Applications: Volcanoes (2)

About ten years after the first feasibility tests of InSAR applied to the study and monitoring of active volcanoes in equatorial Africa, we are here to address the most outstanding results obtained while studying the volcano-tectonic very active region of the Kivu Rift and Virunga region in Central Africa, highlighting the opportunities offered by the new Sentinel mission.

Initially planned for volcanoes monitoring purposes, the SAR dataset acquired over the last decade offered invaluable tools to develop new methodologies and to contribute to other hazard assessments.

1. Volcano monitoring

The first motivation for an InSAR monitoring in the Kivu was the need of understanding the volcanic processes leading to eruptions of the Nyiragongo and Nyamulagira volcanoes. Both located in the East African rift, they are very active and raise deep concerns for the dense population living in the immediate vicinity, exposed e.g. to potentially devastating lava flows.

The initial results that demonstrated the relevance of InSAR for the modelling the deformation associated to the devastating eruption of the Nyiragongo volcano in 2002 (Wauthier et al, 2010) have strongly contributed to program systematic SAR acquisitions campaigns over that area. Indeed, the systematic acquisition increases the chance of capturing an event with pairs of SAR images offering small temporal and spatial baselines, which reduces the effect of the rapid vegetation-induced decorrelation common under these latitudes. These were found indispensable conditions in an operational monitoring perspective combined to ground based systems.

Only two historic eruptions characterize the Nyiragongo in 2002 and 1977, whereas its close neighbour, the Nyamulagira, has recorded more than 30 eruptions over the last Century and no less than 3 eruptions since the beginning of regular InSAR observations in 2005.

As early felt, InSAR guided our research group in the design and location of ground-based networks (such as permanent geodetic (GPS), tilt, seismic and infrasound networks). In parallel to countless efforts deployed to install and maintain those equipments in a tense context of armed conflicts, InSAR provided also a contribution to the study of the eruptions of the Nyamulagira in 2006 (Albino et al, in prep), 2010 (Smets et al, 2014), and 2011-2012 (Albino et al, submitted; d’Oreye et al, in prep).  In particular, the last eruption of Nyamulagira that started in November 2011 and lasted until April 2012, suggested that the activity occurring within the caldera could be linked to more distal fissure activity. Both sites display deformation patterns in interferograms and confirm the complexity of the plumbing eruptive system.

2. New methodologies

More specifically, obtained results enriched progressively with the InSAR archives database build over the last decade from various SAR sensors : initially focused on the source modelling and inversion of co-eruptive deformation, new techniques were developed as for example the Multidimensional Small Baseline Subset (MSBAS) (Samsonov S. and d’Oreye, 2012) which allowed a posteriori detection of deformation signals few weeks before the Nyamulagira 2010 eruption.

SB-InSAR was also developed, taking advantage of the wide band offered by new sensors like TERRASAR-X to split it into sub-bands and to generate (after range inverse Fourier transform of the extracted sub-bands) several lower-resolution images (Derauw et al, this meeting). The technique potentially provides an invaluable solution to the phase unwrapping process and will significantly help in our specific working environment.

3. Earthquakes and landslides

Over time, the framework of these studies also undergone important developments and was enlarged both geographically and thematically.

InSAR indeed allowed in 2008 the study of the Magnitude 6 earthquake that affected the South Kivu and Rwanda Western Province (about 100 km South of the Virunga), causing damages and fatalities (d’Oreye et al, 2010).

More recently, the technique became a key part of projects focused on the study multi-hazards and regional risk assessments. The same datasets are now applied to study landslides that are the major concern in South Kivu and in cities more specifically. Preliminary results using InSAR time series over Bukavu lead us to delineate several landslides including the major Kadutu-Funu one were thousands people are settled and were infrastructures are recurrently damaged.

4 Long-term vision

It is worth saying that interpreting the RS data requires field observations and regular exchanges with our local partners. The development of that research program with a long-term perspective in mind is only feasible with a strong commitment in capacity building and dialogs with local authorities and all stakeholders.

So far, the deployment of RS modern techniques locally has faced unaffordable skill but also technical constraints. Despite the free access to the imagery, poor connexions still represent an obstacle and do not permit the download of such an amount of data and certainly not in a near real time monitoring scheme.

