Effect of Tillage Practices on Infiltration and Soil Strength of a Typic Hapludult Soil AfterTen Years

D. E. RADCLIFFE,* E. W. TOLLNER, W. L. HARGROVE, R. L. CLARK, AND M. H. GOLABI

ABSTRACTHighly weathered soils of the Southeastern USA are poorly struc-

tured and may present special problems under continuous no-tillageproduction. Soil physical properties were examined in a long-termtillage experiment starting in its 10th year to determine if there weredifferences due to tillage. Fall/spring tillage treatments consisted ofmoldboard plow/moldboard plow (CT), moldboard plow/no-tillage(MT), and no-tillage/no-tillage (NT). Cone index measurements inNT exceeded 4 MPa at a depth of 0.10 to 0.20 m, indicating thepresence of a compacted zone. Bulk density was significantly higherthan CT at this depth also (1.60 vs. 1.40 Mg m-'). Spring plantingtraffic compacted the top 0.15 m in all treatments. Infiltration rates,measured with a sprinkler infiltrometer, were significantly higher inNT. The straw mulch (5000 kg ha ') and layer of fine organic litterat the surface of NT prevented the formation of an impermeablesurface crust. Removing the mulch and litter layer from NT sharplyreduced the infiltration rate. Adding a mulch to CT increased theinfiltration rate. We conclude that during short-term summer rain-fall events, infiltration in conventionally tilled soil is controlled bysurface crusting.

Additional Index Words: Bulk density, Cone index, Conservationtillage, Crusting, Infiltration, Macroporosity, Mechanical imped-ance, No-tillage, Reduced tillage.

NO-TILLAGE SYSTEMS offer significant advantagesover conventional tillage in soil, water, and en-

ergy conservation. There is concern, however, thatlong-term continuous no-tillage may lead to high soilstrength due to traffic compaction (Voorhees andLindstrom, 1983).

Long-term tillage experiments have been conductedto determine the effect of continuous no-tillage on soilphysical properties. In the Mid Atlantic USA, therehas been little evidence of any tillage effect. After 6 yrof continuous no-tillage corn [Zea mays (L.)] in Vir-ginia, bulk density at 0.10- to 0.12- and 0.40- to 0.42-m soil depth was not significantly different from thatof conventional tillage (Shear and Moschler, 1969). Ina similar experiment on an Alfisol in Kentucky, therewere no differences in bulk densities at 0.00- to 0.08-m soil depth between no-tillage and conventionaltreatments after 5 yr (Blevins et al., 1977).

Long-term experiments in the Midwest USA, on theother hand, have shown that compaction can be aproblem. Hill and Cruse (1985) reported that in a 7-yr experiment in Minnesota on a Mollisol, bulk den-sity tended to be higher under no-tillage at a depth of0.10- to 0.12-m. Soil strength, measured with a fallingcone penetrometer was significantly higher in no-til-lage plots in the 0.05- to 0.13-m soil depth. In a secondexperiment in Minnesota, after 10 yr of continuousno-tillage on a Mollisol, bulk densities were not sig-nificantly different, but hand-held penetrometer mea-surements showed higher soil strength in no-tillageabove a depth of 0.40 m (Bauder et al., 1981).

D.E. Radcliffe, and R.L. Clark, Dep. of Agronomy, Univ. of Geor-gia, Athens, GA 30602; E.W. Tollner, W.L. Hargrove and M.H.Golabi, Ga. Agric. Exp. Stn., Griffin, GA 30212. Contribution fromthe Univ. of Georgia Agric. Exp. Stn. Supported by State and Hatchfunds allocated to the Georgia Agric. Exp. Stn. Received 12 May1987. Corresponding author.

Published in Soil Sci. Soc. Am. J. 52:798-804 (1988).

