1. INTRODUCTION

The European Plate Observing System (EPOS) ‎[1] is a long-term plan to foster and facilitate the integrated use of data, products, software and services made available through distributed European Research Infrastructures (RI) in the field of Solid Earth Science (SES). In particular, EPOS is a pan-European Research Infrastructure of the ESFRI Roadmap and it has been recently established with an ERIC (European Research Infrastructure Consortium) hosted by INGV in Italy. EPOS is supported by 25 European countries and several international organizations.
EPOS integrates a large number of existing European RIs belonging to several fields of SES, referred to as Thematic Core Services (TCS), from seismology to geodesy, near fault and volcanic observatories, as well as anthropogenic hazards and satellite observations. The EPOS vision is that the integration of the existing national and trans-national RIs will facilitate the access and use of the multidisciplinary data recorded by the solid Earth monitoring networks, acquired in laboratory experiments and/or produced by computational simulations. The EPOS establishment will foster the interoperability of products and services in the Earth science field to a worldwide community of users. Accordingly, EPOS aims at integrating the diverse and advanced European RIs in the field of Solid Earth Science and building on new e-science opportunities to monitor and understand the dynamic and complex Solid Earth System. One of the EPOS TCS, referred to as Satellite Data (SATD), aims at developing, implementing and deploying advanced satellite data products and services, mainly based on Copernicus data (namely Sentinel acquisitions), suitable to be largely used by the SES community.
For a research infrastructure supporting a large community from multiple countries, as the case of EPOS, it is critical that the underlying infrastructure and the computing resources carefully take into account the operational environment for users and its sustainability, both technical and financial. This point is particularly relevant for the SATD RIs that, to deploy robust and effective services, have to properly manage several issues related, for example, to the satellite data access, archive handling and storage, the management of the computing facilities, and the efficient and automatic processing chains. In this framework, the European scenario is rapidly evolving and several pan-European initiatives have been recently fostered. Among all, the European Open Science Cloud (EOSC) ‎[2] and the Copernicus Data and Information Access Services (DIAS) ‎[3] platforms represent the most promising opportunities to reach the long-term technical sustainability of the EPOS TCS SATD.
This work is focused on the technological enhancements and the activities carried out to implement the TCS SATD and to deploy effective EO satellite services in a harmonized and integrated way by benefitting from the current and future European satellite scenario. In particular, we present the advanced Differential SAR Interferometry (DInSAR) techniques implemented to provided robust services to map and investigate the ground motion of local and wide scale deformation phenomena, from regional to continental analysis. Finally, we show the procedures and methods developed or adopted by the TCS SATD to guarantee good data management and stewardship by following the FAIR principles.

2. EPOS TCS SATELLITE DATA
The structure of the EPOS TCS Satellite Data is shown in Figure 1. The scope of this TCS is the implementation of Earth Observation services, based on satellite observations, transverse to the large EPOS community and suitable to be used in several application scenarios. In particular, the main goal is to contribute, with mature services that have already well demonstrated their effectiveness and relevance in investigating the physical processes that control earthquakes, volcanic eruptions and unrest episodes, as well as those driving tectonics and Earth surface dynamics. The development of TCS SATD is supported by 5 European institutions providing different services (Table I), CNR and INGV (Italy), CNRS (France), CSIC (Spain) and University of Leeds (UK), and benefits from the collaboration with the European Space Agency (ESA).
At this stage, two levels of products and services, based on Differential SAR Interferometry (DInSAR) techniques ‎[4] to estimate and analyze Earth surface displacements and terrain motion mapping for geohazards applications, are distributed. The first level deals with “standard” satellite products/tools (e.g., SAR interferograms, LOS displacements maps and deformation time-series generation). The second level concerns value-added satellite products/tools (e.g., modelling analyses, 3D displacement maps, source mechanisms, fault models, strain maps). The TCS services are mainly based on Copernicus data (Sentinel-1/2 datasets); in addition, advanced DInSAR web processing services dealing with ERS-1/2, ASAR-ENVISAT, and Sentinel-1 data are made available by the TCS SATD. Since the services include both access to products and processing utilities, we have to consider two specific functioning modes:
• Continuous mode - systematic and periodic generation of products (e.g. the systematic production of updated surface deformation time series over given areas);
• On-demand mode - users run the tools and process chosen satellite datasets (e.g. ad hoc generation of deformation measurements using satellite observations during a telluric crisis, such as a co-seismic motion map).
The continuous services systematically generate products directly accessible by the users. Such products are relevant to areas of the Earth surface significant for the Solid Earth Science community (volcanoes, faults, seismogenetic areas, geohazard supersites, etc), while the on-demand services make available to users advanced web-tools able to generate satellite products by processing Copernicus datasets (Figure 2).
The TCS community worked to apply FAIR principles [6] to its products and data. Following OGS guidelines, the TCS implements the ISO 19115 standard, that has been adapted to describe interferometric SAR products, both maps and time series. Moreover, all the TCS products are distributed following the open data and open access policy, with CC-BY license.
The TCS has a unique thematic interface towards the EPOS central hub, referred to as Integrated Core Services (ICS) (Figure 1). This interface is represented by the Geohazards Exploitation Platform (GEP) ‎[5], developed with the support of ESA, which is able to provide interoperable access to data products, web processing tools and processing facilities. This unique gateway provides the user interface (GUI) and the interoperability layer of the TCS, establishing unique AAAI and API for the several RIs.

