Modelling Tools for the Prediction of Drainage Quality from Mine Wastes

Improving the environmental properties, utilisation potential and long-term prediction of mining wastes Edited by Päivi M. Kauppila and Timo Tarvainen Geological Survey of Finland, Bulletin 408, 27-42, 2018

MODELLING TOOLS FOR THE PREDICTION OF DRAINAGE QUALITY FROM MINE WASTES

Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

Muniruzzaman, M., Karlsson, T. & Kauppila, P. M. 2018. Modelling tools for the prediction of drainage quality from mine wastes. Geological Survey of Finland, Bulletin 408, 27–42, 7 figures and 2 tables.

The weathering of mine wastes often leads to low quality drainage typically characterised by acidic pH and elevated concentrations of dissolved metals/metalloids. Therefore, prior knowledge and quantitative predictions of drainage quality is crucial during mine planning in order to properly assess the environmental impact in the vicinity of mining activity. In recent decades, a great deal of research attention has been paid to accurately predict the mine waste drainage and that has led to the development of a wide variety of predictive models with different levels of sophistication. Despite the availability of a plethora of modelling approaches and well established tools, there is still a lack of attention towards attempting a rigorous predictive modelling at the planning phase (e.g. environmental impact assessment) of a mine. This work presents a relatively simple predictive model that can be used at such early phase of a mine when data is very limited. The model formulation is based on reactive transport approaches that take into account water flow, gas transport and mineral weathering reactions. Furthermore, this paper also includes example case studies (both in waste rock pile and tailings systems) demonstrating the scope and capability of the presented model and how such approaches can be used effectively at potential mine sites.

Keywords: Prediction, mine waste, drainage quality, AMD, predictive model, reactive transport modelling

Geological Survey of Finland, P.O. Box 1237, FI-70211 Kuopio, Finland E-mail: [email protected]

https://doi.org/10.30440/bt408.2 Editorial handling by Timo Tarvainen. Received 28.4.2018; Received in revised form 30.9.2018; Accepted 22.11.2018

27 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

1 INTRODUCTION

Management of mine wastes is a prescient issue in cess based approaches that quantitatively resolve all mining sectors since uncontrolled waste disposal the relevant physical, geochemical, microbiological, may result in liability for the operators with the risk electrochemical, and thermal processes (e.g. Steefel of financial consequences as well as reputational et al. 2015). damage (e.g. Blowes et al. 2014). Of primary concern Depending on the capabilities, typical predictive is the release of low quality drainage from the waste models in mining environmental simulations can deposits that leads to adverse effects on the envi- be categorised as geochemical models that only ronment, ecosystem, and human health (e.g. Blowes take into account the geochemical processes occur- & Jambor 1990, Blowes et al. 2014, Nordstrom et al. ring in the waste piles (e.g. Parkhurst et al. 1985, 2015). Such drainages are known to be the results Davis & Ritchie 1986, Davis & Ashenberg 1989, Ball of the weathering processes of sulphide-rich waste et al. 1987, Blowes & Jambor 1990, Allison et al. deposits under oxic environments and/or under 1991, Wolery et al. 1992, Alpers & Nordstrom 1999, the influence of microbial activities (e.g. Blowes Tempel et al. 2000, Eary et al. 2003, Ramstedt et & Ptacek 1994, Tremblay & Hogan 2000, Amos et al. 2003, Moncur et al. 2005, Gunsinger et al. 2006, al. 2015, Nordstrom et al. 2015). Therefore, it is of Nordstrom & Campbell 2014), and more sophisti- utmost importance to understand the controlling cated reactive transport models that are capable physicochemical processes leading to toxic drain- of simultaneously capturing hydrogeological pro- age in mining environments to sufficiently predict cesses, multicomponent solute and gas transport, the overall system behaviour in advance (e.g. Dold thermal processes, microbiological and electro- 2017). chemical mechanisms in addition to the geochemi- During mine planning, the estimates of the cal processes (e.g. Pruess 1991, Pantelis 1993, Steefel drainage quality are required to properly assess & Lasaga 1994, Wunderly et al. 1996, Bethke 1997, the environmental influences for the environmen- Lefebvre et al. 2001, Mayer et al. 2002, Saaltink et tal impact assessment and for the environmental al. 2002, Prommer et al. 2003, Parkhurst et al. 2005, permit application to facilitate mine planning and da Silva et al. 2009, Šimunek et al. 2012, Parkhurst to prevent negative impacts on the watersheds. & Appelo 2013, Muniruzzaman et al. 2014, Amos et The prediction of effluent quality is, neverthe- al. 2015, Lichtner et al. 2015, Muniruzzaman & Rolle less, a challenging task. This is mainly because 2016, Nordstrom & Nicholson 2017, Pedretti et al. the mineral weathering reactions responsible for 2017, Rolle et al. 2018). the mine drainage are complex and long term (e.g. Despite the diversified supply of prediction meth- Blowes & Jambor 1990, Blowes & Ptacek 1994). In ods, limited publications exist on how to approach addition, they are site-specific and depend on the the modelling in a mine planning phase for which geology and climatic conditions of each mine site, data on the mine wastes is still limited. Due to these even though the overall chemical processes are the challenges, one of the aims of the KaiHaMe project same (cf. Plumlee 1999). (Management of mining wastes) was to provide In recent decades, a wide range of prediction additional tools for predictive modelling. As a first techniques have been developed including experi- step, a review of the existing prediction methods, mental methods focusing on laboratory and field including typical laboratory and field tests, as well scale tests to characterise different properties of as numerical modelling in particular, was car- waste materials (e.g. Morin & Hutt 1994, Price ried out within this project (Muniruzzaman et al. 2009, Tripathy 2014, Parbhakar-Fox & Lottermoser 2018b). In addition to the methods, the review cov- 2015, Dold 2017), as well as numerical approaches ered aspects such as relevant processes resulting to quantitatively describe and predict the system and occurring in mine drainage, available codes for dynamics by capturing all the key processes (e.g. numerical modelling, and code and model uncer- Mayer et al. 2003, Maest et al. 2005, Amos et al. tainties, and limitations and applicability under 2015). Numerical modelling is instrumental in Nordic climate. Muniruzzaman et al. (2018b) also quantifying the overall system behaviour especially discussed the potential approaches to enhance the where coupling between multi-scale processes leads prediction accuracy by using integrated methodolo- to non-intuitive system dynamics (e.g. Steefel et gies to properly describe the multifaceted processes al. 2005). Such modelling frameworks rely on pro- occurring in mine wastes.

