A preview of this full-text is provided by Wiley.
Content available from Journal of Geophysical Research: Oceans
This content is subject to copyright. Terms and conditions apply.
1. Introduction
Tropical cyclones, also known as hurricanes, are intense weather events which present a significant and
increasing natural hazard in coastal areas (Elko etal.,2019; Field et al.,2014; Knutson et al.,2010; Riss-
er & Wehner,2017; Wahl etal., 2015). Already the most destructive natural disaster in the United States
(Grinsted etal.,2019), these storms are projected to increase in intensity (Wahl etal.,2015), precipitation
(Villarini etal.,2011), and frequency with future climate warming (Knutson etal.,2010). During tropical
cyclones strong winds can produce storm surge, generate large waves, and drive strong currents, resulting in
a multihazard coastal environment. Improving the understanding of how these processes interact to drive
hydrodynamics in estuaries and cause coastal flooding is a vital research area (Chaumillon etal.,2017; Clu-
nies etal.,2017; Elko etal.,2019). This hazard has been highlighted by flooding events during several recent
hurricanes along US coasts, including Hurricane Katrina in 2005 (Fritz etal., 2007), Hurricane Irene in
2011 (Mulligan etal.,2015b), and Hurricane Sandy in 2012 (Bennett etal.,2018; Beudin etal.,2017). These
storms can deliver large amounts of precipitation over short periods, which combined with wind-driven
Abstract During extreme storms, both wind-driven changes in water levels and intense precipitation
can contribute to flooding. Particularly on low-lying coastal plains, storm-driven flooding can cover large
areas, resulting in major damage. To investigate the roles of rainfall and storm surge on coastal flooding,
a coupled flow-wave model (Delft3D-SWAN) that includes precipitation is used to simulate two major
storm events. The modeling system is applied over a domain covering coastal North Carolina, USA,
including the large Albemarle-Pamlico estuarine system, and a long and narrow back-barrier estuary
(Currituck Sound [CS]) that experiences major water level variations is investigated in detail. A high-
resolution (50m) grid with eight vertical layers is used to simulate the conditions during Tropical Storm
Hermine and Hurricane Matthew in 2016. Hindcasts (winds, pressure, and precipitation) from eight
different atmospheric models are used as atmospheric input conditions, and the results are compared
with detailed observations of surface waves, currents, and water levels from sensors mounted on five
monitoring platforms in CS. Results show that major differences exist between wind fields producing
variations coastal conditions. Precipitation directly on the water surface had a large effect on water levels
and produced a larger inundated area. These results help to understand the important contributions of
each physical process (precipitation, wind-driven surge, and waves) to circulation and water levels, and
provide guidance on the impact of atmospheric forcing conditions on back-barrier environments during
hurricanes.
Plain Language Summary Winds and heavy rain can result in coastal flooding during
extreme storms, such as hurricanes. This is a major hazard in coastal areas, where flooding from storms
can cover large areas and produce substantial damage. To help understand the forces that contribute
to this hazard, a model is applied to simulate the waves, water levels, and currents in Currituck Sound
(CS), part of the larger Albemarle-Pamlico estuarine system in coastal North Carolina, USA. Two major
storms are simulated, Tropical Storm Hermine and Hurricane Matthew in September and October 2016.
Observations were collected from five monitoring platforms in CS, and the measurements indicate that
large changes in water levels occurred during both storms. Winds, pressure, and precipitation are used as
model inputs, and the simulation results are compared with observations of surface waves, currents, and
water levels. The results help to understand the roles of precipitation, wind, and waves on the motion of
water in back-barrier estuaries and coastal flooding during tropical storms.
REY ET AL.
© 2020. American Geophysical Union.
All Rights Reserved.
Impacts of Hurricane Winds and Precipitation on
Hydrodynamics in a Back-Barrier Estuary
Alexander J. M. Rey1 , D. Reide Corbett2 , and Ryan P. Mulligan1
1Department of Civil Engineering, Queen's University, Kingston, ON, Canada, 2Integrated Coastal Programs, East
Carolina University, Greenville, NC, USA
Key Points:
• Surface waves and coastal
hydrodynamic were simulated in a
long and narrow estuary during two
tropical storms
• After comparing atmospheric
inputs, model results were in general
agreement with observations from
five monitoring platforms using
Rapid Refresh winds
• Results indicate that both wind-
driven storm surge and precipitation
directly on the water surface
contribute to flooding
Supporting Information:
• Supporting Information S1
Correspondence to:
R. P. Mulligan,
ryan.mulligan@queensu.ca
Citation:
Rey, A., Corbett, D. R., & Mulligan, R.
P. (2020). Impacts of hurricane winds
and precipitation on hydrodynamics
in a back-barrier estuary. Journal
of Geophysical Research: Oceans,
125, e2020JC016483. https://doi.
org/10.1029/2020JC016483
Received 5 JUN 2020
Accepted 30 SEP 2020
10.1029/2020JC016483
RESEARCH ARTICLE
1 of 25
Journal of Geophysical Research: Oceans
storm surge, can produce high total water levels. Recent studies have emphasized the increasing precipi-
tation risks posed by these storms because of climate warming, with Hurricane Harvey (2017) producing
∼38% more precipitation in Houston, Texas, than would have occurred without anthropogenic warming
(Risser & Wehner,2017). In another example, if the ocean surface is warmed by 2°C (RCP 8.5 scenario
[Field etal.,2014]), the simulated landfall from Hurricane Matthew (2016) would produce a 70% increase
in the area flooded along the South Carolina (SC) coast compared to observations (Jisan etal.,2018). De-
spite these increasing risks, flooding hazards can be underestimated, potentially resulting in risk reduction
adaptations are either underdesigned or not implemented (Halstead,2018).
Back-barrier bays and estuaries are found throughout the world, and their low-lying characteristics com-
bined with increasing urbanization makes them particularly vulnerable to flooding from hurricanes
(McGranahan etal.,2007; Miselis etal.,2016; Smallegan etal.,2017; Wahl etal.,2015). A significant amount
of research has been conducted on coastal flooding in these environments, and different numerical models
have been applied to study these systems, with a selection of recent investigations summarized in Table1.
Research has primarily focused on the effects of wind-driven storm surge, and questions remain about the
impacts from precipitation and waves on flooding and circulation (Defne etal.,2019; Mulligan etal.,2015b;
Rey etal.,2019). Notably, an investigation by Peng etal.(2004) used the Princeton Ocean Model (POM) and
the Holland wind model (Holland,1980) to investigate the effects of synthetic tropical storms on the Albe-
marle-Pamlico estuarine system (APES) in North Carolina (NC), USA. This study found that flooding was
highly sensitive to the track and forward speed of the storm, and emphasized the necessity of accurate wind
fields. The potential for different hydrodynamic conditions depending on the selection of atmospheric forc-
ing model was also highlighted by Garzon etal.(2018) during flooding in Chesapeake Bay from several dif-
ferent hurricanes. Using the parametric Holland model and hindcast wind fields, the Rapid Refresh (RAP)
and European Reanalysis (ERA5) atmospheric models were found to provide the most accurate water level
results. Thomas etal.(2019) also found that results were improved when atmospheric model results were
used for model forcing, in an investigation where ADCIRC was used to simulate Hurricane Matthew (2016)
and investigate the nonlinear relationship between tides and wind-driven storm surge on coastal flooding.
Using a combination of the Simulating WAves Nearshore (SWAN), XBeach, and ADvanced CIRCulation
models, a recent investigation by Gharagozlou etal.(2020) accurately simulated offshore wave and water
levels conditions, as well as changes in dune morphology during Hurricane Isabel in the same area as the
present study. As an alternative to parametric models or hindcast wind fields, Clunies etal.(2017) found
that when simulating wind-driven water levels in a large estuary, model performance can be improved by
applying uniform winds measured at a single site over the coastal ocean.
Relatively few studies have investigated the effects of precipitation on coastal flooding in estuaries, and
this has recently been identified by the scientific community as a focus area for research (Elko etal.,2019).
A study by Lin etal.(2010) noted the spatial variation in precipitation that occurs during tropical storms,
and emphasized the need for future studies in this area to improve understanding. This was followed by a
key study by Orton etal.(2012), where precipitation from the North American Mesoscale (NAM) model
was applied in conjunction with the Stevens Institute Estuarine and Coastal Ocean hydrodynamic Model
(sECOM) during Hurricane Irene (2011), and precipitation was found to contribute to a 2% increase in
water levels in the 1.5m deep Hudson River estuary. Notably, precipitation directly onto the hydrodynamic
grid (“Rain-on-grid”) was included in this study, in contrast to the more typical inclusion of precipitation
through river discharges or overland flow (e.g., Brown etal.,2014). By including river discharges, Kumbier
etal.(2018) found a 0.3–1.5m increase in water levels in a large river estuary in Australia during a strong
storm, and Lee etal.(2019) found a 21.7% increase in inundation area around a large bay on the Korean
peninsula compared to a simulated storm surge without river discharges during Typhoon Maemi (2003).
Incorporating river discharges, wind, and waves, Defne etal.(2019) noted a significant increase in the vol-
ume of water in an estuary when offshore swell and storm surge were included in the model, with a smaller
increase from river discharges, as well as a major change in spatial variability of water levels when wind was
included; however, only small effects from including the local wave setup in the model. The hydrodynamic
conditions that occur in estuaries during tropical storms also remain poorly understood, as the typically
shallow and complex bathymetry in back-barrier estuaries makes them particularly complex environments
(Hsu etal.,1999). Bennett etal.(2018) and Clunies etal.(2017) noted that small changes in bathymetry can
REY ET AL. 2 of 25
10.1029/2020JC016483
Journal of Geophysical Research: Oceans
REY ET AL. 3 of 25
10.1029/2020JC016483
Author Year Storm Location Model Atmospheric Forcing
Atmospheric
Resolution Precipitation Resolution
This study 2020 TS. Hermine, H. Matthew APES Delft3D-SWAN 8 Hindcasts 2.5–30km Rain-on-grid/discharge 50m
Defne etal. 2019 H. Sandy Barnegat Bay NJ. COAWST NARR 32km Discharge 40m
Lee etal. 2019 Typhoon Maemi Korean Peninsula Delft3D, HEC-HMS Holland – Discharge 8 m
Thomas etal. 2019 H. Matthew SE US ADCIRC-SWAN WRF+Holland – Discharge 150m
Bennett etal. 2018 H. Sandy New York Delft3D-SWAN WRAMS 18km – 115m
Kumbier etal. 2018 June 2016 storm Shoalhaven Estuary,
Australia
Delft3D Observed – Discharge 25m
Garzon etal. 2018 H. Irene, Sandy, Joaquin,
Jonas
Chesapeake Bay ADCIRC-SWAN 5 Hindcasts+Holland 12–30km – 200m
Clunies etal. 2017 2 Extra-tropical storms APES Delft3D-SWAN NARR/Observed 32km – 250m
Paramygin etal. 2017 H. Andrew, Jeanna, Wilma,
TS. Fay
Southeast US CH3D-SWAN NAM 12km – 10–21m
Miselis etal. 2016 H. Sandy APES COAWST-SWAN Observed – – –
Mulligan etal. 2015b H. Irene APES Delft3D-SWAN Holland – – 250m
Vidal-Ju-arez etal. 2014 8 storm San Quintín, Mexico Delft3D-SWAN Observed – – 50m
Orton etal. 2012 TS. Irene New York sECOM NAM 12km Rain-on-grid NAM/
Discharge
50m
Lin etal. 2010 H. Isabel APES & Chesapeake
Bay
ADCIRC GFS+GFDL 27km WRF/NEXRAD 150m
Bunya etal. 2010 H. Katrina, Rita Southern Louisiana
and Mississippi
ADCIRC+STWAVE H*WIND – Obs. 50m
Peng etal. 2004 10 Cat. 2/3 storms APES POM Holland – – 325m
Abbreviations: ADCIRC, ADvanced CIRCulation; APES, Albemarle-Pamlico estuarine system; GFS, Global Forecast System; POM, Princeton Ocean Model.
