Numerical Integra 00 S Lau

NUMERICAL INTEGRATION EXPERIMENTS WITH
A BAROTROPIC PRIMITIVE EQUATION MODEL
Jimmy Ray Slaughter
VAL POSTGRADUATE SCHOOL

onterey, California
T
Numerical Integration Experiments With
a
Barotropic Primitive Equation Model

by
Jimmy Ray Slaughter
Thesis Advisor
G.
J.
Haltiner
March 1972
Approved
loh.
public. fiaXiah^',
dib&ubutlon unlunotzd.
Numerical Integration Experiments With

a
Barotropic Primitive Equation Model
by
Jimmy Ray Slaughter Lieutenant Commander, United States Navy B.S., United States Naval Academy, 1963
Submitted in partial fulfillment of the requirements for the degree of
MASTER OF SCIENCE IN METEOROLOGY

from the
NAVAL POSTGRADUATE SCHOOL
March 1972
DUDLEY KNOX LIBRARY NAVAL POSTGRADUATE SCHOOL

MONTEREY, CALIFORNIA 93943-5002
ABSTRACT
Experiments are carried out with various time and space
differencing schemes applied to the barotropic primitive

equations using both real data and
a
particular stream func-
tion which is an analytic solution to the nondivergent baro-
tropic vorticity equation.
With both types of data, there
were some significant differences in the forecasts produced

by the various schemes.
Replacement of the widely used leap
frog
(centered)
scheme by others which eliminate or lessen
some of its inherent errors at the expense of more computer time or storage appears to be justified at such time when
computer capacity no longer restricts operational use of

these more time consuming schemes.
TABLE OF CONTENTS
I.
INTRODUCTION
II.
PROGNOSTIC EQUATIONS
A.
10 10 10
SPACE DIFFERENCING
1.
2.
Second Order Scheme
Fourth Order Scheme
11 11
B.
TIME DIFFERENCING
1.
2.
3.
Leap Frog
Runge Kutta
11
11
12
Euler Backward
III.
BOUNDARY CONDITIONS
A.
13
PARTICULAR STREAM FUNCTION

1.
2.
3.
13 13
13
North-South Boundary Conditions

East-West Boundary Conditions
Variation for Fourth Order Space Difference
14 14 18
B.
ACTUAL 500 MILLIBAR DATA
IV.
GRID
A.
B.
A PARTICULAR STREAM FUNCTION
18
18
19
FNWC 63x63 HEMISPHERIC
V.
INITIAL CONDITIONS
A.
B.
19
20
22 22
VI.
RESULTS
A.
B.
C.
22
COMPUTER REQUIREMENTS
23
VII.
CONCLUSIONS
28
35 36
BIBLIOGRAPHY
INITIAL DISTRIBUTION LIST
DD FORM 1473
37
LIST OF TABLES
TABLE
I.
Table of root mean square errors in position and central height of the three low centers at 24, 48, and 72 hours as forecast by various differencing schemes (on actual data) in comparison to the corresponding analysis.
21
LIST OF FIGURES
1.
Illustrating the 63x63 square grid and boundaries as applied to a particular The dimensionless stream function. symbols x and y are used for notation convenience
.
16
2.
Illustrating the 65x65 expansion of the grid in Fig. 1 in order to apply fourth order space differencing over the same area. The dimensionless symbols x and y are used for notation convenience.
17
3.
Initial height field as produced by a particular analytic stream function on a 63x63 grid. Only the central contours of the three features are illustrated.--
21
4.
Relative position and central height of a particular stream function as forecast by various differencing schemes.
24
5.
Forecast and analyzed relative positions and central heights for an actual data 500 millibar feature designated low number one.
25
6.
Forecast and analyzed relative positions and central heights for an actual data 500 millibar feature designated low number two.
Forecast and analyzed relative positions and central heights for an actual data 500 millibar feature designated low number three.
One feature in an initial height field produced from a particular stream function illustrating the central height and 390 meter contour.
26
7.
27
8.
30
9.
Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the leap frog scheme.
31
10.
Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the Runge Kutta scheme. Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the Euler backward scheme.
32
11.
33
12.
Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the leap frog fourth order space differencing.
34
LIST OF SYMBOLS
Phase speed.
Grid distance. Grid distance at 60 degrees latitude.
d'
f
f
Coriolis parameter.
Coriolis parameter at 60 degrees latitude
Fleet Numerical Weather Central.
Force of gravity.
FNWC
g
Height of meteorological surface.

x
L L
Wavelength in the x direction.

