4.2. Hydrochemical Analysis and Principal Component Analysis
Table 2 shows the results of the water chemistry testing of the mine inrush samples and potential end elements (including δ
18O, δD, Na
+, Ca
2+, Mg
2+, Cl
−, SO
42−, and total dissolved solids). The analysis results in
Table 2 show that the concentration of water ions and hydroxide isotopes in the mine has a large range, indicating that there are large differences between the water inrush points in the mines. The total dissolved solids value of the mine inrush far exceeded that of the seawater sample, indicating that the mine inrush contained a certain proportion of bedrock brine, which had an impact on the groundwater in the study area. Stable isotopes have a tracer effect, and from the level of hydrogen-oxygen isotopes, seawater has the largest mean value of hydrogen-oxygen isotopes, while rainwater has the lowest average of hydrogen-oxygen isotopes. This is because rainwater is formed by the evaporation of seawater and fresh water, which directly recharges freshwater, so rainwater is not used as a potential inrush end member. Furthermore, the Quaternary saline brine was recharged by seawater and freshwater, and its chemical characteristics were controlled by the mixing of seawater and freshwater. Thus, when seawater and fresh water are used as potential end elements, the contribution of Quaternary brine to the mixed model cannot be considered. By comprehensive analysis of the test results and the hydrogeological conditions of the study area, it was inferred that there were four potential inrush end elements of mine water: bedrock brine, seawater, freshwater, and Quaternary saline brine. It was found that the composition of seawater, freshwater, and saline water was relatively stable, and the ion concentration basically did not change. Many transgressions and regressions occurred in the study area, forming a variety of bedrock brines. Due to the similar formation conditions of bedrock brine in the study area, the characteristics of the three kinds of bedrock brine were not obvious, and further analysis was required to identify the source of mine inrush water. Mining needs to be considered as an important influencing factor. Mining causes stress redistributions of the surrounding rock, closing the original water-conducting fractures and creating new water-conducting fractures. Thus, even at the same sampling point, the inrush end element will change as the mining work progresses, so it is necessary to consider the brine evolution in the identification of mine water sources.
For the principal component analysis, the Kaiser–Meyer–Olkin (KMO) test and Bartlett test were used to measure the correlation of the data. The Kaiser–Meyer–Olkin value, calculated with the SPSS statistical analysis software, was 0.61, which is greater than 0.6. The
p-value of the Bartlett test was 0.000, which was less than 0.005, indicating that the data used in the paper are suitable for principal component analysis.
Figure 5 shows the results of the hydrochemical analysis and principal component analysis. The circle represents mine water, and the triangle represents potential end members.
Figure 5a shows the correspondence between Cl
− and δ
18O in all water samples. As the figure shows, seawater, fresh water, and bedrock brine are the apex of the area, which can be judged to be potential inrush end elements. However, it is not possible to distinguish the types of bedrock brines that play a major role in different regions. As bedrock brine with a low Cl
− concentration and low δ
18O is replaced by bedrock brine with a high Cl
− concentration and high δ
18O during the inrush source identification process, deep brine (for example) will be obscured by middle brine, and central brine will be obscured by shallow brine (the blue triangle and red triangle, respectively). In the process of bedrock brine identification, three factors need to be comprehensively considered: (1) the difference in the ion concentration of brine end members; (2) the spatial location and dominant flow channel of brine; and (3) the water temperature and its evolution at the sampling point.
Figure 5b shows the principal component analysis results of the main ions in the water sample, and the two principal components, principal component 1 and principal component 2, were obtained by reducing the dimensionality of the main ions and hydrogen-oxygen isotope analysis results. Since the characteristics of many key indicators will be “neutralized” in the process of dimensionality reduction, even though the main contributing end elements of the water inrush point can be preliminarily determined, it is impossible to accurately identify the brine end elements in the different sublevels. For example, the Euclidean distance between the inrush points of the −375 m, −510 m, and −600 m sublevels and the seawater is the closest, indicating that seawater has a high proportion in these water samples. However, the type of brine cannot be determined, so further analysis is required.
Figure 6 shows the correlation between Cl
−, Mg
2+, and Ca
2+ in all water samples. The Cl
− and Mg
2+ in the mine water of the shallow sublevels showed a good linear relationship, but the mine water in the deep sublevels deviated from this linear relationship, and its Mg
2+ concentration was lower than that in the shallow sublevels (as shown in
Figure 6a).
