Improvement of Coal Mining-Induced Subsidence-Affected (MISA) Zone Irregular Boundary Delineation by MT-InSAR Techniques, UAV Photogrammetry, and Field Investigation

Abstract

Coal mining-induced ground subsidence is a severe hazard that can damage property, infrastructure, and the environment in the vicinity when the deformation is not negligible. The boundary of a mining-induced subsidence-affected zone refers to the area beyond which the ground subsidence is less concerned. Accurately measuring mining-induced ground deformation is essential for delineating the irregular boundary of the impacted area. This study employs multitemporal interferometric synthetic aperture radar (MT-InSAR) techniques, including differential InSAR (DInSAR), InSAR stacking, and interferometric point target analysis (IPTA), to analyze coal mine subsidence and delineate the boundaries of the mining-impacted zones. DInSAR accurately reconstructs, locates, and detects the trend in mining-induced subsidence and correlates well with documented mining operations. The InSAR stacking method maps the spatial variation of the ground’s average line-of-sight (LOS) velocity over the mining area, delineating the boundary of the impacted zone. IPTA analysis combining multilook and single-pixel phases achieves millimeter-level surface measurement above tunnel alignments and measures unevenly distributed deformation fields. This study considers an average of 4 cm per year of surface deformation in the LOS direction as the subsidence threshold value for delineating the boundary of the mining-induced subsidence-affected (MISA) zone during the active coal mining stage. Interestingly, there are twin transportation tunnels near the mining area. The twin tunnels completed before the coal mining activities started were functioning well, but damage was observed after the mining began. Our study reveals the tunnels are located within the InSAR-derived MISA zone, although the tunnels approach the MISA boundary. As direct signs of subsidence, ground fissures have been identified near the tunnels via field investigations and UAV photogrammetry. Furthermore, the derived distribution of ground fissures validates and verifies InSAR measurements. The integrated approach of MT-InSAR, UVA photogrammetry, and field investigation developed in this study can be applied to delineate the irregular boundary of the MISA zone and study the accumulating effects of mining-induced subsidence on the performance of infrastructure in areas proximate to coal mining activities.

1. Introduction

Underground coal mining has been practiced for centuries, and coal is expected to be an energy source in the foreseeable future. Ground subsidence, an inevitable consequence of underground mining activities at any depth, can occur immediately or in a prolonged manner and may affect localized or extensive areas [1,2]. As a result, mine subsidence can impact surface buildings, transportation infrastructure within the mining vicinity, and underground facilities [3,4,5,6]. The boundary of the mining-induced subsidence-affected (MISA) zone delineates the surface area affected by mining-induced subsidence. A thorough assessment of the extent of MISA zones is essential for designing a safe and cost-effective underground mining extraction plan.

In practice, a subsidence of less than 2 cm is commonly considered the threshold value of ground deformation; at values less than that, the deformation impact on the ground surface is negligible. Therefore, such a uniform threshold value of subsidence defines the boundary of a mining-impacted MISA zone [7,8,9,10]. However, using the uniform threshold value is oversimplified when applied in different areas because the ground experiences different subsidence processes in various regions due to spatial variation in factors such as geological settings, mining configurations, geomorphic features, overlying and underlying strata, groundwater conditions, etc. In addition, most subsidence threshold values defining the boundaries of impacted MISA zones are based on long-term deformation measurements after mine closures and even beyond. The uniform value approaches are substandard for studying impacted MISA zones for cases where mining activities are still ongoing [11].

Empirical methods, theoretical calculations, and numerical modeling methods are also widely applied to determine the boundary of the MISA zone, with particular emphasis on the substratum movement mechanism [12,13]. As a result, these methods often obtain a regular boundary of MISA zones induced by underground coal mining activities. In reality, the boundary of mining-induced ground movements is typically irregular. Additionally, the predicted area of the MISA zone is usually smaller than the actual extent, and the underestimation of MISA areas in the mining design stage may cause subsequent damage to be overlooked.

In-situ surveying provides the most accurate information on surface deformation, and the boundaries of impacted zones can be assessed using in-situ measurements from onsite monitoring instruments. Although traditional measurement techniques such as GPS, leveling, and total station surveys are reliable and well-established, they come with high costs, lengthy measurement cycles, and limited measurement points. Additionally, these methods cannot provide data on surface deformations that occurred in the past. Synthetic aperture radar (SAR), a remote sensing technique, has revolutionized the measurement of surface displacements. SAR offers extensive spatial measurements covering coal mining areas and their surroundings even without prior leveling data [14,15]. SAR interferometry (InSAR) allows for exploring the displacement field, measuring displacement magnitude, and tracking ground movements over time [16,17,18].

