Académique Documents
Professionnel Documents
Culture Documents
Abstract: The Richards equation was used to numerically model 1D flow within unsaturated topsoil and solved using a two-step
Crank-Nicolson scheme. Under constant water supply, the Darcy velocity obtained at the upper boundary, the ground surface, is regarded
as the correct infiltration rate. Based on this, three infiltration-capacity formulas by Horton, Philip, and Green-Ampt were evaluated for
three types of soil. Results demonstrated that all three formulas provide similar fits to the numerical results, but the Horton formula differs
most as compared to the other two formulas in terms of infiltration rate. The cumulative infiltration curves were not distinguishable from
each other. By perturbing the rainfall intensity, duration, and initial soil moisture content, regression results demonstrated that only the
saturated hydraulic conductivity K s can be used as a good estimator for parameter k or f c in all formulas. The other two parameters,
however, will deviate from their original values. Only the late-time infiltration parameter k or f c can be predicted a priori using measured
k s values.
DOI: 10.1061/ASCE1084-069920027:5373
CE Database keywords: Infiltration; Percolation; Models.
paper is part of the Journal of Hydrologic Engineering, Vol. 7, No. 5,
September 1, 2002. ASCE, ISSN 1084-0699/2002/5-
M k k (1)
373379/$8.00$.50 per page. t z z z
q d Qk
h
z
t0
z0
if u 0 (6)
boundary and infinite depth Segol 1993. Results are shown in
Hsu et al. 1995, which shows a good match between the nu-
merical and analytical solutions.
u ponding if u ponding (7)
u
t
z0
Qq d Qk
h
z
z0
if 0 u ponding (8) Test Problems and Methods of Analysis
zDEP
0 (9)
Three types of soil were tested. For each type of soil, a base case
was determined first for a homogeneous initial water content i0
in soil, a constant rainfall intensity Q 0 , and duration of simula-
where q d Darcy velocity at the ground surface; Qrainfall in- tion as 10 times the ponding period t p . The three base cases were
tensity (L/T); zthe depth from the ground surface positive named SAND, LOAM, and CLAY. For each base case, the value
downward, and negative upward, Lpressure head if posi- of the initial water content was perturbated up and down to de-
tive or suction head if negative; and khydraulic conductivity termine the sensitivity. These cases were named SAND1.5,
(L/T), which is a function of and can be derived from a soil SAND0.67, LOAM1.5, LOAM0.72, CLAY1.25, and
water retention curve; hzhydraulic head; CLAY0.8. Similarly, for each base case, the value of the rainfall
pondingponding depth L; and U value at the soil surface intensity was also perturbated up and down. They were named
and must be evaluated from a numerical solution. In this study, SAND1.2Q, SAND0.9Q, LOAM1.2Q, LOAM0.9Q, CLAY1.2Q,
and CLAY0.9Q. All together, there were 15 cases. Finally, the for the beginning period of infiltration, into a form, Eq. 5 similar
rainfall duration was extended to 20 times the ponding period for to Philips with 2k s in the same place as parameter S.
selected cases to determine its effects. The parameters of these Figs. 1 and 2 compare the infiltration curve using the numerical
cases are given in Tables 2 4, where i0 is the initial soil water method and three infiltration formulas for the base cases in clay
content. For case with i / i0 1.25, its initial soil water content and sand soil, respectively. Results of cumulative infiltration in
was 25% higher than i0 ; Q 0 was the basecase rainfall intensity clay is plotted in Fig. 3. Similar plots were obtained for the rest
at the upper boundary condition. The infiltration rate was equal to cases.
the Darcy velocity obtained at the ground surface.
Once the infiltration curves were obtained from the numerical Figs. 1 and 2 show that Hortons curve deviates from the numeri-
solution, we applied a nonlinear least square regression to capture cal curve most, which can be attributed to the mathematical be-
the best-fitted parameters f 0 , f c and for the Horton formula. havior of exponential decay. On the other hand, the Philip and
After substituting f 0 , f c , and into the Horton formula, the Green-Ampts curves, obtained by only three intersections, fit
cumulative error can be measured for different assumptions of f 0 . very well with the numerical solution. The first point (tt p ) in
The combination of f 0 , f c and with the least error measured are Hortons curve is always lower than the numerical value of Q.
the best-fitted parameters. In all cases, 451 numerical data points Fig. 3 shows that the cumulative infiltration curves are undistin-
from t p to 10t p , with uniform time interval, were used to fit the guishable. In loam and clay cases, results are the same.
