Soil Solution Chemistry: Effects of Bulking Soil SamplesR. Giesler* and U. Lundstrom

ABSTRACTBulking of soil samples is commonly used to obtain a spatially av-

eraged measure of soil properties. This study investigated whetherbulking can affect the chemistry of soil solution extracted by the cen-trifuge drainage technique. The short-range variation within a 1-m-wide pit face was also investigated. Systematic differences were foundbetween the chemistry of soil solutions taken from bulked and un-bulked soil samples. The soil samples studied were taken on severaloccasions from the upper and lower part of a Bs horizon. Samples ofsoil solution extracted from replicate bulked soil samples containedsignificantly (P < 0.05) larger amounts of Si, Na, Mg, and K as wellas significantly (P < 0.05) smaller amounts of Al and Fe than soilsolution from unbulked samples. Bulking reduced the variation in allsolutes measured, although a variation remained in replicates fromthe bulked sample. The variation between individual core samplestaken just a few tens of centimeters apart can be considerable, al-though the magnitude of variation differs depending on the solutestudied. The results suggest that unbulked soil samples may be pref-erable, especially when Al or Fe chemistry is studied and a centrifugedrainage technique is used.

IN RECENT YEARS there has been increasing interestin methods for obtaining soil solution directly from

soil samples. Different techniques have been used,many based on destructive sampling of soil samplestaken from the field. The approach whereby the soilsolution is removed directly from a soil sample offersan opportunity to study the soil solution chemistry insamples that are well defined in time and space. Thisis especially attractive when the soil solution has to

Swedish Univ. of Agricultural Sciences, Dep. of Forest Ecology,Soil Science Section, S-901 83 Umea, Sweden. Contribution fromSwedish Environmental Protection Agency, contract no. Dnr 802-761-91-Fs. Received 26 May 1992. * Corresponding author.

Published in Soil Sci. Soc. Am. J. 57:1283-1288 (1993).

be related to a particular soil horizon or when seasonalchanges are to be studied (Zabowski and Ugolini, 1990;Linehan et al., 1989; Beyme et al., 1987; Manley etal., 1987). Thus, when the purpose of an investigationis to study the soil solution under conditions as closeto natural as possible, consideration has to be givento the choice of extraction technique. Extraction withimmiscible displacement involves the risk of adulter-ating the soil solution (Elkhatib et al., 1987). Thetreatment of the soil sample prior to the extraction isalso important. Ross and Bartlett (1990) and Ed-meades et al. (1985) concluded that freshly collectedfield-moist samples should be extracted within 24 hin order to avoid storage effects. Thus all sample han-dling that involves air drying and rewetting should beavoided due to drastic changes in soil solution chem-istry (Bartlett and James, 1980). The centrifuge drain-age technique, where fresh, field-moist samples areused, seems to be the most desirable when the purposeis to collect soil solutions as close to natural conditionsas possible.

In contrast to lysimeter techniques, however, it isimpossible to resample precisely the same spot on suc-cessive sampling visits, since the soil sampling is de-structive. Short-range spatial variation among varioussoil parameters can be considerable (Skyllberg, 1991;Campbell et al., 1989; Nykvist and Skyllberg, 1989;Alberts et al., 1977). Therefore, knowledge of thevariability is essential when deciding the sample num-ber required, especially for temporal studies.

A widely used sampling method is to bulk severalindividual soil cores by mixing them together into acomposite sample. This is an attractive method forovercoming the problem of spatial variability that avoidsthe expense of analyzing numerous replicates (Mrozand Reed, 1991). However, bulking soil samples also

1284 SOIL SCI. SOC. AM. J., VOL. 57, SEPTEMBER-OCTOBER 1993

Table 1. Morphological description of the soil horizons.Depth Horizon Description-4-0 6 fibric

0-1 E light gray (2.SY 7/0 moist), loamy fine sand,boundary abrupt and irregular to Bs

1-30 Bs reddish brown (SYR 6/1 moist), loamy fine sand,moderately weak medium crumb structure,diffuse transition horizon between Bs andC horizon

