Top-to-bottom Ekman layer and its implications for shallow rotating flows
Cushman-Roisin, Benoit ; Deleersnijder, Eric
DOI
10.1007/s10652-018-9611-y
Publication date
2018
Document Version
Accepted author manuscript
Published in
Environmental Fluid Mechanics
Citation (APA)
Cushman-Roisin, B., & Deleersnijder, E. (2018). Top-to-bottom Ekman layer and its implications for shallow
rotating flows. Environmental Fluid Mechanics, 19 (2019), 1105–1119.
https://doi.org/10.1007/s10652-018-9611-y
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent
of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Takedown policy
Please contact us and provide details if you believe this document breaches copyrights.
We will remove access to the work immediately and investigate your claim.
This work is downloaded from Delft University of Technology.
for Shallow Rotating Flows
BenoitCushman-Roisin
Thayer S hoolof Engineering
Dartmouth College
Hanover, New Hampshire03755, USA
Eri Deleersnijder
Institute of Me hani s, Materials andCivil Engineering(IMMC)
&Earthand LifeInstitute (ELI)
Université atholique deLouvain
4 avenue Georges Lemaître
B-1348 Louvain-la-Neuve,Belgium
Delft Instituteof Applied Mathemati s(DIAM)
Delft Universityof Te hnology
Van Mourik Broekmanweg 6
2628XEDelft, TheNetherlands
June 15,2018
Submittedto Environmental FluidMe hani s
(Spe ial Shallow Flows Issue)
ABSTRACT
Theanalyti alsolutionisderivedfor rotationalfri tionalowinashallowlayerofuidinwhi hthetop
andbottomEkmanlayersjoinwithoutleavingafri tionlessinterior. Thisverti alstru ture hassigni ant
impli ations for the horizontal ow. In parti ular, for a layer of water subje ted to both a surfa e wind
stress and bottom fri tion, the vorti ity of the horizontal ow is a fun tion not only of the url of the
wind stress (the lassi al resultfor deep waterknownasEkman pumping) but also of its divergen e. The
importan eofthisdivergen etermpeaksforawaterdeptharound3timestheEkmanlayerthi kness. This
means that a url-free but non-uniform wind stress on a shallowsea or lake an, through the dual a tion
of rotationand fri tion, generate vorti ity inthewind-driven urrents.
Wealsondthattheredu tionofthree-dimensionaldynami stoatwo-dimensionalmodelismoresubtle
than one ouldhave anti ipated and needs to be approa hed with utmost are. Taking thebottom stress
as dependent solely on the depth-averaged ow, even withsome veering, is not appropriate. The bottom
stress ought to in lude a omponent proportional to thesurfa e stress,whi his negligiblefor large depths
1 Introdu tion
The ee t of fri tion in rotating ows is governed by the Ekman number, and it is well known (Pri e et
al.,1987; Cushman-Roisin&Be kers, 2011) that,for anEkman number mu hsmallerthan unity,therole
of fri tion issigni ant only inthin layers nearthetop andbottom boundaries, alledEkman layers, and
o asionallynearlateralboundaries,inso- alledStewartsonlayers(Bennetts&Ho king,1973). Awayfrom
these boundary layers, inthebodyofthedomain alledtheinterior,theee t offri tion isfeltindire tly.
Theprimary signatureisthe Ekman pumping,averti al velo itythatstarts atthebaseofthetop Ekman
layer as proportional to the url of the surfa e stress and ends at the top of the bottom Ekman layer as
proportionalto the urlofthebottom stress. Onthe
β−
plane,thesurfa e andbottomstressesmaydier, andtheEkmanvelo itymayvaryovertheverti al,butonanf −
plane,theEkmanpumping velo itymust be uniform over the verti al, and the surfa e and bottom stresses be equal (Cushman-Roisin & Be kers,2011).
Thesituation be omesmu hmore omplexwhenthe rotatingdomainisshallowenoughfor thetop and
bottomEkmanlayerstomerge,leavingnofri tionlessinterior. Thebehavioroftheverti alvelo ityismu h
ri her, and the on ept of Ekman pumping is no longer helpful. The bibliography of this work is limited
be ause, although the governing equations are straightforward, the analyti al solution of the linearized
problemis tediousandtends to obs ure thepropertiesof theow.
