Skip to content

An ICA-based detection of volcanic deformation using InSAR data in Ontake volcano, Japan

Yutaro Shigemitsu1, Shungo Tonoyama2, Takeo Tadono1

  • Affiliations: 1Japan Aerospace Exploration Agency, Earth Observation Research Center, Ibaraki, Japan; 2Data Assimilation Research Team, RIKEN Center for Computational Science, RIKEN, Hyogo, Japan

  • Presentation type: Talk

  • Presentation time: Monday 09:30 - 09:45, Room S150

  • Programme No: 3.1.5

  • Theme 3 > Session 1

Abstract

Independent Component Analysis (ICA), a type of machine learning, is a powerful technique that can separate Interferometric Synthetic Aperture Radar (InSAR) data into surface deformation and noise-derived signals without complex assumptions. Recent studies have applied ICA to time-series InSAR data to detect volcanic deformation (e.g. Ebmeir et al., 2016, Gaddes et al., 2019). Most of these studies have used Sentinel-1 C-band data, and few studies have applied ICA to long-wavelength L-band data. L-band data are susceptible to noise from, for example, the ionosphere, so it is expected that ICA can be more useful for extracting surface deformation components from InSAR data that contain such noise. Here, we applied ICA to the ALOS-2 L-band time-series InSAR data from 2014 to 2022 for the Ontake volcano in Japan, where the 2014 phreatic eruption occurred. The results showed that the volcanic deformation component could be separated and extracted from the noise component to the same extent as in Narita et al. 2019. Based on our results, the estimated surface deformation component around the crater could be up to 20 cm smaller than that estimated by some previous studies, in which case our method is expected to provide more accurate estimates of volcanic deformation. In the future, we aim to improve the accuracy of surface deformation estimates by confirming the consistency with the ICA-derived deformation component based on modelling results using analytical and numerical methods.