Infiltration of water in disturbed soil columns as affected by clay dispersion and aggregate slaking

Soil crusting negatively affects the productivity and sustainability of irrigated agriculture, reducing water infiltration and plant emergence, and enhancing surface runoff and erosion. Clay dispersion and slaking of the aggregates at the soil surface are the main processes responsible for crusting. The infiltration rates (IR) of ten arid-zone soils in disturbed soil columns were measured and their relative susceptibilities to dispersion and slaking were determined. It was also examined whether the final soil IRs (FIR) could be estimated from various soil stability indices. The susceptibility to chemical dispersion was determined by measuring the IR of soil columns slowly pre-saturated from below with tap water and subsequently ponded with deionized (DW), canal irrigation (CW), and gypsum-saturated (GW) waters. The susceptibility to aggregate slaking was determined by comparing the IR measured in pre-saturated (slow wetting from below) and air-dry (fast wetting) soil columns ponded with CW. The FIRs of most soils decreased in the order GW > CW > DW. Seven soils were susceptible to clay dispersion induced by DW. Five soils were susceptible to clay dispersion induced by CW. Only two soils were susceptible to slaking. The fast wetting in these soils completely sealed the soil surface, reducing their IRs to zero from the start of leaching. Clay dispersion rather than aggregate slaking was the principal process inducing sealing and decreasing IR in these soils when subject to low-salinity waters. The indices WSA (water stable aggregates), MDC (mechanically dispersed clay) and MWDstir (mean weight diameter of stirring aggregates after a prewetting treatment) gave consistent and significant relationships with FIRs.


Introduction
Chemical dispersion of clay particles and slaking or physical disintegration of soil aggregates are two of the main processes responsible for soil crusting.They release small particles that clogs the conducting pores immediately beneath the surface, developing disrupted layers or seals that form crusts upon drying.Crusted soils generally decrease infiltration rates, plant emergence and yield of crops, and increase surface runoff and erosion.
The relative importance of dispersion and slaking depends on various soil properties, particularly soil exchangeable sodium percentage (ESP), the rate of soil wetting and drying (i.e.irrigation management, and climatic factors such as rain and wind intensities), and the electrical conductivity (EC) of the applied water.Dispersion of soil clays is induced by low electrolyte concentrations (lower than the soil's flocculation value, FV) and high sodium adsorption ratio (SAR) and pH values in the soil and applied water (Shainberg and Letey, 1984).This mechanism is enhanced by the mechanical breakdown of aggregates and the subsequent exposure of new surfaces to chemical dispersion (slaking promotes dispersion) (Rengasamy et al., 1984;Sumner et al., 1998).The negative effects of clay dispersion, soil sealing and crusting on the infiltration rate of soils are well documented (Sumner and Stewart, 1992).
Slaking of aggregates is produced by the pressure build-up of entrapped air inside the micropores when dry soils are subject to rapid wetting.If the pressure is large enough to overcome the cohesive forces between the aggregates, they break into microaggregates of different sizes and strength, releasing the excess air pressure.Slaking is generally limited to the immediate soil surface (So and Cook, 1993), although the depth of slaking depends on aggregate size and stability (Collis-George and Green, 1979).Thus, Farres (1978) indicated that coarse aggregates are less prone to slaking because of the lower compression of the entrapped air.Several authors have suggested that the hydraulic properties of the seals and the susceptibility of soils to crusting depend on the size distribution of detached primary particles and/or aggregate fragments resulting from aggregate breakdown (Mullins et al., 1987;Roth and Eggert, 1994;Bresson, 1995;Le Bissonnais, 1996).
Slaking during wetting of some hardsetting soils may predominate over dispersion, as shown by the greater proportion of slaked fragments (20 to 60 µm) over clay particles (< 2 µm) (Young and Mullins, 1991).The physical disintegration of aggregates due to slaking may be more important than the mechanical effect of the raindrop impact (Le Bissonnais and Singer, 1992;Loch and Foley, 1994).Chan andMullins (1994) andCaron et al. (1996) found that cultivated soils were more prone to slaking and structural decline than uncultivated soils, and Ferruzzi et al. (2000) concluded that most cultivated soils in their studies were prone to slaking during the irrigation events.In summary, various internal and external factors affect the susceptibility of soils to slaking (Le Bissonnais, 1990).
Although the degree of slaking and dispersion of field-undisturbed and laboratory-disturbed soils may differ, laboratory measurements are useful for diagnostic and screening purposes.Infiltration rate (IR) measures the amount of water passing through the soil in a given time and is an indication of the air porosity and pore conductivity.When slaking and/or dispersion occur, seals develop, decreasing the porosity and the IR until steady-state infiltration is reached.Hence, surface sealing may be quantified using infiltration measurements obtained in disturbed soil columns.
The first objective of this work was to determine the susceptibility of ten arid-zone soils (i) to clay dispersion by measuring the IR of pre-wetted soil columns subject to waters of different electrolyte concentrations, and (ii) to aggregate slaking by comparing the IRs measured in pre-wetted (i.e., slow wetting) and airdry (i.e., fast wetting) soil columns.
Although IRs measured in disturbed soil columns are useful for the purposes already mentioned, they are time-consuming and difficult to perform in terms of reproducibility and accuracy (Abu-Sharar et al., 1987).It is desirable to substitute these measurements by simple and reliable soil indices, provided they correlate signif icantly with IR.It was hypothesised that decreased IR, as a consequence of dispersion and/or slaking, might be related to different soil stability indices obtained at the macro-aggregate (i.e., slaking) and micro-aggregate (i.e., clay dispersion) levels, such as those indices proposed by Rengasamy et al. (1984), Kemper and Rosenau (1986), Le Bissonnais (1990Bissonnais ( , 1996)), Amezketa and Aragüés (1995) and Amezketa et al. (1996).For example, Miller and Baharuddin (1986), Minhas and Sharma (1986), So and Cook (1993) and Levy et al. (1993) found that soil infiltration was inversely correlated with soil dispersibility.
The second objective was to determine whether the steady-state infiltration rates of the ten soils could be estimated from various macro-and micro-structural stability indices previously determined by Amezketa et al. (2003 a, b) for the same soils.

