36713728.pdf - core
TRANSCRIPT
Calhoun: The NPS Institutional Archive
Theses and Dissertations Thesis Collection
1981
Development of improved finite elements formulation
for shallow water equations.
Woodard, Edward T.
Monterey, California. Naval Postgraduate School
http://hdl.handle.net/10945/20493
NAVAL POSTGRADUATE SCHOOLMonterey, California
THESISDEVELOPMENT OF IMPROVED
FINITE ELEMENT FORMULATIONSHALLOW WATER EQUATIONS
FOR
by
Edward T. Woodward
September 1981
Thesis Advisors: R.T.
U.R.
WilliamsKodres
Approved for public release; distribution unlimited.
T204559T20A560
SECURITY CL AIII'ICATIOM OF THIS »»GE (Whan Dim Em.r.di
REPORT DOCUMENTATION PAGE READ INSTRUCTIONSBEFORE COMPLETING FORM
1 IIH1T NUMltH 2. GOVT ACCESSION NO. 1 RECIPIENT'S CAT ALOG NUMBER
4 TITLE «n* Submit)
Development of Improved Finite ElementFormulation for Shallow VJater Equations
S. TYPE OF REPORT ft PERIOO COVEREDMaster's ThesisSeptember 1981
«. PERFORMING ORG. REPORT NUMBER
7 A _--.?» J
Edward T. Woodward
1 CONTRACT OR GRANT NUMBERS)
t »(M9UING ORGANIZATION NAME AND AOORESS
Naval Postgraduate SchoolMonterey, California 93940
10. PROGRAM ELEMENT. PROJECT TASKAREA ft WORK UNIT NUMBERS
11 CONTROLLING OFFICE NaUI AMO AOORESS
Naval Postgraduate School
Monterey, California 93940
12. REPORT OATE
September 198111. NUMBER OF PAGES
170
T4 MONlTORlNS AGENCY NiMC k AOOHIIM/ Uttfrmnt tram Contra'tUnf Slltca) IS. SECURITY CLASS. o( thia ripori)
IB* DECLASSIFICATION' DOWNGRADINGSCHEDULE
I*. DISTRIBUTION STATEMENT 'ol thf *rpo"
Approved for public release; distribution unlimited
'7 DlSTRiBU T ION STATEMENT at ina •attract anlarad In Black 20. II dlllararti *•• * apart)
It SuPPL EmEnTARY NOTES
It. <EY WORDS CMilnut on tarataa tlda II nacaaaarr and Idamttty PF Black ->umtbmn
Finite elements piecewise constant basis functionShallow water equations primitive barotropic equationsVoriti city-divergence formulation Galerkin methodSemi -implicit numerical prediction model
piecewise linear basis function variable size elementsJO ABSTRACT fC«iilmM on ravaraa ..a. // nacaaaarr ana Idamiltr kw 'lock mammar)
The basic principles of the Galerkin finite element method are discussedand applied to two different formulations; one using different basis functions
and the other using the vorti city-divergence form of the shallow waterequations. Each formulation is compared to the primitive form of the equations
developed by Kelley (1976). The testing involves a comparison of three finiteelement prediction models using variable size elements. Equilateral elementssignificantly improve the solution and are used in most of the comparisons.
DOFORM
I JAN 7 , 1473 EDITION OF 1 NOV •• IS OBSOLETI
S/N 102-0 14- 4S0 1
SECURITY CLASSIFICATION OF THIS PAOE (Wkmm Data Kntatod)
^cuwt» cl«iiihc»"9ii a* Twit »m>*»»i n»n »«—
«
The formulation using different basis functions produces poorer results than
the primitive formulation. The vorti city-divergence formulation producessuperior results while executing faster than the primitive model. However,it does require more storage and the relaxation parameters are sensitiveto the domain geometry. The computer implementation for the vorticity-divergence model is discussed and the source listing is included.
DD FornQ 1473
Approved for public release; distribution unlimited
DEVELOPMENT OF IMPROVEDFINITE ELEMENT FORMULATION FOR
5 HAL LOW -V. ATE? EQUATIONS
by
Edward T. WoodwardCaptain. United States Air ForceB.A., University of Maine, 1970
Submitted in partial fulfillment of therequirements for the degrees of
FASTER OF SCIENCE IN COMPUTE? SCIENCEand
MASTER OF SCIENCE IN METEOPOLCSY
from the
NAVAL POSTGRADUATE SCHOOLSeoterrber 1981
abstract -°«war*The basic principles of the Galerkin finite element
method are discussed and applied to two different
formulations* one using different basis functions and the
other using the vorticity—divergence forrr of the shallow
water equations. lach f angulation is compared to the
primitive form of the equations developed by lelley (1976).
The testing involves a comparison of three finite element
prediction models using variable size elements. Equilateral
elements significantly improve the solution and are used in
most of the comparisons. The formulation using different
basis functions produces poorer results than the primitive
formulation. The vorticity-divergence formulation produces
superior results while executing faster than the primitive
model. Fcwever, it does require more storage and the
relaxation parameters are sensitive to the domain geometry.
The corputer implementation for the vorticity-divergence
model is discussed and the source listing is included.
TA3LE OF CONTENTS
I . INTRODUCTION 11
A. 3ACKGRCTJNE 12
2. OBJECTIVES 15
:. THESIS STRUCTURE 17
II. FINITE ELEMENTS 19
A. FINITE ELEMENT CONCEPT 2Z
2. GALERKIN APPLICATION 25
C. APIA COORDINATES 31
III. SHALLOW-WATER MOTEL 35
GOVERNING EQUATIONS 3?
ECUATION FORMULATION 39
7T V F DISCRETIZATION 41
COMPUTATIONAL TBCENIQUE 43
GRII GEOMETRY 45
INITIAL CONE IT IONS 47
50UNDART CONDITIONS 49
IV. COMPUTER IMPLEMENTATION 52
A . PROGRAM ARCHITECTURE 53
1. v ain Prograrr 55
2. Initialization Phase 56
3. Forecast Phase 64
3. UTILITY ^OIULES 65
V. PRIMITIVE "ODEL EXPERIMENT 63
A. v 0~EL DESCRIPTION 69
3. RESULTS 73
VI. LINEARIZE! MODEL EXPERIMENT 76
A. EQUATION REFORMULATION 78
P. RESULTS 81
Til. TORTICITY-BIVERGENCE MODEL EXPERIMENTS 34
A. TEST DOMAINS AND INITIAL CONDITIONS 84
B. TEST CASE COMPARISONS 87
1. Regular Case 87
2. Srcooth Case 91
3. Atruut Case 102
C. COMPUTATIONAL SENSITIVITY 105
Till. CONCLUSIONS 109
APPENDIX A 'Source List) Ill
A. DEFINITIONS Ill
3. MAIN PROGRAM 114
C. INITIALIZATION PEASE 115
D. FORECAST PHASE 141
Z. UTILITY PROGRAMS 151
LIST OP REFERENCES 167
INITIAL DISTRIBUTION LIST 159
LIST OF FIGURES
Figure 1
.
F i gu ~e 2
.
Figu re 3.
Figure 4.
Figure r"
.
Figure 6.
Figure n.
Figure 3.
Figure 9.
Figure 12
Figure 11
Figure 12
Figure 13
Figure 14
Figure IS,
Figure 16
Figure 1?
Figure 13
Fiecewise linear basis function 23
Basis and test function interactionluring the piecewise integrationore cess 27
Basis function for node 23 30
"artesian versus area coordinates 32
Transformation to area coordinates 32
Domain divided into equilateraltriangles 46
3 dirensional view of the initialfields 48
Assemble and store the coefficientmatrix for elerrent 3 62
Tomain divided into right triangles 71
Initial fields for the primitive""<)lel 72
46 hour forecast comparison betweenequilateral and right triangularelements 74
Differently shaped basis functions 77
Staggering the geopotential aboutthe velcci ty 79
Initial fields for the linearizedexperiment 32
48 hour forecast comparing theprimitive vs linearized model S3
Test domain geometries 35
Initial field for the Regular Case 86
48 hour forecast comparison for theRegular Case 39
7
r :^':re 19.
Figv.re 22.
Ji^ure 21.
Figure 22.
Figure 23.
Figure 24.
Figure 25.
Figure 26.
Figure 27.
Figure 28.
Initial field for the Smooth CaseA? A = 1.1 ir/s 92
48 hour forecast comparison for theSmooth Case A?A = 1.1 rr/s 93
Initial fields for the Smooth CaseA?A = 5.5 rr/s 95
4? hour forecast com-oarison for theSmooth Case, API = 5.5 m/s 96
96 hour forecast comparison for theSmooth Case, APA = 5.5 m/s 97
Initial fields for the Smooth CaseAFA = 1.1 m/s, 2 waves 1ZZ
46 hci;r forecast comparison for theSmooth Case, 2 waves.". 101
Initial fields for the Abrupt Case 123
43 hour forecast comparison for theAbrvpt Case 104
Computational sensitivity using the*e hour Vorticity field 127
3
LIST OF TABLES
Table 1. -oirparison of corputat ional tiires forthe 46 hour forecast, Regular Case S0
Table ?. Comparison of computational times forthe 48 hour forecast, Sirooth Case 91
Table 3. CoTparison of corputat ional tirres for*he 26 hour forecast, Srooth Case 93
Tahle 4. Comparison of computational times forthe 45 hour forecast, Abrupt Case 105
c
ACKNOWLEDGEMENTS
The author telives that most of the credit should go to
Professor R.T. Williams for the education that he gave me
through his patient teaching and guidance during the entire
research and writing of this paper. Acknowledgement is also
due Professor U.R. Kodres for the many useful suggestions,
encouragement and reading the manuscript.
Special thanks to each department chairman, Professor
B.H. 3radley f Computer Science, Professor R.J. Renard,
Meteorology, and Professor !J.J. Ealtiner, former Meteorology
chairman, for allowing me to pursue the two degrees and
accepting this thesis as partial fulfillment for the
requirement? of "both degrees.
Thanks are extended to Lt. R.G. Kelley whose research
was the forerunner to this investigation and Capt. M. Cider
*or the many useful comments he offered during the
implementation. Particular thanks to Dr. M.J. P. Cullen whose
personal suggestions provided the direction for this work.
I give my love to my wife, Joy, for her Infinite
patience, for being the first guinea pig in reading and
preparing this manuscript. Finally, to my family, Kim, Mark
and Joy thanks for enduring and encouraging me throughout my
graduate studies at the Naval Postgraduate School.
10
I. INTHOrUCTION
Shuman fl978) claims that progress in numerical modeling
cf the general circulation has been to some degree dictated
in the past "by the rate of development in the field of
computer technology. Eowever, the limited ability to
Darameterize the effects of small-scale processes in terms
of large scale motions has been an equally important
limiting factor. Essentially, the major problem of numerical
modeling of the general circulation is simply that of
producing a very long range numerical weather forecast.
Certainly the equations used in the models must be more
sophisticated to include those physical processes- which are
unimportant for a short range forecast, but may become
crucial as the length of the forecast is extended. Another
area where concentrated efforts have improved the forecast
involves the computational techniques employed to
approximate and solve the governing equations of the models.
The motivation behind this thesis is to investigate the
application of a relatively new computational technique to
the field of numerical weather prediction. The finite
element method, long established in engineering, has been
seriously considered only during the past decade In
meteorology. This method has great potential for application
in atmosoheric Prediction models.
11
A. BACKGROUND
The rrost common numerical integration procedure for
weather prediction has been the finite difference method in
which the derivatives in the differential equations of
motion are replaced by finite difference approximations at a
discrete set of points in space and time. The resulting set
of equations, with appropriate restrictions, can then he
solved by algebraic methods. Until recently, the finite
difference method has been the workhorse in atmospheric
prediction models, from their first computer implementation
to the present .
With the introduction of each new generation of
computers. the gap between numerical forecasts and
atmospheric observations has decreased. The rate at which
this gap decreased has slowed down and appears to be
leveling off. This would indicate that computer technology
may not be the prirrary obstruction to better numerical
forecasts. In fact, bigger and faster computers alone have
demonstrated their inability to significantly improve the
numerical forecast.
7cr example, a major limiting factor of finite
difference approximations is the truncation error. The
National leather Service 7 Layer Primitive Equation P-odel
(7LPZ Model), operational from 1966 to 1980. had truncation
errors which increased at a rate proportional to the square
o*" the grid spacing. That is, the smaller the grid interval,
12
the siraller the truncation error. To increase its accuracy
would require increasing the grid matrix density. This would
require increased computer storage and computational tiire.
State of the art computers are capable of providing these
additional resources.
The problem now goes beyond numerical techniques and
romputer technology. Operationally, the National Weather
service is not capable (due to monetary restrictions) of
providing a denser concentration of atmospheric
observations. Therefore, with the present density of initial
data ( observations) and objective analysis techniques
(getting the data for grid points by interpolating from
observed data sources), reducing the grid spacing further on
the 7LPZ Model does not significantly increase the accuracy
of the solution.
This additional computer capability can not be utilized
usin«? finite difference methods. Therefore, new numerical
integration techniques must be investigated, such that given
the same density of observed data, superior solutions are
produced .
Two alternative techniques, the spectral method and the
finite element method, have started to gain attention. Both
the spectral and finite element methods require more
computational time per forecast time step than does the
finite difference method. For example, the finite element
method requires an equation solver to invert a larger matrix
13
at each tirre step for each variable. In this sense, these
methods were held hack by computer technology, hut recent
advances in computer technology (i.e. larger and faster
storage devices, multi-processors, etc) have made these
alternative numerical techniques competitive.
For long range weather predictions, the spectral method
applied over the globe or hemisphere is a natural method,
due to the existence of efficient transforms for the
nonlinear terms on spherical geometry. It also eliminates
the truncation error for the horizontal space derivatives
and the nonlinear instability (aliasing). For these reasons,
global spectral models have been developed, and implemented
on an operational level, replacing the global finite
difference models .
Eowever, because the spectral harmonics are globally
rather than locally defined, it is thought that for problems
of more detailed limited area forecasting, the finite
element method is more suitable. Pioneering work to adapt
finite element methods to meteorological applications has
been done by Cullen (1973,1974 and 1979), Staniforth and
Mitchell '1977), Kinsman (1975) and lelley (1976). The most
recent finite element meteorological model at the Naval
Postgraduate School was written by Kelley (1976) with the
-?llaboration of Tr. R.T. Williams. It is this study that
will serve as a basis for this thesis. The model written by
Kelley will be altered and used for comparative testing with
14
irproved finite element fortrs implemented by this author.
Some of the techniques and coles developed by Keliey are
also employed in this thesis. Older (1981) developed a
technique to smoothly vary the grid geometry in the domain.
This technique is also implemented both on Kelley's model
and vith the new formulation to give greater versatility
when testing the model performance.
3. C3J2CTI7ES
The objectives for this thesis can be divided into two
categories: 1) meteorology, 2) computer science. First, the
meteorological objectives of developing improved finite
element forms for shallow water equations are as follows:
1) - Older (1981) after collaboration with Dr.
M.J. P. Cullen, showed how equilaterally shaped elements
produced significantly better results than did other
triangular elements, ^elley (1976) used right triangular
elements in the implementation of a two-dimension.al finite
element model using the primitive form of the shallow water
equations. A considerable amount of small-scale noise was
observed in the solution. Hereafter, this model, which was
developed by lelley (1975), will be referred to as the
primitive model. This first objective involves
re-implementing the primitive model using equilaterally
shaped elements and comparing the results to those in
Zelley's thesis.
15
2^ - v illiaT>s and Zienlciewicz (1981) presented new
finite elerrent techniques for formulations for the shallow
water equations, which use differently shaped functions to
approximate the different dependent variables, which in
effect stagger the variables. Schoenstadt (1980)
demonstrated the advantage of spatial staggering of
dependent variables in finite difference models. The
application of this technique to finite element models is a
natural extension, and excellent results were obtained by
Williams and Zienkiewicz (1981) from application of these
r.ew formulations on linearized one dimensional cases. The
objective here is to implement the new forms on the
primitive jodel and again do quanitative comparisons of the
results
.
?) - The major emphasis in this study deals with the
implementation and comparison of the vorticity divergence
for'r of the shallow water equations, which is described in
detail in Chapter III. This formulation has the following
advantages. First, the geostrophic adjustment process is
treated better than in the primitive form of the equations.
Secondly, the velocity and height fields are evaluated at
the same grid point, where the best primitive form requires
staggering these dependent variables. And thirdly, a larger
time step is allowed due to the semi implicit form of the
equation. Again comparisons between the results from the
vorticity divergence and primitive model are presented.
16
The computer science aspect of this thesis was primarily
devoted to the implementation of the different models and
the style and architecture of the program. Finite element
methods require more computational time than do finite
difference methods, not only in the solution of the
equations, but also in the amount of computation required to
evaluate each term in the equations.
The implementations of these two dimensional models,
although complex when viewed from the surface, have a lot of
generality and redundancy in the operations required.
Versatile modules can he written to ease the implementation
and facilitate changes. The objective here is to efficiently
implement these new forms and demonstrate tne utility of
these versatile modules for future implementations.
C. THZSIS STRUCTURE
This thesis presents the results obtained from tests of
the various finite element formulations. The results are
compared to those from the primitive model written by Celley
(1976). Accompanying the results is a detailed discussion of
the reformMiation and implementation process.
Chapter II of the thesis presents a tutorial of the
finite element method and the area coordinates system used
in the evaluation of the element integration. The Galerkin
finite element method used in all the models is developed
and applied to the advection equation in one dimension.
1?
rhapter in presents the detailed description cf the
vcr f i ri ty-di vergence shallow water model. Here the equations
are shown and written using the Salerkin method. A
discussion of the computational technique used is presented
along with the -model's physical parameters.
Chapter IV presents a descriptive overview of the
computer implementation. The chapter includes a list of
options available for testing, a brief description of the
iratrij compaction technique and the formulations of the
versatile modules used to implement the complex equations.
Chapters V through VII discuss the results obtained fror
the different experiments. Chapter V briefly describes the
primitive model used for all comparisons and the results
froffl changing the element shape to equilateral triangles.
Chapter VI reformulates the primitive model so that the
geopotential is staggered with respect to the velocity
variable. Jor simplicity, the continuity equation is also
linearized. Chapter VII compares the results from the
vert icity-di vergence model developed in Chapter III to those
:"rom the primitive model.
The last chapter summarizes the results from all the
experiments and identifies what areas require follow on
work. The source code for the vorticity-divergence model is
presented in Appendix A.
16
II - FIN ITS 5LZMINTS
As is often the case with an original development, it is
rather difficult to quote an exact date on which the finite
element method was invented, but the roots of the method can
be traced tack to these separate groups: applied
mathematicians, physicists and engineers. Since the early
developments of the finite element method, a large amount of
research has been devoted to the technique. However, the
finite element method obtained its real impetus by the
independent developments carried out by engineers. Its
essential simplicity in both physical interpretation and
mathematical form has undoubtedly been as much- behind its
popularity as is the digital computer which today permits a
realistic solution of even the most comolex situations.
The name finite element was coined in a paper by
R.V. Clough, in which the technique was presented for plane
stress analysis, as discussed in Bathe (1976). While finite
element methods have made a deep impact via the field of
solid mechanics, where it can be said that today they
represent the generally accepted method of discretizing
continuum problems for computer-based solution, the same
appears not to be true in fluid mechanics or atmospheric
ored ic tion .
19
Numerous finite element formulations are currently
available. Strang (1973), Ncrrie (1973) and Zienkiewicz
11571} present detailed theoretical discussions of each. The
Galerkir. method , the most popular finite element method, is
described in detail below and used in the equation
formulation later.
A. FINITE ILZMINT CONCEPT
The problem of solving partial differential equations
can be specified in one of two ways. In the first, finite
difference methods specify the dependent variables at
certain grid points in space and time, and the derivatives
are evaluated using Taylor series approximations. Secondly,
the calculus of variation requires the minimization of a
functional over a domain, where a functional is defined as a
variational integral over the domain.
The calculus of variation approach creates a purely
physical model where the functional equivalent to the known
differential equations are lenown. Its major disadvantage is
that it limits the method only to those problems for which
fun'Uionals exist. Finite element methods, an extention of
this method. derive mathematical approximat icns directly
fror the differential equations governing the problem. The
advantage here is that it extends the method to a range of
problems for which a functional may not exist, or has not
been ii sc ove red .
2d
The finite eleirent rrethod divides the domain into
subdorains or finite elements (usually of the sarre form).
v.Mes are located along the boundary of the elements,
usually at the element vertices and at strategic positions
(mid! side, centroid, etc.) in the interior and on the sides
of faces of eletrents.
Commonly used elements are triangular, polygonal or
polyhedral in form for two-dimensional problems. The choice
of elements depends on the type of problem, the number of
elements desired, the accuracy required and the available
computing time. To begin with, the element must be able to
represent derivatives of up to the order required in the
solution procedure, and to guarantee continuous first
derivatives across the element boundaries to avoid
singularities. Triangular elements are errployed in this
thesis because they can be used effectively to represent
irregular boundaries, and/cr geometry, and also to
concentrate coordinate functions in those regions of the
domain where rapidly varying solutions are anticipated.
Consider the problem of solving approximately the
differential equation
L(u) = f(x II-l
where I is a differential operator, u the dependent
variable, and f(x) is a specified forcing function. Suppose
that II-l is to be solved in the dorrain a * x - b and that
21
appropriate boundary conditions are provided. The residual P.
is forced fror II— 1 as follows:
L(u) - f(x) = R II-2
The critical step is to select a trial fairily cf
approximate solutions (the members of a trial family are
often called basis functions). The basis function is
prescribed 'functionally) over the domain in a piecewise
fashion, element by element, and are generally a combination
of low order polynomials . A one dimensional example is
shown in Figure 1, wherein the domain (x axis) is divided
into six elements (line segments) A through F. The basis
functions are linear and one is shown for node 4 only in
Figure 1.m he function has a value of unity over node 4, and
decreases linearly to is zero at nodes 3 and 5 and zero
els swhere.
Consider a series of linearly independent basis
functions V. (x), as in Figure 1. Now u(x) can be
approximated with a finite series as follows:
u(x) = Z $. V. (x) = 0.V.\ J J J J
where £ is the coefficient of the jth basis function and hasj
a value equal to u at node j.
Substituting this approximate solution I 1-3 wherever u
appears in the differential equation 1 1 —
1
L^.V.) - f(x) = R
22
II-4
The best solution will be one which in some sense
reduces the residual R to a minimum value at all points in
the domain. By definition, the residual obtained using the
exact differential equation is identically zero everywhere.
The residual P., formed in equation II-4, is minimizied when
multiplied with a weighting function, integrated over the
domain and set equal to zero. This process is known as the
weighted residual method
b
RV dx = 3 II-5\
where W is the weighting function and is referred to as the
test function in the following development. The weighted
residual method minimizes the errors of the residuals, such
that the summation of all the positive and negative errors
add to zero
.
The ^alericin method, the most popular fini te element
method, is more general in application and is a special case
of the method of weighted residuals, as discussed by Pinder
ar.d Gray ^1977). The requirement imposed on the weighted
residual method forming the ^alerkin method is:
* the test (weighting) function be equal to the
basis (trial) function W = V . This process
leads in general to the best approximation of
the solution.
24
The final (Jalerkin form is obtained by substituting II-4
into II-5, yielding
-fi
±l((j}.l. )dx - Vf
1f (x)dx = II-S
If this procedure is repeated for N points in the domain
a system with N equations and N unknowns will be generated.
3. OAimiN APPLICATION
The following example taken in part from Haltiner and
Williams (1980) applies the Galerkin rethod to the advection
equation with linear elements
du du— + c — =2at dx
II-7
This equation is dependent in both time and space. The
treatment of time variation is important for most
meteorological prediction problems. The C-alerkin method is
not applied to the time dependence because it is more
convenient to use finite differences in time f as is done with
this example later. The same treatment is applied tc the
prognostic equations later, where two finite differencing
methods are employed to do the time integration.
The 7alerkin procedure represents the dependent variable
u'x.t) with a sum of functions that have the prescribed
25
spatial structure as in Figure 1. Approximate u(x,t) with
the finite series as follows
u(x f t) = I rf.(t)V .(i) = 0.?,1 = 1^ ^ ^ *J
II-8
where the coefficient . (t), a function of tirre, is the
scalar value of u at node j. The basis functions, V.(x), are
functions of space only and j equals 1 to ? for the example
In Figure 1. The repeated subscript in this forx implies a
sun- ever the repeated subscript.
The Galerkin equation for the advection equation II-7 is
obtained by setting L = c(d()/dl) and substituting in the
aooroximate solution II- 8 wherever u is found.
N
1j=i at
*0Am J
b
r N
V.V. dx + c I (. —JV,
dx]
II-9
where i = 1 to N v V. the test function and V.the basis
function. The domain of integration is given by a * x - b,
and the integration is done in a piecewise fashion, element
by element
.
In this one-dimensional case, an equation like 1 1—9 is
written for each node, i. Considering node 4, what are the
possible non-zero contributions from equation 1 1—9? Figure
2 illustrates the basis and the test function interaction
luring the piecewise integration process. From the
25
j - 1
i
Basis Function
s
Test Function
^t.
J = 2
1A2B3C^D5E6F7
J - 3
3 C ^
= 4
J - 5
= 6
C ^ D
,'%
4 D 5
J - 7
1 A 2B 3 C
^ D 5 E 6 F 7
Figure 2 . Basis and test function interaction
during the piecewise integration process
.
27
definition of the basis and the test function, locally
defined as unity at node j and linearly decreasing to zero
; ± 1 and zero elsewhere, the only non—zero
contributions are made when * = 3 over element C, j - 4 ever
elements ~ and ?, and j = 5 over element D.
The evaluation of 1 1-9 for i = m, which is given in
Haltinep and Williams (1981), leads to the equation:
6 dtm+1
4u + u1 ) + e ( u
1- u m - ) = 11-10
in m-1 m+1 m-12-*x
The boundary points, which in this example are nodes 1
and 7, are evaluated in the same way as the interior nodes,
with the exception that cyclic conditions are imposed.
The time discretization of II—10 is d-one using- a finite
difference scheme. Applying leapfrog time differencing gives
the following equation
f
12 At m+1 m+1 m m
n+1m-1
n-1 ,
m-l
Hi m+1) = n-ii
The resultant equation set, in matrix form, contains an
NxN matrix where N is the number of nodes.
The transition from one-dimension to two is
mathematically identical. The domain is now subdivided into
finite areas, which are triangles in this implementation and
28
the basis functions are linear. However, now they are
pyrarid shaped with value unity at the center and decrease
to zero at the surrounding nodes, and are zero elsewhere.
Figure 2 shows this basis function for node 28 outlined in
heavy black. The value at any node again can be approximated
by II-3, where j ranges over all nodes connected to node i
including i itself. The connectivity for node i = 28 in
Figure 2 is j = 15,16,27,28,29,39 and 40.
The integration is still over the entire domain. With
both the basis and the test function zero over the domain,
except locally over each element, the global integration can
be performed by integrating locally over each element. By
definition, this integration can be expressed as an inner
product of both functions (i.e. basis, test) as follows:
<w =re
))A
V . V, dA
Using this definition and the repeated
notation eauation II—9 becomes
11-12
subscript
»-<?„?!> c iM*.X<> = 11-13
where the dot implies differentiation with respect to time,
and the second subscript implies differentiation with
respect to the second subscript. The local integration may
be calculated directly from exact expressions derived from
area coordinates described in detail in the next subsection.
C6
Figure 3. Basis function for node 28. The
shaded area is the complete basis function
and the V. , where j = 15, 16,27,28,29,39,^0
are jth node basis functions for node 28. The
dashed line at node 28 has length unity.
30
In sunwary, the Galerkin procedure involves subdividing
the domain into finite elements, approximating the dependent
variables by a linear combination of low order polynomials
a~A substituting them into the equations. The equation is
multiplied by a test function, integrated over the entire
domain and finally the resulting system of equations is
solved .
C. APIA COORDINATES
While the Cartesian coordinate system is the natural
choice of coordinates for most two dimensional problems, it
is not convenient when working with triangularly shaped
elements. It is therefore necessary to define a special set
of normalized coordinates for a triangle. Area, or natural
coordinates as they are commonly called, reduce the
formidable task of integrating products between the basis
avA test functions and their derivatives over a triangular
element and result in easily computable and exact
expressions .
The following development is taken in part from the
formulation by Zienkiewics (1971). Consider the triangular
element illustrated in Figure 4. There is a one-to-one
correspondence between the Cartesian coordinates (X,Y) and
the area coordinates (L^Lg.L-a ) for the element. Let A
denote the area of the triangle and Aj_ , ^and A^ the areas of
the svbtriangles in Figure 4 such that A = A^+Ag+A,.
31
2 (X2,Y2 )
(0,1,0)
P(X,Y) or
Cartesian vs. area coordinates
YA
W 71
a3
" X2 " X
l'
W y2
Fig. 5 • Transformation to area coordinates
32
The relationship between a point P(X,I) in Cartesian
rcordinates and ?(L , L , L_ ) in area coordinates can be
seen by the following transformations
X = LjX + L2X * L-X
Y = L1Y + L^Y + L
3Y
1 = Ll
+ L2
+ L3
11-14
where .-* L- = - and* A
andL1
= f2A + b1X + a
1Y) /2A
L2
= (2A + b2X + a
2Y) /2A
L3
= '2A + b3X + a
3Y) /2A
11-15
where 2A is twice the area of the triangle and the a's and
b's are defined as in Figure 5.
It is worth noting that every tuple (L- , L2
, L-» )
corresponds to a unique pair fX,Y) of Cartesian coordinates.
In Figure 4, L- a 1 at vertex 1 and at vertices 2 and 3. A
linear relation exists between the area and Cartesian
coordinates which iirplies that values for L.vary linearly
over the triangle with a value one at vertex 1 and a value
of zero at vertices 2 and 3; and similarly for L2and L-
.
This demonstrates how each component in the tuple (L^, L2 , LJ
behaves over the triangle as do the linear basis and test
functions over the element, as was seen in Figure 4. Clearly
= 7. 11-16
33
where V. is a linear function of the Cartesian coordinates
(i.e. "basis , test ) .
Zienkievicz (1971) shows that it is possible to
integrate any polynomial in area coordinates using the
sirole relationship
~ 1 ^2 ^3 dxd y=
m ! n ! o !
A( m. + n + p + 2 )
!
2A 11-17
where rrf a and p are positive integers and A is the
elementary area. For an example of this integration
technia.ue using inner product notation, equation 11-12 is
evaluated as follows
V^ didy =2! 0! 0!
2A = -
(2 + + 2)! 6
<vv-<f 1! 1! 0!
V.V, dxdy =J
(1 + 1 + +
A2A = —
2)! 12
i = J
i t J
11-18
The differential operations in area coordinates follow
directly from the differentiation of (11-15) where
ar.t
b 3 b. o
bx i=l 2A dL4
b 3 * a
'by i=l 2A bL.
11-19
11-20
34
As explained earlier (see Equation 11-16), Viis a linear
function 'i.e. basis, test) which equals a component L1
of
the srea coordinate tuole. Therefore
5 =if 1 ^ j
1 if i =j
11-21
Consequently BV. dx
jx
dv.
dx
h. I! 1
2 A dl<
for j = 1 is
2A11-22
Is an exarrple, consider the inner product <V . ,v.> at
vertices j = 2, i = 1. This integration is evaluated as
<72x»V =(f b
y > 2A7
1±xdy
11-23
.!*1! 0!
