Evaluating the Interferometric Performance of China’s Dual-Star SAR Satellite Constellation in Large Deformation Scenarios: A Case Study in the Jinchuan Mining Area, Gansu

Abstract

Mining activities can trigger geological disasters, including slope instability and surface subsidence, posing a serious threat to the surrounding environment and miners’ safety. Consequently, the development of reasonable, effective, and rapid deformation monitoring methods in mining areas is essential. Traditional synthetic aperture radar(SAR) satellites are often limited by their revisiting period and image resolution, leading to unwrapping errors and decorrelation issues in the central mining area, which pose challenges in deformation monitoring in mining areas. In this study, persistent scatterer interferometric synthetic aperture radar (PS-InSAR) technology is used to monitor and analyze surface deformation of the Jinchuan mining area in Jinchang City, based on SAR images from the small satellites “Fucheng-1” and “Shenqi”, launched by the Tianyi Research Institute in Hunan Province, China. Notably, the dual-star constellation offers high-resolution SAR data with a spatial resolution of up to 3 m and a minimum revisit period of 4 days. We also assessed the stability of the dual-star interferometric capability, imaging quality, and time-series monitoring capability of the “Fucheng-1” and “Shenqi” satellites and performed a comparison with the time-series results from Sentinel-1A. The results show that the phase difference (SPD) and phase standard deviation (PSD) mean values for the “Fucheng-1” and “Shenqi” interferograms show improvements of 21.47% and 35.47%, respectively, compared to Sentinel-1A interferograms. Additionally, the processing results of the dual-satellite constellation exhibit spatial distribution characteristics highly consistent with those of Sentinel-1A, while demonstrating relatively better detail representation capabilities at certain measurement points. In the context of rapid deformation monitoring in mining areas, they show a higher revisit frequency and spatial resolution, demonstrating high practical value.

1. Introduction

Geological deformation induced by open-pit mining has become a critical factor threatening safe production in mining regions. Large-scale mining activities often trigger geological disasters such as surface subsidence and slope instability, which not only threaten the stability of mining infrastructure but also pose significant threats to human safety [1,2,3]. Therefore, large-scale, high-precision, rapid-response surface deformation monitoring in mining areas is undoubtedly crucial in order to identify safety issues in mining areas [4].

Traditional in situ monitoring methods, such as the Global Navigation Satellite System (GNSS), leveling, and distributed fiber-optic sensing, suffer from limitations including single-point constraints, low efficiency, and elevated costs, rendering them inadequate for large-scale, continuous monitoring in mining areas [5]. The time-series InSAR (TS-InSAR) techniques, represented by methods such as PS-InSAR [6,7], small baseline subset InSAR (SBAS-InSAR) [8], and distributed scatterer InSAR (DS-InSAR) [9], demonstrate significant algorithmic distinctions. In recent years, they have become powerful tools for open-pit mine deformation monitoring, with capabilities including wide area coverage, high precision, and regular temporal sampling. These techniques have been widely adopted in practical applications, marking a significant advancement in the field of mine deformation monitoring. Among various applications, using medium-to-low-resolution Sentinel-1A satellite images in combination with TS-InSAR technology has become one of the most prevalent approaches to monitoring deformation in mining areas. For instance, He et al. [10] investigated the Wannian Mine in the Fengfeng mining area using various TS-InSAR techniques, concluding that DS-InSAR provides higher accuracy compared to other techniques. Similarly, Du et al. [11] utilized both PS-InSAR and DS-InSAR to process Sentinel-1A images covering the PeiBei mining area from 2017 to 2018 and found that DS-InSAR not only proved feasible for mine deformation monitoring but also outperformed PS-InSAR in terms of capability.

However, open-pit mining areas are frequently subjected to large-gradient deformations due to intensive extraction activities [12]. These deformation fields often exhibit strong nonlinearity and spatial discontinuity. They are also influenced by high-frequency anthropogenic activities, such as mining truck operations and mechanical vibrations, dynamic changes in surface conditions, including exposure to bare ground, and the movement of waste piles. During active operations, localized, small-scale deformation bodies may also develop within specific mining sections. The C-band Sentinel-1A satellite, despite its advantages, has a limited spatial resolution and a relatively long revisit interval, which restricts its ability to capture large-gradient deformations that exceed the theoretical detection threshold (i.e., deformation between consecutive acquisitions must remain below one-quarter of the radar wavelength). Furthermore, the sparse distribution of coherent scatterers often leads to phase unwrapping errors in localized regions, undermining the satellite’s accuracy and reliability in deformation monitoring in geologically complex mining environments.

