Specific Effect of Magnesium on Soil Erosion and Water InfiltrationR. Keren*

ABSTRACTThis study was conducted to determine the effects of adsorbed Mg

on erosion of soils exposed to rain. The effects of adsorbed Mg andCa on soil erosion and infiltration rate (IR) of two soils (a CalcicHaploxeralf and a Typic Rhodoxeralf) exposed to rainfall, in theabsence and presence of adsorbed Na, was studied. A drip-type rain-fall simulator and deionized water were used at water-drop kineticenergy of 12.5 kj nr3. The erosion rate of the soils was higher forthe Mg-soils than for the Ca-soils. Both the steady-state IR and thecumulative water depth required to reach a steady-state IR werelower for Mg soils than for Ca soils. Adsorbed Mg by the mont-morillonitic soil increased erosion and lowered IR, regardless ofCaCO3 content. The Ca aggregates were more stable than Mg ag-gregates, even in the presence of Na. The specific effect of Mg onsoil erosion was explained by the presence of these ions on the ex-ternal surfaces of the clay tactoids, and the wider hydration shell ofthe Mg ion, as opposed to that of the Ca ion. It was suggested thatthe steady-state infiltration and soil erosion rates are the result oftwo processes that occur in opposite directions: seal formation andsoil detachment.

THE FORMATION of a seal at the soil surface is acommon feature in many soils. This is particu-

larily evident in soils of the arid and semiarid regions,where the HC controls the IR (Hillel and Gardner,1970; Morin et al., 1989). The HC depends both onthe adsorbed cation composition and on the electrolyteconcentration of the percolating solution (Quirk andSchofield, 1955; McNeal et al., 1968; Russo and Bres-ler, 1977). Soil surfaces are particularly susceptible toadsorbed-ion composition due to the mechanical ac-tion of the falling drops (Keren, 1989) and the relativefreedom of soil-particle movement at the soil surface.Considerable research effort has been expended on anattempt to understand the role of adsorbed Na ionson soil physical properties when Ca was the comple-mentary ion in the adsorbed phase (Agassi et al., 1981;Keren and O'Connor, 1982; Pupisky and Shainberg,1979; Russo and Bresler, 1977; Singer et al., 1982).The effect of Na on the soil physical properties whenMg is the complementary ion is less well known.

Laboratory studies imply that Mg is a deleteriousion in some circumstances. Montmorillonite and illiteseem to be more easily dispersed in the presence ofNa and Mg than in the presence of Na and Ca (Bakkerand Emerson, 1973; Emerson and Chi, 1977; Shain-berg et al. 1988). McNeal et al. (1968) showed thatmixed Na-Mg soils developed lower HC than Na-CaAbbreviations: IR, infiltration rate; HC, hydraulic conductivity; CH,Calcic Haploxeralf; TR, Typic Rhodoxeralf; ESP, exchangeable so-dium percentage.

Institute of Soils and Water, Agricultural Research Organization,The Volcani Center, Bet Dagan, Israel. This research was supportedby Grant no. 1-743-84 from BARD—the U.S.-Israel Binational Ag-ricultural Research and Development Fund. Contribution no. 2812-E, 1989 series, from ARO, The Volcani Center, Bet Dagan, Israel.Received 22 Nov. 1989. 'Corresponding author.

Published in Soil Sci. Soc. Am. J. 55:783-787 (1991).

soils under similar conditions. In contrast, the calcar-eous soils did not show a specific effect of Mg on claydispersion and HC losses (Alperovitch et al., 1981). Inthese soils, the exchangeable Mg enhances the disso-lution of the carbonates and increases electrolyte con-centration, thus preventing clay dispersion andreduction in soil HC. Recently, Keren (1989) showedthat the IR for soils adsorbed by Mg was lower thanfor soils adsorbed by Ca, even when Na was present(Keren, 1990). These results indicated that exchange-able Mg has a specific effect on IR when the soil wasexposed to rainfall, whether or not the soil containedCaCO3.

