Вы находитесь на странице: 1из 6

WATER RESOURCES RESEARCH, VOL. 22, NO.

3, PAGES 423-428, MARCH 1986

Adaptationof the Carter-TracyWater Influx CalculatiOn


to Groundwater Flow Simulation

KENNETH L. KIPP, JR.

U.S. GeologicalSurvey,Denver, Colorado

The Carter-Tracy calculation for water influx is adapted to groundwater flow simulation with ad-
ditional clarifyingexplanationnot presentin the originalpapers.The Van Everdingenand Hurst aquifer-
influencefunctionsfor radial flow from an outer aquifer region are employed.This technique,basedon
convolutionof unit-stepresponsefunctions,offersa simplebut approximatemethod for embeddingan
inner region of groundwaterflow simulation within a much larger aquifer region where flow can be
treatedin an approximatefashion.The use of aquifer-influence functionsin groundwaterflow modeling
reducesthe size of the computationalgrid with a correspondingreduction in computer storageand
executiontime. The Carter-Tracyapproximationto the convolutionintegralenablesthe aquiferinfluence
functioncalculationto be made with an additional storagerequirementof only two times the number of
boundarynodesmore than that requiredfor the inner regionsimulation.It is a goodapproximationfor
constantflow rates but is poor for time-varyingflow rates where the variation is large relative to the
mean. A variety of outer aquifer region geometries,exterior boundary conditions,and flow rate versus
potentiometricheadrelationscan be used.The radial, transient-flowcasepresentedis representative. An
analytical approximation to the functionsof Van Everdingenand Hurst for the dimensionlesspoten-
tiometric head versusdimensionlesstime is given.

INTRODUCTION action of a groundwater flow system with a management


model or a stream flow model rather than the interaction
Many petroleumreservoirsare completelyor partly bound-
ed by aquifers adjacent to their lateral or bottom boundary between one groundwater flow region and another. The
surfaces.As oil is produced,the declinein boundary pressure Carter-Tracy algorithm was first implemented for a ground-
causeswater influx resultingfrom some combinationof ex- water flow simulator in the Survey Waste Injection Program
pansion of the water, compaction of the aquifer porous (SWIP) code by INTERCOMP [1976] which was developed
matrix; artesianflow, and expansionof oil or gas containedin from a petroleum reservoirsimulator.
the aquifer.The aquifer outer boundariesmay be at a finite or Aquifer-influence functions enable the groundwater flow
infinite distance from the reservoir-aquifer boundary. These simulation region to be embedded within a surrounding
outer aquifer boundariesmay be of impermeable media, or region for which the aquifer propertiesare known only in a
they may be rechargeboundaries. generalsenseand where the outer-regionflow field influences
In petroleum-reservoirsimulation,aquiferscan be included the inner regionof interestonly in a generalway. The primary
in the computational grid, but the increasesin computer- benefit of using aquifer-influencefunctionsis the reductionin
storage requirementsand calculationtime are rarely justified number of grid points which resultsin a savingsin computer-
becausethe aquifer-propertydistributionsare seldom as well storagerequirementand computationtime. The use of aquifer
known as those of the reservoir [Aziz and Settari, 1979, p. influence functions representsone approach to the problem of
325-1.Thereforeboundingaquifersare often excludedfrom the an excessivenumber of grid points.
simulation region, and their influenceis taken into account by The purpose of this technical note is to adapt the Carter-
calculating the water influx across the reservoir-aquifer Tracy water-influx calculation techniqueto groundwater flow
boundary in responseto boundarypressurevariationsas the simulation employingthe terminology of groundwater hydrol-
flow simulation proceeds. ogy. In addition, a new, clarified, derivation is presentedthat
Several methods have been used to calculate water influx at facilitatesunderstandingof the conceptsand implementation
reservoir-aquiferboundaries. For a summary, the reader is of this boundary-conditiontreatment. The term water-influx
referredto Craft and Hawkins [1959, chapter 5] and Aziz and calculationwill be replacedby the more generalterm aquifer-
Settari [1979, section 9.6]. The Carter-Tracy algorithm influence function (AIF) boundary condition, because the
[Carter and Tracy, 1960] is basedupon the aquifer-influence latter term better expressesthe ability to handle both inflow
functionsdevelopedby Van Everdingenand Hurst [1949] and and outflow and to select the outer aquifer region size and
Hurst [1958] that describe the transient boundary pressure outer boundary conditions.The sequencewill be to present
and cumulative flow from the aquifer in responseto the tran- the equationsthat lead to the aquifer-influencefunctionsfor
sient flow variation or pressure variation at the reservoir- the unit-step drawdown and the unit-step flow rate, then to
aquifer boundary. Convolutions of unit-step responsesyield give the convolutionintegrals.Finally, the Carter-Tracy algo-
the transient water influx at this boundary. rithm for approximate calculation of the aquifer-influence-
Early uses of the concept of an influence function in function flow rate convolution integral will be derived.
groundwaterhydrologyincludethe works of Maddock [1972],
Hall and Moench [1972], Morel-Seytoux and Daly [1975], and
THEORY
Morel-Seytoux [1975]. These applicationsinvolved the inter-
Aquifer-Influence
Functions
This paper is not subject to U.S. copyright. Published in 1986 by Considerthe simulationof groundwaterflow with constant,
the American GeophysicalUnion. uniform fluid density,and isothermalconditions.Supposethat
Paper number 5W4179. the simulation region can be divided into subregions(Figure
423
424 KIPP.' TECHNICAL NOTE

