Image processing-based study of soil porosity and its effect on water movement through Andosol intact columns

Agricultural Water Management 96 (2009) 1377–1386

Contents lists available at ScienceDirect

Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat

Image processing-based study of soil porosity and its effect on water movement through Andosol intact columns B. Prado a,f,*, C. Duwig b, J. Ma´rquez c, P. Delmas d, P. Morales e, J. James d, J. Etchevers f a

Universidad Nacional Auto´noma de Me´xico, Instituto de Geologı´a, Ciudad Universitaria, 04510, Mexico, D.F., Mexico UMR 5564 LTHE/IRD, BP 53, 38041 Grenoble Cedex 9, France c CCADET, UNAM, A.P 70-186, 04510 D.F., Mexico d Department of Computer Science, The University of Auckland, Private Bag 92019, New Zealand e Laboratorio de Geoquı´mica Isoto´pica, Instituto de Geologı´a, UNAM, D.F., Mexico f Colegio de Postgraduados, Laboratorio de Fertilidad de Suelo, CP 56230 Montecillo, Mexico b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 15 September 2008 Accepted 12 April 2009 Available online 30 May 2009

The soil pore network and marcoporosity are important factors affecting water and solute transport. The transfer of contaminants to water resources is of particular importance in the Valle de Bravo watershed as it provides 10% of the drinking water for the 20 million inhabitants of Mexico City. This watershed is composed mainly of Andosols with unique mineralogical and physical characteristics. Soil porosity is usually examined on thin sections, using various image analysis techniques. We propose a novel methodology combining image analysis and a displacement experiment to study relationships between soil structure and water tracer transport parameters. H218O displacement experiments were conducted through intact soil columns sampled at three depths from a representative cultivated Andosol profile. The soil structure and pore characteristics were obtained by image analysis on thin sections obtained from each column at the end of the displacement experiment. The total 2D porosity (for pores larger than 50 mm) varied from 80% of the total section area in the topsoil to around 60% in the subsoil. Tubular pores were the most abundant in the soil profile, but ploughing of the topsoil had destroyed sections of these pores and replaced them with packing pores. Water transport in the intact subsoil columns was always in physical non-equilibrium, showing the existence of preferential flow pathways. In the topsoil, one column out of three showed no preferential flow, demonstrating that soil ploughing also homogenised pore connections. Pore connectivity was larger in the ploughed topsoil than in their deeper soil horizon counterparts. Our methodology offers a 2D quantitative characterisation of the macroporous network at 50 mm resolution and the determination of water transport parameters on the same intact soil samples. 3D characterisation of soil porosity using X-ray computed tomography (CT) gives a better picture of pore connection but usually has lower spatial resolution and a larger cost. ß 2009 Elsevier B.V. All rights reserved.

Keywords: Macroporosity Image analysis Displacement experiment

1. Introduction Solute transport through soil has become a key research topic since contamination of groundwater sources has been observed worldwide. The Valle de Bravo watershed in Mexico needs to be protected from further contamination as it provides 10% of the drinking water for the 20-million inhabitants of Mexico City. Water and solute transport through soil is a complex process that can be directly related to the pore network. Both soil porosity

* Corresponding author at: Universidad Nacional Auto´noma de Me´xico, Instituto de Geologı´a, Ciudad Universitaria, 04510, Mexico, D.F., Mexico. Tel.: +52 56 22 42 86x159. E-mail address: [email protected] (B. Prado). 0378-3774/$ – see front matter ß 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.agwat.2009.04.012

and other soil characteristics such as structure and texture affect transport processes (Strock et al., 2001). Literature on soil structure characteristics and its relationship with solute transport is ambiguous. Seyfriend and Rao (1987) reported that solute dispersion is related to the soil structure and its water content. Bejat et al. (2000) observed that in an unsaturated soil, there is a linear relationship between the soil water content and the dispersion of a non-reactive solute, as well as between the dispersion and the pore water velocity. They did not find a direct relationship between the soil’s structural properties and hydrodynamic dispersion. The pore network, which depends on the soil structure, plays a decisive role in water and solute movement through soil. Walker and Trudgill (1983) found significant correlations between geometric variables describing soil porosity and solute transport parameters. For example, the dispersivity

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

1378

coefficient is exclusively determined by pore geometry (Yule and Gardner, 1978). Poletika and Jury (1994) demonstrated that the presence of macropores is one of the factors responsible for heterogeneity in water and solute movement through soil. Some authors have related saturated hydraulic conductivity and residual water content to pore size distribution in soil (Holtham et al., 2007). Macropore networks differ depending on the soil type, morphology, agricultural practices, and faunal activity. Active (i.e. functional) pores change with water content and water pore velocity (Andreini and Steenhuis, 1990; Shipitalo et al., 1990; Edwards et al., 1992; Quinsenberry et al., 1994). Soil thin section analysis (e.g. Walker and Trudgill, 1983) and dye application in intact soils (Seyfriend and Rao, 1987; Hatano et al., 1992; Vanderborght et al., 2002; Tarquis et al., 2006) are the most common techniques for studying soil structure, porosity, and transport parameters. The application of image analysis techniques to determine pore characteristics and soil structure in thin sections has become an indispensable tool for research in soil science (Protz et al., 1987). The characterisation of the porous network from the image analysis of soil thin sections allows an independent and direct evaluation of the water dynamic in a structured soil (Hallaire et al., 1997, 1998; Cervantes et al., 2003; Holtham et al., 2007). It has been used to measure the pore size distribution in the various soil horizons (Ismail, 1975), to characterise the orientation, shape, and size of different pores (Murphy et al., 1977), and to quantify dye transport in preferential flow pathways (Forrer et al., 2000; Duwig et al., 2008). Image analysis has also been used to determine solute transport parameters (Persson, 2005) and to explain the shape of breakthrough curves obtained from solute transport experiments in soil columns (Walker and Trudgill, 1983; Sugita et al., 1995). One technique for evaluating solute transport and sorption processes in soil is the displacement of the solute through a soil column. The soil is immobile, and the solute moves through the column only once. Transport and sorption processes are affected by the soil’s structure and other properties. Leachates at the bottom of the column are collected and analysed; the breakthrough curve shape is determined by the different processes occurring during the displacement of the solute through the soil matrix (e.g. adsorption and preferential flow). Experimental studies where the breakthrough curve and soil structure characteristics are determined in the same intact soil sample are scarce, and have only been conducted using expensive non-destructive technologies such as soft X-ray radiography (Mori et al., 1999), and a combination of CAT and SPECT scanning (Perret et al., 2000). In the case of Andosols they are nonexistent. This type of study allows the determination of a direct relationship between solute transport parameters (by applying a water tracer as the solute of interest) and the soil pore network (by analysing soil thin sections) (Sugita and Gillham, 1995; Sugita et al., 1995; Bejat et al., 2000).

