Deliverables
GEM portal will make all available public products and documents publicly available to a user community wider than the one already involved in the projectSINERGISE will setup GEM project page and later manage it throughout the project
Integration of VHR sourcesNew data gateways will be designeddeveloped to enable ingestion of Very HighResolution data by Sentinel Hub Draft list of sources to be supported PlanetScope SPOT Pleiades and WorldView
Project Demo regions 10 and 1 MIO km2 areas will be selected based on analysis of available data A set of sites will be selected to allow benchmarking the methods and algorithms with regard to the diversity of LC context the various landscape patterns ecoregionalisation patterns actual satellite observation conditions atmospheric perturbations and cloud coverage availability of insitu observation and verification capabilities of consortium The challenge is to allow identifying the most appropriate regions which would enable scalingup to the global scale For obvious reason it is also very important to select sites where historical insitu and satellites time series are already available or at least will be ready with the star of the project The answers on who how and when we will prepare certain datasets are the most important ingredients of the documentFor the selected DEMO areas all of the dataset will be collected or collection services established
Front office services will need to be upgraded/developed to expose new capabilities of Sentinel Hub and to make it possible to integrate them into TOMTOM’s and SATCEN’s use case. Three upgrades are planned to complement the existing capabilities:-QGIS; Sentinel Hub QGIS plugin will be developed to enable users to configure and harness the power of GEM services directly in QGIS. QGIS (previously known as Quantum GIS) is a free and open-source cross-platform desktop geographic information system (GIS) application that supports viewing, editing, and analysis of geospatial data. As SINERGISE encourages collaboration and further development, this component will be released as open source to the QGIS community! The plugin will be easily reworked for other WMS data sources, especially those with a time component. -PYTHON & JUPYTER NOTEBOOK; For supporting thematic and temporal exploitations the GEM will provide effective sub-setting and analysis functionality that can be accessed by a Python client which serve to the user only the data amount that is really needed. Jupyter Notebooks are a way to hold this Python code, they can be shared among users (e.g. via GitHub), and executed from within the Web browser. GEM will provide examples/templates of such notebooks to its end-users “with batteries included”, i.e. which will allow quick-start utilization of GEM’s data, processing and retrieval offerings reachable via the mentioned Python WPS client so that user’s algorithms can be deployed to run in a secured way on server-side archives. -JUPYTER HUB HOSTED IN THE CLOUD; An alternative to the user-local option described above, a cloud-based Jupyter service (Jupyter Hub) will be provided on the server side (1) to prevent the need for users to download large amounts of data, (2) to free users from setting-up and maintain ICT infrastructure for processing and (3) to facilitate sharing and commercialising user-generated information products. The open-source Pangeo package, one of the most developed community platforms for BIG DATA (maintained by UK Met Office, US National Center for Atmospheric Research, Lamont Doherty Earth Observatory and many other well-known contributors), will be integrated. With this cloud-based service, when users authenticate to the service, a docker container with a Jupyter lab-session will be spawned. Users will have an immediate access to the aDC and will be provided with the computational resources to explore and exploit it. Auto-scaling functionalities of CLOUD environment will allow for dynamic and automated resource provisioning, thus keeping operating costs at affordable levels.
METEO/CLIMATE serviceMeteo/climate services will be developed using METEOBLUE's existent and in-project algorithms (up/downscale ). The resultw will be a new input data source for Sentinel Hub: METEO/CLIMATE service. Weather models will be based on the NMM (Nonhydrostatic Meso-Scale Modelling) approach, forecasts provided based on METEOBLUE’s proprietary global weather forecast computation. For the purpose of GEM, the meteorological historical databases for all known global meteorological models (GFS; NEMS, ECMWF, ERA-5) will be provided for training and validation purposes, along with validation routines. Meteorological data (historical and forecast) will be retrievable through an API to be used in the operational product and for post-processing. The data will be provided in a form of calibrated artificial hourly “measurements” for every pixel, of variables such as expected radiation, intensity, air temperature, surface temperature, eventually also for derived variables such as precipitation and soil moisture.The GEM project is considered a strategic commercialisation leap – METEOBLUE becoming the first global provider of weather data to integrate its services into EO downstream platform.
