-
Paper 35 - Session title: Terrain subsidence and landslides II
09:00 A Combined Procedure Using Satellite-Based Differential SAR Interferometry And Field Measurements For Landslide Characterization In Small Urban Settings
Confuorto, Pierluigi (1); Angrisani, Anna Claudia (1); Di Martire, Diego (1); Infante, Donato (1); Novellino, Alessandro (2); Plank, Simon (3); Ramondini, Massimo (4); Calcaterra, Domenico (1) 1: Department of Earth, Environmental and Resources Sciences, Federico II University of Napoli (Italy); 2: Geomatic Ventures Ltd, Nottingham, (UK); 3: German Remote Sensing Data Center, Geo-Risks and Civil Security Department, German Aerospace Center (DLR) (Germany); 4: Department of Civil, Architectural and Environmental Engineering, Federico II University of Napoli (Italy)
Show abstract
Landslides represent globally one of the most noticeable and widespread “natural hazards”, responsible for significant economic damage as well as large quantities of fatalities. To this regard, non-structural actions, such as monitoring, assume a key role, in order to prevent and acquire knowledge about these critical phenomena. Differential SAR Interferometry (DInSAR) represents nowadays one of the main and most advanced techniques for satellite data-based landslide monitoring. In this paper, a solid procedure structured in two phases is proposed: the first, based on the exploitation of DInSAR data at municipal scale, and the latter, focused on a slope-scale analysis, integrating SAR and traditional in-situ data. The municipal scale analysis allows the detection and identification of critical landslide-affected areas, through a statistical Persistent Scattererer (PS) clustering process. Such methodology takes into account 4 requirements, used as input to generate cells that can represent evidence of instability : i) threshold of minimum velocity, to select PSs that can be considered as moving targets; ii) maximum inter-point distance between two close targets; iii) density of moving targets; iv) homogeneity index, consisting in the ratio between the total number of PSs within the cell and the number of “moving” PSs. According to the identification of the cells, higher priority sectors (i.e. characterized by higher hazard) can be selected in order to perform the slope-scale analysis. The latter is carried out by means of geomorphological and geological field survey, as to define landslide’s boundaries, geometries, and the potential damage of the area involved. Moreover, field activity is integrated with in-situ measurements, such as inclinometers and piezometers, with geotechnical data, deriving from boreholes and with DInSAR data, in order to comprehend landslide’s dynamic and triggering factors. The above-described procedure has been implemented on the test site of Cirò, a small town of ca. 3000 inhabitants, located in the Crotone province (Southern Italy). For the Cirò case, the municipal scale analysis has been applied using a dataset of 35 TerraSAR-X images, acquired in descending orbit in the time span 2008-2010 and processed through the Coherence Pixels Technique (CPT). Among all the landslides detected in Cirò, an event occurred in February 1, 2011, involving one of the main access roads to the town center and causing severe damage, forcing people to abandon five houses, has been classified of priority importance. Hence, for the slope-scale analysis, PSs deriving from the third edition of the Extraordinary Environmental Remote Sensing Plan (PST-A in Italian) have been exploited. Such data have been obtained through the processing of COSMO-SkyMed data acquired between 2011 and 2014 by means of the Persistent Scatterer Pairs (PSP) technique. To integrate the DInSAR information, field surveys have been carried out among 2013 and 2015 and in-situ data have been collected between 2011 and 2014. The results so far obtained through the proposed procedure showed promising outcomes, which can in future be applied and tested on further landslide affected areas, in order to provide a valuable tool for local administrations and to prevent damage and economic losses.
