VZJ sign up for etocs
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 26 May 2006
Published in Vadose Zone J 5:742-750 (2006)
DOI: 10.2136/vzj2005.0112
© 2006 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Young, M. H.
Right arrow Articles by Oyelowo, L.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Young, M. H.
Right arrow Articles by Oyelowo, L.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Young, M. H.
Right arrow Articles by Oyelowo, L.
Related Collections
Right arrow Recharge
Right arrow Stochastic Processes
Right arrow Disposal Facilities

ORIGINAL RESEARCH

Incorporating Parametric Uncertainty in the Design of Alternative Landfill Covers in Arid Regions

Michael H. Younga,*, William Albrightb, Karl F. Pohlmanna, Greg Pohllb, Walter H. Zachritzc, Stephen Zitzerd, David S. Shafere, Irene Nesterf and Layi Oyelowog

a Div. of Hydrologic Sciences, Desert Research Inst., Las Vegas, NV
b Div. of Hydrologic Sciences, Desert Research Inst., Reno, NV
c National Park Service, Chesapeake Watershed Cooperative Ecosystems Studies Unit, Frostburg, MD
d Div. of Earth and Ecosystem Sciences, Desert Research Inst., Las Vegas, NV
e Center for Environmental Remediation and Monitoring, Desert Research Inst., Las Vegas, NV
f Tybrin Corp., Edwards AFB, CA
g Environmental Management, Edwards AFB, CA
* Corresponding author (michael.young{at}dri.edu)

Received 14 September 2005.



   ABSTRACT
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Monte Carlo simulations and a combination of site-specific data (e.g., soil properties, climatic conditions, and native vegetation) were used to design alternative (evapotranspiration) landfill covers at Edwards Air Force Base, located near Lancaster, CA. Laboratory analyses of site soils indicated the presence of three distinct surface soils, from which statistical distributions were generated. A 10-yr climate sequence (precipitation and potential evapotranspiration) was used for the upper boundary. Potential evapotranspiration was partitioned into potential evaporation and potential transpiration using the phenology of a Mojave Desert plant community. Nearly 1000 realizations were run for each of 72 different combinations of soil type, cover thickness, and plant cover percentage. The results indicate that threshold design parameters, needed to limit deep flux to <0.5 cm yr–1, differ based on the relationship between the Ks (saturated hydraulic conductivity) of the surface soil, cover thickness, and plant cover percentage. In the lower conductivity soils (mean Ks = 20 cm d–1), deep flux was ≤0.2 cm yr–1 for a cover thickness >80 cm with a plant cover >10%. Higher conductivity soils (Ks = 250 cm d–1) required thicker soils covers (>100 cm) and greater plant cover (>20%) to achieve similar fluxes. In all cases, variations in both cover thickness and plant cover percentage indicated threshold values, above which incremental additions added little to cover performance. The methods developed here could be implemented at other sites where conditions are known. Designs can account for uncertainties in site parameters and contribute to improved decision making.

Abbreviations: EAFB, Edwards Air Force Base • ET, evapotranspiration • MAF, mean annual flux • PE, potential soil evaporation • PET, potential evapotranspiration • PT, potential plant transpiration


   INTRODUCTION
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
LANDFILL CLOSURE for most waste streams typically requires some means to predict final cover performance. Conventional designs that use a resistive barrier of low-conductivity materials (i.e., finer grained soil, geomembranes, or both) to impede drainage through the cover typically rely on specification of materials and methods and are not commonly subjected to a performance criterion such as a maximum drainage rate. In contrast, alternative cover designs, often called evapotranspiration covers or ET covers, rely on the development of a functioning system of soil, climate, and plants to minimize deep flux. In arid environments like the Mojave Desert, virtually all water that infiltrates the upper soil layer is used by plants or simply evaporates from the soil surface (Nowak et al., 2004).

