Published March, 1995 Plant and Environment Interactions Soil Water Balance Changes in Engineered Soil Surfaces M. R. Sackschewsky,*C. J. Kemp,S. O. Link, and W. J. Waugh ABSTRACT mechanisms, such as erosion or biointrusion, must not increase the potential for water infiltration through the buried waste. A barrier design being evaluated at the Hanford Site (Wing, 1993) and other DOEfacilities (Nyhan et al., 1990) uses a capillary break as the primary means of preventing infiltration. A capillary break is producedby placing a layer of fine-textured soil, such as a silt loam, over a layer of coarse sands and gravel. The matrix potential within the fine-soil layer prevents moisture flow into the relatively large pores of the course layer (Hillel, 1982). Drainage through the capillary break will occur only as the fine soil layer at the textural break approaches saturation. The soil cover should be thick enough to store all precipitation, but not so deep that water can flow beyond the effective depth of evapotranspiration (ET) (Anderson et al., 1993). If carefully constructed with the proper materials, a capillary break system also can be effective in preventing biointrusion into the buried waste (Cline et al., 1980). The presence of vegetation can greatly increase the amount of ET from the upper soil layers (Anderson et al., 1987) and can greatly decrease the amount of deep drainage (Gee et al., 1994). Using field plots, Waugh et al. (1994) found that a mixture of cheatgrass (Bromus tectorum L.) and wheatgrasses (Agropyron sp.) were able to extract nearly all of the available water in the rooting zone (30-125 cm depth) of a fine soil, even when the plots were irrigated up to twice normalprecipitation. Using large (2 m diam.) lysimeters, Gee et al. (1993) showedthat a mixture of grasses and shrubs significantly increases ET and decreases or eliminates deep drainage, even under three times normal precipitation. Vegetation contributes significant erosion protection to a soil surface (Wischmeierand Smith, 1978). Unfortunately, after catastrophic disturbances such as fire or tilling, the erosion protection provided by vegetation and a large portion of the soil cover could be lost.Under such circumstances, the soil surface can be protected with gravel (Finley et al., 1985). However, a thick (several centimeters) gravel layer reduces the soil surface area available for evaporation, thus decreasing the overall soil-column evaporation rate (Groenvelt et al., 1989) and can increase infiltration (Valentin and Casenave, 1992; Kemperet al., 1994) and deep drainage (Gee al., 1992), at least in the absence of vegetation. Adequate erosion protection can be obtained using a thin gravel veneer (<1 cm) or a gravel admixture rather than a thicker gravel mulch. Gilmore and Waiters (1993) Permanentdisposal of radioactive waste requires the construction of isolation barriers that minimizeboth recharge and erosion. Recharge can be prevented by storing precipitation near the surface so that it will be returned to the atmospherevia evapotranspiration. Erosion can be reducedwith gravel mulch, but thick gravel layers mayincrease recharge. Gravel mixed into the surface soil may provide erosion protection without increasing recharge. To comparethe effects that erosion control has on infiltration, two lysimeter experiments were conductedto examinethe effects of sand and gravel mulchesand gravel admixtures, using two precipitation regimens and with or without vegetation. Sand and gravel mulchincreased soil-column water storage and decreased evapotranspiration comparedwith a plain soil surface. Gravel admixtures did not significantly affect the soil water balance comparedwith plain soil surfaces. Vegetation increased evapotranspiration and decreased soil moisture storage comparedwith nonvegetated treatments. Irrigation greatly increased evapotranspiration but had little effect on soil water storage. Drainage was detected fromsand and gravel-mulchlysimeters, but not fromlysimeters with a plain-soil or gravel-admixture surface. Results are significant for isolation barrier designs in arid sites: (i) a nonvegetated gravel-mulch surface eventually will result in recharge, even under low precipitation (160 mm/yr); and (ii) a soil columnwith a