Soils
Ten arid-zone soils located in the Bardenas I and Monegros I irrigation districts of the middle Ebro river basin (Aragón, Spain) were sampled (0-20 cm depth), air-dried, ground and sieved (< 2 mm).The soil samples were characterised by standard methods (Carter, 1993) and showed a wide range of chemical and physical properties (Table 1).Three soils were saline-sodic, one was saline-non sodic and the remaining six were non saline-non sodic.All the soils were calcareous (CaCO 3 values between 12 and 44%) and had organic matter contents between 1 and 4%.Five textural classes (clay loam, silt loam, silty clay loam, loam, and silty clay) were represented in these soils.X-ray diffraction patterns of the clay fraction indicated high proportions of hydrated micas (> 70% of total clay) and chlorites (5-20% of total clay) and low proportions of kaolinites (< 5% of total clay) and pyrophyllites (< 5% of total clay, except in Callén 1, Tramaced 2, Sariñena 4, Grañén T1 and Grañén 1, with values of 5-20% of total clay).Swelling smectites and vermiculites were not found.
These soils were selected because, despite their similar mineralogy, they differed substantially in several of their stability indices (Table 2) (Amezketa et al., 2003a(Amezketa et al., , 2003b)).Except for the water stable aggregates (WSA) and the mean weight diameter of the aggregates in the slow-wetting treatment (MWD slow ), all the coeff icients of variation (CV) of the means of the stability indices were greater than 24%.Moreover, some of the indices expressing clay dispersion (i.e., spontaneous and mechanically dispersed clay) had CV of the means close to or greater than 100%.

