**E**ff**ectiveness of Fly Ash and Red Mud as Strategies for Sustainable Acid Mine Drainage Management**

**Viktoria Keller <sup>1</sup> , Sre´cko Stopi´c 1,\*, Buhle Xakalashe <sup>2</sup> , Yiqian Ma <sup>1</sup> , Sehliselo Ndlovu <sup>3</sup> , Brian Mwewa <sup>3</sup> , Geo**ff**rey S. Simate <sup>3</sup> and Bernd Friedrich <sup>1</sup>**


Received: 7 June 2020; Accepted: 6 August 2020; Published: 10 August 2020

**Abstract:** Acid mine drainage (AMD), red mud (RM) and coal fly ash (CFA) are potential high environmental pollution problems due to their acidity, toxic metals and sulphate contents. Treatment of acidic mine water requires the generation of enough alkalinity to neutralize the excess acidity. Therefore, red mud types from Germany and Greece were chosen for the neutralization of AMD from South Africa, where this problem is notorious. Because of the high alkalinity, German red mud is the most promising precipitation agent achieving the highest pH-values. CFA is less efficient for a neutralization and precipitation process. An increase in temperature increases the adsorption kinetics. The maximum pH-value of 6.0 can be reached by the addition of 100 g German red mud at 20 ◦C to AMD-water with an initial pH value of 1.9. German red mud removes 99% of the aluminium as aluminium hydroxide at pH 5.0. The rare earth elements (yttrium and cerium) are adsorbed by Greek red mud with an efficiency of 50% and 80% at 60 ◦C in 5 min, respectively.

**Keywords:** acid mine drainage; secondary materials; management; absorption; precipitation

#### **1. Introduction**

AMD is a term for wastewaters from mining processes. AMD contains a high concentration of dissolved metals from ores in a sulphate solution [1,2]. While mining, the mass of the rocks is fragmented. This leads to an increase of the surface area contacting the atmosphere which results in higher acid production rate [3]. As mentioned by Tabelin et al. [4] the contaminated excavation debris/spoils/mucks, loosely referred to as "naturally contaminated rocks", contain various hazardous and toxic inorganic elements like arsenic (As), selenium (Se), boron (B), and heavy metals like lead (Pb), cadmium (Cd), copper (Cu), and zinc (Zn). If left untreated, these naturally contaminated rocks could pose very serious problems not only to the surrounding ecosystem but also to people living around the construction and disposal sites.

Park et al. [5] mentioned that remediation options like neutralization, adsorption, ion exchange, membrane technology, biological mediation, and electrochemical approach have been developed to reduce the negative environmental impacts of AMD on ecological systems and human health. However, these techniques require a continuous supply of chemicals and energy, expensive maintenance and labor cost, and long-term monitoring of affected ecosystems until AMD generation stops.

As mentioned by Igarashi et al. [6], AMD containing Zn, Cu and As was treated by using a laboratory-type continuous ferrite process flow setup. Cu and As were removed in the first sludge, which are stable in standard leaching tests. Magnetic magnesioferrites and magnetite were generated when dissolved silicon (Si) was low. However, in the study by Igarashi et al., information about the recovery of critical metals such as rare earth elements is missing.

AMD also originates from sulphide conglomerates stored on deposits where the rain rinses the acid and metals such as uranium (U) out of the dumps [3]. Finally, the acid infiltrates the groundwater and the environment such as rivers in the affected region around the deposits which contaminates the aquatic life and the soil in the surroundings [7]. Additionally, South Africa is a water-poor country with an average rainfall of under 450 mm per year [7]. Acid mine drainage containing around 3500 mg/L [3] of sulphate has a pH value between 2 and 3 which increases the solubility of certain metals such as heavy metals. Along the river pathways, the iron (Fe) in a sulphide form oxidizes and precipitates, leading to a bright orange trail. The formation of AMD is the result of oxidation of pyrite, FeS2, with oxygen and ferric iron, Fe3+, as shown in Equations (1)–(3) [8]. Consequently, sulphuric acid is formed during this reaction. At this pH, bacteria like *Thiobacillus ferrooxidans* can survive [7] and accelerate the geochemical reactions.

$$\text{FeS}\_{2(s)} + \frac{7}{2}\text{O}\_{2(aq)} + \text{H}\_2\text{O} \rightarrow \text{Fe}^{2+} + 2\text{SO}\_4{}^{2-} + 2\text{H}^+ \tag{1}$$

$$\text{Fe}^{2+} + \frac{1}{4} \text{O}\_{2(aq)} + \text{H}^+ \rightarrow \text{Fe}^{3+} + \frac{1}{2} \text{H}\_2\text{O} \tag{2}$$

$$\text{Fe}^{3+} + 3\text{H}\_2\text{O} \rightarrow \text{Fe}^{3+} + 3\text{OH}^- + 3\text{H}^+ \rightarrow \text{Fe(OH)}\_{3(s)} + 3\text{H}^+ \tag{3}$$

AMD can be classified by the content of acid and dissolved metals with Ficklin diagram [9]. At the surface, the AMD stream is diluted by rain or surface water which leads to an increase of the pH values with white precipitation products in the AMD stream. The precipitation products contain mainly crystalline and amorphous aluminium phases [10] which adsorb other metals. Heavy metals are generally removed by iron precipitation. Aluminum precipitation only becomes important when iron content is low, but this is rarely the case.

In South Africa, coal is extracted by underground mining or open-pit mining with little surface dumping [3]. FeS<sup>2</sup> is contained in the host rock but is more abundant in the coal layers. When mining terminates, the mine will be closed by collapsing the upper layers. Water fills in the voids of the fractured rocks and reacts with the FeS<sup>2</sup> [3]. Rain penetrates through the soil and becomes acidic and influences the natural groundwater. In the dams around coal mines in Middelburg and Witbank in South Africa, the salinity and sulphate concentration amount to about 200 mg/L [3], which is the limit for domestic use.

Regarding the problem of growing deposits, the branch of research on wastewater treatment is of high importance. The primary aim of wastewater treatment is to recover valuable metals and neutralize the acidic solution [11]. An advantage is the higher content of valuable metals in the wastewater compared to the ores used in primary production.

There are further possibilities for water purification like the SAVMIN™ process. This is the simplest technology for reducing high sulphate concentrations is lime precipitation developed by MINTEK in South Africa [9]. Precipitation of metal hydroxides and sulphides was performed using sodium hydroxide, calcium carbonate and hydrogen sulphide. Regarding the costs, calcium carbonate is most suitable for the neutralization experiments. Because the neutralization is exothermic, room temperature was chosen for the removal of metals from waste solution. Pre-oxidation with hydrogen peroxide transforms the bivalent Fe to the trivalent Fe which precipitates at lower pH values [12] than the bivalent Fe. Analogously, As is oxidised from the trivalent state to the pentavalent state to enable hardly dissolving oxides [13,14]. Removal of As(III) and As(V) can occur during neutralization simultaneously. Additionally, because of the high reactivity of sulphides with heavy metal ions, they can be applied to achieve a larger decrease in metal concentration after neutralization with hydroxides [13]. Trivalent Fe cannot be removed by sulphides precipitation since they do not form stable sulphides under

wastewater conditions [13]. Instead, the trivalent will be reduced to the bivalent state which forms iron(II) sulphide (FeS) [14]. Another possibility for selective removal of metals is cementation with nanoscaled zerovalent powders which have a large reaction surface area [15]. Cementation at room temperature and reductive precipitation took a short time in contrast to adsorption. There is a tendency that metals are adsorbed in this sequence: Fe(II)/Fe(III) > Cu(II) > Mn(II) > Zn(II) [16] following the Langmuir model which states that all surfaces of a given solid have the same affinity to adsorb metals [17]. Addition of brine can decrease the kinetics of neutralization, so there is a long-term effect which avoids a too-high pH value [18]. Precipitation of dissolved metals is achieved by the introduction of chemical additives such as sodium carbonate and oxalic acid to mostly affect changes in pH and/or ligand concentration at room temperature.

High standards are applied for drinking water which is chosen as a better reference in comparison to the purified water. Pure water is clear, colour- and tasteless and low in microbiological contamination [14]. Some metals, like Fe3+, change the colour of water. Fe3<sup>+</sup> turns the water into a reddish-brown colour [19]. The optimum pH for drinking water ranges between 6.5–8.5 [19]. The taste threshold for salts depends on the cation (sodium, potassium) and the anion (chloride, sulphate) concentrations. For sodium chloride, the taste threshold concentration is 200–300 mg/L [19]. Turbidity indicates any kind of contamination which can be caused by inorganic and organic matter in the form of suspended particles [19]. Microorganisms are attached to particles [14]; therefore, filtration of particles can reduce their population in treated water. Turbidity of AMD can also be a sign of natural precipitation of iron. In Germany, the most important laws for wastewater treatment are:


The European and German standards are based on the recommendations of the World Health Organization (WHO) [14], which has set up health-based guidelines for chemicals in drinking water [19]. The effects of contaminants on taste, odour and appearance are cited by the WHO, while the most important effects such as toxicity are not mentioned [19]. Mostly, because the concentration in treated water is too low, it is not possible to determine guideline values for some metals such as silver (Ag) and gold (Au). In the case of aluminium, the guideline value is nine-fold higher than the value achievable by current technology [19]. The TrinkwV restricts the concentration of aluminium since there could be a correlation between aluminium consumption and Alzheimer's diseases [14]. The pH value of drinking water is between 6.5 and 9.5. Water with carbonic acid (H2CO3) can have a lower pH value than 6.5 [21].

