Canadian Geotechnical Journal

A method to predict desiccation crack depth in a compacted

clayey soil

Journal: Canadian Geotechnical Journal

Manuscript ID cgj-2020-0575.R2

Manuscript Type: Article

Date Submitted by the

15-May-2021
Author:

Complete List of Authors: Hassanikhah, Arash; The University of Oklahoma - Norman Campus, Civil
Engineering and Environmental Science
Miller, Gerald; The University of Oklahoma - Norman Campus, Civil
Engineering and Environmental Science
Cerato, Amy; The University of Oklahoma - Norman Campus, Civil
Dr

Engineering and Environmental Science

Desiccation cracks, Tensile strength, Matric suction, Moisture loss, Crack

Keyword:
depth
aft

Is the invited manuscript for

consideration in a Special Not applicable (regular submission)
Issue? :

© The Author(s) or their Institution(s)

Page 1 of 45 Canadian Geotechnical Journal

1 A method to predict desiccation crack depth in a compacted clayey soil

2
3 Arash Hassanikhah1,*, Ph.D., Former Postdoctoral Research Fellow
4
5 Gerald A. Miller1, Ph.D., Professor
6
7 Amy B. Cerato1, Ph.D., Professor
8
9 1 School
of Civil Engineering and Environmental Science, University of Oklahoma,
10 Norman, OK 73019, USA
11
12 *Corresponding author: [email protected]
13

14 Abstract

15 Estimation of crack depths and spacing due to desiccation of clayey soils is needed to

16 predict changes in mechanical or hydraulic properties in the cracked layer. These

Dr

17 changes affect the infiltration of water, stability, and deformation of a soil mass.
aft

18 Desiccation cracks are associated with increasing suction due to moisture loss

19 accompanied by restrained shrinkage, which results in tensile stresses in near surface

20 soil layers. Cracks propagate when the developed tensile stresses exceed the tensile

21 strength of soil. A simple analytical method is presented to predict crack depths in

22 compacted clayey soil due to changes in matric suction with depth. The model equation

23 is based on the Hookean elastic equation relating incremental strain to incremental

24 stress and incorporates two stress state variables including net normal stress and matric

25 suction. Input to the model equation includes the tensile strength and elastic

26 parameters, and to complete the prediction of crack depth, the suction change profile

27 of interest is needed. The method validity was investigated by comparing predicted

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 2 of 45

28 crack depths to those observed in soil compacted in a bench scale apparatus for studying

29 desiccation cracking. Tensile strength and elastic properties were determined from tests

30 conducted on soil during desiccation under approximate uniaxial conditions. Predicted

31 crack depths were obtained based on changes in suction interpreted from water content

32 sensors at various depths in the soil bed and compared favorably to observed

33 desiccation crack depths.

34 Keywords: Desiccation cracks, Tensile strength, Matric suction, Moisture loss, Crack

35 depth

36 Introduction

37 Desiccation cracks due to soil shrinkage create substantial challenges in many

Dr

38 geotechnical engineering scenarios where infiltration of water into a soil mass is

aft

39 undesirable, such as for landfill liners and caps, earth-dams, embankments and cut

40 slopes, and around foundations and pavements founded on clayey soils. Formation of

41 desiccation cracks diminishes the strength of a soil mass and allows for rapid ingress of

42 water during rain events leading to loss of suction, soil swelling and reduced shear

43 strength. Thus, being able to predict the depth and spacing of desiccation cracking is an

44 important part of understanding and predicting the mechanical and hydraulic behavior

45 of the cracked layer and its impact on soil below the cracked layer. This paper presents

46 results of a study on desiccation cracking in compacted clay. The study involved

47 experimental investigation of crack depths in a bench scale apparatus, and development

48 of an analytical model to predict the crack depths as the suction profile changes.

49 Background: modeling and predicting desiccation crack depth in unsaturated soil

© The Author(s) or their Institution(s)

Page 3 of 45 Canadian Geotechnical Journal

50 Numerous researchers have studied various aspects of desiccation of clayey soils but

51 relatively few have focused on the prediction of crack depths. In recent years our

52 understanding of the mechanisms involved in desiccation cracking has been greatly

53 enhanced by innovative experimental and theoretical approaches (e.g. Cordero et al.

54 2017; Tang et al. 2018; Tang et al. 2019; Tran et al. 2019; Wang et al. 2018; Bordoloi et

55 al. 2020; Cheng et al. 2020; Tang et al. 2021) and there have been significant

56 advancements in sophisticated mathematical modeling of desiccation cracking (e.g.

57 Peron et al. 2013; Vo et al. 2017 and 2018; Costa et al. 2018; Pouya et al. 2019; Tran et

58 al. 2019); however, there has been less effort in developing practical analytical methods

59 for predicting depth of cracks in desiccating clay under field conditions. Relatively simple
Dr

60 analytical models are useful in practice to assess the impact of desiccation cracking in
aft

61 many important geotechnical problems. Some significant work on development of

62 simple analytical models of desiccation crack depth includes that of Lau (1987); Morris

63 et al. (1992); Konrad and Ayad (1997); and Yin et al. (2016). However, the approaches

64 presented in these works have some practical limitations, particularly in regard to their

65 application to compacted clayey soils as discussed below.

66 Lau (1987) presented two analytical approaches for predicting crack depth that

67 were based on the two-stress state variable model for unsaturated soils. One was

68 derived from elastic equilibrium analysis and one from plastic equilibrium analysis. This

69 work set the stage for further research, but the equations have limitations. They are

70 limited to suction profiles that change from an initial value of zero to a linear variation

71 above the water table. In natural soils, the suction profile during desiccation is typically

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 4 of 45

72 highly non-linear and the position of the water table may not be relevant during the

73 time frame over which desiccation cracks develop. The relevance of the water table

74 would strongly depend on its depth below the fill and the intervening geology and post

75 construction weather conditions. For the purpose of analysis, the water content of

76 compacted soils at the end of construction may be assumed to be constant with depth,

77 as is the corresponding matric suction (see Fig. 1). Thus, the initial suction prior to

78 desiccation is not zero as it is with saturated soils. The initial water content may be

79 assumed equal to some average value relative to the laboratory compaction curve and

80 the corresponding matric suction is determined from direct measurement on

81 compaction samples or from a soil-water characteristic curve (SWCC). During a drying

Dr

82 period following construction, the water content will decrease rapidly near the ground
aft

83 surface and slower at greater depth. This results in a highly nonlinear distribution of

84 matric suction with depth, being very high near the ground surface and decreasing with

85 depth (see Fig. 1). Therefore, incorporating nonlinear suction profiles into models for

86 predicting desiccation crack depth is imperative for the analysis of unsaturated soils.

