Simultaneous Measurement of the Spatial Distribution of Soil Water Contentand Bulk Density

V. K. Phogat, L. A. G. Aylmore,* and R. D. Schuller

ABSTRACTContinuous nondestructive measurements of the spatial distribu-

tions of soil water content and bulk density are essential prerequisitesto the resolution of many problems in the study of soil-plant-watersystems. Computerized axial tomography (CAT) was applied to dual-source (l37Cs and l69Yb) gamma-ray attenuation in soil columns todetermine nondestructively the spatial distributions of volumetricwater content (6,) and bulk density (p,) for two soils that exhibitswelling to different degrees. Beam slice thickness was 2 mm andpixel dimensions 2 by 2 mm. The CAT scanning technique for av-erage p, and 0, provided excellent agreement with the correspondingvalues obtained gravimetrically. Average standard deviations (a) forp, of pixels for the dry soils, although huge, were highly reproducibleand provide a measure of the structural status of the soils. Changesin the mean and a of pixel p, following wetting were related to thetexture, structure, mineralogy, and Ca or Na exchange status of thesoil. Systematic errors arising from the random nature of radioactiveemissions, for count times of 2 s, resulted in large <r in estimated 0,of pixels even for dry soils. Very large counting times (some 169 sfor an individual ray-sum and, hence, 112 h to complete a dual-sourcescan) were required to achieve a values of pixel 6, on the order of0.025 cm'/cm3 for a uniform field of water. Such large count timeslimit this approach to steady-state or only slowly changing systems.Thus, while the application of CAT to dual-energy-level scanninghas the potential to become a major tool for nondestructive studiesof soil water content, structural status, and stability, realization ofthis potential awaits substantial improvements in scanning geometryand counting electronics.

SINGLE- AND DUAL-SOURCE gamma attenuationmeasurements have been extensively used by soil

scientists to obtain nondestructive measurements ofps and Ov in soil columns. Various isotopes have beenused as sources of the gamma radiation including ^Co(Herman and Harris, 1954; Gardner and Calissendorff,1967); 137Cs (Ferguson and Gardner, 1962; Gurr,1962); 241Am (King, 1967; Groenevelt et al., 1969) and144Ce (Gardner and Calissendorff, 1967). Any two gam-ma radiation sources of sufficiently different energiescan be employed for simultaneous and nondestructivemeasurement of ps and 0V (Soane, 1967).

More recently, application of CAT (Hounsfield,1972) to x- and gamma-ray attenuation has providedan exciting new method for nondestructive three-di-mensional imaging within a solid matrix. Hopkins etal. (1981) and Davis et al. (1986) demonstrated theapplication of an x-ray CAT scanning technique toindustrial problems, particularly for nondestructivetesting of timber poles, concrete pillars, and steel-belt-ed automobile tires. Onoe et al. (1983) illustrated theuse of a portable x-ray CAT scanner for measuringannual growth rings of live trees.

The first reported attempts of its use to measureSoil Science and Plant Nutrition, Univ. of Western Australia, Ned-lands, Western Australia 6009. Received 16 July 1990. 'Correspond-ing author.

Published in Soil Sci. Soc. Am. J. 55:908-915 (1991).

spatial changes in soil bulk density was by Petrovic etal. (1982). Subsequently, CAT scanning has been ap-plied using single-energy x- or gamma-ray attenuationto measure the spatial variation in 0V in proximity toplant roots (Hainsworth and Aylmore, 1983, 1986,1988, 1989; Aylmore and Hainsworth, 1988). Severalother investigators (Crestana et al., 1985,1986; Brownet al., 1897; Anderson et al., 1988; Tollner and Verma,1989) have also illustrated the use of x-ray CAT fornondestructive measurement of ps and/or 0V of porousmedia. Phogat and Aylmore (1989) used single-source(137Cs) gamma CAT scanning to examine the spatialdistribution of soil macroporosity and for monitoringthe changes that occur during wetting and drying pro-cesses in a nondestructive manner. Similarly, x-rayCAT scanning has also been used for characterizingthe macropores in soil (Grevers et al., 1989; Warneret al., 1989; Anderson et al., 1990). In addition, roots,seeds, and insects (Tollner et al., 1987) and pesticidegranules (Cheshire et al., 1989) within soils have beensuccessfully detected using an x-ray CAT scanner.

One of the limitations of the use of single-energy x-or gamma-ray CAT scanning systems in studies in-volving measurements of the spatial distribution of 0Vin soils is the assumption of uniform ps. Because theattenuation is a function of both ps and 0V of the soil,an accurate determination of 0V in soils is not possiblewhen changes in ps occur during experiments (Petrovicetal., 1982; Hainsworth and Aylmore, 1983; Andersonet al., 1988; Phogat and Aylmore, 1989).

