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


     

Published online 25 January 2008
Published in Soil Sci Soc Am J 72:412-423 (2008)
DOI: 10.2136/sssaj2006.0429
© 2008 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 Google Scholar
Google Scholar
Right arrow Articles by Augeard, B.
Right arrow Articles by Vauclin, M.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Augeard, B.
Right arrow Articles by Vauclin, M.
Agricola
Right arrow Articles by Augeard, B.
Right arrow Articles by Vauclin, M.
Related Collections
Right arrow Structure and Properties
Right arrow Experiment Design
Right arrow Soil Models

SOIL & WATER MANAGEMENT & CONSERVATION

Dynamics of Soil Surface Bulk Density: Role of Water Table Elevation and Rainfall Duration

B. Augearda,*, L. M. Bressonb, S. Assoulinec, C. Kaoa and M. Vauclind

a UR Cemagref, Hydrosystèmes et Bioprocédés, BP 44, 92163 Antony cedex, France
b UMR INRA/INAPG, Environnement et Grandes Cultures, 78850 Thiverval-Grignon, France
c Institute of Soil, Water and Environ. Sciences, Volcani Center, ARO, P.O.B. 6, Bet-Dagan 50250, Israel
d LTHE, UMR 5564 (CNRS, INPG, IRD, UJF), BP 53, 38041 Grenoble Cedex 9, France

* Corresponding author (benedicte.augeard{at}cemagref.fr).


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Measurements of soil bulk density profiles combined with thin-section analysis have been suggested to assess the structural seedbed degradation caused by rainfall. The effects of water table elevation and rainfall duration on surface sealing and seedbed slumping were studied on a repacked silt loam soil. Two initial water table elevations (0.3 and 0.7 m below the soil surface) and three simulated rainfall durations (15, 30, and 40 min at 30.5 mm h–1 followed by 180 min at 7 mm h–1) were used. Seedbed bulk density profiles were generated using x-radiography of resin-impregnated soil slices. Macroporosity measurements using image analysis and thin-section observations showed that infilling of eroded particles in interaggregate voids and compaction of the infilled particles were the main sealing processes. Below the seal, the seedbed exhibited coalescence and welding of aggregates into larger units, which affected mainly macroporosity. A model of sealing, exponential decrease in bulk density with depth, and slumping, linear increase in bulk density with depth, adequately reproduced the measured bulk density profiles (regression RMSE range 0.057–0.106 Mg m–3). The change in surface bulk density increased with rainfall duration, whereas this factor did not significantly affect slumping. The highest initial water table elevation led to the highest soil surface and internal seedbed bulk densities. It was suggested that high values of soil water content led to a decrease in aggregate cohesion. Moreover, the number of wetting and drying cycles and the water content during these cycles were shown to increase the magnitude of slumping.

Abbreviations: HWT, high water table • LWT, low water table • RI, roughness index • UV, ultraviolet


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Soil surface sealing caused by rainfall impact strongly controls infiltration into most bare soils and influences surface runoff and soil erosion. During the past few decades, a large number of experimental studies have investigated the decrease in surface hydraulic conductivity with rainfall characteristics—intensity, duration, and kinetic energy—and the influence of various factors on surface sealing—like soil mineralogy and texture, aggregate size distribution, aggregate stability, organic matter content, initial bulk density, initial water content, etc. (Assouline, 2004). Several processes of seal formation have been identified: disruption of surface aggregates, slaking, dispersion, compaction by the impact of raindrops, infilling of underlying pores by fine particles, and settling of fine particles carried in suspension by runoff, to name a few. These processes have been studied using various micromorphological techniques such as thin-section observations (e.g., Bresson and Cadot, 1992; Kwaad and Mücher, 1994), the scanning electron microscope (e.g., Chen et al., 1980; Onofiok and Singer, 1984; Tarchitzky et al., 1984; Wakindiki and Ben-Hur, 2002), image analysis (e.g., Rousseva et al., 2002; Fox et al., 2004), x-ray radiography (Roulier et al., 2002; Bresson et al., 2004), or x-ray tomography (Fohrer et al., 1999). The wetting rate and initial soil water content have been shown to influence the mechanisms involved in surface sealing (e.g., Francis and Cruse, 1983; Truman et al., 1990). At the soil surface, prewetted aggregates are prone to partial abrasion or microerosion rather than total disruption (Bresson and Cadot, 1992; Le Bissonnais, 1996).

Structural change during a rainfall event is not restricted to the soil surface but can also affect the underlying soil, which may collapse on wetting (e.g., Gusli et al., 1994; Kwaad and Mücher, 1994; Bresson and Moran, 1995). The resulting compaction, called slumping (Mullins et al., 1990), is caused by several processes where overburden pressure dominates rather than rainfall kinetic energy (Mullins and Panayiotopoulos, 1984; Or and Ghezzehei, 2002; Bresson and Moran, 2003). In a poorly aggregated seedbed, particles agglomerate due to matric suction, whereas a well-aggregated soil exhibits a coalescence of aggregates under plastic conditions partly induced by overburden pressure and enhanced by microcracking and partial slaking (Bresson and Moran, 2003). Coalescence occurs mainly under slow wetting, while fast wetting leads to physical dispersion and aggregate breakdown (Bresson and Moran, 2003). Few studies dealing with the quantification of the respective effects of sealing and slumping have been published, especially in the case of high water content conditions where the slumping effect is expected to play a significant role. These conditions are encountered in soils subject to waterlogging where the water table remains within the first meter of soil, particularly those that require surface or subsurface agricultural drainage for successful crop growth. In Western Europe, seedbeds may be in moist conditions during autumn (e.g., for winter wheat [Tritcum aestivum L.] cultivation) or during early spring (e.g., for horsebean [Vicia faba L. var. equina Pers.] and spring barley [Hordeum vulgare L.] cultivation). Soil physical properties, such as aggregation, penetration resistance, and pore size distribution, have been shown to be altered by this particular hydrological regime as well as water table management (Hundal et al., 1976; Baker et al., 2004). Moreover, saturation surface runoff may occur (Augeard et al., 2005), and any accompanying soil erosion is closely related to surface moisture conditions (Huang et al., 2001). Therefore, structural degradation by rainfall is likely to be influenced by the presence of a perched water table.