Extending Ekman's se ond paper on fri tional rotating ows (Ekman, 1923) and with the fo us on
stormsurgesin shallowseas, Welander (1957) onsidered the asewhen the depthof fri tionalinuen e
is omparable tothe water depthand arrived at the on lusion that thewind-stressdivergen e may be of
the same importan e asthe wind-stress url in governing the streamfun tion of the depth-averaged ow,
but he didnot onsider theverti al velo ityand qui klypro eeded to thetime-dependent ase ofa storm
surge. Inthe1970s, whenele troni omputers rstbe ame availableandusedforstormsurgefore asting,
they were not yet apable of handling three-dimensional grids, and a urry of work ensued to redu e the
3D equations to 2D equations bypre-solving for theverti al prole of thehorizontal velo ity. This work
(Jelesnianski, 1970; Forristall, 1974; Cheng et al., 1976; Nihoul, 1977) largely fo used on the ee ts of
a verti ally varying eddy vis osity to obtain the bottom stress in terms of the depth-averaged horizontal
velo ity. No attention waspaidto the verti al velo ity. Ade adelater, as3Dnumeri al models ameinto
being, Lyn h and O er (1985) paid renewed attention to the verti al stru ture of vis ous rotating ow
in order to provide analyti solutions astest ases for 3D hydrodynami models. Like their prede essors,
they hosedepth-varying eddyvis osities and tookthe bottom stress as aligned with and proportional to
thedepth-averaged velo ity. Theirsolutions are astassolutions to an eigenvalue problemintheverti al,
whi h requiresthenumeri al solution ofa 1Dequation.
In this work, we ask dierent questions: What happens to the on ept of Ekman pumping when the
Ekmannumberis notsmall? Whatisthe verti al proleoftheverti alvelo ity? Howdolo alwind-stress
urland divergen e ae tthedepth-averagedow? And,underwhi h onditions an thebottom stressbe
takenasparallelto the depth-averaged velo ity? Theanswerto thelatterpointsto unexpe ted di ulties
intheredu tionof3Ddynami s toa2Dmodeland might astdoubts onthea ura yofpast2Dmodels.
2 Top-to-Bottom Ekman Dynami s
We are onsidering a shallow ow ofwater ofuniform density subje tto an imposed surfa e stress at the
top andno slip alonga at horizontalbottom.
We solve the lassi al Ekman dynami s problem in the ontext of the following assumptions:
f
-plane (withf > 0
for onvenien e), low-Rossby number (to permit the negle t of nonlinear adve tion terms and fa ilitatetheanalysis),hydrostati balan e (∂p/∂z = −ρg
), onstant densityρ
(no strati ation),at bottom and rigid lid (−H ≤ z ≤ 0
,H =
onstant), eddy vis osity model with uniform eddy vis osityν
,and imposedsurfa e wind stress (
τ
x
,τ
y
fun tionsofx
andy
).Were ognize thatthenegle tof thetemporal derivativesand nonlinearadve tion termsis notjustied
inmany ir umstan es,but wenonetheless ignorethosetermsheretopla ethefo usonthe ombinedrole
of Coriolis and fri tion for es in a mathemati ally tra table fashion, with the underlying hypothesis that
themainndings ofour study will ontinue to holdat leastqualitatively inthebroader dynami alsetting
thatin ludes temporal derivatives and nonlinearadve tion terms. It also happens thatthe assumption of
a lowRossbynumber andtemporal Rossbynumberis justiable wheneverthes ales of theproblem obey
thefollowing relations:
Ek =
ν
f H
2
≥ 1,
Ro =
U
f L
=
τ
ρf
2
HL
≪ 1,
Ro
T
=
1
f T
≪ 1,
(1)in whi h
L
andU
are horizontal length and velo ity s ales,T
is the time s ale, andτ
is a s ale for the windstress. Su hrelationsarenotunrealisti forshallowwaterows. For example,thevaluesf ∼ 10
−4
/s,H ∼ 10
m,L ∼ 10
km,T ∼ 2 × 10
5
s (=several days),τ /ρ ∼ 10
−4
m2
/s2
, fri tionvelo ityu
∗
=
pτ/ρ ∼
10
−2
m/s,U ∼ τ /ρf H ∼ 0.1
m/s,andν ∼ u
∗
H ∼ 10
−1
m2
/s do meet the pre eding riteria (
Ek ∼ 10
,Ro ∼ 0.1
,andRo
T
∼ 0.05
).Webegin theanalysis byexpressing the horizontal pressure gradient for ein termsof geostrophi ow
omponents:
−f v
g
= −
1
ρ
∂p
∂x
(2)+f u
g
= −
1
ρ
∂p
∂y
,
(3)fromwhi h followsthat onan
f
-plane,the geostrophi owiswithout divergen e:∂u
g
∂x
+
∂v
g
∂y
= 0.
(4)It isalso
z
-independent byvirtue ofthe hydrostati balan e with onstant density. The equations governing thethree velo ity omponentsu
,v
andw
are:−f v = −f v
g
+ ν
∂
2
u
∂z
2
(5)+f u = +f u
g
+ ν
∂
2
v
∂z
2
(6)∂u
∂x
+
∂v
∂y
+
∂w
∂z
= 0.
(7)There arethree boundary onditions at ea hhorizontal boundary.
•
Atthe bottom,weimposeno ow onthefri tionaland impermeablebottom:u = v = w = 0
atz = −H.
(8)•
Atthe top,we impose asurfa e wind stressand noverti alowthrough therigid lid:ρν
∂u
∂z
= τ
x
,
ρν
∂v
∂z
= τ
y
,
w = 0
atz = 0.
