Abstract

The project ‘Research and Training on Irrigation with Saline Water in Tunisia’, with the project report published by UNESCO in 1970, was set up to specify Tunisian standards for the use of saline water and to mitigate their effects on crop yields. The objective of this study is to assess the risk of long-term soil salinization by considering the agricultural practices mentioned in the project for the ‘Water Quality’ experiment in the semi-arid region of Cherfech (Tunisia). A Hydrus-1D model was used to simulate the movement of water and the transfer of salts. Soil hydraulic and solute transport parameters were estimated using inverse modeling. Calibration and validation of the model were made for the water and salt profiles carried out with four irrigation qualities QA, QB, QC and QD. Four scenarios over 50 years were studied: (i) S1 with rainfall (400 mm) only; (ii) S2 with rainfall and irrigation (1,400 mm); (iii) S3 with a 2 °C increase in temperature; (iv) S4 with the addition of 100 mm·d−1 of rainfall. The four scenarios highlighted the high risk of soil salinization, especially for the QB, QC and QD qualities after 20 years of irrigation and the deep dynamics of water and salts beyond the root zone which increases the risk of groundwater salinization.

HIGHLIGHTS

  • Saline water in Tunisia is a significant constraint on development.

  • The UNESCO project report Research and Training on Irrigation with Saline Water in Tunisia of 1970 is the only local reference for decision makers.

  • Inverse modeling is an essential tool to estimate soil transport parameters.

  • Assessment of soil salinization under different climate change scenarios.

  • Use of saline water based on the project practices can lead to the destruction of agricultural soils after 20 years.

INTRODUCTION

The scarcity of water and its poor quality in addition to soil salinization are considered to be the major challenges that affect the environment in irrigated areas. Increasingly, the world's water resources are suffering due to climate change, pollution, exploitation and poor management of water supplies (Hu et al. 2017; Qi et al. 2018). As a result, more than one-sixth of the world's land is affected by degradation and deforestation and 6.4% of land is said to be affected by salinization or alkalization (FAO 2015).

Under the effect of the poor quality of irrigation water due to the lack of water resources and the poor drainage of the soil, soil salinity influence increases from one year to another. Land salinization is a major problem worldwide (Besharat et al. 2020; Phogat et al. 2020). It already affects at least 400 million hectares and seriously threatens an equivalent area (Shrivastava & Kumar 2015); 3.2 million hectares are salinized to varying degrees of severity. In Africa 80.5 million ha are affected by salinity. This problem is more present in arid and semi-arid countries because they are characterized by a lack of precipitation and also strong evaporation which implies the concentration of salt in water and the salinization of soil (Singh 2019; Sun et al. 2019). In Tunisia, approximately 80% of water resources are saline waters and in addition approximately 35% of agricultural soils are salted. To address the problems of soil salinization which affect its physical and chemical properties and therefore the crop yields, applied research in the management of saline water for irrigation has been a national priority (Kanzari et al. 2018). One of the first major projects was the research and training on irrigation with saline water in Tunisia from 1962 to 1969 (UNESCO 1970). Several experimental stations were set up in the north, center and south. The main objective of this project was to apprehend the effect of irrigation with saline water on crop production. Despite the international renown of the project report (UNESCO 1970) and its results, whose impact is perfectly transferable to regions similar to Tunisia, very few studies were done to exploit its results. Indeed, two references can be found based on the ‘Water Quality’ experiment carried out at Cherfech (Tunisia):

The objective of this study is to assess the long-term risk of soil salinization in the Cherfech region by considering the agricultural practices recommended in the project (UNESCO 1970) for the ‘Water Quality’ experiment under different climate change scenarios.

MATERIALS AND METHODS

Region, soil and irrigation water

The Cherfech region (36°57′23.0″N 10°02′45.0″E) is located northwest of the city of Tunis and belongs to the irrigated district of the lower valley of the Medjerda River. The latter is the main river in Tunisia and its salinity range is between 0.87 g·l−1 and 3 g·l−1. The soils of this region have generally three distinct levels: a higher clay-loam level 60 to 80 cm thick, a medium silty-clay to loamy level, 40 to 60 cm thick, and a deep clay-loam to clay level. These soils have bulk density of 1.5 g·cm−3 to 1.6 g·cm−3.

