ra

Received 27 December 2010Received in revised form 9 February 2012Accepted 17 February 2012Available online 31 March 2012

Keywords:Soil friabilityX-ray CT scanningDrop-shatter testSoil pore characteristics

1. Introduction b50 m (Elliot and Heck, 2007; Peth et al., 2010; Quinton et al.,

Geoderma 181182 (2012) 2229

Contents lists available at SciVerse ScienceDirect

Geode

j ourna l homepage: www.e lseX-ray computed tomography (CT) scanning technology has, inrecent decades, been shown to be a very powerful technique to vi-sualize and quantify soil structure, as reviewed by Taina et al.(2008). The application was, in the early years after adoption insoil science, concentrated on describing macro-structures such asbiopores (Capowiez et al., 2003; Gregory et al., 2003), tree roots(Pierret et al., 1999) and dense layers (Lipiec and Hatano, 2003).With the development of high-resolution micro-CT scanning overthe last decade, the technique has also been applied to other as-

2009; Schluter et al., 2011).Soil pore characteristics are important for a large range of essential

soil functions such as colloid, water and gas transport, habitat for soilorganisms as well as soil mechanical properties such as soil friability.The latter is a key soil physical property, yielding valuable informa-tion on the ease of producing a favorable seed- and rooting bed dur-ing tillage operations (Munkholm, 2011). Soil friability is related tobrittle fracture of soil as described by Braunack et al. (1979) andDexter and Watts (2000). Brittle fracture results from the progressivedevelopment of fracture planes, resulting in a crack opening and apects of soil micromorphology. It has especitify soil pore space characteristics on ima

Abbreviations: BD, Bulk density; CT, computed tomoT, Total porosity; CT, Air-lled porosity derived from CTdiameter; ROI, Region of interest; SACT, Surface area of CSADS, Surface area generated by drop shatter test. Corresponding author. Tel.: +45 87157727.

E-mail address: lars.munkholm@agrsci.dk (L.J. Munk

0016-7061/$ see front matter 2012 Elsevier B.V. Alldoi:10.1016/j.geoderma.2012.02.024X-ray computed tomography (CT) scanning technology has, in recent decades, been shown to be a very pow-erful technique to visualize and quantify soil structure. The objective of this project was to quantify soil porecharacteristics, on undisturbed eld moist soil, using a high resolution X-ray CT scanner and link then theseresults to soil friability assessed using the drop shatter method. Minimally disturbed soil cores were takenfrom selected treatments in a long-term rotation and tillage treatment experiment located on a silt loam atthe Elora Research Station near Elora, Ontario, Canada. Soil cores varied in porosity and pore characteristics.A drop shatter test was used as a reference procedure to quantify soil friability. The top 40 mm of the 80 mmhigh soil samples were scanned using a X-ray micro-CT scanner. The selected region of interest(363636 mm) was reconstructed with a voxel size of 60 m. Estimated surface area, produced from thedrop-shatter test, varied between 0.2 and 1.62 m2 kg1, and an average of 0.79 m2 kg1. Total and air-lled porosity was determined on the soil cores using traditional methods. Total porosity ranged from 41to 60 m3 100 m3, and an average of 49 m3 100 m3. The air-lled porosity, at sampling/testing, ranged be-tween 5 and 32 m3 100 m3, with an average of 15 m3 100 m3. The porosity determined from CT imageryranged between 1 and 31 m3 100 m3, with an average of 4.5 m3 100 m3. The number of branches, junc-tions and end points averaged 298, 117 and 198 per cm3, respectively. We found signicant and strong cor-relations between the soil pore characteristics assessed on the whole soil cores and the characteristics of theair-lled pores determined using high-resolution X-ray computer tomography (CT). Our study conrmed asignicant correlation between soil friability, expressed by surface area produced by standardized drop-shatter, and soil pore characteristics. The strongest correlations were found with porosity, surface area andnumber of junctions per cm3.