InSAR measures ground deformation by exploiting the phase difference in complex SAR imageries acquired over the same investigated area at different times. The first documentation of InSAR’s application for monitoring mine subsidence dates back to early 1997 [19], which successfully used the differential InSAR (DInSAR) method [20,21]. DInSAR techniques typically detect rapid surface deformation with centimeter-level accuracy and have also proven to be an effective tool for monitoring areas prone to surface subsidence due to mining activities [22]. However, atmospheric effects and temporal decorrelation in DInSAR monitoring can introduce noise into the generated interferograms, which may lead to unexploitable measurements. Instead, multitemporal (MT) InSAR technologies maintain phase and amplitude stability across multiple SAR images, which addresses the issue of spatiotemporal decorrelation and allows for millimeter-level ground deformation measuring [23]. MT-InSAR can reliably measure ground deformation without a high gradient related to the pixel size and wavelength ratio [24]. It is challenging to create successful unwrapped interferometry where the maximum detectable deformation gradient exceeds one fringe per pixel during a short time interval.

The potential of MT-InSAR techniques, such as the small baseline subset (SBAS) and persistent scatterer interferometry (PSI) techniques, have been explored for monitoring mining-induced ground movements [25]. Previous studies have revealed that with enough density of persistent scatterers (5–10 PS/km²), these techniques can accurately estimate multiple phase contributions in the interferometry [26] and retain valuable phase information even after many years in mountainous terrain [27,28,29]. PSI, PSInSAR^TM, and interferometric point target analysis (IPTA) share the same basic principles [30]. IPTA results with available ground control measurements have been compared and validated, demonstrating that such time-series analysis can achieve a measuring accuracy of 2 to 3 mm and that the accuracy of average deformation velocity measuring is on the order of 1 to 2 mm/year [31,32]. However, such techniques usually focus on monitoring slow deformation, which is approximately several to a dozen millimeters per year and is limited by a linear deformation model [33]. Examples of applications include abandoned coal mines, iron ore areas, salt mines, etc. [34]. Comparatively, the differential InSAR (DInSAR) method can measure relatively large-scale ground movements and is not limited by deformation mechanisms.

Instead of using two SAR image acquisitions for configurations, DInSAR’s capabilities can be significantly enhanced using large stacks of coregistered SAR images. Consecutive DInSAR or InSAR stacking methods have gradually become an alternative for detecting ground surface deformation with nonlinear characteristics [35]. The InSAR stacking technique calculates the mean line-of-sight (LOS) velocity via averaging and weighting a stack of unwrapped interferograms, outperforming DInSAR methods for estimating deformation rates [36,37,38].

Thus, this study integrates the DInSAR, InSAR stacking, and IPTA analysis methods to leverage the advantages of each approach and retrieve the boundary of a MISA zone. SAR fringes derived from DInSAR retrospectively construct the mining history. The InSAR stacking method is employed to delineate the irregular boundary of the MISA zone. Furthermore, time-series analysis of single-pixel and multilook phases is conducted to monitor areas around the boundary of the subsidence zone, paying particular attention during the field investigation to the pre-existing tunnels located at the boundary of the identified MISA zone.

InSAR measurements are validated and supplemented by field investigation and unmanned aerial vehicle (UAV) photogrammetry. Specifically, ground fissures induced by coal mining and underground tunnel damage, investigated using field investigation and drone photogrammetry, are considered direct indicators within the MISA zone. Field investigation reveals underground and aboveground damage caused by coal mining, while InSAR can only measure ground surface deformation. UAV photogrammetry and field investigation can identify and provide the distribution of ground fissures [39,40], validating the InSAR-derived MISA boundary.

2. Study Area

Shanxi is a top coal-producing province in China. Throughout the twentieth century, most coal in Shanxi province was obtained through longwall mining, an underground mining technique that extracted a wall or long face of coal. The fundamental concept of longwall mining involves utilizing temporary support systems to ensure a safe working environment for miners. Once the coal from the working section is removed, the supports are withdrawn, permitting overlying rock layers and a roof to collapse into the voids.

Coalfields are widely spread in Shanxi province. Figure 1 presents the distribution of Shanxi coalfields digitized from hardcopy maps created by the Chinese Academy of Geological Sciences and compiled by the U.S. Geological Survey [41]. Coal-mining regions are indicated by buffers with a 10 km radius around mine locations.