To monitor changes in the spatial distribution of psand 0V in situations where the ps of the soil changesdue to swelling or shrinking on addition or removalof water requires independent estimates of attenuationassociated with both ps and 0V. This can only be ob-tained by the simultaneous use of two sources of dif-ferent energies. Gamma-ray-sourced systems aregenerally preferred because monochromatic photonbeams are readily obtained. However, gamma-raysources emit a much smaller photon flux than x-raysources and, unless a very active source is used, scantimes are consequently longer. The objective of ourstudy was to evaluate the feasibility of applying theCAT scanning procedure to dual-source gamma atten-uation to enable simultaneous measurements of thespatial distributions of ps and 0V in swelling soils.

THEORYThe theory and use of the CAT technique for med-

ical purposes have been reviewed in some detail byBrooks and Di Chiro (1975,1976), Budinger and Gull-berg (1974), and Panton (1981), and a complete reviewof various aspects of CAT scanning has been presentedby Newton and Potts (1981, p. 3853-3917). A reviewof CAT-scan theory as it relates to the determinationof 0V was presented in detail by Hainsworth and Ayl-more (1983), Crestana et al. (1985), and Anderson et

908

PHOGAT ET AL.: SIMULTANEOUS MEASUREMENT OF WATER CONTENT AND BULK DENSITY 909

al. (1988). The following provides a brief outline ofthe theory as required for dual-gamma-source CATscanning.

Following Hainsworth and Aylmore (1983), the lin-ear attenuation coefficient (M) for each pixel for drysoil can be described as

Mdry = MsPs [1]

where ns is the mass attenuation coefficient in cm2/g for soil solids and ps is in g/cm3.

Equation [1] can be extended to wet soil asMwet = MsPs + MA [2]

where MW is the mass attenuation coefficient of water.In situations where the ps of the soil does not change

with the addition or removal of water, 0V can be cal-culated from a combination of Eq. [1] and [2] to give

"v = (Mwet - Mdry)/Mw- [3]

For two gamma-ray energies, the attenuation equa-tions for Eq. [2] may be written as

Mwet a = MsaPs + Mwa^v [4]

Mwetb = MsbPs + MwA [5]

where the subscript a refers to the low-energy radiationand subscript b refers to the high-energy radiation.Thus, Mwa» Mwb, Msa> and jtsb are the mass attenuationcoefficients for water and soil solid, respectively.

Equations [4] and [5] can be solved simultaneouslyto givePs = t(Mwb Mwet a) ~ (Mwa Mwet b)]/[(Mwb Msa)

- G^b Mwa)] [6]

#v = [(Msb /"wet a) — (Msa Mwet b)]/[(Msb Mwa)~ (Mwb Msa)] • [7]

Thus, ps an 0V for an individual pixel can be cal-culated by scanning the wet soil column with both theradiation sources at a fixed position.

In general, we are interested in the precision of bothPs and 0V measurements. Therefore, the standard de-viations in the determination of the dry ps [ff(ps)J andthe 0V [<r(0v)] using dual sources can be written, similarto Gardner et al. (1972), as

*(Ps) = (Mwb <Twet a + Mwa *wet b)1/2M [8]

<K0V) = (ML <4* b + M2b *wet a)

1/2M [9]where <rwet a and <7wet b are the standard deviations ofthe attenuations of the wet soil determined by the low-and high-energy sources, respectively, and A is equaltO (MwbMsa) - (MsbMwa)-

The standard deviations in Eq. [8] and [9] refer tothe variation in the wet soil attenuation measurementsMw,,, a and Mwe, b. These are considered to be randomvariables subject to statistical fluctuation and are theonly source of random errors in ps and 0V, while theattenuation coefficients p^, Mwb. Msa, and MSb are as-sumed to be fixed, known constants that do not con-tribute to the random errors of ps and 0V. As these arenot random variables, we cannot refer to means. Inpractice, experimental measurements of Mwa, Mwb. Msa,and Msb are subject to statistical variation, but it isassumed that these uncertainties can be made negli-

gibly small by taking the mean of repeated measure-ments, which, in practice, requires making them morethan an order of magnitude smaller than the errors inM^t a and Mwetb- It should also be noted that, althoughany errors in Mwa, Mwb Msa, and Msb would show them-selves as systematic errors in ps, 0V, o{ps) and <r(0v), theydo not play any part in the randomness of ps and 0V,as they are still treated as constants in Eq. [6] to [9].

