Periodic Reporting for period 2 - CYBELE (FOSTERING PRECISION AGRICULTURE AND LIVESTOCK FARMING THROUGH SECURE ACCESS TO LARGE-SCALE HPC-ENABLED VIRTUAL INDUSTRIAL EXPERIMENTATION ENVIRONMENT EMPOWERING SCALABLE BIG DATA ANALYTICS)
Période du rapport: 2020-07-01 au 2022-03-31
The key outcomes of CYBELE are as follows:
* Deployment of the necessary middleware, Big Data and HPC frameworks and software components, tailored to the respective Testbed
* CYBELE Platform, resulting from the integration of software components and services developed in the project, has been updated with new versions of the software components and a single sign-on functionality
* Implementation of formal, machine-readable, multi-profile definition of the data catalogues. This was exercised for relevance with several agriculture scientific and interoperability working groups and for usability in the cross-domain, emerging standards including OGC APIs, GeoSPARQL, Observation & Measurements
* Publication of 15 more scientific publications in journals and books and 13 more conference papers than initially planned, 26 more articles in industry magazines and blogs than originally planned; organized 25 more events and presented in 72 more events than initially planned, published 4x more articles in the project’s blog, and attracted 40% more visitors and followers than originally planned
* Definition of exploitation paths for each of the project’s assets. See the project’s business plan (D9.5) for full details, including descriptions of future governance structures, future sources of fundings and financial considerations
WP1: Requirements Analysis, Use Cases & Reference Architecture
•Technical requirements and specifications finalised and documented in Deliverable D1.7 (Version c)
•Architecture of CYBELE platform finalised and documented in Deliverable D1.7 (Version c)
WP2: Infrastructure Implementation
•Release of first version of security strategy for CYBELE services
•Initial deployment of most of the CYBELE components on PSNC, USTUTT, WIT and BULL Testbeds
•Initial validation of these components by deploying some preliminary versions of the pilots on some of the testbeds
WP3: Data Services
oData Cleaning and Curation components
oAdvanced Query Builder
oData Brokerage Engine
WP4: Workflow Composition & Services Exposure
•Data-driven algorithms implementation based on demonstrators’ needs
•Design, testing, and validation of data visualisation and user interfaces
•Utilisation of authentication server, designed and developed for the needs of WP2 for the integration within CYBELE platform, for the secure delivery of WP4 envisioned services
WP5: Platform Delivery & Testbed Deployment & Evaluation Framework
•Integration, testing, and release of three versions of CYBELE Platform
•Preparation, configuration, delivery, and maintenance of the CYBELE testbeds
WP6/WP7: Precision Agriculture and Precision Livestock Farming Demonstrators
•Integration of CYBELE platform to demos, when all WP6 demos were focused on activities related to the use of HPC (i.e. parallelisation, setting up Docker / Singularity, etc.)
•Evaluation of the CYBELE platform and PA Demonstrator tools
WP8: Dissemination, Communication & Clustering
•Communicate efficiently the project and its outcomes by meeting and in most cases exceeding the originals targets. Indicatively, we organized 25 more events and presented our work and outcomes in 72 more events than initially planned (in total), and attracted 20% more visitors and followers in the project channels and website than originally planned.
WP9: Exploitation, Business Modelling & Technological Impact Assessment
•Target markets and value estimations, including the breakdowns of Total Addressable Markets, Serviceable Available Markets and Serviceable Obtainable Markets
•IPR Handing and Innovation Management strategies defined
•Finalisation all technological and business impact assessments measured against set KPIs
‘Organic soya yield and protein-content prediction’ was validated on Krivaja, an agricultural partner of BioSense. Those results showed that CYBELE, by implementing a machine learning system that predicts yield based on satellite images and fertiliser content and then finds the optimal NPK amount for max yield/cost ratio, assisted in the decrease of fertilisers costs by 19.24% against the 15% that was set as a benchmark, reducing the corresponding amount from €111.71/ha to €89.64/ha.
‘Predictive models for food safety’ supported the increase of predictive analytics features offered by AGRO to its clients. Four new features were introduced, including the use of a text mining component for the classification of incidents in terms of products and hazards. CYBEKE supported a 58% decrease in the time needed to perform risk estimation, which is even higher than the targeted 50%.
‘Climate services for organic fruit production’ – the following figures are focused on the benefits of the frost forecasting using as reference unit 1 hectare: the affected area can be reduced from 12% to 10%, that means an improvement of 16% (1,600 m2 less affected). This reduction translated into: persimmon: €1,680/ha, citrus €1,239/ha, and stone fruit trees €3,197/ha.
‘Sustainable pig production’: saw an improvement in the average health prediction precision and sensitivity in warning systems for pigs compared to the previous model on the same dataset. This was mainly because the precision of the final 1DCNN model was doubled compared to the old algorithm that was used
‘Open sea fishing’: - forecasts result in a 2.5% increase of catch per unit of effort of sole, the main target species of the Belgian bottom trawl fishery, compared to the status quo catch-per-unit-effort (CPUE), where fishers assume that CPUE on the next day is similar to the CPUE on the current day.
‘Aquaculture monitoring and feeding optimisation’: affected production parameters such as increased growth numbers and lower feed conversion rates. While the actual improvements in terms of KPIs will be more accurate when comparing two full production cycles (spanning 16-18 months), the results already demonstrate significant impact.