Hindcast and validation of Hurricane Ike (2008) waves ... · 1] Hurricane Ike (2008) made landfall near Galveston, Texas, as a moderate intensity storm. Its large wind ﬁeld in conjunction

Hindcast and validation of Hurricane Ike (2008) waves forerunnerand storm surge

M E Hope1 J J Westerink1 A B Kennedy1 P C Kerr1 J C Dietrich12 C Dawson3

C J Bender4 J M Smith5 R E Jensen5 M Zijlema6 L H Holthuijsen6 R A Luettich Jr7

M D Powell8 V J Cardone9 A T Cox9 H Pourtaheri10 H J Roberts11 J H Atkinson11

S Tanaka112 H J Westerink1 and L G Westerink1

Received 26 February 2013 revised 9 July 2013 accepted 12 July 2013 published 13 September 2013

[1] Hurricane Ike (2008) made landfall near Galveston Texas as a moderate intensity stormIts large wind field in conjunction with the Louisiana-Texas coastlinersquos broad shelf and largescale concave geometry generated waves and surge that impacted over 1000 km of coastlineIkersquos complex and varied wave and surge response physics included the capture of surge bythe protruding Mississippi River Delta the strong influence of wave radiation stress gradientson the Delta adjacent to the shelf break the development of strong wind driven shore-parallelcurrents and the associated geostrophic setup the forced early rise of water in coastal baysand lakes facilitating inland surge penetration the propagation of a free wave along thesouthern Texas shelf shore-normal peak wind-driven surge and resonant and reflected longwaves across a wide continental shelf Preexisting and rapidly deployed instrumentationprovided the most comprehensive hurricane response data of any previous hurricane Morethan 94 wave parameter time histories 523 water level time histories and 206 high watermarks were collected throughout the Gulf in deep water along the nearshore and up to 65 kminland Ikersquos highly varied physics were simulated using SWANthornADCIRC a tightlycoupled wave and circulation model on SL18TX33 a new unstructured mesh of the Gulf ofMexico Caribbean Sea and western Atlantic Ocean with high resolution of the Gulfrsquos coastalfloodplain from Alabama to the Texas-Mexico border A comprehensive validation was madeof the modelrsquos ability to capture the varied physics in the system

Citation Hope M E et al (2013) Hindcast and validation of Hurricane Ike (2008) waves forerunner and storm surge J GeophysRes Oceans 118 4424ndash4460 doi101002jgrc20314

1 Introduction

[2] The Louisiana and Texas (LATEX) Gulf Coast is sit-uated in an area of high tropical storm activity Major hurri-canes making landfall along the LATEX coast includestorms in 1886 (unnamed landfall at Indianola TX) 1900(unnamed landfall at Galveston TX) 1915 (unnamedlandfall at Galveston TX) 1961 (Carla) 1965 (Betsy)1969 (Camille) 1980 (Allen) 1983 (Alicia) 2005 (Katrinaand Rita) 2008 (Gustav and Ike) and 2012 (Isaac) Hurri-cane Ike is of significant interest because of its size its var-ied response physics and the quantity and quality of waveand water level data collected

[3] Hurricane Ike entered the Gulf of Mexico after makinglandfall in Cuba Upon entering the Gulf at 2030 UTC 9 Sep-tember 2008 (Table 1) Ike tracked northwest and its windfield broadened and strengthened until reaching a 10 min sus-tained wind speed of 37 m s1 and radius to maximum windsof 148 km at 0000 UTC 12 September 2008 (31 h beforelandfall) when the stormrsquos center was approximately 300 kmsouth of Isles Dernieres LA (Figure 1 Table 2) with tropicalstorm force winds extending 400 km from the stormrsquos centerAt this point significant wave heights were measured at over8 m in the mid-Gulf 6 m to the south of Grand Isle LAand 4 m off of Galveston Island Approximately 13 h before

1Department of Civil and Environmental Engineering and Earth Scien-ces University of Notre Dame Notre Dame Indiana USA

2Now at Department of Civil Construction and Environmental Engi-neering North Carolina State University Raleigh North Carolina USA

3Institute for Computational Engineering and Sciences University ofTexas at Austin Austin Texas USA

4Taylor Engineering Jacksonville Florida USA5Coastal and Hydraulics Laboratory US Army Engineer Research and

Development Center Vicksburg Mississippi USA6Faculty of Civil Engineering and Geosciences Delft University of

Technology Delft Netherlands7Institute of Marine Sciences University of North Carolina at Chapel

Hill Chapel Hill North Carolina USA8Atlantic Oceanographic and Meteorological Labs Hurricane Research

Division NOAA Miami Florida USA9Oceanweather Inc Cos Cob Connecticut USA10New Orleans District US Army Corps of Engineers New Orleans

Louisiana USA11ARCADIS Boulder Colarado USA12Now at Earthquake Research Institute University of Tokyo Tokyo

Corresponding author M E Hope Department of Civil and Environ-mental Engineering and Earth Sciences University of Notre Dame 156Fitzpatrick Hall Notre Dame IN 46556-5637 (markehopegmailcom)

copy2013 American Geophysical Union All Rights Reserved2169-927513101002jgrc20314

JOURNAL OF GEOPHYSICAL RESEARCH OCEANS VOL 118 4424ndash4460 doi101002jgrc20314 2013

landfall Ike began to shift and track north-northwestwardthen making landfall at Galveston Island TX with a maxi-mum wind speed of 41 m s1 Ike generated a maximum

measured surge at landfall of 53 m in Chambers CountyTX located to the northeast of Galveston Island (Figure 1)[FEMA 2008] Across the LATEX coast Ike produced surge

Table 1 Summary of Significant Times and Characteristics of Hurricane Ikea

HoursRelativeto Landfall

UTCTime

UTC Date(2008) Latitude Longitude

Max WindVelocity

Radius toMaximum

Winds (km)

MinimumCentral

Pressure (mb)Saffir-Simpson

Category Notes

289 0600 1 Sep 172 37 13 167 1006 Trop Depression Formation217 0600 4 Sep 224 550 54 28 935 4 Maximum Intensity1945 0430 5 Sep 236 604 50 28 945 4 Enters SL18thornTX33

Domain187 1200 5 Sep 234 620 46 28 954 3 OWI winds start138 1300 7 Sep 210 732 49 947 3 Landfall on Great Inagua

Island Bahamas12475 0215 8 Sep 211 757 50 945 4 Landfall in Holguin Cuba89 1400 9 Sep 226 829 30 - 965 1 Landfall in Pinar del

Rio Cuba825 2030 9 Sep Enters Gulf of Mexico31 0000 12 Sep 261 900 37 148 954 219 1200 12 Sep 269 922 39 93 954 2 Peak in South Plaquemines13 1800 12 Sep 274 930 39 93 955 2 Shift in track WSE peak

in NOLA7 0000 13 Sep 283 941 41 74 952 2 WSE peak in Lake

Pontchartrain0 0700 13 Sep 293 947 41 950 2 Landfall at Galveston

Texas5 1200 13 Sep 303 952 37 56 959 111 1800 13 Sep 317 953 22 74 974 Trop Storm23 0600 14 Sep 355 937 15 93 986 Trop Depression OWI winds end53 1200 15 Sep End of simulation

aWinds are 10 min average [Berg 2009] (Automated Tropical Cyclone Forecast Archive ftpftpnhcnoaagovatcf)

Figure 1 Map of Northern Gulf of Mexico and Louisiana-Texas Coast The black line represents Ikersquostrack ADCIRC grid boundaries and raised features are brown the coastline is solid gray bathymetriccontours (as labeled) are dashed gray Geographic locations of significance are labeled by numbersidentified in Table 2

HOPE ET AL HINDCAST AND VALIDATION OF HURRICANE IKE

levels of 18 m in Lake Pontchartrain 22 m in Lake Borgne18 m at Grand Isle 30 m near Vermillion Bay LA 45 m atthe Sabine Lake Gulf Outlet 33 m at Galveston Island TXand 15 m at Corpus Christi TX

[4] Following Hurricanes Katrina and Rita wave andwater level gages were strengthened to become more reli-

able under hurricane conditions Additionally the use ofshort-term deployable gages placed prestorm nearshore andinland increased the density of recorded data across thecoast As a result of these efforts the number density andextent of wave and water level gages that collected datathroughout the storm surpassed that of any previous storm

[5] The wave measurements describe generation in deepwater transformation nearshore and dissipation onshoreand are summarized in Table 3 NOAA National DataBuoy Center (NDBC httpwwwndbcnoaagov) wavedata at 13 stations includes offshore buoys on the continen-tal shelf as well as in the deep Gulf Louisiana State Uni-versityrsquos Coastal Studies Institute (CSI httpwwwcsilsuedu) recorded wave data at five nearshoregages off the coast of Southern Louisiana Andrew Ken-nedy (AK) from the University of Notre Dame deployedeight gages via helicopter off the Texas coast from SabineLake to San Antonio Bay in depths ranging from 85 to 16m [Kennedy et al 2012] the US Army Corps of Engi-neers Research and Development Center Coastal Hydraul-ics Laboratory (USACE-CHL) deployed six gages in theTerrebonne and Biloxi marshes that were placed to under-stand the dissipation of waves over wetlands

[6] Water level time series Table 3 were collectedthroughout the LATEX shelf and adjacent floodplain by theUS Army Corps of Engineers (USACE) the USACE-CHLthe National Oceanic and Atmospheric Organization(NOAA) the US Geological Survey (USGS) the coopera-tive USGS and State of Louisiana Coastwide ReferenceMonitoring System (CRMS) CSI the Texas Coastal OceanObservation Network (TCOON) and AK Time history dataat these 523 stations describe in detail the development andevolution of surge on the LATEX shelf and its subsequentinland penetration High water marks (HWMs) were col-lected for the Federal Emergency Management Agency(FEMA) following the storm Of the available HWMs dataat 206 locations were deemed as reliable indicators of still-

Table 2 Geographic Locations by Type and Location

River and Channels

1 Mississippi River Birdrsquos foot2 Mississippi River Gulf Outlet (MRGO)3 Inner Harbor Navigation Canal (IHNC)4 Gulf Intracoastal Waterway (GIWW)Water Bodies5 Chandeleur Sound6 Lake Borgne7 Lake Pontchartrain8 Lake Maurepas9 Barataria Bay10 Terrebonne Bay11 Vermillion Bay12 Calcasieu Lake13 Sabine Lake14 Galveston Bay15 Corpus Christi BayLocations16 Chandeleur Islands17 Biloxi Marsh18 Caernarvon Marsh19 Plaquemines Parish LA20 New Orleans21 Terrebonne Marsh22 Grand Isle23 Isles Dernieres LA24 Chambers County TX25 Bolivar Peninsula26 Galveston Island27 Houston TX

Table 3 Summary of Collected Dataa

Data Type

Data SourceWaterLevels

SignificantWave Height

Mean WaveDirection

Mean WavePeriod

Peak WavePeriod

HighWater Mark Winds Currents

NDBC 13 9 13 13CSI 5 5 5 5 5 2 2AK 8 8 8 8USACE-CHL 6 5 5 5NOAA 37 29 2USACE 38 33USGS-PERM 33 24USGS-DEPL 50 40TCOON 25 17 4CRMS 321 235TABS 4FEMA 206

aData sources are as follows NDBC National Data Buoy Center (httpwwwndbcnoaagov) CSI Louisiana State University Coastal Studies Insti-tute (httpwwwcsilsuedu) AK University of Notre Dame Andrew Kennedy [Kennedy et al 2011a] USACE-CHL US Army Corps of EngineersCoastal Hydraulics Laboratory (J Smith personal communication 2009) NOAA National Oceanic and Atmospheric Administration (httptidesand-currentsnoaagov) USACE US Army Corps of Engineers (http wwwrivergagescom personal communication 2011) USGS-PERM US Geo-logical Survey (D Walters personal communication 2009) (httppubsusgsgovof20081365) USGS-DEPL US Geological Survey [East et al2008] TCOON Texas Coastal Ocean Observation Network (httplighthousetamucceduTCOON) CRMS Coastwide Reference Monitoring System(httpwwwlacoastgovcrms2) TABS Texas Automated Buoy System (httptabsgergtamuedu) FEMA Federal Emergency Management Agency[FEMA 2008 2009]

water elevations and resulting solely from Ike An additional393 water level time histories were identified as recordingreliable still water high water levels All water levels are ref-erenced to the North American Vertical Datum of 1988(NAVD88 200465 epoch in Louisiana)

