Prévia do material em texto
WATER RESOURCES RESEARCH, VOL. 27, NO. 6, PAGES 1035-1040, JUNE 1991
Hillslope Infiltration' Divergent and Convergent Slopes
J. R. PHILIP
CSIRO Centre for Environmental Mechanics, Canberra, Australia
This perturbation analysis of divergence/convergence effects on hillslope infiltration and downslope
unsaturated flow indicates that, in general, these effects are unimportant for usual hillslope topogra-
phy. Only where the radius of the contour is less than about 10 times the characteristic infiltration
length/gray (or 10 times the sorptive length) need the perturbation be taken into account. This result
opens up a considerable extension of applicability of the analysis of infiltration on planar hillslopes to
embrace slopes with curved contours, such as ridge noses and hollows.
1. INTRODUCTION
Hillslope infiltration and downslope subsurface unsatur-
ated flow for planar slopes has been studied recently [Philip,
1991]. Here we address the question of how results of that
work are affected by slope divergence or convergence.
Figure 1 indicates schematically what we mean by these
terms. Figure 1 a shows a planar slope, with equally spaced
linear contours; Figure 1 b shows a divergent slope, with the
radius of curvature of the concentric contours increasing
downslope; and Figure l c shows a convergent slope, with
the radius of curvature of the concentric contours decreasing
downslope. Both the divergent and convergent slopes are (at
least part of) conical surfaces with the cone axis vertical.
!ida [1984] studied saturated flow in such conical slopes. He
referred to divergent slopes as "ridge-type" and to conver-
gent ones as "hollow-type". Previously, Beven [1977], using
an essentially two-dimensional model, parametrized diver-
gence/convergence through a "slope width". It will be
understood that our divergent slopes are relevant to the
slopes of isolated hills and to the horizontally convex slopes
at ridge noses, and our convergent slopes model those of
isolated closed depressions and the horizontally concave
hollows at valley heads.
2. PROBLEM FORMULATION:
SLOPE DIVERGENCE PERTURBATION
We consider a long divergent hillslope of homogeneous
isotropic soil with uniform slope angle 3'. We use cylindrical
coordinates (r, z) with the origin of the horizontal radial
coordinate r on the axis of the conical surface which
contains the hillslope. Note that r is the radius of curvature
of the slope contours. The vertical coordinate z is taken
positive downward. In the radial half-plane (r, z) we also use
the rotated coordinates (r,, z.) defined by
r.=r cos •/+z sin 3•
(1)
z, = -r sin •/+ z cos y
(see Figure 2a).
The relevant axisymmetric form of the equation governing
unsaturated soil water flow [cf. Philip, 1969, equation (25)] is
Published in 1991 by the American Geophysical Union.
Paper number 91WR00098.
1035
•=-- D + D -• , (2) at Or •zz r Or dO Oz
where t is time, 0 is volumetric moisture content, D is
moisture diffusivity, and K is hydraulic conductivity. In
general, D and K are strongly varying functions of 0. We
seek the solution of (2) subject to the conditions
t=0 z.>0 0=00
(3)
t>0 Z.=0 0=01 .
Here 00 is the uniform initial moisture content, and 0• is the
0 value corresponding to moisture potential • at which
water is available at the soil surface z, = 0. With free water
present in excess at negligible depth on the hillslope, •1 = 0,
and 0• is the saturated moisture content.
The corresponding equation for the planar hillslope
[Philip, 1991, equation (2)], with x replaced by r, may be
written as
(4)
We see that (4) is identical to (2) except for the term
Dr-•OO/Or on the right of (2). This term represents the
primary effect of slope divergence. For all t this slope
divergence term is at most of order r-• relative to (at least
the largest of) the other terms on the fight of (2). Divergence
thus gives a perturbation of the solution for planar slopes of
order r -• . Evidently, the appropriate perturbation solution
will be usefully convergent for r large enough, that is, far
enough from the cone axis.
Philip [1991] simplified the analysis for long planar slopes
by seeking the solution which depends only on t and the
rotated coordinate z,. Equation (4) was in this way reduced
to the simpler [Philip, 1991, equation (5)]
• = cos y. (5)
The solution of (5) subject to (3) was
Z,(O, t) m E qontn/2( cOS ,y)n-l,
n=l
(6)
1036 PHILIP: HILLSLOPE INFILTRATION' DIVERGENT AND CONVERGENT SLOPES
a Planar slope
Cone
axis
b Divergent slope
!.
