surface marks of the sea bed
a hydrodynamic theory for the imaging of
bottom topography by X- and K-band radar
Stéflingen
behorende bij het proefschrift Surface marks of the sea bed
De instelling van veel oceanografen orn kennis van vissers en zéisters over de zee te negeren, betekent het mislopen van een vette vangst.
Omdat onze 'fysische intü!tie' gebaseerd is op lineair denken is
het, vóór het doen van berekeningen aan een niet-lineaire vergelijking, zinvol orn wat te jongleren met centrale aspecten van die vergelijkiñg.
(dit proefschrift)
In het geval dat zowel korte als lange golven op zee aanwezig zijn, geldt enerzijds op theoretische gronden dat de korte golven
geadvecteerd worden met de orbitaalsnelheid van de lange golven, terwijl anderzijds radarbeelden laten zien dat het patroon van
eriergierijke korte golven zich verplaatst met de fasesnelheid van de lange golven. Het is ten onrechte dat nog ved oceanografen in deze
situatie eèn parádox zien, dié zôu stoelen op fouten. Oer8tijging vañ-dït standpüht wòrdt bereikt door te erkennen dat er spràke i
van twee ,erschillehdesnèlheden: de ¿nélhéid van het patrodn en die yan de korte golven (vergeiijk met de ouderwetse wit met rode
draalende kapperszuilen).
(Shemdln et al. 1986, TOWARD field experiment interim report, JPL, California)
4 Eén van de basisvragen. betrefiende watergolven, namelijk in hoeverre
korte fourler-componenten van het wateropperviak vrije golven zijn, is nog niet opgeiost.
(B.M. Lake & H.C. Yuen 1978, JFM 88, pp. 33-62; H. Mitsuyasu, Y.Y. Kuo & A. Masuda 19?9, ,JF1 92,
Het is wenselijk dat
f bewezén wordt dat de continutteitseisen aan
het grensvlak van twee niet-visceuze vÏoeistoffen als limiet volgen
uit de continúlteitneisen voor visceuze vloeiàtoffen, òf dat het
tegendeel bewezen wordt.
De voorwaarde van Banner & Phillips (19714) voor het breken van korte
golven, nanielljk dat locaa]. de snelheid van de waterdeeltjes groter
is dan de snelheid van het golfprofieì, is niet juist. Zij geldt
slechts als noodzakelijke, maar niet voldoende, voorwaardevoor één
manier van breken, nl. die welke vergelijkbaar is met die van lange
golven dIe het strand op lopen.
(M.L. Banner & 0.M. Phillips 19714, JFM65, pp. 6147-56).
Met verstand zó te gebrulken in de wetenschap als daar gebruikelljk
is, is ongeveer even weinig doeltreffend als je magisch zwaard i-n de
schede laten en met een keukenmesje ten strijde trekken.
(Variant op een uitspraak van Benjamin Hoff.
In: Tao of Pooh, Dutton, New Yòrk, 1982).
Voor
nieuwe
inzichten
in
de
niet-lineaire
eigensohappen
van
watergolven is het denken in termen van fourier-componenten
belemmerend.
Het
gebruik
van
mOmenten ornturbulentie
in
een
stroming.
te
beschrijven is nlet.meer inspirerend. Het is tljd voor een nieuwe
invalshoek, waarbij de structuur van de stronhing voorop staat.
De
rol
van
de
natuur'kunde
als
objectieve waàrheidazoekster
is
uitgespeeld. Haar nieuwe rol zal zij delen met al onze andere kennis
over het zijn op aarde.
Aarinemende dat 'de slag bij Waterloo gewonnen is op de sportvelden
van Eton' (Duke of Wellirigton)o kunnende ver].iezers, orn devOlgende
slag te wnnen, het beste-alle mogelijkhederi en onmogelijkheden
opzij zetten en alsnog gaan sporten op de velden van Eton.
-Golven zijnmooi.
0
150
Ultriodlgthg
voor het bijwonen van mijñ promótie op 3 Juni a.s. orn 1I.3O uur, en de receptie naafloop. Beide vinden plaats in het Academiegeboùw, Domplein
29, Utrecht.. Een roútebeschrijving ¿tàat öp de achterkant van deze bladzijde..
Klaaz-tje van Gâstel Zandhof'sestraat 11 7 3562 GD Utrecht
surface marks of the
sea bed.
a hydrodynamic theory for the imaging of
bottom topography by X- and Kband radar
tekenen van de zeebodem aan het wateropperviak
(met een inleiding en samenvatting ¡n het Nederlands)
FROEFSCH RIFT
ter verkrijging van de graad van doctor in de wiskunde
en natuurwetenschappen aan de Rijksuniversiteit te
Utrecht, op gezag van de rector magnificus prof.dr. J. A.
van Ginkel, volgens besluit van het college van dekanen
in het openbaarte verdedigen opwoensdag 3juni1987
dès hamiddas te 14.30 uur
door
KLAARTJE VAN GASTEL
promotor : Prof.dr. J.T.F. Zimmerman
co-promotor: Dr. G.J. Komen
Dit onderzoek is uitgevoerd met f inanciële steun van de werkgroep Meteorologie en Fysische Oceanograf ie van de Stichting voor Zuiver Weterischappeiijk Onderzoek (ZWO).
Table of contents
i Introduction 13
On phase velocity and growth rate of 17
wind-induced gravity-capillary waves
Abstract 17
.2.1. Introduction 17
2.2. Theory 20
2.2.1. Methods arid equations 20
2.2.2. Basic flow and Perturbation stream 23
function in water
2.2.3. Basic flow and pérturbation stream 25
function in air
2.2»4 Phase velocity and growth rate 27
2.3. Discussion . 31 2.3.1. Growth rates 31 2.3.2. Phase velocities 37 2.l. Main conclusions 40 Appendix 42 References 44
Nonlinear interactions of gravity-capillary waves: 117
Lagrangian theory and effects on the spectrum
Abstract l7
3.1. Introduction 147
3.2. The interaction equations 50
3.3. Multiwave space . 61
3.1!. Symmetries within the triads 70
3.5. Discretization 71 3.6. Spectral devèlopment 711 3.7. Conclusions 81! Appendix 3.A. 85 Appendix 3.B 86 References 87
14 Imaging by X-band radar of bottom topography and
91
internal waves: a nonlinear phenomenon
Abstract
4.1. Introduction 4.2. Model equations
14.3. Quasi-equilibrium spectra at short fetches Spectra in a slowly-varying current field 4.5. Conclusions References 5. Inleiding en samenvattÏng Curriculum vitae 125 91 91 93 96 102 113 115 119
Chapter 2 has been published as:
Van Gastel, K., Janssen, P.A.E.M. & Kornen, G.J.
1985.
J. Fluid Méch.161 pp. 199-216.
2 3 4 g :UNITED . j:KING DOTM l'3D 'E 4.2 A 51'30'N 2E FRANCE
cg
D 230'E/
I
2E ENGLISH CHANNELFIG
I 0.3 'n-0.5Hl
JI
?30 E 0.5 DUNKERQUE OOSTENDE-t
BELGUIM CONTOURS IN METERS CURRENT SPEEDS IN METERS PER SECONDI I I I I I I I
0.9 3'E
Fig. 1.1: (a) A radar image of the southern bight of the North Sea. The different
grey tones can be related to the local depth. To illustrate this a depth chart of
short
seabed
surface
-9 surface -* backscatter
internal waveJ i
current
iwaves
I I I I
I.
measurements
Ivan
¿.astel (1986)
Iampification :1-2
energy balance
Iwind input
nonlinear interactions I
Ch4: solutions of energy.
balance
Iamplificatión:100 .1000
Neasrements
by
Kwoh et al. (1986)
amplification:10
total amplification
1000_10,000
Fig. 1.2: Chain by which modulations of the inteñsity of the radar image can be
retreated to changes in the local depth. Indicated is where and how the links are
treatéd in this thesis, supplemented by the work of Kwoh et al. (1986). and the amplification for each link as found in these studies. The total amplification is in good agreement with data.
