SSSAJ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 22 August 2006
Published in Soil Sci Soc Am J 70:1774-1787 (2006)
DOI: 10.2136/sssaj2005.0335
© 2006 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Crescimanno, G.
Right arrow Articles by Garofalo, P.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Crescimanno, G.
Right arrow Articles by Garofalo, P.
Agricola
Right arrow Articles by Crescimanno, G.
Right arrow Articles by Garofalo, P.
Related Collections
Right arrow Irrigation
Right arrow Solute Transport Models
Right arrow Sustainable Management of the Vadose Zone

Soil & Water Management & Conservation

Management of Irrigation with Saline Water in Cracking Clay Soils

Giuseppina Crescimanno* and Paolo Garofalo

Università di Palermo, Dipartimento ITAF–Sezione Idraulica, Viale delle Scienze, 13 90128 Palermo, Italy

* Corresponding author (gcrescim{at}unipa.it)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY and conclusion
 REFERENCES
 
Management scenarios aimed at optimizing irrigation in a Sicilian vineyard characterized by a cracking clay soil irrigated with saline water were explored for seven soil profiles (Baglio 1–Baglio 7), by using the simulation model soil-water-atmosphere-plant environment (SWAP), which accounts for shrinkage and cracking. Accurate prediction of water content, {theta}, was obtained for the seven profiles by expressing the soil hydraulic properties according to the Brutsaert retention equation coupled with the hydraulic conductivity model proposed by Gardner (B-G model). A satisfactory prediction of the electrical conductivity of saturated extract (ECsat) was obtained using for the dispersivity (Ldis), a calibration value of 20 cm. Different irrigation schedulings and alternating waters of different quality were then explored as viable management options. The results showed that bypass flow determined a favorable water distribution, and that the best irrigation strategy was to make a minimum number of irrigations, by maximizing at the same time the amount of water supplied at each irrigation. Water storage in cracks was found to promote salt-leaching; neglecting cracks and bypass flow was shown to overestimate salinization. Alternating two different irrigation waters proved to be the best strategy, which could be adopted to reduce soil salinization and enhance crop transpiration. Findings concerning the role of cracks in the process of salt-leaching suggested that, under field conditions, application of a leaching solution was more efficient if the soil presented a considerable degree of cracking.

Abbreviations: ADE, advection-dispersion equation • B-G, Brutsaert-Gardner • COLE, coefficient of linear extensibility • DW, Durbin–Watson test • ECsat, electrical conductivity of saturated soil extract • ECw, electrical conductivity of irrigation water • ESP, exchangeable sodium percentage • Ldis, dispersivity • LF, leaching fraction • MSTEP, multi-step • RMSR, root mean squared residual • SAR, sodium adsorption ratio • SCIM, suction crust infiltrometer method • SWAP, Soil-Water-Atmosphere-Plant environment


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY and conclusion
 REFERENCES
 
THE NEED to protect environmental quality has modified the original goal of agricultural producers to maximize yield. Farmers have now more constraints on their management decisions because they have now to face environmental issues as well as production factors. Scientists should provide technical information to guide farmers and policymakers in making decisions that optimize the dual goal of high crop yield and prevent environmental degradation.

In arid regions, irrigation with saline waters, often a consequence of intensive agricultural systems, is one of the main causes of secondary salinization (Szabolcs,1994), resulting in soil degradation (UNEP, 1991). Although accurate worldwide data are not available, vast areas of irrigated land are increasingly threatened by salinization (Crescimanno et al., 2004). The effects of salinity are manifested in loss of land, reduced rates of plant growth, reduced yields, and, in severe cases, total crop failure (Rhoades and Loveday, 1990). Threshold relationships between the ECsat and crop yield have been empirically determined for several crops and can be used to evaluate the influence of saline irrigation waters on agricultural production (Maas and Hoffman, 1977). Salt composition of soil water also influences the composition of cations in the exchange complex of soil particles, affecting soil structural and hydraulic properties (Crescimanno et al., 1995).

Leaching is considered the basic management tool for controlling salinity. Water is applied in excess of the total amount used by the crop and lost by evaporation. The strategy is to keep the salts in solution and flush them below the root zone. The amount of water needed is referred to as the leaching requirement or the leaching fraction, LF (Hoffman, 1990). However, application of the LF can be performed only if water is a nonlimiting factor. If a limited water supply is available, only appropriate irrigation scheduling can be applied to prevent salinization. This means selecting amount and timing of irrigation to make optimal usage of water for the crop and minimize salt-accumulation in the root zone.

When at least two qualities of water exist for irrigating, blending, or cyclic irrigation can be employed to keep salinity under levels compatible with thresholds not affecting crop productivity (Maas and Hoffman, 1977). Grattan and Rhoades (1990) pointed out the many advantages of selecting the cyclic irrigation option, which include: no blending facility would be required, more salt-sensitive plants could be included in the rotation, and soil salinity could be reduced at critical times of physiological growth. In addition, the intermittent leaching that takes place under this strategy can be more effective at leaching salts than continuous leaching, that is, imposing a leaching fraction at each irrigation (Shalhevet, 1984).

Swelling/shrinking clay soils change volume (V) with changes in water content, and during dry periods extensive cracks will form in the field (Bronswijk, 1989). Soil cracks alter the pore-size distribution through intermittent wetting, acting as significant pathways for water and solutes and determining the occurrence of bypass flow, that is, the rapid transport of water and solutes via shrinkage-cracks to subsoil and to groundwater through an unsaturated soil matrix (Beven and Germann, 1982; Bouma, 1991).