Integrating high-resolution SAR data (e.g., TerraSAR-X, ALOS PALSAR) into InSAR technology has dramatically improved deformation monitoring in mining areas. Notable progress has been achieved in terms of both spatial resolution and accuracy in deformation detection. For example, Przyłucka et al. [13] combined D-InSAR with SqueeSAR to process high-resolution TerraSAR-X data, effectively evaluating the influence of mining-induced deformation intensity and mining boundaries on the surrounding urban infrastructure. In a subsequent study, Tang et al. [14] employed a coherence-optimized SBAS approach using ascending and descending Sentinel-1A data in conjunction with descending high-resolution TerraSAR-X imagery. This approach enabled the millimeter-level monitoring of slope deformation in the Rhenish open-pit mine in Germany, facilitating the retrieval of surface displacement within and around the mining area and supporting a detailed analysis of the deformation mechanisms. Zhang et al. [15] proposed an enhanced Stacking-InSAR technique that integrated coherence gradient analysis, leveraging GF-3 and Sentinel-1 data to monitor large-scale mining subsidence while addressing challenges such as decorrelation and displacement underestimation in regions with steep deformation gradients. Although high-resolution SAR images have greatly enhanced our capabilities in mining-related deformation monitoring, the widespread application of international high-resolution satellites such as TerraSAR-X is hindered by high acquisition costs and region-specific imaging priorities, which limit continuous data availability for some areas of China. While the launch of GF-3, China’s first C-band high-resolution SAR satellite, fills the domestic data gap, it is mainly used for national security due to regulatory constraints. Civilian use requires lengthy, strict approval. While medium-to-low-resolution data such as Sentinel-1 can serve as a partial substitute, their limited spatial resolution and precision limit their utility in high-accuracy monitoring tasks, impeding their use both in scientific research and in practical applications in mining deformation monitoring.

The Chinese commercial small satellite industry is experiencing rapid growth [16], with significant advancements in small SAR satellites. These satellites have become a focal point for strategic competition among emerging aerospace enterprises. Representative examples include the “Fucheng-1” and “Shenqi” dual-satellite constellation, developed by the Changsha Tianyi Research Institute, which was launched successfully and has been in operation for several years. These platforms have consistently acquired large volumes of high-quality SAR datasets, showing marked improvements in SAR interferometry performance and capabilities in the Chinese commercial SAR sector [17]. Characterized by a high resolution, low cost, and rapid response, small SAR satellites are well suited for applications including disaster monitoring, urban planning, agricultural assessment, and maritime law enforcement. Additionally, the strict orbit control demands faced by dual-star constellations significantly challenge the Flight Operations Center, especially the Flight Dynamics System. Depending on solar and geomagnetic activity, frequent weekly maneuvers may be required. Flight Dynamics must monitor ground track evolution daily, often outside normal hours, and generate maneuver execution products 2–5 times weekly, with adequate uplink opportunities for timely satellite updates [18].

Despite these technological advancements and challenges, there is still a lack of Chinese case studies utilizing small-satellite SAR data to monitor geological deformation in mining areas. To address this gap, in this study, we employed TS-InSAR technology to monitor surface deformation in complex open-pit mining environments using high-resolution SAR imagery from the “Fucheng-1” and “Shenqi” satellites. Specifically, we analyzed ground deformation in the Jinchuan mining area, Gansu Province, using 15 scenes with high-resolution images from the “Fucheng-1” satellite and 5 scenes from the “Shenqi” satellite with PS-InSAR technology. The results were compared with, and validated using, those obtained from both PS-InSAR and DS-InSAR using Sentinel-1A data. Corner reflectors (CRs) installed within the mining area helped in evaluating the reliability of the InSAR results. This study demonstrates the effectiveness of using “Fucheng-1” and “Shenqi” satellite data in monitoring rapid subsidence caused by mining activities. The study also assessed the stability of the dual-star interferometric capability, imaging quality, and orbit control ability. It also provides technical and data support for Tianyi Research Institute’s application of these satellites in monitoring mining areas.

2. Methodology

2.1. Traditional TS-InSAR Data Processing

To cover the entire Jinchuan mining area from October 2024 to April 2025, a total of 15 SAR scenes acquired by the descending Sentinel-1A satellite along path 33, frame 462, with an average temporal baseline of 12 days, were downloaded from the Alaska Satellite Facility (ASF). Additionally, from November 2024 to April 2025, we collected 15 scenes from the “Fucheng-1” satellite and 5 scenes from the “Shenqi” satellite, both of which were provided by the Tianyi Research Institute. These strip-mode datasets offer spatial resolutions up to 3 m and revisit intervals as short as 4 days. Routine interferometric processing steps, including geometric co-registration, flattening, topographic phase computation based on TanDEM-X DEM, interferogram generation, topographic phase removal, and geocoding, were carried out using Doris 2.5.0 software [19,20]. Due to the characteristics of the Sentinel-1 TOPS imaging mode, a sub-pixel co-registration accuracy in the azimuth direction better than 0.001 pixels is required to avoid phase errors exceeding 3°. Therefore, an enhanced spectral diversity method based on image intensity information was employed to refine the co-registration results based on precise orbit and correlation-based alignment. Subsequently, the persistent scatter processing implemented in StaMPS/MTI [21] was used to derive the spatiotemporal evolution of surface deformation. This included the selection of persistent scatterer (PS) targets, the estimation of the spatial uncorrelated/correlated look angle (SULA/SCLA), 3D phase unwrapping, and master atmosphere and orbit error estimation (AOE) [22]. Ultimately, 14 single-master differential interferograms from Sentinel-1A and 19 from the Tianyi satellite constellation were obtained after the removal of various artifacts. These interferograms were used to derive the average deformation rate and cumulative displacement time-series results across the study area during the observation period using a least-squares inversion approach.