Table 1
Summary of Selected Previous Numerical Modeling Investigations on Estuary Hydrodynamics During Storms
Journal of Geophysical Research: Oceans
produce significant hydrodynamic changes, and Mulligan etal.(2015b) found that locally generated waves
during major storms produce important changes to current patterns in estuaries.
The present understanding of the flooding and circulation patterns in back-barrier estuaries during storms
is limited by a lack of detailed observational data and corresponding three-dimensional numerical mode-
ling. In this study, observations of waves, water levels, and currents in a back-barrier estuary are investigated
during two tropical storms using high-resolution numerical modeling to improve understanding of storm
hydrodynamics. The observational dataset includes measurements from five instrumented platforms in
Currituck Sound (CS), in eastern NC, a water body that is representative of many long and narrow back-bar-
rier lagoons along the east coast of North America. The modeling system is a coupled wave-hydrodynamic
model that includes input from time- and space-varying wind and precipitation. The combined effects from
wind fields and precipitation on coastal flooding and circulation are examined, and the results used to im-
prove understanding of the impacts from tropical cyclones on back-barrier estuaries.
2. Methods
2.1. Study Area
The APES in NC (Figure1) is the second largest estuarine system in the United States, and is a key part
of the regional economy (Wells & Kim,1989). Previous studies in this area have highlighted that the high
variability in wind speed and direction during storm events can have dramatic effects on waves (Mulligan
etal.,2015b), water levels (Clunies etal.,2017), productivity (Corbett,2010; Giffin & Corbett,2003), and
morphology (Mulligan etal., 2015a; Paerl etal.,2006). Flooding in this low-lying area frequently occurs
during storm events (Bales,2003; Peng etal., 2004; Wagner etal.,2016), presenting a risk that is likely
REY ET AL. 4 of 25
10.1029/2020JC016483
Figure 1. Map of the model domain including bathymetry, model boundaries, validation sites, selected meteorological
stations, and RAP forecast tracks for Tropical Storm Hermine and Hurricane Matthew. The study location is indicated
on the inset map, and a map with all sites labeled is shown in FigureS1.
Journal of Geophysical Research: Oceans
growing in conjunction with increased urban development (Miselis
etal.,2016), sea level rise (Field etal.,2014), storm frequency (Knutson
etal.,2010), and intensity (Wahl etal., 2015). CS, in the northernmost
segment of the larger APES system (Figure 2), is an ∼58-km-long and
8-km-wide microtidal estuary with averaged depth of 2.5m (Fine,2008;
Moran etal., 2015). Characterized by muddy to sandy sediments, CS is
typical of back-barrier estuaries along the east coast of North America
(Wagner etal., 2016). CS is protected from the ocean by the northern
Outer Banks, and does not have any inlets connecting it directly to the
ocean. It is located ∼60km from the nearest inlet (Oregon Inlet), and
as a result of this distance and the shallow connection through Croatan
sound to Pamlico Sound, water level fluctuations are primarily wind driv-
en (Fine,2008), and there is a minimal (0.1 m amplitude) contribution
from astronomical tides (Caldwell,2001; Moran etal.,2015). Overwash
can occur along the Outer Banks barrier islands during very strong storm
events, producing widespread flooding and significant damage (Conery
etal.,2018; Gharagozlou etal., 2020). During storms without overwash,
ocean waves, or ocean storm surges do not have a large effect on the con-
ditions in CS (Long & Resio,2007; Moran etal.,2015). Salinity is typical-
ly low (2–5 ppt), reaching up to 10 ppt during large storms with ocean
seawater overwash (Robinson & McBride,2006). Only minor freshwater
flows (∼7m3/s) enter CS, and about one third of the total drainage area
consists of open water (Fine,2008).
2.2. Hurricanes
The impacts of two tropical cyclones that made landfall in 2016 are stud-
ied to gain a more comprehensive understanding of the estuarine re-
sponse to storms. These include Tropical Storm (TS) Hermine and Hurri-
cane (H) Matthew. Forming on August 28, 2016, Hermine reached a peak
of intensity on September 2 as a category 1 storm on the Saffir-Simpson
scale, and traveled across Florida and along the US east coast before weakening to an extratropical storm
over Oregon Inlet, NC (Figure1, site OI). Over the APES, 200mm of rain was measured, and a peak wind
speed of 32m/s was recorded at the Field Research Facility (FRF) pier (Figure1, site FP), producing a storm
surge of 0.6–1.2m above mean sea level on the sound side of the Outer Banks (Berg,2017). In the US, Her-
mine was responsible for 1 death and ∼$550 million USD in damage (Klein,2016). Forming a month later
on September 28, 2016, Hurricane Matthew reached a peak intensity on October 1 as a category 5 storm,
with estimated wind speeds of 75m/s, and set the record for the southernmost category 5 hurricane in the
Atlantic basin (Stewart,2017). Making landfall on October 8 in SC, Matthew passed south of the Outer
Banks on October 9 as a category 1 storm with sustained winds of 36 m/s, producing extensive coastal
flooding and a peak sound side surge of 1.55m above mean sea level at the Hatteras Island US Coast Guard
Station (Figure1, site HT). Matthew delivered heavy rainfall, with total precipitation typically exceeding
250mm along the storm track, and a peak value of 480mm recorded near Evergreen, on the SC/NC border.
Across the US, 34 direct fatalities and $10 billion USD in damages were associated with Matthew, with
88,000 homes damaged in NC. While 24 fatalities were reported in NC because of flooding caused by heavy
precipitation, none were associated with the storm surge (NCRPP,2017; Stewart,2017; Wang etal.,2017).
While overwash did not occur in CS, since H Matthew never made direct landfall in NC, extensive flooding
and damage was observed along estuarine shorelines (Smith & Scyphers,2019).
2.3. Data Collection
Data were collected by the US Army Corps of Engineers (USACE) at five observation platforms (CS 1–CS
5) in CS (Figure2), from January 2016 to January 2018. Observations include winds, water levels, waves,
and currents, measured by an array of sensors on each platform. Winds were measured at a 6m elevation
REY ET AL. 5 of 25
10.1029/2020JC016483
Figure 2. Map of Currituck Sound, including bathymetry and observation
platforms.
415 420 425 430435
UTM X (km)
3990
3995
4000
4005
4010
4015
4020
4025
4030
UTM Y (km)
CS 1
CS 2
CS 4
CS 3
CS 5
CC
-5
-4
-3
-2
-1
0
1
2
3
4
5
h (m)
Water level observations
Sensor platforms
Journal of Geophysical Research: Oceans
above the water surface using RM Young Marine Anemometers sampling at 1Hz and averaged over 10min
intervals. Water levels were measured with Xylem Waterlog sensors by averaging 360 sample bursts taken at
1Hz. Seabird SBE26+ pressure sensors were used for wave measurements, collecting a 4,096 sample bust at
4Hz every 30min. Vertical profiles of horizontal currents were observed using Nortek Aquadopps, operat-
ing at a frequency of 2MHz, by taking 600 ensemble-averaged samples every 30min with a vertical resolu-
tion of 0.4m. Water levels and wave heights were collected at all five sites, while currents and winds were
collected at four sites. Water level data at five other sites shown in Figure1 were collected by the US Geolog-
ical Survey (USGS) and National Oceanographic and Atmospheric Administration (NOAA). Offshore wave
data were collected at six sites by the USACE, NOAA, and the Coastal Data Information Program (CDIP).
Additional meteorological observations of wind and precipitation were collected by the National Weather
Service (NWS) at 12 sites and used for comparison with atmospheric model results. A complete list of all
observation sources is provided in Table2.
2.4. Coupled Numerical Models
Numerical models are commonly applied to improve the understanding of coastal processes during storms;
however, significantly more research has been conducted on model performance in ocean environments com-
pared to back-barrier estuaries. In the present study, Delft3D-SWAN is applied to improve the understanding
of storm impacts in a long and narrow estuary, and this model has been successfully applied in a variety of
coastal environments (Booij etal.,1999; Lesser etal.,2004). In Delft3D, the Navier-Stokes horizontal momen-
tum equations are solved by the hydrodynamic module (FLOW), which simulates water levels and currents
from boundary inputs and spatially varying meteorology. Waves and wave-current interactions are solved us-
ing the SWAN model. SWAN is a third-generation shallow water spectral wave model that simulates wave gen-
eration, propagation, and dissipation, and is coupled to FLOW at specified time steps. In the APES, Delft3D-
SWAN has been previously used to understand storm impacts over short timescales corresponding to a storm
event (Mulligan etal.,2015b), longer timescales covering a series of storms over a month (Clunies etal.,2017),
and the hydrodynamic responses to morphological change over long timescales (Mulligan etal.,2019).
2.4.1. Model Setup
A regular, orthogonal grid was constructed over the domain shown in Figure1 for this study. Covering the
entire APES, the domain includes back-barrier estuaries, inlets, barrier islands, and the coastal ocean to a
depth of ∼20m. The hydrodynamic grid has a horizontal resolution ranging between 50 and 100m, with the
highest resolution areas covering CS and the Outer Banks barrier islands. The NOAA Coastal Relief Model
(NOAA,2020) bathymetry, with a horizontal resolution of ∼30m and referenced to the NAVD88 vertical
datum, was used in combination with high resolution (5m) bathymetry observations from CS collected by
CSA Ocean Sciences Inc. in February 2018.
