G004-0034
Comparison of InSAR time series generation techniques as part of the collaborative GeoSciFramework research project
Comparison of InSAR time series generation techniques as part of the collaborative GeoSciFramework research project
Wednesday, 9 December 2020
Poster
Abstract:
Improved intermediate-to-short term forecasts of catastrophic natural hazards can minimize damage and loss of life. The GeoSciFramework (GSF) project, funded by the NSF Office of Advanced Cyberinfrastructure and EarthCube programs, will allow researchers to detect precursory signals and reveal more suppressed, long-term motions of Earth's surface at unprecedented spatial and temporal scales. These goals will be accomplished by training machine learning algorithms to recognize patterns across various data signals during geophysical events and deliver scalable, real-time data processing proficiencies for time series generation. The algorithm will employ an advanced convolutional neural network method wherein spatio-temporal analyses are informed by physics-based models and continuous datasets, including Interferometric Synthetic Aperture Radar (InSAR), seismic, GNSS, and gas-emission data.
Here, we focus on the Differential InSAR (DInSAR) time series analysis component, which quantifies mm-to-cm level line-of-sight (LOS) ground deformation at ~25 meter spatial resolution. We first compare processing techniques that produce initial interferograms such as the Generic Mapping Tool SAR (GMT5SAR) and the InSAR Scientific Computing Environment (ISCE). Using those results, we compare time series generation programs like the Generic InSAR Analysis Toolbox (GIAnT) and the Miami InSAR Time Series Software in Python (MintPy). We also implement a new, innovative method [Zheng and Zebker, 2016] which removes the topographic phase component of the SAR signal so that simple cross multiplication returns an observation sequence of interferograms in geographic coordinates [Zebker, 2017]. Our results provide high resolution images of ground motions and measure LOS deformation over time. The ultimate goal of this project is to apply machine learning on the time series output, where pixels exhibiting anomalous movement are identified. Consistent, strong residuals in the signal, typically caused by satellite turbulence, can be removed using modeled components [Gabbes et al., 2019]. We showcase preliminary efforts to match modeled DInSAR deformation with our processed results. Finally, we present an integration of these data sets and demonstrate how they will be ingested and streamed through the GSF.
Here, we focus on the Differential InSAR (DInSAR) time series analysis component, which quantifies mm-to-cm level line-of-sight (LOS) ground deformation at ~25 meter spatial resolution. We first compare processing techniques that produce initial interferograms such as the Generic Mapping Tool SAR (GMT5SAR) and the InSAR Scientific Computing Environment (ISCE). Using those results, we compare time series generation programs like the Generic InSAR Analysis Toolbox (GIAnT) and the Miami InSAR Time Series Software in Python (MintPy). We also implement a new, innovative method [Zheng and Zebker, 2016] which removes the topographic phase component of the SAR signal so that simple cross multiplication returns an observation sequence of interferograms in geographic coordinates [Zebker, 2017]. Our results provide high resolution images of ground motions and measure LOS deformation over time. The ultimate goal of this project is to apply machine learning on the time series output, where pixels exhibiting anomalous movement are identified. Consistent, strong residuals in the signal, typically caused by satellite turbulence, can be removed using modeled components [Gabbes et al., 2019]. We showcase preliminary efforts to match modeled DInSAR deformation with our processed results. Finally, we present an integration of these data sets and demonstrate how they will be ingested and streamed through the GSF.