2.2. DS-InSAR Phase Optimization

PS-InSAR technology often suffers from a low density of PS pixels due to environmental complexity and the presence of poor-quality interferograms. In contrast, DS-InSAR technology improves interferogram quality through filtering and phase optimization. It identifies statistically homogeneous pixels (SHPs) via hypothesis testing on time-series amplitude images and subsequently fuses PS candidates (PSCs) and distributed scatterer candidates (DSCs), significantly increasing the density of monitoring points [23,24]. In this study, DS-InSAR processing was applied to the Sentinel-1A dataset to assess the advantages of high temporal revisit frequency and high spatial resolution in improving point density, as well as to verify the consistency of deformation results derived from multiple datasets. Subsequently, these inversion results were compared against those obtained using traditional PS-InSAR techniques on the “Fucheng-1” and “Shenqi” datasets. To enhance the processing efficiency and achieve a higher density of monitoring points, in this study, we first screened image pixels using the amplitude deviation (DA) threshold method: pixels with DA $\le 0.49$ were retained, with those with DA $<0.25$ directly recognized as PSC targets, while pixels with DA values within the range of $0.25-0.49$ underwent a two-sample Kolmogorov–Smirnov (KS) test [7]. Pixels outside of this range were excluded from further testing. The inspection window was set to $31\times 11$ in the range and azimuth directions, respectively, within which all neighboring pixels were compared with the central pixel to assess whether their temporal amplitude distributions were statistically consistent [9].

Phase values of homogeneous pixels were used to calculate the central pixel’s phase value through weighted averaging, with the weights determined based on the correlation coefficients between the central pixel and each homogeneous pixel [25]. A coherence coefficient covariance matrix was constructed for the entire set of interferograms, and phase optimization was conducted on all-combination differential interferograms via the eigenvalue decomposition of the coherence matrix. That is to say, for any distributed target, the coherence matrix $T$ can be expressed as follows [26]:

$$T=E\left[y{y}^H\right]\approx {\displaystyle \frac{1}{N_P}}{\sum }_{\in \mathsf{\Omega }} y{y}^H$$

(1)

where $y=[y_1,y_2,\dots ,y_N]$ is the complex vector after normalization; $y^H$ is the conjugate transpose of $y$; $\Omega$ is the set of homogeneous points of the distributed target; and $N_P$ is the number of these homogeneous points. In addition, $E\left[{\left|y_i\right|}^2\right]=1$; the coherence matrix $T$ captures the coherence and related information of the distributed target’s homogeneous points.

Additionally, as the coherence matrix is positive-definite, it can undergo eigenvalue decomposition, represented as follows [27]:

$$T=U\Lambda U^H={\sum }_{i=1}^N{\lambda }_i{u}_i{u}_i^H$$

(2)

where $\Lambda$ is a diagonal matrix and $\Lambda =diag({\lambda }_1,{\lambda }_2,\dots ,{\lambda }_N)$; ${\lambda }_i$ represents the $N$ eigenvalues of the coherence matrix and $u_i$ is the eigenvector corresponding to each eigenvalue ${\lambda }_i$.

In the comparison of eigenvalues, larger ones indicate that their corresponding eigenvectors are optimal estimates, while others are treated as noise. Normally, the maximum eigenvalue surpasses the rest. Thus, the coherence matrix $T$ can be expressed as the sum of two parts:

$$T=T_{signal}+T_{noise}={\sum }_{i=1}^m{\lambda }_i{u}_i{u}_i^H+{\sum }_{m+1}^N{\lambda }_i{u}_i{u}_i^H\approx T_{signal}$$

(3)

where $T_{signal}$ and $T_{noise}$ are the coherence matrices for signals and noise, respectively; when the eigenvalues of noise are small enough to be ignored, the maximum eigenvalue’s corresponding eigenvector gives the optimized phase for the distributed target.

Subsequently, time-series analysis was carried out using the StaMPS/MTI algorithm as described in Section 2.1. To ensure consistency in subsequent comparisons, we standardized the time-series processing parameters for the two types of SAR satellites, as detailed in

Table 1. Stamps processing parameters [28] for PS-InSAR and DS-InSAR with various SAR datasets.

Parameter	Value
Max_topo_err	35
Unwrap_grid_size	100
Weed_max_noise	0.4
Weed_standard_dev	1
Density_rand	20
Coherence threshold	0.7

, and the overall data processing workflow is illustrated in Figure 1.

3. Study Area and Data Sources

3.1. Study Area Overview

The Jinchuan Cu-Ni mining area, located in Jinchang, Gansu Province, is the largest nickel and cobalt production and reserve base in China and represents a super-large underground mining operation. Situated in the eastern section of the Longshou Mountains along the Hexi Corridor, this region is characterized by an erosional hilly and low-mountain landform, with elevations ranging from 1700 to 2700 m. The area experiences an arid climate, with an average annual precipitation of only 170 $\mathrm{m}\mathrm{m}$ and an evaporation rate of approximately 1700 $\mathrm{m}\mathrm{m}$. Tectonically, the mining zone lies within a high-stress region dominated by horizontal tectonic stress, and both the orebody and surrounding country rock have undergone extensive fracturing due to multiple geological events. This has led to pronounced deformation in the surrounding underground rock; typically, this is rapid and persistent and exhibits significant temporal and spatial heterogeneity.

