Deforestation verification · EUDR
We verify farm-level deforestation risk for companies and consultancies preparing for the EU Deforestation Regulation. Send us your supplier plot data. We check every plot against independent forest datasets and return a documented evidence package — including what we could not determine, and why.
The problem
The regulation asks a narrow question: was this land forest at the end of 2020, and has it been cleared since? Answering it at portfolio scale fails for two reasons that have nothing to do with imagery.
A farm over 4 hectares recorded as a single dropped pin. A boundary that crosses over itself. The same field submitted twice under two cooperative names. Coordinates that land in the wrong country, or in the sea.
None of it can be analysed, and none of it gets better by being ignored. Someone has to go back to the supplier and get it fixed.
Cocoa and coffee are often grown under shade trees. From above, a farm that has been a farm for thirty years reads as closed canopy — and a global forest layer calls it forest.
Clearing that flag is not a matter of asserting it. It takes an imagery timeline, a vegetation record, and a written argument that survives someone disagreeing with it.
Services
Most engagements start with an assessment and continue as monitoring. The rest are bought when they're needed.
Every plot in a supply chain checked against independent forest datasets, with geometry problems separated from genuine findings. You get an evidence package, a list of exceptions, and a stated set of limitations.
A plot flagged as forest that your supplier says has always been farmland. We build the imagery timeline from 2018 onward, read the vegetation record, and write the rebuttal — or tell you the evidence doesn't support one.
We go back to the cooperative with a single link that shows only their own plots and names the problem in plain language, on a phone, in their language. Then we chase until it's corrected.
Scheduled re-checks measured against the previous cycle rather than from scratch. You hear about what changed and what it means — not about the plots that didn't.
Due diligence statement data prepared from a completed assessment, reference numbers recorded against the order, and the evidence behind any past order rebuildable on request — including the rule version that applied at the time.
Method
This is the part worth reading before hiring anyone for this work. A verdict is only as good as what it can show about how it was reached.
Each plot is tested against three independent, openly licensed forest datasets, and they are not interchangeable votes. Whether the land was forest at the end of 2020 is answered by the EU’s own JRC baseline — the only one of the three that can answer it, because the ESA layer’s reference date falls after the cutoff and the UMD layer reports loss rather than cover. UMD dates when loss occurred; ESA corroborates what is on the ground now. A plot clears without a person only when the baseline shows non-forest at the cutoff and no loss is dated after it. Forest at the baseline, loss dated after it, or a layer that was unavailable all send the plot to a person.
The FAO thresholds the regulation relies on: canopy cover of at least 10%, height of at least 5 m, minimum area 0.5 ha. Those are parameters of a versioned rule, not assumptions buried in code, so an order can be reproduced later under the rule that applied when it was issued.
At 9.3 m per pixel the smallest area the data can resolve is about 0.03 ha. A plot too small for the imagery behind it returns inconclusive and says so, rather than producing a verdict the pixels cannot support.
If a dataset was unavailable for a country, or a boundary fell outside its coverage, the report names what was not assessed. Silence is the failure mode that ends a compliance vendor, so we don't allow it.
People do not look at every plot; they look at the ones that need judgement. Every consequential finding is reviewed before it's delivered, and the original machine finding is kept alongside the correction rather than overwritten.
Which datasets were consulted, the pinned version of each, what each one showed, when it was retrieved, and which rule version applied. Enough for someone else to rebuild the conclusion — or to challenge it properly.
| Reference dataset | Pinned version | What it contributes |
|---|---|---|
| JRC Global Forest Cover | GFC2020 · V4 | The EU's own 2020 baseline — the reference for the cutoff date. |
| UMD Global Forest Change | GFC-2025-v1.13 | Annual loss year — the only layer that dates its own loss. |
| ESA WorldCover | v200 | Land-cover corroboration for its own reference date. |
Who we work with
You hold the client relationship and own the due diligence system. We supply verification capacity behind it, without hiring against every new account. Deliverables can carry your branding, and we stay out of your client conversations unless you put us in them.
You file the statement and you carry the responsibility. We produce the plot-level evidence that sits behind it, in a form you can hand to a regulator, an auditor, or a customer asking how you know.
What we don't do
A verification service that will say anything is worth nothing. These are the lines we hold, including when a client would rather we didn't.
Request an assessment
Engagements are scoped by portfolio size, commodity, sourcing countries and cadence. Six questions get us to a real number instead of a range.