87 Morris et al. (1992) analyzed the mechanics of cracking after reviewing the

88 morphology of cracks in the field. They proposed three different solutions based on the

89 elastic theory, shear strength theory, and linear elastic fracture mechanics (LEFM). To

90 develop all three solutions, they considered that the soil subjected to desiccation

91 cracking is initially saturated and essentially normally consolidated, and that under

92 unsaturated conditions the soil behavior is governed by two stress variables (i.e. matric

93 suction and net normal stress variables). These initial conditions were consistent with

© The Author(s) or their Institution(s)

Page 5 of 45 Canadian Geotechnical Journal

94 the nature of mine tailings they were investigating. They concluded that in addition to

95 soil suction, soil properties such as compression modulus, Poisson’s ratio, shear

96 strength, tensile strength, and specific surface energy are factors that affect desiccation

97 crack development in soil. In addition, of the three solutions they proposed, they found

98 that the methods based on the elastic theory and tension-shear strength theory gave

99 the most favorable predictions of desiccation crack depth compared to their limited

100 empirical observations.

101 Konrad and Ayad (1997) proposed a model to predict the average crack depth

102 and spacing for slurries, consolidated natural soil, and compacted clays undergoing

103 desiccation based on linear elastic fracture mechanics (LEFM) theory and work by
Dr

104 Lachenbruch (1961). The model was demonstrated (Ayad et al. 1997) for a sensitive clay
aft

105 based on comparisons to field observations. Model input includes the soil-water

106 characteristic curve to predict the suction distribution at crack initiation, tensile strength

107 to determine crack initiation, and the fracture toughness to calculate the ultimate crack

108 depth. While LEFM may in some ways better represent the physics of the desiccation

109 problem compared to elastic theory, and provides a basis for predicting crack spacing,

110 it has a major practical limitation for predicting crack depth under field conditions. As

111 emphasized by Konrad and Ayad (1997), the model captures crack initiation and

112 unstable crack propagation when the critical suction is reached. However, and quite

113 importantly, it does not model further deepening of the cracks as drying continues.

114 Building upon the work of Konrad and Ayad (1997), Yin and Vanapalli (2016)

115 presented a framework for modeling desiccation crack depth in unsaturated soil based

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 6 of 45

116 on the two-stress state variable model for unsaturated soil and incorporating LEFM.

117 While this model advances understanding of desiccation cracking physics, it bears the

118 same practical limitation as Konrad and Ayad’s (1997) model; it only applies to initial

119 crack propagation and does not incorporate subsequent deepening of the cracks during

120 continued drying.

121 In the research presented in this paper, a method for predicting desiccation crack

122 depth in compacted clayey soils under continuous drying was developed. The method is

123 based on analytical equations developed from incremental Hookean elastic equations

124 that relate suction change to tensile failure and crack depth. By superimposing a plot of

125 predicted suction change to produce cracking versus crack depth from the model over
Dr

126 an actual suction change profile of interest, the predicted depth of cracking is
aft

127 determined by the point of intersection of the two curves.

128 Proposed analytical model for predicting crack depth

129 A simple analytical model is presented for estimating the depth of desiccation cracking

130 in compacted clay soil. However, the method should be generally applicable to other

131 soils provided the model parameters can be determined. The model was developed by

132 employing the theory of elasticity and incorporates mechanical characteristics of soil,

133 including tensile strength and elastic properties, as well as changes in stress conditions.

134 The model considers the stress changes between the initial conditions immediately after

135 compaction and the final conditions after drying when cracking is imminent, as depicted

136 in Figure 1. Initial conditions given in Figure 1a, represent unsaturated, at-rest

137 conditions in the compacted clay at the end of construction with a uniform suction

© The Author(s) or their Institution(s)

Page 7 of 45 Canadian Geotechnical Journal

138 profile (solid line with (ua-uw)o) shown in Figure 1b. With the application of a drying

139 suction profile (dashed line (ua-uw)f) as shown in Figure 1b, and cracking is imminent,

140 the stress conditions at the crack tip depth are shown in Figure 1c, where the horizontal

141 stress in the principal directions is equal to the tensile strength and the vertical stress is

142 unchanged. The depth of cracking for a given matric suction profile corresponds to the

143 greatest depth where the cracking suction is reached, as indicated in Figure 1b. The

144 proposed method enables the determination of this depth for a given suction profile.

145 The initial soil properties represent the state of the soil immediately after

146 compaction. The change in matric suction corresponding to a particular depth of

147 cracking can be estimated using the basic Hookean elastic equation for the incremental
Dr

148 horizontal strain due to an increment of stress change in unsaturated soil with isotropic
aft

149 elastic properties (Fredlund and Rahardjo 1993):

d(𝜎𝑥 ― 𝑢𝑎) d(𝜎𝑦 ― 𝑢𝑎)𝜈 d(𝜎𝑧 ― 𝑢𝑎)𝜈 d(𝑢𝑎 ― 𝑢𝑤)

150 d𝜀𝑥 = 𝐸 ― 𝐸 ― 𝐸 + 𝐻 (1)

151 where:

152 dx=increment of horizontal strain in the x-direction;

153 d(x-ua)=(x-ua)f-(x-ua)o=increment of net-normal stress in the horizontal x-

154 direction, where subscripts “o” and “f” represent initial and final stresses,

155 respectively;

156 d(y-ua)=(y-ua)f-(y-ua)o=increment of net-normal stress in the horizontal y-

157 direction;

158 d(z-ua)=(z-ua)f-(z-ua)o=increment of net-normal stress in the vertical z-direction;

159 d(ua-uw)=(ua-uw)f-(ua-uw)o=increment of matric suction due to desiccation;

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 8 of 45

160 E=modulus of elasticity with respect to net normal stress;

161 =Poisson’s ratio with respect to net normal stress; and

162 H=modulus of elasticity with respect to matric suction.

163 Equation 1 is applied to two situations to simulate stress conditions: first, in the

164 field, and second, in a laboratory bench scale test described in the next section. For the

165 field situation depicted in Figure 2, uniform stress and material conditions are assumed

166 to extend to infinity in the lateral directions. The stresses in the two principal horizontal

167 directions (x and y) are assumed to start at the same at-rest stress condition and

168 simultaneously reach the tensile strength when the cracking suction is reached during

169 desiccation. Cracks then develop to a depth that depends on the properties of the soil
Dr

170 and increment of suction that occurs. Desiccation patterns are complex and result in
aft

171 different types of crack patterns that can be orthogonal or non-orthogonal (e.g. Miller

172 et al. 1998; Kodikara et al. 2000; Vogel et al. 2005; Lakshmikantha et al. 2009). The type

173 of pattern that develops depends on many factors including the type of constraint

174 imposed on the soil during drying, stress and material anisotropy, soil type and other

175 factors. Observations in the literature presented in the form of photographs at the

176 ground surface reveal a variety of cracking patterns including regular appearance of

177 orthogonal cracks mixed with non-orthogonal cracks (e.g. Morris et al. 1992 and Sun et

178 al. 2009). A simple square orthogonal pattern shown in Figure 2 was used in the

179 development of the model presented in this paper.

