Conference Agenda

China has been affected by some of the world's most serious geological disasters and experiences high economic damage every year. Geohazards occur on remote and highly populated areas.

In the framework of the DRAGON4 32365 Project, this paper presents the main results and conclusions derived from an extensive exploitation of available remote sensing data and methods that allow the evaluation of their importance for various geohazards. Therefore, the great benefits of recent remote sensing data (wide spatial and temporal coverage) that allow a detailed reconstruction of past events and to monitor currently occurring phenomena are exploited to study various areas and various geohazard problems, including: surface deformation of the mountain slopes and glaciers; identification and monitoring of ground movements mining areas and; subsidence, landslides, ground fissure and building inclination studies. Suspicious movements detected in the different study areas were verified and validated by field investigation and measurements in the local.

In the framework of the DRAGON-4 Project, the National Institute of Geophysics and Volcanology of Rome (INGV, Italy) and the Northeastern University of Shenyang (China) collaborate to study the surface movement over industrial regions in Northeast China. The traditional heavy industrial base, especially in the Benxi-Anshan-Shenyang-Fushun (BASF) region, is playing an important role in the economic development of the region, although severe consequences on the local environment are taking place due to the continuous mining activities. Various geo-hazards, such as subsidence, landslides, ground breakage and building inclinations, have been occurring for decades. The continuous monitoring of the effects of the mentioned phenomena is thus of great importance for the safety of the local population. Taking advantage of the availability of dense remote sensing dataset it is possible to analyze the geo-hazards and their environmental impacts in the region; and then making forecast about their occurrence in the future and providing support for disaster prevention and damage reduction.

We adopt multi-temporal InSAR methodologies able to estimate the spatial and temporal deformation over large areas. In this study, we use time-series InSAR results from multiple stacks (from ascending and descending orbits) and different sensors to monitor gravitational deformations and subsidence phenomena in urban areas especially effecting underground paths and railways, and in mining regions.

We take advantages from both the Persistent Scatterers Interferometry (PSI) and Small Baseline Subset (SBAS) techniques, for processing SAR data stacks acquired by Sentinel-1A/B (C-Band) and COSMO-SkyMed (X-Band) exploited for several areas in the BASF region.

In comparison to the 2018 already presented mid-term project results, we investigated the same areas of Fushun and Shenyang analyzing Sentinel-1 datasets, provided by ESA, along both the tracks. The considered time interval spans from June 2015 to December 2018 for the descending orbit, and from April 2017 to September 2018 for the ascending case, respectively.

The retrieved mean ground velocity maps confirm the results from the previous exploited CSK data. Subsidence phenomena are still ongoing reaching values higher than -120 mm/yr inside the mine and -60 mm/yr at its edges. Some small areas (i.e. localized groups of pixels) show positive values (uplift) probably due to stockpile of excavation debris and/or processing waste material.

Two descending stacks of COSMO-SkyMed stripmap images and a stack of TerraSAR-X images covering Shenyang city are exploited in a small baselines subset analysis (SBAS) in our recent study. During the processing of COSMO-SkyMed images，a stack of 18 images acquired from March 13^th 2016 to April 17^th 2017 covering eastern Shenyang and a stack of 15 images acquired from March 1^st 2016 to April 21^st 2017 covering western Shenyang are processed respectively. 58 interferograms are generated out of 18 SAR images for the eastern stack, whereas 44 interferograms are generated out of 15 SAR images for the western stack. In the meantime, 68 interferograms are generated out of 20 TerraSAR-X images acquired from April 15^th 2015 to October 5^th 2016 for SBAS analysis. The topographic phase is simulated and removed from the interferograms using the TanDEM-X DEM of 3-arc-second resolution(with spatial sampling of 90 m× 90 m) covering the study area. The SBAS approach has been proposed to overcome the limitation of decorrelation with reduced amount of SAR images by making full use of all possible interferograms with small spatial and temporal baselines. In this study, a modified SBAS approach developed in StaMPS to ensure the temporal continuity by connecting separated subsets of interferograms is implemented for data processing. The displacements acquired in line of sight direction is translated to vertical direction based on a simple assumption that no horizontal ground motion occurs for subsidence monitoring applications.