Purification of wastewaters is a high-complex process. The process line depends on the origin of the wastewater and therefore the contaminations as well as the subsequent use of the purified water. For use as drinking water, the water needs to fulfil special regulations, for example, the TrinkwV [21]. The purification process contains several steps such as mechanical separation of solids and suspended particles (e.g., flotation, sedimentation, filtering) [14,22]; oxidation of dissolved metals to solids [14]; ion exchange processes to neutralize the water from salts and metals [23]; disinfection of water by oxidation media (ClO2, O3, hyper chlorites) or UV-light [24] and biological treatment (aerobic processes and anaerobic processes) [11].

The by-products used in this paper for treating AMD are red mud and CFA. Red mud comes from the Bayer process [25,26], whereas coal fly ash is the inorganic residue from coal combustion. In red mud, there are mainly hematite, other crystalline aluminium silicate phases, silica, titanium oxide, rare earth oxides and un-leached residual alumina [27]. Coal fly ash contains approximately 50% of crystalline phases, mostly quartz and mullite [28].

The neutralization process enables the transformation of several waste streams into valuable products regarding the zero waste guidelines in technology. There are several water treatment methods based on red mud and CFA [29] like acid and base treated powder [30–33] or zeolites as special aluminium silicates [34,35], but not all of them have yet been tested in acid mine drainage. Transforming CFA to zeolites can increase the adsorption of dissolved metals [36].

Our main aim is to perform a neutralization of AMD using secondary materials such as red mud and CFA and to discuss the possibility of the recovery of valuable metals such as Al, Zn, Mn and rare earth elements. The neutralization of AMD in this study is a preparation of wastewater for return into downstream processing or releasing to the environment. Acidic water can damage the plants by corroding or damaging the environment. Another aim of neutralization with alkaline material is the preparation of solid waste materials for further metal winning processes. The final aim of this study is to develop a process for water purification which fulfils the following criteria: economical, good water quality after clarification, resources of materials used for neutralization are as regionally available (<500 km) as possible and waste products are environmentally friendly.

#### **2. Materials and Experimental Procedure**

#### *2.1. Characterisation of the Studied Materials*

The AMD sample was collected from Mpumalanga, South Africa. All sampling and laboratory analysis is performed in accordance with recognized global standards such as the International Standards organization (ISO). After sampling and laboratory analysis in South Africa, all samples were sent to Germany. The AMD water was characterized by using ICP-OES analysis (SPECTRO ARCOS, SPECTRO Analytical Instruments GmbH, Kleve, Germany) and the solid samples by X-ray fluorescence, (Axios FAST, Malvern Panalytical GmbH, Kassel, Germany). The AMD was first filtrated in order to remove the formed precipitate, but AMD was not acidified. The solid samples were ground up before the X-ray diffraction analysis (XRD) analysis. Additionally, characterization of the red mud and fly ash to the X-ray powder diffraction XRD (Bruker AXS, Karlsruhe, Germany) was operated. Bauxite residue, employed during AMD-treatment as the main raw material, was provided by Aluminum of Greece plant, Metallurgy Business Unit, Mytilineos S.A. (AoG). The sample was first homogenized by using laboratory sampling procedures (riffling method) and then a representative sample was dried in a static furnace at 105 ◦C for 24 h. Subsequently, the material was milled using a vibratory disc mill and the sample was fully characterized.

Chemical analyses of major and minor elements were executed via the fusion method (1000 ◦C for 1 h with a mixture of Li2B4O7/KNO<sup>3</sup> followed by direct dissolution in 10% HNO<sup>3</sup> solution) through a Perkin Elmer 2100 Atomic Absorption Spectrometer (AAS) (Waltham, MA, USA), a Spector Xepos Energy Dispersive X-ray Fluorescence Spectroscope (ED-XRF) (SPECTRO, Kleve, Germany), a Thermo Fisher Scientific X-series 2 Inductively Coupled Plasma Mass Spectrometer (ICP-MS) (Waltham, MA, USA) and a Perkin Elmer Optima 8000 Inductively Coupled Plasma Optical Emission Spectrometer (ICP-OES) (Waltham, MA, USA), whereas the loss of ignition (LOI) of the sample was provided by differential thermal analysis (DTA), using a SETARAM TG Labys-DS-C (Caluire, France) system in the temperature range of 25–1000 ◦C with a 10 ◦C/min-heating rate, in air atmosphere.

Mineralogical phases were detected by XRD using a Bruker D8 Focus powder diffractometer with nickel-filtered CuKα radiation (λ = 1.5405 Å) coupled with XDB Powder Diffraction Phase Analytical System version 3.107 which evaluated the quantification of mineral phases via profile fitting specifically for bauxite ore and bauxite residues.

#### 2.1.1. Acid Mine Drainage from South Africa

The pH of the AMD wastewater ranges around 2.0. The ICP-OES analysis given in Table 1 shows the composition from the wastewater which has a dark red–yellow colour as shown in Figure 1 and their limits in the TrinkwV. According to the Ficklin diagram from [9], the AMD sample is defined as highly acidic and as having a high metal content inn the water. The total concentration of the relevant metals (mg/L), amounted to 18.903 mg. The single concentration of elements in AMD (mg/L) amounted to 14.2 Zn, 2.09 Ni, 1.94 Co, 0.65 Cu, 0.02 Cd, and 0.003 Pb. Comparison of the metal concentration with the TrinkwV shows that the critical metals were cadmium, chromium and nickel, iron manganese and the sulphate. The pH of 2.0 was too low (interval is 6.5–9.5).


**Table 1.** Metal containing wastewater (three significant figures).

**Figure 1.** Mine drainage water from South Africa.

#### 2.1.2. Red Mud from Germany and Greece

For the neutralisation experiments, wto sorts of red mud were used. In Figure 2, the red mud shown on the left originates from Germany (Stade, Lower Saxony, Germany), and the right one was from Greece (Aluminium of Greece, Viotia). The Greek red mud is darker due to its higher content of iron as shown in Table 2. This table shows that the Greek red mud contains more iron oxide, alumina and lime than the sample from Germany. In contrast, the German sample contains more sodium oxide, titanium dioxide and silica. Red mud is very alkaline due to the sodium hydroxide from the Bayer process (pH < 12). For further experiments, the s/L-rates of 1:10 (100 g/L AMD) and 1:5 (200 g/L AMD) were chosen.

**Table 2.** Content of oxides in the red mud samples wt.% or ppm (Sc).


−

**Figure 2.** Red mud samples, (**left**) from Germany, (**right**) from Greece.

The XRD analysis of Greek Red Mud is presented in Table 3.


**Table 3.** X-ray diffraction analysis (XRD) analysis of Greek red mud (*d* = 1.87 µm).

The XRF analysis of the rare earth elements in Greek red mud is shown in Table 4.

**Table 4.** Composition of rare earth elements in Greek red mud.


Mineralogical analysis of the German red mud was explained by Kaussen and Friedrich [37]. The missing 1.5% accounts for unaccounted, unknown or amorphous content. There is no sufficient reference data (known crystallographic measurements) to characterize 100% of bauxite residue

mineralogical composition. In the current XDB full profile fitting mineral phase quantification it was not possible to quantify amorphous content. Amorphous content can be determined in phase quantification when a known quantity of an internal standard such as corundum is added to the sample.

#### 2.1.3. Coal Fly Ash from South Africa

The coal fly ash used in this study was obtained from an Eskom thermal power station, Mpumalanga, South Africa. The XRD analyses of fly ashes have been measured with a Bruker D4 powder diffractometer in Bragg–Brentano geometry. The qualitative evaluation was done with the program EVA and the PDF 2 file from ICSD. The quantitative determination was carried out using a Rietveldt refinement with the program TOPAS from Bruker. The analyses were carried out on powders <60 µm which were prepared as backloading compacts. SEM-analysis was performed using GEMINI SEM 300, Car Zeiss Microscopy GmbH; Oberkochen, Germany.

As shown in Table 5, the main components of fly ashes are silica, alumina, lime and hematite with a sum of 96.1 wt.%, and Y2O<sup>3</sup> in small amounts (0.013). X-ray diffraction shows that 49% of the phases are amorphous phases, and the proportion of crystalline phase is (in %): 0.3 CaO, 0.7 Fe2O3, 38 Al4SiO8, 12 SiO2, and 0.1 TiO2. Most of the other oxides are present in traces. The coal fly ash in Figure 3 has a light grey colour (left) and spherical submicron particles (right).

**Table 5.** X-ray fluorescence (XRF) analysis of the content of compounds in the coal fly ash (wt.%).


**Figure 3.** Fly ash powder (**left**) and SEM analysis (**right**).

#### *2.2. Parameters for Neutralization*

  For neutralization, there are four important parameters:


The experiments, see Table 6, were designed for each media separately because coal fly ash is less alkaline, but contains yttrium (0.01%), and it is very stable for leaching. Every experiment was conducted twice to reproduce the results.