REFERENCES
[1] European Plate Observing System (EPOS) [Online] Available: http://www.epos.org
[2] European Open Science Cloud (EOSC) [Online] Available: https://www.eosc-hub.eu
[3] Copernicus Data and Information Access Services (DIAS). [Online] Available: http://copernicus.eu/news/upcoming-copernicus-data-and-information-access-services-dias
[4] A. K. Gabriel, R. M. Goldstein, and H. A. Zebker, “Mapping small elevation changes over large areas: Differential interferometry,” J. Geophys. Res., vol. 94, no. B7, pp. 9183– 9191, Mar. 1989.
[5] Geohazards Exploitation Platform (GEP) [Online] Available: https://geohazards-tep.eu/
[6] Wilkinson, M., Dumontier, M., Aalbersberg, I. et al., “The FAIR Guiding Principles for scientific data management and stewardship,” Sci Data, vol. 3, n. 160018, 2016. https://doi.org/10.1038/sdata.2016.18

Differential Synthetic Aperture Radar Interferometry (DInSAR) has extensively proven in the last decades its unprecedented capability to measure Earth surface displacements at very-large scale and with high accuracy. In particular, DInSAR techniques allow us to retrieve ground deformation related to both natural hazards (earthquakes, tectonic movements, volcanic phenomena, landslides, etc.) and anthropic actions (mining, gas injection/extraction, groundwater exploitation, etc.) with centimeter to millimeter accuracy [1]. Moreover, the current DInSAR scenario is characterized by a huge availability of SAR data acquired by the large number of operating SAR sensors, such as ALOS-2, COSMO-SkyMed, RADARSAT-2, SAOCOM, Sentinel-1, TerraSAR-X, which will be further increased thanks to the already planned NISAR and ROSE-L missions.
Actually, one of the main limitations for correctly retrieving deformation signals from DInSAR results is the Atmospheric Phase Screen (APS) component, which accounts for the presence of atmospheric artifacts contaminating the intereferometric measurements. Indeed, since atmospheric properties such as temperature, pressure and humidity can vary in space and time, the refractivity index of the atmosphere (through which the transmitted microwave pulses and the backscattered signals propagate) may change between the acquisition times of the two SAR images resulting in an interferogram. Consequently, the generated interferogram will present the above mentioned APS component not related to deformation. Moreover, there can be scenarios in which distinguishing the APS from the real deformation is particularly complicated. This is, for instance, the case of areas characterized by the presence of significant topography (since atmospheric properties vary with height producing a topography-correlated atmospheric phase component) and significant deformation as well, as in the case of volcano eruptions. This is particularly true when the displacement component is comparable to the one owing to the atmospheric artifacts [2].
A significant number of methods for APS mitigation have been proposed in literature over the last years, which can be essentially classified in two categories: DInSAR time series and external data based approaches. The first solutions account for the APS statistical properties both in space and in time in order to filter out atmospheric contributions from DInSAR time series, and is typically effective with large datasets. The second techniques rely on the use of external auxiliary data (e.g. Zenith Total Delay from GPS measurements, meteorological models or Global Atmospheric Models (GAM)) in order to directly estimate and remove the atmospheric artifacts from interferograms; the major drawbacks, in this case, are represented by the generally low spatial and/or temporal resolutions of the external data. However, thanks to the development of numerical weather prediction models over recent years, we have now available GAM datasets providing accurate atmospheric parameters measurements having rather high resolutions. In particular, the ECMWF ERA-5 datasets, generated by exploiting the Copernicus Climate Change Service Information [3], are available on a global-scale, covering the Earth on a 30km grid. In particular, the ERA-5 data resolve the atmosphere using 137 levels from the surface up to a height of 80km and are available hourly.