Introduction
In
1969
de Loor discovered that on X-hand radar images of thesea Uem
3 cm) bottom topography could be identified (de Loor1981).
Since then this phenomenon has received much attention, mainly due to its occurrence on SAR-images (Synthetic Aperture Radar) taken from the
satellite SEASAT (Fu & Holt
1982,
Phillips1985,
Valenzuela et al.1985,
Halsema et al.1986).
Figure 1.1 is an example of this imaging. It was quickly recognised that not only bottom topography could be imaged thisway but also internal waves and large-scale current structures like eddies (Fu & Holt
1982).
The imaging cari be performed by K- to L-band radar, A = 8 to200
mm.How this imaging is possible was a mystery. The electromagnetic waves do
not enter the sea but are reflected at the sea surface, thus a direct modulation is impossible. The quantity determining the backscatter at
the sea surface is the energy of surface waves having about the electromagnetic wavelength, thus the radar Images imply an indirect modulation via the short surface waves. However, a modulation of these
short waves was also a surprise, because they are too short to feel the
bottom dirèctly and the current gradients are too small to expect significant refraction effects on the surface waves. Most surprising of
all the modulation of these waves can be extremely large: under light
winds contrasts of a factor 1000 and more have been seen, even when the
change in surface current is lessthan 25% (Valenzuela et al.
1985,
Kwoh et al.1986,
Halsema et al.1986).
The modulation chain can hé split into three steps, indicated in figure
1.2.
Elswhere, I have analysed measurements on the first. step, themodulation of surface current by a change in depth (Van Gastel,
1986).
Two conclusions of this study are important to the imaging mechanism: 1) the surface current can be slightly out of phase with the depth
modulation, arid 2) the arnplication of the modulation factor is small in this step: less than 2.
In this thesis an account is given of my theoretical research on the second step, the modulation of short surface waves by surface current. The parameter space of the second step is generated by the quantities wind speed, surface current speed and wavelength of the surface waves.
Here only low wind speeds, less than 5 m/s, and millimeter and
centimeter wäves are considered. This is the regime for which the very
large contrasts on the radar images occur. A restriction of my analysis is that it is one-dimensional.
The idea behind the treatment of this step is first to find an energy balance for waves of 0.5 to 15 cm and then to solve, this balance.
Effects I want to include in the balance are refraction, input of energy by wind, dissipation and nonlinear interactions between the short waves.
In chapter 2 a term describing the input of energy by wind, in other
words the generation of waves by wind, is found. New s the nearly
analytical description of this term. Chapter 3 deals with the nonlinear interactIons in triads of short surface waves. Many questions and
problems encumbering the incorporation of this term in an energy balance
are being solved here. In chapter I an energy balance is completed, by finding a description of the dissipation, and solved. As not much is
known of the dissipation rough estimates have to be used for this term. New items are the dissipation functions and thé solutions of a nonlinear
balance. Former theories (Hughes
1978,
Lyzenga et al.1983,
Alpers &Hennings 19811, Ermakov & Pelinovsky 19811 and Yuen et al..
1986)
all use balances witIiout the nonlinear term.The solutions of the so-constructed energy balance show modulations of the energy by slowly-varying currents of a factor 100 to 1000. These large modulation factors are only found for k
260
m1. Together with the results of Van Gastel(1.986)
on the first step of the. modulation mechanism and of Kwoh e.t al.(1986)
on the third step, from surfacewaves to backscatter, this thesis gives a description of the imaging of bottom topography by X- and K-band radar, for low windspeeds and
wavenumbers smaller than 260 m1. It is the first theory to predict correct order of magnitude of the modulation. This agreement with data
is due to the incorporatiòn of the nonlinear interactions in the energy balance.
References
Alpers, W. & Hennings, I. 19811. A theory of the imaging mechanism of underwater bottom topography by real and synthetic aperture radar.
J. of Geophys. Research, 89, C6, pp. 1O.529-1O.5'46.
Ermakov, S.A. & Pelinovsky, E.N. 19814. Variations of the spectrum of wind ripple on coastal waters under the action of internal waves.
Dyn. of Atmospheres and Oceans, 8, pp. 95-lOO.
Fu, L. -L. & Holt, B. 1982. Seasat views oceans and sea ice with
synthetic-aperture radar. JPL Pubi. 81-210.
Gastel, K. van 1986. Velocity profiles of tidal currents over sand
waves. Royal Netherlands Meteorological Institute,
De But, The
Netherlands.
Halsema, D. van, Gray, A.L., Hughes, S.J. & Hughes, B.A. 1986. C- and Ku-band scatterometer results from the scattmod internal wave experiment. IGARSS '86 symposium, ESA SP-254, pp. 311-317.
Hughes, B.A. 1978. The effect of internal waves on surface wind waves.
2. Theoretical analysis. J. of Geophys. Research, 83, Cl, pp. 1155-1165.
Kwoh. D.S.W., Lake, B.M. & Rungaldier, H. 1986. IdentIfication of the
contribution of Bragg-scattering and specular reflection to X-band
microwave backscattering in an ocean experiment. TRW, Space & Techn. Group, California.
Loor, G.P. de 1981. The observation of tidal patterns, currents, and bathymetr' with SLAR imagery of the sea. IEEE journal of oceanic
engineering, 0E-6, no. 14
Lyzenga, D.R., Shuchnan, R.A., Kasischke, E.S. & Meadows, G.A. 1983.
Modeling of bottom-related surface patterns imaged by synthetic
aperture radar. IGARSS'83 symposium.
Phillips, O.M. 19814. On the response of short ocean wave components a
fixed wavenumber to ocean current variations. J. of Physical Ocean.
111, pp. 11425-11433.
Valenzuela, G.R., Plant, W.J., Schuler, D.L., Chen, D.T. & Keller, W.C.
1985.
Microwave probing of shallow water bottom topography in the Nantucket Shoals. J. of GeOphysical Research 9O no. C3, pp.14931-49142.
Yuen, H.C. Crawford, D.R. & Saffnan, P.G. 1986. SAR Imaging of bottom
topography in the ocean: results from an improved model. IGARSS'86, ESA SP-254, pp.
807-812.
2
On phase velocity and growth rate of:
wind-induced gravity-capillary waves
Abstract
Generation and growth of gravity-capillary waves (À = 1 cm) by wind are
reconsidered using linear Instability theory to describe the process.
For all friction velocities we solve the resulting Orr-Sommerfeld
equation using asymptotic methods. New elements in our theory, compared
with the work of Benjamin (1959) and Miles (1962), are more stress on
mathematical rigour and the incorporation of the wind-induced shear
current. We find that the growth rate of the initial wavelets, the first waves to be generated by the wind, is proportional to u.
We also study the effect of changes in the shape of the profiles of wind
and wind-induced current. In doing this we compare results of Miles
(1962), Larson and Wright (1975), Valenzuela (1976), Kawal (1979a),
Plant & Wright (1980) and our study. We find that the growth rate is very sensitive to the shape of the wind profile while the influence of changes in the current profile is much smaller. To determine correctly the phase velocity,
the value of current and current shear at the
interface are very Important, much more so than the shape of either wind or current profile.
2 1. Introduction
Recently interest in generation, growth and equilibriúm of gravity-capillary waves has been renewed owing to the growing importance of
remote sensing of the sea surface. Microwave-radar backscatter is
determined largely by the energy density of waves with wavelengths of the order of ¿-'l0 cm (Raney et al., 1985). In order to áscertain the
energy density one needs knowledge of sources and sinks of energy and of kinematical quantities like advection and refraction.
y(mm) 0.8- 0.6-y1 0.4-I I I I 0.6 0.8 1.0 1.2
U(m s-')
Fio. 24 Wind speed and current as a function of height. In air a linear-logaritimió profile is drawn; in water , Miles constant profile (Miles took U, = 0, in this figure we took = 0.75 m - - - -. Valènzuela's linear-logarithmic profile; --. Kawai's error-function-like profile; , our
exponential profile. On the vertical scale typical values for various quantities are indicated. À is a wavelength, g,, a thickness of the viscous sublayer in air, y, a critical height, 'i, a wave amplitude and Yiw a thickness of the viscous sublayer in water as assumed in Valenzuela's profile.
