ISSN 1810-2328, Journal of Engineering Thermophysics, 2011, Vol. 20, No. 1, pp. 13–33.


c Pleiades Publishing, Ltd., 2011.

Dynamics of Local Density Perturbation in Stably Stratified Fluid:

Results of Numerical Experiments
G. G. Chernykh1, 2, 3* and A. N. Zudin1, 2**
1
Institute of Computational Technologies, Siberian Branch, Russian Academy of Sciences,
pr. Akad. Lavrent’eva 6, Novosibirsk, 630090 Russia
2
Novosibirsk State University, ul. Pirogova 2, Novosibirsk, 630090 Russia
3
Siberian State University of Telecommunications and Informatics,
ul. Kirova 86, Novosibirsk, 630102 Russia
Received May 30, 2010

Abstract—Linear and nonlinear numerical models of dynamics of local density perturbation in a

stably stratified medium are constructed. The influence of viscosity on the process of generation
and propagation of internal waves generated by the local density perturbation in a pycnocline is
evaluated. The problem on the dynamics of local density perturbation in the presence of wave
background is considered.

DOI: 10.1134/S1810232811010024

INTRODUCTION

We consider a plane flow generated by a local density perturbation in a stratified medium. Interest in
investigating the dynamics of localized regions of a mixed fluid in a stably stratified medium is concerned
with studying some geophysical effects and with solving practical problems [1–6]. The flow generated
by the local density perturbation is also of interest for studying internal waves induced by wakes in
a stably stratified medium [7–10]. As a rule, laboratory and natural experimental investigations are
expensive, and successful only in the simplest situations. Asymptotic methods have become widely used
in studying the local perturbation dynamics. However, some difficulties associated with nonlinearity
of problem statements in a number of situations of practical importance are also encountered here.
The domain of applicability of asymptotic solutions requires additional analysis and also attraction
of experimental data and results on numerical analysis, which are based on applying complete and
linearized equations of hydrodynamics. Numerical simulation plays an important role in studying the
class of flows under consideration (and many other physical phenomena). A detailed bibliography and
analysis of works concerned with this problem are represented in [3, 11–24]. Analysis of the known
works on numerical simulation of a flow shows that the results on numerical investigation of the local
density perturbation dynamics in a stably stratified fluid [15, 17–19, 24] are not comprehensive. There
are no results on detailed numerical analysis of applicability of linear and nonlinear mathematical models
to calculation of the flow characteristics. In the present paper, we present numerical results that illustrate
evolution of the region of partially mixed fluid in a stably stratified fluid. We also perform numerical
experiments on assessment of applicability of mathematical models and consider the problem on the
local density perturbation dynamics in a nonlinearly stratified fluid in the presence of a wave background.
Our investigations extend the works that were commenced in [15, 17–19, 24].
*
E-mail: [email protected]
**
E-mail: [email protected]

13
14 CHERNYKH, ZUDIN

1. PROBLEM STATEMENT WITH APPLYING EULER EQUATIONS IN THE

OBERBECK–BOUSSINESQ APPROXIMATION
Euler equations in the Oberbeck–Boussinesq approximation are one of the main mathematical
models of the nonstationary stratified flow dynamics. Using the stream function–vorticity variables, the
Euler equations can be written as:
∂ω ∂ω ∂ω g ∂ρ1
+u +v = , (1)
∂t ∂x ∂y ρ0 ∂x

∂2ψ ∂2ψ
+ = ω, (2)
∂x2 ∂y 2

∂ρ1 ∂ρ1 ∂ρ1 dρs

+u +v +v = 0. (3)
∂t ∂x ∂y dy
Herein, x, y are the Cartesian coordinate system, the y axis is directed vertically upward, opposite
to the force of gravity; ρ1 = ρ − ρs , ρs = ρs (y) is the density of the unperturbed medium outside the
mixing zone; ρ0 = ρs (0); u and v are the horizontal and vertical velocity components of the fluid:
u = ∂ψ/∂y, v = −∂ψ/∂x; g is the gravity acceleration. The medium stratification is assumed stable,
i.e., (1/ρ0 )dρs /dy ≤ 0.
The initial and boundary conditions in the problem under consideration are:
ψ = ω = ρ1 = 0, x2 + y 2 → ∞, (4)
 
ρ0 − ρs (y)  −(1.25r)8
ρ = ρ (x, y) = ρ0 1 −
0
1 − εe ,
ρ0

−∞ < x < ∞, −∞ < y < ∞, t = 0, (5)

ψ = ω = 0, −∞ < x < ∞, −∞ < y < ∞, t = 0. (6)

Here, ε = const = 0.75. Condition (6) corresponds to u(0, x, y) = 0, v(0, x, y) = 0.
The above-mentioned initial density distribution corresponds to incomplete fluid mixing (it is known
[11] that the evolution of turbulent spots in a stratified fluid in the turbulent mixing zone is accompanied
by incomplete fluid mixing).
In numerical solving the problem, the system of Eqs. (1)–(3) was preliminary reduced to dimen-
sionless form √ via using the characteristic scales of length R (the mixing zone radius) and the time
scale T∗ = 1/ ag (a = −(1/ρ0 )dρs /dy at y = 0); using additionally ρ1 = ρ0 aRρ̃1 . For analyzing the
calculation results, it is convenient to introduce the time scale T = 2πT∗ , that is, the Vaisala–Brunt
period. Hereafter, we omit the sign for dimensionless variables, (˜), if possible.
Numerical integration of the system of Eqs. (1)–(3) is performed using an approach based on Euler–
Lagrange variables [14, 15, 17, 19]:
t = t, x = ξ = x, y  = η, dη/dt = 0, (7)

η(0, x, y) = η 0 (x, y), −∞ < x < ∞, −∞ < y < ∞, (8)

η(t, x, y) → q1 (y), x2 + y 2 → ∞, t > 0.

