The Effect of Bulk Density and Initial Water Content onInfiltration in Clay Soil Samples1

F. A. GUMBS AND B. P. WARKENTIN2

ABSTRACT

Infiltration measurements were made on swelling clay soilsamples packed into columns. Small increases in bulk densityover the range 1.10 to 1.25 g/cm3 markedly decreased the rateof water movement. The magnitude of the effect was greaterfor confined samples than unconfined samples at all initialwater contents. A 1-cm compact layer in the profile retardedwater movement if the soil was confined. In partially confinedsamples the soil in the compact layer would swell on wetting,and water movement was retarded only when the bulk densityafter swelling still exceeded the bulk density of the remainderof the column. Bulk densities below 1.05 g/cm3, and heat ofwetting in partially confined samples with 0% initial watercontent produced nonlinear distance to wet front vs. squareroot of time relationships. Comparison of horizontal and verti-cal infiltration showed that under these experimental conditionsgravity contributed significantly to water movement at highinitial water content.

Additional Index Words: diffusivity, compacted soil, swellingclay, aggregates.

THE RATE OF ADVANCE of the wet front depends on perme-ability, changes in bulk density, layering in the profile,

and on initial water content of the soil. Olsen (1960) mea-sured an exponential increase in permeability with increasein porosity. Millington and Quirk (1959) and Philip(1957a) found that permeability increases exponentiallywith increase in initial water content. Phillip (1957b) hasshown that the rate of advance of the wet front increases,and rate of infiltration decreases, with increasing initialwater content for both short and long time infiltration.Hanks and Bowers (1962) and Miller and Gardner(1962), found that when there is textural layering in aprofile, infiltration is controlled by the less permeable layer.

There is little information available on the effect of bulkdensity changes and initial water content on water infiltra-tion into swelling clay soils. This study reports such mea-surements on swelling clay soil samples. The rate of ad-vance of the wet front was measured for vertical and hori-zontal infiltration into confined or partially confined soilcolumns with the water entering at small positive pressures.The rates of advance of the wet front, taken from the slopesof the distance to the wet front against the square root oftime, are used to evaluate the effect of bulk density onthe infiltration. Diffusivity-water content relationships arecalculated to evaluate the effect of initial water content.

1 Contribution from the Dep. of Soil Science, MacdonaldCollege of McGill University. Part of the work submitted asan M.S. thesis by the senior author. This study was supportedby a Grant in Aid of Research from the National ResearchCouncil, Canada. Received Jan. 10, 1972. Approved June 27,1972.

2 Graduate Research Student and Professor, respectively, De-partment of Soil Science. The senior author is now Lecturer inSoil Science, University of the West Indies, Trinidad.

No attempt is made in this paper to predict infiltrationfrom the theories proposed recently to describe water infil-tration into swelling soils (Philip and Smiles, 1969; Za-slavsky, 1964). Such predictions require more measure-ments of soil parameters, and more information on theinteraction between volume change and soil water poten-tials, than were available in this study.

Results on infiltration into columns of sieved and packedsoil samples may not be valid for clay soils in the field,because of differences in void size distribtuion. The cracksand natural peds in field soils result in larger units thanused in the laboratory, and in greater nonhomogeneity.Despite these differences, the effects observed for unsatu-rated flow would be expected to occur in the field. Thebulk densities around 1.1 g/cm3 used in the experimentsare slightly lower than the values of 1.2 to 1.3 g/cm3

measured for the plow layer of this soil.

EXPERIMENTAL PROCEDURE

The soil used in this study was the Ste-Rosalie clay, a humicgleysol. A sample was taken from the C-horizon, which has60-70% clay, 25-30% silt, 0-10% sand, and 2-5% organicmatter (Lajoie and Baril, 1954). The minerals present are micaand chlorite, with lesser amounts of quartz, feldspar, and am-phibole, and small amounts of montmorillonite or vermiculite.

The soil sample was dried and ground, first in a mechanicalgrinder to break down the large clods and then by glass-on-glass grinding. In the sample as used, the aggregate sizes 0.60-0.84 mm, 0.42-0.60 mm, 0.30-0.42 mm, 0.19-0.30 mm, andless than 0.18 mm were present in the approximate ratio of2:1.6:2:2:1. These aggregates are relatively stable on wetting,so this ratio of aggregate sizes also represents the sample afterwetting. Compaction to bulk density values up to 1.2 g/cm3 didnot crush the aggregates.

