26
2.3.2. Other Validation Methods........................................................................................27
Chapter 3 Methods ........................................................................................................................ 29
3.1. Methods Overview ............................................................................................................30
3.2. Data ...................................................................................................................................31
3.2.1. Louisiana Parish Boundary Data .............................................................................32
3.2.2. Land Cover Dataset..................................................................................................33
3.2.3. Digital Elevation Model ...........................................................................................34
3.2.4. SLOSH Storm Surge Prediction Data ......................................................................35
3.2.5. ADCIRC Storm Surge Prediction Data ...................................................................36
3.2.6. Observed Flooding Data Acquisition .......................................................................37
3.3. Data Preparation................................................................................................................39
3.3.1. Coordinate System and Projection ...........................................................................40
3.3.2. Land Extent Preparation ..........................................................................................40
3.3.3. SLOSH Preparation .................................................................................................41
3.3.4. ADCIRC Preparation ...............................................................................................44
3.3.5. Observed Flooding Preparation ...............................................................................44
3.4. Data Analysis ....................................................................................................................46
3.4.1. Flood Extent Analysis ..............................................................................................46
3.4.2. SLOSH and ADCIRC Flood Depth Analysis ..........................................................47
Chapter 4 Results .......................................................................................................................... 49
4.1. Pre-Landfall Flooding Predictions: NHC Advisories 9-14 ...............................................50
4.1.1. Pre-Landfall Flood Extent Analysis.........................................................................50
4.1.2. Pre-Landfall 9-14 Depth Analysis ...........................................................................52
4.2. Post-Landfall Flooding Predictions: NHC Advisories 15-17 ...........................................53
vi

4.2.1. NHC Advisories 15-17 Flood Extent Analysis ........................................................53
4.2.2. Post-Landfall Depth Analysis ..................................................................................55
4.3. Overall Extent Analysis ....................................................................................................57
4.4. Overall Depth Analysis .....................................................................................................59
4.5. Locality Analysis ..............................................................................................................60
Chapter 5 Discussion .................................................................................................................... 67
5.1. Storm Surge Analysis .......................................................................................................67
5.2. Research Limitations ........................................................................................................68
5.2.1. Flooding Data...........................................................................................................69
5.2.2. Storm Surge Data .....................................................................................................69
5.2.3. Environment .............................................................................................................70
5.2.4. Temporal Scale ........................................................................................................71
5.3. Future Research ................................................................................................................71
5.4. Conclusion ........................................................................................................................72
References ..................................................................................................................................... 73
Appendix A Surge Extent by Advisory ........................................................................................ 78
Appendix B Surge Depth by Parish .............................................................................................. 82
Appendix C Surge Depth by Advisory ......................................................................................... 90

vii

List of Tables
Table 1 Table with differences between SLOSH and ADCIRC……………….………………..24
Table 2 Timeline series………………………………………………………………………….29
Table 3 Data table………………………………………………………………………………..31

Table 4 Advisories 9-14 areal and percentage result analysis.…………………………….…….52

Table 5 Advisories 15-17 areal and percentage result analysis …………………………………56

Table 6 Depth difference chart in NHC advisories 9-14……..………………………………….83

Table 7 Depth difference chart in NHC advisories 15-17……………………………………….84

viii

List of Figures
Figure 1 Hurricane Ida and Katrina track comparison………………..…………………………...2

Figure 2 Wetlands of southeastern Louisiana ……………..……………………………………...5

Figure 3 Storm surge effects in Golden Meadow, Louisiana….………………………………….6

Figure 4 Total precipitation amounts from Hurricane Ida ………………………………………..7

Figure 5 Study area of southeastern Louisiana and surrounding areas …………………………...9

Figure 6 Bodies of water throughout the study area.…………………………………………….10

Figure 7 SLOSH mesh basins in use……………………………………………………………..20

Figure 8 ADCIRC’s vertically integrated continuity equation.………………………………….22

Figure 9 Example ADCIRC grid mesh……………………….…………..……………………...23

Figure 10 Simplified extent workflow………..…………………………………..……………...30

Figure 11 Simplified depth analysis workflow.………………………………………………….31

Figure 12 Louisiana parish boundaries……….………………………………………………….33

Figure 13 Raw land cover dataset…….………………………………………………………….34

Figure 14 10m DEM of the study area.…………………………………………………………..35
Figure 15 SLOSH advisory 9 dataset.……………………………………………………………36
Figure 16 ADCIRC data reflecting NHC advisory 9.……………………………………………37
Figure 17 Land extent of study area.…………………………………………………………….41
Figure 18 SLOSH preparation workflow………………………..……………………………….42
Figure 19 SLOSH surge simulation data from NHC advisory 9. ……..………………..……….43
Figure 20 Raster to polygon output and clip…………….……………………………………….44
Figure 21 ADCIRC workflow………………………………………………………….………..45
Figure 22 Observed flooding workflow………...………………………………………………..46
ix

Figure 23 Extent of flood waters during Hurricane Ida……..…………………………………...46
Figure 24 FwDET flow chart…………....……………………………………………………….48
Figure 25 Pre-Landfall Advisory 9 Flood Extent …..…………………………………………...52
Figure 26 Depth Analysis for NHC Advisory 9…………………………………....……………53

Figure 27 Pre-Landfall depth analysis in feet……………...…………………………………….54
Figure 28 Post-Landfall Advisory 15 Flood Extent ……………………………………………..55
Figure 29 Depth Analysis for NHC Advisory 15…………………..……………………………57
Figure 30 Post-Landfall depth analysis in feet…………………………………………………..58
Figure 31 SLOSH flood extent percentage analysis……………………………………………..59
Figure 32 ADCIRC flood extent percentage analysis……….…………………………………..59
Figure 33 Overall depth analysis in feet…………………………………………………………60
Figure 34 Locality analysis overview …………………………………………………………...61
Figure 35 Thibodaux, LA and Bayou Lafourche flood analysis...………………………………62
Figure 36 Alluvial City flood analysis………………………….………………………………..63
Figure 37 Jean Lafitte flood analysis ……………………………………………………………64
Figure 38 Highway 90 flood analysis.…………………………………………………………...65
Figure 39 Fontainebleau State Park flood analysis ……………………………………………...66
Figure 40 Advisory 10 Depth Analysis………………………………………………………….79
Figure 41 Advisory 11 Depth Analysis………………………………………………………….80
Figure 42 Advisory 12 Depth Analysis………………………………………………………….80
Figure 43 Advisory 14 Depth Analysis………………………………………………………….81
Figure 44 Advisory 16 Depth Analysis………………………………………………………….81
Figure 45 Advisory 17 Depth Analysis………………………………………………………….82
x

Figure 46 Jefferson Parish Depth Analysis ……………………………………………………...85
Figure 47 Lafourche Parish Depth Analysis …………………………………………………….86
Figure 48 Orleans Parish Depth Analysis ……………………………………………………….86
Figure 49 Plaquemines Parish Depth Analysis ………………………………………………….87
Figure 50 St. Bernard Parish Depth Analysis …………………………………………………...87
Figure 51 St. Charles Parish Depth Analysis …………………………………………………....88
Figure 52 St. John the Baptist Parish Depth Analysis …………………………………………..88
Figure 53 St. Tammany Parish Depth Analysis …………………………………………………89
Figure 54 Tangipahoa Parish Depth Analysis…………………………………………………...89
Figure 55 Terrebonne Parish Depth Analysis …………………………………………………...90
Figure 56 Depth Analysis for NHC Advisory 10......……………………………………………91
Figure 57 Depth Analysis for NHC Advisory 11……….....…………………………...………..92
Figure 58 Depth Analysis for NHC Advisory 12…………..…………………………...……….92
Figure 59 Depth Analysis for NHC Advisory 14…………....…………………………..………93
Figure 60 Depth Analysis for NHC Advisory 16………………………………………………..93
Figure 61 Depth Analysis for NHC Advisory 17………………………………………………..94

xi

Abbreviations
ADCIRC Advanced Circulation

AER Atmospheric and Environmental Research

AI Artificial intelligence

ANN Artificial neural network

AWSR2 Advanced Microwave Scanning Radiometer 2

BOE Brown ocean effect

CERA Climate Emergency Risk Assessment

CRC Climate Resilience Center

DFO Dartmouth Flood Observatory

FEMA Federal Emergency Management Agency

FwDET Floodwater Depth Estimation Tool

GCOM-W1 Global Change Observation Mission- Water 1

GFM Global flood mapping

GIS Geographic information system

GUI Graphical user interface

GWCE Generalized wave continuity equation

HAZUS Hazards US

LaDOTD Louisiana Department of Transportation and Development

MDL Meteorological Development Laboratory

MEOW Maximum envelopes of water

MFED Maximum flood extent depiction

MODIS Moderate Resolution Imaging Spectroradiometer
xii

MOM Maximum of MEOW

MOTF Modeling task force

NAD North American datum

NASA National Aeronautics and Space Administration

NHC National Hurricane Center

NOAA National Oceanic and Atmospheric Administration

NRT Near real time

NWS National Weather Service

P-Surge Probabilistic storm surge

SDP SLOSH display program

SFED Standard flood extent depiction

SLOSH Sea, Lake, and Overland Surges from Hurricanes

WGS World Geodetic System

xiii

Abstract
Hurricane Ida struck southeastern Louisiana with winds greater than Hurricane Katrina, and truly
tested the rebuilt levee systems of New Orleans and lower Louisiana. Weather and governmental
agencies used predictive models to anticipate and predict storm surge locations and severity, as
storm surge from hurricanes is the leading fatality cause globally during tropical events. Current
tropical systems are fueled by climate change that is impacting storm strength and regularity, yet
storm surge models must ensure the highest degree of accuracy. The National Oceanic and
Atmospheric Administration’s Sea, Lake, and Overland Surges from Hurricanes (SLOSH) and
the Climate Resilience Center 's Advanced Circulation (ADCIRC) models are two preeminent
models that simulate and predict storm surge in an effort to publish evacuation orders and save
lives. Hurricane Ida rapidly intensified prior to landfall, which stressed the ability of agencies to
properly predict storm impacts. In this project, SLOSH and ADCIRC were tested and evaluated
against each other, and against observed flooding during Hurricane Ida to determine model
strengths and weaknesses. Results show that both SLOSH and ADCIRC overestimated storm
surge extent, but underestimated surge depth. ADCIRC was more accurate in long range
forecasting, while SLOSH was more accurate in short range forecasting. Recommendations are
posed to enhance the accuracy of current storm surge models. This spatiotemporal analysis can
help validate surge models against climate enhanced storms with the goal of saving as many
lives as possible.
1

Chapter 1 Introduction
The deadliest part of any tropical system is not wind speed, but the flooding that destroys
property and causes hundreds of fatalities every year (NHC 2022). Perhaps the single greatest
flooding event caused by a tropical system is storm surge, the rise in seawater caused by tropical
winds pushing water onshore (Didlake 2020). Tropical systems have increased in regularity and
intensity over the past few decades due to climate change, which is warming the oceans that
provide fuel for the storms (C2ES 2020). Accurate storm surge modeling is crucial for the public
to take steps to protect their property, livelihood, and families, while inaccuracies can have
catastrophic consequences. With climate change fueling ever more dangerous, powerful, and
irregular storms, storm surge models must ensure the highest degree of accuracy.
Hurricane Ida traveled a similar path as Hurricane Katrina (Figure 1) and represented a
true test of New Orleans and lower Louisiana’s rebuilt levee systems in the wake of Katrina.
Hurricane Katrina caused numerous levee breaches across the state and in particular, New
Orleans in 2005. In the years following Katrina, nearly 15 billion dollars was spent to rebuild and
shore up the levee system throughout New Orleans and lower Louisiana (Schleifstein 2020).
Hurricane Ida made landfall on August 29, 2021 and brought with it 150 mph winds and a storm
surge up to 14 feet. Fortunately, the levee system withstood Ida’s onslaught, and the city was
spared the devastation brought by Katrina 16 years earlier. Hurricane Ida was influenced by
climate change in that it rapidly intensified prior to landfall. The irregularity of the storm’s rapid
intensification prior to landfall and the storm’s sustainment of its strength far after making
landfall enabled a swift pounding of the state which caused over 36 billion dollars in damage,
over 3/4 of the state’s total annual budget.
2

Figure 1. Hurricane Ida and Katrina track comparison
The rapid intensification of the storm, the sustainment of maximum wind speeds hours
after making landfall, and a track similar to Katrina makes Hurricane Ida ripe for research for
current modeling techniques in use today. Hurricane Ida’s meteorological history spans from
August 14 to September 1, 2021 (Beven, Hagan, and Berg 2022). This case study focuses on the
three days of August 28 to 30, Hurricane Ida’s peak intensity on the Louisiana coast. This case
study of Hurricane Ida analyzes current storm surge modeling techniques and poses
recommendations for future advancements.
1.1. Hurricane Ida
Hurricane Ida struck Louisiana with strong winds and a devastating storm surge that
flooded vast amounts of lower Louisiana (Beven, Hagan, and Berg 2022). Storm surge is the
abnormal raise in water due to storms and tropical systems (NOAA 2022). Beginning as a
tropical wave off the west coast of Africa, Ida raced westward and maintained its status as a
3

tropical wave until reaching tropical depression status; a tropical system with winds less than 39
mph, on August 26 in the southwestern Caribbean Sea. Six hours later the storm reached tropical
storm strength and the next day reached hurricane strength. The storm struck Cuba as a category
1 hurricane on August 27 and entered the Gulf of Mexico early on the 28th.
Ida began a period of rapid intensification upon entering the Gulf of Mexico, where the
waters were 86 degrees with minimal windshear, a near perfect scenario for intensification.
Windshear is a difference in wind speeds over a short distance in the atmosphere and is a culprit
of weakening tropical systems (Corbosiero and Molinari 2002). After 12 hours in the Gulf, Ida
intensified from 80 to 105 mph. Ida further intensified from 105 to 150 mph over the next 12
hours, a total of 70 mph increases in maximum windspeeds in 24 hours.
Ida made landfall at 11:15 a.m. Central Daylight Time on August 29 at Port Fourchon,
tied with the 1856 “Last Island” hurricane as the strongest to ever hit the state of Louisiana west
of the mouth of the Mississippi River. The storm continued a north-northwest track, with the eye
passing between New Orleans and Houma.
The hurricane maintained category 4 strength for four hours after landfall, and category 3
status a further four hours. This intensity can be attributed to the brown ocean effect (BOE).
BOE is a phenomenon in which a tropical system can maintain strength or strengthen while over
land due to the heat given off by moisture-rich ground (Sheppard and Andersen 2017). BOE can
occur where heated marshes and wetlands (figure 2) give off latent heat similar to the ocean,
allowing storms to retain strength (Andersen and Sheppard 2017). Though not specifically
relating to Hurricane Ida, Andersen and Sheppard show that areas similar to lower Louisiana
contain the necessary geography to allow a hurricane to retain strength after landfall, like
Hurricane Ida. Ida rapidly weakened as it tracked further inland and lost contact with the marshy
4

wetlands. The storm became extra-tropical, or absent its tropical characteristics, where cool
surface air replaces the warm tropical air, as it continued through Mississippi and Tennessee.