Boundary where aquifer- uniform, with impermeable upper and lower boundary sur-
influence
functions
are
applied__•
A faces.Groundwater flow in this outer region is radial, and the
potentiometric head satisfiesthe following equation [Bear et
Inner
region
• Oute.•-•"•-•r al., 1968, p. 405]:
S c•h' c•2h' 1 c•h'
- + - (0
Kb c•t' c•r'2 r' Or'
where

Exterior
S storagecoefficient(-);
boundary
Equivalent, K hydraulic conductivity(L/T);
cylindrical boundary
b region thickness(L);
h' potentiometrichead (L);
t' time (T);
Impermeable
boundary r' radial coordinate (L).
Inner region Initial condition

at t' = 0 h' = ho' (2)

•0•region
Fig. 1. Illustration of inner and outer aquifer regionsand bound-
where ho' is the initial uniform potentiometrichead (L).
Van Everdingen and Hurst used two different boundary
conditions at the interior cylindrical boundary: one was con-
aries of the (a) outer region completelysurroundinginner region and stant potentiometric head and the other was constant flow
the (b) outer regionhalf surroundinginner region. rate. The objective was to derive the responsefunctions for a
unit-step drawdown and a unit-step withdrawal flow rate.
1), where the inner region is of primary interest and the aqui- While it seemsthat only the responsefunction for the unit-
fer properties and geometrical configuration of the outer step withdrawal flow rate would be needed in order to calcu-
region are known only in a general sense.The outer region late flow rates at the AIF boundary in responseto poten-
may completely (Figure la) or partially (Figure lb) surround tiometric head changes,the Carter-Tracy algorithm actually
the inner region. Variable density and nonisothermal flow employsthe responsefunction for the unit-stepdrawdown.
may be simulated in the inner region but not in the outer one.
Boundary conditions
In this case, the dependent variable must be pressure, as in
petroleum-reservoir simulation, rather than potentiometric at r'= r•' h'= h•' t' > 0 (3a)
head. The actual flow-model region may be reduced to only
the inner region, and the boundary condition at the interface where r•' is the interior radius (L), and h•' is the specified
betweenthe two regions(the AIF boundary) is representedby potentiometrichead (L), or
the aquifer-influencefunction for the outer region. Aquifer- Oh' +(2'
influencefunctionsare analytical expressionsthat describethe
flow rate, pressure,potentiometric head, and cumulative flow
at r'= r•' Or'- 2nr/Kb (3b)
at the boundary between the inner and outer regions. For whereQ' is the specifiedflow rate (L3/T) (positiveis from the
purposesof groundwater flow simulation, the pressureand outer to the inner region).At the exteriorcylindricalboundary
cumulativeflow aquifer-influence
functionsare not of interest the condition for an infinite region is
and will not be considered.
as r'-• oo h'-• ho' (4a)
The class of influence functions that describe transient flow
acrossthe interior boundary are based upon analytical solu- and, for a finite region, either no flow:
tions to the groundwaterflow equation in the outer region. To Oh'