2012 Elsevier B.V. All rights reserved.Article history:a b s t r a c ta r t i c l e i n f oSoil pore characteristics assessed from X-and correlations to soil friability

Lars J. Munkholm a,b,, Richard J. Heck b, Bill Deen c

a Aarhus University, Department of Agroecology, P.O. Box 50, DK-8830, Tjele, Denmarkb University of Guelph, School of Environmental Sciences, Guelph, ON, Canada N1G 2W1c University of Guelph, Department of Plant Agriculture, Guelph, ON, Canada N1G 2W1ally been used to quan-ges with voxel size of

graphy; a, Air-lled porosity;imagery; MWD, Mean weightT derived air-lled pore space;

holm).

rights reserved.y micro-CT derived images

rma

v ie r .com/ locate /geodermasudden loss in strength (Hatibu and Hettiaratchi, 1993). The propaga-tion of cracks in an unconned stressed soil depends on the frequencyand the morphology (connectivity, orientation) of the air-lled poresas well as the strength at the crack tips as stressed by Hallett et al.(1995a,b). A strong link between soil fragmentation/soil friabilityand air-lled soil pore space was conrmed by Munkholm et al.(2002), and this suggest a strong correlation between soil friabilityand pore characteristics derived from 3D images of soil structure.

Woolwich silt loam and classied as a Grey Brown Luvisol (CSSC,

130. The samples were scanned at 120 kV, 170 mA and 3500 ms ex-

Table 1Laboratory results from soil core measurements.

Standard soil characteristics Drop shatter

BD T a Water content MWD SADS

g cm3 m3 100 m3 m3 100 m3 m3 100 m3 g 100 g1 mm m2 kg1 cm2 cm3

Mean 1.36 48.6 15.3 33.3 24.5 15.3 0.79 10.4St. dev. 0.13 5.1 7.0 3.5 3.0 5.6 0.34 3.7High 1.57 59.5 32.2 38.1 30.2 28.7 1.62 17.4Low 1.07 40.7 5.2 25.5 17.9 8.0 0.20 3.1

23L.J. Munkholm et al. / Geoderma 181182 (2012) 22291998) or Albic Luvisol (WRB, 2006). The particulate size distributionis on average: 16, 44, 40 and 2.13 g 100 g1 of clay, silt, sand and or-ganic carbon, respectively. Thirty-year average rainfall (19702000)is reported at 920 mm and it is distributed relatively uniformly overthe twelve months. The average monthly temperatures for January,April and July are 7.6, 5.9 and 19.7 C, respectively(Meteorological Services Canada, 2010).

The experimental design is a randomized block split plot with fourreplicates. The main plot treatment is rotation and the plot treatmentis tillage. Seven 4-course rotations are included in the trial. Sampleswere taken from Rotation 1 (R1), designated CCCC, which is con-tinuous corn (Zea mays L.) and Rotation 6 (R6), designated CCO(RC)B(RC), which is corn, corn, oats (Avena fatua L.) and springbarley (Hordeum vulgare L.). Red clover (Trifolium pretense L.), desig-nated (RC), was used as an under seeded cover crop in oats and springThe objective of this project was to quantify soil pore characteris-tics on undisturbed eld moist soil using a high-resolution X-ray CTscanner and then link these results to pore characteristics assessedusing traditional methods and soil friability assessed using the dropshatter method. We hypothesized 1. Signicant and strong correla-tions between soil pore characteristics assessed using traditionaland X-ray CT methods and 2. Signicant and strong correlations be-tween soil pore characteristics of especially the air-lled macroporesand soil friability. Samples were taken from selected treatments in along-term rotation and tillage experiment. The experiment allowedus to test the hypotheses for a given soil type, on samples that variedin soil pore characteristics and friability. In this paper we do not focuson treatment effects per se; they are reported by Munkholm et al.(accepted for publication) together with other data from theexperiment.

2. Materials and methods

2.1. Soil sampling

Minimally disturbed soil cores (=6.4 cm, height=8.0 cm) weretaken from the long-term rotation and tillage trial (initiated in 1980)located at the University of Guelph' s Elora Research Station nearElora, Ontario, Canada (4339 N, 8025W). The soil is mapped as aTable 2Results from image analysis on CT scan images with 60 m voxels.