[7] Wind data were used from four NOAA and twoTCOON stations along the LATEX coast and current datawere used from two CSI and four Texas Automated BuoySystem (TABS httptabsgergtamuedu) stations on thecontinental shelf

[8] The measurement data provide a comprehensivedescription of Ikersquos waves and storm surge Ikersquos expansivewave fields with maximum measured significant waveheights reaching 10 m in the deep Gulf were dominated bylocally generated seas and well-defined swells that reachedshore prior to the storm making landfall Effective attenua-tion occurred on the continental shelf in the nearshore andespecially behind barrier islands and within wetlands

[9] Storm surge was dictated by geography bathymetryand storm track and included a variety of fundamentallydifferent physical processes Steady easterly winds acrossthe Mississippi Sound and over the Biloxi and CaernarvonMarshes persisted as Ike was progressing across the Gulf ofMexico This resulted in the effective capture of surge bythe protruding Mississippi River Delta and river systemwhich projects onto the continental shelf This slow processlasted 2 days created a surge of 15ndash25 m in Lake Borgneand at the convergence of the Mississippi River Gulf Outlet(MRGO) and Gulf Intracoastal Waterway (GIWW) Fromthis point surge flowed into the Inner Harbor NavigationCanal (IHNC) into the heart of New Orleans peaking 13 hbefore landfall with a maximum water level of 25 m Thisregional surge also drove water into Lakes Pontchartrainand Maurepas to the north of New Orleans through theRigolets Chef Menteur Pass and Pass Manchac where 18m of surge was observed within Lake Pontchartrain peak-ing 7 h before landfall The same process occurred to thesouth and east of New Orleans in the marshes and wetlandsof Plaquemines Parish Water from Chandeleur Sound waspushed into the Caernarvon Marsh reaching 3 m at EnglishTurn A 2 m surge was pushed from Breton Sound againstthe protruding west bank Mississippi River levee south ofPoint-a-la-Hache where there is no corresponding levee onthe east bank [Kerr et al 2013a 2013b] peaking approxi-mately 19 h before landfall Having penetrated the riverthis surge propagated upstream The south and west facingportions of the lsquolsquoBirdrsquos Footrsquorsquo developed surge influencedby wave radiation stress gradient induced setup and moder-ate shore normal winds and reached uniform levels of12 m

[10] The region from the Atchafalaya and VermillionBays to Galveston Bay was influenced by a geostrophicallydriven surge forerunner and by shore-perpendicular wind-driven surge Water levels along this coast reached 2ndash25 mmore than 12 h prior to landfall while winds were still pre-dominantly shore parallel or directed offshore Factors con-trolling this Coriolis-driven early setup included the wideLATEX shelf with its smooth muddy bottom Ikersquos largesize and steady northwest track and the concave shape ofthe coast being coincident with the shore parallel winds[Buczkowski et al 2006 Kennedy et al 2011a 2011b]The time scale associated with the forerunner allowed

surge to penetrate far inland into hydraulically connectedwater bodies and adjacent low lying coastal floodplainsFor example Morganrsquos Point within Galveston Bay andManchester Point in the Houston Ship Channel experiencedwater levels of up to 2 m more than 12 h before landfall

[11] The coastal forerunner propagated as a free conti-nental shelf wave from Galveston TX southward on theLATEX shelf reaching Corpus Christi TX with an ampli-tude of 15 m The time of arrival of the continental shelfwave at Corpus Christi approximately 300 km southwestof Galveston coincided with the landfall of the storm atGalveston This was the largest measured continental shelfwave ever reported in the literature [Kennedy et al 2011a2011b]

[12] The region between the Atchafalaya and VermillionBays and Galveston Bay also experienced a peak surgecoincident to peak shore-normal winds ranging from 3 madjacent to the Atchafalaya Bay to 5 m to the west of Sab-ine Lake and to 35 m near Galveston TX Theforerunner-driven higher water levels within GalvestonBay persisted through the arrival of the strong winds atlandfall combining the forerunner and the wind-drivensurge levels within and around the bay

[13] As the storm passed and winds subsided the coastalsurge receded back onto the shelf The abrupt bathymetricchange at the continental shelf break led to an out-of-phasereflection of the surge back onto the shelf The recordshows a cross shelf wave appearing at the coast three timeswith increased damping with each cycle The cross shelfwave has a period of approximately 12 h coinciding withthe resonant period of the shelf The resonant period of theshelf can also be seen in the strong amplification of semi-diurnal tides on the wide portion of the LATEX shelf cen-tered at Lakes Sabine and Calcasieu [Mukai et al 2002]

[14] The scale and complexity of the Gulf coastal fea-tures on the LATEX shelf and the inland floodplain requirethe use of computational models that are basin-scale multi-process and provide a high level of resolution in manyareas A coupled nonphase resolving wave and circulationmodel was used to simulate the waves riverine drivenflows tides and the wave-driven wind-driven andpressure-driven circulation during Ike SWANthornADCIRCis a tightly coupled modeling system that operates on anunstructured mesh allowing for interaction of waves andcirculation and has recently been applied to hindcastKatrina Rita Gustav and Ike [Westerink et al 2008 Die-trich et al 2011a 2011b 2012b] As a means of compari-son ADCIRC has also been coupled to the Wave Model(WAM) and the Steady State Spectral Wave (STWAVE)model [Komen et al 1994 Smith 2000 Smith et al2001 Geurounther 2005 Smith 2007 Bender et al 2013]which evaluate wave conditions on a sequence of struc-tured grids throughout the Gulf and LATEX shelf and hasbeen used to hindcast Katrina Rita and Gustav [Bunya etal 2010 Dietrich et al 2010 2011a]

[15] For Ike the SWANthornADCIRC model uses theSL18TX33 computational domain that encompasses thewestern North Atlantic Gulf of Mexico and CaribbeanSea and provides a very high level of resolution on the LA-TEX shelf and adjacent floodplain from Pensacola FL tothe Texas-Mexico border The SL18TX33 computationaldomain is an evolution of a sequence of earlier Louisiana

models with significant refinements in grid resolution andthe incorporation of the entire Texas coastal floodplain[Westerink et al 2008 Bunya et al 2010 Dietrich et al2010 2011a] Nearshore and onshore maximum elementsize is 200 m with a minimum of 20 m in channels and riv-ers The continental shelf in the Gulf of Mexico is resolvedwith an element size of 500 m to 1 km increasing to 1ndash5km in the deep Gulf of Mexico The SL18TX33 mesh is animprovement over earlier studies because high levels of re-solution are extended from the southern Texas borderthrough Mobile Bay AL and thus it describes the entireregion that was affected as Ike moved onto the shelf andmade landfall

[16] Based on the unprecedented quality and quantity ofmeasured event wave and water level data the multitude ofdriver processes along the LATEX coast the developmentof a highly resolved computational model of the entire LA-TEX coast and adjacent basins and the availability of ahigh-resolution data-assimilated wind input field Ikepresents a unique and highly challenging opportunity tovalidate the performance of SWANthornADCIRC Modelwave and water level responses will be qualitatively andquantitatively evaluated in comparison to measured dataand put into context relative to the component physics

2 Model Description

[17] Significant progress has been made in recent yearsto achieve full dynamic coupling of riverine flow tidesatmospheric pressure wind and waves in simulating hurri-cane waves and circulation Basin-scale to inlet-scaledomains incorporate basins shelves inland water bodieschannels and floodplains and require high spatial meshvariability in order to properly resolve processes at a localscale Large high-performance computing platforms withover 10000 cores in conjunction with highly scalableunstructured mesh codes have allowed theseimprovements

21 Wave and Surge Model

[18] ADCIRC was implemented for this simulation as atwo-dimensional explicit barotropic model and solves themodified shallow water equations for water levels anddepth-averaged velocities in the x and y directions U and Vrespectively [Kolar et al 1994 Dawson et al 2006 West-erink et al 2008 Luettich and Westerink 2004 httpwwwunceduimsadcircadcirc_theory_2004_12_08 pdf]

[19] Sufficient mixing on the continental shelf due towave action has allowed for the two-dimensional depth-integrated version of ADCIRC to be successfully appliedObservations in the Gulf during Hurricane Ivan (2004)indicate a well-mixed layer of 60 m during the passage ofthe storm [Mitchell et al 2005] Numerical studies suggestthat turbulent mixing due to the interaction of windswaves and currents during Hurricane Frances (2004) in theupper ocean boundary layer extends down on the order of100 m [Sullivan et al 2012]

[20] The integrally coupled SWANthornADCIRC modeloperates on a single unstructured mesh with ADCIRC solv-ing for water levels and currents via the shallow waterequations at a 05 s time step ADCIRC passes these solu-tions to the unstructured implementation of SWAN which

solves the wave action balance equation and passes waveradiation stresses back to ADCIRC [Booij et al 1999 Riset al 1999 Zijlema 2010 Dietrich et al 2011b] Infor-mation is exchanged every 600 model seconds equivalentto the time step used in the SWAN computation For theSWAN model wave direction is discretized into 36 regularbins frequency is logarithmically distributed over 40 binsranging from 0031384 to 142 Hz wave growth mecha-nisms due to wind formulation is based on Cavaleri andRizzoli [1981] and Komen et al [1984] and modifiedwhitecapping is based on Rogers et al [2008] In shallowwater depth-induced wave breaking is determined viaBattjes and Janssenrsquos [1978] spectral model with the break-ing index set to frac14 073 [Battjes and Stive 1985] Thesesource term parameterizations are identical to recent stud-ies using SWANthornADCIRC [Dietrich et al 2011a]Within SWAN spectral propagation velocities are limitedin areas where insufficient mesh resolution may cause spu-rious wave refraction [Dietrich et al 2012a 2012b]

[21] Wave hindcasts are also performed with the WAMand STWAVE wave models coupled to ADCIRC WAM isrun on a Gulf-wide structured mesh and generates solutionsthat are forced as boundary conditions for STWAVE on asequence of structured grids along the LATEX coast[Komen et al 1994 Smith 2000 Smith et al 2001Geurounther 2005 Smith 2007 Bender et al 2013] WAM isa third-generation model solving the action balance equa-tion with 28 logarithmically distributed frequency bins and24 equally spaced directional bins run on a structured Gulf-wide mesh with 005 resolution WAM is run independ-ently using default parameters and its solution is used tospecify the wave conditions at the boundary of theSTWAVE nearshore wave model in conjunction withADCIRC-generated winds and water levels STWAVEuses a sequence of structured nearshore meshes with a reso-lution of 200 m STWAVE solves the wave action balanceequation using 45 frequency bins ranging from 00314 to208 Hz and 72 equally spaced directional bins The WAMSTWAVEthornADCIRC paradigm has demonstrated highskill in simulating nearshore waves and surge [Bunya et al2010 Dietrich et al 2010] Because of the loose couplingof ADCIRC to WAMSTWAVE model duration is notrequired to coincide

22 SL18TX33 Mesh

[22] The hindcast of Hurricane Ike applies theSWANthornADCIRC model to the SL18TX33 computationalmesh The mesh domain includes the western North AtlanticOcean Caribbean Sea Gulf of Mexico and coastal flood-plains of Alabama Mississippi Louisiana and Texas (Fig-ure 2) The mesh is the result of merging and refining twomeshes TX2008_R33 [Kennedy et al 2011a 2011b] andSL18 an evolution of the Louisiana SL16 mesh [Dietrich etal 2011a] Grid resolution varies from 20 km or larger inthe deep Atlantic and Caribbean 1ndash5 km in the central Gulfof Mexico 1 km and lower on the continental shelf 100ndash200 m in nearshore wave transformation zones and as smallas 20 m in channels and other similarly sized hydraulic fea-tures The mesh consists of 9108128 nodes (vertices) and18061765 triangular elements At every computationalnode over the 600 s coupling interval SWAN solves 1440unknowns (36 directions 40 frequencies every 600 s) for

every 3600 ADCIRC unknowns (x and y direction currentsand water level every 05 s)