In the hydrodynamic (FLOW) model, a 15s time step is used with a vertical grid of eight equally spaced
bathymetry-following layers, providing a minimum vertical resolution of 0.35m in CS. Simulations are per-
formed over 46h, which includes a 36-h period during the passage of each storm with an additional 10h of
simulation time before each event for spin-up, with initial conditions set from observed water levels. Default
hydrodynamic settings (Cz=65m2/s, Hv=1m0.5s−1) are applied after sensitivity testing had only a minor
influence on currents (<0.01m/s; TableS11), and water levels (<0.02m; TableS9). Parameters used in sen-
sitive testing were selected based on previous studies by Mulligan etal.(2015b) and Bastidas etal.(2016),
including the time step (7.5–15 s), uniform bottom roughness coefficient (Cz=45−95 m2/s), horizontal
eddy viscosity coefficient (Hv=0.1m0.5s−1−10m0.5s−1), and the wind drag coefficient (default, Yelland and
Taylor [1996]). Model sensitivity to river inflows was also examined using time-series observations from the
nine river discharge sites shown in Figure1 and detailed in TableS3. As a result of the relatively low flow
rates, short study period, and the fact that no rivers flow directly into CS, negligible water level impacts were
found in the study area during the simulation period (TableS9), and discharges are therefore excluded from
simulations. A 0.1 m depth is used to define the grid cell flooding threshold, following the application of
Defne etal.(2019).
Waves are simulated over the same domain as hydrodynamics; however, a grid with a coarser horizontal
resolution (100–200m) is used for computational efficiency. The wave frequency space is defined using
REY ET AL. 6 of 25
10.1029/2020JC016483
Journal of Geophysical Research: Oceans
49 logarithmically spaced bins from 0.05–3.00Hz, and the directional space is defined using 36 bins with
10°resolution. Waves are simulated in stationary mode to improve model stability, and coupled “online”
in 1h intervals with hydrodynamics, allowing for results to be communicated between models. After sen-
sitivity tests for the depth induced breaking parameter, nonlinear triad interactions, whitecapping formu-
la, and drag coefficient found only minimal effects (<0.01m) on significant wave heights (TableS10), all
parameters were left at default values. Notably, the van der Westhuysen whitecapping formula (Mulligan
etal.,2008; van der Westhuysen et al.,2007) had a small positive effect (1cm decrease in RMSD) on the
accuracy of significant wave heights, particularly at the more sheltered CS 4 and 5 sites. The focus of the
remainder of this study is on the sensitivity of model results to different wind field inputs.
REY ET AL. 7 of 25
10.1029/2020JC016483
ID Name Parameters
Depth/
elevation
(NAVD
88m) Source
CS 1 Currituck Sound 1 Water level; wind; wave; currents 3.4m USACE
CS 2 Currituck Sound 2 Water level; wave; currents 2.7m USACE
CS 3 Currituck Sound 3 Water Level; wind; wave 2.3m USACE
CS 4 Currituck Sound 4 Water level; wind; wave; currents 2.6m USACE
CS 5 Currituck Sound 5 Water level; wind; wave; currents 2.7m USACE
FP FRF Pier Water level; wind, precipitation 6m USACE
BF Beaufort Duke Marine Lab Water level; wind 0m NOAA/Duke
OI Oregon Inlet Marina Water level 0m NOAA
HT Hatteras Coast Guard Water level; wind 0m NOAA/USCG/ISU/ASOS
AS Albemarle Sound @ Leonards Point Water level 0m USGS
CC Currituck Sound @ Corolla Water level 0m USGS
VB Virginia Beach Wave Wave; wind 47m NDBC/USACE
DS Diamond Shoals Buoy Wave; wind 59m NDBC
OB Oslow Bay Buoy Wave 30m CDIP/USACE
CH Cape Henry Buoy Wave 18m CDIP/USACE
O18 Oregon Inlet Buoy Wave 18m NDBC/UNC
F17 FRF 17m Buoy Wave 17m USACE
FFA First Flight Airport Wind, precipitation 4.00m ISU/ASOS
MDA Manteo Dare Airport Wind, precipitation 4.00m ISU/ASOS
ECG Elizabeth City Coast Guard Wind, precipitation 7.50m NWS
CPM Cherry Point Marine Corps Air Station Wind, precipitation 9.90m NWS
EWN Coastal Carolina Airport Wind, precipitation 6.10m NWS
HSE Billy Mitchell Airport Wind, precipitation 5.2m NWS
NJM Swansboro Bogue Field Wind, precipitation 6.40m NWS
ONX Currituck Country Airport Wind, precipitation 5.50m ISU/ASOS
EDE Edenton Northeast Airport Wind, precipitation 6.10m NWS
OCW Warren Field Airport Wind 11.80m NWS
NBT Piney Island Wind 2 m NWS
CPL Cape Lookout Wind 4.60m NDBC
Note. Data sources highlighted in Figure1 are marked in bold.
Table 2
List of Data Sources
Journal of Geophysical Research: Oceans
2.4.2. Model Forcing
The model is forced using a combination of observations and large-scale atmospheric models. Water levels
are prescribed at the boundaries used spatially interpolated observations at the USACE FRF pier (located
∼500m offshore) and the Beaufort Marine Lab tide gauge stations (Figure1 FP and BF), similar to the
approach used by Clunies etal. (2017) and Mulligan etal.(2019). Observed 2-D wave spectra from four
offshore directional wave buoys (Figure1 OB, DS, VB, and CH) are interpolated along the domain boundary
for offshore waves. Observations from these stations are shown over a 47-day period covering both storms
in Figure3. As previous investigations have highlighted the importance of accurate atmospheric (precipita-
tion, pressure, and winds) forcing (Bennett etal.,2018; Dietrich etal.,2018; Garzon etal.,2018), preliminary
analysis was completed on eight atmospheric models: analysis products from the Global Forecast System
(GFS), NAM Forecast System, and RAP; reanalysis datasets from the Climate Forecast System (CFSv2) and
the ERA5; Zero-hour (initialization) products from the Regional Deterministic Prediction System (RDPS),
and the High Resolution Rapid Refresh Model (HRRR); and the satellite derived Modern-Era Retrospective
analysis for Research and Applications (MER), described in Table S1. To improve understanding of the
impact of precipitation, preliminary analysis was also completed on three additional gridded precipitation
datasets in combination with the RAP wind field: the satellite derived Global Precipitation Measurement,
the Canadian Precipitation Analysis, and the Multisensor Precipitation Estimate; with no increase in accu-
racy observed (TableS5).
From these atmospheric models, three datasets are selected to represent a range of methodologies (hindcast
and nowcast), sources (NOAA and the European Centre for Medium Range Weather Forecasting), and
which produced the strongest agreement with observations: the RAP analysis (Benjamin etal.,2016), with
a 13km domain covering the continental US, the CFSv2 reanalysis (Saha etal.,2010), with a 27km global
REY ET AL. 8 of 25
10.1029/2020JC016483
Figure 3. Regional observations, including (a) wind components at First Flight Airport (Figure1 FFA); (b) water
levels; (c) offshore significant wave heights; (d) 36-h precipitation accumulation. Modeling periods for TS Hermine and
H Matthew are shown in gray.
Journal of Geophysical Research: Oceans
domain, and the ERA5 (Hersbach etal.,2019), with a 30 km global domain. For simulations including
precipitation, only rainfall directly on the water surfaces (i.e., wet grid cells) is included, with precipitation
over land removed (assuming 100% infiltration). Sensitivity testing using 0%, 10%, 25%, 40%, and 55% runoff
was completed by varying the precipitation rate over land areas, with only minimal effects (<0.02m) on
modeled water levels in CS (TableS2), similar to the findings from Dresback etal.(2013).
3. Observations
3.1. Regional Observations
Observations summarizing the atmospheric and ocean conditions are shown in Figure3 over the period
from August 29 and October 15, 2016. These observations include wind components (Uwx, Uwy) and mag-
nitudes (|Uw|), water levels (η), significant wave height in the ocean (Hs), and precipitation accumulation.
Wind speeds were typically below 10m/s prior to and between the two major storm events; however, peaked
to more than 35m/s during both storms, and wind direction rapidly shifted over CS. Water levels were dom-
inated by ∼1.2m semidiurnal tides at stations FP and BF (used for model forcing), and by ∼0.3m semidi-
urnal tides at station OI, with an ∼0.2m subtidal variability on the estuary side (station HT). The combined
tide and storm surge peaked at 1.2m during TS Hermine and 1.9m during H Matthew, with the highest
surge for both storms occurring at station HT, inside the Outer Banks, in Pamlico Sound. Significant wave
heights in the ocean were relatively consistent between stations, with Hs typically less than 2m and peaks
of ∼6m. Precipitation from the storms is evident in Figure3d, particularly during H Matthew, when a total
accumulation of more than 200mm was recorded at station ECG.
3.2. Currituck Sound Observation Platforms
Water levels showed considerable gradients across both the longitudinal and cross-estuary axes of CS, var-
ying by up to 0.45m between CS 1 and 3 and by up to 0.25m between CS 4 and 5. Observations from the
monitoring platforms in CS are shown in Figure4 during 36-h periods covering the two storms. Winds were
generally consistent between stations; however, a small (∼4m/s) decrease in wind speed at the southern-
most observation point (CS 1) occurred at 16:00 UTC on September 3, coinciding with the closest approach
of the eye of TS Hermine to CS. At any location in the sound the significant wave height varied with respect
to the wind direction, due to the limitation of fetch from any direction. For example, the east-west winds
observed during TS Hermine produced smaller waves at CS 5 on the eastern side of the estuary; while the
north-south wind pattern of H Matthew had the largest waves at the southernmost CS 1 site. Notably, the
very sudden increase in wind speed during H Matthew produced a sharp increase in significant wave height
throughout the estuary, rising from below 0.3m–1.4m over a 2-h period. Spatial variability in CS is clearly
visible in observations of the depth-averaged currents and water levels. The current speed varied signifi-
cantly at different sites, particularly during H Matthew, with the highest speeds (1.5m/s) along the longitu-
dinal centerline of the sound (CS 1, 2, 3). A different, complex, temporally and spatially varying water level
pattern occurred during each storm in conjunction with different precipitation patterns, wind speeds, and
wind directions, with water levels rapidly rising and falling (within a 5-h period) during TS Hermine and a
peak water level of 1.0m at several platforms during H Matthew.
3.3. Large-Scale Atmospheric Models
Wind fields from the three (RAP, ERA5, and CFSv2) selected large-scale atmospheric models examined
in this study are visualized, at the time step with the highest wind speed in the study area, over the APES
region and CS for the two storms in Figure5. Previous studies, including Garzon etal.(2018) and Bennett
etal.(2018), have emphasized the importance of accurate wind fields in order to accurately simulate coastal
environments. Qualitative differences between wind fields are apparent in the high variability of the size
and shape of the eye during TS Hermine. During the storm, a relatively small eye located directly over the
Outer Banks is present in the RAP model output (Figure5a), contrasting the larger eye farther offshore in
the ERA5 model output (Figure5e). These differences extend into the CS study area, with the larger eye
in the ERA5 and CFSv2 models producing lower wind speeds over the sound (Figures5b and 5f/5j). The
variations between wind models are less pronounced during H Matthew; however, the enhanced resolution
REY ET AL. 9 of 25
10.1029/2020JC016483
Journal of Geophysical Research: Oceans
of the RAP model (13km) results in wind speed differences over land and water, particularly over Pamlico
Sound, that are not seen in the ERA5 (30km) and CFSv2 (27km). The quantitative comparison of the mean
distances between the track of the modeled minimum sea level pressure (SLP) and the National Hurricane
Center (NHC) best track (Landsea & Franklin,2013) match the qualitative evaluation of the wind fields
(TableS12). Average eye error (location of the minimum SLP and the NHC best track over the 36-h model
period) was 33km in the RAP model, similar to the GFS model (30km), and more accurate than the ERA5
(52km) or CFSv2 (66km) hindcasts. These patterns are mirrored by statistical comparisons of wind field
accuracy shown in TableS13 for TS Hermine and TableS14 for H Matthew, where the RAP model had the
lowest RMSD and highest correlation coefficient for both storms.
