Aggregate breakdown and surface seal development influenced by rain intensity , slope gradient and soil particle size

Aggregate breakdown is an important process which controls infiltration rate (IR) and the availability of fine materials necessary for structural sealing under rainfall. The purpose of this study was to investigate the effects of different slope gradients, rain intensities and particle size distributions on aggregate breakdown and IR to describe the formation of surface seal. To address this issue, 60 experiments were carried out in a 35×30×10 cm detachment tray using a rainfall simulator. By sieving a sandy loam soil, two sub-samples with different maximum aggregate sizes of 2 mm (Dmax2mm) and 4.75 mm (Dmax4.75mm) were prepared. The soils were exposed to two different rain intensities (57 and 80 mm h –1 ) on several slopes (0.5, 2.5, 5, 10, and 20%) each at three replicates. The result showed that for all slope gradients and rain intensities, the most fraction percentages in soils Dmax2mm and Dmax4.75mm were in the finest size classes of 0.02 mm and 0.043 mm, respectively. The soil containing finer aggregates exhibited higher transportability of pre-detached material than the soil containing larger aggregates. Also, IR increased with increasing slope gradient, rain intensity and aggregate size under unsteady state conditions because of less development of surface

the arrangement of soil particles, therefore, the detached particles can clog the soil pores, again reducing the IR. Ribolzi et al. (2011) concluded that the kinetic energy of raindrops and associated risks of soil crusting also decrease on steeper slopes, which might lead to increasing IR. The soils of arid and semiarid regions due to low content of organic carbon are generally susceptible to surface sealing and erosion (Cerdà, 2000;Mahmoodabadi and Cerdà, 2013). Under these conditions, only few studies have investigated aggregate breakdown and surface sealing. The objective of this study was to evaluate aggregate breakdown under different rain intensities, slope gradients and soil aggregate sizes by the determination of aggregate size distribution and to assess the formation and development of surface seal on the basis of obtained data of IR.

Soil preparation and characteristics
In this study, a soil sample was taken from the upper 20 cm of an agriculture land. It was air dried and then passed separately through 2 mm and 4.75 mm sieves. Therefore, two soils with different maximum aggregate sizes were provided (Zamani and Mahmoodabadi, 2013), which were named D max 2mm and D max 4.75mm. Note there were no primary particles coarser than 2 mm in the soils, since the original soil was collected from agricultural land. Some physical and chemical properties were measured for both sub-samples, separately. Texture of the soils was determined using the hydrometer method (Gee and Or, 2002). Aggregate size distribution was determined by wet and dry sieving facility (Kemper and Rosenau, 1986). Also, some chemical properties of the soils including pH and EC were measured in soil: water suspension at the ratio of 1:5.
Organic carbon content was determined as described by Walkley and Black (1934) and the percentage of CaCO 3 equivalent was measured using the titration method (Pansu and Gautheyrou, 2006). The measured physical and chemical properties of the soils are listed in Table 1. The obtained results showed that the mean weight diameter (MWD) in terms of dry and wet for soil D max 4.75mm was 0.78 and 0.3 mm, respectively, while, these parameters for soil D max 2mm had lower values. Both the soils showed a very low organic carbon content (<1%), whereas, the content of CaCO 3 equivalent was higher than 10%, which is dominant in arid and semiarid region soils (Mazaheri and Mahmoodabadi, 2012). The fraction percentage of aggregates for the soils is also shown in Fig. 1. For both soils D max 2mm and D max 4.75mm, the most frequent size classes were found to be in the range of 0.063 to 0.5 mm, respectively with 75.9% and 79.9%, while larger and finer size classes were lower.

Treatments and experimental setup
Totally, 60 experiments were carried out using the prepared soil samples under different rain intensities of 57 and 80 mm h -1 and several slopes (0.5, 2.5, 5, 10, and 20%), each at three replicates. An experiment was done with a rainfall simulator to generate different rain intensities. The nozzle used in the rainfall simulator was a pressurized one which was placed 1.5 m above the soil surface (Fig. 2). In order to measure rain intensity, 16 containers (6.8 cm diameter) were used, which were placed at regular distances under the simulated rains (Mahmoodabadi et al., 2007). To assess the uniformity of rain intensity, the coefficient of Christiansen was calculated (Grierson and Oades, 1977).