The use of ET designs for closure of waste sites in the western USA, particularly in the Southwest, has a history of design validation with the use of large drainage lysimeters that provide a reliable indication of cover performance. Three lysimeter studies are applicable to the environment found in the western Mojave Desert: the Marine Corps Air Ground Combat Center at Twentynine Palms, CA (Woodward-Clyde Consultants, 1998), the U.S. Environmental Protection Agency's Alternative Cover Assessment Program at Sacramento, Apple Valley, and Altamont, CA (Albright et al., 2004; Benson et al., 2002), and the Alternative Landfill Cover Demonstration at Sandia National Laboratories near Albuquerque, NM (Dwyer, 2003). These projects involved side-by-side tests of alternative and conventional designs and found that various configurations of ET designs performed to acceptable regulatory standards and, in some cases, equal to that of the minimum recommended designs. Comparisons of conventional vs. ET cover designs were also done through the development of multidecision support analyses (Paige et al., 1996a, 1996b) that sought to optimize a number of competing criteria (e.g., cost, deep flux, sediment yield, etc.). In this study, we were operating under the assumption that an ET-type cover design is the preferred cover paradigm.

Recent studies (Albright et al., 2002; Scanlon et al., 2002) have compared simulated and field-measured fluxes with the accuracy required for regulatory permitting activities. In general, the results have demonstrated an inconsistency between field results and numerical predictions of deep fluxes. One approach for capturing uncertainty in alternative cover design parameters (Interstate Technology and Regulatory Council, 2003) is the design sensitivity analysis. The design sensitivity analysis entails a pragmatic procedure whereby specific design parameters are systematically varied to allow evaluation of the resulting change in simulated performance. This approach is consistent with the guidance document recently published by the Interstate Technology and Regulatory Council (2003). Evaluation of the trends in predicted performance, estimated as a function of one, two, or more important parameters (e.g., cover thickness, soil hydraulic properties, or plant cover), can provide a stronger foundation for the landfill cover design, especially when deep flux data from lysimeters or natural analog sites are not available for the site or region in question. gather

In this study, we used a Monte Carlo simulation technique to assess uncertainty in predicted performance of alternative landfill covers and to collect additional information regarding the impact of incremental variation in specific design parameters on deep flux. The overall goal was to develop a method of designing alternative landfill covers that maximizes the use of data from a particular study site; it was not the goal, however, to design a particular landfill cover for a particular site. For this study, cover thickness and plant cover percentage were more easily controlled by engineering practices, allowing a target design to be identified and met. These two parameters were thus treated as categories. The soil hydraulic properties, however, vary in a way that cannot be accounted for through typical engineering practices; therefore, varying this parameter around a mean with known variance accounts for the random nature of soil variability. Combinations of soil type, cover thickness, and plant cover percentage were selected and varied according to statistical characteristics or potential engineering criteria. The objectives of the study were to (i) test different combinations of design parameters using numerical simulations to help predict how systematic variation in the design components affected deep flux, and (ii) examine the sensitivity of the hydrologic system to changes in design parameters. This research took place at EAFB (Edwards Air Force Base), located northeast of Lancaster, CA, ~125 km north of Los Angeles.


   MATERIALS AND METHODS
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Site Description
Edwards Air Force Base is situated in Antelope Valley, which is underlain by rock units of three major groups separated by unconformities (Dibblee, 1963): (i) pre-Tertiary crystalline rocks, (ii) Tertiary volcanic pyroclastic and sedimentary rocks, and (iii) Quaternary sedimentary deposits. The surficial and shallow sediments of the basin consist of Pleistocene sedimentary deposits. These deposits range in thickness from very thinly bedded to >22.9 m (Motts and Carpenter, 1968). The soil profile across much of the base ranges from a sandy loam to sand texture of 1 to 3 m in thickness, underlain by a coarse granitic weathered geological material known as grus.