In this study, the field experiment called ‘Water Quality’ is detailed. Four irrigation water qualities (QA, QB, QC, QD) were used on four groups of parcels where there were the same agricultural rotations, and the salt concentration of each water quality was:

  • – QA: 0.2 g·l−1;

  • – QB: 1.4 g·l−1;

  • – QC: 2.5 g·l−1;

  • – QD: 3.6 g·l−1.

The same irrigation regimes were practiced and the same farming techniques were applied. Water supplies by irrigation reached, on average, 1,000 mm, to which was added the precipitation (400 mm), which occured mainly in winter.

GetData Graph Digitizer, developed by Fedorov (2008), is a program for digitizing graphics and figures. It is used to extract the original data (x, y) from scanned graphs in the absence of values and numerical data. This software was used to obtain numerical data concerning the variation of soil water profiles and soil salt profiles.

The extracted data from UNESCO (1970) are:

  • – bulk density from page 260;

  • – soil water profiles from page 339;

  • – soil salt profiles from page 281 to page 289.

The authors will send the full report (UNESCO 1970) upon demand.

Modeling approach by Hydrus-1D

Theory

Water Flow
Variably saturated water flow in an unsaturated soil is commonly described using the Richards equation (Richards 1931):
formula
(1)
where t is time and z is depth (positive upward), and θ and h denote the volumetric water content and the soil water matric pressure head, respectively. The Richards equation can be solved numerically when the initial and boundary conditions are prescribed and two constitutive relations, i.e., the soil water retention, θ(h), and hydraulic conductivity, K(h), functions, are prescribed. The soil water retention curve is described using the van Genuchten analytical expression (van Genuchten 1980):
formula
(2)
The hydraulic conductivity function is described using the capillary model of Mualem (1976):
formula
(3)
where:
  • θr and θs are the residual and saturated volumetric water contents, respectively;

  • Se is the effective saturation;

  • m = 1 − (1/n);

  • Ks is the saturated hydraulic conductivity;

  • l is a pore connectivity coefficient;

  • α and n are shape parameters.

Solute transport
Solute transport in variably saturated soil is described using the advection–dispersion equation:
formula
(4)
where C is the solute concentration, θ is the soil volumetric water content, R the retardation factor, D the dispersion coefficient, and υ is the pore water velocity. The retardation factor R and the dispersion coefficient D are defined as: and , where Kd is the linear adsorption distribution coefficient, ρb the bulk density, λ is the longitudinal dispersivity and is the absolute value of the pore water velocity.
Parameter optimization
The global approach of parameter estimation involves the minimization of an objective function that considers all deviations between the measured and simulated data, with the simulated results controlled by the adjustable parameters to be optimized (Inoue et al. 2000). The objective function OF(b) for the water flow and solute transport is given by:
formula
(5)
where C is the solute concentration; Wh and Wc are normalization factors for matrix pressure head and concentration, respectively, with each factor being inversely proportional to their measurement variance; N1 and N2 are the number of observations for matrix pressure head and concentration, respectively; and b is the vector of optimized parameters. The subscripts m and o refer to the measured and optimized values. The weighted least-squares estimator is a maximum-likelihood estimator as long as the weights contain the measurement error information of particular measurements. The objective function for simultaneous optimization of soil hydraulic properties and solute transport parameters combines objective functions from Equation (5).

The Levenberg–Marquardt algorithm (Marquardt 1963) incorporated in the Hydrus-1D model (Šimůnek et al. 2016) was used to minimize the objective function OF(b). The parameter vector b includes the parameters α, n, θr, θs, Ks, λ, and Kd. Each inverse problem was restarted several times with different initial estimates of optimized parameters and the run with the lowest value of the objective function was assumed to represent the global minimum.

Calibration and validation of the model

The calibration of the model was carried out over a period of three years (1966–1968) on the measured water profile in the report. Hydrus-1D (Šimůnek et al. 2016) calibration for the measured salt profile (A68) was performed for the QA quality. The validation of the model was carried out for the water profiles and salt profiles measured in intermediate periods for QA. The salt profiles measured for other water qualities (A68): QB, QC and QD, were also used to validate the model.