30-60 B/C transition horizon with diffuse boundaries,gradually changing from Bs to C characteristics

60 + C light brownish gray (10YR 6/2 moist), loamy finesand, structureless, consolidated single grain

involves disturbance of the soil matrix. This entails arearrangement of soil particles, a redistribution of thestructure-dependent pore size distribution, and abreakage of aggregated structures. If there is hetero-geneity in the soil solution composition that dependson the pores size, disturbance of the soil matrix couldaffect the soil solution composition. Bulking mightalso generate new equilibrium conditions that can af-fect the solute concentration (Hantschel et al., 1988).

The objectives of this study were to: (i) comparethe soil solution chemistry of bulked and unbulkedsoil samples, (ii) determine the effect of bulking onthe short-range spatial variability, and (iii) determinethe magnitude of short-range spatial variation in soilsolutions obtained from individual soil samples takenfrom the face of a pit.

MATERIALS AND METHODSSite Description

The study area is located in a 70-yr-old Scotch pine (Pinussylvestris L.) forest within the Svartberget Forest ResearchPark (62°14'N, 19°46'E, 225 m above sea level, 55 km north-west of UmeS, in the northern part of Sweden. The meanannual air temperature is 1.1 °C and annual precipitation is593 mm, of which about 40% occurs as snow (Degermark,1992).

The site is dominated by Scotch pine with Norway spruce[Picea abies (L.) Karsten] as an associated species. The fieldlayer consists mainly of dwarf shrubs, Vaccinium myrtillus L.and V. vitis-idaea L., while mosses, Pleurozium schreberi andHylocomium splendens, dominate the bottom layer.

The soil is a Spodic Cryopsamment of mixed mineralogyformed on glaciofluvial sediment. A morphological descriptionis presented in Table 1. The texture in the soil profile is mostlyhomogenous throughout, consisting of approximately 82% finesand, 15% silt, and 3% clay. Soil samples were taken fromthe upper 5 cm of the Bs horizon and the lower part of the Bshorizon (30 cm below the mineral soil surface, measured fromthe center of the sampling device). The E horizon was gen-erally < 1 cm thick where samples were taken.

SamplingIn May 1990, soil samples were collected from three small

pits (1.0 by 1.0 m). The distance between each pit was 10 m.The soil samples were taken from the same face of each pitand the pits were refilled after each sampling. In July andOctober 1990, the sampling was repeated for one pit and newsamples were taken after reexcavation and enlargement of theoriginal pit >20 cm past the previous sampling face. The ex-pansion and sampling of the pit proceeded in the same directionon subsequent sampling occasions.

Individual soil samples were taken by driving a cylindrical

20mm

Soil coreand

sampling device

- Soil cylinderDrilled holes(0 0.5 mm )

Collection cup

Fig. 1. Sampling scheme for unbulked and bulked soil coreson the pit face.

sampling device 6 cm perpendicularly into the soil face. Thesampling device was made of a polyvinyl chloride tube withan inner diameter of 4.6 cm and sealed with a tight-fittingpolyethylene lid at one end. Each sample core contained ap-proximately 100 cm3 of soil. The sampling device was madeto fit into the soil cylinder of the centrifuge assembly, thusenabling centrifugation of undisturbed soil samples. Twentysoil samples were taken at each of two levels (upper and lowerBs horizon), 3 and 30 cm below the surface of the mineralsoil (Fig. 1). Every second core was bulked into one mixedsample, containing about 1000 cm3 of soil in total. The bulkedsamples were immediately sealed in 3-L polyethylene bags.The remaining cores were kept separately in the sampling de-vices, which were sealed at the open end with the same typeof tight-fitting polyethylene cap as the one at the closed end.All samples were kept at 4 °C until centrifugation within 32h.