During both storms, observations of wind speed and direction over CS were generally aligned with atmos-
pheric models. This is detailed in Figure6 for the u and v wind components at CS 1 for the three selected
large-scale atmospheric models. All three models showed broad agreement overall, with RAP providing the
best qualitative match to observations. Despite this agreement, for both storms, observed wind speeds were
higher and varied more quickly than modeled speeds. During TS Hermine, wind directions slowly changes
from westward to southward in both models and observations; however, timing differences are evident,
with a faster change in direction from southwesterly to southeasterly winds in the observations. Differences
REY ET AL. 10 of 25
10.1029/2020JC016483
Figure 4. Observations in Currituck Sound during TS Hermine and H Matthew: (a, b) wind speeds; (c, d) wind
directions; (e, f) water levels; (g, h) depth averaged current magnitudes; (i, j) significant wave heights; (k, l) peak wave
periods. All times are in UTC.
0
10
20
30
|Uw| (m/s)
TS. Hermine
(a)
H. Matthew
150
200
250
300
350
Uw
(deg)
-1
0
1
(m/s)
0
0.5
1
1.5
|u| (m/s)
0
0.5
1
1.5
Hs (m/s)
09/03 00h 09/03 18h 09/04 12h
Date (2016)
0
5
Tp (s)
10/09 00h 10/09 18h 10/10 12h
Date (2016)
CS 1
CS 2
CS 3
CS 4
CS 5
(b)
(c)
(e)
(g)
(i)
(k)
(f)
(h)
(j)
(l)
(d)
Journal of Geophysical Research: Oceans
REY ET AL. 11 of 25
10.1029/2020JC016483
Figure 5. Comparison of hindcast wind fields, interpolated to the model grid, used as model inputs for TS Hermine and H Matthew; close up of wind fields
for TS Hermine and H Matthew. Observations are shown by red vectors and colored circles, adjusted to a 10m measurement height, on the same scale.
Visualizations for all eight wind fields from in TableS1 are shown in FigureS8.
Figure 6. Comparison of modeled and observed (adjusted to a 10m measurement height) east and north wind
components at CS 1 during TS Hermine and H Matthew. Vertical lines indicate selected times for Figures5, 7, and
13–15. All times are in UTC.
09/10 00h 09/10 18h 10/10 12h
Date (2016)
-20
0
20
U
wx
(m/s)
TS. Hermine
CS 1
03/09 00h 03/09 18h 04/09 12h
Date (2016)
-20
0
20
U
wy
(m/s)
CS 1
H. Matthew
(a)
(c) (d)
(b)
Journal of Geophysical Research: Oceans
are also evident during H Matthew, particularly during the first 6h of the storm when a brief northwest
wind was observed but not resolved by the models. Despite the relative similarities between modeled and
observed winds, the orientation and geometry of CS make it highly sensitive to wind direction, especially
on the along-estuary axis, allowing small differences between models to produce large changes in the hy-
drodynamic conditions. Differences in both overall quantity and spatial distribution of rainfall are evident
between the two tropical storms, matching the general trends in surface observation, as shown in Figure7
for the three selected atmospheric models, with spatial distribution visualized in FigureS10 and time series
comparisons in FigureS11 for all models. All three atmospheric models indicate lower rainfall totals during
TS Hermine than H Matthew; however, the distribution of rainfall varies considerably. Overall precipitation
patterns between atmospheric models were broadly similar during H Matthew, with large precipitation
bands over Albemarle and CSs. The local accumulation of precipitation varied considerably between mod-
els, ranging from less than 100–225mm between the CFSv2 and ERA5 models over the southern region of
CS (TableS15).
4. Results
A comprehensive overview of the model performance is shown in Figure8 using modified Taylor diagrams
(Elvidge etal.,2014; Taylor,2001) for water levels (η) and significant wave height (Hs). These diagrams
help to visualize three statistics on a single plot, and show models with stronger agreement to observations
REY ET AL. 12 of 25
10.1029/2020JC016483
Figure 7. Comparison of hindcast precipitation fields interpolated to the model grid: 36-h totals for TS Hermine (between September 3, 00:00 UTC and
September 4, 12:00 UTC) and H Matthew (between October 9, 00:00 UTC and October 10, 12:00 UTC); close up of wind fields for TS Hermine and H Matthew.
Observations are shown by colored circles on the same scale. Visualizations for all eight precipitation fields from in TableS1 are shown in FigureS10.
Journal of Geophysical Research: Oceans
using points that are located closer to the normalized observation point. Correlation coefficients (R) are
shown along the azimuthal angle. Normalized standard deviations
*
, shown on the radial axis, are
calculated by normalizing model standard deviations (
m
) against observed standard deviations (
o
). Stand-
ard deviation normalized Centered-Root-Mean-Square-Differences (CRMSD, or bias corrected RMSD) are
radially distributed from the normalized observation point at
*
=1 and R= 1. Statistics are shown for
hydrodynamic model forced by the three selected atmospheric models (RAP, ERA5, CFSv2), in addition to
the RAP model without precipitation. The overall statistics show that the RAP model with precipitation
has the strongest agreement with observations. Particularly during TS Hermine, both water levels (η) and
significant wave heights (Hs) were more accurate when using RAP winds than either ERA5 or CFSv2. Model
performance was less sensitive to wind field during H Matthew, evident in the tighter cluster of points; how-
ever, the RAP model still outperformed the other atmospheric models in terms of providing more accurate
inputs for the wave and storm surge models. Notably, all atmospheric inputs produced simulated waves and
REY ET AL. 13 of 25
10.1029/2020JC016483
Figure 8. Taylor diagrams showing three important statistics that quantify agreement between model results and observations (correlation coefficient (R: green
lines), standard deviation normalized centered-root-mean-square-difference (CRMSD: blue circles with origin at
*1
and R=1), and normalized standard
deviation (
*/
mo
: radially from black circles with origin at
*0
) over 36h between September 3, 00:00 UTC and September 4, 12:00 UTC (TS Hermine:
a, c) and October 9, 00:00 UTC and October 10, 12:00 UTC (H Matthew b, d) at selected sites for: (a, b) water levels; (c, d) significant wave heights. Black dots
represent observations.
(a)
(c) (d)
(b)
Journal of Geophysical Research: Oceans
water levels with lower normalized standard deviations than observations, indicating smaller fluctuations
in model results than observations during both storms. While effect reduces the overall model agreement
with observations (R=0.8 for the RAP winds), the averaged RMSD values in the RAP forced model were
sufficiently low (0.20m) to provide confidence in additional analysis.
4.1. Water Levels
Modeled water levels at the five CS monitoring platforms (CS 1–5) are shown through time in Figure9 for
selected atmospheric forcings, with statistics included in Table3 for all atmospheric models. Results from
all models are included in FigureS3. For both storms, RAP winds produced water levels that had the best
agreement with observations (RMSD < 0.25m), with a correlation coefficient (R) of 0.84 during TS Hermine
and 0.75 during H Matthew. The Delft3D-FLOW model was partially able to resolve the complex spatial
variations in water levels that occurred during TS Hermine, correctly simulating the falling water levels at
CS 3, and varying water levels at other sites. Using atmospheric forcing from the ERA5 and CFSv2 hindcasts
did not produce as accurate water levels, with consistently elevated water levels compared to observations.
Particularly during H Matthew, simulation accuracy was improved when precipitation was included in
the model. Including precipitation had similar effects at all 5 CS stations, increasing the total water level
by ∼20cm in all models, and decreasing the RMSD by 10cm. Despite the improvement in accuracy from
the addition of precipitation, all wind fields underestimated the peak water levels that occurred during the
storms, and overestimated the minimum water levels during the storms, particularly toward the complex
southern inlet of CS, suggesting that the hydrodynamics in the larger Albemarle Sound may be not be fully
captured by the model. During both storms, the HRRR model was the only model with similar performance
to the RAP model, a notable result as the 3.5km resolution of the HRRR model is sufficient to resolve local
REY ET AL. 14 of 25
10.1029/2020JC016483
Figure 9. Water level observations and model results for simulations with and without precipitation during TS
Hermine and H Matthew. Vertical lines indicate selected times for Figures5, 7, and 13–15. Results from all tested
atmospheric models are shown in FigureS3. All times are in UTC.
(d)
(f)
(h)
09/10 00h 09/10 18h 10/10 12h
Date (2016)
(j)
-1
0
1
(m)
TS. Hermine
(a) CS 3
-1
0
1
(m)
(c) CS 2
-1
0
1
(m)
(e) CS 1
-1
0
1
(m)
(g) CS 4
03/09 00h 03/09 18h 04/09 12h
Date (2016)
-1
0
1
(m)
(i) CS 5
H. Matthew
(b)
Journal of Geophysical Research: Oceans
wind effects in CS. The results from the MER and RDPS forcings were not in good agreement with obser-
vations (Table3).
4.2. Waves
Statistics for significant wave heights are shown in TableS6 for all atmospheric forcings, with time series
results in Figure10 for three selected forcing conditions. During TS Hermine, results are in very good
agreement with observations at the CS monitoring platforms (RMSD=0.12m, R=0.83), capturing both
the steady increase and spatial variation in wave heights across the sound. Aligning with the along-estuary
wind direction, waves increase in height between the northernmost CS 3 site and the southern CS 1 site.
Compared to water level results, the wave model is less sensitive to the input wind field, with broadly sim-
ilar performance between wind fields, except for the lower accuracy (R=0.16) NAM wind model. Outputs
from simulations with and without precipitation are similar. Model agreement is lower during H Matthew
(RMSD=0.17m) due to a 3-h lag in the increase in significant wave height between modeled and observed
results. This aligns with model errors in the speed and direction of the input wind fields (Figure6) com-
pared to observations.
The wave spectra are shown in Figure11 at a time step during each storm at three selected observation sites
(CS 3,2,1). Similar peak frequencies occurred during both storms; however, larger high frequency waves
(>0.5Hz) provided more total energy during H Matthew than TS Hermine. While the overall shape and
spatial trends of the local wind-generated wave spectra are generally in agreement with measurements,
REY ET AL. 15 of 25
10.1029/2020JC016483
RMSD/R
η(m) CS 1 CS 2 CS 3 CS 4 CS 5 OI HT Mean CS 1 CS 2 CS 3 CS 4 CS 5 OI HT Mean
(A) TS Hermine (B) H Matthew
RAP no
precp.
