HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (10) Citing Articles via Google Scholar Google Scholar Articles by Vanderborght, J. Articles by Flühler, H. Search for Related Content PubMed Articles by Vanderborght, J. Articles by Flühler, H. GeoRef GeoRef Citation Agricola Articles by Vanderborght, J. Articles by Flühler, H. Related Collections Soil Methods/Instrumentation Vadose Zone Processes and Chemical Transport

DIVISION S-1—SOIL PHYSICS

Imaging Fluorescent Dye Concentrations on Soil Surfaces

Uncertainty of Concentration Estimates

Jan Vanderborght^*,a, Paul Gähwiller^,b, Hannes Wydler^b, Ute Schultze^b and Hannes Flühler^b

^a Laboratory of Soil and Water, Katholieke Universiteit Leuven, Vital Decosterstraat 102, B-3000 Leuven, Belgium
^b Soil Physics, Inst. of Terrestrial Ecology, Swiss Federal Institute of Technology, ETHZ, Grabenstrasse 11a, CH-8952 Schlieren, Switzerland

* Corresponding author currently at, Inst. for Chemistry and Dynamics of the Geosphere, ICG-IV Agrospere, Research Center Jülich, D-52425 Jülich, Germany; (j.vanderborght{at}fz-juelich.de)

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 
To investigate transport processes in soils, detailed information about the spatial distribution of solutes is required. We describe a method to obtain concentration maps of fluorescent tracers on cross sections of soil cores with a high spatial resolution. The fluorescence signal of two dye tracers, brilliant sulfaflavine (BF; 1H-Benz(de)isoquinoline-5-sulfonic acid, 2,3-dihydro-6-amino-1,3-dioxo-2-(p-tolyl)-, monosodium salt) and sulforhodamine B (SB; xanthylium, 3,6-bis(diethylamino)-9-(2,4-disulfophenyl)-, inner salt, sodium salt), was imaged on the exposed cross sections. The fluorescence signal was corrected for variable illumination light intensity and optical properties of the soil across the exposed surface. Correction factors for varying optical soil properties were derived from the image of the reflected excitation light at the exposed surface. Linear calibration relations related the corrected fluorescence image to the total tracer concentration (C_t) map, that is, mass of dye dissolved in the soil solution and sorbed to the soil particles per unit volume bulk soil. Corrections for varying optical properties of the soil surface were important to reduce the uncertainty of the concentration that was estimated from the fluorescence signal. For BF, the calibration relations were different for different soil materials and a soil specific calibration had to be used. Variations in background fluorescence were an important source of uncertainty of the BF concentration estimates but can be overcome by applying higher concentrations. For SB, variations in calibration relations and in the background fluorescence were considerably smaller, and so is the uncertainty of the estimated SB concentrations.

Abbreviations: BF, brilliant sulfaflavine • CCD, charge-coupled device • SB, sulforhodamine B • TDR, time domain reflectometry

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 
LEACHING OF CHEMICALS in natural undisturbed soils remains an important research topic in soil science. This research is mainly fostered by concerns about the environmental impact of chemicals applied, stored, dumped, or accidentally spilled at or near the soil surface. To investigate leaching in soils, the movement of surface applied tracers was monitored. A general conclusion that can be drawn from leaching experiments is that the transport process in soils is very heterogeneous. However, what can be learned about the leaching process from a leaching experiment depends largely on the monitoring procedure. Commonly used procedures to monitor the leaching process are in situ extraction of soil solution using suction samplers or extraction of soil samples that were taken from the field. An alternative for soil solution samplers is the time domain reflectometry (TDR) technique that infers the concentration from the in situ measured bulk soil electrical conductivity (Vanclooster et al., 1993; Ward et al., 1994). A problem with monitoring solute concentrations in soils is that concentrations may vary across small horizontal distances. In addition, leaching may occur rapidly in only a small fraction of the total pore space, referred to as preferential flow. The probability of intercepting an active flow path with a given number of samplers is often insufficient (Roth et al., 1991). Even when bulk or volume averaged properties are measured in a relatively large sampling volume, for example, of TDR probes, active flow paths that comprise only a small fraction of the total bulk soil volume may go undetected through sampling volume (Vanderborght et al., 2000).

A characterization of the structure of these preferential flow regions and the interaction between the preferential flow and bypassed regions are essential to understand and predict solute leaching (Flühler et al., 1996). However, the spatial resolution that is obtained from soil sampling may be too coarse to identify preferential flow paths accurately (Ritsema and Dekker, 1996).

Nondestructive techniques such as x-ray computer tomography (Peyton et al., 1994), magnetic resonance imaging (Cislerova et al., 1999), electric resistance tomography (Binley et al., 1996), and single photon emission computed tomography (Perret et al., 2000) are currently being developed to monitor preferential flow and transport in soil cores. Application of these techniques is still confined to relatively small soil cores since a trade-off must be made between spatial resolution and object size.