As recommended by the Guidelines of InSAR Monitoring for Geo-hazard of the Chinese InSAR community, areas presenting deformation velocities larger than 5mm/yr in LOS can be categorized as subsidence area. Taking the incidence angle into consideration, vertical deformation rate larger than 5.5 mm/yr suggests subsidence in Shenyang. Generally speaking, most parts of Shenyang are relatively stable. However, there is a large area in Tiexi district showing serious subsidence. According to the geological data in Shenyang, the basal ground in this area is generally composed of sandy soil and fine sand. The permeability of the basal ground in this area is quite strong, and therefore instability and ground subsidence could possibly occur in this area. Subsidence is also detected in Tawan, Yushutai and Xiaonanjie area. They have presented a strong connection to the groundwater funnel in Shenyang.

We also processed a new CSK dataset over the Anshan city along the descending track. Moreover, we updated the processing of CSK image dataset for the western part of the city of Shenyang, thanks some new CSK acquisitions (descending track) provided by the Italian space Agency (ASI),

Our results confirm that the heavy industrial exploitation of mines and water pumping in the BASF region of Northeast China cause clear and strong ground deformation effects of high potential impact on the local infrastructures and population. The use of multiple stacks, from different sensors, of InSAR data allows monitoring such phenomena with an accuracy and temporal sampling not possible earlier.

By now, the use of EO products plays a fundamental role to monitor natural and man-induced hazards and to support Disaster Risk Management providing an important tool for local and national organizations.

Acknowledgments

This work is financially supported in part by the National Natural Science Foundation of China (Grant No. 41601378) and the Fundamental Research Funds for the Central Universities (Grant No. N150103001). The COSMO-SkyMed data is provided by ASI via the ASI-ESA Dragon4 Project ID. 32365_4. The TerraSAR-X data is provided by Airbus Defence and Space.

We present a conceptual framework that integrates data-driven modelling with remote sensing to detect and delineate landslide phenomena. The main objective of the associated MSc thesis is to test and implement the developed methodology within an Alpine study site and the Longnan region (China).

The methodological framework includes (i) an initial screening and collection of available data sets (e.g. on past landslide events, environmental data, satellite products) which can then be used to (ii) explore landslide susceptible terrain using data-driven modelling procedures. In this context, EO-based predictor variables (e.g. SRTM-derivatives, land cover information) as well as available landslide information (landslide inventory) will be included into supervised statistical/machine-learning classification techniques. The resulting spatial information on landslide prone zones allows restricting the main area of interest for the subsequent remote sensing based analysis. More specifically, optical remote sensing data (e.g. change detection based on Sentinel-2) will be tested for their potential to identify and map recent landslide phenomena. The ensuing landslide information is expected to further enhance the knowledge on the spatio-temporal occurrence of recent landslide events and to improve the previously described data-driven landslide susceptibility assessment. Depending on the progress of the previous activities, also the potential of Sentinel-1 data (e.g. SAR Interferometry) may be tested to acquire information on slope deformation. The activities associated to this MSc thesis will start in April 2019 and will be completed within December 2019.

Radar Interferometry (InSAR) can provide measurements of surface displacement from Space, with millimetric accuracy [1, 2]. These measurements are used in natural hazards analyses but also for monitoring anthroprogenic activities. In the last few years, the number of SAR satellites with shorter repeat intervals and higher resolutions make increase significantly SAR data volume. This increase as lead to challenges in terms of manual inspection [3], giving in turn rise to the search of automated ways to process the available data. The point previously described and advances in hardware lead to advances in deep learning, which has already been applied in several areas such as computer vision.

We propose a supervised Deep Learning (DL) approach for multivariate outlier detection in post-processing of multitemporal InSAR (MTI) results. We used a Convolution Neural Network (CNN) to process the data leading to one of the following labels: outliers, inliers or potentially dangerous lower coherence points. The input data were organized in such a way that for each point the model has access to the multivariate features (such as velocity, height, etc.) of the nearest points, as well as its coordinates in a local system (centered on each point).

After training and model evaluation, the accuracy, precision and recall were analyzed (the last two for each label), considering a threshold value of 0.6 applied to the model’s output. Our model achieved a 95% accuracy and a mean value of 89%, respectively in precision and recall.

Our research intends to demonstrate the usefulness of DL to detect deformation patterns in post-processing InSAR data, with the purpose of increasing point densities of Permanent Scatterers (PS) point networks, thus enhancing the reliability of InSAR post-processing data.

References

[1] Crosetto, M., Monserrat, O., Cuevas-González, M., Devanthéry, N., Crippa, B. (2016). Persistent Scatterer Interferometry: A review. ISPRS Journal of Photogrammetry and Remote Sensing, 115, 78-89.

