Two-term infiltration equation is obtainable by integration from an exact solution of the one-dimensional downward Richards equation. The mathematical trial solution is a variables-separable combination of product and additive forms. The absence of both these stipulations in customary ponded-infiltration measurements thus accounts for why the two-term equation frequently fails experimentally.
Two-term infiltration equation is obtainable by integration from an exact solution of the one-dimensional downward Richards equation. The mathematical trial solution is a variables-separable combination of product and additive forms. The absence of both these stipulations in customary ponded-infiltration measurements thus accounts for why the two-term equation frequently fails experimentally.
Two-term infiltration equation is obtainable by integration from an exact solution of the one-dimensional downward Richards equation. The mathematical trial solution is a variables-separable combination of product and additive forms. The absence of both these stipulations in customary ponded-infiltration measurements thus accounts for why the two-term equation frequently fails experimentally.
Exact mathematical derivation of a two-term infiltration equation D. Swartzendruber Department of Agronomy, University of Nebraska-Lincoln Abstract. Previous mathematical developments of so-called two-term infiltration equations have been approximate instead of exact and thus provide little insight to account for any experimental shortcomings of such equations. In contrast, it is here shown that one form of two-term infiltration equations is obtainable by integration from an exact solution of the one-dimensional downward Richards equation subject to the customary initial and boundary conditions. The mathematical trial solution is a variables-separable combination of product and additive forms. Intrinsic to the solution are two further stipulations: (1) the unsaturated hydraulic conductivity function must be linear with the water content, and (2) the ponded-water head on the soil surface must increase as the square root of time from initial water application. The absence of both these stipulations in customary ponded-infiltration measurements thus accounts for why the two-term equation frequently fails experimentally. Nonetheless, the mathematical simplicity of the new solution makes it useful pedagogically. Introduction The two-term water infiltration equation [Philip, 1957c] can be written I = Ss tl/2 q- Mt (1) where I is the cumulative quantity of infiltration after time t of ponded-water application to the soil surface, and S s is the water sorptivity of the soil alone, with subscript s denoting zero ponding (i.e., the applied water layer on the soil surface is infinitesimally thin). As originally presented, parameter M, akin to hydraulic conductivity, could be linked rigorously with Philip's [1957a] exact, t 1/2 series solution only by constraining t from becoming too large. This would mean that 0 < M < Ko [Philip, 1957c, pp. 261-262; 1969b, p. 275], where Ko is the sated (satiated) hydraulic conductivity of Miller and Bresler [1977]. Notwithstanding the early optimistic expectations [e.g., Philip, 1957c; Watson, 1959] for the practical utility of (1) in spite of its intrinsic time constraint, subsequent work did un- cover difficulties, in the sense that the least-squares fitting of (1) (Watsoh's [1959] procedure) to experimental field data yielded negative values of M in half or more of the data sets [Skaggs et al., 1969; Fahad et al., 1982]. Admittedly, such be- havior could be rationalized or excused by appealing to the problematic influence of the time constraint inherent in the origins of (1). Unfolding confusingly alongside these findings was the erroneous claim that M = Ko was rigorous [Miller and Klute, 1967, p. 226]. There was a passing suggestion [Philip, 1969b, p. 284] that (1) fitted to the whole t range of infiltration data would tend to give M = Ko. In the Childs [1969, p. 279] sense of "an empirical formula to which one is led on physical grounds," Swartzendruber and Youngs [1974] set M = Ko in (1) and compared it with two other infiltration equations. Swartzendruber [1987, p. 812] also Copyright 1997 by the American Geophysical Union. Paper number 96WR03906. 