Storm Surge Computations for the North Carolina Sea ......A RENCI Technical Report Adapting Scienti˜c Work˚ows on Networked Clouds Using Proactive Introspection Storm Surge Computations

A R E N C I T e c h n i c a l R e p o r t

w w w . r e n c i . o r g

Adapting Scienti�c Work�ows onNetworked Clouds

Using Proactive Introspection

Storm Surge Computations for the North Carolina SeaLevel Rise Risk Management Study

TR-12-04

Brian Blanton, PhD Renaissance Computing Institute (RENCI)University of North Carolina at Chapel [email protected]

Prepared for:North Carolina Division of Emergency Management, Office of Geospatial and Technology Management

NCSeaLevelRiseRiskManagementStudy Page1

Preparedfor:

NORTHCAROLINADIVISIONOFEMERGENCYMANAGEMENTOFFICEOFGEOSPATIALANDTECHNOLOGYMANAGEMENT

StormSurgeComputationsfortheNorthCarolinaSeaLevelRiseRiskManagementStudy

Preparedby

BrianBlanton,Ph.D.

RenaissanceComputingInstitute

UniversityofNorthCarolinaatChapelHill

November12,2012

Version1.0


CONTENTS

Contents...................................................................................................................................................2

1. OverviewofADCIRCApplication.............................................................................................31.1. TheUseofADCIRCtoEvaluatetheCoastalFloodingHazardfortheNorthCarolinaSeaLevelRiseRiskManagementStudy...............................................................................................................................31.2. ReviewofMethodologyfortheNorthCarolinaFloodInsuranceStudy.......................................3

2. ADCIRC-relatedTASKS................................................................................................................42.1. Task1:SupportJPMSensitivityTest..................................................................................................42.2. Task2:WindFieldGeneration...........................................................................................................62.3. Task3:TidalDatumCode..................................................................................................................62.4. Task4:PerformBaselineSimulations................................................................................................6

2.4.1 Equilibriumtidalsolution.............................................................................................................62.4.2 TidalDatums................................................................................................................................62.4.3 SimulationofHistoricalHurricanes..............................................................................................82.4.4 Productionstormsurgesimulations..........................................................................................92.4.5 Statisticalsynthesis....................................................................................................................11

2.5. Task5:PerformSeaLevelRiseScenarioSimulations......................................................................122.5.1 20cmScenario:..........................................................................................................................132.5.2 40cmScenario:..........................................................................................................................142.5.3 60cmScenario:..........................................................................................................................152.5.4 80cmScenario:..........................................................................................................................162.5.5 100cmScenario:........................................................................................................................17

2.6. Task6:Storminessstatisticalanalyses.............................................................................................182.6.1:20cmSLRandMid21stCentury:..................................................................................................192.6.2:40cmSLRandMid21stCentury:..................................................................................................202.6.3:40cmSLRandEnd21stCentury:..................................................................................................212.6.4:60cmSLRandMid21stCentury:..................................................................................................222.6.5:60cmSLRandEnd21stCentury:..................................................................................................232.6.6:80cmSLRandMid21stCentury:..................................................................................................242.6.7:80cmSLRandEnd21stCentury:..................................................................................................252.6.8:100cmSLRandMid21stCentury:................................................................................................262.6.9:100cmSLRandEnd21stCentury:................................................................................................27

2.7. Overviewofstatisticalresults..........................................................................................................28

3. CapeFearwaterlevels.............................................................................................................303.1. IdealizedModelExperiments..........................................................................................................313.2. IdealizedModelResults...................................................................................................................343.3. ConnectionofIdealizedResultstoNC-SLRIS...................................................................................363.4. IdealizedResultsConclusion............................................................................................................40

4. References....................................................................................................................................41


1. OVERVIEWOFADCIRCAPPLICATION

1.1. TheUseofADCIRCtoEvaluatetheCoastalFloodingHazardfortheNorthCarolinaSeaLevelRiseRiskManagementStudy

Toaddress theevaluationof the coastal hazard component for theNorthCarolina Sea LevelRise Impacts Study (NC-SLRIS), the Renaissance Computing Institute (RENCI) proposed anapproachthatfollowsconceptuallytheapplicationofthetidalandstormsurgemodelADCIRCforcomputingfloodhazardlevelsfortherecentFloodInsuranceStudy(FIS)forNorthCarolina’scoastalcounties.TheFISapproachusesahigh-resolutionnumericalmodelgridforstormsurgeandwavesbasedonrecenttopographicsurveysandbest-availablebathymetricdata,aswellasadvanced statistical techniques for modeling North Carolina’s tropical storm climate. Theapproach addresses conduction and management of needed simulations on RENCI’s high-performancecomputers.

1.2. ReviewofMethodologyfortheNorthCarolinaFloodInsuranceStudy

RENCIrecentlycomputedfloodhazarddatafortheNorthCarolinaFloodplainMappingProgram(NCFMP)FloodInsuranceStudy(FIS)forcoastalcounties.Thecomputationalsystem(Blanton,2008)developedfortheFISusesstate-of-the-artnumericalmodels, includingthestormsurgeandtidalmodelADCIRC(Westerinketal,2008),andusescomputerresourcesatRENCIfortheactual computations. This systemhasbeen testedseveral times in thepast threeyears,andRENCIconsidersthissystemacceptablyrobustforthepurposesofNC-SLRIS.

The NC FIS project developed a comprehensive digital elevation model (DEM) using recentcoastalLIDARdataaswellasbest-availablebathymetricdata. ThisDEMwasusedtodevelopthe ADCIRC grid that is being used for the flood hazard simulations. Additionally, the JointProbabilities Method (JPM) approach to model the current tropical storm population (P.Vickery, Applied Research Associates) represented an advanced application of JPM tosubstantiallyreducethenumericalmodelcomputationalresourcerequirements.