Meanwhile, Eqs. (1)–(3) are written as:
∂ρ 1 ∂ψ ∂ρ

+ = 0, (9)
∂t J ∂η ∂ξ

∂y ∂ψ

=− , (10)
∂t ∂ξ

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 15
   
2
1 ∂ ∂ψ ∂ 1 ∂y ∂ψ ∂ ∂y ∂ψ ∂ ∂y ∂ψ
J + +1 − + = ω, (11)
J ∂ξ ∂ξ ∂η J ∂ξ ∂η ∂ξ ∂ξ ∂η ∂η ∂ξ ∂ξ

∂ω 1 ∂ψ ∂ω ∂ρ 1 ∂ρ ∂y

+ = − , (12)
∂t J ∂η ∂ξ ∂ξ J ∂η ∂ξ

∂(t, x, y) ∂y
J= = .
∂(t , ξ, η) ∂η
In the boundary and initial conditions (4), (5), the functions ρs (y) and ρ0 (x, y) are replaced by ρ̂s (η)
and ρ̂0 (ξ, η), which are obtained from ρs (y) and ρ0 (x, y) by substituting the values y = q2 (ξ, η) that are
the solution to the equation η(0, x, y) = η 0 (x, y) (η 0 (x, y) is assumed a monotone function of y at fixed
x). Conditions (4)–(6) contain the initial condition
y(0, ξ, η) = q2 (ξ, η) (13)
and also the boundary condition
y = q3 (η), ξ 2 + η 2 → ∞, t > 0. (14)
As the coordinate lines η = const in the case under consideration we choose lines ρ = const. With
this choice of the coordinate lines, Eq. (9) is fulfilled identically because ρ = ρ(η). As a result of transition
to the Euler–Lagrange coordinate system, in Eq. (12) there is no convective transfer in the vertical
direction. This makes it possible to prevent the corresponding errors in finite-different approximation.
The finite-difference algorithm based on the concept of the predictor-corrector method, and its testing
are described in detail in [14, 17, 19].

2. INTERNAL WAVES GENERATED BY LOCAL DENSITY PERTUBATION IN FLUID

WITH NONLINEAR DENSITY DISTRIBUTION IN DEPTH

To ascertain the influence of nonlinearity of unperturbed medium stratification on the flow generated
under collapse of the mixed fluid zone, we performed some calculations for distributions ρs (y). Euler
equations in the Oberbeck–Boussinesq approximation (1)–(3) were applied. The initial density distri-
bution was specified in the form of Eq. (5).
Figure 1 presents calculation results that illustrate the pattern of the flow for a zero initial distribution
of vortex and stream function, and for ρs (y) specified as:
ρ̃s (ỹ) = ρ̃0 − ỹ, (15)

ρ̃s (ỹ) = ρ̃0 − β̃ · tanh(ỹ/β̃) β̃ = β/R = 1.0, 0.5, 0.1, (16)

⎧
⎨ ρ̃ − ỹ, −2.5 ≤ ỹ ≤ 1.1,
0
ρ̃s (ỹ) = (17)
⎩ ρ̃ − (tanh((ỹ − 2)/0.1) + tanh(9) + 1.1), 1.1 ≤ ỹ ≤ 3.
0

For the distribution ρs (y) defined by (15), the process of generation and propagation of internal waves is
characterized by formation of a system of vortices in each quadrant of the plane xy [11, 12]. The number
of vortices grows with time. In particular, for variant (15) with t/T = 2, there are four such vortices in
each quadrant; for variants (16) with β̃ = 1, 0.5, 0.1, there are three, two, and one vortex, respectively.
The foresaid is illustrated in Fig. 1, where the stream lines ψ̃ = ψT∗ /R2 = const are depicted at the time
t/T = 2. Hereinafter, the stream lines are represented by the following levels:
1, −0.0540; 2, −0.0420; 3, −0.0300; 4, −0.0180; 5, −0.0060; 6, −0.0000; 7, −0.0060; 8, −0.0180;
9, −0.0300; 10, −0.0420; 11, −0.0540.
Figures 1b–1d correspond to the distribution (16), β = 1, 0.5, 0.1. In variants with β = 1, 0.5, the
coordinate lines η = const coincided with the lines ρ = const with a variable interval specified as follows.

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

16 CHERNYKH, ZUDIN

Fig. 1. Stream lines obtained with the Oberbeck–Boussinesq approach for different profiles of density ρs (y), t/T = 2.

The fist step was construction of a sequence of values yj = yj−1 + h0 q j (y0 = 0, yj > 0, h0 /R = 0.064,
q = 1.035, j = 1, . . . , 25); then lines of equal density ρ = ρs (y) were found and accepted as coordinate
lines at the initial time. For negative values of y, the lines η = const were constructed similarly (it
is appropriate to mention here that in the new Euler–Lagrange coordinate system the grid in both
directions was chosen uniform). In the case of β̃ = 0.1, the lines η = const at the initial time were
constructed as follows. First we found a line η = η∗ such that (for y ≥ 0) the value of the density ρ̃(η)
with η > η∗ differed from the constant one by no more than 0.1 × 10−9 . In the region with a variable
density 0 ≤ η ≤ η∗ , as the lines η = const for t/T = 0 we chose lines ρ/aRρ0 = const with the interval
δρj = ρj − ρj−1 (ρj = ρs (yj ), yj = yj−1 + hj , y0 = 0, hj = 0.074, j = 1, . . . , 10). In the region η > η∗ ,
for each xi = i × h1 , i = 1, . . . , 71, the coordinate lines for t/T = 0 were constructed according to the

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 17

formula
η(0, xi , yi,j ) = ηj , j = 11, . . . , 17,

yi,j = yj (xi ), yi,j = yi,j−1 + h∗ (xi )q (j−10) ,

h∗ (xi ) = (2.5 − yi,10 )(1 − q)/(1 − q 7 ), yi,10 = y(0, xi , η∗ ), q = 1.1.

For the density distribution (16), β̃ = 0.1, we also calculated a variant on a grid with the number
of nodes in the vertical direction twice as greater than that for the considered variant. The obtained
maximum values of the stream function differed by no more than 7%.
If we enter the quantity H that is determined from the relationship
ρs (H) − ρ0 = 0.99(B − ρ0 ), B = lim ρs (y),
y→∞

then for β̃ = 1, 0.5, 0.1 the value of H̃ = H/R is 2.64, 1.32, 0.26, respectively. It is seen that the transition
from the distribution (15) to (16) leads to substantially varying wave pattern that is generated under
collapse of the local density perturbation in a pycnocline: with small values of H, the flow is characterized
only by one vortex in each quadrant of the plane xy (this fact was fairly well known to the authors by the
time of writing [15]; however, detailed numerical experiments based on the Euler–Lagrange approach
are represented in the present work).
The form and intensity of the vortex (consequently, the form and intensity of the wave packet) remain
unchanged up to quite great time values. For the purpose of detailed analysis of the wave pattern, in
the latter case the calculations were also performed in a 200R × 30R region on a 2001 × 301 grid.
Figure 2 shows stream lines ψ̃ = const for β̃ = 0.1, ε = 0.75, t/T = 15 (a), and t/T = 108 (b). The
corresponding distribution ρ = const = ρs (0.0716), which characterizes the internal wave, is depicted
in Fig. 3. It is seen that the intensity and form of the main perturbation do not practically change; this
evidences high reliability and efficiency of the numerical model.