Several authors used dye tracers to visualize preferential flow paths and identify transport processes (van Ommen et al., 1989; Ghodrati and Jury, 1990; Kung, 1990; Flury et al., 1994; Petersen et al., 1997; Vervoort et al., 1999). The structure of the preferential flow paths was, in most cases, characterized from binarized images of the stained patterns using a set of different parameters, for example, relative surface of stained area, number, size, or fractal dimension of stained objects (Hatano and Booltink, 1992; Baveye et al., 1998). To calibrate and validate transport models, dye concentrations rather than dye indicators are required (Forrer et al., 1999). One technique to determine dye concentration on soil surfaces is to quantify light absorption by the dye. Light absorption is quantified by analyzing color pictures of the soil surface. This technique was applied to determine concentration distributions in homogeneous media, for example, glass beads (Schincariol et al., 1993; Richard and Robin, 2000) and silica sand (Tidwell and Glass, 1994; Aeby et al., 1997). To determine concentration distributions in natural and heterogeneous soils, spatial variations in light absorption by the soil material need to be taken into account (Ewing and Horton, 1999; Forrer et al., 2000; Stadler et al., 2000).

Alternatively, dye concentrations can be derived from the intensity of light that is emitted by fluorescent dyes after excitation. Fluorescence of a dye is observed when photons are emitted from electronically excited states following the absorption of light of shorter wavelengths. Fluorescent dyes have been used to stain preferential flow paths in soils (Omoti and Wild, 1979; Jørgensen et al., 1998) and to monitor solute transport during breakthrough experiments (Trudgill, 1987). Recently, an in situ technique using fiber optic mini probes has been developed to monitor fluorescent dye concentrations with high temporal resolution in soils (Campbell et al., 1999; Garrido et al., 1999; Ghodrati, 1999). Using fluorescent dyes with different excitation spectra (fluorescence intensity as a function of the wavelength of the excitation light) and different emission spectra (intensity of the emitted light as a function of the wavelength) the fluorescence signal of the different dyes can be separated by filtering the excitation and emission light. This implies that different dyes can be determined independently and multitracing experiments can be carried out. Besides marking preferential flow paths, the fluorescence signal on soil surfaces can be quantitatively interpreted. In the field of environmental analysis, laser-induced fluorescence is used to determine contaminant concentrations in situ by measuring the emitted fluorescence on a soil surface (Löhmannsröben and Roch, 2000). The quantitative interpretation of the fluorescence signal from dyes on a soil surface is similar to the interpretation of fluorescence from contaminants (Löhmannsröben and Schober, 1999).

In this paper, we discuss a procedure that was introduced by Aeby et al. (2001) to derive tracer concentrations maps from a fluorescence image of an exposed soil surface. This procedure corrects variations in the fluorescence signal that are not caused by variations in dye concentration. We evaluate the effects or importance of different correction steps and discuss the uncertainty on the estimated concentrations. Finally, we propose some changes to the procedure of Aeby et al. (2001) that reduce the uncertainty of the estimated concentrations. Fluorescent tracers were applied to undisturbed soil cores during a leaching experiment. This leaching experiment was carried out to investigate transport of fluorescent tracers in undisturbed soil. Breakthrough data, fluorescent tracer distributions on horizontal surfaces at several depths in the soil cores, and the relation between soil structure and solute leaching are discussed in an accompanying paper (Vanderborght et al., 2002).

	EXPERIMENTAL SETUP

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 
Tracer Application
During a leaching experiment, dye tracers were applied to undisturbed soil cores (107-mm i.d., 110-mm length) that were taken at three different depths in a loamy Stagni-Humic Cambisol under forest (Unterehrendingen, Switzerland). Physical and chemical properties of this soil and of a loamy Eutric Regosol (Bekkevoort, Belgium) are listed in Table 1. Calibration relationships were also determined for the Bekkevoort soil (see below). The soil cores were taken by pushing stainless steel cylinders gently into the soil. In the lab, the samples were pushed out of the steel cylinders and cast with a resin (Araldite CW 2418-1 and hardener HR 5162, Novartis, Switzerland). The samples were saturated with water, put on fritted glass plates in sample holders and placed in a small Plexiglas container with four horizontally installed nozzles that created a mist of fine water droplets (Kasteel, 1997). By turning the nozzles periodically on and off, a quasi steady state water flux was established at the surface of the soil cores. The flow rate that was generated in the Plexiglas container ranged from 33 to 51 mm d^-1 depending on the location in the container. Once a steady flow rate was achieved in the soil cores, 1 mL of a tracer cocktail [2 g L^-1 BF (Sigma Chemical Co., St. Louis, MO), 0.2 g L^-1 SB (FLUKA Chemie AG, Buchs, Switzerland) and 10 g L^-1 CaCl₂·2H₂O] was applied every 30 min to approximate a step input of tracer solution with concentration C₀. C₀ was calculated from the amount and concentration of the applied tracer cocktail and the flux of solute free water at the soil surface. The salt CaCl₂·2H₂O was added to investigate the effect of sorption to soil particles on dye leaching by comparing dye leaching with leaching of Cl^- that is not retarded by sorption (Vanderborght et al., 2002). Depending on the location of the soil core in the container, the following C₀ were applied: 0.2 to 0.32 g L^-1 BF, 0.02 to 0.032 g L^-1 SB, and 1 to 1.6 g L^-1 CaCl₂·2H₂O.

