Conference Agenda

Remote sensing (RS) data is successfully applied since decades for the identification and monitoring of landslide phenomena at different spatio-temporal scales. However, limitations associated with data availability/accessibility (e.g. spatial coverage, low temporal revisit time, high costs) might hampered the development of operational tools.

The results and analyses retrieved in the framework of D4 project 32365 have shown the great benefits of RS in monitoring multi-hazards. The wide spatial and temporal data availability allowed a detailed description of landslide histories even of remote regions. Therefore, the continuous monitoring of such hazards, namely large landslides, is of fundamental importance to minimize and prevent the actual and future risks. In this Dragon-5 proposal, we foresee to continue the monitoring activities started with the Dragon-4 project mainly by means of multi-source RS data at diverse areas located in different countries.

In this first year we applied the InSAR Stacking technique to process and analyze the sentinel-1 data covering the Gilgit research area from October 2019 to October 2020. The research results show that the absolute value of the deformation rate in most areas is less than 10 mm/yr, which is relatively stable. The maximum sedimentation rate of each image frame is 347.6mm/yr, 525.3mm/yr, 455.4mm/yr, 284.5mm/yr, respectively. The deformation results were graded and colored, and displayed in a three-dimensional scene. Several highly suspected landslides located near human settlements in the area were identified. To compensate for the geometric distortion caused by a single imaging geometry a combination of different orbit data was use to effectively avoid "monitoring loopholes". Therefore, the research data adopted the method of combining ascending and descending orbits allowing for comprehensive and accurate early identification of landslide hazards. This work is of great significance for understanding the geological deformation of Gilgit area, especially the identification of some slopes with obvious slip phenomenon, which has great reference value for the follow-up disaster investigation work.

Multi-temporal landslide detection through optical imagery time-series analysis is a second goal of this project. Building upon the results of our Dragon 4 project, we investigate further in automated landslide detection approaches using high-resolution optical imagery (i.e. Senintel-2). Time-series analysis has shown to be an efficient technique for identifying major landslide events both spatially and temporally. Multi-temporal change detection demonstrated to minimize false-positives e.g. through artefacts or agricultural activity that result in bi-temporal change-detection approaches. The next step of our work will consist on the investigation of landslide predisposing conditions through the recognition of preceding land-cover changes and on the explotation of triggering causes by linking landslide events with rainfall data or seismic activity records.

Finally, in the scope of this project we also intend to explore the availability of SAR data with spatial and temporal resolutions at an unprecedented level, associated with the new methods of SAR time series processing to develop an active system for structural risks detecting and alerting. However, only the use of Artificial Intelligence (AI) techniques will allow to deal with the huge amount of data that will be generated. The Vilariça Valley, located in the north of Portugal is crossed by an active fault and will be used as test site to develop the AI-based risk alert system. In this region there is a high number of buildings with historical and patrimonial interests that may be at risk. In this first year, we download all Sentinel-1 data available and start the small and large area processing. In parallel we are also designing the platform to be developed, integrating different data sources and AI techniques.

The seismic deformation monitoring efforts using InSAR in the past 16 years gain fruitful achievements under the Dragon 1-4 cooperation projects. The seismic-related works using InSAR method include interseismic deformation monitoring along big faults, regional-scale deformation detection, major earthquake deformation measurements and postseismic deformation analysis for rheology studies. In recent years, induced seismicity monitoring is also another important task to do for mines or shale gas production.

In Dragon 5, we continue our Dragon 1-4 works on seismic deformation monitoring, in conjunction with detecting abnormal changes of electromagnetic field in the lithosphere. However, new challenges appear on SAR data analysis itself and integration with electromagnetic field to interpret the mechanism of causing seismic deformation. In the past 5 years, Sentinel-1 satellites acquired high-quality data and are still accumulating with fast rate and require high capability for InSAR data processing. To overcome the issues, we developed parallel computation systems for this purpose, which also has a great storage system attached to it. Moreover, with the big forward on artificial intelligence (AI) and machine learning algorithms developed in recent years, we hope to integrate them into data processing system to improve deformation detection precision and data analysis process in aggregation with electromagnetic data. Another piece of work is to deal with the atmospheric delays on InSAR time-series analysis because the current methods all have various kinds of difficulties in the analysis, and prevent further improvements on precisions. The project proposes to use machine learning methods to construct models that could be used to accurately make predictions or simulations of atmospheric delays, as shown by some of the recent tries.

