Proceedings, Slope Stability 2011: International Symposium on Rock Slope Stability in Open Pit Mining and Civil

Engineering, Vancouver, Canada (September 18-21, 2011)

Guidelines for Groundwater Modelling in Large Open Pit Mine Design
J. Hazzard Itasca Consulting Group Inc., Minneapolis, USA
B. Damjanac Geological Itasca Consulting Group Inc., Minneapolis, USA
L. Lorig Itasca Consulting Group Inc., Minneapolis, USA
C. Detournay Itasca Consulting Group Inc., Minneapolis, USA

Groundwater analysis in open-pit mines can be quite complex. To model the groundwater state, simplifying
assumptions must necessarily be made. This paper presents guidelines on what simplifying assumptions can be
made and when it is safe to make them without significant loss of accuracy. A numerical model of a typical
large open-pit mine is used as a test bed. Different groundwater modelling procedures are tested under different
conditions (permeability and porosity). Rules of thumb are suggested for when a steady-state analysis is
acceptable, when a transient analysis is required and when a coupled (undrained-drained) approach should be
used. In addition, the effect of infiltration is examined and the importance of simulating seasonal infiltration
fluctuations is discussed.



Accurate determination of groundwater conditions in large open-pit mines is critical in the assessment of slope
stability. The groundwater conditions are dictated by rock properties, joint properties, extraction rates, recharge
rates, drainage and many other factors. Beale (2009) provides a good overview of hydrogeological analysis in
open-pit slope design. A good review of the current state of the art in the understanding of the interaction
between water and pit slope deformations is given in Sullivan (2007).
To represent this system accurately and to determine a factor of safety (FOS) for a slope, complex numerical
models often are required. However, in many cases, certain assumptions can be made and the modelling can be
simplified without negatively affecting results. The goal of this research is to examine different aspects of the
system and determine when it is appropriate to use simplified models. The result is a set of guidelines that can
be used by practitioners when performing groundwater modelling for the purpose of analyzing the stability of
large open-pits.
A comprehensive suite of numerical models are run to answer the following questions.

Under what conditions can a steady-state pore pressure solution be used and when is it necessary to
perform a transient seepage analysis?
When is a coupled (undrained-drained) analysis required?
What is the effect of infiltration and when is it necessary to consider seasonal fluctuations?

Where possible, rules of thumb are proposed using dimensionless parameters that encompass the rock diffusivity
(which itself includes permeability, and via storage coefficient: porosity, fluid and rock moduli), recharge rates
and extraction rates. These guidelines should enable the engineer to determine what type of groundwater
modelling needs to be performed for a given open-pit mine design.
Similar types of analyses have been attempted before. Hoek & Bray (1977) showed the increase in slope angle
that can be attained through dewatering. Brown (1982) calculates the drop in pressure that can be expected for
different slope geometries and different diffusivities. Unfortunately, simplified one-dimensional models were

1 to 1 % 2500 kg/m3 0. 2008). A series of numerical models are run with different fluid flow properties. ν Porosity. the FOS is calculated and the next block is excavated. In this paper. ρ Cohesion. large open-pit mine. The rock behaves elastically for the simulations. Blocks 1 through 7 then are excavated at 9-month intervals and the FOS is calculated for each excavation. 2 Numerical model 2. The rock stiffness and strength are the same in all models. After 9 months. Rock and fluid properties are shown in Table 1. Each of these excavation blocks are removed instantaneously. The relations between FOS and different parameters are presented. Factor of safety values are calculated for each model.5 MPa 45⁰ 10-9 to 10-6 m/s 1000 kg/m3 2. as is the left boundary (symmetry condition). a two-dimensional model of a typical mining scenario is constructed. c Friction. ρw Fluid bulk modulus. The initial water table is 50 m below the surface. Rules of thumb are proposed to help decide what type of analysis needs to be performed for different mining scenarios. Rock properties used in the numerical model. Kw 5 GPa 0. 2. except when FOS calculations are performed.used that sometimes can drastically overestimate the pressure drop when compared with a two-dimensional model with a phreatic surface. in which case a Mohr-Coulomb constitutive model is considered. The initial pit is excavated gradually over a period of 18 years to set up the initial state for the study (shown). E Poisson’s ratio. Property Value Young’s modulus. Table 1. Permeability and porosity are varied to cover the range of possible scenarios expected to be encountered in the field.1 Model geometry A numerical model of a typical large open-pit mine was constructed using FLAC (Itasca. Rock properties are chosen to represent a typical.25 0. such that an equivalent continuum model can be used.2 Rock properties The rock is assumed to be full of fractures at very close spacing. The problem geometry is shown in Figure 1. The head is fixed on the right boundary at a distance of 2500 m from the middle of the pit. n Density.2 GPa . after which the model is run for 9 months to obtain the pore pressures in the slope. k Fluid density. The model base is an impermeable boundary. φ Permeability.