Susceptibility of soils to clay dispersion: effect of electrolyte concentration on infiltration rate
Four replicate soil columns of each soil were prepared by packing between 40 and 46 g of the < 2 mmsoils into small plastic methacrylate cylinders (4.4 cm in diameter by 12.0 cm long) at bulk densities of 1.3 to 1.5 Mg m -3 , depending on the soil.The cylinders were open at the top and closed at the bottom, except for a plastic outlet for collection of the leachates.The soils were carefully added to the cylinders to a total thickness of 2.0 cm over a 2.0 cm layer of acid-washed  dS m -1 ) (mmol (g kg -1 ) (g kg -1 ) (g kg -1 ) (g kg -1 ) MPa MPa (%) (%) (cmol kg -1 ) (g kg -1 ) (g kg -1 ) (g kg -1 ) L -1 ) 0,5 (g kg -1 ) (g kg -1 )  2), prevented clay dispersion.The saturated columns were left to stand for about 13 h to increase aggregate mechanical strength (Levy et al., 1997).
Following saturation, the flow direction was reversed and the columns were ponded and leached using a 3 cm-constant-head device.Three leaching waters were used: (i) canal irrigation water (CW, EC ≈ 0.38 dS m -1 , SAR ≈ 0.5 for the Bardenas soils, and EC ≈ 0.42 dS m -1 , SAR ≈ 0.8 for the Monegros soils), (ii) deionized water (DW, EC < 0.01 dS m -1 ), which simulates rain water and leads to maximum clay dispersion, and (iii) saturated gypsum water (GW, EC ≈ 2.2 dS m -1 ), which represents conditions without clay dispersion.
The leachates were collected in appropriate volume increments, and their ECs were periodically measured.The infiltration rate (IR, equivalent to the saturated hydraulic conductivity) of the soil columns was calculated using: where ∆Q is the volume of water collected during a given time period ∆t, and A is the cross sectional area of the soil columns.The leaching process was conti-nued until both a final steady-state effluent EC and infiltration rate (FIR) were achieved.These steadystates were reached in about 1-74 pore volumes, depending on the soil type.
A soil was considered susceptible to clay dispersion when the FIRs obtained with DW and/or CW were significantly lower (P < 0.05) than the FIR obtained with GW.For comparison of the susceptibility of soils to DW and CW, the FIRs were expressed in relative terms (RFIR), given by the ratio of FIR for a given solution (DW or CW) to that for GW.Thus, RFIR DW and RFIR CW represent stability indices of infiltration against clay dispersion produced by rainwater (DW) and irrigation canal water (CW), respectively.Low RFIR values indicate susceptibility to clay dispersion and high RFIR values indicate resistance to clay dispersion.

Susceptibility of soils to aggregate slaking: effect of fast wetting on infiltration rate
The preparation of the soil columns and the leaching process for this test was done as described above.The only difference was that air-dry soils were used instead of pre-saturated soils, and that only irrigation canal water (CW) was used for leaching.This fast wetting treatment was compared with the previously described slow wetting-CW leaching treatment.
significantly lower (P < 0.05) than the FIR obtained in the slow-wetting treatment.For comparison of the susceptibility of soils to slaking, the FIRs were expressed in relative terms (RFIR), given by the ratio of FIR for fast-wetting treatment to that for the slow-wetting treatment.

Statistical analysis
Each soil and treatment was replicated four times.Eighty percent of the coefficients of variation (CV) of the mean FIRs were lower than 30%, and the average CV for all the treatments and soils was 28%.According to other studies (Abu-Sharar et al., 1987;Chiang et al., 1987), these values were considered satisfactory.
Results were analysed using the Statgraph Plus 2.1 software.One-way ANOVA was carried out to compare the means of the FIR among soils and treatments.Where the analyses showed significant differences at P ≤ 0.05, Duncan's multiple range tests were conducted to separate FIR values from individual treatments.Also, correlation analysis, a non-parametric statistical test (Spearman rank correlation test), and simple regressions were employed.The Spearman rank correlation test is resistant to outliers since it is based on the ranks of the data rather than on the data itself.Statistical signif icances were reported at the 0.05 (P < 0.05,*) and 0.01 (P < 0.01,**) probability levels.