CentrifugationA modification of the centrifuge technique described by

Wyndham and Bath (1976) was used. The centrifuge assembly(Fig. 2) is constructed of Delrin (polyamide plastic). Delrin iseasily workable and makes it possible to construct the centri-fuge assembly without adhesives. The material is also durableenough to withstand the high forces that are applied. The soilsampling device described above fits into the soil cylinder andenables centrifugation of undisturbed soil cores as well as bulkedsamples. A Beckman J/E 21 centrifuge with a JA 14 rotor wasused. Samples were centrifuged at 14 000 rpm for 80 min. Theaverage relative centrifugal field is 16 500 x g (r = 75 mm).A constant temperature of 5 °C was maintained during centri-fugation.

The bulked sample was split into 10 replicate samples (each

-1 meter

10cm j

O single core samples• bulked core samples

Fig. 2. Centrifuge assembly for extracting soil solution.

GIESLER & LUNDSTROM: BULKING EFFECTS ON SOIL SOLUTION CHEMISTRY 1285

sample containing approximately 100 cm3 of soil) after mixing.The mixing prior to subsampling was done by turning over theplastic bag with the soil sample 10 times. Each subsample wascentrifuged, and the soil solution obtained from each subsam-ple was kept separate prior to analysis. The 10 separate andundisturbed cores were placed in the centrifuge cylinder whilestill within the original sampling device, and the soil solutionobtained from each soil sample was kept separate. On eachsampling occasion, all sampling was performed on the sameday.

Immediately after centrifugation, pH was determined on aportion of each sample extracted, and another portion for in-ductively coupled plasma analysis was acidified with concen-trated HNO3.

AnalysesThe soil solutions were analyzed for Si, Fe, Al, Mg, Na,

Ca, and K using inductively coupled plasma-atomic emissionspectrophotometry (Perkin-Elmer Plasma II emission spec-trometer, Norwalk, CT). A ion chromatograph (Dionex model4000J, Dionex Corp., Sunnyvale, CA) was used to determineCl, F, PO4, NO3, and SO4 concentrations. Insufficient yieldsof soil solution limited the possibilities for analyzing anionsexcept during the May sampling. Soil solution pH was mea-sured with an Orion Research Model 601/digital lonalyser andan Orion 8103 SC Ross combination electrode (Orion Re-search, Cambridge, MA).

StatisticsFor statistical comparisons between bulked and unbulked

samples, Student's f-test with unpooled t values was used, dueto unequal sample variance (Zar, 1984). To test for differencesbetween two regression lines, we have followed the procedureof Zar (1984). Significant differences refer to the 0.05 levelunless otherwise stated.

RESULTSCations and Silica

Bulking affected the soil solution composition. Themean concentration value of replicates from a compositesample was higher for Si on all occasions compared withthe mean of the undisturbed single core samples, al-though the divergence was not always significant (Fig.3). For Al and Fe, bulking yielded significantly lowersoil solution concentrations on the May and Octobersampling occasions (Fig. 3). Soil solution from singlesoil cores yielded mean concentrations of Al and Fe thatwere up to three times higher than the concentrations inthe bulked samples (Fig. 3). The concentrations of Aland Fe in the soil solution from the 30-cm depth (notshown) were close to the detection limits, but the resultsfollowed the pattern seen at the 3-cm depth.

The pH was lower in the soil solution from bulkedsamples. This was most pronounced at the spring sam-pling (Fig. 3). The maximum difference was about 0.5pH units (May sampling, 3-cm depth, Pit 1).

Bulking also affected Mg, Na, Ca, and K. The trend,however, was not as evident as for Al and Fe. For ex-ample, mean Ca concentrations from single core sampleswere significantly higher than bulked samples in somecases, but the opposite was also found (Fig. 4). On theother hand, solutes such as Mg, Na, and K tended tohave higher mean concentrations with bulking. This wasmost obvious for Na at the 30-cm depth, where bulkingproduced significantly higher mean concentrations in four

Si 3cm

bulked D unbulkedSi 30 cm

pit 1 pit 2 pit 3 pit 1 pit 1 " pit 1 pit 2 pit 3 pit 1 pit 1May July October May July October

Fig. 3. Mean and standard deviation for pH and Si, Al, andFe concentrations in soil solution from bulked and unbulkedsamples. Significant differences are indicated for the 0.05(*), 0.01 (**), and 0.001 (***) significance levels. Note thatthe concentration scales are not the same for 3 and 30 cm.

cases out of five (Fig. 4). Soil solutions from the Julyand October resampling did not exhibit significant dif-ferences between bulked and unbulked samples for Ca,Mg, and K (Fig. 4).