0.17 0.16 0.08 0.31 0.19 0.22 0.32 0.21 0.31 0.31 0.24 0.55 0.28 0.20 0.34 0.32
0.78 0.70 0.95 0.36 0.76 0.82 0.90 0.75 0.60 0.75 0.89 0.41 0.72 0.89 0.93 0.74
RAP precp. 0.16 0.16 0.10 0.21 0.23 0.16 0.23 0.18 0.18 0.19 0.12 0.41 0.18 0.14 0.27 0.21
0.85 0.82 0.97 0.56 0.86 0.89 0.93 0.84 0.67 0.74 0.90 0.37 0.70 0.90 0.93 0.75
HRRR precp. 0.13 0.12 0.11 0.21 0.18 0.18 0.23 0.17 0.20 0.22 0.16 0.42 0.20 0.20 0.30 0.24
0.89 0.96 0.98 0.76 0.96 0.86 0.93 0.90 0.62 0.83 0.86 0.48 0.78 0.77 0.94 0.76
CFSv2 precp. 0.28 0.35 0.34 0.30 0.45 0.24 0.25 0.31 0.18 0.21 0.17 0.41 0.21 0.17 0.22 0.22
0.35 0.00 0.25 0.00 0.00 0.60 0.92 0.30 0.55 0.58 0.74 0.15 0.51 0.84 0.94 0.62
RDPS precp. 0.29 0.34 0.43 0.30 0.43 0.48 0.50 0.39 0.26 0.28 0.22 0.49 0.27 0.21 0.20 0.28
0.00 0.12 0.45 0.00 0.03 0.14 0.70 0.20 0.24 0.35 0.64 0.00 0.18 0.54 0.89 0.40
NAM precp. 0.23 0.32 0.33 0.26 0.41 0.21 0.36 0.30 0.21 0.25 0.18 0.42 0.24 0.22 0.44 0.28
0.57 0.00 0.00 0.00 0.00 0.75 0.86 0.31 0.41 0.32 0.66 0.00 0.26 0.65 0.68 0.43
GFS precp. 0.22 0.28 0.26 0.26 0.37 0.21 0.32 0.27 0.30 0.31 0.21 0.54 0.29 0.21 0.35 0.31
0.41 0.00 0.36 0.00 0.00 0.80 0.88 0.35 0.46 0.56 0.84 0.18 0.49 0.84 0.75 0.59
ERA5 precp. 0.24 0.30 0.24 0.31 0.39 0.23 0.29 0.29 0.23 0.26 0.18 0.48 0.25 0.21 0.29 0.27
0.08 0.00 0.19 0.00 0.00 0.66 0.89 0.26 0.51 0.50 0.75 0.09 0.43 0.79 0.94 0.57
MER precp. 0.24 0.31 0.30 0.26 0.40 0.20 0.33 0.29 0.26 0.29 0.20 0.49 0.27 0.26 0.37 0.31
0.43 0.00 0.15 0.00 0.00 0.79 0.80 0.31 0.37 0.34 0.73 0.00 0.26 0.51 0.74 0.42
RMS 0.21 0.26 0.29 0.23 0.36 0.42 0.66 0.35 0.45 0.38 0.28 0.59 0.34 0.50 0.95 0.50
Abbreviations: CFSv2, Climate Forecast System; ERA5, European Reanalysis; GFS, Global Forecast System; HRRR, High Resolution Rapid Refresh Model;
MER, Modern-Era Retrospective analysis for Research and Applications; NAM, North American Mesoscale; RAP, Rapid Refresh; RDPS, Regional Deterministic
Prediction System.
Table 3
Root-Mean-Square-Difference (Upper, Bold) and Correlation Coefficient (Lower) for Selected Atmospheric Models for all Water Level Observation Points During:
(A) TS Hermine and (B) H Matthew. The Observed Root-Mean-Square (RMS) Level is Included on the Bottom Row
Journal of Geophysical Research: Oceans
model results are slightly biased toward higher frequencies, simulating shorter waves with a smaller peak
period (Tp) compared to observations. Model peak wave frequencies (fp) were approximately fp=0.65Hz
(Tp=1.6s) at CS 3 and approximately fp=0.45Hz (Tp=2.3s) at the sites with longer fetch (CS 1 and 2).
This can be contrasted with observed peak frequencies and peak periods of approximately fp=0.45Hz
(Tp=2.2s) at CS 3, fp=0.32Hz (Tp=3.1s) at CS 2, and fp=0.25Hz (Tp=4.0s) at CS 1. Modeled peak wave
frequencies were similar regardless model parameters (FigureS7) or the atmospheric wind field applied
(FigureS2), aligning with significant wave height results.
REY ET AL. 16 of 25
10.1029/2020JC016483
Figure 10. Significant wave height observations and model results for selected atmospheric forcing TS Hermine and H
Matthew. Vertical lines indicate selected times for Figures5, 7, and 13–15. All times are in UTC.
(d)
09/10 00h 09/10 18h 10/10 12h
Date (2016)
(f)
0
0.5
1
1.5
H
s
(m)
TS. Hermine
(a) CS 3
0
0.5
1
1.5
H
s
(m)
(c) CS 2
03/09 00h 03/09 18h 04/09 12h
Date (2016)
0
0.5
1
1.5
H
s
(m)
(e) CS 1
H. Matthew
(b)
Figure 11. Wave spectrum observations and model results for the full model using RAP winds at a selected time step
during TS Hermine and H Matthew. Modeled (red) and observed (black) significant wave heights are indicated at the
same time step.
0
0.2
0.4
0.6
E(f) (m
2
s)
TS. Hermine
03-Sep-2016 14:00
(a) CS 3 Hs = 0.62 m
Hs = 0.41 m
0
0.2
0.4
0.6
E(f) (m
2
s)
(c) CS 2 Hs = 0.83 m
Hs = 0.63 m
0 0.5 1
frequency (Hz)
0
0.2
0.4
0.6
E(f) (m
2
s)
(e) CS 1 Hs = 0.77 m
Hs = 0.71 m
H. Matthew
09-Oct-2016 08:00
(b) CS 3
Hs = 0.98 m
Hs = 0.66 m
Observations
Model
(d) CS 2 Hs = 1.09 m
Hs = 0.90 m
0 0.5 1
frequency (Hz)
(f) CS 1 Hs = 1.42 m
Hs = 0.91 m
Journal of Geophysical Research: Oceans
4.3. Currents
Vertical profiles of horizontal velocity at two selected CS monitoring platform sites (CS 2,5) are shown
through time (x-axis) and depth (y-axis) for the cross-estuary (u) and along-estuary (v) velocity component
in Figure12 for TS Hermine and H Matthew. Observations are shown above model results forced using RAP
winds and precipitation, and velocity magnitude profiles are included in the (FigureS4). Significant spatial
variation is seen across CS in both the direction and magnitude of the observed velocities (Figures4g and
4h), as well as between the two storms.
During TS Hermine, a recirculation pattern occurs in the cross-estuary current profile, with currents trave-
ling in the westward (u) direction on the surface, in line with the prevailing wind direction, and flowing in
the opposite direction near the bed. Near the center of the estuary (site CS 2), a cross-estuary recirculation
pattern emerges early in the storm. However, later in the storm this pattern dissipates in the cross-estuary
direction and strengthens in the along-estuary direction. As further evidence of the hydraulic complexity of
REY ET AL. 17 of 25
10.1029/2020JC016483
Figure 12. Modeled and observed across (east: u) and along-estuary (north: v) velocity component profiles at CS 2 and
CS 5 during H Matthew and TS Hermine. All times are in UTC.
Journal of Geophysical Research: Oceans
this environment, northward flow was observed at CS 5, contrasting southward flow at CS 2. Model agree-
ment with the observation was mixed during TS Hermine (u-component R=0.77, v-component R=0.74),
with the model capturing some vertical variability of the horizontal flow CS 2, but not the depth and
strength of recirculation (TablesS7 and S8). This is emphasized in the stronger agreement in the cross-es-
tuary flows near the bed (u-component R= 0.81) compared to near the surface (u-component R=0.15).
Similar to TS Hermine, a cross-estuary recirculation emerges in the early stages of H Matthew (Figure12).
Later in the storm this pattern changes, with the along-estuary (v) surface flows traveling in an opposite
direction to near-bed flows at CS 2. Model agreement was good during the storm (u-component R=0.65,
v-component R= 0.84), with changes in flow direction well resolved; however, current magnitudes were
consistently underestimated by the model. Overall, velocities were approximately five times stronger in
the along-estuary direction than the cross-estuary direction in the observations and the simulations, and
higher during H Matthew than TS Hermine, a product of the long and narrow geometry of CS and the wind
direction during H Matthew.
5. Discussion
Spatial maps of water levels during the two storms in CS using the RAP wind field with precipitation are
shown in Figure13 at selected times during each storm, illustrating several key hydrodynamic patterns
that can occur in back-barrier estuaries during storm events. At the selected time step during TS Hermine,
strong cross-estuary winds drove water toward the southwest corners of both the northern and southern
reaches of CS. The minimal flow through the narrow center of the sound, evident by the buildup of water
on the north side of the narrow and shallow central area, emphasizes the importance of accurate, high res-
olution bathymetry, as suggested by Clunies etal.(2017) and Bennett etal.(2018). Varying wind directions
resulted in significant changes in water levels. As winds shifted later in the storm and water levels began
increasing in the southern reach due to flow from the northern reach (Figures4e and 4f). This water level
gradient also occurred during H Matthew, producing higher water levels at the southern reach of the sound.
REY ET AL. 18 of 25
10.1029/2020JC016483
Figure 13. Modeled water levels using RAP winds and precipitation in Currituck Sound on (a) September 3 at 14:00
UTC during TS Hermine and (b) October 12 at 12:00 UTC during H Matthew. Ocean results have been removed for
clarity, and observations are shown by colored circles on the same scale, and wind direction is indicated by Uw. Water
levels at the observation stations during the storm periods are shown in Figure9.
Journal of Geophysical Research: Oceans
Water levels also increased throughout the storm due to precipitation on the estuary surface (Figure9). Al-
though the relative wind direction has been shown to play a key role in back-barrier estuary hydrodynamics
in previous studies (Clunies etal.,2017; Garzon etal.,2018; Thomas etal.,2019), the detailed observations
in the present study, with five water level sensors distributed along and across the sound, provides substan-
tial validation of this process.
