Tải bản đầy đủ - 0 (trang)
V. Application of Computer-Assisted Tomography to Soil–Water Studies

V. Application of Computer-Assisted Tomography to Soil–Water Studies

Tải bản đầy đủ - 0trang



relationships between attenuation and bulk density/water content at the

higher values associated with soil and similar porous materials. As might

be expected, bearing in mind the dependence of the mass attenuation

coefficienton the energy of the radiation, the chemical composition of the

soil matrix, and the packing or bulk density, as well as differences in effects

such as beam hardening, source detector geometry, and degree of electronic discrimination, substantial variations in the slope of the linear

regressions have been noted with different soils and scanners. Furthermore, plots of these relationships do not always extrapolate to the origin as

required by Eqs. ( 1 1) and (12).

The mass attenuation coefficient of a chemical compound, or a mixture,

is more sensitive to variations in the chemical composition the lower the

photon energy and the heavier the elements that are subject to abundance

variations. In examining the influence of chemical composition on photon

attenuation by soils, Coppola and Reiniger ( 1974) had earlier demonstrated that variation of the mass attenuation coefficient with soil composition becomes significant at energies below about 200 keV and is at a

maximum below 50 keV (Fig. 12). Above 300 keV any difference in the

value of the attenuation coefficient was negligible. Thus in the case of an

"'Am source (59.6 keV), the pray attenuation by soil was influenced

rather strongly by the particular soil composition (differences as large as

200% were calculated). In contrast, the influence was essentially negligible

at the energy of 13'Cs (662 keV). They also concluded that although

variations in the abundance of heavier elements such as Fe can cause

significant variations in p, rather large changes in the abundance of other

elements, e.g., of Si, have no comparable effect on the p coefficient.

Because most commercially available medical X-ray CT scanners operate

at around 120 keV excitation, significant differences in attenuation resulting from variations in chemical composition could be expected to occur.

The first reported attempt to use commercial medical X-ray CAT scanners to measure spatial changes in soil bulk density was by Petrovic et al.

(1982), who showed that attenuation for an American Science and Engineering CT scanner was linearly related to the bulk density of soil from the

surface horizon of Metea fine sandy loam over the range of 1.2 to 1.6

Mg/m3. The sensitivity to density changes for the Metea soil was about

0.02 Mg/m3, with a maximum observed deviation from predicted of 0.07

Mg/m3. Only a slightly different regression slope was observed for a mixture of glass beads and spheres and this was attributed to the differences in

atomic number of the absorbing materials. Hainsworth and Aylmore

(1983), using an EM1 1007 X-ray CT scanner, demonstrated that spatial

changes in soil water content with time of the order of 0.006 g/cm3 could

be readily resolved by the CAT technique in nonswelling soils (Fig. 13) and



- .-[



Average earth crust

Calcareous clay

Norfolk sandy loam

Cecil sandy loam

Nipe clay






1 .o


y Energy (MeV)

Figure 12. Mass attenuation coefficients for soils of various chemical compositions

versus y-ray energy. [After Coppola and Reiniger, 1974.0 by Williams & Wilkins (1974).]

illustrated the scanner's potential for studies of soil/plant /water relations.

Crestana et al. (1985) also obtained essentially linear calibration curves

between the output from a General Electric T 8800 scanner and soil bulk

density and soil water content for a sandy soil from the Ap horizon from

Trieste, Italy, and a fine sandy loam from Barretos, Brazil. However, the

linear calibration curves for bulk density for the two soils diverged markedly. Changing the bulk density caused a roughly parallel displacement of

the calibration curves for attenuation as a function of water content.

Crestana et al. (1985) used their CT scanner to observe the change in the

X-ray attenuation coefficient with time at a single point in space within a

homogeneous soil core as a wetting front passed. They showed that CT

scanners could be used to measure the movement of water in soils at rates

of 1.6 mm/sec.