The distinction between the sealing and slumping effects requires precise measurements of bulk density profiles at a millimetric scale, i.e., smaller than the mean aggregate size, which generates great heterogeneity in the measurements. Therefore, only trends of bulk density profiles, easily assessed using fitted models, can be analyzed. Assouline and Mualem (1997) considered the seal as a nonuniform disturbed layer with continuous changes in physical properties with depth. They suggested a relation between rainfall characteristics and bulk density profile dynamics, from which hydraulic properties were derived. The corresponding infiltration model was calibrated and validated against data from Morin et al. (1981) and Baumhardt et al. (1990) for saturated and unsaturated flow conditions, respectively. It was found to perform well for a wide range of soil and rainfall conditions. Additionally, recent experimental data (Roth, 1997; Fohrer et al., 1999; Bresson et al., 2004) have validated the model of bulk density distribution with depth suggested by Mualem and Assouline (1989). To include structural change of the whole soil profile, Bresson et al. (2004) proposed a more complex model that took into account the slumping effect.

Scarce experimental data on the dynamic changes in bulk density during rainfall are available (Tackett and Pearson, 1965; Boiffin and Sebillotte, 1976; Heddadj and Gascuel-Odoux, 1999). They seem to support the general trend suggested by Assouline and Mualem (1997) but measurements were taken at a centimetric scale at best.

The objective of this study was to characterize the effect of initial high soil water content conditions induced by different water table elevations on the dynamics of sealing and slumping processes taking place in a repacked seedbed of a silty loam soil. Three durations of simulated rainfall were used to generate different development stages of the soil surface and seedbed degradation. Bulk density and macroporosity profiles were generated using x-radiography and ultraviolet (UV) images of impregnated undisturbed soil slices. Models were fit to experimental data to compare the six treatments (two initial water table elevations and three rainfall durations).


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Experimental Device
The soil material was sampled at 0.11 kg kg–1 water content from the top 0.3 m of the Ap horizon of a silt loam (a mixed, mesic Typic Hapludalf) formed on loess deposits in the western part of the Paris Basin (France) and stored at the sampling water content at 4°C. The main physical and chemical soil characteristics are: 112 g kg–1 clay (<2 µm), 147 g kg–1 fine silt (2–20 µm), 449 g kg–1 coarse silt (20–50 µm), 287 g kg–1 fine sand (50–200 µm), 5 g kg–1 coarse sand (200–2000 µm), 20 g kg–1 organic C, pH 7.5, and 10.5 cmol kg–1 cation exchange capacity. The soil was repacked and two layers were distinguished: a seedbed and an underlying layer. To prepare the underlying soil layer, the soil was passed through a 20-mm sieve and packed in an 0.8- by 0.4- by 0.9-m tank (Fig. 1 ). To stabilize its structure, the repacked soil was subjected to five wetting and draining cycles by raising the water table up to the soil surface and lowering it down to the bottom of the tank. The same underlying layer was used for all rainfall simulations. At the end of the last simulated rainfall, five undisturbed cores (diameter 80 mm, height 50 mm) were collected from various depths of the underlying layer, oven dried, and weighed to estimate the underlying layer dry bulk density, {rho}. The average value of 1.405 ± 0.010 Mg m–3 was similar to those encountered for no-till subsurface horizons of fields in the Paris Basin (Roulier et al., 2002). To simulate a freshly prepared seedbed, a top 50-mm layer was added on the underlying layer without compaction and raked to minimize aggregate segregation. The average bulk density of this layer measured on 10 undisturbed cores was 1.065 ± 0.007 Mg m–3.


Figure 1
View larger version (30K):
[in this window]
[in a new window]

  		Fig. 1. Schematic representation of the experimental setup.

 
A schematic representation of the experimental device is given in Fig. 1. Using a perforated board overlaid with a sheet of geotextile installed 50 mm above the bottom, the tank was hydraulically connected with an overflow measurement device, so that the initial water table elevation could be controlled. Surface runoff and drainage flow were collected and measured by tipping bucket gauges (tip volume, 15 ± 0.025 cm3). The soil water pressure head was monitored by two series of nine ceramic porous cup tensiometers (diameter 8 mm, air-entry pressure head –100 kPa), horizontally inserted at different depths (0.05, 0.15, 0.25, 0.35, 0.45, 0.55, 0.65, 0.75, and 0.85 m), connected to a pressure transducer via a scannivalve, and recorded every 15 min. Pressure head measurements were used to provide information on the water table elevation. The tank was set at a 1% slope.

Experiments
An experiment consisted of two successive rainfall simulations at a given initial water table elevation. Two experiments were conducted corresponding to two initial water table elevations: 0.7 m below the soil surface (–6.87 kPa head) for the low water table (LWT) in the first experiment and 0.3 m below the soil surface (–2.94 kPa head) for the high water table (HWT) in the second experiment. Between the two experiments, the 50-mm seedbed layer was removed and replaced with a new one.

At the beginning of each experiment, the water table was set to a 50-mm depth so that the repacked seedbed was wetted by capillary rise under zero suction using deionized water (electrical conductivity <6 x 10–6 S m–1). Then the water table was slowly lowered back to its initial, pre-rainfall depth (the total duration of this wetting and draining cycle was approximately 1 wk). After this cycle, the volumetric water content of surface aggregates was measured on 10 collected aggregates using the paraffin method (Blake and Hartge, 1986).