Although a considerable number of field-scale studies have shown the feasibility of using the cyclic irrigation strategy on agricultural crops (Ayars et al., 1986; Grattan et al., 1987; Rhoades, 1989; Rhoades et al., 1989; Sharma and Rao, 1998; Schaan et al., 2003), only a few papers investigating the efficiency of cyclic strategies on cracked, clay soils can be found in the literature. Crescimanno et al. (2002), performing laboratory experiments, found that alternating waters with different salinity was effective in determining salt-leaching, and that application of the leaching solutions was more effective if the soil presented a considerable degree of cracking. Tanton et al. (1988), performing field experiments, evidenced that the process of leaching of solutes in clay soils occurred through macropores/cracks, concluding that the macropores structure must be improved if saline clays are to be reclaimed. These results suggested that cracks play a significant role in the process of salt-leaching.

In Sicily, the increasing scarcity of good quality waters coupled with intensive use of soil under semiarid to arid climatic conditions results in irrigation with saline waters in clay soils having a high shrink-swell potential and susceptibility to cracking (Crescimanno and Provenzano, 1999). These soils are irrigated in the summer season, when cracks are open, by sprinkler systems that involve high application rates. Because of these high application rates, bypass flow is prevalent during irrigation (Crescimanno, 2001). Laboratory investigations performed on undisturbed soil columns from these areas showed that salinization or leaching occurred during bypass flow depending on the concentration of the solution applied (Crescimanno and De Santis 2004). Instead, the low values of the Sodium Adsorption Ratio (SAR) of irrigation water, and the low values of exchangeable sodium percentage (ESP) measured in these soils indicated no risk of sodification under the current conditions. These results suggest that, under these conditions, management strategies accounting for cracking and bypass flow should be adopted to prevent salinization and land degradation.

Obtaining viable management options through field research is difficult and expensive because of the number of variables to consider. Simulation models (Simunek et al., 2003) can be used to evaluate the consequences of changes in plant and soil properties or in irrigation strategies, and these provide a more concrete basis to assist agronomists and supporting services. However, the utility of this approach requires that the models adequately depict the real situation.

van Dam et al. (1997) developed a model for fine-textured clay soils containing shrinkage cracks. This model, named SWAP, takes into account shrinking and swelling as a function of the water content. The model assumes that water and solutes can move instantaneously to specified bypass depths once the infiltration capacity of the soil matrix is exceeded by rainfall rate and a critical depth of water has formed at the soil surface (Verburg et al., 1996). SWAP provides as output the water content, {theta}, (and pressure head, h), as well as ECsat (Rhoades, 1996). Reduction in crop yield is calculated as a function of ECsat (Maas and Hoffman, 1977).

Crescimanno and Garofalo (2005) tested the applicability of the SWAP model for prediction of water content ({theta}) and ECsat in a Sicilian cracking clay soil irrigated with saline water. Using {theta} measurements collected from four profiles (Baglio1, Baglio2, Baglio3, and Baglio4) located in a Sicilian vineyard, they found that using the parameter estimation method based on multi-step outflow experiments, and representing the soil hydraulic properties by the Brutsaert retention equation, coupled with the hydraulic conductivity model proposed by Gardner (B-G model), it was possible to obtain an accurate prediction of {theta}.

With reference to solute transport simulated by SWAP, they found that a satisfactory prediction of ECsat could be obtained by calibrating the model with reference to the Ldis parameter, which represents the dispersivity in the advection-dispersion equation (ADE) (Warrick, 2003). Using field measurements of ECsat from the Baglio1 to Baglio4 soil profiles, they found that although different mean ECsat values were measured in the four profiles during the simulation period, almost the same Ldis value of about 20 cm was found as the calibration value. They concluded that "if confirmed for other soils, this calibration procedure would provide an effective dispersion coefficient that reflected the complexities of the flow pathways and heterogeneity in local fluid velocities in the flow direction (Beven et al., 1993; Forrer et al., 1999)." This result could have significant practical implications because a value of Ldis obtained by calibrating SWAP on a limited number of field sites could be used to predict salinization for other sites located in the same field.

The main objective of this paper was to explore management strategies optimizing irrigation, and also reducing the risk of secondary salinization, in a Sicilian vineyard where irrigation with saline water is increasingly practiced, and land degradation represents an ever greater environmental hazard. Being the soil in this vineyard characterized by a considerable susceptibility to shrinkage and cracking, the SWAP model, which takes into account a variable soil volume, was applied to test different management options.

Another objective of this paper was to investigate whether the value of 20 cm previously found for Ldis by calibrating SWAP on four soil profiles (Baglio1–Baglio4) (Crescimanno and Garofalo, 2005), could be applied to accurately predict ECsat for the Baglio5, Baglio6, and Baglio7 profiles, also located in the previously considered vineyard, without calibration.

After checking that accurate prediction of {theta} and ECsat was provided by SWAP, two viable options addressing constraints of limited water availability were simulated for the seven soil profiles. These options were (i) different irrigation schedulings, that is, irrigation with a fixed amount of water but different number of irrigations, and (ii) cyclic strategies, that is, alternating two irrigation waters having different salinity. In addition, to evaluate the influence of cracks on salinization, and to check the consequences of neglecting shrinkage and cracking on prediction of salinization, some additional simulations were performed for the same profiles using SWAP under the assumption of no shrinkage and cracking.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 SUMMARY and conclusion
 REFERENCES
 
Theory of SWAP
One-dimensional, vertical, transient, unsaturated flow in the SWAP model (van Dam et al., 1997) is described by the Richards equation, which is solved numerically. The shrinkage characteristic is expressed by the model proposed by Kim (1992). The shrinkage characteristic allows the calculation, at a certain soil depth or node i, of the relative cross-sectional area of the cracks at the soil surface, Ac (m2 m–2) (Bronswijk, 1989). The model calculates a crack volume if {theta}i is lower than the water content corresponding at the beginning of the structural shrinkage.

