Water and Chloride Movement Through Soil Columns Simulating Pedal Soils1

J. BOUMA AND J. L. ANDERSON2

ABSTRACTInterpretation of dispersion phenomena in soils is complicated by

many complex interactions within the highly heterogeneous pore sys-tem. Model experiments were therefore designed to investigate thespecific effects on hydrodynamic dispersion of (i) vertically continuousmacropores in different types of microstructure and (ii) different flowregimes. Two homogeneous microstructures were formed by mixing,puddling, and drying of sand and silty clay loam materials to sandyloam and clay loam textures. Small vertically continuous cylindricalpores were made in 30-cm long columns that had a diameter of 10 cmto simulate macropores. Breakthrough curves, following a daily 0.5-cm application of chloride solution, showed that identical macroporeshave different hydraulic functions in different soil materials. This in-termittent flow resulted in much better displacement of untracedwater in the more permeable sandy loam columns, but "immobile"water remained in both soils. Dispersion phenomena in both micro-structures (which represented two "ideal" textural porosities) differedsignificantly when saturated flow was compared with flow through acrust. Implications for the physical interpretation of soil structuredescriptions are discussed.

Additional Index Words: hydrodynamic dispersion, miscible dis-placement, microstructure.

WATER MOVEMENT through pedal soils with a silty clayloam texture was examined in undisturbed soil cores

from three soil series, using chloride as a tracer. Soils withsubangular blocky structure conducted water differentlyfrom soils having prismatic structure (2, 3). The differenceswere hypothetically explained by relating differences inhydraulic conductivity to ped dimensions and assumed ver-tical pore continuity patterns. Information derived fromchloride breakthrough curves allowed an elucidation of porepatterns within columns. Previous studies show that water isable to bypass a large portion of the soil matrix by travelingthrough macropores and interpedal cracks while simulta-neously some water moves at a slower rate through the ma-trix of the peds themselves. (11, 12, 15). Explanations ofobserved breakthrough phenomena were hypothetical, be-cause of the very complex and heterogeneous flow pro-cesses involved. These model experiments were thereforedesigned to test the hypotheses that were made to explainobserved differences among breakthrough curves (2,3, 11,12, 15). By standardizing certain soil structural features,such as occurrence of macropores and different types ofmicrostructure, an independent evaluation of the effect ofsingle well-defined factors on soil physical behavior be-comes possible. Use of undisturbed soil material rarelyallows the opportunity for such analyses. However, ar-tificial soil structure may provide information not directly

'Contribution from the Soil Science Dep., Univ. of Wisconsin, Madisonand the Geological and Natural History Survey, Univ. Extension. Fundedin part by the State of Wisconsin and the Environmental Protection Agency(Grant R802874). Received 21 June 1976. Approved 1 April 1977.

2Soil Scientist, Soil Survey Institute, Box 98, Wageningen, The Nether-lands (formerly Associate Professor of Soil Science, Univ. of Wisconsin,Madison), District Soil Scientist Hennepin Soil and Water ConservationDistrict, 250 N. Central Ave., Wayzata, MN 55391 (formerly ResearchAssistant, Univ. of Wisconsin, Madison).

related to natural soils. Structure descriptions attempt tocharacterize structures in heterogeneous natural soils interms of rather specific and exclusive types in spite of allheterogenities observed (7). Model structures, used in thisstudy, represent "ideal" types of structure and their physi-cal characterization can be quite meaningful when com-pared with results for heterogeneous, natural soils of a simi-lar type.

This study specifically focused on hydrodynamic disper-sion in different soil materials, by assessing: (i) the effectsof continuous, relatively large vertical pores; (ii) the impactof microfabrics of different textures, and (iii) the effect ofdifferent methods of liquid application in order to evaluatethe degree to which the physical behavior of a soil with agiven structure may be manipulated by changing the methodof liquid application. An attempt was made to expressresults in quantitative terms, for example in terms of appar-ent dispersion coefficients. However, this was not alwaysfeasible because physical theory for dispersion is still beingdeveloped for "nonideal" types of behavior (17).