The effects of the pre-rainfall wetting and draining cycle on the surface layer bulk density were assessed by conducting three additional experiments. A 50-mm layer of the same soil was added without compaction and raked in a 400- by 400- by 100-mm tank. In the first experiment, five soil cores (diameter 80 mm, height 50 mm) were taken to measure initial bulk density. In the two other experiments, the repacked seedbed was subjected to a single wetting and drainage cycle similar to the one that was measured by the tensiometers located at a depth of 50 mm in the large tank for each water table elevation. Then seedbed bulk density before rainfall was measured on five undisturbed cores.

The rainfall simulation device consisted of pivoting 65–40 Veejet nozzles (Spraying System Co., Wheaton, IL) suspended 5.3 m above the large tank and used stored natural rainwater (electrical conductivity <2 x 10–3 S m–1) with different nozzle water pressures (p). The two successive runs of rainfall simulation presented different characteristics (Table 1 ). The first rainfall had an intensity of 30.5 ± 1.6 mm h–1 and was applied at a water pressure p = 140 kPa. The kinetic energy was estimated from drop size distribution (mean drop size 1.55 mm and velocity range 4 to 6 m s–1) at 16 J mm–1 m–2, a value within the 10 to 20 J mm–1 m–2 range representing natural spring or autumn rainfalls (McIsaac, 1990; Assouline et al., 1997a; Van Dijk et al., 2002). This 30.5 mm h–1 intensity rainfall was used to generate a structural seal on the soil surface. To obtain different seal development stages, one longitudinal third of the soil surface was shielded from rain after 15 min then another third after 30 min, the third one being exposed for 40 min. Then three rainfall durations were considered in each experiment. The shield mimicked an inclined roof to divert rain. The second rainfall consisted of a 7.2 ± 0.3 mm h–1 rainfall applied for 180 min (65–10 Veejet nozzle, p = 200 kPa, mean drop size 0.74 mm, velocity range 4 to 6 m s–1, and kinetic energy 4 J mm–1 m–2). This rainfall of low intensity was applied similarly to all experiments to quantify infiltration through the generated seals and is mentioned here because of its potential to influence the soil structure; however, this was not quantified.


View this table:
[in this window]
[in a new window]

  		Table 1. Characteristics of simulated rainfalls for the different experimental conditions.

 
After the end of each experiment, undisturbed soil surface samples were taken from the areas submitted to the three rainfall durations with an 80- by 50- by 50-mm box (two subsamples per area). Samples were air dried for 1 wk and afterward oven dried at 40°C for 48 h to make changes in soil structure as small as possible (Attou et al., 1998). After drying, no significant soil shrinkage (<3%) was observed for soil volume decrease following the Chertkov et al. (2004) method. Samples were impregnated under a vacuum of –5 kPa with polystyrene resin in which a blue ultraviolet fluorescent dye was incorporated. Three vertical 5-mm-thick slices were prepared from each sample. One thin section per experimental condition was prepared and observed with a stereomicroscope under UV light. Consequently, six 80- by 50- by 5-mm slices were extracted and considered as subsamples for a given treatment (i.e., each one corresponding to an initial elevation of the water table and to a duration of the high-intensity rainfall run).

Bulk Density and Roughness Measurements
Bulk density profiles of the slices were obtained using x-radiography. The x-ray generator device was described in Bresson et al. (2004). Films were digitized on a scanner with a 0.045-mm-square pixel. Bulk density images were generated by the calibration procedure presented in Bresson and Moran (1998). The calibration procedure involved three stages: (i) calibration of the radiography gray levels in terms of glass thickness using a staircase made from glass cover slips (r2 > 0.99), (ii) measurement of the ratio between the soil and resin mass attenuation coefficients and the glass mass attenuation coefficient using compacted bricks of known thickness and bulk density, and (iii) image correction accounting for the heterogeneity of the irradiation field. To calculate the bulk density profile, surface roughness was taken into account by generating equidistant lines from the soil surface, which were smoothed by linearization with depth (morphological distance function of the software Visilog by NOESIS, Les Ulys, France). The resulting average bulk density along every line formed the bulk density distribution with depth.

The roughness was evaluated by comparing the number of pixels of the soil surface line with a horizontal line. The calculated dimensionless roughness index (RI) is adapted from the roller chain method (Saleh, 1993) RI = [100 (1 – soil surface line length/horizontal line length)].

Macroporosity and Microporosity Estimates
Both faces of every slice were photographed under UV illumination with a digital video camera (0.1-mm-square pixel). This image was then turned into a binary representation of macropores (diameter >0.1 mm) using the two-step segmentation procedure suggested by Moran and McBratney (1992). Then the macroporosity distributions with depth were generated using the same procedure as for the bulk density distributions.

Meso- and microporosity (pore diameter <0.1 mm) profiles were determined by the difference between the mean macroporosity of both faces of each slice and the total porosity derived from the bulk density measurement of the same slice, assuming that the mean soil particle density was 2.65 Mg m–3. In the following, we used the term microporosity for pore diameters <0.1 mm. Bulk density measurement (resolution 0.045 mm) was linearly interpolated to reach the same resolution as the macroporosity resolution (0.1 mm). To help compare the densification degree, the void index (defined as the ratio between void and solid volumes) instead of porosity was used.

Modeling
A conceptual model was fit to the profiles to represent their trends and allow comparison between treatments. Following Mualem and Assouline (1989), the bulk density within the disturbed zone, {rho}c(z) was assumed to change with depth z below the soil surface, according to

Formula 1[1]
where {rho}i is the initial bulk density, {Delta}{rho}0 is the increase in bulk density at the soil surface, and {gamma} is a characteristic parameter of the soil–rainfall interaction. The value of {gamma} was used to compare the curvature of the exponential decay function with depth for all the treatments. Note that a modeled seal thickness dc can be estimated from Eq. [1] by truncating the distribution at the depth where {Delta}{rho}c = {rho}c – {rho}i is equal to 10–3 of {Delta}{rho}0. Thus, dc = –ln(10–3)/{gamma} and {gamma} is inversely proportional to the depth where the influence of the modeled seal on bulk density is considered insignificant.