There are no doubts that the newcomer SENTINEL will open the doors of a new world for InSAR; the facilitated access to the data, the systematic acquisition plan, the short re-visiting period will certainly boost the study of geodynamic processes in the Kivu rift Basin and the Virunga where remote sensing plays a particularly crucial role. But implementing InSAR in the region still represents another challenge for tomorrow.

A seven-year time series of satellite radar images over Unimak Island, Alaska-site of Westdahl Volcano, Fisher Caldera and Shishaldin Volcano-was processed using a model-free Persistent Scatterer Interferometry (PSI) technique assisted by Numerical Weather Prediction (NWP) model. In the present study, an advanced InSAR technique, namely a weather model-supported Persistent Scatterer InSAR (PSI) approach, is used for deformation history reconstruction from C-band InSAR time series. Both atmospheric artifacts and other non-deformation phase artifacts are treated through a PSI processing technique that combines information provided by Numerical Weather Prediction (NWP) models with the deformation model-free PSI technique [Hooper et al., 2012]. The choice of using NWP products is due to the fact that the frequent cloudy conditions over the Aleutians limit the usefulness of optical satellite sensor water vapor products. The atmospheric delays simulated by regional NWP forecasts are not directly subtracted from interferograms to reconstruct the subtle deformation in this study because the limited quality in reproducing the spatial pattern of atmospheric delays [Foster et al., 2013; Gong et al., 2013; Cimini et al., 2012; Liu, 2012]. Instead, NWP-derived atmospheric statistics are integrated into the PSI processing to optimize a filter for removing the atmospheric delays. 

In the presentation, we will outline the mathematical concept and algorithm implementation of using statistical information about the APS as a prior to optimize the choice of unknown filter parameters. The geodetic result of Unimak Island case study has been demonstrated and compared with campaign and continuous Global Positioning System (GPS) measurements as well as Small Baseline Subset (SBAS) time series InSAR results for quality assessment and geophysical interpretation. 

We observed subtle surface inflation at Westdahl Volcano that can be fit by a Mogi source located at approximately 3.6 km north of Westdahl peak and at depth of about 6.9 km that is consistent to the GPS estimated depth for the 1998 to 2001 time period. The magma chamber volume change decays during the period of 2003 to 2010. The deformation field over Fisher Caldera is steadily subsiding over time. Its best-fit analytical model is a sill source that is about 7.9 km in length, 0.54 km in width, and locates at about 5.5 km BSL underneath the center of Fisher Caldera with strike angle of N52°E. Very little deformation was detected near Shishaldin peak, however, a region approximately 15 km east of Shishaldin, as well as an area at the Tugamak range at about 30 km northwest of Shishaldin, show evidence for movement towards the satellite, with a temporal signature correlated with the 2004 Shishaldin eruption. The cause of these movements is unknown. The research result has been submitted to Journal of Geophysical Research-Solid Earth.

During continental rifting the majority of active volcanism typically becomes focused in a central rift axis. However, volcanic structures are also commonly found offset from the axis of spreading. Eruptions at such off-rift volcanoes are infrequent, and consequently the crustal processes are poorly understood. Nabro volcano, situated to the east of the Afar Rift Zone (ARZ), erupted on 12 June 2011. The eruption was only the second eruption of an off-rift volcano associated with the ARZ in modern times (since Dubbi volcano in 1861). We have recently published a study [Hamlyn et al., JGR 2014] comparing the post eruption ground displacement observed through TerraSAR-X InSAR with seismicity recorded at a local seismic array that we deployed following the eruption. This has provided new insights into the controls of caldera faults on subsidence. We now present a detailed analysis of the InSAR derived, time series of post-eruption subsidence using both TerraSAR-X and COSMO-SkyMed observations. We test competing hypotheses for the origin of the subsidence and stress the importance of atmospheric correction using independent data when interpreting topographically-correlated deformation at volcanoes.