Since the hydration number of the Mg ion is ap-proximately 50% higher than that of Ca ion (Bockrisand Reddy, 1970), it is expected that the thickness ofthe Stern layer in the diffuse double layer on the ex-ternal surfaces of the tactoid would be wider for Mgthan for Ca ions. Because the linkage among the tac-toids forming an aggregate is through the external sur-faces, and due to the wider hydration shell of the Mgions concentrated on the external surfaces, it is ex-pected that less energy would be required to breakdown Mg soil aggregates than to break down Ca soilaggregates. Thus, it is expected that Mg soil woulddisintegrate more easily than Ca soil when subjectedto rainfall, even in the presence of Na at low levels.This hypothesis was tested in the present study. Theobjective was to determine the effect of adsorbed Mgon erosion and IR for calcareous and noncalcareoussoils.

MATERIALS AND METHODSSoil Preparation

Two soils from the coastal plain of Israel were used. TheCalcic Haploxeralf (loamy soil) from Nahal Oz was devel-oped on loessial windblown material from the Siani Desert.The typic Rhodoxeralf (sandy loam soil) from Morasha wasformed on coastal sand and calcareous sandstone. The maindifference between the soils in their texture, and CaCO3 andsesquioxide contents are given in Table 1.

The soils were packed 2 cm deep in perforated trays (30by 50 cm) and placed in a rainfall simulator (Morin andBenyamini, 1977) over a layer of coarse sand at a 5% slope.The soils were leached with a 0.2 molc L-3 solution of CaCl2,MgCl2, NaCl + CaCl2, or NaCl + MgCl2 using a rainfallintensity of 10.4 nm S"1. Appropriate Na/Ca and Na/Mgratios in solutions were used to obtain ESP 10 only for theCa and Mg Ch soil. In order to prevent seal formation during

Table 1. Chemical and physical characteristics of the soils in thisstudy.

Soil propertyTexture, %

claysiltsand

pHCation-exchange capacity, molc kg'1CaCo3, %

CalcicHaploxeralf

21.741.436.98.4

17.813.4

TypicRhodoxeralf

20.63.2

76.27.6

11.3trace

783

784 SOIL SCI. SOC. AM. J., VOL. 55, MAY-JUNE 1991

the leaching, the rain was applied as a mist for approximately2h (three teachings of 40 min each). After leaching, electro-lytes in soil solution were removed from the soils by onefurther leaching with deionized water and the soils were airdried. The adsorbed-ion composition was determined for allsoils using the NH4OAc method (Black et al., 1965, p. 891),to ensure that the adsorbed-ion composition was as required(values are given in figures).

Rainfall Simulation ExperimentsA drip-type rainfall simulator with a closed water chamber

(75 by 60 cm) was used to generate rainfall through a set ofhypodermic needles (> 1000 needles, spaced in a 2 by 2 cmgrid), forming a fixed, known droplet size. The averagewater-drop diameter was 2.97 ± 0.05 mm, and the rainfallintensity of 9.2 /tm s~' was controlled by a peristaltic pump.The kinetic energy of a raindrop was regulated by changingthe fall height of the droplets. The height used in these ex-periments was 1.6 m resulting in a raindrop kinetic energyof 12.5 kJ m-3 (Epema and Riezebos, 1983).

Soil aggregated <4 mm were packed 2 cm thick in thetray (20 by 40 cm) over a layer of coarse sand. The sandallowed free drainage of water to an outlet pipe. The trayswere placed in the rainfall simulator at a 15% slope. The soilwas first saturated from the bottom with tap water, and thenexposed to simulated rainfall consisting of deionized water(electrical conductance <0.01 dS m-1) until a steady-stateIR was obtained. Standard operating conditions for eachrainfall event were geared to calibrate and check the rainfallrate before and after the run. During each rainfall event,both infiltrated and runoff waters were collected separatelyin graduated cylinders and the water volume was recordedas a function of time. The amount of sediments in the col-lected runoff was measured. Three replicate runs were con-ducted simultaneously.

