Monitoring subsurface fires in coalfield using observations of land subsidence from differential interferometric synthetic aperture radar (DInSAR)

Nishant Gupta, Tajdarul H Syed∗ and Ashiihrii Athiphro Department of Applied Geology, Indian School of Mines, , . ∗Corresponding author. e-mail: [email protected]

Coal fires in the Jharia coalfield pose a serious threat to India’s vital resource of primary coking coal and the regional environment. In order to undertake effective preventative measures, it is critical to detect the occurrence of subsurface coal fires and to monitor the extent of the existing ones. In this study, Differential Interferometric Synthetic Aperature Radar (DInSAR) technique has been utilized to monitor subsurface coal fires in the Jharia coalfield. Results showed that majority of the coal fire-related subsidence were concentrated on the eastern and western boundaries of the coalfield. The magnitude of subsidence observed was classified into high (10–27.8 mm), low (0–10 mm) and upliftment (−10–0 mm). The results were strongly supported by in situ observations and satellite-based thermal imagery analysis. Major subsidence was observed in the areas with repeated sightings of coal fire. Further, the study highlighted on the capability of the methodology for predicting potential coal fire zones on the basis of land surface subsidence only. The results from this study have major implications for demarcating the hazardous coal fire areas as well as effective implementation of public safety measures.

1. Introduction with coal fires add to the growing concern of haz- ardous impact on the masses residing in the study Accounting for nearly 70% of India’s energy pro- area. Thus the monitoring of coal fires, because of duction, coal remains a key contributor to the its socio-economic importance and environmental economic development of the country. The Jharia impacts, present an interesting challenge to scien- coalfield of eastern India is the primary producer tists and policy-makers alike. Concerns over the of high grade coking coal, which is a critical com- degrading health and safety of the local popula- ponent of India’s major steel industry. However, tion justify a detailed assessment of the extent in spite of its economic importance, Jharia coal- and direction of propagation of coal fires. Such field has been burning underground for nearly a an assessment will also contribute towards the century and hosts the maximum number of uncon- identification and implementation of the essential trollable surface and subsurface coal fires in India preventive measures. (Chatterjee et al. 2006). Jharia coal fires lead Previously, Bhattacharya et al. (1991)and to considerable loss of valuable, non-renewable, Mukherjee et al. (1991) utilized airborne thermal reserve of prime coking coal in addition to hav- data for the identification of coal fires and their ing adverse effects on the regional environment depth of occurrence, in the Jharia coalfield, using (Chatterjee 2006). Land subsidence associated linear heat flow equation. The efficacy of thermal

Keywords. Coal fire; Jharia coalfield; DInSAR; thermal remote sensing.

