Dynamic digital elevation models (DEMs) with high accuracy are crucial for interpreting and retrieving deformations from interferometric synthetic aperture radar (InSAR) observations. The inversion deformation accuracy decreases when the DEM is inaccurate. For the traditional multi-temporal InSAR (MT-InSAR) deformation monitoring, it is assumed that the terrain does not be changed during the monitoring time period, and the DEM error is taken as a fixed parameter to be separated. However, if the surface height changes during the monitoring time period, the traditional methods will fail because the DEM error is no longer a fixed parameter. Therefore, this paper proposes a new dynamic estimation method of surface height change (i.e., DEM errors) to solve this problem. The method is based on the principle that the DEM errors before and after the surface height change have two linear relationships with the perpendicular baseline, and solves for the surface height change by finding the time point that minimizes the sum of squares of the residuals when solving the equations. The simulated and real experimental results show that the new method can accurately acquire the dynamic surface height changes and the timeline of surface height changes, which can accurately remove the DEM error from different unwrapped phases, and contribute to the estimation of deformation rates and deformation time series in height change regions.

The three-runway system expansion project of the Hong Kong International Airport (HKIA) began with the land reclamation to the north of the original runway. Understanding its ground deformation is essential for subsequent civil construction and planning at the new land. Synthetic Aperture Radar Interferometry (InSAR) technique is firstly used to investigate the spatiotemporal characteristics of land subsidence after the completion of third runway pavement. Due to the consolidation of underlay materials, the third runway is subject to varying degrees of land subsidence, with the monitored maximum sinking rate to be ~100 mm/year during September 2021 to October 2023. We adopted the Independent Component Analysis (ICA) to separate the underlying sources in order to explore the spatiotemporal characteristics of deformation in the reclaimed land. The results show that there are three distinct deformation sources in the study area, including an exponential decay signal (an exponential decay consolidation process), a periodic signal (thermal effects correlated with buildings and bridges) and a linear signal (a continuous subsiding). Considering the different reclamation methods, the linear deformation component is mainly located in areas with prefabricated vertical drains (PVD), which is strongly associated with the overall subsidence pattern. On the other hand, the land reclaimed by Deep Cement Mixing (DCM) method tends to reach a stable state earlier than those reclaimed by the PVD method, demonstrating the effectiveness of the DCM in reinforcing the reclamations. These results benefit our understanding of the settlement process over the third runway of HKIA and provide reliable suggestions for follow-up reinforcement plans on specific locations if needed.

Airborne light detection and ranging (LiDAR) is a fast ground observation technology that can obtain point cloud data with centimeter precision in height and centimeter resolution in spatial domain, which has been widely applied to construct digital element model (DEM). However, there are still limitations in deformation monitoring, especially in large-scale and high-precision surface deformation monitoring. Namely, when the deformation magnitude is close to the LiDAR monitoring accuracy, the difference results contain obvious strip errors. To solve this problem, a strip error correction method based on the differential results is proposed for airborne LiDAR deformation monitoring including three main steps: quasi-stable adjustment on elevation correction, fixed-axis iterative closest point (ICP), and affine ICP algorithm to correct the strip errors. Experiments show that the proposed method can correct the strip error, and the RMSE of the two-phase point clouds for one flight track was decreased from the original 6.2 and 5.3 cm to 3.5 and 3.3 cm. Deformation patterns and rates are consistent as verified by external results with the small baseline subset InSAR (SBAS-InSAR) of Sentinel-1 SAR data. An internal comparison with the classical ICP algorithm shows this method is more accurate. The effectiveness of the proposed method in monitoring surface deformation is verified.

In recent years, many synthetic aperture radar (SAR) satellites have been launched to provide abundant SAR data, which creates a demand for dynamic monitoring of surface deformation. In general scenarios, we process SAR images within a fixed period to obtain the deformation time series. However, conventional interferometric SAR (InSAR) processing involves manually selecting a specific time period, making it impossible to capture the complete deformation process. Therefore, we propose a novel framework, that is, the sequential small baseline subset InSAR (SBAS-InSAR) backward estimation algorithm, which utilizes sequential least-squares adjustment to dynamically recover previous surface deformation time series. To validate the effectiveness of the proposed method, we conduct both simulated and real data experiments. Also, the simulated results obtained by new method are totally consistent with the traditional SBAS-InSAR method, while the standard deviation of the difference between the deformation time series obtained by InSAR and GNSS results is better than 6 mm. Moreover, the new method has higher computation efficiency.