(9)WenotethatthesetofEquations(5)-(6) -(7) ontains atotalofve
z
-derivativeswhilewehavespe ied six boundary onditions in thez
-dire tion. Thus, the system is over-spe ied by one, and we ought to expe t a onstraint on some of the variables that enter the equations, parti ularly a relationship betweenthegeostrophi velo ity(
u
g
,v
g
),whi h isthehorizontal pressuregradient indisguise,and thewind stress (τ
x
,τ
y
).Thesituation anbeanti ipatedby onsideringthe aseofadeepwater olumn omparedtotheEkman
layerthi kness, dened as
d =
r 2ν
f
.
(10)If
d ≪ H
, fri tion is signi ant only in top and bottom boundary layers that o upy mu h less than the waterdepth,leavingafri tionlessandgeostrophi interior. Inthis ase, surfa eEkmandynami sgeneratesaso- alled Ekmanpumping verti alvelo itythroughthe interior(Cushman-RoisinandBe kers, page254)
proportional tothe wind stress url:
w
interior
=
1
ρf
∂τ
y
∂x
−
∂τ
x
∂y
,
(11)while bottom Ekman dynami s generate an interior verti al velo ity proportional to the vorti ity of the
geostrophi ow(Cushman-Roisin andBe kers, page 249):
w
interior
=
d
2
∂v
g
∂x
−
∂u
g
∂y
.
(12)Given that the verti al velo ity in the interior takes a unique value be ause theinterior geostrophi ow
hasno divergen e, itfollowsthat the pre edingexpressions mustbeequal to ea h other:
∂v
g
∂x
−
∂u
g
∂y
=
2
ρf d
∂τ
y
∂x
−
∂τ
x
∂y
.
(13)Inotherwords,inwaterdeep omparedtotheEkmanlayerthi kness,thevorti ityoftheinteriorgeostrophi
owis onstrainedtobeproportionaltothe urlofthesurfa ewindstress,alo aldependen y. Theprimary
aim of this study is to determine how this relationship is ae ted when thewater depth is omparable to
theEkman layerthi kness(
d ∼ H
,i.e.ν/f H
2
∼ 1
)and fri tion ae tstheentire water olumn.At this point, we ould pro eed to solve Equations (5) -(6)for
u
andv
,but the algebra is very tedious (Welander, 1957). A more e ient path toward the relationship we seek is not to solve for the velo ityomponentsthemselvesbut fortheir divergen e
∂u/∂x + ∂v/∂y
andvorti ity∂v/∂x − ∂u/∂y
. Forthis,we dierentiate Equations (5)-(6) and ombine to obtain the following oupled equations for the divergen eand vorti ity:
f
∂u
∂x
+
∂v
∂y
= f
∂u
g
∂x
+
∂v
g
∂y
+ ν
∂
2
∂z
2
∂v
∂x
−
∂u
∂y
(14)f
∂v
∂x
−
∂u
∂y
= f
∂v
g
∂x
−
∂u
g
∂y
− ν
∂
2
∂z
2
∂u
∂x
+
∂v
∂y
.
(15)In (14),thedivergen e ofthegeostrophi owvanishesbyvirtue of (4) , and thegeneral solution ofthe
equations an be written as:
∂u
∂x
+
∂v
∂y
= cosh ζ (D cos ζ − C sin ζ) + sinh ζ (B cos ζ − A sin ζ)
(16)∂v
∂x
−
∂u
∂y
=
∂v
g
∂x
−
∂u
g
inwhi h thedimensionlessverti al oordinate
ζ
isdened byζ =
z + H
d
.
(18)It vanishes at thebottom and rea hes
H/d
at thesurfa e.The four onstants ofintegration
A
,B
,C
andD
are determined bythe appli ation oftop and bottom boundary onditions. Thehorizontalvelo ityvanishesalongthebottom(ζ = 0
),andsomustitsdivergen e and vorti ity. Thisdeterminesthe oe ientsA
andD
:A = −
∂v
g
∂x
−
∂u
g
∂y
,
D = 0.
(19)Before applying boundary onditions at the surfa e, it is ne essary to determine the verti al velo ity
a rossthewater olumn,whi h isobtained byintegrating the ontinuity equationverti ally,starting from
w = 0
at thebottom:w = −
Z
z
0
∂u
∂x
+
∂v
∂y
dz
′
= −d
Z
ζ
0
−C cosh ζ
′
sin ζ
′
+ sinh ζ
′
(B cos ζ
′
− A sin ζ
′
)
dζ
′
=
d
2
(B + C)(1 − cosh ζ cos ζ) −
d
2
(B − C) sinh ζ sin ζ
+
d
2
A (cosh ζ sin ζ − sinh ζ cos ζ).
(20)
Enfor ing
w = 0
atthe surfa e (z = 0
,ζ = H/d
)yields:A
cosh
H
d
sin
H
d
− sinh
H
d
cos
H
d
+ (B + C)
1 − cosh
H
d
cos
H
d
− (B − C) sinh
H
d
sin
H
d
= 0 .