180 For the bench scale model, the stresses in the two principal horizontal directions

181 are assumed to start at the same at-rest stress condition, but as indicated in Figure 3,

© The Author(s) or their Institution(s)

Page 9 of 45 Canadian Geotechnical Journal

182 tensile stresses are assumed to develop only in the x-direction, while the horizontal

183 stress in the y-direction goes to zero as the soil shrinks away from the sides of test box.

184 This results in parallel cracks of finite length across the width of the soil test bed.

185 In developing the equations that relate desiccation crack depth to change in

186 matric suction, the following assumptions were made:

187 1) Air pressure is equal to zero and unchanged during desiccation.

188 2) The soil is isotropic with respect to the elastic properties E, H, and .

189 3) The tensile strength (t) and modulus ratio (E/H) determined from

190 laboratory tests under uniaxial conditions are applicable to the field and

191 model conditions. That is, t and E/H are not dependent on the net-normal
Dr

192 stress path leading to failure. However, t is dependent on the compaction

aft

193 moisture content and E/H depends on the suction at the moment of

194 cracking, as established in uniaxial tensile strength tests.

195 4) Poisson’s ratio () is constant during desiccation.

196 5) Initial net normal stresses correspond to at-rest conditions immediately after

197 compaction is complete, with (x-ua)o=(y-ua)o=Ko(z-ua)o and z=zc;

198 where, Ko is the at-rest earth pressure coefficient defined in terms of net

199 normal stresses,  is the soil total unit weight, and zc is the desiccation crack

200 depth.

201 6) Initial suction, (ua-uw)o corresponds to the compaction water content as

202 determined from the soil water characteristic curve.

203 7) The final suction profile is known and is applied instantaneously.

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 10 of 45

204 8) Crack formation is instantaneous and occurs immediately following

205 application of the suction profile and mobilization of tensile strength (t) in

206 the principal horizontal directions; i.e. for the field condition: (x-ua)f=(y-

207 ua)f=t, and for bench model conditions: (x-ua)f=t and (y-ua)f=0.

208 9) In the field, horizontal strains, dx and dy, are zero prior to crack formation,

209 and in the bench scale model only dx is zero prior to crack formation. These

210 assumptions are consistent with constraints imposed against shrinkage in

211 the x- and y-directions in each of the two model cases.

212 10) Vertical stress in the z-direction remains unchanged during desiccation up to

213 the point of crack formation. That is, the change in soil unit weight due to
Dr

214 drying is ignored.

aft

215 Based on the preceding assumptions, the initial and final stress conditions,

216 incremental stress changes, and incremental strains to be used in Equation 1 are

217 summarized in Table 1 for each situation. Substituting the incremental stresses and

218 strains noted in Table 1 into Equation 1 results in the following two equations for the

219 change in suction, d(ua-uw)c, that will produce a desiccation crack depth, zc.

𝐾𝑜𝛾𝑧𝑐(1 ― 𝜈) ― 𝜎𝑡(1 ― 𝜈)
220 Case 1 (field conditions): d(𝑢𝑎 ― 𝑢𝑤)𝑐 = (2)
(𝐸 𝐻 )
𝐾𝑜𝛾𝑧𝑐(1 ― 𝜈) ― 𝜎𝑡
221 Case 2 (bench scale model): d(𝑢𝑎 ― 𝑢𝑤)𝑐 = (3)
(𝐸 𝐻 )
222 Noting that for an elastic material, Ko=/(1-), these equations can also be expressed

223 as follows.

© The Author(s) or their Institution(s)

Page 11 of 45 Canadian Geotechnical Journal

𝛾𝑧𝑐𝜈 ― 𝜎𝑡(1 ― 𝜈)
224 Case 1: d(𝑢𝑎 ― 𝑢𝑤)𝑐 = (4)
(𝐸 𝐻 )
𝛾𝑧𝑐𝜈 ― 𝜎𝑡
225 Case 2: d(𝑢𝑎 ― 𝑢𝑤)𝑐 = (5)
(𝐸 𝐻 )
226 Based on these equations, cracking will initiate at the ground surface, where

227 zc=0, when d(ua-uw)=-t(1-)/(E/H) for the field case and when d(ua-uw)=-t/(E/H)

228 for the bench scale model. To increase zc beyond zero, additional increases in suction

229 are required to overcome the lateral stress due to overburden pressure, which is

230 represented by the term containing zc. Equations 2 thru 5 can be used to predict the

231 depth of desiccation cracks for a given suction profile representing drying conditions

232 provided elastic properties of the soil and tensile strength are determined
Dr

233 experimentally or estimated.

aft

234 Application of the cracking depth model to a hypothetical field situation

235 To demonstrate the use of the desiccation crack depth model, consider a 5-m profile

236 of compacted fill with properties as shown in Figure 4a. These soil properties are

237 similar to those for the bench scale model discussed in the next section. At the end of

238 construction, matric suction is assumed to be constant with depth. As the soil dries

239 due to moisture loss from the ground surface, the change in suction with depth will

240 vary in a manner similar to that shown in Figure 4b. Curve 1 represents the initial

241 condition where suction change is zero and Curve 6 represents the driest condition

242 considered for this example. For convenience, the curves in Figure 4b were generated

243 using a simple logarithmic decay function (Table 2); however, in practice the suction

244 change profile could be predicted via unsaturated seepage analysis using historical

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 12 of 45

245 weather data as input or by empirical methods. Also, field measurements of suction

246 can be used to establish typical suction change profiles or to validate predictions from

247 unsaturated seepage models.

248 The depth of cracking that will occur in the field is assumed to be the greatest

249 depth where actual suction in the field equals the theoretical suction estimated from

250 Equation 2. Thus, once a suction change profile or profiles of interest are established,

251 Equation 2 is used to predict the suction change necessary to produce cracking as a

252 function of depth. The resulting dashed line in Figure 4b, is superimposed on the

253 expected suction change profiles to determine the depth of cracking. The points where

254 the suction change profile curves intersect the line given by Equation 2 define the
Dr

255 depths of cracking as noted by the arrows in Figure 4b. As shown in Figure 4b, no
aft

256 desiccation cracks will develop early in the drying process as represented by Curve No.

257 2, which does not intersect the line for Equation 2. The desiccation crack depths

258 corresponding to Curve Nos. 3, 4, 5 and 6 are 0.23, 0.60, 1.20 and 1.63 m, respectively.

