Conference Agenda

Sentinel-1 (S1) data have the potential to measure, by radar interferometry (InSAR), the present-day Earth surface deformation, whether of tectonic origin or otherwise (anthropogenic, hydrological), on the scale of specific local targets (active faults, sedimentary basins, cities) as well as on the continental scale (large lithospheric blocks bordered by mountain ranges and major fault systems). With now more than 4 years of S1 images archive, the global coverage and high-temporal resolution of these images thus allow to investigate the dynamics of slow aseismic slip on faults, a critical step to better understand physical processes involved in the generation of large earthquakes, in various tectonic contexts worldwide. They also allow a refined quantification of strain partitioning across complex fault systems, as well as of the degree of strain localization on faults, which can be confronted to different fault system evolution scenarios and lithospheric deformation mechanisms. In cases where non-tectonic deformation superimpose with tectonic deformation, time-series analysis helps extracting the specific spatio-temporal signature of each phenomena. We will illustrate these different applications with our most recent case studies, in Asia in particular, based on InSAR time-series analysis of S1 data.

With the medium-resolution (~2.0 meter in azimuth and ~13.0 meter in range for TOPS/IW mode) SAR data, it is possible to acquire large scale deformation (>1000 km) in a continuous TOPS scanning. With the temporal sampling of 6-day or times of it, Sentinel-1 SAR data were quickly accumulated since later of 2014. However, processing of the large data set is a challenge, which is useful and/or a requirement for some typical applications, such as tectonic deformation analysis or anthropogenic activity monitoring for a vast region.

We utilize high-performance computation (HPC) for this purpose, which is widely used for scientific applications. To accelerate processing, we adopt Gamma processor for conventional processing with a benefit of multiple-core parallel processing on each node, and it dramatically reduces the time cost for TOPS mode SAR data alignments. On HPC with multiple nodes, the data alignment and interferometric processing procedures were deployed on each node, without communications between nodes required. After preprocessing with Gamma, multiple doppler-deramped and coregistered images are prepared for time-series analysis. In this stage, we adopt the sophisticated processor StaMPS (Hooper et al., 2007) for PS and SBAS analysis, or combine the two methods for hybrid analysis. Due to patch-level parallelization, the large-scale data could be divided into multiple patches with different dimensions and they are processed on each node simultaneously.

We applied our two-level processing approach in multiple challenge areas, the North China Plain, the Longmenshan area and other Tibet regions for both anthropogenic activity monitoring and tectonic deformation detections. Both areas are quite tricky for normal InSAR processing, but with our HPC parallel system, we acquire consistent results compared with GPS observations. The method conducted in these tests confirmed the robustness of our approach for deformation detection with Sentinel-1 large scale InSAR data.

The North China Plain (NCP) is a vital agricultural region and is highly-populated, so the groundwater utilization is quite heavy in this region for irrigation and human beings. This leads to an overdraw of groundwater and fast subsidence over the whole area.

We utilize high-performance computation (HPC) for detection of the related deformation. To accelerate processing, we adopt Gamma processor for conventional processing with a benefit of multiple-core parallel processing on each node, and it dramatically reduces the time cost for TOPS mode SAR data alignments. On HPC with multiple nodes, the data alignment and interferometric processing procedures were deployed on each node, without communications between nodes required. After preprocessing with Gamma, multiple doppler-deramped and coregistered images are prepared for time-series analysis. In this stage, we adopt the sophisticated processor StaMPS (Hooper et al., 2007) for PS and SBAS analysis, or combine the two methods for hybrid analysis. Due to patch-level parallelization, the large-scale data could be divided into multiple patches with different dimensions and they are processed on each node simultaneously.

Our processing shows that in the center of the NCP, farming activity produces widely distributed deformation, not only localized subsidence bowls as in typical subsidence regions. Our results are also consistent with GPS observations on this scale. Tectonic units in the same region could bound the subsidence behavior, hence both kinds of activities may have some kind of interactions.

The global and systematic coverage of Sentinel-1 radar images enables to characterize, by radar interferometry, surface deformations at the scale of large active faults. This represents considerable progress in fault monitoring and opens new perspectives in seismic hazard assessment. Our study focuses on the Yushu - Ganzi - Xianshuihe active fault system (YGX), located on the eastern part of the Tibetan plateau. This left-lateral fault system accommodates the collision between the Indian and the Eurasian plates. The Ganzi segment may represent a 350 km-long seismic gap, unbroken for the past ~120 years. To measure the interseismic deformation across the YGX fault system, we perform a time series analysis of 4 years of Sentinel-1 InSAR data, acquired along ascending and descending orbits, using the New Small Baseline Subset processing chain including the latest adaptations (Doin et al., 2011, Grandin, 2015). The results are presented as mean velocity maps across the faults and compared to previous GPS studies and the long-term fault history. Simple elastic models of velocity profiles are also derived. They show that the Ganzi gap may be the site of aseismic slow slip which, depending on its spatio-temporal characteristics, could contribute to reduce seismic hazard on the fault or, conversely, facilitate the initiation of future major ruptures. The characterization of strain partitioning and strain localization across this fault system enables to precisely evaluate spatial and temporal variations of slip at various depths on the fault and constitutes a key constraint on seismic hazard assessment and lithospheric deformation mechanisms.