No. pores CT Porosity largest CT/a

No. cm3 m3 100 m3 % of CT m3 100 m3

Mean 94 4.5a 61 37S.D. 74 (0.37)b 27 25High 277 30.7 99 114Low 15 1.1 22 11

a Geometric means.b S.D. for log10-transformed data.posure generating an axial sequence of X-ray attenuation imagery. Ahigh-pass copper lter was utilized, between the tube and the sam-ple, to minimize the hardening of artifacts and maximize the contrastbetween sample phases. The individual scans took 3.5 h. Softwarefrom GE Healthcare (GE Reconstruction Utility), specically writtenfor the unique instrumental design, was used to reconstruct the16 bit axial image into a 3D image model. The resulting 3D imageryhad a voxel size of 20 m. To reduce uncorrelated noise from the CTimage, a 3D low-pass Gaussian lter was applied in Micro View (GEHealthcare, 2006), as described by Tarquis et al. (2009). For eachsoil sample, a region of interest (ROI) of 36 mm36 mm36 mmwas generated in Micro View. The ROI was located in the center ofthe sample. The le size was large and we were, due to computingconstraints, not able to make subsequent analysis on the entire data-set. We choose to reduce resolution instead of volume as we weremainly interested in the largest air-lled macropores. These poreswe assumed to be of strongest importance in relation to soilbarley. In 2010, rst-year corn was grown in R6. The tillage treat-ments included no tillage (NT) and conventional tillage with mold-board plowing (MP). Moldboard plowing (to 20 cm depth) was lastcarried out on November 18, 2009. Secondary tillage, in associationwith MP, consisted of two passes by a eld cultivator and packerwithin 1 day prior to crop seeding in the spring. The tillage plots are717 m; and 8 rows of corn were sown in each of the studied plotson May 7 2010.

On May 28, 2010, two samples were taken from the 1020 cmdepth in each of the 16 plots included in this study. One sample wastaken between rows 2 and 3 and another between rows 6 and 7.These rows had not been trafcked in relation to crop establishment.The samples were taken in the center of the plots. Immediately aftersampling, the samples were stored in a refrigerator at 5 C until CTscanning as well as between CT scanning and the drop-shatter test.The CT images from two samples were lost and data analyses were,therefore, carried out on in total 30 samples.

2.2. CT scanning

The top 40 mm of the soil samples were scanned using an EVS(now GE Medical, London, Canada) micro-CT scanner, model MS8X-SACT SA largest Branches Junctions End points

cm2 cm3 % of SACT No. cm3 No. cm3 No. cm3

2.3a 56 279a 117a 198a

(0.3)b 26 (0.34)b (0.34)b (0.33)b

12.9 96 2142 1036 9220.6 17 76 33 53

fragmentation. The ROI was reconstructed with a voxel size of 60 mto create a manageable le size.

2.3. Binary thresholding

CT imagery of soil typically contains a large proportion of mixedvoxels, i.e. voxels containingmore than one of the major constituents:air, water and solid; this is due, in part, to the existence of particlesand pores smaller than the voxel size. Therefore, thresholding is of ut-most importance and the selected way of thresholding has tradition-ally been highly user dependent (Baveye et al., 2010). In this paper,wemodied the standardized and automated thresholding proceduredeveloped by Elliot and Heck (2007) and further described by Tarquiset al. (2009). The open source software program ImageJ was used forbinary thresholding and image analysis (Rasband, 2005). Each imageslice had a depth of 60 m and a length and width of 36 mm. A stan-dard deviation (SD) variance plug-in was applied to the stack of im-ages and this created a modied stack of low variance (LV) voxels.The threshold limit was user set at an upper value of 1.10 and alower value of 0.90 standard deviations. Voxels that had varianceabove 10% from the surrounding voxels neighboring 124 voxels(555 unit volume) were deemed too heterogeneous and weregiven a zero value, and the remaining areas of low variance kepttheir original values; the zeros were changed to become Not-A-Number (NAN). A vast majority of the voxels exhibited varianceslarger than 10% and this will obscure the threshold value. By remov-ing these voxels from the calculation, the peaks of the specic phaseswill be solely represented. The multi-thresholder plug-in was used

for thresholding the images, specically employing the threshold al-gorithm created by Huang and Wang (1995). The resulting thresholdvalue was then applied when thresholding the original stack of CTimages.

2.4. Image analysis

General pore characteristics in the 3D images were generated foreach ROI using the ImageJ plug-in 3D object counter plug-in (Bolteand Cordelires, 2006). This plug-in were used to count the numberof unconnected pores and determine characteristic such as volumeand surface area for the individual pores. The algorithm used thenearest neighbor approach to separate objects. Based on the individ-ual pore results, we calculated the total pore volume and surface area.The In this study, a minimum pore volume lter of 3 voxels was usedin 3D object counter, this is to reduce the possibility of interpretingnoise as an indication of a pore. A skeleton reconstruction was alsoperformed for each sample to determine pore branching of thepores, average length of the pores and maximum pore length. Inthis study the plug-in provided by BoneJ (Doube et al., 2010) isused to achieve these results. Firstly, the binary stack undergoes a3D thinning algorithm in the plug-in skeletonize 3D. The thinning al-gorithm utilizes the 3D erosion function in ImageJ, which erodes thepores surface until only the skeleton remains. Erosion was performedsymmetrically in order to guarantee medial position of the skeletonlines and such that the connectedness of the object is preserved(Lee et al., 1994). This skeletonized binary stack is then put throughthe analyze skeleton algorithm within the BoneJ plug-in, which