100
(1) where xi is the measured intensity in each container, m is the average rain intensity and n is the number of containers. Also, the measurement of average drops size was done using the stain method (Arjmand Sajjadi and Mahmoodabadi, 2015). The average (± standard deviation) drop size for the rain intensities of 57 and 80 mm h -1 was 2.2±0.08 mm and 2.5±0.09 mm with the coefficient of uniformity of 86% and 80%, respectively A detachment tray was used in the experiments, which was a 35×30 cm drainable tray with 10 cm depth (Fig. 2). The washed sediment was collected from the central test area of the tray. On two sides of the test area, a buffer section was provided so that, the soil was not only lost by splash, but it could also be returned from the buffer area (Arjmand Sajjadi and Mahmoodabadi, 2015). Different parts of the applied detachment tray are shown in Fig. 2.

Rainfall simulation experiments
Before every experiment, each soil sample was saturated for 24 hours. Afterward, the drainage water was removed out of the tray. Simulated rainfall lasted until a constant runoff rate was reached (40-45 min). For each rainfall event, the sediment-laden overland flow was sampled at time intervals (2,5,15,20,30  (2) where Ω is stream power (W m -2 ), ρ is water mass density (kg m -3 ), g is the gravitational acceleration (m s -2 ), q (m -2 s) is volumetric flux per unit width and S is the gradient of bed slope (m m -1 ).
During each experiment, infiltrated water was collected from the bottom of detachment tray at different time intervals. Since, the soil was saturated during each run, aggregate breakdown and the resultant size redistribution compared to the original soil was attributed to the seal formation. Therefore, at the end of each experiment, the upper 5 mm of soil surface was sampled for the determination of aggregates size distribution.
Aggregate size distribution of the eroded soil was measured by wet sieving (Kemper and Rosenau, 1986). For this purpose, soil aggregates were submerged and gently sieved into clear water, while each sample was sieved for 2 min. For soil D max 2mm, six sieves with sizes of 1, 0.5, 0.25, 0.125, 0.063 and 0.037 mm and for soil D max 4.75mm, one additional sieve with a size of 2 mm were used. Then, remained aggregates on each sieve were dried in oven at 105 °C for 24 h.
For quantification of aggregate breakdown of the eroded soils, fraction percentage was determined for each size class compared to non-eroded (original) soil. The obtained data from the wet sieving of the original soil was subdivided into 10 size classes using interpolation method, each having an equal mass fraction (10%). Also, both soil samples were subdivided ten size classes. Finally, the fraction of each size class was obtained using the subdivision of equal classes obtained from the original soil as described in Mahmoodabadi and Sirjani (2012). Thereupon, the fraction of the eroded soils for each experiment was calculated based on the size classes of the original soil.
All statistical analyses were performed in the SAS statistical framework and for obtaining the main differences between the treatments; the Duncan's (α=0.05) test was applied 3. Results and discussion