259 Soil parameters used in Equation 2 were determined in the following manner:

260 1) The tensile strength was measured while soil was desiccating under

261 approximate uniaxial conditions using a device developed at the University of

262 Oklahoma (Varsei et al. 2016). Three compacted samples were tested,

263 corresponding to different initial moisture contents, with the highest moisture

264 content corresponding to the initial conditions in Figure 4a. The initial dry

265 density was similar to that in Figure 4a. The resulting plot of tensile strength

266 versus compaction volumetric water content, is shown in Figure 5a. Note that

© The Author(s) or their Institution(s)

Page 13 of 45 Canadian Geotechnical Journal

267 at some water content, lower than the range for the data depicted in Figure 5a,

268 the tensile strength will typically decrease. (e.g. Varsei et al. 2016).

269 Depending on the compaction water content, the tensile strength can

270 vary significantly, which has a significant influence on the predicted crack depth

271 as shown in Figure 6a. It is noted that various physics-based models exist for

272 predicting tensile strength as a function of water content (e.g. Snyder and

273 Miller 1985; Péron et al. 2009; Lu et al. 2009; Tang et al. 2015; Varsei et al.

274 2016; Yin and Vanapalli 2018); however, since actual tensile strength test

275 results were available for the test soil, the simple best fit equation shown in

276 Figure 5a was used to determine the tensile strength based on the initial water
Dr

277 content. For the compacted soil represented by Figure 4a, the tensile strength
aft

278 is -23.3 kPa based on the compaction volumetric water content of 0.308.

279 2) Modulus ratios (E/H) were back calculated from the results of tensile strength

280 tests described above, and are shown plotted against the cracking suction in

281 Figure 5b. For uniaxial failure conditions at the instant immediately prior to

282 crack development, the modulus ratio based on Equation 1 with dx=0, d(x-

283 ua)=t and d(y-ua)=d(z-ua)=0, is given by Equation 6.

𝐸 ― 𝜎𝑡
284 Uniaxial conditions: 𝐻 = d(𝑢𝑎 ― 𝑢𝑤) (6)
𝑐

285 The resulting values, indicated by the circles in Figure 5b, decrease rapidly

286 with increasing cracking suction. The behavior observed indicates the modulus

287 ratio approaches 0.4 as the suction approaches the air entry suction

288 corresponding to saturated conditions, which is close to the theoretical limiting

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 14 of 45

289 value for saturated conditions equal to 1-2 (e.g. Morris et al. 1992). At higher

290 cracking suctions the modulus ratio decreases rapidly to a low value, which is

291 assumed to be limited to 0.05 for the purpose of modeling in this paper. Modulus

292 ratios of 0.15 to 0.2 are typical for clayey soils (Fredlund and Rahardjo 1993), and

293 Lau (1987) reported values in a range of 0.05 to 0.35 for tests on initially

294 saturated slurries.

295 For the field example depicted in Figure 4, a constant E/H of 0.15 was

296 assumed. The influence of E/H on Equation 2 is demonstrated in Figure 6b,

297 which emphasizes the importance of considering the dependency of E/H on

298 suction. For the purpose of determining E/H for application of the model to
Dr

299 bench scale experimental results, the best fit empirical equation in Figure 5b
aft

300 was used. A logarithmic decay function that approaches a limiting value of E/H

301 of 0.05 at high suction and E/H of approximately 0.4 to 0.5 around the air-entry

302 suction, was found to provide the best fit to the test data. The upper and lower

303 limits of this equation corresponding to the air entry suction and high suction

304 values, respectively, are consistent with theoretical considerations and values

305 reported in the literature mentioned above.

306 3) Poisson’s ratio was estimated from =Ko/(1+Ko) and the lateral stress ratio was

307 estimated from Ko=1-sin. To account for the influence of compaction on lateral

308 stresses, one could rationally select a higher Ko than given by this expression.

309 Also, as noted previously, a constant value was assumed and the dependence of

310 Poisson’s ratio on matric suction was not considered. However, the solution

© The Author(s) or their Institution(s)

Page 15 of 45 Canadian Geotechnical Journal

311 method is less sensitive to Poisson’s ratio compared to the tensile strength and

312 modulus ratio as shown in Figure 6c.

313 Observed cracking and suction in bench scale experiments

314 To explore the validity of the analytical equations presented above, a bench scale

315 experiment was designed and conducted to study crack propagation in a compacted clay

316 during desiccation. The experiment was designed to simultaneously monitor changes of

317 suction and crack depth development. As shown in Figure 7, a box with 570 x 100 x 895

318 mm length, width, and height, respectively, was fabricated to contain the soil bed.

319 Plexiglas acrylic was used for the front and back walls of the box to enable observation

320 of the desiccation crack propagation in the soil. The sidewalls of the box were made of
Dr

321 wood.
aft

322 The test soil for this experiment was a lean clay with soil properties summarized

323 in Figure 4a. It was compacted at 95% maximum dry unit weight based on standard

324 Proctor effort, and 2% wet of optimum moisture content (wopt+2%). To avoid moisture

325 loss during compaction, the soil bed was prepared in a moisture room. The soil was

326 compacted in roughly 11 equal lifts using volume-based compaction to achieve a final

327 soil bed thickness of about 0.9 m. Five volumetric water content sensors were installed

328 at depths of 0.04, 0.22, 0.42, 0.59 and 0.84 m during preparation of the soil bed. The

329 lead wires exited the sides of the box and were connected to a data logger and laptop

330 to record moisture changes during drying. Following completion of the soil bed, rows of

331 five screws (120 mm in length) were installed horizontally into the soil at regular

332 vertical intervals along each of the narrow sides of the box. The purpose was to restrain

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 16 of 45

333 the shrinkage of the soil along the length of the box so that the tensile stresses would

334 develop in the x-direction indicated in Figure 7.

335 After the final soil layer was compacted, the test bed was moved to a laboratory

336 with ambient humidity conditions, unsealed, and allowed dry from the top as crack

337 depth and moisture sensor monitoring was conducted. A fan was positioned roughly a

338 m from the model to move the air across the top of the soil bed and increase the rate of

339 drying. The experiment was performed twice with the same soil to investigate the

340 repeatability. Photographs of the top and sides of the soil bed in Figure 8 show the

341 desiccation cracks that developed over 120 days for Model Test 1 and 90 days for Model

342 Test 2. Due to time constraints related to personnel changes, Model Test 2 had to be
Dr

343 terminated after 90 days. Note, cracks in Figure 8 have been traced on the acrylic with
aft

344 a fine black marker to enhance the side view visibility. The top views in Figure 8 indicate

345 that cracks developed transverse to the x-direction consistent with the assumption that

346 the greatest tensile stress would act in the x-direction. Further, a small gap can be seen

347 between the soil bed and the longer box walls, consistent with the assumption that the

348 final horizontal stress in the transverse y-direction was zero. The side view indicates

349 that four large cracks propagated to depths of 0.3 to 0.43 m in Model Test 1 and three

350 large cracks in Model Test 2 propagated to depths of 0.2 to 0.3 m. The cracks are not

351 linear and during development seem to have been influenced by inhomogeneity’s and

352 layer boundaries in the compacted soil bed. Note also, that the appearance of primary

353 cracks occurred over a relatively short period of a few days within the first five days of

© The Author(s) or their Institution(s)

Page 17 of 45 Canadian Geotechnical Journal

354 drying, and propagation to greater depths occurred roughly in parallel, although the

355 final lengths of cracks were not the same.