28 Geological Survey of Finland, Bulletin 408 Modelling tools for the prediction of drainage quality from mine wastes

As a next step, this investigation focuses on the lar, the model formulation, quality of the site-spe- predictive modelling of drainage water quality from cific information, and overall modelling workflow mine waste facilities (both waste rock piles and tail- were treated from the perspective of predictions ings) by means of reactive transport modelling. The in future waste facilities. The ultimate intent of study presents examples of predictive simulations these simulations is to demonstrate the presented at three particular mine sites in Finland illustrat- model capabilities and how this tool can be used ing the specific capabilities and scope of predictive in the planning phase of a mine. The simulation modelling. These simulation examples generally outcomes suggest that numerical tools in combina- demonstrate how reactive transport modelling can tion with good quality data have a great potential be effectively used in predicting the seepage water to interpret the timing and occurrence of the future compositions from mine waste settings under dif- low quality drainages that may be harmful to the ferent conditions (e.g. environmental conditions or surrounding receptors. In the following sections, a closure scenarios). Although the presented exam- brief summary of the results of the predictive mod- ples include predictive modelling in existing sites, elling is provided and the details of the study are they can be considered representative of potential presented in a separate GTK Open File Work Report future sites where data are very limited. In particu- by Muniruzzaman et al. (2018a).

2 STUDY SITES AND DATA COLLECTION

Mine waste and drainage water samples were col- and from the Särkiniemi Ni mine in Leppävirta (in lected from three mine sites around Finland for operation during 2007–2009), and tailings samples the development of predictive models. Waste rock from the Pyhäsalmi Cu-Zn mine in Pyhäjärvi (oper- samples were collected from the Kylylahti Cu-Co- ated since 1962) (Fig. 1). All the sites were metal Zn-Ni-Au mine in Polvijärvi (operated since 2012) sulphide mines.

29 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

Fig. 1. Location of the study areas and sampling points. The sulphidic waste rock pile of Kylylahti has already been backfilled to the underground mine. The investigated tailings pond at the Pyhäsalmi mine contains 3.0 (situation in 2003) – 7.5 (estimated maximum storage capacity) Mt of highly sulphidic waste material (Pohjois-Suomen ympäristölupavirasto 2007). The sulphidic waste rock pile of Särkiniemi contains around 13 000 t of waste rocks (Tornivaara et al. 2018). Basemaps © National Land Survey of Finland and HALTIK 2013.

30 Geological Survey of Finland, Bulletin 408 Modelling tools for the prediction of drainage quality from mine wastes

3 MODELLING APPROACH

Predictive analyses were performed using the avail- time integration method. The resulting system of able site-specific information at the three differ- equations describing the aqueous and gas phase ent mine sites. The model formulation incorporates transport was solved with the direct matrix solver water flow in partially saturated domain, along with UMFPACK (Davis and Duff 1997). Afterwards, the multicomponent solute and gas transport, and geo- transported concentrations at each location of chemical reactions. The flow, transport, and reac- the discretised domain are updated and passed to tion problem was solved with finite volume method reaction step (in PHREEQC) for the calculations of by employing an operator-splitting approach, in gas-water partitioning and chemical reactions. The which the reaction problem was solved within solubility of gases in the liquid phase was calculated PHREEQC (Parkhurst & Appelo 2013). An upwind within PHREEQC, which uses either Henry’s law or differentiation scheme was used for the spatial dis- Peng-Robinson EOS for ideal and non-ideal gases cretisation of the flow and transport equations. For respectively (Parkhurst & Appelo 2013). The entire the integration in time the explicit Euler method formulation of the presented modelling approach is was used in the computation of advective fluxes, implemented in MATLAB®. The fundamental steps whereas the diffusion/dispersion problem in aque- of the modelling approach used in this study are ous/gas phase was solved by using the implicit Euler briefly explained in Figure 2.

Fig. 2. Schematic diagram of the structure and calculation steps of the modelling approach. t = t + dt (time = time + time step size) refers to the update of temporal steps in a time loop.

4 RESULTS AND DISCUSSION

This section briefly presents different aspects of closure scenarios of waste facilities. The presented predictive modelling starting from the conceptu- examples strictly focus on the demonstration of dif- alisation of study sites followed by the demonstra- ferent features of the model rather than accurately tion of different capabilities of the presented model. reproducing real conditions in any particular site. In particular, the following sections focus on how The specific model outcomes should also be treated such modelling may help in predicting the drainage as what kind of insights such predictive modelling compositions and mineralogical evolution, estimat- can offer, especially in the planning phase of a mine. ing long-term behaviour, and analysing different

4.1 Conceptual model

The predictive simulations were performed along a assemblages throughout the domain (Fig. 3). The 1-D vertical simulation domain with homogeneous conceptual models were built based on the measure- distributions of hydraulic properties and mineral ments of mineralogy, drainage water, and from the

31 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila insights of laboratory static tests on waste materi- However, to provide an accurate prediction site- als. The system was conceptualised as a three phase specific kinetic parameters might be required and porous media with water, air, and solid matrix being such parameters can be measured based on labora- the main phases. Although in real settings, a vari- tory batch experiments (e.g. Williamson & Rimstidt able water saturation is likely to occur and influence 1994) or humidity cell tests (e.g. Maest & Nordstrom of capillary fringe might be important, for simplic- 2017). Geochemical reactions were simulated using ity the presented simulation examples considered a customised database, in which PHREEQC data- a constant water saturation along the depth. The base was extended for additional reactions from kinetics of the mineral dissolution-precipitation WATEQ4F, llnl, CrunchFlow, and Sit databases. reactions were modelled using the literature values.

