ESOE 5136 2021 April 26 - Handout Lo

Derivation of the Boussinesq equations in constant water depth

The derivation shown here is based on the textbook by Mei et al. (2005).

1. Characteristic scales
Consider water waves traveling in constant water depth h. The waves have a characteristic
amplitude A, characteristic wavelength
√ L, characteristic wave number k = 2π/L, and
characteristic long wave speed gh, in which g is the gravitational acceleration. The
parameter  = A/h is a measure of how nonlinear the waves are; the parameter µ = kh is
a measure of how frequency dispersive the waves are.

2. Normalization
Using primes to denote dimensional variables, we normalize the variables as follows:

z0

0 0 0
p
(x, y) = k(x , y ), z = , t = t · k gh





 h

η0

 k
, φ = φ0 · A √


 η =
A

h gh



k
√ √k . (1)
 ∂φ ∂φ ∂φ0 Ah gh ∂φ0 Ah gh 1

 (u, v) = ( , ) = ( , ) = (u, v) · √
∂x ∂y ∂x0 k ∂y 0 k A
h gh





 k
√
∂φ0 Ah gh

∂φ kh


0
 w = ∂z = ∂z 0 1 = w · A √



h h gh

3. Laplace equation and the boundary conditions

After normalization, the governing equation is the Laplace equation:

µ2 (φxx + φyy ) + φzz = 0, −1 < z < η. (2)

On the free surface, the KFSBC (kinematic free surface boundary condition) applies:

µ2 (ηt + φx ηx + φy ηy ) = φz , z = η. (3)

In addition, the DFSBC (dynamic free surface boundary condition) also applies on the
free surface:
 2 2 
µ2 (φt + η) + µ (φx + φ2y ) + φ2z = 0, z = η. (4)
2
Lastly, the no-penetration boundary condition is applied on the horizontal bottom:

φz = 0, z = −1. (5)

冪級 數 展 開 )
4. Power series expansion (冪
If we assume that the solution φ is an analytic function (解析函數), which means that its
Taylor series about any point in its domain converges to the function near that point, then
we can express the solution φ as a power series:
∞
X
φ(x, y, z, t) = (z + 1)n φn (x, y, t). (6)
n=0

1
ESOE 5136 2021 April 26 - Handout Lo

Some useful expressions based on (6):


X∞
∇φ = (z + 1)n ∇φn (note: this is a vector)






 n=0
∞


 X
2
(z + 1)n ∇2 φn (note: this is a scalar)




 ∇ φ =
n=0
∞ ∞ . (7)
X X
n−1
(z + 1)n (n + 1)φn+1

φz = n(z + 1) φn =






 n=0 n=0


 X∞ X∞
n−1
(n + 1)(n + 2)(z + 1)n φn+2



 φ zz = n(n + 1)(z + 1) φ n+1 =
n=0 n=0

We recall that the “del” operator ∇ is defined as

n
X ∂
∇[· · · ] = êi [· · · ], (8)
∂xi
i=1

which produces a vector (向量). The Laplace operator ∇2 of some function f is defined as

∇2 f = ∇ · ∇f = fxx + fyy (in 2D space), (9)

which is a scalar (純量).

5. Equations expressed in terms of power series
Laplace equation
Substituting the power series (6) into the Laplace equation (2) gives
∞
X h i
µ2 ∇2 φ + φzz = (z + 1)n µ2 ∇2 φn + (n + 2)(n + 1)φn+2 = 0, −1 < z < η. (10)
n=0

Since (z + 1)n 6= 0 in general, for the above equation to hold, the terms in the square
brackets must equal zero:
h i −µ2 ∇2 φn
µ2 ∇2 φn + (n + 2)(n + 1)φn+2 = 0 ⇒ φn+2 = , n = 0, 1, 2, · · · . (11)
(n + 1)(n + 2)

No-penetration boundary condition

Substituting the power series (6) into the no-penetration equation (5) gives
∞
X
n(0)n−1 φn = 0 ⇒ φ1 = 0 since 00 = 1. (12)
n=0

Using (12) in (11) then gives



 φ1 = φ3 = φ5 = φ7 = · · · = 0,



−µ2 2 −µ2 2


∇ ∇ φ0

 φ2 =
 φ 0 =

 2·1 2!
. (13)
−µ2 2 µ4 2 2
∇ ∇ ∇ φ0


 φ 4 = φ 2 =



 4·3 4!

2 6
 φ = −µ ∇2 φ = −µ ∇2 ∇2 ∇2 φ



6 4 0
6·5 6!