(21) Theappli ationoftheremainingtwoboundary onditionsatthesurfa e,whi hinvolvethesurfa ewindstress,ne essitates the preliminarytaking ofthe
z
-derivative of thedivergen e andvorti ity:∂
∂z
∂u
∂x
+
∂v
∂y
=
1
d
[(B − C) cosh ζ cos ζ − (B + C) sinh ζ sin ζ]
−
1
d
A(cosh ζ sin ζ + sinh ζ cos ζ)
(22)
∂
∂z
∂v
∂x
−
∂u
∂y
=
1
d
[(B + C) cosh ζ cos ζ + (B − C) sinh ζ sin ζ]
−
1
d
A(cosh ζ sin ζ − sinh ζ cos ζ) .
(23)
Thestress boundary onditions (9) an nowbe applied at thetop (
ζ = H/d
):∂τ
x
∂x
+
∂τ
y
∂y
=
ρf d
2
(B − C) cosh
H
d
cos
H
d
− (B + C) sinh
H
d
sin
H
d
−
ρf d
2
A
cosh
H
d
sin
H
d
+ sinh
H
d
cos
H
d
(24)∂τ
y
∂x
−
∂τ
x
∂y
=
ρf d
2
(B + C) cosh
H
d
cos
H
d
+ (B − C) sinh
H
d
sin
H
d
−
ρf d
2
A
cosh
H
d
sin
H
d
− sinh
H
d
cos
H
d
.
(25)Sin e the oe ient
A
is already known in terms of the vorti ity of the geostrophi ow [see (19) ℄, the pre edingtwo equations forma 2×
2systemof equations forB − C
andB + C
. The solutionis:B − C =
2
ρf d∆
∂τ
y
∂x
−
∂τ
x
∂y
sinh
H
d
sin
H
d
+
2
ρf d∆
∂τ
x
∂x
+
∂τ
y
∂y
cosh
H
d
cos
H
d
−
1
∆
∂v
g
∂x
−
∂u
g
∂y
cosh
H
d
sinh
H
d
+ cos
H
d
sin
H
d
(26)B + C =
2
ρf d∆
∂τ
y
∂x
−
∂τ
x
∂y
cosh
H
d
cos
H
d
−
2
ρf d∆
∂τ
x
∂x
+
∂τ
y
∂y
sinh
H
d
sin
H
d
+
1
∆
∂v
g
∂x
−
∂u
g
∂y
cosh
H
d
sinh
H
d
− cos
H
d
sin
H
d
,
(27)after
A
wasrepla ed byits valuegivenin(19) andwith∆
,thedeterminant ofthesystem, dened as∆ = cosh
2
H
d
cos
2
H
d
+ sinh
2
H
d
sin
2
H
d
= sinh
2
H
d
+ cos
2
H
d
.
(28)Substitutionin(21)yieldsthesought-afterrelationshipbetweenthegeostrophi owandthewindstress:
∂v
g
∂x
−
∂u
g
∂y
cosh
H
d
sinh
H
d
− cos
H
d
sin
H
d
=
2
ρf d
∂τ
y
∂x
−
∂τ
x
∂y
cosh
2
H
d
+ cos
2
H
d
− cosh
H
d
cos
H
d
− 1
+
2
ρf d
∂τ
x
∂x
+
∂τ
y
∂y
sinh
H
d
sin
H
d
.
(29)This last equation is the relation between the geostrophi ow (
u
g
,v
g
), and the wind stress (τ
x
,τ
y
) that extends Equation(13) to the ase of shallowwater(d ∼ H
,i.e.ν/f H
2
∼ 1
). Itwasderived byWelander (1957, Equation (17)) albeit asa more ompli ated mathemati al expression. We note that thevorti ityof the geostrophi ow is no longer simply proportional to the url of the wind stress but is now a linear
ombination of both url and divergen e of thewind stress. In thelimit
H ≫ d
,Equation (29) reverts to (13)asit should.To better dis ern the dependen y of the geostrophi vorti ity on the url and divergen e of the wind
stress,we rewrite (29) as:
∂v
g
∂x
−
∂u
g
∂y
=
2
ρf d
F
H
d
∂τ
y
∂x
−
∂τ
x
∂y
+
2
ρf H
G
H
d
∂τ
x
∂x
+
∂τ
y
∂y
,
(30)in whi h
F
andG
are dimensionless fun tions of the dimensionless variableH/d
. These are plotted on Figure 1. We note that, forH > 7d
,F ≃ 1
andG ≃ 0
, orresponding to the deep-water situation with geostrophi vorti ity proportional to the wind-stress url only. In the limitH/d = 0
,F
vanishes,G =
3
4
,and the situation has ipped with the geostrophi vorti ity now purely proportional to the wind-stress
divergen e. Note that in the transition regime (
0 < H < 7d
),F
rea hes a maximum of1.225
, whi h ex eeds its asymptoti value1
, andG
rea hes a minimum of−0.1113
, whi h falls below its asymptoti valueof0
.Figure 1: The dimensionless fun tions
F
andG
of thethi kness ratioH/d
. The fun tionF
provides the dependen eof the geostrophi vorti ityonthe wind-stress url,andG
onthewind-stress divergen e.Caution is in order. When the water olumn is deep, the interior geostrophi ow is asymptoti ally
equalto thedepth-averagedow,but isno longersointhe aseofshallowwater be auseEkmandynami s
a tovertheentire water olumn. Thus,weneed tomake adistin tionbetween thegeostrophi omponent
of theowand the depth-averaged ow.
