Forest screening and field review
Forest screening is an additional signal for reviewing a production plot. It helps organize evidence for follow-up; it is not a legal determination or EUDR clearance.
Point or polygon: the 4 ha rule
For regular EUDR geolocation of crop plots, such as cocoa or coffee, one location point is sufficient up to and including 4 ha, using latitude and longitude with at least six decimal places. Above 4 ha, a polygon with sufficient boundary points is required. Cattle geolocation refers to the establishments where they were kept. Due diligence does not impose boundary polygons on every smaller plot; deforestation-free and legal production must still be substantiated. Source: EUDR, Article 2(28) and Article 9.
The app distinguishes between checking a location point and checking a boundary. For a point, check that it identifies the correct production plot. Satellite screening uses estimated surroundings based on the point and declared area, not a surveyed field boundary. A polygon can improve spatial analysis for smaller plots too, without being a blanket additional EUDR requirement.
Under Plot boundary, enter coordinates as latitude, longitude, for example 6.123456, 1.234567. Use coordinates as point sets that exact position without another map click, including offline. Address search and new satellite images need a connection. Search only positions the map. An existing polygon must be explicitly deleted before replacing it. Then enter the plot size (at most 4 ha) and save. GeoJSON stores values in reverse order: [longitude, latitude].
Screen several plots together
Under Origin → Plots, select individual plots or all plots on the current page. Selection persists across list pages. Select all matching plots also includes later pages and respects the current search and producer filter. Individual plots can then be excluded.
Screen selection creates a durable run that continues in the background when the page is closed. Queued work for the same plot version is reused. Membership and versions are fixed at the start; later additions are not silently included.
The report shows No signal, Needs clarification, Loss indication, In progress, Not assessed and Boundary changed since start. Click a category to list its plots. These are satellite findings, not legal approvals. Changed boundaries cannot inherit a green result from an earlier version. Reopen saved reports under Screening runs. Outstanding dispatch can be resumed after a queue outage.
After a plot import, Screen plots from this import includes newly created plots and existing plots recognised as duplicates in the file. Other skipped or invalid rows are excluded. Importing alone does not start satellite requests. Explicit selection is limited to 1,000 plots per run; all matches is limited to 20,000. Narrow larger selections.
Identify the plot and record evidence
Individual screening and boundary editing have dedicated pages; the plot list no longer remains underneath a review. Change boundary opens the editor and returns to screening. A new boundary needs a new screening.
Plot identification means comparing the saved point or polygon with reliable existing field maps, GeoJSON/GPS surveys or supplier records. For a point geometry, verify its location first. A new site visit is not a blanket extra app step; unclear or conflicting evidence may require further investigation. What else needs checking — and how? links to the map, editor, historical sources, evidence and review record.
Land-use history, legal production and the link to the actual product or delivery also need assessment. Depending on role and supply chain, due diligence covers information collection, risk assessment, necessary mitigation and the required EU-system declaration. The guide is orientation, not a fixed legal checklist. Source: European Commission: due diligence and roles.
Saved geometry supports overlay, side-by-side at one shared scale and individual versions. The previous dashed outline is drawn over the current fill, including in the print report, so overlap does not hide it.
Traffic light and interpretation
The colour describes automatic satellite screening. The plot list, result and print report use the same logic. Manual review is documented separately.
| Colour | Meaning | Next step |
|---|---|---|
| Green | No Hansen loss pixels in 2021–2025 with sufficient data | Check the location or existing boundary, cut-off land use and origin evidence; document review |
| Amber | Loss without JRC forest overlap, incomplete data or missing measurements | Resolve the loss indication or data gap using dated imagery and field evidence |
| Red | Loss overlaps JRC forest mapped for 2020 | Investigate timing, cause and subsequent use |
New screens require at least 98% usable coverage for JRC and Hansen land data. This is a technical data threshold, not a legal deforestation tolerance. Missing data is not zero loss. Optional WorldCover or historical-canopy gaps do not erase the measured loss result. Older stored results remain unchanged and explicitly show their unrecorded Hansen coverage; an old zero-loss result may be shown as favourable without inventing a retrospective coverage measurement.
Example: 88% tree class, 0 ha JRC forest and zero loss pixels. With sufficient data this produces green screening. WorldCover includes tree plantations, and perennial woody crops may be mapped as trees or shrubs. The 88% is an area-class share, not canopy closure or the EUDR forest share. Confirm the crop and land-use history.
100% coverage means all considered raster points or area fractions contain usable values, not 100% accuracy. JRC counts valid raster points; Hansen compares valid mapped-land area with the sampled footprint inside the geometry. Its 2000–2012 data mask does not measure observation quality in every loss year. WorldCover uses valid classified areas; class shares under partial coverage refer to the usable portion only.