2
ESOE 5136 2021 April 26 - Handout Lo

Since φ0 = O(φ) = O(1), (13) implies that

φ2 = O(µ2 ), φ4 = O(µ4 ), φ6 = O(µ6 ). (14)

Keeping up to O(µ4 ) terms – which means that anything O(µ6 ) or smaller can be omitted
– we can express the solution φ using the power series (6) as

µ2 µ4
φ(x, y, z, t) = φ0 (x, y, t)− (z+1)2 ∇2 φ0 (x, y, t)+ (z+1)4 ∇2 ∇2 φ0 (x, y, t)+O(µ6 ). (15)
2 24

KFSBC
Expressing φz up to O(µ4 ),

µ4
φz = −µ2 (z + 1)∇2 φ0 + (z + 1)3 ∇2 ∇2 φ0 + O(µ6 ), (16)
6
and defining the total depth as

H = 1 + η, ⇒ Hx = ηx , Hy = ηy , (17)

we can write the KFSBC (3) as

hH
t
i hH
t µ2 i
µ2 + ∇H · (∇φ|z=η ) = µ2 + ∇H · (∇φ0 − H 2 ∇2 ∇φ0 )
  2
µ4 3 2 2
= −µ2 H∇2 φ0 + H ∇ ∇ φ0 + O(µ6 ) . (18)
6
Ht µ2 µ2
⇒ + ∇H · (∇φ0 − H 2 ∇2 ∇φ0 ) = −H∇2 φ0 + H 3 ∇2 ∇2 φ0 + O(µ4 )
 2 6

DFSBC
Expressing ∇φ and ∇2 φ up to O(µ2 ) (because higher-order terms will not be needed),
2

 ∇φ = ∇φ0 − µ (z + 1)2 ∇2 ∇φ0 + O(µ4 )

2 , (19)
2 2 2 2 2 4

(∇φ) = (∇φ) · (∇φ) = (∇φ0 ) − (∇φ0 )µ (z + 1) ∇ ∇φ0 + O(µ )


we can write the DFSBC (4) as

" #
h µ 2 i 1 h i
µ2 φ0t − H 2 ∇2 φ0t + η + µ2 (∇φ0 )2 − µ2 H 2 (∇φ0 ) · ∇2 (∇φ0 )
2 2
1
+ µ4 H 2 (∇2 φ0 )2 = O(µ6 )
2
" # . (20)
µ2 2 2 1 h i
⇒ φ0t − H ∇ φ0t + η +  (∇φ0 )2 − µ2 H 2 (∇φ0 ) · ∇2 (∇φ0 )
2 2
1
+ µ2 H 2 (∇2 φ0 )2 = O(µ4 )
2

6. Equations expressed in terms of horizontal flow velocities

Since the velocities ~u can be calculated from the velocity potential φ as ~u = ∇φ, we have
the leading-order velocity solution as

~u0 = ∇φ0 . (21)

3
ESOE 5136 2021 April 26 - Handout Lo

We can then use (21) to rewrite (18) as

1 µ2 µ2
Ht + ∇H · (~u0 − H 2 ∇2 ~u0 ) + H∇ · ~u0 − H 3 ∇2 (∇ · ~u0 ) = O(µ4 ), (22)
 2 6
and rewrite (20) as
h µ2 i 1 h i 1
φ0t − H 2 ∇ · ~u0t + η +  (~u0 )2 − µ2 H 2 ~u0 · (∇2 ~u0 ) + µ2 H 2 (∇ · ~u0 )2 = O(µ4 )
2 2 2
.
1 h 1 1 1 i
= φ0t + (~u0 )2 + η + µ2 − H 2 ~u0 · (∇2 ~u0 ) + H 2 (∇ · ~u0 )2 − H 2 ∇ · ~u0t
2 2 2 2
(23)
Next, we take the gradient ∇ of (23) to obtain the vector equation (which corresponds to
two scalar equations):
∇H h 1 1 1 i
~u0t +~u0 ·∇~u0 + +µ2 ∇ − H 2 ~u0 ·(∇2 ~u0 )+ H 2 (∇·~u0 )2 − H 2 ∇·~u0t = O(µ4 ). (24)
 2 2 2
7. Equations expressed in terms of depth-averaged horizontal velocities
Accurate up to O(µ2 ), the velocity solutions are
2

 ~u = (u, v) = ∇φ = ~u0 − µ (z + 1)2 ∇(∇ · ~u0 ) + O(µ4 )

