Abstract
AbstractAutomated systems for detecting deformation in satellite interferometric synthetic aperture radar (InSAR) imagery could be used to develop a global monitoring system for volcanic and urban environments. Here, we explore the limits of a convolutional neural networks for detecting slow, sustained deformations in wrapped interferograms. Using synthetic data, we estimate a detection threshold of 3.9 cm for deformation signals alone and 6.3 cm when atmospheric artifacts are considered. Overwrapping reduces this to 1.8 and 5.2 cm, respectively, as more fringes are generated without altering signal to noise ratio. We test the approach on time series of cumulative deformation from Campi Flegrei and Dallol, where overwrapping improves classification performance by up to 15%. We propose a mean‐filtering method for combining results of different wrap parameters to flag deformation. At Campi Flegrei, deformation of 8.5 cm/year was detected after 60 days and at Dallol, deformation of 3.5 cm/year was detected after 310 days. This corresponds to cumulative displacements of 3 and 4 cm consistent with estimates based on synthetic data.
Highlights
Interferometric synthetic aperture radar (InSAR) can be used to measure ground displacement over large geographic areas
Previous studies have shown that deep convolutional neural networks (CNNs) have the capability to identify volcanic deformation signals in wrapped interferograms when trained on large data set (Anantrasirichai et al, 2018, 2019; Valade et al, 2019)
Shifting the wrap boundaries produces a small reduction in detection threshold for noise-free data but no change in the overall detection threshold when noise is included in the simulation as shown in Figure 2
Summary
Interferometric synthetic aperture radar (InSAR) can be used to measure ground displacement over large geographic areas. Previous studies have shown that deep convolutional neural networks (CNNs) have the capability to identify volcanic deformation signals in wrapped interferograms when trained on large data set (Anantrasirichai et al, 2018, 2019; Valade et al, 2019). The approach of (Valade et al, 2019) only tested short-term interferograms that showed deformation of >10 cm Global Volcanism Program (2019). These high rates are typically only observed for very short periods associated with dike intrusions or eruptions (Biggs & Pritchard, 2017). There are many deformation signals that occur at lower rates but for longer duration, such as sustained uplift at silicic volcanoes (Henderson & Pritchard, 2017; Lloyd et al, 2018; Montgomery-Brown et al, 2015; Remy et al, 2014; Trasatti et al, 2008), subsidence and heave in former coalfields (McCay et al, 2018), engineering projects such as tunnelling, and landsliding of natural and engineered slopes (Whiteley et al, 2019)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