J. Earth Syst. Sci. 122, No. 5, October 2013, pp. 1249–1258 c Indian Academy of Sciences 1249 1250 Nishant Gupta et al. data from spaceborne Landsat-5 Thematic Map- (Wegmuller et al. 1999; Tom´as et al. 2005). How- per (TM) for the detection of coal fires in the ever, the DInSAR technique is limited by the fact Jharia coalfield was first established by Reddy et al. that it does not provide statistics of deformation (1993) and Saraf et al. (1995). Then on, Thermal rate at individual points within the interferogram Infra-Red (TIR) and Short Wavelength Infra-Red as is the case with the recently developed Persis- (SWIR) bands of Landsat-5 TM have been effec- tent Scatterer InSAR (PSInSAR) technique. The tively used to identify surface and subsurface coal PSInSAR technique allows highly accurate mea- fires (Prakash et al. 1997) and to calculate the sub- surements of deformation that occurs at individual, pixel area and temperature (Prakash and Gupta phase-dependent radar targets called permanent or 1999). In order to differentiate between surface and persistent scatterers (PS) (Jiang et al. 2011). subsurface coal fires and to study its lateral prop- Previously, DInSAR technique has been exten- agation, Chatterjee (2006) implemented the tech- sively used for analysing temporal and spatial nique of pixel integrated temperature modelling on characteristics of land subsidence due to various Landsat TM thermal IR images. Additional coal geophysical phenomena (e.g., Carnec et al. 1996; fire studies utilizing thermal bands were carried Galloway et al. 1998; Amelung et al. 1999; out by Chatterjee et al. (2007) (using images from Chatterjee et al. 2006). In the recent past, DIn- Landsat-5 TM and Landsat-7 ETM+), Gautam SAR technique has been used for the assessment of et al. (2008) (using NOAA/AVHRR) and Martha land subsidence caused by mining activities (e.g., et al. (2010) (using images from ASTER). Most Colesanti et al. 2005; Herrera et al. 2007; Jung recently, Mishra et al. (2011), empirically, esti- et al. 2007). While Prakash et al. (2001)andVoigt mated a scaled variation between the temperature et al. (2004) devised an integrated approach with of surface and subsurface coal fires obtained from the combination of optical, thermal and microwave the thermal band of Landsat ETM+ data and data, Jiang et al. (2011) utilized advanced InSAR in situ temperature observations using thermal techniques to study coal fires. The current study imaging camera. presents the most recent and comprehensive assess- In the past, Synthetic Aperture Radar (SAR) ment of land surface subsidence caused by coal fire interferometry (InSAR) has been effectively utilized in the Jharia coalfield of India. The methodology to monitor subsidence caused by various natu- involved the implementation of repeat pass DIn- ral and anthropogenic factors (such as earthquake SAR to make precise measurements of vertical land (e.g., Pathier et al. 2006), mining activities (e.g., subsidence primarily attributed to underlying coal Herrera et al. 2007) and groundwater withdrawal fires. Here we validated the DInSAR-based subsi- (e.g., Motagh et al. 2008)). This technique presents dence measurements with independent results from a method to obtain highly precise measurements thermal imagery and in situ observations. The of surface displacement, most prominently in the study, further hypothesized that areas with dis- vertical direction. InSAR involves the computa- cernible vertical subsidence but without any ther- tion of an interferogram using SAR image pairs mal signature or in situ sighting of coal fire are of the same area acquired at different times. In potentially hazardous zones with high probability this study an advanced InSAR technique based on of subsurface coal fire occurrence. small perpendicular and temporal baseline, two- pass Differential Interferometric SAR (DInSAR) was used. DInSAR provides information on two- 2. Study area dimensional deformation in the line of sight (LOS) of radar by calculating a differential interferogram Jharia coalfield, one of the leading coal producing for the image pair acquired during repeat pass of belts in India, is bounded by the latitudes 23◦39– the radar over the same area at different times 23◦50N and longitudes 86◦05–86◦30E and located (Massonnet and Feigl 1998). DInSAR provides within the state of (figure 1a). The a cost effective way of obtaining information on sickle shaped coalfield (figure 1b) contains the only two-dimensional deformation, with wide and con- remaining reserve of prime coking and tinuous spatial coverage, when compared to Dif- occupies an area of approximately 450 km2.Major ferential Global Positioning System (DGPS) and portions of this coalfield is operated by Bharat other instrumental methods, which are only able Coking Coal Limited (BCCL 2008). Geologically to measure ground deformation at discrete points. the rocks of the study area, belonging to the While the spatial coverage of levelling techniques Gondwana Supergroup, range in age from Upper is comparable to that of DInSAR, the bench- Carboniferous to Lower Cretaceous age, lie uncon- mark density is significantly lower than that formably over the older Archaean rocks. In Jharia of DInSAR. Apart from being cost intensive, coalfield, the major coal bearing formation is the levelling surveys are significantly constrained in Barakar Formation consisting of bands of coarse- both space and time when compared to DInSAR to-medium grey and white sandstones, shales and Monitoring subsurface coal fires in Jharia coalfield 1251

70°0'0"E 75°0'0"E 80°0'0"E 85°0'0"E 90°0'0"E

35°0'0"N 35°0'0"N

30°0'0"N 30°0'0"N

25°0'0"N 25°0'0"N

(a) 70°0'0"E 75°0'0"E 80°0'0"E 85°0'0"E 90°0'0"E

(b)

