Monitoring Volcanic Deformation Using InSAR: An Optimised InSAR Time-Series Approach for Seasonally Snow-Covered Volcanoes

Monitoring Volcanic Deformation Using InSAR: An Optimised InSAR Time-Series Approach for Seasonally Snow-Covered Volcanoes

¹Tianyuan Zhu*, ¹Juliet Biggs, ¹Alison Rust, ²Milan Lazecký, ³Loreto Cordova

¹School of Earth Sciences, University of Bristol, Bristol, United Kingdom

²COMET, School of Earth and Environment, University of Leeds, Leeds, United Kingdom

³Servicio Nacional de Geología y Minería (SERNAGEOMIN), Santiago, Chile
*tianyuan.zhu@bristol.ac.uk

 

Satellite-based Interferometric Synthetic Aperture Radar (InSAR) has been widely used for monitoring volcanic deformation, especially since Sentinel-1 launched in 2014, providing an unprecedented volume of routinely acquired, open-access data. Automated systems now continuously process interferograms and regularly update deformation time series, providing a valuable dataset for monitoring volcanoes globally. However, seasonal snow leads to coherence loss and subsequent unwrapping errors in interferograms, causing gaps in the network of the automated time-series analysis and reducing deformation accuracy. As ~41% of subaerial Holocene volcanoes exhibit seasonal snow cover (with snow persistence of 7-90%), optimising InSAR processing for seasonally snow-covered volcanoes would substantially improve monitoring active volcanoes, especially in high-latitude and high-altitude areas.  

 

In this study, we developed an optimised InSAR time-series processing workflow using MODIS 8-Day Snow Product, which has been successfully applied to Laguna del Maule (LdM), a caldera with strong seasonal snow cover in Chile (Snow Persistence=51%). At LdM, the default product from the LiCSBAS auto-processing system underestimates the average line-of-sight deformation by 28% at GNSS station MAU2 between 10/2014 and 06/2023. To improve the accuracy of time series, we adapt the LiCSBAS time-series processing strategy using the quantified relationship between MODIS 8-Day Snow Products and Sentinel-1 InSAR coherences. The optimised workflow, including an algorithm based on Graph Theory for network selection, reduced data requirements by ~90% and LiCSBAS processing time by ~80%, while improving the accuracy of the LiCSBAS-processed deformation to match GNSS observations.

 

Vegetation is another crucial factor in coherence loss, and cloud cover affects optical satellite data. Using MODIS products, we also show that over 50 of 484 seasonally snow-covered volcanoes have lower Normalized Difference Vegetation Index (NDVI) and cloud-obscured duration than LdM (NDVI=0.16; Cloud Duration=144 days), confirming that seasonal snow is their dominant source of coherence loss and MODIS products are applicable.

 

We applied our validated workflow to seasonally snow-covered volcanoes across a range of environments with different cloud and vegetation cover to produce a long-term deformation (2014–present) using Sentinel-1 data. The optimised workflow has implications for the accuracy and efficiency of global volcano monitoring, improving the quality of modelling and forecasting.