MATERIALS AND METHODSTwelve model columns 10 cm (4 inches) in diameter and 30 cm

(12 inches) high were prepared from mixtures of silty clay loamand sand (1). The silty clay loam material was obtained from theB2t horizon of a Batavia silt loam (Mollic Hapludalf) and the me-dium sand material was obtained from the B2 horizon of a Plain-field loamy sand (Typic Udipsamment). Two series of six columnseach were constructed. The first series was constructed from amixture of 60% silty clay loam and 40% sand resulting in a mate-rial with a textural classification of clay loam. The second serieswas constructed from an 80% sand, 20% silty clay loam mixturewith a textural classification of sandy loam.

The mixtures were thoroughly blended while dry. They werethen moistened until the mixture was pliable and of a consistencesuch that a column could be formed in a mold which had internaldimensions similar to those of the columns. The columns werethen allowed to air dry for 2 weeks. Following the drying, the col-umns were slowly remoistened. Two columns of each series hadone 0.5-cm diam vertical cylindrical pore drilled through the cen-ter. Three 0.5-cm-diam pores were drilled in a triangle pattern intwo other columns. The final two columns of each series were leftwithout any large cylindrical pores. The columns were then al-lowed to air dry before being permanently placed in plastic cylin-ders. These procedures induced the formation of stable microstruc-tures and of unpuddled surfaces adjacent to the vertical pores (4).Thin sections of the two soil materials are shown in Fig. 1 and 2.The basic fabric of the clay loam mixture in Fig. 1 can be de-scribed as porphyroskelic where soil plasma fills the simple pack-ing voids between skeleton grains. The basic fabric of the sandyloam mixture in Fig. 2 can be described as intertextic where theplasma occurs as bridges between the skeleton grains, leavingpartly open simple packing voids in between. Figures 1 and 2 illus-trate the homogeneous nature of the mixtures and the unpuddledcharacter of the walls of the cylindrical pores.

Moisture retention characteristics of the soil mixtures were de-termined to assess the amount of pore space and moisture contentspresent within the columns. Results are not separately reported,but were used to calculate volumes of water-filled pores in Tables1 through 4. Calculated porosity values in Table 1 are lower thanthose found in natural soils of identical textures, because of thelack of larger pores.

766

BOUMA & ANDERSON: WATER & CHLORIDE MOVEMENT THROUGH SIMULATED PEDAL SOILS 767

Fig. 1—Thin section image of the clay loam soil material, which has ahomogeneous prophyroskelic basic fabric. The walls of the largepore have a microfabric which is not different from that of the s- ma-trix.

Tensiometers were installed at depths of 10 cm (4 in) and 20 cm(8 in) in the columns to monitor the moisture regimes.

Each of the columns without large pores were slowly moistenedto saturation and a 300-ppm KC1 solution was applied to the uppersurfaces of the columns where infiltration occurred. Measurementof the breakthrough curves established patterns of water movementthrough the two matrix materials without the influence of any largecontinuous pores or cracks.

The columns that had large vertically continuous pores drilledthrough them were slowly saturated and were allowed to reachequilibrium moisture contents after drainage as determined fromthe in situ tensiometry. A 0.5-cm daily dose of 300-ppm chloridesolution was distributed evenly over the infiltrative surfaces of thecolumns during a 7-min period. The 7-min interval was arbitrarilychosen because this is the average length of time it takes to evenlydistribute septic tank waste over a soil seepage bed in the field (9)."Short-circuiting" of septic tank effluent along vertical largerpores in pedal soils is a practical problem, and by choosing thisapplication regime, data are generated that have this particularpractical relevance.

Column effluents were monitored in terms of cumulative out-flows and chloride concentrations to determine chloride break-through curves. (Fig. 3-6).

Following the attainment of the concentration of 300 ppm chlo-ride in the column effluent the columns were flushed with deion-ized water.

