Study area.
The study area is located in the central part of Xingtai Mining Area (Fig. 1). The coalfield in this area is the Carboniferous-Permian concealed coalfield, and the surface of the study area is all covered by Cenozoic strata.
According to the borehole and roadway surveys, the strata developed in the mining field rank from old to new as follows: Majiagou Formation and Fengfeng Formation of the middle Ordovician system, the middle Benxi Formation and upper Taiyuan Formation of the Carboniferous system, the Lower Shanxi Formation, Lower Shihezi Formation and upper Shihezi Formation of the Permian system, and the Quaternary system. The Quaternary pore aquifer, Permian sandstone fissure aquifer, Carboniferous thin limestone fissure karst aquifer and Ordovician limestone fissure karst aquifer have deposited in the study area. Taiyuan Formation is the main coal-bearing stratum, where No.9 coal seam, with good stability, is one of the main minable coal seams. No.9 coal seam is relatively close to the underlying confined aquifer, with an average distance of about 15m. The underlying confined aquifer, the Benxi formation thin limestone fissure karst confined aquifer, is the research object in this study. There are great differences between different parts of the aquifer in terms of water-richness, karst and fissure are relatively developed in some parts, which are strongly hydraulically connected with the lower Ordovician limestone aquifer, posing a major threat to the mining of No.9 coal seam.
Data.
Aquifer thickness. The thickness of aquifers is one of the important factors influencing the water-richness of aquifers (Fig. 2a). Under normal circumstances, the thicker the aquifer, the greater the water content in the aquifer of unit thickness.
Permeability coefficient. as a significant hydrogeological parameter, is a commonly used index to indicate the permeability of rock strata (Fig. 2b). Permeability coefficient is not only determined by the properties of rock, e.g. particle size, composition, grain arrangement, filling condition, property and development degree of fractures etc., but is also related to the physical properties of fluid such as volume weight and viscosity. When the physical properties of regional groundwater are similar, the greater the permeability coefficient, the stronger the permeability of rock stratum. Given the same physical properties of groundwater in this area, the greater the permeability coefficient, the stronger the permeability of rock strata.
Flushing fluid consumption. Flushing fluid consumption is an important index indicating hydraulic properties of rock strata (Fig. 2c). Flushing fluid not only plays the function of cleaning and lubrication. The variation of flushing fluid consumption represents the lithology and permeability of rock strata. During geological drilling, flushing fluid consumption should be observed in real time. If flushing fluid consumption changes all of a sudden, the permeability of rock strata has changed. Hence, it is of great significance to study the water-richness of aquifers with flushing fluid consumption as one of the influencing factors.
Water level difference. The variation of water level elevation at the observation well during the underground dewatering test is one of the important indices to indicate the recharge intensity of the aquifer in the current area. If the water level at the observation well does not vary drastically during throughout the test, it means that the recharge conditions are favorable in this area, and when the coal mine water disaster occurs, there will be a large amount of water and the disaster will last for a long time; a drastic variation of the water level at the observation well implies poor recharge conditions in this area, which indicates that even if a coal mine water disaster occurs, the losses will be comparatively small (Fig. 2d).
Core recovery. Core recovery refers to the ratio of the total length of the core directly obtained in the borehole (with the length of broken core and weak mud deducted) to the drilling depth (Fig. 2e). It is stipulated that when calculating the core length, only hard and complete cores with a diameter of greater than 10cm should be calculated. Core recovery is a quality index to represent the integrity of rock mass. The lower the core recovery, the more smashed the rock is and the better its connectivity is.
Hydrochemical properties. During the dewatering test, the water in the target aquifer was sampled for the regular hydrogeologic analysis every 24 hours. The hydrochemical properties of groundwater in the discharge aquifer of the mining area and its hydraulic connections with the Ordovician limestone aquifer were explored and analyzed based on the change of the content of main ions in the water.
Methods
CRITIC Weight Method
In this study, the CRITIC weight method was used to determine the weight of each influencing factor on the water-richness of aquifers.15 Compared with the entropy weight method and the standard deviation method, the CRITIC weight method gives consideration to both the variability and conflict of indices, which, to a great extent, avoids the weight distortion caused by only considering the degree of confusion among indices.16,17 If the standard deviation between the indices is relatively large, these indices are highly variable and the index weight is comparatively high as well. If the correlation between the indices is weak, it means that these indices strongly conflict with each other, and the index weight is low. The CRITIC weighting model was established in the following specific steps:
Establish the evaluation index matrix. m evaluation indexes were given for n influencing factors. \({\text{r}}_{\text{i}\text{j}}\) represents the j-th evaluation index of the i-th factor, and the evaluation index matrix was expressed as:
$${\text{R}}_{\text{n}\text{m}}=\left[\begin{array}{ccc}{\text{r}}_{11}& \cdots & {\text{r}}_{1\text{m}}\\ ⋮& \ddots & ⋮\\ {\text{r}}_{\text{n}1}& \cdots & {\text{r}}_{\text{n}\text{m}}\end{array}\right]$$
Calculate the variability of evaluation indices. The variability of the j-th evaluation index was denoted by the standard deviation:
$$\text{E}\left({\text{r}}_{\text{j}}\right)=\frac{1}{\text{n}}\sum _{\text{i}=1}^{\text{n}}{\text{r}}_{\text{i}\text{j}}$$
$${{\sigma }}_{\text{j}}=\sqrt{\frac{\sum _{\text{i}=1}^{\text{n}}{({\text{r}}_{\text{i}\text{j}}-\text{E}\left({\text{r}}_{\text{j}}\right))}^{2}}{\text{n}-1}}$$
Where, \(\text{E}\left({\text{r}}_{\text{j}}\right)\)stands for the mean of the j-th evaluation index; \({{\sigma }}_{\text{j}}\)represents the standard deviation of the j-th evaluation index.
Calculate the conflict between evaluation indices. The conflict between the \({\text{j}}_{1}\)-th and the \({\text{j}}_{2}\)-th factors was expressed by the Pearson Correlation Coefficient (PCC):
$${{\rho }}_{{\text{j}}_{1}{\text{j}}_{2}}=\frac{\text{C}\text{o}\text{v}({\text{j}}_{1},{\text{j}}_{2})}{{{\sigma }}_{{\text{j}}_{1}}{{\sigma }}_{{\text{j}}_{2}}}=\frac{\text{E}\left({\text{r}}_{{\text{j}}_{1}}\bullet {\text{r}}_{{\text{j}}_{2}}\right)-\text{E}\left({\text{r}}_{{\text{j}}_{1}}\right)\text{E}\left({\text{r}}_{{\text{j}}_{2}}\right)}{\sqrt{\text{E}\left({{\text{r}}_{{\text{j}}_{1}}}^{2}\right)-{\text{E}}^{2}\left({\text{r}}_{{\text{j}}_{1}}\right)}\sqrt{\text{E}\left({{\text{r}}_{{\text{j}}_{2}}}^{2}\right)-{\text{E}}^{2}\left({\text{r}}_{{\text{j}}_{2}}\right)}}$$
$${\text{R}}_{{\text{j}}_{1}}=\sum _{{\text{j}}_{2}}^{\text{m}}(1-{{\rho }}_{{\text{j}}_{1}{\text{j}}_{2}})$$
Where, \({{\rho }}_{{\text{j}}_{1}{\text{j}}_{2}}\)refers to the correlation coefficient between evaluation indices \({\text{j}}_{1}\) and \({\text{j}}_{2}\); \({\text{R}}_{{\text{j}}_{1}}\) denotes the conflict of the \({\text{j}}_{1}\)-th evaluation index; \(\text{C}\text{o}\text{v}({\text{j}}_{1},{\text{j}}_{2})\)stands for the covariance between two evaluation indices.
Calculate the information quantity of evaluation indices. The information quantity \({\text{i}}_{\text{j}}\) of the j-th evaluation index was expressed as follows:
$${\text{I}}_{\text{j}}={{\sigma }}_{\text{j}}\bullet {\text{R}}_{\text{j}}$$
Calculate the weight of evaluation indices. The weight \({\text{W}}_{\text{j}}\) of the j-th index factor was expressed as follows:
Data Normalization Method
To eliminate the influences of evaluation indices in different dimensions on the evaluation results, the data must be normalized. The maximum (Eq. 1) or minimum (Eq. 2) method was used to normalize data about the evaluation indices. The maximum method was applied to normalize the influencing factors positively correlated with the water-richness of aquifers, and the minimum method was used to process the influencing factors negatively related to the water-richness of aquifers.
$${\text{Y}}_{\text{i}}=\frac{{\text{y}}_{\text{i}}-{\text{y}}_{\text{m}\text{i}\text{n}}}{{\text{y}}_{\text{m}\text{a}\text{x}}-{\text{y}}_{\text{m}\text{i}\text{n}}}$$
1
$${\text{Y}}_{\text{i}}=\frac{{\text{y}}_{\text{m}\text{a}\text{x}}-{\text{y}}_{\text{i}}}{{\text{y}}_{\text{m}\text{a}\text{x}}-{\text{y}}_{\text{m}\text{i}\text{n}}}$$
2
Where, \({\text{Y}}_{\text{i}}\) is the index value of the influencing factor nondimensionalized at point i; \({\text{y}}_{\text{i}}\) denotes the quantified index value of the influencing factor at point i; \({\text{y}}_{\text{m}\text{a}\text{x}}\) represents the maximum quantified index value of the influencing factor in the study area; \({\text{y}}_{\text{m}\text{i}\text{n}}\) stands for the minimum quantified index value of the influencing factor in the study area.
Water-richness Index Method
The water-richness pattern and distribution of aquifers are controlled by various factors, with a complicated control mechanism, diverse combination types and influencing conditions. Therefore, it is irrational to determine the water-richness of aquifers in the study area based on a single influencing factor. Based on the multi-source data fusion technology, the water-richness index (WI) method fuses multiple factors that influence the water-richness of aquifers, calculates the weight of each influencing factor, and superposes the constructed thematic maps of evaluation indices to obtain the WI index of aquifer of each evaluation unit in this study area. Natural Jenks, the standard classification method in ArcGIS, was then used to divide all the WI index data into 5 grades, namely weak, relatively weak, medium, relatively strong and strong. The measured data about the specific yield and water inrush points in the target aquifer in the study area were verified in the water-richness zoning map to finally gain the water-richness evaluation model of the target aquifer in the study area.
$$\text{W}\text{I}=\sum _{\text{i}=1}^{\text{m}}{\text{W}}_{\text{i}}{\text{f}}_{\text{i}}\left(\text{x},\text{y}\right)$$
Where, WI is the water-richness index of the aquifer; \({\text{W}}_{\text{i}}\) denotes the weight of influencing factors; \({\text{f}}_{\text{i}}\left(\text{x},\text{y}\right)\) refers to the single-factor influencing value function; \(\left(\text{x},\text{y}\right)\) represents the geographic coordinates.