Appendix C-2

Case Study 2 - Underground Mine in Northwestern British Columbia

C2-1 Project Overview

Case Study 2 is situated in northwestern British Columbia. The project site ("Site") comprises glaciated alpine terrain that slopes into a glacial outwash and till valley. Several significant streams and one lake occupy the valley immediately downstream of the Site (Figure 1). The underground mine, as proposed, will produce several thousand tonnes of molybdenum per day with access from one historic adit and one newly driven adit. Waste rock will be processed and used as backfill; hence above-ground, long-term storage facilities will not be required. An ore load-out facility and temporary rock dump will occupy the lower reaches of the Site.

Figure C2-1: Case Study 2 location map illustrating location of main creeks (blue lines), access roads (red and brown lines) and project site (green switchbacks and red portal).

The operation of the mine will induce groundwater discharge to the underground workings and, as a result, has the potential to adversely impact fish-bearing surface waters by decreasing stream baseflows. Seepage from the mine workings may also threaten surface water and groundwater drinking sources for local residents. Water sampled from the historic adit demonstrates existing elevated molybdenum and arsenic concentrations. Due to the proximity of the project site to residential drinking water sources and fish-bearing habitat, this project carries a high degree of risk to the local receiving environment (high VEC).

The project has been modeled at two levels. First, a detailed numerical model was developed to simulate discharge to the mine workings and changes to groundwater flow and quality over the course of mine development and closure. Secondly, after initial review, the numerical model was updated and the modelling approach refined through improved conceptualization and uncertainty analysis to more confidently quantify the discharge to the underground workings and to assess potential impacts to surface water discharge processes. Furthermore, the improved numerical model was used to simulate the transport of adit seepage over 100 years after mine closure.

C2-2 Modelling Objectives

Modelling was undertaken at two distinct periods during the application for an environmental assessment certificate. During the initial modelling stage, the model was used to estimate discharge from the mine workings over the 10 year mine life and for 1000 years post-mining. The purpose of the model was to establish groundwater flow directions and quantities, and to assess the potential for mine-affected groundwater to impact downstream water sources. This initial modelling study was, however, based on very limited data and a relatively simple modelling approach that did not adequately address the high levels of uncertainty encountered. As a result, a second phase of modelling was undertaken.

The objectives of the second, more comprehensive, modelling effort were twofold:

• To provide an updated estimate for discharge to underground workings over time, and
• To assess the impacts of the inflow into the mine on surface water discharge processes, and
• To assess the impacts on well users downhill of the proposed mine.

This second model was constructed on the basis of the initial model but included several key modifications, namely:

• Revision of the model domain to allow interactions between creeks in the adjacent watersheds and the mine,
• Simplification of the representation of mine workings,
• Extensive sensitivity analysis of key parameters, and
• More thorough transient simulations related to all stages of the mine plan

However, it should be noted that the field data supporting the model remained the same.

C2-3 Data Review

Geologic data were available from the exploration logs of the Proponent and residential well drilling logs from the provincial government. This data aided in the conceptualization of the Site geology and of hydrogeologically significant units. Borehole logs were reviewed to aid in site conceptualization and hydraulic testing was performed in select exploration holes to establish hydraulic properties of the subsurface.

Hydrogeologic data comprises seasonal static water levels from five existing residential wells and from five additional monitoring wells distributed across the watershed. Baseline water quality samples were also analyzed from these wells. Well locations are plotted in Figure 2. The dataset is very limited but spatially distributed across the Site and in downstream receiving environment (upstream of Kathlyn Lake).

Hydraulic testing of the bedrock was accomplished by packer testing in eight exploration holes. However, these tests did not address horizontal anisotropy of hydraulic conductivity. Slug testing was conducted in the shallow alluvium. Seep estimates from the historic adit were also available at the time of model preparation.

No assessments were reported on the nature of bedrock fractures at the site to support assumption of equivalent porous media for the numerical model. There were also no assessments of effective porosity or specific yield of the bedrock units available to the modeller, parameters used to estimate groundwater velocities and travel times.

Figure C2-2: Location of monitoring wells (red) and residential wells (green) in the project area.

C2-4 Conceptual Model

Initial Conceptual Model

A three dimensional, finite difference model code (MODFLOW) was initially selected to meet the number of modelling objectives. The selected code is widely accepted and includes a parameter estimation and optimization sub-routine (PEST) for calibration.

The initial model was conceptualized using drilling records and geologic data from a number of sources, including:

• Geologic mapping/model,
• Exploration drilling and testing, and
• Residential well logs from MOE's WELLS database

The conceptual model includes delineation of hydrostratigraphy, hydrogeologic properties of those strata, relationships between groundwater and surface water features, regional hydrologic (climatic) conditions, the mine development and closure plan, and an understanding of potential receiving environments (Figure 3).