View larger version (24K): [in this window] [in a new window]   Fig. 1. Schematic of the image acquisition setup.

	METHODS

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 
For a dye solution, the fluorescence intensity, I_f, is a function of the intensity of incident light, I₀, and the concentration C:

[1]

where _f is the fluorescence efficiency, is the light absorption efficiency of the dye according to the Beer Lambert law, and d is the thickness of the absorbing layer. For 2.303Cd < 0.05, Eq. [1] can be linearized:

[2]

For fluorescence measurements on soil surfaces, the product Cd can be interpreted as the dye mass that is illuminated per unit area. Due to scattering and absorption of incident light at the soil surface, the excitation light intensity at the soil surface, I₀, is different from the intensity of incident light. Even for a uniform intensity of incident light, the excitation light intensity may vary over the soil surface due to variations in soil properties over the soil surface. The procedure that we use to derive dye concentration maps from fluorescence measurements on soil surfaces consists of three steps. Figure 2 gives an overview of the different steps: correction for variable incident light intensity (Step I), correction for variations in soil surface properties that influence the local excitation light intensity and that are linked to variations in light reflection at the soil surface (Step II), and application of calibration relations (Step III).

View larger version (53K): [in this window] [in a new window]   Fig. 2. Outline of the image processing steps to derive dye concentrations from the fluorescence image. Left-hand side panels are images of the fluorescence signal of brilliant sulfaflavine (BF) on horizontal cross sections of four soil cores from the Unterehrendingen soil (upper and lower cores in the panel are taken from the topsoil and subsoil, respectively): A(x) is the raw fluorescence signal; Aff(x) is the flat field corrected fluorescence signal; Affr(x) is the background subtracted fluorescence signal corrected for variations in soil surface reflection; and Ct(x) is the total concentration pattern. Right-hand side panels represent correction factors <F(x)>/F(x) for inhomogeneous illumination (flat field correction); correction factors R0/Rff(x) for variations in excitation light reflection at the soil surface; and calibration relations between corrected fluorescence and total dye concentration, Ct.

[3]

where A_ff(x) is the flat field corrected fluorescence image and <F(x)> is the mean value of the flat field.

Corrections for Spatial Variations of Soil Surface Properties (Step II)
The fluorescence signal that is recorded by the camera depends, in addition to the concentration of the fluorescent dye at the soil surface, on soil surface properties that influence the local excitation light intensity. Two different causes for spatial variability of the excitation light intensity over the soil surface can be distinguished. First, geometrical irregularities of the soil surface at different scales, that is, surface roughness due to irregular particle shapes, irregular arrangement of particles, and depressions at a larger scale, lead to differences in local excitation light incidence. Rougher soil surfaces contain more shadowed regions where the local excitation light intensity is lower than the intensity of the incident light. Second, variations in the chemical and physical composition of the soil surface (mineralogy, organic carbon content, and water content) lead to differences in local light absorption and, hence, differences in local excitation and emission light intensities. Variations in soil surface roughness are seen as gradations of darkness, whereas variations in chemical and physical composition of the soil surface are seen as gradations of soil color.

Assuming a chemically and physically homogeneous soil surface, the effect of spatial variability of soil surface roughness on the fluorescence signal can be corrected as:

[4]

where A_ffr(x) is the fluorescence image corrected for variations in soil roughness, R₀ is a reference reflection, and R_ff(x) [= R(x) <F(x)>/F(x)] is the flat-field corrected reflection image (the index ⁽¹⁾ discriminates the different correction procedures for variations in optical properties of the soil surface). This correction term follows directly from the linear relations between the local excitation light intensity, the fluorescence signal, and the reflection signal, R_ff(x).