MATERIALS AND METHODSComputer-Assisted Tomography Scanning System

A prototype CAT scanning system constructed in the SoilScience and Plant Nutrition Laboratories of the Universityof Western Australia (Hainsworth and Aylmore, 1988) wasused for these studies. Commercially available 137Cs (1.85X 10'° Bq) and 169Yb (7.4 X 10'° Bq) gamma sources weremounted in a Pb castle opposite to a Nal(Tl) scintillationdetector attached to a photomultiplier tube. The signal fromthe photomultiplier tube passed through a tube base-pream-plifier (Model 2007P, Canberra-Packard Instrument PtyLtd., Mt. Waverly, Victoria, Australia1) before being fed toan amplifier (Model 2012, Canberra-Packard). Other com-ponents of the detection system included a high-voltage pow-er supply (Model NE 4646, Nuclear Enterprises, Reading,England), a single-channel analyzer (SCA) (Model 2030,Canberra-Packard) and a dual counter (Model 2072 A, Can-berra-Packard) with an RS-232 interface for communicationwith an IBM PC. The counts in the desired energy rangeswere discriminated by the SCA and subsequently countedseparately by the dual counter. The beam was collimated togive a slice thickness of 2 mm and pixel size of 2 by 2 mm.The Pb castle contained a mechanism by which the twosources could be alternately brought into line with the col-limator and detector. The data acquisition system as well asthe movements of translation and rotation of the samplewere controlled by the PC.

As the source and detector were fixed, the object wasmoved across the beam and scanned at 2-mm intervals. Ineach linear scan, two air counts were required on both sidesof the object to allow the calculation of linear attenuationcoefficients for the object. Successive linear scans were madeafter rotating the object progressively in 5° incrementsthrough 180°. Once this process was completed, the linearscans were back-projected by using filtered back-projection(Hermon, 1980) for a given number of rotations to recon-struct an image of the scanned cross-sectional slice. Oncethe back-projections were completed, the gamma attenua-tion values for each pixel in the slice were determined. Thewhole process was repeated for both sources. The values ofps and 0V for individual pixels were then calculated using Eq.[6] and [7], where M .̂,,, and Mwctb correspond to the individualpixel attenuations due to l69Yb and 137Cs, respectively, ineach of the scans. This gave separate maps of ps and 0V inthe scanned slice.

Choice of SourcesA combination of 137Cs and241 Am sources has most com-

monly been used in conventional dual-source scanning be-cause of the tenfold difference in gamma-ray energy (662and 60 keV, respectively) and their long half-lives. However,both the time required to successfully complete a CAT scanand the precision obtained depend on the transmission in-tensity and, hence, count rate. Gamma-ray sources generallyemit much smaller photon fluxes than x-ray sources. While

1 System identification is provided solely for the benefit of thereader and does not imply the endorsement of the University ofWestern Australia.

910 SOIL SCI. SOC. AM. J., VOL. 55, JULY-AUGUST 1991

high-strength 137Cs sources are readily available, the beamstrength obtainable from 24IAm has a practical limit becauseof self-absorption (Miller, 1955). Scan times required withthis source are not sufficiently rapid to follow the rapidchanges in soil 0V that may be associated with, for example,water extraction by plant roots or water infiltration into thesoil surface. Clearly in dual-source CAT scanning, the min-imum scanning time commensurate with adequate precisionwill generally be desirable. Hainsworth and Aylmore (1988)suggested the use of 169Yb as an alternative source to 241Am,since this provides a similar energy level (63.1 keV) andattenuation coefficient for water, but emits much higher pho-ton outputs (more than 20 times) at high activities (e.g., 3.7-7.4 X 1010 Bq). The major disadvantage of the 169Yb sourceis its relatively short half-life of 31 d, resulting in a workinglife of approximately 2 mo. In our study, the combinationof a 7.4 X 10'° Bq source of 169Yb and 1.85 X 10'° Bq sourceof 137Cs was used in an attempt to reduce scan times andincrease precision and accuracy.

