UBC Faculty Research and Publications

Generation of internal waves on the continental shelf by Hurricane Andrew. Keen, Timothy R.; Allen, Susan E. Nov 30, 2000

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-Alllen_AGU_2000JC900137.pdf [ 1.87MB ]
JSON: 52383-1.0041904.json
JSON-LD: 52383-1.0041904-ld.json
RDF/XML (Pretty): 52383-1.0041904-rdf.xml
RDF/JSON: 52383-1.0041904-rdf.json
Turtle: 52383-1.0041904-turtle.txt
N-Triples: 52383-1.0041904-rdf-ntriples.txt
Original Record: 52383-1.0041904-source.json
Full Text

Full Text

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 105,NO. Cll, PAGES 26,203-26,224,NOVEMBER 15, 2000  The generationof internal waveson the continentalshelfby Hurricane  Andrew  TimothyR. Keen NavalResearch Laboratory, Oceanography Division,Stennis SpaceCenter,Mississippi  Susan E. Allen  Department ofEarthandOceanSciences, University of BritishColumbia, Vancouver, Canada  ABSTRACT. Observed currents, temperature, andsalinityfrommooredinstruments onthe Louisiana continental slopeandshelfrevealmultiplebaroclinic oscillations duringHurricane Andrewin August1992.Thesemeasurements aresupplemented by numerical modelsin orderto identifypossible internal wavegeneration mechanisms. ThePrinceton OceanModelisrunwith realistic topography, stratification, andwindforcingto extendtheobservations toMississippi Canyonandotherareasontheshelf.A two-layerisopycnal modelisusedwithidealized topography andspatially uniformwindstoisolateinternalwavesgenerated in andaround thecanyon. The combinationof theobservations andtheresultsfrom the numericalmodelsindicatesseveralpossible mechanisms for generating longinternalwaves:(1) near-inertial internalwavesweregenerated across theslopeandshelfby dislocation of thethermocline by thewindstress; (2) interaction of inertialflowwithtopography generated internalwavesalongtheshelfbreak,whichbifurcated into landward andseaward propagating phases; (3) downwelling alongthecoastdepressed the thermocline; afterdownwelling relaxes,aninternalwavefrontpropagates asa Kelvinwave;and(4) Poincar6 wavesgenerated withinMississippi Canyonpropagate seaward whilebeingadvected westward overthecontinental slope.Theseprocesses interactto producea three-dimensional internal wavefield,whichwasonlypartlycaptured by theobservations. 1. Introduction  temperature, salinity,andcoastalwaterdepthduringthehurricane  The dominantcontinental shelfresponse to a tropicalcycloneis  havebeendiscussed by C94.  The oceanographic response to HurricaneAndrewwas both barotropic, includingthegeneration of a Kelvin-likecoastalstorm barotropic and baroclinic. Keen and Glenn [1999] (hereinafter  surgeandcontinental shelfwaves[Fandryet al., 1984;Hearnand referredto as KG99) discussshallowwater currentsduring the  Holloway,1990;Fandryand Steedman, 1994].Kelvin wavesare directlyforcedstageof the stormanddemonstrate the importance restricted to the transientwavefront [Grimshaw,1988; Tang and of baroclinicflow. Theyfocuson fourprocesses directlyrelatedto Grimshaw,1995].The variablewind stresses duringhurricanes also thegeneration of baroclinic oscillations: (1) turbulent mixing,(2) generate subinertial shelfwaves(frequency to< If I, wheref is the trappedcoastalwaves,(3) near-inertial oscillations, and (4) localinertialfrequency), whichdominate thelarge-scale barotropic upwellinganddownwelling coastalflows.Turbulentmixingwas response [Huthnance,1978; Tang et al., 1997]. In examining strongwithin1 Rw(radiusof maximumwinds)of thehurricane eye. observationsof storm flows, however, the influence of baroclinic  This distancewas about 40 km immediatelyprior to landfall. processes on continental shelfoceanography is alsoapparent [e.g., Consequently, stratification waseliminated nearthecoast,whereas Hazelworth,1968;Smith,1982].The mostcompletedocumentation thepycnocline wasundisturbed at distances greater than2 R•,. The of theseeffectsis availablefor HurricaneAndrew [Cardoneet al., resulting internalwavefield washighlyvariablewithinthe storm 1994] (hereinafterreferredto asC94).  Hurricane Andrew made landfall southwest of New Orleans,  region.  Large-scalemeteorological forcing such as extratropical Louisiana, at approximately 0900 UT on August26, 1992(Figure cyclones canbe treatedas linearfronts[seeKundu,1986].The 1). Whiletransiting thecontinental shelf,thehurricane eyecrossed coastalresponse to suchan impulsivewindconsists of a locally overmooredinstrument arraysmaintainedby the Louisiana-Texas forcedoscillationwithinthe upperlayer and long barotropicand (LATEX) oceanographyprogram, National Oceanic and baroclinic gravitywavesgenerated at thecoast,whichdominate in Atmospheric Administration (NOAA) meteorological buoysand thelowerlayer[Millotand Crepon,1981].Consequently, thenearCoastal-Marine Automated Network stations, and commercial oil  surfaceinertialwave field may be two dimensionalin plan view platforms.The resultingobservations of winds,waves,currents, [Smith,1989]. The wind fields associated with tropicalcyclones cannotbe treatedas lines,however,becauseof their smallspatial scale. Thus the internal wave field will be three dimensional even  Copyright2000 by theAmericanGeophysical Union.  near the surface.  Papernumber2000JC900137.  Louisianacontinentalshelf and slopeduringHurricaneAndrew  0148-0227/00/2000JC900137509.00  This paperexamines the generation of internalwaveson the 26,203  26,204  KEEN AND ALLEN: INTERNAL WAVES DURING HURRICANE ANDREW  A  :!'LOuisianaii½:'.':?'?11iil;,::: ,:•::.::::,•:,:: :':!:: ........... <•;i:ili!?i• .............. ß  /  ,  ,  28øN  26ON  -  24øN  I  94øW  I 92øW  I  I 90øW  I '"" I  I  88øW  I 86øW  84øW  B i i i i i i i i i Ji i i i i i i i i i i i i i i i i i i J 1 t i i i i i i i i i i 1 i i t i I J I i i i I I i I I I I I • I I I I i I  _  29ON  -  _;  _  40  _-  -50  _  2 --  25ON  -  -7•ø' $ø90  -  -  _  B  i i i i i i I i i i i  g2øW  91øW  90øW  i i i i i i i i I  89 ø  Figure 1. (a) Map showingtrackof HurricaneAndrewand locationof the eye (labelsare Julianday). GDI is Grand Isle. (b) Insetmap showingLATEX mooringsand cross-section A-B referredto in text and Plates2 and 3 The cross sectiongoesdowntheaxisof MississippiCanyon.  using both observationsand numericalmodels.The observational reproducesintemal waves within the frequencyband of interest evidencefrom the LATEX mooringsis presentedin section2. sufficientlyto be a usefultool to examinegenerationmechanisms However,becauseof the limitedareacoverageof the dataand the where observations are unavailable. This model is then used to complexityof thecoastalresponse, a baroclinic,primitiveequation examine possible internal wave generation within Mississippi numericalmodelwith realistictopography, stratification, and wind Canyon,which is locatednearthe stormtrack.The resultsfrom the forcingis usedto examineinternalwavegeneration in section3. A realisticnumericalmodelare complex,however,makingit difficult comparisonof the model with the observationsshows that it to evaluatesmaller-amplitude internalwaves.Thus in section4 we  KEEN AND ALLEN:  INTERNAL  WAVES  DURING  HURRICANE  ANDREW  26,205  presentresultsfrom an idealized numericalmodel. This model isolatesthe influenceof alongshorevariationsin topographyand the time dependence of the wind field. The implicationsfrom the  are not directlycorrelatedat mooringslocatedmorethan 1 R•. east of the storm track. We thus expect to find the maximum nearinertial oscillationamplitudesin the observationsfrom near the  observations,the realistic numerical model, and the idealized model  storm track.  are discussed in section 5.  This sectionpresentsevidencefor baroclinicoscillationsusing time seriesof currents,temperature,and salinity at the moorings listed in Table 1. The frequencydependenceof these data is 2. Baroclinic Oscillations Observed on the Shelf examinedby using power spectraof the observations.Previous and Slope observational and modelingstudiesshowthat the followingwaves shouldbe present:(1) barotropicedgewavesand continentalshelf The LATEX moorings(seeTable 1 for depthsandFigure 1 for waves[Fandryand Steedman,1994], (2) barotropicandbaroclinic locations)are locatedalongtwo lineson the Louisianacontinental Kelvin waves [Fandry et al., 1984; Beletskyet al., 1997], (3) shelfandslope.The easternarrayconsists of fourmooringsin water inertial waves [Chen et al., 1996], and (4) superinertialinternal depthsfrom505 m to 20 m. A secondarraylocated150 km to the west has mooringsin water depths of 51 m and 18 m. The waves[e.g.,Niwa and Hibiya, 1997].It is not expectedthat all of these oscillations will be discernible from the observations, temperatureand salinity profiles at mooring 12 (Figure 2) on however, becauseof the presenceof dominant waves and a August24 (Julianday (JD) 237) are representative of stratification background spectrum.It therefore will be necessaryto use before the storm (C94). There was a thermocline at 20 m, but the numerical models to furtherelucidatethe generationof baroclinic mixedlayerdepthindicatedby the salinityprofilewasonly 10 m. oscillations during the storm. The buoyancyfrequencyN is greatestat the thermocline,where  high-frequency internalwaves(N = 25 cyclesperhour(cph))would be generated. However,the samplingintervalof 30 min doesnot resolveanyof theseinternalwaves. The inertialperiod(IP) at the latitudeof the mooringsranges from 25.9 hours(frequencyof 0.927 cycleper day (cpd)) to 26.4 hours (frequencyof 0.909 cpd). In this paper, we refer to frequencies within 20% abovethe localinertialfrequency f asnear inertial(seeKundu[1976] for a discussion). The inertialperiodis very similarto the diurnal O• astronomicaltidal period of 25.8 hours (frequencyof 0.93 cpd). Becauseof the similarity in the inertialand O• tidal frequency,all observedtime seriesof currents had the five largesttidal components(M2, S2, N2, K•, and O•)  The time seriesof currentsmeasuredat the LATEX moorings are presentedin Figure3. The powerspectraof currentmagnitudes (Figure4) were calculatedby usingfast Fouriertransforms(FFD after removingthe mean.The samplingintervalis 30 min. These datawill be usedto analyzebaroclinicoscillations generated by the hurricane'spassage. The lower frequencies (lessthan0.35 cpd) are expectedto be aliasedbecauseof the recordlengthof only 10 days. This verylow frequencymotionwill not be discussed. The currentmagnitudes at bothmetersat mooring12 (Figure3a) reveala strongoscillationat nearthe inertialfrequency(Figure4a).  