Introduction

The Jharia coalfield, India is known for its high grade coal and associated coal fires. This coalfield hosts the maximum number of known coalmine fires among all the coalfields in India, which is burning in underground since nearly a century. Jharia coalfield was nationalized in 1970 and the coal fire has spread in many areas. Unplanned coal exploitation without fire-prevention arrangements prior to the nationalization was responsible for these fires (Vaish and Pal 2013, 2015a, 2016). The unplanned old mining of the area has raised some serious geotechnical concerns viz, foundation problems on buildings placed on these mining areas and the damage to roads resulting in potholes, etc (Pal et al. 2016; Bharti et al. 2014, 2015; Kumar et al. 2015; Singh 2015; Srivardhan et al. 2016). The old mining infrastructure facilities such as shafts, mining wells, mining goaf, mining galleries, and hidden mining holes are potential locations of hazards to those who are unaware of the existence of cavities. The hazard may be caused due to the collapse of the mining cavities from natural alteration processes during the course of time by wear and tear.

The term subsurface cavity is used to denote all subsurface features such as caves, caverns, voids, karst, pothole, and sinkhole etc. The voids, potholes, and sinkholes are very common in Jharia coalfield, which are caused by coal fire and old unplanned/illegal mining activity. Underground hidden cavity pose a great threat to the local environment, agricultural land, ecology and the health of those living in their proximity; eventually resulting in loss to the national economy.

Generally, the delineation and mapping of subsurface cavities remain a challenging task in many scientific and environmental fields (Van Schoor 2002; Abu-Shariah 2009). The traditional commercial techniques use drilling to locate cavities. A much more effective and economic solution is the use of geophysical survey, which helps in optimizing the location of drill hole. The geophysical techniques used for cavity detection are (1) induced polarization tomography (Brown et al. 2011; Martínez-Moreno et al. 2014 among others), (2) 2D electrical resistivity tomography (Zhou et al. 2004; Cardarelli et al. 2006, 2010; Pánek et al. 2010; Martínez-Pagán et al. 2013; Metwaly and AlFouzan 2013; Cardarelli et al. 2014 among others), (3) vertical electrical soundings (Rodríguez Castillo and Reyes Gutierrez 1992), (4) electromagnetic (Lange 1999), (5) ground penetration radar (Mochales et al. 2008; Leucci and De Giorgi 2010; Brown et al. 2011 among others), (6) self potential (Lange 1999), (7) magnetometry (Mochales et al. 2008 among others), (8) magnetic resonance sounding (Guérin et al. 2009), (9) gravimetry (Lange 1999; Mochales et al. 2008 and Gambetta et al. 2011 among others), (10) seismic methods: seismic refraction tomography (Guérin et al. 2009; Valois et al. 2010; Cardarelli et al. 2010, 2014 among others) .

The electrical resistivity tomography (ERT) method has wide applications in environmental, engineering and shallow subsurface investigations (Van Schoor 2002). Out of all various geophysical methods, the ERT has been proved to be very effective and suitable technique for mapping and delineation of cavities (Abu-Shariah 2009; Pánek et al. 2010; Martínez-Moreno et al. 2014; Martínez-Pagán et al. 2013; Metwaly and AlFouzan 2013). It is based on the assumption that various entities like solid bedrock, sediments, air, and water filled structures would have detectable electrical resistivity contrast relative to the host medium (Pánek et al. 2010). The present study deals with delineation and mapping of subsurface voids which are very common in and around Patherdih colliery of Jharia coal field, India.

Study area

The study area, Patherdih colliery (Fig. 1a) is situated in the southern part of the Jharia coalfield. The rock formations of Jharia coalfield unconformably overlying the Archean basement, mainly belong to the Lower Gondwana group rocks of Permian age comprising Talchir, Barakar, Barren measures and Raniganj formations, from bottom to top (Fig. 1a.). The study area lies within the geographical coordinates 23.66°N–23.67°N latitude and 86.43°E–86.44°E longitude. All the working coal seams lie in the Barakar formation of Lower Gondwana group rocks and are of early-Permian age. The rocks of Barakar formation comprises predominantly of sandstone of variable grain size, argillaceous sandstone, intercalation of sandstone and shale, carbonaceous shales, jhama, mica-peridotite and coal seams (Chandra 1992; Vaish and Pal 2015b). All parts of active coal fire zones have been covered by dumping of overburden material to prevent and combat against further exaggeration of coal fire in the surroundings.

Fig. 1
figure 1figure 1

a Location map of the study area (Patherdih colliery) along with generalized geological map of Jharia Coal field. b Active fire smoke has been observed near RD 90 m of AA/, the perpendicular distance is about 15–20 m from AA/

Full size image

There are 14 major coal seams in the area up to a total depth of ~700 m with varying thickness of 2 m to ~19 m. The major coal bearing seams at shallow depths are affected by fire, which are coal seam-XIVA and coal seam-XIV. The average thicknesses of these seams are 2.1 and 8.6 m. The average depths of these seams are 25 and 50 m, respectively. The seams affected by fire are partially filled up and blanketed by soil. Active fire smoke has been observed near RD 90 m of profile AA/ (Fig. 1b) with the perpendicular distance of about 15–20 m. Borehole log showing different coal seams at different depth with their physical status is shown in Fig. 2.

Fig. 2
figure 2

Borehole log showing different coal seams at different depth with their physical status. The location of this borehole is shown in Fig. 1

Full size image

Methodology

The burning of coal leads to the formation of voids caused by volume reduction due to the transformation of coal to ashes. The surface displacement related to mine fires, together with mining induced subsidence leads to subsequent subsidence of overlying strata. These subsidence leads to formation of cracks and fissures. The cracks and fissures helps in creation of ventilation paths for oxygen circulation. It further supports the internal combustion thus aggravating the underground coalmine fires (Jiang et al. 2011). A conceptual model of void formation caused by coal fire is shown in Fig. 3. In the present study it is supposed to have considerable resistivity contrast between bedrock i.e. mainly sandstone/clay stone/shale etc. and void space possibly filled with ashes generated from burned coal seam. The results of laboratory measurements on the samples of coal and surrounding formation show that the sandstone and shale samples have resistivity of 100 Ωm whereas the coal samples have resistivity greater than 700 Ωm at full saturation (Verma et al. 1982). The results of field investigation show that generally, the coal bearing strata exhibit relatively high resistivities in the range of about 70–350 Ωm whereas, sandstone, sandy shale and shale contribute moderate to low resistivities in the range of about 25–100 Ωm (Verma and Bhuin 1979; Verma et al. 1982).

Fig. 3
figure 3

A conceptual model of void formation caused by coal fire

Full size image

ERT data have been acquired using Wenner, Schlumberger, dipole–dipole and gradient arrays to compare the results and to delineate features using joint inversion combining all array data of same profile. These joint inversion of all array data would further optimize the resolution capability and signal-to-noise ratio (Zhou and Greenhalgh 2000; Dahlin and Zhou 2006). Among the common arrays, the Wenner array has the strongest signal strength. This can be an important factor if the survey is carried in areas with high background noise. Wenner array is relatively sensitive to vertical changes in the subsurface resistivity below the center of the array. However, it is less sensitive to horizontal changes in the subsurface resistivity. In general, the Wenner array is good in resolving vertical changes i.e. horizontal structures, but relatively poor in detecting horizontal changes i.e. narrow vertical structures (Loke 1999). The Schlumberger array is moderately sensitive to both horizontal and vertical structures. In areas where both types of geological structures are expected, this array might be a good compromise between the Wenner and the dipole–dipole array (Loke 1999). The dipole–dipole array is sensitive to horizontal changes in resistivity, but relatively insensitive to vertical changes in the resistivity. Thus, it is good in mapping vertical structures, such as dykes and cavities, but relatively poor in mapping horizontal structures such as sills or sedimentary layers. This array has a better horizontal data coverage than the Wenner (Loke 1999). Dahlin and Zhou (2004, 2006) have shown that the gradient array with multiple current-electrode combinations is best among the electrode arrays in terms of resolution of subsurface structures and it is clearly superior to the commonly used Wenner, Schlumberger, dipole–dipole, pole–dipole and pole–pole arrays in most of the modeled cases.

Present study utilizes a state-of-the-art 61-channel 64 electrode FlashRES-Universal electrical resistivity tomography (ERT) data acquisition system (ZZ Resistivity Imaging Pty Ltd, Australia). It offers an advanced technique for collection of data with maximum AB and MN combinations to obtain extremely large readings without being limited to standard configurations (such as Wenner, Schlumberger and dipole–dipole) than the other traditional methods. 61-channel or 61-voltage data are collected in parallel for each AB current pair and the amount of data collected reach more than 60,000 with 64-electrode layout within an hour. This is far superior to similar electric instruments in terms of the amount of data collected and data collection time. Stummer et al. (2004) performed synthetic data tests to demonstrate that the large amount of data results in better accuracy. Zhe et al. (2007) have concluded that the larger number of data points acquired using multielectrode and multichannel ERT techniques would delineate anomalies with better position, shape and accurate resistivity values after inversion. They monitored the noise through full waveform display. The additional important feature of the equipment is that there is no need to collect the data separately for different array. The user can collect the data with numbers of specified arrays in a single run and the data collected by different arrays could be used separately for interpretation. This makes data collection more effective and uses less time than the conventional systems (Zhe et al. 2007). The total 64 electrodes used for entire data acquisition system are (1) two current electrodes (A and B), (2) one reference electrode (M), and (3) 61 potential electrodes (VMN1–VMN61). The principle of ERT data acquisition consists of application of constant direct current imposing into the ground via two current electrodes (A and B) and one as common reference electrode (M). It then simultaneously measures 61 potentials VMN1, VMN2…VMN61 (relative to M) on remaining 61 electrodes (Fig. 4). Position of different possible electrode pair alternatively can acts as current and remaining 61 electrode as potential electrodes keeping one electrode as common reference depending on different geometry of electrode arrays (Zhe et al. 2007). The acquired data are processed using FlashRES Universal survey data checking program for removing noisy data. The filtered output data is then inverted using a 2.5D resistivity inversion (Zhou and Greenhalgh 2000). Different anomalous resistive features observed in the inverted resistivity sections have been compared based on their approximate locations, depths, dimensions, and resistivity values (Song and Kuenzer 2014, Cardarelli et al. 2006 among others). The color scale bars of all 2D resistivity sections are given separately. The range of each color scale is 0–500 Ωm.

Fig. 4
figure 4

61 Channel ERT data acquisition field setup using 64 electrodes, of which two as current electrodes (A, B) and one as common reference electrode (M) and 61 potentials relative to M on remaining electrodes

Full size image

Results and discussions

Following, Ezersky (2008) before carrying out the survey a situation similar to the field has been simulated (Fig. 5). The sandstone of resistivity 2–80 Ωm extends from surface of the earth to 22 m. The coal seam/shaly coal of thickness 5 m overlies the sandstone of resistivity 100 Ωm at a depth of 27 m. The coal seam/shaly coal resistivity varies from 100 to 350 Ωm. Two voids of 50 m by 5 m and 40 m by 5 m at RD 160 m and RD 470 m, respectively, have been considered. The forward modeling has been carried out for two situations, (1) the resistivity of the void at RD 160 m varied from 500 to 2000 Ωm, and (2) the resistivity of the void at RD 470 m varied from 500 to 10,000 Ωm. In both the situation the considered void at RD 160 m has more resistive elements than that of RD 470 m. The forward modeling has been carried out with 10 m electrode spacing for 64 electrodes as shown in Fig. 5a. The obtained data has been corrupted with 5 % of random noise. Figure 5b, d shows the obtained results from dipole–dipole and gradient arrays for the voids with resistivity varying from 500 to 2000 Ωm. Whereas, Fig. 5c, e illustrates the results from dipole–dipole and gradient arrays with resistivity of void varying from 500 to 10,000 Ωm. Interestingly, all the results recover the void location and dimension. However, the resistivity of the void is not recovered and is about 170–300 Ωm. In fact, the resistive voids (Fig. 5a) generates conducting voids (Ezersky (2008) as shown in Fig. 5b–e. In the considered model the void at RD 160 m is more resistive than that of the RD 470 m. However, the obtained resistivity of the both the voids are approximately same.

Fig. 5
figure 5

a Resistivity model section, b, d inverse resistivity section obtained from dipole–dipole and gradient array for the void resistivity varying from 500 to 2000 Ωm, c, e inverse resistivity section obtained from dipole–dipole and gradient array for the void resistivity varying from 500 to 10,000 Ωm

