During the second and final reporting period, the project fully achieved its objectives delivering significant results:
•The functionalities provided by DIAS cloud platforms (CreoDIAS and Mundi) were utilised to set up the backbone of the earth observation (EO) based area monitoring system of DIONE. The system was designed and implemented in a cloud agnostic manner, ensuring that there is no hard dependency on a specific cloud. Sentinel Hub was made available on the different deployments and it was properly configured to fit into DIONE’s EO based CAP compliance monitoring system.
•Deep neural network architectures have been designed and implemented, aiming to spatially augment the resolution of coarser Sentinel 2 bands to the resolution of the finer band (10m), towards establishing a higher resolution basis in the service of the project’s Common Agriculture Policy (CAP) monitoring objectives. A multi-temporal super-resolution (SR) neural network was also trained on VHR imagery obtained through ESA’s Data Warehouse mechanism, enabling monitoring of smaller parcels than with the original bands. The imagery predicted by the multi-temporal SR model described above was used to train a model for classification of non-productive Ecological Focus Areas in both Lithuania and Cyprus. Further to that, an additional feasibility study was done on using orthophoto and drone imagery for this purpose.
•A fully functional version of the DIONE area monitoring system relying on markers has been released that includes a set of image analysis and machine learning techniques, made available via a REST API, which utilise Sentinel signals towards the identification of various crop types, land types as well as the detection of agricultural activity (mowing, ploughing, harvesting) at several times during the growing season. The implemented area monitoring system using Sentinel-2 markers was performed on large scale both in Lithuania and Cyprus – by producing markers on a full country scale.
•The final version of the DIONE geotagged photos framework has been produced. The framework includes a series of technical innovations aiming at assisting and guiding users to capture efficiently representative photos of their parcels (i.e. Augmented Reality, enhanced positional accuracy through the use of multiple location differentiators) while adhering to current technical recommendations and ensuring the security, validity and reliability of the collected photos.
•A soil scanning system that is able to measure and assess soil quality has been delivered. The system is based on a suitably selected low-cost spectral sensor (records the diffuse reflectance spectrum in the near infrared) employing microelectromechanical systems technology along with a mobile application. Novel machine learning tools have been developed and applied in order to transform the raw data collected through the in-situ soil scanning system to the appropriate soil properties (SOC, clay, pH and CaCO3) as well as a second toolset of machine learning algorithms that efficiently combined these point measurements with EO imagery and produced SOC maps for the areas of interest.
•The DIONE compliance monitoring tool, which decides on beneficiaries’ compliance and is integrated with the existing tools of paying agencies, has been implemented. Some of the tool’s key functionalities include the visualisation of parcel boundaries, geo-tagged photos, biophysical indices, area monitoring markers, the ability (for Paying Agency inspector) to request geo-tagged photos through dedicated forms, the assessment at parcel level of the compliance with CAP greening/eligibility rules (including the configuration of complicated decision trees), filtering farms on the map by compliance status and/or by crop group used for compliance decision, and management of assignments of farmers to inspectors or of consultants to farmers.
•A machine learning based Environmental Performance tool has been implemented, including different agri-environmental indicators with a focus on addressing environment and climate priorities within the framework of CAP implementation. More specifically, the tool supports the following indicators: i) land cover change ii) organic farming, iii) land irrigation, iv) greenhouse gases emissions v) water quality vi) soil erosion, vii) soil organic matter viii) natura 2000, and ix) high nature value.
•All DIONE Toolbox components have been successfully deployed and tested in real life conditions in Cyprus and Lithuania during both 2021 and 2022. During the demonstration activities, hundreds of users have been engaged with DIONE toolbox innovations whilst an ample of data have been collected and produced.
•Last but not least, the project has applied various steps towards the efficient dissemination and commercial exploitation of the project’s results. DIONE partners signed a Memorandum of Understanding with the intention of establishing a DIONE Alliance joint venture, which will serve as a leader in the DIONE Toolbox commercialisation effort. A commercial use case, outside the pilot demonstration activities, was already explored and realised in the duration of the project.
