by Nikos Kalatzis | Technology Director | GreenSupplyChain DIH
Among the core objectives of the STELLA project is to develop a Pest Surveillance System (PSS) that will act as a holistic digital infrastructure to aid in the early warning and detection of quarantine and regulated non-quarantine pests. The STELLA Pest Surveillance System (PSS) represents an ambitious and forward-looking effort to transform the way pests and diseases are monitored, detected, and managed across agriculture and forestry. Designed as a centralized, digital hub, the STELLA PSS integrates advanced technologies-including IoT devices, remote and proximal sensing, AI models, and citizen-science tools-into a unified ecosystem that delivers early warning, precise detection, and targeted response guidance. By supporting stakeholders ranging from farmers and agronomists to policymakers and researchers, the platform aims not only to improve operational decision-making but also to foster collaboration, reduce reliance on chemical pesticides, and enhance sustainability.
Beneath this sophisticated, user-facing system lies a core challenge that must be addressed for STELLA to fulfil its promise: the fragmentation of agricultural data. The STELLA PSS is developed to facilitate pest surveillance in various agricultural contexts and will be initially assessed in the context of the STELLA Use Case pilots, however the overall architecture aims to be generic and easily expandable for monitoring additional geographical regions and types of pest infestations. The data harmonisation mechanisms of the STELLA PSS are designed and implemented by the GreenSupplyChain Digital Innovation Hub (GSC DIH), with a strong emphasis on the extensive use of data modelling standards as the primary integration approach. Pest prediction models, satellite-based detection systems, proximal sensing technologies, citizen-science applications, and IoT sensors all produce heterogeneous datasets—often in incompatible formats, using different classifications, terminologies, spatial references, and descriptors. Without a robust strategy for harmonizing these disparate data streams, any attempt to produce unified analytics, cross-system visualizations, or interoperable services would be severely constrained.
Handling Data from Diverse Monitoring Systems: Why Interoperability Matters
Modern pest surveillance depends on multiple complementary technologies. Remote sensing platforms provide large-scale vegetation anomaly detection; proximal sensors like the EDEN Viewer deliver high-resolution plant-level diagnostics; IoT stations generate real-time microclimatic data; and pest prediction models rely on weather inputs, biological thresholds, and phenological indices. Each system has distinct data structures, formats, units, and semantic assumptions. For example:
- Satellite imagery produces raster datasets with specific spatial resolutions, coordinate reference systems, and spectral band combinations.
- IoT weather stations generate time-series data with fine temporal granularity.
- AI-driven detection algorithms output confidence scores, geolocated polygons, and classification attributes.
- Phenology-based early warning models rely on crop-specific BBCH codes or threshold-based degree-day calculations.
This diversity is both a strength and a challenge. While multiple data sources enrich the robustness of analyses, they inherently create data silos that prevent seamless integration, and hence datasets remain underexploited. Without harmonization, each subsystem might operate effectively in isolation, but their combined value – cross-validated detections, multi-modal insights, data-driven recommendations – would remain inaccessible.
This problem becomes even more pronounced when considering that STELLA spans multiple countries, agronomic contexts, and technological ecosystems. Data collected in France, Greece, Italy, Belgium, or New Zealand must be processed consistently to allow the platform’s centralized services to function reliably. Therefore, interoperability is not optional; it is foundational to the platform’s design and operation.
The STELLA Approach: Ensuring Semantic Interoperability Through Standardized Data Models
To break these silos, the STELLA PSS adopts a multi-layered strategy for ensuring semantic and technical interoperability. The first mechanism focuses on embedding interoperable and standardized semantics directly into the outputs of the pest prediction and detection systems. By aligning all datasets with widely recognized agricultural and geospatial standards, STELLA ensures that data from diverse technologies can be understood, compared, and combined without ambiguity.
Three semantic pillars support this approach:
1. EPPO Codes for Biological Consistency
Die European and Mediterranean Plant Protection Organization (EPPO) provides a standardized coding system for pests, pathogens, and plant hosts. Using these codes ensures that all subsystems refer to organisms in a consistent, unambiguous manner, regardless of language, country, or data source.
