Академический Документы
Профессиональный Документы
Культура Документы
Stabilityofspatialinterpolationfunctionsinfiniteelement
onedimensionalkinematicwaverainfallrunoffmodels
LuongTuanAnh1,*,RolfLarsson2
1
ResearchCenterforHydrologyandWaterResources,
InstituteofHydrometeorologicalandEnvironmentalSciences
2WaterResourcesEngineeringDepartment,LundUniversity,Box118,S22100Lund,Sweden
Received27May2008;receivedinrevisedform5July2008
Abstract. This paper analyzes the stability of linear, lumped, quadratic, and cubic spatial
interpolationfunctionsinfiniteelementonedimensionalkinematicwaveschemesforsimulationof
rainfallrunoff processes. Galerkins residual method transforms the kinematic wave partial
differentialequationsintoasystemofordinarydifferentialequations.Thestabilityofthissystemis
analyzedusingthedefinitionofthenormofvectorsandmatrices.Thestabilityindex,orsingularity
ofthesystem,iscomputedbytheSingularValueDecompositionalgorithm.Theoscillationofthe
solution of the finite element onedimensional kinematic wave schemes results both from the
sources,andfromthemultiplicationoperatorofoscillation.Theresultsofcomputationexperiment
and analysis show the advantage and disadvantage of different types of spatial interpolation
functionswhenFEMisappliedforrainfallrunoffmodelingbykinematicwaveequations.
Keywords:Rainfallrunoff;Kinematicwave;Spatialinterpolationfunctions;Singularvaluedecomposition;
Stabilityindex.
1.Introduction1
berelaxedbyconsideringthetotalflowtobe
the result of the flow from many small plots
drainingintoafinenetworkofsmallchannels.
Theactualphysicalflowprocessesmaybe
quite complicated, but for practical purposes
there is nothing to be gained from
introducingcomplexityintothemodels.Asa
common way of getting optimal results, the
onedimensional kinematic wave models [2,
5, 8, 11] are often selected. These can be
solved by different methods, one ofwhich is
the finite element method (FEM) which is
analyzedinthispaper.
TheFEMmodelsarenormallyderivedby
the weighted residuals method, which is
_______
*Correspondingauthor.Tel.:844917357025.
Email:tanh@vkttv.edu.vn
57
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
58
((h) f )Wi = 0 ,
where:
(h) = f :partialdifferentialequationofh;
h ai N i :estimatedsolution;
i
Wi:setofweightingfunctions;
Ni:functionsofspatialordinate;
ai :functionsoftime.
According to Peyret and Taylor [9], the
weighted residual method is a general and
effective technique for transforming partial
differential equations (PDE) into systems of
ordinary differential equations (ODE). When
hi , ai and Ni are functions defined on a
spatialinterval(element)themethodiscalled
FEM.Thespecialcaseofweightingfunctions
Wi = N i iscalledGalerkinsresidualFEMand
it is often used for solving onedimensional
kinematicwaverainfallrunoffmodels.
The numerical solutions of the finite
element schemes for overland flow and
groundwater flow in one dimensional
kinematic wave rainfallrunoff models may
often run into problems with stability and
accuracy due to oscillation of the solution.
The scheme may be considered stable when
smalldisturbancearenotallowedtogrowin
the numerical procedure. The reasons for
oscillation of the Galerkins FEM method for
kinematic wave equations have been
discussedbyJaberandMohtar[5].
Oneimportantfactorwhichinfluencesthe
stability characteristics of the method is the
choice of spatial interpolation function. Jaber
and Mohtar [5] used linear, lumped and
upwind schemes for spatial approximation
and the enhanced explicit scheme for
temporal discretization. They analyzed the
stabilityofdifferentschemesthroughFourier
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
(1)
q = h
(2)
where:h:flowdepth(m); q :unitwidthflow
(m2/s); r ( x, t ) :effectiverainfallorlateralflow
(m/s); = So1 / 2 / n ; = 5/3; n : Manning
roughnesscoefficient( m1/3 /s ); S o :thesurface
orbottomslopethatequalstofrictionslopein
thecaseofkinematicwaveapproximation; x :
spatialcoordinate(m);and t :time(s).