Following the June 2011 eruption at Nabro, TerraSAR-X and COSMO-SkyMed were both tasked to prioritize the acquisition of SAR data over the volcanic centre. During the following 15 months, Nabro was imaged 129 times by these satellites, with an acquisition every 5 days on average. Using a Small Baseline Subset (SBAS) approach, we processed the 25 images acquired by TerraSAR-X between 1 July 2011 and 5 October 2012 on descending orbit 046, to create 34 interferograms. We complemented these with 19 images from ascending orbit 130 spanning 6 July 2011 to 10 October 2012 from ascending orbit 130, which we used to create 21 interferograms. We produced velocity ratemaps and time series using π-RATE, showing subsidence of up to 25cm/yr centred on Nabro [Hamlyn et al., JGR 2014]. COSMO-SkyMed satellite also imaged the volcano on a descending track between 26 June 2011 and 18 July 2012 within the Italian Space Agency project SAR4Volcanoes: a total of 64 images were acquired and used to produce 171 interferograms. Each dataset were used to create a detailed time series of incremental deformation of the Nabro caldera using an SBAS approach. We combine these data sets using a modelling approach to produce a detailed time series of the deflation of a Mogi source at ~7 km depth. The raw and smoothed (temporal window of 70 days, spatial window of ~3km) time series show that the volcano continued to subside for the entire period, with a rapid subsidence in the first 12 weeks, followed by a slowly declining rate. In addition there are three apparent pauses in the subsidence.

As the subsidence rate is approximately correlated with the topography of the volcano, we were concerned about the influence of atmospheric delays on our geophysical signal. To assess the impact of atmosphere delays, we use the outputs from ERA-Interim (ERA-I), a global atmospheric model computed by the European Centre for Medium-range Weather Forecasting (ECMWF), to correct each SAR acquisition. The atmospheric correction noticeably reduced the scatter in the time series, and removed the three apparent atmospheric artefacts. This result stresses the importance of applying atmospheric corrections using independent sources of information – the standard approach of filtering in space and time did not remove these atmospheric errors and without the ERA-I correction we would have produced a significantly different interpretation of the InSAR observations.

Finally, we explore alternative mechanisms that might explain the long-lived subsidence at Nabro volcano. In particular, we test models of thermal contraction and cooling, visco-elastic relaxation, and degassing. The Nabro eruption was one of the largest volcanic emissions of SO2 since the 1991 eruption of Mt. Pinatubo. Therefore we compare the InSAR derived subsidence to the SO2 emissions observed by the IASI instrument. A correlation between the rate of SO2 emission and the rate of subsidence suggests a common underlying process, possibly magma cooling.

Campi Flegrei caldera (South Italy) is among the areas with the highest volcanic risk in the world due to about 1 million people living within and around the caldera borders. In recent years Campi Flegrei caldera has experienced an accelerating uplift rate of the ground deformation. In particular, during the April 2012 - January 2013 time interval the caldera has shown a rapid uplift of about 6 cm with a peak rate of about 3 cm/month. This event led the Italian Civil Protection to raise the alert level from green to yellow.

In this scenario, we applied a joint inversion technique to DInSAR and GPS measurements, recorded between 2012 and 2013 at Campi Flegrei caldera, to image the kinematics of the emplacement of a magmatic sill beneath the caldera. In particular, we exploited the displacement time series that we obtained by processing 90 SAR images acquired from the COSMO-SkyMed sensor constellation along ascending orbits via the well-known DInSAR algorithm referred to as SBAS algorithm, and the displacement measurements provided by 14 continuous GPS stations deployed within the caldera and belonging to the permanent INGV-OV monitoring network.

We applied a geodetic imaging technique to determine the spatial and temporal evolution of the ground deformation source in the selected period. The preliminary results show that the most likely source has a planar geometry and is located at a depth of about 2700 m; this is in good agreement with earlier studies relevant to previous intervals proposing that the ground deformation source at Campi Flegrei consists in an inflating sill. The retrieved temporal pattern of the source geometry reflects that of a growing sill that, at the end of the considered period, has a roughly elliptical geometry with an extension of about 6 km in the EW direction and about 4 km in the NS one. The maximum aperture of the sill is of about 30 cm at its center.

To understand the dynamics of this phenomenon we used a numerical model of the emplacement of a magmatic sill, to fit the retrieved geometry. The parameters to be determined are: the average magma viscosity, the amount of magma already present in the sill before the 2012-2013 episode and the magma injection rate.

The achieved results show that the most likely value for the viscosity is between 10³-10⁴ Pa∙s and that to justify the observed deformation pattern it is required that the reservoir should have contained at least 10¹⁰ kg of liquid magma before 2012. The injection rate has two main peaks on September and December 2012, and a smaller one on March 2013. The first two peaks have a value of about 400 kg/s and a duration of 3-4 months. The total amount of injected magma is of about 8.2∙10¹⁰ kg. The magma viscosity value suggests that it has a femic composition; this agrees with recent findings in the analysis of gas geochemistry, which have shown that its magmatic source is a trachybasalt. 