As illustrated in Figure 2, the Jinchuan deposit consists of four mining areas—IV, II, I, and III, from east to west. Specifically, Areas I and III are managed by Longshou Mine, Area II is jointly managed by Mine 2 and Mine 3, and Area IV is under the jurisdiction of Mine 3 [29]. The region’s stratigraphy has undergone multiple phases of metamorphism and magmatic intrusion, resulting in a lithological assemblage dominated by migmatite, gneiss, and marble. It has also been subject to repeated tectonic movements, which have further compromised the stability of the rock mass [30]. In recent years, deep-mining engineering stability has become a growing concern due to mining activities, geological faults, and complex rock formation structures. This has led to severe surface subsidence, roof falls, rib failures, and fractures in the Jinchuan mining area, posing a significant threat to safe and efficient production. Effective surface deformation monitoring is thus crucial in ensuring safe mining operations [31].

3.2. Available Datasets

3.2.1. Sentinel-1A

Launched on 3 April 2014, Sentinel-1A offers open-source SAR images with world-leading orbital and attitude control, ensuring excellent and stable interferometric capabilities. Its IW imaging mode boasts a 250 km swath width, with a spatial resolution of 20 m in the azimuth direction and 5 m in the range direction. Due to damage to Satellite B, its revisiting period is currently 12 days, making it a top-tier data source for InSAR research. In this study, 15 Sentinel-1A images (from 2 October 2024 to 12 April 2025) underwent time-series PS/DS-InSAR processing to capture the spatial–temporal deformation distribution in the study area. Additionally, four images (from 23 February 2025 to 31 March 2025) were used for CR identification and deformation monitoring based on CR.

Table 2. “Fucheng-1”, “Shenqi”, and Sentinel data parameter information.

Satellite Name	Band	Acquisition Mode	Orbit/View Angle	Pixel Spacing (m)	Incidence Angle (°)	Heading Angle (°)	Time Span
Shenqi	C	Strip	Ascending/right	1.3/1.8	38.18	12.30	20241121–20250413 (Total: 5 scenes)
Fucheng	C	Strip	Ascending/right	1.3/1.8	38.18	12.30	20241026–20250409 (Total: 15 scenes)
Sentinel-1A	C	IW	Descending/right	2.3/14.0	30.6	193.37	TS-InSAR: 20241002–20250412 (Total: 15 scenes)
							CR-InSAR: 20240223–20250331 (Total: 4 scenes)

provides the image parameters.

3.2.2. Fucheng-1/Shenqi

Developed by Tianyi Research Institute, “Fucheng-1” and “Shenqi” are two operational SAR satellites launched on 7 June 2023 and 24 September 2024, respectively. They deliver stable imaging, interferometry, and orbital control performance. With a strip-map mode resolution better than 2 m, they are the leading global commercial interferometric SAR satellites. On 20 November 2024, they achieved repeat-track dual-star interferometry with a vertical baseline within 150 m, and their shortest revisiting period was 4 days, demonstrating enhanced imaging and coherence capabilities. Their high-frequency revisiting and high resolution are beneficial in monitoring applications in defense, urban management, and disaster response. As shown in Figure 3, their high-resolution SAR images cover the Jinchuan mining area. “Shenqi” is based on the technology “Fucheng-1”, offering better SNR and ambiguity control. Table 2 lists their specific parameters.

3.2.3. CR

As shown in Figure 4, to meet the deformation-monitoring needs of the Jinchuan mining area and verify the accuracy of time-series InSAR results, nine 1 m triangular CRs were installed in the mining area. On 3 March 2025, researchers went to the Jinchuan mining area in Jinchang to adjust the pitch and azimuth angles of seven CRs. Based on the orbital parameters of the Sentinel-1A satellite (descending orbit, Path 33, Frame 463), the azimuth angle was set to around $13°$ from true north, and the pitch angle was set to about $20°$. The specific locations of the CRs were recorded with positioning equipment and are listed in

Table 3. Information on CR deployment locations.

ID	$\mathbf{Latitude} \mathbf{(}\mathbf{°}\mathbf{)}$	$\mathbf{Longitude} \mathbf{(}\mathbf{°}\mathbf{)}$	Height (m)	Mode
CR1	38.482	102.171	1558.54	Descending
CR2	38.479	102.170	1563.60
CR3	38.485	102.166	1559.92
CR4	38.479	102.162	1616.83
CR5	38.477	102.179	1629.61
CR8	38.464	102.166	1651.77
CR9	38.481	102.171	1551.08
CR10	38.481	102.166	1594.23
CR11	38.470	102.172	1608.33

.

4. Results and Discussion

4.1. Analysis of Coherence Coefficient Results

In this study, we systematically investigated the Jinchuan mining area in Jinchang, Gansu Province, through advanced TS-InSAR analysis, utilizing multi-source satellite data from “Fucheng-1”, “Shenqi”, and Sentinel-1A. In this research, we comprehensively evaluated their interferometric performance and rigorously examined the advantages and reliability of high-frequency small satellite constellations for enhanced mining area surveillance and deformation detection. As illustrated in Figure 5, in this study, we designated the 22 January 2025 image as the reference master image, with the “Shenqi” and “Fucheng-1” satellites maintaining an exceptionally vertical baseline control within 70 m, demonstrating superior orbital precision and remarkable collaborative operational capabilities. Our analysis demonstrated that as the spatiotemporal baseline increased, all interferometric pairs showed an average coherence coefficient degradation of approximately 0.15 over a three-month period relative to the master image. Notably, “Fucheng-1” and “Shenqi” maintained consistently high coherence coefficients ranging from 0.6 to 0.8 and exhibited temporal distribution characteristics in mean coherence coefficients similar to those of Sentinel-1A interferometric pairs with similar temporal baselines. These coherence characteristics provide critical advantages in optimal candidate point selection in TS-InSAR processing while substantially improving the precision and reliability of subsequent deformation inversion results.