CT=3.8 m3 100m-3; SADS=0.82 m2 kg-1 CT=2.8 m3 100m-3; SADS=0.43 m2 kg-1

24 L.J. Munkholm et al. / Geoderma 181182 (2012) 2229CT=3.0 m3 100m-3; SADS=0.96 m2 kg-1Fig. 1. Visualization of pore structure from the X-ray CT imagery oCT=2.1 m3 100m-3; SADS=0.56 m2 kg-1f four different samples (598598560 voxels total volume).

provides a summary of number, length (mean, maximum and total)and junction of branches in each ROI. Visualization of the 3D imageswas carried out using the binary stack the plug-in 3D image viewerin ImageJ. This plug-in generates a 3D representation of the stackand allows for visual assessment of the ROI.

2.5. Drop shatter test

Soil fragmentation behavior was evaluated in the eld by adrop shatter test modied from the methods described by Hadasand Wolf (1984) and Schjnning et al. (2002). Once the CT imag-ing was completed, the undisturbed cores were gently pressed outof the tubes and then dropped from 2.0 m depth onto a concreteoor. The soil was subsequently air-dried and passed through anest of sieves with the openings of 19, 9.2, 4, 2 and 1 mm. Theamount of material in each size class was recorded. Based on theresults, the mean weight diameter (MWD) and the approximatespecic surface area (m2 kg1) of the samples were calculatedaccording to Hadas and Wolf (1983). These parameters havebeen used as indices of soil friability as reviewed by Munkholm(2011).

2.6. Standard soil core measurements

All soil samples were weighed before CT scanning and drop-shatter test, (eld moist) condition, in air-dry condition after drop-shatter and in oven dry condition (105 C, 24 h) after sieving. Thisallowed us to calculate total porosity and air-lled porosity at sam-pling as well as bulk density when assuming a particle density of2.65 g cm3.

2.7. Statistical analysis

All statistical analyses were carried out using SAS (Version 9.2, SASInstitute, Cary, NC) (SAS Institute, 2005). We used PROC INSIGHT totest data for normality. The pore characteristics derived from CT im-agery were log-transformed to yield normality. We used PROC GLMto test interaction between soil friability and pore characteristics.There was no signicant effect of management observed on the rela-tionships between soil friability and pore characteristics.

3. Results and discussion

3.1. Standard soil characteristics

The rotation and tillage experiment allowed us to test the hypothe-ses on samples with a wide range in basic soil pore characteristics for agiven soil type. Total porosity ranged from 41 to 60 m3 100 m3 withan average of 49 m3 100 m3 (Table 1). These values are in correspon-dence with previous measurements from the experiment (e.g. Muelleret al., 2009). The air-lled porosity at sampling/testing displayed aneven larger relative variation, i.e. from 5 to 32 m3 100 m3. Thewater content at sampling was, on average, 24 g 100 g1. This indi-cates a matric potential around 333 hPa, at sampling, according towater retention data from the Elora site (Parkin, unpublished data).Thus the soil could be considered within the range for optimum friabil-ity (300 to 1000 hPa) according to Utomo and Dexter (1981) aswell as Munkholm and Kay (2002).

3.2. Drop shatter test

The average mean weight diameter (MWD) was 15.3 mm, but itranged from 8.0 to 28.7 mm (Table 1). The surface area produced

25L.J. Munkholm et al. / Geoderma 181182 (2012) 2229Fig. 2. The 3D skeleton reconstruction for the four samples shown in Fig. 2 (598598560 voxels total volume).

26 L.J. Munkholm et al. / Geoderma 181182 (2012) 2229from the drop-shatter test varied between 0.2 and 1.62 m2 kg1,with an average of 0.79 m2 kg1. This is comparable to the valuesreported for a drop shatter test on a sandy loam reported byMunkholm et al. (2002). The lowest degree of fragmentation wasfound for high density samples. The surface area produced decreasedsignicantly and linearly with increasing BD (R2=0.62).

3.3. CT scanning data

The average number of pores for the ROIs was 89 per cm3. Howev-er, only a few inter-aggregate pores dominated the pore volumedetected from the CT imagery. The pore volume of the largest poreaccounted for between 11 and 99% of the total CT volume. The porevolume, CT, displayed a geometric mean of 4.6 m3 100 m3 and ran-ged between 0.9 and 30.7 m3 100 m3 (Table 2). Examples of thepore network for four samples are shown in Fig. 1.