. an FOS can be calculated. Most models are run in plane strain (2D flow) mode. The pore pressures calculated in step 2 are used to calculate effective stress. and pore pressures at the surface of the excavation are set to zero. With the approach taken.1. however. Some of the models described employ a simplified two-way coupling (undrained-drained approach). See Dawson et al. (3) A mechanical calculation is performed. the undrained response is calculated first (in which the mechanical deformations affect the fluid pressures) and then the calculation proceeds as above for the one-way coupling. (2011). although some axisymmetric analysis is performed as described in Section 3. From the amount of reduction required. The transient analysis relies on specific storage to obtain the pore pressures at 9 months after the excavation. This is achieved with these steps: (1) Rock is excavated. This approach is described in more detail in Hazzard et al. Geometry of the numerical model used in the studies. (1990) for details.200 m 50 m 1 2 1000 m 3 4 5 6 45⁰ 7 50 m 2500 m Figure 1. which is used in the failure analysis to calculate the FOS. Note.3 Modelling approach Most of the models shown here employ a one-way fluid-mechanical coupling. the pore pressures influence the mechanical behaviour (failure) but the mechanical deformations do not influence the pore pressures directly. Factors of safety are calculated using the shear-strength reduction method in which the cohesion and friction are reduced gradually until failure occurs. that the stiffness of the rock is accounted for in the transient flow calculation through specific storage. (2) A steady-state or transient fluid flow calculation is performed. 2. In these models.

This may be slightly inaccurate for models with very low porosity. Sketch of mining excavation showing quantities used to define excavation rate. This assumes that diffusion (rather than water table movement) is the dominant mechanism for pore pressure adjustment. . : The other factor that affects the pore pressures is the rate of mining. is a measure of fluid storage in the rock.2. Mr. The volumetric mining rate (per unit model thickness). and the storage coefficient. [4] as a number to be used to quantify the effect of different model parameters on pore pressures and slope stability. Two possible modes of storage are identified for this problem: specific storage (associated with water and rock compressibility) and specific yield (phreatic storage associated with effective porosity). In this study. The specific storage coefficient is given by Sv  1   1    n  Kw E 1    [2] where symbols are as defined in Table 1. but the error introduced will be small. with reference to Figure 2.4 Dimensionless excavation rate The rate at which excess pore pressure dissipates depends on permeability. is defined. as [3] In our model. S. specific storage will be used to calculate diffusivity. porosity. Figure 2. We therefore can propose a dimensionless excavation rate metric based on the mining rate and the diffusivity. the rate of mining for stages 5 to 10 is Mr = 200 m × 133 m / 9 months. γw is the unit weight of water. These parameters are encompassed in the diffusivity: c k S w [1] where k is the permeability. fluid bulk modulus and rock stiffness.

