On 15 November 2017 a Mw 5.5 earthquake struck the southeast part of the Republic of Korea (South Korea), injuring many people and causing extensive damage in and around the city of Pohang. This event was preceded by the September 2016 Mw 5.5 Gyeonju earthquake, occurred ∼30 km farther south on one of the largest fault of the region, the Yangsan Fault. These two earthquake were the largest earthquake ever recorded since the last century in Korea. The November 2017 earthquake has triggered intense public debate in South Korea about its potential link with geothermal stimulation operations at a nearby EGS site. The proximity of this industrial facility with the epicentre of the earthquake has raised public concerns in South Korea.
In the absence of data from a local seismic network, we use advanced full-waveform seismological techniques, applied to regional and teleseismic network data, to locate the Pohang earthquake sequence and determine the source parameters of the largest events. We also use space borne radar interferometry to quantify the coseismic deformation and obtain an independent estimate of the mainshock source parameters. Seismicity and geodetic results are in agreement regarding location, depth and fault geometry. The hypocentral depths are shallower than the characteristic background seismicity of the area and mainly in the range 3-7 km, similar to the depth of injection (∼4 km). Moreover, the mainshock and most of the aftershocks occurred within 2 km of the EGS site. Our seismological and geodetic analysis rules out a re-activation of the Yangsan fault; however, we found that the earthquake transferred static stress to its northern branch, potentially increasing the seismic hazard in the area. Our results demonstrates the extent to which a suspected case of induced seismicity can be investigated without local data from site operators.

Existing international and Norwegian infrastructure projects, e.g., the GlobCurrent project under the European Space Agency (ESA), NorDataNet, NMDC and NORMAP under the Research Council of Norway, provide open data access through the OPeNDAP protocol following the conventions for Climate and Forecast (CF) metadata. The CF conventions are designed to promote the processing and sharing of files created with the NetCDF application programming interface (API). This approach is now also being implemented in the Norwegian Sentinel Data Hub (satellittdata.no) to provide satellite EO data to the user community. Simultaneously with providing simplified and unified data access, these projects also seek to establish and use common standards for discovery metadata allowing development of standardized tools for data search and (subset) streaming to perform actual scientific analysis.

A combination of software tools and data access, which we call the Geo-Scientific Platform as a Service (SPaaS), takes advantage of these opportunities to harmonize and streamline the search, retrieval and analysis of integrated satellite and auxiliary observations of the oceans in a seamless system. The core part of the Geo-SPaaS is a metadata catalog to store granular metadata describing the structure, location and content of available satellite, model, and in situ datasets. Data analysis tools include software for visualization, interactive in-depth analysis, and server-based processing chains. The Geo-SPaaS components are integrated in virtual machines (e.g., VirtualBox, VMware), of which provisioning and deployment are automatized using existing state-of-the-art open-source tools (e.g., Docker, Vagrant, Ansible, Git, Python). The software, including scientific tools and virtual machine configurations, is available on GitHub (https://github.com/nansencenter/), and is coupled to an online continuous integration system (Travis CI).

The presented software systems facilitate the development of geophysical algorithms based on synergetic use of satellites and in-situ measurements, and models. In this presentation, we demonstrate example use cases of the system within ocean and sea ice research, including exploitation of in-situ data (e.g., from ARGO floats) and model simulations available from the Copernicus Arctic Marine Services (ARC-MFC).

The Geo-SPaaS enables researchers to more quickly develop and test scientific algorithms, which then can be operationalized on server systems. Future prospects of Geo-SPaaS include the creation of shared remote metadata repositories (Geo-SPaaS nodes). This involves an extension of the existing metadata catalog solution in Geo-SPaaS that will allow users to store and search geospatial metadata on a remote host, enabling users to reduce the time spent on defining and ingesting own metadata in the catalog. The initial results of a study to combine ElasticSearch with the existing Python/Django based Geo-SPaaS solution will thus be presented.

Area C - Board 296

Technology, EO infrastructure, and science have been advancing rapidly in several key areas relevant to understanding the Earth and its biogeochemical processes. Together, these advances create new opportunities to monitor biodiversity change, ecosystem condition, and the ecosystem services upon which humans depend. Coincidentally, these advances are occurring at a time when the climate is changing rapidly and human population is reaching new levels, both of which increase pressure on the world’s natural systems and its biodiversity, which has been declining for decades. How can we take advantage of these opportunities to better monitor and understand changes in biodiversity and ecosystems?

The traditional approach to develop EO algorithms, process data, and generate products follows rather fixed paths where specific algorithms are developed and approved and then used to generate a mostly static list of standard products. These products are then available for download to individual users to utilize on their local machines. Advances in computing technology, including decreased cost and the availability of the cloud, now enable more flexible approaches. These approaches are actually necessitated because of the increase in the amount and diversity of remote sensing data available, by advances in science, and by the need for more inter-disciplinary collaboration that the advances in science require.

Recently a diverse group of biodiversity scientists, ecologists, and remote sensing experts got together to assess the opportunities these advances in technology provide for improving monitoring of biodiversity and ecosystems at a global scale. A suite of needs and opportunities were identified, including:

• Facilitation of collaboration and cross-discipline interaction
• Increased agility and flexibility in algorithm development and data processing approaches
• Convenient access to unprecedented volumes and types of data and support for the related processing
• Support for collaborative algorithm development as well as experimental research and applications using EO data
• More integration, including more fusion of different types of sensors as well as better integration of remote sensing and in situ data
• Development and availability of Analysis Ready Data to simplify processing
• More derived products and tools to increase product usability and understanding, particularly for applied users
• Increased access and usability of in situ and airborne remote sensing data

With these in mind the general architecture of a system to utilize EO in support of global biodiversity and ecosystem monitoring was conceptualized. The result is similar to the high-level architecture of the Multimission Algorithm and Analysis Platform (MAAP), a new cloud-based joint NASA-ESA project that will support the ESA BIOMASS and NASA NISAR and GEDI missions. A second workshop will refine these ideas and assess practical ideas on implementation and its challenges. In this paper we will explore the needs for an EO-based global biodiversity monitoring system and tie them to the opportunities that the rapid changes in technology and science provide.

This work had its inception at the “Unlocking a New Era in Biodiversity Science: Linking Integrated Space Based and In-Situ Observations” study funded by the W.M. Keck Institute for Space Studies, and the research described in this paper was in part carried out at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA; government sponsorship acknowledged.

Area C - Board 297

The incremental surge of remote sensing data has encouraged the phenomenal growth of Earth Observation based applications addressing the needs of diverse communities. This incredible growth comes with the increasing complexity of discovering and processing these large datasets in a scalable, robust and reliable service-oriented platform infrastructure. NextGEOSS platform is committed to address the challenges encountered by the Earth observation driven applications by offering a suite of innovative services. The NextGEOSS is incubating a set of pilot applications to build capability and mature these services with the good level of granularity and tailorability to be reused by several pilot applications. NextGEOSS is encouraging other European Earth Observation applications and projects to extract the benefits from these NextGEOSS platform services to achieve scalability and mature into an operational phase. The users are on-boarded with a 5 step user experience starting from the user engagement, followed by a meticulous analysis and preparation of the data, setting up of the platform and the integration of the application on the platform and with the useful NextGEOSS services. The applications enter then into operational phase after a successful integration.

The NextGEOSS project presents a new European vision of GEOSS data exploitation for innovation and business. It focuses on a fundamental change, where we optimize the connectivity of the European and global Earth observation data centers with new discovery and processing services. The NextGEOSS platform offers easy and seamless access to all relevant data repositories, as well as efficient operations (search, retrieval, processing/re-processing, visualization and analysis). It is an Earth observation data hub that helps develop and deploy new services requiring a wide variety of data sources.The NextGEOSS platform with its services creates a solid foundation for capacity building. The platform services comes with relevant step-by-step capacity building activities that also can be transformed and used to support the various domain specific GEOSS community platforms,

The NextGEOSS Data Discovery service facilitates and harmonizes the access to diverse data sources with continuous and optimized harvesting of the metadata to ensure that the NextGEOSS pilot applications enjoy the availability of latest datasets to satisfy different operational scenarios. This service is complemented with a quality of service enforcing reliability and a very high quality of service level by automatic detection of unavailability of the data providers and performance-bottlenecks related to data access providing chance to perform remedy actions proactively.
The NextGEOSS data processing service facilitates the integration of data processing algorithms onto the NextGEOSS Cloud platform and also to configure community portals using standard geospatial web services. This service offers a collaborative workplace to co-design NextGEOSS pilot applications. Application developers have the flexibility to create and manage containers and the virtual machines supporting the integration and deployment in any federated cloud infrastructure.
The NextGEOSS user management (UM) service provides NextGEOSS users with Single-Sign-On (SSO) authentication for accessing GEOSS data and services in a federated System of Systems environment. It Provides SSO capability that enables a registered user to log in once, and access multiple GEO/GEOSS applications where the user is signed-up already, without being required to authenticate for each application separately. The service supports the integration of other GEO-related user management systems based of diverse protocols.
The NextGEOSS analytics service collects and analyzes performance health indicators of the services (CPU, Memory, Disk usage) using Big Data analytic engines to assess, report and notify the users of the real time performance of the applications and infrastructures. This service also continuously extracts critical analytics from various data providers for the users to choose the preferred data providers based on selected criteria (response speed, download time etc.). It provides a critical service to go towards operations.
The user feedback service developed in NextGEOSS provides users an open and collaborative approach to share feedback on any resources. This is of paramount importance as it is a direct mechanism to raise users’ voices and promote interaction between users and data providers. Users will be more easily engaged with GEOSS activities if they can see a real opportunity to create a community and establish social links on the data hub around the data resources, they are interested in. The data producers may take advantage of this situation, being able to respond to users’ demands, in creating new versions of the resources or addressing their concerns as new feedback items. This service capitalizes on previous European projects and is aligned with current available standards.

NextGEOSS services are enabling the NextGEOSS applications to reach their full potential by facilitating access, processing and other supporting services in an open and collaborative manner. These services are based on standards and made available “a la carte” for the integration of new applications into the European GEOSS for the benefit of the global Earth observation community.