In this study önly part of this intricate process is considered. We concentrate on the initial generation and growth of gravity-capillary
waves under the influence of the wind.
Most of the recent studies on this subject use the linear-instability
theory as presented by Miles (1957). In 1959 Benjamin made an analytical
study of the flow over a wavy boundary. He looked at the flow over a
rigid surface, in this way decoupling the flows in the two media. Although he noted the possibility of generalizing this theory to the
0.2-Y0 'io
-0.2_
-0.4- -0.6- -0.8-y -1.0-i 0.2 I 0.4 y(mm)-3
-4
-5
-6
-7
flow over a fluid and determining growth rates, he did- not càrry out
such a programme. Miles (1962) did apply Benjamin's theory to the growth of gravity-capillary waves by wind. He used a linear-logarithmic flow in air, the profile of which is drawn in figure 2.1. In accordance with
Benjamin he assumed the water to be at rest. It may be noted that these two flows do not satisfy the equation for continuity or shearing stress at the boundary of two fluids.
Valenzuela (1976) numerically solved the equations using a coupled wind-current system satisfying the continuity equations. For the wind as well. as for the current he assumed a linear-logarithmic profile.
Kawai (1979a) extended the research to the generation of
gravity-capillary waves by combining numerical and experimental work. He measured the flow at the moment the initial wavelets- appeared and their growth rate, phase velocity and frequency. His numerical work describes these measurements. He uses a coupled wind-current system, an error-function-like current profile (drawn In figure 2.1) and the usual wind
profile.
In the next section our analytical analysis is presented. We use the linear-logarithmic profile ±n air and an exponential profile in water (figure 2.1). We have chosen this profile because it closely resembles, Kawai's profile and because it allows for an exact solution of the
Rayleigh equation. We briefly discuss the derivation
of the
Orr-Sommerfeld equation plus boundary conditions as a description of the
growth of gravity-capillary waves. We then solve these equatioñs
asymptotically. Asymptotic analysis makes sense becaUse the density of air is small compared to the density of water and because reasonably
large Reynolds numbers can be defined in air and- water. We find expressions for the phase velocity and growth rate of the waves.
It is inherent with the asymptotic methods that we are able to indicate to what order each expression is correct. This is an improvement ori the
wave-growth theories of both Benjamin and Miles. We are even able to Indiôate the order of the errors in Miles' expression for the growth.
Another improvement of our analysis is that the main flows satisfy the
continuity equations. This means that formal justification for our expressions exists, in contrast to the cases of' Benjamin and Miles.
Numerical results of our añalysis are given in section
2.3.
We alsostudy how sensitive the phase velocity and growth rate are to changes in
shape of the profiles of wind and current, by comparing the results of Miles, Valenzuela, Kawai and our study: each is based on a different
profile. For verification we use the experimental laboratory results of Kawal (1979a), Larson & Wright (1975) and Plant & Wright (1980).
2.2. Theory
2.2.1.Methods and equations
The growth of waves on the interface of water and air can be seen as the
perturbation of the equilibrium consisting of a plane interface and uniform basic flow in air and water. Physically, the description would
be as follows. The wind sets in and after a few seconds the upper layer
of the water starts to drift with the wind. These flows, both strongly sheared near the interface, are unstable and after another few seconds ripples start to appear (see Kawai 1979a). In this initial stage the growth of the waves is exponential; after a further few seconds other
mechanisms come into effect and saturation sets in. In a final stage the
wind and current profile would be modified by the constant flow of
energy from, the air towards the waves. In this chapter we confine ourselves to the initial stage of wave growth, where instability and viscous damping are the only energy sources. Keeping this in mind a
mathematical description of the growth of the waves can be given. Growth
is then desöribèd as an instability of the equilibrium in the
normal-mode analysis. For simplicity the situation is assumed to be uniform in
one horizontal direction: as we are interested in plane-parallel flow
this does not diminish the possibility of finding growth (Drazin & Reid 1982, p. 155). In a later section the description of the wind and the
to be shear flows satisfying at the interface the usual continuity
equations of normal and tangential velocity, shearing stress and normal pressure (Batchelor i981 pp. 11815O). The equation for the continuity
of shèaring stress will be of importance. It reads:
a 0) 'w U'(0)
, (2.1)
where p is the viscosity U is the velocity of the basic flow in the horizontal direction and a prime denotes differentiation with respect
to , a dimensionless height coordinate equal to the product of
wavenumber (see eq. (2.6)) and height y. The subscripts a and w stand
for air and water respect-ively.
We are interested in the deviation (x,t) of the interface from equilibrium, where x is the non-trivial horizontal coordinate and t the time. To calculate r we introduce a perturbation stream function p. it
is assumed to have a wave-like nature:
= (2.2)
It will be seen from eq. (2.6) that r has the same x- and t-dependence; thus k is the wavenumber and c the phasè velocity.
The equations for , the height-dependent part of the stream funètion,
can be easily derived from the five conservation laws and the e4uation of state (Batchelor 1981, p. 16l) which together govern the fluid motion. Additional assumptions we made are: temperature and density are
constànt; gravity is the only body force present; and turbulence and other nonlinear features are neglected, although they enter indirectly
through the background profile. Here it
may be ädded
that surface tension, which canñöt be neglected for the short. wavès of intérest hère, is a conservative force like gravity. Thus we know bèforèhand thàt thesetwo forces influence the frequency and not the growth of the waves. Surface tension is not a body force but enters through the boundary
conditions. The governing equations then become the Orr-Sommerfeld
express the vanishing; of the wave-induced disturbance at large heights and the continuity conditions at the Interface.,
In dimensionless form the equations rad (see Valénzuela 1976; Kàwai 1979à):
(w' +
W (2.3) (2.1a) (2.lth) (2»c) (2. )4'd) (2.5)Here u, is the friction vèlocity in air (see seö-tion
2.2.3),
W is a, dimensionless velocity, W = U - c, and W0 is the value of' W at 0,the air-watér interface. The. gravitational acceleration is g, T is the surface tension and p the density. and c are dimensionless constants:
a'1w an E: = vk/u, trie inverse ofa Reynolds nurnber
Finding the growth of a particular wave is now trañslated into solving
thé problem ôutlinèd abOyé.- To completely determine the probleh we need boundary conditions or iñitial- values i-n x and t-. As we aré interested
in temporal growth rates we take periodic boundary-conditions in x and
initial vaLles for 4, at a given time.
m-la
implies that k is real and that e is salved as a function of k. When one is interested in spatialgrowth, i.e. growth with fetch, tlie roles of x and t are reversed and initial valu'es for given xand boundary conditions i t are taken (Kawai 1979a; Drazin & fleid 1982, pp. 152-153). The imaginary part of c determines he growth rate, as can be seen from the equation for the
2 2
-1)
- W'' n, = -1(-
4,;
=- w
'= w'
- w
'as
o-aww
ow
E: W' W'' & [ (i- + '' J = - "w + [(w' +
't'a - - 4, - ic. 4,'''] = u kW -(W03)
-
w''
+ukW
w o+ O,
4,'. + 0
a°
ik(x-ct)e o
To find o as a function of k the profiles of wind and current have to be specified; then the problem
Orr-Sömmerfeld equation in
methods (see, för instance,
conditions at infinity are the aid of the continuity
that S, and £ are small
a w
(2.6)
is completely determined. We next solve the
air and water separately using asymptotic
Drazin & Reid 1982, chapter 1!). The boundary
applied and the solutions are coupled with equations. Throughout we will use the fact
parameters:
We take the water flow to be time independent. This is consistent with the experimental results of Kawai (19T9a); he found the flow to depend
on time but on a much larger scale than the growth of the waves.