Full size image

Two ERT profiles (AA/ and BB/) have been considered over known fire affected area. The electrode spacing of 10 m and total profile length of 630 m have been selected for delineation and mapping of subsurface voids. The starting point i.e. 1st electrode position considered as reduced distance (RD) as 0 m (A/B) and end of the profiles i.e. 64th electrode position as RD 630 m (A//B/). The ERT data have been collected using Wenner, Schlumberger, dipole–dipole and gradient arrays. Total 651, 1665, 3013 and 5368 current electrode pairs have been used for Wenner, Schlumberger, dipole–dipole and gradient arrays, respectively. The collected data have been processed using FlashRES Universal survey data checking program (FlashRES Universal, User manual 2014). The acquired data over the profiles have been processed using three different thresholds for decreasing noise level and vice versa enhancing the quality of the acquired data (FlashRES-Universal user manual). Quality factor 5 means that the 95 % of the acquired data is of good quality. Figures 6, 7 and 8 illustrates the 2D ERT sections of profile AA/ generated using (a) Wenner, (b) Schlumberger, (c) gradient, (d) dipole–dipole and (e) joint inversion of the all combined arrays with three different thresholds viz., (1) 10 mA current and quality factor 17; (2) 40 mA current and quality factor 12 and (3) 60 mA current and quality factor 5, respectively. These thresholds correspond to 94, 84 and 43 %, respectively, of the acquired data. Similarly, 2D ERT sections of profile BB/ with the same threshold combinations of currents and quality factors have been generated and shown in Figs. 9, 10 and 11, respectively. For this profile the above three thresholds corresponds to 84, 64 and 40 %, respectively, of the acquired data. The details of various features delineated from the inverted resistivity sections of various arrays have been presented in Tables 1 and 2, respectively, for both AA/ and BB/ profiles. Out of three different thresholds, the best result is obtained when the quality factor is 5 with current threshold of 60 mA. The quality of the results naturally decreases with the increase in the quality factor and decrease in the current threshold. The quality of results may changes with varying current source, but this would vary if spontaneous potential are absent which depends on environment noise. Figure 12a, b compares the root mean square (RMS) error for different arrays with three thresholds for both the profiles. Tables 1 and 2 provide the details of RMS errors of inversion results generated using different array for both the profiles. It has been observed that RMS error increases with the increase in quality factor. RMS error is minimum for the joint inversion of all the combined arrays. RMS error of the Gradient array is least of all the other arrays viz., Wenner, Schlumberger and dipole–dipole arrays.

Fig. 6
figure 6

2D ERT section along AA/ using a Wenner, b Schlumberger, c gradient, d dipole–dipole arrays and e joint inversion of all the arrays with current threshold of 10 mA and quality factor 17

Full size image
Fig. 7
figure 7

2D ERT section along AA/ using a Wenner, b Schlumberger, c gradient, d dipole–dipole arrays and e joint inversion of all the arrays with current threshold of 40 mA and quality factor 12

Full size image
Fig. 8
figure 8

2D ERT section along AA/ using a Wenner, b Schlumberger, c gradient, d dipole–dipole arrays and e joint inversion of all the arrays with current threshold of 60 mA and quality factor 5

Full size image
Fig. 9
figure 9

2D ERT section along BB/ using a Wenner, b Schlumberger, c gradient, d dipole–dipole arrays and e joint inversion of all the arrays with current threshold of 10 mA and quality factor 17

Full size image
Fig. 10
figure 10

2D ERT section along BB/ using a Wenner, b Schlumberger, c gradient, d dipole–dipole arrays and e joint inversion of all the arrays with current threshold of 40 mA and quality factor 12

Full size image
Fig. 11
figure 11

2D ERT section along BB/ using a Wenner, b Schlumberger, c gradient, d dipole–dipole arrays and e joint inversion of all the arrays with current threshold of 60 mA and quality factor 5

Full size image
Table 1 Details of distinct resistive anomalous features delineated in AA/ profile using (a) Wenner, (b) Schlumberger, (c) gradient, (d) dipole–dipole arrays, and (e) joint inversion of all the arrays corresponding to different current threshold and quality factor
Full size table
Table 2 Details of distinct resistive anomalous features delineated in BB/ profile using (a) Wenner, (b) Schlumberger, (c) gradient, (d) Dipole–dipole arrays and (e) joint inversion of all the arrays corresponding to different current threshold and quality factor
Full size table
Fig. 12
figure 12

Comparison of root-mean-square (RMS) error of inversion results obtained from Wenner, Schlumberger, gradient, dipole–dipole arrays and four combine array for different current threshold and data quality factor

Full size image

It has been observed that the ERT data collected over profile AA/ have better current injection (proper grounding of electrodes) and low RMS error (quality factor low) of resistivity readings than that of the ERT data collected over profile BB/. Due to scarcity of space and for mapping of the feature encountered in AA/, the profile BB/ has been taken partially over rugged topography filled with overburden which could not establish proper grounding in some electrodes, despite watering with bentonite clay to the electrodes. It results in degradation of the data quality of the profile BB/.

The comparative study of Figs. 6, 7, 8, 9, 10, 11 and Tables 1 and 2 proves that different resistive anomalous features delineated using joint inversion of combined data collected by all arrays gives best suitable results. Because it provides 2D inverted section with minimum RMS error among all. The different anomalous resistive features delineated by this technique could be correlated well with same in all individual arrays. Similarly, it is observed that features delineated using gradient array provides next suitable results with less RMS error than the dipole–dipole, Schlumberger and Wenner arrays. Whereas, features delineated using dipole–dipole array provides second next suitable results with less RMS error than the Schlumberger and Wenner arrays.

Figures 6, 7, 8, 9, 10, and 11 shows the presence of voids associated with coal fire zone with location, dimension and resistivity. Generally, the gradient, dipole–dipole and joint inversion of all combine arrays brings up more than one voids associated with coal fire. The imprints of entire coal seam fires have been mapped by using joint inversion of combined array only (Figs. 6e,7e,8e,10e and 11e). The delineated voids seem to be conducting, in comparison to the results obtained of model study (Fig. 5). The observation regarding resistivity of the voids may also be related to the heating of water-saturated coal samples. Duba (1977) studied the electrical conductivity/resistivity of coal and coal char. Revil et al. (2013, Fig. 7) discussed the shape of this curve. Initially, a large increase of resistivity from 1000 Ωm at 24 °C to ~6.3 × 108 Ωm at 110 °C is observed (Duba 1977). Then the resistivity reaches to ~3.1 × 109 Ωm at 300 °C with a relatively smooth increment owing to water loss on drying or desaturation (Revil et al. 2013). Subsequently, resistivity decreases slowly to ~2.5 × 106 Ωm at 515 °C and then decreases rapidly to ~0.01 Ωm at 800 °C. Revil et al. (2013) termed this path as dry path.

The voids delineated at depth of about 23–25 m is caused by XIVA coal seam fire. Whereas voids delineated at depth of about 45–54 m is caused by XIV coal seam fire. These observations support the conceptual model of void formation caused by coal fire (Fig. 3). Finally, a model (Fig. 13) of fire propagation has been established by correlating different equivalent anomalous resistivity zones/cavities between the profiles AA/ and BB/ using 2D ERT sections obtained from joint inversion of all combine arrays. The coal seam-XIVA affected by fire is resulted in void at RD 89 m and at depth of 25 m (profile AA/) which is characterized by relatively high resistivity of about 300 Ωm. This fire activity is assumed to be connected with the profile BB/ at RD 324 m and at depth of 25 m through fracture plane. It is interesting to mention that coal fire smokes have been observed on the ground near RD 89 m and RD 324 m of profiles AA/ and BB/, respectively. The surface manifestation of coal fire smokes support the model established in Fig. 13. In addition, a distinct anomalous resistive zone of about 200 Ωm has been delineated at RD 278 m and at depth of 46 m (XIV seam coal, Fig. 2) in profile AA/ which is assumed to be connected with the profile BB/ at RD 520 m and at depth of 48 m through fracture plane. This zone is assumed to be previously burned and filled with moist coal ashes. Similar, anomalous resistive zones with resistivity of about 200 Ωm have been delineated, one at RD 595 m and at depth of 24 m in profile AA/ and another at RD 126 m and at depth of 44 m in profile BB/.

Fig. 13
figure 13

Model of fire propagation and relationship of anomalous resistivity zones/cavities established from 2D ERT section of profile a AA/and b BB/using joint inversion of all the arrays with current threshold of 60 mA and quality factor 5

Full size image

Conclusions

The ERT study has been carried out for delineation of hidden near-surface cavities over known coal fire affected area. Initially, field situations have been simulated for two voids with varying resistivity through forward modeling for gradient and dipole–dipole arrays. The data for both the arrays have been corrupted by 5 % random noise. The inversion of the noise corrupted data sets for both the arrays bring up the location and dimension of the voids within reasonable accuracy unlike their resistivities. Infact, the considered resistive void seems to be relatively conducting void.

Subsequently, 2D ERT data have been acquired along two profiles over coal fire affected area of Patherdih colliery using Wenner, Schlumberger, dipole–dipole, and gradient arrays. The acquired field data have been analyzed for three different combinations of quality factor and current threshold i.e. (1) 5 and 60 mA, (2) 12 and 40 mA, and (3) 17 and 10 mA, respectively. The minimum RMS error has been obtained with the quality factor 5 and threshold 60 mA for all the arrays including joint inversion of all combine arrays. In accordance with the synthetic model study, the location and dimension of the voids have been well identified using field ERT data except their resistivities. It has been established that the different resistive anomalous features delineated in both profiles are generated by burning of coal seams XIVA and XIV which leads to the void formation at the depths of about 23–25 and 45–54 m, respectively. The obtained results matches well with the available borehole log and surface manifestation of fire over the study area. These observations support the conceptual model (Fig. 3) of void formation caused by coal fire and final model of fire propagation (Fig. 13). The results proves the efficacy of the ERT technique for detection of voids associated with coal fire.