Half
a
wavelength in the
.
direction.
m
u
Map factor
Velocity component in the x direction.

Velocity component in the
y
direction.
v
a
C
Initial meridional wind speed.
Relaxation coefficient.
Relative vorticity.
Geopo tential
<j>
ty
Stream function.
Initial stream function amplitude.
I.
INTRODUCTION
In selecting a particular differencing scheme for appli-
cation to numerical weather prediction,
major consideraTherefore,
tion is computer time and storage requirements.
the selected scheme is a compromise between verification ac-
curacy and computer limitations.
At present,
the leap frog
(centered) scheme is widely used in numerical weather prediction and will be used as
son.
a
standard for purposes of compari-
Present computer limitations and methods of verifica-
tion tend to preclude seeking more accurate solutions based

on mathematical considerations alone.
The purpose of this
study is to look beyond present computer capabilities and
examine several other differencing schemes which reduce or
eliminate some of the inaccuracies in the leap frog scheme.

This scheme has three time levels and,
therefore,
spurious
computational mode arises from the finite differencing

scheme.
of
This problem can be eliminated by the application
two
level time schemes such as the Euler backward or

A
Runge Kutta methods which have on computational mode.
fourth order space differencing scheme is applied to reduce
truncation error and improve phase speeds.
Should any of
the schemes produce a significant change in forecast rela-
tive to the leap frog method,
it must be determined if
the
difference represents
sufficient increase in accuracy to
warrant the added computer time and storage requirements.
II
PROGNOSTIC EQUATIONS
The equations of motion and the continuity equation for

a
barotropic fluid in the momentum flux form are

+ m + m
.
9(uh)
fr-r
dt
-r
dx
8
9(vh) ~
9 t
-r
9
+ m (vuh) +
(uuh)^
J
m m
9
"^
ay
8
9
(uvh) m
fvh =
v 9h mgh-x
dx
(1)
(2)
tt
y
J
(vvh) , ,9h + fuh = - mgh-r 9y m
9h
9t
+ m
[it
(uh)
9xm
9
-z
(vh),
dy
(3)
where
is
the depth of
the fluid
g
(or height of
pressure
f
surface in the atmosphere),
is
the force of gravity,
is
the coriolis parameter and u and v are horizontal velocity
components
A.
SPACE DIFFERENCING
1
.
Second Order Scheme
With u, v, and
given at each grid point,
the
differential equations may be approximated by

(uh)
9t
it 3
.
.
2 //jr/ /Ad[(u. -m /4df (u.
+1>
.,
_l + .+
.
uh uh 1 + 1>3. J 1J\ J u..)(; u..)( + )
'
(4)
uh
-
(u.
l-l,
.+
j
u. .)( m 13
1,3 +
.
.
uh
1J
+ (u. .._+ * 13 1,3+1

-
)t-J-* +1 +
.
vh
m vh
m
vh m
m vh
1J
)
. .
'
(u.
1,3-1
n
.
+ u..)(
13
.
1,J
1J
)]
+ f..(vh).. '13 13
.
mgh.
(h.
13
1+1,3
x
- h.
. .)/2d = F. i-l3 13
where
1)
increases with
and
decreases with
y.
(See Fig.
Similar expressions are used for the other two equa-
10
j J
tions.
This method of space differencing, devised by A. Ara^

[1]),
kawa (see, for example,
prevents nonlinear instability
in the continuous time case
(sometimes referred to as the
semi-discrete equations).
2
.
Fourth Order Scheme
Fourth order space differencing which takes into account

a
greater number of grid points and reduces trunca-
tion error approximates uh tendencies by
f^l.3
- */3(,
1J
uh
li
JJ
m
+ (u
i+2, uh n 1-2, )-(u, , 4 + u. 4 )< i-2,j ij m

