Transparency Report

Four analysis presets are available. The fao preset is the Article 2(6)-aligned default applied to all standard analyses. The hansen_30 preset applies a 30% canopy threshold appropriate for analyses focused on closed-canopy humid tropical forest where open-canopy ecosystem conversion is not the primary risk. Operators may select an alternative preset at submission time.

* Agricultural-use exclusion applies to the fao and eudr_strict presets only — FDP commodity maps (cocoa, palm, coffee, rubber) ∪ WorldCover Cropland/Built-up (EUDR Art. 2(13)).

Extended table including GEE asset IDs, spatial resolution, DOIs, and known coverage gaps:

• Deforestation verdict (convergence rule): the final deforestation area is the refined union of Hansen GFW tree-cover loss and JRC-TMF post-cutoff transition, both intersected with the FAO forest mask (agricultural-use pixels already excluded). Loss pixels attributed to non-conversion by the causality engine (e.g. wildfire recovery, urban expansion) are subtracted from the union before the compliance verdict is issued. Degradation (JRC-TMF partial canopy loss without full transition) is tracked separately and does not affect eudr_compliant. Accuracy is validated through an independent per-country confusion-matrix assessment against held-out reference plots — not solely by internal reproducibility hashing.
• De minimis threshold: default 0.0 ha (strict). Any detected loss triggers a non-compliant verdict.
• Sentinel value handling: pixels returning −1 from GEE (out-of-coverage mask) are mapped to null and excluded from the aggregation.
• Soft-fail behaviour: if a dataset is temporarily unavailable, analysis continues on remaining sensors; the unavailable dataset is excluded from sensors_run.
• Shadow fields: computed in parallel but never influence the verdict or appear in sensors_run.
• Job idempotency: submitting the same geometry + commodity + cutoff date returns the cached result.

Shadow context signals — computed for every job but excluded from the compliance verdict and from sensors_run:

sbtn_natural_land_2020float | nullcontext signal

Mean fraction of the plot classified as natural land in the SBTN Natural Lands Map v1.1 (2020 baseline).

Range: 0.0–1.0, or null if out of coverage

Context signal only — not used in eudr_compliant or sensors_run. Useful for identifying plots that overlap with non-forest natural ecosystems (cerrado, wetlands, shrubland) where Hansen GFW tree-cover-loss detection is less relevant. Sentinel value −1 from GEE is mapped to null.

deuce_attribution_hafloat | nullcontext signal

Hectares of deforestation attributed to this commodity in this country in the most recent available year, according to the DeDuCE dataset.

Range: ≥ 0.0, or null if country/commodity not covered

Country-level signal derived from national trade statistics. Does NOT measure deforestation on the specific plot — it is a national context indicator. Commodity normalisation: palm_oil → palm, cattle_beef → cattle. Source: Trase / Chalmers University, doi:10.5281/zenodo.10674962.

deuce_attribution_yearint | nullcontext signal

Calendar year of the DeDuCE attribution figure in deuce_attribution_ha.

Range: 2001–2022, or null

Always paired with deuce_attribution_ha. null when deuce_attribution_ha is null.

fdp_cocoa_prob_2020float | nullcontext signal

Mean cocoa suitability probability for the plot at the 2020 snapshot (Forest Data Partnership).

Range: 0.0–1.0, or null

Context signal. High probability indicates the plot is climatically and geographically suitable for cocoa cultivation. Does not confirm cocoa is being grown.

fdp_cocoa_prob_2023float | nullcontext signal

Mean cocoa suitability probability for the plot at the 2023 snapshot.

Range: 0.0–1.0, or null

Paired with fdp_cocoa_prob_2020. Difference between 2020 and 2023 indicates suitability change.

fdp_coffee_prob_2020float | nullcontext signal

Mean coffee suitability probability for the plot at the 2020 snapshot.

Range: 0.0–1.0, or null

Context signal only — not used in compliance verdict.

fdp_coffee_prob_2023float | nullcontext signal

Mean coffee suitability probability for the plot at the 2023 snapshot.

Range: 0.0–1.0, or null

Context signal only — not used in compliance verdict.

fdp_palm_prob_2020float | nullcontext signal

Mean oil palm suitability probability for the plot at the 2020 snapshot.

Range: 0.0–1.0, or null

Context signal only — not used in compliance verdict.

fdp_palm_prob_2023float | nullcontext signal

Mean oil palm suitability probability for the plot at the 2023 snapshot.

Range: 0.0–1.0, or null

Context signal only — not used in compliance verdict.

fdp_rubber_prob_2020float | nullcontext signal

Mean rubber suitability probability for the plot at the 2020 snapshot.

Range: 0.0–1.0, or null

Context signal only — not used in compliance verdict.

fdp_rubber_prob_2023float | nullcontext signal

Mean rubber suitability probability for the plot at the 2023 snapshot.

Range: 0.0–1.0, or null

Context signal only — not used in compliance verdict.

fdp_declared_cropstring | nullcontext signal

The commodity declared by the operator at job submission, as used to select the FDP probability layers.

Range: cocoa | coffee | palm_oil | rubber | null

Stored for audit trail purposes. Determines which fdp_*_prob_* pair is populated; other commodity pairs will be null.

carbon_co2e_loss_mgfloat | nullcontext signal

Estimated CO₂-equivalent carbon stock lost within the plot boundary, in megagrams (tonnes).

Range: ≥ 0.0, or null

Derived from above-ground biomass density (Spawn et al. 2020) intersected with detected forest loss area. Context signal — not used in compliance verdict. Uncertainty ±20–40% typical.

jaxa_fnf_2020float | nullcontext signal

Mean JAXA PALSAR-2 Forest/Non-Forest classification score for the plot at the 2020 epoch.

Range: 0.0–1.0, or null

SAR-derived forest classification. Context signal for dense humid tropical forest presence. Sentinel −1 mapped to null.

esa_worldcover_classstring | nullcontext signal

Modal ESA WorldCover land cover class within the plot at the 2020 snapshot.

Range: Tree cover | Shrubland | Grassland | Cropland | Built-up | Bare / sparse vegetation | Snow and ice | Permanent water bodies | Herbaceous wetland | Mangroves | Moss and lichen | null

Modal class across all 10 m pixels within the plot. Context signal — not used in compliance verdict.

gedi_canopy_height_mfloat | nullcontext signal

Mean GEDI canopy height within the plot, in metres.

Range: 0.0–70.0, or null

Useful for verifying FAO 5 m minimum tree height criterion. Sparse GEDI sampling outside ±52° latitude means null is common at higher latitudes. Context signal only.

• ·Cerrado and chaco: conversion of open-canopy natural ecosystems (cerrado, chaco, savanna) is captured primarily by the SBTN Natural Lands Map and MapBiomas, not by Hansen GFW, which detects tree cover loss only.
• ·SE Asia peatland: peatland loss is captured by KLHK for Indonesia only. No equivalent national dataset is currently available for Malaysia or Papua New Guinea.
• ·JRC TMF artefacts: known cloud shadow artefacts in equatorial Africa may produce spurious degradation signals in a small fraction of plots.
• ·No forward-looking model: all analysis is retrospective against the 31 December 2020 cutoff. No predictive deforestation model is currently active.