Long-term simulation and scenarios

Four scenarios were studied over a period of 50 years:

  • – S1: rainfall only was considered;

  • – S2: rainfall and irrigation were considered;

  • – S3: effect of an increase of 2 °C in the air temperature;

  • – S4: effect of exceptional rainfall of 100 mm·d−1.

The simulated soil salt profiles were calculated each ten years.

Statistical evaluation of the model

The modeling result was evaluated by two methods: graphically and statistically. In the graphical approach, the measured and simulated volumetric water content and soil salinity were plotted as a function of soil depth. The statistical approach involved the calculation of the normalized root mean square error (RMSE):
formula
(6)
where are the predicted values, are the measured values, is the average value of observed data and n is the number of observations. For a perfect fit between observed and simulated data, values of RMSE should equal 0.

RESULTS AND DISCUSSION

Water and salt dynamics

Soil water movement

According to the project's report (UNESCO 1970) only three soil water profiles were carried out in the parcel. A first was made in summer 1966 (01/07/1966), a second profile in winter 1967 (03/01/1967) and a last profile in winter 1968 (30/01/1968). Figure 1 shows the variation of these water profiles.

Figure 1

Measured soil water content profiles for each date.

Figure 1

Measured soil water content profiles for each date.

The water state starts from a relatively dry state, since the profile was measured in a summer period of high evaporation. Alternating between irrigation/rainfall and evaporation does not seem to have a significant effect on the other water profiles carried out in the wet period (winter). The soil humidity values varied between 25% and 28% during the three years of the study. The land parcels had been cultivated throughout the year; water consumption by roots can be the cause of this stabilization of the soil water state.

Salt movement

The soil salt profiles measured during the experiments carried out at Cherfech for the four different water qualities are illustrated in Figure 2. The variation of the salt profile has shown that:

  • – The initial salt profile is an unbalanced profile, the surface soil salinity (about 1 g·l−1) is much less than at depth, which exceeds 4 g·l−1; this can be explained by the capillary rise of salts from the bottom.

  • – An absence of desalination of the initial profile except for QA quality in the spring of 1967.

  • – Alternating salinization/desalination in the surface layer under the effect of irrigation/rainfall and evaporation; QB and QC soil salt profiles present an intermediate soil salinity state ranging between 2 g·l−1 and 5 g·l−1.

  • – A deep accumulation of salts which can reach 6 g·l−1 with QD quality.

Figure 2

Measured soil salt profiles for irrigation water quality (QA, QB, QC and QD) from spring (Sp) 1965 to autumn (A) 1968 along with corresponding summer (Su) and winter (W) for each year.

Figure 2

Measured soil salt profiles for irrigation water quality (QA, QB, QC and QD) from spring (Sp) 1965 to autumn (A) 1968 along with corresponding summer (Su) and winter (W) for each year.

The characterization of the salt profile highlights a significant risk of salinization in the soil surface and at depth which increases with the irrigation water quality: QA < QB < QC < QD.

Simulation of soil water and salt transport

Model calibration

Inputs

The simulation period is 579 days, from 01/07/1966 until 30/01/1968 with a daily time step. The profile is modeled for a depth of 150 cm, composed of two layers (0–40 cm and 40–150 cm) depending on the texture of the soil. The soil water profile (01/07/1966) and the soil salt profile (summer 1966) data for QA quality were used as initial data during the simulation. The upper conditions of the soil profile correspond to atmospheric boundary conditions (BC), with a surface layer where evaporation has been specified (Figure 3), and the lower limit is free drainage. In the absence of any information concerning the irrigation schedules, the values of the water supplies were split up at a daily time step. For solute transport, the boundary conditions were the type of concentration flux BC.

Figure 3

Mean values of the monthly reference evapotranspiration ET0 (mm·d−1).

Figure 3

Mean values of the monthly reference evapotranspiration ET0 (mm·d−1).

The soil hydraulic properties (α, n, θr, θs, Ks) were estimated by inverse modeling. The calibrated soil hydraulic parameters are shown in Table 1.

