In short. The coloured zones show the legal status of the land — what each area is officially allowed to be used for. NoFilter's coffee grows on land that is legally farmable, right beside protected forest, acting as a buffer. The red overlay marks where tree cover has been lost since 2000, spotted by satellites — it shows where loss happened, not why. Everything below is the detailed, auditable version.
What this map shows. Where NoFilter sources coffee, the legal forest classification of the surrounding land, and where tree cover has been lost since 2001. This is the evidence layer for the brand thesis that smallholder coffee on legally cultivable land at the forest edge can hold the line on adjacent protected forest.
What Hansen detects. The red overlay is the Hansen / UMD Global Forest Change dataset via Global Forest Watch. It records tree cover loss at 30-metre pixel resolution, annually since 2001, against a fixed year-2000 canopy baseline. A pixel flags as lost when a substantial portion of its canopy is removed in a given year. Only the first year of loss is recorded; subsequent disturbances on the same pixel are not captured.
What Hansen doesn't detect. Hansen captures loss but cannot identify cause. A red pixel might represent illegal clearing, legal timber harvest, plantation rotation, agroforestry canopy management, landslide, or fire. Selective logging that leaves canopy mostly intact is mostly invisible. False-positive rates in Indonesia are estimated at 8–12 percent, concentrated on steep slopes, cloud edges, and areas of marginal canopy density.
The 30 percent canopy threshold. Numbers in the Forest Change panel filter to count only loss on land that had at least 30 percent canopy in the year 2000. This is the GFW default and aligns with most national forest definitions. Lower thresholds capture more loss including agroforestry-scale thinning. The 30 percent setting is the standard GFW convention; lower thresholds would count more canopy change, including agroforestry-scale thinning. We report the standard figure and flag that it is not the only defensible one. The map tile and the calculator may use different thresholds, which is why some red pixels visible on screen are not counted in the headline tally.
Where the polygons came from. Cooperative footprints are not legal property boundaries. Sampean, Tanjung Dolok, Aek Sabaon, and Sitaratoit are convex hulls drawn around GPS plot points supplied by PT BAS / Koperasi Maju Bersama (October 2024). The hulls are deliberately generous, enclosing more than the actual cultivated area. Plot perimeter shapefiles will replace these when received. Cinta Raja is placeholder geometry pending the OIC restoration boundary. Batang Gadis polygons are coffee distribution zones across five sub-districts in Mandailing Natal, supplied by SRI via Koperasi Kopi Mandailing Jaya (May 2026); these are not legal property boundaries but indicate the spread of cooperative member plots. 290 anonymised farmer plot points within these zones are also displayed.
What this map cannot prove. This map shows what did or did not happen on cooperative land. It does not establish a counterfactual — what would have happened to the surrounding landscape without NoFilter. Showing zero or low loss inside cooperative polygons is necessary but not sufficient evidence of conservation impact. A defensible counterfactual requires comparison against matched control villages or adjacent forest of the same legal class. We do not yet display that comparison.
What we still owe. Plot perimeter shapefiles from cooperative partners. Loss broken down by KLHK legal class within each polygon (loss in CA conservation forest is qualitatively different from loss in APL agricultural land). Regional comparison numbers showing landscape-scale loss alongside polygon-scale loss. A second figure at 10 percent threshold alongside the 30 percent headline. Year-by-year disaggregation of the Hansen tile overlay. (Hover-to-identify on the legal-status overlay is now live.)