356 Figure 9 depicts the observed maximum crack depth versus elapsed time and the

357 change in volumetric water content versus elapsed time corresponding to each sensor

358 depth for Model Test 1 and Model Test 2. Note, the volumetric water content is

359 expressed in decimal form. As indicated by the sensor at 0.04 m, the near surface soil

360 dried relatively fast for roughly the first 20 to 50 days in both model tests, and then

361 exhibited more gradual drying thereafter. Sensors below a depth of 0.04 m showed a

362 much more gradual response and much less moisture loss overall. The lowest sensors at

363 depths of 0.59 m and 0.84 m exhibited negligible moisture loss over the entire
Dr

364 experiment for both model tests. The crack propagation was relatively quick at first,
aft

365 more so for Model Test 1, and appeared to slow somewhat out to about 70 days in both

366 model tests. However, after 70 days the rate of crack growth increased, which may be

367 due to the increased surface area exposed to drying as the crack development

368 increased. Comparing the volumetric water contents at similar elapsed time, for

369 example 80 days, it is apparent that soil in Model Test 1 dried out more than in Model

370 Test 2 for the same period. This is likely due to variations in the ambient humidity and

371 temperature in the laboratory; no attempt was made to control these conditions

372 beyond what was provided by the HVAC system in the building.

373 Suction profiles at various times were developed using volumetric water

374 content determined from the moisture sensors and the soil water characteristic curve

375 (SWCC) shown in Figure 10. To generate the SWCC, samples of the compacted clay were

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 18 of 45

376 allowed to dry while monitoring the moisture content change. Corresponding total

377 suction was determined using a Decagon WP4 Potentiameter, which is a chilled mirror

378 hygrometer device. The osmotic suction corresponding to each moisture content was

379 estimated (Wei and Miller 2019) and subtracted from the total suction to give the values

380 of matric suction plotted in Figure 10 (solid circles). The method of Wei and Miller (2019)

381 involves measuring total suction and water content of a sample, then mixing the sample

382 with water to saturate it and taking another total suction measurement and water

383 content. The total suction of the saturated specimen represents the osmotic suction and

384 can be related to the osmotic suction in the original sample using phase relationships.

385 The calculations depend on the type of salt assumed to be present in the pore water.
Dr

386 More details can be found in the paper by Wei and Miller (2019). The best fit curve was
aft

387 generated using the Fredlund and Xing (1994) equation. An empirical equation (Zapata

388 et al. 2000) based on the PI and percent of fines was used to generate an SWCC for

389 comparison. Superimposed on the SWCC in Figures 10a and 10b are the suctions

390 corresponding to moisture contents determined from sensor measurements (open

391 squares) in bench scale Model Tests 1 and 2 after 120 and 90 days, respectively. As

392 indicated, the suction at shallow depth is very high and at the lowest two depths is

393 essentially unchanged in both model tests.

394 Suction values at four and three different elapsed times, respectively, for Model

395 Test 1 and 2 are plotted against depth in Figure 11, with suction on a log scale. The

396 elapsed times correspond to observed maximum crack depths of 0.1, 0.2, and 0.3 m for

397 each model test; an additional elapsed time is included for Model Test 1 corresponding

© The Author(s) or their Institution(s)

Page 19 of 45 Canadian Geotechnical Journal

398 to a maximum crack depth of 0.43 m. There is a very large suction gradient with depth

399 particularly from 0.04 to 0.22 m. The curves converge to a matric suction of

400 approximately 134 kPa at a depth of 0.59 m, below which very little suction change

401 occurred. The actual variation in suction between sensor locations is probably highly

402 nonlinear, particularly at shallow depth. However, due to experimental uncertainty, the

403 authors simply chose to represent the variation in suction with a piecewise linear curve

404 connecting the data points. Experimental uncertainty in the sensor measurements

405 emanates in part from non-uniform drying due to crack development and sensor

406 position relative to cracks.

407 Comparison of observed and predicted crack depths in bench scale experiments
Dr

408 To assess the validity of the proposed crack depth model, predicted changes in suction
aft

409 to produce cracking (Equation 3) were plotted along with actual suction profiles

410 corresponding to observed crack depths in the bench scale models. The depth at which

411 the actual suction profiles intersect the predicted suction changes represents the

412 predicted crack depth. In Figure 12, the suction change profiles corresponding to

413 observed crack depths of 0.2, 0.3 and 0.43 m for Model Test 1 and 0.2 and 0.3 m for

414 Model Test 2, are plotted along with the predicted cracking suction (Equation 3). Suction

415 profiles corresponding to crack depths of 0.1 m were not included because of

416 insufficient number of moisture sensors to define the suction profile in the upper 0.1 m.

417 The positions of the predicted cracking suction lines shift because the modulus ratio

418 (E/H) varies as the suction increases with time. Modulus ratios used in Equation 3

419 represent a depth-weighted average value determined using the equation in Figure 5b

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 20 of 45

420 for the range of matric suction above the depth of cracking. Implicit in the use of the

421 equation in Figure 5b to model E/H is that the dependency on matric suction is the same

422 for different soils compacted to different initial moisture conditions. Tables 3 and 4

423 present a summary of the observed and predicted depth of cracking indicated by the

424 arrows in Figure 12, for Model Tests 1 and 2, respectively. Corresponding suction and

425 change in suction at the predicted cracking depths, as well as tensile strength and

426 modulus ratios used in predictions are also summarized in Tables 3 and 4. Predicted

427 cracking depths are also plotted against time along with observed cracking depths in

428 Figures 9a and 9c. Reasonable agreement is observed between predicted and actual

429 depths of cracking for the three suction profiles shown.

Dr

430 Discussion
aft

431 In Figure 13, a comparison of the suction and suction change profiles from the first and

432 second bench scale model is presented for maximum crack depths of 0.2 and 0.3 m.

433 Figure 13a shows the data on a log scale while Figure 13b shows the data below 0.2 m

434 in the zone of interest corresponding to observed crack depths of 0.2 and 0.3 m. To

435 reach a maximum crack depth of 0.2 m and 0.3 m for Model Test 1, it took 29 and 70

436 days, respectively. For Model Test 2, it took 63 and 90 days, respectively, for the same

437 maximum crack depths. Thus, on average the rate of drying was greater for Model Test

438 1 than Model Test 2. As mentioned, differences in the rate of drying may be the result

439 of differences in the ambient humidity of the laboratory atmosphere during the testing

440 or the extent of cracking that occurred in each model test. As shown in Figure 8, there

© The Author(s) or their Institution(s)

Page 21 of 45 Canadian Geotechnical Journal

441 was a greater number of primary and secondary cracks in Model Test 1, which would

442 tend increase the surface area exposed to atmospheric drying.