#### *2.3. Setup and Test Performance*

The aim of the following experiments is to prove the necessity of by-products to neutralize 500 mL of metallic wastewater. The quantity of the substances is:


The acid mine drainage was heated up to between 20 ◦C and 60 ◦C and agitated for 120 min up to 350 rpm. After precipitation, 6 samples with approximately 10 mL (from 500 mL samples) or 50 mL (from 1 L samples) were taken out of the suspension. The filtered 10 mL samples were diluted with deionised water, from a 5 mL solution samples to 50 mL, without new dilution. The solid residues were analysed after precipitation in order to determine the metal content and the absorption kinetics regarding the consumed red mud and fly ash. The precipitation and absorption fraction were calculated using Equations (4) and (5), respectively.

#### **3. Results and Discussion**

With respect to iron ions as the greatest fraction in the AMD wastewater, white and yellow flakes can be found in the samples after 5 and 10 min. In the samples with the red muds, there are white flakes, as seen Figure 4. In the samples with coal fly ash there is a yellow suspension due to a lower pH than 3 as seen in Figure 5. Yellow-orange powders are left after a second filtration of the solution. These yellow–orange flakes indicate that iron has precipitated as iron(III) hydroxide from the solution. In the red mud samples, the iron hydroxide is already removed from the red mud. The reformation of Fe(OH)<sup>3</sup> in the filtrate of coal fly ash indicates that the formation of Fe(OH)<sup>3</sup> at constant pH has a long term kinetic of several minutes. –

**Figure 4.** Typical flakes in the 5 min and 10 min samples after treatment with red mud.

–

**Figure 5.** Suspension in the 5, 10 and 20 min samples after treatment with coal fly ash.

The XRF analysis showed that the yellow powder after treatment with coal fly ash, filtration and drying mainly contains approximately


We supposed that iron and aluminium precipitate as iron(III) and aluminium(III) hydroxide, respectively. Additionally, some compounds precipitate as sulphates, e.g., CaSO4, because it is slightly soluble, what is presented in Figure 6. Additionally, the presence of aluminium oxide in the product was revealed.

**Figure 6.** XRD analysis of the precipitation product.

#### *3.1. pH Values as Function of the Time*

The initial pH value of the AMD water was 1.973 ± 0.031 (the calculated relative error ∆pH/pH = 15.7%) for all 12 experiments. After the measurement of pH-Value, AMD solution was added in a glass reactor and heated to 60 ◦C. Red mud and fly ash (50–100 g) were added at 60 ◦C in the solution of AMD. The pH-Values were measured in time of 120 min. The same procedure was performed at room temperature and results were compared with ones obtained at 60 ◦C.

The pH was plotted as a function of time as seen in Figure 7. All samples have a strong pH increase in the first 5 min. The maximum pH of 6 can be reached by 100 g German red mud at 20 ◦C. The pH values from the Greek red mud tend to be lower than the ones from the German red mud, the maximum pH being 5.9, which is in the same range as the German red mud. The samples with coal fly ash indicate that in the first 5 min, the average pH was 2.78 ± 0.14 (∆pH/pH = 5%) and most of the iron ions were still dissolved in the solution. After 10 min, the average pH was 2.93 ± 0.11 (∆pH/pH = 4%) and there was still some iron(III) dissolved. For full iron(III) removal, the pH needs to be at least higher than three, which was mentioned by Stiefel [11].

— — — **Figure 7.** pH values in time (GerRM—German red mud; GrRM—Greek red mud; CFA—coal flying ash).

– = 1 − Concentration of metal at time *t* Initial concentration of metal in AMD = 1 − 0 . Coal fly ash had pH-values between 8 and 11 depending on the quantity of the added sample, in comparison to red mud (pH > 12). Especially because of 49% of amorph structure and 38% of very stable mullite in this structure, as shown at Figure 3, the fly ash has smaller neutralisation efficiency in comparison to used red mud. The temperature is very important in regards to pH measurements. As the pH value changes with the change in temperature, the new measured pH value is technically the true pH value. Under laboratory conditions, a note of the temperature and pH value should be made together. There is only one major temperature effect in pH measurement that can cause errors in readings. It is the only reasonably predictable error due to changes in temperature, and is the only temperature-related factor that a pH meter with temperature compensation can correct. This temperature error is very close to 0.003 pH/ ◦ . Generally in our experiments, Because of an influence of the diffusion of the acidic solution, an increase of temperature from room temperature to 60 ◦C increases the neutralization efficiency of AMD during the addition of fly ash and red mud. It seems that adding FA and RM can increase pH of AMD at the scale of 120 min. Long-term neutralization efficiency is required, as confirmed in subsequent work, and in the work of Paradis et al. [38].

#### *3.2. Comparison of Precipitation E*ffi*ciency per Material*

Regarding the literature [8], the critical pH for iron(III) precipitation using CFA can be assumed as approximately 3.0. At this point, 99% of the dissolved iron is precipitated. The maximum concentration of dissolved iron is 44.2 mg/L. However, it must be considered that the iron concentration in the solution was higher direct after the filtration since iron(III) hydroxide has precipitated after filtration. The following six figures (Figures 7–12) show the precipitation efficiency as a function of the pH of aluminium, manganese and zinc. The symbols represent the time of sampling (5, 10, 20, 40, 60, 120 min). The precipitation efficiency *x<sup>P</sup>* was calculated by Equation (4). Negative values for *x<sup>P</sup>* indicate leaching effects and enrichment in the acid mine drainage.

$$\chi\_P = 1 - \frac{\text{Concentration of metal at time } t}{\text{Initial concentration of metal in AMD}} = 1 - \frac{c\_t}{c\_0} \tag{4}$$

#### 3.2.1. Manganese

In Figure 8, by using the German red mud, manganese had a lower precipitation efficiency at a higher temperature (60 ◦C) than the sample with more mass at 20 ◦C at the same pH-value. There is a tendency of reaching a maximum precipitation efficiency of approximately 75% as the sample with 100 g at 20 ◦C, which shows a light equilibrium state. Using a Greek 50 g red mud results in leaching in acid mine drainage up to 10% at a higher temperature. Greek red mud with higher mass at room temperature leads to the precipitation of manganese up to 30%. Manganese tends to be leached in both coal fly ash experiments. In the 100 g sample at 60 ◦C it is leached from 10% to 22%. In the 250 g lower heated experiment, manganese was leached to 20% in the first 5 min and then precipitated until it was at a value of 6% of the initial concentration. It follows that manganese may precipitate after pH 4 by using red mud.

**Figure 8.** Participation efficiency of manganese.

#### 3.2.2. Aluminium

Figure 9 shows the precipitation efficiency of aluminium for all precipitation media compared to the initial concentration of 370 mg/L. German red mud removes 99% of the aluminium as aluminium

–

–

hydroxide at pH 5.0. Higher temperature leads to an earlier precipitation pH and therefore, to higher efficiency at the same pH compared to the experiment with higher mass at 20 ◦C. As a contrast, higher mass results in faster pH increase so that the pH is higher after the same time. Experiments with Greek red mud showed a similar tendency, although the precipitation efficiency of the sample in heated acid mine drainage was just 14% after 5 min. In all coal fly ash samples, the aluminium was not precipitated under pH 3.5. In the experiment with heated acid mine drainage, a leaching from coal fly ash up to 89% at pH 2.8 occurred at first, then re-precipitation. As a conclusion, all samples showed similar correlation of pH and precipitation efficiency, like an s-curve with a strong increase between pH 3.0–3.6 in all experiments, see the red circle. Positive precipitation efficiency starts at pH 3.3–3.5 until pH-Value of 5. After the pH of 5.0, precipitation of aluminum is complete.

**Figure 9.** Participation efficiency of aluminium.

**Figure 10.** Participation efficiency of zinc.

∙ AMD

∙ AMD

)

)

RM/CFA

RM/CFA

=

=

<sup>0</sup> <sup>0</sup>

<sup>0</sup> <sup>0</sup>

=

=

AMD RM/CFA

AMD RM/CFA (<sup>0</sup> −

(<sup>0</sup> −

**Figure 12.** Kinetics of manganese adsorption for both types of red muds [mg/g].

#### 3.2.3. Zinc

 **100/20**  In German red mud, the zinc is completely precipitated, to under the detection limit. Zinc, as shown in Figure 10, can reach a minimum precipitation of 45% in both combinations of Greek red mud. The maximum precipitation efficiency can be at least 65% for the sample at 20 ◦C with 100 g and 59% for the pre-heated acid mine drainage experiment with 50 g Greek red mud. In coal fly ash, zinc was precipitated between 20% and 60% in both cases, whereas the non-heated sample had a better efficiency of 60% after 120 min. There is a linear correlation between pH and precipitation efficiency. As a conclusion, pre-heated acid mine drainage results in leaching of all metals, which is also expected of zinc but only for the first 40 min, longer neutralization leads to stronger zinc back-dissociation.