Susceptibility of soils to clay dispersion: effect of electrolyte concentration on infiltration rate
The mean IRs as a function of cumulative time of leaching for the ten soils pre-saturated with tap water and subjected to deionized water (DW), canal irrigation water (CW) and gypsum-saturated water (GW) are presented in Fig. 1.Changes in IR were attributed to clay dispersion, since slaking was prevented (the soils were pre-saturated before leaching; i.e., no soil matric potential gradient and no entrapped air in the soil columns).Decreases in IR GW during leaching were generally negligible, indicating the lack of clay dispersion when the soils where equilibrated with the gypsum-saturated water.On the other hand, some soils exhibited pronounced decreases in their IRs, especially when subject to DW, indicating clay dispersion and sealing.In most soils the IR DW were lower than the IR GW from the start of leaching, suggesting that clay dispersion induced by DW was almost instantaneous.
It took 1 to 5 h to achieve the steady-state or final infiltration rate (FIR).Approximately 1 to 145 pore volumes were passed through the columns in order to equilibrate the exchange phase of the soil with the composition of the solution and to obtain a constant EC of the leachates (data not given).The mean FIRs of the soils studied varied from 955 mm h -1 (FIR CW of Sariñena 4) to 0.1 mm h -1 (FIR DW of Callén 1).
The final infiltration rates in most soils decreased in the order: FIR GW > FIR CW > FIR DW (Table 3), following the well-established effect of increased clay dispersion and partial plugging of conducting pores (> 50 µm) with decreasing electrolyte concentrations.
The FIR CW was significantly lower (P < 0.05) than FIR GW for five of the ten soils (Table 3).These five soils had RFIR CW values lower than 70% (Table 3).
Seven out of the ten soils leached with deionized water (i.e., simulated rainwater) had FIR DW significantly lower (P < 0.05) than the corresponding FIR GW (Table 3).These seven soils had RFIR DW values lower than 60% (Table 3).Two (i.e., SA 3/1 and Sariñena 4) out of the three remaining soils also exhibited FIR DW values 35 to 45% lower than the corresponding FIR GW values, but they were not significantly different due to their relatively large standard errors.

Susceptibility of soils to aggregate slaking: effect of fast wetting on infiltration rate
The susceptibility of soils to slaking was analysed by comparing the IRs of the presaturated (i.e., slow wetting) and air-dry (i.e., fast wetting) soil columns subject to CW (Fig. 1).The IRs in five of the ten airdry soil columns (IR CWns ) increased during the first minutes of leaching, probably due to dissolution and/or escaping of the entrapped air within the pores and the rearrangement of the soil particles during the fastwetting process.Thereafter, IR CWns remained constant or decreased slowly during the leaching process.Except for the first 60 to 90 min of leaching, the IR curves of the fast wetting and slow wetting soils were parallel and quite similar in eight of the ten soils.Only soils SA 31/1 and Callén 1 had unparalleled IR CWns and IR CW curves because of the impervious character of their air-dry soil columns (i.e., IR CWns = 0 mm h -1 ) from the start of leaching, indicating that slaking took place immediately after ponding with CW.It is noted that the IR curves were parallel in Grañén T1, but IR CWns was greater than IR CW from the start and during the leaching process, indicating that the fast wetting in this soil increased and maintained the proportion of conducting pores.
The final infiltration rate (FIR) of the slowly prewetted (i.e., no slaking) and the air-dry (i.e., potential slaking) soils were similar (i.e., not significantly different at P > 0.05) in eight of the ten soils studied (Table 3).

Predicting steady-state infiltration rates from macro and micro-aggregate stability indices
We first analysed potential relationships between the relative final infiltration rates (RFIR) of the ten soils studied and the corresponding stability indices presented in Table 2.No clear relations were found in general between RFIR and these stability indices, although WSA (Water Stable Aggregate) vs. RFIR CW , and MDC (Mechanically Dispersed Clay) vs. RFIR DW showed an apparent threshold behaviour.Thus, RFIR CW > 50% were only obtained when WSA > 75%, and RFIR DW > 30% when MDC < 45%.
In addition, RFIR DW was satisfactorily estimated from the mechanical dispersed clay (MDC) through: ; R 2 = 0.692 ** This equation indicates that the most dispersible soils on the basis of the MDC (a measure of the stability of clays against the mechanical and chemical dispersion; Rengasamy et al., 1984) are also those presenting the lowest RFIR DW .In addition, the Spearman correlation coefficient (r s = -0.71* ) shows that RFIR DW and MDC ranked the soils in a similar inverse order.
Secondly, we analysed potential relationships between the final infiltration rates (FIR) of the ten soils and the corresponding stability indices presented in Table 2.This analysis was carried out using the Ln FIR in order to normalise their log-normal distribution values.WSA was the only index with some significant correlation coefficients (r) when the ten soils were included in the analysis.Thus, Ln FIR CW and Ln FIR CWns were positively and significantly (P < 0.05) correlated with WSA, although the R 2 only explained about 50% of the Ln FIR variances.
The graphical representations of Ln FIR vs. WSA (Fig. 2) show that the Callén soil was a clear outlier in these relationships (i.e., its WSA of 83% is much higher than the expected value given by the general linear tendency).This is probably because of its very high SAR (111), which promoted the complete breakdown of aggregates in the WSA test, so that the detached particles totally blocked the 250 mm-sieve, giving an erroneous result.When the Callén soil (represented in Fig. 2 by a black dot) was excluded from the regressions, the correlation coefficients increased to greater than 0.8 (significant at P < 0.01) for Ln FIR GW , Ln FIR DW and Ln FIR CW (Table 4).The Spearman rank correlation test lead to the same correlations (data not shown), and thus to similar qualitative conclusions.These positive correlations are conceptually consistent, since both WSA and FIRs were obtained under similar treatment conditions (slow wetting, certain mechanical stress and certain chemical dispersion, except in the gypsum leaching treatment).The mechanical stress consisted of a light stirring of the soil samples in the WSA test, and of a light soil disturbance due to the water flow in the infiltration studies.In summary, the more stable the aggregates, the greater the steady-state soil infiltration rates.WSA may therefore be used to estimate FIR with a high level of confidence.