3. THE INFLUENCE OF VISCOSITY ON THE DYNAMICS OF LOCAL DENSITY

PERTURBATION IN A PYCNOCLINE

It is of interest to investigate the influence of viscosity on the characteristics of internal waves

generated by the local density perturbation in a pycnocline. The calculations were based on Eqs. (9)–
(12). Meanwhile, Eq. (12) was replaced by the following:
 
∂ω 1 ∂ψ ∂ω ∂ρ 1 ∂ρ ∂y 1 1 ∂ ∂ω ∂ 1 ∂y 2 ∂ω

+ = − + J + +1
∂t J ∂η ∂ξ ∂ξ J ∂η ∂ξ Re J ∂ξ ∂ξ ∂η J ∂ξ ∂η


∂ ∂y ∂ω ∂ ∂y ∂ω
− + . (18)
∂ξ ∂ξ ∂η ∂η ∂ξ ∂ξ

The structure of the numerical algorithm remained the same (as in [14]). At the stage of predictor, the
viscous terms were allowed for according to the scheme of stabilizing correction [25].
Some calculation results corresponding to the initial density distribution (5), the distribution ρs (y)
of the form (16), β̃ = 0.1, are represented in Figs. 4–7. The computational grid was the same as that
in Figs. 2 and 3. Figures 4a–4c show stream lines ψ̃ = ψT∗ /R2 = const, which correspond to the time
t/T = t/2πT∗ = 8 for an ideal fluid (Fig. 4a) and a viscous fluid with Reynolds numbers Re = R2 /T∗ ν
equal to 104 (Fig. 4b) and 103 (Fig. 4c). Herein, ν is the kinematic coefficient of viscosity. The stream
lines are represented by levels: 6, 0.0000, 7, 0.0060, 8, 0.0180, 9, 0.0300, and 10, 0.0420. The sign ×
in Figs. 4a–4c shows position of markers placed at the initial time into the mixing region. It is seen
that for the ideal fluid the horizontal size of the mixing zone and the vortex intensity are maximal; the

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

18 CHERNYKH, ZUDIN

Fig. 2. Preserving vortex form in ideal fluid for different times: (a) t/T = 15; (b) t/T = 108.

flow patterns for Re−1 = 0 and Re = 104 differ slightly at t/T = 8. Substantial viscosity is illustrated in
Fig. 4c.
In addition to Fig. 4, Fig. 5 represents lines ρ̃ = const = ρ̃s (0.0717) and ρ̃ = ρ/aRρ0 . Solid, dashed,
and dash-dotted curves correspond to Re−1 = 0, Re = 104 , and Re = 103 , respectively. Results in Fig. 5
agree with those in Fig. 4.
The wave pattern is represented in more detail in Figs. 6a–6c as isolines ρ̃ = ρ̃s (ỹ0 ) and ỹ0 = y0 /R =
0.0717 for different times. Curves 2, 4, 6, 8, 10, 12, and 14 correspond to t/T = 2, 4, 6, 8, 10, 12, and 14.
Figures 6a–6c were obtained for Re−1 = 0, Re = 104 , and Re = 103 , respectively.
The data on variations of the maximal internal wave amplitude as a function of time and Reynolds
number are also listed in the table. Columns I–III contain dimensionless maximal values of internal
wave amplitudes for Re−1 = 0, Re = 104 , and Re = 103 . It is seen that in the ideal fluid, for t/T ≥ 4,
the maximal amplitude of internal waves is almost constant. As in Figs. 5 and 6, the data from the table
demonstrate the influence of viscosity on the generated internal waves.
The time variations of the abscissa of the maximum stream function modulus as a function of
Reynolds number are represented in Fig. 7a; variations of the horizontal size of the mixing zone are
plotted in Fig. 7b. For small time values, the plots of functions depicted in Fig. 7 are similar; at t/T ≥ 6,

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 19

Fig. 3. Preserving wave form in ideal fluid for different times: (a) t/T = 15; (b) t/T = 108.

we see a lag of the horizontal size of the mixing zone (Fig. 7b) as a result of incomplete fluid mixing in it
at the initial time. At larger time values, the horizontal size stops to grow; in an ideal fluid corresponding
to the principal solitary wave, the convective vortex goes on moving with a constant velocity.
We will dwell on accuracy control of calculations for Re = 104 . The calculations were performed in
the region 0 ≤ x/R ≤ 50, 0 ≤ y/R ≤ 30. It was obtained, in particular, that for the times t/T = 4, 6, and
8, the maximal value of the stream function ψm on a 501 × 301 grid was 0.458 × 10−1 , 0.434 × 10−1 , and
0.409 × 10−1 ; on a 1001 × 601 grid it was 0.423 × 10−1 , 0.414 × 10−1 , and 0.391 × 10−1 , respectively.
The deviation of the arrays of grid values of the functions y, ψ, and ω in a uniform norm does not exceed
deviations of ψm .
We will note that in the case of an ideal fluid, the displacement velocity of the first (primary) ridge of
internal waves in the direction of the x axis for t/T ≥ 4 is ≈ 0.150R/T∗ . This velocity was calculated
as a velocity of changing abscissa of the maximum stream function modulus (Fig. 7). If we use the
asymptotic relationship (5.36)–(5.37) borrowed from [26]
 
 2 β̃ 2 3 Ã
(1)
c̃ = 1+ , (19)
2 5 β̃

then for the maximal amplitude of the waves, Ã = A/R ∈ [0.32, 0.33], the wave displacement velocity
c̃(1) = c(1) T∗ /R ∈ [0.144, 0.145]. The quantity c̃(1) is calculated for the first (primary) wedge of a wave
that is similar to that shown by a solid curve in Fig. 5. Comparing the calculated data with results
obtained in [26] shows that that they are in satisfactory agreement. In the case of viscous fluid, by analogy
with [27] (where the author considered solitary internal waves at the interface of two fluids of different
density), the internal wave propagation velocity c̃(1) for the calculated maximal amplitude A = A(t) was

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

20 CHERNYKH, ZUDIN

Fig. 4. Correlation of stream lines for ideal (a) and viscous fluid: Re = 104 (b), Re = 103 (c); t/T = 8. Particles that
were in the region of mixed fluid at the initial time are marked.

determined both from direct processing of the calculated data and from (19). It has been obtained, in
particular, that for Re = 104 , Re = 103 , and t/T = 10, the direct processing of the calculated data yields
c̃(1) = 0.138 and c̃(1) = 0.0953, and relationship (19) yields c̃(1) = 0.136 and c̃(1) = 0.102. For other
values of t/T , the results are similar.
Along with the “narrow” pycnocline (16), β̃ = 0.1, we considered the “wide” pycnocline (16), β̃ =
0.5. The calculated data in the form of isolines ρ̃ = ρ̃s (0.477) are given for the time t/T = 8 in Fig. 8.
Solid, dashed, and dash-dotted curves correspond to Re−1 = 0, Re = 104 , and Re = 103 , respectively.