Infiltration was studied with soils packed in columns madeup of Lucite rings 1 cm in height. The first measurements weremade using columns 4.5 cm inside diameter. Most of the meas-urements were with columns 3.2 cm inside diameter, becausethe smaller samples were more convenient. The work of Lalet al. (1970) indicates that the rate of advance of the wet frontin horizontally confined columns increases as column size in-creases. This is due to the different degree of swelling whichtakes place in the top layers of the soil column. In the experi-ments reported below, comparisons are therefore made onlybetween soil columns of the same size.

The columns were packed by compressing known weightsof soil to predetermined heights. For bulk densities below 1.15g/cm3 the compression was done by hand in the already assem-bled columns; for the higher bulk densities the soil was com-pressed in the 1-cm sections with a hydraulic press and thesections assembled and squeezed together. Care was taken toachieve the best possible contact between sections by disturbingthe surface of the soil to a depth of about 1 mm with a manu-ally rotated wire brush before the next increment of soil or thenext 1-cm section was applied.

Infiltration measurements were made on duplicate samples.If the difference in the slopes of the distance to the wet front(x) vs. square root of time (f%) of successive infiltration runswas greater than about 5%, a third replicate was used and theaverage of the three taken (provided no difference was greaterthan 10%). Abnormal results (i.e. greater than a 10% differ-

720

GUMES & WARKENTIN: EFFECT OF BULK DENSITY AND WATER CONTENT ON INFILTRATION IN CLAY 721

Table 1—Summary of slopes of x vs. f& lines (\) under different conditions of infiltrationVertical infiltration

Watercontent

0.00.00.04.03.03.03.06.57.48.18.48.55.59.01919

Partially confinedBulk density

g/cm1

1.1.1.1.1.1.1.

1.

1.

1.1.1,1,

142184208115115137169

111

146

.208

.178

.134,183

Xcm/minW

1.411.050.971.501.541.340.98

1.96

1.67

0.530.852.102.04

AX/ADb

2.28

0.33

0.83

0.12

Horizontal InfiltrationConfined

Bulk density1st cm

1.1791.279

1.1981.255

1.2491.262

Average

1.1751.184

1.1611.174

1.1841.204

A AX/ADbcm/minVi

0.87 ,0.64 2'61

0.74 . _,0.54 1>M

2.151.72 2-19

ConfinedBulk density X AX/ADp

1st cm Average——— g/cms ———

1.1.

1.179 1.

1.

1.

1.

1.197 1.1.275 1.

219246154

106

161

224

179196

cm/minV2

0.600.520.83

1.09

0.81

0.52

1.501.27

0.30

0.46

1.35

• AD,, = 0.1 g/cm'

ence), which occurred in about 10% of the trials, were dis-carded.

The "confined" samples had a porous stone fixed at thewater entry end of the column. The "partially confined" sam-ples were free to swell in the vertical direction at the top ofthe column. A 1.5 cm layer of water was ponded on the surface,protected by a loose disc of filter paper. The base of the col-umn had a fixed porous plate.

Samples with different initial water contents, Wi, of 0%,3%-4%, 7%-9%, and 19% by weight were prepared asfollows:

1) 0% Water Content—Samples were oven dried at 105-110C for 24 hours.

2) 3%-4% Water Content—This was the water content ofthe air-dry soil.

3) 7%-9% Water Content—Samples in thin layers werekept for periods of up to 10 days in an atmosphere of 100%relative humidity.

4) 19% Water Content—The soil samples were spread thinlyin a narrow band on a plastic sheet and sprayed with the calcu-lated weight of water. The samples were then mixed thor-oughly and stored in double walled plastic bags for 1 week.

The start of any infiltration experiment was taken at / — 0when the first drops of water contacted the soil surface. There-after, readings were taken of the time, the distance to the wetfront, and the volume of water infiltrated.

At the end of the experiment the water supply was stopped,the head of water drained away, and the 1 cm sections sepa-rated by pushing a thin (0.25 mm) rigid shim between eachpair of rings. The entire sectioning took 5-10 min, dependingon the length of the column. The soil slices were then quicklyremoved from the rings and the water content determinedgravimetrically.

