Multi-factor evaluation for the water yield properties of coal floor aquifers, based on GIS

Water hazards in mines threaten coal production safety in several countries. A good example is the highly confined Ordovician limestone aquifer in North China, where karst has developed intensively and aquifer yield distribution is highly anisotropic. The abundance of structures and the impacts of coal mining in the area are likely to lead to water inrush incidents. In recent years, several serious and severe water inrushes have occurred in North China, resulting in casualties and enormous financial losses.

Aquifer yield plays a crucial role with respect to water inflow in mines, and large water inrush accidents only occur in areas where yields are high. If aquifer yields are low, even in areas where pressures are high or where fault zones link aquifers, significant water inrush accidents cannot happen. Only a clear understanding of aquifer yield can help to reduce water hazards in mines. Consequently, many studies have been carried out into aquifer yields. Meyer & Winstanley (2003) predicted groundwater yield for the Cambrian–Ordovician aquifer in northeastern Illinois using uncertainty parameters in inferred conceptual models. Zhao & Chang (2007) divided a middle Ordovician aquifer in the Handan-Fengfeng mining area into four different yield zones according to the degree of development of karst. Bwalya (2011) estimated the specific yield of an aquifer in Idaho, USA, by alternative linearization of water table kinematic conditions. Wang et al. (2012) evaluated the yield of limestone in the Upper Taiyuan Formation in the Cheji coal mine, on the basis of specific yield. Peng et al. (2014) predicted the yield of a sandstone aquifer at the base of the upper Shihezi formation, in the Permian of the Wolong Lake coal mine, using fuzzy clustering, and comprehensive prediction. Their predictions were based on the characteristics of the lithologic and geological structure. After selecting evaluation indices, Liu et al. (2014) used a probabilistic neural network to predict the yield of the sandstone aquifer in the coal seam roof at the Xieqiao coal mine.

Although recent research into the yield of aquifers in coal mines has produced significant achievements, there are drawbacks including incomplete evaluation indices, difficulty in extracting index data and insufficient graphic representation. Because of this, this research into the Ordovician limestone aquifer of the Danhou coal mine in Hebei, China, was based on evaluation indices for coal-seam floor aquifer yield. These indices were used to build a comprehensive evaluation model for the aquifer's yield. Geographic information system (GIS) technology was integrated into the processing, storage and graphic representation of each index datum, to realize the evaluation analysis of the Ordovician aquifer. This method can effectively reflect the fact that coal-seam floor aquifer yield is influenced by many factors. At the same time, the evaluation results are objective and visible, providing a theoretical basis for aquifer yield research.

The stratigraphic structure in this area is well developed beneath the Coal Measures strata. It includes the Sanggan group (Archaean), and the Changcheng, Jixian, Paleozoic-Cambrian and Ordovician systems of the middle-upper Algonkian. The Coal Measures strata belong to the Xiahuayuan formation in the lower-middle series of the Jurassic. There are Jiulongshan, Tiaojishan, and Houcheng formation strata of the Middle Jurassic, Zhangjiakou formation strata of the Upper Jurassic, and Tertiary and Quaternary strata overlying the Coal Measures.

Danhou mine is in the southern part of the central mining area of Yuxian, and its structural configuration is determined by the paleo-topography of the basement and the amplitude of crustal subsidence. Of the fractures developed in the mining area, three groups distributed around the edge are aligned, respectively, approximately south/north, east/west and northeast/southwest. Regional dip is quite variable, with a range between about 5 and 15°.

The research area can be divided, effectively, into four aquifer zones. The separate, karst-fracture confined aquifers of the Cambrian and Ordovician, the pore-fissure confined aquifer of the Jurassic, and the overlying, unconsolidated, porous Quaternary aquifer. The main aquifer threatening safety in Danhou coal mine is the Ordovician karst-fracture confined limestone, which, as the basement of the Coal Measures, underlies the coal seam floor directly. The principal lithology includes a number of limestones that are, variously: gray, dolomitic and brownish-yellow, aphanitic, and leopard-like, interlayered with thin celadon marls. Karst developed dissolution pores and corrosion fissures are accompanied by significant numbers of paleo-caverns. The aquifer's specific yield and permeability coefficient are, respectively, between about 0.02 L/s/m and 2.3 L/s/m, and 0.5 m/d and 7.2 m/d.

The research was carried out from the perspective of the representativeness of the indices and convenient data acquisition of data, and based on national regulations and previous research. Specific yield, permeability, drilling fluid consumption, aquifer thickness, and fault and fold distribution were selected as key indices for aquifer yield evaluation.

• (1) Specific yield

  • Specific yield is the yield of a well when the pumped drawdown is 1 m (Li et al. 2014). This is information that can be obtained easily in pumping tests in coal mines, and also important evidence for determining the water yield grade in Provisions for Mine Water Control (State Administration of Work Safety Supervision 2009). The higher the specific yield, the greater the yield from the aquifer.

• (2) Aquifer thickness

  • Thickness is the most intuitive factor influencing aquifer yield. The greater the thickness of a block, the larger the unit volume of storage, and, hence, the higher aquifer yield. Data regarding the thickness of an aquifer can be obtained from geological drilling logs in mining areas.

• (3) Permeability coefficient

  • The permeability coefficient is a constant characterizing a rock's permeability. A high permeability coefficient implies better interconnection between aquifers, better rock permeability, and higher aquifer yield.

• (4) Drilling fluid consumption

  • In drilling, the consumption of fluid (down-hole) can reflect the permeability of the strata, which is determined by the geological conditions. Variations in the quality and consumption of drilling fluid, indicate changes in the permeability and leakage of individual strata (Wu et al. 2009). Thus, it is an important indicator of the hydraulic properties of drilled strata and adopted as important evidence of aquifer yield.

• (5) Fault and fold distribution

  • Generally speaking, the fracture zones of faults and the axes of folds display diverse lithologies. Fractures develop intensively and are likely to become water conducting channels, collecting surrounding groundwater and forming high yield zones. The distribution of faults and folds is, thus, an important index for judging potential aquifer yield.

W_k is determined by the analytic hierarchy process (AHP) suggested by the American operational researcher Saaty (1977). It is a multifactorial decision method combining both qualitative and quantitative analysis. Its basis is the decomposition of a complex problem into several levels and factors, so that the relative importance of two indexes can be determined. After the judgment matrix has been established, the weights (degrees) of importance of different schemes are obtained by calculating its maximum eigenvalue and corresponding eigenvector (Saaty 1977, 1980). The specific steps are:

• (1) Establishing the analytic hierarchy model

  • The major influencing factors for each research object were determined according to the basis of judgment and system analysis concerning it. The sub-factors determining each major factor were then resolved. Thus, the factors that interact are gradually acquired, and a hierarchical relationship is constructed from top to bottom.

• (2) Constructing the judgment matrix

  • If the influences of n factors X = {x₁, x₂, x₃ … x_n} on factor Z are to be compared, Saaty (1980) propose establishing a pairwise comparison matrix, comparing all factors in pairs. Thus, two factors x_i and x_j are selected each time, and a_ij represents the influence ratio of x_i to x_j on Z. All of the results, A = (a_ij)_n×n in the comparison, are exhibited as a matrix, called the ‘judgment matrix’. With respect to determination of a_ij, Saaty et al. (1980) proposed that the scales be represented by the numbers 1 to 9 and their reciprocal. The connotations of the scales 1–9 are listed in Table 1.

