4.5. Cross-Correlations
Table 4 illustrates the cross-correlation of time series for D33 (corresponding to the discharge in drain 33, located near section I, as shown in
Figure 1) and NBL (corresponding to the reservoir water level) during the dam’s elevation stages. The positive lags correspond to the cross-correlations of the variables of D33 versus NBL, the negative lags correspond to the cross-correlations of the variables of NBL versus D33. The zero lag is the instant correlation.
Considering that significant correlations are greater than 10%,
Table 4 presents five significant correlations (for lags 0, 1, 6, and 7), highlighted in red.
The tailings dam underwent seven upstream elevation stages (
Table 1). Thus, cross-correlations between monitoring instruments in these sections, starting from the second elevation phase of the dam, were modeled.
In sections H and I of the dam, drainage was carried out from the second elevation stage to the seventh elevation stage; in sections A and B, it was only carried out in the seventh elevation stage.
Table 5 shows the cross-correlations during the second elevation stage in section H utilizing drain D35 and the variables PLUV (rain gauge), NBL (reservoir water level), INA8H (water level meter 8H), and INA14H (water level meter 14H) to assess the drainage effect.
Table 6 displays the cross-correlation during the second elevation stage in section I using drain D33 and the variables PLUV (rain gauge), NBL (reservoir water level), INA15I (water level meter 15I), PZ6I (piezometer 6I), and PZ11I (piezometer 11I) for the drainage effect evaluation.
In both section H (
Table 5) and section I (
Table 6), drainage begins in the second elevation stage. In both cases, significant instantaneous cross-correlations predominantly occur, indicating the rapid response of the instruments at the onset of drainage.
Table 5 highlights relevant instantaneous cross-correlations between drain D35 and the reservoir water level (NBL), between D35 and the water level meter 8H (INA8H), and between D35 and the water level meter 14H (INA14H). In section H, apparently, the drain, piezometers, and water level meter inside the dam were responding to the water accumulated in the reservoir rather than to precipitation. This behavior is likely due to water accumulation during the first elevation stage, when the drainage system had not yet been implemented (
Table 3).
Table 6, on the other hand, shows significant instantaneous cross-correlations between drain D33 and the rain gauge (PLUV), in addition to those between D33 and the water level meter 15I (INA15I), as well as between D33 and the piezometers PZ6I and PZ11I. Drain D33 was probably releasing water from rainfall as well as from the reservoir. Additionally, in both sections, the lagged cross-correlations maintain a consistent pattern in terms of intensity and frequency, suggesting balanced drainage in both sections despite differences in the variables considered.
Table 7 and
Table 8 present the cross-correlations among the monitoring instruments in sections H and I, respectively, during the third elevation stage.
In section I, during the third elevation stage (
Table 8), instantaneous correlations predominate, except for the PLUV vs. D33 correlations and vice versa. The significant lagged cross-correlations occur at minimal frequencies and exhibit medium to low intensities, suggesting stability in this part of the dam.
In section H (
Table 7), both instantaneous and lagged cross-correlations are minimal in both intensity and frequency, which may indicate that drainage has been more effective in this section, contributing to greater overall stability compared to section I.
As shown in
Table 3, drained flux at this elevation stage was 0.36 L/s in section H and 0.17 L/s in section I, further supporting this interpretation.
Table 9 illustrates the cross-correlation during the fourth elevation stage in section H, utilizing, in addition to the instruments already mentioned in previous elevation stages—namely, drain D35, PLUV, NBL, INA8H, and INA14H (
Table 5 and
Table 7)—the INA26H (water level meter 26H), PZ5H (piezometer 5H), and PZ26H (piezometer 26H) to assess the drainage effect.
Table 10 presents the cross-correlations between the monitoring instruments in section I during the fourth elevation stage.
In the fourth elevation stage, in section H (
Table 9), instantaneous correlations do not predominate, indicating that drainage in this stage was very effective. In contrast, in section I (
Table 10), instantaneous correlations are present but with low values, suggesting that drainage was also effective in this stage. Notably, the lagged cross-correlations exhibit lower intensity and frequency compared to the second elevation stage in both sections, which may indicate that the dam became dehydrated due to drainage. This same trend is generally observed in the third elevation stage for both sections H and I. As shown in
Table 3, from the second elevation stage onward, drainage gradually decreases until the fourth elevation stage in both sections, further supporting this interpretation.
Table 11 presents the cross-correlations between the monitoring instruments in section H during the fifth elevation stage. In addition to the instruments already mentioned in previous elevation stages, namely, drain D35, PLUV, NBL, INA8H, INA14H, INA26H, PZ5H, and PZ26H (
Table 9), the water level meter 9H (INA9H) was considered to assess the drainage effect.