Zn

2+ + 2OH

−

→ Zn(OH)<sup>2</sup>

#### *3.3. Adsorption Kinetics of Adsorbed Metals*

The adsorption kinetics can be compared by using the rate of adsorbed material per gram precipitation media, see Equation (5).

$$q\_t = \frac{(c\_0 - c\_t)\_{M\ell} \cdot V\_{\text{AMD}}}{m\_{\text{RM}/\text{CCA}}} \tag{5}$$

with:

*q<sup>t</sup>* rate of adsorbed material [mg/g] *c*0, *c<sup>t</sup>* concentration of the metal ions [mg/L] (*c*0: initial, *c<sup>t</sup>* : after *t* min) *V*AMD volume of the tested AMD water [L] *m*RM/CFA mass of the used precipitation media [g]

#### 3.3.1. Aluminium

Figure 11 shows the adsorption kinetics for the aluminium in the acid mine drainage precipitated by the both types of red muds. The coal fly ash samples were missing due to enrichment in the AMD. The aluminium in the ash may be a soluble phase. The concentration of aluminium in the solution was 370 mg/L, therefore there were 185 mg dissolved in 500 mL. It can be recognised that the adsorption potential was higher using red mud at 60 ◦C than at 20 ◦C. This indicates that the aluminium adsorption is a temperature-activated process and therefore, it is more like chemisorption than physisorption. The increase to 60 ◦C doubles the adsorption potential per gram red mud from 1.8 mg/g (0.9% of the total amount of dissolved aluminium in solution) to 3.7 mg/g (2%). The difference between the use of the German red mud and the Greek red mud was relatively small, Greek red mud was a little bit better than the German type. In case of the 50 g 60 ◦C sample, the mass of the adsorbed part was 185 mg, which was the maximum, whereas the 100 g 20 ◦C adsorbed only 180 mg. In the case of aluminium, both combinations achieved equal precipitation efficiency of 100%. After 20 min (average pH = 4.751 ± 0.495), the precipitation can be assumed as completed.

#### 3.3.2. Manganese

The concentration of manganese was 82 mg/L which means 41 mg in 500 mL acid mine drainage water. The adsorption kinetics for manganese in Figure 12 show that the German red mud had better adsorption potential than the Greek one. Because of enrichment, the 50 g sample at 20 ◦C with the Greek red mud is missing, see Equation (6).

$$\text{MnO} \rightarrow \text{Mn}^{2+} + \text{O}^{2-} \tag{6}$$

None of the samples have achieved the equilibrium state. The German red mud 100 g and 20 ◦C sample could have an equilibrium at 0.31 mg/g manganese. One gram can adsorb 0.7% of the dissolved manganese which is in total 30 mg (73% in total). For the German red mud 50 g at 60 ◦C sample, it can be assumed that the equilibrium is approximately 0.6 mg/g red mud which means an adsorption grade of 1.4% and therefore, 30 mg in total (also 73% in total). This means that both combinations can work equally for manganese, see Table 7. The equilibrium for the 50 g Greek red mud sample at 60 ◦C was negative due to leaching effects. Therefore, the 120 min adsorption rate cannot be determined exactly. Under recognition of instrumental errors, the equilibrium can be near zero.


**Table 7.** Adsorption kinetics of manganese by red mud with the percentage at assumed equilibrium.

#### 3.3.3. Zinc

Figure 13 shows the sorption kinetic of zinc. The concentration of dissolved zinc was 14.3 mg/L, therefore, there were 7.15 mg dissolved in acid mine drainage. The equilibrium of zinc for coal fly ash was between 0.01 and 0.02 mg/g which were at a maximum 0.28% of the total amount of the dissolved zinc per gram in coal fly ash. The 100 g of coal fly ash could adsorb a maximum 2 mg which is 28% of the total amount and 250 g maximum 5 mg zinc (nearly 70% of the total zinc). For Greek red mud, the equilibrium at 20 ◦C was approximately 0.047 mg/g which is a grade of 0.66%. A 100 g portion of this sample can adsorb 4.7 mg zinc which is 66% of the total zinc in solution. The best sample with 50 g and 60 ◦C had a short equilibrium 0.067 mg/g for 15 min which was 0.9% adsorption per gram with total adsorption of 3.35 mg (47%) zinc. After 20 min, an increase of adsorption can be seen. This indicates that there could be another mechanism of adsorption, like precipitation. After 40 min, the mixture achieved a pH of 4.1. Regarding the results, the Greek red mud had a better adsorption potential than the coal fly ash but overall, the 250 g coal fly ash sample at 20 ◦C can be as useful as the 100 g of Greek red mud at 20 ◦C, as shown in Table 8. The precipitation follows Equation (7).

$$\text{Zn}^{2+} + 2\text{OH}^- \rightarrow \text{Zn}(\text{OH})\_2 \tag{7}$$

**Figure 13.** Kinetics of zinc adsorption for Greek red mud and coal fly ash [mg/g].

)AMD

RM/CFA

=

( − <sup>0</sup>

NaOH → Na

<sup>+</sup> + OH −


**Table 8.** Kinetics of zinc adsorption for Greek red mud and coal fly ash.

#### *3.4. Leaching Kinetics*

Analogously, the leaching kinetics can be calculated with Equation (8) as well as the adsorption kinetics in the previous chapter.

$$q\_t = \frac{(c\_t - c\_0)\_{M\epsilon} V\_{\text{AMD}}}{m\_{\text{RM}/\text{CFA}}} \tag{8}$$

#### Sodium

Due to a high concentration of sodium hydroxide, it is the main component for neutralization of acid mine drainage. The consumption of sodium from coal fly ash for neutralization of AMD was not very high, only approximately 0.05 mg/g coal fly ash since the content of sodium in coal fly ash was just 0.35 wt.% (<1 mg/g), whereas the German red mud contains 8.9 wt.% (66 mg/g) and the Greek red mud 1.8 wt.% (13 mg/g). Regarding the average leaching kinetics of sodium in Figure A1 in Appendix A, German red mud at 60 ◦C has the highest leaching ability with 50 mg/g sodium which is 76% of total sodium in the sample as seen in Table 9 and 20 mg/g in the non-heated sample which was 30%. The heated Greek red mud had a similar leaching ability like the non-heated German red mud with 20 mg/g. The non-heated Greek red mud had a leaching ability of 10 mg/g which was 77% of the total sodium in the sample. As a conclusion, the non-heated German red mud and the heated Greek red mud sample have a similar ability to achieve the same neutralization pH per gram.

$$\mathrm{NaOH} \rightarrow \mathrm{Na^{+}} + \mathrm{OH^{-}} \tag{9}$$

**Table 9.** Content of sodium at the beginning and after leaching in the types of red mud.


#### *3.5. Rare Earth Concentration in the One Litre Experiments*

The initial concentration of cerium is 5.51 mg/L and of yttrium was 2.05 mg/L. Figure 14 and Table 10 shows the concentration of the rare earth elements yttrium and cerium during the experiments. The detection limit for the rare earths was 1.0 mg/L, which marks the start of the *y*-axis. The rare earth elements (yttrium and cerium) were adsorbed by Greek red mud with an efficiency of 50% and 80% at 60 ◦C in 5 min, respectively. After 5 min, using German red mud can reduce the concentration of the rare earth to nearly under the detection limit (Ce: 1.12 mg/L, Y: <1 mg/L). Using Greek red mud reduces cerium to under the detection limit after 60 min, yttrium after 40 min. Using coal fly ash can result in re-dissolving into acid mine drainage which indicates an equilibrium of the concentration in the solid and in the acid mine drainage. After 60 min, the concentration of the rare earths does not change much. There is a steady-state in concentration.

**Figure 14.** Concentration of yttrium and cerium in time.


**Table 10.** Yttrium and cerium in precipitation experiment [mg/L].

#### *3.6. Sodium Concentration Regarding the Guidelines*

– The initial concentration of the sodium in acid mine drainage is 52.4 mg/L. The guideline for sodium in drink water is 200 mg/L [21], see the red line in Figure A2 in Appendix A. The German red mud leads to a final concentration of 4–5 g/L which is twenty-five times that of the recommended concentration. The final concentration of the Greek red mud was approximately 2 g/L which is tenfold of the recommended concentration. Only the coal fly ash sample fit into the guideline, with a final concentration of approximately 70 mg/g. Therefore, for the red mud samples, it is necessary to distinguish if the water is used as drink water after treatment.

With respect to the results, there is a possibility of purifying the acid mine drainage by precipitation with red mud and coal fly ash. Overall iron is removed nearly completely, and aluminium also precipitates to 99% in most samples. Same samples as Greek red mud and coal fly ash have a potential for selective removal of metals which can be useful for multistage process designing. In all red mud samples, the concentration of sodium has increased strongly.

Comparison of the adsorption kinetics shows that in some cases, the combinations of higher mass or higher temperature could give similar results like those in the experiment with German red mud and manganese (both 73% of total) or the one with both kinds of red muds and aluminium. Additionally, for zinc precipitation, the 100 g, 20 ◦C Greek red mud and the 250 g, 20 ◦C coal fly ash have equal precipitation efficiency. For economical use, higher mass and low temperature are preferred. Additionally, the red mud and the coal fly ash can replace the soil, which enables outside application for neutralisation. The neutralised red mud and coal fly ash can be used for further metal-winning treatment.

The main cause for the increase of pH is the sodium hydroxide. Since the red mud can release up to 80% of the sodium ions into acid mine drainage, a pH increase up to 6 is possible. A combination of

higher mass and higher temperature can achieve the neutral water state. The coal fly ash causes pH increase by dissociation of metal oxides in acids, but the pH increase is not as high as using red mud since coal fly ash does not contain sodium oxide. The oxides release oxygen to acid mine drainage, which influences the potential of acid mine drainage, resulting in the formation of hydroxyl ions and shifts the precipitation pH to lower values.