Edwards Air Force Base is located in the western Mojave Desert, which is characterized by long, hot summers and short, cool winters. The mean annual estimated temperature for Rosamond Lake, within EAFB boundaries, is 16.8°C. A short period of low temperatures during the winter months is followed by long summers and daytime temperatures that often reach and exceed 37.8°C. Average annual precipitation (P) since 1942 is 14.83 cm. Summer (May–September) precipitation averages 1.82 cm and is generally in the form of convective storms. Winter (October–April) precipitation averages 12.98 cm and is generally in the form of regional frontal storms. Annual potential evapotranspiration (PET) is ~200 cm, which defines the region as arid (0.03 < P/PET ≤ 0.2) by the United Nations Educational, Scientific, and Cultural Organization (1979).

Site Characterization
Soil Material
Soils from 13 available borrow sources at EAFB (nine existing and four proposed) were grab sampled and then analyzed in the laboratory. Three samples were collected from each borrow location at depths that best characterized the current excavation depth of each borrow pit. Particle size distribution was determined for each of the 39 soil samples (three samples from each of the 13 borrow pits) using a laser light scattering technique (Saturn DigiSizer 5200, Micromeritics Instruments, Norcross, GA), located at the Desert Research Institute's Soil Characterization and Quaternary Pedology Laboratory in Reno, NV. Standard Proctor check point data (Method D698–00ae1, American Society for Testing and Materials, 2005) were obtained by Terracon Laboratories (Reno, NV). The results of density tests were expressed as the moisture content needed to achieve 80% of the Standard Proctor density in the field. The 80% criterion was chosen (Interstate Technology and Regulatory Council, 2003), rather than the maximum dry density, because it was anticipated that a final design would specify lower soil densities to enhance plant viability.

Soil hydraulic properties were determined for each of the 39 repacked core samples. Although the overall sample size clearly was small and probably not sufficient to capture the spatial heterogeneity of soil properties, the samples illustrate the overall methodology of using site data whenever possible to more realistically represent the variability of the hydraulic properties. Recompacted samples were analyzed for water content as a function of water potential [{theta}({psi})] and hydraulic conductivity as a function of water potential [K({psi})], both of which can be expressed through closed-form equations. The first equation relates water content to water pressure (van Genuchten, 1980), which is expressed as:

Formula 1[1]
where {Theta} is relative volumetric water saturation, {theta}r and {theta}s are residual and saturated volumetric water contents, respectively, {psi} is the soil water potential, {alpha} and n are fitting parameters, and m = (1 – 1/n) for 0 < m < 1. The second equation relates hydraulic conductivity to water pressure [K({psi})] (Mualem, 1976; van Genuchten, 1980). The saturated hydraulic conductivity obtained from the laboratory tests of core samples is used along with results from Eq. [1]. The resulting equation is:

Formula 2[2]
where Ks is the saturated hydraulic conductivity.

Samples were analyzed for Ks using the constant head method of the American Society for Testing and Materials (2000) for time periods ranging from several hours to 2 d. Immediately following the Ks tests, soil samples were analyzed for unsaturated hydraulic properties using the multistep outflow method of Eching and Hopmans (1993) and Hopmans et al. (2002) in the wet range (i.e., water potential greater than –50 kPa). Soil samples in the dry range were analyzed using a pressure plate extractor (Model 1600, Soil Moisture Equipment, Goleta, CA) at 98 kPa (~1 bar) and 489 kPa (~5 bar) pressure, and using a water activity meter (Model CX-2, Decagon, Pullman, WA) for measurements near air-dry water content.

Data from laboratory tests were used to determine soil hydraulic functions using inverse modeling (Simunek et al., 1998). In this case, the data are the responses of a tensiometer installed midway along the column length, the cumulative flux from the column, and paired values of water content and water tension taken at the end of the experiment. The program (HYDRUS-1D) solves the governing flow equations, and then changes the fitting parameters ({theta}r, {theta}s, {alpha}, n, and Ks) in Eq. [1] and [2] until differences are minimized between observed data collected during the experiment and predicted data obtained from the model (Hopmans et al., 2002). Further treatment of the optimized parameter values is discussed below.