Area C - Board 298

The Centre for Environmental Analysis (CEDA) delivers long term curation of scientifically important climate and Earth Observation data and facilitates the use of data by the UK and European environmental science community. CEDA currently holds over 6PB of EO data and this archive includes data from the Sentinel and METOP series of satellites as well as MODIS data from Aqua and Terra. The archive also includes ATSR1/2 and ENVISAT data as well as airborne and in-situ datasets and related derived products. These EO datasets compliment other satellite derived climate data as well as UKMO data in addition to CMIP5 and forthcoming CMIP6 data. CEDA also provides access to the research outputs from the UK National Centre for Earth Observation (NCEO) as well as the National Centre for Atmospheric Science (NCAS) and the Natural Environment Research Council.

CEDA users can access this wealth of data by using the JASMIN “super data cluster” co-located with CEDA at STFC Rutherford Appleton Laboratory. Users of the JASMIN e-infrastructure can access the multi-Petabyte fast storage and excellent parallel processing data analysis capabilities upon the archive. JASMIN is enabling our community to create long term large scale datasets as part of international initiatives such as the ESA Climate Change initiative (CCI) and the Copernicus Climate Change service (C3S). These diverse datasets range from lake surface temperatures (Sentinel 3) to analysis of geohazards (Sentinel 1). All these derived data products depend on the timely retrieval of large volumes of data from ESA, NASA and EUMETSAT amongst numerous other sources to the CEDA archive.

CEDA constantly strives to improve access by providing a range of methods to query, find and access the data archive. The Satellite Data Finder (SDF) allows users to find and download contemporaneous data from a range of EO products on the archive. The SDF also provides an OpenSearch interface to the catalogue allowing users to gain access either directly on JASMIN or via a range of download endpoints such as HTTP, FTP, OpenDap or GridFTP. The SDF is complimented by the Flight Finder providing a similar interface for the FAAM. CEDA also provides access to a detailed top level metadata catalogue and hese metadata services support CEDA involvement in a variety of projects.

Area C - Board 299

Satellite-based geodetic data are critical for measuring the accumulation of elastic strain across active geologic structures, which is increasingly important for assessing the seismic hazard associated with the eventual release of this strain, and for monitoring volcanic deformation, which can be a key indicator of the risk of eruption. In many parts of the world current strain rate estimates rely primarily on Global Navigation Satellite System (GNSS) networks, with sparse and heterogeneous spatial coverage limiting both accuracy and resolution. The rapid improvement in Interferometric Synthetic Aperture Radar (InSAR) data quantity, quality and processing techniques now enables routine measurements of ground displacements at different spatial and temporal scales: from motions across entire tectonic plate boundary-zones to volcanic unrest prior to an eruption.

Within the NERC-funded LiCS project (Looking inside the Continents from Space), the Centre for the Observation and Modelling of Earthquakes, Volcanoes, and Tectonics (COMET) is developing the LiCSAR platform consisting of algorithms for an automated generation of interferograms and derived products, including velocity and strain rate fields. The platform is based on automatic generation of Sentinel-1 (S-1) interferograms at a relatively high spatial and temporal resolutions with the goal of detecting interseismic deformation across faults with slip rates in the order of a few millimeters per year as well as monitoring activity at the 1,300 active volcanoes around the world. LiCS products are provided in open access to the geoscience community via an online portal (http://comet.nerc.ac.uk/COMET-LiCS-portal/).

To date, our online portal includes around 100,000 interferograms and derived products. Our recent research includes activities towards a joint application of S-1 InSAR and GNSS data to measure strain rates in the order of 10e-7 to 10e-8 per year at active fault systems across the Alpine-Himalayan belt; the work includes generation of global maps of motion velocities, e.g. over Turkey. On volcanoes, we have produced with LiCSAR about 30,000 interferograms on 900 different volcanoes, and we have started to use machine learning techniques to automatically detect ground deformation signals on our dataset. Preliminary results are promising; our system was able to detect 2017 unrest signals at Cerro Azul and Etna.

Our final aims are to deliver an early detection system capable to track volcanic deformation in near real-time (with a user interface that enables scientists in volcano observatories to use the data for monitoring purposes) and to generate and automatically update maps of average velocity of the tectonic motion including time series estimation and visualization.

This contribution will demonstrate our current progress in the data processing approach, results to date, and some of the challenges and technical solutions associated with reduction of atmospheric delay errors in the interferograms (applying corrections based on GACOS system), deriving large-scale crustal velocity and strain-rate fields at the plate tectonic scale, and the automatic detection of volcanic displacements.

Area C - Board 300

Interferometric Synthetic Aperture Radar (InSAR) is widely used for generating deformation maps. However, often there are several driving mechanisms behind the estimated deformation time-series. For example, there may be shallow and deep mechanisms, where the shallow are caused by changes in groundwater level, the dynamics of structures, polder drainage or natural compaction of shallow layers, while deep deformation can refer to mining, gas, salt, or oil extraction. Disentangling the different deformation regimes is not trivial, and currently typically the challenge in time-series InSAR processing.

Using gridding, here we not only assign a velocity to each grid cell, but we analyze the entire probability density function of the deformation behavior, estimated via histograms of, e.g., velocity for each grid cell. This enables us to estimate the distribution and basic summary of statistics such as mean, median, dispersion, skewness, kurtosis, and percentiles. From the central limit theorem, we expect a near-Gaussian distribution if the distribution stems dominantly from noise in the observations. Significant deviations from this Gaussian are a indication of the possibility of a superposition of different driving mechanisms in estimated deformation. For example, the skewness can be an indicator of multiple superposed driving mechanisms.

Here we analyze the options for reliably estimating skewness for grid cells with varying sizes, varying number of points and the consider the effect of the spatial variability of the signal within the grid cell. We emphasize the estimation of the precision of the skewness, as an important quality metric for this approach for disentangling superposed deformation signals, and provide a multi-resolution quadtree-based scheme for optimal estimation.

We apply our method for deformation time-series resulting from InSAR processing of six Sentinel-1 stacks over the Netherlands for more than three years time-period. The result of this procedure provide an improvement to the national deformation of the Netherlands.

Area C - Board 301

Earth’s surface deformation that occur as a consequence of an earthquake is a crucial information for investigating the causative source of the seismic event. In this context, the space-borne Differential Synthetic Aperture Radar Interferometry (DInSAR) has proven to be one of the key methods for the quantitative measurement of the Earth’s surface deformation, with centimetres to millimetres accuracy [1]. DInSAR relies on the evaluation of the phase difference between two SAR images, acquired from different orbital positions and at different times [1]. Depending on the system configuration, the footprint of space-borne SAR acquisitions can span from a few kilometres up to hundreds of kilometres, making it particularly suitable for accurate investigations of wide areas at relative low cost. In these sense, according to USGS records [2], from 1992 to 2016, about 3700 earthquakes with significant magnitudes (Mw > 6.0) have occurred, while only a limited number of them has been successfully investigated through DInSAR [3]. This is mainly due, apart the intrinsic limitation of the DInSAR technique, to the lack of a satellite program with a systematic and global acquisition policy, which are fundamental characteristics to allow creating DInSAR operational services at global scale. However, since the launch of the Copernicus Sentinel-1 SAR satellite missions in 2014 and 2016, the availability of SAR images dramatically increased. Indeed, this constellation acquires, with global coverage policy, radar images every 6/12 days over the same area, allowing us to dispose of a huge archive of SAR data that can be processed for obtaining co-seismic displacement maps in a short time frame and anywhere in the world.

Considering the relevance of the satellite interferometric analysis for the hazards monitoring, as well as the availability of new radar systems as Sentinel-1, which are characterized by a high reliability level, is it therefore possible the development of operational services for the generation of DInSAR products, some of them being already in place [4, 5]. In this work an unsupervised and automatic tool for the generation of DInSAR co-seismic displacement maps is presented. Benefiting from the mostly global availability of Sentinel-1 SAR data and the on-line earthquake catalogues, the tool retrieves information about the depth and magnitude of recent earthquakes and triggers, if necessary, the interferometric process over the area affected by the seismic event.

The workflow process is the following (Figure 1). First, the extraction of earthquake information (epicenter location, magnitude, time, ...) from the on-line public available web catalogues, as those provided by main international geophysical institutions (e.g. USGS [2], INGV [6]), is performed (Block A of Figure 1). The retrieved information is provided according to different standard formats (QuakeML, geoJSON, ...) and is accessible via subscription feeds that are updated with a defined frequency. The system is not limited to a single earthquake catalog interface. The relevant earthquake information is collected in accordance to an empirical magnitude and depth relation, which considers that only high magnitude (> Mw 6.0) and relatively shallow earthquakes (typically < 20 km) very likely induce a surface deformation that is detectable via DInSAR [7] (Block B). Among the earthquakes that respect the relation, only those with the epicentre on land (or even on water but that can likely induce detectable deformation on land) are processed.

Once the occurred earthquake has been selected, the SAR data retrieval is performed via an automatic query to the open access Sentinel-1 catalogue (Block C). The query is performed over an area whose extension depends of the relation between magnitude, depth and epicenter location, which is derived from theoretical and empirical considerations and is susceptible of further tuning and refinement. Once all the tracks covering the earthquake area have been identified, the system retrieves all the available SAR Sentinel-1 data (from both ascending and descending passes) up to 30 days before the event (or at least 1 pre-event image even in a larger time span), in order to allow the generation of the co-seismic interferograms. The data retrieval, and accordingly the subsequent DInSAR processing, remains active up to 30 days after the event.

Once the data are downloaded, they are processed through an efficient DInSAR algorithm [8] (Block D). According to this scenario and taking benefit from the operational capability of the Sentinel-1 constellation, the processing of the different tracks can be carried out in parallel, while actually their execution depends on the available computing resources and on the effective temporal acquisition of the SAR data. A processing prioritization of the different tracks on the basis of the post-event acquisition time has been implemented (according to a First come-First served policy).

The tool provides wrapped interferograms and displacement maps (unwrapped interferograms converted in centimetres) in the satellite Line of Sight (LOS). The output data are provided according to the specification of the European Plate Observing System (EPOS) [9] research infrastructure, and will be made openly available through the EPOS portal, to be investigated and interpreted by the scientific community. The system has been implemented on in-house computing facilities and has been tested through a controlled experiment with several significant earthquakes.

