Enhancing Thin Coal Seam Detection in Eastern Indian Coalfields Using ICWT-Decon-Based Seismic Attributes and Acoustic Impedance Inversion

Abstract

A high-resolution seismic survey (HRSS) is often used in coal exploration to bridge the data gap between two consecutive boreholes and avoid ambiguity in geological interpretation. The application of high-resolution seismic surveys in the Indian context is challenging as the delineation of thin non-coal layers within the coal layer requires a very high seismic data resolution. However, conventional seismic processing techniques fail to resolve thin coal/non-coal layers and faults, which is crucial for the precise estimation of coal resources and mine economics. To address these issues, we applied the inverse continuous wavelet transform deconvolution (ICWT-Decon) technique to post-stack depth-migrated seismic sections. We examined the feasibility of the ICWT-Decon technique in both a synthetic post-stack depth-migrated model and 2D/3D seismic data from the North Karanpura and Talcher Coalfields in Eastern India. The results offered enhanced seismic sections, attributes (similarity and sweetness), and acoustic inversion that aided in the precise positioning of faults and the delineation of a thin non-coal layer of 4.68 m within a 16.7 m coal seam at an approximate depth of 450 m to 550 m. This helped in the refinement of the resource estimation from 74.96 MT before applying ICWT-Decon to 55.92 MT afterward. Overall, the results of the study showed enhancements in the seismic data resolution, the better output of seismic attributes, and acoustic inversion, which could enable more precise lithological and structural interpretation.

1. Introduction

Core drilling is the most commonly used method in coal exploration worldwide, and, in India, it is especially used for the estimation of the coal quality and resources. According to the latest update of the Indian Standard Procedure (ISP) norms in 2022 for different stages of coal exploration, such as prospecting (G3), general exploration (G2), and detailed exploration (G1), the average distance between two consecutive boreholes is 1600 m, 800 m, and 400 m, respectively. Borehole data offer point information and the interpolation of the data between two consecutive boreholes is performed for geological interpretation using certain assumptions. Moreover, due to their drifted origins, Indian coals are banded in nature. The thickness of the non-coal bands within coal layers ranges from tens of centimeters to meters. Obtaining information about thin coal seams, i.e., thicknesses of >0.9 m and >1.0 m [1], is essential for resource estimation for the planning of underground and opencast mining. The lack of geological information between the two consecutive boreholes, especially in complex geological conditions, may cause difficulties in the identification of locations of possible seam abnormalities, faults, washout due to meandering paleochannels, zones of seam splitting, thinning, and merging, etc., which may result in ambiguity in geological interpretation, requiring additional boreholes to improve it [1]. In order to bridge the data gap between two successive boreholes and to optimize the number of boreholes in an area, high-resolution seismic surveys (HRSSs) are widely used worldwide in coal exploration [1,2,3,4,5]. These datasets also aid in the identification of faults [6,7,8,9], dykes [9], and mine voids [10,11,12], etc. However, the application of HRSSs in the Indian context is challenging due to the requirement of a seismic data resolution beyond λ/4 of the seismic wave to delineate thin coal seams, which is essential for mine planning. The enhancement of the seismic data resolution through advanced data processing and interpretation techniques is the primary objective of the present research work.

Among the different techniques of seismic data processing, deconvolution is an important process to enhance the resolution of seismic data. The convolutional deconvolution model was proposed based on the assumptions of a minimum phase stationary wavelet and white reflectivity function [13,14]. As an improvement to the earlier technique, Margrave et al. [15,16] proposed the short-time Fourier transform (STFT)-based Gabor deconvolution technique. Li et al. [17] further advanced the method by using spectral modeling and variable-step sampling with hyperbolic smoothing of the Gabor spectrum to derive the magnitude of the attenuation function and eliminate the source wavelet’s effect. The reflectivity series is then estimated using the source wavelet and attenuation function.

Gholami and Sacchi [18] proposed a blind deconvolution technique that maximizes the sparsity of the reflectivity by concurrently estimating the wavelet and reflectivity. This approach generates the overall reflectivity series using an average stationary wavelet. Gholami [19,20] extended this by simultaneously estimating the reflectivity and Q-factor, as well as the reflectivity and the time-varying phase of wavelets, to increase the seismic signal’s temporal resolution. Lari and Gholami [21] introduced a spatial and temporal time-varying wavelet-based blind deconvolution approach without pre-assumptions of the wavelet’s bandwidth and phase, effectively modeling and removing the blurring effects of the source signatures and the Earth’s filtering effects. Radad et al. [22] developed an S-transform with an energy concentration measure, applying an optimum window instantaneously on each time–frequency (TF) sample to generate a TF map, which was then compared with maps obtained through other methods, such as the STFT and standard S-transform. Yu et al. [23] adjusted the seismic signal’s phase before and after the synchrosqueezing transform by obtaining the dominant frequency value from the instantaneous frequency attribute to improve the time–frequency resolution.