To determine the conditions under which a steady-state solution would be acceptable. As the constant head boundary is moved farther from the pit. Performing a steady-state analysis outside of this range will produce unconservative results (calculated FOS will be too high). Three different approaches were examined: steady-state flow. (2011) show that the value R at which the steady-state solution becomes approximately equal to the transient solution (Rs) depends on the distance to the boundary squared. (2011). Hazzard et al. A comparison was made between the FOS calculated with each scheme for the different values of porosity and permeability. Excavation stage 7 was excluded because the different volume of this excavation results in a different R. 3. For this study.3 Results 3. we can expect Rs to reduce by a factor of 4. Clearly there is a significant difference that will influence the stability of the slope.05 at n = 0. The FOS values calculated for each excavation stage are shown in Figure 4. To obtain this plot. The FOS of the transient solution relative to the steady-state solution for the different porosities and values for R are shown in Figure 5. Therefore. transient fluid analysis and coupled (undrained-drained) calculations.1 Steady state versus transient Figure 3 shows a comparison of the pore pressures 9 months after excavation step 3 and at a large time after excavation 3 (steady-state solution). it takes longer to reach the steady-state solution.1 Steady state. This result depends on the location of the constant head (right) boundary. if the distance to the boundary is doubled. Changing k changes the diffusivity in Equation [1] and. The goal of the study was to determine when it is acceptable to use a simple approach and when it is necessary to consider more complicated analyses.1%). the model was solved for different values of porosity and permeability over the ranges shown in Table 1. Figure 5 (bottom) shows the result when the distance to the constant head boundary is doubled. models were run with different values for permeability (k) and porosity (n).1. the average FOS values for excavation stages 1 to 6 were calculated for each model. transient and fully coupled This section briefly summarizes results from Hazzard et al. also changes the dimensionless excavation rate (R) in Equation [4]. Figure 3. The steady-state solution exhibits an FOS 20 – 30% higher than the transient solution. Transient pore pressures after excavation stage 3 (left) compared to steady-state pore pressures (right) for k = 10-8 m/s and n = 1%. . Figure 5 (top) shows that the FOS for a transient analysis is within 2% of the FOS for a steady-state analysis only for low R (< 0. therefore.

Comparison of FOS for transient and steady-state analysis (k = 10-8 m/s and n = 1%). The bottom shows results for a constant head boundary at 5000 m. . Factor of safety in the transient (one-way coupled) solution relative to the steady-state solution for different dimensionless excavation rates. Figure 5.Figure 4. The top shows the results for a constant head boundary at a distance of 2500 m from the pit centre (as in Figure 1).

. Stage 1 is excavated at 216 months.30% higher than in the two-dimensional (plane strain) models. This figure shows that the FOS for two-way coupling is almost always greater than that for one-way. The increase in FOS is related to the permeability. a one-way coupled solution will yield conservative results (calculated FOS is too low). Figure 6 shows the different behaviours. The pore pressure then gradually recovers with time. for most large. permeability. there is an initial drop in pore pressure when the rock is excavated. The degree of conservativeness depends on the value of R as shown in Figure 7.3. It was found that pore pressures were higher in the axisymmetric models due to the geometry of the drainage face. the FOS in the axisymmetric models was 5 . etc. with the greatest increases being observed for the lowest permeability models.3 Undrained analysis As described above. k = 10-8 m/s and n = 1%. open-pit analyses. simplified two-way coupling can be simulated by first calculating the undrained response. Figure 6. Therefore. For the two-way coupling.1. If FOS values are plotted relative to the steady-state solution in a porosity-versus-R diagram. due to increased confinement.1. Pore pressures recorded at a point 100 m from the slope face and 400 m below the ground surface.2 Effect of dimension Axisymmetric models were compared to two-dimensional flow models. due to the stress decrease. depending on the rate of excavation. However. then results similar to those of the 2D models (Figure 5) are obtained 3. These pore pressure drops will result in a slope that is more stable. and then calculating the transient (drained) solution. Figure 7 shows the FOS for two-way coupled analysis relative to the FOS for one-way coupling. The pore pressure drop due to the undrained response does not fully recover to equal the pore pressure in the transient analysis except for very slow excavation rates or high permeabilities (low R).