The total amount of the rare earth elements was 9.83 mg/L, which can be reduced after neutralization under 4 mg/L using German red mud.

The German red mud performs the best results of all media. The Greek red mud also has good results, but they were not as good as the German ones. The precipitation efficiency of coal fly ash was low in some samples. Coal fly ash is usable when iron, aluminium and zinc needs to be precipitated but manganese would be dissolved in solution.

#### **4. Conclusions**

The following conclusions are found in this work:


Kinetic study shows that contrast combinations of mass and temperature can achieve the same precipitation adsorption. Increasing of temperature to 60 ◦C can double the adsorption capacity of material like aluminium, which is equivalent to the twofold mass of material. The alternative with the higher amount of material should be preferred because of the higher pH increase due to easily-soluble sodium hydroxide. As the results indicated, there is much less metal removal from solution using the adsorption method.

**Author Contributions:** Conceptualization, B.X., V.K. and S.S.; Funding acquisition, S.N. and B.F.; Investigation, V.K., Y.M. and B.M.; Methodology, V.K., S.S. and B.X.; Supervision, S.N., G.S.S. and B.F.; Writing—original draft, S.S., B.X., S.N., B.M., G.S.S. and B.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the International Office of the BMBF in Germany, grant number. 01DG17024, and by NRF in South Africa (grant number: GERM160705176077). The APC was funded by the International Office of the BMBF in Germany, grant number. 01DG17024.

**Conflicts of Interest:** The authors declare no conflict of interest.

—

**Figure A1.** Kinetics of sodium adsorption for German and Greek red mud [mg/g].

**Figure A2.** The concentration of sodium in treated acid mine drainage regarding the guidelines [18].

–

–

–

–

–

–

–

–

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **E**ff**ects of Backfilling Excavated Underground Space on Reducing Acid Mine Drainage in an Abandoned Mine**

**Kohei Yamaguchi 1,2,\*, Shingo Tomiyama <sup>3</sup> , Toshifumi Igarashi <sup>3</sup> , Saburo Yamagata <sup>4</sup> , Masanori Ebato <sup>5</sup> and Masatoshi Sakoda <sup>6</sup>**


Received: 14 June 2020; Accepted: 28 August 2020; Published: 31 August 2020

**Abstract:** Three-dimensional groundwater flow around an abandoned mine was simulated to evaluate the effects of backfilling the excavated underground space of the mine on reducing the acid mine drainage (AMD). The conceptual model of the groundwater flow consists of not only variable geological formations but also vertical shafts, horizontal drifts, and the other excavated underground space. The steady-state groundwater flow in both days with high and little rainfall was calculated to calibrate the model. The calculated groundwater levels and flow rate of the AMD agreed with the measured ones by calibrating the hydraulic conductivity of the host rock, which was sensitive to groundwater flow in the mine. This validated model was applied to predict the flow rate of the AMD when backfilling the excavated underground space. The results showed that the flow rate of the AMD decreased by 5% to 30%. This indicates that backfilling the excavated space is one of the effective methods to reduce AMD of abandoned mines.

**Keywords:** abandoned mine; groundwater flow analysis; acid mine drainage; backfilling

#### **1. Introduction**

Acid mine drainage (AMD) is generated at many active, closed, and abandoned mines throughout the world. AMD is a serious environmental issue in the mining industry [1–9], which is generally characterized by low pH and high concentrations of sulfate, heavy metals (e.g., copper (Cu), lead (Pb), zinc (Zn), and cadmium (Cd)) [10,11], and metalloids (e.g., arsenic (As), [12]). Low pH is caused by oxidation of sulfide minerals and dissolves heavy metals in the host rock. Moreover, the contribution of bacteria is important under in situ conditions for AMD formation [13,14]. The chemical reactions (1)–(4) [15] causes low pH when the oxygenated rainwater contacts with pyrite in the unsaturated zone. Backfilling the excavated underground space may reduce the amount of AMD due to the decrease in

hydraulic gradient [16,17]. In addition, the contact area between oxygenated rainwater and pyrite decreases due to the rise of groundwater levels.

$$\text{FeS}\_2\text{(s)} + 7/2\text{O}\_2 + \text{H}\_2\text{O} = \text{Fe}^{2+} + 2\text{SO}\_4^{2-} + 2\text{H}^+ \tag{1}$$

$$\text{Fe}^{2+} + 1/4\text{O}\_2 + \text{H}^+ = \text{Fe}^{3+} + 1/2\text{H}\_2\text{O} \tag{2}$$

$$\text{Fe}^{3+} + 3\text{H}\_2\text{O} = \text{Fe(OH)}\_3\text{(s)} + 3\text{H}^+ \tag{3}$$

$$\text{FeS}\_2\text{(s)} + 14\text{Fe}^{3+} + 8\text{H}\_2\text{O} = 15\text{Fe}^{2+} + 2\text{SO}\_4^{2-} + 16\text{H}^+\tag{4}$$

Sulfide minerals in the host rock exposed to shallow groundwater or rainwater are oxidized during the operation of the mine. In addition, AMD is continuously generated for more than several decades after closing or abandoning the mines [18].

In Japan, there are over 5500 closed or abandoned non-ferrous metal mines and 79 mines of these are treating AMD. AMD from closed or abandoned mines is commonly treated by neutralization with hydrated lime or sodium hydroxide. In this application, most toxic heavy metals are precipitated and removed [19], and then the treated water is released into nearby rivers. The treatment of AMD induces a large load from an economic perspective although lime neutralization has effectively been used over the last 40 years in Japan. Total subsidiary aid cost for about 50 years is ~70 billion JPY (~650 million USD) in Japan [20].

Active treatments by adding alkaline reagents are costly and should last for decades. On the other hand, passive treatments are expected to be applied to mines with a relatively lower flow rate of AMD [21]. For both treatments, it is necessary to reduce the amount of produced AMD. Thus, it is important to clarify the processes of generating AMD. There have been a variety of studies of AMD monitoring and characterization in not only Japan (e.g., [22–24]) but also other countries [25–34]. Some studies pointed out that groundwater flow patterns in and around mines should be evaluated to purpose countermeasures against AMD reduction. There are several methods of the mitigation, such as covering of ground surface with low-permeable layers and impoundment of land subsidence [35–41]. In this study, the effects of backfilling the underground space in a mine on produced AMD were examined by 3-D groundwater flow model to reduce the AMD produced because the mine selected has a huge volume of underground space already excavated.

#### **2. Geology and Mining of Study Area**

The selected mining area is located in valley terrain at an altitude of 300 to 400 m (Figure 1). The basement rock mostly consists of pre-Neogene granite. Neogene tuff, andesite, and rhyolite were deposited on the granite basement. They erupted and deposited on the granite in the marine. The type of mine is a vein-type deposit formed in faults and fractures in the Neogene strata. Distribution of the mined area was recorded in the pit map created during the operating period when drifts were excavated at a depth of 60, 90, and 150 m from the ground surface (−2 L, −3 L, and −5 L levels, respectively; Figure 1). Among them, the −5 L level drift is used as drainage of mine water to the mine mouth. Mining was carried out by the shrinkage method at the beginning of the operation because the host rock of the Neogene strata was solid and because the veins in the Neogene strata were inclined with slopes of 70◦ to 90◦ . Since the mine is producing sludge generated from AMD neutralization, backfilling the formed sludge mixed with cement (sandy slime) into excavated underground space was adopted to prevent collapses of the space. This is because the host rock to be excavated became crumblier with the progress of operation.

*Minerals* **2020**, *10*, x FOR PEER REVIEW 3 of 13

*Minerals* **2020**, *10*, x FOR PEER REVIEW 3 of 13

**Figure 1.** Distribution of old mining levels and excavated underground space of the study area. **Figure 1.** Distribution of old mining levels and excavated underground space of the study area. **3. Conceptual Model of Groundwater Flow** 

#### **3. Conceptual Model of Groundwater Flow 3. Conceptual Model of Groundwater Flow** Groundwater flows through aquifers consisting of surface soil and weathered layers at this mine site. Groundwater in the aquifers flows to mining areas through faults and veins as shown in

Groundwater flows through aquifers consisting of surface soil and weathered layers at this mine site. Groundwater in the aquifers flows to mining areas through faults and veins as shown in Figure 2 [42]. Groundwater in the mined area flows through the mining levels (−2 L, −3 L, and −5 L) to the adit mouth of the drainage level (−5 L). Thus, the total head of groundwater decreases from the ground surface to the deeper underground, and thereby duplicate aquifers, a shallower aquifer, and a deeper aquifer, are formed. Groundwater flows through aquifers consisting of surface soil and weathered layers at this mine site. Groundwater in the aquifers flows to mining areas through faults and veins as shown in Figure 2 [42]. Groundwater in the mined area flows through the mining levels (−2 L, −3 L, and −5 L) to the adit mouth of the drainage level (−5 L). Thus, the total head of groundwater decreases from the ground surface to the deeper underground, and thereby duplicate aquifers, a shallower aquifer, and a deeper aquifer, are formed. Figure 2 [42]. Groundwater in the mined area flows through the mining levels (−2 L, −3 L, and −5 L) to the adit mouth of the drainage level (−5 L). Thus, the total head of groundwater decreases from the ground surface to the deeper underground, and thereby duplicate aquifers, a shallower aquifer, and a deeper aquifer, are formed. The flow rate of AMD is about 4 to 10 m<sup>3</sup> /min from this mine. The AMD from the adit mouth of –5 L level accounts for about 0.04 to 0.16 m<sup>3</sup> /min.