RADCLIFFE ET AL.: EFFECT OF TILLAGE PRACTICES ON INFILTRATION AND SOIL STRENGTH 799

Highly weathered Ultisols of the Southeast USA maypresent special problems for continuous no-tillagepractice. Because of the low organic matter content,coarse texture, and generally poor structure, these soilscan be susceptible to compaction. The predominanceof 1:1 nonexpanding clays and the shallow depth ofwinter freezing limit natural alleviation of compactedzones. Several researchers in the region have reportedreduced yield under no-tillage systems (Gallaher, 1984;Hargrove and Hardcastle, 1984; Thurlow et al, 1984;Touchton and Johnson, 1982). Nelson et al. (1977),however, found that yield was higher under no-tillageaveraged over 3 years, and attributed this to betterinfiltration and less evaporation. Hargrove (1985) alsoreported higher yields with no-tillage corn, and showedwater contents after irrigation that suggested higherinfiltration in no-tillage compared to conventional til-lage. Langdale et al. (1983), using simulated rainfall,found that run-off was much lower in no-tillage within-row chiseling than in conventional tillage.

Tollner et al. (1984) described soil physical prop-erties after 6 yr in a long-term tillage experiment inGeorgia. Infiltrability appeared to be higher under no-tillage in that surface-applied Ch leached to a deeperdepth than in the conventional treatment. No-tillageplots had a higher bulk density at the 0.00- to 0.15-msoil depth, but conventional tillage was higher at 0.15to 0.30 m. Soil strength, measured with a hand-heldpenetrometer was higher in no-tillage and reached amaximum in the 0.15- to 0.25-m zone of soil depth.

The experiment described by Tollner et al. (1984)had been conducted for 10 yr in 1985. Our objectivein this paper was to determine if long-term continuousno-tillage significantly altered the soil physical prop-erties of a Hapludult. Chemical properties and cropperformance will be described in a subsequent paper.

METHODS AND MATERIALSThe long-term randomized complete block experiment

described by Tollner et al. (1984) was established in the fallof 1975 in Pike County, GA (Lat 33° 10'N and Long 84°20'W) and consisted of three tillage treatments, replicatedfour times, in a winter wheat (Triticum aestivum L.) andsoybean (Glycine max L. Merrill) double-crop system. Thefall/spring tillage treatments were as follows: moldboard plowand disk/moldboard plow and disk (CT, conventional til-lage), moldboard plow and disk/no-tillage (MT, minimumtillage), and no-tillage/no-tillage (NT, no-tillage). The mold-board plow and disk narrow had draft depths of about 0.25and 0.10 m, respectively. Wheat was planted each fall witha standard grain drill (0.18 m row width) and all soybeanswere planted in the spring with a Cole no-till planter (0.76-m row width). Tillage treatments were repeated in the sameplots each year. The soil series is a Cecil sandy clay loam(clayey, kaolinitic, thermic Typic Hapludult). The test areahad been in tall fescue (Festuca arundinacea L. Schreb.) sodfor an undetermined number of years before this study wasestablished.

In 1985 soybeans were planted on 14 June. On 18 and 19July, cone index was measured using a tractor-mounted, hy-draulically driven cone penetrometer (Clark and Reid, 1984).The base diameter of the cone was 20 mm and the anglewas 30°. The instrument measured mean cone index in 0.025-m depth increments to 0.60 m below the soil surface. Thesampling scheme for each plot consisted of three transects1.52 m in length, perpendicular to, and crossing, two soy-

Soybean Rows

Spring Planting.Traffic

— 0.76m—-o

1 o penetrometer 'measurements

Fig. 1. Sampling scheme for cone index measured on July 18-19,1985.

bean rows (Fig. 1). Five measurements were made alongeach transect at 0.38-m intervals starting where the tractorwheel passed during planting in the spring of 1985. Tran-sects were separated by 0.5 to 3.0 m. A total of 15 mea-surements were made in each plot consisting of six in rowmeasurements, six measurements in traffic inter-rows, andthree measurements in nontraffic inter-rows. At the sametime, soil water content was measured gravimetrically, sam-pling randomly from inter-row and in-row positions, in 0.15-m depth increments to 0.75 m below the surface.

Four soil cores (54-mm in diam and 60-mm in length)were taken from each plot in the row in December 1985,centered on a depth of 0.20 m. Saturated conductivities weredetermined for each core using tap water and the constanthead method (Klute and Dirksen, 1986). Soil moisture char-acteristic curves were determined using a pressure mem-brane apparatus and air pressures of 0.01, 0.05, 0.075, and0.100 MPa. Pore size distributions were calculated from thecharacteristic curves (Vomocil, 1965). The cores were oven-dried (105 °C) and bulk density and total porosity (assuminga particle density of 2.65 Mg m-3) were calculated. Particlesize analysis was determined by the hydrometer method(Anonymous, 1972) on samples collected in January 1986from the 0.00- to 0.15- and 0.15- to 0.30-m depths in eachplot.