Figure 1. (a) Location map of Jharia coalfield in the overall Indian perspective. (b) Magnified view of the study area represented by a false colour composite (FCC) illustrating the Jharia coalfield with some of the major collieries affected by coal fires. 1252 Nishant Gupta et al. coal seams. Tectonically, this area consists of NE– two SAR images with similar imaging geome- SW-trending faults with exposures of lamprophyre tries, acquired at different times, for the same and dolerite dykes in a criss-cross manner (Saraf area. These two images are precisely coregistered et al. 1995). Figure 1(b) shows a false colour com- before the phase difference is computed for each posite (FCC) image of the study area prepared pixel. Interferograms used in this study were pro- using Landsat ETM+ bands 7, 5 and 3 (RGB). duced using the Repeat Orbit Interferometry Pack- The FCC (figure 1b) prepared shows coal bands age (ROI PAC), which was developed at Jet and coal dumps in the shades of brown to brownish Propulsion Laboratory and the California Insti- black colour. tute of Technology (JPL/Caltech). The interfer- ogram thus generated, from two SAR images, consists of a phase difference (Φ) (equation 1) 3. Data used which is an aggregate of phase difference related to topographic information of the terrain (Φtopo), Several key-independent satellite and field-based the line-of-sight surface displacement between two datasets have been utilized in this study. For acquisitions (Φdisp), delay in radar signal due to example, the Envisat Advanced Synthetic Aper- variations in the tropospheric water content in ture (ASAR) data for the study area was obtained atmosphere (Φatm), noise in the interferometric sig- from the European Space Agency (ESA). Two nal (Φnoise) and due to the difference in the relative images acquired on 13 October 2007 and 22 Decem- positions of the satellite at the time of acquisition ber 2007 were processed to generate a deforma- (Φorbital) (Chang et al. 2004). tion map using the DInSAR technique. These images have a temporal baseline of 70 days Φ = Φtopo + Φdisp + Φatm + Φnoise + Φorbital. (1) and a perpendicular baseline of 154 m. Landsat While using the DInSAR technique, the Φ ,was Enhanced Thematic Mapper Plus daytime images disp estimated by removing the phase differences con- over the study area (Path/Row: 140/44) were also tributed by other sources from the total on the LHS acquired for 19 November 2007 at 04:33:17.4738125 of equation (1). The Φ , was removed using a 3 GMT and 5 December 2007 at 04:33:23.8403750 topo arc-second (approx 90 m) photogammetric SRTM- GMT from Glovis (http://glovis.usgs.gov). Due DEM in two pass DInSAR technique (Massonnet to the unavailability of night time ETM+ imagery and Feigl 1998). The Φ , can either be accounted over the study area, scenes acquired during local atm for by using independent observations such as GPS winter were selected for this study. While the (Ge et al. 2003), or neglected if the tropospheric Digital Elevation Model (DEM) used in this study delay is considered to be homogeneous at the time was obtained from Shuttle Radar Topography Mis- of the radar image acquisition. In this study, the sion (SRTM) (Rodriguez et al. 2006), the required Φ term was neglected following the assumption orbit data was acquired from ESA Doris orbits atm that the magnitude and spatial extent of phase (DOR). Specific sightings of coal fire and polygonal delay produced by atmospheric fluctuations will be boundaries of surface and subsurface coal fire areas negligible in arid and semi-arid areas (Zhou et al. in the Jharia coalfield were produced from a digi- 2013), which is also the regional setting applica- tized version of the field survey map published by ble for the Jharia coalfield. Precise orbit data was Limited in their master plan conjunctively used to remove orbital phase differ- (BCCL 2008). ences (Φorbital) and to estimate the spatial baseline in order to register the DEM to SAR image coordi- nates. The phase difference Φnoise, is an artifact of 4. Methodology spatial and temporal decorrelation and the thermal noise of the radar instrument (Mora et al. 2003). 4.1 Processing of microwave data Errors in the removal of these phase differences In general, C band Envisat ASAR data have been account for the total error in the DInSAR-based proven to be very effective in some of the key estimates of subsidence. However, the magnitude missions outlined by the ESA. DInSAR technique of uncertainty depends on the terrain characteris- precisely quantifies the line-of-sight (LOS) dis- tics as well as on the quality of the datasets used placement, thus becoming an important tool for to remove these phase differences. the measurement of surface displacement from The displacement phase Φdisp, due to the differ- space. This displacement gets encoded in the phase ences in ground elevation in the two SAR images difference between two SAR images, which can is represented by equation (2). be measured at each point (pixel) of a phase dif-   4π ference image, which is termed as an interfero- Φdisp = − δR (2) gram. An interferogram is generally created from λ Monitoring subsurface coal fires in Jharia coalfield 1253 where δR is the height displacement along the coal fire. The selection of threshold temperature is slant-range direction and radar wavelength is rep- dependent on the geology of the area since radiant resented by λ. The measured phase in an inter- energy has to pass through these less conductive ferogram is a multiple of 2π. subsurface materials (Saraf et al. 1995). Here, the Thus a C-band microwave signal (λ = 5.6 cm), threshold temperature was refined in the context a complete cycle of 2π phase change will repre- of threshold temperature estimates established in sent a displacement of λ/2, i.e., 2.8 cm between the previous studies (Martha et al. 2010;Mishraet al. radar and each target on the surface. In this study, 2011). The final thermal anomaly map, referred phase unwrapping of interferograms, required to to as coal fire map, was produced on the basis generate a displacement map, was performed, of the consideration that pixels with temperature using Statistical-Cost, Network-Flow Algorithm <33◦C would represent background temperature, for Phase Unwrapping (SNAPHU) (Chen and pixel temperatures lying between 33◦ and 40◦C Zebker 2002). would represent subsurface coal fires and pixels with temperature >40◦C would represent surface coal fires. Actual land surface temperature was not taken into consideration as our primary aim was to 4.2 Processing of Landsat ETM+ image qualitatively determine the location and extent of Emitted thermal radiations from the Earth’s sur- surface and subsurface fire areas. face are precisely recorded by thermal sensors The Scan Line Corrector (SLC) of Landsat and the information is stored as digital numbers ETM+ instrument failed on 31 March 2003, since (0–255). Thermal anomaly of the Earth’s surface then all Landsat ETM+ images have wedge-shaped created by surface and/or subsurface coal fires can gaps on both sides of each scene. The net effect is be detected and mapped by converting these digital that approximately 22% of the pixels in a Landsat- numbers to radiant temperature and subsequently 7 scene are lacking any kind of information. A tech- to kinetic temperature. Here, scene specific radi- nique developed by Scaramuzza et al. (2004)which ance values were calculated from raw digital values fills the gaps in one scene with data from another using equation (3) (Markham and Barker 1986), Landsat scene is used to correct for the missing data pixels. In this technique, a local linear his- − Lmax(λ) Lmin(λ) togram matching is carried out to find a linear Lλ = Qcal (3) Qcal max transformation from one image to another based on the standard deviation and mean values of each where Lmin(λ) = minimum detected spectral radi- band, of each scene. To obtain quality results, cloud ance of the scene, Lmax(λ) = maximum detected free images, with minimal time gap were used for spectral radiance of the scene, Qcal = grey level of the purpose of gap filling. the analysed pixel, Qcal max = maximum grey level (255). Once the digital numbers for the thermal band have been converted to radiance values, the inverse of Planck’s function was applied to derive 5. Results its corresponding temperature values. The radiant temperature was determined using the following 5.1 Identification of areas undergoing subsidence equation: using DInSAR technique

