What we check, how we reach a result, and what the result cannot tell you.
Where the EU publishes a reference map or definition, we use it. Each row names the source.
| EU reference | What we do | Source |
|---|---|---|
| Cutoff date | We look for change after 31 Dec 2020. | Regulation 2023/1115, Art. 2(13) |
| Forest definition | Our baseline is the Commission's 2020 forest map, built to the EU forest definition. | EU Forest Observatory, GFC2020 |
| Forest types | Primary, regenerating and planted forest are shown as context. | EU Forest Observatory, Art. 2(8) to 2(11) |
| Degradation | It changes the result for wood and timber only. | Regulation 2023/1115, Art. 2(13)(b) |
| Several datasets | We add Hansen and JRC change data to the 2020 map, as the JRC advises. | JRC guide to GFC2020 |
The EU does not prescribe one tool, and authorities may use other data. This is a screening, not a forecast of an authority's conclusion.
We define the area, compare it with the forest baseline and later change data, apply consistent rules, then save the evidence with the result.
A boundary is used as the plot area. A point is checked within a 500 m surrounding area. If the location is coarse, suspect or unusable, it is not presented as plot-level precision.
Location precision The area that can be assessed depends on the geometry provided.
The Commission's GFC2020 map is our reference for likely forest extent in 2020. The system then checks later change in Hansen tree-cover loss and JRC Tropical Moist Forest data. Imagery supports review.
Data compared The same area is checked against the 2020 forest map and the change layers after it.
Before cutoff
After cutoffThe result follows the checked change layers, not a subjective score. A signal must cover more than 0.5 hectares before it changes the result. This is our own screening threshold, not an EU-prescribed one.
Context fields Recent alerts and country risk are useful context, but they do not create a Finding or Caution.
Every result keeps the submitted location, checked area, outcome, reason, imagery and data-coverage note. It also records the screening date and the methodology and dataset versions used.
Re-screening A later check creates a new record. The earlier result stays intact.
Screening record
Each result says what we found in the submitted area, shows the supporting evidence, and suggests the next useful step.
Tree cover was lost after the cutoff on land the EU's 2020 forest map shows as forest. It is a screening signal, not confirmed conversion to agriculture.
Confirm that the boundary is the production plot. Then review the detected area in the imagery alongside supplier and traceability records.
Either a wood or timber plot shows degradation, or tree cover was lost where the EU's 2020 forest map shows no forest. That map can miss forest.
Check the imagery and the land-use history of the plot. For wood, review the supply chain too.
No signal does not mean compliant. It means no deforestation signal, or for wood and timber no degradation signal, passed the rule for the area in the datasets assessed.
Keep the result with supplier, origin and traceability evidence. A later check can be run if the location is refined or newer inputs become available.
There was not enough usable location information to assess an area. This happens when no point or boundary is supplied, or when the location cannot be used.
Add or correct the point or, preferably, the plot boundary. A new check will create a separate result record.
For wood and timber, forest degradation affects the EUDR result. For other commodities, the same signal remains visible as review context. It may be worth assessing, but it does not become a Finding or Caution.
Review it with the imagery and your supply-chain evidence. It stays visible instead of being hidden behind No signal.
It is a careful snapshot of 2020, not a perfect one. It finds about 92 of every 100 forest areas, and it leaves out cocoa, coffee, palm oil, rubber and soy plantations on purpose.
So when trees vanish where the map shows no forest, we do not ignore it. We mark it as a Caution and ask you to check.
Every dataset we use is public. You can open the same 2020 forest map, look at your plot, and compare it with recent satellite images.
A result is useful evidence, but it is not the whole due-diligence picture.
Shade-grown canopy, agroforestry and cloud can make change harder to interpret. The main change layers update annually, so very recent clearing may not yet appear. Tree loss is not proof of conversion to agriculture.
Interpretation noteBefore-and-after imagery and a Finding help focus the next investigation.A boundary is the best way to represent a plot. A point is checked within 500 m around it, and a signal is not tied to the exact point. An address represents a wider area.
Interpretation notePlot boundaries provide the most precise result. Coarse locations are marked so they are not mistaken for plot-level results.The satellite check assesses forest change at a location. It does not show product origin, legality, land rights or supplier records. That evidence comes from you and your suppliers.
How it fitsKeep the result with your supplier, traceability and legality records as part of your due-diligence file. The workspace records your own legality assessment under the eight headings of Article 2(40), with your evidence and signed conclusion.We use named public datasets. Every export records the methodology and dataset versions used for that specific result.
| Source | Publisher | How we use it | Effect on the result | Update cycle |
|---|---|---|---|---|
| JRC Global Forest Cover 2020 V4 | EU Commission (JRC) | 10 m forest map, reference year 2020 | Splits loss into Finding or Caution | 2020 baseline |
| Hansen Global Forest Change 2025 v1.13 | University of Maryland | Annual tree-cover loss | Finding on 2020 forest, otherwise Caution | Annual |
| JRC Tropical Moist Forest v1_2025 | EU Commission (JRC) | Deforestation and degradation | Finding; Caution for wood | Annual |
| JRC Global Map of Forest Types V1 | EU Commission (JRC) | 2020 forest context | Context only | 2020 context |
| RADD and GLAD alerts | Wageningen University, University of Maryland | Recent nearby pressure | Context only | Alert feeds |
| WWF HydroSHEDS Free-Flowing Rivers V1 | WWF | Watercourse proximity | Context only | Static reference |
| EU country benchmarking (Regulation 2025/1093) | EU Commission | Country due-diligence context | Context only | Reviewed when updated |
No single method exists. The regulation lets authorities use any suitable data. We use the Commission's own 2020 forest map and definitions and the JRC's advice to combine datasets. An authority may still reach a different view.
Post-cutoff tree loss passed the screening threshold on land the EU's 2020 forest map shows as forest. Start with the plot imagery, submitted location and supply-chain evidence. It is not a legal finding.
EUDR treats forest degradation separately for wood and timber, so that signal is a Caution. Tree loss where the 2020 forest map shows no forest is also a Caution, because that map can miss forest.
No signal means we checked the location and no change signal passed the rule. No location means there was not enough usable location information to run the check.
Coordinates must be valid and are kept to six decimals, no two plots may share a location, and outlines must be closed and must not cross themselves. A single point is accepted for plots up to 4 ha, and for cattle. We do not check that a location lies on the supplier's farm or that an answer is true: a plausible but wrong location passes every check, so keep supplier and traceability evidence with the result.
Use a boundary whenever you have one. It shows the area under review; a point identifies a location but cannot show the full extent of a plot.
They do not overwrite them. Each export keeps the methodology and dataset versions used when that check was run.