Baroclinic Tidal Flows and Inundation Processes in Cook Inlet, Alaska: Numercal modeling and...

8/14/2019 Baroclinic Tidal Flows and Inundation Processes in Cook Inlet, Alaska: Numercal modeling and satellite observations

1/17

Ocean Dynamics (2007) 57: 205221DOI 10.1007/s10236-007-0103-8

Lie-Yauw Oey. Tal Ezer. Chuanmin Hu.

Frank E. Muller-Karger

Baroclinic tidal flows and inundation processes in Cook Inlet,Alaska: numerical modeling and satellite observations

Received: 7 September 2006 / Accepted: 26 January 2007 / Published online: 3 March 2007# Springer-Verlag 2007

Abstract A wetting and drying (WAD) algorithm isimplemented in a baroclinic three-dimensional ocean circu-lation model of Cook Inlet, Alaska, where large tidal ranges

(10 m) regularly expose extensive mudflats. The modelincludes tides and wind- and buoyancy-induced flows. Inthe upper Inlet, the model successfully simulates largeamplification of tides and propagation of fast (34 m s1)tidal bores over shallow mudflats. The simulated returnflows during ebb expose large areas (100 km2) of themudflats. Medium-resolution (250- and 500-m) imagesobtained from the moderate resolution imaging spectro-radiometer (MODIS) instruments aboard the Terra and Aquasatellites were used to verify the model results by identifyingthe location, extent, and temporal changes of the exposedmudflat regions. The results demonstrate the value ofoperational, medium-resolution remote sensing data in

evaluating the WAD model. Sensitivity tests show thatWAD produces approximately 20% larger tidal amplitudeand 10% slower phase than the corresponding experimentwithout WAD. In the deep channel of the central Inlet, theconfluence of saline water of the lower Inlet with brackishwater from rivers and melting ice from land around theupper Inlet produces a salinity front. At the simulated front,

strong vertical circulation cells and surface convergence andcurrents develop, especially during the flood. The char-acteristics resemble those of rip tides often observed in

this region.

Keywords Numerical model. Tides. Inundation.

Satellite data . MODIS. Alaska

1 Introduction

Wetting and drying (WAD) is a common and importantphenomenon of the coastal ocean and estuaries, especiallyover low-lying coastal zones. In bays and inlets with largeamplitude tides, such as in the Bay of Fundy (Gulf ofMaine) and the Cook Inlet (Alaska), WAD is an essential

part of the local environment and ecosystem. In extremecases, the effects of WAD are dramatic and tragic, asdemonstrated recently during the Indian Ocean tsunami(December 2004) and the flooding of New Orleans byhurricane Katrina (August 2005). WAD modeling is clearlyof practical importance, but until recently, WAD schemeshave been implemented mostly in hydraulic and in coastal/estuarine models (Stelling et al. 1986; Balzano 1998;Lynch and Gray1980; Flather and Hubbert1990; Casulliand Cheng1992; Lin and Falconer1997; Ji et al.2001; Xieet al.2004; Chen et al.2006). However, as various regionaland large-scale ocean general circulation models (OGCMs)and forecast systems evolve (Chassignet et al.2003; Drillet

et al.2005; Ezer and Mellor2000; Fan et al.2004; Ko et al.2003; Marchesiello et al. 2003; Romanou et al. 2004;Sheng and Tang2003; Smith et al.2000; Oey et al.2003,2005), the distinction between the coast and open ocean

becomes blurred, and there is a need to include dynamic (ormovable) coastal boundaries to model inundation pro-cesses. The first such OGCM using the Princeton oceanmodel (POM; Blumberg and Mellor 1987; Mellor 2004)has recently been developed and described in Oey (2005,2006). In these latter papers, idealized one-, two-, andthree-dimensional cases such as the dam-break, waterfall, tsunami, tidal, and estuarine-circulation problems

Responsible editor: Jrg-Olaf Wolff

L.-Y. OeyProgram in Atmospheric and Oceanic Sciences,Princeton University,P.O. Box CN710, Sayre Hall,Princeton, NJ 08544-0710, USA

T. Ezer (*)Department of Ocean Earth and Atmospheric Sciences,Old Dominion University,Norfolk, VA 23529, USAe-mail: [email protected].: +1-609-2581318Fax: +1-609-2582850

C. Hu. F. E. Muller-KargerInstitute for Marine Remote Sensing,College of Marine Science, University of South Florida,St. Petersburg, FL 33701, USA


2/17

were tested. The POMWAD implementations rely onthe vast experiences of coastal and estuarine modelers(cited above), but are unique in two important ways.Firstly, the WAD algorithm is implemented in a three-dimensional baroclinic ocean forecast system that isequipped with data assimilative tools, nesting capabil-ities, and large-scale and regional forcing (see http://www.aos.princeton.edu/WWWPUBLIC/PROFS/and pub-

lications listed therein). Secondly, the WAD conditionsacross the cellsinterfaces are fully dynamic using the fullset of primitive equations. In this paper, we describe themodel implementations and results for Cook Inlet, Alaska,

where the tidal range reaches 10 m and active WAD physicsexist. We will also evaluate the simulated WAD processeswith repeated medium-resolution remote sensing imagery.