RFIR
Similarly, the graphical representations of Ln FIR vs. MWD stir and FV show that the SA 92/1 was an outlier (black dots in Fig. 2).No apparent reasons were found for the abnormal behaviour of this soil.When excluded, all the regressions (except those with Ln FIR CWns ) were significant at P < 0.01 (FV) or P < 0.05 (MWD stir ), indicating that these stability indices could be apparently used to estimate the FIRs in this set of soils.The Spearman rank correlation coefficients also indicated that these indices ranked the soils in the same order as the infiltration parameters (data not given).
The positive correlations between the infiltration parameters (Ln FIR GW , Ln FIR CW , Ln FIR DW ) and the MWD stir index are conceptually consistent since they were obtained after presaturation and certain mechanical stress and chemical dispersion (except in the gypsum leaching treatment).Presaturation was performed by capillarity in the infiltration studies, and with ethanol in the Le Bissonnais test.The mechanical stress consisted of a light stirring of the soil samples in DW in the Le Bissonnais test, and of a light disturbance of the soil due to the water flow in the infiltration studies.In addition, certain chemical dispersion occurred in those tests, except in the gypsum leaching treatment.Thus, infiltration increased as aggregate stability (i.e., MWD stir ) increased.
On the other hand, the positive correlations between the steady-state infiltration values and the FV index (Fig. 2, soil SA 92/1 excluded) were opposite to the expected ones, since the higher the FV of a soil, the higher its tendency to disperse and, therefore, the lower its FIR.Although these results are not conclusive since only nine (seven for Ln FIR CWns ) soils were included in this analysis, they raise doubts about the use of the FV as an index of the stability of soils against clay dispersion and the subsequent reductions in soil infiltration rate.
It is remarkable to note the lack of significant correlations between the infiltration rates and the rest of the stability indices (MWD fast , MWD slow , MWD microag and SDC).We expected to f ind reliable and positive relationships between FIR CWns (i.e., fast wetting of soils with canal water) and MWD fast values, since the last index measures the stability of aggregates versus fast wetting (i.e., potential slaking of aggregates).However, only two soils were susceptible to slaking according to the FIR CWns , whereas all soils (except Barbués 3/1) were very sensitive to slaking according to MWD fast (MWD fast < 0.5 mm; Amezketa et al., 2003a).This suggests that MWD fast overestimates the actual slaking observed in the infiltration experiments.

Susceptibility of soils to clay dispersion: effect of electrolyte concentration on infiltration rate
No clear relationships were found between the FIRs obtained with the three solutions and the soil properties presented in Table 1.Nevertheless, the two soils exhibiting the highest FIRs (SA 3/1 and Sariñena 4) had the lowest SAR and EC in the saturation extract, whereas the two soils exhibiting the lowest FIRs (SA 31/1 and Callén 1) had the highest SAR and EC.The low FIRs of SA 31/1 and Callén 1 soils, close to 1 mm h -1 with the canal irrigation water (Table 3), indicate that they are not cultivable without amelioration.The reclamation is feasible using gypsum-saturated waters in soil SA 31/1, since its FIR GW increased to 7.8 mm h -1 , but not in soil Callén 1, whose FIR GW of 1.7 mm h -1 was still insufficient for leaching purposes.
A remarkable result is that even though the electrolyte concentration of the canal water (3.8 to 4.2 mmol c L -1 ) was greater than the soils' FVs (except in soil SA 92/1, with a FV of 5.0 mmol c L -1 ), FIR CW was significantly lower (P < 0.05) than FIR GW for five of the ten soils (Table 3).Since there were no swelling clays in these soils and aggregate slaking was negligible in the slow-presaturated soil columns, clay dispersion should be mainly responsible for the decreased FIRs.This implies that FV was not an appropriate index of dispersion for those soils.Nevertheless, it should be noted that the electrolyte concentration of the canal water was only slightly greater than the FVs.Furthermore, the SAR value of the canal water was between 0.5 and 0.8, whereas the FVs were obtained with CaCl 2 solutions (i.e., SAR = 0).The evidence that clay dispersion was induced by the low EC of the canal water is substantiated by the fact that eight of the ten soils had FIR CW values not significantly different (P > 0.05) from the FIR DW values (Table 3).
The susceptibility of soils to clay dispersion and sealing induced by CW may be classif ied in three different groups on the basis of their relative final infiltration rates (RFIR CW = 100FIR CW /FIR GW ) (Table 3): (i) very susceptible soils (SA 31/1 and Grañén T1), with RFIR CW < 25% (i.e., reductions greater than 75% The substantial reductions in FIR DW with respect to the FIR GW values for the seven soils (Table 3) were attributed to clay dispersion induced by the very low electrolyte concentration (< 0.1 mmol c L -1 ) of deionized water, well below the 1.4 to 5.0 mmol c L -1 soils' FVs (Table 2).The dispersed clay particles were retained within the soil columns, since no clay was collected in the leachates (data not given).The lodgement of clay particles in soil columns leached with distilled water were also reported by Felhendler et al. (1974), Frenkel et al. (1978), Pupinsky and Shainberg (1979), Chiang et al. (1987), and Amezketa and Aragüés (1995).Even soils with very low SAR levels (i.e., SA 20/1, SA 92/1, Grañén T1) exhibited these reductions in FIR DW , in agreement with Amezketa and Aragüés (1995).Similar results were obtained by Agassi et al. (1981), Shainberg et al. (1992) and Mamedov et al. (2000) in sprinkler-irrigated soils, where the beating action of raindrops enhanced clay dispersion and exacerbated the negative effect of SAR.However, less documentation is available concerning the negative effects of very low SAR levels on infiltration in flood-irrigated soils, where the drop-impact mechanism is irrelevant.
The susceptibility of soils to clay dispersion induced by DW may be classif ied in three different groups on the basis of their relative final infiltration rates (RFIR DW = 100FIR DW /FIR GW ) (Table 3): (i) very susceptible soils (SA 20/1, SA 31/1, Callén-1, Barbués 3/1 and Grañén T1), with RFIR DW ≤ 30% (i.e., reductions greater than 70% compared to the corresponding FIR values obtained with GW, and significantly different at P < 0.05), (ii) susceptible soils (SA 3/1, SA 81/1, SA 92/1 and Sariñeña-4), with RFIR DW between 55 and 65% (i.e., reductions of between 45 and 35% of FIR GW , and depending on soils significantly different at P<0.05 or not significantly different at P > 0.05), and (iii) resistant soils (Tramaced 2) with RFIR DW = 94%, not significantly different to the FIR GW at P>0.001.In summary, nine (seven from a statistical point of view) out of the ten soils examined were very sensitive or sensitive to clay dispersion induced by simulated rainwater.The addition of chemical (i.e., gypsum, soil conditioners, etc.) and physical (i.e., soil mulching) amendments to the surface of these soils are recommended practices to avoid or minimise clay dispersion and soil crusting under rainfall conditions.