Figure 6b shows that the mine water in the deep sublevels has a high Ca
2+ concentration. From this, it can be inferred that the brine with a low Mg
2+ concentration and high Ca
2+ concentration (deep brine represented by 1140-4 and 1140-5) being mixed into the mine water, or ion exchange, occurs in the mine water during the seepage process, so the strength of the water–rock reaction needs to be evaluated. The red triangle and blue triangle in
Figure 6 are the middle brine represented by 960-DX and the shallow brine represented by 375-20, respectively. As shown in
Figure 6, the ion concentration and total dissolved solids of the shallow brine and middle brine are much higher than the deep brine, so deep brine has obvious differences from the other two brines. Thus, deep brine can be identified according to characteristic parameters such as ion concentration. The shallow brine is similar to the middle brine with little difference in the main characteristic parameters, but the shallow brine has a higher total dissolved solids, Mg
2+ concentration, and a lower SO
42−concentration, so it can be distinguished according to this characteristic and spatial location. In
Figure 6b, there is a good linear relationship between the Cl
− and Ca
2+ of shallow and middle mine water, and the mine water in the deep sublevels that is affected by deep brine deviates from this mixing line (i.e., the mixing line determined by seawater, saline water, middle brine, and shallow brine).
Figure 6b preliminarily identified the inrush end elements of mine water in the sublevels but was limited by the number of analysis indicators. This meant it was impossible to accurately identify the inrush end elements. Thus, the hierarchical multi-index analysis method was proposed to identify the mine water source.
4.3. Hierarchical Multi-Index Analysis
According to the distribution of the mine water, the study area was divided into three levels: shallow sublevels (−375 m and −510 m), middle sublevels (−600 m, −1005 m, −1035 m, −1050 m, and −1065 m), and deep sublevels (−1095 m and −1140 m). Under the influence of mining, the brine end elements of the same sampling point may change over time, so the mine water in middle and deep sublevels are plotted on the same map for comprehensive analysis. To identify the mine water sources, four pairs of indicators (Cl−–δ18O, Cl−–Ca2+, Cl−–Mg2+, Cl−–Na+) were analyzed by the hierarchical multi-index analysis method.
4.3.1. The Shallow Sublevel
In
Figure 6,
Figure 7 and
Figure 8, the triangle and circle represent the potential end element and mine water, respectively, and the difference in color indicates the type of water sample. The mixing lines and mixing boundaries in the figure indicate the mixing mode. When the potential end element is not clear, the dashed line is the mixing boundary, as is shown in
Figure 7a. When multiple end elements are on a straight line, the dashed line is a mixing line, as shown in
Figure 7b. In
Figure 7a, the δ
18O of the seawater, freshwater, shallow brine, and middle brine are all regional extremes, which are judged to be potential end elements. Meanwhile, the possibility of deep brine and saline water is excluded, but the type of brine cannot be determined.
Figure 7d shows that shallow brine is more likely a potential end element. In addition, when combined with the spatial location and dominant seepage channel of brine, it can be determined that shallow brine is the inrush end element. In
Figure 7c,d, the mine water in the red circle (375-4 and 375-15) deviates from the mixing line and is presumed to have undergone ion exchange. There are three main possible ion exchange methods in the mining area: CaX + 2Na = Na2X + Ca, MgX + 2Na = Na2X + Mg, and CaX + Mg = MgX + Ca. A good linear relationship between Cl
− and Na
+ (
Figure 7b) proves that no ion exchange occurs between Na
+ and Mg
2+, and Ca
2+. Thus, high Ca
2+ and low Mg
2+ concentrations in the 375-4 and 375-15 mine water indicate that ion exchange occurs between Mg
2+ and Ca
2+ (
Figure 7c,d). As shown in
Figure 7, the Euclidean distance between the mine water and the seawater in shallow sublevels was the smallest, so it can be assumed that the proportion of seawater is the highest in the end members. In summary, the mixed model of the shallow sublevels is a ternary hybrid model of seawater, freshwater, and shallow brine.
4.3.2. The Middle Sublevel
The middle brine represented by 960-DX is distributed in the middle sublevels of −960 m.
