AUDIT OF MODEL PREDICTIONS OF DEWATERING REQUIREMENTS FOR A LARGE OPEN PIT GOLD MINE

Transcription

1 AUDIT OF MODEL PREDICTIONS OF DEWATERING REQUIREMENTS FOR A LARGE OPEN PIT GOLD MINE Joanna Moreno, URS Corporation, Denver Colorado Peter Sinton, URS Corporation, Denver Colorado Richard Vogwill, URS Corporation, Perth, Australia

2 AUDIT OF MODEL PREDICTIONS OF DEWATERING REQUIREMENTS FOR A LARGE OPEN PIT GOLD MINE Joanna Moreno, Peter Sinton, and Richard Vogwill URS Corporation Denver, Colorado and Perth, Australia ABSTRACT Groundwater models are often used to estimate and optimize dewatering and groundwater supply schemes for mines. Projections of future dewatering rates are highly dependent on the conceptual model used as a basis for the numerical model. This paper presents an audit of modeled dewatering rates versus actual data and an examination of the reasons for discrepancies. The setting is a gold mine, in a collapsed volcanic crater breached by the ocean. Mining will occur to about 200m below sea level within a remnant geothermal system. A detailed hydrogeologic investigation was conducted to support dewatering and geothermal depressurization design for future mining operations. These studies included development of a three-dimensional model to simulate heat and densitycoupled groundwater flow, using software developed by URS/Dames & Moore. In this paper, the groundwater modeling results prepared 5 to 10 years previously by others was audited using operational dewatering data. This model was then extended using the new mine schedule and actual locations for dewatering wells to predict future dewatering requirements. In addition, a comparison between hot and cold model results was used to assess the need to use a heat-coupled versus a standard groundwater model for dewatering predictions. HYDROGEOLOGIC SETTING The setting is a gold mine on an island, in a collapsed volcanic crater breached by the ocean. The deposit being mined has a fairly high hydraulic conductivity (10 to 40 m/d), with a flat hydraulic gradient prior to dewatering. The surrounding caldera walls have a significantly lower hydraulic conductivity (about 0.01 m/d), with steeper hydraulic gradients and higher water levels than within the caldera itself. Therefore the center of the mined area can be considered a bath tub of groundwater, which, during dewatering, is fed by inflow from the caldera walls, geothermal upwelling, some meteoric recharge, and increasingly by seawater inflow. Groundwater in the orebody is above boiling point in many areas. Mining will occur to about 200m below sea level within the remnant geothermal system. A detailed hydrogeologic investigation was conducted to support dewatering and geothermal depressurization design for future mining operations. These studies included development of a three-dimensional model to simulate heat and densitycoupled groundwater flow, using software developed by URS/Dames & Moore and extended to couple heat transport by the authors and others. In this paper, the groundwater modeling results prepared 5 to 10 years previously by others were audited using operational dewatering data. This model was then extended using the new mine schedule and actual locations for dewatering wells to predict future dewatering requirements. This paper describes how the modeled boundary conditions and assumptions about the geothermal system have affected modeled results. From a modeling standpoint, the site poses an interesting combination of difficulties: saltwater intrusion, heat-coupled flow, and time-dependent boundary conditions from a multiphase model, as well as the usual difficulties of simulating a geologically complex and fractured medium. The following sections describe some of these aspects of the model. MODEL DEVELOPMENT Due to the need to maintain pit working faces at safe temperatures and pressures without the release of noxious gases, geothermal management (Figure 1) and groundwater management are both required. A multiphase geothermal model (using a code related to TOUGH2), prepared by others, was used to estimate geothermal depressurization needs and temperature and pressure boundary conditions over time for the groundwater model. Geothermal upwelling rates were calculated by calibrating the multiphase model to the observed temperature distribution. The geothermal model could not be used to evaluate the dewatering requirements due to limitations on the feasible number of model cells, at that time. Areas where higher temperatures indicate connection with geothermally active zones were assumed to be areas of geothermal upwelling, approximated in the model by either constant head (CH) or constant flux (CF) boundary conditions derived from the multiphase model results. Upwelling was assumed to increase over time as the mine is deepened and hotter groundwater is encountered. The groundwater model base boundary condition was changed from a CF to CH condition in year 5 of mining (allowing 5-fold greater geothermal inflows) to represent response to increasing upward hydraulic gradients as the mine progresses.