The matrix and crack infiltration at a given rainfall intensity, P (cm h–1), are calculated as follows:

Formula 1[1a]

Formula 2[1b]

Formula 3[2a]

Formula 4[2b]
where Imax is the maximum infiltration rate of the soil matrix, (cm h–1), Ic is the infiltration rate into the cracks, (cm h–1), Am and Ac (m2 m–2) are the relative areas of soil matrix and cracks, respectively.

Solute transport in the model is described with the ADE, (Warrick, 2003):

Formula 5[3]
where {theta} (m3 m–3) is the volumetric water content, c is the solute concentration (g m–3) in soil water, t (s) is the time, q is the soil water flux (positive upward) (m s–1), z is the vertical coordinate with the origin at the soil surface (positive upward) (m) and D is the apparent diffusion coefficient (m2 s–1).The root water uptake is semi-empirically described by a sink term, which is a function of the maximum root water uptake, the soil water pressure head and the salt concentration.

The maximum possible root water extraction rate, integrated over the rooting depth, is equal to the potential transpiration rate, Tm (m d–1), which is governed by atmospheric conditions. The potential root water extraction rate at a certain depth, Sp(z) (d–1), may be determined by the root length density, Iroot(z) (m m–3), at this depth as fraction of the integrated root length density:

Formula 6[4]
where Droot is the root layer thickness (m). Stresses due to dry or wet conditions and/or high salinity concentrations may reduce Sp(z). The water stress in SWAP is described by the function proposed by Feddes et al. (1978).

For salinity stress the response function of Maas and Hoffman (1977) is used. SWAP assumes water and salinity stress to be multiplicative. This means that the actual root water flux, Sa(z) (d–1), is calculated from:

Formula 7[5]
where {alpha}rw (-) {alpha}rs (-) are the reduction factors due to water and salinity stresses, respectively, and Sp(z) (d–1) is the potential root water extraction rate at a certain depth.

The {alpha}rs reduction factor is variable between 0 and 1 depending on ECsat:

Formula 8[6]

Formula 9[7]
where EClim is the threshold salinity value, that is, the ECsat level below which there is no salt stress, and EC% is the percentage decrement value for unit increase of salinity in excess of the threshold. For grapes, EClim is equal to 1.5 dS m–1, and EC% is equal to 9.6%.

At the bottom of the system, boundary conditions can be described with various options, e.g., water table depth, flux to ground water or free drainage.

Field and Soil Characteristics
Data collection was performed in a 25 by 25 m field located in Sicily (37° 40' 55''N; 12° 38' 50'' E) where irrigation with saline waters is performed on grapes by a sprinkler system, which allows high application rates at the soil surface. Irrigation water is taken from the Trinità artificial reservoir. The electrical conductivity of irrigation water, ECw, is about 2.1 dS m–1. However, when rainfall is particularly low, and water stored in this reservoir is not enough to cover irrigation needs, water from wells is used for irrigation, with ECw values up to about 6.0 dS m–1.

Seven soil profiles (Baglio1—Baglio7) were considered in this field. Four of the seven profiles (Baglio1–Baglio4) had been described in a previous paper (Crescimanno and Garofalo, 2005). However, since scenarios were performed for all the seven profiles, the complete set of soil physical, shrinkage and chemical properties was reported in Table 1.


View this table:
[in this window]
[in a new window]
  		Table 1. Classification, physical and chemical properties, COLE and shrink-swell potential of the soils considered.

 
Soil Shrinkage and Hydraulic Characteristics
The soil shrinkage curve was determined by measuring vertical and horizontal shrinkage on undisturbed soil cores (diameter d = 8.5 cm, height H = 11.5 cm) (Crescimanno and Provenzano, 1999). The shrinkage characteristic was expressed by the model proposed by Kim (1992; Crescimanno and Garofalo, 2005):

Formula 10[8]
where e = Vp/Vs is the void ratio, (m3 m–3), u = Vw/Vs is the moisture ratio, (m3 m–3) and {alpha}sh, ßsh, {gamma}sh are dimensionless fitting parameters; Vp is the total pore volume, Vs is the solid volume and Vw is the water volume.

Bulk density ({rho}b) was determined from the shrinkage curve and used to calculate the volumetric water content, {theta}, which accounted for a variable soil volume. The coefficient of linear extensibility, COLE (Grossman et al., 1968), indicating the shrink-swell potential (Parker et al., 1977), was also calculated.

The saturated hydraulic conductivity of the soil matrix, ksat, and some k(h) values close to saturation, were determined in soil columns (d = 20 cm, H = 20 cm), by the suction crust infiltrometer method, SCIM (Booltink et al., 1991). The soil hydraulic parameters/functions were determined by inverse method based on multi-step (MSTEP) outflow experiments performed on replicated soil cores (d = 8.5 cm, H = 5 cm). The saturated water content, {theta}s, was assumed to be equal to the water content value measured by a hanging water column apparatus (Burke et al., 1986) at a pressure head value of –2 cm. This {theta}s value was found to be consistent with the calculated porosity [{theta}s = us/(1+es)]. The MSTEP experiments were performed in pressure cells by applying three successive steps with pneumatic pressures ranging from 10 to 40 cm, from 40 to 70 cm, and from 70 to 800 cm (Crescimanno and Iovino, 1995). After the MSTEP experiments, the cores were put in a pressure plate apparatus to measure the water content at h = –15300 cm, that is, wilting point, {theta}wp. Independent measurement of {theta}s and {theta}wp was necessary because only the pressure range from 10 to about 1000 cm can be explored in the pressure cells used for the MSTEP.

