INFLUENCE OF CRUST FORMATION UNDER NATURAL RAIN ON PHYSICAL ATTRIBUTES OF SOILS WITH DIFFERENT TEXTURES

One of the main negative anthropic effects on soil is the formation of crusts, resulting in soil degradation. This process of physical origin reduces soil water infiltration, causing increased runoff and consequently soil losses, water erosion and/or soil degradation. The study and monitoring of soil crusts is important for soil management and conservation, mainly in tropical regions where research is insufficient to explain how soil crusts are formed and how they evolve. The purpose of this study was to monitor these processes on soils with different particle size distributions. Soil crusts on a sandy/sandy loam Argissolo Vermelho-Amarelo (Typic Hapludult), sandy loam Latossolo Vermelho-Amarelo (Typic Hapludox) and a clayey Nitossolo Vermelho eutroferrico (Rhodic Kandiudalf) were monitored. The soil was sampled and data collected after 0, 3, 5 and 10 rain storms with intensities above 25 mm h-1, from December 2008 to May 2009. Soil chemical and particle size distribution analysis were performed. The changes caused by rainfall were monitored by determining the soil roughness, hydraulic conductivity and soil water retention curves and by micromorphological analysis. Reduced soil roughness and crust formation were observed for all soils during the monitored rainfall events. However, contrary to what was expected according to the literature, crust formation was not always accompanied by reductions in total porosity, hydraulic conductivity and soil water retention.


INTRODUCTION
The erosion process consists in the detachment, transport and deposition of soil particles (Pimentel et al., 1995).As in other tropical countries, the main agent responsible for soil erosion in Brazil is water (Pagliai et al., 2003).Changes and the intensification of land use in agricultural areas can affect the soil structure, porosity, aeration, infiltration and productivity, as well as expose the soil to the direct impact of raindrops, promoting the appearance of soil surface crusts, one of the main factors of soil degradation (Valentin & Bresson, 1997;Jones, 2002).Some soil types are more susceptible to the crusting process, e..g, soils with higher silt and fine sand content and lower aggregate stability (Le Bissonnais & Bruand, 1993).
The influence of soil crusting on water infiltration was first studied by Duley (1939) and McIntyre (1958) and more recently some authors have looked into the phenomenon in tropical climate (Faria et al., 1997;Schaefer et al., 2002;Brandão et al., 2006).Brandão et al. (2006) determined multiple regression equations to estimate the hydraulic resistance of crusts originated by the direct impact of raindrops as well as the rate of infiltration into crusted soils.
Studies on loamy soils in Northeastern France showed that the formation of structural crusts can reduce the soil hydraulic conductivity from 45-60 mm h -1 to nearly 6 mm h -1 , and depositional crusts can reduce the hydraulic conductivity to nearly 1 mm h -1 (Martin et al., 1997).Reduction in soil infiltration of 50 to 100 % can occur in a single rainfall event and infiltration rates are higher in sloped areas where high erosion rates minimize crusting (Morgan, 2005).
Soil surface sealing that occurs during the formation of surface crusts and the loss of roughness seem to be the subprocesses responsible for high runoff rates and soil erosion (Le Bissonais & Bruand, 1993;Bajracharya & Lal, 1999;Darboux et al., 2001).The purpose of this study was to monitor the evolution of surface crusts in three soil types in relation to the accumulative effect of natural rainfall events, and to evaluate the effect of surface crusts on soil porosity, roughness and water infiltration and retention in the surface soil layer.
Micromorphometrical, soil roughness and hydrophysical studies were used to monitor the formation and evolution of the surface crusts on the studied soils.The undisturbed soil samples used in the micromorphometrical studies were collected in three replicates in cardboard boxes (0.12 x 0.07 x 0.05 m).The samples were air-dried for 15 days and in a ventilated oven at 40 ºC for 48 h.The samples were impregnated with polyester resin and UVITEX OB (Cyba-Geigy®) fluorescent pigment (Murphy et al., 1977).After hardening, polished blocks were prepared and digital images captured for pore analysis and quantification.The distribution and pore morphology characteristics proposed by Murphy et al. (1977) and Ringrose-Voase (1991) were assessed using Noesis® Visilog 5.4 and macros developed in Microsoft Excel®.The pore quantity, size and type, as well as total porosity of the samples were calculated based on digital image analysis (Cooper et al., 2005;Juhász et al., 2007), and the porosity graphs constructed using SigmaPlot® software.
For the determination of the characteristic soil water retention curve and soil bulk density, undisturbed samples were collected in 50 10 -6 m 3 stainless steel cylinders from the layers 0-0.05 m from each plot, in three replicates.The soil water retention curves were drawn for the matric potentials -1, -3, -5, -6, -8 and -10 kPa using the Haines funnel, and the potentials -33, -50, -70 and -100 kPa determined by Richards´ porous plate apparatus.
Soil water infiltration and hydraulic conductivity were calculated using the TRIMS multidisc infiltrometer and the protocol developed by Ankeny et al. (1991), based on Wooding's (1968) equation.This protocol estimates hydraulic conductivity K(ψ) using only one disk in the measurements performed with the infiltrometer.Hydraulic conductivity was determined for the matric potentials -1, -3, -5 and -10 kPa.
Soil surface roughness was measured with a laser distancemeter mounted on a portable roughness scanner adapted from Arvidsson & Bölenius (2006) with a capacity of measuring an area of 1 m 2 , with a vertical resolution of 0.05 mm and a horizontal grid of 2.00 mm, totaling 500,000 sampling points per plot.