obtain an analytical solution, the aquifer and fluid properties at r' = re' - 0 (4b)
c•r'
of the outer region must be assumed to be constant and uni-
form, and the geometry of the boundaries must be approxi- where re' is the exterior radius (L), or specifiedhead:
mated by simple shapes.
The Carter-Tracy calculation technique for the aquifer-
at r'= re' h'= ho' (4c)
influence function employs analytical solutions presented by
Van Everdingenand Hurst [1949]. Van Everdingen and Hurst
actually developed the solutions for pressureand cumulative Equations (1) through (4) can be cast into dimensionless
volumetric flow, but potentiometric head and volumetric flow form with the followingvariableand parameterdefinitions:
rate are more useful for our applications. The remainder of h'- ho'
this section outlines the development of these solutions, which h= • (5)
ho' - h•'
will be denoted as aquifer-influence functions for poten-
tiometric head and boundary flow rate. for the specifiedhead boundary conditions.
Let the boundary between the inner flow-simulation region
and the outer aquifer region be approximated by a cylinder of (h' - ho')2nKb
h= (6)
a given radius and height (Figure la). For a simulation region Q,
that is a rectangular prism, this boundary cylinder will be a for the specifiedflow rate boundary conditions.
severeapproximation of the actual boundary shape.The outer
r'
boundary of this outer region is a cylinder at a finite or infi-
r -- (7)
nite radius. The thicknessof the outer region is assumedto be Fi •
KIPP' TECHNICAL NOTE 425

t'Kb and Hurst actually solved for the cumulative-flow response


t-- (s) function:
rI'2S
r E'
R- (9)
W(t)
= d, (16)
Equations (1)-(4) become, in dimensionlessform, instead of (14); but we do not need this function for our appli-
cation.
•2h 1 c9h c9h
(10) The concept of superposition or convolution is used to
•r 2 r c•r c•t
derive the aquifer-influence function from the unit-step re-
Initial condition sponsefunctions for a transient potentiometric-headfunction
at the AIF boundary between the inner and outer aquifer
at t =0 h=0 (11) regions. The convolution interval comes from Duhamel's theo-
Boundary conditions rem, which states that the solution, u(x, t), to a linear differ-
ential equation with a time-varying boundary condition, It(t),
at r= 1 h= - 1 (12a) at one boundary, and homogeneousinitial conditions and ho-
mogeneousboundary conditions at all other boundaries pres-
unit-step drawdown, or ent, can be expressedby the following integral:
•h
at r= 1 --=
c•r
1 (12b) u(x,
t)=[o
•(z)aU(x,
t- ,) dr
c•t (17)
unit-step withdrawal flow rate, where U(x, t) is the solution of the correspondingboundary-
as r--• oo h•0 (13a) value problem for U(0, t) = 1, i.e., the unit-step responsefunc-
tion [Tychonov and Samarski,1964, p. 209]. Thus, the flow-
or
rate responseof this groundwater flow system to a time-
varying potentiometric head at the inner boundary of the
at r=R •=0 (13b) outer region is obtained by letting x equal r in (17), taking the
c•r
radial derivative, and evaluating it at r-- 1. It can be written
or
as

at r=R h=0 (13c)

The definitions selectedfor the dimensionlesspotentiometric Q(t)