Nangou Coalfield in Shanxi has been selected as the target case study (Figure 1). Available reports on the Nangou coalfield provide detailed geological and technical information about this longwall underground mine. The coal seam lies at a depth of approximately 476 m, with an inclination angle of approximately 7° towards the south. The mining thickness of this coal panel is approximately 4.2 m. The surface topography of this mine site is mostly undulating, with elevation ranging from 1522 to 1804 m. According to documented coalfield mining activities, this coal panel has been mined since July 2020.

Figure 2a presents the detailed location of the Nangou coalfield and the westward mining direction. Faults have been identified on the southeastern side of the mining panel. Furthermore, twin highway tunnels were built completely and open to use before the mining excavation.

3. Methodology

3.1. Field Investigation

Field investigation presents first-hand evidence of mining-affected ground behaviors, which is close to the ground truth. Field investigations also provide validation and verification for the InSAR-delineated MISA zone.

A field investigation was conducted in May 2021, coinciding precisely with the conclusion of this coal panel’s mining and aligning with the acquisition time of the last SAR image. Field investigation indicates that ground fissures, subsidence, small-scale landslides, and collapse occurred during underground coal mining in the study site (Figure 2). Figure 2d shows how the relative displacement on two sides of a fissure is measured by setting pairs of survey stations. The survey station method achieves high-precision measurements, which is an advantage of the in-situ survey. Ground fissures around fault outcrops run parallel to the fault lines and extend up to 200 m outward.

Subsurface investigation serves as a supplementary method to explore underground impacts, for remote sensing techniques can only measure ground surface movements. Subsurface investigations find that mining-related subsidence has caused severe damage to the existing tunnels within the MISA area. The twin tunnels experience considerable damage, including lining cracks, cracks on the wall, roof collapse, tunnel invert damage, damage to steel supports, etc. (Figure 2e–h). The most severe damage to the tunnel occurred at the location intersecting the fault zones.

Field investigations on underground and aboveground conditions are critical and informative before applying remote sensing techniques, as they provide ground truth and validate InSAR measurements since mining activities started.

3.2. UAV Photogrammetry

The maximum ground subsidence during the active mining stage can reach meters, resulting in trenches and collapsing ground fissures. Ground fissures are a direct indicator of mining-induced subsidence. Although advanced InSAR techniques can be introduced to monitor ground fissures, InSAR-derived measurements are commonly affected by temporal and spatial decorrelations in mountainous areas. As a complementary method, UAV photogrammetry can map landforms and ground fissures at a small scale, which field investigation can further verify.

UAV measurement was conducted when the coal panel mining fished. The procedure of UAV aerial mapping starts with determining the flight path and altitude. Then, multiple overlapping images of complete coverage of the target area are taken using a drone equipped with high-resolution cameras. A camera mounted on a drone captures a total of 387 images covering the study area. High-resolution photos of the study area overlapped to make the same point on the ground visible from different angles and elevations, and they are stitched into a single high-resolution orthomosaic aerial map using the Pix4D mapper software. As a result, drone photogrammetry produces surveyor-level measurements of landscapes, provides valuable insights into ground surface conditions, and obtains detailed information on ground fissure distribution.

3.3. InSAR Processing

3.3.1. Dataset Collection

This study utilized 25 level-1 single-look complex SAR images obtained from the European Space Agency (ESA) Sentinel-1 satellite. These SAR images encompassed the mining period from July 2020 to May 2021. SAR images acquired from ascending orbit offered a ground resolution of 5 m in the range direction and 20 m in the azimuth direction [42].

In order to geocode SAR data and remove the topographic phase contribution, this study employed external digital elevation model (DEM) data with a 30 m spatial resolution provided by the National Aeronautics and Space Administration (NASA). The NASADEM improves the original Shuttle Radar Topography Mission (SRTM) DEM by minimizing data voids and expanding spatial coverage [43]. NASADEM reprocesses interferometric SAR data from the SRTM using an optimized hybrid processing technique.

3.3.2. DInSAR

Two or more SAR acquisitions collected from the same location in space but at different times can create InSAR interferograms to examine ground changes over time. There are distinct variations between DInSAR and InSAR in terms of their applications and processing techniques. Although both methods utilize phase differences between SAR images to identify surface changes, DInSAR is specifically designed to measure ground deformation by reducing or eliminating topographic phase contributions.