Parameter estimation was performed according to Crescimanno and Baiamonte (1999), representing the soil hydraulic functions by:

— the equation proposed by Brutsaert (B) (1966), for the water retention curve:

Formula 11[9]

— coupled with the model proposed by Gardner (1958) (G) for the hydraulic conductivity function k(h):

Formula 12[10]

where h (cm) is the pressure head, {theta}s is the volumetric water content at saturation, {theta}r is the residual water content, k is the unsaturated hydraulic conductivity (cm h–1), ksat is the saturated hydraulic conductivity (cm h–1), {alpha}', n', ß, and {lambda} are empirical parameters. The hydraulic model represented by Eq. [5] and [6] (B-G model), which couples a {theta}(h) function with a closed-form equation not derived from the {theta}(h) function, was demonstrated to provide accurate estimation of the {theta}(h) and k(h) functions (Crescimanno and Baiamonte, 1999; Crescimanno and Garofalo, 2005). Optimization was performed on the outflow volumes, supplemented by four {theta}(h) values obtained during the MSTEP experiments ({theta} values at –10, –40, –70, and –800 cm) and by the k(h) values obtained by the SCIM method. Parameter estimation was performed by fixing both {theta}s and ksat, at the measured values. Optimized parameters were therefore {theta}r, {alpha}', n',{lambda}, and ß.

Accuracy of Predicted {theta} and ECsat
Gravimetric water content, U, was determined on undisturbed soil cores sampled at 30 and 45 cm in the selected profiles at different dates (from 14 July 1998 to 31 Dec. 2000). U and {rho}b(U) were used to calculate the volumetric water content, {theta}.

Soil saturated extracts were prepared from the soil collected in the field at the same dates as those sampled for determining U. Soil electrical conductivity, ECsat, was determined on these extracts by a conductivimeter (Crison, Micro CM 2002).

The accuracy of the predicted {theta} values was evaluated by calculating the root mean squared residual, RMSR{theta}, between measured and predicted {theta}:

Formula 13[11]
where N is the number of measurements.

To check systematic errors between measured and predicted {theta}, the predicted {theta} values were regressed against the measured values, and the hypotheses (i) that the slope (b) of the regression line was not significantly different from 1, and (ii) that the intercept (a) of the regression line was not significantly different from 0, were checked. The Durbin-Watson (DW) test was used to check if the random errors in the regression line exhibited autocorrelation.

With reference to solute transport, we checked if a good match between measured and predicted ECsat could be obtained for the Baglio5 to Baglio7 profiles using for Ldis the calibration value of 20 cm previously found for the Baglio1 to Baglio4 profiles (Crescimanno and Garofalo, 2005). The RMSRECsat between measured and predicted ECsat, was calculated by Eq. [11]. The predicted ECsat values were regressed against the measured ECsat, and the hypotheses (i) and (ii) were checked. The DW test was also performed.

Management Scenarios
Climatic data (rain intensity, maximum, and minimum temperature, rainfall height) recorded daily from 8 July 1998 to 31 Dec. 2000 by a rain gauge located in the field were used as input in SWAP. Annual rainfall in 1998, 1999, and 2000 was 390 mm in average and the annual reference evapotranspiration in 1998, 1999, and 2000 was 1450 mm in average. Although an annual amount of irrigation water equal to 120 mm is supplied under normal conditions, due to the constraints of limited water availability, the annual irrigation amount supplied from 1998 to 2000 was very low, and equal to 66 mm in 1998, to 48 mm in 1999, and to 24 mm in 2000. The irrigation season in this vineyard usually ranges from mid June to mid September.

A root distribution characterized by 60% roots in the 30- to 70-cm layer, and by 20% both in the 0- to 30-cm and in the 70- to 100-cm layers, was assumed (Crescimanno and Garofalo, 2005). Simulations were performed by using a bottom boundary condition of freely draining profile, and the B-G hydraulic parameters were used to simulate water transport.

The following management scenarios were considered:

– Scenario 1—Irrigation scheduling. Irrigation with a fixed annual volume of 1120 m3 ha–1 or 112 mm, and electrical conductivity of irrigation water equal to 6.2 dS m–1 (the most critical possible salinity value), but testing different options in terms of number of water applications, that is: 1a: eight water applications, which means weekly irrigation; 1b: four water applications, which means irrigation every 2 wk; 1c: two water applications, which means a monthly water application. The 1c is the irrigation scheduling more often used in this irrigated area.

To explore how cracks may affect the process of salt-accumulation and/or leaching, scenario 1c was repeated under the hypothesis of no shrinkage, which means no cracking and bypass flow. This scenario was indicated with 1c'.

– Scenario 2c—Cyclic strategy. Irrigation with a fixed annual volume of 1120 m3 ha–1 or 112 mm, but alternating two waters of different salinity. The saline irrigation water is the one used in Scenario 1, the less saline water, with ECw = 2.1 dS m–1, which is the value measured during the winter season, is used when the crop is more sensitive to salinity according to the crop physiology.

A performance indicator (Smets et al., 1997) was used to evaluate the impact of management scenarios on salinization:

Formula 14[12]
where Si and Sf (mg cm–2) represent the quantity of salts accumulated in the soil profile at the starting date and to the end of simulation, respectively.