Experimental ProcedurePhysical and chemical properties of two surface (0-7 cm)

soils collected from Kulin and York, Western Australia aregiven in Table 1. The principal differences between the twosoils are the higher silt and clay contents in the York soiland a dominant clay mineral of smectite in the York andkaolinite in the Kulin soil. Mass attenuation coefficients forthese soils were determined as follows. Soil samples wereoven dried, passed through a 2-mm sieve, and packed to aknown p, in acrylic columns of 5.4-cm i.d. and 3.0-cm height.An additional column was filled with deionized water. Eachcolumn was scanned with each source at four successivelayers of 2-mm thickness using a counting time of 2 s foreach ray-sum. Average /t over the entire sample cross-sec-tions were obtained for each column and for each sourceand substituted into Eq. [1] with the known p, values toobtain ̂ and /*sb for the two soils. The values of n^ and/^b were taken to be equal to the average values of n for thewater column scanned with 169Yb and with 137Cs, respec-tively. The density of water was taken as 1.0 g/cm3.

To obtain soil samples with different 0V, segmentable acryl-ic columns of 5.4-cm i.d. consisting of four 1.0-cm-long seg-ments were used. Whatman no. 50 filter paper was fitted tothe bottom of each column. Using oven-dry soil, nine col-umns of each soil were packed at a p, of 1.41 ± 0.020 g/cm3 for Kulin soil and 1.36 ± 0.023 g/cm3 for York soil, toa height of 3.0 cm. One centimeter was left empty to allowfor any swelling. All columns were saturated from the bottomand then equilibrated at appropriate suctions for a period of5 d. Three successive slices of 2 mm each were scanned ineach column through the second segment from the bottomwhen dry and rescanned at equilibrium moisture contentwith each source. Mean /t values over the entire samplecross-sections of all the scanned slices were determined. Theaverage ps and 0V for each slice were calculated using Eq. [6]and [7], and results were averaged for each column. Afterscanning, each column was carefully dismantled and averageps and 0V in the scanned segments were determined.

Table 1. Selected physical and chemical properties of the soils usedin this study.______ _

Minerals Cation-Textural Coarse Fine in clay exchange

Location class______sand sand Silt Clay fraction! capacity—————— (%) —————— cmoiykg

8.20

18.93

Kulin Sandy Kaolinite,loam 47.72 24.52 10.5 17.31 illite

York Sandy clay Smectite,loam 17.15 32.75 26.5 23.58 kaolinite

To study the mean and spatial distribution of p, and 0Vand to monitor the spatial changes in their distribution onwetting or application of salt solutions, three columns (Yl,Y2, and Y3) of 5.4-cm i.d. and 35-cm length were packedwith York soil aggregates (2.8-4.0 mm). In addition, onecolumn (Kl) was packed with Kulin soil (<2.0 mm). Ini-tially, the four columns were scanned dry at depths of 0.5,1.5,4.5, and 9.5 cm from the soil surface with both radiationsources. Deionized water was applied at the soil surface ofColumns Kl and Yl, through a coarse porous glass disk (5.1-cm o.d.). A positive head of 0.5 cm of water was maintainedat the surface by appropriate positioning of a Mariotte bub-bling tube. Water supply to both the columns was stoppedafter the wetting front had travelled to a depth of 18 to 20cm. Columns Y2 and Y3 were treated in a similar mannerusing 0.1 M CaCl2 and 0.1 M NaCl solutions, respectively.The four columns were again scanned at the same depths.The following day, deionized water was applied to the sur-face of the columns previously treated with CaCl2 and NaClsolutions, i.e., Y2 and Y3, and these were scanned at thesame positions to observe the relative effects of the divalentand monovalent cations on structural status. The mean andstandard deviations of ps and 0V in each soil slice scannedwere determined from the calculated pixel ps and 0V acrossthe entire cross-section of the slice, using Eq. [6] and [7].

The precision of the dual-gamma-ray CAT scanning tech-nique for obtaining the spatial distributions of ps and 0V wasstudied by scanning columns filled with deionized water andYork soil (<2.0 mm) with both the sources using a rangeof counting times. Means and standard deviations of pixelps and 0V were calculated for the sample cross-sections foreach counting time.

RESULTS AND DISCUSSIONMean and standard deviations of the mass atten-

uation coefficients obtained from four slices for thetwo soils and three for water, using each radioactivesource, are given in Table 2. These values of massattenuation coefficients were used in Eq. [6] and [7]to calculate the unknown ps and 0V of these columns.The maximum errors associated with the dual-gam-ma-source CAT scanning technique in the determi-nation of average ps and 0V were calculated from thevalues of standard deviation of mass attenuation coef-ficients for the two soils and deionized water. Usingthe latter values, Eq. [8] and [9] were solved for thescan of the uniform field of water to yield a value of0.009 g/cm3 for <r(ps) and 0.011 cm3/cm3 for (r(0v). Thevalues of a(ps) calculated in a similar way for Kulinand York soils were 0.039 and 0.038 g/cm3, respec-tively. The value of <r(0v) for both the soils was 0.042cm3/cm3. The larger values of <r(ps) and a(8v) for thesoils arise from the much greater variation in ps and0V in the soil columns compared with those for thewater column.