ComputationsAnalysis of each set of IR data was carried out separately

with nonlinear regression, using the method proposed byMarquardt (1963), and computerized by Goodnight and Sail(1982). The infiltration equation used was described by Mor-in and Benyamini (1977);7ti = I, + (A - If) exp(—4 = P t,<tp

[1][2]

where 7ti is the IR rate as a function of time; I{ is the initaland I{ the steady-state infiltration rates; r is a soil coefficient;p is the rainfall intensity; t-t is the time from beginning ofthe storm; and tp is the time of incipient ponding. Valuesfor the infiltration-equation parameters were obtained bynonlinear regression using PROC SYSNLIN of the SASpackage. The lines in the figures of IR were calculated usingthis procedure, and the bars of the dots (experimental results)indicate population standard deviations as determined forthe replication of each treatment.

RESULTS AND DISCUSSIONThe IR of the two soils saturated with Ca and Mg

ions, as a function of the cumulative rainfall, is givenin Fig. 1. The IR of the CH soil decreased graduallywith an increase in cumulative rainfall when the soilwas saturated with Ca ions, whereas a sharp drop wasobserved for the Mg system. Furthermore, the steady-state IR valve for the Mg form of this soil (1.1 ± 0.1/urn s"1) was lower than the value for the Ca form (2.2± 0.2 MOI s"')> and the cumulative water depth re-quired to reach a steady-state IR was lower for the Mgsoils than for the Ca soils (Fig. 1).

Similar behavior was observed for the TR soil, butthe IR values were lower at any given rainfall depth,and the effect of Mg was less pronounced (Fig. 1). Thesharp drop in IR with the increase in cumulative rain-fall and the low steady-state value of IR indicate thatseal formation is taking place on these two soil surfacesexposed to raindrops at a kinetic energy of 12.5 kJnr3, even when the soils are saturated with either Caor Mg ions.

The erosion rate for the CH and the TR soils sat-urated with Ca and Mg ions as a function of cumu-lative rainfall is given in Fig. 2 and 3, respectively.The erosion rate increased with rainfall depth for bothsoils, adsorbed either by Ca or Mg ions, approachinga steady-state value. The increase in erosion rate forthe Mg soils was steeper than the rate for the Ca soils.Although the IR for both Mg soils was lower than forCa soils at any rainfall depth, the erosion rate for theMg soils was higher. Similar to the IR data, the effectof Mg on soil erosion was more pronounced for theCH soil than for the TR soil. These results show thatthe erodibility of Mg soils is higher than that of Casoils. The higher erosion-rate values for the Mg soilmay suggest that the average aggregate size of Mg soilin the seal is smaller than the size for the Ca soil. Thishypothesis is supported by the fact that the averagenumber of platelets in a tactoid of Mg montmorilloniteis smaller than that of Ca montmorillonite (Banin andLahav, 1968).

The lower IR values (Fig. 1) and the higher erosionrate (Fig. 2 and 3) for the Mg soil can be explained asfollows. The potential of the soil clay to disperse de-pends on the adsorbed-ion composition and the soilsolution concentration. As the flocculation value ofMg and Ca montmorillonite at pH 7 is approximately1 molc m-3 (Goldberg and Glaubig, 1987), the clay ofboth soils dispersed when exposed to rainfall of deion-ized water. Electrical double-layer repulsion and vander Waals attraction are the two major forces actingbetween two clay platelets. Israelachvili and Adams(1978) observed an additional short-range repulsionforce. This force is caused by the hydration of thecounterions (Low, 1987; Pashley, 1981). It is expected,therefore, that an additional repulsion force would berequired to remove some fraction of the hydration

0.025 0.050 0.075 0.100CUMULATIVE RAINFALL (m)

0.125

Fig. 1. Infiltration rate of Calcic Haploxeralf (CH) and Typic Rho-doxeralf (TR) soils saturated with Mg and Ca ions, as a functionof cumulative rainfall at a water-drop kinetic energy of 12.5 kJ nr3.