Roth (1997) showed that the exponential decay function (Eq. [1]) seems to adequately reflect the earlier stage of seal formation, whereas a sigmoidal function could be more appropriate for the later stages. Bresson et al. (2004) suggested that the bulk density profiles depend not only on the soil material characteristics but also on the processes involved in seal formation: for the slaking and the coalescing seals, the sigmoidal function described the bulk density profile better, while for the infilling seals, both models (exponential and sigmoidal) were found to correctly fit the data. It should be noted that the exponential function in Eq. [1] can also represent the shape of a sigmoidal distribution when a power function of depth z is considered as the independent variable (Assouline, 2004). To account for slumping, Bresson et al. (2004) suggested modifying Eq. [1] by adding a linear term with z:

Formula 2[2]
where m is a fitting coefficient depending on the rheological properties of the soil material.

The influence of rainfall duration on seal thickness can be assessed by fitting Eq. [2] on bulk density profiles obtained after different rainfall durations separately and by comparing the values of {gamma}. The time evolution of the slumping slope m and the change in surface bulk density {Delta}{rho}0 can also be considered. Actually, following Assouline and Mualem (1997), the surface bulk density is expected to change with rainfall duration, t, during the dynamic phase of the developing seal according to

Formula 3[3]
where {Delta}{rho}0m is the maximal increase in bulk density, which depends on soil characteristics, and β is a soil–rainfall parameter.

Model Calibration and Statistical Methods
Initial bulk density was set at the measured value: {rho}i = 1.065 Mg m–3. The three unknown parameters of Eq. [2] ({Delta}{rho}0, {gamma}, m) were fitted on the six subsample bulk density profiles of one treatment (i.e., with a given initial water table elevation and after a given rainfall duration) by a statistical procedure that maximizes the log-likelihood (Nonlinear Mixed-Effects Models, software R, R Development Core Team, 2004). Mixed models integrate possible random effects on parameters between each subsample profile. The goodness of the fit was evaluated by the error made on each parameter estimate calculated with the regression procedure (standard error) and by the RMSE.

Profile analysis using analysis of covariance of two nonlinear regressions was made to assess whether the two bulk density data sets were significantly different. The two data sets correspond to bulk density profiles obtained for (i) various rainfall durations and the same initial water table elevation and (ii) various initial water table elevations and the same rainfall duration. The objective was to test the hypothesis H0 that the two regression curves are similar. The two data sets were therefore fitted independently with two distinct regressions of Eq. [2] and the residual sum of squares RSSA for both fits was calculated. Then Eq. [2] was fitted to both data sets together and RSSB was calculated. Letting nA and nB be the degrees of freedom from the two distinct regressions and the single regression, respectively, then the relative improvement by two distinct regressions vs. one was quantified by calculating the value of the Fisher–Snedecor test:

Formula 4[4]
If F is greater than FnBnA, H0 is rejected and the fit is significantly better with two distinct regressions. The Student's t-test was also used to assess the significance of the difference between two treatments in water content of aggregates and macrovoid and microvoid index evolution. Before using the Student's test, the equality of variance was tested using the F test.


   RESULTS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Effects of Initial Water Table Elevation and Rainfall Duration on Bulk Density Profiles
Figure 2 shows a typical measured bulk density profile and the corresponding fitted model (Eq. [2]). Sealing and slumping effects can clearly be detected by the decrease in bulk density in the first few millimeters at the top of the soil profile followed by a linear increase with depth. The measurements present a few autocorrelated variations in depth around the fitted model, with a periodic trend of approximately 5 mm per period in relation to the presence of coarse aggregates. In Fig. 3 , the fitted and measured bulk density profiles are plotted for the two initial water table elevations and the three rainfall durations (15, 30, and 40 min of the 30.5 mm h–1 intensity rainfall). At each depth, the average of the six measured values of each treatment is presented with the corresponding standard deviation. Mean standard deviation is equal to 0.11 Mg m–3. For nearly all cases, as in Fig. 2, the bulk density increased near the soil surface because of seal formation, and increased linearly with depth due to the slumping effect. Note that for the initial high water table elevation after 40 min of the 30.5 mm h–1 intensity rainfall (HWT3), bulk density values remained high in the entire profile (Fig. 3f) compared with the other treatments.


Figure 2
View larger version (13K):
[in this window]
[in a new window]

  		Fig. 2. Example of a bulk density profile and the corresponding fitted model (Eq. [2]) accounting for sealing and slumping processes. Sample treatment: 40 min of the 30.5 mm h–1 rainfall and low initial water table depth, 0.7 m below the soil surface. The parameters of the model ({rho}i, {Delta}{rho}0, m, dc, {gamma}) are schematically reported.

 

Figure 3
View larger version (14K):
[in this window]
[in a new window]

  		Fig. 3. Mean measured values and standard deviation of bulk density profile and the corresponding fitted model for low (LWT, 0.7 m below the soil surface) and high (HWT, 0.3 m below the soil surface) initial water table conditions after 15 (1), 30 (2), and 40 min (3) of the 30.5 mm h–1 rainfall event.

 
Equation [2] is a good description of the general trend in bulk density evolution. For the HWT1 treatment, the model fit was of poorer quality after the 20-mm depth because three of the subsamples were only 20 mm thick, increasing the weight of this layer in the regression. The regression RMSEs ranged between 0.057 and 0.106 Mg m–3 and were lower under LWT conditions (Table 2 ). These values are higher than those found by Bresson et al. (2004), where RMSEs varied from 0.044 to 0.079 Mg m–3. In Bresson et al. (2004), however, the fit concerned only one profile, whereas in the present study, the RMSE calculation includes both the variability among six subsamples of each treatment and the model errors in representing the profile evolution. Parameter standard errors calculated by the regression procedure reflect this variability, which mainly concerned the HWT conditions.