[23] Bathymetric data for the Atlantic Caribbean anddeep Gulf of Mexico was obtained from the ETOPO1 dataset [Amante and Eakins 2009] Nearshore areas werespecified using Coastal relief digital elevation models(httpwwwngdcnoaagovmggcoastal) with data forinland water bodies including lakes channels and riverscoming from recent USACE and NOAA surveys Marsh to-pography was specified based on marsh type with the Loui-siana Gap Analysis Program (LA-GAP httpatlaslsuedurasterdownhtm) land-cover databases withnonmarsh topography based on LiDAR (httpatlas-lsuedulidar) [Dietrich et al 2011a] In all cases bathym-etrytopography was applied to the mesh using a localelement-scale averaging to avoid discontinuities Relevanthydraulic barriers such as levees roads and coastal dunesthat lie below minimum mesh resolution are represented inthe mesh as lines of raised vertices or submesh-scale weirs[Westerink et al 2008] All coastal features are set to ele-vations consistent with post-Ike conditions Bathymetricvalues and element sizes for the portion of the SL18TX33domain that include the LATEX shelf and coast aredepicted in Figures 3a and 3b

[24] The use of the SL18TX33 mesh captures the basinshelf-scale and inland response physics of tides wavesand surge generated by Ike The broad spatial scale of theprocesses driven by Ike necessitates a computational do-main encompassing the entire Gulf of Mexico and LATEXcoast

23 Winds

[25] Ikersquos core wind field was developed by NOAArsquosHurricane Research Division Wind Analysis System(HWIND) To create the wind field data were assimilatedfrom in situ monitoring systems (buoys and wind towers)remote sensing by satellites and active measurement byaircraft [Powell et al 1996 1998 2010] HWIND analy-sis is provided for an 8 8 area centered on the centralposition of the storm HWIND analysis is provided at 3 hintervals starting at 1930 UTC 5 September 2008 until1630 UTC 13 September 2008 HWIND analysis isblended with Gulf scale winds produced by the InteractiveKinematic Objective Analysis (IOKA) system [Cox et al1995 Cardone and Cox 2009] Final wind fields representthe conditions of 30 min sustained wind speeds at a heightof 10 m with marine exposure Gulf-wide winds are appliedat a resolution of 01 with a finer resolution of 0015 near

Figure 2 The SL18TX33 domain and grid bathymetry (m) of the SL18TX grid Ikersquos track is shownwith the black line for reference

the landfall location Final wind fields are provided at 15min intervals starting at 1200 UTC 5 September 2008 until0600 UTC 14 September 2008 It should be mentioned thatthe analyzed high resolution OWI HWINDIOKA datainput into ADCIRC differs slightly from the data thatappears in Berg [2009] resulting in slight discrepanciesbetween modeled winds and reported winds

[26] ADCIRC reads these marine wind fields and appliesa wind gust factor of 109 to convert the 30 min sustainedwinds to 10 min sustained winds to be consistent with itsair-sea drag formulation as well as a directional wind

reduction factor representing the reduction in 10 m windspeed as the atmospheric boundary layer evolves due tosurface roughness on land [Bunya et al 2010] ADCIRCapplies a wind drag coefficient that is data-driven windspeed limited and directional [Powell et al 2003 Powell2006 Dietrich et al 2011a]

24 Vertical Datum Adjustment

[27] At the initiation of the simulation at 0000 UTC 8August 2008 water levels are increased to correspond to thedatum shift from local mean sea level to NAVD88 updated

Figure 3 (a) Bathymetrytopography (m) (b) grid size (m) and (c) Manningrsquos n of the SL18TX33grid on the LATEX shelf and coast

to the 200465 epoch to account for the intraannual sea sur-face variability driven by effects such as upper layer warm-ing and seasonal riverine discharges and the measured sealevel rise from 2004 to 2008 The sea surface is raised 0134m to adjust computed values to NAVD88 200465 [Garsteret al 2007 Bunya et al 2010] and 0025 m due to sealevel rise from 2004 to 2008 Then 0121 m is added due tothe intraannual variation creating a total adjustment of0134 mthorn 0025 mthorn 0121 mfrac14 0280 m (httptidesand-currentsnoaagovsltrendssltrends shtml)

25 Bottom Friction

[28] Hydraulic friction is parameterized in the ADCIRCmodel using a spatially varying Manningrsquos n value [Bunyaet al 2010] These values are applied based on data sup-plied from the following land cover databases LA-GAPMississippi Gap Analysis Program (MS-GAP httpwwwbasicncsuedusegapindexhtml) and the CoastalChange Analysis Program (C-CAP httpwwwcscnoaa-govdigitalcoastdataccapregional) The land classifica-tions have standard Manningrsquos n values associated withthem that are assigned to the nodes via pixel averagingwith values detailed in Dietrich et al [2011a] Offshoreareas with sandygravel bottoms such as the Florida shelfare set to nfrac14 0022 and areas with muddy bottoms like theLATEX shelf are set to nfrac14 0012 [Buczkowski et al2006] The lower LATEX shelf friction is critical to devel-oping fast flows that generate the large forerunner observedduring the storm [Kennedy et al 2011a 2011b] These val-ues are applied at depths gt5 m and they are increased line-arly to nfrac14 0022 toward the shoreline Manningrsquos n valuesfor a portion of the SL18TX33 domain including the LA-TEX shelf and coast are depicted in Figure 3c

[29] SWAN utilizes a roughness length formulated byMadsen et al [1988] based on Manningrsquos n values used inADCIRC and water depths computed in ADCIRC

z0 frac14 Hexp 1thorn H1=6

nffiffiffigp

where frac14 04 (Von Karman constant) Hfrac14 total waterdepth computed in ADCIRC and gfrac14 gravitational constant[Bretschneider et al 1986] SWAN computes a newroughness length at each time step based on updatedADCIRC water level values To avoid unrealistically smallroughness length values the minimum Manningrsquos n valuepassed to SWAN is nfrac14 002 (minimum n is set to 003 forSTWAVE)

26 Rivers

[30] River inflow into the domain occurs at two loca-tions Baton Rouge LA representing the Mississippi Riverand Simmesport LA representing the Atchafalaya RiverBoth locations use a river-wave radiation boundary condi-tion in order to allow tides and storm surge to propagateupstream past these boundaries [Westerink et al 2008Bunya et al 2010] River flow is ramped up from zerousing a hyperbolic ramp function for a period of 05 daysFollowing the ramping period river levels are given 3 daysto reach equilibrium After 35 days river levels at theinflow boundaries are held constant and tidal forcing com-mences with meteorological forcing starting at a later

specified time River discharges were determined usingdata from the US Army Corps of Engineers New OrleansDistrict (httpwwwmvnusacearmymil) for the periodbetween 5 September 2008 and 15 September 2008 Riverflow rates used were 12210 m3s and 5233 m3s for theMississippi and Atchafalaya Rivers respectively

27 Tides

[31] Periodic conditions are applied at the open oceanboundary along the 60W meridian Astronomical tides(K1 O1 Q1 P1 M2 S2 N2 and K2) are forced on the openocean boundary using the TPXO72 tidal atlas [Egbert etal 1994 Egbert and Erofeeva 2002] Nodal factors andequilibrium arguments are computed and applied for thesimulation start time Tides are ramped using a hyperbolictangent function for 12 days to avoid exciting spuriousmodes in the resonant Gulf of Mexico and Caribbean Seabasins reaching full amplitude 25 days before the start ofmeteorological forcing

3 Recorded Data

[32] Following Katrina and Rita existing gages werestrengthened to assure data records were produced for theduration of tropical storms Additionally temporary gageswere placed in nearshore areas such as marshes creeks and1ndash5 km offshore to produce a composite understanding ofwave and surge generation evolution and dissipation andprovide a wealth of validation data (Table 3) Each time se-ries was reviewed and assessed for accuracy and reliabilitywith range limited or failed periods of data being removedto assure appropriate comparison to model solutions

4 Synoptic History and Validation

[33] The evolution of Hurricane Ike winds waves andsurge fields as simulated by the coupled SWANthornADCIRCmodel and qualitative and quantitative comparisons to datausing the extensive wave and water level data are pre-sented The simulation is started from a cold start on 0000UTC 8 August 2008 with a 35 day riverine spin-up periodallowing river levels to reach equilibrium followed by a 12day tidal spin allowing the tides in the Gulf of Mexico toattain a dynamic equilibrium A 105 day Gustav simula-tion is run from 0000 UTC 26 August 2008 to 1200 UTC 5September 2008 to establish ambient water level conditionsprior to Ike which is simulated over a 10 day period from1200 UTC 5 September 2008 to 1200 UTC 15 September2008 Wind wave water level and current fields through-out the period of 18 h prior to landfall to 12 h after landfallare shown in Figures 4ndash8 Time series and locations ofselect wind wave water level and current stations are pre-sented in Figures 9ndash25

41 Winds

[34] Ike crossed the 60oW meridian at 0430 UTC 5 Sep-tember 2008 entering the SL18TX33 domain Before enter-ing the Gulf of Mexico Ike made landfall in eastern andwestern Cuba Upon entering the Gulf at 2030 UTC 9 Sep-tember 2008 Ike moved northwest and grew in size [Berg2009] Tropical storm force winds (10 min sustained surfacewinds of at least 15 m s1) first reached the MississippiRiver Delta in Southern Louisiana at 1500 UTC 11

September 2008 40 h before landfall and persisted for morethan 36 h Winds over the Mississippi Breton and Chande-leur Sounds were consistently easterly and southeasterly anddirected toward the protruding Mississippi River Delta sig-nificantly impacting surge development in the regionAccording to OWI HWINDIOKA reanalysis Ike reached

its peak wind speed of 41 m s1 in the Gulf of Mexico at0430 UTC 12 September 2008 At this point Ikersquos tropicalstorm force and stronger winds produced an integrated ki-netic energy of 154 TJ corresponding to a 54 out of a possi-ble 6 on the Surge Destructive Potential Scale [Powell andReinhold 2007] with tropical storm force winds and

Figure 4 Wind speeds m s1 on the LATEX shelf and coast during Ike Vectors representing windspeed and direction are displayed Plots represent the following times (a) 1300 UTC 12 September2008 approximately 18 h before landfall (b) 1900 UTC 12 September approximately 12 h before land-fall (c) 0100 UTC 13 September approximately 6 h before landfall (d) 0700 UTC 13 Septemberapproximately at landfall (e) 1300 UTC 13 September approximately 6 h after landfall and (f) 1900UTC 13 September approximately 12 h after landfall

hurricane force winds extended out 400 km and 140 kmrespectively from the center of the hurricane After slightlyweakening later on 12 September 2008 Ike would againreach a peak wind speed of 41 m s1 before and at landfallat Galveston TX at 0700 UTC 13 September 2008

[35] During the period from 1300 UTC 12 September2008 18 h prior to landfall until 0100 UTC 13 September2008 6 h prior to landfall much of the LATEX shelf andcoast experienced shore-parallel winds as a result of thelarge size of the storm and large-scale circular coastal ge-ography of the region Figures 4andash4c Winds shifted slowly

as the storm progressed and areas in the immediate vicinityof landfall such as Galveston Island and the Bolivar Penin-sula did not experience a shift in wind direction until im-mediately before the stormrsquos center had made landfall Atlandfall (Figure 4d) Ikersquos maximum wind speed was 41 ms1 occurring at the coast of the Bolivar Peninsula As Ikeapproached the coast and made landfall winds transitionedto shore-normal orientation blowing onshore northeast oflandfall and offshore southwest of landfall The stormtracked through the east side of Galveston Bay which atlandfall was already filled with more than 2 m of additional

Figure 4 (continued)

water caused by the forerunner surge and was impacted bynear-maximum-strength winds before landfall and 30 ms1 winds immediately after landfall

[36] Following landfall winds over Galveston Bay and inthe area of landfall remained oriented onshore Six hours af-ter landfall winds over Galveston Bay were 20 m s1 still

tropical storm force (Figure 4e) These persistent onshorewinds impeded the recession of water out of Galveston Bayand the marshes to the northeast of Bolivar Peninsula wheremaximum recorded water levels during Ike occurred