Infiltration measurements were conducted during thesummer of 1987 using a sprinkler infiltrometer similar indesign to that described by Peterson and Bubenzer (1986).A square metal enclosure with dimensions the same as rowwidth (0.76 m) was used. Water was applied for 1 h and anyrunoff that accumulated at the downslope end of the enclo-sure was pumped off and the amount recorded during al-ternate minutes. One set of measurements were made duringthe period 9 to 16 July, 3 weeks after planting soybeans. Asingle measurement was made in each of the four replicatedplots of the CT and NT treatments. A position along a rowwas selected randomly and the enclosure was placed so thatit straddled the row. With this arrangement, half of the areaenclosed was from a nontraffic inter-row, and half from atraffic inter-row. The soybean plants, which were approxi-mately 0.25 m high at this stage, were not removed. Carewas taken to not disturb the soil surface within the enclo-sure. A sprinkling rate of 42 mm h ' was used in this set ofmeasurements. Infiltration rate was expressed as a percentof the sprinkling rate.

A second set of measurements were conducted during theperiod 11 to 20 August. The enclosure was placed entirelyin the nontraffic inter-row area in these runs, so that plantswere excluded from the enclosure. Two measurements were

800 SOIL SCI. SOC. AM. J., VOL. 52, 1988

made in each CT and NT plot, one with a straw cover andone without. Before these measurements, the surface soilwas carefully raked to destroy any crust that might havebeen present. In the NT plots, for the covered run, the strawwas removed by hand, the surface was raked, and the strawreturned. For the uncovered run in NT, the straw was re-moved along with the top 20 mm of soil and fine organiclitter so that the mineral soil was exposed. The surface wasthen raked. In the CT plots, the uncovered run was accom-plished after raking the surface. For the covered run, thesurface was raked and approximately 300 g of straw (5194kg ha~')> gathered by hand from adjacent NT plots, wasplaced on the surface. A sprinkling rate of 70 mm h~' wasused in the second set of measurements.

In the statistical analysis of the penetrometer data, a ran-domized complete block strip-strip design was used (Gomezand Gomez, 1984) with tillage treatments as whole plots,positions as strip plots, and depths within positions as strip-strip plots. Mean cone index values for different tillages ata given position were compared at each depth by computinga LSD for each position. The infiltration data was analyzedusing a randomized complete block split-split design withtillages as main plots, cover (with or without straw mulch)as split plots, and time as split-split plots (Gomez and Gomez,1984).

RESULTSGravimetric water contents of soil samples taken at

the time cone index was measured are shown in Fig.2. Water contents increased with depth reflecting thetransition from a sandy clay loam Ap horizon (0.00-0.20 m), through a sandy clay loam Btl horizon (0.20-0.37 m), to a clay Bt2 horizon (Perkins et al., 1985).Within each depth, differences between tillage treat-ments were not statistically significant (0.05 level), butwater content tended to be higher in NT and MT,compared to CT.

Soil Strength and DensityStatistical analysis of the penetrometer data was

confined to the 0.00- to 0.40-m depth because it seemedunlikely that tillage or traffic would have an effect oncone index below 0.40 m and including data for thelower depths might mask treatment effects. Analysisof variance (Table 1) indicated that cone index wassignificantly affected by tillage and position (in-row,nontraffic between row, or traffic between row). Coneindex was also affected by depth and there was a sig-nificant interaction between tillage and depth and be-Table 1. Overall analysis of variance for cone index.

Source df SS F value

Replication (Rep.)TillagePositionTillage x positionRep x tillage x position (error a)DepthRep x depth (error b)Tillage x depthRep x tillage x depth (error c)Position x depthRep x position x depth (error d)Tillage x position x depthResidual errorCorrected total

3224

2415

3090309060

180575

113.672.128.017.7

272.7

204.354.952.510.218.116.0

945.4

13.82**5.36**1.69

36.18**

11.17**

15.48**

3.40**

0.00

SOIL WATER (kg-kg"1)0.10 0.20 0.30 0.40

0.10 -

• 0.20 -

0.30O_JO

0.40

0.50

0.60Fig. 2. Soil water content by weight on July 18, 1985.

tween position and depth. Because of these interac-tions, separate analysis of variances were computedfor each position (Table 2) and cone index was com-pared at each depth using a LSD test (Fig. 3 and 4).Tillage had a significant effect on cone index in therow and nontraffic between row positions, but not inthe traffic position (Table 2).