View this table:
[in this window]
[in a new window]

  		Table 2. Estimated value (± standard error) of the bulk density model parameters (Eq. [2]) fitted on six measured profiles for each initial condition (low and high initial water table) and each 30.5 mm h–1 rainfall duration (15, 30, and 40 min).

 
Estimations of the fitted model parameters (Eq. [2]) and the corresponding standard errors are presented in Table 2 for each initial water table elevation and rainfall duration. All the absolute values of the coefficient of correlation between parameters are <0.75. A correlation coefficient >0.95 would have indicated that the corresponding parameters may not be uniquely identified (Hill, 1998). Consequently, the nonlinear fits did not reveal limitations in the determination of the bulk density profile model parameters.

Statistical tests were performed to determine if the fits of Eq. [2] to each treatment were significantly different (Table 3 ). These tests infer that each initial water table elevation and each rainfall duration treatment have a different bulk density profile. The value F/FnBnA was calculated to assess the sensitivity of the test. The effect of the 30.5 mm h–1 rainfall duration was more pronounced for the HWT conditions than for the LWT conditions. Moreover, the influence of initial water table elevation increased with the 30.5 mm h–1 rainfall duration.


View this table:
[in this window]
[in a new window]

  		Table 3. Statistical ratios between calculated and theoretical values of the F test at the 0.05 probability level to compare the model fitted on the two initial conditions (LWT, low water table; HWT, high water table level) and the three 30.5 mm h–1 rainfall durations (1, 15 min; 2, 30 min; 3, 40 min). If F/FnBnA > 1, the two data sets are significantly different (nA and nB are the degrees of freedom of the numerator and denominator of the F value).

 
The volumetric water contents of surface aggregates varied from 0.17 ± 0.04 m3 m–3 for LWT to 0.22 ± 0.04 m3 m–3 for HWT. The Student's t-test with uniform variance showed that they were significantly different at the 0.05 probability level. Table 4 shows the mean bulk density of the 50-mm surface layer before and after the first wetting and draining cycle for the two initial water table conditions. The increase in bulk density is significantly greater for HWT (20%) than for LWT (13%), a difference of 7% (P < 0.01, Student's t-test). Assuming that the bulk density profile within the core is linear with depth as modeled in Eq. [2], the slopes m corresponding to the bulk density measured in 50-mm-high cores (Table 4) are 0.0054 and 0.0085 Mg m–3 mm–1 for LWT and HWT conditions, respectively. Then, before the rainfall experiments, a first slump process, whose magnitude depended on the initial water table level, was observed.


View this table:
[in this window]
[in a new window]

  		Table 4. Bulk density of the 50-mm-thick surface layer measured before ({rho}i) and after ({rho}f) the first wetting and drying cycle for the low water table (0.7 m below the soil surface) and high water table (0.3 m below the soil surface) initial conditions. Mean of five subsamples, with standard deviation in parentheses.

 
The evolution of the model parameters m, {Delta}{rho}0, and {gamma} as a function of the 30.5 mm h–1 rainfall duration for the LWT and HWT conditions is presented in Fig. 4 . The HWT experiment led to a thicker seal (lower value of {gamma}) and a greater change in surface bulk density at the end of the experiment. Table 5 shows the parameters obtained by fitting Eq. [3] to the change in surface bulk density, {Delta}{rho}0, with the 30.5 mm h–1 rainfall duration. The value of {Delta}{rho}0m under HWT conditions was greater than under LWT conditions (60%, approximately), but the time required to reach this value was much longer (i.e., βHWT < βLWT; see Eq. [3]). The initial water table elevation also affected the seedbed slumping, as shown in Fig. 5 , where the model parameter m is plotted as a function of the 30.5 mm h–1 rainfall duration: the estimated values of m are greater under HWT than under LWT conditions. The effect of rainfall duration (15, 30, and 40 min) was not observed under the LWT condition, whereas the slumping slope increased under the HWT condition after 40 min of the 30.5 mm h–1 rainfall.


Figure 4
View larger version (10K):
[in this window]
[in a new window]

  		Fig. 4. Evolution of (a) the exponential curvature, {gamma}, and (b) the change in bulk density at the soil surface, {Delta}{rho}0, as a function of the 30.5 mm h–1 rainfall duration for the low (LWT) and high (HWT) initial water table conditions. Vertical bars correspond to the standard error of the fit. Equation [3] is fitted on {Delta}{rho}0 to model its change in time.

 

View this table:
[in this window]
[in a new window]

  		Table 5. Fitted parameters of the maximal increase in bulk density {Delta}{rho}0m and the soil–rainfall parameter β for the exponential increase of change in surface bulk density, {Delta}{rho}0, with the 30.5 mm h–1 rainfall duration (Eq. [3]) for the low water table (LWT) and high water table (HWT) initial conditions.

 

Figure 5
View larger version (8K):
[in this window]
[in a new window]

  		Fig. 5. Evolution of the slope of the slumping model (m) as a function of the 30.5 mm h–1 rainfall duration for the low (LWT) and high (HWT) initial water table conditions. Vertical bars correspond to the standard error of the fit except for the values of m before rainfall, which were obtained from core bulk density measurements and whose vertical bars correspond to the standard deviation.

 
Microscopic Analysis of Soil Thin Sections and Micro- or Macrovoid Index
As observed in thin sections, most aggregates kept their integrity and were surrounded by a more porous soil material, which means that no microcracking or slaking of the structural elements was observed within either the surface seal or the underlying seedbed. Close microscopic analysis, however, revealed differences within both the seal and the seedbed between samples exposed to differing durations of the 30.5 mm h–1 rainfall, and mainly within the slumping process between samples submitted to different water table elevations.