On the other hand, many researchers have applied wavelet transform-based processing techniques to enhance the seismic data resolution. Sinha et al. [24] used a continuous wavelet transform (CWT) to identify faults, paleochannels, and potential hydrocarbon zones. Li and Liner [17] validated the Holder exponent attribute to identify geological features such as wedges and stratigraphic pinch-outs using a wavelet-based multiscale analysis algorithm. Li and Liner [25] demonstrated that wavelet-based multiscale singularity analysis enhances stratigraphic features and acoustic impedance inversion. Smith et al. [26] applied a CWT to recover lost wavelet signatures by computing the phase and amplitude spectra of harmonic and sub-harmonic frequencies and convolving them with raw data, thereby enhancing the signal’s bandwidth. This not only increased the seismic resolution but also reduced the ringy nature of the data, aiding in the enhancement of structural and stratigraphic features. Munadi et al. [27] successfully applied a CWT to seismic data for accurate fluid type identification in reservoir rock. Saeid et al. [28] applied CWT-based spectral decomposition to resolve the distribution of fluvial sediments and potential reservoirs in the Southwestern Cusiana giant oilfield’s footwall. Ali et al. [29] demonstrated the CWT’s efficiency on seismic data even in the presence of noise, and Zhao et al. [30] used a 3D CWT for pre-stack 3D seismic data de-noising.

All of the previous works have significantly advanced the deconvolution process, producing effective high-resolution results and demonstrating the techniques’ capabilities in oil and gas exploration. However, only a few researchers have implemented CWT techniques in coal seismic data, such as Haris et al. [31], who did so to estimate the coal bed methane properties in East Kalimantan, Indonesia. In this study, we applied a CWT to each seismic trace to calculate the wavelet transform modulus maxima lines (WTMML) and reconstructed each seismic trace with a Ricker wavelet for the enhancement of the frequency bandwidth of the seismic data. This approach helped to improve the overall structural interpretation, seismic attributes, and acoustic inversion to properly image the non-coal bands within the coal layers, which is a very common challenge in India for the accurate estimation of resources.

2. Geology of the Study Area

The Gonadwana Master Basin of Peninsular India is a group of four intra-cratonic sedimentary basins, namely the Son-Mahanadi Basin, Damodar Valley Basin, Satpura Basin, and Godavari Valley Basin, that contains lower Gondwana sequence rocks from the Upper Carboniferous to the Lower Triassic periods. These are graben- and half-graben-type basins formed by normal faults and growth faults resulting from basin rifting and sediment depositional overburden. Such fault-controlled subsidence created extensional basins that accommodated the deposition of thick piles of continental sediments [32]. The depositional sequence of the formations, from younger to older, of the Gondwana basins is Kamthi, the Barren Measures, Barakar, Karharbari, Talchir, and the Precambrian basement [33]. Surrounded by Precambrian terrains, these basins are significant for their huge coal reserves, resulting from the drifted deposition of Gondwana flora under favorable temperate and fluvial–lacustrine conditions.

Due to the variable lithological and sedimentary attributes that lead to difficulties in inter-basinal correlation, the formations of the Gondwana sequence, except Talchir and Barakar, have different names in different basins [32]. Glaciogenic Talchir is the oldest formation of the Gondwana sequence. The consistency of Talchir/Karharbari and the overlaying coal-bearing Barakar Formation, which are the lower part of the Gondwana succession, is one of the main features of all of the Gondwana basins [32,34,35,36,37]. A glacial environment with no flora and a glacial to temperate environment with less flora during the Upper Carboniferous and Early Permian lead to the Talchir Formation being devoid of coal. However, during the Lower Permian period, the presence of huge Gondwana flora and the fluvial-depositional conditions resulted in thick coal deposits in the Barakar Formation. The thick and consistent deposits of coal in the Barkar Formation across all of the Gondwana basins makes it the principal source of power-grade coal in India.

Out of these lower Gondwana basins, the Son-Mahanadi Basin consists of three coalfields, namely the Talcher, Ib Valley, and Korba coalfields, situated in Odisha and Chhattisgarh in the eastern part of India. The Damodar Valley Basin consists of the Jharia, Raniganj, West Bokaro, Giridih, East Ramgarh, Bokaro, North Karanpura, and South Karanpura coalfields.

In the present study, 2D and 3D seismic data acquired in two different coalfields, namely North Karanpura and Talcher, respectively, are used for the application of the ICWT-Decon technique. These two coalfields have contrasting sub-surface geological features, such as complex faults and coal quality, etc. The regional stratigraphic succession of the North Karanpura and Talcher coalfields is given in

Table 1. Regional stratigraphy of North Karnapura and Talcher coalfields.

North Karanpura Coalfield			Talcher Coalfield
Stratigraphic Divisions	Formation	Major Lithologies	Stratigraphic Divisions	Formation	Major Lithologies
Recent	Soil/Alluvium	Soil, sandy soil	Recent	Soil/Alluvium	Colluvial fills, sand, silt, and clay sandy soil
Unconformity			Unconformity
Jurassic (Post-Gondwana)	Igneous Intrusives	Mica peridotite Dolarite	Jurassic
Unconformity			Unconformity
Triassic (Upper Gondwana)	Mahadeva	Ferruginous sandstone with shale intercalations Massive coarse to conglomeratic Feldspathic sandstone	Triassic (Upper Gondwana)	Undifferentiated Kamthi	Fine to medium-grained light grey to reddish sandstone and shale Pale greenish sandstone with rare shale and pink clay bands, ferruginous coarse-grained to pebbly sandstone at top.
	Panchet	Shales, sandstones, conglomerates
Unconformity			Unconformity
Permian (Lower Gondwana)	Raniganj Formation	Sandstones, shales Coal seams	Permian (Lower Gondwana)
	Barren Measures	Siltstones Shales Fine sandstones		Barren Measures	Siltstones, shales, coarse to medium sandstones
	Barakar	Coarse sandstones, shales, coal seams		Barakar	Medium to coarse sandstones, grey to dark grey shales Coal seams
	Karharbari	Sandstones, shales, coal seams		Karharbari	Massive medium to coarse sandstones containing clasts Shales Coal seams
Upper Carboniferous to Lower Permian (Lower Gondwana)	Talchir	Sandstones, diamictite (glacial tillite), shales, boulder beds	Upper Carboniferous to Lower Permian (Lower Gondwana)	Talchir	Sandstones, diamictite (glacial tillite) Needle shales, boulder beds
Unconformity			Unconformity
Precambrian (Pre-Gondwana)	Archaean	Metamorphic Igneous rocks (schists and gneisses)	Precambrian (Pre-Gondwana)	Archaean	Metamorphic igneous rocks (schists and gneisses)

. The table is based on GSI Bulletin Series A No. 45, “Coalfields of India, 1983”.

2.1. North Karanpura Coalfield

The North Karanpura coalfield is situated in the upper reaches of the Damodar river and covers an area of 1230 sq km. This coalfield is elliptical in shape and has a maximum width of 64 km in the east–west and 32 km in the north–south directions (Figure 1). The major sedimentary succession of the coalfield comprises Raniganj, the Barren Measures, Barakar, Karharbari, and the Talchir Formation, which occur uncomfortably over the Precambrian metamorphic basement. Among these formations, Raniganj, Barakar, and Karharbari are the coal-bearing formations of the area. Similar formations are encountered in the study area where the 2D seismic survey is conducted. Among the above coal-bearing formations, there is high potential for coal seams in Barakar compared to the others in the area. The deeper sedimentary column of the study area suggests that the Dakra coal seam of the Barakar Formation is the thickest compared to the other coal seams, such as Karkata, Bishrampur, and Bhukbhuka. Further, the study area is intersected by a few faults having throws ranging from 20 m to 50 m. A representative cross-section of the study area in the North Karanpura coalfield is shown in Figure 2.

2.2. Talcher Coalfield

The Talcher coalfield is in the southeastern part of the Mahanadi Valley in the Gondwana Basin, situated within 20°50′ N to 21°15′ N latitudes and 84°21′ E to 85°25′ E longitudes [32], extended in an area of ~1860 sq km. The location of the coalfield is shown in Figure 1. The major lithoforms of the Talcher coalfield are Talchir, Karharbari, Barakar, and Kamthi, which lay over the Archaean basement rocks. Talchir has glacial and also pre-glacial sediments; on the other hand, potential coal, fine- to coarse-grained sandstones, shale, and conglomerates, etc., are the part of the Karharbari, Barakar, and Kamthi Formations [38,39,40]. The Barakar Formation comprises most of the major coal seams, except the deeper coal seam, which belongs to Karharbari, in the Talcher coalfield. A representative cross-section of the study area in the Talcher coalfield is shown in Figure 3.