[37] Figure 9 shows the locations of six observation sta-tions on the LATEX shelf and onshore that recorded wind

Figure 5 SWAN significant wave heights (m) on the LATEX shelf and coast during Ike Vectors rep-resenting wind speed and direction are displayed Plots represent the following times (a) 1300 UTC 12September 2008 approximately 18 h before landfall (b) 1900 UTC 12 September approximately 12 hbefore landfall (c) 0100 UTC 13 September approximately 6 h before landfall (d) 0700 UTC 13 Sep-tember approximately at landfall (e) 1300 UTC 13 September approximately 6 h after landfall and (f)1900 UTC 13 September approximately 12 h after landfall

velocity and direction during Hurricane Ike Figures 10 and11 compare the OWI HWINDIOKA-based wind speedsand directions as adjusted by ADCIRC (10 min averagewinds overland directional wind boundary layer adjust-ments adjustment for water column height relative tophysical roughness element scale) to the observed dataUnfortunately many data recording stations failed at orbefore peak winds near landfall leaving fewer points ofcomparison for the maximum winds It should be notedthat the OWI wind fields used as ADCIRC input representlarge-scale synoptic wind patterns and exclude local and

short time scale phenomena such as the diurnal cycle seenin the observed data This diurnal cycle is particularlyprominent at station TCOON 87730371 In regard to thesynoptic cyclonic winds the OWI winds capture well thegrowth peak and reduction of wind velocities Of particu-lar note is the capture of the passing of the eye at stationTCOON 87710131 One particular source of error in theOWI winds is the underprediction of winds on the LATEXshelf before landfall as seen in stations TCOON 87713411and TCOON 87710131 between 3 and 15 h GMT on 12September These moderate velocity shelf parallel winds

drive the forerunner surge and underprediction of thesewinds leads to a lower shore parallel current and lowerwater levels prelandfall In regard to wind direction theOWI winds capture the shifting of winds as Ike made land-fall but fail to capture some of the short-time scale shifts inwind direction Because these short-duration localized phe-

nomena are not captured in the OWI winds they will notappear in the ADCIRC circulation response

42 Waves

[38] As Ike progressed through the Gulf of Mexico thelargest waves were generated by the stormrsquos most intense

Figure 6 SWAN peak period (s) on the LATEX coast during Ike Vectors representing wind speedand direction are displayed Plots represent the following times (a) 1300 UTC 12 September 2008approximately 18 h before landfall (b) 1900 UTC 12 September approximately 12 h before landfall (c)0100 UTC 13 September approximately 6 h before landfall (d) 0700 UTC 13 September approxi-mately at landfall (e) 1300 UTC 13 September approximately 6 h after landfall and (f) 1900 UTC 13September approximately 12 h after landfall

winds located to the east of the eye as illustrated in Figures5 and 6 In the northeastern Gulf deep water NDBC buoys42036 and 42039 recorded significant wave heights of 4 mand 8 m respectively and maximum mean wave periods of10 s and 12 s respectively (Figures 12ndash14) Ike passed justto the east of NDBC buoy 42001 generating a maximumsignificant wave height of almost 10 m before the stormpassed and 8 m afterward with a maximum mean period of12 s as the storm center passed over the buoy (Figures 12ndash14) Maximum computed SWAN significant wave heightsin the Gulf of Mexico exceeded 15 m occurring in the

deep Gulf to the south of the Louisiana continental shelfbreak Far to the west of the track at NDBC buoys 42002and 42055 significant wave heights reached 6 m and 3 mrespectively and mean periods reached 13 s at both buoys(Figures 12ndash14)

[39] To the east of New Orleans on the Alabama-Mississippi Shelf the shallow bathymetry and the associ-ated depth-limited breaking attenuated the large oceanswell (Figures 5 and 6) Furthermore the ChandeleurIslands prevented these large long waves from entering theChandeleur Sound limiting wave heights in the Sound to

lt2 m In the Biloxi Marsh friction and even shallowerdepths limited wave heights to 05 m and peak periods to 5s This rapid transformation from deep water to land isobserved by NDBC buoys 42040 and 42007 andCHL gages 2410510B 2410513B and 2410504B (Figures12ndash16 and 17)

[40] The narrow shelf to the south and west of the Mis-sissippi River Delta allows large swell waves to propagateclose to the delta and bays to the west (Figures 5 and 6)Rapid wave attenuation occurs as depths become shallowand wetlands are penetrated Offshore from TerrebonneBay CSI gages 06 and 05 recorded significant wave

Figure 7 ADCIRC water surface elevation (m) on the LATEX shelf and coast during Ike Vectorsrepresenting wind speed and direction are displayed Plots represent the following times (a) 1300 UTC12 September 2008 approximately 18 h before landfall (b) 1900 UTC 12 September approximately 12h before landfall (c) 0100 UTC 13 September approximately 6 h before landfall (d) 0700 UTC 13 Sep-tember approximately at landfall (e) 1300 UTC 13 September approximately 6 h after landfall and (f)1900 UTC 13 September approximately 12 h after landfall

heights of 6 m and 3 m respectively and a maximum peakwave period of 16 s (Figures 12 16 and 17) CHL wavegage 2410512B in the marshes to the north of TerrebonneBay recorded significant wave heights of 1 m and peakwave periods reached a maximum of 3 s demonstrating thedepth limited and bottom friction induced breaking thatoccurs in the bay and marsh system

[41] The broad Texas shelf also limited the propagationof the large swell waves generated in the central deep Gulf(Figures 5 and 6) NDBC buoys 42019 and 42020 are bothpositioned on the outer Texas shelf southwest of landfall

and recorded significant wave heights of up to 7 m andmaximum mean wave periods of 12 s and 14 s respectivelyOn the inner Texas shelf NDBC buoy 42035 (which wasdislodged from its mooring as the storm passed httpwwwndbcnoaagovstation_pagephpstationfrac1442035) wasinitially located just to the south of Ikersquos track and recordeda significant wave height of 6 m and maximum mean waveperiod of 13 s before being dislodged in the hours before Ikepassed On the nearshore Texas shelf Andrew Kennedyrsquos(AK) gages Z Y X W V S and R shown in Figures 1216 and 17 recorded wave heights and peak periods in mean

water depths of 85ndash16 m covering a section of coast fromBolivar Peninsula north of landfall to Corpus Christi southof landfall Stations AK Z and Y to the north of landfallexperienced the strongest landfalling winds and recordedsignificant wave heights of 5 m and peak wave periods of 16

s prior to landfall and 6ndash12 s at landfall indicating the transi-tion from swell dominance to wind-sea dominance as Ikepassed To the south of landfall AK stations X V S and R(Figure 12) recorded maximum significant wave heights of58 m 5 m 3 m and 45 m respectively (Figure 16) Based

Figure 8 ADCIRC currents (m s1) on the LATEX shelf and coast during Ike Vectors representingwind speed and direction are displayed Plots represent the following times (a) 1300 UTC 12 Septem-ber 2008 approximately 18 h before landfall (b) 1900 UTC 12 September approximately 12 h beforelandfall (c) 0100 UTC 13 September approximately 6 h before landfall (d) 0700 UTC 13 Septemberapproximately at landfall (e) 1300 UTC 13 September approximately 6 h after landfall and (f) 1900UTC 13 September approximately 12 h after landfall

on the timing of the maximum significant wave height andpeak period at the time of maximum significant wave height(Figure 17) the largest waves at stations V S and R werethe result of swell generated offshore

[42] SWAN WAM and STWAVE wave characteristicsare compared to measured values at representative stationsin Figures 12ndash17 At the deep water NDBC buoys 4203942036 42001 42002 and 42055 are shown in Figures 12ndash15 both SWAN and WAM capture the growth of swellwaves as Ike progresses through the Gulf At nearshorebuoys SWAN more accurately captures the maximum sig-

nificant wave heights as seen at NDBC buoy 42007 nearthe Mississippi-Louisiana coast (Figures 12 and 13) AtNDBC buoy 42002 a dramatic departure is seen betweenthe recorded and computed mean wave direction and themean wave direction modeled by SWAN beginning atlandfall This is due to the measurement range limitation ofhigh wave frequencies at NDBC buoys due to the nature ofthese large wave gages By landfall at buoy 42002 the seastate had transitioned to locally generated wind waveswhich are not accurately captured by the large NDBCbuoys Therefore the mean wave direction is based on the

dominant wave period that can be captured by the buoywhich in this case does not align with the local wind waves

[43] In the Biloxi Marsh SWAN captures the smalllocally generated waves as seen at stations USACE CHL2410510B 2410523B and 2410504B (Figures 16 and 17)At the CSI gages 05 and 06 south of Terrebonne BaySWAN accurately captures the arrival of swell generatedoffshore (Figures 16 and 17) North of Terrebonne Bay atCHL gage 2410512B SWAN accurately models the small1 m significant wave height but slightly overestimates thepeak wave period of 3 s (Figures 12 16 and 17) As in theBiloxi Marsh wave solutions in this area are highly sensi-tive to water depth and bottom friction

[44] On the outer TX shelf at NDBC buoys 42020 and42019 both SWAN and WAM capture the development ofswell and peak significant wave heights At nearshoreNDBC buoy 42035 WAM severely underpredicts the de-velopment of swell and peak significant wave heightwhereas SWAN captures the peak as well as wave growth(Figures 12ndash14) At AKrsquos inner shelf gages along the TX

coast both SWAN and STWAVE capture maximum sig-nificant wave heights as well as wave growth prior tolandfall (Figure 16) At AK stations X Y and Z peak sig-nificant wave heights were wind-seas generated by stronglandfalling winds This is opposed to stations V S and Rwhere winds were weaker and maximum wave heightswere generated by swell in the deep Gulf Figure 16 showsa late arrival of the peak significant wave height at AKstations X V S and R This late arrival of maximum sig-nificant wave heights at the inner shelf stations away fromlandfall and underprediction of waves prior to landfall atstations near Ikersquos landfall location indicates an artificialretardation of swell across the TX shelf Despite thisSWAN models the quick transition from swell to wind-sea at landfall as shown in Figure 17 STWAVE also cap-tures this transition but it is more gradual in comparisonto SWAN

[45] For all measured time series agreement of modeledresults to measured data can be quantified via the ScatterIndex (SI)

Figure 9 Locations of NOAA and TCOON stations on the LATEX shelf NOAA in red TCOON inblue Ike track is in black the coastline is in gray and SL18TX33 boundary and raised features in brown

Figure 10 Time series (UTC) of wind velocities (m s1) at NOAA and TCOON stations ADCIRCoutput in black Observation data in gray Dashed green line represents landfall time

Figure 11 Time series (UTC) of wind direction () at NOAA and TCOON stations ADCIRC outputin black observation data in gray Dashed green line represents landfall time

Figure 12 Locations of NDBC CSI CHL and AK gages in the Gulf of Mexico NDBC in blackCSI in red CHL in green and AK in blue Ike track is in black the coastline is in gray andSL18TX33 boundary and raised features in brown NDBC 42058 lies outside the frame in the Carib-bean Sea

SI frac14

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1N

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

where N is the number of observed data points Si is themodeled data value Oi is the measured value Eifrac14 SiOiand E is the mean error [Hanson et al 2009] The SI is theratio of the standard deviation of model error to the meanmeasured value Tables 4 and 5 summarize SI and bias forall measured wave data It should be noted that WAM andSTWAVE are subject to slightly different wind forcingthan SWAN SWAN receives its winds from ADCIRCwhere overland winds are reduced due to directionalonshore roughness Thus a narrow zone of offshore

Figure 13 Time series (UTC) of significant wave heights (m) at 12 NDBC stations SWAN results arein black WAM results are in blue and STWAVE results are in red Dashed green line represents landfalltime

directed winds adjacent to noninundated land areas will bedifferent However the offshore marine winds with no landboundary layer adjustments are the same for all threemodels