IntheNCFISsystem,theparametricboundarylayermodelHBL(Vickeryetal,2009)modelsthetropical stormwind and pressure fields. Stormparameters forHBL are derived through theabove-mentionedJPMapproach.TheextratropicalcomponentismodeledwithanalyzedwindandpressurefieldsfromOceanWeather,Inc.

IntheNCSLRRMScontext,theprimarycomponentsoftheNCFISwillbeused.Thisincludesthe model ADCIRC, a comprehensive ADCIRC grid for the region, and a modification of thestorm tracks developed for the FIS, considered the baseline storm population for SLR RMS.Thesestorms,andmodificationsthatrepresentfuturestormclimates,aredescribedelsewhereintheTask2approach(ARA,PeterVickery).TheexistingextratropicalstormsetfortheFISwillbeusedforanyrequiredextratropicalsimulations.


2. ADCIRC-RELATEDTASKS

TheRENCI/UNCscopeconsistsof6 tasks,numbered the sameas in theRFDO. Task3 (TidalDatumCode)wasnotpartofthefinalRENCIproject;however,weretainthesamenumberingas in theRDFO for clarity. The specific tasks are as follows, eachofwhich is detailed in thesubsequentsectionsofthereport.Task1:SupportJPMSensitivityTestswithARATask2:WindFieldGenerationTask3:TidalDatumCodeTask4:PerformBaselineSimulations–present-dayADCIRCgridandstormclimateTask5:PerformFutureSLR/GeomorphologicalSimulations–ADCIRCgridsand/orincreasedmeansealevelthatcorrespondto20cm,40cm,60cm,80cm,and100cmscenarios.Task6:StorminessIncorporation

Tasks 1, 2, and 6 are to compute specific analyses and/or inputs to themain computationaltasks (4and5). Tasks4and5requiredthemajorityof theeffortandconsumedsubstantialcomputationalresources.

2.1. Task1:SupportJPMSensitivityTest

Task 1 of RENCI’s component of the NC-SLRRIS project provides computational support toAppliedResearchAssociates (ARA) todeterminea reduced setof storm tracks foruse in theBaselinecomputationofcoastalhazards.Tothisend,RENCIhasconductedADCIRCsimulationsonacoarseADCIRCgridusingstormtrackdataprovidedbyARA. RENCIprovidedthestormsurgeresultsbacktoARAforfurtheranalysis.

RENCIreceivedtwosetsofcandidatestormtracks from ARA: NC-Reduced-set-1-B-Caseand NC-Reduced-set-2-BandC-Case. Thesesets contain 450 and 294 tracks,respectively. Each set was converted fromtheARA“hur”formattotheADCIRCfort.22format, and the surge response wascomputed on a coarse grid that covers theNorth Carolina region and surroundingwaters(Figure1). Thisgridhas4636nodesand 8703 elements, does not include land(i.e.,doesnotsupportlandinundation),andexecutesveryfast inaserialcomputationalmode. Maximum surge levels were Figure 1: Coarse ADCIRC grid used for storm

trackselectionscreening.


recordedateachof117coastalnodesinNorthCarolina(indicatedbythereddotsinFigure1).Inadditiontothetworeducedtracksets,RENCIalsocomputedthecoarsegridsurgeresponsetothefullNCFIStrackset,whichcontains675tropicalstorms.

Each simulation computes amaximum storm surge, an example ofwhich is shown inError!Reference source not found. for a relatively intense landfalling storm (dp4r3b2c3h1l1). Foreach“scenario”orsetoftracks,themaximumsurgeresultsaregatheredandthereturnlevelsarecomputedusingtheJointProbabilitiesMethodasdescribedbyARA.Figure3showsthe1%water level at each coastal node in the coarse grid,with the numbering as indicated above.Changesincoastlineorientationandcontinentalshelfwidthareevident.Comparedtothe1%surgelevelscomputedwiththefullFIStrackset(NCFMP_Final_EQSpacing),thermsdifferenceforbothreducedsetsis11cm.ThisissummarizedintheError!Referencesourcenotfound..

Table 1: Basic statistics for comparison of surge results from the three track sets. Thecomparisonismadeforthe1%returnlevelatthecoastalnodes.

Rms min maxSet-1-B-Case .11 -.29 .27Set-2-BandC-Case .11 -.35 .25

Figure 2: Example of coarse grid surgeresponse to a landfalling storm. The stormtrack is shown with the black line, and thewaterlevelisshownincolorinmeters.

Figure 3: 1% Surge return level at coastalnodelocationsforthestormsurgecomputedfromthethreetracksets.Theabscissaisthenumbering of the coastal nodes (1-117) asindicated in Error! Reference source notfound..


2.2. Task2:WindFieldGenerationEachtropicalstormdefinedas inputtothecomputationalsystemrequireswindandpressurefieldstoforcethemodels.Inthistask,RENCIcomputedthewindandpressurefieldsassociatedwitheachofthe294tropicalstormsdefinedbyARA,usingARA’sHBLwindmodel.Thisstormsuitewasused for themaincomputational scenarios (baseline,20cm,40cm,60cm,80cm,and100cmADCIRCgrids).

2.3. Task3:TidalDatumCodePrior to initiation of the RENCI tasks, the explicit task to develop a tidal datum code wasremovedfromthescope.However,thisoriginaltaskismaintainedinthescopingandreportingdocumentstomaintainconsistentnumberingbetweenthetaskorder,proposals,andcontracts.Thetidaldatumcodewasdevelopedindependentof,andexternalto,theNC-SLRISproject.