Plant Communities
Mojave Desert plant communities were characterized by a combination of site visits and review of existing literature (e.g., Charlton, 1994). Native Mojave Desert plant communities consist of various proportions of evergreen shrubs, drought-deciduous shrubs, perennial forbs and grasses, succulents, and winter annuals. One of the most common communities occurring from California to southern Nevada, and present at EAFB, is dominated by creosotebush [Larrea tridentata (Sessé & Moc. ex DC.) Coville], an evergreen long-lived shrub, and white bursage [Ambrosia dumosa (A. Gray) W. W. Payne], a short-lived drought-deciduous shrub (Shreve, 1942; Ackerman and Bamberg, 1974; Axelrod, 1979; MacMahon, 1988). These two species can contribute up to 30 and 20% of the total perennial plant cover, respectively. Other common shrubs are Lycium spp., Ephedra spp., Hymenoclea salsola Torr. & A. Gray and Krameria erecta Willd. ex Schult. & Schult. f. and may contribute another 30 to 40% to total cover. The perennial grasses Achnatherum hymenoides (Roem. & Schult.) Barkworth and Pleuraphis rigida Thurb. can contribute as much as 10% to total cover. Total perennial plant cover in Mojave Desert communities varies from 5 to 40% (Romney et al., 1973). The amount of perennial plant cover will vary based mainly on variations in soil properties that influence soil water infiltration and water holding capacities (Hamerlynck et al., 2002). Thus, for the systems at EAFB, a sustainable plant cover design could consist of a creosotebush–white bursage community with plant cover ranging from 15 to 20%. For our analyses, we used a wider range of plant cover: from 0 to 30%.

Root zone distributions for the above canopy were calculated for each of the modeled geometries using the approach of Vrugt et al. (2001). Roots were limited to surface soil only, and were prohibited from penetrating the subsurface (grus) layer. Placing roots only in the surface soil layer biases the deep flux toward higher, more conservative, values. A non-uniform root-zone distribution was used, with higher mass of roots near the soil surface and lower mass with depth. This non-uniform distribution was approximated using a root zone distribution function (Vrugt et al., 2001):

Formula 3[3]
where ß(z) is the dimensionless spatial root distribution with depth z (z ≥ 0); Zm is the maximum rooting depth [L], taken at the base of the upper soil horizon; Pz (unitless) is an empirical parameter; and Z* is the depth at which root distribution is maximized [L], in this case 30 cm.

Climate
The precipitation dataset chosen for the numerical simulations consists of daily values for a period of 10 yr. The daily precipitation records cover the period from 1974 through 1983, which had the highest 10-yr mean annual precipitation for any combination of 10 consecutive years for the period of record (1942–2000). For the selected period from 1974 through 1983, the mean annual precipitation was 20.46 cm, compared with the long-term average of 14.83 cm. The highest recorded annual total was 39.42 cm in 1983, the lowest recorded annual total was 4.60 cm in 1975, and precipitation during eight of the 10 yr was greater than the long-term mean.

Potential evapotranspiration data were obtained from monthly pan evaporation data collected at the Mojave weather station, located immediately northwest of EAFB, between 1964 and 1978. Pan evaporation was converted to PET by using a ratio of 0.70 (Farnsworth and Thompson, 1982). The magnitude of the correction factor is similar to data collected at Lake Mead, NV, and Davis, CA (Linsley et al., 1982). Monthly averages ranged between 8.4 and 43.7 cm and were divided by the number of days in each month to derive daily PET values.