[46] Table 4 summarizes model performance at everystation within each wave modelrsquos domain while Table 5summarizes error statistics only at stations shared by atleast two wave models In general good agreement is seenbetween SWAN and WAMSTWAVE to measured data atNDBC CSI and AK gages SI and bias values for signifi-

cant wave heights mean and peak periods and mean direc-tion at NDBC CSI and AK gages are similar to thosefound in previous SWANthornADCIRC validation studies[Dietrich et al 2011a] Table 4 provides an overall assess-ment of model performance but to understand how thewave models performed in relation to one another Table 5must be examined Overall SWAN and WAMSTWAVEperform comparably but some regional and model differ-ences can be discerned by looking at model performance indiffering coastal geographies at common stations At

Figure 14 Time series (UTC) of mean wave period (s) at 12 NDBC stations SWAN results are inblack WAM results are in blue and STWAVE results are in red Dashed green line represents landfalltime

stations common to both SWAN and WAMSTWAVEwave heights are overestimated for all geographic locationsand models with the exception of WAMSTWAVE atNDBC buoys SI and bias increase as stations are located inincreasingly shallow water implying a trend of overesti-mated wave heights in very shallow water A slight advant-age is seen with WAMSTWAVE in peak wave period incoastal waters For inland waters SWAN performs betterfor peak period It should be noted that wave heights aresmall at inland stations Mean wave direction represents

the wave parameter where one model clearly outperformsthe other SWAN has significantly lower bias and SI com-pared to WAMSTWAVE in deep water Unfortunatelymean wave direction was only recorded at NDBC deepwater stations making it impossible to see if the spatialtrend of increasing accuracy in deep waters extends to shal-lower water The spatial trend of decreasing accuracy anddiffering model results in shallow water may be due toeach modelrsquos representation of bathymetry mesh resolu-tion and the general increased sensitivity of wave

Figure 15 Time series (UTC) of mean wave direction () at 12 NDBC stations SWAN results are inblack WAM results are in blue and STWAVE results are in red Dashed green line represents landfalltime

parameters to shallower depths WAM STWAVE andSWAN operate on different grids with very different levelsof grid resolution in the nearshore affecting process resolu-tion and the depiction of bathymetry in the various modelsThis is likely the cause of some of the differences in thesolutions of the models at NDBC station 42007 and 42035Specifically the WAM grid is poorly resolved at these sta-tions where large gradients in bathymetry and wave charac-teristics occur Particular attention must be given to theinland gages where both SWAN and WAMSTWAVE per-formed poorly in proportionally based errors due to thesmall wave amplitude values The inland sample size issmall (4 gages) but the poor results indicate a deficiency in

modeling waves in shallow inland waters This deficiencystems from the large sensitivity of small inland waves towater levels and bottom friction parameterization The factthat both models produce poor results and the water surfaceelevation results produced by ADCIRC which are used toforce the wave models are quite accurate (Table 6) wouldindicate that both the SWAN and WAMSTWAVE modelssuffer from poor bottom friction parameterization for shortinland wind waves

43 Surge and Currents

[47] Ikersquos unusually large wind field in the Gulf of Mex-ico resulted in a myriad of surge processes occurring over a

Figure 16 Time series (UTC) of significant wave heights (m) at 12 CSI CHL and AK gages SWANresults are in black and STWAVE results are in red Dashed green line represents landfall time

1000 km stretch of the LATEX shelf and coast The stormsurge response is regional and depends on the geographyand orientation of the shelf and the characteristics of thestorm

[48] Due to Ikersquos large wind field and track across theGulf of Mexico easterly winds persisted over the Missis-sippi Chandeleur and Breton Sounds for over 36 h Whilethe winds over these Sounds never exceeded 20 m s1 thelong duration and steady direction allowed for effectivepenetration of surge generated over these waters and intothe lakes and marshes surrounding New Orleans (Figure 7)

NOAA gages 8761305 and 8761927 (Figures 18 and 19)located on the south shore of Lakes Borgne and Pontchar-train recorded a maximum surge level of 21 m and 18 mrespectively 15 and 7 h before landfall The similarity inthese hydrographs (with a time lag as the water movesinland) demonstrate the large-scale spatial response in theregion and the slow time scale of the response allowingLake Pontchartrain to be effectively filled To the southeastof New Orleans winds over the Chandeleur and BretonSounds forced surge into the Biloxi and CaernarvonMarshes to the east of the Mississippi River and against the

Figure 17 Time series (UTC) of peak wave period (s) at 12 CSI CHL and AK gages SWAN resultsare in black and STWAVE results are in red Dashed green line represents landfall time

associated levee systems The Delta and levee systemextends far onto the continental shelf effectively capturingthe locally generated surge coming from the shallow watersto the east of the Mississippi River CHL gages 2410504Band 2410513B and CRMS gage CRMS0146-H01 (Figures20 and 21) located in the Biloxi and Caernarvon Marshesall recorded maximum surge levels of 2 m Water levelsrise as the surge is blown over the shallow Caernarvonmarsh and against the river levee south of English Turn inPlaquemines Parish where surge reached 3 m indicatingthat no attenuation in surge occurred over the CaernarvonMarsh In fact the steady winds allowed water levels toincrease across the marsh

[49] The buildup of surge to the east of the MississippiRiver combined with the lack of surge buildup to the westof the river created a water surface gradient that produced astrong current around the Delta (Figure 8) This 2 m s1

current persisted to the south of the Delta for over 24 h[50] To the west of the Delta the narrow continental shelf

allowed large swell generated in the deep Gulf to propagateclose to coastal wetlands The coast in this area experiencedslightly onshore moderate velocity (not exceeding 20 ms1) winds and when combined with the wave setup causedby large breaking waves nearshore a slow rise of water wasobserved Simulations where the wave interaction wasneglected indicated that up to 50 of storm surge on theDelta was generated by wave setup This is consistent withthe steep bathymetry in the region and previously validatedstorms [Dietrich et al 2011a 2010 Bunya et al 2010Kerr et al 2013a 2013b] USACE gage 82260 and USGS-Perm gage 292800090060000 (Figures 20 and 21) locatedin this region to the northwest of Barataria Bay recordedmaximum water levels of 2 m and 16 m respectively

[51] To the west of Barataria Bay the continental shelfprogressively broadens to over 200 km at its widest pointin the vicinity of Lakes Sabine and Calcasieu This wideshallow shelf and large scale concave coastal geography ofthe LATEX coast combined with Ikersquos steady prelandfallwinds to generate a strong long-lasting shore-parallel cur-rent (Figure 8) Figure 23 shows the location of observedcurrents on the LATEX shelf and Figures 24 and 25 showADCIRC and observed current velocity and direction On

the Louisiana shelf CSI station 3 shows a gradual increasein current speed beginning on 12 September reaching itsrecording maximum value of 1 m s1 12 h before landfallOn the Texas shelf (Figures 23ndash25) at the Texas AutomatedBuoy System (TABS) current data buoys show the devel-opment of the forerunner driving current Unfortunatelythe TABS buoys are not able to record currents in excess of1 m s1 as is seen in the plateau in the velocities As thestorm approached the steady shore-parallel current createda geostrophic setup that caused a rise in water at the coaststarting 24 h before landfall [Kennedy et al 2011a2011b] This geostrophic setup typically identified as aforerunner surge is only possible due to the strong (morethan 1 m s1) shore parallel current driven by shore parallelwinds as seen in Figures 4 8 10 and 24 The low bottomfriction and wide shelf are vital components to the largeamplitude of the forerunner Figure 7a shows 1 m of surgehas developed on the entire LATEX shelf 18 h before land-fall when winds on the coast did not exceed 20 m s1 andwere generally shore parallel or directed offshore Thisgeostrophic setup is illustrated at several gages across theLATEX coast UND Kennedy Z and Y (Figure 19) bothshow a gradual rise in water beginning early on 12 Septem-ber 2008 24 h before landfall Similar to the flooding pro-cess to the east of the Mississippi River the long time scaleof the geostrophic process allowed water to penetrate farinland Onshore in Texas TCOON gages 87704751 and87707771 (Figures 18 and 19) located inland in Lake Sab-ine and Galveston Bay respectively both recorded a rise inwater starting early on 12 September 2008 when winds atthe coast were still strongly shore-parallel or slightly off-shore These two TCOON gages are of particular interestbecause they demonstrate the inland penetration of theforerunner surge into coastal lakes and bays in advance oflandfall This early inundation only weakly affects peaksurge at the coast because the forerunner surge propagateddown the LATEX shelf prior to the landfall of the stormand the primary surge in open water is generated by land-falling shore perpendicular winds However the forerunneris critical to inland coastal estuarine and wetland areas par-ticularly regions that would later experience the strongwinds associated with landfall The early penetration

Figure 18 Locations of AK NOAA TCOON and CSI gages on the Louisiana-Texas coast AK inblack NOAA in red TCOON in blue and CSI in green Coastline is in gray and SL18TX33 boundaryand raised features in brown

occurs at a slow time scale and is retained by the inlandwaters after the propagation of the open water forerunnerdown the shelf toward Corpus Christi This early inlandinundation and retention exacerbated the impact of thelocally generated surge within inland coastal lakes andbays at landfall

[52] Following generation the geostrophically driven fore-runner surge propagated down the shelf as a free continentalshelf wave To the southwest of landfall the forerunner surgecan be seen in Figures 7dndash7f and 8andash8f Propagating downthe shelf the peak of the free wave reached Corpus Christi

and TCOON gage 87758701 (Figures 18 and 19) as Ike wasmaking landfall on the Bolivar Peninsula

[53] Prior to landfall (Figures 7andash7c) water levels in theregion near landfall are driven by a predominantly shore-parallel process the forerunner surge Starting in Figure7d surge at the coast of the Bolivar Peninsula has transi-tioned to a shore-normal process driven by strong shore-normal winds Figure 7d shows the buildup of water againstthe Bolivar Peninsula and Figure 7e shows the wind-drivenprogression of water over the Peninsula inland and onto thecoastal floodplain while the surge at the coast has rapidly

Figure 19 Time series (UTC) of water surface elevations (m) at 12 AK NOAA TCOON and CSIgages ADCIRC output in black observation data in gray Dashed green line represents landfall time

recessed back into open water Figure 7f shows the over-land recession process which is impeded by the persistentbut weakening shore normal winds following landfall andalso by the frictionally dominated coastal floodplain AKstations Z and Y are offshore from the Bolivar peninsulaand recorded a peak surge of 46 m and 43 m respectively(Figures 18 and 19) Onshore FEMA high water marks andUSGS-Temp gages extensively covered the area near land-fall USGS-DEPL gages USGS-DEPL_SSS-TX-GAL-001and USGS-DEPL_SSS-TX-GAL-002 were located on theGulf side and bay side of Bolivar Peninsula and recordedmaximum water levels of 48 m and 4 m respectively (Fig-ures 20 and 22) This lower bayside elevation relates to bayand inland penetration time scale lag due to frictional re-sistance Inland sides of barrier islands will typically lagbehind the open coast side due to overland frictional resist-ance and other processes such as wave radiation inducedsetup at the coast from large swell waves To the northeastof landfall a consistent water level of 5 m was measuredby USGS-DEPL gages USGS-DEPL_SSS-TX-JEF-001USGS-DEPL_SSS-TX-JEF-004 and USGS-DEPL_SSS-TX-JEF-005 (Figures 20 and 22) Further inland FEMAmeasured two still-water high water marks exceeding 51 min Jefferson County over 15 km from the coast represent-ing the highest recorded surge elevation during the event

[54] Water recessed rapidly back onto the shelf and intothe deep ocean from the almost 48 m mound of water

driven against the shore at landfall This flux of water to-ward the deep Gulf was reflected back toward the coast asan out-of-phase wave due to the abrupt bathymetric changeat the continental shelf break This process can be seenacross the LATEX coast between the Atchafalaya Basinand Galveston Island This cross-shelf reflection can beseen in Louisiana at CSI gage 03 and in Texas at AK gagesZ Y and W (Figures 18 and 19) This reflection almostcertainly relates to the shelf resonance as the post-stormsecondary and tertiary peaks occur at approximately 12 hintervals during the reflections According to Sorensen[2006] the length of an open resonant basin at the basicmode is computed as