2.4. Task4:PerformBaselineSimulationsAbaselinescenariowascomputedonthefirstprojectADCIRCgrid(Version4.3.2).Thisscenarioreflectsthepresent-daystormclimateandthegridisareducedversionofthelargerFISgrid.All simulations in thisandsubsequentSLR scenariosused thecoupledADCIRC+SWANmodel,version 49.60. For this present-day scenario, as well as the subsequent SLR scenarios, thefollowingsetofactivitieswasperformed. Resultsforthebaselineactivitiesareshowninthissection.ResultsfromtheSLRsimulationsareshowninSection2.5below.

2.4.1 EquilibriumtidalsolutionAnequilibriumtidalsolutionwascomputedonthescenariogrid,withthesametidalelevationboundary conditions thatwereused in theNC FIS tidal validation study (Egbert et al, 1994).The main model outputs from this simulation include the global elevation and velocityharmonic analysis files and a station velocity file. The stations for velocity output werespecifiedacrosseachtidalinletandrivermouth.Thisoutputwasusedbysubsequentprojectanalyses.Themainsimulationparametersareasfollows:

2.4.2 TidalDatumsTidaldatumswerethencomputedateachADCIRCgridnodefromtheglobalharmonicanalysisfileoftheequilibriumtidalsolution.ThisincludestheusualtidaldatumsofMeanHigherHighWater (MHHW),Mean HighWater (MHW),Mean LowWater (MLW), andMean Lower LowWater(MLLW),aswellascumulativedistributionsoftidalheightsneededforthesubsequentJPMandESTstatisticalanalyses.GIS-compatiblefilesforeachdatumwerealsoproduced.

TimeStep 0.5secRunLength 77.625days(150M2tidalcycles)RampupLength 10.35days(20M2tidalcycles)TidalConstituents M2,S2,N2,K2,K1,O1,P1,Q1HarmonicAnalysisPeriod Days10.35through77.625


Eachsurfaceisinitiallydefinedonlyoverwater,sincetheharmonicanalysisresultsaresensitivetothepercentoftimethatanode iswettedduringtheharmonicanalysisperiod. Inordertouse the tidal datums over land for the surge statistical analyses, the datum surfaces areextended inlandtocoverthearealextendofthesurgeresults. Thedatumsarecomputed inMSL and converted toNAVD88 as the last step in the analysis. Figure 4 showsMHHWandMLLWwaterlevelsaboveNAVD88forthebaselinescenario.

Figure4:Tidaldatumsfrombaselinescenarioequilibriumtidalsimulation.Left)MeanHigherHighWater(MHHW). Right)MeanLowerLowWater(MLLW). UnitsaremetersrelativetoNAVD88.The computed datum values were compared to the NOAA-published values (Table 2). TheNOAAvaluesarepublishedrelativetothestationdatum,sotheMSLlevelhasbeensubtractedfromtheNOAAvaluesforcomparisontothecomputeddatums,whicharerelativetoADCIRC’smeansealevel.Overall,thecomputedtidaldatumsarewithin2-6cmoftheobservedvalues.

Table2:TidalDatumComparisonforBaselineEquilibriumTidalSolution.UnitsaremetersrelativetoMSL.

NOAA ADCIRC NOAA ADCIRC MHHW-MSL MHHW MLLW-MSL MLLW

Duck,NC 0.58 0.59 -0.54 -0.51OregonInletMarina,NC 0.18 0.21 -0.18 -0.18

CapeHatterasFishingPier,NC 0.56 0.55 -0.49 -0.46Beaufort,NC 0.56 0.59 -0.52 -0.51

Wilmington,NC 0.68 0.72 -0.74 -0.69WrightsvilleBeach,NC 0.69 0.71 -0.62 -0.62

Southport,NC 0.74 0.74 -0.71 -0.66SunsetBeach,NC 0.87 0.83 -0.81 -0.76


2.4.3 SimulationofHistoricalHurricanesTwo historical events (Fran 1996; Isabel 2003) were simulated. These simulations includedtides,witha45-daytidalspinupperiodpriortotheonsetofthewindandpressureforcing,andstartingequilibriumadjustment factors for thespecific startdate. Themaximumwater levelandwaveheightswereretainedforvisualizationaswellasotherprojectanalyses.

Table3:Startandendtimesforthevalidationsimulationcomponents,andrunlengthsindays.

Storm TidalStartDate MetStartDate SimulationEndDate Metlength,TotalRunLength[days]

Fran(1996) 1996-07-1600Z 1996-08-3000Z 1996-09-0700Z 7.8,52.8

Isabel(2003) 2003-07-3100Z 2003-09-1400Z 2003-09-1912Z 5.5,50.5

The maximumwater levels from the two historical tropical storm simulations are shown inFigure5. HurricaneFranmaximumwater levelsare largestalongthecoast fromWrightsvilleBeachsouthwardtoFryingPanShoals,withthehighestlevelsreaching3.25meters.CapeFearRiverlevelsareabout1.0to1.25meters.HurricaneIsabelcausedlargerinlandeffects,mainlyinthePamlicoSoundarea,withthemaximumlater levelsgenerally inthelowerNeuseRiver.ThesesolutionswerecomparedtotheobservedhighwatermarksthatwereanalyzedfortheNCFIS.TheerrordistributionsareshowninFigure6.Bothsolutionsarerelativelyskillful,withrmserrorsof0.38and0.30meters,andareconsistentwiththevalidationresultsfromtheNCFIS.

Figure5:Historical simulationmaximumwater level for thebaseline (present-day)scenarioforHurricanesFran(1996)andIsabel(2003).Thestormtracksareshownwiththeblackline.WaterlevelunitsaremetersNAVD88.