Presentation
[Authors] [ Overview programme] [ Keywords]
-
Paper 86 - Session title: Terrain subsidence and landslides II
10:20 Monitoring the change of soil seismic response through the InSAR-derived ground subsidence: application to the Mexico City subsidence
Albano, Matteo (1); Polcari, Marco (1); Bignami, Christian (1); Moro, Marco (1); Saroli, Michele (2,1); Stramondo, Salvatore (1) 1: Istituto Nazionale di Geofisica e Vulcanologia (INGV), Italy; 2: Dipartimento di Ingegneria Civile e Meccanica, Università degli Studi di Cassino e del Lazio meridionale
Show abstract
Subsidence phenomena have been widely observed through remote sensing SAR Interferometry (InSAR) data. The large area coverage and the satellite viewing geometry make the InSAR a reliable tool in constraining the surface displacements induced by natural or anthropic subsidence. The InSAR technique is particularly suitable in detecting ground movements in urban environments because these areas are generally less affected by temporal decorrelation problems and provide dense and coherent scatterers for estimating the ongoing deformation. The ground subsidence rate provided by satellite measurements has been successfully integrated with geological/geotechnical information to generate subsidence hazard zoning maps and for ground rupture evaluation. The information content provided by the well-known Persistent Scatterers Interferometry (PSI) technique has been exploited to assess the health of engineering structures at urban and local scale. Moreover, the InSAR interseismic velocities have been correlated with the thickness and the resonant period of quaternary soft lithologies, thus allowing to estimate the change of the soil resonant period through a systematic analysis of InSAR time series. This correlation could allow to monitor the long-term subsidence and the consequent change in soil dynamic properties for wide areas and with relatively low costs, however, the InSAR-derived ground subsidence has never been used as a tool for seismic hazard assessment.
In this work, we developed an empirical procedure to evaluate the effect of the ground subsidence on the spatial and temporal seismic response of soils. The proposed method exploits the capabilities of the spaceborne SAR Interferometry technique to detect and map the ground subsidence with unprecedented spatial and temporal coverage. The information provided by satellites is combined with a-priori geological/geotechnical information to assess the soil compaction and the shortening of the soil vibration periods.
The developed procedure was applied to estimate the shortening of the soil resonant period of Mexico City. Mexico City lies on a basin of volcanic origin, filled with Plio-Quaternary lacustrine soils deposited when the whole territory was occupied by a series of interconnected lakes. The poor geotechnical properties and the extremely high water content and compressibility make the lacustrine deposits of zone susceptible to compaction if loaded. In fact, the intensive exploitation of the groundwater in the deep granular soils has caused the consolidation of the clayey deposits and an exceptional ground subsidence that reached approximately 8 – 10 m between 1891and 1995. Today the subsidence is still present, with values ranging between 5 and 35 cm/year. The high subsidence rate, the wide spatial extent and the extreme subsidence gradients represent a great hazard for buildings and infrastructures of Mexico City.
We exploited the InSAR-derived ground subsidence in the period between 2005 and2013 to estimate the change in soil resonant period. SAR data coming from ENVISAT and RADARSAT-2 missions were processed on-line with Geohazard Exploitation Platform (GEP) and locally with the multi-baseline IPTA approach, respectively. The obtained ground subsidence has been combined with a-priori information of soil thickness and resonant period to calculate the soil resonant period change in the observed timespan.
The results show that in approximately nine years the ground surface has subsided by approximately 0.5–3.5 m and the soil resonant period has decreased by approximately 0.1–0.4 s. The obtained results, validated with field measurements, highlight the effectiveness of the proposed procedure for the continuous monitoring of the soil resonant periods. The estimated change in resonant period on Mexico City has a great impact on the response spectra used for design, it is then necessary to update the map of the soil resonant period in order to account for the change of dynamic properties of soils caused by subsidence.
The proposed procedure is suitable for the on-line implementation thus allowing a continuous upgrade of the soil resonant period map and the direct usability of updated data by stakeholders for the design of new structures and the rehabilitation of existing ones.