PET data were further partitioned into potential soil evaporation (PE) and potential plant transpiration (PT) processes. The partitioning was estimated based on seasonal changes in PT, the degree of plant cover, and the uptake phenology of the plants that typically populate the southwestern Mojave Desert. Growth and annual gas exchange patterns for Mojave Desert plant communities are closely related to rainfall patterns, especially relative to season and the intensity of rainfall events (Beatley, 1974). More than 90% of total annual production (0–250 g m–2) by all five major growth forms (evergreen shrubs, drought-deciduous shrubs, perennial grasses, perennial forbs, and winter annuals) is produced from December to May (Bowers, 1987; Turner and Randall, 1989). Despite significant variation in total plant cover percentage, relative contributions of plant growth forms (Tueller et al., 1991), plant rooting depths, and species distributions within Mojave Desert plant communities (Wallace et al., 1974), net annual storage of water in most Mojave Desert soils is zero (Nowak et al., 2004).

Using a mixed canopy structure with fixed percentages of evergreen shrubs, drought-deciduous shrubs, perennial forbs and grasses, and winter annuals, similar to that observed at EAFB, an aggregate of individual leaf area indices, LAIt, was calculated using

Formula 4[4]
where i is the counter for growth form, canopy% is the percentage of total canopy populated by the ith growth form; cover% is the total ground cover; and LAIm is the multiplier that converts leaf biomass for each growth form in LAI (Kemp et al., 1997). Four different cover percentages were used (0, 10, 20, and 30%). Total PET was partitioned into PT and PE using the method of Kemp et al. (1997), who partitioned the solar radiation component of the energy budget by canopy interception:

Formula 5[5]
and,

Formula 6[6]
where k is an empirical parameter related to radiation extinction by the canopy (assumed to be ~0.63 for LAI of 0.20–2.0). Table 1 shows the LAIt and partitioning percentages for the static canopy structure for EAFB.


View this table:
[in this window]
[in a new window]
  		Table 1. Partitioning of potential evapotranspiration into components of plant transpiration (PT) and soil evaporation (PE) for a specific canopy structure.

 
Using the description given above of desert plant phenology, water uptake was assigned as percentages of PET to evergreen shrubs (25% of PET), winter annuals (20–37% of PET, depending on time of year and winter rainfall), and drought-deciduous shrubs (17% of PET, also depending on the time of year). Winter precipitation is important for seed germination, so uptake by winter annuals and drought-deciduous shrubs was reduced and zeroed, respectively, for those years when winter rainfall (1 November–28 February) was <26.8 mm (~1 inch). Table 2 shows the activity period for the growth forms and Fig. 1 illustrates how PT was distributed throughout the year for a plant cover scenario of 30%.


View this table:
[in this window]
[in a new window]
  		Table 2. Activity period for components of potential evapotranspiration (PET).

 

Figure 1
View larger version (29K):
[in this window]
[in a new window]
  		Fig. 1. Time series of potential plant transpiration (PT) for a 10-yr period, assuming 30% ground cover.

 
Numerical Implementation
Model and Dataset Descriptions
HYDRUS-2D (Simunek et al., 1999), which numerically solves Richards' equation for variably saturated porous media, was used for the numerical testing of covers and the design sensitivity analysis. The model was assigned a free drainage (unit gradient) bottom boundary condition that represents the base of the cover. The atmospheric upper boundary includes the 10-yr precipitation record and the estimates of PET, partitioned into PE and PT.

Six geometries of the cover were selected for the Monte Carlo simulations. Each geometry combined different thicknesses of the surface soil and underlying grus layer (Table 3). For each of the six geometries, the number of nodes was likewise increased [nodes = 2(soil thickness) + 1], keeping similar the average node thickness (~0.5 cm). A variable grid density was used so that the upper boundary had a node thickness of <1 mm and the bottom boundary had a node thickness of ~1 cm.


View this table:
[in this window]
[in a new window]
  		Table 3. Cover configurations used in the simulations.