=;j-h(•:)
r3Q,,(t-
z)d•: (18)
head, given by (5) and (6), mean that the inner boundary Equation (17) can be shown to apply also to a linear differ-
conditionsfor the two caseswill be a unit-stepdrawdown and ential equation with a derivative boundary condition in the
a unit-step withdrawal flow rate for any specifieddimensional form r3u/c•xl•=o= It(t) by following a developmentparallel to
potentiometrichead or withdrawal flow rate respectively.The that of Churchill [1958, p. 229-230]. In this case U(x, t) is the
solutionsto (10)-(13) for an infinite region are given in Van solution to the correspondingboundary value problem with
Everdingenand Hurst [1949] for the pressureand the cumula- r3U/c•xl•=o- 1. The potentiometric-headresponseto a time-
tive flow unit-stepresponsefunctions.They were derived using varying flow rate is obtained by letting x = r in (17), and
Laplace transform techniquesand Mellin's inversion integral evaluatingthe boundarycondition at r - 1. This gives
with integration in the complex plane [Churchill, 1958, p.
176]. For our needs, the flow rate response to a unit-step
drawdown, •,,, is obtained by solving (10) (subject to con-
' h(1,
t)= Q(O
OH.(t
- •:)dz (19)
dition equations (11), (12a), and (13a)), then taking the radial
The convolution integral of (19) will be used in the devel-
derivativeand evaluatingit at the inner boundary
opment of an approximate calculation technique for the
convolution integral of (18).

Q.(t) Oh
I 4fo/'•[Jo2(/'•)
=•rrr=• - •2 e-X2tdJ'
+ Yo2(;0] (14) In principle, (18) could be used at the AIF boundary to
calculate the flow rate into the simulation region from the
outer region of the aquifer. However this is not practical when
The potentiometric-headresponseto a unit-step withdrawal h(t) is not known a priori, as in the caseof numerical simula-
flow rate, H., for an infinite region, is obtained by solving (10) tion of groundwaterflow in the inner aquifer region.The inte-
(subject to condition equations (11), (12b), and (13a)), and gral must be recomputed repeatedly from the initial time to
evaluatingthe solutionat the inner boundary, the current time level, since time is a parameter in the inte-
grand as well as being the upper limit of the integration vari-
H.(t) t)= •-4fo
= h(1, ©;taEj•2(;
(1--e
0-x2')
+ y•dit
2(•)3 (15) able. Carter and Tracy [1960], using an approach given by
Hurst [1958], developedan approximate algorithm to avoid
where Ji and Y/ are Besselfunctions of the first and second the repeatedcomputationof (18) as the simulationprogresses.
kind of order i, respectively.Equation (14) was presented by
Jacob and Lohman [1952] in a different form as a solution to The Carter-Tracy Approximation
the constant-drawdownproblem for flow to a well (Boundary- The discretization of time means that the flow rate defined
condition equation (3a)). These two aquifer-influencefunctions by the aquifer-influencefunctionboundary condition also will
will be referred to as the flow-rate response,Q., to a unit-step be discretizedin some fashion. Let this flow rate be approxi-
drawdown, and the potentiometric-head response,H., to a mated by a sequenceof flow rates Qi, that are constant over
unit-step withdrawal flow rate respectively. Note from (15) each time step interval ti to t•+ •, as is shown in Figure 2. The
that H, is always lessthan or equal to zero. Van Everdingen link between the continuous flow rate and the discontinuous
426 KiPP: TECHNICAL
NO•