[2] Bakon, M., Oliveira, I., Perissin, D., Sousa, J., Papco, J. (2017). A Data Mining Approach for Multivariate Outlier Detection in Postprocessing of Multitemporal InSAR Results. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 10, NO. 6., 2791-2798.

[3] Anantrasirichai, N., Biggs, J., Albino, F., Hill, P., & Bull, D. R. (2018). Application of Machine Learning to Classification of Volcanic Deformation in Routinely Generated InSAR Data. Journal of Geophysical Research: Solid Earth, 123, 6592–6606.

Landslide is a hazard that threaten the people who lives in the mountain area, it comes active especially rainy seasons and causes a large number of casualties every year. The movement of the slope is an indicator of activity of the landslide, it is helpful to capture the precursor of the activity, the monitoring of the movement of the slope is very important. Subsidence is a "slowly vary geological hazard". Because of its lagging response and slow progress, subsidence is in mm and not easy to detect. It has the characteristics of long formation time, wide influence range, difficult prevention and control, and difficult to recover. High-accuracy displacement monitoring can help us obtain improved knowledge on the subsidence and landslides. In this work we will show the capability of up-to-date Advanced Land Observing Satellite-2 (ALOS-2) Synthetic Aperture Radar (SAR), Envisat ASAR On-the-Fly Data, Archived Sentinel-1 Data in monitoring the movement of the landslide and subsidence in China, which can capture the fast and slow movement with different spatial and temporal baseline combination, the results shows that the SAR data has its advantage in monitoring the movement of the landslides and subsidence in mountain and city area.

Abstract

Dagushan Iron Mine is the deepest open-pit iron mine in Asia, with abundantiron ore resources. With continuous open-pit mining activities, the stairs extend to underground step by step, and engineering geological conditions are gradually revealed. The factors affecting slope stability are also changing gradually, e.g., exposure of surface water and groundwater. The lithological structure and composition of the slope body are also changing, as well as the effect of blasting on the orebody during mining process, along with the change of the slope safety and stability. As a huge artificial loose accumulation body, instability of the dump will lead to disasters and major engineering accidents for the mine, which not only affectingproductivity, but also causing huge economic loss. Therefore, in order to ensure the safe operation of the mine, it is necessary to conduct slope stability monitoring with non-contact strategy. This kind of non-contact monitoring doesn’t need to install measurement points on the dangerous slope, and thus no need to worry about sliding problems of the measurement points.

As an effective non-contact deformation monitoring tool, SAR interferometry has good potential in displacement monitoring of mines. With a stack of SAR images, time series InSAR is able to overcome spatial and temporal decorrelation problems, as well as the atmospheric phase artifacts, resulting in high precision deformation estimates[1][2]. Amongst various time series InSARalgorithms, small baseline subsets analysis (SBAS) is able to estimate deformation using all the high quality interferograms, which improves the utilization of SAR data and is suitable for analysis on long time series[3]. Therefore, the SBAS method is used to monitor the displacements in Dagushan open-pit iron mine. In this paper, 117 sentinel-1 images acquired from 2017 to 2019 are used, as well as the 3-arc-second DEM generated by the German TanDEM-X mission[4][5]. With height accuracy of approximately 1m, TanDEM-X DEM can be used to remove the topographic phase from the interferograms.

During data processing, a super master is first selected according to the spatial and temporal baselines. All the slave images are coregistered to the super master image during coarse coregistration and fine coregistration. Then, high quality interferograms with small spatial and temporal baselines are generated following a multi-master strategy. With the high density in time and space, as many interferograms as possible participate in displacement estimation. The short spatial baselines can reduce the influence of DEM error on deformation estimates. In order to improve the quality of interferograms, Goldstein filter is applied on all interferograms. Then, phase unwrapping based on minimum cost flow is conducted for each interferogram.The residual topographic artifacts, as well as the atmospheric phase screen (APS) signals, are also estimated and filtered out. Based on the unwrapped interferograms, the average displacement rate and displacement time series are estimated using singular value decomposition method. The estimated displacements map in line-of sight direction show that the northern slope, western part and the northern part of the dump suffer from severe displacements. In order to assess the precision of the displacement estimates, a comparison with on-site date collected by measurement robots is carried out. There is a very good consistency between the two results. The outcome of this study can help with mine disaster prevention and mitigation, and provide technical support for ensuring safe mining activities.

Keywords: small baseline subsets analysis, displacement monitoring, open-pit mine

References

[1]Ferretti, A.; Prati, C.; Rocca, F. Nonlinear subsidence rate estimation using permanent scatterers in differential SAR interferometry. IEEE Trans. Geosci. Remote Sens. 2000, 38, 2202–2212.