The values of mean 0V for slices of the two soils

Table 2. Means and standard deviations of mass attenuation coef-ficients obtained by computerized axial tomography (CAT) scan-ning for various experimental materials._____________

Mass attenuation coefficient137Cs

Material Mean SD of mean Mean SD of mean

t Minerals in clay fraction are by decreasing dominance.

WaterKulin soilYork soil

0.1756500.2073040.222639

0.0002760.0012560.001483

0.0814400.0741990.077526

0.0001980.0006170.000714

PHOGAT ET AL.: SIMULTANEOUS MEASUREMENT OF WATER CONTENT AND BULK DENSITY 911

determined by the dual-source gamma CAT scanningtechnique (Fig. 1) showed good agreement (R2 =0.991) with the values obtained gravimetrically (butexpressed volumetrically). The differences in 0V deter-mined by the two methods fall within the range of± 0.024 g/cm3. Also shown in Fig. 1 are the relation-ships between 0V and the water content that wouldhave been obtained if the calculation had been madeon the basis of Cs attenuation alone for both the soils(rather than the dual-source CAT scanning technique),which neglects ps changes. For both soils, the single-source (137Cs) CAT scanning technique markedlyunderestimated 0V at or near saturation, undoubtedlyas a result of a decrease in ps due to swelling, andslightly overestimated 0V at or below values of 0.27cm3/cm3 in the Kulin and 0.05 cm3/cm3 in the Yorksoil. The latter effect is undoubtedly due to an increasein ps because of shrinking of the initially saturated soilmatrix with decrease in 0V. The differences in behaviorof the two soils during the desorption cycle with re-spect to 0V, compared with the 0V calculated using 137Csalone, can be attributed to the differences in the claymineralogy.

The results for the mean ps in slices of the two soilsobtained using dual-gamma-ray CAT scanning andcalculated from Eq. [6] are compared with the gravi-metrically obtained values in Fig. 2. The agreement isagain good, with an R2 value of 0.995. The variationin the two determinations was approximately ±0.015g/cm3.

It is apparent from Fig. 1 and 2 that the interceptsfor ps and 0V do not extrapolate to the origin, as wouldbe anticipated. This may be largely attributed to errorsin soil volume measurements arising during disman-tling and segmenting of the columns.

It is clear from these results that the dual-gamma-ray CAT scanning technique is capable of successfullymeasuring, simultaneously and nondestructively, themean ps and 6V in both swelling and nonswelling soilswith excellent accuracy. However, the primary objec-tive of the CAT technique is not to measure average

ps and 0V but to reveal the spatial distributions of psand 0V in soils.

Spatial Distribution of Water Contentand Bulk Density

The means and standard deviations of ps and 0V forthe pixels in each scan of the Kulin and York soilsamples under various treatments and at variousdepths are shown in Table 3. The average <r(ps) forpixels in the initial dry scans for the Kulin and Yorksoil samples were 0.33 and 0.78 g/cm3, respectively.The larger value of <r(ps) for pixels for the York soilsamples reflects the greater heterogeneity arising fromthe spatial distribution of soil and pore spaces in thesecolumns packed with 2.8- to 4.0-mm size aggregates,compared with the Kulin soil samples, which werepacked with <2.0-mm size aggregates. Thus, the dif-ferent values of <r(ps) for pixels illustrate the differentstructural status of the two soils.

When water was applied to the surface of these soilsamples, <r(ps) for pixels in scans for the York (Yl) soilsamples was reduced from 0.796 to 0.560 g/cm3, in-dicating a decrease in heterogeneity of the scannedslices. Presumably, this was because of slaking orswelling of the soil aggregates. In the case of the Kulinsoil, the value of <r(ps) for pixels remained essentiallythe same, presumably because of the smaller size (<2.0-mm) aggregates used and the lower inherent swell-ing associated with kaolinite clay. However, the av-erage ps of both the soils decreased on wetting and thisis undoubtedly due to the swelling that occurs. In bothsoils, the magnitude of the reduction in ps was greaterin the surface layers compared with the layers at otherdepths.