3 dewatering. Early in the prediction, the amount of groundwater released from storage is about the same as the amount pumped. By year 5, the ratio of seawater coming in at the harbour to pumped groundwater is about 0.5. For the base of the model, the ratio is nearly zero. By year 15, the ratio of seawater inflow to pumped groundwater is about For the base of the model, the ratio is about 0.2. These ratios indicate how pumping is likely to affect the flow of groundwater into the model, but do not indicate how much seawater, if any, may be have to be extracted in the future. Processes of Heat Energy Transport Figure 1. Geothermal Conditions Encountered During Exploration Drilling The heat/chloride/groundwater flow model comprises approximately 25,000 cells in 9 layers. Areas where prepumping groundwater levels were close to sea level and/or tidal responses were observed were assumed to be connected to the ocean. Ocean levels were approximated by a constant head at sea level. Flow balance studies were used to derive estimates for recharge and lateral inflows. HEAT AND SALTWATER EFFECTS Groundwater temperatures in the pit and caldera vary from 40 to 220 degrees C. To simulate dewatering options, a groundwater flow model that also simulates heat flow is required because: Viscosity changes due to temperature changes can be simulated. Viscosity changes by a factor of about 6 for the range of temperatures observed within the zone impacted by dewatering at this site. The hydraulic conductivity, and hence the groundwater flow rates, change by the same factor due to the change in viscosity. This effect resulted in dewatering pumping rates varying by a factor of 2 for the same drawdown in the pit, for the heat-coupled versus cold models. Density changes due to temperature variations can be simulated. This effect is similar to, but smaller than effects caused by changes in viscosity. Consequently, a density-coupled solute transport model (TARGET) was modified to simulate heat transport, as described below. Chloride concentrations were used for additional model calibration. Modelling predictions indicate that substantial amounts of seawater may begin to flow inland during The transport of heat energy in groundwater occurs by processes the same as, or analogous to, those involved in solute transport. Solute transport processes and their mathematical formulations are described, for example, by Bear (1979). Heat transport occurs by advection and mechanical dispersion of hot groundwater in exactly the same way as solute transport occurs by advection and dispersion of contaminated groundwater. Whereas molecular diffusion is also a factor in solute transport, the analogous process of thermal conduction is important in heat transport. Thermal conduction also occurs into or out of the rock or soil matrix, a process analogous to sorption/desorption of solute. Mathematical Approach The primary variable used in the TARGET code to model heat energy transport is temperature. Each model cell is assumed to be at thermal equilibrium (i.e. the groundwater and rock or soil are at the same temperature as each other). This assumption may be untenable at high groundwater flow rates, but in that case energy transfers between the groundwater and the rock or soil can be neglected by setting the specific heat capacity of the matrix to zero. In this way the matrix stores no heat energy and its temperature is irrelevant. The heat transport equation in TARGET is constructed from consideration of the heat energy content of a cell and temporal changes to that quantity. The heat energy content of a cell (in joules) is given by: (n S w c w T V) + ( b c s T V) where n = porosity of matrix (-) S = water saturation (-) w = density of water (kg/m 3 ) b = bulk density of matrix (kg/ m 3 ) c w = specific heat capacity of water (J/kg.deg.C) c s = specific heat capacity of matrix (J/kg.deg.C) T = temperature (deg.c) V = cell volume (m 3 )