[2] Ferretti, A.; Prati, C.; Rocca, F. Permanent scatterers in SAR interferometry. IEEE Trans. Geosci. Remote Sens. 2001, 39, 8–20.

[3] Berardino P, Fornaro G, Lanari R, et al. A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms[J]. IEEE Transactions on Geoscience & Remote Sensing, 2003, 40(11):2375-2383.

[4] Torres R ,Snoeij P , Geudtner D , et al. GMES Sentinel-1 mission[J]. Remote Sensing of Environment, 2012, 120(6):9-24.

[5] Huber, M.; Gruber, A.; Wendleder, A.; Wessel, B.; Roth, A.; Schmitt, A. The Global TanDEM-X DEM: Production Status and First Validation Results. In Proceedings of the 2012 XXII ISPRS Congress International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Melbourne, Australia, 25 August–1 September 2012; Volume XXXIX-B7, pp. 45–50.

Glacier motion represent a significant reference for the hazard assessment of glacier and glacial lakes. GF-3, as the first civil spaceborne synthetic aperture radar satellite in China, has important advantages in monitoring glacier motion due to its characteristics of all-weather, all-time capabilities and high spatial resolution. In this paper, based on five GF-3 images with FSⅡ imaging modes, the surface velocities of the Yiga Glacier, located in Nyenchen Tonglha Mountains, are estimated over five time periods using offset tracking technique during November 2017 to March 2018. The results were compared with the offset tracking results of sentinel-1 images which have a similar time with GF-3 image and based on the assumption that the velocity of the bedrock in the study area should be 0, the velocity residuals of the bedrock in each period are calculated, then the applicability of GF-3 image in monitoring glacier surface motion was evaluated. The results of GF-3 images show that the distribution of Yiga Glacier motion is similar in four periods, and the maximum surface velocities are all distributed in the central part of the glacier where the elevation changes dramatically. Meanwhile, the results are consistent with the results of sentinel-1 based on two images. The RMSEs of velocity residuals in the bedrock area in four periods are 1.4 cm/d, 2.0 cm/d, 1.7 cm/d and 2.3 cm/d, respectively, which validate the reliability of the deformation estimated used GF-3 images in this paper. Based on the above analysis, GF-3 SAR data can be used as one of the conventional data sources for monitoring glacier surface movement. Because of its high spatial resolution and high cost performance, GF-3 can play a unique role in monitoring the motions of glaciers.

Land subsidence is one of the most common environmentalproblems inurban areas around the world ^[1,2]. It has been hindering social stability and sustainable development for a long time.The deformation of the earth's surface and the structures upon it is usually a long-term gradual process. As the economic and cultural center of northeast China, Shenyang is developing rapidly in recent decades.With continuous above-ground and under-ground construction, Shenyang is suffering from continuous subsidence during a long time span.Therefore,continuous subsidence monitoring is essential in Shenyang.

As aspaceborne geodetic technology, synthetic aperture radar Interferometry (InSAR) is widely used in surface topography measurement and deformation monitoring.Using a stack of SAR images, Time Series InSAR is capable of overcoming decorrelation problems and monitoring land subsidence with very high accuracy. Several Time Series InSAR technologies such as Persistent Scatterer SAR Interferometry (PSI) ^[3], Small Baselines Subsets Analysis(SBAS) ^[4],Pixel Offset Tracking (POT) ^[5] and otherInSARtime series analysis algorithms have been widely used to monitor surface deformation.In this paper, three stacks of high-resolution TerraSAR-X and COSMO-SkyMed datasets are used to monitor the ground subsidence of Shenyang by means of SBAS. The COSMO-SkyMed images are acquired in descending orbit, including 15 images covering western Shenyang and 18 images covering eastern Shenyang. Both stacks are acquired during March 2016 and Apirl 2017. The 20 TerraSAR-X images are acquired in ascending orbit from August 2015 to October 2016. Besides, TanDEM-X DEM of 3-arc-second resolution(with spatial sampling of 90 m × 90 m) covering the study area is used to simulate and remove topographic phase from the interferograms ^[6,7].