This paper comprehensively evaluates temporal coherence characteristics between “Fucheng-1” and “Shenqi” satellites through comparative analysis with Sentinel-1A, employing both visual interpretation and quantitative statistical approaches. Figure 6 demonstrates the three groups’ temporal coherence coefficient time-series results across identical geographical regions using the RD model, revealing that the “Fucheng-1” and “Shenqi” dual-satellite constellation exhibits spatial distribution characteristics remarkably similar to those of Sentinel-1A, despite their differing satellite parameters.

To systematically quantify and evaluate the temporal coherence characteristics, specific interferogram pairs were selected with controlled baselines: for “Fucheng-1”, slave images were selected from 26 October 2024 and 11 January 2025, and for “Shenqi”, slave images were selected from 11 March 2025, with these images demonstrating temporal baselines of −88, 11, and 48 days, respectively. For Sentinel-1A analysis, we utilized slave images from 26 October 2024, and 6 January and 7 March 2025, corresponding to temporal baselines of −84, 12, and 48 days. As illustrated in Figure 7, the cumulative distribution of the interference coherence coefficient for “Fucheng-1” slave imagery with an 11-day temporal baseline exceeds 0.7 at $78.96\mathrm{\%}$. Similarly, for Sentinel-1A with an 11-day temporal baseline, the coherence coefficients reaching 0.7 comprised $75.67\mathrm{\%}$. Overall, the “Fucheng-1” and “Shenqi” dual-satellite constellation exhibits mean coherence coefficients slightly superior to those of Sentinel-1A at similar temporal baselines. Moreover, the constellation’s higher spatial resolution enhances its sensitivity to detailed variations in distributed targets, resulting in a marginally higher standard of deviations than those seen in the Sentinel-1A results. These coherence properties provide robust data support for subsequent phase unwrapping, the density of monitor points in time-series processing, and the reliability and precision of the final results.

4.2. Analysis of Differential Interferometric Effects

In Figure 8, the time-series differential interferograms for the “Fucheng-1” and “Shenqi” satellites, derived from the stacked intensity images, are compared with the Sentinel-1A differential interferograms. Due to its higher resolution, the interferometric performance of the “Fucheng-1” and “Shenqi” dual-star constellations is exceptional, with the differential interferograms exhibiting smooth and continuous phase variations. Except for the mining areas, where phase changes are more pronounced, the phase variations remain stable, and the interferometric capability is consistent over a long temporal baseline.

Using the “Fucheng-1” image from 22 January 2025 as the reference master image and the “Shenqi” image from 21 November 2024, with a 62-day temporal baseline, we conducted detailed monitoring for five mining areas in the Jinchuan mining district, taking advantage of the satellite’s high resolution. Figure 9 presents the differential interferometric effects from the dual-satellite constellation. The interferometric fringes are visible in the mining areas, and the deformation patterns are evident in the differential interferograms. Upon analyzing the interferometric fringes, it can be observed that the deformation in the P1, P2, and P5 mining areas is relatively small, while the P3 and P4 areas show significant deformation. In particular, the P4 mining area exhibits dense interferometric fringes over a 2-month span, suggesting a deformation of over 10 cm. Subsequently. TS-InSAR technology will be used for deformation monitoring and the analysis of mining activities in areas P1 to P5.

Beyond the visual assessment of differential interferograms, in this paper, we implement two quantitative evaluation metrics—SPD and PSD—to objectively compare the interferometric performance between the dual-star constellation and Sentinel-1A. The SPD quantifies the phase deviation between central pixels and their eight neighbors [32], while PSD measures phase smoothness across local regions or entire images [33]. In theory, lower SPD and PSD values indicate better phase smoothness, and as shown in Figure 10, account for resolution differences between the Sentinel-1A and “Fucheng-1”/”ShenQi” systems, even when the overlapping study area covers 172.01 km², leading to significant differences in interferogram dimensions. Consequently, in this study, we employed the mean values for both the SPD and PSD metrics to comparatively evaluate the quality of interferograms generated by the dual-satellite constellation and Sentinel-1A. Quantitative analysis reveals that the “Fucheng-1”/”Shenqi” constellation demonstrates a superior interferogram quality, showing 21.47% lower mean SPD values with 70.99% reduced standard deviation compared to Sentinel-1A. When using the mean PSD with a 19 × 19 sliding window, the dual-star constellation exhibits 35.47% lower mean PSD values accompanied by 70.36% lower standard deviation, conclusively demonstrating its enhanced phase quality in differential interferograms relative to Sentinel-1A.