Figure 8a indicates that seawater and freshwater are potential end members, but due to the similar chemical characteristics of shallow brine and middle brine, it is impossible to determine the type of brine. In
Figure 8b–d, except for the mine water (1050-1, 1065-1) in the red circle, a good linear relationship between Cl
− and Na
+, Ca
2+, and Mg
2+ is presented; thus, the dashed lines in the figures are all mixing lines. It can be seen from
Figure 8c that the middle brine is more likely a potential end element. Moreover, the early mine water in the−600 m sublevels was high-temperature hot water (deep hot water from the F3 fault), which prompted speculation that F3 created a hydraulic link between the mine water of the −600 m sublevels and the middle brine [
25,
29]. According to the above analysis, it was determined that the end element in the middle sublevels is seawater, middle brine, and freshwater, but the mine water in the red circle does not conform to this law and needs to be further analyzed. The mine water in the red triangle and the red circle are water samples collected from the same sampling point at different times. The water samples in the red triangle were sampled in March 2022, and the water samples in the red circle were sampled in June and September 2022. Thus, the possibility of ion exchange was excluded; the essential reason for this was that the brine changed over time. The analysis results of
Figure 8c,d show that the brine type of the mine water in the red circle is deep brine. Under the action of mining disturbances, stress redistribution was generated in the surrounding rock, and the stress concentration and stress state changed to cause the original water-conducting fracture to close and the new water-conductive fracture to occur, which further led to the change in brine type. Therefore, in the process of identifying the water sources, the impact of mining activities should be fully considered. In summary, the mixed model of the middle sublevels is a ternary hybrid model of seawater, freshwater, and middle brine.
4.3.3. The Deep Sublevel
The deep mine-water data distribution is more discrete than those of shallow and middle mine water. There was no tendency toward the concentrated distribution of individual end elements, and most of them were distributed discretely along the mixing line (
Figure 9). A comprehensive analysis of
Figure 9 shows that freshwater, seawater, and deep brine are potential inrush end elements. In
Figure 9c,d, the mine water (from the −1095 m sublevel) in the red circle deviates from the mixing line. The mine water in the −1095 m sublevels has a high Mg
2+ concentration and a low Ca
2+ concentration, but its sum is constant. It was thus inferred that ion exchange occurs between Mg
2+ and Ca
2+, but the direction of ion exchange is opposite to the shallow sublevels. This may be due to differences in the lithology between the shallow sublevels and deep sublevels, or because the high-temperature conditions in the deep sublevels change the direction of ion exchange. In addition, the Euclidean distance between deep brine and mine water reflects that deep brine is the main source of the mine water. The mixed model of the deep middle mine water mainly includes deep brine, seawater, and freshwater.
The water source identification results show that the shallow, middle, and deep sublevels are all ternary mixing models. Seawater, brine, and freshwater are potential end members of mine water, and only the type of brine differs. The Euclidean distance between the mine water and the three end elements preliminarily indicates that the main sources of mine water are seawater and brine, which is consistent with the analysis results of
Figure 4.
4.4. Analysis on Water–Rock Reaction
In the process of water source identification, it was found that the mine water had a water–rock reaction during the seepage process. The water–rock reaction methods were, mainly, divided into three ways: rock dissolution, mineral precipitation, and ion exchange. Many experts have analyzed the ion activity of the mine water in the Sanshandao Gold Mine at different temperatures, and they have proved that there is no mineral precipitation and rock dissolution in the mine water [
30]. The possibility of ion exchange is evaluated by analyzing the correspondence between the ions (
Figure 10). Cl
− is stable in most environments, so changes in Cl
− concentration are controlled by mine water mixing.
Figure 10a indicates that no ion exchange has occurred between the Na
+ and other ions. In
Figure 10b, there was a good linear relationship between the Cl
− and the sum (molar concentration) of Ca
2+ and Mg
2+ in the mine water. However, the Ca
2+ and Mg
2+ of some mine water in the shallow and deep sublevels deviated from the mixing line. This phenomenon indicates that ion exchange occurs between the Ca
2+ and Mg
2+ in mine water, which is consistent with the research results of other studies [
8].
4.5. Mixing Ratio Calculation and Deviation Analysis
This article adopts a ternary mixing ratio calculation model based on mass conservation, which considers that the water inflow in the tunnel is composed of three types of end elements and assumes that the concentration of the three end elements is constant. The calculation formula is as follows:
Here, a1, a2, and a3 are the concentrations of analysis indicators a for water inrush end element 1, end element 2, and end element 3, respectively; b1, b2, and b3 are the concentrations of analysis indicators b for water inrush end element 1, end element 2, and end element 3, respectively; asp is the concentration of indicator a for the p-th water inrush sample; bsp is the concentration of indicator b for the p-th water inrush sample; and δ1, δ2, and δ3 are the proportion of each water inrush end-element of the p-th inrush sample.
The end-member mixing ratio of the mine water in the Sanshandao gold mine was calculated based on the water source identification result. As Cl
− and δ
18O were conserved under most conditions, and as the ion exchange had little effect on their concentration, Cl
− and δ
18O are used as the analytical indicators in the calculation of mixing ratios. The ternary calculation model was used to calculate the end-element mixing ratio of mine water, and the reliability and accuracy of the new method were evaluated based on the results of the deviation analysis. The model bias can be calculated according to the following formula:
where
represents the calculated concentration of the factor and
represents the measured concentration of the factor. The factor calculation concentration is obtained by linear weighted superposition according to the end element type and mixing ratio. A positive deviation indicates that the calculated value is greater than the measured value, while a negative deviation indicates that the calculated value is less than the measured value.
Gu et al. calculated the mixing ratio and deviation of the mine water in the −375 m, −510 m, and −600 m sublevels of the Xishan mining area [
25]. The reliability of the hierarchical multi-index analysis method was evaluated by comparing the deviation of the method adopted in this paper with a general method (
Table 3). As shown in
Table 3, the hierarchical multi-index analysis method greatly reduces the calculation bias of mine water, which proves that the hierarchical multi-index analysis method can effectively identify water sources and optimize the mixing model. Due to the large number of sampling points in the −600 m sublevels and the continuous sampling, the sublevels of −600 m were selected as the research object. The evolution trends of seawater proportion over time between the two research methods were compared. As shown in
Figure 11a, the proportion of the seawater calculated decreases with the mining process. However, the rate of reduction gradually slows down, and finally, the proportion of seawater is stable at 0.6~0.7. Other experts have analyzed the changes in the proportion of seawater in the early stages of mining in the Xinli mining area and found that the proportion of seawater showed a trend of increasing year by year [
23].
Thus, the analysis results of the two papers show that the proportion of seawater first increases and then decreases with the progress of mining. The reason for this is that in the early stages of mining, water conduction fractures germinated in the surrounding rock under the action of disturbance stress, while in the later stage of mining, some of the original water conduction fractures connected with seawater were closed under the action of ground stress. In addition, the proportion of seawater calculated by this study method is lower than that which was calculated by [
25]. Moreover, it can accurately reflect the state of water inrush points and can reasonably assess the risk of water inrush in mines. Two continuous monitoring points in the −375 m sublevels were selected to analyze the proportion change in the seawater. In 2019, the proportion of seawater in the −375 m sublevels were much higher, reaching 0.95, than those in the −600 m sublevels. However, following the progress of mining, the proportion of seawater fluctuated in the range of 0.6~0.8.
Figure 12a shows the changing trends of the mean seawater proportion in the sublevels of −375 m, −510 m, and −600 m. The variation trend of seawater proportion is the same in the three sublevels, which all gradually decrease with time. The curves of the −375 m and −600 m sublevels almost coincide, and the average seawater proportion of the −510 m sublevel is much lower than that of the other two sublevels. Since the water inrush points in the sublevels of −375 m and −600 m are affected by the F3 fault and the northwest water-conducting fissure zone, they are subject to the vertical recharge of seawater (
Figure 3). As the F3 fault is the largest water-conducting fault in the mining area, it is necessary to analyze its impact on mine water inrush. The inrush points in the −600 m sublevels were divided into five categories according to their distance from the F3 fault, and the typical inrush points in each category were selected to analyze the changing trends of the seawater proportions in 2022. With the increase in distance from the F3 fault, the overall trend of seawater proportion decreases. This indicates that the F3 fault has a significant impact on the proportion of seawater at the inrush point; thus, the monitoring of the inrush point near the F3 fault should be strengthened (
Figure 12b).
Under mining conditions, the type of brine at the same sampling point may change. The deviation calculation results show that the brine evolution model can effectively improve the calculation accuracy of the end-element mixing ratio (
Table 4).