Figure 6 shows microscopic images of the seal structure. Samples collected after 15 min of the 30.5 mm h–1 rainfall showed some remaining coarse aggregates near the soil surface and a high surface roughness: RI = 52 ± 6 for LWT and 43 ± 7 for HWT. Infillings of clean washed silt particles occurred in the compound packing void between the aggregates (Fig. 6a and 6c). After 30 and 40 min of rainfall, the surface roughness index significantly decreased to 41 ± 7 for LWT and 30 ± 6 for HWT (P < 0.05, Student's t-test with uniform variance). Moreover, qualitative analysis of the microscopic images showed that the particles infilled in the interaggregate voids were more densely packed (Fig. 6b and 6d).


Figure 6
View larger version (168K):
[in this window]
[in a new window]

  		Fig. 6. Thin-section images of structural surface seal under low (LWT) and high (HWT) initial water table conditions obtained by ultraviolet light after (a and c) 15, (b) 30, and (d) 40 min of the 30.5 mm h–1 rainfall event. Vertical bar gives the scale (0.8 mm). Initial aggregates are surrounded by bold solid lines.

 
Below the surface seal, the seedbed exhibited a loss of its original granular structure by coalescence and welding of aggregates into larger units (Fig. 7 ). The fine material, i.e., small aggregates and basic particles, was reorganized between the aggregates, and the resulting macropores displayed many convexities, which followed a meniscus-like outline. Samples corresponding to the HWT condition after 40 min of rainfall (HWT3) presented a substantial collapse and fewer macropores than other samples.


Figure 7
View larger version (138K):
[in this window]
[in a new window]

  		Fig. 7. Thin-section image obtained by ultraviolet light showing the coalescence and welding of aggregates below the surface seal for high initial water table conditions (HWT) after 15 min of the 30.5 mm h–1 rainfall event. Vertical bar gives the scale (0.8 mm). Initial aggregates are surrounded by bold solid lines.

 
Microvoid and macrovoid index changes within the seal are presented in Fig. 8a , where the differences between the void index at the soil surface and the void index at a 2-mm depth are plotted as a function of the 30.5 mm h–1 rainfall duration for each initial water table elevation. The macrovoid index increased significantly (P < 0.01) with depth from 0.21 ± 0.12 to 0.55 ± 0.24, whereas the microvoid index decreased from 0.75 ± 0.11 to a significantly lower (P < 0.01) 0.59 ± 0.16. The influence of the initial water table elevation on micro- and macrovoid changes within the seal was not detected (P > 0.05, Student's t-test with nonuniform variance), except for the 40-min duration of the 30.5 mm h–1 rainfall, where the HWT experiment led to a smaller difference in the macrovoid index than the LWT experiment.


Figure 8
View larger version (12K):
[in this window]
[in a new window]

  		Fig. 8. Differences between (a) 0- and 2-mm-depth and (b) 10- and 20-mm-depth micro- and macrovoid indices as a function of the 30.5 mm h–1 rainfall duration for low (LWT) and high (HWT) initial water table elevations. Vertical bars correspond to the standard deviation between six subsample slices.

 
Micro- and macrovoid index changes were also compared at greater depths to investigate the effect of slumping. Figure 8b shows the differences between the macrovoid and microvoid indices at 10- and 20-mm depths for all treatments. Slumping led to a decrease in the macrovoid index between the 10- and 20-mm depths from 0.50 ± 0.21 to 0.24 ± 0.16 (mean difference 0.26 ± 0.16 significantly positive for each treatment, P < 0.01, Student's t-test). The microvoid index of all the profiles remained quite constant as depth increased (difference –0.02 ± 0.11 not significantly different from 0, P > 0.05, Student's t-test with uniform variance).

Changes in Soil Hydraulic Conditions during Rainfall Events
Tensiometer measurements revealed that the water table elevation increased during rainfall simulations. Figure 9a shows the time course of the water table depth during the 30.5 mm h–1 rainfall event for the HWT and LWT initial conditions. In both cases, the water table rose to the soil surface. For the HWT initial conditions, the saturation occurred in the last 30 min of the event, generating surface runoff (Fig. 9b), contrary to the LWT conditions, where saturation was observed only at the end of the rainfall and did not generate runoff.


Figure 9
View larger version (12K):
[in this window]
[in a new window]

  		Fig. 9. Time course of (a) water table depth and (b) surface runoff measured during the 30.5 mm h–1 rainfall experiment for the low (LWT) and high (HWT) initial water table elevations.

 

   DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
CONCLUSIONS
 REFERENCES
 
Seal Formation and Initial Water Content
Initial soil water content influences the sealing processes (Le Bissonnais and Singer, 1992) and therefore the change in surface bulk density during seal formation (Fohrer et al., 1999; Bresson et al., 2004): slaking seal under initial dry conditions and infilling seal for wet conditions. In the present study, whatever the initial elevation of the water table, the water content was very high, which explains why most aggregates kept their integrity and only infilling seals were formed. Differences in the initial soil water content in relation to the water table elevation, however, induced differences in seal development. Under the wettest conditions, the seal was found to be denser ({Delta}{rho}0m HWT > {Delta}{rho}0m LWT; Table 5) and thicker (Fig. 4), but formed more slowly (βHWT < βLWT, Table 5). The total kinetic energy of the first rainfall, 325 J, was much greater than the second application, 86 J. We therefore assume that the main differences in the structure and bulk density of the seal sampled belonging to the same experiment, i.e., the same initial water table elevation, can mainly be ascribed to the high-intensity rainfall.