Table 1. Particle size distribution of the studied soils
To ensure homogeneity of the surface conditions of all soils when setting up the experiment, the plots were tilled with a heavy disk and leveling harrow, simulating cropping conditions.A 16 disk harrow was used (depth 0.3 m) and tractioned by a Ford 6610 tractor.The initial soil cover was eliminated by a broad spectrum herbicide (glyphosate) to maintain the soil bare during the entire experimental period.A herbicide dose of 3 L ha -1 was applied by a manual backpack sprayer.
Surface crusting and soil roughness evolution were observed in the rainy season (November 2008 to May 2009).To monitor the evolution, the soil surface was maintained bare throughout this period.Field measurements and soil sampling were carried out four times, according to the sequence of rainfall events considered erosive.In this study, rainfall events were considered erosive at an intensity of < 25 mm h -1 (Hudson, 1965).The four measurements and samplings were determined: the first after soil tillage (T0), the second after the occurrence of three erosive rainfalls (T3), the third after the occurrence of five erosive rainfall events (T5) and the last after the occurrence of 10 erosive rainfall events (T10).Measurements and sampling were always performed in three replicates and 24 h after the end of each rainfall, to permit the drainage of any excess water from the study area.
To characterize the surface horizons, soil particle size distribution and chemistry were analyzed in disturbed samples collected from the layers 0-0.05m and 0.1-0.15m.Five samples per plot were collected Variograms were created using the GS+® software package to analyze the roughness evolution and the variances between the measured points.
Means were compared in tests to detect the differences between the data.The experiment was evaluated in a randomized block design considering the soil type as blocks and rainfalls as treatments.Tukey's test at 5 % probability and the statistical analyses were performed using the SAS® package.