Although tested with Sentinel-1 data, the implemented tool is independent from the exploited SAR acquisitions, thus increasing the number of data to be processed. Indeed, the only dependency is on the catalog interface that, if does not respect an Open standard, requires the implementation of an appropriate wrapper. It is also worth noting that the presented tool, since it takes benefit from efficient and scalable DInSAR algorithms, can be exploited to perform large processing campaigns of all the co-seismic DInSAR pairs acquired by the Sentinel-1, and even ERS and ENVISAT, since their respective launch. To do this, disposing of proper computing facilities, such as those provided by the DIAS [10] platforms where data and processing are co-located, is strongly envisaged.

Bibliography
[1] Massonnet, D. et al., “The displacement field of the Landers earthquake mapped by radar interferometry,” Nature, vol. 364, no.6433, pp. 138– 142, Jul. 1993.
[2] USGS. United States Geological Survey, Earthquakes hazard program, https://earthquake.usgs.gov/earthquakes/feed
[3] Funning, G., Garcia A. 2018 “A systematic study of earthquake detectability using Sentinel-1 Interferometric Wide-Swath data”, Geophisical Journal International, publish in: https://doi.org/10.1093/gji/ggy426.
[4] SARVIEWS.A, SAR-based hazard monitoring service, http://sarviews-hazards.alaska.edu
[5] COMET, COMET, Centre for the Observation and Modelling of Earthquakes, Volcanoes and Tectonics, Lics. http://comet.nerc.ac.uk/COMET-LiCS-portal/
[6] INGV, National Institute of Geophysics and Volcanology, http://cnt.rm.ingv.it/feed/atom/all_week
[7] Dawson, J., and P. Tregoning (2007), Uncertainty analysis of earthquake source parameters determined from InSAR: A simulation study, J. Geophys. Res., 112, B09406, doi:10.1029/2007JB005209.
[8] Casu et al. 2014 SBAS-DInSAR Parallel Processing for Deformation Time-Series Computation” in IEEE journal of selected topics in applied earth observations and remote sensing, VOL. 7, NO. 8, AUGUST 2014.
[9] EPOS, European Plate Observing System, https://www.epos-ip.org/tcs/satellite-data.
[10] DIAS. The upcoming Copernicus Data and Information Access Services (DIAS), http://copernicus.eu/news/upcoming-copernicus-data-and-information-access-services-dias

Area C - Board 302

Damages produced by earthquakes are caused by permanent surface deformation due to fault offset, as well as dynamic shaking due to the pass of seismic waves. The investigation of the relationship between this two physical processes may provide a better understanding of earthquake hazard. To this aim, complete and homogeneous catalogs are necessary. Regarding dynamic changes, data are constantly acquired by seismic stations installed on the ground surface, and the so called “ShakeMaps” are generated after a seismic event by relying on their measurements. Instead, co-seismic deformation can be measured with different geodetic techniques. For example, surface deformation maps can be generated via Differential SAR interferometry (DInSAR). The DInSAR analyses performed so far have consolidated this technique as extremely important to map and characterize co-seismic deformation. However, DInSAR analyses are generally performed for specific cases, by relying on different software solutions, as well as different processing parameters. Thus, results are often not complete and/or extremely heterogeneous. The advances on cloud-based processing platforms allows nowadays to process (and/or re-process) a large amount of SAR data available at global scale to generate homogeneous results in a relatively short time.
In this work, we show a cloud-based implementation designed to produce a homogeneous catalog of co-seismic deformation and changes in SAR coherence associated with earthquakes with Magnitude larger than Mw 5 by relying on different SAR datasets. The seismic events are selected over a specific time windows and area of interest are by relying on the USGS global catalogue. The final goal of this study is to perform a back analysis on a statistically relevant number of events to derive empirical relationships and descriptors between earthquake physical parameters (e.g., Magnitude, depth, fault plane dimensions, rupture mechanism, etc.) and the co-seismic deformation observed at the surface (e.g., amplitude and extension). Details on the implementation and preliminary results on test areas will be presented. This work is done in the framework of European Commission's Horizon 2020 project "BETTER", with the main objective to facilitate the usage of large volume and heterogeneous datasets by downstream users, so that they can focus on the analysis of the extraction of the potential knowledge within the data and not on the processing of the data itself. More information is available on the website https://www.ec-better.eu/.

Area C - Board 303

Space-borne observation is becoming more and more crucial for the analysis of Earth surface and subsurface processes. In this context, Synthetic Aperture Radar (SAR) imagery is an important source of information, at different spatial and temporal scales. For example, SAR imagery helps geoscientists to study the co-seismic and post-seismic effects of earthquakes, to analyze volcanic unrest and eruptions, to observe displacements due to landslides and land subsidence, and to monitor the stability of infrastructures. While topics on the exploitation of other types of satellite imagery (e.g., optical/thermal/multispectral) are integrated in geoscience teaching programs at different levels, SAR is traditionally constrained to the remote sensing community. The reason is mainly because a strong engineering and IT backgrounds are necessary to fully exploit this technology.
To fill this gap, we developed at ETH Zurich a pilot course tailored to master students in Earth Sciences, aiming at training the new generation of geoscientists to fully take advantage of SAR data. The course offers the chance to learn a cutting-edge technique and to independently apply the methods to real scenarios relevant for their future activities as scientists and/or practitioners. Students learn background information on SAR through a mix of frontal lectures, games, and develop independently projects on real case-studies during workshops by leveraging on open source data and software. The students’ performance is assessed by peers and by an international steering committee during a mini-conference.

Area C - Board 305

Large and permanently moving landslides are widespread in many landscapes of the Alpine Space, with significant impacts on erosion, sediment transfer and human settlements. While in situ geophysical methods and terrestrial remote sensing are indispensable for a detailed monitoring and understanding of individual landslides, their area-wide mapping and monitoring is still challenging. SAR interferometry has proven useful for the detection and monitoring of very slow movements (< 1 m.yr-1) but limitations are encountered for the investigation of slow-moving landslides (1 m.yr-1 – 30 m.month-1). Such limitations can be addressed through the analysis of time series of optical remote sensing images.
We implemented the Multiple Pairwise Image Correlation technique -MPIC- for the analysis of nearly two years of Sentinel-2 datasets (28 tiles) covering the entire Alpine Space. The processing technique generates stacks of partially redundant horizontal displacement fields and computes multi-temporal indicators for a detection and quantification of surface displacement. The MPIC service (parallelized algorithm, processing chain, user modes) will be presented in detail. Wide-area (region) and local (slope) examples of processing will be presented. The accuracy of the derived inventories and displacement time-series and their implications for the understanding of the seasonal landslide dynamics are discussed.
The processing technique is implemented both on the A2S processing and service platform of University of Strasbourg and on the ESA Geohazards Exploitation Platform (GEP).

Area C - Board 306

In 2014, the European Space Agency (ESA) has started the Earth Observation (EO) Exploitation Platforms (EPs) initiative, a set of R&D activities that in the first phase aims to create an ecosystem of interconnected Thematic Exploitation Platforms (TEPs) on European footing.
The Geohazards Exploitation Platform (GEP, https://geohazards-tep.eo.esa.int/#!) is an activity originated by ESA in November 2015 to directly support the geohazards community exploiting satellite EO data to assess geohazards. The GEP supports data storage and dissemination of many different EO resources (e.g. Synthetic Aperture Radar (SAR) missions, Very High Resolution Optical missions); provides on-line hosted processing for geohazard assessment (e.g. terrain motion monitoring based on InSAR or stereo-optical data); and promotes and strengthens collaborative work (e.g. sharing results with the community) and community building.
It includes six partnerships for pilot services with TRE Altamira, CNR-IREA, DLR, INGV, ENS-CNRS/NOA and EOST-CNRS. In parallel, a number of users have been invited or selected in order to test the platform services under one of the following scenarios: (i) EO data exploitation, (ii) new EO service development or (iii) new EO product development. 
More than 20 on-demand services using both Optical & SAR data are available on the recently released version of GEP (v2), such as for instance the SAR based SBAS multi-baseline processing chain of CNR IREA for InSAR measurements and the MPIC-OPT chain of CNRS EOST for Optical measurement using Sentinel-2 & VHR data. 8 systematic services delivering continuously updated information layers on GEP, including DLR's large scale production of Sentinel-1 InSAR browse images at both 100m and 50m resolution over tectonic regions. New basic services providing full resolution GeoTIFF imagery (e.g. RASTER on-demand service process many EO missions data) and change detection imagery (e.g. the SNAP based COIN & SNAC tools) for rapid on-line visualization are also available on GEP.
The GEP primarily focuses on expert users i.e. InSAR specialists and end-users i.e. geological surveys and mandated Disaster Risk Management (DRM) organisations. Today, there are more than 80 users, primarily from the public sector in Europe, Asia, Africa and the Americas. Furthermore, users of the Committee on Earth Observation Satellites (CEOS) Working Group on Disasters (http://ceos.org/ourwork/workinggroups/disasters/) and the Geohazard Supersites and Natural Laboratories (GSNL http://www.earthobservations.org/gsnl.php) community have been experimenting with the GEP functionalities such as data delivery, on-line processing and e-collaboration since 2015.
Looking forward, the GEP has been selected as the gateway for the Satellite Data Thematic Core Service of the European Plate Observing System (EPOS) and contributes to the CEOS Geohazards Lab initiative, a number of interoperable platforms with federated resources from CEOS agencies (e.g. ASI, CNES, DLR and ESA) and community partners (e.g. CNR IREA) that provides data delivery mechanisms, on-line processing tools and services and an e-collaboration environment to exploit EO data to assess geohazards and their impact. The Geohazards Lab will support primarily, the CEOS and GSNL users, but also users from the entire geohazards community. 