Substitution of (2.8) in the Orr-Sommerfield equation enables us to find
the perturbation stream function $ in water. We write as a sum öf
two independent solutions. These will be called the inviscid and viscid solution respectively, and $, as one is in first order a solution of the inviscid Rayleigh equation and the other is relatively large in
regions where viscosity is important. We normalize the solution to unity at the surface:
interface (which can derived from the kinematical boundary condition):
«1
a
«1,
£ « 1
(2 . 7a, b, e)
2.2.2 Basic flow & perturbation stream function in water
As mentioned in the introduction the basic flow in the water is taken to
be -U = U (2.8a) w o S u À
eU
(2.8b)wo
(2.9)
4,(0) 1
Normalization of is possible because the set of equations (2.3)-(2.5) does riot depend on the amplitude of the perturbation (substitution of = a4 yields exactly the same equations). For convenience we also
normalize the two independent solutions:
= .i(0) = 1
The inviscid solution can be found by a formal expansion of and in . To first order this yields:
iw = iwo + iw iwl
U F(p,q;r;) e F(p,q;r;-2) iwl = K . Iwo + 4'iwi + (2.10)p = 1/A +
(+ 1/A2),
q =1/A -
(1 + 1,A2), r = 1 + 2/AThe function is a complicated expression. Upon inspection we find
iw10
= (0) =which is the only result rêquired below. Here K is an integral over the
total depth of a differential operator working on
jw the exact form
of the operator and an explicit expression för K are given in the
appendix (2.A1-2.A3). F is the hypergeometric function (see Abramowitz & Stegun 1965).
The viscid solution varies on a scale of (Drazin & Reid 1982). It can be found by a WKB approximation:
= exp
fd'
(2.11)f =
i)
(-W) - .+
0(c)
Note that (2.9)-(2.11) give the complete solution of the Orr-Sommerfjeid equation satisfying the boundary conditions at infinite depth.
2.2.3 Basic flow and perturbation stream function in air
We have taken the usual linear-logarithmic wind profile (figure 2.1):
U =+U
a ca o u*U =U +U
(-tanh)
ai
O K sinh = (2.12) = r e,K=O»1,
U1=ru.
The value of r determines the thickness of the viscous sublayer. All
that is known about its value is that it is of the order of unity (Monín & Yagloni 1971 ). In the literature on growth of gravity-capillary waves values of 5 and 8 prevail (Miles 1962; Valenzuela 1976; Kawai 1979a). We have taken r = 5. All the essentials of the analysis are independent of this choice; only the values of the growth rates are larger for larger
r.
From (2.8) and (2.12) it can be deduced that the condition for
continuity of shearing stress at the interface, ie. equation (2.1), is
fulfilled.
To solve the Orr-Sommerfield equation with (2.12) as the basic flow we
again separate the solution
a into an inviscid and a viscid part:
A
ia + B ,
=
The results of Kawai (1979a) concerning the phase velocity indicate that
it is reasonable to assume that the critical height
c (defined
by Ua(c) = c) is beneath the top of the viscous sublayer:
<
(2.14)
This implies that the Rayleigh equation has no singularities. The zeroth
order expansion of the inviscid solution (which is the solution of the
Rayleigh equation) can then be calculated numerically without
complications. We used the méthod described by Janssen & Peeck (1985).
It will be important in the following to note that:
for n odd
and for n even
0(1/c ) for 0.007 c 0.02
n a a
=
io(i)
for 0.06 caia
= 0(1).
(2.15)
Note the two ranges for the order of magnitude for when n is odd. For intermediate ca , i.e. 0.02 ca 0.06, the order of
a is also
intermediate. For u = 0.15 ms the minimum value for ca of 0.007
corresponds to wavelengths where capillary effects become unimportant.
It is the smallest value for c that we have considered.
a
Eq. (2.1) also implies that in the inner viscous layer ( a thin layer
around the critical height; for an exact definition see Drazin & Reid
1982) the profile can be approximated by Ua
= - +
U. In the inner
viscous layer, which includes the interface, the iscid solution varies2/3 2/3
on a scale of ca . It is given to order Ca by the second integral of the Airy-function (Drazin & Reid 1982):
va = Ai (, 2)/Ai (i13cahh/3Wo2) 2/3 + o a .1/3
Ç -
+ c
W J , phase (i1"3) = 1/6 irao
caThe function q,'(Ç) has often been tabulated and plotted (e.g. Benjamin (2.16)
1959). It is related to the Tietjens function DT:
2/3 DT
(k01) =
-
(La va2.2.!! Phase velocity and growth rate
Application of the remaining boundary conditions (2.'4b-d) to the stream functioiis in air and water yields expressions for A, B, C1 D and c. We find for 0.007 E 0.02
B=
10(1) for 0.06 £ (2.17)D=
fo
(6L2"3)
for 0.007 L 0.02 för 0.06Lp
A=1-B, C=1-D.
The cause of the two distinct ranges in the order of magnitude of A, B and D is their dependence on
ia'
. In principle a third range also occurs when - Wj = O (1/c). However, we have checked that this
does not occur for 0.007.
To find the phase velocity o we introduce an expansion for it. Many
powers and cross-terms of the small parameters ó La and c appear in
the equations. However, we find that it is possible to define
e=c0+c1
c2 c--=O(E
J,
E:),
e a w c2 1/2 2/3- = O (c
, C1 w a (2.18)The order estimates depend on
the relative magnitude of the smallparameters. We have indicated several possibilities; the largest of
these determines the accuracy of the expression.
The approximations to the phase velocity cO and cl can be expressed as
follows*:
= U0 A U ((A U0)2 + k + (2.19a)
icH+ ? i ni75
The terms
h',
, mp7"
and AVare given in the appendix(2.A3-2.A8).
H
represents the effects of viscosity and shear flow in water; its value is approximately '.R
andT
are respectively the complexamplitude of normal pressure and shear stress of the air on the surface.
2
-They have been de dimensionless by dividing by _Pak u i- is
dominatéd by the term -i B tp''' and T'by
t
B ''. in, the coefficientof T,deperlds oh properties df the flow in the water, for the case of
no flow it eqüals 1 . IYalso depends on the water flow and has a value df about 2.
The growth rate of the energy B is given by
B = 2 k Imc . (2.20a)
As c = c0 + c1 and c0 is real this implies
B = 2 k Imc1. (2.20b)
Eq. (2.20a) fòllows immediately from (2.6) and the definition of energy density for surface waves.
* We are oñly interested in gravity-capillary waves and neglect the possibility of finding Tollmien-Schlichting waves, although they are also solutions of the equations (Miles 1962).
cl = u*.
To get some feeling of what (2.19) implies we first consider thé case Ua
= const. Then (2.19) simplifies to
co = Uo +
U =+U
a. ca o+ I'2 Ji= i. 2=
g k p , .,'
2 w ku o 1)W , m=i,T=o andlY=2.C0 is now the familiar expression for the phase veloóity of free waves. /Xdescribes viscous damping and?/IVthe correction to the phase
velocity due to the renormalization of the gravity force (see Whitham
197k, p. 41l5). Here it may also be noted that when 6 = O growth s
impossible, as C0 is real. This was found numerically by Kawai (1977),.
who studied the possibility of the instability of a sheared current
generating waves.
The effec.t of the shear in the water is to decrease e0 and the damping
due to changes in
H
and IV.. The shear in the air together with itsviscosity have two effects. One is that the pressure is shifted in phase relative to the surface waves, theréby making growth possible (Miles in
1957 was the first to determine the phase shift of the pressure). The
other is that the tangential stress r is now non-zero.