Essentially, the image analysis separated air-lled volume fromsolid and water and, therefore, the CT values can be comparedwith the air-lled pore space value (a) from the standard coremeasurements. On average we were able to indentify 37% of theair-lled pore space in the CT imagery. The relatively large differ-ence between a and CT is not surprising, as the CT image voxelsize (60 m) was larger than the expected minimum size of poresdrained at the specied water potential, i.e. pores down to

Fig. 3. Correlations between soil pore characteristics. T: total porosity a: airl=10 m will be emptied at 333 hPa when using the La Placeequation. Thus, even when using the high-resolution micro-CTtechnique applied in this study, we were only able to detect air-lled macropores. In one case, CT was larger than a. This is likelydue to the fact that ROI accounted for only about 25% of the totalsample volume. The surface area showed a geometric mean of2.3 cm cm3, with values ranging between 0.6 and 12.9 cm2. Theresults from the skeletonized procedure further showed a geomet-ric mean number (per cm3) of 279, 117 and 198, respectively forbranches, junctions and end points. The range between lowestand largest was large, e.g. 76 to 2142 for the number of branchesper cm3. Examples of skeletonized pore network for four samplesare shown in Fig. 2. Skeletonization of 3D soil pore network haspreviously been shown to provide valuable information on earth-worm burrows network (Capowiez et al., 1998, 2003) and porenetwork in aggregates (Peth et al., 2008).

3.4. Correlation between soil pore characteristics

Signicant linear and positive correlations were found betweenmost of the different soil pore characteristics (Fig. 3). Strongest corre-lations were found between parameters assessed at similar level ofobservation. That is, the correlations were high between parametersdetermined on the whole samples, i.e. T and a, (R2=0.78), the

led porosityCT: airlled porosity CT images, SACT: surface area CT images.

Fig. 4. Relationship between pore characteristics on whole soil cores and friability assessed by drop shatter. *** indicates signicant difference at Pb0.001 level.

27L.J. Munkholm et al. / Geoderma 181182 (2012) 2229parameters assessed from the X-ray CT imagery, i.e. log CT and logSACT, (R2=0.82) and the parameters assessed from the skeletonizedimages (R2>0.72 for any of the combinations of log branches, logjunctions and log end points). Total and air-lled porosity was ratherstrongly correlated to porosity and surface area assessed from the X-ray CT imagery, (R2=0.510.64) but weakly correlated to the param-eters measured on the skeletonized images (R2b0.33). Porosity andsurface area assessed from the X-ray CT imagery, also correlated rath-er strongly to parameters assessed from the skeletonized images, es-pecially with log branches (R2=0.440.60) and log junctions(R2=0.520.68). Log SACT correlated typically better to these proper-ties than log CT.

3.5. Correlation between soil friability and pore characteristics

In general, the surface area parameter, SADS calculated from thedrop-shatter test, was best correlated to the pore characteristicsdata. Therefore, for simplicity, we only show correlations includingSADS in the following. There was a signicant positive correlation0.1 1 100.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

0.1

a

Surfa

ce a

rea,

m 2

kg-1

Porosity,, m3 100m-3

SA DS

= 0.

39 +

0.62lo

g 10

(Por CT

) R2 =0

.40**

Fig. 5. Relationship between pore characteristics from the X-ray CT imagery and friability aslevels, respectively.between the surface area produced by the drop shatter test andtotal porosity, T, and air-lled pore space, a (Fig. 4). Surprisingly,the correlation to total porosity was stronger than to a, i.e.R2=0.66 compared with 0.52. We had expected a stronger corre-lation to a, as the air-lled pores are expected to play a key rolein brittle fracture as described earlier. However, fracture not onlydepends on the abundance and characteristics of the air-lledpore space but also on the strength at the crack tips. Total porosityyields information on the average density and thus the packing ofsoil elements. Thus, T may to a larger extent than air-lled poros-ity, reect both key issues of brittle fracture (i.e. stress concentra-tion at crack tips and strength at crack tips).

The pore characteristics from the CT imagery displayed positiveand signicant correlations to soil friability. The correlation to thepore volume, specically log (CT), was weaker than to a (R2=0.40and 0.52, respectively) (Fig. 5b and a). It is possible that we mayhave been able to improve the t to the drop shatter data if we hadbeen able to quantify a larger proportion of the air-lled pore space.Remember, that pores with diameter in the range between 10 and1 10 1000.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0b

Surfa

ce a

rea,

m 2

kg-1

Surface area,CT, m2 cm-3

SA DS

= 0.