Wedene thedepth-averaged velo ityas
¯
u =
1
H
Z
0
−H
u dz ,
¯
v =
1
H
Z
0
−H
v dz .
(31)Verti alintegration of Equations(14) -(15) fromtop to bottom
†
followed byutilization ofboundary
ondi-tions (9)and useof (22)-(23)provides:
∂ ¯
u
∂x
+
∂¯
v
∂y
=
1
ρf H
∂τ
y
∂x
−
∂τ
x
∂y
−
d
2H
(B + C)
(32)∂¯
v
∂x
−
∂ ¯
u
∂y
=
∂v
g
∂x
−
∂u
g
∂y
−
1
ρf H
∂τ
x
∂x
+
∂τ
y
∂y
+
d
2H
(B − C).
(33)Further repla ement of
B + C
andB − C
bytheirvalues givenin(27) -(26) leads to:∂ ¯
u
∂x
+
∂¯
v
∂y
= −
d
2H∆
∂v
g
∂x
−
∂u
g
∂y
cosh
H
d
sinh
H
d
− cos
H
d
sin
H
d
+
1
ρf H
∂τ
y
∂x
−
∂τ
x
∂y
1 −
1
∆
cosh
H
d
cos
H
d
+
1
ρf H
∂τ
x
∂x
+
∂τ
y
∂y
1
∆
sinh
H
d
sin
H
d
(34)∂¯
v
∂x
−
∂ ¯
u
∂y
=
∂v
g
∂x
−
∂u
g
∂y
1 −
d
2H∆
cosh
H
d
sinh
H
d
+ cos
H
d
sin
H
d
+
1
ρf H
∂τ
y
∂x
−
∂τ
x
∂y
1
∆
sinh
H
d
sin
H
d
−
1
ρf H
∂τ
x
∂x
+
∂τ
y
∂y
1 −
1
∆
cosh
H
d
cos
H
d
.
(35)†
Finally, eliminationof the vorti ityof the geostrophi ow byuseof (29)provides:
∂ ¯
u
∂x
+
∂¯
v
∂y
= 0,
(36) and∂¯
v
∂x
−
∂ ¯
u
∂y
=
2
ρf d
F
¯
H
d
∂τ
y
∂x
−
∂τ
x
∂y
−
2
ρf d
G
¯
H
d
∂τ
x
∂x
+
∂τ
y
∂y
.
(37)The rst of these equations returns the expe ted result that the depth-averaged ow has no divergen e
(in ompressible owbetween two parallel impermeableboundaries), while these ond equationis the
on-strainton theow[Equation (29) ℄nowexpressedintermsofthevorti ityofthedepth-averagedow. This
equation was rstderived by Welander (1957, Equation (26)), but no dis ussion wasprovided otherthan
pointingout thatthere isa dependen y onthewind-stressdivergen e.
Figure 2: The dimensionless fun tions
¯
F
and¯
G
of thethi kness ratioH/d
. The fun tion¯
F
provides the dependen eof the depth-averaged vorti ityon the wind-stress url,andG
¯
on thewind-stressdivergen e.Figure2showsthedependen yofthe
F
¯
andG
¯
fun tionsonthethi knessratioH/d
. Thedependen yof thedepth-averaged vorti ityonthewind-stress url(F
¯
fun tion) isfairlysimilartothatofthegeostrophi vorti ity(F
fun tion),startingfrom0
andasymptotingto1
;theintermediatemaximumnowis0.9259
,and there isan intermediate minimum of0.8903
. In ontrast, thedependen y on thewind-stressdivergen e is quitedierent. First,thereisareversalinsign(notshowninFigure2but madeexpli itinEquation(37) ).Indeep water (
H/d → ∞
),itslowly vanishes;intheshallowlimit (H/d → 0
),thefun tionG
¯
rea heszero morerapidlythanthefun tion¯
F
,implyingthat,whenH << d
,thewind-stress urlremainsthedominant ontribution to the depth-averaged vorti ity (as shown in the next se tion). The fun tionG
¯
rea hes a maximum of0.1730
forH/d = 3.17
. In other words, the ontribution of the wind-stress divergen e to the depth-averaged vorti ity is only signi ant at intermediate depths. Figure 3 displays the ratio of theoe ients of the divergen e and url terms in (37) (without the minus sign). We note that it rea hes a
maximumof
0.1891
(19%) atH/d = 3.265
.3 Case of Very Shallow Flow
Figure3: The ratio ofthe oe ientsof the divergen e and url termsinEquation(37) .
thelimit
H/d → 0
inboth (29)and (35) . At the leading order, these equations be ome, respe tively:∂v
g
∂x
−
∂u
g
∂y
≃
3
2ρf H
∂τ
x
∂x
+
∂τ
y
∂y
(38)∂¯
v
∂x
−
∂ ¯
u
∂y
≃
H
ρf d
2
∂τ
y
∂x
−
∂τ
x
∂y
.
