**4. Result and discussion**

#### **4.1 Geochemical analysis**

A total of 16 water samples were collected from the study site, including 12 coal mine waters, two surface waters, and two carbonate waters, respectively. Concentrations of major ions are drawn in a piper plot (**Figure 2**). **Figure 2** suggests that the carbonate water and coal mine water belong to medium-mineralized water, and surface water belongs to low-mineralized water, respectively. The surface water is Na-Mg-Ca-Cl− -SO4 2−-HCO3 − -type water, the carbonate water is Na-Mg-Ca-SO4 2− type water, and the coal mine water is Na-Ca-SO4 2−-, Na-SO4 2−-, or Na-HCO3 − -type water, respectively. Coal mine waters showed characteristics of high-soluble minerals. [SO4 2−] of most coal mine water samples were higher than USEPA and Chinese highest limit, 250 mg/L. Besides [SO4 2−], [Cl− ], TDS, and hardness were also higher than the Chinese-regulated limit. The combination of higher levels of Ca2+, Mg2+, HCO3 − , and SO4 2− concentrations in the groundwater suggests that the coupled reactions involving sulfide oxidation and carbonate dissolution largely control the solute acquisition processes in the study area [52].

**Figure 2.** *Piper plot of the water samples.*

The PCA analysis is used to reduce the dimensions of the water matrix. In this study case, dimension means water parameters. Water samples are represented by 10s of conventional inorganic and organic parameters, some of which are an indicator of the environment and reaction pathways, and some others a redundant or collinear. The PCA method could solve problems of not only parameter redundant and collinear, but also shows principal components in the data matrix, and relationships between parameters and among the parameters and samples could also be shown by using the parameters' loading and samples' score, respectively.

In this study, the traditional method of PCA calculation was applied, and principal components and variance that the PC explained were calculated. In the original table, 16 parameters were tested, and the PCA calculation used 16 new components to represent the original parameters, which explain the variance of samples, in descending order. The head six components explained 29, 21, 17, 10, 9, and 5% of the variance, respectively. Considering the balance of more variance explained and less components, we chose two principal components to stand for the sample data. The GM method was used to group the ions and trace elements in the water sample, which is shown in **Figure 3**. The parameters were clustered into four groups. Group 1 includes K+ + Na+ and Cl− ; group 2 includes Ca2+, Mg2+, Cl− , SO4 2−, TDS, and hardness; group 3 includes HCO3 − , CO3 2−, and pH; group 4 includes As, Hg, Se, Cd, Pb, respectively. The samples were collected in or around the coal mine district, so the clustering result is representative, and the groups were separated from others distinctly. From the clustering result, it is suggested that group 2 stands for the dissolution of carbonate, and group 4 stands for the trace element. The trace element contaminant could be identified from this result.

#### **4.2 Leaching mechanism of trace elements from the coal host rock**

To investigate the leaching mechanism of trace elements from the coal host rock, both the rock sample and water sample were tested. The rock samples were those

**Figure 3.** *Loadings of the multivariate analysis and clustering result of water samples.*

#### *Leaching Mechanisms of Trace Elements from Coal and Host Rock Using Method of Data Mining DOI: http://dx.doi.org/10.5772/intechopen.100498*

collected from coal roof, which then was processed in a standard treatment to decide its content. The milled rock samples were mixed with deionized water in the batch experiments to observe and evaluate the leaching behavior and mechanism of the trace elements from rock to water. The major and trace element concentrations in host rock and leachate are listed in the Table 1 in Shan et al. [53]. A hypothesis was that the occurrence and leaching mechanism of the trace elements in the solid samples were related to their concentrations in the water samples. Therefore, the PCA was applied to reduce dimensions of the rock and water samples, and then, the analytical results of solid and liquid samples are discussed parallelly.

For the rock samples, 18 elements were tested, and then, the PCA method was applied. The first two components explained 91% of all variance; therefore, the two PCs were used to stand for information of the data. For the water samples, 16 ions and trace elements were tested. The same analytical process was applied. The first two PCs explained 87% of all variance, which were used to stand for information in the water samples. By using the new PCs, parameters were assigned loadings on every new component. Then, the parameters of rock and water samples can be drawn in a two-dimensional (2D) scatter diagram. **Figure 4** shows the elements of rock samples, and **Figure 5** shows the ions and elements of water samples in a 2D scatter diagram, respectively.

The PCA-treated data were clustered using the expectation maximization (EM) algorithm. The EM algorithm could make several clustering results. By considering the BIC score and conciseness of every clustering model, the parameters in the rock samples were clustered into three groups. The first group includes Mo, Pb, Cr, V, Ti, and Al, which are marked in solid circles; the second group includes Zn, Ba, Mn, Fe, Mg, As, Hg, Se, and Cd, which are shown in hollow squares; the third group includes Cu, Sr, and Ca, which are shown in solid triangles. As mentioned before, the clustering could help to analyze the elements' occurrence in solid samples. Cr has a high affinity of clay and ash yield in gangue [3]. Zhou et al. [2] reported a high relationship

**Figure 4.** *Loadings of the multivariate analysis and clustering result of rock samples.*

**Figure 5.** *Loadings of the multivariate analysis and clustering result of rock leachate.*

of Pb and Se and with Fe in gangue, so high-sulfide mineral affinity was observed. Zn and Cd were found to have a high association with pyrite and sphalerite. Xiong et al. [26] found that Cd is mainly in sulfide form in the coal host rock. As and Mo are mainly carbonate- and silicate-related form. Finkelman et al. [3] found that Mo, Pb, Cr, Ti, and Al are mainly in clay minerals, As, Hg, Cd, and Zn mainly occur in sulfide form, and Ca and Sr are mainly carbonate-related. The PCA analysis corroborates the previous studies. As the **Figure 5** shows, the first group stands for clay affinity elements, the second group stands for elements with sulfur-mineral affinity, and the third group stands for the carbonate-related elements.