Figure 2. Wetlands of southeastern Louisiana
Hurricane Ida produced a devastating storm surge (figure 3) that reached 14 feet above
ground level, and in some instances, was able to overcome levee systems. The largest storm
surge was recorded in Plaquemines parish on the eastern bank of the Mississippi River. Around
New Orleans and Lake Pontchartrain, a high storm surge of 8.5 feet was recorded, far less than
Katrina but very dangerous, nonetheless.
5

Figure 3. Storm surge effects in Golden Meadow, Louisiana (Photo Credit: Luke Sharrett)
Hurricane Ida also brought massive amounts of precipitation throughout southeastern
Louisiana and Mississippi. The largest rainfall total measured over 15 inches in Ponchatoula,
Louisiana. The storm further produced large precipitation totals throughout the southern United
States, and even produced upwards of 10 inches of rain in Pennsylvania and New Jersey while
tracking as an extra-tropical system (Figure 4).
6

Figure 4. Total precipitation amounts from Hurricane Ida (David Roth/ NOAA 2022)
Climate change can be considered a factor with Hurricane Ida, due to its rapid
intensification and strength sustainment post-landfall. Most hurricanes intensify gradually, even
over warm water, however, Hurricane Ida’s wind speed increased from 105 to 150 mph in a
matter of hours, an almost unseen intensification rate (Zhe et al. 2022). Hurricane Ida also
maintained its strength for 6-8 hours after landfall, another aspect that can be attributed to
climate change A study conducted of 71 North Atlantic hurricanes found that 50 years ago
hurricanes would lose 75% of their windspeed 24 hours post landfall. Current hurricanes are now
losing just 50% of their windspeeds 24 hours post landfall, suggesting hurricane decay is slowing
in the warming oceans (Li and Chakraborty 2020). Another climate change anomaly observed
during Hurricane Ida was the BOE, which can be exacerbated by the warming climate as tropical
7

systems can maintain or even strengthen over land, an anomalous event within nature (Sheppard
and Andersen 2017).
In total, Hurricane Ida was directly responsible for 55 fatalities and a further 32 indirectly
throughout Louisiana and the rest of the United States (Beven, Hagan, and Berg 2022). Over
60% of fatalities were related to drowning. Nearly 20% of deaths were reported after Ida’s
passing and were due to power outages that lead to heat exhaustion and stroke in the hot climate.
The storm’s name was retired on April 22, 2022, due to massive damage and loss of life. The
name Ida will never be used for an Atlantic storm again.
1.2. Study Area
Louisiana is no stranger to tropical storms and hurricanes. The state is bordered by Texas
to the west, Arkansas to the north, Mississippi to the east, and the Gulf of Mexico to the south.
The warm waters of the Gulf are a main reason Louisiana ranks fourth in the United States for
total number of hurricane landfalls, and third in major hurricane landfalls from 1851 through
2018 (Heil 2019). The specific study area for this research project is the southeastern portion of
the state, containing 10 parishes, including the major city of New Orleans, depicted by Figure 5.
8

Figure 5. Study area of southeastern Louisiana and surrounding areas
New Orleans is situated between Lake Pontchartrain and the Mississippi River, which has
made the city prone to flooding throughout its history. With a population of just under 400,000,
New Orleans is the largest city in the United States below sea level, with only man-made levees
and flood walls protecting it from sustained flooding.
The study area’s proximity to the Gulf of Mexico, numerous lakes, sounds, and wetlands
provides geographic vulnerabilities and ample opportunity for tropical events and extreme
flooding. The water surroundings increase potential water volume during storm surge events.
Figure 6 represents the study area and the numerous bodies of water that influence storm surge
and flooding.
9

Figure 6. Bodies of water throughout the study area
1.3. Storm Surge Modeling
The United States’ premier weather agency, the National Oceanic and Atmospheric
Administration’s (NOAA) National Hurricane Center (NHC) creates, publishes, and updates
storm surge maps for the United States, Mexico, and Caribbean. Such maps, which are not tied to
a particular storm, show storm surge by a hurricane’s maximum sustained windspeed category,
the Saffir-Simpson scale, from 1 to 5, and shows the likelihood of flooding caused by the storm’s
intensity. Though storm winds are a major factor in storm surge intensity, other factors influence
storm surge, such as coastline orientation, storm intensity, size, speed, and local bathymetry.
Users can view the storm surge maps interactively, allowing users to manipulate the model to
find and observe flooding likelihood for any area of interest. At 48 hours prior to landfall,
NOAA publishes site- and storm-specific surge maps, with 6-hour updates afterwards, otherwise
known as an advisory.
10

A NHC public advisory contains information about a tropical system to warn the public
about the storm. The advisory contains a summary of the storm, any watches or warnings
associated with the storm, outlook about the storm, and hazards associated with the storm (NHC
2022). NHC publishes advisories every 6 hours and SLOSH outputs are generated based on
those advisories.
NOAA utilizes the SLOSH model exclusively for all its storm surge maps, while other
agencies have created separate storm surge models and maps to increase accuracy and public
awareness of flooding. The Climate Resilience Center (CRC), a joint academic, industrial, and
governmental effort, created the Advanced Circulation (ADCIRC) model to enhance the
resilience of the nation’s people, infrastructure, economies, and the natural environment
(Leuttich and Westerink 2012). ADCIRC simulations are made publicly available on the Coastal
Emergency Risk Assessment (CERA) website, with similar interactive functions as NOAA’s
SLOSH site.
The SLOSH modeling approach used in this project is P-Surge, which accounts for
uncertainty in wind forecasts that primarily drives storm surge. P-Surge modeling parameters
include the NHC’s official storm forecast, parametric wind model, and historical 5-year error
statistics in track, size, and intensity. The ADCIRC model in this project uses basic parameters
such as wind forecasts, atmospheric pressure, and rain. ADCIRC also incorporates other
parameters like coastal bathymetry, temperature, and water salinity. The SLOSH and ADCIRC
models are described extensively in the following chapter.
1.4. Hypothesis
The hypothesis for this thesis project was that the SLOSH and ADCIRC storm surge
models would underestimate flooding during Hurricane Ida due to the idiosyncrasies caused by
11

climate change. A spatiotemporal analysis was conducted in and around New Orleans to show
predicted SLOSH and ADCIRC storm surge extents and depths of eight NHC advisories,
compared with observed flooding. Within the study area, flooded regions were calculated per
square mile and by percentage of the overall land area within the study area. Additionally, flood
depth analysis of SLOSH and ADCIRC was performed in the study area, to compare surge depth
averages across the study area. Comparing and contrasting flood areas by model and by advisory
determined which model produced more accurate results.
1.5. Thesis Overview
This thesis project is structured with four additional chapters. Chapter Two describes
previous related literature on climate change, storm surge modeling, and validation of surge
models. Chapter Three describes the data and methods employed. Chapter Four details the
results of the technical work completed. Chapter Five is oriented towards a discussion on storm
surge techniques with recommendations for future endeavors.

12

Chapter 2 Related Work
Three main areas have been identified for prior research on this thesis project. The first section
details the linkage between natural disasters and climate change, which is an influencer on
tropical systems and their intensity and regularity. The second section describes storm surge
modeling and predictions, which must be relayed to show how agencies have used various surge
models in the past to warn the public. Storm surge modeling can take many different forms and
processes, but three main techniques are commonly used: numerical models, data-driven models,
and artificial intelligence models. The third section covers spatial analysis and validation, which
will show previous endeavors to validate surge predictions from historical storms and the steps
taken to achieve greater accuracy.
2.1. Natural Disasters and Climate Change
A consensus is forming in the academic community that changing climate is influencing
natural disasters, their regularity, and their intensity. This section overviews climate change’s
impact on the environment and the effect it is having on tropical systems. Increases in global
temperatures, earlier than usual tropical system development, increases in storm precipitation,
and damage proliferation are covered as a part of climate change effects.
2.1.1. Temperature Increase
The International Panel on Climate Change (IPCC) releases data periodically that
simulates future climate scenarios based on Representative Concentration Pathways (RCPs).
RCP figures estimate changes to future climate trends based on four different scenarios of future
greenhouse gas emissions. RCP rates range from 1.9 at the lower end, assuming green gas
emissions stop almost immediately, to 8.5, assuming current greenhouse gas emissions stay on
13

their current course, with intermediate RCPs in between. The RCP rates directly correlate with
warming temperatures and warming ocean temperatures, which will affect hurricane strength
(IPCC 2022).
The current IPCC report, the sixth edition published in 2022, gives an assessment on an
increase of 1.5 degree Celsius in the near term by 2040 for RCP 1.9. Though this represents a
best-case scenario, the temperature increase still comes with dire consequences. A 1.5 degree
increase would cause a nearly unavoidable increase in climate hazards (IPCC 2022). The IPCC
report further emphasizes that at current warming trends, hurricanes will continue to increase in
frequency through at least 2050, which will increase economic loss in the United States and
globally and may cause migration from coastal areas inland as people flee yearly destruction.
The IPCC report provides case studies and describes future tropical system effects caused
by an RCP of 1.9. First, the report studies Hurricane Harvey, which struck Houston on August
26, 2017, and reports that the record-breaking precipitation recorded was three to ten times more
likely to have occurred due to climate change (IPCC 2022). The report also states that damage
along the Gulf of Mexico will increase in the future due to hurricane frequency, strength, and
coastal development. Finally, the report estimates that flooding will increase due to flood zone
expansion as a result of increasing hurricanes and tropical systems (IPCC 2022).
The IPCC report presents grave scenarios, even if greenhouse emissions are curbed, at
least in the short term. Above all, the earth’s temperature is warming, which will affect the
frequency and intensity of hurricanes over the next few decades at a minimum.
A separate study conducted by researchers at NOAA highlights the differences in average
annual global temperatures from 1880-2021. Earth’s average temperature has steadily increased
about 0.08 degrees Celsius per decade from 1880-1991. However, the past four decades have
14

seen increases double, to the rate of about 0.18 degrees Celsius (Dahlman and Lindsey 2022).
Further, when accounting for sea surface temperature in addition to dry surface temperature,
2021 average temperatures are 0.84 degrees Celsius warmer than the twentieth-century average.
Looking at the globe’s average yearly temperature, the most recent decade holds some of
the warmest years on record. In fact, 2020 was the third warmest year since 1900, and 2021 was
the sixth warmest (Dahlman and Lindsey 2022). A trend is developing, particularly after the
industrial revolution, that is seeing temperatures skyrocket. Projections of current warming
trends show that by the year 2100, the Earth may be nearly 3 degrees Celsius warmer than
averages in the mid twentieth century.
2.1.2. Hurricane Season
Tropical systems derive their strength from warm ocean water. Climate change has been
warming the oceans for decades and is thus causing more disastrous hurricanes throughout the
globe. Not only are warming oceans increasing strength and regularity in tropical systems, but
historical date ranges in which they occur have been expanding.
The Atlantic hurricane season runs from June 1 to November 30 yearly, a range date that
was chosen in 1965 by the NHC. The June to November hurricane season was chosen due to
favorable conditions for tropical system development (Truchelut et al. 2022). The warming
oceans are experiencing an increase in sea surface temperatures, which is expanding the
favorable condition windows in the North Atlantic, particularly on the front and back end of
hurricane seasons (Truchelut et al. 2022). Sea surface temperatures have risen nearly 1 degree
Celsius since 1980, potentially making the months of May and April favorable for tropical
development. In fact, from 2012-2020 seven tropical systems formed outside of the hurricane
15

season range and were close enough to the US mainland to warrant watches and/ or warnings
(Truchelut et al. 2022).
By observing the first and last tropical development since 1950 and then assessing the
99% confidence interval of their date ranges, new potential dates are discovered that could
expand the current Atlantic hurricane season. Tropical system development, with the influence of
warming oceans, has a favorable condition range of mid-May to mid-December, with those dates
slowly expanding due to climate change. Truchelut’s article shows that warming oceans are
expanding the time tropical systems have favorable conditions for development in the Atlantic
Ocean.
A study of storm distribution identified anomalies of storm development outside of the
hurricane season. The study focused on storm development from 1851-2007, at a region south of
30 degrees north latitude and east of 75 degrees west latitude, which is the main formation region
of Atlantic hurricanes (Kossin 2008). Trends show development range dates are increasing
outside of the hurricane season, to the rate of about 1 day per decade on both the front and back
end of the season.
Correlation to sea surface temperature can be a reason for hurricane development ranges
increasing as the ocean is becoming warmer yearly, particularly in the date range not usually
associated with hurricane development. It is further posited that a one-degree Celsius increase in
sea surface temperature can correspond to a 20 day increase on the front and back end of
hurricane season. A 20-day change would see the hurricane season shift from May 10 through
December 20.
16