In this paper we take in considerations three sites particularly challenging from the point of view of the APS estimation and removal, because of the complexity of the investigated scenarios:
i) the Canary island of La Palma (Spain), located in the Atlantic Ocean, which is a volcanic complex characterized by a long eruptive activity, focusing on the last eruption occurred on 19 September 2021 that is still ongoing and to date has caused extensive damage to homes, infrastructures and many productive activities on the island;
ii) the volcanic area located in the Napoli bay area (Italy) that includes the Vesuvio Volcano and the caldera of Campi Flegrei, the latter being characterized by the bradyseism phenomenon that caused in the last years an uplift with a rate reaching 10 cm/year;
iii) the Etna Volcano (Sicily), which is the Europe's largest and most active volcano, characterized by frequent eruptions often accompanied by large lava flows.

The three above-mentioned sites are all characterized by relevant deformations and significant heights, therefore it is often difficult to separate the APS topography-related interferometric component from the actual ground displacement causing an underestimation of the last one.
The aim of the work is to evaluate and compare the performances of the APS filtering solutions:
1) exploiting the join availability of spatial and temporal information given by long deformation time series generated through Sentinel-1 data acquired over the sites of interest,
2) applying to the DInSAR interferograms stacks the APS corrections directly calculated by exploiting the external ERA-5 data [4].
In the final presentation we will show a comprehensive analysis of the results obtained through the above mentioned two different approaches, and achieved both on individual interferograms and on deformation time series. Moreover, we will also investigate the possibility to combine the two techniques, in order to estimate and remove the APS interferometric component in a more accurate way. Finally, the impact of using the experimental Sentinel-1 ETAD products, which account for tropospheric and ionospheric corrections [7], will be possibly analyzed.



References

[1] D. Massonnet and K. L. Feigl, “Radar interferometry and its application to changes in the Earth’s surface,” Rev. Geophys., vol. 36, no. 441-500, 1998
[2] Zebker, H. A.,Rosen, P. A., and Hensley, S. (1997), Atmospheric effects in interferometric synthetic aperture radar surface deformation and topographic maps, J. Geophys. Res., 102( B4), 7547– 7563, doi:10.1029/96JB03804
[3] https://cds.climate.copernicus.eu/
[4] Jolivet, R., P. Agram and C. Liu, Python-based Atmospheric Phase Screen estimation - User Guide (2012), http://earthdef.caltech.edu .
[5] Casu, F., Elefante, S., Imperatore, P., Zinno, I., Manunta, M., De Luca, C. , and Lanari, R., SBAS-DInSAR Parallel Processing for Deformation Time-Series Computation. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 7, no. 8, pp. 3285–3296, 2014
[6] Manunta, M.; De Luca, C.; Zinno, I.; Casu, F.; Manzo, M.; Bonano, M.; Fusco, A.; Pepe, A.; Onorato, G.; Berardino, P.; De Martino, P.; Lanari, R., The Parallel SBAS Approach for Sentinel-1 Interferometric Wide Swath Deformation Time-Series Generation: Algorithm Description and Products Quality Assessment, IEEE Trans. Geosci. Remote Sens., 2019, 57
[7] https://sentinels.copernicus.eu/web/sentinel/missions/sentinel-1/data-products/etad-dataset

The “Pas de l’Ours” landslide, located in the Queyras valley (Southeast France) is undergoing periods of fast deformation that began in the Spring 2017. The total moving mass is estimated at 17 million cubic meters, with a width of 1 km and a length of 600 m, which makes it currently one of the largest active landslides in the French Alps. In addition to the large deforming mass, numerous rockfalls and mudflows have occurred and have severely damaged the road that is located at the foot of the landslide.

