http://wjst.wu.ac.

th Article

Laboratory Method for Estimating Water Retention Properties of

Unsaturated Soil
Sabri KANZARI1,2,*, Mohamed HACHICHA1 and Rachida BOUHLILA2
1
Laboratory of Valorization of non Conventional Water, National, Institute of Research of Rural
Engineering, Waters and Forestry of Tunis, Tunisia
2
Laboratory of Modeling of Hydraulic and Environment, National Engineering School of Tunis, Tunisia

(*Corresponding author; e-mail: [email protected])

Received: 17 January 2012, Revised: 18 April 2012, Accepted: 2 November 2012

Abstract
Soil hydraulic properties are necessary for modeling water flow and solute transport in the vadose
zone. However, direct measurement of these characteristics in field conditions is tedious, time-consuming
and expensive. In this study, a laboratory method was used to characterize soil water retention curves of
three soil samples in the region of Bouhajla (Central Tunisia). For experimental purposes, volumetric
water content and pressure head values were measured using the gravimetric method and Watermark
sensor, of a small disturbed soil core, respectively, during a drying cycle under the effect of evaporation.
The van Genuchten model was fitted to the measured retention curves with the RETC software to
determine residual water content (θr), saturated water content (θs) and the two shape parameters; α and n.
Strong correlations were found between the fitted and measured retention curves. The van Genuchten
model was also fitted to the retention curves measured by pressure chamber (as the reference method).
The results were evaluated by calculating the root mean square error (RMSE) and the geometric mean
error ratio (GMER). Statistical analysis proved the success of the proposed method for estimating van
Genuchten soil retention parameters of the studied soils. A Mann-Whitney test performed at the
significance level of 0.05 showed no significant difference between the two methods.
Keywords: Evaporation method, parameter optimization, RETC, soil retention curve, Bouhajla

Introduction
Salinization risk assessment of soils and pedotransfer functions (PTFs) [1,2] or determined
aquifers in arid and semi-arid regions requires in the laboratory [3,4], which allow higher spatial
knowledge of the evolution of water movement and temporal resolution. The evaporation method
and solute transport in subsurface flow. During the is one of the most widely and easily used methods
last decades, a large number of numerical models to determine the retention curve and hydraulic
have been developed for the simulation of water conductivity of unsaturated soils. This method is
flow and solute transport in the unsaturated zone. based on measuring both soil moisture and
Nevertheless, their use in field conditions is often pressure head during a soil drying cycle under the
limited by the lack of characterization of retention effect of evaporation. The method developed by
properties. In situ field measurements of soil Wind [5] introduced an iterative graphical
retention properties are tedious, costly, time procedure to estimate, firstly, the water retention
consuming and are not accurate because of curve from average soil moisture and pressure
experimental shortcomings and high spatial and head readings, and to determine hydraulic
temporal variability. Therefore, the retention conductivities from measured pressure head profile
properties of unsaturated soils are often estimated and variations in water content distribution.
indirectly from other soil properties using Although, in general, five tensiometers were used

Walailak J Sci & Tech 2012; 9(4): 361-367.

Water Retention of Unsaturated Soils Sabri KANZARI et al
http://wjst.wu.ac.th

in measurements ranging from −50 cm to at least an evaporation laboratory method. This method is
−850 cm, in evaporation methods, several authors to monitor the water content by the gravimetric
have proposed to reduce the number of method and the pressure head by a Watermark
tensiometers to 2 [6-8]. However, Wessolek et al. sensor which allows a wider measurement range
[9] and Simunek et al. [10] have used only one than the conventional tensiometer (0 cm to −1,990
tensiometer in small soil cores and showed that cm [13,14], during a drying cycle of a small soil
this method is able to accurately estimate soil container under the effect of evaporation and using
hydraulic characteristics. Furthermore, as an the RETC program to determine the van
alternative to the Wind Algorithm, the analysis of Genuchten model parameters from measured
water flow during an evaporation experiment can retention curves.
be performed by using optimization algorithms.
The RETC software [11] which is based on the Material and methods
Levenberg-Marquardt optimization algorithm is
Laboratory experiment
often used for estimating soil hydraulic parameters
Three soils presented in Table 1 were
by fitting water retention and hydraulic
sampled from a land parcel (35°15’47.58’’N;
conductivity models to measured data.
10°4’17.16’’E) in the village of Bouhajla (Central
The overall objective of our study is the
Tunisia). The soil samples were crushed and then
numerical simulation of water movement and salts
placed in small clear plastic containers (6 cm
transfer in Bouhajla (Central Tunisia),
diameter / 12 cm long). A Watermark sensor
characterized by saline soils [12], to try to assess
(Irrometer Inc., USA) was implanted in the middle
groundwater contamination risk. The specific
of each soil layer (Figure 1).
objective of this paper is to estimate soil water
retention properties of three soils from Bouhajla by

Table 1 Soil particle size analysis of the three soils.