Area C - Board 307

Since January 2017, operational monitoring of 22 worldwide volcanoes has been available on the Geohazards Exploitation Platform of ESA (GEP), thanks to the activities conducted in the VEGAN project and in partnership with the GEP project, an activity originated by ESA concerning hosted processing for risk management (https://geohazards-tep.eo.esa.int).
The VEGAN project, led by NOVELTIS in partnership with the Italian Istituto Nazionale di Geofisica e Vulcanologia (INGV) has been supported by ESA under the Value Adding Element (VAE) program. VEGAN offers operational decision-making services to monitor the volcanic eruptions and study their impact on agriculture and vegetation. All eruption stages are considered: early detection, crisis, and post-crisis. For each stage, innovative products derived from satellite data are available.
VEGAN services benefits from the GEP pre-operational facilities: High IT capacities and state-of-the-art platform allowing massive and automatic data processing, as well as useful visualisation and download functionalities. GEP is part of the Thematic Exploitation Platform offered by ESA to bring the user closer to new EO-based products and services.
The VEGAN products that are available on GEP are as follows:
• Hot Spots Detection Maps for monitoring active eruptions and burnt areas as lava flow, lava lakes or fires during and after the eruptions. This product is continuously available at high resolution (20 meters) and relies on the use of Sentinel-2 data during and after a crisis. With Hot Spots Detection Maps, it is possible to highlight the high thermal event related to volcanic activities in not monitored remote areas, as in Turrialba (Costa Rica) in January 2017, and hence prevent major events. By means of this product it is also possible follow the lava flows and burnt areas evolutions.
• Surface Temperature Maps (STEMP) before, during and after a crisis. This product is also available continuously by using Sentinel-2, Sentinel-3 and Landsat-8 data, which offers both spatial and temporal coverage of the regions of interest.
• Vegetation Vigor Maps systematically derived from Sentinel-2 at high resolution (10 meters) over the monitored regions. The service offers also the possibility to perform multi-temporal comparisons of vegetation vigor maps to evaluate the impact of lava or volcanic ash on surrounding vegetation and agriculture production. For example this product shows the impact of the eruption of Turrialba volcano in January 2017 on the vegetation. One can see that two months later, the vegetation is partially back.
All VEGAN products have been successfully tested according to the end-users in different countries and landscapes, i.e. tropical forests, islands, temperate areas, deserts, boreal regions…
The VEGAN services are suitable for operational deployment worldwide and can be provided on demand in order to monitor volcanic eruptions and their impact on agriculture and vegetation.

Area C - Board 308

To address the needs of the Haitian community involved in recovery and rehabilitation after the impact of Hurricane Matthew in October 2016, the Committee on Earth Observation Satellites (CEOS) triggered the 4 year-long Recovery Observatory pilot project, led by the National Center for Geo-spatial Information (CNIGS) of Haiti with technical support from CNES [www.recovery-observatory.org]. In this work, we present a number of scientific experiments carried out within ESA’s Geohazards Exploitation Platform (GEP) to process big SAR data stacks acquired by the Sentinel-1, COSMO-SkyMed and TerraSAR-X constellations to retrieve useful information to support land planning and monitoring recovery in the Haitian departments most affected by Hurricane Matthew, i.e. Grand’Anse, Sud and Nippes. The cross-agency research team of ASI, DLR, ESA and CNES with support from ARGANS, Terradue and Athena Global, has started to focus on the development of a SAR-based workflow, which encompasses steps ranging from satellite data acquisition, to image processing in GEP and generation of value-added geohazard products. The experiments in GEP are based on available hosted processing services capable to generate SAR change detection products (e.g. radar amplitude change and coherence maps), as well as Interferometric SAR (InSAR) ground motion products (e.g. differential interferograms, annual deformation velocity maps and deformation time series) with Persistent Scatterers (PS) and Small BAseline Subset (SBAS) methods. In addition to the wide-area coverage regularly provided by Sentinel-1 in C-band, a tailored SAR acquisition campaign in X-band has been tasked to support disaster recovery. Since the end of 2016, DLR provides full coverage of the region of interest every ~4 months with a mosaic of 3 m resolution TerraSAR-X StripMap scenes and, for three hotspots selected by the stakeholders, since the end of 2017 ASI is acquiring a bespoke 1 m resolution COSMO-SkyMed SpotLight time series with 16 days revisit. Our experiments in GEP prove the value of thematic exploitation platforms available or currently being developed to process big SAR data to efficiently generate derived products useful to support geohazard analysis and management. In Haiti, in particular, the retrieval of detailed, multi-sensor and multi-scale geohazard information from SAR-based products generated in GEP will be valuable to help the end-users in their decision-making processes and recovery progress monitoring after disasters.

Area C - Board 309

Urbanization and climate change, representing two of the most relevant global trends related to the human presence on the planet, challenge our environmental, societal and economic development. The availability of and access to accurate, detailed and up-to-date information will impact decision making processes all over the world. The suite of the Copernicus Sentinel satellites in combination with their free and open access data policy contributes to a spatially and temporally detailed monitoring of the Earth’s surface. At the same time a multitude of additional sources of open geo-data is available – e.g., from other earth observation (EO) missions (Landsat, etc.), from national or international statistics or land surveying offices, volunteered geographic information or social media. Detailed and reliable information on global human settlements can directly contribute to the monitoring and decision making regarding the 2030 Agenda for Sustainable Development and provides relevant data for the Sustainable Development Goals (SDG), specifically SDG 11, as well as other global and national policy frameworks. However, the capability to effectively and efficiently access, process, and jointly analyze the heterogeneous and often large-volume data collections poses a key technical challenge.

The Urban Thematic Exploitation Platform (U-TEP) is developed to provide end-to-end and ready-to-use solutions for a broad spectrum of users (scientists, researchers, planners and policy-/decision makers) to extract unique information/ indicators required for urban management and sustainability. Key components of the system are an open, web-based portal connected to distributed high-level computing infrastructures and providing key functionalities for:

i) high-performance data access and processing,
ii) modular and generic state-of-the art pre-processing, analysis, and visualization,
iii) customized development and sharing of algorithms, products and services, and
iv) networking and communication.

U-TEP aims at opening up new opportunities to facilitate effective and efficient urban management and the safeguarding of livable cities by systematically exploring the unique EO capabilities in Europe in combination with the big data perspective arising from the constantly growing sources of geo-data. The capabilities of participation and sharing of knowledge by using new media and ways of communication will help to boost interdisciplinary applications with an urban background. The services and functionalities are supposed to enable any interested user to easily exploit and generate thematic information on the status and development of the environment based on EO data and technologies. In order to assure sustainable developments, U-TEP follows an Open Source strategy committing that the platform is based on Open Source components and modules, with an online presence at the world's leading software development platform GitHub (https://github.com/urban-tep).

U-TEP provides access to various new products, such as the TimeScan layer, geotagged Tweets, the Global Urban Footprint (GUF) and its successor the World Settlement Footprint (WSF), as well as datasets of other research groups (Nighlights, WorldPop, etc.). Moreover, U-TEP provides processing services that the users can deploy applications to conduct analyses or generate certain products on-demand. U-TEP’s TimeScan on-demand service, for instance, has already been used to successfully generate three global TimeScan layers for 2015, 2010 and 2000 from a total of more than one million Landsat images (Esch et al. 2018). The production was solely based on an implementation of the TimeScan on-demand service at the HPC cluster of U-TEP’s IT4I processing center. The resulting TimeScan-Landsat layers are available at U-TEP’s Geobrowser (https://urban_tep.eo.esa.int/geobrowser/?id=portfolio#!&context=TimeScan%2FLandsat).

The Visualization and Analysis Tool (VISAT) allow experts and non-experts to conduct geospatial analyses and queries. Figure 1 illustrates the development of settlement area for the greater Ho Chi Minh region in Vietnam, based on the WSF Evolution product for the years 1985, 1995, 2005 and 2015. The charts show multisource and multitemporal statistics for selected countries in SE-Asia.

U-TEP recently successfully completed its development phase and entered a pre-operational phase. The user feedback and the lessons learnt from the 8-months pre-operations phase, identified some key properties and functionalities that can be considered as major determinates for the usability, benefit and attractiveness of an instrument such as U-TEP. A systematic enhancement and streamlining of the platform capabilities and product/service functionalities will be conducted until Q2 2019 and will enable the operational phase of the platform. However, the innovative character of U-TEP platform in terms of available data and processing and analysis functionalities attracted already a large user community (>340 users and institutions from >40 countries) of diverse users (i.a. from science, public institutions, NGOs, industry) so far.

Area C - Board 310

The THEIA Land Data and Services Centre (www.theia-land.fr) is a consortium of 12 French public institutions involved in Earth observation and environmental sciences (CEA, CEREMA, CIRAD, CNES, IGN, INRA, CNRS, IRD, Irstea, Météo France, AgroParisTech, and ONERA). THEIA has been initiated in 2012 with the objective of increasing the use of space data by the scientific community and the public actors.
THEIA structured the French scientific community 1) through a mutualized Service and Data Infrastructure (SDI) distributed between several centers, allowing access to a variety of products; 2) through the setup of Regional Animation Networks (RAN) to federate and animate users (scientists and public / private actors) and 3) through Scientific Expertise Centres (SEC) clustering virtual research groups on a thematic domain. One of this SEC is the "Urbanization and Artificialization” Centre clustering experts in multi-sensor urban remote sensing. THEIA in collaboration with ODATIS (Data and Service for the Ocean), ForM@Ter (Data and Service for the Solid Earth), and AERIS (Data and Service for the Atmosphere) form the "Earth System" Research Infrastructure.
The objective of this poster is to present recent (>2016) innovations of the URBAN SEC in terms of 1) development of algorithms useful for urban remote sensing using optical and SAR sensors, 2) validation of the urban products provided by the THEIA Land Service and Data Infrastructure, and 3) demonstration of user-tailored applications for urban studies. The Urban Expert Centre brings together researchers and engineers from several institutes: LIVE - Strasbourg, IGN-LaSTIG - Univ. Paris Est, CESBIO – Toulouse, LETG - Rennes, Irstea – Montpellier, TETIS - Montpellier, INP Bordeaux, IRD, ESPACE-DEV - Montpellier, ESPACE - Nice, ONERA. Research results and methods linked to (1) the detection and mapping of the urban footprint at an annual frequency; (2) the identification of urban fabrics, (3) the mapping of green networks within the cities. In parallel, the group proposes to summarize and identify relevant indicators (parameters) dedicated to urban planning and management.