plain-soil or gravel-admixture surface is capable of recycling all water back to the atmosphere,even under high-precipitation (450 mm/yr). E RGEQUANTITIES of radioactive and hazardous waste are currently buried in shallow landfills at several U.S. Department of Energy (DOE)sites in the western USA. A major challenge in the final disposal of this waste is to design and construct earthen covers that will isolate the buried waste and prevent dispersion of the contaminated material for hundreds or thousands of years (Reith and Thompson,1992). Environmental factors that mayinduce contaminantspread include water infiltration, wind and water erosion, and biological transport caused by burrowing animals or plant uptake. Prevention of deep infiltration is the primary factor considered in the design of protective barriers. Waterthat infiltrates through a cover has the potential to transport contaminants to the saturated groundwater system and subsequently to humans. Measures that are employed to minimize the influence of other contaminant transport M.R. Sackschewsky and C.J. Kemp, Westinghouse Hanford Company, P.O. Box 1970, Richland, WA99352; S.O. Link, Pacific Northwest Lab., P.O. Box 999, Richland, WA99352; and W.J. Waugh, Environmental Sciences Lab., U.S. Dep. of Energy GrandJunction Projects Office, P.O. Box 14000, Grand Junction, CO81503 (M.R. Sackschewsky, present address: Pacific Northwest Lab., P.O. Box 999, Richland, WA99352. C.J. Kemp,present address: IT Hanford Company,Richland, WA99352). Received 25 Mar. 1994. *Corresponding author. Abbreviations: DOE,U.S. Department of Energy; ABS, acrylonitrilebutadiene-styrene; PVC,polyvinyl chloride. Published in J. Environ. Qual. 24:352-359 (1995). 352 SACKSCHEWSKY ET AL.: SOIL showed that a gravel admixture reduces sediment loss via runoff erosion. Ligotke and Klopfer (1990) found that a gravel admixture provides nearly as muchprotection from wind erosion as a thicker gravel mulch. An important difference between a gravel admixture and a gravel mulch is the effect on water infiltration. Using field plots, Waughet al. (1994) showed that a gravel admixture slightly increased near-surface soil water storage (30 cm depth), but the increase was nullified the presence of vegetation. Likewise, lysimeter results presented by Gee et al. (1993) suggest that an admix gravel has little effect on soil water storage. Deposition (i.e., the opposite of erosion) can also influence soil water balance. Sand layers can develop on a protective barrier over time, especially if vegetation is present that can trap wind-blown sand particles to produce coppice dunes (Gile, 1966; Stuart et al., 1971; Woodet al., 1978). Modaihshet al. (1985) and Kemper et al. (1994) showedthat a surface sand layer significantly increased infiltration and soil water storage, and that the influence of a sand layer is similar to that of a gravel layer. Twolysimeter experiments were designed to test how several infiltration and erosion control mechanisms interact to influence the hydrologic performanceof capillary barriers. Wehypothesized that: (i) sand and gravel mulches, and to a lesser extent the gravel admixture, wouldincrease soil water storage, decrease evaporation, and increase drainage; (ii) plants would extract most the soil moisture, and that drainage would be minimal in lysimeters with a vegetated surface compared with nonvegetated (bare) lysimeters; and (iii) an increase precipitation would have a greater effect on gravel- or sand-mulchlysimeters than on plain-soil lysimeters, and have a greater effect on nonvegetated lysimeters than on vegetated lysimeters because of the greater potential ET of the nonmulchedand/or vegetated surfaces. MATERIALS AND METHODS The experimentswere conductedat the Small-TubeLysimeter Facility adjacent to the HanfordMeteorologicalStation, on the DOEHanfordSite in south-central WashingtonState. Thefacility consists of an array of 105lysimeters, 70 of which were used to study the effects of gravel or sand mulchesand gravel admixtureson soil-columnwater balance, as influenced by vegetation and increasedprecipitation. Eachlysimeter comprisesan acrylonitrile-butadiene-styrene (ABS)well casing (169 cm long, 30.4 cm i.d.) placed inside a polyvinyl chloride (PVC)sleeve (175 cmlong, cmin