Soil-Water Characteristic Curves for Unsaturated Soils - Unsaturated Soil Mechanics in Engineering Practice - page 260

30

Experimental
drying curve

Proposed PTF

20

Log PDF

10

0

10 6 0.1 1 10 100 1000 10,000 100,000

Soil suction, kPa

Figure 5.106 Comparison of experimental and predicted SWCC and logarithmic probability
density curves for silt (after Vereecken et al., 1989).

Silty clay loam
Silty clay loam PTF

40

Loam
Loam PTF
Sand
Sand PTF

30

Sandy loam

Sandy loam PTF

Silt loam
Silt loam PTF

20

PTF - Pedo transfer
function

10

0 0.01

0.1 1

10 6 10 100 1000 10,000 100,000

Soil suction (kPa)

Figure 5.107 SWCCs computed from grain-size distribution curves for variety of soil types
(from M.D. Fredlund, 2000).

the minimum particle size is too large, there is an absence
of smaller particles which resulted in the soil drying
prematurely.
The initial packing porosity n p is the only volume-mass
variable used in the estimation of the SWCC. The packing
porosity is important to the estimation of the SWCC and
two techniques have been proposed for its estimation: (i) a
statistical method and (ii) a neural net method.
The statistical method involved examining the frequency
distribution of the packing porosities for various textural
categories. A mean and variance value was computed for
the packing porosity for different soil types.
The use of a neural net appeared to improve the estimation
of the SWCC. The neural net has an algorithm that is trained
to respond to various input conditions. The effect of the
packing porosity is shown in Fig. 5.108 for sand and in
Fig. 5.109 for silt. These examples show that the packing

porosity influences the entire SWCC. Fig. 5.109 shows that
the estimated SWCC does not reach 100% saturation with
packing porosity n p = 0 . 36. This condition occurs when the

packing porosity falls too far below the actual porosity of
a soil.

5.12.14 Comparison of Various Pedo-Transfer
Function Methodologies
The M.D. Fredlund (2000) estimation technique can be com-
pared to other estimation models, namely, Arya and Paris
(1981), Scheinost et al., (1997), Rawls and Brakensiek (1985),
Vereecken et al., (1989), and Tyler and Wheatcraft (1989).
Arya and Paris (1981). The Arya and Paris (1981)
pedo-transfer function was originally developed from a mod-
est database and then extrapolated to larger databases. A
value of 1.38 is commonly accepted as a reasonable estimate
for the α variable. Later investigations by Arya et al. (1982)