A 50% gypsum crust was prepared by mixing equal volumes ofdry sand and gypsum. The mixture was moistened to a pliablepaste-like condition. This was applied to the infiltrative surfaces ofthe columns and allowed to set thus creating a slowly permeablebarrier which induces unsaturated subcrust conditions. The 300-ppm chloride solution was continuously ponded to a depth of 0.5cm on the crust. Column effluents were monitored as describedabove.

Chloride concentrations were determined by a Buchler-Cotlovechloride titrator Model no. 4-2008. (10). Results were recorded inboth tabular and graphical form for each column.

Hydrodynamic dispersion coefficients were calculated only forthose situations where one apparent dispersion coefficient woulddescribe the entire displacement process (2, 6, 16). Unsteady un-saturated flow through uncrusted columns with the large verticalpores could not be characterized in terms of a dispersion coeffi-cient by applying dispersion theory. No attempt was made in thisstudy to express this nonideal behavior in terms of coefficientspresented recently (17). However, it is possible to estimate thevolume of "immobile" (untraced) water, left in the column at

Fig. 2—Thin section image of the sandy loam soil material, which hasa homogeneous intertextic basic fabric. The walls of the large porehave a microfabric which is not different from that of the s-matrix.

Ce/Cf= I , with a simple graphical integration technique to beapplied to the breakthrough curve. Any small (Ce/Cf) interval(with a known average Cl concentration) corresponds with a vol-ume of liquid that has left the column. The volume of untracedwater in that volume of liquid can thus easily be estimated and canbe summed up for all small intervals constituting the entire rangeof CJCf from zero to one. The estimated volume of immobilewater at CJCf = 1 (which is zero for "ideal" behavior) can befound by subtracting this sum from the total volume of initiallypresent untraced liquid in the column and can be expressed as apercentage.

RESULTS AND DISCUSSION

Saturated Flow

Saturated flow was studied only in the columns withoutthe continuous vertical pores. This allowed an assessmentof the flow mechanisms of the two microstructures undersaturated conditions. Due to thorough mixing the micro-structures were homogeneous (Fig. 1 and 2, not consideringthe large pores). Lack of the larger macropores allows anunambiguous assessment which often cannot be made in un-disturbed soil samples with heterogeneous microfabrics andvarying numbers of macropores in the form of planes andchannels of unknown continuity. Apparent dispersion coef-ficients were calculated (2,3,6) and are presented in Table1. Breakthrough curves are presented in Fig. 3. Columns 1and 2 (clay loam texture) had low K saturation values andlow dispersion coefficients. In columns 3 and 4 (sandy loamtexture) the K saturation values were larger which is re-flected in the larger dispersion coefficients. As indicated bythe breakthrough curves (Fig. 3) dispersion followed idealbehavior. Excellent agreement between the duplicate mea-surements indicates the reliability of the method used toprepare the columns.

Unsteady, Unsaturated flowThe unsteady, unsaturated flow regime was induced by

applying a daily 0.5-cm dose of chloride solution to drainedcolumns containing untraced water. During this intermittentflow moisture tensions in the column were reduced, but

768 SOIL SCI. SOC. AM. J., VOL. 41, 1977

Table 1—Cumulative outflows and times of chloride appearance in two sets of pedal models without large vertically continuous pores (saturatedflow). Data are included in this and following tables on the textural composition of the columns, K saturation, porosity and calculated D values

(see text).

Column

1234

Sand /siltyclay loam

%40/6040/6080/2080/20

First trace of chloridewater-filled Cumulative

K(sat)

1.231.19

18.1620.00

Porosity%40403232

pores

942942765765

outflow

580560420390

Time

660.290.25

150-ppm chlorideCumulative

outflow

940930740725

Time

10100.540.50

300-ppm chlorideCumulative

outflow

1,6501,6901,3101,295

Time

18190.920.88

D

44

5862

these had reached equilibrium again before addition of thenext dose.