For a similar soil, rainfall intensity, and bulk density measurement method, Bresson et al. (2004) measured a maximal bulk density of 1.53 Mg m–3, which remains close to our measured values (mean value of 1.37 Mg m–3, SD = 0.08 Mg m–3, and 1.56 Mg m–3, SD = 0.13 Mg m–3, for the LWT and HWT conditions, respectively). The total rainfall kinetic energy was greater in their simulation (810 J m–2 for their study vs. 411 J m–2 for ours), however, and the initial water content maintained as sampled in the field was probably smaller than in our experiment. This might suggest that the more disturbing effect of kinetic energy was balanced by the greater cohesion of initial aggregates (Rajaram and Erbach, 1999). Using another measurement method, Roth (1997) found a maximum bulk density at the top 2 mm of loamy silt and silty loam seals varying between 1.44 and 1.64 Mg m–3. Our results are in the low range of Roth's data set, probably because of the great difference in rainfall kinetic energy (1547 J m–2 in Roth's study) not compensated by the moisture effect (equivalent to the field capacity, as in Bresson's work). Using the x-ray computed tomography measurement method, Fohrer et al. (1999) found an increase in surface bulk density under moist conditions at the magnitude of 0.37 Mg m–3, which falls within the range of our results (0.305–0.495 Mg m–3). Unfortunately, kinetic energy was not reported in that study.

Seal formation led to a decrease in the macrovoid index (–62%) and a simultaneous increase in the microvoid index (25%) near the soil surface in the upper 20 mm compared with the underlying layer (Fig. 8a). Therefore, the increase in bulk density near the soil surface ({Delta}{rho}0) corresponds mostly to a reduction in macroporosity. This can be explained by the infilling process, during which particles were packed in the interaggregate macrovoids, as checked by thin-section observations. The increase in the microvoid index near the soil surface is also consistent with the high packing porosity of infilled particles compared with aggregates. In other words, whereas the initial soil structure was composed of aggregates with low microporosity and interaggregate macrovoids, the structure of the infilling seal was characterized by particles with high packing microporosity filling the former macrovoids, which can explain the changes in both the micro- and macrovoid indices near the soil surface.

Because the seal formation process is the same whatever the initial water table elevation, the greater development of seals in the HWT than the LWT experiments should result from the lower cohesion of aggregates under wetter conditions. Cruse and Larson (1977) showed that soil particle detachment after a raindrop fall was negatively correlated to the shearing stress on the plane failure of a soil core, as defined by the Mohr–Coulomb theory and measured by a triaxial compression test. Based on their results, Assouline and Mualem (1997) suggested an empirical function to relate the shear strength per unit area of soil {tau} (g cm–2) to initial bulk density {rho}i (g cm–3) and initial water pressure head hi (cm):

Formula 5[5]
Since initial surface bulk densities are similar in the LWT and HWT experiments, {tau} at the soil surface will mainly be related to differences in hi and determined by the initial depth of the water table. As the profile was under hydrostatic conditions at the beginning of rainfall (hiHWT = –0.3 m and hiLWT = –0.7 m of water for high and low water tables, respectively), Eq. [5] leads to the ratio ({tau}HWT/{tau}LWT) = (hiHWT/hiLWT)0.13 = 0.89. Consequently, as {tau} is lower for the HWT conditions than for the LWT conditions, more aggregate destruction is expected under the HWT condition, as observed.

This estimated effect of the initial water table elevation on {tau} can be exploited to validate its measured effect on the dynamics of seal formation expressed by means of the parameter β in Eq. [3]. The parameter β is related to the rainfall properties as well as to {Delta}{rho}0m and {tau}({rho}i,hi) (Assouline and Mualem, 1997). Therefore, for similar rainfall conditions, β is inversely proportional to the terms [{Delta}{rho}0m, {tau}({rho}i,hi)]. According to the {Delta}{rho}0m and {tau}({rho}i,hi) estimates presented above, one can predict that βLWT > βHWT, in agreement with the observed trend. Moreover, the fitted value of β corresponding to the LWT experiment, βLWT, is 1.7 times the corresponding value βHWT obtained for the HWT experiment (Table 5), while the value for βLWT estimated using the {Delta}{rho}0m and {tau}({rho}i,hi) estimates is 1.5 times the corresponding value for βHWT. Considering that Eq. [5] is an empirical relationship that was not fitted to the soil used in this study, an error of only 12% on the predicted βLWT/βHWT ratio indicates that the dynamic model of soil surface sealing developed by Assouline and Mualem (1997) realistically accounts for the initial water content effect on infilling seal processes.

Thin-section observations of surface seal obtained after 40 min of the 30.5 mm h–1 rainfall showed that, when interaggregate packing voids were filled, raindrops compacted the infilled particles, reducing its porosity. Consequently, for these treatments, the increase in the microvoid index near the soil surface due to the infilling process was compensated by the effect of compaction, which explains why the difference between the microvoid index at 0- and 2-mm depths was down 68% (LWT) and 53% (HWT) between 15 and 40 min of rainfall duration (Fig. 8a). Measurements of rheological properties on wet soil aggregates taken by Ghezzehei and Or (2001) showed that, under wet conditions, yield stress and the plastic viscosity coefficient both decreased with increasing water content. Analyzing the effects of moisture content and load on compression of natural aggregate beds under uniaxial stress with oedometer tests, Guérif (1982) also illustrated that, after a moisture content threshold, the compressive behavior of particles depends mainly on internal cohesion, which decreases with increasing water content. In the present study, compression behavior could then be one factor explaining the difference in the maximum changes in surface bulk density, {Delta}{rho}0m, between the HWT and LWT treatments. The variability in aggregate size, distribution, and internal cohesion probably explains the variability between subsample bulk density profiles.

Slumping and Water Content Dynamics
Slumping did not affect microporosity, which remained rather constant below the surface seal (Fig. 8b). As revealed by thin-section observations, no aggregate microcracking occurred, which would have created microporosity, as observed on hard-setting soils (Bresson and Moran, 2004). On unstable soils developed on loess deposits in Western Europe, the slumping process involved was rather a coalescence under plastic conditions (Bresson and Boiffin, 1990; Kwaad and Mücher, 1994; Bresson et al., 2001) intensified by the high water content conditions, and which affected only macroporosity. Thin sections also displayed meniscus-like macropores, which suggest local reorganization of fine fragments and basic particles by capillary forces. This process is likely to occur during the draining phases (Mullins et al., 1990; Gusli et al., 1994).