Soil Clay (g kg-1) Silt (g kg-1) Sand (g kg-1) Texture (USDA*)

1 9.5 4.5 85 Sand
2 12.5 30 57 Loam
3 35 5 60 Clay
* Scheme: United States Department of Agriculture

Figure 1 Schematic and photo of soil laboratory experiment.

362 Walailak J Sci & Tech 2012; 9(4)

Water Retention of Unsaturated Soils Sabri KANZARI et al
http://wjst.wu.ac.th

Each soil was saturated from the top with where θr is the residual water content [L3 L-3], θs
distilled water and was left to evaporate. During 3 -3
the drying cycle, no device was used to accelerate is the saturated water content [L L ], h is the
evaporation. Monitoring volumetric water content water pressure head [L], α [L-1] and n [-] are shape
was performed by a gravimetric method (weighing parameters.
scale) and the pressure head by a dielectric method Eq. (1) contains four independent
(Watermark sensor). Upon conversion of coefficients, represented by the vector b = {θr, θs,
gravimetric water content to volumetric humidity, α, n}. The different parameters are essentially
the values of bulk density were measured using the empirical coefficients without much physical
cylinder method [15]. The measurements were significance [17]. Their values were determined by
made daily until the digital meter indicated h = fitting the retention model to the observed data
using the parameter optimisation RETC software
−1,990 cm which corresponds to the Watermark
[11]. This program uses Marquardt’s maximum
sensor limit.
neighbourhood method to minimize the objective
function, O(b):
Parameter estimation
The van Genuchten model [16] was used to 2
set the water retention curve θ (h) , which relates
N
 ∧

min Ob (b) = ∑  θ − θ
 i i ( b)   (2)
the volumetric water θ [L-3L-3] content to pressure i =1  
potential h [L]. This function is
∧

 θs − θr where θi and θi are the observed and fitted

θ r + m
h<0 water contents, respectively, and N is the number
 1 + αh n
θ ( h) =  ;
of retention data. Initial values for the soil
θ hydraulic parameters θr, θs, α and n were estimated
h ≥ 0
 s with the ROSETTA [18] pedotransfer function
 using measured values of sand, silt, and clay
content (Table 1 and 2).
1
m = 1− ,n >1 (1)
n

Table 2 Initial values of van Genuchten soil retention parameters estimated by Rosetta.

Soil θr (cm3 cm-3) θs (cm3 cm-3) α (cm-1) n (-)

1 0.0547 0.372 0.0298 1.8898
2 0.0466 0.3870 0.0213 1.4091
3 0.0762 0.3859 0.0272 0.0272

Pressure chamber was used to validate the values obtained by the

Soil samples were placed in a pressure proposed laboratory method. Soils samples were
chamber. The same pressures measured by left 48 h in the pressure chamber. Pressure was
Watermark sensors were applied to soil cores. For changed successively on the same sample each 48
each value of pressure (from 0 cm to −1,990 cm), h.
water content was measured gravimetrically. The
mass of soil sample was determined by subtracting Statistical analysis
the mass of the container and the probe from total To evaluate retention curves measured by the
weight. The pressure chamber (reference method) proposed laboratory method, two statistical
parameters were used: the root mean square error

Walailak J Sci & Tech 2012; 9(4) 363

Water Retention of Unsaturated Soils Sabri KANZARI et al
http://wjst.wu.ac.th

(RMSE) and the geometric mean error ratio whether there is a significant difference between
(GMER). These statistical parameters are the proposed method and the pressure chamber
calculated as follows: method or not.

j Results
∑ (L i − Pi )²
During the drying cycle, the pressure head
RMSE = i =1 (3)
(h) reduced from a saturated point (h = 0 cm) to a
j
partially saturated point (h = −1,990 cm) for all the
 1 j  Li 
GMER = exp  ∑ ln  (4) soils. However, the values of volumetric water
 j i =1  Pi  content (θv) at the saturation and at the dry end are
different between the soils. Sandy soil has the
lowest values of θv at saturation and at the end of
where Li is the value measured by the laboratory drying than other soil samples. Soil water retention
method, Pi is the value measured by the pressure increased with soil enrichment by fine particles
(clay + silt). Soil water retention curves of clay
chamber, P is the average value of pressure soil showed the higher values of θv at saturation
chamber data and j is the number of and at drying and dry while loamy soil showed an
observations. The RMSE and the GMER when intermediate moisture state between them and the
equal to 0 and 1, respectively, correspond to an surface layer.
exact match between observed and fitted data. The Measured water retention data obtained from
GMER value less or greater than 1 indicates that the laboratory method were fitted by RETC
the corresponding model underestimates or software to determine the van Genuchten equation
overestimates fitted data. The smaller (closer to 0) parameters (Figure 2). Strong correlations were
the RMSE value was, the better the model was. found between the measured and fitted curves, the
Statistical processing was achieved by the correlation coefficient R² ranged between 0.87 and
STATISTICA software, Version 5 (Statsoft 0.97. The values of van Genuchten’s equation
France, 1997). The non-parametric Mann-Whitney parameters and the values of the objective function
test was also performed at the significance level of O(b) are presented in Table 3.
0.05 (test is significant at p < 0.05) to find out

Table 3 Estimated van Genuchten soil retention proprieties and values the objective function from the
laboratory method.