Differences in the solute concentration from single corestaken just a few tens of centimeters apart at the samedepth were very large. For example, the Al concentrationin the soil solution from single core samples ranged from

<ccc

UJOoO

100-

Na 200-

100-

Na

60-

pit 1 pit 2 pit 3 pit 1 pitlMay July October

D unbulked0 bulked

pit 1 pit 2 pit 3 pit 1 pit 1May July October

Fig. 4. Mean and standard deviation for Ca, Mg, Na, and Kconcentrations in soil solution from bulked and unbulkedsamples. Significant differences are indicated for the 0.05(*), 0.01 (**), and 0.001 (***) significance levels. Note thatthe concentration scales are not the same for 3 and 30 cm.

1286 SOIL SCI. SOC. AM. J., VOL. 57, SEPTEMBER-OCTOBER 1993

300

-150

Si

p 100! Al

tr so

oZ 40! Fe

O 20

" 0 0 . 5 1 0 0 . 5 1 0 0 . 5 1pit 1 pit 2 pit 3

DISTANCE (meters)Fig. 5. Spatial variation in soil solution concentrations for Si,

Al, and Fe at the 3-cm depth; May 1990 samples.

20 to 98 nM within 30 cm in Pit 3 (Fig. 5). Similarobservations were found for Fe and Si (Fig. 5) and theother cations but not always as pronounced as for Al.

Bulking and mixing single core samples into a com-posite sample reduced the variation within the compositerelative to the short-range spatial variability between sin-gle core samples (Table 2). The coefficient of variationwas determined for each pit and sampling occasion, andan average for the five values was calculated. The av-erage coefficient of variation ranged from 18 to 90% inunbulked samples and from 4 to 26% in bulked samples(excluding pH). Thus, the average coefficient of varia-tion was reduced in bulked compared with unbulkedsamples by approximately a factor of 0.3.

Table 3. Comparison of Cl and SO4 concentrations in soilsolution from bulked and unbulked samples.

Depthcm

3

30

Pit

2323

ClUnbulked

253 ± 90t161 ± 42*»*44 ± 19*62 ± 13***

SO4

Bulked Unbulked———— pM ———————264 ± 4 2 48 ± 16233 ±18 45 ± 9***

74 ± 11 5 ± 497 ± 14 7 ± 5

Bulked

37 ± 268 ± 56 ± 36 ± 2

*,**,*** Significant at the 0.05, 0.01, and 0.001 probability levels,respectively.t Mean ± standard deviation.

The relationship between cations and anions was ex-amined in both bulked and unbulked samples for the May1990 sampling at 3-cm depth. The sum of cations in-cluded H, Ca, Mg, Na, K, Al, and Fe, whereas anionsincluded SO4 and Cl. It was assumed that Al and Fe eachcontributed a (3 + ) charge. The difference between thesum of cations and the anions is the anion charge deficitand was assumed to be due to dissociated and complexedorganic ligands.

The anion charge deficit was pronounced for bothbulked and unbulked samples. A simple linear regressionbetween the sum of cations and the anions for the bulkedand unbulked samples did not give significantly differentslopes for the regression lines. However, the elevationsof the lines were significantly different (P < 0.001),indicating a decrease in the anion charge deficit of about78 mmolc m~3 (Fig. 6) with bulking. The decrease inanion charge deficit corresponded with an average de-crease for Al and Fe in the bulked samples of about 79mmolc m~3.