In the row position (Fig. 3), cone index was low inCT and MT above 0.25 m but in NT there was a highstrength zone at 0.15 to 0.25 m. Cone index exceeded4 MPa at the center of the high strength zone in NT,well above the 2 to 3 MPa range that is reported toprevent root growth (Taylor and Gardner, 1963; Tay-lor and Burnet, 1964). Water content at this depth inNT corresponded to a matric potential of about —0.01MPa, using the characteristic curve from the samplescollected in December. Since water content in NT wasthe same or higher than in CT at each depth (Fig. 2),the difference in cone index between these two treat-ments could not be attributed to soil water. Differ-ences in soil texture can also affect cone index (Spiveyet al., 1986), but at the 0.15- to 0.30-m depth wherethe largest differences in cone index occurred, therewas little difference in particle size distribution amongtreatments (Table 3). The higher cone index in NTcompared to CT at the 0.15- to 0.25-m was thereforeTable 2. Analysis of variance for cone index by position.

In rowsNontraffic Traffic

between rows between rows

Source df SS F value SS F value SS F value

Replication (rep)TillageRep x tillage

(error a)DepthRep x depth

(error b)Depth x tillageResidual errorCorrected total

32

615

453090

191

52.452.3 7.17*

21.9129.8 22.55**

17.373.3 7.18**30.6

377.6

47.225.3 6.61*

11.5102.9 30.79**

10.069.3 8.21**25.3

291.7

23.112.2 1.82

20.192.5 50.47**

5.579.9 16.08**14.9

248.1

** Significant at the 0.01 level. *,** Significant at the 0.05 and 0.01 levels, respectively.

RADCLIFFE ET AL.: EFFECT OF TILLAGE PRACTICES ON INFILTRATION AND SOIL STRENGTH 801

0.00

CONE INDEX (MPa)1 2 3 4 5 6 7 0.00

CONE INDEX (MPa)1 2 3 4 5 6 7 8

0.60Fig. 3. Cone index by depth, in-row position, on July 18-19, 1985.

an indication of a compacted zone of higher bulk den-sity.

This was confirmed by the measurements on phys-ical properties of soil cores taken at a depth of 0.17to 0.23 m in the row (Table 4). Bulk density was sig-nificantly higher at this depth in NT. The value of1.60 Mg m~3 in NT was in the range of bulk densities(1.55-1.65 Mg m-3) that should prevent root growthin a soil with a particle size distribution of the 0.15-to 0.30-m depth increment, according to Daddow andWarrington (1983). Bulk density in CT (1.40 Mg m-3)was well below the growth limiting value of 1.55 Mgm~3.

It is difficult to compare cone index in MT and NT(Fig. 3) because, in this case, the treatment with thehigher cone index (NT) had a 0.01 to 0.02 kg kg~'lower water content compared to MT, so that at leastpart of the difference in soil strength may have beendue to soil water differences, and not differences incompaction. Below 0.30 m, the MT and NT treatmentcurves converged and were consistently lower than inCT. The lower cone index was probably due to highersoil water content in NT and MT, compared to CT(Fig. 2).

Tillage was not a significant factor in the traffic po-sition (Table 2) because spring planting traffic in-creased cone index in the CT and MT treatments (Fig.4), narrowing the differences in cone index betweenthese treatments and NT. It is not apparent why thesame amount of traffic could have caused greater corn-Table 3. Particle size analysis from samples taken in January 1986

at two soil depths.

Tillage

No-tillageMinimumConventionalNo-tillageMinimumConventional

Depth

m0.00-0.150.00-0.150.00-0.150.15-0.300.15-0.300.15-0.30

Sand

544846

474444

Silt

—— % ——272728

252627

Clay

192626

293029

0.60Fig. 4. Cone index by depth, traffic between-row position, on July

18-19, 1985.

paction in MT compared to CT. Soil water contentcould have been higher in MT at the time of plantingdue to a mulch effect, or a significant portion of thewinter season traffic in MT could have occurred in thesame area as the spring planting operation. In the NTtreatment cone index was high regardless of positionas we would expect, in that the entire plot area prob-ably experienced some traffic during the 10 yr of theexperiment.