Area C - Board 311

Phenology is the science that deals with the study of the timings of periodic plant and animal life cycle events and how these are influenced by environmental conditions. These timings, which vary from year to year and from place to place, are excellent indicators of climate change. The main sources of phenological data are (volunteered) ground observations, cameras pointing at vegetated canopies and Earth observation sensors. Time series of images acquired by the latter sensors are often used to retrieve land surface phenological metrics, or phenometrics, by analyzing the dynamics of vegetation indices such as the normalized difference vegetation index (NDVI) or the enhanced vegetation index (EVI). Here we focus on the validity and coherence of one of these phenometrics, namely the Start of Season (SOS). This phenometric represents the start of the green up period of vegetated canopies after their “winter break”, and it can be calculated from either NDVI or EVI. Since we are interested in performing our analysis at large-scale, we designed and implemented a Spark-based platform that uses Hadoop for storage and Jupyter notebooks for the analysis of the data. TIMESAT, a well-tested and known software to retrieve phenometrics, is used to analyze the time series of vegetation indices. Our platform was tested using data over continental USA. In particular, we used the full Copernicus Global Land Service dataset, which combines all available decadal composites from SPOT-VEGETATION (1998-2014) and PROBA-V (2014-2017). We compared the SOS products retrieved using the full time series of NDVI and EVI images, and three TIMESTAT fitting functions: Asymmetric Gaussian, Savitzky-Golay and Double Logistic. The validity of these SOS products was subsequently evaluated using volunteered phenological observations from the USA National Phenological Network. Results show that there are significant differences (more than two months) in the SOS values depending on the vegetation index and the fitting function used. Our work confirmed the difficulty of linking land surface phenometrics with ground observations because we found root mean square errors of about 3 to 4 weeks between the various SOS products and the volunteered observations. Further work will focus on the use of our platform to efficiently retrieve phenometrics from Copernicus (e.g., Sentinel 2 and 3) data.

Area C - Board 312

We will present the Ag|knowledge platform, an operational, cloud-based, big EO data processing & analytics platform designed for the easy integration of EO data & analytics into operational IT systems in the agricultural domain. We will illustrate the working paradigm of this system by explaining its contribution to the BigDataGrapes platform – a big data platform for the Grapevine-powered industries.

There is demand to embed EO analytics into operatonal working practices from a wide variety of application domains. In a lot of these domains operational IT systems exist that handle spatial data and could benefit from EO analytics. Examples are administration systems that deal with the CAP regulatory frameworks (e.g. IACS), Farm Management Systems, or financial and insurance systems. Typically the system operators would benefit from the integration of EO data & analytics but do not want to care about EO data storage, processing or management due to always increasing data amounts and the lack of EO expertise.

This is what the Ag|knowledge platform was designed and built for. It was launched in 2015 and is designed as an easily integratebale cloud-based, big EO data processing & analytics platform. It handles EO data storage, processing, management & analytics and is easily integrateable into any IT system through APIs. It works object based and delivers timeseries of cloud free, readily processed satellite images and statistics for registered objects (e.g. agricultural fields). Therefore the client system doesn’t have to care about satellite observation dates, mosaicking data or cloud masking. On top of that advanced analytics are available that enable the client to analyse the data with time series based algorithms, perform object comparisons, do classifications or find anomalies in the data sets with machine learning and other techniques. Thereby the data volumes that need to be transferred, stored and analyzed decrease from Terabytes to Kilobytes which eases data handling for the client system.

Ag|knowledge currently delivers information based on Copernicus Sentinel & Landsat data and is operationally used by a variety of IT systems around the world, e.g. IACS systems in several EU member states, Farm Management Systems and financial or insurance service systems.

In this presentation we will use the BigDataGrapes platform, a distributed, cloud based big data analytics platform that utilizes Ag|knowledge as an example application to explain its working principle. The BigDataGrapes platform specifically tries to help companies across the grapevine-powered value chain ride the big data wave, supporting business decisions with real time and cross-stream analysis of very large, diverse and multimodal data sources. It connects, visualizes and allows the combined analysis of EO data with ground based observations (e.g. weather, soils, farm management) and large amounts of metadata about e.g. grapevine varieties, wine & table grapes quality or economic figures. The BigDataGrapes platform is currently being set up in the frame of the H2020 project BigDataGrapes.

Area C - Board 313

Ending hunger and malnutrition caused by a lack of steady access to nutritious food is the prominent goal number two of the UN Sustainable Development Goals. The challenge of increasing food supply to feed a growing global population puts sustainability of agriculture and aquaculture on a level with ensuring food security. EO data play a major role in understanding and managing the global food supply on all scales, for instance using Sentinel-2 data to extract information about crop status, from chlorophyll content to leaf area to water content, and monitor growth trends. Specialised service providers have developed algorithms to extract the most valuable information for farming decision making, e.g. regarding optimal usage of fertilisers, or to be able to predict crop growth or even to compare different productivity across fields. This information is mostly extracted using time series analyses and often needs to be assessed over large areas, e.g. to cover a whole country.
So, the big data challenge consists in:
• Easily identifying and accessing relevant data for a given region and time range;
• Elaborating such amount of data with the chosen processing functions rapidly and at minimum processing cost;
• Performing post-processing analytics;
• Flexibly compose the above activities in customised workflows.

The Food Security Thematic Exploitation Platform (FS-TEP) ESA project, has developed a Platform as a Service (PaaS) which addresses all of the above, exploiting an Infrastructure as a Service (IaaS) to manage access to cloud processing resources and Data as a Service (DaaS) to access the necessary input data. This paper covers the main technological highlights for such solution.


The value of cloud-based massive processing is best achieved if resources are only consumed when needed. This requires the ability to autoscale computing capacity based on the current and foreseeable demand, which is at the core of the FS-TEP elasticity function illustrated in the following picture.
Let’s imagine that the platform receives a request (via the user interface or an API) to process the Leaf Area Index (LAI) over 100 Sentinel-2 tiles. The Job Handler will add 100 processing jobs in the Job Queue. These jobs are executed by the Worker Nodes, running on IaaS processing resources instantiated on demand, which retrieve the jobs from the queue and manage the processing workflow. The autoscaler sits in the middle and optimises the resource capacity, by increasing or decreasing the number and type of worker nodes on the basis of the characteristics of the job queue, e.g. number of pending jobs, relative processing weight, etc. This allows optimising resources usage elastically in response to varying needs. A set of configuration parameters allows tuning the logic for activating/de-activating nodes on the basis of the billing features of the chosen IaaS provider. For instance if the provider bills nodes usage at a granularity of one hour, the autoscaler only switches off nodes when a relevant fraction of the hour has been used, so that the processing resources can still be used by other processing tasks. In essence the autoscaler supports fast and cost-effective processing of large data sets.
As an example of a massive processing scenario, the FS-TEP has been used to derive several biophysical parameters over 6,000 Sentinel-2 tiles, covering an area of approximately 200,000 Km2 (the whole Belgium and Netherlands and vast areas of Germany) and a time range of 1.5 years, worth 32,000 cumulative hours of cloud processing. A degree of parallelism of 40 processing jobs was used and the processing was executed in incremental chunks, thus allowing performing all tasks in few weeks.
The FS-TEP is currently running on top of Creodias (https://creodias.eu/) which supports easy access to a large variety of EO data and processing nodes which are instantiated in elastic manner using the logic described above.

The post-processing analysis is enabled by geospatial analytics features, which at the current stage include the following GeoServer modules:
• ImageMosaic Plugin;
• ncWMS support;
• WPS+WCS combination.

The ImageMosaic data store allows the creation of a mosaic from a number of georeferenced raster data, which can be used to create a multi-temporal layer that can be easily interrogated for time series analyses. The ncWMS module implements an extension to the WMS protocol to support extracting and plotting time series. The WCS+WPS combination (which embed the Java Topology Suite – JTS - library) allows users to easily implement useful geospatial analysis functions to be performed on-the-fly, such as:
• Intersection of vector layers;
• Calculation of features length and area;
• Features aggregation processes;
• Downloading cropped areas of a given layer.


The Platform provides a common API to allow access to its functionality. One of the design goal is to adopt well–known standards whenever available, with a special emphasis on OGC/INSPIRE-compliant standards (e.g. for data discovery, access, description and processing).
The architecture of the FS-TEP platform follows the principles of modern cloud native applications such as cloud optimised workflows, cloud optimised content and container orchestration.


Currently some evolutions are in progress, including:
• Implementing the auto-scaling using Kubernetes: the FS-TEP autoscaler uses similar principles to those adopted by Kubernetes for auto-scaling, e.g. monitoring a given set of resources and responding by adequately adapting the amount of resources deployed. FS-TEP uses Docker containers for deploying processing jobs, thus enabling Kubernetes container orchestration, based on pod management.
• Using cloud-optimised geotiff (https://www.cogeo.org/) as output data format: this allows improving storage scalability, e.g. using object storage solutions, while keeping the geotiff capabilities (e.g. access to overview, etc.).

And others are envisioned, including:
• Augmenting the geospatial analytics features to support more powerful analytics use cases;
• Expanding the API concept with full-fledged API Gateway solutions that provide the technical element to enables future commercial exploitation (including marketplaces for data, algorithms and services) and increase the platform interoperability with the wider EO (and non-EO) ecosystem.

Area C - Board 314

Following many large oil spills, such as the 2010 Deep Water Horizon incident in the Gulf of Mexico, the Oil and Gas (O&G) industry has intensively reviewed their approaches to oil spill management. An Oil Spill Response Joint Industry Project (OSR-JIP) was initiated, which highlighted the need of equipping O&G companies with a Common Operating Picture (COP), i.e. a geo-information system providing accurate and timely information to support oil spill response by providing actionable information during all its phases, including preparedness, emergency response and restoration.

Satellite Earth Observation data is an important contributor to this COP, as highlighted by the previous ESA/EARSC EO4OG project, but there is still potential to expand its use when responding to oil spill incidents. Limitations highlighted by the industry include the wide diversity of data sources and their access methods, uncertainty around future acquisitions, and the need to comply with strict regulatory regimes.

The Oil Spill Incident Data Exploitation Platform (OSI-DEP) project, which is funded by ESA as part of its “Science for Society: Expand Demand” initiative, aims to overcome some of these challenges. It will provide the industry with a platform through which multiple EO data streams can be integrated and visualised to build up a full picture of an incident. These data are made available for use in a company’s own COP, to enable compliance with regulations mandating in-house processing. Initially the data provision will be focussed on open data such as Sentinel products but we will also explore the integration of commercial services within the platform. The project is driven by a user advisory group from the oil and gas sector and seeks to involve additional users as the project evolves.

The platform has been built as an evolution of the Food Security Thematic Exploitation Platform (FS TEP), which provides a scalable architecture, sufficiently elastic to process large amounts of data in response to an incident. This allows the OSI-DEF platform to efficiently provide services such as mapping of oil spills and monitoring their development as new images arrive.

A key innovation within this project is the development of a simplified user interface, such that the majority of users will be presented only with the actionable information that they require, while retaining the full backend TEP functionality. It therefore serves as a case study of how the TEP concept can be adapted for a specific user group, balancing the needs of the users with those of service developers.

The project seeks to expand the use of EO data within the Oil and Gas sector by developing demonstration services that can be provided through the platform itself. For the offshore oil sector there are already a wide range of services available for oil spill mapping and drift modelling, and we will integrate a selection of such services directly into the platform.

For the onshore oil sector, such services are less mature than for the offshore oil sector. The project will therefore develop new services targeted at this community, allowing the monitoring of pipeline leakages through impacts such as damage to nearby vegetation.

Area C - Board 315

The Copernicus Emergency Management Service (EMS) provides valuable information for emergency response in relation to different types of disasters. Once triggered the EMS is providing rapid maps after a natural hazard event. Of course high map quality is not the major task of the service, the first priority is to support the efficient management of emergency situations.
But the EMS resulting information can also be used as reference data for further remote sensing research activities. Therefore, it would be interesting to know about the map quality. The validation methodology is based on the validation protocol developed by the Joint Research Centre (JRC). Anyways, an independent quality assessment of the results are valuable in order to evaluate the application of the data.
The German land cover change detection service “Laverdi” will automatically and regularly provide land cover change information by 2019. The detection of changes in the earth's surface on the basis of satellite-based remote sensing data has been a very active field of research for several decades. The aim of Laverdi is to use free Copernicus satellite data for an automatic derivation of land cover changes (with focus on forests, wetlands, cropland, bare soil, water, grassland and urban land cover) and to regularly provide this information for individual landscape elements via a web service. To achieve high processing performance the “Copernicus Data and Exploitation Platform – Deutschland” (CODE-DE) for data processing and analysis. Finally the most up-to-date and consistent information about land cover changes can be used to continue and update geodata from different sources in a comprehensive and uniform quality. A further objective of the project is the regular, standardized extraction of statistically relevant geo-information for the quantitative description of several UN-SDG (United Nation - Sustainable Development Goals) indicators. So far, the landscape change detection service Laverdi was used for supporting the calculation of the SDG indicators 11.7.1, 15.1.1 and 15.4.2 for the year 2015.
Forest land cover changes after aregional storm events are planned to be derived automatically. During the development phase of Laverdi windfall areas were predicted for the Friederike storm event in January 2018. The comparison of the results of the EMS results, as well as in-situ observations for some of the affected areas are shown in this study.
Due to the very high resolution (VHR) remotes sensing images EMS is using for its maps the minimum capture size of the windfall areas are < 0.1 ha. The rapid provision and the individual user specification of the maps are of high valuable for decision makers. However, the dates of the VHR pre event images for Friederike storm were not always appropriate. This results in inaccurate derivation of storm affected areas in comparison with the Laverid approach. Laverdi detects only areas larger than 0.2 ha, but the comparison with EMS and in-situ observation indicates that the quality of the detected areas of Laverdi is higher than for the EMS. Mainly due to the inappropriate pre event date and the detection methods.

Area C - Board 316

ImClass consists in a generic and user–tailored active learning ensemble classifier for the detection and mapping of a broad range of land surface classes. To have relevant operational results, the main features of your service are 1)the use of medium resolution imagery (e.g. S2 & L8 datasets), 2)the use of a mono-class approach taking several ill-posed problem (e.g. class imbalance) as use cases (landslide, urban works, burnt areas), 3)the use of the existing ALADIM generic classification system (developed at EOST by Stumpf et a.l, 2018) as the base classifier, 4)the involvement of operational users (e.g. ONF-RTM, EuroMetropole Strasbourg/EMS) at all steps of the algorithm development, 5)the use of the A2S processing infrastructure of University of Strasbourg to carry out and implement the calculation.

The algorithm is able to combine multi-sensor optical images, and is optimized for rapid processing. The algorithm combines a Random Forest classifier integrated in an active learning cycle in order to select iteratively the most relevant training samples for the classification step and increase the accuracy of the model. The method uses innovative approaches as 1) a large number of image features are computed, 2) several data type can be combined (optical sensors, topography), 3) image time series can be integrated to reduce noise.

Results obtained for the classification of urban works and forest fires are presented, using both VHSR and MSR optical images. The tests, performed on large regions highlight 1) the effect of segmentation on the quality of the results, and 2) the influence of the training samples as a diversity of mapped objects is needed for accurate object classification. Results of extensive testing on cities of the Grand Est Region (France) and over rural areas in Portugal are discussed.

As the method has been optimized using open standards for high performance computing, the algorithm is fast.

Area C - Board 317

One of the trends subscribing to the big data theme in remote sensing is the combination of increasing satellite data sets. In land monitoring, traditionally, data from optical satellite has been used. With the availability of new and open optical and radar satellites data (e.g. Sentinel-1 and -2), multi-sensor methods combining both data sets are advancing. In general, Sentinel 1 and Sentinel 2 observation times are not identical and optical and RADAR observations are not directly compatible (e.g. optical reflectance vs RADAR backscatter) (Reiche et al., 2015). The immediate effect of misregistration in land change monitoring would be higher commission errors for the supposed detected change. Still, automated and scalable coregistration, although necessary and usually part of the preprocessing workflow, remained a challenge left for big, global scale, near real time applications. Precisely coregistered radar and optical data sets are crucial, in order to apply a multi-sensor radar-optical algorithm. While geocoding of individual optical and RADAR images can be done with a set of software packages e.g. by Werner & Strozzi (2000), no coregistration method targets optical-radar multi-modality, while being scalable and open-source, within the context of land change monitoring.

Here, we will present and discussion first results and proof of concepts towards the development of an open-source and scalable method for coregistering Sentinel 1 and Sentinel 2 data in order to support near real time land change monitoring applications. We will illustrate and discuss the follow key requirements. Firstly, accurate coregistration at pixel level is critical to avoid false change detections, which can defeat the initial purpose of the application. Secondly, the coregistration method has to be unsupervised, yet robust enough, as it is must support global monitoring algorithms, which use the entire Sentinel 1 and Sentinel 2 datasets. Thirdly, it must be performant enough to support near real time applications, which need to process by themselves the same large volumes of data as well. Based on these requirements, some properties of the coregistration method can be extracted. Optical and radar imagery have inherently different viewing angles due to their acquisition modality. Considering this and to ensure pixel level overlay, the coregistration has to be non-rigid. To achieve the second and third requirements, the Sentinel 1 (radar) and Sentinel 2 (optical) images are manually analysed in order to collect as many assumptions as possible to optimize the generalized coregistration methods.

The following characteristics have been observed in the Sentinel-1 and Sentinel-2 datasets, in order to be used as assumptions for the developed coregistration. Both datasets are georeferenced with an estimable accuracy. Sentinel-2 optical imagery has a more precise internal coregistration over time. Internal misregistration precision is not arbitrary, but depends on various metadata (path, ascending or descending acquisition, etc). Changes are assumed to be rare. Changes not related to land features are not of interest, thus considered noise and assumed to follow a normal distribution.

In summary, we will present first results of the developed coregistration method dedicated to Sentinel-1 (RADAR) and Sentinel-2 (optical) data fusion for supporting global near real time land change detection applications.

Keywords: coregistration; radar; multimodality; Sentinel; big data; near real time; land monitoring

Reiche, J., Verbesselt, J., Hoekman, D., & Herold, M. (2015). Fusing Landsat and SAR time series to detect deforestation in the tropics. Remote Sensing of Environment, 156, 276–293. http://doi.org/10.1016/j.rse.2014.10.001.

Werner, C. & Strozzi,T. (2000). Gamma SAR and interferometric processing software. In Proceedings of the 2000 ERS-ENVISAT Symposium. Gothenburg, Sweden.

Area C - Board 318

It has now become commonly understood that the Earth Observation by remote sensing, although being an unprecedented means of Earth’s environment monitoring (especially in this Copernicus era), has to be seen and exploited as a complement to other means of observing our environment instead of a stand-alone system to get the most out of it. The physical and biological modeling ingests the EO and in situ available data to provide an as-complete-as-possible picture of Earth for today and allows interpolation in space and time, extrapolation and hopefully trends detection in the frame of evolving climate.
In such a context, although still be a technological challenge, having access, in a two-click approach, to all available environmental information from multiple sources at a given place and time has become the next challenge that would ease and definitely change the daily duties and scientific investigations work of environmentalists. With the venue of new IT platforms like DIAS that aim to streamline Copernicus data fluxes, this target appears to be achievable and is the object of our new endeavor presented in this talk: the development of a service that rely on Copernicus data access streams (TEPs, DIAS, Sentinels hub, Coll GS…) for the collection of EO data and model data, and on multiple additional sources, for in situ observations as brought by open participative observations.
All these ‘facets’ of the environmental multiplex are gathered under the same platform – similar to what has been the Cubism art approach in the start of the 20th century, and that would be technologically emphasized 100 years later . A business approach has been designed and is presently under deployment to foster the use of EO data (but not only) and to identify and drain new needs for observations allowing a better start for the NewSpace. First results of this platform will be presented at LPS2019.

Area C - Board 319

Since the launch of the first Sentinel-2 satellite in 2015, a large amount of Earth Observation (EO) data has been acquired and delivered for various thematic applications. The Sentinel-2 satellite along with the Landsat missions provide medium-resolution multispectral information about the earth’s surface, and by continuously orbiting earth, they enable the extraction of additional information in the form of, e.g., time series data. It can be expected that by using automated methods to analyze the time-series data, new high-level services can be produced.
In order to achieve the high-level services, there is need to develop efficient methods to extract information from the huge EO data storage. Several tools for information extraction from Sentinel data have been introduced in recent years; however, we have experienced that some of the potential end-users of Sentinel-2 data either lack of resources or skills to start exploiting the EO data. To be flexible with FGIs various project requirements and to be able to adapt to collaborator (e.g. other national research institutes) requests, we have built our own toolkit, the EODIE (Earth Observation Data Information Extractor). EODIE was developed at the Finnish Geospatial Research Institute (FGI) utilizing freely available tools.
In the current development stage, EODIE is set up to work with Sentinel-2 and up to a certain level also Sentinel-1 data. The user provides an (ESRI) shapefile with polygons of interest or the extent of an area of interest and a time range (starting from 2016 until today) and what kind of output is needed. The output options range from atmospherically corrected products (L2A) to cloudmasked vegetation index or single band raster files and statistics given per polygon of interest. Additionally, users can extract polygon-wise time series data from the processed statistics. The toolkit is based on command line (bash) and Python scripts that call several python libraries and ESA provided open access software, like SNAP and sen2cor. Several open Python libraries are utilized to, among others, read and reproject the shapefile, read and write raster files, calculate per polygon statistics and visualize the results.
The toolkit can be used on both Linux and Windows systems. It can be deployed on virtual machines running on remote servers, with data on site or used locally with the data downloaded from, e.g. the Sentinel open data hub.
Pilot projects utilizing the toolkit have been started in forestry and agriculture related studies and more projects including the toolkit are being planned. The EODIE development plan includes adding support for additional Earth Observation input data and setting up a machine learning pipeline. One future goal is to utilize the open topological database managed by the National Land Survey of Finland together with EODIE outputs in developing machine learning models on national scale.

Area C - Board 320

The Data Cube Service for Copernicus project (DCS4COP, EU_H2020) addresses the challenges of handling the big volumes of Copernicus data, integrating data streams from different sources and generating high-quality information. The project was set-up in 2018 to demonstrate and establish a novel service for the value-adding Earth Observation industry: the EODataBee service.
DCS4COP objectives
• To provide a novel service to intermediate business users.
• To exploit the scientific excellence of precursor projects.
• To develop and stablish a DataCube environment for Copernicus data.
• To exploit European EO infrastructure.
• To prepare a comprehensive service portfolio.
• To integrate the services into user’s workflows.
• To consolidate the commercial cooperation framework.
• To establish the self-sustained business.
DCS4COP wants to demonstrate first the value of EO data for the market segment of coastal and inland water services, capitalising on the scientific achievements of the previous HIGHROC research project. The vast experience of the DCS4COP team in water quality ensures the high quality of the environmental information service. The project will also evaluate other thematic areas for the long-term plan and business expansion of the service, like atmosphere, land or emergency.
EODataBee service model
EODataBee will be a commercial service with a portfolio comprising a great variety of environmental information layers, easy-to-integrate user interfaces and applications, tailored trainings and consultancy services. This will be offered as a Platform as a Service (PaaS) or Software as a Service (SaaS) to intermediate business users (IBUs) acting on the market of environmental information products and services.
EODatabee will offer individual solutions tailored to customer requirements, allowing each IBU to detail the specific thematic, geographical and temporal requirements while offering great flexibility with respect to the technical infrastructure and user-defined processing. The service will cover a large part of the EO value chain and will seamlessly integrate environmental data from other sources.
It will enable IBUs: 1) who already provide services to end users to add additional spatial information layers to their portfolio and to optimise their production process in terms of efficiency and costs; and 2) who are not yet on the market to enter this market at a very low entry level and with high-quality services right from the start.
EODataBee is a novel capability to ease the integration, preparation and processing of numerous data sources for EO downstream applications.

Area C - Board 321

Expansion of urban areas is taking place at twice the rate of urban population growth (Seto et al., 2014). The conversion of agricultural lands and the cropland, grasslands or wetlands making up the agricultural area to urban settlements results in a change in carbon density of the land area and results in either net emissions or removals to the atmosphere. Generally, changes will result in losses of carbon but in cases where the area of vegetation cover has increased or the soil management practice has changed (e.g. from intensive tillage to perennial vegetation such as parks, trees or golf courses) there could also be carbon gains in above ground biomass, dead organic matter and soils. In fact, some studies (e.g. Huang, Robinson, & Parker, 2014) suggest that carbon storage has increased after urban development. Emission and or removal estimates from this source, defined as Cropland, Grassland or Wetland converted to Settlements, according to the Intergovernmental Panel on Climate Change (IPCC) nomenclature, have not yet been developed for Canada. The lack of estimates from this source are recognized as a gap in Canada’s reporting of emissions and removals from the land sector in Canada. In order to fill this gap, a consistent time-series of activity data describing the changes in vegetation cover during urban expansion in Canada needs to be compiled. Environment and Climate Change Canada (ECCC) is currently working with PCI Geomatics in developing a Big Geospatial Data Analysis framework (i.e. Data Cube) including the creation of Analysis Ready Data (ARD) within PCI’s GXL system for the thematic classification of a variety of earth-observation imagery (high resolution orthophotos, Landsat, Sentinel 1 and 2, RADARSAT-2). This system will enable a time-series analysis of earth-observation data over the Census Metropolitan Areas (CMAs) of Canada and provide an assessment of land cover changes resulting from urban expansion in Canada since 1990. These assessments of land cover change will then be incorporated with models of carbon from biomass gain/loss to estimate emissions and removals within this rapidly changing landscape. This presentation will discuss the solution under development with a focus on the creation of at scale multi-sensor Analysis Ready Data, earth observation indices and time series analysis techniques that are most useful for detecting change in land cover and biomass.

References
Huang, Q., Robinson, D. T., & Parker, D. C. (2014). Quantifying spatial–temporal change in land-cover and carbon storage among exurban residential parcels. Landscape Ecology, 29(2), 275-291. doi: 10.1007/s10980-013-9963-0

Huffman, T., Leckie, D., McGovern, M., Olesen, M., Green, M., Hill, D. A., Liu, J. (2015). Integration of multiple spatial datasets in the development of a temporal series of high-accuracy, high-resolution land use maps 35th EARSeL Symposium. Stockholm, Sweden.

Seto, K., C., S. D., Bigio, A., H. Blanco, G., Delgado, C., Dewar, D., Ramaswa, A. (2014). Human Settlements, Infrastructure and Spatial Planning. In S. K. (Ed.), Climate Change 2014: Mitigation of Climate Change. Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change New York, NY, USA: Cambridge University Press.

Area C - Board 322

The Big Data revolution in satellite Earth Observation started when large EO data archives were released publicly. However, at that time the full potential to turn data into information was heavily limited by the unavailability of sufficient ICT resources to handle full mission archives enabling data analytics, multi-temporal analysis, etc. The Big EO Data revolution continued with the progressive enlargement of commercial and free&open satellite data archives which were improving performances in spatial/spectral resolution and revisit capabilities. The latest step (chronologically) in this sense is the major contribution of the Copernicus programme through the Sentinels. In parallel, the value of satellite images has been slowly shifting from the traditional use (single image analysis to derive cartographic layers) towards innovative data exploitation strategies based on Big Data paradigms such as large scale multi-temporal analysis, rapid detection of objects with a frequent revisit, fusion of EO and non-EO data. The new Geoinformation paradigm is now based on the intimate combination of satellite time series, scalable ICT (e.g. cloud) and EO data processing and analysis capabilities to deliver actionable information (e.g. analytics, reports, maps). In this context, modern Geoinformation service providers cannot avoid to adopt novel technologies such as Data Cubes deployed in highly scalable cloud environments to build effective service chains that are in line with vertical market expectations which are more and more demanding in terms of spatio-temporal coverages, delivery time and accuracy. CUBEO is e-GEOS scalable pre-processing and Data Cube platform for Geoinformation application services developed in cooperation with MEEO and exploiting MEEO’s well recognized technology and experience in EO data access platforms. CUBEO allows its users to build on demand Data Cubes from multi-source optical (e.g. Sentinel-2, Landsat-8, MODIS) and SAR (e.g. Sentinel-1, COSMO-SkyMed) data over any Area of Interest, any Time of Interest, worlwide in just a few steps. The main concept behind CUBEO is to provide expert and non expert users with professional EO Data Preparation Pipelines (based on free SW such as SNAP as well as on e-GEOS proprietary solutions) capable to pre-process raw data (i.e. Level 1 optical/SAR data) to generate Analysis Reeady Data (i.e. Level 2 / Level 3 products) accessible though Data Cubes (i.e. extended WCS interfaces) to enable further data analysis operations (e.g. classification, anomaly detection, extraction of descriptive/predictive analytics). CUBEO has been developed using a cloud-oriented approach (containers, process orchestration, autoscaling) to maximize the horizontal and vertical scalability of the Data Preparation Pipellines and of the WCS data access endpoints to be able to cope with on demand Data Cube requests largely varying in size. CUBEO is currently deployed on AWS, with minor migration effort required to deploy it in a different cloud environment (e.g. Copernicus DIAS). CUBEO has already supported large scale geoinformation analysis tasks such as, for example, the analysis of MODIS 2000 – 2018 Time Series at continental scale as well as the preparation and analysis of combined Sentinel-1/Sentinel-2/Landsat-8 yearly Time Series over more than 2 million km2 for agricultural monitoring purposes.

Area C - Board 323

Through a secure cloud computing architecture, the sobloo DIAS platform provides a broad access to Copernicus data, as well as complementary data collections, ranging from Earth Observation commercial imagery to Mobile and IoT databases.

Going beyond the data, sobloo offers a suite of tools including APIs and services to process data, including on-the-fly, without having to manage volume and timing constraints. All users, public or private are invited to take advantage of sobloo’s universal docking station to plug in their data, services and applications.