diam.). The tops of the sleeves and lysimeter pipes are placed approximately2.5 cmabovegrade. A rubber insulating collar (36-cminflatable, whiteinnertube)is placedat the upper extremity of each lysimeter pipe to minimizeheat transfer betweenthe atmosphereand the airspace betweenthe outer sleeve and inner lysimeter pipe. Eachlysimeter is fitted with a recessed cap at the bottomand an aluminumlifting collar at the top. The sealed pipe serves as a combinedweighing and drainage lysimeter. Drainage is measuredby collecting water froma clear, flexible polymertube that is fitted to a threaded drain at the lowend of the end cap. Changesin water storage are estimated as the weight changefrom the initial WATER BALANCE CHANGES 353 lysimeter weight, measuredby suspendingthe lysimeters from a load cell attachedto a gantrycrane. Theload cell is calibrated at each measurementdate, and has an accuracy of +0.1 kg (equivalent to 1.4 mmwater). The lysimeters were filled by hand, starting with a 15-cm drainage layer graded from 1-cmpea gravel at the bottomto no. 20/30 (1.27-0.85 mm)sand at the top. Overlying the drainage layer is 1.5 mof silt loamsoil. In the lysimeters with a gravel-admix surface, 30%(by weight) gravel was uniformlypremixedinto the upper 20 cmof silt loambefore placementin the lysimeter. In the lysimeters with a gravelmulchsurface, the upper7 cmof soil wasreplacedwith gravel ranging in size from1.0 to 3.0 cm. In lysimeters with a sand mulch surface, the upper 20 cm of soil was replaced with local dunesand. Soil moisturelevels weredeterminedfor each layer at the timeof filling. Acompletedescriptionof the initial conditions for each lysimeter can be found in Relyeaet al. (1990). The treatments were distributed randomlythroughout the lysimeter array. Constructionand filling werecompleted in mid-September1988. Experiment1 was designedas a 3 x 2 x 2 factorial analysis of varianceto test the effects of erosion control practices. It included three different surface treatments (plain soil, 30% gravel admix, and surface gravel mulch),two levels of water input (ambient precipitation and ambientprecipitation plus supplemental irrigation), andthe presenceor absenceof vegetation (cheatgrass). Therewerea total of 12 treatmentcombinations, with five replicates of each combination. Experiment2 was designed as a 2 x 2 factorial analysis of variance to examinethe effects of a sand mulchcompared with a gravel mulchunderboth bare and vegetated conditions. All of the lysimeters in Experiment2 received supplemental irrigation as well as ambientprecipitation. Therewerea total of four treatment combinations,with five replicates of each treatment. Thesupplementalirrigation treatmentreceived natural precipitation plus enoughirrigation (0-41 mm)to bring the total water input to a predeterminedtotal. Duringthe first 2 yr of the experiments, the amountof water added (monthly) increased the total water input to twice the long-termmonthly average (obtained from Stone et al., 1983). FromNovember 1990throughSeptember1992, this wasincreasedto three times the long-termmonthlyaverage, with water addedbimonthly.At the end of the 4-yr course of the experiment, the ambient precipitation lysimeters receiveda total of 644mmof water, while the irrigated lysimeters received a total of 1561mmof water. Cumulativeamountsof normal and ambient precipitation, the irrigation goal (twoor three times normalprecipitation), and the total amountapplied to the irrigated lysimeters throughoutthe course of the experimentsare shownin Fig. 1. The lysimeters with a vegetated surface were seeded (1000 seeds/lysimeter) with cheatgrass in early October1989. Seed was collected from the HanfordSite during early June 1989. Seeds were air-dried and maintained at roomtemperature in paperbagsuntil planting. Just beforeplanting,cheatgrasslitter was collected to serve as mulch. On the lysimeters with a surfaceof plain soil or graveladmix,the soil crust wasbroken, seeds were distributed evenly over the disturbed soil, and the seeds were packed lightly by hand and mulched with approximately1 cmof cheatgrasslitter. Lysimeterswith gravel mulchon the surface were not disturbed before planting. The cheatgrass was replanted on the lysimeters in October 1990 using the sameprocedures, with the addition of 0.53 g ammoniumnitrate per lysimeter (equivalent to 25 kg N/ha). The amountof plant biomasswas not quantified, but there was a 354 J. ENVIRON.QUAL., VOL. 24, MARCH-APRIL 1995 1700.