SYLVAN-AI
Chat-enabled forest management, grounded in real European geodata
- Programme
- Horizon Europe
- Open call
- PoliRuralPlus Enhance
- Grant agreement
- No. 101136910
- Duration
- Six-month action
- Domain
- Forestry / GeoAI
- Status
- Delivered, publicly accessible
Overview
SYLVAN-AI is a chat-enabled forest manager that puts the PoliRuralPlus JackDaw tool-using GeoLLM behind a single conversational interface. It was selected under the PoliRuralPlus Enhance Open Call to validate JackDaw in a real, high-complexity European domain, forestry, over a six-month action.
Europe's forests are largely managed by small-scale owners who lack the GIS tools or training to meet EUDR and Green Deal reporting obligations. The spatial data needed to manage a forest plot exists, cadastre, Natura 2000, terrain, soil, fire risk, land cover, but is scattered across separate national geoportals. SYLVAN-AI removes that barrier: a user draws a plot on a map and asks, in plain language, “Are there Natura 2000 areas on my land?” or “What should I plant here?”, and receives a spatially accurate, source-cited answer in seconds.
The project delivered a working, publicly accessible platform, secured behind PoliRuralPlus single sign-on.
Results
Hybrid retrieval: reasoning on top of exact spatial data
The defining technical contribution is a hybrid retrieval design. JackDaw's GeoLLM reasoning and tools are combined with a purpose-built local PostGIS grounding layer holding 265 GB of standardised EU geodata. Every drawn-parcel query is grounded in exact spatial data, and a no-hallucination guardrail forces the assistant to say “I don't have data for this area” rather than invent a value.
| Layer | Coverage |
|---|---|
| Cadastral parcels | 442.3 million parcels across 27 countries |
| EEA Natura 2000 | 27,173 protected sites |
| Copernicus DEM GLO-30 | Terrain and elevation |
| ISRIC SoilGrids | Soil properties |
| NASA GFWED | Fire-weather conditions |
| Copernicus CORINE Land Cover 2018 | Land cover, with the 2012–2018 change layer |
| Czech national forest typology | 949,358 site polygons, 2.72 million hectares |
Five guided user journeys
Each journey is live in the product and mapped to a real forestry decision, so the assistant answers the question a user actually has rather than the one the data happens to fit.
- The Compliant OwnerMeeting EUDR and Green Deal reporting obligations without GIS training.
- The Climate-Aware PlannerDeciding what to plant, given terrain, soil and a changing climate.
- The InvestorAssessing a plot before committing capital to it.
- The ConservationistUnderstanding protected habitats and what they permit.
- The Carbon EstimatorPutting a defensible number on a plot's carbon position.
Built to be evaluated, and to respect privacy
- A built-in validation frameworkA standardised PoliRuralPlus form with System Usability Scale scoring, an accuracy-audit dashboard, a Golden-Set benchmark harness and session recording, so the platform is instrumented for rigorous, comparable evaluation rather than assessed by impression.
- GDPR by designPersonal data is stripped at the input boundary, before any text reaches the language model.
- A resilient architectureThe platform degrades gracefully when upstream services are unavailable, rather than failing outright.
- Feedback back into JackDawA systematic endpoint audit produced concrete, actionable feedback to the JackDaw team, converted into an itemised hardening backlog.
Benchmarked against the geoportals it replaces
Annex B documents five success cases in Czech, Greek, French, German and Spanish forests. Each is benchmarked step by step against the public geoportals a user would otherwise have to consult: about 296 minutes of work by traditional means against roughly 11 minutes in SYLVAN-AI, a time-to-insight reduction of some 96%, with every individual case above 90% and all of them far beyond the 50% target.
Launch and reception
SYLVAN-AI was launched in June 2026 and shown at the PoliRuralPlus JackDaw Code Camp on 16–18 June, including to FPG and the European Forest Institute (EFI). The action met its core objectives and validated them directly with stakeholders, and reception was consistently positive, including direct feedback from those stakeholders.
What it demonstrates
SYLVAN-AI shows that conversational, grounded GeoAI is a credible route to democratising spatial intelligence for the people who manage Europe's land, while advancing the New European Bauhaus values of sustainability, inclusion and aesthetics and aligning with the EU Deforestation Regulation and Green Deal.
Funding
PoliRuralPlus has received funding from the European Union's Horizon Europe research and innovation programme under Grant Agreement No. 101136910. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.