**Figure 2. Figure 2.** Conceptual model of groundwater flow and its discharge. Conceptual model of groundwater flow and its discharge.

The flow rate of AMD is about 4 to 10 m<sup>3</sup> /min from this mine. The AMD from the adit mouth of −5 L level accounts for about 0.04 to 0.16 m<sup>3</sup> /min.

#### **4. Methods**

#### *4.1. In Situ Survey*

Groundwater levels were continuously measured at B-2 and B-3 from August 4, 2014 to December 17, 2016 at intervals of 60 min. The strainer pipes were installed from 15.0 to 35.0 m deep of B-2 well and from 3.0 to 15.0 m deep of B-3 well. B-2 is located less than 10 m away from B-3 and both wells are at the same ground level. This means that B-2 is to monitor the deeper groundwater level whereas B-3 is to monitor the shallower groundwater level. The location of these boreholes corresponds to upstream of groundwater into underground space, sensitive to the AMD production.

The flow rate of AMD was measured at M1 in Figure 1 using triangular weirs installed in the drain of the tunnel. The flow rate of AMD was measured from 4 August 2014 to 17 December 2016 at intervals of 60 min. The daily precipitation was calculated by accumulating hourly precipitations measured at the observatory 2 km away from the study area.

#### *4.2. Theoretical Equation*

Saturated–unsaturated groundwater flow analysis was applied to evaluate groundwater flow using Dtransu-3D-EL software [42,43] together with G-TRAN/3D pre- and post-processing softwares for Dtransu-3D [36,40]. Dtransu-3D-EL software solves the equation for saturated–unsaturated groundwater flow derived from the mass preservation and Darcy's equation, which can be written as

$$
\rho\_f \Theta \mathbf{y} \frac{\partial \mathbf{c}}{\partial t} + \rho [\mathbb{A} \mathbf{S} \mathbf{s} + \mathbb{C} \mathbf{s}(\boldsymbol{\theta})] \frac{\partial \boldsymbol{\varrho}}{\partial t} = \frac{\partial}{\partial \mathbf{x}\_i} \langle \rho \mathbf{K}\_{ij}^S \mathbf{K}\_r(\boldsymbol{\theta}) \frac{\partial \boldsymbol{\varrho}}{\partial \mathbf{x}\_j} + \rho \mathbf{K}\_{\mathbb{G}}^S \mathbf{K}\_r(\boldsymbol{\theta}) \rho\_r \rangle \tag{5}
$$

where ϕ is the pressure head, θ is the (volumetric) water content, *Ss* is the specific storage, *Cs*(θ) is the specific water capacity, *K S ij* is the directional components of the saturated hydraulic conductivity function, *K*r is the relative hydraulic conductivity, *t* is time, ρ*<sup>f</sup>* is the density of solvent, ρ is the density of fluid, ρ*<sup>r</sup>* is the ratio of ρ*<sup>f</sup>* to ρ, β = 1 is the saturated zone, β = 0 is the unsaturated zone, and γ is the solute density ratio [43].

#### *4.3. Numerical Model*

Basic configuration of the numerical model is shown in Table 1. The model domain had an area of 0.58 km<sup>2</sup> with a total elevation of 520 m and bounded by topographic ridges as shown in Figure 3. The mined area above the mining levels (−2 L, −3 L, and −5 L levels) was assumed to be in an unsaturated zone. The ground surface topography was reproduced by using the digital elevation model (DEM) created based on numerical maps by aerial survey. Ore body, mining levels (−2 L, −3 L, and −5 L) and excavated area (backfilled with sandy slime) were constructed in the mesh diagram of the numerical model based on the handwritten ore map and level map with 1:3000 scale. First, the ground plane was divided into squares with a side of 20 m. Next, a two-dimensional square on the ground surface were extended in the underground direction to form quadrangular prisms. Quadrangular prisms were added downward to create a three-dimensional mesh model.

**Table 1.** Basic configuration of the numerical model.

**Figure 3.** Simulation model of three-dimensional groundwater flow. **Figure 3.** Simulation model of three-dimensional groundwater flow.

The hydraulic conductivities (*K*) and unsaturated properties of numerical blocks representing the surface soil, embankment, and Neogene strata as well as the excavated areas were based on either in situ measured results or estimated from the results of other papers as listed in Table 2. Hydraulic conductivities were obtained from in situ permeability tests or estimated from rock permeability data collected throughout Japan [44]. The hydraulic conductivities (*K*) and unsaturated properties of numerical blocks representing the surface soil, embankment, and Neogene strata as well as the excavated areas were based on either in situ measured results or estimated from the results of other papers as listed in Table 2. Hydraulic conductivities were obtained from in situ permeability tests or estimated from rock permeability data collected throughout Japan [44].

The properties of various geological strata under unsaturated conditions were obtained by the van Genuchten model [45], which are given by The properties of various geological strata under unsaturated conditions were obtained by the van Genuchten model [45], which are given by

$$S\_{\varepsilon}(\boldsymbol{\upvarphi}) = \frac{\boldsymbol{\Theta} - \boldsymbol{\Theta}\_{r}}{\boldsymbol{\Theta}\_{s} - \boldsymbol{\Theta}\_{r}} \in \left[\frac{1\_{\text{pr}}}{1 + |d\boldsymbol{\upvarphi}|^{n}}\right]^{m} \tag{6}$$

$$\left\{\boldsymbol{\upleft}\right\}\right\} \qquad \left[\begin{array}{cccc} \left(\boldsymbol{\upleft} & \boldsymbol{\upleft}^{\boldsymbol{\upleft}} \end{array}\right) & \right]\right]^{2}$$

degree of pore connectivity (no unit). The van Genuchten parameters used for the current simulations are presented in Table 3, and the corresponding functions are depicted in Figure 4.

$$\mathcal{K}\_{\mathbb{P}}\{\mathcal{S}\_{\varepsilon}(\boldsymbol{\varphi})\} = \mathcal{S}\_{\varepsilon}\left[1 - \left(1 - \mathbb{S}\_{\varepsilon}^{1/m}\right)^{m}\right]^{2} \tag{7}$$

where *S<sup>e</sup>* is effective water saturation, θ*<sup>s</sup>* is the saturated water content, θ*<sup>r</sup>* is the residual water content, *a*, *m*, and *n* are the van Genuchten parameters [45], and *l* is a parameter representing the degree of pore connectivity (no unit). The van Genuchten parameters used for the current simulations are presented in Table 3, and the corresponding functions are depicted in Figure 4. *Minerals* **2020**, *10*, x FOR PEER REVIEW 6 of 13


**Table 2.** Hydraulic conductivities of different geological strata. **Table 2.** Hydraulic conductivities of different geological strata.

**Table 3.** Residual water content θ*r*, saturated water content θ*s*, and van Genuchten parameters *a*, *l*, and *n* [45]. **Table 3.** Residual water content θ r, saturated water content θ s, and van Genuchten parameters *a*, *l*, and *n* [45].


**Figure 4.** Water content and hydraulic conductivity curves using the van Genuchten's model for sandy slime, tailings, and andesite. See Table 3 for the corresponding parameters of the van Genuchten model [45]. **Figure 4.** Water content and hydraulic conductivity curves using the van Genuchten's model for sandy slime, tailings, and andesite. See Table 3 for the corresponding parameters of the van Genuchten model [45].

The distribution of mined areas and tunnels (−2 L, −3 L, and −5 L levels) in the numerical model was based on the records of the operating period. At that time, the thickness of the mining area was set at 3 m based on the actual space size. Similarly, the height and width of mining levels were set at 2 m. The average rainfall in August 2014 was 13.5 mm/day (hereafter, days with high rainfall) and the average in August 2015 was 3.5 mm/day (hereafter, days with little rainfall). The infiltration rate was calculated based on the water balance analysis, days with high rainfall: 5.0 mm/day (recharge rate 0.37) and days with little rainfall: 0.47 mm/day), (recharge rate 0.13). Both mining level (−2 L level) and drainage level (−5 L level) were assumed to be seepage boundary The distribution of mined areas and tunnels (−2 L, −3 L, and −5 L levels) in the numerical model was based on the records of the operating period. At that time, the thickness of the mining area was set at 3 m based on the actual space size. Similarly, the height and width of mining levels were set at 2 m. The average rainfall in August 2014 was 13.5 mm/day (hereafter, days with high rainfall) and the average in August 2015 was 3.5 mm/day (hereafter, days with little rainfall). The infiltration rate was calculated based on the water balance analysis, days with high rainfall: 5.0 mm/day (recharge rate 0.37) and days with little rainfall: 0.47 mm/day), (recharge rate 0.13). Both mining level (−2 L level) and drainage level (−5 L level) were assumed to be seepage boundary conditions.

conditions. The boundary conditions of the numerical simulations are as follows:


The flow chart of the simulation is the same as Nishigaki (1995) [49]. **5. Results and Discussion** 

#### **5. Results and Discussion** *5.1. Monitored Data*

#### *5.1. Monitored Data* The groundwater level of B-2 in the deeper aquifer was almost constant at the elevation of

The groundwater level of B-2 in the deeper aquifer was almost constant at the elevation of about 287 m while that of the B-3 in the shallow aquifer was changed in response to rainfall. This means that the groundwater level of B-3 is sensitive to rain whereas that of B-2 is not so sensitive to rain. about 287 m while that of the B-3 in the shallow aquifer was changed in response to rainfall. This means that the groundwater level of B-3 is sensitive to rain whereas that of B-2 is not so sensitive to rain.