Cook Inlet, Alaska, is named after the English explorerJames Cook (17281779) who, in 1778, sailed along thenorthwestern coast of North America and the Bering Strait(Beaglehole1974). The Inlet extends approximately 250300 km from the Gulf of Alaska in the south into the city of

Anchorage in the northeast where it branches into twoshallower extensions, the Knik Arm north of Anchorageand the Turnagain Arm southeast of Anchorage (Fig. 1).The water levels and currents in Cook Inlet are strongly

Fig. 1 Map of the Cook Inlet region in the northeastern Pacific (small inset) and the major (crude) topographic and geographic features

206
http://www.aos.princeton.edu/WWWPUBLIC/PROFS/http://www.aos.princeton.edu/WWWPUBLIC/PROFS/http://www.aos.princeton.edu/WWWPUBLIC/PROFS/http://www.aos.princeton.edu/WWWPUBLIC/PROFS/


3/17

influenced by tides from the Gulf of Alaska. The tidalresonant length scale can be estimated by (gHave)

1/2P/4250 km (Gill1982), whereg=9.8 m s2 is the gravitational

constant,P12.42 h is the M2tide period, andHave50 mis the Inlets averaged water depth. This resonant scale isclose to the Inlets length; as a result, tides are significantly

Fig. 2 Observed hourly sealevel (in meters) ata Anchorageand b Kodiak Island during2000

207


4/17

amplified from about 1 m near the Gulf of Alaska,opening to about 5 m in the northern Inlet (Fig. 2). Thelarge tide produces strong currents and tidal bores withspeeds 4 m s1 and peak heights 3 m in the TurnagainArm during a flood (http://www.tidal-bore.tripod.com/usa/turnagain.html), and exposes extensive mudflats through-out the upper Inlet in the subsequent ebb. WAD is,therefore, a common and important phenomenon in the

Cook Inlet. The extensive mudflats are visible fromspecially processed satellite images that will be used tovalidate the model. In the central portion of Cook Inlet,strong oscillatory tidal currents in channels are oftenassociated with the so-called rip tides (Haley et al. 2000;Okkonen and Howell 2003); using the model calcula-tions, we will investigate possible mechanisms for this

phenomenon.In addition to tides and WAD, currents induced by

buoyancy (from rivers) and winds also constitute importantcomponents of the circulation and mixing in the Cook Inlet.Figure3 shows the monthly discharges from seven majorrivers obtained from the US Geological Survey (USGS).

However, tributary discharges are significant, and thecombined high discharge during spring/summer snow meltamounts to over 5 103 m3 s1, or approximately 23times the combined peak discharge shown in Fig. 3. Theriver freshening, as well as the tidal and wind mixing,results in a partially mixed estuary. The surface salinitySinthe lower Inlet near the StevensonKennedy entrance is3032 psu, which is a little less than the salinity valuesfound over the adjacent shelf/slope in the Gulf of Alaska(Xiong and Royer1984). The Cook Inlet/Gulf of Alaskaclimate is extreme. Winter storms associated with theAleutian Low propagate to the Gulf of Alaska from thenorthwestern Pacific (Wang et al.2004). These storms and

continental drainage winds bring high wind speeds (over20 m s1) and frigidly low air temperatures. The meanwinter wind is generally westward. In summer, theAleutian Low is replaced by a high-pressure system(Wang et al.2004), resulting in a generally eastward (andmuch weaker) wind (Royer1975).

The paper is organized as follows. Section 2describesthe Cook Inlet model setup. In Section3, we discuss tides,tidal bores, and WAD processes in the upper Inlet, andcompare the modeled WAD regions with remote sensingdata. Section4examines the role of buoyancy and tides in

producing rip tides in the central Inlet, and Section 5concludes the paper.

2 The Cook Inlet model setup

The model is based on POM: three-dimensional withcurvilinear horizontal grid, terrain-following (sigma)vertical grid and Mellor and Yamadas (1982) mixing. Itincorporates a WAD algorithm that is described in greatdetails in Oey (2005, 2006). At each time step, a cellsface is blocked if the local water depth is less than a

prescribed value (=5 cm is used) by setting the correspond-ing fluxes=0. The requirement that cells must be slightly

wet is consistent with Wangs (1996) numerical stabilitycondition for shallow-water equations. The model solves the

full set of governing equations over wet as well as dry cellsand flows across cell faces satisfy the balance equations,which dictate if the cells become wet or dry at the next timestep. Dry cells are, thus, temporarily dormant, to beactivated depending on volume and momentum conserva-tions in that cell as well as in the neighboring cells (heat andsalt fluxes are also activated for general stratified flows).

WAD simulations require bottom topography from deepwaters to absolute land regions where the orography issufficiently high that grid cells are always dry. The

Fig. 3 a USGS monthly mean river discharges (plotted in theoriginal unit in ft3 s1; 1 m3 s127 ft3 s1) from seven major riversinto the Cook Inlet. b Location of the seven rivers marked on themodel topography (in meters below a land datum; see Oey 2006).The gray represents the absolute land region that is always dry; themagenta represents the WAD region that can be wet or dry

208
http://www.tidal-bore.tripod.com/usa/%20turnagain.htmlhttp://www.tidal-bore.tripod.com/usa/%20turnagain.htmlhttp://www.tidal-bore.tripod.com/usa/%20turnagain.htmlhttp://www.tidal-bore.tripod.com/usa/%20turnagain.html


5/17

transition (of the potential WAD region) is not well defined

for Cook Inlet especially over the mudflats. On the otherhand, the National Oceanic and Atmospheric Administra-tion/National Ocean Services (NOAA/NOS) nauticalcharts (sounding data) do delineate potential WAD regionsfor boaters. We, therefore, interpolated the topography dataset without WAD regions (which was on a 11-km gridkindly provided by Dr. Proshuntinsky, Woods HoleOceanographic Institution, WHOI) with NOAAs potentialWAD regions so that the depth tapers to zero at the nearestabsolute land points. The final depths for all water andWAD regions are then interpolated onto the models

curvilinear grid cells (Fig. 3b). There are 401151 grid

points and grid sizes vary from 1 km near the southernopen boundaries of the model domain to less than 0.5 km inthe north near Anchorage, and in the Knik and TurnagainArms. Sixteen sigma levels are used in the vertical.

Comprehensive climatology that covers the entire CookInlet is not available, yet temperature and salinity (T,S)arerequired for initial and boundary conditions. Therefore, theinitial S is set to 32.4 psu, which corresponds approxi-mately to the shelf/slope estimates off the Stevenson/Kennedy entrances (e.g., Xiong and Royer1984; Okkonenand Howell 2003). With river discharges and mixing by

0 10 20 30 40 50 60-10

0

10

(m/s)

Wind Vectors (6h Avr.)

0 10 20 30 40 50 60-10

0

10

(m/s)

0 10 20 30 40 50 60-10

0

10

(m/s)

0 10 20 30 40 50 60-10

0

10

(m/s)

Day of Year (2000)

a Anchorage

b Nikiski

c Drift River

d Augustine Island

Fig. 4 The observed wind vec-tor stick plot at four stationsfor the first 60 days of theyear 2000; six hourly averageswere calculated from the NOAAmeteorological stations

Table 1 Model experiments

Experiment Stratification Wind River Cz Integration period WAD

2Dpom No No No 0.0025 30 days No

2Dwad No No No 0.0025 30 days Yes

3Dtidehom No Yes No Variablea 120 days Yes

3Dtide Climatology Yes No Variablea 120 days Yes

3Driv Climatology Yes Yes, high Variablea 1 yearb, year 2000 Yes

3Drivsea Climatology Yes Seasonal Variablea 1 yearb, year 2000 Yes

aSee Mellor (2004) and Oey (2006)bOne year without wind, then continued with year 2000 wind

209


6/17

tides, we expect that, after some time (months), the modelInlet will establish its own salinity distribution. Fortemperature, data from the generalized digital environ-mental model (GDEM climatology; Teague et al. 1990)

provides the first guess, which we modify according toNOAA/NOS measurements at Anchorage, Seldovia, andKodiak Island.

Cook Inlets sea level contains large tidal components

particularly M2, S2, K1, and O1 (Fig. 2). The largest ofthese, M2, has an amplitude of about 1 m near theStevenson and Kennedy entrances. To simplify the analysiswithout sacrificing essential WAD physics, most of theresults shown below are from experiments that use a round-number period of 12 h and an average amplitude o=1.2 mat the models southern open boundaries. However, we alsohave repeated some experiments including all tidalcomponents. Open boundary specifications consist of acombination of normal velocity specified as a sinusoid

(amplitude Uno offiffiffiffiffiffiffiffiffiffiffiffiffiffigHmsl

q , whereHmslis the average

depth at 70 m in the lower inlet) and radiation (Oey and

Chen 1992). The same specification is applied to theShelikof Strait open boundary except that when there areriver dischargesQriver, the same outflow is allowed to exitthe model domain at the strait. The zero-phase assumption

between the Shelikof Strait and Stevenson/Kennedy en-trances is an excellent one. Other open-boundary specifica-tions are as follows: one-sided advection for tangentialvelocities, temperature and salinity (velocities= 0 andclimatological T and S during inflow), Orlanskis (1976)radiation for both components for the three-dimensionalvelocity, and radiation based on external gravity phase speedffiffiffiffiffiffiffiffiffiffiffigHmsl

p on the relatively unimportant (because normal

velocity is specified).

We assume the Inlet is ice-free throughout the year, andat the surface, wind stresses are specified and heat and saltfluxes are zero. In the midst of winter (February), the ice-freeassumption is a poor one near shore and in the upperInlet (Moore et al. 2000), and the model results in theseregions in winter are less reliable. That surface heat and saltfluxes are zero is also inaccurate especially in late fall andwinter when strong evaporative losses are expected due tostrong winds and low air temperatures. To partiallycompensate for the missing forcing, model surface tem-

perature is relaxed to climatology with a timescale of1.5 months. The surface salt flux is still zero, however, andthe distribution of salinity inside the model Inlet is assumed

to be dominated by discharge and mixing with more salinewaters from the Gulf of Alaska. We use the NOAA/National Data Buoy Center (NDBC) buoy data and NOSwinds at four stations: Anchorage, Nikiski, Drift River, andAugustine Island; examples are shown in Fig. 4 as windstick plots for the first 60 days of year 2000. These windsare converted to stresses using Large and Pond (1981) andlinearly interpolated onto the model grid. The wind isgenerally weaker in the upper Inlet (Anchorage) than in thelower Inlet (Augustine Island); the channeled wind alongthe Inlet dominates with additional short-period fluctua-tions, as can be seen in Fig.4. There exist, however, finer

spatial variability and transient wind events due to air flowsover mountain ranges and through valleys (Orographiceffects; D. Prentki, 2004, private communication); thesewind events are not captured well by the use of only fourstations.

a Anchorage

b Nikiski

c Kodiak Island

Fig. 5 Example of 5 days (in 2000) sea level variations at (fromnorth to south) a Anchorage, b Nikiski, and c Kodiak Island. Solidlines are observations and dash lines are from model simulationsusing observed tidal forcing at the open boundaries

210


7/17

Discharges from seven of the larger rivers were obtainedfrom USGS (Fig.2). As pointed out above, the values arelow estimates because USGS does not include additionalcontributions from tributaries. Corrections provided by Dr.Prentki were used in the model. The corrected dischargetotals to about 5 103 m3 s1 during the summer maximum.Peak discharges occur in late spring through summer, andminima are from late fall through winter. Okkonen and

Howell (2003) show that salinities measured in the lowerthrough the middle Inlet reach lowest values in fall(OctoberNovember). The lag (between peak dischargeand times of freshest Inlets waters) suggests a tide-riverinemixing time scale of about 23 months. Discharge isspecified as a downward vertical volume flux per unit area,wRiver(Oey1996). For each river, we spread wRiverover a

(somewhat subjectively determined) area in the vicinity ofthat river. For the rivers in the upper Inlet, the spreadingisoften over potential WAD cells (Fig. 3)a situation thatmimics the actual spreading of fresh waters over shallowstreams.

The initial field is one of rest, velocity U=0, S isconstant= 32.4 psu (as mentioned previously), and T iseither stratified in the vertical, area-averaged from clima-

tology, or homogeneous depending on model experiments(below). The water level is initially at the mean sea level(MSL). Tidal velocity forcing and/or wind and/or river arethen turned on. On lateral (absolute) land boundaries,normal fluxes are nil, and no-slip conditions are applied.Bottom friction is computed either by using a bottom dragcoefficient Cz obtained by matching the near-bottom

Fig. 6 Simulation of tides andriver run-off in Cook Inlet, ex-periment 3Driv. Shown here isan enlarged region of the upperInlet. Velocity vectors (black) at

the first grid nearest thesurface are shown superimposedon color images of the corre-sponding salinity at four differ-ent phases of a tidal cycle. Redvectors indicate wind stressesused in the model at the indi-cated dates. Blackcontour in-dicates the coastline and whiteregions show dry (and land)areas

211


8/17

modeled velocity to the Law of the Wall, or simply a constantCz (see Mellor2004and the WAD implementation in Oey2006). Smagorinskys (1963) shear and grid-dependentformula for horizontal viscosity and diffusivity are usedwith their ratio=5 and the Smagorinskys constant=0.1.

We have carried out various experiments to system-atically test the model (Table1). Experiments 2Dpom and2Dwad are two-dimensional barotropic experiments forced

only by tides at the open boundaries, but with the WADscheme turned off (2Dpom) and on (2Dwad), respectively.For these two-dimensional calculations, quadratic bottomfriction formula using the depth-averaged velocities isused, with Cz as a constant=0.0025. Experiment 3Dtide-hom is the full three-dimensional POMWAD but thewater is homogeneous. The integration is for 120 days.Experiment 3Dtide is the full three-dimensional (barocli-nic) POMWAD in which stratification exists because ofinitial, surface, and open-boundary T/S climatology, butriver discharge is nil. Experiment 3Driv turns on riverrunoffs. River forcing is intentionally extreme (to testPOMWAD): Corrected peak discharges (late spring

through summer) from the seven rivers (Fig. 3) arespecified, and they are held steady for 2 years. The totalis 5103 m3 s1. Year 1 is without wind, but with tide, sothat the model establishes a quasi-equilibrium state (insalinity distribution in particular). In year 2, six hourlywinds that correspond to the year 2000 (linearly inter-

polated as described above) are also used. Finally, theexperiment 3Drivsea uses seasonal river inputs instead ofkeeping them steady. The experiments 3Driv and 3Drivseawere conducted with both the single-component and all-component tidal forcing along the open boundariesdescribed previously.

3 Tides and WAD processes in the upper Cook Inlet

3.1 Model simulations of tides and salinity

We focus on experiment 3Driv and discuss other caseswhen appropriate to draw contrasts. Because tidal reso-nance is an essential feature of the Cook Inlets WAD andcirculation dynamics, we first show in Fig.5an example ofthe phenomenon with plots of 5-day modeled and observedsea-levels at the lower (Kodiak), central (Nikiski), andupper inlets (Anchorage). The simulated sea levelcompares well with that which is observed, both in terms

of the phasing and the amplification of sea level fromKodiak to Anchorage. In addition to inadequate modelphysics, discrepancies are in part due to inexact knowledgeof the open-boundary conditions (the Kodiak Island tidestation in Fig.5c is located outside our domain, so a slightadjustment of the imposed tidal phase on the open

boundary was needed). An optimal refinement of thetidal forcing as described in Chen and Mellor (1999)should improve the simulations, but this has not been donein this study.

Figure6 shows salinity distributions at four tidal stagesin the upper Inlet. At the beginning of flood (Fig. 6a),

extensive dry areas (they are really mudflats) can beseen south and southeast of Anchorage in the wide entranceto Turnagain Arm. Dryareas are also seen in the northern

portion of the upper Inlet, especially in the upper KnikArm. Three hours later (Fig.6b), more saline flood waters

begin to fill the upper Inlet. The peak flood currents arestrong, up to 34 m s1 at some location. The flooding iscomplete in Fig. 6c, in which one sees the filling-up of flood

waters in the upper Inlet (all model grid cells are wetatthis stage). Saline waters can clearly be seen to intrude intothe two Arms. Three hours later (Fig. 6d), strong ebbingcurrents ensue with peak speeds of 56 m s1. The strongspeeds (at peak flood in particular) occur in local regionswhere the water layer is thin (


9/17

such as the coastal zone color scanner (CZCS, 19781986)and the sea-viewing wide field-of-view sensor (SeaWiFS,1997present) provide coarse resolution data at about 1-km/pixel. This is insufficient for the purpose of delineatingsmall-scale WAD features. On the other hand, the moderate

resolution imaging spectroradiometer (MODIS) instru-ments aboard the Terra (morning pass, 1999present) andAqua (afternoon pass, 2002present) satellites areequipped with several medium-resolution bands at about250 or 500 m/pixel. These medium-resolution data haveshown unprecedented capability for coastal studies, forexample, in the detection of oil spills (Hu et al.2003) andassessment of coastal/estuarine water quality (Hu et al.

2004; Miller and McKee2004). These data are, therefore,preferred in the present study not only because of theirresolution, but also because of the repeated coverage that isavailable at high latitudes due to the polar orbit of the Terra

Fig. 8 MODIS/Aqua (a) and MODIS/Terra (b) true-color (RGB)images in the upper Cook Inlet: (a) near low tide (23:15 GMT) atAnchorage on July 22, 2005, and (b) near high tide (21:05 GMT) onAugust 27, 2005. The RGB images were composed using the threeMODIS bands at 645-nm (R, 250-m), 555-nm (G, 500-m), and 443-nm (B, 500-m)

Fig. 9 Enlarged MODIS images in the region shown in Fig.8a, fordifferent times as indicated: (a) approximately 2 h before low tide atAnchorage and (b) near low tide. These compare the changes (areasindicated bydashed ellipses) in mudflat exposure during the last 2 hof low tide

213


10/17

Fig. 10 Tidal (M2) amplitudes and phases for the experiment without WAD, 2Dpom ( dashed contours) and with WAD, 2Dwad (solidcontours; also color image) for the upper Cook Inlet

Fig. 11 Central and upper CookInlet topography (colored, scaleis the same as in Fig. 3) andlocations of the sections wherevarious contours are plotted anddiscussed

214


11/17

and Aqua satellites, which allows multiple views of theCook Inlet region within repeated tidal cycles. At thelatitude of Cook Inlet, MODIS instruments provide 24images per day between 21:00 and 23:50 Greenwich meantime (GMT) under ambient sunlight and during some

phases of the tidal cycle. Water absorbs most of the light inthe red and near infrared and is completely opaque in the

mid-infrared. This makes it easy to delineate water fromany other non-water surface. The absorption coefficient(aw) of water in the blue-green wavelengths is less than0.1 m1, but is significantly larger in the longerwavelengths. For example, at 645 and 859 nm (theMODIS 250-m bands), aw=0.325 and 4.19 m

1, respec-tively. Because remote sensing reflectance, R, is inversely

Fig. 12 Along-section velocity in Turnagain Arm (section B inFig.11). Shown are hourly plots starting 1 h after the beginning offlood (Hour=04).Redindicates peak flood (i.e., southeastward flow

toward the shallower region) and blue indicates the peak ebb. Thetidal elevation (m) is also shown and the mean sea level (MSL) isindicated

215


12/17

proportional to aw,R(859 nm) is very small (100 mg l1) in the water or if the bottom is shallow(


13/17

coast of the upper Cook Inlet (upper center of the image).Despite the uncertainty in the detailed model topographyover the mudflat areas, the satellite images are in goodagreements with the predicted dry regions of Fig. 6a.MODIS images are useful also in delineating details of themudflat morphology at different times. Figure9 shows anenlarged image of the mudflat region south/southeast ofAnchorage at two different times: 23:15 GMT on July 22,

2005 (lower panel; same time as in Fig. 8a) and 1 h and45 min earlier at 21:30 GMT (upper panel). While theformer shows almost-completely exposed mudflats (darkshades), the latter shows intertwined light and dark shadesindicative of a more complex morphology due to unevenlysubmerged mudflats (the ellipses in Fig. 9 highlight theregions with significant changes; the larger ellipse is15 km across).

During high tide (Fig.8b, which can be compared withFig. 6c), most of the mudflats mentioned before aresubmerged and covered by murky turbid waters (brownishcolor) and more saline waters (clear and bluish colors) thathave now penetrated from the lower Inlet. The saline

intrusion is particularly pronounced along the deep channelsouth of the Forelands (east of Kalgin Island, lower-leftcorner of Fig. 8b) and will have important dynamicalimplications, to be discussed later in conjunction with themodel results.

3.3 WAD effects on tidal amplitudes and phases

WAD processes alter the surface area and total volume ofan estuary so that tidal amplitude and phases change whencompared to the case without WAD. Figure 10 compares(harmonically analyzed) M2 tidal amplitudes and phasesfor the two cases (2Dpom and 2Dwad), focusing on theupper Inlet only (in the lower Inlet, south of the Forelands,there are only slight differences between the two cases).The comparisons are made only for those grid cells thatare always wet. Large differences can be seen: Amplitudesfor 2Dwad are approximately 20% larger, and the cor-responding phases lag 2Dpom by about 10%. The slower

phase for 2Dwad may be explained by the water-

Fig. 14 Stream function (con-tours) and salinity (colored) at1 h before peak (high) tide in thevertical sections I=180(upperpanel) and I=140 (lowerpanel) in the central Cook Inlet(see Fig.11 for locations). Theview is southwestward towardsthe lower Inlet. Solid (dash)stream-function contours indi-cate clockwise (anti-clockwise)motion in the vertical plane asindicated in the inset. Thestream function is calculated

from the cross-channel compo-nents (Vand W, see the corre-sponding velocity values inFig.15)

217


14/17

retention property of the WAD physics that creates tidalasymmetry between flood and ebb (cf. Oey 2005).Because tidal energy fluxes into the upper Inlet throughthe Forelands are approximately equal for the two cases,the larger amplitudes for 2Dwad are due to its effectivelysmaller storage volume in the upper Inlet due to the

presence of WAD areas.

3.4 Tidal bores in the Turnagain Arm

In the Science Forum website of the University of AlaskaFairbanks (http://www.gi.alaska.edu/ScienceForum/water.html), Larry Gedney observed: Driving the SewardHighway along Turnagain Arm...one may occasionallyobserve a white, frothy line of turbulent water extendingcompletely across the arm and moving slowly upstream...This is a tidal bore. It is created when rising tides in CookInlet encounter the constricted entrance and diminishingwater depths of Turnagain Arm... Because of large tidalranges, currents in the upper Inlet are very strong. During a

flood, the tail portion of the rising tide propagates fasterthan its front; the forward characteristics intersect produc-ing a hydraulic jump that constitutes the model bore.Toexamine this phenomenon, we plot vertical sectioncontours of velocities along the Turnagain Arm (sectionB in Fig.11). Figure12shows the along-section velocity inTurnagain Arm every hour, beginning with 1 h after theflood begins (labeled Hour=04). In the next 6 h (panels),one can see a bore-like structure in which the sea surface

develops a dip of about 2 m, propagating up theTurnagain Arm at about 34 m s1 (1015 km/h). Thesevalues are consistent with those given previously inSection 1. The tourist information (for those waiting toview the bore along the Arm), for example, indicates thatthe bore arrives at Bird Point (the middle of the shallowArm, X6070 km in Fig. 12) about 1 h after it is seenentering the shallow Arm in Beluga Point (X50 km). The

simulations also show detailed variations in the flowvelocity and wave propagation speed along the Arm. In theTurnagain Arm, for example, flood starts in the deeper part(hour=4) when the shallow region is still draining (bluecolors), which produces a salinity front nearX45 km (seeFig.13 below).

3.5 Tidal advection and mixing of salinity in the upperCook Inlet

Figure13contrasts the salinity at the three sections (A, B,and C in Fig.11) during low (left column) and high tides.

The dry areas at low tide are clearly seen for section A.Those at sections B and C cannot be seen, as rivers arespecified and grid cells are always wet along the centeraxes of the Turnagain and Knik Arms. One sees, however,

pockets of water left in local topographic bowlsin theseArms at low tides. In general, water along the Knik Arm isless saline, and the salinity front propagates farther into thedeeper inlet than in the Turnagain Arm. There are tworeasons for that asymmetry between the two Arms. First,

Fig. 15 Vertical sectional con-tours of (a) U, (b) W, (c) V(across-section), and (d) stream

function at the Kalgin Islandsection (I=180; see Fig.11forlocation)

218
http://www.gi.alaska.edu/ScienceForum/water.htmlhttp://www.gi.alaska.edu/ScienceForum/water.htmlhttp://www.gi.alaska.edu/ScienceForum/water.htmlhttp://www.gi.alaska.edu/ScienceForum/water.html


15/17

the river inputs from the Knik Arm are larger than the riverinputs into the Turnagain Arm, and second, the topographyalong the Knik Arm is steeper compared with the flat end ofthe Turnagain Arm, allowing faster flushing out of watersfrom the Knik Arm during an ebb. Note also the strongvertical mixing produced by the tides and strong gradientsacross the entrance to the Knik Arm (upper right panel ofFig. 13). A similar phenomenon of strong cross-channel

density gradients also occur in the central Inlet, which wenow discuss.

4 Processes in the central Cook Inlet and rip tides

Figure11shows the three sections in the vicinity of KalginIsland where strong tidal currents have been observed.Drifters launched in the central Inlet show strong tidallydriven oscillatory flow, whereas the drifters are oftentrappedover the deep channel, moving back and forth fordays along the channels (Dr. Mark Johnson, 2005, personalcommunication; http://www.ims.uaf.edu/research/johnson/

cmi). The strong currents are often referred to as rip tides(Haley et al. 2000; Okkonen and Howell2003), and theyhave scales of a few hundred meters. In the model, thesefeatures are diffused but are still discernible. We now showthat these rip tides are due to strong fronts caused byconfluences of saline (i.e., ocean) and less-saline (i.e.,rivers) water masses. Stratification (horizontal and vertical)is essential for their existence. Figure 14 shows thecontours of stream function superimposed on color imagesof salinity at the Kalgin Island section (marked I=180forthe model grid line number) and another section some40 km to the south (I=140). These contours correspondto 1 h before the peak (high) tide, and one can see the

intrusion of high-salinity water due to the flooding currentson the left side of the I=140 section, as well as in thedeep channel on the left side of the Kalgin Island section(the I=180section). Such salinity intrusion during floodwas previously noted in MODIS image in Fig. 8b. Figure 14shows recirculating cells at both sections but particularly atthe section off the Kalgin Island (I=180); the convergentregion between the two opposite-signed recirculating cellsdefines the front at this stage of the tidal cycle. The salinitydifference across the front is 11.5 psu. Figure 15 givesmore details of the flow field at the Kalgin Island section.The figure shows strong flood (U, the through-section oralong-channel component from southwest to northeast up

the Inlet; Fig. 15a) with a velocity of U

2 m s1

at thefrontal zone and clear convergence shown by the cross-channel velocity (V, Fig. 15c). The upwelling and down-welling velocities (Fig. 15b) of W25 m day1 createcross-channel circulation cells seen by the stream function(Fig.15d).

To test the importance of rivers and associated stratifi-cation in the dynamics of these circulation cells, Fig. 16compares the results of the realistic model case withseasonal river input (3Drivsea) against the experiment(3Dtide) in which river runoffs are null (i.e., there is a weakvertical stratification from the initial GDEM climatology,

but no horizontal gradients associated with fresh waterinput) and also against the experiment without stratification(3Dtidehom). Note that only the wider channel south of theKalgin Island (the left part of Fig.15b) is shown. The frontover the deepest part of the channel is formed only in thecase when there is river discharge (Fig. 16a) and also at

Fig. 16 Comparison of the vertical velocity contours at 1 h beforehigh tide at the Kalgin Island section (I=180; see Fig. 11 forlocation). For clarity, only the eastern portion of that section isshown. The three cases being compared are 3Drivsea, 3Dtide, and3Dtidehom (see Table1for details)

219
http://www.ims.uaf.edu/research/johnson/cmihttp://www.ims.uaf.edu/research/johnson/cmihttp://www.ims.uaf.edu/research/johnson/cmihttp://www.ims.uaf.edu/research/johnson/cmi


16/17

times of flood when large amount of saline water intrudesinto the Inlet through the deep channel (Fig.17). The modelresults in this study are qualitatively similar to theobservations of Okkonen and Howell (2003) taken acrossKalifornsky Beach (Fig. 11), although a more detailedcomparison is not warranted due to the idealized nature of

the model forcing we have used. Haley et al. (2000) alsonote that rip tides occur most often during flood, withstrong current speeds of about 23 m s1 east of KalginIsland. It is quite possible that the higher observed speedsmay be at times caused by the wind forcing, but thecombined mechanism of tidal current and buoyancy,

proposed above, is robust, i.e., the phenomenon repeatsfor each tidal cycle.

5 Discussion and conclusions

A WAD algorithm for POM in its most general three-

dimensional settings with stratification and forcing isimplemented and applied to study the WAD and circulation

processes in the Cook Inlet, Alaska. The model incorpo-rates high-resolution and realistic topography with WADregions. The model includes tidal forcing, river discharge,temperature and salinity fields along the open boundaries,and wind forcing. Cook Inlet is an environmentallysensitive region with harsh conditions; it is an excellent

place to test a model with WAD under extreme forcing thatincludes large tidal ranges and strong currents. Over a tidalcycle, the model simulates complex WAD regions thatcompare quite well with MODIS satellite images despitethe uncertainty of the detailed mudflat regions. In this

regard, we have demonstrated the usefulness of the fine-resolution MODIS images in detecting dry and wetregions, hence, also in evaluating the WAD processes. Inthe future, one may use multiple images to map the mudflatareas at different stages of the tidal cycle (with moreextensive temporal and spatial coverage than that shown inthis paper in Figs. 8 and 9 and with the possible usage ofthe near-IR and mid-IR bands) to improve the representa-tion of the mudflat topography in the model. Identifying

wet and dry pixels from the satellite images (using differentchannels) will provide a more quantitative data set tocompare with wet and dry model cells. The samemethodology may lead to improved flood predictions inother circumstances (e.g., tsunamis, storm surges, etc.). Inthe upper Inlet, the model simulates WAD processes that

affect tidal amplitudes and phases. The model simulates thepropagation of bore-like features over the shallow mudflatsof the Turnagain Arm, with height and propagation speedsin rough agreements with the experiences of localobservers. In the central Inlet, the model shows upwellingand downwelling cells with strong horizontal and verticalvelocities. These strong currents are caused by fronts

produced by the convergence of saline water from thelower inlet through the deep channel with fresher water ofthe upper Inlet. River inputs coupled with tides areessential to the existence of these fronts, as shown bysensitivity experiments with and without rivers. This studydescribes complex dynamics of an inlet with extreme tides

using a model that includes the WAD algorithm. Futureresearch may include further model evaluations anddevelopments of new model components such as sedimenttransports associated with WAD and also wave-currentinteractions (Mellor2003).

Acknowledgment The study was supported by the MineralManagement Service (Contract no. 1435-01-03-CT-72021); Oeyand Ezer were also partly supported by Office of Naval Research(ONR) grants; Hu and Muller-Karger were supported by NationalAeronautics and Space Administration (NASA) grants. The NOAA/Geophysical Fluid Dynamics Laboratory (GFDL) provided compu-tational resources.

References

Balzano A (1998) Evaluation of methods for numerical simulationof wetting and drying in shallow water flow models. Coast Eng34:83107

Beaglehole JC (1974) The life of Captain James Cook. StanfordUniversity Press, Stanford, CA, p 760

Fig. 17 Comparison of thevertical salinity contours at 1 hbefore peak high tide (left panel)and 1 h before peak ebb tide(right panel) at the KalginIsland section (I=180; seeFig.11for location). For clarity,only the eastern portion of thatsection is shown. This casecorresponds to experiment

3Drivsea (see Table1fordetails)

220


17/17

Blumberg AF, Mellor GL (1987) A description of a three-dimensional coastal ocean circulation model. In: Heaps N(ed) Three-dimensional coastal ocean models. AmericanGeophysical Union, Washington, DC, p 208

Casulli V, Cheng R (1992) Semi-implicit finite difference methodsfor three-dimensional shallow water flow. Int J Numer MethodsFluids 15:629648

Chassignet EP, Smith LT, Halliwell GR, Bleck R (2003) NorthAtlantic simulations with the hybrid coordinate ocean model(HYCOM): impact of the vertical coordinate choice, reference

pressure and thermobaricity. J Phys Oceanogr 33:25042526Chen P, Mellor GL (1999) Determination of tidal boundary forcing

using tide station data. In: Mooers CNK (ed) Coastal oceanprediction, coastal and estuarine studies, vol. 56. AmericanGeophysical Union, Washington, DC, pp 329351

Chen C, Beardsley RC, Cowles G (2006) An unstructured grid, finite-volume coastal ocean model (FVCOM) system. Oceanography19(1):7889

Drillet Y, Bourdall-Badie R, Siefridt L, Le Provost C (2005)Meddies in the Mercator North Atlantic and Mediterranean Seaeddy-resolving model. J Geophys Res 110:C03016

Ezer T, Mellor GL (2000) Sensitivity studies with the North Atlanticsigma coordinate Princeton ocean model. Dyn Atmos Ocean32:185208

Fan SJ, Oey L-Y, Hamilton P (2004) Assimilation of drifters andsatellite data in a circulation model of the northeastern Gulf of

Mexico. Cont Shelf Res 24(9):10011013Flather RA, Hubbert KP (1990) Tide and surge models for shallow-

water-Morecambe Bay revisited. In: Davies AM (ed) Modelingmarine systems, vol. I. CRC Press, Boca Raton, FL, pp 135166

Gill AE (1982) Atmosphere-ocean dynamics. Academic, New York,p 662

Haley B, Tomlins G, Smith O, Wilson W, Link M (2000) MappingCook Inlet rip tides using local knowledge and remote sensing.MMS Report, OCS Study, MMS 2000025

Hu C, Muller-Karger FE, Taylor C, Myhre D, Murch B, OdriozolaAL, Godoy G (2003) MODIS detects oil spills in LakeMaracaibo, Venezuela. Eos AGU Trans 84(33):313, 319

Hu C, Chen Z, Clayton TD, Swarzenski P, Brock JC, Muller-KargerFE (2004) Assessment of estuarine water-quality indicatorsusing MODIS medium-resolution bands: Initial results fromTampa Bay, Florida. Remote Sens Environ 93:423441

Ji ZG, Morton MR, Hamrick JM (2001) Wetting and dryingsimulation of estuarine processes. Estuar Coast Shelf Sci53:683700

Ko D-S, Preller RH, Martin P (2003) An experimental real-time intraAmericas sea ocean nowcast/forecast system for coastalprediction. Proceedings, AMS 5th conference on coastal atmo-spheric and oceanic prediction and processes. AMS, New York

Large WG, Pond S (1981) Open ocean momentum flux measure-ments in moderate to strong winds. J Phys Oceanogr 11:324336

Lin B, Falconer RA (1997) Three-dimensional layer-integratedmodeling of estuarine flows with flooding and drying. EstuarCoast Shelf Sci 44:737751

Lynch DR, Gray WG (1980) Finite-element simulation of flow indeforming regions. J Comp Phys 36:135153

Marchesiello P, McWilliams JC, Shchepetkin A (2003) Equilibriumstructure and dynamics of the California current system. J Phys

Oceangr 33:753

783Mellor GL (2003) The three-dimensional current and surface waveequations. J Phys Oceanogr 33:19781989

Mellor GL (2004) Users guide for a three-dimensional, primitiveequation, numerical ocean model. Program in atmospheric andoceanic sciences. Princeton University Press, Princeton, NJ, p 42

Mellor GL, Yamada T (1982) Development of a turbulent closuremodel for geophysical fluid problems. Rev Geophys SpacePhys 20:851875

Miller RL, Mckee BA (2004) Using MODIS Terra 250 m imageryto map concentrations of total suspended matter in coastalwaters. Remote Sens Environ 93:259266

Moore SE, Shelden KEW, Litzky LK, Mahoney BA, Rugh DJ(2000) Beluga, Delphinapterus leucas, habitat associations inCook Inlet, Alaska. Mar Fish Rev 62(3):6080

Oey LY (1996) Simulation of mesoscale variability in the Gulf of

Mexico. J Phys Oceanogr 26:145175Oey L-Y (2005) A wetting and drying scheme for POM. Ocean

Model 9:133150Oey LY (2006) An OGCM with movable land-sea boundaries.

Ocean Model 13:176195Oey LY, Chen P (1992) A model simulation of circulation in the north-

east Atlantic shelves and seas. J Geophys Res 97:2008720115Oey LY, Lee H-C, Schmitz WJ (2003) Effects of winds and

Caribbean eddies on the frequency of loop current eddyshedding: a numerical model study. J Geophys Res 108(C10):3324, DOI10.1029/2002JC001698

Oey LY, Ezer T, Forristall G, Cooper C, DiMarco S, Fan S (2005)An exercise in forecasting loop current and eddy frontalpositions in the Gulf of Mexico. Geophys Res Lett 32:L12611,DOI10.1029/2005GL023253

Okkonen SR, Howell SS (2003) Measurements of temperature,

salinity and circulation in Cook Inlet, Alaska. Report, OCEStudy MMS 2003-036. Mineral Management Service, p 28

Orlanski I (1976) A simple boundary condition for unboundedhyperbolic flows. J Comput Phys 21:251269

Romanou A, Chassignet EP, Sturges W (2004) The Gulf of Mexicocirculation within a high resolution numerical simulation of theNorth Atlantic Ocean. J Geophys Res 109:CO1003, DOI10.1029/2003CJ001770

Royer TC (1975) Seasonal variations of waters in the northern Gulfof Alaska. Deep-Sea Res 22:403416

Sheng J, Tang L (2003) A numerical study of circulation in thewestern Caribbean Sea. J Phys Oceanogr 33:20492069

Smagorinsky J (1963) General circulation experiments with theprimitive equations. Part I: the basic experiment. Mon WeatherRev 91:99164

Smith RD, Maltrud ME, Bryan FO, Hecht MW (2000) Numerical

simulation of the North Atlantic Ocean at 1/10 deg. J PhysOceanogr 30:15321561

Stelling GS, Wiersma AK, Willemse JBTM (1986) Practical aspects ofaccurate tidal computations. J Hydraul Eng (ASCE) 9:802817

Teague WJ, Carron NJ, Hogan PJ (1990) A comparison betweenthe generalized digital environmental model and levitusclimatologies. J Geophys Res 95:71677183

Wang J (1996) Global linear stability of the 2-D shallow waterequations: an application of the distributive theorem of roots forpolynomials on the unit circle. Mon Weather Rev 124(6):13011310

Wang J, Jin M, Musgrave DL, Ikeda M (2004) A hydrological digitalelevation model for freshwater discharge into the Gulf of Alaska.J Geophys Res 109:C07009, DOI10.1029/2002JC001430

Xie L, Pietrafesa LJ, Peng M (2004) Incorporation of a mass-conserving inundation scheme into a three-dimensional storm

surge model. J Coast Res 20:1209

1223Xiong Q, Royer TC (1984) Coastal temperature and salinity in thenorthern Gulf of Alaska. J Geophys Res 89:80618066

221
http://dx.doi.org/10.1029/2002JC001698http://dx.doi.org/10.1029/2005GL023253http://dx.doi.org/10.1029/2003CJ001770http://dx.doi.org/10.1029/2002JC001430http://dx.doi.org/10.1029/2002JC001430http://dx.doi.org/10.1029/2003CJ001770http://dx.doi.org/10.1029/2005GL023253http://dx.doi.org/10.1029/2002JC001698

Baroclinic Tidal Flows and Inundation Processes in Cook Inlet, Alaska: Numercal modeling and...

Documents

Transcript of Baroclinic Tidal Flows and Inundation Processes in Cook Inlet, Alaska: Numercal modeling and...