The compacted zone in NT may have been due totraffic or it may have been a relic tillage pan. Trafficproduced a compacted zone that had a peak cone in-dex at a depth of about 0.10 m in MT and CT, andthere was little effect below 0.16 m (Fig. 4). The com-pacted zone in NT, on the other hand, had a peakcone index at a depth of about 0.15 m and the layerextended to a depth of about 0.25 m. Although trafficprobably contributed to compaction in NT, the factthat the zone of high strength occurs at a deeper depthin NT may indicate that the compacted layer was pres-ent throughout the experimental area at the beginningof the study. The test area had been in fescue sod forat least 10 yr before the experiment was established.It is very likely, however, that at some earlier timethe area was extensively tilled during cotton produc-tion. The low contents of organic matter and expand-ing clay minerals in these soils, combined with a shal-low depth of freezing during winter may have alloweda compacted layer, once formed, to persist for manyyears (Elkins et al. 1983).

Table 4. Properties of soil cores taken in the row at a depth of0.17 to 0.23 m on 11 Dec. 1985.

Tillage

No-tillageConventional

Bulkdensity

Mgnr8

1.60**1.40

Pore size distibution, ion

>29

0.29*0.37

29-5.8

———— m3

0.080.08

5.8-3.9,

0.020.02

<3.9

0.600.54

Saturatedhydraulic

conductivity

mm h"1

81134

' Significant at the 0.05 and 0.01 levels, respectively.

802 SOIL SCI. SOC. AM. J., VOL. 52, 1988

Fig.

10 20 30 40 50 60

TIME (minutes)5. Infiltration rate with time measured in July, 1987.

$!UJ

CC

O

100:

9080

70

60

50

40

30

2010

CT No Cover

/67mm h"'VCT Cover

NT Cover V,

45mm h"

August 1987Sprinkler rate = 70 mm h -1

Irnmh"

10 20 30 40 50

TIME (minutes)60

Fig. 6. Infiltration rate with time measured in August, 1987.

Pore Size DistributionIt appeared that earthworm (Lumbricus terrestris)

activity increased markedly in the NT treatment dur-ing the 10-yr course of the experiment and old rootchannels had become increasingly apparent. Wethought that the NT treatment might have more ma-cropores despite a denser soil matrix, but pore sizedistribution at 0.20 m indicated a significantly greaterpercentage of >29jum pores in CT compared to NT(Table 4). Saturated conductivity, an indirect measureof macroporosity, was also higher in CT but the dif-ference was not statistically significant due to the highcoefficient of variation (108%). It is possible that ourbulk density sampler (54 mm diam) sampled too smalla volume to represent earthworm and root channelmacroporosity. However, infiltration measurementsalso failed to indicate more macropores in NT.

InfiltrationThe first set of infiltration measurements, con-

ducted in July 1987, showed that infiltration was higherin NT (Fig. 5). Using a sprinkling rate of 42 mm h~',run-off occurred in only one of four replicate NT plots,producing a mean final infiltration rate (infiltrationrate after 60 min) of 37 mm h~' for NT compared to16 mm h~' for CT. At this point, it was not clearwhether the difference in infiltration was due to thepresence of large macropores in NT or a surface crustin CT. Several thunderstorms occurred during the 3-week interval between planting and the time we ini-tiated the infiltration measurements. The highest rain-fall rate, recorded on an hourly basis, was 36 mm h~'.The surface of the CT plots appeared to have a crustand there was evidence that water had runoff fromthese plots. To determine if surface crusting was re-sponsible for the low infiltration rate in CT, the sec-ond set of infiltration measurements were made inAugust 1987.