Rain-induced particle size redistribution
The fraction percentage in ten size classes for soil D max 2mm due to different rain intensities and slope gradients compared to the original soil is presented in Fig. 3. The fraction percentage of the original soil was indicated in Fig. 3 by uniform fraction of 10% in each size class. If the fraction percentage of each size class (10 size classes of eroded soil) was greater than 10%, indicates that the size class increased on the soil surface. Generally, the fraction percentage in the size class of 0.02 mm was the highest among all the rain intensities and slope gradients. This size class was affected by decreasing in the fraction percentage of coarser size classes. Therefore, the fraction percentage in coarser size classes decreased and for finer size classes the opposite case was found.
For the rain intensity of 57 mm h -1 and 0.5% slope gradient, the fraction percentage of eroded soil in the range of 0.055-0.092 mm was slightly greater than that of the original soil ( Fig. 3a). The fraction percentage in the range of 0.121-0.411 mm decreased and in the coarsest size class (1.5 mm) it was slightly higher than the original soil. At 2.5% slope gradient, the fraction percentage in the size class of 0.055 mm was higher than the original soil, instead, the fraction percentages decreased in size classes coarser than 0.073 mm (Fig 3b). At 5% slope gradient, the fraction percentage in the range of 0.055-0.092 mm was higher, whereas in the size classes ranged from 0.121 to 0.411 mm, it was less than the original soil (Fig 3c). However, for rain intensity of 57 mm h -1 , the fraction percentage of the coarsest size class (1.5 mm) increased compared to the original soil. At 10% and 20% slope gradients, the fraction percentages increased in size classes ranged from 0.055 to 0.092 mm, while those size classes coarser than 0.121 mm decreased compared to the original soil (Figs. 3d and 3e).
In comparison case, for the rain intensity of 80 mm h -1 and in all slope gradients (Fig.   3), the fraction percentage in the range of 0.055-0.092 mm was higher than the original soil (except 5% slope gradient). In contrast, in the size classes coarser than 0.121 mm, the fraction percentage decreased compared to the original soil for all slope gradients (except 5% slope gradient). At 5% slope gradient, the fraction percentage in the range of 0.055-0.073 mm was higher and in size classes coarser than 0.092 mm, it was less than the original soil.
The obtained results for soil D max 2mm exhibited some differences in the two applied rain intensities. The first difference can be referred to the fraction percentage in the size class of 0.02 mm, which was higher in rain intensity of 57 mm h -1 than that obtained in rain intensity of 80 mm h -1 . This means that in rain intensity of 57 mm h -1 , however, the aggregates were broken down by raindrops impacts during the rainfall event and produced finer particles, the resultant surface flow did not have enough transportability to carry detached particles way out of the test area. Therefore, the fraction percentage of the finest size class (0.02 mm) was enhanced in the eroded soil under the lower rain intensity (57 mm h -1 ). In contrast, the higher rain intensity of 80 mm h -1 caused to more detachability of soil aggregates and higher flow rates, which intensified transportability of finer pre-detached materials as well. Asadi et al. (2011) reported that with increasing flow stream power, sediment size distribution became coarser, finally becoming similar to or even coarser than the original soil, therefore, finer sediment remained on the soil surface.
The second difference can be related to the coarsest size class (1.5 mm), which showed higher fraction percentage in rain intensity of 57 mm h -1 than that observed in 80 mm h -higher than rain intensity of 57 mm h -1 , as a result, much more larger aggregates have been broken down. Consequently, the coarser particles size percentage was reduced under 80 mm h -1 rain intensity compared to the lower rain intensity. In addition, the rain intensity of 80 mm h -1 , generated higher flow rates leading more transportability of aggregates. Meyer et al. (1980) found that the percentage of coarser particles in eroded sediment was higher for more intense rainstorms. Beuselink et al. (2000) reported that in lower stream powers, finer particles are transported selectively and considerably, instead, large particles remained on soil surface, but with increasing stream power, larger particles also were transported.
The obtained result for soil D max 4.75mm and rain intensity of 57 mm h -1 showed that the fraction percentage for size class of 0.043 mm found to be the highest, which implied a considerable increase compared to the original soil in all slope gradients (Fig.   4). For this lower rain intensity, the fraction percentage in the coarsest size class (3.375 mm) was more than the original soil for all slope gradients. Also, a reduction trend in the fraction percentage was found from the size class of 0.064 mm to 0.433mm (Fig.   4a). Similarly, for rain intensity of 80 mm h -1 , the most fraction percentage was placed at the finest size class (0.043 mm) and the size classes coarser than 0.064 mm showed less fraction percentages than the original soil in all the slopes (Fig. 4c).
Comparison of the fraction percentages for soil D max 4.75mm under different rain intensities (Fig. 4) showed that in both rain intensities, the most fraction percentage compared to the original was the finest size class (0.043 mm). However, for the rain intensity of 80 mm h -1 , the fraction percentage of the finest size class was higher than that obtained for the intensity of 57 mm h -1 . In contrast, in the coarsest size (3.375 mm), it was reduced at higher rain intensity (80 mm h -1 ) compared to the lower intensity (57 mm h -1 ). The result for the finest size class is in contradictory with soil D max 2mm . This may be partly due to the fact that the soil containing larger aggregates exhibited higher infiltration rate and lower flow rates (Mazaheri and Mahmoodabadi, 2012). The result showed that the flow stream power generated on soil D max 2mm and soil D max 4.75mm ranged from 0.0007 to 0.0346 and from 0.0004 to 0.0313 W m -2 , respectively. In other words, the higher rain intensity was introduced on soil D max 4.75mm, the greater amounts of finer particles were produced. Nevertheless, because of higher infiltration rate of this soil, the stream power of generated flow seems not to be enough to transport and move out all the pre-detached materials from the test area (Arjmand Sajjadi and Mahmoodabadi, 2015). This finding implies that the redistribution of particles or aggregates on the surface of eroding soil depends on aggregate size distribution as well as rain intensity and the resultant flow stream power.