The tectonic environment of China and surrounding regions depend mostly on the collision of Indo and Eurasia plates. The Dragon 5 project will still focus on faults, such as the Haiyuan, Kunlun, Altyn Tagh, Xianshuihe, Tianshan fault systems etc. In addition, we will also integrate InSAR and GPS data to develop an inversion model for regional strain distribution in particular regions such as Tibet, North China Plain, to prepare for seismic hazard mitigation, and assess the risk for national key projects, such as the Sichuan-Tibet railway construction project. Furthermore, the recent hot topic on induced seismicity is the new field for InSAR working with other traditional approaches, in particular for the Sichuan basin, so we will also address this new topic in our Dragon 5 project.

The project aims to develop and apply innovative data analytic methods underpinned with machine (deep) learning technology to analyze and detect seismic anomalies from electromagnetic data observed by the SWARM and CSES satellites along with CSELF network.

Both the ground-based and satellite-based observations have shown that a range of frequency band of electromagnetic signals had been recorded in ionosphere around strong earthquakes. These phenomena conjectured that when earthquakes occurring electromagnetic waves generated could penetrate from lithosphere around the epicentre areas of earthquakes into ionosphere, which could be supported by the simulated results of penetrating process of electromagnetic wave from ground to ionosphere.

Since the DEMETER satellite launched by the French CNES on 29 June 2004, a large number of papers have been published in respect of ULF/ELF/VLF/LF electromagnetic perturbations in topside ionosphere, and they were inferred to be possibly related to earthquakes. On 22 November 2013, the SWARM satellite constellationbegan to operate, mainly focusing on observing geomagnetic field in ULF band. Some interesting phenomena around earthquakes have been reported using the SWARM satellite data. The first Chinese Seismo-Electromagnetic Satellite (CSES) was launched on 2^nd February 2018, some perturbations related to earthquakes were also detected in electromagnetic field. The DEMETER operation ended in 2010, but SWARM and CSES both are still in orbit at present. SWARM has delivered data for more than 7 years, and CSES for longer than 3 years, thus the measurements obtained by these satellites provide an unprecedented opportunity for conducting stereo investigations into earthquakes at different altitudes and local times, especially on the electromagnetic waves at different frequency bands.

In China, a new CSELF network was constructed with more than 30 stations since 2012, dedicated to record electromagnetic waves below kHz. All these ground electromagnetic observations can be compared with the satellite data at the respective frequency bands, and the same frequency signals could be distinguished and traced from the ground to the ionosphere. By accounting for the multi-frequency bands and a large amount of data, machine learning-based approaches will help scan and extract all the disturbed signals and construct prediction models by incorporating the relation between electromagnetic signals and earthquakes.

In Dragon 3 and 4, five algorithms for anomaly detection have been developed, including Wavelet Maxima (WM), Geometric Moving Average Martingale (GMAM) based on the Martingale theory, the integration of Cumulative Sum (CUSUM) and Exponentially Weighted Moving Average (EWMA) called CUSUM-EWMA, a Fuzzy Inspired Approach to Seismic Anomaly Detection (FIAD), and Enhanced Martingale. These algorithms have been used to analyze NOAA, SWARM and CSELF data for distinguishing anomalies in relation to the Wenchuan, Lushan, Puer, Jinggu, Taoyuan, Ludian and Peloponnese earthquakes occurred in China and Greece, and the preliminary results have been achieved.

In this report we will present a brief summary of the results obtained from the Dragon 3 & 4 projects, and then introduce the aims and objectives of this Dragon 5 project, particularly emphasizing on the challenges head when addressing possible approaches for verifying the relationship between electromagnetic disturbances with earthquakes, and developing pragmatic and sophisticated anomaly detection algorithms underpinned with Deep Neural Networks, we will report the preliminary results achieved to date.