removed(C94).  The current vectors rotated CW, with the surface and bottom  Inertialflows are correlatedwith synopticwind forcingon the Louisianashelf [Chen et al., 1996; Chen and Xie, 1997]. Forced near-inertialoscillationsare generatedwhen variationsin the wind stressare in phasewith near-inertialoscillationsin the mixedlayer [Schott, 1971]. When this occurs, the oscillations will reach maximumamplitude.During HurricaneAndrew, the wind vectors rotated counterclockwise(CCW) on the western side of the storm track, and thus no correlationis expected.The wind rotated  2.1.  Current  Vectors  currents180ø out of phase(Figure 3b). Plottingthesecurrentsby using hodographs(not shown) indicatesa mean southwesterly barotropiccurrentparallelto isobaths,with surfacecurrentsgreater  than0.3m s-• afterthestorm peak.Thepower peakat0.375cpd indicatessubinertialbarotropicmotion with a period of 64 hours. The currentvariancedecreases steadilyat superinertialfrequencies. For most of the moorings, this high-frequencymotion is significantlyless than the near-inertial and subinertialbands. Althoughtheseinternalwavesare important,it is not practicalto closely examinethem in the presentstudy. Thus we limit our  clockwise(CW) at neartheinertialfrequencyalongthestormtrack. The CW rotationincreasedaway from the trackto the eastand was twicetheinertialfrequencyat GrandIsle (seeFigure1 for location). discussion to internal waves with near-inertial Consequently, the wind rotationand the near-inertialoscillations periodsonly.  Table 1. Locationof Louisiana-Texas CurrentMetersfor This Study [Mooring  Longitude,øW  Latitude,øN  Meter Depth,rn  WaterDepth,rn  12  90.494598  27.923870  12,  100  13  90.485878  28.057529  12,  100,  14  90.492867  28.394569  11,  37,  47  15  90.491577  28.608299  10,  17,  20  18  91.982719  28.962730  10,  19,  22  19  92.034798  28.465170  3,  47,  51  505 190  200  and subinertial  26,206  KEEN AND ALLEN:  DEGREES 5 0  10  15 ß  20  INTERNAL  C 25  WAVES  DURING  32 i  ANDREW  kg/m •  PSU 30  HURRICANE  33  34  35  36  I  i  I  I  I  I  1020  N (CPH)  1025  0  5  10  15  20  25  I  I  I  I  I  I  •  lOO  200  300  4oo  5oo  I  I  Figure 2. Profilesof temperature, salinity,density,andbuoyancyfrequencyat mooring12 onJD 237 (August24).  Using 8 monthsof observations at the LATEX moorings,Chen impliesthat the storm wind field simultaneously generatedthe et al. [1996] found that near-inertialoscillationsare greatestat initialmixedlayerperturbation at thesemooringson August26. mooring13 anddecreaseboth landwardand seaward.This appears Theuppercurrentmeterfailedat mooring15 (Figure3a) andthe to be the casefor the hurricaneflow aswell. The strongestandmost recordis incomplete. The currentsappearto be barotropicprior to recognizablenear-inertialoscillationswere generatedat mooring failure,however.The dominantpowerpeak (Figure4d) at 0.468 13.Thesurface currents reached 1.25m s-• after1 IP, atwhichtime cpdis alsopresentat mooring14.  bottomcurrents exceeded 1 m s'•. After2 IPs, the surface and bottom currentamplitudeswere in phasefor the durationof the measurements. Like the currentsat mooring 12, a CW inertial rotationcontinuedthroughoutthe observationperiod (Figure 3b). The powerspectrum(Figure4b) indicatesthat the largestvariance is associatedwith a strong near-inertial peak at 0.93 cpd, corresponding to a periodof 25.8 hours.Note the superposition of this motion on the O• tide. The current variance in the subinertial  band is broadly distributedand significantlylessthan that in the near inertial.  The surfacecurrentsat mooring 14 containoscillationsthat are initially in phasewith mooring 13. The surfacecurrent variance (Figure4c) is greatestat the near-inertialfrequencyof 1.03 cpd (periodof 23.3 hours).The inertialcurrentsbelow the mixedlayer have an initial phaselag of 6 hours,but they are dampedwithin a few days.The subinertialpeak at 0.468 cpd (periodof 51 hours)is the dominantfrequencyat the lower current meter instead.This peak appearsto representa barotropicflow. Moorings 13 and 14 are 37.5 km apart,whichis very closeto R•. for the hurricane.The correlationbetweenthe initial responseat moorings13 and 14 thus  The along-shelfcorrelationdistancefor the LATEX shelf is about300 km [Chenet al., 1996], but thelengthscalefor the storm, R•,,is only40 km (KG99). Thusthe correlationbetweencurrentsat the easternand westernmooringsis expectedto be insubstantial. Mooring18 is locatedapproximately 2 R},,westof the stormtrack, and consequently, the storm currents(Figure 3a) were weaker. Surface flow was to the west (Figure 3b) for most of the measurement intervalexcepton JD 239, whenthe strongest currents were eastwardin responseto the local wind stress.The current variance(Figure 4e) at the surfaceis distributedbetweennearinertialand subinertialfrequencies. The dominantpower peak (Figure 4f) for surfacecurrentsat mooring19 is at 0.75 cpd (periodof 32 hours).Inertial motionis weakerthan subinertial,but CW rotationis apparentin the current direction at the bottom after JD 236 (Figure 3b), becoming  important at thesurface afterJD 240.Thecurrent variance at the bottomis alsogreatestat 0.75 cpd,asat mooring13, butthe inertial peak is smaller.Subinertialmotion at both metersalsooccursat 0.468 cpd,whichis dominantat the lowercurrentmeter.This lowfrequencymotionis presentat both westernmoorings,as well as  KEEN AND ALLEN: INTERNAL WAVES DURING HURRICANE  ANDREW  A  B  JULIAN 234  236  238  26,207  JULIAN  DAY  240  242  244  Mooring 12:  246  234  236  238  DAY  240  242  Mooring 12:505 m  --12m  505 m  •oorn  246  244  --  12m 100 m  360.  180.  ,  ,  ,  Mooring 13' 200 m  Mooring 13' 200  ,  ,  .  -- 12rn 100 rn  360-  --12m  rn  100 m  180  Mooring 14'  ,  --•  Mooring14' 47 m  m  47 rn  --' 37TM m rn  36O  37m  180  ,  ,  ,  Mooring 15: 20 rn ,,•  --•o  Mooring 15:20 rn  m 17rn  ---1om 17m  360.  180  '••"•'5'•?"'"'"'" ,.,%..-.•¾"'.....-'"'•"•W..., ß  ,  .  ,  .  ,  .  ,  .  ,  .  ,  .  ,  .  ,  .  ,  .  ,  .  ,  Mooring18'  --•o m  22 rn  ....  .  ,  Mooring 18' 22 rn 360  19m  180  Mooring 19'  --3m  51m  --  360,  Mooring 19'51 rn  47 rn  180-  o  0-  234  2•6  2•8  '  2•0  JULIAN  2•2 DAY  2•4  2•6  234  2•6  2•8  '  2•0  JULIAN  2•2  2•4  2•6  DAY  Figure 3. Time seriesof observations at the LATEX moorings.(a) Currentmagnitudes in metersper second.(b) Currentdirection,in degreesclockwisefrom north.  26,208  KEEN AND ALLEN: INTERNAL  WAVES DURING HURRICANE  Frequency (cpd) 0.4  0.6  0.8  I  2  4  12 m  A) Mooring12:  2.0x10's 1.0x10's 0.0  4.0x10's'1  12m  /x,,,  There is no power peak (Figure 6a) at a frequencyof 0.905 cpd, however.Instead,the largestpeakis locatedat 0.6525 cpd (period of 36.78 hours).  I  •  ANDREW  I  B)Mooring 13:  The uppermeterat mooring13 (Figure 5b) recordeda steady drop of 6øC betweenJD 237 and JD 242, and thereis no distinct near-inertialpeak (Figure 6b). Instead, internal wave energy is distributedacrossthe spectrum. The initial temperature response at 100 m wasalsoweak,but near-inertialoscillationsdevelopedon JD 239. A near-inertialpeak at 1.03 cpd dominatesthe temperature powerspectrumbelowthe mixedlayer.  2.0x10'•  JULIAN •  ............. 11m  2.0x10's  234  _  236  238  I  /x,,,,C)Mooring 14:  I  DAY  240 •  I  242 ,  244  I  ,  246  I  A) Mooring 12' 505 rn  3O  1.0x10's 25  0.0  1.0x10's  ............. 17 m  D) Mooring 15: 2O  20 m  .... •,,,,,,i. ....... i- 0.0'""i . '..................... /.... i......... '"'"  I • 10 m  1.0xl 0'•] ............. 19 m  --  E)Mooring 18'  22 m  15  12m  '  I  '  I  '  I  '  I  '  I  '  ing 13:200 rn  3O  25  12m'• ............. 100m  Mooring 19:  2O  1.0x10'•  ;;,  •  ;,",  ',, •: ,.,'•, ; '?•  15  i  '  i  ,  i  ,  i  ,  i  ,., '•,,,;•:  ,  i  ,  G) Mooring 14:47 rn  0.0  0.4  0.6  0.8  I  2  Frequency (cpd)  Figure 4.  :,  "-' ..... :•' '"" ..... •"':-..?; : i : ". i :;".',.':'"",, • "., •,:',:  25  Power spectra for observedcurrents at LATEX  moorings. Units arem2s-2.SeeFigure 1forlocations.  20  .................................. '" '";  11 rn  ............. 37m  moorings14 and 15. It may be associatedwith a barotropicshelf waveasdescribedby Hearn and Holloway [ 1990].  15  2.2. Temperature  The astronomicaltidal signal cannot be easily removed from temperatureand salinitytime seriesbecauseof nontidalvariability. There is, therefore,no way of unambiguously differentiatingnearinertial frequencyvariabilityat around0.93 cpd from the O• tidal signal in most cases.This is partly causedby the choice of frequencybins usedin the FFF method.Thus this discussion will not closelyexaminefrequencies of either0.93 cpd or 0.9375 cpd. Higherandlower frequencies are consideredsafelydistancedfrom the O• signal,however. The near-surfacetemperatureat mooring 12 (Figure 5a) was increasingasthe stormapproached on JD 238 (August25). A rapid decrease of 4øC on JD 239 followed the maximum  surface current  25  2O  D) Mooring18: 22m  -10 rn ............. 