2.1.3. Precipitation
The warming climate and warming sea surface temperature are supercharging hurricanes
and increasing their lethality (Trenberth et al. 2018). In addition to making hurricanes bigger and
stronger, climate change is affecting precipitation levels in hurricanes, thereby increasing
flooding and surge potential.
A study conducted on Hurricane Harvey analyzes the link between sea surface
temperature and rainfall totals in 2017. The 2017 hurricane season is known as a record-breaking
season due to sheer numbers of storms and the damage inflicted. NOAA measures hurricane
seasons by accumulated cyclone energy (ACE), a measuring tool which accounts for different
metrics to analyze a storm season. 2017 measured at 225% ACE, well above normal for typical
hurricane seasons (Trenberth et al. 2018).
Hurricane Harvey dumped up to 60 inches of rain in several locations in southern Texas,
creating massive flooding over large inland areas. Trenberth conducted a study which analyzed
sea surface temperatures in the Gulf of Mexico, along with rainfall potential and discovered that
climate change affected Hurricane Harvey’s precipitation totals by 20-26% and rainfall totals in
most other tropical systems by 5-15%.
Extreme precipitation events have shown a correlation with the warming climate. By
observing recordings of precipitation over land areas, and by combining climate model outputs,
precipitation sensitivity can be calculated. Intensification of precipitation extremes has been
increasing in recent years, correlating with the rise in surface temperature (O’Gorman 2015). In
addition, extreme precipitation in tropical areas appears to be the most affected by climate
change, which could be a reason tropical systems are being observed with more precipitation, as
tropical regions fuel hurricane growth.
17

2.1.4. Damage
A study conducted by Pant and Cha on future damage potential for hurricanes highlights
the increasing danger due to continuing effects of climate change. Using the IPCC’s future
climate scenario, along with input into the Hazard US (HAZUS) software, loss potential was
calculated at select major cities across the United States (Pant and Cha. 2019).
Using the IPCC’s representative concentration pathway (RCP) of 8.5, which shows
climate change continuing its current track, ocean temperatures are set to increase anywhere
from about 2 to 4 degrees Celsius (Pant and Cha. 2019). This ocean temperature increase will
increase the lethality of tropical systems, which in turn will increase damage across coastal cities
in the United States.
Maximum windspeeds for hurricanes projected in the year 2100, show up to 200mph
being a semi-common occurrence, which put those future storms well above the category 5
threshold. Results suggest economic loss to be more than double today’s losses in 2100 with
future hurricane scenarios. In particular, coastal cities south of Virginia Beach and along the Gulf
of Mexico show three- and four-fold increases in damage loss in year 2100 future storm
scenarios (Pant and Cha. 2019).
2.2. Numerical Storm Surge Modeling
Numerical storm surge modeling, sometimes referred to as physics modeling, is
performed by a set of physics equations or algorithms applied to a grid that computes wave
heights within the grid. In numerical surge modeling, a few governing equations form the basis
for surge predictions. The Shallow Water Equations (SWE) and the Navier-Stokes equation of
motion are two equations widely used in surge modeling (Dube et al. 2010). Those equations can
then be applied to a grid mesh system of an area’s coastline, which can form the basis of surge
18

predictions. Two different techniques of mesh grids are commonly used in conjunction with the
governing equations, structured and unstructured.
One of the first statistical surge models was developed by Venkatesh (1974) in a study of
storm surge on the Great Lakes. Venkatesh employed a multiple regression model, with
dependent variables being vertically averaged water velocity and water elevation departure from
undisturbed depth, where historical water pressure and temperature data on grid points
throughout the Great Lakes were input into an equation to predict wave heights (Venkatesh
1974).
Early numerical techniques employed finite difference (FD) schemes to model the
Navier-Stokes equation of motion. An FD scheme is a numerical technique for solving equations
by approximating derivatives with finite differences (Grossmann, Roos, and Stynes 2008). FD
schemes are widely used in surge modeling and perform well in most circumstances (Dube et al.
2010). FD models employ structured grids, which can hinder the grid’s ability to refine more
complex coastal geometry and local flood patterns near the coast (Dube et al. 2010). Structured
grids refer to the mesh basins within a surge model where surge calculations are performed.
Structured grids are discrete shapes where surge calculations occur. Essentially, FD schemes in
conjunction with structured grids are adequate, but are not as accurate in complex coastal
geometry, like areas in lower Louisiana.
Somewhat newer unstructured grids have been used with a Finite Element (FE) scheme
to solve the SWE. An FE scheme is a numerical method where differential equations are solved
in multiple space variables (Bickford 1994), like an unstructured mesh grid. FE and unstructured
grids allow for high resolution of small coastal features and can map complex coastal geometry
well (Westerink et al. 1994). Unstructured grids are continuous and can change based on the
19

surge model employed, which can give more accurate resolutions in areas as defined by the user.
FE schemes, though possessing more resolution than FD, require large computational resources
and time constraints, both disadvantages in surge forecasting (Dube et al. 2010).
The two storm surge models employed herein, SLOSH and ADCIRC, are numerical
models that employ FD and FE schemes within their modeling code. SLOSH employs a
structured grid mesh with FD and is thus not as accurate in complex coastal areas. ADCIRC
utilizes FE schemes with unstructured grids, giving ADCIRC higher resolution in coastal inlets
and barrier islands. As an FE, ADCIRC requires more resources and time to run simulations as
compared to SLOSH.
2.2.1. SLOSH
SLOSH is the current model used by NOAA and the NHC. At its core, SLOSH is a
computerized physics-based model that predicts storm surge based on past and hypothetical
predicted storms by considering certain parameters of the tropical system (NHC 2022). Those
parameters are used to create a model of the storm’s wind field which drives the storm surge.
SLOSH predicts storm surge by using three different approaches: deterministic, probabilistic,
and composite. Though SLOSH is a single model, the different approaches used within it can
make SLOSH seem like a family of related models.
Simply put, the deterministic approach with SLOSH involves mathematical equations
that are applied to an area’s coastline, by incorporating the unique terrestrial and aquatic micro
terrain and other physical features (NHC 2022). The equations are applied to a coastline by
incorporating a grid mesh, known as a basin, covering the forecast area. The model itself solves
the Navier-Stokes equation of motion.
20

The mesh basins used in the SLOSH model are based upon polar, hyperbolic, and
elliptical grids with momentum points at the corners of each grid and wave heights at the center
of each grid. The mesh basin used for this project is ms8, which covers the Louisiana coastline.
Figure 7 denotes the most updated mesh basins in the United States, Mexico, and Caribbean, as
of January 1, 2020.

Figure 7. SLOSH mesh basins in use (NHC 2022)
The probabilistic method, also known as P-Surge, incorporates statistics of past forecast
performances to generate an ensemble of different SLOSH model runs. P-Surge also accounts for
the uncertainty in wind speed forecasts, mitigating the inherent error in wind speed predictions.
Each ensemble member’s input within P-Surge is derived from current hurricane forecasts along
with the associated five-year averages of track, cross track, and intensity errors. Essentially,
21

within a P-Surge model, a five-year statistical error weight is added to the model to output a
statistical storm surge prediction.
The composite method involves running the SLOSH model several thousand times with
hypothetical hurricanes under varying storm conditions. Two outputs generated by the composite
method are Maximum Envelops of Water (MEOW) and Maximum of MEOWs (MOMs).
MEOW provides a worse case basin snapshot for a particular storm category and incorporates
the uncertainty in landfall locations. MOM provides a conservative snapshot of where flooding
may exist during a particular storm. The uncertainty that is accounted for in the composite
approach makes it the best method for predicting vulnerability from storm surge and is the basis
for the development of evacuation zones.
A heavy influencer of SLOSH and its model variations is historical hurricane behavior,
which may be inadequate due to climate change affecting storm patterns. The past 10-20 years
has seen an increase in hurricane frequency, strength, and unique steering with storm surge
becoming ever more devastating.
SLOSH is the preferred model used by the National Hurricane Center due to the low cost
and quick time generation for the model. Accuracy issues have been noted, however, SLOSH
gives government agencies proper lead-time to decide and act on evacuation orders, where time
is key over accuracy.
SLOSH has shown inaccuracies in forecasting when a storm is unusually large and when
a storm is unusually slow (Ratcliff 2020). Slow storms can transport more water through the
open ocean and generate forerunner storm surge due to the Earth’s rotation, which can generate
surge many hours prior to landfall (Snowcroft et al. 2020). Even with the low cost and quick
22

speeds, SLOSH is still accurate to within 20% of a storm’s predictive surge, another reason for
its preference (Snowcroft et al. 2020).
2.2.2. ADCIRC
The Advanced Circulation (ADCIRC) model is a computer program that solves equations
of motion for water on the Earth (Leuttich and Westerink 2013). In other words, ADCIRC can
simulate and predict water movement over the ocean and land. The model is used by numerous
academic, industrial, and governmental entities, most notably the Department of Homeland
Security and the U.S. Army Corps of Engineers. The Federal Emergency Management Agency
(FEMA) has validated ADCIRC and uses the model as a basis for flood risk analysis and
mapping. The basis of ADCIRC is solving forms of shallow water equations and solving a
vertically integrated continuity equation (Figure 8), which measures vertical flow of water for
water surface elevation (Leuttich and Westerink 2004). ADCIRC computes water levels by
solving the Generalized Wave Continuity Equation (GWCE) and uses an algorithm to solve
GWCE.

Figure 8. ADCIRC’s vertically integrated continuity equation (Leuttich and Westerink 2004)
23

ADCIRC also uses a mesh system (Figure 9) for its wave height calculations, but uses
triangular unstructured grids instead of structured grids, which gives ADCIRC the ability to
compute localized resolution within each grid mesh. ADCIRC, once hindered by long
computational speeds, now uses parallel algorithms that make surge simulations rapid and
capable of solution on personal computers.

Figure 9. Example ADCIRC mesh (Leuttich and Westerink 2013)
ADCIRC is more complex and capable than SLOSH mainly due to its highly flexible
mesh system (Snowcroft et al. 2020), however; the model requires more time than feasible for
storm surge evacuation, where time is of greatest importance. ADCIRC is better at simulating
tides propagated from the open ocean and can resolve very detailed bathymetry in coastal
regions. SLOSH and ADCIRC differ in three main areas: mesh shape and resolution,
mathematical methods, and physics in the model.
ADCIRC uses a flexible unstructured grid mesh system for its basins. The unstructured
mesh allows for concentration in areas, like coastal regions, allowing for higher resolution. The
physics in ADCIRC can generate three-dimensional solutions for ocean surface wave impacts,
24

whereas SLOSH computes at a two-dimensional level. Table 1 shows key differences in the two
models.
Table 1. Key differences between SLOSH and ADCIRC (Snowcroft et al. 2020)
SLOSH ADCIRC
Mesh Structured: Shape is curvilinear, and
resolution can change gradually
Unstructured: Shape is
triangular, and there is large
capability to vary resolutions.
Mathematical
Methods
Finite difference: Use the rate of change
of a quantity between two neighboring
grid points to represent the continuous
gradient in the equations. (Discretization
of the continuous equations)
Finite element: Approximate the
true answer to the equations with
a combination of simpler
functions. (Discretization of the
true solutions)
Physics 2-dimensional: No ocean surface wave
impacts
2-dimensional/ 3-dimensional:
With ocean surface wave
impacts
Computational
Cost
Low High

2.2.3. Data-Driven Models
Data-driven storm surge models use a combination of statistical and machine learning
techniques for surge predictions (Tadesse et al. 2020). Data-driven models use predictor data
sets, which are historical data that can be applied for future predictions, whereas numerical
models rely on solving equations, however some numerical models incorporate data-driven
methods. Meteorological and remotely sensed wind speed, sea level pressure, and sea surface
temperature are common data used in data-driven modeling techniques as predictor data. Further
25

datasets used in data-driven models involve historical surge data as derived by tidal gauges or
other sources (Tadesse et al. 2020).
The data-driven model predictors are then fed into an algorithm or equation, which is the
statistical portion of the data-driven model. The equation computes maximum surge height at the
location of the specific tidal gauge being calculated (Tadesse et al. 2020). The equations can also
be used across a grid or mesh system as with major statistical models to generate surge
predictions across a large spatial area.
Machine learning algorithms, like the ones stored in ArcGIS Pro, can be used as a
substitute to equations used to predict water levels. A classification technique like Random
Forest can be used to simulate surge. Random Forest algorithm works by using decision trees
that split based on variables. In the storm surge modeling realm, random forest decision trees
consider parameters like wave heights, sea temperature, and wind speed among others.
2.2.4. Artificial Intelligence Models
An artificial neural network (ANN) is another type of machine-learning algorithm that
has been used for forecasting storm surge. It is based on human brain biology and is represented
by three components: input, hidden, and output layers (You and Seo 2008). The ANN storm
surge model works by using 24 different storm parameters and is trained through cycles of
learning where network weights are continuously updated (Ayyad et al. 2022).
Another machine-learning model that has been used to model storm surge is the cluster
neural network model, which works by combining the functionality of ANN with a cluster
analysis technique. This technique calculates similar distances within unsupervised data that
correlate to one another (You and Seo 2008). Unsupervised data techniques find patterns in data
based on relationships between data points (Kotu and Deshpande 2019). Data is then grouped
26