Surface waves have been previously shown to influence water levels and runup on shorelines in shallow
estuaries (Mulligan etal.,2015b), and in this investigation, significant wave heights showed considerable
spatial variation throughout CS (Figure14). Significant wave heights increased along the length of the es-
tuary (from CS 3 to CS 1), producing larger waves during H Matthew from the longer fetch along the longi-
tudinal axis of the sound generated by the northerly winds, compared to the northeast winds that occurred
in TS Hermine. Modeled wave heights are highly sensitive to wind direction due to this variability in fetch,
creating an environment that presents considerably challenges for spectral wave models. This aligns with
the finding from Long and Resio(2007), where the complex bathymetry and geometry of CS were observed
to produce very complex wave spectra, and supports their suggestion that third-generation spectral wave
models may not accurately simulate wave conditions in enclosed estuaries. Modeled wave directions were
influenced by wind direction, primarily traveling in the same direction as the prevailing wind.
Maps of the spatial distribution of horizontal velocity in CS are shown in Figure15 for the surface (a, b) and
near bed (c, d). Modeled velocities were predominately in the along-estuary direction, in general alignment
with the observations (Figure12). Despite similar peak wind speeds during the storms, stronger current
velocities occurred during H Matthew compared to TS Hermine, a result of the along-estuary wind direc-
tion earlier in the storm. This indicates the importance of fetch, and therefore wind direction as a driving
factor in back-barrier estuary hydrodynamics, as well as the sensitivity of these systems to variations in
atmospheric conditions. Despite the predominantly north-south direction of flow, the cross-estuary winds
during TS Hermine produced variability in the vertical profile of horizontal flow, with modeled velocity
generally coincident with the wind direction at the surface and recirculation at the bed, modified by local
bathymetry. This trend was also observed during H Matthew, although the effect was less pronounced.
REY ET AL. 19 of 25
10.1029/2020JC016483
Figure 14. Model results for significant wave height and direction using RAP winds and precipitation in Currituck
Sound on (a) September 3 at 14:00 UTC during TS Hermine and (b) October 12 at 12:00 UTC during H Matthew. Ocean
results have been removed for clarity, and observations are shown by colored circles on the same scale, and wind
direction is indicated by Uw.
Journal of Geophysical Research: Oceans
These complex hydrodynamic patterns have previously been observed in ponds with similar depths (Rey
etal.,2018; Sweeney etal.,2007), and represent an important aspect of estuarine hydrodynamics.
Coastal flooding is a major hazard that can be caused by tropical cyclones; however, the relative contribu-
tion of various storm processes (winds, waves, and precipitation) to flooding remains poorly understood in
back-barrier estuaries (Cyriac etal.,2018; Elko etal.,2019; Halstead,2018; Wahl etal.,2015), with most
research focusing on the flooding from offshore surge and swells (Gharagozlou etal.,2020). To investigate
REY ET AL. 20 of 25
10.1029/2020JC016483
Figure 15. Modeled velocities in Currituck Sound using RAP winds and precipitation at the (a, b) surface and (c, d)
bed on September 3 at 14:00 UTC during TS Hermine (a, c) and October 12 at 12:00 UTC during H Matthew (b, d).
Ocean results have been removed for clarity, and observations are shown by red vectors and colored circles on the same
scale, and wind direction is indicated by Uw.
Journal of Geophysical Research: Oceans
this, the model was run with various processes enabled or disabled, allowing the relative contribution to
flooding in CS to be evaluated. Statistics from these runs are included in TableS4. As waves were found to
have a minor (<1cm) impact on water levels, time-series results (Figures9, 16c, and 16d) and maps of the
flooded extent at selected time steps (Figures16a and 16b) are only shown for the wind and precipitation
processes, as well as the full model that includes all processes. Overall, this analysis emphasizes the im-
portance of winds on water levels, as none of the observed changes in water levels occurred in simulations
without winds, and excluding winds dramatically reduced the extent of flooding, aligning with results from
previous studies (Clunies etal., 2017; Garzon etal.,2018; Thomas etal., 2019). Wind driven flooding in-
creased and decreased over the storm period due to changing wind speeds and directions, in contrast to the
more consistent increase in water levels seen in the precipitation only simulations.
Considerably less research has been conducted into the impact of precipitation, despite increasing risks
due to climate change (Villarini etal.,2011). During H Matthew, the simulated flooded area increased
considerably when precipitation was included, inundating an additional 4km2 of the land surface in the CS
area. This includes inundating the frequently flooded coastal areas on the northwest corner of the estuary,
as well as urban areas on the east bank of CS, generally matching observations during the storm from Hal-
stead(2018) and NCRPP(2017). This effect was considerably more pronounced in H Matthew due to twice
the volume of rainfall compared to TS Hermine. Notably, the precipitation effect was entirely a product of
“Rain-on-grid” precipitation, which is only infrequently included in previous investigations (e.g., Orton
etal.,2012). Moreover, the increase in flooded area was nonlinear, and increased by more than the sum of
each process. The strong effects of precipitation on water levels are likely due to the relative isolation of CS
from the larger estuarine system. This illustrates the importance incorporating accurate spatial distribu-
tions of rainfall into coastal models, and the wide variations in precipitation patterns between models, with
REY ET AL. 21 of 25
10.1029/2020JC016483
Figure 16. Simulated flooding extent indicated by the shoreline positions for selected RAP processes in Currituck
Sound on (a) September 3 at 14:00 UTC during TS Hermine and (b) October 12 at 12:00 UTC during H Matthew. Inset
histograms show the total area in the shown domain that was flooded for each combination of processes at the same
time steps, while time series plots (c, d), show the change in flooded area for each process during over a 24-h period.
Ocean results have been removed for clarity.
415 420 425 430 435
UTM X (km)
3990
3995
4000
4005
4010
4015
4020
4025
4030
UTM Y (km)
TS. Hermine
03-Sep-2016 14:00
Uw
(a)
415420 425 430 435
UTM X (km)
H. Matthew
09-Oct-2016 08:00
Uw
(b)
Precp.
Winds
Full model
Coastline
Sensor platforms
0
5
10
15
Af (km2)
0
5
10
15
Af (km2)
00:00 06:00 12:00 18:00 00:00
Date Sep 03, 2016
0
5
10
15
Af (km2)
(c)
00:0006:00 12:00 18:00 00:00
Date Oct 09, 2016
0
5
10
15 (d)
Journal of Geophysical Research: Oceans
the average error between observed and modeled rainfall accumulations varying between 120mm in the
CFSv2 model and 35mm in the RAP model.
While the overall agreement for all parameters (winds, waves, and water levels) between model results and
observations are generally in good agreement when using winds from the RAP model with precipitation,
the accuracy is limited in some areas. Notably, modeled water levels did not fully capture the rapid changes
in water levels that were observed during both storms. Agreement at the northern CS 3 station was very
good; however, the model did not capture the magnitude of set-down in water levels that occurred during
TS Hermine in conjunction with a change in wind direction after the peak in wind speed. This also occurred
during H Matthew, where model results matched the water level trend, but not the peak elevations (Fig-
ure9). This effect is likely due to a combination of errors in the meteorological forcing data, bathymetry,
and/or the model's hydrodynamic accuracy in Pamlico Sound. Significant wave heights were generally ac-
curate, especially during TS Hermine, but lagged the very rapid increase in wave heights during H Matthew,
and modeled wave frequencies were overestimated. While the current magnitudes and directions were sim-
ulated effectively, the observed bidirectional vertical profile of horizontal flow did not occur in the model.
Due to the important implications resulting from this complex flow pattern (e.g., sediment transport), addi-
tional research is required to better understand this phenomenon.
6. Conclusion
As tropical storms increase in frequency and intensity, understanding the hydrodynamics and the processes
that drive coastal flooding during these storms is vital. To investigate this, the impacts of two tropical cy-
clones on CS, a back-barrier estuary that is part of the Outer Banks region of NC, were analyzed. A high-res-
olution coupled hydrodynamic-wave model (Delft3D-SWAN) was applied to simulate the hurricane-driven
processes in this long and narrow estuary. After testing eight atmospheric wind models and three precipita-
tion input fields, results from runs completed with RAP, ERA5, and CFSv2 were selected for further detailed
analysis. Overall results show that when RAP winds and precipitation were used as atmospheric inputs, the
Delft3D-SWAN model produced results that generally agreed with observations of waves, water levels, and
currents from five monitoring platforms in CS.
Despite both storms having a similar peak wind speed of ∼30m/s over CS, the surface waves, currents, and
water levels observed during TS Hermine and Hurricane Matthew had very different patterns at five moni-
toring platforms in the sound, related to the fetch and wind direction. During TS Hermine, a smooth varia-
tion in wind direction (from the east to from the north) resulted in rapidly varying water levels throughout
the sound, while the sudden change in wind direction during H Matthew produced higher water levels
across the sound. Significant wave heights were also influenced by wind direction, with higher significant
wave heights aligning with the strong along-estuary winds during H Matthew. The hydraulic complexity
of back-barrier estuary environments is indicated by the velocity measurements, with bidirectional flow
occurring during both storms throughout the estuary. Differences between observation and model results
are caused in part by the differences in the magnitude and direction between input wind fields and the true
wind conditions, with the long and narrow geometry of CS highly sensitive to variations in wind direction,
particularly for the along-estuary axis. Additional investigation is required in order to fully characterize the
primary drivers of this model error.
An analysis of the impacts of selected storm processes (i.e., wind, waves, and precipitation) emphasizes
the contribution from each process on hydrodynamic patterns. These results show the importance of wind
direction, as slight changes in direction relative to the orientation of the estuary resulted in very different
fetch and hydrodynamics. In addition to winds, precipitation directly on the water surface had an important
effect, raising water levels by ∼20cm during H Matthew and inundating an additional 4km2 on the land
surrounding CS, generally aligning with observations during the storm. While coastal flooding from precip-
itation during tropical storms is well documented in literature, relatively few studies have investigated the
effects from rainfall directly into an estuary. Although these findings are applicable in small estuarine en-
vironments that are relatively isolated from larger systems, results suggest that in these environments pre-
cipitation directly on the water surface can have a major impact on total water levels and coastal flooding.
Research should continue to investigate hydrodynamic patterns and storm processes using high-resolution
REY ET AL. 22 of 25
10.1029/2020JC016483
Journal of Geophysical Research: Oceans
numerical modeling in similar back-barrier estuary environments during other tropical storms. Additional
analysis could also be conducted to find techniques to further reduce the model error, with a particular fo-
cus on reproducing the very rapid changes in water levels. The results presented here suggest that although
model results are very sensitive to atmospheric forcing, Delft3D-SWAN can effectively model coastal con-
ditions during tropical storms, and could be applied in a forecast configuration to predict future conditions
in the coastal ocean.
Data Availability Statement
The data used in this study are archived in the ”Hydrodynamic Model Results for Hurricane Impacts in Coastal
North Carolina” Dataverse repository at Queen's University (https://dataverse.scholarsportal.info/dataverse/
CoastalModelling). The model results for Tropical Storm Hermine are accessible at https://doi.org/10.5683/
SP2/74K6EZ and the results for Hurricane Matthew are accessible at https://doi.org/10.5683/SP2/SUEBDN.