The AMD flow rate is shown in Figure 5. The maximum value was 0.16 m<sup>3</sup> /min in April during the snowmelt season. Although the AMD flow rate tended to increase during the snowmelt season, the AMD during the summer varied from year to year, high in 2014 and low in 2015. This means that the infiltration of rainwater directly affects the flow rate of the AMD. The AMD flow rate is shown in Figure 5. The maximum value was 0.16 m<sup>3</sup> /min in April during the snowmelt season. Although the AMD flow rate tended to increase during the snowmelt season, the AMD during the summer varied from year to year, high in 2014 and low in 2015. This means that the infiltration of rainwater directly affects the flow rate of the AMD.

**Figure 5.** Comparison between the simulated and measured acid mine drainage (AMD) flow rate of days with high (**right**) and little (**left**) rainfall. **Figure 5.** Comparison between the simulated and measured acid mine drainage (AMD) flow rate of days with high (**right**) and little (**left**) rainfall.

#### *5.2. Calibration of Hydraulic Conductivity of the Neogene Strata 5.2. Calibration of Hydraulic Conductivity of the Neogene Strata*

The study area is snowy in winter and snowmelt increases the amount of AMD in spring (March to May). However, the rain depends on the year during summer. Thus, the AMD flow rates in days with high rainfall (August 2014) and days with little rainfall (August 2015) were selected for calibration. In order to evaluate seasonal variations of AMD flow rates, analysis of steady state groundwater flow was performed for both days with high and little rainfall. The study area is snowy in winter and snowmelt increases the amount of AMD in spring (March to May). However, the rain depends on the year during summer. Thus, the AMD flow rates in days with high rainfall (August 2014) and days with little rainfall (August 2015) were selected for calibration. In order to evaluate seasonal variations of AMD flow rates, analysis of steady state groundwater flow was performed for both days with high and little rainfall.

The simulated AMD flow rates by changing the hydraulic conductivity of the Neogene strata ranging from 1.0 × 10−9 m/s to 9.4 × 10−8 m/s are compared with those measured rates as shown in Figures 5 and 6. Since measured hydraulic conductivity of the Neogene strata ranged from 1.0 × 10−9 m/s to 9.4 × 10−8 m/s, the hydraulic conductivity was parametrically changed in the simulation. The calculated AMD flow rate strongly depended on the hydraulic conductivity of the Neogene strata. The AMD flow rate was the highest in case of the highest hydraulic conductivity of 9.4 × 10−8 m/s, and lowest in case of the lowest hydraulic conductivity of 1.0 × 10−9 m/s. When the hydraulic conductivity of the Neogene strata was 9.2 × 10−9 m/s, the flow rate of AMD agreed with the measured ones during both days with high and little rainfall. This means that the same hydraulic conductivity of the Neogene strata is applicable to any season. The simulated AMD flow rates by changing the hydraulic conductivity of the Neogene strata ranging from 1.0 × 10−<sup>9</sup> m/s to 9.4 × 10−<sup>8</sup> m/s are compared with those measured rates as shown in Figures 5 and 6. Since measured hydraulic conductivity of the Neogene strata ranged from 1.0 × 10−<sup>9</sup> m/s to 9.4 × 10−<sup>8</sup> m/s, the hydraulic conductivity was parametrically changed in the simulation. The calculated AMD flow rate strongly depended on the hydraulic conductivity of the Neogene strata. The AMD flow rate was the highest in case of the highest hydraulic conductivity of 9.4 × 10−<sup>8</sup> m/s, and lowest in case of the lowest hydraulic conductivity of 1.0 × 10−<sup>9</sup> m/s. When the hydraulic conductivity of the Neogene strata was 9.2 × 10−<sup>9</sup> m/s, the flow rate of AMD agreed with the measured ones during both days with high and little rainfall. This means that the same hydraulic conductivity of the Neogene strata is applicable to any season.

**Figure 6.** Change of precipitation and AMD flow rate. **Figure 6.** Change of precipitation and AMD flow rate. **Figure 6.** Change of precipitation and AMD flow rate.

The sensitivity analysis of hydraulic conductivity of the ore body was also conducted to investigate whether the ore body distributed on the upper part of −5 L affected the flow rate of AMD. When the hydraulic conductivity of the ore body was reduced to 1.9 × 10−5 m/s, the AMD increased by 20%, and when the hydraulic conductivity was increased to 1.3 × 10−5 m/s, the AMD decreased by 10%. This indicates that the effect of hydraulic conductivity of the Neogene strata on AMD flow rate was more significant than that of the ore body. It was also found that the amount of AMD from the deeper −5 L decreased when the shallower ore body collected more groundwater. The sensitivity analysis of hydraulic conductivity of the ore body was also conducted to investigate whether the ore body distributed on the upper part of −5 L affected the flow rate of AMD. When the hydraulic conductivity of the ore body was reduced to 1.9 × 10−<sup>5</sup> m/s, the AMD increased by 20%, and when the hydraulic conductivity was increased to 1.3 × 10−<sup>5</sup> m/s, the AMD decreased by 10%. This indicates that the effect of hydraulic conductivity of the Neogene strata on AMD flow rate was more significant than that of the ore body. It was also found that the amount of AMD from the deeper −5 L decreased when the shallower ore body collected more groundwater. The sensitivity analysis of hydraulic conductivity of the ore body was also conducted to investigate whether the ore body distributed on the upper part of −5 L affected the flow rate of AMD. When the hydraulic conductivity of the ore body was reduced to 1.9 × 10−5 m/s, the AMD increased by 20%, and when the hydraulic conductivity was increased to 1.3 × 10−5 m/s, the AMD decreased by 10%. This indicates that the effect of hydraulic conductivity of the Neogene strata on AMD flow rate was more significant than that of the ore body. It was also found that the amount of AMD from the deeper −5 L decreased when the shallower ore body collected more groundwater.

Calculated groundwater levels are compared with observed ones in boreholes as shown in Figure 7. The measured groundwater levels agreed with the calculated ones, irrespective of seasons when the hydraulic conductivity of the Neogene strata of 9.2 × 10−9 m/s was used. The measured and simulated levels of B-3 increased during the days with high rainfall and decreased during the little rain days. The simulated groundwater level of B-2 had a difference of about 4 m between the days with high and little rainfall. However, the measured values of B-2 were constant during both seasons. This indicates that the groundwater flow model can simulate the shallower groundwater flow and not deeper groundwater flow well. Calculated groundwater levels are compared with observed ones in boreholes as shown in Figure 7. The measured groundwater levels agreed with the calculated ones, irrespective of seasons when the hydraulic conductivity of the Neogene strata of 9.2 × 10−<sup>9</sup> m/s was used. The measured and simulated levels of B-3 increased during the days with high rainfall and decreased during the little rain days. The simulated groundwater level of B-2 had a difference of about 4 m between the days with high and little rainfall. However, the measured values of B-2 were constant during both seasons. This indicates that the groundwater flow model can simulate the shallower groundwater flow and not deeper groundwater flow well. Calculated groundwater levels are compared with observed ones in boreholes as shown in Figure 7. The measured groundwater levels agreed with the calculated ones, irrespective of seasons when the hydraulic conductivity of the Neogene strata of 9.2 × 10−9 m/s was used. The measured and simulated levels of B-3 increased during the days with high rainfall and decreased during the little rain days. The simulated groundwater level of B-2 had a difference of about 4 m between the days with high and little rainfall. However, the measured values of B-2 were constant during both seasons. This indicates that the groundwater flow model can simulate the shallower groundwater flow and not deeper groundwater flow well.

**Figure 7.** Change of precipitation and groundwater levels at B-2 and B-3. **Figure 7.** Change of precipitation and groundwater levels at B-2 and B-3. **Figure 7.** Change of precipitation and groundwater levels at B-2 and B-3.

Figure 7 shows changes in groundwater levels of monitoring wells of B-2 and B-3. The simulated results agreed with the observed ones. This implies that the calibrated model is effective Figure 7 shows changes in groundwater levels of monitoring wells of B-2 and B-3. The simulated results agreed with the observed ones. This implies that the calibrated model is effective in evaluating groundwater levels of both rainwater-sensitive and rainwater-insensitive wells. Figure 7 shows changes in groundwater levels of monitoring wells of B-2 and B-3. The simulated results agreed with the observed ones. This implies that the calibrated model is effective in evaluating groundwater levels of both rainwater-sensitive and rainwater-insensitive wells.

in evaluating groundwater levels of both rainwater-sensitive and rainwater-insensitive wells. The simulated value of B-2 responds to the increase/decrease in rainfall with a water level difference of 3.8 m between the days with high rainfall and little rainfall. Even if the amount of rainfall is large, if it runs off the surface layer, it may not contribute to the rise in water level. Therefore, we conducted a sensitivity analysis in which the hydraulic conductivity of the surface The simulated value of B-2 responds to the increase/decrease in rainfall with a water level difference of 3.8 m between the days with high rainfall and little rainfall. Even if the amount of rainfall is large, if it runs off the surface layer, it may not contribute to the rise in water level. Therefore, we conducted a sensitivity analysis in which the hydraulic conductivity of the surface The simulated value of B-2 responds to the increase/decrease in rainfall with a water level difference of 3.8 m between the days with high rainfall and little rainfall. Even if the amount of rainfall is large, if it runs off the surface layer, it may not contribute to the rise in water level. Therefore, we conducted a sensitivity analysis in which the hydraulic conductivity of the surface soil was increased by 10%. As a

soil was increased by 10%. As a result, the water level difference between days with high rainfall

result, the water level difference between days with high rainfall and little rainfall was 3.7 m. Although it had the effect of suppressing the increase/decrease in water level difference, it was small. and little rainfall was 3.7 m. Although it had the effect of suppressing the increase/decrease in water level difference, it was small.