0043-1397/97/96WR-03906509.00 found that (1) with M - Ko would emerge from a quasi- solution, multiparameter exponential form when the number of parameters was reduced from six to just two: S and K o. This, in turn, ultimately prompted the question, Can we pose an exact-solution boundary value problem that will produce (1) with M = Ko and thus bestow exactness upon the two-term infiltration equation (in Ko) without constraint on time t? The answer is yes, and we now turn to the details of this problem, the exact solution of which also affords insight into the reasons for failure in I - Ss tl/2 q- Kot when fitted by least squares to infiltration data as customarily obtained. Mathematical Analysis The Problem Consider one-dimensional downward infiltration of water into an infinitely long, uniform, rigid soil (Figure 1). The gov- erning partial differential equation is that of Richards [1931], which when written in transformed manner [Philip, 1957a, p. 350; Swartzendruber, 1969, p. 229] is OZ O[D] dK Ot = O0 oZ/00 + dO (2) wherein Z - Z(0, t) is the depth (positive downward, below the soil surface) at which the volumetric water content is 0 at time t. D = D (0) and K = K(0) are the soil water diffusivity and hydraulic conductivity functions, respectively, and (2) is taken subject to Z>0 0= 0n t=0 (3) Z=Z0 0= 00 t>0 (4) Zo = roh/(dI/dt - ro) (5) where 0n is the uniform initial water content; 0o is the sated water content (constant) corresponding to K(0o) = Ko result- ing from the water depth h being ponded instantaneously on the soil surface (Z - 0) from t = 0 + onward; Z o is the depth 491 492 SWARTZENDRUBER: TECHNICAL NOTE Soil Surface, Z = 0 time t < 0 time t = 0 time t > 0 (a) (b) (c) Figure 1. Schematic diagram of the one-dimensional down- ward infiltration problem showing three time stages in the application and maintenance of ponded water. to which 0o (at zero pressure or suction head) has penetrated in time t (for zero air entry value); dI/dt is the volumetric infiltration flux of the water across (into) the soil surface; and cumulative infiltration I is the total volume of water infiltrated (in time t) per unit of bulk soil cross-sectional area perpen- dicular to the downward direction of water flow. Figures lb and lc depict conditions (3) and (4), respectively. Boundary condition (4), which limits (2) to the region Z _> Z o with Z o specified by [5), was first given in essence by Philip [1958a, p. 280]. He noted for the region 0 <- Z <_ Z o that the water content is the constant 0o, so that K(0o) = Ko, and that the pressure head is h at Z = 0 and is zero at Z = Z o. With the total difference in hydraulic head across Z o thus being h + Z o, he applied Darcy's equation to the Z o region and solved for Z o to obtain (5). We here shall also explicitly inject the possibility of a more general, time-dependent expectation for h, namely h =hc+p(t) (6) where h c is constant and p(t) is a function only of time, with p(0) = 0. This means that the initial ponded head h c is applied instantaneously on the soil surface at t = 0. In the special case p (t) = 0 for all t, (6) would then reduce to the constant ponded-head case h -- h c [Philip, 1958a]. Last, note that the instantaneous imposition of ponded-head h c at t = 0 will in turn raise the water content instantaneously from 0,, to 0o at the soil surface (Z = 0). Hence the position coordinate Z of any 0 between 0,, and 0o will originate from zero at t = 0; analytically Z=0 t=0 for On< 0--< 00 (7) The Solution As a trial solution of (2) subject to conditions (3), (4), and (5), along with stipulation (6) and corollary (7), we introduce Z = O(o)r(t) + U(t) (8) where 4>(0) is a function only of 0, and T(t) and U(t) are each functions only of t, but T and U are different functions. The form of (8) is a variables-separable combination of product and additive types. From partial differentiation of (8) with respect to t and 0, respectively, og/ot = O(o)r'(t) + U'(t) (9) og/00 = O'(o)r(t) (t0) where T' = dT/dt, U' = dU/dt, and 4>' = dO/dO. Using (9) and (10) in (2)yields or' + U' = -O(D/O'T)/00 + g' (lt) where K' = dK/dO, but O(D/O'T)/00 = T-O(D/O')/00 = r-d(D/O')/dO (12) so that the use of (12) in (tt) gives or' + U' = -r-d(D/O')/dO + g' (13) As it stands, the ordinary differential equation (13) is not of variables-separable form but can be made so by setting U' =K' (14) and then (13) becomes, after rearrangement, rr' = -O-d(D/O' )/d O (15) Beginning with (14), note that the left-hand side is a function only of t while the right-hand side is a function only of 0, thus meaning that both sides can be set equal to the separation constant a, or d U/dt = a (16) dK/d 0 = a (17) Integrating (16) and (17) yields, respectively, U = at + C (18) K=a0+C2 (19) where C and C2 are