References
Bales, J. D. (2003). Effects of hurricane floyd inland flooding, September-October 1999, on tributaries to Pamlico Sound, North Carolina.
Estuaries, 26, 1319–1328. https://doi.org/10.1007/BF02803634
Bastidas, L. A., Knighton, J., & Kline, S. W. (2016). Parameter sensitivity and uncertainty analysis for a storm surge and wave model. Nat-
ural Hazards and Earth System Sciences, 16, 2195–2210. https://doi.org/10.5194/nhess-16-2195-2016
Benjamin, S. G., Weygandt, S. S., Brown, J. M., Hu, M., Alexander, C. R., Smirnova, T. G., etal. (2016). A North American hourly assimila-
tion and model forecast cycle: The rapid refresh. Monthly Weather Review, 144, 1669–1694. https://doi.org/10.1175/MWR-D-15-0242.1
Bennett, V. C., Mulligan, R. P., & Hapke, C. J. (2018). A numerical model investigation of the impacts of Hurricane Sandy on water level
variability in Great South Bay, New York. Continental Shelf Research, 161, 1–11. https://doi.org/10.1016/j.csr.2018.04.003
Berg, R. (2017). Hurricane Hermine (Technical Report No. AL092016, p. 63). Miami, FL: National Hurricane Center. Retrieved from
https://www.nhc.noaa.gov/data/tcr/AL092016iJHermine.pdf
Beudin, A., Ganju, N. K., Defne, Z., & Aretxabaleta, A. L. (2017). Physical response of a back-barrier estuary to a post-tropical cyclone.
Journal of Geophysical Research: Oceans, 122, 5888–5904. https://doi.org/10.1002/2016JC012344
Booij, N., Ris, R. C., & Holthuijsen, L. H. (1999). A third-generation wave model for coastal regions 1. Model description and validation.
Journal of Geophysical Research, 104, 7649–7666. https://doi.org/10.1029/98JC02622
Brown, M. M., Mulligan, R. P., & Miller, R. L. (2014). Modeling the transport of freshwater and dissolved organic carbon in the Neuse
River Estuary, NC, USA following Hurricane Irene (2011). Estuarine, Coastal and Shelf Science, 139, 148–158. https://doi.org/10.1016/j.
ecss.2014.01.005
Bunya, S., Dietrich, J. C., Westerink, J. J., Ebersole, B. A., Smith, J. M., Atkinson, J. H., etal. (2010). A high-resolution coupled riverine
flow, tide, wind, wind wave, and storm surge model for southern Louisiana and Mississippi. Part I: Model development and validation.
Monthly Weather Review, 138, 345–377. https://doi.org/10.1175/2009MWR2906.1
Caldwell, W. S. (2001). Hydrologie and salinity characteristics of Currituck Sound and selected tributaries in North Carolina and Virginia
1998-99 (Vol. 1, 01-4097). US Department of the Interior, US Geological Survey, Raleigh, North Carolina.
Chaumillon, E., Bertin, X., Fortunato, A. B., Bajo, M., Schneider, J. L., Dezileau, L., etal. (2017). Storm-induced marine flooding: Lessons
from a multidisciplinary approach. Earth-Science Reviews, 165, 151–184. https://doi.org/10.1016/j.earscirev.2016.12.005
Clunies, G. J., Mulligan, R. P., Mallinson, D. J., & Walsh, J. P. (2017). Modeling hydrodynamics of large lagoons: Insights from the Albe-
marle-Pamlico Estuarine System. Estuarine, Coastal and Shelf Science, 189, 90–103. https://doi.org/10.1016/j.ecss.2017.03.012
Conery, I., Walsh, J. P., & Reide Corbett, D. (2018). Hurricane overwash and decadal-scale evolution of a narrowing barrier island, Ocra-
coke Island, NC. Estuaries and Coasts, 41, 1626–1642. https://doi.org/10.1007/s12237-018-0374-y
Corbett, D. R. (2010). Resuspension and estuarine nutrient cycling: Insights from the Neuse River Estuary. Biogeosciences, 7, 3289–3300.
https://doi.org/10.5194/bg-7-3289-2010
Cyriac, R., Dietrich, J. C., Fleming, J. G., Blanton, B. O., Kaiser, C., Dawson, C. N., etal. (2018). Variability in Coastal Flooding predictions
due to forecast errors during Hurricane Arthur. Coastal Engineering, 137, 59–78. https://doi.org/10.1016/j.coastaleng.2018.02.008
Defne, Z., Ganju, N. K., & Moriarty, J. M. (2019). Hydrodynamic and morphologic response of a back-barrier estuary to an extratropical
storm. Journal of Geophysical Research: Oceans, 124, 7700–7717. https://doi.org/10.1029/2019JC015238
Dietrich, J. C., Muhammad, A., Curcic, M., Fathi, A., Dawson, C. N., Chen, S. S., etal. (2018). Sensitivity of storm surge predictions
to atmospheric forcing during Hurricane Isaac. Journal of Waterway, Port, Coastal, and Ocean Engineering, 144, 1–24. https://doi.
org/10.1061/(ASCE)WW.1943-5460.0000419
Dresback, K. M., Fleming, J. G., Blanton, B. O., Kaiser, C., Gourley, J. J., Tromble, E. M., etal. (2013). Skill assessment of a real-time forecast
system utilizing a coupled hydrologic and coastal hydrodynamic model during Hurricane Irene (2011). Continental Shelf Research, 71,
78–94. https://doi.org/10.1016/j.csr.2013.10.007
Elko, N., Dietrich, J., Cialone, M., Stockdon, H., Bilskie, M., Boyd, B., etal. (2019). Advancing the understanding of storm processes and
impacts. Shore and Beach, 87, 41–55.
Elvidge, S., Angling, M. J., & Nava, B. (2014). On the use of modified Taylor diagrams to compare ionospheric assimilation models. 2014
XXXIth URSI General Assembly and Scientific Symposium (URSI GASS) (pp. 1–4). Beijing, China: IEEE. https://doi.org/10.1109/
URSIGASS.2014.6929835
Field, C., Barros, V., Dokken, D., Mach, K., Mastrandrea, M., Bilir, T. et al. (Eds.). (2014). Climate change 2014: Impacts, adaptation, and
vulnerability. Part A: Global and sectoral aspects. Contribution of working group II to the fifth assessment report of the intergovernmental
panel on climate change. Cambridge, UK and New York, NY: Cambridge University Press.
REY ET AL. 23 of 25
10.1029/2020JC016483
Acknowledgments
The authors thank Heidi Wadman,
Spicer Bak, Pat Dickhudt, and Ian
Conery at the USACE Field Research
Facility. Research funding for this
project was provided by the Queen's
University Engineering Dean's
Graduate Research Award held by A.
Rey. R. P. Mulligan also acknowledges
support from the Natural Science and
Engineering Research Council of Can-
ada Discovery Grant Program under
award number RGPIN/04043-2018. D.
R. Corbett would like to acknowledge
funding from NCDoT that provided
additional bathymetry information
(Agreement 2018-05). Computational
support was provided by SHARCNET
(www.sharcnet.ca), Compute Canada
(http://computecanada.ca), Cory Wyatt
at Queen's University, and Maria Aristi-
zabal Vargas at Rutgers University.
Journal of Geophysical Research: Oceans
Fine, J. M. (2008). Documentation of data collection in Currituck Sound, North Carolina and Virginia, 2006–2007 (U.S. Geological Survey
Open-File Report 2008-1147, 11 p). Retrieved from http://pubs.water.usgs.gov/ofr2008-1147/
Fritz, H. M., Blount, C., Sokoloski, R., Singleton, J., Fuggle, A., McAdoo, B. G., etal. (2007). Hurricane Katrina storm surge distribution
and field observations on the Mississippi Barrier Islands. Estuarine, Coastal and Shelf Science, 74, 12–20. https://doi.org/10.1016/j.
ecss.2007.03.015
Garzon, J. L., Ferreira, C. M., & Padilla-Hernandez, R. (2018). Evaluation of weather forecast systems for storm surge modeling in the
Chesapeake Bay. Ocean Dynamics, 68, 91–107. https://doi.org/10.1007/s10236-017-1120-x
Gharagozlou, A., Dietrich, J. C., Karanci, A., Luettich, R. A., & Overton, M. F. (2020). Storm-driven erosion and inundation of barrier
islands from dune-to region-scales. Coastal Engineering, 158, 103674. https://doi.org/10.1016/j.coastaleng.2020.103674
Giffin, D., & Corbett, D. R. (2003). Evaluation of sediment dynamics in coastal systems via short-lived radioisotopes. Journal of Marine
Systems, 42, 83–96. https://doi.org/10.1016/S0924-7963(03)00068-X
Grinsted, A., Ditlevsen, P., & Christensen, J. H. (2019). Normalized US hurricane damage estimates using area of total destruction 1900-
2018. Proceedings of the National Academy of Sciences, 116, 23942–23946. https://doi.org/10.1073/pnas.1912277116
Halstead, A. (2018). Assessing and communicating flood risk in Currituck County, North Carolina. Master of environmental management.
Durham, NC: Duke University.
Hersbach, H., Bell, B., Berrisford, P., Horányi, A., Sabater, J. M., Nicolas, J., et al. 2019. Global reanalysis: Goodbye ERA-Interim, hello
ERA5. ECMWF Newsletter, 159, 17–24. https://doi.org/10.21957/vf291hehd7
Holland, G. J. (1980). An analytic model of the wind and pressure profiles in hurricanes. Monthly Weather Review, 108, 1212–1218. https://
doi.org/10.1175/1520-0493(1980)108(1212:AAM0TW)2.0.C0;2
Hsu, M. H., Kuo, A. Y., Kuo, J. T., & Liu, W. C. (1999). Procedure to calibrate and verify numerical models of estuarine hydrodynamics.
Journal of Hydraulic Engineering, 125, 166–182. https://doi.org/10.1061/(ASCE)0733-9429(1999)125:2(166)
Jisan, M. A., Bao, S., Pietrafesa, L. J., Shen, D., Gayes, P. T., & Hallstrom, J. (2018). Hurricane Matthew (2016) and its impact under global
warming scenarios. Modeling Earth Systems and Environment, 4, 97–109. https://doi.org/10.1007/s40808-018-0420-6
Klein, A. (2016). Hurricane hermine. New Scientist, 231, 4.