2.4.4 ProductionstormsurgesimulationsTheproductionstormsurgesimulationsare themaincomputational step in thebaselineandSLRscenario.ThisstepcarriedouttheproductiontropicalandextratropicalstormsimulationsrequiredfortheESTandJPMreturnperiodanalyses.Generally,theextratropicalstormswererunfirstinorderfortheESTstatisticalprocessingtooccurconcurrentlywiththetropicalstormsimulations.Thecombinednumberofstormsimulations(294tropical+21extratropical=315)tookabout4weeks to compute. Anoverviewof theproduction simulation results is shownnext. Therankedstormsurgevaluesareshown inFigure7 for threecoastal locations. Bothextratropicalandtropicalstormsareincluded.Generally,theextratropicalstormsurgesareoflarger magnitude north of Cape Hatteras (Duck). The 1993 extratropical storm causedsignificantsurgesintheCapeFearRiver,seenasthegreencircleataboutstormnumber275.Themaximumofall stormmaxima is shown inFigure8. The largestvaluesexceed5metersalongthecoast in theWrightsvilleBeacharea,with3+meters in the lowerCapeFear,upperNeuse,NewandWhiteRivers. Since theseare themaximumwater levels reachedacrossallstorms,theyrepresentannualoccurrencelevelsnear0.1%(1in1000years).

Figure 6: Error distribution for historical baseline simulations. Left) Hurricane Fran (1996).Thermserroris0.38m.Right)HurricaneIsabel(2003).Thermserroris0.30meters.

00.050.1

0.150.2

0.25

-1.00

-0.80

-0.60

-0.40

-0.20

0.00

0.20

0.40

0.60

0.80

1.00

Relativ

eFreq

uency

Error[m]

FranBaselineError

00.050.1

0.150.2

0.25

-1.00

-0.80

-0.60

-0.40

-0.20

0.00

0.20

0.40

0.60

0.80

1.00

Relativ

eFreq

uency

Error[m]

IsabelBaselineError


Figure 7: Ranked storm surges for three coastal locations for the baseline productionsimulations.Circlesmarktheextratropicalsimulationresults.Thereare315totalstormsinthestormpopulation.

Figure8:MaximumateachADCIRCnodeacrossallbaselinesimulations.Unitsaremeters

NAVD88.


2.4.5 StatisticalsynthesisThefinalstepisthestatisticalsynthesisofthetropicalandextratropicalresults,wheretheJPMprobabilities are used with the tropical production simulation results to compute the waterlevels associated with the 10%, 4%, 2%, 1%, .2%, and .1% annual chances of occurrence.Likewise,theEmpiricalSimulationTechnique(EST) isusedtoderivereturn levelsatthesamefrequencies. All EST analyses were carried out at Dewberry. The final return levels are astatisticalcombinationofthe(independent)JPMandESTresults. AllstatisticalmethodsusedareconsistentwiththeNCFIS,andeach(JPM,EST,combinations,andincorporationoftides)arefullydescribedinpriorprojectdocuments,theNCFISIDS#3document,aswellastheESTUser’s Guide (Scheffner et al, 1999). The final data set from this step is the combined(JPM+EST)returnlevelsforthe10%,4%,2%,1%,.2%,and.1%annualchancesofoccurrence.SummaryresultsareshowninFigure9.Theleftpanelshowsthe1%waterlevelfortheNorthCarolinacoastalwaters.Thecolorscalemaximumissetto5metersforeasiercomparisontowater levels for theSLRscenariosdescribedbelow. Water levelsarehighestalong theopencoast in theWrightsville Beach area and lowest in the sounds. The Neuse River levels arehigherthanthesoundlevels,consistentwiththeFISresults.TherightpanelshowstheannualchancelevelsatDuck,WrightsvilleBeach,andWilmingtonnodes.

Figure9:Left)1%waterlevel(mNAVD88)forthebaselinescenario.Right)Waterlevelsatstandardoccurrencefrequenciesatthreelocations.


2.5. Task5:PerformSeaLevelRiseScenarioSimulationsThe sea level rise scenarios were computed in the same manner as the baseline (Task 4)scenario. In addition to the equilibrium tides, historical storms, production simulations, andstatistical synthesis, one additional step was carried out. In order to determinegeomorphologicalchangesforascenario’sADCIRCgrid,RENCIconductedanequilibriumtidalsimulationonthepreviousscenario’sgridbutwiththecurrentscenario’sSLRamount;inotherwords, the effect of the SLR amount on the tides on the previous geomorphologicalconfiguration/grid. Tidal datums and tidal velocities across the region’s tidal inlets werecomputedandprovidedtoDewberryforfurtherapplicationtothescenario’sADCIRCgrid.Each project sea level rise amount (20, 40, 60, 80, 100 cm) represents a “loop” through thebaselineprocedure. Themaximumstormsurgesurfaces for thehistoricalstormsareshownfirst, and then 1% surface and the return levels at the three selected coastal locations areshowninFigure10throughFigure14.The historical storm results show the generally expected behavior of increased water levelsalongtheopencoastintheareaswherethestormshadthelargestimpacts.ThesurgeimpactsinthelandareabetweenPamlicoandAlbemarleSoundsbecomesincreasinglyfloodedasMSLincreases,withwaterlevelsreaching1.5metersforFran(1996).Forexample,thisareafloodsinthe100cmcaseandhaswater levelsataboutthe1%level (comparetheFranwater leveland1%waterlevelinFigure14).Thisareadoesnotfloodinthebaselinethrough40cmcases.Open-coastwater levels tend to increaseatamountsconsistentwith the increasedsea level.Physically,thisisbecausetheaveragewaterdepthandcontinentalshelfwidthdonotchangedappreciably as themean sea level increases. Since themeandepth and shelfwidth are theprimaryfactorsthatcontrolregionalsurgeresponse,andthereislittlechangeisthedepthandshelf width, surge increases essentially linearly with the mean sea level increase. In thesheltered waters, however, the simple linear response seen along the open coast is notexpected,generallyduetononlineareffects.Particularly,thearealextentoflandbelowmeansea level increasesrapidlyasmeansea level increases,andthusthe lower-lying, flatterareas(e.g., land west of Pamlico Sound) exhibit substantially increased inundation in the end-memberscenario(100cm).