The wetting and draining cycle of the seedbed before rainfall simulations contributed to the total slumping observed, and the resulting decrease in the 5-cm top layer mean bulk density was higher under the HWT than the LWT condition (Table 4). The effect of water table elevation on slumping can be related to the mechanical properties of soil aggregates: as water content remained higher during the HWT cycle than during the LWT cycle, internal aggregate cohesion was weaker (Rajaram and Erbach, 1999) and overburden pressure led to a greater coalescence. After the two rainfall applications, which also included two wetting and draining cycles, the seedbed was more affected by slumping than before rainfall simulations: the slope of the slumping model approximately doubled or tripled (Fig. 5). Since the period of high water content in the seedbed was shorter during the rainfall simulation than during the pre-rainfall wetting treatment, this greater structure collapse might be ascribed to the higher wetting and draining rates (Bresson and Moran, 2004). Moreover, seedbed wetting from above by infiltrating water may exert different forces on the aggregates than saturation from below by the rising water table. Again, the slope of the slumping model was relatively higher under the HTW condition than under the LWT condition. This difference also appears consistent because both initial water content and saturation duration were higher under the HWT condition. The respective roles of drying and wetting cycles and the water content, however, cannot be clearly delineated in this study.

The 30.5 mm h–1 rainfall duration did not significantly impact the slumping slope (Fig. 5), but this was not the case for the HWT3 condition, where the slumping parameter modeled was significantly higher. For this treatment, thin-section observations revealed a particularly large collapse, which may correspond to more slumping. This was also highlighted by the high bulk density values and the small macrovoid indices in the whole profile. For this reason, the macrovoid index at a depth of 10 mm was close to the value at the 20-mm depth (Fig. 8b). Indeed, the increase in bulk density due to slumping may not be linear with depth along the whole seedbed as assumed in Eq. [2]. It is assumed to reach a maximum corresponding to a macroporosity close to zero, as the relatively low overburden pressure near the soil surface does not affect microporosity. Then, when aggregate cohesion is very weak (unstable soil and high water content), substantial global slumping can lead to bulk density that is constant with depth, as observed in the coalescing seal (Bresson et al., 2004). The HWT3 sample presents a bulk density certainly close to this maximal value (Fig. 2f), approximately 1.6 Mg m–3, which corresponds to an increase of 0.53 Mg m–3 over the initial bulk density. As suggested by Or and Ghezzehei (2002), the proposed model could be improved to account for realistic dynamics of slumping under wet conditions. The modeled bulk density profile would depend on the wetting rate, the number of wetting and drying cycles, and the soil properties (aggregate cohesion) and would tend to a maximal bulk density value in cases where slumping is the main processes.

At the 25-mm depth, slumping led to a bulk density comparable to that resulting from soil surface sealing (Fig. 3). Since bulk density is expected to be related to the soil hydrodynamic properties (Assouline et al., 1997b; Stange and Horn, 2005; Assouline, 2006a, 2006b), infiltration would be controlled not only by the surface seal but also by the structural degradation of the whole seedbed. It would be interesting to apply such approaches to account for the effect on the infiltration process of the structural changes in the whole seedbed where the macroporosity is mainly affected.


   CONCLUSIONS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
REFERENCES
 
Bulk density profiles at the surface of a loamy seedbed exposed to simulated rainfalls were generated by calibrated x-radiography of resin-impregnated soil slices for two initial relatively high water content conditions (water table set at 0.3 and 0.7 m deep). The combined sealing and slumping model proposed by Bresson et al. (2004) (Eq. [2]) and the dynamic model of Assouline and Mualem (1997) (Eq. [3]) adequately reproduced the bulk density profiles: regression RMSE range 0.057 to 0.106 Mg m–3. A high variability between the subsamples was observed, however, particularly in the initial high water table experiment.

For the wettest conditions, sealing and slumping processes led to the highest soil bulk densities at the soil surface and within the underlying seedbed. Detachment of small particles from aggregates and compaction of infilled elements, which successively occurred at the soil surface, were actually directly influenced by the initial water content through the soil aggregate mechanical properties involved, namely aggregate cohesion and the compressive behavior of infilled particles under compaction. Within the seedbed, slumping, which involved aggregate coalescence and reorganization of fine particles by capillary forces, was influenced not only by the mean water content (or water pressure head), but also by the number of draining and wetting cycles. Microporosity (<0.1 mm in diameter) was not significantly affected by slumping; however, it increased at the soil surface during the infilling process and decreased during compaction by raindrops. In the high water content conditions studied here, soil moisture influences the degree of surface sealing and slumping but not the physical processes involved.


   ACKNOWLEDGMENTS
 
The study was conducted within the RIDES project "Ruissellement, Infiltration et Dynamique des Etats de Surface" (Surface Runoff, Infiltration and Dynamics of Soil Surface Characteristics), funded by the French "Programme National de Recherches en Hydrologie" (PNRH). We are grateful to the Soil Science Unit of INRA in Orléans for having provided the opportunity to use the rainfall simulator of their laboratory. The skilled technical assistance of C. Chaumont (Cemagref, Antony), L. Prudhomme, B. Renaux, and C. Le Lay (INRA Orléans) was very helpful. We also thank C. Gascuel for fruitful scientific discussions.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.

Received for publication December 12, 2006.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 





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 Google Scholar
Google Scholar
Right arrow Articles by Augeard, B.
Right arrow Articles by Vauclin, M.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Augeard, B.
Right arrow Articles by Vauclin, M.
Agricola
Right arrow Articles by Augeard, B.
Right arrow Articles by Vauclin, M.
Related Collections
Right arrow Structure and Properties
Right arrow Experiment Design
Right arrow Soil Models

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome