Risk assessment of Cretaceous water inrush in the Ordos Basin based on the FAHP-EM

The Ordos Basin is currently the largest coal-producing area in China, and the total amount of coal buried less than 2 km from the surface of the basin is about 2 trillion tons, which accounts for more than 40% of the country's total and is projected to be the country's main energy supply production area over the next 50–100 years (Wang, 2017). Cretaceous sandstone and mudstone strata are distributed in the central region of the Ordos Basin with an area of 13.21 × 10⁴km² and a general thickness of 300–800 m; the region with a thickness greater than 1 km accounts for about 20%. The central region contains the largest aquifer in the arid and semi-arid regions of the Ordos Basin. The total amount of water in the aquifer is about 7 billion cubic meters (dynamic replenishment) per year, which accounts for 70% of the total amount of groundwater resources in the basin (Hou et al., 2008). This aquifer is the most important water resource for the development of energy resources, economic and social development, and the maintenance of the environment in the Shaanxi, Inner Mongolia, Ningxia, and Gansu regions. Production mines built in the early stages of development (since 2000) are mainly distributed in the shallow coal seam area around the basin, which is not covered by Cretaceous strata. In recent years, some large-scale mines have been built and started coal production in areas that extend to the area of Cretaceous coverage in the hinterlands of the basin, resulting in deformation and destruction of the overlying strata. It is easy to cause water leakage in the Cretaceous system (Adam & Paul, 2000; Zhu et al., 2020), water inrush accidents in mines (Andreas & Nikola, 2011), and death of surface vegetation and aggravation of desertification (Wang & Hang, 2010). Therefore, it is crucial to analyze risk factors such as the water-bearing capacity of Cretaceous aquifers and to evaluate the risk of water inrush in order to improve safety during coal production and to conserve the environment.

In recent years, scholars have proposed a series of water inrush prediction methods according to observations gathered from operating coal mines, which enriched the theory of water inrush forecasting and prevention and guided the prevention and control of water inrush. In the 1960s, Chinese scholars proposed a water inrush coefficient method for predicting water inrush from the coal seam floor during the ‘Hydrogeology Battle’ in the Jiaozuo mining area (Hu et al., 2019). Wu et al. (2014) proposed a ‘three maps-two predictions’ method based on years of research theory, experience, and achievements, which solved three major problems addressing the water-filling source, water passageway, and strength of the coal seam roof. Afterward, Wu et al. (2015) proposed a vulnerability index method with multi-level zoning characteristics which considered the complex interactions among the main factors controlling water inrush from the coal seam floor and their relative weight ratio. Yang et al. (2019) used an in situ water-level monitoring system and an artificial neural network model to determine the source of the mine water. Zhao et al. (2018) proposed a water inrush evaluation model based on the random forest method; this model is a powerful intelligent machine learning algorithm with high classification accuracy and the ability to evaluate the importance of variables. Chu et al. (2017) considered the risk of water inrush in karst tunnels using a feasible and accurate fuzzy comprehensive evaluation method. Zhang & Yang (2018) and Wu et al. (2013) used an analytic hierarchy process (AHP) to study the risk zoning of water inrush from the roof of a shallow coal seam and the factors influencing water inrush from Ordovician limestone underlying a coal seam. Li (2010) used the theory of multiple information fusion to study the prediction and prevention of coal mine water hazards. Ali et al. (2020) assess and analyze the vulnerability and mitigation capacity of biosphere reserves based on GSI. Li et al. (2010) established the cusp catastrophe model of water inrush from a collapsed column by using catastrophe theory. Lu et al. (2017) evaluated the risk of water inrush from the separation layer based on the entropy method (EM).

Although the above research has made progress on evaluating the risk associated with water inrush, the methods also have shortcomings. The water inrush coefficient method only considers the factors of aquifer water pressure and the thickness of the aquifuge, and the results are conservative, resulting in wasted resources. The artificial neural network model has some disadvantages, such as different structure selection, and the weight converging to the local minimum leads to the failure of network training. When the AHP judges the consistency of a matrix, it needs to be adjusted many times, and the judgment process is very complex. Furthermore, the above studies mainly involve water inrush problems on high-pressure Ordovician limestone, subsidence column, old gob area, borehole, and bed separation. There are few studies on Cretaceous systems in the Ordos Basin. Therefore, the FAHP-EM was proposed to evaluate the risk of water inrush in coal mines, based on the fuzzy AHP and EM (FAHP-EM). FAHP-EM is a method of subjective and objective comprehensive weighting, which combines the advantages of both methods and complements the procedures. The combined weighting uses existing data to reflect the potential regularity, resulting in accurate and reasonable values. Taking the Yingpanhao coal mine as the study area, this paper uses the FAHP-EM and the water inrush coefficient method to evaluate the risk of Cretaceous water inrush in the mine and compares the evaluation results of the two methods, and it further shows the FAHP-EM has strong applicability.

The Yingpanhao coal mine is located in Wushenqi county in the southwest of Ordos city in the Inner Mongolia Autonomous Region (Figure 1(a)). It is adjacent to Bayanchaidamu minefield in the east, Galutu minefield in the west, an exploratory area in the north, and Baijiahaizi minefield in the south (Figure 1(b)). The mine is situated in the middle of the Maowusu desert in the Ordos Basin. The minefield is about 113.41 km² and the elevation ranges from 1,173.00 in the east to 1,317.40 m in the southwest.

The main landform of the study area is sandy beach, with sand dunes, longitudinal dunes, and sandy lands distributed widely. Most of the surface is covered by modern aeolian and lacustrine sand, and Quaternary loess is found sporadically. According to borehole exposures and geological mapping data, the strata in this area from older to more recent systems are upper Triassic Yanchang formation (T_3y), middle Jurassic Yan'an formation (J_2y), Zhiluo formation (J_2z), Anding formation (J_2a), lower Cretaceous Zhidan group (K_1zh), Neogene Pliocene (N₂), Quaternary upper Pleistocene Malan formation (Q_{3 m}), Residual-slope (Q₃^dl+pl), Holocene eluvial-slope wash (Q₄^al+pl), marsh sediment (Q_{4 h}), and Aeolian sand (Q₄^eol) (Figure 2).

The first coal seam in the study area was 2⁻² coal with a thickness ranging from 3.16 to 10.24 m and an average thickness of 6.29 m. The hydrogeological profile of the first exploration line in the study area is shown in Figure 3. From top to bottom, the aquifers in the coal seam are Quaternary phreatic aquifer, confined aquifer of the lower Cretaceous Zhidan group, and a confined aquifer of the middle Jurassic Zhiluo formation. Among them, the Cretaceous Zhidan group has the largest water reserves, and the maximum water inflow reached 18,787 m³/d when the main shaft, auxiliary shaft, and air shaft were excavated. To prevent serious water inrush accidents and ensure safety during future coal mining, the risk of water inrush in the Cretaceous system must be evaluated, and the risk zoning map of water inrush must be drawn.

A disaster involving Cretaceous water inrush from a coal seam roof is affected by many factors, and the magnitude of water released varies with both the mining and geological conditions. The selection of factors influencing the inrush is very important in order to accurately predict water hazard zoning. Considering the difficulty of obtaining actual factor indices and the geological conditions of the mining area, this paper chose coal seam mining thickness, rock quality designation (RQD), the distance between the top boundary of the water-conducting fracture zone and the bottom boundary of the Cretaceous system, the thickness of the Cretaceous aquifer, and the height of the water head as the key factors affecting the risk of Cretaceous water inrush. The selected factors were analyzed as follows.

After excavation of the coal seam, the original stress of the roof rock changes, and the stress is redistributed. The deformation occurs to varying degrees in the range affected by the rock mass. The bigger the coal seam mining thickness, the higher the development height of the water-conducting fracture zone, the closer the fracture-conducting zone is to the bottom of the aquifer, the more likely that water inrush will occur in the aquifer.

After coal seam mining, the deformation and failure degree of overburdened roof rock are closely related to the structural characteristics of the overburdened rock. RQD is the ratio of the cumulative rock core length of greater than 10 cm to the length of footage per run. The greater the RQD, the better the integrity of the rock. After coal mining, the higher the development height of the water-conducting fracture zone, the greater the possibility of water inrush in Cretaceous systems.

This index is the factor directly affecting water inrush. The greater the distance between the top boundary of the water-conducting fracture zone and the bottom boundary of the Cretaceous system, the more difficult it is to form the water-conducting channel, reducing the possibility of water inrush. The main coal seam in the study area is a Jurassic coal seam. Because of the inappropriateness of the existing empirical formula (Liu et al., 2018), this paper collected measured data of the height of the water-conducting fractured zone in the Jurassic coalfield under similar mining conditions and overburdened conditions (Table 1). The ratio of the height of the fractured zone to the mining height was 11.30–28.20, with an average value of 20.50. As can be seen from Table 1, the highlighted measured values in the mine area are very close to the average value, which indicates that the formula calculated according to the average value is feasible. Based on this, the distance data between the top boundary of the water-conducting fracture zone and the bottom boundary of the Cretaceous system is calculated.

The thicker the aquifer, the greater the aquifer water content per unit area, and the greater the risk consequences of a water hazard.

Water head height refers to the confined head value in the Cretaceous aquifer. Generally speaking, the higher the confined water head of an aquifer, the greater the hydrostatic pressure on the lower strata of the aquifer, and the easier the aquifer is to break, increasing the risk of water inrush.

The AHP was proposed by American operation researcher Saaty (1977). After nearly half a century of development, it has developed into a more mature and effective method for solving complex problems with multiple objectives. Zadeh (1965) proposed the use of fuzzy sets and introduced the concept of membership in fuzzy mathematics. Zhang (2000) combined the AHP with fuzzy mathematics and gave the principles and steps of the FAHP method to reduce the influence of human factors on a comprehensive evaluation. We can see from the research results of Zhang (2000) that the introduction of fuzzy matrices in the AHP makes it easier to test whether the judgment matrix is consistent; as well as can quickly make the fuzzy inconsistent matrix have fuzzy consistency; and the standard for judging whether the matrix is consistent is also more scientific. This method can also effectively solve the ambiguity in the comprehensive evaluation process (Lu et al., 2017).

To build the model, it is necessary to divide the goals of the decision, the factors considered, and the objects of the decision into goals according to their mutual relationship layer, middle layer, and decision layer, the factors of the same layer are subordinate to the factors of the previous layer and at the same time dominate the factors of the next layer. The target layer of this study is the Cretaceous water inrush risk layer and there are five indices, including the coal seam mining thickness (U1), the RQD (U2), the distance between the top boundary of the water-conducting fracture zone and the bottom boundary of the Cretaceous system (U3), the Cretaceous aquifer thickness (U4), and the water head height (U5). Using these factors, the hierarchical structure model of the Cretaceous water inrush risk in Yingpanhao Coal Mine was determined.

The fuzzy consistent judgment matrix represents the comparison of the relative importance between an element of the previous layer and the related elements of this layer; according to the digital scale in Table 2, the relative importance of any two-layer elements with respect to a criterion can be quantitatively described, thereby constructing a fuzzy consistent judgment matrix. The fuzzy index scale method quantifies each factor on a scale of 0.1–0.9 and makes pairwise comparisons about the factors (Table 2).

The EM is an objective weighting method, and it determines the weight of the main control factors according to the variation of each index value and avoids the deviation caused by human factors. The EM first appeared in thermodynamics and was later introduced into information theory (Shannon, 1948). It is widely used in various fields (Zou et al., 2006; Zhang et al., 2010; Xia et al., 2018). According to the following three steps, the weighting is evaluated using the EM.

According to Equation (4), since the obtained compatibility index as given by I(R, W_R*) = 0.098 < 0.1, the fuzzy complementary judgment matrix had satisfactory consistency. Therefore, the weight matrix of each factor was w₁ = 0.23, w₂ = 0.21, w₃ = 0.20, w₄ = 0.18, and w₅ = 0.18.

Five evaluation indicators and the evaluation objects were selected, based on the quantitative results of key influencing factors (Table 2). From Equations (6)–(8), the standardized matrix was obtained. According to Equations (9) and (10), the entropy was e₁ = 0.9788, e₂ = 0.9787, e₃ = 0.9805, e₄ = 0.9920, and e₅ = 0.9815. According to Equation (11), the entropy weights were u₁ = 0.24, u₂ = 0.24, u₃ = 0.22, u₄ = 0.09, and u₅ = 0.21.

Based on the subjective (w_i) and objective (u_i) weights, the comprehensive weight (t_i) is determined per Equation (12) and presented in Table 3.

In the standardization process, the positive and negative correlations between the factors and target event must be considered. Four of the factors (the coal seam mining thickness, proportional coefficient of hard rock, height of the water-conducting fracture zone, and water head height) were positively correlated with the Cretaceous water inrush potential. For these variables, the larger the value, the greater the risk of Cretaceous water invasion. The greater the aquifuge thickness, the less the risk of a Cretaceous water inrush. Hence, the distance from the top boundary of the water-conducting fracture zone to the bottom boundary of the Cretaceous system was negatively correlated. We used the EM to standardize the data and establish an attribute database of each evaluation index, then GIS software was used to draw a five-factor normalized thematic map (Figure 5(a)–5(e)). As can be seen from Figure 5(a)–5(e), the five impact indexes have different influence areas on the Cretaceous water damage in Yingpanhao mining area.

The normalized thematic maps are processed according to the weight of each evaluation index.

According to the above steps, the risk assessment map of Cretaceous water inrush in Yingpanhao mine was established. Using the natural grading method in GIS, the area is divided into safe, relatively safe, transitional, less fragile, and fragile (Figure 6).

Comparing Figure 6 with Figure 7, Figure 6 shows that the western and northern margins of Yingpanhao mining area are mostly safe area, relatively safe area, and transitional area; the southeast is divided into fragile area and less fragile area. Figure 7 shows that the western margin of Yingpanhao mining area is the safe area and the eastern margin is the fragile area. In particular, it needs to be noted that in Figure 6, the auxiliary shaft and the main shaft are located in the fragile area, and the air shaft is located in the less fragile area. In Figure 7, they are all in the safe area. The water inrush coefficient method only considers two factors: aquifer thickness and hydrostatic pressure; but in the FAHP-EM evaluation method, five key influencing factors are adopted. According to the well construction report of Yingpanhao coal mine, the maximum instantaneous water inflow of Zhidan group is 18,787 m³/d; the shaft location is located in the Cretaceous water inrush risk area (as shown in Figure 6), which verifies the accuracy of the water inrush prediction method in this paper. Therefore, compared with the water inrush coefficient method, the FAHP-EM can provide a viable option to predict the water inrush risk index in different areas of the entire mining area.

In summary, the FAHP-EM proposed in this paper comprehensively considers factors such as mining thickness, aquifuge lithology and thickness, aquifer thickness, and water pressure. The calculation and evaluation results are viable accurate option than those of the water inrush coefficient method, effectively avoiding the waste of resources caused by conservative calculation results. At the same time, the FAHP-EM combines the FAHP and the EM, which reflects the actual experience of experts, and makes full use of the potential regularity reflected by the existing data, effectively avoiding the complexity of the AHP in judging the consistency of the matrix. The FAHP-EM can not only predict and evaluate coal mine roof water hazards, but compared to many natural disasters and other industries' risk assessments, this method can also provide a viable option as long as reasonable risk impact indicators are selected. This article only makes a comparison between the FAHP-EM and the water inrush coefficient method of the coal mine water prevention and control standard. In the future, the FAHP-EM can be compared with the existing coal mine water hazard evaluation method. This is where this article needs to be improved.

The results of this paper can be summarized as follows:

• 1.

  Based on the FAHP and the EM, using ArcGIS data processing, weight extraction, data normalization, and other functions, a method for evaluating the risk of water inrush from coal roof is proposed.

• 2.

  Take the Cretaceous water disaster of Yingpanhao coal mine as an example, comparing the evaluation results of water inrush coefficient method and the FAHP-EM, the mine area is divided into five categories according to the safety degree of the FAHP-EM, and the mine area is divided into two categories according to the traditional water inrush coefficient method. The water inrush point in the well construction report verifies that the FAHP-EM can provide a viable option to evaluate the water inrush risk of the Cretaceous system.

• 3.

  This study provided an important guiding significance for coal mining for Yingpanhao coal mine and the west China coal fields, and provided a theoretical basis for the prevention and control of Cretaceous water disasters.

The study was supported by the State Key Program of the National Natural Science Foundation of China (Grant No. 41931284) and the Postgraduate Research and Innovation Projects of Jiangsu Province (Grant No. KYCX21_2328).

All relevant data are included in the paper or its Supplementary Information.