2A = -
2A (1+0+0+2)! 6
Therefore any inner product in the formulation can be
readily evaluated using area coordinates. Another benefit of
using this coordinate system is that all of the inner
products are functions of space only and need be computed
only once.
i><j
III. SHALLOW WATIR ^Clil
The governing equations for this model are derived by
raking several simplifying assumptions on the primitive
equations of motion, whish then give the barctropic shallow
water equations. However, as mentioned previously, the
shallow water equations describe many significant features
of the large-scale motion of the atmosphere, and therefore
have been used in numerous experiments over the years.
The vcrticity-divergence form of the equations has
several advantages. Williams (1981) has shown that the
frecstrophic adjustment process is treated much better with
the vcrticity divergence formulation than with a direct
treatment of the primitive form of the shallow water
equations, such as was used by Kelley (1976). This
formulation also allows the~ velocity components and the
height to be carried at the same nodal points, whereas the
best scheme for the primitive form of the equations requires
staggering of the fields, as seen in Schoenstadt (1980). The
orticity divergence form of the equations is also
convenient for the application of semi-implicit
differencing, which saves considerable computer time.
A. SOYIRNING EQUATIONS
The primitive form of the shallow water equations in
Tirtesiac coordinates is
dt dx dy
du 30 dK
dt ox dx
dt dy dy
III-l
III-2
III-3
equation (III-l) is the continuity equation and the III—2
ani 1 1 1-3 are the momentum equations, respectively. The
variables are defined as follows:
0,1
i
i
t
c
5
the spatial coordinates o* the domain
components of the wind vector
geopotential = (gravity x free surface height)
2 2mean geopotential = 49,000 meters /seconds
time
kinetic energy
absolute vorticity = (-9 + fj
relative vorticity
ccriolis force (mid-channel f -plane)
divergence
The shallow water equations can
vorticity divergence form as follows:
oe written in
3?
— + B(j) =
dt
dt
— T2^ =
&t
Ox(^u J
-
—(uQ)dx
d— (vC)dx
— (0v)By
— (vC)oy
d— (uQ)By
v2K
III-4
III-5
III-6
where III -4 is the sare continuity equation as III—1 , I I 1—5
is the vorticity equation and III— 6 is the divergence
equation .
3ecause of the vorticity divergence fortr of the
equations, it becomes necessary to solve the tirre dependent
variables "S and I in terms of 4/, the strearr function
(rotational part of the wind), and*X, the velocity potential
(divergent part of the wind). The initial fields for the
model will be in terms of ¥, TL and $,
The following diagnostic relationships are defined and
used later in the solution of the equation set.
u = "V*x> III-7
III-8
where the subscript implies differentiation,
2 2u * V
K = kinetic energy, I II—9
uC = u(* fj, 111-10
vC = v(S + fj , III-ll
Of = 0u, III 12
P = pv, I I 1-13
38
-8 - v2 V, 111-14
D - v2*. 111-15
3. IOUATION FORMULATION
The G-alerkin method described in Chapter II is now
applied to eauations III-4 through 111-15. For ease of
comprehension » the shorthand inner product notation as in
11-12 will be used to simplify the equations. The detailed
Galerkir. formulation will be shown for equation III-?, the u
component of motion. The method follows directly from the
example in Chapter II of this thesis, which in turn follows
in part from £elley (1976) and Ealtiner and Williams (1981).
Consider equation I I 1-7 and assume that each variable u,
^ and 1 is approximated by
- - W* = ¥. v. , in-ie
where the repeated subscripts indicate summation over the
range of the subscript. Substituting these approximate
solutions into 1 1 1-7 yields
»T =* ~^J.) + —(* 7.) 1 1 1-17
Since only the basis function 7. is a function of space,
TII-17 may be further simplified by factoring out the time
dependent coefficients.
The next step requires multiplying by a test function V.
as discussed in Chapter II, and integrating over the area
domain
( av.—J V, dAay
iv. v, dA = - ¥
A
J J
(( 5V.—JV, dA
dxx
III-1S
'he final form in inner product notation is
<uj
Tj- T i>
= " «J 7JT'
T1> + <V±C' Ti> nI - 1SJ JX
where the double subscript implies differentiating with
respect to the second subscript.
The three prognostic equations (III-4, III—5 and III-6)
are similarly advanced using the Galerkin technique to
become, respectively:
<Vr Ti>
+ l <rjW =- <^vjx« v i> ' <*jTjrV IXI " 20
<tjfj.fi> = -<(ug).V vi>-<(vc).V v
i>in-2i
,2,<r.V. f 7, > - <0.V*V .,7,> = <(vQ).V. ,V,> - <(uC) .V . ,T, >J J i *J J i J Jx i J jy i
+ <£.v2 7 .,V. >J J i
TT T-??
where v is the Laplacian operator and the dot implies
differentiation with respect to the time dependence in
III-4, III- 5 and III-6.
Similarly, Galerkin equations are formulated for
equations III-7 through 1 11-15
.
40
C. TIMZ riSCP.ETIZATION
The equation set 1 1 1-20, 111-21 and 111-22 is arranged
so that all the tern's on the left hand side can "be treated
Implicitly, and all the terms on the right hand side can he
treated explicitly. The explicit time integration will be
done by the leapfrog difference method. To start the time
integration, two forward half steps are taken, after which
the full leapfrog scheme is used for the remainder of the
forecast period.
The vorticity equation III- 21 is solved independently
frcr 111-20 or 111-22. However, 111-20 and 111-22
'continuity and divergence equations, respectively) are
coupled. To explicitly solve either, decoupling of the
equations is necessary. In this thesis this is done through
algebraic substitution of II 1-22 (solved for D(n+1)) into
IIT-20. Once the time integration is performed on 111-20,
111-22 can he solved 'or E(n+1) using the i (n+1) value.
The final prediction equations are
- [3EHT]n+1
- [BEEYJ°~1
" 2E j[< T jx- V ix>+ < V jy Viy>]
111-23
whe-e A = 4/(2*t) f E = A/J f C = B/(2*t) and [BERY] is the
geostrcphic boundary contribution, see Section 3.
.n+1v?j.f1 >
n-1*j <V
j,Vi >
J3(^Wi >] I I 1-24
rj
T'j' 7i'- D
n
J1<7jf 7i > - (At/2)[0
n
j
+1<v
2v
j,V
i>
* ^S'V * 2(TQ)3<fja,? 1 >
111-25
After these three elliptic equations are solved, the
history of the variables I I 1-7 through 111-15 is updated.
A large time step can he applied to this forrr of the
shallow water equations due to the semi-implicit nature of
the equations, -his is very important since finite element
methods generally require more computer time per time step.
The vcrticity-divergence formulation acts as a filter, which
slows down the high frequency waves in the solution. The
two-dimensional advective stability criterion for a linear
element, derived by Cullen (1973), was used to determine the
correct time step,
4 x
^t =
Id i/e
111-26
whereat is the time step in seconds, ax the shortest grid
spacing in meters and c the fastest phase velocity.
42
D. COMPUTATIONAL TECHNIQUES
The final prognostic equation set requires the solution
of a Felmholtz equation for $ and Poisson equations for ¥
and J.. The rrost common method of solution used by
meteorologists has been the successive over relaxation
method (SOH) in which an initial guess of the solution is
trade and then progressively improved until an acceptable
level of accuracy is reached. SOH is employed in the
solution of the equations, where 111-23 can be represented by
v2[M]{x} - C[M]{x} * {b} 111-27
and III 24, III 25 by
v2[M]{x} = {b} 1 1 1-28
2where v the Laplacian operator, [M] = <V-,V.> matrix, {x}
- the dependent variable in vector notation, C - constant as
in III -23 and {b} the right hand side of the equation or the
forcing function.
The mass matrix [M] , dimensioned (nxn), is a matrix of
coefficients whose rows are the equations of the system to
be solved. There exists a one to one correspondence between
the rows of the mass matrix and the nodes of the domain.
Each equation has a term (column) for each node, where a
non-zero term represents connectivity. Nodes are connected
if they are both vertices of the same element. Obviously [M]
is a sparse matrix containing the inner products for the
43
left hand side. Chapter 4 of this thesis will describe the
matrix compaction procedure.
The forcing function {b}, dimensioned (nil), involves
only variables at the current time step and is easily
computed using four very versatile subroutines described in
detail in the next chapter.
The initial guess to start SOR is the previous time step
solution. An average of 30 passes per equation are needed
for each time step. The solution is considered to have
converged to its final value when the residual for each node
has beer, reduced to some acceptably small value.
The diagnostic equations III-7 through 111-15 must also
be solved every time step. Eowever, the same technique is
not used for these equations. Dr. V .J.P. Cullen suggested an
ur.de- relaxation scheme for which three passes over the
domain should produce a solution of acceptable accuracy,
since the coefficient matrix is so strongly diagonally
dominant. .
yass lumping of the coefficient matrix is used for
the first guess. This technique requires replacing the mass
mat-ix [«] by the identity matrix [I] . A first guess of this
type i? *ble to describe most of the large scale features,
whi~h in turn reduces the number of iterative passes ever
the field. Successive passes converge to solutions which
describe smaller scale motion, approximately to the same
order of magnitude as introduced by computational error, so
that further iterations are not needed.
44
I. GRID GIOPSTRT
The dor-air. of this trodel is a cylindrical channel, with
total length of 4245 Km and width of 3503 Km. The channel
simulates a belt around the earth and it proves to be an
excellent test bed for comparing with the finite element
formulations used by Selley (1576) and Older (1981).
The domain is subdivided into' equilateral triangles as
shown in Figure 6. Post of the test runs for this thesis use
a 12x12 mesh which has 156 nodes and 289 elements. This
implementation is not restricted to one grid pattern. The
technique developed by Older (1931) to vary the nodal
geometry smoothly to achieve areas of denser and coarser
resolution is also implemented, as in a third grid pattern
that varies the nodal geometry abruptly. A short discussion
of these nodal geometries with accompaning illustrations of
each is presented in Chapter VII, where the different test
cases are described.
Cyclic continuity is assured in the x direction by
wrapping the domain around the earth to form a cylindrical
domain. This has the advantage of eliminating the east-west
boundaries and it simulates the flow around the earth. The
only boundaries en this domain are the north-south walls and
their treatment will be discussed shortly.
45
F. INITIAL CONDITIONS
As mentioned previously, the reformulation of the
governing equations into the vorticity-divergence
shallow water equation set requires solving the tirre
dependent variables in terms of the stream function and
velocity potential. The continuity equation is not altered,
so that its solution is expressed in terms of i.
For the basic testing of the model's performance, simple
analytic sinusoidal initial conditions are used to insure
the most accurate analysis possible and to simplify the
comparisons .
The sinusoidal initial fields are graphically shown in
Figure 7 as 3-dimensional surfaces. The geopotential field
consist of a half sine wave in the y direction and a single
cosine wave in the x direction. The stream function ¥,
calculated by dividing the geopotential field by the
coriolis force, has the same physical structure as 0. The
velocity potential 1 has a single sine wave in the x
direction and a half sine wave in the y direction.
These initial conditions are computed as follows
= fQ Asinoc1cosoc
2- f U(y - y^) - $
¥ = 0/f o111-29
1 = CsinocjSinop quasi-geostrophic divergence
where A - arbitrary amplitude
& - coriolis value for mid-channel latitude
47
3
mean flow
m id-la t i tude value of y
| - rear, free geopotential height
= 49,200 m2/s
2
2^xr/L
wave number
channel width
*i-
*2
r
V
L - channel length
- -(f Uoc?BA)/(f§ + $3)
*1 + *2
3. 5CUNEARY CONDITIONS
3oundary conditions are only required on the north and
south walls of the grid domain. Eue to cyclic continuity,
the domain is wrapped around creating a cylinder eliminating
the east and west boundaries. However, careful attention to
detail is needed during the implementation to assure this
continuity. Separate boundary conditions are applied to each
of the predictor equations 111-23, 111-24 and 111-25. These
conditions are computed for the wall nodes only and are
applied during each pass through the relaxation scheme.
The vorticity eauation 111-24, the most sensitive of the
predictor equations to solve, requires ^ on the north-south
boundaries to remain constant for the entire forecast
period. Since this equation is solved in terms of ¥, the
Initial north-south ¥ values are saved and assigned to the
49
"boundary points after each pass through the relaxation
subroutine.
-he oroper boundary condition for the divergence
equation III —25 would be dX/dn = C. However, for the purpose
of this study, there is more interest in the sinusoidal
variation in the y direction and not in the region of the
walls. Therefore I = 2 is appropriate.
The continuity equation 111-23, the most complex
predictor equation, requires that there be no mass flux
through the north-south walls. The geostrophic boundary
condition
— = -uf 111-30dy
is aoplied to the north south boundary nodes for the terms
[BEET] in equation I 11-23 . Integrating the inner product
<£ • V.,V. > by Darts produces the boundary terms
If v2 (0.V )V. dxdy = f / * • (* $ .V. ) V .dxdy
yx J J yx J J
= \[ [v.(Viv(^V
j)) - v^V^-vV^ dxdy
= i V. v(0 .7. )-S ir - // W. .v(^.V, ) dxdy
= r3ERTl - ^[<v jx .v ix > - <Vjy
.V ly >] 111-31
where n is a unit vector normal to the domain and dr is the
differential distance along the path of integration on the
perimeter of the domain.
52
"he geostrophic boundary condition 111-38 is substituted
into the contour integral in equation 1 1 1-31 and put into
laleriir forrr, in the sare way as in the one-dimensional
advective equation in Chapter II. The resulting term is
derived as follows
i V.v'0.7.).n dr = & —(#.V.) V, dxoy
v*x(u . . + 2u. u. . ), 111-32
3 J+l J J-l i
Equation 1 1 1-32 appears twice in the continuity equation
III 23, for *ime levels (n+1) and (n-1). All values of u are
known "or time (n-1), since they are saved from the previous
calculations. Fowever, u(n+l) has not been computed. To
solve for u(n+l), both¥(n+l) and T(n+1) are needed, ^(n+l)
is solved first from the vorticity equation. 1(n+l) needs
0(?»+l) as part of its solution and 0(n+l) needs u(n+l) in
its solution. To avoid this problem, it is assumed that
X(b+1) has a negligible contribution to the solution of
ufn+1) and only ¥( n*l ) is used.
51
iv . gcrruTiR ieplimintation
The formulation and general theory of the finite element
method was presented in the previous chapters. The objective
In this chapter is to discuss sere Important computational
aspects pertaining to the implementation of the finite
element prediction system.
The nain advantage that the finite element method has
over other prediction techniques is its generality.
Conceptually, it seems possible by using many elements, to
approximate virtually any surface with complex boundaries
iz 1! initial conditions to such a degree that an accurate
solution can be obtained. In practice, however, obvious
engineering limitations arise, a most important one being
the cost of the computation. As the number of elements is
increased, a larger amount of computer time is required for
a forecast. Furthermore, the limitations of the program and
the computer may prevent the use of a large number of
elements. These limitations may be due to the computer speed
and stcrage availability, or round-off errors propagated in
the computations because of finite precision arithmetic.
Also, the malfunction of a hardware component, if the
prediction is carried out using many computer hours to
execute, can be a serious problem. It is therefore desirable
to use efficient finite element programs.
KO
The effectiveness of a program depends essentially or.
the following factors. Firstly, the use of efficient finite
element techniques is important. Secondly, efficient
programming methods and sophisticated use of the available
computer hardware and software are important. The third very
iirportfint aspect in the development of a finite element
oro^ram is the use of appropriate numerical techniques.
The vcrtici ty-divergence model described in the previous
chapter is implemented on the 13m" 3333 computer located at
the Naval Postgraduate School. Some notable features of its
architecture are the three trillion bytes of virtual mass
storage, of which eight mega bytes are available to each
user, and the c7 nanosecond machine cycle time. The model is
executed rostly using a 12x12 element domain requiring 4201c
bytes of storage and 30 seconds of CPU time to execute.
Irceeding execution time and/or available storage is not a
problem, in fact the system allowed a lot of flexibility
during the implementation phase of the model.
The source code is written using FORTRAN IV and compiled
on an optimizing Fortran H compiler. Appendix A contains the
source cede listing, which is divided into five subdivisions
delineating the logical structure of the program.
A. PROGRAM ARCEIT£CTURZ
Program features incorporated in the model are:
1) Modularity. With only a few exceptions, each
module is limited to one page of FORTRAN code. This makes it
53
easier to comprehend the program. Zach module performs only
o-^e task. For example, subroutine CONTIC computes the value
jf the forcing terms for the continuity equation. likewise
there is also one module for the divergence and vcrticity
equations. To implement a new set of equations, only these
modules would have to be altered.
"asily controllable switches. Switches may be set
to either print, plot or tabulate harmonic analysis data for
most of the available fields. The ability to display
intermediate results allows each portion of the algorithm to
be monitored for computational adjustments. This also makes
it easier for unfamiliar users to become acquinated with the
computational model.
3 ^ Forcing term subroutines. In previous
irpiementaticns • each forcing term was calculated by a
special subroutine. In this implementation, the calculations
are accomplished by general purpose routines which simplify
the implementation of the complex prognostic equations:. Hits;
allows implementation of different equation sets (i.e.
^aroclinic y odel) over the same domain with minimal effort.
4< T cementation . Zach variable is defined by a
short phrase 'Appendix A, A.). The function of each module
is described in an introductory paragraph. Shared data is
placed in named common blocks and identified with each
subroutine which uses them. A subroutine index is given.
1 . "air, Program
The main program is short, calling only six modules
which reflect the basic sequential flow of the model. It
starts with initialization of all model parameters (i.e.
model options, domain, finite element arrays, inner
products). It then initializes the input fields (i.e.
secpctential heights, stream function and velocity
potential) and is followed by initialization of all
remaining dependent variables. At this time the model is
totally initialized and time integration begins. As
mentioned previously two techniques are employed for time
integration, each having its own module. Upon completion of
the last forecast, the program terminates.
Arrays are the only data structures used and are
grouped using 19 different common blocks. Several arrays are
use! as static link lists, as described in detail later,
which simplified the algorithms. The common block format has
the advantage of reducing the overall execution time of the
program. Most of the arguments passed during a call to a
subroutine are contained in conmon. This requires less time
to execute since no parameter passing is required for the
arguments. Another benefit of this format is that the code
becomes less cumbersome and more readable. Each variable and
array is defined in the first subsection of Appendix A along
with a oage index for the subroutines.
2. Initialisation Phase
Appendix A, Section C contains all the subroutines
used during the initialization phase of the program. Frorr
the user point of view, the most important suoroutine is
INI-S3, the first subroutine called, which contains all the
global variables that control the different options
available per run. This is the only subroutine that is
changed to run the different experiments, assuming that no
new computational technique is introduced. The selection of
options ere:
1) - channel location - the channel may be
shifted north or south by presetting the
north/south latitude limits in INITG3.
2) - variable geometry - the node positions may
be grouped for more dense node patterns to
yield higher resolution. Two variables HI
and H2 set the ratio used to vary the
spacing -along the x .and y. axes,
respectively.
3) - initial field wave length and amplitude can
be altered to produce various effects.
4) change the initial mean flow.
5) - diffusion can be entered for any of the
three prognostic equations.
6) - maximum length of forecast period may be
changed and a print, plot or harmonic
c p.
analysis of any dependent variable may be
requested for any tire interval.
Or.ce the experiment is determined, the options
listed above are set. The program is ready to be executed.
The largest part of the initialization phase
consists of establishing the domain and producing ail the
finite elerent computational vectors that remain constant
throughout the experiment.
"he first several steps in setting up the domain are
concerned with indexing. Subroutine COF.RES is called first,
where all the nodes (grid points) and elements (triangular
areas' 1 are numbered consecutively starting at the southwest
corner of the domain and moving eastward across each row or
latitude. lor earch element, a record- -of all of its nodes
(ertlces) *re stored in array ELMENT (M,2), where M is the
total number of elements. To facilitate the inner product
evaluation later, a local numbering system is required for
each element. That is, for each element, its nodes are
stored counterclockwise in a positive sense. The first node
however, is arbitrary.
With the domain divided and numbered, a connectivity
list ( the correspondence between each node and the neighbor
nodes) is constructed for each node by subroutine CORRIL.
lach node is adjacent to four or six other nodes depending
on whether it is a boundary or interior node, respectively.
These adjacent nodes, plus itself, make up the connectivity
57
list for one node. The connectivity lists are then
concatenated sequentially starting with the first nodes
connectivity list into the vector NAME (NN), where NN is the
sur of the nodes in each connectivity list. (i.e. for a
12x12 dorain with 156 nodes, and equilateral elements NN =
1044). 3or the first tiire during the initialization phase,
special attention is given to cyclic continuity. As
discussed earlier, cyclic continuity is the joining of the
east and west "boundaries to create a cylindrical channel.
The connectivity list for these east/west boundary nodes
must be complete to insure proper continuity for the
calculations later.
The connectivity vector NAM is frequently used
during most computations. Two utility vectors ISTART
(containing the starting location in NAM for a particular
node' and MUM (containing the number of nodes in its
connectivity list) are used to locate and indei through the
vector NAMI, as will be seen shortly. This same technique is
used to index through the coefficient matrices and used
during most of the node interaction computations.
The physical properties of the channel are
calculated next in subroutine CHANAL. Here the north and
south latitude boundaries, which were pre-set in INITG£ by
the user, are used to compute the grid spacing along the x
and y axis. Since this channel simulates a belt around the
earth, the magnitudes of both LIITAX and DELTAY (meters) are
£3
proportional to the width of the channel divided by the
number of ,?rid points in the y axis.
The Cartesian coordinates for each node are computed
by subroutine LCCATI using the DILTAX and EILTAY calculated
in CEANAL. If the option to use varying grid geometry is
i-sired, subroutine TRANS transforms the grid geometry.
TRANS also computes the corresponding new Cartesian
coordinate values for each prid point and calculates the
minimum E5ITAX and IELTAY within the domain. When the
secmet-y is changed to create a smaller DELTAX or DELTA!,
the two dimensional advective stability criteria is also
charged. A new time step DT has to be computed using
equation II 1-26. Since TRANS transforms the geometry, it
also computes the new CT
.
Another transformation is required as discussed in
Chapter II. The transformation from Cartesian coordinates to
area coordinates is needed to perform the area integration
of the inner products. Sub-routine AREA computes these
transformations exactly as outlined in Chapter II, Section
C. Again cyclic continuity is very important and special
care is needed to insure proper transformation.
Following the area transformation is the computation
of all the inner products that are required to solve the
equations. The advantage of using area coordinates is that
the inner products (function of space coordinates only) are
computed and stored once and used repeatedly without
53
recalculation- Subroutine INN2P. computes and stores these
products using the formulas derived In Chapter II.
The 3oefficlent matrix, dimensioned NxN , where N is
the total number of nodes, is a matrix of coefficients whose
rows are the equations of the system to he solved. As
discussed In the computational technique section, the
members of this sparse matrix are the inner products for the
left hand sides of the equations. Three coefficient matrices
are used in the solution of the equations. The diagnostic
equations (III-? through 111-15) use a coefficient matrix
with the inner product <V ,,L > which is constructed by
subroutine AMTHXl and stored in compacted form in vector
GfNN) by subroutine ASEMBL. Eowever, when solving the
prognostic equations, these coefficient matrices have a DT
ftlire step in seconds) term, so that these matrices are not
assembled until the time integration begins. The vorticity
and divergence equations (111-24,111-25) use the coefficient
matrix H(N'N) with inner products < vjx
.? lx > + <vjy*\y> in
solving the ?oisson equations for the stream function and
velocity potential, respectively. The continuity equation
[111-23) uses a combination of inner products in its
coefficient matrix J(N'N) as follows
jx» v ix^ V¥jy *
v iy'J(At)
<vj,v
i >iv-i
60
to solve the Helirholtz equation. At the start of each tirre
Integration module, subroutine AMTHX2 is called to construct
the two coefficient matrices E and F.
These banded and sparse rratrices are ccrrpacted into
vectors to save storage during their assemblage by ASEM3L.
The vectors are dimensioned NN, as is NAME, the connectivity
vector, and both use ISTART and NUM to index through them.
This compaction routine was used by Kelley (1976) and Older
'1981) in their models, but was developed by Hinsman (1975).
To illustrate matrix assemblage using an element by
element technique, consider Figure 8. Note that this
illustration is for element number 3, but all elements are
treated in a similar maner. The computational technique
required that for each point (node) describing element 3,
namely nodes Z, 3 and 14 stored in array ELMENT, the inner
nro-uct <TT .,7 J
> between those points be distributed to theirJ i
proper location in the coefficient matrix.
Subroutine AMTRI1 builds the inner, product nodal
interaction and stores it in matrix 3, dimensioned 3x3.
Figure 6 illustrates the 3 matrix for element 3, where the
inner product <V . ,V.> is the multiplicand of the
corresponding basis and test functions, respectively.
The local dispensing of interactions is done in
ASZ VBL. Consider the second row of [3] in Figure 8. These
are the interactions between node 3 of element 3 to the test
61
AMTRX - builds J3J ^OT one element, passes [3] and the element number
associated with [2] to ASEK3L. This process continues till all element
node interactions are assembled in the coefficient matrix.
H,
V2
V3V2
V3
V3V3
V14
V2
V1V3
Vl4
Yl*V14V14
ELMJ-'BR
ASEMBL - assembles the node interactions into the coefficient matrix
[Cj , which has the same structure as HAKE. The following diagram assembles
inner product '•jj. into the coefficient matrix [c] , for element J.
EIKENT
1
1
2 3
1 1 2 3
2 2 14 13
ZLKNBR -» 3 2 ill*]
4 3 15 1
1^
"rH287 144 145 |156
288 Vtk 133 1*5
ISTART
1 1
2 6
3 ©*-^
4 16^
5 21
6 26
7 31
155 1035
156 1040
KUM
5 12
SH
ASZTBL pseudo-code
-do i - 1 -• 3 2
11 - SLMENT(l) 1 3
u-do j - START(ii) -* LAST(li)
jj - NAME(j)
-do k • l-»3 2
kk - 2LMEKT(k) 1 3
if ( kk - jj )
14
then
C(J) - G(j) * B<i,k)
else continue
L end do
-end do
-end do
sg
^s-
5 151
5 152
5 153
5 154
5 155
5 156
LAST (3)
NAME
1 1
2
LAST (3) 13
12
24
6 2
1
13
14
10 3
3) -m 3
2
13 14
15
3) *15 4
4
3
15
16
20 5
143
155
1043 144
1044 145
C(l)
:(z)
2(3:
:(io)
c(n)
liizi
10
11
c(i3):+ y^:(14)
c(l5)
c(l6)
0(20:
15
16
G(1042)
C(1043)
C(1C44)
20
1043
1044
Figure 8. Assembling and storing the coefficient matrix for element 3.
62
functions. AS1KBI locates nodes 3's connectivity list in
NAM! using ISTARI and NUM. In Figure 8, this list is
delineated ty START(3) and LAST(3). Now ASEMBL steps through
the connectivity list for three iterations. Turing each
pass. ASI.V 3L is searching for one of the three node numbers
for element 3. When a match is found with one of element 3's
nodes 'i.e. 2,3 or 14) and node 3's connectivity list (i.e.
3,2,14,15 or 4) the proper position, to which this
interaction is to he added has been found in the coefficient
matrix. Since MAPE and vector C, the compacted coefficient
rat Mi, are dimensioned identically, the same pointer (i.e.
J in Figure 3) is used to index through both arrays. This
procedure is repeated for all elements in the domain to
assemble the coefficient matrix of the equations. The
pseudo-code for ASEMBL is shewn in Figure 8 to facilitate
stepping through this exarrple.
The domain and all finite element work: vectors are
initialized at this point. Subroutine ER-MS3T is called- .later
to compute interpolation points for the harmonic analysis
subroutines.
The last phase of the initialization process is the
initialization of the dependent variables. The three input
fields geopotentlal heights, stream function and velocity
potential are computed in subroutine IC using the equation
set II 1-30 . However, the variables calculated from the
diagnostic equations have to be computed using the input
fields. These varieties are used during each tiire step while
solving the prognostic equations.
The diagnostic equations are solved in subroutine
DIPTAB, first during the initialization phase and later
during the time integration phase. Each diagnostic equation
calls its own module to compute the value of the forcing
function and stores the computed values in the vector RES.
These equations all use the same coefficient matrix when
solving the diagnostic equations. Subroutine SOLVER is
sufficiently genereal to solve each equation. SOLVER uses
vector RES and coefficient matrix G to under-relax towards
the solution. As mentioned previously, the coefficient
matrix is strongly diagonally dominant so that three passes
over the domain are sufficient. At the end of rEFVAR, output
is generated depending on what print, plct, or harmonic
analysis controls were preset.
This completes the initialization pnase of the model
and the program, for the forecast phase will he described
next
.
2 . Forecast Phase
The forecast phase is accomplished in two steps. The
first time step is made using two half steps by subroutine
r*£TZNO. Eere the prognostic coefficient matrices are
constructed using half the ET value by calling AMTRX2.
Ar"TFX~ uses the same computational technique to construct
the ccefficient matrices as described for AMTRX1
.
64
lach of the prognostic eauations 111-23, 111-24 and
111-25 calls its own subroutine (CONTIC, VORTEQ and TIVEC
respectively' to compute all the terms on the right hand
side, which are stored in the vector HHS. After computations
for RES are completed, subroutine RELAX solves the equations
by over-relaxation as described in the computational
technique in Chapter 3. Once the solutions for the (n+1)
time step are completed, DEPAB is called to update the
variables from the diagnostic equations. Two oasses through
rtATZNO advances the solution fields one time step.
The remainder of the forecast period is integrated
using the leapfrog scheme. Subroutine L2APFR performs this
integration using the identical format as vATZNO, except
that IT equals two DT. At preset times as specified in
INITG5, the different fields are saved for printing. This
process continues until the final forecast time is reached.
B. UTILITY YOTULZS
Ir.ce the equation formulation is completed, as in
Chapter III, all the inner products and types of
integrations are inown. Versatile modules can oe written to
per^crm these computations. Consider a term of general form
<^A . V . ,V . > where i is the node about which the term is
evaluated and the j's are the nodes connected to node i, or
the surrounding nodes. The inner product values <V.,V > are
already computed and stored for all the nodes, during the
Initialization nhase of the model.
65
-he regaining computation to complete trie evaluation of
this term is the mult iplicat ion of the scalar coefficient of
A it node 1 with the corresponding inner product <V.,V. > for
node «. This requires indexing through node i 's connectivity
list stored in vector NAME, and for each node in the list
irnltiply and total the products. The cumulative sum of these
multiplications is assigned as node i 's contribution for
this term. Subroutine TER V 3 performs this exact computation.
All that is passed to 7ERM3 is the scalar field A and the
sign of the inner product, TEBM3 then computes the
contribution for each node in the domain and accumulates it
in the wcric vector RES.
Three other utility modules are; TERM1, which computes
the first scalar multiplication for triple inner products
'i.e. /Aj Vj3k Vk ,Vi >) . The product <V.V
k,V£
f is again already
computed and stored by subroutine INNER. TERM1 computes <\ ?.