443 While the rate of drying was on average greater for Model Test 1, the suction

444 profiles corresponding to maximum crack depths of 0.2 and 0.3 m were reasonably

445 similar, and so too are the predicted crack depths (Tables 2 and 3). For illustration

446 purposes, in Figure 13, this theoretical relationship is presented for the average E/H

447 values corresponding to depths of 0.2 and 0.3 m determined for both model tests. The

448 intersection of the theoretical curve for cracking suction change with the experimental

449 curves (bold arrows) is seen to occur at similar depths for both model tests, giving similar

450 predictions of cracking depth in each case. It is noted that the predicted cracking depth
Dr

451 corresponding to the observed maximum cracking depth of 0.2 m is about 0.22 m. For
aft

452 the observed maximum cracking depth of 0.3 m, the predicted cracking depth for both

453 model tests is about 0.38 m.

454 The greater inaccuracy of the predicted crack depths for the observed maximum

455 crack depth of 0.3 m is likely due partly to the assumption of a linear variation in cracking

456 suction between the sensor locations. To emphasize this point, a 2nd order polynomial

457 was used to fit the trend in the experimental data points for Model Test 1 below 0.2 m

458 depth, as shown in Figure 14. For this nonlinear curve, the predicted crack depth is about

459 0.33 m as compared to 0.38 m for the linear trend (Figure 13b) and compares better to

460 the actual maximum crack depth of 0.3 m.

461 There are other sources of error that influence the predicted crack depths

462 including but not limited to the uncertainty associated with: sensor measurements,

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 22 of 45

463 visual interpretation of the crack depth, laboratory measurements of soil properties,

464 and theoretical assumptions. As mentioned, the volumetric water content determined

465 by the sensors will be impacted by its proximity to the cracks as they develop, and the

466 non-uniform drying associated with crack development. The estimation of the matric

467 suction is dependent on the interpreted SWCC, which in turn depends on the laboratory

468 measurements and fitting procedures employed. Finally, there are a number of

469 assumptions involved in the theoretical development of the proposed method. Among

470 these is the assumption that the soil behaves as an elastic material. While it is known

471 that soils in fact do not behave as linear elastic materials, there are numerous examples

472 of using the theory of elasticity to interpret soil behavior including cracking (e.g. Morris
Dr

473 et al. 1992; Fredlund and Rahardjo 1993). The authors’ intent was to investigate the
aft

474 viability of a relatively simple analytical model that could be readily implemented in

475 practice. In spite of the limitations, the proposed analytical model shows great promise.

476 However, like any new analytical model, additional validation is needed through further

477 implementation.

478 The proposed model addresses the estimation of the maximum depth of a single

479 desiccation crack, but does not address the spacing. While crack spacing was not the

480 focus of this paper, a practical suggestion is offered for those endeavoring to analyze

481 the impact of desiccation cracking on a particular problem. Extensive laboratory

482 observations of crack spacing are available (e.g. Costa et al. 2013) and considerable work

483 has been performed on the modelling of crack spacing in thin clay layers validated

484 against laboratory experiments (e.g. Kodikara and Choi 2006; Trabelsi et al. 2012;

© The Author(s) or their Institution(s)

Page 23 of 45 Canadian Geotechnical Journal

485 Sánchez et al. 2014; Tran et al. 2019); however, relatively little work has focused on

486 predicting crack spacing in thicker deposits under field conditions. From a practical

487 perspective, a simple way to estimate crack spacing in thicker soil deposits under field

488 conditions is to rely on the relatively limited empirical observations of crack spacing in

489 the field published in the literature, available predictions of crack spacing from

490 computational models reported in the literature, and local observations of crack spacing

491 in subject soils.

492 Conclusions

493 An analytical method based on linear elastic theory and the two-stress state variable

494 approach for unsaturated soil was developed to predict crack depth associated with
Dr

495 changes in suction profiles in desiccating soil. The model equation is developed from the
aft

496 Hookean incremental elastic strain equations for an unsaturated soil based on two

497 stress state variables including net normal stress and matric suction. The model

498 equation gives the change in suction associated with a given crack depth. By

499 superimposing a plot of cracking suction change versus cracking depth from the model

500 equation over a suction change profile of interest, the depth of cracking coincides with

501 the point of intersection of the two lines.

502 Model parameters include the tensile strength, Poisson’s ratio and modulus ratio

503 (E/H) of the soil. The model and sensitivity to these parameters was demonstrated with

504 a hypothetical field situation, and its validity was investigated using a bench scale

505 experiment where desiccation crack development and moisture content changes were

506 monitored over time. Soil tensile strengths and modulus ratios were determined from

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 24 of 45

507 results of desiccation tensile strength tests conducted under approximate uniaxial

508 conditions and were found to be sensitive to initial compaction moisture content and

509 matric suction. Predicted crack depths corresponding to suction change profiles at three

510 different times compared favorably to observed crack depths.

511 References

512 Ayad, R., Konrad, J.M., and Soulie, M. 1997. Desiccation of sensitive clay: application of
513 the model CRACK. Canadian Geotechnical Journal, 34(6), 943-951.
514
515 Bordoloi, S., Ni, J., and Ng, C.W.W. 2020. Soil desiccation cracking and its
516 characterization in vegetated soil: A perspective review. Science of the Total
517 Environment, 138760.
518
519 Cheng, Q., Tang, C. S., Zeng, H., Zhu, C., An, N., and Shi, B. 2020. Effects of microstructure
520 on desiccation cracking of a compacted soil. Engineering Geology, 265, 105418.
Dr

521
522 Cordero, J.A., Useche, G., Prat, P.C., Ledesma, A., and Santamarina, J.C. 2017. Soil
523 desiccation cracks as a suction–contraction process. Géotechnique Letters, 7(4), 279-
aft

524 285.
525
526 Costa S., Kodikara J., Barbour S.L., Fredlund D.G. 2018. Theoretical analysis of
527 desiccation crack spacing of a thin, long soil layer. Acta Geotech, 13(1):39-49.
528
529 Costa, S., Kodikara, J. K., and Shannon, B. 2013. Salient factors controlling desiccation
530 cracking of clay in laboratory experiments. Géotechnique, 63(1), 18-29.
531
532 Fredlund, D.G. and Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soils. John Wiley
533 and Sons, New York.
534
535 Fredlund, D.G., and Xing, A. 1994. Equations for the soil-water characteristic curve.
536 Canadian geotechnical journal, 31(4), 521-532.
537
538 Kodikara, J.K., Barbour, S.L., and Fredlund, D.G. 2000. Desiccation cracking of soil layers.
539 Balkema, Unsaturated soils for Asia, 90 (5809), 139.
540
541 Kodikara, J.K., and Choi, X. 2006. A simplified analytical model for desiccation cracking
542 of clay layers in laboratory tests. In Unsaturated Soils 2006, ASCE, 2558-2569.
543

