【正文】 e 6. UDEC model of flexural toppling in a surface coal mine slope. The engineer must again be cautious that the structural input into the distinctelement analysis is representative. Hencher et al. (16) illustrated the importance of bedding spacing on predicted failure mechanism. Stead and Eberhardt (17) showed the importance of discontinuity orientation on failure modes observed in surface coal mine slopes. It is stressed that tailoring the structure of the model to acmodate the low random access memory of a laptop puter, for example by using unrepresentative discontinuity spacing, may lead to unrepresentative results. Simulations must always be verified with field observations and wherever possible instrumented slope data. This bees even truer with the development of 3D discontinuum codes such as 3DEC (11). Only when a confident portrayal of the 3D characteristics of a slope has been obtained, can the results be considered representative. This in turn requires the undertaking of an extensive and indepth site investigation beforehand. The discontinuous deformation analysis, DDA, developed by Shi (18) has also been used with considerable success in the modelling of discontinuous rock masses, both in terms of rockslides (19) and rockfalls (20). An important recent development in discontinuum codes is the application of distinctelement methodologies and particle flow codes, . PFC2D/3D (11). This code allows the rock mass to be represented as a series of spherical particles that interact through frictional sliding contacts. Clusters of particles may be bonded together through specified bond strengths in order to simulate joint bounded blocks. One major advantage of such an approach is that high stresses induced in the rock slope will break the bonds between the particles simulating, in an approximate manner, the intact fracture of the rock. The use of discontinuum methods in association with continuum methods has been shown by several authors to provide an instructive approach to rock slope analysis. Board et al. (21) illustrate the analysis of plex deformation within the 650m high Chuquicamata pit slope, Chile, using a bined approach which utilises both FLAC and UDEC analyses. Similarly, Benko and Stead (14) used an approach adopting FLAC for the initial investigation of Frank Slide and UDEC for the indepth analysis. The latter study integrated results from limit equilibrium, continuum and discontinuum analyses using each technique as a tool to provide a step in the overall rock slope analysis.Hybrid Techniques Hybrid approaches are increasingly being adopted in rock slope analysis. This may include bined analyses using limit equilibrium stability analysis and finiteelement groundwater flow and stress analysis such as adopted in the GEOSLOPE suite of software (6). Hybrid numerical models have been used for a considerable time in underground rock engineering including coupled boundary/finiteelement and coupled boundary/distinctelement solutions. Recent advances include coupled particle flow and finitedifference analyses using FLAC3D and PFC3D (22). These hybrid techniques already show significant potential in the investigation of such phenomena as piping slope failures, and the influence of high groundwater pressures on the failure of weak rock slopes. Coupled finite/distinctelement codes are now available which incorporate adaptive remeshing. These methods use a finiteelement mesh to represent either the rock slope or joint bounded block. This is coupled with a discrete element model able to model deformation involving joints. If the stresses within the rock slope exceed the failure criteria within the finiteelement model a crack is initiated. Remeshing allows the propagation of the cracks through the finiteelement mesh to be simulated. Hybrid codes with adaptive remeshing routines, such as ELFEN (23), have been successfully applied to the simulation of intense fracturing associated with surface mine blasting, mineral grinding, retaining wall failure and underground rock caving (24). The authors are currently exploring the use of this code in the modelling of varied rock slope failure processes.FUTURE DEVELOPMENTS The analysis of plex landslides can now be undertaken routinely using stateoftheart numerical modelling codes on desktop puters. If the benefits of these methods are to be maximized then it is essential that field data collection techniques are more responsive to advances in design capabilities. Much of current data collection methodology has changed little over the last decade and is aimed towards limit equilibrium analysis. Data including rock mass characteristics, instrumentation and groundwater must be collected in order to allow more realistic modelling of rock slope failure mechanisms. The practising engineer and the research scientist must make efforts to think beyond the use of standalone desktop puters and embrace the rapidly developing technology of parallel puters. Several decades ago engineers in industry accepted the need to run slope analyses on mainframe puters, as they could not be done by hand or calculator. The analogy exists today that in order to fully exploit the developments in 3D coupled models with adaptive remeshing we must use parallel processors where personal puters are no longer sufficient. Such an approach has now been accepted by industry in the 3D modeling of underground potash mines and in the simulation of petroleum reservoirs. The next decade holds enormous potential in our ability to model the plete failure process from initiation, through transport to deposition. This will provide a far more rigorous understanding on which to base risk assessment. The advent of virtual reality programming will allow the engineer to convey the results of simulations in a powerful and graphically efficient manner. It is essential however that quality quantity of both input data and instrumentation data for model