ß .:.;.:,::. -.::;:.:.::'-"
c Convergent slope
Fig. 1. Schematic figures showing (a) a planar slope, •b) a
divergent slope, and (c) a convergent slope. The arrows indicate the
direction of maximum slope. In Figures lb and lc the vertical axis
of the conical surface is shown. In both cases the contours are
centered on this axis.
with the coefficients •p,(`0) precisely the functions of 0
entering the series solution for infiltration into a horizontal
soil surface [Philip, 1957a, b; 1969].
In the present work we develop the perturbation analysis
by injecting into (5) the divergence term Dr -] O0/Or in an
appropriate approximate form. With (6) taken as the leading
term of the required solution, we find that, to order r -I ,
D 00 D 00
_ m = sin y. (7)
r Or r o Oz,
Here r 0 is the value of r at the surface point (r,, z,) = (r0
sec y, 0). Using the right side of (7), we thus obtain the
perturbed form of (5)'
.... D - sin y+•cos y az, (8) •t Oz, '
This equation, accurate to order r• • , enables us to evaluate
the leading term of the divergence perturbation.
We therefore seek the solution of (8) subject to (3) of the
form
z,(0, r 0, t)=f0(0, t)+ r•f•(O, t). (9)
The expansion may be taken to higher powers of rg l, but
this is not required for present pu•oses. It is convenient to
write f0 for the z,(0, t) of (6), which is z,(0, m, t) in the
extended notation of (9).
Putting (9) into (8) and equating coefficients of rg • gives
the required equation in f], namely,
•=• D +D sin (10)
3t 80 30] Y'
This is subject to the conditions
a Divergent slope
Cone
axis
Cone
axis b Convergent slope
Fig. 2. Schematic figures illustrating space coordinates (r, z)
and rotated coordinates (`r,, z,)' {a) divergent slopes, and {b)
convergent slopes.
o, f , dO = D 0-•o 0 o 0=0 !
dt
+ D dO t sin y
0
t>-0 fl=0 0= 0•. (11)
The first condition in (11) preserves the material balance to
order r• -]. The second follows from the identity of the
second condition in (3) (this paper) with the second condition
in (3) of Philip [1991]. Integrating (10) with respect to both 0
and t, and using (11), we obtain the numerically convenient
integrodifferential form
fl dO- D •dt + D dO t sin y,
O0
o o
(,12)
PHILIP: HILLSLOPE INFILTRATION: DIVERGENT AND CONVERGENT SLOPES 1037
subject to the condition
t :> 0 f• = 0 0 = 01. (13)
3. DIVERGENT SLOPES: SERIES SOLUTION
Putting (6) into (12) and (13), we see by equating coeffi-
cients of powers of t 1/2 that f• is of the form
f •(O, t) = Z •n(O )tn/2(cos 7) n- 2 sin T,
n=2
(14)
CONVERGENT SLOPES
The analyses for convergent hillslopes go similarly. In this
case we consider a long convergent hillslope of homoge-
neous isotropic soil with uniform slope angie y (see Figure
2b). We take cylindrical space coordinates (r, z) as before,
and again we use the rotated coordinates (r., z.) in the
radial half-plane. Here, however, the sense of rotation is
reversed, and we therefore replace 7 in equation (1) by -7-
In what follows we shall not usually distinguish between
the divergent and convergent cases; it will be understood
that 7 > 0 for slope divergence and 7t • in (12) gives the equation for se2:
/o • d • 2 f o • + DdO •2 dO = P dO '
o o
(15)
with P the function of 0 defined by
P = D(dO/dqo •) 2. (16)
Equation (15) is subject to the condition
0 = 0• •2 = 0. (17)
Similarly, equating coefficients of t 3/2 gives the equation for
•3. It is
• •:3 dO = P d•3/dO - Q,
o
(18)
with Q the function of 0 defined by
00 0½2 0•2
Q = 2D .... . (19)
0q0 1 Oc.P 10qO 1
Equation (18) is subject to the condition
0 = 0• •3 = 0. (20)
Repeating this procedure gives the integrodifferential
equations for •4, •5, etc. The method of solution and the
form of the resulting equations closely follow those of the
infiltration series solutions [Philip, 1957a, b; 1969]. The
simple and rapid numerical technique for solving the se-
quence of linear integrodifferential equations [Philip, 1957a]
applies here also. Note that (15) is similar to equation (10) of
Philip [1966] for m = 2. It tums out that (in present notation)
-•2(0) was evaluated there in the context of radial two-
dimensional absorption.
