Automated methods for detecting volcanic deformation using Sentinel-1 InSAR time series

Radar satellites, such as Sentinel-1, are now able to routinely produce time-series of ground deformation around the world. This represents a remarkable opportunity for operational use, but poses new research challenges associated with the handling of large data volumes and the production of consistent and reliable datasets across multiple environments. This presentation focusses specifically on volcano monitoring, but most considerations also apply to other applications. Firstly, we discuss the importance and use of atmospheric corrections in volcanic environments using examples from Indonesia and Turkey, where we demonstrate that previously reported deformation signals can be attributed to tropical atmospheric conditions over high relief edifices. We compare the use of global weather models and phase-elevation corrections and find that each is suited to different atmospheric conditions, but that a combined approach works best for large datasets and timeseries. Next, we consider how detection or flagging algorithms can be applied to large datasets to reduce the volume of data that requires expert inspection. We assess the ability of InSAR time-series to detect deformation anomalies using either a fixed threshold or a cumulative sum control chart, using the 2017-2018 crisis at Agung volcano as a case example. We assess the Receiver Operating Characteristics (ROC) of several methods and find the average area under the ROC curve to be 0.5 for the uncorrected data (corresponding to no discrimination capability), 0.8 after combined atmospheric corrections (weather model and phase-elevation approaches), and 0.98 using a cumulative sum control chart (corresponding to ideal separation between classes).