Fig. 3. Conceptual model used in the predictive simulations in the waste piles and the general conceptualisation approaches of physicochemical system at different porous compartment (Q in, Q out represent the general source/ sink terms in aqueous and gas phase, and q is the recharge rate).

4.2 Prediction of drainage compositions and evolution of mineralogical assemblages

Figure 4 shows the evolution of the predicted ing based on the static tests. A constant recharge drainage chemistry over time at the outlet of the rate of 300 mm/y, which is ~50% of the mean Särkiniemi waste rock pile. The waste rock pile at annual precipitation in that region, was applied this site has approximately a height of 10 m and the at the top boundary of the domain. The recharge waste rocks are composed of biotite, plagioclase, water contained a generic rainwater composition quartz, hornblende, and sulphides (mainly pyrrho- (Reimann et al. 1997) and it was also assumed to tite with minor pyrite, chalcopyrite and pentland- be in equilibrium with atmospheric O2 and CO2. ite) (Table 1). Only minor carbonates were present The gas transport was also simulated applying a and the waste rock was classified as acid produc- constant boundary condition at the top and only

32 Geological Survey of Finland, Bulletin 408 Modelling tools for the prediction of drainage quality from mine wastes diffusive transport (gas phase diffusion coefficient, elevated concentrations of metals, particularly Ni -5 2 Dg=1.75×10 m /s) was considered. The drainage (cf. also Karlsson et al. 2018). from the waste rock was acidic and it contained

Table 1. Initial mineral contents, reactive surface areas, and reaction rate coefficients.

Mineral Mineral Content Surface Area, Rate Reference A coefficient, k -1 a 2 -1 -2 -1 [wt%] [mol Lw ] [m Lw ] [mol m s ] Biotite 33.56 2.12 0.65 10-10.97 Nagy (1995) Hornblende 6.07 0.58 0.05 10-8.10 Palandri & Kharaka (2004) Serpentine 3.52 0.33 0.05 10-9.08 Declercq & Oelkers (2014) Albite 1.45 0.15 0.60 10-10.16 Palandri & Kharaka (2004) Chlorite 1.35 0.06 0.50 10-11.11 Palandri & Kharaka (2004) Pyrrhotite 1.52 0.50 0.30 b 10-8.19 Williamson & Rimstidt (1994) Anthophyllite 0.45 1.53×10-2 0.50 10-11.94 Palandri & Kharaka (2004)

Pentlandite 0.03 1.03×10-3 0.70 b 10-8.19 Williamson & Rimstidt (1994) Pyrite 0.02 4.42×10-3 0.30b 10-8.19 Williamson & Rimstidt (1994)

c -10.5 SiO2(a) - - - 10 Rimstidt & Barnes (1980) a -1 Moles of mineral per L of pore water, calculated from wt% by using a solid density, ρs = 2.65 [kg L ] and an average porosity, θ = 0.50

b 2 -1 -1 Surface area per moles of minerals per liters of pore water [m mol Lw ] c Both kinetic parameters and equilibrium constant was slightly adjusted to be in the consistent drainage concentration range as the measured values

Besides primary minerals, the model for the colour for each species in Figure 4. Please note that Särkiniemi waste rock pile also considered ferri- unlike the simulated profiles, dotted lines represent hydrite, gypsum, jarosite, gibbsite, and amorphous only a single drainage measurement (not the evo- silica as secondary minerals. Based on the model, lution with time), which was performed approxi- the pH value drops quite fast in the drainage water mately after 8 years of the construction of the waste after the disposal of the waste rock started, due rock pile. Although a few of these species (e.g. Fe, to the oxidation of pyrrhotite, minor fractions of Na) show slight discrepancy between the simulated pyrite, and pentlandite and because of the absence and measured concentrations, the predicted values of sufficient amount of carbonate minerals to effec- are evidently in the similar order of magnitude for tively buffer the pH (Fig. 4a). These oxidation pro- all the dissolved species as well the overall conduct- cesses also lead to elevated concentrations of the ance of the effluent. This suggests that the model dissolved ions in the drainage during the modelled is able to reasonably predict the drainage behaviour period: SO4, Mg, Fe, Ni, Al, Si, Ca, K, Na (Fig. 4a-c). even though a simple conceptual model was used This front of the increased concentrations of dis- for the waste rock pile system. The simulated value solved species is directly correlated to the drop in of pH at the late time plateau in Figure 4a is also pH front (Fig. 4a) as well the increase in electrical consistent with the NAG (net acid generation) pH conductivity values (Fig. 4d). The simulated con- (2.7-2.9) value obtained from the NAG test per- centration of each species also seem to be consistent formed with the waste rock samples. The measured with the measured drainage water composition rep- data through time can be used for updating such a resented by the dotted lines with the corresponding pre-mine model once the actual waste rock pile is

33 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

Fig. 4. Simulated (solid lines) and measured (dotted lines) compositions of the drainage water from the Särkiniemi waste rock pile as a function of time: (a) pH (primary y-axis), SO4, Mg, Ca (secondary y-axis), (b) Fe, Al, Si, (c) Na, K, Ni, and (d) electrical conductivity of drainage water. Please note that the dotted lines represent only a single measurement point (and not a series of collected data vs. time) that was performed at ~8 years.