V k ,7 i > and constructs a compacted vector similar to the
coefficient matrices. This reduces the effort of multiplying
the second scalar to a TERf3 computation. TZRP2 computes
node interaction of the following type <Aj Vjxt 7^x> , where both
the basis and the test functions are derivatives. Lastly,
subroutine TIRK4 computes node interaction for terms as <A.V.3 Jx
,v\ , where only the basis function is a derivative.
When examining the right hand side of the equation sets
111-23, 111-24 and 111-25, it is obvious this implementation
is a subscripting nightmare; however, the use of the utility
66
modules TESM1, TERM2 , TERM3 and TERM4 simplifies the
irplementation to only determining what order to call the
utility modules. Examination of subroutines CONTEQ, VCRTEQ
and EIVEC, which compute the right hand sides for 111-23,
111-24 and 1 1 1-25 respectively, illustrates this fact. No
other subroutines or calculations were required.
Implementation of these equations required minimal effort.
67
V. P3IMITIYI MOBIL EXPERIMENT
The previous two chapters presented the detailed
formulation and implementation of the vorticity-di vergence
,
shallow water equation rrodel. The results frorr this rrodel
will be compared with the results from the primitive model
in Charter 711. To facilitate interpretation of the
comparisons, a "brief description of the primitive model
follows. See lelley (1976) for a detailed discussion of the
entire model
.
Also presented in this chapter is an experiment which
demonstrates significant improvement of the solution from
the primitive model. lelley's implementation used elements
which were right triangles. Cider (1981) showed that
eauilateral elements are far superior to triangular
elements. This experiment re implements the primitive model
using equilateral elements and a comparison is made between
the results cf both implementations.
A. ,
V C-IL DESCRIPTION
A form of the barctropic, shallow water, primitive
equations developed by Phillips (1959) is used as the
£Overnin£ equation set for this model. In Cartesian
coordinates the equation set is
68
— = (u#) (v0)3t dx dy
V-l
3u d0 du du— = - — -u — -v — « vf3t ox dx c-y
3v d2f dv dv
^t dy dx dy
V-2
V-3
The finite element formulation of this set of equations
evaluates the height and velocity components at the same
nodal points. This is an important consideration, "because
the other models ( the linearized model, see Chapter VI, and
the vorticity-divergence model, see Chapter III) either
stagger the dependent variables or have the property of a
staggered formulation. When comparisons are made between the
models, it is this lattice structure that is being compared.
This form of the shallow water equations includes
gravity waves as a solution. Gravity waves have a maximum
phase soeed of about 300 meters/second. When the correct
tine step is calculated using equation 111-26, a
considerably smaller time step is obtained compared to the
larger time step permitted in the vorticity-divergence
formulation. This is an important feature. If solutions from
all models are equally as good, the best formulation would
be determined using the computational time required to
produce the desired forecast.
69
All Todels use the same domain structure. In fact, the
domain described in Chapter III was patterned after the
iomain implemented by Kelley. Again, this domain simulates a
belt around the earth, with cyclic continuity which
eliminates the east-west boundaries. Rigid boundary vails
exist at the equator and at 30 degrees north latitude. The
domain is composed of a 12x12 point mesh and subdivided into
the right triangular elements illustrated in Figure 9.
Notice that the grid points are not shifted as in the
equilateral element implementation shown in Figure 6.
The following boundary conditions are imposed:
1) - no cross channel flow at the latitude
boundaries .
2) - a geostrophic balance at the channel walls
imposed on the continuity equation V-l.
This model has a simple second order diffusive term in
the equations of motion V-2 and V-3. However, for the
purpose of evaluating these different formulations, this
option was not implemented during the comparison phase.
Initial conditions consist of a single wave in the x
direction and a half wave in the y direction. The initial
fields for the three dependent variables are shown in Figure
12. The maxixum zonal wind perturbation of 5.5 meters/second
is superimposed on a mean zonal flow of 12 meters/second.
70
3
Si
^.oo 9j.oo .jo. jo .mu ga >.na
A) INITIAL FIELO (CPM
§
In
u.ca . si:. 30
!3Ii0^iHuvr,/mm*
-*' « yx a x ^*— 8.e < Sc A « -«r-^
§' U.O-' . 3.3"^ ^-tt-O—
J
gt XXXXXXXXXXK*
^ao so.:: .oo.oa 2-to.:a 320. 33 *oo.oo uoo.ao
3) INITIAL U FIELD (M/S)
:i INITIAL V FIELD IM/S)
Figure 10. Initial fields for the primitive model. Both the x and y
axes are multiplied "oy 10 Km.
72
This cursory description of the primitive model mentions
only those significant features that will weigh heavily in
the ccirparisions later. The Galerkin implementation of this
model is similar to that presented in Chapters II and III
and the system of equations is solved using a Gauss-Seidel
iterative procedure. Further details concerning this
primitive model are given in Kelley (1976).
2. RZSU1TS
This experiment involves the shape of the elements. As
mentioned previously, Kelley's implementation subdivides the
domain into right triangular elements, as illustrated in
Figure 9. Considerable small-scale noise was observed by
Kelley in f he 43 hour forecast solution.
The transition from right triangles tc equilateral
triangles changes the size of the domain. With right
triangles, the ax arday grid spacings are equal (300 KM). A
12x12 grid matrix has a length and width of 3600 KM. With
equilateral triangles, the*x and ay grid spacings are no
longer equal. Arbitrarily, the *y grid spacing is held
constant (300 KM) and a new^x grid spacing computed by
u = Ay/ cos (30) V-4
A 12x12 grid matrix with equilateral elements has a width of
3600 KM and a length of 4045 KM.
Figure 11 contains the 48 hour forecasts produced using
both types of elements. The three dependent variables fields
73
°VC3 JO. 30 '.60.00 2H0.00 320.30 400.00
A) t (GPM) RIGHT TRIANGLES
c3 so. -J ;sj.gj c40.oo 120. aa 4J0.00 4aa.ao10" Ml
31 J [GPM] EQUILATERAL TRIANGLES
!
" y
I - . . j . , « . L . ^ ,
°Voo so. oa .60.00 240.:: !20.;o .00. oo
C) U tM/S) RIGH r TRIANGLES
1
a» oSri
.|Hf7\U fA^1
9 - }H
tvy
X:"~"~— ,—r^
,"T^r-'T"^ .'
.
^.M SJ.;0 163. iO ..*.', .<1..u 400.3010*"
El V IH/Sl RIGHT TRIANGLES
D) U (M/S) EQUILATERAL TRIANGLES
°3 30 SO. 00 160. 00 2'lQ.OO VI. 00 400.30 490.00»10' •"
F) V IM/S) EQU:L^ r ERAL T RIANCLES
Figure 11. US hour forecast comparison from the primitive model using
both right triangles and equilateral triangles for element shapes.
Arbitrary perturbation amplitude of 5*5 m/s.
7^
are compared for each. The small -scale noise that Xelley
observed is present. The three fields shew varying degrees
cf distortions, vhich are especially noticeable along the
"boundaries .
Cider (1981) found that theroot rrean square error (RMSE)
was reduced 22 percent by using equilateral shaped elements.
This improvement is apparent on viewing Figures lib, d and
f. The contours are smooth and the boundaries are
noise free. Eelley showed excellent treatment of wave
propagation by this primitive model. The lowest resolution
grid '6x6) tested by Kelley was within four percent of the
true phase velocity. Changing the element shape had no
apparent effect on the phase velocities.
recause the outcome cf this experiment was a
significantly improved forecast solution, future comparisons
with the primitive model will be made using equilateral
elements .
75
- - 7- V - r T vt S
"": e previous chapter demonst rated how the shape of the
element can significantly improve the solution. Williams and
?ienktewicz (1951} used differently shaped basis functions
c - a linearized e o ua t i o n set to o r od uc e excellent solutions
when applied to the gecstrophic adjustment problem.
Tra'ial staggering of dependent variables in finite
difference formulations has given rucn better solutions to
*'re geostrophic adjustment orocess, and these forms are
widely used in meteorology. Schoenstadt (1930) found similar
result? with finite element formulations with piece wise
linear basis functions. However, staggering nodal points is
net a convenient rethod to implement, especially in
two-dimensions with irregular boundaries* so alternative
schemes are needed .
The implementation of the alternative scheme introduced
by 'illia-s and Zienkiewicz (1931! are presented in this
chapter. As mentioned above, this formulation uses different
basis functions for the height and the velocity fields. One
of r he basis functions is piecewise linear, while the other
is piecewise constant, as is illustrated in ? igu
r
o r e ^TO'ciT^ 1 do ma in •
S 1~
for a
Yfc
1-
a) Piecewise linear basis function.
-I h
^-A X ->I
b) Piecewise constant basis function
Figure 12 . Different shaped basis functions.
77
A. ICUATION REFORMULATION
The priritive forrr of the shallow water equations
^resented in Chapter V (V-l.V-2 and V-3) is used to derive
the linearized equations needed for this experiment. The
velocity equations V-2 and V-3 reirain unchanged and a linear
basis function (V . ) is used to approximate the u and v
variables .
The continuity equation 7-1 is linearized as follows:
VI-1
where f is the average geopotential over the domain. A
plecewise constant basis function (V . ) is used tou
approximate the geopotential. This linearization is
reasonable in this case because the Rossby radius of
Reformation jj f\ is much larger than ax [see Williams and
Zie^kievlez (1981)]. The Galerkin method: is applied to this
linearized equation set using a plecewise linear test
function for V-2 and V-3, and a plecewise constant test
function for VI-1.
The oiecewise constant basis function has the property
cf displacing the geopotential to the centroid location of
the elements, which should give the same effect as
staggering the grid points, as seen in Figure 13. The
density of geopotential data in the domain is now greater
7P
Figure 13 . Staggering of the geopotential
about the velocity. The a symbol are the
actual grid point locations and the • symbol
the centroid location of each element.
79
than the density of the velocity data. Instead of a
gecoctent ial value for each node, there is one averaged
val*.te over each element.
""he final form of the Galerkln equation for VI-1, 7-2
aid 7-3 after performing the tire differencing is:
**l, v - - ..n-1u ; <v 4f 7 i> - u <vj-';i> - •* »$<**, -V *Kay"*x'h>
+ v.u , vV . V. V "> - f vn<V V >1 VI -3
r^/^ n n . „V V >
71-4
This linearized set of equations 71-2, 71-3 and 71-4 is
solved using a ^auss Seidel iterative procedure. It is worth
mentioning that the coefficient matrix <W. ,V.> in equation
7!-2 has all non zero coefficients equal to one, since the
integration of the inner product <W.,Wi> involves piecevise
constant functions.
This equation set is implemented rather than the
equation set 7-1, 7-2 ar.i 7-3 using all of the existing
primitive model cede. The major modification involved the
way that the geopotential was referenced. With an average
geopotential over each element instead of a value at each
node fcr a 12x12 mesh, there are 233 geopotential points
versus the 156 velocity (u,v) points.
E. RESULTS
The results from this linearized model are compared to
those from the primitive model. The initial field for this
experiment, Figure 14 has a maximium perturbation zonal wind
that is one fifth of the value used in Chapter V, Figure 10
.
The mean zonal wind remains 10 meters/second. The 48 hour
forecast solutions for each field are compared in Figure 15.
This alternative formulation shows some promise,
although there are seme minor perturbations in these
contours compared to those in the primitive solution. No
explanation is offered for this small-scale noise, although
possibly the other formulations presented b/ Williams and
Zlenkievlcz (1981) would improve the solutions.
31
^.00 80.00
R] INITIAL FIELD (GPM)
oo iMO.o" i.m.ca tOU.OO 480.00
\1 !!h; "/v.' .l.'.rir'r/<t • .« I . i i h j. 41 i l ium „ I ..
^.oo so.oo .60.00 iiO.;c ;2:.;o ,00.00 -lao.oo-IO'di
B) INITIAL U FIELD (M/Sl
CJ INITIAL V FIELD (M/S)
Figure 1A. Initial fields for both the primitive model and the
linearized model.
82
NOT AVAILABLE
'"b 00 SO. 00 160.00 iMO.CO 3J3.BO iOO.OO »ao.oo
fl) t tGPMJ PE MCOEL B) LINEARIZED
aj.JO SU.JO 16J.3U My.
• 10" »CJ U (M/S) PE MCOEL
KiU.UU -190.00
0) U (M/S) LINEARIZED
.1
sir/ \ 11 i
£/'VJuo.su
V (M/S) PE MODEL
b'
.ao ''io. m ' ;so.B3 ;
'
m.:'
3 '
t»;.: 3 ' < 6
'
o.ao Tao.oo"'• 3° 80 -°°
'
60;
00 J"°-°.1Q
.3J0 - 00 <00 - 00 "°- 00
F) V (M/S! LINEARIZED
Figure 15. ^8 hour comparison between the primitive model and the
Linearized model. (APA = 1.1 m/s)
33
tf 1 1 . v:?.TICITY-riVERGZNCI ^CTZL ZXPZRIMZNT
.: s in the previous two chapters, a comparison between
two models will be presented. The results from the
vort ici ty-di vergence model are ccmnared to the results fror1
the primitive model. To fully exploit the differences
between the models performances and indicate the strengths
and weaknesses of each formulation, three domain geometries
i?A three initial conditions are used. All solutions are at
4S hour, except for one case which was extended to 96 hours.
Zrom these two finite element formulations some
additional insight is obtained concerning the execution time
required as the grid resolution changes. Lastly, a "brief
discussion on the sensitivity of the computational technique
is p i en
.
A. TZST DOMAINS ANl INITIAL CONEITIONS
The three domain geometries used in the model evaluation
are illustrated in Figure 16. All domains consist of a 12x12
element -resh with equilateral shaped elements (156 grid
points and cyclic continuity is imposed on the east and
vest boundaries. The domain has dimensions of 4045 KM along
the x axis and 35^3 KM along the y axis.
The regular domain 'Figure 16a) has a uniform
distribution of -rid ooints, with a minimum grid spacing
1$:
3
X X X X X
X X X X
X X X X X X
X X X X X X
X X X X X X
X X X X X X
X X X X X X
XXXX X X
XXXXXX
X X X X X
X X X X X
X X X X
X X X X X
X X X X
XXXXX*
^.00 90. 30 I SO. CO 2*Q. GO 2^0. CO 400. CO *80. 00
R] REGULAR GRID
* o
2a
X X X X X X X X K X X
X X X X X X X X X
X X XXXXXXX X X X
X X X X
X X X X
X X X X
X X X X
X X X X X
X X
X X X X X X X
X XXXXXXX X X X X X X
X X X X X X
X X xxxxx i
X XXXXXXX X xxxxx
X XXXXXX
^), 00 30 . 00 i 60. JO 2*0. 00 U1 . 00 400. 00 460. 00-13* ra
3) SMOOTh GRID
a 3
b
: x x x x
X X
1 £»« X X X X
X X X X
XXXXXX X X X
X X X X X
XXX
t « « x
XXXXXX X X X
xxxxxX X X X
xxxxxX X X X
xxxxx
xxxxx
X X
X X
X X
X X
X X
X X
X X
X X
3.oo bo. ca .53.;:
C! rsrupt : a
; . t'o.;o .^o.oo *ao.oo«10* «*
Figure 16 . Test domain geometries
.
35
along the x axis of 337 £M . This is the most congenial
loiraia ar.d produces the best results.
The smooth domain [Figure 16b) has a smoothly varying
iistrioution of grid points. This technique allows a smooth
variation of resolution. It was developed by Cider (1981),
who showed how it significantly reduced noise at all
frequencies oo^pared with other variable grid domains.
The degree of resolution variation is accomplished by
selecting an appropriate value for the ratio of maximum
Stretch to minimum shrink: along both axes. The experiments
presented in this chapter use a 2.5 ratio for both
directions. This produces a grid point concentration in the
right -enter of Figure 16b and coarse resolution at the top
srd bottom left of the domain. The minimum grid spacing
along the x axis is 159 Km: in the dense grid point area.
The third domain was the least hospitable geometry for
both models. The abrupt domain (Figure 16c) has a dense grid
point concentration on the left and coarse spacing en the
right of the domain. The minimum grid spacing along the j
axis is 15S KM in the area on the left. The grid spacings
are uniform except for the abrupt change along the center.
Although these three domains are simple in structure,
they are adequate to test both models' performance. To
further enhance some differentiating characteristics between
the two formulations, three initial conditions are used. Two
have previously been described in Chapters V and VI. Their
36
rain iistinguishing feature is the arbitrary perturbation
airolitu<?e AVA) magnitude. The nearly linear case Las an APA
= 1.1 m/s compared tc 5.5 m/s for the more nonlinear case.
All initial conditions have a 12 sn/s mean flew component.
tfith the nearly linear case both models behave well. The
introduction of the rrore nonlinear initial disturbance
illustrated the boundary and computational technique
sensitivity. The third initial condition is the nearly
linear initial field with the wave length equal to half the
domain length. This has the effect of producing two waves
with the domain .
2. ^IST CASI COMPARISONS
The comparisons between models are divided according to
the domain geometry. The primitive model has three fields!
geopotential , a and v. The vorticity - divergence model has
seven fields: geopotential, stream function, velocity
potential, u, v, vorticity and divergence. The geopotential,
u ^tA v will be the only fields used for the comparison.
1 . Regular Case
The regular domain geometry (see Figure 16a) is used
for this first comparison. The initial conditions have an
A?A = 1.1 m/s, as shown in the contour plots in Figure 17.
The 4:9 hour solutions for both models are presented
in Figure 16. As anticipated, all of the solutions have
37
8aSi
NlTIfiL $ FIELD ICPM)
UU.00 490.00
1o
1—V '"
i
,
|\[
«sL^ > /A v\^i^/,/ /A V *m
tl ,11 I ^ I'
ll
^1.00 30.30 150.30 2WO.0G :20.3G <00.30 480.0010*M
B) INITIAL U FIELD (M/SJ
x - — i.i \ >
tfjl - :,^-. yi
< » * . . < , « .
!-
,1* « '^i/ ' ' » . ty
^.OO SO. CO '60.00 210.33 3i3.C ioO.OO -.dCCO• JO'm
Ci INITIAL V FIELD (M/SJ
Figure 17. Initial fields for both the primitive and vorticity
—
divergence models using the regular domain and perturbation amplitude
of 1.1 m/s.
88
15U-*
^).30 ?a.oa i6o.oi' (40.00 vo.oo noo.oa 430.0-
fl) (GPM) =E MODEL
.03 io.oo 153.00 Jta.oa ::o.oo 400.00 hso.io
B) (GPM) V-0 MCOEL
^.30 50.33 '.53.30 <43.30 JiO.CO ,00.00 430.0010* n
E) V (M/S) ?E MCDEL
Figure 18. ^8 hour forecast comparison between the primitive model
and the vorticity-divergence model using the regular domain. (APA =
1 . 1 m/s )
.
89
smoothly "ontoured fields with no noticeble noise and their
phase velocities are comparable.
v -th the nearly identical solutions, the
distinguishing feature between the models is the
computational tire. The computational time is determined by
the size of the time step each model is allowed to use as
indicated on equation 111-26. For this domain the gri
i
snacin* is 33? £* . The maximum phase velocity possible is
different for each model. The primitive model allows gravity
waves so c - 300 m/s whereas the vorticity-di vergence model
filters out these high frequency waves. A c = 12 m/s is used
for this formulation.
^odel
D -
?"]
Table 1
<3X c
(m/s)At
(sec)
it 7-z-77
3?010
45513,755
* of steps(iterations)
37c13
CPU units( sec)
145.- p.
Coinoarisons of the computational times for a 4B hourforecast between the primitive model and thevortinity-divergence model.
Table 1 co moares the results of the computational
times for both models. The vorticity divergence model
produced as accurate results over the regular domain using
22*- of the computational time needed by the primitive model.
In fa^t from geostrcphic reasoning the vort icity-divergence
model should produce better forecasts when small scale
features are oresent.
2. Smooth Case
c rp -\ r\ t V omain geometry (Figure 16c s used in
this second comparison. The initial fields shown in Ji^rure
IS have a disturtar.ee amplitude of 1.1 tr/s.
The 48 hour solution comparison is presented in
Ji^'jre 22. All fields again have srrooth contoured plots.
Close inspection of the two u fields, Figures 20c and i.
shows that the urimitive u field has small kinks along sore
contours and a weak tilt near the central boundary ncdes,
although this may he a function of the plotting routine.
Notice the £ood symmetry for the vort ic ity-di vergence u
field.
As in the regular case, this smooth experiment
produced twe acceptable solutions. Again, the computational
time is the differentiating criteria.
Table 2
vodel AX[EM)
c
(T/S)At(sec)
199199
32212
271£124:
n of steps(iterations )
c3821
CPU ujnits( seic)
324 .3
49
Comparison of the computational times for a 43 hour forecast"etwee", the primitive model and the vort icity-divergencemodel using the smooth domain.
Table 2 ccm-oares the computational times for both models.
The vcrtici tv-diver*ence has a 33.8^ saving of CPU time.
_iJ.
_^
^"
3 —
.
a
-T-_
1- « — «9S39-
^.OO 30. :o
PI) INITIAL t FIELD CGPMJ
.60. 33 JHO.CO 32J.00 400.00 480.00«10*M
i \*\ x * * V'Z IS v
#-'v\1
. . J ... - . 1
^J.JO 30.hu . »••.'.
• 10*'-"
3) INITIAL 'J FIElD (M/S)
• < « «
• - . - « f
-] u . ,
i>
^a.oa lo.ca iso.co
C: I'.:-:-- r :E_: (M/:
Figure 19. Initial fields for both the primitive and vorticity-
ilvergence models using the smooth domain and perturbation amplitude
of 1.1 m/s.
92
•
oo jo. 30 16O. JO Jho.j: >20.00"10*"
RJ Z !GPM) de mgoel
jo. jo «ao.30 u do.oo :gu.uu riu.uo i.'u.un K'U.uo xmo.oo«|0" ««
3) (GPM) V-0 >1U0EL
?j.oo so. co ;so.:o no. oa ija.aa .oo.:oIC'w
C) U [M/SJ oc- MODEL
Tj.00 80.00 !60.00
0) U (M/S) V-0 MCCEL
J. 00 100.00 400.00
.-
< «
i [ \'
I' *. • ' *
'
H. M
E) v iM/si pe v\::e.
ifU". N J,
;'•'-•' '
. .v.'..-.','
! \" /Will I
.03 dC.33 160.00
Fl V [M/SJ V-0 M CGE
•33 120.00 .33.30 -aO.OO«I0 «
Figure 20. ^8 hour forecast comparison between the primitive model
and the vorticity-divergence model using the smooth domain. (APA =
1.1 m/s)
93
In an effort to contrast the computational accuracy
at the rodels, this experiment is repeated using the rore
nonlinear initial case. Figure 21 shews the initial fields
with a?A = 5.5 rr/s . This larger initial disturbance is
reelected in the greater geopotential amplitude and
magnitudes of the contour lines.
Figure 22 shows the 45 hour solution comparison. As
in the rore linear case presented above, all of the plots
have smooth contours with no noticeable noise. The
vorticity-di vergence geopotential field, Figure 22b, has the
rid^e extending farther north, and flattening of the
southernmost contour, than does the primitive geopotential
field, Figure 22a. The mean geopotential heights for both
models have also increased. The primitive mean geopotential
is now 49£8<? gum and the vortici tv-divergence geopotential
is 49630 gpm.
These two discrepancies indicate that the boundaries
are not handled accurately. Turing the implementation of the
vortici ty-di vergence model, treatment of the boundaries was
the most troublesome phase. The vortici ty-divergence
formulation is a complex equation set and time limitations
restricted further investigation of mere sophisticated
boundary conditions.
This same initial condition is now extended to a 96
hour forecast, which is shown in Figure 23. The mean
vcrticity-divergence 2-eopotentiai increased to 49S00 gprr
,
*1
^>.oo so. co 160.30 2*o.:: no. ao tao.oo 7ao.:o
fi] INITIAL FIELd" IGPM)
-! sa >/
« « # '« \*3.o i^« > x \* * a.o_i
:i x n n x x k « x x » < x
>-H'>.v. • /*— "-—»—« « *>^**-1
a i ...... a. . t . —t—. *-
^.jo 30.oo iso.oo <4a.;3 j;o.go -.co.oo »oo.oo• 10*"
B) INITIAL U FIELD IM/SJ
•.-..-.'j a « « < I la... Jlj[\\\\
i*. x\ *--»"xy » ,
|fx / x x * art
ft I::•'""<
> a K all a
« i < « » < - k
/. . .
":. BO 80. -0 ;. j 00
'10' UK
.00 .80.00
o initirl v r:
::_: im/sj
Figure 21 . Initial fields for both the primitive and the vorticity-
divergence models using the smooth domain and perturbation amplitude
of 5.5 m/s.
95
^.m aa.oo ;sa.30 243.33 323.30 «cc.30 480. oo
lO'nfi) (GPMJ PE MODEL
^J.00 SO. 30
O U (M/S)
iso. os ;*o.oo 323.00 -.co.so
lO'nPE MODEL
1•
. ww.w;'
T.30 93. 03 .SJ..
E) V (M/Sl 3 - MODEL
^.00 30 30 160.00 240.00 320.00 400.00 480.00*10* a
3) $ IGPM) V-D MODEL
HZK f^rU.Oj I-t
"3.00 S0.3C
0) U IM/S)
JlL-1'.£0.00 240.00 370.00 4110.00 uSO.OO
c 10* KM
V-D MODEL
: / .
' .
' :. /vMVNUf
m •
-1
I
l
i7
80.00 '50.30
V [M/S)
!40. 30 320. 3010" K»
V-D MODEL
400.30 -.80. 30
Figure 22. <*8 hour forecast comparison between the primitive model
and the vorticity-divergence model using the smooth domain. (APA = 5>5
m/s)
96
J
^.00 50.30
3) # (CP^l 3E MCOEL
.an jno.no 480. oo10' <«
^j.oa ao.oo :so.oc 24.0.00 320.00 -.00.00 490.00«1Q' **
B) (3PM) V-0 MODEL
o u :m/si =e model
^).C3 30.00
D) U iM/S) V-0 mqoEL
SO. CO JHO.ai J20.00 -CO. 00 <dO..lO'w
. 01
J .
.' .-J
..... .
§ « « - - 4 < -
i' ' " ' ' f--. « « * < 4 . . < 4
\\\ •
'•. '"
k,; si
.30 so. as 160.1
E) V (M/S)
,j.:a ^30.00
3E MODEL
, \.....
.. , *.«. ./..U\,
bO.OO <r4C. .:n VII. CO 4UU.00 480.00a.oa ao.oo
F) V (M/Sl v'-l'-y!ODEL
Figure 23 • 96 hour forecast comparison "between the primitive model
and the vorticity-divergence model using the smooth domain. (APA =
5-5 m/s)
9?
whereas the mean primitive geooo tential regained constant.
All fields have smooth contours. Close inspection of both u
? * e .. z A — -4 . W '2c and *i, shows a slight skewing of the
contours along the central channel grid points. The
vorticity divergence u field has the rost pronounced
deviations. A hypothesis for this skewing, explained further
below, is that it is caused by the computational technique
employed. The relaxation scherres are extremely sensitive and
fine tuning of the relaxation coefficient would nave
rehired more time than was available.
'able 3
Vodel A x r
(m/s)4 t
(sec)
199199
2Z212
2718124
# of steps( iterations )
127642
CPU units(sec)
594.291.0
Comparison of the computational times for a 9€ hour forecastbetween the primitive model and the vortici ty-divergenceT ccel using the smooth domain.
Table 3 shows the comparison between troth models' for
the 96 hour forecast. A savings of 35 percent is realized
with the vorticity divergence model.
The above experiment points out the two areas where
the vorticity — divergence model is presently weak, the
Increase of the geopotential and the sensitivity :f the
relaxation coefficients. 3oth these weaknesses can be
improved and are not a result of the formulation but of the
implementation . At present, their influence is not detected
93
ex-ep' In extremely long forecasts, such as in the 96 hour
exaxpie. Further experiments ray still be required if they
demonstrate significant differences in the solutions cf the
two models.
The last experiment on the smooth domain uses the
mere linear initial condition AFA =1.1 m/s, tut its wave
length is divided by two, so that two shorter waves are
propagated through the channel. The initial conditions are
shown in Figure 24. Decreasing the wave length has the same
effect as decreasing the density of grid points. In this
case six grid points are used to describe the wave structure
versus the 12 used in the previous cases.
The 46 hour comparisons are shown in Figure 25. As
in all previous cases, a computational time saving of S4% is
rained with the vcrticity-divergence model. With fewer grid
points describing the wave structure, more small-scale noise
is introduced into the solution. Comparing the primitive
geopotential field. Figure 25a, tc the initial geopotential ,
Figure 24a, shows a dampening cf the wave amplitude, whereas
the vo rtici ty-divergence geopotential. Figure 25b,
correlated well with the initial geopotential.
The high frequency noise is evident on both u
fields, Figures 25c and d. The primitive model u field is
poorly defined along the boundary and becomes irregular over
the interior grid points.
Si
IO'kii00.00 430.00
fi) INITIRL p F [£L3 ;pmi
.Jib.'. ] :- :%2f
/j//L\\ // J\\ s? '•
i
/lfl\*xi«•J/ k\\x/ «. < C- < « .< jl> M-oi
- .>*
it • r iwlj " it— Maui~.oo ao.oo tsa.oo 2<<o.oo ..;..; 400.00 4SO.00
B) INITIAL U FIElO (M/S)
i- . . . - . . . . 1 . . ,
- „..«... - .
I
-J i • £ \
;
=b. 30 SO. 00 1 60. CD
C) INITIAL V [ELC (M/S)
0.30 :<C30 400.00 490.00
Figure 2fr. Initial fields for both the primitive and the vorticity-
divergence models using the smooth domain. There are two waves embeded
in the flow and A?A = 1.1 m/s.
100
*1
_ P il « KB
T.JO 30. JO ISO. 30 2>.0. JO 323.30 -.00. 30-10* «
P) IGPM] PE MODEL
- »i
ii I.i
ii i« i
t.oo au.oo if.fl.oo .jo. lw i 'o.iio nnu.oo aao.oo10" a
8) ZJ (GPM) V-D MODEL
3
*]
——,-_—._,? - 1
(J i \ fcl r -
r ?* * / x \ x / J \ V x 'xl * V_i
S** yi x1—-T^_. « -.-—^* J v/r^T
IS / .,-0 .^^7^1 f, TV x/./TX:
ji t ii i .i . . . , /// i— .
—
j_i i_ui , 1,! ii —imU i i., ii .n i,
i i Uuil'VoO M.JU IfiU tHU.UU ....ju .0O..1O
C) U (M/Sl PE MODEL
80.30 ^.30 dfl.00 Itill.jU .10.00 1^0 OU 400.30 480.3010' n«
D) U (M/Sl V-D MODEL
*1
"^x^\ < / / *_^ xi -x « i k A x i \ *A
i
if* I r . > . . .
:
ui . .
' «
- . . . , 4 ,.,...,M y-w •.:'".::.S. .......