AnionsThe difference between the mean concentration of an-

ions from bulked and unbulked samples was relativelylarge compared with the other solutes examined. The Clconcentrations were higher in the bulked samples, andthe difference was significant in three out of four cases(Table 3). A significant difference in SO4 concentrationswas found in only one case (Table 3) where the bulkedsamples gave a higher mean concentration. Chloride andSO4 dominated the anions analyzed, whereas PO4, F,and NO? were close to the detection limit. The degreeof variation in Cl and SO4 concentrations within the bulkedsample and between single core samples followed thepattern observed for the cations.

Table 2. Average (n = 5) coefficient of variation from fivesamplings for bulked and unbulked samples.

Solute

SiAlFeCaMgNaKPH

3 cmUnbulked

294890533131436.9

Bulked————— °,

41026188

10121.2

30cmUnbulked

fc —————————18

404122

3.6

Bulked

4

188

15

1.6

Soil WaterThe mean water content and yields from bulked and

unbulked samples did not differ significantly (Fig. 7).Bulking and mixing single core samples reduced the var-iation in water content considerably, but the mean water

600

CO

E

~0 400E

CO

o<o

200

—— Y.0.96X +112.11 r2=0.63----Y=1.00X +34.02 r2 = 0.96

©

O

O

O single core samples• bulked core samles

200 400 600

ANIONS (mmolcFig. 6. Sum of cations (Na, K, Mg, Ca, H, Al, and Fe) vs.

sum of anions (SO4 and Cl) for bulked samples and singlecore samples; May 1990 samples.

GIESLER & LUNDSTROM: BULKING EFFECTS ON SOIL SOLUTION CHEMISTRY 1287

content was still the same as the mean of the single coresamples. The variation between single core samples wasless at the 30-cm depth than at the 3-cm depth.

DISCUSSIONBulking soil samples involves destruction of the nat-

ural soil structure. Larger pores will be disrupted, andthere will be some degree of reorganization of soil par-ticles. Aggregated structures might also be broken. Phys-ically, a change in the pore size distribution will changethe water-retention properties of the soil. This might af-fect the soil solution yields. When high-speed centrifu-gation is used, however, the applied tension can reachvalues of > 3000 kPa. At such tensions, the soil structureis of less importance for water retention (Hillel, 1980).Instead, it is the texture of the soil sample that deter-mines the soil solution yield. This is indicated by ourresults, where the yields were similar for both bulkedand unbulked samples. It may be assumed, however, thatbulking has involved at least some degree of mixing be-tween the soil solution extracted and the soil solutionstill retained in the soil sample after centrifugation. Thismight contribute to a difference in the origin of the soilwater due to the sample handling, even though the yieldswere the same.

Several processes might contribute to the differencesbetween the chemical composition of the bulked and un-bulked soil solution samples. Preferential flow pathwaysand nonequilibrium conditions in the chemical transferbetween different pore classes may affect the soil solu-tion composition in different pore classes (Luxmoore etal., 1990; Cozzarelli et al., 1987). In a bulked sample,where the soil solution is homogenized, these differenceswill probably have diminished. Considering that the or-igin of the soil solution may differ depending on samplehandling, this might influence the soil solution compo-sition in the extract. The reorganization of the soil par-ticles and a breakage of aggregated structures may alsoexpose mineral surface areas and ion-exchange sites tothe soil solution different from those under undisturbedconditions (Hantschel et al., 1988; Jardine et al., 1988).Most likely bulking and mixing single samples into acomposite sample will create new combinations of thepore water composition and pore surfaces. This mightgenerate new equilibrium conditions that can affect thesoil solution concentrations.

