PicterraTrace’s six-month forestry demonstration ended in September after trials at eight sites. ESA’s project record, updated on 16 October, reports a revealing discovery: a satellite-based wood estimate disagreed with harvested material partly because the two measurements counted different things.[1]
The technology starts with geographic inputs. Picterra’s platform documentation describes importing imagery alongside plot boundaries, training models to detect objects or changes, and returning analysis through integrations with mapping and business systems. The boundary identifies the area being assessed. The model’s output then needs a definition that tells its recipient what the reported quantity represents.[3]
The timber measurement problem
In the harvest validation, the estimate covered total standing biomass; the measured output was merchantable wood. ESA says recognised conversion factors resolved the mismatch. Its public summary does not name those factors, so it cannot support reproducing that correction.[1]
Three quantities a forest report can contain
| Quantity | What it describes |
|---|---|
| Above-ground biomass | Living material above the soil, including stems, branches, bark and foliage. Normally expressed as dry mass.Source 2 |
| Merchantable volume | The stem volume suitable for sale. It excludes material outside that commercial definition.Source 2 |
| Carbon stock | The mass of carbon in a specified pool, such as living biomass or soil.Source 2 |
FAO’s glossary describes conversion and expansion factors that relate merchantable volume to above-ground biomass. They account for material beyond the commercial stem and can differ for standing stock, growth and removals. A conversion must therefore match the quantity being transformed. The factor for wood leaving a forest is not automatically the right factor for everything still growing there.[2]
This gives a buyer a useful first question when an AI estimate and a harvest record disagree: do they cover the same material? A model error requires one kind of investigation. A mismatch between an inventory boundary and a commercial specification requires another. The comparison becomes informative once its two sides refer to the same quantity.
Accuracy at the scale of the decision
The carbon checks were separate. At two sites, estimates agreed with independently certified inventories within a few per cent at management-unit level, according to ESA. The summary supplies no precise error measure or tree-level result.[1]
A total and a map answer different questions. As a simple illustration, overestimating one parcel by ten tonnes and underestimating another by ten leaves the combined total unchanged. Someone buying material from just one parcel would still care about the local error. This arithmetic is not a reconstruction of the pilot; it explains why the assessment’s geographic scale matters.
Pilot users identified integration with their geographic information systems as the principal adoption obstacle. ESA also reports commercial contracts before the demonstration closed, without disclosing their value.[1]
For a forestry team, the useful output is an estimate attached to the correct boundary, date, unit and inventory definition, delivered where the next decision is made. Those details let a timber buyer compare available wood with an order, or an inventory specialist compare successive measurements. They turn a number on a map into something that can be checked and used.
Sources & further reading
- Sustainable forestry with PicterraTrace ESA Space Solutions · Accessed 19 Oct 2026
- Appendix 1: Glossary of supporting terms Food and Agriculture Organization of the United Nations · Accessed 19 Oct 2026
- The GeoAI foundation behind real-world impact Picterra · Accessed 19 Oct 2026