constants of integration. The successive use in (19) ofK = K(On) = Kn at 0,, and ofK = K(0o) = K o at 0o enables the evaluation of a by subtraction and rear- rangement to be written a = (go- gn)/(0o- On) (20) Turning next to (15), the variables again are separable with separation constant b, namely, rr' = b (21) d(D/O')/dO = -bO (22) Solution of (21) and (22) is akin to that of Swartzendruber [1966] but with some change in details. Integrating (21) and solving for T yields T = (2bt + C3) /2 (23) where C3 arises from the constant of integration. Putting (23) and (18) back into the trial solution (8) gives Z = O(O)(2bt + C3) 1/2 --[- at + C (24) wherein a is retained for economy in writing symbols until the right-hand side of (20) is expressly needed. Using corollary (7) in (24) provides, after rearrangement, C 1 = -C31/20(0 ) (25) SWARTZENDRUBER: TECHNICAL NOTE 493 in which finite nonzero C and C3 are not permissible, since that would force qb(0) to be constant over the whole range of 0 ( 0, to 0o). Neither could qb(0) be chosen as zero for the general nontrivial case. It might seem consistent with (25) to choose C and C3 to be large without bound (i.e., c), but such choices would not be useful in (24). Hence we select C3 = 0 for the general finite qb(0) > 0, which from (25) makes C = 0, so that (24) becomes g = &(O)(2bt) /2 + at (26) Although (2b)/2&(O) in (26) could be defined as a new function of only 0, the same end is achieved without loss of generality by choosing b = 1/2 and continuing the use of qb(0), which makes (26) and (22) become, respectively, Z = c(O)t /2 + at (27) d(D/&')/dO = -qb/2 (28) Note that the consequence embodied in (14), arising out of the variables-separation process, has forced the K( 0 ) relation- ship to be the linear (19). Such linearity is indeed a noteworthy limitation, but has nonetheless been employed theoretically before by Parlange and Fleming [1984, equation (6)], taking 0, = 0, D(0 ) as a power function, and allowing 0 o to be time dependent. Although they hypothesized a similarity expression (their equation (7)) reminiscent of (27) but with t /2 replaced by t -m, they did not pursue our problem here in which 0o (condition (4)) is constant rather than time dependent. To determine whether the initial condition (3) is satisfied, rewrite (27) as git /2= 49(0) + at /2 (29) Putting condition (3) into (29) by letting 0 0, as t - 0 for nonzero finite Z yields 4'(0) (30) which thus supplies the first condition to which the ordinary differential equation (28) becomes subject. Next, putting boundary condition (4) into (27) gives Z o = Co t/2 + at (31) where qb o = qb(0o). To investigate whether (31) can be matched with (5) will require I for use in (5), which as a general for- mulation [Philip, 1957a, p. 352] for fixed t: I I = rnt + Z dO (32) If (27) is the solution to the problem, the Z from it can be substituted into (32) to obtain i00 I = Knt + t /2 c(O) dO + at dO (33) where t /2 and t are taken through their respective integral signs because t is fixed. In (33), carry out the simple evaluation of the second integral and substitute for a from (20), to yield ultimately I I = t /2 &(O) dO + rot (34) In (34), since 4> is a function only of 0 and the limits of integration are fixed, the integral is fixed. For an infinitesimally thin layer of ponded water (h = 0), Philip [1957c, p. 257; 1969b, p. 237] defined a similar fixed integral as the soil sorp- tivity, Ss, in (1). Philip [1958a] extended the sorptivity concept to include the effect of nonzero ponding, which is here ex- pressed explicitly as total sorptivity S, namely s = 0(o) ao (35) although Philip [1958a, b] did not write (35) explicitly in the sense of total sorptivity. Nonetheless, it is to be emphasized in the present context that the S of (35) is conceived to include the condition of nonzero water ponding on the soil surface. Using (35) in (34) hence yields I: St /2 + Kot (36) To determine whether (36) will satisfy (5) and (6), we find by differentiation of (36) that dI/dt = K o + S/2t /2, which when substituted into (5) along with the h of (6) yields Zo = 2(Kohct/2/S) + 2Kop(t)t/2/S (37) The Z o of (31) matches exactly the Z o of (37) if we make the identifications Oo: 2gohc/S (38) p(t) = Sat/2/2Ko (39) where (39) provides the particular form of p(t) in (6), con- forming as well to p(0) - 0 as mentioned. Also, (38), the second condition to which (28) is subject (along with (30)), is the same as that found by Philip [1958a, p. 280, second of conditions (15)] for constant-head nonzero ponding