Lecture 8: Infiltration Modelling

8.1 General
In the hydrological cycle infiltration is a major dividing process at the soil surface atmospheric
water interphase because it determines how much water penetrates the land surface. The soil's
infiltration capacity (i.e. its steady infiltration rate) is also an important factor in selecting and
planning irrigation systems, in deciding appropriate water conservation techniques on
agricultural lands and in the hydrological modelling of runoff processes.

Owing to the difficulty of measuring infiltration rate in the field, a lot of interest has therefore,
been shown in studying and modelling the infiltration process. Several infiltration models have
been developed to model this vital property of soil.

8.2 What is modelling?

Modelling is the use of models (e.g., physical, mathematical, or logical representation of
a system, entity, phenomenon, or process) to develop/predict data utilized for decision making.
For developing a model, the parameters which are affecting the variable of interest are
considered. Once the model is developed and calibrated, the actual experimentation can be
avoided, which are generally costly and time-consuming.
8.3 Infiltration Modelling:
Several infiltration models have been developed by considering two main forces viz.
gravitational and capillary besides other soil characteristics. Among the many models
developed for monitoring the infiltration process, Kostiakov model, Horton model, Green-
Ampt model and Philip Two Term model have been studied in detail because of their simplicity
and the ease of estimating their fitting parameters. Based on the parameters involved in the
model, these models can be grouped under two categories
(i) Empirical models – Kostiakov and Horton
(ii) Physical based models – Green-Ampt and Phillip Two Term

The constants of the models are either determined by graphical method or by regression
equation. Now, with the help of computers it is common practice to use graphical method
(plotting of curve between infiltration rate and time). For the estimation of constants of the
models, the observed infiltration rate, cumulative depth of infiltration and elapsed time is
required.

8.3.1 Kostiakov Model (1932): it is a simple power function model and most commonly used
due to its simplicity to represent infiltration when time is small. However, model does not fit
well for larger value of time. The model takes a form
𝑓 = (𝑎𝑏)𝑡 (𝑏−1) (8.1)
Where,
f = infiltration rate at time t
a > 0 and 0 < b < 1 are constants of the model.
t = elapsed time, min
It is clear from the relationship that as 𝑡 → 0, 𝑓 → ∞ and as 𝑡 → ∞, 𝑓 → 0. In reality f does not
attain these values. This model cannot predict the start of runoff after the occurrence of rainfall.
By taking log on both sides of the equation, the non-linear form can be converted in to linear
form as
𝑙𝑜𝑔𝑓 = 𝑙𝑜𝑔(𝑎𝑏) + (𝑏 − 1)log (𝑡) (8.2)
Determination of constants of Kostiakov model: Graphical method
The following steps are involved for the computation of constants
(i) Open an excel sheet and enter f and t. Compute log f and log t
(ii) Plot a graph between log f and log t and fit a straight line (log t on x-axis and log f on y-
axis)
(iii) Compute slope of the straight line i.e b-1
(iv) Compute intercept i.e. log (ab)=Value = ab = log inverse
(v) Substitute values of the constants in equation (8.1) and predict the infiltration rate

8.3.2 Horton Model (1939, 1940): This model is applicable for overland flow. Horton has
observed that infiltration capacity decreased with time until it approached a more or less
constant value. He hypothesized that infiltration is similar to the exhaustion process according
to which the rate of performing work is proportional to the amount of work remaining to be
performed. In the case of infiltration, the work remaining to be performed at any time t is equal
to that of changing the infiltration capacity f to its ultimate constant value fc. The rate of
performing work is df/dt. The amount of work remaining to be performed is f – fc. Since f
decreases with time,

𝑑𝑓
= −𝑘 ∗ (𝑓 − 𝑓𝑐 ) (8.3)
𝑑𝑡
where, k is a proportionality factor dependent on soil type and initial moisture content. The
initial condition can be taken as f = f0 (initial infiltration rate) at t = 0.
The model takes the following form:

𝑓 = 𝑓𝑐 + (𝑓0 − 𝑓𝑐 ) ∗ 𝑒 −𝑘𝑡 (8.4)