A comparison of the values of <r(ps) for pixels forscans of the York soil (Yl and Y2) demonstrates thatthe soil retained its structure to a greater extent whentreated with 0.1 M CaCl2 (Y2) than when treated withwater (Yl). The value of a(ps) for pixels of the surfacelayer treated with 0.1 M CaCl2 (Y2) decreased from0.783 to 0.718 g/cm3 whereas, in the case where water

0.6

0.5

o> 0.4

Oo0.3

CO

I-O

COECOO

0.2

0.1

0.0

o Dual-Yorko Dual-Kulin• Cs-York• Cs-Kulin

Y = 0.0041 + 0.9821 XRA2 = 0.991

0 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6Volumetric water content (cm3/cm3)

Fig. 1. Relationship between volumetric water content and watercontent determined by dual- and single-energy gamma comput-erized axial tomography (CAT) scanning for soils from Kulin andYork, Western Australia.

0)•o

1.5

1.4

•" JT*

!««o sCOECOO)

CO

Q

1.3

1.2

1.1

• Yorko Kulin

Y = 0.0079 + 0.996 XR"2 = 0.995

1 .2 1 .3 1 .4Bulk density (g/cm3)

1 .5

Fig. 2. Relationship between gravimetric bulk density and bulk den-sity determined by dual-energy gamma computerized axial to-mography (CAT) scanning for soils from Kulin and York, WesternAustralia.

912 SOIL SCI. SOC. AM. J., VOL. 55, JULY-AUGUST 1991

Table 3. Means and standard deviations of pixel bulk density and pixel water content in scans for soils from Kulin and York, WesternAustralia, under various treatments at different depths.

Soil Depth

Initial dry After solution After waterBulk density Water content Bulk density Water content Bulk density

Mean SD Mean SD Mean SD Mean SD Mean SDWater contentMean SD

-g/cm3 - cmVcm' —— ——— g/cm3 - - cm3/cm3 - -g/cm3 - cm3/cm3 -Kulin (Kl)

York (Yl)

York (Y2)

York (Y3)

0.51.54.59.5

Average0.51.54.59.5

Average0.51.54.59.5

Average0.51.54.59.5

Average

1.4241.4301.4031.3931.4121.0601.0811.0661.0311.0601.0721.0681.0631.0121.0541.1091.0811.0751.0011.067

0.3440.3420.3250.3200.3330.8360.7570.7740.8150.7960.7830.7760.7740.7440.7690.7640.7210.7790.8060.767

-0.0030.005

-0.0010.0010.0000.003

-0.0010.0010.0010.0010.006

-0.0020.0140.0010.0050.0050.001

-0.0030.0020.001

0.3620.3620.3400.3410.3510.7080.6330.6330.7290.6760.7630.6760.6120.7410.6980.6380.6720.6900.6190.655

1.0511.0351.0671.0111.0410.9540.9220.9590.9000.935

0.7180.7580.5720.6450.6730.5100.6470.6790.7050.635

0.3460.3180.2970.2750.3090.4140.3910.3620.3150.371

0.5340.5230.6440.6830.5960.4370.4220.5450.5150.480

.221

.242

.256

.393

.2780.8680.8800.9311.0010.9200.9240.9220.9871.0080.9600.8890.8740.9530.9970.928

0.3360.3610.3620.3230.3460.5060.5650.5230.6470.5600.6470.7010.5870.6850.6550.4440.5460.5840.5640.534

0.5180.4560.3780.0060.3400.4910.4700.4290.3700.4400.4360.4030.3450.3110.3740.4960.4290.3340.3120.393

0.3450.3680.3820.3440.3600.4770.4800.5180.5400.5040.4230.4870.5580.5820.5130.3970.4380.4800.4920.452

was applied (Yl), the decrease was from 0.836 to 0.506g/cm3. As could be expected, the data illustrate thatthe presence of Ca2+ reduces the swelling or slaking ofthe York soil aggregates. Under the application of 0.1M NaCl (Y3), there was a greater reduction in pixelo-(ps) in the surface layer (0.764 to 0.510), comparedwith the CaCl2 treatment, indicating that aggregates ofYork soil were more susceptible to slaking or swellingunder the NaCl treatment. When water was appliedto the columns previously treated with CaCl2 andNaCl solutions (Y2 and Y3, respectively), <r(ps) for pix-els for the surface layer in the NaCl-treated soil de-creased to 0.444, compared with a value of 0.647 forthe CaCl2 treatment. Thus, as expected, a(ps) for thesurface layer demonstrates that the susceptibility ofaggregates of York soil to disruption is increased whenprewetting with NaCl is followed by water application,because of the dispersive effects of the Na+ ion.