When a point on the Earth’s surface moves, the distance between a point on the Earth’s surface and the SAR sensor changes after the movement, corresponding to a signal phase shift in the SAR signal. For instance, if the sensor captures a SAR image for the initial time $t_0$ with a measured phase ${\varphi }_M$ and a subsequent SAR image at time $t$ with a phase ${\varphi }_S$. The DInSAR technique exploits the phase difference (${\varphi }_{int}={\varphi }_S-{\varphi }_M$), which contains information on surface displacements (${\varphi }_{displ}$), topographic phase (${\varphi }_{topo}$), orbit errors (${\varphi }_{oribit}$), atmospheric delays (${\varphi }_{atm}$), and random noise (${\varphi }_{noise}$) (Equation (1)). Surface displacements can be retrieved by removing the unwanted components from an interferogram. Thus, phase-based SAR techniques could be used to measure ground displacements. The measured deformation signal reflects motion between the acquisitions along the sensor–target LOS direction. DInSAR can detect rapid surface deformation and deliver elevation change with a detection limit of ~1 cm [44].

$${\phi }^{ij}={\phi }_{displ}^{ij}+{\phi }_{topo}^{ij}+{\phi }_{orbit}^{ij}+{\phi }_{atm}^{ij}+{\phi }_{noise}^{ij}$$

(1)

3.3.3. InSAR Stacking

DInSAR’s capabilities can be significantly enhanced by stacking SAR images. Instead of relying on the two-image configuration, stacking multiple unwrapped interferograms makes separating the deformation signal from random noise possible. Since atmospheric noise varies randomly across independent interferograms, this stacking approach, which increases the number of SAR observations, can help reduce the impact of atmospheric phase delay. Given $N$ wrapped interferograms, each pixel’s mean velocity ($V_{mean}$) is estimated using Equation (2). The N number of interferograms enhances the signal N times more significantly than a single interferogram, while the atmospheric noise increases by $\sqrt{N}$. Therefore, stacking N interferograms can improve the signal-to-noise by a factor of $\sqrt{N}$.

$$V_{mean}=\sum _{i=1}^N{\varphi }_i\Delta T_i/\sum _{i=1}^N\Delta T_i^2$$

(2)

where $\mathsf{\Delta }{T}_i$ represents the time interval between SAR images that forms the ith interferogram and ${\varphi }_i$ denotes the unwrapped phase. In the least-squares approach, the phases in each interferogram are considered equally weighted. In addition, the average LOS displacement velocity can be estimated by weighting each interferogram according to its time interval [36].

When interferograms maintain adequate coherence, particularly those with short intervals and minimal perpendicular baselines, several temporally adjacent interferograms can be stacked to enhance the signal-to-noise ratio. This stacking can create a pseudo-interferogram, which can be analyzed to interpret rapid displacement over extended periods, even when no coherent interferogram is available [45,46]. Therefore, this study uses the image stacking method to map ground movement that co-occurred with underground coal mining activities.

3.3.4. IPTA Time-Series Analysis

Theoretically, the time-series InSAR method can get reliable measurements in ground movements without a high deformation gradient [47] and with a deformation rate not exceeding cm/month. The boundary of a MISA area typically presents slow and stable deformation velocity. Hence, time-series analysis with millimeter accuracy is suitable for measuring ground deformation around the MISA zone boundaries over time.

Time-series analysis of InSAR data monitors Earth’s surface displacement over time using a network of interferograms. Considering the rolling terrain and mountainous study area, time-series analysis uses a combined SAR image stack that includes single-pixel and multilook phases. First, a single reference interferogram network is established as a baseline to extract deformation signals at individual pixels through interferometric point target analysis (IPTA). IPTA leverages interferometric signals’ temporal and spatial properties from pointlike targets, enabling precise estimation of deformation, atmospheric, topographic, and orbital phase components. Then, multilook elements are incorporated to extend spatial coverage in gravel or sparse-/short-vegetation areas. Multilooking reduces phase noise in differential interferograms, enhancing the multilook phases’ quality. Furthermore, a multireference stack is more suitable for scenarios involving rapid and potentially nonlinear motion, as linear regression may fail when deviations from linear motion exceed about 1 radian. A multireference stack also enhances spatial coverage for elements with intermediate coherence.

A high-density and evenly distributed set of coherent pixels is essential to measure ground movements reliably. This study achieves the necessary pixel density by integrating coherent pixels from single-pixel interferograms with less coherent ones using a multireference network of interferograms. As a result, the spatial coverage and monitoring capabilities are enhanced.

3.3.5. SAR Dataset Processing

DInSAR, InSAR stacking, and time-series analysis methods are comprehensively used to retrospectively analyze the ground deformation history, measure the boundary of the MISA zone, and reveal unevenly distributed ground deformation above the pre-existing tunnel alignments. SAR datasets are analyzed using the GAMMA Remote Sensing AG software through a computer-aided process.