4 The mathematical formulation of the heat transport equation is the same as for solute transport (Bear 1979, p. 239) with the following substitutions: SOLUTE HEAT m j c w T D l w / w c w + (1-n)l s /n S w c w M j c w T* Kd c s / c w where D = diffusion coefficient (m 2 /s) m j = solute concentration (kg/ m 3 ) M j = source/sink solute concentration (kg/ m 3 ) l w = water thermal conductivity (J/s.m.deg.C) l s = matrix thermal conductivity (J/s.m.deg.C) T* = source/sink temperature As for solute transport, the heat transport equation is solved by neglecting the non-diagonal diffusive terms of the equation. Fluid Density - Temperature Auxiliary Relationship: The variation of water density ( w ) with temperature (T) was derived from data given in CRC (1975) and is given by: w = x 10-5 ( T-3.98 ) This relationship is valid over the temperature range 0 to 100 degrees Centigrade. Fluid Viscosity - Temperature Auxiliary Relationship: The variation of water viscosity with temperature is calculated in TARGET from the following relationships (CRC, 1975): For temperatures between 0 and 20 degrees Centigrade: log( T ) = /x where T = temperature (deg.c) T = water viscosity at temperature T x = (T-20) (T-20)(T-20) and for temperatures between 20 and 100 degrees Centigrade: log( T / 20 ) = [1.3272(20-T) (20-T)(20-T)] (T + 105) The resulting numerical model was successfully tested versus several of the test cases (analytical and numerical results) presented by Voss (1984). The model presented in Voss (1984) is inapplicable to this application because it is two-dimensional (or radial). As hot groundwater is drawn into the dewatering wellfield, temperatures increase, viscosity decreases and hydraulic conductivities increase. Consequently, flow gradients through the pit are decreased and pumping efficiency is effectively increased. A comparison between hot and cold model results, for a cross section through the dewatering wellfield, is shown in Figure 2. Dewatering Wells Figure 2: Hot (black) versus Cold (gray) Groundwater Flow Vectors in the Area of the Dewatering Wellfield (isotherms in degrees C) This figure shows that hot model results predict higher flow rates at distance from the wells and lower flow rates adjacent to the wells, consistent with an effectively higher overall hydraulic conductivity in the hot model case. However flow directions are also affected, with the hot model results showing greater

5 zones of capture, and black and gray arrows in opposite directions in some parts of the flow field as a result. Further hot and cold model results are compared later. SOURCES OF UNCERTAINTY Uncertainties in the model include: the base boundary condition, the degree of anisotropy in the aquifers, the hydraulic conductivity and storage coefficient of the aquifers, the degree of connection with the ocean, and the adequacy of an equivalent porous medium assumption. Although the model base boundary condition was specified using the multiphase model results, this boundary condition was also tested. Hydraulic conductivities are fairly well known from many tests, but variations in temperature may be caused by heterogeneity in formations and spatial variation in hot upwelling. In addition, properties in previously unpumped parts of the pit carry a greater uncertainty. Due to the consistency in responses to pumping from surrounding monitoring wells it appears that the aquifer responds as an equivalent porous medium at least on the scale of the separation distance between wells. Sensitivity analyses in which other parameters were systematically varied are presented later. AUDIT OF PREDICTED VERSUS OBSERVED DEWATERING RATES Approximately one year of dewatering data have been collected and were available for the model audit. Both the mine plan and the dewatering schedule/well placement were different from the original plan. When the original (FS) model was updated with actual pumping rates and locations, the results were as shown in Figure 3. This figure shows that the FS model tends to under predict groundwater drawdown by about 10 to 20m. However, the simulated trends and pattern of drawdown were reasonable. MB14D (Central Pit) Water and Mine Level (mrl) 1,040 1,030 1,020 1,010 1, ,000 40,500 36,000 31,500 27,000 22,500 18,000 13,500 9,000 4,500 Pumping Rate (L/sec) Jun-97 Aug-97 Sep-97 Nov-97 Jan-98 Feb-98 Apr-98 Jun-98 Jul-98 Sep-98 Nov-98 Date Actual Water Level Lowest Mine Level, Actual FS Model, Water Level Actual Pumping Rate Figure 3: Predicted and observed hydrograph for dewatering well near the main pit REASONS FOR DISCREPANCIES Uncertainties in model boundary conditions, aquifer properties, and heterogeneity all potentially contribute to discrepancies between predicted and observed groundwater levels. In addition some wells monitor shallow sections of the aquifer and may not respond in the same way as deeper wells. Also, wells closest to the ocean may show less drawdown than predicted due to lower conductivity zones or greater lateral connection to the ocean (conclusions with opposite consequences). Therefore a series of sensitivity tests was undertaken. The sensitivity cases tested the following parameter changes. These variations do not represent observed or expected ranges, but rather extreme variations to test model responses: Case 1: lower storativity in most permeable aquifer Case 2: half hydraulic conductivity Case 3: introduce anisotropy