4.3. TS-InSAR Deformation Monitoring Results

4.3.1. Sentinel-1A PS/DS-InSAR Deformation Inversion Results

Figure 11 presents the deformation results derived from the PS-InSAR and DS-InSAR processing of 15 Sentinel-1A scenes using January 18, 2025 as the master image, identifying 265,704 PS points with deformation rates ranging from −63.10 to 22.18 $\mathrm{m}\mathrm{m}/\mathrm{y}\mathrm{r}$ in the study area. Following phase optimization, the interferogram quality and monitoring point density significantly improved, with DS points increasing by a factor of 6.75 to reach 1,794,067 points, showing deformation rates from −62.74 to 19.14 $\mathrm{m}\mathrm{m}/\mathrm{y}\mathrm{r}$, while achieving point densities of 1544.66 $\mathrm{p}\mathrm{o}\mathrm{i}\mathrm{n}\mathrm{t}/\mathrm{k}{\mathrm{m}}^2$ for PS measurements and 10,429.77 $\mathrm{p}\mathrm{o}\mathrm{i}\mathrm{n}\mathrm{t}/\mathrm{k}{\mathrm{m}}^2$ for DS measurements. The deformation analysis demonstrates strong consistency between PS-InSAR and DS-InSAR results, particularly in terms of spatiotemporal distribution, with 87.03% agreement for PS-InSAR and 91.34% agreement for DS-InSAR in areas exhibiting deformation rates between −5 and 5 $\mathrm{m}\mathrm{m}/\mathrm{y}\mathrm{r}$, indicating relative stability in the Jinchuan mining district periphery. However, active mining areas displayed substantial deformation, causing decorrelation and phase unwrapping challenges due to large-scale ground displacement.

As shown in Figure 12, DS-InSAR demonstrates a significant improvement in monitoring point density within the mining areas compared to PS-InSAR. However, due to the relatively long temporal baseline of Sentinel-1A and the limitations of C-band satellites in rapid deformation monitoring, which leads to large deformation during the revisiting period, the P4 mining area experiences complete coherence loss, resulting in there being no deformation monitoring points in this area. In the P1 and P5 areas, the deformation gradually increases from the mining boundary toward the mine center, but no monitoring points are found at the center of the mining area. For P3, the deformation near the mine center is large enough to cause phase unwrapping errors, manifesting as apparent uplift errors in the central regions. This indicates that the relatively longer revisiting period and coarser resolution of Sentinel-1A lead to limitations in monitoring mining areas. These limitations cause complete decorrelation or phase unwrapping errors in the mining center, highlighting the challenges in accurately monitoring mining regions using Sentinel-1A data.

4.3.2. “Fucheng-1” and “Shenqi” PS-InSAR Deformation Inversion Results

Time-series PS-InSAR processing was conducted using 15 “Fucheng-1” images and 5 “Shenqi” images with 22 January 2025 as the reference master image, revealing significantly denser monitoring points due to the higher-resolution imagery, as demonstrated in Figure 13. The analysis yielded tens of millions of monitoring points with a remarkable density of 60,712.75 $\mathrm{p}\mathrm{o}\mathrm{i}\mathrm{n}\mathrm{t}/\mathrm{k}{\mathrm{m}}^2$, showing deformation rates ranging from −115.28 to 55.40 $\mathrm{m}\mathrm{m}/\mathrm{y}\mathrm{r}$. Notably, 86.99% of measured deformation rates fell within the stable range of −5 to 5 $\mathrm{m}\mathrm{m}/\mathrm{y}\mathrm{r}$, confirming the overall stability of non-mining areas while effectively capturing the substantial deformation characteristics in active mining zones through the enhanced point density and measurement precision enabled by the high-resolution satellite data.

4.3.3. Comparative Analysis of TS-InSAR Results: “Fucheng-1”, “Shenqi”, and Sentinel-1A

Figure 14 demonstrates that the Tianyi Institute’s dual-star constellation, with its shorter revisiting period and higher-spatial-resolution imagery, captures larger deformation magnitudes and higher monitoring point density in the five mining areas compared to Sentinel-1A while maintaining consistent deformation trends, which can enable a more detailed understanding of subsidence in mining areas.

During the early observation task, due to the limitations of C-band satellites in rapid deformation monitoring, the “Fucheng-1” and “Shenqi” satellites had a revisiting period of 11 days, with most of the imagery captured solely by “Fucheng-1”. After 20 February 2025, the observation density of “Shenqi” was increased in this region, leading to deformation trends in the P3 and P4 areas that were consistent with the Sentinel-1A results. P4 is located in the West No. 2 Mining Area of Longshou Mine, which is used for the pillarless sublevel caving mining method [34]. Under the influence of large-scale goaf areas and ground pressure, the roof develops cracks and fails, leading to a significant increase in surface subsidence. In the P4 central mining area, intensive mining activities caused significant deformation, with dense deformation fringes in the interferogram, resulting in an obvious decorrelation. As a result, there were no monitoring points in the internal regions of the mining area, except for the outer part of the mining area. In the P3 mining area, phase unwrapping errors also occurred, but for the P1 and P2 mining areas, the calculation results showed a relatively noticeable improvement compared to the Sentinel-1A time-series results. Backfill mining is employed at P1 and P2 [35,36]. The backfill bears the surrounding rock pressure, and its settlement primarily results from the compression of the backfill and the absence of contact between the backfill and the roof. The deformation magnitude is relatively smaller compared to that for P4. P3 has a mining history spanning several decades and a mining depth exceeding 300 m. The monitoring results are affected by recent intensive underground mining activities, high slope gradients, and severe surface collapse phenomena, specifically in the P2 mining center area, where the Sentinel-1A results exhibit severe decorrelation, resulting in the absence of monitor points and the presence of unwrapping errors. However, based on the “Fucheng-1” and “Shenqi” satellite data, the maximum deformation rate in the center of this area reached −115.28 $\mathrm{m}\mathrm{m}/\mathrm{y}\mathrm{r}$, indicating a more accurate and detailed deformation measurement compared to Sentinel-1A. This demonstrates the advantage of using higher spatial resolution datasets from multiple satellites, particularly in regions with moderately intense mining activities where deformation is more significant. The increased observation frequency and the combination of data from different satellites enable better monitoring and more precise deformation rate estimation, especially in challenging areas with phase unwrapping issues.