K2 The displacement map produced using a pair T = (4) (2007/10/13–2007/12/22) of SAR acquisitions for ln (K1/Lλ − 1) monitoring vertical ground movement in the study 2 where K1 = calibration constant (666.09 W/m /sr/μm) area is shown in figure 2. Results, in general, and K2 = calibration constant (1281.71 K) (Landsat-7 show that the study area has experienced vary- Science Data Users Handbook 2008). The radiant ing degrees of subsidence as well as upliftment. temperature was then converted to kinetic tem- Since the focus of the present work is on analysing perature with an emissivity (ε) value of 0.96 using subsidence related to coal fires, we have followed equation (5). the convention of assigning positive and negative 1 values, in the displacement map, as respective T = T × . (5) kin rad 1/4 measures of subsidence and upliftment in the ε study area. Computation of thermal anomaly maps reflect- The displacement map (figure 2) shows that ing the presence of surface and subsurface coal both the eastern flank and western flank of the fires, using Landsat-7 ETM+ thermal band 6, also Jharia coalfield have witnessed major subsidence, involved the selection of a threshold temperature primarily due to coal fires. Areas such as Kujama, to discriminate between pixels with and without Kusunda, , North , Bararee and 1254 Nishant Gupta et al.

86º12'0"E 86º18'0"E 86º24'0"E 86º30'0"E N 23º48'0"N 23º48'0"N 23º42'0"N 23º42'0"N

Subsidence (mm) 28

-25

23º36'0"N 86º12'0"E 86º18'0"E 86º24'0"E 86º30'0"E

Figure 2. Displacement map corresponding to the SAR image pair (2007/10/13–2007/12/22). Illustrated is vertical dis- placement measured in millimetres. A positive value of vertical displacement indicates subsidence whereas negative vertical displacement values indicate upliftment.