~ ~ Normal Cumulative Ppt. 1500/ ~ IrrigationGoal(2-3XNormalPpt.) 1300- ~ Cumulative Ambient ept. E~1100- ,< ¯~ 7oo~ 500300100-100 ~ ~ > ~- ~ ~ ~> ~ o Date Fig. I. Cumulativeamountsof normalprecipitation, recorded ambient precipitation, irrigation goal (two times normalprecipitation from Oct. 1988 to Oct. 19~0, three times normalprecipitation from Nov. 1~ to Oct. 1992), and the total amo~tof precipitation ÷ irrigation applied to the irrigated lysimeters. visible increase in biomassduring 1991and 1992, eventhough no additional seed was provided. Eachlysimeter was weighedand examinedfor drainage (D) on a monthlybasis during the first 2.5 yr of the experiment and every 2 to 3 moover the last 1.5 yr. Changesin soil water storage (S) were calculated as the difference between the monthlylysimeter weightand the initial (September1988) weight. Total water input (P) was knownfrom the measured amountsof precipitation andirrigation. Thesoil surfaceis below the lip of the lysimetertube to preventrunoff. Becauseevaporation and plant transpiration cannotbe separatedin this design, they are combinedinto the term evapotranspiration(ET), which is calculated using the equation: ET= P - D - S. Differences in S, ET,and D after 4 yr of measurement were examinedusing a factorial analysis of variance. Comparisons amongindividual treatment meanswere performed using the TukeyHSDmultiple comparisonprocedure, because it is the most powerfultest for unplannedcomparisonsof equal sample size means(Sokal and Rohlf, 1981). Statistical significance was set at P = 0.05. RESULTS All treatments showedseasonal fluctuations in S. The amountof water stored increased during the wetter winter months, when water input was greatest and ET was lowest. During the drier, hotter summer months, the soil columns dried out from a decrease in water input and an increase in ET (Fig. 2 and 3). The seasonal fluctuations were greater in the irrigated lysimeters than in the ambient precipitation lysimeters. The amplitude of the seasonal fluctuations of S in the ambientprecipitation treatments were less than 50 mmfrom winter to summer, whereas the amplitude of the irrigated lysimeters was approximately 100 mm.In general, the amplitude of the seasonal fluctuations increased during the last 2 yr of the experiment, in concert with an increase in plant biomassin the vegetated lysimeters and after the increase in irrigation to three times normal precipitation. Vegetation Effects At the end of the study, the presence of cheatgrass had significant effects on both S (F = 67.4, P < 0.001) and cumulative ET (F = 106.5, P < 0.001). In the ambient precipitation lysimeters, the vegetated soil and admix lysimeters had average S values of 67 and 103 mm lower than the nonvegetated soil and admixlysimeters by the end of the fourth year (Fig. 2). In the absence vegetation (Fig. 2A) the admix treatment consistently had approximately 20 mmhigher S than the plain soil; with vegetation (Fig. 2B) these two treatments had virtually identical S. Therefore, vegetation caused a greater decrease in S in the admixtreatments than in the plain-soil treatments. Storage in the bare, gravel-mulch lysimeters steadily increased during the course of the experiment (Fig. 2A), but the presence of vegetation in the gravel mulch surface treatment (Fig. 2B) reduced the amount of water storage by approximately 250 mmat the end of the experiment compared with the nonvegetated, gravel mulch lysimeters. In the irrigated lysimeters, vegetation had differing effects on S, depending on the soil surface configuration (Fig. 3). On most of the measurement dates the effect of vegetation on S was more pronounced in the admix treatments than in the plain soil treatments. During the first 2 yr, vegetation reduced S by approximately 60 to 80 mmin the admixlysimeters, but at most 30 mmin the plain soil lysimeters, comparedwith their nonvegetated counterparts. After 4 yr, vegetation reduced S by 143 355 SACKSCHEWSKY ET AL : SOIL WATERBALANCECHANGES 400 ~ _~. 300-14 2oo -.