Assuming a constant surface roughness, correction factors for variations in chemical and physical composition of the soil surface can also be derived from the reflection image, R_ff(x). Assuming that most of the excited dye molecules are at or close to the surface and directly exposed to incident light, the excitation light intensity is the sum of direct excitation light incidence and the diffuse reflected excitation light at the soil surface (Aeby et al., 2001). In the same vein, the emitted light intensity is the sum of the directly emitted light and the reflected emitted light at the exposed soil surface. This assumption leads to the following correction for spatial variations in light absorption that result from variations in chemical and physical composition of the soil surface:

[5]

where f_ex0 and f_em0 are the reference diffuse reflectance at the excitation and emission wave lengths, respectively, and f_ex(x) and f_em(x) are the diffuse reflectance of the exposed soil surface at location x, which depend on the chemical and physical composition of the soil surface at this location. Because of the random orientation of soil particles, the light reflection from a macroscopically flat surface is diffuse, that is, the light is reflected in all directions with the same intensity. The diffuse reflectance is the ratio of the intensity of the reflected light from a macroscopically flat surface to the intensity of the incident light. It is measured using diffuse reflectance spectroscopy by comparing the light reflection from a soil sample with the reflection from a white reference standard (BaSO₄), which is placed at exactly the same position with respect to the light source and detector as the soil sample and which reflects >98% of the incident light at every wavelength of the visible spectrum. The diffuse reflectance is closely related to the light reflection that is measured using the image acquisition setup. At a certain point it can be derived from the reflection image as:

[6a]

and

[6b]

where R₀ is the reflection of a reference surface with diffuse reflectance f_ex0 and f_em0.

There are two problems related to the application of this correction. First, the correction factor for variations in surface roughness is different from the correction factor for variations in chemical and physical composition of the surface. But, the reflection image, R_ff(x), reflects variations in surface roughness as well as variations in chemical and physical composition of the surface. Since the roughness and chemical and physical composition vary independently across the soil surface, it is impossible to distinguish them based on the reflection image, R_ff(x). Therefore, Aeby et al. (2001) suggested using an averaged correction term, which they defined as the arithmetic average of Eq. [4] and [5]. Second, for rough soil surfaces, it might be questioned whether most of the dye molecules are directly exposed to the incident light. As an alternative working hypothesis, we postulate that most dye molecules are excited by scattered or diffuse reflected light from randomly oriented microscopic surfaces. The intensity of this local scattered light depends on the local light absorption by the soil minerals and the roughness of the surface. It is proportional to the local intensity of the reflected light that is recorded by the camera in the reflection image. If it is assumed that dye molecules are excited by the scattered light rather than by direct incident light, then the intensity of the scattered excitation light and the fluorescence signal are proportional. In the same vein, the local fluorescence light and the local diffuse reflected excitation light that are recorded by the camera are proportional. As a consequence, variations in both soil surface roughness and chemical and physical composition of the surface could be linked to variations in local light reflection and corrected for by the following correction:

[7]

Notice that this equation is identical to the correction for variations in soil roughness assuming a chemically and physically homogeneous soil surface (Eq. [4]). Löhmannsröben and Schober (1999) used and experimentally verified an identical correction of laser-induced fluorescence of diesel in soils with differing light absorption and reflection. Since the light absorption by a certain compound depends on the wavelength, reflection images, R_ff(x), have to determined at the excitation light frequency.

It is evident that also fluorescent dye molecules absorb excitation light and influence the reflection image. However, correction factors were determined assuming that variations in the reflection image result only from variations in soil surface properties. Therefore, the correction factors can only be used when light absorption by fluorescent dye molecules is negligible compared with the light absorption by the soil surface. In addition to dye molecules, other chemical substances at the soil surface, such as organic matter, may emit a background fluorescence signal. This background fluorescence signal, A_bg, may be negatively correlated, due to low fluorescence efficiency, to the reflection image. For a constant background or a background signal that is negatively correlated to the reflection image, Eq. [7] generates a noise on the corrected image. The background fluorescence is derived from the average fluorescence signal in selected regions on the surfaces in which apparently no dye was present. The standard deviation of the background fluorescence within the nonstained regions, _Abg, defines the minimal dye fluorescence that can be discriminated from the background signal. Twice the standard deviation of the fluorescence signal in nonstained regions can be interpreted as a detection limit of the dye fluorescence. Background subtracted fluorescence signals (A_ff(x) - A_bg) that are larger than 2_Abg or detection limit, come with a probability >95% from dye molecules. To avoid the introduction of noise, an average value of the background signal is subtracted from the original fluorescence signal and all values of the background subtracted fluorescence [A_ff(x) - A_bg] that are smaller than 2_Abg are set equal to zero before the fluorescence image is multiplied by the correction factor:

[8a]

[8b]

It should be noticed that for the calculation of A_ffr(x) or A_ffr(x), a flat field image F(x) is not required since replacing A_ff(x) and R_ff(x) in Eq. [4] and [7] by A(x) and R(x), respectively, would lead to exactly the same results. However, a flat field image is required to determine the background signal A_bg and its standard deviation _Abg and calculate A_ffr(x) and A_ffr(x).