6 Case 4: reduction in specific yield Case 5: vertical hydraulic conductivity half the horizontal, and conductivity reduced to 80% of original values Case 6: lesser connection to sea Case 7: as case 6 with lesser conductivity connection to the sea Case 8: infiltration rates reduced to 30% of original values Case 9: as case 8 and specific yield reduced by 5% Case 10: as case 9 and vertical and transverse hydraulic conductivity reduced to half the longitudinal value Case 11: as case 10 and hydraulic conductivities reduced 10 fold Case 12, best fit model: All infiltration rates reduced by 60% Horizontal (X, model east-west) hydraulic conductivity of the two properties representing the aquifer were reduced 80% (from 7 to 1.4 and from 15 to 3 m/d) Horizontal (Y) hydraulic conductivity of the two properties representing the aquifer were reduced one order of magnitude (from 7 to 0.7 and from 15 to 1.5 m/d) Vertical (Z) hydraulic conductivity of the two properties representing the aquifer were reduced one order of magnitude (from 7 to 0.7 and from 15 to 1.5 m/d) Specific yield of all zones with horizontal hydraulic conductivity greater than 0.5 m/d reduced by 5% (aquifer zones went from 15% and 20% to 10% and 15%) Connection of aquifer to ocean reduced by introducing a band of one order of magnitude lower hydraulic conductivity where aquifer contacts the ocean. Case 13: as case 8 but specific yield and porosity reduced 100 fold Case 14: Geothermal upflow rates reduced to 10% of the initial values. Sensitivity results are summarized in Figure 4. Lower hydraulic conductivities, storage properties, and upwelling rates were the most sensitive input parameters tested, with predictions improved by the introduction of anisotropy and a reduction in the lateral connection to the sea (Figure 5). All of these changes point to conservatism in the original model, which therefore provide an upper estimate for future dewatering requirements. It is possible that as deeper levels are mined, zones of the mine previously unpumped and less densely monitored may result in unexpected conditions. Therefore some degree of conservatism is warranted. MB14D (Central Pit) 1,040 1,030 Case 11 Water and Mine Level (mrl) 1,020 1,010 1, Case 14 Case 4 Case 13 Case 7 Case 5 Cases 3 & 6 Case 1 Cases 2 & 14 Case 4 Case 8 Cases 9 & 11 Case 10 Revised Model Case Aug-97 Sep-97 Nov-97 Jan-98 Feb-98 Apr-98 Jun-98 Jul-98 Sep-98 Nov-98 Revised Model (Case 12) Actual Water Level Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8 Case 9 Case 10 Case 11 Case 13 Case 14 Date