3. Data and Methodology

The present research endeavor mainly focused on three aspects. The first was the enhancement of the seismic data resolution to demarcate the thinnest possible lithological layer by using inverse continuous wavelet transform deconvolution (ICWT-Decon). Further, it aimed to validate the proposed ICWT-Decon using synthetic seismic data and also both 2D field seismic data from the North Karanpura Coalfield and 3D seismic survey data from the Talcher Coalfield, Eastern India. Finally, it aimed to improve the results of spectral decomposition, seismic attributes, and seismic acoustic impedance inversion.

3.1. Inverse Continuous Wavelet Transform Deconvolution (ICWT-Decon)

Grossmann and Morlet [41] were the first to formulate the continuous wavelet transform (CWT) (a detailed description is given in Appendix A) in 1984, and Mallat and Zhong [42] demonstrated the use of the CWT to detect multiscale boundaries through the calculation of the WTMML. Further, the application of complex wavelets was shown by Tu and Hwang [43]. With the help of a logarithmic scale for the WTMML, the Holder or Lipschitz coefficient can be obtained. Herman and Stark [44] and Li and Liner [25] presented the use of the Holder exponent in seismic processing to characterize acoustic impedance contrasts, while the whole WTMML was used to cluster different seismic facies by de Matos et al. [45]. Mallat and Zhong [42] approximately reconstructed the raw trace using a multiscale edge depiction, while Teolis and Benedetto [46] showed the reconstruction of the seismic trace with the help of coefficients of a CWT.

Inverse continuous wavelet transform deconvolution (ICWT-Decon) was proposed by de Matos and Marfurt [47]. In this method, first, a continuous wavelet transform (CWT) is applied on every seismic signal of the PSDM section for the calculation of the WTMML from the CWT magnitude. Further, each seismic trace is reconstructed. This step resembles an inverse filtering operation. Thus, the entire process is called inverse continuous wavelet transform deconvolution (ICWT-Decon). In comparison with conventional deconvolution techniques, ICWT-Decon leverages the CWT’s ability to highlight or enhance the minute singularities present in the seismic data that are not visible to the human eye or through conventional data processing.

Processed PSTM/PSDM 2D seismic sections or 3D seismic volumes, along with geophysical log data—especially density and sonic logs—are prerequisites for the application of ICWT-Decon. The seismic data are initially subjected to a CWT to generate a wavelet scalogram of each seismic trace. From the wavelet scalogram, WTMM lines are extracted. These WTMM lines are convolved with a source wavelet of the desired frequency to reconstruct each seismic trace. A flow chart of the methodology followed in the present study is given in Figure 4.

3.2. Synthetic and Filed Seismic Data

To generate synthetic seismic data, we considered the density and sonic log data from well 01, with key horizons and high-angle faults corresponding to the geological model envisaged in the area, as mentioned in Section 2. The envisaged model is in the depth domain and is converted to the time domain using a velocity model generated from the time–depth relationship derived from the sonic data. Further, the model is convolved with the wavelet to generate data with a broadband frequency range of 5–10–50–60 Hz to replicate typical field seismic data. A depth section generated using the synthetic data is shown in Figure 5.

In our study, these synthetic data are considered as actual data. The resolution of such data is limited to 15 m for an average velocity of 2400 m/s. However, with the same velocity, the desired bandwidth of seismic data should be 5–10–150–160 Hz to resolve a thin bed of 5 m. This is considered the desired data. A comparison of a single seismic trace from the actual and desired seismic data is shown in Figure 6.

Further, the CWT using the complex Morlet wavelet (details of the selection of the wavelet are given in Appendix A) as a basis function is applied to the synthetic seismic traces. The WTMM lines, which were identified from the wavelet scalogram, are used to draw impedance contrasts down each trace. Further, each seismic trace is reconstructed by convolving it with the desired wavelet (Figure 7).