58 +

0.67*

log 10

(SA CT

) R2 =0

.40**

sessed by drop shatter. *** and ** indicate signicant difference at Pb0.001 and Pb0.01

100ctioroce

28 L.J. Munkholm et al. / Geoderma 181182 (2012) 222960 mwere not expected to be included in CT and, therefore, we onlydetected in average 37% of a with the described scanning and imageanalysis procedure. The correlation to surface area of the pore space,SACT was at the same level as for the pore volume, CT.

Signicant correlations were found between SADS and the log tothe number of branches, junctions and end points per cm3

(R2=0.35, 0.44, and 0.15, respectively) (Fig. 6a, b, c). This meansthat number of branches and junctions explained variation in SADSas good as log (CT) and log (SACT). Good correlation between frag-mentation and density of branches and junctions corresponded withwhat we had expected. A high density of branches and junctions sig-nies an extensive, well-connected and complex pore network,which implies a high probability of crack propagation and interaction.Given a random crack orientation, there will be a high chance of crackpropagation irrespective of the orientation of the applied force. Thehigh density of branches means short distances between individualbranches and thus a high chance of crack interact according toHallett et al. (1995b). There was a tendency to improved t by com-bining log (CT) and log (No. of junctions per cm3) (R2=0.48).We also found signicantly improved t when combining air-lledpore space (a) with log (No. of junctions per cm3) (R2=0.62).This indicates that the morphology of the air-lled pore space playsignicant role in relation to soil friability.

3.6. Potential for improved tting to drop shatter data?

The pore characteristics derived for high resolution CT scan-ning explained, at best, 48% of the variation in friability as

10 100 10000.00.20.40.60.81.01.21.41.61.82.0

1 10# jun# branches cm-3

a bSu

rface

are

a, m

2 kg

-1

Fig. 6. Relationship between pore characteristics from X-ray CT imagery skeletonized pat Pb0.001, Pb0.01 and Pb0.05 levels, respectively.expressed by SADS. Better correlation may have been achieved ifwe had scanned the whole 80 mm long sample and not just thetop 40 mm. However, a much improved correlation cannot beexpected from increasing the volume scanned. The selected(36*36*36 mm=46,656 mm

3 or 47 cm3) was around 10 timeslarger than the representative volume (REV) value of 35 cm3

for porosity and specic surface area estimated by Elliot et al.(2010) in another X-ray microCT imagery study. Improved t todrop shatter data may also have been achieved by obtainingmore detailed information on the air-lled pore space, i.e. nerspatial resolution and more sophisticated information on poremorphology. Information on the density and variation in densityof the solid material may also be useful in improving t to dropshatter data. This may add valuable information on the strengthof the soil at the crack tips. General indices reecting connectivityand entropy such as Eulers number (Vogel and Roth, 2001) andfractal dimension (Tarquis et al., 2008, 2009) derived from CT im-agery may also be helpful in improving ts to experimental fria-bility data.4. Conclusions

We found signicant and rather strong correlations between thesoil pore characteristics assessed on the whole soil cores and thecharacteristics of the air-lled pores determined using high-resolution X-ray computer tomography (CT). Our study revealed asignicant correlation between soil friability, expressed by surfacearea produced by standardized drop-shatter, and soil pore character-istics. The strongest correlations were found with pore volume, sur-face area and number of junctions per cm3. At best, the CT porecharacteristics explained 48% of the variation which was at thesame level as for bulk soil air-lled porosity but poorer than fortotal porosity (R2=0.66). There is a need for further studies to im-prove the correlations between CT soil characteristics and soil friabil-ity. This study only included a soil from one soil type and at a singlewater content. More studies are needed to validate our ndings forother soil types and for a range of water contents.

Acknowledgments

This work was supported by the OECD Co-operative Research Pro-gramme Fellowship, OECD Trade and Agriculture (TAD/PROG) andthe Danish Ministry for Food, Agriculture and Fisheries through theSustainable Plant Production Systems to Mitigate Global Warming(CoolCrop) Project. Funding for the acquisition of X-ray CT scannerwas provided to Heck and Deen from the Canadian Foundation for In-novation and the Ontario Innovation Trust. The help received from

1000# end points cm-3ns cm-3

1 10 100 10000.00.20.40.60.81.01.21.41.61.82.0

c

Surfa

ce a

rea,

m 2

kg-1

dure and friability assessed by drop shatter. ***, ** and * indicate signicant differenceTatiana Rittl during sampling and CT scanning is gratefullyacknowledged.