2.5.1 20cmScenario:

Figure10:20cmscenariohistoricalstormmaximumsurge,1%combinedwaterlevel,andreturnlevelsforthethreeselectedstationlocations.


2.5.2 40cmScenario:



2.5.3 60cmScenario:

Figure12:60cmscenariohistoricalstormmaximumsurge,1%combinedwater level,andreturn levelsforthethreeselectedstationlocations.


2.5.4 80cmScenario:



2.5.5 100cmScenario:



2.6. Task6:StorminessstatisticalanalysesTheeffectsofpossiblechangesintropicalcycloneareincorporatedintothescenariostatisticalsynthesesbyreweightingtheJPMstormweights.Noadditionalordifferentstormsimulations(eitherwind/pressureorADCIRCsimulations)wererequired. Thereweightingprocedureandresultsisdescribedelsewhereintheprojectdocumentation.ARAprovidedtwosetsofrevisedstormweights; SetA for themid 21st century, and Set B for the end 21st century. For eachcombinationofSLRscenarioandstorminesssetinTable4,numbered1-9,theJPMreturnlevelswererecomputedusingtherevisedstormweightsandthepreviouslycomputedsurgeresultsfromtheSLRscenario. Results forthe1%surfaceandthereturn levelsatthethreeselectedcoastal locationsareshownbelow,foreachofthestorminessscenarios, inFigure15throughFigure23.

Table4:Storminessandscenariocombinations. StorminessApplication

Value,m TimeFrame a b0 2010-2100 0.1 2050-2100 0.2 2050-2100 1 0.3 2075-2100 0.4 2100 2 30.6 2100 4 50.8 2100 6 71 2100 8 9

StormSeta: Mid21stcenturyStormSetb: Endof21stcentury


2.6.1:20cmSLRandMid21stCentury:

Figure15:1%water leveland return levels for storminess combination1of20 cmSLRandmid-centurystormclimate.


2.6.3:40cmSLRandEnd21stCentury:

Figure17:1%waterlevelandreturnlevelsforstorminesscombination3of40cmSLRand

end-centurystormclimate.



Figure18:1%waterlevelandreturnlevelsforstorminesscombination4of60cmSLRand

mid-centurystormclimate.



Figure19:1%water leveland return levels for storminess combination5of60 cmSLRandend-centurystormclimate.



Figure21:1%water leveland return levels for storminess combination7of80 cmSLRandend-centurystormclimate.



Figure22:1%waterlevelandreturnlevelsforstorminesscombination8of100cmSLRandmid-centurystormclimate.



Figure23:1%waterlevelandreturnlevelsforstorminesscombination9of100cmSLRandend-centurystormclimate.


2.7. OverviewofstatisticalresultsThe following figures show the return levels for the different scenarios for each stationseparately. TheDuck andWrightsville locations are representativeof theopen coasts northandsouthofCapeHatteras, respectively. For theSLRscenarios (Figure24), return levelsaregenerallylower,foraspecificrecurrencefrequency,alongtheOuterBanks,andtheincreaseinwaterlevelatlowerfrequenciesisless.Thisisdueprimarilytotherelativelysmallerimpactofthetropicalstormsurgesresultsinthecombinedwaterlevels.Essentially,theoppositeoccursinthelowerpartoftheNCcoastalwaters;waterlevelsatspecificfrequenciesarelargerthanthose in theOuter Banks, and levels increasemore as the frequencydecreases. ThewaterlevelbehavioratWilmingtoniscomplexanddiscussedfurtherinSection3.

Figure24:SLRScenarioreturnlevelsatthethreemodelnodesforDuck,WrightsvilleBeach,andWilmington.


Figure 25 shows the return levels for the storminess scenarios. Generally, the qualitativeresults are similar to the SLR scenarios in terms of the behavior north and south of CapeHatteras,aswellasintheupperCapeFear.TheimpactsofthestorminessJPMmodificationsare most notable at the Wrightsville Beach location, where the return levels at lowerfrequenciesarehigherthanthoseforthepresent-daystormclimate(asshowninFigure25).

Figure25:StorminessScenarioreturnlevelsatthethreemodelnodesforDuck,WrightsvilleBeach,andWilmington.


3. CAPEFEARWATERLEVELS

During analyses of theNC-SLRISmodel-generated products for the statistical flood elevationsurfaces,itisnoticedthat,intheWilmingtonarea,SWELvaluesdonotincreaseinproportionto the increasedMSL increment imposed on the scenario. This area is of particular concerngiventheplacementofaNOAAtidegaugeneartheturningbasinandthedataanalysisofthewaterlevelrecordinthecontextofchanneldredgingactivitiesandtidalamplitudes.At face value, smaller increases in SWEL than expected seem problematic and concern wasplacedonRENCI’snumericalmodelresults.ThisunexpectedbehaviorisshowninFigure26fortheJPMresults,alongapproximatelyequidistantpointsfromtherivermouth(0km)toabout30 km upstream of Wilmington, taking the westward channel branch to the west of EagleIsland.Theupperplot shows the1%SWELvaluesalong theCapeFearRiver including tidesand theerrortermfromtheNCFISproject.ThelowerplotshowstheJPMlevelswithoutthetidalanderror contributions. Considering first the JPM+Tides+Error results, the1%values increasebyaboutthesea level increaseattherivermouth. Asthescenario“increases”,thewater levelsincrease in the lower river region,butdonot increaseasmuchasexpectedupstreamof theWilmingtonarea(atabout45kmfromtherivermouth).Additionally,Itappearsthatthe40cmwater levels are actually larger than the 3 later scenarios, at least in a limited region nearWilmington.