2 . (25)
2 2 2 4

w = φz = −µ (z + 1)∇ φ0 = µ (z + 1)∇ · ~u0 + O(µ )


Let’s consider the depth-averaged horizontal velocities:

Z η
1 η h µ2
Z
~u = 1
i
∇φdz = ~u0 − (z + 1)2 ∇(∇ · ~u0 ) + O(µ4 ) dz
H −1 H −1 2
µ2 2 2
= ~u0 − H ∇ ~u0 + O(µ4 ) . (26)
6
µ2 2 2 ~
⇒ ~u0 = ~u + H ∇ u + O(µ4 )
6
Substituting (26) into (22) gives
Ht + ∇ · (H ~u) = 0, (27)
which is the continuity equation expressed in terms of depth-averaged horizontal velocities.
(27) can be shown to be exact.
Substituting (26) into (24) gives
2 2
~ut +~u·∇~u+ ∇H + µ (H 2 ∇2~u)t +µ2 ∇ −  H 2~u·(∇2~u)+  H 2 (∇·~u)2 − H ∇·~ut = O(µ4 ),
h i
 6 3 2 2
(28)
which is the momentum equation expressed in terms of depth-averaged horizontal veloci-
ties.
Altogether, (27) and (28) are commonly referred to as the “fully nonlinear and weakly
dispersive wave model,” in constant water depth. The derivation process can be extended
to consider variable water depths, and the results can be found in Lynett and Liu (2002);
Mei et al. (2005).
We recall that in the derivation, the waves are assumed to be moderately long (and hence
weakly dispersive) so that O(µ4 )  1 is negligibly small. On the other hand, no as-
sumptions on wave nonlinearity are made. Therefore, the waves are allowed to be highly
nonlinear; i.e., O() = 1.

4
ESOE 5136 2021 April 26 - Handout Lo

8. Pressure equation
The pressure field, normalized by ρgh, in which ρ is the fluid density, can be calculated
from the normalized Bernoulli equation as
( ) (
h 1 i h µ2 i
−P = z +  φt + (∇φ)2 + 2 φ2z = z +  φ0t − (z + 1)2 ∇ · ~u0t
2 µ 2
) . (29)
h 2 i
+ ~u0 − µ2 (z + 1)2 ~u0 · ∇2 ~u0 + µ2 (z + 1)2 (∇ · ~u0 )2 + O(µ4 )
2

(23) provides means to rewrite φ0t in (29), which then becomes

( )
µ2 h 2 2
i
2 2
+ O(µ4 ). (30)
 
P = (η − z) − H − (z + 1) ∇ · ~u0t +  ~u0 · (∇ ~u0 ) − (∇ · ~u0 )
2

It is worth pointing out that the hydrostatic pressure in (30) is O(1). The leading-order
dynamic pressure due to water waves is O().
布新 內 斯 克 方 程 )
9. Boussinesq equations (布
When written out fully in scalar form, the fully nonlinear and weakly dispersive wave
model is really lengthy and not easy to work with. A more practical but slightly less
accurate version is the Boussinesq equations.
The Boussinesq theory assumes that O() = O(µ2 ) < 1, which means that the waves are
weakly nonlinear and weakly dispersive – O() and O(µ2 ) are equally small but not com-
pletely negligible. Therefore, terms of O(, µ2 ) need to be kept, but terms of O(2 , µ2 , µ4 )
can be neglected.
Discarding terms of O(µ2 , µ4 ) in (28), and replacing H with 1 + η in (27) and (28) results
in  h i
 t
 η + ∇ · (1 + η)~
u =0


µ2 , (31)
~ ~ ~ ~ 2 4

 t
 u + u · ∇ u + ∇η − ∇(∇ · u t ) = O(µ , µ )
3


which are called the Boussinesq equations in constant water depth. Again, the full
version applicable in variable water depths can be found in Lynett and Liu (2002); Mei
et al. (2005).
The corresponding pressure field for the Boussinesq equations is
µ2 2
P = η − z + (z + 2z)∇ · ~ut + O(2 µ2 , µ4 ), (32)
2
in which the fact that
~u0 = ~u + O(µ2 ) (33)
is utilized. From (32) we see that the pressure field is no longer hydrostatic in the Boussi-
nesq theory.

References
Lynett, P. and P. L.-F. Liu (2002), “A numerical study of submarine-landslide-generated waves
and run-up.” Proc. R. Soc. Lond., A 458, 2885–2910.
Mei, C., M. Stiassnie, and D. K.-P. Yue (2005), Theory and Applications of Ocean Surface
Waves. World Scientific.