constructed. Model refinement at this stage is very This mineral dissolution is directly related to likely but the precision of a pre-mine model within the rise in the concentration of other ions (Al, Si, an order of magnitude should be good enough for Ca, Mg, Na, and K) as presented in Figure 4. It is initial mine planning. apparent from the figure that the silicate minerals Capability of the model to simulate weathering (biotite, serpentine, chlorite or hornblende) dissolve and precipitation of key minerals in the waste rock much slower compared to typically fast dissolving pile is demonstrated in Figure 5, which illustrates minerals (such as carbonates) as only a very small the simulated profiles of the mineral contents in the fractions of these minerals were removed from the domain at different times (t = 0–10 years) at the 10 simulation domain after 10 years (Fig. 5, middle m high Särkiniemi waste rock pile. The first col- rows). The slow dissolution rate of these silicates umn in this figure (Fig. 5a,e,i) shows the dissolution suggests that these particular minerals may con- fronts of the different sulphide minerals represent- tribute to the acid buffering in the long term (e.g. ing a sequential depletion of these minerals over Jambor et al. 2002). The last column of the figure time. The acidic conditions induced by the sulphide depicts the precipitation of secondary minerals in oxidations lead to the potential dissolution of the the system (Fig. 5d,h,l). The model results reveal silicates in the domain (second and third columns, that a significant amount of Fe(III) and silica spe-

Fig. 5b-c,f-g,j-k). cies precipitate as ferrihydrite and SiO2(a) phases,

34 Geological Survey of Finland, Bulletin 408 Modelling tools for the prediction of drainage quality from mine wastes

Fig. 5. Mineral contents versus depth and their temporal evolution at the Särkiniemi waste rock pile.

respectively (Fig. 5d,h). The concentrations of pro- vations of the evolution of mineral weathering in tons and Fe are also limited by the precipitation of mine wastes (e.g. Blowes & Jambor 1990, Blowes et the ferrihydrite phase. In contrast, the amount of al. 1995) and thus indicate that the developed model precipitated sulphate phases is negligible (Fig. 5l). seems to be able to capture the mineral weathering The results are in line with the overall field obser- processes in this waste rock pile.

4.3 Sensitivity of the model input parameters

A sensitivity analysis can be performed by varying predictions. The simulations in Figure 6 are per- different model input parameters and looking at the formed by changing the values of input parameters effect on the model outcome. Figure 6 demonstrates (Table 2) with respect to a “base case” scenario sensitivities of different parameters from the reac- (blue lines). This base case scenario was selected tive transport simulations. This analysis enhances based on the available information at Kylylahti the system understanding as well as identifying waste rock pile, which is characterised by a rather possible improvements in future data collection by high sulphide content (>9%), some fractions of focusing on the most sensitive parameters. This carbonates (~2%), and significant amount of sili- approach may help increasing the accuracy of the cates (~25%). The waste rock pile has a height of

35 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

20 m and a constant recharge rate (~300 mm/y) was oxidation rate (orange dash-dot line) leads to a sig- applied as the boundary condition. Gas transport nificantly earlier breakthrough (and vice versa for was considered to occur only by diffusive mecha- lower oxidation rate) of the low pH front compared nisms and a constant concentration boundary, rep- to the base case. Impact of the oxidation rate is also resenting atmospheric conditions, was employed at apparent both at early (a1) and late (a2) times. In the top of the pile. For more detailed description of contrast, the recharge rate does not show a signifi- all these cases, interested readers are referred to cant impact on the timing of the acid rock drainage Muniruzzaman et al. (2018a). for this scenario (Fig. 6b1). In fact the application Figure 6a1-a2 shows the impact of the sul- of a higher (dash-dot orange line) or lower (dotted phide oxidation rate on drainage pH profiles at the yellow line) water flux through the domain does not Kylylahti waste rock pile. It is evident that a higher lead to significantly different arrival times of the

Fig. 6. Impact of the model input parameters on the drainage water quality of the Kylylahti waste rock pile: effects of oxidation rate (a1, a2), recharge rate (b1, b2), gas diffusion (c1, c2), and domain size (d1, d2). The left panel shows smaller time scale behaviour whereas the right panel presents the profiles in longer time scale.

36 Geological Survey of Finland, Bulletin 408 Modelling tools for the prediction of drainage quality from mine wastes

Table 2. Initial parameters used in the sensitivity analysis.

Parameter Base Case Additional Case 1 Additional Case 2 Oxidation rate (reactive surface area [m2/mol/L]) 1.43 0.143 14.3 Recharge rate [mm/y] 300 30 3000 Pore gas diffusion coefficient2 [m /s] 5.63×10-6 5.63×10-7 5.63×10-8 Domain length [m] 20 5 10

low pH fronts compared to the base case (blue line). slow diffusion case (orange lines) compared to the Such phenomena indicates that the early behaviour base case (blue lines, Fig. 6c1-c2). The acid mine of the drainage chemistry of this waste rock pile is drainage occurs significantly later (at ~60 years) effectively controlled by the geochemistry rather for the case of 100-fold smaller diffusion coefficient than the water flow velocity. However, at late times, (dotted yellow lines, Fig. 6c1). Similar pattern is the impact of the recharge rate is significant (b2). also observed at late times, when slower diffusive Figure 6c1-c2 shows the impact of gas transport case leads to a later rise in pH (c2). The bottom row into the Kylylahti waste rock pile. The model pre- (Fig. 6d1-d2) shows the effects of the domain size dicts that the gas diffusion influences the timing of on the drainage chemistry. The simulations show the acidic drainage (pH breakthrough) at the end of that the total height of the 1-D domain is perhaps the waste rock pile. By applying a 10-fold smaller the least sensitive parameter among these four diffusion rate, the temporal pH profiles at the outlet parameters. Both at early (d1) and late (d2) times, of the domain are very close to the base case with the profiles for all scenarios are quite close. only a slight delay (by only a few years) for the