I 4: ' »i X1 M « •
• —A *9\U k j - j ->-
LJ^« 1 . 1 . < .\^ . '
' / ' ^ , ,
i "..0,
„ I 4 ^.r ., -'. t| . . ,,y .„J i' ;;
Vx
*.(*-
iL-w,J%.0I H.OC [60.;
EI V (M/Sl10' oi
PE MODEL
.30.30 .-iO.30 30 -!..:: iSO.OO Jv.O.00 3J0.00 .00.30 .80.3010* w
F) V (M/SJ V-0 MODEL
Figure 25. ^ hour forecast comparison "between the primitive model
and the vorticity-divergence model using the smooth domain. Two waves
are em ceded in the flow and A?A = 1.1 m/s
.
101
The smooth domain allows a variable resolution of
grid points and produces excellent results. As the wave
iensth sets shorter, or the forecast length gets larger,
sere s^all-scale noise is apparent, especially with the
orimitive f ormulat ion .
3. Arruot Case
The abrupt case comparison uses the abrupt dorrain
geometry 'see Figure 16c). This grid ooint configuration is
v.sed to -"urther illustrate the effects of spatial
resolution. The previous case using the smooth domain and
half wave length introduced noise into the solution, but the
spatial resolution changed slowly and gradually.
Consider the transition necessary in an operational
model, where the luxury of havi-n^- a- long smooth- transition
into the region of high resolution may not be possible. The
abrupt domain is an example of the results obtained when
spatial resolution is decreased rapidly.
The initial fields are shown in Figure 26. The more
linear case, A?A = 1.1 m/s , is used to eliminate effects due
to the initial field, so that only the effects due to the
£ri J ^ec^etry are seen.
The 46 hour comparisons are shown in Figure 27. 3oth
solutions are affected by this geometry, but the primitive
solutions are totally disorganized and unacceptable.
Table 4 shows the comparison of computational times
for both models for a 48 hour forecast. An 3££ savings in
122
p) ;Ni T ;ai c field (gpmi
B) INITIAL U FIELD JM/S
/ >~^ \ r
1W-WiHrrr"""]Ur......
* + » "^-C.i.J'"* <5^ \^7^.30 ic ja 240.30 920. ca ,00. ao -lao.ao
ilO'w
C) INITIAL V FIELD IM/S)
Figure 26. Initial fields for both the primitive and vorticity-
divergence models using the abrupt domain and perturbation amplitude
of 1.1 m/s. Note that the element shapes are not equilateral triangles
103
JO 80. CO 160.30 cVO.JO 3.M.0O10'
w
R) IGPMJ PE MODEL
CO. 00 48.0.00 ~J.uO 90.00
B) iCPM) V-D MOOEL
100.00 ."'40.00 VO.UO 4UO.0O *ao. 0010' KK
-k^^r^r /f r*'ft N*
3
S9^ ; '
A w />"> A o' \ J
U.J
t.jo ao.:o*.—I— * —
.. ;: ,; o.co 400.00 <ao.oc tj.ou so. go :so.oo jho.oo jjo.oo too.oo <eo.ooi n" n »lC'a10'
«
"1C a
C) U (M/S: PE MODEL 0) U (M/S) V-D MODEL
V0" ' ' '
six J < i a
b Hy> «, »\
«—>, « i « • -
.
. < .»
°a.w so. oo i60. :o j^o.jo :>t.do <oo.:o .?o.oc«10'«
E) V (M/S) PE MODEL
f in ao.jo ji.ij.no i»MO.nn
F) V (M/5J
n vino. 00 gflo.oo
10 «»
V-D MODE!
Figure 27. ^Q hour forecast comparison between the primitive model
and the vorticity-divergence model using the abrupt domain. (APA =
1.1 m/s).
10^
CPU tire is obtained using the vortici ty-divergence model.
only is there a computational savings with the
vortici ty-divergence model, but the solution is
significantly better than the nrimitive model.
Table 4
A X'7v)
c
(m/s)At
(sec)
16?168
3201?
2286558
v o^el ax c At # of steps CPU unitsat ions) (sec)
756 368.925 55.7
Comparison of the computaional times for a 48 hour forecastbetween the primitive model and the vcrticity-di vergenceroc" el using the abrupt domain.
C. COMPUTATIONAL SENSITIVITY
This section will offer an explanation for the skewing
of the contours in the 96 h~our forecast solution- over the
smooth domain 7i<?ure 23). As mentioned previously, there
was not enough time available for fine tuning the
c verrelaxation coefficient, which is used while solving the
system of equations. The overrelaxation coefficient may be
very sensitive and small changes can, on occasion, radically
change the rate of convergence. The optimal value of the
overrelaxation coefficient depends on the specific form of
the coefficient of the equation and the error distribution.
The eouation set to be solved consists of three
equations and each equation required its own relaxation
coefficient. When solving the equations over the regular
domain, the entire system is well behaved and an optimum
125
relaxation coefficient is easily determined. However, as the
ioirain geometry changes, the system cf equations do net
converse as rapidly and the relaxation coefficient requires
further fine tuning.
The 96 hour forecast uses the smooth domain. The
f*i d -la ti tudinal grid points are compacted, creates a denser
belt in the riddle cf the channel. The coefficients
originally computed using the regular domain need
adjustments to properly solve the equations.
To illustrate the significance for fine tuning the
relaxation coefficient, consider the series cf plots in
Figure 28. The vorticity divergence equation set can "be
simplified by assuming the flow is non-divergent, so that
only the vorticity equation needs tc be solved. Figure 23a
is the 46 hour vorticity field using this equation over the
regular domain with an o verrelaxa tion coefficient of 1.3.
The field is well defined with smooth contours.
Figure 28b is similar to the case in Figure 28a except
that the smooth domain is used. Notice the V-shaped k:i nic in
the pattern wi*h a steeper slope in the upper half. This
in creased bias in the upper half is caused by relaxing the
field in the same direction during each pass over the
domain. When the direction is reversed after each pass, the
exaggerated bias in the upper half disappears, as is seen in
Figure 2£c. However, the V-shaped kink is net eliminated.
1Z6
3
1
*1
.lyi//.
^i.ao ao.30 i»o.co nQ.:3 ko.oo 400. :o 480.co
R) VORT 48 HOUR PROG
^b.oo 'o.oo 60.20 240.33 ?20.co ,00. oo 490.00
B) VORT u.8 HOUR PROG
W/T?\) ::::::.
! • '
-- -"-
T.30 iO.JO :S0.30 240.30 323.33 ,30.30 <30.3G-10*
•' = 0) VORT 48 HOUR PROG
Figure 28. Computational sensitivity using the 48 hour vorticity fields.
Fig. 28 A) the 48 hour forecast using the regular domain, 28 B) the same
forecast using the smooth domain relaxing in one direction, 28 C) same as
28 B) except alternate the direction of relaxation after each pass over
the domain, and 28 D) same as 28 C) except the relaxation coefficient was
fine tuned form 1.3 to 1.297.
10?
Varying the relaxation coefficient from 1.3 to 1.297
produces the much improved solution in Figure 28d . If the
relaxation coefficients for the other two equations could
alsa he fine tuned, it appears that improved solutions would
result. There are also other relaxation techniques available
that have potential for improving the solution while also
sonrerging at a faster rate. Some of these techniques will
he tested in the futuer usin^: this model.
128
VIII. CONCLUSIONS
T'his research investigated different finite element
formulations for the shallow water equations. The
two—dimensional dorrain was a channel which simulated a belt
around the earth. Analytic initial conditions were used to
simplify the comparisons. Two formulations were examined;
one using different shaped "basis functions and the other
using a different form of the equation. Zach was compared to
the primitive fornr of the shallow water equations that was
developed by Kelley (1376).
The use cf equilateral shaped elements which was
su^ested by ~r. .
V .J.F. Cullen significantly improved the
solutions compared to lelley's model, which originally used
ri?ht triangles as basis functions, *ost of the other
studies in this thesis used the equilateral triangles.
Williams and Zienkiewicz (1981) suggested the use of
piecewise linear basis functions for the velocity field and
piecewise constant functions for the height field. This
formulation was tested with a linearized continuity
equation. The results were poorer than those obtained with
lelley's model .
v ost of the effort in this thesis was devoted to
ir-plement at ing and testing a vcrticity-divergence model
1 7Cx C H
sirrilar to the ones developed by Staniforth and Mitchell
fig 7 ?} and Cullen and Hall (1979). Several tests were
presented which compare this formulation with Kelley's
rod el. It was found that this model executes approximately
one order of magnitude faster than does the primitive
formulation ^sed by Kelley. Secondly, as the spatial
resolution between grid points decreases, this formulation
oroduces a solution that is far superior to the primitive
form. A disadvantage is its computational sensitivity, which
requires fine tuning in solving the elliptic equations for
certain geometries. It also requires 25 percent more
computer storage, due to the more complex equation set and
the additional variables that are treated.
Implementation of finite element models is not easy.
Fowever, there is a lot of generality and redundancy
imbedded In the computations. Versatile modules were written
which significantly reduced the effort in implementing the
vortici ty-d
i
vergence model.
further research is suggested using this finite element
formulation. It has accurate phase propagation , is able to
handle variable grid geometry, reduce the small-scale noise
and decrease the model's execution time. Specifically more
ad.va.r.cei methods of solving the elliptic equations should be
investigated. Finally, the formulation should be tested with
small-scale forcing, where its advantages should be most
evident .
112
.
. . . I .
I - .
L • • i
.
r-iF i m
* j ^
' i
i- -
L. .
.>
1 -
< \ X I i ( > t
I
F ; . t '. _ - ~ ~
r. . . .
-
, \ [;..{• j >
i
ii.jj T - ~ 1 •(
-.'.,)
: .
r
ili j /ml >E !l.) C iA.\l\El LAT! T >F
: _ .
* -.-:;.; ~ ~:
r-
i i j k s j
r :.,r FLAo ISEO r irJiTiALU u icn u n , p s t f C --i I i
.^- ••• f it: ATI
'
. L"
It J t ] cc ip L E FCST
j
: . ;
_~ r. r • th thsj le^/el : .
i: :: t .
5^ ^i*|TS G.'j The r AXI 5:f ; x : : f ; . its j r i- c jxisr: t.- : i\-i j t :
-- levelj::
-:f r -- compacted j. inectivity a.-.say
; iTAL i
> j
;
'
. i ^ [.i ruF 13 HA INf • =s i ,
r.-1 ; j c y a I j
: 9. Pr L A r I C J P • 3 IN T ire Tn z
( "ii- 1 ) : :• -. lev el
r~TAL NU
3
»IO .--- - -
r I-J T J ; \ A < \X I
";
tg;a:<yya<ISl>y: - f j \ c t i - : efficient
INITIAL MI ) C -1 Ai T I \ ij I t y '_
; j ; r ;j
'
JY
F 5 \JT )-:'' 3ild I )A*Y
L ; T IT jjij c . F • -
: -.,_• j ..;:-: Y
. r I T JD ELATITUDE
. r-Uk [Z J iT.-,L l 1
'
f" - rne : j . ; . i :r[ : ;;]
r A L _ - F r W E L E
'
"* y> T n 3 i L IW ) E L
th - r - -•]- i•'- 1_ •
* i
j- . . . i
• - . . J . /"
' J
iGT a ; jT -
. jjv)a;;y Fk;; i ;-j jatgI [ 3T; CE " W T i- < ! jJUf Jm -.'r :" -
. '
-.
. . - iirr : :: f r
; C I T /£ ; L A T I
-[T
I
- n1 >:
t i
- -
J •
j <
/ 1 j ;
i j
• \ . ;r «t :
i
• ".
J
-\ : [
-t i :
I . i J T I-
.j \JCT Ic'i i F \ j i I A T r» [ C F 5.
"
. ; it > T" -.r [ .... fi :lj
:e r: 'l :
:•
r;.- :
::
~ •: '.: : ; - each ele il\t
.33 '•'•: \ JSEC E'l SJLVi io C..,i: UIT> lQUATI-jIj
'i3S 'ATX IX jjc. I I SCLVI ivj 3EPE-NJE IT /AF. IABLES
« _) o "«"»< Jjl'Ij IN 5 ^> L / 1 io j i ri T "» ) I V L ) .
T : : -. 3 r • • -~
.
-.-. - < jje s
JT:>:JT 3.ELFC7 IG\3'
•
' ITAK7ING .TiCATICr. I j CCXNECT I 7 I TY
; : c t i 7 i t v '• \ t f ' <
".".:. .,•..- f jes :...e:i.:j t: each e l £'-1 i i t
F Fhl-VJGL I IfcKPUL \7£C
I SLY S IV E J 'HI ,3 SI •: :hi r
1- !
.
1"
r.
• [ A l :
I E L : ( D C ;
J
. ,'
cj <:~: .:: '-: eld
('
• S ' L , T - R T I C I T Y ) F I :f L D
\-t\ h: A ,*I J
.. ; _ . < s 7IEI7Y) rlcL
. . ,: •- J3F0 I : c ;lc Jlat riG r ie I
: I ;'--') IF ALL - ,/A, ;M3
?. HS /AL JES • /CxTE j AN J j i
:)
S-r. J r\ •: E F . J \C T I I / A L J E 3
J . : •Jn 'i_ .T }F * IT! ; . F I f=l_l3
H - L I FI :LJ
;h r
' p- "i*
i E i r f---:•
f I e l J5
v j ; j c - < v j # / 1 >: .: »h :j jot < vj, •</•<, yi>
-> ;'
: ICT <VJ*,VI. < V J < t / I .<>
... <7J \, 71 >
: jc t <. v j , / ! >
r> ,.:"<; j » /
.
' N
•.
; t : : E E '.G V F I EL ]
:' i JlS7A.iCF -" Z ' ";IGIN \l1 ITE 5GLATE r
J ; XL.i )I IA7E3
: • S J I > 7 A.* 1 C E.^ . ; ,7: ' .'_ \ "E
r ! C [TV = I E L J
. - _
] Jj - I .
1-- I j
I # • * • •
.-
;
. j i
-
J1 1 : i
IS
: r .
/)• -
. » .
.
L * 1• >
-~
,
j'-; . - * I
-i ; : L > < L h
z\ • . iiL • j .»...: :
:
,
r ir s l?t «: I- - " "
: :- i.
j... :.E .
• r : _ ;_ u v i r i
\i
[
t \ fi vti •-
.
.. . rv • r i
; .-•: r s . , . ,
> • t L O
ITZZ"
'. r .'
. ; .\ n l FjC I^l _ J J -«
'
I."
' . . ..._ + '
i "J J» 5 / '- U \ i_L: S =ijh 1.. . i.il. WVLYJIS L6J; •
. : .a t i i r i l -. i i i y s i s :
r
: ,: : lat.es > j i .~>
:j.< ia^ij.uc \ ialysi s , j :
:'
"i vl i : . ; >hi, r> s i uj ;;-»! - i -:l ; ,.
.
i t : \u i : -; i )el 5
. - » if r:
;
. [ ri XLIZ^S U.L 3LJQAL /A-*I \.iLtS1
-
"S *LL
r Ti :>c ;-JC F3
: > jtes rHs i it-: .
i
:i *t :- ;ki j ^chts
• •«••••*
• * jt
: s .-..«: i j•
:- < i \ : r r L ;
7
? m sr
•:
!. E i ' - <' ~ "
: i c : . : e ; <\
~ i ;\i l 5 o>LCJLATiS Hie" X, f j [J > . i 'J
r CISTA'ICES E23
' '» \ N H I* j ;.j '4 T '*
I 3 J T I C N I i :* 7 !
': L \ LL Y . . . L 3 c: '4 T 3 ' T 1 -1 2 E i T 2 3 3 A T I L 4 L
J T E5LJS ru
. ' : .:t i" : f : f ^ u f e l l
. .lLj - - . ! \ J A J L JT JF FIELJS .1
, T. \'Y iOUE
• _>-
I ..> TG FH '
Ji_ - T -. , > %
res )iJ : / F < .1 _ T.: !
. : ::
.•
>: ; e :
:
: w . / . -iflax\tiu:j; _ / : s F' '
-
'. 3Y jvc - :^-^ r. : i.3
Lt'MJb ! C v-n- I . . . ^ • \/j
i :*!;-.::•) r i? : ; : , >.<
1
1
+
12 1
3 . : : j p -, -.Lc 3 -'_..:. r y -.-:'• -. l
- n|
- - -
; )
> r > .> I J j jiiL v / -« r» Y I !' t j ........ ...^
^»<j i^^' "." I j . j j ^ I i ^ \ i ...... ..........
r - i
! / ~Nr
. . [ i : ; . r ] , ;l /•: -'
'
3 ;: l-t s
- r i > J
/ ;
'./ ;
. ;
-
L -
- —-
L *
L:'
lln'I >
> »
7 f
» _ . < ' t '_
\ l-> o
-U|
i
< t
) i
'. '»
'.7
/.5 j<,' ' J
'
\ / / J
/ . Jl
/ < II
IT/
I
r
_
I
t
i » J
: < i
.1 i
L 3 fa
r . 'i t r f
t I i j i
ri .
-, > -i •' ; I L I r ^ J
> 3 J ( i d j ) , \ '•! A S 3 I ( I 5-
. IT - 1 I L : ») f .<n U J f
y~ l » v t j p , y o : r , < l a a ':
* l a r { i s ^ ) , \ ' , 3 p c j c:
,:.•>)» . - •:.:(:.• i
j / ( 3 » 2 :i i ) » ; IJ /. ( 3 1 3 t
L 1 • t
V i [ L
f , s l ;
:
)
-. >
) "i
J IN ( 3 f '
... < :-
Lift) ,
'
3 , i
J »
ill-.')1 I ( L 3
(I r - !'R]( I
^ ![".:•), i.ave;
) * i
n , / j / : x ' $ , ) i m n2 j <n , v j <" : y ( 3 , 3 f z i nLdo ) t ALFA i I5o ) » ii" r AuST Irs
i L56
li I
)
,
1
L i L 3 ;: )
I ;.'...:.." .j ph\3 =
:"
: al :.: - '", v_ L - . ^ • » -
: . it i iLiLf.. __ i [T
'. , 1 F 7 - S S
: :"..:_" [ s i,:j T
f i .~> s > s i f ; . i : * \ •> ^
. i
;hl ic
....
\3 t j•
., :
-.- :
•
"i •
:rEOf'N 10 i£ Fb.s F Ir3T :
: •i * . ST
; n. :
.
T : : . ( )T
)
S - L - ... !
' '
. .
'
', , r it*.\ui j::lt ; < ,:,_ • . i. atv.il 3,':j
-- ! . . r ~ 1,3) ,'HI Oif't /T/i }| . . IF
/ : ••
. i•
I:
•- ,Yl_AT i L T : I , A.y : -
t { L J ~i
» .. r i_ a
. ,• .' - L J
) * "I
_
_ r = 1 "i L , - L i T
:= ,L IM LATH)- -. •
' .L - I i)
1
; ! . T .:. : jr .'.-
I;- .: - 1 => -i * - »
> = P•
.•
"
1 . ' * .
"
( C A , ; T j £ L - 5 : T •
"
•
. > r jki.j USE u = -l.o_ -> r
L.O)
Iri-ji i.o re '.C5
• - : = ) .
:
uiLirv
- : iTj-.r;;. j •~
. :.:. is [ te? ui ;
" = V3 hp "^f
L-.'j; ^._; ' = _-; 11= i ; L 5 = V ^ ; [)=(,); 21= 1Z\\ •
:• =
r. c l l \
~ :
' .
,
r -i j a : _ >
r- =t = i.i.- - t - ~> -. 1
(L.I ,.« = ).<:
: = --S . 3 7 » E )h
i : ; = l ) . o
- _ . - > •
- . ) fe
. ' = .• ^ • 3 . ) / j . 7
.'-» = \ «P --S .O/L .2
ir.^;;,v = ) . .1 '-
— .
J
= ). ]
1, 1
- J I
• - = CJ A . .1
= /).0= l
)? -) ).
:
' •r { i ) =
-ij - t : ,-
[Ty 'A .
•
-
3 ) =-< ' j l .• \ r £
-i = ;
i.«.-U
* - - - - T . — . ,
.) =
} ". I
:--,')=;
Ir
i v . J t v i ! . f ) •
r
. ) i i „ J i iv'»\LFifiETJ» - I E L "3 5:
• -; o = i
'.' '-: ,ui it = iei_ts afte :\cn ire^ATirj ;\ : i ;"-
:r{ v) = o
kLYTICAL /ALJC3 F - „,;,\-:,/; 7 , Q 1 J , ) J , ; V , -I - . , >E \ A
1 [, 5 S I UC CHi F«: » R I 4 1 TI -IE r.r -- ; \i. i i I jT t 12 )
: :r(in= i
"'
i
TI
'
i T E • /AL .J H J-- - ( IUoT JL S : IE KJLT I LF ;r )
j \ /'
:-_
; ; j >*i t ti.*e ; ,:i- ;. / \i : . i
- :? n iz j
: T( L3)= -
. t . : • , , - r , j i >/ t }j , }y , al f • » .j l r ; fields \ s [ : [ ^n \ t ( 1 3
)
" L4)= L
r( L5) aI
. ' : ._ field >u i r
:
: >« »i ;
• jj = :
.
'- iich fields rr ?L.i r ; 7 digit f , \
• y nu.-i jE-;
L/3 : ..' :ri7F J j:*;,. :P U: PS I : C--
1
1 : ;J:V:VG v T : j 1 7 :
1 >3 i ,u , / [;3 T : IELui; I?U( L£) = JL'UOLOJ
'
. .i = L001100
; -i r ( i ? i
• l r b ; >
= L
= a> : • [ : • ; i :. \ ;:,:.:,.• 7 : ; I ; :
J ) a 1 J KM
. . j
• :T
: -liz ,= r :
:l > •. < -.> t ; s. this r
• cll. ,.-.>
i A.J.L . , : 7.: t
- j•
i
L L ;
. r
. . , r i . i
ui: :"
'
i c 1 1
J
- c-iT j - . J iU IjE.\ r IE
: i
N >rc n vi n ia tk i x
T m i ''
,
- -_ ;-'. \CTE-xIS7ICll : . -
CC -1 Mr. <iv :C~*DI 4 7E5 -.n r i£ V, )ES
1 -LL L iCMc
T < \ .SF J* IS ' •_ N IF- 1*1 iRIO T J Ui"U?7 J-sIO *r- 5 i
LL
;c " \LL I .
--. 3 ' )UCTS: - . . : .
" \l ; ... \ r:
s %35 'AT
? : <
... ,T•
i
j ? > {. * s r : : .-zj-j f j .< i * r e x ? j la 7 1 • j n. l _ i > ms c
r
r .•
.
- ; ; : < : s
l :
: il ,r
AL. ',:):., > : eler: j E
)
• • '
f N - ' t . ' f ' ' -"
»/ . 'L k/ . " .
,
" ie r ,- j, )Ri.r\ < , )ti. r.;v, ilat , il
:i.'
" r t :.,,;},-•. : iiSzta [F, iv-j if
j i». , 1
1
Li » J
CH
" /
: i ../:/.:.; /}
, t i / i / , b L .^i "i
/ L / ,i . < L F T / : /
/ 1 /
1 L L r T = L .it
L • 3
ELL 'cM. r _
:
.: '-l : . l '. r t . ir sor
:;> j rj .__ jse._ :
L •» .* L F T = L -.-Li- " * L
l. .. !1 = i_.JPL.F1 f i
.FT = L .*LFT * ILC JG• T =
J .: T * ',
l\ = E L * 1
•
.- » . ) F. L ]
s l*RLFTB2 5 = L.JF • iT
ifcT
( -.».-., .)[': = JPKLFT
lr v ........ .•'_ :) EL :E:NT(£L-1Nci.^ i-\CCl2)
= Tl '
. j- * 1%
: T ( E L I • EL) = L MR K rl
T
. '
I Fl . ," -DE? ' = J >S . iT
r ( e i !'n , ) e :j i = j > s l f r
~. . . . )o) ;n r l o
-.
"'.F ( : L . , JEl )
a L L 1
.• T I U L 1 . - - = L - T
ir
3
\ S S I . i m .•» E r
, T
p p ,- H T
• - . J b L E ' 1 E .\ T ; _, Y C L I I I i . \_: t j j f ,
•.
"
LAST = ELM.«ir - 1
EL *.E::T( LAST, [_')E2) = EL'lE. iT { LA oT ,,\C )l2 ) - 'TLCiVG...... ; I
'- !"( EL •{Nri^t.4 JOE1 |arL^r;jT( LAST, iuDE: L U i T { R L i . E 2 J = L : « i T ( E L ? \ 2 < , m o •!# E ?. J - . L U \ GI
•
! .. _...h7E*!)ELME ; T(- L 1MB*, I IDE 3 ) = EL IF- IT ( LAST , fiCO
[ TV
_ )
E2 )
...T .
L
.:>_.. i
T: .:
• :: l -_ : riSTaucTS T^
~ w;ut cctmecti vitv iatrix i i array» ACKj IT I .NT \h -"
. ISTA.U C.JNTA I."
I S STARTING1 T I
:•: \ ; i •,
. ,TI / 1rv. -
K ; .1 5TCI- ES
5 C
-< i
s
. I v i T Y ii J 2 S F
T,\ j.!i 1ST \KT ,!
']Jc .
_ -
' : -• EL »3R,\GCELi NC^E2tNG )SJ3 t N33CHK,C !0L, I )EX• / : '1/ ,i: . ti'J'icUTf'HtH JR, PHI 1AP ,.' •1,N I \PL, [FLG
: : • •
. / c <:/...< -: ^ r i: -i 3 » 3 i
,
p h i u i f »
z
t,vj i f t o i vd i
f
\- l/CMl/ ;U.I(156) f ! >TA4T( lio ) t.-' .(1^--!
1/ IPK I-NT/IPRNT 2))t.<AVE? )» -U t R •? ,LE APHR RUN'>!8 r
<
IS I j: j'
I E : L 3 j ,'
)
INITIALIZE ,'JI TG I, lA'lc IG l a.\l :3NSGC'GT I VEL>.
" • - - . ST : !. i '. ,
- iA.lE.l 1 i GO E = L i i C ;N j ) E
\ l • 1 ( r ! > E J = i
•\*E(.'kJDEtL J = iGCE
__ TrlE REGAINING IHLU.MNS OF 4A^EL^ COL = 2 1 7^'n I )E »CCLJ =
CC.JSTKUCT Tf-E CUMNECTIVITY 4ATRIX M\.4Ej.: 50 EMNaa = i , \ jcl.-i
t
jC '> icosja = i, ^
IG0E1 = EL -1 c :iT( ELM NnRtNGiJSU J J
;HECi\ MODE iJJSCRI pTS; IF iJjAL LGC?i ELSE CONTINUE3C 5'» 'iGOCHK = I, 3
if r cjchk. c }.<::> s j >) gu t, 50
)C2 = c^ IEMT( cL WIS a r\~ )CHK )
: ; E <z\ T \,:';:l 1 cC ;.N 4ECTIV ITY ZCLt*
r
NvJ*CiO-)ElJ = NJM( NCDEl) f I
rC ' r - UPLICATE 1GGE IUM3fc.<S IN I41E>_ -«:t = \um(nol)E1)
: 3C ]L = 2 r LAST( i.. E( JEl.COL). EJ.NJ0E2) \JM( N 3UE1) «NUM( MJDEi )-l
[F ( !AME(MU3E1,CQL ).E^.M0E2 J 5G TC 50IF I IA1E( .^H.OJD.^ :.,;) UME(,>lCCELf CJL 1 = MCDL2; : : \ t i . j _
i°act the ; . i .= :t; / ity ^a:^ ix[$T.\KT( 1)»1DC " ! MODE = U NCNHOEIc
! IGOE .,4E. MGN IDE) I START ( NGQE+l ) = l START (NODE ) *NJM ( MGCE)L \ST = MJM( MODE)j.: ro I \J EX = I, LAST1
> T- = [$TA IT (MODE J * I IOEX - I
. . •'•£( IPTR ) = 'A -IE (v
! )E, I 1DEX)CC'iT i
THE I ? Ua T IE C INNEC TIVITY MAT U x
1-( £P«>.NT( 11. EJ.il CALL uJTP'JT(L)
IET: .
)
12 )
3 J i » T I'll: ., J T 1
: in r i .: - : it : \ l : :
;
: r -" in.j'ijiro . , \y
: .-:.. * :: .
rj in - : 10 ^o five.."."•
..: "
rr .
-: , . j ,•: \
i;;:- [j ,• > : ' '
: -: • eu. '
: . .-
I b n t n z z p. c F :
T ^-~ >:'
iO , ! y
•' , -: r - ' ,JR
r : :.j ?".
c
.\ .. ML .
: - ;t -. : • )
:C-1 1 i/ : ' 1/ • - < 'E » . ELMT , i;i,Huja f Ph
: r.
1. 1J. i /C
'1 5 / [ d J ( L 5 6 ) r .
J-3 P r i > I T f< I ( i 3 o ) »
I 3 w »
v
<
' I f it «Pr-w, u. AT , sil_f ;
<i( i 2 ) »Y1 ( 1 w
» .
r " L >J
,3T
j _ r SQ'JTHrlc * d NOA^.Y PCI !lj rd- : = 1, IL3 4i".
;nu:Ej = 5
i s r< t » ; j l ; + l
!G - L1 5 T > = 1 3 r ; r ^ m Ln; . JO UDE = tSTRT, ISTu?I . |:J : ) * 4
s: r if3-duuj :a*y pc i ;ts n o
I 3 T T = [ S T 3 P + L
1ST 3 = N •." )E - 2*NL It GISr^T, I 3T iP
SET . . rn 1C HJArf Y - 1 RJ[ S T* T = I S T C P * L
I5T3? = I^<T +• 'ILTIj - L
>0 IU0E = ISTRT, I3T. J
I c C ( M J E J = '*
ro
j- T <THREN 3C JNCA^Y P
[ S T? T = I 3 T : p t i
1 1
5 - - [jTU )EE) - 5
e r j
131
j * .
" j r:
•.
- : ,\ \n a
l
-V4AL - : \LC ,LA T -;> KEY fA.UABLES [N TIE IHaNNEL
. / : ! 1 \ / TH E T A , T «E TAU , 0^ L T ; < f ;cl T W t , l A T , N L ''.I G , 1 f
, L 7 / s m I ,-:.,:.. j I I.
1: = .i*i
f f 5 L » < L \ t A t ) i a .
- "• - EAri
r<• 3 »',:;l\" VELJC LTY
.-7 .27220 5c- )5_ .'= iL . r
'>._.. j = f L . .•
; s = p i / l i .
7 - i r . - -- - . <
c l« "> L:)viek l\t itudinal jcui ::a>\
r•
'-J \l - THE MAINEL ' S IP PER LATIT'JJI IAL BOUNDARY
' - :: rV J=TH6T kU"* " \DC : •
IC /ERT FRU1 SPHERICAL T3 IERCATC8 4AP PRO JECTI JNS< = C J3(THETAj)/{ l.-3IN{TriETAU) )
•' 3 = -^l :j ( o *\
K=COS(ThETA)/(l.-SI J(THETA) )
y- ::-al ] , (
<
»*a
. - IIOTH 3F r IE C 1 XMNEL IN HE rt <3. = r
rj >-v^ :r
ELTAY - 3ECMETRIC SPACE BETWEEN NORTH/SOUTH GRID ^TS_TAY=w/ UNLAT
'--.. \Q - CHANNEL 1 S 113 LATITUDE VALUE IN RADIANSr -1
r \0~ ( THETAU-T'HETA) /2. f rHETA
)tL r ^ - iECMET^IC SPACE BETWEEN EAST/WEST 3RIO PTS)HLTAX= )ELTAY/COS ( T IETAU)CI Al = 5ELTAXf.'I 4 = )ELT*Y
iL - THE HORIZONTAL DISTANCE BETWEEN GRID PCINTb (,<M)
< L « A.N L JNG *D E LT UXL1 - rHE HORIZONTAL DISTANCE BETW.JEEN ThF 110'MiEL iRIO POINTS<L1=XL*CUS( THETAO
)
ii*l ? C.:S( T IETAO I
(s(^.*Pl/XLL)*'*2+( lJ l/.U)lt*2
:.;riolis value for ruE iid channel, latitude= : = 2.3 * :.me;a * >im( thetao j
< ETUP.N_ 1
122
5 ,
: \ i
IJT [ .E LOCATE
- : 1 L C J L A T E S 7 H ; < , ( , -. ; ) > I .\T
j
u it.
; ic ,_
),0)
j i 3
;'
.
7 . ,
J E S f-
1 ! ^R"1 > ." I
- L
;
z S
' :. \':.i
( \ >LC
"
• . M 'U .
:L.M f f iVi f ttJ J*: T
i / . : . ^ E ,
Ji J/C'H A /THcT A, T.-tE TAJ, OELT AX ,
/ T H ET A i Xu , .i , V b L , Y T 3 ? ,
; L.I - i ' 1 / C 1? / a LO ' J j (L5o) f YL4r(l'5 6j; • : ;/ i
-; i .r/: »a ;
t: 2 n , ,a; -\., ,
>
» 1 = >. Ltl J3j . . .
= J. / : * )
.
-. . = ).j ' JELTAX
; l ; "l .r : r - c
; 1 Q ,r ESc7 ; u_ ieaAi )a= a , • ^ j,oc ;,
HC*V*XH.-WA. iT<=NLAT + l
r h : r
1 ,-:
i l T
rLEA
I 1I » 1
L A T , \
-"
L .0 » J
t j 1 m r »•'
" ( 1 5 a
1 J
j : >o i = l t l-
'
10 J =i , ii )N:
<LG iG ( <) = XXt X^= X XO EL T * A". \ r (
=v
: = x 1
r :
i
. jer = Y >E L TA YFACT = -PACT< : = < - FACT:
.
'
" 1 .
.
» ; .
r.
:tE :nc j* \ d cartesian coo^o; .
J - IT ( 3 ) . E «. 1 ) CALL JUT PUT (3 )
i . : :ui ;
i
= r-. av j
,S - T ' - ;SF .-.
'S THE JNIFOR.l C )Gf< C INATF. SYSTE* ( < , i )
V 5 *T <L V /ARY -NIG CGGPJINATt i . >TE ! . THIS ROUTINE: •
. r*L ! .u ^ " c. , iMLY THE /* [ AOLF "'A F c
L \= F 'I
J r _,- /n
.: ) SUjaJ'JTI e
: \ r /
i
: A L - L X , L Y » X i. , X 2\L*"J - : .
;~*
:' '2 - L ' - T , NU 1 , I S T A R T , h A i c , I H ')
' / CM 1 / . )0 £ , it ! L L:
r , J . , ri - ) 3 iD H [ i A * , ] 11 , N ,
> 1 , [ F i.
. I/C II V/THFTA , TH6TAU, }ELT AX t -DSL TAY , ML AT ,NL'J 1G.JT; 16/ I lETAQ . <L , .-. , v -L, YT j^ ,Y = C* , < L a
m :a , Z, JGAri, FOM/C'7/XLING U^o) ,YLAT(15o) tAMJ,S-*CC]EF, SU1-M iSo)
; :* ' i/c:;i7/ vip;» y-iin,f :gthr
/[-•I T/IPRNT{20) ,.-jAVE>nOi Xl,R2,LEAP:iR ,RU-\N3S- ibl.- UEE( 11)
ii (XI. £-.).- 1.0] RETURN[ f
( ( Rl .E j . 1 .3 » . A \'0. ( R2 . E J . 1 . J ) J R ET'JSL< = ;ELTAX * FL j A T ( MLGNGJL r = j_lT W « FLOAT! JLAT)XI = > . 2 8 3 1 iCt I L <
< I = 6 . 2 3 3 1 So / L Y1*((:U-L-1)/(P1 + L.3))/K1i = (I R2 - l.) ) / I P2 * 1.0 )) / K2
Ji 10 IQOE = I, NGNHDE< < = xl :\ .ii ^goe)^ i = YL^T(N :JE)< L C I 1 00 E ) = < < +- \ * CO S ( X 1 * X X J - AXLG iG(NODE) = (X - ; u*xx
)
YLAT{ MOOc) = YY «• 3 * Sl>l< X.2* i' < )
i- i :J i it (^ ) . i }.i ) ;all jjivuim
'
: .= L.a :+) j
»0 < < = It IL )NG= <L JNGtftX) -XL' JG( N <-l )
II 1 . 3 T . X ] X . 1 1' = ) X
: . I T I : JU E
: = i.o e+ -j
)C 2b JY = L t a. ATJY = i'lATI IY*r\JLCNG+l) - /LAT( \|Y«NLONG)[f tYMIfJ.GT.OY) YMI.N = Ji
iTliNUE
F^k X j,< Y iXIS TRA:1SFCR-1AT1C.M CC3PPUTE lu^ l:T• r TI D , ')/ (L2APHR i h 3K 3 4P> f:s t
< r 1 i
v ' = XMIN[F ^v «II .'..Li.Ali'.l KY 4IN = r -II ,
,' - ( X Y M I N / ( \ o S ( 'J 3 A R ) « ( n . 0« * . 5 ) ) 1*0. \
FACT. • = FCSTiR / (JT/3oOO.) ) * 1.0L : •
' rl = [FIXtFACTCP J
<
:r
( iPk:iT (^) .Ei.l )
'. . us I . \TE S AFTEw r,<A.N«SF jR,
*«LL JUT?UT(4)TIG .
:.-.-. i
Ti \ 3 .-; i -> t
1 JT - r .- c T•
\ PT '
I-
\
.- L » "i
- • » \ J' _ * L_ * - ) •\
rr
;
.
-•
- if N T f
*
'./ l i 1/ . J . J )E »
_
•' '
-- 1 w TA ?
: ' •./.:'/. iet . .
./.'.' .... ,
: i / cm 1 7 / x -i i ,
;-••../:'-; ,t/- >
. j: iiii :m sc^lf i
• > C «L. E _ / 1 . " * I
; : \l '.-/ .5 f l
IF ( 1. .'"..- L. -)) :
:"
' » <
T:: SRI J /ALU
. ,' - =
'IN = DELTAY«.
'
: = DELTAX1L \T?1 = ML \T » 1
IC 20 J « Li IL >T
iCDE = MC DE 1
(i::Gl jUl-I = KLO)E1 = 100EL I = L i LI
;e = MODE * 1
.r * = : : .l
; :(
i
)NG ( M E ) = F A CDIFX « <LCMG(N
[ F X . L T . X -1 1 ; m )
TIN.JEr ZNJE
1 - 'I-
\ ,-ilC lSTA,DL ISHFO; mi; •:>.-. r*c hffe^pnt 1
-'
[. .
•. f
r — : . - r s .
J ,:.:', T ,MA IE f IB J
:L.\ r t !\N f riOUR t Ph[ ;.i-\i< » i.M 1 ,M f
' ^..M^Ll^tjCL'*^ iLAT,NL<l i ,VE L t Y r J.3 , Y3CT f XLA IDA »6»L5o I »YLmT( 15 !>) ,A..V>, S :QCF,SH'II . » f-
. j i i
r 20) , wav 'E.i - , .,•:,-. : *? '•" ,a ''
jhb
»li
j A r>
i L 3
GATE5, >< E
l :
6 )
5 1 1. i ) i • j > .5 •!>»+3,3.5/) f l o . n /
r j <•;
pi
NG( NODE)
)*iJELTAXXLCNG(-IDDEI)
) - XL )HG( if.JF-1 )
CM I • = J I F X
< P,R Y »XIS TRANSPUT IATICN CC V RJTE NEW - T
(LEAPHK) FIJI A »3 •. FCST« D
I r-• \T I CM NO 1 IE
• •
: . = (iint f ( y 1in.lt .
\'a v.)) t » ( x y.m i n / ( abs ( '.j
:; T * FCSTH /
l . '. >H < = IF IX (F AC
->"[ ,t x , y :ooroi
X <"' IN = Y M I
BAP )*{6,0**.5) ) I *0. 5
( J T / 3oOO . J + 1.0T'^ )
( [PkNT ( -) . Jj. I ) CALL OUTPUT (4)
* JKN_ ^
125
)T :>
.
:- \
-- :alc : jlatgs i*he area gves each fle -ient, using- L
• (AL3 : . \ \ ..:> r . XJKI\J T.
z T^ A , j ri
;' M 1G N • c7 • EEN
. • TLSIA ,. H' V 1-4 Ci iRuI.\aTcS (LI iL2 i L3 ) 1 r,:E '»
-- " : ? j~
. "\CH . L ' ELE IF. Nl iR: " T _"";
1
'
:
DNS FRG' 4 CARTES I
= I, V ELMTT < ANSH IRMAT ll
: 1
1
;•. 3JE2 = :L 'E IT! :L"'i <, iJ0E2)X.\G0E3 = El *E -»r( :; L^NB <
, jUOE > )
- » lONGI v>! IDE! )
Ac = <LCNG(<NQOE2
)
< ; = <LONG( <V. >=3 )
lu A . E - C G G R C I .ME
mt: r he a cA i NOJE1 »EL 1 idP )
\\ IC )E2 ,EL"1N3R)lu ;E3 » EL INdR
)
;EFFICIE.>lTS= O - x2= XI - X3= <2 - XI
(HORIZONTAL TRANSFORM AT IuN )
-
•
1- 11
:
\
1-
! F
* ECT E.Md C3EFF ICIENTS;( IC J T.EQ.(?*NL "4-3-1 ) J
( : : >'.:. EQ.l 2*nlcn ;-ij )
I
r
: . . r .- .. (2*nlgn j) ;
. ; JN T . E Q . ( 2 * N L CN *M(ICO JNT .
: }.(4*NL 1N3-1) )
( IC\JUMT.5 3.(4*NL0 :4>1))( [ COUNT. c-J.( 4*M MG) J
( (C iUNT.Sj.i **nl ING) J
( iCGUNT.E3.( ^*.\LG'>IGJ ) I ;
)je rj cyclicMl f EL .'IN }R) =M 3 , E L MN ' ? )
=
\(1
,
= lmn-3K ;=W2 ,EL*N3K>~A( L i EL 'N i<) =
\( i * EL IN }R )=
, E L V \3 R ) =4M 1R) =1
i
QU IT =
C G ,*i T I N
X3 - X< - x 1
<3 - '
X - X 3
X3 - <
X - XLXL - X
X + X2
I TY
- A C
- XL
CC'P'JTc THE 3 CC6\ IGOE1 fFL INBP)
E2 , E L ' 1 \ 8 R J
ri I IJUE3 ?ELJ
.l>R J
EFFICIENTS (VERTICAL TRAMSFCR= YLAT(KNGDE2) - YL AT ( K.NOPF3 )
= YLATU 1 DE3J - YLAT( KN 5DE L)= YL UU i J jEL ) - ri_AT( K IODE2)
i i
- i -.2 is :" :
! i N A T E.2(EL.'4i\a
-: ue
iVEs rHE ELEMENT, THE: e .
T ' "•
: exFI rC THE '.• _ ,
)F THE : :
r F - !
.
- WfvL
OE, !L IflBR ) *3( '. tc L
L.N J
r ) )
-Ei. I
E
- >t :
• •
-
.) > ^lEEDF'CJEL. TMESE \ = ML : 3EPE"UE>n GF J['~: SH THEY At
. -,. .L > - • JiljT •
'--1 [| , rr«E i 1 1 T UL[7.\ri \ PHA3E
r :•
. t ; jc'M7at:c f "; _ . . , , i ; : ,s
: 'OL!'". ATE , ji C'iUCIAL T:- T-^. _ - : .
" :; i a jj - •. . : tk fim :te -••_.:' - .r >
r: rHcn .
-r r
_<_- iTi \ [; ' :•. ' Tr MPLEX FG INCLUDE KE2E. S»_ E THE
iC :; T F . ) E T A I L 3
.
k L * 3 ru:Z IJ3/C.U/ iO'JO-^E i UEL.MT1 iN,H,iUK f PHI"3ARtNMl l J , \ P 1 , IFLj... . / ; a > / ,[j,;j^,it5 fi:-o),A<E^(:;))
• :n/c.*°/v JX i 3 1 2 i 1 ) f vj
y
( 3 1 2 i h ) f v u x ( 3 1 3 , 3 , 2 1 a
)
. / ; L 'V V : J ( 2 * -3 ) , V I j K, ( 3 , 3 f 3 , 2 d .5 ) , V U Y ( 3 , 3 f 3 , 2 o 3 )
'" . : L 5 / 7J X I X ( 3 • i f 2 J 3 ) f y J ri X ( 3 f 3 » 2 3 3 )
' ' 1/ C 11 3 A/ 7J < I Y ( 3 , 3 f 2 3 3 ) f V J V I Y { j , 3 , 2 3 3 )
}^/ I?Rt\T/! MNTI20) , .,AVEMJ, L L *! 2 , LeAPhR f RU !
J.\ 3H
[NJtX = .J
3 I LOO [ = 1, \CEL '.
r
A4 ^ AREA2( I )*?/IJ(I )=AR5A2( I)/2t.PVIJK - AREA2U ) / 12 3.0
3 C » J * 1 ,
3
/JX(
J
f I »=d( JtI)/6 .
/JY(J,I)=A(J f I)/S.
)C j < = 1 ,l
/JXIX(JfKfI)•/ j v : x ( j , n f i )
/JXIY(J Y KtII/JYIY(JtKtl)< \ = 1 .
IF ( J . F. Q . K
J
< < = <-,/:-.
UJfD*B(i<fI)/A4k( J,
I
) *BlX, IJ/A4)( Jt Il*A(KtI )/A4
I *A ( K , I ) / \ 4= \ < J ,
<N=2.
I b ) 1 = 1 t 3/UXl J,K, 4 f I )»0C*f i)«\X/I JY( J,K, M, I)=A(4 ,1) *XX< 1 = L.I: [ ( J . EQ . K ) .GR -
(
J . EQ . M ) . GR . ( \ . EIF i'j.E,.n).Vi'.(^.F].!)) XI =
. "/I JK( J v Kt'lf 11 = 'VI JK * XI:; ( n-:iE
r ICiTINJE: :
r in je
;.•')) < 1 = 2
F ( IP* .
T( 5 1 ) LL JJTPUT (3)
in
J -, l) \ JT i\ : * 1^ XL
<i - :C*-i5TPJCT5 T ^~ COEFFICIENT v ATrUX F03 THE L.h.S..T-^tX EJ..JATICN3- THESE COEFFICIENTS IGNSIST IF
:
: : ;T <VJ, •/[>. FGK -' • ELE-1EMT I i l"HE I A I
N
"• :
•i T I h ! ^
1' IX,
_LL
: Z F F I C [ E > -
E /A Li
cv r t v j > = )
<v
I
, V J> = AHt42/12c ; 1 1 v j > = ---A:/i*
f c •> J.
: = j
I <> j
-
s
E S-L
»
E F F I C: :
: H.LE
: :NTS
].
5-. L
TCPI
I*" -:
<-'
1 EN T
)UB \kE PASSED T
1* bbBPIATRIX I I TO T ^E SHAPE
T t J
[ :\ TE } c E L *t i\6 RI- '
• ". j/C.U/ ,..^ti', ELMT , 'M,rCNrh; mm vL ,N ,\" L r
IC-UCN/C^/ELIEV'T (2^3,3) ,PH IDI F , ZTAO IF , I VD IFI C 'IG.N/ C.UC / V I J ( 2 3a ) , V I JK ( 3 , 3 , 3 , 233 ) , V I J Y ( 3 , 3 , 3 , 2d
; i/C %U3/H( 1044) ,0(1 )44) ,M L-i4<t):: I ICN/C'U**/ 33(3, 3), C( 10+4)
-, , ! l - 'ASS 'ATkIX GJ I = 1, .
,
3(1) = 3.0J! 50 ELMN3R = 1 , Mi iL M T
) C ) I G E I = 1 , 3
L ) ICOE J = 1, 3
= L.
T - iT F I« )I4 iCN a -
,:: nE-^SL F (NO Jcl . E .- ,N JOE J ) <N = 2. ")
B :. I . J )z I f IdQEJ )
s tf[J(EL-1ftt)K)*X.NI ITLNJEC -LL AS E i3L (EL IN BR , 5 )
c cT
: i j
e
. E : .
;
IFLG
l i_ -
Si ' " . r : .= ^s : ".l ( :i \hi r,c )
\A V FC?;t..\ J',, tmE
l r i ;r .- T a,\g ~
; •
• - - ,u thi i . I "... :
ELMNdR ~ ELEMENT 1 3E FORKED Ci
'•1ASS 1A T? I X
i IGOES CF -L \U .-
I VjT A < jj <ENT 5
C - rtuKK \KRAY
[\TEGG02 ELME.-T, N!U I, I START ,NAHEINTEGER EL 1.\3R ? STA,U: '
'
, I
' : ' 2 /zL -'4 iT(233 f 3JtPHIDIFiZTA")[FfDIV0I
CC^M/C'i/,JM!^),I STAR! { L5o) f NA 't I I W-+): ;h .. ,/c u v/ 3 •(.>,?) , : <( 10*4 )
: : •'. oi gn :(i >*4j
LCCATE THE :. . ECTIVIH LIST FT* TH50 [\00E = It 3
:0E1 = ELHENT( ELMN JR f UiCOE).AJT = \UM(.\G0EI )
' - T = 1ST iRTtNGOEI ) - 1
.
.
,~ d:;e ;f elmnsr* s ngjes, : ml• ) J
' " ) E = L , 3
>E J = E L 1 E.'JT ( 3 L r*N 3R , J, IJO E )
... 5 THRU ..jH'j C NNtCTlVITY, FI'IO lATCh *ITH30 LO I JOEX = 1 , LAST
. "... = iT ^RT + INJEX
~. . .
= >JA1E( iXTNOJ )
: IGJEJ .NE.TiT.'JCO J iO Tu 10
a 1ATCH [S FCJ.VO BETWEEN MGGEJ *.NC rST\HO; ITS INjouct stdred in v, is :,,fli<4; re ruE iass *iat.<i
CiNXTNGOl = 53( INCQE, JNGOE) C(NXTNOJ);; rn jo
, i
, v: j
iGi5ej
1NE *
a ;
lc c : .r
I ! JE•
• n-NJE: ; . i r i
•
j e
: t j
l£ i
JT
i f
-IScT
;lt - r;< . . ;: fctly
. : I- IS 3 i- ]
; ro i'.T
\: -
.J .-- , JLxL )
iIS " . •: L . I-
G 4 3L0E U IS-jI J^" r
POL ml * /A" I ABLE.-
i HE i a'
i
)
' . I ; A N A
i
hjuT GaLGJLaTES T h E 3L.K^\.^LAr>= ). I *G
UP CcKT J i
5 JL JT ION F I
A i.AL fllZl «
r
. mT ( :
)
r> r \ T
rE " ' : . - r t \-js istart, .'iA.iEt 130'
./ : '•: / . ' e , el'*»t , "... , - .j- t ?rVC'Ua/THETA , T !ETA.j,OElTAX rOEL
- .. ./: 12/ ll : .t c ' > ,3 j ,. j-t i . ;r- , it a
' ' l/C «3/> JM( 1561 , i s T ( i 5 6 ] , N A
'4 Y t*
I L A T , \ LIF, 31 vj IF
1 , I F L GG i j i
1) )
IZ i/CMS/ IdJ( l«56),>nPT,NT4l( 156) ,XL( 1
: : «. i/c 17/ XL T : ^ ( 1 56 ) , YL AT ( 1 bo ) , V«P , 5
: '
• 'J / CM 3 / A ( 3 , 2 3 8 ) , 5 ( 3 , 2 o 6 ) ,
JI.^EMSI Jf <XX(3),YYY(3)\ <. ( I . .; )
) , Y 1 { 13)E F , SJ P. ( L 56 )
DO C:
: >s : = ..) ) >ooiX L = ) E L T A X *N L ! \Ga y * XL - >£LTA x +• p ; c r
= iLAT 1
3 T= NGNGOE
v; = 0.3
10 J * L,V'V U
fliJl - ^sv S = irs ; f l ^ \ Y< s = ). :
20 I = 1 , NL ': J
A L ( [) = XS< i = < "> •- j f L T \
1 I=
Jl = I
100 I = 1 , ' »p r
< < = <1(U),' V = Y 1 ( J 1 1
- 3 = 1, i-.l "L V T• = LOO I * kF E^ :( J )
. JO j< - L, 3
i- xl ::. ;( El <- ;r (j 'J KJ )
•
( X 2 . GT . XA. A. J D. 11 • ) . 1 ) A 2=X2-XL1 -
( <2.
L
r .-E J SI . A N G . i l.Lil. 1 L . ! G ) < 2 = <2 + <LX <- <
: ( J S ) = < 2•
( JKJ = * LAT( •_ L '* 2 IT ( J , J.N) )
, «. = AM A X 1 ( X> ML) rX XX12) , a X X ( 3 ) )"
i = A ' •
j i ( < x X ( 1
J
, X X X { 2 ) , X X < ( 3 ) )
y •'!.a * AMAXKYY Y( 1) ,Y ^ v (2) , YY r ( 3 ) )
v ,; . = AMINKYYY ,YYY{2) , t\ U 3 ) )
; (XX.GT.X.*AX) j U'
^Di=
( •• A . L r. A ' I i) J o TT 4.
j
IF (YY.GT.Y'1\X1 DG T 1 -O[ Y Y . LT . i ,* I M
)
-
; J TO 4 3'
4 L = VJ S ( XX M YY Y ( 3) _ / YY( 1) )- Y Y •*
( X < <( ) )-<-<<( 1 ) )
f < s < ( i
)
^v vv( 1 ) - XX X ( i ) * Y Y Y ( 3) )
' = '. , j ( «. \ '(v v Y{ 2) -Y fY(3)l - rr"*(xx 3(3 j - x x x { u )
f \ ^ * 1 2
)
*YYY{ : ) - .< s X ( 3 )fe Y Y Y ( ?) )
• = AVS(XXMYW -V YY{2J )- Y Y * ( X < X ( 1 ) - X X X ( 2 ) )
f x x .< 1 1 j *YYY (2 ) - XX X( 2 ) *YYY
(
1) )
- Ai * A 2 * \3I- ' ' - . JT.AREA) JC r -i ^
:
i\- ;
1 ( 1 ) = J
, -
r' 5
"
. JE= 11+1[1 .GT • »LU • ) Ji = ji f i
'
( I 1 . G •.
'_ ) I i = l
; n je. r
E\3.
1 ^0
i L .: IT TIE
jp r -: : .
i
t i \ll H I .
~L T,^ )E D C'I0£MT VAa I A3L c i
. _
-. .
ii/Ci^/ iS( 15;. .
'/'& /thetao/ C '1 7 / < L ' '•
i
•: :' • /cui/uii
' ; :/ 'lc r (::•'" / IPF I -iT/I PF
in — ..
;fc : : :i re^TiAL h C ;~i T
-. :
. -
\ N = _. ' * 5. L«tl5
MIJ - H \Lr In] C- ID = n/2.0 « ' J
> : A = vn-')0 • J. =.4LAT+l
: 4PJT6 rHE Dili. .
' I'HTIALIZE AL1 = 1,3
K = ]
f a VLATl
U
/XL
HViilEL li-JTh
AL FIELDS FOP, J, V UJO FI P E r^ |UDEL r^E: sjv< v_CTC :
S FO'-? EiCh F1EL.J
FA i i JQK I Z CNT A L I L A T I TJ J I N A L )
)c io :=l,-,t. L=YK*Y
L. J - .. f * t_ J I )
< = <L"
' J(K)= X K X * * A
•Vii . , l) = F-">*AMPMJ [ . • = >HltlAi
[ < A ) = PHlUr'M" - • = - I i
"; - j - a
* ( f i >
:- = - • - i/(y
= :te-*h*si: - i ( \ j = ) . o)iv(k, n = ^-, '*
>AST(X, iJ a PHI (K» 4 S T ( < , ?. 1 = *S I ( <>\ST(!<,3) b :nl(K
= K+lL C : u M T I IE
[Fui + D.Gr. r i :
f = YLAT( (I + L ) * 1L "
2: ...::,•:
ea ;.-i r-i :"jf pe ; nc i :: j iT a ' - N
SI H =1 LJ I «C0S(FI2I-F0*lPH I ( K , <
)
/Fl•'.- (K MYK**2*XK**2) i *AAP >/k 2 ) * P l *K * I YK * *2 * X*.* *2 ) J
K**2 i-XK**2 )
\( --I i ) "*S I J (F 12 )
51 I { F 1
1
) * S I I ( F I 2 )
,•')
)
)
C TO 2 1
»+( Y-Y-1 ID) «-GH
•
= >4
i i \- / < .
ro
; in jej :
'
I = i , 3)C 70 J b i,«!C
-: j , I ) = > i : ( j ,
[r (:.-- j [ n .
- 1 r i-..i
• i
i )- > a a s
) : A L L J j T > .r
( 7 )
1 3 L
I i i \-
^ 7.
3 .
- - L i _T •
:
:5 iLL T -\Z )j=PEAND .stT:n EACH TIME EMTE3." ATICN.
: . . .: L'
'
v
/ C ". */ •
: 1 5 h ) t A '' \ - i J
I I ) J t
I 5 .5 ,.'.
! hi s ;k ,'•'! ,N L , [FLU
i.>ii L so) ,?AjT( 1-5
; ;u d i
'
\
-. /
:« i •> / : .'
(
: i/
1
3 k i \ t /
:
)i _.„;.:. ztu I
1[ Z i T A L X
: 2 : = i,1
I JOE) = J'
: • l .,-•:
[ P 5 I
CALL SOLVE.- ( U i
W L 5.->, J),JlV(I56fj), JHHl:S2lFS.Wt;(l
> ) , ; 7 ( 156) i A LF A ( 1 5 -> ) , 3 E T A { L 56 J , U. iA
1
56 ) tJIVKG( 15 )
1 15 . )
CCMPQ jE .r
IGCE)r C H 1 , 2 , n J
/ , • - : Z ,x
i T AL V ^ ! 1
:ali jvrhs tc.iCALL ; l ; E a (v,
,PSI,2,1)It 2)
! T I C i\
'I .
TI c
*
KE i i\ INET IC E IE
CALL KE3HSCALL S J L / 3 •'. ( XK
Z3 r., 7CRTICITV
C-LL - IS I3 s
;
call j"L; c ^ c t
317, j I VER :. v c:CALL V ) RH S ( : 1
1
:all ; jl ( )i
[\itial he •:-:-
IU3E = I,:: r -. i socEf 'U)
IQEfflPl) =
: fl ;. E j. i)IL = 1, 3
I 20 ICOE » I
i
•: i , )
•: i = j (
LzTAi lOOEt <L J =
IJ.JEtNL) =
JJ, J*( £2TA f c
: A L L JU VR H i (• J
)
; ' - . SOLVE- ( , J
. / /*(ZETA * FI A L L * J VR H S ( y J
:4LL SILVER ( v'/
ALFA, J*PHI: i LL A 6 RH S ( J
:
: A L L 3 1 L / E: (A
r A f . v * D1
1
. •__ \3RHS (7 )_CALL SJLVER ( *3E
IF (IFl3.cv.1I[ F i
:-> < iT ( -i ) . : 3
( ( I > <'
: T ( 1 5 ) .
= 1
RGY (J**2 * 7**2 J/2
2 , , 3 )
[ DEL**2)*PSI, 1 F L 3 , L )
A t 1 t * )
( CEL**2 ) »oHi, I
C L3, 2 )
,/P GiOf 5J
SSAP.Y TI IE LEVEL j n-;NCNt: )E
= ZTU ICDE)31 /• 3(\.,JE)
;o : ) 30
IC.NJJE^ODEJZ T A { i )0E))I yR3(N .JE)
Z^ TA AND )I V
t n ,
-
3 )
)
rlf 7)
= A , , 1 )
TA,i,9J
GO T~ 50. L ) ;UL jJTPUT ( n )
ZQ .1) . ANJ.( I PRNT ( L9) :;.i)) CALL PLCT
J) c
So3 >T I" E JV».HS ( DLt 02 i II » I i I i < )
»S - :^LCJLAT > E IT iC ME .j jr v c ; ip . if
1.4? UT \ ' 3u if IT S ,
; = D C ri I / . j X «• 3 ? S I
\ T C F i C- 3= !T .
>o>.. - - . . _
' c .-> r . - r v a ": ' ; I
-
ISIG.N
- iJ E '
• l C ; T V >*< ; CHI
"IT L c v'LL F "JR PS I
- sign :f r ic
>
a\'i; c i
:
»
I c •
.
ix
" cge <*2 - r
'ic el: i » i <
»
, i IIJR , PH I >AK r'
I .
. ML ./:!?/ J. i - If (2 38 ,3 J ,PnlOIF, ZTAMF, )IV ! IF,P1, IFL
. " U/C / ' 5c ) , a Im.oJ i L56) » -> * j SI I 15e) ~) A•
I
".
' Jli/ !9/ V JX ( 3 t 2-3 < ) , V J Y I 3 1 23 3 ) » W.J X ( 3 , 3 , 3 , 2 E
I «t 2SICN )1( L5fa) , )2( L56)
:all • -s.t
3j
SIGN = L.OIF ( ISIGN.EG.O) S I V. = -L.0
iLCJLATE FOR EACH !O0E 3 ca LLE4ENTLT I L '.. - = Li n jelmt
! = =L"4EMT(FLMN3:<tll• = EL 4ENTI EL •'•. >- ,2)
:DE: = EL IENTI rL 1N U,3 )
:C-<P ; JTE )E*WATIVE FG3 BOTH THE STREAM FUNCTION ANJ/ELJI I TV j |TE ITIAL \N 3 ADD" • > = SI j • * I Jl (NGJcl) *VJY(1,EL.M N3R)
- !1 (NO 3 1Z ) *VJ • ( 2tELMNBR)* *0L (NC0E3)*VJY(3,EL.MN3ft ) 1
t- ( l( Kr )ELJ *VJX( L »EL V N3 !U* )2 ( N JE2 ) *VJ <l 2 » ELHN Ji<)
»-02(NG0E3) :*VJX(i ,ELMNir{) J
VCJ ESULTS I, *CP'< 7ECT JR <h3'..- j( . J " L ) = <HS l N( EL J
* r 2-\?
?HS( JO0E2J = «?HS( NOO E2) TEMP•S(" J0E3 ) = sf-j ( MO :3) « T E.MP
-.
' [ f : J E
•: TUH ,
E
13 5
i L. ?. N ;:.i T I M E kHS'J
. . f
,
;J - IK ik I Sd \ iiUE TE~ '*$ CJ It .
»>.»"> J f
1 _ L,\ ,'-.. Lf I
1 . ,,•»I )lL
'IC") f= = 1 » iujjuc
LI i )£) = ). 1
" "- !
LJ *
J I : HLJdR (Z, IV, I TITLE)
'
J
:'. ^<' z :-' me <£laxatic--- fqr scve*al>T 3J : S3 r'] T i- i-.lL^TLGM IS T!-E FjRClNiO
: i v [
;
f o • ":
= l j ip e j u>3 { \~- i \ . r - : .
- , ] ) - J'JJ J A < V
: r r v. ; ; •: r : - 5 AJT IuN cIC L )
C A CL
ir m -•: Ru*ri3 r?
~ .'
) / C" i / t
»
:-lr IN»H I'J* t
a hl IAK f M'^l » i r
I I"LE
i?l t irL• ./:•;/ i!J.KL5o), I START ( l->? ) ,
! |^E( 1044).':*/ IS ( 156 > , A USSJ { I 36 ) , AMASS I { 1 56 ) , ? AS T ( i 56 , 3 )
' ' ' /C-15/I JO( L56), lOPTtNTRI ( L5a ) , <l( 12 ! , < i( 13)I'j. j/ CM 1 3/rf (
I
04-4 ) , G ( 1 0<+4 ) , F ( 10-»4 )
•:/;•'• I iT/[ ' IT (2 1) , ,a/ Ei\G ,i< L , <:, LEAPFS , t^UMNiCKHME'iSIC.N ZQ56) , C JE( L5o )
<c\L* 3 TITLc(^):
A- r ; ;'JE/I3o*i.o/
: ATA TITLE/ 1 J* • 3 1 V « , « j J
,r = 3
AL >HA = 0.65667
»
J cZETA ',
1/FA
CALCJLATE rHE '
: IV5 TS-JESS
:all iaslhs (crict gi)£ 21 12J2 = 1, IONJOE: ; je i = i.j
( ( I30( !OOE).EO .5) .Ai\'0.( IY.EJ.1 )
)
- lOflEJ = ^J 3l , )E) / ( AH \SSJ( j :
:
r- J
... IT I »UE
- - "•M ,|T M'.n.;; ;n ti 3
,-
:
T:( 3 ,3.lj TI T LE ( ITI H _)
z .r ( /// , 1 )x, - ,• ~ r <st ;uess • t/)
;o tj 2^« A.MASSK MCJE ) )
[ T E ( 6 ,40 ) ( : ( .
UT( LX,12E1L.4)! , sf:::" = l , U IGCE )
\tLAA ^; y 3 IT=- A t: is.
'' W ITER = 1, I TE • AT
C '•_.- \SLHS I 1 1 >)IQDE = L, NCN ' )E
[F • I : SD( . )JEJ .EO .5) . A.* J. ( I Y.E ). 1 ) ) Gil TO 60<:;:^ = uhsimooe) -a.-iassjmgoej - vmasskngoe) j /
Gl I START ( "jL )E) )
ITOE) = l ( \'0 iE ) \LPHA*KESI )
"i 7 1
»C I njE) « 0.0' ecru i
"( <>.-;.) ) M TC ; )
,30) riTLE ( ITITLE) t'.'•'-'
.
-
!"( /// , • . v ,.'.•:: :••
[x :io,4ojI : je
;et ,
:
:.c :: 3c1 t 1
<- = i.
»
LJC)
.3^
iL*»t\JuT PIE 'AiiLHS OtHY , A )
i j - j i jr
F jr\ ! A S S I \ T < I A FC* THE l_ E F T \ A >j C S TJr
i3 C" r1 IAS -J
•".:< I J T «C »Ah T j . A.** \3 3J...:.'. " ! 7 [ T : T-1f= S»K .CL'- )I,iJ *inoEs ( : <. • . . . : . .
"•
' r " ; . . r % . ! j
T: ;. , f : :m > r h
. » . , / 1 -2 i 2 J . i - J Y ; * [ j .
I .J «ki,J IE IT ") !Y - )EPF.,\JENT 7-\,iI ABLE
: rrj'ic;*2 j-ji, tiY v.t, !/ :
; • "/C'l/. i-TJEtN :L"r,-. I,ri IS , PH [ JAR , Vl , M , N J 11 j/C U/N'J 1{ 156), I 3TAHK 156) , NA.*!=( LO-'+A)
::•. : w/ G(i56) ,a iao3J( i5o) , a.vassm L56) ,34311
) ; : ; 1 c j••< v ( 1 5 6 ) , a ( 1 aa j
: : 1^3.3 1 a t ;. 1 : c sj : . ije » 1, ncwjek > A 3 3 J I ,£ E ) = ) .
4 .'m 3 3 1 ( :MG :) t J = 1 .
>i r.iE 1 oj : : it*i ;u ric «3\c = L, NCnlQuE
3 - 'j = 0.0[STRT = I 3TART(N0DE) ^ I
L ' ^T
= I3TkT f NiJYl JQOE) - 1~
) I MOE ( = i >T" T, LAST
JV3JE = I.V«1E( I.IDEX)i = ; i -i i » i i -i v / i
• if, •
"\ « \ i
, [FLG
L 5 6 , 3 )
1 ]
SJ.-iJ = iJ'J * ) J-' v ( INDUC) *A ( I UcX )
rHE : + ) c itki iutijus4VA531 ' E) = iMASSI (NOJE) *J-.J^Y(\C:1E) *A( ISTA3\.'ASSJ( NOE ) = ^.^:j('::j.j^J'jc c\ r i i j e
IT(NC3 = ) )
^ET'JP.'i
L 3 .»
\
: . r :- ;e \ >
i") - . « _ l . A T i£ 3 T i
- * Z)/2.L i *4E i I C c IE - 3Y OT- [ 3UTICN r .. <£ *E
,'. . ,' L
,P kST(,CHI i i
. _ - . . \ . .
• .i : . i t . i j | . ' - . i -. . t i
' L
» J I I ST;;<T ( 15h) »:\.l "-J 1 )i-4I
.' ' i/CUl/ !l:>o ) , ( I be) ,a.<E( LSo) , >$[( 15oJ' \/C !
I-t-/.>j( 3, 3) ,C( i Kw); I I ) < ( L 5 o , 3 )
: all xiso
: . > r\ j* ;
l j : = l ,
L -3 C l I ) = " .
.". VLl "- - l 11 fJt jUMY, L ,NJ
: . . soe -= Li ncn : >e[F133T = [ START {NCOE)LAST = I
"I ST * IlJ-'I Mlc) - L
; i ) i . jo ex = i-
: ^ ir
, L'or=
1 = IrlS(NCOE) « J(.NA.4E< INCEX) )*C( INHEX)' i : : r
;
; . v jr
'i /* -
D-: 30 I = L,v,.
•
; ( i j = ) .
:
;all rERMi (2tV, July, i , iJ
: 7 ) luDE = LivK .j )E
[FIRST = [ STAR r( NO )E)
LAST = IF IR3T * MU l( ICOE) - i
)C 53 I NDEX = [FI RST, LAST<ri$(N jc) - RriS(NOLOZ) * /{ Na.ME( I \ub<) ) *C ( INCEX)
?c :c:it I
,•; : c n t r
•j j E
3C loo vk: )e = it n no 3E
] > r. S I M 'J '3 E I = ) . 5 * R H S ( M 13 E J
TU °. *J
5o j
L37
>o - jut ; i:- / )rs is ci.i^Mi'To
.
HS - CALCJLATEi f jC /HRTICITY .1 : I ' = - V'f :CF. >EPCN0
. . .
JT •< N J .J
i..-L' : ) p s 1
1
.-.:,. TIL lALCUHTIt[No
•
:
. Lj - h L A G •'
' - I- 1 Si f I 1 F ST.f !_i \ : =li:
fr-i/C.'UJ/IEr.'. ( 135, 3) ,31 V( 15 6,3) ,3Ht ( 15i»2) ,S«VE I 156,2)' /CM 15//JX IX (3,3,2:v3) t V J Y i X ( 3 » •> t
?i \/C %1I5A/VJXI Y( 3 , 3 1 233 J , i/JYI Y( 3, 3, 2o 3)' /C'Uo/ ) J( 136), )V( 156 J ,ALrA< 15:0 ,3ETA( 156) ,U i
L -lENSlGM )l(15->) , )JHY( L56t3)11 ( 156)
I L L 1 S
;'.<.> fcp i o TI» JTE /ALJc iF FG GIN*IP ( IFLG. - . )) )Q H LO-
i
Ji i 3 a : 1 1
) *< vjx, ;: <>s ; ; ,
= -l.olALL TERM2 (DLtJlJMYf LfhLfVJXlXtSIGNil)
-(ps I ja chii *</j y, /i y>LL "- - 12 (OLf 3UMY*ltNLfVJYIYf SIS<-<4t2)
j c ro .3
!PJT = V^LUE ]F FG*C I.,o TE ;.'.S FCP /ORTICITY( : '['.'J? .3) ;q r
, 50. I0DE = I, N0N00E
)EJ = 5WEI \00Etll;- ( !gj(nc : ).•=;. 5) ihs(nqt>ei = o.oC : : T I ,J:
[flg.."iE. i) ;a rj joL L SG L V E 3 (Z T \,L,t)
r : :jeJl T
Tc?
JTE /\eJE OF FC CliJG IcVi FJ-^ DIVERGENCE
1 :
51 1U3EI = SA/P( NO )Et 21( i : ( )C).E J.5 ) 3 IS1M00E ) = ).0
3 C * T I , j t1 .-
i iFLG-Nt.i) ;o rn 7 ^
:
A
l w S I V E r t3lVRGt)f5)
P £ r
: 1
SL3 i'Jri.'JE JUVRrtS (OJ 1Y)
ini - ;alc jla r : s r-ij ;sj > j
.
= J * ( Z = TA + F ) X -J J iJ = \J*
{ i = f F
\ J;T
< j - : r l-J-'Y - J- >; . ,£ ,T
<HS /AL
i
•
i > i
*HE - E
L 5 6 1 2 )
CALL - i SO
;T
- iCT [j ":
L I = L , . '
,
C ( I J = 3.01ALL TERM1 (l v JJMYfZETAf2fiNPU
20 1 00 E = 1 , I C N JUL:
-i "JT = 1ST WKNQDE )
LAST = IF Ik ST * Vj:i IJ0E1 - i
IC J ) INDEX = [FIRST, La3T>E) = HS(NC)E) * JIMYMA
_ ", iTl.UE
; s rw jct ; j jR v )*fo ter-i1C 30 :oe « It NCNODE" M • J E ) = ) . 1
3C *J ICDE = 1, V.:j).)F.i
-: st = : starknc )Ej
last = [first • numi modtj - l•
. o in') ex = [first, last'"' ~ )C1 = TF_ APi viQ JEI *D JMY(N4
t : C IT1 jE) C '"•
) I C : = I , I : . J £a ) RHS( . £) = RHS(NCOE) * F )*TL IP
< --
ic( I IDFX) ) *C( I IDcX)
'»:( I i ;r<) )
3 G ( IK :tX )
( \CJi)
i^
- :alcjlu;:> th= ihs r -- .,L C \ . -'"T,» - : - IT -•: •
T i_ j : j / ctm '" ei"
.'
._
.:
1. : ) i'th me iiii^-ir fiel:>.
j :' - ) - ' . ; - i ; .;!_: (J )
:.":;.: s 2 zl -ie it,.\u 1, ista.sT ,i\.\ Ac t 13)/ C U / iJ IJ E f
'! - L -1 T
, i: j t i I
'
J -. ,J
i I UK , >m , N , M3 1 , I r L
G
: •' /C-U/VJ '( 136), ISTA.Ut 13a) , -J A *Z( I •)«->)
f> *HS ( 15 6 I i A'4Ai S J ( 1!> S ) f A.* * J 3 I ( 1 ?< ) ,
P
AST I L 56 , 3 )
:/ - •L :/IET„( L56, 3) t JlVt 15o,3) »^HI( L56 f ?) i SAVCt L56i2)
LVCU^ ; 3 » 3 ) t C ( 1 1«*
)
' * -is : : : .• • 1
1
5 ••) , f ill: u L3o j
"L . I SI
:
:
.
r_ (u :s / ) * oh i
IC 10 I = 1,C ( I ) = 0.0IALL TER.-U (i,-!LLE-, JHl,2, i'UI
jc 20 icde = l, ict'ieiftst = i start (node)..jF = IFIRST ili'-M *J JPP ) - L
: 20 INDEX = I FI 5 ST , LAST•S; DOE J = r^HS(NQJE) JUMYl NA4 c( I'JOEX ) )*C ( I NDE X )
CC iTI <JE
. it)
L t :•
). . -> >MASf:
If]
• _ . -
I )J I
-.-,;:: : :t
1 HI 4 J C J i - .- r . r j i
-
" ,-'->m- r • .
J z? J; Ftn-i ] - F{ i) » jr/2 *
- •.'_." JT - ucLTA t ti
.
[•ME? / a L
i A L r
'cu/\c ".:,::_ ,r,.,-,,-; jjr ,phi^a,<,\ l ,• , . u , i ,-l :
: . . / - ( :: o » , / ( L 5u ) , XK t ( i .
; -» , > s T
; L 5 c ) t C H I ( 1 5 o )
'
.•' V'L if I'.' ••
( L56, : ! , , 1 / ( [5 .,, 5 ) , >I 1 ( I b , 3 ) , 3 A V E ( L 56, 2
)
;" *:/c '. L3/h( i )V4) ,:;{ 1 0^.^+) ,r ( 10-+A)
: -• i/ ip . : if/ 1
'' ir (2 )) ,.xave:.j, u r :
,
l t aph;< ,ru"j j -3;<
::• ;
si '. »hi .• ( i v.)
PilTI \L EZE THE TI ME LEVcLS
= ]
. .- L = >
: :< = 1
; ; IALF = JT/2.3
iCNST^JCT IASS 1ATRIX >!EEDE0 FOR SCLUTION CFXCNITES:
% L_ * 1TKX2 [OTHALF)
3 :^ : IT "! HALF ST- 3 3
) L 20 1ST-- = L , 2
jp ;m: fc 3 r ti ieHUJK = -1
'
; n )TH A LF/ 3600.0
L /TRTEJ (JTHALF)C * L l = L *. X ( P S I , H , 2 )
; .l :s
.
T: ; ( jt ha l f j
ELAXi [PH I lEw,F,l)J . LO ICDE = L, MCN006Ph [ i lJJE,ViPl) = PHI ;
c ,,(-iaO£ )
C * L L ) I VE •3 ( THA L F , L J
Z A L L : LA X ( CH I , H , 3 )
: - ll )E j/ \
TEST F JR JiJTPUT CGNO ITIu.iS!
- Jl = I HOUR * 3( iPRNT ( L2)*2 ) .."JE . IHUja J 3D TG L i
::
( iPK.'IT (L D.EQ, 1) CALL uUTP JT (7 )
f i :> = Mi l ,ij
. - . . i) :all jutp jt ; ;
)
L°^r.T (i^).fo. D :all i > rp jt m »
iip ."( l "o . e
r
. i ) :all 'l.Tj i
= i
! ,
T•' THE T I IE LE ;
c L Pol i r F^ 3,'1=3" = \
. = Z
L'+l
r I i ..
!
T 9. x 2 (0 T )
«• ' - -A - 'r.i- j;i> T-< : CCEFFIC IEIT *ATRI '< F ?1 "' TFF L .
w. S.
•"-: :rj\ric v
js jsfj ... r ie sll'Jtioh jf the three"":
-S . t., ,: : ; = ; a <c ;- iTsijCTtD; .t
: r
TiCITY •
« L * O O
A* "> F , iSE :
M3L :o i - rn: ME i UT.v i .< , ., , L \ * -1TR* 1 . Fr: il FFE-i ENC
.• VJCUCTj JjE'). H 15 CJ ISTPIJC r:.: n ; 7 1
<VJ',7lX> > ;vJ ,r,V[Y> INNC-i PSQOJCT S . F IS. _.; o' j ; the i niu .j •ijK!;:Si j-l'- u Ij
... - .-. '
. c .st •,
r• iicn c j it \ ;• ; r.HE current _,fl t a
F L T : r
: .h j
F iFLTh t c \ .:•;• es . .j it ;.
LIST 1Y - ECU IP JT-jt , :::. half
r "- 'AT ; J , .
r "
i ;-V . ,kjc * in t * [ :e the. : in the lea v • , te ;hn i ;
. :sr )T - :jk l . r :el t a t
:- : ..- *_ el.hent—';l ;
. .
-
i• ' _ - • : l \
• l/CU/NUNOJE tNf!EL.MT,,i.N,H Jul ,PHt iAh, i.M 1 f M ,NP L , 1FLG12/ EL -1ENT t 23.3,3 ) ,PHI -I F , ZT A3 IF, I VD IF
CC 4• /C ,JU3/ 1(1 >^A) , G( 10'-+**) ,F( LO^A)N/C 11 t/ j.»{ ; , 3 ) , C( I )-«)
IC 'CM L5/v/J\IX( 3,3, 2d3) , VJY IX (3 , >,23 3)C L ' ' ./ C U 5 A / VJ X I Y ( 3 , 3 , 23 3 ) , V J Y I ({ * , 3 , 2 8 3 )
:* 1 . • i lT/IPk.\|T(2 0) , .M7enu,Ki,K?.,LEAPHR, *UNN BR
10 INDEX = L , NN: : MI ' ) = 3.0
5j [lj IASS UTRIX H, .ilTH i;.N E; PRC3JCT <VJX,VIX>
I 60 EL - i, I0ELMF)C j'l CO EI = L , 3
j ' : 3 e j = L , 320 3o(.l:JUEI , JUCEJ) = /JX IX { JUDE I , NJJE J, ELMNFjh )
: c unjeLL ASEMBL ( tl '\ ::^, H)t>0 1 CO E I = I , 3
JC AO ICUE J = I » 3-: i :
,;•;', ,:.ii) = / jy iyuuoei, iJCEJ,el>inhr)>: :c.jTihjE
;ALL kSE 13L ( EL,vi3^, - )
2" * .^ . r i. i j e
3^IL J 1A3S ^AT: IX F :R C NTE 3
CC oF = >. )/(PHl i AR M 3T**2 ) )
: Ii [M'J EX = L,
'0 "(I.)E<) = MI.4JEX) * CiJNSr*G( INDEX)
:. < IPHNT [2).E").i ) :all DUTPUTI 3 I
LA?
;M.CJLAT = 5 Th- RH'i CJNTk13uTI '. (F'.'G-
.* ": : i
rv >h i'ii >Tic c jj;ncN. :ach rz. . .
-: . :
:•
r s c i w u rr d a sr
. c : I . ^k.«.ay
i
~ 3 a
•1)
:i - y _ _ I « 1
: i/C-11/ 'iO^Dt , ' : L'M , , I'J J •
/C'12/£L :
r i- u . :
.
:, it •.:;-,: 'o
:.'../: •' t/ <S( 156} , A •lASSJt 156 ) , A MS ; [ ( I 5o !
:: ' ./c n/yjxi , _ i ,vjy( 3^ s.n ,vu <( 3,3,3"
." ' /C''12/ZE TM 15 3, 3 ),..)! v{ 15o, 3) , »ri t ( 15o,' i/C-U J/7JX [X ( 3,3 ,2 j 1* ,VJYI.<(3, J, 2.1:i)
» - , i
I F
, P A S T (
, 2 j•:•
)
3) , SAV
L5o,3)
E ( 1 5 6 , 2 J
. ' ' / C 1 1 5 A / V J a I V < , j t c j
;, /;"i J /. K^ci, , , .3 j ) ,,_- . t
I ) , V J Y I V ( 3 , .3 , 2 : )
. iSlCH 3 LI* ' Y ( 1 5 6 )
.',( L 5 *S } , U ,
' l ( L 5 - )
CALL i^)
-UT*<*.J U.VJX, M >
3 ; 3 i = - T
3 ALL T ;R"^ (QU,VJX,5I 3^ , i)
- 2- T «< W J . / J Y , V I >: LL rcRM4 UV,VJ Y,3IGN,2)
*<Z?TA( HI) J.VJ,VI>3:3. = 1.0lALL I.M3 ( 3J-1Y, IE'\ f Z , J 11 , ~> 1 j •
-11 A 1 1 F *0T < CZETA . VJ K , VI X>+<ZE T A . V J Y , V I Y>
)
SIGN = -ZTADIF*DTCALL Tz^V (OJMY, ZETA ,2,NMl ,VJXIX,SIS.VW1 )
CALL r=R12 O'JIY, ZETA,2,4rU,VJYlY,S['3M,2 )
i") \ ~J-z = ngnh )i>ave . . ;.: ,1 ) = -.HS( 1 JOE)
:t
t
: .! -I < ( Z , A , i
-> - . ^ :
-"
1: _ \a \ )F ALL PP. OS T [C
;
j>~
r 1
. :
•:
) T _ i
i -T^rs rj r ^ : : fL
i•
iU:i,I START
-..-' 'C ./ : '1 ,/T - r \ ,7 IET; J, JuLIaX, )E L T *Y , NL AT ,\'L • io
>1, ifdt
' ' '.
' : ' -( L5o) tA^A'iSJ( L5.-i»} i U I { LS6) i
' SI J*l - I I t EP3L M( j) ,ALF ( 3) ,A( 10h'+) , FI FL )( J J
< L"
n> » )
)
> . T -, IX I T < / 1 0/ , F I _ .
i 3 h I'
,' J 3 E
:v
. r . :'.)L i/i. " _- )i f 5.oe + )2 f x. 5<£*»oi/ f alf/ l,
i' r R . E J . 2 ) I R [ T E ( o 1 5 ) \<2j <
:K UT( /, L0X,« • ,F^.L,« 4RSDC 10 ICDE = I, 1CNC0E
( [PTR.E v.l) »HS(N'J0FJ = -- IS(.\'CC=II v )E) ^ MSTriODE, IPTk)EFSILM = EP SL;N( IPTR)\L^ H - = A LF { I P T R J
[FIRST = \L~\3 + 1
LAST = \L :v
l jx IL \T
' - Utci = 1
]/•".- -LAX TO r
, LVE SHLJTI J.'J
[ r £- = I T E < f 1
. . E = IFIR3T[ r ( L jI^.Ew--L) 1 OF. = LASTr r :tal = a
• i C n I
•j
JTE J • C l! ?'
- 1 6J T i ) . >
) iNfjJE = [pn:r, >., ^>j = ).0
[ S T .\ T = [ S T A R T ( .;l c! L
[L^ST = IST.xT 'JUMI 'I IDl ) - 2.N
J J = [STf.T f [LASTo-:) L = lA'E(JJ)
J = X,VJ » Z( JNJDE)*MJJ )
XVI = £{ivjC )E)*A (I START! . ).;t ) i
CC V )~ i < £ > IDUAL< c s : .. = ( S ( , ? J E ) x •
j X4I )/M [START (NCCE) )
[F(ABS( RESIO).LT.EPS[ CN ) ITj t\l = [TJTAL * i
Zi .. :cl = ZtNOOE) - \LPHA*RES[
J
. )E = MODE * ID I
R
C-" r : j J E= -:
:
iD A ...01 T: _ l
• v( z , I & rr j
is
• - H I *P LE T I C
N
l i" ;tal. ::.(.'.. z -^~ <l j . ,) i :-c rj 50:
-(
i
T ;-'..-if . 1AXU rj s j t ; 2;
50 5C >0 jCDE = 1, MCNO :; \jJ i ., ;.^E, I »Tf ) = Z(N !DE).- i r
"( -. ,7 )) [TE , FISLOC [PTRJ t »LPhA»HPSILN
7: ••'lilMlj,' I T r ^ 4T 1
1 j :S ' f A i t
. ( t ' AL 5I A = » , F = • t F L 5 . 6 )
• ' J
^ V ( 1 1 1 3 T cj
-: "5.; j i ) .,,> c i [
. < c L - <:I EL )S t
:
l-> T;i [ c \
.
r , the 1 1
»
j L
j or. - > :
HI> i >
CH i
i
.
J n
P,) I
1
'J / C 1 6 / T : i 6 T 4n , <L »
-«i i V £ L r Y T 1
/'Z'W/ALL>:Cj{ L5o) ,YL \T( L5o./ : . . K 15-b ) ,;< i5o) f x\t(
: i
s ;./:.. : - e+ i / , > s i s / i . i .)
:, 7 = \L C Jo *i iL A Tr ( l ) , i o , 5 ) » , : ^ -
j e l~
v i j r ,aL \
-
SKI
1 )/
T .» i
A , E , U ..)
i SO! (I
L , Ii-L
--> i \ t F
b > )
;) II (
: INO UY Ci ,)IT! 31 jK PH C'!j v- ^
>
: i
; i
:(
Z (
j-
.
-: . j :
re 70
• 1 :• r
i
t i j is r: <
- ) IGOE = Li ML D IGI JO 6) = PS ISLAST + MOJd) = =>SI J
r jC'JNDArtY C MJITt DMS FORi ) I CDF: = It NLG.'JG
E J - ) .= N )c » LAS^
IJDE2 ) = 0.0
EL )
PS I FIEL'J
U FILL,
n- ->
- :ALC JlATES th- >hi CJ.4T:< IllUTI j
. .7 : ijIt
t PP.-jGN3ST !C : < J \T ;.;\.j (F'jkcimg r
r ;'')
:"-,>--:•
_ » - . i
» • / C ' U / . . E » I.i E L •'' T , .V i f r ' j : J * , i
) -.
. / : ' j / : . ;iT{2ii f 3) f pr»tJiF l zT 1n.
"' ' /c r/ <l-."::-i-;( 156) ,YL.\rt 15-j) tA-vp,
i / C ."! 9 / 7 J X ( , 2 3 ) , V J Y ( 3 , if 3 J ) , 7 I J/C HL/ui i
c.6 ) ,Vt 15 o) , XXL( irio) ,->
• • /: 'i j '
: "
' .( i;
., n , j:v( ii j, i) ,»
'CML5/7JXIX ( 3, 3 ,2 j;1) , 7JY1X ( 3, ^
i- " • / C . 4 / J [ Y ( 3 , 3 f 2 i ) ) t v J < I / { 3 i
: 'i
' /•: i / iivjoj, (i5o) ,
- . . » J * Y { L -3 6 ) » JO r 2 ( 1 >o f 3 } f X ij
j
1
* < 1
3
»'
1 t
"5
1 » [ r L 5
1 , j I / :'
J c - p t sc '( i
: >J )
,^( 3 » 3 »
i-
] 3)
i 1 ( L 7 ") »
L
H : ( L ) )
1 [ ( L 3 s t 3) » > i J .-;( I 3 J T
)
»•<
i )
;
1]
/ ?•
i.
/
I': .-.V
/ t (1
L K L 5 cz
r
L \ i I t;1
L ( i,n
j l
I ML JVRHS (PSI tXCHI ,2 f 0l: iu i :l; c ^ cjnp i , i, i
)
:all *hso
CCOPUTE THc kiUNDARY CO-lTRIdU!" UN. = 1 . J
;aLl PH IBOY (UNP1 ,5IGN)-> rllQDY (U7M1 iSIGM
»-CG.MST*<PHI (N-l) J.VJ,VI>S 1 G.N = 4.0/ (PHI^AR*( JT*-2 ) )
IaLL ' - a -'3 (0J1Y,PHI, 2, lit SIGN)
-M.3/0T*<ni7< l-l) J.YJ r VI>S i
3" = —4 . ) /QT7 LL 73R 13 ' )UMY, DI7, 2, . !1,S1 i J )
-t. )*{<PHI (N-l) J. VJX, VIX>+<PHI ( .-1 ) J. 7 J'r
; : : = -i ,oCALL T----12 ZU-iY, PHI r 2 f .\Ml,VJXlX,SI 30, 1 J
CALL TER.M2 (OJ">Y, PHI , 2,.^.U, 7JYIY, SI3*Jf 1)
-2.o*<3vj. yjx, ;i>;ign = -2. )
: cL. TERM4 ( ;v ,
V
'J <, SIGN, 1 )
y> )
+ .:..)«< jjj. /jy, / 1>> : 3 , = 2.3; vLL "
: ?M4 (O-JfVJYfSI i-Nf 1 )
- . . )-M<XK£J.VJX,VIX> + <XixEJ. /JY, /[Y>)j :
">• i = -2.o
Call 7 : 3 < : ( s < : , i j i ^ z , l , * ,
/
j \ : <
»
s r j ;
. ."": < 12 (XKct OU'lY^f It It 7JY IYtSIG I , L J
-
/ J :;
{ <j! . . 7 J < , /Ia>*<JI /. /JY, /[ '>= 2.0*017 [ F
:all '- i ;: (oo »y ,::/,:, \ -i , yj< : < F 3 : ;
._ rER.-12 (0U1Y,0I /,2, HI, /JYIy,SI3
- -. )/{PHirj^-J*OT)*<ALFAJ. /JX ,VI>SI GO = -«*. J/ (
J il 3Ar »0T J
14 ( ALF A , V J < , S I GO , 1 )
--. . )/ ,
J ^ I ^., k0T ! * <3t r A . /JY , 7I>;i}0 = -4. I/(P;II3A.V*0T):ALL ' ;
.'-*
( j3 r; , / J ', 3 I , ! , L )
en :
yt -
L -
r 1 . »Mi iJY ( j.j, SIG;;)
.'
' l/C.-U/
r = \l j:
iU = ,L
s i : i=
A . iQOE:
: = .
. _J
. = L \ S; < i = l .
A >ES)
" :- .
: L.T, .'.,,« ] ] , >H UK r N M l ,N,' U ,
/T IE T A?T HETAJ,OELT k < ,OEL T -'\Y , "IL AT f NLC'lG (
''_-"' - ' • * s\jjJ { l')6 ) i k.v A S IS I I 1 !>••> j , r> A ST ( i.
*0 » \L » . , . - . ' > , Nr jXL \.; J (»•:»! )-,
Tl M ,..-..:. ,
J I L
"
: IL *T. . - 1
IrLG;
r
)X1 = K LO^G) X 2 = «. L
t 1 w-3( Mu )E.\)L
( 1 = a -
= 1 , L C • S' I
f f )E
»cS ) - <L IG( . Jts-ni ...'•. S + 1 ) — XLO tG i i-jD ES J
- • ;n*f )*( jxi • J J( IC CES-1 i
+ !. 0*0X1 *UJ{MGOES ) + 2.0*DX2*Ul('f )X2*U H 'it OES + 1 ) ) /o.O(VCJEN J - <L.1NG< \ r
_ -\-i )
EN + L) - XLG.N SI MODEN )
* -SIGN*FO*luXi*UJ( IGCEN-1 )
* 2.0*DXl*U :J(.\»UJtN) * 2.0*0X2*UL{N0X2 *U K \GOE ^i ) >/6.C
G J E S )
(1 j EN i
IY.JITY F j! o . i JDARY PQI NTS-> r > -= <LGNhS( 1 )
)X2 = ^Lt3(LAST-H)
;yclic cj.ni- = < LO i
) x 2 = M. -
<HS( NL 3NG )
3<i = CL0X2 = <L -- t-S( lAS I>NL
<L
'i'lG ( N L IN G J
ID - <L IGQ)= S IGN*P ^ >(0X1 *UU< 'JLCNG)
2.0*0X1*UJ( 1) 2. )*DX2*UU( 1 )
)X2*U K 2) )/o. )
<LGMO< N ]M00F)I LAS T+ 2 ) - X LJNG ( L A S T + 1
)
= - SIGN«FOMuXl* JL( IGOEN-1)2.0 *0Xl*iJ(J{.N:J0EJ) 2.0*DX2*ULi(N0JEn)
• I ;; J J( \G0E i+L ) ) /o.G
II TY F ,.< WES r aoU.N :ary p. i -its(NL j\S ) - KlUNG( lLGNG-1
)
XL 3IG( JL VI S)
5IGN*F0*( >X1 *UU( NLGNG-L)2.3*DXL-*UU(NLUfiG) * 2. *L)X2*UU (NL.1NG )
« )X2*UJ( L ) )/b.OI10EJ - <L j iG( .
..'.' : :e-l )
XL ING( NQMC )EJGNGJ = -SI GN*F0*(0X l*UL{ IJNGOE-1
)
2.0*0Xl*UU(Na,NG0EJ + 2 .0*OX2*UU (M0NG3E
J
>X2*'JU( LA3T+L) )/6.0-'
:t
Ja.'^
i\.2A- < E L \ < 1 ( *1 , A , I ? T ft )
jA TI
_ r
L 5 CALLE
)
*EL T I UN r r | -
E A L * 3 F I E _4 «
.O T
» .<•
. T -\ 1 i T*-E iwLVEC rI 'L
. j j ! » t • ;<
^.i ;i- r«'J TnE FIEL3
i.j/CMl/ 4;j:i JE , . !
:.l 'T,.„, ,irjJS ,PHI 5Ai , .: 1 , , ,'.M , I .
:rL 3
•
~ / C -1 1 .
/TILTA f THETAJ » JELTAX t JELT *Y , UL AT ,\LO'iG » L T
.:'' 'j/CMj/Ni K 15o), I 3TART{ lob ) f NA,M=( 1 )<<*)
. .
' :../C •' W'
- IS( 15 6) , A IASSJ1 156 I , AMASSI { 15< » ,3 AST ( 15o, 3 J
ij I :-
. u( 15 i ) , EPSLMJ) tALl- l 3 J f M L )— ) » FIELD (3 )
xii,7:^/, : ;-L-/' >hi • , • ? si » , • chi •
EPSL ./ L.3E--H, 5.0F.+ )2
)>.' LO CDE = 1, '-ICMOOE( iptr.ekii "•-
S( :hej = -khs(lTIDCl) = ?AST(r CDE, IPTfv): j
s : l < = z? s l •
i ! : ? j r )
XLPHA = AL' (ijT
-l: LRST 1
L VST = NLC'IS* .^ \T
.L : .^IF IP ST = NLCttG 1
LAST = MC I3*NL \T:
r " = oi.i = i
j/»ALF/l. 4 ">,1.Z -7/ 21,1.520/
fte l \x t aI TEft = L
TE-
1- = [FIRSTIF (IOIR.E ;.-i):
' rAL = t
5 ~L YERI
SGLJT ICN
j ODE = LAST
. ' > ;T •_ J \ \ J I C C N TR 1 3UT
1
1 NS. -, r \- ip = rr
I ^3T , LmSTJ - 1.0
ISTRT s[ STAR T( ID J: ) I
L LAST = IST^T * NUM(NODE) - 212 30 J J = ISTRT , 1 1 AST
E = . * ' _ ( J J )
X 4 J = < *J f Z( J*JODE)*A( j J )
S = 1 ( IODE }*A(I START( IOOEJ )
'/vit; RESIDUAL:SID = ( RHS(NQDE) * X A J + M I ) /M I START (NCDE ) )
n: (MdS(RESIJ).LT.£DSI LK) I TOTAL = [TOTAL » L
' ijoei = :(\;) c) - -«lpi!4*aE3;j
I . • E = "1L D E [ 1 R; iTri je
'- ,7 F - : iPLETICf
r L.E0..I ;JCC:
•" :"_•..-.ia a : t - j : :
tI 2 )
: i J C = It ' . 3 . IE
;C 'AS"!" - 3E , 1 >T? ) = U NGDE),^TEL t
7 D ITER, FISLO( iPTR) , ALPHA ,5P SI LN7, F 'J H 4 A T ( 1 X t 1 3 ,
'
* ' ! a, • \L ^ , r- 7 . <+
ITERAT1G j: i -*, J
E 3 S I L L N = • f F 1 5 . 6 J
. - i
SwK 'r
: : :i (OT, I^ATZJ)
. - :AiCJL\r£S r^: ihs ; jtp. ijjticn (facingr.iL :.:-;: ;e prognostic =<jjaticn.
T = S .«
i• • ~
.:T - Z L T A 1 /ALU:
[« \ r z j - f l a j f c r
: -3 r 2 rr :
. .- - :
:.':.-' *2 ;L IT
u.j/C 12/ EL IF IT (243,3 ),PmIO{F , /TA MF, 3 IVO IF'C <A •• _5o) ,A 1ASSJI Lj . i,v*. i5 I i L 3 c ) ,
D
/ c 1 v * j a ( 3 , : «. ) , v j y ( 3 , 2 j ) , v 1 j x ( 3 3 , 3 ,
2
. '' i/ CM 1 L / J ( 1 56 ) , 7 ( 1 5 :> ) , x X si ( 1 5 6 ) , > S I ( I 5 6 ) , C
. _ Mi /C 112/Z5TA( 155, 3) ,3I7( 15 o, 3) ,3H I ( 136,3)
. / -:' l : / / j * : < n , 3 , 2 , j j , / j m ^ ( 3 , > , : t >
)
: ': L3A/VJXI Y( 3,3,2 id) ,VJY1 Y( 3,3,2 io)
' / 2 '
L r> / i 1(15 5), ) 7 ( 1 5 o ) i a L f a ( 1 a o ) , J E T A ( 15I
•-. 1 I i
'
3 II ( 156) ,0u !Y ( 15 o ) ,DLiNY2( 15o,3 )
AST { I;
?)
-1(13r Sm i E
IFuG
56,3)
I I 3 -, 2 )
n ( 1 5 c
)
A lUu ISO
T* ( <? HI ( . ) . /J < , V I X > *-< r'; i I I . ) j . ; J > , ; I * > )
)C 10 IU0E = 1, NONOQE\\ \
': : ( \u l)E ) = ( ( P H I ( NC D E , NP 1 )+PH I ( N J t , NM 1 ) )
/
If I MATZli.EQ.l) &YGPHI UGDE) = PH I ( NODE t N)ri.UE
JIG>1 =:-iLL T-" '?. (AVo?HI,J :J
,,Y2t ltN'U , Vo X IX , SI GN , 1 )
:ML rE^.M2 (AVGPHI,3J*YZ, l,.-i;'l,7JY IYfSIG'1,2)
t- i'-s./J. / J a , 7 I >5
'
j I = "^T
lALL t: *M4 v }7i VJXt SI V., 1 )
- .: r - < )u.vjy,vi>" * " - — ) tCALL "Z « '-
( y J, 7 J Y, SIGN, 2
)
" " ' CX K E J . / J < , V I X>+ < XKc J • 7 J Y , J I Y > )
SIGN = DT_L t:k'2 (X!\£,DUMY2 ,l,.N,7JXiX, SIG J, 1 )
:ALL r 2^'2 (XKE,0U 1Y2,l,NtVJYlY ,SIG'i,2)
< 1 1 v I .- 1 ) j . ; j , j : >
SIGN = 1.0CALL x 2- :
? (iXW ,017,2, N 11, SIGN)
- : : ro
i
f *ot ( oi / .
7
jx , v 1 x>+< o iv . /
J
y , v 1 y >
)
.= -DI7DIF*DT
CALL TES.-12 (•JlMY,0I7,2,.VU,VJXIX,SIG'J, 1)CaLL r ~ '2
( 0'JNY,DI7,2 ,Nrtl,VJYIY,SIGN,2)
"
: in UDE = Li lONJDE5A7E(NOGG,2) = rS( . jt'i
_
? • )
c
i" IM E L u A .» F 3 ( 1 f )
IA -C hPi>,»°.~< - * j I i
3 LE IAKCHPIG SuHG : .HE £ F (
;
N !+ i ) JSE5. » ,
T": ' ) : F{ .<- i ) = Fi i-i i -j.:r > fi . ;• . n*i s ;c
r F :?..' \k.J ZTc i \S : ":
! PLE T E:3"
.' vLJ'E is ». .-i • : . ;
r '• jl :
'
.
r•.- r ' \l . . :.- l j r Jvj . .
~.y. r p\-Pj
' S . I 7 1 Tj:»S F T'-tt- i <". .". ;ACH.
T L.
: . • ,.j IE n IT - >"l: i
T .'! V . |
/: 41/ . .
:, LMT,rl i.njji- ,
"!: iAK , ; : ,.\ ,-.
>
/ CVH/J( L«?6 ) ,-/( l? 3 ) , .<,\t( 15-3) ,PSI < L5 o) ,CHH
C, • i/CMl2/Z=TA( 15b, Jl i -3 IV (15 >, J ) t^n I ( 1 56 , 2 ) , SA; • */z:* ii/;j(io'4^i t .}( i j-+t) ,f < i -)4<»
)
""'•./I-
3 :\T/I?SNT(20) , A-A7S ' »'
L »a> , LEAP*- - ,-/.
:'-...:.' [ lE./( 156)
rrr .. Wr T-i" T' 1= I P \/ 1= I ;.
1•"
V E
I r l Gc I
( iSfct 2)
THE r I IE LEVELi! ;ITUL IZE: = i
>.- :u * - DT / 3600 .")
7 » ]DT = 2.0*DT
CCNSTRJCT 4
* S 5 IATRIX NEE0E0 FCR CCNTEJ E DCAll AURA2 (TWCDT)
LE 4PFRJG T ] LEAPHR -\1 )* 3
1C 30 ISTEP = L t LEAPHR
ICJ = HOUR + )T/ 36 30.0
C-LL 'JRTEd (1.^0T)CALL ~ .z L A X (
o s I , H , 2 )
: \ ll : ) \
r
e j ( t * jd t j
C *LL k ELAX1 I PH IN EW,F ,1
J
) : L3 IODE = L, NCNHOE> -. ;
. E,.mP11 = 5 HHE.<l:-iiGuEJ
: \Lu JIVEO (T < TjT t ^i
CALL : £ LAX (CHI ,H,3)
:all J z ?VA >
IV
J'.
'JUE THE T IIE' I a
i' L
= !PLj F 1 = . .
•' L
: T J M I
15 .
! L : rv MS
r i:.e""
: ': '
t ::a •
, )l :vi, h.*y2, .:
l ; .
• '. s \lf'\,
:_\ , \:n .\c. re < 1 1 i
" ."1 - )F fi : F I TST 'JLTIPHC.ATI
•» t
, l )
••>
: » l . l m i t '-;: -
.
:-'.a. r = ^ , i ,. . iti t
: = < j j . v j , j k . •/ ;< , v i > . \ 3 \ y 5 i \ t : < - •. is. iw/ 1
_">•'_ JAY \ -, JSEJ IN SJi3kCu tI '.'E - U^.
j-Ii£LI_Y F E T
"
: V\ [j ^S3E <bLEO I.'JTO TH£ IGF KING VFCTu:all '
: i _>y »sh irji.. :
r ~- n :.';: u . st \te -»e-4T /cmio/
ThE IE A"-j .'.-:.« [S I a.-Ill) p. i ; FAC1 T tl s is T :--- / I j.\
.'.-. )uct fhai is .'
;
"? ; i this sul»:\ jjt i ic. 3UT F <
CTI i.r
it \-v iErn >
E r L I N r IT ••« \ S -* E '
-\ 1 L >r
IY2. [ 1 F >
- r-
L
Avj J S t
C
- jepemde r
- i-^i )EN7- FLA j FJK
/Al t A-3LE- CUkRENT Tl 'E
/ r- I \jl.- ( L?6 )
M3I \3LE ( 156 ,3 )
iHICH DEPENDENTIS T ^ h E ij S E
J
LEVEL FQS D U 1 Y _
: ;TE 5E**2 EL -5ENT, i 3:3
C 1 v;* / ; 1 1/ JCf !DE ,N.~)ELMTfMN,HOiM,PH-I JAK , |M 1 ,N,NP1,1-2 ^Mu.Ni/CM2/ EL IS NiT ( 2o3 ,3 ) , Ph ID I F , ZT A 3 1 F , l> I VO [FI I *1 f]f i / CM 5 / I dO ( L 5 6 ) r >l OP T , N T K I ( 1 5 6 ) , X 1 ( 1 2 ) , Y 1 { 1 3 )
CT * 13*1 /CM 1 )/VIJ(2 T 5) i '(3 tit it 2 3 U,VIJY(3,3,3.2 ' 5)
: '10.-J/CMl*/3=.(3, 3) ,C( 134*). ENSI f J
,v l( I 56) , OL !Y2( L56, 3)
, 100 1 = 1 ,,\CELMTL -ELME NT ( I , 1
J
,: = ^L ,r \r (i,2)=cL ;NT( I ,3)
. '. [ 10,20) , 101 •'":
l
) i = 1J MY 1 ( i L )
)2 = 0U IY 1 ( "2 )
)2" 1 ^ 1( >43 )
3 1 " " 30Dl = IMY2 ( lit iL))2 = 0JMY21 l2t ID33 = ^J ^Y2( 13* ID
:lRi\ECt fcr 3::vJ,
i da?.y ^hen ir^<jc rn ( 50,40 ), iter *
[r I I n 1 ( I 1 ) . E ) . 5 ) 31 = .
[F [Ib:3(N2) .S'J.5) >:=!.-)IF (I 3J(N3 ) .E^. 5) 33= ).0
;
.
:
'P'JTE I
•
i L
I _i -
t2 )
,3)tl)t2l
( 2 1 3
)
( ?. , I )
i .,213,3)
A ^[' J
Li_
— i
==
: IBL
L*p(i *p i
L ? P (
L^PI
L*P([ up
{
( I ,
TICIt i
I
I
.
1
)
t
i
»
i
- »
I- t
ItI t
C)
» ^ »
» t ,
t _ r
I »
t 3 ,
:
I
)
I
)
I
)
I
)
I
)
» i t it r
EL) "JuJOc*?( 2
¥ul*?\ 2^ 2 • M 2+ J2*P( 2t2,2, I
+ 2'»:fJ
( 2 ,3,2, I )
*D2*P( 2, 1,3, I)*0Z *P(2
t 1 1 2 ,
I !
I !
I )
I )
3 ¥ °{
3 * P {
ASi
23
+ )2*P( 2,3,3,1) +">:*°( 3,3,3I)
+
h I3*P( 5 ,Ji- 3* o
( 3 ,
1
+03*P(3,20 J*P< 3,3+ ) 3 * P I 3 , 1
»-D3*P( 5,2
t _ ,
t
i
t
t
ETJF i
j
. ) i
•: - :
j- v)
\ :
~> r He i :A i <T
: .>;
\ T i ;.> l'/cK -iCh ELcV'"
7 [^ I1
•:- 3 * : ) iC T 3ET .
_ : THE U- . - L w. i iR t ) L7\r[/Ejf I • - . v , J < t . I
- i-
. : it , :•
j
/ ; '1/ JN03E , " E^ :
- <i
LAC F - i i 1 1 C i J
7 \i\l \?.Lr.S IS JSEDiL - TI.'lc l _ y LL - . I J 4Y2» - ;.n.ne^ 5 <• :olct
F I!-; r :
i
j, 3 )
- -LA jSi- ) v- y
: .
* , 3 J i ? H [ 3 I f , n i ) I P , J 'i V I F'
7.1 ' / . ' / . -i",l L'i^i , V *ASSJ( lr >) ,4 I ( ) ,J
v ; r (
CC V-ii I/CM5/Id0< 15o)i nPT,,ir.U ( LJo) , <1( 12 ) , Yl( mI
I 13 I C.N . .> IYK1;-)), :o -!Y2 ( L56 , 3 J , P
, I F L
L =5 o , 3 J
IAuC'JLATE THE I ITEP \C TI . ,- }k E \CH'
E L 4 .-i tJ 1 = L t . I E L -17
' :jcl = EL I-i
ri EL*M3 3 , L)
)52 = El t": IT{ "L -IN ' ;,2 )
IC )E3 = EL'^EMT(=LN!:VH^ ,3 J
;: ro (20,^0) , idimen)i = )j:*yi< <uje :
)
J 2 = . 4Y L( NO )c2)J W L( .
j -53)5C TO 5Cjl = )j 1y2( n00e1 , ml)>2 = ) jhy2 ( .:)-:-'» \ljJj = )J*Y2( N )Uc3,NL)
. 3 , 3 , 2 i 3 )
JUOE PER ELF IE,NT
:c;-i^jte i itek vcti cn i et ,eei = EL« or
•
;
L = D 1 * P ( 1 , i , I )
t - : = : l * J( : , i , :
)
T.'ri = 31*3(3, 1, I)
)
* L>2*?(2,2, I )
f ^J - J( 3 , 2 , [ )
< ) 3 * P ( 1 , 3 . I )
» )3*P (2,3,1)< J 3 * o
( 3,3,1)
ZC"> EC T F'JS i:i';J^v ,HEN [ BO*Y = 2
32 TO (70,60), [QOPYir
( I J0( NU JED.E3.5I TEMPI = T.OI r i I i ;( NOJE2) .2 :. 5 ) Tc 2 = 0.0".
I I 1 :( JO0E3) .£3.5) r E.V 3 = ).0
\:: <e;ult:> t. r.i2 ins ( .jl)r<< /ectc*)*H3( MJOEL ) = vH5( NC0E1 ) ^ T -
-- L ^ 3 I ^ -
• 1 1.
. i : > = < N S ( NC E2 ) « r 2 i ° 2 * S I G
N
i3< )OE3) = ' ("Cl^En <- TE1P3*SIGNr -*
it" t
. - . . . . > :
v P T J r '
: \o
.52
! . . ' J ~
Z > <
r r
i )
T~< '3 (OJ.IYli VJ 1/2, U P'E i, !L,S)
• E;
" y 3 - j i ): t -. E r the usi
-> r x ,'J , / I> .
; r t - j . . < . . .
) I -1Y2 - Jil^ENDE VM 'r ; - fl \ 3
L:
So )/ \ : \ • i.
.
.MICH DE l3 -~ iOt-\T
j i j.iL J V d L - ! J
i: j f r i E r
;
.~. >
-,
W *_-
-
-1
'
]
JI • Z
* > IJ 1, I >TAST,MA IE
l/CU/ iJ iJJE » I'JEL.MTt^NtHJUR, PHI JAi , Nf- I , I ,.' -' 1 , I F L. / : I :>/ . JM ( 1 5 6 ) , I ST A* T ( 1 5 6 J , i\ A * c
( 1 )4 4;/C 14/ '!iS( L 5 6 ) t k IASSJI L5f>) , \ JASSI ( 1 5c) ,PAS'l/C 1L3/H( L 044), i(l !044) ,F( 10 1 4 )
SI GN JU »YI [ 156) i )U -1Y2I L5o ,3 )
:., " : (20,60), iD HEN
3C 40 4 CD c£ = L, iC'J IDEI FUST = I START (NCDE)LAST = LF [3 jT <- M H{ IQDE ) - 1
0 INDEX = t F I «ST, L ;ST-
. . JOE) - RHS(NCOE) J j wi( , ,
:jntinje;*: ro loo
1 5 j , j )
=( INDEX) )*G( [NjE.X)
3 ,
L n
: , 30 JCOE = 1 , ICNODEI FUST = I STA^T(NOOc).AST = IF I • jT * TJ-w >JQDE ) - 11
: JO INDEX = IFIRST, LASTIS(NQDc) = *HS(NCDi£) )LMY2( NA -IE ( INDEX ), ML ).«G( INDEX )*S
: : r i iu e
- :
uj
.Or, jTIiiE OiPf Si-JN» [.30RY )
- IMPUTES THE. . i
•
-
E\ : ir E v- \ r i ;.\
. : r _i .Si ;
.
e
i . .
r *CH ELdME • r
c EM THE >AS [SiVE**ET:vjx,vi>.
m : , - l•
1 - t .-: . >n ' ILCT;!Si"-i - ii'V-l = T.IE TE^'!IviJpJY - FL-ao -.Sf: ) F IP 3CUrj 3A3Y
I r c .j E -
:
» 1T O T A .
T, \ ,1
• •' i/c-u/ ;c io )£• in=LMT t :-i: »na j-. ,phi 3a • . imi » : » ipi, ifl".
.
' '.' -l •- :t ( 2 ; t : ) ,t ; .: r- , <:t~ ?- :
. , ; :v) :h:'. ' -/ ? HS( 156) , A iASiJ< L5n) , AMASS I ( 15M ,PAST( l=-i, 1
:/ C 5/1 J> ( L 5 o ) . 1 Jw T , i T .< I i L 5 o J , / L ( L 2 ) 1 Y 1 ( L
3
)
LCJL\T [.MTEK-iCTI I
J 30 EL VM3H = 1 , . ;L"T\CJE1 = EL"^l:MT(PLMNJ^tl )
:2 = EL !E i
r( ELMN6 1,2)
:3 = EL ' EMI ( -L *N ; ,,3)
\Ch "; JOE PER ELE IT
. ^ D{ NUDE1). ; = 3 ( \ E 2
)
j 3 = ;( v:jc3)
:C'-.3 ,
TE TE '" CQNT* I >U TIGNS
I = EL IN8 5
rF.iPi = ( :l * 3( l , r ) D2*p{2, i
r !P2 = I J1*P(1,I ) 02*P12,\1 . I
; J = 101*P(1,I ) 4- D2*P(2,I
L J *P<3» I ) )*SHi->J«• C 3* P ( 3 r I ) ) *S I GMf D3*P( 3» I ) ) *SI5ii
6G rt Y =; : • ECT F : ; 30 JNO \li HHE J
^ T'
( 20 r l )), [3 DRYCC.NTI :jf[E(I33< NjC )E 1) .£ ,. -
) TEMPI = ). )
:- :i .,:.( \coE2 ) .: ). 5 > te -u'2 = i.o
* F< : j >< . )DE i) .... 5) TEMP3 = o.O
*C0 tESUL TS TJ rHE O' i i rtGP. ;
, 7E«. :I ) ^ "vrii ( NC'lEll r E.i,J 1
)E2) = •->{ \' jJE ?) + TEMP2>< luJ E3 ) = CHS{ N3DEJJ * TE M
: : it ;•; ie
r : ^
- - ;
jt
J" IlE TJT (IF
.-T 3 ,r -_>f 1 NT j_r .- coitents JF i.r-
- • ' ' ' £ 3 -viL T S • '
j H \ C . 1 i T I j GALTHr ARRAYS
IT I MO
. -j r. »
u' ./
:? -
•i/
;•\ 2 /
«
'
: . i
^ _
:
:
5:
i/•./
/
L, </
j ,/
M/
' « 1/
'E IS [
IF
lb
l(
5/1 30(
) c
T V
T
j r
t I
H £ 1
1
» -3
ST
' 7/i >/' V4 L )
111M<L1 L3114•1 15^15•I 1 6!L7?" I
•J I
v( >,
IJ<(/ y I j/ j( :
/ : r r
/H{1/33</ ;J <
A/VJ/ iU(/< 11
NIT/ I
I (12
6 )
-
if
( 2- ~i
• (
^43 t
IX<I1 o
N t
- »
) , I T 2
) ,
/('>
\
»C,33 t
T j
: -
(2s
»
3
yi' q:')
(
i
(
l
j »
/(,F))
T(
• i i-j.-i
, J , J> b) ,PH• -. r i
T, .1
,2 » j
-if ;
:
JK( j->) , XT J I 7
-144),4 )
J3 ) ,
2 3 3)L5o )
JoT i
7 )
t ' -
LIV
.
: )i13o
: ;
YT ]
; i i
J ,A
,3 t
Kcl( 15t
;
:
(
J < t
i 'V» «-
),.•;
P , Y
^ ) ,
. t A 2 (
3 ) , V I
15?) ,
o , 3 ) ,
1044)
1 1<
'"(
< 1rI »
t
l ^
JX) t
". » • ;l\ 1. , , <
,it.. r, ml :
, ) : v 3 I
F
}44 )
»V
PI, IFLSi G i !
r
) A rs » r I
)
( L 5 6 )
« 1 t i , ;
ViJY(3f3,5ftiv 3 )
I ( L--) ,CHI (156)»HI( L5o, 3) riAVEi 1
/ J Y I X ( 3 , 3 , 2 3 3 )
, V J Y I Y ( 3 , 3 , 2 3 3
)
,ALrA( 15o) ,3ETA( 156) ,Uri.MK 15KEi\G,Rl ,"R2,LF \9H9, ,RU . I3H
TO ( 1 , 3 J 1 40 , 5 , 60 , 100 , 70 , 8 , 9 C ) » I FP
'•;.r FOR EACH ELEMENT ALL 2F ITS N.'IOES < CCRRES >
' r E (6,11)lOES 1 ,/)11 FC3'i.\T( ///,10X, • ELEMENT *
J : 12 :. L IN >< = 1 , . IELMT
L2 iRlTt 16,13) LL^MdR, ( EL 1ENT ( ELMN3R , NOOE ) , NC0E = 1 , 3
)
13 FCP 1AT( 10X, I5,2X, 31ft)
THE : .:•. .- ; T I VI ' \ T ^ I A U « r: L >
J J''
E C T I V I TV
•RITE (6,21)21 =CR'-1AT(///, 1)X, 'THE : _"; '2 ; r I V ! r Y MATRIX L t S T c .; ',
1 «SY kIQ3E' )
* R I T E ( i,22)12 rZR UT( /, 10 X, »\iQDE I START
L • \ JOE S' )
3C 24 I JOE = 1, ICNJJESTART = I STAKT( SUOE)l =
: i i c. )t) • sta.-;t - :
.-
:
"" (6,2 3) ICDc ,STAU, Id -1 (
:JMAT(U3,Ii,I7,2<,3l4): it ; jj ei R I T f= (6,25) INIR. 1 AT (///, 10X, • TOTAL NU113ER JF CCNNEC TI VI TY MOOES ',
L • = • , I o
)
- : " : (6,26)_. 3 FCR 1AT( /,10X f 'CnMNEC TIVITY SECTOR sA.ME •
)
I'
: (6,27) (MAME(K) ,K=1,7
r- . . . : (25 1 5)M)
L->
' ~ ' _ » :j _> I \:itl r
. . - ,' .::•.;-;•.
L' )2?c 132'ii / CABLES' , /)[ T c (
•j » 3 3 J ( .} ( J ) , J = L , J , i J
3 - ' ' ' XT (•£ i i. .
)
: : < , 3 * )
- A T ( / / / , L ) < , ' M A S o ' A T a I v
l \\j ) rv t ; •f /
)
. I' ,3 3) ( i[ j) ,J=1 ,\ «)
: *r ( ///, i •::•<, • ia sj 'ati ix
'7 LOG JOJ = 1 , f . 1
-. 5 _ J
.
IS fE T j
\ •» t a x 2 >
5 o l v i ; t
! °i J T e V'^Tc
I 3 U i i :.
' , /
)
. _ ,.: , . \v :a ea>
R ITE( j ,41
J
- '. FCrl 1AT( /// , 10X, • A*EA2 Ai.RAL • E L i
' vI T S • , / J
[Tc (6,^3) (A.-*EA2( I ) , 1=1
v _\
•. m
1 _
. [Tc ( 0,-3 ) (ARE \2( I ) , [=1 , 50)MITEI ) ,<*5 )
FCR.'I AT ( /// , 3X , • A ( I , -K ) • ,dX , ' A ( 2 , K )• , 1 X , • A { 3 , K )
•, d X ,
L ' »J ( 1 1 K) • , 3X , • i3 ( 2 , X. ) ' , IX , • d { 3 , \ ) ., IE a E K = L , 50 '
)C f«! C = 1 , 2 3 <3
- [T£ 16,431 (A( I , <), 1 = 1 ,3) , ( j ( I , K] , 1=1 , 3 >, A,?EA2 COF ." % !A T ( 6 E 1 t . ^ » i. ) X ,
-1 -r . j )
c: : T a t r
- %,:?ITE( » ,5i J
5i -. 1 A T ( / / / ,
ITE ,52
)
; FC* 1AT( IX, 1
.KITE! » f 53)> - : - i at ( / / /
,
> [ T E ( .•> 1 5 2 J
5 5 V T ( / / /
L • Y 30 T • , 1 1 X i
: ' T - £ T ';1
• v 4 X: r E ( 6 » 5 6
56 F C 2 :U T ( 3 X ,
:^ ro loooo
CJC3 ll NATE THi [>1TS <L0CATc>
L ix, 'x .: , i •.. i jates 1, / j
I (LJNG( I ) ,1 = 1, IGN IDE )
2F10.1 )
1 )X, ' v C JO .0 1 IATES' , / )
v VLAT (11,1 = 1, IUMC IE )
)
, 6X y • C HA I *r L LE.4 j F i «
, LX , • : 'ANNE L WIDTH 1, 6
X
'Vi'j' , 1 )X ,• JE L r - O , IXi 1 )EL T A X ' , SX ,
, •CGSIGLIS VALJE ' )
) X L , W , YT'.l P , Y '6 j T , L)E L T A Y , >E I T A X , T h E T V} , \/ E L} E 1 5 . 5 )
r THE J- DJ JCTS < r .
' e ;- >
"* " J TC I All1 F : AT{ ///, 1'IXf » IMME? J ~ ;jjCT Vlj =
: ; r ELE ENTS' ,/J
/ t
.tr ITE (6 ,62 ) (VI J( EL 1N6K ) ,cL -'•' iR = I ,3 ))
i2 FCS'1AT( 1X,6E15.6 )
ITE(b,63J;3 " ' \T( ///, LOX, • I VNER PK ?J^CT
d: 5 - [=i,3J. [ r E ( 6 , 1 5 J [
L - : " kT( /tl')<» •=LEME*IT)C - v = L ,
'
: j /•r = i,3
17 ,
r T: i ,-ni ( VI JK C\, ^,J, I ) , J=l , 3 )
,j FC3.'IAT( lX f 1F16.4J
C V J f V I >
< / j , / K , V I > ', / )
, 15 ********£*******<:-» i •f / )
->o
:
'.
i
*1
j i
.;
"
.
":
f-;r '
i
;
.-.'
: r
L I 7 X
,
>/ )
s
-
-
;cntJ _
: i
=(
I
\ r
» > 7)
/// ,
= 1 ,-t57JU 3
.5 n
L X, • INNF. ' 3'JUCT ,o .
[ < -iJ P * 3DjCT /J\ - Cv'JV,/)
( v J v ( J , I ) . J = L , j ) , ( / J Y ( I J , I ) ,
F 1 6 . 4 , 1 OX , • ** « * « ' , 5 X , 5 f I -. . - )
L ) X , • I N iM -E R -
11 ; / J X I < =
; - : } J C T V J Y IV = <^ / J Y , V I Y >
7 J X ,
rVI>
JJ= i
/ J X ,
t -
7 I >
i3)
7 r <>ELS lEi
-(
. J
= 1 ,4,43 i
112)= !,.'>
,57)-
jooo
( VJX IX(.<| J i I ) , J=l, 3) » (V JYIY( :<i J J
rs« ,
= 1,3;
CHI Af.J ? i I
73
76
,RIT
Fl
i r•
[T
I
'
IT: .,
IT
E( SAT(E (
E(6
E I
F ( >
AT(_ (
ATI
,71)///,6,72( ix,73///5, 7
,74///6, 7
LX,
IOX, »PHI LJ
L QO)0
) (PHI U,NP1) t I = 1,*J.MCCE)L2EU. 4)HC'JR
i o x , • o s i field r i m E =
) ( >SI(I ) ,I = 1,MCN0DEJ•
. J-.
LOX, 'CHI F I EL J TIME =) ( CH I ( I ) ,1 = 1 ,MONGOE
)
2 E 1 1 . •+ )
CIO
T I -1 E = • , F 4 . 1 , « f- S •, / )
, i"
4.1,
4 . L ,
• 3R S
I ui ; c I
,/
)
,/
)
i
:
13
-
..;
..
c
••
.. <
- F
A
1 F
'
r ._
•. <
F
1/ )
I IT T
[TE(5I 'AT (
ITF. i
[ T E ( 65 I AT I
[TI I
ITElo
I f E (
o
. . _> ,
I:
I
1' _ i >
< MT(«?.S« ,
ITS (
i
" -'
i "_ i
[ T E ( 6• m
-. 14 T<
HE ^E 'E 35 NT 7.\? I AoLE F I ELC S
,81) HOUR///, IOX, »U FIELD0,32) (
!
J(klQOEJ ,,\
IX,12E11.4).3-5) HOUR
TIME -,.,jd=lt iuNCCE)
,3 3) HOUR///,nx,'V FIELDo, 12 ) (V(NODE ) r N !0E= 1, i 3NCDE),84) HOJR
r I m E = '
r fl4) ritlJS///tlOXi'KEFIELJ TIve=6, 32 ) ( XKE { 100 E) , I )JE = l , '.CNF )E), 35 ) >
, I-
///, IOX, WCkTI city fi.l;/Ij , 3 .; ) ( 7. E T A ( iJ 3 E i i ^ 1 ; , j j J c - 1 , N l)
:iT Fi-uJ
, id JE = l , : j^i
t; i
v -_=
<['I T
, F 4 . 1 ,
, F 4 . 1 ,
' ,r <+. 1
r: ie =
/// i l:)X, •OIVER-3/)5,32) (
,37)) I V ( NOD E ,
M
P
1
ri
in, ,87 ) 10 J?
I///, IOX, »0L FIELD T rv- =
5,32) I j J( M-JOE ) rx!Ou«I = L, JC «ICDE)
, 3) HUi J ^/// , LOX, »QV r
IZ LJ T I^E =
1 6,32) ( i.'
[ MO )E I ,'jO )E = 1 , i : jCDE )
,c ') H.. J7
[ ///, 10X,'ALFi FI_^^ THEj » 3 2 ) i \ L F A ( ^
.
j 3 F } , Ji3 J E= 1 , < C N E
3= )
4P =
E)
F4 . 1
F4. I
3P>
i HRS'
• HHS'
, ' HRS
» -4 .
,F4.
,' HR S
, ' 3KS
. I ,'
1
,/)
,/)
' ,/ '
L,
',/ )
• ,/)
3 /
/
)
the = »r f * . i ,
' r i rv 5 •i
: r ; torn
' : 5 )E 3lrMOE\T i :-l: <LcAPi-K>
• - * ! • 1-' < « /I!ATi///inx,»j F:ac r
:
'
e -
-: : ..<».: z :i .
)
• IT
<:( > , -.J >'1 - J<
: =I AT ( / / / , 10x , • V P
i ri^s - »F*,l, iI •< s • , / )
. I TE ( o, >2 J (7 MODE) ,N^OE=L, Jj iJCE)
L I
''
. i a
~:\: >l rsL
PLLFSi - ,: T , M \LL TpE IFzCcwiarY \ — V 3 ' F 3 ISiEJ r
;, • . . T-<E ;/•..->-. |." rriC VE^SATEC 3 LCTTER.
i L r , -n i , H - , P H I > Ar. ,* L , < .'-'!, I F L .:
!I.w'T : FA, P-'E TAJ, 3rLl"\X ,JELFAY, .LA T, ML hUiJT> 6 ) , V ( I 5 c ) t A .< c ( 1 3 6 ) , ^ 5 I ( 1 5 6 ) , C H I ( 1 3 c )
J'
J L 1/ :ETA ( L "n , 3) , J I / { L 3 6
,
3 ) ,PH I ( L 3 o , 2 ) , SA VE (
L
5 6
,
2 )
:.'. /
:-•
; '.;/ :»
•. . r i i : ; , . v* _ j , .-. 1, , :,le vj < ,.> j . in*".
:.
'' (
'•I , \ 3 ( 1 » f : i A ( 1 -^
J » C L ( 1 2 J
iSic.M *• i it ; ),,%,.<: i —),.> n/( : >> ), n est( 7
)
ti T: ! i . . r:
__ :: i
7; , : ra :i; uj
v\r,\ rr TLE/ 1 jcI
/
:m\ titlel/ 1 j h i •,
• •,
• chi •
,
r• .
» /.j.\r *,
' )i\ • /; '.
" ,
r; 'l:.: /• )•,' i.5« , ' : •
,
13', '
; ' •
:
•" . '!*/i,i,o,o,i ,o, c,0,0,0/
J ATA ITEST/1 )O0 )00,1 KOO ), 1000C , ICC , 100 1 10 , 1 /
l' = ?
I --L T s [?K\T(16))KV = 3j
J ? F S = L o 3r. * L _ >L CT S2 ( • 3
~ K < » H * k r , \ P T S » I 3 , N B VY )
r I TL E ( 3 J = ^ ) ...
[ M E < = 1 , 7
I' ( IPLCT.LF,ITEST( I iOEX) ) jG TG 100
3 : r: (10,20,30, vo, " ),6o,7 )) , index
10 : A LL 3L CTS "J ( M Y , PH I , 2 ».'i? 1 , *0.* K , N P T S , 1 . , W w AX ,* M I \ )
j rG j ^
«;] t -LL_ JLCTi3 (PS I , 3U 1Y,1, IP1 ,.- iKK ,NPTS, 3 . )0 H , .
' <X , .-, ^ I m)
ji ' i
30 CALL 'LETS., (CHI , CU 1 Y , 1 ,4P I , V.jRK ,NPTS , 1 . , WVAX ,,,'M 1 N )
V TO JO
^0 :^Ll :,LoT53 (J, jU:MY, l fNPL,wJRK,NPTSf l.OfrtMAXfWViNJJC TO ?
5 3 E\LL PLOT S3 ( 7, )UtMY , 1 ,MP1 , * CP. < , N°TS, 1 .0 , WM AX ,WM[ N )
)C T( [
I CTS'-J (Q'JMY ,ZETA,2 , <
J . , i'CkX ,.\2TS ,0 . 1E+03 ,.i !AX, wMIM )
.CTS.3 ;' < , 71 / ,Z , :' i , JKj.NPTSfl.OfwMAX^MINU
. ; )F = • S( .1 IAX-f 1IN)/11.;l( u = «"!\3 r ;> I = 2, 12
$5 CL( I ) = :l< 1-1 I +""
: 1 fLti - ) = 7ITI.E I ( [?!l)EX)
1 )'
. .LE. 0. 1+) T ITL 51 j•
» G G • L • V ) TITL1- 4 ^
. |UC c • * ) ti rt_ ; w'
• o • • •
)
T I T L- - • JL 1 • ' )
": n ~
£.-
. • O ' • . . • ! (_
• 'j E • L -t- r i i'u
-
_
i . j E . '- 7 . T I" L
~2
riTL.- • o c • . ' • _
• G E • 2 J • i TITLE•
- .",-..: .• 'I T
:-
'
lr '
. J - . 1 5
.
f ITL rE;
'
• j E . + L
.
TITLE ->
.
_ i '
• G E • <+ 7 •* T[ "L E U
n tl -i < i_
j
r in ::2(•> 1
titl 12 i
r i tl >4. t
mi - ;;
T)
T:~ l
i, )
11 Tl-
i
_ ^
7|
- f .
:2 ( 1 )
i ; ri _ T / 3)
UTL E2( : ))
TITL E2( 1 1 >
1 I Tl E2( '.2)
T [r L E2l L 3 J
I T L E 2 ( L 4-
)
:- = i.o
< ; = -oooooo.o/
- ^ -3 000 )0.0
:ALL ;j:USJ (w.)KKX,W3RKY,rfOP;<,:NlPT3,Ii, If wC r i (S , Y S f T I TL5 , X A X f V AX i iA f I ER )
I PLOT = I PLCT - I TE5T( I.JOEX J
« * I T E { o , 9 5 ) (EhC CP1AT( LX, • IES = » * 13 1
..; .'i ije
ENJ
» i i w »
PL IT;
\ . nurpjE 5 l :: "2 u^kx,*orky ,
; > - i'
(i L>i r r < T if -
i'^
. ; v ;>> ; .3 jr ES t ie - ' J id . .;t ?oi
-' - : :' J • ;'= '
''
• ''
- •
- _-'_:.- »'•» J
r i i '_ t . \ , t"".
f _L"i-.t h c r \ o , < l f , .
:
l »- : :
J
:;• l ^:.:( l'-i-.i t\L ;rt l^i j
r
1 :i «•- i » i » > it ^Y )
T) i j ;
-.p IL L 3
[uti TH ! M - -
F I :U •
1 1. - •
« 1 » t3l»
•
L
» ~ ri — r _•_ -.
» 1
T* » ; l : G , u 1
' r 3
.
r ,<
l_ < E » 1 • ».
.
1
t•
-"0
» A '.'r J r> _
i
•'
» -j
j l i- )
J I'-'
c i ) ,i "IY )
; : > j t:- x , t
[ !-. i= L -\ T = iLAT/2: -l r
J
: = iu ii/:IT = :L AT 4- 1
=
[ -: j : = :
10 LAT =
L5
:
v LAT = Li ITn LC.V3 = L, NL-JMG•'•.. )E = NGDE + 1
. = M0JE1 1>.:-'. 30 E ) = < LO i\ J ( NGf, E i
<C <Y(\|QO£) = vlU! IOOE1J: [iujc.ej.ij :>o t j is
) = i* ): -= node + :
. . LX('ICDE) = <LONG( i 3Dc L
iC 3 < X ( J CD E J = <
L
.CRKYCICDE] = fLAT(*!nOElJ:l rj 20
) + ) E L T .W
:oq = oCCNT !' JE
3UILJ ihE I a 50 INDA^Y POINTSICCJ'iT = 1
Hj
30
j '-
IT = ICC JMT' i : - » in r\ c _-J = NODE - ( .L : IG*2 f i )
MT) = JulEJT = iLCNG - L
i ) : IDE* = i , <T
: ; ] /._
= i
:
:unt + i
: ,i = -_ 3E - :
rmrr.iviMTi = j- n?( ICD-JNT) = WOE
"
LSI
PLJTiS ( 31 ,2
• , T h:ij - r =% a "
1 5 »- £
R
IK • F - -
'•/cm 1 / . :
;'
: ' t - •
' / c /, - . r •
' ' = MAT + I
( 10,50) ,
FTELJ T2
j ( L rfs) , YL » T ( .
PE, LiUCKK,IDi
be PLcrrto T
JvJ ,PHT:3 \.'<, "
... ui':me: )
en,f
He
\X, *N!IV
J)
j i
[TYPf;
f :q -
. .
.",
"":
'-
: N! v; :t ;•
•'
- • = Jl C 1 1*';
i ' I = J L ( L J
: 3 F 1 =
. )0 =JL 30 LA" = Li IT
32 2 3 long = i, nlcng)£ = NCOE * I
MCOEl = ilJJEl * I
: < i . )E) ^ 31 ( NJUU51 ) - FIF (/tC K(V| :E) .GT ... ' \o i.MAX = mC3K(NC0E)IF ( -4
.
"< '<( V. 3 £ ) . L T . i 1 1 M ) iM I 4 - v^;\i (,j jF)
-
:
IF (lC33.Ej.ll 3' T^ 25: : : ) = l
3C
o •>
-2 = ^ODEl - jL: .3 L
.i2« <( NODE) = Jl ( \C3 C 2 ) * Fir (..--. <{.Nin ;e).gt«4-ux) < ux =
IF !..'<(i.::).LT..'M -r-»I ^ =
3 : : 3 3
i : 3 3 = n
3- 7
st : <e !'.l t, dmensi jnec u<<< w
1 1* A X = ) ^ ( 1 t L) v
F
•': : = D2 ( 1 f i_ ) * F
IC0E1 = 3
Zjc -ICJ-3 = 3
^3 L A T = 1 , \
T
JG .3 LONG = L, ML0M3: 3 E = ' ? 3 - 1
ICJE1 = 100 El « I
• CriK ( MODE) = JZ( M0QE1 , L) * F[F (hOkK(NMCE) .GT .W UX) MAX =
: - r« -j-wijdei.lt.wmim; . it ri'jsif ( ; , ;
:. -. / .1 ) j 3 t -, o5
i = : o c + l
)Ei - v_; . 3 *
i i=
L
< A !Jt) = 321if (.;-.<(:: =
) .gt ..••
:
;
;
( aOk k ( >i-j j E ) . l t . vi
K 7' 73tCJO = 3
: ...:: ; JE
2,L) c r
IAXJ v -UX =: ; ) rfM i n
«• Q 9 K ( \C\3E)«CRK( IQDE J
«CRK( j"3F)rtCKK( jOJE )
h: C S K (
" ! E )
*CRK< MODE
)
. E i J ? .
SUJ^'TJT IMc HA'JAL
IAL - S f T S P r C HA'J '1GN1C A.NAM.YS I S » JONE 3Y HARM A.N. EACH• [S .r INDEPENDENTLY. I ,T^ J IS CMLEO TO I JTc K pr._ A IF.
kisl AJLc N ^ J> L J T 1 3 N F I 5 L D i "J v JiS 1 n.
LCi. 5
-;
:ai_* !
;
: . .
/ ;<: '.:..:, .
: l i r , . ,
,
h i.
•,p «- : v « , \ * i , \ , :
> i , r
L VT ItTAtT !ETAj,DELTAX, JELTA* ,ML A T ,\LC 1G , Jl 5 6 ) , j .: p r , .
«; r k [ ( 1 5 ^ ) , a ( l 2 ) , y l ( 1 3
)
'"/ : 11/ H 156) ,V( L56) ,XKE( L5a) , >SI ( L~6) ,CHI( 155
: M ;/ C U J / ZE r A ( L5r> , 3 ) , D I V ( 15 o , 3 ) , PH I ( 1 5 o t 3 ) , SA VE (
: L6/.JJ ( 1 5 6 ) » ! V ( 156 ) , \L F A ( i 56 ) , OFT A < L56 ) , ON 4
:C-1 fl l/I^.Mir.O) f .i«Vc: % - L , ;>,LiAPH • , "J'• i.3-
DI »E ISID.N I TEST (7 ) , F I EL J( i.3o }, .(12)
2 A T4 [ T EST/ 1 OOO 1 , I > C DO , L ) ) ) » 1 C C , L 00 , 1 , 1 /[ANAL = IP<i\TU3)
LOO I' ;.-.-x = L , 7
IF IIA,«jAL.LT.ITEST(IIOEX) ) Hi rj LOO
DC TC ( 10, 20, 30 f 40,50,60,70) , [MCE*
DC 15 ICDE = L, MCNOOEFIGLD( IGOE 1 = PHI ( 'ICDE,MM )
DC 25 MCDE = It NCNODE•=
I si d( i cos) = ps I (nude)jc r j -^ c
DC 3 5 iCOe = L, ^CfiUDEF [ cLClN^JE i = CHI ( :;j E): ; r ; \ o
l:
L ( 15
: as •of = : ,C: U 5
3 5
7-:
75
; j
Jl
-
j 3
.. - ..
FIELD! ICDE ) = J( "ICJE )
i c ro io
JC 5 5 ICDE = L, NC IGOEFicLDt ICDE) = / ( MGOEJ;i rc jo
: >5 ICDE - L, NCNODEFIELD! NUDE) = ZETAP! TDE , J? 1
)
3C TO JO
)C f5 I ODE = Li MCNDDEr
i: t_J( ICDE) = )IV (
'.' IDE, jPI )
miTc ( 6,3L) AUNN 3- , HOUR= C R 4 AT ( » L « , / / / 1 2 X , A 3 , 1 OX » • T I It - • , F 6 . 1 , ru UR *
, / )
.•
"> = . L A T ¥ L
. . >5 J = . i .' 3
J,"' = J - L
< = j 1 * ILlMG + I
I I - I
;ALL INTE -^ (XL ( I ) ,Y1 (J),,NTRI(i<) ,FI5L0,W, II )
3 ALL iA R.M.N 1 [ J, JM ,*, INDEX)I ANAL = I ANAL -ITcST { INDEX): : .i ; riE
• :r
j- -,
i. ->->
:UTIME [NTEaP (XXfVYt JF.sIf FIELJ»w,£ J
- >r
: jc- ' • ./
'' /
"''' /
; .1 1lm /
[ 'E .31
J =
i IE
.
L
ELFI
= (L( A 6 3= FI= ^L= /L= fL
( T{ TJ)
L )
= ( r= F I :
I I J =<ITE(b: -,-
.
: <at(
?:jT
3 jFhe '
;
-: l
: l / .
: -< i * /c • _ /
-
C 17/ <
2\ FI
L T A X *
/ 2.lE'in<£ JT<*E'iT (
ELO< i
3N «i I I
( r3el )( :
ONG( I
VT ( I I
AT { II4* r
j
. _ j .
i - :
L3< it
1 1 J D )
INE I
EVE 1Li
•' '
'
'..
T,
:
-
TuETAL ' - •
! Tl . : ; (
ELJ( 1
ILL I
J
-
j r<
: ,
J T ':I ,
: )-
i )-
>.GT.I ) -I ) -) - Y) - y_ ™ *
*r^ )
*TC )
)- :
« * <. «
If jtk
^1 T E K P L A T £ 3 * /
v ii» -\CcO L^r\
' -
' ( :;Xt YY J
r I P I C A T I i n)A6
'
I \i>LE kESCLUTIGN F Ic L
C L E S i a . AS i \ ? JTh: ani ! T TQ J
,E i \T = VP
2 'j I
v r
r , ii' L A T c j ,
! r ElE
\i 1 1, is r,\ . : , ..,::,
;
» 4JELMT , a . ,-«.
] J.-, ,-J • . > ak , iM L ,i
vJ ,.\'P 1 , IFL.!
,T IE TAJ, ,J.m.-L t
,v
. iLi r ,\L :"^,.T> J 1,3) ,
3 iilJlF,ZT \MF,D! "OIF1
; o ) ,
x- :_ -", r ( L 5o ) , V'J P(ji : . i- f iOK ( 1 5 o >
56) , i( 12)
L )
2 )
3)F I ELXL', nI
< I
FIEL<l 3r
j
LATILAT{T ) )
) v:/
T
/ r
*YL A
I =YI , X X
)(JJ
)Td}(ixK
Jj )
<K)/ (
TQJ
3
T( i I
,YY,
)
)
= Tj XL)
T J*T : - [;*tj)
5
) - J * X L C M ; ( 1 1 )
FIEL i (ID, FIEL MJJ),FIELQ(KK), ( I ) .
^5
14, 7F LT. 3 )
I )-*
J im : 1 (' '1 , J i , Y , I F I £ L 3 )
i/CML/\CNO -
:
, . =LMT,>^4fHru;A,PHI3AK, \iMl,\ f
i/ cu a/ rnt r a , r = ta j , oe lt ax ,oelt"4v , il at ,.\l)i *•: " :
' /( L2) ,fc( 12, 7 ,: iu:,'j
> 1 , I r L
3 1 J r
- I
-'
. .
Jl 71, -L )
, -I T
3 : /
•
• /
j r ( L 1 2 , 3 , - , 5 , 6 ,7
) , [ F I i L J- ; = F I E L ( 1
)
it T J -»
_: .J = rIELJC)
3 . : = = I E L ( 3 )
> T
• Flo = fielj(^)
~ - . ; = F I E L J ( 5
)
:r r j ->
b F L J = F I E L 3 { 6 I
' FLO = FIELTC7)
-.
-'= iLONGJ * s i LA T +
1
_ - )l L TA X[ 1 =IF ( J 1 ) 11,10,27
. : j l = jl i
Ml = (11/2) * L
112 = 111-1. = b.23 31 j,53/fl:a t
( : :
)
__ = L.0/FLCAT(M12)
."!ITE(6 f l9I FL^, I L » JM ,Z,N12. - f . v i \t( // , \:;, '
z -i :, : j points = •1
1
«' ,-i i.s i J J JI ITS • ,I3t-3X,- •
. : i .(---'_ .' = — ' cL't.opXi , I
: jT , \ /= kl.MBE"* MJiSIBLE" 1, 13
3,5X,
,//)
: 3 = 1,1-1
FCil ,JJ = E2*C0S( FLOAT< {
I
_ :r
; ; : ,j] = c j *-3i m< Ft iat( (
i
: : 5 i = i,mF C ( I , 1 1 = ) . n* F C ( I , 1
J
- FC( W Nil) = 0. 5 *FC( I,.llll)
_7 LI = Ii/2) + 1
.12 = 111-1«•: +0 J = L,M1
. ,' = KCiO : = l,i -i
;:
5l.i = jj; * y ( : i *fu i , j )
- F ( j ) = j ~»'
>*( J-l)) * (J - i )
> * E 1 )
i r { i j = o.oj F ( . 1 1 1 ] = ) . }
: . .0 J = 2, \12= ) .
JC 50 I = L , I l
id 5 J "4 = iijM Y ( I )"F >{ I , J J
J F ( J ) = j j I
Lo5
L I- J =
: ( J ) = ) . 1
f ; j > .:
; . \ m ; -. n
1 1 ;> P ( J ) . ; e . ) . o ) . AN !i . ( A f ( J ) . c . o . o ) ) ; 3 t\hc ( J ] = \ T-\; j ( i3 F I J ) / ...
:( J ; )
j
1 L*
L - c
I 1
12M
*. -1
L
)"
•
; J = L f Ml: \ s (
?
h :,
:
e ( j ) ) - ) . 3 ) >OioOi'
L. ( o) = ^ri J) /CC5(?MAiii J) )
; : r j > 5•L( J) =
: : , r r \ 1=
no j => h \ ;> if 1 j ) =
( J ) / S I N ( P HA S c (J ))
1
»
1
,
7. 1
1 1
7 r:?HAi5( yj /
. . . J = 2
,
.-
( v?u j) ) no, 12), i:IF (PHASE (J) ) 112,112,115
xsc( j ) = ?hase( jj * 1 )
3C rc 113>HA >t( J )
= JLI"» b ( J ) - uo
A ' J L( J) = - V'PL ( J )
C C N T I IJ 6
[ T 2 ( b 9 1 2 1
J
= C* IAT< / ,• ' ,r=,
«
n r r« , rz ), • .*. <
:
:t[c :
'Ea.T o , • p i * S E
»r)
122i^ITE(r»,122) 1M,AMPL(L)i ( ( A.4PL ( I > i PHASE ( I ) ) , I =2 ,M 1 )
F 13 IAT(T6, I2,U6, E11.4-o{T32,E11.4,12X,Fd.2,/) )
• :-t j
LIST OF REFERENCES
Bathe, E. and Wilson, E.L., Numerical Methods In FiniteElement Analysis , p. ?1-123, Prentice-Hall, Inc., 1976.
Cullen, M.J.?., "a Simple Finite Element Method forv eteorological Problems," Journal Institute of Math s
Applies , v. 11, p. 15 31, 1973.
,"A Finite Element Method for a Non-Linear InitialValue Problem." Journal Institute of Math Apply , v. 13,p. 233-247, 1974.
and Hall, C .Z . , Forecasting and General CirculationResults frorr Finite Element Models," Quarterly Journalof the Royal Meteorology Society, v. 105, p. 571-593,July 1979.
Haltiner, G.J. and Williams, R.T., Numerical Prediction andEyaajPlc Meteorology , 2nd ed. , John Wiley and Sons, Inc.,1982.
Einsran, I.E., Application of a Finite Element Method to thejarotropic Primitive Equations , M.S. Thesis, NavalPostgraduate School, Monterey, California, 116p., 1975.""
lelley, R.U., A Finite Element Prediction Model withVariable Element Size , ^.S. Thesis, Naval PostgraduateSchool, Monterey, California, 109 p., 1976.
Norrie, T.E. and de Tries, &., The Finite Element Methods .
Academic Press, 19 73.
Older, M., A Two Timensional Finite Element Advection Modelwith Variable Resolution ! M .S . Thesis, NavalPostgraduate School, Monterey, California, 84 p., 1981.
Phillips. N.A., "Numerical Integration of the PrimitiveEquations en the Hemisphere," Monthly Weather Review , p.333 345, September 1959.
Pinder, G.F. and Gray, W.G., Finite Element Simulation inSurface and Subsurface Hydrology , Academic Press, 1977.
Schoenstadt , A., A Transfer Function Analysis of NumericalSchemes Used To Simulate Geostrophic Adjustment,"'••Qnthly Weather Review
,v. 10e, no. 8, p. 1248 1259,
August 1980.
157
ShUFdB
,
Numerical Weather Prediction, BulletinrTericg-i vetecrolcgy Society , v. 59, no. 1, o. 5-17,January 1973.
Staniforth, A.N. and Mitchell, H.L.,„ "A Semi-ImplicitFinite Element Earotropic Model," vonthly WeatherPer lev , 7. 105, p. 154-169, February 1977.
Strang, 3. and Fix, 5. J., An Analysis of the Finite ElementMethod , Prentice Hall, Inc., 1973.
Villiars. H.T., "Ont?the Formulation of Finite Element
Prediction M odels," Monthly Weather Review , v. 109, no.3. p. 463-4c6, March 1981.
and Zienkievi.cz, O.C., Improved Finite Element FormsFor The Shallov Water Equations,' International Journal:or Numerical Methods in Fluids , v. 1, no. 1, p. 81-97,January varch 1981.
Zlenkievlcz, O.C., The Finite Element Method in Engineeringcler.ce, Mc'rav Hill, 1971.
168
IMTI.AL ? 1 3 C ?. I i U T I C ^ LIST
C . .CUES
1 Information CenterC are ron StationAlerandria, Virginia 22314
2. Library, ~:ce ?1421
r. -'a 1 Postgraduate Schoolv cotere;-, California 9 2-5 4c
2. Dr. R.T. Williams, Code 53WuEepartrreat of v eteorolc<?yNa^al Postgraduate SchoolMonterey, California 93542
4. Commander, Air Weather ServiceMilitary Airlift Command"cited States Air TcrceS""ott Mr Force 3ese, Illinois 62:226
5. Certain Edward T. Woodward44 Felene T r .
3ellevue, Nebraska fcZcS
f. Certain Erian "an Crran.-jit ::??V right-Patterson AF5, Ohio 4^432
Dr. David A. ArcherTouglas ErPont Rachford, Inc.£150 Chevy Chaserouston, Texas 77027
?. Frcf. Gordon Bradley, Code 523zCepert^ent of Ccrruter Sciencefcaval Postgraduate SchoolYcnterey, California 93940
12
1*. ° . S . E u rge r , Code 35Department o: MeteorologyNaval Postgraduate SchoolMonterey, California £3940
Prof. Silles Cantin, Cede 69CiDepartment of Mechanical EngineeringNaval Postgrad n ate SchoolMonterey. California 5394?
1-
VC -.i lien
Vptpff .nl '>»' <a1
i •CI.'.j i li-i.lt.C-k Li. X .1 c, 111- A
17
J". ?.L. "Isberry, Code c3:ls-r7^r r ""£.it of Meteorolo gy-" = 1 ? os t rrac'ia t e School
v r:e:e;-, California 9394J
Dr. J .A , GeltD:o V - ~. c ai l.iv ii j_a-
7r. iv e r s i t y of WashingtonSeattle, Washington 98105
14. It. ?..L. Eaney, Cede 63H-/Department of Meteorology'aval Postgraduate Schoolv :-terey, California 9394:,:
15. LCE5 Tcnali E. Einsman, Cede 35repertment of Meteorology,\'a-al Postgraduate Schoolv cnterey, California 9394£
If. Prof. Uno ?.. Kodres, Cole 52KrDepartment cf Computer Science"aval Postgraduate School^OTjterey, California 9394C
r
.
ert L. LeeAtmospheric end Geophysical Science livisioTr.iversity of California?.C. 2 ox 3C8Llrer^ore, California 945S<?
18. Pro*. .-..'£. MacPhersonLehigh UniversityDepartment of Mechanical EngineeringBethleem, ?a 18015
oo
- - « -- 2 M ;' vtp-c Code ^nM
r
Chairman* Tepartn-ent of OceanographyVaval Postgraduate SchoolMonterey i California 93942
Prof. ?.. Newt en, Cede 69Nerepartirent of Mechanical EngineeringNaval Postgraduate SchoolMonterey, California 9394.3
1 7 Z
21. rr. V-
; .A. Phillips 1
optional vetecrolc*icai Center/NOAAk o rl d » ea t he r Vj i 1 d i n g
c^ . . r . - . F o smcnc! 1
Naval Environjental Prediction Research Facility"onterey. Tali for- id 9 39 40
2?. Prof. 7. Salinas, Code £9Zc 1
Department of v e c nacical Engineeringv f-^l Postgraduate Schoolvorterey, California 93942
24. Dr. T. Sasaki 1
Pepartrrent of MeteorologyUniversity of OklahomaNorman « Oklahoma 73069
c - Prof. A.L. Schoenstadt, Cede 53Zhrepa-:~ent of MathematicsV'c'-al Postgraduate SchoolVorterey, California 93940
26. Dr. Andrew StaaiforthHecherce en Prevision NumeriqueWest Isle Office Power, 5 ie^e eta^e2121 route Trans-CanadaPorval, Cuetec RSP1J3, Canada
Mational Oceanic and Atmospheric Administrationi r ckenoacker -a usewayiairi, Tlorida 33149
-* *
P r . .- . 'AeinsteinKaval Environmental Prediction Research FacilityMonterey, California 9394^
- T V 4 ry n A n ifh n f *> r ^ d a T r
repartrent of v et aerology' a T'~l Postgraduate SchoolMonterey, California 93943
D rof. j . C . Zienkiewiczread of Civil Engineering Pepart~entApplied Science Building,:Singleton Par te
>,' 3 - c £ £ S-~- QPP"ni'ai Kingdom
171
Thesis l'^/OW834 Woodardc.l Development of
improved finite ele-ment formulation for
DEC £.^hallow water equa-tions. 3 3 2 5 9
15 3 3 5 3 6
Thesisa . Jo . > . 'tc
W834 Woodardc.l Development of
improved finite ele-
ment formulation for
shallow water equa-tions.