(39)Wenotefrom Equation(38) thatthevorti ityofthe geostrophi owis now proportionalto the
diver-gen eofthewind stress,a ompleteipfrom(13)where itwasproportionalto the urlofthewindstress.
Equation(39) forthe depth-averaged ow issimilar to(13) but witha oe ient thathasan extrapower
of theratio
H/d
.4 Impli ations for 2D Modeling
Let us suppose that we had not bothered about 3D Ekman dynami s and insteadhad usedthe following
two-dimensional model withlinearfri tion:
−f ¯
v = −f v
g
+
τ
x
ρH
− r¯
u
(40)+f ¯
u = +f u
g
+
τ
y
ρH
− r¯
v
(41)∂ ¯
u
∂x
+
∂¯
v
∂y
= 0,
(42)inwhi h thewindstress isapplied asabodyfor eand
r
isafri tion oe ient,whi hshouldberelatedto thevis osityν
somehow. We useoverbars onthevelo ity omponentsto indi atethat theseareverti ally uniform, i.e. theyaredepth averages. Thus, this modelassumes a bottom stress ounter-alignedwiththedepth-averagedvelo ity. Su ha 2Dmodel,oftenextendedto in lude timederivatives, nonlinearadve tion
terms, and aquadratri expressionfor the bottom stress haslongbeen astaple ofwind-driven ir ulation
modeling for lakes and oastal areas (Heaps, 1969 - Equation (5); Li k, 1976 - Page 58; S hwab, 1983
-We ask what value should be given to the fri tion oe ient
r
in terms of the kinemati vis osityν
and the other parameters in the formalism,f
andH
, in su h a wayas to render this 2Dmodelan exa t representation ofthe 3Dmodelaftereliminationofthe verti al stru tureoftheow. For this,webeginbysolving themomentumequations (40) -(41) for
u
¯
andv
¯
,whi h isstraightforward:¯
u =
f
2
f
2
+ r
2
u
g
−
rf
f
2
+ r
2
v
g
+
1
ρH(f
2
+ r
2
)
(rτ
x
+ f τ
y
)
(43)¯
v =
f
2
f
2
+ r
2
v
g
+
rf
f
2
+ r
2
u
g
+
1
ρH(f
2
+ r
2
)
(rτ
y
− f τ
x
) ,
(44) andthenpro eedbytakingthedivergen e∂ ¯
u/∂x+∂¯
v/∂y
,whi hwesettozeroas ontinuity(42)demands. Theresult is:∂v
g
∂x
−
∂u
g
∂y
=
1
ρrH
∂τ
y
∂x
−
∂τ
x
∂y
+
1
ρf H
∂τ
x
∂x
+
∂τ
y
∂y
,
(45)whi h is isomorphi to (30) in that the vorti ity of the geostrophi ow is a linear ombination of the
divergen e and url ofthe wind stress. Mapping(45) onto (29) demands thefollowing two equivalen ies:
r
f
=
d
2H
cosh
H
d
sinh
H
d
− cos
H
d
sin
H
d
cosh
2
H
d
+ cos
2
H
d
− cosh
H
d
cos
H
d
− 1
(46)2H
d
sinh
H
d
sin
H
d
= cosh
H
d
sinh
H
d
− cos
H
d
sin
H
d
.
(47)Whiletherstoftheseequationsservestosetthevalueforthefri tion oe ient
r
intermsofthevis osityν
(subsumedind
),the se ondequationought to be satised,too. A glan eatthis equationrevealsthatit is impossible to satisfyit for an arbitraryvalue†
of theratio
H/d
. Thus, we arebrought to on lude that the redu ed two-dimensional model (40) -(41) -(42) is not a legitimate redu tion of the three-dimensionalmodel(5)-(6) -(7) .
One may think at this point that the veering involved in Ekman dynami s should ause the bottom
stress tobe ata non-zero angleto themean velo ityandthat thepresen eofthis angle wouldresolvethe
matter. Surprisingly, however, it does not: Equation (47) arises on eagain. A non-intuitive remedy is to
turnthe surfa ewindstressbya ertainangle. Thisdoesnot seemtohave aphysi albasisuntilonethinks
of it as a parameterization of the bottom stress as a linear ombination of the mean velo ityand surfa e
stress ratherthan beingdependent onthe meanvelo ityalone.