19m  234 236 2•8 2•0 2•2 2•4 246 JULIAN  DAY  Figure 5. Temperaturetime seriesof observationsat LATEX (Figure3a) by 8 hours.A secondpulseoccurred26.5 hourslater. moorings.UnitsaredegreesCelsius.  KEEN AND ALLEN: INTERNAL  WAVES  DURING  HURRICANE  ANDREW  JULIAN  Turbulentmixingreducedthe surfacetemperature at mooring14 by 5øC (solid line in Figure 5c) on JD 238. This mixing was not immediatelyseenat the lower meter(dashedline), however,where the temperaturedid not begin to increase until the surface temperaturereacheda minimumof 24øC. The temperatureat the lower metersubsequently increasedrapidlyby 4øC on JD 239 just as a 1øC spikewas observednear the surface.The occurrenceof this peak at both metersindicatesstrongdownwarddeflectionof  234  236  238  26,209  DAY  240  242  35  30  isotherms, but the source of this disturbance cannot be identified  A) Mooring 12' 505 rn  from these data alone. It may representan internal wave as 25  Frequency (CPD) 0.4 i  0.6 i  0.8  1  246  244  12m  20  2  4  '  I  '  I  '  I  '  I  '  I  '  i  4.0x10 '4  A)Mooring 12' 505  -  35  -  30  rn  3.0x10-4  B) Mooring 13:200 m  2.0xl 0-4  25  12m  ............. 100m  1.0x10'4  20  ,  8.0xl 0'4-  35  !i B)Mooring 13' -  i!  200 rn  ,  6.0xl 0-4  30  : ,,,  ,  ,  ,  i i  4.0xl 0-4  ,  C) Mooring 14'  12m  ,  0  :  25  i  11m  47m  .............100m  ............. 37m  ,  ,,'  2.0xl 0'4  ,, ,  0  ,  20 '  0.0  C) Mooring 14:47m  4.0x10'4-  •11m  2.0x10'4  I  '  I  '  I  '  I  '  I  D) Mooring 18:22 m  35  30  25  20 •  0.0  4.0x10'5-  2.0x10'5  234  _/•_,_...•  '•  D)Mooring 18: _  •  I  236  '  I  '  238  I  240  JULIAN  22 rn  ,,.....•,,,,'""i• • -- 10 m  '  I  242  '  I  244  246  DAY  Figure 7. Salinitytime seriesobservedat LATEX moorings.See Figure1 for locations.Unitsarepracticalsalinityunits.  hypothesized by Keenand Glenn[1999] or a response to thelocal wind stress. There was a net increase of 2øC at the lower meter on  0.0  0.4  0.6  0.8  1  2  4  Frequency (CPD) Figure 6. Powerspectraof observedtemperature time seriesfrom  theLATEXmoorings. Unitsare(øC)2.  JD 239, whichis verylikely causedby localturbulentmixing.The dominantpeakfor thelowergauge(dashedline in Figure6c) is at 1.125 cpd, whichis 18% abovethe local inertialfrequency.It is thereforewithin the near-inertialband [Kundu, 1976]. The lack of a  spectral peakat thisfrequency for theuppermetersuggests thatthe initialtemperature response at mooring14 wasonlyweaklyinertial. The uppermeterat mooring18 (Figure5d) revealsturbulent mixing and overshoot,with subsequentrelaxationto local conditions.The time seriesat the lower meter(dashedline in Figure  26,210  KEEN AND ALLEN: INTERNAL  WAVES DURING  0.6  0.8  1  ANDREW  2.3. Salinity  Frequency (cpd) 0.4  HURRICANE  2  4  The observednear-surfacesalinityrecord(Figure7a) at mooring 12 is similarto temperature but with lesshigh-frequency variability (Figure8a). Thereis a subinertialpeakat 0.375 cpd,a broadnearinertial peak, and a smaller peak at 2.4375 cpd (period of 9.8 hours).The response is visiblefor only 2 days,however. The salinity at the upper meter at mooring 13 (solid line in Figure7b) oscillatedby morethan 1 practicalsalinityunit (psu) throughJD 239, and remainedrelativelyconstantthereafter.This prestormvariabilityis reflectedin the near-inertialpeak(Figure8b) at 1.03 cpd. This samefrequencyis presentat the lower meter, althoughthe amplitudeof salinityvariationsis muchsmaller. The reductionin near-inertialoscillationsafter the storm suggeststhat the salinitygradientdiminishedbecauseof turbulentmixing. The surfacesalinityat mooring14 (Figure7c) increased by 3 psu on JD 238, followed by oscillationsof less than 1 psu. This variabilityshowsup as a powerpeak at 0.9375 cpd (solidline in Figure8c) that may be tidally influenced.The salinityat the lower meter decreasedslightly and oscillated weakly thereafter.This variability is representedby a near-inertialpeak at 1.03 cpd in Figure 8c. The dampingof this inertialmotionat the surfacemeter may be associatedwith the transportof salty water from offshore within the mixed layer, as indicatedby the higher salinityat the upper meter. A similar processis indicatedat mooring 19 (not shown)but theresultinggradientwassmaller. The near-surface salinityat mooring18 (Figure7d) decreased by 4 psuon JD 235. This changehad little to do with the approaching hurricane,however. There is no correlatedtemperaturechange either,and this may be causedby eitherinstrumenterror or local advectionprocesses. The spikesof low salinityafter the stormare alsoproblematical becauseof surfacerunoff and naturalvariability alongthe Louisianashelf. The hypothesisthat an internalKelvin wave propagatedwestwardon JD 239 is supportedby a rapid decreaseof 2 psu at the lower meter. This single peak is not accompanied by otheroscillationsor with a net changein salinity.  4.0x10 -$ I A) Mooring 12' 505 m m  2.0x10-5-  0.0  B) Mooring 13:200 m  6.0x10'5 4.0x10'5  12m  100 m  2.0x10'5 0.0  3.0x10-4C)Mooring 14:47 m 2.0x10 -4  I 11m  ............. 37m  1.0xl 0-4 0.0  4.0x10 '3 '• D) Mooring 18:22 rn •  2.0x10'3  ---  10m  3. SimulatingInternal Wave GenerationWith RealisticTopography and Forcing The observations discussed in section 2 indicate that internal  wavegeneration on theLouisianacontinental shelfandslopewasa complex processthat extendedbeyond the availablemooring arrays. The limited measurements prevent different generation 0.0 mechanisms frombeingfully exploredwithouttheuseof numerical 0.4 0.6 0.8 1 2 4 models.Consequently, a numericalhindcasthasbeencompleted by using realisticforcing and bathymetryin order to resolvethe Frequency (cpd) interactionof the stormflow andthe coastaltopography better.It is Figure 8. Power spectraof observedsalinitytime seriesat LATEX not feasibleto use this model to explain the detaileddynamical moorings. SeeFigure 1forlocation. Units are(psu) 2. relationshipsbetween observationsat the different moorings. Beforeproceedingto an analysisof internalwave generation,the 5d) containsa transientsignal like the one at mooring 14 but model results will be compared with the observationsto without a net increasein temperature.Neither meterhas a strong qualitativelydeterminehow confidentwe can be in usingthem are lacking.The modelwill thenbe usedto near-inertialpeak (Figure 6d). In fact, the spectraare biasedto wheremeasurements examine regional internal wavegeneration duringthehurricane. subinertialtYequencies. There was a net cooling of the water column,possiblycausedby localrunoffor advectionfromoffshore. 3.1, Princeton Ocean Model Hindcast Using water level and temperaturedata, KG99 show that an The wind and wave fields duringHurricaneAndrewhave been internal Kelvin wave generatedat the coastduring downwelling couldproducethe transienttemperature responseobservedon the hindcastfor 1000 UT August24 to 0000 UT August27. The 30shelf.This mechanism is discussed by Csanady[ 1984], Millot and min wind fields are generatedby first reanalyzingtraditional Crepon[1981], andBeletskyet al. [1997]. Strongforcingsuchas cycloneparameterssuch as track and intensity(in terms of The moredifficultstormparameters, suchastheshapeof that during a hurricanewould producea nonlinearwave with a pressure). theradialpressure field andtheambientpressure field withinwhich steepfront[Bennett,1973],asobserved at moorings14 and 18.  KEEN AND ALLEN: INTERNAL  WAVES DURING  the storm is embedded, are then estimated. The time histories of  theseparametersare specifiedfor the entirehindcastperiod.The stormtrackand stormparameters arethenusedto drive a numerical primitiveequationmodelof the cycloneboundarylayer [Thompson and Cardone, 1994]. The solutionis comparedwith measured surfacewinds, and if necessary,the storm wind parametersare variedandthe modelrerun.This iterativeprocedureis continuedto minimize error. The wind speedat Grand Isle (see Figure 1 for  HURRICANE  ANDREW  26,211  three-dimensionalbaroclinic flows. The present choice of simulationparametersreflects the resultsof previous work on improvingmodel skill for HurricaneAndrew [Keen and Glenn, 1998]. The model domain covers the Gulf of Mexico with a  horizontalresolutionof 0.05ø (approximately 5000 m). A minimum waterdepthof 8 m is usedwith 50 (5 levelsandgrid stretching near the surface.The resultingverticalresolutionat the surfaceis 4 m for a maximumdepthof 3656 m. The modelequationsare integrated location) is consistently 3-4 rn s'• higherthanmeasured but by usingan extemaltime step(to solvefor the barotropicflow) of directionsagreeclosely.Furtherdetailsareavailablein C94. 16.67 s and an intemaltime step(for baroclinicflow) of 304.44 s. The ocean circulationhindcastbeginsat 1000 UT August 24 Temperatureand salinity profiles measuredalong the eastem with the ocean at rest and continuesuntil 0000 UT August 31. mooringarray(moorings12 through15) on August24 wereusedto Since no winds are available after 0000 UT August 27, the construct a depth-dependenttemperatureand salinity initial remainderof the hindcasthas no atmosphericforcing. The winds conditionfor the model[Keenand Glenn, 1998]. Wave-breaking were rampedup for 12 hours to reduceinitial oscillations.The turbulenceis incorporatedinto the turbulenceclosuremodel by Louisiana continental shelf thus has a spin-up period of usingthe hindcastwave fields. The inertialperiodfor the northemGulf of Mexico is between approximately 36 hoursbeforehurricanewindsarepresent. This studyusesthePrincetonOceanModel (POM) [seeOeyand 24 and26 hours.The dominanttidal periodsin the Gulf of Mexico Chen, 1992; Mellor, 1993; Mellor and Yamada, 1982] to calculate arethediumalK• (23.9 hours)andO• (25.8 hours),whichoverlap  B. Mooring 13:12 rn  A. Mooring 12:12 rn 1,5-  OBSERVED ..........  