Presentation
[Authors] [ Overview programme] [ Keywords]
-
Paper 226 - Session title: Terrain subsidence and landslides II
10:00 Landslide Detection Based on DEM Matching
Rui, Jie (1,2); Wang, Chao (1); Zhang, Hong (1); Jin, Fei (2); Zhang, Zhanmu (2); Liu, Zhi (2); Wang, Fan (2); Tang, Yixian (1) 1: Key Laboratory of Digital Earth, Institute of Remote Sensing and Digital Erath,CAS, China, People's Republic of; 2: Department of Remote Sensing Information Engineering, Zhengzhou Institute of Surveying and Mapping Zhengzhou, China
Show abstract
In this paper, a landslide detection method based on two-step robust DEM-matching algorithm in variable mountainous areas was presented. DEM matching presents two challenges: determining the alignment between the surface features of the two DEMs and estimating the matching transformation parameters. Firstly, we utilized a shape context descriptor to compare contour lines and detect invariant terrain peaks as control points based on contour line shapes. Secondly, a least squares surface-matching method was used for optimization. Finally, landslides were detected in the elevation change map. The experimental result indicates that large-scale landslides can be detected based on robust DEM matching.
Presentation
[Authors] [ Overview programme] [ Keywords]
-
Paper 312 - Session title: Terrain subsidence and landslides II
09:20 Optimizing the deformation signal caused by slow moving landslides in the Northern Apennines of Italy
Bayer, Benedikt (1); Schmidt, David (2); Simoni, Alessandro (1); Mulas, Marco (3); Bertello, Lara (1); Corsini, Alessandro (3); Bonacini, Francesco (3) 1: University of Bologna, Italy; 2: University of Washington, Seattle; 3: University of Modena, Italy
Show abstract
Pelitic turibidites and mélange type rocks are common lithologies in the Northern Apennines of Italy. These formations have a high landslide susceptibility and host different landslide types that often undergo steady or seasonal deformation. The study area is located South of Bologna and Modena and is characterized by a Mediterranean climate with two wet seasons in autumn and spring respectively. Due to the presence of vegetation and agricultural land use, interferograms are often decorrelated, which is a particular problem for X-band satellites (like COSMO SkyMed) or satellites with low acquisition frequency (like Envisat). With the launch of Sentinel 1, data from the first C-band satellite with a high acquisition frequency became accessible at no cost for scientific applications. In this work we present interferometric results for the period between 2014 and 2016 from Sentinel over two river catchments in the Northern Apennines. We used GMTSAR (Sandwell, 2011) to process interferograms from four Sentinel swaths. The Small Baseline module of the Stanford method of Persistent Scatterers (Hooper et al., 2007) was used for post processing. We show how processing parameters influence the final interferometric results both on the regional scale and for single landslides. To that end we chose a small subset of our study area and established a framework to systematically test the influence of parameters that govern pixel selection and unwrapping. One focus area is the Camugnano landslide, a complex rockslide that is composed of several nested landslide bodies. The landslide suffered from repeated reactivations, resulting in the installation of a monitoring system in 2014. GPS and inclinometer data are available, and hence it is possible to compare the InSAR derived displacements to independent measurements. It also gave us the possibility to evaluate the range of InSAR processing parameters that yielded reasonable results.
In addition, we present a simple strategy to optimize the backprojection of the InSAR results on the downslope vector, given different viewing geometries and illustrate how gaussian low pass filters and smoothing operations can be used to improve the displacement signal in space and time. The approach consists of three main steps: i) slope and aspect maps are derived from an external digital elevation model and for each pixel this slope and aspect information is used to derive the downslope unit vector. Then pixels that are common to all datasets are selected. ii) Assuming that the real displacement vector may differ up to 15 degrees in azimuth and 5 degrees in slope, we start with a simple back-projection of the LOS range-change in the downslope direction. For each viewing geometry we invert the interferograms seperately in order to solve for mean velocities. Then we modify slope and aspect values until the differences between the mean velocities from the separate viewing geometries are minimized in order to obtain a new unit vector of minimal differences. iii) The interferograms from the different satellite tracks are projected on this unit vector of minimal differences. In each dataset a date at the beginning of the time series is chosen and modified in order to form a closed network. The modified dates are chosen to be as close as possible in time (typically < 8 days). Then the combined network that includes data from different viewing geometries with different look angles is inverted to solve for a smooth displacement time series. We discuss the advantages and drawbacks of this strategy by comparing the three dimensional displacements to the line of sight measurements for a single landslide case. We also evaluate the accuracy of the technique by comparing them to independent monitoring data from GPS and inclinometers.