.
-Hij
m 8d
[<u
+
.
u..)( i+2 xj m
uh
+
:
'
(5)
vh
vh
m
.
i,j+2
+
-<u.
iJ-2
xj
.
u..)( X J ij m
.
vh
vh m
1J
)]
ij
(vh)..
xj
-mgh.
(h, ,_
x + 2,j
h.
i-2,
,)/4dl
j
where
F.
i,J
is
the second order space difference.
B.
TIME DIFFERENCING
U
represents uh
r)
vh
or h;
is
time; n is time step
number and
s
is
-^
o t
TT
in the following time differencing
chemes
1
Leap Fro g
U
n+1
11
-1
2AtF
Runge Kutta
AUi AU
2
=
=
At F[U
n
,
nAt]
At F[U + AU 1 /2,
(n+j)At]
11
AU
=
3
At F[U
AU
(n +
p
1)
At]
AU
=
4
At F[U
n
+ AU
(n +
At]
n+1
+ rtÛ- + 2AU + 2AU_ + AU.] l 1 O J 4
3.
Euler Backward
n U* - U n
= =
At F
n+1
-U
At F*
Denotes trial values.
12
. .
Ill
BOUNDARY CONDITIONS
A.
The boundary conditions in this case are applied to an
initial stream function which is an analytic solution to the
nondivergent barotropic vorticity equation and after each

time step during the forecast.
1
North-South
Boundary Conditions
a
For the 63x63 grid
wall exists between y=l and

(See Fig.
1).
y=2 and also between y=63 and y=62

is
No flux
permitted across this rigid boundary.

x,63
x
,
This is accom=
plished by setting v
making the average
v
-v
x,62
,_
and v
x,l
-v
x,2
thereby
=0
y
alone the wall.
The other para-
meters are set equal across the wall.
Summarizing, the
boundary conditions are

North
v
U
South
V
n x,l
x, 63
x
,, x , 62
= - V
x,2
(LI 63 =
x, 62
n 1
=u
=
*x,63
*x,62
is
*x,l
*x,2
where
2
(j)
the geopo tent ial
East-West Boundary Conditions

The East-West boundary conditions are cyclic as
suggested by the function plotted in Fig.

the boundary conditions are
1.
Summarizing,
13
East
l,y
62,
v
West
63, y
2,y
and
d>
where A represents parameters

3
.
x,y
x,y
x,y
Variation for Fourth Order Space Difference

In order to apply fourth order space differencing
and still maintain the boundary conditions of the original grid,

it was
convenient to create an additional outside row

grid.
(See Fig.
2)
of points and relabel as a 65x65
The
boundary conditions are summarized as follows:

North
x,65
v
South
x,62
x,l
v
x,4
A
=
x,64 x,65
x , 54
,.
=-v
x,63
x,62
,_
x,2
-v
x >y J
represents
j and
x,3 x,4
x,3
a m
x,y
x,y
x,l
x,2
x,63
East
A
West
A
62, y
l,y
65, y
4,y
3,y
A
v
represents
>
y
,
>
2,y
A-
63, y
A,. 64, y
A_
x,y
and
(f>
x,y
B,

When the equations were integrated using real data, the
"spongy" boundary conditions of the FNWC model were applied

After each time step, the prognostic results were modified
by setting
uh =
t
(1-k )uh_ + k
f
uh
vh
h
=
=
(l-k')vh + k' vh
(l-k')h
+ k'
h
14
where the
subscript indicates initial values.
Below lat-
itude four degrees north set k'=4.

set k'=0.
Above 17 degrees north
Between four degrees north and 17 degrees north,

'
there is linear variation in k
from zero to one.
This re-
stores the variables to their initial values south of four
degrees north, while north of 17 degrees north, full varia-
bility is permitted.
degrees north.
There is no restoration north of 17

there is partial restoration.
Between,
15
^_
/ /
1
y = 3
*wfl[*
>
/it/
it/
zrnxz&L
ytf2
^xxaxxxzznxzn
"
zzzzzzzzzzgafe
Xl
*2
y-2.
iu/i/i//i//ii ///tuf f** ^JJft
tg./i mui/ uiiiinii wtr/u,//

i
Illustrating the 63x63 square grid and boundaries as applied to a particular stream function. The dimens ionless symbols x and y are used for notation convenience
.
Figure
16
~- H=GS
rn/i >//> jf/t
j/j ///jrfZ>
/VORTH
WfiU
j--63
?&/)} // /
j/f sit
> >
//
fti-r? / /JtJj/7
y--62
\/\
X'G3
$
X=l
X2
*3
i
X sif
f
I
X-S2
X s C
X'fer
1/1
!/
y=*
y-3
gzzzzzgzzzzzzzz zzzzzzzzzz^c 5$JJ ^zazzzzzzzzzzzzzzzzzzzzzzzzgzzzzzzz
yr2.
Illustrating the 65x65 expansion of the grid in Fig. 1 in order to apply fourth order space difThe dimens ionles ferencing over the same area. symbols x and y are used for notation convenience
Figure
2
17
IV.
GRID
A.
A 63x63 square grid as in Fig.