Calibration Procedure (Step III)
The corrected and background subtracted fluorescence signal, A_ffr(x), is related to the concentration pattern using a calibration relation. For relatively low concentrations, a linear relation can be assumed (Eq. [2]). The definition of the concentration units depends on how the concentrations are determined or preset in the calibration experiment. The calibration relation can be determined in different ways. In a first procedure, soil material could be sampled and the tracer extracted to determine the local concentration at number of points on a soil surface of which the fluorescence signal was recorded. Using the extraction procedure, the concentration is most naturally defined in terms of the extracted tracer mass per mass of dry soil material. However, two problems arise with this extraction calibration procedure. First, the fluorescence signal is defined by the number of dye molecules per surface area. Thus, the local soil bulk density must be known to link the fluorescence signal to the mass of dye per mass of dry soil. Second, the fluorescence signal depends on the concentration at the soil surface whereas the concentration in the soil extract represents the average concentration in the sampled soil volume. Since local concentrations vary a lot with a short distance, this may lead to a discrepancy between the concentration in the extract of the sampled soil volume and the concentration at the soil surface. Therefore, we used an alternative procedure and prepared a soil paste mixing preset amounts of dry soil and tracer solution with known concentration. A sample container with known volume was filled with the soil paste and weighted so that the total dye concentration, C_t, that is, dye mass per volume bulk soil, in the soil samples could be calculated. The total concentration C_t comprises the mass of dye dissolved in the soil solution and sorbed to soil particles. It should be noted that a unique calibration relation between fluorescence signal and total dye concentration for soil materials with different dye adsorption isotherms (for the same total dye concentration a different fraction of dye mass is adsorbed to soil particles) implies that fluorescence of dye molecules is not influenced by adsorption to soil particles. A calibration relation was established between C_t in the prepared soil samples and the fluorescence signal from these samples. To prepare the samples, we took nonstained soil material from the 0- to 0.1-m and from the 0.5- to 0.6-m soil layers of the Unterehrendingen soil. In addition, we also used nonstained soil material from the 0- to 0.1-m and the 0.5- to 0.6-m soil layers of the Bekkevoort soil (Eutric Regosol) in order to evaluate the variability of the calibration relationships between different soil materials. The properties of the different soil layers of the Unterehrendingen and Bekkevoort soils are listed in Table 1. The soil material was dried for 24 h at 70°C, ground, and sieved with a 0.5-mm sieve. For the Unterehrendingen soil, we mixed 12 g of the dried and sieved soil with 3.6 mL dye solution. For the Bekkevoort soil, a mixture of 12 g soil and 3.6 mL dye solution was too liquefied and therefore we mixed 12 g soil with only 2.4 mL dye solution for this soil. Dye solutions of 10, 20, 40 (50 for the 0.5- to 0.6-m layer in Unterehrendingen), 100, and 200 mg L^-1 SB and 50, 100, 200, 500, 1000, 2000 (Unterehrendingen), and 5000 (Unterehrendingen, 0.5- to 0.6-m soil layer) mg L^-1 BF were used to create a calibration series of soil pastes. Small Polystyrole-cylinders of 6.49-mL volume were filled with soil paste and sealed to prevent evaporation that would alter the dye distribution within the sample and the optical properties of the soil surface. Water flow within the sample towards the soil surface accumulates dyes at the soil surface and dry surfaces are brighter than wet ones (Gimmi et al., 1999). The resulting volumetric water content, , in the samples with the Unterehrendingen and Bekkevoort soil, was, respectively, 0.44 and 0.34. The total dye concentrations, C_t, in the calibration samples ranged for SB from 3 to 70 mg L^-1 bulk soil (Bekkevoort) and from 4 to 90 mg L^-1 bulk soil (Unterehrendingen), and for BF from 15 to 350 mg L^-1 bulk soil (Bekkevoort) and from 20 to 2000 mg L^-1 bulk soil (Unterehrendingen). Exact C_t were calculated for each calibration sample knowing the weight and volume of the sample, the ratio of dye solution and dry soil of the soil paste, and the dye concentration in the dye solution.

The fluorescence and reflection images of the calibration samples were recorded under the same illumination conditions as the horizontal cross sections of the Unterehrendingen soil cores and corrected for variations in illumination light intensity (Eq. [3]). Averages of the fluorescence and reflection across the surface of a calibration sample were used in the correction equations for variations in optical properties between samples from different soil layers (Eq. [4]–[7]) and in the calibration relations. The reference reflection, R₀, is defined as the average reflection of the samples from the four different soil layers. However, since the spectrum of the excitation light is relatively wide, and since the diffuse reflectance of the soil cannot be measured at the emission wavelength using the image acquisition setup, the diffuse reflectances of a certain soil material at the excitation and emission wave lengths, f_ex and f_em, were measured using diffuse reflectance spectroscopy. The reference diffuse reflectances, f_ex0 and f_em0, were defined as the average reflectances of the different soil materials, consistent with the definition of R₀. Differences in optical properties of the surfaces of calibration samples and soil cores lead to a different light reflection and fluorescence signal for the same total concentration C_t. Equation [5] or Eq. [7] corrects these differences so that, in principle, the factors that determine the optical properties (water content, surface roughness, and soil material) need not to be identical in the calibration samples and soil cores.