ary flow rate will disappear.To avoid this, one would have to
reset the time origin for the calculation of Qi to the time at
which the flow rate changedsign.
Thus the Carter-Tracy approximation is most applicable
when the amplitude of the flow rate variation is small relative
to the average value. Of course the actual flow rate is not
ce Qi• known a priori and only the approximate flow rate is com-
_.e Qi-1
o c Q3 puted during a simulation. However one can obtain an indica-
'• Q2
tion from the variation of the boundary pressureas to whether
ß Q1
.E the Carter-Tracy approximation can be expectedto perform
poorly. The inaccurate calculation of the aquifer-influence
tO t I t2 t3 ti-1 ti ti+l tn tn+l function for time-varyingflow rates is the major shortcoming
Dimensionless time, t of the Carter-Tracy approximation. For application to
Fig. 2. Continuousflow rate function,Q(t), and its approximating petroleum-reservoirsimulation, water influx from the outer
discontinuous flow rate function,Qi, chosenso that the cross-hatched region during petroleum production is usually being con-
areasare equal.Other discontinuous flow ratesare shown. sideredso Q doesnot changesignduring the calculationsand
is often approximatelyconstant.
flow rate (Figure 2) is made by requiring that the cumulative Taking the Laplace transform of (22) using the property of
net inflow over the entire period from the beginningof the the transformof an integral [Churchill, 1958,p. 40], we obtain
simulation to the current time level be the same. The cumula-
tive net inflow is given by --h(s)
Q.(s)
= W.
S
+Q.
S2
Q.t.
S
(25)
This equation can be solved for the transform of the head
w(t)= = - he) function h.
for the continuous function and by In order to derive the Carter-Tracy algorithm for aquifer-
influence function calculation, we need the relation between
W.+
a= • Qi(t•+
a- t,) (21a)the flow-rate responseto a unit-stepdrawdown and the head
i=0
responseto a unit-stepwithdrawal flow rate. This relation is
basedon the fact that a given transientdrawdown producesa
= • Q,(ti+•- t,) + Q.(t- t.) (2lb) flow-rate responsedefinedby (18), while a given transientflow
i=0
rate produces a drawdown response defined by (19). The
for the discontinuousfunctionwhere t•+ • = t. Settingthe cu- mathematical relation is clearer in Laplace transform space.
mulative net inflows equal yields We take the Laplace transforms of (18) and (19) and use the
property that the Laplace transform of a convolution integral
is the product of the Laplace transforms of the two functions
fo•fo
•- h(z)•Q"(2-Z)
dzd2=W•+Q•(t-t•)(22) being convoluted [Churchill, 1958, p. 37]:

•= • Q,•t•+x-t•) (23)
i=0

where• is the cumulativenet inflow at time t,.


L f(z)g(t-
)
z) dz = f(s) a(s) (26)

Another approachis to usethe intermediate-value theorem


from integralcalculus[Thomas,1960,p. 192],whichsaysthat where f(s) and g(s) are the Laplace transformsof f and g.
Using the property that the transform of the derivative of a
if a functionQ is continuousand positivevaluedover domain
function is [Churchill, 1958, p. 8]
(a, b) and W• is the area under the graph of Q over this
domain, then there is at least one number, ½,betweena and b L(f (t)) = sf(s) - f (O) (27)
such that
we obtain
= Q(½)(b- a) (24)
Q(s)= -h(s) sQ.(s) (28)
The flow rate, Q, at the beginningof eachinterval in time, is
and
selected so that (23) is satisfied. The restriction that Q be
positivemeansonly that Q cannotchangesignover the inter- h(s)= Q(s)sH.(s) (29)
val, to to t, The graph of the continuousflow rate and the
discontinuousapproximateflow rate shownin Figure 2 illus- sinceh(0)= 0 and (2(0)= 0. For the same flow-rate function,
trates the selectionof Q• so that the cross-hatchedareas are Q(t), (28) and (29) can be combined to give the desiredrelation
equal. This is equivalentto making the areasunder the con- between the two unit-step responsefunctions'
tinuous and discontinuouscurves equal for each cumulative --1 = s2H.(s)Q.(s) (30)
time range, to to t•. This method of determiningthe flow rate
is exactwhenQ is a constant.The effectof a varyingflow rate It will be easierto invert the transform function if (30) is used
appearsin Figure 2 as a lag of the Q• in approximatingQ(t), to convert from the flow-rate unit-responsefunction to the
and a smoothingof the variationin Q(t). For a flow rate with potentiometrichead unit-response function.Thus
periodicvariation,the approximateflow rate approachesthe
h(s)= sH.(s)(W•-- Q.t.) q- H.(s)Q. (31)
averageflow rate thus loosingthe variationalbehaviorof the
actual flow rate. If the actual flow rate has an averagevalue of Inverting the Laplace transform gives the equation relating
zero, the approximateflow rate will becomezero after a suf- the potentiometrichead to the approximate flow rate at the
ficiently long time and the aquifer-influence-function bound- AIF boundary as a function of time
KIPP: TECHNICALNOTE 427

dHu(t) grid for the inner region would have an exact cylindrical
h(t)= dt (Wn
-- Qntn)
+ Hu(t)Qn (32) boundary.For a cartesian-coordinate systemwith the x-y axes
horizontal, the equivalent radius for the approximate inter-
which can be solved for the desired AIF boundary flow rate facialboundary
is calculated
sothattherectangular
area,
Qn: and the equivalent circular area are the same; i.e.,

li:rl '2 = Ax:


v (35)
h(tn+
•)- •- Wn For boundariesbetweenthe inner and outer aquifer regions
Qn= tn+x (33) that do not completelysurround the inner aquifer laterally,
the apportionment factor 0•m,must contain an angle of influ-
ence factor, fo. This factor is the fraction of a full circle sub-
Equation (33), for the flow rate acrossthe AIF boundary is a tended by the boundary betweenthe inner and outer aquifer
function of quantitiesat time levelst• and t•+ • and represents regions.For example,an outer-aquiferregion that surrounds
the Carter-Tracy [Carter and Tracy, 1960] method for water- half of the inner simulationregion (Figure lb) would have an
influx calculation adapted to groundwater flow simulation. angleof influencefactorof «.
Note that it involvesonly the potentiometric-headresponseto The final equationfor the aquifer-influence-function bound-
a unit-step withdrawal flow rate and not the flow-rate re- ary condition flow rates is obtained by discretizing(33) in
sponseto a unit-stepdrawdown. space.The temporalsubscriptswill now be made superscripts
NUMERICAL IMPLEMENTATION
and the spatial location will be denotedby subscripts.For
boundary node or cell, m, we have the dimensionalflow rate,
The results to this point are for a uniform flow rate across Qm'n,at time tn,in preparation for calculationof conditionsat
the AIF approximate cylindrical boundary between the inner time, tn+ 1.
simulation region and the outer aquifer region. To apply the
transient-flow-aquifer influence function calculation to the
boundary of grid-cell faces that arise when a finite-difference
discretization is used for the inner region, the aquifer Qm'n = O•mA
m -•-In+
1Wren
-Jr-
(hm
n+•--hm
n)
influence-functionflow rate must be suitably apportioned
among the boundary-cell faces. The common method is to Hun
+1 dHu
• In+ltn
make the fractionof the total flow rate that is apportionedto
a givenboundarycell the sameas the ratio of that boundary- ß2•Kb(ho'-h,') (36a)
cell facial area to the total boundary-facialarea betweenthe
inner and outer aquifer regions. For cases where the inner
dHu
--hø')mn---•-I Wm
tn
n+l2r•r•
'2S+ •hm
h' dt
In+
aquifer region is strongly heterogeneous, apportionmentby Qm,n = O•mAm211;Kb
the hydraulicconductivity-facialarea productshouldbe more
realistic.The apportionmentfactor is then definedfor cell m as
Hun
+1 dHu tn

•m -- •b (34) (36b)
• KiAi where
i=1

whereAmis the boundaryfacialareaof cell m (L2),K mis the 6hm,__hm,n+


1 _ hm,n; (37)
hydraulicconductivityat the boundaryface of cell re(L/T), hm'n potentiometrichead in boundarycell rn at time, P, (L);
and % is the numberof boundarycellsbetweenthe inner and Hun value of the aquifer-influence
functionH u,at time,
outer aquifer regions. tn, (--);
Equation (34) is an empirical approximation for a global K homogeneoushydraulic conductivity of the outer
parameter in terms of local values.It is probably most useful region, (L/T);
in providing a good initial estimate for the allocation factor b uniform thicknessof the outer region, (L);
which can then be adjustedduring the model calibration. S storagecoefficientof the outer region,(-);
The derivationof the flow rate givenby (33) wasalsobased W' cumulativenet inflow,(L3).
upon a uniform potentiometrichead, h, over the cylindrical
approximating, AIF interface. A flow simulation in the dis- Note that the storage requirements for this calculation are
cretized inner region will yield a nonuniform distribution of only the cumulative flow, W', the potentiometric head at the
headsat the boundary nodes,exceptin specialcases.In the end of the nth time step,h', and the flow-rate allocation factor,
numerical implementationof this aquifer-influence-function •, for eachnodeon the AIF boundary.Valuesof hmn'are from
calculation,the head at each interfacialboundary node is the flow simulation,so the additionalstorageamountsto only
taken to be the value computedby the discretizedcalculation. two timesthe numberof aquiferinfluenceboundarynodes.
This introducesan additionalapproximationbecauseany lat- Equation (36b) is of the form
eral or verticalflow in the outer aquiferregioninducedby a Qm'n-- a•(t') + a2(t') •h m' (38)
nonuniform potentiometric-head distribution over the inter-
facialboundaryis neglected. wherea• is the known flow rate term addedto the right-hand
The third approximation,mentioned at the outset,is that sideof the discretizedflow equationfor nodem, and a2 is the
the boundarybetweenthe inner and outer aquiferre•ons is implicit term added to the left-hand side factor.
approximatedby a cylindricalinterface(Figure la), whereas The values of Hu(t) and dHu/dt are usually obtained from
the actual boundary is a set of rectangularfacesfor a finite- tables by interpolation. Values of the dimensionless
difference grid in cartesian coordinates of a three-dimensional potentiometric-headfunction in responseto a unit-withdrawal
inner-aquiferregion.Naturally, a two-dimensional,cylindrical flow rate at the AIF boundaryhave been tabulatedby Van
428 KiPP: TECHNICAL
NOTE

TABLE 1. Coefficients for the Analytical Approximations to the ing of transients.Errors can be seriousfor flow rates whose
Van Everdingenand Hurst AquiferInfluenceFufictions variationsare large relative to the averagevalue. In particular,
if the average flow rate is zero, an adjustment to the time
R bo bx b2 b3
origin for the aquifer influencefunction computationshould
oo -0.82092 3.68 x 10 -4 -0.28908 -0.02882 be made each time the flow rate changessign.
1.5 -0.10371 - 1.66657 0.04579 0.01023 In general,a variety of outer-aquiferregion geometriesand
2.0 -0.30210 -0.68178 0.01599 0.01356 flow rate versuspotentiometrichead relationshipscan be em-
3.0 -0.51243 -0.29317 -0.01534 0.06732
4.0 -0.63656 -0.16101 -0.15812 0.09104
ployed.The radial, transient-flowcasepresentedin thisnote is
5.0 -0.65106 -0.10414 -0.30953 0.11258 representative. The influencefunctionapproachto boundary
6.0 -0.63367 -0.06940 -0.41750 0.11137 flow simulationcan be regardedas a predecessorto the devel-
8.0 -0.40132 -0.04104 -0.69592 0.14350 opment of mixed methodssuch as combiningboundary ele-
10.0 -0.14386 -0.02649 -0.89646 0.15502
ments with finite-differencetechniques.
Adapted from Fanchi [1985].
REFERENCES

Everdin•ten and Hurst [1949] for the infinitecylindricalregion Aziz, K., and A. Settari, PetroleumReservoirSimulation,Applied Sci-
ence Publishers, London, 1979.
and for regionswith a finiteouter-boundaryradius. Bear,J., D. Zaslavsky,and S. Irmay, Arid zone research,in Physical
However, it is much more convenient to use the approxi- Principlesof Water percolationand seepage,vol. 29, UNESCO,
mate analytical representationdevelopedby Fanchi [1985]. Paris, 1968.
He employedlinearregression analysisto obtainthe following Carter, R. D., and G. W. Tracy, An improvedmethod for calculating
water influx, Trans. Soc.Pet. œng.AIMœ, 219, 415-417, 1960.
equation that approximatesthe Van Everdin•tenand Hurst
Churchill,R. V., OperationalMathematics,McGraw-Hill, New York,
[1949] aquifer influencefunctionswith very small errors.In 1958.
the notation of the presentwork, Craft, B.C., and M. F. Hawkins, Applied PetroleumReservoirœngi-
neering,Prentice-Hall, EnglewoodCliffs, N.J., 1959.
Hu(t)= bo+ bat+ b2In (t) + b3 In2 (t) (39) Fanehi, J. R., Analytical representationof the Van Everdingen-Hurst
aquifer influencefunctionsfor reservoir simulation, Soc. Pet. Eng.
Table 1, adaptedfrom Fanchi[1985], containsvaluesof the bi d., 25(3), 405-406, 1985.
coefficientsfor severalcases.The first line givesthe coefficients Hall, F. R., and A. F. Moench, Application of the convolution equa-
for the analytical approximation to (15) for the case of an tion to stream-aquiferrelationships,Water Resour.Res.,8(2), 487-
493, 1972.
infinite outer aquifer region. The subsequentlines are for Hurst, W., The simplificationof the material balanceformulasby the
various values of R for the case of a finite outer aquifer region Laplacetransformationto flow problemsin reservoirs,Trans.Soc.
with no flow at the exterior boundary. R is the ratio of ex- Pet. Eng. AIME, 213, 292-303, 1958.
terior to interior radius for the outer aquifer region. INTERCOMP Resource Development and Engineering, Inc., A
model for calculatingeffectsof liquid wastedisposalin deepsaline
CONCLUSION aquifers,I, Development,and II, Documentation,U.S. Geol. Surv.
Water Resour. Invest., 76-61, 1-253, 1976.
The Carter-Tracy calculationfor AIF boundary flow rates, Jacob, C. E., and S. W. Lohman, Non-steady flow to a well of con-
using the Van Everdingen and Hurst aquifer-influence- stant drawdown in an extensiveaquifer, Eos Trans. AGU, 33, 559-
functions for radial flow from an outer aquifer region, has 569, 1952.
been derived in the terminology of groundwater hydrology Maddock III, T., Albegraictechnological
functionfrom a simulation
model, Water Resour.Res.,8(1), 129-134, 1972.
with additional clarifyingexplanation.It can be seenthat this Morel-Seytoux,H. J., A combinedmodel of water table and river
technique offers a simple but approximate method for em- stageevolution,Water Resour.Res.,1•(6), 968-972, 1975.
bedding a region of groundwater flow simulation within a Morel-Seytoux,H. J., and C. J. Daly, A discretekernelgeneratorfor
muchlarger aquiferregion.The larger regioncan be treatedin stream-aquiferstudies,Water Resour.Res.,11(2),253-260,1975.
Thomas,G. B., Jr., Calculusand Analytic Geometry,3rd ed., Addison-
a more approximatefashionbecauseof its remotenessfrom
Wesley,Reading,Mass., 1960.
the primary region of interest. Tychonov,A. N., and A. A. Samarski,Partial DifferentialEquationsof
The useof aquifer-influence-functions reducesthe sizeof the MathematicalPhysics,vol. 1, translatedby S. Radding, Holden-
computationalgrid, with a correspondingreductionin com- Day, San Francisco,Calif., 1964.
Van Everdingen,A. F., and W. Hurst, The applicationof the Laplace
puter storageand executiontime. The Carter-Tracy[Carter transformation to flow problems in reservoirs, Trans. $oc. Pet. Eng.
and Tracy, 1960] approximationto the convolutionintegral AIME, 186, 305-324, 1949.
enablesthe aquifer-influence-function
calculationto be made
with modest additional storage over that required by the K. L. Kipp, Jr., Water ResourcesDivision, U.S. GeologicalSurvey,
Box 25046, MS 413, Denver Federal Center, Denver, CO 80225.
inner-region simulation. The major disadvantage of the
Carter-Tracy approximation is the inaccuracyof the compu- (ReceivedMay 31, 1985;
tation of the discretized flow rate for time-varying boundary revised October 28, 1985;
flow rateswhich appearsas a significanttime lag and smooth- acceptedNovember 12, 1985.)

Вам также может понравиться