Fig.
2
and as modified in
is used
to
for fourth order space differencing
minimize the number of changes required to adapt programs

to the FNWC 63x63 hemispheric grid.
The boundaries in
latitude extend from 30 to 90 degrees.
Grid length is
determined by dividing the geographic distance represented

by this latitude difference into 62 equal parts.
B.
FNWC 63x63 HEMISPHERIC

In conjunction with the actual 500 millibar data,
the
FNWC 63x63 hemispheric grid is used.
This is
square re-
presentation of
hemispheric polar stereographic grid in

d
which the grid distance

,
is determined by
,1
1
"m
, .
' (
+ sin 60 + sin Y
is
where m is the map factor,
d'
the grid distance at 60

y
=
latitude (381 kilometers), and sin

sxn y
=
is
,_
defined by
_
973.71
R2
2
wnere D R
,
(I-I
P
.2,
)
_ (J-J ,2 )
,_
973.71 + R
I, J
I
arbitrary grid point.

grid point for the north pole
=
,J
P
P
32,32.
18
V.
INITIAL CONDITIONS
A.

In order
to obtain a smooth initial stream function
field,
an analytic solution to the nondivergent barotropic
vorticity equation is used, namely

]b
lb
sin
L
y
(y
-r)
cos
(x-ct)
where
is
the phase speed and the wavelength in the x diry
ection and the half wavelength in the

to
direction are equal
61d
y
i
(L = L = 61d) a
y =
i
The initial stream function is

d>
-r-)
i>
sin
(y -
Li
cos
2tt
L X
Here,
I
\p
is
z
determined by using the relation

o
v L
ô
2tt
where v
is
the initial specified amplitude of
the meri-
dional wind speed derived from

V =
drp
2tt
9^
L~ ô x
[sin
r
IT
L~
y
(y "
2>
COS
tt
Z~ x
X]
Using the initial field thus obtained, the linear balance
equation
V
2
<f>
fV
2
ijj
+ Vf
ViJ;
which in finite difference form may be written

?
2
<f>
.
.=
1J
1J
iK. Y
1J
+t[^ X f 4
^+ v
y
V
y
i>] yj
F. ij
.
19
where
*2
<
>
<
>
i+ i (j
<
<
>
t -i. i
<
>i, j+ i
<
>i.j-i-
<
>u
\-
<
>i + i.3
h-l,i
The preceding equation may now be solved as a Poisson
equation for the geopotential field.

V 2 <K ij
.
Thus,
F. ij
.
The initial guess for the relaxation procedure is

6 T =
o
f
\b
where
is a mean coriolis parameter. r
Subsequent "guesses" are made according to the relation

ij
where
R. ij
(n)
a(? 2 <J>.f
ij
n)
F. .) ij
and a is
a
=
relaxation coefficient.
4>/g,
Using
computations for the initial height field

Additionally,
a
are completed.
constant height is added

to
to each grid point value
in order
eliminate negative
heights, which would otherwise cause computational instability.

(See Fig.
3)
B.

The initial height field is obtained from the FNWC 500
millibar analysis.
The linear balance equation is solved

ip=
<J)/f
for the stream function as a Poisson equation using

as a first guess
20
Differencing Scheme
Leap Frog Leap Frog Fourth Order Space
RMSE For Position (Nautical Miles)

97
76
RMSE For Central He i ght (Me t e r s

30
31
Runge Kutta
93
31
25
Euler Backward
88
Table of root mean square errors in position and central height of the three low centers at 24, 48, and 72 hours as forecast by various differencing schemes (on actual data) in comparison to the corresponding analysis. TABLE
1
Latitude
90
{ Goo 580
WO
tOO
520
600
J-j
Latitude
30
Initial height field as produced by a par t icular analytic stream function on a 63x63 grid. Only the central contours of the three features are illustrated.
Figure
3
21
VI.
RESULTS
The results shown in the following figures were obtained
from the computer subroutine mapping of the 63x63 height

field obtained by the various differencing schemes using 10
minute time steps.

simply as
a
One grid squares are used in the figures

a
convenience to provide
ready reference scale
to compare positions of low centers.
A.
PARTICULAR STREAM FUNCTION

Results for the forecasts made using
a
particular
stream function for the initial values are summarized in

Fig.
4
for central height and relative position.
The fore-
cast results for the individual differencing schemes are
presented in greater detail in Fig.