	RESULTS AND DISCUSSION

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 
In this section we discuss the importance of different steps in the image analysis procedure (Fig. 2) to derive concentration patterns from a fluorescence image and the uncertainty of the derived concentration pattern. The importance of a certain step is evaluated in terms of the changes that it makes to the original image. Although the derivation of concentrations from the fluorescence signal using calibration relations is the last step in the procedure, we start with discussing this step, since it was decided based on the calibration experiments which correction factor was used to correct for variations in soil surface properties.

Calibration Relations
In Fig. 3 , the fluorescence of the calibration samples is plotted vs. the total dye concentration in the samples, C_t. Calibration relationships are shown for fluorescence signals that are, A_ffr, and that are not, A_ff, corrected for differences in light reflection between soil samples with different soil material. Either Eq. [5] or Eq. [7] were used to correct for differences in light reflection. In order to facilitate the interpretation of the double logarithmic plots, the background fluorescence signal (fluorescence signal for C_t = 0 mg L^-1), which was determined for each soil material and correction procedure, was subtracted. The relative reflection coefficient, R_ff/R₀ and the diffuse reflectance, f_ex and f_em, of the soil material in the calibration samples at the excitation and emission wavelengths are listed in Table 3, together with the correction factors derived from Eq. [5] and [7]. Notice in Table 3 that the correction factors derived from Eq. [5] are close to 1, whereas Eq. [7] leads to larger corrections of the fluorescence signal. Except for BF in the Unterehrendingen soil, the darker color of the top soil layer due to its higher organic carbon content is reflected in larger correction factors. The correction factors are also larger for the Bekkevoort soil that appears to be darker than the Unterehrendingen soil.

View larger version (26K): [in this window] [in a new window]   Fig. 3. Calibration relations for different soil layers and the two dyes, brilliant sulfaflavine and sulforhodamine B: total dye concentration in the calibration sample, Ct, is plotted vs. the noncorrected fluorescence signal Aff, and vs. the fluorescence signal corrected for variations in light reflection at the soil surface by Eq. [5] Affr, and by Eq. [7] Affr.

[9]

where n is the number of calibration points. The obtained slopes for the different soil materials and different fluorescence corrections are listed in Table 4. For BF, the slopes of the calibration relations are quite different for the different soil types and correcting the fluorescence for differences in light reflection resulted in an even larger spreading of the slopes. Therefore, differences in the fluorescence cannot be explained by differences in local excitation light intensities due to differences in optical properties of the soil surfaces. The different calibration relationships between fluorescence signal and total BF concentration for the two layers from the Unterehrendingen soil can neither be explained by assuming that only dissolved BF molecules contribute to the fluorescence signal. For the same total BF concentration, the fluorescence signal is a factor 3.3 higher in the 0- to 0.1-m layer than in the 0.5- to 0.6-m layer, whereas the concentration in the liquid phase is only a factor 1.3 higher. This factor was calculated from the water content, bulk density of the calibration samples, and the linear distribution coefficient, K_d, between liquid and solid phase concentrations (see Fig. 1 in Vanderborght et al., 2002). This suggests that the chemical composition of the soil solution and the sorption mechanisms of BF to soil particles, which may differ for different soil materials, influence the fluorescence of BF on soil surfaces. The large differences in the calibration relations between the different soil materials imply that a soil-specific calibration is required to derive the absolute concentrations from the fluorescence signal. In a given soil material, relative concentration distributions can be inferred directly from the fluorescence signal. However, to interpret the fluorescence signal of BF on a vertical soil surface with different soil layers and corresponding different calibration relations, a soil type or depth dependent calibration relation must be used.

The differences in light reflection between the calibration samples were mainly due to differences in chemical and physical composition of the soil surfaces since nearly flat surfaces were obtained by pushing the soil paste to the Polystyrene container surface. However, unlike Eq. [5], Eq. [7] also corrects for variations in surface roughness. Therefore, we use Eq. [7] from this point forward to correct A_ff(x) for variations in light absorption and roughness of the soil surface.