in hori- zontal water adsorption. Integration of (28) from 0, to 0 [Philip, 1955] leads to o dO ck dO = -D(O) dck (40) In principle, this integration between limits also yields the term D(On)[dO/dck]o:o, added to the right-hand side of (40), but this term vanishes because condition (30) implies d O/d ck -- 0 at 0 - 0,. The ordinary integrodifferential equation (40) remains subject only to 4> = qbo at 0 = 0o, with qb o given by (38). For D(0), 0,, 0o, and h c supplied as inputs, (40) is solvable (for qb(0)) by the numerical method of Philip [1955] as modified [Philip, 1958a, p. 280] for the nonzero qb o (for h c > 0) of (38). If, however, h c = 0, meaning that initial ponding (at t = 0) is infinitesimally thin because h (0) = 0 in (6) by virtue of p (0) = 0, then qb o of (38) is also zero, and the numerical method for determining qb(0) reverts back to the original one of Philip [1955]. Also, of course, with qb o = 0 (from h c = 0), S would then be replaced by S s in (35) through (39), thus reducing (36) to (1). With the mathematical proof thus completed, the two-term infiltration equation (36) is therefore seen to arise exactly from the two-term solution embodied in (27), namely, the solution of (2) subject to conditions (3) and (4) as supplemented by (5), (6), and (7). Included in (36) are both zero and nonzero initial surface ponding, for hc = 0 and hc > 0, respectively, although even the case of zero initial ponding becomes both ponded and 494 SWARTZENDRUBER: TECHNICAL NOTE progressively more deeply ponded for t > 0, by virtue ofp (t) from (39) appearing in (6). Results and Discussion From the two-term (27), it is clear that the new exact solu- tion is simply the additive combination of b() t/:, the clas- sical horizontal solution [Philip, 1955], plus the linear term at. This same additive linear term also appears in the classical, constant-ponding infiltration solution for asymptotically large times [Philip, 1957b]. Graphical illustration of the new two-term solution requires numerical values of b(O) and a in (27). Such values were given by Philip [1958a] for the now classical Yolo soil material under nonzero ponding (h = 0.25 m of water) and also for zero ponding (h = 0) [Philip, 1957a, b], along with K o = 123.00 rim/s, K = 0.12 rim/s, o = 0.4950, and = 0.2376, to yild a = 477.39 nm/s from (20). We recognize from the outset, however, that these numerical values of Philip's were deter- mined from a nonlinear K() rather than from a linear K(O), as in (19). Nonetheless, it is still deemed of interest to utilize these values for illustrating profiles in this first-time explor- atory sense. For h = 0.25 m of water, a water-content profile, or set of paired values (, Z), was determined from (27) at each se- lected time of 0.01, 0.04, 0.10, and 0.20 Ms. A second sequence of profiles for the same times was determined from Philip's [1958a] solution under constant pondcd head of 0.25 m (i.e., p(t) = 0 in (6) for all ! > 0), using the first three terms of his exact t / series solution. A third sequence of profiles at these same times was determined from just the first term, b(O), on the right-hand side of (27), which thus constitutes the profiles for horizontal water absorption (gravity free). Graphic profiles for these three cases are shown in Figure 2. At very small times, all three proilia cases merge as they should, because b(O) predominates. As time increases, the gravity-free profiles lag progressively behind their gravity- aided infiltration counterparts, again, as would be expected. Also, the forward portions of the constant-ponding (p (!) = 0) profiles become steeper (as viewed from the Z axis) in com- parison with their gravity-free and new-solution counterparts, once more as expected. The new-solution profiles clearly dem- onstrate their progressive translational advancement beyond their respective gravity-free counterparts, by the distance +at in (27). Also, these new-solution profiles arc progressively ahead of even their respective constant-ponding profiles (p(t) = 0), in response to the time buildup of pondcd-watcr head expressed by p(t) 0 of (39) in (6). Profiles for zero initial ponding (h - 0) on the Yolo soil material [Philip, 1957a, b], using the scaled data of Swartzen- dmber [1987, p. 813], are shown in Figure 3. To avoid confusing overlappings, the zero-gravity profiles are omitted. Obviously, the behavior and trends of Figure 2 are not only sustained but even accentuated in Figure 3, especially the substantially deeper penetration of the new-solution profiles as compared with their respective constant-ponding profiles. For example, at 3.5 Ms (17.5 times the maximum 0.2 Ms in Figure 2), the constant-ponding profile is everywhere more shallow than even the 3.0-Ms new-solution profile. Also, as time increases over the range shown in Figure 3, the new-solution profiles become progressively less steep than their respective constant-ponding profiles, as viewed from the Z axis. This means that the new- solution profiles do not approach the constant-shape, linearly 0.1 0.4 0.5 / WATER CONTENT 0 h c =0.25m 0.01 0o0'