based on similar function or type. AI storm surge modeling has shown promise and has been as
accurate or more accurate than current numerical-based models. (You and Seo 2008).
2.3. Spatial Analysis and Validation
Storm surge models must be validated to determine the veracity of the model’s
predictions. No storm surge model will ever achieve total accuracy but verifying that model’s
performance is a step that will save future lives. Different techniques exist that assess storm
surge model accuracy, but one main avenue exists, hindcast comparison. Hindcast simulations
input observed storm parameters, like wind speed, direction, and atmospheric pressure, into the
surge model to compare performance.
2.3.1. Hindcast
Most hindcast simulations do not map out simulated versus observed flooding, but
instead employ statistical methods to compare ground truth surge levels to simulated surge levels
in a series of charts and graphs (Dinapoli et al. 2020). Dinapoli et al. employed several different
equations to summarize and validate surge near Buenos Aires. The hindcast model that was
created was a numerical model that estimated surge based on observed windspeed and
atmospheric pressure during surge events (Dinapoli et al. 2020). Those readings were input into
the model itself via open-source programming languages like python. Though using a pre-
operational model, results show accuracy to within 8-13% of observed surge heights during the
temporal scale.
FEMA also employs hindcast simulations to validate its surge models. As with most
hindcast simulations, FEMA uses recordings from tidal gauges and high-water marks observed
during a storm and compares those with the levels as computed in its surge model. FEMA
possesses a system called the interactive model evaluation and diagnostic system (IMEDS) to
27

validate its models and assess performance. Inputs to IMEDS include wind, wave, water levels,
and water current parameters that can generate statistical analysis of a past storm. IMEDS
includes an additional step than most hindcast simulations and shows spatially specific tidal
gauges that were deemed accurate or inaccurate. IMEDS is accurate to within NOAA’s standards
of 80% surge height (MDL 2018) and possess robust error metrics like scatter index and skill
score.
Another hindcast simulation on SLOSH and ADCIRC data performed on Hurricanes
Andrew and Irma is a study by Turan et al. Turan et al. deduced that ADCIRC and SLOSH
underpredicted surge during Hurricane Andrew (Turan et al. 2018). For Hurricane Irma,
ADCIRC was shown as the more accurate model (Turan et al. 2018). The study produced a
series of charts and graphs detailing each model’s accuracy, but as with most hindcast
simulations, no visualization showing SLOSH and ADCIRC differences were produced.
2.3.2. Other Validation Methods
A study completed by Veeramony et al. conducted a validation project of the Deltf3D
surge model on Hurricane Ike in 2008. The Delft3D model is employed by the Naval
Oceanographic Office to model surge for their prediction system (Veeramony, Condon, and
Ormondt 2017). The study focused on wave height, water level, and high-water marks as
observed at various recording stations. Further, the study used best forecast wind speeds based
on the NHC’s best advisory track, as opposed to hindcast windspeeds. Results show the Deltf3D
model simulated surge heights to within 96% of observed heights at 6 of 9 recording stations.
(Veeramony, Condon, and Ormondt 2017).
Another validation study was conducted on Hurricanes Ike and Wilma by Kelly et al.,
however, this study validated the fully adaptive storm tide (FAST) model. The FAST model,
28

newer than SLOSH and ADCIRC, is a numerical model that solves shallow water equations like
the two former models. FAST does not require the forecaster to know where fine resolution is
needed in the model, unlike SLOSH and ADCIRC (Kelly et al. 2016). The study used observed
wind and water figures from both hurricanes and input those figures into FAST to determine
differences in the model versus ground truth. Results show FAST predicted surge heights to
within 10-20% of observed surge (Kelly et al. 2016). Additionally, FAST showed quick
computational speeds that even matched SLOSH, which is around 20 minutes (Kelly et al. 2016).
However, additional computer cores are needed, which would inhibit FAST versus SLOSH.

29

Chapter 3 Methods
This research project focuses on two separate analyses to determine model accuracy. First, an
analysis of flood extent throughout the study area was performed to assess the predicted extent
using the SLOSH and ADCIRC models as compared to the flooding extent observed. Second, a
depth analysis is conducted by comparing SLOSH, ADCIRC, and observed flooding depths
across the study area. The chapter is broken down into three sections overviewing data
acquisition, preparation, and analysis.
The methods and results in this chapter are organized into a time series that correlates to
pre- and post-landfall of Hurricane Ida. The eight NHC advisories in this research project form
the time series, with the first five occurring prior to Hurricane Ida’s landfall, advisories 9-14. The
final three NHC advisories, 15-17, cover post landfall. Advisory 13 is skipped during this project
as it was an intermediate 3-hour advisory. Table 2 depicts each advisory along with its associated
timeline.
Table 2. Timeline series
Hurricane Ida
Reference
Event Time (local)

Pre-Landfall
NHC Advisory 9 August 28, 10:00 am
NHC Advisory 10 August 28, 4:00 pm
NHC Advisory 11 August 28, 10:00 pm
NHC Advisory 12 August 29, 4:00 am
NHC Advisory 14 August 29, 10:00 am
Landfall Landfall August 29, 11:15 am

Post-Landfall
NHC Advisory 15 August 29, 4:00 pm
NHC Advisory 16 August 29, 10:00 pm
NHC Advisory 17 August 30, 4:00 am

30

3.1. Methods Overview
In simplified form, the workflow for extent analysis involves taking each raw raster of
the storm surge prediction data and converting it into a polygon of the same surge data. The next
step is clipping the polygon to the study area, and then dissolving the polygon to rid unnecessary
boundary lines within the data. Next, the polygons are summarized within the study area to
compute square mileage and a percentage of flooding within the study area. Results suggest both
SLOSH and ADCIRC vastly overpredicted flood extent when compared to observed flooding.
SLOSH over predicted flood extent by an average of 5% across the study area as compared to
ADCIRC, however, SLOSH showed steady improvement throughout the temporal scale. Figure
10 shows a condensed workflow as completed with each raster dataset for flood extent analysis.

Figure 10. Simplified flood extent workflow
Depth analysis is performed to determine surge depth across the entirety of the study
area. Additionally, the average surge depth per model and NHC advisory is also calculated. In
simplified form, the depth analysis was performed using two different sets of geoprocessing
tools. An ArcGIS tool developed by researchers at the University of Alabama can estimate flood
depth by utilizing a flood extent polygon and a DEM. The Floodwater Depth Estimation Tool
(FwDET) is a GIS-based tool that can estimate flood depth by using a flood extent polygon and a
DEM (Cohen et al. 2019). This FwDETv 2.1 tool was utilized to generate a flood depth raster for
the study area. Next, the zonal statistics tool was used for each surge model and the observed
31

depth raster to generate average surge depth for each advisory and model. Figure 11 represents
the workflow performed to conduct depth analysis across the study area.

Figure 11. Simplified depth analysis workflow
3.2. Data
This section describes the data employed in this project. Table 3 provides a summary of
the data while the following subsections offer more detailed information.
Table 3. Data table
Data Original
Coordinate
System
Description Resolution Type Source
Louisiana
Parish
Boundary
WGS 1984 Administrative
boundaries of the
study area (10
parishes)
Vector Louisiana
Department of
Transportation
and
Development
(LaDOTD)
Land Cover NAD 1983
Albers
Land extent of the
study area
Vector USGS
Digital
Elevation
NAD 1983
UTM Zone
15N
Elevation data of
the study area
10 meters Raster Louisiana State
University
Center for
32

Model
(DEM)
GeoInformatics
(C4G)
Maximum
Flood Extent
Depiction
(MFED)
WGS 1984 Maximum extent
of flooding data as
acquired via
remote sensing
90 meters Raster AER FloodScan
Standard
Flood Extent
Depiction
(SFED)
WGS 1984 Standard extent of
flood data as
acquired via
remote sensing
90 meters Raster AER FloodScan
MODIS
Flood (14-
day
composite)
WGS 1984 Flooding
observed via
remote sensing for
the study area
250 meters Raster NASA NRT
GFM
SLOSH
storm surge
prediction (8
sets, advisory
9-17)
NAD 1983 SLOSH
simulation data
showing predicted
flooded areas
10 meters Raster NHC
ADCIRC
storm surge
prediction (8
sets, advisory
9-17)
WGS 1984 ADCIRC
simulation data
showing predicted
flooded areas
10 meters Raster Coastal
Emergency Risk
Assessment
3.2.1. Louisiana Parish Boundary Data
The state of Louisiana’s Department of Transportation and Development’s open data
portal was utilized to acquire the administrative boundaries for this research project. Specifically,
the Louisiana parish dataset was downloaded that contained all parishes in the state. Figure 12
represent the parish boundaries downloaded for this project.

33

Figure 12. Louisiana parish boundaries
3.2.2. Land Cover Dataset
This research project utilized land cover data for the purpose of creating a boundary of
land extent within the study area. Parish boundaries include lakes and maritime zones, which can
extend many miles into the Gulf of Mexico. A dataset containing just land extent was required so
that flooding extent analysis of land throughout the study area could be completed. The USGS’s
land cover download portal was utilized to acquire a dataset containing land cover only. The
landcover dataset acquired contained the various land cover classes that exist in Louisiana
(Figure 13).
34

Figure 13. Raw land cover dataset
3.2.3. Digital Elevation Model
A DEM is key for use when determining surge depth, as elevation can directly
corresponds to flood depths and locations. For this research project, a 10-meter DEM was
acquired from the Louisiana State University’s Center for GeoInformatics website. A 10-meter
DEM was chosen due to both surge models also possessing the same resolution. Finer resolution
DEMs were experimented with, at both 5 and 3 meters, however, the computational time was not
feasible for the many datasets used in this research. Figure 14 denotes the DEM used in this
project.
35

Figure 14. 10m DEM of the study area
3.2.4. SLOSH Storm Surge Prediction Data
Historical SLOSH data was acquired from the NHC’s tropical system data archive
website, where data on each named storm is kept for research purposes. By choosing a year and
storm name, users can access a bevy of information like hurricane warning and watch locations,
hurricane track locations, and storm surge warnings, among other information. Storm surge data
is published and available for download for each named storm dating back to 2015. The NHC
data archive only stores one type of surge data, P-Surge, a newer form of SLOSH modeling that
accounts for 5-year error averages in past storms.
The SLOSH datasets are 8-bit raster TIFFs. The P-Surge data represents a flood
inundation prediction for a particular area in the path of a storm. Two types of P-Surge data are
available, inundation and tidal mask. The inundation dataset, used in this project, represents
surge flooding inundation with past error weights assigned, which is SLOSH’s newest
36

simulation. The other dataset also includes a tidal mask layer which assigns value to areas along
coastlines that inherently flood during high tides. This project worked exclusively with the
inundation dataset due to same data availability of the ADCIRC surge model covered in the next
section. The inundation and tidal mask datasets are divided by time and are tied to an NHC
advisory. Figure 15 shows SLOSH data as pulled from NOAA’s repository for NHC advisory 9.

Figure 15. SLOSH advisory 9 dataset
The data shows surge extent and maximum inundation depth based on the storms current
position and likely future landfall. The categories are symbolized in the exact way NOAA
symbolizes their advisory data. The levees are included in the data and are included above for
visualization purposes.
3.2.5. ADCIRC Storm Surge Prediction Data
ADCIRC data was found on the CERA website. CERA contains a free web-based GUI
that can run ADCIRC simulations based on storm parameters or based on historical storms.
37

ADCIRC inundation data can be chosen by storm from 2011 to present, and by advisory.
Inundation data for advisories 9 through 17 were downloaded. Each is a 32-bit raster TIFF at 10-
meter resolution. Figure 16 represents ADCIRC data from NHC advisory 9.

Figure 16. ADCIRC data reflecting NHC advisory 9
3.2.6. Observed Flooding Data Acquisition
Numerous sources provide historical flooding data. The reputable Dartmouth Flood
Observatory (DFO) pieces together several different remote sensing datasets to build its flood
extent packages for storms. Flood layers used by DFO include: 250 m MODIS, NASA’s ARIA
SAR, and the European Union’s Copernicus emergency mapping service. Though DFO was the
first-choice source for flood data, as of November 2022, DFO had yet to publish flood extent
data from their spaceborne sensors specifically for Hurricane Ida. Two other reputable sources
were identified, however, which had Hurricane Ida flood data.
38

The Atmospherics and Environmental Research’s (AER) FloodScan program is a
satellite-based remote sensing platform that provides daily maps of large-scale inland flooding.
Flooding data is acquired by an Advanced Microwave Scanning Radiometer 2 (AWSR2)
onboard Global Change Observation Mission-Water Satellite 1 (GCOM-W1), a Japanese
satellite providing long-term Earth observation imagery. FloodScan works by applying an
algorithm to its raw remotely sensed data to provide users with flood inundation mapping
(Galantowicz, Picton, and Root 2021). FloodScan tracks over 99% of the Earth’s surface every
two days, and flood data is sensed at 90m resolution worldwide.
FloodScan provides five different datasets, and all are raster GeoTiffs: persistent water,
fractional flooding, woody wetlands, MFED (maximum flood extent depiction) flooding, and
SFED (standard flood extent depiction) flooding. Persistent water represents bodies of water that
are permanent fixtures, like rivers, lakes, and oceans. Fractional flooding is a dataset, at 22km
resolution, which represents the unfiltered raw data acquired by the AWSR2. Woody wetlands
represents areas like swamps and marshes that area heavily vegetated and moist, but not flooded.
SFED is designed to generate low false positive readings for large-scale flooding events
(Galantowicz, Picton, and Root 2021). Its algorithm is consistent over large temporal scales and
minimizes false positive readings. Its consistency does produce conservative estimates of flood
extent. MFED is designed for quick flood readings for disaster response, and as such, does not
incorporate the steps in its algorithm to minimize false positives (Galantowicz, Picton, and Root
2021).
The FloodScan user’s guide recommends SFED for singular extreme flooding events, and
MFED for flood occurrence over time periods (Galantowicz, Picton, and Root 2021). Both
categories fit with this project as Ida was a singular extreme event, and its impacts required study
39

over a multi-day temporal scale. For those reasons, both SFED and MFED were used. Data for
August 31 was chosen as that date represents the time at which flooding from Hurricane Ida was
at its peak.
Coastal flooding data was acquired from National Aeronautics and Space
Administration’s (NASA) Near Real-Time (NRT) global flood product, remotely sensed from
both Aqua and Terra satellites. Aqua and Terra are part of NASA’s earth observation research
project that studies precipitation, evaporation, and the cycling of water around the globe. The
Moderate Resolution Imaging Spectroradiometer, better known as MODIS, generates flood data
globally at 250m resolution in 10-degree tiles. Three types of data are available: MODIS flood
water, which shows only flooded areas within the extent chosen; MODIS surface water, which
depicts flood water before reference water is filtered within an extent; and MODIS water
product, which is a combination of the two.
The MODIS water product was chosen to give a combination dataset of flooding within
the study area. For this project, 3-day composite data was downloaded, because a total
representation of Ida’s impact was desired. Though covering a different temporal scale than
FloodScan, MODIS’s 3-day composite captures surge prior to and during landfall, while
FloodScan captures post-landfall flood stages. Storm surge can be a multi-day event, and it was
so with Hurricane Ida. To properly study the surge effects from the storm, a multi-day composite
is needed of water data. Four separate 10-degree tiles were required for complete coverage of the
study area.
3.3. Data Preparation
After acquiring the data, the data needed to be cleaned in preparation for analysis. This
section describes the process required for each data set.
40