Soil θr (cm3 cm-3) θs (cm3 cm-3) α (cm-1) n (-) r² O(b)10-4

1 0.01678 0.25305 0.0104 2.62368 0.97 38.20
2 0.0093 0.39429 0.00854 1.71022 0.87 69.90
3 0.20436 0.41129 0.03364 2.40498 0.98 185.30

364 Walailak J Sci & Tech 2012; 9(4)

Water Retention of Unsaturated Soils Sabri KANZARI et al
http://wjst.wu.ac.th

Figure 2 Measured and fitted soil retention curves of Bouhajla unsaturated soils.

The values of these parameters (θr, θs, α and water movement and solute transport in the
n) are very heterogeneous between the different unsaturated zone of Bouhajla.
soils. Soils 1 and 3 have the highest values of n Good agreement between the fitted and
and α, the most sensitive parameters to water flow measured (by the pressure chamber) retention
[19]. These layers may be particular areas for curves are shown in Figure 2. The RMSE and

Walailak J Sci & Tech 2012; 9(4) 365

Water Retention of Unsaturated Soils Sabri KANZARI et al
http://wjst.wu.ac.th

GMER calculated for the different soil samples are three soils: sand, silt and clay were, respectively,
close to 0 and 1, respectively, as shown in Table 4. 0.8302, 0.5493 and 0.1761 above 5 %. So there is
The GMER values were greater than 1 meaning no significant difference between the proposed
that the proposed laboratory method may slightly laboratory method and the reference method
overestimate the soil water retention curve. (pressure chamber) for measuring the soil water
Calculated values of the Mann-Whitney test of the retention curve.

Table 4 Statistical analysis of measured retention curve and pressure chamber values.

Soil RMSE GMER

1 0.1126 1.15
2 0.0882 1.03
3 0.1189 1.26

Discussion recommendations will be taken into consideration

in future work.
The evaporation method is a widespread
experimental method for estimating soil hydraulic
Conclusions
properties. In this research, we have demonstrated
the success of estimating soil retention parameters A simple evaporation method was advanced
by RETC from an evaporation experiment on in this study for estimating soil water retention
small soil cores and using a single sensor for proprieties of an unsaturated zone. The obtained
measuring the pressure head in line with the work estimation results are acceptable and have shown
of Simunek et al. [10] and Wessoleck et al. [9], that the van Genuchten retention curve parameters
and we have extended the range of measurement to are different from one layer to another. However,
−1,990 cm. Estimated values of van Genuchten these results allowed us to get an idea of the range
equation parameters, especially α and n, are of the of these parameters for each soil layer, especially
same order as the parameters estimated by other for the shape parameters α and n. These results are
authors using the evaporation method, for instance essential for modeling of water flow and salts
Bruckler et al. [20] and Fujimaki and Inoue [6] for transfer in the Bouhajla vadose zone.
sandy loam soils and Basile et al. [21] for sandy
clay soils. However, the hydraulic conductivity Acknowledgements
curve has not been determined assuming that it can
This study was done under the support of
be estimated from the equation of van Genuchten
INRGREF/ACSAD project “Technology Transfer
[16]. Simultaneous estimation of the retention
of Brackish/Saline Waters for Farmers” and the
curve and unsaturated hydraulic conductivity using
PISEAU project “Salinity Management for Small
the Wind algorithm [5] could be an interesting
Farmers and Long-term Salinization Risk in
perspective of this study. According to Abbasi et
Central Tunisia”. We thank Mr. Selim BEN
al. [1] salinity indirectly affects soil hydraulic
HAMADI for his assistance in the lab
properties by acting on the porosity and
measurements. We also would like to thank the
permeability, the study of the effect of salts on
valuable comments of the anonymous reviewers,
these properties is recommended.
which helped us to improve the manuscript.
Finally, the Levenberg-Marquardt
optimization algorithm implemented in RETC
References
presents some difficulties to optimize certain
parameters of water content and hydraulic [1] Y Abbasi, B Ghanbarian-Alavijeh, AM
conductivity from collected data in field conditions Lighat and M Shorafa. Evaluation of
(Wesseling et al.) [4], use of other optimization pedotransfer functions for estimating soil
methods is also suggested. All these