Similar more detailed linear relationships between attenuation coeffi-




















Distance (mm)


Figure 13. Half-slice three-dimensional plots showing changes in soil water content with

time due to the infiltration of water into a soil column from an artificial root (alundumtube).

Representations at (a) 1 min, (b) 10 min, (c) 15 min, and (d) 20 min of inEiltration are

illustrated. (After Hainsworth and Aylmore, 1983.)

cients (using a Philips Tomoscan 3 10) and volume fractions of soil solids

were obtained by Anderson et al. (1988) for two silt loams from Missouri.

Over 99% of the variation in CT attenuation coefficients for 40 dry soil

cores for each of the two soils could be accounted for by linear regression

relationships with the volume fraction of soil solids (fs). Approximately

98% of the variation in CT attenuation coefficients for the 40 wet cores for

each of the two soils was accounted for by regression relationships with the

volume fraction of soil water (fw), after correcting for swelling effects and

differences in bulk density. Differences in attenuation coefficients for the

two soils were shown to be largely due to differences in Fe content.

Parameter values for the Mexico silt loam compared favorably with those

determined for Metea fine sandy loam used by Petrovic et al. (1982) but

differed markedly from values given in the earlier work by Crestana et al.

(1985), where larger sampling volumes had been used. They suggested the

possibility of developing a universal relationship between X-ray CT data

versus bulk density and water content if differences in the electron densities of the soils are known and the effect of soil core size on computed

tomography results could be characterized.



Tollner and Murphy ( 1991) also reported linear relationships among the

absorption coefficientsfor solids and liquid portions in five soils ranging in

texture from sand to loam and concluded that, for many applications and

providing that zero swelling and shrinkage could be assumed, one calibration relationship could be applicable to a wide range of soils. For all soils

tested, except for the Wilcox clay, which contained 66% clay of 2 : 1 clay

mineralogy and exhibited significant shrinkage and swelling, the predicted

soil density term was constant within 5% when the water term was fixed.

The assumption of zero shrinkage during drying was largely met for all

soils except the Wilcox clay.

Equally good linearity of response between the y attenuation coefficient

and the average bulk density/water content was observed by Phogat and

Aylmore (1989) using a 137Cssource (Fig. 14), but again significant variation in slope between materials is evident. Furthermore, although the

relationshipsbetween y attenuation and water content for both a sand from

Bassendean, Western Australia, and a kaolinite-dominated sandy loam

from Kulin, Western Australia, were also essentially linear, a degree of

swelling resulted in the slope of the linear regression for the Kulin soil

being substantially less than that for Cs radiation as measured for the

nonswelling sand (Fig. 14B). Phogat and Aylmore (1989) suggested that

variations in the mean and standard deviations of pixel attenuation coefficients for a scanned layer could be used to assess the structural status of the

soil, because these changes reflect changes in the uniformity of the layer

arising from changes in the spatial distribution of pore volume and soil

matrix (soil aggregate). However, although some correction for bulk density changes on swelling can be made on the basis of appropriate calibrations, the difficulty in accounting for swelling and bulk density changes, in

general, particularly when using polychromatic X-rays, remains a major

impediment in studies of soil/plant/water relations. Taking into account

the number of variables that may influence the regression for bulk density,

it seems unlikely that a truly universally applicable relationship for attenuation at the energy levels used in medical CT scanners can be derived, and

independent calibration for specific soils will remain an essential prerequisite to their use. More generally applicable relationships may, however, be

possible using higher energy monochromatic prays (e.g., from lS7Cs).

Differences in soil water salinity or chemical composition of practical

interest in soil science have no appreciable effect on the y-ray attenuation

(Coppola and Reiniger, 1974). In most of the previous studies using Xrays, values close to the standard measured absorption value for water of

0.0191 mm-l at energies around 120 keV (McCullough, 1975) have been

observed, implying that one need not determine this coefficient for each

soil. However, this value will of course change with the energy of the






0 Glasslubes


0 Kulin SOH



Bulk Density (g/cm3)





Kulin Soil




0.16 -

Y=O 1167+0 0592X










Volumetric Moisture Content (cm 3/cm 3)

Figure 14. Linear relations between attenuation coe5cient and (A) bulk density for

Kulin soil and glass tubes and (B) volumetric moisture content for Kulin soil and pure sand.