RESULTS AND DISCUSSION
The evolution of the soil surface roughness along the sequence of the 10 rainfall events can be observed in figures 1 to 6, which show the roughness evolution at time zero (T0), after 3 (T3), after 5 (T5) and after 10 rainfalls (T10).The initial roughness of the Typic Hapludox was high, with a rugged microtopography in spite of the high pulverization of the aggregates caused by tillage (Figures 1 and 2).After the third rainfall (T3), changes were observed in surface conditions, with a less rough soil and a less undulating microrelief than at T0. Between the 3 rd and 5 th rainfall, the roughness measurements were similar, showing that in spite of some changes in the microtopography, the soil surface changed little between those rainfalls.After the 10 th rainfall (T10) there was a significant reduction in roughness, forming a surface with a high smoothness degree (Figures 1 and 2).
The reduction in roughness of the Typic Hapludult (Figure 3) followed the same patterns observed in the Typic Hapludox, with a marked reduction from T0 to T10 and less accentuated between T0, T3 and T5.On the other hand, the concentration of soil aggregates at the soil surface after tillage (T0) was higher in the Typic Hapludult (Figure 4) than the Typic Hapludox.
The greatest changes were observed in the Rhodic Kandiudalf.At T0, the soil surface of the Rhodic Kandiudalf was rough with clods of various sizes (Figure 6) that were reduced with the sequence of rainfalls (Figures 5, 6a,b).The surface was smoothened after the rainfall sequence, but some of the clods remained and the appearance of fissures due to contraction and expansion processes of the clayey material by wetting and drying cycles was observed.
In the three soils studied, the roughness reduction was accompanied by a lowering of the soil surface evidenced by the modifications in the height measurements (Figures 1, 3 and 5).This process may have occurred due to the removal and redeposition of soil material by the natural accommodation of the soil particles, reducing the soil pore volume.deposit material (Darboux et al., 2001).A loss in surface roughness results in a loss of depressions of the microrelief that can alter the hydraulic resistance, runoff and increase water retention (Croft et al., 2009).The loss of soil surface roughness can be considered an important factor contributing to the time reduction of runoff formation.The smoothening of the soil surface diminishes the capacity of the soil surface to retain water and runoff can initiate as soon as the rainfall intensity surpasses the soil water infiltration rate (Darboux, 2001;Assouline, 2004;Croft et al., 2009).
The impact of raindrops on the soil surface, apart from modifying the surface roughness, can affect the soil structure and some of the soil physical properties.Among these modifications, changes in porosity caused by compression and remobilization of soil material can be cited.To evaluate how the raindrop impact affects soil porosity, micromorphometrical studies were performed to quantify and qualify the types of pores present in each sampling period of the experiment.
The images of the studied soils showed the formation of a crust made up of a apparently finer material than the material immediately below the crust.The Typic Hapludox presented high total porosity and number of pores in T0.In all monitored rainfalls, a low number of complex pores occupying a great area was observed.This indicates the high connectivity between pores in this soil.After three rainfall events (T3), few changes in total porosity, number of pores, and in the pore morphology and Soil surface roughness plays an important role in the control of runoff and is known as determinant of the capacity of a soil surface to retain, transport and   distribution were observed, whereas a loss in small elongated pores (20-50 µm) was recorded.At T5 and T10, the tendency was the same as at T3; the pore morphology and distribution were little altered, although some irregular and elongated pores were lost.This evolution did not affect total porosity.No significant differences were detected in the changes in porosity (Table 2) that occur due to the redistribution of soil particles caused by the raindrop impact.The modifications in the soil structure induced by soil tillage that favor soil pulverization, together with the low clay contents, promote the detachment and mobilization of soil particles during rainfall (Morgan, 2005).The sequence of rainfall events caused  remobilization of soil material, but this remobilization caused no alterations in pore shape and organization, or in total porosity.The total porosity of the Typic Hapludult was high at T0 (Figure 8) with well interconnected pores.As in the Typic Hapludox, the sequence of erosive rainfalls caused few changes in pore morphology, distribution and occupied area.The changes in porosity of the Typic Hapludult were not significant (Table 2).The mobilization and reassembly of the sandy material due to the rainfall sequence caused few changes in the surface porosity of this soil.
The greatest changes in porosity were observed in the Rhodic Kandiudalf, with a decrease in the area occupied by pores and modification of the pore morphology and distribution each rainfall event, except at T10 (Figure 9).This reduction can be explained by the redistribution of the detached soil particles and soil compression due to raindrop impact on the soil surface (Huang et al., 1983), reducing the quantity of irregular and elongated pores.At T10, an increase in irregular pores was noted that can be explained by the appearance of elongated pores (fissures) which, as they coalesce due to intense fissuration of this material caused by wetting and drying cycles that occur along the rainfall sequence, act as irregular pores.Despite the changes in porosity, no significant differences were observed for this property in the Rhodic Kandiudox (Table 2).
The low significant changes observed in soil porosity are confirmed by the few alterations observed in hydraulic conductivity.Different from results in the bibliography (McIntyre, 1958;Valentin & Bresson, 1992;Schaefer et al., 2002;Fox et al., 2004), the hydraulic conductivity measured between the analyzed periods (T0, T3, T5 and T10) did not decrease, showing that the crusting effect on the hydraulic conductivity for these soils is low.
The hydraulic conductivity of the Typic Hapludox and Typic Hapludult is considered high and continued high during the whole experimental period (Figure 10a,c), without significant differences between rainfalls (Table 3).The slight rearrangement of porosity at T3, T5 and T10 (Figures 7 and 8) caused a minor decrease inhydraulic conductivity in these soils compared to T0, due to a small loss in the hydraulically active pores (Souza et al., 2007).
The greatest modifications in hydraulic conductivity were observed in the Rhodic Kandiudalf (Figure 10e), although no significant differences were observed among rainfalls, regardless of despite the high variability between measurements (Table 3).
The measured values for hydraulic conductivity at T0 were slightly higher (although without significant differences) than the measured values for the rainfall sequence and in relation to the other soils, although the Rhodic Kandiudalf is classified as a clayey soil (nearly 50 % clay).Great quantities of dispersible clay can have negative effects on hydraulic conductivity, once these particles are fine, hydrophilic and form micropores that favor water retention.However, when these fine particles form aggregates, the porosity and permeability of these soils increases, due to the formation of structural pores between the aggregates that promote soil drainage and hydraulic conductivity.Aggregates with high clay quantities are stable and take longer for changes aggregates formed in loamy texture soils (Le Bissonnais, 1996).
For the Rhodic Kandiudalf, the sequence of rainfalls caused changes in the total area occupied by pores, in the total number of pores (Figure 9) and the pore size and type, causing changes in soil hydraulic conductivity.Due to the modifications caused at the soil surface by the rainfall events, the hydraulic conductivity of this soil passed from a soil with high permeability to a soil with intermediate permeability (Olson et al., 1996).Although the differences in hydraulic conductivity of the Rhodic Kandiudalf were not significant, they were greatly reduced as the rainfall events proceeded, leading to lower values of hydraulic conductivity at T3, T5 and T10 than at T0.
In most cases, surface crusts are formed by the orientation and redistribution of soil material, with higher density and smaller pores than uncrusted soil (Souza et al., 2007).This increase in microporosity increases the soil water retention capacity and restricts water flux in these areas (Silva & Ribeiro, 1997).
The few changes in porosity observed in the Typic Hapludox and the Typic Hapludult also caused little changes in soil water retention (Figure 10b,d), once the pore distribution was not significantly modified.On the other hand, for the Rhodic Kandiudalf the changes in porosity affected soil water retention.
The soil water retention curves for the Typic Hapludox (Figure 10b) showed no significant differences in pore distribution and water retention patterns.This soil water retention behavior along the rainfall events was due to the pore morphology and distribution described in the image section.The dominance of irregular pores in this soil explains the course of the water retention curve along the rainfall events.The crust formation on the Typic Hapludox caused no changes in the water retention patterns.
The water retention measurements of the Typic Hapludult showed no differences (Figure 10d).The water retention curves show a well-drained soil with well-developed macroporosity with retention characteristics that were unaffected during the rainfall events.Similar to the Typic Hapludox, no changes were observed in the water retention patterns of the Hapludult due to crust formation.The evidences observed in pore morphology studies explain the shape of the water retention curves as the rainfall events occurred.
The variability of the water retention curves for the Rhodic Kandiudalf was higher than for the other studied soils (Figure 10f).Similar to the behavior of pore morphology after 10 rainfall events, the soil water retention curves also indicated changes during the rainfall sequence.The curves showed increased water retention at all matric potentials measured between T0 and T10.The pore morphology evolution from T0 to T10 showed a decrease in irregular pores from T0 to T5 and an increase at T10.At the same time, a higher quantity of elongated and rounded pores was formed.The increase in water retention, mainly at the matric potentials between -30 kPa and -1000 kPa, is a result of the modified pore morphology.The transformation of conducting (irregular and large elongated) pores to retention (rounded and small to medium elongated) pores would explain the higher retention at these matric potentials.On the other hand, the increase in the quantity of elongated pores and their probable coalescence, forming irregular pores after T10, caused increased retention at lower matric potentials (saturation up to -10 kPa), though below the increase observed in matric potentials above -30 kPa (Silva & Ribeiro, 1997).