Figure 3 illustrates the workflow of dataset processing implemented in this study. Interferograms are generated based on the constraints of perpendicular and temporal baselines for stacking and time-series analysis. Collected SAR images and external DEM are preprocessed before the generation of interferogram networks, including geocoding, co-registration, deramping, and cropping of the area of interest.

The InSAR stacking method maps ground displacements caused by coal mining, which is generally effective when capturing events that happened rapidly between two measurements. For this purpose, multireference interferograms are generated for stacking analysis. Spatial phase unwrapping is conducted using a minimum-cost flow (MCF) phase unwrapping algorithm before stacking analysis. As a result, an average velocity map is derived from the stacked interferograms, which helps to depict the overall ground movements and delineate the MISA zone.

Time-series analysis measures the ground deformation at the MISA zone boundary, including surface measurements above pre-existing tunnels. First, a single-pixel differential interferometric phase is generated in vector data format for a selected list of persistent scatter candidates. PSI is applied to full-resolution interferograms and serves as the baseline method for retrieving the time-series deformation. Next, multilook differential interferometric phases are generated and extracted into vector data format, compatible with the corresponding single-pixel phases to be extracted. Finally, a combined multireference stack that contains single-pixel and multilook differential interferometric phases is created. This approach, where multiple master SAR images are used instead of a single reference image for all interferometric pairs, increases the number of interferograms generated with small spatial and temporal baselines, potentially maximizing pixel coherence.

Dataset processing uses a combined point list, including unwrapping differential phases, estimating the atmospheric phases, calculating a height correction to the DEM-derived point height, and calculating a mask that discards decorrelated points. Phase unwrapping is executed in the spatial domain using the MCF algorithm, which is designed for sparse data to correct errors in time-dependent phase unwrapping. Once reliable unwrapped phases are obtained for interferometric pairs, and baseline refinements are performed through a least-squares method, it becomes possible to differentiate atmospheric phases, nonlinear deformations, and residual phase errors based on their unique temporal and spatial characteristics. In order to address the challenges posed by steep topography, filtering, and unwrapping are carried out in map coordinates to avoid placing values with significantly different terrain heights and atmospheric phases in close proximity. A multireference stack is also utilized to estimate and update atmospheric phases and height corrections.

4. Results

4.1. Mining History Reconstruction

Archived SAR images allow for retrospective tracking of surface displacement and provide valuable information for reconstructing mining history. Figure 4 reconstructs the mining-induced displacement, indicated by SAR fringes, with two red lines representing the pre-existing twin tunnels.

According to documented coalfield mining activities, this panel was mined since July 2020. However, the interferogram 20200725–20200806 presents no displacements that occurred during this period. The first SAR fringe appeared in August 2020, indicating subsidence occurred approximately one month after the underground mining operation began. SAR interferometry from 20200806 to 20210427 tracks westward mining activities and surface subsidence. As mining progressed, the center of subsidence shifted westward, and the subsidence range gradually expanded. The subsidence process is dynamic, as surface movement begins with undermining and continues until some maximum displacement has occurred. At the end of the subsidence process, equilibrium is achieved, and the resulting surface deformations become relatively static. The ground surface tends to be stable after active mining and residual subsidence began to develop (after 20210427) with no prominent fringe center.

Moreover, impacts on the twin tunnels around the boundary of the MISA zone can be recognized intuitively from interferometry during 20200911–20201005, 20201005–20201110, and 20201216–20210202, as the twin tunnels, represented by red lines, are intricated in the SAR fringes.

The InSAR-mapped mining subsidence is highly consistent with documented working faces, indicating that subsidence from longwall mining typically occurs within some days following coal extraction. Ground subsidence can correlate to the progression of the working face. The subsidence associated with longwall coal extraction happens correspondingly with mining progress, generally completed within weeks to months.

4.2. Mean LOS Velocity of Mining-Induced Ground Subsidence/Surface Mean Velocity Fields

Detecting rapid and substantial subsidence with SAR interferograms remains challenging because of the limitations imposed by the SAR wavelength. InSAR struggles to accurately measure subsidence when the displacement of a pixel in the radar LOS direction exceeds half the wavelength. Furthermore, longwall mining often causes substantial subsidence over a short period, leading to decoherence and preventing the InSAR technique from identifying high-coherence points around the subsidence center. The most significant subsidence typically occurs at the center of the subsidence trough. Field observations in this study area reveal deformations of several dozen to hundreds of centimeters within a relatively small area (approximately 2 km by 2 km). InSAR measurements cannot capture subsidence with steep gradients, resulting in data gaps in the mean velocity maps.