For an angle
α
of rotation, positive inthetrigonometri sense, we try2Dequations of theform:−f ¯
v = −f v
g
+
cos α τ
x
− sin α τ
y
ρH
− r¯
u
(48)+f ¯
u = +f u
g
+
sin α τ
x
+ cos α τ
y
ρH
− r¯
v
(49)∂ ¯
u
∂x
+
∂¯
v
∂y
= 0.
(50)of whi h thesolution for
u
¯
andv
¯
is:¯
u =
f
2
u
g
− rf v
g
f
2
+ r
2
+
(f sin α + r cos α)τ
x
+ (f cos α − r sin α)τ
y
ρH(f
2
+ r
2
)
(51)¯
v =
f
2
v
g
+ rf u
g
f
2
+ r
2
+
(f sin α + r cos α)τ
y
− (f cos α − r sin α)τ
x
ρH(f
2
+ r
2
)
,
(52)†
withthe following zero-divergen e ondition:
∂v
g
∂x
−
∂u
g
∂y
=
f cos α − r sin α
ρrf H
∂τ
y
∂x
−
∂τ
x
∂y
+
f sin α + r cos α
ρrf H
∂τ
x
∂x
+
∂τ
y
∂y
.
(53)Mapping(53) onto (29) demands thefollowing two equivalen ies:
f
r
cos α − sin α =
2H
d
cosh
2
H
d
+ cos
2
H
d
− cosh
H
d
cos
H
d
− 1
cosh
H
d
sinh
H
d
− cos
H
d
sin
H
d
(54)f
r
sin α + cos α =
2H
d
sinh
H
d
sin
H
d
cosh
H
d
sinh
H
d
− cos
H
d
sin
H
d
.
(55)
Both onditions an now be met as the two onditions now form a
2 × 2
systemof equations for the two unknownsr/f
andα
. Figure 4 displays the variations of the ratior/f
(magnied by a fa tor100
) and angleα
as fun tions of the thi kness ratioH/d
. As an be expe ted, both vanish for great depths (withr/f ≃ d/2H
). As the waterdepthde reases, fri tionbe omesmore important, asexpe ted, and theangleα
swings between a minimum of−8.910
◦
and a maximum of
+48.17
◦
. The angle
α
be omes appre iable (|α| > 5
◦
) when
H/d
fallsbelow6
.Figure 4: The fri tion fa tor
r
(s aled and magnied) and angleα
of veering (in degrees) that are the solutions to Equations (54) -(55) .Nowthatwehaveidentiedana eptableredu tion, we anextra tthe orrespondingexpressionofthe
bottomstress. WritingEquations (48) -(49)as
−f ¯
v = −f v
g
+
τ
x
ρH
−
τ
bx
ρH
(56)+f ¯
u = +f u
g
+
τ
y
ρH
−
τ
by
ρH
(57)to serve asthedenitionsof the bottomstress omponents (
τ
bx
, τ
by
),we obtain:τ
bx
= (1 − cos α) τ
x
+ sin α τ
y
+ ρrH ¯
u
(58)τ
by
= (1 − cos α) τ
y
− sin α τ
x
+ ρrH ¯
v .
(59) Thus, theproperwayto parameterizethebottom stressis toadd asurfa e stress omponent toa lassi alforin reasing depthbe ausetheangle
α
vanishesasH/d → ∞
,asFigure4shows. Theneedtoin ludethe surfa e stress inthe formulation of the bottom stress was already re ognized byNihoul (1977 - Equation(61)), but without re ourse to the pre eding developments, he did not make provision for veering and
obtained a oe ient (
m
inhis notation)witha weak andmonotoni dependen e on thewaterdepth.5 Con lusions
We investigated the ase of rotating shallow water when Ekman dynami s extend throughout the water
olumn insteadofbeingrestri tedtothinboundarylayers undersurfa e andabove bottom. Although this
had been investigated in parts some time ago, the omplexity of the algebra obs ured some of the ow
properties. Chiey,Welander (1957) did identify that thedepth-averaged ow vorti ity dependsnot only
onthesurfa estress url(asitdoesindeepwater)butalsoonitsdivergen e;however,hedidnotdo ument
the importan e of the new term. Using less ompli ated algebrai manipulations, we show here that the
divergen e term vanishes in both limits of very deep (
H/d → ∞
) and very shallow water (H/d → 0
), withlargest signi an e on the orderof 20% for awaterdepth around3 times theEkman layer thi knessd =
p2ν/f
.Thephysi alinterpretationofthedependen eoftheverti alvelo ity,andhen eoftheowvorti ity,on
both urlanddivergen eofthesurfa estressisasfollows. Inredu eddepth,wherethetopandbottom
Ek-manlayersoverlap,ea hoftheselayers
†
is, inasense,in omplete. Thisimpliesthattheverti allyaveraged
transportinthetoplayerisatlessthanthe full
90
◦
to the surfa estress. The divergen e ofthistransport,
whi hindu estheverti alvelo ity,thus ontainstwoterms: Thetransport omponentperpendi ulartothe
surfa e stress ontributes the lassi al term proportional to the url, while the omponent parallel to the
surfa estress ( ausedbythedeparture from
90
◦
) ontributes ase ondterm proportionaltothedivergen e
of thesurfa e stress. Thelatter ontribution vanishes whenthe depthbe omessu ient large to reate a
de oupling of both Ekman layers, with the transportin the top Ekamn layer then oriented at
90
◦
to the
surfa e stress.