Presentation
[Authors] [ Overview programme] [ Keywords]
-
Paper 347 - Session title: Terrain subsidence and landslides II
09:40 Monitoring Fast Motion of Guobu Slope near Laxiwa Hydropower Station by Point-like Targets SBAS Offset Tracking
Li, Menghua (1); Zhang, Lu (1); Liao, Mingsheng (1); Shi, Xuguo (1,2) 1: State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China; 2: Department of Information Engineering, China University of Geosciences, Wuhan, China
Show abstract
Laxiwa Hydropower Station located in the upstream Yellow River is the largest hydropower station in Qinghai Province of China. It is constructed since 2001 and the impoundment started in 2009. The slope located 500 m upstream of the dam on the right bank of the reservoir was found deforming greatly and continuously since the impoundment. The deforming slope is about 700 m high and 1000 m wide. Collapse of such a huge slope would damage the dam and other facilities, greatly threatening the safety of people and infrastructures downstream. Therefore it is important to monitor and analyze the mechanisms of the unstable slope for early warning and prevention purpose.
Spaceborne SAR observation provides a promising geodetic tool that can be supplementary to traditional measurements, especially in vast inaccessible mountainous area. Methods for deriving deformation from SAR images can be classified into two categories: phase based and amplitude based methods. Due to the advantage in precision, countless studies focused on phased based time series methods such as PSInSAR, SBAS and DSInSAR. But they are limited by the deformation gradient and only suitable for studying slow-moving landslide. The amplitude based methods such as pixel offset tracking and point-like targets offset tracking (PTOT) are more suitable for large displacement measurement with precision of more than 1/10 pixels if high correlations were guaranteed.
In most of the applications of amplitude based method, a single master image is adopted to derive time series deformation. Localized offset between the reference and slave images induced by topographic variations is usually compensated by a simple linear relationship with additional topography data. However, the existing Digital Elevation Models (DEMs, e.g. SRTM and ASTER) exhibit height differences as big as 50 m in our study area. For the TerraSAR-X data, the latitude of the satellite above Earth is 514 km. A 50 m difference in topography can induce about 0.04 m range offset for an image pair with a perpendicular of 200 m and a look angle of 24°. This offset is about 1/10 of the range pixel size for TerraSAR-X 300MHz High-resolution Spotlight (HS) images. Therefore, range offset due to topographic relief cannot be fully eliminated by computing the direction function with an external DEM and the SAR imaging geometry.
We follow the same strategy of the phased based Small BAseline Subset (SBAS) approach based on PTOT. The point-like target (PT) offset tracking method focuses on the stable point-like targets. It is more efficient and reliable when compared with the original pixel offset tracking technique. Point targets are first selected from the mean amplitude image by correlation with a 2-D sinc function template. Then we calculate the relative offset between image pairs formed with small separation in time and space. Range offset due to topographic will be firstly compensated with an external DEM before the final inversion of time series offset. Correlation of PTs in each pair will be used as a tool to weed out the unreliable result. Finally, we apply the SBAS inversion strategy to retrieve the range and azimuth displacement and the offset induced by inaccurate DEM.
In our study, two stacks of TerraSAR-X 300MHz High-resolution Spotlight (HS) descending images were collected in different look angles for displacement extraction. Each stack has 10 images from December 2015 to October 2016. Phase based SBAS method failed due to the large deformation gradient. Preliminary results shows displacements of more than 1.2 m in range direction happened during 9 months. Point-like targets SBAS offset tracking method will be applied to retrieve the displacement histories of Guobu Slope. Detailed analysis will be given after we collected ground measurement as well as water level and rainfall data.
Presentation
[Authors] [ Overview programme] [ Keywords]