• (3) Weight solution and consistency check

  • The weightings are calculated first by determining the maximum eigenvalue λ_max of the judgment matrix, A, and then, using AW = λ_maxW, the corresponding eigenvector, W, of λ. The values of W were normalized so that they were between 0 and 1. Establishing the judgment matrix helps to reduce interference from other factors and reflects the difference between the impacts of a pair of factors objectively. However, as with all comparisons, there is inevitably some inconsistency and so this needs to be checked. This is done by:

  • 1. Calculating the consistency index, CI

    

    2

  • 2. Searching for the corresponding mean random consistency index RI. For n (n = 1, 2 …, 9), the values of RI, as given by Saaty (1980), are shown in Table 2:

  • 3. Calculating the consistency ratio, CR

    

    3

  • When CR < 0.1, the consistency of the judgment matrix is acceptable; otherwise, modification is required. The AHP judgment matrix (Table 3) for evaluating the yield of the coal-seam floor aquifer in this area was established on the basis of the Table 1. Subsequently, the factor weights were calculated using the judgment matrix and these are shown in Table 4.

As can be seen, the value of the mean random consistency index, CR, is 0.01308, which is <0.1. The judgment matrix therefore satisfies the consistency check. The influencing factor weights obtained are shown in Table 4.

The red area in Figure 8 is that with high aquifer yield. This arises from the great thickness of aquifer (65 to 105 m), and its high specific yield (1.3 to 2.3 L/s/m) and permeability coefficient (5.6 to 7.2 m³/d). The orange-red area has slightly lower yields than the red area. Here, the relatively large aquifer thickness (50 to 80 m), specific yield (0.8 to 2.3 L/s/m) and permeability coefficient (4.2 to 5.7 m³/d) play dominant roles in the water yield of the southwest zone; while the water yield of the northeast zone is determined by the high permeability coefficient (5.6 to 7.27 m³/d) and consumption of drilling fluid (0.3 to 0.39 m³/d). The dark green zone denotes the low yield area arising from low values of aquifer thickness (25 to 26 m), specific yield (0.02 to 0.05 L/s/m) and permeability coefficient (0.7 to 2.5 m³/d). So, this area has the lowest water yield in the evaluation area. The lighter green zone represents a relatively low yield area arising from the relative thinness of the aquifer (27 to 40 m), and low specific yield (0.05 to 0.36 L/s/m), permeability coefficient (0.7 to 2.5 m³/d) and consumption of drilling fluid (0.14 to 0.29 m³/d). The yellow area has moderate water yield, and is the transition zone between the relatively high- and low- yielding areas. The faulted and folded areas are found mainly in the red zones, where yields are high. Value ranges of the factors in the different aquifer yield zones are shown in Table 5.

• (1) By combining AHP, a comprehensive model for evaluating coal-seam floor aquifer yield properties was established. In terms of representativeness and convenience of data acquisition, five factors – specific yield, permeability coefficient, drilling fluid consumption, aquifer thickness, and fault and fold distribution – were selected to evaluate aquifer yield. The model reflected the fact that aquifer yield is influenced by many factors.

• (2) Taking the Ordovician limestone aquifer at Danhou coal mine, Hebei, China, as the data source, GIS technology was used to develop thematic contour maps of the distribution of factors influencing aquifer yield. The thematic GIS maps for the different factors were subsequently superimposed, producing the comprehensive aquifer yield properties evaluation model discussed in this paper. The CI value for each yield property was calculated for every superimposed cell, and the research area was divided into zones with different yield characteristics, so that a map showing zones with different coal-seam floor aquifer yields could be drawn. GIS technology is helpful because it enables the storage, processing and graphic representation of relevant information, and can provide reference for aquifer yield analysis and evaluation in the future.

This research was financially supported by China National Natural Science Foundation (Grant no. 41430318, 41272276), the China National Scientific and Technical Support Program (Grant Numbers 201105060-06, 2012BAB12B03), Guizhou Province Science and Technology Agency Foundation (Qian Ke He LH Zi [2014]7617), Guizhou University Introducing Talents Research Foundation (2014-61), the Innovation Research Team Program of the Ministry of Education (IRT1085), and the State Key Laboratory of Coal Resources and Safe Mining.