RESULTS AND DISCUSSIONRate of Wetting

Infiltration theory summarised by Philip (1969) pre-dicts a linear x vs. f*- relation for horizontal infiltration,and a linear x vs. t^ relation for the early phases of infil-tration in the vertical direction with a gradual shift to anonlinear x vs. f% at longer times. These relationships holdif the boundary and initial conditions can be transformedby the use of a variable X = x/i*. In these studies, straightline relations of x vs. t^ were obtained for both horizontaland vertical infiltration at all initial water contents greaterthan 0%, and at all bulk densities greater than 1.05 g/cm3.An occasional replicate horizontal or vertical infiltrationtrial gave a nonlinear relation of x vs. f%. Since this oc-

curred only occasionally, such trials were considered tobe due to nonreproducibility of the initial and boundaryconditions demanded by the theory, and were thereforediscarded. The x vs. ?& straight lines did not always gothrough the origin. This was believed to be due to inaccu-racies in determining zero time at the start of infiltration.

At the low bulk density of 1.05 g/cm3, subsidence ofthe sample took place on wetting, and this probably alteredthe boundary conditions sufficiently to produce a nonlinearx vs. f% relationship. These results for low bulk densityare not used in this paper.

Table 1 summarizes the measured A (slope of the x vs.fa relation) values for different bulk densities and dif-ferent initial water contents. The A values for 0% initialwater content are the initial slopes of the A: vs. t^ relation-ship, since the slopes changed slightly near the end ofwetting. Oven-dry samples (0% initial water content)showed linear x vs. t^ plots during wetting of about % ofthe column, after which deviations from linearity couldbe detected.

An increase in bulk density over the range 1.10 to 1.25g/cm3 markedly decreased the rate of advance of the wetfront. At a given water content, there was an exponentialrelationship between increase in bulk density and decreasein A. It was found that (i) vertical water movement inconfined samples was more affected by a change in bulkdensity than water movement in unconfined samples, (ii)the effect of a change in bulk density on the horizontalmovement of water in confined samples increased with anincrease in initial water content, (iii) the effect of gravitywas small if the samples were confined. This minor effectof gravity is expected for a clay soil for short time ofinfiltration. A soil with 3% initial water showed little dif-ference between vertical and horizontal rates of wet frontadvance but with 19% initial water content a gravity effectis apparent, AV:AH — 1.72:1.27.

These results can be compared with those of Jackson(1963) for Pine silty clay soil, the one closest to Ste-Rosalie clay in grain size distribution. He found that anincrease in porosity of 0.047 resulted in a 40% increasein the distance to the wet front at 900 min for a sampleat an initial water content of 5%. The results in Table 2show that the changes were of the same order of magnitude

722 SOIL SCI. SOC. AMER. PROC., VOL. 36, 1972

Table 2—The effect of change in porosity on the advance of thewet front in horizontal columns

Soil

Pine*Pine'Ste-RosalleSte-RosalleSte-RosalleSte-RosalleSte-RosalieSte-Rosalle

Initialmoisturecontents

%5.04.50.00.09.09.0

19.019.0

Timeminutes

900900900900900900200200

Distanceto

wet frontcm

22:531.517.2

19.717.523.519.023.1

Porosity

%0.5090.5560.5530.5880.5470.5900.5630.557

IncreaseIn

porosity

0.047

0.035

0.043

0.006

Increasein distanceto wet front

%

40.0

14.5

34.3

21.6

' Results from Jackson (1963).

for Ste-Rosalie clay at an initial water content of 9%, butthere is a strong dependence on the initial water content.The higher the initial water content, the greater is theresponse to a change in porosity. At 0% there was abouta 15% increase, at 19% water it was about 22% for aporosity change of 0.006 and an infiltration time of 200min. This indicates an increase of over 100% in distanceto wet front when compared at the same change in porosity.

Phillip (1957b) stated that the increased rate of advanceof the wet front with increase in initial water content wasmainly due to a water storage effect. The results here indi-cate that in vertical columns there is in addition an in-creased gravity effect due to the increased saturation ofthe top part of the column. At a bulk density of 1.14g/cm3, the saturation increased from 87% to 94% as theinitial water content increased from 0 to 19%.

It was also observed that the percent saturation of thetop part of the column decreased from 90% to 85% withan increase in the size of the column from 3.2 cm to 4.5cm inside diameter, and from 87% to 82% with a decreasein bulk density from 1.2 g/cm3 to 1.1 g/cm3. Lal et al.(1970) showed that more swelling takes place in the sur-face layers of a large than of a small soil column. This willproduce lower bulk densities at the top of larger columnsand a lower percent saturation.