Besides identifying the high-gradient subsidence center, the InSAR stacking method effectively determines the spatial coverage and boundary of the MISA zone over the entire data acquisition period (Figure 5). The mean velocity map reveals that the longwall mining panel is not at the geometric center of the MISA zone. In addition, the MISA area is much larger than the mining panel. Subsidence in excavated areas results in a smooth profile that extends beyond the mining panel boundaries in all directions.

Subsidence analysis is essential during mine planning and permitting stages, and a subsidence threshold value must be estimated. As mentioned in the introduction section, a subsidence of less than 2 cm is commonly considered the threshold value for long-term deformation induced by underground mining. However, considering the case study is still in the active mining stage and based on the pattern of the InSAR-derived mean velocity map (Figure 5), a 4 cm/year ground surface deformation in the LOS direction is selected as the subsidence threshold, which signifies the limit beyond which subsidence effects become of less concern, marking the boundary of the MISA area.

4.3. Ground Fissures Derived from Field Investigation and UAV Measurements

A highly accurate, geometrically corrected orthomosaic map is derived using the UAV photogrammetry technique, presenting a bird’s-eye view of the investigated area. UAV photogrammetry captures a series of overlapping images from different angles. The images are aligned and stitched into a seamless mosaic, and distortions are corrected. As a result, the orthomosaic map provides centimeter-level resolution, which is essential for measurements and observations of ground fissures. Special attention is paid to the distribution of ground fissures between the excavation panel and tunnel alignments. Identified ground fissures, indicated by blue lines in Figure 6, are validated by field investigation.

These identified ground fissures are located within the InSAR-derived MISA zone boundary. Their distribution between the mining excavation panel and tunnel alignments provides possible evidence of the existence of the MISA zone.

4.4. Angles of Draw of Unevenly Distributed Ground Movements

InSAR stacking measurements and drone photogrammetry determine a 4 cm/year ground surface deformation in the LOS direction as the subsidence threshold for delineating the MISA zone for this case study. The extent of the irregular subsidence boundary can be described using the angle of draw, which represents the angle at which subsidence propagates from the edge of the underground extraction panel to the places where the mining-induced subsidence is less than the threshold value.

Accurate evaluation of these limit angles becomes more complex with inclined seams. The coal seam in this case study has an inclination angle of approximately $7°$. The schematic representative profiles of the subsidence trough (not to scale) are plotted in Figure 7. The three trough subsidence profiles for sections A–A’, B–B’, and C–C’ show nonuniform movements across the panel. The angle of draw on the dip side of the panel was greater than that on the rise side. Specifically, the angle of draw on the panel dip side (${\delta }_1$) ranges from $61°$ to $77°$, whereas the angle of draw on the rising side (${\delta }_2$) varies between $26°$ and $41°$, as summarized in

Table 1. Angles of draw (in degrees) of the inclined coal panel.

	$\mathbf{Angle} \mathbf{of} \mathbf{Draw} \mathbf{Coal} \mathbf{Panel} \mathbf{on} \mathbf{Dip} \mathbf{Side} \left({\mathit{\delta }}_1\right)$			$\mathbf{Angle} \mathbf{of} \mathbf{Draw} \mathbf{of} \mathbf{Coal} \mathbf{Panel} \mathbf{on} \mathbf{Rise} \mathbf{Side} \left({\mathit{\delta }}_2\right)$
	Profile A–A’ $({\delta }_{1A}$)	Profile B–B’ $({\delta }_{1B}$)	Profile C–C’ $({\delta }_{1C}$)	Profile A–A’ $({\delta }_{2A}$)	Profile B–B’ $({\delta }_{2B}$)	Profile C–C’ $({\delta }_{2C}$)
Values in degrees	61	64	77	41	29	26

. Accordingly, InSAR measuring results indicate that the inclination of the seam reduces the limit angle on the rise side and increases it on the dip side (i.e., ${\delta }_1>{\delta }_2$).

Coal extraction in the westside unfaulted zone results in a trough-shaped subsidence that extends outward in all directions. In contrast, ground movement in the neighborhood of faults is unevenly distributed between the faults’ two sides. Figure 7 shows that the spatial distribution of subsidence in section A–A’ is influenced by faults. These faults act as boundaries that regulate the lateral extent of the subsidence trough. This process is conceptually represented in Figure 8. The parameter $∆\delta$ in Figure 8 represents a decreased angle of draw affected by existing faults. Correspondingly, Figure 7 presents the value of the parameter ${\delta }_{2A}$ is larger than that of parameters ${\delta }_{2B}$ and ${\delta }_{2C}$.