4.4 Long-term predictive analysis and closure scenarios

Predictive reactive transport simulations are valu- this simulation is merely to demonstrate the model able for assessing mine closure scenarios. Figure 7 capabilities and how this model can be used upon depicts the predicted drainage quality from the top cover applications. However, accurate reproduction surface (i.e. top 2 m layer) of tailings impoundment of the current conditions at the Pyhäsalmi mine site at the Pyhäsalmi mine site for 1000 years (light is beyond the scope of this simulation. green lines) with and without a cover on top of the Due to the reduced penetration rate of the waste facility. These simulation scenarios consider recharge water and lower supply of oxygen in the the beginning of waste disposal (i.e. t = 0 year) as tailings, the drainage pH stays higher for a longer initial condition. In the simulations, the cover con- period (t = ~0-80 years) at the beginning of the ditions were mimicked by employing a lower (10- profile under the cover condition compared to the fold) recharge rate (q = 30 mm/y) as well as gas base case (Fig. 7a and figure inset). Afterwards, the -7 2 -1 diffusion coefficient D( p = 4.38×10 m s ) compared drainage pH stays around 6 for almost another two to that of uncovered base case. centuries (dark red line, Fig. 7a) before it drops to The model predicts that it would take around four acidic conditions (pH < 2). This early time buffering centuries (as indicated in the low pH conditions and for a couple of 100 years is probably due to a com- elevated sulphate concentrations) for the sulphide bined effect of the limited supply of the reactants minerals to be completely depleted from this 2 m (i.e. water and oxygen) into the system for the sul- domain under the “semi-open conditions” as used phide oxidation reactions as well as the equilibrium in the uncovered base case simulation (Fig. 7a). of the carbonate minerals. The simulations also The model prediction is quite different for predict that even though the acidic drainage will not a covered tailings facility. This can be seen as a occur before ~400 years under closure conditions, hypothetical scenario that attempts to analyse the the net duration of the acid mine drainage will be effects of a cover on a waste facility compared to the comparatively longer for this condition relative to analogous uncovered case. The ultimate objective of the base case (Fig. 7a).

37 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

Fig. 7. Prediction of the long-term behaviour of drainage water quality from the tailings impoundment at the Pyhäsalmi site by considering the application of a potential cover on the tailings facility.

4.5 Uncertainties in the predictions

Depending on the unknowns and level of under- • Interactions between the pore water species and standings about the key processes, modelling mineral surfaces due to surface complexation or results may be subject to a considerable extent of ion exchange processes uncertainties. The major sources of uncertainties in • Insufficient information about the seasonal vari- this study can be briefly summarised as: ations and the key effects on reactive transport • Heterogeneity associated with the physical and processes chemical processes and the related parameters • Insufficient information on solute fluxes into • Missing data on water saturation and moisture multiple dimensions content profiles • Inadequate information regarding the exact dis- • Unknowns related to the presence of preferential tributions of gas contents in the waste systems flows and fractures • The presented results did not consider heat • Kinetic reaction rate coefficients under the inves- transport processes which may have some influ- tigated settings ence on the overall outcome • Influence of microbial activities on the overall geochemistry

38 Geological Survey of Finland, Bulletin 408 Modelling tools for the prediction of drainage quality from mine wastes

These sources of uncertainties occur mainly better closure scenarios. While predictive results can because these models simulate future sites where be considered an “order of magnitude” approach, waste disposal has not been started yet (i.e. planning these results still provide information on potential phase). Rigorous modelling of such future settings outcomes with various mine planning and closure can be uncertain due to inadequate site-specific scenarios. In addition, the sensitivity analyses dis- information necessary for a realistic conceptuali- cussed in section 4.3 allow for an understanding of sation. However, predictive modelling during the the most important model parameters. Thus, addi- mine planning can provide valuable information on tional efforts can focus on these specific parameters potential environmental impacts. Understanding during the modelling and mine planning process, these impacts, even with some uncertainty, can help as needed. avoid or reduce environmental impacts by choosing

5 CONCLUSIONS

Numerical modelling is an instrumental tool not on the results of the current study and the litera- only for effectively predicting the effluent quality ture review (Muniruzzaman et al. 2018b), the key from mine waste dumps but also to perform more points and limitations of the procedures are briefly robust environmental impact assessment. This summarised as: work presents a relatively simple reactive transport • Successful demonstration of the presented mod- model that can be used in predictions of drainage elling approach involving operator-splitting quality from potential future waste facilities where scheme site-specific data is scarce. In order to demonstrate • Replication of conceptual models the model capabilities, modelling examples were • Simulation results were tested at three different presented at three different study sites based on mine sites the available information. Although the exercises • Model can be easily used to test alternative waste were performed at existing sites, they are repre- designs (e.g. cover, buffer addition) sentative scenarios involving future waste facilities • Identification of the most sensitive parameters and workflow in modelling was followed from the which help managing future data collection perception of these representing future sites. Based

6 RECOMMENDATIONS

Guidelines for the improvement of prediction qual- spatial variability) and to reasonably character- ity at the planning phase of a mine are briefly sum- ise the physical and chemical heterogeneity of the marised as: waste facility. Model sensitivity can be a useful Detailed site-specific data should be collected tool to identify the most important data. Analogues that are representative of the study site to develop from previous sites could be used for proposed sites a realistic conceptual model. The extent of the col- where these data are not yet available. lected information should be sufficient to identify Accurately capturing the scale dependence of at least the key mechanisms for that site. reactive transport processes is arguably the most Kinetic parameters, which are crucial in the over- challenging aspect in numerical simulations (e.g. all prediction scheme, should be experimentally Steefel et al. 2005), since most reactive transport measured by means of laboratory or field scale test- parameters are measured in the laboratory. When ing (e.g. humidity cell tests, column/flow-through using this scaling approach, the scaling factors tests). should be treated as site-specific and they should More efforts, budget, and systematic approaches be experimentally determined applying systematic should be dedicated to detailed data collection (e.g. investigations. hydrogeology and geochemistry along with their

39 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

For predictive modelling in a planned waste and optimisation of the required monitoring/data facility (early stage of a mine), all the detailed collection scheme, the waste facility itself or the data regarding the “proposed” waste pile may not water treatment requirements to provide a better be available at hand. However, in such circum- management of the wastes. stances, a systematic monitoring scheme should In addition to the information only at the drain- be established allowing more detailed data collec- age water, spatial profiling of different quantities tion through time in order to continuously update at the waste pile systems should also be considered the conceptual model as well as to better con- in the future data collection scheme. strain the numerical model with more site-specific Besides the characterisation of the geochem- information. istry, emphasis should also be given on the fluid In addition to the prediction of effluent quality, flow and transport processes that are sometimes the predictive simulations incorporating reactive relatively overlooked in dealing with mine waste transport models can also be used in the design management.