Horton parameters. The corresponding parameters for the Horton If we follow the trends of the infiltration formulas out of the
formula are listed in Table 5. regression region tt p or t10t p , Hortons curve is lower than
The processes in determining the numerical values of the the other two for tt p region. As t approaches 0, Philips 1/t
Green-Ampt and Philip parameters are less rigorous. Three inter-
section points (t i , f i ) between the Hortons best fitted curve and
the numerical infiltration curves were utilized. The Green-Ampt Table 5. Regression Results of Hortons Parameters
parameters , k, t 0 and the Philip parameters S, k, t 0 were f0 fc Standard error Mean error
obtained by solving three simultaneous equations with three un- Parameters cm/h cm/h cm/h cm/h
knowns; t 0 artificial time-shift parameter. Note that all formulas
Clay 1.2Qa 0.110 0.0559 0.0229 1.6E-03 7.5E-03
have three parameters to be fitted, which makes the comparison
fair. The results are listed in Table 6. In Table 6, the DIV% is the Clay 0.0924 0.0512 0.0139 1.2E-03 1.2E-03
deviation percentage between the Philip parameter S and Clay 0.9Qa 0.0845 0.0490 0.0106 1.0E-03 1.4E-02
2k s , because the Green-Ampt formula can be simplified, Clay 1.25b 0.0929 0.0519 0.0275 1.2E-03 1.4E-02
Clay 0.0924 0.0512 0.0139 1.2E-03 1.2E-03
Clay 0.8b 0.0926 0.0510 0.0101 1.2E-03 1.8E-03
Table 3. Parameters of Numerical Experimental Cases in Loam Soil Loam 1.2Qa 1.59 1.04 0.177 1.4E-02 3.6E-02
Loam 1.36 1.01 0.115 6.8E-03 7.4E-02
Loam Loam Loam Loam
Loam 0.9Qa 1.24 1.00 0.0964 2.8E-03 1.5E-02
Variable Loam 1.2Qa 0.9Qa 1.5b 0.72b
Loam 1.5b 1.35 1.01 0.266 7.4E-03 9.2E-03
Q/Q 0 1 1.2 0.9 1 1 Loam 1.36 1.01 0.115 6.8E-03 7.4E-02
i / i0 1 1 1 1.5 0.72 Loam 0.72b 1.36 1.01 0.0869 6.4E-03 1.9E-02
tp (h) 6.4 3.508 9.896 2.684 8.632 Sand 1.2Qa 22.9 16.1 7.28 1.7E-01 3.9E-01
Duration h 64 35.08 98.96 26.84 86.32 Sand 19.5 15.6 4.12 8.0E-02 1.2E-01
DEPmin (cm) 420 282 576 516 394 Sand 0.9Qa 17.8 15.5 2.89 3.9E-02 2.0E-03
t max (h) 0.0128 0.00672 0.0126 0.0718 0.00557 Sand 1.5b 19.5 15.6 4.91 8.1E-02 2.8E-01
t (h) 0.004 0.004 0.004 0.004 0.004 Sand 19.5 15.6 4.12 8.0E-02 1.2E-01
z (cm) 3.0 1.5 3.0 3.0 2.0 Sand 0.67b 19.5 15.6 3.71 8.0E-02 1.2E-01
a a
Rainfall
intensity. Rainfall
intensity.
b b
Soil water content. Soil water content.
function soars up faster than the natural-log behavior of the will overestimate infiltration rate while the Green-Ampt and
Green-Ampt formula, which also contributes to a higher t 0 value Philip formula will underestimate it after a longer time period.
in the Philip parameter. The trends in Figs. 1 and 2, extended to According to Table 2 4, under the same rainfall intensity, t p is
t10t p , indicate that Hortons final infiltration value is larger proportional to ( s i ). The depth touched by the wetting
than the saturated hydraulic conductivity of soil k s . The Green- front (DEPmin) is proportional to the accumulated rainfall
Ampts and Philips curves are lower than k s after t10t p . Based (Qduration). The maximal time step (t max) feasible at the
on this, extrapolation of the Horton formula with fitted parameter, beginning decreases for larger Q and increases for larger i .