Effect of Compact Layers

When columns were packed so that there were differ-ences in bulk density between the top half and the bottom

25

20

15

10

Confined,- «=I9%Unconfined, w=0%Confined, w=9%

10 15 20, mins*

25 30

Fig. 1—Advance of wet front for vertical infiltration into col-umns with two layers of different bulk density. 1—higherDb on top; 2 and 3—lower Db on top.

half of the columns, the rate of advance of the wet frontwas different in the two sections and the x vs. f% relation-ship showed the expected nonlinearity (Fig. 1). Whenwater moves from a lower bulk density at the top of thecolumn to a higher bulk density at the bottom, (bulkdensity profile Fig. 2, numbers 2 and 3) x vs. t^ is con-cave downwards (Fig. 1, numbers 2 and 3). When move-ment is from a higher bulk density to a lower bulk density,x vs. f^ is concave upward (Fig. 1, number 1). Over therange of bulk densities tried, this nonlinearity in the x vs. t*relationship was manifested regardless of the initial watercontent (0% to 19%) and the degree of confinement.

The effect of a higher bulk density layer at the surfaceof a confined column of soil is shown in Table 1. At 0%initial water content when the bulk density of the surface1 cm layer was increased from 1.179 to 1.279 g/cm3

(average density of the column being 1.180 g/cm3) therate of advance of the wet front decreased by 0.235cm/min^ (a 27% decrease). These wet front advanceswere still linear functions of t^. These observations con-firm that the rate of infiltration depends to a large extenton the diffusivity near the surface. This is in agreementwith the analytical results of Hanks and Bowers (1963).The results also show that the linearity of x vs. t^ is nota sensitive criterion for uniformity of bulk density whenresistance to flow occurs near the water entry boundary.Philip (1967) showed analytically that t~* behavior ofsorption rates may occur even in the absence of homo-geneity. From this work it seems that the nature of theheterogeneity affects the linearity of the x vs. t^ relation.Probably the nature of the variation of the diffusivity-watercontent relationship with depth in the column, as influencedby heterogeneity, is important in determining the linearityof x vs. (to.

An increase in the bulk density of a soil column from1.1 to 1.2 g/cm3, or a 1-cm compact layer in a confinedcolumn, markedly decreased the rate of movement ofwater. However, a compact layer (1 cm thick) of bulkdensity, 1.25 g/cm3 placed at one of several depths (0-1,2-3, 5-6, or 8-9 cm) in a profile of air dry soil of rela-tively uniform bulk density of 1.11 to 1.12 g/cm3 did notretard the advance of the wet front (Fig. 3) when the

10 I.Q5Bulk Density

MO MS 120 125

10

20

25

Fig. 2—Bulk density profile after infiltration for samples shownin Fig. 1.

GUMES & WARKENTIN: EFFECT OF BULK DENSITY AND WATER CONTENT ON INFILTRATION IN CLAY 723

E is

at surface averageUniform packing 0984 UI5I" cm compacted to Dt=L2 0.910 UI82" « » " Db=l2 0.900 U196"1 » " " Db=l2 0306 11099th " " « Db=t2 0366 LI08$' « » " Dt=l5 0.867 UO6

2.5i

10 15l",

Fig-. 3—Advance of wet fronffor vertical infiltration in sampleswith 1-cm compact layer in the profile.

sample was not confined at the water entry end. The swell-ing of the compact layer was enough to make it unrecog-nizable when final bulk densities were measured at the endof the experiment. However, a compact layer of bulk den-sity 1.50 g/cm3 placed anywhere in the profile decreasedthe rate of water movement. Figure 3 shows the case forthe compact layer at the 2-3 cm depth. The final bulkdensity of this layer was about 1.3 g/cm3. A certain mini-mum value of bulk density in the compact layer is there-fore necessary for water movement to be affected. Whenthe column of soil is confined, the bulk density does notchange on wetting and a smaller difference between theinitial bujk densities of the compact layer and the rest ofthe profile^Nvill influence the rate of water movement.

Effect of Confining the SampleAt any given bulk density (taken as the average bulk

density of the column at the end of infiltration) the rateof advance of the wet front was always greater in the par-tially confined than in the confined samples (Table 1).The volume change took place rapidly during infiltrationwhen the soil was free to swell in the vertical direction atthe top of the column. About 75% of swell occurred in10-15 min, where the total swell of about 1 cm increasein height of the column at the end of the experiment wastaken as 100%. It is reasonable to suggest that the in-creased infiltration rate in partially confined samples maybe partly due to the increased conductivity in the top of thecolumn as a result of a decrease in the bulk density. Anydispersion that may have taken place on the surface of thepartially confined samples was not enough to offset theincreased infiltration from the decrease in the bulk densityof the surface layers.