© The Author(s) or their Institution(s)

Page 25 of 45 Canadian Geotechnical Journal

544 Konrad, J.M., and Ayad, R. 1997. An idealized framework for the analysis of cohesive
545 soils undergoing desiccation. Canadian Geotechnical Journal, 34(4), 477-488.
546
547 Lachenbruch, A.H. 1961. Depth and spacing of tension cracks. Journal of Geophysical
548 Research, 66(12), 4273-4292.
549
550 Lakshmikantha, M.R., Prat, P.C., and Ledesma, A. 2009. Image analysis for the
551 quantification of a developing crack network on a drying soil. Geotechnical Testing
552 Journal, 32(6), 505-515.
553
554 Lau, J.T.K. 1987. Desiccation cracking of clay soils. M.Sc. Thesis, Department of Civil
555 Engineering, University of Saskatchewan, Canada.
556
557 Lu, N., Kim, T.H., Sture, S., and Likos, W.J. 2009. Tensile strength of unsaturated sand.
558 Journal of engineering mechanics, 135(12), 1410-1419.
559
560 Miller C.J., Mi H., Yesiller N., 1998. Experimental analysis of desiccation crack
561 propagation in clay liners. Journal of the American water resources association, 34(3),
562 677-686.
563
Dr

564 Morris, P.H., Graham, J., and Williams, D.J. 1992. Cracking in drying soils. Canadian
565 Geotechnical Journal, 29, 263-277.
aft

566
567 Peron, H., Laloui, L., Hueckel, T., and Hu, L.B. 2009. Desiccation cracking of soils.
568 European journal of environmental and civil engineering, 13(7-8), 869-888.
569
570 Peron, H., Laloui, L., Hu, L.B., and Hueckel, T. 2013. Formation of drying crack
571 patterns in soils: A deterministic approach. Acta Geotechnica, 8: 215-221.
572
573 Pouya A., Vo T.D., Hemmati S., and Tang A.M. 2019. Modeling soil desiccation cracking
574 by analytical and numerical approaches. Int J Numer Anal Methods Geomech, 43:738-
575 763.
576
577 Sánchez, M., Manzoli, O.L., and Guimarães, L.J. 2014. Modeling 3-D desiccation soil crack
578 networks using a mesh fragmentation technique. Computers and Geotechnics, 62, 27-
579 39.
580
581 Snyder, V.A., and Miller, R.D. 1985. Tensile strength of unsaturated soils. Soil Science
582 Society of America Journal, 49(1), 58-65.
583
584 Sun, J., Wang, G., and Sun, Q. 2009. Crack spacing of unsaturated soils in the critical
585 state. Chinese Science Bulletin, 54(12), 2008-2012.
586

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 26 of 45

587 Tang, C.S., Pei, X.J., Wang, D.Y., Shi, B., and Li, J. 2015. Tensile strength of compacted
588 clayey soil. Journal of Geotechnical and Geoenvironmental Engineering, 141(4),
589 04014122.
590
591 Tang, C.S., Wang, D.Y., Zhu, C., Zhou, Q.Y., Xu, S.K., and Shi, B. 2018. Characterizing
592 drying-induced clayey soil desiccation cracking process using electrical resistivity
593 method. Applied Clay Science, 152, 101-112.
594
595 Tang, C.S., Zhu, C., Cheng, Q., Zeng, H., Xu, J. J., Tian, B.G., and Shi, B. 2021. Desiccation
596 cracking of soils: A review of investigation approaches, underlying mechanisms, and
597 influencing factors. Earth-Science Reviews, 103586.
598
599 Tang, C.S., Zhu, C., Leng, T., Shi, B., Cheng, Q., and Zeng, H. 2019. Three-dimensional
600 characterization of desiccation cracking behavior of compacted clayey soil using X-ray
601 computed tomography. Engineering geology, 255, 1-10.
602
603 Trabelsi, H., Jamei, M., Zenzri, H., and Olivella, S. 2012. Crack patterns in clayey soils:
604 experiments and modeling. International Journal for Numerical and Analytical methods
605 in geomechanics, 36(11), 1410-1433.
606
Dr

607 Tran, H.T., Wang, Y., Nguyen, G.D., Kodikara, J., Sanchez, M., and Bui, H.H. 2019.
608 Modelling 3D desiccation cracking in clayey soils using a size-dependent SPH
aft

609 computational approach. Computers and Geotechnics, 116, 103209.

610
611 Varsei, M., Miller, G.A., and Hassanikhah, A. 2016. Novel approach to measuring tensile
612 strength of compacted clayey soil during desiccation. Int. J. Geomechanics.
613
614 Vogel, H.J., Hoffmann, H., Roth, K. 2005. Studies of crack dynamics in clay soil I.
615 Experimental methods, results, and morphological quantification. Geoderma, 125, 203-
616 211.
617
618 Vo T.D., Pouya A., Hemmati S., Tang A.M. 2017. Numerical modelling of desiccation
619 cracking of clayey soil using a cohesive fracture method. Comput Geotech, 85:15-27.
620
621 Vo T.D., Pouya A., Hemmati S., Tang A.M. 2018. Modelling desiccation crack geometry
622 evolution in clayey soils by analytical and numerical approaches. Can Geotech J.
623 91:89-101.
624
625 Wang, L.L., Tang, C.S., Shi, B., Cui, Y.J., Zhang, G.Q., and Hilary, I. 2018. Nucleation and
626 propagation mechanisms of soil desiccation cracks. Engineering Geology, 238, 27-35.
627
628 Wei, Y., and Miller, G.A. 2019. Determining Osmotic Suction Using a Chilled Mirror
629 Device. Geotechnical Testing Journal, 42(6), 1457-1474.
630

© The Author(s) or their Institution(s)

Page 27 of 45 Canadian Geotechnical Journal

631 Yin, P., Qi, S., and Vanapalli, S.K. 2016. A Framework for Predicting the Depth of
632 Desiccation-induced Cracks in Clayey Soils. Proceedings of Geo Vancouver 2016.
633
634 Yin, P., and Vanapalli, S.K. 2018. Model for predicting tensile strength of unsaturated
635 cohesionless soils. Canadian Geotechnical Journal, 55(9), 1313-1333.
636
637 Zapata, C.E., Houston, W.N., Houston, S.L., and Walsh, K.D. 2000. Soil–water
638 characteristic curve variability. In Advances in unsaturated geotechnics, ASCE 84-124.
639

Dr
aft

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 28 of 45

Table 1. Assumed stress and strain conditions for field and bench scale model situations applied
in Equation 1.
Parameter Field Bench
( x -u a )o Kozc Kozc
( x -u a )f t t
d( x -u a )  t -K o  z c  t -K o  z c
( y -u a )o Kozc Kozc
( y -u a )f t 0
d( y -u a )  t -K o  z c -K o  z c
( z -u a )o zc zc
( z -u a )f zc zc
d( z -u a ) 0 0
Dr