Correction for Variations in Soil Surface Properties
In Fig. 4 , a noncorrected fluorescence image, A_ff(x), and three differently corrected images for variations in light reflection at the soil surface, A_ffr(x), A_ffr(x), and _ffr(x), are shown. _ffr(x) is calculated using Eq. [8] but on the basis of the median filtered (9- by 9-pixel median filter) fluorescence and reflection images. Not subtracting the background before the correction led to a considerable introduction of noise A_ffr(x), Fig. 4b] that nearly completely blurred the signal in the original fluorescence image. In Fig. 5 , the background fluorescence, detection limit (which was defined as twice the standard deviation of the background fluorescence in nonstained regions), and the typical signal of the dye fluorescence are graphically presented. The typical signal merely represents a typical background subtracted fluorescence signal in stained regions on the soil surfaces. The total dye concentrations, C_t, that correspond with this background, detection, and typical dye signal and that are derived using the calibration relations are listed in Table 5. For BF, the background is as large as the dye signal and the detection limit is 20 and 30% of the typical dye signal for the 0- to 0.1-m and 0.5- to 0.6-m layer, respectively. To increase the signal to detection limit ratio, higher BF concentrations should be applied. For SB, the background is smaller compared with the typical dye signal and the detection limit is 5% of the typical dye signal. It should also be noted that the variability of the background fluorescence in nonstained regions of the soil core surfaces and the calibration plots (Fig. 3) point at similar absolute values of the detection limit.

View larger version (61K): [in this window] [in a new window]   Fig. 4. Corrections of the brilliant sulfaflavine fluorescence signal for variations in light reflection at the soil surface (soil core from the 0.5- to 0.6-m layer of the Unterehrendingen soil): a) noncorrected signal, Aff(x); b) corrected signal without subtracting the background fluorescence before the correction, Affr(x); c) corrected signal with background fluorescence subtracted before correction, Affr(x); d) corrected signal with background fluorescence subtracted and fluorescence and reflection images smoothed by a 9 by 9 median filter before correction, ffr(x). Right-hand side panels are pixel values along a transect (marked in Panel b).

View larger version (17K): [in this window] [in a new window]   Fig. 5. Illustration of the background fluorescence signal, detection limit, and typical dye fluorescence signal.

In Fig. 6 , the range of correction factors that is used to correct the fluorescence image for variations in soil surface roughness and light absorption are shown for the different soil layers and the two dyes. The range is represented by the minimum and maximum correction factors and an interval around the mean correction factor. This interval is defined as:

[10]

where is the standard deviation of the log_e transformed correction factors, and R₀/R_mean is the average correction factor for a soil surface. For lognormally distributed variables, this interval represents the range in which 90% of the correction factors are found. In general, this interval ranges from 0.65 to 1.5 R₀/R_mean.

View larger version (12K): [in this window] [in a new window]   Fig. 6. Range of correction factors for variations in excitation light reflection at the surface of soil cores from the 0- to 0.1-m, 0.2- to 0.3-m, and the 0.5- to 0.6-m layer of the Unterehrendingen soil for the excitation light wavelengths of brilliant sulfaflavine (BF) and sulforhodamine B (SB). Horizontal bars mark the interval that is defined in Eq. [10] and that contains 90% of the correction factors.

	SUMMARY AND CONCLUSIONS

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

View larger version (15K): [in this window] [in a new window]   Fig. 7. Summary of uncertainty factors of the estimated brilliant sulfaflavine (BF) and sulforhodamine B (SB) concentrations and of correction factors normalized to one with which the flat field corrected fluorescence signal, Aff(x), is multiplied to obtain the concentration estimates on soil surfaces in cores from the Unterehrendingen soil.

	IN MEMORIAM Paul Gähwiller

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 
The experimental work of this contribution was a first step in the dissertation of Paul Gähwiller, who died in the mountains on 7 Oct. 1999. This is to commemorate his spirit of a profound scientific curiosity and his virtually unlimited willingness to help whenever help was needed.

	ACKNOWLEDGMENTS

 
We wish to thank Dr. L. Schober (Institut für Physikalische und Theoretische Chemie, University Erlangen, Germany) for measuring the diffuse soil reflectances. The assistance of Dr. S. Bujukova, H.-P Läser, J. Leuenberg and H. Wydler with the experimental setup, the measurements of the breakthroughs, and the analysis of the fluorescence images was essential and very much appreciated. We are also indebted to reviewers for their high quality reviews.

The corresponding author is grateful to the Belgian Fund for Scientific Research, Flanders, which funded his research stay at the Soil Physics Laboratory of the Institute of Terrestrial Ecology (ETHZ). While carrying out this research, the corresponding author was a post-doctoral research assistant of the Belgian Fund for Scientific Research, Flanders.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 
(deceased).

