Subsidence feature discrimination using deep convolutional neural networks in synthetic aperture radar imagery

Schwegmann, C. P.; Kleynhans, W.; Engelbrecht, Jeanine; Mdakane, L.W.; Meyer, R. G. V.

doi:10.1109/igarss.2017.8128031

Cited by 24 publications

(14 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Once each new SAR image is acquired, the time needed to evaluate the subsidence signal and verify whether it is indeed precursory, is short (1-2 days). Our next step for the near future is to apply machine learning techniques to the interferometric analysis (e.g., [49]) in order to provide hazard information to stakeholders in an even shorter time.…”

Section: Towards An Automatic Early Warning Systemmentioning

confidence: 99%

SAR Interferometry for Sinkhole Early Warning and Susceptibility Assessment along the Dead Sea, Israel

et al. 2019

View full text Add to dashboard Cite

During the past three decades, the Dead Sea (DS) water level has dropped at an average rate of ~1 m/year, resulting in the formation of thousands of sinkholes along its coastline that severely affect the economy and infrastructure of the region. The sinkholes are associated with gradual land subsidence, preceding their collapse by periods ranging from a few days to about five years. We present the results of over six years of systematic high temporal and spatial resolution interferometric synthetic aperture radar (InSAR) observations, incorporated with and refined by detailed Light Detection and Ranging (LiDAR) measurements. The combined data enable the utilization of interferometric pairs with a wide range of spatial baselines to detect minute precursory subsidence before the catastrophic collapse of the sinkholes and to map zones susceptible to future sinkhole formation. We present here four case studies that illustrate the timelines and effectiveness of our methodology as well as its limitations and complementary methodologies used for sinkhole monitoring and hazard assessment. Today, InSAR-derived subsidence maps have become fundamental for sinkhole early warning and mitigation along the DS coast in Israel and are incorporated in all sinkhole potential maps which are mandatory for the planning and licensing of new infrastructure.

show abstract

Section: Towards An Automatic Early Warning Systemmentioning

confidence: 99%

SAR Interferometry for Sinkhole Early Warning and Susceptibility Assessment along the Dead Sea, Israel

et al. 2019

View full text Add to dashboard Cite

show abstract

“…It is furthermore recognized that classifying sinkhole features (on imagery or elevation models) is typically done manually, and is an intensive process with accuracy and reproducibility limitations [63,65,71,77]. There is therefore a need to investigate automated extraction techniques, with one of the promising growing fields of research being artificial neural networks [78].…”

Section: Discussion and Perspectives-towards Operational Sinkhole Earmentioning

confidence: 99%

“…Irrespective of the choice of data, the visual inspections of imagery or terrain models in an effort to identify sinkholes are time consuming, labour intensive and highly dependent on operator expertise which may lead to incomplete, unreproducible or biased results [63,65,71,77,78]. Automated feature extraction algorithms can therefore play a critical role in sinkhole inventory collection.…”

Section: Compilation Of Sinkhole Inventoriesmentioning

confidence: 99%

The Role of Earth Observation, with a Focus on SAR Interferometry, for Sinkhole Hazard Assessment

Theron

Engelbrecht

2018

Remote Sensing

View full text Add to dashboard Cite

Sinkholes are global phenomena with significant consequences on the natural-and built environment. Significant efforts have been devoted to the assessment of sinkhole hazards to predict the spatial and temporal occurrence of future sinkholes as well as to detect small-scale deformation prior to collapse. Sinkhole hazard maps are created by considering the distribution of past sinkholes in conjunction with their geomorphic features, controlling conditions and triggering mechanisms. Quantitative risk assessment then involves the statistical analysis of sinkhole events in relation to these conditions with the aim of identifying high risk areas. Remote sensing techniques contribute to the field of sinkhole hazard assessment by providing tools for the population of sinkhole inventories and lend themselves to the monitoring of precursory deformation prior to sinkhole development. In this paper, we outline the background to sinkhole formation and sinkhole hazard assessment. We provide a review of earth observation techniques, both for the compilation of sinkhole inventories as well as the monitoring of precursors to sinkhole development. We discuss the advantages and limitations of these approaches and conclude by highlighting the potential role of radar interferometry in the early detection of sinkhole-induced instability resulting in a potential decrease in the risk to human lives and infrastructure by enabling proactive remediation.

show abstract

“…Of note are recent developments applied to classify InSAR data in order to detect ground uplift and subsidence, and specifically to identify volcanic unrest. [29][30][31] Although promising, these developments do not make use of the different temporal signatures of signals of interest to reconstruct de-noised deformation patterns. Our auto-encoder takes as input a noisy InSAR time series reconstructed from successive SAR acquisitions, and outputs accumulated ground deformation taking place during the time series interval, with the atmospheric noise removed.…”

Section: Introductionmentioning

confidence: 99%

Autonomous Extraction of Millimeter-scale Deformation in InSAR Time Series Using Deep Learning

Rouet-Leduc,

Jolivet,

Dalaison

et al. 2020

Preprint

View full text Add to dashboard Cite

Systematic characterization of slip behaviours on active faults is key to unraveling the physics of tectonic faulting and the interplay between slow and fast earthquakes. Interferometric Synthetic Aperture Radar (InSAR), by enabling measurement of ground deformation at a global scale every few days, may hold the key to those interactions. However, atmospheric propagation delays often exceed ground deformation of interest despite state-of-the art processing, and thus InSAR analysis requires expert interpretation and a priori knowledge of fault systems, precluding global investigations of deformation dynamics. Here we show that a deep auto-encoder architecture tailored to untangle ground deformation from noise in InSAR time series autonomously extracts deformation signals, without prior knowledge of a fault's location or slip behaviour. Applied to InSAR data over the North Anatolian Fault, our method reaches 2 mm detection, revealing a slow earthquake twice as extensive as previously recognized. We further explore the generalization of our approach to inflation/deflation-induced deformation, applying the same methodology to the geothermal field of Coso, California.

show abstract

Subsidence feature discrimination using deep convolutional neural networks in synthetic aperture radar imagery

Cited by 24 publications

References 5 publications

SAR Interferometry for Sinkhole Early Warning and Susceptibility Assessment along the Dead Sea, Israel

SAR Interferometry for Sinkhole Early Warning and Susceptibility Assessment along the Dead Sea, Israel

The Role of Earth Observation, with a Focus on SAR Interferometry, for Sinkhole Hazard Assessment

Autonomous Extraction of Millimeter-scale Deformation in InSAR Time Series Using Deep Learning

Contact Info

Product

Resources

About