3.3. Seismic Acoustic Impedance Inversion

The process of converting reflection seismic data to impedance is called acoustic impedance inversion. This process is based on the assumption that all recorded seismic reflections are normal incident. The basic forward model for all seismic inversions is defined as the convolution of the seismic wavelet with the Earth’s reflectivity series, which can be expressed as follows [48]:

$$T\left(t\right)=R_i\left(t\right)\ast W\left(t\right)+\mathrm{ambient}\mathrm{noise}$$

(1)

where T(t) is the seismic trace, $R_i\left(t\right)$ is the Earth’s reflectivity, and W(t) is the source wavelet. The Earth reflectivity series is defined as follows:

$$R_i={\displaystyle \frac{{\left(AI\right)}_{i+1}-{\left(AI\right)}_i}{{\left(AI\right)}_{i+1}+{\left(AI\right)}_i}}$$

(2)

where

$R_i$ is the reflection coefficient of the i^th layer;

${\left(AI\right)}_i$ is the acoustic impedance (product of acoustic velocity and density) of the i^th layer.

Post-stack acoustic impedance for the present work is carried out using the maximum likelihood amplitude inversion algorithm from the quantitative seismic interpretation module of the Paradigm software (Aspen Technology, Inc., Bedford, MA, USA). The algorithm directly works with the impedance model. The impedance model is convolved with the wavelet to obtain the synthetic seismic trace, which is compared with the actual seismic trace to identify the differences between these traces. Subsequently, the impedance model is updated and the process is repeated. The iteration process ends when the defined optimization value between the two models is achieved or we reach the defined maximum iterations.

To carry out post-stack inversion, three main inputs are required, i.e., the initial Earth model, seismic data (stack section or 3D volume), and wavelet. The initial Earth model is also called a low-frequency model (LFM) and is prepared using the filtered P-impedance log in all available boreholes and grids of different horizons as per the order of deposition in the subsurface. The seismic data and a suitable single wavelet or the average of the wavelets extracted during the well to seismic tie from different locations in the study area are used to carry out post-stack seismic inversion and obtain the acoustic impedance of the area. In the present research, post-stack seismic inversion is carried out using field data from the Talcher and North Karanpura coalfields.

4. Results and Interpretation

4.1. Synthetic Seismic Data

The process of the reconstruction of seismic traces is able to preserve the high-frequency content as there are no external factors that affect the frequency content of the source wavelet with the depth. To appreciate the output of ICWT-Decon, a comparison of the seismic sections before deconvolution, after predictive deconvolution, and after the application of ICWT-Decon on the PSDM synthetic data, along with their respective frequency spectra, is shown in Figure 8. From Figure 8a, it is observed that the frequency of the synthetic seismic section before deconvolution was 10 Hz to 100 Hz, with a dominant frequency of 50 Hz. On the other hand, after the application of conventional predictive deconvolution, the spectrum became flat and there was an enhancement in the spectrum power, with no notable change in the frequency range. However, this change led to an improvement in the resolution of the seismic section, highlighted in Figure 8a,b with an ellipse-shaped mark. Further, with the application of ICWT-Decon on the same synthetic data (Figure 8c), the frequency range was drastically augmented to 15 Hz to 220 Hz, with a corresponding improvement in the seismic section that was very distinct, shown in the encircled part. This improvement in the strength of the signature of the seismic reflectors and the frequency range after the application of ICWT-Decon on the synthetic data encourages the adoption of the technique for field seismic data in the study area in the North Karanpura coalfield (NKCF).

4.2. Two-Dimensional Seismic Field Data from North Karanpura

In one of the coal blocks of the North Karanpura coalfield, 2D seismic data of 48 folds are acquired with a group interval of 10 m, a shot interval of 10 m, 96 active channels, and a sampling interval of 0.5 ms. The seismic source used for this survey is, using the Sercel NOMAD-15. A linear sweep with a length of 12 s, with a listening time of 1.5 s, is used for data acquisition. In the study area, the zone of interest is between 548.73 m and 559.67 m in the Barakar Formation, wherein the top of coal seam II is at 548.73 m and the bottom is at 565.43 m, while the top and bottom of the non-coal band are at 554.99 m and 559.67 m, respectively. This is confirmed using multi-parametric geophysical log data, where the density and natural gamma have lower values and the resistivity and P-slowness have higher values, while the opposite is observed for coal zones and non-coal zones, respectively (Figure 9).

For the present study, 2D seismic data acquired along four profile lines, namely PL-01 to PL-04, are considered. The PSDM sections are subjected to the ICWT-Decon process. The output seismic sections after the application of the ICWT-Decon process are labeled as the SPE-PSDM sections.

The comparison of the PSDM sections with the SPE-PSDM sections of the seismic data along profile PL-01 show (Figure 10) an enhancement in the frequency spectrum of the PSDM-SPE section over the PSDM section. Before the application of ICWT-Decon, the range of the frequency bandwidth is around 15 Hz to 160 Hz, with a dominant frequency of around 25 Hz (Figure 10c). After the application of ICWT-Decon, the frequency bandwidth changes to 15 Hz to 180 Hz, with a dominant frequency of around 50 Hz (Figure 10d), with a relatively flat spectrum. Notable changes in resolution are marked with cyan-colored dotted ellipses. This enhancement in the frequency increases the vertical resolution of the seismic data. The deflection of the wiggles of the seismic traces at 548.73 m and 565.43 m, respectively, in the PSDM section helps to delineate the top and bottom of the coal seam (Figure 10a). However, there is no deflection in the wiggles within the top and bottom of the coal horizon to trace different horizons. On the other hand, in the SPE-PSDM section (Figure 10a), deflection in the wiggles at the top and bottom of the coal seam and within the coal seam is observed, which enables us to delineate the top and bottom of the non-coal horizon at 554.99 m and 559.67 m, respectively. From the overlay of the lithological markers from well 01 and the intersections of PL-02 on the PSDM section of PL-03 (Figure 11 and Figure 12), it is evident that there is no distinct deflection in the seismic wiggles within a depth range of 550 m to 565 m. Meanwhile, the seismic wiggles in the SPE-PSDM section have clear signatures at the top and bottom of the non-coal horizons at 554.99 m and 559.67 m, respectively. Similar observations can also be made for the seismic lines at PL-04 (Figure 13).

4.3. Three-Dimensional Seismic Field Data from Talcher Coalfield

In one of the coal blocks of the Talcher coalfield, 3D seismic data are acquired with a group interval of 12 m, a shot interval of 24 m, 340 active channels per receiver line, a sampling interval of 0.5 ms, receiver line spacing of 150 m, 14 active receiver lines, and shots per salvo of 35. The seismic source used for data acquisition in this survey is explosive. From these data, we use the 3D-PSDM seismic volume of 1.89 sq km and the log data from borehole 1 for the application of ICWT-Decon. The output seismic sections after the application of the ICWT-Decon process are labeled as the SPE-3D-PSDM sections. From the multi-parametric geophysical log interpretation of borehole 01, the top and bottom of the coal seam were identified at depths of 533.90 m and 567.25 m, respectively. Additionally, the non-coal zone within the coal seam was pinpointed at depths of 541.85 m (top) and 555.0 m (bottom) (Figure 14).

The application of ICWT-Decon on the 3D seismic data of the Talcher coalfield resulted in an enhancement in the bandwidth of the frequency spectrum. Specifically, the bandwidth, which was initially 10–90 Hz in the 3D-PSDM section, saw an increase to 10–160 Hz in the SPE-3D-PSDM (Figure 15). Notably, when the SPE-3D-PSDM was employed in conjunction with borehole 01, significant improvements in the lithological interpretations were noted.

A comparison between the 3D-PSDM and SPE-3D-PSDM sections, as depicted in Figure 15, reveals notable differences and also an improvement in the frequency spectrum, as the dominant frequency rose from around 55 Hz to 90 Hz. In the 3D-PSDM section (Figure 15a), the seismic traces’ wiggles show minimal deflection within the top and bottom of the coal horizon, i.e., within the depth range of 533.90 m to 567.25 m. The interval velocity is around 2400 m/s, making it challenging to trace distinct horizons. Conversely, in the SPE-3D-PSDM section (Figure 15b), the wiggles of the seismic traces prove instrumental in delineating both the top and bottom of the non-coal horizon, at a depth of 541.85 m and 555.00 m, respectively, within the coal horizon. These findings underscore the efficacy of the ICWT-Decon technique in coal exploration, particularly in enhancing the resolution.

4.3.1. Seismic Attribute Analysis

The spectral decomposition applied to both the PSDM and SPE-PSDM sections of the PL-01 to PL-04 2D seismic data is shown in Figure 16. The PSDM sections of all four profiles show that the signatures of the horizons are not sharp enough on either side of the faults at SPE F1-F4, whereas the prominent signatures of the horizons can be seen in the SPE-PSDM sections. This improvement in the results of spectral decomposition enables the accurate positioning of the fault in the SPE-PSDM section.

Similarly, the application of spectral decomposition to both the 3D-PSDM and SPE-3D-PSDM slices, illustrated in Figure 17, reveals noteworthy insights. In the 3D-PSDM sections (Figure 17a), an NW–SE-trending feature indicates a major fault, which is also corroborated with the displacement of the horizons in the 3D-PSDM sections at different intervals. However, in terms of tracking the continuity of the fault in the 3D-PSDM sections, the signature is not prominent; in particular, the marked feature lacks sharpness. On the other hand, the same feature shows a crisp and distinct signature in the SPE-3D-PSDM section (Figure 17b).

