Monitoring protected nature is, at scale, mostly a screening problem. Municipalities are responsible for large areas, field visits are expensive, and working through maps and aerial photos by hand takes time. Changes are often caught late, or only when someone reports them.
VERNA does that screening automatically. Nature is notoriously complicated, and no single model handles all of it well. What sets VERNA apart is that it does not rely on one. It runs several deep-learning architectures side by side, for segmentation, change detection, embedding-based comparison and vegetation-height estimation, and turns what they find into results a caseworker can review and act on.
The inputs are aerial orthophotos and near-infrared imagery, airborne LiDAR and elevation models, and Sentinel imagery. The output is a shortlist of places where protected nature has probably changed: new water bodies, terrain that has been dug out or filled in, shifts in vegetation, habitats that are not registered yet.
That last step is the hard one. A prediction on a map is not useful by itself. A caseworker has to see why an area was flagged, compare it against earlier years, record an assessment and stand behind it later. VERNA is built around that whole process. In Denmark it has already identified more than 2,000 potential changes and 2,500 potential new ponds and lakes across 15 municipalities.
The goal is not to replace fieldwork. It is to make fieldwork more targeted. When authorities know where change is most likely, limited field resources can be used on the places where they matter most. That means earlier detection, better documentation and a more consistent basis for prioritising nature protection and restoration.
Denmark is a good place to build this because protected nature is already mapped in a structured way. The registrations of §3 nature, protected-area boundaries and habitat types give us unusually strong reference data to train models on and validate against. That is what deep learning needs to work well. It also lets us do something most monitoring cannot: cover large areas at high resolution and repeat it.
We have already taken VERNA beyond §3 nature into tree and canopy mapping. This works at the level of the individual tree: each tree is detected, located and measured. Combined with canopy-cover estimates, this makes 3-30-300 assessments more practical, because municipalities need both tree locations and canopy information to assess the rule consistently.
The next step is habitat screening for Annex IV species – the strictly protected species that municipalities must account for in planning and casework. Here, the model does not replace field surveys. It predicts where suitable habitat is likely to occur, so a municipality knows where to look first.
There is a reason to do this now. The EU Nature Restoration Regulation has been in force since August 2024, and every member state has to submit a draft National Restoration Plan by September 2026. For urban ecosystems, the regulation also makes baselines unavoidable: Member States must ensure no net loss of urban green space and urban tree canopy cover by 2030, compared with 2024. All of it depends on a baseline you can defend and monitoring you can repeat over large areas.
Denmark is a good place to prove this, but the problem is not Danish. Across Europe, authorities are being asked to establish baselines, monitor change and prioritise restoration with limited capacity. The local rules, habitats and data differ, but the operational challenge is familiar: where should people look first?
That is where we think VERNA can travel. The approach is transferable, but it has to be adapted locally: finetuned on national data, mapped to local habitat definitions and validated with people who understand the landscape. We are looking for partners who can bring that local knowledge; authorities, EO companies, research groups and organisations working directly with nature monitoring.
About Koordinat
Koordinat is a Copenhagen-based AI and geospatial technology startup. We build tools that help authorities use Earth Observation data to monitor nature, identify changes and prioritise fieldwork.
Contact
Thomas Lykke Rasmussen
Co-founder, Koordinat