366 Walailak J Sci & Tech 2012; 9(4)

Water Retention of Unsaturated Soils Sabri KANZARI et al
http://wjst.wu.ac.th

water retention of saline and saline-alkali U.S Department of Agriculture, Agricultural

soils of Iran. Pedosphere 2011; 21, 230-7. Research Service Riverside, 1991, p. 117.
[2] B Ghanbarlan-Alavijeh, A Liaghat, GH [12] S Kanzari, M Hachicha, R Bouhlila and J
Huang and MTh van Genuchten. Estimation Battle-sales. Characterization and modeling
of the van Genuchten soil water retention of water movement and salts transfer in a
properties from soil textural data. semi-arid region of Tunisia (Bou Hajla,
Pedosphere 2010; 4, 456-65. Kairouan) - Salinization risk of soils and
[3] J Simunek. Estimating soil hydraulic aquifers. Comput. Electron. Agric. 2012; 86,
parameters from transient flow experiments 32-42.
in a centrifuge using parameter optimization [13] UK Singh, L Ren and K Shaozhong.
technique. Water Resources Res. 2005; 41, 1- Simulation of water in space and time using
9. an agro-hydrological model and remote
[4] JG Wesseling, CJ Ritsema, J Stolte, K sensing techniques. Agr. Water Manage.
Oostindie and K Dekker. Describing the soil 2010; 97, 1210-20.
physical characteristics of soil samples with [14] B Cardenas-Lailhacar and MD Dukes.
cubical splines. Transport Porous Med. Precision of soil moisture sensor irrigation
2008; 71, 289-309. under field conditions. Agr. Water Manage.
[5] GP Wind. Capillary Conductivity Data 2010; 97, 666-72.
Estimated by a Simple Method. In: RE [15] C Mathieu and F Pieltain. Analyse Physique
Rijtema and H Wassink (eds.). Water in the des Sols, Méthodes Choisies. In: Lavoisier
Unsaturated Zone, Proc. UNESCO/IASH (ed.). Tec et Doc, Paris, 1998, p. 275.
Symp. Wageningen, The Netherlands, 1968, [16] MTh van Genuchten. A closed-form equation
p. 181-91. for predicting the hydraulic conductivity of
[6] H Fujimaki and M Inoue. A transient unsaturated soils. Soil Sci. Soc. Am. J. 1980;
evaporation method for determining soil 44, 892-8.
hydraulic properties at low pressure. Vadose [17] JB Kool, JC Parker and MTh van Genuchten.
Zone J. 2003; 2, 400-8. Determining soil hydraulic properties from
[7] U Schindler, W Durner, G von Unold and L one-step outflow experiments by parameter
Muller. Evaporation method for estimating estimation: I. theory and numerical studies.
measuring unsaturated hydraulic properties Soil Sci. Soc. Am. J. 1985; 49, 1348-54.
of soils: extending the measurement range. [18] MG Schaap, FJ Leij and MTh van
Soil Sci. Soc. Am. J. 2010; 74, 1071-83. Genuchten. Rosetta: A computer program for
[8] H Schelle, SC Iden, A Peters and W Durner. estimating soil hydraulic parameters with
Analysis of the agreement of soil hydraulic hierarchical pedotransfer functions. J.
properties obtained from multistep-outflow Hydrol. 2001; 251, 163-76.
and evaporation methods. Vadose Zone J. [19] Z Lu and D Zhang. Stochastic analysis of
2010; 9, 1080-91. transient flox in heteregeneous variably
[9] G Wessolek, R Plagge, FJ Leij and MTh van saturated porous media: the van Genuchten-
Genuchten. Analysing problems in Mualem Constitutive Model. Vadose Zone J.
describing field and laboratory measured soil 2002; 1, 137-49.
hydraulic properties. Geoderma 1994; 64, [20] L Bruckler, P Bertuzzi, R Angulo-Jaramillo
93-110. and S Ruy. Testing and infiltration method
[10] J Simunek, O Wendrorth and MTh van for estimating soil hydraulic properties in the
Genuchten. Parameter estimation analysis of laboratory. Soil Sci. Soc. Am. J. 2002; 66,
the evaporation method for determining soil 384-93.
hydraulic properties. Soil Sci. Soc. Am. J. [21] A Basile, A Coppola, R De Mascellis and L
1998; 62, 894-905. Randazzo. Scaling approach to deduce field
[11] MTh van Genuchten, FT Leji and SR Yates. unsaturated hydraulic properties and
The RETC Code for Quantifying the behavior from laboratory measurement on
Hydraulic Functions of Unsaturated Soils. small cores. Vadose Zone J. 2006; 5, 1005-
16.

Walailak J Sci & Tech 2012; 9(4) 367