Equations(1)and(2)arepartialdifferential
equations which have no general analytical
solution.However,withgiveninitialcondition
h(t=0)andboundaryconditionh(x=0),numerical
solutions can be found. The kinematic wave
resultsfromthechangesinflowandsinceitis
unidirectional(fromupstreamtodownstream),
onlyoneboundaryconditionisrequired.
Principles of spatial discretization for the
onedimensional kinematic wave model
using the FEM method have been presented
byRossetal[11].Thesurfaceareaoftheriver
basin is divided in the crossflow direction
into strips. Each strip is then divided into
computational elements based on the
characteristics(e.g.slope)ofthebasinsothat
eachelementisapproximatelyhomogeneous.
For each computational element, the
variables h(x,t)and q(x,t)areapproximatedin
theform:
59
n
h(x,t) h = N i (x)hi (t);
i =1
(3)
n
q(x,t) q = N i (x)qi (t)
i =1
where: N i ( x) : space interpolation function
(shapefunctionorweightingfunction).
Itisnotedthattheexpressions(3)should
satisfy not only Equation (1) but also the
initialconditionandtheboundarycondition.
The
Galerkins
residual
method
normalizes the approximated error with
shapefunctionoverthesolutiondomain:
M dh
N i
i N i + qi
ri N i dx = 0 .(4)
dt
x
i =1
Theapproximation(3)combinedwiththe
integral (4) transforms the partial differential
Equation (1) into a system of ordinary
differentialequations,whichforeachelement
(4)takestheform:
(e)
(e)
(e) dh
A
(5)
+B q f
= 0 .
dt
For the linear scheme, the spatial
interpolationfunctionscanbedefinedas:
N1(x) = 1 y ,and N 2 (x) = y ,
where y = x / l ; l isthelengthoftheelement.
In this case, the matrices of Equation (5)
arewrittenas:
1 1 1
B (e ) =
;
2 1 1
l l
l
3 6
(e)
A =
; f = 2 r ( x, t )
l
l l
2
6 3
The lumped scheme [5] is based on the
spatial interpolation functions expressed in
theforms:
l
l
N *j 1 = 1 H s ; N *j = H s
2
2
TheheavysidefunctionH(x)isdefinedas:
H(x) = 0ifx < 0;
H(x) = 1ifx 0;
s:distancefromnodej1.
(e)
60
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
N1 = 1 3 y + 2 y 2 ;
N2 = 4 y 4 y2;
N3 = y + 2 y 2 .
The matrices for one element are defined
asfollowing:
l
l
2l
15 15 30
l
l
8l
;
A (e) =
15 15 15
l
2l
l
30 15 15
2
1
1
l
2
6
3
6
2
2l
2
; f ( e ) = r ( x, t )
B ( e ) =
0
3
3
3
2 1
1
l
6 3
6
2
For cubic scheme (one element, four
nodes), spatial interpolation functions can be
expressedintheforms:
N1 = 1 5.5 y + 9 y 2 4.5 y 3
N 2 = 9 y 22.5 y + 13.5 y
2
N 3 = 4.5 y + 18 y 2 13.5 y 3
N 4 = y 4.5 y 2 + 4.5 y 3
The matrices for one element are
integratedandarepresentedas:
33
3
19
8
l
l
l
105 l
560
140
1680
33
27
27
3
l
l
l
l
(e)
560
70
560
140
;
A =
27
27
33
3
l
l
140 l 560 l
70
560
19
3
33
8
l
l
l
l
140
560
105
1680
1
2
57
= 80
3
10
7
80
B(e)
57
80
7
l
8
80
3l
3
10 ; f (e) = 8 r ( x, t )
57
3l
8
80
l
1
8
2
3
10
81
80
81
0
80
3
57
10
80
A=
l1
2
ll l2
+
2 2
l2 l3
+
2 2
l3 l4
+
2 2
l4 l5
+
2 2
l5 l6
+
0
2 2
.
.
ln2 ln1
+
2
2
ln1 ln
+
2 2
.
0
0 .
.
0
ln
2
B = 0
.
0
1
2
1
2
1
2
1
2
.
.
.
.
0
.
.
1
2
1
2
1
2
2
.
.
0
.
.
1
2
.