B.
through Fig. 12.

The results of forecasts made on three low systems from
the 500 millibar FNWC analysis are summarized in Fig. 5,

Fig.
6
and Fig.
7.
Although only low systems are illustrasimilar results were obtained from the
The time frame of
ted in this study,
movement and intensification of highs.
this study begins with the 1200Z March 24,
1966 500 milli-
bar heights and terminates with the 1200Z March 27, 1966
forecast.
Low number one (See Fig. 5) was located over
central Siberia.
Low number two (See Fig.

7)
6)
and low num-
ber three (See Fig,
were located over Southwestern
22
Canada and Labrador respectively.
Root mean square error
computations for position and central height as forecast

by each differencing scheme when compared to the correspond-
ing analysis are contained in Table 1.
C.
COMPUTER REQUIREMENTS
Although peak program efficiency was not attempted dur-
ing the application of the various numerical schemes,
con-
sistency in application was carefully maintained.

the leap frog scheme as a standard,
Using
the leap frog fourth
order space differencing and Euler backward schemes require
approximately
35
percent increase in run time.

a
The Runge
Kutta scheme requires approximately
170 percent increase
in run time and a 25 percent increase in computer storage.
23
N
n
72hr
f""
I
72hr
48hr
| I 1
24hr
i 1
I
I
I
&
l:::J
7*
58 Nautical Miles
ijL.
One grid square
Differencing Scheme Forecas t

x
Central Height (Meters)

24
hr.
48 hr.
72
hr.
Leap Frog
382
383
383
oLeap Frog Fourth Order Space

iRunge Kutta
383
382
384
383
380
385
383
384
Euler Backward
383
Initial Central Height
383 Meters
Relative position and central height of a particular stream function feature as fore cast by various differencing schemes.
Figure
4
24
N
A
72hr
L
I
48
hr
~ 5 A
n
,
IT
!
i
L
24 hr
206 Nautical Miles
1.
One. grid square
Differencing Scheme Forecast

x

24
hr.
48 hr
72
hr.
Leap Frog
4930
4940 4960
4990 5030
4940
4930
vRunge Kutta
4940
4940
4990
4990
Euler Backward
Corresponding FNWC Analysis
4940
5020
5090
5040
<g)5110
Initial Height
FNWC Analysis
Forecast and analyzed relative positions and central heights for an actual data 500 millibar feature designated low number one.
Figure
5
25
72hr
o
O/ll
1
48hr
72 hr
I (
0)
r
i
\oS
L-L-J
i
l
L.__J
i-x "
I
206 Nautical Miles
One grid square
Differencing Scheme Forecast

*
Central Height
24 hr.
(Meters)
72 hr.
48 hr
Leap Frog
5020
5070
5080
5050

ARunge Kutta
5060 5020 5020
5080
5080 5080
5060 5080
Euler Backward
5020
5120
5100
5160
Initial Height
FNWC Analysis
Forecast and analyzed relative positions and central heights for an actual data 500 millibar feature designated low number two.
Figure
6
26
N
72hr
(
1
48hr
| I
*A O
24hr]
o
! I i
IT
',206
Nautical
Mies
(2)
J,
One grid square
Differencing Scheme Forecas t

x

24 hr
48 hr.
72 hr.
Leap Frog
Leap Frog Fourth Order Space
4960 4960 4960
4960 4980
4910
4960
4920 4930
ARunge Kutta
4950
4960
Euler Backward
4960
4960
5110
4990
Filled
Initial Height
FNWC Analysis
Forecast and analyzed relative positions and central heights for an actual data 500 millibar feature designated low number three.
Figure
7
27
VII.
CONCLUSIONS
The various differencing schemes were applied to the
smooth analytic solution in order that differences in forecasts produced by these schemes would be more readily dis-
tinguishable than with actual data.