3.3.1. Coordinate System and Projection
The datasets in this research project were in varying geographic coordinate systems, but
mainly in World Geodetic System (WGS) 1984 and North American Datum (NAD) of 1983. To
ensure the highest degree of accuracy for this project, all datasets needed to be projected into the
same coordinate system.
The study area is entirely encompassed within the southeastern portion of Louisiana, a
relatively large scale. Every dataset used for this project was projected into the NAD 1983
(2011) State Plane Louisiana South FIPS 1702 (US feet) projected coordinate system. This was
chosen because it employs a projection specifically created for the southern portion of Louisiana.
3.3.2. Land Extent Preparation
The raw raster land cover data was in a Geo TIFF format and contained different land
classifications for the entire state of Louisiana. The dataset needed to be transformed into a
polygon to act as the study area boundary for clipping other data. The water features within the
study area, such as lakes and rivers, needed to be cut as to not skew flooding results in the study
area. The raster data contained categorical values of land cover classification. It was reclassified
to change any cell with a value for water to “no data”.
The reclassified raster was converted into a polygon feature class and its features were
dissolved to eliminate the numerous boundary lines within. Next, the data was clipped to the
study area so the land extent would represent only the study area. Figure 17 represents the land
extent of the study area, which is used for all flood analysis throughout this project.
41

Figure 17. Land extent of study area
3.3.3. SLOSH Preparation
Figure 18 represents the preparation workflow that was performed, specific to the flood
extent analysis. First, the dataset was reclassified to eliminate the levees in the data, as the levees
will give erroneous flood extent results. Next, the data was converted to a polygon and clipped to
the study area. The data was then dissolved to eliminate internal lines in the data for a cleaner
look. All eight SLOSH advisory workflows are the same throughout.
42

Figure 18. SLOSH preparation workflow
Value 7 in all SLOSH datasets represented levees, and thus needed to be excluded from
spatial analysis. Value 7 would show as a flooded area during analysis and thus would produce
an erroneous conclusion of SLOSH flooded areas. To eliminate value 7 from SLOSH, the
reclassify tool was utilized by inputting the SLOSH dataset as the input raster, value as the
reclass field, and changing value 7 to no data in the reclassification box. All eight SLOSH
datasets were reclassified prior to conducting analysis. After reclassifying, only flooded values
remained, and analysis could be conducted to determine flood extent percentage with SLOSH
data. Figure 19 depicts SLOSH inundation data for advisory 9 after reclassification.
43

Figure 19. SLOSH surge simulation data from NHC advisory 9
The raster to polygon geoprocessing tool was utilized to convert the SLOSH data into a
polygon by choosing the dataset and then value for input field (Figure 20). Next, the polygon
was clipped to the study area to eliminate any excess data (Figure 20). The dissolve tool was then
employed to eliminate boundary lines within the dataset.

Figure 20. Raster to polygon output (left), polygon clip output (right)
44

3.3.4. ADCIRC Preparation
ADCIRC data could not be converted to a polygon utilizing the standard geoprocessing
tool, due to the dataset not being an integer raster, and therefore raster functions were utilized.
Under raster functions, the conversion set was picked, which converts the ADCIRC raster to a
convertible dataset. After conversion to a colormap, the data could be converted to a polygon
utilizing the raster to polygon geoprocessing tool. The new ADCIRC polygon was then clipped
to the study area to eliminate excess data with the pairwise clip tool. The dissolve tool was then
employed to eliminate boundary lines within the dataset. Figure 21 represents the ADCIRC
preparation workflow as performed with a dataset. The same steps were performed for all eight
advisories.

Figure 21. ADCIRC workflow
3.3.5. Observed Flooding Preparation
Additional steps were taken for the six observed flooding datasets, with the goal of
combining all six into one usable shapefile for the purposes of showing total flood extent, and
ease of use throughout the project. After converting all six raster sets to polygons and clipping to
the study area, the merge tool was input with the four MODIS tiles to combine those, and then
the two FloodScan datasets to combine those. Next, the two remaining datasets were merged to
45

create one observed flooding polygon. The dissolve tool was then utilized to eliminate polygon
boundaries within the dataset. Figure 22 represents the preparation workflow as performed for
the observed flooding datasets.

Figure 22. Observed flooding workflow
Figure 23 depicts all six flood datasets combined, which represents total flooding observed form
Hurricane Ida.

Figure 23. Extent of flood waters during Hurricane Ida
46

3.4. Data Analysis
After cleaning and preparing data, analysis could be conducted to determine flooded
areas within the study area. This section describes how the SLOSH and ADCIRC estimates were
compared against each other and the observed flooding data.
3.4.1. Flood Extent Analysis
The summarize within geoprocessing tool was utilized to calculate the area SLOSH data
is contained within the land extent portion of the study area. For the summarize within tool, the
study area is chosen for input polygon, the clipped SLOSH dataset is chosen for summary
features, and id and sum chosen for field and statistic, respectively. Finally, square miles are
chosen for the shape unit and the tool can be performed.
For ADCIRC data, the summarize within geoprocessing tool was utilized with input
polygon being the clipped ADCIRC polygon, id and sum chosen for field and statistic, and
square miles chosen for shape unit. The geoprocessing tool can be run successfully.
After conducting the technical work, the summarized flooded area in square miles is
contained in each model layer’s attribute table, as created running the summarize within tool.
First, however, the study area’s square mileage must be determined to determine percentage
calculations of the surge and flood datasets. The summarize within tool was run with the study
area’s land extent as the input and summary data, which calculates the square mileage of the
study area. The land extent area was calculated to be 5,632.91 miles.
After running the summarize within tool, each model’s advisory layer has summarized
flood data in its attribute table. After dividing the summarized flood data with the study area’s
square mileage, a percentage calculation is obtained.
47

3.4.2. SLOSH and ADCIRC Flood Depth Analysis
A further comparison of the two storm surge models was computed by comparing surge
depth throughout the study area. The FwDET tool calculates water depth by subtracting
calculated flood water elevation from topographic elevation at each grid cell within the flooded
area (Cohen et al. 2019). Figure 24 represents the flow chart utilized with FwDET.

Figure 24. FwDET flow chart (Cohen et al. 2017)
The flood extent polygon, as created earlier in this chapter, was input into the tool via
ArcGIS Pro, along with a 10-meter DEM of the study area. The output was a raster dataset with
48

estimated flood depth throughout the study area. This step was repeated with the SLOSH and
ADCIRC flood extent datasets to determine their flood depth as well.
The ADCIRC dataset’s value can be used as is, meaning if the average value for a raster
is 2, then 2 feet is the average depth for the study area. For SLOSH, further work is involved to
generate the actual flood depth of the raster. SLOSH datasets have preset values in the raster as
downloaded from NOAA. A value of 1 does not mean a depth of 1 foot, but instead means a
depth of 0-1 foot. A value of 2 denotes depths of 1-3 feet, a value of 3 denotes depths of 3-6 feet,
a value of 4 denotes depths 6-9 feet, and a value of 5 denotes depths over 9 feet. Thus, for each
SLOSH output raster generated, the surge depth must be calculated against its preset value.
The zonal statistics geoprocessing tool was then utilized to compute average flood depth
for the study area by surge model and advisory. SLOSH and ADCIRC predicted depth were
computed per advisory using the FwDET tool and compared against the observed flooding depth.
The zonal statistics tool was used for each of the 8 NHC advisories with both surge models and
the observed flooding depth.
The next chapter covers results from extent and depth analysis. The extent and depth
analysis provide a comparison of SLOSH and ADCIRC against observed flooding.

49

Chapter 4 Results
This chapter overviews the results from chapter three’s workflow for the eight NHC advisories
and their surge predictions. Comparisons of SLOSH, ADCIRC, and observed flooding extent
and depth are included.
For all eight advisories, SLOSH and ADCIRC overestimated flood extent significantly,
to the tune of about 24-31% throughout the study area over observed flooding. SLOSH and
ADCIRC extent predictions, however, remained around 2-7% of each other. This indicates that
both models predicted similar surge extent, albeit in different locations. For all eight advisories,
ADCIRC was the more accurate model in terms of being closer to the observed flooding extent
percentage, however, its dominance waned in the final four advisories. Though the less accurate
model throughout this extent analysis, SLOSH showed accuracy improvement across all eight
advisories, whereas ADCIRC accuracy showed slight decreases as the storm approached and
passed landfall. This can suggest that ADCIRC performs better overall than SLOSH, but SLOSH
performs well in short range forecasting.
In depth analysis, visual representations give the appearance of both surge models
predicting more depth due to more flood extent across the study area. Once the flood data is
broken down and compared by average flood depth, the two models predicted less average surge
depth than the observed flooding dataset at a rate of about 20%. Both models were similar in
their depth predictions with each other with ADCIRC being slightly more accurate as it was
closer to observed depth in 4 of 8 advisories, and SLOSH more accurate in 3 of 8. The final
advisory both models predicted the same depth. This result indicates that both models
underestimated surge depth across the totality of the study area.
50

The chapter is organized into three sections. The first section covers NHC advisories 9-
14, or pre-landfall, along with its flood extent and depth analysis. The second section covers
NHC advisories 15-17, or post-landfall, along with its flood extent and depth analysis. The final
section overviews the entire temporal scale and produces charts and Figures that suggest trends
in the models.
4.1. Pre-Landfall Flooding Predictions: NHC Advisories 9-14
NHC advisories 9-14 represents August 28 at 10:00am (advisory 9) to August 29 at
10:00am (advisory 14), with advisories published every 6 hours in between. During this time
Hurricane Ida’s strength intensified rapidly. Ida’s maximum sustained winds were recorded at 86
mph at advisory 9, 104 mph for advisories 10 and 11, 138 mph for advisory 12, and 150 mph for
advisory 14.
4.1.1. Pre-Landfall Flood Extent Analysis
Figure 25 represents comparisons in both surge models and the observed flooding dataset
for NHC advisories 9, with the idea of highlighting flood extent for the three datasets. Since both
surge models overpredicted surge, the vast majority of observed flooding was successfully
predicted by SLOSH and ADCIRC. There were still slight areas where observed flooding went
unpredicted with both models. Though covering similar extent percentages across the study area,
SLOSH and ADCIRC differed in flood locations. The other pre-landfall advisory maps are in the
appendix.
In all five advisories, SLOSH showed two areas where it predicted flooding more than
ADCIRC: the Mississippi River southwest of Lake Pontchartrain and Bayou Lafourche,
otherwise known as the Chetimachas River (figure 25).
51

Figure 25. Pre-Landfall Advisory 9 Flood Extent
Table 4 shows the areal and percentage of flood extent within the study area. As
previously discussed SLOSH and ADCIRC overestimated flooding significantly, with ADCIRC
being more accurate over advisories 9-14.
Table 4. Advisories 9-14 areal and percentage result analysis

Sq Mi within Study
Area
Study Area Sq Mi
Total Flooded Percentage
Study Area 5,632.91
Observed Flooding
Extent
1,024.3 “ “ 18.18%
SLOSH 9 3,089.4 “ ” 53.37%
ADCIRC 9 2,920.8 “ “ 45.60%
SLOSH 10 3,098.3 “ “ 53.53%
ADCIRC 10 3,067.8 “ “ 46.97%
SLOSH 11 3,024.5 “ ” 52.23%
ADCIRC 11 2,917.3 “ “ 45.53%
SLOSH 12 3,073.3 “ “ 53.09%
ADCIRC 12 3,035.8 “ “ 47.61%
Bayou
Laforche
Miss.
River
52

SLOSH 14 2,845.6 “ ” 50.52%
ADCIRC 14 2,664.9 “ “ 47.31%

4.1.2. Pre-Landfall 9-14 Depth Analysis
Visually, observed flooding shows deeper closer to the coast, with pockets scattered near
Lakes Pontchartrain and Maurepas. ADCIRC predicted similar results to observed flooding,
albeit at a greater extent throughout the study area. SLOSH, on the other hand, shows light
flooding closer to the coast and deeper flooding inland concentrating at Bayou Lafourche, an
area ADCIRC negated. Visually speaking, ADCIRC’s flood depth was more in line with
observed flooding than SLOSH. Figure 26 represents flood depths for advisory 9. The other
advisory depth analysis maps are located in the appendix.