Figure C2-3: Conceptual model used for MODFLOW modelling.

The sequence of modelling included several key steps, namely:

• baseline calibration to static water levels and seep estimates from the historic adit
• steady state simulation of mine operations
• transient simulation of post-closure conditions

Updates to the Conceptual Model

For the updated numerical model, MODFLOW was selected by the proponent as the most appropriate code for the modelling objectives. Furthermore, it was agreed that the code was commonly applied in practice and could be easily reviewed, hence lending to transparency during the decision making process. Although continuum models, like MODFLOW, do not address discrete bedrock fractures, the modeller notes that the applicability of an equivalent porous media approach to simulations of flow and transport in heterogeneous, fractured bedrock settings has been well proven and documented in other studies elsewhere. However the justification did not appear to be based on assessments of the conditions of the bedrock units at the site. The assumption of equivalent porous medium became a significant point of uncertainty from the perspective of the government reviewers.

The original MODFLOW model was updated in several important areas. First, the model domain was both expanded to account for flow to adjacent watersheds, and simplified with respect to simulation of the complex underground workings. Second, a more thorough sensitivity analysis was conducted to assign ranges of key parameters to the model. Finally, transient simulations were used to revise estimates of previous steady state simulations during mine operations. Transient simulations were also used to predict flow and transport under different closure scenarios. These updates significantly improved the model.

The sequence of simulations for the updated model comprised the following steps:

• steady state calibration to heads and flow
• transient simulation of adit construction
• transient simulation of mine operation
• transient simulation of post-closure rebound, flow, and transport

The modelling approach used for the updated model is outlined in Table 1.

Table C2-1: Modelling approach for updated model.

Modelling progressed from steady state calibration, to transient simulations of mine development, operation, and closure. To account for uncertainty in the conceptual model (in this case, the delineation of higher-K fractured bedrock), two distinct conceptual models were calibrated and used for model predictions. For both conceptual models, sensitivity analyses were performed on key parameters, namely:

• Drain conductance
• Fault location(s) and conductivity
• Hydraulic conductivity of key strata

During the sensitivity analyses, the parameters were adjusted within their reasonable limit, or until the calibration targets were not satisfied.

Pathline analysis and a solute transport simulation were carried out to assess the post-closure impacts of mine-affected groundwater on the receiving environments.

C2-5 Numerical Model Setup

Initial Model Setup

For the initial model setup, the Site was discretized into 45 rows, 130 columns, and 40 layers. Grid spacing was refined near the underground workings and adits. The model grid was bounded by a major river to the east, a topographic divide (Coast Mountains) to the west, and to the north and south to include potentially sensitive creeks. Boundaries to the north and south were assigned no flow boundaries, while the Coast Mountains to the west were treated as a general head boundary. The river to the east was represented with a river boundary with elevation and discharge values determined from field investigations. Stream boundaries were assigned to the two streams of interest. A constant head boundary was assigned to the lake representing its measured elevation. Recharge was applied to the model domain with values representing the balance of precipitation, runoff, evapotranspiration, contributions from up-gradient sources (faults or fractures), and sub-glacial seepage where appropriate. Table 2 provides a summary of these assigned boundary conditions and Figure 4 provides a visual representation of the model boundaries for the initial model.

Table C2-2: Summary of assigned boundary conditions.

Figure C2-4: Boundary conditions and model discretization for the initial model.

To simulate the discharge to the underground mine, inactive cells (representing the workings) were surrounded by drain cells. The drain cells simulated bedrock seepage faces. Drains on the sides of the workings (representing vertical faces) were assigned heads at the middle of the cell, while drains above and below the workings were assigned head values at the top of the cells. Figures 5 and 6 illustrate the drain set-up.

Figure C2-5: Drain and inactive cell set-up for representing mine workings.

Figure C2-6: Mine working and drain set-up for initial model.

Updated Model Setup

In the updated model, the domain was expanded to the north and south to allow for interactions between the adjacent watersheds and the mine. The updated model domain and boundary conditions are illustrated in Figure 7. Fourteen surface water features were included as constant head and general head boundary conditions (lakes and marshes, respectively).

Figure C2-7: Model domain and boundary conditions for updated model.

The vertical discretization was reduced from 40 layers to 17 layers without compromising the delineation of hydrostratigraphic zones. To represent the underground mine workings, simple drains replaced the more complicated inactive zone-drain combination used in the updated model (Figure 8).