Another interesting simplification is that of a wind profile that is linear up td infinity. This is quite realistic
for the very short
waves (c«1 )
Moreover, this flow allows for an exact solution of theOrr-Sommerfeld equation in terms of the Airy-function. This was already known by Mises (191.2 a, b) and Hopf (191k) and perhaps the solution is
even of earlier date. For the profile
(2.21)
expression (2.16) for
va becomes valid at all heights and ia becomes
exactly
Equation (2.19) remains the same but the pressure can now be given explicitly: - + ) + 'va (2.22)
kuW0
a o vaHere = U/u; 4j' and
"V
can be found In the appendix (2.A21-2.A22).Both the shear stress and the normal pressure appear i-n eq. (2.19b). However, the effect of the stress on the growth is much smaller than the effect of the pressure. This can be deduced by noting that m = 0(1) and
comparing the leading terms of? and
Y:
_IaBcb'
and Then notethat for bth c'' and ''' the real and imaginary part are of the same
order and that i,'j is an order smaller than
Miles (1962) also expressed the growth in terms of approximations to the pressure and shear stress:*
-Ic1I+(
-I
)w M M
°1M - 2
However, theré are two important differences between Miles' expression
and our's. Firstly, Miles' expression is formally invalid since the
shearing stress of the basic flows is discontinuous at the surface; Miles supposed Uw 0. In the next section it will be seen that the
numerical resülts of (2.23) are, however, satisfactory (this was also
shown by Välenzuela 1976). Secondly, Miles approximated the pressure and shearing stress on physical rather than mathematical grounds (actually, Benjamin (1959) made these approximations and Mlles adopted them).
* Miles studied the generalized situation of finite depth. We have taken the limit of his results for infinite depth and written his
Therefore he was not able to indicate to what order his expressions were
correct. Using our mathematical analysis we can estimate the order of the error in Mlles' expression for the pressure. For the case U 0,
assuming the expressions for the streamfunctions to be exact (though
Miles obtained approximations), it is Ca For comparison, based on the assumption of infinite precision of the c,'s, our expression for the
-1/2 1/3
pressure is correct to order ISEw Ca for 0.007 0.02 and to order . -1/2 2/3 for 0.06
W C a
Another conclusion to be drawn from (2.19) is that, as growth by wind
input and viscous damping nearly cancel, the growth rate of the
gravity-capillary waves is very sensitive to slight changes of the oceanic and
atmospheric Parameters
a'w"'a
and2.3.Discussion of results
2.3.1. Growth rates
To calculate the growth rate numerically we neglect Y.This
can bejustified by noting that the leading term of
T
is of thé same order as2/3
the error inr (as
va is correct to order Ca ). We use the following numerical values; all in S.l.: g 9.806, T/Pw = 7.25 10e,
= 1.2 io, "a = 1.510
and =io6.
In figure 2.2 the curves of the growth rate as a function of wavenumber are shown for several windspeeds. For u 0.05 rn/s all curves show a single positive maximum. We f ïnd a critical value for wave generation near u* = 0.05 rn/s. This value is in accordance with Miles (1962) and
Kawai (1979b). For u between, roughly, 0.10 and 0.30 mIs the top of the
curve occurs at such wavenumbers that Ra 36. Growth at a certain
wavenumber strongly increases with windspeed; there is no simple scaling law. The growth at the top of the curve increases even faster with increasing windspeed; this will be discussed later.
3.2
- I I
-.t
I i I I I - lOO 200 300 400 500 600 700 800 900k(m)
Fm. Z.2. Growth rate as a function of wavenumber for various wind speeds: = 0.248 ms, U5 = 0.102m s'; --. 0.214,0.098;--, 0.170, 0.096; 0.136,0.075; - - - -, 0.050. 0.025. 4.0 3.6
ß(s)
3.2 2.8 2.4 2.0 1.6 1.2 0.8 0.4/
/
\
-.'
I -- I I I I I I I I lOO ÔÓ 300 400 500 600 700 800 900k(m)
Fia. t3. Growth rate as a function ofwavenumber; a comparison between different studies, each
using a different flow in water: -, Miles (1962) u. = 0.23 m 5i; Valenzuela (1976)
0.25 m s'; - - - -, Kawai (1979a) 0.248 m s' ...our study 0.248 m s'. Experimental results
of Larson & Wright (1975) at u = 0.27 m s are indiòated with crosses.
2.8
fi (s)
2.4 2.0 1.6 1.2 0.8 0.410
fi(s).
10
10-2
I0
l0I
u. (ms)
Fio .. .2.4. Maximum growth rute as a function of frjctioii viIoeitv. 'Ilnuiv . Milis (1962): S. Valenzuela (1976) - - - -. Kawai (1979e): oùrstudy. Measurements: fl. Larson & Wright.
(1975):
0
Kawai (1979e).To compare our results with those of others we have chosen one specific
friction velocity. The main features are the same for other values of
u*, which we have checked for u 0.14 rn/s. We find the same type of growth curves as Valenzuela (1976) and Kawai (1979a), as can be seen in figure
2.3.
This implies that the decomposition of the stream function into a viscid and an inviscid part, which we have used throughout, isvalid at all friction velocities; this point was left in doubt by Valenzuela (1976). The height of the top of our curve differs by about
10% from that of either Kawal or Valenzuela, in agreement with our estimated accuracy of
c2"3
= 7%. The difference between the öurves ofKawai and Valenzuela is 20%. As each curve is based on a different
current profile these differences determine the sensitivity of the
growth to small changes in the current profile. Miles (1962) studied smaller wavenumbers than we have. Therefore not much more can be said
than that the results of the two studies do not disagree. The measurements of Larson & wright (1975) on B give values which are
reproduced by our theory within 25%*.
Those waves for which the growth rates are largest are the first to be
generated by the wind (Kawal 1979a). Therefore the maximum growth rates
as a function of u are of interest and we have plotted them in figure
2.11. This plot also offers another method to compare the effects of the various curreit profiles.
As a result of our calculations we find, for the range of friction
velocities 0.05 rn/s u 0.11 m/s (roughly, this range coincides with i rn/s °io 12 m/s), that
B
max (2.211)
This is a surprisingly simple result considering the intricate expression (2.1gb).
* There is some uncertainty concerning the friction velocity; Larson & Wright give the value u in the steady state while our computations
are for the transient state. According to Kawai (1979a) u in the transient state is considerably less (up to 50%) than in the steady state at the same value of, for instance, U2. Another uncertainty is introduced by the fact that u determinations in the laboratory are,
as a rule, exact up to not more than 5 - 10%. However, we have neglected these complications and simple compared data at the same
4.4 4.0 fi(s) 3.6 3.2 2.8 2.4 2.0 1.6 1.2 0.8 0.4 I I I I I -- I t I I 100 200 300 400 500 600 700 800 900
k(m')
FIG. . 5. Growth rate as a function of wavenumber at u = 0.214 m s. effect ofchanges in profiles s
Our theory: ----. linear wind profile; -.-, wO . (25) as approximation to the
l)'eSSIIre.
In the range considered we again find that our results are close to those of Miles, Valenzuela and Kawai; deviations are within 20%. The
values of the measurements of Kawal and Larson & Wright are higher than the theoretical values; the largest difference is 100% of the
theoretical maximum growth rate. It must be noted that the functional
dependence on u differs in the various theories and experiments; e.g. Kawai finds numerically that
max Thus1 for
given u,
B isindependent of the current profile within 20% but relation (2.21) is different for the various profiles.
,__
To study the effect of the shape of the wind profile we have taken r = 8 (see eq. (2.12)) and we have compared the linear-logarithmic profile with the lineär profile (2.21). Note that these changes occur above the critical height. We have also studied separately the effects of the two
features which distinguish Miles' theory from ours. We took a Benjamin
type of approximation to the pressure - equal to to
order
a - together with the exponential profile in water; that is, we
used eq. (2.19b) with replaced by
/ 2 ku* W o À
- W C.