Figure 26. Depth Analysis for NHC Advisory 9
53

Averaging surge depth across the entire study area for both models and observed flooding
gives another avenue to compare results. Both models predicted less average surge depth than
observed flooding. SLOSH was more accurate in 3 of the 5 pre-landfall advisories, with
ADCIRC more accurate in 2 of 5. Figure 27 shows the average depth across the pre-landfall
advisories.

Figure 27. Pre-Landfall depth analysis in feet
4.2. Post-Landfall Flooding Predictions: NHC Advisories 15-17
NHC advisories 15-17 represents the time following landfall, but specifically covers
August 29 at 04:00pm to August 30 at 04:00am, with advisories published every 6 hours in
between. During this time Ida would slowly weaken with maximum sustained wind speeds of
132 mph, 104 mph, and 58 mph for advisories 15-17, respectively.
4.2.1. NHC Advisories 15-17 Flood Extent Analysis
Figure 28 represents the flood extent analysis for NHC advisory 15. As with the previous
advisories, observed flooding is predicted well with the two models. A few areas exist where
observed flooding was not predicted, but they are isolated west and north of Lake Pontchartrain,
54

along with small pockets towards the coast. Two main areas still exist that show SLOSH extent
different than that of ADCIRC. Those areas are the same as the first five advisories, the
Mississippi River southwest of Lake Pontchartrain, and Bayou Lafourche. The other post-
landfall advisory extent maps are located in the appendix.

Figure 28. Post-Landfall Advisory 15 Flood Extent
The areal and percentage table from the study area analysis suggests differences in the
results from the previous five advisories. NHC advisories 9-14 show ADCIRC as the more
accurate model over SLOSH to around 5% in most cases. With NHC advisories 15-17, ADCIRC
was still the more accurate model, but down to around 2.5%. Table 5 shows the areal and
percentage analysis with SLOSH and ADCIRC. This analysis can suggest that ADCIRC is the
more accurate model overall, especially in long range forecasting, whereas SLOSH can be
accurate in post-landfall applications.
55

Table 5. Advisories 15-17 areal and percentage result analysis

Sq Mi within Study
Area
Study Area Sq Mi
Total Flooded Percentage
Study Area 5,632.9
Observed Flooding
Extent
1024.3 “ “ 18.18%
SLOSH 15 2834.2 “ “ 50.31%
ADCIRC 15 2742.5 “ “ 48.69%
SLOSH 16 2776.7 “ ” 49.29%
ADCIRC 16 2655.2 “ “ 47.14%
SLOSH 17 2790.9 “ “ 49.55%
ADCIRC 17 2623.1 “ “ 46.57%

4.2.2. Post-Landfall Depth Analysis
Surge depth post landfall shows similar results to pre landfall in that ADCIRC appears to
be the more accurate of the two in predicting depth across the study area. SLOSH does predict
far less depth in advisories 15 and 16 as the storm moves further inland. ADCIRC predicts far
deeper surge in post landfall towards the coast but appears deeper than observed flooding. One
notable anomaly is with the SLOSH model at advisory 17, where areas around the Mississippi
River and Bayou Lafourche show significant surge depth. This could be attributed to the rivers
surpassing their flood stages post landfall. Figure 29 represents flood depths for advisory 15. The
other post-landfall depth analysis maps are located in the appendix.
56

Figure 29. Depth Analysis for NHC Advisory 15
For the three advisories post-landfall, ADCIRC was the more accurate in predicting
average surge depth. Both models still predicted less average depth than actual observed depth.
Figure 30 shows the post landfall depth analysis for the surge models and observed flooding.
57

Figure 30. Post-Landfall depth analysis in feet
4.3. Overall Extent Analysis
Figures 31 and 32 show bar charts of the flooding extent analysis throughout the study
area with SLOSH and ADCRIC during all eight NHC advisories. The results show that SLOSH
consistently improved its extent forecasting against the 18.18% observed flooding extent
percentage. SLOSH’s own extent percentage improved as Ida crept closer to landfall and even
improved after landfall.
58

Figure 31. SLOSH flood extent percentage analysis
ADCIRC shows a pattern nearly opposite of SLOSH, as accuracy decreased slightly
throughout the temporal scale. ADCIRC was more accurate through all eight advisories but is
not as accurate in the final four. ADCIRC still produced better results prior to and after landfall,
even with Ida’s unforeseen intensification.

Figure 32. ADCIRC flood extent percentage analysis
45.00%
47.00%
49.00%
51.00%
53.00%
55.00%
2400
2500
2600
2700
2800
2900
3000
3100
SLOSH 9 SLOSH 10SLOSH 11SLOSH 12SLOSH 14SLOSH 15SLOSH 16SLOSH 17
SLOSH Extent: NHC Advisories 9-17
Sq Mi within Study Area Flooded Percentage
45.00%
47.00%
49.00%
51.00%
53.00%
55.00%
2400
2500
2600
2700
2800
2900
3000
3100
ADCIRC 9 ADCIRC
10
ADCIRC
11
ADCIRC
12
ADCIRC
14
ADCIRC
15
ADCIRC
16
ADCIRC
17
ADCIRC Extent: NHC Advisories 9-17
Sq Mi within Study Area Flooded Percentage
59

Overall, the percentage analysis over the entire temporal scale suggests ADCIRC is the
better model for forecasting. However, SLOSH shows performance nearly on par with ADCIRC
in the time just prior to and during landfall. Coupled with the monetary and time advantages
SLOSH possesses over ADCIRC, SLOSH can predict surge adequately enough for planning and
evacuation. Both models appear to maintain high enough accuracy, at least with surge extent, to
work against modern tropical systems, as Ida’s rapid intensification did not affect SLOSH and
ADCIRC as was hypothesized.
4.4. Overall Depth Analysis
Overall, both surge models predicted less average depth across the study area when
compared with observed flooding. When factoring which model was more accurate, ADCIRC
comes slightly out on top. ADCIRC was more accurate in 4 of the 8 advisories, whereas SLOSH
was more accurate in 3 of 8. The final advisory both models predicted the same average surge
depth. Figure 33 denotes the depth analysis chart across the entire temporal scale.

Figure 33. Overall depth analysis in feet
60

While adequately predicting flood extent, caution should be exercised when claiming it
adequately predicted flood depth. Both models underpredicted average surge depth, at a rate of
about 20% or about 0.15 feet. While not appearing to be a big difference, even finite
underestimations of depth can be disastrous. While not affecting flood extent, the surge models
did underestimate surge depth as hypothesized.
4.5. Locality Analysis
Using satellite imagery and zooming into areas within the study area can help shed light
on the differences in the models. The locality analysis was used to determine spatial patterns
within the surge and flood datasets. Identifying spatial patterns within the flooded areas can help
determine model differences based on the spatial environment. Additionally, deciphering if
certain topographic areas hinder or help surge is useful. Figure 34 denotes an overview of a
locality analysis with each location’s figure letter that is referenced later.

Figure 34. Locality analysis overview
61

The following figures represent a close-up of five separate locations, along with a
discussion for each. Each location was chosen due to its unique environmental characteristics
and the presence of flooding in each dataset. Figure 35 denotes an area centered on Thibodaux,
Louisiana, and a section of Bayou Lafourche. Of note, the red lines on the maps represent the
study area boundary.

Figure 35. Thibodaux, LA and Bayou Lafourche flood analysis
As represented in the figure, SLOSH predicted heavy flooding in this section of Bayou
Lafourche’s basin. By comparing the models with the DEM used in this study, a correlation can
be made between flooding and elevation. Essentially, SLOSH predicted a deeper surge in this
62

area, which allowed its surge extent to be greater and cover more land in this section of the study
area. Also note that both models do a poor job in predicting overall surge in this location. That
could be due to the inland nature of this location, as it is nearly 50 miles from the coast. Figure
36 represents an area centered on Alluvial City, Louisiana.

Figure 36. Alluvial City flood analysis
All three satellite imagery datasets contain flooding, with SLOSH showing less extent.
All three datasets do follow the elevation as depicted in the center of each map which shows an
outflow canal with a flood levee at higher elevation. On visual inspection however, ADCIRC
appears to be the better model at predicting flood extent and depth, especially in the area south of
63

Alluvial City. Even though this area is closer to the coast, SLOSH does produce results similar to
ADCIRC and observed flooding. Figure 37 represents Jean Lafitte, a small town just south of
New Orleans.

Figure 37. Jean Lafitte flood analysis
The two surge models accurately predicted zero flooding within the city of New Orleans,
and that can be attributed to the levees surrounding the city. Both models also accurately
predicted minimal flooding on the banks of the Mississippi as the higher elevation denotes in the
DEM. The areas immediately north and west of Jean Lafitte show both models predicting heavy
flooding, which is absent in observed flooding. Lesser flood depth due to higher elevation can be
64

a simple reason, with the two models predicting inland lake surge as a more complicated one.
Overall, ADCIRC’s prediction is more in line with observed flooding, which saw SLOSH
underpredict flooding in the southern area of the map. Louisiana State Highway 90 is the focus
of figure 38, a highway that cuts through the southern part of the state.

Figure 38. Highway 90 flood analysis
This location shows a significant difference in what the two models predicted against
observed flooding. Scarcely any areas show observed flooding, where both surge models
predicted heavy flood amounts with SLOSH predicting the heaviest and most widespread. It is
clear that both models expected surge from the two large bodies of water to inundate the region.
65

ADCIRC’s extent is contained locally to the two large lakes in the area, whereas SLOSH shows
a complete flood of the entire map. However, the small area that did flood was not foreseen by
SLOSH at all. ADCIRC, though heavily overpredicting flood in this location, did successfully
predict the flood that was observed, another indication that ADCIRC was better in this specific
location. Figure 39 represents the final flood analysis location, Fontainebleau State Park on the
north side of Lake Pontchartrain.

Figure 39. Fontainebleau State Park flood analysis
The north side of Lake Pontchartrain saw scarce flooding, which was adequately
predicted, albeit in larger amounts. Observed flooding only shows areas in the very low-lying
66

elevations near the banks of the lake as flooded, where the two models extended flooding further
inland. SLOSH shows more flood depth and extent and goes deep into the various river and
stream systems that flow into Lake Pontchartrain, which can be easily seen on the DEM. A
visual inspection confirms ADCIRC was more accurate with flood predictions in this location.

67

Chapter 5 Discussion
This chapter discusses the outputs, limitations, and future research of storm surge modeling.
Highlights include a discussion of the results that were discovered and the intricacies of studying
storm surge. Then, discussions on limitations inherent to modeling environmental phenomenon.
Next, further avenues of research are posed that can provide additional ways to analyze surge
flooding. Finally, recommendations for future modeling techniques are posed that may be able to
provide more accurate surge data.
5.1. Storm Surge Analysis
In terms of this thesis’ hypothesis, flood extent was successfully predicted, albeit at a
much larger scale. For flood depth, the two models underestimated depth, as hypothesized. The
results in both extent and depth analysis show ADCIRC as the more accurate model overall, with
SLOSH being more accurate at certain times along the temporal scale. Ironically, both models
performed the best at advisory 14, just prior to landfall.
NHC uses SLOSH due to its quick simulation speeds and easy computational strain.
SLOSH, though not as accurate in this particular study, produces accurate enough results for the
evacuation of people in surge predicted areas. The NHC’s reasoning for using SLOSH is also the
reason ADCIRC is not employed by the NHC. High computational needs, which in turn costs
more money, along with slower surge simulation results.
Overall, in this study, ADCIRC was the more accurate model predicting flood extent and
depth. SLOSH appeared to show a significant lack of coastal flooding but did predict heavy
inland flooding. ADCIRC was more in line with observed flooding, heavier coastal flooding with
limited inland flooding. The biggest difference with SLOSH was the prediction of heavier surge
from inland lakes and Bayou Lafourche spilling its banks and inundating its surroundings.
68

ADCIRC on the other hand predicted flood better than SLOSH but did anticipate more coastal
flooding than what was observed and predicted some surge from inland lakes.
A reason for SLOSH’s inability to adequately predict coastal flooding in this project
could be attributed to its use of structured mesh grids. The structured mesh grids are a bad match
for Louisiana’s complex geography with its many barrier islands, marshes, and tidal swamps.
SLOSH’s mesh grids cannot vary resolution in those shallow coastal regions that inundate
Louisiana’s coastline, and as such, struggled to replicate surge in coastal areas. ADCIRC, with
its unstructured mesh grids, can produce fine resolution which can capture Louisiana’s coastal
regions and bathymetry that can assist with surge predictions in those areas.
The goal in surge modeling is to provide necessary lead-time to issue evacuation orders.
With evacuation orders being issued many days in advance of storm surge, long range prediction
accuracy is required for proper evacuations, which hints at ADCIRC being the best model in that
sense. Once other factors are included, such as time and money, SLOSH may be the more
preferred model, depending on time and money tolerance.
5.2. Research Limitations
Limitations influence research projects, but also open opportunities for more thorough
research in future endeavors. The large scale of this research project produced limitations in
certain areas, and they are included below.
One of the aims of this research project is to compare observed flooding against storm
surge model flooding predictions. No observed flooding dataset will be totally accurate due to
the idiosyncrasies of how the data is acquired, like remote sensing in this case. Hurricane Ida’s
precipitation influenced and contributed to flooding throughout the study area, in addition to
69

storm surge. This disclaimer acknowledges the fact that precipitation from Ida probably skewed
the observed flooding data throughout this project.
Resolution differences in the data used in this project can also skew results. The surge
models and the DEM are in 10 meters, whereas most of the flooding data is 90 meters. Using a
flood dataset that is closer to 10 meters can improve results throughout the project. The observed
flooding, with its coarse resolution, can skew flood areas throughout the study area.
5.2.1. Flooding Data
The ability to create an accurate dataset depicting flooding from an event is nearly
impossible even with today’s technology. Satellites and other remote sensing technologies have
the ability to capture flood areas, but they do come with limitations. For instance, the observed
flooding in this dataset was pieced together from several sources, sources that can contain false
positives with flood acquisition, along with the inability to predict flood in heavily wooded areas
that permeate the study area.
The observed flooding dataset compiled in this project lacked flood depth in the data. A
somewhat experimental ArcGIS tool, FwDET, was used to create the depth in the flood data. The
tool has produced accurate results in past research projects, but the accuracy of the flood depth
derived and used in this project has not been studied.
5.2.2. Storm Surge Data
This research project focused on maximum inundation for both SLOSH and ADCIRC,
mainly due to data availability. For SLOSH, maximum inundation surge means P-Surge, one of
the three methods SLOSH predicts surge. However, the composite method is known as the best
surge predictions method, meaning perhaps SLOSH would have produced more accurate results
as such.
70

