Do we gain information by systematically correcting tropospheric noise in InSAR time series on volcanoes ? Insights from Piton de la Fournaise volcano (Reunion)

Satellite radar interferometry (InSAR) is a useful tool for measuring surface deformation at large-scale, but in tropical volcanic environment, tropospheric delay is a major source to the phase delays observed in the interferograms. Different methods have been proposed to mitigate atmospheric noise applied either on the time series with temporal filtering and Principal Component or on single interferograms, using empirical method or external datasets. At the regional scale of a volcanic island (>50 km), ECMWF-type meteorological models have already shown good performance for correcting long-wave signals and for enabling the detection of pre-eruptive inflation prior to Agung 2017 eruption (Albino et al. 2020). However, the performance of such 10-km resolution meteorological models is not adapted for the scale of volcanic edifices (<10 km).

At basaltic volcanoes, such as the Piton de la Fournaise (Reunion Island), most of the ground deformations signals are related to magma intrusions (Peltier et al., 2009), as shown by recent episodes of unrest in April 2021, December 2021 and September 2022. In this scenario, the signal over noise ratio is relatively high and such signals are detected on several interferograms without atmospheric corrections. However, even at small-scale, the atmospheric noise may change the amplitude and the pattern of the signal of displacement. This could have an effect on the values and uncertainties of the parameters of the intrusion retrieved during the data inversion but so far, we do not have a quantification of this effect. In addition to strong deformation signals related to dyke intrusions, GNSS measurements have shown a slow persistent inflation signal related to the re-alimentation of the magma reservoir between the emplacement of the intrusion in December 2021 and the recent intrusion of September 2022. Despites the good coverage of the GNSS stations ( 24 on the volcano), the information about the spatial extension of the signal is limited. However, this type of signal is not yet detected on InSAR time series as atmospheric noise is likely to mask such small amplitude signals.

The first objective of the internship will be to evaluate the effect of different atmospheric corrections for retrieving source parameters. First, we will generate a large dataset of synthetic signals combining ground deformation and different level of atmospheric noise. For the deformation signal and atmospheric noise, we will use a range of values that is characteristics of those observed at Piton de la Fournaise. Then, we will model each synthetic signal produced with GBIS, a Matlab code using the method of Monte-Carlo Markov-Chain (MCMC) for the data inversion (Bagnardi et al., 2018) that will allow us to estimate the optimal values and the level of confidence for each source’s parameter. The goal is to estimate how different corrections affect the retrieval of the source’s parameters and if some geometries of intrusion are more prone to this effect. Finally, we will apply the same method on real datasets, using the signals associated with the last three dyke intrusions emplaced successively in April 2021, December 2021 and September 2022.

As a second objective, the study will help to better constraint the limit of detection of InSAR time series at Piton de la Fournaise and therefore evaluate in which conditions slow deformation signals related to the re-pressurization of the magma reservoir can be detected. Sentinel-1 SAR images have been already processed for the period (2021-2022) using FAST-SAR (Fully Automated processing of Small Targets), a new service based on NSBAS processing chain and developed by the SNO ISDeform. The first task here will be to apply different strategies for the atmospheric corrections and to evaluate their relative performance. A preliminary study showed that the atmospheric noise is statistically reduced for only half of the interferograms using ECMWF high-resolution weather models. A way of improvement will be to improve the performance of regional weather models corrections by assimilating local information derived from the Zenithal Total Delay (ZTD) of the GNSS stations installed on the volcano, an effort that already started within the framework of an ongoing INSU project “ATMOVOLC”.

Albino, F., Biggs, J., Yu, C., & Li, Z. (2020). Automated Methods for Detecting Volcanic Deformation Using Sentinel‐1 InSAR Time Series Illustrated by the 2017–2018 Unrest at Agung, Indonesia. Journal of Geophysical Research : Solid Earth, 125(2), e2019JB017908.

Bagnardi, M., & Hooper, A. (2018). Inversion of surface deformation data for rapid estimates of source parameters and uncertainties : A Bayesian approach. Geochemistry, Geophysics, Geosystems, 19(7), 2194-2211.

Peltier, A., Bachèlery, P., & Staudacher, T. (2009). Magma transport and storage at Piton de La Fournaise (La Réunion) between 1972 and 2007 : A review of geophysical and geochemical data. Journal of Volcanology and Geothermal Research, 184(1-2), 93-108.