 
Monte Carlo Implementation
The design sensitivity analysis incorporates an uncertainty approach to evaluate how slight changes in cover design and material properties might affect the long-term deep water flux. We considered different cover designs by classifying some parameters as categorical and others as stochastic. The categorical parameters were soil type, cover thickness, and plant cover percentage. These were varied within categories because, generally, they can be controlled by field practices (e.g., thickness of a known soil material can be specified and field constructed). The results of the laboratory analyses yielded four soils (three surface soils and the underlying grus) with distinctly different saturated hydraulic conductivities. The surface soils thus fell into one of three classes. The cover thicknesses were discussed above and are presented in Table 3, and the degree of plant cover was varied from 0 to 30% in 10% increments. Therefore, the number of possible combinations of categorical variables is determined from the surface soils (three types), geometries (six thicknesses), and plant cover percentages (four percentages), or 72 combinations total. The stochastic parameters included three hydraulic parameters from Eq. [1] and [2]: the two shape parameters ({alpha} and n) for the water retention curve and Ks.

Using the hydraulic property values obtained from laboratory analyses for each of the four soil types, a mathematical transformation was then selected for each variable such that the transformed variable yielded a normal distribution function. Lognormal and log ratio transformations were used for {alpha} and Ks, respectively, for all soils. The lognormal transformation is given as

Formula 7[7]
and the log ratio transformation is given as

Formula 8[8]
where X is the untransformed variable, with limits of variation from A to B (A < X < B) (Johnson, 1987); A and B were obtained from the laboratory analyses. The van Genuchten shape parameter n did not require a transformation to yield a normal probability density function. After the appropriate transformation was established, the laboratory values were used to determine the mean and sample covariances. Because each transformed variable is normally distributed, a multivariate normal distribution was selected to represent the joint probability density function for the uncertain variables. Random deviates for use in the Monte Carlo simulations were generated by applying

Formula 9[9]
where µ is the vector of mean values, T' is the transpose of the factored covariance matrix, and z is a vector of independent standard normal deviates. The statistics used for the Monte Carlo simulations are shown in Table 4. A detailed discussion of the Monte Carlo algorithm can be found in Carsel and Parish (1988).


View this table:
[in this window]
[in a new window]
  		Table 4. Statistical description of the four soils under study.

 
Figure 2 shows the frequency distribution of the simulated Ks for the surface soils and underlying grus used in the Monte Carlo simulations. Differences in the mean Ks can be seen for each of the three surface soil types (Fig. 2a), which are designated as Surface Soils 1 through 3. Though an insufficient number of samples for the surface soils were available to verify the lognormal distributions shown in Fig. 2a, for the purposes of illustrating the effects of different soil properties on deep flux, we assumed that the samples followed a lognormal distribution and generated Ks values accordingly. The distribution of the grus soil (Fig. 2b) did show a more distinct lognormal-type distribution, with hydraulic conductivity results that ranged from <10 to >1800 cm d–1 (mean value = 36 cm d–1). The multivariate normal distribution for the transformed variables yielded a partial correlation when the uncertain variables were randomly generated. As an example, Fig. 3 shows the relationship for 1000 randomly sampled values of {alpha} and Ks for Surface Soil 2. The weak correlation found between {alpha} and Ks has also been reported by other researchers (e.g., Smith and Diekkrüger, 1996; Avanidou and Paleologos, 2002).


Figure 2
View larger version (23K):
[in this window]
[in a new window]
  		Fig. 2. Probability distribution functions (PDF) of lognormalized Ks (saturated hydraulic conductivities) for (a) surface soils and (b) underlying grus.

 

Figure 3
View larger version (18K):
[in this window]
[in a new window]
  		Fig. 3. The relationship between shape parameter {alpha} and Ks (saturated hydraulic conductivity) for 1000 randomly sampled values for Surface Soil 2.