The ions and trace elements in the rock leachate could be clustered into three groups, the first group includes Al, Si, Cr, Mn, Fe, Cd, and Pb; the second group includes Ti, V, As, Se, Mo, and Hg; and the third group includes Zn, Sr, and Ba, respectively. The coexisting pattern of ions and elements in the water are controlled not only the occurrence in rock, but also the water-rock interaction, and adsorption behavior. Therefore, the clustering result of solid and liquid results was not exactly the same. However, two results are comparable to find out certain or probable reaction mechanisms in the water-rock interaction pathway. The three groups clustered for the water samples can be compared with those of the solid samples. Therefore, a primary deduction could be made. The first group of elements in the water samples suggests the reaction pathway of clay reaction with water. When the clay mineral reacts with water, the transformation of illite to kaolinite could happen, and some minerals, such as Cr, could be released. Cd was clustered to the second group in the rock analysis but was clustered to group 1 in water analysis. The result could be explained by two reasons: first, Pb and Cd embedded in both sulfur minerals and clay minerals, and second, Pb and Cd were controlled not only by dissolution, but also by adsorption. When the water has a low pH value, metal elements tend to release, while they could be adsorbed in a higher pH environment. According to our observation, the concentration of Pb and Cd in the surface water in the coal mine district was evidently higher than that in the non-coal mine district. As, Hg, and Se have a similar

*Leaching Mechanisms of Trace Elements from Coal and Host Rock Using Method of Data Mining DOI: http://dx.doi.org/10.5772/intechopen.100498*

pattern in the solid and liquid samples. It is apparent that they were controlled by the dissolution of sulfur minerals. The content of the sulfur mineral in the rock was not high in our samples. However, the oxidation and dissolution processes were distinct, leading to the release of toxic trace elements.

#### **4.3 Leaching mechanism of trace elements from coal**

The major and trace element concentrations in coal and leachate are listed in the Table 1 in Shan et al. [53]. The same analytical method with rock was applied to the coal and coal leaching analysis. And the PCA and clustering analytical results of coal and coal leaching water are shown in **Figures 6** and **7**. Two principal components could explain 96 and 91% variance for the coal and leachate, respectively. As **Figure 6** shows that elements are clustered into four groups, the group 1 includes Mo, Pb, Cr, V, Cu, Ti, Al, Hg, and Se; group 2 includes Zn and Cd; group 3 includes Ba, Mn, Sr, Mg, and Ca; group 4 includes Fe and As, respectively. The ions and trace elements in coal leachate, as shown in **Figure 7**, were grouped into three groups. Group 1 includes Al, Se, and Pb; group 2 includes Si, As, Sr, Mo, and Hg; group 3 includes Ti, Cr, Mn, Fe, Zn, Cd, and Ba, respectively. Finkelman et al. [3] investigated the occurrence of most of the trace elements, it is found that 65% of Ti, 90% of Al, and 75% of Cr 25% and 30% of Cu and Mo are in clay minerals, little Pb and Se are in clay form, 75 and 65% of Zn and Cd formed in mono-sulfide form, and 70 and 90% of As and Hg are sulfide form. Pumure et al. [39] argued that As and Se usually occur in clay minerals. Pb was found to be sulfide form as pyrite and galena [54] and organic form [55].

Combining the literature review and PCA-clustering analysis, group 1 for the coal samples stands for clay affinity, groups 2 and 4 are sulfur-mineral elements, and group 3 is related to carbonate minerals. Group 2 has two elements, Zn and Cd. This result is consistent with some previous studies [2, 56]. It is concluded the main occurrence of trace elements: As, Hg, Cd occurred in sulfide minerals, and Pb, Cr, and Se occurred in clay minerals, respectively. Zn and Cd are the primary elements

**Figure 6.** *Loadings of the multivariate analysis and clustering result of coal samples.*

**Figure 7.** *Loadings of the multivariate analysis and clustering result of coal leachate.*

in sphalerite. Compared with the host rock, the sphalerite is more probably to form an independent mineral in coal.

The coal leachate clustering results were relatively different with that of the analytical results of coal. Compared to the rock samples, coal is a more complex matrix and consists of organic and mineral matter, the latter including crystalline minerals, non-crystalline mineraloids, and elements with non-mineral associations [55]. However, some patterns could be concluded. Group 1 includes Al, Se, and Pb, which is similar to group 1 in the coal analysis. Therefore, group 1 stands for the elements that originated from clay minerals. Group 2 stands for the elements related to sulfur-bearing minerals. As and Hg had similar behavior patterns in solid and liquid matrices. So the leaching product in water was mainly from the dissolution of its bearing mineral, the sulfide mineral. Similar to the host rock analysis, low content of sulfur-mineral may lead to trace element concentration.

The trace elements Se, Cr, and Pb have similar behavior patterns in solid and liquid matrices, suggesting a dissolution progress of its bearing minerals. According to the literature research and coexisting analysis, these elements usually occur in continental facies minerals, such as clay minerals.