(u a -u w )o (u a -u w )o (u a -u w )o
(u a -u w )f (u a -u w )c (u a -u w )c
aft

d(u a -u w ) d(u a -u w )c d(u a -u w )c

d x 0 0
d y 0 
d z  

© The Author(s) or their Institution(s)

Page 29 of 45 Canadian Geotechnical Journal

Table 2. Parameters used to create hypothetical drying curves in Figures 4 and 6. The curves are
based on a logarithmic decay function of the form d(ua-uw)=e(z-b)/-a where z=depth, and a and
b are constant in the table below.
Curve No. 1 2 3 4 5 6
a * 1.2 1 0.9 0.75 0.65
b * 5 5 5 5 5
*-Curve No. 1 respresents zero suction change with depth.
Dr
aft

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 30 of 45

Table 3. Parameters used in Eq. 3, predicted and observed crack depths for bench Model 1.
d(u a -u w )c (kPa) (u a -u w )c (kPa)  t (kPa) (E/H )avg Predicted z c (m) Observed z c (m)
165 299 -23.4 0.152 0.22 0.20
200 334 -23.4 0.130 0.37 0.30
253 387 -23.4 0.105 0.45 0.43
Dr
aft

© The Author(s) or their Institution(s)

Page 31 of 45 Canadian Geotechnical Journal

Table 4. Parameters used in Eq. 3, predicted and observed crack depths for bench Model 2.
d(u a -u w )c (kPa) (u a -u w )c (kPa)  t (kPa) (E/H )avg Predicted z c (m) Observed z c (m)
170 304 -23.4 0.153 0.22 0.20
Dr

170 304 -23.4 0.148 0.38 0.30

aft

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 32 of 45

Dr
aft

Figure 1. a) Initial stress conditions immediately after construction of compacted clay fill at
depth zc; b) initial suction profile and that when crack development to zc is imminent; and c)
final stress conditions when desiccation crack development to zc is imminent.

© The Author(s) or their Institution(s)

Page 33 of 45 Canadian Geotechnical Journal

Dr
aft

Figure 2. Schematic of desiccation cracks developed under field conditions: a) hypothetical

crack geometry, b) model representation and stress conditions when cracking is imminent.

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 34 of 45

Dr
aft

Figure 3. Schematic of desiccation crack developed under test box conditions: a) hypothetical
crack geometry, b) model representation and stress conditions when cracking is imminent.

© The Author(s) or their Institution(s)

Page 35 of 45 Canadian Geotechnical Journal

a)
0m
compacted lean clay (CL)
LL=39, PI=17, %fines=95
Gs=2.75
dmax=18.8 kN/m3 (std. Proctor)
wopt=14.9% (std. Proctor)
=20.1 kN/m3 (R=95%)
d=17.9 kN/m3 (R=95%)
initial w=16.9 %=OMC+2%
Dr

initial w=0.308
initial (ua-uw)o=134 kPa
aft

=34o
Ko=1-sin=0.44
= Ko/(1+ Ko)=0.31
E/H=0.15
t=-23.3 kPa

5m
Figure 4. a) Field example soil profile and properties. These are also the properties of the soil
used in the experimental bench scale model. b) Example drying profiles and cracking suction for
the field example.

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 36 of 45

Dr

Figure 5. a) initial volumetric water content versus tensile strength (t), and b) cracking suction
versus modulus ratio (E/H) determined from uniaxial tensile strength measurements during
aft

desiccation.

© The Author(s) or their Institution(s)

Page 37 of 45 Canadian Geotechnical Journal

Dr
aft

Figure 6. Influence of a) tensile strength, b) modulus ratio, and c) Poisson’s ratio on predicted
crack depth.

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 38 of 45

a)

Dr

Figure 7. a) Schematic drawing of the test box for desiccation crack experiments showing
aft

location of moisture sensors at depths of 0.04, 0.22, 0.42, 0.59 and 0.84 m. b) Photograph of
the test box after compaction, prior to initiating the test.

© The Author(s) or their Institution(s)

Page 39 of 45 Canadian Geotechnical Journal

a) c)

b) d)

Dr
aft

Figure 8. Desiccation crack development after 120 days for Model 1: a) top view, b) side view;
and 90 days for Model 2: c) top view, d) side view.

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 40 of 45

Dr
aft

Figure 9. a) & c) Observed maximum crack depth versus elapsed time for Test Models 1 & 2,
respectively. Note, predicted values of crack depth discussed later are also presented. b) & d)
Change in volumetric water content corresponding to different depths versus elapsed time for
Test Models 1 & 2, respectively.

© The Author(s) or their Institution(s)

Page 41 of 45 Canadian Geotechnical Journal

Dr

Figure 10. Soil water characteristic curve (SWCC) for the test soil. Solid circles represent
experimentally determined values of matric suction, while the solid line is a best fit curve based
aft

on the Fredlund and Xing (1994) equation. The dashed line is an empirical curve based on the
method of Zapata et al. (2000). Open squares correspond to the water contents from moisture
sensors in: a) bench scale Model 1 after 120 days and b) bench scale Model 2 after 90 days.

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 42 of 45

Dr
aft

Figure 11. Matric suction profiles estimated from water content measurements and the SWCC
for: a) Model Test 1 at elapsed times of 9, 29, 70 and 120 days corresponding to observed crack
depths of 0.1 0.2, 0.3 and 0.43 m, respectively; and b) Model Test 2 at elapsed times of 27, 63
and 90 days corresponding to observed crack depths of 0.1, 0.2, and 0.3 m, respectively.

© The Author(s) or their Institution(s)

Page 43 of 45 Canadian Geotechnical Journal

Dr
aft

Figure 12. Suction change profiles and predicted cracking suction lines (Equation 3) for a) bench
scale Model 1 corresponding to observed crack depths of 0.2, 0.3 and 0.43 m, and b) bench
scale Model 2 corresponding to observed crack depths of 0.2, and 0.3 m.

© The Author(s) or their Institution(s)

 Canadian Geotechnical Journal Page 44 of 45

Dr
aft

Figure 13. Comparison of suction profiles from bench Model Test 1 and Model Test 2,
corresponding to crack depths of 0.2 and 0.3 m: a) change in suction on a log scale versus
depth; b) change in suction versus depth.

© The Author(s) or their Institution(s)

Page 45 of 45 Canadian Geotechnical Journal

Dr
aft

Figure 14. Comparison of suction profiles from bench Model Test 1 and Model Test 2, for
depths below 0.2 m corresponding to observed maximum crack depths of 0.3 m. A non-linear
polynomial curve is fitted to the data from Model Test 1.

© The Author(s) or their Institution(s)