Precipitable water vapor (PWV) from numerical weather models, such as the latest generation of European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA5) and the ECMWF High RESolution (HRES) models, are important to meteorological studies and to error mitigation of geodetic observations such as Interferometric Synthetic Aperture Radar. In this study, we provide global validations of these new weather models with respect to Global Positioning System (GPS, ∼13,000 stations) and Moderate Resolution Imaging Spectrometer (MODIS, ∼1 km resolution) using data from January 2016 to December 2018 of every 1 h. The global standard deviations of the Zenith Tropospheric Delay (ZTD) differences (DSTDs) between weather models and GPS are 1.69 cm for ERA5 and 1.54 cm for HRES. The global PWV DSTDs between weather models and MODIS are 0.34 cm for ERA5 and 0.32 cm for HRES. The two weather models generally perform better in western North America, Europe, and Arctic by having low ZTD DSTDs (<1.3 cm) or PWV DSTDs (<0.3 cm). HRES also has a low ZTD DSTD of less than 1.3 cm in Antarctic, Japan, New Zealand, and Africa and outperforms ERA5 in most regions of the world, despite the fact that 83% of the HRES PWV values are temporally interpolated (from 6 to 1-h). However, under extreme weather conditions, ERA5 performs better owing to its high temporal resolution (1 h). We have developed a new generation of the Generic Atmospheric Correction Online Service for InSAR (GACOS) which can utilize ERA5, HRES and GNSS products to generate high resolution tropospheric delay maps for InSAR atmospheric correction. In this study, we also demonstrate some successful applications of the GACOS to a variety of geophysical studies.

References:

Yu, C., Z. Li, and G. Blewitt (2021), Global Comparisons of ERA5 and the Operational HRES Tropospheric Delay and Water Vapor Products With GPS and MODIS, Earth and Space Science, 8(5), e2020EA001417, https://doi.org/10.1029/2020EA001417.

Yu, C., Z. Li, L. Bai, J.-P. Muller, J. Zhang, and Q. Zeng (2021), Successful Applications of Generic Atmospheric Correction Online Service for InSAR (GACOS) to the Reduction of Atmospheric Effects on InSAR Observations, Journal of Geodesy and Geoinformation Science, 4(1), 109-115.

In the framework of Dragon 5 project, Northeastern University (NEU) from China, the National Institute of Geophysics and Volcanology (INGV) from Italy and Changbaishan Volcano Observatory of China have been collaborating to analyze the multiple geohazards over the heavy industrial base and Changbaishan Volcano in Northeast China using multi-source remote sensing data provided by ESA and third party missions.
The heavy industrial base in Northeast China has been playing an important role in the economic development of China. However, the hard mining activities have a strong impact on local environment due to continuous ground excavations for coal and iron ore. Therefore, mining areas in Northeast China are subject to a multi-hazard exposure including subsidence, landslides, ground fissures and building inclination. In order to keep safety of local environment and mining activities, continuous monitoring of the spatial and temporal variations of multiple geohazards are carried out for early-warning and risk assessment by the research team.
The Changbaishan active volcano region (Jilin Province, about 300 km east from Shenyang city) is also affected by landslides, earthquakes and ground deformation related to volcanic and hydrothermal processes. The last deformation related to such phenomena occurred during the 2002-2006 unrest episode, in 2011 and in 2017, when a nuclear test explosion in North Korea triggered landslides on the steep slopes surrounding the caldera lake. The frequency of earthquake swarms has also increased since December 2020. The multi-hazard exposure of Changbaishan is also high, because a population of about 135000 in China and 31000 in North Korea lives within 50 km distance from the volcano. In addition 2000000 tourists visit the Changbaishan Volcano National Reserve, a part of the UNESCO Man and Biosphere program each year. Therefore, surveillance of the dynamics of Changbaishan Volcano is also crucial for disaster prevention and risk mitigation.
In the framework of Dragon 4 and Dragon 5, the research team has collected more than 300 COSMO-SkyMed images, hundreds of Sentinel-1 images, 30 TerraSAR-X images and 19 ALOS-2 PalSAR images over the study areas. We focused investigation over some of the most important sites and towns located in Northeastern China, such as Shenyang city, Anshan and Fushun open pit mines, and Changbaishan volcano as well. Time Series InSAR analysis is carried out over the study areas with assistance of TanDEM DEM and precise LiDAR DEM using both the Persistent Scatterers and the Small Baseline Subsets technique. The results from multiple stacks, covering different temporal interval and operating in different frequency bands, have shown a very good consistency over each study area. The results are compared with terrestrial measurements, precipitation data and geological data for the purpose of accuracy assessment as well. Our research results have demonstrated that displacement time series retrieved through the advanced InSAR technologies can be used as a routine tool for deformation monitoring of multiple hazards and provide nowadays a fundamental technical support for disaster prevention and mitigation.