Both SLOSH and ADCIRC data for this research project represented maximum surge and
water inundation based on the NHC forecast advisory time the data represents. For example,
NHC advisory 9 data from both SLOSH and ADCIRC represents the maximum inundation the
study area may receive based on the NHC track of the storm at advisory 9’s release. The data
does not represent flooding at the NHC advisory release date and time. The data does not
necessarily represent surge at landfall either, but is a representation of maximum inundation,
which could occur at any time, although periods near landfall are usually the occasion.
5.2.3. Environment
Predicting storm surge is an art, in a way in which it can never be perfect. No computer
can possibly mimic the environment or simulate completely accurate storm surge. Surge models
can incorporate a bevy of environmental parameters intended to simulate nature’s storm surge
and its effects. The very nature of being unable to simulate nature means that storm surge
predictions will never achieve complete accuracy. Additionally, surge models create their
predictions based on weather forecasting, which has errors in its own right. Ultimately, this
research project provides a baseline showing which surge model performed best under the
chosen circumstances. Tweaking one parameter even slightly, will give totally different outputs
and interpretations to a different researcher.
The study area itself can be considered a limitation in this project. Louisiana’s coastal
geography lends itself to ADCIRC being the more accurate model due simply to the use of
unstructured mesh grids. A separate study in an area with less complex coastal geography may
make for a more thorough analysis, as one model won’t necessarily start with a disadvantage.
71

5.2.4. Temporal Scale
The temporal scale in this project started about 24 hours prior to landfall. Considering
evacuations are published multiple days in advance of a hurricanes landfall, a study of increased
temporal scale may produce better results.
At the start of the temporal scale, evacuations are already in place, and for the most part,
people have moved out of harm’s way. A study comprising a temporal scale starting at 96 hours
before landfall will give a better view of which surge model predicted better results compared to
actual surge flooding. A 96 hour prior to landfall study will give better indications on whether
SLOSH or ADCIRC’s predictions were adequate with what surge the storm produced.
Additionally, hurricanes can have drastic changes in movements in the final 96 hours, further
complicating surge model study.
5.3. Future Research
A main area research can be expounded upon is in the flooding data. A few different
avenues exist to acquire and/ or create more data. DFO posts excellent flooding datasets on its
website depository but has not completed the Hurricane Ida data as of this thesis. It is unsure
whether the flood data will ever be compiled by DFO for Ida but having a complete set of flood
data without having to piece it together would be beneficial.
Using tidal and flood gauges throughout the study area would be an additional help to
confirm the legitimacy of flood depth during this research. The city of New Orleans and other
locations throughout the study area have numerous gauges that are designed to detect flooding
levels, and that would have helped this project gauge how accurate the FwDET tool is.
Incorporating more storm surge models can give a wider scope of discovery and learning.
SLOSH and ADCIRC are but two of numerous other surge models, and others such as SWAN,
72

FAST, and Delft3D could give a greater sense of modeling accuracy. This author is unaware of
research comparing multiple surge models past three, so expanding a surge comparison to five or
more would be exciting and eye opening.
Finally, these results have real-world policy implications. Storm surge modeling directly
drives evacuation orders that can potentially impact millions of citizens. While an evacuation
order is simple on the surface, the decision to issue that order cannot be taken lightly. Plainly, an
inaccurate surge prediction can lead to the wrong areas evacuated, and areas that were not under
evacuation orders can be devastated. The logistical burden of evacuating millions of people can
strain the socioeconomic balance of the locale under evacuation orders and can make such
evacuation extremely difficult. Storm surge prediction must be accurate, but the evacuation
orders that accompany it must be shrewd.
5.4. Conclusion
Storm surge will likely never be predicted perfectly, even with newer technologies like
artificial intelligence taking hold. However, providing the public with surge models that can give
an edge in accuracy is the goal. Though a conclusion was made of what surge model is more
accurate for the particular time and place of this research project, that does not necessarily mean
that the surge model is more accurate in every application, as the weather and environment
change at a moment’s notice. Storm surge modeling is not an exact science, but research of this
type can save lives, and that is worth the research time.

73

References
Ayyad, M., M.R. Hajj, and R. Marsooli. 2022. “Artificial Intelligence for Hurricane Storm Surge
Hazard Assessment.” Ocean Engineering 245 (January): 110435.
doi:10.1016/j.oceaneng.2021.110435.
Andersen, T., and M. Shepherd. 2017. “Inland Tropical Cyclones and the ‘Brown Ocean’
Concept.” Hurricanes and Climate Change 3 (February): 117–34. doi:10.1007/978-3-319-
47594-3_5.
Benevolenza, M.A., and L. DeRigne. 2018. “The Impact of Climate Change and Natural
Disasters on Vulnerable Populations: A Systematic Review of Literature.” Journal of
Human Behavior in the Social Environment 29 (2): 266–81.
doi:10.1080/10911359.2018.1527739.
Benincasa, S. 2020. “A Spatial Investigation of New York City’s Historical Shoreline.” Thesis,
Los Angeles, CA: University of Southern California. Thesis / Dissertation ETD.
Beven, J., A. Hagan, and R. Berg. 2022. Rep. National Hurricane Center Tropical Cyclone
Report: Hurricane Ida. National Weather Service.
https://www.nhc.noaa.gov/data/tcr/AL092021_Ida.pdf
Bickford, W.B. 1994. First Course in the Finite Element Method. Burr Ridge: Irwin.
C2ES. 2020. “Hurricanes and Climate Change.” Center for Climate and Energy Solutions.
December 15. https://www.c2es.org/content/hurricanes-and-climate-change.
Cohen, S., G.R. Brakenridge, A.K., B. Bates, J. Nelson, R. McDonald, Y.F. Huang, D.
Munasinghe, and J. Zhang. 2017. “Estimating Floodwater Depths from Flood Inundation
Maps and Topography.” JAWRA Journal of the American Water Resources Association
54 (4): 847–58. doi:10.1111/1752-1688.12609.
Cohen, S., A. Raney, D. Munasinghe, J.D. Loftis, A. Molthan, J. Bell, L. Rogers, et al. 2019.
“The Floodwater Depth Estimation Tool (Fwdet v2.0) for Improved Remote Sensing
Analysis of Coastal Flooding.” Natural Hazards and Earth System Sciences 19 (9): 2053–
65. doi:10.5194/nhess-19-2053-2019.
Corbosiero, K.L., and J. Molinari. 2002. “The Effects of Vertical Wind Shear on the Distribution
of Convection in Tropical Cyclones.” Monthly Weather Review 130 (8): 2110–23.
doi:10.1175/1520-0493(2002)130<2110:teovws>2.0.co;2.
Dahlman, L., and R. Lindsey. 2022. “Climate Change: Global Temperature.” Edited by Jessica
Blunden. NOAA Climate.gov. NOAA. June 28, 2022. https://www.climate.gov/news-
features/understanding-climate/climate-change-global-temperature.

74

Didlake, A.C. 2020. “What Is a Hurricane Storm Surge?: Earth.” What Is a Hurricane Storm
Surge? Eatrthsky. September 20. https://earthsky.org/earth/definition-what-is-a-
hurricane-storm-surge/.
Dietrich, J.C., J.J. Westerink, A.B. Kennedy, J.M. Smith, R.E. Jensen, M. Zijlema, L.H.
Holthuijsen, et al. 2011. “Hurricane Gustav (2008) Waves and Storm Surge: Hindcast,
Synoptic Analysis, and Validation in Southern Louisiana.” Monthly Weather Review 139
(8): 2488–2522. doi:10.1175/2011mwr3611.1.
Dietrich, J.C., S. Tanaka, J.J. Westerink, C.N. Dawson, R.A. Luettich, M. Zijlema, L.H.
Holthuijsen, J.M. Smith, L.G. Westerink, and H.J. Westerink. 2011. “Performance of the
Unstructured-Mesh, SWAN+ADCIRC Model in Computing Hurricane Waves and Surge
- Journal of Scientific Computing.” SpringerLink. Springer US. November 10.
https://link.springer.com/article/10.1007/s10915-011-9555-6.
Dinápoli, M.G., C.G. Simionato, and D. Moreira. 2020. “Development and Validation of a Storm
Surge Forecasting/Hindcasting Modelling System for the Extensive Río De La Plata
Estuary and Its Adjacent Continental Shelf.” Natural Hazards 103 (2): 2231–59.
doi:10.1007/s11069-020-04079-5.
Dube, S.K., T.S. Murty, J.C. Feyen, R. Cabrera, B.A. Harper, J.D. Bales, and S. Amer. 2010.
“Storm Surge Modeling and Applications in Coastal Areas.” World Scientific Series on
Asia-Pacific Weather and Climate, March, 363–406. doi:10.1142/9789814293488_0012.
Emanuel, K. 2005. “Increasing Destructiveness of Tropical Cyclones over the Past 30 Years.”
Nature 436 (7051): 686–88. doi:10.1038/nature03906.
Galantowicz, J., J. Picton, and B. Root. 2001. FloodScan Data Users Guide. Atmospheric and
Environmental Research, Inc. (AER). March 26.
https://www.aer.com/contentassets/22663ebfdb7c467599f363466b32770f/floodscan_data
_users_guide_v05r01_r01.pdf.
Glahn, B., W.A Shaffer, N. Kurkowski, and A. Taylor. 2009. “The Role of the Slosh Model in
National Weather Service Storm Surge ...” Weather.gov. Meteorological Development
Laboratory, National Weather Service, NOAA. August.
https://www.nws.noaa.gov/mdl/pubs/Documents/Papers/Vol-33-Nu1-Glahn.pdf.
Grossmann, C., H.G. Roos, and M. Stynes. 2008. Numerical Treatment of Partial Differential
Equations. Berlin: Springer.
Heil, J. 2019. “10 U.S. States Where Hurricanes Hit Most Often.” Universal Property &
Casualty Insurance Company. Universal Property & Casualty Insurance Company.
https://universalproperty.com/united-states-where-hurricanes-hit-most/.
Houston, S.H., W.A. Shaffer, M.D. Powell, and J. Chen. 1999. “Comparisons of HRD and Slosh
Surface Wind Fields in Hurricanes: Implications for Storm Surge Modeling.” Weather
and Forecasting 14 (5): 671–86. doi:10.1175/1520-
0434(1999)014<0671:cohass>2.0.co;2.
75

Jelesnianski, C.P., J. Chen, and W.A. Shaffer. 1992. Slosh: Sea, Lake, and Overland Surges from
Hurricanes. Rockville, MD: U.S. Dept. of Commerce, National Oceanic and
Atmospheric Administration, National Weather Service.
Kelly, D.M., Y.C. Teng, Y. Li, and K. Zhang. 2016. “Validation of the Fast Forecast Model for
the Storm Surges Due to Hurricanes Wilma and Ike.” Natural Hazards 83 (1): 53–74.
doi:10.1007/s11069-016-2301-5.
Kossin, J. P. 2008. “Is the North Atlantic Hurricane Season Getting Longer?” Geophysical
Research Letters 35 (23). https://doi.org/10.1029/2008gl036012.

Kotu, V., and B. Deshpande. 2019. Data Science: Concepts and Practice. Cambridge: Morgan
Kaufmann Publishers, an imprint of Elsevier.
Leuttich, R., and J. Westerink. 2004. “Formulation and Numerical Implementation of the 2D/3D
ADCIRC Finite ...” Formulation and Numerical Implementation of the 2D/3D ADCIRC
Finite Element Model Version 44.XX . University of North Carolina at Chapel Hill .
December 8. https://adcirc.org/wp-
content/uploads/sites/2255/2018/11/2004_Luettich.pdf.
Leuttich, R., and J. Westerink. 2012. “User's Manual - V50 - ADCIRC.” ADCIRC Users Manual
V50 Comments. The University of North Carolina at Chapel Hill . April 10.
https://adcirc.org/home/documentation/users-manual-v50/.
Li, L., and P. Chakraborty. 2020. “Slower Decay of Landfalling Hurricanes in a Warming
World.” Nature 587 (7833): 230–34. doi:10.1038/s41586-020-2867-7.
Lin, N., K. Emanuel, M. Oppenheimer, and E. Vanmarcke. 2012. “Physically Based Assessment
of Hurricane Surge Threat under Climate Change.” Nature Climate Change 2 (6): 462–
67. doi:10.1038/nclimate1389.
MDL. 2018. “SLOSH Model.” SLOSH Model - MDL - Virtual Lab. Meteorological
Development Laboratory. https://vlab.noaa.gov/web/mdl/slosh.
Mohammed, S. F. 2019. “Storm Surge Hazard Database for the United States Using Stochastic
Hurricane Model: Development, Bias Correction and Applications.” TigerPrints.
Clemson University. May. https://tigerprints.clemson.edu/all_theses/3077/.
NHC. 2022. Hurricane Preparedness - Hazards. National Hurricane Center/ NOAA. May 10.
https://www.nhc.noaa.gov/prepare/hazards.php.
—, 2022. “Sea, Lake, and Overland Surges from Hurricanes (SLOSH).” Sea, Lake, and
Overland Surges from Hurricanes (Slosh). NOAA. Accessed June 14.
https://www.nhc.noaa.gov/surge/slosh.php.
NOAA, US Department of Commerce. 2018. “Hurricane Florence: September 14, 2018.”
National Weather Service. NOAA's National Weather Service. November 27.
https://www.weather.gov/ilm/HurricaneFlorence.
76

O’Gorman, P. A. 2015. “Precipitation Extremes under Climate Change.” Current Climate
Change Reports 1 (2): 49–59. https://doi.org/10.1007/s40641-015-0009-3.