when the permeability is low. the infiltration increases the pore pressures and raises the water table.Figure 7. For lower permeabilities. The maximum decrease in FOS relative to models with no infiltration is about 20% when porosity is low and permeability is approximately equal to the infiltration rate. One-way coupling was assumed for all simulations. This plot also shows that infiltration has less effect as porosity increases.2 Infiltration This section uses the numerical model to explore the effect of meteorological infiltration (rain) on slope stability. 3. When permeability is high (e. The average FOS values for excavation stages 1 to 6 were calculated for models with different permeabilities and porosities. Both constant and seasonal infiltration were simulated. This is attributed to relatively lower level of resaturation. It was found that the effect of infiltration on pore pressure is a function of the infiltration rate relative to the permeability of the material. Complete resaturation occurs in this case. The net infiltration rate is rain minus evaporation. 3. This plot shows that the lowest FOS values tend to correspond to infiltration rates approximately equal to permeability. Figure 9 shows the results. whereas for k = 10-8 m/s. Other net infiltration rates were also tested (q = 1 m/ year and q = 0.2. the rain runs off and the FOS is affected less. the model quickly becomes saturated and much of the rain runs off. For the model with k = 10-6 m/s. k = 10-6 m/s). This was thought to represent the maximum infiltration rate likely to be observed in the field. . the maximum decrease in FOS is about 6%. However. The pore pressure changes are reflected in the FOS as shown in Figure 8.3 m / year). Different net infiltration rates were tested. FOS for two-way coupled analysis relative to one-way coupled. For permeabilities much higher than the infiltration rate. These values then were compared to the FOS for the case of no infiltration.1 Constant infiltration rate An infiltration rate of q = 3 m / year (~10-7 m/s) was simulated.g. the rapid drainage ensures lower pressures and higher FOS. Results were similar to those shown here. the decrease in FOS due to infiltration is up to 15%.

and two times the average infiltration rate for six months. Porosity = 1%. and compare the results to the FOS calculated for a constant infiltration rate. It is clear that the high permeability model is affected greatly by the seasonal variations. Example pore pressure histories are shown in Figure 10 for models with k = 10-6 m/s and k = 10-8 m/s. To examine the effect of the seasons on FOS. this may or may not correspond to peak pore pressures. four different scenarios were simulated to consider different offsets for the start of the wet season. then we should pick up the peak pressure state for each excavation stage. However. FOS for an infiltration rate of 3 m / year relative to no infiltration. Figure 9. Depending on the start of the wet season. We then can calculate the average FOS over stages 1 to 6 for this minimum envelope.2 Seasonal infiltration rate The effect of seasonal infiltration was examined by alternating the rate of infiltration between zero for six months. 3. whereas the low-permeability model is not. . if we construct an envelope encompassing the minimum FOS for the four different scenarios. Factors of safety were calculated nine months after each excavation.2. The FOS for the different seasonal simulations in one model are shown in Figure 11. FOS for model with k = 10-6 m/s (left) and k = 10-8 m/s (right) at different net infiltration rates.Figure 8.

the maximum decrease in FOS caused by seasonality is only about 2. n = 1%.Figure 12 shows the results for q = 3 m / year.003).3 m / year. unless R < 0. Figure 10. the effect of seasonality becomes even less severe. Factors of safety for different seasonal simulations in the model with k = 10-6 m/s. Figure 11. . for high permeabilities and/or low excavation rates (R < 0. For q = 0.5 % compared to constant q. However. This plot clearly shows that the seasonal variations essentially have no effect on the FOS. Example pore pressures for seasonal infiltration variation compared with constant rate for the model with k = 10-6 m/s (left) and k = 10-8 m/s (right). n = 1%. except that as q decreases. the seasonality may result in a decrease in FOS of up to 8%. The history point is located 400 m below the ground surface and 150 m from the slope face. Similar results are obtained for the other infiltration rates.01.