[1/c - W0 -o o iwoW ('-')
o va la (2.25)Also we took = O together with our expression for the pressure. The results for one value of u are shown in figure 2.5.
When the current is set uniformly equal to zero the growth rate becomes about 15% läwer than when the exponential profile is used. This deviation is in accordance with the sensitivity to the current profile we fàund above. When the linear wind profile is used we still obtain
growth but the growth rates are incorrect. The results are best for high wavenumbers but even then the growth is two times too large. The growth
curves as a function of wavenumber no longer show a maximum. When the
linear-logarithmic profile is used with a thicker viscous sublayer, r = 8 instead of r 5, growth is nearly, though nog quite, 8/5 times as
much. When (2.25) is used for the pressure the growth corresponds within 20% with the valuès found using our full expression for ; the growth is
always tod large.
Our analysis shows that the growth is very sensitive to changes in the
shape of the wind profile, even when these changes occur well above the critical height. This is not as surprising as it may seem. Miles (1957) was able to express first the growth rate in terms of an integral of the stream function and the wind pröfile from the interface up to infinity, and then in terms of values of these functions and their derivatives at
the stream function and its derivatives at 'the critical height still depend on all parts of the wind profile because the stream function is the solution of a differential equation containing this profile at all heights. Note that the second step of Miles as sketched above is not
possible when viscosity is taken into account.
2.3.2 Phase velocities
In principle, the phase speed of the growing waves depends in two ways
on the wind: directly; but also indirectly through the wind-induced
current. First we treat the effect' 'of the current. The current can be
characterized by its value at the surface U0 and by the shape of its
profile. To study the effect of these two we have varied both. First, to
investigate the effect of the shape of the current profile we compared
our results with those of Valenzuela and Kawai using their values of U0
and our profile. We find that the phase speed is insensitive to the
choice of linear-logarithmic, exponential or Kawai's profile. Differences between values of the phase speed are in the order of a few
percent. Using a constant profile, either U O or U U0, leads to
errors of about 20% at u* = 0.136 m/s and 50% at u = 0.6 rn/s.
As Valenzuela uses U0 = O.8u* while Kawai used his measured values, which are near U0 = 0.5u, it is difficult to present illustrations of
the foregoing in a figure. In figure 2.6 we show results for u = 0.136 m/s: apart from Valenzuela's results all theoretical values are based on
Kawai's value for U0. Measurements of' Kawai and Plant & Wright (1980)
are also presented in figure 2.6.' The effect of the value of U0 depends on wavenumber and u. At higher wavenumbers the phase speed becomes less
sensitive to U0. For k = 155 m1 the dependence on U0 for different friction velocities is shown in figure 2.7. At low friction velocities
the influence of U0 on o is small. However, at u = 0.6 m/sValenzuela's value based on U = 0.8 u* is 50% larger than our value, obtained by
using U0 0.65 u. If we use U0 = 0.5 u in our exponential profile we obtain a phase speed '40% lower than that at U0 = 0.65 u. Using U0 =
0.45 0.40
c(m s-')
0.35 0.20 0.15 0.10 I I I I I I I_ I 100 200 300 400 500 600 700 800 900k (m')
F10. 2.6. Phase velocity as a function of wavenumber at u = 0.136 ms'. Theory:
Valenzuela (1976) . Kawai (1979a):
or study. (19a): -.... our study.
= 0; -.-,
our study, (J 0. Measurement: x. Kawai (1979a): O. Plant & Wright (1980).
Plant & Wright (1980), see figure 2.7. In their experiment U0 was not measured, however, they suggested U0 = 0.6 u. This analysis shows the
importance of U0 in comparing various theoretical methods and in comparing experimental and theoretical data.
Next we studied the direct effect of the wind. The wind has no direct
effect on c0 but has on c1 (see eq. (2.18) and (2.19)); the real part of c1 is the first order correction in
a and on the phase speed.
Thus the maximum possible direct effect is given by
0.I 0.2 0.3 0.4 0.5 0.6
u. (mis)
2.7. Phase velocity as a function of friction velocity for different values of U0. Theory
Valenzuela (1976) U0 = 0.8u; - -: our study: e0, U0=0.65u .c. U0=
O.65u: --.
c.U0 = O.5v.Measurements: . Plant & Wright (P180).
In practice (1/c) Re e1 is even smaller than this. For example1 for
u
=0.136 rn/sand lOOm1
<k
<500m1,
(1/c) Re c1< 0.01
a"< 0.014),
while for u = 0.6 m/s and k = 700 rn
14/3
and for k = 155 ni1
(1/c) Ree1 0.30 =
Incidentally, Re e1 is always negative (see figure 2.7).
As well as this order analysis we have calculated the effect on the phase speed of changes in the shape of the wind profile by varying the thickness of the viscous sublayer; we compared r = 5 and r = 8. As is expected from the foregoing, for u < 0.25 rn/s the difference is less
than 1% for all wave-ñumbers. For u = 0.6 rn/s the difference is about 15% of e (± 30% of Re e1).
Summarizing, we can say that the wind-induced current has a large effect ori the phase velocity; this can be 50% of the phase speed of free waves.
The exact shape of the profile, linear-logarithmic, exponential or Kawai's, is not of importance, though a constant profile leads to errors. However, the value of the current at the surface certainly is of importance, an error in U0 of 25% can lead to an error in the value of e of 50%. This suggests that the value of the current and its shear at the surface are the two most important features of the current profile.. The
direct effect of the wind on the phase velocity is only noticeable at
larger friction velocities: for u = 0.6 rn/s it can be 15% of e.
2.!!. Main conclusions
It is possible to describe the Initial growth of gravity-capillary waves with asyrnptdtic methods. Our analysis results in expressions for phase
-2/3 2/3
velocity and growth rate accurate to order 6Ca and Ea respectively for 0.007 .02 and to order
-1/3 2/3
6E and c resp. for E 0.06, where c is the inverse of a
a a a a
Reynolds number: E \a'u*. For intermediate Ca the accuracy is also intermediate. Our analysis confirms the validity of Miles' (1962) expression fpr the growth rate.
The wind speed strongly influences their growth; we find that the growth rate of the initial. wavelets is proportional to u3. There is no simple scaling law for the growth rate at fixed wavenumber but it alsò depends strongly on u. Changes in the shape of the wind profile, even above the
critical height, can change the growth rate by a factor of more than a
three. The influence of the current profile on the growth is within ±20%.
The phase velocity is more sensitive to the wind-induced current than to the wind itself; the effect of the current can be 50% of the phase speed while the direct effect of the wind is less than 15%. A linear-logarithmic, exponential or error-function-like profile all lead to the same results. Good agreement with experimental data can be obtaine when
the value of the current at the surface is known; errors in U0 of 25%
lead to errors in c of up to 50%.
Finally we checked that the perturbation normál pressure of the air on the surface causes the growth; the effect of the perturbation shear
stress is ari order Ra213 smaller.
Acknowledgements
This work was partly supported by the Netherlands Organization for the
Advancement of Pure Research (ZWO). Wé would like to thank G.R.
Valenzuela for making his original data available to us, Anne de Baas
and Wim Verkleij for comments on early versions of this paper, and Fred Dobson for an extensive review.
Appendix K
= -
:!-.
-1)210d'
A3Uc
14U2
U-
+ a(2)m[2++2)F(3,rn+2+2/A;
m+3+2/A; .2) r(r) 2a=[r()-t()
a prime stañds for differentiation with respect to the argumnt.
E.
'-w
c
. ' w ' + C(3!
-
''')
-o 1 iwo o
iwl
i
w o vwl iwo Iwo vwi(2.A11) w Terms -(2. A2) U
sas
eD ' and - D ''' also appear in (2.A14) but these cancel.
lE VWO VWO
w
+ Ati - w0c
Iwo +
3aA
- a (A ''' +B ''')
(2.A5)u*
-r=
c.B
(2.A6)+
(2Aml)
F (, m+3+2/A; m+11+2/A;2)]
(2.A1)r(p+j) .r(q+j) i(p+m-j) 1(q+mej) r(r+j)
j!
r(r+m-j) (m-j)!Here and in the following F has p,
q and r as parameters and
argument when these are not specified.