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 21

Fig. 5. Lines of equal density ρ̃ = const = ρ̃s (0.0717) for ideal (solid curve) and viscous fluid: Re = 104 (dashed curve),
Re = 103 (dash-dotted curve); t/T = 8.

Maximal amplitude of internal waves versus time and Reynolds number

t/T I II III
2.0 0.3025 0.3017 0.2987
4.0 0.3239 0.3096 0.2358
6.0 0.3277 0.2977 0.1840
8.0 0.3278 0.2869 0.1439
10.0 0.3279 0.2755 0.1124
20.0 0.3286 0.2210 0.0491
30.0 0.3291 0.1738 0.0478
40.0 0.3298 0.1323 0.0458
50.0 0.3306 0.0989 0.0439
60.0 0.3310 0.0720 0.0420
70.0 0.3287 0.0514 0.0407

The main impact of viscosity manifests itself in “absorption” of the short-wave section of the line
ρ̃ = const. The calculations were performed in the first quadrant of the plane xy in a 30R × 5R rectangle
with 300 × 50 cells. Figure 8 represents only part of the computation domain.

4. LINEAR NUMERICAL MODELS OF INTERNAL WAVES

Since the linear model of internal waves was widely used in studying the wave fluid motions [2, 5, 16,
28], it is of interest to perform numerical modeling of a flow generated by local density perturbation in a
stratified medium for distributions ρs (y) of the form (15), (16), based on Euler linear equations [5]. Using
independent variables x, y, t, and dependent ψ, ω, ρ1 = ρ − ρs (y), the linear model of wave motions in a
stratified medium can be written in a standard manner (in a dimensionless form):
∂ω ∂ρ1
= , (20)
∂t ∂x
∂ρ1 ∂ψ dρs
= , (21)
∂t ∂x dy

∂2ψ ∂2ψ
+ = ω. (22)
∂x2 ∂y 2

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

22 CHERNYKH, ZUDIN

Fig. 6. Correlation of lines of equal density ρ̃ = const = ρ̃s (0.0717) for ideal (a) and viscous fluid: Re = 104 (b),
Re = 103 (c); t/T = 2, 4, . . . , 14.

The system of Eqs. (20)–(22) is supplemented with initial and boundary conditions
ω(0, x, y) = 0, ψ(0, x, y) = 0, ρ1 (0, x, y) = ρ0 (x, y) − ρs (y), (23)

ω(t, x, y) = ψ(t, x, y) = ρ1 (t, x, y) = 0, r 2 = x2 + y 2 → ∞, t ≥ 0. (24)

The corresponding symmetry conditions were set on the axes of symmetry.
The problem (20)–(24) was solved using the simplest finite-difference algorithm of the predictor-
corrector type:
n+1/2
ωi,j − ωi,j
n ρn1i+1,j − ρn1i−1,j
= , (25)
Δt/2 2h1

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 23

Fig. 7. Time variation of the abscissa of the maximum stream function (a) and horizontal size of the mixed region (b)
for ideal (solid curve) and viscous fluid: Re = 104 (dashed curve), Re = 103 (dash-dotted curve).

Fig. 8. The wide pycnocline—the density distribution (16), β̃ = 0.5. Isolines ρ̃ = ρ̃s (0.477) are presented for the time
t/T = 8. Solid, dashed, and dash-dotted curves correspond to Re−1 = 0, Re = 104 , and Re = 103 , respectively.

n+1/2
ρ1i,j − ρn1i,j n
ψi+1,j − ψi−1,j
n
dρs
= , (26)
Δt/2 2h1 dy i,j

n+1/2 n+1/2
(Δh ψ)i,j = ωi,j , (27)
n+1/2 n+1/2
n+1
ωi,j − ωi,j
n ρ1i+1,j − ρ1i−1,j
= , (28)
Δt 2h1
n+1/2 n+1/2
1i,j − ρ1i,j
ρn+1 ψi+1,j − ψi−1,j
n
dρs
= , (29)
Δt 2h1 dy i,j

(Δh ψ)n+1 n+1

i,j = ωi,j . (30)

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

24 CHERNYKH, ZUDIN

In Eqs. (25)–(30), Δh ψ is the finite-difference analog of the Laplace operator, applied to the function
ψ; the index (n + 1/2) [(n + 1)] means that the corresponding quantities refer to time moment (n +
1/2)Δt [(n + 1)Δt]. We used finite-difference grids, nonuniform along the applicate. Sizes of the grid
domains with the use of the linear model are similar to those in Fig. 1. The nonuniform distribution of
grid nodes along the y axis coincided with the distribution of values of the function y for x/R = 7 in
calculations of variants for the density distribution (15), (16) by the Oberbeck–Boussinesq model. The
algorithm was preliminary tested on a problem having the exact analytical solution:
ψ(t, x, y) = c(t) · sin ax · sin by,

ω(t, x, y) = −c(t)(a2 + b2 ) · sin ax · sin by,

ρ1 (t, x, y) = d(t) · cos ax · sin by,

l2 = a2 /(a2 + b2 ), c(t) = (1/l) sin lt, d(t) = (a/l2 ) cos lt,

0 ≤ t ≤ 1, a = b = 1, 0 ≤ x ≤ π, 0 ≤ y ≤ π. (31)

Calculations on the sequence of dense uniform grids have shown convergence of grid solutions to the
exact one with the order O(Δt + h2 ) at norm that is a grid analog of the norm of space of continuous
functions (Δt is the value of step in time variable; h = const is the grid step along the axes x and y).
Results of calculating the variants of the linear density distribution of unperturbed fluid of the
form (15):
ρ̃s (ỹ) = ρ̃0 − ỹ

and the nonlinear density distribution of the form (16):