EO-LEARN GatewaysTUM will contribute their existent data loaders interfaces and wrapper functions helping SINERGISE to accelerate the development of ML gateways towards external ML frameworksThe exact list of ML frameworks will be fixed after the data modelling requirements will be finalised Current list of potential ML frameworks to be integrated consists ofTENSORFLOW Googles TensorFlow is the most widely used framework for machine learning based on Theano see below KERAS is an open source neural network library written in Python It is designed to enable fast experimentation with deep neural networks it focuses on being userfriendly modular and extensible APACHE MXNet is a modern opensource machine learning framework used to train and deploy deep neural networksPyTorch primarily developed by Facebooks AI Research lab FAIR is free and opensource software ermarked by two highlevel features Tensor computing and Deep neural networks
Cloud maskingCloud masking algorithms will be included into EOLEARN to enable pseudoprobability cloud masking of Sentinel dataFollowing it Sentinel Hubs preprocessing chains will be updated to integrate new pseudoprobability Cloud masking and make it readily available for complete Sentinel2 archive
Continuous monitoring servicesContinuous monitoring services will be programmed to expose full set of GEM services: MONITORING = MAPPING + ALERTING + DRILL DOWN, all of them forming a foundation for GEM’s “spot mode” monitoring.Contiunous monitoring service will be realised through the orchestrated functionality of main GEM's components:1.INPUT DATA principles promote data fusion concepts where all of the existing Sentinel Hub’s data sources: ARD-processed EO data (Sentinel 1, 2, 3 and 5P, MODIS, LANDSAT, commercial VHR imagery; ARD: Analysis Ready Data) and Sentinel-Hub’s EO indices will be complemented with weather data, in-situ data (ground-truth training and validation data), vehicle positional data and ancillary data sets.2.ICT PROCESSING INFRASTRUCTURE capable of (1) facilitating sustainable (cost-effective) processing of data on a global scale (or regional scale) and (2) supporting spot mode monitoring processing (discovery of and interpretation of relevant details). 3.MACHINE LEARNING METHODS to produce continuous, global, cross-scale, multi-purpose, incremental data-interpretation services (mapping and change detection services).4.DATA MODELLING will exploit input data and employ ML methods to develop processing chains (interpretation models) capable of (1) interpreting data of global/regional magnitude, (2) scale independent interpretation of data and (3) early detection AND INTERPRETATION of changes.5.Processing chains (previous paragraph) will be exposed in a form of GEM SERVICES (mapping and change monitoring services), exposing GEM capabilities to be used as standalone service or to be integrated into domain-specific processing chains.
DemonstrationDemonstration run will be performed with the employment of all use-case services in a period of 6-months of continuous operation on DEMO areas of 1 and 10 MIO km2. Data services will provide constant delivery of new data, continuous change detection monitoring will be performed as data arrives, mapping services will be performed on-demand (alert) or in regular periods.
Publications
Author(s):
M. Schneider, A. Broszeit, M. Korner
Published in:
2021
Publisher:
TUM
Author(s):
S. Albani, Luna, M. Lazzarini, N. Baselovic, O. Barrilero, P. Saameno, M. Madrid, A. Patrono
Published in:
2023
Publisher:
IEEE
Author(s):
S. Albani, O. Barrilero, M. Lazzarini, A. Luna, P. Saameño
Published in:
2021
Publisher:
satcen
Author(s):
S. Albani, O. Barrilero, M. Lazzarini, A. Luna, P. Saameño
Published in:
2021
Publisher:
satcen
Searching for OpenAIRE data...
There was an error trying to search data from OpenAIRE
No results available