The 31 December 2020 cutoff, and why tree-cover loss is not deforestation

A plot converted from forest to agricultural use after this date can never produce EUDR-compliant goods. There is no remediation, no grace period and no offset. The plot is out, and so is everything grown on it.

Why that date

It is deliberately retrospective. The regulation entered into force in June 2023, but the cutoff sits two and a half years earlier, so that the announcement of the law could not itself trigger a wave of pre-emptive clearing. The practical effect is that you are being asked to evidence land-use history from before anyone knew they would need to.

This is the root of the whole compliance problem. Nobody was keeping plot-level records in 2020 for a regulation that did not yet exist, which is why the evidence has to come from satellite imagery rather than from documents.

What counts, and what does not

• This is deforestation Forest converted to cropland or pasture after the cutoff Conversion that nobody intended, forest lost to fire, then farmed Clearing by a previous owner or occupant of the land For wood only: converting primary or naturally regenerating forest into plantation forest. This is degradation, and it also breaches the regulation
• This is not deforestation Harvesting a plantation that is replanted and stays forest Storm, disease or fire damage where the land remains forest Selective logging that does not convert the land use Any clearing completed before 31 December 2020, however recent Conversion of land that was not forest to begin with

The two definitions doing the work are narrow. Deforestation is the conversion of forest to agricultural use, whether or not human-induced. Forest degradation applies to wood and covers structural change, primary or naturally regenerating forest becoming plantation forest, or becoming other wooded land.

Note what is absent from the deforestation definition: intent, and the identity of whoever did it. If the land was forest on 31 December 2020 and is farmland now, the commodity fails the test. It does not matter that the current farmer inherited the situation.

Why your screening will flag plots that are fine

Satellite systems detect tree-cover loss. The regulation prohibits deforestation. These overlap heavily but they are not the same measurement, and the gap between them is where most of the practical work sits.

A canopy-loss alert tells you that trees that were there are no longer there. It does not tell you whether the land use changed, whether the trees grew back, whether the loss was a managed harvest, or whether the area met the definition of forest in the first place. A coffee plot that shaded out and was replanted can trigger an alert. So can a windthrow event. So can cloud artefacts and sensor error.

An alert is an obligation to look, not a verdict. The correct response to a flagged plot is investigation, not exclusion. Excluding every alerted plot will cut out compliant farmers, often smallholders with the least ability to argue back, and will not make your file more defensible. What makes it defensible is showing that you looked and recorded what you found.

How to resolve a flagged plot

1. 1.Confirm the geometry. A surprising share of alerts are boundary problems: the polygon includes a neighbouring parcel, a road, or a river margin. Check the plot outline before you check the forest.
2. 2.Establish the baseline. What was on that land on 31 December 2020? This is the question that matters, and it is answered with imagery from around that date rather than with recent alerts.
3. 3.Characterise the loss. Date it, size it, and look at what followed. Regrowth within a season or two points to harvest or natural disturbance. Conversion to visible crop rows or pasture points the other way.
4. 4.Corroborate on the ground. Ask the producer or cooperative what happened. A photograph, a planting record or a local explanation is legitimate evidence and often resolves an alert quickly.
5. 5.Record the determination. Write down what you found, what you concluded, and why, with the imagery attached. This record is your compliance position. An undocumented correct conclusion is worth nothing in an inspection.

Data sources that carry weight

No single source is authoritative, and the regulation does not mandate one. In practice a defensible screen combines several.

• The EU forest baseline: the Commission's own map of forest cover at the cutoff, which is the reference an inspector is most likely to reach for.
• WHISP (Whisp for EUDR), an open convergence-of-evidence tool that combines multiple global datasets into a single plot-level assessment, developed specifically for this regulation.
• Sentinel-2 imagery: 10 m optical data with a revisit of a few days, good enough to date a clearing event and to see what replaced the trees.
• Global forest-change datasets: useful for detecting loss, weaker at distinguishing harvest from conversion. Read them as alerts, not conclusions.

Where sources disagree, and they will, the disagreement itself is evidence you have to resolve and record. Convergence across independent datasets is what makes a negligible-risk conclusion credible.

The cutoff and the legality test are separate

A plot can be entirely forest-free since 2020 and still fail. The second condition in Article 3 requires production in accordance with the relevant legislation of the country of production, land tenure, environmental law, labour and human rights, tax and customs. Clean satellite history does not answer that question, and satellite data cannot.

Related

• Geolocation: coordinates, polygons and the four-hectare rule
• The Article 10 risk assessment
• What is the EUDR?