ρ̃s (ỹ) = ρ̃0 − β̃ · tanh(ỹ/β̃), β̃ = 1, 0.5, 0.1,

by the model (20)–(22) are represented in Figs. 9a–9d. Figure 9a corresponds to the linear density
distribution of unperturbed fluid (15); Figs. 9b–9d correspond to distributions (16), β̃ = 1, 0.5, 0.1.
Comparing Fig. 1a and Fig. 9a shows similarity of wave patterns in the case of linear stratification
when using nonlinear and linear models (in the case of regions with completely mixed fluid, this fact
was noted, in particular, in [29]). The wave patterns differ only in the neighborhood of local density
perturbation. The patterns differ much more with decreasing β̃ in distribution (16). For β̃ = 0.1 (Fig. 9d),
the linear model yields a physically implausible solution: convective vortices do not displace with time.
Results of calculating the variant with β̃ = 0.1 by the linear model are represented in more detail in
Figs. 10a and 10b. The calculations were done in the domain 0 ≤ x/R ≤ 28, 0 ≤ y/R ≤ 10, hx = 0.1,
hy = 0.05, Δt = 0.01T (the figure shows part of the calculation domain). It is seen that as the time
grows, the center of the convective vortex remains immobile (contrary to the calculations in the nonlinear
statement, see Figs. 2 and 7). A similar result was also obtained on a more detailed grid (Fig. 10c, hx =
0.05, hy = 0.025, and Δt = 0.0025T ). Since the obtained result is nontrivial, we performed calculation
based on an algorithm like (25)–(27) (without using the corrector stage). The obtained results were
similar to those in Figs. 10b and 10c. They are represented in Fig. 10d. Let us try to comment the
calculation results (Fig. 10) via analyzing the energy conservation law.
From the Euler linear equations, assuming sufficiently rapidly decreasing integrands as R → ∞, we
can obtain the following relationship (the sign for dimensionless variables is omitted; Ω : −∞ < x < ∞,
−∞ < y < ∞):
 
∂ u2 + v 2
+ yρ1 dxdy = − ρ1 vdxdy. (32)
∂t 2
Ω Ω

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 25

Fig. 9. Stream lines calculated using the linear models for stratification of the form (15), (16); t/T = 2.

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

26 CHERNYKH, ZUDIN

Fig. 10. The linear model, unperturbed fluid density distribution (16), β̃ = 0.1, hx = 0.10, hy = 0.05, Δt = 0.01T ;
(a) t/T = 3; (b) t/T = 5; (c) hx = 0.05, hy = 0.025, Δt = 0.0025T , t/T = 5; (d) scheme (25)–(27); t/T = 5.

In the case of linear stratification (or, generally speaking, nondegenerate stratification, i.e., −(1/ρ0 ) ·
(dρs /dy) = a(y) > 0), for the linear model the energy conservation law [30]:

∂ u2 + v 2 ρ21
+ dxdy = 0. (33)
∂t 2 2a(y)
Ω

In the nonlinear model written in the Oberbeck–Boussinesq approach, the energy conservation law
looks as (a(y) ≥ 0):

∂ u2 + v 2
+ yρ1 dxdy = 0. (34)
∂t 2
Ω

Relationship (32) contains the right-hand side stipulated by linearization of equations. It can lead to
“generation” of energy in a pycnocline (contrary to the case of linear stratification, where the right-hand
side (32) is small due to specificity of the flow generated by local perturbation). For analyzing the fact of
“generation” in the linear model, relationship (32) was rewritten as:
⎛ ⎞
t   
u2 + v 2
E(t) = E(0) + Π(t), Π(t) = − ⎝ ρ1 vdxdy ⎠ dt, E(t) = + yρ1 dxdy.
2
0 Ω Ω

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 27

In the case of nonlinear stratification with β̃ = 0.1, for the times t/T = 0, 0.5, 2.0, and 4.0, the E(t) took
on values 0.202 · 10−1 , 0.232 · 10−1 , 0.589 · 10−1 , and 0.132 · 100 , respectively. These values agree with
the above-mentioned assumption on energy “generation” in a pycnocline.
Along with the mathematical model (20)–(22) and its finite-difference analog (25)–(30), as the linear
model we also used an equation being the consequence of (20)–(22):
∂2 dρs ∂ 2 ψ
Δψ = . (35)
∂t2 dy ∂x2
In the case of linear stratification, this equation transforms into the known Sobolev equation [31]. The
initial and boundary conditions were the following equations compatible with (20)–(24):
∂Δψ ∂ρ1
Δψ = 0, = , t = 0, (36)
∂t ∂x

ψ = 0, r 2 = x2 + y 2 → ∞, t ≥ 0. (37)
From (36), find ψ and ∂ψ/∂t for t = 0. We used the simplest finite-difference approximation of Eq. (35):
 h n+1  n  n−1
Δ ψ i,j − 2 Δh ψ i,j + Δh ψ i,j dρs n
ψi+1,j − 2ψi,j
n + ψn
i−1,j
= · 2 . (38)
Δt2 dy j h1

The quantity (Δh ψ)1 was determined from the second relationship (36):
0 0
∂ρ1 Δt2 ∂ 2 ψ dρs
(Δ h
ψ)1i,j = (Δ h
ψ)0i,j + Δt + · . (39)
∂x i,j 2 ∂x2 dy i,j

The values of (ψ)n+1

i,j were found using an iterative scheme of stabilizing correction. The numerical
experimental results based on comparing with the exact analytical solution (31) have demonstrated
convergence (on a sequence of grids) of grid solutions to the exact one with the order O(Δt2 +
h2 ) corresponding to the order of approximation (38). Results of calculating the variants for density
distributions (15), (16) by Eq. (35) are represented in Figs. 9e–9i. They are practically identical to those
in Figs. 9a–9d for the linear model (20)–(22).
Of course, using the linear model of internal waves, one always stipulates its domain of applicability.
The presented calculated data are only an illustration of using the linear model. It is hardly probable
that its applicability to the problem on the local density perturbation dynamics in a pycnocline will be
successful.