Figure 15 and

Table 4. Time-series PS-InSAR deformation rate and cumulative deformation results at points G1-G5 based on the “Fucheng-1” and “Shenqi” dual-star constellation.

ID	$\mathbf{Deformation} \mathbf{\left(}\mathbf{m}\mathbf{m}\mathbf{\right)}$	$\mathbf{Rate} \mathbf{(}\mathbf{m}\mathbf{m}\mathbf{/}\mathbf{y}\mathbf{r}\mathbf{)}$
G1	−52.45	−83.00
G2	−65.32	−115.28
G3	−30.81	−62.84
G4	−18.31	−37.70
G5	−13.29	−24.22

presents the cumulative deformation analysis for the five mining areas P1–P5; the central areas of P3 and P4 are excluded due to decorrelation or phase-unwrapping errors, necessitating the use of the peripheral points G3 and G4 for these regions, while points G1, G2 and G5 are selected in P1, P2 and P5, respectively, based on maximum subsidence rates. The analysis of annual deformation rates and cumulative deformation values reveals consistent subsidence trends from October 2024 to April 2025 across all monitoring points G1-G5, with G1 and G2 demonstrating linear deformation patterns and G2 accumulating the maximum deformation of −65.32 $\mathrm{m}\mathrm{m}$, whereas points G3–G5 exhibit stabilization tendencies after March 2025. We recommend that subsequent monitoring of the mining areas be carried out, with frequent imaging tasks.

A profile analysis of annual deformation rates was performed for areas P1 and P2 using 50-m buffer zones along the line segments L1–L3, with the buffer zones containing 21,100, 7900, and 33,700 PS monitoring points, respectively, as shown in Figure 16. The deformation rate profiles reveal a funnel-shaped subsidence pattern in both P1 and P2 areas, while demonstrating relative stability at the 2200-m position along the profile line L3 between these areas, indicating that mining activities in P1 and P2 have not induced large-scale deformation in the intervening region.

4.4. CR-InSAR Deformation Monitoring Results

Nine artificial triangular CRs measuring 1 m in side length were deployed in the Jinchuan mining area to ensure operational safety, with heading and elevation angle adjustments performed on 3 March 2025 to enable the CR-InSAR technology to perform localized deformation monitoring and the validation of TS-InSAR results. Four Sentinel-1A scenes acquired from February to March 2025 and downloaded from the ESA website were utilized for corner reflector identification and deformation analysis, taking both ascending and descending orbit characteristics into account to facilitate reflector detection and intensity variation assessment.

The RD model was employed to transform nine geographic CRs coordinates obtained through field measurements into SAR coordinates, with the initial visual identification of CRs performed using the Sentinel-1A intensity mean image, revealing an average offset of approximately 2 pixels following coordinate transformation. The pixels exhibited regular patterns, prompting the systematic adjustment of the CR SAR coordinates using this offset, with final precise positioning achieved through the application of the maximum-intensity principle within a $7\times 7$ pixel window, as illustrated by the CR intensity images and deployment locations presented in Figure 17. The weak background scattering energy enabled the clear visual identification of all nine triangular CRs, with Figure 17 displaying zoomed-in mean intensity images of the CR targets annotated with their corresponding intensity values, while quantitative analysis of the CR echo-signal strength using dB values revealed that the central reflector points exhibited intensities exceeding 60 dB and occupied between 4 and 6 pixels in the imagery.

Figure 18 demonstrates the intensity time series for all nine CRs, showing that after heading and elevation angle adjustments were made to seven CRs (excluding CR01 and CR09, which required no adjustment), their dB values increased substantially from the initial 50–60 dB levels to 68–75 dB, representing a nearly 15 dB enhancement, with the mean intensity images revealing that all nine triangular CRs achieved approximately 70 dB intensity.

The accuracy estimation of phase deformation monitoring using corner reflectors is typically based on their signal-to-clutter ratio (SCR) [37,38].