Susceptibility of soils to aggregate slaking: effect of fast wetting on infiltration rate
In eight of the ten soils studied, the similar final infiltration rate (FIR) of the slowly pre-wetted (i.e., no slaking) and the air-dry (i.e., potential slaking) soils (Table 3) indicated that slaking did not determine their FIR values.Although slaking took place in soil SA 31/1, its FIR CWns value was similar (i.e., not significantly different at P > 0.05) to the FIR CW value (which was very low because of the high clay dispersion produced by the CW in the slowly pre-wetted treatment; Fig. 1).Of the two soils with significant differences (P < 0.05) between FIR CWns and FIR CW , Grañén T1 had a higher FIR CWns value (i.e., as previously indicated, fast wetting and particle rearrangement promoted infiltration) whereas Callén 1 had a lower FIR CWns (i.e., fast wetting induced slaking and decreased infiltration).In summary, slaking provoked by fast wetting was relevant in two soils (SA 31/1 and Callén 1), but only in Callén 1 the final infiltration rate of the slaked soil was significantly lower (P < 0.05) than in the unslaked soil.
The RFIR CWns (i.e., 100FIR CWns /FIR CW ) values (Table 3) were: (i) greater than 100% in three soils (i.e., fast wetting promoted inf iltration), although this increase was only significant (P < 0.05) in Grañén T1, (ii) very high (close to 100%) in two soils (i.e., infiltration was not affected by fast wetting), (iii) high (between 70 and 90%) in three soils (i.e., infiltration was low to moderately affected by fast wetting) and (iv) zero (i.e., fast wetting induced slaking and lead to impermeable soils) in two soils characterised by their already low or very low permeability irrespective of the leaching solutions.Thus, in general, slaking was not a significant process of aggregate breakdown, seal formation and plugging of conducting pores in the soils studied.In contrast, Le Bissonnais and Singer (1992) found that the infiltration rate in two initially air-dry soils was 20 times lower than that for the prewetted soils after 40 mm rainfall due to slaking.Levy et al. (1997) and Mamedov and Levy (2001) also observed that slaking reduced significantly the final infiltration rate in two and f ive soils respectively.The above authors concluded that the rate of surface aggregate prewetting prevails in determining aggregate slaking, susceptibility to sealing and low IR.These studies were performed under simulated rainfall, where the raindrop impact exacerbated the slaking effect on infiltration.In contrast, there are very few studies quantifying the effect of slaking on IR under flood irrigation, despite the evidences of slaking and slumping of aggregates during irrigation in field and in column studies (Ferruzzi et al., 2000;Mace and Amrhein, 2001).Auerswald (1995) observed that percolation (i.e., the amount of water that percolated in 10 min) of air-dry aggregates flood-irrigated with DW was much smaller than that of the same slowly-wetted aggregates, due to slaking.In general, percolation linearly increased and slaking was reduced with increasing antecedent moisture contents.Levy et al. (1997) ascribed the sensitivity of the aggregates to slaking to the high silt-to-clay ratio (1.63), whereas Mamedov and Levy (2001) found that soil susceptibility to slaking increased with increases in clay content up to 40%.None of these reasons could explain our results, since half of our soils had silt-toclay ratios greater than 1.6 and only two of them slaked.In addition, the only two soils that slaked had clay contents lower than 40%, while the only soil with clay content higher than 40% (SA 81/1) did not slake.
In terms of water management, the two soils susceptible to slaking should be irrigated using systems of low application rates (i.e., drip, sprinklers of low pluviometry, etc.).If furrow or flood irrigation systems must be used, slaking could be prevented or minimised by forming raised-beds (Chan and Mullins, 1994) and adding organic matter (Chenu et al., 2000;Ferruzzi et al., 2000) or hydrophobic polymers such as polyacrylamides (Ferruzzi et al., 2000) to the soil surface.These polymers prevent or slow down the entry of water into the aggregates, and increase the internal cohesion and binding of aggregates.
Finally, considering that seven and five soils were susceptible to clay dispersion induced by DW and CW, respectively, and that only two soils were susceptible to aggregate slaking, it can be concluded that clay dispersion rather than aggregate slaking was the principal mechanism controlling the infiltration rates of these arid-zone soils when subject to rainfall and canal irrigation waters.