Lodna in the eastern flank and Phularitand, These observed upliftments can be in-part due to Damuda and in the western flank are some the preventive measures adopted to control the of the major areas showing high magnitudes of sub- propagation and remedy the immediate hazardous sidence. On an average, displacement values ranged implications of coal fire. These measures include between −25 mm (upliftment) and 28 mm (subsi- surface sealing and blanketing with cohesive and dence) with a mean value of 10 mm measured over incombustible materials like sand and soil. It has a period of 70 days in Jharia coalfield. been documented that over 22 million cubic metres The primary cause of subsidence observed here is of surface blanketing work has been carried out in the occurrence of subsurface coal fires. Since min- the area (BCCL 2008). Furthermore, these areas ing activities are limited both in terms of spatial of upliftment are also noted to be in close proxim- extent and magnitude, use of small temporal base- ity to major opencast mines and hence it is most line between the SAR image pairs limit the impact likely that these upliftments are caused due to the of mining-related subsidence in the displacement dumping of overburden materials from the nearby map (figure 2) thus produced. In addition, the mag- opencast mines. nitude of subsidence induced due to mining activi- In order to validate the cause of subsidence ties is generally assumed to be negligible since areas observed in the displacement map (figure 2), the with coal fires normally have insignificant mining DInSAR-based results were further corroborated activities because of the associated hazards. Varia- with a field survey map (BCCL 2008). This field tions in the magnitude of subsidence (high to low) survey map was prepared on the basis of demo- are caused by the heterogeneity in the intensity of graphic as well as scientific field surveys. Figure 3 coal fires. This varying intensity of coal fires can be shows the displacement map overlaid with georef- attributed to diversity in the coal seam layer and erenced polygonal boundaries (pink colour) rep- its proximity to land surface, via cracks and/or fis- resenting areas with surface and subsurface coal sures, that act as pathways for supplying oxygen fires (BCCL 2008). This was done to emphasize required for the propagation of coal fires. the direct correlation between the spatial distribu- Also observed in displacement map are minor tion of surface subsidence and the co-occurrence areas of upliftment such as in Tisra and Bararee. of subsurface coal fires. For the ease of visibility, Monitoring subsurface coal fires in Jharia coalfield 1255

86°12'0"E 86°18'0"E 86°24'0"E 86°30'0"E N 23°48'0"N 23°48'0"N 23°42'0"N 23°42'0"N

Magnitude of Displacement

-10mm - 0mm (Upliftment) 0mm - 10mm (Subsidence) 10mm - 27.8mm (Subsidence) Jharia coal field

23°36'0"N 86°12'0"E 86°18'0"E 86°24'0"E 86°30'0"E

Figure 3. Displacement map overlaid with georeferenced polygonal boundaries (pink) representing in situ observations of coal fire. The colours represent varying degrees of subsidence classified as high (red), low (blue) and negative/upliftment (white). displacements observed in this map were fur- the two Envisat ASAR images, which were pro- ther classified into three classes and colour coded cessed to obtain a displacement map (figure 2). accordingly. Red and blue colours represent high This was done so as to achieve maximum corre- (10 mm–27.8 mm) and low (0 mm–10 mm) mag- lation between the results obtained from Landsat- nitude of subsidence respectively, whereas white 7 ETM+ thermal band data and Envisat ASAR colour represents upliftment (−10 mm–0 mm) (fig- data. Figure 4 reveals that the eastern flank of ure 3). Figure 3 reveals a strong positive correlation Jharia coalfield is more affected by surface and sub- between areas showing subsidence and established surface coal fires than the western flank. Collieries coal fire sightings. Majority of the high magnitude like Rajapur, Tisra, , Ena industry, Kusunda subsidence, with a maximum value of 27.8 mm, and Kujama in the eastern flank and Shatabdi were observed in parts of Block II, Shatabdi OCP, OCP, Block II and Damuda show strong thermal , Mudidih, , Bassuriya, Rajapur and anomaly signals due to coal fires. Lodna collieries (figure 3), which are also areas The effectiveness of the DInSAR technique to renowned for repeated instances of coal fire related monitor coal fires is highlighted by the compari- hazards. son of observed land subsidence and merged anal- ysis of ancillary datasets depicting coal fire zones. 5.2 Validation using thermal IR Landsat-7 Results revealed that most of the areas affected ETM+ data by subsidence, in the displacement map (figure 2), are also areas with large thermal anomalies and Figure 4 shows the coal fire map overlaid by reported sightings of coal fire (figure 4). But, per- georeferenced polygonal boundaries (purple colour) haps most importantly, land subsidence, measured representing the spatial extent of surface and sub- as a surrogate of coal fires, was also observed in surface coal fire obtained from the field survey areas outside those demarcated in figure 4. Hence, map (BCCL 2008). As mentioned earlier, the coal these areas were classified as potential subsurface fire map illustrated in figure 4 was generated coal fire areas. These potential subsurface coal from Landsat-7 ETM+ thermal band 6. This ther- fire areas in the displacement map (figure 2)show mal anomaly map represents pixel integrated tem- subsidence signatures but does not show anoma- perature from surface and subsurface coal fires. lous thermal values and are thus rendered unde- The image selected for this purpose was acquired tectable in the merged analysis of in situ and during the period between the acquisition dates of remotely sensed observations (figure 4). This is also 1256 Nishant Gupta et al.