Pant, S., and E.J. Cha. 2019. “Potential Changes in Hurricane Risk Profile across the United
States Coastal Regions under Climate Change Scenarios.” Structural Safety 80 (May):
56–65. doi:10.1016/j.strusafe.2019.05.003.
Pörtner, H.O, D.C. Roberts, M. Tignor, E.S. Poloczanska, K. Mintenbeck, A. Alegría, M. Craig,
et al. 2022. Climate Change 2022: Impacts, Adaptation and Vulnerability. Geneva,
Switzerland: WMO, IPCC Secretariat.
Ratcliff, J. 2020. “Modeling the Storm Surge and Compound Flooding from Hurricane Florence
Using ADCIRC.” Carolina Digital Repository. March 30.
https://cdr.lib.unc.edu/concern/scholarly_works/2b88qh85q.
Schleifstein, M. 2020. “15 Years after Katrina, New Orleans Levees Are in the Best Shape Ever.
Experts Say It's Not Enough.” NOLA.com. NOLA.com. August 24.
https://www.nola.com/news/environment/article_80c27be8-e3e7-11ea-bbf9-
1731ebdd9171.html.
Snowcroft, G., I. Ginis, C. Knowlton, R. Yablonski, H. Morin, and D. McIntire. 2020.
“Numerical Models of Storm Surge, Wave, and Coastal Flooding.” Hurricanes: Science
and Society . University of Rhode Island Graduate School of Oceanography.
http://www.hurricanescience.org/science/forecast/models/modeltypes/stormsurge/.
Stott, P.A., D.A. Stone, and M.R. Allen. 2004. “Human Contribution to the European Heatwave
of 2003.” Nature 432 (7017): 610–14. doi:10.1038/nature03089.
Tadesse, M., T. Wahl, and A. Cid. 2020. “Data-Driven Modeling of Global Storm Surges.”
Frontiers in Marine Science 7. doi:10.3389/fmars.2020.00260.
Trenberth, K.E., L. Cheng, P. Jacobs, Y. Zhang, and J. Fasullo. 2018. “Hurricane Harvey Links
to Ocean Heat Content and Climate Change Adaptation.” Earth's Future 6 (5): 730–44.
doi:10.1029/2018ef000825.
Truchelut, R.E., P.J. Klotzbach, E.M. Staehling, K.M. Wood, D.J. Halperin, C.J. Schreck, and
E.S. Blake. 2022. “Earlier Onset of North Atlantic Hurricane Season with Warming
Oceans.” Nature Communications 13 (1). doi:10.1038/s41467-022-31821-3.
Turan, C.K., Y.P. Kinfu, M.A. Samad, A. Farhadzadeh, and K. Ng. 2018. “Comparison of
ADCIRC and Slosh Model Simulations for Hurricanes Andrew and Irma near Miami,
Florida.” World Environmental and Water Resources Congress 2018.
doi:10.1061/9780784481424.019.
Veeramony, J., A. Condon, and M. van Ormondt. 2017. “Forecasting Storm Surge and
Inundation: Model Validation.” Weather and Forecasting 32 (6): 2045–63.
doi:10.1175/waf-d-17-0015.1.
77

Venkatesh, S. 1974. The Development of Manual Techniques for the Real Time Prediction of
Storm Surges on the Great Lakes. Unpublished.
Westerink, J.J., R.A. Luettich, and J.C. Muccino. 1994. “Modelling Tides in the Western North
Atlantic Using Unstructured Graded Grids.” Tellus A 46 (2): 178–99. doi:10.1034/j.1600-
0870.1994.00007.x.
You, S.H., and J.W. Seo. 2009. “Storm Surge Prediction Using an Artificial Neural Network
Model and Cluster Analysis.” Natural Hazards 51 (1): 97–114. doi:10.1007/s11069-009-
9396-x.
Zhu, Y.J., J.M. Collins, P.J. Klotzbach, and C.J. Schreck. 2022. “Hurricane Ida (2021): Rapid
Intensification Followed by Slow Inland Decay.” Bulletin of the American
Meteorological Society. doi:10.1175/bams-d-21-0240.1.

78

Appendix A Surge Extent by Advisory

Figure 40. Advisory 10 Extent Analysis
79

Figure 41. Advisory 11 Extent Analysis

Figure 42. Advisory 12 Extent Analysis
80

Figure 43. Advisory 14 Extent Analysis

Figure 44. Advisory 16 Extent Analysis
81

Figure 45. Advisory 17 Extent Analysis

82

Appendix B Surge Depth by Parish
Table 6. Depth difference chart in NHC advisories 9-14
Parish Advisory SLOSH Depth
per ft
ADCIRC Depth
per ft
Delta (ft) Observed
Depth (ft)
Jefferson 9 8.19 3.2 4.99 6.18
10 8.52 3.93 4.59
11 6.75 3.75 3
12 8.88 4.61 4.27
14 6.21 5.28 0.93
Lafourche 9 7.38 2.92 4.46 1.55
10 6.51 3.47 3.04
11 3.48 2.7 0.78
12 7.38 4 3.38
14 3.96 4.11 -0.15
Orleans 9 7.62 2.91 4.71 5.83
10 8.55 4 4.55
11 8.61 4.87 3.74
12 8.61 4.77 3.84
14 3.15 4.9 -1.75
Plaquemines 9 6.72 3.46 3.26 2.28
10 7.74 4.16 3.58
11 7.74 5.27 2.47
12 7.02 5.13 1.89
14 5.1 5.57 -0.47
St. Bernard 9 6.75 3.69 3.06 2.52
10 8.82 4.7 4.12
11 8.85 5.52 3.33
12 6.66 5.29 1.37
14 3.42 5.5 -2.08
St. Charles 9 8.1 1.81 6.29 4.73
10 7.41 2.28 5.13
11 5.55 2.11 3.44
12 8.22 2.61 5.61
14 5.85 2.65 3.2
St. John the
Baptist
9 6.06 1.7 4.36 3.00
10 6.12 2.13 3.99
11 4.47 2.08 2.39
12 5.91 2.71 3.2
83

14 3.24 2.64 0.6
St. Tammany 9 4.98 2.63 2.25 10.1
10 5.85 3.38 2.47
11 6.03 4.01 2.02
12 4.14 3.95 0.19
14 2.96 4.03 -1.07
Tangipahoa 9 6.93 2.68 4.25 5.92
10 7.89 3.24 4.65
11 7.47 3.7 3.77
12 6.36 3.9 2.46
14 4.08 4.12 -0.04
Terrebonne 9 7.65 2.73 4.92 1.75
10 5.52 1.64 3.88
11 2.9 1.3 1.6
12 4.56 1.44 3.12
14 1.3 1.25 0.05

Table 7. Depth difference chart in NHC advisories 15-17
Parish Advisory SLOSH Depth
per ft
ADCIRC Depth
per ft
Delta Observed
Depth (ft)
Jefferson 15 6.42 5.75 0.67 6.18
16 5.07 5.54 -0.47
17 4.47 5.32 -0.85
Lafourche 15 6.84 4.28 2.56 1.55
16 2.84 2.68 0.16
17 2.78 2.13 0.65
Orleans 15 2.6 5.01 -2.41 5.83
16 2.26 4.55 -2.29
17 2.12 2.41 -0.29
Plaquemines 15 3.78 5.27 -1.49 2.28
16 2.56 4.34 -1.78
17 2.62 3.57 -1.05
St. Bernard 15 3 5.67 2.67 2.52
16 2.84 4.5 -1.66
17 2.68 3.11 -0.43
St. Charles 15 5.94 2.99 2.95 4.73
16 6.12 2.98 3.14
17 4.59 2.84 1.75
St. John the
Baptist
15 2.98 3.43 -0.45 3.00
16 3.09 3.07 0.02
84

17 1.82 2.23 -0.41
St. Tammany 15 1.76 4.2 -2.44 10.10
16 1.76 4.15 -2.55
17 1.6 3.75 -2.15
Tangipahoa 15 3.9 5.04 -1.14 5.92
16 4.77 4.51 0.26
17 2.9 4.43 -1.53
Terrebonne 15 1.02 1.31 -0.29 1.75
16 0.78 1.04 -0.26
17 0.98 1.02 -0.04

Figure 46. Jefferson Parish Depth Analysis

85

Figure 47. Lafourche Parish Depth Analysis

Figure 48. Orleans Parish Depth Analysis

86

Figure 49. Plaquemines Parish Depth Analysis

Figure 50. St. Bernard Parish Depth Analysis

87

Figure 51. St. Charles Parish Depth Analysis

Figure 52. St. John the Baptist Parish Depth Analysis

88

Figure 53. St. Tammany Parish Depth Analysis

Figure 54. Tangipahoa Parish Depth Analysis

89

Figure 55. Terrebonne Parish Depth Analysis

90

Appendix C Surge Depth by Advisory

Figure 56. Depth Analysis for NHC Advisory 10
91

Figure 57. Depth Analysis for NHC Advisory 11

Figure 58. Depth Analysis for NHC Advisory 12
92

Figure 59. Depth Analysis for NHC Advisory 14

Figure 60. Depth Analysis for NHC Advisory 16
93

Figure 61. Depth Analysis for NHC Advisory 17

Abstract (if available)

Abstract Hurricane Ida struck southeastern Louisiana with winds greater than Hurricane Katrina, and truly tested the rebuilt levee systems of New Orleans and lower Louisiana. Weather and governmental agencies used predictive models to anticipate and predict storm surge locations and severity, as storm surge from hurricanes is the leading fatality cause globally during tropical events. Current tropical systems are fueled by climate change that is impacting storm strength and regularity, yet storm surge models must ensure the highest degree of accuracy. The National Oceanic and Atmospheric Administration’s Sea, Lake, and Overland Surges from Hurricanes (SLOSH) and the Climate Resilience Center 's Advanced Circulation (ADCIRC) models are two preeminent models that simulate and predict storm surge in an effort to publish evacuation orders and save lives. Hurricane Ida rapidly intensified prior to landfall, which stressed the ability of agencies to properly predict storm impacts. In this project, SLOSH and ADCIRC were tested and evaluated against each other, and against observed flooding during Hurricane Ida to determine model strengths and weaknesses. Results show that both SLOSH and ADCIRC overestimated storm surge extent, but underestimated surge depth. ADCIRC was more accurate in long range forecasting, while SLOSH was more accurate in short range forecasting. Recommendations are posed to enhance the accuracy of current storm surge models. This spatiotemporal analysis can help validate surge models against climate enhanced storms with the goal of saving as many lives as possible.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Validating the HAZUS coastal surge model for Superstorm Sandy

PDF

Tracking trends in earthquakes and tropical storms: a web GIS application

PDF

Exploring the pernicious effects of redlining and discriminatory policies on an American city: a spatio-temporal case study of New York City

PDF

Projecting vulnerability: a combined analysis of sea-level rise, hurricane inundation, and social vulnerability in Houston-Galveston, Texas

PDF

The impact of severe coastal flooding on economic recovery disparities: a study of New Jersey communities following Hurricane Sandy

PDF

A spatial investigation of New York City's historical shoreline

PDF

Climate conservation priorities: using MCDA to identify future refuge for the Joshua Tree Forest in California

PDF

A critical assessment of the green sea turtle central west Pacific distinct population segment utilizing maxent modeling on nesting site locations

PDF

Development of a historical urban land use web application for the city of Hong Kong

PDF

Human suffering during wartime: a StoryMap of violations of international law during the Russo-Ukrainian War

PDF

Advancing Redwood City's bicycle infrastructure through a geodesign workflow

PDF

The impact of definition criteria on mapped wildland-urban interface: a case study for ten counties along the Oregon-California border

PDF

A comparison of pixel-based, object-based, deep learning, and data fusion image classification approaches in delineating urban tree canopy in City Terrace

PDF

Impacts of vegetation management on wildfire severity: a study of the 2021 Caldor fire

PDF

Developing a replicable approach for the creation of urban climatic maps for urban heat island analysis: a case study for the city of Los Angeles, California

PDF

Spatiotemporal hotspots of 2018-2020 crime in Houston, Texas

PDF

A spatial analysis of beef production and its environmental and health impacts in Texas

PDF

A comparison of weighted and fuzzy overlays in mapping landslide susceptibility, south-central front range, Colorado

PDF

The use of site suitability analysis to model changes in beach geomorphology due to coastal structures

PDF

The community wildfire protection plan repository: using VGI to create a national collection of wildfire management information

Asset Metadata

Creator Bowers, John Carlson, Jr. (author)

Core Title Spatiotemporal analysis of the SLOSH and ADCIRC storm surge models: a case study of hurricane Ida

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Degree Conferral Date 2022-12

Publication Date 01/04/2023

Defense Date 01/04/2023

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag analysis,climate change,flood,hurricanes,OAI-PMH Harvest,storm surge

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sedano, Elisabeth (committee chair), Huang, Guoping (committee member), Ruddell, Darren (committee member)

Creator Email jcbowers@usc.edu,john.c.bowersjr@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC112710935

Unique identifier UC112710935

Identifier etd-BowersJohn-11399.pdf (filename)

Legacy Identifier etd-BowersJohn-11399

Document Type Thesis

Format theses (aat)

Rights Bowers, John Carlson, Jr.

Internet Media Type application/pdf

Type texts

Source 20230111-usctheses-batch-1000 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu