Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/237509304
CITATIONS READS
291 182
2 authors, including:
Werner Alpers
University of Hamburg
206 PUBLICATIONS 6,297 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Ecosystem Approach to the management of fisheries and the marine environment in West African
waters View project
All content following this page was uploaded by Werner Alpers on 14 October 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 89, NO. C6, PAGES 10,529-10,546,NOVEMBER 20, 1984
INGO HENNINGS
ing the radar reflectivity modulation is a viable first-order ing of bathymetry is not confined to low wind speeds,as is
theory, which could be the starting point of a future, more often stated.Therefore the observedstrongcross-sectionmod-
complete(multiparameter)theory. ulation
cannot
beattributed
to a threshold
effect
forripple
For real aperture radar the imaging mechanismis deter- wave generationby the wind.
mined only by cross-sectionmodulation. However, for syn- 5. Radar imagingof large-scalesubsurfacefeaturesseems
thetic apertureradar, phasemodulation,i.e., velocity bunch- to depend only weakly on wind direction. This implies that
ing, also contributesto the imagingprocess[seeLarsonet al., blockingof Braggwavesin an adversecurrentcannotbe the
1976; Alpers and Rufenach,1979; Alpers et al., 1981; Alpers, dominant imagingmechanism,becausethis effectstronglyde-
1983]. This contribution to the SAR imaging mechanismis pends on the angle between wind and current direction
briefly reviewedin section2.3. [Gordon
et al.,1983;Lyzenga
et al.,1983].B]oc.
king.0ccurs
In section3 we confront our theory with existingexperi- onlyunderveryspecialconditions,
and thusradarimagingof
mental data, includingvisualobservationsdocumentedin sea bottom topography should exhibit a resonant-typebehavior.
charts,X band RAR imageryof asymmetricsandwavesin the However,the data we have analyzedseemnot to supportthis
North Sea off the Dutch coast,and L band SeasatSAR imag- hypothesis.
ery of sandbanksin the Southern Bight of the North Sea 6. In general,L and X band radar signaturesof sandbanks
(northeasternapproachto the English Channel).The bathy- and sand waves have a double sign. This means that on the
metry information is extracted from depth charts, and the radar image, one side of the subsurfacefeaturesis associated
current velocityfrom a tidal streamatlas. with increased,and the other sidewith reducedgray level(see,
Finally, section4 containsa summaryand discussionof the for example,Figures5a and 6a). Often the positiveand nega-
proposedimaging theory. tive excursionsfrom local averagegray level are almost sym-
metric. This implies that turbulent interaction of the current
2. THEORY OF THE IMAGING MECHANISM
with the ripple wavescannot be significant,becausethis pro-
A theory of the radar imaging mechanismof underwater cesswould always lead to an attenuation of the short waves
bottom topography(bathymetry)has to beginwith the follow- and thus a reduction of the spectralenergy density of the
ing facts. Bragg waves,but never to an enhancement.
1. Underwater bottom topography is imaged by real and 7. The radar signatureof bathymetryrespondsvery rap-
syntheticaperture radar. This implies that imaging of bathy- idly to changingtidal flow conditions[de Loot, 1978;McLeish
roetry by SAR cannot be causedexclusivelyby phaseor veloc- et al., 1981]. For example,in the SouthernBight of the North
ity bunchingmodulation.Thereforeamplitudeor cross-section Sea no bathymetricallyinducedradar signaturesare observed
modulationmust be an important factor in radar imagingof at the turning of the tide, i.e., at slackwater [Lodge, 1983b].
bathymetry. This implies that the vorticity in the residualcurrent field
2. Microwavesemitted by the radar penetrateinto seawa- generatedby tidal flow over bottom topography [Zimmer-
ter only to a depth which is small in comparisonto the elec- mann,1978; Kornenand Rieprna,1981; Robinson,1981] cannot
tromagnetic wavelength. Consequently,underwater bottom contribute significantlyto the imaging mechanism(G. J.
topography can only be sensedindirectly via surfaceeffects. Komen, personalcommunication,1983).By definition,the re-
Sincethe radars co,3sidered here operate at incidenceangles sidualcurrentis both stationaryand largescale.
between20ø and 70ø, the radar reflectivityof the seasurfaceis Theoretical estimates and measurements of the residual vor-
dominated by Bragg scattering (see, for example, Wright ticity in the SouthernBight of the North Sea yield valuesof
[1968, 1978], Valenzuela[1978], and Alperset al. [1981]). For the orderof 10-7 to 10-6 S-• [Komenand Rieprna,1981],
Braggscattering,the radar backscattering crosssectionis pro- whereasthe velocitygradientsassociatedwith the mean tidal
portional to the spectralenergydensity of the Bragg waves, flow over bathymetryare typicallyof the order of 10-4 to
whichhavea wavelengthof 20/2 sin {D,where20 is the radar 10-3 S-x. Thusin the SouthernBightof the North Seathe
wavelengthand {D the incidenceangle. Consequently,cross- modulation causedby residualvorticity should be at least 2
sectionmodulationresultsfrom modulationof the Braggscat- ordersof magnitudesmallerthan the modulationcausedby
tering waves. the mean tidal flow (seeequation(34)). However,the situation
3. Radar imaging of bathymetry is only observedwhen may be different in the Nantucket Shoals(G. R. Valenzuela,
strong (tidal) currentsare present.Therefore one is led to the personalcommunication,1984).
hypothesisthat underwaterbottom topographymodifiesthe
current field at the surface and that the radar senses these 2.1. Current-Bottom TopographyInteraction
current variations. (Note that wind-generated currents are The interaction of a three-dimensiona/time-variable current
usually confinedto the upper layers of the ocean and there-fieldwith a three-dimensional underwaterbottomtopography
fore, in general,do not interactwith bottom topography.)In can sometimesbe a very complex processwhich does not
the caseof RAR imaging,the currentfield is sensedindirectlyallow a simple mathematical description [Loder, 1980].
via its effecton the short-scalesurfaceroughness.
The currentNevertheless, in this paper we make the simplestpossibleas-
fieldinteractswith the surfacewavesand thuscausesa spatial sumptionthat the currentflow abovethe bathymetryis lami-
modulation of the Bragg scatteringwaves. However, in the nar, free of any vertical current gradient, and only weakly
case of SAR imaging, in addition to the above mentioned time-dependent.Furthermore, we assumethat the tidal veloci-
amplitudeor cross-sectionmodulation,phasemodulationmay ty component U. normal to the direction of the underwater
also contributeto the imaging.The phasemodulationmecha- ridgeor bank obeysthe continuityequation
nism does not require amplitude modulation of the Bragg
waves(seesection2.3). U_•(x_•)
d(x_•)= const= c (la)
4. Underwater bottom topography has been observedby
andthattheparallelcomponent
U IIremains
constant,
Ka bandreal apertureradar evenat wind speedsof 10 m s-•
(Beaufort4) [McLeish et al., 1981]. Consequently,RAR imag- U II = const (lb)
ALPERS
ANDHENNINGS'IMAGINGMECHANISMS
OFBOTTOM
TOPOGRAPHY 10,531
Here d(xO denotesa depth profile along a line perpendicular In the specialcasewhere S = 0, (2) togetherwith (4a) and
to the ridgedirection,x•_. (4b)statesthat action densityN is conservedalong a ray path
Current measurements on and off the sandbank South Falls in four-dimensionalphasespace.This is a well-known equa-
by Vennand Olier [1983] in the SouthernBight of the North tion in many fields of physicsand is applicablewhen waves
Sea (northeasternapproach to the English Channel), where propagatefreely in a slowly varying medium.As an example
the tidal currentflows acrossthe bank at an obliqueangle, we mentionherethe propagationof Alfv6nwavesin interplan-
have confirmedtlaat (la) and (lb) are acceptablefirst-order etaryspace[Vb'lket al., 1974].
approximations.However, deviationsfrom thesesimplerela- In the caseconsideredherethe shortBraggscatteringwaves
tions are observed,and a more refinedtheory shouldaccount are subjectnot only to a variable surfacecurrent but also to
for them. Note that (la) and (lb) imply that the tidal stream the action of the wind, to nonlinearenergytransferto other
vector changesdirection when the flow crossesa sandbank at waves, and to nonlinear dissipationprocesses[Hasselmann,
oblique angles: when approaching the crest it is deflected 1972; Hasselmannet al., 1973, 1976]. If no variable current
away from the bank direction,and when leavingthe crestit is were present, then the short surfacewaves would be in local
deflected back toward the bank direction. equilibrium,and their spectrumwould be the Phillipsequilib-
Having reducedthe bathymetry-currentinteraction to the rium or saturation range spectrum[Phillips, 1977, pp. 140-
verysimpleequations(la) and (lb), we havenow to developa 159-].However, a horizontal current gradient perturbs this
theoreticalmodel to relate the amplitudemodulationand the equilibriumand causesa modulationof the energyspectrum
phase modulation to the surface current variation. With re- of the short waves.
spectto cross-sectionmodulation,we again make drasticsim- We assumethat the variable surfacecurrent leads only to
plifying assumptionswhich we believeare justified in a first- smalldeviationsof the actiondensityfrom equilibrium.There-
order theory. With theseassumptionsthe current-short-wave fore we write the action densityN and the surfacecurrent U
hydrodynamicinteractioncan be describedby a simpleequa- as sumsof a constantequilibriumterm and a time-dependent
tion which is easilytractablemathematically. Such a hy- perturbation term
drodynamicmodulationtheoryis presentedin section2.2. The
phase modulation, however, is a theoretically well-known N(x, k, t) = No(k) + 6N(x, k, t) (6)
imagingmechanism[Larsonet al., 1976;AlpersandRufenach, U(x, t)= Uo + •U(x, t) (7)
1979'Alperset al., 1981'Alpers,1983]andis brieflyreviewed
in section 2.2. Furthermore,we approximatethe sourceterm S by a diagonal
operatorand keep only the linear term in a perturbationseries
2.2. HydrodynamicModulation expansion [Alpers et al., 1981]. In this approximation the
Sincethevariation
of thesurface dueto interaction transportequation(2) reads
current
with bottom topography has spaceand time scalesthat in dN/dt = - !•gN (8)
generalare small in comparisonto the spaceand time scales
of the shortripple waves,the current-Braggwaveinteraction where# is a parameterwith dimension (time)-1.The parame-
can be describedby a Wentzel-Kramers-Brillouin-type inter- ter p is calledthe relaxationrate,and z, = p- • the relaxation
action theory [Schiff, 1955; Longuet-Higgensand Stewart, time. One refersto (8) as the relaxationtime approximationto
1964; Whitham,1965; Bretherton,1970; Willebrand,1975]. In (2) [Keller and Wright, 1975; Wright, 1978; Alpersand Hassel-
this theory the transportequation,which describesthe vari- mann,1978, appendixB]. Physically,z, is the responsetime of
ation of the spectralenergydensityof shortwavesin a slowly the wave systemto current variations.It is determinedby the
varying current field, is the action balance or radiation bal- combinedeffect of wind excitation,energy transfer to other
ance equation [Hasselmannet al., 1973; Keller and Wright, waves due to conservative resonant wave-wave interaction,
1975;AlpersandHasselmann, 1978;Wright, 1978].This equa- and energyloss due to dissipativeprocesseslike wave break-
tion reads ing. No measurementsof the relaxation time in the open
oceanexist.However,from theory we expectthat z, is of the
dN
order of 10-100 wave periods.Applied to SeasatSAR Bragg
dt -Ze[N]=•+:i•xx+ •-• N=S(x,k,t) (2) waves,which have a wavelengthof 34 cm and a wave period
where of 0.47 s, this meansthat r, shouldlie in the rangebetween4.7
and 47 s. In this paper we considerz, (or p) as a free parame-
N(x, k, t) = E(x, k, t)/co' (3) ter.
where
c•No
ZeENo]= •k
co(x,k, t) = co'(k)+ k. U(x, t) (5)
c•U c•No
denotesthe wave frequencyin the movingmediumwith vari-
able velocityU(x, t).
-k.
c•x
• ki•Ui
i,j=1,2
c•No (11)
•Xj cqkj
10,532 ALPERS
ANDHENNINGS:
IMAGINGMECHANISMS
OFBOTTOM
TOPOGRAPHY
Thesecond
identityfollows
wheninserting
(4b)and(5)for1•. caseof short-longsurfacewaveinteraction.The dominanceof
The secondterm of the left-hand side of (9) containingthe the local time is the reasonwhy the modulationof the short
perturbedactiondensity6N gives wavesby the long surfacewavesis relativelyweak: the dwell
time of short wavesin flow regionswith positiveor negative
strain rates OU/Oxis determinedby the period of the long
= + (%+ u0). ON surfacewaves.Therefore the weak hydrodynamicinteraction
has not sufficienttime to build up a strong modulation. Note
whereeg= Oc0'/Ok
is theshort-wave
groupvelocity.
Herethe
terms that after half the period of the long surfacewave the short
wavesenter from a convergentinto a divergentflow regime
O• O O(k. gU) OgN and vice versa. Consequently,the modulation of the short
.... 6N= (13)
0x 0k 0x 0k waveschangessignat twicethefrequencyof the longwaves.
However, the situation is different for the present case of
and short-wave interaction with surface currents over bottom to-
pography.Here the dwell time of short wavesin regionsof a
6U.•xx6N (14) givensign of the strain rate is determinedby the advection
time, which typically has valuesrangingfrom severaltens of
havebeenneglected
because
they are of secondorder.Thus(9) seconds to several minutes. Thus the weak interaction can act
reduces to longer on the short waves and can potentially generatea
strongermodulation.However,in the caseof tidal flow over
[0 0x+#I6N=+k.OU
•+(%+U0).•x Ox'ONo
Ok (15) large-scalesubsurface features,like sandbanks,the advection
time can be so large that it is evenlarger than the relaxation
The time scales of the three terms on the left-hand side are time. In this case the relaxation time determines the short-
wave modulation. The relaxation time is a measure of how
given by the local time T, the advectiontime %, and the
relaxationtime z,. The local time is of the orderof the period much disequilibriumthe short-wavesystemcan endure.The
of the semi-diurnal tide divided by 2•r, which is 12.5 short wavescannot be strainedinfinitely; the limit is given by
hours/2n• 2 hours.The advectiontimeis givenby •r'
We now solve(17) by Fourier transformation.By introduc-
z• = I(%+ U0). Kl-' (16) ing the Fourier representations
where K is the wave number of the bottom topography.The
advection time and the relaxation time are always small in
comparisonto the local time (xa and zr are typicallyof the
•N(x,
k)=;;o•
n(K,
k)e
iKx
dK (18a)
order of 1 min or smaller).Therefore the first term of the
gU(x)= u(K)e
'KxdK (18b)
left-handsideof (15) can be neglectedagainstthe secondand
third term. Thus (15) reducesto
(17) becomes
I(%+ U0)'•xx
c91
+ # 6N= k. cgU
O'•'cgNø
O'•- (17) n(K, k) =
K. (% + Uo)+ i#
[K. (%+ Uo)]2 + #2 ß[k.u(K)](K.
0No•
In an explicit tensor notation the right-hand side of (17) is (19)
given by (11). Equation (17) statesthat the sum of advection
and relaxation of the action density perturbation is balanced If we assumea power law dependenceof the short-wave
by "straining" exerted on the wave systemby the spatially energyspectrumœo= co'Noof the form œo~ Ikl-• (• is equal
varying current U. to 4 if a Phillips equilibrium spectrumis assumed),then (19)
As has alreadybeennoted in connectionwith (11), the right- becomes
hand side of (17) representsa refraction term. It originates
from the fact that the spatially variable current U refractsthe n(K, k) K. (%+ Uo)+ i# K. k
short waves,i.e., changestheir wave number. A wavenumber No [K.(%+Uo)]
2+#2(•+7)-•- [k.u(K)]
changecausesa local perturbation of the equilibrium action (20)
(or energy)spectrumof the short waves,becausethe equilibri-
um action densityspectrumNo varies as a function of wave where7 is given by
number.Thus the steeperNo is as a functionof wave number, k 8co'
the largeris the modulationgN/No = gE/Eo. -
co' Ok
Furthermore, the modulation is proportional to the velocity
gradient(or the strain rate) OU/Ox.Note that the left-hand with k = Ikl.
side of this approximationto the spectraltransportequation Inserting the dispersionrelation for water waves
(17) is differentfrom the correspondingleft-handside of the
approximation used for describingthe weak hydrodynamic
co'= gk +- k3 (22)
interaction betweenshort and long surfacewaves [Alpers and P
Hasselmann,1978]. In the latter casethe local time T is given
by T = f•-•, wheref• is the radian frequencyof the long where •/denotes the accelerationof gravity, z the surfaceten-
surfacewaves.This impliesthat T is typically of the order of 2 sion,and p the densityof water, 7 reads
s, which is short in comparisonto % and zr. Therefore the
partial time derivative term cannot be neglected in (15). 11 + 3k2. z/•tp
Indeed, it is just this term which dominatesthe othersin the 7=• k2z/gp (23)
ALPERSAND HENNINGS'IMAGINGMECHANISMS
OF BOTTOMTOPOGRAPHY 10,533
•a(x)
_f••Mhy&(K,
k•)u•(K)eiKx
dK(29)ationtimezr = #- •, i.e.,
advectiontime % (see(16))is smallin comparisonto the relax-
then,according
to (27)and(28),Mhydr(K
, kx)isgivenby ß• • IK. (%+ Uo)l-• <<• (32)
4+7 If condition(32) applies,then (30) reducesto
Mhydr(K'
kx)
= --E++E_
4+7
Mhydr(K
) = -iK•, (33)
K.
+ EK.(Uo
--
(Uo + g- (3o) c•a(x)
ao
4 + 7 c•U•,(x)
# c•x
(34)
where
This relationshipbetweencross-sectionmodulation and cur-
E+ = Eo(+2ko) (31a) rent gradientconstitutesthe main result of this paper. Note
that in this approximationthe modulationis independentof
%+ = %(+2ko)= -%(- 2ko) (3lb) wind direction.
This MTF relates the current variations in the look direction Now we want to investigatein more detail the parameter
of the antenna to the cross-section modulation. rangefor which(34) is a valid approximation.Restrictingour-
Equation (29) together with (30) allows the calculation of selvesto the SeasatSAR case,we expectthat the relaxation
10,534 ALPERSAND HENNINGS'IMAGINGMECHANISMS
OF BOTTOMTOPOGRAPHY
rate of the 34-cm SeasatSAR Bragg waves is larger than 0.02 wheregrad• d is the gradientof the depth profile in the direc-
s-x (see the discussion
following(8)). Thereforeif the in- tion perpendicularto the bank crest,IUol do cos 'v-
equality do= c is the constantdefinedby (la), IUolis the modulusand
1 0.02 Uoñ is the normal componentof the undisturbedcurrent ve-
I(%(+
2ko)
+ Uo)'KI< (1%
+1+ IUol) locity,do is the water depth outsidethe bank area, and • is
the anglebetweenthe (undisturbed) flow and the x•_direction
=5 x 10-3s-X (35) (seeFigure 1). If we insertthe gradientof the depthprofilein
is fulfilled,then the relationship(32) is approximatelysatisfied. flow directiond' into (39), we obtain
Here L denotesthe length scale(not the wavelength)of the
6{7 4+7 d'
slope of the subsurfacefeature in the cross-trackor x direc-
tion.Inserting
thegroupvelocity
leg(___
2ko)l= 0.36m s-• and {70 - It lgol
doCOS2
• •'• (40)
a typicalvaluefor thesurface
currentvelocity[Uol= 1 m s-x, Thisequationwill be usedwhencomparingSeasatSAR image
we obtain from (35) the condition L >• 300 m. This condition intensityvariationswith bottomtopography.
is fulfilled for the sandbanks in the Southern Bight of the If the real apertureradar is a linear imagingsystem,then
North Sea,which we are considering. the relative RAR image intensitymodulation(6I/Io)RARis
However,the generalexpressionfor the cross-sectionmodu- equal to the relative variation of the radar crosssection
lation due to hydrodynamic interaction contains both the
relaxation and the advection term. The effect of the advection (C•I/Io)RA
R= (c•l/Io)hydr
= 6{7/6
0 (41)
term on the imagingmechanismamountsto the addition of a Sincethis modulationis due to hydrodynamicinteraction,we
low-passfilter to the imaging. The transfer function of the alsousethe subscripthydr,synonymously with RAR.
low-passfilter in two-dimensionalK spacereads Accordingto (39), RAR shouldimagebottom topography
best when the current flow is in the cross-track direction and
K. (% + Uo) + iit It
r = ß- (36)
whenthe topographicfeaturesare alignedparallelto the flight
[K. (%+ Uo)]2 + ]22 i
direction.However,bottomtopographyshouldnot be imaged
The bandwidthKc of this low-passfilter can be definedby
by RAR when the flow is parallel to the flight direction
IKc' (% + Uo)l= It (37) (• = 90ø)or whentopographicfeaturesare alignedin the flow
direction(•b -- 90ø).
A typical value for the cutoff wavelengthAc = 2•/IKcl is Ac = However, some caveatsare associatedwith the predicted
250m, whichis obtained wheninserting
I%+ Uo]= 1 m s-a strongdependence of the modulationon azimuthangle•b(see
andIt = 0.025s- • into(37),andassuming
K (%+ Uo).Thus (40)).This strongazimuthaldependence is a consequenceof
the cross-sectionvariations due to hydrodynamicinteraction the fact that we have neglectedin our model the interaction
representa low-passfiltered replica of the current gradient betweenshortsurfacewavestravelingin differentdirections.If
field OU,,/Ox. this interaction were included, then we would obtain a weaker
In this paper we shallrestrictourselvesto thosecaseswhere azimuthaldependence
of the modulation.Thereforewe expect
(34) is applicable,i.e., where the advectiontime is large in that bottom features become detectable even when the current
comparisonto the relaxation time. flows parallel to the flight direction(•b= 90ø).However,for
In termsof the componentsof the flow velocityperpendicu- •b= 90ø the radar signatureshould be much weaker than for
lar andparallelto thebankcrest,Ul and U II,(34)reads 4,-0 ø.
ao
6•= 4+7cøs
2• OU•
•
the targethas a velocitycomponentUr in the rangedirection,
then zero Doppler frequencyis encounteredat an azimuth
angle of
OU,
---•hydr
0X.i
- (38b)
• • tan • = Ur/V (42)
However, a word of caution should be added here. It is
awayfrom the beamcenter(V is the platformvelocity).If the
permissibleto replace (38a) by (38b) only if the simple one-
range of the target is R, then a SAR tuned for stationary
dimensional model for describing the bottom topography-
targetsdisplacesa movingtargetin azimuthby
currentinteractionis applicable.Often this interactionis more
complex, and in this casea full three-dimensionaldescription 6y = -R(U/V) (43)
is required.
If the radial velocity of scatter elementsvaries in azimuth
Inserting the continuity equation (la) into (38b) yields
direction(c•U,./c•y
v• 0), then the displacement6y of the scatter
_ grad elementsin the SAR image is nonuniform.There are regions
{706{74+•IUol
It do
cos
• COS
2• d2 (39) in the image which have a deficiencyand others which have
ALPERS
ANDHENNINGS:
IMAGINGMECHANISMS
OFBOTTOM
TOPOGRAPHY 10,535
4+78U,
=--
(5l(y)•
Rsin •cos4(
O{sin •U• •U•l.
•'• v3xH
) # 8x
0.3 (51)
•I = Rsin
(•) v.b.OIUol
do
cos
• cos
• sin
V • gradx•
de (6I/Io)sAR
= (6I/Io)hy,•+ (6I/Io)v.b.
Note that both modulationscan have positiveas well as nega-
(54)
(47) tive signs.This implies that the total SAR image intensity
modulationis either enhancedor reducedas comparedto the
RAR imageintensitymodulation.
currentprofile currentprofile Now we want to obtain an estimateof the relative mag-
type1 (•) type2 nitude of hydrodynamicand velocity bunchingmodulations.
Comparisonof (38b) and (46b) showsthat both modulations
I t,dol
flow
direction
flow
direction are proportionalto 8U,/Sx, and that the proportionalityfac-
U,{Yo)
4,
tt'f ]Ux(Yo) tors are
Phydr
= 4+7cøs
24) #
imo,
ge plone
Y
in the caseof hydrodynamicbunchingand
R
/•v.•.=- sin O cos 4) sin 4)
V
/ imoge • õI RdUx•
/ intensitylI I T•=V•-I I I
in thecaseof velocitybunching
modulation.
ThefactorJfihy•rJ
has its maximumat 4• = 0ø (4• is the anglebetweenthe flight
flig?direction and the sandbank direction), and at •b= 45ø. With
V # = 0.025s-•, 7 = 0.5, R/V = 130s, and (} = 20ø (applicable
Fig. :2. Schematicplot of syntheticaperture radar (SA•R)image to SeasatSAR) we obtain the following maxima of the pro-
intensitymodulation duo to velocity bunching:(a) two current pro- portionality factors:
files of oppositesigns,which are typical for tidal flow over sandbanks;
(b) corresponding imageintensityprofilesin flight direction. Ifihyarlmax
= 180s Ifiv.•.lmax
= 22 s (55)
10,536 ALPERS
ANDHENNINGS'IMAGINGMECHANISMS
OFBOTTOM
TOPOGRAPHY
ALPERSAND HENNINGS.'IMAGING MECHANISMSOF BOTTOMTOPOGRAPHY 10,537
,::::: .
ß .
:..; ::
ß.. i;'
:'.::..
..... .......
filled. If this is not the case,then the relative image intensity I = Io{1+ $I/Io) i•--•-• I {divergent)
modulation due to velocitybunchingis given by [Alpers and
Rufenach,1979; Alpers,1983]
6!
(6l(y)'• p. exp-- Io dx d2 •'o•1 <o
F.
...
0 .
"'
..
"5 '1'0km.
English..
• --.- coast
f{:ight direction
•.*.:... '%.....
ß
! -,. /
$1'00'N Z'OO'E
Fig. 5b. Simplifiedsea chart of the sea area shown in Figure 5a togetherwith the tidal velocityfield encountered
during the Seasatoverflighton August19, 1978,0646 UT. The numbersabovethe tidal streamvectorsdenotethe velocity
in centimetersper second,and the numbersbelow denotethe directionrelativeto N. The dashedlinespresentthe 20-m
depthcontours.
The windwasblowingfromtheSE at 4 m s-:.
These images, which were taken by the Dutch EMI side- downstream of the sand wave crests,which are associatedwith
looking airborne radar (SLAR) (X band real aperture radar), weakly convergentcurrents,exhibit a slightly increasedsur-
are reproducedin Figures3a and 3b. face roughness.Therefore they appear on the image with a
In this area of the North Sea the water depth is 20-25 m, slightlybrightergray tone level.
and the range of the tide is 1.5-2 m. During the times of the In the secondcase(Figure 4b), where the flow direction is
overflights,'
tidal currentvelocities
were+(0.4-0.5)(Figure3a) reversed,the steep slopesare located downstreamfrom the
and -(0.6-0.7) m s-• (Figure3b),respectively.
The heightof sand wave crests.The steepsloperegionsare now associated
the sand waves above seaflooris typically 4 m, and the dis-
tance between successivecrests,lying at about 90ø to the
coast,is 300-500m. Therefore
I(%+ Uo).Kt-x _>50 s, and
tidal flow
(34) shouldbe applicablefor describingthe cross-section
mod- [•]
ulation. J$ direction
Furthermore, the height profiles of the sand waves are Td[mid' ./'•
asymmetric:they have their steepslopesto the N-NE [Hou- 0.5
with stronglyconvergent surfacecurrentsin whichthe spectral The above exampleclearly showsthat suchradar imagery
energydensityof the Braggwavessharplyincreases aboveits can be used to infer the direction of the asymmetryof sand
averagevalue.Consequently, thesesteepsloperegionsappear waves.However,the presentRAR imagerydoesnot allow the
as thin, distinctive,bright streaksand not as dark streaksas derivation of a quantitative relationship between gray tone
before.However, the regionsupstreamof the sandwave crests level and cross-sectionvariation, becauseof unknown nonlin-
are now associatedwith weakly divergentcurrents.Therefore earities in the radar system and the photographic repro-
they appear on the image with a slightly darker gray tone duction process.
level. An orderof magnitudeestimateof the cross-section
modu-
10,542 ALPERS
ANDHENNINGS:
IMAGINGMECHANISMS
OFBOTTOM
TOPOGRAPHY
2' •'N
."?'
,,, _,-- • •,:--'. .... - &:.,'
• ,-' ..- ,-' r2} l # .•
lation can be obtainedfrom our hydrodynamicinteraction tidalvelocityvectorsat the timeof theoverflightinserted. The
model.If we insertd • do= 20 m, d' = 0.07for the steepest tidal current fields were derived from the tidal stream atlas of
slope,IUol= 0.6 m s-•, •b= 0ø, q•= 0ø, # = 0.025s-•, and the BritishAdmiralty(Tidal StreamAtlas,Dover Strait, 1975;
• = 0.5 into (39), then we obtain North Sea,southernportion,1976).Accordingto themeteoro-
logicalmap the wind wasblowingfrom 135øNat 4 m s-•.
(50'/0'
0 • 0.38 (57)
In Figure 5a the Englishcoastnear Ramsgateis visiblein
This seems to be a reasonable value which is consistent with the lower left-hand corner. The dark feature on the left-hand
the radar imagery. sideon the bottomis GoodwinSands,whichfallsdry at low
tides.The V-shapedfeaturein the centeris a pair of two
3.3. SeasatSyntheticApertureRadarImageryof Sandbanks underwater ridgescalledSouthFalls(a thin line to the left)
The predictionof our imagingtheorythat radarimagein- and Sandettie(the broaderline to the right).SouthFalls is
tensityvariationscorrelatewith the slopeof the bottom to- about 30 km long and 600-800 m broad and risesto within 7
pographydividedby the squareof the depthprofileis sub- m of the seasurface.Its slopeis steeperto the westthan to the
stantiatedby an analysisof digitallyprocessed SeasatSAR east.Sandettie, however,hasa verygentleslopeto the north-
imageryof large sandbankslocatedin tidal channels.We have westand a very steepslopeto the southeast.Note that both
chosenfor our investigation SeasatSAR imagescontaining banksare not alignedparallelto the meantidal flowdirection.
landmarks.Theseimageshave a sufficiently high geometric South Falls has an anticlockwise and Sandettie a clockwise
accuracy( • 100 m) that it is possibleto establishsucha corre- offsetwith respect
to the regionaldirectionof the peaktidal
lation.The analyzedareaslie in the easternapproachto the flow [Kenyonet al., 1981; Kenyon,1983]. The seafloorbe-
English Channel. tweentheridgeshasa depthtypicallybetween30 and40 m.
The SeasatSARimagespresented
in thissectionare digi- Sandwaves(megaripples)
are alsovisibleon theimage,es-
tally processed
and have a spatialresolutionof 25 m x 25 m pecially at the southernend of South Falls and on Sandettie.
(fourazimuthallooks).Theywereprocessed by the Deutsche Thesandwavecrestsarealignedapproximately perpendicular
Forschungs-und Versuchsanstaltffir Luft- und Raumfahrt to themeantidalflowdirection. The maximumheightof these
(DFVLR) in Oberpfaffenhofen, WestGermany,on the Mac- sandwavesis about 10 m aboveseafloor,but the typical
DonaldDettweilerand Associates digitalSeasatSAR pro- averageheightof sandwavesin this regionis 4 m. Typical
cessorfor theEuropeanSpaceAgency(ESA). wavelengthsare 300-500m [Van Veen,1936].
Figures5a, 6a, and 7a showdigitallyprocessedSeasatSAR Figure6a showsthecontinuation of theSeasatSARimage
imagesof theSouthernBightof theNorth Sea(thenortheast- of Figure5a towardthe northeast,and Figure7a showsthe
ern approachto the EnglishChannel)fromorbit 762(August continuation towardthe'southeast. By comparing Figures6a
19, 1978, 0646 UT), and Figures5b and 6b show the corre- and6bit canbe seenthat thosesandbanks whicharealigned
sponding(simplified)seachartsto Figures5a and 6a with the approximatelyparallel to the tidal flow direction are not
ALPERSAND HENNINGS:IMAGINGMECHANISMS
OF BOTTOMTOPOGRAPHY 10,543
-,:½
.....
. :.:;.•
•½..½
'•:. ...
., ..
'....•......
•; ,, ..
.-•..
•..:.,. ß ,.•
...
..%:..:•
...:
...
:-.
"flightdirection
.....
-"• .:,•'..
;•";.'•.•.•E...;
;.;,•.•:.-
passfilteredreplicaof the currentgradientfield as discussed in Although the velocity gradient (strain rate) generatedby
connectionwith (36).The inclusionof the advectiontermsinto tidal flow over bathymetryis typically 1 order of magnitude
the imaging theory is requiredwhen the subsurfacefeatures smaller than the velocity gradient generatedby the orbital
containhigh wavenumbercomponents and whenthe current motion associatedwith long surfacewaves,the modulation
velocityis large(see(35)). can neverthelessbe larger, becausethe "interaction time" is
If the relaxationtime is not constantbut dependson wind longer. In the case of tidal flow over bathymetry the cross-
speed,thenthe imagingof large-scale bottomtopographicfea- sectionmodulationdue to hydrodynamicinteractionis pro-
turesis alsowind-dependent. Experimentsin wind wave tanks portionalto the relaxationtime. This parameterdescribeshow
show that r, decreases with wind speed[Keller and Wright, much deviationfrom equilibriumthe short-wavesystemcan
1975;Wright, 1978].We expecta similarbehaviorin the open endurewhen beingstrainedby a variablecurrent.
ocean, and therefore the modulation depth should decrease
with wind speed.Our theorydoesnot supportthe existenceof
a critical wind speedabove which radar imagingof under- Acknowledgments. We thank the ESA (Earthnet,Frascati)and the
waterbottomtopographybecomes basicallyimpossible. DFVLR for providingthe digitallyprocessed SeasatSAR images,and
G. P. de Loor for making availableto us the EMI SLAR images.
Furthermore,we expectthat to first orderthe radar imaging
Furthermore,we are grateful to many colleaguesfor very fruitful
mechanismis independentof wind direction. However, this discussions,
especiallyto R. H. Stewart,G. J. Komen, and K. Hassel-
statementdoes not apply for radar imaging of small-scale mann. This research was supported by the Deutsche For-
10,546 ALPERS
ANDHENNINGS:
IMAGING
MECHANISMS
OFBOTTOM
TOPOGRAPHY