(𝑓 − 𝑓𝑐 ) = (𝑓0 − 𝑓𝑐 ) ∗ 𝑒 −𝑘𝑡 (8.5)
Where,
f0 = initial infiltration rate, cm/h at t = 0
fc = steady state infiltration rate prior to attaining constant infiltration rate, cm/h
e = exponential
k = constant that determines the rate at which f0 reaches fc
t = time, min
By taking natural log on both sides of the equation
ln(𝑓 − 𝑓𝑐 ) = 𝑙𝑛(𝑓0 − 𝑓𝑐 ) − 𝑘𝑡 (8.6)
The limitation of this model is the determination of reliable values of its constant f0 and fc
Determination of constants of Horton model: Graphical method
The following steps are involved
(i) Open an excel sheet and enter f and t. Plot a curve between f and t and select fc
(ii) Compute ln (f - fc) for each value of t
(iii) Plot values of ln (f - fc) against t on graph paper and fit a straight line
(iv) Compute slope of the straight line i.e. k
(v) Intercept is (f0 - fc) as f0 - fc=ln-1 (value), and f0=fc+ln-1 (value)
(vi) Substitute values of the constants in equation (8.4) and predict the infiltration rate

fc curve
30

25
Infiltration rate (cm/h)
20

15
f (cm/h)
10

0
0 50 100 150 200
Time (min)

8.3.3. Green-Ampt (GA) model (1911): A simple model, based on Darcy’s law was proposed
for infiltration into uniform soil with uniform moisture content due to a pool of negligible depth
of water. The water is assumed to infiltrate into the soil. The infiltration water defines a sharply
wetting front separating the wetted and unwetted zones as shown in Fig. 8.1.

Fig. 8.1. Schematic diagram of the Green-Ampt model

The GA model results from application of Darcy’s law to the wetted zone as
𝐿+𝑆𝑐
𝑓=𝐾 (8.7)
𝐿

Where,
K = hydraulic conductivity of the wetted zone
L = distance from the ground surface to the wetting front
Sc = capillary suction at the wetting front.
If η is porosity, or fillable pore space, then F = ηL, and equation 8.7 can be expressed as
 𝐹+𝜂𝑆𝑐 𝜂𝑆𝑐
𝑓=𝐾 = 𝐾 (1 + ) (8.8)
𝐹 𝐹
𝑎𝐾 1
Or 𝑓 = 𝐾 + = 𝐾 + 𝑎𝐾 ∗ 𝐹 (8.9)
𝐹

Where,
F = Accumulated or Cumulative depth of infiltration, cm
a = constant = ηSc
Determination of constants of Green-Ampt model: Graphical method
The following steps are involved for computation of constants
(i) Compute the values of 1/F
(ii) Plot a graph between 1/F and f and fit a straight line
(iii) Compute slope i.e. aK
(iv) Compute intercept i.e. K
(v) Substitute values of the constants in equation (8.9) and predict the infiltration rate
8.3.4. Phillip Two Term (PTT) Model (1957): For a uniform soil with a uniform soil-moisture
content and excess water supply rate at the surface, Phillip has found a solution to the flow
equation in the form of an infinite series. The PTT model is as given below:
1
𝑓 = 2 𝑆𝑡 −0.5 + 𝐴 (8.10)
Where,
S = Soil sorptivity, a function of initial and surface water contents of the soil and soil water
diffusivity;
t = time, min
A = parameter depending upon soil properties
Since, determination of S and A in the field is difficult, Phillip has suggested that A is
approximately K/3 (where K is hydraulic conductivity of the soil).
Determination of constants of Phillip Two Term model: Graphical method
The following steps are involved for computation of constants
(i) Compute t-0.5
(ii) Plot f against t-0.5 and fit a straight line
(iii) Compute slope i.e. S/2
(iv) Compute intercept i.e. A
(v) Substitute values of the constants in equation (8.10) and predict the infiltration rate