Instead of free-text labels like olive fruit fly or Bactrocera oleae, systems store a canonical EPPO code that can be universally interpreted. This avoids mismatches, typos, and inconsistencies – critical for automated processing and cross-dataset analytics.
2. GeoJSON for Geospatial Consistency
Given that pest surveillance is inherently spatial, ensuring geospatial interoperability is essential. GeoJSON is a lightweight, widely adopted geospatial data format that supports coordinates, polygons, bounding boxes, and feature collections.
By requiring detection systems (e.g., satellite-based anomaly detection) and prediction models to express outputs in GeoJSON, STELLA ensures:
- consistent coordinate handling
- smooth integration with web-based mapping frameworks
- compatibility with INSPIRE, OGC, and modern GIS pipelines
GeoJSON also supports additional semantic attributes, allowing confidence scores, timestamps, and phenological information to be embedded directly in spatial features.
3. BBCH Codes for Phenological Alignment
Crop development stage is a critical variable in pest monitoring. The BBCH scale provides a universal coding system describing phenological stages of agricultural crops. By incorporating BBCH metadata, STELLA ensures that detection and prediction results can be interpreted in the correct agronomic context—e.g., distinguishing between early infection risks during flowering versus late-season symptoms.
Together, EPPO, GeoJSON, and BBCH create a shared semantic backbone across systems, eliminating ambiguity and allowing automated pipelines to interpret and correlate data seamlessly.
Using DCAT Metadata to Facilitate Automated Data Processing
While standardized semantics ensure consistency within datasets, metadata is essential for enabling interoperability between datasets. STELLA adopts the W3C DCAT (Data Catalog Vocabulary) for describing datasets in a structured, machine-readable, and searchable manner.
DCAT metadata provides:
- dataset title, description, creator, and provenance
- spatial and temporal extent
- licensing and access constraints
- links to the actual data files
- thematic classifications and keyword tags
- references to EPPO codes, BBCH stages, or sensing technologies
By enriching every dataset with DCAT-compliant metadata, the platform enables automated workflows such as:
- dataset discovery in catalogs
- integrity checks and validation
- linkage between related datasets
- spatiotemporal indexing
- machine-to-machine API integration
- automated ingestion by analytics pipelines
This metadata-driven architecture supports FAIR data principles – Findable, Accessible, Interoperable, Reusable – ensuring that datasets are not merely stored, but can be meaningfully reused across the platform’s subsystems and by third-party services.
A Robust Data Repository: CKAN as the Backbone of Metadata and Dataset Management
To operationalize this metadata-first approach, STELLA employs a data management repository designed specifically for dataset cataloging and interoperability: CKAN.
CKAN is an open-source platform widely used in European and international open data portals. It supports DCAT natively, making it an ideal backbone for STELLA’s centralized repository. CKAN enables:
- hosting and versioning of datasets and metadata
- full DCAT and DCAT-AP compatibility
- API-based data discovery and retrieval
- role-based access control for different stakeholders
- integration with third-party applications and analytic engines
Through CKAN, STELLA gains a scalable and interoperable environment where datasets from satellite AI models, IoT weather stations, proximal sensing devices, and early warning systems can coexist harmoniously. CKAN’s metadata catalog ensures that every dataset – no matter its origin – is discoverable, interpretable, and usable by the platform’s analytics, visualization dashboards, and decision-support engines.
Conclusion
Breaking data silos is fundamental to the success of the STELLA PSS. The STELLA PSS is developed to facilitate pest surveillance in various agricultural contexts, so the objective is to be generic and easily expandable for monitoring additional geographical regions and types of pest infestations. To achieve this objective and to allow the integration of diverse digital monitoring systems with the STELLA PSS, widely accepted standards (EPPO, BBCH, and GeoJSON) will be used and the datasets will be provided via the CKAN-based data repository. The result is a robust, interoperable, and future-proof digital infrastructure that enables accurate pest surveillance, fosters cross-stakeholder collaboration, and drives informed decision-making across agriculture and forestry.
If you are interested in more technical details, we are happy to share and to discuss further. Just send us an email: .