Figure26:Along-channel1%JPMwaterlevelsfortheNCSLRISscenarios.Top)1%JPMlevelswithtidesanderrorterm.Bottom)1%JPMlevelswithsurge-only.


Thelowerplot(JPM,notides,noerror)indicatesthat,evenwithoutthetides,thefloodwaterlevelsdonotbehaveinastrict“bathtub”sense.Inresponsetothisconcern,RENCIhasscrutinizedthemodeloutputanddeterminedthatthemodelresponsesarewellwithinexpectations,giventheextentofgeomorphologicalchangestothe underlying scenario topography and bathymetry, which determines the ADCIRC gridelevations. Additionally, the unexpected behavior of SWEL levels can be explained byconsideringthetidalresponsetoariverchannelthat1)stronglyconverges intermsofcross-channelwidthanddepthconvergingand2)thathasasubstantial intertidalstorageareathatalso increases as sea level increases. This storage volume (defined as the volume of watercontainedbetweenthechanneltopo/bathyandtheMHWandMLWwatersurfaces)generallyincreases as sea level increases. However, the storage volume may change due tomorphological considerations that are themselves a complicated response of increased sealevels,modifiedtidalregime,etc.Ofcourse,inADCIRC,thetopo/bathyisstatic,meaningthatforagiven scenarioand tidal regime, theeffectsof the sea level increasecanbe reasonablyunderstood.Apurely linear,“bathtub”perspectivemotivatestheexpectationthatSWEL,tidalamplitudes,and tidal datums increase relatively monotonically with the imposed static sea levelincrement. This would be largely correct if geometric considerations of the river could beignored and (more importantly) if therewere no imposed geomorphological changes on theriverbasin. In the limitofachannelwithverticalwalls (i.e.,no floodplain), thenthebathtubexpectationisessentiallycorrect.Foradetailedanalysisofidealizedchannelresponsetotidalforcingundersealevelincreases,seeFriedrichs,Aubrey,andSpeer(1990). In rivers with complex geometries, the tidal water level response a sequence of sea levelincreases depends critically on how the intertidal storage volume increases. As this volumeincreases, tidal amplitudes may decrease as the incoming tidal wave is reflected less. Theconsequenceofdecreasingreflectionisthattheprogressivecharacterofthewaveincreases,inresponse to increased conveyance upstream to fill the increased volume. If the volumeincreaseissmall,thentheincreasedmassfluxneededtofillthatvolumeisalsorelativelysmalland tidal amplitudeswill decrease a small amount. However, if the volume change is large,thenthetidepropagationcharacteristicscanchangesubstantially.

3.1. IdealizedModelExperimentsTo illustrate this dependency, and in addition to the engineering and coastal oceanographicliterature,wehaveconductedasequenceoftidalsimulationswith idealizedrivers. ThebasicCapeFeargeometriccharacteristicsareusedtodefine thealong-channeldependenceonthecross-channelwidthanddepthprofiles.Inparticular,cross-channeldepthprofilesweretakenupstreamintheNCSLRISBaselineand100cmgridsandusedtodefinethetopo/bathy intheidealized grids. Themouth of the Cape Fear River is about 1 kmwide, but the river widthexpandssubstantiallyimmediatelyupstream.Thisisindicatedinthegridfiguresbelowasthewideningoftheriverbetween0and3km.


The idealized river is 70 km longwith a constriction in channel and bankwidth at about 40km. Thisplaces“Wilmington”atabout45kmupstream. Thisbaselinegeometry is shown inFigure27.

Figure27:Idealizedbaselinegridgeometry.Left)Planviewofgrid,with2,0,and-10metergrid elevation contours. The thick black line is the grid’s exterior boundary. Right) Cross-channeldepthsat10kmintervalsalong-channel.Theprofileat50km(blacksquares)isthenarrowestpartofthechannel.

To represent the end-member SLR scenario of a substantially altered storage volume in theupper(northofWilmington)CapeFearRiver,thebaselinegriddepthprofilesaremodifiedtobroadenANDdeepenthestorageupstreamofabout50km.ThisgridisshowninFigure28.Inadditiontoa1-meterincreaseinmeansealevel,theupperbasinisdeepenedtoincreasethevolume.Eachconfigurationincludesadeepcentralchannelwithvariable-widthalong-channelfloodplainbanks.


Figure28: Idealizedgridgeometrywithincreasedstoragecapacityontheupperbasin.Left)Planviewofgrid,with2,0,and-10metergridelevationcontours.Thethickblacklineisthegrid’sexteriorboundary.Right)Cross-channeldepthsat10kmintervalsalong-channel.Theprofileat50km(blacksquares)isthenarrowestpartofthechannel.

Threesimulationswereconductedtoshedlightonthetidaldynamicsinchannelswithvaryingtidalstoragevolumesinanupperbasin:

1. Baselinegeometry(Figure27)withnosealevelincrease.2. Baselinegeometrywitha1-metersealevelincrease-ThisisthegridinFigure27with

adatumoffsetof+1meter.3. Deeperupperbasingeometrywitha1-metersealevelincrease(Figure28).