Incomplete displacement of the initially present untracedwater was observed in all columns at the time when columneffluents reached the concentration of the added chloridesolution (CeICf = 1). For example, only 130 cm3 of liquidhad left the clay loam column, with one channel, whereas809 cm3 of untraced liquid was initially present in the col-umn (Table 2). This must imply that pockets with untracedwater (immobile water) remained inside the columns. Thedaily applied chloride solution will flow down the large ver-tical pore or pores, but will also be pulled into the surround-ing soil by (low) lateral gradients of the pressure potentialthereby displacing untraced water. Due to the low perme-ability of the clay loam, the process occurs very slowly andis apparently effective only in a relatively small zone aroundthe vertical channels. The effluent concentration is thereforeequal to the influent concentration long before all untracedwater is displaced. An estimated 94% (one vertical channel)and 92% (3 vertical channels) of the initially present un-traced water is still present in the cores at the time whenCjCf = 1 (graphically derived from Fig. 4 and 5). Thisphenomenon, but less well expressed, was also observed fornatural soils with peds of several centimeters size (2,3), andcontrasts with data on"tailing" in so-called aggregatedsoils with artificial aggregates of 1-2-mm diameter (14). Inaggregates of such small diameter the distances over whichdiffusion must take place are much shorter. This results in amore effective dilution of applied traced water over a longerperiod of time than that found when the size of the soilaggregates is on the order of centimeters rather than milli-meters. Results of dispersion experiments using such smallaggregates may not be representative for many natural pedal

NUMBER OF DISPLACEMENTS OF WATERFILLED PORES

.Fig. 3—Breakthrough curves for two sets of pedal models withoutlarge vertically continuous pores under saturated flow. Columns 1and 2 are constructed of 40% sand and 60% silty clay loam. Col-umns 3 and 4 are constructed of 80% sand and 20% silty clay loam.

soils. Theoretically, it is not possible to reach CjCf = \before all untraced water has been displaced. However,very slow diffusion, as observed, results in very small dilu-tions which are difficult to measure with standard tech-niques and which are of theoretical, but not of practical in-terest.

Flow processes in the more permeable sandy loam col-umns are quite different. The volume of liquid that left thecolumn at the time when Ce = Cf is higher than the volumeof untraced water which was initially present in the column(Tables 2 and 3). However, this does not imply that all ini-tially present untraced water has been displaced from thecolumn. The liquid that leaves the column is composed ofchloride solution flowing (intermittently) through the largerpores and of untraced water which is slowly displaced fromthe soil. An estimated 28% (one vertical channel) and 36%(three vertical channels) of the initially present untracedwater is still present in the cores at the time when CJCf = 1(graphically derived from Fig. 4 and 5). Breakthrough datafor columns with one or three vertical channels were notvery different for each of the two soil materials. The firsttrace of chloride in the column effluents occurs within a fewminutes after the application of the traced water (Tables 2and 3). Columns with three holes allow a somewhat betterlateral absorption of the chloride solution than those withone hole. This is evident from the slightly longer times untilthe appearance of the first trace of chlorides and from thelarger cumulative outflows at the time when CJC{ = 1. Inother words, when three channels are present the chloridesolution is diluted somewhat more efficiently by untracedwater being displaced simultaneously from the column.However, these differences are minor when compared withthe differences present between results for the two soil ma-terials. This can be illustrated by comparing the estimatedvolumes of "immobile" water remaining in the columns atCJCf = 1 for each of the soil materials. The values, whichexpress the effect of the number of vertical channels, showvery little difference between columns with one channel andthose with three channels respectively, in each of the soilmaterials. The average value of 93% "immobile" waterTable 2—Cumulative outflows and times of chloride appearance in

pedal models with one large vertically continuous pore(equal 0.5-cni distribution).