Ìwo ¡ c F
=
+ __2 r
AU
J=Csiw-Wc_îs t
o c iwo + B = W (, - p L i/c - W 5 ' + W 5 o va ja a o ja o IwoA =1-B
D =b+
'
'co =
(-i A2t vw w C = C + ic C owi
st,
iwo c , (W 5 ' + W5'
+ ic5ttt)
o iwo o vwi. w vwl w vw A W o= - -
AU vwl 1E w (+ Th
22k
puW
w o = 0 (1) Co C=1
C = (5 o 1 1 2 1 It5''
w vw !=K
iwl AU , A2U A3I A3 2 t'' (1 + __.2 !_) (1 + ___-..2+ _._2
______ Iwo e F W W 2 o o W o -W '' vwo ic AU k? '4vw1 (2.A15) (2.A16) (2.A17) (2.A18) (2.A19) (2.A7) (2.A8) = A o (2. A9) (2.A1 0). -s)
(2.A11 ) (2.Al2)10
(2.A13) ic w (2.Ai LI)A.?'?
!
(at the surface!) "la -va 2 Eca Wo 'va - 1] 2i.
'=__..2r_!LL+
(!.L'2___E]
aclwo
3 F F U FT
M --(1/c- W0
W ('
wo) (1/ca_ + W0)-ova
ia (2.A20) (2.A22)This last expression is a translation of Miles' result into our notation; Miles did not substitute U' = Uk/c and 1h his case W = - c/ui.
References
Abramowltz, M. & Stegun, l.A., 1965. Handbook of Mathematical Functions, Dover Publ., New York.
Batchelor, G.K., 1981. An Introduction to Fluid Dynamics, Cambridge University Press, Cambridge.
Benjamin, T.B., 1959. Shearing flow over a wavy boundary, J. of Fluid
Mech. 6, pp. 161-205.
Drazin, P.G. & Reid, W.H., 1982. Hydrodynamic Stability, Cambridge University Press, Cambridge.
Hopf, L., 191k. Der Verlauf kleiner Schwingungen auf einer Strömung
reibender Flüssigkeit. Ann. Phys., Lpz. (14) 1114, pp. 1-60.
Janssen, P.A.E.M. & Peeck, H.H., 198k. On the quasilinear evolution of
the coupled air-flow, water wave system. Submitted for publication.
Kawai, S., 1977. On the generation of wind waves relating to the shear flow in water - A Preliminary Study, Tôhoku University Series 5,
Vol. 214, nos 1/2. Al (i113 £ wo, 1) - 2/3 DT
awo)
(2.A21) (?.A23) (2 A2 k)Kawai, S., 1979a. Generation of Initial Wavelets by Instability of a Coupled Shear Flow and their Evolution to Wind Waves, J. of Fluid
Mech., 93, pp. 661-703.
Kawai, S., 1979b. Discussion on the Critical Wind Speed for Wind-Wave Generation on the Basis of Shear-Flow Instability Theory, J. of Oceanographical Society of Japan, 35, pp. 179-186.
Larson, T.R. & Wright, J.W., 1.975. Wind-generated gravity capillary waves: laboratory measurements of temporal growth rates using microwave backscatter, J. Fluid Mech., 3, pp. 1417-1436.
Miles, J.W., 1957. On the Generation of Surface Waves by Shear Flows, J. of Fluid Mech., 3, pp. 185-2014.
Miles, J.W., 1962. On the Generation of Surface Waves by Shear Flows,
Part 14, J. of Fluid Mech., 1.3, pp. 2433-14148.
Mises, R. von, 1912a. Beitrag zum Oszillationsproblem. In: Féstschrift
H. Weber, pp. 252-282. Teubner, Leipzig.
Mises, R. von, 1912b. Kleine Schwingungen und Turbulenz. Jber. Deutsch. Math.-Verein 21, pp. 2141-2148.
Monin, A.S. & Yaglom, A.M., 1971. Statistical Fluid Mechanics: Mechanics of Turbulence. Vol. I, M.I.T. Press, Cambridge, Massachusetts. Plant, W.J. & Wright, J.W., 1980. Phase Speeds of Upwind and Downwind
Travêlliñg Short Gravity Waves. J. of Geophysical Research, 85, no. C6, pp. 33014-3310.
Raney, R.K., Hasselmann, K., Plant, J.W., Alpers, W., Shuchman, R.A. Jam, A. & Shemdin, O.H., 1983. Theory of SCAR Ocean Wave Imaging:
the Marsen Consensus, to be published.
Valenzuela, G.R., 1976. The Growth of Gravity-Capillary Waves in the
Coupled Shear Flow, J. of Fluid Mech., 76, pp. 229-250.
Whitham, G.B., 19714. Linear and Nonlinear Waves, John Wiley & Sons, Néw York.
3
Nonlinear interactions of gravity-capillary waves:
Lagrangian theory and effects on the spectrum
Abstract
A weakly nonlinear inviscid theory describing the interactions within a continuous spectrum of gravity-capillary waves is developed. The theory
is based on the principle of least action and uses a Lagrangian in
wavenumber-time space. Advantages of this approach compared to the method of Valenzuela & Laing (1972) are much simplified mathematics and
final equations and validity on a larger time scale. It is shown that much of the development of the spectrum under influence of nonlinear
terms can be understood without actually having to integrate the
equations. To this end multiwave space, a new concept comparable to phase space, is introduced. Using multiwave space the magnitude of the nonlinear transfer is estimated and it is shown how the energy goes through the spectrum. Also it is predicted that at fixed wavenumbers, the smallest being 520 m , finite peaks will arise in the spectrum. This is confirmed by numerical integrations. From the integrations it is
also deduced that nonlinear interactions are at least as important to
the development of the spectrum as wind growth. Finally it is shown both
analytically and numerically that the near-Gaussian statistics of the
sea surface are unaffected by nonlinear interactions.
3.1. Introduction
Remote sensing of our seas, which has taken such a flight over the last decade, gives a lot of information which the scientific world still has
difficulty in interpreting correctly. One example of this is formed by
the images of bottom topography of shallow waters taken by microwave
radar. An essential link in the imaging process is formed by
bottom topography images is that not all is known about the behaviour of these wavelets while part of what is known is still too intricate to be handled in a compound model (Phillips, 19814).
This paper deals with gravity-capillary waves. I will first give a brief
review of the present level of our knowledge of them. This is done most
easily in the context of the energy balance equation (Willebrand,
1975):
+5
+S
3t 3k 3x 3x 3k wind vise nl br'
Q = k.0 + w
Here A(k, x, t) denotes the action density, which is defined by
A = E/w; E being the local energy density and w the intrinsic frequency. Also, Q is the apparent frequency, U the surface current and S stands for source or sink.; resp. acting due to wind, viscosity,
nonlinear interactions and breaking events. Actually all sources are coupled, but as an approximation they are dealt with separately (Komen
et al., 19814).
The damping due to viscosity is well known: Sviso = -14vk2A (Phillips,
1977,
p. 52).
The equations governing the energy input by wind are also familiar (Miles, 1962, Valenzuela, 1976), but it is only recently that a quick and accurate way to solve them has been found (Van Gastel et al.,1985).
We know very little of the causes and frequency of breaking of the waves (Phillips, 19814); Banner & Phillips (19714) have nade sometheoretical predictions on this subject. The nonlinear interactions in a spectrum of gravity-capillary waves have been calculated by Valenzuela & Laing,
(1972).
A drawback of their method which is based on Hasselmann's(1962)
perturbation analysis, is that the resulting expressions for the interaction coefficients are complicated and cannot be understoodphysically; also the numerical computations necessary for quantitative results are delicate and lengthy. This is probably one of the reasons
why the nonlinear interactions have never yet been included when solving the energy balance. As Longuet-Higgins
(1976)
puts it: "there isobviously a need for a much simpler approach, more amenable to physical interpretation".