KEREN: MAGNESIUM EFFECTS ON EROSION AND INFILTRATION 785

shell around the adsorbed ions, thus increasing thedegree of clay-platelets association. Since the hydra-tion number of the Mg ion is approximately 50% great-er than that of the Ca ion (Bockris and Reddy, 1970)it is expected that less energy would be required tobreak down the linkages among aggregates of Mg tac-toids than those of Ca tactoids. This is due to a thickerdiffuse layer associated with the clay platelets in theStern layer on the external surfaces of the tactoidswhen Mg is present. This hypothesis is supported bythe fact that Mg soil aggregates are weaker than thosecontaining adsorbed Ca ions (Emerson and Smith,1970).

The IR of the Na/Ca-CH soil is given in Fig. 4. Asexpected, the presence of adsorbed Na significantly

0.025 0.050CUMULATIVE RAINFALL

0.075(m)

J0.100

Fig. 2. Erosion rate of a Calcic Haploxeralf soil saturated with Mgand Ca ions, as a function of cumulative rainfall at a water-dropkinetic energy of 12.5 kJ m'3.

15

-E,0

Mg Soil

o8

I S D

TYPIC RHODOXERALF

J0.025 0.050

CUMULATIVE RAINFALL (m)0.075

Fig. 3 Erosion rate of a Typic Rhodoxeralf soil saturated with Mgand Ca ions, as a function of cumulative rainfall at a water-dropkinetic energy of 12.5 kJ m"3.

affected the IR of both Ca and Mg systems. The IRvalues were lower than those for the Ca and Mg soils(Fig. 1). For example, the steady-state IR values forthe Ca- and the Mg-CH soil in the presence of Na were1.7 ± 0.1 and 0.8 ± 0.2 urn s~l, respectively, whereasthe values in the absence of Na were 2.2 ± 0.2 and1.1 ± 0.1 Mm s"1, respectively.

The erosion rate for Na/Ca- and Na/Mg-CH soil isgiven in Fig. 5. In accordance with Singer et al. (1982),the presence of adsorbed Na increased erosion of thesoil. Furthermore, results indicate that Mg had a spe-

10.0 r

CALCIC HAPLOXERALFESP 10

0 0.025 0.050 0.075 0.100CUMULATIVE RAINFALL (m)

Fig. 4. Infiltration rate of a Calcic Haploxeralf soil adsorbed withNa-Mg and Na-Ca ions at exchangeable sodium percentage (ESP)10, as a function of cumulative rainfall at a water-drop kineticenergy of 12.5 kJ m-3.

3.0 r

2.5

'2.0

COOa:

CALCIC HAPLOXERALFESP 10

I S D

J0.025 0.050

CUMULATIVE RAINFALL ( m)0.075

Fig. 5. Erosion rate of a Calcic Haploxeralf soil adsorbed with Na-Mg and Na-Ca ions at exchangeable sodium percentage (ESP)10, as a function of cumulative rainfall at a water-drop kineticenergy of 12.5 kJ m"3.

786 SOIL SCI. SOC. AM. J., VOL. 55, MAY-JUNE 1991

cific effect on soil erosion in the presence of adsorbedNa at ESP 10. The erosion rates for the Na/Mg soilwere higher than those for the Na/Ca soil, indicatingthat the erodibility of soils is higher in the presenceof Mg as a complementary cation. It seems that Na/Ca aggregates are more stable than those adsorbed byNa/Mg when exposed to water drops with this kineticenergy. These results suggest that the forces involvedin stabilizing the aggregates of the Na/Mg soil at ESP10 are weaker than those opperating in the Na/Ca sys-tem. The specific effect of Mg in the presence of ad-sorbed Na ions at low levels can be explained asfollows. At this ESP value, Ca and Mg adsorbed notonly on sites exhibiting a strong preference for bivalentcations—such as those in the interior of clay tactoids—but also on external surfaces. The attraction forcesbetween clay platelets are great at relatively short dis-tances from the clay surface, e.g., in the range of theStern layer of the diffuse double layer (Keren et al.,1988). The presence of Mg ions on external surfacesmay lower the association strength among soil aggre-gates due to the wider hydration shell (in comparisonwith Ca ions), thus enhancing aggregate disintegrationand clay dispersion.