86°12'0"E 86°18'0"E 86°24'0"E N 23°48'0"N 23°48'0"N 23°42'0"N 23°42'0"N

Temperature < 33°C 33°C - 40°C > 40°C

86°12'0"E 86°18'0"E 86°24'0"E 23°36'0"N 23°36'0"N

Figure 4. Coal fire map overlayed by georeferenced polygonal boundaries (purple) representing in situ observation of coal ◦ ◦ fire. The colours depict thermal anomaly with respect to reference temperatures of 33 Cand40C and thus represent locations of surface (red), subsurface (yellow) coal fires within a background temperature (blue). true for some of the areas within the polygonal 2007) and Landsat ETM+ band 6 thermal data boundaries marked in figure 3. This lack of resolve (Mishra et al. 2011) used in association with can be attributed to the coarse geometric resolu- ASTER data for mapping of coal fires in Jharia tion (60 m) of band 6 (thermal) of Landsat-7 due coalfield (Martha et al. 2010). However, it is to which the thermal anomalies generated by coal important to note that the use of DInSAR tech- fires on the surface occupy only a fraction of the nique presented a unique opportunity to explore pixel size of Landsat-7 ETM+ band 6 image (Voigt the capability of locating potential subsurface et al. 2004; Kuenzer et al. 2007). As a consequence coal fires in the vicinity of those specified in of which, initiation of fresh subsurface coal fires, previous studies. The relevance of this study is due to the progression of opencast or underground embellished by its capability of demarcating the mining activities, having a thermal anomaly spread boundaries of hazardous areas undergoing high over a small area (Chatterjee et al. 2007)couldnot magnitudes of subsidence, as a consequence of coal be detected in the coal fire map. Alternatively, sub- fires, as well as those with potential hazardous surface coal fires present at great depths are also consequences. This has major implications for the unable to produce surface thermal anomaly that management, selection and application of preven- are significant enough to be detected by the ther- tive measures in order to mitigate the environmen- mal band of Landsat-7 ETM+ or during field sur- tal and socio-economic impacts of coal fires. vey. Further, the accuracy of detection of thermal anomalies from daytime Landsat-7 ETM+ images is affected by many false anomalies related to topo- graphic and solar irradiation effects (Voigt et al. 6. Conclusion 2004; Kuenzer et al. 2007). The results presented here are also in concur- Jharia coalfield is the primary source of coking coal rence with some of the earlier studies carried out in India. It is, however, associated with surface to map the surface and subsurface coal fire in and subsurface coal fires which lead to considerable the Jharia coalfield using various remote sensing wastage of this valuable non-renewable resource of data. The results obtained here are comparable fossil fuel. In addition, coal fires are almost always with those obtained from the analysis of NOAA/ associated with environmental hazards and have a AVHRR data (Agarwal et al. 2006), Landsat TM major impact on socio-economic structure of the thermal IR data (Chatterjee 2006; Chatterjee et al. region. Monitoring subsurface coal fires in Jharia coalfield 1257

In the study we used observations of vertical sub- Vegas: InSAR reveals structural control of land subsi- sidence from Envisat satellite to characterize and dence and aquifer-system deformation; Geology 27(6) quantify the subsurface coal fire occurring in the 483–486. Bharat Coking Coal Limited 2008 Master plan for deal- Jharia coalfield. Results revealed that major belts ing with the fire, subsidence and rehabilitation in the of subsidence are located in the eastern and western leasehold of Bharat Coking Coal Limited, Project and flanks of the Jharia coalfield. The estimates thus Planning Division, Dhanbad, India. obtained were further validated with information Bhattacharya A, Reddy C S and Mukherjee T 1991 Multi- from in situ field surveys and satellite based ther- tier remote sensing data analysis for coalfire mapping in Jharia coalfield of Bihar, India; Proc. of the 12th Asian mal infrared data from Landsat ETM+ satellite. Conference on Remote Sensing, Singapore, 30th October– Major areas of subsidence correlated very well with 5th November 1991 (Singapore: National University of the field survey map as well as the coal fire map Singapore) 22 1–6. derived from thermal remote sensing. The results Carnec C, Massonnet D and King C 1996 Two exam- demonstrated the potential of DInSAR technique ples of the use of SAR interferometry on displacement fields of small spatial extent; Geophys. Res. Lett. 23(24) for identification and mapping of coal fire zones 3579–3582. using vertical subsidence as a surrogate of coal Chang H C, Ge L and Rizos C 2004 DInSAR for Mine fire. The areas with significant subsidence outside Subsidence Monitoring Using Multi-source Satellite SAR those which have been correlated with ancillary Images; http://www.fig.net/pub/jakarta/papers/ts 19/ data were classified as potential coal fire zones. ts 19 3 chang etal.pdf. These potential coal fire zones were hypothesized Chatterjee R S 2006 Coal fire mapping from satellite thermal IR data – A case example in Jharia coalfield, Jharkhand, to show subsurface signatures of coal fire but cur- India; ISPRS J. Photogramm. 60 113–128. rently undetected by in situ and thermal remote Chatterjee R S, Fruneau B, Rudant J P, Roy P S, Frison P, sensing techniques. The method and results pre- Lakhera R C, Dadhwal V K and Saha R 2006 Subsi- sented here have major implications for classifying dence of Kolkata (Calcutta) City, India during the 1990s and marking boundaries of coal fire zones leading as observed from space by Differential Synthetic Aper- ture Radar Interferometry (D-InSAR) technique; Remote to improved implementation of pre-emptive meth- Sens. Environ. 102 176–185. ods of fire control and safety. In addition, the Chatterjee R S, Wahiduzzaman Md, Shah A, Raju E V R, combination of InSAR and Thermal IR techniques Lakhera R C and Dadhwal V K 2007 Dynamics of coal provide a high potential for an improved monitor- fire in Jharia Coalfield, Jharkhand, India during the 1990s ing of areas affected by coal fires. But, it is impor- as observed from space; Curr. Sci. 92 61–68. Chen C W and Zebker H A 2002 Phase unwrapping for tant to note that subsidence due to coal fires is large SAR interferograms: Statistical segmentation and most likely a discontinuous process, thus the pro- generalized network models; IEEE T. Geosci. Remote 40 cess itself is hard to monitor. While the utilization 1709–1719. of DInSAR technique has proved to be an impor- Colesanti C, Le Mouelic S, Bennani M, Raucoules D, Carnec tant tool for monitoring coal fires in the region, the C and Ferretti A 2005 Detection of mining related ground instabilities using the Permanent Scatterers technique – socio-economic significance of the coal fire warrants a case study in the east of France; Int. J. Remote Sens. detailed GPS-based field investigations to validate 26(1) 201–207. the potential fire areas demarcated in this study as Galloway D L, Hudnut K W, Ingebritsen S E, Phillips S P, well as further validation of this technique in other Peltzer G, Rogez F and Rosen P A 1998 Detection of coal fire prone zones. aquifer system compaction and land subsidence using interferometric synthetic aperture radar, Antelope Valley, Mojave Desert, California; Water Resour. Res. 34(10) Acknowledgements 2573–2585. Gautam R S, Singh and Dharmendra Mittal A 2008 An effi- cient contextual algoirthm to detect subsurface fires with Envisat original data is copyright European Space NOAA/AVHRR data; IEEE T. Geosci. Remote 46(7) Agency and was provided under the Category-1 2005–2015. Proposal 9370. Authors would also like to thank Ge L, Chang H C, Janssen V and Rizos C 2003 The inte- the anonymous reviewer for the comments and gration of GPS, radar interferometry and GIS for ground deformation monitoring; Int. Symp. on GPS/GNSS, suggestions, which has significantly improved the Tokyo, Japan, 15–18 November, pp. 465–472. quality of the manuscript. Herrera G, Tomas R, Lopez-Sanchez J M, Delgado J, Mallorqui J J, Duque S and Mulas J 2007 Advanced DInSAR analysis on mining areas: La Union case study References (Murcia, SE Spain); Eng. Geol. 90 148–159. Jiang L, Lin H, Ma J, Kong B and Wang Y 2011 Poten- Agarwal R, Singh D, Chauhan D S and Singh K P 2006 tial of small-baseline SAR interferometry for monitoring Detection of coal mine fires in the Jharia coal field using land subsidence related to underground coal fires: Wuda NOAA/AVHRR data: Detection of coal mine fires in the (Northern China) case study; Remote Sens. Environ. 115 Jharia coal field using NOAA/AVHRR data; J. Geophys. 257–268. Eng. 3 212–218. Jung H C, Kim S W, Jung H S, Min K D and Won J S Amelung F, Galloway D L, Bell J W, Zebker H A and 2007 Satellite observation of coal mining subsidence by Laczniak R J 1999 Sensing the ups and downs of Las persistent scatterer analysis; Eng. Geol. 92 1–13. 1258 Nishant Gupta et al.