- |L A -- p . Measurement Date 400-~[ ~ Soil zoo ~"~ -~--,A(~m,× B 100 ~ -~00 MeasurementDate Fi~. ~. C~ulatJve cb~ in ~il water storage over t~e ia (~) b~e aad (B) ve~e~ted lys~eters that r~eiv~ ~bieat pr~ipi~tioa. ba~ represeat plus or min~ the m~m~obse~ed s~a~d error tot ~¢b tr~tmeat. mmin the plain soil and 170 mmin the admix lysimeters comparedwith the nonvegetated lysimeters. In the gravel mulch lysimeters the presence of vegetation reduced the amount of soil moisture storage by about 50 mmcompared with the nonvegetated lysimeters during the first 2 yr, but had no significant effect by the end of the experiment. Surface Configuration Effects The surface configuration had significant effects on S (F = 131.8, P < 0.001) and cumulative ET (F = 226.1, P < 0.001). Without vegetation, the admix lysimeters consistently had about 50 mmgreater water storage than plain soil lysimeters (Fig. 2A and 3A). With vegetation, these two surface treatments were virtually identical in S (Fig. 2B and 3B). Gravel mulch significantly (P 0.05, Tukey HSD)increased S and decreased ET compared with the plain soil or admix surfaces except for the ambient precipitation, vegetated, gravel-mulch treat- Ve~! ment, which was more similar to all of the soil and admix treatments than to any of the other gravel mulch treatments (Fig. 2B). Irrigation Effects Irrigation had significant effects on both S (F = 21.1, P < 0.001) and ET (F = 4315.5, P < 0.001). In plain-soil and admix surface lysimeters, irrigation increased S during the winter months. Vegetation dried the soil of irrigated lysimeters to the same or lower levels as the ambient precipitation lysimeters at the end of each growing season (Fig. 2B and 3B). Irrigation significantly increased the amountof soil water storage in the vegetated gravel-mulch lysimeters, but had little effect on the nonvegetated gravel-mulch lysimeters by the end of the experiment. After 4 yr, S in the ambient precipitation, nonvegetated, gravel-mulch lysimeters was not significantly different from S in the irrigated, bare, gravel-mulch lysimeters. 356 J. ENVIRON.QUAL., VOL. 24, MARCH-APRIL 1995 400 300200100- o -iooSoil -200- ---=-- Admix --b-- Gravel Sand -300 MeasurementDate 400 - B ~ E ...... 300 ~200O 1000._> _~-~ooo -200 - -- -~ Admix ~ Sand -300 MeasurementDate Fig. 3. Cumulativechanges in soil water storage over time in (A) bare and (B) vegetated lysimeters that received supplementalirrigation. Vertical bars represent plus or minus the maximum observed standard error for each treatment. WaterBalance Partitioning The analysis of variance of S and ET at the end of the experiment indicated that all of the maineffects (irrigation, vegetation, and surface type) were highly significant. Twoof the interaction effects (irrigation surface type and the three-way interaction) were also significant for both S and ET. In general, vegetation decreased,andirrigation increasedS, althoughthe plainsoil and admixlysimeters that were both irrigated and vegetated had the lowest values of S at the end of the experiment(Table 1). The gravd-mulchsurface lysimeters had higher S and lower ETthan the plain-soil or gravel-admixlysimeters. Bothof the interaction effects are causedprimarily by the ambientprecipitation, vegetated gravel-mulchlysimeters having muchlower levels of soil water storage than any of the other gravel mulch treatments.Thelarge effect of vegetationin the irrigated soil and admixlysimeters relative to the nonirrigated counterpartsalso influencesthese interaction terms. Mul- tiple comparisons of the different treatmentsreveal that all of the gravel-mulchtreatments (except the ambient precipitation, vegetated, gravel-mulchtreatment) were significantly different in S fromall of the plain-soil and admixtreatments (Table 1). Underambientprecipitation, the presenceof vegetation did not significandy affect S in the soil or admixtreatments. Whenirrigated, however, the vegetatedsoil and admixlysimeters had significantly lower levels of S than their nonvegetatedcounterparts (Table1). Both vegetation and irrigation increased the amount of cumulativeETcomparedwith the bare and/or ambient precipitation treatments. Gravel-mulchlysimeters had significantly loweramountsof ETthan comparableplainsoil or gravel-admixlysimeters, except in the ambient precipitation, vegetated