5. COMPLETE EULER EQUATIONS

In the Cartesian coordinate system with the use of the stream function ψ and vorticity ω̂ (u = ∂ψ/∂y,
v = −∂ψ/∂x, ω̂ = ∂/∂y (ρu) − ∂/∂x (ρv)), the complete Euler equations take the form [32, 33]:
∂ρ ∂ρ ∂ρ
+u +v = 0, (40)
∂t ∂x ∂y

∂ ω̂ ∂ ω̂ ∂ ω̂ ∂ρ ∂ρ ∂ u2 + v 2 ∂ρ ∂ u2 + v 2
+u +v =g − · + · , (41)
∂t ∂x ∂y ∂x ∂x ∂y 2 ∂y ∂x 2

∂ ∂ψ ∂ ∂ψ
ρ + ρ = ω̂. (42)
∂x ∂x ∂y ∂y
The initial and boundary conditions for the problem with the use of complete Euler equations are similar
to the corresponding conditions for the Oberbeck–Boussinesq approach (4)–(6):
ψ = ω̂ = 0, ρ = ρs (y), x2 + y 2 → ∞, (43)

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

28 CHERNYKH, ZUDIN

ρ = ρ0 (x, y), ρ0 (x, y) = ρs (y), x, y ∈ A, t = 0, (44)

ψ = ω̂ = 0, −∞ < x < ∞, −∞ < y < ∞, t = 0. (45)

As in the foregoing,
√ the variables are made dimensionless using the length scale R and the time
scale T∗ = 1/ ag, (a = −(1/ρ0 )dρs /dy, y = 0). Moreover, we use the representation ρ = ρ0 aRρ̃; the
quantity g in Eq. (41) is replaced by g̃ = 1/aR. Note that ρ̃s (0) = 1/aR.
The system of equations was numerically solved using Euler–Lagrange approach similar to the one
applied in Section 1. Then Eqs. (40)–(42) are written as:
∂ρ 1 ∂ψ ∂ρ

+ = 0, (46)
∂t J ∂η ∂ξ

∂y ∂ψ

=− , (47)
∂t ∂ξ

∂ ω̂ 1 ∂ψ ∂ ω̂ g̃ ∂y ∂ρ ∂y ∂ρ 1 ∂y ∂ρ ∂y ∂ρ ∂ u2 + v 2

+ = − − −
∂t J ∂η ∂ξ J ∂η ∂ξ ∂ξ ∂η J2 ∂η ∂ξ ∂ξ ∂η ∂η 2

1 ∂ρ ∂y ∂ u2 + v 2 ∂y ∂ u2 + v 2
+ 2 − , (48)
J ∂η ∂η ∂ξ 2 ∂ξ ∂η 2
   
2 2
1 ∂ ρ ∂y ∂ψ ∂ ρ ∂y ∂ψ
+ +1
J ∂ξ J ∂η ∂ξ ∂η J ∂ξ ∂η

∂ ρ ∂y ∂y ∂ψ ∂ ρ ∂y ∂y ∂ψ
− + = ω̂. (49)
∂ξ J ∂η ∂ξ ∂η ∂η J ∂η ∂ξ ∂ξ
The finite-difference analog of the system of Eqs. (46)–(49) is similar to the algorithm given in [17].
In numerical integration of Eq. (49) with the use of the iterative scheme of stabilizing correction,
the experimentally chosen iteration parameter substantially differs from the Oberbeck–Boussinesq
approach model. The latter is caused by appearing in the equation the ρ that can vary over a wide range.
The dynamics of the local density perturbation with applying complete Euler equations was initially
calculated for the case of linear stratification. The calculations were done in the domain 0 ≤ x ≤ 40R,
−20R ≤ y ≤ 20R. Analyzing the calculation results corresponding to 1/aR = 50, we can conclude that
the far wave fields are substantially different. The Euler equations in the Oberbeck–Boussinesq approach
lead to a symmetrical wave pattern. Using the complete Euler equations with 1/aR = 50 substantially
disturbs the symmetry. However, detailed analysis of the near wave field (r ≤ 8R) has shown that the
complete Euler equations, Euler equations in the Oberbeck–Boussinesq approach, and the linear model
yield similar results. The main distinction of the calculations by the linear model manifests itself only in
the immediate vicinity of the origin of coordinates (r/R ≤ 2). The other two mathematical models yield
quite similar results. For a smaller value of 1/aR (e.g., for 1/aR = 5), the change in the density along
the vertical within the domain of integration becomes substantial.
This causes greater distinctions in calculations by the complete model and by the Oberbeck–
Boussinesq model. The aforesaid is illustrated in Fig. 11 that represents stream lines obtained by
complete Euler equations (Fig. 11a) and by Euler equations in the Oberbeck–Boussinesq approach
(Fig. 11b). One can see considerable distinctions in the pattern of stream lines even in the vicinity
of the origin of coordinates. The quantity 1/aR = 5 is sufficiently small, 1/aR ∼ 103 ÷ 104 in the real
ocean. However, one can observe rather arbitrary values of 1/aR due the presence of the so-called thin
microstructure of hydrophysical fields in the ocean [1, 4]. For greater 1/aR (e.g., 1/aR = 103 ), as the
calculations show, there are practically no distinctions.
We analyzed thoroughly the distribution (16) for β̃ = 0.1 (a narrow hyperbolic tangent). The calcula-
tions were done for 1/aR = 103 , 102 , 101 , and 100 . A substantial deviation is observed for 1/aR = 1. This

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 29

Fig. 11. (a) Stream lines, Euler equations (1/aR = 5); (b) Oberbeck–Boussinesq approach, t/T = 2, linear stratifi-
cation.

Fig. 12. Stream lines for Euler equations, 1/aR = 100 , t/T = 8.

version for the time value (t/T = 8) is represented in Fig. 12. One can see a substantial deviation from
the calculations by the model in the Oberbeck–Boussinesq approach (see Fig. 4). The nonsymmetry
is explained as follows. The Oberbeck–Boussinesq approach is true under conditions when the value
of ρ differs slightly from its characteristic value (e.g., ρ0 ). However, as the 1/aR decreases, the
density variation cannot be considered a small variation. The complete Euler equations do not allow
for symmetry (antisymmetry) of the solution. In the case of strongly varying ρ, symmetry breakdown
can become appreciable. We also thoroughly analyzed the distribution (16) for β̃ = 1, 2, 5 and for
different values of 1/aR. For illustration, Fig. 13 correlates stream lines calculated by the complete Euler
equations and by the Oberbeck–Boussinesq model for β̃ = 1.0 and 1/aR = 2. There is a substantial
difference in the distribution of stream lines, which, as before, is caused by substantially changing density
in depth.

6. THE PRESENCE OF WAVE BACKGROUND

Using Euler equations in the Oberbeck–Boussinesq approach and Euler–Lagrange variables, we
constructed a numerical model of the local density perturbation dynamics in the presence of wave

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

30 CHERNYKH, ZUDIN

Fig. 13. Stream lines at the time t/T = 2 for Euler equations, 1/aR = 2, β̃ = 1.0 (a); for Oberbeck–Boussinesq
approach, β̃ = 1.0 (b).

background. The background wave perturbation was generated by the following initial data for the
density field [15]:
⎧
⎨ ρ̃ − ỹ, −2.5 ≤ ỹ ≤ 1.1,
0 0
ρ̃s (x̃, ỹ) =
⎩ ρ̃ − tanh((H (x̃, ỹ) − 2.0)/0.1) + tanh(9)) − 1.1, 1.1 ≤ ỹ ≤ 3.0,
0 1

⎧
⎪ ỹ − 3A · cos(ϕx̃)
⎪
⎨ 1 − A · cos(ϕx̃) , 2 + A · cos(ϕx̃) < ỹ ≤ 3.0,
H1 (x̃, ỹ) =
⎪
⎪ 0.9ỹ + 1.1A · cos(ϕx̃)
⎩ , 1.1 ≤ ỹ ≤ 2 + A · cos(ϕx̃).
0.9 + A · cos(ϕx̃)

Here, A = 0.15 and ϕ = 3π/7. The size of the calculation domain was 0 ≤ x̃ ≤ 63, −20 ≤ ỹ ≤ 5.
The grid contained 631 × 251 nodes. The integration domain was chosen sufficiently large in order to
decrease the influence of the boundary conditions; for this reason, the figure represents only part of
the calculation domain (0 ≤ x̃ ≤ 21, −10 ≤ ỹ ≤ 5). Figure 14 demonstrates a nearly linear interaction.
Similar results were obtained for the case of A = 0.15 and ϕ = 6π/7.

CONCLUSIONS
The main results of the work are the following. Using Navier–Stokes and Euler equations in the
Oberbeck–Boussinesq approach, we represented linear and nonlinear numerical models of the dynamics
of local density perturbation in a stably stratified medium. Construction of the nonlinear numerical
models was based on transition to Euler–Lagrange variables. The calculated results have demonstrated
substantial influence of the fluid density distribution on the internal wave pattern generated under
local density perturbation collapse. In the case of the local density perturbation dynamics in a narrow
pycnocline (an ideal fluid), we have demonstrated the known fact of generation of soliton-type stationary
internal waves. The propagation velocity of the principal wave and its amplitude satisfy the known
Benjamin asymptotic relation.
Numerical experiments on evaluating the influence of viscosity on the generation and propagation
of internal waves in a pycnocline have been carried out. We have proposed a simple algebraic relation
between the maximal wave amplitude and the wave propagation velocity in a viscous fluid in a “narrow

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 31

Fig. 14. Stream lines at the time t/T = 2; (a) interaction between wave background and internal waves generated
by local density perturbation; (b) only external wave background; (c) only internal waves generated by local density
perturbation in nonlinearly stratified fluid; (d) the difference of fields (a) and (b).

pycnocline”: One of the quantities being known from calculations, the other is found from the above-
mentioned Benjamin relation.
The linear numerical model of local density perturbation dynamics in a stably stratified medium has
been constructed. We used linear Euler equations and an equivalent fourth-order equation for stream
function, which transforms into Sobolev equation for the case of linear stratification. Applicability of the
linear model has been evaluated for a series of density distributions of an unperturbed fluid. The results
of calculations using complete Euler equations (without the Boussinesq approach) have been presented.
Using the wave perturbation of a pycnocline, we have considered the problem on interaction between
the internal waves, generated by the local density perturbation, and the background perturbations. For
the considered class of pycnocline perturbations, we have obtained a nearly linear interaction.

ACKNOWLEDGMENTS
This work was supported by the Russian Foundation for basic Research (project no. 10-01-00435)
and by the Grants Council (under RF President) for State Aid of Leading Scientific Schools (grant NSh-
6068.2010.9). The authors are grateful to O.F. Voropaeva for critical comments and useful discussions.

REFERENCES
1. Fedorov, K.N., Tonkaya termokhalinnaya struktura vod okeana (Thin Thermohaline Structure of Ocean
Water), Leningrad: Gidrometeoizdat, 1976.