First trace ofchloride 150-ppm chloride 300-ppm chlorideSand/ Volume of

silty Total water-clay pore filled Cumulative Cumulative Cumulative

Column loam volume pores outflow Time outflow Time outflow Time

1234

%

40/6040/6080/2080/20

966966777777

cm3

809809566566

12142424

min

55

1213

cm3

2428

480410

10 min10 min12 days10 days

cm3

120135

1,000965

days33

2524

BOUMA & ANDERSON: WATER & CHLORIDE MOVEMENT THROUGH SIMULATED PEDAL SOILS 769

Table 3—Cumulative outflows and times of chloride appearance inpedal models with three large vertically continuous pores

(equal 0.5-cm distribution).

Column

1234

Sand/

clayloam

%40/6040/6080/2080/20

Volume of

porevolume

984984801801

filledpores

cm3

764764520520

First trace ofchloride

Cumulativioutflow

12131913

150-ppm chloride 300-ppm chloridee Cumulative

Timemin

105

4336

outflowcm3

2223

240160

Time

15 min10 min6 days4 days

Cumulativeoutflow

cm3

200200

1,2301,280

Timedays

55

3132

remaining in the clay loam columns is much larger than theaverage of 32% for the sandy loam columns.

In natural soils, structures with small peds will have moreplanar voids per unit horizontal soil surface than those withlarge peds. Vertical channels provided within each of thecolumns was to simulate this difference in the number ofnatural vertical voids, such as worm or root channels andplanar voids between peds. Results for this experimentshow very small differences. This probably does not reflectconditions which are found in naturally occurring soils sincethe channels provided were all of the same size and continu-ous throughout the sample; generally the channels occurringin natural samples are not continuous (5).

Steady, Unsaturated flowApplication of a 50% sand-gypsum crust induced mea-

sured moisture tensions of 9 ± 1.3 cm in the six clay loamcolumns and of 14 ± 1.7 cm in the six sandy loam columns.This resulted in steady infiltration rates of 0.27 ± 0.06cm/day and 0.38 ± 0.05 cm/day, respectively. The applica-tion of the crust did not allow flow through the large ver-tically continuous pores, resulting in almost identical mois-ture potentials and infiltration rates for the six columns ofeach soil material. Breakthrough data representative of thecrusted columns are presented in Table 4 and Fig. 6. Calcu-lated values of the coefficients of dispersion, which express"ideal behavior" with no "immobile" water at CJCf = 1,are presented in Table 4 for all of the columns. Steady flowrates through the crusts were somewhat < 0.5 cm/day dos-ing rate, used to test the effect of small intermittent flows.This small reduction in the flow rate does not explain thelarge differences in outflows and times, for the columns

NUMBER OF DISPLACEMENTS OF WATERFILLED PORES

Fig. 4—Breakthrough curves for two sets of pedal models with onelarge vertically continuous pore under an intermittent flow of 0.5cm/day. Columns 1 and 2 are constructed of 40% sand and 60% siltyclay loam. Columns 3 and 4 are constructed of 80% sand and 20%silty clay loam.

with macropores before the appearance ot chloride in thecolumn effluent. For example, initial chloride breakthroughin the clay loam columns with three vertically continuouspores occurred after only 7 min following the application ofa 0.5-cm dose (Table 3), but the initial breakthrough is after15 days for the same column with a crust applied to the in-filtrative surface (Table 4 lists corresponding data for col-umns without large pores). Similar, but less extreme, dif-ferences were found for the sandy loam models, wherecorresponding times were 40 min and 8 days, respectively.The large differences can be explained by assuming that thetraced water flows through only the finer pores of the ho-mogeneous matrices with the larger macropores being airfilled. When the 0.5-cm daily dose is applied to an un-crusted soil surface flow into the macropores occurs, result-ing in much earlier breakthrough of chlorides (2,3). Thesedifferences occur, even though moisture tensions at equilib-rium were higher in the uncrusted columns and higher ten-sions are associated with higher lateral capillary forcespulling the water into the soil matrix.