condition (Table 1). Multiple comparisonsof the individual treatments showthat the soil and admixsurfaces always had similar amountsof cumulativeET, althoughthe total amountvaried signifi- 357 SACKSCHEWSKY ET AL.: SOIL WATERBALANCECHANGES Table 1. Values for cumulative storage change, evapotranspiration (ET), and drainage for the combinations of surface type, irrigation level, and vegetation treatments. Treatmentcombination Surface Irrigation Vegetation Soil Soil Soft Soil Admix Admix Admix Admix Gravel Gravel Gravel Gravel Cumulativeamount(ram) (Mean + 1 SE) Storage ET Drainage Ambient Bare - 77 + 19 abc* 720 + 19 b 0 5:0 a Ambient Cheatgrass - 144 + 6ab 788 5:6b 0 5: 0a Irrigated Bare - 33 5: 21bc 1594 5: 21d 0 5: 0a Irrigated Cheatgrass - 177 + 27a 1738 5: 27e 0 5: 0a Ambient Bare - 38 5: 12bc 682 5: 12b 0 + 0a Ambient Cheatgrass - 141 5: 20ab 785 5: 20b 0 5: 0a Irrigated Bare 12 5: 37c 1548 5: 37d 0 5: 0a Irrigated Cheatgrass - 158 5: 1lab 1718 5: 10e 0 5: 0a Ambient Bare 201 +44d 431 5:38a12 + 7a Ambient Cheatgrass - 62 5: 18abc 708 5: 19b 0 5: 0a Irrigated Bare 247 5= 45d 1237 5:15c77 + 31b Irrigated Cheatgrass 260 5:11 d 1271 + 14 c30 5= 10 ab * Values within a columnfollowed by the sameletter are not significantly different (Tukey HSD,P _< 0.05). cantly with irrigation and vegetation, and the amountof total ETin the soil and admixlysimeters was consistently higher than the gravel-mulch counterpart. After 4 yr of measurements, no drainage was detected in any of the lysimeters with a plain-soil or an admix surface, or from any of the ambient-precipitation, vegetated gravel-mulch lysimeters. The other three gravelmulch treatments eventually produced detectable amounts of drainage (Fig. 4). The irrigated, nonvegetated gravelmulch lysimeters started to produce drainage after the winter of 1990. The irrigated and vegetated gravelmulch lysimeters began producing drainage during the winter of 1991; the ambient precipitation, nonvegetated lysimeters began to produce drainage during the winter of 1992. The rate of drainage from the irrigated lysimeters increased after the amount of supplemental irrigation was increased to three times normal precipitation. Significant effects of irrigation (F = 8.307, P = 0.006) and surface type (F = 12.44, P < 0.001) were observed for cumulative drainage at the end of the experiment, as well as the irrigation x surface type interaction (F 8.155, P = 0.001). The surface type and surface irrigation effects are caused by the lack of drainage by the soil and admix lysimeters, whereas the irrigation effect is caused by the differences observed between the irrigated and ambient precipitation gravel mulchlysimeters. Because of the lack of drainage in the admix and plain soil lysimeters, and the variability observed among the lysimeters within each of the gravel mulchtreatments, the vegetation main effect was not significant. Because of the large variability in the drainage data, only the irrigated, nonvegetated gravel-mulch lysimeters had total drainage values that were significantly different from zero (Table 1). Comparisons of Gravel and Sand Mulches In general, a sand mulchbehaved similarly to a gravel mulch, especially during the first 2 yr of the experiment (Fig. 3). During the last 2 yr of the experiment, the lysimeters with a sand-mulch surface had approximately 50 mmless storage than the lysimeters with a gravelmulch surface. This coincides with the increase in the Table 2. Values for cumulative storage change, evapotranspiration (ET), and drainage for the irrigated sand and gravel mulch lysimeters. Treatmentcombination Cumulative amount (mm) (Mean5= 1 SE) Surface Vegetation Storage Gravel Gravel Sand Sand Bare Cheatgrass Bare Cheatgrass 247 5:45 a* 260 5:11 a 177 + 57 a 181 5:22 a ET 1237 5:15 1271 5:14 1268 5:17 1348 5:23 Drainage a a a b 77 5:31 30 + 10 116 5:42 33 5:18 a a a a * Values within a columnfollowed by the sameletter are not significantly different (Tukey HSD,P < 0.05). amountof irrigation at the start of the third year. The presence of vegetation decreased S by up to 100 mm during the summermonths in all years except for the final summer. The analysis of variance at the end of the experiment indicated neither vegetation nor surface type had a