Turning data into insight, sobloo platform allows accessing and creating thematic services to extract added-value information based on Copernicus data. This is what the expert community has been wishing for and will equally allow non-experts to find sound information. First of a series of upcoming thematic services, soCAP and Muscate are helping users to keep concentrated on their domain of expertize in directly accessing actionable information:

• soCAP – Biophysical parameters from sobloo
sobloo leveraging its partnership with Airbus to be first to propose soCAP agricultural service.
sobloo will make dedicated online services available to analyse crop type, development, agricultural practices, and production over single parcels, regions or countries. Result of years of “field” experience in agriculture, the new Airbus service soCAP is 100% cloud-based, offering systematic declouding and processing in real-time. It provides 100% consistent and reliable crop profiles series regardless of the imagery type used, thanks to the automatic generation of biophysical parameters. It is also 100% accessible, thanks to a scalable API, for a seamless integration in your current and future CAP monitoring tools. Behind the online services, sobloo adds the massive and up-to-date Sentinel data collections that soCAP will process in real-time, completed by commercial HHR imagery where needed. The analysis of both longer time-series as well as very fast processing of the most recent archive, via a dedicated data warming functionality, will be no challenge for the user who can focus on further steps of analytics. The resulting robust crop profiles support the transition to a policy control focusing on fostering a more sustainable and profitable farming.

• Muscate – multi-mission processing service on-the-fly
Promoted and developed by the French Space Agency CNES, Muscate is a multi-satellite, multi sensor ground segment for multi temporal data. Muscate offers trustable products ready-to-use on land use monitoring applications based on imagery products from Sentinel 1, Sentinel-2, Landsat 5, 7 and 8, Spot World Heritage, Formosat, Kalideos.
Within Sobloo, the following Muscate products will be made available:
- Snow ON/OFF (L2B/L3A);
- Surface reflectance (L2A);
- Cloud Free (L3A)
- Land Cover (annual and monthly maps)


Supporting creative grounds, a dedicated so-called front-end team has equally been set up. This team accompanies users make the best out of platform’s capabilities to drive their business or institutional needs. Users are therefore invited to take advantage of sobloo’s partner’s geo-information, IT integration, consultancy and cloud expertise.

In addition, users can benefit from a suite of professional services at different steps of the data value chain, ranging from business support features from the development of algorithms to the setup of billing as a service. True to our collaborative and innovative roots, sobloo encourages co-creation and developing its data and services offer in order to enrich Copernicus based applications.

sobloo supports the creation of new services and new uses and will lead to an increase in:
• Visibility: by offering worldwide visibility to showcase services and applications
• Collaboration: by fostering a collaborative world to enrich data collections and services
• Growth: by staying focused on what third parties do best and letting sobloo do the rest, such as commercial exposure, billing and rating services, managed services, etc
• Exploration: by offering a unique environment to test and innovate with new business models.

Early adopters and future users are encouraged to give their feedback in order to further improve the overall user experience.

sobloo is thus uniquely positioned to contribute to the development of a living and innovative data ecosystem and give answers to the Big Data era.

Area C - Board 324

ABSTRACT
Mundi Web Services

A digital platform providing better access and use of earth observation data to create and operate innovative apps whatever the market thematic.

1. INTRODUCTION
The European Union Copernicus Programme relies on a family of dedicated Earth Observation missions called the Sentinels. The data acquired from these missions are systematically downlinked and processed to operational user products. Copernicus data and information includes the user products from the Sentinels missions, but also value-added information generated by the Copernicus Services and data and information provided by the Copernicus in-situ component.

The Copernicus DIAS (Data and Information Access Services) consists mainly of virtual systems hosted on a cloud computing environment and providing IT services. This Cloud infrastructure makes available for processing and operations a data offer, which includes Copernicus data and information but may also include other geospatial data (e.g. EO ESA missions, National missions data, etc).

2. MUNDI
The Atos DIAS is called Mundi
www.mundiwebservices.com
Combining simple access to Copernicus and third-party data, cloud and big data technologies, Mundi accelerates the time to market for innovative services based on earth observation data. Mundi allows to deliver a service that is distinctive, reliable, and immediately monetizable.

3. THE CONSORTIUM
Led by Atos, the Mundi team is based on a strict European consortium with company widely recognized on their respective specialties.
Mundi consortium ensures access to Copernicus data and information and are responsible for the provision of the DIAS underlying ICT infrastructure and interfaces and offers well-documented and scalable ICT resources to Third-parties.
The Mundi consortium is composed of Atos, T-System, DLR, Thales Alenia Space, GAFag, e-Geos, EOX, Spacemetric and Synergize

4. THEMATIC USE CASES
Mundi supports many thematic markets. For example:
• Agriculture
• Urban monitoring
• Services industry (Finance, Insurance, retail…)
• Tourism and transport
• Environment
• …
This, in hosting and enhancing the creation and operations of these many application fields; e.g. Public land reserves follow up for local communities

5. SOLUTION
Thanks to its Marketplace, Mundi allows delivering an application that is visible from the market, distinctive, reliable, and monetizable.

Mundi provides Cloud platform capacities to access a curated Data Collection and a selected Tools Collection.

6. CLOUD
These components are accessible at the Infrastructure layer to implement directly a customized Cloud tenant. Open Telekom Cloud provided by T-System in Germany bring a secured, robust and performant Cloud infrastructure for Mundi users.
-Compute with virtual computing servers (Elastic Cloud Server - ECS) through virtual network computing console (VNC)

-Storage in block level storage capacities (Elastic Volume Service -EVS) or in Object Storage Service that offer a highly simplified access mechanism and a high level of scalability
-Load balancing for distribute traffic on multiple ECS and balances loads.
-Big Data functions (Map Reduce Service) offers a range of tools that allows for big data analysis. This includes storage capacities and methods as well as analysis functions. Run with Hadoop, HBase, Spark, Hue, Loader, Kafka, Storm, Flume and Kerberos.
-Database: mySQL, PgSQL, SQL Server

7. DATA COLLECTION
Mundi offers Copernicus data and information to be used right on the cloud. Of course, it covers Sentinel data, Copernicus Core Services but also Landsat-7, Landsat-8, and many more satellites to come (Cosmo-Skymed, IRS.). Data are made available through the standardised interfaces on the Mundi platform.
In addition, other Earth Observation data sets and more thematic data can be used on Mundi to maximize opportunities to quickly and easily develop a distinctive, reliable, and immediately monetizable service.

8. TOOLS COLLECTION
Mundi allows to access 2 level of tools:
-A set of free of charge services to reach a first level of Earth Observation Data management,
-Many advanced tools are available on Mundi provided by a third party insuring a more advanced data usage.

9. SUPPORT
Atos and its Mundi consortium can assist Mundi users in a very wide range of support activities, complementing their own skills.
From technical or functional supports to more thematic or business consulting, Mundi team is designed to help Mundi users to reach new opportunities.

Area C - Board 325

Earth Observation applications are benefitting from a constantly growing amount of data from satellites. Especially the Sentinel family is contributing since 2014 to this gigantic amount of information that needs to be analysed. The increasing complexity of algorithms and models is additionally demanding higher computing resources. Cloud computing gets more and more accessible even for SMEs and other users and developers that do not have an IT-only focus but want to solve problems within multiple disciplines benefitting from the huge amount of information contained in the data available.

The company VISTA is facing the challenge to exploit the available remote sensing data, making use of sophisticated algorithms that demand high computing power for a growing number of customers and thus increasing areas. Research projects dealing with large areas, such as the whole Zambesi catchment in the Horizon 2020 project DAFNE, also have needs for high computing power to process needed datasets. Besides ESA’s Food-Security Thematical Exploitation Platform (FS-TEP), that is already used to process Satellite data, a second solution is desired to build up a redundant infrastructure. Experience that has been acquired during multiple projects dealing with cloud computing, such as Apps4GMES and the FS-TEP, has been used to develop a high-performance scalable cloud environment based on the Copernicus Data and Access Information Services Platform (C-DIAS). MUNDI, as the DIAS platform that has been developed by a consortium led by Atos is called, is based on the Open Telekom Cloud (OTC) as computing infrastructure. The MUNDI Platform is offering the needed infrastructure for fast and easy access to browse Sentinel and also Landsat datasets based on open search or the OGC Catalogue Service for the Web (CSW) Standard. Direct access to the datasets from within the OTC is serving the cloud paradigm of bringing the processes closer to the data.
VISTA applied the state-of-the-art cloud technologies Kubernetes and Docker to develop applications that are easy to run and scale on almost every available Cloud Platform on the market. In the presented use-case a Docker image containing all the needed software for pre-processing of Sentinel-2 data has been build. This image is performing an atmospheric correction, computation of cloud & cloud shadow geometry, specific methods of radiative transfer modelling for classification and vegetation parameter retrieval. The outputs of this pre-processing are needed as input for the agricultural products produced in a next step. The whole pre-processing requires a lot of computing power, resulting in 3h runtime on a standard desktop PC with 8 cores and 8GB memory for a complete scene without clouds.
The developed Docker image is in a next step used as a blue print to build so called containers. Each container has the same set-up defined in the image, ready to perform a task, in the use case implemented, the pre-processing of one Sentinel-2 scene. A Kubernetes cluster consisting of 10 worker nodes, each equipped with 32 CPUs and 64GB of RAM has been set up on OTC. A master node is performing the job and resource management of the whole cluster, allowing easy distribution and orchestration of jobs. In the described set-up, the whole processing power of 320 Cores and 640GB of RAM can be accessed, which could also easily be further extended.
The workflow implemented by VISTA on its own premises to search, download & process Sentinel-2 data has been adapted to send jobs for pre-processing to the kubernetes master node, which is taking care of setting up containers for each scene that needs to be processed. Each container is loading the scene from the archive and starting the pre-processing. Depending on the number of scenes in the job queue, all 10 nodes or only a fraction is powered up and after successful processing and delivering of outputs of all jobs automatically shut down again. This leads to the desired scaling effect of cloud resources. Accessing the whole processing power of the cluster, 80 Sentinel-2 scenes are processed simultaneously within 4 hours which results in a processing time of ~4 hours for one coverage of Germany or sixty hours for whole Europe. Given the repeat cycle of Sentinel-2 a&b of 5 days, the data that sums up over Europe within one month can be processed within 15 days.

The presented technological solutions can be adapted to multiple processing requirements, while giving access to the almost infinite computing resources that is offered by the cloud. Direct access to satellite archives, as offered by the FS-TEP and DIAS platforms, as well as automatically constant adapting of computing resources to the current needs is implementing the two important cloud computing paradigms of bringing processes closer to data and high scalability.

Area C - Board 326