Simulationparametersinclude:1. 10dayslongwitha2dayrampup.2. Harmonicanalysisoverdays6-10.3. ElevationboundaryconditionofanM2tidewith1-meteramplitudeand0phase.4. Norotation.5. ConstantManning’sN.


3.2. IdealizedModelResultsThenatureofthetidesintheseidealizedscenariosisinvestigatedbylookingattheprimarytideelevation response in the along-channel direction. Figure 29 shows the along-channel M2elevationamplitude(black)andphase(red)forthethreescenarios.Forallscenarios,thetidalamplitudedropsslightlyduetotheexpansionoftheriverwidthimmediatelyupstreamoftheopenboundary. Case1and2 (bothon thebaselinegridgeometry)amplify slightly fromtheminimum amplitude at 10 km. However, the case1 amplitude increases relative to theboundary condition by about 10%. The distance at which the amplitudes begin to dropsubstantiallyismostupstreamincaseone,atabout55km.Case3(deeperupperbasinAND1mSLR)exhibits this characteratabout45km,which is in thevicinityof “Wilmington” in theidealizedgeometry.

Figure29:Along-channelM2Elevationamplitudeandphaseforthe3idealizedcases.

In the case 1 scenario, the tide amplifies due to the convergent natureof the channel. ThisamplificationdecreasesastheaveragewaterdepthincreasesANDtheriverwidthswiden(lessconvergent). The M2 elevation phase increases upstream (later arrivals) and undergoessubstantial lags upstream of about 50 km. The phases are all about 60 deg in the idealizedWilmington area, but as the water deepens and storage volume increases, the tide arrivesearlierintheupperbasinarea.


Next,weconsiderthealong-channeltidalvelocitycomponent.ThisisshowninFigure30.Thevelocity amplitudes are all relatively smaller and relatively the same in the lower riverarea.However,justdownstreamofthechannelconstriction,thevelocityincreasesduetotheconvergingchannel.Forthebaselinegrid(case1and2),theupstreamvelocitypeaksatabout1.35m/s,withthemaximumoccurringmoreupstreamincase1.Theabruptchangesinspeedare due to the transition of the channel depths. In case 3 (deeper upper basinwith 1m SLincrease),thepeakvelocityexceedsthatattheopenboundaryandoccursmoredownstreamrelativetothebaselinegridcases.Themorerapiddropinspeedisduetotherapidlyincreasingstoragearea.

Figure30:Along-channelM2velocityamplitudeandphaseforthe3idealizedcases.

While the elevation and velocity amplitude and phase are illustrative of the impacts ofincreased storage volumes in the upper river basin, they themselves do not reflect thetransition of the incoming tidewave frommore standing (case 1) tomore progressive (case3).Recallthatforastandingwave,thewaterlevelandvelocityareoutofphaseby90degrees,andforapurelyprogressivewavetheelevationandvelocityareexactly inphase. Soit isthephasedifferencebetweenelevationandvelocitythatismostrelevanttothenatureofthetidalwave.Figure31showsthisdifferenceforthe3idealizedcases.Inthebaselinecase1,thetidebecomes more standing in character (relative to the open boundary phase difference) andshifts toward a mixture of standing and progressive upstream of the narrowest part of thechannel.Thesmallestphasedifferenceisabout30degrees.Thecase2(baseline+1m)phase


difference is relatively similar tocase1,withan“earlier”minimumnear50kmand lessofadifferenceinthelowerchannelarea.Theminimumisstillabout30degrees.However, in case 3, the character is substantially different from the baseline cases. Thepartitioningofthewavebetweenstandingandprogressivealmostimmediatelymovestowardsprogressive,withthewavebecomingalmostentirelyprogressiveinthe“Wilmington”area,anda“rapid”transitiontowardalmostfullystandinginthedeeperupperbasin. Ofcourse,atthehead of the channel in all cases, there is significant reflection and hence an almost entirelystandingwavephasedifferencenear90deg.

Figure31: Along-channelphasedifferencebetweenelevationandthealong-channelvelocityfortheM2tide

3.3. ConnectionofIdealizedResultstoNC-SLRISIntheNC-SLRISproject,asequenceofSLRamounts isaddedtotheBaselinegridoragridtowhich some level of geomorphological adjustments have been imposed. The 100cm end-memberscenariohasasubstantiallyincreasedstoragevolume,ascomparedtojustaddingsealevel increases to the baseline geometry. This is illustrated in Figure 32, which shows thechangeinstoragevolumewithincreasingSLforthe6projectscenarios.Theredlineisthetidalstoragevolume(incubickm)belowthescenario’sMSL,using thegrid for thatscenario. Thegreenlineisthevolumebelowmeanhighwater.Thevolumeincreasesarelarge,witha5-foldincreaseinthestoragevolumebelowMSL.Forcomparison,thebluelineisthevolumebelowMSLforeachscenariobutcomputedrelativetothebaselinegrid. Noteaslightlyexponential


increaseassea level increases,anexpectedfeatureofvolume increasewith increasingwaterlevelsinbasinswithgentlyslopingsidewalls.

Figure32: TidalstoragevolumesintheCapeFearRiver,upstreamofClarksadEagleIslands,fortheprojectscenarios.Thebluelineindicatesstoragecomputedwiththesealevelincreasebutwiththebaselinegridtopography/bathymetry.

The impactof tidal storagechanges (increases)on theM2 tidepropagatingup theCapeFearRiverisshowninFigure33.Thegeneralcharacteristicsareverysimilartotheidealizedresultsdiscussed above. For the scenarios that are close to the Baseline (00 cm, 20 cm, 40 cm), intermsofupperbasinstoragevolume, there isa relativelysmallchange in thetidalelevation,althoughthereisaslightamplitudeincreaseduetoslightlydeeperwaterandlittlegeometricchangeupstreamofWilmington. However, as the storage volume increases (red line, Figure32) upstream ofWilmington, tidal amplitudes drop and the phase is lagged. The amplitudedifferencebetweenthescenarioend-membersisabout60cm.