Perusal of the values of o-(0v) for pixels in scans ofthe dry soil samples (Table 3) indicates that the av-erage 0V in scans for the dry Kulin soil approximatedto zero, but with a o-(0v) for pixels of ±0.351 cm3/cm3.In the case of scans for the dry York soil samples, theo-(0v) for pixels was as high as 0.676 cm3/cm3, with anaverage 6V of 0.003 cm3/cm3. The magnitude of o-(0v)for pixels in scans for both the dry soils are surprisinglyhigh. These systematic errors can undoubtedly be at-tributed to propagation of errors associated with theinherent randomness of gamma emissions from theradiation sources. Dual-source scanning is subject tothe multiplicative variation applicable to measure-ments of gamma emission and attenuation for bothsources. Clearly, individual pixel values are subject tosubstantial variation and there are errors in the esti-mation of both individual pixel ps and 0V in all scans.As demonstrated above, however, the mean values ofps and 0V provide accurate estimates of these quantities

co 0.055 1 2 1 44 6 8 1 0

Counting time (sec)Fig. 3. The standard deviations in dual-gamma computerized axial

tomography (CAT) scan determinations of pixel bulk density fora soil from York, Western Australia, and pixel water content ina uniform field of water at different counting times.

and <r(ps) and o-(0v) for pixels are highly reproduciblebetween slices.

The errors introduced by the random nature of theemission from the radioactive sources can, however,be reduced by increasing the counting time. The valuesof <r(ps) or o-(0v) for pixels in scans of a uniform fieldof water and dry York soil (<2.0 mm) at differentcounting times are given in Fig. 3. This clearly indi-cates that the maximum <r(0v) for pixels associated withthe dual-gamma-ray CAT scanning system (at thecounting time of 2 s used in the determinations) isaround 0.18 g/cm3, as this value was obtained for auniform field of water. Furthermore, Fig. 3 shows thatfurther improvement in the determination of the spa-tial distributions of ps and 0V can be achieved by usinglonger counting times, as a(0v) and <r(ps) for pixels de-crease with increasing counting time. The <r(ps) for pix-els for York soil scans remained above the valuesobtained for water scans at each counting time because

PHOGAT ET AL.: SIMULTANEOUS MEASUREMENT OF WATER CONTENT AND BULK DENSITY 913

0.1

u1

I 2.0

OU

13.0

Cs-137 Yb-169 Dual source

Gamma radiation source

Fig. 4. Reconstructed half-slice three-dimensional computerized axial tomography (CAT) scanning images showing the spatial distributionof water content in slices of a uniform field of water scanned with 137Cs and 169Yb at different counting times and calculated for both single137Cs- and "9Yb-sources as well as for the dual-gamma-source CAT scanning procedure.

of the inherent variations in the soil samples. A count-ing time of 13 s gave a <r(0v) of 0.073 cm3/cm3 for auniform field of water, which could be considered asthe highest precision using this system and countingtime. In contrast, the value of <r(ps) for pixels for theYork soil at a counting time of 13 s was 0.141 g/cm3,showing an increase of 0.068 g/cm3 due to inherentvariations in ps across the entire cross-section of thescanned slice.

To illustrate the effect of counting time on o(0v) ander(ps) for pixels associated with dual-source gammaCAT scanning, three-dimensional representations ofthe distribution of pixel attenuation and estimated 0Vin a uniform field of water, scanned with 137Cs and169Yb and calculated using Eq. [7] with counting timesof 0.1, 2.0, and 13.0 s are shown in Fig. 4. The heightof the surface represents 0V for each pixel in the slice.These images illustrate that, as the counting time isincreased, the variation in pixel 0V decreases and, as

a result, the surface of the images becomes more uni-form for all three methods. The higher photon outputof the 169Yb and the fact that the average attenuationdue to Yb is greater than that due to Cs reduces therelative error (the absolute error in attenuation beingroughly the same) and explains the greater uniformityof the Yb scans compared with the Cs scans. At acounting time of 13 s, 137Cs gave a «r(0v) for pixels of0.016 cm3/cm3, whereas the corresponding value forthe 169Yb scan was ±0.007 cm3/cm3. When the scandata of both the sources were used in Eq. [7], it yieldeda <r(0v) for pixels of 0.073 cm3/cm3. This variation inpixel 0V using both the sources is high in spite of thevery low values for scans of the individual sources(137Cs and 169Yb). This multiplicative propagation oferrors is most evident when due to random emissions,a particular pixel estimated to have a low attenuationusing 137Cs is estimated to have a high attenuationusing 169Yb or vise versa. As a result, small variations

914 SOIL SCI. SOC. AM. J., VOL. 55, JULY-AUGUST 1991

in the Cs and Yb scans can give rise to large and un-acceptable variations when Eq. [6] and [7] are applied.The observed errors are confirmed by Eq. [8] and [9],showing that errors are solely attributable to the ran-dom variation in the single-source scans resulting fromthe inherently random nature of gamma emission.