Thanks to their unique physical characteristics (e.g. low bulk density, high water retention capacity, and usually a high content of organic matter), which derive from the presence of amorphous materials, Andosols usually offer good conditions for agriculture and can support high population densities. However, the presence of these amorphous materials renders their study quite complex and requires the use of adapted methodologies. Andosol structure and porosity has to be studied at field moisture to avoid the irreversible formation of aggregates, when drying. Its dark colour makes it impossible to use dyes other than fluorescent ones. In the current study, displacement experiments were conducted with the water tracer H218O through intact columns sampled at different soil depths of an Andosol profile. The soil structure and pore network were obtained by image analysis on thin sections obtained from each column once the displacement experiment concluded. The image analysis technique consisted of three steps: image segmentation, identification of pores (i.e. labelling), and the calculation of geometric and morphologic parameters, namely their perimeters, areas, shapes (through their shape factor), bi-dimensional connectivity and tortuosity. The specific objectives of this work were to analyse the soil pore network and its variation with depth within the Andosol profile, and to evaluate the relation between these soil properties and water transport processes. 2. Materials and methods 2.1. The soil studied The soil studied is located in the elementary catchment la Loma, part of the Valle de Bravo basin in Mexico State, 150 km west of Mexico City, at a height of 2500 m (198160 48.600 N and 998580 13.700 W). La Loma has been the location for various studies on agrochemical transport through the soil vadoze zone (e.g. Prado et al., 2006; Duwig et al., 2006, 2008; Mu¨ller and Duwig, 2007). The Valle de Bravo basin is the most important reservoir of the Cutzamala system, which provides a significant amount of drinking water (19 m3 s1 or 21% of the daily supply, Tortajada and Castela´n, 2003) to Mexico City. A plot under maize in the la Loma catchment was selected, and the water balance, runoff, erosion, and nutrient losses monitored. The soil was characterised and classified as a Pachic Andosol (WRB, 2001). The physical, chemical, and mineralogical characteristics are described in detail in Prado et al. (2007). The whole soil profile presents andic properties: bulk density < 0.9 g cm3, phosphate retention  70%, Alox + (1/2)Feox  2%, volcanic glass content in the fine earth fraction < 10%. Table 1 shows some selected properties of this soil. The unsaturated soil hydraulic conductivity (at h = 100 mm) and saturated conductivity vary with depth: they decrease from 0.05 and 0.11 cm min1 at the soil surface to 0.012 and 0.08 cm min1 at 55 cm depth (Prado, 2006).

Table 1 Selected soil properties. Depth (cm)

SOCa (g/kg)

0–15 15–20 20–45 45–65 65–85 85–110

54 53 56 53 47 51

a b c d

Texture Sand

Silt (%)

Clay

29 45 23 25 26 22

62 50 66 66 63 68

9 5 11 9 11 10

SOC, Soil Organic Carbon. CEC, cation exchange capacity (cmolc kg1). WC, 15 bar water content. Allophane = 6  Si extracted by oxalate (Parfitt, 1990).

CECb

WCc %

pH H2O

Allophaned (%)

Horizon

22.3 23 20 24 23.1 23.6

25.5 26.9 30.5 31.1 37.9 33.2

5.5 6.1 6.2 6.3 6.3 6.5

18.5 23.1 22.5 25.5 26.0 27.7

Ap A1 A2 2A1 2A2 3A

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

1379

Table 2 Experimental column parameters. Column

Depth (cm)

Length (cm)

rsa (Mg m3)

rb (g cm3)

Porosityc (%)

ud (cm3 cm3)

Darcy flow (cm cm1)

Voe (cm3)

C1 C2 C3 C4 C5 C6 C7 C8 C9

5–30 5–20 5–20 30–45 30–55 30–53 80–95 80–105 80–95

25 15 15 15 25 23 15 25 15

2.44 2.44 2.44 2.60 2.60 2.60 2.39 2.39 2.39

0.79 0.74 0.63 0.61 0.79 0.82 0.70 0.69 0.60

0.68 0.70 0.74 0.77 0.70 0.68 0.71 0.71 0.75

0.66 0.83 0.70 0.82 0.77 0.79 0.88 0.77 0.84

0.032 0.008 0.005 0.004 0.016 0.013 0.012 0.016 0.005

392 296 250 291 459 424 314 458 298

a b c d e

rs = soild density estimated by Mercury Intrusion Porosimetry. r = bulk density. Porosity = 1  (r/rs). u = water content. Vo = pore volume.

2.2. Water displacement experiments Displacements experiments using the water tracer H218O (normalized water with 95.1% H218O, Euriso-top, Gif-Sur-Yvette France) were conducted on intact soil columns. Nine soil columns were excavated in the field at three different depths: between 5 and 30 cm, between 30 and 55 cm, and between 80 and 105 cm. The columns were excavated by carefully pushing a cylindrical PVC tube with a length of 25 cm and an internal diameter of 5.5 cm. The good contact and the absence of any artificial pore space in the interface between the PVC tube and the soil was verified using a CAT scan (Delmas, personal communication). The bottoms of five out of nine soil columns were cut to obtain columns 15 cm long in order to reduce the flow velocity without increasing the experiment time. The different column lengths are specified in Table 2. The columns were stored at 4 8C until the experiments. Columns were sampled across several horizons depending on the length and depth of the columns: column C1 across Ap, A1 and A2 horizons, columns C2 and C3 across Ap and A1, column C4 across A2, columns C5 and C6 across A2 and 2A1, columns C7, C8 and C9 across 2A2 and 3A. During the displacement experiments, a rainfall simulator made of nine hypodermic needles and a reservoir was used at the top of the soil columns. The columns were oriented vertically and two peristaltic pumps (one at the top feeding the solution into the simulator and one at the bottom sucking the leachates) created a constant flow through the columns. At the base of the column there was a grid to maintain the soil. The space between the pump and the soil core was hermetically closed to maintain the suction. In order to analyse the effect of the flow on the hydro-dispersive parameters, different Darcy flows were applied, varying from 0.004 to 0.03 cm min1. The same Darcy flow was applied to two columns out of three per depth. However, due to the heterogeneity of the structure and porous network of the intact soil, it was not possible to inject exactly the same flux in each column of the same depth and thus have exact repetitions. Flux was checked at the entry and exit of the columns by regularly weighing the applied solution and the collected samples. Table 2 presents the length, the bulk density, the soil porosity, the final water content, the Darcy velocity, and the pore volume of each column studied. For seven out of the nine columns, the volumetric water content was larger than the porosity calculated from the bulk and solid density. This result means that the soil in the columns was saturated and in some cases, water was probably entrapped due to soil stratification (Table 1). At the beginning of the experiment, a solution containing the same major cations (2.4 mM of CaCl2, 0.2 mM of MgCl2 and 0.1 mM of KCl) as the field soil solution was injected during two to three pore volumes so as not to alter the natural ionic strength of the soil.