Although exchangeable Mg (in the presence of Na)does not have a specific effect on the HC of a calcar-eous soil (Alperovitch et al., 1981), these results showthat Mg has a significant effect on both erosion andIR when this soil is exposed to rainfall. During leach-ing of a calcareous soil, the existence of exchangeableMg enhanced dissolution of CaCO3, and the increasein the electrolyte concentration in the soil solutionprevented clay dispersion and a reduction in the soilHC. This effect of CaCO3 can occur in the soil profile.However, electrolyte concentration in soil solution atthe soil surface is determined solely by the electrolyteconcentration in the applied water, as the contact timebetween an incremental volume of water and the sur-face area of CaCO3 is too short. Thus, the low elec-trolyte concentration at the soil surface exposed torainfall, coupled with the energy impact of the waterdrops, render the Na/Mg soil surface more susceptibleto dispersion and erosion than the Na/Ca soil, evenin the presence of CaCO3.

As shown above, the IR in both soils reached steady-state values, despite the fact that soil erosion occurredduring the entire range of the storm. This indicatesthat the seal that formed due to the impact of theraindrops is not stable. Two processes operating inopposite directions are possibly taking place simul-taneously at the soil surface exposed to rainfall. Thefirst process is seal formation. This process could beaccounted for by three possible mechanisms: (i) break-down of soil aggregates by raindrop impact; (ii) com-paction of the soil surface to form a thin soil film onthe surface that restricts further entry of water andmovement of particles in the soil pores; and (iii) claydispersion and movement of fine particles into theupper layer (few millimeters) and deposition into soilpores (McIntyre, 1958). It was suggested (Keren, 1989)that all three mechanisms mentioned above occur inCa and Mg soils. The second process is soil detach-ment and sheet erosion. This process destroys the sealon the soil surface and enhances IR. Thus, the steady-

state infiltration and soil-erosion rates are a result ofthe net effect of these two processes.

CONCLUSIONSThe results presented here show that adsorbed Mg

ions have a specific effect on soil erosion and infiltra-tion rates for montmorillonitic soils. The presence ofadsorbed Na affects soil erosion, but the erosion ratefor Na/Mg soil was greater than the rate for Na/Casoil at any given cumulative water volume. Magne-sium has a significant effect on soil erosion and IR,despite the fact that soil contains CaCO3. The con-centration of the soil solution at the soil surface isdetermined solely by the concentration of electrolytein the applied water. Thus, the low electrolyte con-centration near the soil surface exposed to rainfall,coupled with the energy impact of the water drops,render the Mg soil surface more susceptible to sealingthan the Ca soil, regardless of the presence of CaCO3.The Ca- and Na/Ca-soil aggregates were more stablethan those adsorbed by Mg and Na/Mg ions, respec-tively. The specific effect of Mg on soil erosion wasexplained by the presence of Mg ions on external sur-faces of the clay tactoids, and the wider hydration shellof Mg as opposed to that of the Ca ion. It was suggestedthat the steady-state IR and soil-erosion rate are a re-sult of two processes that occur in two opposite di-rections: seal formation and soil detachment.

ACKNOWLEDGMENTSThe author wishes to thank Mr. Y. Gian for his technical

assistance.

MANRIQUE ET AL.: PREDICTING CATION-EXCHANGE CAPACITY FROM SOIL PROPERTIES 787