Infiltration was sharply reduced in NT when themineral soil was exposed to raindrop impact (Fig. 6).The final infiltration rate of NT without cover (1 mmh~') was identical to that of CT without cover. Sta-tistical analysis showed that cover was a significantfactor but tillage was not. Adding a surface mulch to

CT increased infiltration to the point that runoff wasminimal after 1 h with a sprinkling rate of 70 mm h~'.These results indicate that a surface crust rapidly de-veloped when the mineral soil was exposed to rain-drop impact. The final infiltration rate in CT withoutcover (1 mm h~') was lower than that observed in theJuly measurements (16 mm h~'). This is because thehigher sprinkling rate used in August produced a moreimpermeable crust. The surface of all of the treat-ments in the August measurements were raked beforestarting the sprinkler, so we were not measuring theeffect of a crust that was already in place in this study.A straw mulch at a rate of approximately 5000 kg ha~'was sufficient to prevent the formation of a crust, asshown in the CT with cover treatment. The surfacelayer of fine organic litter is also sufficient to preventraindrop impact. In preliminary measurements, we didnot see a drop in infiltration in NT if we removed thestraw mulch, but did not remove the top 20 mm oforganic litter and soil. This layer may have also con-tained a greater percentage of stable aggregates com-pared to the mineral soil and this may have preventedcrust formation. The decrease in infiltration in NTwith cover after about 45 min may indicate that suf-ficient water had been applied at this point (53 mm)for the compacted zone in NT to impede infiltration.

The measurements in August were made in the non-traffic inter-row. In one CT plot using a sprinkling rateof 42 mm h~', we compared infiltration in the non-traffic inter-row with that in the traffic inter-row andfound that the final infiltration rate was lower wherespring planting traffic occurred (12 vs. 19 mm h~').These measurements were not replicated so we do notknow if the effect was statistically significant. The val-ues do, however, bracket the mean final infiltrationrate for conventional tillage measured in July (16 mmh~') when the enclosed area consisted half of traf-ficked and half of nontrafficked inter-rows. The wheeltracks formed depressed areas where it appeared thatwater had ponded in the weeks previous to our mea-surements. If infiltration is lower in the traffic posi-tion, this may be due to a compacted surface soil, orto a more impermeable crust that formed as orientedsilt and clay particles settled out under intermittentponding.

RADCLIFFE ET AL.: EFFECT OF TILLAGE PRACTICES ON INFILTRATION AND SOIL STRENGTH 803

g3.0

O^ 2.0>LJ>

fc 1.0

tr0.0

un__D_ .D.

1976 1978 1980 1982

YEAR1984 1986

Fig. 7. Relative yield of NT soybeans or grain sorghum comparedto CT from 1976 to 1986. Yields were not measured in 1977 and1980.

DISCUSSIONThe fact that yields of NT summer crops have been

as high as or higher than CT in all but one year (Fig.7) indicates that the effect of the compacted zone inNT has been offset by other factors, such as the higherinfiltration rate. It has been suggested that the full ben-eficial effect of NT may not be evident until severalyears after establishment in the Southeast (Hargroveet al. 1982). We have shown that the higher infiltrationrate in NT is largely an effect of the surface mulch.Therefore one would not expect to see the full extentof improved infiltration until surface residue cover ap-proaches 100%, a process that may take several yearsin this region. A crust, formed in the early years, wouldprobably deteriorate due to freezing, thawing, wetting,and drying, provided it was protected from furtherraindrop impact. This may explain why the single yearin which NT summer crop yields were lower than CToccurred 2 yr after establishment.

Another factor that could reduce the effect of thecompacted zone in NT would be periods when matricpotential in the layer exceeded —0.01 MPa, the valueat the time we measured cone index. These periodswould be likely to occur for short intervals early inthe growing season and might allow root penetrationof the compacted zone.

Our infiltration data imply that surface effects aremore important than underlying soil permeabilityduring short-term summer rainfall events. Burch et al.(1986), working on an Alfisol in Australia, reached thesame conclusion. Recent work has shown that manysoutheastern soils are easily dispersed and prone tocrusting, despite the predominately kaolinitic miner-alogy (Chiang et al. 1987; Miller and Bahruddin 1986).In a double-crop system in the Southeast, the summercrop is usually planted in mid-June, just prior to theonset of a period of summer thunderstorm activity. Asurface crust may form in cultivated soils shortly afterplanting that controls infiltration for the remainder ofthe growing season. The paucity of 2:1 expanding clayin these soils reduces cracking of crusts and probablycontributes to their persistence. The formation of aclosed canopy in this instance would not improve in-filtration if the crust was already formed. Cultivationjust prior to canopy closure may be a desirable man-

agement practice. The use of no-tillage to improve in-filtration under irrigation may also be desirable.

804 SOIL SCI. SOC. AM. J., VOL. 52, 1988