Ground-based instruments including a GBSAR, GNSS, and seismometers have been deployed on-site to monitor the landslide evolution. In addition, we are monitoring the landslide motion using SAR data from the Sentinel-1(S1) satellite and optical images from Sentinel-2 (S2) and Planet. These various datasets provide us with multiple measurements that can be combined to better understand the complete landslide behavior.

We analyze the landslide deformation pattern measured by the satellite images and ground-based InSAR techniques between 2015 and 2019. The SAR acquisitions are processed using the SqueeSAR algorithm and ISCE. The optical images are processed with the MPIC-OPT-SLIDE service available on the ESA Geohazards Exploitation Platform (GEP). These measured derived from satellite acquisitions are compared with the in-situ measurements (GB-InSAR and GNSS’s). We show the complementarity of these techniques to measure different kinematic regimes from tens of millimeters per year to tens of meters per year. We are able to detect the onset of the acceleration of 2016-2017 and discuss the triggering factors. We also investigate the progressive decrease of the magnitude of the following accelerations (2018, 2019). The location of the active parts of the landslide also varies through time. This work highlights how complementary monitoring techniques can be combined to retrieve the evolution of the slope instabilities in various kinematic regimes, and offers a new perspective into the physical processes controlling landslide dynamics.

Accurate and timely observation of mountain processes is considered one of the most important scientific activities in the field of environmental sciences in recent years, as it allows to detect indications and/or precursors of changes potentially impacting ecosystems at regional and global scales. One of the peculiarities of geologically young mountain ranges is the possibility to directly observe and study the evolution of paraglacial and periglacial morphological processes, i.e. all earth surface modifications that are directly conditioned by cyclic glaciation and deglaciation periods. Because of the current climatic changes, modifications of the mountain landscape due to large mass movements (i.e., glaciers, landslides of different size and typology and rock glaciers) are more and more observed and their impacts are expected to increase. In some cases, the rapid and potentially catastrophic evolution of such mass movements might directly affect anthropogenic infrastructure, economic activities, and also human lives. Hence, systematic mapping and monitoring of existing slope instabilities as well as the investigation of past catastrophic slope failures are essential for an effective hazard assessment, risk management and disaster response.

In this contribution, we show the initial results obtained by processing and analyzing available satellite radar datasets acquired from the ESA Sentinel-1 mission in the period 2018-2021. We focused on the Bhagirathi Valley, Uttarakhand, India, a high alpine area more and more threatened by large and catastrophic slope collapses. Currently different hydropower projects are under construction or in the planning phase within the region, increasing infrastructure at risk and the potential for cascading disasters such as the Chamoli rock/ice avalanche in 2021. We use standard radar interferometry to first detect and classify areas affected by potential instability. In addition, we focus on specific locations to determine the spatial and temporal evolution of surface displacements and to identify potential changes of trends associated to climatic variables. The result of this large scale and systematic investigation will be the base to test and calibrate numerical models of mass movements. With these models, the simulation of rock-, ice- and snow avalanches, as well as processes combining these input materials, enables the assessments of the hazard intensities and the generation of hazard indication maps for the region. These are essential tools for the planning of effective mitigation measures such as hazard zonation.