ACKNOWLEDGEMENTS

The authors would like to thank Raymond Johnson which helped improving the quality of this manu- (U.S. Department of Energy) and Andrew Barnes script. The author also acknowledge the funding (Geochemic Ltd) for reviewing this manuscript from the European Regional Development Fund and their constructive comments and suggestions, (ERDF).

REFERENCES

Allison, J. D., Brown, D. S. & Novo-Gradac, K. J. 1991. poundment, Timmins Area, Ontario. Geomicrobiology MINTEQA2/PRODEFA2, A Geochemical Assessment Journal 13, 13–31. Model for Environmental Systems – Version 3.0 Us- Blowes, D. W., Ptacek, C. J., Jambor, J .L., Weisener, C. er’s Manual. U.S. Environmental Protection Agency G., Paktunc, D., Gould, W. D. & Johnson, D. B. 2014. Report EPA/600/3-91/021. 106 p. The Geochemistry of Acid Mine Drainage. In: Holland, Alpers, C. N. & Nordstrom, D. K. 1999. Geochemical mod- H. D. & Turekian, K. K. (eds) Treatise on Geochem- elling of water–rock interactions in mining environ- istry (Second Edition), Volume 11. Oxford: Elsevier, ments. In: Plumlee, G. S. & Logsdon, M. J. (eds) The 131−190. Available at: https://doi.org/10.1016/B978-0- Environmental Geochemistry of Mineral Deposits, 08-095975-7.00905-0 Reviews in Economic Geology, vol. 6A. Littleton, CO: da Silva, J. C., Vargas, Jr., E. D. A. & Sracek, O. 2009. Mod- Society of Economic Geologists Inc., 289–323. elling multiphase reactive transport in a waste rock Amos, R. T., Blowes, D. W., Bailey, B. L., Sego, D. C., Les- pile with convective oxygen supply. Vadose Zone Jour- lie, S. & Ritchie, I. A. M. 2015. Waste-rock hydroge- nal 8, 1038–1050. ology and geochemistry. Applied Geochemistry 57, Davis, A. & Ashenberg, D. 1989. The aqueous geochemis- 140–156. try of the Berkeley Pit, Butte, Montana, USA. Applied Ball, J. W., Nordstrom, D. K. & Zachman, D. W. 1987. WA- Geochemistry 4, 23–36. TEQ4F – A Personal Computer FORTRAN Translation Davis, G. B. & Ritchie, A. I. M. 1986. A model of oxida- of the Geochemical Model WATEQ2 with Revised Data tion in pyritic mine waste: Part 1 Equations and ap- Base. U.S. Geological Survey, Open-File Report 87-50. proximate solution. Applied Mathematical Modelling 108 p. 10, 314–322. Bethke, C. 1997. Modelling transport in reacting geo- Davis, T. & Duff, I. 1997. An unsymmetric-pattern mul- chemical systems. Comptes Rendus de l’Académie des tifrontal method for sparse LU factorization. SIAM Sciences 324, 513–528. Journal on Matrix Analysis and Applications 18, 140– Blowes, D. W. & Jambor, J. L. 1990. The pore-water geo- 158. Available at: doi:10.1137/S0895479894246905 chemistry and the mineralogy of the vadose zone of Declercq, J. & Oelkers, E. H. 2014. CarbFix Report 4 sulfide tailings, Waite Amulet, Quebec, Canada. Ap- PHREEQC mineral dissolution kinetics database. plied Geochemistry 5, 327–346. CarbFix Project no. 281348. Geoscience Environement Blowes, D. W. & Ptacek, C. J. 1994. Acid-neutralization Toulouse, Belin, Toulouse. Available at: https://www. Mechanisms in Inactive Mine Tailings. In: Jambor, J. or.is/sites/or.is/files/kinetic_database.pdf L. & Blowes, D. W. (eds) The Environmental Geochem- Dold, B. 2017. Acid rock drainage prediction: A critical re- istry of Sulfide Mine Wastes. Mineralogical Associa- view. Journal of Geochemical Exploration 172, 120–132. tion of Canada, Short Course Series, Vol. 31, 95–116. Eary, L. E., Runnells, D. D. & Esposito, K. J. 2003. Geo- Blowes, D. W., Al, T., Lortie, L., Gould, W. D. & Jambor, J. chemical controls on ground water composition at the L. 1995. Microbiological, Chemical, and Mineralogical Cripple Creek Mining District, Cripple Creek, Colora- Characterization of the Kidd Creek Mine Tailings Im- do. Applied Geochemistry 18, 1–24.