Figure33:Along-channelM2tidalelevationamplitudeandphaseintheCapeFearRiver.

The phase angle between the elevation and velocity, indicative of the mixture betweenstanding(toward90deg)andprogressive(toward0deg),isshowninFigure34fortheBaselineand100cmscenarios.Duetotherealisticandmorecomplexriver/channelcharacteristics,thevelocity exhibits much more variability than in the idealized cases. Nonetheless, the phaseangle for the 100 cm scenario is substantially more toward 0 (progressive, in phase) in thelower part of the river than in the Baseline scenario. The baseline scenario is generally anapproximatebalancebetweenstandingandprogressiveandoveralldoesnotchangeasmuchasinthe100cmscenario.


Figure34:Along-channelphasedifferencebetweenM2tidalelevationandvelocity.OnlytheBaseline(blue)and100cm(red)scenariosareshown.

Thestoragevolumeincreasesalsoaffectstormsurgesinasimilarway,althoughthediagnosticquantities like phase angle not meaningful in this context. To illustrate the impact of thestoragevolumeonthesimulatedwaterlevels,thealong-channelmaximumwaterlevelforthe1993 extratropical storm (Storm of the Century) is shown in Error! Reference source notfound.Figure35.ThewaterlevelresponseintheBaselinescenarioisthatthesurgeisamplifiedup the channel to the Wilmington area, upstream of which there is a reduction of waterlevel.Thischaracteristicisgenerallythecaseforthefirst3scenarios.However,asthestoragevolumeincreases,twothingshappen.1)WatervolumeflowspasttheWilmingtonarea,fillinginthestorageincrease,andraisingwaterlevelsintheupperbasin.2)This volume of water is no longer stored in the lower part of the river, as indicated by thegenerallyflatwaterlevelinthe80and100cmscenarios.Inotherwords,thewaterthatwouldotherwise fill in the lower river floodplain is, in later scenarios, being driven into the largerbasinsupstreamofWilmington.


Figure35:Along-channelwaterlevelfortheextra-tropicalstorm1993.

3.4. IdealizedResultsConclusion Theimplicationsofthisdecreasedtidalamplitudewithincreasingstoragevolumearethatthetidal contributions to the statistical flood levelswill be less in “later” scenarios, inwhich thelarger storage volumes occur. In these idealized experiments the tidal amplitude at“Wilmington”decreasesfromabout1.1mto0.6m, implyingthat if thenon-tidalSWEL levelsincreaseby1meterbetweentheNCSLRISend-members,thenitcanbeexpectedthatthetotalcombinedSWEL+tideswillnotincreaseatthesamelevelasthesealevel increase. Itseemsunlikelythateffectsofchanneldredgingcanhaveanappreciableimpactonthetidalandsurgedynamics in the Cape Fear River system, although local (to the gauged location) effects areprobable.


4. REFERENCES

Blanton, B.O., 2008, North Carolina Coastal Flood Analysis System: Computational System.TechnicalReportTR-08-04,RENCI,NorthCarolina,September2008.

Egbert,G.D.,Bennett,A.,andForeman,M.G.G.(1994).TOPEX/Poseidontidesestimatedusing

aglobalinversemodel.J.Geophys.Res.,99:24821–24852.Friedrichs, C. T.,D. G. Aubrey, andP. E. Speer(1990), Impacts of relative sea-level rise on

evolution of shallow estuaries, inResidual Currents and Long-Term Transport,CoastalEstuarine Stud., vol. 38, edited byR. T. Cheng; 122, AGU, Washington, D. C.,doi:10.1029/CE038p0105.

Scheffner, N.W., L.E. Borgman, and D.J. Mark, 1996, Empirical simulation technique based

storm surge frequency analyses, Journal ofWaterway, Port, Coastal and Ocean, Vol.122,No.2,93-101.

Scheffner,N.W.,Clausner,J.E.,Militello,A,Borgman,L.E.,Edge,B.L.,andGrace,P.E.,1999.Use

andapplicationoftheempiricalsimulationtechnique: User’sGuide. TechnicalReportCHL-99-21. U.S. Army Engineer Research and Development Center, Coastal andHydraulicsLaboratory,Vicksburg,MS,194p.

Vickery, P.J., D.Wadhera,M.D. Powell and Y. Chen, 2009, A Hurricane Boundary Layer and

WindFieldModel forUse inEngineeringApplications, JournalofAppliedMeteorologyandClimatology,Vol.48,381-405,DOI:10.1175/2008JAMC1841.1.

Vickery,P.J. andB.O.Blanton,2008,NorthCarolinaCoastal FloodAnalysisSystem:Hurricane

ParameterDevelopment.TechnicalReportTR-08-06,RENCI,NorthCarolina,September2008.

Westerink,J.J.,R.A.Luettich,Jr.,J.C.Feyen,J.H.Atkinson,C.Dawson,M.D.Powell,J.P.Dunion,

H.J.Roberts,E.J.Kubatko,H.Pourtaheri,2008.“ABasin-toChannel-ScaleUnstructuredGrid Hurricane Storm SurgeModel as Implemented for Southern Louisiana”,MonthlyWeatherReview,136:833-864,DOI:10.1175/2007MWR1946.1.

StormSurge-tech-report-2012NCSLRIS_Adcirc_FinalReport

Documents

Storm Surge Computations for the North Carolina Sea ......A RENCI Technical Report Adapting Scienti˜c Work˚ows on Networked Clouds Using Proactive Introspection Storm Surge Computations