In addition, faults limit and determine the extent of the subsidence trough, resulting in ground fissures occurring at the fault outcrops, which are found in the field investigation. Faults in regions affected by longwall mining subsidence are prone to reactivation, though this is not always observed. Reactivation of faults co-occurs with mining subsidence because of the release of accumulated strain along the fault zone [48], leading to shear displacements, surface cracks, and irregular subsidence. While field investigations identified ground fissures aligned with the fault direction, it cannot be conclusively determined that the faults are reactivated without additional data or in-situ monitoring.

4.5. Time-Series Measurements Around Tunnel Alignments

The mean LOS velocity map (Figure 5) reveals that the pre-existing twin tunnels are intricated within the area affected by coal mining. Damage to twin tunnels is a known environmental consequence of differential movements. In order to examine ground deformation over time above the tunnels, a time-series analysis of SAR data is performed using IPTA methods. Traditionally, the number of pixels monitoring ground deformation in mountainous areas is limited since InSAR provides time-series deformation only for coherent pixels. This study achieved a relatively high pixel density by combining single-pixel and multilook phases for point selection (see Figure 9), gathering enough measuring points or persistent scatterers to estimate multiple phase contributions in the interferometry and provide accurate deformation measurements.

Furthermore, retrieved time-series analysis data are interpolated using radial basis function (RBF) interpolation to display the overall deformation field and trend as a contour background. The results indicate that the interaction location between pre-existing faults and tunnel alignments is the loci of subsidence, showing concentrated abnormal surface deformation. Faults located far from the mining panel but close to the MISA zone’s boundary and intersecting with existing tunnel alignments act as the boundary controlling the lateral subsidence extent and the loci of concentrated ground displacements where the tunnel experienced the most severe damage. Correspondingly, the ground subsidence field can be divided into four sections (I, II, III, and IV) based on the distribution of cumulative settlements along tunnel alignments.

Time-series analysis enables the examination of point-specific ground deformation over time, with temporal samples corresponding to the acquisition dates of SAR images. Figure 10 displays the results of time-series analysis along the tunnel alignment for representative monitoring points within sections I, II, III, and IV. Ground subsidence of all four sections began in September 2020, consistently with the first detected SAR fringes in the mining history reconstruction illustrated in Figure 4, which demonstrates again that the tunnels are affected by coal mining activities.

Mining-induced ground subsidence plays less of a role in sections I and IV, located at the two ends of the tunnel alignments. Specifically, section IV, which is on the right-hand side of the intersection between the faults and tunnels, is rarely affected by coal mining. From this perspective, the evidence that existing faults laterally control mining-induced ground displacement is demonstrated again. In contrast, sections II and III experienced relatively large amounts of ground deformation and high deformation gradients, especially section III, which intersects with faults.

In order to explore the trend in cumulative surface settlements with time, the moving average (MA) method is applied to process the pointwise measurements to present the trend in ground subsidence, indicated by the solid black lines in Figure 10. Accordingly, for sections II and III, the coal mining-induced ground subsidence at a measuring point follows a waved S-shaped curve with time: it initially progresses slowly (initial phase), then accelerates rapidly (primary phase), and eventually stabilizes (residual phase). Time-series analysis revealed that mining-induced ground movements caused continuous damage to the existing tunnels.

The coal seam was excavated parallel with the tunnel alignment and has gradually moved westward since July 2020. In August 2020, approximately one month after coal mining operations began, SAR fringes indicating the center of surface movements occurred, as illustrated in Figure 4. Time-series analysis (Figure 10) reveals the ground movements above the tunnel alignment, representing the ground deformation at the MISA zone boundary. Figure 10 illustrates that ground settlements at the surface boundary occurred in September 2020, two months after the beginning of coal mining, with a one-month delay from the surface subsidence center occurrence. It took approximately two months for surface deformation to occur around the MISA zone boundary following the beginning of coal mining operations.

5. Discussion

While InSAR interferometry provides high-precision deformation measurements, achieving accuracy to the centimeter or even millimeter level, its effectiveness is constrained by spatial and temporal coherence and deformation gradients. The phase gradient between two adjacent pixels in an interferogram must be less than π. Otherwise, phase decorrelation or discontinuity occurs when the surface deformation gradient surpasses this detectable limit. Ground displacements from underground coal mining often involve large deformation gradients and rapid changes, resulting in significant interferogram decorrelation and phase discontinuity. Consequently, the InSAR technique is challenging in identifying stable points with relatively high coherence, leading to the loss of information regarding the maximum subsidence.