significant effect on S. Evapotranspiration is significantly affected by both vegetation (F = 10.39, P = 0.005) and surface type (F = 9.43, P = 0.007). Both of these effects are due primarily to the significantly higher amountof cumulative ET observed in the vegetated sand-mulch treatment comparedwith the other three treatment combinations (Table 2). Drainage was detected in 9 of the 10 sand-mulch lysimeters. Both the vegetated and nonvegetated sandmulch lysimeters began producing drainage following the winter of 1990 (Fig. 4). An analysis of variance drainage from the sand- and gravel-mulch lysimeters showed that there was a significant effect caused by vegetation (F = 5.36, P = 0.034), but no difference caused by surface type. However,none of the individual treatments were found to be significantly different from each other in the multiple comparison of treatment means (Table 2). This is because of the large within-treatment variation in all of the treatment combinations. DISCUSSION The first hypothesis, that a thick gravel- or sandmulchwouldresult in an increase in soil water storage, a decrease in ET, and an increase in the amountof deep drainage, compared with a plain-soil surface, proved true. Wealso found that a gravel admixture surface had little effect on the soil columnwater balance when comparedwith a plain-soil surface. The admixture results are similar to those of Waughet al. (1994), whofound that vegetation will withdrawall of the available soil moisture from the rooting zone, regardless of the presence of a gravel-admix surface. Waughet al. (1994) found that nonvegetated admix plots had higher levels of water storage just below the admix layer (about 30 cm). They also found that in the absence of vegetation, the water contents below the zone of seasonal wetting and drying eventually rose above the constant levels observed in the vegetated plots, indicating that deep drainage was more likely in the bare rather than in the vegetated conditions. The field plots of Waughet al. (1994) did not have a capillary break within the soil column, which allowed moisture to flow deeper into 358 J. ENVIRON.QUAL., VOL. 24, MARCH-APRIL 1995 Gravel,Nonveg (Arab.) Gravel,Nonvegetated Gravel,Vegetated Sand,Nonvegetated Sand,Vegetated 3XIrrigation 2XInigalion -20 = _ ~ _ ~ > _ o~ o ~ ~ -- o 9, MeasurementDate Fig. 4. Cumulative drainage over time, collected fromSeptember 1988throughSeptember 1992, fromlyslmeters with a gravel- or sand-mulch surface. Vertical bars represent + 1 SE for each treatment on the last measurementdate. the soil than in the small tube lysimeters. Using larger lysimeters, Geeet al. (1993) found that vegetated, admix surface lysimeters had virtually identical seasonal soilmoisture storage patterns as vegetated plain-soil surface lysimeters under ambient precipitation conditions at the Hanford Site; Gee et al. (1993) did not have nonvegetated, admix surface lysimeters available to compare with the vegetated treatments. The hypothesis that vegetation would increase ET and decrease soil water storage and deep drainage was also shownto be true, especially whenthe lysimeters received supplementalirrigation. Becauseof the low annual precipitation and high potential ET rate at the HanfordSite, even the bare admixand plain-soil lysimeters were able to recycle all of the water they received back to the atmospherevia evaporation. Therefore, the effects of vegetation were much smaller in the ambientlysimeters comparedwith the lysimeters that had supplementalirrigation. Wefound two results that are highly significant to the design of protective covers in arid environments such as the Hanford Site. First, a nonvegetated, 7-cm thick gravel-mulch surface eventually results in deep drainage, even under low ambient precipitation conditions. This indicates that thick gravel mulches should not be used for erosion control because the establishment and maintenance of an adequate vegetative cover cannot be guaranteed over the functional lifetime of an isolation barrier. The second significant finding is that the nonvegetated, plain-soil and admix surface lysimeters recycled all of the annual water input via evaporation alone, even when subjected to three times the normalprecipitation rate (up to 450 mm/yr). This