Implications for the Physical Interpretation of StructureDescriptions

A comparison will first be made between data presentedin Tables 1 and 4. Both relate to movement of the tracerthrough "the fine soil pores." The soil material within peds

NUMBER OF DISPLACEMENTS OF WATERFILLED PORES

Fig. 5 —Breakthrough curves for two sets of pedal models with three large vertically continuous pores under an intermittent flow of 0.5 cm/day. Col-umns 1 and 2 are constructed of 40% sand and 60% silty clay loam. Columns 3 and 4 constructed of 80% sand and 20% silty clay loam.

770 SOIL sci. soc. AM. j., VOL. 41, 1977

Table 4—Cumulative outflows and times of chloride appearance in two sets of pedal models without vertically continuous pores (50% gypsum-sandcrust applied to intially drained columns).

... . First trace of chloride 150 ppm chloride 300-ppm chlorideVolume of ———————————— ————————— ——————————Sand/silty Inf. Moist, content Subcrust water-filled Cumulative Cumulative Cumulative D, No D, 1 D, 3

Column clay loam rate___below crust tension___pores____outflow Time outflow Time outflow Time milli pores mill! pore milli pores% cm day"' % cm ———— cm3——— days cm3 days cm3 days ———— cm2 day"1 ————

1 40/60 0.24 37.5 8 883 460 24 850 45 1420 76 <1 <1 12 40/60 0.23 37.0 9 871 500 28 830 46 1480 82 <1 <1 13 80/20 0.34 25.0 15 590 265 10 530 20 1132 42 , 2 2 34 80/20 0.32_____25.0_____15_____590______250 10_____550 22____1170 47______2______2_____6

is characterized morphologically in terms of the basic soil planar voids between peds are part of the structural porositystructure, which is defined as the size, shape, and arrange- (8) or of the secondary structure (7). The conceptually at-ment of the elementary particles and voids (7). Voids which tractive separation of textural and structural porosity mayare part of the basic structure are considered textural poros- have less physical meaning than is often implied since theity in soil physical literature (8). Movement through "the two seem irrevocably tied together (13).fine soil pores" occurs here either because of lack of macro-pores (Table 1, Fig. 3) or because of the presence of a crust Acknowledgmentwhich did not allow flow through the macropores that were The authors acknowledge the preparation of thin sections in the labora-present (Table 4, Fig. 6). Each of the two microstructures tory of micropedology of the Soil Survey Institute, Wageningen, Thehad a significantly different behavior in terms of dispersion Netherlands (A. Jongerius, Head),as a result of the two flow regimes. This illustrates that aspecific, well-defined textural porosity or basic soil struc-ture still may exhibit a wide range of physical behavior, as afunction of different physical boundary conditions. Thismay seem obvious. However, distinction of "fine" texturalpores (basic soil structure) often seems to suggest a ratherdistinct, and as such rather constant, level of physical be-havior.

The effect of the large continuous pores was very appar-ent with the application of a small intermittent dose. In theclay loam columns, flow was primarily through the macro-pores, bypassing a large portion of the homogeneous soilmatrix. To a lesser extent this was also true in the sandyloam columns. Soil morphological studies emphasize thedescription of larger soil pores as individuals in terms ofsize, shape, and arrangement (7). This study has demon-strated that identical large pores have a different hydraulicfunction in different types of soil materials, due to the dif-ferent hydraulic behavior of the surrounding soil matrix.This conclusion is based on an analysis of simulated smallintermittent flows, occurring in reality as rainfall, wasteapplication, or irrigation, and does not, of course, apply tosaturated flow where large vertically continuous pores willconduct very high quantities of liquid in all soil materials.Macropores such as tubular root and worm channels and

20————————————————————/^^^"

\ /^^i- ^^o = JFI S j&§ a as <#u S /%3 S // Column 1 ———9— ,<// Column 2 ........£ . ,/?/ Column 3 ———X t,y // Column 4 ———

°i————'———"&'——'———ft————^————fr————'————JoNUMBER OF DISPLACEMENTS OF WATERFILLED PORES

Fig. 6 —Breakthrough curves for two sets of pedal models without ver-tically continuous pores after application of a 50% gypsum-sandcrust to the infiltrative surfaces.