The first peak is associated with a seismic swarm, located beneath the town of Pozzuoli. The swarm consisted in about 200 earthquakes (maximum magnitude 1.8) occurring within an interval of about 1.5 hours. The hypocenters were located outside the area usually affected by microearthquakes in the previous years.

Our finite element structural mechanical modeling shows that the inferred source caused a marked increase in the maximum shear stress along the rim of the sill. Indeed hypocenters were located very close to the northern edge of the growing magmatic reservoir. Our findings suggest a key to interpret the caldera unrest that, started about 60 years ago, has led to a maximum uplift in the area of more than 3 m. Consequently, the observed uplift phenomenon could be interpreted in terms of injection of limited magma batches feeding the growth of a shallow magmatic reservoir. Similar mechanisms have been inferred for other calderas, where the repeated emplacement of magmatic sills has been recognized having an important role in the evolution of the volcano. Accordingly, the observation of short evolution of volcanic precursory phenomena as well as the development of innovative real-time analysis techniques should be taken into account and fostered for an effective surveillance of the Campi Flegrei caldera.

This work has been partially supported by the Italian Department of Civil Protection and by the Italian Space Agency under theSAR4Volcanoes project (agreement n. I/034/11/0).

Mauna Loa is a basaltic shield volcano, located in in the Big Island of the Hawaii archipelago (central Pacific Ocean), which rises over 4 km above sea. It is one of the Earth's most active volcanoes, having erupted 33 times since its first well-documented historical eruption in 1843. Most recent eruption occurred in 1984, from March 24 to April 15. After this event, the volcano entered in a phase of quiescence, interrupted by three main episodes of unrest. The first one occurred in mid-2002, when an inflation phase started just after a brief swarm of deep Long-Period (LP) earthquakes. In late 2004, a dramatic increase in the inflation rate was immediately preceded by a more intense swarm of several thousands LP earthquakes. This Inflation slowed again in 2006, ceased altogether in late 2009, and resumed slowly in late 2010. The latest episode began in April 2014 and continues to date, with a variable rate of inflation behavior.

In this work, we investigate the deformation process affecting the Mauna Loa volcano from 2003 to 2014 by exploiting the advanced Interferometric Synthetic Aperture Radar (InSAR) technique referred to as Small BAseline Subset (SBAS) algorithm. This technique relies on the use of a large number of SAR acquisitions and implements an easy combination of small spatial and temporal separation (baseline) interferograms computed from satellite images. This approach allows preserving the characteristics of wide spatial coverage typical of satellite SAR systems since it is able to mitigate the noise effects (referred to as decorrelation phenomena) and, at the same time, to maximize the number of coherent targets.

The time series of line-of-sight (LOS) displacements derived from the multi-temporal and multi-platform SAR data were obtained using ENVISAT dataset, acquired by ascending and descending orbits over the 2003 - 2010 time period and COSMO-SkyMed images, taken from both orbital passes in the 2012-2014 time span. All the available SAR data used in this study were obtained through the GEO Geohazards Supersites and Natural Laboratories (GSNL) initiative.  

For each coherent pixel of the radar images, we compute time-dependent surface displacements as well as the average LOS deformation velocity. We also benefit from the use of the multi-orbit (ascending and descending) ENVISAT and CSK data to discriminate the vertical and east-west components of the displacement field.

From the analysis of the space-time deformation trend in the last 10 years, we observe a butterfly - shaped pattern of fringes centered on the volcano summit, which indicates an uplifting trend of about 2 cm from April 2014 to August 2014; this may suggest an inflation of the complex magma reservoir beneath Mauna Loa. Moreover, we compare the latest episode of unrest with those of 2002, 2004 and 2006 to better understand the long-term behavior of Mauna Loa volcano. The final step of this work consist in the inversion of the retrieved DInSAR results, in order to model both deep geological structures and magmatic sources to better understand the dynamics that drive on the volcano process. To characterize the geometry and the evolution of the volcano deformation source, we apply both forward and inverse modeling techniques. The former technique allowed defining roughly the geometrical parameters (shape, position and orientation) of the source, while the latter allowed retrieving its temporal evolution.

The obtained results have direct implication in the interpretation of the last unrest episode, providing a useful tool to aid the forecast of possible future eruptions at Mauna Loa.

SAR data used in this study were obtained through the GEO Geohazards Supersites and Natural Laboratories (GSNL) initiative.