indicates that a gravel admix can be used for erosion control, because, even if the vegetation is lost through somedisturbance such as fire, the likelihood of deep drainage is low. The results of this experiment differ from those of Waughet al. (1994), who found that soil moisture storage increased over time with the lack of vegetation. The differences observed between these lysimeter experiments and the field-plot experiment maybe caused by the relatively small size of the tube lysimeters. Using 2-mdiam. lysimeters, Gee et al. (1993) found that some drainage did occur from nonvegetated, plain soil lysimeters, underthe relatively stressful conditions of three times normal irrigation coupled with high snow cover. They did not detect drainage from vegetated lysimeters. Vegetation was found to be a highly desirable component of the simulated isolation barrier system. In all of the plain-soil and gravel-admix treatments, the presence of vegetation decreased the amount of moisture stored in the soil profile, and increased the amount of ET. Vegetation also was found to alter the moisture balance of the nonirrigated gravel-mulch lysimeters. The effects of vegetation, though present, were less pronounced in the irrigated gravel-mulch lysimeters. This was probably caused by the less favorable conditions for plant establishment and growth in these lysimeters. Especially following the increase in irrigation to three times normalprecipitation, the surfaces of the irrigated gravel-mulch lysimeters were constantly wet, or even flooded, during the late fall and winter months whencheatgrass usually germinates and initiates growth. This resulted in less plant biomass later in the growing season. Even with this constraint, the presence of vegetation greatly reduced the amount of deep drainage from the lysimeters, again indicating the desirability of vegetation as part of the SACKSCHEWSKY ET AL.: SOIL WATER BALANCE CHANGES isolation barrier system. The presence of perennial vegetation, such as shrubs or bunchgrasses, would help to circumvent year to year, and between treatment variation in plant biomass. We also found that the presence of a sand layer on top of the simulated barrier system drastically reduces the effectiveness of the barrier to recycle water to the atmosphere and to prevent deep drainage. Over time, sand layers could develop on top of an isolation barrier, and the deposition of sand could actually be accentuated by the presence of vegetation (Stuart et al., 1971; Wood et al., 1978) or gravel-admix surfaces (Ligotke, 1993). It may be difficult to completely prevent sand deposition over long periods, but the effects of a sand layer may be alleviated by the presence of deep rooted vegetation, especially shrubs and perennial grasses, that can extract water from deeper soil layers over longer periods during the growing season than cheatgrass alone (Gee et al., 1992; Cline et al., 1977). At the Hanford Site, cheatgrass is able to extract soil water only from depths above 100 cm in silt loams (Cline et al., 1977) and above 50 cm in sandy soils (Link et al., 1990a), while native perennial grasses such as bluebunch wheatgrass [Pseudoroegneria spicata (Pursh) Love] are capable of extracting water from depths of at least 125 cm (Link et al., 1990b). The experiments described here may be continued using perennial grasses instead of cheatgrass. In summary, we found that a gravel admixture surface does not significantly affect the soil water balance of a simulated isolation barrier. A gravel or sand surface mulch does significantly reduce the ability of the simulated barrier to recycle water to the atmosphere via ET. Systems with a sand or gravel mulch eventually will produce deep drainage from the soil, especially in the absence of vegetation. Vegetation significantly increases the amount of ET, decreases the amount of water storage in the soil, and decreases the amount of drainage. ACKNOWLEDGMENTS We would like to thank Michael Thiede, Dr. Larry Cadwell, Dr. John Relyea, and Steven Phillips for their assistance during this project. Glendon Gee and Richard Wing provided many useful suggestions and support. This project was funded by the U.S. Department of Energy through contracts DE-AC0687RL10930 and DE-AC06-76RLO-1830. 359