Field Data and Model Predictions for a Monolithic Alternative Cover , PH: (608) ; FAX: (608) ;

Transcription

1 Benson, C., Bohnhoff, G., Ogorzalek, A., Shackelford, C., Apiwantragoon, P., and Albright, W. (25), Data and Model Predictions for an Alternative Cover, Waste Containment and Remediation, GSP No. 142, A. Alshawabkeh et al., eds., ASCE, Reston, VA, Data and Model Predictions for a Monolithic Alternative Cover By C. H. Benson 1, G. L. Bohnhoff 2, A. S. Ogorzalek 3, C. D. Shackelford 4, P. Apiwantragoon 2, and W.H. Albright 5 1 Professor, Geological Engineering, Univ. of Wisconsin-Madison, Madison, WI, USA 5376, PH: (68) ; FAX: (68) ; benson@engr.wisc.edu 2 Graduate Research Asst., Geological Engineering (Bohnhoff) or Civil and Environ. Engineering (Apiwantragoon), Univ. of Wisconsin-Madison, Madison, WI, 5376, USA, glkrupp@students.wisc.edu, papiwantrago@students.wisc.edu 3 Graduate Research Asst., Dept. of Civil Engineering, Colorado State Univ., Fort Collins, CO, 8523, USA, ogie1@engr.colostate.edu 4 Professor and Director, Rocky Mtn. Hazardous Substance Research Center, Dept. of Civil Engineering, Colorado State Univ., Fort Collins, CO, 8523, USA, shackel@engr.colostate.edu 5 Hydrogeologist, Desert Research Institute, Reno, NV, USA, bill@dri.edu Abstract Water balance data from a test section simulating a monolithic alternative cover were compared to predictions made with two numerical models: UNSAT-H and Vadose/W. Onsite data were used as model input to the greatest extent practical. More accurate predictions were obtained with Vadose/W than UNSAT-H. Surface runoff was overpredicted appreciably by UNSAT-H, which affected all subsurface hydraulic processes. In contrast, Vadose/W accurately predicted surface runoff, evapotranspiration, and the temporal variations in soil water storage. However, neither model predicted percolation accurately. Both models also failed to capture a key change in the transpiration pattern during the last winter-summer period of the study. Differences in the method used to simulate precipitation intensity appear partly responsible for the difference in accuracy by which the two models predict surface runoff. Simulations were conducted with the lower boundary condition as a unit gradient or a seepage face to evaluate how this boundary condition affects predictions of the water balance. Essentially the same predictions were obtained regardless of the lower boundary that was used. Introduction Final covers employing water balance principles are being considered with increasing frequency for closure of waste containment facilities in lieu of conventional covers employing resistive barriers such as compacted clay layers and geomembranes (Benson et al. 21, Dwyer 21, Zornberg et al. 23, Albright et al. 24). Covers employing water 1

2 balance principles rely on the ability of fine-textured soils to retain infiltrating water with minimal drainage during periods of elevated precipitation and minimal evapotranspiration as well as the capability of plants and the atmosphere to remove the stored water via transpiration and evaporation. These covers are referred to by names such as store-andrelease covers, evapotranspirative covers, water balance covers, and alternative covers. The alternative cover nomenclature is used herein. Design of an alternative cover generally consists of identifying the thickness of cover needed to store water that infiltrates during the cooler and wetter months or extreme climatic events and evaluating whether evapotranspiration can remove the stored water during the growing season. During preliminary design, hand calculations employing approximate methods are used to size the thickness of the cover (Khire et al. 2, Benson and Chen 23). The design is then checked and refined using numerical models and, in some cases, a field demonstration is conducted to verify that the design is adequate (e.g., Chadwick et al. 1999, Khire et al. 1997, 1999, Zornberg et al. 23, Albright et al. 24). There are a variety of numerical models that can be used, but model accuracy has only been evaluated in a limited number of cases (Fayer et al. 199, Khire et al. 1997, 1999, Roesler et al. 22, Scanlon et al. 22). Moreover, in many of the model evaluations, critical input parameters (e.g., hydraulic properties, characteristics of the vegetation) have not been measured or have been used as calibration parameters. In this paper, predictions made with two numerical models developed specifically for evaluating the hydrology of covers (UNSAT-H and Vadose/W) are compared to field data from an instrumented test section simulating a monolithic alternative cover. Input to the models was based on measurements made in the field and laboratory to the greatest extent possible so as to provide a direct assessment of the predictive capabilities of the models. Materials and Methods Site The field site is located in Altamont, CA, USA, which is approximately 65 km east of San Francisco. The climate at the site is semi-arid, with an average annual precipitation of 358 mm and a precipitation-to-potential-evapotranspiration ratio (P/PET) of.31. Precipitation at Altamont occurs almost exclusively as rain. The average daily temperature ranges from 2 o C in January to 32 o C in August (Albright et al. 24). An alternative cover is being evaluated at the field site as part of the US Environmental Protection Agency s (USEPA) Alternative Cover Assessment Program (ACAP). The cover profile consists of a 15-mm-thick surface layer, a 91-mm-thick storage layer, and a 3- mm-thick layer of interim cover soil comparable to the existing soil overlying waste at the site (Albright et al. 23, 24). Vegetation on the cover consists of a mixture of grasses. Albright et al. (24) describe the species of grass being used. A test section was constructed in 2 to evaluate the effectiveness of the cover under near full-scale conditions. All three layers were constructed with crushed claystone from an on- 2

3 site borrow area that classifies as low plasticity clay in the Unified Soil Classification System (Gurdal et al. 23). The storage layer serves as the primary reservoir for infiltrating water, although the surface layer and interim cover layer provide storage capacity as well. The storage layer was compacted at an average water content 2.3% dry of optimum and at an average relative compaction of 87%, based on standard Proctor (ASTM D 698). A detailed description of methods used to construct the test section can be found in Bolen et al. (21) and Roesler et al. (22). A large (1 m x 2 m) pan-style lysimeter lined with a geomembrane and drainage composite was used for monitoring the water balance of the test section (Benson et al. 21, Albright et al. 24). The drainage layer at the base of the lysimeter is known to form a capillary break, which is unrealistic for hydrology problems involving a deep vadose zone. However, a capillary break commonly exists between interim cover soil and the waste because the hydraulic properties of waste mimic those of coarse-textured soils (Benson and Wang 1998). Three nests of low-frequency (4 MHz) time domain reflectometry (TDR) probes are used to measure water content and one nest of thermal dissipation sensors is used to monitor matric suction. A weather station is used for meteorological monitoring. A more detailed description of the instruments used to monitor the lysimeter can be found in Benson et al. (21) and Albright et al. (24). The water balance is assumed to consist of five components: precipitation (P), runoff (R), evapotranspiration (ET), soil water storage (SWS), and percolation (P r ). Soil water storage is computed by spatially integrating the water content data. Evapotranspiration (ET) is computed as the residual of the water balance: ET = P R P r ΔSWS (1) ET computed with Eq. 1 includes actual ET and the net error in the other water balance quantities. An interval of 1 d was used when computing ET with Eq. 1. Hydraulic Properties Undisturbed thin-wall (71-mm) tube and block (2-mm diameter) samples were collected from each lift of the test section during construction. Each sample was tested in the laboratory to determine the saturated hydraulic conductivity (K s ) and the soil water characteristic curve (SWCC) for desorption. Flexible-wall permeameters (ASTM D 584) were used to measure K s, with an average applied effective stress of 21 kpa and a hydraulic gradient of 1. SWCCs were measured as per ASTM D 6836 using a pressure-plate extractor and a chilled-mirror hygrometer. van Genuchten s equation was fit to the water retention data and the van Genuchten parameters are used herein to describe the SWCCs. Methods used to sample the soils and conduct the tests are described in detail in Gurdal et al. (23). All of the test data are compiled in Gurdal et al. (23) and a summary is included in Albright et al. (24). 3

4 Evaluation of the hydraulic properties suggested that the cover profile can be described with three layers: (i) the surface layer and an upper portion of the storage layer, (ii) the lower portion of the storage layer, and (iii) the interim cover layer (Gurdal et al. 23). Hydraulic properties of these layers are summarized in Table 1 in terms of geometric means (K s and van Genuchten s α parameter) and arithmetic means (saturated water content, θ s ; residual water content, θ r, van Genuchten s n parameter). Geometric means were used for K s and α because these parameters were determined to be log-normally distributed. Arithmetic means were used for the other parameters because they were determined to be normally distributed (Gurdal et al. 23). Standard deviations for K s, α, and n are shown in parentheses. Table 1. hydraulic properties of cover soils. Std. deviations in parentheses. Saturated Hydraulic SWCC Parameters Conductivity (cm/s) Layer Surface Layer and Upper Portion of Storage Layer Lower Portion of Storage Layer Thickness (mm) θ r θ s α (m -1 ).5 (1.9) n (.11) (.65) (.13) Interim Cover (.9) (.39) Note: standard deviations for α and n are of naturals logarithms. 5.3x1-7 (-) 2.2x1-6 (-) 4.5x1-7 (2.53) 3.x1-6 (2.57) 1.1x1-4 (-) Two samples were collected from the surface layer as blocks each year after construction to evaluate how weathering and vegetative growth affected K s and the SWCC. No change in the SWCC has been observed, but K s has increased over time due to formation of macropores caused by wet-dry cycling and biota intrusion. Thus, three different K s (geometric means) are reported for the surface layer and upper portion of the storage layer (one per year since construction) in Table 1. SWCCs representing in situ conditions were also constructed using the water contents and matric suctions measured in the field (i.e., with the TDR probes and thermal dissipation sensors). van Genuchten parameters for these field-fit SWCCs are in Table 2. Table 2. Parameters of field-fit soil water characteristic curves. -Fit SWCC Parameters Layer θ r θ s α (m -1 ) Surface Layer and Upper Portion of Storage Layer Lower Portion of Storage Layer Interim Cover n 4

5 Vegetation Properties Measurements were made annually at the site to define the peak leaf area index (LAI) and the root density profile. LAI measurements were made in the laboratory on clippings using a Li-Cor LI-31C area meter and in the field using a Li-Cor LAI-2 plant canopy analyzer. Rooting depth and the root density (R) profile were measured in the laboratory on field specimens collected using the Weaver-Darland box method. The peak LAI was determined to be approximately 1.5. Roots extended to the bottom of the storage layer. The root density data were fit with the exponential model: R = ae -bz +c (2) where a, b, and c are empirical parameters and z is depth (m). Average values for these parameters were determined to be as follows: a =.44, b = 8. m -1, and c =.53. Models Two numerical models were used in this comparison: UNSAT-H (Fayer 2) and Vadose/W (Krahn 24). Both models were developed specifically for evaluating the hydrology of earthen covers with vegetation. The UNSAT-H simulations were conducted with WinUNSAT-H, the Microsoft Windows implementation of UNSAT-H ( that uses UNSAT-H v2.4 at its core. The Vadose/W simulations were conducted with the version of Vadose/W incorporated in GeoStudio 24 ( UNSAT-H and Vadose/W simulate water flow under variably saturated conditions by solving a modified form of Richards equation: θ ψ ψ t = K z T ψ + K ψ + q vt S(z, t) z (3) where θ is volumetric water content, t is time, z is the vertical coordinate (z = at the ground surface), ψ is matric suction, K T is a combined liquid and vapor unsaturated hydraulic conductivity, K ψ is the unsaturated hydraulic conductivity, q vt is water vapor flux, and S is a sink term for extraction of pore water by plant roots. UNSAT-H solves Eq. 3 using the finite-difference method, whereas Vadose/W uses the finite element method. Only one-dimensional simulations were conducted in this study. Boundary and Initial Conditions A flux boundary condition was applied at the surface for both models to simulate infiltration of precipitation and evaporation from the soil surface. The evaporation component depends in part on the amount of vegetative cover, as characterized by LAI. A detailed description of the algorithms used to implement the flux boundary in UNSAT-H and Vadose/W can be found in Fayer (2) and Krahn (24). A unit gradient condition was used for the lower boundary condition in UNSAT-H. For Vadose/W, the lower boundary was a unit gradient or a seepage face. The seepage face 5

6 boundary was used based on recommendations in Scanlon et al. (22), which indicate that a seepage face boundary more accurately represents the capillary break formed by the drainage layer at the base of a lysimeter. A seepage face boundary was not used with UNSAT-H, because the model does not include this option. All simulations began on January 1, 21. Initial conditions were defined using matric suctions (UNSAT-H and Vadose/W) and soil temperatures (Vadose/W) existing in the field on this date (Albright and Benson 23, Albright et al. 24). Meteorological Input Daily meteorological data were input to the models. Except for precipitation, the data were obtained from on-site instruments. The precipitation data were obtained from a local weather station operated by the US National Oceanic and Atmospheric Administration (NOAA). Precipitation data from NOAA were used because of issues related to shielding of the on-site rain gage. All of the meteorological data are summarized in Albright and Benson (23). For the simulations with UNSAT-H, precipitation was applied using the default condition (1 mm/h beginning at midnight until the total precipitation is reached). For Vadose/W, precipitation was applied with a daily sinusoidal pattern. Hydraulic Properties The mean hydraulic properties in Table 1 and the field-fit SWCCs were used as input for both models. For UNSAT-H, K s and the van Genuchten parameters describing the SWCC were input directly. The van Genuchten-Mualem model was used for the unsaturated hydraulic conductivity function with the pore interaction parameter set at.5. For Vadose/W, the spline-function option was used to describe the SWCC. The spline function was fit to the SWCC as defined by the van Genuchten parameters in Table 1 and 2. The unsaturated hydraulic conductivity was then estimated using the van Genuchten estimation method provided in Vadose/W. Vegetation The peak LAI in both models was set at 1.5 based on the field and laboratory measurements. LAI was assumed to increase linearly from to 1.5 over the first 3 d of the growing season, decrease linearly from 1.5 to during the last 3 d of the growing season, and remain constant in between. For the first growing season, however, the peak LAI was set at 1. to account for the immaturity of the vegetation. The growing season for the vegetation (September 1 July 1) was obtained from the Soil Conservation Service. Root growth was defined based on recommendations in Roesler et al. (22). Roots were assumed to grow at 3 mm/d from the start of the growing season until the interim cover soil was reached. In UNSAT-H, an exponential distribution was assigned for the root density (Eq. 2) using the field-measured parameters described previously. The triangular root density option was used in Vadose/W using the peak root density measured in the field. 6

7 A plant limiting function is used in both models to relate water availability and transpiration rate as a function of the anaerobiosis point, limiting point, and wilting point (Fayer 2, Krahn 24). The anaerobiosis point was set at 8 kpa and the limiting point was set at 5 kpa based on recommendations in Winkler (1999). The wilting point was set at 6.25 MPa, which is the largest matric suction measured in the deeper (> 3 mm) portion of the root zone during the monitoring period. Spatial and Temporal Discretization The spatial-and-temporal discretization used in the simulations was set to achieve a target mass balance error of less than 1 mm/yr. For both models, this criterion required a nodal spacing or element thickness of 1 mm at the ground surface and layer boundaries. For UNSAT-H, the maximum time step was set at.25 h and the minimum time step was set at 1-5 h. For Vadose/W, the maximum time step was set at 24 h and the minimum time step was set at.24 h. The resulting run times for a three-year simulation were on the order of 15 h for WinUNSAT-H and 12 h for Vadose/W. Results and Discussion UNSAT-H Simulations Simulations were conducted with both models for the period between January 1, 21 and September 3, 23 (nearly three years and three growing seasons), which was the extent of the field data at the time the modeling study commenced. Predictions obtained with UNSAT-H using the mean and field-fit hydraulic properties are shown in Fig. 1 in terms of the five primary water balance quantities: cumulative precipitation (Fig. 1a), cumulative runoff (Fig. 1a), cumulative evapotranspiration (Fig. 1b), soil water storage (Fig. 1c), and cumulative percolation (Fig. 1d). The field data are also shown in Fig. 1. Surface runoff (SRO) is over-predicted for both parameter sets, with the over-prediction being larger for the mean parameters. Less than 1 mm of SRO was measured in the field during the monitoring period, whereas UNSAT-H predicted 25 mm of SRO with the fieldfit parameters and more than 4 mm with the mean parameters. This error is significant because an inaccurate prediction of SRO corresponds to an inaccurate prediction of infiltration (infiltration is the compliment to surface runoff for UNSAT-H and Vadose/W, since both models ignore interception by the plant canopy). Consequently, the volume of water to be managed by the cover is incorrect, and all subsequent water balance quantities are incorrect. A similar observation was reported by Roesler et al. (22). They attributed the overprediction of SRO to the surface layer with a K s that was too low. They obtained better predictions by increasing K s of the surface layer. By making this adjustment, however, the K s effectively became a calibration parameter rather than a soil property. In the simulations described here, the saturated hydraulic conductivity of the surface layer was based on measurements on relatively large specimens (2 mm diameter) trimmed from hand-carved block samples. These specimens included numerous macropores formed by weathering and 7

8 Precipitation and Runoff (mm) Evapotranspiration (mm) (a) (b) Precipitation Precipitation and Runoff (mm) Cumultaive Evapotranspiration (mm) Soil Water Storage (mm) (c) Soil Water Storage (mm) 4 (d) 4 Percolation (mm) Percolation (mm) 1/1/1 7/1/1 1/1/2 7/1/2 1/1/3 7/1/3 1/1/4 Fig. 1. water balance data and model predictions from UNSAT-H using mean or field-fit () hydraulic properties: (a) precipitation and surface runoff, (b) evapotranspiration, (c) soil water storage, and (d) percolation. 8

9 biota intrusion, both of which contributed to increases in K s. Thus, the K s of the surface layer used as input is believed to be representative of the field condition. Consequently, other factors besides K s may be responsible for the over-prediction of surface runoff. One of these factors appears to be precipitation intensity, as described in the subsequent section on Vadose/W. Evapotranspiration (ET) obtained from the water balance data and predicted with UNSAT- H is shown in Fig. 1b. UNSAT-H under-predicts ET for both sets of hydraulic properties, with a greater under prediction for the mean parameters. ET is under-predicted because too little water entered the soil and was available for evaporation or transpiration. The over-prediction of SRO is also reflected in the predictions of soil water storage (SWS). In the field, SWS varies approximately 1 mm as water is stored in the cover and then extracted by evaporation and transpiration. In contrast, during the first two years SWS predicted with the mean parameters exhibits a much smaller change in storage throughout the year because too little water is entering or leaving the cover. Only in the third wet season does the SWS predicted with the mean parameters increase appreciably. The increase in SWS during Winter 23 occurs in response to very heavy precipitation events at the end of 22 along with the higher K s of the surface layer. Even though runoff was over-predicted during this period, the model allowed an appreciable fraction of precipitation to enter the cover, resulting in a large increase in SWS relative to that occurring during the same period in the previous two years. More accurate predictions of SWS were obtained with the field-fit parameters, primarily because the prediction of SRO was more accurate using the field-fit parameters. The field SWS also exhibits a different pattern in Spring and Summer 23 than in the previous two years, and this change in behavior is not captured by UNSAT-H. Rather than diminish nearly monotonically, the field SWS continues to increase through March 23 and then remains high during the summer months. This continual increase in SWS is attributed to elevated precipitation in Winter 23, which was approximately twice that received during the previous years. Apparently, the larger amount of precipitation received during 23 exceeded the transpiration capacity of the plants and allowed soil water to accumulate. The phenology of the vegetation also appears to have been altered, resulting in cessation of transpiration much earlier than normal (hence, the lack of a drop in SWS during spring and summer). Changes in phenology in response to changes in climate are not accounted for in most cover models in use today (including UNSAT-H and Vadose/W). Consequently, the model predictions exhibit a similar pattern of water extraction each year regardless of the effect of climate on phenology, which can result in gross deviations between predicted and measured SWS, such as those shown at the end of the record in Fig. 1c. Comparison of measured and predicted percolation in Fig. 1d indicates that UNSAT-H under-predicts percolation for both sets of parameters. Over-prediction of SRO is one reason for the under-prediction of percolation (too little water is available to migrate to the base of the cover). Preferential flow may be partly responsible because the rapid rise in 9

10 cumulative percolation occurs before the SWS peaks. Another possibility is that the field K s of the storage layer may be higher than the K s being input to the models. Even though relatively large samples were collected from the storage layer for K s testing, these samples may not have been large enough to truly represent the K s operative in the field. Vadose/W Simulations Simulations were conducted with Vadose/W over the same period used for UNSAT-H. Predictions obtained using the mean and field fit hydraulic properties and a unit gradient lower boundary are shown in Fig. 2 in terms of the five primary water balance quantities: cumulative precipitation (Fig. 2a), cumulative runoff (Fig. 2a), cumulative evapotranspiration (Fig. 2b), soil water storage (Fig. 2c), and cumulative percolation (Fig. 2d). The field data are also shown in Fig. 2. Vadose/W predicts surface runoff (SRO) more accurately than UNSAT-H even though the same hydraulic properties, vegetation data, and meteorological data were input to both models. The prediction is particularly accurate using the mean hydraulic properties. The method used in Vadose/W to distribute the precipitation appears to be responsible for the better accuracy. In these simulations, Vadose/W distributed the precipitation with a daily sinusoidal pattern. This approach results in precipitation with less intensity than the default application method in UNSAT-H (1 mm/h). Consequently, the infiltration capacity of the surface layer is exceeded only during the largest precipitation events and these events generate SRO (similar behavior occurs in the field, Fig. 2a). In contrast, when UNSAT-H is run using the default application method, the infiltration capacity is exceeded more frequently and SRO occurs too often. Good agreement also exists between the measured and predicted evapotranspiration (ET) as shown in Fig. 2b, particularly for the mean parameters. The only appreciable deviation is during the end of the record, when little ET occurred in the field despite an abundance of available water (see previous discussion in section on UNSAT-H and discussion below). The relatively good agreement between measured and predicted ET reflects the relatively accurate prediction of SRO. When SRO is predicted accurately, water can enter the cover and become available for evaporation and transpiration. Reasonably good agreement also exists between the predicted and measured soil water storage (SWS), especially when the mean hydraulic properties were used as input (Fig. 2c). The peak and minimum SWS are reproduced relatively accurately. The rate at which SWS is depleted is predicted more accurately during the second year than the first year, which may reflect the immaturity of the vegetation during the first year of the study. The good agreement between the measured and predicted SWS is most likely due to the accurate prediction of SRO. However, as with UNSAT-H, the unusual behavior at the end of the record is not captured. Vadose/W predicts a large depletion of SWS due to ET in Spring and Summer 23, whereas the SWS remained relatively constant in the field. 1

11 Precipitation and Runoff (mm) (a) Precipitation Precipitation and Runoff (mm) Evapotranspiration (mm) (b) Evapotranspiration (mm) Soil Water Storage (mm) (c) (d) Soil Water Storage (mm) Percolation (mm) Percolation (mm) 1/1/1 7/1/1 1/1/2 7/1/2 1/1/3 7/1/3 1/1/4 Fig. 2. water balance data and model predictions obtained using Vadose/W with mean or field-fit hydraulic properties: (a) precipitation and surface runoff, (b) evapotranspiration, (c) soil water storage, and (d) percolation. 11

12 Comparison of the measured and predicted percolation (Fig. 2d) shows that Vadose/W under-predicts percolation in a similar manner as UNSAT-H. Under-prediction of percolation concurrent with accurate predictions of the other water balance quantities suggests that differences likely exist between the hydraulic properties existing in the field and the laboratory-measured properties used as input to the models. Effect of Lower Boundary Simulations were conducted with Vadose/W with the lower boundary set as a unit gradient (UG) or a seepage face (SF) to evaluate how this boundary affects water balance predictions. For the UG condition, flow is forced to occur continuously under gravity driven conditions at a rate corresponding to the unsaturated hydraulic conductivity at the boundary. In contrast, the SF boundary permits flow to occur only when saturated conditions exist at the boundary. When saturation exists, flow from the SF boundary occurs under unit gradient conditions at a rate equal to the saturated hydraulic conductivity at the boundary. The UG boundary generally is used during design because the continuous flux from the boundary provides a conservative estimate of the percolation rate. The SF boundary has been suggested for lysimeters to more accurately represent the effect of the drainage layer (Scanlon et al. 22). Predictions made using both boundary conditions are shown in Fig. 3 along with the field data. hydraulic properties were used as input in both simulations. Changing the lower boundary condition has virtually no effect on the water balance predictions. When a SF boundary is used, predicted SRO is slightly lower, ET is slightly lower, and SWS is slightly higher than when the UG boundary condition is used. Percolation is slightly lower when the SF boundary is used because flow is not permitted until saturation occurs. However, the difference in cumulative percolation predicted by both models over the threeyear period is less than.4 mm, which is smaller than the mass balance error (1 mm/yr or 3 mm total). Thus, in contrast to the findings in Scanlon et al. (22), there is no practical difference between predictions made using a UG or SF boundary when simulating the water balance of the lysimeter at this site. Conclusions and Recommendations This paper has described a comparison between water balance predictions made with the numerical models UNSAT-H and Vadose/W and field data from a test section in a semiarid environment simulating a monolithic alternative cover. On-site meteorological data, vegetation properties, and hydraulic properties were used as input to the models to the greatest extent practical. UNSAT-H over-predicted surface runoff (SRO) using the measured properties and the default rainfall intensity as input. The over-prediction of SRO resulted in under-predictions of evapotranspiration (ET) and percolation, and smaller changes in soil water storage (SWS) than occurred in the field. In contrast, Vadose/W predicted SRO, ET, and changes in SWS relatively accurately using the measured properties as input. However, percolation was under-predicted by Vadose/W. 12

13 Runoff (mm) Evapotranspiration (mm) Soil Water Storage (mm) Percolation (mm) (a) (b) (c) (d) Unit Gradient Unit Gradient Unit Gradient Unit Gradient Seepage Face Seepage Face Seepage Face Seepage Face 1/1/1 7/1/1 1/1/2 7/1/2 1/1/3 7/1/3 1/1/ Runoff (mm) Percolation (mm) Evapotranspiration (mm) Soil Water Storage (mm) Fig. 3. water balance data and predictions from Vadose/W using mean hydraulic properties as input and unit gradient and seepage face boundaries at the base: (a) surface runoff, (b) evapotranspiration, (c) soil water storage, and (d) percolation. 13

14 A key factor contributing to the difference in accuracy in the SRO predictions appears to be the method by which the precipitation is applied, with Vadose/W applying precipitation at a lower intensity that more closely resembles conditions typically occurring in the field. This finding suggests that models should simulate the field precipitation intensity as closely as possible if the water balance is to be predicted accurately. The under-prediction of percolation by both models may be related to the saturated hydraulic conductivity input for the storage layer. More study is needed on the sensitivity of water balance predictions to the hydraulic properties used as input. A comparison was also made between predictions obtained from Vadose/W using a unit gradient or a seepage face as the lower boundary. These predictions were nearly identical, with the differences in percolation rate being smaller than the mass balance error. Thus, in contrast to findings reported by others, the lower boundary is not always a critical factor when simulating the water balance of lysimeters. Other factors, such as precipitation intensity and hydraulic properties, are likely to be more important. Both models failed to capture unusual behavior in the SWS in 23. Nearly twice as much precipitation was received in Winter 23 than in the previous two years and the additional precipitation apparently affected the phenology of the vegetation. Transpiration ceased in Spring 23, resulting in elevated SWS in Spring and Summer 23. In contrast, both models removed nearly all of the stored water via ET. The inability of these models to adjust phenology in response to changes in climate is a serious shortcoming that needs to be addressed through future research. Acknowledgements Financial support for this study was provided through the USEPA Science to Achieve Results (STAR) Program (Grant No. R ) as part of the USEPA s Rocky Mountain Hazardous Substance Research Center (HSRC) and by USEPA s Alternative Cover Assessment Program (ACAP). Dr. Mitch Lasat is the USEPA program manager for the HSRC and Mr. Steven Rock is the USEPA program manager for ACAP. The software packages used in this study were provided by Geo-Slope International Ltd. and the University of Wisconsin-Madison Geotechnics Laboratory. The findings in this paper have not been reviewed by USEPA or Geo-Slope. Endorsement by USEPA or Geo-Slope is not implied and should not be assumed. References Albright, W. and Benson, C. (23). Alternative Cover Assessment Program 22 Annual Report, Desert Research Institute, Reno, Nevada. Albright, W., Benson, C., G. Gee, Abichou, T., Roesler, A., and Rock, S. (23). "Evaluating the alternatives." Civil Engineering, 73(1), Albright, W., Benson, C., Gee, G., Roesler, A., Abichou, T., Apiwantragoon, P., Lyles, B., and Rock, S. (24), " water balance of landfill final covers." J. of Environmental Quality, in press. 14

15 Benson, C. (21). "Waste containment: Strategies and performance." Australian Geomechanics, 36(4), Benson, C., Abichou, T., Albright, W., Gee, G., and Roesler, A. (21). " evaluation of alternative earthen final covers." International J. of Phytoremediation, 3(1), Benson, C. and Chen, C. (23). "Selecting the thickness of monolithic earthen covers for waste containment." Soil and Rock America 23, Verlag Gluck auf GMBH, Essen, Germany, Benson, C. and Wang, X. (1998). Soil Water Characteristic Curves for Solid Waste. Environmental Geotechnics Report 98-13, Dept. of Civil and Environmental Engineering, University of Wisconsin-Madison. Bolen, M., Roesler, A., Benson, C., and Albright, W. (21). Alternative Cover Assessment Program: Phase II Report, Geo-Engineering Report No. 1-1, University of Wisconsin-Madison. Chadwick, D., Ankeny, M., Greer, L., Mackey, C., and McClain, M. (1999). " test of potential RCRA-equivalent covers at the Rocky Mountain Arsenal." Proc. SWANA 4th Annual Landfill Symp., SWANA Publication No. GR-LM-4, Silver Spring, MD, Dwyer, S. (21). "Finding a better cover." Civil Engineering, January Fayer, M., Rockhold, M., and Campbell, M. (1992). "Hydrologic modeling of protective barriers: comparison of field data and simulation results." Soil Sci. Soc. Am. J., 56, Fayer, M. (2). UNSAT-H Version 3.: Unsaturated Soil Water and Heat Flow Model - Theory, User Manual, and Examples, Report No. PNNL-13249, Pacific Northwest National Laboratory, Richland, WA, USA. Gurdal, T., Benson, C., and Albright, W. (23). Hydrologic Properties of Final Cover Soils from the Alternative Cover Assessment Program. Geo Engineering Report 3-2, University of Wisconsin-Madison. Khire, M., Benson, C., and Bosscher, P. (1997). "Water balance modeling of earthen landfill covers." J. of Geotech. and Geoenvironmental Eng., ASCE, 123(8), Khire, M., Benson, C., and Bosscher, P. (1999). " data from a capillary barrier in semiarid and model predictions with UNSAT-H." J. of Geotech. and Geoenvironmental Eng., ASCE, 125(6), Khire, M., Benson, C., and Bosscher, P. (2). "Capillary barriers: Design variables and water balance." J. of Geotech. and Geoenvironmental Eng., ASCE, 126(8), Krahn, J. (24), Vadose Zone Modeling with VADOSE/W, GEO-SLOPE International Ltd, Calgary, Alberta, Canada. Roesler, A., Benson, C., and Albright, W. (22). Hydrology and Model Predictions for Final Covers in the Alternative Cover Assessment Program 22. Geo Engineering Report 2-8, University of Wisconsin-Madison. Scanlon, B., Christman, M., Feedy, R., Porro, I., Simunek, J., and Flerchinger, G. (22). "Intercode comparisons for simulating water balance of surficial sediments in semiarid regions." Water Resources Research, 38, 59. van Genuchten, M. (198). "A closed-form equation for predicting the hydraulic conductivity of unsaturated soils." Soil Sci. Soc. of Am. J., 44,

16 Winkler, W. (1999). Thickness of Monolithic Covers in Arid and Semi-Arid Climates. MS Thesis, University of Wisconsin-Madison. Zornberg, J., LaFountain, L., and Caldwell, J. (23). "Analysis and design of evapotranspirative cover for hazardous waste landfill." J. of Geotech. and Geoenvironmental Eng., ASCE, 129(6),