#### *5.3. Prediction of the E*ff*ect of Backfilling 5.3. Prediction of the Effect of Backfilling*

Figure 8 shows the vertical distribution of total head and Darcy velocity along the drainage tunnel simulated by 3-D groundwater flow model. Before backfilling the underground space, the total head decreased from the ground surface to the mining level (−2 L level) as shown in Figure 8b. Both groundwater surfaces of the shallower and lower levels around the mining level (−2 L level) were clearly observed in the simulation (Figure 8b). The AMD flow rate exceeded 1.0 m/day as Darcy flow velocity around the mining level (−2 L) (Figure 8c). This flow rate corresponds to 0.091 m<sup>3</sup> /min of AMD produced. After backfilling the mining levels (−2 L, −3 L, and −5 L), AMD flow rate was calculated at 0.059 m<sup>3</sup> /min, 64.9% of AMD flow rate before backfilling, during days with high rainfall (35.1% reduction). In days with little rainfall, AMD flow rate was calculated at 0.059 m<sup>3</sup> /min after backfilling, which was 95.1% of the flow rate before backfilling (4.9% reduction). This indicates that the AMD flow rate was reduced after backfilling whether it was the rainy season or not. In particular, backfilling the underground space was more effective during the rainy season. On the other hand, the groundwater level rose after the mined area was backfilled (Figure 8b). In addition, the groundwater flow around the −2 L level was reduced after the backfilling (Figure 8c). Higher groundwater level could mitigate the reaction between atmospheric oxygen and sulfide minerals in the underground. Thus, the rise in groundwater level may prevent AMD formation. If DO (Dissolved Oxygen) is consumed in the saturated zone, there is no new supply of oxygen and the oxidation reaction of pyrite does not occur. The geochemical modelling like previous theoretical model and experiments [50–52] is necessary to estimate the amount of acidification. Figure 8 shows the vertical distribution of total head and Darcy velocity along the drainage tunnel simulated by 3-D groundwater flow model. Before backfilling the underground space, the total head decreased from the ground surface to the mining level (−2 L level) as shown in Figure 8b. Both groundwater surfaces of the shallower and lower levels around the mining level (−2 L level) were clearly observed in the simulation (Figure 8b). The AMD flow rate exceeded 1.0 m/day as Darcy flow velocity around the mining level (−2 L) (Figure 8c). This flow rate corresponds to 0.091 m<sup>3</sup> /min of AMD produced. After backfilling the mining levels (−2 L, −3 L, and −5 L), AMD flow rate was calculated at 0.059 m<sup>3</sup> /min, 64.9% of AMD flow rate before backfilling, during days with high rainfall (35.1% reduction). In days with little rainfall, AMD flow rate was calculated at 0.059 m<sup>3</sup> /min after backfilling, which was 95.1% of the flow rate before backfilling (4.9% reduction). This indicates that the AMD flow rate was reduced after backfilling whether it was the rainy season or not. In particular, backfilling the underground space was more effective during the rainy season. On the other hand, the groundwater level rose after the mined area was backfilled (Figure 8b). In addition, the groundwater flow around the −2 L level was reduced after the backfilling (Figure 8c). Higher groundwater level could mitigate the reaction between atmospheric oxygen and sulfide minerals in the underground. Thus, the rise in groundwater level may prevent AMD formation. If DO (Dissolved Oxygen) is consumed in the saturated zone, there is no new supply of oxygen and the oxidation reaction of pyrite does not occur. The geochemical modelling like previous theoretical model and experiments [50–52] is necessary to estimate the amount of acidification.

**Figure 8.** *Cont.*

*Minerals* **2020**, *10*, x FOR PEER REVIEW 10 of 13

(**c**)

**Figure 8.** Vertical section of simulation model and simulated results of groundwater flow. (**a**) Vertical section of simulation model. (**b**) Distribution of total heads before (left) and after backfilling (right). (**c**) Distribution of Darcy flow velocity before (left) and after backfilling (right). **Figure 8.** Vertical section of simulation model and simulated results of groundwater flow. (**a**) Vertical section of simulation model. (**b**) Distribution of total heads before (left) and after backfilling (right). (**c**) Distribution of Darcy flow velocity before (left) and after backfilling (right).

The measured amounts of AMD showed a difference of about twice in the days with high rainfall and the days with little rainfall as shown in Figure 8 and 1.5 times in the simulation. The amount of AMD in the days with high rainfall and that in the days with little rainfall were almost the same after backfilling because AMD was not significantly affected by the infiltration of The measured amounts of AMD showed a difference of about twice in the days with high rainfall and the days with little rainfall as shown in Figure 8 and 1.5 times in the simulation. The amount of AMD in the days with high rainfall and that in the days with little rainfall were almost the same after backfilling because AMD was not significantly affected by the infiltration of rainwater.

rainwater. The obtained results imply that backfilling underground space is effective in reducing the flow rate of AMD and preventing AMD formation is also expected. The backfilling method is promising when inclined water-conductive zones, such as excavated underground space and fractures, are located in the mine. Based on the outcome of this study, an appropriate design of backfilling underground space can be performed to mitigate the AMD in closed and abandoned mines. The effects of backfilling underground space on AMD production depend on the geology and The obtained results imply that backfilling underground space is effective in reducing the flow rate of AMD and preventing AMD formation is also expected. The backfilling method is promising when inclined water-conductive zones, such as excavated underground space and fractures, are located in the mine. Based on the outcome of this study, an appropriate design of backfilling underground space can be performed to mitigate the AMD in closed and abandoned mines. The effects of backfilling underground space on AMD production depend on the geology and hydrogeology of mine sites, so applying this method to a variety of sites is required to quantitatively evaluate the performance.

#### hydrogeology of mine sites, so applying this method to a variety of sites is required to quantitatively evaluate the performance. **6. Conclusions**

advice, and cooperation during this study.

publish the results.

**6. Conclusions**  3-D numerical model consisting of a variety of geological formations and underground tunnels was constructed. By using the calibrated model, both groundwater levels around the mine and flow rates of AMD agreed with the measured values, irrespective of the season. In this mine, AMD after backfilling the underground space was reduced to more than 30% in days with high rainfall and to 5% in days with little rainfall. In addition, the acidification of groundwater may be expected due to 3-D numerical model consisting of a variety of geological formations and underground tunnels was constructed. By using the calibrated model, both groundwater levels around the mine and flow rates of AMD agreed with the measured values, irrespective of the season. In this mine, AMD after backfilling the underground space was reduced to more than 30% in days with high rainfall and to 5% in days with little rainfall. In addition, the acidification of groundwater may be expected due to the rise in groundwater levels. These results imply that backfilling the underground space is effective in reducing AMD in this mine.

the rise in groundwater levels. These results imply that backfilling the underground space is effective in reducing AMD in this mine. **Author Contributions:** Conceptualization, K.Y. and S.T.; software, K.Y. and S.T.; validation, K.Y., S.T. and **Author Contributions:** Conceptualization, K.Y. and S.T.; software, K.Y. and S.T.; validation, K.Y., S.T. and S.Y.; investigation, M.E.; writing—original draft preparation, K.Y. and S.T.; writing—review and editing, T.I.; visualization, K.Y. and S.T.; supervision, M.S. All authors have read and agreed to the published version of the manuscript.

S.Y.; investigation, M.E.; writing—original draft preparation, K.Y. and S.T.; writing—review and editing, T.I.; visualization, K.Y. and S.T.; supervision, M.S. All authors have read and agreed to the published version of the **Funding:** This work was supported in part by Development of advanced technology for mine drainage treatment from 2012 to 2014 of Ministry of Economy, Trade, and Industry.

manuscript. **Funding:** This work was supported in part by Development of advanced technology for mine drainage treatment from 2012 to 2014 of Ministry of Economy, Trade, and Industry. **Acknowledgments:** The authors wish to thank the editor and anonymous reviewers for their constructive comments that improved this manuscript. We would also like to thank the staffs of Mitsubishi Materials Corporation, Eco-Management Corporation, and Mitsubishi Materials Techno Corporation for their help, advice, and cooperation during this study.

**Acknowledgments:** The authors wish to thank the editor and anonymous reviewers for their constructive comments that improved this manuscript. We would also like to thank the staffs of Mitsubishi Materials Corporation, Eco-Management Corporation, and Mitsubishi Materials Techno Corporation for their help, **Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

**Conflicts of Interest**: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