For targets such as corner reflectors, the complex echo signal from a resolution cell can be considered as the vector sum of the corner reflector’s echo signal and other clutter signals. The greater the signal energy of the corner reflector, or the higher its energy ratio to clutter, the smaller the resulting phase error. The relationship between SCR and phase error can be expressed using Formula (4) [37]:

$${\sigma }_{{\varphi }_P\approx }{\displaystyle \frac{1}{\sqrt{\mathrm{S}CR}}}$$

(4)

This relationship can be converted to deformation error based on wavelength:

$$d_{{\varphi }_P}={\displaystyle \frac{\lambda }{4\pi }}\ast {\sigma }_{{\varphi }_P}$$

(5)

It can be observed that with the same accuracy requirements, sensors with longer wavelengths demand a higher SCR. For instance, to achieve 1 mm monitoring accuracy, a C-band sensor requires an SCR of 10 dB. As shown in Figure 18, the CRs deployed in this study all exhibit SCR values around 20 dB, which can meet the requirements for high-precision monitoring.

Following topography residual compensation in differential interferometric phases using GPS-measured corner reflector heights, significant atmospheric delays persist in certain differential interferograms, potentially distorting deformation measurements. This study adopts the same strategy as PS-InSAR, performing large-scale spatial filtering on unwrapped phases to remove these spatially correlated low-frequency atmospheric phase components. Figure 19 provides a comparative visualization of differential interferograms (master, 20250319; slave, 20250331) before and after atmospheric phase correction.

As shown in Figure 18 and Figure 20, CR08 served as the reference point due to its relatively high signal-to-clutter ratios and more stable cumulative deformation, which were observed through the TS-InSAR analysis of Sentinel-1A and the dual-satellite constellation. Figure 21 and Figure 22 display the triangulation network established using the nine CRs, with MCF phase unwrapping and least-squares adjustment revealing subsidence across all nine points, particularly at CR09, which exhibited the most significant deformation with a cumulative LOS rate of −65.27 $\mathrm{m}\mathrm{m}/\mathrm{y}\mathrm{r}$ and total deformation of −4.29 $\mathrm{m}\mathrm{m}$, as detailed in

Table 5. Deformation inversion results for CR (with CR08 as the reference point).

ID	$\mathbf{Last} \mathbf{Deformation} \mathbf{\left(}\mathbf{m}\mathbf{m}\mathbf{\right)}$	$\mathbf{Rate} \mathbf{(}\mathbf{m}\mathbf{m}\mathbf{/}\mathbf{y}\mathbf{r}\mathbf{)}$
CR01	−0.92	−14.04
CR02	−3.74	−56.91
CR03	−0.62	−9.44
CR04	−2.09	−31.77
CR05	−2.68	−40.82
CR08	0.0	0.0
CR09	−4.29	−65.27
CR10	−2.95	−44.89
CR11	−1.25	−19.00

.

As shown in Figure 23, the PS-InSAR LOS deformation results were projected vertically using the incidence angle parameters of the “Fucheng-1” and “Shenqi” dual-star constellation, with CR08 as the reference. The results show reasonable consistency between the time-series results of the nine CR points and the CR-InSAR results. This demonstrates the effective time-series monitoring capability of “Fucheng-1” and “Shenqi”.

5. Conclusions

For this paper, we used time-series PS-InSAR technology to monitor and analyze deformation in the Jinchuan mining area in Jinchang City, using 15 scenes with “Fucheng-1” images and 5 scenes with “Shenqi” images. Currently, the research employing dual-star configurations for bistatic interferometric and time-series analysis remains limited, which establishes the significance of this study in quantitatively assessing the interferometric characteristics and time-series monitoring capabilities of the “Fucheng-1”/”Shenqi” dual-star constellation. The conclusions are presented as follows:

• The “Fucheng-1” and “Shenqi” SAR satellites, with their higher spatial resolution and shorter revisiting period characteristics, maintained a spatial vertical baseline within 70 m in this study. Compared to the Sentinel-1A satellite, the interferometric coherence exhibited similar temporally and spatially distributed characteristics, and the overall interferometric effect showed relatively clear improvements. In terms of time-series processing, the high-resolution characteristics of the dual-star constellation resulted in a denser monitoring point distribution, which stood out against the prominent consistency of the spatial deformation characteristics in the Sentinel-1A results. Overall, the calculation results align with the regional outcomes from Sentinel-1A, with certain mining areas demonstrating a relatively superior performance compared to the results obtained from Sentinel-1A.
• Nine CR points were deployed in the Jinchuan mining area to verify the reliability of the results from “Fucheng-1” and “Shenqi”. A comparison of CR-InSAR results with Sentinel-1A PS-InSAR results showed a relatively high degree of consistency in the monitoring outcomes.
• We used “Fucheng-1” and “Shenqi” images from 21 November 2024, to 13 April 2025. Before 24 February 2025, only one “Shenqi” image was available. Most image revisiting periods were 11 days due to the limited number of “Shenqi” images. This caused decorrelation and unwrapping errors in some areas. To enhance the result accuracy, future work should maintain dual-star imaging and keep the temporal baseline within seven days.

The results demonstrate that the “Fucheng-1” and “Shenqi” satellite constellation exhibits superior orbit control accuracy, enhanced bistatic interferometric capability, and improved time-series processing performance, with their high revisit frequency proving particularly advantageous in monitoring rapid mining subsidence, suggesting that their high-frequency and high-resolution characteristics should be fully utilized in future research and engineering applications to enhance the effectiveness of regional deformation monitoring.