The similarity and sweetness attributes were applied to both the 3D-PSDM and SPE-3D-PSDM slices and the results are shown in Figure 18 and Figure 19. The comparison of the similarity and sweetness attribute results before and after the application of the ICWT-Decon process demonstrates the improvement in the seismic image in terms of continuity and sharpness. This enhancement in the seismic attribute results contributes to the precise placement of faults and the identification of structural complexities, etc.

4.3.2. Analysis of Acoustic Impedance Inversion

Due to the drifted origins of Indian coal seams, subtle changes within the seams are very common and have impact on coal mining, especially in underground coal mining. These changes are prominent in the borehole data (core or geophysical logs), which are a type of information that needs extrapolation/interpolation to build a 3D geological model [48]. Conversely, in the case of seismic data, this lateral continuity is good, with an average/poor temporal resolution. Therefore, bringing these two geophysical measurements into one platform can lead to an Earth model with a good lateral and vertical resolution for better geological outputs. This can be achieved through seismic inversion. The prerequisites for seismic inversion are geophysical logs (density and sonic), well-processed 2D/3D PSTM/PSDM seismic data, and their interpretation. In this study, we demonstrate the effect of the seismic data resolution on the impedance results through the PSDM data before and after the application of ICWT-Decon.

The superimposition of the impedance log of well 01 and the acoustic impedance inversion results of both the PSDM and SPE-PSDM sections of profile PL-02 is shown in Figure 20. The low acoustic impedance, depicted in the magenta color, indicates a pure coal zone. The other colors indicate non-coal zones. The inversion results for the SPE-PSDM section clearly show a change in acoustic impedance within the coal layer, which indicates the presence of a non-coal layer; this is also corroborated by the log data signature. Similarly, the presence of a coal layer within the non-coal layer can be identified.

The superimposition of the impedance log of borehole 01 and the acoustic impedance inversion results of both the 3D PSDM and 3D SPE-PSDM sections of profile PL-02 is shown in Figure 21. At the marked portion in Figure 21b, it can be observed that there is no sign of splitting of the reflector. Similarly, no such change is observed in Figure 21a to indicate splitting. On the other hand, the result is boosted through the ICWT-Decon process and the splitting of the reflector signature is very distinct at the marked position in Figure 21d. This leads to the development of a splitting signature in the corresponding inversion result (Figure 21c). This shows the tremendous scope of application of ICWT-Decon in the seismic inversion domain.

5. Discussion

Although seismic surveys are a well-established technique for the mapping of the layout and disposition of coal seams and the understanding of their structural complexities, their application to the coal scenario in India is challenging due to the requirement of thin bed delineation for the improvement of structural interpretation and ungraded resource estimation. To showcase the usefulness of the present work, a comparison among three entrance geological sections is developed based on the interpretation of only borehole data, PSDM before the application of ICWT-Decon, and PSDM after the application of ICWT-Decon, as shown in Figure 22.

In the first scenario, presented in Figure 22a, the distance between well 01 and well 02 is 800 m and the top and bottom of the coal seam are encountered at approximately 550 m and 565 m, respectively, in well 01. In well 02, they are at 570 m and 585 m, respectively. This is a clear indication of a fault at 20 m. However, the placement of a fault based on only well data is insufficient and ambiguous. Under these circumstances, for coal exploration in India, geological interpretation is performed based on certain assumptions, such as the assumption that the faults in study area are usually 70° to 80° and should not intersect either of the boreholes. Accordingly, for the present condition, the fault can be marked anywhere between 232 m and 549 m from well 01 or 251 m to 568 m from well 02. This zone of ambiguity in fault placement is shown with a blue arrow in Figure 22a. Based on the fault placement, on either side of the fault, the top and bottom of the coal seams and non-coal seams are projected as straight lines. However, in reality, these horizons and faults may not be straight.

To bridge this data gap between two successive boreholes, seismic data are useful. After seismic well tying, the correlated reflectors present in the seismic sections guide the tracing of the top and bottom of the litho-horizon. For such tracing, the reflector should be thick enough or the traces must have prominent deflections to resolve the top and bottom of a particular horizon. This is the second scenario, presented in Figure 22b, where the top and bottom of the coal horizons can be traced as deflections in the seismic traces, which are prominent. Meanwhile, no deflection is visible within the coal zone to demarcate the non-coal part present, as per the well 01 data. The singularities related to the non-coal part are also recorded in the seismic data, which may not be enhanced to the desired extent through conventional processing techniques. Therefore, through the present research, the applicability of ICWT-Decon has been successfully demonstrated, enhancing the frequency, which in turn augments the results of seismic attributes and inversion. After applying ICWT-Decon, the seismic section became much clearer and also developed the desired deflections in the traces in the non-coal zone, along with changes in the top and bottom, which made it possible to trace all four desired horizons (Figure 22c). Therefore, from the third scenario, it is clear that ICWT-Decon is a powerful tool to enhance the seismic data resolution and it can be used in routine coal exploration to harvest the maximum available information from the acquired seismic data.