 

   RESULTS AND DISCUSSION
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
The results of the simulations have been reduced, compiled, and represented in three nomograms that show how deep flux changes as a function of cover thickness and plant cover percentage (Fig. 4 ). The general shapes of the nomograms are very similar, though some differences are apparent between soil types. For example, a large percentage (75%) of the total combinations of cover thickness and plant cover percentage for Surface Soil 1 led to deep fluxes of <0.5 cm yr–1, whereas fewer combinations of input parameters with Surface Soils 2 and 3 (65 and 25%, respectively) with higher hydraulic conductivities resulted in this low level of deep flux. Each nomogram has specific regions where the ensemble average of the MAF (mean annual flux) is high (>1 cm yr–1), and regions where the MAF is significantly reduced (i.e., <0.5 cm yr–1). The nomograms show clear trends toward decreasing MAF for thicker covers and those that have greater plant cover. Thicker covers clearly reduce deep flux because the water holding capacity of the thicker soil is higher, allowing the cover to act as a "store and release" feature that holds the water within the profile and makes it available for soil evaporation or plant transpiration. In very thin covers, like those at 46 cm, flux is higher regardless of the presence or absence of plants; in this case, water percolates below the base of the cover before ET processes can act. As the covers thicken, the effect of the higher water holding capacity can be seen in every combination of plant cover percentage and surface soil type, although it appears that adding soil to the cover and increasing the water holding capacity provides marginal improvements on cover performance when thickness reaches a certain threshold, ranging between 70 and 85 cm depending on soil type.


Figure 4
View larger version (29K):
[in this window]
[in a new window]
  		Fig. 4. Nomogram showing average flux for 10-yr simulations. Units are centimeters per year for all three graphs.

 
Deep flux was dependent on the presence of plants; without plants as a mechanism for removal of water, deep flux was always >0.5 cm yr–1 (Fig. 4). Increasing the plant cover from 0 to 10% significantly reduced deep flux regardless of the cover thickness or the soil type, but additional increases in plant cover beyond 10% had a diminishing effect on flux. Using Surface Soil 1 as an example, Fig. 5 shows that more intensive revegetation yields progressively fewer benefits when considered in terms of flux reduction only. Here, adding vegetation to the 46-cm-thick cover reduced the deep flux only marginally (from 1.25 to 0.59 cm yr–1). Presumably the water percolated below the root zone and out of the model domain before being transpired. Thicker soil covers, for example 107 cm thick, experienced a greater reduction in deep flux as the plant cover was increased from 0 to 30% (from 0.59 to 0.02 cm yr–1). For all cases with different cover thicknesses, the reduction in MAF was largest as the plant cover increased from 0 to 10%, with further reductions diminishing as plant cover increased.


Figure 5
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 5. Deep flux as functions of plant cover percentage and cover thickness for Surface Soil 1.

 
Although the nomograms in Fig. 4 provide a rapid tool for estimating MAF as field conditions change, deviations from mean values are also important. Figure 6 shows the trends in CV for flux, calculated as the quotient of the standard deviation and the mean. The results show that the CV increases as the MAF decreases. The hydrologic system shows higher sensitivity to field conditions as the cover thickness increases from 45 to 72 cm (especially for plant cover exceeding ~15%), and as plant cover increases from 0 to 10% for covers exceeding ~90 cm. Although a higher CV for lower fluxes might indicate the need to better characterize field soil, it is interesting to note the relationship between the MAF and standard deviation, shown in Fig. 7 . This plot shows a nearly constant relationship between mean and standard deviation (i.e., constant CV) for MAF values greater than ~0.25 cm yr–1; however, as the MAF is reduced further, from ~0.25 to 0 cm yr–1, the nearly linear relationship breaks down and the change in standard deviation dominates the change in CV. This outcome may not have a mathematical significance other than the likelihood that it follows the input distribution of the stochastic variables (lognormal for Ks and {alpha}, and normal for n); however, it does become significant if relating modeling results to enhanced field practices. For example, if the extent of field characterization (i.e., a larger or smaller number of samples would be needed to account for spatial variability) is dependent on the magnitude of the MAF values and their associated uncertainties, then the nonlinear relationship between MAF and standard deviation might become important if the cover is required to prevent deep fluxes of <0.5 cm yr–1. Or, if the differential of MAF from 0.1 to 0.2 cm yr–1 (e.g., 100%) is associated with a differential of SD much greater than 100%, then field practices might again be tied closely to the modeling results. If the cover design is engineered to reduce MAF several times less than a regulatory standard, however, then the variance becomes less of an issue.