as
However, the choice

(61
to resolution was probably
too high in this study
grid lengths per wavelength) and shorter wavelengths are

yet to be tested.
Differences produced in location and
central values were not conclusive when compared to the results produced using actual data.
tion for the analytic initial field
A reduction in resolu-
(approximately six grid

a
lengths per wavelength) would possibly produce
more mean-
ingful separation in the accuracy produced by the various
schemes
The negligible increase in accuracy
on a root mean square error comparison)
(four percent based

of
the Runge Kutta
scheme over the leap frog scheme does not justify the ex-
tremely large increase in computer time and moderate increase in computer storage.
Both the leap frog fourth
order space and Euler backward differencing schemes show
promise for improved accuracy with
moderate increase in
.
computer run time (approximately 35 percent)
The leap
a
frog fourth order space differencing scheme gave

cent increase in position accuracy.
scheme produced
a
24
per-
The Euler backward
nine percent increase in position accur-
28
acy and
17
percent increase in central height.
With both
the Euler backward and the leap frog fourth order space
schemes,
the 35 percent increase in computer time could be
reduced somewhat by more efficient programs.
The small
statistical sample of pressure systems examined in this

study does not permit firm conclusions to be drawn regarding
the various
schemes tested.
However,
there is reason to
believe that the fourth order space differencing of the
nonlinear advection terms will reduce the phase error in the
prediction of pressure systems and this method should be

tested further.
Williams
[2]
has shown that the pressure
force need only be differenced with second order accuracy
which permits
somewhat larger time step without computa-
tional instability than when fourth order differences are

used throughout.
29
206
N.
Mi.
390
meters
One grid square
One feature in an initial height field pro duced from a particular stream function illustrating the central height and 390 me ter contour
Figure
30
L
206
N.
Mi.
24 hr. 48 hr.
72 hr.
One grid square
Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the leap frog scheme Figure 9
31
24 hr. 48 hr.
One grid square
72 hr.
Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the Runge Kutta s cheme
Figure 10
32
/
N
24 hr. 48 hr.
72 hr. One grid "square
Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the Euler backward scheme
Figure 11
33
/
/
V*.
%,
Immmm
24 hr.
48 hr.
72 hr.
N.
k.
Mi.
One grid square
Forecast height in meters of central values and positions of centers and 390 meter contours for a particular stream function by the leap frog fourth order space differencing.
Figure 12
34
BIBLIOGRAPHY
1.
Haltiner, G. J., Numerical Weather Prediction

1971.
Wiley,
2.
Naval Postgraduate School Report 5lWu72031A, Phase Speed Errors with Second and Fourth Order Space Differences with Staggered and Unstaggered Grids by R. T. Williams, March 1972.
35
INITIAL DISTRIBUTION LIST

No
1.
.
Copies
2
Defense Documentation Center Cameron Station Alexandria, Virginia 22314

Library, Code 0212 Naval Postgraduate School Monterey, California 93940
2.
3.
Professor G. J. Haltiner, Code 51Ha Department of Meteorology Naval Postgraduate School Monterey, California 93940
4.
Lieutenant Commander J. R. Slaughter Carrier Airborne Early Warning Training Squadron 110 (RVAW-110) Naval Air Station, North Island San Diego, California 92135
Department of Meteorology Naval Postgraduate School Monterey, California 93940
Naval Weather Service Command Washington Navy Yard Washington, D. C. 20390
5.
6.
36
Security Classification
DOCUMENT CONTROL DATA

(Security classification of
title,
-R&D
is
body of abstract and indexing annotation mux/ be entered when the overall report
2a.
classitied)
ORIGINATING ACTIVITY
(Corporate author)
REPORT SECURITY CLASSIFICATION
Naval Postgraduate School Monterey, California 93940

REPORT TITLE
Unc lassified
2b.
GROUP
Numerical Integration Experiments with Barotropic Primitive Equation Model

4.
DESCRIPTIVE NOTES (Type

AUTMORiS)
of report and.inclusive dates)
Master's Thesis, March 1972

5-
ffirsl name, middle initial, last
name)
Jimmy
6.
R.
Slaughter
7a.
REPORT DATE
TOTAL
NO.
OF PAGES
7b.
NO. OF REFS
March 1972
8a.
38
NO.
9a.
CONTRACT OR GRANT
PROJEC
T NO.
ORIGINATOR'S REPORT NUMBER(S)
6.
96.
OTHER REPORT
this report)
NO(S) (Any other numbers that may be assigned
10.
DISTRIBUTION STATEMENT
Approved for public release, distribution unlimited