To compare the different scenarios in terms of crop transpiration, and of evaporation, the following ratios were calculated:

Formula 15[13]

Formula 16[14]
where Tscen (cm) and Escen (cm) were the actual crop transpiration and evaporation of the considered scenario, and T1c (cm) and E1c (cm) represent transpiration and evaporation obtained by scenario 1c, which is the commonly applied irrigation scheduling in the irrigated area.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
SUMMARY and conclusion
 REFERENCES
 
Soil Shrinkage and Hydraulic Characteristics
A good fit of the Kim model to the experimentally measured values of void ratio, e, and moisture ratio, u, was obtained for the Ap horizon of the Baglio5 profile (Fig. 1 ). The same good fit, not shown for brevity's sake, was found for the other soil horizons and profiles. This result confirmed the suitability of the Kim model to accurately represent the soil shrinkage curve (Crescimanno and Garofalo, 2005). The COLE values calculated for the different horizons made it possible to classify the soils as having a shrink-swell potential (Parker et al., 1977) from medium to very high (Table 1). This indicated that the soils had a considerable susceptibility to cracking.


Figure 1
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 1. Shrinkage characteristic obtained for the Ap horizon of Baglio5 profile. The continuous line represents the Kim model fitted to the measured (u, e) values.

 
The soil hydraulic parameters obtained according to the B-G model were reported in Table 2. The parameter estimation procedure based on the B-G model provided an estimated k(h) function in close agreement with the k(h) values measured by the SCIM (Fig. 2b , Ap horizon of the Baglio5 profile). A good agreement was also observed between the predicted {theta}(h) function and the measured ({theta}, h) values obtained by the MSTEP experiments (Fig. 2a, Ap horizon of the Baglio5 profile). The water retention curve predicted by the B-G model accurately matched the water content independently measured at –15300 cm, that is, wilting point, {theta}wp (Fig. 2a). This demonstrated the good prediction of the water retention function at the lowest {theta} values. A good match between the measured {theta}wp and that predicted using the B-G parameters was observed for all the other profiles and horizons. The similar results obtained for the Baglio 6 and 7 profiles, not reported for brevity's sake, perfectly agreed with those previously obtained on the Baglio1 to Baglio4 profiles (Crescimanno and Garofalo, 2005). This confirmed the suitability of the B-G model to accurately represent the hydraulic properties of the clay, structured soils considered in this investigation.


View this table:
[in this window]
[in a new window]
  		Table 2. Hydraulic parameters determined by the parameter estimation method, using the Brutsaert retention equation coupled with the hydraulic conductivity equation proposed by Gardner (B-G model).

 

Figure 2
View larger version (15K):
[in this window]
[in a new window]
  		Fig. 2. (Baglio5 soil profile, Ap horizon). Water retention curve (a) obtained by the parameter estimation method using the B-G model, including the SCIM measurements in the optimization procedure and fixing ksat. {theta}wp is the water content independently measured at h = –15 300 cm. Hydraulic conductivity function (b) obtained by the parameter estimation method using the B-G model including the SCIM measurements in the optimization procedure and fixing ksat.

 
Accuracy of Predicted {theta} and ECsat
The predicted {theta}, using SWAP with the B-G parameters as input, was in close agreement with that measured (Fig. 3 , Baglio5 profile). The good match between the predicted {theta} and the lowest measured {theta} ({theta} ~ 0.31 m3 m–3 in the Ap horizon, and {theta} ~ 0.32 m3 m–3 in the A1 horizon, Fig. 3), confirmed that a reliable prediction of the water content was obtained using the B-G model in the {theta} range from saturation to wilting point. The low values of the RMSR{theta} between measured and predicted {theta} (Table 3), indicated the good prediction of {theta}. The a and b parameters of the equation found by regressing the predicted {theta} against those measured (Table 3) were not significantly different from 0 and 1 respectively at the 0.05 probability level when the B-G model was used. The DW values (Table 3) showed that there was neither positive nor negative autocorrelation and that it was therefore possible to exclude internal dependence of errors. Similar results were obtained for the Baglio5 and Baglio6 profiles, which confirmed those previously reported by Crescimanno and Garofalo (2005), indicating that the accuracy of {theta} predicted by SWAP depended on the use of soil hydraulic properties, which reflected the hydraulic behavior of the soils considered. This is consistent with previous results (Schaap and Leij, 2000) and confirms that parameters in soil hydraulic functions characterizing water retention and hydraulic properties are the most important input variables for models based on numerical solutions of the variably saturated flow (Richards) equation. Since no calibration was performed to adjust the estimated hydraulic properties, the good match between the measured and the simulated {theta}, obtained using the B-G soil hydraulic properties, also proved that the parameter estimation procedure adopted provided a reliable estimation of the {theta}(h) and k(h) functions.


Figure 3
View larger version (27K):
[in this window]
[in a new window]
  		Fig. 3. Baglio5 soil profile: daily volumetric water content, {theta}, predicted by SWAP at (a) 30 cm and at (b) 45 cm.

 

View this table:
[in this window]
[in a new window]
  		Table 3. Parameters indicating the agreement between measured and predicted {theta}.

 
With reference to prediction of ECsat, the values of the root mean squared residual, RMSRECsat, between measured and predicted ECsat (4), obtained for Baglio5 to Baglio7 using the Ldis = 20 cm previously obtained by calibration on the Baglio1 to Baglio4 profiles, were comparable with those previously found for the Baglio1 to Baglio4 profiles (Crescimanno and Garofalo, 2005). The a and b parameters of the equation found by regressing the ECsat predicted against those measured (Table 4) were not significantly different from 0 and 1 respectively, at the 0.05 probability level, both at 30 and at 45 cm. This indicated that no systematic errors were associated with the predicted ECsat. The DW statistic (Table 4) also proved that the random errors in the estimated ECsat were independent (the significance level was always 0.05), excluding internal dependence of errors. An example of the good match between simulated and predicted ECsat can be observed in Fig. 4a and 4b for the Baglio5 profile, at 30 and 45 cm, respectively.