Table 1

Calibrated soil water hydraulic properties

Layer (cm)θr (cm3.cm−3)θs (cm3.cm−3)αnKs (cm·d−1)
0–40 0.0100 0.4000 0.0568 1.2145 28.39 
40–150 0.0918 0.4100 0.0261 1.4341 12.92 
Layer (cm)θr (cm3.cm−3)θs (cm3.cm−3)αnKs (cm·d−1)
0–40 0.0100 0.4000 0.0568 1.2145 28.39 
40–150 0.0918 0.4100 0.0261 1.4341 12.92 

The dispersion coefficients ‘Disp’ and the adsorption coefficients ‘Kd’ of the overall soil salinity were also estimated by inverse modeling. The initial and minimum value used is equal to 1 and the maximum value is equal to 100 during the optimization process. The final values of the solute transport parameters are assigned in Table 2. Meddahi et al. (1993) determined the values of the dispersion coefficient for the two layers as well as the values of Kd. These values are respectively 50 and 1.

Table 2

Calibrated solute transport parameters.

Layer (cm)Disp (cm2.d−1)Kd
0–40 6.37 5.58 
40–150 1.00 1.91 
Layer (cm)Disp (cm2.d−1)Kd
0–40 6.37 5.58 
40–150 1.00 1.91 
Calibration results

Figure 4 shows the measured and simulated soil water and salt profiles for the final date of the simulation (the 579th day) with the calibrated parameters. According to the graphical evaluation the values measured and simulated are very close (Figure 4). The model underestimates the measured data. The RMSE value for the soil water content is 7.1% and for the soil salinity is 4.6%. These values are very low and indicate the reliability of the model to reproduce the water and salt dynamics.

Figure 4

Measured and simulated soil profiles for water content and salt after model calibration for 579 days from 01/07/1966 until 30/01/1968.

Figure 4

Measured and simulated soil profiles for water content and salt after model calibration for 579 days from 01/07/1966 until 30/01/1968.

Model validation

The Hydrus-1D model was validated on the only available release date (winter 1966–1967) for soil water profiles. For the salt profiles the model has been validated for QA quality on the profiles carried out between the initial profile (summer 1966) and the final profile (winter 1967–1968) whose dates of output are the 63rd, 185th, 244th, 336th, and 428th day of the simulation. In addition, validation of the Hydrus-1D model was carried out for the other qualities QB, QC and QD on the final salt profiles of each experiment. The input parameters were retained during model validation except for the initial and the water quality specified according to each case.

Figure 5 shows that the simulated soil water content values are very close to the values measured for the output date winter 1966–1967. The RMSE value is 1.15%, which is very low and indicates the success of the model validation.

Figure 5

Measured and simulated soil water content profiles (winter 1966–1967).

Figure 5

Measured and simulated soil water content profiles (winter 1966–1967).

In the absence of water profiles measured during the studied period (1966–1968), the success of the validation made it possible to trace the dynamics of the water in the soil. Indeed, the variation in water profiles during the simulation period showed (Figure 6):

  • – in the top soil humidification/drying cycles under the effect of rainfall irrigation and evaporation;

  • – an important variation of soil water content at 150 cm in depth highlighting a deep dynamic of water.

Figure 6

Simulated soil water content profiles for the missing years from autumn 1966 (A66) to autumn 1967 (A67).

Figure 6

Simulated soil water content profiles for the missing years from autumn 1966 (A66) to autumn 1967 (A67).

The simulated salt profiles for the various release dates during the QA water quality experiment are fairly close to the soil salinity values measured except for the autumn 1967 date when the simulated salt profile underestimated the measured profile (Figure 7). The values are quite low and indicate the good performance of the model to reproduce the dynamics of the salts during the QA test except for the date autumn 1967 when the value of the error is high.

Figure 7

Measured and simulated soil salt profiles for QA water quality and the corresponding root mean square error (RMSE) values.

Figure 7

Measured and simulated soil salt profiles for QA water quality and the corresponding root mean square error (RMSE) values.

The Hydrus-1D model was evaluated for the numerical simulation of the variation in soil salinity during other ‘water quality’ experiments: QB, QC and QD. Figure 8 illustrates the simulated salt profiles for the winter date 1967–1968 (579th day of the simulation) and the measured profiles. The simulated soil salinity values are very close to those measured.

Figure 8

Measured and simulated soil salt profiles for QB, QC and QD water qualities.

Figure 8

Measured and simulated soil salt profiles for QB, QC and QD water qualities.

The RMSE statistical index calculated for the different water qualities used is 2.09% for QB, 5.32% for QC and 3.42% for QD. The values are low and demonstrate the good performance of the Hydrus-1D model for reproducing the dynamics of salts with the different water qualities.

Long-term evaluation of soil salinization risks

Calibration and validation of the Hydrus-1D model allowed the studying of scenarios (Karandish & Šimůnek 2018) in order to assess the risk of long-term salinization of soils with the agricultural practices of the project (UNESCO 1970).

The long-term simulation was carried out over 50 years (1966–2015). The World Bank database contains monthly values of rainfall and average monthly temperatures for most countries in the world for the period 1901–2015. The Thornwaite formula estimates the reference evapotranspiration values from the average temperatures. The comparison between the values of ET0-Thornwaite and ET0-UNESCO showed a good correlation between the two methods (see the Supplementary Material). The monthly ET0 and rainfall values re-recorded in the Supplementary Material were used as input parameters to the Hydrus-1D model for the long-term simulation over 50 years and for each scenario.

Figure 9 presents the long-term variation of the soil water content over a period of 50 years for each scenario studied and shows:

  • – The amounts of rainfall, which average 400 mm·year−1, have failed to restore the initial water content of the soil. Indeed, a constant drying of the water profile is noted under evaporation.

  • – The quantities of water added by irrigation, with an average of 1,000 mm·year−1, have oversaturated the soil at the surface while deep percolation is noted.

  • – The increase in temperature by 2°C has the effect of increasing evaporation, resulting in greater drying of the soil at the surface.

  • – The torrential rains of 100 mm·d−1 managed to maintain a water state close to saturation during the first 20 years and which gradually dried up. At depth, the water state has varied very little.

Figure 9

Long-term variation in soil water content over a period of 50 years for each scenario studied.

Figure 9

Long-term variation in soil water content over a period of 50 years for each scenario studied.

Figures 1013 illustrate the variation of soil salt profiles every ten years of the studied scenarios for each water quality. The variations of the soil salt profiles show that:

  • – In S1: The rainfall alone without irrigation has no effect on the variation of the soil salt profile for the four used water qualities; the soil salinity reaches a static equilibrium.

  • – In S2: The infiltration (rain/irrigation)/evaporation have a cyclic effect of salinization/desalination in the topsoil for QA and continuous desalination at depth. The soil salinity profile is reversed for QA. On the other hand, for the other water qualities, continuous salinization is noted for all the layers of the soil and more importantly in the root zone and which from the 20th year reaches enormous thresholds along with water quality. The worst case is QD in which soil salinity is near 60 g·l−1 with QA < QB < QC < QD.

  • – In S3: The increase in temperature of 2 °C slowed down the salinization process of the topsoil under the effect of evaporation and accelerated it at depth for the four qualities of water.

  • – In S4: The exceptional rainfall of 100 mm·d−1 increased the displacement of the salt amounts towards the deepest layers (Saâdi et al. 2018) and increased the salinization risks of the groundwater in all the studied water qualities.

Figure 10

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QA water quality.

Figure 10

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QA water quality.

Figure 11

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QB water quality.

Figure 11

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QB water quality.

Figure 12

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QC water quality.

Figure 12

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QC water quality.

Figure 13

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QD water quality.

Figure 13

Long-term variation in soil salinity over a period of 50 years for each studied scenario and for QD water quality.

CONCLUSION

Tunisia is among the arid and semi-arid countries which are characterized by a lack of rainfall and strong heatwaves. As a solution to water shortage, the decision makers were forced to seek other sources of water such as saline water, which potentially is quite important in Tunisia. The UNESCO (1970) project was carried out to study the effect of saline irrigation water on crop production. Despite the reputation of the results obtained within the framework of this project, the recommendations of which have been considered until today, the risk of soil salinization has not been detailed and has been limited to the four years of study.

The results of characterization of the soil water profiles and soil salt profiles carried out in the ‘water quality’ experiments showed a deep dynamic of the water and a cyclic and reversible salinization in the topsoil and an accumulation of salts especially with the poorest water qualities (QC and QD). The calibration and validation of the Hydrus-1D model allowed the study of scenarios for the evaluation of the risks of long-term soil salinization. Irrigation with doses of approximately 1,000 mm·year−1 with rainfall of 400 mm·year−1 revealed a deep water dynamic beyond 1.5 m, which has the consequence of a deep dynamic, too, of salts. The increase in temperature by 2 °C and rainfall confirmed these results, hence the increased risk of groundwater salinization. After about twenty years and in the case of all scenarios, the soil salinity reached very high values (>20 g·l−1) for qualities QB, QC and QD, which highlights the important risk of long-term application of the project recommendations in the absence of strategies to mitigate progressive soil salinization in the Cherfech region.

For mitigating the effect of the use of saline waters in irrigation according the recommendation of UNESCO (1970) it is suggested to use more efficient irrigation techniques such as trickle irrigation. Such a technique allows the control of irrigation doses and irrigation frequencies in order to reduce the stress induced by salt accumulation in the root zone. Also, it is recommended to bring additional water amounts to leach salt after the irrigation season and to use salt-tolerant crops.

DATA AVAILABILITY STATEMENT

All relevant data are included in the paper or its Supplementary Information.

REFERENCES

REFERENCES
FAO
2015
Status of the World's Soil Resources
.
FAO
,
Rome, Italy
, p.
364
.
Fedorov
S.
2008
GetData graph digitizer
. http://www.getdata-graph-digitizer.com.
Inoue
M.
,
Šimůnek
J.
,
Shiozawa
S.
&
Hopmans
J. W.
2000
Simultaneous estimation of soil hydraulic and solute transport parameters from transient infiltration experiments
.
Advances in Water Resources
23
(
7
),
677
688
.
Marquardt
D. W.
1963
An algorithm for least-squares estimation of nonlinear parameters
.
Applied Mathematics
11
(
2
),
431
441
.
Meddahi
M.-E.
,
Mallants
D.
,
Feyen
J.
&
Vereecken
H.
1993
Modélisation de l’évolution de la salinité dans la zone racinaire (Modeling of the evolution of salinity in the root zone)
.
Science du sol
31
(
1–2
),
59
76
.
Phogat
V.
,
Mallants
D.
,
Cox
J. W.
,
Šimůnek
J.
,
Oliver
D. P.
&
Awad
J.
2020
Management of soil salinity associated with irrigation of protected crops
.
Agricultural Water Management
227
,
105845
.
Richards
L. A.
1931
Capillary conduction of liquids through porous mediums
.
Journal of Applied Physics
1
,
318
333
.
Šimůnek
J.
,
van Genuchten
M. T.
&
Šejna
M.
2016
Recent developments and applications of the HYDRUS computer software packages
.
Vadose Zone Journal
15
(
7
). doi: 10.2136/vzj2016.04.0033.
Sun
G.
,
Zhu
Y.
,
Ye
M.
,
Yang
J.
,
Qu
Z.
,
Mao
W.
&
Wu
J.
2019
Development and application of long-term root zone salt balance model for predicting soil salinity in arid shallow water table area
.
Agricultural Water Management
213
,
486
498
.
UNESCO
1970
Research and Training on Irrigation with Saline Water
,
Technical Report of UNDP Project, Tunisia 5
.
UNESCO
,
Paris, France
.
Valles
V.
,
Bourgeat
F.
&
Guiresse
M.
1988
Calcul des doses d'irrigation pour les sols salés : application d'une méthode géochimique de calcul à un sol tunisien (Calculation of irrigation doses for salty soils: application of a geochemical calculation method to Tunisian soil)
.
Cahiers ORSTOM Série Pédologie
24
(
2
),
115
122
.
van Genuchten
M. T.
1980
A closed-form equation for predicting the hydraulic conductivity of unsaturated soils
.
Soil Science Society of America Journal
44
,
892
898
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Supplementary data