The last years, impacts of natural disasters on populations and infrastructures are rising worldwide. All efficient solutions are therefore required to support hazard and risk assessments, land-use planning, public risk financing and disaster forecasting through integrated landslide risk management systems. For landslide disasters, effective and extensive products are challenging at regional and country scales, especially when robust landslide inventories, appropriate destabilization factors and triggers or exposed populations and infrastructures must be captured and post-processed.
In the same time, the acquisition, quality and access to satellite optical and radar observation data (with for example Copernicus, Pléiades or TerraSAR missions) and rainfall measurements (such as Global Precipitation Measurement or Global Forecast System missions) significantly progressed the last decade. In addition, robust and efficient models are now trained enough to support near-real time landslide and risk assessments, with for example the Landslide Hazard Assessment for Situational Awareness (LHASA), the FLOW path assessment of gravitational hazards at a Regional scale (FLOW-R) or the Potential Impact Index (PDI) models.
This presentation aims then to present the prototype App LHIS, for LANDSLIDE HAZARD INFORMATION SYSTEM. Indeed, LHIS promotes the landslide awareness and disaster risk financing by providing an App to anticipate, forecast and respond to incipient landslide events in near-real time. LHIS is based on integrated landslide-related models (ALADIM, LHASA, FLOW-R, PDI) automatically executed online and applied on HR satellite imagery. Its implementation as a web-service within the Geohazard Exploitation Platform (GEP) allows an easy access, processing and visualization of these EO-derived products. LHIS targets two main usage modes: LHIS-Nowcast and LHIS-Impact.
LHIS-Nowcast aims to forecast in near-real time the landslide hazard triggered by extreme rainfall events and identify and quantify related exposed infrastructures and populations. Based on the chained LHASA, Flow-R and PDI models, its indeed models the failure susceptibilities of potential landslide sources identified according to rainfall nowcasts and computes their maximum propagations. Then it identifies related exposed populations and infrastructures that can be reached and estimates the potential damage costs.
LHIS-Impact aims to map and assess damages in the aftermath of a major landslide event. Based on the chained ALADIM and PDI models, it can detect changes on HR to VHR optical satellite images before/after a specific selected event and automatic delimitation of the impacted area. Then it identifies the affected infrastructures and populations and computes the real damage costs.
Together, both LHIS usages modes can therefore support land-use mapping, for cost/benefit analyses of prevention measures), as well as estimation of the insurance coverage needed for reconstruction, by providing pertinent results for parametric insurance calculations, including landslide inventories, susceptibility and hazard maps, potential damages and costs analyses in near-real time, and real damages and cost after a major landslide disaster.
The prototype was developed and tested in Morocco in close collaboration with the FSEC (solidarity fund against catastrophic events) and the World Bank. The input data, processing details and results, calibrated for the pilot study on the Rif Tangier-Tetouan peninsula (Northern Morocco), will be presented.

Following a seismic swarm that started on 11 September 2021 and gradually intensified, a magma pathway propagated along the Cumbre Vieja rift zone of La Palma. On 19 September 2021, the eruption began by the opening of an 800 m long fissure located in the area of Cabeza de Vaca, El Paso, on the mid-western flank of Cumbre Vieja. The eruption intensified over the next weeks and was characterized by lava fountaining from multiple vents, Strombolian explosions, and advancing lava flows towards the western coast of the island. Thereby, over 2,600 houses were destroyed. On 28 September, the lava flows entered into the ocean and initiated the formation of a lava delta that is episodically growing. Due to the high impact hazards, the area is hardly approachable by field sensors allowing estimation of the dimension and evolution of surface changes. Therefore, we have been acquiring and analyzing multiple sensor satellite data. During the initial dike intrusion period, multi-temporal differential SAR interferometry analysis of C-band (Sentinel-1) and X-band SAR sensor data (PAZ, TerraSAR-X/TanDEM-X, Cosmo-SkyMed) showed over 40 cm deformation of the affected slope towards the sea. The following eruption was monitored by SAR amplitude data, VHR and HR optical satellite data (Pléiades, GeoEye, Sentinel-2, Landsat-8, hyperspectral DESIS) to study the spatio-temporal evolution of lava flows and ash deposition. At the time of writing (15 of November 2021), the lava flows of the still ongoing eruption covered an area of over 10 km². The growing lava delta accumulated to an area of ~0.4 km². Thermal data of Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS) have been jointly analyzed to estimate the temporal evolution of the lava effusion rate and the total volume (6.22 × 10^7 ± 3.11 × 10^7 m³ by mid of November 2021). Although this thermally identified volume is already close to those of other historical eruptions on La Palma, it is largely underestimating the real volumes that erupted in 2021. Monitoring by the mid-resolution thermal sensors MODIS and VIIRS has the advantage of a high observation frequency, but only lava emplaced at the surface during the time of the satellite overpass can be detected. The difference of the thermally identified lava volumes to estimates arising from field observations may provide a hint of the importance of hidden lava flows. Lava flowing underground and in tubes and directly entering the ocean cannot be detected by these sensors. Therefore, change analysis between pre-eruption digital elevation models (DEMs) and newly created co-eruption DEMs from bi-static TanDEM-X data as well as from Pléiades stereo data are essential to understand the subaerial and hidden lava flow dynamics and to derive the total erupted lava volume.