S 2 iK/2
t p Philip (11)
2i iK 2
Fig. 3. Cumulative infiltration curves for base case in clay soil t p Horton
1
i
f 0 i f c ln
f 0 f c
i f c (12)
Table 7. Comparison of t p and t 0 in Green-Ampt and Philips Formulas between Numerical Fitting and Formula Prediction
GREEN-AMPT PHILIP
t p (h) t 0 (h) t p (h) t 0 (h)
t p by
numerical t 0 by t0 t 0 by
fitting t p by Deviation numerical by Deviation t p by Deviation numerical t 0 by Deviation
Cases h formula % fitting formula % formula % fitting formula %
Clay 1.2Qa 25 25.2 1 10.85 9.64 13 22.2 11 9.17 11.4 20
Clay 41 42 2 17.39 14.80 17 35.2 14 13.70 18.8 27
Clay 0.9Qa 55.6 56.7 2 22.86 19.50 17 46.4 17 17.24 25.6 33
Clay 1.25b 21.4 21.5 1 8.87 7.60 17 18 16 6.89 9.67 29
Clay 41 41.8 2 17.33 14.80 17 35.2 14 13.70 18.8 27
Clay 0.8b 56.4 57.7 2 23.96 20.70 16 48.7 14 19.02 26.2 27
Loam 1.2Qa 3.5 3.25 7 1.20 1.42 15 2.4 32 0.72 1.99 64
Loam 6.4 5.04 21 1.56 3.05 49 3 53 0.47 4.39 89
Loam 0.9Qa 9.9 6.60 33 1.67 5.29 69 3.9 61 0.32 7.28 96
Loam 1.5b 2.7 2.20 18 0.69 1.15 40 1.4 48 0.27 1.72 84
Loam 6.4 5.04 21 1.56 3.05 49 3.2 50 0.57 4.39 87
Loam 0.72b 8.6 6.64 23 2.04 4.30 52 4.2 52 0.70 6.07 88
Sand 1.2Qa 0.087 0.07 15 0.03 0.03 22 0.05 37 0.01 0.0503 72
Sand 0.18 0.14 25 0.04 0.07 49 0.08 57 0.01 0.13 93
Sand 0.9Qa 0.314 0.19 40 0.04 0.17 78 0.09 70 0.00 0.257 100
Sand 1.5b 0.15 0.11 24 0.03 0.07 56 0.06 57 0.01 0.109 93
Sand 0.18 0.14 25 0.04 0.07 49 0.08 57 0.01 0.13 93
Sand 0.67b 0.2 0.15 23 0.04 0.09 55 0.09 57 0.01 0.145 93
a
Rainfall
intensity.
b
Soil water content.
Table 8 compares the sensitivity of the parameters of the three from the simplification of the wetting front assumption in the
formulas for change in initial water content in soil and rainfall derivation of the Green-Ampt formula.
intensity. The standards are based on the base cases. We defined Considering the difficulty in assigning formula parameters, it
the term stable if the deviation was less than 5% from the is easy to select the parameter K for both the Green-Ampt and
original values. For the Horton formula, parameter f c is stable. Philip formulas and parameter f c in the Horton formula; t 0 and
Initial water content affects only the parameter ; both parameters in the Green-Ampt formula as well as t 0 and S in the Philip
f 0 and increase if Q increases; K values for both the Philip and formula can be estimated by either formulas, tables, or physical
Green-Ampt formulas are stable; and S are stable for clay parameters with moderate errors; in the Horton formula is al-
cases and positively related to Q for loam and sand cases. Sorpo- most impossible to prescribe.
tivity S decreases when i increases, which is physically reason-
able; is not affected by i and is superior to S in this sense.
Notation
The following symbols are used in this paper:
Summary and Conclusions D K() / diffusivity;
F cummulative infiltration;
To assess three infiltration formulas using the numerical solutions f c final infiltration rate in Horton for-
of the Richards equation is well based theoretically, because the mula cm/h;
three formulas are all related to or just a simplification of the f 0 maximum infiltration rate in Horton
Richards equation. For a fair comparison between the three for- formula
mulas, a time shift parameter t 0 is added for both the Philip and cm/h;
Green-ampt formulas to become a three-parameter formula. f p infiltration-rate capacity when abun-
The ponding period t p is essential when rainfall intensity is dant rainfall supply cm/h;
larger than the saturated hydraulic conductivity k s . The numerical h total head cm;
solution can produce with t p automatically, while for formulas, k hydraulic conductivity cm/h;
artificial adjustment is necessary and, in some cases, for example k(),k() hydraulic conductivity function of
Hortons, if the rainfall intensity is larger than k s , no t p can be or ;
estimated. k s saturated hydraulic conductivity cm/
The Green-Ampt and Philip formulas fit better than the Horton h;
formula in terms of the infiltration rate curves. The deviation of M () /; specific moisture capacity;
cumulative infiltration in the three formulas is negligible. The key n parameters of van Genuchten formula;
to the use of these formulas is the values prescribed for formula Q rainfall intensity cm/h;
parameters. Q 0 base case rainfall intensity cm/h;
For the Horton formula, parameter is very difficult for users q d Darcy velocity cm/h;
to determine. It changes with rainfall intensity, initial water con- S soil sorptivity (cm/ h);
tent in soil, and the duration of the rainfall event. Parameter f 0 is t time h;
always a little bit lower than the rainfall intensity. Parameter f c is t p ponding time h;
stable and is a little larger than k s . z depth from ground surface cm;
For the Philip formula, the K value is stable and can be as- parameters of van Genuchten formula
signed as the k s value of soil; t 0 estimated by formula can have 20 1/cm;
to 100% error. Parameter S is stable for rainfall intensity but can s i , deviation of saturated and
be affected by the soil water content. If S is prescribed using initial soil water content (cm3 /cm3 );
2k s , the error is 10 to 150% overestimated. deviation between saturated and initial
For the Green-Ampt formula, the condition of parameters K pressure head cm;
and t 0 are similar to the Philips. Values of the best-fitted param- soil water content cm3/cm3;
eter are fairly stable but smaller than the deviation of the 0 base case soil water content
initial and wetting front saturated . This probably originates cm3/cm3;