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

32 CHERNYKH, ZUDIN

2. Lighthill, J., Volny v zhidkostyakh (Waves in Fluids), Moscow: Mir, 1981.

3. Maderich, V.S., Nikishov, V.I., and Stetsenko, A.G., Dinamika vnutrennego peremeshivaniya v stratifit-
sirovannoi srede (Dynamics of Internal Mixing in a Stratified Medium), Kiev: Naukova Dumka, 1988.
4. Monin, A.S. and Ozmidov, R.V., Okeanskaya turbulentnost’ (Ocean Turbulence), Leningrad: Gidrome-
teoizdat, 1981.
5. Turner, J.S., Buoyancy Effects in Fluids, Cambridge Univ., 1973.
6. Munroe, J.R, Voegeli, C., Sutherland, B.R., Birman, V., and Meiburg, E.H., Intrusive Gravity Currents from
Finite-Length Locks in a Uniformly Stratified Fluid, J. Fluid Mech., 2009, vol. 635, pp. 245–273.
7. Gilreath, H.E. and Brandt, A., Experiments on the Generation of Internal Waves in Stratified Fluid, AIAA J.,
1985, vol. 23, no. 5, pp. 693–700.
8. Voropaeva, O.F. and Chernykh, G.G., Internal Waves Generated by Momentumless Turbulent Wake in a
Linearly Stratified Fluid, Matem. Mod., 1998, vol. 10, no. 6, pp. 75–89.
9. Voropaeva, O.F., Moshkin, N.P., and Chernykh G.G., Internal Waves Generated by Turbulent Wakes in a
Stably Stratified Medium, Dokl. Phys., 2003, vol. 48, no. 9, pp. 517–521.
10. Moshkin, N.P., Chernykh, G.G., and Fomina, A.V., On the Influence of Small Total Momentum Imbalance on
Turbulent Wake Dynamics in the Linearly Stratified Medium, Matem. Mod., 2005, vol. 17, no. 1, pp. 19–33.
11. Vasilyev, O.F., Kuznetsov, B.G., Lytkin, Yu.M., and Chernykh, G.G., Development of the Region of Turbu-
lized Fluid in a Stratified Medium, Izv. AN SSSR, Ser. Mekh. Zhid. Gaza, 1974, no. 3, pp. 45–52.
12. Lytkin, Yu.M. and Chernykh, G.G., On the Internal Waves Induced by Collapse of Mixing Zone in a Stratified
Fluid, in Matematicheskie voprosy mekhaniki, Dinamika sploshnoi sredy (Mathematical Issues in
Mechanics, Dynamics of Solid Medium), Novosibirsk: Inst. Gidr. SB RAS, 1975, iss. 22, pp. 116–132.
13. Gushchin, V.A., Separation Method for Problems of Inhomogeneous Fluid Dynamics, Zh. Vych. Mat. Mat.
Fiz., 1981, vol. 21, no. 4, pp. 1003–1017.
14. Zudin, A.N. and Chernykh, G.G., On One Algorithm of Calculating Nonstationary Stratified Flows, in
Chislennye metody mekhaniki sploshnoi sredy (Numerical Methods in Mechanics of Solid Medium),
SB USSR Acad. Sci., Computing Center, Inst. of Theoretical and Applied Mechanics, Novosibirsk, 1982,
vol. 13, no. 5, pp. 32–47.
15. Zudin, A.N. and Chernykh, G.G., Examples of Calculating Nonstationary Stratified Flows Using Euler–
Lagrange Coordinate System, Preprint of Inst. of Theoretical and Applied Mechanics SB USSR Acad. Sci.,
Novosibirsk, 1985, no. 9.
16. Stepanyants, Yu.A., Sturova, I.V., and Teodorovich, I.V., Linear Theory of Generation of Surface and Internal
Waves, Itogi Nauki Tekhn., Ser. Mekh. Zhidk. Gaza, VINITI, 1987, vol. 21, pp. 93–179.
17. Zudin, A.N. and Chernykh, G.G., Internal Waves Generated by Local Density Perturbation in a Nonlinear
Stratified Fluid, in Modelirovanie v mekhanike (Modeling in Mechanics), SB USSR Acad. Sci., Comput-
ing Center, Inst. of Theoretical and Applied Mechanics, Novosibirsk, 1988, vol. 2, no. 4, pp. 49–74.
18. Voropaeva, O.F. and Chernykh, G.G., Evolution of the Turbulent Mixing Zone in a Nonlinear Stratified Fluid,
in Modelirovanie v mekhanike (Modeling in Mechanics), SB RAS, Computing Center, Inst. of Theoretical
and Applied Mechanics, Novosibirsk, 1998, vol. 3(20), no. 5, pp. 3–29.
19. Chernykh, G.G., Voropayeva, O.F., and Zudin, A.N., Numerical Simulation of Internal Waves Generated by
Local Density Perturbation, Stratified Medium, Mathematical and Numerical Aspects of Wave Propa-
gation (Proc. 3d Int. Conf.), Cohen, G., Ed., Philadelphia, PA: Soc. Industrial and Applied Mathematics,
1995, pp. 119–128.
20. Smirnov, G.V., Veden’kov, V.E., and Galkovskii, A.N., On the Conditions of Generation of Nonlinear Internal
Waves under Collapse of Mixed Spot in a Stratified Fluid, Okeanologiya, 1997, vol. 37, no. 3, pp. 338–344.
21. Terez, D.E. and Knio, O.M., Numerical Simulations of Large-Amplitude Internal Solitary Waves, J. Fluid
Mech., 1998, vol. 362, pp. 53–82.
22. Maderich, V.S., van Heijst, G.J., and Brandt, A., Laboratory Experiments on Intrusive Flows and Internal
Waves in a Pycnocline, J. Fluid Mech., 2001, vol. 432, pp. 285–311.
23. Chernykh, G.G. and Zudin, A.N., Dynamics of Local Density Perturbation in Shear Flow of a Linearly
Stratified Medium, Russ. J. Num. Anal. Math. Modeling, 2003, vol. 18, no. 2, pp. 117–134.
24. Chernykh, G.G. and Zudin, A.N., Linear and Nonlinear Numerical Models of Local Density Perturbation
Dynamics in a Stable Stratified Medium, Russ. J. Num. Anal. Math. Modeling, 2005, vol. 20, no. 6, pp. 513–
534.
25. Yanenko, N.N., Metod drobnykh shagov resheniya mnogomernykh zadach matematicheskoi fiziki (The
Subincremental Method of Solving Multidimensional Problems of Mathematical Physics), Novosibirsk:
Nauka, 1967.
26. Benjamin, T.B., Internal Waves of Permanent Form in Fluids of Great Depth, J. Fluid Mech., 1967, vol. 29,
pt. 3, pp. 559–592.
27. Gavrilov, N.V., Solitary Internal Waves in a Two-Layer Fluid, Zh. Prikl. Mekh. Tekhn. Fiz., 1986, no. 5,
pp. 49–54.

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011

 DYNAMICS OF LOCAL DENSITY PERTURBATION IN STABLY STRATIFIED FLUID 33

28. Gorodtsov, V.A. and Teodorovich, E.V., Linear Description of Evolution (Collapse) of Symmetrical Pertur-
bation Distributions of a Uniformly Stratified Fluid, in Metody gidrofizicheskikh issledovanii (Methods of
Hydrophysical Investigations), Gorky: Inst. of Applied Physics USSR Acad. Sci., 1984, pp. 134–147.
29. Chernykh, G.G., Lytkin, Yu.M., and Sturova, I.V., Numerical Simulation of Internal Waves Induced by the
Collapse of Turbulent Mixed Region in Stratified Medium, Proc. Int. Symp. on Refined Modeling Flows,
Paris, 1982, pp. 671–679.
30. Gabov, S.A. and Sveshnikov, A.G., Zadachi dinamiki stratifitsirovannykh zhidkostei (Problems of
Dynamics of Stratified Fluids), Moscow: Nauka, 1986.
31. Sobolev, S.L., On One New Problem of Mathematical Physics, Izv. AN SSSR, Ser. Matem., 1954, vol. 18,
no. 1, pp. 3–50.
32. Granberg, I.G. and Dikii, L.A., Stationing in a Nonlinear Problem of Airflow around Mountains, Izv. AN
SSSR, Ser. Fiz. Atm. Ok., 1972, vol. 8, no. 3, pp. 264–269.
33. Mallik, D.D., Some Comments on Equations of Hydromechanics and Their Exact Solutions, Vestnik LGU,
Ser. Mat., Mekh., Astr., Leningrad: LGU, 1967, iss. 1, pp. 98–105.

JOURNAL OF ENGINEERING THERMOPHYSICS Vol. 20 No. 1 2011