11.
SUPPLEMENTARY NOTES
12.
SPONSORING
MILI
TAR Y ACTIVITY
Naval Postgraduate School Monterey, California 93940

13.
ABSTR AC T
Experiments are carried out with various time and space differencing schemes applied to the barotropic primitive equations using both real data and a particular stream function which is an analytic solution to the nondivergent barotropic vorticity equation. With both types of data there were some significant differences in the forecasts produced by the various schemes. Replacement of the widely used leap frog (centered) scheme by others which eliminate or lessen some of its inherent errors at the expense of more computer time or storage appears to be justified at auch time when computer capacity no longer restricts operational use of these more time consuming schemes.
FORM NOV 65
1473
1
(PAGE
1)
37
S/N 0101 -807-681
A-31408
key wo R DS
barotropic primitive equations

numerical forecast
DD
F
,
o1".,1473 <back
a3
1
S/N 0101-507-6321
409
Sl
*9hter
pef-iment!
1 3427n
. **
êrf Ca/
4U
. '
nte 9r
at!
525
Thesis S57025 Slaughter c- 1 Numerical integration experiments with a barotropic primitive equation mode?
134270
thesS57025
N ^rical
integration experiments with

III Hill
3 2768 002 01141 3 DUDLEY KNOX LIBRARY

Numerical Integra 00 S Lau

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Numerical Integra 00 S Lau

Enviado por

Direitos autorais:

Formatos disponíveis

NUMERICAL INTEGRATION EXPERIMENTS WITH

A BAROTROPIC PRIMITIVE EQUATION MODEL

Jimmy Ray Slaughter

VAL POSTGRADUATE SCHOOL

Barotropic Primitive Equation Model

Jimmy Ray Slaughter

Numerical Integration Experiments With

Barotropic Primitive Equation Model

Submitted in partial fulfillment of the requirements for the degree of

MASTER OF SCIENCE IN METEOROLOGY

NAVAL POSTGRADUATE SCHOOL

DUDLEY KNOX LIBRARY NAVAL POSTGRADUATE SCHOOL

differencing schemes applied to the barotropic primitive

particular stream func-

tion which is an analytic solution to the nondivergent baro-

tropic vorticity equation.

With both types of data, there

were some significant differences in the forecasts produced

Replacement of the widely used leap

scheme by others which eliminate or lessen

computer capacity no longer restricts operational use of

Second Order Scheme

Fourth Order Scheme

PARTICULAR STREAM FUNCTION

North-South Boundary Conditions

Variation for Fourth Order Space Difference

ACTUAL 500 MILLIBAR DATA

A PARTICULAR STREAM FUNCTION

FNWC 63x63 HEMISPHERIC

A PARTICULAR STREAM FUNCTION

ACTUAL 500 MILLIBAR DATA

A PARTICULAR STREAM FUNCTION

ACTUAL 500 MILLIBAR DATA

Height of meteorological surface.

Wavelength in the x direction.

Velocity component in the x direction.

Initial meridional wind speed.

In selecting a particular differencing scheme for appli-

cation to numerical weather prediction,

tion is computer time and storage requirements.

the selected scheme is a compromise between verification ac-

curacy and computer limitations.

the leap frog

standard for purposes of compari-

Present computer limitations and methods of verifica-

tion tend to preclude seeking more accurate solutions based

The purpose of this

study is to look beyond present computer capabilities and

examine several other differencing schemes which reduce or

eliminate some of the inaccuracies in the leap frog scheme.

computational mode arises from the finite differencing

This problem can be eliminated by the application

level time schemes such as the Euler backward or

Runge Kutta methods which have on computational mode.

fourth order space differencing scheme is applied to reduce

truncation error and improve phase speeds.

the schemes produce a significant change in forecast rela-

tive to the leap frog method,

sufficient increase in accuracy to

warrant the added computer time and storage requirements.

The equations of motion and the continuity equation for

barotropic fluid in the momentum flux form are

(vvh) , ,9h + fuh = - mgh-r 9y m