Finally, we combine (6), (9), and (14) to secure the
required solution to order r• -• , namely,
) •2 tan 7 z,(O, r0, t) = qOl tl/2 + qøn + •n
_ r0
ß (COS 7) n- ltnl2. (21)
5. FURTHER CALCULATIONS
5.1. Flow Velocity Components and
Integrated Flows
Philip [1991] used two different means of resolving soil
water flow velocity: First, resolution into horizontal and
vertical components, u and v, with u positive directed into
the hillslope and second, resolution into components parallel
and normal to the slope, u d and V n, with u d positive
downslope. We employ both sets of components here also.
See section 3 and Figure 2 of Philip [1991]. We note, in
particular, that in the present symbols
u = D 0 O/Or, (22)
ud = K sin 7, (23)
and, in the infiltration rate norma/to the slope,
v,•0 = K• cos 7- (D O0/Oz.)z.=o. (24)
Philip [1991] also introduced the integrated horizontal flow
U, defined in the present symbols by
U = u dz, (25)
sin y
and the integrated downslope flow U a'
Ud = (u d--Ko sin 7) dz. = (K-Ko) dz. sin 7
(26)
We study these integrated flows here also.
5.2. Normal Infiltration
The computations rapidly grow complicated, but it suffices
to treat only the leading term of the perturbation, calculable
from •2 for the series solution. We denote the result for
planar slopes by Vno(O•, t) (i.e., r o -, oo) [Philip, 1991]. We
find that the series solution yields
Vno(ro, t)- Vno(O•, t)- D dO- •2 , (27)
ro
o
where
dO.
1038 PHILIP: HILLSLOPE INFILTRATION: DIVERGENT AND CONVERGENT SLOPES
Note that asymptotic large t analysis is more complicated
here than for planar slopes [Philip, 1991] and is not pursued.
However,
lim Vno(ro, t)= Kt cos y, (28)
so that
lim [vno(r o, t) - Vno(O•, t)] = 0. (29)
Divergence decreases v•0 and convergence increases it,
though, as we shall see, the effects are often trivial in the
applications.
5.3. Horizontal Flow
K1 -- Ko K• - Ko
a = = ß (34)
q'• K d• D dO
o o
•0 is the moisture potential corresponding to moisture
content 00. For a given soil situation the two lengths are
connected by the relation
2a-l= 2b Igrav, , (35)
where, as shown by White and Sully [1987], the constant b
ranges from 0.5 to 0.6 under many conditions of practical
interest. We develop the argument here in terms of l•av.
The magnitude of the perturbations may be inferred from
the magnitude of the leading term of the series expansion of
f•, and specifically from the magnitudes of •2, E2, and
together with that of f 0ø0 • D d 0.
It may be readily shown that, to order r• 'l, the total
horizontal flow
6.1. Magnitude Estimates and Computed Values
for the Example
• O0 •= O0 U= D • dz = - D • dz tan ¾
Or sin • sin 7
= D dO tan 7-
o
(3O)
To this order, the total horizontal flow U is unaffected by
divergence or convergence. It is constant and independent of
t, as for planar slopes [Philip, 1991].
5.4. Integrated Downslope Flow
Here also we limit ourselves to the leading term of the
perturbation. U d(OO, t) denotes the result for planar slopes
[Philip, 199!]. Then the series solution gives
Ud(ro, t) -- Ud(o:, t) = r•-lA 2 t sin 2 y, (31)
with
The integrodifferential equation for •2, (!5), and the con-
dition governing it, (equation (17)), are identical to those for
qo2 [Philip, 1957a, equations (3.17) and (3.18); Philip,
1957b, equations (33) and (34); Philip, 1969, equations (88)
and (92)] except that the term K - K0 is replaced by f 00 D
dO. Since (15) is linear, it follows that
q>2(0-----• = 0 D dO/(K 1 - Ko) (36)
0
[Io •2= 0•
cb: 0 D dO/(K• - Ko) . (37)
• 0
0• We recall that •: = f Oo •o: d 0. Now it follows from (34) and
(35) that
S' D dO/(K• - Ko) = b /•rav, o
so (36) and (37) become
(38)
O= 0o A 2 = (K(O) - Ko) ds•2 . (32)
dO=Or
Note that divergence increases the integrated downslope
flow, though it decreases the normal infiltration rate. Con-
vergence gives reverse effects.
6. THE MAGNITUDE OF THE PERTURBATIONS
On a priori physical grounds one expects the perturbation
due to divergence/convergence to be trivially small provided
r is large compared with characteristic lengths of the flow
process. One such length is/•rav, defined [Philip, 1991] by
S 2
/gray = (K1 ' ro)(Ol - 00)' (33)
Another is the sorptive length 2a -• , where the sorptive
number a is defined [Philip, !983] by
•2( 0 )
qo2(0'-• = O[b /srav], (39)
• = O[b Igrav]- (40)
(I) 2
We may check these relations for the illustrative example
of Philip [1991] for Yo!o light clay with 00 = 0.2376, 0• =
0.4950. For that soil, b = 0.557 and l•rav = 0.497 m, so (40)
becomes •2/(I) 2 = 0[0.277 m]. Calculation gives ,E2 =
!.3808 x 10 -8 m 2 s -• . Now, since (I)2 = 4.654 x 10 -8 m s -•
[Philip, 1957b], we find •2/(I) 2 = 0.297 m, agreeing well
with (40).