CONCLUSIONS
1.The formation of surface crusts was observed for all soils analyzed, and soil texture was an important factor in the differentiation of the formation of these crusts.
2. The hydraulic conductivity in the Rhodic Kandiudalf was reduced and pore morphology and water retention patterns varied due to the crust formation as the rainfall events succeeded.
3. The crust formation on the Typic Hapludox and the Typic Hapludult did not significantly affect pore morphology, soil hydraulic conductivity rates and water retention patterns.4. Soil surface roughness was reduced with the occurrence of rainfalls, independently of the clay content of the soil. 5. Our results suggests that the eventual runoff formation from tropical soils with loamy and clayey texture seems to be more related to a reduction in the soil surface roughness, that would reduce the surface water storage, than to changes in porosity and reduction of the soil hydraulic conductivity.

Figure 7 .
Figure 7. Number of and total porosity distribution in pore diameter classes for the Typic Hapludox after T0, T3, T5 and T10 rainfall events.TNP: total number of pores and TP: total porosity.

Figure 8 .
Figure 8. Number of pores and total porosity distribution in pore diameter classes for the Typic Hapludult after T0, T3, T5 and T10 rainfall events.TNP: total number of pores and TP: total porosity.

Figure 9 .
Figure 9. Number of pores and total porosity distribution in pore diameter classes for the Rhodic Kandiudalf after T0, T3, T5 and T10 rainfall events.TNP: total number of pores and TP: total porosity.

Figure 10 .
Figure 10.Hydraulic conductivity for the three soils after T0, T3, T5 and T10 rainfall events (a,c,e) and soil water retention curves for the soils after T0, T3, T5 and T10 rainfall events (b,d,f).

Table 2 . Total porosity analysis of the studied soils
Means in each column do not differ statistically by Tukey's test (p < 0.05).

Table 3 . Hydraulic conductivity in the studied soils
Means in each column do not differ statistically by Tukey's test (p < 0.05).