Determination of constants of Infiltration models: Least square method

The constants of the models are determined as follows:
𝑌 = 𝑎 + 𝑏𝑋

∑ 𝒀 ∑ 𝑿𝟐 −∑ 𝑿 ∑ 𝑿𝒀
𝒂= (8.11)
{𝑵 ∑ 𝑿𝟐 −(∑ 𝑿)𝟐 }

𝑵 ∑ 𝑿𝒀−∑ 𝑿 ∑ 𝒀
𝒃 = {𝑵 ∑ 𝑿𝟐 −(∑ 𝑿)𝟐 } (8.12)
Example 8.1: The following infiltration data are collected from double ring infiltrometer.
Determine constants of all the infiltration models using both the approaches.
Time (min) 2 4 6 8 10 15 20 25 30 35
f (cm/h) 25.2 18.3 9.9 8.4 7.8 6.84 5.64 4.92 4.68 4.2
Time (min) 40 50 60 70 85 100 115 130 145 160
f (cm/h) 4.08 3.96 3.6 3.42 3.32 2.96 2.84 2.8 2.68 2.68

Solution: The detailed calculation is as given below:

(a) Kostiakov Model

(b) Horton Model:

(c) Green-Ampt Model:

(d) Phillip Two Term Model:

8.4 Evaluation of Infiltration Models

The infiltration models are evaluated by comparing the difference between predicted
infiltration rate and the measured infiltration rate. The following model performance indices
are generally used.
8.4.1 Maximum Absolute Error
The maximum absolute error is used to measure success of numeric estimation. The maximum
absolute error (MAE) is computed as

𝑀𝐴𝐸 = 𝑚𝑎𝑥 (|𝑓𝑚 − 𝑓𝑝 |) (8.13)

Where
fm = measured infiltration rate.
fp = predicted infiltration rate.

8.4.2 Bias
The bias of an Infiltration model is the average difference between the predicted value and
the measured value of the infiltration models. The value of zero bias is called unbiased. Its
value is defined by:

∑𝑛
𝑖=1(𝑓𝑚 −𝑓𝑝 )
𝐵𝑖𝑎𝑠 = (8.14)
𝑛

8.4.3 Root Mean Square Error (RMSE)

Mean-squared error is the most commonly used measure of success of numeric estimation, and
root mean-squared error is the square root of mean-squared-error after we give it the same
dimensions as the estimated values themselves. This method exaggerates the estimated error -
the difference between predicted value and measured value (actual value).
The root mean squared error (RMSE) is computed as:

1 2
𝑅𝑀𝑆𝐸 = √𝑛 [∑𝑛𝑖=1(𝑓𝑚 − 𝑓𝑝 ) ] (8.15)

8.4.4 Nash–Sutcliffe Model Efficiency

The Nash–Sutcliffe model efficiency coefficient is used to assess the predictive power of
hydrological models. It is defined as:
2
∑𝑛 (𝑓𝑚 −𝑓𝑝 )
𝑀𝑜𝑑𝑒𝑙 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 = 1 − ∑𝑛𝑖=1(𝑓 ̅̅̅̅ 2
(8.16)
𝑖=1 𝑚 −𝑓 𝑚)

A value of 90% and above indicates very satisfactory performance, a value in the range of 80–
90% indicates fairly good performance, and a value below 80% indicates an unsatisfactory fit.
8.4.5 Willmott’s Index of Agreement (W)
W illustrates the agreement degree between predicted values and measured values of
infiltration rate. The range of W is from 0 to 1. A value of 1 for W shows perfect agreement
between the predicted infiltration and the measured infiltration at the same time, and on the
other extreme end, a value of 0 for W implies a complete disagreement. Willmott index can be
simplified as:
2
∑𝑛
𝑖=1[𝑓𝑝 −𝑓𝑚 ]
𝑊 =1− 2 (8.17)
∑𝑛 ̅̅̅̅ ̅̅̅̅
𝑖=1[|𝑓𝑝 −𝑓𝑚 |+|𝑓𝑚 −𝑓𝑚 |]

The best fit model may be selected on the basis of minimum the Maximum absolute error, Bias,
Root mean square error (RMSE) and maximum the Model efficiency criteria.
8.5 Prediction of Infiltration rate:

Kostiakov Model Green Ampt Model

Using Phillip Two Term Model

8.6 Performance Evaluation of Infiltration Models

8.6.1 Kostiakov Model

8.6.2 Green-Ampt Model

8.6.3 Phillip Two Term Model

, cm/h
, cm/h

, cm/h