Figure C2-8: Updated model with simplified expression of mine workings as drains.

Recharge was applied to four distinct regions of the model domain to represent the various hydrologic regimes present (Figure 9).

Figure C2-9: Recharge zones for updated model.

C2-6 Model Calibration

Initial Model Calibration

The model was initially run with steady state conditions for 40 years to calibrate the simulated historic adit discharge to the observed adit discharge. This time period represents the age of the historic adit.

Water level calibration data are notably limited in this study. The steady state flow model was calibrated to water levels in just five boreholes and a water level from a hole in the floor of the adit. The Parameter Estimation and Optimization (PEST) sub-routine for MODFLOW was used to meet calibration targets within the 95% confidence interval (Figure 10).

The calculated root mean square (RMS) error was approximately 5% and was judged by the modeller to be acceptable for this simulation.

Packer testing showed significant variability in estimated bedrock hydraulic conductivity, hence multiple cases were simulated in the initial round of modelling. Both isotropic and anisotropic conditions (vertical only; hydraulic conductivities of the bedrock units were assumed to be isotropic in the X-Y plane) were applied to three variations of hydraulic conductivity distributions, for a total of six model scenarios. This approach provided a range of discharge estimates corresponding to the uncertainty in the magnitudes and spatial distribution of hydraulic conductivity within the conceptual model. From the simplistic isotropic models, the scenario with the tightest distribution of conductivity that excluded apparent outlier values was considered most reasonable. From the more realistic anisotropic cases, it was determined that the most probable estimate of bedrock hydraulic conductivity (based on the resultant discharge values) lies between ½ order of magnitude above the geometric mean to 1 order of magnitude below the geometric mean of the packer test estimates. From this rationale, a plausible upper bound and lower bound on bedrock conductivity was proposed with a corresponding range of plausible discharge values.

Figure C2-10: Observed vs. simulated water levels after model calibration.

Updated Model Calibration

The calibration was carried out in two distinct phases of parallel modelling. Due to uncertainty in the conceptual model, two simulations (Model A and Model B) were carried through the steady state head and flow calibration phases. Both models are based on the hydrogeologic units presented in Figure 11. These simulations did not initially include the major fault zone inferred to run through the Site.

Figure C2-11: Distribution of hydrogeologic units for First Calibration.

Head and flow values available for calibration of the steady state model were very limited. The model was calibrated to head measurements in eight wells, one fracture zone in the area of the new adit, and long-term, steady flows from the historic adit. The hydraulic conductivity values of the various rock formations were adjusted within a reasonable range to calibrate the simulated adit seepage to the actual observed outflow ("1066 Adit"). Table 3, Figure 12, and Figure 13 present the results of this first round of calibration.

Table C2-3: Results of First Calibration runs.

Figure C2-12: Results of First Calibration (Model A).

Figure C2-13: Results of First Calibration (Model B).

Calibration results for Model A and Model B show normalized root mean square error (NRMSE) of less than 10%, which the modeller judged as acceptable given the size of the dataset.

C2-7 Model Predictions & Sensitivity Analyses

Initial Model Predictions

The initial model was run at steady state for 40 years to calibrate adit discharge and subsequently for 10 years to simulate the life of the project. The groundwater flow field for the Site is presented Figure 14. Following operations, the model was run transiently for one thousand years to simulate post-closure conditions.

Figure C2-14: Groundwater equipotentials and flow directions.

The steady state runs were used to estimate a range for discharge to the mine workings (Figure 15) and to generate travel time estimates (based on assumed porosities) for mine-related particles travelling to downstream receptors.

Figure C2-15: Groundwater flow to mine workings.

The transient runs were used to assess the infilling rate of mine workings after closure and the change in flow field as a result of mine flooding. Transport modelling runs were also built off the transient simulations.

Updated Model Predictions

The calibration models derived from the steady state calibration were used as initial conditions for the transient simulations of mine development, operation, and closure.

The two models, slightly different in concept, were used to provide a range of estimates of discharge over the mine life (Figure 16).

Figure C2-16: Discharge estimates for adit construction and mine operation.

As Model B provided higher estimates of discharge (more conservative), it was selected for a sensitivity analysis of drain conductance. Drain conductance in this model represents, in effect, the efficacy of grouting employed to prevent seepage into the underground workings. This is effectively an operational parameter that may be adjusted during construction. A higher drain conductance in the simulation represents a less effective grout, while a low conductance would represent a grout that precludes seepage into the workings. Figure 17 presents the results of this sensitivity analysis.

Figure C2-17: Sensitivity analysis of simulated drain conductance.