40 Geological Survey of Finland, Bulletin 408 Modelling tools for the prediction of drainage quality from mine wastes

Gunsinger, M. R., Ptacek, C. J., Blowes, D. W., Jambor, media: Experiments and model-based interpretation. J. L. & Moncur, M. C. 2006. Mechanisms controlling Geochimica et Cosmochimica Acta 141, 656–669. acid neutralization and metal mobility within a Ni- Muniruzzaman, M., Karlsson, T. & Kauppila, P. M. rich tailings impoundment. Applied Geochemistry 21, 2018a. Prediction of the Drainage Water Quality from 1301–1321. Mine Wastes with Reactive Transport Modelling. Jambor, J. L., Dutrizac, J. E., Groat, L. A. & Raudsepp, Geological Survey of Finland, Open File Work Report M. 2002. Static tests of neutralization potentials of 12/2018. Available at: http://tupa.gtk.fi/raportti/ silicate and aluminosilicate minerals. Environmental arkisto/12_2018.pdf Geology 43, 1–17. Muniruzzaman, M., Kauppila, P. M. & Karlsson, T. 2018b. Karlsson, T., Kauppila, P. M., Lehtonen, M., Tiljander, Water quality prediction of mining waste facilities M., Forsman, P. & Lahtinen, T. 2018. Laboratorio- based on predictive models. Geological Survey of Fin- analyysit kaivannaisjätteiden käyttäytymisen ennus- land, Open File Research Report 16/2018. 65 p. Availa- tamisessa. Geological Survey of Finland, Open File ble at: http://tupa.gtk.fi/raportti/arkisto/16_2018.pdf Work Report 13/2018. (in Finnish) Nagy, K. L. 1995. Dissolution and precipitation kinetics of Lefebvre, R., Hockley, D., Smolensky, J. & Lamontagne, sheet silicates, in Chemical Weathering Rates of Sili- A. 2001. Multiphase transfer processes in waste rock cate Minerals, Reviews in Mineralogy 31, 173– 233. piles producing acid mine drainage 2: Applications Nordstrom, D. K. & Campbell, K. M. 2014. Modelling of numerical simulation. Journal of Contaminant Low-Temperature Geochemical Processes. In: Hol- Hydrology 52, 165–186. land, H. D. & Turekian, K. K. (eds) Treatise on Geo- Lichtner, P. C., Hammond, G. E., Lu, C., Karra, S., Bisht, chemistry, Second Edition, Vol. 7, 27–68. G., Andre, B., Mills, R. T. & Kumar, J. 2015. PFLOTRAN Nordstrom, D. K. & Nicholson, A. (eds) 2017. Geochemi- User manual: A Massively Parallel Reactive Flow and cal Modeling for Mine Site Characterization and Re- Transport Model for Describing Surface and Sub- mediation. Colorado: Society for Mining, Metallurgy & surface Processes. Los Alamos National Laboratory, Exploration. 181 p. United States. LA-UR-15-20403. 195 p. Nordstrom, D. K. Blowes, D. W. & Ptacek, C. J. 2015. Hy- Maest, A. & Nordstrom, K. 2017. A geochemical examina- drogeochemistry and microbiology of mine drainage: tion of humidity cell tests. Applied Geochemistry 81, An update. Applied Geochemistry 57, 3–16. Available 109–131. at: https://doi.org/10.1016/j.apgeochem.2015.02.008 Maest, A. S., Kuipers, J. R., Travers, C. L. & Atkins, D. Palandri, J. L. & Kharaka, Y. K. 2004. A compilation of A. 2005. Predicting Water Quality at Hardrock Mines: rate parameters of water-mineral interaction kinetics Methods and Models, Uncertainties, and State-of- for application to geochemical modeling: USGS Open the-Art. Kuipers & Associates and Buka Environ- File Report 2004-1068, Menlo Park, California, Na- mental. 77 p. Available at: http://www.ceaa-acee. tional Energy Technology Laboratory – United States gc.ca/050/documents_staticpost/cearref_3394/hear- Department of Energy. ings/SM09.pdf Pantelis, G. 1993. FIDHELM: Description of Model and Mayer, K. U., Blowes, D. W. & Frind, E. O. 2003. Advances Users Guide. Australian Nuclear Science and Technol- in reactive-transport modelling of contaminant re- ogy Organization Report, ANSTO/M123. lease and attenuation from mine-waste deposits. In: Parbhakar-Fox, A. & Lottermoser, B. G. 2015. A critical Jambor, J. L., Blowes, D. W. & Ritchie, A. I. M. (eds) review of acid rock drainage prediction methods and Environmental aspects of mine wastes. Ottawa: Min- practices. Minerals Engineering 82, 197–124. eralogical Association of Canada, Short Course Series Parkhurst, D. L. & Appelo, C. A. J. 2013. Description of 31, 283–302. input and examples for PHREEQC version 3 – a com- Mayer, K. U., Frind, E. O. & Blowes, D. W. 2002. Mul- puter program for speciation, batch-reaction, one ticomponent reactive transport modelling in variably dimensional transport, and inverse geochemical cal- saturated porous media using a generalized formu- culations. U.S. Geological Survey, Techniques and lation for kinetically controlled reactions. Water Re- Methods 6-A43, p. 497. Available at: http://pubs.usgs. sources Research 38, 1174–1195. gov/tm/06/a43/. (Accessed 5th June 2017) Moncur, M. C., Jambor, J. L., Ptacek, C. J. & Blowes, D. W. Parkhurst, D. L., Kipp, K. L., Engesgaard, P. & Charl- 2009. Mine drainage from the weathering of sulfide ton, S. R. 2005. PHAST - A program for simulating minerals and magnetite. Applied Geochemistry 24, ground-water flow, solute transport, and multicom- 2362−2373. ponent geochemical reactions. U.S. Geological Survey, Moncur, M. C., Ptacek, C. J., Blowes, D. W. & Jambor, J. Techniques and Methods 6–A35. 235 p. L. 2005. Release, transport and attenuation of metals Parkhurst, D. L., Thorstenson, D. C. & Plummer, L. N. from an old tailings impoundment. Applied Geochem- 1985. PHREEQE – A computer program for geochemi- istry 20, 639–659. cal calculations. U.S. Geological Survey, Water Re- Morin, K. A. & Hutt, N. M. 1994. An empirical technique sources Investigations Report 80–96. 195 p. for predicting the chemistry of water seeping from Pedretti, D., Mayer, K. U. & Beckie, R. D. 2017. Stochas- mine-rock piles. In: Proceedings, America Society of tic multicomponent reactive transport analysis of Mining and Reclamation, 1993, 12–19. Available at: low quality drainage release from waste rock piles: http://www.asmr.us/Portals/0/Documents/Confer- Controls of the spatial distribution of acid generat- ence-Proceedings/1994-Volume-1/0012-Morin.pdf. ing and neutralizing minerals. Journal of Contami- (Accessed 4th July 2017) nant Hydrology 201, 30–38. Available at: https://doi. Muniruzzaman, M. & Rolle, M. 2016. Modelling mul- org/10.1016/j.jconhyd.2017.04.004 ticomponent ionic transport in groundwater with Pohjois-Suomen ympäristölupavirasto 2007. Pyhäsal- IPhreeqc coupling: Electrostatic interactions and men kaivoksen ympäristö- ja vesitalouslupa, Pyhä- geochemical reactions in homogeneous and hetero- järvi. Lupapäätös 85/07/02, Dnro Psy-2004-y-199, geneous domains. Advances in Water Resources 98, 70 p. (in Finnish) 1–15. Plumlee, G. S. 1997. The Environmental Geology of Min- Muniruzzaman, M., Haberer, C. M., Grathwohl, P. & eral Deposits. In: Plumlee, G. S. & Logsdon, M. J. (eds) Rolle, M. 2014. Multicomponent ionic dispersion dur- The Environmental Geochemistry of Mineral Deposits: ing transport of electrolytes in heterogeneous porous

41 Geological Survey of Finland, Bulletin 408 Muhammad Muniruzzaman, Teemu Karlsson and Päivi M. Kauppila

Part A: Processes, Techniques, and Health Issues. Steefel, C. I. & Lasaga, A. C. 1994. A coupled model for Reviews in Economic Geology, Vol. 6A, 71–116 transport of multiple chemical species and kinetic Price, W. A. 2009. Prediction Manual for Drainage Chem- precipitation dissolution reactions with application to istry from Sulfidic Geologic Materials. Natural Re- reactive flow in single-phase hydrothermal systems. sources Canada. MEND Report 1.20.1. 579 p. American Journal of Science 294, 529–592. Prommer, H., Barry, D. A. & Zheng, C. 2003. MODFLOW/ Steefel, C. I., Appelo, C. A. J., Arora, B., Jacques, D., MT3DMS-based reactive multicomponent transport Kalbacher, T., Kolditz, O. & Yeh, G. T. 2015. Reactive modelling. Ground Water 41, 247–257. transport codes for subsurface environmental simu- Pruess, K. 1991. TOUGH2: A general numerical simulator lation. Computational Geosciences 19, 445–78. Avail- for multiphase fluid and heat flow. Berkeley: Lawrence able at: doi: 10.1007/s10596-014-9443-x Berkeley National Laboratory, Report LBL 29400. Steefel, C. I., DePaolo, D. J. & Lichtner, P. C. 2005. Reac- Ramstedt, M., Carlsson, E. & Lövgren, L. 2003. Aqueous tive transport modeling: An essential tool and a new geochemistry in the Udden pit lake, Northern Sweden. research approach for the Earth sciences. Earth and Applied Geochemistry 18, 97–108. Planetary Science Letters 240(3–4), 539−558. Avail- Rimstidt, J. D. & Barnes, H. L. 1980. The kinetics of sil- able at: https://doi.org/10.1016/j.epsl.2005.09.017 ica-water reactions. Geochimica et Cosmochimica Tempel, R. G., Shevenell, L. A., Lechler, P. & Price, J. Acta 44(11), 1683–1699. Available at: https://doi. 2000. Geochemical modelling approach to predict- org/10.1016/0016-7037(80)90220-3 ing arsenic concentrations in a mine pit lake. Applied Reimann, C., Caritat, P. D., Halleraker, J. H., Volden, T., Geochemistry 15, 475–492. Äyräs, M., Niskavaara, H., Chekushin, V. A. & Pav- Tornivaara, A., Räisänen, M. L., Kovalainen, H. & Kaup- lov, V. A. 1997. Rainwater composition in eight arc- pi, S. 2018. Suljettujen ja hylättyjen kaivosten kai- tic catchments in northern Europe (Finland, Nor- vannaisjätealueiden jatkokartoitus (KAJAK II). Suo- way and Russia). Atmospheric Environment 31(2), men ympäristökeskuksen raportteja 12/2018, Helsinki 159–170. Available at: https://doi.org/10.1016/1352- 2018, 157 p. (in Finnish) 2310(96)00197-5 Tripathy, D. P. 2014. Prevention and Treatment of Acid Rolle, M., Sprocati, R., Masi, M., Jin, B. & Muniruzzaman, Mine Drainage: An Overview. In: Sengupta, D. (ed.) M. 2018. Nernst‐Planck based description of transport, Recent Trends in Modelling of Environmental Con- Coulombic interactions and geochemical reactions in taminants. India: Springer, 95–117. Available at: doi: porous media: Modeling approach and benchmark ex- 10.1007/978-81-322-1783-1_4 periments. Water Resources Research 54. Available at: Tremblay, G. A. & Hogan, C. M. 2000. MEND Manual, https://doi.org/10.1002/2017WR022344 Volume 3 – Prediction. Natural Resources Canada. Saaltink, M. W., Domenech, C., Ayora, C. & Carrera, J. MEND Report 5.4.2c. 105 p. 2002. Modelling the oxidation of sulphides in an un- Williamson, M. A. & Rimstidt, J. D. 1994. The kinetics and saturated soil. In: Younger, P. L. & Robins, N. S. (eds) electrochemical rate-determining step of aqueous py- Mine water Hydrogeology and Geochemistry. Geolog- rite oxidation. Geochimica et Cosmochimica Acta 58, ical Society of London, Special Publication 198, 187– 5443–5454. 205. Wolery, T. J. 1992. EQ3/6 – A Software Package for Geo- Šimunek, J., Jacques, D., Šejna, M. & Van Genuchten, M. chemical Modelling of Aqueous Systems. Lawrence T. 2012. The HP2 program for HYDRUS (2D/3D): A Livermore National Laboratory, CA. coupled code for simulating two-dimensional varia- Wunderly, M. D., Blowes, D. W., Frind, E. O. & Ptacek, C. bly-saturated water flow, heat transport, and biogeo- J. 1996. A multicomponent reactive transport model chemistry in porous media. In: Version 1.0. Prague: PC incorporating kinetically controlled pyrite oxidation. Progress. 76 p. Water Resources Research 32, 3173–3187.