L frac14 TffiffiffiffiffiffiffigHp

where L is the length of the open basin T is the resonantperiod and H is the water column depth Assuming an av-erage depth on the shelf of 30 m the resonant basin lengthis approximately 185 km This is consistent with the widthof the LATEX shelf along western Louisiana and easternTexas which varies between 160 and 220 km The 12 hresonant period of the broad LATEX shelf is also evi-denced by the strong amplification of semidiurnal tides onthis shelf [Mukai et al 2002] The fluctuating current fieldsin Figures 8dndash8f represent the signature of the cross shelfresonance

Figure 20 (a) Locations of USACE (black) USACE-CHL (red) USGS (green) CRMS (blue) andUSGS-DEPL (purple) gages on the LATEX coast (b) subset of locations shown in Figure 20a for whichhydrographs are shown Coastline is in gray and SL18TX33 boundary and raised features in brown

[55] ADCIRC water surface elevations and currents arecompared to measured values at representative stations inFigures 18ndash25 To the east of the Mississippi River theADCIRC model accurately captures the rise of water in thelakes and bays surrounding New Orleans shown at NOAAgages 8761927 and 8761305 in Figures 18 and 19 The skillshown in modeling the surge generated on the MississippiSound that penetrated into Lakes Borgne and Pontchartrainindicates that the SL18TX33 model has adequate resolutionin the small scale channels and passes hydraulically con-necting the sound and lakes In the Biloxi and Caernarvon

marshes the early rise in water and associated inland pene-tration process are captured by the model shown at CHLgages 2410513B and 2410504B (Figures 20 and 21)ADCIRC slightly overpredicts the peak surge at CRMSgage CRMS_CRMS0146-H01 in the Caernarvon Marsh(Figures 20 and 21) however based on the recorded datait appears that the gage has an upper limit of measurementof 2 m Model accuracy in this region indicates that the uni-versally applied air-sea drag and bottom friction in marshesand wetlands in the region are correctly parameterizedbecause the peaks are correctly captured and the flooding

Figure 21 Time series (UTC) of water surface elevations (m) at 12 USACE CHL USGS and CRMSgages ADCIRC output in black observation data in gray Dashed green line represents landfall time

and recession curves in this slow process are also well rep-resented During the recession of surge from the marshesbottom friction is the controlling process in the shallowoverland flow that occurs

[56] To the west of the Delta the complex interaction oflarge swell waves breaking nearshore shore-normal windsand a strong shore-parallel current is captured with a slightunderprediction of peak surge at USACE gage 82260 (Fig-ures 20 and 21)

[57] The forerunner surge is a shelf scale process that iseffectively captured by ADCIRC as shown at gages USGS-

PERM_07381654 USGS-DEPL_SSS-LA-VER-006 USGS-DEPL_SSS-LA-CAM-003 and USGS-DEPL_SSS-TX-GAL-002 AK gages Z Y W V and U and TCOON gages87704751 and 87707771 (Figures 18ndash20 and 22) but themodeled rise in water is slightly lower than the measureddata at some gages The model lag is most pronounced atgages AK Z and Y (Figures 18 and 19) This is also seen atTABS station B (Figures 23 and 24) where we note thatbetween 3 and 15 h UTC on 12 September there is an under-prediction in the shore parallel current speed by ADCIRCThis lag in forerunner surface elevations and the associated

currents is likely associated with the low bias in the OWIwinds during this time in the region as seen at stationsTCOON 87710131 and 87713411 (Figures 9 and 10) Theforerunner surge process is reliant on the generation of thesteady strong shore-parallel current necessary for geostro-phic setup Due to the cap on the measured currents on theshelf at the CSI and TCOON gages it cannot be determinedif ADCIRC accurately modeled the peak currents howevergood agreement is seen between ADCIRC and observedvelocities during most of the storm (Figures 23ndash25)ADCIRC also accurately captures the change in currentdirection as Ike moved across the shelf with an exception atTABS B to the southwest of landfall where ADCIRC failedto model the quick change in direction that occurred right atlandfall Note that on the shelf currents are likely quite uni-form over depth due to the vigorous wave induced verticalmixing [Mitchell et al 2005 Sullivan et al 2012]

[58] The propagation of the free wave down the coast iscaptured by ADCIRC as seen at TCOON station 87758701and AK stations V and U (Figures 18 and 19) In additionthe currents generated by the shelf wave are well repre-sented in the model as is shown by the comparisons atTABS station W (Figures 23ndash25)

[59] To the northeast of landfall where the maximumsurge levels occur ADCIRC accurately models the peakshore normal wind-driven surge ADCIRC shows goodagreement to peak surge offshore at AK stations Z and Yand inland at stations USGS-DEPL_SSS-TEX-JEF-005USGS-DEPL_SSS-TEX-JEF-004 USGS-DEPL_SSS-TX-JEF-001 USGS-DEPL_SSS-TX-GAL-001 and USGS-DEPL_SSS-TX-GAL-002 (Figures 18ndash20 and 22)

[60] The 12 h resonant wave on the LATEX shelf result-ing from coastal surge waters recessing into the deep Gulfis captured by ADCIRC This is seen at water surface

Figure 23 Locations of CSI and TABS stations on the Louisiana-Texas coast TABS in black CSI inred coastline is gray Hurricane Ikersquos track in black and SL18TX33 boundary and raised features inbrown

Figure 24 Time series (UTC) of current velocities (m s1) at CSI and TABS stations ADCIRC outputin black observation data in gray and dashed green line represents landfall time

elevations at AK stations Y and Z (Figures 18 and 19) andTABS current stations B and F (Figures 23ndash25)

[61] Maximum high water during the storm event is pre-sented in Figure 26 A spatial analysis of differencesbetween measured and modeled maximum high water atthe 206 FEMA HWMs and at 393 hydrograph-derived highwater values is shown in Figure 27 A comparison of modelto measurement high water at these same 599 locations isshown in Figure 28

[62] Table 6 summarizes the SI and bias of ADCIRCmodel results to recorded data for time series of water lev-els The overall time series scatter index (SI) equals01463 and indicates generally good agreement with thedata The bias of 00114 m indicates globally a smallunderprediction of water levels by ADCIRC Examiningboth time series and HWM error statistics in Table 6 indi-cates that coastal stations are generally more accuratelyhindcast than inland stations Coastal stations are slightly

overpredicted while inland stations are slightly biasedlow This is due to the general geographic simplicitylarger depths and homogeneity of frictional resistance ofthe open coast as compared to the nearshore and espe-cially floodplain As hurricane-driven storm surgeencroaches inland any number of complexities includingtopography bathymetry heterogeneous frictional resist-ance and subgrid scale impedances to flow can be foundsignificantly complicating the flow The poorest R2 valueis found at the CRMS stations (06833) The CRMS gagesare located in Louisiana with the majority of stationslocated in inland marshes Water levels in inland marshesare highly dependent on the level of connectivity tocoastal water bodies and while the SL18TX33 mesh accu-rately represents major channels and connections avail-able bathymetric and topographic data is not of highenough resolution to properly represent all relevantconnections

Figure 25 Time series (UTC) of current direction () at CSI and TABS stations ADCIRC output inblack observation data in gray and dashed green line represents landfall time

Table 4 Summary of Scatter Index (SI) and Normalized Error Bias for Wave Data Time Series for All Data Stations in a Modelrsquos Geo-graphic Coverage

Data Source Model

Sig Wave Height Peak Wave Period Mean Wave Period Mean Wave Direction

NumberofData Sets SI Bias

Number ofData Sets SI Bias

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

CSI STWAVE 4 01764 02641 4 01384 00678 4 01939 03831 0 na naSWAN 5 01478 01393 5 01601 00851 5 02106 03376 2 02197 01934

USACE-CHL STWAVE 4 04252 04632 4 09701 11156 4 15295 03000SWAN 5 08827 07621 5 04844 02645 5 28319 03580

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

[63] Figure 27 indicates that there is a small low bias insoutheastern Louisiana and a small high bias to the east ofthe storm track in southwestern Louisiana and easternTexas It is also important to note that the OWI HWINDIOKA blended winds utilized by ADCIRC are consideredthe best available representation of Ikersquos wind field butmay lack small-scale localized variations that may influ-ence local water levels In summary the overall ScatterIndex of 01463 bias of 00114 estimated ADCIRCHWM indicators of R2frac14 091 absolute average differenceof 017 m (equal to 012 m once measured HWM error esti-mates are incorporated) 94 of modeled HWMs within 50cm of measured HWMs and standard deviation of 022 m(equal to 019 m once measured HWM error estimates areincorporated) support the accuracy of this hindcast espe-cially for a storm with maximum high water levels exceed-ing 5 m

5 Conclusions

[64] Hurricane Ike made landfall as a strong category 2storm at Galveston TX Due to Ikersquos large wind field andthe LATEX shelf and the coastrsquos unique geography a num-ber of regional and shelf-scale processes occurred beforeduring and after landfall The extensive data set collectedduring Ike captures the multitude of processes that occurredduring Ike on the LATEX shelf and coast and provides avaluable opportunity to validate the ADCIRC SWAN andWAMSTWAVE models to measured data In the deepGulf Ike produced significant wave heights exceeding 15m that radiated from the stormrsquos center and transformedupon reaching the continental shelf At deep water NDBCstations WAM and SWAN performed comparably for allerror measures with the exception of mean wave directionfor which SWAN outperformed WAM As the swell propa-gates across the shelf SWAN shows a lag in the arrival of

Table 5 Summary of Scatter Index (SI) and Normalized Error Bias for Wave Data Time Seriesa

Data SourceGeographicLocation Model

NDBC WAMSTWAVE 1110b 01891 00840 1110b 01544 00373 109b 01204 00556 87b 04037 06475SWAN 01809 00465 01725 00456 01102 00385 02449 00177

CSI WAMSTWAVE 4 01951 02641 4 01384 00678 4 01939 03831SWAN 01416 01221 02002 01343 02200 03243

USACE-CHL WAMSTWAVE 4 04652 04632 4 09701 11156 4 15295 03000SWAN 05201 08589 04638 03024 18304 03125

AK WAMSTWAVE 8 02421 01727 8 01380 01597SWAN 02892 01807 02156 01497

Deep Water WAMSTWAVE 1110b 01891 00840 1110b 01544 00373 109b 01204 00556 87b 04037 06475SWAN 01809 00465 01725 00456 01102 00385 02449 00177

Coastal Water WAMSTWAVE 12 02264 02032 12 01382 01290 4 01939 03831SWAN 02400 01611 02105 01446 02200 03243

Inland WAMSTWAVE 4 04652 04632 4 09701 11156 4 15295 03000SWAN 05201 08589 04638 03024 18304 03125

All WAMSTWAVE 2726b 02466 01247 2726b 02680 02378 1817b 04499 00493 87b 04037 06475SWAN 02603 02244 02348 01308 05407 00231 02449 00177

aStatistics incorporate only stations that are shared between at least two model geographic coverages Bolded and italicized model name indicateswhich model was active within a data set Data sets are grouped geographically as follows Deep Water NDBC Coastal Water AK amp CSI and InlandUSACE-CHL

bOne station NDBC 42007 lies within both the WAM and STWAVE model domains Consequentially for WAMSTWAVE statistics an additionaldata set is used in the computation of statistics

Figure 26 Extent and elevation of storm event maximum water levels (m) on the LATEX Coast dur-ing Hurricane Ike

peak wave heights indicating a small artificial retardationof swell on the continental shelf This retardation has beenshown to be heavily dependent on bottom friction In thesensitivity tests it was determined that the Madsen formu-lation utilized by SWAN requires the minimum Manningrsquosn to be set equal to 002 avoiding unrealistically smallroughness lengths A minimum Manning n value gt002does not allow for accurate swell propagation Effectivewave breaking is seen in coastal marshes where waveheights are severely limited by bottom friction and shallowwater column depths Model performance in these shallowmarsh areas is in a relative sense worse than in deep andcoastal waters although the waves are small here and abso-lute error measures are correspondingly small Howeverthe errors do reflect the sensitivity of waves in shallow

waters to bottom friction and water levels A thorough ex-amination of bottom friction and its influence on wavemodels in shallow marsh regions would assist in betterparameterizing the role of bottom friction in nearshorephysical processes in wave models The SWAN modeloffers a multitude of bottom friction parameterizations ofwhich the Madsen formulation was selected for this studybased on previous success in validating Gulf of Mexicohurricanes [Dietrich et al 2011a 2012b] An in depthstudy of the modelrsquos sensitivity to other bottom frictionparameterizations may provide insight into better treatmentof bottom friction in complex wave environments such asthose seen in Ike [Kerr et al 2013a 2013b]

[65] As Ike progressed across the Gulf steady and mod-erate intensity winds over the Mississippi Chandeleur andBreton Sounds persisted for over 36 h These persistentwinds drove surge into the lakes bays and marshes sur-rounding New Orleans ADCIRCrsquos capture of the surgersquosgrowth peak and recession in this area indicates the mod-elrsquos data driven parameterization of air-sea drag and bottomfriction in marsh and wetland-type areas is accurate To thewest of the Mississippi River Birdrsquos Foot Delta ADCIRCaccurately captures the complex interaction of a strong sur-face water level gradient driven current shore normal winddriven surge and large wave breaking nearshore drivensetup

[66] The strong shore parallel current on the LATEXshelf produced by Ikersquos large wind field drove a forerunnersurge that increased water levels at the coast and intohydraulically connected inland lakes and bays well beforelandfall While this forerunner surge propagated to thesouthwest along the LATEX shelf and had little effect onthe peak surge in open waters in Louisiana and easternTexas it had a significant impact on high water marks con-tained in hydraulically connected inland water bodiesADCIRC captures this shelf scale process with a slightunder prediction in the early rise of water at the coast andconsequently water levels lag in the coastal lakes and baysThis slight underprediction appears to be associated with alow bias in the shore parallel winds in the region duringthis prelandfall phase of the storm Advection also plays a

Figure 27 Spatial analysis of high water marks in Louisiana and Texas Green represents points atwhich modeled water level was within 05 m of measured water level Redyellow represent points whereSWANthornADCIRC overpredicted water levels bluepurple represent points where SWANthornADCIRCunder-predicted water levels Coastline is in gray and SL18TX33 boundary and raised features in brown

Figure 28 Scatterplot of high water marks presented inFigure 27 Y axis is modeled HWMs plotted against meas-ured HWMs on the X axis

role in the forerunner generation as has been shown byKerr et al [2013a 2013b] Additionally a flow regime-based bottom friction similar to that applied in riverineenvironments in Martyr et al [2012] may further enhancethe early generation of the forerunner surge

[67] Following generation by shore parallel winds theforerunner surge propagated down the Texas shelf as a freecontinental shelf wave that reached Corpus Christi as Ikemade landfall This shelf wave is modeled by ADCIRCbut slightly lags in its propagation down the coast This islikely related to the slight low bias in the generation of theforerunner ADCIRC also accurately represents the peaksurge and positive inland water level gradient seen over theBolivar peninsula As Ike made landfall maximum windsimpacted inland lakes and bays that were still filled withadditional water from the forerunner surge An R2 value of091 when comparing all measured high water marks tomodeled high water marks indicates ADCIRCrsquos high levelof skill in modeling peak water levels Overall opencoastal water levels were better predicted than inland val-ues The large role that small-scale features and bottomfriction can play in the flooding and recession processesparticularly in complex coastal marshes and wetlands war-rants studies that further improve resolution in addition toimproved bottom and lateral friction parameterizations

[68] Diminishing winds following landfall allowed for therelease of water driven against the coast back toward thedeep Gulf When the mass of water pushed against the coastduring landfall was released by slowing winds it wasreflected by the abrupt bathymetric change at the continentalshelf break resulting in a cross-shelf resonant wave with aperiod of 12 h This cross-shelf resonant process is capturedin ADCIRC Surge driven into inland water bodies andcoastal wetlands experienced a prolonged recession processdominated by bottom friction that occurred over a much lon-ger time scale than the surge that developed in open water

[69] ADCIRCrsquos performance when compared to thewealth of data collected during the storm demonstrates this

modelrsquos ability to effectively model basin and regionalscale storm surge processes and the SL18TX33 computa-tional domainrsquos accurate portrayal of the complex LATEXshelf and coast This performance can be quantified by anoverall SI of 01463 and bias of 00114 m for 523 meas-ured water level time series and an estimated average abso-lute difference of 012 m for 599 measured high watermarks including 94 of modeled high water marks within050 m of the measured value (Figure 28 and Table 6)These qualitative assessments are similar to those found inother studies of Hurricane Ike utilizing ADCIRC and theSL18TX33 domain [Kerr et al 2013a 2013b]

[70] Acknowledgments This project was supported by NOAA viathe US IOOS Office (NA10NOS0120063 and NA11NOS0120141) andwas managed by the Southeastern Universities Research Association theUS Army Corps of Engineers New Orleans District the Department ofHomeland Security (2008-ST-061-ND0001) and the Federal EmergencyManagement Agency Region 6 The National Science Foundation (OCI-0746232) supported ADCIRC and SWAN model development Computa-tional facilities were provided by the US Army Engineer Research andDevelopment Center Department of Defense Supercomputing ResourceCenter The University of Texas at Austin Texas Advanced ComputingCenter The Extreme Science and Engineering Discovery Environment(XSEDE) which is supported by National Science Foundation grantOCI-1053575 Permission to publish this paper was granted by the Chief ofEngineers US Army Corps of Engineers The views and conclusions con-tained in this document are those of the authors and should not be inter-preted as necessarily representing the official policies either expressed orimplied of the US Department of Homeland Security The authors thankProfessor Chunyan Li and the Coastal Studies Institute at Louisiana StateUniversity for providing the CSI water level and current data

ReferencesAmante C and B W Eakins (2009) ETOPO1 1 arc-minute global relief

model Procedures data sources and analysis NOAA Tech Memo NES-DIS NGDC-24 19 pp Natl Geophys Data Cent Boulder Colo

Battjes J A and J P F M Janssen (1978) Energy loss and set-up due tobreaking of random waves in Proceedings of 16th International Confer-ence on Coastal Engineering Am Soc of Civ Eng HamburgGermany

Table 6 Summary of ADCIRC Water Levels for All Measured Water Level Dataa

Data SourceNumber of Timeseries Data Sets SI Bias

ADCIRC to Measured HWMs Measured HWMsEstimated ADCIRC

Errors

Number ofHWMs Slope R2

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

AK 8 02409 00887 8CSI 5 01771 00281 2NOAA 37 01137 00470 29 09710 09566 01458 01799USACE-CHL 6 02409 00517 5USACE 38 01111 00133 33 09896 08842 01575 02061USGS-Depl 50 01091 00413 40 10066 09704 01710 02098 00300 04898 01410 02040USGS-PERM 33 01086 01433 24 09329 09144 01941 01938CRMS 321 01651 00251 235 09727 06883 01785 02260 00491 01007 01293 02024TCOON 25 01080 00825 17 10016 09755 00988 01246FEMA 206 09850 08497 02845 03575 01249 02331 01596 02711Inland 448 01497 00235 543 09790 08186 01786 02250 00511 01093 01275 01967Coastal 75 01260 00606 56 10294 09426 01156 01457 00650 01216 00506 00802All 523 01463 00114 599 09844 09083 01727 02176 00524 01104 01203 01858

aScatter index and bias were calculated for time series only Average absolute difference and standard deviation have units of meters To accuratelydetermine error in measurements sets must contain at least 10 HWMs and be hydraulically connected and spatially limited Not all time series containedusable HWM data Inland data are defined by data sets USACE-CHL USACE USGS-Depl USGS-Perm and CRMS Coastal data are defined by datasets AK CSI NOAA and TCOON

Battjes J A and M J F Stive (1985) Calibration and verification of adissipation model for random breaking waves J Geophys Res 90(C5)9159ndash9167 doi101029JC090iC05p09159

Bender C J M Smith A B Kennedy and R Jensen (2013) STWAVEsimulation of Hurricane Ike Model results and comparison to dataCoastal Eng 73 58ndash70 doi101016jcoastaleng201210003

Berg R (2009) Tropical Cyclone Report Hurricane Ike 1ndash14 Sep pp55 NOAANational Hurricane Cent Miami Fla

Booij N R C Ris and L H Holthuijsen (1999) A third-generation wavemodel for coastal regions 1 Model description and validation J Geo-phys Res 104(C4) 7649ndash7666 doi10102998JC02622

Bretschneider C L H J Krock E Nakazaki and F M Casciano (1986)Roughness of typical Hawaiian terrain for tsunami run-up calculationsA users manual J K K Look Lab Rep Univ of Hawaii HonoluluHawaii

Buczkowski B J J A Reid C J Jenkins J M Reid S J Williams andJ G Flocks (2006) usSEABED Gulf of Mexico and Caribbean (PuertoRico and US Virgin Islands) offshore surficial sediment data releaseUS Geological Survey Data Series 146 version 10 US Geol SurvReston Va

Bunya S et al (2010) A high-resolution coupled riverine flow tidewind wind wave and storm surge model for southern Louisiana and Mis-sissippi Part I Model development and validation Mon Weather Rev138 345ndash377 doi1011752009MWR29061

Cardone V J and A T Cox (2009) Tropical cyclone wind field forcingfor surge models Critical issues and sensitivities Nat Hazards 51 29ndash47 doi101007s11069-009-9369-0

Cavaleri L and P M Rizzoli (1981) Wind wave prediction in shallowwater Theory and applications J Geophys Res 86(C11) 10961ndash10973 doi101029JC086iC11p10961

Cox A T J A Greenwood V J Cardone and V R Swail (1995) Aninteractive objective kinematic analysis system paper presented at theFourth International Workshop on Wave Hindcasting and ForecastingBanff Alberta

Dawson C J J Westerink J C Feyen and D Pothina (2006) Continu-ous discontinuous and coupled discontinuous-continuous Galerkin finiteelement methods for the shallow water equations Int J Numer Meth-ods Fluids 52(1) 63ndash88 doi101002fld1156

Dietrich J C et al (2010) A high-resolution coupled riverine flow tidewind wind wave and storm surge model for southern Louisiana and Mis-sissippi Part IImdashSynoptic description and analysis of HurricanesKatrina and Rita Mon Weather Rev 138(2) 378ndash404 doi1011752009MWR29071

Dietrich J C et al (2011a) Hurricane Gustav (2008) waves and stormsurge Hindcast synoptic analysis and validation in southern Louisi-ana Mon Weather Rev 139(8) 2488ndash2522 doi 1011752011MWR36111

Dietrich J C M Zijlema J J Westerink L H Holthuijsen C N Daw-son R A Luettich Jr R E Jensen J M Smith G S Stelling and GW Stone (2011b) Modeling hurricane waves and storm surge usingintegrally-coupled scalable computations Coastal Eng 58(1) 45ndash65doi101016jcoastaleng201008001

Dietrich J C et al (2012a) Limiters for spectral propagation velocities inSWAN Ocean Modell 70 85ndash102 doi101016jocemod201211005

Dietrich J C S Tanaka J J Westerink C N Dawson R A Luettich JrM Zijlema L H Holthuijsen J M Smith L G Westerink and H JWesterink (2012b) Performance of the unstructured-mesh SWANthornADCIRC model in computing hurricane waves and surge J Sci Com-put 52 468ndash497 doi101007s10915-011-9555-6

East J W M J Turco and R R Mason Jr (2008) Monitoring inlandstorm surge and flooding from Hurricane Ike in Texas and LouisianaSeptember 2008 US Geol Surv Open File Rep 2008-1365

Egbert G D A F Bennett and M G G Foreman (1994) TOPEXPOS-EIDON tides estimated using a global inverse model J Geophys Res99(C12) 24821ndash24852 doi10102994JC01894

Egbert G D and S Y Erofeeva (2002) Efficient inverse modeling ofbarotropic ocean tides J Atmos Oceanic Technol 19(2) 183ndash204doi1011751520-0426(2002)019lt0183EIMOBOgt20CO2

FEMA (2008) Texas Hurricane Ike rapid response coastal high water markcollection FEMA-1791-DR-Texas Washington D C

FEMA (2009) Louisiana Hurricane Ike coastal high water mark data col-lection FEMA-1792-DR-Louisiana Washington D C

Garster J K B Bergen and D Zilkoski (2007) Performance evaluationof the New Orleans and Southeast Louisiana hurricane protection sys-

tem Vol IImdashGeodetic vertical and water level datums Final Report ofthe Interagency Performance Evaluation Task Force US Army Corpsof Eng Washington D C 157 pp

Geurounther H (2005) WAM cycle 45 version 20 Inst for Coastal ResGKSS Res Cent Geesthacht Germany

Hanson J L B A Tracy H L Tolman and R D Scott (2009) Pacifichindcast performance of three numerical wave models J Atmos Oce-anic Technol 26(8) 1614ndash1633 doi1011752009JTECHO6501

Kennedy A B U Gravois B Zachry R A Luettich T Whipple RWeaver J Reynolds-Fleming Q J Chen and R Avissar (2010) Rap-idly installed temporary gauging for waves and surge during HurricaneGustav Cont Shelf Res 30(16) 1743ndash1752 doi 101016jcsr201007013

Kennedy A B U Gravois B C Zachry J J Westerink M E Hope JC Dietrich M D Powell A T Cox R A Luettich Jr and R G Dean(2011a) Origin of the Hurricane Ike forerunner surge Geophys ResLett 38 L08608 doi1010292011GL047090

Kennedy A B U U Gravois and B Zachry (2011b) Observations oflandfalling wave spectra during Hurricane Ike J Waterw Port CoastalOcean Eng 137(3) 142ndash145 doi101061(ASCE)WW1943ndash54600000081

Kerr P C et al (2013a) Surge generation mechanisms in the lower Mis-sissippi River and discharge dependency J Waterw Port Coastal OceanEng 139 326ndash335 doi101061(ASCE)WW1943ndash54600000185

Kolar R L J J Westerink M E Cantekin and C A Blain (1994)Aspects of nonlinear simulations using shallow water models based onthe wave continuity equations Comput Fluids 23(3) 523ndash538doi1010160045ndash7930(94)90017-5

Komen G L Cavaleri M Donelan K Hasselmann S Hasselmann andP A E M Janssen (1994) Dynamics and Modeling of Ocean WavesCambridge Univ Press New York

Komen G J K Hasselmannm and K Hasselmann (1984) On the existenceof a fully-developed wind-sea spectrum J Phys Oceanogr 14 1271ndash1285 doi1011751520-0485(1984)014lt1271OTEOAFgt20CO2

Madsen O S Y-K Poon and H C Graber (1988) Spectral wave attenu-ation by bottom friction Theory paper presented at the 21th Int ConfCoastal Engineering Torremolinos Spain

Martyr R C et al (2012) Simulating hurricane storm surge in the lowerMississippi River under varying flow conditions J Hydraul Eng 139492ndash501 doi101061(ASCE)HY1943ndash79000000699

Mitchell D A W J Teague E Jarosz and D W Wang (2005) Observedcurrents over the outer continental shelf during Hurricane Ivan GeophysRes Lett 32 L11610 doi1010292005GL023014

Mukai A J J Westerink R A Luettich and D J Mark (2002) A tidalconstituent database for the Western North Atlantic Ocean Gulf of Mex-ico and Caribbean Sea Tech Rep ERDCCHL TR-02ndash24 US ArmyEng Res and Dev Cent Vicksburg Miss

Powell M (2006) Drag coefficient distribution and wind speed depend-ence in tropical cyclones Final Report to the National Oceanic andAtmospheric Administration (NOAA) Joint Hurricane Testbed (JHT)Program Atl Oceanogr and Meteorol Lab Miami Fla

Powell M D and T A Reinhold (2007) Tropical cyclone destructivepotential by integrated kinetic energy Bull Am Meteorol Soc 88(4)513ndash526 doi101175BAMS-88-4-513

Powell M D S H Houston and T A Reinhold (1996) HurricaneAndrewrsquos landfall in South Florida Part I Standardizing measurementsfor documentation of surface wind fields Weather Forecast 11 304ndash328 doi1011751520-0434(1996)011lt0304HALISFgt20CO2

Powell M D S H Houston L R Amat and N Morrisseau-Leroy(1998) The HRD real-time hurricane wind analysis system J WindEng Ind Aerodyn 77ndash78 53ndash64 doi101016S0167ndash6105(98)00131-7

Powell M D P J Vickery and T A Reinhold (2003) Reduced dragcoefficient for high wind speeds in tropical cyclones Nature 422(6929)279ndash283 doi101038nature01481

Powell M D et al (2010) Reconstruction of Hurricane Katrinarsquos windfields for storm surge and wave hindcasting Ocean Eng 37 26ndash36doi101016joceaneng200908014

Ris R C N Booij and L H Holthuijsen (1999) A third-generation wavemodel for coastal regions 2 Verification J Geophys Res 104 7667ndash7681 doi1010291998JC900123

Rogers W E P A Hwang and D W Wang (2003) Investigation ofwave growth and decay in the SWAN model Three regional scaleapplications J Phys Oceanogr 33 366ndash389 doi1011751520-0485(2003)033lt0366IOWGADgt20CO2

Smith J M (2000) Benchmark tests of STWAVE paper presented at theSixth International Workshop on Wave Hindcasting and ForecastingMonterey Calif

Smith J M (2007) Full-plane STWAVE II Model overview ERDC TN-SWWRP-07-5 Vicksburg Miss

Smith J M A R Sherlock and D T Resio (2001) STWAVE Steady-state spectral wave model userrsquos manual for STWAVE version 30Tech Rep ERDCCHL SR-01-1 Eng Res and Dev Cent VicksburgMiss

Smith J M R E Jensen A B Kennedy J C Dietrich and J J Wester-ink (2010) Waves in wetlands Hurricane Gustav in Proceedings of the32nd International Conference on Coastal Engineering (ICCE) Shang-hai China

Sorensen R M (2006) Basic Coastal Engineering Springer N YSullivan P P L Romero J C McWilliams and W K Melville (2012)

Transient evolution of Langmuir turbulence in ocean boundary layersdriven by hurricane winds and waves J Phys Oceanogr 42 1959ndash1980 doi101175JPO-D-12ndash0251

Westerink J J R A Luettich J C Feyen J H Atkinson C Dawson HJ Roberts M D Powell J P Dunion E J Kubatko and H Pourtaheri(2008) A basin- to channel-scale unstructured grid hurricane stormsurge model applied to southern Louisiana Mon Weather Rev 136(3)833ndash864 doi1011752007MWR19461

Zijlema M (2010) Computation of wind-wave spectra in coastal waterswith SWAN on unstructured grids Coastal Eng 57(3) 267ndash277doi101016jcoastalengsss200910011

landfall Ike began to shift and track north-northwestwardthen making landfall at Galveston Island TX with a maxi-mum wind speed of 41 m s1 Ike generated a maximum

measured surge at landfall of 53 m in Chambers CountyTX located to the northeast of Galveston Island (Figure 1)[FEMA 2008] Across the LATEX coast Ike produced surge

Table 1 Summary of Significant Times and Characteristics of Hurricane Ikea

HoursRelativeto Landfall

UTCTime

UTC Date(2008) Latitude Longitude

Max WindVelocity

Radius toMaximum

Winds (km)

MinimumCentral

Pressure (mb)Saffir-Simpson

Category Notes

289 0600 1 Sep 172 37 13 167 1006 Trop Depression Formation217 0600 4 Sep 224 550 54 28 935 4 Maximum Intensity1945 0430 5 Sep 236 604 50 28 945 4 Enters SL18thornTX33

Domain187 1200 5 Sep 234 620 46 28 954 3 OWI winds start138 1300 7 Sep 210 732 49 947 3 Landfall on Great Inagua

Island Bahamas12475 0215 8 Sep 211 757 50 945 4 Landfall in Holguin Cuba89 1400 9 Sep 226 829 30 - 965 1 Landfall in Pinar del

Rio Cuba825 2030 9 Sep Enters Gulf of Mexico31 0000 12 Sep 261 900 37 148 954 219 1200 12 Sep 269 922 39 93 954 2 Peak in South Plaquemines13 1800 12 Sep 274 930 39 93 955 2 Shift in track WSE peak

in NOLA7 0000 13 Sep 283 941 41 74 952 2 WSE peak in Lake

Pontchartrain0 0700 13 Sep 293 947 41 950 2 Landfall at Galveston

Texas5 1200 13 Sep 303 952 37 56 959 111 1800 13 Sep 317 953 22 74 974 Trop Storm23 0600 14 Sep 355 937 15 93 986 Trop Depression OWI winds end53 1200 15 Sep End of simulation

aWinds are 10 min average [Berg 2009] (Automated Tropical Cyclone Forecast Archive ftpftpnhcnoaagovatcf)

Figure 1 Map of Northern Gulf of Mexico and Louisiana-Texas Coast The black line represents Ikersquostrack ADCIRC grid boundaries and raised features are brown the coastline is solid gray bathymetriccontours (as labeled) are dashed gray Geographic locations of significance are labeled by numbersidentified in Table 2

River and Channels

Data Type

Mean WaveDirection

Mean WavePeriod

Peak WavePeriod

2 Model Description

22 SL18TX33 Mesh

23 Winds

25 Bottom Friction

nffiffiffigp

26 Rivers

27 Tides

3 Recorded Data

41 Winds

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

River and Channels

Data Type

Mean WaveDirection

Mean WavePeriod

Peak WavePeriod

2 Model Description

22 SL18TX33 Mesh

23 Winds

25 Bottom Friction

nffiffiffigp

26 Rivers

27 Tides

3 Recorded Data

41 Winds

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

2 Model Description

22 SL18TX33 Mesh

23 Winds

25 Bottom Friction

nffiffiffigp

26 Rivers

27 Tides

3 Recorded Data

41 Winds

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

2 Model Description

22 SL18TX33 Mesh

23 Winds

25 Bottom Friction

nffiffiffigp

26 Rivers

27 Tides

3 Recorded Data

41 Winds

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

23 Winds

25 Bottom Friction

nffiffiffigp

26 Rivers

27 Tides

3 Recorded Data

41 Winds

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

25 Bottom Friction

nffiffiffigp

26 Rivers

27 Tides

3 Recorded Data

41 Winds

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

25 Bottom Friction

nffiffiffigp

26 Rivers

27 Tides

3 Recorded Data

41 Winds

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

42 Waves

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

SI frac14

ifrac141Ei E 2

ifrac141jOij

and normalized bias

bias frac141N

ifrac141Ei

ifrac141jOij

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Data Source Model

NDBC WAM 10 01893 00737 10 01571 00410 9 01248 00780 7 04379 07207STWAVE 1 01871 01870 1 01271 00011 1 00812 01463 1 01638 01348

WAMSTWAVE 11 01891 00840 11 01544 00373 10 01204 00556 8 04037 06475SWAN 13 02150 01031 13 02034 01265 13 01138 01177 9 02209 01364

AK STWAVE 8 02421 01727 8 01380 01597SWAN 8 02892 01807 8 02156 01497

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

5 Conclusions

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Errors

Avg AbsDiff

StdDev

Avg AbsDiff

StdDev

Avg AbsDiff Std Dev

Hindcast and validation of Hurricane Ike (2008) waves ... · 1] Hurricane Ike (2008) made landfall near Galveston, Texas, as a moderate intensity storm. Its large wind ﬁeld in conjunction

Documents

2009-02-17 Seminar - Hurricane Ike Galveston Lessons...

Hindcast and validation of Hurricane Ike (2008) waves...

Hurricane Ike Pics

The aftermath of Hurricane Ike

Hurricane Ike: 5 Years Later September 24 -25, 2013 SSPEED.....

Hurricane Ike Wind Investigation Report

Mitigation Assessment Team Report Hurricane Ike in … ·.....

Dynamical Seasonal Hurricane Hindcast Simulations

Mitigation Assessment Team Report Hurricane Ike in Texas...

Hurricane Ike 2008

Lessons From Hurricane Ike

JSC Hurricane Ike Lessons Learned Report - NASA

Hurricane Ike Impact Report - The Storm...

Contrasting Hurricane Ike washover sedimentation and ...

Hurricane Ike Slideshow

Hurricane Ike :