Interferometric synthetic aperture radar (InSAR) technology is the most advanced and effective method for monitoring large-scale slow-moving landslides. However, automatic landslide detection based on InSAR products regarding landslide samples and deep learning models is still challenging. Different InSAR products are inconsistent for slow-moving landslide detection due to the fuzzy boundaries of potential landslides and complex deformation characteristics. In addition, the accuracy of existing landslide detection models is not high because multiscale factors in feature abstract and feature fusion are rarely considered. This article proposes a multiscale Swin Transformer InSAR products detection network (MSIDNet) to detect slow-moving landslides automatically. First, we adopt an advanced Swin Transformer as the backbone to extract features and multiscale spatial-temporal attention blocks in the neck to improve the feature fusion capability. Meanwhile, to train the new model, we built a landslide dataset based on visual interpretation, including deformation rate, phase gradient, and C-index (GRCI). Experiments demonstrate that our proposed method outperforms the existing deep learning models such as Faster R-CNN and Yolov3. The GRCI dataset is more efficient and less biased than the traditional deformation rate and phase gradient datasets. The precision of detected landslides in a given testing area is 0.89, and it has good generalization ability. This study provides a new dataset and method for slow-moving landslide detection based on InSAR products.

Deep learning networks have achieved great success in many areas, such as in large-scale image processing. They usually need large computing resources and time and process easy and hard samples inefficiently in the same way. Another undesirable problem is that the network generally needs to be retrained to learn new incoming data. Efforts have been made to reduce the computing resources and realize incremental learning by adjusting architectures, such as scalable effort classifiers, multi-grained cascade forest (gcForest), conditional deep learning (CDL), tree CNN, decision tree structure with knowledge transfer (ERDK), forest of decision trees with radial basis function (RBF) networks, and knowledge transfer (FDRK). In this article, a parallel multistage wide neural network (PMWNN) is presented. It is composed of multiple stages to classify different parts of data. First, a wide radial basis function (WRBF) network is designed to learn features efficiently in the wide direction. It can work on both vector and image instances and can be trained in one epoch using subsampling and least squares (LS). Second, successive stages of WRBF networks are combined to make up the PMWNN. Each stage focuses on the misclassified samples of the previous stage. It can stop growing at an early stage, and a stage can be added incrementally when new training data are acquired. Finally, the stages of the PMWNN can be tested in parallel, thus speeding up the testing process. To sum up, the proposed PMWNN network has the advantages of: 1) optimized computing resources; 2) incremental learning; and 3) parallel testing with stages. The experimental results with the MNIST data, a number of large hyperspectral remote sensing data, and different types of data in different application areas, including many image and nonimage datasets, show that the WRBF and PMWNN can work well on both image and nonimage data and have very competitive accuracy compared to learning models, such as stacked autoencoders, deep belief nets, support vector machine (SVM), multilayer perceptron (MLP), LeNet-5, RBF network, recently proposed CDL, broad learning, gcForest, ERDK, and FDRK.

Multitemporal interferometric synthetic aperture radar (InSAR) technology has become an important tool for digital elevation model (DEM) generation and surface deformation monitoring. The prerequisite is to obtain the correct unwrapped phase of the interferograms. However, it is still challenging to obtain reliable results by traditional 2-D phase unwrapping (PU) algorithms and even 3-D PU algorithms, when the phase continuity assumption fails due to large deformation gradients or abrupt topographic changes. This letter proposes an${L}^{ {1}}$-norm sparse 2-D PU algorithm based on reliable redundant network in the phase discontinuous region, which is obtained by implementing ambiguity detector on free network within a given distance. Experiment on real synthetic aperture radar (SAR) interferograms covering urban tall buildings verified that the algorithm can obtain more reliable results in phase discontinuous regions than traditional PU algorithms.

Digital elevation models (DEMs) are vital in the geosciences and many other fields. Interferometric synthetic aperture radar (InSAR), an advanced earth observation technology, has shown its potential in DEM reconstruction. Multi baseline InSAR (MB-InSAR) is currently improving the precision of DEM reconstruction by combining multiple interferograms. However, MB-InSAR for DEM generation can result in severe decorrelation, which may cause significant gaps in the final DEM product. To solve this problem, an improved MB-InSAR DEM reconstruction method is proposed in this study, which we term as the dynamic DEM calculation algorithm. The proposed method can estimate the DEM pixel-by-pixel, which allowed us to select the interferograms dynamically and therefore minimize the void values. For the performance test of the proposed method, 25 ascending and 20 descending TerraSAR-X images over Heifangtai (China) were collected to form repeat-pass interferograms and produce the DEM using the proposed method. Results showed that the number of valid pixels increased by approximately 20% compared with the traditional MB-InSAR DEM reconstruction method without loss of precision, thereby illustrating the feasibility of the proposed method.

The synthetic aperture radar (SAR) interferometry (InSAR) has been developed for more than 20 years for historical surface deformation reconstruction. In particular, the onboard Sentinel-1/A/B satellite, newly planned NASA-ISRO SAR (NISAR), and Germany Tandem-L will continue to provide unprecedented SAR data with an increased number of acquisitions. However, processing of real-time SAR data has been experiencing challenges regarding the InSAR deformation parameter estimation over a long time with the small baseline subsets (SBAS) InSAR technology. We use sequential adjustment for the estimation of the deformation parameters, which uses Bayesian estimation theory under the least square criteria to inverse long time-series deformation dynamically. Finally, both simulated and real Sentinel-1A SAR data verify the performance of the sequential estimation. It can be regarded as an effective data processing tool in the coming era of SAR big data.

Multi-source Synthetic Aperture Radar (SAR) datasets are used to investigate the deformation history and failure mechanism of small-scale loess landslide in the Heifangtai loess terrace, Gansu province, China. A total of 65 SAR datasets acquired by L-band ascending ALOS/PALSAR, X-band descending TerraSAR-X and C-band descending Sentinel-1A/B covering the different periods of Heifangtai terrace are fully exploited. In addition, groundwater level data are also involved to analyze the failure mechanism of loess landslide. The displacements occurred during the past eleven years were quantitatively identified for firstly by InSAR technique. The results show that three slopes, failed on October 1, 2017 had large cumulative deformations from January 2016 to November 2016. The acceleration dates of the deformation for the three slopes were successfully captured by Sentinel-1A/B data. Furthermore, the result shows that the magnitude of the landslide deformation is closely correlated to the groundwater level variation.

Automatic mapping of potential landslides is proposed by integrating optical remote sensing imagery, high-resolution DEM, and InSAR-derived deformation map. First, suspected landslide areas are mapped with an automatic landslide mapping algorithm based on two-dimensional continuous wavelet transform (2D CWT) to high-resolution DEM. Then, object-based image analysis (OBIA) method is introduced to extract suspected landslide areas with optical remote sensing imagery and deformation map. Last, the potential landslides are identified by determining the intersection of two suspected landslide areas. The method is tested in Heifangtai loess terrace, China and the identified potential landslides show a strong consistency with the landslides inventory mapping.

Aiming to mitigate the landslide hazard after event and to investigate the potential landslide over large area, the three-step landslide investigation strategy based on Synthetic Aperture Radar Interferometry (InSAR) is proposed, which includes (1) fast landslide mapping is calculated with small volume SAR data over specific landslide region on demand. (2) Pre-landslide deformation are precisely recovered with all archived SAR data to analyze the spatiotemporal landslide characteristics. (3) The potential landslides are fully investigated in the catchment scale with large volume SAR data. Sentinel-1A SAR data are continuously involved to give fast landslide monitoring and investigation over Baige landside and Jinsha River catchment.