View this table:
[in this window]
[in a new window]
  		Table 4. Parameters indicating the agreement between measured and predicted ECsat.

 

Figure 4
View larger version (23K):
[in this window]
[in a new window]
  		Fig. 4. Baglio5 soil profile: daily electrical conductivity of saturated soil extract, ECsat, predicted (a) at 30 cm and (b) at 45 cm assuming for Ldis the value of 20 cm.

 
To further evaluate prediction of ECsat provided by SWAP, the RMSRECsat values obtained for the Baglio5 to Baglio7 profiles were compared with the ECsat variation determining a variation of 5% in crop yield, which is equal to 0.521 dS m–1 for grapes (Maas and Hoffman, 1977). The RSMRECsat values, reported in Table 4, were always lower than this value. As a consequence, the predictive errors associated with the simulated ECsat could be considered acceptable. These results confirmed that the calibration procedure previously adopted for determining Ldis (Crescimanno and Garofalo, 2005) provided an "effective" dispersion coefficient, which reflected the complexities of the flow pathways and heterogeneity in local fluid velocities in the flow direction (Beven et al., 1993; Forrer et al., 1999). The significant practical implication of this result is that for long-term simulation of solute transport on a large scale, ECsat measurements collected from a limited number of field sites could be used to find the calibration value of Ldis, and this value could be used to make predictions for other points located in the same field. However, these results were obtained for a non sodic soil, under a condition of SAR of irrigation water close to the soil ESP, and consequently sodium in the solution and in the exchange complex should be almost in equilibrium. Under this condition, the simplifying assumption that the salts are not adsorbed to soil solids could be considered acceptable. Prediction of ECsat provided by SWAP should be carefully checked when irrigation is performed on sodic soils, or when sodication can be the consequence of using irrigation waters with SAR higher than soil ESP (Crescimanno and De Santis, 2004).

Management Scenarios
Irrigation Scheduling (Scenario 1)
Decreasing {Delta}S values were obtained for the seven profiles passing from Scenario 1a to Scenarios 1b and 1c (Fig. 5 ). This result can be explained by the fact that reducing the number of irrigations, and increasing the amount of water applied, determined a higher application intensity (I). Since in all the three scenarios irrigation was performed in summer (starting date June 15), when cracks were open and the hydraulic conductivity (HC) of the soil matrix was low, bypass flow of water was prevalent. According to Eq. [1] and [2], at increasing I, an increasing amount of water, and of dissolved salts, bypasses the upper layers, rapidly reaching the bottom layers. This is the reason why, when irrigation was performed according to Scenario 1c, a higher percentage of water and salts bypassed the surface layers compared with Scenarios 1a and 1b.


Figure 5
View larger version (21K):
[in this window]
[in a new window]
  		Fig. 5. Amount of solutes accumulated, {Delta}S (mg cm–2), in the seven soil profiles according to the three considered irrigation schedulings (Scenario 1).

 
Concerning relative transpiration, the highest RT values were associated with Scenario 1c (Table 5). This was not only a consequence of the lower {Delta}S, which determined a lower {alpha}rs in Eq. [5], but was also the consequence of water content distribution in the profile after irrigation (Fig. 6 , Baglio1 profile). As can be seen in the figure, {theta} values higher than those obtained by the 1a and by the 1b scenarios were obtained by Scenario 1c in the 5- to 60-cm layer, where the maximum percentage of roots was concentrated, 1 d after irrigation. This water distribution was the consequence of bypass flow, which promoted storage of water in the deepest layers. Consistently with this water distribution (Fig. 6), the lowest evaporation was also obtained in Scenario 1c, as demonstrated by the RE values (Table 6).


View this table:
[in this window]
[in a new window]
  		Table 5. Ratio (RT) between transpiration (T) obtained by scenarios 1a, 1b, 1c, and 2c, and T obtained by scenario 1c. Actual transpiration provided by scenario 1c.

 

Figure 6
View larger version (23K):
[in this window]
[in a new window]
  		Fig. 6. Water distribution along the soil profile 1 d after irrigation (Baglio1 profile).

 

View this table:
[in this window]
[in a new window]
  		Table 6. Ratio (RE) between evaporation (E) obtained by scenarios 1a, 1b, 1c and 2c, and E obtained by scenario 1c. Actual evaporation provided by scenario 1c.

 
The water distribution observed 1 d after irrigation (Fig. 6) was confirmed in the course of the whole irrigation season, as can be seen in Fig. 7 (Baglio1 profile), where the average volumetric water content, {theta}m, is represented as a function of time. The highest {theta}m values were associated with Scenario 1c, which justify the higher RT and the lower RE obtained in this scenario. For clarity's sake, only the {theta}m calculated during the irrigation season of the year 2000 (from June to end of September) was represented in Fig. 7. However, similar results were obtained for the 1999 as well as for the other profiles.


Figure 7
View larger version (17K):
[in this window]
[in a new window]
  		Fig. 7. Average volumetric water content, {theta}m (m3 m–3), during the irrigation season of year 2000 according to Scenario 1 (Baglio1 profile).

 
These results indicated that when a limited annual or seasonal amount of saline irrigation water can be provided to the crop, irrigation scheduling may considerably affect salinization and transpiration. Under our conditions, reducing the number of irrigations, and increasing the irrigation amount at each application, proved to be the best strategy to prevent salinization, also enhancing crop transpiration. However, this result can be considered valid only for crops with deep root distribution. On the contrary, for crops with roots concentrated in the upper layers, bypass flow could be an unfavorable process, owing to the water distribution causing higher {theta} values in the deepest layers.