Satisfaction of (40) implies satisfaction of (39), so long as
the •:2 (0) and q•2 (0) functions are reasonably similar in shape.
This condition is tested in Figure 3, which compares se2(0)
and b/•rav ½2(0) for the Yolo light clay example.
We need also the magnitude of A2/t{ 2, where K2 is the
second coefficient in the series expansion for Uo(oo, t)
[Philip, 199!]. Comparing the relation
PHILIP: HILLSLOPE iNFILTRATION' DIVERGENT AND CONVERGENT SLOPES 1039
01 0 o
o
i -
2 -
4
I I I I I I ...... I ,,, I I ,, I
Fig. 3. Comparison of the divergence/convergence perturbation
function •2(0) (solid curve) with bigray •2(0) (dashed broken curve),
both computed for Yolo light clay. The order of magnitude agree-
ment is consistent with (39).
"O= 0o 12(A - K0 in the symbolism
of that paper) is given for Ko/Ki small by
•2 = c(K1 - Ko). (43)
The constant c ranges from 0.3 to 0.4 for common soil
situations. Combining (38) and (43) then gives
f o•' D dO/•2 = b lgrav/c. o
Then (44) and (40) yield
(44)
• D dO - •2 o
(I) 2
= O[b(c-i _ 1)/gray], (45)
an order statement requiring that (40) be sufficiently accu-
rate. The cited ranges of b and c imply that the factor
b(c -1 - 1) lies in the range 0.75 to 1.4. For the Yolo light
clay example, c = 0.378, so (45) gives (f
= 0[0.456 m]. We have, further, that the calculations for
Yolo light clay [Philip, 1991] give f •o • D dO = 3.404 x 10 -8
ß ot D dO- •2)/(I)2 is thus m 2 s -1 The actual value of (œOo •
0.435 m, agreeing well with (45).
6.2. Normal InfiltrationRate
We see from (27) that the primary effect of the perturba-
tion is to change the second term of the series expansion for
Vn0- The perturbation is established for small and moderate
values of t, but (29) indicates that it decreases to zero as t --->
o•. It is apparently of greatest significance at moderate values
of t.
The perturbation is evidently negligible for all t when r0 is
sufficiently large. We take as our criterion for this that the
perturbation should not exceed 10% of the t o term of the
unperturbed expansion for V no [Philip, 1991, equation (27)].
This requires that
D dO-E2 tan 7
0
r0 > 10 . (46)
•2
Combining this with (45) gives
r0 > 100[b(c -1- 1)/•rav] tan 3'- (47)
Adopting the upper bound on b(c -1 - 1), 1.4, and remov-
ing the order sign gives the suggested criterion
r o > 14 Igrav tan 7, (48)
which reduces, for 3' = 30 ø, to
r 0 > 8.08 lgrav- (49)
Detailed calculations for the Yolo light clay example with 3'
= 30 ø give
r 0 > 5.05 lgrav = 2.51 m, (50)
consistent with (48).
6.3. Integrated Downslope Flow
Here also the primary effect of the perturbation is to
change the second term of the series expansion for U a. It is
evidently negligible for all t when r 0 is large enough. We
adopt the criterion that the perturbation should not exceed
10% of the t 1 term of the unperturbed expansion for Ua
[Philip, 1991, equation (45)]. This requires that
3,2
r0 -> 10 • tan 3'. (51)
I 10 O[b /grav] tan 3'- (52)
Adopting the upper bound on b, 0.6, and removing the order
sign gives
r 0 -> 6/gray tan 3', (53)
a criterion less stringent than (48). The detailed calculation
for Yo!o light clay gives
1040 PHILIP: HILLSLOPE INFILTRATION: DIVERGENT AND CONVERGENT SLOPES
r 0 -> 5.5 /grav tan y = 2.75 tan •, m,
in agreement with (53).
(54)
6.4. Discussion
The foregoing results give the r0 ranges for which diver-
gence/convergence perturbations may be neglected. We see
that, provided r0 exceeds these various bounds, the analysis
for planar slopes [Philip, 1991] is applicable. Since/gray tends
to lie in the range 0.1 to 2 m [cf. White and Sully, 1987], the
bounds we have established imply that typically the pertur-
bation is negligible so long as r0, the radius of the contour,
exceeds a few meters.
These bounds, however, must be taken in conjunction
with the condition for assured accuracy of the planar analy-
sis on which this perturbation study is based. This [Philip,
1991 ] is that
X. > 40(K• - Ko)t sin ?/(0• - 00), (55)
with X. the downslope distance from the slope crest (or a
point of slope change). This quantity translates into 40(K• -
Ko)t cos ), sin ?/(0• - 00) as a horizontal interval in r. It
tends to be of the same order of magnitude as the bounds
examined above.
For divergent slopes the bounds (48) and (55) both apply at
the upper end of the slope, with the larger overriding, and
there is no bound at the lower end. For convergent slopes,
on the other hand, (55) fixes the bound at the upper end, and
(48) does so at the lower end.
7. CONCLUDING DISCUSSION
7.1. The Character of the Perturbations
Although the perturbations are small, it is useful to sum-
marize their signs and their physical basis. For divergent
slopes the depth of moisture penetration is increased
slightly. This is accompanied by a slight reduction in the
surface moisture gradient and hence also in the infiltration
rate Vno. On the other hand, the slightly increased normal
depth of penetration gives a small increase in the cross
section for downslope flow and therefore also in the total
downslope flow Ud. The signs of these various modifications
are reversed for convergent slopes.
7.2. Extended Applicability of Planar Analysis
This perturbation analysis of the effects of hillslope diver-
gence/convergence has indicated that, in general, they are
unimportant for usual hillslope topography. Only where the
radius of the contour is less than about 10 times the charac-
teristic infiltration length lgrav (or 10 times the sorptive
length) need the perturbation be taken into account. This
result opens up a considerable extension of applicability of
the analysis of infiltration on planar hillslopes [Philip, 1991]
beyond purely planar slopes to slopes with curved contours
(e.g., on ridge noses and in hollows).
7.3. Physical Remark
In physical terms the explanation of this result is clear. It
is that divergence/convergence adjustments to the flow oc-
cur on the scale of hillslope topography (the contour radius
r), whereas the unsaturated flow processes driven by capil-
larity and gravity can respond on the generally much smaller
scale of/gray (or the sorptive length). In consequence, the
divergence/convergence perturbations to the planar slope
results tend to be negligibly small.
Acknowledgtnents. I am grateful to Ian White for helpful discus-
sions and to Cathy Wilson for drawing my attention to Beven (1977)
and Iida (1984). I also thank the Australian Water Research Advi-
sory Council for the Eminent Researcher Fellowship which sup-
ported this work.
REFERENCES
Beven, K., Hillslope hydrographs by the finite element method,
Earth Surf. Processes, 2, 13-27, 1977.
Iida, T., A hydrological method of estimation of the topographic
effect on the saturated throughflow, Trans. Jpn. Geomorphol.
Union, 5, 1-12, 1984.
Philip, J. R., Numerical solution of equations of the diffusion type
with diffusivity concentration-dependent, II, Aust. J. Phys., I0,
29-42, 1957a.
Philip, J. R., The theory of infiltration, 1, The infiltration equation
and its solutions, Soil Sci., 83,345-357, 1957b.
Philip, J. R., Absorption and infiltration in two- and three-
dimensional systems, in Proceedings, lASH/UNESCO Sympo-
sium, vol. 1, Water in the Unsaturated Zone, edited by R. E.
Rijtema and H. Wassink, pp. 503-525, UNESCO, Paris, 1966.
Philip, J. R., Theory of infiltration, Adv. Hydrosci., 5, 215-296,
1969.
Philip, J. R., Infiltration in one, two, and three dimensions, in
Advances in Infiltration, Proceedings of the National Conference
on Advances in Infiltration, Chicago, U.S.A., pp. 1-13, American
Society of Agricultural Engineers, St. Joseph, Mich., 1983.
Philip, J. R., Inverse solution for one-dimensional infiltration and
the ratio A/K•, Water Resour. Res., 26, 2023-2027, 1990.
Philip, J. R., Hillslope infiltration: Planar slopes, Water Resour.
Res., 27, 109-117, 1991.
White, I., and M. J. Sully, Macroscopic and microscopic capillary
length and time scales from field infiltration, Water Resour. Res.,
23, 1514-1522, 1987.
J. R. Philip, Centre for Environmental Mechanics, CSIRO, GPO
Box 821, CANBERRA ACT 2601, Australia.
(Received March 9, !990;
revised October 3, 1990;
accepted January 11, 1991.)