Although the surfa e stress divergen e ee t annot be dominant in the presen e of a nite surfa e
stress url, it will dominate wherever and whenever the surfa e stress url is weak or nil in water with
depth omparable to the Ekman layerthi kness. Thismeans thata url-free but non-uniform wind stress
on a shallow sea or lake an, through the dual a tion of rotation and fri tion, generate vorti ity in the
wind-driven urrents. To the knowledgeof theauthors,this had not been identied earlier.
This analysis also has impli ations for the redu tion of three-dimensional hydrodynami s to a
two-dimensional model. We show that, for su h a redu tion to be faithful to 3Ddynami s, itis inappropriate
to alignthebottom stresswiththe depth-averaged velo ity,and this astdoubts onnumeroussimulations
performedoverthelastvede adeswithtwo-dimensionalmodels. Wefurthershowthatinvoking aveering
anglebetween bottom stress anddepth-averagedowstill doesnot uretheproblem. Instead,thebottom
stress must be taken as onsistingof two omponents, one depending on the surfa e stress (with veering)
and theotherdependingon the depth-averaged ow(without veering).
ACKNOWLEDGEMENTS
The rst author expresses his gratitude to Prof. GertJan van Heijst for having organized the 2017
Symposiumon Shallow Flowsin Eindhoven. These ond authorwishes to a knowledge past supportfrom
theBelgianFund forS ienti Resear h(F.R.S.
−
FNRS)inre ognition ofthefa tthatsome elementsof thepresent paperoriginated while heserved asResear h Asso iateof theFNRSearlier inhis areer.†
Thereissomesubje tivityastowherethetoplayerendsandthebottomonebegins,obviously,butthisdoesnotneedto
REFERENCES
Bennetts,D.A.,andL.M.Ho king,1973: OnnonlinearEkmanandStewartsonlayersinarotatinguid.
Pro . R. So . Lon. A,333,469-489.
Cheng, R. T., T. M. Powell, and T. M. Dillon, 1976: Numeri al models of wind-driven ir ulation in
lakes. Appl. Math. Modelling, 1,141-159.
Cushman-Roisin, B., and J.-M. Be kers, 2011: Introdu tion to Geophysi al Fluid Dynami s Physi al
andNumeri al Aspe ts. A ademi Press,2nd ed., 828 pages.
Ekman,V.W.,1923: ÜberHorizontalzirkulationbeiwinderzeugtenMeeresströmungen. ArkivMat. Astr.
Fysik,17, No. 26.
Forristall,G. Z., 1974: Three-dimensional stru ture of storm-generated urrents. J. Geophys. Res., 79,
No. 18,2721-2729.
Heaps,N.S.,1969: Atwo-dimensional numeri al sea model. Philos. Trans. R. So . A,265,93-137.
Jelesnianski, C.P., 1970: Bottomstress time-history in linearizedequations of motion for storm surges.
Mon. Weather Rev., 98,No. 6,462-478.
Józsa,J., 2014: Ontheinternalboundarylayerrelatedwindstress urlanditsroleingeneratingshallow
lake ir ulations. J.Hydrol. Hydrome h., 62(1), 16-23. doi: 10.2478/johh-2014-0004.
Lenz,S.,R.T. Guza,S.Elgar, F. Feddersen,T. H.C.Herbers, 1999: Momentum balan esontheNorth
Carolina innershelf. J. Geophys. Res. O eans,104,18,205-18,226.
Li k, W.,1976: Numeri al modeling oflake urrents. Ann. Rev. Earthand Planetary S i., 4,49-74.
Lyn h, D.R., and C. B. O er, 1985: Analyti test ases for three-dimensional hydrodynami models.
Int. J.Numer. Methods in Fluids,5, 529-543.
Nihoul, J. C. J., 1977: Three-dimensional model of tides and storm surges in a shallow well-mixed
ontinental sea. Dyn. Atmos. O eans, 2,29-47.
Pri e,J.J., R.A.Weller,and R.R.S hudli h,1987: Wind-driven o ean urrentsandEkman transport.
S ien e, 238,No. 4833, 1534-1538.
S hwab, D. J., 1983: Numeri al simulation of low-frequen y urrent u tuations in Lake Mi higan. J.
Phys. O eanogr., 13, 2213-2224.
Weisberg, R.H., and L. Zhang, 2008: Hurri ane storm surge simulations omparing three-dimensional
withtwo-dimensional formulations basedon an Ivan-like stormover theTampa Bay,Florida region.
J. Geophys. Res. O eans, 113,doi:10.1029/2008JC005115.
Welander, P., 1957: Wind a tion on a shallow sea: Some generalizations of Ekman's theory. Tellus, 9,