Therefore for D = 5000 m. Rs ~ 0.02 at n = 0.  The initial water table is close to the surface. If delays occur. 4 Conclusions This paper presents results for a ‘typical’ open-pit mine.1. Rs will decrease proportional to D2.005 at n = 1% (see Figure 5). The value of R below which steady-state modelling is acceptable (Rs) depends on porosity and the distance to the vertical recharge boundary.e. Factors of safety for models seasonal infiltration (minimum) relative to constant infiltration (q = 3 m/year). transient one-way coupled models can be used.02). the one-way coupled analysis will produce FOS values approximately 10% too low.5 pit radii from the pit centre (2500 m in this example). As the porosity increases.1%. transient and coupled analyses. however. For values of R > Rs.  Axisymmetric models produce higher FOS values but the calculated values for Rs are similar to the twodimensional models. If these assumptions are approximately valid. At R = 1. Rs ~ 0. Several simplifying assumptions have been made including the following. the underestimation will be approximately 20%. Rs decreases (at n = 1%.1% and Rs ~ 0.  The rock mass can be represented by an equivalent continuum. then the reduction in pore pressure .. Twoway coupled (undrained-drained) solutions will yield higher FOS values due to the initial drop in pore pressure caused by the stress drop that occurs when material is removed. a steady-state solution will produce unconservative results (a false sense of stability). For a model with the vertical recharge boundary at approximately 2. where D is the distance to the boundary. low excavation rates or high permeabilities).  For R > Rs.  Steady-state flow modelling is only acceptable for low dimensionless excavation rates (i. care should be taken to ensure that the pit is excavated in the predicted manner. This study suggests that two-way coupled analysis is the best approach. The degree of conservativeness depends on R.Figure 12.07 for n = 0. Rs ~ 0. For R = 0.  The equivalent continuum is isotropic and homogeneous. then the following conclusions can be drawn regarding steadystate. As the vertical recharge boundary is moved farther from the pit. but the results will be conservative.

The effect of inhomogeneity and anisotropy in groundwater properties currently is being investigated. Proceedings. 5 Acknowledgements The authors would like to acknowledge the support of CSIRO on this project.. (2009). USA. Géotechnique 49: 835840. Drescher. J. In Proceedings of the 64th Canadian Geotechnical Conference. Slope stability analysis by strength reduction. k. then water runs off and the increasing infiltration rate has no effect. Collingwood. London.  The factor of safety (FOS) decreases by about 20% when q ≈ k. pp. 19-41. Itasca Consulting Group. If q >> k. pp.). The influence and control of groundwater in large slopes. Spon Press.  Seasonal variation in infiltration rate does not affect the FOS dramatically. (2011). Damjanac. (2007). (1977). E. MN. In addition. Hoek. 141-200. Brown. Hydromechanical coupling and pit slope movements. . Slope Stability 2007. In Read & Stacey (eds. in press. Perth. Continuing research will address the three assumptions stated at the beginning of this section. A.  Infiltration has the most dramatic effect on slope stability when the rate of infiltration.. SME/AIMMPE. Dawson. (1982).. FLAC (Fast Lagrangian Analysis of Continua). W. Minneapolis. Bray. (1999). q.. CSIRO Publishing. M. Detournay.D. Hydrogeological model. pp. is approximately the same as the rock permeability. models with discrete fracture networks are being examined to determine when it is appropriate to use an equivalent continuum and what the properties of that continuum should be. H.).0. L. 3-43. the seasonal variations can cause a decrease in FOS up to 5%. A. resulting in unconservative predictions. In Potvin (ed.. Developing rules of thumb for groundwater modelling in large open-pit mine design. If q << k. Sullivan. Lorig. Toronto. Hazzard. T. B. then low level resaturation ensures the rain has little effect. Guidelines for Open-pit Slope Design. G. J. C. Australian Centre for Geomechanics. Version 6. 6 References Beale.). Roth. In Brawner (ed..(and resulting stability increase) caused by the undrained phenomenon may not be realized. In this case. E. except for (very) highpermeability models. 3rd International Conference on Stability in Surface Mining (Vancouver. New York. 2nd ed. Rock Slope Engineering. June 1981). 2008. Inc.