Table 12 presents the cross-correlations between the monitoring instruments in section I during the fifth elevation stage. In addition to the instruments already mentioned in previous elevation stages, namely, drain D33, PLUV, NBL, INA15I, PZ26I (
Table 6,
Table 8 and
Table 10), the water level meter 27I (INA27I) was considered to assess the drainage effect.
In the fifth elevation stage, section I (
Table 12) exhibits a different behavior to section H, as there are no significant instantaneous correlations except between D35 and PLUV and vice versa. This may be associated with the higher average precipitation recorded during this stage (126.6 mm) compared to 80.1 mm in the fourth elevation stage, as shown in
Table 3. The difference observed between section H (
Table 11) and section I (
Table 12) suggests that section H was more saturated with water, leading to more intense drainage.
Table 13 and
Table 14 present the cross-correlations among the monitoring instruments in sections H and I, respectively, during the sixth elevation stage.
In the sixth elevation stage, section H (
Table 13), there are no significant instantaneous correlations between drain D35 and the NBL and PLUV variables. However, significant correlations are observed between D35 and the water level meters (INA8H, INA9H, and INA14H), as well as between D35 and the piezometer PZ5H. Additionally, the presence of significant lagged correlations suggests that drainage is more associated with accumulated water in the dam rather than with rainfall-induced inflows, due to the lack of precipitation during this stage. As shown in
Table 3, rainfall was lower in the sixth elevation stage compared to the fifth and seventh elevation stages.
In the sixth elevation stage, section I (
Table 14), in general, maintains the same behavior as section H, which may also be the result of smaller precipitation of water in this stage than in the previous one.
Table 15 and
Table 16 illustrate the cross-correlations among the monitoring instruments in sections H and I, respectively, during the seventh elevation stage.
In sections H and I of the seventh elevation stage (
Table 15 and
Table 16) there are practically no instantaneous correlations, and the significant lagged correlations are of low frequencies and intensities, which shows that the dam is very stable, despite this being the rainiest period, according to
Table 3, with 135.5 mm on average.
As explained in item 2, during the seventh elevation stage, in addition to sections H and I, drainage was also carried out in sections A and B (
Figure 3). Apparently, the drainage in these two sections starts in these stages as these two sections are more distant from the recycling water dam than sections H and I, where drainage started in the second elevation stage (
Figure 3).
Table 17 illustrates the cross-correlation during the seventh elevation stage in section A utilizing drain D44 and the variables PLUV (rain gauge), NBL (reservoir water level), INA10A (water level meter 10A), and INA19A (water level meter 19A) to assess the drainage effect.
Table 18 presents the cross-correlation during the seventh elevation stage phase in section B utilizing drain D43 and the variables PLUV (rain gauge), NBL (reservoir water level), INA3B (water level meter 3B), INA20B (water level meter 20B), PZ29B (piezometer 29B), and PZ20B (piezometer 20B) to evaluate the drainage effect.
In section A (
Table 17) at the seventh elevation stage, there are significant instantaneous correlations between D44 and NBL, and between D44 and INA19 A, and the significant lagged correlations are well balanced, suggesting that the drainage is related to water inside the reservoir and inside the dam.
In section B (
Table 18), significant instantaneous correlations prevail between drain D43 and the water level meters (INA3B and INA20B), as well as between D43 and the piezometers (PZ20B and PZ29B). However, there are no significant instantaneous correlations between drainage and the NBL or PLUV variables, which may indicate that, in this section, drainage is not influenced by rainwater or the reservoir but rather by water confined within the dam. It is observed that the lagged cross-correlations are of the same intensities as the instantaneous drainage correlations and the INA and PZ variables, also the frequencies of the lagged cross-correlations are relatively stable. In the case of drainage and the NBL and PLUV variables, the significant lagged cross-correlations are of low frequency and intensity, which confirms that water release is fundamentally due to the interstitial water of the dam.
As shown by the analysis of the results presented in
Table 15 and
Table 16, which examine the cross-correlations in sections H and I, respectively, the tailings dam in its seventh elevation stage can be considered highly stable, even during the rainiest period, with an average precipitation of 135.5 mm (
Table 3).
This stability is largely attributed to the intense drainage carried out during this pre-decommissioning stage, not only in sections A and B but also in other areas not covered in this study (
Table 2). Notably, drainage rates in sections A and B reached 0.79 L/s and 0.86 L/s, respectively (
Table 3), the highest recorded values.