Knutson, T. R., McBride, J. L., Chan, J., Emanuel, K., Holland, G., Landsea, C., etal. (2010). Tropical cyclones and climate change. Nature
Geoscience, 3, 157–163. https://doi.org/10.1038/ngeo779
Kumbier, K., Carvalho, R. C., Vafeidis, A. T., & Woodroffe, C. D. (2018). Investigating compound flooding in an estuary using hydrodynam-
ic modelling: A case study from the Shoalhaven River, Australia. Natural Hazards and Earth System Sciences, 18, 463–477. https://doi.
org/10.5194/nhess-18-463-2018
Landsea, C. W., & Franklin, J. L. (2013). Atlantic hurricane database uncertainty and presentation of a new database format. Monthly
Weather Review, 141(10), 3576–3592. https://doi.org/10.1175/MWR-D-12-00254.1
Lee, C., Hwang, S., Do, K., & Son, S. (2019). Increasing flood risk due to river runoff in the estuarine area during a storm landfall. Estuarine,
Coastal and Shelf Science, 221, 104–118. https://doi.org/10.1016/j.ecss.2019.03.021
Lesser, G. R., Roelvink, J. A., van Kester, J. A., & Stelling, G. S. (2004). Development and validation of a three-dimensional morphological
model. Coastal Engineering, 51, 883–915. https://doi.org/10.1016/j.coastaleng.2004.07.014
Lin, N., Smith, J. A., Villarini, G., Marchok, T. P., & Baeck, M. L. (2010). Modeling extreme rainfall, winds, and surge from Hurricane Isabel
(2003). Weather and Forecasting, 25, 1342–1361. https://doi.org/10.1175/2010WAF2222349.1
Long, C. E., & Resio, D. T. (2007). Wind wave spectral observations in Currituck Sound, North Carolina. Journal of Geophysical Research,
112, 1–21. https://doi.org/10.1029/2006JC003835
McGranahan, G., Balk, D., & Anderson, B. (2007). The rising tide: Assessing the risks of climate change and human settlements in low
elevation coastal zones. Environment and Urbanization, 19, 17–37. https://doi.org/10.1177/0956247807076960
Miselis, J. L., Andrews, B. D., Nicholson, R. S., Defne, Z., Ganju, N. K., & Navoy, A. (2016). Evolution of mid-Atlantic coastal and back-bar-
rier estuary environments in response to a hurricane: Implications for barrier-estuary connectivity. Estuaries and Coasts, 39, 916–934.
https://doi.org/10.1007/s12237-015-0057-x
Moran, K. L., Mallinson, D. J., Culver, S. J., Leorri, E., & Mulligan, R. P. (2015). Late Holocene evolution of Currituck Sound, North Car-
olina, USA: Environmental change driven by sea-level rise, storms, and barrier island morphology. Journal of Coastal Research, 314,
827–841. https://doi.org/10.2112/JC0ASTRES-D-14-00069.1
Mulligan, R. P., Bowen, A. J., Hay, A. E., van der Westhuysen, A. J., & Battjes, J. A. (2008). Whitecapping and wave field evolution in a
coastal bay. Journal of Geophysical Research, 113, C03008. https://doi.org/10.1029/2007JC004382
Mulligan, R. P., Mallinson, D. J., Clunies, G. J., Rey, A., Culver, S. J., Zaremba, N., etal. (2019). Estuarine responses to long-term changes in
inlets, morphology, and sea level rise. Journal of Geophysical Research: Oceans, 124, 9235–9257. https://doi.org/10.1029/2018JC014732
Mulligan, R. P., Mallinson, D. J., Clunies, G. J., Zaremba, N., Leorri, E., Culver, S. J., et al. (2015a). Sediment dynamics and morphological
evolution of a large back-barrier estuary. In C. Jun & W. Ping (Eds.), The Proceedings of the Coastal Sediments 2015 (pp. 1–8). San Diego,
CA: World Scientific. https://doi.org/10.1142/9789814689977_0205
Mulligan, R. P., Walsh, J. P., & Wadman, H. M. (2015b). Storm surge and surface waves in a shallow lagoonal estuary during the crossing of a
hurricane. Journal of Waterway, Port, Coastal, and Ocean Engineering, 141, 1–11. https://doi.org/10.1061/(ASCE)WW.1943-5460.0000260
NCRPP (2017). Hurricane Matthew Resilient Redevelopment Plan Dare County, Manteo, NC: North Carolina Resilient Redevelopment
Planning.
NOAA National Centers for Environmental Information (NCEI) (2020). U.S. Coastal Relief Model. Boulder, Colorado: Retrieved from
http://www.ngdc.noaa.gov/mgg/coastal/crm.html
Orton, P., Georgas, N., Blumberg, A., & Pullen, J. (2012). Detailed modeling of recent severe storm tides in estuaries of the New York City
region. Journal of Geophysical Research, 117, 1–17. https://doi.org/10.1029/2012JC008220
Paerl, H. W., Valdes, L. M., Joyner, A. R., Peierls, B. L., Piehler, M. F., Riggs, S. R., etal. (2006). Ecological response to hurricane events in the
Pamlico Sound system, North Carolina, and implications for assessment and management in a regime of increased frequency. Estuaries
and Coasts, 29, 1033–1045. https://doi.org/10.1007/BF02798666
Paramygin, V., Sheng, Y., & Davis, J. (2017). Towards the development of an operational forecast system for the Florida coast. Journal of
Marine Science and Engineering, 5, 8. https://doi.org/10.3390/jmse5010008
Peng, M., Xie, L., & Pietrafesa, L. J. (2004). A numerical study of storm surge and inundation in the Croatan-Albemarle-Pamlico Estuary
System. Estuarine, Coastal and Shelf Science, 59, 121–137. https://doi.org/10.1016/j.ecss.2003.07.010
Rey, A., Mulligan, R. P., Boegman, L., Filion, Y., da Silva, A., & Champagne, P. (2018). Impact of control structures on hydraulic retention
time in wastewater stabilization ponds. In Y. Filion & M. Hulley (Eds.), 1st International WDSA/CCWI 2018 Joint Conference (p. 7).
Kingston, ON.
REY ET AL. 24 of 25
10.1029/2020JC016483
Journal of Geophysical Research: Oceans
Rey, A., Mulligan, R. P., Corbett, R. D., & Wadman, H. M. (2019). Numerical modelling of storm-driven sediment transport in Currituck
Sound, NC. In J. D. Rosati, P. Wang, & M. Vallee (Eds.), Coastal sediments 2019 (pp. 1708–1719). St. Petersburg, FL: World Scientific.
https://doi.org/10.1142/9789811204487_0147
Risser, M. D., & Wehner, M. F. (2017). Attributable human-induced changes in the likelihood and magnitude of the observed extreme
precipitation during hurricane Harvey. Geophysical Research Letters, 44, 12457–12464. https://doi.org/10.1002/2017GL075888
Robinson, M. M., & McBride, R. A. (2006). Benthic foraminifera from a relict flood tidal delta along the Virginia/North Carolina Outer
Banks. Micropaleontology, 52, 67–80. https://doi.org/10.2113/gsmicropal.52.1.67
Saha, S., Moorthi, S., Pan, H. L., Wu, X., Wang, J., Nadiga, S., etal. (2010). The NCEP climate forecast system reanalysis. Bulletin of the
American Meteorological Society, 91, 1015–1058. https://doi.org/10.1175/2010BAMS3001.1
Smallegan, S. M., Irish, J. L., & van Dongeren, A. R. (2017). Developed barrier island adaptation strategies to hurricane forcing under rising
sea levels. Climatic Change, 143, 173–184. https://doi.org/10.1007/s10584-017-1988-y
Smith, C. S., & Scyphers, S. (2019). Past hurricane damage and flood zone outweigh shoreline hardening for predicting residential-scale
impacts of Hurricane Matthew. Environmental Science and Policy, 101, 46–53. https://doi.org/10.1016/j.envsci.2019.07.009
Stewart, S. R. (2017). Hurricane Matthew (Technical Report No. AL142016, p. 96). Miami, FL: National Hurricane Center. Retrieved from
http://www.nhc.noaa.gov/data/tcr/AL142016iJMatthew.pdf
Sweeney, D. G., Nixon, J. B., Cromar, N. J., & Fallowfield, H. J. (2007). Temporal and spatial variation of physical, biological, and chemical
parameters in a large waste stabilisation pond, and the implications for WSP modelling. Water Science and Technology, 55, 1–9. https://
doi.org/10.2166/wst.2007.336
Taylor, K. E. (2001). Summarizing multiple aspects of model performance in a single diagram. Journal of Geophysical Research, 106,
7183–7192. https://doi.org/10.1029/2000JD900719
Thomas, A., Dietrich, J. C., Asher, T. G., Bell, M., Blanton, B. O., Copeland, J. H., etal. (2019). Influence of storm timing and forward speed
on tides and storm surge during Hurricane Matthew. Ocean Modelling, 137, 1–19. https://doi.org/10.1016/j.ocemod.2019.03.004
van der Westhuysen, A. J., Zijlema, M., & Battjes, J. A. (2007). Nonlinear saturation-based whitecapping dissipation in SWAN for deep and
shallow water. Coastal Engineering, 54, 151–170. https://doi.org/10.1016/j.coastaleng.2006.08.006
Vidal-Juarez, T., de Alegría-Arzaburu, A. R., Mejía-Trejo, A., García-Nava, H., & Enriquez, C. (2014). Predicting barrier beach breaching
due to extreme water levels at San Quntín, Baja California, Mexico. Journal of Coastal Research, 71, 100–106. https://doi.org/10.2112/
si71-012.1
Villarini, G., Smith, J. A., Baeck, M. L., Marchok, T., & Vecchi, G. A. (2011). Characterization of rainfall distribution and flooding associated
with U.S. landfalling tropical cyclones: Analyses of Hurricanes Frances, Ivan, and Jeanne (2004). Journal of Geophysical Research, 116.
https://doi.org/10.1029/2011JD016175
Wagner, C., Fitzgerald, S., & Antolino, D. (2016). Characterization of water-quality and bed-sediment conditions in Currituck Sound,
North Carolina, prior to the Mid-Currituck Bridge construction, 2011‒15 (ver. 1.1, July 2016) (U.S. Geological Survey Open-File Report
2015‒1208, 84 p.). https://doi.org/10.3133/ofr20151208
Wahl, T., Jain, S., Bender, J., Meyers, S. D., & Luther, M. E. (2015). Increasing risk of compound flooding from storm surge and rainfall for
major US cities. Nature Climate Change, 5, 1093–1097. https://doi.org/10.1038/nclimate2736
Wang, A., Issa, A., Bayleyegn, T., Noe, R. S., Mullarkey, C., Casani, J., etal. (2017). Notes from the field: Mortality Associated with Hur-
ricane Matthew – United States, October 2016. MMWR. Morbidity and Mortality Weekly Report, 66, 145–146. https://doi.org/10.15585/
mmwr.mm6605a3
Wells, J. T., & Kim, S. Y. (1989). Sedimentation in the Albemarle-Pamlico lagoonal system: Synthesis and hypotheses. Marine Geology, 88,
263–284. https://doi.org/10.1016/0025-3227(89)90101-1
Yelland, M., & Taylor, P. K. (1996). Wind stress measurements from the open ocean. Journal of Physical Oceanography, 26(4), 541–558.
https://doi.org/10.1175/1520-0485(1996)026(0541:WSMFTO)2.0.CO;2
REY ET AL. 25 of 25
10.1029/2020JC016483