POM  •  OBSERVED  ..........  POM  1.0 1.0  0.5  0.0  0.5  ,  '  i  '  238  ' 2•8 ' 2•0 ' 2•2 ' 2•4  2.0  ,•  OBSERVED 0.5  -......... POD  1.0  i  '  242  i  244  D. Mooring 19:3 rn  C. Mooring 14:11 rn  1.5  '  2,•0  OBSERVED  i  .......... POM  u)  ',,ii,,,,, t:  ;'•  1',  •',  -  I  ',  ;  ! / '...:  0.5  0.0  .,  238  ' 2,•0' 242' 2•4  0.0  .'  ,  238 2•0 242 2,•4 Julian Day  Julian Day  Figure 9. Time seriesof observedand POM-predictedcurrentmagnitudesat the upper metersat the LATEX moorings.  26,212  KEENAND ALLEN:INTERNALWAVESDURING HURRICANEANDREW  A -92  29  •  -91  -90  -89  -- • •  29  28  28  -92  -91  -90  -89  -92  -91  -90  -89  B  29  28  28  -92  -91  -90  -89  -92  -91  -90  -89  C  29  28  L-  o•./..?• .  28  ./  -92  -91  -90  -89 I  -60  -48  -36  -24  -12  0  12  24  36  m  48  60  Plate 1. Model deviation(m) of the 26øC isothermfrom its initial depth.The 200-m isobathis shownas well as the stormtrack and positionof theeye.The eyewasoutof view in Plates1a and 1c.  KEEN AND ALLEN: INTERNAL WAVES DURING HURRICANE ANDREW  7  9  11  13  15  17  19 21  23  25  27  29  31  7  9  11 13  TEMPERATURE(C)  15 17  19 21  26,213  23  25  27  29  31  29  31  TEMPERATURE(C) ß  .  .,  500  5OO  ,  -\  1000  '  '  "  1 ooo  ,  A)JD237.583  '"• ••?'•' ;•' !"i "• 50  B) JD 238.917  ,.  100  1•;0  0  1•0  50  Km  9  11  13  15  17  19 21  23  25  27  29  31  9  TEMPERATURE(C)  11  13  15  k  ,  17  19 21  23 25  27  TEMPERATURE(C)  ß  500  150  Km  .  !  5OO  1000"  1000  "D) JD 239.583  "ß C) JD 239.25  50  100  Km  150  0  50  100  Km  Plate2. Crosssections of POM-predicted temperature alonglineAB fromFigure1.  150  26,214  KEEN AND ALLEN: INTERNAL  WAVES DURING HURRICANE  ANDREW  the inertialperiod.Thereforetidal forcingis neglectedin the observations.This peak is somewhatstronger,however. The simulationso that the inertialresponsecan be evaluatedmore subinertialpeakis muchlargeras well. Severalhigherfrequency effectively. During the spin-up interval, near-inertialoscillationswere  peaksare absentin the observations at this mooring.The model resultsincludethesepeakswithreducedamplitudes. generated at the continental slope.The dampingtime for nearThe hindcastsurfacecurrentsat mooring19 (Figure9d) match inertialoscillations in themodel(seeFigure9) is morethan6 days, the observationsduring the hurricane'spassagebut diverge although thisvarieswithlocation. However,thisis nota significant significantly afterward.Thisbehavioris probablydueto thelackof measureof model stabilitybecausenear-inertialoscillationsare wind forcingafterthe storm.The currentsare out of phaseby JD ubiquitouson this shelf. What is more importantis that the 240.0 becausethe near-inertialpeak(Figure 10d) is very weakin magnitudeof the prestormcurrentsis similarto the observations or the observations. The observations reveala strongsubinertialpeak smallenough to beinsignificant. Comparisons usinga modelwitha at 0.82 cpd(periodof 29.2 hours).The modelpredictsa maximum spin-upintervalof 5 days showeddegradation in model skill, peakat 0.586 cpd and a smallerpeakat a near-inertialfrequency. because of thelackof near-inertial oscillations priorto thestorm. Notealsotheobserved superinertial peakat 1.406cpd(periodof 17 hours),which is reducedin the model.The model doespredict someof thehigher-frequency motion,however. 3.2. Comparisonof POM ResultsWith Observations This studyfocuses on mechanisms of internalwavegeneration. The wind and wave fields used to force the POM end at 0000 Thereforepredictingtemperature and salinitydistributions is not a UT August27 (JD 240.0).Thusthemodelfreelyoscillates during primaryconcernof the numericalhindcasteffort. Nevertheless,it is therelaxation stageof the hurricane. Oneconsequence is thatthe usefulto havethemostaccurate initialdensitydistribution possible model-predicted posthurricane flow doesnot perfectlymatchthe in order to capturelocal baroclinicprocesses in differentwater observations. Keen and Glenn [1996, 1998] use quantitative depths.Despite using depth-dependent initial temperatureand methods to examinethe model'sskill duringthe directlyforced salinityfields,we expectthe modelto predictlessvariabilityin the stage.A skill analysisbasedon the stormpeak,however,is not higher-frequency pan of the spectrumbecauseof local effects.In applicable because of thenatureof theobservations; thestormpeak fact,the observedcurrents(Figure10) do revealmoresuperinertial is followedby freeoscillations superimposed on coastalflowsthat motion than that reproducedby the model. Local processes are independentof the storm.This combinedsignal makes significantly affectedthetemperature andsalinitydistributions prior quantitativemodel skill analysisproblematical. A qualitative to the storm. Consequently,even when the most timely comparison of the model-predicted andmeasured flowsis madein measurements were used, the observations were insufficient to KG99 andis notrepeatedhere. generatethree-dimensional fields that accuratelyreproducedthe Thepurpose of thisstudyisto examine internalwavegeneration. temperatureand salinity records at all of the moorings. Therefore themostimportant indexof modelperformance is in the Furthermore, the inclusionof the tidal signalin the measurements frequencydomainasdiscussed above.Thusthissectionwill usethe makesa directcomparisonbetweenthe modeland the observations FFT methodto comparethe model'sfrequency response to the problematical. Otherfactorsthatpreventa quantitative comparison observations.As was demonstratedin section2, the currentsare during the total studyinterval are local rainfall, runoff from the moreusefulthantemperature and salinityfor examining internal coast, and local heat fluxes. Becauseof these problems,a waves becauseenvironmental factors not included in the model can quantitative skill analysiscannotbe appliedto the model-predicted significantly affecttemperature and salinity.The robustness of the temperature andsalinityresults. hindcast will be demonstrated by visuallycomparing the current The depth-dependent initial conditionfor temperatureand magnitudes at several moorings. Beforecomparing theobservations salinityusedin thePOM resultedin a thermaltrendat mooring12 and model predictions, the observedcurrentmagnitudes were (Figure11a) thatis oppositeto the observedtendency.The modelinterpolated to themodeltimesbyusinga polynomial interpolation predictedtemperature in the mixedlayer(dashedline in Figure11) [Presset al., 1992].However, themeans werenotremoved priorto increasedratherthan decreased.Warmer water from the upper analyzing with the FFF in orderto evaluate the modelresponse slope,as represented by 29.5øC water from mooring13 (Figure morecompletely. l lb), was advectedseawardby storm currents.The hindcast The hindcastsurfacecurrents(dashedlinesin Figure9a) at temperature (Figure1lb) at mooring13 dropsrapidlyas the storm mooring12 areverysimilarto theobservations (solidlines)during passes,before stabilizingat 2øC lower than the initial condition. the stormpassage, but the peaksare low. A disparityin currents The responseat mooring 14 (Figure 11c) is similar to the afterthemodelwindsceased isapparent. Themodelpredicts strong measurements,with rapid mixing as the eye passed over. near-inertialoscillations are seenin both measuredand oscillations (Figure10a) with a near-inertial peakmatchingthe Subsequent observations. A second near-inertial peakis predicted at a higher predictedtime series. The model predictionsat the western frequency but with lesspowerthanthat observed. The modelhas moorings(Figures1ld and 1l e) arealsosimilarto the observations. morepowerat thesubinertial frequencies. The discrepancies at moorings12 and 13 are very likely partly Themodel-predicted surface currents at mooring13 (Figure9b) attributableto an eddybecausethereis no analogoustrendin the havelow amplitudes and a slowlyincreasing phaseerror.The measuredsalinitytimeseries(Figures7a and7b). modelspectrum (Figure10b)is similarto the observations, but with A direct comparisonof the time seriesand power spectraof a lowernear-inertial peakanda strongsubinertial peak.Thephase measured and predictedcurrentmagnitudes hasbeenpresented in erroris causedby a shift of the near-inertial peakto a higher this section.The resultsindicatethat the POM is capturingthe frequency. Note, however,that severalsuperinertial peaksare dominantbaroclinicsignalsat the moorings.The model skill is predictedby themodel. greatestat the easternmooringswhere the wind forcing was The hindcastsurfacecurrentsat mooring14 (Figure9c) have simplestand strongest.The responseat the westernmoorings,  smallphaseerrors,but the maximumcurrentis 50% high.This which are located more than 2 Rw from the storm track, is not as behavior is explained by thepowerspectra (Figure10c).Themodel good.The relaxationflow is well predictedby the model,despite has a near-inertialpeak that almost exactly matchesthe the lack of poststorm wind forcing,becauseof the dominantnear-  KEEN AND ALLEN: INTERNAL WAVES DURING HURRICANE ANDREW  1.0x10-5-  --  •  B. Mooring 13:12 rn  2.0X10'5  A. Mooring 12- 12 rn  26,215  • .......... Observed  Observed  ..........POM  5.0X10-6  1.0x10'5  0.0 0.0  0.4 0.60.8 • 2.0X10'5-  2  4 3.0X10'6-  C. Mooring 14:11 rn --  i  ;,,  0.4 0.6 0.8 i  2  4  D. Mooring 19:3 rn  Observed  ..........POM  •  Observed  ..........  POM  ,, ,,  2.0xl 0'6,  ,  ,  ,  1.0x10'5  1.0x10-6-  0.0 ,  0.4 0.60.8•  2  ,  4  CPD  0.4  0.6  0.8  I  '  1  2  4  CPD  Figure 10. Powerspectraof observed andPOM-predicted currentmagnitudes at LATEX moorings. The verticalaxis unitsarem2s-].  The POM predictssurfaceflow to the northwestand parallelto isobathsnear the MississippiRiver delta as the hurricaneeye crossedthe continentalslope(northeastcomerof Figure 12a). This flow is part of a cycloniceddy (labeled1c in Figure 12a) drivenby the wind stress.Bottomcurrentsat thistime (Figure 13a) aretoward the shelf break and into MississippiCanyon (see Figure 1 for location), forming a downwelling-favorableflow regime. Thus 3.3. Baroclinic Flow on the Louisiana Shelf near-surfaceflow east of the canyon does not cross isobaths, The shallow-water currents on the Louisiana continental shelf whereasbottomflow crossesisobathsto within a few grid pointsof during HurricaneAndrew have been discussedby KG99. The the coast.This flow is a significantdeviationfrom previousreports observations were used in combination with a numerical model to of downwellingflowsnearcanyons[Klinck, 1996; Hickey, 1997]. It examine coastally trapped waves, near-inertial currents, and is a geometriceffectcausedby the presenceof the deltato the east. Flow on thewesternsideof the canyonis moretypical,with bottom upwelling- and downwelling-favorable currents.However, KG99 focusedon the buildup and directlyforced stagesonly. In this currentsfollowing isobaths.Bottom currents are predominantly by KG99, study,we extendthe analysisto the relaxationstage.A detailed seawardwithin the canyonat this time. As was discussed examination of the baroclinic flow associated with internal waves weak upwelling-favorableflow is present near the western moorings. duringthisintervalhasbeencompletedin the frequencydomainin Justbeforelandfall, a cyclonicbarotropiceddy (labeled2c in sections2 and3.2. This sectionattemptsto identifysalientfeatures of the three-dimensional currentstructurethat accompanied the Figures12b and 13b) is presentat the head of the canyon,while baroclinic oscillations seen in the measurements and the model. flow on the shelffollowsisobaths.This flow patternis similarto inertial and subinertialresponseof the shelf and slope waters. Higher-frequency motionis alsoreproduced, however.The greatest discrepancyis found in the subinertialrange, where the model predictionis too large.This error is largeston the shelfwestof the stormtrack.The model-predicted thermalhistoryat the mooringsis in generalagreementwith the observations on the shelf.  26,216  KEEN AND ALLEN: INTERNAL JULIAN 234  236 ,  I  238 a  I  DAY  240 t  WAVES DURING  I  242 •  I  246  244 I  I  I  HURRICANE  ANDREW  rotationas the inertial flow. The upwelling-favorable flow west of the stormtrack continuesat this time. The barotropicanticyclonic eddylabeled4a is coupledto the cycloniceddy labeled5c. Eddy pairssuchas this pair are generated throughthe interactionof the  A) Mooring • 2' • 2 m  3O  A) JD 238. 917 Max Vector= 1.94 m/s 25  -92 2O  ß  i  ,  i  ,  I  ,  i  '  i  -90  -89  ,  29  3O  29  .............  25  '  i  ,  i  ,  i  ,  '•.:'.5. •->•*-"-x":-",-%•• * I  28  B) Mooring 13' 12 rn 2O  -91  i  ,  i  28  ,  3O  -92  25  -91  -90  -89  B) JD 239.25: Max Vector= 1.93 m/s C) Mooring 14:11 rn  -92 I  2O  -91 I  I  -90 I  •  -89 I,  •,  3O  29 25  D) Mooring 18' 10 rn 28  28  2O 3O  -92  -91  -90  -89  25  E) Mooring 19:10 rn  C) JD 239.583: Max Vector= 1.15 m/s -92  234 2•6 2•18 2,i0 242 2,i4 246 JULIAN  -90  -89  •,.i,,,.,,•,,•....•.,..•l • i_! i t_  DAY  Figure 11. Time seriesof observed(solidline) andPOM-predicted (dashedline) temperature(in degreesCelsius)at upper metersat LATEX moorings.  -91  29 ,._-...-...-..-.•-C...•.,, •..•  •  • • • • •''  ' '•'""•:'•"'•••.  •-,  •.': • -- r 29  '••/•///I  !  thatfor a narrowcanyon[Allen, 1996]. Thereis alsoevidenceof an anticycloniceddy at the bottom(labeled3a in Figure 13b). The surfaceexpression of this weakgyreis maskedby the wind-driven currents,however,as indicatedby the dashedline usedfor eddy 3a in Figure 12b. Paired eddiesnear canyonssimilar to thesehave -92 -91 -90 -89 been observedin laboratoryexperiments duringstrongoscillatory flows [Perdnneet al., 1997]. The model-predicted eddiesmay be Figure 12. Snapshots of surfacecurrentspredictedby POM. The generatedby combinedinertialandwind-drivenoscillatoryflow on labeled circles are eddies referred to in text: a, anticyclone;c, the outer shelf, where the wind has the samesenseof anticyclonic cyclone.Seetext for explanation. ,  ,  KEEN AND ALLEN: INTERNAL WAVES DURING HURRICANE ANDREW  currentspersistat midshelf.Bottom currentsare variable near the shelfbreak,but strongflow bothto theeastandto the westfrom the canyonis evident.  A) JD 238.917 Max Vector= 1.38 m/s -92  29  -91  -90  ............  -89  -,---/.-.....  ..........  This discussion shows several characteristics of the storm flow  29  •' :N'"-"-"-'---'"---'--"--:-"-' • ,"' '• "/I  .............  "//••-'"  ..............  : :-_-_ -.  • ........  /''1  , •,.,  28  28  ,  ,  -92  ,  -91  -90  -89  B) JD 239.25 Max Vector= 1.24 m/s -92  -91  -90  26,217  -89  that impact the generationof internal waves. First, the flow is stronglyrotational.Much of this energy occursat near-inertial frequencies.Eddies predictedby the model are generatedby mechanismsthat have been studied in field, laboratory, and modelingstudies.The complextopographyof the Louisianashelf simply places these mechanismsin proximity to one another. Baroclinicoscillations are seenin boththe openshelfsettingto the west of the stormtrack and in the environmentof Mississippi Canyon.A strongbarotropicflow dominateson the inner shelf. These currentsreverse direction after landfall. The barotropic Kelvinwavethatdrivesthisflow is not the topicof thispaper.Nor do we pursuethe continentalshelf waves that also contributeto subinertialbarotropicflow. Instead,the remainderof this paper focuseson near-inertialand superinertial internalwave generation due to thermoclinedisplacement on the open shelfand within the canyon.  3.4. ThermoclineDisplacement The generationof internal waves at the thermoclinecan be examinedby using the predicteddeviationof the 26øC isotherm (Plate 1) from its initial depthof approximately 35 m. Positive anomalies indicate downward  deflection of the isotherm. When the  hurricaneeye was approachingthe continentalshelf (Plate l a), downwellingwas occurringnear MississippiCanyon where the 26øCisothermwasdisplaceddownwardby as muchas 35 m. The possiblecauseof the variationsin the isothermanomalyalong the shelf break  -92  -91  -90  -89  C) JD 239.583: Max Vector= 0.52 -92  -91  i  •1  -90  will  be discussed in section 4.2.  There  are both  landwardand seawardpropagating internalwavessuperimposed on trappedcoastalwaves.Thesewaveswere generatedduringspin-up of the model. However, thesemodesare also generatedby the storm, as will be discussedin section4. A baroclinic  -89  i  29  29  front located at 91øW is the result of thermocline  deepening.This front subsequentlypropagateswestward as a trappedcoastalwave (Plates lb and l c), which follows the inner shelf topographyratherthan the continentalslope.This transient wave appearsto be an internalKelvin wave asproposedby KG99.  However, theestimated frontal speed of 0.7m s-1is significantly '  '"  '"  •" t•'  ;;;;;;;,/// l L/  ••'•K//)tLf  ',.'--.-.  k'..d'.•  ,  /  •  -'- .....  28  28  ,-•  ...................... ............. ,  -92  ,  -91  ,  -90  -89  Figure 13. Snapshots of bottomcurrents predicted by POM. The labeledcirclesare eddiesreferredto in text: a, anticyclone;c, cyclone.Seetextfor explanation. largecycloniceddy(lc in Figure12a)with thebottom[Fandryand Steedman,1994]. The windweakenedsubstantially afterthe eyemadelandfalland the relaxationstagebegan.The couplededdy pair 4a-5c (Figure 12c) has propagated westward,and eddy 5c is no longervisible belowthe mixedlayer(Figure13c).Flow is barotropicon the inner shelfon thewesternsideof thestormtrack,but upwelling-favorable  lessthanthevalueof 4 m s-• estimated fromtheobservations. This kind of discrepancy betweenmodel-predicted andobservedinternal frontshasbeennotedby Beletskyet al. [ 1997] for internalKelvin wavesin Lake Michigan.Amongthe possiblecausesare numerical dispersion,errorsin stratification,heat and salt fluxes,bathymetry errors,andwind forcing. Severalhours after the eye made landfall (Plate lb), the 26øC isothermanomalyindicatesthat multipleinternalwavesare being predictedby the model. Superinertialedgewavesare propagating alongthe 200-m isobath,after being generatednear Mississippi Canyon.Theseedgewaveshavea wavelengthsimilarto the storm Rw.A near-inertialwaveis generatedwestof the canyon.This wave originatesat the largepositiveanomaliesnear90.1øW,betweenthe stormtrack and the 200-m isobath.It is seen as a large 26øC anomalygradientlandwardof the 200-m depthcontour.Note the similarityof this wave in Plate 1c, which is 24 hourslater. There is also evidenceof internalwaves within the canyon.Subsequent internalwavegenerationconsistsof shelfwavespropagatingto the west and oscillationswithin the canyon.The near-inertialwaves generated at the slope,whicharenot apparentin theseplots,will be discussed in section 4.  26,218  KEEN ANDALLEN: INTERNAL WAVES DURING HURRICANE ANDREW  3.5. InternalWaves WithinMississippi Canyon  themaincanyon,at distances between50 and150km.Thisreflects  thegeneration of downwelling currents along theeastern canyon Displacement of thethermocline wasimportant forgenerating rather thannearthecanyon head, where flowwasprimarily internal waves during Hurricane Andrew. Thelargest anomalies in margin Aftertheinitialpulse,thecurrents are the depthof the 26øCisotherm arepredicted by POM near parallelto depthcontours. inertial.However,thereis evidence in thecrosscurrents Mississippi Canyon because ofdownwelling intothecanyon prior dominantly values in Plate3a) of a wavepropagating seaward at to landfall.Furthermore, eddiesnearthe canyon influenced the (positive generation and propagation of near-inertial wavesobservedat the  approximately 0.5ms-•.Itisalso seen intheaxis currents although  it ismasked bythelargevariations atnear-inertial frequencies. This moorings. Because of thelackof observations withinMississippi  to exitthecanyon. Canyon, however, it isnecessary torelyonthenumerical modelto phaseappears The previous discussion of temperature andcurrents within examineinternalwaveprocesses in thisarea.The modeldoesnot Mississippi Canyon demonstrates that different internal waves were predict thecurrents, temperature, andsalinity exactly. Nevertheless,  in thisarea.However, thecomplexity of the it accurately reproduces theoverall spatial andtemporal evolutionbeinggenerated  wind, and stratification in the POM hindcast of the stormresponse. Thus it can be used to examinethe realistictopography,  it difficult toseparate thewavephases generated during the generation of internal waves withfrequencies of 0.7 to 2 cpdin makes  hurricane. In orderto isolatethe internalwaves,it is usefulto areaswhereobservations arelacking. Mississippi Canyon is approximately 30 kmwideat the200-m simplifythe problem,especiallyto eliminatethe subinertial  continental shelfwavesassociated withthe spatial isobath (seeFigure 1).Thusit is significantly larger thantypicalfrequency  canyons discussed in the literature. For example, Juande Fuca variabilityof the wind field. Canyon [Cannon, 1972],Hudson Canyon [Hotchkiss andWunsch,  1982],andGrand Rh6ne Canyon [Durrieu deMadron, 1994]are 4. SimulatingInternal Wave GenerationOver between 10and15kmwide.Astoria Canyon [Hickey, 1997]and  IdealizedTopography Theinternal deformation radius Ro- N hf-• (where histhedepth) Thepurpose ofusing a simplified model inthepresent study is MontereyCanyon[Petruncio, 1996]arelessthan10 km in width.  for Mississippi Canyon is approximately 36 km,whichis of the to isolatethe mechanisms for generating internalwaveson the same orderasthecanyon width.Consequently, thesurface currentscontinental shelfandslope, especially in Mississippi Canyon. The in Figure12aresimilar to Klinck's [1996]weakly stratified case results fromthePOMindicate thatthisfeature isa significant siteof (i.e., wide canyon).Bottomcurrents (Figure13a) indicatewave generation.The observations indirectlysupportthis downwelling ontheupstream margin of thecanyon asthestorm conclusion,although direct measurementsare not available. approached. Upwelling subsequently begins ontheupstream sideof Idealized models canbeusedto examine physical mechanisms for  thecanyon (Figure 13b)andspreads to thedownstream marginwave generation without thecomplexities introduced bycontinuous  after landfall (Figure 13c). This flowisingeneral agreement with stratification, spatially variable windstresses, andtopographic theresuits ofprevious model studies [Allen, 1996; Chen and Allen,irregularities. However, it is important thatthesimplifying 1996], butfordownwellingrather thanupwelling-favorable winds.assumptions donotinvalidate theiruse.  This general similarity extends tothegeneration ofeddies discussedThewater column onthecontinental shelf andslope prior to above. Hurricane Andrew canbeapproximated asatwo-layer system. As Thegeneration ofinternal waves within Mississippi Canyon can isshown byPlates 1 and2, internal waves aregenerated atthe beexamined byusing cross sections along thecanyon axis(seethermocline bythewind. These oscillations arereadily studied by Figurelb for location). Theinitialtemperature (Plate2a) using anisopycnal model [e.g.,Cooper andThompson, 1989]. distribution wasgenerated byusing hydrographic datafromthe Internal waves generated bytopography commonly travel asbeams easternmooringarray.Warm waterfromthe shelfis foundon the  closeto thegeneration region[Petruncio, 1998].Farther away, slope atdepths of400m. Deepening ofthemixed layeroccurs as thesewavesalso tend toward a mode 1 baroclinicstructure. the storm intensifies(Plate 2b), with maximumdownwardIdealizedmodelshaveprovenusefulto examinethe interaction of deflection of the thermocline at a distance of 80 km from the stratified shelfandslope flowswithtopography [e.g.,Allen,1996; canyon head.Thisdownwelling peaked justbefore theeyemade Carrasco, 1998].Thusit is advantageous to simulate internal wave landfall(Plate2c).At thistime,warmwaterfromtheinnershelfhas generation during Hurricane Andrew byusing atwo-layer isopycnal  penetrated to a depthof 100m withinthecanyon, andthe 18øC numerical isothermhas been depressed from 150 m to 400 m.  This  downwarping pushed the8øCisotherm toa depth of 650m. After landfall(Plate2d), the isotherms haverebounded to neartheir  model.  4.1. Shallow Water Model  TheShallow WaterModel(SWM)[Allen,1996]asusedin this  original depth, andaninternal wavewitha wavelength of 50kmis studyconsists of two-layers. It incorporates an enstrophyand seenalongthe8øCisotherm.  energy-conserving formulationfor the advectionand Coriolisterms  Internal waves within Mississippi Canyon canbeidentified by [Arakawa andLamb,1981],except wherethedepthof a layer  using phase plotsof currents fromthelowest sigma levels fromthe becomesvery small.When this occurs,the modelusesthe firstPOMmodel (Plate3).Thecurrents inthefigure havebeenrotatedorderupwindscheme for themassconservation equation. The to alignwiththecanyon axis.In addition, themagnitudes ateach vertical viscosity takestheformof a lineardragbetween thetwo station alongthecanyon havebeennormalized separately. This layers anda linearEkman-type bottom dragatanangleof 20øon method focuses onthevariability ateach station without biasing the thebottom layer. Because oftheshallow water (minimum depth of result withthelargest magnitudes, which occur nearthecanyon30m)used in thesimulation, a moderately largevertical viscosity head.Negative valuesof across-axis andalong-axis currentshadto be usedto keepthe modelstable.Parameters usedin the  indicate flow to the southwest andsoutheast, respectively. simulation aregivenin Table2. Maximum bottomcurrents arepredicted at JD 239 (0000UT  Themodelgrid(Figure14a)contains a straight coastline, shelf, August 26) throughout thecanyon. Axialflowwasseaward within andslope withawidthsimilar tothestudy area. Theslope iscutby  26,219  KEENANDALLEN:INTERNAL WAVESDURINGHURRICANE ANDREW Table2. Parameters Usedin theShallowWaterModel Value  Parameter  Gridsize  3 kmx 3 km  Domainsize  600kmx 600km  Upper layerdepth  20m  Reducedgravity  0.038 ms-' 2  500m  Horizontaleddyviscosity  -1  s  Verticaleddyviscosity  0.05 m2s"  Ekman layerdepth  14m  Coriolisparameter  6.83x10-5 s-'  Beta  Northemboundary  30 m deep,no slip  Southemboundary  1000mdeep, zerogradient velocity, constant gradient surface and interface  periodic plus40 gridpointsponge to noeast/west  East/west boundaries  variations  Time step  10s  Totalsimulation  4 days  (Figure 14b)withmagnitudes andtimedependence a canyon similar in scaleto Mississippi Canyon. Thissimple windstress representative of mooring 14, which lay directly along thestorm topography reduces alongshore variations in theflowfieldand track.The idealizedmodeleliminates the effectsof spatially permits theflowinteraction withthecanyon tobeexamined in windsandcontinuous stratification. A uniformwindstress detail.This simulation considers the effectof a spatiallyuniform variable  B W•nd used in 2-Layer  A Bathymetry for 2-Layer Model 360  Model  .............................  268  218  Y 16½  118  68  Contou,red from50m[o950m bv50m-  15½  21½  278  33½  39½  45½  x  Simulation Hour  Figure 14.(a)Bathymetry used forShallow Water Model simulation. Units are kilometers. (b)Time series ofuniform wind stress used inSWMsimulations. Units arem2s'2(stress/water density).  26,220  KEEN AND ALLEN:  INTERNAL  WAVES  DURING  HURRICANE  B)  A)  -1.0  -0.5  0.0  0.5  1.0  -1.0  244  244  243  243  >,242  >,242  •- 241  c: 241  -•  240  --3 240  239  239  238  238  0  ANDREW  50  100  Km  150  0  -0.5  50  0.0  100  0.5  1.0  150  Km  Plate 3. Phaseplotsof normalized(a) cross-canyon currentsand (b) along-canyon currentsThe crosssectionis line AB in Figure 1.  KEENANDALLEN'INTERNALWAVESDURINGHURRICANE ANDREW  26,221  thewindwasnotmeasured. Nevertheless, by using will not generate continental shelfwaves,whichwouldobscure either,because the hindcast wind (C94) and the measured temperature, the lag lower-amplitude wavesthat cannotbe identifiedfrom the between the peak wind and maximum downward deflection is observations. The simplestratification precludeshigher-mode to be 10 hours.The SWM predicts an interface rebound internalwaves.The idealizedmodel simulationis not intendedas a estimated rebound, butthelag hindcastof HurricaneAndrew. This model is useful to examine of 12m, whichis lessthanhalftheestimated error (relative to the peak wind) is only 2 hours. The usefulness of fundamental processes only,andthusnoattempt ismadetovalidate this comparison is in demonstrating that the response on the slope is or tunetheSWM. Theuseof datesin describing theSWM resultsis  dominantly inertial, which istofirstorder captured bytheidealized  for convenienceonly.  SWM.  4.2. Model Results  4.2.2. Trappededgewaves. The canyonin the model  generates superinertial edge waves overtheshelf. This The response of the simplified modelto the stormwinds topography response to thecanyon in SWM.These includes surface velocities ashighas2 m s-• anda coastal storm is themajorbarotropic havea period of 100minanda wavelength of about340km. surgeof 0.32 m. Upwelling predicted duringday 1 produceswaves  Themaximum amplitude of 0.18m occurs at modelday1.5(JD 239.0). They are generated when surges within thecanyon reach the interface byasmuchas31m. Theinternal waves generated bythe coast. These waves travel both eastward and westward from the SWM are of particular interest to the present study.Sincethe withtheexpected modesplitting andfaster propagation to idealizedwind forcingis spatially uniform,the SWM doesnot canyon, andSteedman, 1994;TangandGrimshaw, 1995]. generate the subinertial shelfwavesdiscussed by Tangand theeast[Fandry associated withthiswavetumscyclonically in the Grimshaw[1995].The SWM generates coastally trappedwaves, Thevelocity simulation, partly because of the large value of the bottom drag however,because of thepresence of thecanyon. for numerical stability. Thesehigh-frequency wavesare 4.2.1. Near-inertial oscillations.The inertial responsein the required resolved bytheobservations andthePOMresults. deepocean consists ofabaroclinic oscillation withvelocities of 1.5 barely 4.2.3. Internal waves. The resultsfrom the POM (Plate 1) m s-• in theupper layer(notshown). Thisoscillation decays to 1 m a persistent ridgein the26øCisotherm overthecontinental s'l duringthe following2 days.This primarilyfirst-mode,show slope. Internal waves propagate both landward andseaward away deepwater oscillation generates flowatthesame frequency overthe Thisprocess is mostapparent duringthePOM slope andshelf. TheSWM-predicted near-inertial oscillation isseen fromthisfeature. interval, beforethe stormgenerates otherbaroclinic in theinterface height(Figure15)ontheslope(waterdepthsimilar spin-up Because of thesimpler topography, stratification, and tomooring 12),whichhasa maximum downward displacement of oscillations.  interfaceelevations of 6 m. Subsequent downwellinglowersthe  8 m at 12 hoursafterthe peakwind.The observed mixed-layer windfield usedin the idealizedmodel,theseinternalwavescanbe  in detailwithSWM. temperature atmooring 12(Figure 5a)increased by 1øCduring the analyzed The SWM resultsshow that interactionof the near-inertial localpeakwinds.The subsequent decrease of 5øCrepresents an with the continental slopeis the dominantgenerating upward movement of approximately 30 m (seeFigure2). This oscillations for the internalwavesover the slopeand shelf.An deflection can only be estimated, however,because the actual mechanism  in theinterface height (Figure16)formsattheshelf temperature profile wasnotmeasured during thestorm. Theprecise initialtrough overtheshelfand lagin timing of theresponse atmooring 12cannot bedeterminedbreak(Y = 310km)whentheflowis westward  16  E 18 o  20 22  ..-  24  o  26  28 3O  0  24  48  72  Simulation Hour  Figure 15.Time series ofinterface depth onthecontinental slope calculated bytheSWM.  96  26,222  KEENANDALLEN:INTERNALWAVESDURINGHURRICANE ANDREW  A. Hour 80  B. Hour 84  360.  310.  260.  210  Y  r.._-. ..................................'7...-7.............................................................--:  1•o  -  _  _  110  60  I  t  I  I  [  150  I  I  I  I  I  I  210.  I  I  I  t  i  I  270.  I  I  330.  i  390.  451  150  210  i  I  270 .  330 .  i  390  450 .  x  C. Hour 92  D. Hour 96  360.  ---• ....... ;---'----'--~'_..L. '._.:--':---' -._•--•---'---'----'---:--:---•. 260  2101 Y  160'I,  11o.  60./i  i  150.  i  i  i  '::::  -  ::..::::: .................................. • -- --_  I  210.  .. '-':• ..................  i  I  270 .  _  i  I 330 .  x  I  I  I  I  I 390 .  I  450.  150.  210.  270.  330.  300  450.  x  Figure 16.Interface height computed bySWMat(a)32 hours after windstopped, contoured from -3.75mto1.75m  by0.25m,(b)36hours after wind stopped, contoured from -5.8mto1.2mby0.4m,(c)44hours after wind stopped, contoured from -4mto1.5mby0.25m,and(d)48hours after wind stopped, contoured from -3.6mto2.8mby0.4 m.Thesolid lines arepositive, thedashed lineisthezerocontour, anddotted lines arenegative contours. Notethatthe  windwasstopped at hour48.  southwestward in deepwater.Thisdivergent flow is seenin the primarilybetween90øWand92øW,a distanceof lessthan200 km. POM results for surface and bottom currents as the storm However, thislengthis significantly largerthanthestormscaleRw,  approaches (Figures 12aand13a).Thetrough subsequently divides whichis40 km;thusthestrong oscillations atmooring 19(Figure (Figure16b),withonewavetraveling shoreward (Y= 320km)and lOd). the other travelingseaward(Y = 260 km). The shoreward Theuniform westward flowalongtheshelfandslopeassociated propagating wavehasa phase speed of0.8ms-•.Theoffshore wave withthedeepwater inertialoscillation generates a waveform withits travels ataphase speed of1.9ms'• withagroup speed of0.7ms-•. crestlocated overthecenter of theslopeat Y = 270km (Figure It isimportant tonotethattheidealized model doesnotpredict 16c).Thecrestpropagates seaward ata phase speed of about1.2m theregional response of theLouisiana shelfandslopeto the s'•asaninternal wave. However, itdoes notsplit (Figure 16d) and hurricane. The internalwavesare generated only wherethe is somewhat broader thantheseaward traveling troughdiscussed deepwater storm flowinteracts withtopography. Asissuggested by above.It hasaboutone-halftheamplitude. Thiswavecannotbe the POM results,this generation mechanism wouldoperateidentifiedin thecomplexresultsfromPOM or in theobservations.  KEEN AND ALLEN: INTERNAL  WAVES DURING  HURRICANE  ANDREW  26,223  4.2.4. Poincar6waves. Downwellinginto MississippiCanyon Holloway[ 1996],because baroclinicratherthanbarotropiccurrents depresses theinterfacedepthin the SWM, just asthe 26øCisotherm force it. flexes downward in POM. After the wind stress is removed, the  interfacerelaxesand generatesa transientflow within the canyon. The wave front associatedwith this flow propagates seawardas a Poincar6wave, followed by an oscillatorytail due to dispersion [Gill, 1982]. Eliminating the internal waves generatedover the sloperevealsthesewaves in the SWM results.Since the internal wavesare parallelto the isobathsfor the simplemodelbathymetry, the interfaceheightfar from the canyon,but at the sameY location, can be subtractedat every grid point. The result is a map of interface dislocationsassociatedwith the Poincar6 waves only (Figure17). The Poincar• waves travel seawardalong the canyon axis at  5. Discussionand Summary  Observedcurrents,temperature,and salinity from moored instrumentson the Louisianacontinentalslope and shelf reveal multiple baroclinicoscillationsduring Hurricane Andrew. In additionto turbulentmixing,thecurrentmetersnearthe stormtrack recordedinternalwavesthat were dominantlynear inertial on the shelfand slope.A barotropicsubinertialsignalalsoindicatesthat continentalshelfwaveswere generatednear the stormtrack.More than 2 Rw west of the track, the responsewas dominantly about 0.5ms-•. Afterexiting thecanyon, thecrests andtroughs are subinertial.Superinertialinternalwaveswere alsopresent.These are supplemented by numericalmodelsin orderto advectedto the west by the near-inertial oscillationsdiscussed measurements in areasnot above,producinga scallopedpath. The group velocity of these identifypossibleinternalwavegenerationmechanisms covered by the observations. waves isestimated tobe0.5to0.7ms-• at150-25 øwestofdirectly The PrincetonOceanModel was run with realistictopography, offshore. Thephase velocity is approximately 1 m s-1,at 250-45 ø westof directlyoffshore.The maximumamplitudeof the interface stratification,and wind forcing. The model-predictedcurrents that deviationof the wavesdecreases from 1 m to 0.5 m duringthe 2.5 revealpowerpeaksat near-inertialand subinertialfrequencies match the observations. A comparison between model-predicted daysafterthewind ceases. currentsand temperaturereveals the presenceof significant The idealized model demonstrates three mechanisms for thatdoesnot appearto be causedby isopycnaldisplacements duringHurricaneAndrew:(1) near-inertial variabilityin theobservations oscillationsin deep water are generateddirectlyby the wind field, local stormwind forcing.These effectsincludeprecipitationand (2) internalwavesparallelto the shelfbreak are producedby the runoff, internal waves and continental shelf waves originating interactionof deepwaterinertial flow with the continentalslope, before the stormand from other areas,deepwaterflows such as and (3) a transientwave is generatedin MississippiCanyonby the eddies, and astronomical tides. Temperature and salinity areespeciallysensitiveto thesefactors. strongup- anddown-canyonflowsdirectlyforcedby the hurricane. measurements The modelpredictseddiesgeneratedby interactionof the storm The wavesinducedwithin the canyondisperseafter exiting the of these canyon and lose amplitude. It is unlikely, therefore,that they flow with MississippiCanyon.However,the development contributeto the oscillationsobservedat the moorings,which are eddies differs substantiallyfrom previousreports [e.g., Klinck, probablydue to the other mechanisms.The secondmechanism 1996;Allen, 1996;Hickey,1997], becauseof the highlyrotational differs subtly from the internal tide generationdescribedby hurricanewindsand the presenceof the MississippiRiver deltato the east. An examinationof the depth anomaly for the 26øC isothermshowsthat an internalwave front is a likely contributorto the observations at the westernmoorings.The modelalsoshowsthe complexinteractionof internalwavesgeneratedat the thermocline with the shelf topography.The thermoclineanomaly, bottom currentsin thecanyon,andtemperature withinthecanyonall reveal 310 internal waves generated within the canyon. However, the complexityof the model predictionspreventsthe unequivocal identification 260  of these waves.  