p(f) > O, new solution. / p(t) = O, Philip [1958a] J ...... zero gravity. Figure 2. Three types of water-content profiles at different times of downward infiltration into the Yolo soil material of Philip [1958a], for an initial ponded-water head of h c - 0.25 m of water on the soil surface. translating, traveling-wave profiles of the constant-ponding case [Philip, 1957b, pp. 447-448], even though the infiltration flux dI/dt from (36) does approach K o as t - . This behavior is qualitatively akin to the linearized, constant-diffusivity solu- tion of Philip [1969a], which also does not produce traveling- wave profiles, although the profiles of that linearized solution WATER CONTENT 0 0.2Ms N 0.5Ms "' 1.0Ms 1.5Ms 2.5Ms / I 3.0Ms . I 3.5Ms p(f) > O, new solution p(t) = O, Philip [1957a]. Figure 3. Two types of water-content profiles at different times of downward infiltration into the Yolo soil material of Philip [1957a], for an initial ponded-water head of h c - 0 on the soil surface. For times of 1.0 Ms and larger, the two types of profiles at a given same time are connected by a vertical line. SWARTZENDRUBER: TECHNICAL NOTE 495 K 0 K(e) / i K n 0 ' e On eo Figure 4. Schematic curve (solid) of K(0) for a coarse- textured soil. Also shown for the high-0 range (0n -< 0 -< 0o) is a least-squares linear approximation (dashed line) of the true curve. are far less satisfactory in shape than even those of the new solution in Figures 2 and 3. We now turn to the two major restrictive assumptions resi- dent in the new two-term solution (27) and hence also in (36). The first is the requirement in (19) of a linear K( 0 ). Notwith- standing the employment of such linearity both here and by ?arlange and Fleming [1984], K(0) is generally strongly non- linear. For coarse-textured soils, however, the nonlinearity might be dealt with approximately, as portrayed in Figure 4 by a least-squares straight line fitted to the high-0 range of the actual curvilinear K( 0 ). The second and perhaps even greater restrictive assumption is that of the time-dependent buildup of ponded-water head on the soil surface, as embodied in the functionp (t) given by (39). Once again, using Philip's [1957a, 1958b] values of S, a, and K o to evaluate p(t) from (39), h was determined from (6) and plotted in Figure 5 for h c values of 0, 0.25, and 0.50 m of water. ' I ' I ' I 1.25 - - E 1.00 - hc = 0.50m _ -r 75 -- n,' = 0.25m .5o - I .25 - hc = 0 _ z 0 0 0 1.0 2.0 5.0 TIME f (Ms) Figure 5. Time curves of ponded head h from (6) with p(t) given by (39) for three values of initial ponded head h c = 0, 0.25, and 0.50 m of water, for the Yolo soil material of Philip [1957a, 1958a]. Even for h c -- 0, the value of h is 71.5 mm of water after 86.4 ks (1.00 day), and 455 mm after 3.5 Ms (40.51 days). These values of h are in substantial and changing disagreement with a constant h of, say, 15 mm in typical infiltration experiments, in which the time-dependent ponded-head buildup for the validity of the two-term (36) is not being met, even apart from any failure of K(0) to be linear as in (19). If an experimental assessment of the new solution were undertaken in the sense of Figure 4, a corresponding buildup of time-dependent ponded head in accordance with (39) and (6) should be feasible experimentally. To carry this out, how- ever, would require prior experimentation for determining S, a, and K o of (39) before conducting the infiltration experiment portrayed in Figure 1. Unfortunately, though, the overall de- finitiveness of the experimentation would still be conditioned by the extent to which K(0) in the high-0 range were truly linear. Even though the assumption of a linear K(0) is an impor- tant limitation, it is nonetheless the only constraint on the three soil-characterizing functions K(0), D (0), and r(0), where r is the soil-water suction head. Specifically, there is no need for the artificiality of any of these three functions being held constant, in contrast with the