The Linfen-Yuncheng Basin (LYB) is one of the most serious geo-hazards regions in China, which have been experiencing severe geo-tectonic movement, seismic, land subsidence and ground fissures. To monitor the complex surface deformation at LYB, Interferometric Synthetic Aperture Radar (InSAR) is employed. Forty-nine scenes acquired from three SAR tracks from 2007 to 2011 are used to obtain the ground deformation over LYB based on four multi-temporal InSAR analysis methods (i.e. PI-RATE, Stacking, SBAS and TCP). The maximum displacement is observed in Jishan County with the subsidence rate up to 120 mm/yr in the vertical direction. Two visually triangular areas are recognized at Jishan and Taocun-Xiaxian, which are controlled by local normal faults. In addition, cross-validation of the results in overlapping areas between adjacent tracks, ascending and descending tracks are conducted, which show good consistency among four methods.

Taiyuan graben basin has been suffering serious ground deformation due to the combined effects of groundwater pumping, ground fissures and faults. However, most ground deformation monitoring mainly focus on the single city, thus it is difficult to reveal the regional extent of the deformation. Fifty-five ALOS/PALSAR images acquired from three adjacent tracks from 2007 to 2011 are used to obtain annual deformation rate based on Small Baseline Subsets (SBAS-InSAR) technique. The standard deviations of deformation differences between two adjacent tracks are 3.7mm/a and 4.2mm/a, respectively. Large subsidence funnels mainly occur in the regions, where the groundwater pumping is intensive. The maximum subsidence is observed in Qingxu with the land subsidence rate up to 240mm/a, while regions with subsidence rates ranging from 50 to 90mm/a are detected in other locations. Meanwhile, significant deformation differences are retrieved in both sides of the faults, which could be explained as the faults-controlled deformation.

Underground mining induced displacements in most areas of China amount to meter-level while with small spatial coverage, spatially discontinuous and temporally nonlinear features. Traditional phase-based InSAR methods can hardly obtain large deformation in the center of land subsidence area due to the phase noise, maximum monitoring ability. This paper systematically studies the offset tracking technique based on SAR image intensity maps with and without pre-installed Corner Reflectors (CRs).The results of this experiment indicate that the high resolution SAR data with the aid of pre-installed CR points can better solve the large gradient deformation monitoring problem. In addition, offset tracking method has the potential to achieve two-dimensional deformation field along the line-of-sight and along the azimuth directions, which can provide more detailed information regarding mining induced deformation, which is complimentary to the traditional phase-based and intensity-based techniques.

InSAR has been widely used in monitoring land subsidence over large area. However, many factors in InSAR processing, such as decorrelation error, atmosphere error, height error and thermal noise limit the accuracy of InSAR measurements. The height error over urban area is particularly the most difficult issue in TerraSAR-X data processing for its shorter wavelength and higher spatial resolution. We employed a Multi-temporal InSAR (MTI) method based on the PS and SBAS method proposed by Hopper to estimate height of urban building and ground deformation simultaneously. By means of MTI method, the first raw urban DSM with 1,500 km2areas over urban area has been mapped with a height accuracy of about 5 m. The MTI-derived deformation shows that the established TerraSAR DSM reduced the height error influence on deformation phase effectively. GPS and leveling measurements are applied to calibrate the InSAR results. Precision of our InSAR annual subsidence can reach 6 mm.

Taiyuan Basin is one of the regions with complex geological structure and frequent seismic activities in China. Because of groundwater exploitation and tectonic activity, land subsidence has occurred widely in this region. For disaster prevention and mitigation, obtaining ground deformation characteristics within the basin is an urgent requirement. The Interferometric Synthetic Aperture Radar (InSAR) was used for mapping the ground deformation in this area. The ground deformation characteristics over Taiyuan Basin from 2007 to 2011 were obtained and the cross-correlation between the regional ground subsidence and fault activity were analyzed.

Three InSAR products and DEM data are involved to detect the potential landslides over Wudongde Hydropower Station Section, Jinsha River. More than ten known and unknown landslides are successfully detected. Meanwhile, Small Baseline Subsets (SBAS) InSAR technique is applied to calculate the time-series deformation of the Jinpingzi landslide. Then, in-situ georobot measurements are used for point-wise comparison. The precision in slide direction is 1.8 cm by comparing with in-situ georobot measurements. At last, the Erpingcun landslide is taken as an example to calculate the vertical and horizontal deformation components by fusing the ascending and descending line-of-sight results.