Predicting steady-state infiltration rates from macro and micro-aggregate stability indices
The significant correlations found between RFIR DW and MDC indicate that MDC could be a reliable screening index of the relative susceptibility of soils to clay dispersion induced by rainwater (DW) if these preliminary results are validated in a wider spectrum of soils.Sumner (1993) suggested that MDC could reflect the behaviour of clays when the velocity of water in the soil pores is high enough to cause clay particles to become dispersed.Miller and Baharuddin (1986) found that a number of south-eastern USA soils were dispersible when shocked in DW, and obtained significant (P < 0.01) and negative correlations between several measures of soil dispersibility and soil infiltration.Particularly, the dispersible clay measured after 36 h of shaking at an 8:1 water:soil ratio, and a timeweighted dispersible clay index had the highest correlation coefficients (r ≈ -0.5 to -0.6).Levy et al. (1993) also found a significant (P < 0.01) and negative relation between the percentage of dispersed clay (i.e., clay dispersed in DW after 1 h-shaking at an 15:1 water:soil ratio) and the logarithm of the f inal infiltration rate in 23 soils, although the correlation only explained 55% of the FIR variance.Minhas and Sharma (1986) found a negative relation between the relative hydraulic conductivity of soils leached with DW (RHC DW ) and the degree of clay dispersion (CD) calculated as the percent of total clay that dispersed after shaking (end-over-end, twenty times in a 50:1 water:soil ratio).The correlation coefficients were -0.67 ** and -0.78 ** for sandy loam and clay loam soils, respectively.However, the prediction capability was improved when log RHC DW was plotted as a function of log CD (r increased to -0.90 and -0.97 for the two soils, respectively).So and Cook (1993) showed for 152 soils that the hydraulic conductivity values of slaked and dispersed soils were strongly dependent on the amount of dispersed clay (i.e., percent of total clay that dispersed in DW after shaking end-over-end for 30 min in a 20:1 water:soil ratio) or the amount of dispersed silt+clay.The most suitable relationships were logarithmic with a significant level at P < 0.001.These results are consistent with those found in our work.
The positive correlations found in our work between WSA and FIRs may indicate that WSA may be used to estimate FIR with a high level of conf idence.Reanalysing the data of Ramos and Nacci (1997), we also found a significant correlation between WSA and the hydraulic conductivity of 11 mulched soils.Roth et al. (1986) (cited in Roth, 1992) also found a positive and significant correlation between WSA and FIR obtained under simulated rainfall.
Similar to our f indings, Amezketa et al. (1996) obtained significant correlations between MWD stir and the inf iltration rates of 10 California soils.Reanalysing the data of Ramos and Nacci (1997), we also found significant correlations between MWD stir and the hydraulic conductivity of 11 mulched soils.Le Bissonnais and Arrouays (1997) also found a significant positive correlation between wet stirring MWD (similar to MWD stir ) and an infiltration coefficient (i.e., the ratio between inf iltration and rainfall) under simulated rainfall.
In summary, of the eight stability indices examined as possible estimators of FIR, the WSA (water stable aggregates), MDC (mechanically dispersed clay) and MWD stir (mean weight diameter of stirring aggregates after a prewetting treatment) indices gave consistent and significant relationships with FIRs.The validity of the WSA, MDC and MWD stir indices to rank and predict the inf iltration rate of soils subject to clay dispersion and aggregate slaking should be ascertained in a wider spectrum of soils.

464E.SAFigure 1 .
Figure 1.Infiltration rate (IR) as a function of cumulative time for the ten soils pre-saturated with tap water and leached with deionized water (DW, EC < 0.01 dS m -1 ), canal irrigation water (CW, EC ≈ 0.4 dS m -1 , SAR < 1), and gypsum-saturated water (GW, EC = 2.2 dS m -1 ) and of ten non pre-saturated soils leached with CW (CW ns ).Vertical bars represent standard errors of the means of four replications.

Figure 2 .
Figure 2. Relationships between Ln of final infiltration rates (FIR GW , FIR DW , FIR CW and FIR CWns ) of the soils studied and the corresponding stability indices WSA, FV and MWD stir , with indication of their linear correlation coefficients.Solid lines are for the ten soils (eight for LnFIR CWns ), and dashed lines are for the nine soils (seven for LnFIR CWns ) (i.e., black-dot soils excluded from regressions).

Table 3 .
Absolute (FIR) and relative (RFIR) steady-state or final infiltration rates measured in four replicate slowly-prewetted soil columns leached with gypsum-saturated water (GW), deionized water (DW), and canal irrigation water (CW), and in four replicate air-dry soil columns leached with CW (CWns).For each soil, FIR values with different letters were significantly different at P < 0.05 DW = 1 (0.0032 + 0.0013MDC )

Table 4 .
Correlation matrix between the natural logarithms of the final infiltration rates (Ln FIR) of the soils studied and the corresponding stability indices that gave significant correlation coefficients at P < 0.05 (*) and P < 0.01 (**).