What the plots show: declarations and mapped observations
Five of the 17 supplied plots carry a supplier declaration of an "indication of deforestation". Hansen and JRC map post-2020 loss on different plots. Two variants (an older Hansen release, a different cell-inclusion rule) do not reconcile the two, and the comparison does not establish the cause.
- Five plots carry a declared indication, recorded mitigation with evidence, and a compliant final status: P-0003, P-0005, P-0008, P-0010, P-0013. These are declarations.
- Hansen alone maps loss after 2020 on 7 of 17 plots. With JRC's 2020 forest map, only P-0001 has mapped loss on mapped 2020 forest: 2 cells in 2022. Region-wide, 341 of 80,429 post-2020 loss cells (0.4%) are on mapped 2020 forest.
- The declared plots and the mapped plots differ, and neither an older Hansen release nor a different boundary rule reconciles them. The mapped observations are not compliance findings.
Five plots declare an indication, then a compliant status.
The supplier's record covers 17 plots (plot IDs only are used here). For five, the record states that an "indication of deforestation" was found using Hansen Tree Cover Loss 2021–2023, that on-site verification with evidence was recorded, and that the final status is "Deforestation-free" and "Compliant" for EUDR.
This pattern is a review lead. It does not establish a violation, an error in the record or a legal status.
◆ = declared an indication.
Hansen alone maps loss after 2020 on 7 of 17 plots.
Hansen's map records the year of any tree-cover loss for every ~28 m cell. Counting the cells whose centre falls inside each plot gives the barcode: plot by year, with count as colour intensity. The supplier's record names the 2021–23 window; the current Hansen release also covers 2024 and 2025.
Aggregate counts only, no geometry; counts are cells, not hectares. Barcode years before 2021 are left out because the decision does not read them.
Loss in the supplier's 2021–23 window appears on P-0001 (38 cells, all coded 2022), P-0016 (1 cell, 2023) and P-0018 (1 cell, 2021). The other four plots with later loss (P-0005, P-0006, P-0013, P-0014) show it only in 2024 or 2025. None of the three plots with in-window loss is one of the five that declared an indication.
Almost all post-2020 loss is on land JRC does not map as 2020 forest.
EUDR asks about forest on 31 December 2020, not about any tree loss. Hansen loss also records harvests on land that was not forest in 2020. The decision layer therefore combines Hansen with JRC's 2020 forest class, and counts only mapped loss on mapped 2020 forest.
Step 1 of 4
The dot positions are random; only the counts are real.
Plot by plot
- P-0001: 60 of its 121 cells are JRC 2020 forest. Of its 38 loss cells in 2022, 2 are on mapped 2020 forest (mapped loss on mapped 2020 forest). The other 36 are on land JRC did not map as forest.
- The other 16 plots: no mapped 2020 forest, so no mapped loss on mapped 2020 forest, with complete support. Six of them (P-0005, P-0006, P-0013, P-0014, P-0016, P-0018) do have Hansen loss after 2020. All of it falls outside JRC's mapped 2020 forest class.
Why those cells are not forest in the JRC map is not established; a plantation harvest cycle is one possibility.
Detail: why one layer alone overstates post-2020 loss technical
On the region, 42.9% of land cells show Hansen loss at some point since 2001 (319,011 of 743,418). A Hansen-only forest proxy (canopy above 10%, no loss to 2020) marks 58.4% of land as forest and counts 71,030 post-2020 loss cells on it; JRC marks 15.9% as forest. That gap is why a rubber and oil-palm mosaic needs the second layer. At JRC's native ~9 m resolution, P-0001 has 554 forest cells out of 1,083 JRC cells.
The JRC rule reads one JRC cell per Hansen cell (the centre). Reading a 3×3 majority or "any forest in the block" instead changes no plot decision, but only P-0001 has forest, so it is a weak test. "Any forest" moves the regional total of loss-on-forest cells from 341 to 506 and P-0001's from 2 to 4.
The declared plots and the plots with mapped loss do not coincide.
The sets are compared in three views: the supplier's 2021–23 window, all Hansen loss after 2020, and mapped loss on mapped 2020 forest.
Step 1 of 5
Dots show the last two digits of the plot ID; purple = declared, orange ring = mapped loss.
No declared plot has mapped loss in the window the supplier named, and the one plot with mapped loss on mapped 2020 forest is not a declared plot. Two declared plots (P-0005 and P-0013) show later Hansen loss, in 2024 and 2025, after the 2021–23 window the supplier cites.
Neither an older release nor a different rule reconciles the declarations with the maps.
Two candidate explanations were tested: the supplier used an older Hansen release, or assigned cells to plots by a different rule. The method was fixed before the results were examined. In every variant, none of the five declared indications is reproduced in the 2021–23 window.
Switching from "centre inside" to "any overlap" changes five plot-period answers on other plots, so the inclusion rule is part of the decision, yet it creates no 2021–23 signal on the five declared plots. These tests cannot establish what the supplier did: manual review, a threshold, another layer or a buffer are all possible. A plot-level decision can be checked only if the layers, versions and rules behind it can be replayed; Plotline's receipts record these.
"No loss seen" and "could not be read" are different answers.
Every Plotline answer carries one of four truth states. Missing data is never reported as zero.
Click a cell to cycle through the four kinds.
Support
Mapped loss after 2020
The grid is invented. In the app, unknown is drawn hatched and never as 0.
Every selected cell was read and none shows loss. Shown as "none" with complete support.
At least one loss cell was seen. It stays positive even if other cells are missing: missing data cannot cancel loss that was seen.
Some cells could not be read, and none of the rest shows loss. The answer cannot be "no loss", only "none seen".
Nothing can be computed: no cell centres, no valid land, or a required layer is missing.
Design limits
- Edge cells are all-or-nothing. Under the centre rule a cell counts for a plot when its centre lies inside the boundary; cells are about 0.076 ha. Partial edge coverage and the 0.5 ha minimum area are not modelled.
- JRC is sampled once per Hansen cell, at the centre of its 3 × 3 block. The alternatives (majority, any forest) give the same decision for all 17 plots, but only one plot has mapped forest.
- The layer is rule-specific. A changed rule or an added layer means re-deriving from the lossless audit tier: seconds per region, about 7 CPU-minutes per tile (estimated).