Further sensitivity analysis was performed with respect to the inclusion of a prominent fault structure through the Site. Inclusion of the fault zone in the model only slightly reduced discharge to the mine workings by effectively lowering the water table in the mine area. The only scenario that would result in increased discharge to the mine is a highly transmissive fault that effectively connects the underground workings to the area of sub-glacial recharge west of the mine; however, this scenario was simulated to increased discharge only slightly.

In addition to Model A and Model B of the First Calibration stage, a Second Calibration phase involved the inclusion of a new hydrogeologic unit in the mine area to represent a zone of (relatively) higher hydraulic conductivity. The distribution of hydrogeologic units of this third model is presented in Figure 18. Furthermore, the rate of sub-glacial recharge was increased in order to assess model sensitivity to this parameter.

Figure C2-18: Model from Second Calibration phase with new hydrogeologic unit in the mine area.

Again, the calibration targets were achieved for this conceptual model and are presented in Figure 19.

Figure C2-19: Calibration results for Second Calibration model.

The third model predicts higher discharge values compared to the models of the First Calibration. These results are compared in Table 4.

Table C2-4: Comparison of simulated discharge.

Further sensitivity analyses were undertaken to observe the effect of hydraulic conductivity perturbations on the solutions from the Second Calibration. The hydraulic conductivity values assigned to each hydrogeologic unit were systematically varied to assess which have the greatest impact on mine discharge values. Through this process, it was determined that the model was most sensitive to the rock unit that hosts the mine workings.

Additionally, the model sensitivity to recharge was studied. Sub-glacial recharge west of the mine was shown to dramatically influence discharge estimates.

From the combination of models and sensitivity runs, an upper bound estimate of mine discharge was estimated. This estimate represents the worst-case (most reasonably conservative) scenario under the range of simulated conditions.

Further estimates of streamflows were provided, all of which are similar across the three main model runs.

C2-8 Transport Modelling

Initial Model Transport Simulations

During the initial round of numerical modelling, a particle tracking module (MODPATH) was used to estimate the advective path and travel time of a group of particles introduced to the system (Figure 20). Particles were assigned to the new adit, the underground stopes, and beneath the load-out facility. The time required for a particle to reach a common downstream receptor (given upper and lower bound conditions) was used to assess the risk to the receiving environment.

Figure C2-20: Pathline analysis using MODPATH.

Following the MODPATH simulations, the solute transport module, MT3DMS, was used to simulate the transport of total organic carbon (TOC) to the receiving environment. A single value representing the initial TOC concentration was assumed based on a consultant's report. The transport simulation was conducted as two separate scenarios: (1) where TOC does not react or degrade with time, and (2) where TOC is reactive and decays with time.

Three observation wells were created in the model domain to track the break-through of TOC over time (Figure 21). These wells were located upstream of the sensitive receiving environments and distributed spatially (across the domain as well as with depth). The results are presented as break-through curves in Figure 22.

Figure C2-21: Location of simulated observation wells (purple) along with existing residential and monitoring wells (green).

Figure C2-22: Break-through curves for TOC.

Updated Model Transport Simulations

During operation, the mine will act as a sink and groundwater flow to receptors is unlikely to occur. However, after mine closure and rebound of the water table, mine-affected groundwater will migrate and may impact the downgradient environment. Particle tracking (MODPATH) confirmed that mine CoCs may reach groundwater discharge points several decades after closure. The particle tracking exercise helped to identify potential receptors, while not able to provide estimates of CoC concentration. At this point, an additional sensitivity analysis was performed to assess the effect of the shallow till hydraulic conductivity. The effect was to deflect particles towards different areas of the receiving environment (e.g. away from a lake and towards a river) which could have significant implications regarding receptor sensitivity and dilution capacity.

An analysis of geochemical data from the historic adit provided evidence to support the enlargement of the hydrogeologic unit hosting the mine workings. This was performed as the Third Calibration phase and this final model - which still agreed with flow simulations from the First and Second Calibration models - was used for transport simulations.

The transport module MT3D was used to simulate non-reactive transport of a nonspecific mine CoC from the mine workings (representing 100% concentration). The default values for dispersivity were adopted. Simulations of post-closure transport indicate that supply wells and surface water will not receive a significant percentage of the solute in the 100 year timeframe simulated (Figures 23, 24, 25).

Figure C2-23: Predicted concentration contours (blue) at 100 years after mine closure for depths represented by Layers 2 and 5 of the model. Contours at 50, 10, 1, and 0.1% of mine concentration.

Figure C2-24: Vertical distribution of mine solute at 30 and 100 years after mine closure. Contours at 50, 10, 1, and 0.1% of mine concentration.

Figure C2-25: Mine-affected groundwater concentrations at local supply wells.

Additional analyses were performed with a subsequent model to show that the downstream supply wells would receive only 1% more groundwater from the mine area compared to pre-mining conditions.

On the basis of transport simulations and the predicted path of the mine-affected plume, additional groundwater monitoring well locations were recommended by the modeller.

C2-9 Case Study Evaluation

The initial numerical model was limited by scarce calibration data and a wide range of interpreted hydraulic parameters from hydraulic testing - this was deemed to be a significant point of uncertainty by government reviewers. A statistical approach was used to assess the plausible range of hydraulic conductivity values; however, in doing so the entire range of hydraulic testing results were not adopted (some values were deemed as outliers).

The updated model was limited by the same sparse dataset and uncertainty in the conceptual model. However, a thorough sensitivity analyses provided a more defensible range of flow and transport solutions to meet the modelling objectives. In this case, the modeller acknowledged that the accuracy of the model is limited by the quality and quantity of data and the timeframe over which it was collected. Furthermore, the modeller demonstrated that model calibration is non-unique in that more than one set of parameters may result in a solution that meets calibration targets. The aspect of non-uniqueness was addressed quite thoroughly during this modelling exercise.

In both modelling phases, prior field investigations were acknowledged and data was incorporated into the conceptualization and calibration procedures. This data was, however, limited spatially and temporally.

The conceptualization of the hydrostratigraphy and groundwater flow is well presented and, in the case of the updated model, the modeller has made an attempt to capture the conceptual uncertainty in two baseline models.

The modelling code was selected for its transparency and for ease of review. Furthermore, the updated model was constructed in the likeness of the former simulations which allowed for some continuity and comparison between models. A 3D numerical model was essential for simulating the complex groundwater flow and discharge associated with the underground mine workings and spatial distribution of hydrogeologic units. The level of risk associated with the valued ecosystem components merited a detailed numerical model on which solute transport simulation could be based.

Transport simulations were conducted during both phases of modelling. Simulations included simple steady state pathline analyses as well as transient transport simulations of both reactive and non-reactive species. The updated model provided a sensitivity analysis regarding the permeability of aquifer units nearest the environmental receptors and a thoughtful analysis of the impact of mine-affected groundwater on supply wells in the downstream environment. A thorough analysis of geochemical data was also undertaken in the second modelling attempt which led to refinement of the conceptual model.

Calibration targets were met with the aid of an automated calibration sub-routine in the first instance; however, this process was not fully and transparently explained in the model documentation reviewed for this case study. Furthermore, the model sensitivity to boundary conditions and the impact of preferential pathways was not presented clearly in the model documentation as reviewed. During calibration of the updated model, multiple conceptual models were calibrated to head and flow measurements. This calibration process was revisited throughout the modelling process to ensure that the model was calibrated with the project-specific objectives in mind. In all instances of modelling, the dataset available for calibration was extremely limited given the complexity of the site and the sensitivity of the receiving environment. This limited dataset should not be considered exemplary.

Sensitivity analyses were undertaken at nearly every stage of modelling. Sensitivity of the model to operational parameters (grouting), preferential flow paths (faults), and boundary conditions (e.g. sub-glacial recharge) was examined and incorporated into upper bound flow and transport estimates.

Predictions of solute transport carry a significant uncertainty because site-specific transport parameters were not available and the transport model could not be calibrated. The limitations in applying a solute transport model to this complex bedrock environment are not discussed in the modelling report.

The updated groundwater flow and solute transport models were applied to meet the modelling objectives and to provide plausible upper bound estimates on the impact of mine operations in this environment. However, the results are prefaced by the extremely limited dataset on which model conceptualization and calibration were based.

In summary, there were several major points of uncertainty related to the numerical model. Primarily, data was limited for calibrating the model. As well, major assumptions of equivalent porous medium for fractured bedrock, horizontal isotropy of hydraulic conductivity of the various fractured bedrock units and effective porosities did not appear to be justified by data or assessments from the site. These issues introduced significant uncertainties to assessing how the proposed mine will affect changes in groundwater flow and contaminant travel times, and impacts of these changes to local high VEC water users' streams. As recommended in these guidelines (see Figure 2-2), consultation with government technical staff during the development of the modelling objectives, the conceptual model, and the data collection requirements could have helped to identify these issues and questions much earlier in the modelling process.

Finally, it is emphasized that a numerical groundwater model developed at the EA stage for assessing impacts of proposed mines is most useful for calibrating baseline conditions and performing sensitivity analyses related to predictions, as was done here. As mining operations proceed and additional data become available, the model could be verified.