New screens add loss pixels by year (2001–2025), pre-2021 loss counts and mean canopy closure in 2000. These come from the same Hansen product, are not independent confirmations and do not establish forest status in 2020. Missing values appear as —. Run a new screening to obtain these additional measurements. Annual loss data ends in 2025; changes from 2026 are not assessed.
Technical source: Hansen/UMD dataset definitions.
Start a screening
- Open Origin → Plots and find the production plot.
- Select Forest screening. The job runs in the background.
- Keep the assessment view open or reopen it later to see the completed result.
- Check that the saved plot version and geometry match the location in the field.
Each assessment is bound to an immutable plot version. Start a new assessment after changing a boundary. If the saved geometry is a point, Via does not show a true field boundary: the raster check derives a screening footprint from the point and declared area.
Read the automated signals
The assessment combines the JRC 2020 forest cover layer with Hansen tree-cover-loss data for 2021–2025. Overlapping loss pixels are reported as a Loss candidate. They prompt review, but do not establish agricultural conversion, its cause or its legality. No match in the evaluated data does not prove that no risk exists. Incomplete data coverage cannot support a reliable no-match conclusion.
ESA WorldCover 2021 is included separately as a 10 m land-cover map. The assessment shows its data coverage and the percentage of each mapped class. The plot list displays the mapped cropland and tree-cover shares as a brief hint. These values describe mapped land cover in 2021; they do not show when a conversion happened.
The ESA product manual reports independently validated overall accuracy of 76.7 ± 0.5% globally and 76.5 ± 1.3% for Africa for WorldCover 2021 v200. These are aggregate validation figures, not an accuracy guarantee for an individual field or class. The product can be used free of charge under CC BY 4.0; the ESA product manual requires attribution, which Via includes in its report.
The EUDR defines forest using area, tree height, canopy cover and predominant land use. For the EUDR definition of deforestation-free, relevant products must be produced on land that has not been subject to deforestation after 31 December 2020. A WorldCover class such as Cropland or Tree cover does not test all of the legal criteria. In particular, a 2021 cropland class does not establish that the land was not forest before the cut-off date; a tree-cover class is not automatically EUDR forest. WorldCover is useful supporting information, but is not enough on its own to determine EUDR status. See the current regulation text and consider all available evidence.
How EU checks work and where its map data come from
The EU does not issue a blanket clearance based on one global map. The market operator must collect the geolocation of production plots and carry out due diligence. Competent authorities in EU Member States check due diligence statements and overall compliance. They may cross-check plot coordinates against satellite imagery or forest-cover maps; the Regulation names Earth-observation data such as Copernicus as a possible source. In relevant cases, field audits in third countries are also possible with the country’s cooperation.
The EU Observatory combines existing monitoring sources, including Copernicus and other public and private data. Its global forest-cover map for the 2020 cut-off was created to support risk assessment and maps forest presence at 10 m resolution. The Commission describes Observatory maps as voluntary, non-exclusive and non-binding. No single map is prescribed; authorities and operators can use more suitable national or regional data. The Commission also cautions against treating a map as final proof of compliance: plots outside mapped forest areas may still be checked.
This explains the difference from WorldCover. WorldCover is a valid and useful 2021 land-cover signal, but it does not show forest status on the 31 December 2020 cut-off date, when a conversion happened or whether it was legal. It can help interpret a plot and plan a field review; a decision must consider other relevant evidence.
Sources: European Commission EUDR FAQ, especially 1.19, 1.20 and 9.10 · EU Forest Observatory: global forest maps.
Conduct a human review
In Manual review, compare the saved geometry, the previous plot version and a satellite view. Open supplementary sources that fit the region:
- JRC Tropical Moist Forest maps annual change in moist tropical forests; dry forests and savannas are not fully covered.
- GLAD-L and RADD provide forest alerts with different sensors, coverage and time limits. An alert is a clue to examine.
- BNETD provides a national land-cover source for Côte d’Ivoire; it does not cover southern Burkina Faso or Togo.
- WorldCover is already included in the automated result as 2021 land cover.
Save your interpretation with a rationale. Signal confirmed, No supporting evidence and Inconclusive record the review without changing the machine result. Reviews remain in the history and cannot be overwritten.
Print or save the report as PDF
Select Print report / save as PDF, then choose Save as PDF or a printer in the browser dialog. The report includes the plot ID and version, screening metrics, WorldCover shares, candidate locations, sources, a geometry preview, review history, limitations and the result checksum.
The geometry preview is a schematic of the saved screening geometry, not satellite imagery. If the saved geometry is only a point, the report shows that point instead of inventing a field boundary.
Dated Sentinel-2 plot crops can be compared side by side and retained as private image evidence. Acquisition date, geographic bounds, exact plot version and checksum are retained. Access to the external catalogue and public image archive requires explicit confirmation. Clouds, shadows and land use still require professional interpretation.