A two-layerisopycnal model(SWM) [Allen,1996] is usedwith idealizedtopography andspatiallyuniformwindsto isolateinternal wavesgeneratedin and aroundMississippiCanyon.This model reproducesmode 1 baroclinicoscillationsassociatedwith the canyonandthe continental shelfand slopeonly. The resultsare in agreement with the POM simulation, but thereis significantly less noise.Internalwavesgenerated overthe shelfbreakin bothmodels resultfrom divergenceof shelfand deepwaterstormflows. These wavespropagateseawardin both models.The SWM resultsalso reveala landwardpropagating phaseand a secondinternalwave locatedoverthe slope,whichpropagates seaward.Along-axisflow within the canyongeneratesPoincar6wavesin both models'the  210  Y  160  11o  Contoured i  150.  i  i  i  !  I  210.  i  i  i  i  i ii  I  I  from i  27f3.  ,i  i  -.5m ]  3313.  i  i  seaward propagation speed is predicted to be0.5 m s-• by both  to 1.1m by .lm ,  i  I  !  390.  I  i  i  i  ,i  459.  x  Figure 17. Normalizedinterfacevariationpredictedby the SWM at hour72. The Poincar6wavesgeneratedin the canyonduringthe stormpassage arevisibleasa curvedcrestandtroughin the middle of thefigure.Seetextfor explanation.  models.The interactionof thesewaveswith the deepwaterflow and otherinternalwavesis revealedin the POM-predictedthermocline anomaly. The combination  of the observations and the results from the  numericalmodelsindicateseveralmechanisms for generatinglong internal waves: (1) near-inertial internal waves were generated acrossthe slopeand shelfby dislocationof the thermoclineby the  26,224  KEENAND ALLEN:INTERNALWAVES DURINGHURRICANEANDREW  wind stress;(2) interaction of inertialflow with topographyHeam,C. J.,andP. E. Holloway, A three-dimensional barotropic modelof generated internalwavesalongthe shelfbreak,whichbifurcated  theresponse of theAustralian northwestshelfto tropical cyclones, J.  intolandward andseaward propagating phases; (3) downwelling Phys.Oceanogr.,20, 60-80, 1990. B. M., The response of a steep-sided, narrowcanyonto timealongthe coastdepressed the thermocline; afterdownwellingHickey, variablewindforcing,J. Phys.Oceanogr., 27, 697-726,1997.  relaxes, aninternalwavefrontpropagates asa Kelvinwave;and(4)  Holloway, P.E.,A numerical modelof intemal tideswithapplication tothe  Poincar6 wavesgenerated withinMississippi canyonpropagate Australian northwestshelf,J. Phys.Oceanogr., 26, 21-37,1996. seaward whilebeingadvected westward overthecontinental slope. Hotchkiss, F. S., andC. Wunsch, Intemalwavesin Hudson Canyonwith  possible geological implications, DeepSeaRes.,29, 415-442,1982. Theseprocesses interactto producea three-dimensional internal J. M., On coastaltrappedwaves:Analysisand numerical wavefield,whichwasonlypartlysampled by theobservations. The Huthnance, use of numerical models with different characteristics has made it  calculation by inverse iteration, J. Phys.Oceanogr., 8, 74-92,1978. Keen,T. R., andS. M. Glenn,A quantitative skillassessment of numerical  possible toidentifyindividual internal wavegeneration mechanisms hydrodynamicmodelsof coastalcurrents,in Estuarineand Coastal andexamine theircomplex interaction. Theprediction of several of Modeling, vol.IV, edited byM. L. Spaulding andR. T. Cheng, pp.2640, Am. Soc.of Civ. Eng.,Reston,Va., 1996. theseinternalwavesby bothmodelssupports theuseof numerical modelskill for modelsto assistin analyzingcomplexbaroclinic flowsin coastal Keen,T. R., and S. M. Glenn,Factorsinfluencing hindcasting shallow water currents during Hurricane Andrew, J. Atmos. regions. Because of thiscombined approach, thisstudyshows that OceanicTechnol.,15, 221-236, 1998.  the baroclinicresponseto hurricanes in shallowwatercan be as important asthebarotropic response.  Keen,T. R., andS. M. Glenn,Shallowwatercurrents duringHurricane Andrew,J. Geophys. Res.,104, 23,443-23,458,1999.  Klinck,J. M., Circulation nearsubmarine canyons: A modeling study,J.  Acknowledgments. The first author was funded under the Naval Geophys.Res.,101, 1211-1223,1996. Research LaboratoryVery High Resolution 4-D CoastalOceanCurrents Kundu, P.K.,An analysis ofinertial oscillations observed neartheOregon project,ProgramElement62435N of the Office of Naval Research.The coast,J. Phys.Oceanogr.,6, 879-893,1976.  second author wasfunded undertheNSERCResearch program. LATEX mooring dataweresupplied bytheMinerals Management Service (MMS). References  Kundu,P. K., A two-dimensional modelof inertialoscillations generated byapropagating windfield,J. Phys.Oceanogr., 16, 1399-1411, 1986. Mellor,G. L., Usersguidefor a three-dimensional, primitiveequation, numerical ocean model,report, 35pp.,Atmos. andOcean Sci.Program, PrincetonUniv., Princeton,N.J., 1993.  of a turbulence closure model Allen,S. E., Topographically generated, subinertial flowswithina finite Mellor,G. L., andT. Yamada,Development forgeophysical fluidproblems, Rev.Geophys., 20, 851-875,1982. lengthcanyon,J. Phys.Oceanogr., 26, 1608-1632,1996. on the continental shelfof Arakawa, A., and V. R. Lamb,A potential energyand enstrophyMillot, C., andM. Crepon,Inertialoscillations the Gulf of Lions-Observations andtheory,J. Phys.Oceanogr., 11, conserving scheme fortheshallow waterequations, Mon.Weather Rev., 109, 18-36, 1981.  639-657, 1981.  processes of energytransfer from Beletsky, D., W. P. O'Conner, D. J. Schwab, andD. E. Dietrich,Numerical Niwa,Y., andT. Hibiya,Nonlinear traveling hurricanes to thedeepoceanintemalwavefield,J. Geophys. simulation of intemalKelvinwavesandcoastal upwelling fronts,J. Phys.Oceanogr.,2 7, 1197-1215,1997.  Res.,102, 12,469-12,477,1997.  ocean model:Withapplication to the Bennett, J. R., A theoryof largeamplitude Kelvinwaves,J. Phys. Oey,L., andP. Chen,A nested-grid simulation of meanders andeddies in theNorwegian coastal current, J. Oceanogr.,3, 57-60, 1973. Cannon, G. A., Wind effects on currentsobservedin Juan de Fuca  Geophys.Res.,97, 20,063-20,086, 1992.  Per6nne, N., J. Verron,andD. Renouard, Rectified barotropic flowovera submarine canyon, J. Phys.Oceanogr., 27, 1868-1893,1997. Petruncio, E. T., Observations and modelingof the internaltide in a submarine canyon,Ph.D. thesis,176 pp., Nay. Postgrad. Sch., in Hurricane Andrew, Gulfof Mexico,finalreport toMiner.Manage.  submarine canyon, J. Phys.Oceanogr., 2, 281-285,1972. Cardone, V. J.,A. T. Cox,J.A. Greenwood, D. J.Evans,H. Feldman, S. M. Glenn,andT. R. Keen,Hindcast studyof wind,waveandcurrentfields Serv., Hemdon, Va., 1994.  Monterey,Calif., 1996.  E. T., Observations of theintemaltidein Monterey Canyon, J. Carrasco, A. C., Intemaltidesin DixonEntrance, Ph.D.thesis, 173pp., Petruncio, Univ. of B. C., Vancouver,B. C., Canada,1998.  Phys.Oceanogr.,28, 1873-1903,1998.  W. H., S. A. Teukolsky, W. T. Vetterling, andB. P. Flannery, Chen,C., andL. Xie, A numericalstudyof wind-induced, near-inertial Press, Numerical Recipes, 963pp.,Cambridge Univ.Press, NewYork,1992. oscillations overthe Texas-Louisiana shelf,J. Geophys. Res.,102, 15,583-15,593, 1997.  Chen,C., R. O. Reid, andW. D. Nowlin Jr.,Near-inertialoscillations over theTexas-Louisiana shelf,J. Geophys. Res.,101,3509-3524,1996.  Chen,X., andS. E. Allen,Theinfluence of canyons on shelfcurrents: A theoretical study,J. Geophys. Res.,101, 18,043-18,059, 1996. Cooper, C., andJ. D. Thompson, Hurricane-generated currents ontheouter  Schott, F.,Spatial structure of inertial-period motions in a two-layered sea, basedonobservations, J. Mar. Res.,29, 85-102, 1971.  Smith,N. P.,Response of FloridaAtlanticshelfwatersto Hurricane David, J. Geophys. Res.,87, 2007-2016,1982.  Smith,P. C., Inertialoscillations nearthecoastof NovaScotiaduring CASP, Atmos.Ocean,27, 181-209, 1989.  A modalanalysis of coastally trapped continental shelf,1, Modelformulation andverification, J. Geophys. Tang,M. T., andR. Grimshaw,  Res.,94, 12,513-12,539, 1989.  Csanady, G. T., Circulation in the CoastalOcean,279 pp., D. Reidel,  wavesgenerated by tropicalcyclones, J. Phys.Oceanogr., 25, 15771598, 1995.  Tang,M. T., P. Holloway, andR. Grimshaw, A numerical studyof the stormsurgegenerated by TropicalCycloneJane,J. Phys.Oceanogr., DurrieudeMadron,X., Hydrography andnepheloid structure in theGrandNorwell, Mass., 1984.  Rh6necanyon,Cont.ShelfRes.,14, 457-477, 1994.  27, 963-976, 1997.  E. F., andV. J. Cardone, Practical modelling of hurricane Fandry,C. B., and R. K. Steedman, Modellingthe dynamics of the Thompson, surface windfields,J. Water.,PortCoastalOceanEng.,122(4), 195transient, barotropic response of continental shelfwatersto tropical cyclones,Cont.ShelfRes.,14, 1723-1750,1994.  205, 1994.  Fandry, C.B.,L. M. Leslie, andR.K. Steedman, Kelvin-type coastal surges generated by tropicalcyclones, J. Phys.Oceanogr., 14, 582-593,1984.  Gill, A. E., Atmosphere-Ocean Dynamics, 662 pp.,Academic Press,San  S.E. Allen,Department ofEarthandOcean Sciences, University of Diego,Calif., 1982. BritishColumbia, Vancouver, BC,Canada V6T 1ZA.(allen@ocgy.ubc.ca) Grimshaw, R., Large-scale, low-frequency response onthecontinental shelf T. R. Keen,Oceanography Division, NavalResearch Laboratory, Code dueto localized atmospheric forcingsystems, J. Phys.Oceanogr., 18, 7322,Stennis SpaceCenter,MS 39529.(keen@ndssc.navv.•nil) 1906-1919, 1988.  Hazelworth, J. B., Watertemperature variations resulting fromhurricanes,(Received May 21, 1999;revisedMay 24, 2000; J. Geophys.Res.,73, 5105-5123, 1968.  accepted July26, 2000.)  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items