Time changes of infiltration rate (IR)
Time changes of IR for soil D max 2mm under different rain intensities and slope gradients is presented in Fig. 5. For both rain intensities, at the beginning of event, infiltration values were at the highest rates, meanwhile, the fluctuations of IR for different slope gradients were relatively high. Due to the time changes of IR at these first minutes, this period can be considered as unsteady state conditions. Under these conditions, higher IR values were obtained for the steepest slope (20%). Towards the end of event, the variations of IR were minimal. Also, it reduced to reach steady state conditions as the changes of IR found to be negligible with time. The highest fluctuation of IR with time was found when the IR was at the maximum value, therefore this value for each experiment assumed as unsteady IR. To compare these two conditions, results of variance analysis for measured IR under unsteady and steady state conditions are presented in Table 2. As is obvious, the single effects of rain intensity and soil particle size distribution on IR were significant under both unsteady and steady conditions. In contrast, the influence of slope gradient on IR was just significant under unsteady state whereas, under steady state conditions no significant effect was found.
Since, the studied soils remained saturated during the rainfall, the time changes of IR can be only attributed to seal formation.  (1986) inferred that increased IR on steeper slopes can be resulted from reduced surface sealing. In some studies, no significant relationship has been found between slope gradient and IR (e.g. Singer and Blackard, 1982;Mah et al., 1992), whereas in some others, a reduction in IR with increasing slope gradient has been reported (e.g. Chaplot and Le Bissonnais, 2000;Essig et al., 2009). Fox et al. (1997) observed a reduction in IR with increasing slope gradient until a critical threshold was reached, thereafter, IR was found to be irrelevant to slope gradient. More counterintuitive are the studies that showed an increased in IR with increasing slope gradient (e.g. Poesen, 1986;Assouline and Ben-Hur, 2006).
Under steady state, lower rates of infiltration were observed compared to the unsteady state conditions. In addition, the effect of slope gradient on steady IR was insignificant (Table 2). According to Fig. 3, the aggregate breakdown due to raindrop impact produced finer aggregates, which were used to form a surface seal with lower hydraulic conductivity than the original soil. Freebairn et al. (1989) attributed the reduction in IR during rainfall in both laboratory and field conditions to the formation of surface seal.
Similarly, Moss and Watson (1991) reported that the reduction of IR is likely related to the obstruction of surface pores due to aggregate breakdown and seal formation.
Comparison of IR between the simulated rain intensities for soil D max 2mm (Fig. 5) implied that the higher rain intensity (80 mm h -1 ) led to greater IR values than those obtained for the lower rain intensity (57 mm h -1 ), particularly under unsteady state conditions. A plausible reason is that as rain intensity increased, the transportability of flow enhanced to carry detached particles way out of the test area. As discussed above, the finest size class (0.02 mm) showed higher fraction percentage in rain intensity of 57 mm h -1 than that obtained in 80 mm h -1 . However, some researchers (e.g. Foley and Silburn, 2002) reported an increase in IR due to higher rain intensities. In this regard, some inconsistent results have been reported (Liu et al, 2011;Schmidt et al 2010). Liu et al. (2011) believed that the relationship between rain intensity and IR is reverse. Schmidt (2010) verified that higher rain intensities with more erosive impacts can increase the amount of runoff as a result of IR reduction. In our study, it shows that in spite of the higher erosivity of more intense rain, the surface seal was not developed completely under unsteady state conditions because of washing out and removing fine soil particles. Figure 6 shows the changes of IR with time for different rain intensities and slope gradients for soil D max 4.75mm. The results of this soil are similar to those obtained for soil D max 2mm. At the start of rain event, the unsteady IR fluctuated highly among different slope gradients, while over time, it approached to a nearly constant value for all slopes. The result indicated that the unsteady IR increased with increasing slope gradient. Also, increasing rain intensity increased IR under unsteady state conditions.
A considerable point observed in both soils (Figs. 5 and 6) is that the measured IR in soil D max 4.75mm was higher than in soil D max 2mm. The reason for higher IR values in soil D max 4.75mm can be attributed to existing of larger aggregate sizes and the subsequent larger pores. In addition, larger aggregate create a relatively rough surface therefore, the generated runoff have more enough time to infiltrate into the soil.