References

Baveye, P.C., Laba, M., Otten, W., Bouckaert, L., Dello Sterpaio, P., Goswami, R.R., Grinev,D., Houston, A., Hu, Y., Liu, J., Mooney, S., Pajor, R., Sleutel, S., Tarquis, A., Wang, W.,Wei, Q., Sezgin, M., 2010. Observer-dependent variability of the thresholding stepin the quantitative analysis of soil images and X-ray microtomography data. Geo-derma 157, 5163.

Bolte, S., Cordelires, F.P., 2006. A guided tour into subcellular colocalization analysis inlight microscopy. Journal of Microscopy 224, 213232.

Braunack, M.V., Hewitt, J.S., Dexter, A.R., 1979. Brittle fracture of soil aggregates and thecompaction of aggregate beds. Journal of Soil Science 30, 653667.

Capowiez, Y., Pierret, A., Daniel, O., Monestiez, P., 1998. 3D skeleton reconstruction ofnatural earthworm burrow systems using CAT scan images of soil cores. Biologyand Fertility of Soils 27, 4159.

Capowiez, Y., Pierret, A., Moran, C., 2003. Characterisation of the three-dimensionalstructure of earthworm burrow systems using image analysis and mathematicalmorphology. Biology and Fertility of Soils 38, 301310.

CSSC, 1998. The Canadian System of Soil Classication, third ed.Dexter, A.R., Watts, C.W., 2000. Tensile strength and friability. In: Smith, K.A., Mullins,

C.E. (Eds.), Soil and Environmental Analysis Physical Methods. Marcel Dekker Inc.,New York, pp. 405433.

Doube, M., Kosowski, M.M., Arganda-Carreras, I., Cordelires, F., Dougherty, R.P.,Jackson, J., Schmid, B., Hutchinson, J.R., Shefelbine, S.J., 2010. BoneJ: free and exten-sible bone image analysis in Image. Journal of Bone 47, 10761079.

Elliot, T.R., Heck, R.J., 2007. A comparison of optical and X-ray CT technique for voidanalysis in soil thin section. Geoderma 141, 6070.

Elliot, T.R., Reynolds, W.D., Heck, R.J., 2010. Use of existing pore models and X-ray com-puted tomography to predict saturated soil hydraulic conductivity. Geoderma 156,133142.

GE Healthcare, 2006. MicroView 2.1.2 MicroCT Visualization and Analysis. London,Canada.

Gregory, P.J., Hutchison, D.J., Read, D.B., Jenneson, P.M., Gilboy, W.B., Morton, E.J., 2003.Non-invasive imaging of roots with high resolution X-ray micro-tomography.Plant and Soil 255, 351359.

Hadas, A., Wolf, D., 1983. Energy efciency in tilling dry clod-forming soils. Soil andTillage Research 3, 4759.

Hadas, A., Wolf, D., 1984. Renement and re-evaluation of the drop-shatter soil frag-mentation method. Soil and Tillage Research 4, 237249.

Hallett, P.D., Dexter, A.R., Seville, J.P.K., 1995a. Identication of pre-existing cracks onsoil fracture surfaces using dye. Soil and Tillage Research 33, 163184.

Hallett, P.D., Dexter, A.R., Seville, J.P.K., 1995b. The application of fracture mechanics tocrack propagation in dry soil. European Journal of Soil Science 46, 591599.

Hatibu, N., Hettiaratchi, D.R.P., 1993. The transition from ductile ow to brittle failurein unsaturated soils. Journal of Agricultural Engineering Research 54, 319328.

Huang, L.K., Wang, M.J.J., 1995. Image thresholding by maximizing the index of nonfuz-ziness of the 2-D grayscale histogram. Pattern Recognition 28, 4151.

Lee, T.C., Kashyap, R.L., Chu, C.N., 1994. Building skeleton models via 3-D medial surfaceaxis thinning algorithms. CVGIP: Graphical Models and Image Processing 56,462478.

Lipiec, J., Hatano, R., 2003. Quantication of compaction effects on soil physical proper-ties and crop growth. Geoderma 116, 107136.

Meteorological Services Canada, 2010. Canadian Climate Normals: 19712000. Avail-able at Environment Canada, Ottawa, ONwww.climate.weatherofce.ec.gc.ca/climate_normals. (accessed Dec. 2010).