Received for publication November 10, 2000.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION EXPERIMENTAL SETUP METHODS RESULTS AND DISCUSSION SUMMARY AND CONCLUSIONS IN MEMORIAM Paul Gahwiller REFERENCES

 

• Aeby, P. 1998. Quantitative fluorescence imaging of tracer distributions in soil profiles. Ph.D. Thesis no. 12951. Eidgenössische Technische Hochschule, Zurich, Switzerland.
• Aeby, P., J. Forrer, C. Steinmeier, and H. Flühler. 1997. Image analysis for determination of dye concentrations in sand columns. Soil Sci. Soc. Am. J. 61:33–35.[Abstract/Free Full Text]
• Aeby, P., U. Schultze, D. Braichotte, M. Bundt, F. Moser-Boroumand, H. Wydler, and H. Flühler. 2001. Fluorescence imaging of tracer distributions in soil profiles. Environ. Sci. Technol. 35:753–760.[Medline]
• Baveye, P., C.W. Boast, S. Ogawa, J.Y. Parlange, and T.S. Steenhuis. 1998. Influence of image resolution and thresholding on the apparent mass fractal characteristics of preferential flow patterns in field soils. Water Resour. Res. 34:2783–2796.
• Bedrick, A.E., J.L. Painter, and D. Lutringer Bonitz. 1979. Spectral properties and relative quantum yields of new fluorescent stains. Immunopharmacology 2:33–37.[Medline]
• Binley, A., S. Henry-Poulter, and B. Shaw. 1996. Examination of solute transport in an undisturbed soil column using electrical resistance tomography. Water Resour. Res. 32:763–769.
• Campbell, C.G., M. Ghodrati, and F. Garrido. 1999. Comparison of time domain reflectometry, fiber optic mini probes, and solution samplers for real time measurement of solute transport in soil. Soil Sci. 164:156–170.
• Cislerova, M., J. Votrubova, T. Vogel, M.H.G. Amin, and L.D. Hall. 1999. Magnetic resonance imaging and preferential flow in soils. p. 397–411. In M.T. van Genuchten, F.J. Leij, and L. Wu (ed.) Proc. Int. Workshop, Characterization and Measurement of the Hydraulic Proporties of Unsaturated Porous Media, Riverside, CA. 22–24 Oct. 1997. University of California, Riverside, CA.
• Ewing R.P., and R. Horton. 1999. Discriminating dyes in soil with color image analysis. Soil Sci. Soc. Am. J. 63:18–64.[Abstract/Free Full Text]
• Flühler, H., W. Durner, M. Flury. 1996. Lateral solute mixing processes—A key for understanding field-scale transport of water and solutes. Geoderma 70:165–183.
• Flury, M., H. Flühler, W.A. Jury, and J. Leuenberger. 1994. Susceptibility of soils to preferential flow of water. Water Resour. Res. 30:1945–1954.
• Forrer, J., R. Kasteel, M. Flury, and H. Flühler. 1999. Longitudinal and lateral dispersion in an unsaturated field soil. Water Resour. Res. 35:3049–3060.
• Forrer, J., A. Papritz, R. Kasteel, H. Flühler, and D. Luca. 2000. Quantifying dye tracers in soil profiles by image processing. Eur. J. Soil Sci. 51:313–322.
• Garrido, F., M. Ghodrati, M. Chendorain, and C.G. Campbell. 1999. Small scale variability in solute transport processes in a homogeneous clay loam soil. Soil Sci. Soc. Am. J. 63:1513–1522.[Abstract/Free Full Text]
• Ghodrati, M. 1999. Point measurement of solute transport processes in soil using fiber optic sensors. Soil Sci. Soc. Am. J. 63:471–479.[Abstract/Free Full Text]
• Ghodrati, M., and W.A. Jury. 1990. A field study using dyes to characterize preferential flow of water. Soil Sci. Soc. Am. J. 54:1558–1563.[Abstract/Free Full Text]
• Gimmi, T., N. Ursino, and H. Flühler. 1999. Estimation of water saturation in a heterogeneous sand tank. p. 164–173. In J. Feyen and K. Wiyo (ed.) Proc. of Int. Workshop of EurAgEng's Field of Interest on Soil and Water, Modelling of Transport Processes in Soils at Various Scales in Time and Space, Leuven, Belgium. 24–26 Nov. 1999. Wageningen Pers, Wageningen, The Netherlands.
• Hatano, R., and H.W.G. Booltink. 1992. Using fractal dimensions of stained flow patterns in a clay soil to predict bypass flow. J. Hydrol. 135:121–131.
• Haugland, R.P. 1996. Handbook of fluorescent probes and research chemicals. 6th ed. Molecular Probes, Eugene, OR.
• Jørgensen, P.R., L.D. McKay, and N.H. Spliid. 1998. Evaluation of chloride and pesticide transport in a fractured clayey till using large undisturbed columns and numerical modeling. Water Resour. Res. 34:539–553.
• Kasteel, R. 1997. Solute transport in an unsaturated field soil: Describing heterogeneous flow fields using spatial distribution of hydraulic properties. Ph.D. Thesis no. 12477. Eidgenössische Technische Hochschule, Zurich, Switzerland.
• Kung, K-J.S. 1990. Preferential flow in a sandy vadose zone: 1. Field observation. Geoderma 46:51–58.[ISI]
• Löhmannsröben, H.-G., and T. Roch. 2000. In