Kuenzer C, Zhang J, Li J, Voigt S, Mehl H and Wagner W Prakash A, Gupta R P and Saraf A K 1997 A Landsat T M 2007 Detecting unknown coal fires: Synergy of auto- based comparative study of surface and subsurface fires mated coal fire risk area delineation and improved ther- in the JCF, India; Int. J. Remote Sens. 18 2463–2469. mal anomaly extraction; Int. J. Remote Sens. 28(20) Prakash A, Fielding E J, Gens R, Van Genderen J L and 4561–4585. Evans D L 2001 Data fusion for investigating land subsi- Landsat-7 Science Data Users Handbook-Level 1G dence and coal fire hazards in a coal mining area; Int. J. 2008 Product-Band 6 conversion to temperature; Remote Sens. 22 921–932. http://landsathandbook.gsfc.nasa.gov/pdfs/Landsat-7 Reddy C S S, Srivastav S K and Bhattacharya A 1993 Appli- Handbook.pdf. cation of Thematic Mapper short wavelength infrared Markham B L and Barker J L 1986 EOSAT Landsat Tech- data for the detection and monitoring of high temper- nical Notes 1. Earth Observation Satellite Co, Maryland, ature related geoenvironmental features; Int. J. Remote August 1986. Sens. 14 3125–3132. Martha T R, Guha A, Kumar K V, Kamaraju M V V and Rodriguez E, Morris C S and Belz J E 2006 A global assess- Raju E V R 2010 Recent coalfire and land-use status of ment of the SRTM performance; Photogramm. Eng. Rem. Jharia coalfield, India from satellite data; Int. J. Remote Sens. 72(3) 249–260. Sens. 31(12) 3243–3262. Saraf A K, Prakash A, Sengupta S and Gupta R P 1995 Massonnet D and Feigl K L 1998 Radar interferometry and Landsat-TM data for estimating ground temperature and its application to changes in the earth’s surface; Rev. depth of sub-surface coal fire in the Jharia coalfield, India; Geophys. 36 441–500. Int. J. Remote Sens. 16 2111–2124. Mishra R K, Bahuguna P P and Singh V K 2011 Detection of Scaramuzza P, Micijevic E and Chander G 2004 SLC gap- coal mine fire in Jharia coal field using Landsat-7 ETM+ filled products: Phase one methodology; Available at landsat. data; Int. J. Coal Geol. 86 73–78. usgs.gov/documents/SLC Gap Fill Methodology.pdf. Mora O, Mallorqui J J and Broquetas A 2003 Linear and Tom´as R, M´arquez Y, Lopez-Sanchez J M, Delgado J, nonlinear terrain deformation maps from a reduced set Blanco P, Mallorqu´ı J J, Martinez G H and Mulas J of interferometric SAR images; IEEE T. Geosci. Remote 2005 Mapping ground subsidence induced by aquifer over- 41(10) 2243–2252. exploitation using advanced Differential SAR Interfero- Motagh M, Walter T R, Sharifi M A, Fielding E J, Schenk A, metry: Vega Media of the Segura River (SE Spain) case Anderssohn J and Zshau J 2008 Land subsidence in Iran study; Remote Sens. Environ. 98(2) 269–283. caused by widespread water reservoir overexploitation; Voigt S, Tetzlaff A, Zhang J, Kunzer C, Zhukov B and Geophys. Res. Lett. 35.16 L16403. Strunz G 2004 Integrating satellite remote sensing tech- Mukherjee T K, Bandopadhyay T K and Pande S K 1991 niques for detection and analysis of uncontrolled coal Detection and delineation of depth of subsurface coalmine seam fires in North China; Int. J. Coal Geol. 59 fires based on an airborne multispectral scanner survey in 121–136. a part of the Jharia Coalfield, India; Photogramm. Eng. Wegmuller U, Strozzi T and Bitelli G 1999 Validation of ERS Rem. Sens. 57 1203–1207. differential SAR interferometry for land subsidence map- Pathier E, Fielding E J, Wright T J, Walker R, Parsons B E ping: The Bologna case study; In: Geoscience and Remote and Hensley S 2006 Displacement field and slip distribu- Sensing Symposium; IGARSS’99 Proceedings. IEEE 1999 tion of the 2005 Kashmir earthquake from SAR imagery; International 2 1131–1133. Geophys. Res. Lett. 33 L20310. Zhou L, Zhang D, Wang J, Huang Z and Pan D 2013 Map- Prakash A and Gupta R P 1999 Surface fires in JCF, ping land subsidence related to underground coal fires in India–their distribution and estimation of area and the Wuda coalfield (Northern China) using a small stack temperature from TM data; Int. J. Remote Sens. 20 of ALOS PALSAR differential interferograms; Remote 1935–1946. Sens. 5(3) 1152–1176.

MS received 7 December 2012; revised 9 April 2013; accepted 13 April 2013