Mueller, L., Kay, B.D., Hu, C., Li, Y., Schindler, U., Behrendt, A., Shepherd, T.G., Ball, B.C.,2009. Visual assessment of soil structure: evaluation of methodologies on sites inCanada, China and Germany: Part I: comparing visual methods and linking themwith soil physical data and grain yield of cereals. Soil and Tillage Research 103,178187.

Munkholm, L.J., 2011. Soil friability: a review of the concept, assessment and effects ofsoil properties and management. Geoderma 167168, 236246.

Munkholm, L.J., Kay, B.D., 2002. Effect of water regime on aggregate-tensile strength,rupture energy, and friability. Soil Science Society of America Journal 66, 702709.

Munkholm, L.J., Schjnning, P., Kay, B.D., 2002. Tensile strength of soil cores in relationsto aggregate strength, soil fragmentation and pore characteristics. Soil and TillageResearch 64, 125135.

Munkholm, L.J., Heck, R. and Deen, B. accepted for publication. Long-term effects of ro-tation and tillage on visual evaluation of soil structure, soil physical properties andcrop yield. Soil and Tillage Research.

Peth, S., Horn, R., Beckmann, F., Donath, T., Fischer, J., Smucker, A.J.M., 2008. Three-dimensional quantication of intra-aggregate pore-space features usingsynchrotron-radiation-based microtomography. Soil Science Society of AmericaJournal 72, 897907.

Peth, S., Nellesen, J., Fischer, G., Horn, R., 2010. Non-invasive 3D analysis of local soil de-formation under mechanical and hydraulic stresses by CT and digital image corre-lation. Soil and Tillage Research 111, 318.

Pierret, A., Capowiez, Y., Moran, C.J., Kretzschmar, A., 1999. X-ray computed tomogra-phy to quantify tree rooting spatial distributions. Geoderma 90, 307326.

Quinton, W.L., Elliot, T., Price, J.S., Rezanezhad, F., Heck, R., 2009. Measuring physicaland hydraulic properties of peat from X-ray tomography. Geoderma 153, 269277.

Rasband, W., 2005. ImageJ. National Institute of Health, Bethesda MD.SAS Institute, 2005. SAS Institute, SAS/STAT Software: Language Reference and Con-

cepts Release 9.1.3. SAS Institute, Cary, NC.Schjnning, P., Elmholt, S., Munkholm, L.J., Debosz, K., 2002. Soil quality aspects of

humid sandy loams as inuenced by organic and conventional long-term manage-ment. Agriculture Ecosystems and Environment 88, 195214.

Schluter, S., Weller, U., Vogel, H.J., 2011. Soil-structure development including seasonaldynamics in a long-term fertilization experiment. Journal of Plant Nutrition andSoil Science 174, 395403.

Taina, I.A., Heck, R.J., Elliot, T.R., 2008. Application of X-ray computed tomography tosoil science: a literature review. Canadian Journal of Soil Science 88, 120.

Tarquis, A.M., Heck, R.J., Grau, J.B., Fabregat, J., Sanchez, M.E., Anton, J.M., 2008. Inu-ence of thresholding in mass and entropy dimension of 3-D soil images. NonlinearProcesses in Geophysics 15, 881891.

Tarquis, A.M., Heck, R.J., Andina, D., Alvarez, A., Anton, J.M., 2009. Pore network com-plexity and thresholding of 3D soil images. Ecological Complexity 6, 230239.

Utomo, W.H., Dexter, A.R., 1981. Soil friability. Journal of Soil Science 32, 203213.Vogel, H.J., Roth, K., 2001. Quantitative morphology and network representation of soil

pore structure. Advances in Water Resources 24, 233242.WRB, 2006. World Reference Base for Soil Resources 2006. Food and Agriculture Orga-

nization of United Nations, Rome.

29L.J. Munkholm et al. / Geoderma 181182 (2012) 2229

Soil pore characteristics assessed from X-ray micro-CT derived images and correlations to soil friability1. Introduction2. Materials and methods2.1. Soil sampling2.2. CT scanning2.3. Binary thresholding2.4. Image analysis2.5. Drop shatter test2.6. Standard soil core measurements2.7. Statistical analysis

3. Results and discussion3.1. Standard soil characteristics3.2. Drop shatter test3.3. CT scanning data3.4. Correlation between soil pore characteristics3.5. Correlation between soil friability and pore characteristics3.6. Potential for improved fitting to drop shatter data?

4. ConclusionsAcknowledgmentsReferences