.
l1 r1
2
l r
l r
1 1
+ 2 2
2
2
l r
l 2 r2
+ 3 3
2
2
f =
l r l r
n 1 n 1 + n n
2
2
l n rn
= Cr
1
2
1
2
.
0
2
1
2
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
Foroverlandflow,thesystemofordinary
differential equations (6), can be written in
theform:
A
dh
+ Bq Cr = 0 ,
dt
(7)
where: C:sparsematrixcontainingthesizeof
elements;r:vectorofeffectiverainfall.
The solution of Equation (7) can be
obtained by various numerical methods, one
ofwhichisthestandardRungeKuttamethod
and Successive Linear Interpolation for
solutionofODEwithboundaries[4,10].
In order to analyze how the stability and
accuracyofthesolutionschemesdependson
the choice of spatial interpolation functions,
equation (7) has been transformed into a
systemoflinearalgebraicequations:
Ax = y ,
(8)
h
= x :unknownvector;
where:
t
y = Cr Bq :givenvectorforexplicit
temporal differential scheme and estimated
vectorforimplicitinteractiveschemeforeach
timestep.
3.Stabilityanderroranalysis
In order to evaluate the stability of
various finite element schemes, the Singular
ValueDecomposition(SVD)algorithmwillbe
applied. It will be introduced and described
below together with the definition of some
essentialvectorandmatrixconcepts:
(i)AccordingtotheSVDalgorithm[4.10],
the matrix A (mm) can be expressed in the
form:
(9)
A = UV T ,
where U, V: square orthogonal matrices
(mm), :diagonalmatrixwith ii 0 called
singularvaluesofmatrixA.
(ii)Thenormofthevectorxisdefinedas:
x = (x T x)1 / 2
(10)
61
Ax
Cond( A) =
max
x x
Ax
min
x x
max
x
x
U V T x
min
x
= max ,(12)
min
A(x + x) = y + y y min x ,
where: x , y : oscillation vector of solution
andoscillationvectoroferrorsrespectively.
Thismeansthat:
x
y
Cond ( A)
(14)
x
y
The relationship (14) shows that the
stabilityofthesolutionofsystem(8)depends
on the stability index of the matrix A with a
high value of the index indicating lower
stability.Therelationship(14)alsomeansthat
the stability index (or singularity of A) may
be considered as the multiplication of
oscillationy:
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
y = Cr Bq .
(15)
matrixC.
Expression (16) shows that the source of
oscillation include oscillation in the source
termr(effectiverainfall)aswellasoscillation
intheadvectiontermaccumulatedduringthe
computation process. The upper limits of
these oscillations depend on the chosen
spatial interpolation function, and they are
related with the structure of the matrices B
and C respectively. These values will be
computed and the results will be discussed
below for the selected types of interpolation
functions.
The solution of the system (8) normally
requires to inverse matrix A [5, 12]. We can
show that the singularity of the (square)
matrixAhasthesamevalueasthesingularity
oftheinversematrixA1byusingEquation(9):
1
'T
A = U V .
(18)
The decompositions (9) and (18) are
'
'
Inordertoverifythemethodology,some
basic investigations are made for different
types of interpolation schemes in section 4.1.
In section 4.2, the effect of using elements of
various lengths is investigated. Finally, in
section4.3,theinfluenceofdifferentdisturbance
sourcesisanalyzed.
4.1.Stabilityindexofmatrix Afordifferenttypes
ofspatialinterpolationfunctions
Now we assume that the studied strip of
surface area is divided into elements of
(equal) unit length. The index of stability of
matrix A has been computed for various
numbers of elements for each type of
interpolation function. The results of the
computationsarepresentedinFig.1.
12.0
Linear
Quadratic
Cubic
Lumped
10.0
8.0
6.0
4.0
A = V U .
(17)
ApplicationofSingularValueDecomposition
ofA-1gives:
1
4.Numericalexperiments
Cond(A)
62
min
) /(
max
(19)
2.0
0.0
1
4
Elements
Fig.1.ThechangeofstabilityindexofmatrixA.
Thenumericalexperimentsshowthatthe
indexofstabilityisvirtuallyconstantforeach
type of interpolation scheme when the
numberofelementsistwoorhigher.Itisalso
clearthatthelumpedschemegivesthelowest
value of stability index, while linear,
quadratic and cubic schemes give 2, 3 and 4
times higher values respectively. In
conclusion, the lumped scheme has the
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
63
Furthermore,itshouldbepointedoutthatthe
efficiency of the algorithm is an important
aspect with regards to the choice of
interpolationschemeforpracticalapplications.