In this paper, the modified SBAS approach in StaMPS is used for time-series InSAR analysis, due to its ability to ensure temporal continuity by connecting separated subsets of interferograms with larger baselines. Theoretically, a complex multilook operation to mitigate the effects of the decorrelation noise should be independently carried out before generating interferograms^[8].In this study, the spatial resolution of COSMO-SkyMed and TerraSAR-X are similar in size, so we could skip this step.The residual topographic artifacts, as well as the atmospheric phase screen (APS) signals, are also estimated and filtered out^[9-11].Based on an assumption that subsidence only happens in vertical direction, the estimated deformation in Line of Sight (LOS) is translated to vertical displacements.

Targeting at revealing the long-term ground subsidence, a fusion method based on nonlinear curve fitting is implemented using the overlapping time period between the TerraSAR-X and COSMO-SkyMed datasets from March 2016 to October 2016.It is revealed that the synergistic results of COSMO-SkyMed and TerraSAR-X datasets can obtain a more comprehensive understanding of the slow-moving subsidence.The subsidence results in this paper show a very good consistency with geological conditions and ground water funnel distribution in Shenyang City.Generally speaking, most parts of Shenyang are relatively stable. However, there’s a large area in Tiexi district showing serious subsidence. According to the geological data in Shenyang, the basal ground in this area is generally composed of sandy soil and fine sand. The permeability of the basal ground in this area is quite strong, and therefore instability and ground subsidence could possibly happen to this area. Ground subsidence is also detected in Tawan, Yushutai and Xiaonanjie area. They have presented a strong connection to the groundwater funnel in Shenyang.

IndexTerms—Time Series InSAR, Subsidence,SBAS, Fusion

5.REFERENCES

[1]. Pradhan, B.; Abokharima, M.H.; Jebur, M.N.; Tehrany, M.S. Land subsidence susceptibility mapping atKinta Valley (Malaysia) using the evidential belief function model in GIS. Nat. Hazards 2014, 73, 1019–1042.

[2].YusupujiangA , Fumio Y , Wen L . Multi-Sensor InSAR Analysis of Progressive Land Subsidence over the Coastal City of Urayasu, Japan[J]. Remote Sensing, 2018, 10(8):1304-.

[3]Ferretti, APermanent scatterers in SAR interferometry.IEEE Trans Geosci Remote Sens 39(2001).

[4] Berardino, Paolo , et al. A New Algorithm for Surface Deformation Monitoring Based on Small Baseline Differential SAR Interferograms. IEEE Transactions on Geoscience and Remote Sensing 40.11(2002):2375-2383.

[5] T., Strozzi , et al. Glacier motion estimation using SAR offset-tracking procedures.Geoscience& Remote Sensing IEEE Transactions on 40.11(2002):2384-2391.

[6] Huber, M.; Gruber, A.; Wendleder, A.; Wessel, B.; Roth, A.; Schmitt, A. The Global TanDEM-X DEM: Production Status and First Validation Results. In Proceedings of the 2012 XXII ISPRS Congress International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Melbourne, Australia, 25 August–1 September 2012; Volume XXXIX-B7, pp. 45–50.

[7] The TanDEM-X 90m Digital Elevation Model. Available online: https://geoservice.dlr.de/web/dataguide/tdm90/ (accessed on 31 October 2018).

[8] Rosen, P. A., et al. Synthetic aperture radar interferometry. Proceedingsofthe IEEE. 88.3(2002):333-382.

[9] Lanari, Riccardo , et al. An Overview of the Small Baseline Subset Algorithm: a DInSAR Technique for Surface Deformation Analysis. Pure and Applied Geophysics164.4(2007):637-661.

[10] Manzo, Mariarosaria , et al. A Quantitative Assessment of DInSAR Measurements of Interseismic Deformation: The Southern San Andreas Fault Case Study.Pure& Applied Geophysics 169.8(2012):1463-1482.

[11] Pepe, A., et al. The study of the deformation time evolution in coastal areas of Shanghai: A joint CX-band SBAS-DInSARanalysis.Geoscience& Remote Sensing Symposium 2015.

[12]Jing,L, The study of Physical and Mechanical Properties of Soil and Engineering Geological In Shenyang City Center.[D]. 2015.

The application of the traditional InSAR time series technology is often limited by the little measure points on the surface of the landslides, especially in the region with dense vegetation. In order to overcome its disadvantages corresponding to the surface characteristics of landslides, the DS-InSAR time series technology was presented and employed in monitoring of Jiaju landslide status. Compared with the SBAS-InSAR technology, the presented DS-InSAR time series approach could yield much more high dense measure points on the surface of landslide. The distributed location and the motion variation of landslide were apparently shown in the final deformation results. Therefore, the DS-InSAR time series approach would be valuable and has great potential in landslide hazard monitoring.