Electrical conductivity of the leachates was monitored at the bottom of the column. Once a constant value of this parameter was reached and the flux became stable, about half a pore volume of the water tracer H218O diluted in the same cationic solution was injected at the top of the column. This was followed by several pore volumes of the cationic solution, until the electrical conductivity reached its initial value. Samples were automatically collected for H218O analysis by mass spectroscopy (Thermo Finnigan MAT 253 according to the method of Epstein and Mayeda, 1953). The code CXTFIT 2.1 (Toride et al., 1999) was used in inverse mode to obtain the hydrodynamic parameters of the convection dispersion equation (CDE) coupled with the mobile–immobile model. An analytical solution of the CDE was fitted to the experimental data by the least-squares optimisation method. By setting the retardation factor R to 1, with H218O being an inert tracer, values for the dispersion D (cm2 min1), the mobile water fraction f(um/u), and the dimensionless mobile–immobile region exchange coefficient (v) were obtained. The first-order mass transfer coefficient (a, s1) that represents the rate of solute exchange between the mobile and immobile liquid regions is estimated by the equation:

a¼

vun L

;

(1)

where n = average pore water velocity (cm min1), L = column length (cm). The characteristic exchange time ta (min) between the mobile and immobile regions is calculated as the ratio between immobile water content (uim) and a. The resident water time tw (min) is defined as the time for one pore volume to travel through the soil column. 2.3. Thin sections preparation and image acquisition Soil structure and porosity were analysed on thin sections extracted from the same intact columns used in the displacement experiments. Once the experiments finished, the soil column was cut into 5 cm long segments. The intact samples were then placed in a closed chamber on a 7 cm high grid placed above a plastic container filled with a 20% acetone solution. Every week the acetone solution was replaced by a new acetone solution 20% more concentrated and the acetone concentration in the old solution was analysed. The water contained in the soil pores was gradually replaced by acetone, and the whole pore network was considered to be filled with acetone when the acetone concentration in the plastic container did not vary for one whole week. The entire process took about 6 weeks. The intact samples were then impregnated with a polyester resin by applying a vacuum to ensure the resin penetrated every pore. The impregnated samples

1380

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

were dried at room temperature for 4 weeks before being cut into vertical sections. These were polished down to a thickness of 30 mm. A set of three thin sections were obtained and analysed for the intact columns C1, C5, and C8 (see Table 2). The thin sections were digitized with an optical scanner (Hewlet Packard Scanjet 7400c) to obtain a digital image with an 11 mm-per-pixel resolution. This pixel length was experimentally confirmed using a measuring scale, placed on the thin section during scanning. The minimal size of the pores visible on the image was about 11 mm. Due to image processing segmentation (low-pass 3  3 filter) and mathematic morphology filters (opening/closing procedures) most parameters could only be computed for pores which spanned more than five pixels across. Therefore only pores larger than 50 mm (the widely accepted lower threshold characterising macroporosity) could be analysed. 2.4. Digital image processing Digital image processing of the thin slices was completed in several stages (Fig. 1). In the first stage the scanned images of soil sections (Fig. 1a) were segmented (Fig. 1b). They were separated into binary components, according to a criterion based on intensity and colour thresholds. Automatic segmentation algorithms were tried, but failed as the grey levels of pixels representing pores in similar materials changed considerably among images. Classical techniques of image segmentation did not provide useful results. For this reason we employed a semi-automatic procedure known as ‘‘Colour Prediction’’ (Barton and Delmas, 2002). First, pixel samples from pores and non-pores were selected manually (using the mouse to choose specific pixels), in order to define a 2D matrix of colour prediction for the Hue and Saturation colour channels (Duwig et al., 2008). Empirical tests showed that the HSL colour space (Barton and Delmas, 2002) offered better results than the original RGB colour space. A large Gaussian window was applied as a low-pass filter incrementing the neighbourhood of each pixel classified as pores, while a narrow, negative-weighted window decremented the neighbourhood of pixels classified as non-pores. The Gaussian smoothing, as well as the inclusion of pixels not

belonging to the porous network (by means of negative values) considerably enhanced the segmentation results. The colour predictor in the Hue-Saturation space generated a table of the positive and negative values of Hue-Saturation. Separation of the pore and non-pore regions was achieved by thresholding the HueSaturation colour predicate plane (as positive–negative) to create a binary mask image. The colour white was assigned to pores and black to no-pores (soil) for further processing (Fig. 1b). The mathematical morphology operators (erosion, dilation, opening, closing, etc., see Serra, 1982) were applied to these binary images to filter out misclassified pixels and compute the parameters which characterise the pore network. Next, the two-pass sequential labelling algorithm (Haralick and Shapiro, 1985) was applied to classify all pore pixels into labelled connected components (represented by different grey-level values in Fig. 1d). The iterative algorithm propagated labels from top to bottom, starting at the left superior corner and moving down to the right inferior corner of the image. A new label was assigned when, for a given pixel, its top, top-left, and left neighbours (on a 3  3 neighbourhood window) were not labelled (i.e. did not belong to the pore pixels). Otherwise the smallest label of the three neighbourhood pixels was assigned to the pixel studied. Following this, a second propagation from bottom to top replaced redundant labels. The algorithm converged after two iterations. Once all pores were labelled, their individual perimeter and area were calculated. Other parameters such as the shape factor and the bi-dimensional connectivity were also calculated. Discrete border tracking using the so-called criterion of ‘‘8-connectivity’’ was used to calculate the perimeter of each pore. This considers that boundary pixels are connected either by edge or by vertex. The tracking took into account eight possible orientations, summing up two different distances between neighbours: the vertical–horizontal distance (equal to 1) and the pffiffiffi diagonal distance (equal to 2). 2.5. Analysis at the micro-morphological scale The parameters selected to describe the soil porosity are those reported in the bibliography: the superficial porosity in 2D, the