The Role of Cracks in the Process of Salinization and Salt-Leaching (Scenario 1c')
To evaluate the influence of cracks on salinization, and to check the consequences of neglecting shrinkage and cracking on selection of irrigation strategies preventing salinization, Scenario 1c was repeated with the assumption of no shrinkage and cracking, that is rigid soil (Scenario 1c').

The {Delta}S values obtained by Scenario 1c' (Fig. 8 ) were always higher than those obtained by Scenario 1c, in which cracks were taken into account. Since the only difference between Scenarios 1c and 1c' was that in this latter scenario the soil was considered as nonshrinking, with no cracks, the lower {Delta}S obtained by Scenario 1c certainly depended on the fact that the cumulative water flow from the cracks into the matrix, CWF (cm) (Fig. 9 ), was taken into account. Significantly different CWFs were found for the different profiles, with the lowest value for Baglio3, and increasingly higher values for Baglio1, Baglio6, Baglio5, Baglio2, and Baglio7 (in the order of increasing CWF). A significantly higher CWF was found for Baglio4. It is interesting to notice that cracks differently affected the difference in {Delta}S between Scenarios 1c and 1c'. For Baglio1 and Baglio3 this difference (diff1 = {Delta}S1c – {Delta}S1c') was negligible; for the other profiles, the effect of the cracks was more pronounced, especially for Baglio2 (diff1 = 8.0 mg cm–2), Baglio7 (diff1 = 8.2 mg cm–2), and Baglio4 (diff1 = 16.6 mg cm–2). The increasing order of diff1 corresponded to a decreasing order of CWF (Fig. 9).


Figure 8
View larger version (16K):
[in this window]
[in a new window]
  		Fig. 8. Amount of solutes accumulated in the soil profiles, {Delta}S (mg cm–2), in Scenarios 1c and 1c'.

 

Figure 9
View larger version (20K):
[in this window]
[in a new window]
  		Fig. 9. Cumulative water flow from the cracks into the matrix, CWF (cm), obtained for the different profiles.

 
To understand the relationship between diff1 and CWF, the shrinkage and hydraulic properties of the different profiles were analyzed. If we compare Baglio3 and Baglio1 (lowest CWFs) with the other profiles in terms of shrinkage behavior, calculating the crack volume as a percentage of volume at saturation (Vcr/V), and representing it vs. matric potential (h) (Fig. 10 ), we can see that for Baglio3 and Baglio1 (Fig. 10b) structural shrinkage is evident, with a null crack volume for h ranging between 0 and –300 cm, and with a Vcr/V of about 4% at h = –15300 cm. On the contrary, for the other profiles (Fig. 10a), structural shrinkage was evident only between h = 0 and h = –30 cm, with a higher Vcr/V, ranging from 6% to about 8%, at h = –15300 cm. As a consequence, the higher diff1 values corresponded to the profiles with the highest CWFs, that is, with more susceptibility to cracking. This is confirmed by analysis of the overall flux of solutes leaving the bottom profile, CWQ (Fig. 11 ). In the figure, the flux of solutes leaving the matrix (CWQm), which depends on HC of soil matrix, can be distinguished from the flux of solutes from the cracks (CWQcr), which depends on shrinkage and cracking.


Figure 10
View larger version (30K):
[in this window]
[in a new window]
  		Fig. 10. Crack volume as percentage of volume at saturation, Vcr/V (%), vs. pressure head (a) for Baglio2, Baglio4, Baglio5, Baglio6, and Baglio7, and (b) for Baglio1 and Baglio3.

 

Figure 11
View larger version (35K):
[in this window]
[in a new window]
  		Fig. 11. Overall flux of solutes, CWQ (mg cm–2 d–1), from the soil profiles, flux of solutes leaving the matrix (CWQm), and the cracks (CWQcr).

 
Comparison between Baglio4 and Baglio2, for which the maximum CWQs were obtained, showed that a higher CWQcr was obtained for Baglio4 than for Baglio2. Since comparable CWQm values were found for Baglio4 and Baglio2, which reflected the similar k(h) values in the range close to saturation (Fig. 12 , A1 horizon), the higher diff1 observed for Baglio4 was the consequence of the higher cracks' contribution of this soil to the process of solute leaching compared with that obtained for Baglio2. This demonstrated that cracking significantly contributed to the process of salt-leaching, and was in agreement with previous results obtained by laboratory experiments in soil columns. These experiments indicated that application of a leaching solution to a salinized, cracking soil, determined a greater efficiency of salt-leaching if the soil presented a considerable degree of cracking (Crescimanno et al., 2002).


Figure 12
View larger version (22K):
[in this window]
[in a new window]
  		Fig. 12. Hydraulic conductivity functions, k (cm h–1), obtained for the seven profiles for the A1 horizon.

 
Comparing results of Scenarios 1c and 1c' on the different soil profiles suggested that if cracks and water storage in cracks were not taken into account, the risk of salinization was overestimated. The magnitude of this overestimation depended on the soil shrinkage behavior, being greater at increasing shrinkage and cracking. For soils showing a considerable susceptibility to shrinkage and cracking, management strategies optimizing irrigation should therefore be obtained by physically based models taking into account a variable soil volume.

Cyclic Strategies (Scenario 2c)
With reference to alternating two waters of different salinity, expressed as ECw, (first water with lower salinity, then water with higher salinity) (Scenario 2c), the {Delta}S values obtained by Scenario 2c (Fig. 13 ) were significantly lower that those provided by Scenario 1c for all seven profiles. Negative {Delta}S values, indicating salt-leaching, were obtained only for Baglio4 and Baglio2; negligible {Delta}S values, indicating no solute accumulation, for Baglio1; and positive and increasingly higher {Delta}S, indicating salt-accumulation, for Baglio7, Baglio3, Baglio5, and Baglio6 (in order of increasing {Delta}S). The negative {Delta}S values found for Baglio4 and Baglio2 were certainly due to the highest CWQs (Fig. 14 ); the difference (diff2 = {Delta}S1c – {Delta}S2c) between the {Delta}S values obtained with Scenarios 1c and 2c decreased from 30.40 mg cm–2 (Baglio4) to 26.90 mg cm–2 (Baglio6), following the same order in which CWQ decreased (Fig. 14).


Figure 13
View larger version (14K):
[in this window]
[in a new window]
  		Fig. 13. Amount of solutes accumulated in the soil profiles, {Delta}S (mg cm–2), according to Scenarios 1c and 2c.

 

Figure 14
View larger version (33K):
[in this window]
[in a new window]
  		Fig. 14. Overall flux of solutes, CWQ (mg cm–2 d–1), from the soil profiles in the 2c scenario.

 
The efficiency of cyclic strategy was due to the lower amount of salts added to the soil alternating the water with lower salinity to the water with the highest salinity (Rhoades, 1989), and to the fact that irrigation was performed in July-August, when the soil was dry and cracking was significant. Since cracks promoted salt-leaching, as discussed in the previous paragraph, this result was consistent with previous investigations performed under laboratory conditions on soil columns from the same vineyard, which showed the favorable influence of cracks on salt-leaching determined by cyclic strategies (Crescimanno et al., 2002; Crescimanno and De Santis, 2004). Similar results had been found by Tanton et al. (1988) under field conditions.

Higher RT values corresponded to Scenario 2c compared with those obtained by Scenario 1c (Tab 5), which can be explained by the lower average values of ECsat in the soil profile (Fig. 15 ). According to Eq. (6) and (7), a lower ECsat determines a higher root water flux, Sa. The same RE values (Table 6) were found in Scenarios 1c and 2c. The reasons for this are that RE was not influenced by salinity, and that Scenarios 1c and 2c determined the same water distribution along the profile.


Figure 15
View larger version (17K):
[in this window]
[in a new window]
  		Fig. 15. Average electrical conductivity of the saturated extract, ECsat (dS m–1), vs. time (Baglio1 profile).

 

   SUMMARY and conclusion
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY and conclusion
REFERENCES
 
Application of the SWAP model to three soil profiles located in a Sicilian vineyard characterized by a cracking, clay soil confirmed results obtained on four previously considered profiles located in the same field, indicating that the SWAP model provided a satisfactory prediction of {theta} when the soil hydraulic characteristics were represented by the B-G model.

With reference to solute transport, application of SWAP to the three soil profiles, without calibration, using for Ldis the determined calibration value of 20 cm, led to predicted ECsat in good agreement with those measured. This result confirmed that the calibration procedure previously adopted to determine Ldis using ECsat measurements from the Baglio1 to Baglio4 profiles, provided an effective dispersion coefficient, which reflected the complexities of the flow pathways and heterogeneity in local fluid velocities in the flow direction. This finding could be significant as in field application of SWAP it could be sufficient to calibrate this parameter using measurements from a limited number of profiles to simulate solute transport for other profiles of the same soil.

Although this paper did not evaluate the accuracy of SWAP in simulating crackings dynamics, the good match between measured and predicted values of {theta} and of ECsat, for these three additional profiles, as well as for the four profiles previously considered, indicated that SWAP accurately simulated water and solute transport in cracking soils. This means that, if properly used, SWAP is suitable to explore management scenarios optimizing irrigation, and preventing salinization, in cracking soils.

Analysis of three different irrigation schedulings evidenced that the best scheduling was to make a limited number of irrigations (two in our case), using larger application volumes. This result was found to depend on the water distribution in the soil profile, which in turn depended on bypass flow of water, determined by the higher water application intensity involved in this scenario. As a consequence in our case, bypass flow was a mechanism determining a favorable water distribution. However, this result can be considered valid only for crops developing a deep root distribution. The best irrigation scheduling is therefore a function of soil type and crop characteristics, and the best option is to be found by simulations taking specific site conditions into account.

With reference to the role of cracks in the process of salt-leaching (Scenario 1c'), the simulations performed indicated that water stored in cracks promoted leaching of the accumulated solutes, and that neglecting the presence of cracks led to overestimating salinization. This overestimation was significant for the soils having a considerable susceptibility to shrinkage and cracking. When irrigation is performed in cracking soils, simulation models taking into account cracks should therefore used to explore sustainable irrigation strategies.

Cyclic strategy proved to be the best management option to be suggested to reduce the risk of salinization (Scenario 2c). Findings concerning the role of cracks in the process of salt-leaching suggested that, under field conditions, application of a leaching solution was more efficient if the soil presented a considerable degree of cracking.

A considerable variability in the amount of solutes accumulated in the different profiles was found in relation to the shrinkage and hydraulic behavior of the profiles considered. This stresses the importance of using shrinkage and hydraulic parameters/functions reflecting the soil physical behavior when SWAP is applied for predictive and/or management purposes.


   ACKNOWLEDGMENTS
 
Research supported by MURST (Rome, Italy), PRIN 2003: "Strategie per l'utilizzazione di acque salino/sodiche in terreni argillosi con crepacciature" The helpful comments and suggestions of two anonymous reviewers are gratefully acknowledged.

Received for publication October 6, 2005.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY and conclusion
 REFERENCES