7 Figure 4: Sensitivity of Predicted Drawdowns to Aquifer Parameters and Boundary Conditions MB14D (Central Pit) Water and Mine Level (mrl) 1,040 1,030 1,020 1,010 1, ,000 40,500 36,000 31,500 27,000 22,500 18,000 13,500 9,000 4,500 Pumping Rate (L/sec) Jun-97 Aug-97 Sep-97 Nov-97 Jan-98 Feb-98 Apr-98 Jun-98 Jul-98 Sep-98 Nov-98 Date Revised M odel, Water Level Actual Water Level Lowest M ine Level, Actual Actual Pumping Rate Figure 5. Best Fit Model Predictions Since the best-fit model contained non-conservative assumptions about aquifer properties and saltwater intrusion rates, this model estimated a significant reduction in required pumping rates over time as groundwater in storage is pumped out and slowly replenished. Therefore two alternative, more conservative, cases were extended out over the life of the mine (15 years). The predicted results are shown in Figure 6. Results showed that current pumping rates (actual and predicted) are higher than forecast previously because of the delay in initial dewatering relative to mining commencement. Relatively lower pumping rates are predicted for times when the lowest mine level does not change, and Relatively higher rates are predicted when the lowest mine level drops rapidly. It can be seen that the heat-coupled model produces a dramatically different result than the cold-water model, especially near the pumping center where upwelling and flow rates are greatest. However, CH versus CF model base boundary condition resulted in little change in the predicted drawdowns. This is probably because geothermal upwelling is only about 10% of the total inflow to the mine, with the balance provided by seawater inflow, inflow from the caldera, reduction in storage, and meteoric recharge. CONCLUSIONS The hot model produces significantly different predictions than the cold model, suggesting that for the observed tempreature ranges a standard groundwater flow model would provide inaccurate results. This is particularly important close to the dewatering center where the effects of heat transport, and the flow rates, are the greatest. Alternative explanations are available to explain observed discrepancies between predicted and observed flow patterns. The most sensitive parameters for dewatering predictions are hydraulic conductivity, storage, and upwelling rates, with anisotropy and connection to the ocean also playing a part. It is likely that geothermal upflow rates have been conservatively overestimated, contributing to upper bound pumping predictions.

8 1,100 Predictions Using the FS Model with Actual and Planned Mine Level 1,600 1,050 1,333 1,000 1,067 Elevation (mrl) Heat Flow Not Simulated Pumping Rate (L/s) Heat Flow Simulated 850 Heat Flow Simulated (CF) /7/95 12/6/97 12/6/99 12/5/01 12/5/03 12/4/05 12/4/07 12/3/09 12/3/11 12/2/13 Date Lowest Mine Level, Actual and Planned (October 1998) FS Model, Groundwater Elevation FS Model, Groundwater Elevation, Heat Flow Not Simulated FS Model, Groundwater Elevation (CF) FS Model, Pumping Rate Figure 6. Predicted Future Dewatering Requirements Inaccuracies of the model are likely to introduce increasing errors into estimated dewatering requirements further into the future. For this reason, modeling predictions are more accurate for the first few years, relative to predictions late in the dewatering simulations. By the end of the 15-year simulation, there is as much as 20 m difference in predicted groundwater elevations in the deepest part of the pit for the two heat-coupled cases. The model cannot be used to differentiate between drawdowns of less than about 10 meters, due to remaining discrepancies between simulated and actual water levels. It is possible that the vertical grid would need to be refined, to more accurately simulate actual drawdown. The case that would most accurately represent future drawdowns can be identified by comparison with future dewatering data. Audit of model predictions can provide new insight on how the hydrogeological system responds to pumping and mining. This insight can in turn be used to design better monitoring networks and pumping systems to improve and streamline operations. CRC, 1975, Handbook of Chemistry and Physics 56th Edition, ed. R.C. Weast, CRC Press, Inc., Cleveland, Ohio. Voss, C.I., 1984, A finite-difference simulation model for saturated-unsaturated, fluid-density-dependent groundwater flow with energy transport or chemically-reactive single-species solute transport U.S. Geological Survey, Water-Resources Investigations Report Number ACKNOWLEDGEMENTS Stephanie Williamson, formerly of Dames & Moore (Perth, Australia), provided the scope, technical oversight and support for this work; her support is gratefully acknowledged. John Forth, formerly of Dames & Moore (Brisbane, Australia), performed the original model setup and calibration briefly described here. His pioneering application of the heat code is gratefully acknowledged. Jonathon Larkin, formerly of Dames & Moore (London, UK), performed the heat transport code revisions and validation testing described here. His work is gratefully acknowledged. REFERENCES Bear, J., 1979, Hydraulics of Ground Water McGraw Hill Series in Water Resources Environmental Engineering.