requirement D(0) = con- stant in the solutions of Philip [1969a] and of Knight [Philip, 1974, p. 261]. Note also that these two solutions, along with those of Philip [1957a, b, 1958a], are far more complicated mathematically than the new solution embodied in (27) and (36). Hence this new and relatively simple exact solution for downward infiltration is useful pedagogically for teachers and students of vadose-zone hydrology, a setting in which such simple but rigorous examples are few. In summary, the exact derivation of the two-term equation (36) includes the two important stipulations: (1) that the hy- draulic conductivity function K(0) be linear for 0n -< 0 -< 0o and (2) that the ponded-water head p(t) on the soil surface build up with time in accordance with equation (39), namely p(t) = Sat/2/2Ko . Since neither of these stipulations is met in the usual measurements of ponded-water infiltration, this accounts for why the two-term infiltration equation frequently fails experimentally. This reminds us that the mathematical exactness of equation (27), desirable and worthy though it is, does not of itself ensure the applicability of equation (36). Acknowledgments. Published as paper 11476, Journal Series, Ag- ricultural Research Division, University of Nebraska-Lincoln. Thanks are expressed to John R. Philip, CSIRO Fellow Emeritus, Center for Environmental Mechanics, Canberra, Australia, for kindly supplying data used in Figure 2. References Childs, E. C., An Introduction to the Physical Basis of Soil Water Phe- nomena, John Wiley, New York, 1969. Fahad, A. A., L. N. Mielke, A.D. Flowerday, and D. Swartzendruber, Soil physical properties as affected by soybean and other cropping sequences, Soil Sci. Soc. Am. J., 46, 377-381, 1982. Miller, E. E., and A. Klute, The dynamics of soil water, I, Mechanical forces, in Irrigation of Agricultural Lands, edited by R. M. Hagan, H. R. Haise, and T. W. Edminster, pp. 209-244, Am. Soc. of Agron., Madison, Wis., 1967. Miller, R. D., and E. Bresler, A quick method for estimating soil water diffusivity functions, Soil Sci. Soc. Am. J., 41, 1020-1022, 1977. Parlange, J.-Y., and J. F. Fleming, First integrals of the infiltration equation, 1, Theory, Soil Sci., 137, 391-394, 1984. Philip, J. R., Numerical solution of equations of the diffusion type with 496 SWARTZENDRUBER: TECHNICAL NOTE diffusivity concentration dependent, Trans. Faraday Soc., 51, 885- 892, 1955. Philip, J. R., The theory of infiltration, 1, The infiltration equation and its solution, Soil Sci., 83, 345-357, 1957a. Philip, J. R., The theory of infiltration, 2, The profile of infinity, Soil Sci., 83, 435-448, 1957b. Philip, J. R., The theory of infiltration, 4, Sorptivity and algebraic infiltration equations, Soil Sci., 84, 257-264, 1957c. Philip, J. R., The theory of infiltration, 6, Effect of water depth over soil, Soil Sci., 85, 278-286, 1958a. Philip, J. R., The theory of infiltration, 7, Soil Sci., 87, 333-337, 1958b. Philip, J. R., A linearization technique for the study of infiltration, in Proceedings of the Symposium on Water in the Unsaturated Zone, vol. 1, pp. 471-478, Int. Assoc. of Hydrol. Sci. and UNESCO, Gent- brugge, Belgium, 1969a. Philip, J. R., Theory of infiltration, Adv. Hydrosci., 5, 215-296, 1969b. Philip, J. R., Recent progress in the solution of nonlinear diffusion equations, Soil Sci., 117, 257-264, 1974. Richards, L. A., Capillary conduction of liquids through porous medi- ums, Physics, 1,318-333, 1931. Skaggs, R. W., L. E. Huggins, E. J. Monke, and G. R. Foster, Exper- imental evaluation of infiltration equations, Trans. ASAE, 12, 822- 828, 1969. Swartzendruber, D., Variables-separable solution of the horizontal flow equation with nonconstant diffusivity, Soil Sci. Soc. Am. Proc., 30, 7-11, 1966. Swartzendruber, D., The flow of water in unsaturated soils, in Flow Through Porous Media, edited by R. J. M. De Wiest, pp. 215-292, Academic, San Diego, Calif., 1969. Swartzendruber, D., A quasi-solution of Richards' equation for the downward infiltration of water into soil, Water Resour. Res., 23, 809-817, 1987. Swartzendruber, D., and E.G. Youngs, A comparison of physically based infiltration equations, Soil Sci., 117, 165-167, 1974. Watson, K. K., A note on the field use of a theoretically derived infiltration equation, J. Geophys. Res., 64, 1611-1615, 1959. D. Swartzendruber, Department of Agronomy, University of Ne- braska-Lincoln, 133 Keim Hall, East Campus, Lincoln, NE 68583. (Received April 8, 1996; revised December 9, 1996; accepted December 16, 1996.)