Additionally, higher spatiotemporal-resolution acquisitions and longer wavelengths (such as L-band) are preferred to achieve the highest detectable deformation gradient between pixels. The available C-band Sentinel-1 dataset itself inherently determines the detectable ground movements.

This study employs multiple InSAR techniques to investigate the MISA zone boundary comprehensively. Specifically, DInSAR interferograms are used to reconstruct the mining history, with SAR fringes representing surface movements induced by mining activities. Based on the known mining history, the InSAR stacking method is applied to map the mean LOS velocity field, delineate the boundary of the MISA zone, and identify pre-existing tunnels within the affected area. Tunnels happen to be located in proximity to the MISA zone boundary. Furthermore, time-series analysis integrating single-pixel and multilook phases revealed ground deformation above the tunnels, reaching millimeter measuring accuracy.

Multiple InSAR methods with different measuring abilities comprehensively retrieve the mining-induced ground movements and demonstrate the measuring results with each other. Comparisons among DInSAR, InSAR stacking, and IPTA time-series analysis are summarized in

Table 2. Comparisons among various InSAR methods applied in measuring mining-induced ground movements and subsidence boundary.

Methodology	Objectives	Measurement Accuracy	Interferograms
DInSAR	Mining history reconstruction	Centimeter	9
InSAR stacking	Surface mean velocity fields;	Centimeter	146
Time-series analysis	High-resolution time-series analysis of ground movement above the existing tunnel and around the boundary of the mining-impacted area	Millimeter	24 single reference 68 multireference

. SAR fringes from the DInSAR and InSAR stacking methods effectively measured the extent and boundaries of the mining-impacted area. Still, they are limited in capturing the maximum subsidence because of constraints imposed by satellite wavelength and decoherence. For engineering projects in mining-impacted areas, millimeter-level monitoring accuracy is essential. Time-series analysis combining single-pixel and multilook phases has demonstrated high accuracy and a relatively dense measurement point distribution, even in mountainous regions.

6. Conclusions

This study delineated the irregular boundary of a MISA zone using integrated MT-InSAR techniques, UAV photogrammetry, and field investigation. These methods were applied comprehensively to measure ground movements, validate each other, and determine the area of the MISA zone and its changes with the progress of the mining face.

The InSAR techniques retrospectively analyzed the history of ground movements and revealed unevenly distributed ground deformation. InSAR methods successfully measured a centimeter-level trough subsidence profile and millimeter-accuracy time-series deformation, besides the center of ground subsidence, limited by satellite wavelength, high-gradient subsidence, and decoherence. Specifically, DInSAR-derived SAR fringes reconstructed the exact locations of work, which helped understand the mining history and progress. In the case of poor historical documentation and ground measurements, InSAR is an indispensable technique for reconstructing mining history and measuring ground movements. Its effectiveness was demonstrated from this perspective. Then, InSAR stacking methods mapped the mean velocity field, revealed the subsidence pattern, and identified the boundary of the mining-affected zone. Mining-impacted surface areas were more extensive than the extraction area. Pre-existing tunnels happened to be located within the mining-affected subsidence boundary and suffered severe damage. Furthermore, IPTA time-series analysis of SAR datasets was conducted to investigate further the ground deformation occurring above tunnel alignments, providing valuable quantitative information on the subsidence rate. Time-series analysis achieved a high monitoring accuracy, reaching the millimeter level for an engineering project (i.e., existing tunnels) involved in the mining-impacted area.

SAR fringes and InSAR stacking results indicated that subsidence from active mines induced damage and caused disorder to the existing tunnels. Faults happened to be located at the boundary of the affected zone and intersected with the tunnel alignment. Faults acted as boundaries that restricted the lateral spread of the subsidence trough, leading to significant differential subsidence within the zone near the faults. Additionally, faults served as focal points for concentrated ground displacements along the tunnel alignment, resulting in the most severe damage to the tunnel. Field investigation revealed that unevenly distributed ground deformation and settlement rates corresponded well with the magnitude of tunnel damage.

This study utilized various InSAR methods to capitalize on their strengths, address the limitations of each approach, and obtain a comprehensive deformation pattern. SAR interferometric analysis successfully reconstructed mining history, mapped mean velocity fields, determined the irregular boundary of mining-impacted zones and investigated the effects on pre-existing tunnels within coal-mining-impacted zones. In addition, the distribution of mining-induced ground fissures was identified using UAV photogrammetry and field investigation, which validated and supplemented the InSAR-derived boundary of the coal-mining-affected areas. The findings and results from this study offer valuable insights for coal mine planning and disaster prevention.