Seismic data do not provide coal quality information; therefore, the coal resource estimated from the data has only ungraded prognostic value. Furthermore, unresolved thin non-combustible dirt bands exist within the coal horizons, which leads to the overestimation of ungraded potential coal resources. Therefore, thin lithological layers and four types of faults from the SPE-PSDM sections, as well as the spectral decomposition results, were used to generate a faulted grid for all four horizons (Figure 23) to compute the ungraded coal resources. The resources calculated from the grids of the coal seam amounted to 74.96 MT, which includes 25% of resources from the non-coal portion (55.92 MT). From this, it is evident that the resolution of non-coal horizons within thick coal horizons is crucial for the improved estimation of ungraded coal resources.

Similarly, utilizing data from the SPE-3D-PSDM volume, spectral decomposition, the similarity and sweetness attributes, and the acoustic inversion results, four horizons—namely the coal top, non-coal top, non-coal bottom, and coal bottom—along with one fault were employed to generate final 3D seismic volume of the study area (Figure 24) and s. These grids wer then utilized to compute the ungraded coal resources. The calculated resources from the coal seam grids amounted to 51.10 million metric tons (MT), including approximately 30% derived from the non-coal portion (15.42 MT). This underscores the critical importance of resolving the non-coal horizons within thick coal horizons for the improved estimation of ungraded coal resources.

Seismic data processed through the ICWT-Decon method play a significant role in the G3/G2 stage of coal exploration by offering insights into the layout and disposition of coal seams, including thin beds, and providing structural information about the subsurface. In the absence of seismic data, planning for the G1 stage of coal exploration relies on anticipating the target depths of boreholes based on the available G4 stage borehole data. Consequently, the estimations of the drilling work, project costs, machinery requirements, and timelines are highly tentative, necessitating additional boreholes to address the geological uncertainties arising from this limited information.

Having prior seismic information enhances the authenticity of planning in the G1 stage of exploration. The structural information and prognostic estimates of ungraded coal resources obtained using seismic data at the G3/G2 level guide management in evaluating the project’s viability for further planning in the G1 stage of exploration. Specifically, leveraging the output of SPE software, version 1.0, which offers enhanced subsurface imaging, it is possible to reduce the geological uncertainties during project execution, optimizing the required number of exploratory boreholes.

6. Conclusions

In the present study, the application of the ICWT-Decon method to enhance the resolution of seismic data was demonstrated on synthetic 2D seismic data from the North Karanpura coalfield, Eastern India, and 3D seismic data from the Talcher coalfield, Eastern India, using in-house developed spectral enhancement (SPE) software. The major conclusions drawn from this study are as follows.

• Following the application of ICWT-Decon, a significant improvement in the seismic data resolution was observed, resulting in more robust attribute analysis outcomes.
• The heightened seismic resolution in the SPE-PSDM and SPE-3D-PSDM sections facilitated the identification of thin non-coal beds within the coal seams, a task not achievable through conventional processing and interpretation techniques. The spectral decomposition, similarity attribute, and sweetness attribute results obtained with the SPE-PSDM and SPE-3D-PSDM data also enabled us to accurately interpret lateral discontinuities and features such as faults.
• The acoustic impedance results of ICWT-Decon using the processed 2D and 3D seismic data highlight the heterogeneities, such as the presence of dirt bands, seam splitting and merging, structural disturbances, etc., present within the coal seams. The results also indicate the tremendous scope of ICWT-Decon in the seismic inversion domain.
• The resource estimation derived from the 2D seismic data in the North Karanpura coalfield after applying ICWT-Decon (74.96 MT) displayed a 25% variation from the conventional seismic output (55.92 MT). Similarly, for the 3D seismic data in the Talcher coalfield, the coal resource was overestimated by 30%. Consequently, the successful application of the SPE technique in coal exploration using 3D seismic data suggests the potential application of ICWT-Decon in seismic inversion, reservoir characterization, and the study of mechanical properties in different formations, etc.