Figure 6
View larger version (29K):
[in this window]
[in a new window]
  		Fig. 6. Nomogram showing the coefficient of variation on flux for 10-yr simulations. Units are in percent for all three graphs.

 

Figure 7
View larger version (15K):
[in this window]
[in a new window]
  		Fig. 7. Relationship between mean annual flux and the standard deviation of flux for each of the three soil types.

 

   CONCLUSIONS
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
Alternative covers for landfills are ecologically engineered systems that combine attributes of ecosystem function and engineering design to meet specific performance goals. The nomograms show clear trends toward decreasing MAF for covers that thicken and those that have higher vegetative cover. In the former case, thicker covers clearly reduce estimates of deep flux, because the water holding capacity of the cover increases as it becomes thicker. In covers that exceed 61 cm thick, the water holding capacity of the profile was high enough to store more of the winter-dominated precipitation for later transpiration; however, in the thinner covers (i.e., 46 cm) with lower water holding capacity and a shallower root zone, a larger amount of flux was predicted—more than twice the rate predicted for thicker covers. As expected, a significant dependence on the presence of plants was observed and no combination of soil type and thickness tested without plants reduced estimates of flux below 0.5 cm yr–1. Most of the increase in performance due to the presence of plants was realized by increasing plant cover from 0 to 10%, which could be considered a threshold vegetative cover. Using the field data collected at this site and the results from the numerical modeling, an optimum design would be to use Surface Soil 1 at a thickness exceeding ~80 cm with at least 10% plant cover.

These observations show that reductions in flux are maximized only by considering the interactions between soil hydraulic properties, cover thickness, and plant cover percentage. It is apparent that relative benefits of adding more soil to a landfill cover will reduce deep flux only to a point. Likewise, establishing some vegetative cover will significantly reduce deep flux, but marginal improvements in flux reduction were seen when plant cover exceeded 10%. By constructing a cover thick enough to store winter-dominated precipitation, and then establishing a plant canopy capable of removing that water, deep flux can be minimized. Of course, vegetative communities take many years to establish, so deep flux may therefore be elevated for some period of time until that community develops sufficiently.

The amount of data used in the analyses is arguably sparse in some areas (e.g., soil hydraulic properties); however, the intent of this research was to develop and illustrate an approach for designing effective landfill covers, not to design any cover per se. These procedures could thus be used for any site where soil properties, plant community characteristics, and climate are appropriate for ET covers, simply by identifying design parameters (material thickness and plant cover), and then incorporating the material variability into the analyses. As with any numerical prediction of hydrologic processes, however, predictions need to be viewed in the context of limitations in the model itself, the validity of the conceptual model, and the data used as input.


   ACKNOWLEDGMENTS
 
This study was funded under Contract DACA05-02-C-0009 by the U.S. Corp of Engineers, Sacramento District. Further support was provided by the Center for Arid Lands Environmental Management and the Division of Hydrologic Sciences at the Desert Research Institute. Thanks to Brad Lyles for his assistance in sample collection, John Goreham for assistance analyzing soil samples, and Darren Meadows for assistance in analyzing multistep outflow experiments.


   REFERENCES
TOP
 EXECUTIVE SUMMARY
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Young, M. H.
Right arrow Articles by Oyelowo, L.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Young, M. H.
Right arrow Articles by Oyelowo, L.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Young, M. H.
Right arrow Articles by Oyelowo, L.
Related Collections
Right arrow Recharge
Right arrow Stochastic Processes
Right arrow Disposal Facilities

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome