earth pressure theories and modified slope stability methods have been used to carry out analysis and de- sign of geosynthetic reinforced soil walls. These methods have been extended for the case of earth- quake loading by simply introducing inertia force components equated to horizontal and vertical ground acceleration coefficients. These methods are limit equilibrium approaches that assume simplified failure mechanisms and employ either global factors of safety (AASHTO 1998, FHWA 1996) or partial factors (i.e. limit states design, BSI 1995) to design for in-service as opposed to collapse conditions. The performance of structures designed with factors of safety against col- lapse states (i.e. working stress conditions) are con- trolled by deformation limits that cannot be accounted for explicitly using limit equilibrium based methods. In addition, reinforced soil walls are complex systems typically involving a facing (such as concrete panels or stacked modular concrete blocks) soil backfill and horizontal layers of polymeric reinforcement. Limit equilibrium and semi-empirical analysis approaches require simplifying assumptions regarding the me- chanical properties of polymeric reinforcement prod- ucts and the interaction between components of rein- forced soil wall systems. Finally, these approaches provide very limited insight into the fundamental be- havior of reinforced soil walls since failure mecha- nisms are assumed a priori. Advanced numerical methods hold promise as a de- sign and research tool to investigate the entire re- sponse of reinforced soil retaining walls. An advan- tage of related computer codes is that appropriate constitutive models for polymeric materials and soils can also be implemented. Furthermore, numerical models can be used to carry out parametric analyses of the influence of wall geometry, facing type and me- chanical properties of the constituent materials on wall behavior. Calibrated numerical models can also be used to extend the database of carefully instrumented field or laboratory-scale structures and hence contrib- ute to the development of rational design methods based on conventional concepts of earth pressure the- ory. This paper provides a review of numerical model- ing of geosynthetic reinforced walls by the writers and others. The survey is limited to the finite element method (FEM) and finite difference method (FDM) in numerical codes used to simulate the behavior of geo- synthetic reinforced soil walls under static and dy- namic loading conditions. Review of numerical modeling of geosynthetic reinforced soil walls R.J. Bathurst & K. Hatami Royal Military College of Canada, Kingston, Ontario, K7K 7B4 Canada ABSTRACT: The use of geosynthetics to construct reinforced soil walls has increased dramatically in recent years. While the economic benefits of modern geosynthetic reinforced soil structures is proven, the mechanical behavior of these structures is less well understood. Correct numerical modeling offers the possibility to investi- gate the influence of a wide range of geometry and component material properties on structure performance. A challenge faced by modelers has been the treatment of the visco-elastic-plastic properties of polymeric reinforce- ment materials and the interactions between reinforcement layers, soil and structural facings. The paper reviews numerical modeling efforts of the writers and others to simulate the behavior of reinforced soil walls under both static and dynamic loading conditions. The paper is focused primarily on numerical modeling of geosynthetic re- inforced soil walls using the finite element and finite difference techniques. Implications of the results of some numerical modeling studies to current design practice are identified. The few cases in which the results of nu- merical models have been compared against physical tests are highlighted. Bathurst, R.J. and Hatami, K. Review of numerical modeling of geosynthetic reinforced soil walls, Invited theme paper, Computer Methods and Advances in Geomechanics: 10th International Conference of the International Association for Computer Methods and Advances in Geomechanics, 7-12 January 2001, Tucson, Arizona, USA, Vol. 2, pp. 1223-1232. 2 STATIC ANALYSIS 2.1 Numerical modeling studies without calibration The studies reviewed in this section are focused on parametric analyses of uncalibrated numerical models (i.e. model predictions are not compared against be- havior of physical measurements). Hermann and Al-Yassin (1978) described two dif- ferent FEM approaches to model the behavior of rein- forced-soil walls: discrete and composite approaches. In the composite approach, the reinforced soil struc- ture comprising the reinforced soil and metallic rein- forcement is homogenized and the metallic facing modeled with a structural element with bending ca- pacity. They argued that both the discrete and com- posite approaches could model the slippage and yielding of reinforcement, buckling of facing plate and nonlinear material behavior of the backfill material. They concluded that both approaches could be applied with equal accuracy. The advantage of the composite approach is reduced computational effort. However, this advantage is no longer a concern with modern computing platforms. A disadvantage of the composite approach is that there is no explicit information on the interaction of soil and reinforcement including bond stresses, stress concentration, edge effects and local- ized transfer of stress between the components. Con- sequently, the composite approach has been aban- doned by researchers. Naylor (1978) studied the effect of metallic rein- forcement strip-facing connection rigidity on the be- havior of a 10 m high reinforced-soil wall system us- ing a finite element model with a constant linear- elastic law for the soil and a linear elastic stiffness model with a Mohr-Coulomb (M-C) yield criterion for the soil-reinforcement interface. Naylor adopted an equivalent composite approach to model the rein- forced soil as a homogeneous mix of soil and metallic strips. The composite approach was justified by the high reinforcement density (i.e. small reinforcement spacing) of the model. Naylor included the flexibility of the reinforcement-facing connection, the density of reinforcement layers and wall foundation condition as variables in his parametric study. He concluded that a non-rigid reinforcement-facing connection allows the soil lateral pressure behind the facing to be reduced. This, in turn, reduces the maximum reinforcement load and shifts the peak reinforcement load location away from the facing and into the soil mass. He ob- served that increasing the reinforcement stiffness by a factor of 6 increased the maximum reinforcement load by 20% at end of construction. He concluded that Rankine theory underestimates the reinforcement load in heavily reinforced soil walls by a greater amount than for lightly reinforced soil wall systems. Naylor found that a compliant foundation (as opposed to a rigid foundation) results in a modest increase in the maximum reinforcement load and a change in the dis- tribution of vertical stresses at the bottom of the wall. The reinforcement load at the bottom of the wall as well as vertical soil pressures acting at the bottom of the reinforced soil mass were less with a compressible foundation but lateral movements were greater. Short- comings of the Naylor finite element study were that it did not simulate the staged construction of the wall or soil plasticity. Ho and Rowe (1993, 1996) and Rowe and Ho (1993, 1996, 1997, 1998) modified the finite element program AFENA to study the response of a hypotheti- cal 6 m high propped-panel wall under self-weight. The soil was modeled using a non-linear elastic model for the soil and linear elastic interface elements with M-C yield criterion. The wall configurations were constructed in 24 lifts. They noted that three different regions could be de- fined within the backfill. One region is contained be- tween the Rankine failure plane and the facing (active zone). A second region is a transition zone between the Rankine failure plane and a zero force line. The zero force line is a theoretical line beyond which no reinforcement force is required for equilibrium and corresponds to the theoretical stable slope defined by the angle of repose of the backfill frictional soil. The area under the zero force line was defined as the third zone. Numerical results by Ho and Rowe showed that the lateral earth pressures behind the wall were less than predicted values from Rankine theory and oscil- lated about the pressure line predicted by Coulomb theory. This observation was in agreement with ex- perimental data reported by Andrawes et al. (1990). Ho and Rowe noted that curtailment of reinforcement within the transition zone (i.e. zone 2) results in sharp increase of soil strain within the retained zone. They concluded that the effect of reinforcement is to prevent the development of horizontal strain in the soil. How- ever, they noted that the stiffness range of geosynthetic reinforcement products is not large enough to prevent soil elements from developing into a plastic state. Therefore, geosynthetic reinforced soil walls can be expected to develop a clear active zone within the re- inforced region behind the wall. Ho and Rowe found that the locus of maximum reinforcement load along the wall height was closer to the Rankine failure plane than a log-spiral shape. A very important practical ob- servation was that the variation of maximum rein- forcement load with depth was closer to a uniform distribution than that predicted by Rankine theory. As a result, the magnitude of reinforcement load was larger than values predicted by Rankine theory close to the top of the wall and considerably less near the bot- tom. The same conclusion has been reached by Bathurst and Hatami (1998b) who carried out similar parametric analyses using a very different numerical approach (i.e. FDM implemented by program FLAC - Itasca 1998). Ho and Rowe concluded that relatively large reinforcement loads at the top of reinforced soil walls that have been observed in field measurements should not be attributed to compaction effects. Their numerical simulations which did not include compac- tion of the backfill indicated a significant reinforce- ment load at the top of the wall which was attributed to interaction between the soil, reinforcement and facing. Ho and Rowe compared the horizontal soil stress with reinforcement load at each reinforcement level. They found that the sum of horizontal soil stresses is not entirely supported by the reinforcement layers but is also carried in part by the toe. They fur- ther concluded that a rigid foundation has a significant effect on the magnitude of reinforcement load, espe- cially close to the wall base. As a result of their nu- merical predictions, Ho and Rowe questioned the no- tion of back-calculating soil lateral pressure from reinforcement loads. They argued that the reinforce- ment load was not controlled by stress in the soil un- der active conditions. Ho and Rowe also found that the curtailment of reinforcement in the transition zone (i.e. zone 2) may contribute significantly to wall lateral displacements. They concluded that calculating lateral displacements of the wall from integration of rein- forcement strains along the length of the reinforcement will underestimate wall deformations. Rowe and Ho (1993) found that a rigid toe can in- crease the vertical soil stress at the bottom of the wall by as much as 25% higher than the nominal expected value. They also found that compaction had little ef- fect on soil horizontal stresses in geosynthetic rein- forced soil walls due to wall yielding. Rowe and Ho maintained that the locus of maximum reinforcement load in different layers is not unique in walls with ex- tensible reinforcement under working conditions and may vary depending on construction method, rein- forcement stiffness and length, facing type, etc. How- ever, they found that the locus of maximum rein- forcement load approaches the Rankine failure plane as the wall approaches a collapse state. Rowe and Ho (1996) found that continuity of the wall facing panel (i.e. propped versus incremental panel) has a negligible effect on reinforcement load and wall deformations. The effect of facing rigidity was found to be more important in the case of a rigid foundation. Their numerical results suggested that for the case of a rigid foundation, a more rigid facing would be beneficial to reduce maximum reinforcement loads and wall lateral displacements. Rowe and Ho concluded that the magnitude of elastic modulus of the backfill soil does not have a significant effect on the reinforced zone because this zone typically develops into a plastic state in walls with extensible reinforce- ment. Rowe and Ho (1996) found negligible influence of foundation flexibility on the magnitude of rein- forcement loads and wall deformations. However, the flexible foundation layer in their finite element model was thin (i.e. 0.15H where H is the wall height) and may not apply to thicker foundation layers. Rowe and Ho (1996, 1997, 1998) showed that the soil- reinforcement interface friction angle has essentially no influence on wall response for values larger than 2/3() where is the friction angle of the cohesionless backfill soil. The reinforcement load in the lower rein- forcement layers and wall lateral displacements in- creased with proportionately lower interface friction angle values. Ho and Rowe (1996) concluded that the reinforce- ment length to wall height ratio, L/H, is the most im- portant geometric parameter influencing the wall re- sponse. They found little variation in the magnitudes of reinforcement load and soil stress for L/H values larger than 0.7. It is worth noting that the ratio L/H=0.7 is the minimum reinforcement length ratio recommended by current design codes (e.g. FHWA 1996, AASHTO 1998) for static stability of reinforced soil walls. Mitchell and Christopher (1990) concluded from the results of a similar FEM investigation that wall displacements were also largely influenced by the L/H ratio. A design chart to predict wall deformations as a function of L/H and reinforcement type (i.e. ex- tensible or in-extensible) appears in current AASHTO (1998) and FHWA (1996) guidelines. Ho and Rowe (1996) concluded that placing equally spaced reinforcement layers with L/H=0.7 is an efficient reinforcement distribution and is recom- mended over other distribution patterns of reinforce- ment in reinforced soil walls. They expressed caution in extrapolating the response of low-height walls to taller retaining wall systems because of the greater in- fluence of the combined rigid facing and hinged toe boundary condition on low height walls. Ho and Rowe (1999) noted that their conclusions with respect to the optimal L/H ratio and reinforcement distribution ar- rangement within the backfill was consistent with the results of centrifuge physical modeling (Porbaha 1999). Rowe and Ho (1997) found that the total force re- quired for internal equilibrium of a reinforced soil wall is essentially independent of material properties other than the backfill friction angle. They noted that the magnitude of this equilibrium force can be determined using Rankine theory. However, a higher total force can be expected for the case where the backfill-facing friction angle is small. In addition, they found that the total force for external equilibrium of the facing is practically independent of material properties adopted in the model and can be calculated using a Coulomb active wedge analysis. The exception to good corre- spondence between Coulomb theory and numerically predicted results occurs when the facing rigidity or the soil-reinforcement interface friction angle is small. Rowe and Ho (1998) showed that the magnitude of wall lateral displacement is influenced by the soil fric- tion angle and a reinforcement stiffness factor, , de- fined as =J/K a HS v , where J is the reinforcement stiffness, K a is the Rankine active earth pressure coef- ficient, is the soil unit weight, H is wall height and S v is the vertical spacing between reinforcement lay- ers. They argued that parameters such as soil-facing friction angle, soil-reinforcement friction angle, soil elastic modulus and facing rigidity are not as impor- tant. Rowe and Ho noted that reducing the reinforce- ment length in the lower half of the wall height may result in a substantial increase of wall lateral dis- placement. Hatami et al. (2000) investigated the influence of the simultaneous use of reinforcement layers with dif- ferent stiffness values (non-uniform reinforcement) in wrapped-face (very flexible facing) reinforced soil walls. The numerical simulation of wall models was carried out using the finite difference based program FLAC. The simulations were carried out using a linear elastic M-C model for the soil and linear elastic (FLAC) cable elements. The simulations included staged construction of the walls. The wall lateral dis- placements and horizontal earth pressures behind the facing in all non-uniform reinforced wall models showed a clear dependence on relative stiffness values of the individual reinforcement layers. The parametric study showed that wall displacements were less for configurations with alternating primary and secondary reinforcement layers of equal length compared to nominal identical walls with the two different rein- forcement stiffness materials grouped separately in the bottom and top halves of the wall structure. 2.2 Numerical modeling studies with calibration This section describes studies in which physical rein- forced soil walls were simulated using numerical modeling and the results of numerical and physical models compared. Karpurapu and Bathurst (1991a,b) modeled the re- sponse of the 3 m high Denver Test Wall using the non-linear large strain finite element program GEO- FEM. They discussed the sources of discrepancies between their prediction of the wall response (class A prediction) and the observed (measured) performance. They noted that the boundary constraint at the top of the test facility significantly reduced the soil and panel deformations. They reported Class C predictions using the same method of analysis and material properties as used in the class A prediction analysis but with boundary conditions modeling the actual physical test. Karpurapu and Bathurst (1991b) concluded that proper boundary conditions (including side wall friction) and correct choice of soil properties were essential for suc- cessful numerical simulation of the performance of reinforced soil test walls. Karpurapu et al. (1991), Bathurst et al. (1992), Karpurapu and Bathurst (1992, 1994, 1995) and Ra- jagopal and Bathurst (1994) used the program GEO- FEM to model the behavior of carefully instrumented full-scale walls constructed at the Royal Military Col- lege of Canada (RMCC). The material properties used in the numerical simulations were obtained from inde- pendent small-scale laboratory experiments. They used the hyperbolic model proposed by Duncan et al. (1980) for the backfill soil and included the soil dila- tion angle inferred from numerical simulation of the results of direct shear box tests. Karpurapu et al. noted that soil dilation has a significant effect on the normal pressures at the reinforcement-soil interface for well- compacted dense granular soils. For the particular sand used at RMCC, the soil strength was about 20% higher when soil dilation was included compared to the case where soil dilation was neglected. They found it necessary to use appropriate load-strain-time (iso- chronous) data to model the reinforcement stiffness in order to predict correct strain magnitudes (and hence tensile forces) in the reinforcement layers at soil fail- ure. Karpurapu et al. concluded that their numerical simulation was successful in predicting the peak val- ues and the trend in reinforcement strains at incipient collapse of an incremental panel wall under surcharge loading. However, the numerical model over-predicted the length of the reinforcement in the soil anchorage zone. Nevertheless, this result was viewed as a con- servative (i.e. safe) result for the design of reinforced soil walls using the GEOFEM code. The numerical simulation study by Karpurapu et al. was quantita- tively consistent with experimentally measured data that showed significant forces generated at the hori- zontally restrained toe of the wall facing and con- firmed that these forces contribute additional stabiliz- ing reactions for low height walls. This additional capacity may account for a large portion of the conser- vatism inherent in current limit-equilibrium methods of analysis and design. The internal soil failure surface predicted by the finite element simulation was ob- served to be in good agreement with the excavated failure surface and consistent with failure geometry predicted by conventional Coulomb wedge theory. Andrawes and Yogarajah (1994) used the finite element program CRISP to study the effect of rein- forcement-facing connections on the response of three 2 m high physical reinforced-soil wall models. Equivalent linear reinforcement stiffness values were determined from reinforcement isochronous creep curves. The interface element introduced by Goodman et al. (1968) was used between soil, reinforcement and the facing. Andrawes and Yogarajah found that in the case where the reinforcement is rigidly attached to the facing, the strain distribution along the reinforcement layer is approximately linear with a maximum value at the connection to the facing. If some flexibility was allowed in the reinforcement-facing connection, the maximum reinforcement load was observed to move away from the facing. In this case, the distribution of reinforcement load along its length was nonlinear with negligible values at the connection. Andrawes and Yogarajah concluded that correct modeling of the connection boundary conditions is essential for accu- rate numerical predictions of reinforced soil wall re- sponse. Ochiai and Fukuda (1996) and Nakane et al. (1996) reported finite element studies of an experimental 6 m high geotextile reinforced retaining wall with a con- crete block facing. They concluded that their finite element model satisfactorily replicated experimental measurements of wall lateral displacement, reinforce- ment strains, soil lateral pressure behind the wall and the facing column toe load. Both groups used a linear elastic-plastic constitutive model with a M-C failure criterion for the soil elements and constant modulus linear elastic bar elements for the reinforcement layers. However, little information was provided on how they determined model input parameter values for backfill soil, reinforcement and joint elements/interfaces. Nakane et al. (1996) also carried out comparative finite element analyses of geotextile reinforced soil retaining wall models constructed with different facing types. They observed that for the case of a rigid facing, maximum reinforcement forces developed at the con- nection to the facing panel. In addition, they found a significant foundation reaction at the base of the rigid facing (footing). Helwany et al. (1999) used the program DACSAR and calibrated their finite element model against the measured response of the Denver Test Wall. The non- linear soil model proposed by Duncan et al. (1980) was used to model the soil. They obtained satisfactory agreement between the measured results of facing lat- eral displacement and reinforcement strain and results from of the finite element simulation. Helwany et al. carried out parametric analyses to investigate the ef- fects of wall height, backfill type and reinforcement stiffness on wall response. They concluded that back- fill type is the most important parameter in the per- formance of reinforced soil wall systems. They found that the stiffness of the geosynthetic reinforcement has an important influence on wall displacement response when the backfill shear strength and stiffness are low. Helwany et al. developed a series of design charts for selection of backfill type and reinforcement stiffness to meet code requirements for the performance of rein- forced soil wall systems. Walters et al. (2000) reported the monitored be- havior of several full-scale reinforced segmental (modular block) retaining walls constructed at RMCC. Program FLAC was used to simulate the wall response during construction and was calibrated against the measured history of wall lateral displace- ment, reinforcement strain and toe reaction forces. The predicted values from the FLAC program and the measured wall response with respect to all the above response histories were found to be very close. How- ever, the numerical results were very sensitive to the choice of interface properties assigned to the discrete block units during simulated construction. Ling et al. (2000) modeled the construction of a full-scale, concrete-block, geosynthetic-reinforced soil retaining wall using the finite element method. The soil and geosynthetic reinforcement were both mod- eled using hyperbolic formulations. Ling et al. com- pared the numerical and measured experimental re- sults. While the magnitude of recorded wall displacements, lateral earth pressures and bearing pressures were in general agreement, the distribution of values was observed to deviate from measured val- ues. 3 DYNAMIC ANALYSIS 3.1 Seismic loading This section is focussed on seismic modeling of rein- forced soil walls. Because of the limited number of studies on geosynthetic reinforced walls under simu- lated earthquake loading some early studies of metallic reinforced soil walls are included. Richardson and Lee (1975) carried out dynamic fi- nite element analyses using the computer program QUAD-4 to interpret shaking table test results of small-scale (i.e. 300 mm high) model walls subjected to sinusoidal base excitation. Their analyses included strain dependent soil modulus and damping. The me- tallic reinforcement was modeled using elastic bar elements. Richardson and Lee found that using non- linear soil properties resulted in underestimation of measured reinforcement loads. They found that a satis- factory agreement between measured reinforcement load and predicted values from their finite element models could be obtained by using the actual meas- ured soil properties from independent laboratory tests. Segrestin and Bastick (1988) calibrated the finite element program SUPERFLUSH against the dynamic response results of a 3 m high wall reported by Chida et al. (1985). They reported a good agreement between calculated and measured values of reinforcement in- cremental load (i.e. load in excess of static value) at a given frequency of input harmonic base motion. They then used the calibrated model to investigate the dy- namic response of two (6 m and 10.5 m high) metallic reinforced soil wall models to input ground motion. A frequency domain analysis approach was adopted in the program which included an iterative procedure to account for varying soil elastic modulus using the cal- culated soil deformation. Segrestin and Bastick carried out a set of parametric analyses which included three different foundation soil conditions and two recorded earthquake ground motions, scaled to various peak ground accelerations. The results of Segrestin and Bastick showed that a soft foundation reduces the magnitude of reinforcement load in the wall subjected to ground motion to lower values than the same wall constructed on a rigid foundation. The magnitude of reduction was particularly significant for the rein- forcement layers in the bottom half of the wall. Their results showed that the reinforcement incremental load was reasonably uniform along the reinforcement length for most layers and the location of the maxi- mum reinforcement load in each reinforcement layer was the same as that for static loading. Segrestin and Bastick found that the reinforcement incremental load was greater in the lower reinforcement layers. They concluded that lower reinforcement layers would de- velop higher incremental loads when steel reinforce- ment strips are wider or there are a greater number in a horizontal plane. This conclusion is consistent with the results of numerical modeling reported by Bathurst and Hatami (1998b, 1999a) who showed that rein- forcement incremental loads increased with increasing reinforcement stiffness over a wide range of values that included relatively soft geosynthetic reinforce- ment materials as well as metallic reinforcement (i.e. 100 J 69,000 kN/m). Segrestin and Bastick found an acceleration amplification factor greater than 2 between the top and the base of their wall models. This result is in agreement with the calculated accel- eration amplification by Bathurst and Hatami (1998b) for a comparable model wall height. Segrestin and Bastick proposed a simplified approach to calculate the reinforcement incremental load in reinforced soil walls subjected to ground motion. Their proposed method has been adopted in AASHTO (1998) seismic design guidelines for reinforced earth structures. How- ever, as pointed out by Bathurst and Alfaro (1996) the Segrestin and Bastick simplified procedure is based on a limited analysis that does not include a sufficiently wide range of wall types, material properties and ground motion characteristics. Specifically, the rein- forcement stiffness values in their analyses are much greater than values associated with typical geosyn- thetic reinforcement materials. The subsequent nu- merical modeling results reported by Bathurst and Hatami (1998b) confirmed that for panel walls with a fixed toe the incremental reinforcement load was es- sentially constant with depth for geosynthetic rein- forcement materials but increased linearly with depth for metallic reinforcement. Cai and Bathurst (1995) studied the dynamic re- sponse of a modular block geosynthetic reinforced soil retaining wall using the finite element method. The dynamic finite element program TARA-3 was modi- fied to include a hyperbolic stress-strain relationship with Masing hysteretic unload-reload behavior to model the cyclic shear behavior of the backfill soil. The reinforcement material was modeled using a similar hysteretic model which accounted for the measured response of cyclic load-extension tests per- formed on unconfined geogrid specimens. Interface shear between wall components was simulated using slip elements. Cai and Bathurst found that the relative displacement and interface shear force between modular units increased with the duration and magni- tude of base excitation during simulated seismic events. Relative displacements and shear forces were greatest at interfaces where a geosynthetic was present and at some locations the shear capacity between modular units was exceeded. They concluded that an accurate estimate of interface shear properties is par- ticularly important for the seismic design of segmental geosynthetic reinforced soil retaining walls. Cai and Bathurst also observed that dynamic tensile forces in reinforcement layers accumulated during the simulated seismic events and peak tensile forces increased with the magnitude of base peak acceleration. They re- ported that reinforcement forces calculated using pseudo-static earth pressures (Mononobe-Okabe method) were consistently greater than the forces cal- culated by the dynamic FEM analyses. Reinforcement forces resulting from static loading conditions prior to base excitation were greater that the forces predicted using conventional Coulomb earth pressure including the topmost layer. The over-prediction of reinforce- ment forces using the Coulomb approach is likely ex- acerbated by the low height concrete wall investigated (i.e. 3.2 m high) in combination with the restrained footing condition. The horizontal accelerations pre- dicted at different locations in the unreinforced soil mass, reinforced soil mass and the facing column showed that peak accelerations were time coincident. Cai and Bathurst found little acceleration amplifica- tion in their simulation results which, again, can be attributed to the low wall height. They concluded that the conventional practice of using a single acceleration factor in pseudo-static methods of analysis could therefore be justified for the case investigated. Bathurst and Hatami (1998a,b,c,1999a,b) and Ha- tami and Bathurst (1999, 2000a,b) used the finite dif- ference based program FLAC to study dynamic re- sponse of retaining wall systems to input ground motion. Bathurst and Hatami (1998a,b) investigated the influence of reinforcement stiffness, length and base boundary condition on the seismic response of a geosynthetic reinforced full-height panel wall. They found program FLAC to be a useful research tool that could be used to investigate the response of retaining walls under static and dynamic loading conditions. Several advantages of the program include a fast algo- rithm capable of tackling large distortions and struc- tural collapse. User-defined constitutive models can be incorporated in the program in addition to the avail- able library of constitutive models. Dynamic loading can be readily applied to the model in terms of accel- eration, velocity, stress or force histories. A number of built-in boundary conditions are available to account for truncated modeling of unbounded media (e.g. non- reflecting boundary). The model can also include vari- ous structural elements and interfaces within the con- tinuum domain. Bathurst and Hatami found that both reinforcement stiffness and reinforcement length in- fluence the magnitude of wall lateral displacement when subjected to ground motion. However, dynamic response of the simulated walls was most sensitive to the toe boundary condition (i.e. fixed pinned toe and pinned sliding toe). The magnitude of permanent wall displacement was observed to decrease with increas- ing reinforcement stiffness. Permanent wall displace- ments were relatively insensitive to changes in rein- forcement length to wall height ratios in the range L/H = 0.7 to 1. Bathurst and Hatami (1998b,c) found that signifi- cant reinforcement incremental loads can develop when reinforced soil walls are subjected to input ground motions in the vicinity of wall fundamental frequency. The maximum reinforcement load in all simulation runs was consistently observed to occur at the rigid connection of reinforcement layers to the facing. They found that reinforcement load distribu- tion over the wall height is very different for geosyn- thetic reinforced walls as compared to the walls with metallic reinforcement as noted earlier. A horizontally restrained toe reduced the total and incremental force in the lower layers for cases with reinforcement stiff- ness values associated with geosynthetic products (e.g. J 2000 kN/m). In contrast, the total and incremental reinforcement load increased monotonically with depth for the case of metallic reinforcement. The lin- ear trend of the dynamic load increment with depth below the crest used in the AASHTO (1998) method was not observed in any simulations with reinforce- ment stiffness values comparable to geosynthetic ma- terials. Bathurst and Hatami (1998c) proposed that a bi-linear dynamic load increment distribution be used in empirical models for reinforcement stiffness values typically associated with geosynthetic reinforced soil walls. Bathurst and Hatami (1998b, 1999b) carried out parametric analyses to investigate the quantitative in- fluences of backfill width, damping ratio and type of far-end truncated boundary on calculated wall re- sponse. Bathurst and Hatami (1998b) found that sig- nificant acceleration amplification (e.g., more than twice the amplitude of base acceleration) was ob- served for walls subjected to ground motions in the vi- cinity of the fundamental frequency. However, they also showed that the magnitude of amplification was influenced by the magnitude of damping ratio used in the numerical models. Bathurst and Hatami showed that the soil in the retained soil zone develops into the plastic state during shaking and the inclination of the failure surface in this region can be predicted satis- factorily using Mononobe-Okabe theory considering the acceleration amplification in the backfill. Hatami and Bathurst (1999, 2000a) studied the in- fluence of different structural design parameters on the fundamental frequency of reinforced-soil retaining wall models. The design parameters in the study in- cluded the wall height, backfill width, reinforcement stiffness, reinforcement length, backfill friction angle and toe restraint condition. The intensity of ground motion, characterized by peak ground acceleration, was included in the study. Hatami and Bathurst con- cluded that the fundamental frequency of reinforced- soil wall models with sufficiently wide backfill sub- jected to moderately strong input accelerations can be estimated with reasonable accuracy from linear elastic wave theory using the shear wave speed in the backfill and the wall height. Their numerical analyses showed no significant influence of the reinforcement stiffness, reinforcement length or toe restraint condition on the fundamental frequency of wall models. The strength of the granular backfill, characterized by its friction angle, also did not show any observable effect on the fundamental frequency of reinforced-soil retaining wall models. However, the resonance frequencies of wall models were dependent on the ground motion intensity and to a lesser extent, on the width to height ratio of the backfill. Hatami and Bathurst (2000b) studied the dynamic response of a segmental (modular block) retaining wall model to recorded ground motions. They com- pared the magnitude and characteristics of wall re- sponse to those obtained under harmonic input base acceleration. They found that the maximum lateral displacement of the facing column and maximum rein- forcement load of the segmental retaining wall model subjected to a single frequency, harmonic input accel- eration were considerably larger than the correspond- ing values obtained using a number of earthquake ac- celerograms with comparable predominant frequencies. They concluded that the random charac- teristic of actual ground accelerations would partly ex- plain the relatively good performance of reinforced- soil retaining wall systems that were designed without seismic considerations or at best, using simple pseudo- static limit equilibrium methods (Bathurst and Alfaro 1996). Hatami and Bathurst found that low-frequency ground motions with high intensity values can result in significant structural response magnitude of short- period retaining wall systems. 3.2 Blast loading Yogendrakumar et al. (1991) investigated the dy- namic response of geosynthetic reinforced soil struc- tures subjected to air blast loading. They adopted a method that was based on an estimate of the pressure- time response at the boundaries of target structures using a 2D gas dynamic code SPLIT2D. The soil- reinforcement response was simulated by modifying the finite element computer program RESBLAST. They compared the response of reinforced soil wall and slope models with and without reinforcement. They used the results of rapid cyclic tensile loading tests to obtain mechanical properties needed for the nonlinear reinforcement model implemented in the fi- nite element code. Yogendrakumar and Bathurst showed that the polymeric reinforcement improved the performance of the structure under air blast loading by reducing the magnitude of dynamic and permanent soil deformations by as much as 30%. 4 CONCLUSIONS The paper summarizes many examples of the use of computer codes that use the finite element and finite difference methods to predict the response of rein- forced-soil wall models under static and dynamic loading conditions. The majority of the studies re- viewed by the writers investigated the response of ide- alized reinforced soil walls. Relatively few studies are available in the literature that report direct compari- sons of numerical results and experimentally measured results from carefully instrumented walls. In the few cases where direct comparisons are reported there are some significant discrepancies between predicted and measured values. A more serious shortcoming of many studies that have included physical test results is that the physical experiments were only lightly in- strumented and hence boundary forces which are an important test of the accuracy of numerical results cannot be evaluated or boundary conditions in the physical tests are over-simplified from typical field cases. The stress-dependent hyperbolic soil model pro- posed by Duncan et al. (1980) has been used by many modelers. This model has the advantage that parame- ters can be determined from independent standard tri- axial tests. The model has been used in combination with the M-C failure criterion for granular soils under static conditions. Constitutive models that include soil dilation have been shown to improve the prediction of the response of reinforced soil walls constructed with compacted granular soils. Reinforcement models have been generally linear elastic-type but efforts to incor- porate non-linear models are reported by a number of researchers. A practical approach to the problem of visco-elastic (creep) behavior of geosynthetic rein- forcement materials is to use modified linear elastic models with parameters interpreted from in-isolation isochronous load-strain data. For dynamic modeling, most researchers have re- stricted soil models to linear elastic-type to minimize computational time. This is also true for the rein- forcement materials. Nevertheless, hyperbolic stress- strain models with hysteretic unload-reload behavior have been used for the backfill soil and polymeric re- inforcement materials. More work is required to in- vestigate whether or not the response of reinforced soil walls under dynamic loading is significantly influ- enced by the choice of constitutive model for the soil and reinforcement. A common deficiency of the results of reinforced soil wall modeling is that tensile strains in extensible reinforcement elements propagate deeper into the rein- forced soil mass than has been observed in some care- fully instrumented physical tests. The modeling of segmental retaining walls constructed with a discrete dry-stacked column of concrete blocks has proven to be a particular challenge. The writers and co-workers have shown that interface mechanics of the block units are complicated by the presence of reinforcement layer inclusions. More effort is required to develop more accurate soil-reinforcement interface models that can accurately simulate the construction-induced response of segmental reinforced soil walls. The response of reinforced soil walls is further complicated by non-quantifiable factors such as qual- ity of construction during building of the structures. Most of the wall deformation that occurs during the in-service life of a modern reinforced soil wall system occurs during construction. Quality of construction may control the behavior of complex multi-component systems such as reinforced segmental retaining wall systems and render Class A predictions problematic. Despite the shortcomings in numerical modeling of geosynthetic reinforced soil walls some important les- sons have been learned that have practical implica- tions to development of empirical rules for the design of structures in static and earthquake environments. It is clear that the type of connection between the rein- forcement and the wall facing plays a major role in determining the location and magnitude of peak loads in the reinforcement. In addition, the boundary condi- tion that exists at the toe of a hard-faced wall has a major influence on wall response. More work is re- quired to investigate the influence of depth of com- pressible foundation layer on wall response under both static and earthquake loading. Results of numerical modeling have identified or confirmed empirical rules for the optimum vertical arrangement and length of reinforcement layers in typical reinforced soil wall structures. The numerical simulation work by the writers has shown that the relationship of the fundamental fre- quency of a reinforced soil wall to the frequency con- tent of base input ground motion may control the seismic response of a wall to a much larger degree than the mechanical properties of the components. The paper identifies FEM studies that have been used to develop empirical guidelines for static and seismic design of reinforced soil walls. An example is given illustrating how the results of numerical model- ing of metallic reinforced soil retaining walls have been used to develop seismic design guidelines. How- ever, a subsequent more exhaustive numerical para- metric study has shown that empirical rules for metal- lic reinforced soil walls may not apply for nominal identical walls constructed with relatively extensible geosynthetic reinforcement materials. 5 REFERENCES AASHTO 1998. Interims: Standard specifications for highway bridges. Am. Assoc. Sta. Highw. Trans. Offic. Washington, DC, USA. Andrawes, K.Z., K.H. Loke & K.C. Yeo 1990. Application of boundary yielding concept to full scale reinforced and unrein- forced soil walls. In A. McGown, K.C. Yeo & K.Z. Andrawes (eds), Performance of Reinforced Soil Structures. Brit. Geo- tech. Soc. Glasgow: 79-83. Andrawes, K.Z. & I. Yogarajah 1994. Effects of reinforcement connections on the behaviour of reinforced soil retaining walls. In H.J. Siriwardane & M.M. Zaman (eds), Computer Methods and Advances in Geomechanics. Proc. 8th Int. Conf. Num. Meth. Geomech. Morgantown, WV, USA. 2: 1313-1318. Rotterdam: Balkema. Bathurst, R.J. & M.C. Alfaro 1996. Review of seismic design, analysis and performance of geosynthetic reinforced walls, slopes and embankments. In H. Ochiai N. Yasufuku and K. Omine (eds), Earth Reinforcement, Proc. Int. Symp. Earth Reinf., Fukuoka, Kyushu, Japan, 2: 887-918. Rotterdam: Balkema. Bathurst, R.J. & K. Hatami 1998a. Influence of reinforcement stiffness, length and base condition on seismic response of geosynthetic reinforced retaining walls. Proc. 6th Int. Conf. Geosyn. Atlanta, GA,USA. 2: 613-616. Bathurst, R.J. & K. Hatami 1998b. Seismic response analysis of a geosynthetic reinforced soil retaining wall. Geosyn. Inter. 5(1- 2): 127-166. Bathurst, R.J. & K. Hatami 1998c. Influence of reinforcement properties on seismic response and design of reinforced soil retaining walls. Proc. 51st Can. Geotech. Conf. Edmonton, AB, Canada. 2: 479-486. Edmonton. Bathurst, R.J. & K. Hatami 1999a. Numerical study of the influ- ence of base shaking on geosynthetic reinforced retaining walls. Proc. Geosynthetics99 2: 963-976. Boston. Bathurst, R.J. & K. Hatami 1999b. Earthquake response analysis of reinforced-soil walls using FLAC. In C. Detournay & R. Hart (eds), FLAC and Numerical Modeling in Geomechanics. FLAC Symp. Num. Mod. Geomech. Minneapolis, MN, USA: 407-415. Bathurst, R.J., R.G. Karpurapu & P.M. Jarrett 1992. Finite ele- ment analysis of a geogrid reinforced soil wall. In R.H. Bor- den, R.D. Holtz & I. Juran (eds), ASCE Geotech. Special Publ. no. 30: Grout. Soil Improv. Geosyn. 2: 1213-1224. BSI 1995. Code of practice for strengthened/reinforced soils and other fills. Brit. Stan. Inst. BS 8006: 1995. Cai, Z. & R.J. Bathurst 1995. Seismic response analysis of geo- synthetic reinforced soil segmental retaining walls by finite element method. Comp. Geotech. 17(4): 523-546. Chida, S., K. Minami & K. Adachi 1985. Tests with regard to the stability of the fill constructed by the reinforced earth tech- nique. (unpublished report translated from Japanese) Duncan, J.M., P. Byrne, K.S. Wong & P. Mabry 1980. Strength, stress-strain and bulk modulus parameters for finite-element analysis of stresses and movements is soil masses. Rep. UCB/GT/80-01, UC Berkeley, CA, USA. FHWA 1996. Mechanically stabilized earth walls and reinforced soil slopes design and construction guidelines. In V. Elias and B.R. Christopher (eds). Federal Highway Administration (FHWA) Demonstration Project 82, Washington, DC, USA. Goodman, R.E., R.L. Taylor & T.L. Brekker 1968. A model for the mechanics of jointed rocks. ASCE J. Soil Mech. Foun. Div. 94(SM3): 637-659. Hatami, K. & R.J. Bathurst 1999. Frequency response analysis of reinforced-soil retaining walls. Proc. 8th Can. Conf. Erthq. Eng. Vancouver, BC, Canada: 341-346. Hatami, K. & R.J. Bathurst 2000a. Effect of structural design on fundamental frequency of reinforced-soil retaining walls. Soil Dyn. Erthq. Eng. 19(3): 137-157. Hatami, K. & R.J. Bathurst 2000b. Investigation of seismic re- sponse of reinforced-soil retaining walls. Proc. 4th Int. Conf. Rec. Adv. Geotech. Erthq. Eng. Soil Dyn. San Diego, CA, USA. (in press) Hatami, K., R.J. Bathurst, P. Di Pietro & P.M. Bianco 2000. Nu- merical study of retaining walls with non-uniform reinforce- ment. Proc. EuroGeo 2. Bologna, Italy. (in press) Helwany, S.M.B., G. Reardon & J.T.H. Wu 1999. Effects of backfill on the performance of GRS retaining walls. Geotex. Geomem. 17(1): 1-16. Hermann, L.R. & Z. Al-Yassin 1978. Numerical analysis of rein- forced soil systems. ASCE Symp. Earth Reinf. Pittsburgh, PA, USA: 428-457. Ho, S.K. & R.K. Rowe 1993. Finite element analysis of geosyn- thetic-reinforced soil walls. Proc. Geosynthetics93, Vancou- ver: 203-216. Ho, S.K. & R.K. Rowe 1996. Effect of wall geometry on the be- havior of reinforced soil walls. Geotex. Geomem. 14(10): 521-541. Ho, S.K. & R.K. Rowe 1999. Effect of wall geometry on the be- havior of reinforced soil walls - Reply to the discussion by A. Porbaha. Geotex. Geomem. 17(2): 125-126. Itasca Consulting Group 1998. FLAC - Fast Lagrangian Analysis of Continua. Ver. 3.40. Itasca Cons. Gr., Inc. Minneapolis, MN, USA. Karpurapu, R.G. & R.J. Bathurst 1991a. Finite element predic- tions for the Colorado geosynthetic-reinforced soil retaining walls. Proc. Int. Symp. Geosynth.-Reinf. Soil Retaining walls: 329-341. Denver: Balkema. Karpurapu, R.G. & R.J. Bathurst 1991b. Discussion and class C predictions for the Colorado trial walls. Proc. Int. Symp. Geosynth.-Reinf. Soil Retaining walls: 343-346. Denver: Balkema. Karpurapu, R.G., R.J. Bathurst & P.M. Jarrett 1991. Finite ele- ment analysis of an incrementally constructed geogrid rein- forced soil wall. Proc. 44th Can. Geotech. Conf.: 82/1-82/8. Calgary. Karpurapu, R. & R.J. Bathurst 1992. Analysis of geosynthetic reinforced soil walls by finite element methods. In H.J. Siri- wardane & M. Zaman (eds), Computer Methods and Ad- vances in Geomechanics. Proc. 8th Int. Conf. Num. Meth. Geomech. Morgantown, WV, USA. 1: 861-870. Karpurapu, G. & R.J. Bathurst 1994. Finite element analysis of geosynthetic-reinforced soil walls. Proc. 13th ISSMFE Conf.: 1381-1384. New Delhi. Karpurapu, R.G. & R.J. Bathurst 1995. Behaviour of geosyn- thetic reinforced soil retaining walls using the finite element method. Comp. Geotech.. 17(3): 279-299. Ling, H.I., C.P. Cardany, L.X. Sun & H. Hashimoto 2000. Finite element study of a geosynthetic-reinforced soil retaining wall with concrete-block facing. Geosyn. Int. 7(2): 137-162. Mitchell. J.K. & Christopher, B.R. 1990. North American prac- tice in reinforced soil systems. In P.C. Lambe and L.A. Han- sen (eds) ASCE Geotechnical Special Publication No. 25, Proc. Conf. Design and Performance of Earth Retaining Structures, Cornell University, Ithaca, NY, USA June 1990; 322-346. Nakane, A., M. Taki, Y. Yokota & H. Miyatake 1996. FEM comparative analysis of facing rigidity of geotextile- reinforced soil walls. Proc. Int. Symp. Earth Reinf. 1: 433- 438. Fukuoka: Balkema. Naylor, D.J. 1978. A study of reinforced earth walls allowing strip slip. ASCE Symp. Earth Reinf. Pittsburgh, PA, USA: 618-643. Ochiai, Y. & N. Fukuda 1996. Experimental study on geotextile- reinforced soil walls with different facings. Proc. Rec. Cas. Hist. Devel. Des. Geosyn.-Reinf. Soil Retain. Walls, Tokyo, Japan, Nov. 1996: 140-147. Porbaha, A. 1999. Discussion on: Effect of wall geometry on the behavior of reinforced soil walls. Geotex. Geomem. 17(2): 121-123. Rajagopal, K. & R.J. Bathurst 1994. Parametric finite element in- vestigation of reinforced soil retaining walls. Proc. 5th Int. Conf. Geotex. Geomem. Rel. Prod.: Singapore, September 1994: 193-196. Richardson, G.N. & K.L. Lee 1975. Seismic design of reinforced earth walls. Proc. ASCE J. Geotech. Eng. Div. 101(GT2): 167-188. Rowe, R.K. & S.K. Ho 1993. A review of the behavior of rein- forced soil walls. Proc. Int. Symp. Earth Reinf. Prac. 2: 801- 830. Fukuoka: Balkema. Rowe, R.K. & S.K. Ho 1996. Some insights into reinforced wall behavior based on finite element analysis. Proc. Int. Symp. Earth Reinf. Prac. 1: 485-490. Fukuoka: Balkema. Rowe, R.K. & S.K. Ho 1997. Continuous panel reinforced soil walls on rigid foundations. Proc. ASCE J. Geotech. Geoenvi- ron. Eng. 123(10): 912-920. Rowe, R.K. & S.K. Ho 1998. Horizontal deformation in rein- forced soil walls. Can Geotech. J. 35: 312-327. Seed, H.B. & I.M. Idriss 1970. Soil moduli and damping factors for dynamic response analysis. Rep. EERC 70-10, UC Ber- keley, CA, USA. Segrestin, P. & M. Bastick 1988. Seismic design of reinforced earth retaining walls - the Contribution of Finite Element Analysis. In T. Yamanouchi N. Miura and H. Ochiai (eds), Theory and Practice of Earth Reinforcement, Proc. Int. Geo- tech. Symp. Theo. Prac. Earth Reinf. Fukuoka, Kyushu, Ja- pan, October 1988: 577-582. Walters, D., K. Hatami, R.J. Bathurst, N. Vlachopoulos, D. Saunders, G.P. Burgess & T.M. Allen 2000. Full-scale testing and modeling of geosynthetic reinforced soil retaining walls. Proc. 53rd Can. Geotech. Conf. Montreal. (in press) Yogendrakumar, M. & R.J. Bathurst 1992. Numerical simulation of reinforced soil structures during blast loadings. Trans. Res. Rec. 1336: 1-8.