Vertical lines represent standard deviations. (After Phogat and Aylmore, 1989.)

radiation, and for prays from "Am, la9Yb,and 13'Cs the coefficients are

around 0.201, 0.176, and 0.083 cm-*, respectively.



Of considerableinterest is the ability of CAT scanners to characterizethe

internal structure and the nature of components present in the soil. A

knowledge of the size and distribution of pores is, for example, relevant to



an understanding of many important processes that take place in soils,

including water entry and redistribution, aeration, and root penetration. In

particular, the ability to monitor root proliferation and distribution by a

noninvasive technique is an essential prerequisite to detailed studies of

water uptake by plant roots.

The ability of the Scanner to resolve voids and objects of different sizes

and spacing is obviously a function of the pixel dimensions, which for most

commercial scanners is of the order of 1 X 1 mm. Spatial resolution of

objects is based on the difference in attenuation of the transmitted X-rays

by the object and the adjoining soil mass. Good spatial resolution can best

be achieved when there is a large difference in H values between the subject

(e.g,, a void) and the background (e.g., the soil mass). Most workers (e.g.,

Petrovic et al., 1982; Anderson et al., 1990) have found that under such

conditions CT scanners are able to detect holes of the order of 1 mm in

diameter but are not necessarily able to separate adjoining holes of similar

separation. Although the spatial resolution as determined by the scanner

and image reconstruction algorithm used by Warner et al. (1989) was

approximately 0.5 mm, the practical resolution was about twice this value

due to boundary effects. If the boundary of an air-filled pore 0.5 mm in

diameter exactly coincides with the pixel edges, a scan image can theoretically show that pore. However, if the pore boundary falls in the middle of

the pixel, the scanner will produce an attenuation value that is an average

value for the materials within the pixel (e.g., soil and air). In practice, the

scanner actually monitors a volume of soil (in this case voxel dimensions

were 0.5 X 0.5 X 8 mm). Hence if any part of the voxel is filled with air

and the rest is filled with solid, the attenuation value for the voxel will be

an average between that for the solid and that for the air. Consequently,

studies of porous structure have been largely limited to macropores of

dimensions similar to pixel size.

The CAT image of an acrylic cylinder containing various sizes of holes is

shown in Fig. 15. The CAT scanner accurately depicted the location and

size (in millimeters) of the air-filled holes. Holes of 2.0 mm and larger in

diameter could easily be distinguished when the image was enlarged to

actual size or larger. Smaller pores, down to approximately 1.0 mm, were

distinguishable from the images but were often greyish in tone. Holes 0.5

mm diameter could not be distinguished with the pixel dimensions used.

The scanner used in this study had a scan field of 5 10 mm with 5 12 pixels,

i.e., one pixel corresponds to an object size of approximately 1 mm (0.996

mm actual). Any part of the measured field can be reconstructed with a

pixel size between 1 and 0.1 mm by selecting the zoom factor between 1

and 10. For example, a zoom factor of 10 corresponds to a measured field

5 10 mm in diameter with an object size of 0.1 mm per pixel. Thus, the use



Figure 15. X-Ray CAT scanning image of an acrylic cylinder containing various sizes of

holes (in millimeters).

of an object size of 5 1 mm in diameter would result in a higher resolution

(0.1 mm). However, large objects may have the advantage of greater

statistical significance.

Difficulties may also arise in high-contrast resolution because a zone of

drastically changing density, such as an air-filled hole or a stone in the soil,

can cause the destruction or blurring of information elsewhere in the scan

due to beam hardening and other mathematical anomalies. These include

lines and streaks (Fig. 16) developed in the plane of symmetry at a density

gradient interface and the Gibbs phenomena (Bracewell, 1956; Brooks and

Di Chiro, 1976), where large changes in attenuation from high-density to

low-density areas cause the calculated values in the adjacent low-density

area to be less than the actual attenuation coefficient in that region, i.e., an

overshoot. The size of object required to cause this type of artifact or the

extent of the perturbation will depend mainly on the attenuation differential and the characteristics of the CT scanner used. Many scanners are

better able than others to compensate for this artifact (Petrovic et af.,


When only small differences in attenuation exist between an object and

the surrounding soil, the ability to resolve the object accurately is reduced.

Thus Petrovic ef al. (1982) found that objects within the scanning plane

that differed in attenuation by I%, or about 10 Hounsfield units, had to be

at least 6.4 mm in diameter to be detectable using a pixel size of 1 X 1 mm.

Grevers et af.(1989) observed that spatial resolution in CT scans between



Figure 16. Streaking artifact in a CAT image of a vertical slice through a soil column.

soil pores and the soil matrix decreased as a result of impregnation with

resin because of the lower contrast in attenuation coefficient between pores

and the soil matrix than in the nonimpregnated samples. The difference in

Hounsfield values between the soil matrix and air-filled pores was around

1900H, whereas that between resin-filled pores and soil matrix was around

900H. Consequently, classification of the image into pore and soil matrix

was more difficult and led to an underestimation of macroporosity. These

CT images were, however, analyzed by applying an image-analyzing computer to grey-scale images from the scanner, which would have contributed

to the difficulty in spatial resolution. The use of actual Hounsfield values

for pixels from the CT scanner would undoubtedly have reduced this

discrepancy. Grevers ez al. (1989) concluded that the images of a soil

macropore system obtained from CT scanning compared favorably with

those obtained from thin sections and had the advantage of being nondestructive and less time consuming. Similarly, Anderson et al. ( 1990) examined the influence on effective resolution of blurring at the boundary of

constructed macropores by comparing measured hole diameters on the

basis of 50 and 75% differencesin mean bulk density (MBD) as determined

by the CT scanner. Although good agreement ( r = 0.97) was found between the CT-measured hole diameters and the wire diameters used in

constructing the holes for both the 50 and 75% MBD methods, fewer holes








+ B (0.5-1.0mm)


+ D (4.7-6.3mm)

+ E (6.3-8.0mm)

+ F Clod (Massive structure)

+ C (2.8-4.Omm)






















Macroporosity (%)

Figure 17. Macroporosity distribution for Scans of different sizes of awegates of Kulin

soil. (After Phogat and Aylmore, 1989.)

were detected using the more restrictive 50% MBD criteria as compared to

the 75% MBD method.

Several attempts have been made to develop quantitative characterizations of porosity and water content in soils on the basis of the distribution

of pixel attenuation. Phogat and Aylmore (1989) determined average macroporosity and the spatial and frequency distribution of macroporosity for

soil samples by assigning the value of zero macroporosity to pixels having y

attenuation coefficients corresponding to the bulk density of a soil aggregate, 100%macroporosity to pixels with zero attenuation coefficients, and

proportional values to pixels with intermediate coefficients (Fig. 17). CAT

scanning of soil samples before and after a wetting and drying cycle provided a quantitative illustration of the greater reduction in macroporosity

for soil wetted under flooding compared with that wetted under capillary

action (Fig. 18). A similar analysis of the distribution of pixel attenuation

values was used by Sawada et af. (1989) to characterize the degree and

extent of dispersion of soil water content and hence to evaluate the efficiency of soil wetting agents in overcoming water repellence. The possibility of relating the hydraulic conductivity of a porous glass bead system

directly to the spatial distribution and continuity of pore space as measured

by the CAT scanning procedure was examined by Phogat and Aylmore

(1993). A useful preliminary relationship for glass bead systems was obtained, indicating considerable potential usefulness in this approach.

Because of their inherent limitations, commercially available medical

X-ray scanners have thus far proved of most use in visual studies of soil

structure (Anderson el af., 1988; Jenssen and Heyerdahl, 1988; Tollner

and Murphy, 1991; Grevers et af.,1989),the advancement and stability of





















Macroporosity (‘A)

Figure 18. Changes in frequency distributions of macroporosity for York soil after

wetting under flooding or capillary action, followed by drying.

wetting fronts, and the structural changes following wetting and drying

(Phogat, 1991). In addition, roots, seeds, insects (Tollner et al., 1987), and

pesticide granules (Cheshire et al., 1989)within soils have been successfully

detected using an X-ray CAT scanner. In the author’s laboratory CAT

scanning has also been used successfully in continuous monitoring of the

activity of worms and dung beetles under different soil conditions as well

as in studying the intensity of root nodule development. The quantitative

applications of CAT to soil water studies have, however, been largely

limited to statistical assessments of macroporosity distributions before and

after complete wetting and drying cycles and to the measurement of water

drawdowns in proximity to plant roots in nonswelling soils (Hainsworth

and Aylmore, 1986; Aylmore and Hamza, 1990; Hamza and Aylmore,






The flow of water to plant roots is part of an overall catenary process of

water transport from the soil through the plant and into the atmosphere. It

is a predominantly passive process whereby water moves in response to

gradients in its total water potential. Van der Honert (1948) applied an

analog of Ohm’s Law to the steady-statetranspirational flux, and although

he did not examine water movement through the soil to the root surface,

this can be easily incorporated because the whole pathway from soil to

atmosphere forms a thermodynamic continuum (Slatyer, 1967).



Consequently, the pathway of water flow can be conveniently represented by a series of resistances: the flux of water (F)from the soil to the

plant and out into the atmosphere under steady-stateconditions is given by

the gradient in water potential through the soil/plant /atmosphere continuum divided by the total resistance. Thus


- Y,)/R,


where Ysis the average soil water potential, Yl is the average leaf water

potential, and Rt is the total resistance to water flow. Thus as a first

approximation, Rt can be assumed to be the sum of the soil and plant

resistances in series,



Rt = R, iR,


with the soil resistance & and the plant resistance 4 being defined by






respectively, in which Yois the average water potential at the outside

surface of the root.

In recent years numerous models aimed at a quantitative formulation

and prediction of the processes involved in the movement of soil water and

its extraction by plant roots have been developed (Hillel, 1982). Not only

do these models differ in aim and level of detail, but also in approach. Two

alternative approaches have generally been taken in modeling water uptake

by plant roots. The first can be described as the microscale or single-root

approach based on the radial flow of water to a cylindrical sink. Solutions

to this approach have been attempted both by analytical means (Gardner,

1960; Cowan, 1965), usually requiring restrictive assumptions (Hillel,

1982), and by numerical means (Molz and Remson, 1970; Hillel et al.,

1975). However, such models have differed widely in their quantitative

prediction of water extraction. Thus, as pointed out by Hillel (1982), there

are numerous theoretical models, lacking in consensus, which remain

largely untested and hence unproved. It has become easier to formulate

than to validate models of water extraction by plant roots (Belmans et al.,


There has, in particular, been considerable controversy as to the relative

magnitudes of the soil and plant resistances (Newman, 1969a,b; So et al.,

1976, 1978; Reicosky and Ritchie, 1976). When soil moisture is at field

capacity, the influence of soil resistance on water uptake by the plant is

likely to be small. As soil moisture approaches the wilting point, the

Tài liệu bạn tìm kiếm đã sẵn sàng tải về

V. Application of Computer-Assisted Tomography to Soil–Water Studies

Tải bản đầy đủ ngay(0 tr)