Archaeological prospection and protection of cultural and natural heritage sites are important applications of remote sensing. The key goal of the Dragon-5 project SARchaeology is to exploit satellite SAR imagery and multi-sensor approaches to detect objects of archaeological significance, and monitor the status and stability of cultural and natural heritage sites and their assets.

The project focuses on arid areas, e.g. paleo-channels around Lop-Nor in China, kurgans (Iron Age burial mounds) in the Altai mountains in China and Tuva region in Russia, and partly buried archaeological ruins in the larger province of Rome in Italy, as well as natural heritage of the Jiuzhaigou site in China.

Image analysis methods that are used in the project include – but are not limited to – feature extraction, image classification, change detection, and multi-temporal Interferometric SAR (InSAR). The latter is essential to address the goal of site protection, via estimation and monitoring of surface deformation due to geological processes (e.g. subsidence, landslides), which can endanger natural and cultural heritage sites.

SAR datasets exploited for archaeological prospection include ALOS-1 L-band, as well as shorter wavelengths, namely ERS-1/2, ENVISAT, RADARSAT-1/2 and Sentinel-1 C-band, and TerraSAR-X, potentially Iceye and Paz X-band data, in order to test signal penetration capabilities at the different wavelengths and spatial resolutions. Thanks to the upcoming wider availability of long-wavelength data from various L-band missions and BIOMASS P-band mission, sub-surface target detection is also becoming possible, thus opening new perspectives for the use of SAR for archaeological prospection and identification of hidden paleo-channels and linear structures. Long-term surface motion monitoring and site surveillance are guaranteed with Sentinel-1 SAR data and their abundant stacks acquired over the study sites since 2014. Optical imagery from Deimos-2, WorldView, GeoEye, QuickBird, CBERS-4 and Jilin-1 will be used to provide very high resolution basemap layers to aid the SAR image interpretation and identify the main archaeological features. SPOT, Pleiades and RapidEye imagery from ESA collections will also be used.

Where possible in-situ measurements will be collected at the times of SAR satellites passes.

During the first year, the project has been setup and preliminary work has started. Collaboration between the European and Chinese teams in the framework of other projects, such as the long-term monitoring of surface deformation in Wuhan and research on kurgans, has formed the basis for kicking-off a much stronger partnership in Dragon 5.

The focus of the initial project activities has been on study site selection based on existing literature review, consolidation of scientific objectives for each heritage site, EO datasets identification and access.

For the research on burial mounds, the work has also been focused on improving the methodologies and better monitoring the sites with respect to climatological factors. This is important as the most valuable burial mounds are to be found in or close to permafrost areas. Global warming and thawing of permafrost endangers the organic remains in some of the sites in question that are currently still frozen and therefore extremely valuable for archaeological analysis. Learning more about the current extent of permafrost, monitoring spatial changes and hopefully being able to predict the spatio-temporal patterns of future changes will be highly important for the planning and prioritization of archaeological excavations.

Regarding dissemination and teaching activities, Prof. Fu organized the International Workshop on Space Technologies for Disaster Mitigation of World Heritage on 13-16 October 2020, in Jiuzhaigou, China, and Prof. Balz gave lectures on SAR remote sensing.

Field data collection and ground truthing for the main sites in Central Asia has been postponed due to the pandemic situation, though site observations from previous cooperation in the field of the detection and mapping of kurgans and other burial mounds in Central Asia will be exploited to address the current difficulties, with a view to future dedicated field work in Russia and Mongolia when the situation will allow.

The second year of the project will be focused on intensive EO data processing and initial analysis and interpretation, for both archaeological prospection and heritage protection. Moreover, it is planned to conduct field work in the area of Rome, and to resume the work in the Altai. Furthermore, a plan to extend the use of data from different sources, especially the combination of European and Chinese remote sensing data sources, is also in place.

Regarding the level of training and involvement of young scientists in the project, a PhD student from the Chinese team is currently preparing an application to submit to the Chinese Scholarship Council to support a research visit in Rome to work with the European team on long-term monitoring of heritage sites with multi-temporal InSAR and ground truthing, tentatively in 2021-2022 for a period of 1.5 years. On the other hand, the European team is assessing opportunities to identify and recruit MSc students and graduates to work on the project starting from 2022.