Large variation in pixel ps calculated from Eq. [6]for scans of York soil (Fig. 5) at the three countingtimes (0.1,2.0, and 13.0 s), when compared with waterscans (Fig. 4), reflects the inherent variation in pixelps of the soil scans. The distribution of ps in a slice ofYork soil before and after wetting, as determined bythe dual-gamma-ray CAT scanning technique using acounting time of 13.0 s is shown in Fig. 6. The CAT

Fig. 5. Reconstructed half-slice three-dimensional computerized ax-ial tomography (CAT) scanning images showing the spatial dis-tribution of pixel bulk density obtained by dual-gamma CAT scan-ning for counting times of (a) 0.1 s; (b) 2.0 s, (c) 13.0 s for aninitially dry soil, and (d) 13.0 s for the soil after wetting.

120

at 100a>X

0)A£

Initial dry (dual)After wetting (dual)Cs-aloneYb-alone

0.8 1 .0 1 .2 1 .4 1 .6

Bulk density (g/cm3)1.8 2.0

Fig. 6. The distribution of pixel bulk density in a slice of soil obtainedby dual-energy gamma computerized axial tomography (CAT)scanning before and after wetting, using a counting time of 13.0s, and the distribution of bulk density in the slice that would havebeen calculated if single-energy 137Cs or 169Yb scanning had beenused.

images of these scans, showing the spatial distributionof pixel ps before and after wetting are shown in Fig.5c and d, respectively. The shift in the distributioncurve for ps towards the left in Fig. 6 shows the de-crease in ps of the York soil on wetting. The distri-butions of pixel ps using single radioactive sources(137Cs and 169Yb) at the same counting time representsthe actual variation in pixel ps within the scanned slice.The excessive spread of the distribution of ps calcu-lated from the dual-gamma-ray CAT scanning, bothbefore and after wetting of York soil, is the result ofpropagation of the single-source errors describedabove. The systematically higher ps of York soil nearthe column walls before and after wetting in Fig. 5cand 5d illustrates the fact that the packing of the soilcolumns on the vibrator resulted in an uneven ps, sincethis effect was not visible in scans of the column filledwith water (Fig. 4).

Even at the counting time of 13 s used here, thevalues of <r(0v) and <r(ps) for pixels remain unacceptablylarge for most purposes. Estimates of the counting timerequired to provide values in a more acceptable rangecan be made using the relationship between a(6v) forpixels in a scan of a uniform field of water and count-ing time (Fig. 3). Use of this relationship indicates that,to obtain a er(0v) on the order of 0.05 cm3/cm3, anindividual ray-sum counting time of 35 s would berequired, resulting in a total scan time of 23 h for a5.8-cm o.d. column. This includes the time taken toobtain air counts on either side for calibration pur-poses. Assuming a normal distribution of counts, acounting time of 169 s would be required to ensurethat 95% of the pixels in the scan for 0V were ±0.05cm3/cm3 (i.e., er(0v) = 0.025 cm3/cm3). This would re-quire a total of 112 h to complete one dual-source scan.

These calculations clearly illustrate that, using dual-source gamma CAT scanning, it is possible to measurethe spatial distributions of ps and 0V in soils simulta-neously and nondestructively with a satisfactory levelof precision. The total scanning time required to ob-tain this level of precision, however, severely limitsthe speed and flexibility of the system and, hence, itsusefulness in studying these properties in situationswhere they change with time, for example, during in-filtration and redistribution processes.

CONCLUSIONSApplication of CAT to dual-energy-level scanning

clearly has the potential to become a major tool fornondestructive studies of the structural status and sta-bility of soils and for spil-plant-water relationships.However, the use of existing gamma source-detectorsystems is likely to be limited to steady-state or onlyslowly changing systems by the scanning times re-quired to obtain satisfactory precision in such mea-surements. Realization of the full potential of thisexciting new technique awaits substantial improve-ments in scanning geometry and counting electronics.

ACKNOWLEDGMENTSV.K. Phogat is grateful to the Australian International De-

velopment Assistance Bureau for the award of an Australian

PHOOAT ET AL.: SIMULTANEOUS MEASUREMENT OF WATER CONTENT AND BULK DENSITY 915

Commonwealth Fellowship and to Haryana AgriculturalUniversity, Hisar, India, for granting study leave.