Unsteady IR
The result of Table 2 indicated that the influence of slope on IR was significant just under unsteady state conditions. The effect of slope gradient and rain intensity on the unsteady IR for soil D max 2mm and D max 4.75mm is shown in Fig. 7. In general, the obtained unsteady IR increased as slope steepness increased, especially under the higher rain intensity. For soil D max 2mm, the unsteady IR ranged from 19 mm h -1 at 10% slope to 24.7 mm h -1 at 20% slope under 57 mm h -1 rain intensity. In higher rain intensity (80 mm h -1 ), it varied from 32.4 mm h -1 to 45.2 mm h -1 as slope gradient increased from 0.5% to 20%. Therefore, the unsteady IR under 80 mm h -1 was higher than 57 mm h -1 rain intensity. This finding was consistent with the results of Assouline and Ben-Hur (2006) who reported that infiltration rate and soil loss increased at higher rain intensities. This was attributed to thinner and less developed seal layer resulting from higher erosion of the soil surface and lower component of drop impact. Thus, the probable reason for the difference between the applied rain intensities in the present study may be partly as a consequence of greater stream power due to the higher rain intensity of 80 mm h -1 in removing fine soil particles and underdevelopment of surface seal.
For soil D max 4.75mm as slope gradient increased from 0.5% to 20%, the unsteady IR values due to rain intensities of 57 and 80 mm h -1 ranged from 25.7 mm h -1 to 30.6 mm h -1 and from 32.6 to 45.1 mm h -1 , respectively. Therefore, for soil D max 4.75mm similar to soil D max 2mm, the unsteady IR was higher under rain intensity of 80 mm h -1 than that under 57 mm h -1 . In both rain intensities, the unsteady IR values were higher at steeper slopes for both soils. This means that at steeper slopes and under unsteady state conditions due to faster depletion of pre-detached soil particles, seal layer was lessdeveloped, which enhanced the infiltration of water into the soil.

Conclusion
Considering the obtained fraction percentage in size classes for both eroded soils, the percentage of the finest particles was found to increase compared to the original soil, whereas, the reverse result was found for larger aggregates. Also, an increase in rain intensity led to an intensification of aggregate breakdown, however, the effect of rain intensity on the contribution of fraction percentage in size classes depends on the aggregate size. In addition, the soil containing finer aggregates exhibited relatively easy transportability of the pre-detached material than the soil containing larger aggregates.
Since, the studied soils remained saturated during the rainfall event, the change of infiltration rate with time was only attributed to seal formation. The surface seal was found to be less-developed during the first minutes, while with the progress of time, it was established to form a more developed seal layer. Furthermore, the result showed that the measured infiltration rate increased with increasing rain intensity, aggregate size and at the steepest slope under unsteady state conditions because of less development of surface seal. But under steady state conditions, no significant relationship was found between slope and the measured infiltration rate, which were attributed to the development of surface seal. Under steady state, lower rates of infiltration were observed compared to the unsteady state conditions. In addition, the soil containing larger aggregates exhibited higher rates of infiltration as this soil was less sensitive against raindrop impact and seal formation. The finding of this study highlights the importance of rain intensity, slope steepness and soil aggregate size on aggregate breakdown and seal formation which can control infiltration rate and the consequent runoff and erosion rates.