Fig. 1. Stages of digital image processing of the soil core thin sections: (a) scanned image of soil thin slices; (b) segmented image: white colour was assigned to pores and black to soil (no-pores); (c) distance function regarding the bi-dimensional connectivity; (d) individual pore labeling; (e) labelling last eroded (Nu)

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

1381

pores size and shape, their connectivity, and tortuosity (Hallaire et al., 1997, 1998; Cervantes et al., 2003; Holtham et al., 2007). 2.5.1. The 2D porosity Two-dimensional porosity corresponds to the ratio between the total area occupied by the pores and the total area of the crosssection in the image. It is the count of pore pixels divided by the overall count of pixels in the image. 2.5.2. The pore size The pore size was estimated from the area of its cross-section in the image (first in pixels, then converted to mm2). The pores were classified into four size classes: class 1, area > 1 mm2; class 2, 0.09 mm2 < area < 1 mm2; class 3, 0.01 mm2 < area < 0.09 mm2; class 4, area < 0.01 mm2. Using the pore size computed from the image, the equivalent diameter (assuming all pores to be circular) was derived from their computed area. All pores larger than 50 mm in diameter or with an equivalent area of 0.002 mm2 were considered to belong to the macropore class. 2.5.3. The pore shape The shape of each pore was estimated with the longitudinal index Ia (Coster and Chermant, 1985), estimated from the area A and the perimeter P of each pore in the image with the equation:

Ia ¼

P2 4pA

Fig. 2. Tortuosity. Geodesic length (plain line), direct path (dotted line).

paths coincide) and takes larger values, for pores that produce complex paths. 3. Results and discussion 3.1. Water displacement experiments: breakthrough curves analysis Fig. 3 shows the adimensional experimental breakthrough curves (BTCs), the simulated ones, and the injected H218O pulse for three columns: (a) at 5–30 cm (C1), (b) at 30–55 cm (C5) and (c) at 80–105 cm (C8) depth (see Table 2). The H218O pulse duration

(3)

This index also known as shape factor takes the minimum value of 1 for a perfect round pore and increases as the pore is longer or has an irregular profile. It is suited for the distinction of different categories of pores (Hallaire and Cointepas, 1993). Considering the classification proposed by Ringrose-Voase (1996), three categories of pore shapes were defined: tubular pores, Ia < 5; fissure pores, 5 < Ia < 10; and packing pores, Ia > 10. 2.5.4. The pore bi-dimensional connectivity This parameter, introduced by Hallaire et al. (1997), was estimated with an original index, exploiting the notion of the ‘‘ultimate eroded’’, proposed in mathematical morphology by Serra (1982). On a binary image (Fig. 1b), pores are treated as particles (in white) and the material as background (in black). The connectivity number Nc corresponds to the number of ‘‘connected particles’’. The number of the last eroded, Nu, corresponds to the number of convex components in porosity (Fig. 1c and e). These two numbers allow the definition of the index of bi-dimensional connectivity Ic by Ic ¼

1  Nc Nu

(4)

Ic varies between 0 and 1; it is small if the pores are isolated and increases its value when the pores are interconnected or forming links. 2.5.5. Tortuosity Tortuosity is defined as the length of the true path followed (geodesic length) between two points (Fig. 2, plain line) divided by the apparent or direct path (Fig. 2, dotted line) between those two points. It is the shortest path entirely contained inside the pore, divided by the shortest Euclidean path (that is, the straight line), between the ends of the first path. The mean tortuosity gives an idea of the sinuosity of the pore network. This parameter is equal to 1 for a circular or rectilinear pore (when Euclidean and Euclidean

Fig. 3. Injected H218O pulse, and observed and simulated dimensional H218O breakthrough curves for three soil columns (a) 5–20 cm depth (C1), (b) 30–45 cm depth (C5), and (c) 80–95 cm depth (C8).

1382

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

varied from one column to the other, because of the difficulty in establishing the same flux in every column. Before injecting the pulse, the Darcy velocity was calculated from the flux at the column entry. This flux was later corrected as the mean flux between the column entry and exit. The BTCs were asymmetrical in all the extracted columns except for column C1. Column C1’s BTC was symmetrical (Fig. 3a), with a gravity centre located at one pore volume after half of the H218O pulse was injected. This signifies that the water tracer movement in C1 column was not retarded and was in physical equilibrium. Asymmetrical BTCs for an inert water tracer (retardation factor equals 1) means the existence of physical non-equilibrium, or preferential fluxes. The two types of water transport, physical equilibrium and nonequilibrium were only found at the soil layers surface (columns C1–C3). This was due to the structural differences observed at this depth. In column C1, water movement in physical equilibrium means that the soil was homogeneous. In columns C2 and C3, water movement in physical non-equilibrium signifies the existence of preferential pathways caused by well connected macropores (fissures, pores due to soil fauna and decaying roots). One explanation for these two types of water transport is the effects of soil ploughing. Columns C1–C3 were excavated in a plot sown with maize where the ploughing affected the first 20 cm of the soil profile. Ploughing destroys the aggregates and mixes and homogenises the soil; process that can be mimicked by sieving the soil at a uniform particle size. Prado et al. (2006) demonstrated that water movement in packed columns with the same soil occurred in physical equilibrium, due to sieving and homogenising the soil before packing it. The excavation of the intact columns in the present study was done 5 months after the soil was left bare and nearly a year after the last ploughing. Roots from the previous maize crop and from weeds growing afterwards, plus the reorganisation of soil fauna during these 5 months are factors that can explain the formation of preferential pathways in some places in the first 20 cm of the soil profile. Another explanation of the contrasting results between columns C1 and C2/C3 can be related to soil stratification. Soil stratification was more important in the 5–30 cm length C1 column, which included three soil horizons (Ap, A1, A2) than in the 5–20 cm C2 and C3 columns that contained only two soil horizons (Ap and A1). These two horizons are characterised by similar physical characteristics (Table 1). However, the last 10 cm of the C1 column came from the 20–45 cm A2 soil horizon which retains much more water than its Ap/A1 topsoil counterpart, because of its larger clay content (Table 1) (Bartoli et al., 2007): this bottom horizon of the C1 column therefore regulated water transport in this column. 3.2. Modelling parameters Table 3 shows the parameters obtained from inverse modelling of the experimental data. 3.2.1. Dispersivity We calculated the Peclet number (Pe) from the equation nL/D (where n is the pore water velocity in cm min1 and L is the column length in cm) to compare the relative importance of molecular diffusion and convective dispersion in the columns. Kutı´lek and Nielsen (1994) presented four zones for the Pe values, defining the relative importance of both processes: zone 1: Pe < 0.3, zone 2: 0.3 < Pe < 5, zone 3: 5 < Pe < 20 and zone 4: Pe > 20. In zones 1 and 2, molecular diffusion is the most important process, while in zones 3 and 4, it is weak. The Pe values obtained in this study were superior to 5 for all columns (Table 3). We concluded that in the range of pore water velocities obtained in our experiments, the molecular diffusion was weak compared to the convective

Table 3 Model fitted column parameters. Column

C1 C2 C3 C4 C5 C6 C7 C8 C9 a b c d e f

Depth (cm)

Pea

5–30 5–20 5–20 30–45 30–55 30–53 80–95 80–105 80–95

25 8 5 5 15 11 16 20 5

lb

f = um/uc

tad (min)

twe (min)

Modelf

1.00 0.80 0.70 0.66 0.73 0.78 0.75 0.76 0.72

– 1657 2272 2379 795 685 667 577 2352

– 1026 1254 1084 534 487 461 544 1073

1 2 2 2 2 2 2 2 2

(cm) 0.98 1.9 3.1 2.9 1.6 2.4 1.9 1.2 2.8

Pe: Peclet number. l: Dispersivity. f = um/u: mobile water fraction. ta: exchange characteristic time. tw: water resident time. Model 1: in physical equilibrium, 2: in physical non-equilibrium.

dispersion and that the dispersivity coefficient (l) could be calculated from the equation: D/n. When comparing adimensional breakthrough curves, wider curves with lower peaks indicate larger soil dispersivity. Fig. 3 compares the C5 curve (30–55 cm depth) with the C8 curve (80–105 cm depth). The dispersivity values are shown in Table 3. The structural differences found in the surface columns are also evidenced by the differences in dispersivity values. In column C1 where water movement was in physical equilibrium, the soil dispersivity was 0.98 cm (Table 3), a value which stands at the lower limit for an intact soil, according to Magesan et al. (1995). All the other columns had dispersivity values between 2.3  0.66 cm (at 30–55 cm depth) and 1.96  0.8 cm (at 80–105 cm depth). These results are similar to the dispersivity found by Rao et al. (1980). These authors found a dispersivity value of 2.14 cm for a well structured and aggregated soil. The soil from la Loma catchment had a homogeneous loamy texture through the whole profile. However soil ploughing of the first 20 cm affected the structure as well as some physical and chemical characteristics. Prado et al. (2007) observed that the predominant structure was subangular blocky, with a loose structure (fluffy) at the surface. 3.2.2. Mobile water fraction Table 3 shows the different f values in all studied columns. In column C1 physical equilibrium water transport meant the whole soil water content was mobile (f = 1), while in the other columns the mobile fraction varied from 0.66 to 0.80. This parameter explains the exit of most of the water tracer H218O before one pore volume in columns C5 and C8 (see Fig. 3). Columns C5 and C8 had f values of 0.73 and 0.76, respectively. This means the gravity centre of the H218O BTC was located before one pore volume. The differences in l and f values also modified the breakthrough curves’ shapes. The f value depends on the soil structure and texture. Vanderborght et al. (2000) found that the volume fraction of the preferential flow region is roughly in the order of 1–3% in a loam soil profile. For a silt loam soil with subangular blocky structure, Lee et al. (2001) reported values of f ranging between 0.42 and 0.82; Kamra et al. (2001) found an f value of 0.75 in a silt loam soil. 3.2.3. Mobile–immobile region exchange coefficient The mobile–immobile region exchange coefficient (a) is affected by the pore water velocity (n), the aggregates’ size, and the length of the pathway the water molecules have to travel (Pallud, 2000). We found the exchange coefficient increased proportionally with the pore water velocity. The correlation coefficient between a and n was 0.93. Despite being affected by

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

1383

the pathway length, the column length did not affect the a value. For example, at 80 cm depth, a varies from 0.00033 min1 for a 15 cm long column to 0.00032 min1 for a 25 cm long column. This result signifies that the length of porous pathways was small compared to the length of the columns studied, confirming the volume representativeness of the columns. Table 3 shows that for the columns in physical non-equilibrium, the exchange characteristic time (ta) was greater than the water residence time (tw). The water did not stay inside the soil column long enough to reach physico-chemical equilibrium with the soil. Preferential flows reduce the contact time between the water and soil matrix. 3.3. Analysis at the micro-morphological scale: porous network characteristics from thin section analysis Pore size and shape factors (averages and standard deviations) for each pore class (as defined in Section 2) are presented in Table 4 and discussed hereafter. These values were obtained from three thin sections per column: only columns C1, C5 and C8 were studied. We calculated the overall number of pores (with a diameter larger than 11 mm) found in the thin sections for each depth, and the percentage of macropores (pores with an equivalent diameter larger than 50 mm or with an area larger than 0.002 mm2). The maximum pore count occurred at 5–30 cm (80% of the thin section), decreased at 30–55 cm (40%), and increased slightly at 80–105 (57%). The macroporosity variations followed the same pattern. The soil parameters estimated through 2D image analysis are representative only if the soil is an isotropic medium. Indeed, the porosity values obtained from the 2D image analysis of the thin layers and those determined in a three-dimensional space (estimated from the bulk and solid density) are of the same order of magnitude. This result validates the hypothesis of the studied soil being an isotropic medium. Table 4 shows average pore size and shape factors for each depth studied. Across all depths the percentage of pores in class 1 (largest pores) was the lowest and the percentage in class 4 (smallest pores) was the highest. The soil macroporosity was characterised by a majority of relatively small pores (area < 0.01 mm2) and 96% of all

Table 4 Average and standard deviation (in bracket) of the pore size and shape for each depth studied and for each pore class. Depth (cm) 5–30 a

c

30–55

80–105

0.78 (0.56) – – 100

0.41 (0.15) – – 100

Class1 (%) tpb fpb ppb

0.84 (0.23) – – 100

Class2a (%) tpb fpb ppb

3.2 (0.48) 7 (1.2) 25 (5.3) 68 (6.5)

4.4 (1.6) 18 (2.6) 36 (4.3) 46 (16)

4.1 (0.97) 7 (4.5) 40 (3.4) 53 (7.9)

Class3a (%) tpb fpb ppb

34.8 (2.6) 77 (2.3) 21 (2.6) 2 (0.8)

38.5 (3.9) 81 (4.5) 17(3.9) 2 (0.7)

39.8 (8.9) 76 (4.5) 21 (3.1) 2 (0.1)

Class4a (%) tpb fpb ppb

61.2 (2.69) 100 – –

56.3 (4.45) 100 – –

56.1 (13.5) 100 – –

a Class 1, pores area > 1 mm2; class 2, 0.09 mm2 < pores area < 1 mm2; class 3, 0.01 mm2 < pores area < 0.09 mm2; class 4, pores area < 0.01 mm2. b tp: tubular pores; fp: fissure pores; pp: packing pores. c Standard deviation.

Fig. 4. Pore size distribution expressed as area in mm2, for each depth and each class.

pores were smaller than 0.09 mm2. These results are similar to results found by Oleschko and Chapa Guerrero (1989) and Cabrera Carvajal and Okeschko (1995), which are the only studies on pore sizes of Andosols we are aware of. They studied the topsoil of a mollic Andosol either on thin sections or using the water retention curve and found that the majority of macropores (detection limit: 0.5 mm in pore radius) had equivalent pore radii between 15 and 150 mm (i.e. our pore size classes 3 and 4). In general, pore size did not change significantly with depth. Fig. 4 shows the pore size distribution expressed as area in mm2, for each depth and class. Fig. 5 shows the shape distribution expressed in number of pores according to the classification of Ringrose-Voase (1996) presented previously. Tubular pores are by far the most abundant across the three depths. At 80–105 cm, the fissure pores are more abundant than in the other depths. By determining the pore shape distribution in each pore size class we made the following observations (Table 4). In class 1 (the largest pores) there are only packing pores. In class 2 packing pores are the most abundant, along with some fissure pores and a minimal amount of tubular pores. In class 3, we observe the contrary; tubular pores are the most abundant, with some fissure pores and a lower proportion of packing pores. In class 4 (the most abundant class), there are only tubular pores.

Fig. 5. Pore shape distribution according to the classification of Ringrose-Voase (1996), for each depth and each class.

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

1384

Table 5 Summary of thin section and column parameters for columns C1, C5 and C8. Column

Depth (cm)

la (cm)

fb (um/u)

ac (min1)

nd (cm min1)

Ice

Tortuosity

C1 C5 C8

5–30 30–55 80–105

0.98 1.6 1.2

1 0.73 0.76

– 0.00026 0.00032

0.05 0.02 0.02

0.23 (0.01) 0.18 (0.09) 0.12 (0.03)

1.27 (0.26)g 1.17 (0.19) 1.2 (0.24)

a b c d e g

l: dispersivity. f: um/u: mobile water fraction. a: first-order mass transfer coefficient. n = pore velocity. Ic = bi-dimensional connectivity. Standard deviation.

The tortuosity did not vary between pore size classes (results not shown). The average value was 1.2 (Table 5) for all depths. The bi-dimensional connectivity Ic decreased significantly between the surface and the deepest layers. 3.4. Comparison between soil properties, porous network characteristics and hydro-dispersive parameters Tubular pores were the most abundant pores found on thin sections from all depths. The existence of amorphous minerals and the high organic matter content in Andosols make their structure granular and very porous (Nanzyo et al., 1993), with an important proportion of intra-aggregate pores belonging to size classes 3 and 4. In these classes, we found mainly tubular pores. Tubular pores have a lower mean area (the majority are in class 4, thus the area is lower than 0.01 mm2) than fissure pores and packing pores. During water transport, these tubular pores of small size act like a water reservoir (Walker and Trudgill, 1983), contributing to the large plant available water holding capacity (Shoji et al., 1993). The results obtained from the displacement experiments at 5–30 cm depth showed that soil ploughing increased the pore connection, as we found that in one column (C1) all the water was mobile. A result that coincides with a larger value of the bi-dimensional connectivity at 5–30 cm than for the deeper layers (Ic, Table 5). In the deeper layers, there is a smaller amount of macropores and we found preferential flow in all columns at 30–55 and 80–105 cm depth. Preferential pathways are created by well connected pores that by-pass areas of immobile water where pores are not well connected. This can be related to the greater amount of fissure pores found at 80–105 cm. It did not affect the Ic value. Hallaire et al. (1997) observed that the method for calculating Ic was not as well adapted to fissure pores as to pores with other shapes. The method can potentially consider several connected fissure pores as a single pore, which would artificially decrease the Ic value for a fissured soil.

Fig. 6. Relationship between the mobile water fraction coefficient and the Darcy flow.

In spite of the small variation in the number of packing pores between the three depths studied (Table 4, classes 1 and 2), these are still more important in the 5–30 cm layer. Packing pores are the result of the destruction of well defined pores such as fissures or tubular pores by agricultural practices (Gutie´rrez, 2004). Hubert et al. (2007) evaluated the relative influence of biological and mechanical processes on the structure of cultivated soils and found that mechanical soil ploughing produces continuous packing pores. We mentioned above that the mobile–immobile region mass exchange coefficient (a) is affected by the pore water velocity, the soil aggregates size, and the transport pathway (which can be described by the tortuosity). At the macroscopic level we have shown that the column length did not affect the a value. The similar tortuosity values found in columns at 30–55 and 80–105 cm depth with similar velocity and different column lengths resulted in similar a values. Similar tortuosity values were also related to similar soil structure. A positive linear relationship (R2 = 0.8) was found between the mobile water fraction coefficient and the Darcy flow: the f value increases when Darcy flow increases (Fig. 6). The immobile water in the soil column is retained by capillarity, absorbed and trapped in the air bubbles or in the pores. The effect of the partition between mobile and immobile water fraction is accentuated by the desaturation process, as the soil water content decreases, the immobile water fraction increases because the mobile water is the moving water. 4. Conclusions The objective of this study was to investigate potential relationships between hydro-dispersive parameters (obtained from the displacement experiments in intact columns), and the morpho-geometrical parameters describing the soil core porous network (as obtained from thin section image analysis). The soil’s hydro-dispersive parameters were determined by displacement experiments in intact columns conducted with the water tracer H218O. Using image analysis, the morphological and geometrical parameters (surface porosity, pore size distribution, pore shape, tortuosity, and connectivity) of the soil porous network were computed on the soil columns thin sections. The total 2D porosity (for pores larger than 11 mm) varied from 80% of the total section area at 5–30 cm depth to around 60% at 80 cm depth. Tubular pores were the most abundant in the whole soil profile. This can be linked to the important water retention capacity of Andosols. Soil ploughing increased the porosity and pore connection in the first 30 cm compared to the deeper layers. Soil ploughing was associated with the disappearance of preferential flow in one surface column, and a greater bi-dimensional connectivity in the first depth studied. It also destroyed the pores of well defined shapes

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

(tubular and fissures), replacing them with packing pores. Preferential flow, found in all but one column, was associated with a greater occurrence of fissure pores in the deepest layers. The combined method proposed in this study allows quantitative characterisation of the macroporous network and the evaluation of its role in water and solute transport. This article is the first to present a methodology coupling image analysis and displacement experiments on an Andosol. This method is particularly suited to evaluating the impact of agricultural practices on contaminant leaching, and could be of great interest in Andean countries where volcanic soils are common and agricultural practices are being intensified. Acknowledgements The authors thank the personnel of ‘‘Laboratorio de Fertilidad de Suelo’’ of the ‘‘Colegio de Postgraduados, Montecillo’’ for the help with soil analyses. The research was funded by the ‘‘Institut de Recherche pour le De´veloppement’’ (IRD), France and The University of Auckland, New Zealand (SRF 3608705). References Andreini, M.S., Steenhuis, T.S., 1990. Preferential paths of flow under conventional and conservation tillage. Geoderma 46, 85–102. Bartoli, F., Regalado, C.M., Basile, A., Buurman, P., Coppola, A., 2007. Physical properties in European volcanic soils: a synthesis and recent developments. ´ ., Bartoli, F., Buurman, P., O ´ scarsson, H., Stoops, G., Garcı´a-Rodeja, In: Arnalds, O E. (Eds.), Soils of Volcanic Regions in Europe. Springer-Verlag, Berlin, Heidelberg, pp. 515–537. Barton, G., Delmas, P., 2002. A semi-automated colour predicate for robust skin detection. In: Proceedings of IVCNZ 2002, Palmerston North, New Zealand, pp. 121–125. Bejat, L., Perfect, E., Quisenberry, V.L., Coyne, M.S., Haszler, G.R., 2000. Solute transport as related to soil structure in unsaturated intact soil blocks. Soil Sci. Soc. Am. J. 64, 818–826. Cabrera Carvajal, F., Okeschko, K., 1995. Efecto de la labranza sobre la estructura interna de dos tipos de suelo. Agric. Tec. Mex. 21, 139–158. Cervantes, G.G., Sa´nchez-Cohen, I., Rossignol, J.P., 2003. Soil water dynamics studies using image analysis. In: Proceedings of the First Interagency Conference on Research in the Watersheds (ICRW), Benson, Arizona, USA, 27–30 October 2003, pp. 263–271. Coster, M., Chermant, J.L., 1985. Pre´cis d’analyse d’images. CNRS, Paris, France. Duwig, C., Mu¨ller, K., Vogeler, I., 2006. 2,4-D movement in allophanic soils from two contrasted climatic regions. Commun. Soil Sci. Plant Anal. 37, 2841–2855. Duwig, C., Delmas, P., Mu¨ller, K., Prado, B., Morin, H., Ren, K., 2008. Quantifying fluorescent tracer distribution in allophanic soils to image solute transport. Eur. J. Soil Sci. 59, 94–102. Edwards, W.M., Shipitalo, M.J., Dick, W.A., 1992. Rainfall intensity affects transport of water and chemicals through macropores in no-till soil. Soil Sci. Soc. Am. J. 56, 52–58. Epstein, S., Mayeda, T.K., 1953. Variations of the 18O/16O ratio in natural waters. Geochem. Cosmochim. Acta 4, 213–224. Forrer, I., Papritz, A., Kasteel, R., Flu¨hler, H., Luca, F., 2000. Quantifying dye tracers in soil profiles by image processing. Eur. J. Soil Sci. 51, 313–322. Gutie´rrez, C., 2004. Micromorfologı´a de suelos. Colegio de Postgraduados, Montecillo (Mexico), Soil Science lecture notes (unpubl.) 40 p. Hallaire, V., Curmi, P., Widiatmaka, 1997. Morphologie de la porosite´ et circulations pre´fe´rentielles en sature´. Cas des horizons d’un syste`me pe´dologique armoricain. E´tude et Gestion des Sols 4, 115–126. Hallaire, V., Hachicha, M., Cheverry, C., 1998. E´volution structurale d’un horizon de surface argileux sous irrigation. Caracte´risation de la macroporosite´ par analyse d’images. E´tude et Gestion des Sols 5, 107–116. Hallaire, V., Cointepas, J.P., 1993. Caracte´risation de la macroporosite´ d’un sol de verger par analyse d’image. Agronomie 13, 155–164. Haralick, R., Shapiro, L., 1985. A survey: image segmentation techniques. J. Comput. Vis. Graph. Image Process. 29, 100–132. Hatano, R., Kawamura, N., Ikeda, J., Sakuma, T., 1992. Evaluation of the effect of morphological features of flow paths on solute transport by using fractal dimensions of methylene blue staining pattern. Geoderma 53, 31–44. Holtham, D.A.L., Matthews, P.G., Scholefield, D.S., 2007. Measurement and simulation of void structure and hydraulic changes caused by root-induced soil structuring under white clover compared to ryegrass. Geoderma 142, 142–151. Hubert, F., Vincent, H., Sardini, P., Caner, L., Heddadj, D., 2007. Pore morphology changes under tillage and no-tillage practices. Geoderma 142, 226–236. Ismail, S.N.A., 1975. Micromorphometric Soil Porosity Characterization by Means of Electro-optical Image Analysis (Quantimet 720). Soil Survey Institute, Wageningen.

1385

Kamra, S.K., Lennartz, B., van Genuchten, M.Th., Widmoser, P., 2001. Evaluating non-equilibrium solute transport in small soil columns. J. Contam. Hydrol. 48, 189–212. Kutı´lek, M., Nielsen, D.R., 1994. Soil Hydrology. GeoEcology. Catena Verlag, Cremlingen-Destedt, Germany. Lee, J., Horton, R., Noborio, K., Jaynes, D.B., 2001. Characterization of preferential flow in undisturbed, structured soil columns using a vertical TDR probe. J. Contam. Hydrol. 51, 131–144. Magesan, G.N., Vogeler, I., Scotter, D.R., Clothier, B.E., Tillman, R.W., 1995. Solute movement through two unsaturated soils. Aust. J. Soil Res. 33, 585–596. Mori, Y., Iwama, K., Maruyama, T., Mitsuno, T., 1999. Discriminating the influence of soil texture and management-induced changes in macropore flow using soft Xrays. Soil Sci. 164, 467–482. Mu¨ller, K., Duwig, C., 2007. The transport and sorption of 2,4-D in allophanic soils. Soil Sci. 72, 333–348. Murphy, C.P., Bullock, P., Turner, R.H., 1977. The measurement and characterization of voids in soil thin sections by image analysis. I. Principles and techniques. J. Soil Sci. 28, 498–508. Nanzyo, M., Shoji, S., Dahlgren, R., 1993. Physical characteristic of volcanic ash soils. In: Shoji, S., Nanzyo, M., Dahlgren, R. (Eds.), Volcanic Ash Soils, Genesis, Properties and Utilization. Elsevier, The Netherlands, pp. 189–207. Oleschko, K., Chapa Guerrero, J.R., 1989. Structural dynamics of a mollic Andosol of Mexico under tillage. Soil Till. Res. 15, 25–40. Pallud, C., 2000. Etude multi-e´chelles du fonctionnement hydrodynamique et microbiologique d un sol soumis a un apport de solute´s: application a` l’ammonium et au 2,4-D. Ph.D. diss. Joseph Fourier Univ., Grenoble, 259 p. Parfitt, R.L., 1990. Allophane in New Zealand—a review. Aust. J. Soil Res. 28, 343– 360. Perret, J., Prasher, S.O., Kantzas, A., Hamilton, K., Langford, C., 2000. Preferential solute flow in intact soil columns measured by SPECT scanning. Soil Sci. Soc. Am. J. 64, 469–477. Persson, M., 2005. Accurate dye tracer concentration estimations using image analysis. Soil Sci. Soc. Am. J. 69, 967–975. Poletika, N.N., Jury, W.A., 1994. Effects of soil management on water flow distribution and solute dispersion. Soil Sci. Soc. Am. J. 58, 999–1006. Prado, B., 2006. Etude du mouvement de l’eau et du transfert re´actif du nitrate dans les sols volcaniques du bassin versant e´le´mentaire de la Loma, Mexique. Doctoral dissertation, University Joseph Fourier, Grenoble, France. Prado, B., Duwig, C., Escudey, M., Esteves, M., 2006. Nitrate sorption in a Mexican allophanic Andisol using intact and packed columns. Commun. Soil Sci. Plant Anal. 37, 2911–2925. Prado, B., Duwig, C., Hidalgo, C., Go´mez, D., Prat, C., Etchevers, J.D., Esteves, M., 2007. Characterization, classification and functioning of two profiles under different land uses in a volcanic soil sequence in Central Mexico. Geoderma 139, 300– 313. Protz, R., Shipitalo, M.J., Mermut, A.R., Fox, C.A., 1987. Image analysis of soils— present and future. Geoderma 40, 115–125. Quinsenberry, V.L., Philliphs, R.E., Zeleznik, J.M., 1994. Spatial distribution of water and chloride macropore flow in a well structured soil. Soil Sci. Soc. Am. J. 58, 1294–1300. Rao, P.S.C., Rolston, D.E., Jessup, R.E., Davidson, J.M., 1980. Solute transport in aggregated porous media: theoretical and experimental evaluation. Soil Sci. Soc. Am. J. 44, 1139–1146. Ringrose-Voase, A.J., 1996. Measurement of soil macropore geometry by image analysis of sections through impregnated soil. Plant Soil 183, 27–47. Seyfriend, M.S., Rao, P.S.C., 1987. Solute transport in undisturbed columns of an aggregated tropical soil: preferential flow effects. Soil Sci. Soc. Am. J. 51, 1434– 1444. Shipitalo, M.J., Edwards, W.M., Dick, W.A., Owens, L.B., 1990. Initial storms effects of macropore transport of surface applied chemicals in no-till soil. Soil Sci. Soc. Am. J. 54, 1530–1536. Shoji, S., Nanzyo, M., Dahlgren, R.A., 1993. Volcanic Ash Soils. Genesis, Properties and Utilization. Elsevier Science Publishers, The Netherlands. Serra, J., 1982. Image Analysis and Mathematical Morphology. Academic Press, London. Strock, J.S., Cassel, D.K., Gumpertz, M.L., 2001. Spatial variability of water and bromide transport through variably saturated soil blocks. Soil Sci. Soc. Am. J. 65, 1607–1617. Sugita, F., Gillham, R.W., Mase, C., 1995. Pore scale variation in retardation factor as a cause of nonideal reactive breakthrough curves. 2. Pore network analysis. Water Resour. Res. 31, 113–119. Sugita, F., Gillham, R.W., 1995. Pore scale variation in retardation factor as a cause of nonideal reactive breakthrough curves. 1. Conceptual model and its evaluation. Water Resour. Res. 31, 103–112. Tarquis, A.M., McInnes, K.J., Key, J.R., Saa, A., Garcı´a, M.R., Dı´az, M.C., 2006. Multiscaling analysis in a structured clay soil using 2D images. J. Hydrol. 322, 236– 246. Toride, N., Leij, F.J., van Genuchten, M.Th., 1999. The CXTFIT code for estimating transport parameters from laboratory or field tracer experiments. Version 2.1. U.S. Salinity Laboratory Research Report 137, Riverside, California. Tortajada, C., Castela´n, E., 2003. Water management for a megacity: Mexico City metropolitan area. Ambio 32, 124–129. Vanderborght, J., Timmerman, A., Feyen, J., 2000. Solute transport for steady-state and transient flow in soils with and without macropores. Soil Sci. Soc. Am. J. 64, 1305–1317.

1386

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386

Vanderborght, J., Ga¨hwiller, P., Wydler, H., Schultze, U., Flu¨hler, H., 2002. Imaging fluorescent dye concentrations on soil surfaces: uncertainty of concentration estimates. Soil Sci. Soc. Am. J. 66, 760–776. Walker, P.J.C., Trudgill, T.S., 1983. Quantimet image analysis of soil pore geometry: comparison with travel break breakthrough curves. Earth Surf. Process. Landforms 8, 465–472.

WRB (World Reference Base for Soil Resources), 2001. Lecture Notes on the Major Soils of the World. In: Driessen, P., Deckers, J., Spaargaren, O., Nachtergaele, F. (Eds.), FAO World Soil Resources Reports 94, Rome. Yule, D.F., Gardner, W.R., 1978. Longitudinal and transverse dispersion coefficients in unsaturated plain field sand. Water Resour. Res. 14, 582–588.