The linear and lumped schemes require n+1
equations,whilequadraticandcubicschemes
require 2n+1and 3n+1equationsrespectively
forsolvingaproblemwithnelements.
Table1.StabilityindexofmatrixA
andmaximumcoefficientofoscillation
Casesof
study
B
Option1 max
max
Option2
Option3
0.866
Lum Quad
Cubic
ped ratic
0.866 1.29 1.67
404.5
Linear
Cond(A) 3.73
B
0.866
max
2.00 5.83
0.866 1.29
max
452.8
8.13
1.67
Cond(A) 14.6
B
0.866
max
10.0 41.2
0.866 1.29
63.1
1.67
max
495.0
Cond(A) 149.6
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
64
max
0.500
B
max
0.866
C
max
0.809
B
max
1.0
1.0
C
max
0.901
B
max
0.951
0.951 1.34
C
max
0.940
B
max
1.0
1.0
C
max
0.960
B
max
0.975
0.975 1.35
C
max
0.971
B
max
1.0
1.0
C
max
0.978
2
3
6
7
1.33
1.35
1.35
1.67
1.71
1.73
1.74
1.75
1.75
2.0
1.5
1.0
Lumped/Linear
0.5
Quadratic
Cubic
0.0
1
4
Elements
Fig.2.Thechangeofmaximumextensioncapacity
ofmatrixB.
5.Conclusions
This paper analyses the sources and
causes of oscillation of solutions for finite
element one dimensional rainfallrunoff
models when different types of spatial
interpolation functions is applied for
overland flow kinematic wave simulation. It
doessobyapplyingthedefinitionofnormof
vectors and matrices and the Singular Value
Decomposition(SVD)algorithm.
Thestructureofmatrix A,whichcontains
sizes of the finite elements, is related to the
typeofspatialinterpolationfunctionwhichis
applied. From the above presented results
and discussions, itcan be concludedthat the
stability index or singularity of matrix A can
beconsideredasaneffectofmultiplicationof
oscillation occurring during computation
process. It will affect the stability and
accuracyofthesolutionoffiniteelementone
dimensional kinematic wave schemes, and it
is actually one of the main causes of
oscillationofsolutions.
The results of computation experiment
show the advantage and disadvantage of
different types of spatial interpolation
functions when FEM is applied for rainfall
runoff kinematic wave models. If the reason
for growing oscillation is seen as the most
important criterion for assessing stability of
numerical schemes, the lumped and linear
schemes have higher order of stability than
the quadratic and cubic schemes. However,
whenthelumpedschemeisused,thematrix
A becomes a diagonal matrix and then the
algorithmismoreefficientthanallotherthree
typesofschemes.
The results also show that the finite
element onedimensional kinematic wave
schemes can be improved by choosing the
most suitable spatial interpolation function
fordecreasingthesingularityofmatrixAand
LuongTuanAnh,RolfLarsson/VNUJournalofScience,EarthSciences24(2008)5765
minimizethesourceofoscillation.Thespatial
interpolationfunctionsofhigherorderdonot
always give improved results when finite
element method is used for kinematic wave
rainfallrunoffmodels.
[6]
[7]
References
[1] M.B. Abbott, J.C. Bathurst, J.A. Cunge, P.E. O
Connel,J.Rasmussen,Structureofaphysically
baseddistributedmodelingsystem,J.Hydrol.87
(1986)61.
[2] G.E. Blandford, M.E. Meadows, Finite element
simulation of nonlinear kinematic surface
runoff.J.Hydrol.119(1990)335.
[3] V.T.Chow,D.R.Maidment,L.W.Mays,Applied
hydrology,McGrawHillBookCompany,1998.
[4] G.E. Forsythe, M.A. Malcolm, C.B. Moler,
Computer method for mathematical computations,
PrenticeHall,NewJersey,USA,1977.
[5] F.H. Jaber, R.H. Mohtar, Stability and accuracy
of finite element schemes for the one
[8]
[9]
[10]
[11]
[12]
65