Effect of Initial Water Content

Figure 4 summarizes the results on rate of advance ofthe wet front at different initial water contents of the soil.The rate of advance of the wet front was always greaterin samples with 0% initial water than in samples with

5 10 15 20Initial Water Content,%

Fig. 4—Dependence of X (slope of * vs. (V4 line) on initialwater content at different bulk densities for vertical, uncon-fined infiltration.

W, Db

c——e 102 % LI08*——a 0 % 1.114a——= 3.6% UI7

Iff3

10'

0.2 0.4 0.6 0.8 1.0

Fig- 5—Diffusivity vs. degree of saturation for different initialwater contents.

initial water content up to about 10%. Except for 0% ini-tial water A. shows the expected increase with increase ininitial water content.

Preliminary measurements of capillary rise in samplesdried by heating or over P2O5 indicated that oven dryingincreased rate of movement. This may be an influence onsurface properties. This aspect needs further investigation.

Figure 5 shows the calculated diffusivity/ water contentrelationships at a bulk density of about 1.11 g/cm3/andthree initial water contents. Similar results were obtainedat other constant bulk densities for initial water contentsbetween 0% and 19%. The diffusivities were calculatedfrom the infiltration results, by the method of Bruce andKlute (1956). At water contents less than saturation thereis a dependence on initial water content. This may resultfrom changes in structure during packing of the samplesat different moisture contents. Because of this effect, it isimportant to use the relevant diffusivity/water contentrelationship in the prediction of infiltration into soil col-umns at different initial water contents.

The effect of bulk density on diffusivity is shown in Fig.6 for an initial water content of 8 to 9%. The valuesdepend upon initial water content, but at any one initialwater content larger than 0%, an increase in bulk densityin the range 1.10-1.25 g/cm3 had the same effect on dif-fusivity. If the diffusivity/water content (D(0)) functionat one bulk density is known, then the D(6) relationship

724 SOIL SCI. SOC. AMER. PROC., VOL. 36, 1972

10'

Sio5

£•

flOl

I0f

1.161

02 0.4 0.6(e-em-ej

0.8 1.0

Fig. 6—Diffusivity vs. degree of saturation for different bulkdensities at 8-9% initial water content.

at any other bulk density in the above mentioned rangecan be found from a knowledge of the diffusivity at asingle water content.

When water wets soil samples with increasing initialwater contents, two phenomena with opposite effects onwater movement occur; the heat of wetting decreases andcapillary conductivity increases. Anderson and Linville(1961) and Anderson et al. (1963), found that the tem-perature increases can be quite high for oven dry soils, andfall off very rapidly as water content is increased. This,they found, can have a measurable effect on moving waterforward in the vapor phase. On the other hand, capillaryconductivity is not increased much by an increase in initialwater content in this low moisture range. The nonisother-mal conditions produced by the heat of wetting in the sam-ples with 0% initial water content in this experiment maybe responsible for the nonlinear x vs. tV2 observed. Tem-perature differences due to heat of wetting were not mea-sured in these experiments, but differences were obviousto the touch.

SUMMARY

This paper describes some of the ways in which infiltra-tion of water into swelling soils differs from infiltrationinto rigid porous materials. Infiltration into confined sam-ples is lower than into samples unconfined at the surfacebecause swelling increases porosity and conductivity of thesurface layers which have a large influence on infiltrationrate. Thin compacted layers in unconfined columns havelittle effect on infiltration because after wetting and swellingthe bulk density was decreased to that of the remainder ofthe column. The decrease in infiltration rate with increasein bulk density is larger at higher initial water contents.This is probably again an effect of lower swelling at higher

initial water contents. The calculated diffusivity at anywater content was dependent upon initial water content ofthe sample. Samples with 0% initial water content showedan anomalous behavior in that the rate of wetting washigher than at other initial water contents up to about 10%.

For most conditions, uniformly packed clay soil columnsshowed linear x vs. t1A relationships on wetting. The excep-tions were low bulk density where the soil collapses onwetting, and zero initial water content for samples uncon-fined at the surface where hydration and swelling effectswere greatest. Columns with compacted layers alsoshowed linear x vs. r% wetting patterns if swelling couldreduce the bulk density of the compact layer to that ofthe remainder of the column. If the difference in bulk den-sity was sufficiently large to remain after swelling, non-linear x vs. tv* was obtained.