Our results indicate a simultaneous reduction of Al

Remaining soilsolution aftercentrifugation Yielded soil solution

\unbulked [™

bulked

unbulked

bulked

depth 3 cm

i — 1depth 30 cm

20 40-1WATER CONTENT (kg kg 'x100)

Fig. 7. Total and remaining water content after centrifugation,for bulked and unbulked samples, mean and standarddeviation; October 1990 samples.

and Fe concentrations and the anion charge deficit in soilsolution from bulked samples. Aluminum and Fe appearmainly in organically complexed forms in the upper min-eral part of podzols (Berggren, 1992; Lundstrom, 1990;Dahlgren and Ugolini, 1989; Manley et al., 1987).Lundstrom (1990) demonstrated in lysimeter studies fromthe Svartberget site (where our study was performed) that> 80% of the Al present in the percolate from the upper10 cm of the mineral soil was in the monomeric, organ-ically complexed form. Given that the B horizon is aneffective arrest for dissolved organic C as well as for Aland Fe (Berggren, 1992; Dahlgren and Ugolini, 1989;McDowell and Wood, 1984) it might be hypothesizedthat the same precipitation-adsorption processes that nat-urally occur in the soil might also contribute to the re-duction of Al and Fe in the bulked samples. This wouldalso explain the simultaneous reduction of the anion chargedeficit, assuming that the major contributors to that def-icit are complexed and dissolved organic anions.

The contact time between the soil and the soil solutionis an important factor in all of the above processes. Itmay be assumed that equilibrium conditions are morelikely during periods with low flow and long contact timebetween soil and soil solution. This may explain the less-pronounced effects of bulking in the soil solution com-position obtained from the soil sampled during low-flowconditions (i.e., the July sampling occasion). Residencetimes of water in the soil were shorter during the May(snowmelt period) and October (fall rains) sampling oc-casions. The effects of sample handling were also morepronounced at those times.

Large spatial variation in various soil parameters iswell known from several studies (Litaor, 1988). Not sur-prisingly, our investigation also found a considerable short-range spatial variation. The solute concentrations in soilsolution varied significantly in single core samples col-lected just a few tens of centimeters from each other.Mean soil solution concentration values and the short-range spatial variation may also differ considerably within10 m, despite efforts to choose similar sampling sites.The results raise the question of how many replicatesamples have to be taken to provide a representativespatial mean value. In this study, 10 replicate sampleswere sufficient in most cases to detect significant differ-ences at the 0.05 level when comparing the differentsampling occasions for Pit 1. Nonetheless, the differ-ences between the three pits at the May sampling ex-ceeded those found between the three sampling occasionsin several cases. Similar findings were reported byCampbell et al. (1989), who used a centrifuge techniquewith an immiscible liquid.

Although no general conclusions on sample size canbe made, it is obvious that knowledge of the spatial var-iability is essential when deciding on sampling scheme.Several researchers (Zabowski and Ugolini, 1990;Campbell et al., 1989; Zabowski, 1989; Beyme et al.,1987; Manley et al., 1987; Edmeades et al., 1985)have used different centrifuge techniques to study thesoil solution composition in soil profiles. However, inonly one case (Campbell et al., 1989) was the spatialvariability discussed. Zabowski and Ugolini (1990)studied temporal trends in a podzolic profile, usingduplicate samples from the face of a pit on each sam-pling occasion. According to our results, this might be

1288 SOIL SCI. SOC. AM. J., VOL. 57, SEPTEMBER-OCTOBER 1993

too few samples for attempting to draw conclusions ontemporal trends.

CONCLUSIONSTwo major conclusions can be drawn from this study.

Bulking prior to centrifugation did not give reliable es-timates of the areal mean, especially for dissolved Si,Al, and Fe, when compared with the mean concentra-tions extracted from undisturbed soil cores. Thus, it ispreferable to extract the soil solution from undisturbedsoil cores. If the cost of chemical analyses necessitatesbulking, this should be done by mixing the soil solutionextracted from undisturbed soil cores rather than by mix-ing the soil prior to extraction.

There was considerable short-range variation in thesoil solution chemistry, which must be known when de-termining the number of replicate samples to be taken ina field study, especially when time trends are to be stud-ied.

ACKNOWLEDGMENTSWe thank Professor Nils Nykvist, Drs. Harald Grip, Kevin

Bishop, Mikael Olsson, Mats Nilsson, and Ulf Skyllberg forvaluable advice and for their scientific comments on the man-uscript, and Inger Bergman for her assistance in the field andin the laboratory. This work is financially supported by theSwedish Environmental Protection Agency.