The aim of this
paper is to raise our knowledge of the nonlinearinteractions in a continuous spectrum of gravity-capillary waves to such
a level that an accurate description of these interactions can be used when solving the energy balance (3.1) for this part of the spectrum.
This work can be divided into three steps: 1) mathematical derivation of
the interaction equations, 2) construction of a physical image of how the wavenumbers are related to each other by the resonance conditions
and
3)
study of the symmetries present within a triad. Point 2) helps to make interpretations of numerical integrations possible and greatlyaffects their efficiency.
Analytical expressions for the nonlinear interactions in a continuous spectrum of gravity-capillary waves are obtained as the Euler-Lagrange equations for the Lagrangian in wavenumber-time space. An expansion is
ide in powers of the wave steepness. This method can be seen as a
generalization to the continuous case of the method of Simmons
(1969)
orof that described by Whitham
(1967)
for resonant interactions. Theadvantages of using a Lagrangian or Hamiltonian formalism in this Oase are simpler mathematics, simpler final expressions and validity on a
longer time scale. These advantages have been pointed out frequently in recent years (Miles & Salmon
1985,
Henyey1983).
To achieve the second point I introduce a new concept. This is the
multiplet-wavevector space, multiwave space for short (a multiplet is a
set of waves that
together fulfill the resonance conditions). Thisconcept can be compared to phase space in classical mechanics. Phase
space tells us at a glance how physical space
is interconnected bytrajectories. Similarly, in multiwave space we see the interconnections
in wavevector space due to the nonlinear interactions. The multiwave space is constructed explicitly for triad resonances between parallel
gravity-capillary waves. In this case it is two-dimensional.
The third point., the use of symmetries, is straightforward. It is
comparable to the way Hasselmann & Hasselmann
(1981)
use symmetries for four-wave interactions.One of the gains or using multiwave space and symmetries is that one has to solve resonance conditions and interaction coefficients only for part
of wavevector space, in the present case only for k
I
(g/2T). Another gain is that it allows for a natural construction of a grid for numerical calculations, implying high accuracy. This grid follows thepaths of the energy through the spectrum.
An energy balance containing nonlinear interactions is actually
integrated. As initial states the spectra measured by Liu & Lin (1982) are used. A surprising phenomenon is encountered: the existence of
preferred wavelengths, i.e. peaks in the energy spectrum. I explain the occurrence Of these peaks using multiwave space.
As a sideliné the issue of near-Gaussianity of the seasurface is considered. The Gaussian approximation is essential to the weakly nonlinear theory underlying multiplet interactions. I show analytically
that the Gaussian distribution is a stable equilibrium. In other words,
under influence of nonlinear interactions the third cumulant, commonly
related to phase-locking, is constantly being reduced. This result was
already obtained by Davidson in 1972 for three-wave as well as four-wave interactions in general. Thus also for gravity waves nonlinear interactions drive the spectrum to a Gaussian distribution.
3.2. The interaction equations
The equations for the surface elevation are derived using the
principle of least act-ion. Nonlinear effects are included up to first order. The surface elevation is supposed to consist of a continuum of
free gravity--dapillary waves. Viscosity is neglected.
The action J can be written as
For water waves the Lagrangian L is a function of the elevation and the potential .For infinitely deep water it is given by (Simmons,
1969,
Luke,1967):
L + T [(1 +
.C)-1] +
(Vcp.V
+ (3.3)Here.g is the acceleration of gravity and T the surface tension divided
by the density. The derivatives are defined as follows:
V
(±ì î)
'r
Bt ' Bx' ay' Bz
=(
ax' By
The Lagrariglan as presented in (3.3) is a function and time. To be able to perform the integration
I apply Fourier transformation and substitute the
vertical for each mode:
(x, t)
j:;
f c
(k, t) e k.x dk 4,(z, z,
t) = j.- j (k, t) e k.x + kz dk k Iki(3.11)
of horizontal space in the Lagrangian dependence on the (3.5)A new Lagranglan C depending on wavenumber and time is defined by
= ff
L d k d t (3.6)It can be given explicitly by assuming the waves to be of small, though finite amplitude, or, equivalently:
V k, t c (k,t) = c 4,(k,t) = (3.7)
= 0 (1)
where c is a small parameter proportional to the wave steepness. This
the following. c being small enables the Lagrangian C to be expanded in powers of c . Using the identity
f
eik
dx = 1hrk)
(3.8)this expansion becomes (here and in the following the dependence on t of all functions is not stated explicitly when no confusion can arise):
L (k, t) = C2 (g + k2T) (k) (-k) + k4,(k)Q (-k) +
+ C3
ff[(icic'
+ kk')(k);(k) C(k'') + k(k)
(k')(k'')]& (k+ k'+ k'') dk'dk''
+ O ().
(3.9)
The principle of least actIon states that physical realizations of the system are given by SJ = O. This condition can be transformed into a
condition on the Lagrangian, as in the familiar Euler-Lagrange equations. To illustrate the method I use a simplified Lagrangian t.; for
the actual Lagrangian given by (3.9) the procedure is analogous. Let t.
be a function of (k) and p (-k). Then:
j =f
t {(k)
+ u(k), (-k) + u(-k)]dk -f
t. (cp(k), q(-k))dk.
Here is a small parameter and u is a smooth function of
k
disappearing at the integration boundaries. Continuing:=
f
k) i ((k), 4(-k))) u (k) +t. ((k),
(-k)))
u (-k) ] dk = ( t. ((k), k)) +t. (k), $(k)))]
u (k) dk.Thus
= ° <>
(k) ((k), 4(-k)) + (q(-k), q(k))} =
o.
Applying this procedure to the present Lagrangian C yields two equations as there are two independent functions c and .
Using the substitution k -* -k these equations can be written as:
(k) = (g + k2T) - c
j-
ff[-k'k" + k'k") (k') (k")+ k" (k") (k')] 6 (k-k'-k")dk'dk"
c (k) = + c j ¡f{(kk' + kk') c
(k')
(k")
(3.10)-
k (k')c (k")]
6 (k-k'-k") dk' dk"The zeroth order or linear solutions are two independent functions oscillating with frequencies
(k)
= I
(gk + k3T). (3.11)These functions are the normal modes of the system. By eliminating from (3.10) and transforming to these normal modes according to
c (k,
t) = a (t) iw(k)t + a (t) -iw(k)t(3.12)
c (k,
t) = i(k) a(t)
i(k)t
-i(k)
a (t)equations for the amplitudes of the modes are derived which are uncoupled in lowest order:
-ike
'f
kk' + kk' U)()
+ U)!!)
--
-' kk' a' a' a" ot G a 'a" kk" + kk" , ,, -k'k"+ k'k" V' kk" ma" + a' + k'k" + 2(ù') + 2 (w")2 + 2w, i i(w w' ,-w" )t a a a" 6 (k-k'-k")dk'dk" c&', a,, + O (:2)The following abbreviations are used:
w = w(k),
(3.114)a = i
ra
(k, t) o = iw =1
.0
-wa=-1
a a.(k,t) a=-1
In the summation a' and a" take on the values -1 and +1, a can be both +1 or -1.
Equation (3.13) governs the nonlinear interactions of water waves. One piece of information has not been used so far and should be added; the elevation and potential are real functions. This leads to a relation
between a+ and a- . Two conventions for stating this relation are in use (Hasselmann, 1962, Davidson, 1972). I will have use for both of them here; the first convention is
Y k a(k) =
-k) , w1 = w ,w1 =
w , (3.15)and the other:
Yk, a = -1, 1 ; a(k) (-k) , w(k)
w(k)
(3.16)k>o
:'1='
The interactions described by (3.13) are cyclic in time except when the
exponent vanishes, or, more explicitly, when